Acta Orthopaedica Vol. 90 Issue 4, Aug 2019 by Acta

4/19 ACTA ORTHOPAEDICA

Element of success in joint replacement

www.heraeus-medical.com

Vol. 90, No. 4, 2019 (pp. 297–409)

Proven for 60 years in more than 30 million procedures worldwide. *OREDObOHDGHU LQ FOLQLFDO HYLGHQFH ZLWK PRUH WKDQ VWXGLHV 7KLV makes PALACOS® ERQH FHPHQW ZKDW LW LV 7KH JROG VWDQGDUG DPRQJ bone cements, and the element of success in joint replacement.

Volume 90, Number 4, August 2019

Acta Orthopaedica is owned by the Nordic Orthopaedic Federation and is the official publication of the Nordic Orthopaedic Federation

E DITORIAL O F FICE

Acta Orthopaedica Department of Orthopedics Lund University Hospital SE–221 85 Lund, Sweden E-mail: acta.ort@med.lu.se Homepage: http://www.actaorthop.org

EDITOR

THE FOUNDATION BOARD OF

Anders Rydholm Lund, Sweden

THE NORDIC O RTHOPAEDIC F EDERATION AND A CTA O RTHOPAEDICA

DEPUTY EDITOR

Peter A Frandsen Odense, Denmark CO-EDITORS

Li Felländer-Tsai Stockholm, Sweden Nils Hailer Uppsala, Sweden Ivan Hvid Oslo, Norway Urban Rydholm Lund, Sweden Bart A Swierstra Wageningen, The Netherlands Eivind Witsø Trondheim, Norway Rolf Önnerfält Lund, Sweden

Peter Frandsen Denmark Ragnar Jonsson Iceland Heikki Kröger Finland Anders Rydholm Sweden Kees Verheyen the Netherlands

WEB EDITOR

Magnus Tägil Lund, Sweden S TATISTICAL EDITOR

Jonas Ranstam Lund, Sweden P RODUCTION MANAGER

Kaj Knutson Lund, Sweden

Vol. 90, No. 4, 2019

SUBSCRIPTION INFORMATION SUBSCRIPTION INFORMATION Acta Orthopaedica [print 1745-3674, online 1745-3682] is a peerActa Orthopaedica [print 1745-3674, is a peerreviewed journal, published six times online a year 1745-3682] plus supplements by reviewed journal,on published timesOrthopaedic a year plus supplements Taylor & Francis behalf ofsix Nordic Federation. by Taylor & Francis on behalf of Nordic Orthopaedic Federation. Annual Institutional Subscription, Volume 90, 2019 Annual Institutional Subscription, Volume 90, 2019 $1,218 £753 €976 $1,218 £753 €976 The subscription fee purchases an online subscription. The price The subscription purchases online The 1997 price includes access to fee current contentanand backsubscription. issues to January includes access to current content backare issues to January 1997 (if available). Printed copies of theand journal provided on request (if aavailable). Printed copies of the journal are provided request as free supplementary service accompanying an onlineon subscripas a free supplementary an online tion. Supplements to theservice journalaccompanying are also included in the subscription.price. Supplements to the journal are also included in the subscription tion price. For more information, visit the journal’s website: For more information, visit the journal’s website: http://www.tandfonline.com/IORT http://www.tandfonline.com/IORT Manuscripts should be uploaded at Manuscripts should be uploaded at http://www.manuscriptmanager.com/ao/ http://www.manuscriptmanager.com/ao/ for further handling at: for further handlingEditorial at: Acta Orthopaedica Office, Acta Orthopaedica Editorial Office, Department of Orthopaedics, Department of Orthopaedics, Lund University Hospital, Lund University SE-221 85 Lund,Hospital, Sweden SE-221 85 Lund, Sweden Correspondance concerning copyright and permissions should be Correspondance concerning copyright and permissions should be sent to: sent to:Pårup Håkan Håkan Pårup P.O. Box 3255, SE-103 65 Stockholm, Sweden, P.O. +46 Box(0)8 3255, SE-103 Stockholm, Tel: 440 80 40.65 Fax: +46 (0)8 Sweden, 440 80 50. Tel: +46hakan.parup@informa.com. (0)8 440 80 40. Fax: +46 (0)8 440 80 50. E-mail: E-mail: hakan.parup@informa.com. Ordering information: Please contact your local Customer Service Ordering information: contact to your Customer Service Department to take out aPlease subscription the local Journal: USA, Canada: Department to takeInc., out a530 subscription to theSuite Journal: Canada: Taylor & Francis, Walnut Street, 850,USA, Philadelphia, Taylor & Francis, Inc., Street, 850,207 Philadelphia, PA 19106, USA. Tel: +1530 800Walnut 354 1420; Fax:Suite +1 215 0050. UK/ PA 19106, USA. Tel: +1T&F 800 Customer 354 1420; Services, Fax: +1 215 207 0050. Europe/Rest of World: Informa UK UK/ Ltd, Europe/Rest of Colchester, World: T&FEssex, Customer Informa UK Ltd, Sheepen Place, CO3 Services, 3LP, United Kingdom. Tel: Sheepen Place, 3LP, United +44 (0) 20 7017Colchester, 5544; Fax:Essex, +44 (0)CO3 20 7017 5198; Kingdom. Tel: +44 (0)subscriptions@tandf.co.uk 20 7017 5544; Fax: +44 (0) 20 7017 5198; Email: Email: subscriptions@tandf.co.uk Dollar rates apply to all subscribers outside of Europe. Euro rates Dollartorates apply to allinsubscribers outside Europe. Euro rates apply all subscribers Europe except theofUK and Republic of apply to Ifallyou subscribers in Europe except the UK and Republic of Ireland. are unsure which applies, contact Customer SerIreland. you are unsure which applies, contact Customer Services. AllIfsubscriptions are payable in advance and all rates include vices. AllJournals subscriptions are payable in advance all ratesMexico, include postage. are sent by air to the USA,and Canada, postage. Journals sent by air to the USA,areCanada, India, Japan and are Australasia. Subscriptions enteredMexico, on an India, Japan and January Australasia. Subscriptions aremay entered on by an annual basis, i.e., to December. Payment be made annual basis, to December. Payment may be made by sterling check,i.e., USJanuary dollar check, euro check, international money sterlingNational check, Giro, US dollar check, check, international money order, or credit cardeuro (Amex, Visa and Mastercard). order, National Giro, or credit card (Amex, Visa and Mastercard). Back issues: Taylor & Francis retains a two-year back issue stock of Back issues: Taylor & Francis retains a two-year issuetostock of journals. Older volumes are held by our official back stockists whom journals. volumes are heldbebyaddressed: our officialPeriodicals stockists toService whom all ordersOlder and enquiries should all orders and enquiriesAve., should be 300, addressed: Periodicals Company, 351 Fairview Suite Hudson, New YorkService 12534, Company, Fairview Ave., fax: Suite+1 300, Newe-mail: York 12534, USA. Tel: 351 +1 518 537 4700; 518Hudson, 537 5899; psc@ USA. Tel: +1 518 537 4700; fax: +1 518 537 5899; e-mail: psc@ periodicals.com. periodicals.com.

Airfreight and mailing in the USA by agent named Air Business and mailing mailing in the the USA byagent agent named Air2nd Business Ltd, c/o Worldnet Shipping Inc., 156-15, 146th Avenue, Floor, Airfreight and in USA by named WN Shipping Ltd, c/o Worldnet Shipping Inc., 146thpaid Avenue, 2nd Floor, Jamaica, NY 11434, USA. Periodicals Jamaica NY USA, 156-15, 146th Avenue, 2nd156-15, Floor,postage Jamaica, NYat11434, USA. Jamaica, NYpostage 11434,paid USA. 11431. Periodicals at Periodicals Jamaica NYpostage 11431.paid at Jamaica NY 11431. US Postmaster: Send address changes changes to to Acta ActaOrthopaedica, Orthopaedica,WN Air US Postmaster: Send address changes to2nd Acta Orthopaedica, Air Business Ltd, Worldnet Shipping Inc., 156-15, 146th Avenue, Shipping USA,c/o156-15, 146th Avenue, Floor, Jamaica, NY Business Ltd, c/o Worldnet Shipping 2nd Floor, Jamaica, NY 11434, USA. Inc., 156-15, 146th Avenue, 11434, USA. 2nd Floor, Jamaica, NY 11434, USA. Subscription records are maintained at Taylor & Francis Group, maintained at OX14 Taylor4RN, & Francis 4Subscription Park Square,records Miltonare Park, Abingdon, UnitedGroup, King4 Park Square, Milton Park, Abingdon, OX14 4RN, United Kingdom. dom. Copyright © 2019 The Author(s). Published by Taylor & FranCopyright © 2019 Author(s). Published by Taylor & Francis on behalf of theThe Nordic Orthopedic Federation. This is an cis onAccess behalfarticle of the distributed Nordic Orthopedic is an Open under theFederation. terms of theThis Creative Open Access article distributed under the terms of the Creative Commons Attribution-Non-Commercial License (https://creCommons Attribution-Non-Commercial License (https://creativecommons.org/licenses/by-nc/3.0 . ativecommons.org/licenses/by-nc/3.0 . Informa UK Limited, trading as Taylor & Francis Group makes Informa UK to Limited, as Taylor Group makes every effort ensure trading the accuracy of & all Francis the information (the every effortcontained to ensureinthe of allHowever, the information “Content”) its accuracy publications. Informa (the UK “Content”) contained in its&publications. However, Informa Limited, trading as Taylor Francis Group, our agents, and UK our Limited, trading Taylor & Francis Group, ourwhatsoever agents, andasour licensors make noasrepresentations or warranties to licensors makecompleteness no representations or warranties to the accuracy, or suitability for anywhatsoever purpose ofasthe the accuracy, completeness or suitability purpose of are the Content. Any opinions and views expressedfor in any this publication Content. Any and opinions expressed the opinions viewsand of views the authors, and inarethis notpublication the views are of theendorsed opinionsby and views UK of the authors, and are not the&views of or Informa Limited, trading as Taylor Francis or endorsed by Informa UKContent Limited,should trading Francis Group. The accuracy of the notasbeTaylor relied& upon and Group. be The accuracy of the Content notsources be relied and should independently verified withshould primary of upon informashould be independently primary sources Group of information. Informa UK Limited,verified tradingwith as Taylor & Francis shall tion.beInforma UKany Limited, Taylorproceedings, & Francis Group shall not liable for losses,trading actions,asclaims, demands, not beexpenses, liable for any losses,and actions, proceedings, costs, damages, otherclaims, liabilities whatsoeverdemands, or howcosts, expenses, damages, and or other liabilities or howsoever caused arising directly indirectly in whatsoever connection with, in soever caused arisingout directly or indirectly in connection in relation to or arising of the use of the Content. Terms &with, Condirelation or arising out of thebeuse of theatContent. Terms & Conditions of to access and use can found http://www.tandfonline. tions of access and use can be found at http://www.tandfonline. com/page/terms-and-conditions com/page/terms-and-conditions Indexed/abstracted in: Allied and Complementary Medicine Indexed/abstracted in: (Automatic Allied andSubject Complementary Medicine Library (Amed); ASCA Citation Alert); BioLibraryAbstracts; (Amed); ASCA (Automatic Citation Alert); Biological Chemical Abstracts;Subject Cumulative Index to Nurslogical Chemical Abstracts; Cumulative Index to Nursing andAbstracts; Allied Health Literature(CINAHL); Current Advances in ing and Allied Health Literature(CINAHL); Current Advances in Ecological and Environmental Sciences; Current Contents/Clinical Ecological Current and Environmental Sciences; Current Contents/Clinical Medicine; Contents/Life Sciences; Developmental MediMedicine; Current Contents/Life Medicine and Child Neurology; EnergySciences; Research Developmental Abstracts; EMBASE/ cine and Child Neurology; Energy Research Abstracts; EMBASE/ Excerpta Medica; Faxon Finder; Focus On: Sports Science & MedExcerpta Medica; Faxonand Finder; Focus On: Sports & Medicine; Health Planning Administration; Index Science Medicus/MEDicine; Health Planning and Administration; Index Medicus/MEDLINE; Index to Dental Literature; Index Veterinarius; INIS AtomLINE; Medical Index toDocumentation Dental Literature; IndexNuclear Veterinarius; INIS Atomindex; Service; Science Abstracts index; Medical Documentation Service; NuclearLife Science Abstracts (Ceased); Periodicals Scanned and Abstracted. Sciences Col(Ceased); Periodicals Abstracted. Life Sciences Collection; Research Alert;Scanned Scienceand Citation Index; SciSearch; SportSlection;Uncover Research Alert; Science Citation Index; SciSearch; SportSearch; Veterinary Bulletin. earch; Uncover Veterinary Bulletin. Printed in England by Henry Ling Printed in England by Henry Ling

Acta Orthopaedica

ISSN 1745-3674

Vol. 90, No. 4, August 2019 Editorial Hypothesis-generating and confirmatory studies, Bonferroni correction, and pre-specification of trial endpoints Hip Optimal duration of anticoagulant thromboprophylaxis in total hip arthroplasty: new evidence in 55,540 patients with osteoarthritis from the Nordic Arthroplasty Register Association (NARA) group Minor influence of patient education and physiotherapy interventions before total hip replacement on patient-reported outcomes: an observational study of 30,756 patients in the Swedish Hip Arthroplasty Register Survival of 11,390 Continuum cups in primary total hip arthroplasty based on data from the Finnish Arthroplasty Register Patient-reported outcomes in hip resurfacing versus conventional total hip arthroplasty: a register-based matched cohort study of 726 patients ASA class is associated with early revision and reoperation after total hip arthroplasty: an analysis of the Geneva and Swedish Hip Arthroplasty Registries Outcome of 881 total hip arthroplasties in 747 patients 21 years or younger: data from the Nordic Arthroplasty Register Association (NARA) 1995–2016 Lower 5-year cup re-revision rate for dual mobility cups compared with unipolar cups: report of 15,922 cup revision cases in the Dutch Arthroplasty Register (2007–2016) Centenarian hip fracture patients: a nationwide population-based cohort study of 507 patients Hip-fracture osteosynthesis training: exploring learning curves and setting proficiency standards Knee Medial unicompartmental knee arthroplasty: increasingly uniform patient demographics despite differences in surgical volume and usage—a descriptive study of 8,501 cases from the Danish Knee Arthroplasty Registry Patient-reported 1-year outcome not affected by body mass index in 3,327 total knee arthroplasty patients Marker-based versus model-based radiostereometric analysis of total knee arthroplasty migration: a reanalysis with comparable mean outcomes despite distinct types of measurement error Inadequate evaluation and management of suspected infections after TKA surgery in Lithuania: a retrospective study of 2,769 patients with 2-year follow-up An infrapatellar nerve block reduces knee pain in patients with chronic anterior knee pain after tibial nailing: a randomized, placebo-controlled trial in 34 patients Hand Non-union of the ulnar styloid process in children is common but long-term morbidity is rare: a population-based study with mean 11 years (9–15) follow-up Revision of trapeziometacarpal arthroplasty: risk factors, procedures and outcomes Artificial intelligence detection of distal radius fractures: a comparison between the convolutional neural network and professional assessments Fibrous dysplasia Pain in fibrous dysplasia: relationship with anatomical and clinical features

297

J Ranstam

298

A B Pedersen, I T Andersen, S Overgaard, A M Fenstad, S A Lie, J-E Gjertsen, and O Furnes

306

C Torisho, M Mohaddes, K Gustafsson, and O Rolfson

312

M Hemmilä, M Karvonen, I Laaksonen, M Matilainen, A Eskelinen, J Haapakoski, A-P Puhto, J Kettunen, M Manninen, and K T Mäkelä A Oxblom, H Hedlund, S Nemes, H Brismar, L Felländer-Tsai, and O Rolfson

318 324

R J Ferguson, A J Silman, C Combescure, E Bulow, D Odin, D Hannouche, S Glyn-Jones, O Rolfson, and A Lübbeke

331 338

V Halvorsen, A M Fenstad, L B Engesæter, L Nordsletten, S Overgaard, A B Pedersen, J Kärrholm, M Mohaddes, A Eskelinen, K T Mäkelä, And S M Röhrl E M Bloemheuvel, L N van Steenbergen, and B A Swierstra

342

M Mosfeldt, C M Madsen, J B Lauritzen, and H L Jørgensen

348

A Gustafsson, P Pedersen, T B Rømer, B Viberg, H Palm, and L Konge

354

C Henkel, M Mikkelsen, A B Pedersen, L E Rasmussen, K Gromov, A Price, and A Troelsen

360

A Overgaard, L Lidgren, M Sundberg, O Robertsson, and A W-Dahl K T van Hamersveld, P J Marang–van de Mheen, L A Koster, R G H H Nelissen, S Toksvig-Larsen, and B L Kaptein

366 373

E Terteliene, K Grigaitis, O Robertsson, J Stucinskas, S Tarasevicius, N Porvaneckas, and A Venalis

377

M S Leliveld, S J M Kamphuis, and M H J Verhofstad

383

L Korhonen, S Victorzon, W Serlo, and J-J Sinikumpu

389

S Mattila and E Waris

394

K Gan, D Xu, Y Lin, Y Shen, T Zhang, K Hu, K Zhou, M Bi, L Pan, W Wu, and Y Liu

401

B C J Majoor, E Traunmueller, W Maurer-Ertl, N M Appelman-Dijkstra, A Fink, B Liegl, N A T Hamdy, P D S Dijkstra, and A Leithner

Technical note The development of an online implant manufacturer application: a knowledge-sharing platform for the Swedish Hip Arthroplasty Register Information to authors (see http://www.actaorthop.org/)

406

J Vinblad, D Odin, J KĂ¤rrholm, and O Rolfson

Acta Orthopaedica 2019; 90 (4): 297

297

Editorial

Hypothesis-generating and confirmatory studies, Bonferroni correction, and pre-specification of trial endpoints

A p-value presents the outcome of a statistically tested null hypothesis. It indicates how incompatible observed data are with a statistical model defined by a null hypothesis. This hypothesis can, for example, be that 2 parameters have identical values, or that they differ by a specified amount. A low p-value shows that it is unlikely (a high p-value that it is not unlikely) that the observed data are consistent with the null hypothesis. Many null hypotheses are tested in order to generate study hypotheses for further research, others to confirm an already established study hypothesis. The difference between generating and confirming a hypothesis is crucial for the interpretation of the results. Presenting an outcome from a hypothesis-generating study as if it had been produced in a confirmatory study is misleading and represents methodological ignorance or scientific misconduct. Hypothesis-generating studies differ methodologically from confirmatory studies. A generated hypothesis must be confirmed in a new study. An experiment is usually required for confirmation as an observational study cannot provide unequivocal results. For example, selection and confounding bias can be prevented by randomization and blinding in a clinical trial, but not in an observational study. Confirmatory studies, but not hypothesis-generating studies, also require control of the inflation in the false-positive error risk that is caused by testing multiple null hypotheses. The phenomenon is known as a multiplicity or mass-significance effect. A method for correcting the significance level for the multiplicity effect has been devised by the Italian mathematician Carlo Emilio Bonferroni. The correction (Bender and Lange 2001) is often misused in hypothesis-generating studies, often ignored when designing confirmatory studies (which results in underpowered studies), and often inadequately used in laboratory studies, for example when an investigator corrects the significance level for comparing 3 experimental groups by lowering it to 0.05/3 = 0. 017 and believes that this solves the problem of testing 50 null hypotheses, which would have required a corrected significance level of 0.05/50 = 0.001. In a confirmatory study, it is mandatory to show that the tested hypothesis has been pre-specified. A study protocol or statistical analysis plan should therefore be enclosed with the study report when submitted to a scientific journal for publication. Since 2005 the ICMJE (International Committee of Med-

ical Journal Editors) and the WHO also require registration of clinical trials and their endpoints in a publicly accessible register before enrollment of the first participant. Changing endpoints in a randomized trial after its initiation can in some cases be acceptable, but this is never a trivial problem (Evans 2007) and must always be described to the reader. Many authors do not understand the importance of pre-specification and desist from registering their trial, use vague or ambiguous endpoint definitions, redefine the primary endpoint during the analysis, switch primary and secondary outcomes, or present completely new endpoints without mentioning this to the reader. Such publications are simply not credible, but are nevertheless surprisingly common (Ramagopalan et al. 2014) even in high impact factor journals (Goldacre et al. 2019). A serious editorial evaluation of manuscripts presenting confirmatory results should always include a verification of the endpoint’s pre-specification. Hypothesis-generating studies are much more common than confirmatory, because the latter are logistically more complex, more laborious, more time-consuming, more expensive, and require more methodological expertise. However, the result of a hypothesis-generating study is just a hypothesis. A hypothesis cannot be generated and confirmed in the same study, and it cannot be confirmed with a new hypothesis-generating study. Confirmatory studies are essential for scientific progress. Jonas Ranstam, Statistical Editor jonas.ranstam@gmail.com

Bender R, Lange S. Adjusting for multiple testing: when and how? J Clin Epidemiol 2001; 54: 343-9. Evans S. When and how can endpoints be changed after initiation of a randomized clinical trial? PLoS Clin Trials 2007; 2: e18. Goldacre B, Drysdale H, Milosevic I, Slade E, Hartley P, Marston C, PowellSmith A, Heneghan C, Mahtani K R. COMPare: a prospective cohort study correcting and monitoring 58 misreported trials in real time. Trials 2019; 20: 118. Ramagopalan S, Skingsley A P, Handunnetthi L, Klingel M, Magnus D, Pakpoor J, Goldacre B. Prevalence of primary outcome changes in clinical trials registered on ClinicalTrials.gov: a cross-sectional study. F1000Research 2014, 3: 77.

© 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group, on behalf of the Nordic Orthopedic Federation. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. DOI 10.1080/17453674.2019.1612624

298

Acta Orthopaedica 2019; 90 (4): 298–305

Optimal duration of anticoagulant thromboprophylaxis in total hip arthroplasty: new evidence in 55,540 patients with osteoarthritis from the Nordic Arthroplasty Register Association (NARA) group Alma B PEDERSEN 1, Ina Trolle ANDERSEN 1, Soren OVERGAARD 2, Anne Marie FENSTAD 3, Stein Atle LIE 4,5, Jan-Erik GJERTSEN 3,4, and Ove FURNES 3,4 1 Department

of Clinical Epidemiology, Aarhus University Hospital, Denmark; 2 Department of Orthopaedic Surgery and Traumatology, Odense University Hospital, Department of Clinical Research, University of Southern Denmark, Odense, Denmark; 3 The Norwegian Arthroplasty Register, Department of Orthopedic Surgery, Haukeland University Hospital, Bergen, Norway; 4 Department of Clinical Medicine, University of Bergen, Norway; 5 Department of Clinical Dentistry, University of Bergen, Bergen, Norway Correspondence: abp@clin.au.dk Submitted 2018-11-30. Accepted 2019-02-21.

Background and purpose — The recommended optimal duration of the thromboprophylaxis treatment in total hip arthroplasty (THA) patients has been a matter of debate for years. We examined the association between short (1–5 days), standard (6–14 days), and extended (≥ 15 days) duration of thromboprophylaxis, with regards to the risk of venous thromboembolism (VTE), major bleeding, and death in unselected THA patients. Patients and methods — We performed a cohort study using prospectively collected data from the populationbased hip arthroplasty registries, prescription databases, and patient administrative registries in Denmark and Norway. We included 55,540 primary THA patients with osteoarthritis Results — The 90-day cumulative incidence of VTE was 1.0% for patients with standard treatment (reference), 1.1% for those with short-term treatment (adjusted hazard ratio [aHR] of 1.1, 95% confidence interval (CI) 0.8–1.5) and 1.0% for those with extended treatment (aHR of 0.9, CI 0.8–1.2). The aHRs for major bleeding were 1.1 (CI 0.8–1.6) for short and 0.8 (CI 0.6–1.1) for extended vs. standard treatment. In addition, patients with short and extended treatment had aHRs for death of 1.2 (CI 0.8–1.8) and 0.8 (CI 0.5–1.1) vs. standard treatment, respectively. Patients who started short treatment postoperatively had an aHR for death of 1.8 (CI 1.1–3.1) and absolute risk difference of 0.2%, whereas patients who started short treatment preoperatively had an aHR for death of 0.5 (CI 0.2–1.2) and absolute risk difference of 0.3% compared with patients who had standard treatment with post- and preoperative start, respectively.

Interpretation — In routine clinical practice, we observed no overall clinically relevant difference in the risks of VTE and major bleeding within 90 days of THA with respect to thromboprophylaxis duration. However, our data indicate that short-term thromboprophylaxis started postoperatively is associated with increased 90-day mortality. The significance of these data should be explored further.

The incidence of total hip arthroplasty (THA) procedures increases annually worldwide (Nemes et al. 2014). Risk of symptomatic venous thromboembolism (VTE) within 90 days of THA are reported to range from 1% to 4% (Pedersen et al. 2012, Huo 2012, Wolf et al. 2012) in the presence of thromboprophylaxis, and is furthermore elevated up to 1 year postoperatively (Pedersen et al. 2012). Given the high risk of VTE in the absence of thromboprophylaxis and high mortality following symptomatic VTE (Pedersen et al. 2017), anticoagulant thromboprophylaxis for THA patients is preferred treatment in most countries. However, the recommended optimal duration of the treatment has been a matter of debate for years. The American College of Chest Physicians (ACCP) guidelines from 2012 recommend a minimum of 10 to 14 days of thromboprophylaxis and suggest extending the treatment to 35 days in the outpatient period (Falck-Ytter et al. 2012). The American Academy of Orthopaedic Surgeons (AAOS) guidelines from 2011 recommend individual assessment of the optimal duration of thromboprophylaxis (AAOS 2013). Since a number of concerns have been identified with these guidelines (Budhiparama et al. 2014), and due to considerable change in the THA course with introduction of fast-track programs in

Acta Orthopaedica 2019; 90 (4): 298–305

orthopedic departments, several national guidelines have been published since. Danish national guidelines recommend anticoagulant thromboprophylaxis for 6–10 days in THA patients, and less than 5 days if fast-track THA surgery was performed (Danish Council for the Use of Expensive Hospital Medicine [RADS] 2016). In Norway, thromboprophylaxis is recommended for 10 postoperative days (Granan 2015). The latest paper from the Cochrane database of systematic reviews concluded that there is moderate quality evidence for extended duration of thromboprophylaxis to prevent VTE in THA patients (Forster and Stewart 2016). Neither of the guidelines suggests risk stratification in order to provide specific duration of thromboprophylaxis for specific THA patients. A Danish cohort study observed no overall difference in the risk of VTE or bleeding with respect to thromboprophylaxis duration in THA patients from routine clinical practice (Pedersen et al. 2015), but this study lacked statistical power to analyze data on the subgroup level. We examined the association between duration of anticoagulant thromboprophylaxis for the prevention of VTE in patients undergoing elective THA in Denmark and Norway. As a safety outcome, we consider bleeding and death. We also aimed to identify THA patients who could benefit from extended prophylaxis without increase in bleeding events.

Patients and methods Study design and setting We conducted this population-based cohort study using prospectively collected data available from the Nordic Arthroplasty Register Association (NARA) database, established in 2009. All Swedish, Norwegian, Danish, and Finnish citizens are assigned a unique civil registration number, permitting unambiguous linkage between hip registries and other medical databases in each country. This also enables tracking of deceased and emigrated patients (Schmidt et al. 2014). The healthcare system in Scandinavian countries provides taxsupported healthcare for all citizens; free medical care is guaranteed for emergency and general hospital admissions, as well as for outpatient clinic visits. The study is reported according to the RECORD guidelines. Study population We used the NARA database to identify all patients operated in Denmark and Norway. Data from Sweden and Finland were not included, since individual-level data on duration of anticoagulant thromboprophylaxis were not available from these countries. We included all primary THAs between January 1, 2010 and December 31, 2014 from Denmark (n = 34,625) and between January 1, 2008 and December 31, 2013 from Norway (n = 34,801), and accessed their preoperative and surgery-related records. Primary THA was defined as insertion of a unilateral total hip prosthesis due to primary osteoarthritis.

299

We excluded 269 and 261 patients from Denmark and Norway, respectively, since these were registered in NARA as not receiving any anticoagulant thromboprophylaxis, leaving 34,356 Norwegian and 34,540 Danish THAs in the study population. We restricted the study population to first-time THA due to primary OA, leaving 26,250 THA patients in Denmark and 31,096 patients in Norway. We excluded 1,806 patients from Norway due to registration of several thromboprophylaxis drugs as the first choice. Finally, 2,310 procedures in Denmark and 2,748 in Norway lacked information on duration of thromboprophylaxis, although they were registered as receiving thromboprophylaxis, which resulted in 23,940 and 26,542 procedures for the complete-case analyses. Duration of anticoagulant thromboprophylaxis We categorized the duration of anticoagulant thromboprophylaxis prescribed by surgeon in relation to THA surgery as short-term (1–5 days), standard (6–14 days), or extended (≥ 15 days) based on available international guidelines for thromboprophylaxis and clinical practice in Denmark and Norway. Allocation of duration of treatment is dependent on the local guidelines at the individual departments. Hence, most standard patients operated at one department will receive the same length of treatment irrespective of their risk. However, high-risk patients such as those with previous VTE or cancer have a higher likelihood of receiveing extended prophylaxis. We included the following anticoagulant agents approved for use in the Denmark and Norway for prevention of VTE in THA patients: parenteral low-molecular-weight heparin (including enoxaparin, dalteparin, and tinzaparin) and fondaparinux, dabigatran, and rivaroxaban, initiated both pre- and postoperatively. Mechanical thromboprophylaxis, when used, is combined with anticoagulant thromboprophylaxis. Outcome The outcomes of interest were VTE, major bleeding, and death. These data were obtained from the Danish National Patient Registry (DNRP) (Schmidt et al. 2015) and the Norwegian Patient Register (NPR) (Bakken et al. 2004). The effectiveness outcome in our analyses was VTE, including deep venous thrombosis (DVT) and pulmonary embolism (PE), whereas safety outcome was death and major bleeding events, including intracranial bleeding, gastrointestinal bleeding, and urinary/lung bleeding. The proportion of hospitalized patients correctly registered in the DNRP and NPR with cardiovascular events and major bleeding has been reported as 75% to 95% (Adelborg et al. 2016, Sundboll et al. 2016). Both primary and secondary diagnoses, which were coded by the physician at the discharge from the hospital or during the outpatient visit, were included in our study. Outcome data from general practitioners were not available. Since any suspicion of VTE and major bleeding would lead to hospitalization or outpatient clinic visit of the majority of

300

patients to confirm the diagnosis and initiate treatment, we may lacked only few patients from general practitioners. Information on death, available from the Danish Civil Registration System (Schmidt et al. 2014) and Statistics Norway, was linked to each patient in the national registers before transferring the data to NARA. Covariates Information on comorbidity was collected from the DNRP and NPR before transferring to the NARA database. For all included THA patients, we identified all primary and secondary discharge diagnoses for all hospitalizations and hospital specialist outpatient visits that occurred 1 year prior to the time of surgery. We computed the Charlson Comorbidity Index (CCI) score for each patient at the time of surgery. We defined 3 comorbidity levels: a score of 0 (low), given to patients with no previous record of diseases included in the CCI; a score of 1–2 (medium); and a score of 3 or more (high). Further variables were included as confounders: age and sex at the time of THA, year of primary THA, fixation type (cemented, uncemented, and hybrid/inverse hybrid fixation), and start of thromboprophylaxis (pre- or postoperative). In addition, we obtained information on the use of acetylsalicylic acid (both low-dose and high-dose), vitamin K antagonists, and platelet inhibitors (clopidogrel, prasugrel, and ticagrelor) from the Danish National Database of Reimbursed Prescriptions (Johannesdottir et al. 2012) and Norwegian Prescription Database (Furu 2008) during the 2 years preceding the primary THR. Statistics We used the cumulative incidence function to compute the 90-day risk of VTE, major bleeding, and death, considering death as competing risk. We used Cox regression to compute crude and hazard ratios (HRs) adjusted for potential confounding factors with 95% confidence intervals (CIs). By application of log–log plots and Cox proportional hazards, the statistical model was found suitable. Follow-up started on the day of primary THA and ended either on the day of discharge from the primary hospitalization (if outcome occurred in the hospital), on the day of readmission for VTE/bleeding, date of death, emigration, or 90 days after surgery, whichever came first. All analyses were repeated stratifying on age, sex, CCI, thromboprophylaxis start, and country of origin in order to study the potential difference in effect of duration of thromboprophylaxis on outcomes in subgroups of patients. Patients were included in each exposure category according to initial treatment assignment. Therefore, in the main analyses we started follow up at the time of THA surgery to examine the effect of initial assigned treatment on VTE/bleeding/ death risk (according to intent to treat approach). Sensitivity analyses We tested the robustness of the results in a series of sensitivity analyses. 1st, the risk of all-cause mortality was calcu-

Acta Orthopaedica 2019; 90 (4): 298–305

lated. 2nd, the risk of VTE and all-cause mortality defined as any DVT, non-fatal PE or all-cause mortality was calculated. 3rd, we analyzed data only for patients that were alive at day 7 or day 14 post-surgery without any VTE event during the first 7 or 14 postoperative days. These data will mimic randomization start in some clinical trials, which compares extended prophylaxis with placebo given after the 7th day (Eikelboom et al, 2001). In addition, these analyses should contribute to understanding of immortal time bias. 4th, we changed exposure definition to short-term (1–7 days), standard (8–14 days), or extended (≥ 15 days), since some hospitals provide prophylaxis using weeks as guideline. 5th, we ignored revision as a competing risk. 6th, we performed cluster analyses at hospital level in order to consider unmeasured hospital specific confounders. 7th, we analyzed data only for patients without use of acetylsalicylic acid before surgery. Analyses were performed using SAS V. 9.4 (SAS Institute Inc., Cary, NC, USA). All diagnosis codes used to identify VTE and major bleeding after THA are presented in Appendix 1 (see Supplementary data). Ethics, funding, and potential conflicts of interest The study was approved by the Danish Data Protection Agency (Region of Central Denmark Jr. nr. 1-16-02-743-14) and Regional Ethical Committee of western Norway (Ref. 2015/880/REK Vest). No funding was received for this study and the authors report no conflicts of interest.

Results Among the 55,540 THA patients operated in Denmark during 2010–2014 and Norway during 2008–2013, 15% received short, 31% received standard, and 45% received extended thromboprophylaxis treatment. Data regarding duration of treatment were missing in 9% (Table 1). Compared with patients who received standard duration of treatment, patients with short duration of treatment were slightly younger, included fewer females, included fewer patients with CCI score medium and high, included fewer patients who received acetylsalicylic acid but more patients who received vitamin K antagonists and platelet inhibitors, and were more likely to undergo uncemented or hybrid THA in Denmark. On the other hand, patients with extended duration of treatment were also slightly younger, included more females, included more patients with CCI score medium and high, included fewer patients who received vitamin K antagonists and platelet inhibitors, and were less likely to undergo uncemented or hybrid THA. In the group with short duration of treatment, 8% were prescribed prophylaxis for 1 day, 20% were prescribed treatment for 2 days, 34% of patients were prescribed prophylaxis for 3 days, 17% were prescribed treatment for 4 days, and 17%

Acta Orthopaedica 2019; 90 (4): 298–305

301

VTE/death were 1.1 (CI 0.9–1.4) for patients prescribed short-term and 0.9 (CI 0.8–1.1) for extended treatment, Duration of thromboprophylaxis compared with standard thromboprophy Missing Short Standard Extended Total laxis treatment. Factor n (%) n (%) n (%)n (%) n (%) n (%) Within 90 days of surgery, cumulaTotal 5,058 (100) 8,333 (100) 17,009 (100) 25,140 (100) 55,540 (100) tive incidences of major bleeding were Age group 0.7% for patients treated with both short 10–59 757 (15) 1,487 (18) 2,790 (16) 3,811 (15.2) 8,845 (16) and standard treatment, and 0.5% for 60–69 1,670 (33) 2,819 (34) 5,646 (33.2) 8,665 (35) 18,800 (34) 70–79 1,877 (37) 2,890 (35) 6,129 (36) 8,971 (38) 19,867 (36) extended thromboprophylaxis treatment 80+ 754 (15) 1,137 (14) 2,444 (14) 3,693 (15) 8,028 (15) (Figure). The adjusted HRs for major Sex bleeding were 1.1 (CI 0.8–1.6) for shortFemale 2,967 (59) 4,586 (55) 10,326 (61) 15,975 (64) 33,854 (61) Male 2,091 (41) 3,747 (45) 6,683 (39) 9,165 (36) 21,686 (39) term treatment and 0.8 (CI 0.6–1.1) for Charlson Comorbidity Index extended duration of treatment compared Missing data 27 (0.5) 2 (0) 26 (0.2) 215 (1) 270 (1) with standard treatment. Low 4,158 (82) 7,352 (88) 14,175 (83) 20,371 (81) 46,056 (83) Medium 770 (15) 864 (10) 2,486 (15) 3,975 (16) 8,095 (15) The risk of death within 90 days of surHigh 103 (2) 115 (1) 322 (2) 579 (2) 1119 (2) gery was 0.5% among patients prescribed Acetylsalicylic acid short-term treatment, 0.4% for standard No 3,815 (75) 6,594 (79) 13,108 (77) 19,202 (76) 42,719 (77) Yes 1,243 (25) 1,739 (21) 3,901 (23) 5,938 (24) 12,821 (23) treatment, and 0.3% for extended treatClopidogrel/prasugrel/ticagrelor ment. The adjusted HRs for death were No 4,949 (98) 8,107 (97) 16,635 (98) 24,727 (98) 54,418 (98) 1.2 (CI 0.8–1.8) for short-term duration Yes 109 (2) 226 (3) 374 (2) 413 (2) 1,122 (2) Marevan/marcoumar/dabigratran/apixaban/rivaroxaban of treatment, and 0.8 (CI 0.5-1.1) for No 4,550 (90) 7,699 (92) 15,764 (93) 23,970 (95) 51,983 (94) extended duration of treatment, compared Yes 508 (10) 634 (8) 1,245 (7) 1170 (5) 3557 (6) with standard duration of treatment. Type of fixation Missing data 72 (1) 16 (0) 178 (1) 343 (1) 609 (1) Analyses stratified on country of Cementeret 1,530 (30) 481 (6) 5,164 (30) 8,971 (36) 16,146 (29) origin showed similar results to those of Uncementeret 1,989 (39) 5,934 (71) 8,625 (51) 9,123 (36) 25,671 (46) the overall population for all outcomes Hybrid 491 (10) 1,691 (20) 855 (5) 774 (3) 3,811 (7) Reverse hybrid 976 (19) 211 (3) 2,187 (13) 5,929 (24) 9,303 (17) (Appendix 2, available from the authors). Start of thromboprophylaxis Stratification on sex showed that females Missing data 1,056 (21) 51 (1) 555 (3) 1,866 (7) 3,528 (6) receiving short treatment duration were Preoperative 1,293 (26) 2,677 (32) 4,892 (29) 6,334 (25) 15,196 (27) Postoperative 2,709 (54) 5,605 (67) 11,562 (68) 16,940 (67) 36,816 (66) associated with increased risk of VTE and VTE/death (adjusted HR were 1.6, CI 1.1–2.2 and 1.6, CI 1.2–2.2) comwere prescribed treatment for 5 days. In the group with stan- pared with standard treatment. However, absolute risk differdard duration of thromboprophylaxis, 47% of patients were ences were 0.4% and 0.5%, respectively. No increased risk prescribed treatment for 7 days whereas 22% and 24% were for VTE and VTE/death was observed for male patients with prescribed treatment for 10 and 14 days, respectively. In the short treatment duration, or for female/male patients receiving group with extended treatment duration, 12%, 20%, and 46% extended treatment compared with standard treatment. Stratiof patients received thromboprophylaxis for 28 days, 30 days, fication on different age and CCI score groups produced the and 35 days, respectively. In the group with short-term treat- estimates similar to overall risk estimates for any outcome. Stratification on start of thromboprophylaxis showed that ment duration, 44% received low-molecular-weight heparin, compared with 86% and 71% in the group with standard treat- short treatment started postoperatively was associated with increased risk of dying (adjusted HR was 1.8, CI 1.1–3.1 and ment and extended duration of treatment. Within 90 days of surgery, cumulative incidences of VTE the absolute risk difference was 0.2%), whereas short treatwere 1.1% for patients prescribed short-term thrombopro- ment started preoperatively was associated with reduced risk phylaxis, and 1.0% for both standard and extended treatment of dying (adjusted HR was 0.5, CI 0.2–1.2 and the absolute (Figure). This correspond to the adjusted HRs for VTE of 1.1 risk difference was 0.3%) compared with patients with stan(CI 0.8–1.5) and 0.9% (CI 0.8–1.2) for patients prescribed dard treatment with post- and preoperative start, respectively. short-term and extended thromboprophylaxis compared with No difference in risk of dying was observed for patients startstandard thromboprophylaxis treatment (Table 2). ing extended treatment either post- or preoperatively. The Cumulative incidences of VTE/death as one outcome within adjusted HR for dying for short treatment started postopera90 days were 1.5% for patients prescribed short-term throm- tively vs. standard treatment started postoperatively was 2.2 boprophylaxis, 1.4% for standard treatment duration, and (CI 1.1–4.2) among patients without use of acetylsalicylic 1.3% for extended treatment duration. The adjusted HRs for acid before surgery. Table 1. Patient characteristics, medication use, and surgery-related factors of 55,540 patients undergoing THA in Denmark in 2010–2014 and Norway in 2008–2013

302

Acta Orthopaedica 2019; 90 (4): 298–305

Cumulative incidence function of major bleeding

Cumulative incidence function of VTE

Cumulative incidence function of 90-day mortality

0.010 0.006

0.004

0.008 0.003 0.004

0.006

0.002

0.004 0.002

0.001

0.002 Short Standard Extended

0.000

Days after index operation

Cumulative incidence function of venous thromboembolism (VTE), major bleeding, and mortality within 90 days of total hip arthroplasty.

Table 2. Effect of thromboprophylaxis treatment duration on the risk of death, thromboembolism (VTE), major bleeding, VTE, and death, and risk of revision in total hip arthroplasty patients within 90 days of surgery, with standard treatment as the reference Outcome Death Death Death VTE VTE VTE Bleeding Bleeding Bleeding VTE/death VTE/death VTE/death

Duration of 90 days thrombo- Outcome Number Crude Adjusted cumulative prophylaxis (n) at risk HR (95% CI) HR (95% CI) incidence (95% CI) Short Standard Extended Short Standard Extended Short Standard Extended Short Standard Extended

39 62 72 89 176 245 56 116 128 125 233 315

8,333 17,009 25,140 8,333 17,009 25,140 8,333 17,009 25,140 8,333 17,009 25,140

1.3 (0.9–1.9) 1.0 0.8 (0.6–1.1) 1.0 (0.8–1.3) 1.0 0.9 (0.8–1.1) 1.0 (0.7–1.4) 1.0 0.7 (0.6–1.0) 1.1 (0.9–1.4) 1.0 0.9 (0.8–1.1)

1.2 (0.8–1.8) 1.0 0.8 (0.5–1.1) 1.1 (0.8–1.5) 1.0 0.9 (0.8–1.2) 1.1 (0.8–1.6) 1.0 0.8 (0.6–1.1) 1.1 (0.9–1.4) 1.0 0.9 (0.8–1.1)

0.5% (0.3–0.6) 0.4% (0.3–0.5) 0.3% (0.2–0.4) 1.1% (0.9–1.3) 1.0% (0.9–1.2) 1.0% (0.9–1.1) 0.7% (0.5–0.9) 0.7% (0.6–0.8) 0.5% (0.4–0.6) 1.5% (1.3–1.8) 1.4% (1.2–1.6) 1.3% (1.1–1.4)

HR = hazard ratio; CI = confidence interval. HR adjusted for age, sex, Charlson Comorbidity Index score, type of fixation, start of thromboprophylaxis, as well as acetylsalicylic acid, clopidogrel/prasugrel/ticagrelor, marevan/marcumar/dabigatran/ apixaban/rivaroxaban used before surgery.

We found similar risk estimates derived from the cluster analyses and those derived from analyses without clustering for hospitals (data not shown). In addition, analyses with slight change of exposure categories definition, analyses not including revision as competing risk into appropriate models, or analyses among THA patients who were alive at day 7/14 without any VTE event during the period from the day of primary THA to 7/14 days after, did not affect overall study results (data not shown).

Discussion In this large population-based study including 55,540 THA patients from routine clinical practice, we found no overall

difference in risk of VTE, VTE/death, or major bleeding with respect to the duration of pharmacological thromboprophylaxis. However, differences between duration of treatments were observed when stratifying on sex and start of thromboprophylaxis treatment, suggesting that female patients with short treatment have increased risk of VTE and patients with short treatment initiated postoperatively have increased risk of dying. This could be used to guide preventive treatment for subgroups of patients; however, low absolute risk difference estimates should be taken into consideration. Strengths and weaknesses Completeness and validity of data in the Danish Hip Arthroplasty Registry and Norwegian Arthroplasty Register are documented to be more than 95% (Pedersen et al. 2004, Arthurs-

Acta Orthopaedica 2019; 90 (4): 298–305

son et al. 2005). The positive predictive value of VTE and major bleeding is 75–95% with an accuracy of 94–100% of the comorbid diagnoses included in the CCI (Bakken et al. 2004, Thygesen et al. 2011, Sundboll et al. 2016). Due to prospective registration of data, misclassification of VTE/bleeding is not likely to be related to registration of exposure status (thus, duration of treatment). Therefore, our HRs should be unbiased. However, our absolute cumulative incidences of VTE/bleeding are underestimated. Information on death and migration was available from the Civil Registration System, allowing for nationwide cohort studies with virtually complete follow-up (Schmidt et al. 2014). Further, selection bias in our study is low since we included all THA patients treated in many departments irrespective of specific patient characteristics or willingness to participate in the study. We lack data on duration of treatment for 9% of patients. However, the median age, sex, and comorbidity level for the excluded patients were similar to those of patients included in the study population. In light of prospective collection of data on duration of treatment and subsequent outcomes of interest, it is unlikely that missing data on duration of treatment occurred systematically. Hypothetically, if patients’ compliance with the treatment after discharge had changed, which is most worrisome for patients included in the extended treatment group, this could have affected our results in any direction. Although we were able to adjust for a number of potential confounders, including preadmission anticoagulant medication, unmeasured and residual confounding, as well as confounding by indication, cannot be accounted for entirely in an observational study, in particular when studying the effect of treatment. We have, however, performed a number of sensitivity analyses in order to elucidate any confounding bias. We have used the cumulative incidence method to estimate crude failure (crude VTE or crude bleeding risk), which is the likely number of VTE or bleeding incidents we see in clinical practice and consists of both the failure of the VTE/bleeding and the mortality process (Sayers et al. 2018). In Table 2 we present both number of failures (VTE/bleeding) and number of patients at risk, which allows for estimation of net failure (the Kaplan–Meier estimates with immortal patients). It is worth noting that the difference between cumulative incidence and Kaplan–Meier estimates is very small, most likely due to short-term follow-up and low mortality in the exposure groups. Comparison with other studies 2 recently published systematic reviews based on clinical trials (Forster and Stewart 2016, Trevisol et al. 2018) both concluded that extended duration thromboprophylaxis should be considered to prevent VTE, when undergoing THA. Forster et al. furthermore specified that the evidence for extended duration was moderate, and that the benefit should be weighed against the increased risk of minor bleeding. Likewise, 2 systematic reviews based on older clinical trials (Hull et al. 2001, Sobieraj et al. 2012) also concluded that extended thromboprophylaxis is more efficacious

303

than shorter thromboprophylaxis for preventing VTE or VTErelated deaths while increasing the risk of minor bleeding. These results do not correlate with our findings. However, in a systemic review, O’Donnell et al. (2003) concluded that the absolute reduction in symptomatic VTE attributed to extended prophylaxis in some clinical trials seems to have been overestimated. The authors suggested that the assessment of symptomatic VTE in clinical trials did not occur independently of the routine venography results, which could have led to an overestimation of the symptomatic VTE risk and the absolute risk reduction attributed to extended prophylaxis. A prospective, observational, non-interventional, phase IV study of 3,914 consecutive patients who underwent total hip or knee arthroplasty during 2010 to 2012 reported that treatment with rivaroxaban for 15 days was protective against symptomatic VTE without excess risk of major bleeding (Gomez et al. 2017). The absolute risk of VTE and bleeding in that study was 0.5% and 0.3%, respectively. In the same study, the authors calculated that treating all 1,444 primary THA patients for 35 days with rivaroxaban could theoretically prevented 1–3 symptomatic proximal DVTs. Based on 16,865 THA patients operated from 2010 through 2012 in Denmark, we have previously reported that there is no difference in the risk of VTE and bleeding with respect to duration of thromboprophylaxis (Pedersen et al. 2015). A study including only patients from fast-track departments where thromboprophylaxis is provided for 1–2 days reported an absolute risk of VTE within 90 days of 0.5% (Petersen et al. 2018). These absolute VTE risk estimates are similar to estimates observed in the current study and estimates from clinical trials. Extended prophylaxis is further challenged by aspirin as potential prophylaxis treatment. Azboy et al. (2017) performed a review on aspirin studies, concluding that there is convincing evidence that aspirin is an effective, inexpensive and safe VTE prevention treatment in THA patients without increased risk of bleeding. Safety of aspirin was reported when administered as part of a hybrid strategy that follows a short course of anticoagulation in select low-risk populations, but also when administered alone in appropriately screened patients (Anderson et al. 2018). Since the rates of VTE and bleeding are very low, the potential benefits of extended prophylaxis may not be worth discussing. The orthopedic surgeons are currently discussing not how long, but how short a duration of thromboprophylaxis should be given, targeting extended thromboprophylaxis to high-risk patients. Our stratified analyses suggest that the start of thrombo prophylaxis might play a role since patients who started short treatment postoperatively had an increased risk of dying. This is in line with the latest findings from Norway, showing that THA patients who started thromboprophylaxis preoperatively vs. those who started postoperatively have adjusted odds ratio for dying of 0.7 (CI 0.5–1.1), whereas the risk of other postoperative complications was similar in the two groups (Borgen et al. 2017).

304

Conclusions In this large cohort study including 55,540 THA patients from a routine clinical practice in Denmark and Norway, we found no overall clinically relevant difference in risk of VTE and bleeding with respect to the duration of pharmacological thromboprophylaxis. However, higher relative risk of dying for short treatment initiated postoperatively was observed. The significance of these data should be explored further, and short-term thromboprophylaxis started preoperatively should be considered. Supplementary data Appendix 2. Stratified analyses: cohort of first primary THA operation in Denmark in 2010–2014 and Norway in 2008– 2013 stratified on age groups, CCI, country, sex, and start preoperatively versus postoperatively is available on request from the authors. Appendix 1 is available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674. 2019.1611215

ABP, SO, and OF conceived the idea for this study. ITA performed the analyses. All co-authors participated in drafting and revising the paper, data design, data interpretations and conclusions, and final approval of the paper. ABP and OF take responsibility for the integrity of the work. Acta thanks Banne Nemeth for help with peer review of this study.

AAOS Preventing venous thromboembolic disease in patients undergoing elective hip and knee arthroplasty. Evidence-based guideline and evidence report. http://www. aaos. org/research/guidelines/VTE/VTE_full_guideline. pdf; 2013 (last accessed February 14, 2019). Adelborg K, Sundboll J, Munch T, Froslev T, Sorensen H T, Botker H E, Schmidt S. Positive predictive value of cardiac examination, procedure and surgery codes in the Danish National Patient Registry: a population-based validation study. BMJ Open 2016; 6: e012817. Anderson D R, Dunbar M, Murnaghan J, Kahn S R, Gross P, Forsythe M, Pelet S, Fisher W, Belzile E, Dolan S, Crowther M, Bohm E, MacDonald S J, Gofton W, Kim P, Zukor D, Pleasance S, Andreou P, Doucette S, Theriault C, Abianui A, Carrier M, Kovacs M J, Rodger M A, Coyle D, Wells P S, Vendittoli P A. Aspirin or rivaroxaban for VTE prophylaxis after hip or knee arthroplasty. N Engl J Med 2018; 378: 699-707. Arthursson A J, Furnes O, Espehaug B, Havelin L I, Soreide J A. Validation of data in the Norwegian Arthroplasty Register and the Norwegian Patient Register: 5,134 primary total hip arthroplasties and revisions operated at a single hospital between 1987 and 2003. Acta Orthop 2005; 76: 823-8. Azboy I, Barrack R, Thomas A M, Haddad F S, Parvizi J. Aspirin and the prevention of venous thromboembolism following total joint arthroplasty: commonly asked questions. Bone Joint J 2017; 99-B: 1420-30. Bakken I J, Nyland K, Halsteinli V, Kvam U H, Skjeldestad F E. Norsk pasientregister: Administrativ database med mange forskningsmuligheter. Norsk Epidemiologi 2004; 14: 65-9. Borgen P O, Pripp A H, Dybvik E, Leistad L, Dahl O E, Reikeras O. Similar clinical outcomes with preoperative and postoperative start of thromboprophylaxis in THA: a register-based study. Clin Orthop Relat Res 2017; 475: 2245-52. Budhiparama N C, Abdel M P, Ifran N N, Parratte S. Venous thromboembolism (VTE) prophylaxis for hip and knee arthroplasty: changing trends. Curr Rev Musculoskelet Med 2014; 7: 108-16.

Acta Orthopaedica 2019; 90 (4): 298–305

Danish Council for the Use of Expensive Hospital Medicine (RADS) Baggrundsnotat: Tromboseprofylakse til ortopædkirurgiske patienter; 2016. https://rads.dk/media/2086/lmr-tromboseprofylakse-knae-og-hoftekirurgimarts-2016.pdf Eikelboom J W, Quinlan D J, Douketis J D. Extended-duration prophylaxis against venous thromboembolism after total hip or knee replacement: a meta-analysis of the randomised trials. Lancet 2001; 358: 9-15. Falck-Ytter Y, Francis C W, Johanson N A, Curley C, Dahl O E, Schulman S, Ortel T L, Pauker S G, Colwell C W. Prevention of VTE in orthopedic surgery patients: Antithrombotic Therapy and Prevention of Thrombosis, 9th ed., American College of Chest Physicians Evidence-Based Clinical Practice Guidelines. Chest 2012; 141: e278S-e325S. Forster R, Stewart M. Anticoagulants (extended duration) for prevention of venous thromboembolism following total hip or knee replacement or hip fracture repair. Cochrane Database Syst Rev 2016; 3: CD004179. Furu K. Establishment of the nationwide Norwegian Prescription Database (NorPD)—new opportunities for research in pharmacoepidemiology in Norway. Norsk Epidemiologi 2008; 18: 129-36. Gomez D, Razmjou H, Donovan A, Bansal V B, Gollish J D, Murnaghan J J. A phase IV study of thromboembolic and bleeding events following hip and knee arthroplasty using oral Factor Xa inhibitor. J Arthroplasty 2017; 32: 958-64. Granan L P. Prevention of VTE in orthopedic surgery patients: a Norwegian adaptation of the 9th ed. of the ACCP Antithrombotic Therapy and Prevention of Thrombosis Evidence-based Clinical Practice Guidelines. Norsk Selskap for Trombose og Hemostase; 2015. Hull R D, Pineo G F, Stein P D, Mah A F, MacIsaac S M, Dahl O E, Butcher M, Brant R F, Ghali W A, Bergqvist D, Raskob G E. Extended out-ofhospital low-molecular-weight heparin prophylaxis against deep venous thrombosis in patients after elective hip arthroplasty: a systematic review. Ann Intern Med 2001; 135: 858-69. Huo M H. Post-discharge venous thromboembolism and bleeding in a large cohort of patients undergoing total hip or total knee arthroplasty. J Clin Outcomes Management 2012; 19: 355-63. Johannesdottir S A, Horvath-Puho E, Ehrenstein V, Schmidt M, Pedersen L, Sorensen H T. Existing data sources for clinical epidemiology: the Danish National Database of Reimbursed Prescriptions. Clin Epidemiol 2012; 4: 303-13. Nemes S, Gordon M, Rogmark C, Rolfson O. Projections of total hip replacement in Sweden from 2013 to 2030. Acta Orthop 2014; 85: 238-43. O’Donnell M, Linkins L A, Kearon C, Julian J, Hirsh J. Reduction of outof-hospital symptomatic venous thromboembolism by extended thromboprophylaxis with low-molecular-weight heparin following elective hip arthroplasty: a systematic review. Arch Intern Med 2003; 163(11): 1362-6. Pedersen A, Johnsen S, Overgaard S, Soballe K, Sorensen H, Lucht U. Registration in the Danish Hip Arthroplasty Registry: completeness of total hip arthroplasties and positive predictive value of registered diagnosis and postoperative complications. Acta Orthop Scand 2004; 75(4): 434-41. Pedersen A B, Johnsen S P, Sorensen H T. Increased one-year risk of symptomatic venous thromboembolism following total hip replacement: a nationwide cohort study. J Bone Joint Surg Br 2012; 94: 1598-1603. Pedersen A B, Sorensen H T, Mehnert F, Johnsen S P, Overgaard S. Efficacy and safety of short and extended thromboprophylaxis in 16,865 unselected hip replacement patients from routine clinical practice. Thromb Res 2015; 135(2): 322-8. Pedersen A B, Ehrenstein V, Szepligeti S K, Sorensen H T. Excess risk of venous thromboembolism in hip fracture patients and the prognostic impact of comorbidity. Osteoporos Int 2017; 28(12): 3421-30. Petersen P B, Kehlet H, Jorgensen C C. Lundbeck Foundation Centre for Fast-track & G. Knee Replacement Collaborative. Safety of in-hospital only thromboprophylaxis after fast-track total hip and knee arthroplasty: a prospective follow-up study in 17,582 procedures. Thromb Haemost 2018; 118(12): 2152-61. Sayers A, Evans J T, Whitehouse M R, Blom A Q. Are competing risks models appropriate to describe implant failure? Acta Orthop 2018; 89: 256–8. Schmidt M, Pedersen L, Sorensen H T. The Danish Civil Registration System as a tool in epidemiology. Eur J Epidemiol 2014; 29: 541-9.

Acta Orthopaedica 2019; 90 (4): 298â&#x20AC;&#x201C;305

Schmidt M, Schmidt S A, Sandegaard J L, Ehrenstein V, Pedersen L, Sorensen H T. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7: 449-90. Sobieraj D M, Lee S, Coleman C I, Tongbram V, Chen W, Colby J, Kluger J, Makanji S, Ashaye A O, White C M. Prolonged versus standard-duration venous thromboprophylaxis in major orthopedic surgery: a systematic review. Ann Intern Med 2012; 156: 720-7. Sundboll J, Adelborg K, Munch T, et al. Positive predictive value of cardiovascular diagnoses in the Danish National Patient Registry: a validation study. BMJ Open 2016; 6(11): e012832.

305

Thygesen S K, Christiansen C F, Christensen S, Lash T L, Sorensen H T. The predictive value of ICD-10 diagnostic coding used to assess Charlson comorbidity index conditions in the population-based Danish National Registry of Patients. BMC Med Res Methodol 2011; 11: 83. Trevisol A R, Coppi E F M, Pancotte J, Bortoluzzi E C, Knop G P. Use of oral anticoagulants for the prevention of thromboembolic events in the postoperative period of hip arthroplasty: a systematic review. Rev Bras Ortop 2018; 53(5): 515-20. Wolf B R, Lu X, Li Y, Callaghan J J, Cram P. Adverse outcomes in hip arthroplasty: long-term trends. J Bone Joint Surg Am 2012; 94: e103.

306

Acta Orthopaedica 2019; 90 (4): 306–311

Minor influence of patient education and physiotherapy interventions before total hip replacement on patient-reported outcomes: an observational study of 30,756 patients in the Swedish Hip Arthroplasty Register Christopher TORISHO 1, Maziar MOHADDES 1,2, Kristin GUSTAFSSON 3, and Ola ROLFSON 1,2 1 Department

of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg; 2 Swedish Hip Arthroplasty Register, Gothenburg; 3 Division of Physiotherapy, Department of Medical and Health Sciences, Linköping University, Linköping, Sweden Correspondence: christopher.torisho@vgregion.se Submitted 2018-11-30. Accepted 2019-03-08

Background and purpose — It is unclear whether physiotherapy interventions or patient education before total hip replacement (THR) is beneficial for patients postoperatively. Utilizing the Swedish Hip Arthroplasty Register (SHAR), we retrospectively studied the influence of preoperative selfreported exposure to physiotherapy and/or patient education on patient-reported outcomes 1 year after THR. Patients and methods — Data covering all THRs performed in Sweden for osteoarthritis, between the years 2012 and 2015, was obtained from SHAR. There were 30,756 patients with complete data. Multiple linear regression modelling was performed with 1-year postoperative PROMs (hip pain on a visual analogue scale [VAS], with the quality of life measures EQ-5D index and EQ VAS, and surgery satisfaction VAS) as dependent variables. Self-reported physiotherapy and patient education (yes or no) were used as independent variables. Results — Physiotherapy was associated with slightly less pain VAS (–0.7, 95% CI –1.1 to –0.3), better EQ-5D index (0.01, CI 0.00–0.01), EQ VAS (0.8, CI 0.4–1.2), and better satisfaction VAS (–0.7, CI –1.2 to –0.2). Patient education was associated with slightly better EQ-5D index (0.01, CI 0.00–0.01) and EQ VAS (0.7, CI 0.2–1.1). Interpretation — Even though we found statistically significant differences in favor of physiotherapy and patient education, the magnitude of those were too small and inconsistent to conclude a truly positive influence. Further research is needed with more specific and demarcated physiotherapy interventions.

Physiotherapy, in the form of supervised exercise, has been shown to reduce pain and improve function as assessed with patient-reported outcome measures (PROMs) for hip OA patients (Hernandez-Molina et al. 2008, Fransen et al. 2014), including later stages of the disease (Rooks et al. 2006, Villadsen et al. 2014). Self-management for knee and hip osteoarthritis improves pain (Chodosh et al. 2005) up to 21 months after intervention (Kroon et al. 2014), although both studies questioned the clinical relevance due to limited effect size. A recent meta-analysis including 13 RCTs reported a positive effect of preoperative exercise and patient education on postoperative pain for hip OA (Moyer et al. 2017). The meta-analysis found large differences in published studies with regard to interventions and minimal reporting on confounders. In Sweden, core treatment of OA is standardized in a national educational self-management programme for hip and knee OA patients, the Supported Osteoarthritis Self-Management Programme (SOASP) (Thorstensson et al. 2015). SOASP has the intent to increase quality of life during the course of the disease. Patients participating in the programme meet at group sessions and are taught about their disease, how to manage and cope with OA symptoms, and the rationale for exercising, by physiotherapists or occupational therapists. Participants are also offered individually adapted physical exercises, to be carried out in group training sessions or individually. The National Board of Health and Welfare of Sweden recommends non-surgical treatment options before listing OA patients for total joint replacement (Socialstyrelsen 2014). According to data from the Swedish Hip Arthroplasty Register 2015, the proportion of hip OA patients that have visited a physiotherapist (47–89%) or taken part in the SOASP (10–63%) prior

Acta Orthopaedica 2019; 90 (4): 306–311

to THR, varies widely between different regions in Sweden (Kärrholm et al. 2016). We investigated the influence of self-reported exposure to physiotherapy and/or patient education before THR on patient-reported outcomes 1 year postoperatively. In addition we explored demographic differences in the patient groups receiving or not receiving physiotherapy and/or patient education.

Patients and methods This is an a posteriori study of observational routinely collected health data from the Swedish Hip Arthroplasty Register (SHAR) analyzing the influence of preoperative self-reported exposure to physiotherapy and/or patient education on patientreported outcomes 1 year after THR. The Swedish Hip Arthroplasty Register (SHAR) Data were obtained from the SHAR. This national joint registry has a coverage of all the hospitals performing hip replacements in the country and had a completeness of 98.3% of all total hip replacements performed in 2015 (Kärrholm et al. 2018). All data on primary THRs, including PROMs, are collected by the participating units, and entered into the register database using 2-factor authentication (Kärrholm et al. 2008). Patient-reported outcome measures in SHAR Since 2002, SHAR has gathered PROMs from THR patients. In conjunction with the preoperative visit, patients are asked to complete a short survey (paper and digital version available according to the unit’s preference) including the EuroQol 5 dimensions (EQ-5D), a hip pain visual analogue scale (pain VAS), and self-reported Charnley classification. At 1-year postoperative follow-up, the same pen-and-paper survey is sent by ordinary mail with the addition of a satisfaction item on a VAS. The SHAR PROMS program has been described in detail previously (Rolfson et al. 2011). EQ-5D measures health-related quality of life and consists of 2 parts. For the first part, we used the British value set to calculate the EQ-5D index, which ranges from –0.59 to 1.0, where 1.0 corresponds to perfect health and negative results to a state worse than death (Dolan and Roberts 2002). In the second part, the patient estimates his or her current health status on a 100-degree scale, where 0 corresponds to worst imaginable health and 100 to best imaginable health. With pain VAS, the patient estimates his or her current pain on a visual 100-degree scale, where 0 corresponds to no pain and 100 to worst imaginable pain. Satisfaction VAS measures patient satisfaction with the outcome of surgery on a visual 100-degree scale, where 0 corresponds to a completely satisfied patient and 100 to an unsatisfied patient.

307

In 2012, 2 questions (yes/no) were added to the preoperative survey: (1) “Have you been to see a physiotherapist for your hip during the period of hip problems?” and (2) “Have you taken part in a so-called SOASP (may have been many years before the operation for a shorter period of time) during the period of hip problems.” The response rate in 2014 was 85% for preoperative PROMs and 90% for the 1-year follow-up postoperative PROMs (Garellick et al. 2015). Patient selection Data retrieved from SHAR covered hip OA (ICD-10 codes M16*) patients who had undergone primary THR surgery (NOMESCO codes NFB29, 39, and 49) between 2012 and 2015, with the years covering all available data including physiotherapy, patient education, and 1-year postoperative PROMs. Data retrieval was done in March 2018 and included age, sex, surgery side, first or second surgery, BMI, ASA class, Charnley class, incision, fixation, patient education, and physiotherapy. In addition, all PROMs collected preoperatively and/or postoperatively (pain VAS, EQ-5D and EQ VAS, and postoperative satisfaction VAS) were included. We selected the surgeries where patients had their first primary THR, i.e., they had no previous hip replacement of their contralateral hip. Additional selection criteria were applied to exclude patients missing data (BMI, ASA class, incision, type of fixation, patient education, physiotherapy, preoperative and 1-year postoperative PROMs). Patients with extreme values (BMI < 15 and BMI > 50) were excluded since these were probably erroneously recorded. Statistics The software SPSS statistics version 25 (IBM Corp, Armonk, NY, USA) was used for all statistical analyses. The null hypothesis was rejected when p < 0.05. Continuous variables were compared by using paired t-test. Categorical variables were analyzed by conducting Pearson’s chi-square tests to check for statistical significance between the 2 groups. 95% confidence intervals (CI) were calculated when appropriate. Linear regression analysis The linear regression analyses were made by using a generalized linear model. A 95% confidence interval was used. The dependent variables used in the model were the postoperative PROMs: pain VAS, EQ-5D index, EQ VAS, and satisfaction VAS. The independent variables included were: age, sex, BMI, ASA class, Charnley class, incision, fixation, patient education, and physiotherapy. Also, for the preoperative pain VAS, EQ-5D index, and EQ VAS the corresponding preoperative variables were used as independent variables in the models. Non-respondent analysis 3 different non-respondent analyses were performed. First, patients with missing data on preoperative PROMs,

308

Acta Orthopaedica 2019; 90 (4): 306–311

All THR for OA 2012–2015 n = 54,167 Excluded (n = 23,411): – had second hip surgery, 13,537 — ”Missing 1” – missing preoperative PROMs and/or data on patient education and physiotherapy, 5,282 — ”Missing 2” – extreme BMI or missing data on BMI, ASA classification, surgical approach and/or fixation method, 1,203 – missing postoperative PROMs, 3,389 — ”Missing 3” Study group n = 30,756

Figure 1. The flowchart describes the selection of patients. As defined here, excluded patients form 3 groups: missing 1, missing 2, and missing 3, further investigated in the missing data/non-respondent analysis (see Appendix). Abbreviations: THR = total hip replacement, OA = osteoarthritis, BMI = body mass index, ASA = American Association of Anesthesiologists’ classification.

physiotherapy, and patient education were compared with the patients included in the current analysis (study group). Second, cases excluded due to patients having second hip surgery and, third, the group with missing postoperative PROMs were compared with the study group. The method used was ANOVA with post-hoc Tukey with continuous variables and Pearson’s chi-square test with categorical variables. Ethics, funding, and potential conflicts of interest This study is a part of a larger research project which has been reviewed and approved by the Regional Ethical Review Board in Gothenburg (2014-04-09, 271-14). The study was partly financed by grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement (ALFGBG–522591). The authors declare no conflicts of interest.

Results Demographics (Table 1) Of all the 54,167 cases obtained from SHAR, 30,756 (59%) met the selection criteria and were included in the regression analyses (Figure 1). Of these, 71% reported exposure to physiotherapy, patient education or both, prior to surgery. Among the study group, 68% reported exposure to physiotherapy and 27% reported exposure to patient education, prior to surgery. At baseline, patients exposed to PT/SOASP had a statistically significantly lower age, BMI, preoperative EQ-5D index, and preoperative EQ-VAS, but higher preoperative pain VAS when compared with patients not exposed, on average. The PT/ SOASP group also had a higher proportion of women, ASA I– II, and Charnley class A, and a lower proportion of cemented surgery.

Table 1. Demographics Variable

Study group

Physiotherapy and/ or patient education No Yes

p-value a

Total numbers 30,756 9,040 21,716 Age b 68 (9.9) 70 (10) 68 (9.7) < 0.01 Female c 17,127 (56) 4,250 (47) 12,877 (59) < 0.01 BMI b 27.3 (4.3) 27.4 (4.4) 27.2 (4.3) < 0.01 ASA I–II c 26,315 (86) 7,356 (81) 18,959 (87) < 0.01 Charnley class c < 0.01 A 14,946 (49) 4,372 (49) 10,574 (49) B 4,125 (13) 1,129 (13) 2,996 (14) C 11,685 (38) 3,539 (39) 8,146 (38) Incision c 0.4 Posterior 16,316 (53) 4,743 (53) 11,573 (53) Lateral 14,205 (46) 4,227 (47) 9,978 (46) Other 235 (0.8) 70 (0.8) 165 (0.8) Fixation c < 0.01 Cemented 19,339 (63) 5,967 (66) 13,372 (62) Uncemented 6,165 (20) 1,673 (19) 4,492 (21) Other 5,252 (17) 1,400 (16) 3,852 (18) Pain VAS b 63.2 (15.3) 62.7 (16.3) 63.4 (14.9) < 0.01 EQ-5D index b 0.42 (0.3) 0.43 (0.3) 0.42 (0.3) < 0.01 EQ VAS b 57.9 (22.1) 59.0 (21.8) 57.5 (22.2) < 0.01 a A 2-column Pearson’s chi-square test was used on the categorical variables. An independent sample t-test was used on the continuous variables. b Continuous variables, presented with frequency (standard deviation). c Categorical variables, presented with frequency (percentage).

Linear regression analysis (Figures 2–5) Physiotherapy was associated with 0.7 units lower pain VAS (CI –0.3, –1.1), 0.01 units higher EQ-5D index (CI 0.00, 0.01), 0.8 units higher EQ VAS (CI 0.4, 1.2), and 0.7 units lower (= better) score on the satisfaction VAS (CI –1.2, –0.2) postoperatively. Self-reported patient education was associated with better EQ-5D index and EQ VAS. Patient education was associated with 0.006 units higher EQ-5D (0.01, CI 0.00, 0.01) and 0.7 units higher EQ VAS (0.7, CI 0.2, 1.1). Patient education did not influence pain VAS (–0.3, CI –0.7, 0.2) or satisfaction VAS (0.1, CI –0.4, 0.6) postoperatively. Non-respondent and missing data analysis Compared with the study group, patients excluded due to having THR on their second hip had on average higher age, higher BMI, and higher preoperative pain VAS. They also had a higher proportion of females, Charnley class A, and cemented fixation, but a lower proportion of ASA I–II. Compared with the study group, patients with missing data on preoperative PROMs, physiotherapy, or patient education had on average higher age. They also had a higher proportion of cemented fixation as well as a lower proportion of ASA I–II and posterior incision. As for patients missing postoperative PROMs, they had on average lower age, higher BMI, higher preoperative pain VAS, lower EQ-5D index, and lower EQ-VAS, compared with the study group (Table 2, see Supplementary data).

Acta Orthopaedica 2019; 90 (4): 306–311

309

1-year postoperative EQ-5D

1-year postoperative pain Variable

Variable

Coefficient (CI)

Age BMI Sex Men Women ASA class 1–2 3–4 Charnley class A B C Surgical approach Posterior Lateral Other Fixation Cemented Uncemented Other Preoperative pain VAS Physiotherapy No Yes Patient eduction No Yes

Age BMI Sex Men Women ASA class 1–2 3–4 Charnley class A B C Surgical approach Posterior Lateral Other Fixation Cemented Uncemented Other Preoperative EQ-5D Physiotherapy No Yes Patient eduction No Yes

0.1 (0.0 to 0.1) 0.2 (0.1 to 0.2) ref 0.8 (0.4 to 1.2) ref 2.3 (1.8 to 2.9) ref 3.4 (2.8 to 3.9) 5.2 (4.7 to 5.6) ref 2.3 (1.9 to 2.7) 1.6 (–0.6 to 3.8) ref –1.8 (–2.4 to –1.2) –0.7 (–1.3 to –0.2) 0.1 (0.1 to 0.1) ref –0.7 (–1.1 to –0.3) ref –0.3 (–0.7 to 0.2)

–5

–3

–1

1-year postoperative EQ VAS

Variable

–0.2 (–0.2 to –0.1) –0.3 (–0.4 to –0.3)

Age BMI Sex Men Women ASA class 1–2 3–4 Charnley class A B C Surgical approach Posterior Lateral Other Fixation Cemented Uncemented Other Physiotherapy No Yes Patient eduction No Yes

ref –1.3 (–1.7 to –0.9) ref –6.2 (–6.8 to –5.7) ref –4.5 (–5.1 to –3.9) –8.9 (–9.3 to –8.5) ref –2.0 (–2.3 to –1.6) –2.6 (–4.8 to –0.5) ref 1.2 (0.7 to 1.8) 0.9 (0.4 to 1.5) 0.2 (0.2 to 0.2) ref 0.8 (0.4 to 1.2) ref 0.7 (0.2 to 1.1)

–8

–6

–4

–2

ref –0.057 (–0.064 to –0.050) ref –0.057 (–0.063 to –0.050) –0.11 (–0.11 to –0.10) ref –0.028 (–0.033 to –0.023) –0.013 (–0.041 to 0.015) ref 0.022 (0.014 to 0.029) 0.0093 (0.0022 to 0.016) 0.13 (0.12 to 0.14) ref 0.0061 (0.0006 to 0.012) ref 0.0060 (0.0003 to 0.012)

Figure 4. Linear regression results with the dependent variable EQ VAS 1 year postoperatively.

Discussion In this study based on a national registry, patients visiting a physiotherapist at some point during the course of their disease had statistically significantly better 1-year postoperative PROMs. However, the positive influence was only on a par with or below the smallest factors the model was adjusted for, making the clinical relevance of the results uncertain. For patients with hip OA, physiotherapy in the form of supervised exercise has been shown to reduce pain and improve function (Hernandez-Molina et al. 2008, Fransen et al. 2014). A few authors have also reported beneficial preoperative effects of supervised exercise for patients awaiting surgery (Wallis et al. 2011, Gill et al. 2013). A recent meta-analysis by Moyer et al. (2017) evaluating postoperative effects after preoperative supervised exercise concluded that there were improvements for pain, function, and length of stay. One of the included studies (Rooks et al. 2006) did not show any positive postoperative effects on PROMs and another RCT found no effects lasting past 6 weeks postoperatively following total

–0.1

0.1

0.2

1-year postoperative satisfaction

Coefficient (CI)

Age BMI Sex Men Women ASA class 1–2 3–4 Charnley class A B C Surgical approach Posterior Lateral Other Fixation Cemented Uncemented Other Preoperative EQ VAS Physiotherapy No Yes Patient eduction No Yes

–10

ref –0.031 (–0.036 to –0.026)

Figure 3. Linear regression results with the dependent variable EQ-5D 1 year postoperatively.

Variable

–12

–0.0012 (–0.0015 to –0.0009) –0.0042 (–0.0047 to –0.0036)

–0.2

Figure 2. Linear regression results with the dependent variable pain VAS 1 year postoperatively.

–14

Coefficient (CI)

Coefficient (CI) 0.1 (0.1 to 0.2) 0.2 (0.1 to 0.2) ref 1.9 (1.5 to 2.4) ref 2.6 (2.0 to 3.2) ref 2.1 (1.4 to 2.8) 4.4 (3.9 to 4.8) ref 4.0 (3.6 to 4.4) 0.5 (–2.0 to 3.0) ref –2.2 (–2.9 to –1.6) –1.1 (–1.7 to –0.5) ref –0.7 (–1.2 to –0.2) ref 0.1 (–0.4 to 0.6)

–7

–5

–3

–1

Figure 5. Linear regression results with the dependent variable satisfaction 1 year postoperatively.

knee and hip arthroplasty (Villadsen et al. 2014). There is 1 small RCT with 21 participants that showed better postoperative PROMs for THR patients following preoperative supervised exercise (Ferrara et al. 2008). In that study, patients in the exercise group had a statistically significantly (0.97 points) lower pain VAS (scale 0–10) 3 months postoperatively compared with controls, though also had a non-significant lower pain VAS of 0.62 at baseline. Compared with our postoperative VAS difference of 0.69 (scale 0–100), their result was more than 10 times larger. Our study shows a statistically significant positive association with health-related quality of life as measured with EQ-5D and EQ VAS but not for pain and satisfaction in patients participating in preoperative patient education. There is a lack of larger RCTs (Wang et al. 2016) investigating the role of patient education during the course of the disease in postoperative outcomes. A few review articles (Ibrahim et al. 2013, McDonald et al. 2014, Aydin et al. 2015) have tried to analyze association between surgery-oriented preopera-

310

tive patient education and postoperative outcomes with only 1 (Ibrahim et al. 2013) concluding that preoperative patient education resulted in better postoperative PROMs. The contradictory findings in the literature are partly reflected by our findings with a slight improvement in quality of life but no effect on pain or satisfaction in patients operated with THR following preoperative patient education. There are limitations in this study. First, the physiotherapy interventions were not defined. Some patients reporting they have visited a physiotherapist might have received only non-exercise-based treatment, which potentially may reduce the association we found between physiotherapy and postoperative PROMs. The recommended physiotherapy-based treatment by the Swedish National Board of Health Welfare is long-term exercise (Socialstyrelsen 2012). However, the patients might have visited their physiotherapist before those recommendations were published and/or before receiving other interventions. In addition, we are not aware of how quickly and to what degree those recommendations have been implemented in Sweden. Second, we do not know when the patients received their physiotherapy interventions, how regularly, to what intensity, or their compliance. This could affect our results. While there is a lack of validated evidence of preoperative exercise-based before joint replacement (Hoogeboom et al. 2012), the RCT in the field that has seen postoperative effects from preoperative physiotherapy has administered the interventions within a month from surgery and 5 times a week (Ferrara et al. 2008). The recruitment rate for physiotherapy within weeks before total joint replacement can be as low as 12% (Rooks et al. 2006) or 34% (Hoogeboom et al. 2010), with difficult transportation to the sessions a common complaint (Rooks et al. 2006, Hoogeboom et al. 2010). As 68% of the patients in our study group had answered “yes” to having been exposed to physiotherapy interventions, it is more likely that they have been exposed during earlier stages of the disease, with the intent of reducing OA symptoms. The third limitation is the lack of information regarding to what degree patients have taken part in rehabilitation following THR. Geographic areas that have a higher availability of physiotherapy and SOASP might also have different availability of postoperative rehabilitation, potentially affecting our results. The fourth limitation pertains to the demographic differences between the no PT/SOASP group and PT/SOASP group. 6 preoperative variables were favorable for the PT/ SOASP group according to the factors’ coefficients on our regression analysis: age, BMI, ASA I–II, Charnley class, incision, and fixation. 4 preoperative variables were favorable for the no PT/SOASP group: sex, preoperative pain VAS, EQ-5D index, and EQ VAS. Though all those factors are adjusted for in the linear regression model, there is still a risk of having the results derive from confounding factors not accounted for in the modelling. Finally, there is also

Acta Orthopaedica 2019; 90 (4): 306–311

always a risk, when excluding patients, that the results do not represent the reality. The 3 excluded groups had worse baseline variables compared with the analysis group, predicting worse postoperative PROMs. If the patients in the excluded groups had had complete data and had been included in the analysis group, we would probably have seen overall worse postoperative PROMs. How the factors in the linear regression models would have been affected is, however, hard to forecast. While much of the earlier focus has been on education and physiotherapy in close proximity to THR, our study indicates that interventions at some point during the course of OA have a positive influence on PROMs after surgery. Due to this study being observational, we cannot establish the causal relationships. Although earlier studies have not demonstrated lasting effects of physiotherapy post-THR, the influence observed in our study may be explained by increased compliance with supervised exercise after surgery. In OA patients without joint replacement, a previous systematic review article demonstrated the effect of supervised exercise lasting past 6 months with additional “booster” sessions with physiotherapists (Pisters et al. 2007). This could possibly be translatable for patients who have undergone THR. Further studies with more specific questions of supervised exercises before and after surgery could increase our understanding. Larger RCTs further exploring specific preoperative exercise interventions and their effect postoperatively are also needed. In summary, our study indicates that exposure to physiotherapy at some point during the course of OA has a small positive influence on 1-year postoperative PROMs after THR. Due to demographic differences, and uncertainties regarding the type of physiotherapy interventions and time frame, the clinical relevance of this small influence is uncertain. Therefore, the results should be interpreted with care. Further research is needed with more specific and demarcated physiotherapy interventions. Supplementary data Table 2 is available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674.2019. 1605669

The authors would like to thank all the surgeons and secretaries reporting to and coordinators maintaining the high quality and integrity of data being reported to the Swedish Hip Arthroplasty register. CT, OR, and MM conceived and planned the study. CT performed statistical analyses. CT drafted the manuscript with subsequent substantial inputs from all co-authors. Acta thanks Allan Abbot and Siri Bjørgen Wintherfor help with peer review of this study.

Acta Orthopaedica 2019; 90 (4): 306–311

Aydin D, Klit J, Jacobsen S, Troelsen A, Husted H. No major effects of preoperative education in patients undergoing hip or knee replacement: a systematic review. Dan Med J 2015; 62(7). pii: A5106. Chodosh J, Morton S C, Mojica W, Maglione M, Suttorp M J, Hilton L, Rhodes S, Shekelle P. Meta-analysis: chronic disease self-management programs for older adults. Ann Intern Med 2005; 143(6): 427-38. Dolan P, Roberts J. Modelling valuations for EQ-5D health states: an alternative model using differences in valuations. Med Care 2002; 40(5): 442-6. Ferrara P E, Rabini A, Maggi L, Piazzini D B, Logroscino G, Magliocchetti G, Amabile E, Tancredi G, Aulisa A G, Padua L, Aprile I, Bertolini C. Effect of pre-operative physiotherapy in patients with end-stage osteoarthritis undergoing hip arthroplasty. Clin Rehabil 2008; 22(10-11): 977-86. doi: 10.1177/0269215508094714. Fransen M, McConnell S, Hernandez-Molina G, Reichenbach S. Exercise for osteoarthritis of the hip. Cochrane Database Syst Rev 2014(4): Cd007912. doi: 10.1002/14651858.CD007912.pub2. Garellick G, Kärrholm J, Lindahl H, Malchau H, Rogmark C, Rolfson O. The Swedish Hip Arthroplasty Register Annual Report 2014; 2015. doi: 10.18158/B1OyzZ00Z. Gill S D, McBurney H. Does exercise reduce pain and improve physical function before hip or knee replacement surgery? A systematic review and meta-analysis of randomized controlled trials. Arch Phys Med Rehabil 2013; 94(1): 164-76. doi: 10.1016/j.apmr.2012.08.211. Hernandez-Molina G, Reichenbach S, Zhang B, Lavalley M, Felson D T. Effect of therapeutic exercise for hip osteoarthritis pain: results of a metaanalysis. Arthritis Rheum 2008; 59(9): 1221-8. doi: 10.1002/art.24010. Hoogeboom T J, Dronkers J J, van den Ende C H, Oosting E, van Meeteren N L. Preoperative therapeutic exercise in frail elderly scheduled for total hip replacement: a randomized pilot trial. Clin Rehabil 2010; 24(10): 901-10. doi: 10.1177/0269215510371427. Hoogeboom T J, Oosting E, Vriezekolk J E, Veenhof C, Siemonsma P C, de Bie R A, van den Ende C H, van Meeteren NL. Therapeutic validity and effectiveness of preoperative exercise on functional recovery after joint replacement: a systematic review and meta-analysis. PLoS One 2012; 7(5): e38031. doi: 10.1371/journal.pone.0038031. Ibrahim M S, Twaij H, Giebaly D E, Nizam I, Haddad F S. Enhanced recovery in total hip replacement: a clinical review. Bone Joint J 2013; 95-b(12): 1587-94. doi: 10.1302/0301-620x.95b12.31303. Kroon F P, van der Burg L R, Buchbinder R, Osborne R H, Johnston R V, Pitt V. Self-management education programmes for osteoarthritis. Cochrane Database Syst Rev 2014(1): Cd008963. doi: 10.1002/14651858. CD008963.pub2. Kärrholm J, Garellick G, Rogmark C, Herberts P. The Swedish Hip Arthroplasty Register Annual Report 2007; 2008. doi: 10.18158/B1OyzZ00Z.

311

Kärrholm J, Lindahl H, Malchau H, Mohaddes M, Nemes S, Rogmark C, Rolfson O. Swedish Hip Arthroplasty Register Annual Report 2016; 2018. doi: 10.18158/SJy6jKyrM. McDonald S, Page M J, Beringer K, Wasiak J, Sprowson A. Preoperative education for hip or knee replacement. Cochrane Database Syst Rev 2014(5): CD003526. doi: 10.1002/14651858.CD003526.pub3. Moyer R, Ikert K, Long K, Marsh J. The value of preoperative exercise and education for patients undergoing total hip and knee arthroplasty: a systematic review and meta-analysis. JBJS Rev 2017; 5(12): e2. doi: 10.2106/ jbjs.rvw.17.00015. Pisters M F, Veenhof C, van Meeteren N L, Ostelo R W, de Bakker D H, Schellevis F G, Dekker J. Long-term effectiveness of exercise therapy in patients with osteoarthritis of the hip or knee: a systematic review. Arthritis Rheum 2007; 57(7): 1245-53. doi: 10.1002/art.23009. Rolfson O, Karrholm J, Dahlberg L E, Garellick G. Patient-reported outcomes in the Swedish Hip Arthroplasty Register: results of a nationwide prospective observational study. J Bone Joint Surg Br 2011; 93(7): 867-75. doi: 10.1302/0301-620x.93b7.25737. Rooks D S, Huang J, Bierbaum B E, Bolus S A, Rubano J, Connolly C E, Alpert S, Iversen M D, Katz J N. Effect of preoperative exercise on measures of functional status in men and women undergoing total hip and knee arthroplasty. Arthritis Rheum 2006; 55(5): 700-8. doi: 10.1002/art.22223. Socialstyrelsen S. Nationella riktlinjer—Utvärdering—Vård vid rörelseorganens sjukdomar—Indikatorer och underlag för bedömning.: Socialstyrelsen; 2014. Socialstyrelsen S. Nationella riktlinjer för rörelseorganens sjukdomar 2012: osteoporos, artros, inflammatorisk ryggsjukdom och ankyloserande spondylit, psoriasisartrit och reumatoid artrit : stöd för styrning och ledning. Socialstyrelsen; 2012. Thorstensson C A, Garellick G, Rystedt H, Dahlberg L E. Better management of patients with osteoarthritis: development and nationwide implementation of an evidence-based supported osteoarthritis self-management programme. Musculoskeletal Care 2015; 13(2): 67-75. doi: 10.1002/msc.1085. Villadsen A, Overgaard S, Holsgaard-Larsen A, Christensen R, Roos E M. Postoperative effects of neuromuscular exercise prior to hip or knee arthroplasty: a randomised controlled trial. Ann Rheum Dis 2014; 73(6): 1130-7. doi: 10.1136/annrheumdis-2012-203135. Wallis J A, Taylor N F. Pre-operative interventions (non-surgical and nonpharmacological) for patients with hip or knee osteoarthritis awaiting joint replacement surgery: a systematic review and meta-analysis. Osteoarthritis Cartilage 2011; 19(12): 1381-95. doi: 10.1016/j.joca.2011.09.001. Wang L, Lee M, Zhang Z, Moodie J, Cheng D, Martin J. Does preoperative rehabilitation for patients planning to undergo joint replacement surgery improve outcomes? A systematic review and meta-analysis of randomised controlled trials. BMJ Open 2016; 6(2): e009857. doi: 10.1136/bmjopen-2015-009857.

312

Acta Orthopaedica 2019; 90 (4): 312–317

Survival of 11,390 Continuum cups in primary total hip arthroplasty based on data from the Finnish Arthroplasty Register Matias HEMMILÄ 1, Mikko KARVONEN 1, Inari LAAKSONEN 1, Markus MATILAINEN 2, Antti ESKELINEN 3, Jaason HAAPAKOSKI 4, Ari-Pekka PUHTO 5, Jukka KETTUNEN 6, Mikko MANNINEN 7, and Keijo T MÄKELÄ 1 1 Department

of Orthopaedic Surgery, University of Turku and Turku University Hospital, Turku; 2 Department of Biostatistics, University of Turku, Turku; 3 Coxa Hospital for Joint Replacement, Tampere; 4 National Institute for Health and Welfare, Helsinki; 5 Department of Orthopaedics and Traumatology, Oulu University Hospital, Oulu; 6 Department of Orthopaedics and Traumatology, Kuopio University Hospital, Kuopio; 7 Orton Hospital, Helsinki, Finland Correspondence: matias.hemmila@utu.fi Submitted 2018-11-29. Accepted 2019-03-12.

Background and purpose — The use of trabecular metal (TM) cups for primary total hip arthroplasty (THA) is increasing. Some recent data suggest that the use of TM in primary THA might be associated with an increased risk of revision. We compared implant survival of Continuum acetabular cups with other commonly used uncemented cups. Patients and methods — Data on 11,390 primary THAs with the Continuum cup and 30,372 THAs with other uncemented cups (reference group) were collected from the Finnish Arthroplasty Register. Kaplan–Meier survival estimates were calculated; the endpoint was revision for any reason, for infection, or for dislocation. Revision risks were assessed with adjusted Cox multiple regression models. A subgroup analysis on the use of neutral or elevated liners in the Continuum group was made. Results — The 7-year survivorship of the Continuum group was 94.6% (95% CI 94.0–95.2) versus 95.6% (CI 95.3–95.8) in the reference group for revision for any reason. The risk for revision was higher in the Continuum group than in the reference group both for revision for any reason (HR 1.3 [CI 1.2–1.5)]) and for revision for dislocation (HR 1.9 [CI 1.5–2.3]). There was no difference in the rates of revision because of infection (HR 0.99 [CI 0.78–1.3]). Use of a neutral liner increased the risk for revision due to dislocation in comparison with the use of an elevated rim liner in the Continuum group (HR 1.7 [CI 1.2–2.5]). Interpretation — THA with Continuum cups is associated with an increased risk of revision compared with other uncemented cups, mainly due to revisions because of dislocation. Our results support the use of an elevated liner when Continuum cups are used for primary THA.

Trabecular metal (TM) acetabular components were initially indicated in particular for cup revisions after total hip arthroplasty (THA) (Levine et al. 2006). TM cups provide increased bone ingrowth, better modulus of elasticity, and better stability due to their porous structure compared with conventional uncemented cup devices made of titanium alloy (Meneghini et al. 2010). Currently, TM revision cups are used frequently worldwide. Besides revision surgery, TM cups have demonstrated promising mid- to long-term survivorship in primary THA (Baad-Hansen et al. 2011, Howard et al. 2011) and hence the use of Continuum (ZimmerBiomet, Warsaw, IN, USA) TM cups in primary THA increases (Wegrzyn et al. 2015, De Martino et al. 2016). However, a recent register study showed that the early and mid-term revision rate of TM cups was slightly higher compared with other uncemented cups when used in primary THA in Sweden and Australia (Laaksonen et al. 2018). The revision rate due to periprosthetic joint infection has been slightly increasing during recent decades (Dale et al. 2012). Some data suggest that the use of a TM acetabulum component in hip revision arthroplasty might be associated with a lower infection rate (Tokarski et al. 2015), but this finding has not been confirmed by register data and thus far there is, to our knowledge, no other evidence that the material of TM would protect patients from prosthetic joint infection (PJI) (Laaksonen et al. 2017, 2018). It has also been suggested that there might be an increased risk of dislocations associated with the use of Continuum cups due to a decreased jumping distance of the femoral head (Pakarinen et al. 2018). To compensate for this circumstance, elevated or hooded acetabular liners are currently widely

Acta Orthopaedica 2019; 90 (4): 312–317

313

Table 1. Acetabular cups included in the study Cup design

n (%)

Continuum (ZimmerBiomet, Warsaw, IN, USA) 11,390 (27) Reference group 30,372 (73) Exceed (ZimmerBiomet, Warsaw, IN, USA) 1,550 (4) G7 (ZimmerBiomet, Warsaw, IN, USA) 1,121 (3) Pinnacle (DePuy, Warsaw, IN, USA) 14,844 (36) R3 (Smith & Nephew, Andover, MA, USA) 7,289 (18) Trident (shell) (Stryker, Mahwah, NJ, USA) 4,279 (10 Vision Ringloc (ZimmerBiomet, Warsaw, IN, USA) 1,280 (3)

available for the purpose of decreasing the dislocation rate. The use of elevated liners may, in theory, improve hip stability and decrease the revision rate (Insull et al. 2014), but the routine use of elevated liners has been questioned (Krushell et al. 1991, Insull et al. 2014). We compared the revision rate of Continuum cups used in primary THA with the most commonly used uncemented cups made of titanium alloy. We specifically compared the revision rates (1) for any reason, (2) for infection, and (3) for dislocation. We also studied whether use of elevated liners in primary THA decreases the revision rate due to dislocations compared with the Continuum cup with neutral liners.

Patients and methods This study is based on data from the Finnish Arthroplasty Register (FAR). The FAR data include nearly all hip and knee implants operated in Finland since 1980 (Paavolainen et al. 1991). Orthopedic units are obligated to provide all information essential for maintenance of the register to the Finnish National Institute for Health and Welfare. The register gathers information from most total hip implantations in the entire country and data coverage on primary THA exceeds 95% and on revision THA coverage is 81% (FAR 2018). Dates of death are obtained from the Population Information System maintained by the Population Register Centre. The data content of the register was scrutinized and revised in May 2014. The updated data now include detailed information on items like ASA class, BMI, and surgical approach. Study population Between January 2009 and December 2017, 133,488 primary THAs were reported to FAR. In 11,390 of these the Continuum primary cup was used. The reference group consisted of the 6 most commonly used other uncemented cups made of titanium alloy (n = 30,372) (Table 1). A head size other than 28 mm, 32 mm, or 36 mm, dual mobility, and constrained liners were excluded. The number of patients with bilateral hip prostheses was 4,407 and in 658 patients both hips were operated simultaneously. 498 patients had the Continuum cup in one hip and a control group cup component in the contra-

Table 3. Demographic data of the time period after data content revision in the Finnish Arthroplasty Register starting May 15, 2014. Values are frequency (%) unless stated otherwise Data

Continuum Reference group group

Mean age (SD) 67 (11) 66 (11) BMI (SD) 28 (5) 28 (5) Male sex 3,609 (42) 7,547 (46) Diagnosis Primary osteoarthritis 7,324 (85) 13,852 (85) Rheumatoid arthritis 137 (2) 195 (1) Other a 1,113 (13) 2,278 (14) Femoral head size of prosthesis 28 mm 29 (0.3) 107 (1) 32 mm 1,832 (21) 3,369 (21) 36 mm 6,713 (78) 12,849 (79) Status at end of follow-up Not revised 8,202 (96) 15,792 (97) Revised 372 (4) 533 (3) Liner material Ceramic 619 (7) 2,249 (14) Highly cross-linked polyethylene 7,955 (93) 14,041 (86) Elevated liner No 4,385 (55) 8,648 (62) Yes 3,570 (45) 5,393 (38) Approach Posterior 6,654 (78) 12,884 (81) Anterolateral (modified Hardinge) 1,667 (20) 2,864 (18) Anterior (Watson-Jones) 15 (0.2) 11 (0.1) Anterior (Smith-Peterson) 143 (2) 137 (1) Trochanteric osteotomy performed 1 (0.01) 1 (0.01) ASA class 1 1,281 (15) 2,163 (14) 2 4,132 (49) 8,260 (52) 3 2,992 (35) 5,308 (33) 4 104 (1) 189 (1) Femoral stem fixation Uncemented 5,502 (65) 13,209 (81) Cemented 3,030 (36) 3,057 (19) a Fractures (5% Continuum group vs. 4% control group), avascular necrosis (3% vs. 3 %), osteoarthritis due to hip dysplasia (2% vs. 2 %), tumors (1% vs. 1 %), congenital hip dislocation (0.5% vs. 0.3%), inflammatory arthritis (0.3% vs. 0.4%), Legg–Perthes–Calve disease (0.3% vs. 0.2%), femoral head epiphyseolysis (0.2% vs. 0.1%), status post purulent arthritis (0.1% vs. 0%).

lateral hip. Tables 2 (see Supplementary data) and 3 show the demographic data hip-wise separately for the whole study period and after the data content revision in May 2014. Mortality during the study period in the Continuum group was 4% and 5% in the control group. Surgery In the Continuum group, 36 mm femoral heads were used in 79% of cases. The corresponding proportion in the reference group was 80%. A ceramic liner was used in 14% of cases in the Continuum group and in 27% of cases in the reference group. The rest were highly cross-linked polyethylene liners in both groups. From May 2014 surgical approach data have been available from the register. Since then the majority of the operations have been performed via the posterior approach in

314

Table 4. Indication for revision prior to data content revision (May 15, 2014) of the Finnish Register. Values are frequency (%) Continuum Reference Main reason for revision a group group Aseptic loosening Cup and stem 0 (0) 2 (1) Cup 2 (4) 15 (5) Stem 3 (6) 17 (5) Infection 7 (13) 56 (18) Dislocation 21 (39) 88 (28) Component malposition 3 (6) 31 (10) Fracture 13 (24) 72 (23) Component breakage 0 ( 4 (1) Other 5 (9) 28 (9) a

No data available concerning indication for revision from 83 revisions.

both groups (79% in the Continuum group and 81% in the reference group). Uncemented femoral stems were used in 71% in the Continuum group compared with 83% in the reference group. The average follow-up time was 3 years (0–9) in the TM group and 4 years (0–10) in the reference group. Statistics Kaplan–Meier survival estimates were calculated for both groups and the log rank test was used to compare the survival curves. Revision was described as change or removal of at least 1 component (Tables 4 and 5). To reduce the risk of selection bias we adjusted the estimated revision risks in the Cox multiple regression model by sex, age group, diagnosis, femoral head size, operated side, operation year, and fixation of the femoral stem. An additional cup revision analysis was performed and the type of approach, ASA, BMI, and elevation status of the liner were added to the Cox model as possible confounders for cup revision for any reason as the endpoint. The analysis was done with the data of primary operation after register update in May 2014. In the Continuum elevation subgroup analysis sex, age group, diagnosis, side, stem fixation, and operation year were added to the Cox model (head size was stratified) and other than polyethylene liners were excluded. If the proportional hazards assumption for a variable was not fulfilled in the Cox model, the model was stratified by it instead. Stratification in Cox models means that the hazard functions can be estimated for all level combinations of the stratified variables, and the hazard ratios for the other variables (those that meet the proportional hazard assumption) are then optimized for all these hazard functions. Without stratification we would assume that hazards were the same for all levels of such variables. The primary outcome was revision for any reason and the secondary outcomes were revision for periprosthetic infection, revision for dislocation, and cup revision for any reason. Patients were censored for any event other than the outcome, or at the end of the follow-up. After the register update in May 2014 it has been possible to assess separately which component has been changed or removed in connection with the revi-

Acta Orthopaedica 2019; 90 (4): 312–317

Table 5. Indication for revision after new indications for revision were added at the data content revision (May 15, 2014) of the Finnish Register. Values are frequency (%) Continuum Reference Main reason for revision a group group Aseptic loosening Cup 5 (1) 10 (2) Stem 15 (4) 26 (4) Osteolysis Cup 2 (1) 8 (1) Stem 1 (0.3) 11 (2) Liner wear 0 (0) 2 (2) Component breakage Cup 0 (0) 1 (0.2) Liner 1 (0.3) 11 (2) Head 1 (0.3) 1 (0.2) Modular neck 0 (0) 1 (0.2) Infection 100 (26) 194 (30) Dislocation 132 (34) 153 (24) Component malposition Cup 12 (3) 23 (4) Stem 1 (0.3) 14 (2) Periprosthetic fracture Acetabulum 6 (2) 2 (0.3) Femur 73 (19) 105 (17) Adverse reaction to metal debris 2 (1) 5 (1) Squeaking 2 (1) 5 (1) Unexplained pain 10 (3) 32 (5) Leg length discrepancy repair 4 (1) 10 (2) Other 17 (4) 24 (4) a

No data available concerning indication for revision from 83 revisions.

sion. Therefore, a subgroup analysis for cup-only-revisions was performed only for the newest FAR data. In addition, a subgroup analysis was performed for Continuum cups by liner type (neutral or elevated liner) with dislocation revision as the endpoint. Survival data are presented as percentages with the 95% confidence interval (CI). Cox regression analysis is presented with the hazard ratio (HR) and the CI. All analyses were performed using the SAS software (Version 9.3; SAS Institute, Cary, IN, USA). Ethics, funding, and potential conflicts of interest Ethical approval: June 13, 2017, Dnor THL/926/5.05.00/2017. This research received no funding. The authors declare no conflicts of interest.

Results Revision for any reason The up to 7-year survivorship for the Continuum group was 94.6% (CI 94.0–95.2) and the survival for the reference group was 95.6% (CI 95.3–95.8) for revision for any reason as an endpoint (Figure 1, Table 6; see Supplementary data). By Cox regression analysis the Continuum group had an increased risk of revision for any reason compared with the reference group (HR 1.3, CI 1.2–1.5).

Acta Orthopaedica 2019; 90 (4): 312–317

315

Revisionfree survival probablity

Survival probablity – endpoint revision for infection

Survival probablity – endpoint revision for dislocation

1.00

1.000

CUP TYPE Continuum Others

0.998

0.98

CUP TYPE Continuum Others

Elevation Yes No

0.995

0.996

0.990 0.994 0.990

0.96

0.985

0.992 0.980

0.990

0.985

0.94

0.975

0.988

0.986

0.92 0

Years after index operation

Figure 1. Kaplan–Meier survival for Continuum group and reference group with revision for any reason as the endpoint. 95% CI levels presented in blue and red.

0.980 0

Years after index operation

Figure 2. Kaplan–Meier survival for Continuum group and reference group with revision for infection as endpoint. 95% CI levels presented in blue and red.

Cup revision for any reason In the cup-only-revision analysis performed with the data from May 15, 2014 to December 31, 2017, the 3-year survivorship was the same in the Continuum group as in the reference group: 99.4% vs. 99.6% (CI 99.2–99.6 vs. 99.5–99.7). These figures are not statistically different (Cox regression analysis HR 1.3, Cl 0.8–2.0). Revision due to infection The 7-year survivorship for the Continuum group was 98.9% (CI 98.6–99.1) and for the reference group 99.1% (CI 99.0– 99.2), when revision because of infection was the endpoint (Figure 2). The risk of revision for infection was the same in the groups (HR 1.0, CI 0.8–1.3) (Table 7; see Supplementary data). Revision due to dislocation The 7-year survivorship for the Continuum group was 98.3% (CI 98.0–98.6) and for the reference group 99.0% (CI 98.8– 99.1), when revision because of dislocation was the endpoint (Figure 3). The Continuum group had an increased risk of revision for dislocation (HR 1.9, CI 1.5–2.3) compared with the reference group (Table 7; see Supplementary data). Subgroup analysis: Continuum THA with or without liner elevation The 5-year survivorship for the Continuum group with elevated liners was 98.9% (CI 98.4–99.2) and for the Continuum group with neutral liners 97.8% (CI 97.3–98.2), when revision because of dislocation was the endpoint (Figure 4). After adjustments of the statistical data, the Continuum group with neutral liners had a higher risk of revision for dislocation compared with the Continuum group with elevated liners (HR 1.7, CI 1.2–2.5).

0.970 0

Years after index operation

Figure 3. Kaplan–Meier survival for Continuum group and reference group with revision for dislocation as endpoint. 95% CI levels presented in blue and red.

Years after index operation

Figure 4. Kaplan–Meier survival by subgroup analysis of Continuum THA with or without elevated liner. Endpoint: revision for dislocations. 95% CI levels presented in blue and red.

Discussion This study shows that use of the Continuum THA is associated with a slightly higher risk of revision than use of other uncemented titanium alloy cups. The Continuum study group and the reference group had a similar risk of revision due to infection, but the risk of revision due to dislocation was higher in the Continuum group. Further, the use of elevated liners in the Continuum THA reduced the risk of revision for dislocation compared with neutral liners. Trabecular metal was first introduced to the market in 1997. Since then, TM cups have shown reliable results when used for hip revision arthroplasty and are currently used routinely worldwide (Davies et al. 2011, Mohaddes et al. 2015). Their routine use in primary THA is increasing. Implant survival of primary TM cups has been comparable or even superior compared with uncemented devices made of titanium alloy (Baad-Hansen et al. 2011, Howard et al. 2011, Wegrzyn et al. 2015, De Martino et al. 2016). However, a recent collaborative register study reported that TM cups have a 1.5 times higher risk for revision than other frequently used uncemented cups in primary THA (Laaksonen et al. 2018). These results were somewhat surprising and at variance with previous literature. Our study supports the previous finding from the Swedish and Australian registries of a higher risk of revision of TM cups. The use of TM cups in primary THA is increasing in Sweden and Australia (Laaksonen et al. 2018). Continuum was the 2nd most common cup design in the FAR data of the present study. Due to the good gription of and high primary stability of TM, Continuum cups have been preferred in more demanding THAs. To reduce the risk of selection bias towards more difficult cases being treated with Continuum cups, we adjusted the revision risks in the Cox regression models. Our data suggest that the use of the Continuum cup in primary THA does not give superior results compared with other uncemented devices. However, TM cups are a reliable option when

316

treating large bone defects in revision or complex primary THA and results in these cases have been excellent (Weeden and Schmidt 2007, Macheras et al. 2010). The revisions in the Continuum group in the current study were mainly due to dislocations, and the number of revisions for early lack of osteointegration or aseptic loosening was low. PJIs are a growing challenge as an increasing number of joint replacements are being performed and the life expectancy of patients is increasing (Huotari et al. 2015). Indeed, the cumulative incidence of PJI in USA and the Nordic countries is reportedly growing (Dale et al. 2012, Kurtz et al. 2012). A recent study presented promising results of TM components possibly having a protective effect against PJI (Tokarski et al. 2015). These results were not confirmed in a register study (Laaksonen et al. 2018), and were similar to our results: the risk for revision due to PJI was similar in the Continuum and in the reference group (Table 7; see Supplementary data). Continuum cups with the neutral liner used have been associated with a reduced jumping distance of the femoral head and possibly with a higher dislocation risk due to this circumstance (Pakarinen et al. 2018). In an earlier large register study based on Australian and Swedish data, the revision risk due to dislocation was not assessed separately, although the overall revision risk of TM cups was increased compared with the other uncemented cups (Laaksonen et al. 2018). We found that the risk of revision due to dislocation of the Continuum THA is increased compared with reference THAs. This difference is largely explained by the difference in the revision rate due to dislocation. In the subgroup analysis of the Continuum group we found that cups with a neutral polyethylene liner are associated with 1.7-fold dislocation revision risk compared with Continuum cups with an elevated liner. This is in line with the previous finding by Pakarinen et al. (2018). Elevated liners were first introduced by Charnley in the early 1970s to decrease the tendency for posterior dislocation by providing more coverage (Charnley 1979). The improved stability in primary THA while using an elevated rim liner was first reported in 1996 and, although these liners are widely used, there is only limited clinical evidence to support their use (Cobb et al. 1996, Sultan et al. 2002, Carter et al. 2011). Also, the benefit of routine use of elevated-rim liners in instances in which the acetabular component otherwise is positioned satisfactorily has been questioned (Krushell et al. 1991). In addition, there might be potentially harmful side effects. The elevated liners may predispose the neck of the prosthesis to impinge on the acetabular rim, forcing the head out of the cup anteriorly, but such a risk has not been confirmed in clinical studies (McCollum and Gray 1990, Sultan et al. 2002). Despite these suspicions, elevated liners have not been associated with increased revision rates during 5 years of follow-up (Cobb et al. 1997). Also, the use of lipped liners with modular uncemented acetabular components has been associated with a decreased rate of revision due to instability after primary THA, according to a register study from New Zealand

Acta Orthopaedica 2019; 90 (4): 312–317

(Insull et al. 2014). Our data support these findings: we did not observe any trend toward an elevated risk of revision due to increased wear. It is nevertheless prudent to remember that these problems may appear in a longer follow-up. Our study has some limitations. First, we were not able to assess radiographs to evaluate preoperative bone loss. It is possible that Continuum cups have been used in more demanding cases. However, Continuum being the second most used uncemented cup during our study time does suggest that it is used routinely for primary THA. Second, we were able to analyze only factors included in the register dataset. It is possible that patients might have comorbidities that could influence their dislocation risk that we are not aware of. Third, we were only able to use revision as the outcome. Some of the patients might have experienced pain, dislocations, or other implant-related problems without having a revision, for example, due to poor general health contraindicating risky revision surgery. In summary, this large nationwide study shows that the use of the Continuum cup for primary THA does not provide an advantage over traditional uncemented cups. On the contrary, the use of Continuum cups was associated with an increased revision risk compared with other uncemented cups. This enhanced risk was largely due to revisions for dislocations. If the Continuum cup is used, our results support the use of the elevated rim liner, rather than the neutral rim liner, as the primary choice. Supplementary data Tables 2 and 6–9 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1603596

MH, IL, KM: planning the study, analysis of the data, and writing the manuscript; MK: analysis of the data and revision of the manuscript; MaM and JH: calculating the statistics and revision of the manuscript; AE, A-PP, MiM, and JK: analysis of the data and revision of the manuscript. Acta thanks Henrik Bodén and Ola Rolfson for help with peer review of this study.

Baad-Hansen T, Kold S, Nielsen P T, Laursen M B, Christensen P H, Soballe K. Comparison of trabecular metal cups and titanium fiber-mesh cups in primary hip arthroplasty: a randomized RSA and bone mineral densitometry study of 50 hips. Acta Orthop 2011; 82(2): 155-60. Carter A H, Sheehan E C, Mortazavi S M J, Purtill J J, Sharkey P F, Parvizi J. Revision for recurrent instability: what are the predictors of failure? J Arthroplasty 2011; 26(6): 46-52. Charnley J. Low friction arthroplasty of the hip: theory and practice. New York: Springer Verlag; 1979. Cobb T K, Morrey B F, Ilstrup D M. The elevated-rim acetabular liner in total hip arthroplasty: relationship to postoperative dislocation. J Bone Joint Surg Am 1996; 78(1): 80-6. Cobb T K, Morrey B F, Ilstrup D M. Effect of the elevated-rim acetabular liner on loosening after total hip arthroplasty. J Bone Joint Surg Am 1997; 79(9): 1361-4.

Acta Orthopaedica 2019; 90 (4): 312–317

Dale H, Fenstad A M, Hallan G, Havelin L I, Furnes O, Overgaard S, Pedersen A B, Kärrholm J, Garellick G, Pulkkinen P, Eskelinen A, Mäkelä K, Engesæter L B. Increasing risk of prosthetic joint infection after total hip arthroplasty. Acta Orthop 2012; 83(5): 449-58. Davies J H, Laflamme GY, Delisle J, Fernandes J. Trabecular metal used for major bone loss in acetabular hip revision. J Arthroplasty 2011; 26(8): 1245-50. De Martino I, De Santis V, Sculco P K, D’Apolito R, Poultsides L A, Gasparini G. Long-term clinical and radiographic outcomes of porous tantalum monoblock acetabular component in primary hip arthroplasty: a minimum of 15-year follow-up. J Arthroplasty 2016; 31(9): 110-14. FAR. Finnish Arthroplasty Register. www.thl.fi/far; 2018. Howard J L, Kremers H M, Loechler Y A, Schleck C D, Harmsen W S, Berry D J, Cabanela M E, Hanssen A D, Pagnano M W, Trousdale R T, Lewallen D G. Comparative survival of uncemented acetabular components following primary total hip arthroplasty. J Bone Joint Surg Am 2011; 93(17): 1597-604. Huotari K, Peltola M, Jämsen E. The incidence of late prosthetic joint infections. Acta Orthop 2015; 86(3): 21-5. Insull P J, Cobbett H, Frampton C M, Munro J T. The use of a lipped acetabular liner decreases the rate of revision for instability after total hip replacement: a study using data from the New Zealand joint registry. Bone Joint J 2014; 96-B(7): 884-8 Krushell R J, Burke D W, Harris W H. Elevated-rim acetabular components: effect on range of motion and stability in total hip arthroplasty. J Arthroplasty 1991; 6 Suppl: S53-8. Kurtz S M, Lau E, Watson H, Schmier J K, Parvizi J. Economic burden of periprosthetic joint infection in the United States. J Arthroplasty 2012; 27(8 Suppl): 61–5.e1. Laaksonen I, Lorimer M, Gromov K, Rolfson O, Mäkelä K T, Graves S E, Malchau H, Mohaddes M. Does the risk of rerevision vary between porous tantalum cups and other cementless designs after revision hip arthroplasty? Clin Orthop Relat Res 2017; 475(12): 3015-22. Laaksonen I, Lorimer M, Gromov K, Eskelinen A, Rolfson O, Graves S E, Malchau H, Mohaddes M. Trabecular metal acetabular components in pri-

317

mary total hip arthroplasty: higher risk for revision compared with other uncemented cup designs in a collaborative register study including 93,709 hips. Acta Orthop 2018; 89(3): 259-64. Levine B, Della Valle C J, Jacobs J J. Applications of porous tantalum in total hip arthroplasty. J Am Acad Orthop Surg 2006; 14(12): 646-55. Macheras G A, Kateros K, Koutsostathis S D, Tsakotos G, Galanakos S, Papadakis S A. The trabecular metal monoblock acetabular component in patients with high congenital hip dislocation: a prospective study. J Bone Joint Surg Br 2010; 92(5): 624-8. McCollum D E, Gray W J. Dislocation after total hip arthroplasty: causes and prevention. Clin Orthop Relat Res 1990; (261): 159-70. Meneghini R M, Ford K S, McCollough C H, Hanssen A D, Lewallen D G. Bone remodeling around porous metal cementless acetabular components. J Arthroplasty 2010; 25(5): 741-7. Mohaddes M, Rolfson O, Kärrholm J. Short-term survival of the trabecular metal cup is similar to that of standard cups used in acetabular revision surgery: analysis of 2,460 first-time cup revisions in the Swedish Hip Arthroplasty Register. Acta Orthop 2015; 86(1): 26-31. Paavolainen P, Hämäläinen M, Mustonen H, Slätis P. Registration of arthroplasties in Finland. Acta Orthop 1991; 241: 27-30. Pakarinen O, Neuvonen P, Eskelinen A. Luksaatioiden ilmaantuvuus ja riskitekijät lonkan ensitekonivelleikkauksissa—1381 leikkauksen aineisto Tekonivelsairaala Coxasta [article in Finnish, abstract in English]. Suom Ortop ja Traumatol 2018; 41(2): 142-8. Sultan P G, Tan V, Lai M, Garino J P. Independent contribution of elevatedrim acetabular liner and femoral head size to the stability of total hip implants. J Arthroplasty 2002; 17(3): 289-92. Tokarski A T, Novack T A, Parvizi J. Is tantalum protective against infection in revision total hip arthroplasty? Bone Joint J 2015; 97-B(1): 45-9. Weeden S H, Schmidt R H. The use of tantalum porous metal implants for Paprosky 3A and 3B defects. J Arthroplasty 2007; 22(6 Suppl.): 151-5. Wegrzyn J, Kaufman K R, Hanssen A D, Lewallen D G. Performance of porous tantalum vs. titanium cup in total hip arthroplasty: randomized trial with minimum 10-year follow-up. J Arthroplasty 2015; 30(6): 1008-13.

318

Acta Orthopaedica 2019; 90 (4): 318–323

Patient-reported outcomes in hip resurfacing versus conventional total hip arthroplasty: a register-based matched cohort study of 726 patients Alexander OXBLOM 1, Håkan HEDLUND 1,2, Szilard NEMES 3,4, Harald BRISMAR 1, Li FELLÄNDER-TSAI 1, and Ola ROLFSON 3,4 1 Divison of Orthopaedics and Biotechnology, CLINTEC, Karolinska Institutet; 2 Visby Lasarett; 3 Swedish Hip Arthroplasty 4 Department of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy University of Gothenburg, Sweden

Correspondence: alexander.oxblom@ki.se Submitted 2018-06-14. Accepted 2019-03-25.

Background and purpose — The theoretical mechanical advantages of metal-on-metal hip resurfacing (MoM-HR) compared with conventional total hip arthroplasty (THA) have been questioned. Studies including measures of patientreported function, physical activity, or health-related quality of life have been sparse. We compared patient-reported outcomes in MoM-HR patients with a matched group of patients with conventional THA at 7 years post-surgery. Patients and methods — Patients and patient data were retrieved from the Swedish Hip Arthroplasty Register. The case group, consisting of 363 patients with MoM-HR, was matched 1:1 with a control group, consisting of patients with a conventional THA. Patients were sent a postal patientreported outcome measures (PROM) questionnaire including the Hip Disability and Osteoarthritis Outcome Score (HOOS), EQ-5D, and VAS pain. We used multivariable linear regression analyses to investigate the influence of prosthesis type. Results — 569 patients (78%) returned the questionnaire with complete responses (299 MoM-HRs and 270 conventional THAs). MoM-HR was associated with better scores in HOOS function of daily living (4 percentage units) and HOOS function in sport and recreation (8 percentage units) subscales. Type of prosthesis did not influence HOOS quality of life, HOOS pain, HOOS symptoms, EQ-5D index, hip pain, or satisfaction as measured with visual analog scales. Interpretation — At mean 7 years post-surgery, patients with hip resurfacing had somewhat better self-reported hip function than patients with conventional THA. The largest difference between groups was seen in the presumed most demanding subscale, i.e., function in sport and recreation.

Hip arthroplasty in young and active patients is an orthopedic challenge. In 2011, the Finnish Arthroplasty Register (Mäkelä et al. 2011) reported a 15-year prosthesis survival rate of about 70% in patients younger than 55 years operated with conventional total hip arthroplasty (THA) compared with about 90% in patients older than 60 years in the combined Nordic Arthroplasty Registers (Havelin et al. 2009). Young patients have higher expectations following THA (Scott et al. 2012) and are more active, a patient-factor highly related to polyethylene wear (Schmalzried et al. 2000). They are also more prone to participate in high-impact sports following THA (Williams et al. 2012), which has been correlated with both increased wear (Ollivier et al. 2012) and higher revision rates (Flugsrud et al. 2007). Alternative surface bearings and prosthesis designs have therefore been developed to meet the demands of younger patients. Metal-on-metal hip resurfacing (MoM-HR) gained popularity in the mid-1990s due to advances in metallurgy and tribology, allowing manufacturing of thin acetabular cups accepting large-diameter components (Grigoris et al. 2006). It was believed that the wear-associated disadvantages seen with metal-on-polyethylene thereby could be solved. The method was expected to provide a sustainable arthroplasty for young and active patients with hip osteoarthritis (Amstutz and Le Duff 2012). Besides a bone-preserving surgical technique, MoM-HR was also claimed to restore hip mechanics with a better range of motion (Vail et al. 2006). However, there was a major setback when some MoM-HR implants and THAs with MoM articulations were reported to have unacceptably high failure rates (De Steiger et al. 2011, Smith et al. 2012). As a result, there was a dramatic decline in numbers of MoM-HR implanted worldwide and, in many countries, sur-

Acta Orthopaedica 2019; 90 (4): 318–323

geons promptly stopped using the technique, due to perceived risks and the uncertainty regarding the long-term results of the implants (Cohen 2011). There are, though, some long-term follow-ups of certain brands of MoM-HR implants with acceptable implant survival in a selected group of patients (Matharu et al. 2013). It is evident that cautious patient selection is crucial, quite apart from implant design and surgical technique (Daniel et al. 2014). Reports on benefits of MoM-HR in terms of patient-reported function, physical activity, and health-related quality of life are sparse (Jiang et al. 2011). We compared patient-reported outcomes in patients operated with MoM-HR with a matched group of patients operated with conventional THA at mean 7 years post-surgery.

Patients and methods Patient selection This is an arthroplasty register-based matched cohort study. Patient data were retrieved from the Swedish Hip Arthroplasty Register. The case group, consisting of a consecutive group of all patients operated on with MoM-HR (all Birmingham Hip Resurfacing System, Smith & Nephew, Andover, Massachusetts, USA) at a single institution (Karolinska Huddinge) between the years 2002 and 2013, was matched 1:1 with a control group, consisting of patients with a conventional THA selected from the Register. In the case of bilateral MoM-HR (n = 105) or bilateral THA (n = 102) during the study period, we included data regarding the first operation. Patients deceased by December 2015 (n = 6) were excluded. The groups were matched by baseline characteristics: age, sex, surgical approach, year of surgery, and preoperative EQ-5D score when available. Outcome measures 726 patients (363 MoM-HRs, 363 conventional THAs) were selected for the study (Table 1). In December 2015, patients were invited to participate by mail and asked to complete a patient-reported outcome measures (PROM) questionnaire including the Hip Disability and Osteoarthritis Outcome Score (HOOS) (Nilsdotter et al. 2003), the EQ-5D (EuroQol Group 1990), hip pain measured with a visual analogue scale (VAS), and a VAS addressing satisfaction with the outcome of surgery. In addition to the postal questionnaire we used information from the Swedish Hip Arthroplasty Register covering surgical data, demography, data on subsequent reoperations and, when available, pre- and postoperative PROMs data including hip pain and the EQ-5D (Garellick et al. 2015). Statistics Subject-matter knowledge was used to identify and measure adjustment variables. The goal was to identify a sufficient set

319

Table 1. Patient demographics Characteristics

Case Control group group p-value

Number of patients 363 ( 363 ( Women, n (%) 90 (25) 86 (24) Age at primary operation, mean (SD) 52 (8.8) 51 (8.7) Year of surgery, mean (SD) 2008 (2.9) 2008 (2.9) Follow-up time, mean (SD) 7.3 (2.9) 7.3 (3.0) Distribution of diagnoses, n (%) Primary osteoarthritis 315 (87) 325 (90) Childhood hip disease 41 (11) 31 (8.5) Other hip joint disorders 7 (2.0) 7 (2.0) PROMs preoperatively, n 206 ( 363 ( VAS hip pain, mean (SD) 74 (16.4) 69 (18.4) EQ-5D index, mean (SD) 0.52 (0.29) 0.43 (0.32) Patients reoperated, n (%) 13 (3.6) 16 (4.4)

0.8 0.7 0.9 0.9 0.7

0.002 0.001 0.6

SD = standard deviation; PROMs = patient-reported outcome measures; VAS = visual analog scale; EQ-5D = EuroQol 5 dimensions.

for confounding adjustment for prosthesis type. This set was defined as a set of non-descendant variables for prosthesis type that block all backdoor paths. Confounder identification was based on Rubin’s 3 conditions (Robins 1999, Greenland et al. 1999). By matching we constructed a subset of the population in which the background has the same distribution in both the MoM-HR and the conventional THA groups. In observational studies, there is no guarantee that the treatment groups are conditionally exchangeable given the exposure only. Matching generally exploits the conditional exchangeability; however, matching cases and controls does not achieve unconditional exchangeability. Ignoring the matching variables in a cohort study can leave bias if there are additional confounders, even with adjustment for the additional confounders (Sjölander and Greenland 2013). Based on these 2 facts the final analysis included the variables used for matching. We identified age, sex, preoperative EQ-5D index, and time from surgery. Neither variable is on the path between the exposure and outcome and can block important backdoor paths (Figure 1, see Supplementary data). Using the Directed Acyclic Graph from Figure 1 and d-separation to infer associational statements (Textor et al. 2011) we could conclude that the minimal sufficient adjustment sets for estimating the direct effect and total effect is age, sex, and preoperative EQ-5D index. Time for surgery was included to reduce bias (Sjölander and Greenland 2013). We used multivariable linear regression analyses to investigate the influence of prosthesis type (MoM-HR versus conventional THA) adjusting for age, sex, preoperative EQ-5D index, and time from surgery. R (R Core Team 2017) and IBM SPSS Statistics version 25 (IBM Corp, Armonk, NY, USA) were used for statistical analyses. Missing covariate data were imputed using full-conditional specification (FCS) multiple imputation with the inclusion of the outcomes and matching variables (Seaman and Keogh 2015). The imputed data

320

Acta Orthopaedica 2019; 90 (4): 318–323

Table 3. Postoperative functional outcomes. Values are mean (SD) Variables HOOS index (%) Symptoms Pain ADL Sport/Rec QoL EQ-5D index VAS hip pain VAS satisfaction

Case Control group group p-value 85 (17) 90 (15) 90 (15) 77 (24) 77 (21) 0.90 (0.17) 11 (16) 13 (20)

83 (19) 87 (18) 84 (19) 68 (29) 74 (22) 0.87 (0.21) 12 (18) 12 (21)

0.09 0.01 < 0.001 < 0.001 0.07 0.2 0.4 0.7

SD = standard deviation; HOOS = Hip Disability and Osteoarthritis Outcome Score; ADL = Activity in Daily Living; Sport/Rec = Sport and Recreation; QoL = Quality of Life; EQ-5D = EuroQol 5 dimensions; VAS = Visual analog scale.

were used as input for regression analyses and estimates from each imputed dataset were combined into 1 overall estimate and associated variance, incorporating both the within and between imputation variability using Rubin’s rules (Marshall et al. 2009). Regression estimates (coefficients) were reported with 95% confidence intervals (CI). Observational studies are by nature subjected to unmeasured confounding. We postulate that the possible unblocked backdoor paths are weak. Confounding bias requires a strong confounder treatment and a strong confounder outcome association. Generally, baseline variables explain a low amount of variance of postoperative PROMs (Bengtsson et al. 2017, Nemes et al. 2018) and expectedly the residual confounding bias is low. Ethics, funding, and potential conflicts of interest The study was approved by the Regional Ethical Review Board in Gothenburg (Dnr 407-14). This research did not receive any specific grants from commercial funding agencies or bodies. The study was supported by public funding from the Swedish Hip Arthroplasty Register and research funds from Stockholm County Council. No competing interest declared.

Results 569 patients (78%) returned the questionnaire with complete responses. Mean follow-up time (F-U) was 7 years (IQR 2.2–13 years). The proportion of patients who had undergone any reoperation was similar between groups (Table 1). The preoperative demographics of the patients who did not answer the questionnaire did not demonstrate statistically significant difference from those who answered (Table 2, see Supplementary data). The case group had better unadjusted outcomes in all subscales of HOOS whereas EQ-5D index, VAS pain, and VAS satisfaction were equal between the groups (Table 3).

Figure 2. Graphic representation of postoperative PROMs after multivariable linear regression analyses. Bars represent 95% CI of the adjusted estimates (regression coefficients). For abbreviations, see Table 3.

Both the crude and adjusted estimates (Figure 2) showed that MoM-HR was associated with better scores in HOOS ADL (4.3, CI 1.8–6.9), and Sport/Rec (7.8, CI 3.8–12). We found no statistically significant association between type of prosthesis and remaining HOOS subscales, EQ-5D index, hip pain VAS, or satisfaction VAS.

Discussion Patients who underwent hip resurfacing reported better postoperative functional outcomes (HOOS subscales ADL and Sport/Rec) at mean 7 years post-surgery compared with a group of matched patients with conventional hip arthroplasty. We found no statistically significant differences in EQ-5D index, hip pain, or satisfaction. The largest difference between the groups was seen in the presumed most demanding subscale, i.e., function in sport and recreation. Our observation is in accordance with the study of Haddad et al. (2015), showing that hip resurfacing yields better results regarding return to sports compared with conventional THA. The results also conform to a retrospective study of 215 resurfacing arthroplasties (mean F-U 2 years) (Girard et al. 2013), which showed that 41 of the 50 patients who participated in high-impact activity before the operation and onset of pain, returned to highimpact activity whilst 48 patients returned to any kind of physical activity. Although the last-mentioned study did not include a control group, other studies have demonstrated that only up to 40% of high-activity patients return to sport activity after conventional THA (Del Piccolo et al. 2016, Schmidutz et al. 2012). When functional outcome scores were compared prospectively in 89 consecutively operated hips it was found that the resurfacing patients had greater improvement in Harris Hip scores, in UCLA activity score, and had a higher postoperative UCLA activity score than those operated with conventional THA (Fowble et al. 2009). On the other hand, the groups were not matched regarding overall health or preoperative functional outcome scores.

Acta Orthopaedica 2019; 90 (4): 318–323

Tan et al. (2015) found that functional outcome scores and activity level from short to long-term follow-up were timedependent. Among 100 patients with unilateral MoM-HR, they reported UCLA and SF-12 scores preoperatively, in the short term (mean F-U 2 years), and at a minimum of 10 years after the operation (mean F-U 12 years). They found no decrease in UCLA pain and walking scores between shortterm and long-term follow-up, but a decrease in function and activity scores. With this in mind, when evaluating functional outcomes after hip arthroplasty, the results do not seem to be dependent only on functional outcome validation instruments, age, and sex but also on the time of the follow-up. There are only a few previous studies comparing functional outcome scores between hip resurfacing and THA patients (Pollard et al. 2006, Mont et al. 2009, Costa et al. 2012). A retrospectively matched (sex, age, BMI, and activity level) study with a 7-year follow-up showed no difference in Oxford Hip Score but a higher level of activity as measured by UCLA score, and higher percentage (7% MoM-HR vs. 33% conventional THA) of patients participating in sports in the MoM-HR group (Pollard et al. 2006). Despite matching and medium– long follow-up, that study consisted of a rather small group of patients (53 MoM-HRs, 51 conventional THAs) making it difficult to draw certain conclusions. In another matched casecontrol study comprising 100 patients (50 MoM-HRs, 50 conventional THAs), the authors found no differences in mean Harris Hip Score (90 HR vs. 91 THA) or in patient satisfaction scores (9.2 HR vs. 8.8 THA) in short-term follow-up (Mont et al. 2009). As Harris Hip Score is limited to functional criteria, such a measure does not give an appropriate description of the patients’ functional outcome. In an assessor-blinded randomized controlled study (Costa et al. 2012) with 1:1 treatment allocation, hip function was similar between MoM-HR and THA at 12 months’ follow-up as measured with Harris Hip Score (88 MoM-HR vs. 82 THA) and Oxford Hip Score (40 MoM-HR vs. 38 THA). Furthermore, disability rating and activity level were similar in the first year after surgery. In that study, the long-term effects of HR were not studied. In the meantime, a 5-year F-U report is available that also shows similar hip function or health-related quality of life following a total hip arthroplasty vs. hip resurfacing (Costa et al. 2018). When analyzing the “Forgotten Joint” Score-12 (78 MoM-HR vs. 76 THA) between MoM-HR and conventional THA, it was concluded that the choice of implant should not be based solely on any expectation that either yields superior clinical outcomes compared with the other at short-term follow-up (Ortiz-Declet et al. 2017). Our study has some limitations. The collecting of PROMs did not reach nationwide coverage until 2008, which explains why preoperative data were not available for all of the patients (n = 157 had missing data preoperatively). However, missing preoperative EQ-5D data were successfully imputed and the EQ-5D scores were subsequently used for case-mix adjustment based on preoperative health status. Another limitation

321

pertains to the lack of prospective HOOS data. Although groups were matched based on demography and baseline EQ-5D index, level of functioning in ADL and sports and recreation may have differed preoperatively. The occurrence of reoperations could be a potential source of bias albeit repeat surgeries were evenly distributed between the groups. Whilst conventional THA is performed in most orthopedic units in Sweden, hip resurfacing was only performed in a few specialist centers during the study period. Therefore, all patients operated with HR either actively searched for institutions performing resurfacing prosthesis or were referred from other orthopedic units. Patients operated with conventional THAs likely did not actively request a certain implant, suggesting a biased selection that cannot be adjusted for. Moreover, almost all HR surgeries were performed by 2 experienced surgeons following well-established principles of surgical innovation in contrast to the control group, which was selected from the registry not considering surgeon experience. It must be constantly emphasized that introduction of new devices should follow a systematic approach even if the theoretical basis or preclinical results are excellent. Recently, Reito et al. (2017) described the anti-stepwise introduction of metal-on-metal hip replacements. The strengths of our study include the careful 1:1 matching of the groups for the various demographic factors, surgical approach, time of surgery, and preoperative EQ-5D scores, which reduced many confounding factors. Our study also comprised a fairly large number of patients in the groups and with a satisfactory response rate. To our knowledge no study comparing functional outcome scores between MoM-HR and conventional THA has been undertaken with such a large number of patients followed for a comparable period of time. Although the type of hip prosthesis did not influence the level of satisfaction, postoperative pain relief, or quality of life, MoM-HR patients had better postoperative HOOS scores in the function of daily living and function in sports and recreation domains. Translating the adjusted regression estimates of these 2 HOOS subscales into effect sizes, the influence of MoM-HR was moderate (0.25 and 0.30, respectively). Furthermore, there was no statistically significant difference in reoperation rates using a Birmingham Hip Replacement (BHR) compared with a conventional implant in these 2 ageand sex-matched patient groups. As MoM-HR was developed to address the special demands of a younger and more active population, our results support the rationale for using the technique in this group of patients. Choice of hip arthroplasty for young and active patients with high expectations is still challenging, mostly due to higher risks of wear, dislocation, and need of revision surgery. In summary, by comparing MoM-HR with conventional THA in a matched study design (mean 7 years F-U) of a selected group of patients we have shown MoM-HR to yield better functional outcome scores in 2/5 HOOS subscales; all other outcome measures were similar. When a BHR implant is considered,

322

patients should be informed of the risk of developing of adverse reactions and uncertain long-term results. We highly recommend subsequent close follow-up for this matter. Supplementary data Figure 1 and Table 2 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1604343

OR, HH, and LFT conceived and designed the study. OR and HH obtained ethical approval. OR and SN collected data. SN performed statistical analysis. AO drafted the manuscript. All authors interpreted the results and reviewed, edited, and approved the final version of the manuscript. Acta thanks Nina Mathijssen and Marc Nijhof for help with peer review of this study.

Amstutz H C, Le Duff M J. Hip resurfacing: a 40-year perspective. HSS J 2012; 8(3): 275-82. Bengtsson A, Donahue GS, Nemes S, Garellick G, Rolfson O. Consistency in patient-reported outcomes after total hip replacement: a 6-year registry follow-up of 15,755 patients. Acta Orthop 2017; 88(5): 484-9. Cohen D. Out of joint: the story of the ASR. BMJ 2011; 342: d2905. Costa M L, Achten J, Parsons N R, Edlin R P, Foguet P, Prakash U, Griffin D R. Total hip arthroplasty versus resurfacing arthroplasty in the treatment of patients with arthritis of the hip joint: single centre, parallel group, assessor blinded, randomised controlled trial. BMJ 2012; 344: e2147. Costa M L, Achten J, Foguet P, Parsons N R. Comparison of hip function and quality of life of total hip arthroplasty and resurfacing arthroplasty in the treatment of young patients with arthritis of the hip joint at 5 years. BMJ 2018; 8(3): e018849. Daniel J, Pradhan C, Ziaee H, Pynsent P B, McMinn D J W. Results of Birmingham hip resurfacing at 12–15 years: a single-surgeon series. Bone Joint J 2014; 96-B: 1298-306. Del Piccolo N, Carubbi C, Mazzotta A, Sabbioni G, Filanti M, Stagni C, Dallari D. Return to sports activity with short stems or standard stems in total hip arthroplasty in patients less than 50 years old. Hip Int 2016; 26(Suppl. 1): 48-51. De Steiger R N, Hang J R, Miller L N, Graves S E, Davidson D C. Five-year results of the ASR X-large Acetabular System and the ASR Hip Resurfacing System: an analysis from the Australian Orthopaedic Association National Joint Replacement Registry. J Bone Joint Surg Am 2011; 93(24): 2287-93. EuroQol Group. EuroQol: a new facility for the measurement of healthrelated quality of life. Health Policy 1990; 16(3): 199-208. Flugsrud G B, Nordsletten L, Espehaug B, Havelin L I, Meyer H E. The effect of middle-age body weight and physical activity on the risk of early revision hip arthroplasty: a cohort study of 1535 individuals. Acta Orthop 2007; 78(1): 99-107. Fowble V A, dela Rosa M A, Schmalzried T P. A comparison of total hip resurfacing and total hip arthroplasty: patients and outcomes. Bull NYU Hosp Jt Dis 2009; 67(2): 108-12. Garellick G, Kärrholm J, Lindahl H, Malchau H, Mohaddes M, Rogmark C, Rolfson O. Swedish Hip Registry: Annual Report 2015. Girard J, Miletic B, Deny A, Migaud H, Fouilleron N. Can patients return to high-impact physical activities after hip resurfacing? A prospective study. Int Orthop 2013; 37(6): 1019-24.

Acta Orthopaedica 2019; 90 (4): 318–323

Greenland S, Robins J M, Pearl J. Confounding and collapsibility in causal inference. Stat Sci 1999; 14:29-46. Grigoris P, Roberts P, Panousis K, Jin Z. Hip resurfacing arthroplasty: the evolution of contemporary designs. Proc Inst Mech Eng H 2006; 220(2): 95-105. Haddad F S, Konan S, Tahmassebi J. A prospective comparative study of cementless total hip arthroplasty and hip resurfacing in patients under the age of 55 years: a ten-year follow-up. Bone Joint J 2015; 97-B(5): 617-22. Havelin L I, Fenstad A M, Salomonsson R, Mehnert F, Furnes O, Overgaard S, Pedersen A B, Herberts P, Kärrholm J, Garellick G. The Nordic Arthroplasty Register Association: a unique collaboration between 3 national hip arthroplasty registries with 280,201 THRs. Acta Orthop 2009; 80(4): 393-01. Jiang Y, Zhang K, Die J, Shi Z, Zhao H, Wang K. A systematic review of modern metal-on-metal total hip resurfacing vs. standard total hip arthroplasty in active young patients. J Arthroplasty 2011; 26(3): 419-26. Marshall A, Altman D G, Holder R L, Royston P. Combining estimates of interest in prognostic modelling studies after multiple imputation: current practice and guidelines. BMC Med Res Methodol 2009; 9: 57. Matharu G S, McBryde C W, Pynsent W B, Pynsent P B, Treacy R B. The outcome of the Birmingham Hip Resurfacing in patients aged < 50 years up to 14 years post-operatively. Bone Joint J 2013; 95-B(9): 1172-7. Mont M A, Marker D R, Smith J M, Ulrich S D, McGrath M S. Resurfacing is comparable to total hip arthroplasty at short-term follow-up. Clin Orthop Rel Res 2009; 467: 66-71. Mäkelä K T, Eskelinen A, Pulkkinen P, Paavolainen P, Remes V. Results of 3,668 primary total hip replacements for primary osteoarthritis in patients under the age of 55 years. Acta Orthop 2011; 82: 521-29. Nemes S, Rolfson O, Garellick G. Development and validation of a shared decision-making instrument for health-related quality of life one year after total hip replacement based on quality registries data. J Eval Clin Pract 2018; 24: 13-21 Nilsdotter A K, Lohmander L S, Klassbo M, Roos E M. Hip Disability and Osteoarthritis Outcome Score (HOOS): validity and responsiveness in total hip replacement. BMC Musculoskelet Disord 2003; 4: 10. Ollivier M, Frey S, Parratte S, Flecher X, Argenso J. Does impact sport activity influence total hip arthroplasty durability? Clin Orthop Relat Res 2012; 470(11): 3060-6. Ortiz-Declet V R, Iacobelli D A, Yuen L C, Perets I, Chen A W, Domb B G. Birmingham Hip Resurfacing vs. Total Hip Arthroplasty: a matched-pair comparison of clinical outcomes. J Arthroplasty 2017; 32(12): 3647-51. Pollard T C B, Baker R P, Eastaugh-Waring S J, Bannister G C. Treatment of the young active patient with osteoarthritis of the hip: a 5- to 7-year comparison of hybrid total hip arthroplasty and metal-on-metal resurfacing. J Bone Joint Surg Br 2006; 88-B: 592-600. R Core Team (2017). R: a language and environment for statistical computing. R Vienna, Austria: Foundation for Statistical Computing. https:// www.R-project.org/ Reito A, Lehtovirta L, Lainiala O, Mäkelä K, Eskelinen A. Lack of evidence: the anti-stepwise introduction of metal-on-metal hip replacements. Acta Orthop 2017; 88(5): 478-83. Robins J M. Marginal structural models versus structural nested models as tools for causal inference. In: Halloran E, Berry D. Statistical models in epidemiology: the environment and clinical trials. New York, SpringerVerlag; 1999. pp. 95-134. Schmalzried T P, Shepherd E F, Dorey F J, Jackson W O, dela Rosa M, Fa’vae F, McKellop H A, McClung C D, Martell J, Moreland J R, Amstutz H C. The John Charnley Award: Wear is a function of use, not time. Clin Orthop Relat Res 2000; (381): 36-46. Schmidutz F, Grote S, Pietschmann M, Weber P, Mazoochian F, Fottner A, Jansson V. Sports activity after short-stem hip arthroplasty. Am J Sports Med 2012; 40(2): 425-32.

Acta Orthopaedica 2019; 90 (4): 318â&#x20AC;&#x201C;323

Scott C E, Bugler K E, Clement N D, MacDonald D, Howie C R, Biant L C. Patient expectations of arthroplasty of the hip and knee. J Bone Joint Surg Br 2012; 94(7): 974-81. Seaman S R, Keogh R H. Handling missing data in matched case-control studies using multiple imputation. Biometrics 2015; 71(4): 1150-9. SjĂślander A, Greenland S. Ignoring the matching variables in cohort studies: when is it valid and why? Stat Med 2013; 32: 4696-708. Smith A J, Dieppe P, Vernon K, Porter M, Blom A W; National Joint Registry of England and Wales. Failure rates of stemmed metal-on-metal hip replacements: analysis of data from the National Joint Registry of England and Wales. Lancet 2012; 380(9855): 1759-66.

323

Tan T L, Le Duff M J, Takamura K M, Amstutz H C. Do clinical and quality of life scores change over time after hip resurfacing? Hip Int 2015; 25(2): 146-51. Textor J, Hardt J, Knuppel S. Dagitty: a graphical tool for analyzing causal diagrams. Epidemiology 2011; 22(5):745 Vail T P, Mina C A, Yergler J D, Pietrobon R. Metal-on-metal hip resurfacing compares favorably with THA at 2 years followup. Clin Orthop Relat Res 2006; 453: 123-31. Williams D H, Greidanus N V, Masri B A, Duncan C P, Garbuz D S. Predictors of participation in sports after hip and knee arthroplasty. Clin Orthop Relat Res 2012; 470(2): 555-61.

324

Acta Orthopaedica 2019; 90 (4): 324–330

ASA class is associated with early revision and reoperation after total hip arthroplasty: an analysis of the Geneva and Swedish Hip Arthroplasty Registries Rory J FERGUSON 1, Alan J SILMAN 1, Christophe COMBESCURE 2, Erik BULOW 3, Daniel ODIN 3, Didier HANNOUCHE 4, Siôn GLYN-JONES 1, Ola ROLFSON 3, and Anne LÜBBEKE 1,4 1 Nuffield

Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, UK; 2 Division of Clinical Epidemiology, Geneva University Hospitals, Switzerland; 3 The Swedish Hip Arthroplasty Register and the Department of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Sweden; 4 Division of Orthopaedics and Trauma Surgery, Geneva University Hospitals, Switzerland Correspondence: rory.ferguson@ndorms.ox.ac.uk Submitted 2018-11-28. Accepted 2019-03-31.

Background and purpose — Data from several joint replacement registries suggest that the rate of early revision surgery after primary total hip arthroplasty (THA) is increasing. The ASA class, now widely recorded in arthroplasty registries, may predict early revision. We investigated the influence of ASA class on the risk of revision and other reoperation within 3 months and within 5 years of primary THA. Patients and methods — We used data from the Geneva and Swedish Hip Arthroplasty Registries, on primary elective THAs performed in 1996–2016 and 2008–2016, respectively. 5,319 and 122,241 THAs were included, respectively. Outcomes were all-cause revision and other reoperations evaluated using Kaplan–Meier survival and Cox regression analyses. Results — Within 3 months after surgery, higher ASA class was associated with greater risk of revision and other reoperation. 3-month cumulative incidences of revision by ASA class I, II, and III–IV respectively, were 0.6%, 0.7%, and 2.3% in Geneva and 0.5%, 0.8%, and 1.6% in Sweden. 3-month cumulative incidences of other reoperation were 0.4%, 0.7%, and 0.9% in Geneva and 0.2%, 0.4%, and 0.7% in Sweden. Adjusted hazard ratios (ASA III–IV vs. I) for revision within 3 months were 2.7 (95% CI 1.2–5.9) in Geneva and 3.3 (CI 2.6–4.0) in Sweden. Interpretation — Assessment of ASA class of patients prior to THA will facilitate risk stratification. Targeted riskreduction strategies may be appropriate during the very early postoperative period for patients identified as at higher risk. Systematically recording ASA class in arthroplasty registries will permit risk adjustment and facilitate comparison of revision rates internationally.

Data from several joint replacement registries suggest that the rate of early revision surgery after primary total hip arthroplasty (THA), widely defined as within 5 years of primary THA, is increasing (Thien et al. 2014, Cnudde et al. 2017). Recent data have shown that a high proportion of early revision surgeries are performed within 3 months (Swiss National Joint Registry 2018). Patients requiring such early revision surgery may share particular characteristics that put them at risk, such as preoperative health status. Evidence from the New Zealand Joint Registry suggests that poor preoperative health status, assessed by ASA class, places patients at increased risk of revision within 2 years of surgery (Hooper et al. 2012). However, there are few data on the influence of poor preoperative health status on very early revision, specifically within 3 months of primary THA. Understanding its influence on the rate of very early revision surgery would be beneficial for 3 reasons: 1st, enhanced preoperative risk stratification would support surgeons and patients; 2nd, riskreduction strategies could be identified and implemented for patients most at risk within 3 months postoperatively; and 3rd, risk adjustment would facilitate comparisons of outcomes between datasets. Other reoperations after primary THA include, but are not limited to, debridement of infection, osteosynthesis of periprosthetic fracture, and drainage of hematoma. Data on other reoperations are not widely collected by arthroplasty registries. Evidence on incidence and causative factors is limited (Ferguson et al. 2018). Many methods exist to measure preoperative health status. The ASA classification system is now the most widely collected system for measuring physical health status by arthroplasty registries worldwide (Lübbeke et al. 2018).

Acta Orthopaedica 2019; 90 (4): 324–330

We investigated the influence of ASA class on the risk of revision and other reoperation within 3 months and within 5 years of primary THA. In cases of revision and other reoperation, we investigated the indication for surgery.

Patients and methods We conducted a retrospective analysis of data from 2 arthroplasty registries. We performed a preliminary study in a hospital registry (Geneva Arthroplasty Registry, GAR), and compared the results with those from a national registry (Swedish Hip Arthroplasty Register, SHAR). GAR collects data on all THAs performed at Geneva University Hospitals, the only public hospital of the canton of Geneva, Switzerland serving a population of 500,000 inhabitants (Geneva Joint Arthroplasty Registry 2017). Completeness of recording THAs is > 99%. SHAR collects data on all THAs performed in Sweden, covering 80 clinics (Swedish Hip Arthroplasty Register 2016). Completeness of recording THAs in the registry is 98.3%. The completeness of capture of revision surgery following primary THAs recorded in the GAR was 100% in 2013–2016, based on revisions performed within Switzerland. It was not possible to directly calculate the completeness of capture of revision surgery prior to 2013 in GAR; however, loss to follow-up in GAR after 5 years was 6% during 1996–2012, hence we estimate the completeness of capture of revision surgery was ≥ 94%. In SHAR the completeness of capture of revision surgery was 93%, based on revisions performed within Sweden. Eligible procedures were elective primary THAs performed during the period that registries collected data on ASA class. This period was March 1996 through December 2016 for GAR and January 2008 through December 2016 for SHAR. THAs in patients with missing data on ASA class were excluded. Bilateral cases were included. 2 groups of cases were excluded: 1st, we excluded metal-on-metal THAs because they have a substantially higher revision rate than other bearings (Swedish Hip Arthroplasty Register 2016, Geneva Joint Arthroplasty Registry 2017, National Joint Registry 2017). Moreover, patients with lower ASA class received metal-onmetal prostheses more than other patients, meaning inclusion of such cases could have biased our results. 2nd, we excluded THAs for which the indication was trauma or malignancy. The ASA classification system classifies patients into 6 categories (classes I [normal health]–VI [brain death]). ASA classes V and VI are not appropriate to patients undergoing elective THA, leaving a range of ASA classes I–IV. We evaluated 2 outcomes: incidence of revision and of any other reoperation. Revision surgery was defined as any surgery that involved the addition, removal, or replacement of 1 or more components of the prosthetic hip. Other reoperation was defined as any surgery to the prosthetic hip that did not involve the addition, removal, or replacement of any

325

components of the prosthetic hip. Closed reduction of dislocation was not included as a reoperation. Indications for surgery were also extracted. Covariates were age at surgery, sex, BMI, and diagnosis (primary or secondary osteoarthritis [OA]). The GAR records revision, other reoperation, and mortality data continuously and actively follows patients up at 1, 5, 10, 15, and 20 years. The SHAR records revision and other reoperation data continuously. Mortality data are obtained from the Swedish Board of Health and Welfare. The end of followup was December 2016. Statistics Analyses were conducted independently for each registry. Baseline characteristics were described using frequencies, proportions, means, and standard deviations (SDs). The proportion of THAs in patients of ASA class IV (0.6% in GAR and 0.4% in SHAR) was too small for meaningful analysis on its own. Thus, ASA was categorized into 3 groups: class I, II, and III–IV. The cumulative mortality was assessed with Kaplan–Meier survival estimates. Cumulative incidence of revision and other reoperation by ASA class over 5-year follow-up after index THA was assessed using non-parametric models with death as competing event. As a sensitivity analysis, the survival analyses were re-run including only the 1st THA procedure in each patient. Cause-specific Cox proportional hazard models (presented as cause-specific hazard ratios [HRs] with 95% confidence intervals [CIs]) were used to assess the association between ASA class and risk of revision and other reoperation. Death was considered as a competing event. ASA class I was defined as the referent category. Details on the assumption of the models are presented in the Appendix (see Supplementary data). With the proposed models, the HRs for the associations were potentially different within 3 months following primary THA and after 3 months. Multivariable models with a pre-specified set of adjustment factors (age, sex, BMI, diagnosis) were conducted. Complete case analysis was used for adjusted models. Data were analyzed using SPSS Version 23 software (IBM Corp, Armonk, NY, USA) and R (R Foundation for Statistical Computing, Vienna, Austria) with alpha of 0.05 as the statistical threshold for significance (all tests were 2-sided). Ethics, funding, and potential conflicts of interests The registry data collection was approved by the Geneva University Hospital Institutional Review Board and the Gothenburg Regional Ethical Review Board. No funding was received for the study. The authors declare no potential conflicts of interest.

Results In GAR, 5,319 procedures in 4,501 patients were eligible for inclusion. In SHAR, 122,241 procedures in 106,522 patients were eligible for inclusion (Table 1). In both cohorts the pro-

326

Acta Orthopaedica 2019; 90 (4): 324–330

Table 1. Preoperative patient characteristics by ASA class GAR SHAR ASA I ASA II ASA III–IV Total ASA I ASA II ASA III–IV n = 481 n = 3,496 n = 1,342 n = 5,319 n = 29,280 n = 72,857 n = 20,104

Total n = 122,241

Women, n (%) 259 (54) 2,106 (60) 767 (57) 3,132 (59) 16,112 (55) 41,460 (59) 9,885 (53) 67,149 (57) Age (mean, SD) 60.1 (12.3) 68.8 (11.7) 76.1 (10.1) 69.9 (12.2) 62.5 (10.7) 69.2 (9.7) 72.9 (9.8) 68.2 (10.5) Age categories, n (%) < 55 150 (31) 393 (11) 42 (3.1) 585 (11) 6,325 (22) 5,152 (7.1) 863 (4.3) 12,340 (10) 55–64 127 (26) 667 (19) 109 (8.1) 903 (17) 9,679 (33) 15,756 (22) 2,701 (13) 28,136 (23) 65–74 152 (32) 1,226 (35) 357 (27) 1,735 (33) 9,766 (33) 29,660 (41) 6,999 (35) 46,425 (38) 75–84 49 (10) 1,024 (29) 550 (41) 1,623 (31) 3,225 (11) 19,423 (27) 7,648 (38) 30,296 (25) ≥ 85 3 (0.6) 186 (5.3) 284 (21) 473 (8.9) 285 (1.0) 2,866 (3.9) 1,893 (9.4) 5,044 (4.1) BMI (mean, SD) 24.8 (3.3) 26.8 (4.7) 27.3 (5.6) 26.8 (4.9) 26.1 (3.8) 27.4 (4.4) 28.7 (5.6) 27.3 (4.6) BMI categories, n (%) < 18.5 13 (2.7) 63 (1.8) 43 (3.3) 119 (2.3) 195 (0.6) 531 (0.7) 240 (1.2) 966 (0.8) 18.5–24 245 (51) 1,280 (37) 437 (33) 1,962 (37) 11,405 (38) 20,933 (29) 5,008 (26) 37,346 (31) 25–29 193 (40) 1,303 (38) 441 (33) 1,937 (37) 12,868 (43) 31,374 (44) 6,955 (36) 51,197 (43) 30–34 27 (5.6) 642 (19) 288 (22) 957 (18) 3476 (12) 14,512 (20) 4,520 (23) 22,508 (19) 35–39 2 (0.4) 156 (4.5) 85 (6.4) 243 (4.6) 416 (1.3) 3,234 (4.5) 2,115 (11) 5,765 (4.8) ≥ 40 0 (0.0) 26 (0.7) 27 (2.0) 53 (1.0) 66 (0.2) 529 (0.7) 629 (3.2) 1,224 (1.0) Missing data 1 26 21 48 854 1,744 637 3,235 Diagnosis, n (%) Primary OA 375 (78) 2,817 (81) 1,007 (75) 4,199 (79) 26,644 (91) 67,221 (92) 17,666 (88) 111,531 (92) Secondary OA 106 (22) 679 (19) 335 (25) 1,120 (21) 2,636 (9.0) 5,636 (7.7) 2,438 (12) 10,710 (8.8)

Table 2. Incidence of revision and other reoperation within 5 years of primary THA by ASA score

Total number (%)

Revision Cumulative incidence (CI) 3 months 5 years

Total number (%)

Other reoperation Cumulative incidence (CI) 3 months 5 years

Geneva All patients 126 (2.4) 1.1 (0.8–1.4) 2.6 (2.1–3.0) 95 (1.8) 0.7 (0.5–0.9) ASA I 10 (2.1) 0.6 (0.0–1.3) 2.3 (0.9–3.8) 9 (1.9) 0.4 (0.0–1.0) ASA II 73 (2.1) 0.7 (0.4–1.0) 2.3 (1.8–2.8) 58 (1.7) 0.7 (0.4–0.9) ASA III–IV 43 (3.2) 2.3 (1.5–3.1) 3.3 (2.3–4.3) 28 (2.1) 0.9 (0.0 –1.4) Sweden All patients 2,353 (1.9) 0.9 (0.8–0.9) 2.3 (2.2–2.4) 878 (0.7) 0.4 (0.4–0.4) ASA I 444 (1.5) 0.5 (0.4–0.6) 1.9 (1.7–2.1) 145 (0.5) 0.2 (0.2–0.3) ASA II 1,364 (1.9) 0.8 (0.7–0.9) 2.3 (2.2–2.4) 522 (0.7) 0.4 (0.3–0.4) ASA III–IV 545 (2.7) 1.6 (1.4–1.7) 3.3 (3.0–3.6) 211 (1.0) 0.7 (0.6–0.8)

portions of cases in obese patients (BMI ≥ 30), in those over 85 years of age, and in patients with secondary OA were highest in ASA classes III–IV. In GAR, 126 cases of revision were recorded within 5 years, with 59 (47% of total) within 3 months (Table 2). The incidence of death within 5 years was 12.8% (CI 11.8–13.8). In SHAR, 2,353 cases of revision were recorded within 5 years, with 1,030 (44% of total) within 3 months. The incidence of death within 5 years was 8.3% (CI 8.1–8.5). In both cohorts, the cumulative incidence of revision within 3 months and within 5 years was higher in ASA classes III–IV than in ASA class I (within 3 months, GAR: 2.3% versus 0.6%; SHAR: 1.6% versus 0.5%; within 5 years, GAR: 3.3% versus 2.3%; SHAR: 3.3% versus 1.9%). The cumulative incidence was lower for other reoperation than for revision in both cohorts.

1.9 (1.5–2.3) 2.0 (0.7–3.4) 1.8 (1.3–2.2) 2.2 (1.4–3.0) 0.8 (0.8–0.9) 0.6 (0.5–0.7) 0.8 (0.8–0.9) 1.2 (1.0–1.4)

There was a positive association between ASA class and the risk of revision within 5 years (GAR: p = 0.02 for the comparison between ASA class I or II versus III or IV; SHAR: p < 0.001 for the comparison among all ASA classes). Results were unchanged by including only the first procedure in each patient. An association between ASA class and risk of other reoperation within 5 years was detected only in SHAR (GAR: p = 0.6; SHAR: p < 0.001) (Figure). In GAR, ASA classes III–IV were associated with a higher risk of revision (Table 3). However, the association was restricted to within 3 months after primary THA (unadjusted HR: 3.4, CI 1.6–7.4). The association persisted after adjustment for differences in the preoperative baseline characteristics. The risk of revision was also higher in patients with a diagnosis of secondary OA and in obese patients.

Acta Orthopaedica 2019; 90 (4): 324–330

327

Cumulative revision rate (%) in GAR

Cumulative revision rate (%) in SHAR

Cumulative reoperation rate (%) in GAR

Cumulative reoperation rate (%) in SHAR

Years after index operation

Cumulative incidence of revision in (A) GAR and (B) SHAR and cumulative incidence of other reoperation in (C) GAR and (D) SHAR by ASA class (95% CI shown in shading). The association with ASA class was statistically significant in SHAR (p < 0.001 for both revision and other reoperation). In GAR, the p-value was 0.07 for revision but the difference was statistically significant between ASA class III or IV and ASA class I or II (p = 0.02). No association with ASA class was detected in GAR for other reoperation (p = 0.6). Table 3. Associations with the risk of revision (time-invariant HR unless specified) Model

GAR cohort HR (CI)

SHAR cohort HR (CI)

Univariable model ASA I 1 (ref) 1 (ref) ASA II 1.0 (0.5–2.0) 1.6 (1.4–2.0) a,f 1.1 (0.9–1.3) b,f ASA III–IV 3.4 (1.6–7.4) a,c 3.2 (2.6–3.9) a,f 0.7 (0.3–1.7) b,c 1.3 (1.1–1.6) b,f Multivariable model ASA I 1 (ref) 1 (ref) ASA II 1.0 (0.5–1.9) 1.7 (1.4–2.1) a,g 1.2 (1.0–1.3) b,g ASA III–IV 2.7 (1.2–5.9) a,d 3.3 (2.6–4.0) a,g 0.7 (0.3–1.7) b,d 1.4 (1.1–1.6) b,g Sex Male 1 (ref) 1 (ref) Female 0.8 (0.6–1.1) 0.7 (0.6–0.7) Diagnosis Primary OA 1 (ref) 1 (ref) Secondary OA 2.4 (1.7–3.5) 1.4 (1.3–1.6) BMI < 35 1 (ref) 1 (ref) ≥ 35 3.7 (2.3–6.0) 2.6 (2.2–3.1) a,h 1.2 (1.0–1.5) b,h Age < 85 y 1 (ref) 1 (ref) ≥ 85 y 1.6 (0.8–3.3) a,e 1.9 (1.5–2.4) a,i 0.2 (0.1–1.7) b,e 0.6 (0.4–0.8) b,i a HR within the first 3 months. b HR after 3 months and within 5 years. c The change in HR within the first 3 months

and after was statistically significant (p < 0.001). d The change in HR within the first 3 months and after was statistically significant (p = 0.002). e A change in HR within the first 3 months and after was suspected (p = 0.07). f The change in HR within the first 3 months and after was statistically significant (ASA II: p < 0.001, ASA III–IV: p < 0.001). g The change in HR within the first 3 months and after was statistically significant (ASA II: p < 0.001, ASA III–IV: p < 0.001). h The change in HR within the first 3 months and after was statistically significant (p < 0.001). i The change in HR within the first 3 months and after was statistically significant (p < 0.001).

In SHAR, the association of ASA classes III–IV with revision within 3 months was confirmed (unadjusted HR: 3.2, CI 2.6–3.9). The association decreased after 3 months but remained statistically significant (unadjusted HR: 1.3, CI 1.1–1.6). In contrast to GAR, an association with ASA class II was also detected within 3 months (unadjusted HR: 1.6, CI 1.4–2.0). Adjustment for differences in the preoperative baseline characteristics did not importantly modify the associations. In contrast to GAR, sex and age were also associated with the risk of revision. Women had a lower risk of revision. Patients aged over 85 years had a higher risk within 3 months but a lower risk thereafter. The associations between the risk of revision and the diagnosis of both secondary OA and obesity were confirmed in SHAR. In GAR, ASA class was not associated with the risk of other reoperation (unadjusted HR ASA III–IV vs. ASA I: 1.2, CI 0.6–2.5) (Table 4, see Supplementary data). In SHAR, ASA class was associated with a greater risk of other reoperation within 3 months following primary THA than after 3 months and within 5 years (unadjusted HR ASA III–IV vs. ASA I within 3 months: 3.2, CI 2.3–4.3; after 3 months and within 5 years: 1.6, CI 1.2–2.1). The most frequent indications for revision in both cohorts were dislocation, infection, and periprosthetic fracture, and for other reoperation in both cohorts were infection, periprosthetic fracture, and hematoma (Table 5).

Discussion Our study had 3 important findings on outcomes within 3 months after primary THA. 1st, preoperative ASA classes III– IV compared with ASA class I were associated with a more than 3 times higher risk of very early revision. 2nd, preoperative ASA classes III–IV compared with ASA class I were associated with a more than 2 times higher risk of very early

328

Acta Orthopaedica 2019; 90 (4): 324–330

Table 5. Indications for revision and other reoperation within 3 months by ASA grade Geneva Sweden ASA I ASA II ASA III–IV Total ASA I ASA II ASA III–IV Total Factor n = 481 n = 3,496 n = 1,342 n = 5,319 n = 29,280 n = 72,857 n = 20,104 n = 122,241 Indications for revision, n (%) Dislocation 1 (0.2) 5 (0.1) 12 (0.9) 18 (0.3) 30 (0.1) 103 (0.1) 57 (0.3) 190 (0.2) Infection 1 (0.2) 8 (0.2) 7 (0.5) 16 (0.3) 81 (0.3) 356 (0.5) 195 (1) 632 (0.5) Periprosthetic fracture 0 6 (0.2) 7 (0.5) 13 (0.2) 24 (0.1) 80 (0.1) 42 (0.2) 146 (0.1) Aseptic loosening 0 4 (0.1) 3 (0.2) 7 (0.1) 8 (< 0.1) 21 (< 0.1) 8 (< 0.1) 37 (< 0.1) Implant malposition 0 1 (< 0.1) 1 (0.1) 2 (< 0.1) 3 (< 0.1) 8 (< 0.1) 3 (< 0.1) 14 (< 0.1) Other a 1 (0.2) 1 (< 0.1) 1 (0.1) 3 (0.1) 1 (< 0.1) 2 (< 0.1) 3 (< 0.1) 6 (< 0.1) Unknown 0 0 0 0 2 (< 0.1) 3 (< 0.1) 0 5 (< 0.1) Total 3 (0.6) 25 (0.7) 31 (2.3) 59 (1.1) 149 (0.5) 573 (0.8) 308 (1.5) 1,030 (0.8) Indications for reoperation, n (%) Infection 1 (0.2) 11 (0.3) 3 (0.2) 15 (0.3) 45 (0.2) 219 (0.3) 106 (0.5) 370 (0.3) Periprosthetic fracture 0 3 (0.1) 6 (0.4) 9 (0.2) 4 (< 0.1) 16 (< 0.1) 10 (< 0.1) 30 (< 0.1) Hematoma 1 (0.2) 4 (0.1) 1 (0.1) 6 (0.1) 1 (< 0.1) 9 (< 0.1) 8 (< 0.1) 18 (< 0.1) Abductor avulsion 0 2 (0.1) 1 (0.1) 3 (0.1) 1 (< 0.1) 1 (< 0.1) 0 2 (< 0.1) Dislocation 0 0 0 0 3 (< 0.1) 8 (< 0.1) 1 (< 0.1) 12 (< 0.1) Cement problem 0 0 0 0 1 (< 0.1) 3 (< 0.1) 2 (< 0.1) 6 (< 0.1) Other b 0 3 (0.1) 1 (0.1) 4 (0.1) 2 (< 0.1) 7 (< 0.1) 1 (< 0.1) 10 (< 0.1) Unknown 0 0 0 0 1 (< 0.1) 3 (< 0.1) 2 (< 0.1) 6 (< 0.1) Total 2 (0.4) 23 (0.7) 12 (0.9) 37 (0.7) 58 (0.2) 266 (0.4) 130 (0.6) 454 (0.4) a Other includes: infection suspected but not confirmed; hematoma; other material left in joint; nerve injury; delayed healing; b Other includes: infection suspected but not confirmed; pain; allergy to suture; other material left in joint; aseptic loosening.

other reoperation in SHAR. These risks were independent of age, sex, BMI, and diagnosis. 3rd, a substantial proportion of early revision and other reoperation procedures in patients with ASA classes III–IV were performed within 3 months of primary THA. We also found that beyond 3 months and within 5 years after primary THA patients with increased ASA class were not at increased risk of revision in GAR and were only at a slightly higher risk of revision in SHAR. This study has limitations. 1st, the ASA classification system has been criticized because of the subjective nature of the assessment, which has poor inter-observer correlation (Ranta et al. 1997, Mak et al. 2002). Despite this, as noted by Hooper et al. (2012), the ASA classification system has remained the most widely used anesthetic preoperative assessment and the most widely collected tool for measuring comorbidity by arthroplasty registries worldwide. We accept that there may be poor inter-observer reliability when determining between ASA class I and II, but we agree with Hooper et al. that the difference between ASA class I and III is so profound (a normal healthy patient compared with a patient with severe systemic disease) that we believe that the significance of our results, when comparing ASA class I with ASA classes III–IV, was unlikely to be affected by this potential error. 2nd, the ASA class in our data represents only a snapshot of the physical health status of each patient, taken immediately prior to primary THA. This is a general drawback of using ASA class because it is assessed only in the context of surgery. We do not know whether physical health status changed subsequent to THA, and if this had an influence on revision and other reoperation rates. With increasing follow-up, other health

pain.

changes might intervene that would attenuate the influence of a single baseline measure. However, we aimed to determine the influence of preoperative physical health status, to enable the identification of high-risk patients preoperatively, so the possibility of subsequent changes should not detract from our results. 3rd, our results were adjusted for age, sex, BMI, and diagnosis, but other factors may have had a confounding effect. Here, we elected to focus on patient factors that are routinely measured before THA, and so may be used when counselling patients considering THA. We did not adjust for surgical factors, such as surgical approach and implant fixation, which are known also to influence early revision and other reoperation rates (Jämsen et al. 2014, Meneghini et al. 2017). The reason for this decision is that each of these factors is chosen by the surgeon, and these choices may be influenced by ASA class, age, and BMI. Therefore, the inclusion of surgical factors in the models may lead to over-adjustment. We included both procedures in patients who had undergone bilateral THA. This is because the ASA class may change between the 2 operations. Indeed, a sensitivity analysis in GAR and SHAR including only the first procedure in each patient showed similar results to the analyses, including bilateral cases. Nevertheless, since ASA class is associated with outcome, the preoperative ASA class of the 2nd THA would influence the outcome of both the 1st and 2nd THAs. Thus, a degree of correlation in patients with bilateral THAs cannot be excluded. 4th, GAR is a small registry and we included cases since 1996. This might limit the applicability to modern patients;

Acta Orthopaedica 2019; 90 (4): 324–330

however, we compared the results with results in the larger SHAR with recent data and the results were similar. Many studies have suggested that patients with poor overall physical health have a higher risk of early revision. A systematic review found 5 papers that reported greater preoperative comorbidity was associated with a higher risk of revision, with outcome ranging from 6 months to 8 years (Prokopetz et al. 2012). The Charlson Comorbidity Index was used as the measure of comorbidity in these papers. Whilst the Charlson score is widely validated, its limitation is that it simply considers the presence or absence of certain diseases and does not account for their severity. Furthermore, it requires more information than the ASA classification system to complete and is not routinely calculated before THA, or widely collected by registries. To our knowledge, 2 reports of arthroplasty registry data have investigated the influence of ASA class on early revision rate. Hooper et al. (2012) found an adjusted hazard ratio of 1.4 (CI 1.0–2.0) when comparing revision rates of patients with ASA class I versus III within 2 years postoperatively in the New Zealand Joint Registry. This agrees with our findings; however, the effect of increased ASA class on revision rate was less than in our study. This difference may be because Hooper et al. studied the longer period of 2 years postoperatively. Our data indicate that the effect of increased ASA class is highest within 3 months and decreases thereafter. The 2018 annual report of the Dutch Arthroplasty Registry reported graphically the revision rate after primary THA stratified by ASA class (Dutch Arthroplasty Register 2018). Although numerical data are not reported, the survival curve demonstrates a higher revision rate within 3 months of primary THA in ASA classes III– IV cases than ASA class I cases, similar to the trend observed in GAR and SHAR. Data on the rate of other reoperations are scarce and we could find no previous study that had investigated the influence of preoperative health status on other reoperation rate with which to compare our results. This likely reflects several factors. 1st, the previous lack of a formal definition of what constitutes reoperation after arthroplasty. 2nd, study of revisions has taken precedence because they are seen as a more serious complication. 3rd, very few arthroplasty registries collect data on other reoperations. Our study is the first to demonstrate that poor preoperative physical health status as measured with the ASA class is associated with increased risk of early other reoperation. We note the cumulative incidence of other reoperation was higher in GAR. Whilst there may be a real difference in other reoperation rate between the 2 registries, this observed difference alternatively may reflect greater completeness of capture of other reoperations in GAR. That a substantial proportion of early revision and other reoperation procedures, performed in patients with ASA classes III–IV, occurs within 3 months of primary THA is a clinically important finding. It identifies this period of 3 months as critical to efforts to reduce revision and other reop-

329

eration rates. Infection, periprosthetic fracture, and dislocation were the most frequent indications for very early revision and other reoperation. Strategies to reduce the risk of revision targeted to these complications during this very early postoperative period may be designed and implemented for patients with ASA classes III–IV most at risk, and are a key focus for future work. These may include preoperative, perioperative, and postoperative interventions. Such intensive strategies may be appropriate and acceptable in this cohort of patients over this time frame with regards to patient preference and resource constraints. In summary, our study has identified that within 3 months of primary elective THA patients with preoperative ASA classes III–IV have a higher risk of revision and other reoperation. The proposed benefits are improved patient counselling, targeted risk-reduction strategies, and improved risk adjustment between datasets. Supplementary data Table 4 and Appendix are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1605785 RJF, AJS, OR, and AL contributed to the conception and design of the study. RJF, CC, EB, and DO performed the statistical analyses. RJF, AJS, and AL drafted the manuscript. All authors critically revised the manuscript. The authors would like to thank all GAR and SHAR registry staff, and surgeons and patients who have contributed to the registries. Acta thanks Marianne Hansen Gillam and Liza N van Steenbergen for help with peer review of this study.

Cnudde P, Nemes S, Bülow E, Timperley J, Malchau H, Kärrholm J, Garellick G, Rolfson O. Trends in hip replacements between 1999 and 2012 in Sweden. J Orthop Res 2017; 36(1): 432-42. Dutch Arthroplasty Register. Landelijke Registratie Orthopedische Implantaten Annual Report 2018. Ferguson R J, Palmer A J, Taylor A, Porter M L, Malchau H, Glyn-Jones S. Hip replacement. Lancet 2018; 392(10158): 1662-71. Geneva Joint Arthroplasty Registry. Geneva Joint Arthroplasty Registry: Annual Report 2017. Available on request: christophe.barea@hcuge.ch. Hooper G J, Rothwell A G, Hooper N M, Frampton C. The relationship between the American Society of Anesthesiologists physical rating and outcome following total hip and knee arthroplasty: an analysis of the New Zealand Joint Registry. J Bone Joint Surg Am 2012; 94(12): 1065-70. Jämsen E, Eskelinen A, Peltola M, Mäkelä K. High early failure rate after cementless hip replacement in the octogenarian. Clin Orthop Relat Res 2014; 472(9): 2779-89. Lübbeke A, Silman A J, Barea C, Prieto-Alhambra D, Carr A J. Mapping existing hip and knee replacement registries in Europe. Health Policy (New York) 2018; 122(5): 548-57. Mak P H K, Campbell R C H, Irwin M G, American Society of Anesthesiologists. The ASA physical status classification: inter-observer consistency. American Society of Anesthesiologists. Anaesth Intensive Care 2002; 30(5): 633-40. Meneghini R M, Elston A S, Chen A F, Kheir M M, Fehring T K, Springer B D. Direct anterior approach: risk factor for early femoral failure of cement-

330

less total hip arthroplasty. J Bone Joint Surg 2017; 99(2): 99-105. National Joint Registry. National Joint Registry for England, Wales, Northern Ireland and the Isle of Man: 14th Annual Report 2017. Prokopetz J J, Losina E, Bliss R L, Wright J, Baron J A, Katz J N. Risk factors for revision of primary total hip arthroplasty: a systematic review. BMC Musculoskelet Disord 2012; 13: 251. Ranta S, Hynynen M, Tammisto T. A survey of the ASA physical status classification: significant variation in allocation among Finnish anaesthesiologists. Acta Anaesthesiol Scand 1997; 41(5): 629-32.

Acta Orthopaedica 2019; 90 (4): 324–330

Swedish Hip Arthroplasty Register. The Swedish Hip Arthroplasty Register: Annual Report 2016. Swiss National Joint Registry. Swiss National Joint Registry: Annual Report 2018. Thien T M, Chatziagorou G, Garellick G, Furnes O, Havelin L I, Mäkelä K, Overgaard S, Pedersen A, Eskelinen A, Pulkkinen P, Kärrholm J. Periprosthetic femoral fracture within two years after total hip replacement: analysis of 437,629 operations in the Nordic Arthroplasty Register Association database. J Bone Joint Surg Am 2014; 96(19): e167.

Acta Orthopaedica 2019; 90 (4): 331–337

331

Outcome of 881 total hip arthroplasties in 747 patients 21 years or younger: data from the Nordic Arthroplasty Register Association (NARA) 1995–2016 Vera HALVORSEN 1, Anne Marie FENSTAD 2, Lars B ENGESÆTER 2,3, Lars NORDSLETTEN 1,4, Søren OVERGAARD 5,6, Alma B PEDERSEN 6,7, Johan KÄRRHOLM 8,9, Maziar MOHADDES 8,9, Antti ESKELINEN 10,11, Keijo T MÄKELÄ 11,12, and Stephan M RÖHRL 1 1 Division

of Orthopedic Surgery, Oslo University Hospital Ulleval, Oslo, Norway; 2 Norwegian Arthroplasty Registry, Department of Orthopaedic Surgery, Haukeland University Hospital, Bergen, Norway; 3 Department of Clinical Medicine, University of Bergen, Bergen, Norway; 4 University of Oslo, Oslo, Norway; 5 Department of Orthopaedic Surgery, Traumatology and Odense University Hospital, and Department of Clinical Research, University of Southern Denmark, Odense, Denmark; 6 Danish Hip Arthroplasty Register, Denmark; 7 Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; 8 The Swedish Hip Arthroplasty Register, Sweden; 9 Department of Orthopaedics, Institute of Surgical Sciences, Sahlgrenska University Hospital, University of Gothenburg, Sweden; 10 Coxa Hospital for Joint Replacement, and Faculty of Medicine and Life Sciences, University of Tampere, Tampere, Finland; 11The Finnish Arthroplasty Register, Finland; 12 Department of Orthopaedics and Traumatology, Turku University Hospital, Turku. Correspondence: uxvbha@ous-hf.no Submitted 2018-11-19. Accepted 2019-04-01.

Background and purpose — The literature is scarce on the outcome of the youngest patients with total hip arthroplasties (THAs). We analyzed register data, revision risk, and related factors in patients 21 years or younger with THAs in the Nordic Arthroplasty Register Association (NARA). Patients and methods — We included all THA patients 21 years or younger reported during 1995 through 2016 to the Danish, Finnish, Norwegian, and Swedish hip arthroplasty registers and merged these into the NARA dataset. Primary outcome was any implant revision. Results — We identified 881 THAs in 747 patients. Mean age at primary surgery was 18 years (9–21). The indications for THA were pediatric hip diseases (33%), systemic inflammatory disease (23%), osteoarthritis (4%), avascular necrosis (12%), hip fracture sequelae (7%), and other diagnoses (21%). Unadjusted 10-year survival for all THAs was 86%. Comparison between indications showed no differences in survival. Uncemented implants were used most frequently. Survival for uncemented and cemented implants was the same adjusted for sex, indication, head size, and time period for primary surgery. Aseptic loosening was the main cause of revision. Interpretation — Both cemented and uncemented fixations seem to be a viable option in this age group, but with a lower implant survival than in older patient groups.

In children and adolescents small size of the hip bones and anatomic changes due to underlying disease may make total hip arthroplasty (THA) technically demanding, and in the youngest high activity level is a risk factor for revision (Munger et al. 2006, Flugsrud et al. 2007, Prokopetz et al. 2012). It is reasonable to expect revision surgery throughout the patient’s lifespan. In all ages the reduced bone stock after revision surgery may cause later problems, the implant survival may be poor and recovery of function is more strenuous (Lie et al. 2004, Bischel et al. 2012, Adelani et al. 2014, Goodman et al. 2014). Reports decades ago on THAs in patients younger than 21 years were mostly on patients with juvenile chronic arthropathies and most of the components were cemented (Ruddlesdin et al. 1986, Witt et al. 1991, Cage et al. 1992). Historically, long-time cohorts of cemented THAs demonstrate down to 50% implant survival after 12–19 years (Torchia et al. 1996, Wroblewski et al. 2010). Later studies have found more promising results with uncemented implants. Hannouche et al. (2016) found an estimated 90% survival after 10 years in patients aged less than 20 years with almost exclusively uncemented implants with ceramic-on-ceramic bearings. Tsukanaka et al. (2016) found a 10-year survival rate of 70% in a Norwegian register study of 96 cemented and uncemented hips in 81 patients aged less than 20 years. A recent register study of 769 THAs in patients 20 years or younger from England, Wales, Northern Ireland, and the Isle of Man reported a 5-year implant survival of 96%. The patients included were from the period 2003–2017 (Metcalfe et al. 2018).

332

Acta Orthopaedica 2019; 90 (4): 331–337

We present an epidemiologic overview on more recent primary THAs regarding indications and fixation concepts. Implant survivorship and reasons for revision were evaluated. Patients operated in the first years of the study period may have limited applicability to contemporary THAs because of older implant design. Our hypothesis was that newer implants might have better survival compared with those used in the earlier cohort.

Annual frequency 70

Other Pediatric SIDs Hip fracture AVN OA

Patients and methods Data source This study was based on data from the Nordic Arthroplasty Register Association (NARA), which is a collaboration between the national joint replacement registers in Denmark, Finland, Norway, and Sweden (Mäkelä et al. 2014). Data from the registers have been merged into a common NARA minimal dataset combining available parameters in all four countries. The study is reported according to RECORD guidelines. Study population Using the NARA database, we identified all patients 21 years or younger (n = 747) reported to have had 881 primary THAs during 1995–2016. Covariates Information on country of origin, age, sex, indication for THA, calendar year of surgery, type of fixation, implants and articulations, approach, cause of any revision, and date of death were collected. No THAs were excluded even if there were missing values in some categories in the dataset. The diagnoses for primary THA were grouped into 6 categories (Figure 1). Outcome The primary outcome measure was time to 1st revision. A revision was defined as any removal or exchange of components. Revisions were categorized as change of cup, change of stem, change of both stem and cup, or removal of components. In the NARA minimal dataset changing of liner is recorded as cup revision. Only the 1st revision was reported. Statistics Categorical data were cross-tabulated by the chi-square test. Continuous data were described using means (SD), and possible differences were tested with Student’s t-test. All tests were 2-sided and the significance level was set to 0.05. The Kaplan–Meier method was used to calculate unadjusted survival functions and estimates with 95% confidence interval (CI). In order to examine the association between sex, diagnosis, type of fixation, femoral head size, calendar period, and risk of revision, we calculated the hazard ratio (HR) using Cox regression analyses, crude HR, and HR mutually adjusted for

1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 2015

Figure 1. Indications for THAs in patients 21 years or younger in NARA countries 1995–2016. Other: tumors, sequelae after infection, pharmaceutically induced femoral necrosis. Pediatric: developmental dysplasia of the hip (DDH), Perthes, slipped capital femoral epiphysis (SCFE). SIDs: systemic inflammatory diseases including rheumatoid arthritis, ankylosing spondylitis, and other inflammatory diseases. AVN: avascular necrosis. OA: osteoarthritis.

all covariates. We used log–log plots and Schoenfeld residuals for each covariate to test that the Cox proportional hazard model was fulfilled. Bilateral observations do not introduce significant dependency problems in register studies, hence these were included (Ranstam et al. 2011). Ethics, funding, and potential conflict of interest Ethical approval: Denmark: Danish Data Protection Agency nr. 1-16-02-54-17, Finland: National Institute of Helath and Welfare (THL): Dnro THL/1743/5.05.00/2014, Norway: Norwegian Data Protection Authority 03/00058-20/CGN, Sweden: DNR 804-17 Regional ethical committee Gothenburg. Research was funded by NordForsk Grant for NARA. No conflict of interest was reported by the authors.

Results Demographic data and diagnosis The number of primary THAs in patients 21 years or younger was 881 (0.1%) compared with 745,827 primary THAs in patients of all ages during the 22-year study period. Of the 881 THAs, 134 were bilateral procedures (18%). 43% of the bilateral cases were performed in systemic inflammatory disease patients (SID patients). The male:female ratio was close to 1:1, except in Sweden where the ratio was almost 1:2. Mean age at primary THA was 18 (9–21) years (SD 2.4). Mean age at primary surgery was similar for males and females. Pediatric hip disease was the most common indication for THA accounting for 33% of the patients; the second largest group was SID with 23%, OA accounted for 4%, AVN for 12%, hip fractures for 7%, and other for 21%. The indications for THA varied during the

Acta Orthopaedica 2019; 90 (4): 331–337

333

Table 1. Patient demographics for each country with a total of 881 primary THAs in 1995–2016 (numbers in parentheses) Country THAs per country, n (%) Revisions, n (%) Number of deaths Sex, % male Age, mean (SD) Indications, % Osteoarthritis Avascular necrosis Hip fracture SID a Pediatric Other

Denmark

Finland

Norway

Sweden

Total

253 (29) 171 (19) 207 (24) 250 (28) 881 28 (11) 30 (18) 17 (8) 43 (17) 118 (13) 6 2 7 8 23 53 52 50 36 47 17.9 (2.3) 18.3 (2.4) 17.8 (2.4) 18.1 (2.4) 18.0 (2.4) 2.4 11 5.6 18 48 15

8.8 15 7.6 21 11 36

1.4 11 5.3 17 51 15

4.8 12 7.6 35 18 23

4.1 12 6.5 23 33 22

In 1.2% of the operations information on indication was missing. Systemic inflammatory disease.

Table 2. Fixations, head sizes, articulations, surgical approach, and trochanteric osteotomies (n = 881). Values are frequency (%) Fixations Cemented Uncemented Hybrid Reverse hybrid Resurfacing Missing Head size < 32 mm 32 mm > 32 mm Missing Articulation Metal/metal Metal/ceramic Ceramic/ceramic PolyXL/metal PolyXL/ceramic Poly/metal Poly/ceramic Missing Posterior approach a Yes No Missing Trochanteric osteotomy Yes No Missing a

62 (7.0) 659 (74) 36 (4.1) 78 (8.9) 31 (3.5) 15 (1.7)

Table 3. Causes of revisions and revision procedures performed (n = 118). Values are frequency (%) Cause Aseptic loosening Deep infection Periprosthetic fracture Dislocation Pain only Other Procedure a Total replaced Only stem replaced Only cup or liner replaced Girdlestone Other Missing

61 (52) 6 (5.1) 3 (2.5) 11 (9.3) 1 (0.8) 36 (31)

Fixations, head size, articulations, implants, and surgical approach In total, 74% of the implants were uncemented (Table 2). Hip resurfacing arthroplasty was rare (3.6%). We found no obvious association between diagnosis and type of fixation, which was equally distributed among different diagnoses. 25% of the heads were 32 mm and 46% were smaller. Metal on highly crosslinked polyethylene was the most frequent articulation (18%). Type of articulation was missing in 19% of cases. A posterior approach was used in 47% of the operations (Table 2). Implant brands The number of different brands varied from 9 to 22 for cups and 10 to 21 for stems for each of the participating countries. The variety of brands was similar in the latter period (2012– 2016) compared with the first two (1995–2004 and 2005–2011). Implant survival estimates linked to brands cannot be performed due to the heterogenicity of the material and the relatively small number of THAs.

Implant survival and revisions 118 (13%) of the 881 THAs were revised during the study period (Table 1). With any 11 (9.3) reason for revision as endpoint, the 5-year 405 (46) 6 (5.1) unadjusted survival was 94% (CI 92–96), the 221 (25) 39 (33) 180 (20) 3 (2.5) 10-year survival was 86% (CI 83–89), and the 75 (8.5) 29 (25) 15-year survival 73% (CI 68–78) (Figure 2). 30 (25) Cups had a higher revision rate than stems 145 (17) a Finland did not report procedure. Den1 (0.1) (Figure 3). There were 4 types of fixations: 97 (11) mark did not differ between cup, stem, cemented, uncemented, hybrid (cup unce206 (23) or total revision and data are reported as mented, stem cemented), and reverse hybrid 135 (15) “other.” 78 (8.9) (cup cemented, stem uncemented) and in 54 (6.1) addition 31 resurfacing arthroplasties (3.6%). 165 (19) Adjusted data showed no statistically significant survival dif418 (47) ference between cemented and uncemented implants (p = 0.2, 262 (30) Table 4). 201 (23) The numbers per disease were too small to make meaning21 (2.4) ful estimates of survivorship related to disease. 658 (75) Hazard ratio (HR) for revision was analyzed in different time 202 (23) periods (Table 4). Unadjusted HR was 0.5 (CI 0.3–0.8) for the a Finland did not report approach. years 2005–2011 and 0.4 (CI 0.1–1.0) for 2012–2016 compared with 1995–2004. Adjusted for sex, indication, fixation, and head size the HR in the period 2005–2011 was 0.5 (CI 0.3– study period; there was a trend towards declining frequency 0.9) and 0.6 (CI 0.2–1.8) in the period 2012–2016 compared of SID and increasing frequency of pediatric diseases (Figure with 1995–2004. In 2012–2016 only 5 revisions were recorded. 1, Table 1). There were differences among the Nordic coun- Using the period 1995–2011 as HR reference, HR in the period tries between indications, particularly the pediatric and SID 2012–2016 was 0.5 (CI 0.2–1.3) (data not shown in table). The groups, which varied from 11% to 51% and 17% to 35% hazard ratio for revision was 2.5 (95% CI 1.4–4.5) higher for respectively (Table 1). reverse hybrid than for uncemented fixation (Table 4).

334

Acta Orthopaedica 2019; 90 (4): 331–337

Arthroplasty survival (%)

Table 4. Patient- and procedure-related risk factors for THA revision in patients 21 years and younger at surgery adjusted for sex, indications, head size, implant fixation, and time period (n = 881)

100

Variables

Total No. of Unadjusted number revisions HR (95% CI) p-value

Adjusted HR (95%CI) p-value

Sex Men 418 46 1 1 Women 463 72 1.1 (0.8–1.6) 0.6 1.1 (0.7–1.7) Indication (n = 879): Pediatric 290 30 1 1 Osteoarthritis 36 1 0.2 (0.1–1.7) 0.2 – Avascular necrosis 104 7 0.6 (0.2–1.3) 0.2 0.5 (0.2–1.3) Hip fracture 57 6 1.2 (0.5–2.8) 0.7 1.5 (0.5–3.9) SIDs 203 52 1.2 (0.8–1.9) 0.4 0.9 (0.5–1.7) Other 189 22 1.0 (0.5–1.7) 0.9 1.1 (0.6–2.1) Fixation (n = 866): Uncemented 659 69 1 1 Cemented 62 20 1.9 (1.2–3.2) 0.01 1.6 (0.8–3.2) Hybrid 36 5 0.7 (0.3–1.8) 0.5 0.5 (0.2–1.8) Reverse hybrid 78 17 2.4 (1.4–4.0) 0.002 2.5 (1.4–4.5) Resurfacing 31 7 1.8 (0.8–4.0) 0.1 0.9 (0.2–4.7) Head size (n = 806): < 32 mm 405 76 1 1 32 mm 221 4 0.3 (0.1–0.8) 0.01 0.4 (0.1–1.3) > 32 mm 180 8 0.4 (0.2–0.8) 0.02 0.7 (0.3–1.9) Time period: 1995–2004 274 90 1 1 2005–2011 311 23 0.5 (0.3–0.8) 0.007 0.5 (0.3–0.9) 2012–2016 296 5 0.4 (0.1–1.0) 0.05 0.6 (0.2–1.8)

0.8

20 Uncemented Cemented

0.2 0.5 0.9 0.7

0 0

Years from surgery

Arthroplasty survival (%) 100

0.2 0.3 0.002 0.9

0.1 0.5

0.02 0.4

20 Uncemented Hybrid 0 0

Arthroplasty survival (%)

100

20 Endpoint: stem revision Endpoint: cup revision

Uncemented Reverse hybrid

50 0

Years from surgery

Figure 2. Kaplan–Meier revisionfree survival curve. Confidence intervals are shaded. The 10-year survival was 86%.

Years from surgery

Figure 3. Kaplan–Meier survival curves for cups (red) and stems (blue). Confidence interval are shaded. Data from Finland are not included because surgical procedure at revision is not registered.

Patients at risk in Figures 2–4 Years from surgery: 0 2 4 6 8 10 12 14 16 18 20 Figure 2: 881 738 597 474 357 272 196 132 99 60 27 Figure 3: 710 584 462 361 269 201 145 101 72 43 19 Figure 4: Uncemented 659 540 422 330 243 185 127 85 70 45 24 Cemented 62 56 50 45 40 36 28 19 11 3 1 Hybrid 36 30 28 25 22 20 19 16 10 6 1 Reverse hybrid 78 71 63 45 29 14 9 6 4 2 0

0 0

Years from surgery

Figure 4. Kaplan–Meier unadjusted survival curves with confidence intervals (shaded areas) for different fixation methods with uncemented fixation as reference. (a) Uncemented versus cemented; (b) uncemented versus hybrid; (c) uncemented versus reverse hybrid. For adjusted survival see Table 4. 31 resurfacing arthroplasties were not included.

Acta Orthopaedica 2019; 90 (4): 331–337

There were no revisions recorded where removal of components took place. That would probably have been the case if more than the first revision had been reported. Most of the revisions, 52%, were due to aseptic loosening (Table 3).

Discussion The overall 10-year implant survival was 86%. There was no difference in adjusted survival for cemented and uncemented implants. Reverse hybrid fixation had a higher hazard ratio for revision than other fixations. We could not show a convincing trend towards better survivorship in the latter time period. One-third of the patients had pediatric hip disease. The increase in the number of pediatric hip disease indications cannot be explained by changes in DDH screening since the screening programs have been unchanged over the years. Neither have any changes in the incidences of Perthes disease or slipped capital femoral epiphysis been reported in the literature. There are also differences between the countries concerning pediatric hip disease as indication, but a study of hip radiograms at skeletal maturity showed that the prevalence of developmental dysplasia of the hip is on the same level in the Nordic countries (Engesaeter et al. 2013). Moreover, Lohmander et al. (2006) has found for all ages that there were similar THA indications in the Nordic countries. A more stringent diagnostic approach might have taken place in some countries and over the years, explaining the development in this indication. Systematic inflammatory disease as indication has declined over the years. The decline in SID as indication for surgery was expected since powerful disease-modifying drugs (DMDs) have been on the market for more than 20 years. The decline in THA patients with SID might even continue during the next decade. The differences in SID as indication between the different countries may be reflecting that the incidence and prevalence of inflammatory arthropathies have been varying in different reports. Berntson et al. (2003) found an incidence of juvenile arthritis in the Nordic countries varying between 5/100,000 and 36/100,000 in different areas. In accordance with Metcalfe et al. (2018), we did not find associations between indications for THA and implant survival. Hannouche et al. (2016) also found the same when he studied 91 patients, 113 hips, in patients younger than 20 years with THAs with ceramic-on-ceramic bearings. Several authors have found inferior implant survival in SIDs patients, especially in older series (Roach et al. 1984, Chmell et al. 1997, McCullough et al. 2006). One could expect SID patients to be less physically active than other young people and thus that the prosthesis would last longer. On the other hand, using DMDs may have helped patients to be physically more active over the years, hence increasing the risk of wear. We assume that our data on the youngest SID patients showed better survival because DMDs have had a favorable effect on

335

morphological changes in the joint before surgery took place, therefore making the surgery less demanding. NARA data for THA patients in all ages have previously shown low revision rates in patients with pediatric hip disease. Engesaeter et al. (2012) found a 10-year survival of 94% after pediatric hip disease treated by THA for all ages. We could not find such a favorable trend for pediatric hip disease patients, which may be explained by different ages in the populations. Unadjusted 5-year implant survival was 94% in our material; 10-year survival was 86% and 15-year survival 73%. This is a 5-year implant survival comparable to the 96% survival in the recent national register study by Metcalfe et al. (2018). Our result is also comparable to a recent publication from the University of Utah with 145 THAs in patients 30 years and younger included from 2000 to 2015 with a 10-year implant survival of 89% (Makarewich et al. 2018). Our implant survival is considerably better than results reported in a Norwegian study with surgery performed 1987–2010, which had a 10-year survival of 70% with endpoint any revision (Tsukanaka et al. 2016). The difference in survival might be explained by our more recent study period (1995–2016), which did not include inferior implants and bearing surfaces from the 1980s (Havelin et al. 2002). During recent years there has been a trend in the literature towards using uncemented implants, which are thought to perform better in younger patients, but there are diverging results. In a systematic review Gee et al. (2013) found that among 450 primary THAs in patients aged less than 30 years uncemented stems did better than cemented. Schmitz et al. (2013), conversely, found a 10-year survival of 86% with 69 cemented THAs in patients younger than 30 years in a Dutch study. Wroblewski et al. (2010) reported a 65% implant survival at 19 years in 39 hips with cemented Charnley arthroplasties. Mean age at surgery was 18 years in his study. Data from the Australian Orthopaedic Association National Joint Replacement Registry collected from September 1999 through December 2012 of 297 THAs in patients younger than 21 years unveiled a 5-year revision rate of 4.5%. Many of the implants were resurfacings (Sedrakyan et al. 2014). We found that cemented and uncemented implants had similar survivorship (Table 4). Hybrid fixation had an adjusted HR of revision of 2.5 but, due to the relatively small numbers of THAs in each fixation group, the confidence intervals are wide and tend to overlap. Larger numbers would have given us more precise estimates and hence a clearer picture of survivorship for the different fixations (Figure 4). We found similar adjusted HRs for revision between the 3 time periods analyzed (Table 4) and for the period 2012–2016 compared with 1995–2011 (data not shown). The lack of a favorable trend over time might be due to the fact that in 2015 there were only 27 hips at risk (Figure 2) and only 5 revisions were recorded in the period 2012–2016. The main cause of revision in our study was aseptic loosening, which is in accordance with reports from earlier studies

336

(Dudkiewicz et al. 2003, Pakos et al. 2014, Sedrakyan et al. 2014, Hannouche et al. 2016). There were only 9.3% dislocations as cause of revision, even though 46% of the patients were operated with head sizes smaller than 32 mm (Table 4) and 46% of them were combined with posterior approach. 37% had head size smaller than 28 mm combined with a posterior approach (data not shown in table). A speculation might be that younger patients have stronger muscles around the hip preventing dislocations to a greater extent than older patients. The number of deep infections was low, but is under-reported in registers. There might have been soft tissue debridements that were not reported. Older patients often have comorbidities that make them more frail and susceptible to infection. Although our dataset is the largest to date, it should be interpreted with caution. The NARA minimal dataset contained only information common to all the registers in Denmark, Finland, Norway, and Sweden. A weakness in the study is that data on articulations, head sizes, surgical approaches, and revision procedures were not recorded during the entire study period for all countries. Complementing and harmonizing the Nordic register data is an ongoing process. All countries used many different acetabular and femoral components. The combinations of components made the material as a whole highly heterogeneous. The wide diversity of component designs jeopardizes a more detailed analysis of cemented and cementless components. In summary, analyzing data from the NARA dataset on 881 total hip replacements on patients 21 years or younger there was a decline in systemic inflammatory disease as indication for THA, and the overall survival at 10 years was 86% with reverse hybrid fixation showing less favorable survival. Survival for cemented and uncemented implants was the same adjusted for sex, indication, head size, and time period for index surgery.

All the authors made a substantial contribution to the conception of the study. AMF created the dataset with the NARA study group and conducted the statistical analyses. All the authors critically revised the draft prepared by VH. Acta thanks Vincent Busch and Richard N de Steiger for help with peer review of this study.

Adelani M A, Mall N A, Nyazee H, Clohisy J C, Barrack R L, Nunley R M. Revision total hip arthroplasty with retained acetabular component. J Bone Joint Surg Am 2014; 96(12): 1015-20. doi: 10.2106/JBJS.L.01177. Berntson L, Andersson Gare B, Fasth A, Herlin T, Kristinsson J, Lahdenne P, Marhaug G, Nielsen S, Pelkonen P, Rygg M, Nordic Study G. Incidence of juvenile idiopathic arthritis in the Nordic countries: a population based study with special reference to the validity of the ILAR and EULAR criteria. J Rheumatol 2003; 30(10): 2275-82. Bischel O, Seeger J B, Kruger M, Bitsch R G. Multiple acetabular revisions in THA: poor outcome despite maximum effort. Open Orthop J 2012; 6: 488-94. doi: 10.2174/1874325001206010488.

Acta Orthopaedica 2019; 90 (4): 331–337

Cage D J, Granberry W M, Tullos H S. Long-term results of total arthroplasty in adolescents with debilitating polyarthropathy. Clin Orthop Relat Res 1992; (283): 156-62. Chmell M J, Scott R D, Thomas W H, Sledge C B. Total hip arthroplasty with cement for juvenile rheumatoid arthritis: results at a minimum of ten years in patients less than thirty years old. J Bone Joint Surg Am 1997; 79(1): 44-52. Dudkiewicz I, Salai M, Israeli A, Amit Y, Chechick A. Total hip arthroplasty in patients younger than 30 years of age. Isr Med Assoc J 2003; 5(10): 709-12. Engesaeter L B, Engesaeter I O, Fenstad A M, Havelin L I, Karrholm J, Garellick G, Pedersen A B, Overgaard S. Low revision rate after total hip arthroplasty in patients with pediatric hip diseases. Acta Orthop 2012; 83(5): 436-41. doi: 10.3109/17453674.2 Engesaeter I O, Laborie L B, Lehmann T G, Fevang J M, Lie S A, Engesaeter L B, Rosendahl K. Prevalence of radiographic findings associated with hip dysplasia in a population-based cohort of 2081 19-year-old Norwegians. Bone Joint J 2013; 95-B(2): 279-85. doi: 10.1302/0301-620X.95B2.30744. Flugsrud G B, Nordsletten L, Espehaug B, Havelin L I, Meyer H E. The effect of middle-age body weight and physical activity on the risk of early revision hip arthroplasty: a cohort study of 1,535 individuals. Acta Orthop 2007; 78(1): 99-107. doi: 10.1080/17453670610013493. Gee M J, Ajuied A, Shah Z, George M, Bankes M J. Systematic review of total hip arthroplasty in patients under 30 years old. Hip Int 2013; 23(4): 345-51. doi: 10.5301/hipint.5000002. Goodman S B, Hwang K, Imrie S. High complication rate in revision total hip arthroplasty in juvenile idiopathic arthritis. Clin Orthop Relat Res 2014; 472(2): 637-44. doi: 10.1007/s11999-013-3326-7. Hannouche D, Devriese F, Delambre J, Zadegan F, Tourabaly I, Sedel L, Chevret S, Nizard R. Ceramic-on-ceramic THA implants in patients younger than 20 years. Clin Orthop Relat Res 2016; 474(2): 520-7. doi: 10.1007/s11999-015-4546-9. Havelin L I, Espehaug B, Engesaeter L B. The performance of two hydroxyapatite-coated acetabular cups compared with Charnley cups: from the Norwegian Arthroplasty Register. J Bone Joint Surg Br 2002; 84(6): 83945. Lohmander L S, Engesaeter L B, Herberts P, Ingvarsson T, Lucht U, Puolakka T J. Standardized incidence rates of total hip replacement for primary hip osteoarthritis in the 5 Nordic countries: similarities and differences. Acta Orthop 2006; 77(5): 733-40. Lie S A, Havelin L I, Furnes O N, Engesaeter L B, Vollset S E. Failure rates for 4762 revision total hip arthroplasties in the Norwegian Arthroplasty Register. J Bone Joint Surg Br 2004; 86(4): 504-9. Makarewich C A, Anderson M B, Gililland J M, Pelt C E, Peters C L. Tenyear survivorship of primary total hip arthroplasty in patients 30 years of age or younger. Bone Joint J 2018; 100-B(7): 867-74. doi: 10.1302/0301620X.100B7.BJJ-2017-1603.R1. Mäkelä K T, Matilainen M, Pulkkinen P, Fenstad A M, Havelin L I, Engesaeter L, Furnes O, Overgaard S, Pedersen A B, Karrholm J, Malchau H, Garellick G, Ranstam J, Eskelinen A. Countrywise results of total hip replacement: an analysis of 438,733 hips based on the Nordic Arthroplasty Register Association database. Acta Orthop 2014; 85(2): 107-16. doi: 10.3109/17453674.2014.893498. McCullough C J, Remedios D, Tytherleigh-Strong G, Hua J, Walker P S. The use of hydroxyapatite-coated CAD-CAM femoral components in adolescents and young adults with inflammatory polyarthropathy: tenyear results. J Bone Joint Surg Br 2006; 88(7): 860-4. doi: 10.1302/0301620X.88B7.17046. Metcalfe D, Peterson N, Wilkinson J M, Perry D C. Temporal trends and survivorship of total hip arthroplasty in very young patients. Bone Joint J 2018; 100-B(10): 1320-9. doi: 10.1302/0301-620X.100B10.BJJ-20171441.R2. Munger P, Roder C, Ackermann-Liebrich U, Busato A. Patient-related risk factors leading to aseptic stem loosening in total hip arthroplasty: a casecontrol study of 5,035 patients. Acta Orthop 2006; 77(4): 567-74. doi: 10.1080/17453670610012629.

Acta Orthopaedica 2019; 90 (4): 331–337

Pakos E E, Paschos N K, Xenakis T A. Long term outcomes of total hip arthroplasty in young patients under 30. Arch Bone Joint Surg 2014; 2(3): 157-62. Prokopetz J J, Losina E, Bliss R L, Wright J, Baron J A, Katz J N. Risk factors for revision of primary total hip arthroplasty: a systematic review. BMC Musculoskelet Disord 2012; 13: 251. doi: 10.1186/1471-2474-13-251 Ranstam J, Kärrholm J, Pulkkinen P, Mäkelä K, Espehaug B, Pedersen A B, Mehnert F, Furnes O, group Ns. Statistical analysis of arthroplasty data, I: Introduction and background. Acta Orthop 2011; 82(3): 253-7. doi: 10.3109/17453674.2011.588862. Roach J W, Paradies L H. Total hip arthroplasty performed during adolescence. J Pediatr Orthop 1984; 4(4): 418-21. Ruddlesdin C, Ansell B M, Arden G P, Swann M. Total hip replacement in children with juvenile chronic arthritis. J Bone Joint Surg Br 1986; 68(2): 218-22. Schmitz M W, Busch V J, Gardeniers J W, Hendriks J C, Veth R P, Schreurs B W. Long-term results of cemented total hip arthroplasty in patients younger than 30 years and the outcome of subsequent revisions. BMC Musculoskelet Disord 2013; 14: 37. doi: 10.1186/1471-2474-14-37.

337

Sedrakyan A, Romero L, Graves S, Davidson D, de Steiger R, Lewis P, Solomon M, Vial R, Lorimer M. Survivorship of hip and knee implants in pediatric and young adult populations: analysis of registry and published data. J Bone Joint Surg Am 2014; 96(Suppl 1): 73-8. doi: 10.2106/ JBJS.N.00541. Torchia M E, Klassen R A, Bianco A J. Total hip arthroplasty with cement in patients less than twenty years old: long-term results. J Bone Joint Surg Am 1996; 78(7): 995-1003. Tsukanaka M, Halvorsen V, Nordsletten L, Engesaeter I O, Engesaeter L B, Marie Fenstad A, Rohrl S M. Implant survival and radiographic outcome of total hip replacement in patients less than 20 years old. Acta Orthop 2016; 87(5): 479-84. doi: 10.1080/17453674.2016. 1212180. Witt J D, Swann M, Ansell B M. Total hip replacement for juvenile chronic arthritis. J Bone Joint Surg Br 1991; 73(5): 770-3. Wroblewski B M, Purbach B, Siney P D, Fleming P A. Charnley low-friction arthroplasty in teenage patients: the ultimate challenge. J Bone Joint Surg Br 2010; 92(4): 486-8. doi: 10.1302/0301-620X.92B4.23477.

338

Acta Orthopaedica 2019; 90 (4): 338–341

Lower 5-year cup re-revision rate for dual mobility cups compared with unipolar cups: report of 15,922 cup revision cases in the Dutch Arthroplasty Register (2007–2016) Esther M BLOEMHEUVEL 1, Liza N van STEENBERGEN 2, and Bart A SWIERSTRA 2 1 Department of Orthopaedic Surgery, Sint Maartenskliniek, Nijmegen; 2 Dutch Arthroplasty Register (LROI), ’s Hertogenbosch, the Netherlands Correspondence: esther.bloemheuvel@gmail.com Submitted 2019-02-08. Accepted 2019-04-02.

Background and purpose — During revision hip arthroplasty the dual mobility cup (DMC) is widely used to prevent dislocation despite limited knowledge of implant longevity. We determined the 5-year cup re-revision rates of DMC compared with unipolar cups (UC) following cup revisions in the Netherlands. Patients and methods — 17,870 cup revisions (index cup revision) were registered in the Dutch Arthroplasty Register during 2007–2016. Due to missing data 1,948 revisions were excluded and the remaining 15,922 were divided into 2 groups: DMC (n = 4,637) and UC (n = 11,285). Crude competing risk and multivariable Cox regression analysis were performed with cup re-revision for any reason as endpoint. Adjustments were made for known patient characteristics. Results — The use of DMC (in index cup revisions) increased from 23% (373/1,606) in 2010 to 47% (791/1,685) in 2016. Patients in the index DMC cup revision group generally had a higher ASA score and the cups were mainly cemented (89%). The main indication for index cup revision was loosening. In the DMC group dislocation was the 2nd main indication for revision. Overall 5-year cup re-revision rate was 3.5% (95% CI 3.0–4.2) for DMC and 6.7% (CI 6.3–7.2) for UC. Cup re-revision for dislocation was more frequent in the UC group compared with the DMC group (32% [261/814] versus 18% [28/152]). Stratified analyses for cup fixation showed a higher cup re-revision rate for UC in both the cemented and uncemented group. Multivariable regression analyses showed a lower risk for cup re-revision for DMC compared with UC (HR 0.5 [CI 0.4–0.6]). Interpretation — The use of DMC in cup revisions increased over time with differences in patient characteristics. The 5-year cup re-revision rates for DMC were statistically significantly lower than for UC.

Instability and dislocation after total hip arthroplasty (THA) is a common reason for revision surgery according to the implant registers of the Netherlands (22%) and Australia (23%) (LROI 2018, AOANJRR 2017). The dual mobility cup (DMC) is a “cup in a cup” and was developed in the 1970s to combine the low-friction arthroplasty principle of Charnley with the advantage of a big femoral head principle of McKee to increase implant stability (Philippot et al. 2009). Second, the aim of this product was to decrease polyethylene rim damage from contact between femoral neck and acetabular liner and to restore near-normal range of motion. Nowadays, the DMC is a well-accepted treatment option for patients with an increased risk for instability in primary and secondary THA (De Martino et al. 2014). However, most literature has focused on dislocation rates rather than on longevity of the implant. In the Dutch Arthroplasty Register we found a 5-year cup revision rate for DMC of 1.5% (95% CI 1.0–2.3) after primary THA (Bloemheuvel et al. 2019). In the Swedish arthroplasty register Hailer et al. (2012) found a 2-year overall survival percentage of 93% (CI 90–97) for DMC after revision THA. We studied the cup re-revision rates of DMC using data from the Dutch Arthroplasty Register (LROI) and compared these results with unipolar cup (UC).

Patients and methods The Dutch Arthroplasty Register (LROI) started in 2007 and has a completeness of 98% for hip revision arthroplasty (www.lroi-report.nl). The LROI database contains patient, procedure, and prosthesis characteristics. For each component a product number is registered to identify the characteristics of the prosthesis, such as dual mobility or conventional cup.

Acta Orthopaedica 2019; 90 (4): 338–341

Index cup revisions in LROI 2007– 2016 n = 17,870 Excluded: Missing cup data n = 1,948 Index cup revisions analyzed n = 15,922

339

Annual number of DMC

Cumulative cup re-revision rate (%)

900

800

700

Unipolar cup Dual mobility cup

600

500

5 400 4 300

DMC n = 4,637

UC n = 11,285

200

100

Re-revision n = 152

Re-revision n = 814

0 2010

1 0 2011

2012

2013

2014

2015

2016

Years of index cup revision

Figure 1. Patient flow. DMC: dual mobility cup; UC: unipolar cup.

Figure 2. Trend in use of the dual mobility cup (DMC) in revision hip arthroplasty in the period 2010–2016 in the Netherlands (n = 4,637).

The vital status of all patients is obtained on a regular basis from Vektis, the national insurance database on health care in the Netherlands, which records all deaths of Dutch citizens. For this study we included all index cup revisions in the period 2007–2016. An index cup revision was defined as the 1st registered cup revision, isolated or as part of a total hip revision. A cup re-revision was defined as a procedure where at least the cup was exchanged or removed. Closed reduction after a dislocation or incision and drainage for infection without component exchange were not included in the LROI. Information from the primary (index) procedure is only known when the procedure was performed after 2007 and registered in the LROI. Records with a missing cup product number (n = 1,948) were excluded from the 17,870 index cup revisions registered. Thus, 15,922 index cup revisions were analyzed and divided into DMC (n = 4,637) or UC (n = 11,285) (Figure 1). The median follow-up was 6 (2–11) years. Statistics The index UC and DMC revisions were described separately concerning patient and procedure characteristics. Survival time was calculated as the time from index cup revision to cup re-revision for any reason, death of the patient, or end of the follow-up (January 1, 2018). Cumulative crude incidence of cup re-revision was calculated using competing risk analysis, where death was considered to be a competing risk (Lacny 2015). In addition, Kaplan–Meier survival analyses were performed. Multivariable Cox proportional hazard analyses were performed to compare DMC and UC. Adjustments were made for sex, age at surgery, ASA score, and type of fixation to discriminate independent risk factors. BMI, Charnley score, and smoking status were not included as covariates, as these were only available in the LROI database since 2014.

Years after index cup revision

Figure 3. Cumulative incidence of cup re-revision according to type of cup in the period 2007–2016 in the Netherlands (n = 15,922).

Table 1. Types of dual mobility cups used in index cup revision in the period 2007–2016 in the Netherlands (n = 4,637) Type

Cemented Cementless

Biomet Avantage 3,492 Biomet Avantage Reload Biomet Avantage Rev HA Smith & Nephew Polarcup 211 Amplitude Saturne 250 Mathys SeleXys DS Cup 106 Groupe LEpine Cupule Quattro 32 Groupe Lepine Cupule HAP Press-F

86 167 19 194 43 35 2

For all covariates added to the model, the proportional hazards assumption was checked by inspecting log-minus-log curves and met. Reasons for cup re-revision were described and compared using a chi-square test. P-values below 0.05 were considered statistically significant. For the 95% confidence intervals (CI), we assumed that the number of observed cases followed a Poisson distribution. Ethics, funding, and potential conflicts of interests The dataset was processed in compliance with the regulations of the LROI governing research on registry data. No external funding was received. No competing interests were declared.

Results The use of DMC (in index cup revisions) increased from 23% (373/1,606) in 2010 to 47% (791/1,685) in 2016 (Figure 2) with 8 different types of DMC used (Table 1).

340

Acta Orthopaedica 2019; 90 (4): 338–341

Table 2. Patient characteristics in index cup revisions according to type of cup. Values are frequency (%) unless otherwise specified Factor

DMC UC n = 4,637 n = 11,285

Male sex Age, mean (SD) ASA, n (%) I II III–IV Fixation cup Cemented Uncemented Type of revision Partial (cup only) Total revision Reason for index revision a Loosening acetabular component Dislocation Infection Loosening femoral component Girdlestone/spacer Periprosthetic fracture Cup/liner wear Peri-articular ossification Symptomatic metal-on-metal bearing Other

1,445 (31) 74 (10)

3,692 (33) 71 (12)

478 (11) 2,642 (60) 1,287 (29)

1,824 (18) 6,307 (60) 2,308 (22)

4,057 (89) 487 (11)

6,468 (59) 4,554 (41)

3,203 (69) 1,434 (31)

6,411 (57) 4,874 (43)

1,728 (37) 1,619 (35) 185 (4) 673 (15) 167 (4) 223 (5) 665 (14) 157 (3) 234 (5) 707 (15)

5,320 (47) 1,301 (12) 634 (15) 2,120 (19) 534 (5) 445 (4) 1,278 (11) 387 (3) 818 (7) 2,592 (23)

Numbers do not add up to total due to missing data. DMC: dual mobility cup; UC: unipolar cup. a The total proportion is over 100% since more than 1 reason for revision can be registered.

Table 3. Reason for cup re-revision according to type of acetabular cup. Values are frequency (%) Reason for re-revision a Loosening acetabular component Dislocation Infection Loosening femoral component Girdlestone/spacer Periprosthetic fracture Cup/liner wear Peri-articular ossification Symptomatic metal-on-metal bearing Other a

DMC UC n = 152 n = 814 79 (52) 28 (18) 48 (32) 12 (8) 20 (12) 8 (5) 7 (5) 2 (1) 2 (1) 15 (10)

423 (52) 261 (32) 127 (16) 61 (8) 44 (5) 43 (5) 27 (3) 17 (2) 11 (1) 88 (11)

See Footnote under Table 2

Patients who received a DMC had a higher ASA score and 89% of the DMC group was cemented versus 59% of the UC group. The most frequent indication for index cup revision was loosening of the acetabular component (37–47%) in both groups. Dislocation was more frequently registered as reason for revision in DMC (35% vs. 12%), while (suspicion of) infection was more frequently registered in the UC group (15% vs. 4%) (Table 2).

Table 4. Crude 5-year cumulative incidence (%) of cup re-revision according to type of acetabular cup. Competing risk was used Factor Overall Cup fixation Cemented Uncemented

5-year cumulative incidence of cup-re-revision Dual mobility cup Unipolar cup n % (CI) n % (CI) 4,637

3.5 (3.0–4.2)

11,285 6.7 (6.3–7.2)

4,057 487

3.6 (3.0–4.4) 3.7 (2.3–6.0)

6,466 7.4 (6.7–8.1) 4,554 5.7 (5.1–6.5)

Table 5. 5-year cumulative incidence (%) of cup re-revision according to type of acetabular cup using Kaplan–Meier survival analyses Factor Overall Cup fixation Cemented Uncemented

5-year cumulative incidence of cup-re-revision Dual mobility cup Unipolar cup n % (CI) n % (CI) 4,637

3.8 (3.2–4.4)

11,285 6.9 (6.3–7.5)

4,057 487

3.9 (3.1–4.6) 3.8 (3.6–4.0)

6,466 7.7 (6.9–8.5) 4,554 5.9 (5.1–6.7)

Over half of the cup re-revisions were performed for loosening of the acetabular component. Dislocation was the 2nd most frequent reason for cup re-revision (32%) in the UC group, while this was 18% in the DMC group. Suspicion for infection was the 2nd most frequently registered reason (32%) for cup re-revision in the DMC group, compared with 16% in the UC group (Table 3). From the 79 DMC cup re-revisions that loosened, 67 were cemented. The 5-year crude re-revision rate of DMC was 3.5%(CI 3.0–4.2) and 6.7% (CI 6.3–7.2) for UC (Figure 3). Stratified analyses according to type of cup fixation (cemented versus uncemented) showed comparable differences in 5-year crude cumulative incidence of re-revision in favor of the DMC group, both using competing risk analysis (Table 4) and Kaplan-Meier survival analysis (Table 5). Multivariable survival analyses showed an adjusted hazard ratio of 0.5 (0.4–0.6) for re-revision of DMC compared with UC. Adjustments were made for sex, age at surgery, ASA score, and type of fixation to discriminate independent risk factors.

Discussion This large register study in the Netherlands showed lower cup re-revision rates of DMC compared with UC. Currently, DMC is increasingly used in both primary and revision hip arthroplasty (Darrith et al. 2018, Bloemheuvel et al. 2019). A recent systematic review from Darrith et al. (2018) containing all English-language articles dealing with dual

Acta Orthopaedica 2019; 90 (4): 338–341

mobility (primary and revision) arthroplasty between 2007 and 2016 showed low rates of dislocation (primary 0.5% and revision 2%). The overall survival of the DMC in revision THA was 97% at a mean of 5 years. A limitation of this study is that it could not distinguish between total and partial revisions. The number of register studies of revision DMC is scarce. Gonzalez et al. (2017) compared DMC and UC THA for prevention of dislocation after revision THA. In this prospective hospital registry-based cohort including all total and cup-only revision THAs (n = 316) they found a lower incidence of dislocation in the case of a DMC (2.7% versus 7.8%) but did not study the longevity of the implant. In 2012 a register study based on 228 patients from the Swedish Hip Arthroplasty Register showed 7% overall rerevision rates for any reason after a DMC at 2 years follow-up (Hailer et al. 2012). Until our study, this was the only register study focusing on re-revision rates according to type of cup. We cannot compare their outcome with our results, as our endpoint was cup re-revision and not overall re-revision. A limitation of register studies is the risk for selection bias. It is possible that different cup designs were used for different types of revisions or different types of patients. Therefore, we examined the patient characteristics in detail. We found higher ASA scores in the DMC group, but after correction for casemix factors DMC still showed lower 5-year revision rates compared with UC. Recent annual reports from the Swedish and Australian hip registers found higher ASA scores in case of revision surgery. (AOANJRR 2016, SHAR 2016). However, they did not distinguish between types of cup. Besides differences between patient characteristics we also examined differences in fixation method. In our study 89% of DMC were cemented, compared with 59% in UC. The amount of cup re-revision because of loosening was the same in the DMC and UC group (52%). We performed stratified analyses to correct for difference in fixation method between DMC and UC and still found a lower cup re-revision rate for DMC compared with UC. The annual report from Sweden showed a trend towards an increased use of cemented DMC in cup revisions (34% of the revision cases received a cemented Avantage cup) (SHAR 2016). However, these revision data were not analyzed in subgroups, for example type of fixation. It is also interesting to analyze differences between various DMC designs as the choice of implant might depend on doctor or hospital preferences. Hopefully, after a few more years the numbers will increase and we shall be able to do further analyses. Nevertheless, register studies have a limited possibility to analyze differences in patient characteristics as this depends strongly on the number of registered variables. Therefore, registries should be taken along with prospective cohort studies, in order to collect a more extensive set of patient variables. Our database on revision hip arthroplasties does not contain information on the procedures performed before the start of the LROI in 2007. Therefore, we do not know the type and fol-

341

low-up of the primary procedure as well as the primary diagnosis of the patient. We do not know whether the 1st revision procedure (defined as index revision) included in our revision hip arthroplasty database was really the 1st revision of a hip or a consecutive revision procedure. On the other hand, including all revision hip arthroplasties available in the LROI resulted in the largest population-based study to date of almost 18,000 cup revisions with a median follow-up of 6 years. In summary, the use of DMC in cup revisions increased over time with differences in patient characteristics and indications. The 5-year cup re-revision rates for DMC were statistically significantly lower than for UC. This promising mid-term result justifies continued use of DMC in revision hip arthroplasty in anticipation of longer term results.

All authors contributed to the conception of the study, data analysis, and preparation of the manuscript. Acta thanks Aare Märtson and Petri Virolainen for help with peer review of this study.

Australian Orthopaedic Association National Joint Replacement Registry. Annual Report. Adelaide: AOA; 2016. https://aoanjrr.sahmri.com/ documents/10180/576950/Hip%2C%20Knee%20%26%20Shoulder%20 Arthroplasty Australian Orthopaedic Association National Joint Replacement Registry. Annual Report. Adelaide: AOA; 2017. https://aoanjrr.sahmri.com/ documents/10180/397736/Hip%2C%20Knee%20%26%20Shoulder%20 Arthroplasty Bloemheuvel E M, van Steenbergen L N, Swierstra B A. Dual mobility cups in primary total hip arthroplasties: trend over time in use, patient characteristics, and mid-term revision in 3,308 cases in the Dutch Arthroplasty Register (2007–2016). Acta Orthop 2019; 90(1): 11-14. Darrith B, Courtney P M, Della Valle C J. Outcomes of dual mobility components in total hip arthroplasty: a systematic review of the literature. Bone Joint J 2018; 100-B (1): 11-19. De Martino I, Triantafyllopoulos G K, Sculco P K, Sculco T P. Dual mobility cups in total hip arthroplasty. World J Orthop 2014; 5(3): 180-7. Gonzalez A I, Bartolone P, Lubbeke A, Dupuis Lozeron E, Hoffmeyer P, Christofilipoulus P. Comparison of dual-mobility cup and unipolar cup for prevention of dislocation after revision total hip arhtroplasty. Acta Orthop 2017; 88(1): 18-23. Hailer N P, Weiss R J, Stark A, Karrholm J. Dual-mobility cups for revision due to instability are associated with a low rate of re-revisions due to dislocation: 228 patients from the Swedish Hip Arthroplasty Register. Acta Orthop 2012; 83 (6): 566-71. Lacny S, Wilson T, Clement F. Roberts D J, Flaris P D, Ghali W A, Marshall DA. Kaplan–Meier survival analysis overestimates the risk of revision arthroplasty: a meta-analysis. Clin Orthop Relat Res 2015; 473(11): 3431-42. LROI. Annual report 2018. www.lroi-report.nl Philippot R, Camilleri J P, Boyer B, Adam P, Farizon F. The use of a dualarticulation acetabular cup system to prevent dislocation after primary total hip arthroplasty: analysis of 384 cases at a mean follow-up of 15 years. Int Orthop 2009; 33(4): 927-32. Swedish Hip Arthroplasty Register. Annual Report; 2016. https://registercen trum.blob.core.windows.net/shpr/r/Annual-Report-2016-B1eWEHmHM.pdf

342

Acta Orthopaedica 2019; 90 (4): 342–347

Centenarian hip fracture patients: a nationwide population-based cohort study of 507 patients Mathias MOSFELDT 1, Christian M MADSEN 2, Jes B LAURITZEN 3, and Henrik L JØRGENSEN 4 1 Department of Orthopedic Surgery & Department of Molecular Medicine and Surgery, Karolinska University Hospital & Karolinska Institute, Stockholm, Sweden; 2 Department of Clinical Biochemistry, Herlev Hospital, University of Copenhagen, Herlev, Denmark; 3 Department of Orthopedic Surgery, Bispebjerg Hospital, University of Copenhagen, Copenhagen, Denmark; 4 Department of Clinical Biochemistry, Hvidovre Hospital, University of Copenhagen, Hvidovre, Denmark Correspondence: mathiasmosfeldt@gmail.com Submitted 2018-09-10. Accepted 2019-02-24.

Background and purpose — Several studies suggest a global increase of centenarians during the 21st century. We describe temporal trends of hip fracture incidence and mortality in this group and compare these patients with a group of younger hip fracture patients with regards to comorbidities and mortality. Patients and methods — The full study population included all hip fractures that occurred in Denmark (n = 154,047) between 1996 and 2012. Patients aged 100 or above were identified (n = 507) and hip fracture patients between the ages of 70 to 99 years (n = 124,007) were used for comparison. Data were accessed from national registries. Trends in incidence over time were analyzed using a loglinear regression model, mortality was analyzed using the Kaplan–Meier estimator and trends in mortality over time were analyzed using a log-binomial regression model to obtain relative risk estimates. Results and interpretation — The centenarian patients had fewer comorbidities than the younger comparison group, but mortality was higher at all timepoints. There was no statistically significant change in mortality over time but the incidence of hip fracture among centenarians decreased during the same time period. Our findings describe the characteristics of an emerging group of hip fracture patients and could be of use in the planning of healthcare in the years to come.

As populations are growing increasingly older, new groups of frail patients are emerging. Studies of these patient groups provide insight that might be useful in treatment but also describe characteristics of a population that has survived longer than their peers. It has been estimated that there will be a global increase from almost half a million centenarians in 2015 to between 13 and 50 million during the 21st century (Robine and Cubaynes 2017). Most currently available studies on centenarian hip fracture patients are based on patient populations that are too small to describe the evolution of the group. Previous studies on hip fracture patients in general have showed that while the age-adjusted incidence in patients aged 50 and older declined between 1993–1996 and 2001–2005, the incidence rate for patients aged above 90 more than doubled (Bergstrom et al. 2009). We describe temporal trends of incidence and mortality for hip fractures in centenarian patients and compare this group with younger hip fracture patients with regards to comorbidities and mortality in order to gain a greater understanding concerning a group of patients that is expected to increase rapidly during the 21st century. The study was performed using information on the entire Danish population including all hip fractures that occurred over a 17-year period.

Patients and methods Study population The Danish National Patient Registry was searched for all patients in Denmark aged 18 years or above admitted with a hip fracture (ICD-10 codes DS720 [femoral neck], DS721 © 2019 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group, on behalf of the Nordic Orthopedic Federation. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. DOI 10.1080/17453674.2019.1602386

Acta Orthopaedica 2019; 90 (4): 342–347

[pertrochanteric], and DS722 [subtrochanteric]) from January 1, 1996 to December 31, 2012. The patients were only included once with the first hip fracture that occurred in this period. The full cohort consisted of 154,047 patients. From this cohort we identified 507 patients aged 100 years or above with a hip fracture during this period. We included all 124,007 hip fracture patients between the ages of 70 and 99 years as a comparison group. Aggregated data on annual mortality rates for all individuals aged 99 years or older in the Danish population were available from Statistics Denmark (Danmarks Statistik—statistikbanken.dk) and were included for additional comparison. Data sources: national registries All Danish citizens are registered in the Civil Registration System (CRS) using a unique 10-digit civil registration number (CRN). Demographic information such as vital status and emigration is available from the CRS on everyone residing legally in Denmark. Information from national registries can be linked for the unique individual using the CRN as it is used in all public records. In addition, birth date and sex can be extracted from the CRN. In Denmark, all contacts and admissions to Danish hospitals are registered in the Danish National Patient Registry (DNPR). This includes data on all somatic hospital admissions dating back to 1977 and has since 1995 also included data on outpatient visits. The DNPR contains information on discharge diagnosis (only 1 is registered) or other secondary diagnostic codes in the form of International Disease Classification (ICD) codes (version 8 and 10). For this study, data from the DNPR were available from 1995 and onwards. A detailed description of the databases used can be found in a previous article (Madsen et al. 2016). Data from the national registries mentioned was accessed via Statistics Denmark. Outcomes and covariates For the endpoint mortality, date of admission was used as the index date for mortality rates. The patients were followed until death, emigration, or end of follow-up (November 13, 2014), whichever came first. For classification of cause of death in the centenarian cohort, we included information on the underlying cause of death as reported to the Danish Register of Causes of Death (Helweg-Larsen 2011). This information was available only for the 321 deaths occurring in the period 2002 to 2012. Underlying causes of death were grouped according to the overall chapters of the ICD-10 classification as indicated. Comorbidity was included in the form of the Charlson Comorbidity Index (CCI) and the individual comorbidities making up the index. This was based on data from the DNPR, which records all hospital contacts. If a patient had a hospital contact registered in the DNPR before the time of the fracture with 1 of the ICD-10 codes constituting the comorbidities in the CCI, the patient was listed as having that comorbidity. Hence, a CCI of 0 indicates that the patient has not had a hos-

343

pital contact registered in the DNPR with any of the comorbidities constituting the CCI. The coding of comorbidities and the ICD-10 codes used for the individual comorbidities were done as described by Quan et al. (2005). All prescription drugs sold in Denmark are registered in the Danish National Prescription Database (Kildemoes et al. 2011). Use of antiosteoporotic medication on admission was defined as retrieving a prescription for antiosteoporotic medication (ATC code: M05B) during the last 6 months before the hip fracture date. Statistics Differences in baseline characteristics were analyzed using Pearson’s chi-squared test for categorical variables and the Mann–Whitney U-test for continuous variables. Yearly crude absolute incidence rates were calculated as the number of centenarian hip fracture cases each year divided by the total number of centenarians in the Danish population by January 1 each year and expressed as events per 1,000 persons, as done in similar previous publications (Rosengren et al. 2017). This gives a close estimate of the incidence rate obtained by summing up the individual’s person-time at risk, as the competing risk of dying is balanced by the new individuals that enter the group of persons above 100 years of age during the year. Only the first hip fracture during the study period after age 100 was included. Trends in incidence over time were analyzed using a log-linear regression model assuming a Poisson distribution. Including a dispersion parameter did not affect the model fit, and hence there was no overdispersion. The model was constructed as number of cases as a function of calendar year with number of persons at risk each year as offset. Mortality is given as the proportion of patients dead at certain timepoints and compared in univariate analysis using Pearson’s chi-squared test. Mortality was further analyzed using the Kaplan–Meier estimator and compared using the log-rank test. Median survival times were similarly estimated with the Kaplan–Meier estimator. Trends in 30-day and 1-year mortality over time were analyzed using a log-binomial regression model to obtain relative risk estimates. The model included the dichotomous variable death as a function of calendar year or calendar year, age, and sex. As the aim was to describe change in mortality over time and not to elucidate possible mechanisms or adjust for possible changes, only age and sex were included as additional covariates. This means that the reported estimates could be biased by residual confounding. 2-sided p-values < 0.05 were considered statistically significant. All analyses and data management were conducted using SAS version 9.3 (SAS Institute, Cary, NC, USA) through a secure remote connection provided by Statistics Denmark. Ethics, funding, and potential conflicts of interest The study was approved by the Danish Data Protection Agency (j.nr. 2012-58-0004.) whereas ethical committee approval is not required for this type of observational study according to

344

Acta Orthopaedica 2019; 90 (4): 342–347

Table 1. Basic characteristics for centenarians and comparison group. Values are frequency (%) unless otherwise specified Comparison Centenarians group

p-value

n 507 124,007 NA Women 431 (85.0) 91,439 (73.7) < 0.001 Age (years), median (range) 101 (100–111) 83 (70–99) NA Charlson Comorbidity Index, median (range) 0 (0–7) 1 (0–19) < 0.001 Charlson Comorbidity Index < 0.001 0 344 (67.9) 56,451 (45.5) 1 87 (17.2) 27,421 (22.1) 2 46 (9.1) 19,796 (16.0) ≥ 3 30 (5.9) 20,339 (16.4) Renal disease 2 (0.4) 2,192 (1.8) 0.01 Liver disease, mild 1 (0.2) 873 (0.7) 0.3 Liver disease, severe 0 217 (0.2) 1.0 Congestive heart failure 85 (16.8) 27,568 (22.2) 0.003 Myocardial infarction 21 (4.1) 7,622 (6.2) 0.06 DM wIith complications 3 (0.6) 2,918 (2.4) 0.005 DM without complications 8 (1.6) 8,567 (6.9) < 0.001 Cerebrovascular disease 50 (9.9) 18,322 (14.8) 0.002 Peripheral vascular disease 6 (1.2) 7,480 (6.0) < 0.001 Pulmonary disease 12 (2.4) 12,183 (9.8) < 0.001 Ulcer disease 18 (3.6) 7,074 (5.7) 0.04 Malignancy 22 (4.3) 13,836 (11.2) < 0.001 Solid metastatic tumor 1 (0.2) 1,488 (1.2) 0.04 Rheumatic disorders 6 (1.2) 4,670 (3.8) 0.002 Dementia 16 (3.2) 10,641 (8.6) < 0.001 Paralysis 0 335 (0.3) 0.7 Fracture type 0.06 Femoral neck 273 (53.9) 73,125 (59.0) Pertrochanteric 203 (40.0) 43,569 (35.1) Subtrochanteric 31 (6.1) 7,299 (5.9) Use of antiosteoporotics 6 (1.2) 7,921 (6.4) < 0.001 DM: Diabetes mellitus

Table 2. Cause of death for centenarian hip fracture patients occurring in the period 2002–2012 from the Danish Register of Causes of Death. Values are frequency (%) Cause of death

ICD-10 codes

Diseases of the circulatory system External causes of morbidity and mortality (including falls and fractures) Ill-defined and unknown causes of mortality Diseases of the respiratory system Mental and behavioral disorders Diseases of the digestive system Endocrine, nutritional, and metabolic diseases Neoplasms Diseases of the blood and blood-forming organs and certain disorders involving the immune mechanism Diseases of the musculoskeletal system and connective tissue Certain infectious and parasitic diseases Diseases of the skin and subcutaneous tissue Diseases of the genitourinary system Diseases of the nervous system

I00–I99

94 (29)

V01–Y98 R95–R99 J00–J99 F00–F99 K00–K93

75 (23) 65 (20) 24 (7.5) 22 (6.9) 11 (3.4)

E00–E90 C00–D48

8 (2.5) 7 (2.2)

D50–D89

5 (1.6)

M00–M99 A00–B99

3 (0.9) 2 (0.6)

L00–L99 N00–N99 G00–G99

2 (0.6) 2 (0.6) 1 (0.3)

Danish law. No funding was received for this study and authors have no potential conflicts of interest regarding this investigation.

Results Patient characteristics Comorbidities (Table 1) A higher percentage of centenarian hip fracture patients had a Charlson Comorbidity Index (CCI) of 0 than was the case for the comparison group of patients between 70 and 99 years of age. Of the centenarians, 68% had a CCI of 0 versus 46% in the comparison group and, similarly, a lower percentage of centenarians had scores of 1, 2, and ≥ 3 as well (17% vs. 22%, 9.1% vs. 16%, and 5.9% vs. 16% respectively), indicating fewer comorbidities among the centenarian hip fracture patients than their younger peers.

Mortality Of the 507 centenarian patients, 484 died during the follow-up period (median: 924 days [385–2121]). Only 23 centenarians were alive at the end of follow-up. The proportion of centenarian patients that had died within 30 days, 90 days, 1 years, 2 years, and 5 years were 34%, 49%, 66%, 78%, and 94% respectively. Mortality was lower in the comparison group at all timepoints as expected: 11%, 19%, 31%, 42%, and 64% (all p < 0.001). In comparison, for all individuals aged 99 years or above in the Danish population from 1996 to 2012 the annual mortality rate was 48% (Figure 1). The median estimated survival time was 97 days and 1,029 days in the centenarian and comparison group respectively (logrank p < 0.0001). The difference in median survival between male (70 days) and female (102 days) centenarians was not statistically significant (logrank p = 0.35). The underlying causes of deaths in the 321 centenarians dying in the period 2002 to 2012 are given in Table 2. Temporal trends in incidence and mortality The annual number of centenarian hip fracture patients in Denmark remained relatively stable and varied between 22 (2003 and 2004) and 39 (2007). The overall incidence rate for the entire period was 47.0/1,000 persons. There was an estimated yearly decrease in the incidence rate of 3.4% (1.7–5.1%, p < 0.001) over the entire study period (Figure 2). The mean age of all centenarians in the Danish population of 101 years was constant from 1996 to 2012. There was no statistically significant change in mortality per calendar year for either 30-day (–1.0% [–3.3 to 1.5], p = 0.4) or 1-year (–0.1% [–1.3 to 1.2], p = 0.9) mortality (Figure 3). Adjusting for the possible confounders age and sex had no statistically significant effect on the estimated yearly changes in

Acta Orthopaedica 2019; 90 (4): 342–347

345

Survival probability

Incidence rate (per 1,000 persons)

Proportion dead (%)

1.0

100

0.8

Comparison group

80 60

1-year 60

0.6 40 Centenarians

0.4

30-day

20 20

0.2 Median survival time 97 days 0.0

100

150

200

250

300

350

400

0 1996 1998 2000 2002 2004 2006 2008 2010 2012

Days after fracture

Figure 1. Kaplan–Meier survival curves after hip fracture for centenarians and comparison group of patients between 70 and 99 years of age.

0 1996 1998 2000 2002 2004 2006 2008 2010 2012

Year

Figure 2. Trend in hip fracture incidence for centenarians.

Figure 3. Temporal trends in mortality for centenarians.

30-day (–0.5% [–2.9 to 1.9], p = 0.7) and 1-year (0.2% [–1.1 to 1.4], p = 0.8) mortality. The data were also aggregated into 4, 5-year time periods to account for yearly variations. There was no difference in 30-day and 1-year mortality between any of the time periods. Specifically, for the period 1996–2000 compared with the period 2009–2012 the relative risk was 1.1 [0.82–1.6, p = 0.4] for 30-day mortality and 0.98 [0.82–1.2, p = 0.8] for 1-year mortality.

Discussion The incidence of hip fractures in centenarians in Denmark has decreased over the last 2 decades, but the mortality rate following hip fracture among centenarians has remained unchanged aside from negligible yearly variations. In comparison with hip fracture patients between 70 and 99 years of age, the centenarians had fewer recorded comorbidities. Studies of these patient groups provide insight that might be useful in treatment but also describe characteristics of a population that has survived longer than their peers. Studies on the limits of aging in humans have suggested that the maximal lifespan is approximately 115 years (Dong et al. 2016). In that study, they investigated data from the Human Mortality Database with data from 40 countries and territories and found that while previous increases in life expectancy had been mainly due to decreases in early-life mortality, recent data show a decline in late-life mortality and it would seem that there is still potential for further development of increasing life expectancy. Strengths and limitations The primary limitations of the present study include the lack of information on covariates such as BMI, smoking, frailty, and other variables not available from the national registries. The

information on comorbidities is based on hospital visits and hence does not consider comorbidities that have not led to a hospital visit. Therefore, comorbidities solely treated by general practitioners and not leading to hospital contact will not be captured by the Charlson Comorbidity Index based on the registries. Another limitation includes the fact that the included patients were a mix of first and subsequent hip fractures, which could affect the mortality and comparability over time. The strengths include the above-mentioned populationbased design and the use of the high-quality Danish registries. These allow extensive information with high validity on hip fracture diagnoses and mortality. Unlike retrospective chart reviews, the information can be collected in a completely unbiased way from the registries. Comparison with literature Previous studies have shown that the incidence of hip fractures has shifted towards older age (Bergstrom et al. 2009) and that the group of nona- and octogenarians as hip fracture patients is increasing. In some respects, this has created a relatively new population of frail patients with this diagnosis. As is noted in the article by Bergström et al., this makes the population of hip fracture patients even less homogeneous and presents an increasing challenge in management. As centenarian hip fracture patients are an emerging patient population, there are not many previous studies regarding incidence and outcome; however, it is noted in 1 of the articles that hip fractures were the leading cause of hospital admission in centenarians (Rodriguez-Molinero et al. 2010). It was also noted that centenarians had a 7 times higher incidence of hip fractures than any other age group. The study included all 162 centenarian hip fracture patients in Spain in 2005. The incidence rate was 3.8 %, similar to the 4.7% in our study. However, they found a 16% 1-month mortality rate, approximately half of what we found.

346

Italy and Spain are both among the countries with the highest life expectancies in the world. The centenarian age group in Italy has the highest rate of increase in numbers of the geriatric population. In a cohort of 7,830 centenarians collected between 2004 and 2011, the number of patients who suffered hip fractures was 259, giving an incidence rate of 23/1,000 per year in contrast to the overall incidence rate of 47/1,000 in our study (Mazzola et al. 2016). This is in line with a previous study where the incidence rates have been found to be much higher in Scandinavia than in the rest of Western Europe (Dhanwal et al. 2011). However, only patients who were discharged alive were included in that study and there is no information on in-hospital mortality. Patients were followed up for 2 years and the probability of survival was 32% in the hip fracture cohort compared with 48% in the reference group of centenarians that did not suffer a hip fracture. In comparison, 2-year survival in our study was considerably lower at 23%. The hazard function in the study by Mazzola et al. showed that the greatest risk was 3–4 months postoperatively with an approximation of the trends afterwards in line with previous literature stating that the greatest risk for mortality is in the first 3 months after surgery (Haentjens et al. 2010). In the study by Mazzola et al. (2016), 81% of the centenarians had a Charlson Comorbidity Index of 0 compared with 68% in our study. In a single-center study from Spain on risk factors and mortality rates comparing 33 centenarians with 99 nonagenarian controls (Barcelo et al. 2018), they found that these groups had similar in-hospital outcomes but the centenarians had more postoperative complications. 3-month and 1-year mortality rates were approximately double with 41% and 21% at 3 months, and 62% and 30% for centenarians and nonagenarians respectively. CCI was not different between the groups and while it is difficult to draw conclusions from such a small number of subjects it is noteworthy that the centenarians had fewer readmissions than the nonagenarians after 3 months and after 1 year. Concerning temporal trends, previous studies have shown regional differences in hip fracture incidence, which in the Western world seems to have stabilized and decreased during the latest decades whereas it has increased in the developing world (Morin et al. 2013). In Denmark specifically, the incidence of hip fractures declined by about 20% between 1997 and 2006 (Abrahamsen et al. 2010). Our study period coincides with the widespread introduction of antiosteoporotic treatment and an increased focus on fracture prevention in the elderly, which may contribute to the declining incidence rates. Improvements in nutrition might also influence declining rates of hip fracture incidence during the study period. Also during the study period, there has been a decrease in smoking habits in Denmark (Ng et al. 2014), which might contribute positively to bone health. Furthermore, and also in line with our results, showing no change in mortality rates over time, a Swedish populationbased study found that mortality following hip fracture had

Acta Orthopaedica 2019; 90 (4): 342–347

remained unchanged from 1995 to 2012 (Karampampa et al. 2015). This also included the age group 95+, but no specific information was given for those aged 100 years or above. Conclusions and implications Our study is one of the largest conducted to date concerning centenarian hip fracture patients. Mortality was higher in centenarians than in younger patients but seemed stable over time aside from slight variations from year to year. There was a yearly decrease in incidence rates for centenarian hip fractures over the entire study period. The centenarians in our study seemed to have fewer comorbidities than their younger counterparts. Our findings support the “compression of morbidity” hypothesis of aging (Fries 1980) for this patient group, as it seems that many have reached an advanced age while having few or no comorbidities. This would suggest that for many of the centenarians in this study, onset of morbidity occurred late and was relatively short before their death, occurring not long after their hip fracture. This could be an important consideration in the planning of healthcare for the coming decades as patients are expected to reach increasingly higher ages, but the patients in this study who lived for more than 100 years had fewer registered diagnoses from hospital contacts than the younger hip fracture patients.

CMM, HLJ, and JBL designed the study. MM and CMM wrote the initial draft of the manuscript, CMM and HLJ performed statistical analysis and calculations, and all authors were involved in revisions of the manuscript. Acta thanks Margareta Hedstrom and Rami Madanat for help with peer review of this study.

Abrahamsen B, Vestergaard P. Declining incidence of hip fractures and the extent of use of anti-osteoporotic therapy in Denmark 1997–2006. Osteoporos Int 2010; 21(3): 373-80. doi: 10.1007/s00198-009-0957-3. Barcelo M, Francia E, Romero C, Ruiz D, Casademont J, Torres OH. Hip fractures in the oldest old: comparative study of centenarians and nonagenarians and mortality risk factors. Injury 2018; 49(12): 2198-202. doi: 10.1016/j.injury.2018.09.043. Bergstrom U, Jonsson H, Gustafson Y, Pettersson U, Stenlund H, Svensson O. The hip fracture incidence curve is shifting to the right. Acta Orthop 2009; 80(5): 520-4. doi: 10.3109/17453670903278282. Dhanwal D K, Dennison E M, Harvey N C, Cooper C. Epidemiology of hip fracture: worldwide geographic variation. Indian J Orthop 2011; 45(1): 15-22. doi: 10.4103/0019-5413.73656. Dong X, Milholland B, Vijg J. Evidence for a limit to human lifespan. Nature 2016; 538(7624): 257-9. doi: 10.1038/nature19793. Fries J F. Aging, natural death, and the compression of morbidity. N Engl J Med 1980; 303(3): 130-5. doi: 10.1056/nejm198007173030304. Haentjens P, Magaziner J, Colón-Emeric C S, Vanderschueren D, Milisen K, Velkeniers B, Boonen S. Meta-analysis: excess mortality after hip fracture among older women and men. Ann Intern Med 2010; 152(6): 380-90. doi: 10.7326/0003-4819-152-6-201003160-00008. Helweg-Larsen K. The Danish Register of Causes of Death. Scand J Public Health 2011; 39(7 Suppl.): 26-9. doi: 10.1177/1403494811399958.

Acta Orthopaedica 2019; 90 (4): 342â&#x20AC;&#x201C;347

Karampampa K, Ahlbom A, Michaelsson K, Andersson T, Drefahl S, Modig K. Declining incidence trends for hip fractures have not been accompanied by improvements in lifetime risk or post-fracture survival: a nationwide study of the Swedish population 60 years and older. Bone 2015; 78: 55-61. doi: 10.1016/j.bone.2015.04.032. Kildemoes H W, Sorensen H T, Hallas J. The Danish National Prescription Registry. Scand J Public Health 2011; 39(7 Suppl.): 38-41. doi: 10.1177/1403494810394717. Madsen C M, Jantzen C, Lauritzen J B, Abrahamsen B, Jorgensen H L. Temporal trends in the use of antithrombotics at admission. Acta Orthop 2016; 87(4): 368-73. doi: 10.1080/17453674.2016.1195662. Mazzola P, Rea F, Merlino L, Bellelli G, Dubner L, Corrao G, Pasinetti G M, Annoni G. Hip fracture surgery and survival in centenarians. J Gerontol A Biol Sci Med Sci 2016; 71(11): 1514-8. doi: 10.1093/gerona/glw016. Morin S N, Lix L M, Majumdar S R, Leslie W D. Temporal trends in the incidence of osteoporotic fractures. Curr Osteoporos Rep 2013; 11(4): 263-9. doi: 10.1007/s11914-013-0168-x.

347

Ng M, Freeman M K, Fleming T D, Robinson M, Dwyer-Lindgren L, Thomson B, Wollum A, Sanman E, Wulf S, Lopez A D, Murray C J, Gakidou E. Smoking prevalence and cigarette consumption in 187 countries, 1980â&#x20AC;&#x201C;2012. JAMA 2014; 311(2): 183-92. doi: 10.1001/jama.2013. 284692. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi J C, Saunders L D, Beck C A, Feasby T E, Ghali W A. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005; 43(11): 1130-9. Robine J M, Cubaynes S. Worldwide demography of centenarians. Mech Ageing Dev 2017; 165(Pt B): 59-67. doi: 10.1016/j.mad.2017.03.004. Rodriguez-Molinero A, Yuste A, Banegas J R. High incidence of hip fracture in Spanish centenarians. J Am Geriatr Soc 2010; 58(2): 403-5. doi: 10.1111/j.1532-5415.2009.02706.x. Rosengren B E, Bjork J, Cooper C, Abrahamsen B. Recent hip fracture trends in Sweden and Denmark with age-period-cohort effects. Osteoporos Int 2017; 28(1): 139-49. doi: 10.1007/s00198-016-3768-3.

348

Acta Orthopaedica 2019; 90 (4): 348–353

Hip-fracture osteosynthesis training: exploring learning curves and setting proficiency standards Amandus GUSTAFSSON 1,2, Poul PEDERSEN 3, Troels Boldt RØMER 1, Bjarke VIBERG 4, Henrik PALM 5, and Lars KONGE 1,6 1 Copenhagen

Academy for Medical Education and Simulation; 2 Orthopedic Department, Slagelse Hospital, Region Zealand; 3 Orthopedic Department, Hvidovre Hospital; 4 Orthopedic Department, Kolding Hospital; 5 Orthopedic Department, University Hospital Bispebjerg; 6 Faculty of Health and Medical Sciences, University of Copenhagen, Denmark Correspondence: amandusgustafsson@gmail.com Submitted 2018-10-08. Accepted 2019-03-07.

Background and purpose — Orthopedic surgeons must be able to perform internal fixation of proximal femoral fractures early in their career, but inexperienced trainees prolong surgery and cause increased reoperation rates. Simulationbased virtual reality (VR) training has been proposed to overcome the initial steep part of the learning curve but it is unknown how much simulation training is necessary before trainees can progress to supervised surgery on patients. We determined characteristics of learning curves for novices and experts and a pass/fail mastery-learning standard for junior trainees was established. Methods — 38 first-year residents and 8 consultants specialized in orthopedic trauma surgery performed cannulated screws, Hansson pins, and sliding hip screw on the Swemac TraumaVision VR simulator. A previously validated test was used. The participants repeated the procedures until they reached their learning plateau. Results — The novices and the experts reached their learning plateau after an average of 169 minutes (95% CI 152–87) and 143 minutes (CI 109–177), respectively. Highest achieved scores were 92% (CI 91–93) for novices and 96% (CI 94–97) for experts. Plateau score, defined as the average of the 4 last scores, was 85% (CI 82–87) and 92% (CI 89–96) for the novices and the experts, respectively. Interpretation — Training time to reach plateau varied widely and it is paramount that simulation-based training continues to a predefined standard instead of ending after a fixed number of attempts or amount of time. A score of 92% comparable to the experts’ plateau score could be used as a mastery learning pass/fail standard.

The incidence of proximal femoral fractures (PFF) has been estimated at 0.1% in industrial countries (Dorotka et al. 2003). These patients take up 1.5% of total hospital capacity and constitute a large part of procedures undertaken in orthopedic departments (http://statbank.dk. Accessed 2018). Patients with PFF are on average > 80 years old and often have comorbidities (Roche et al. 2005), making a strong need for timely and definitive surgery. Internal fixation of hip fractures is a common procedure that orthopedic surgeons must master early in their career. Inexperienced trainees can contribute to prolonged length of surgery and higher rate of reoperation (Palm et al. 2007) and training on virtual-reality (VR) simulators has been proposed to reduce the burden of surgeons’ early learning curve on patients (Thomas 2013). Several studies describe different types of hip fracture VR simulators and their metrics’ ability to distinguish between novices and experienced surgeons (Tillander et al. 2004, Blyth et al. 2008, Mabrey et al. 2010, Froelich et al. 2011, Pedersen et al. 2014). However, there is very limited evidence on how to set up structured training programs or on setting credible pass/fail standards using PFF osteosynthesis VR simulators. Training for a certain amount of time or on a certain number of cases is a poor predictor for proficiency and will inevitably lead to trainees performing on variable levels after training. Hence, there is a move away from time-dependent learning and toward proficiency-based learning within medical education. It is therefore prudent to find a benchmark for proficiency (i.e., a pass/fail standard) in a simulated setting before the trainee performs actual surgery under supervi-

Acta Orthopaedica 2019; 90 (4): 348–353

Included novices n = 51

Included experts n=9

Excluded (n = 13): – more than 10 procedures, 7 – failure to train to plateau, 6 Participating novices n = 38

349

Excluded due to failure to train to plateau n=1 Participating experts n=8

Figure 1. Participants inclusions and exclusions.

sion (Stefanidis et al. 2012a, Goldberg et al. 2017). This will optimize resources and patient safety by ensuring that each individual trainee spends exactly the required amount of time training on the VR simulator. Earlier attempts to establish proficiency-based criteria have applied the test a few times or sometimes only once. Such criteria based on data early in the novices’ learning curves and on experts not familiar with the simulator do not add much to the validity argument (Cook 2015). When attempting to support a mastery learning proficiency-based criterion it is more prudent to address the level where participants perform consistently. This is the final, autonomous stage of learning motor skills that is the trademark of experts (Magill 2010). One way to do this is to explore learning curves. Individual trainees have their own learning curve that is typically negatively accelerated, i.e., performance improves considerably in the initial part before entering the plateau phase where additional improvement requires a lot of training/repetitions (Madsen et al. 2014). Another credible way to determine a performance standard is to assess the performance of experts on the same simulator metrics as the trainees (Dyre et al. 2015, Thinggaard et al. 2016) We determined the characteristics of learning curves for novices and familiarization curves of experts to establish a credible pass/fail mastery-learning standard for junior trainees.

Methods The study was conducted from May 2015 to July 2017 at Copenhagen Academy for Medical Education and Simulation, Copenhagen University Hospital Rigshospitalet (Konge et al. 2015a). All training was done during the simulation center’s opening hours and participation was voluntary. 51 novices in their 1st year of specialization were included from 7 different departments. 7 novices were excluded as prior to training they had performed more than 10 osteosyntheses of proximal femur fractures under supervision. 6 novices were subsequently excluded as they discontinued training before reaching plateau. 9 experts from 3 different departments who were all consultants with specialization in orthopedic trauma surgery were included. 1 expert was excluded because of failure to test to plateau (Figure 1). We do not know the reasons as to why some participants discontinuing training before plateau in either group.

Figure 2. TraumaVision during training.

We used previously validated software on the Swemac TraumaVision simulator (STV; Swemac Osmedic ApS, Nivå, Denmark) to explore learning curves of orthopedic surgeons (Pedersen et al. 2014). The STV simulator consists of a computer with 2 screens and software TraumaVision 5.12. A force feedback device (Phantom Omni; Delft Haptics Lab, Delft, the Netherlands) that mimics the surgery tools and generates haptic feedback is connected to the computer. Either the right or the left hand, according to the preference of the user, can handle the device. The movements are visualized on one of the screens. The fluoroscopy is administered by a foot-paddle and can be displayed in either a standard A-P or lateral view on the other screen (Figure 2). The software contains a variety of orthopedic procedures and the 3 used in our program were cannulated screws, Hansson hook-pins, and a sliding hip screw. The individual score on each procedure is a percentage of maximum of metrics deemed clinically relevant by the manufacturer and supported by validity evidence (Pedersen et al. 2014). A combined score was produced as a mean of the individual scores. The individual metrics, individual score, and combined score were used to give feedback after completion of the 3 procedures. Only the combined score was used for data analysis as the scores of the individual parameters have insufficient validity (Pedersen et al. 2014). All participants were naive to the simulations. They were introduced to the simulator and instructed in the correct operation technique for the 3 procedures prior to training. An orthopedic surgeon experienced in hip fracture surgery conducted the introduction and presentation. After completing the introduction, the participants completed a warm-up session containing the 3 procedures. Subsequently the participants trained in 2-hour sessions and received simulator feedback after finishing each round of procedures. The participants were not allowed to train for more than 2 hours per day due to risk of fatigue causing reduced learning (Andersen et al. 2015). A trained simulator assistant oversaw all training and helped the

350

Acta Orthopaedica 2019; 90 (4): 348–353

Total score

Plateau score

100

Distribution of plateau scores Novices Experts

60 Novices Experts

Test number

Figure 3. Learning curves for the first 10 attempts of novices and experts.

Novices Experts 50

9 10 11 12 13 14 15 18

Test number

Figure 4. Plateau scores of novices and experts illustrating the large variation in attempts needed to train to plateau. Line at 88% illustrates the consequences of pass/fail standard of contrasting groups method with many novices plateauing well below and above.

participants interpret the simulator feedback if needed. The training stopped when the participant reached the plateau phase indicated by 3 consecutive attempts without improvement. The plateau criterion was based on a compromise to reduce risk of participant dropout due to prolonged training, while at the same time having an estimated low risk of plateau underestimation. The participants were aware they were training to plateau, but unaware of the plateau criterion. Statistics Levine’s tests were performed and independent samples Student’s t-tests with equal variances assumed/not assumed as appropriate were used to (1) compare the performance of the novice and expert group for variables with normal distribution and (2) to compare continuous data for variables for the novice group. Either Pearson’s chi-square or Fisher’s exact test was used as appropriate to compare categorical data for variables for the novice group. For comparison of performance of the novice and expert group for variables with nonnormal distribution, bootstrapped independent samples t-test was used. 95% prediction intervals (PI) for the novices’ training time, best score, and plateau score was calculated using linear regression analysis adjusting for age, sex, dominance, performed procedures, span of orthopedic employment, and previous simulation-based training. The plateau score was defined as the average of the participant’s last 4 scores. The mean plateau score distribution of the 2 groups was plotted using the contrasting groups method (Downing and Yudkowsky 2009). The intersection between the 2 groups was set as a pass/fail standard and the consequences of the pass/fail standard in comparison with the pass/fail mastery criterion were explored. The statistical analysis was performed using SPSS version 22 (IBM Corp, Armonk, NY, USA). Differences in metrics were considered statistically significant when the p-value was < 0.05. 95% confidence intervals (CI) are used.

100

110

Score

Figure 5. Distribution of plateau scores for novices (red) and experts (black). Using the contrasting groups method, a pass/fail standard for the test can be determined from the intersection of distributions (88%).

Ethics, funding, and potential conflicts of interests Ethical approval was obtained prior to commencement of the study from the Regional Ethical Committee of the Capital Region in the form of an exempt letter (21/11/2014, No. H-42014-FSP). The participants gave informed consent and could opt to drop out at any time. There was no external funding for the study. None of the authors have any competing interests to declare.

Results For novices the median performed osteosynthesis of proximal femoral fracture was 1 (0–10), employment at an orthopedic department was 7 months (0–22), and age 29 (26–54). The median number of years working full time as orthopedic traumatologist after specialization for the experts was 4 (3–15). The novices had hands-on simulation training for an average of 169 minutes (CI 152–187, PI 162–177) to achieve their learning plateau while experts tested on average for 143 minutes (CI 109–177). The highest achieved scores were 92% (CI 91–93, PI 91–93) and 96% (CI 94–97) for the novices and the experts, respectively (Figure 3). The plateau scores were 85% (CI 82–87, PI 84–86) and 92% (CI 89–96) for the novices and the experts, respectively (Figure 4). When examining demographic and previous experience of the novices who failed to score within 1 standard deviation of the experts’ plateau scores compared with novices with more than 1 SD, no statistically significant difference with regard to age (p = 0.1), sex (p = 1.0), dominance (p = 0.7), performed procedures (p = 0.2), span of orthopedic employment (p = 0.7), or previous exposure to simulation-based training (p = 0.5) was found (Table). A pass/fail standard for the plateau score was defined as 88% using a contrasting groups method (Figure 5). A pass/fail mastery criterion was defined as the experts’ mean plateau score of 92%.

Acta Orthopaedica 2019; 90 (4): 348–353

351

Demographics and previous experience for the novices grouped by score within and more than 1 SD of experts’ plateau score n Mean age (CI) Within 1 SD 13 More than 1 SD 25 p-value

27.9 (27.2–28.6) 30.4 (28.2–32.6) 0.1

Male/ Dominant Previous simulation female hand (R/L) Mean OE (CI) Mean SP (CI) (yes/no) 9/4 16/9 1.0

10/3 21/4 0.7

6.8 (4.2–9.4) 2.3 (0.9–3.8) 5.5 (6.7–11.1) 2.7 (1.4–4.0) 0.7 0.2

7/6 10/15 0.5

R/L = right/left. OE = orthopedic employment. SP = supervised procedures. SD = standard deviation.

Discussion This study demonstrates that the time needed to train to a plateau of consistent performance is highly variable (Figure 4). This makes it essential that a simulation-based training program for novices is not based on time spent training or numbers of repetitions as this inherently will create a substantial risk of either premature termination of training or training with little or no improvement for a high proportion of the trainees. Training must be continued until a predefined criterion based on solid evidence is reached. To our knowledge there are no former studies setting credible mastery-learning pass/fail standards that ensure basic proficiency in simulated hip fracture surgery. We found that experts performed better both measured by highest obtained score and by plateau score. As expected, experts also had a higher score on the 1st attempt and hence their familiarization curves have a high onset and are relatively flat, improving only slightly with increasing number of repetitions. The novices on the other hand produced a steep learning curve followed by a curve comparable to the experts with more slight progression, but well below the expert curve (Figure 3). The ability of the simulation test to discriminate between novices training to their learning plateau and experts testing to a familiarization plateau widely amplify previously established (Pedersen et al. 2014) validity evidence of the test. It was unexpected that the experts spent similar time to the novices to reach plateau. This finding indicates that the experts need quite some time to get accustomed to features of the simulator that do not resemble their daily clinical life. Change of direction is not allowed when the lateral cortex of the femur is penetrated by the K-wire, making it necessary to retract and reintroduce the wire in the correct trajectory. This feature can be advantageous for motor learning of the novice but produced challenges to the experts who were all accustomed to penetrating the lateral cortex at the correct entry point and then changing trajectory as needed. Another challenge was the distance from the tip of the K-wire to the femoral head joint. The simulator parameter has an acceptable distance of 1–3 mm to reduce the risk of K-wire pullout when the cannulated drill is retracted. Many of the experts were used to inserting the K-wire with a larger distance to the

joint surface. As the procedure is autonomous for the experts, they exerted some effort to change strategy to comply with the simulation and parameters. When examining a passing competence criterion, Pedersen et al. (2014) found a pass/fail standard of 58% using a contrasting groups method, but suggested a score of 75%, based on data from a single performance on the simulator. In our study all novices but 1 achieved a maximum score above 75% and all but 4 achieved plateau scores above this level, indicating that a pass/fail standard for proficiency must be higher to exploit the maximum training effect of the VR simulator in a mastery-learning program. The contrasting groups standardsetting method sets a cut-off where the combination of passing novices and failing experts is at its lowest and is a commonly used method to set a standard for a test and a pass/fail standard for proficiency before supervised practice on patients (Jørgensen et al. 2018, Russell et al. 2018). We used the participants’ plateau scores indicating consistent maximal obtainable scores for the participants to calculate the pass/fail score with this method. The method indicated a score of 88%. However, from Figure 3 it is apparent that a sizeable proportion of the novices achieve plateau scores well below 88% after a few (4–7) attempts. Though it cannot be ruled out that this is due to lack of ability, we would argue that it is likely to be an example of arrested development as described by Ericsson (2009). He argues that most learners, after achieving a standard of performance that can be elicited with reduced concentration, no longer attempt further improvement and development will be prematurely arrested. Our novices trained without the motivation to achieve a predetermined pass score and their plateau score might not represent their best obtainable score. When no standard is set the trainees must rely on their own selfassessment that can be poor in skills training. Andersen et al. (2017) studied novices doing VR simulation-based selfdirected training in mastoidectomy and found that training was terminated well before a set time limit when additional time would have permitted better performance. Their learning curve plateaued (too) early as seen for a large subgroup in our study. Likewise, Jowett et al. (2007) demonstrated no superior skills retention with further simulation-based training in knottying after the trainees had reached self-assessed proficiency and propose the explanation could be arrested development

352

due to lack of intrinsic motivation. Hence, it can be precarious to base proficiency on data from self-assessed training of novices. Yudkowsky et al. (2015) argued that standards derived from other novices have no place in a proficiency-based curriculum or mastery setting and emphasize overlearning and automaticity as conveyors of long-term retention. In accordance, Stefanidis et al. (2012b) have shown that novices training above and beyond standard proficiency levels set by expert performance have superior skill acquisition and transfer compared with training to proficiency alone. Madsen et al. (2014) explored consequences of performance standards based on the contrasting groups method and experts’ levels on a transvaginal ultrasound simulator test and found that the novices were readily able to train to expert-level scores and that using lower standards should not be recommended. In our opinion the raison d’être for simulation-based training is promoting mastery learning through deliberate practice and we therefore suggest a pass/fail plateau score of 92%, comparable to the average for experts before trainees progress to supervised practice in the clinical setting. The simulator allows for anatomical variation with 4 skeletons integrated in the software and allows for training on suboptimally reduced fractures. A limitation to our test is that it was chosen to base the training on exclusively anatomical reduced fractures on one skeleton’s left side and hence without variation of anatomy. This could lead to higher scores and faster plateauing for the novices compared with the experts, as it can be argued that it is technically more challenging to place the implants optimally in less than perfectly reduced fractures with anatomical variation. This supports the validity argument of the test, but the competency criterion proposed cannot necessarily be transferred to a training setup where variation is introduced. This can be desirable, as variation during training has been shown to enhance learning outcomes in simulation-based training (Zendejas et al. 2013). Another important implication to consider when interpreting consequences of the competency criterion is the inherent limitation of data based solely on simulation. Even though studies on other simulators and settings have explored consequences of the mastery criterion as a pass/fail standard and found it feasible, it is not axiomatically so in the present setting. The expert group sample is small and may represent a level of skill on the simulator that is unobtainable for some novices regardless of interventions to improve learning during training. The impact can be undesirable training that has no improved effect in the clinical setting—the intrinsic factor of application of simulation-based training. To that end, the present mastery criterion is suggestive as transference of skills to the clinical setting and optimal training for this effect is still unknown. Skills obtained from VR training have been shown to be transferable to the clinical setting within many medical specialties (Konge et al. 2015b, Tolsgaard et al. 2017, Thomsen et al. 2017). However, to our knowledge, only Howells et al.

Acta Orthopaedica 2019; 90 (4): 348–353

(2008), on a bench-top arthroscopic knee model, and Cannon et al. (2014) on VR knee arthroscopy simulator have shown transfer of skills to the operating theatre within the field of orthopedic surgery. It is essential that future research within simulation-based training focus on transfer of skills to the clinical setting to optimize training and gauge the consequences of competence criteria of simulation-based training. In summary, this study found that the time training to plateau displayed a high degree of variability. Experts achieved higher scores through all phases of the learning curve compared with novices, supporting enhanced validity evidence of the test, and we suggest a credible pass/fail score of 92% as an average of 4 consecutive attempts before novices proceed to supervised practice on patients. It is important in future research to address the transferability of skills obtained from this simulator to clinical practice and the consequences of passing criterions. AG: study design, data collection, statistical analysis, data analysis and interpretation, writing of the manuscript. PP: study design and revision of the manuscript. TBR: data collection and revision of the manuscript. BV: study design and revision of the manuscript. HP: study design and revision of the manuscript. LK: study design, statistical analysis, data analysis and interpretation, revision of the manuscript. Acta thanks Katre Maasalu and Sari Ponzer for help with peer review of this study.

Andersen S A, Konge L, Cayé-Thomasen P, Sørensen M S. Learning curves of virtual mastoidectomy in distributed and massed practice. JAMA Otolaryngol Head Neck Surg 2015; 141(10): 913-18. Andersen S A, Konge L, Mikkelsen P T, Cayé-Thomasen P, Sørensen M S. Mapping the plateau of novices in virtual reality simulation training of mastoidectomy. Laryngoscope 2017; 127: 907-14. Blyth P, Stott N S, Anderson I A. Virtual reality assessment of technical skill using the Bonedoc DHS simulator. Injury 2008; 39(10): 1127–33. Cannon W D, Garrett Jr W E, Hunter R E, Sweeney H J, Eckhoff D G, Nicandri G T, Hutchinson M R, Johnson D D, Bisson L J, Bedi A, Hill J A, Koh J L, Reinig K D. Improving residency training in arthroscopic knee surgery with use of a virtual-reality simulator. J Bone Joint Surg Am 2014; 96: 1798-806. Cook D A. Much ado about differences: why expert–novice comparisons add little to the validity argument. Adv Health Sci Educ 2015; 20: 829-34. Danish National Council on Statistics: Statbank. Available at: http://statbank. dk. Accessed 2018. Dorotka R, Schoechtner H, Buchinger W. The influence of immediate surgical treatment of proximal femoral fractures on mortality and quality of life: operation within six hours of the fracture versus later than six hours. J Bone Joint Surg Br 2003; 85: 1107-13. Downing S M, Yudkowsky R. Standard setting. In: Downing S M, Yudkowsky R, editors. Assessment in health professions education. New York: Routledge; 2009. p. 119-48. Dyre L, Nørgaard L N, Tabor A, Madsen M E, Sørensen J L, Ringsted C, Tolsgaard M. Collecting validity evidence for the assessment of mastery learning in simulation-based ultrasound training. Ultraschall in Med 2015; 36: 1-7. Ericsson K A. Enhancing the development of professional performance: implications from the study of deliberate practice. In: Ericsson K A, editor. Development of professional expertise. Cambridge: Cambridge University Press; 2009. pp. 405-31.

Acta Orthopaedica 2019; 90 (4): 348–353

Froelich J M, Milbrandt JC, Novicoff W M, Saleh K J, Allan D G. Surgical simulators and hip fractures: a role in residency training? J Surg 2011; 68(4): 298-302. Goldberg M G, Garbens A, Szasz P, Hauer T, Grantcharov T P. Systematic review to establish absolute standards for technical performance in surgery. Br J Surg 2017; 104: 13-21. Howells N R, Gill H S, Carr A J, Price A J, Rees J L. Transferring simulated arthroscopic skills to the operating theatre. J Bone Joint Surg Br 2008; 90-B: 494-9. Jowett N, LeBlanc V, Xeroulis G, MacRae H, Dubrowski A. Surgical skill acquisition with self-directed practice using computer-based video training. Am J Surg 2007; 193: 237-42. Jørgensen M, Konge L, Subhu Y. Contrasting groups’ standard setting for consequences analysis in validity studies: reporting considerations. Adv Simul (Lond) 2018; 3: 5. Konge L, Ringsted C, Bjerrum F, Tolsgaard M G, Bitsch M, Sørensen J L, Schroeder T V. The Simulation Centre at Rigshospitalet, Copenhagen, Denmark. Jour Surg Educ 2015a; 72(2): 362-5. Konge L, Clementsen P F, Ringsted C, Minddal V, Larsen K R, Annema JT. Simulator training for endobronchial ultrasound: a randomised controlled trial. Eur Respir J 2015b; 46(4): 1140-9. Mabrey J D, Reinig K D, Cannon W D. Virtual reality in orthopaedics: is it a reality? Clin Orthop Relat Res 2010; 468(10): 2586–91. Madsen M E, Konge L, Nørgaard L N, Tabor A, Ringsted C, Klemmensen Å K, Ottesen B, Tolsgaard M G. Assessment of performance measures and learning curves for use of a virtual-reality ultrasound simulator in transvaginal ultrasound examination. Ultrasound Obstet Gynecol 2014; 44: 693-9. Magill R A. The stages of learning. In: Motor learning and control. New York: McGraw-Hill; 2010. pp. 265-88. Palm H, Jacobsen S, Krasheninnikoff M, Foss, N B, Kehlet H, Gebuhr P. Influence of surgeon’s experience and supervision on re-operation rate after hip fracture surgery. Injury 2007; 38: 775-9. Pedersen P, Palm H, Ringsted C, Konge L. Virtual-reality simulation to assess performance in hip fracture surgery. Acta Orthop 2014; 85(4): 403-7.

353

Roche J J W, Wenn R T, Sahota O, Moran C G. Effect of comorbidities and postoperative complications on mortality after hip fracture in elderly people: prospective observational cohort study. BMJ 2005; 331: 1374. Russell L, Østergaard M L, Nielsen M B, Konge L, Nielsen K R. Standardised assessment of competence in Focused Assessment with Sonography for Trauma. Acta Anaesthesiol Scand 2018; 62(8): 1154-60. Stefanidis D, Arora S, Parrack D M, Hamad G G, Capella J, Grantcharov T, et al. Research priorities in surgical simulation for the 21st century. Am J Surg 2012a; 203(1): 49–53. Stefanidis D, Scerbo M W, Montero P N, Acker C E, Smith W D. Simulator training to automaticity leads to improved skill transfer compared with traditional proficiency-based training. Ann Surg 2012b; 255: 30-37. Thinggaard E, Bjerrum F, Standbygaard J, Gögenur Ismail, Konge L. Ensuring competency of novice laparoscopic surgeons: exploring standard setting methods and their consequences. J Surg Educ 2016; 73(6): 986-91. Thomas M P. The role of simulation in the development of technical competence during surgical training: a literature review. Int J Med Educ 2013; 4: 48-58. Thomsen A S, Bach-Holm D, Kjærbo H, Højgaard-Olsen K, Subhi Y, Saleh G M, Park YS, la Cour M, Konge L. Operating room performance improves after proficiency-based virtual reality cataract surgery training. Ophthalmology 2017; 124(4): 524-31. Tillander B, Ledin T, Nordqvist P, Skarman E, Wahlström O. A virtual reality trauma simulator. Med Teach 2004; 26(2): 189-91. Tolsgaard M G, Ringsted C, Rosthøj S, Nørgaard L, Møller L, Freiesleben N C, Dyre L, Tabor A. The effects of simulation-based transvaginal ultrasound training on quality and efficiency of care: a multicenter single-blind randomized trial. Ann Surg 2017; 265(3): 630-7. Yudkowsky R, Park Y S, Lineberry M, Knox A, Ritter EM. Setting mastery learning standards. Acad Med 2015; 90: 1495-1500. Zendejas B, Brydges R, Wang T, Cook D A. Patient outcomes in simulationbased medical education: a systematic review. J Gen Intern Med 2013; 28(8): 1078-89.

354

Acta Orthopaedica 2019; 90 (4): 354–359

Medial unicompartmental knee arthroplasty: increasingly uniform patient demographics despite differences in surgical volume and usage—a descriptive study of 8,501 cases from the Danish Knee Arthroplasty Registry Cecilie HENKEL 1, Mette MIKKELSEN 1, Alma B PEDERSEN 2, Lasse E RASMUSSEN 3, Kirill GROMOV 1, Andrew PRICE 4, and Anders TROELSEN 1 1 Department of Orthopaedic Surgery, Clinical Orthopaedic Research Hvidovre (CORH), Copenhagen University Hospital Hvidovre, Denmark; 2 Department of Clinical Epidemiology, Aarhus University Hospital, Denmark; 3 Department of Orthopaedic Surgery, Vejle Hospital, Denmark; 4 Nuffield Department of Orthopaedic Surgery, Rheumatology and Musculoskeletal Science, Nuffield Orthopaedic Centre, University of Oxford,

Correspondence: cehenkel@gmail.com Submitted 2018-11-30. Accepted 2019-03-04.

Background and purpose — Using contemporary indications, up to 50% of patients undergoing knee arthroplasty are eligible for unicompartmental knee arthroplasty (UKA), and lower UKA use likely reflects a restrictive approach to patient selection. Since broader indications have been successfully introduced, and low surgical volume and UKA percentage (usage) are associated with higher revision rates, it is of interest whether the actual use of UKA has changed accordingly. We explored this by assessing time trends in patient demographics and whether these are associated with center UKA volume and usage. Patients and methods — From the Danish Knee Arthroplasty Registry, we included 8,501 medial UKAs performed for primary osteoarthritis during 2002–2016. Using locally weighted regression, we examined changes—both overall and by center volume and usage (low vs high)—in sex distribution, age, weight, and preoperative American Knee Society Score (AKSS-O). Results — Over the last 20 years, UKA use in Denmark has been increasing steadily. Age, weight, and proportion of men all increased regardless of volume and usage. AKSS-O showed an initial increase followed by a decrease. In lowusage and low-volume centers, the proportion of women was higher, patients were younger, weighed less, and had higher AKSS-O scores; however, for age and AKSS-O, the groups were converging during the last part of the period. Interpretation — Characteristics of UKA patients have changed in the last 15 years irrespective of center volume and usage. We found between-group differences for both volume and usage, though with convergence for age and AKSS-O, which suggests an increasingly uniform approach to patient selection.

Unicompartmental knee arthroplasty (UKA) is a viable alternative to total knee arthroplasty (TKA) for patients with pronounced, isolated anteromedial osteoarthritis of the knee, resulting in lower mortality and morbidity (Liddle et al. 2014) as well as better patient-reported outcomes (Liddle et al. 2015a, Burn et al. 2018). However, revision rates are higher for UKA than for TKA (Liddle et al. 2014, Chawla et al. 2017). Kozinn and Scott (1989) proposed a set of strict contraindications for UKA, leaving just 6% of patients eligible for the procedure (Stern et al. 1993). These have now proven unnecessary (Pandit et al. 2011, Hamilton et al. 2017a), and broadening the indications increases the proportion of patients eligible for UKA to as much as 50% (Willis-Owen et al. 2009). Historically, an increase in UKA use was considered to be problematic as the revision burden was assumed to grow. However, low surgical UKA volume and usage—defined as the total annual number of UKAs, and the percentage of all primary knee arthroplasties that are UKA, respectively—have both been associated with higher revision rates (Liddle et al. 2015b, Liddle et al. 2016, Badawy et al. 2017). Interestingly, Hamilton et al. (2017b) found that the positive effect of high UKA usage was independent of UKA volume. Since higher usage is obtained through more liberal patient selection, these findings suggest that a restrictive approach to patient selection is unnecessary. Therefore, it is of interest whether the practice regarding UKA has changed in accordance with the shift towards wider indications. To investigate this, we explored time trends in UKA use and patient demographics among UKA patients registered in the Danish Knee Arthroplasty Registry (DKR). In addition, we assessed whether they are associated with center UKA volume and usage.

Acta Orthopaedica 2019; 90 (4): 354–359

Patients and methods We conducted a descriptive study using data from the Danish Knee Arthroplasty Registry (DKR), which has been collecting pre- and perioperative data on knee arthroplasty procedures performed in Denmark since 1 January 1997 (Pedersen et al. 2012). Study population Our data extraction was done on 4 December 2017 and included all primary medial UKAs reported until this date. Due to possible confounding, only medial UKAs performed for primary osteoarthritis were included in the evaluation of patient characteristics. Procedures from 2017 were excluded in assessment of volume and usage as data were incomplete for this year, which hindered the calculation of meaningful values. Due to a low number of procedures (< 100 annually) during the first 5 years, procedures done in the period 1997– 2001 were excluded as well. Patient demographics To describe changes in patient demographics, we included information on sex, age, weight, and preoperative American Knee Society Score (AKSS-O) (Insall et al. 1989). Weight values < 45 kg and > 200 kg were considered registration errors and were treated as missing values. Volume and usage We defined center volume as the total number of medial UKAs performed at the center in 1 calendar year. Likewise, we defined usage as the percentage of all primary knee arthroplasty procedures at the center that were medial UKAs in the given calendar year. Volume and usage were assessed independently for each year, making it possible for centers to change groups from year to year. Upon calculation of center volume and usage, patients were assigned the given values of the center responsible for their procedure, and all subsequent calculations were done at the patient level. For investigation of volume and usage, both were divided into 2 groups: low and high. For volume, we based the grouping on the distribution of our data, as is often done (Robertsson et al. 2001, Lau et al. 2012, Baker et al. 2013, Badawy et al. 2017). We made the cut at the median value to obtain 2 groups of approximately equal size, thus making centers performing < 52 UKAs annually low volume and centers performing ≥ 52 UKAs annually high volume. For usage, we based our categorization on the works of Liddle et al. (2015b) and Hamilton et al. (2017b), which show that a usage of ≥ 20% yields acceptable results. Hence, centers with a usage of < 20% were categorized as low usage and centers with a usage of ≥ 20% were categorized as high usage regardless of their volume.

355

Records from the Danish Knee Arthroplasty Registry between January 1, 1997 and December 4, 2017 n = 129,253 Excluded (n = 120,752): – revisions, 10,186 – unknown surgery date, 343 – other indications than primary osteoarthritis, 18,912 – not medial unicompartmental knee arthroplasty, 89,867 – medial UKA before January 1, 2002, 216 – medial UKA after December 31, 2016, 1,228 Medial UKAs performed between January 1, 2002 and December 31, 2016 n = 8,501 (n = 7,460 patients)

Figure 1. Selection of the final study population.

Statistics Unless otherwise specified, reported values are mean (SD). In addressing missing values, we chose to omit the patients for the variable(s) at issue but included them for all other investigations. All graphical explorations relating to patient characteristics were performed using locally weighted regression (Cleveland 1979). For each of the 4 variables, both an overall and separate locally estimated scatterplot smoothing (LOESS) curves for the 4 volume and usage groups were fitted. With the exception of weight, all variables were normally distributed. In a subgroup assessment, we excluded all bilateral procedures and repeated the LOESS curves. Calculations and graphs were made in R (version 3.3.2, the R Foundation for Statistical Computing, Vienna, Austria). Ethics, funding, and potential conflicts of interest The study was approved by the Danish Data Protection Agency (J No 2012-58-0004). No external funding was received, and the authors declare no conflicts of interest.

Results Selection of the final study population is mapped out in Figure 1 and patient characteristics are summarized in Table 1. Data completion was high, with a complete set for sex and age, 310 (3.1%) missing values for weight, and 178 (1.8%) missing values for AKSS-O. The use of UKA in Denmark has been steadily increasing since 1997 (Figure 2). The median center volume was 52 (1–234) and the median center usage 22% (0.2–100), both with time trends analogous to the overall increase in use. In 1997, 10 centers reported UKA procedures, which increased steadily to 17 in 2002 and further to 25 centers in 2006 (Table 2). From 2006 to 2008, there was a rapid increase to 35 centers, which was matched by an increase of 10 low-volume centers, but only by 2 additional low-usage centers. From

356

Table 1. Characteristics of the final study population. Values are mean (standard deviation) unless otherwise specified Number of patients 8,501 Low volume 4,097 High volume 4,404 Low usage 3,313 High usage 5,188 Number of women a 4,697 (55) Low volume 2,274 (56) High volume 2,423 (55) Low usage 1,885 (57) High usage 2,812 (54) Age (years) 65.0 (9.4) Low volume 64.2 (9.6) High volume 65.8 (9.2) Low usage 64.5 (9.6) High usage 65.3 (9.3) Weight (kg) b 82 (45–190) Low volume 82 (45–188) High volume 83 (45–190) Low usage 82 (45–188) High usage 83 (45–190) Knee score (AKSS-O) 42 (14) Low volume 44 (15) High volume 40 (13) Low usage 43 (15) High usage 41 (14) Overall values and grouped by center volume (low < 52 per year, high ≥ 52 per year) and usage (low < 20%, high ≥ 20%). a Frequency (%) b Median (range)

Acta Orthopaedica 2019; 90 (4): 354–359

Table 2. Number of centers (C) performing medial unicompartmental arthroplasties (UKAs) each year and the total number of patients (P) a Low volume High volume Low usage High usage Total C P C P C P C P C P 1997 10 23 0 0 10 23 0 0 10 23 1998 15 32 0 0 14 23 1 9 15 32 1999 11 41 0 0 9 17 2 24 11 41 2000 13 81 0 0 11 57 2 24 13 81 2001 13 87 0 0 12 39 1 48 13 87 2002 16 132 1 65 13 48 4 149 17 197 2003 20 188 2 125 18 150 4 163 22 313 2004 23 283 1 85 19 163 5 205 24 368 2005 26 261 2 132 24 194 4 199 28 393 2006 22 245 3 209 20 233 5 221 25 454 2007 27 348 3 187 22 317 8 218 30 535 2008 32 469 3 178 22 269 13 378 35 647 2009 29 344 5 394 22 367 12 371 34 738 2010 31 405 5 411 25 481 11 335 36 816 2011 25 217 7 612 21 315 11 514 32 829 2012 24 268 5 468 19 240 10 496 29 736 2013 18 262 4 476 15 278 7 460 22 738 2014 20 258 5 556 15 346 10 468 25 814 2015 17 289 5 790 12 321 10 758 22 1,079 2016 17 263 8 1,054 15 373 10 944 25 1,317 a Both grouped by center volume (low < 52 per year, high ≥ 52 per year) and usage (low < 20%, high ≥ 20%).

Medial UKA proportion (%) 20

2008 to 2016, the total number of reporting centers decreased to 25, primarily caused by a decrease in low-volume and lowusage centers. The 8,501 procedures in 2002–2016 were performed in 7,460 patients, and subgroup assessment with exclusion of bilateral procedures showed similar distribution of the demographic variables. The most common implant type was the mobile-bearing Oxford knee (91%; n = 7,693), followed by the fixed-bearing implants ZUK (2%; n = 194) and Link (2%; n = 154). Patient characteristics Sex The proportion of females has been steadily decreasing from 67% (n = 115) in 2002 to 55% (n = 622) in 2016. The same pattern was seen in both low- and high-volume centers. Lowvolume centers had a lower proportion of females in the years 2002–2007 and 2013–2016 but a higher proportion in 2008– 2012. Both usage groups also showed an overall decreasing trend, and low-usage centers had a higher proportion of females throughout the study period. Age The age has been increasing, from 64 years (9.6) in 2002 to 66 years (9.2) in 2016. Both volume groups shared this tendency, but patients from low-volume centers were generally younger.

Primary knee arthroplasties

9,000 Medial UKA proportion (%) Annual number of primary knee arthroplasties

8,000 7,000

16 6,000

14 12

5,000

4,000

3,000

6 2,000

1,000

2 0

1997 1999 2001 2003 2005 2007 2009 2011 2013 2015 2017

Figure 2. Annual use of medial UKAs. The national percentage of all primary knee arthroplasties accounted for by medial unicompartmental arthroplasties (UKAs) each year and the annual numbers of all primary arthroplasties. The numbers account for all registered knee arthroplasty procedures except revisions. Note that the dataset contains only procedures up to December 4, 2017, which is why the numbers for 2017 should be interpreted with caution.

In 2007–2011 the low-volume centers expressed a temporary decrease, which resulted in a more pronounced difference between the groups. The decrease was followed by a substantial increase, eliminating the difference between the groups by 2016. The usage groups showed a similar pattern, though less pronounced (Figure 3). In 2016 the groups seem to have switched, presenting a higher age for the low-usage group.

Acta Orthopaedica 2019; 90 (4): 354–359

Age

Volume low high

357

Usage low high

scores for high-usage centers started increasing while the scores for low-usage centers continued to decrease (Figure 4).

Discussion

In our national registry-based study we found that the use of UKA has increased substantially in the past 20 62 62 years accompanied by a change in patient characteristics. Patient characteristics differed with center volume 60 60 and usage, though recent converging trends were noted. 2004 2006 2008 2010 2012 2014 2016 2004 2006 2008 2010 2012 2014 2016 Year A B Year The total number of medial UKAs has increased Figure 3. Age. LOESS curves (with 95% confidence intervals) depicting time steadily since 1997 but most markedly in 2015– trends in the age of UKA patients in: 2016. The total number of TKAs has also increased, A: centers with low (< 52 per year) and high (≥ 52 per year) volume. though with a notable drop in 2015–2016. LookB: centers with low (< 20%) and high (≥ 20%) usage. ing at the annual sums of medial UKAs and TKAs during 2011–2016, they have generally been stable in Knee score (AKSS–O) Knee score (AKSS–O) this period. Therefore, the drop in number of TKAs 50 50 Usage Volume in 2015–2016 may be explained by the simultaneous low low high high increase in medial UKAs, thus resulting in the consid45 45 erable increase in UKA percentage during the last few years. The recent stagnation in the total annual number 40 40 of knee arthroplasties in Denmark might be in contrast to other countries, e.g. Sweden where the number of 35 35 procedures is still increasing (SKAR 2018). However, in Denmark as well as in Sweden, the incidence of knee arthroplasty was increasing until levelling off in 2009. 30 30 In 2011–2016, the incidences were stable, in Denmark 2004 2006 2008 2010 2012 2014 2016 2004 2006 2008 2010 2012 2014 2016 at an average of 147/100,000 inhabitants (DKR 2017) Year Year A B and in Sweden 137/100,000 inhabitants (SKAR 2018). Figure 4. Knee score (AKSS-O). LOESS curves (with 95% confidence intervals) depicting time trends in preoperative knee score (AKSS-O) for UKA patients in: Hence, the different trends in annual number of procedures between the countries do not necessarily reflect A: centers with low (< 52 per year) and high (≥ 52 per year) volume. B: centers with low (< 20%) and high (≥ 20%) usage. differences related directly to arthroplasty surgery. External factors such as population growth rate might also play a part, and hence, direct comparison between the Weight The weight has generally shown an increasing trend, from a countries contains possible pitfalls. Regarding stagnation of median weight of 82 kg (47–150) in 2002 to 85 kg (47–186) the number of procedures in Denmark, this followed a 4-fold in 2016. Though the males were generally heavier, there was increase from 2000 to 2009. This increase was possibly facilino marked difference in the overall tendencies between the tated by the introduction of fast-track programs, resulting in sexes. Both volume groups also shared the overall increas- shorter hospital stays and thereby increased capacity. In the ing tendency. The low-volume group had a lower weight in later years, an increasing focus on nonoperative treatment, 2005–2015 but higher both before and after this period. For e.g., the nationwide implementation of the initiative Good usage, both groups expressed the overall increasing tendency Life with osteoArthritis in Denmark (GLA:D) (Skou and Roos as well, but with lower weights in patients from low-usage 2017), may have led to postponement of surgery. Indicative of this is the peak in number of knee arthroplasties in 2010, coincenters during the entire period. cident with the lowest mean age in the history of the registry (DKR 2017), which was followed by an increasing mean age Knee score (AKSS-O) The general trend in knee score was bell-shaped with an ini- concurrent with the stabilization in the number of procedures tial increase from 39 (15) in 2002 to its peak at 44 (16) in (DKR 2017). Altogether, the recent stagnation in number of 2006, followed by a decrease to 40 (13) in 2016. The low-vol- knee arthroplasties could indicate that we have reached an ume group had a higher score throughout the period, but the appropriate level of treatment with an adequate capacity. trends were converging and in 2016 the difference was minor. In Sweden, the pattern of UKA use has been markedly Low-usage centers had higher scores until 2014, where the different from that in Denmark (SKAR 2015, SKAR 2018).

358

This may reflect controversy in how to weigh the advantages against the risk of UKAs compared with TKAs. From the 1990s until 2014, the use of UKA in Sweden was reduced dramatically, and in 2014 the Swedish medial UKA percentage was just 3.4%. However, from 2014 to 2017 the medial UKA percentage in Sweden more than doubled to 7.7%. In 2014, 53% of the UKA implants in Sweden were Oxford and 40% were ZUK or Link. In 2017, this distribution changed to 66% Oxford and 23% ZUK or Link. With an increasing use of UKA and a higher proportion of Oxford implants at the expense of fixed-bearing implant types, the recent trends in Sweden might indicate a change equivalent to that we have seen in Denmark. In our study population, the proportion of female patients has been decreasing. This is similar to—though more pronounced than—the trend in the overall population of knee arthroplasty patients in Denmark (DKR 2017). A possible explanation for the greater decrease is reporting of higher revision rates among female compared with male UKA patients in the national registry (DKR 2017). Weight trends were increasing for both male and female patients, and the magnitude of the increments is comparable to that of the general population (Christensen et al. 2016). Thus, it is plausible that the weight trend we observe is mainly a reflection of changing demographics in the Danish population along with an increasing proportion of male patients. This is further supported by the analogous time patterns for sex and weight when grouped by volume and usage. Regarding age, the increasing trend we observed differs from the trend in the general knee arthroplasty population, which has been rather stable at 67–68 years (DKR 2017). Though UKA patients are still younger than the overall knee arthroplasty population, the difference is diminishing. This may be indicative of increasing uniformity in patient selection between UKAs and TKAs. Previous studies have reported that patients who received a UKA from a low-volume (Liddle et al. 2016) or a low-usage (Liddle et al. 2015b, Hamilton et al. 2017b) surgeon tended to be younger, which is in line with our findings. It has been hypothesized that this could be associated with a tendency to perform UKA in patients with earlier-stage disease (Hamilton et al. 2017b, Murray and Parkinson 2018). If this is indeed the case, it could explain the higher knee scores among patients from low-volume and low-usage centers. This hypothesis is further supported by a study from Jones et al. (2016) reporting that patients with early radiographic osteoarthritis were younger than patients receiving UKAs for boneon-bone osteoarthritis. Overall for the study period, the knee score in our study population is approximately 7 points higher than reported for knee arthroplasty patients in the latest annual report from DKR (DKR 2017). This is not surprising as the indications for the Oxford implant include a flexion deformity less than 15 degrees and an intact anterior cruciate ligament (Goodfellow 2011), both factors that are represented in the knee score (Insall et al. 1989). Notably, the bell-shaped time trend

Acta Orthopaedica 2019; 90 (4): 354–359

is also detectable in the overall knee arthroplasty population (DKR 2017), and therefore it is unlikely that the pattern can be attributed to factors relating uniquely to the UKA procedure. It is noteworthy that reporting of procedures to the DKR became mandatory in 2006 (Pedersen et al. 2012), which resulted in an increase in completeness in the registry from 82% in 2005 to 93% in 2007 (DKR 2017). This calls into question whether the trends seen before 2006 are distorted by reporting bias. When comparing the effects of volume with the effects of usage, it is notable that the time trends are generally similar for the low- and high-usage groups while the volume groups show more varying patterns. For sex, age, and weight, the lowvolume centers tend to deviate from the overall trend around 2007. This is coincident with the increase in the total number of centers performing UKA from 25 in 2006 to 35 in 2008. Data from the DKR annual report show that, in the same period, 14 new private knee arthroplasty centers appeared (DKR 2017). It is plausible that patient demographics differ in these new private centers; and as the increase in the total number of centers was accompanied by the occurrence of 10 additional low-volume centers but only 2 additional low-usage centers, this could explain the more fluctuating trends in the volume groups compared with the usage groups. In 2007 there was a change in legislation, giving patients the right to have their tax-financed surgery performed at a private center if the waiting time at their public hospital exceeded one month. This is a plausible explanation for the aforementioned changes and, as such, the changes in patient characteristics could represent either surgeon proclivity or selection bias in the patients making use of this new opportunity. As our study is based on an unselected population from a national registry, the external validity is generally high. However, as discussed above, the trends we see are results of complex interactions of changes in demographics and structure as well as factors relating to centers and surgeons, possibly including surgical technique. This complicates interpretation of our findings and might impede generalizability. Another limitation of the study is the categorization of volume and usage. Procedures in DKR are linked to centers and not surgeons, which is why our categorization is center-based. Baker et al. (2013) demonstrated that both surgeon and center volumes were associated with revision but, in addition, that surgeon volume was the more valuable measure of the 2. Hence, the operating surgeon is a possible confounder in our study. For both volume and usage, it is questionable whether our cut-off values are clinically relevant, and choosing other cut-off values might affect the results. In addition to this, the calculation of volume and usage was done for each calendar year. This ensured that centers would not be at a disadvantage if they either started or ceased performing UKA during the study period. But it also means that the volume or usage group does not necessarily express a center’s level of experience altogether. Our findings indicate how external factors can influence our data, emphasizing that aspects such as structural changes

Acta Orthopaedica 2019; 90 (4): 354–359

should always be considered when interpreting data. Volume seemed to be more vulnerable to this, and therefore we suggest that usage is a more robust measure and should be preferred in future research. In conclusion, there has been a considerable increase in use of UKA in Denmark, accompanied by a shift in patient demographics toward an older population with an increasing proportion of males, higher weights, and lower knee scores. Patient characteristics differed with center volume and usage, though for age and knee score the groups were converging. This suggests an increasingly uniform approach to patient selection in accordance with the more permissive view on candidacy.

AT, AP, MM, and CH conceived the study. MM performed data extraction. CH conducted data analyses and drafted the manuscript. All authors contributed to data interpretation and revision of the manuscript. The authors would like to thank Håkon Sandholdt (MSc Statistics) for advice and technical assistance, which has been a great help in data processing and presentation. Acta thanks Ola Rolfson and Annette W-Dahlfor help with peer review of this study. Badawy M, Fenstad A M, Bartz-Johannessen CA, Indrekvam K, Havelin L I, Robertsson O, A W D, Eskelinen A, Makela K, Pedersen A B, Schroder H M, Furnes O. Hospital volume and the risk of revision in Oxford unicompartmental knee arthroplasty in the Nordic countries: an observational study of 14,496 cases. BMC Musculoskelet Disord 2017; 18(1): 388. doi: 10.1186/s12891-017-1750-7. Baker P, Jameson S, Critchley R, Reed M, Gregg P, Deehan D. Center and surgeon volume influence the revision rate following unicondylar knee replacement: an analysis of 23,400 medial cemented unicondylar knee replacements. J Bone Joint Surg Am 2013; 95(8): 702-9. doi: 10.2106/ JBJS.L.00520. Burn E, Sanchez-Santos M T, Pandit H G, Hamilton T W, Liddle A D, Murray D W, Pinedo-Villanueva R. Ten-year patient-reported outcomes following total and minimally invasive unicompartmental knee arthroplasty: a propensity score-matched cohort analysis. Knee Surg Sports Traumatol Arthrosc 2018; 26(5): 1455-64. doi: 10.1007/s00167-016-4404-7. Chawla H, van der List J P, Christ A B, Sobrero M R, Zuiderbaan H A, Pearle A D. Annual revision rates of partial versus total knee arthroplasty: a comparative meta-analysis. Knee 2017; 24(2): 179-90. doi: 10.1016/j. knee.2016.11.006. Christensen A V, Davidsen M, Johnsen N F, Juel K. Vægtøgning i den danske befolkning 1994–2013. The National Institute of Public Health, University of Southern Denmark; 2016. Available at https://www.sdu.dk/da/ sif/rapporter/2016/vaegtoegning_i_den_danske_befolkning_1994-2013 (accessed November 16, 2018). Cleveland W S. Robust locally weighted regression and smoothing scatterplots. J Am Stat Assoc 1979; 74(368): 829-36. doi: 10.1080/01621459.1979.10481038. DKR. Annual report 2017. The Danish Knee Arthroplasty Registry (DKR). Available at https://www.sundhed.dk/content/cms/99/4699_dkr-rapport-2017_ final_til_offentligg%C3%B8relse.pdf (accessed November 16, 2018). Goodfellow J. Unicompartmental arthroplasty with the Oxford Knee. New ed. Oxford: Goodfellow Publishers; 2011.

359

Hamilton T W, Pandit H G, Jenkins C, Mellon S J, Dodd C A F, Murray D W. evidence-based indications for mobile-bearing unicompartmental knee arthroplasty in a consecutive cohort of a thousand knees. J Arthroplasty 2017a; 32(6): 1779-85. doi: 10.1016/j.arth.2016.12.036. Hamilton T W, Rizkalla J M, Kontochristos L, Marks B E, Mellon S J, Dodd C A F, Pandit H G, Murray D W. The interaction of caseload and usage in determining outcomes of unicompartmental knee arthroplasty: a meta-analysis. J Arthroplasty 2017b; 32(10): 3228-37 e2. doi: 10.1016/j. arth.2017.04.063. Insall J N, Dorr L D, Scott R D, Scott W N. Rationale of the Knee Society clinical rating system. Clin Orthop Relat Res 1989; (248): 13-14. Jones L D, Bottomley N, Harris K, Jackson W, Price A J, Beard D J. The clinical symptom profile of early radiographic knee arthritis: a pain and function comparison with advanced disease. Knee Surg Sports Traumatol Arthrosc 2016; 24(1): 161-8. doi: 10.1007/s00167-014-3356-z. Kozinn S C, Scott R. Unicondylar knee arthroplasty. J Bone Joint Surg Am 1989; 71(1): 145-50. Lau R L, Perruccio A V, Gandhi R, Mahomed N N. The role of surgeon volume on patient outcome in total knee arthroplasty: a systematic review of the literature. BMC Musculoskelet Disord 2012; 13: 250. doi: 10.1186/14712474-13-250. Liddle A D, Judge A, Pandit H, Murray D W. Adverse outcomes after total and unicompartmental knee replacement in 101,330 matched patients: a study of data from the National Joint Registry for England and Wales. Lancet 2014; 384(9952): 1437-45. doi: 10.1016/S0140-6736(14)60419-0. Liddle A D, Pandit H, Judge A, Murray D W. Patient-reported outcomes after total and unicompartmental knee arthroplasty: a study of 14,076 matched patients from the National Joint Registry for England and Wales. Bone Joint J 2015a; 97-B(6): 793-801. doi: 10.1302/0301-620X.97B6.35155. Liddle A D, Pandit H, Judge A, Murray D W. Optimal usage of unicompartmental knee arthroplasty: a study of 41,986 cases from the National Joint Registry for England and Wales. Bone Joint J 2015b; 97-B(11): 1506-11. doi: 10.1302/0301-620X.97B11.35551. Liddle A D, Pandit H, Judge A, Murray D W. Effect of surgical caseload on revision rate following total and unicompartmental knee replacement. J Bone Joint Surg Am 2016; 98(1): 1-8. doi: 10.2106/JBJS.N.00487. Murray D W, Parkinson R W. Usage of unicompartmental knee arthroplasty. Bone Joint J 2018; 100-B(4): 432-5. doi: 10.1302/0301-620X.100B4.BJJ2017-0716.R1. Pandit H, Jenkins C, Gill H S, Smith G, Price A J, Dodd C A, Murray D W. Unnecessary contraindications for mobile-bearing unicompartmental knee replacement. J Bone Joint Surg Br 2011; 93(5): 622-8. doi: 10.1302/0301620X.93B5.26214. Pedersen A B, Mehnert F, Odgaard A, Schroder H M. Existing data sources for clinical epidemiology: the Danish Knee Arthroplasty Register. Clin Epidemiol 2012; 4: 125-35. doi: 10.2147/CLEP.S30050. Robertsson O, Knutson K, Lewold S, Lidgren L. The routine of surgical management reduces failure after unicompartmental knee arthroplasty. J Bone Joint Surg Br 2001; 83(1): 45-9. SKAR. Annual report 2015. The Swedish Knee Arthroplasty Register (SKAR); 2015. Available at http://www.myknee.se/en/publications/ annual-reports (accessed January 25, 2019). SKAR. Annual report 2018. The Swedish Knee Arthroplasty Register (SKAR); 2018. Available at http://www.myknee.se/en/publications/ annual-reports (accessed January 25, 2019). Skou S T, Roos E M. Good Life with osteoArthritis in Denmark (GLA:D): evidence-based education and supervised neuromuscular exercise delivered by certified physiotherapists nationwide. BMC Musculoskelet Disord 2017; 18(1): 72. doi: 10.1186/s12891-017-1439-y. Stern S H, Becker M W, Insall J N. Unicondylar knee arthroplasty: an evaluation of selection criteria. Clin Orthop Relat Res 1993(286): 143-8. Willis-Owen C A, Brust K, Alsop H, Miraldo M, Cobb J P. Unicondylar knee arthroplasty in the UK National Health Service: an analysis of candidacy, outcome and cost efficacy. Knee 2009; 16(6): 473-8. doi: 10.1016/j. knee.2009.04.006.

360

Acta Orthopaedica 2019; 90 (4): 360–365

Patient-reported 1-year outcome not affected by body mass index in 3,327 total knee arthroplasty patients Anders OVERGAARD 1, Lars LIDGREN 2,3, Martin SUNDBERG 2,3, Otto ROBERTSSON 2,3, and Annette W-DAHL 2,3 1 The Parker Institute, Copenhagen University Hospital Fredriksberg, Copenhagen, Denmark; 2 The Swedish 3 Lund University, Faculty of Medicine, Department of Clinical Sciences Lund, Orthopaedics, Lund, Sweden

Knee Arthroplasty Register, Lund, Sweden;

Correspondence: anders.fohrby.overgaard@regionh.dk Submitted 2018-12-12. Accepted 2019-03-09

Background and purpose — Patient-reported outcome (PRO) in total knee arthroplasty (TKA) patients with high body mass index (BMI) is controversial. We compared pain, function, quality of life, general health, and satisfaction among different BMI categories preoperatively and 1 year after primary TKA. Patients and methods — 4,318 patients were operated with a TKA for knee osteoarthritis in the Region of Skane in 2013–2015. In all, 3,327 patients (77%) had complete PRO data and information on BMI and were included. Preoperatively the patients filled in the Knee injury and Osteoarthritis Outcome Score (KOOS) and EQ-VAS (general health). 1 year postoperatively the same questionnaires were filled in together with a question asking whether they were satisfied with the surgery. Information on age, sex, BMI, and ASA grade were obtained from the Swedish Knee Arthroplasty Register. Each patient was classified as Outcome Measures in Rheumatology– Osteoarthritis Research Society International (OMERACT–OARSI) responder or not based on a combination of absolute and relative changes in scores. Welch’s t-test and a chi-square test were used in the statistical analysis. Results — Both preoperatively and 1 year postoperatively the obese patients reported somewhat worse scores than the normal weight and overweight. The differences were small with 1 exception, the KOOS sport- and recreation function postoperatively, where normal-weight and overweight patients reported fewer problems than obese patients with a BMI over 35 (40 and 39 points vs. 31 points, p < 0.001). Similar proportions of patients were satisfied and categorized as OMERACT–OARSI responders in the different BMI categories. Interpretation — The degree of improvement in PROs 1 year after TKA surgery does not seem to be affected by BMI.

As the number of primary knee arthroplasties, as well as the number of obese patients undergoing total knee arthroplasty (TKA), continues to increase, there has been more interest in the role of obesity as a risk factor for poor outcomes after TKA. In the literature, the influence of obesity on knee arthroplasty outcome diverges. Some studies show that obesity has no influence on TKA outcomes such as pain and function (Deshmukh et al. 2002, Stevens-Lapsley et al. 2010, Yeung et al. 2011, Baker et al. 2012, Collins et al. 2012), patient satisfaction (Yeung et al. 2011), early complications (Patel and Albrizio 2008, Yeung et al. 2011, Collins et al. 2012), and mid-term survival of the knee arthroplasty (Bourne et al. 2008, Bordini et al. 2009, Yeung et al. 2011, Collins et al. 2012). However, others have found worse outcomes regarding pain and function (Mulhall et al. 2007, Järvenpää et al. 2012, Issa et al. 2013, Liljensoe et al. 2013), satisfaction (Järvenpää et al. 2012), complications (Yasunaga et al. 2009, Järvenpää et al. 2010, Issa et al. 2013) and an increased risk of infection (Namba et al. 2005). If orthopedic surgeons hesitate to operate on obese patients, because of a suspected greater risk of worse outcome after arthroplasty surgery, this may lead to disparity in surgical treatment among the general population. Considering the rising prevalence of obesity, it is of importance to evaluate whether the knee arthroplasty surgery benefits the patients in order to help patients and surgeons decide on treatment. A rising public health concern about the influence of the obesity epidemic on the healthcare system furthers a need to substantiate the effect of obesity on treatments (King et al. 2013). We compared pain, function, quality of life, and general health preoperatively and 1 year postoperatively, as well as improvement and satisfaction 1 year postoperatively in patients operated on with TKA for knee OA, and stratified our analyses to investigate the effects obesity had on patientreported outcomes (PROs).

Acta Orthopaedica 2019; 90 (4): 360–365

Table 1. Preoperative demographics and preoperative PROs for patients included and excluded in the study and the TKA/OA patients in the SKAR 2013–2015. Values are mean (95% CI) unless otherwise specified Variable Women, n (%) Age BMI ASA ≥ 3, n (%) KOOS pain symptoms ADL sport/rec QoL EQ-VAS

Included Excluded a SKAR (n = 3,327) (n = 991) p-value (n = 35,932) 1,912 (58) 593 (60) 0.2 20,389 (57) 69 (69–70) 69 (69–70) 1.0 69 (69–69) 29 (29–29) 29 (29–29) 0.02 29 (29–29) 513 (15) 186 (19) 0.01 6,035 (17) n = 509 41 (40–41) 38 (37–39) < 0.001 48 (47–49) 45 (43–46) < 0.001 46 (46–47) 43 (42–44) < 0.001 12 (12–13) 11 (10–12) 0.07 24 (23–24) 20 (19–22) < 0.001 n = 527 70 (69–70) 64 (6–66) < 0.001

a Excluded or lost to follow-up. BMI = body mass index, ASA = American Society of Anesthesiologists, KOOS = Knee injury and Osteoarthritis Outcome Score, ADL = activity in daily life function, Sport/rec = sport and recreation function, QoL = quality of life, VAS = visual analogue scale.

Patients and method The study population consisted of 4,318 TKA patients (5,065 knees) operated on for knee osteoarthritis (OA) between 2013 and 2015 in the most southern region of Sweden (Region Skane). Of the 4,318 patients, 318 had bilateral simultaneous TKA and 429 bilateral staged TKA during the study period. For patients having bilateral simultaneous TKA the right TKA was considered and for patients with staged surgery, the later surgery was considered. Further, we excluded patients who did not have both preoperative and 1-year postoperative PRO data and those who had died during the follow-up year. Of the 4,286 available patients 77% had complete PRO data and information on BMI (954 patients had not complete PRO data and in 5 patients BMI was missing). The patient characteristics and the preoperative PROs available for the 991 patients excluded or lost to follow-up were similar to those included without clinically relevant differences (Table 1). Patient characteristics such as sex, age, BMI, and ASA classification were obtained from the Swedish Knee Arthroplasty Register (SKAR). BMI was categorized according to the WHO classification underweight (< 18.5), normal weight (18.5–24.9), overweight (25–29.9), obese I (30–34.9), obese II (35–39.9), and obese III (≥ 40). Preoperatively the patients filled in the disease-specific Knee injury and Osteoarthritis Outcome Score (KOOS) (Roos et al. 1998) and the generic instrument EQ-VAS (general health) (EuroQol Group 1990). 1 year postoperatively the same questionnaires were sent to the patients together with a question as to whether they were satisfied with the surgery.

361

The patients were informed of the planned 1-year follow-up, but no reminders were sent if they did not respond. The KOOS consists of 42 questions and includes 5 subscales. Each question is allotted a score from 0 to 4. A normalized score (100 indicating no symptoms and 0 indicating extreme symptoms) is calculated for each subscale. The patients reported their self-perceived general health using the EQ-VAS on a scale (0–100) from the best (100) to the worst imaginable health status (0) and their satisfaction with the arthroplasty surgery using a 0–100 scale (VAS) in which 0 was the highest imaginable degree of satisfaction and 100 was the worst imaginable degree of satisfaction. The satisfaction (VAS) score was categorized into 5 groups: very satisfied (0–20), satisfied (21–40), moderately satisfied (41–60), dissatisfied (61–80), and very dissatisfied (81–100). The KOOS was converted to Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) to be able to classify each patient as an Outcome Measures in Rheumatology–Osteoarthritis Research Society International (OMERACT–OARSI) responder or not at 1 year based on a combination of absolute and relative changes in WOMAC pain, function, and total scores (Pham et al. 2004). The outcome at 1 year was categorized into responders (high and low) and non-responders according to these criteria (Pham et al. 2004). Statistics PRO, age, and BMI are presented as mean value (SD) and/or 95% confidence interval (CI). Welch’s t-test was used for comparisons of the KOOS and the EQ-VAS between the different BMI categories considering unequal variances and unequal sample sizes with assumption of normal distribution. Due to few patients in the underweight and obese III groups, these were analyzed together with the normal weight and obese II groups respectively. For analysis of proportions of sex, ASA grade, and OMERACT–OARSI responder rate the chisquared test was used for comparisons. A difference between the groups of ≥ 8 points in KOOS and ≥ 15 mm in EQ-VAS was considered a clinically relevant difference for statistically significant results (p < 0.05). Multiple linear regression analysis was used to evaluate the relationship between BMI and change in KOOS pain and ADL function preoperatively to 1 year postoperatively adjusted for age and sex. In a further analysis ASA grade (ASA I, ASA II, and ASA ≥ III) and preoperative KOOS pain and ADL function respectively were included in the model in addition to age and sex. Statistical analyses were carried out using Stata version 14 (StataCorp, College Station, TX, USA).

Results Of the 3,327 patients, 58% were women, the mean age was 69 years, mean BMI 29, and 15% were classified as ASA ≥ III (Table 1). 0.2% of the included patients were underweight,

362

Acta Orthopaedica 2019; 90 (4): 360–365

Table 2. Patient reported outcome preoperatively and 1 year postoperatively in the different BMI categories. Values are mean (SD) CI

Normal weight Overweight Obese I Obese II + III p-value p-value p-value p-value p-value p-value (N) (O) (I) (II+) N vs O N vs I N vs II+ O vs I O vs II+ I vs II+ Preoperatively KOOS pain 43 (15) 42–45 41 (15) 40–41 39 (15) 38–40 37 (15) 35–38 0.002 < 0.001 < 0.001 0.004 < 0.001 0.009 symptom 50 (19) 48–51 48 (18) 47–49 47 (17) 46–49 45 (17) 43–47 0.04 0.006 < 0.001 0.3 0.004 0.04 ADL 49 (16) 48–50 47 (15) 47–48 44 (15) 43–45 42 (16) 40–44 0.01 < 0.001 < 0.001 < 0.001 < 0.001 0.08 sport/rec 14 (15) 13–16 13 (14) 12–14 11 (14) 10–11 9 (14) 7–11 0.03 < 0.001 < 0.001 < 0.001 < 0.001 0.1 QoL 25 (14) 24–26 24 (14) 24–25 23 (14) 22–24 21 (13) 19–22 0.3 0.003 < 0.001 0.008 < 0.001 0.02 EQ–VAS 72 (21) 70–74 71 (21) 70–72 67 (21) 65–68 64 (22) 61–66 0.4 < 0.001 < 0.001 < 0.001 < 0.001 0.05 Postoperatively KOOS pain 80 (19) 79–82 80 (19) 79–81 78 (19) 77–79 78 (19) 77–79 0.4 0.02 0.02 0.06 0.07 0.6 symptom 77 (17) 76–78 76 (17) 75–77 75 (17) 74–76 75 (17) 74–76 0.3 < 0.001 0.1 0.06 0.4 0.8 ADL 81(18) 79–82 79 (19) 78–80 76 (19) 75–77 74 (20) 72–76 0.5 < 0.001 < 0.001 < 0.001 < 0.001 0.1 sport/rec 40 (27) 38–42 39 (27) 38–40 34 (27) 32–36 32 (28) 32–35 0.3 < 0.001 < 0.001 < 0.001 < 0.001 0.2 QoL 66 (23) 64–68 65 (24) 64–66 62 (24) 60–63 61 (25) 60–63 0.4 < 0.001 0.008 0.002 0.02 0.8 EQ–VAS 78 (19) 77–80 77 (19) 76–78 74 (20) 73–75 72 (20) 69–75 0.3 < 0.001 < 0.001 < 0.001 < 0.001 0.2 Change KOOS pain 32 (21) 30–34 34 (21) 33–35 35 (22) 33–36 36 (23) 34–39 0.08 0.02 0.006 0.3 0.07 0.2 symptom 27 (22) 25–29 28 (22) 27–29 27 (22) 26–29 28 (22) 27–29 0.3 0.7 0.05 0.5 0.1 0.5 ADL 31 (19) 30–33 32 (20) 31–33 32 (20) 31–34 32 (20) 31–33 0.8 0.3 0.6 0.4 0.6 0.9 sport/rec 26 (27) 24–28 26 (27) 25–28 23 (27) 21–25 23 (29) 19–26 0.8 0.05 0.09 0.006 0.04 0.7 QoL 41 (23) 39–43 41 (25) 40–42 39 (26) 38–41 41 (25) 38–44 0.8 0.1 0.8 0.1 1.0 0.3 EQ–VAS 5 (25) 3–7 5 (26) 4-–7 8 (26) 5–11 8 (25) 5–11 0.09 0.2 0.5 0.1 0.1 0.5 For abbreviations, see Table 1

19% normal weight, 45% overweight, 27% obese I, 7% obese II, and 1% obese III. Both preoperatively and 1 year postoperatively the obese patients reported statistically significant worse KOOS scores than the normal weight and overweight in most of the subscales without clinically relevant differences. The only exception was in Sport/Rec function postoperatively were normalweight and overweight patients reported better outcome then obese II–III patients (40 (CI 38–42) and 39 (CI 38–41) vs. 31 (CI 28–34)) (Table 2). The improvements were comparable in the KOOS subscales pain, symptoms, ADL function, and knee-related QoL without any clinically relevant differences. In Sport/Rec function the normal weight and overweight improved somewhat more than the obese patients although without any clinically relevant differences (Figure 1). We could not show any effect of BMI on change in KOOS pain (0.1 [–0.05 to 0.3]) and ADL function (0.03 [–0.1 to 0.2]) when adjusting for age and sex. When we included ASA grade and preoperative KOOS pain and ADL function respectively in the models, BMI was not found to have any effect on change in KOOS pain but a statistically significant effect on change in ADL function (2 points less improvement/10 higher BMI units) (Table 3). The normal-weight and overweight patients reported somewhat better general health (EQ-VAS) preoperatively than the obese patients without statistically or clinically significant

KOOS, mean value 50 45 40 35 30 25 20 15 10 5 0

normal weight overweight obese I obese II + III Pain

Symptoms

ADL

Sport/rec

QoL

Figure 1. Mean changes preoperatively to 1 year postoperatively in KOOS 5 subscales in the different BMI categories.

differences. The improvement in general health, among the different BMI categories, preoperatively to 1 year postoperatively was small: 5–8 points (Figure 2). The proportion of satisfied patients varied between 80% and 83% in the different BMI categories (Figure 3). Similar proportions (85–88%) of patients in the different BMI categories were classified as OMERACT–OARSI responders and the majority of the patients were classified

Acta Orthopaedica 2019; 90 (4): 360–365

363

Table 3. Relationship between potential confounding factors to change in the KOOS pain and ADL function Variable

Change in KOOS pain coefficient (CI) p-value

Change in KOOS ADL coefficient (CI)

p-value

Age 0.1 (0.02–0.2) 0.02 –0.1 (–0.2 to –0.1) 0.02 Sex Male Ref Ref Female –1.6 (–2.9 to –0.3) 0.02 –1 (–2.2 to –0.3) 0.1 BMI –0.3 (–0.2 to 0.1) 0.7 –0.2 (–0.4 to –0.1) 0.004 ASA classification 1 Ref Ref 2 –2.3 (–4 to –0.6) 0.008 –2.3 (–3.9 to –0.6) 0.008 3 –3.7 (–6.1 to –1.4) 0.002 –4.7 (–6.9 to –2.4) < 0.001 Preoperative KOOS pain –0.7 (–0.8 to –0.7) < 0.001 ADL –0.6 (–0.7 to –0.6) < 0.001 For abbreviations, see Table 1.

as high responders: 73–80%. Patients with BMI ≥ 35 had the highest proportion of responders (Figure 4).

Discussion We found that obese primary TKA patients in southern Sweden reported similar knee-related pain, function, quality of life, and satisfaction as non-obese patients at 1 year after surgery, with a comparable proportion of OMERACT–OARSI responders. This may be valuable information for the knee arthroplasty surgeons when considering obese patients for TKA surgery. Earlier studies have reported disadvantageous patient/surgeon-reported outcome in obese patients when compared with non-obese patients after TKA surgery (Järvenpää et al. 2012,

Liljensoe et al. 2013, Issa et al. 2013) or similar results (Collins et al. 2012). However, the most recent studies from the United States and Western Europe are in line with our results (Daniilidis et al. 2016, Li et al. 2017, Giesinger et al. 2018). The reasons for dissimilarities may be differences in study size, loss to follow-up, time to follow-up, patient selection, and methods of scoring. The majority of these studies include relatively few patients and are from single centers (Järvenpää et al. 2012, Collins et al. 2012, Issa et al. 2013, Liljensoe et al. 2013, Daniilidis et al. 2016, Giesinger et al. 2018). Further, the follow-up time in the above-mentioned studies varies between 6 months and 11 years, which may influence the results. One of the strengths of our study is a large patient base that resembles the national data in Sweden, though gathered from only 5 centers. In contrast to the other studies, our Swedish cohort included a relatively low proportion of obese patients (35%) and especially obese III patients (1%). The US FORCE-TJR cohort (2,964 TKA patients) included 53% obese patients, and 9% of these were obese III (Li et al. 2017), the UK cohort (402 TKA patients) included 51% obese patients of whom 7% were obese III (Giesinger et al. 2018), and the German cohort (199 TKA patients) consisted of 63% obese patients including 9% obese III (Daniilidis et al. 2016). The number of obese III patients is few in all studies except for the US FORCE-TJR cohort, which included 272 obese III patients. Data support an increase in morbidity and mortality associated with severe obesity (Krushell and Fingeroth 2007, Vaishya et al. 2016) among patients receiving TKA, but for non-morbidly obese (overweight and obesity I–II) patients, the data are not clear. The non-obese patients reported somewhat better general health than the obese patients but the differences were small (4–8) preoperatively with small changes 1 year after the TKA surgery in the different BMI categories. The small change in

EQ-VAS mean value

Distribution of patient satisfaction

Distribution of responders

100

Preoperative 1-year postoperative

Normal weight

Overweight

Obese I

Obese II + III

Figure 2. EQ-VAS (general health) mean value preoperatively and 1 year postoperatively in the different BMI categories.

Very dissatisfied Not satisfied Moderate satisfied Satisfied Very satisfied

Normal weight

Overweight

Obese I

Obese II + III

Figure 3. Patient satisfaction in the different BMI categories.

Non-responder Low responder High responder

All

Normal Over- Obese I Obese weight weight II + III

Figure 4. Responder classification in the different BMI categories.

364

general health preoperatively to 1 year postoperatively indicate that the patients did not experience significant general health improvement from the TKA surgery and that a general health measure alone may not be a suitable outcome measure to evaluate the knee arthroplasty surgery. We found no clinically relevant differences in the patient reported outcome between obese and non-obese patients measured by the KOOS at the 1-year follow-up after knee arthroplasty surgery except for the subscale sports and recreation function, which had a clinically and statistically significant different outcome when comparing obese II + III and normal weight and overweight patients. Severe obesity is associated with functional disability (Anandacoomarasamy et al. 2008). The subscale sport and recreation function includes more demanding functions such as squatting, running, jumping etc., which might explain the difference in this outcome measure. Satisfaction has become a common measure to define a successful outcome when evaluating elective arthroplasty surgery. Though satisfaction does not always correlate with PROderived responder rates and it has been advocated that a poor outcome, as well as a good one, might be hidden when reporting the mean pain and function scores (Roos 2018). Our data show 80–82% of the patients in the different BMI categories were categorized to be “very satisfied” or “satisfied” with the surgery while 85–87% were classified as OMERACT–OARSI responders. Bourne et al. (2008) showed that the relationship between patient-reported outcome and patient satisfaction is multifactorial, which may to some extent explain the discrepancy in proportions between satisfied patients and OMERACT–OARSI responders in our study though our study clearly shows similar satisfaction rates among BMI categories. Our patients had surgery in a region in the south of Sweden. These patients presented similar patient characteristics to patients operated on for OA with TKA in the whole country (Table 1). The patients excluded/lost to follow-up in our study for different reasons consisted of a somewhat higher proportion of women although without a statistically significant difference. However, we found a statistically significant difference in the mean BMI and proportion of ASA ≥ 3 patients but the difference was small (0.4% and 3% respectively) between the included patients and those excluded/lost to follow-up, though of no real clinical importance. In 2015, almost 13,000 knee arthroplasties were performed in Sweden resulting in an age-standardized incidence of 132/100,000 inhabitants (SKAR 2016). That more than onethird of the patients having primary knee arthroplasty surgery for OA in Sweden 2015 were obese (BMI ≥ 30) may reflect on an increased risk of progression of OA in the obese population. In summary, considering the similar patient-reported outcome in the different BMI categories, with reservations for the low number of obese III patients, BMI seems to have little effect on patient-reported outcome 1 year postoperatively in patients having a TKA for OA.

Acta Orthopaedica 2019; 90 (4): 360–365

The study was conceived by OR, LL, MS, and AWD. AO, AWD, and OR performed the analyses. AO wrote the initial draft. All the authors contributed to the interpretation of the data and to a revision of the manuscript. Acta thanks Hannu Miettinen for help with peer review of this study.

Anandacoomarasamy A, Caterson I, Sambrook P, Fransen M, March L. The impact of obesity on the musculoskeletal system. Int J Obes 2008; 32: 211-22. Baker P, Petheram T, Jameson S, Reed M, Gregg P, Deehan D. The association between body mass index and the outcomes of total knee arthroplasty. J Bone Joint Surg Am 2012; 94: 1501-8. Bordini B, Stea S, Cremonini S, Viceconti M, De Palma R, Toni A. Relationship between obesity and early failure of total knee prostheses. BMC Musculoskelet Disord 2009; 10: 29. Bourne R B, Chesworth B M, Davis A M, Mahomed N N, Charron K D J. Patient satisfaction after total knee arthroplasty: who is satisfied and who is not? Clin Orthop Relat Res 2008; 468(1): 57-63. Collins R A, Walmsley P J, Amin A K, Brenkel I J, Clayton R A. Does obesity influence clinical outcome at nine years following total knee replacement? J Bone Joint Surg Br 2012; 94: 1351-5. Daniilidis K, Yao D, Gosheger G, Berssen C, Budny T, Dieckmann R, Höll S. Does BMI influence clinical outcomes after total knee arthroplasty? Technol Health Care 2016; 24(3): 367-75. Deshmukh R G, Hayes J H, Pinder I M. Does body weight influence outcome after total knee arthroplasty? A 1-year analysis. Outcome of total knee arthroplasty in obese patients. J Arthroplasty 2002; 17: 315-19. EuroQol Group. EuroQol—a new facility for the measurement of healthrelated quality of life. Health Policy 1990; 16: 199-208. Giesinger J M, Loth F L, MacDonald D J, Giesinger K, Patton J T, Simpson A H R W, Howie C R, Hamilton D F. Patient-reported outcome metrics following total knee arthroplasty are influenced differently by patients’ body mass index. Knee Surg Sports Traumatol Arthrosc 2018; 26(11): 3257-64. Issa K, Pivec R, Kapadia B H, Shah T, Harwin S F, Delanois R E, et al. Does obesity affect the outcomes of primary total knee arthroplasty? J Knee Surg 2013; 26: 89-94. Järvenpää J, Kettunen J, Kroger H, Miettinen H. Obesity may impair the early outcome of total knee arthroplasty. Scand J Surg 2010; 99: 45-9. Järvenpää J, Kettunen J, Soininvaara T, Miettinen H, Kroger H. Obesity has a negative impact on clinical outcome after total knee arthroplasty. Scand J Surg 2012; 101: 198-203. Li W, Ayers D C, Lewis C G, Bowen T R, Allison J J, Franklin P D. Functional gain and pain relief after total joint replacement according to obesity status. J Bone Joint Surg Am 2017; 99(14): 1183-9. King L K, March L, Anandacoomarasamy A. Obesity & osteoarthritis. Ind J Med Res 2013; 138(2): 185-93. Krushell R J, Fingeroth R J. Primary total knee arthroplasty in morbidly obese patients. J Arthroplasty 2007; 22(6): 77-80. Liljensoe A, Lauersen J O, Soballe K, Mechlenburg I. Overweight preoperatively impairs clinical outcome after knee arthroplasty: a cohort study of 197 patients 3–5 years after surgery. Acta Orthop 2013; 84: 392-7. Mulhall K J, Ghomrawi H M, Mihalko W, Cui Q, Saleh K J. Adverse effects of increased body mass index and weight on survivorship of total knee arthroplasty and subsequent outcomes of revision TKA. J Knee Surg 2007; 20: 199-204. Namba R S, Paxton L, Fithian D C, Stone M L. Obesity and perioperative morbidity in total hip and total knee arthroplasty patients. J Arthroplasty 2005; 20(7 Suppl. 3): 46-50. Patel A D, Albrizio M. Relationship of body mass index to early complications in knee replacement surgery. Arch Orthop Trauma Surg 2008; 128: 5-9. Pham T, van der Heijde D, Altman R D, Anderson J J, Bellamy N, Hochberg M, Simon L, Strand V, Woodworth T, Dougados M. OMERACT–OARSI

Acta Orthopaedica 2019; 90 (4): 360â&#x20AC;&#x201C;365

initiative: Osteoarthritis Research Society International set of responder criteria for osteoarthritis clinical trials revisited. Osteoarthritis Cartilage 2004; 12(5): 389-99. Roos E M. 3 steps to improve reporting and interpretation of patient reported outcome scores in orthopedic studies: Using the KOOS as an example. Acta Orthop 2018; 89(1): 1-2. Roos E M, Roos H P, Lohmander L S, Ekdahl C, Beynnon B D. Knee Injury and Osteoarthritis Outcome Score (KOOS): development of a self-administered outcome measure. J Orthop Sports Phys Ther 1998; 28: 88-96. Stevens-Lapsley J E, Petterson S C, Mizner R L, Snyder-Mackler L. Impact of body mass index on functional performance after total knee arthroplasty. J Arthroplasty 2010; 25: 1104-9.

365

Swedish Knee Arthroplasty Register: Annual Report; 2016. http://www. myknee.se Vaishya R, Vijay V, Wamae D, Agarwal A K. Is total knee replacement justified in the morbidly obese? A systematic review. Cureus 2016; 8(9): e804. doi:10.7759/cureus.804. Yasunaga H, Tsuchiya K, Matsuyama Y, Ohe K. Analysis of factors affecting operating time, postoperative complications, and length of stay for total knee arthroplasty: nationwide web-based survey. J Orthop Sci 2009; 14: 10-16. Yeung E, Jackson M, Sexton S, Walter W, Zicat B, Walter W. The effect of obesity on the outcome of hip and knee arthroplasty. Int Orthop 2011; 35: 929-34.

366

Acta Orthopaedica 2019; 90 (4): 366–372

Marker-based versus model-based radiostereometric analysis of total knee arthroplasty migration: a reanalysis with comparable mean outcomes despite distinct types of measurement error Koen T VAN HAMERSVELD 1, Perla J MARANG–VAN DE MHEEN 2, Lennard A KOSTER 1, Rob G H H NELISSEN 1, Sören TOKSVIG-LARSEN 3, and Bart L KAPTEIN 1 1 Department

of Orthopaedics, Leiden University Medical Center, Leiden, the Netherlands; 2 Department of Biomedical Data Sciences, Leiden University Medical Center, Leiden; the Netherlands; 3 Department of Orthopaedics, Hässleholm Hospital, Hässleholm, Sweden and Department of Clinical Sciences, Lund University, Lund, Sweden Correspondence: ktvanhamersveld@lumc.nl Submitted 2018-10-30. Accepted 2019-03-25.

Background and purpose — Pooling data of studies evaluating total knee arthroplasty migration using radiostereometric analysis (RSA) may be compromised when the RSA method used would influence estimated differences between groups. We therefore reanalyzed a marker-based RSA study with model-based RSA to assess possible limitations of each RSA method, including insert micromotions in modular TKA and their effect on estimated group differences. Patients and methods — All patients had received a cemented Triathlon implant (Stryker, Mahwah, NJ, USA) with either an all-polyethylene (n = 29) or a metal-backed (n = 28) tibial component. The latter group was reanalyzed with model-based RSA. Precision of each RSA method was calculated using double examinations. Bland–Altman plots were constructed to determine the limits of agreement between the 2 RSA methods. Polyethylene insert micromotion was quantified by measuring migration with respect to the metal tray. Finally, analyses of the original study were repeated with the model-based RSA results. Results — Systematic differences were found in translations between marker-based and model-based RSA as a result of different reference origins being used for migration calculations. Micromotions of the polyethylene insert within the metal tray were negligibly small. Mean migration results were comparable between marker-based and model-based RSA when using the same reference origin, even though conclusions on individual patients may differ between RSA methods due to various types of measurement error (e.g., marker occlusion and model-fit inaccuracies).

Interpretation — At least for the studied TKA design, pooling mean migration data of different RSA methods appears justified. For translations, however, adjustments should be made to correct for differences in reference origin. Migration patterns of individual patients may differ as a result of distinct types of measurement error.

Due to the high accuracy and precision of radiostereometric analysis (RSA), late loosening of new implants can already be predicted with 2-year RSA results on small patient numbers (Ryd et al. 1995, Valstar et al. 2005, Nelissen et al. 2011). RSA requires the bone and prosthesis to be accurately defined in 3 dimensions, usually achieved by inserting tantalum markers in the bone and by attaching or inserting markers (in)to the prosthesis (i.e., marker-based RSA). Prosthesis markers are generally inserted during surgery in the polyethylene of the implant (Kaptein et al. 2007). Alternatively, in model-based RSA the need for prosthesis markers is eliminated by matching a virtual projection of a 3D model with the contours of the radiographic projection of the implant (Kaptein et al. 2003). Results of model-based RSA are suggested to be comparable with conventional marker-based methods on a group level (Pijls et al. 2018), but direct comparisons on individual patient data are scarce (Kaptein et al. 2007, Hurschler et al. 2009). We recently published the 2-year results of a randomized controlled trial (RCT) on implant migration of cemented metal-backed versus all-polyethylene tibial components in total knee arthroplasty (TKA) using the Triathlon TKA system

Acta Orthopaedica 2019; 90 (4): 366–372

367

Figure 1. RSA images showing the biplanar (lateral and anteroposterior) views with the polyethylene markers and tibial bone markers encircled in red, the fiducial markers in yellow, and the control markers in green. (a) Only 3 of 5 polyethylene markers are visible due to over-projection of 2 markers, in most cases, by the femoral component, which may reduce or invalidate the marker-based accuracy of the RSA measurement. However, migration can also be measured by fitting a model using the contours of the metal-backed tibial component as shown in orange. (b) Migration of the radiolucent all-polyethylene tibial component can only be measured with marker-based RSA.

(Stryker, Mahwah, NJ, USA) (van Hamersveld et al. 2018). Higher migration was found after 2 years for the metal-backed components, even though the difference was small. However, as migration measurements were based on markers inserted in the polyethylene, apparent migration of the modular metalbacked components may partly result from micromotion of the polyethylene insert with respect to the metal tray, a phenomenon that has been shown to occur in older fixed-bearing designs (Nilsson et al. 2003, Hansson et al. 2005). In this study, we reanalyzed the metal-backed components with model-based RSA to eliminate any influence of modularity on migration results and thus investigate whether methodological differences between RSA methods would affect migration results. Second, we quantified movements of the polyethylene insert within the locking mechanism of the metal tray. Finally, we investigated whether the use of model-based RSA would result in different conclusions of the RCT as compared with the marker-based results.

Patients and methods Full details of the original RCT regarding patients, randomization, follow-up, prosthesis, and surgical techniques have been described previously (van Hamersveld et al. 2018). Briefly, 2 surgeons implanted cemented, condylar-stabilizing, cruciate-retaining Triathlon total knee prostheses with either allpolyethylene (n = 29) or modular fixed-bearing metal-backed tibial components (n = 30). The metal tray was designed with a full peripheral capture locking mechanism and an anti-rotational central island (Łapaj et al. 2017). 2 patients with metalbacked components were analyzed with model-based RSA in the original RCT due to polyethylene marker occlusion, which precluded marker-based measurements. Hence, no marker-

based results were available for comparison and these were thus excluded in the present study. Radiostereometric analysis The first RSA examination, performed on the first postoperative day, served as the reference for the migration measurements. Subsequent examinations were performed at 3 months, 1 year, and 2 years after surgery. RSA radiographs were performed in supine position with the knee in a biplanar calibration cage (cage 10, RSA Biomedical, Umeå, Sweden) and analyzed using Model-based RSA software version 4.1 (RSAcore, LUMC, Leiden, the Netherlands). For markerbased RSA analysis 5 tantalum markers (0.8 mm in diameter) were inserted during surgery, after drilling appropriate holes, at standardized positions in the polyethylene of both tibia designs. 2 markers were placed posteriorly, 2 anteromedially/anterolaterally, and 1 anteriorly. The number of markers available for migration calculations could differ over time due to marker occlusion (Figure 1a). Marker-based results of the metal-backed group were based on all 5 polyethylene markers in only 3 patients. As a result of marker occlusion in 1 or more follow-up moments, marker-based results were based on 4 polyethylene markers in 8 patients and on 3 markers in 17 patients. Model-based reanalysis was performed only in the metal-backed group, as the all-polyethylene components are radiolucent (Figure 1b). In the RSA analysis of the original report, a triangulated surface model (from reversed engineering, reduced to 5,000 triangles) was added for the tibial component and its virtual projections were matched with the contours of the radiographic projection of the implant. All other aspects of the analysis, such as insert markers, bone markers, and calibration markers, remained unchanged. Migration of the 28 metal-backed tibial components, by means of the 3D surface model, was calculated twice: with

368

Figure 2. Lateral views showing the different reference origins (center of migrating model at reference time point T0) in (a) model-based and (b) marker-based RSA. The longitudinal axis is represented by the yellow line, the sagittal axis by the green line, and the red hexagon represents the origin. To fully compare model-based and marker-based RSA data using the same reference origin, a second model-based analysis was performed with the reference origin fixed in the center of the polyethylene markers as shown in b.

the reference origin for migration calculations (1) in the geometric center of the model, which is the standard position for model-based RSA analysis, and (2) in the geometric center of the polyethylene markers, which is the standard position for marker-based RSA analysis (Figure 2). In addition, migration of the polyethylene insert markers was determined to assess whether the insert moved with respect to the metal tray. Lastly, method 2 allowed us to compare model-based metalbacked results with marker-based all-polyethylene results using the same reference origin. The precision of each RSA method was determined by means of double examinations at 1-year follow-up. The precision is expressed as the upper limits of the 95% confidence interval (CI) around zero motion (ISO 16087:2013(E) 2013). The primary outcome measure used in the original report is the maximum total point motion (MTPM), which is the length of the translation vector of the marker that moved the most. For model-based RSA, MTPM is the length of the translation vector of the point on the model that moved the most. We also report the number of individual components showing “continuous migration,” defined by Ryd et al. (1995) as an increase in MTPM of ≥ 0.2 mm in the second postoperative year. The limits of marker stability (mean error) and scatter values (condition number) were set at 0.35 mm and 120, respectively, complying with the RSA guidelines (Valstar et al. 2005). Statistics We first estimated differences in model-based analyses with 2 different reference origins, i.e., the reference origin in the geometric center of the model versus the geometric center of the polyethylene markers, using regression analysis. Bland–Altman plots were constructed to determine the limits

Acta Orthopaedica 2019; 90 (4): 366–372

of agreement between the two RSA methods (Bland and Altman 1986). The limits of agreement, defined as the mean ± 1.96×SD, should be within ± 0.5 mm of translation or ± 0.8° of rotation for the measures to be considered equivalent. These thresholds were chosen as these are considered the smallest values of clinically relevant early migration when used as a predictor of aseptic loosening (Hurschler et al. 2009, Gudnason et al. 2017, Pijls et al. 2018). Boxplots were constructed to investigate micromotion of the polyethylene markers with respect to the metal tray along and about each orthogonal axis. Finally, an identical linear mixed-effects model as described in the original report (van Hamersveld et al. 2018) was used to analyze differences in migration between (model-based) metal-backed and (marker-based) all-polyethylene components while using the same reference origin (center of the polyethylene markers). As in the original report, log-transformation of outcome measures was applied when necessary to obtain normal distributions, and the same sensitivity analysis was performed given the unevenly distributed baseline characteristics sex and surgeon as possible confounders by adding these variables to the linear mixed-effects model (van Hamersveld et al. 2018). Significance was set at p < 0.05 (IBM SPSS Statistics 24.0; IBM Corp, Armonk, NY, USA). Ethics, registration, funding, and potential conflicts of interest The original study was approved by the Regional Ethical Review Board in Lund (entry no. 2013/434) and registered at isrctn.com (ID: ISRCTN04081530). All patients gave informed consent. The costs of the RSA radiographs made for the original study were supported by Stryker. The sponsor did not take part in the design, conduct, analysis, or interpretations stated in both the previous and current manuscript. The authors declare no competing interests.

Results Double examinations were performed in 21 metal-backed components at 1-year follow-up to determine the precision of the RSA measurements. Model-based results were less precise in rotations, especially about the longitudinal axis (Table 1). Marker-based versus model-based RSA Regression analysis revealed that with (1) routine modelbased RSA versus (2) model-based RSA with the reference origin in the geometric center of the polyethylene markers, the transverse, longitudinal, and sagittal translations were overestimated by 29% (CI 25–32), 7% (CI 0–13) and 26% (CI 24–28), respectively (illustrated for transverse translations in Figures 3a and 3b). As expected (for mathematical reasons, see Appendix), rotations and MTPM values were not influenced by the position of the reference origin and therefore identical between both model-based analyses. For fair com-

Acta Orthopaedica 2019; 90 (4): 366–372

369

Table 1. Precision of RSA measurements (upper limits of the 95% CI around zero motion unless otherwise stated) Group RSA method

Translations (mm) Transverse Longitudinal Sagittal

All-polyethylene (n = 26 double examinations) Marker-based 0.11 Metal-backed (n = 21 double examinations) Marker-based 0.07 Model-based 0.08 Polyethylene micromotion 0.06

Rotations (°) Transverse Longitudinal Sagittal

MTPM (mm) Mean Upper limit of CI

0.15

0.09

0.17

0.11

0.14

0.11 0.11 0.06

0.13 0.13 0.11

0.08 0.19 0.14

0.14 0.64 0.68

0.11 0.15 0.16

0.12 0.25 0.19

0.11 0.32 0.30

Table 2. Differences between marker-based and model-based translations and rotations with the reference origin fixed at the geometric center of the polyethylene markers Factor

Translations (mm) Transverse Longitudinal Sagittal

Mean (SD) –0.01 (0.05) –0.03 (0.05) –0.05 (0.10) 95% CI a –0.11 to 0.09 –0.12 to 0.07 –0.25 to 0.16

Rotations (°) Transverse Longitudinal

Sagittal

–0.02 (0.13) 0.09 (0.29) –0.06 (0.18) –0.28 to 0.24 –0.48 to 0.66 –0.41 to 0.29

MTPM (mm) –0.03 (0.21) –0.45 to 0.39

a The

values represent the limits of agreement (interchangeability) between the 2 methods (Bland and Altman) and are based on all (n = 28) patients.

Figure 3. Scatter-plots showing (a) that marker-based transverse translation values are generally larger than model-based values due to the difference in position of the geometric center (which is either in the geometric center of the markers inserted in the polyethylene or in the geometric center of the model), also indicated (in b) by the proportional bias observed in the Bland–Altman plot (i.e., the difference between methods is proportional to the level of the measured variable) (Ludbrook 1997). (c) If model-based analysis is performed with the reference origin fixed at the geometric center of the polyethylene markers, results are nearly identical between methods, as also indicated (in d) by the absence of proportional bias and the small limits of agreement in the Bland–Altman plot. Solid lines in a and c: regression line. Dashed lines in a and c: line of equality. Solid horizontal lines in b and d: mean of differences. Dashed horizontal lines in b and d: 95% limits of agreement.

parison of marker-based and model-based translations, the reference origin for the model-based analysis was thus fixed at the geometric center of the polyethylene markers for the remaining analyses described below. This resolved the proportional bias (shown in Figure 3b and absent in Figure 3d) (Ludbrook 1997). Comparing marker-based with model-based RSA, translations showed small limits of agreement indicating that both methods can be used interchangeably (Table 2). The limits of agreement for the rotations and MTPM were larger, especially for rotations about the longitudinal axis (Table 2).

Micromotion of the polyethylene insert with respect to the metal tray Boxplots were constructed to investigate micromotion of the polyethylene insert with respect to the metal tray along and about each orthogonal axis at 3, 12, and 24 months’ followup (Figure 4). The majority of the measurements were within the 95% confidence interval of zero motion (i.e., the precision of the RSA method, indicated by the shaded areas in Figure 4) and group median values did not appear to increase over time. A few outliers depicted in Figure 4 were evaluated to determine the nature of the extreme values, all of which were

370

Acta Orthopaedica 2019; 90 (4): 366–372

MTPM (mm) 1.2 All-polyethylene Metal-backed marker-based Metal-backed model-based

1.0 0.8 0.6 0.4 0.2 0.0 0

Months after operation

Figure 5. RSA analysis results of maximum total point motion (MTPM). The mean and 95% confidence interval for the metal-backed group is shown for both the marker-based (greeen line) as well as the modelbased (blue line) analysis.

the tibial component or by another marker; slightly adjusting the position of these markers resulted in values close to zero in all directions.

Figure 4. Box-and-whisker plots showing the polyethylene insert translations and rotations with respect to the metal tray at each follow-up (n = 28). The line in boxes indicate group median, the box the interquartile range (IQR); the whiskers the maximum values and outliers are depicted as circles (> 1.5×IQR) and stars (> 3×IQR). Shaded blue areas represent the 95% confidence intervals of zero motion (i.e., RSA precision, determined with double examinations), numbers of the outliers are patient study numbers.

found to be due to measurement error as a result of instability or occlusion of the polyethylene markers. The error of patient 6 was due to one polyethylene marker moving posteriorly close to the periphery of the drilled hole where it was inserted (resulting in a mean error between 0.31 and 0.33 at 3, 12, and 24 months, close to the limit of 0.35). This marker stabilized within 3 months, as the polyethylene micromotion values were close to zero when 3 months’ follow-up was taken as the reference (mean error between 0.02 and 0.03 at 12 and 24 months). A similar cause was found in the analysis of patient 58, but in this case 2 anterior markers moved anteriorly; results were also close to zero when 3 months’ follow-up was taken as the reference. In the analysis of patient 22, patient 32, and patient 40, only 3 polyethylene markers were available of which 1 was partly occluded in 1 or more follow-up moments by either

Change in results of original trial When repeating the analysis of the primary outcome (MTPM after 2 years of follow-up) of the original report (van Hamersveld et al. 2018) with the model-based migration values, comparable group differences were found: the all-polyethylene group had an MTPM (CI) of 0.61 (0.49–0.74) versus 0.81 (0.68–0.95) for the marker-based metal-backed group; and versus 0.82 (0.68–0.96) for the model-based metal-backed group (Figure 5). In the original paper, continuous migration of ≥ 0.2 mm in MTPM in the second postoperative year was seen in 4 components in both groups. These 4 individual components of the metal-backed group showed similar migration patterns using model-based analysis (i.e., continuous migration in the second postoperative year). However, 2 additional metal-backed components showed continuous migration based on the modelbased analysis. In both cases, the increase in MTPM in the second postoperative year was likely the result of a sudden increase in rotation about the longitudinal axis due to modelfit inaccuracies, as all other parameters remained stable (data not shown). The other RSA parameters showed comparable betweengroup results when repeating the analysis with model-based migration values, except for translations along and rotations about the longitudinal (y-)axis, again, due to model-fit inaccuracies (Table 3, see Supplementary data). In this trial, 2 surgeons performed the surgeries. When stratifying the results by surgeon as performed in the post hoc sensitivity analysis of the original report, the observed difference in MTPM in favor of the all-polyethylene design was smaller and not statistically significant. Repeating this sensitivity analysis with the model-based measurements resulted in similar conclusions (Table 4, see Supplementary data).

Acta Orthopaedica 2019; 90 (4): 366–372

Discussion We investigated whether model-based RSA, utilizing a different reference origin as compared with marker-based RSA, would affect migration outcomes. By doing so, we were also able to quantify movements of the polyethylene insert within the locking mechanism of the Triathlon metal tray and explore the disadvantages of each RSA method. If the results differed systematically, pooling and comparing RSA data from studies using different RSA techniques would be impaired unless adjusted for the methods being used. However, if the insert moves with respect to the metal tray in modular TKA, then marker-based migration values of the tibial component in the transverse plane are unreliable (Nilsson et al. 2003), and likely produce random error that cannot be corrected for when comparing with model-based RSA studies. Now that an increasing number of RSA studies are available with long-term followup, meta-analysis becomes possible—but one must ascertain pooling of data is justified when different RSA methods have been used. Our study demonstrated systematic differences in translations but not rotations between model-based RSA and markerbased RSA. These differences are caused by the difference in reference origin that is used for migration calculation (Hurschler et al. 2009). As compared with the tibia 3D surface model, the origin in the center of the polyethylene markers overestimated the model-based transverse, longitudinal, and sagittal translations of the tibial component by 29%, 7%, and 26%, respectively. Correcting for this proportional bias, by using a factor or by using the same reference coordinate system in both analysis methods, resulted in nearly identical translations between model-based and marker-based analysis. For the rotations and MTPM values, the limits of agreement between marker-based and model-based RSA were larger because of the reduced precision of model-based rotations, particularly about the longitudinal axis. This is known and due to the relatively round, symmetrical shape of the tibial component in the transverse plane (Kaptein et al. 2007). Still, the limits of agreement between methods were within ± 0.5 mm and ± 0.8° and conclusions on the primary outcome of the RCT regarding group differences in MTPM remained unchanged. Furthermore, we found no evidence for the presence of insert micromotion and excluded this as a cause of unreliable marker-based migration measurements for the modular Triathlon TKA system. For the individual patient, however, use of a different method may result in substantial differences due to various types of measurement error (e.g., marker occlusion and model-fit inaccuracies). Therefore, one must not put too much weight on strict migration thresholds in individual patients (e.g., 0.2 mm of MTPM migration in the second postoperative year). Our findings are in line with an earlier comparison between marker-based and model-based RSA (Hurschler et al. 2009).

371

However, in that study, among other methodological differences, a uniplanar RSA setup was used resulting in marked differences in accuracy between “in-plane” and “out-ofplane” translations and rotations. In the present study we used a biplanar technique, and we did not find such a dichotomy in accuracy. Nevertheless, our findings further support their conclusion that model-based RSA can be used interchangeably with marker-based RSA, at least for the Triathlon TKA, provided that the same reference origin is used or corrected for using a factor when analyzing translations. Previous studies evaluating insert micromotion relative to the metal tray in modular TKAs found small movements in Nuffield fixed-bearing TKAs (Corin Medical Ltd., UK) (Hansson et al. 2005) and NexGen fixed-bearing TKAs (Zimmer, USA) (Nilsson et al. 2003). In the latter study, these movements were closely examined and found to be greater in the transverse plane, which corresponds to the polyethylene–metal tray interface (Nilsson et al. 2003). This contrasts with our results and may be explained by the different designs of the locking mechanisms that were used. In a recent retrieval study of Łapaj et al. (2017), backside damage as a result of abrasion following micromotion of the polyethylene was found in designs with dovetail locking mechanisms, especially in the NexGen trays. Contrarily, they found no evidence for abrasion in the Triathlon knees owing to the full peripheral capture locking mechanism. Furthermore, the anti-rotational central island of the Triathlon design has been shown to effectively reduce micromotion to a minimum for a given reacted torque as compared with other TKA designs, including NexGen (Bhimji et al. 2010), although this mechanical study was performed by the research and development department of Stryker. It should be noted, however, that random error as a result of the reduced precision of model-based RSA limits firm conclusions on the presence of (longitudinal) rotations of the polyethylene within the locking mechanism. Nevertheless, the found translations were minimal and all outliers were found to be caused by polyethylene marker instability or occlusion, thus unlikely to be the result of micromotion in the polyethylene–metal tray interface. A limitation of this study is that we compared the results of only one tibial component design. As the precision of modelbased RSA depends on the shape and accuracy of the fitted model (Kaptein et al. 2003), differences between markerbased and model-based RSA results may be smaller or larger depending on the TKA design and also depending on the location of the prosthesis markers, either in the insert, or attached to the metal tibial component. In summary, systematic differences in translations between marker-based and model-based RSA analysis disappeared when adjusted for the different reference origins being used for migration calculations. Micromotions of the polyethylene insert within the Triathlon metal tray were at most negligibly small. Mean migration results of model-based and markerbased measurements were comparable between groups when using the same reference origin, even though migration pat-

372

terns of individual patients may differ between RSA methods due to various types of measurement error. Supplementary data Tables 3–4 and the Appendix are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2019.1605692 The study was designed and coordinated by KH and STL. Data collection was performed by KH. Statistical analysis was done by KH, PM, and BK. KH, PM, LK, RN, STL, and BK interpreted the data and wrote the initial draft manuscript. All authors critically revised and approved the manuscript. The authors are grateful to Håkan Leijon for providing the marker-based RSA measurements and for his valuable help in performing the additional model-based RSA measurements. Acta thanks Raed Itayem, Maiken Stilling and Lars Weidenhielm for help with peer review of this study.

Bhimji S, Wang A, Schmalzried T. Tibial insert micromotion of various total knee arthroplasty devices. J Knee Surg 2010; 23(3): 153-62. Bland J M, Altman D G. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1(8476): 307-10. Gudnason A, Adalberth G, Nilsson K G, Hailer N P. Tibial component rotation around the transverse axis measured by radiostereometry predicts aseptic loosening better than maximal total point motion. Acta Orthop 2017; 88(3): 282-7. Hansson U, Toksvig-Larsen S, Jorn L P, Ryd L. Mobile vs. fixed meniscal bearing in total knee replacement: a randomised radiostereometric study. The Knee 2005; 12(6): 414-18.

Acta Orthopaedica 2019; 90 (4): 366–372

Hurschler C, Seehaus F, Emmerich J, Kaptein B L, Windhagen H. Comparison of the model-based and marker-based Roentgen stereophotogrammetry methods in a typical clinical setting. J Arthroplasty 2009; 24(4): 594-606. ISO 16087:2013(E). Implants for surgery: Roentgen stereophotogrammetric analysis for the assessment of migration of orthopaedic implants. Geneva: International Organization for Standardization; 2013. Kaptein B L, Valstar E R, Stoel B C, Rozing P M, Reiber J H. A new modelbased RSA method validated using CAD models and models from reversed engineering. J Biomech 2003; 36(6): 873-82. Kaptein B L, Valstar E R, Stoel B C, Reiber H C, Nelissen R G. Clinical validation of model-based RSA for a total knee prosthesis. Clin Orthop Relat Res 2007; 464: 205-9. Łapaj Ł, Mróz A, Kokoszka P, Markuszewski J, Wendland J, Helak-Łapaj C, et al. Peripheral snap-fit locking mechanisms and smooth surface finish of tibial trays reduce backside wear in fixed-bearing total knee arthroplasty. Acta Orthop 2017; 88(1): 62-9. Ludbrook J. Comparing methods of measurements. Clin Exp Pharmacol Physiol 1997; 24(2): 193-203. Nelissen R G, Pijls B G, Karrholm J, Malchau H, Nieuwenhuijse M J, Valstar E R. RSA and registries: the quest for phased introduction of new implants. J Bone Joint Surg Am 2011; 93(Suppl. 3): 62-5. Nilsson K G, Henricson A, Dalen T. In vivo determination of modular tibial insert micromotion. Trans Orthop Res Soc 2003; 28: 1402. Pijls B G, Plevier J W M, Nelissen R. RSA migration of total knee replacements. Acta Orthop 2018; 89(3): 320-8. Ryd L, Albrektsson B E, Carlsson L, Dansgard F, Herberts P, Lindstrand A, et al. Roentgen stereophotogrammetric analysis as a predictor of mechanical loosening of knee prostheses. J Bone Joint Surg Br 1995; 77(3): 377-83. Valstar E R, Gill R, Ryd L, Flivik G, Borlin N, Karrholm J. Guidelines for standardization of radiostereometry (RSA) of implants. Acta Orthop 2005; 76(4): 563-72. van Hamersveld K T, Marang-van de Mheen P J, Nelissen R G H H, ToksvigLarsen S. Migration of all-polyethylene compared with metal-backed tibial components in cemented total knee arthroplasty: a randomized controlled trial. Acta Orthop 2018; 89(4): 412-7.

Acta Orthopaedica 2019; 90 (4): 373–376

373

Inadequate evaluation and management of suspected infections after TKA surgery in Lithuania: a retrospective study of 2,769 patients with 2-year follow-up Egle TERTELIENE 1, Kazimieras GRIGAITIS 2, Otto ROBERTSSON 3, Justinas STUCINSKAS 2, Sarunas TARASEVICIUS 2, Narunas PORVANECKAS 1, and Algirdas VENALIS 1,4 1 Vilnius University, Faculty of Medicine, Vilnius, Lithuania; 2 Department of Orthopedics, Medical Academy, Lithuanian University of Health Sciences, Kaunas, Lithuania; 3 Department of Clinical Sciences and Department of Orthopedics, Lund University and Lund University Hospital, Lund, Sweden; 4 State Research Institute Center for Innovative Medicine, Vilnius, Lithuania Correspondence: egle.terteliene@gmail.com Submitted 2018-12-14. Accepted 2019-04-02.

Background and purpose — The evidence-based algorithms for treatment of periprosthetic joint infection (PJI) recommend surgical intervention in combination with the use of systemic antibiotics. However, still it is not unusual to treat total knee arthroplasty (TKA) patients with suspected infection using only antibiotics. We investigated treatment pathways for TKA patients with suspected infection in Lithuania. Patients and methods — Of the 4,069 TKA patients (4,269 knees) registered in the Lithuanian Arthroplasty Register (2013–2015) 2,769 patients (2,825 knees) were interviewed 2 years after the surgery. The patients were asked if they had been subject to antibiotic treatment after the TKA surgery and/or if any additional surgical interventions on the operated knee had been performed. The number of patients treated with antibiotics due to problems in the operated knee was identified and cumulative revision rates (CRR) were calculated. Results — 180 (7%) patients of the total 2,769 reported that they had been prescribed antibiotics after the primary TKA; 132 of these patients (70%) said they had received antibiotics due to problems with the operated knee. The 2-year CRR after TKA in patients not treated with antibiotics was 0.7% (95% CI 0.4–1), as compared with 24% (95% CI 17–32) in those who had used antibiotics due to the problems in the operated knee for more than 1 week. Interpretation — In Lithuania there seems to be a lack of adherence to evidence-based treatment guidelines when infection is suspected after primary TKA.

Periprosthetic joint infection (PJI) after total knee arthroplasty (TKA) is recognized as the most frequent reason for revisions, especially in the early postoperative stage (Kurtz et al. 2010). Most studies report a 1–2% incidence of PJI about after primary TKA (Peersman et al. 2001, Phillips et al. 2006, Kurtz et al. 2010, Matsen Ko et al. 2016). Accurate and early diagnosis of postoperative PJI and adequate treatment is the key to success. Currently, the evidence-based algorithms concerning the diagnosis and treatment of periprosthetic joint infections of the hip and knee indicate that only surgical treatment such as a debridement, antibiotics, irrigation, and retention of the prosthesis (DAIR) procedure or a 1- or 2-stage revision combined with systemic antibiotic treatment is to be recommended (Azzam et al. 2010, Parvizi et al. 2010, Osmon et al. 2013, Ghanem et al. 2014, Frank et al. 2017, Grammatopoulos et al. 2017). However, in “real life” some patients are still prescribed antibiotics without having surgical intervention in the hope that redness, tenderness, or wound leakage is not a serious infection and that surgical intervention can be avoided (Wagenaar et al. 2017). However, such usage of antibiotics may lead to increased bacterial resistance and more complicated treatment of an infected prosthesis, where matured biofilm on the prosthetic surface can no longer be eradicated with antibiotics only (Bjarnsholt et al. 2013). We evaluated how suspected infection after TKA was treated in “real life” in Lithuania with respect to adherence to guidelines, and investigated the outcome of antibiotic treatment without surgical intervention.

374

Acta Orthopaedica 2019; 90 (4): 373–376

Patients with TKA registered in the Lithuanian Arthroplasty Register September 2013 – September 2015 n = 4,069 (4,269 knees) Excluded (n = 1,300): – contact information not provided by hospitals, 533 – wrong contact information, 438 – dead before end of follow-up period, 67 – refused to participate, 262 Patients participating in the survey n = 2,769 (2,825 knees)

Table 1. Outcome of survey in TKA patients who received antibiotics due to problems with the operated knee (Group 1). Values are frequency (%) Received antibiotics due to problems with the operated knee (Group 1) Number of TKA patients/total no. of patients As prophylaxis As treatment Antibiotics prescribed by orthopedic surgeon Antibiotics for more than 1 month Diagnostic knee aspiration performed

Patients (%) 132/2,769 (4.8) 68/132 (52) 64/132 (48) 96/132 (73) 34/132 (26) 32/132 (24)

Figure 1. Description of material and patients interviewed regarding their use of antibiotics after surgery.

Patients and methods Data on patients having primary TKA procedures was derived from the Lithuanian Arthroplasty Register (LAR) (Tarasevicius et al. 2014) in order to be able to contact operated patients with an inquiry regarding their use of antibiotics during the first 2 years after the primary procedure. The completeness in the LAR was investigated in 2016, by comparing the register with State Patients fund data, and was 95% for primary TKA and 98% for revisions. 4,269 primary TKAs operated in 22 hospitals were registered in LAR between September 1, 2013 and September 1, 2015. 2,825 TKAs (2,769 patients) were included in the study (Figure 1). The patients were approached by 1 of the researchers at 2 years after the primary TKA. The following questions were asked: Have you used an antibiotic after your primary TKA? When did you start using antibiotics? For how long did you use antibiotics? What was the reason for the antibiotic’s usage? Who prescribed the antibiotics? Patients who responded as having used antibiotics for problems in the operated knee were additionally asked if they had been the subject of puncture. Finally, we asked whether the respondents had undergone revision at any time during the 2 years after the primary TKA. After the interview the hospital that had performed the procedure was asked to provide the relevant medical charts to ascertain that the additional surgery performed was a true revision according to the LAR definition. Revision in the LAR was defined as addition, exchange, or removal of 1 or all components. The patients were divided into 3 groups. Group 1 comprised those who received antibiotic treatment because of problems with their knee for a period of more than 1 week during the first 2 years after the primary TKA. Group 2 included those who received antibiotic treatment for more than 1 week due to problems not related to the operated knee and Group 3 patients were those having not had antibiotic treatment or who had treatment for 7 days or less.

Statistics For descriptive statistics, we used frequencies and ranges. Statistical evaluation included 95% confidence intervals (CI). The cumulative revision rate (CRR) was calculated with Kaplan– Meyer statistics and graphs plotted with CI for all groups; a p-value < 0.05 was considered significant. STATA v13 (StataCorp 2013) was used for calculations. Ethics, funding, and potential conflicts of interests The study was approved by the national ethical committee (No. 158200-16-832-371, approved on 2016-06-15). No funding were received to conduct the study and no conflict of interests needs to be declared.

Results 188 (7%) of 2,769 patients responded “yes” to the question: “Have you used antibiotics after the primary TKA?” When asked for the reason why antibiotics had been prescribed, 132 (Group 1) of the 188 patients (70%) said they had received antibiotics due to problems with the operated knee, while 56 (Group 2) (30%) had received the antibiotics for reasons other than the operated knee (pneumonia, bronchitis, urinary tract infection, tonsillitis). Of the 132 patients (Group 1), 68 (52%) reported that the reason for the antibiotic treatment had been infection prophylaxis, while the remaining 64 patients (49%) reported that the reason for the treatment had been that the physician had suspected a prosthetic joint infection (redness, pain, swelling of the operated knee, wound leakage). Patients receiving antibiotic treatment either for prophylaxis or due to suspected infection did not differ significantly from nonantibiotic users’ TKA with regard to their age, sex, and preoperative diagnosis. Among those 132 TKA patients who were prescribed antibiotics because of knee problems the prescribing physician was an orthopedic surgeon in 96 cases (73%) and 34 (26%) reported having used antibiotics for more than 1 month. Of the patients in Group 1, 32 reported that they had had a knee aspiration. Of these, 23 were subsequently revised, 21 because of infection. 100 of the patients in group 1 were not aspirated (Table 1).

Acta Orthopaedica 2019; 90 (4): 373–376

375

Table 2. Reasons for revision in Groups 1, 2, and 3. Values are frequency (%) Revision diagnosis

Group 1 n = 132

Group 2 Group 3 n = 56 n = 2,581

Infection 22 (17) 0 3 (0.1) Loosening of tibial component 5 (3.8) 0 0 Dislocation of the patella 0 0 1 (0.04) Pain in the patella 0 0 5 (0.2) Pain for unknown reason 1 (0.8) 0 2 (0.08) Limited range of motion 0 0 3 (0.1) Loosening of femoral component 1 (0.8) 0 2 (0.08) Instability 1 (0.8) 0 2 (0.08) Technical mistake in TKA 0 0 3 (0.1) Other reasons 2 (1.5) 0 3 (0.1) Total 32 (24) 0 23 (0.9) Group 1: Patients prescribed antibiotics due to problems in the operated knee. Group 2: Patients prescribed antibiotics for other reasons. Group 3: Non-antibiotic group.

Among the 132 patients who had antibiotic treatment (Group 1), 32 had been subject to revision surgery within 2 years of the primary operation. None had been revised in Group 2 (using antibiotics for other reasons) and 23 patients among the 2,581 (Group 3) who reported no antibiotic usage had undergone revision. The reason for revision was infection in 22 patients in Group 1 and in 3 among the non-antibiotic users (Group 3) (Table 2). The 2-year CRR after TKA in antibiotic users due to problems in the operated knee (Group 1) was 24% (95% CI 17–32) as compared with 0.7% (95% CI 0.4–1) among the no-antibiotics group (Group 3) (Figure 2).

Discussion Our results showed that 188 of the 2,769 TKA patients reported that they used antibiotics for more than 1 week, within 2 years after the primary procedure, and 132 of these antibiotics users reported that this was due to problems in operated knee. There are only a few reports in the literature investigating the success rate in curing periprosthetic infection using antibiotic therapy alone. Pavoni et al. (2004) used a non-operative approach to treat 34 patients with prosthetic joint infection (12 patients with early, 16 with delayed, and 6 with late infection). Most of the infections were initially treated with intravenous or intramuscular teicoplanin ± ciprofloxacin or rifampicin, followed by oral ciprofloxacin or minocycline plus rifampicin. 3 patients did not respond to therapy, and the infection was initially controlled in the remaining 31 patients. However, after longer follow-up (up to 5 years) less than half of the infected patients remained unrevised. In another study, Drancourt et al. (1997) reported a success rate of 52% for hips and 73% for knees when treating periprosthetic infection with a combination of antibiotics only, but the follow-up was short (up to 1

Figure 2. Cumulative revision rate of TKA due to infection in antibiotics users for reasons related to operated knee (Group 1), antibiotics users for other reasons (Group 2) and non-antibiotics group (Group 3).

year after the therapy). Further, Drancourt et al. (1997) found that fusidic acid plus rifampicin cured 11 of 21 hip prosthesis infections and 8 of 11 knee prosthesis infections; in only 5 of 19 cured cases was removal of the device necessary. However, these studies are small, the success rate is not impressive, and they were performed before guidelines/consensus concerning the diagnosis and treatment of periprosthetic joint infections became commonly accepted. There are no national guidelines regarding treatment of PJI in Lithuania; however, in the orthopedic departments dealing with PJI there is substantial knowledge on the topic, which is used as a basis for treatment decisions. According to the guidelines, the strategy in the treatment of PJI should be surgical intervention in combination with systemic antibiotics and not antibiotic treatment alone. These treatment pathways should be considered as a “gold standard” in the orthopedic community, but our study showed that this was not the case in Lithuania. Among the 132 TKA patients being treated with antibiotics because of problems with their knee, an orthopedic surgeon was the prescriber in 96 of the cases (74%). Considering our finding that only 24% of the patients receiving antibiotics for more than 1 week became subject to revision within 2 years, it is probable that at least some of the patients did not have a true PJI because otherwise it is unlikely that 74% had escaped further surgery. Of 100 unrevised patients who received antibiotics for more than 1 week, only 9 had been subject to knee aspiration and cultures. That more than 1 week of antibiotic use must be considered treatment but not prophylaxis shows that antibiotics treatment was prescribed without relevant evaluation. The problem is that antibiotic therapy without proper diagnosis of a PJI, inclusive of cultures, not only reduces the possibility of choosing proper surgical and antibiotic treatment but also risks exposing patients to the wrong or unnecessary treatment and increasing bacterial resistance.

376

Of the 132 TKAs with signs of infection, only 22 had a subsequent revision for infection, while the other 110 remained unrevised. One can speculate that 110 knees might have avoided revision surgery. On the other hand, those not cured may have developed more severe infection, with resistant bacteria requiring more extensive surgery. Furthermore, widespread empiric treatment with broad-spectrum antibiotics will probably have a disadvantageous environmental impact, despite few patients being saved from revision surgery (Inabathula et al. 2018). Thus, targeted antibiotic therapy based on proper bacterial sampling is an essential part of appropriate treatment of PJI (Parvizi et al. 2011). A drawback of our study is that we could not approach all the patients registered in the LAR. However, the proportion of interviewed patients was around 70% of the total number, which is why we assume that the results are a reasonable representation of the situation in Lithuania. Another drawback is that the follow-up was only 2 years, as some patients with PJI who were treated with antibiotics only might still have low-grade infection with low symptom expression, thus being unrevised but not cured when the study ended. We must bear in mind that it can be difficult to diagnose infection after TKA surgery, especially in non-hospital healthcare facilities. Thus, providing antibiotic treatment in the hope of the infection being “superficial” may be tempting, despite not being in accordance with widely accepted infection treatment protocols. This is also supported by Wagenaar et al. (2017) who made a questionnaire-based online evaluation of current Dutch orthopedic care for persistent wound leakage after joint arthroplasty. Among 127 orthopedic surgeon respondents, 57% used a protocol for diagnosis and treatment of persistent wound leakage although only 27% utilized the protocol in every patient. However, 24% of orthopedic surgeons prescribed antibiotics due to wound problems. This suggests that improper use of antibiotics is not only a Lithuanian problem. In summary, in Lithuania there seems to be a lack of adherence to evidence-based treatment guidelines when infection is suspected after primary TKA. By highlighting the problem and the spreading of information to both primary care and hospital staff the situation can be improved nationally and internationally. ET, ST, AV: conception of study, interpretation of data, and manuscript preparation. JS, NP: interpretation of data and manuscript preparation. OR, KG: statistical analyses, interpretation of data, and manuscript preparation. Acta thanks Trude Basso and Håvard Dale for help with peer review of this study. Azzam K A, Seeley M, Ghanem E, Austin M S, Purtill J J, Parvizi J. Irrigation and debridement in the management of prosthetic joint infection: traditional indications revisited. J Arthroplasty 2010; 25(7): 1022-7. doi: 10.1016/j.arth.2010.01.104.

Acta Orthopaedica 2019; 90 (4): 373–376

Bjarnsholt T, Ciofu O, Molin S, Givskov M, Hoiby N. Applying insights from biofilm biology to drug development: can a new approach be developed? Nat Rev Drug Discovery 2013; 12(10): 791-808. Drancourt M, Stein A, Argenson J N, Roiron R, Groulier P, Raoult D. Oral treatment of Staphylococcus spp. infected orthopaedic implants with fusidic acid or ofloxacin in combination with rifampicin. J Antimicrob Chemother 1997; 39(2): 235-40. Frank J M, Kayupov E, Moric M, Segreti J, Hansen E, Hartman C, Okroj K, Belden K, Roslund B, Silibovsky R, Parvizi J, Della Valle C J. The Mark Coventry, MD, Award: Oral antibiotics reduce reinfection after two-stage exchange: a multicenter, randomized controlled trial. Clin Orthop Relat Res 2017; 475(1): 56-61. Ghanem E, Heppert V, Spangehl M, Abraham J, Azzam K, Barnes L, Burgo F J, Ebeid W, Goyal N, Guerra E, Hitt K, Kallel S, Klein G, Kosashvili Y, Levine B, Matsen L, Morris M J, Purtill J J, Ranawat C, Sharkey P F, Sierra R, Stefansdottir A. Wound management. J Orthop Res 2014; 32(Suppl. 1): S108-19. Grammatopoulos G, Kendrick B, McNally M, Athanasou N A, Atkins B, McLardy-Smith P, Taylor A, Gundle R. Outcome following debridement, antibiotics, and implant retention in hip periprosthetic joint infection: an 18-year experience. J Arthroplasty 2017; 32(7): 2248-55. Inabathula A, Dilley J E, Ziemba-Davis M, Warth L C, Azzam K A, Ireland P H, Meneghini R M. Extended oral antibiotic prophylaxis in high-risk patients substantially reduces primary total hip and knee arthroplasty 90-day infection rate. J Bone Joint Surg Am 2018; 100(24): 2103-9. Kurtz S M, Ong K L, Lau E, Bozic K J, Berry D, Parvizi J. Prosthetic joint infection risk after TKA in the Medicare population. Clin Orthop Relat Res 2010; 468(1): 52-6. Matsen Ko L J, Yoo J Y, Maltenfort M, Hughes A, Smith E B, Sharkey P F. The effect of implementing a multimodal approach on the rates of periprosthetic joint infection after total joint arthroplasty. J Arthroplasty 2016; 31(2): 451-5. . Osmon D R, Berbari E F, Berendt A R, Lew D, Zimmerli W, Steckelberg J M, Rao N, Hanssen A, Wilson W R. Diagnosis and management of prosthetic joint infection: clinical practice guidelines by the Infectious Diseases Society of America. Clin Infect Dis 2013; 56(1): e1-e25. Parvizi J, Della Valle C J. AAOS Clinical Practice Guideline: diagnosis and treatment of periprosthetic joint infections of the hip and knee. J Am Acad Orthop Surg 2010; 18(12): 771-2. Parvizi J, Zmistowski B, Berbari E F, Bauer T W, Springer B D, Della Valle C J, Garvin K L, Mont M A, Wongworawat M D, Zalavras C G. New definition for periprosthetic joint infection: from the Workgroup of the Musculoskeletal Infection Society. Clin Orthop Relat Res 2011; 469(11): 2992-4. Pavoni G L, Giannella M, Falcone M, Scorzolini L, Liberatore M, Carlesimo B, Serra P, Venditti M. Conservative medical therapy of prosthetic joint infections: retrospective analysis of an 8-year experience. Clin Microbiol Infect 2004; 10(9): 831-7. Peersman G, Laskin R, Davis J, Peterson M. Infection in total knee replacement: a retrospective review of 6489 total knee replacements. Clin Orthop Relat Res 2001; (392): 15-23. Phillips J E, Crane T P, Noy M, Elliott T S, Grimer R J. The incidence of deep prosthetic infections in a specialist orthopaedic hospital: a 15-year prospective survey. J Bone Joint Surg Br 2006; 88(7): 943-8. StataCorp. Stata Statistical Software: Release 13. College Station, TX: StataCorp LP; 2013. Tarasevicius S, Cebatorius A, Valaviciene R, Stucinskas J, Leonas L, Robertsson O. First outcome results after total knee and hip replacement from the Lithuanian arthroplasty register. Medicina (Kaunas, Lithuania) 2014; 50(2): 87-91. Wagenaar F C, Lowik C A M, Stevens M, Bulstra S K, Pronk Y, van den Akker-Scheek I, Wouthuyzen-Bakker M, Nelissen R, Poolman R W, van der Weegen W, Jutte P C. Managing persistent wound leakage after total knee and hip arthroplasty: results of a nationwide survey among Dutch orthopaedic surgeons. J Bone Joint Infect 2017; 2(4): 202-7.

Acta Orthopaedica 2019; 90 (4): 377–382

377

An infrapatellar nerve block reduces knee pain in patients with chronic anterior knee pain after tibial nailing: a randomized, placebocontrolled trial in 34 patients Mandala S LELIVELD 1, Saskia J M KAMPHUIS 2, and Michael H J VERHOFSTAD 1 1 Trauma

Research Unit, Department of Surgery, Erasmus MC, University Medical Center Rotterdam, Rotterdam; 2 Department of Surgery, Gelderse Vallei Hospital, Ede, The Netherlands Correspondence: m.leliveld@erasmusmc.nl Submitted 2018-11-19. Accepted 2019-03-12.

Background and purpose — Anterior knee pain is common after tibial nailing. Its origin is poorly understood. Injury of the infrapatellar nerve is a possible cause. In this randomized controlled trial we compared changes in knee pain after an infrapatellar nerve block with lidocaine or placebo in patients with persistent knee pain after tibial nailing. Patients and methods — Patients with chronic knee pain after tibial nailing were randomized to an infrapatellar nerve block with 5 ml 2% lidocaine or placebo (sodium chloride 0.9%), after which they performed 8 daily activities. Before and after these activities, pain was recorded using a numeric rating scale (NRS; 0–10). Primary endpoint was the change in pain during kneeling after the infrapatellar nerve block. Secondary outcomes were changes in pain after the nerve block during the other activities. Results — 34 patients (age 18–62 years) were equally randomized. A significant reduction of the NRS for kneeling pain with an infrapatellar nerve block with lidocaine was found compared with placebo (–4.5 [range –10 to –1] versus –1 [–9 to 2]; p = 0.002). There were no differences between the treatments for the NRS values for pain during other activities. Interpretation — Compared with placebo, an infrapatellar nerve block with lidocaine was more effective in reducing pain during kneeling in patients with chronic knee pain after tibial nailing. Our findings support the contention that kneeling pain after tibial nailing is a peripheral nerve-related problem.

The common treatment for tibial shaft fractures is intramedullary nailing. A drawback of this procedure is anterior knee pain (Katsoulis et al. 2006). Persisting knee pain after more than 8 years post-nailing is reported with restrictions in daily and leisure activities (Lefaivre et al. 2008, Leliveld and Verhofstad 2012, Larsen et al. 2014). Removal of the nail does not alleviate pain in all patients and can even initiate anterior knee pain in some (Boerger et al. 1999). The cause of this phenomenon is unknown. Among the structures at risk for injury during tibial nailing through an infrapatellar approach is the infrapatellar branch of the saphenous nerve. Injury to this nerve usually results in numbness on the anterior aspect of the knee and the proximal lateral part of the lower leg. This complication has been reported after several other surgical procedures around the knee, such as knee arthroscopy (Mochida and Kikuchi 1995) and anterior cruciate ligament reconstruction (Spicer et al. 2000). In addition, development of postprocedural neuropathic pain has been described (Dellon et al. 1996). Since the infrapatellar nerve runs perpendicular to the patellar tendon, the nerve is at risk for transection during tibial nailing (Mochida and Kikuchi 1995, Kerver et al. 2013). Injury to the infrapatellar nerve after tibial nailing has been reported (Lefaivre et al. 2008, Leliveld and Verhofstad 2012), and sensory deficits of the infrapatellar nerve have been associated with chronic anterior knee pain after tibial nailing (Leliveld and Verhofstad 2012). However, studies to examine a causative relation between infrapatellar nerve injury and anterior knee pain after tibial nailing have not yet been conducted. We hypothesized that if knee pain after tibial nailing is indeed caused by neuropathic pain due to injury or entrapment of the infrapatellar nerve, an anesthetic block of this nerve with lidocaine will reduce knee pain in these patients.

378

Acta Orthopaedica 2019; 90 (4): 377–382

Patients treated with intramedullary nail 2000–2016 n = 601 Excluded (n = 194): – lost to follow-up (e.g. in other hospital), 39 – dead, 28 – age > 65 years, 70 – age < 18 years, 5 – ipsilateral fracture, 19 – pre-existing knee pain, 4 – no comprehension of Dutch language, 5 – wound on knee, 5 – intraarticular fracture (proximal tibia), 2 – wheelchair, 4 – amputation, 1 – history of complex regional pain syndrome, 3 – previous knee surgery, 9 Eligible patients n = 407

trial participation. These patients were sent a numeric rating scale (NRS) to rate knee anterior pain during 8 different daily activities (kneeling, squatting, prolonged sitting with bent knees, jumping, walking on stairs, running, walking, and rest). If the patient did not reply, telephone calls were attempted. Pain scores were returned by 233 patients. Criteria for inclusion in the trial were an NRS of 4–6 (moderate pain) during at least 3 out of 8 activities or an NRS of 7 or higher (severe pain) during 1 or more activities. 79 patients met these criteria, of whom 34 agreed to participate in the study.

Study design and assessment Eligible patients who agreed to participate in this study were seen at the outpatient clinics of the participating hospitals. Trauma characReturned VAS teristics, data concerning the initial procedure n = 233 and nail removal were gathered retrospecInclusion criteria: tively. Length of the longitudinal incision Patients with > 1 activity with VAS > 7 was measured on a flexed knee in mm using or > 3 activities with VAS > 4 a tape measure and localization of the inciRandomized sion was noted (on patellar tendon or medial n = 34 to patellar tendon). Sensory disturbances (numbness, hypesthesia, allodynia) in the Infrapatellar nerve block with lidocaine (n = 17) Infrapatellar nerve block with placebo (n = 17) Pain score (NRS) for 8 activities Pain score (NRS) for 8 activities area of the infrapatellar nerve (anterior and lateral aspect of the knee) were tested using Cross-over a cotton swab, comparing the non-operated Infrapatellar nerve block with placebo (n = 17) Infrapatellar nerve block with lidocaine (n = 17) leg with the operated leg and the surrounding Pain score (NRS) for 8 activities Pain score (NRS) for 8 activities dermatomas. Baseline pain (T0) was scored Patient selection, allocation and study design. using an NRS during 8 activities, which were all supervised by an examiner (SJMK or MSL). The NRS measures pain severity by asking the patient to select a number (from 0 to 10) to represent how severe the pain is, where 0 represents “no pain” and 10 represents “worst Patients and methods pain possible.” This rating scale has shown to be valid, reliPatients able, and appropriate for use in clinical practice (Williamson From the medical record systems and charting database and Hoggart 2005, Hjermstad et al. 2011), is responsive in patients between 18 and 65 years old, treated with an intra- patients with chronic nociceptive or neurogenic pain (Lundemedullary nail for an isolated traumatic unilateral tibial shaft berg et al. 2001), and the minimally clinical important change fracture (AO/OTA 42 A–C) between June 2000 and Decem- has been determined in patients with chronic musculoskeletal ber 2016, were selected from St Elisabeth Hospital (Tilburg, pain (Farrar et al. 2001, Salaffi et al. 2004). If a patient was The Netherlands, level 1 trauma center and teaching hospi- not able or willing to perform an activity it was noted as a tal), Gelderse Vallei Hospital (Ede, The Netherlands, level 2 missing value. Equal randomization of the treatment sequence was pertrauma center and teaching hospital), and Erasmus Medical Center (University Medical Center Rotterdam, The Nether- formed with use of a random-number generator. The allolands, level 1 trauma center and teaching hospital). Nailing or cated sequence was kept in sealed envelopes. Randomization nail removal had to be more than 6 months ago. 601 patients and preparation of the envelopes by a secretary who had no with a tibial shaft fracture were treated with an intramedullary involvement in the trial. Upon each patient’s enrollment into nail introduced through a longitudinal infrapatellar incision the study, the next consecutively numbered envelope was during the specified period (Figure). After application of the opened by an outpatient nurse. Lidocaine 2% and sodium chloexclusion criteria, 407 patients were potentially eligible for ride 0.9% (saline) were used for the nerve blocks. 2 syringes Excluded (n = 174): – lost to follow-up (e.g. moved), 25 – refused to participate, 12 – no response, 137

Acta Orthopaedica 2019; 90 (4): 377–382

379

Table 1. Baseline characteristics

ences between the 2 groups were tested using Student’s t-test (normal distribution) or Mann–Whitney U test (skewed distribution) for continuous variables. The method described by Hills and Armitage for two-period cross-over clinical trials was applied (Hills and Armitage 2004). Treatment effect was calculated with (T1–T0) – (T2–T1) for the lidocaine–placebo sequence group and (T2–T1) – (T1–T0) for the placebo–lidocaine sequence group. Period effect (the response to a treatment during the second period is not influenced by the treatment which was given during the first period) was calculated with (T1–T0) – (T2–T1) for both groups. Both treatment effect and period effect were analyzed using Mann–Whitney U test (Hills and Armitage 2004). In cases where a period effect was present, the results from the first nerve block only were analyzed. p < 0.05 was considered statistically significant. Data analysis was performed using SPSS version 25.0 for Windows (IBM Corp, Armonk, NY, USA).

Factor

Lidocaine Placebo group group (n = 17) (n = 17)

Male sex Age, median (range) AO/OTA fracture classification type A B C Months after tibial nailing, median (range) Tibia nail removed Length of longitudinal infrapatellar incision (mm), mean (SD) Placement of incision medial to patellar tendon midbundle of patellar tendon Sensory disturbance infrapatellar nerve

9 10 40 (22–62) 38 (18–60) 7 8 9 7 1 2 80 (6–168) 67 (11–168) 9 12 58 (17)

58 (12)

2 15 15

2 15 14

were prepared, marked with number 1 or 2 according to the allocation, and checked by a doctor not involved in the trial. The name and date of birth of the participant were written on the envelope. As both fluids were colorless and odorless, both patient and examiner remained unaware of which treatment was administered. An infrapatellar nerve block was performed freehand by depositing 5 mL of the solution in a fan-like manner subcutaneously between the medial surface of the medial femoral condyle and the medial aspect of the patellar tendon, including the incision. After 5 minutes patients completed the 8 activities under supervision and NRS scores were subsequently obtained (T1). Thereafter, each patient crossed over and was injected with the alternate treatment (T2). Time between injections (wash-out period) was approximately 30 minutes. Kneeling is the most frequently and painful activity reported after tibial nailing (Court-Brown et al. 1997, Toivanen et al. 2002, Cartwright-Terry et al. 2007, Vaisto et al. 2008). Therefore, the primary endpoint was the change in pain intensity during kneeling after infrapatellar nerve block with lidocaine and placebo, measured using an NRS. Secondary outcomes were changes in pain intensity after each nerve block as measured using an NRS during the 8 activities. Sample-size calculation A mean NRS of 7 for kneeling pain in patients with chronic anterior knee pain after tibial nailing was used for sample-size calculation (Salaffi et al. 2004). A change in pain intensity of > 30% was considered clinically meaningful (Farrar et al. 2001, Salaffi et al. 2004). Using a 2-sided test, an α level of 0.05, and a power of 80%, 34 patients were needed to be enrolled. Statistics Normally distributed continuous data are presented as mean (SD). Skewed data are presented as median (range). Differ-

Ethics, registration, funding, and potential conflicts of interest Approval was obtained from the central medical research ethics committee and the institutional board of all participating hospitals (NL34510.008.11/P1142 2016/07/20). The study was registered with the Dutch trial registry (NTR4628; Nederlands Trial Register; http://www.trialregister.nl). Written informed consent was obtained from all patients. Participants did not receive compensation of any kind. The study was not funded by any source. The authors have no competing interests to declare.

Results Patient characteristics Baseline demographics, length of the incision, placement of the incision, and sensory disturbances of the infrapatellar nerve are displayed in Table 1. Median age was 46 years (18–62). Median follow-up was 86 months (6–168). 85% of the patients showed signs of injury to the infrapatellar nerve (numbness, hypesthesia, or allodynia). Pain scores All patients received the infrapatellar nerve block according to group allocation. Pain scores at baseline (T0), after the first nerve block (T1), and after the second nerve block (T2) are displayed in Table 2. Kneeling was the most painful activity, followed by squatting. Some participants were not able or willing to perform all 8 activities. Treatment effects (decline in median pain scores) were significant for kneeling (p = 0.02), squatting (p = 0.03) and sitting with bent knees (p = 0.001). However, a period effect was present for the primary endpoint kneeling (Table 2), meaning the intervention exerted a different effect in the first period (T1–T0) than in the second period (T2–T1). We therefore

380

Acta Orthopaedica 2019; 90 (4): 377–382

Table 2. Change in pain scores after nerve block with lidocaine and placebo. Values are presented as median (range) Treatment Period Activity n a Lidocaine Placebo n a Placebo Lidocaine effect b effect b Kneeling Squatting Sitting with bent knees Jumping Walking on stairs Running Rest Walking

14 –4.5 (–10 to –1) 14 –3 (–9 to 1) 15 –3 (–7 to 0) 11 –1 (–6 to 2) 17 0 (–6 to 2) 11 0 (–5 to 2) 17 0 (–1 to 1) 17 0 (–5 to 1)

a Not all patients performed b Mann–Whitney U.

0 (–4 to 1) 0 (–4 to 9) 0 (–1 to 1) 0 (–1 to 1) 0 (–1 to 1) 0 (–3 to 8) 0 (–1 to 1) 0 (–1 to 2)

16 15 17 12 17 11 17 17

–1 (–9 to 2) –1 (–9 to 1) –2 (–6 to 0) –1 (–3 to 1) 0 (–5 to 2) 0 (–3 to 1) 0 (–1 to 6) 0 (–4 to 4)

–1.5 (–8 to 2) 0 (–7 to 2) 0 (–7 to 9) –1 (–3 to 1) 0 (–4 to 3) 0 (–5 to 0) 0 (–3 to 1) 0 (–5 to 3)

0.02 0.03 0.001 0.09 0.4 0.6 0.2 0.5

0.004 0.09 1.0 0.3 0.4 0.8 0.1 0.4

all activities.

chose to additionally analyze the results from the first nerve block (T1–T0), like a randomized trial comparing 2 groups. For kneeling a significant decline in median pain scores remained after a nerve block with lidocaine compared with placebo (–4.5 [–10 to –1] versus –1 [–9 to 2]; p = 0.002). There was no statistically difference between the groups during squatting, sitting with bent knees, jumping, walking stairs, running, walking, and rest (data not shown).

Discussion The purpose of this study was to compare changes in knee pain after a subcutaneous lidocaine block of the infrapatellar nerve or placebo in patients with chronic anterior knee pain after tibial nailing. For kneeling a significant reduction in pain scores was found after an infrapatellar nerve block with lidocaine. The effect of lidocaine usually lasts about 1–2 hours and with a wash-out period of about 30 minutes one can presume that the effect of the lidocaine injection persists during the second treatment period. We expected that pain scores in the lidocaine-first group would reach their utmost lowest levels after injection and only minimal changes would occur after the second injection with saline. Pain scores in the placebofirst group were expected to decline only minimally after the first injection and decline further after lidocaine injection; a difference in change scores would then still be observed. However, data analysis showed a period effect for the primary endpoint, meaning the effect of the treatment was different in the first period (T1–T0) from the effect in the second period (T2–T1). This can easily be explained by the short wash-out period. Also, both the patient and the examiner were blinded to the treatment given. Due to the local effect lidocaine has on the skin, patients may recognize this effect. This affects true blinding and might also have affected the pain scores. Although pain during kneeling was reduced in both groups, pain was not totally diminished and no statistically significant reduction was seen for pain during the other activities (squat-

ting, sitting with bent knees, jumping, running, walking on stairs, walking, and at rest). A possible explanation is that not all patients were able or willing to do these activities, which affected the statistical power. Moreover, the starting pain level was lower than in other activities, thus a smaller effect size can be expected. The study could be underpowered for these activities; however, they were not the primary outcome. In some patients pain can be multi-modal and might as well have originated from intra-articular injury (Hernigou and Cohen 2000) or irritation of Hoffa’s fat pad (Jankovic et al. 2013). Knee pain is a common complaint after intramedullary nailing for tibial shaft fractures. In this study 79 of 233 patients (34%) who returned their NRS indicated they had either moderate or severe knee pain during several activities after a median follow-up of 7.1 years (0.5–14). Although there might be some selection bias due to selective response to the initial questionnaire, this percentage is in concordance with the long-term results of Lefaivre et al (2008) and Leliveld and Verhofstad (2012), who respectively reported 29% and 38% of chronic knee pain after tibial nailing after a median follow-up of 14 and 7 years. Kneeling pain is frequently mentioned to be the most painful activity (Court-Brown et al. 1997, Toivanen et al. 2002, Cartwright-Terry et al. 2007, Vaisto et al. 2008). In a randomized trial comparing 2 different incisions from Toivanen et al. (2002), 62% of the patients stated kneeling pain as being most painful. The mean visual analogue score (0–100 mm) for kneeling pain in these patients was 31 mm (transtendinous approach) and 44 mm (paratendinous approach). In a retrospective study Court-Brown et al. (1997) even reported kneeling pain as the most painful activity in 92% of their patients, followed by squatting (61%). The average scores for these activities on a 10-point analogue scale were respectively 3.1 and 3.3. The fact that kneeling pain scores in our study are higher (median NRS of 8.0) is due to the fact that patients were selected based on their pain scores (scores of 4 or higher for at least 3 activities or 7 and higher for at least one activity). We detected sensory disturbances in the area of the infrapatellar nerve (anterior and lateral aspect of the knee) in 29/34

Acta Orthopaedica 2019; 90 (4): 377–382

of the patients. Iatrogenic injury to the infrapatellar nerve is one of many concepts regarding the origin of anterior knee pain after tibial nailing. The course of the infrapatellar nerve makes it susceptible to iatrogenic injury during nail insertion, especially when longitudinal infrapatellar medial and midline incisions are used (Kerver et al. 2013). Long-lasting sensory deficits at the anterior aspect of the knee are described after tibial nailing (Karladani et al. 2007, Lefaivre et al. 2008, Leliveld and Verhofstad 2012) and a correlation was found with anterior knee pain after tibial nailing (Leliveld and Verhofstad 2012). Nahabedian and Johnson (2001) performed a selective infrapatellar nerve denervation in 9 patients with chronic knee pain after blunt trauma to the knee and total knee replacement. Median pain scores (NRS 0–10) reduced from 8.0 (range 7–10) at baseline to 3.0 (range 0–6) after the denervation and they conclude that selective denervation is a beneficial procedure in selected patients with neuromatous knee pain. An infrapatellar nerve block with lidocaine in our study showed a significant difference in change of pain intensity during kneeling in patients treated with an intramedullary nail, compared with a nerve block with placebo. Because the infrapatellar nerve solely provides sensation of the skin at the antero-lateral aspect of the knee, an anesthetic block with lidocaine can diminish cutaneous neuropathic pain in this region (Nahabedian and Johnson 2001, Hsu et al. 2013) but not pain related to intraarticular injury. Although pain scores declined for all activities at the end of the study, actual function and effect on function were not assessed. Sudden improvement of functional outcome was, however, not expected in patients who sustained knee pain for several years. Long-term improvement in function after infrapatellar nerve block has been reported though (Hsu et al. 2013), as has long-term pain relief after denervation of the infrapatellar nerve (Dellon et al. 1996, Nahabedian and Johnson 2001). The incidence of persisting anterior knee pain after tibial nailing is high and we provide arguments for the hypothesis that iatrogenic injury to the infrapatellar nerve contributes to this problem. Patients suffering from this complication who response well to an infrapatellar nerve block with lidocaine might benefit from denervation (Dellon et al. 1996, Nahabedian et al. 1998, Nahabedian and Johnson 2001). Based on anatomical studies a transverse or oblique incision would yield the least chance of injury to or entrapment of the infrapatellar nerve (Mochida and Kikuchi 1995, Ebraheim and Mekhail 1997, Kerver et al. 2013). Alternatively, the suprapatellar approach for tibial nailing avoids the risk zone for infrapatellar nerve injury (Kerver et al. 2013) and studies have reported low knee pain scores and good functional results after this approach (Chan et al. 2016, Sun et al. 2016, Rothberg et al. 2019). In summary, compared with placebo, an infrapatellar nerve block with lidocaine was more effective in reducing pain during kneeling in patients with chronic knee pain after tibial

381

nailing through a longitudinal infrapatellar incision. Our data support the contention that kneeling pain after tibial nailing is a peripheral nerve-related problem.

The study was designed and initiated by MSL and MHJV. MSL and SJMK were responsible for identifying and contacting eligible patients. Patients were seen at the outpatient clinic by SJMK. MSL analyzed the data and drafted the manuscript. MHJV critically reviewed the manuscript. The authors would like to thank Dr E. Verhagen (Amsterdam Collaboration on Health and Safety in Sports, Public and Occupational Health, VU University Medical Center, Amsterdam, The Netherlands) for his contribution to the statistical analysis. Acta thanks Michael Brix and Jan Erik Madsen for help with peer review of this study.

Boerger T O, Patel G, Murphy J P. Is routine removal of intramedullary nails justified. Injury 1999; 30(2): 79-81. Cartwright-Terry M, Snow M, Nalwad H. The severity and prediction of anterior knee pain post tibial nail insertion. J Orthop Trauma 2007; 21(6): 381-5. Chan D S, Serrano-Riera R, Griffing R, Steverson B, Infante A, Watson D, Sagi H C, Sanders R W. Suprapatellar versus infrapatellar tibial nail insertion: a prospective randomized control pilot study. J Orthop Trauma 2016; 30(3): 130-4. Court-Brown C M, Gustilo T, Shaw A D. Knee pain after intramedullary tibial nailing: its incidence, etiology, and outcome. J Orthop Trauma 1997; 11(2): 103-5. Dellon A L, Mont M A, Mullick T, Hungerford D S. Partial denervation for persistent neuroma pain around the knee. Clin Orthop Relat Res 1996; (329): 216-22. Ebraheim N A, Mekhail A O. The infrapatellar branch of the saphenous nerve: an anatomic study. J Orthop Trauma 1997; 11(3): 195-9. Farrar J T, Young J P, Jr, LaMoreaux L, Werth J L, Poole R M. Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale. Pain 2001; 94(2): 149-58. Hernigou P, Cohen D. Proximal entry for intramedullary nailing of the tibia: the risk of unrecognised articular damage. J Bone Joint Surg Br 2000; 82(1): 33-41. Hills M, Armitage P. The two-period cross-over clinical trial 1979. Br J Clin Pharmacol 2004; 58(7): S703-16; discussion S717-19. Hjermstad M J, Fayers P M, Haugen D F, Caraceni A, Hanks G W, Loge J H, Fainsinger R, Aass N, Kaasa S. European Palliative Care Research Collaborative. Studies comparing numerical rating scales, verbal rating scales, and visual analogue scales for assessment of pain intensity in adults: a systematic literature review. J Pain Symptom Manage 2011; 41(6): 1073-93. Hsu L P, Oh S, Nuber G W, Doty R, Jr, Kendall M C, Gryzlo S, Nader A. Nerve block of the infrapatellar branch of the saphenous nerve in knee arthroscopy: a prospective, double-blinded, randomized, placebo-controlled trial. J Bone Joint Surg Am 2013; 95(16): 1465-72. Jankovic A, Korac Z, Bozic N B, Stedul I. Influence of knee flexion and atraumatic mobilisation of infrapatellar fat pad on incidence and severity of anterior knee pain after tibial nailing. Injury 2013; 44(Suppl 3): S33-9. Karladani A H, Ericsson P A, Granhed H, Karlsson L, Nyberg P. Tibial intramedullary nails: should they be removed? A retrospective study of 71 patients. Acta Orthop 2007; 78(5): 668-71. Katsoulis E, Court-Brown C, Giannoudis P V. Incidence and aetiology of anterior knee pain after intramedullary nailing of the femur and tibia. J Bone Joint Surg Br 2006; 88(5): 576-80.

382

Kerver A L, Leliveld M S, den Hartog D, Verhofstad M H, Kleinrensink G J. The surgical anatomy of the infrapatellar branch of the saphenous nerve in relation to incisions for anteromedial knee surgery. J Bone Joint Surg Am 2013; 95(23): 2119-25. Larsen P, Lund H, Laessoe U, Graven-Nielsen T, Rasmussen S. Restrictions in quality of life after intramedullary nailing of tibial shaft fracture: a retrospective follow-up study of 223 cases. J Orthop Trauma 2014; 28(9): 507-12. Lefaivre K A, Guy P, Chan H, Blachut P A. Long-term follow-up of tibial shaft fractures treated with intramedullary nailing. J Orthop Trauma 2008; 22(8): 525-9. Leliveld M S, Verhofstad M H. Injury to the infrapatellar branch of the saphenous nerve, a possible cause for anterior knee pain after tibial nailing? Injury 2012; 43(6): 779-83. Lundeberg T, Lund I, Dahlin L, Borg E, Gustafsson C, Sandin L, Rosen A, Kowalski J, Eriksson S V. Reliability and responsiveness of three different pain assessments. J Rehabil Med 2001; 33(6): 279-83. Mochida H, Kikuchi S. Injury to infrapatellar branch of saphenous nerve in arthroscopic knee surgery. Clin Orthop Relat Res 1995; (320): 88-94. Nahabedian M Y, Johnson C A. Operative management of neuromatous knee pain: patient selection and outcome. Ann Plast Surg 2001; 46(1): 15-22. Nahabedian M Y, Mont M A, Hungerford D S. Selective denervation of the knee: experience, case reports, and technical notes. Am J Knee Surg 1998; 11(3): 175-80.

Acta Orthopaedica 2019; 90 (4): 377â&#x20AC;&#x201C;382

Rothberg D L, Stuart A R, Presson A P, Haller J M, Higgins T F, Kubiak E N. A comparision of the open, semi-extended parapatellar versus standard entry tibial nailing techniques and knee pain: a randomized controlled trial. J Orthop Trauma 2019; 33(1): 31-6. Salaffi F, Stancati A, Silvestri C A, Ciapetti A, Grassi W. Minimal clinically important changes in chronic musculoskeletal pain intensity measured on a numerical rating scale. Eur J Pain 2004; 8(4): 283-91. Spicer D D, Blagg S E, Unwin A J, Allum R L. Anterior knee symptoms after four-strand hamstring tendon anterior cruciate ligament reconstruction. Knee Surg Sports Traumatol Arthrosc 2000; 8(5): 286-9. Sun Q, Nie X, Gong J, Wu J, Li R, Ge W, Cai M. The outcome comparison of the suprapatellar approach and infrapatellar approach for tibia intramedullary nailing. Int Orthop 2016; 40(12): 2611-17. Toivanen J A K, Vaisto O, Kannus P, Latvala K, Honkonen S E, Jarvinen M J. Anterior knee pain after intramedullary nailing of fractures of the tibial shaft: a prospective, randomized study comparing two different nail-insertion techniques. J Bone Jt Surg Ser A 2002; 84(4): 580-5. Vaisto O, Toivanen J, Kannus P, Jarvinen M. Anterior knee pain after intramedullary nailing of fractures of the tibial shaft: an eight-year follow-up of a prospective, randomized study comparing two different nail-insertion techniques. J Trauma 2008; 64(6): 1511-16. Williamson A, Hoggart B. Pain: a review of three commonly used pain rating scales. J Clin Nurs 2005; 14(7): 798-804.

Acta Orthopaedica 2019; 90 (4): 383–388

383

Non-union of the ulnar styloid process in children is common but long-term morbidity is rare: a population-based study with mean 11 years (9–15) follow-up Linda KORHONEN 1,2, Sarita VICTORZON 3, Willy SERLO 1,2, and Juha-Jaakko SINIKUMPU 1,2 1 Department

of Children and Adolescents, Pediatric Surgery and Orthopedics, Oulu University Hospital, Oulu; 2 Medical Research Centre Oulu; PEDEGO Research Group; Oulu Childhood Fracture and Sports Injury Study; University of Oulu, Oulu; 3 Department of Radiology, Vaasa Central Hospital, Vaasa, Finland Correspondence: korhonen.lindamaria@outlook.com Submitted 2018-06-19. Accepted 2019-02-25.

Background and purpose — Fracture of the ulnar styloid process (USP) is common in children in connection with distal radius fracture. The long-term morbidity of USP nonunion following a childhood distal radius fracture is unclear. We evaluated long-term clinical and radiographic findings of USP non-union. Patients and methods — All 208 children (< 16 years) who had suffered from distal radius fracture with or without a diagnosed concomitant ulnar fracture during 1992–1999 in the study institution were invited to follow-up at mean of 11 years (9–15) after the injury. Radiographs of both wrists of all 139 participants (67%) were taken; 22 patients showed USP non-union and they made up the study population. Distal radioulnar joint (DRUJ) instability, decreased range of motion (ROM), and weakened grip strength as compared with the uninjured side were the main functional outcomes. Elements of the “Disability of Arm, Shoulder and Hand” questionnaire were used for subjective symptoms. Results — The rate of USP non-union following childhood distal forearm fracture was 16% (22/139) and only 9 of the ulnar styloid fractures were visible in the radiographs primarily. At follow-up wrist flexion–extension ROM and ulnar and radial deviation ranges did not differ between the injured and uninjured sides. Grip strengths were similar. 6 patients reported pain during exercise. 7 had ulna minus (mean 2.3 mm) but none showed degenerative radiographic findings. Interpretation — The long-term clinical results of USP non-union following a childhood wrist fracture were good. However, one-third of the patients with USP non-union had ulnar shortening, which may predispose them to degenerative processes later in life.

The distal radius is the most common site of fracture in children and its incidence has increased during the past 40 years (Khosla et al. 2003, de Putter et al. 2011, Kazemian et al. 2011). Displacement in the fracture may be associated with interruption of the distal radioulnar ligaments and avulsion of the ulnar styloid process (USP). USP fracture is a common finding in connection with distal radius fracture in children (up to 30¬–50% of all cases) (Gogna et al. 2014, Wijffels et al. 2014). In adults, the reported rate of USP fracture in connection with distal radius fracture is even higher (Kramer et al. 2013). However, the distal ulnar epiphysis is cartilaginous and apparent not earlier than at the age of 5–9 years, with the result that USP fracture may be under-diagnosed in children (Bae and Waters 2006, Abid et al. 2008). A USP fracture seldom justifies operative treatment (Logan and Lindau 2008, Souer et al. 2009, Chen et al. 2013, Zoetsch et al. 2013), which is mostly based on the need for surgery of the fractured radius. Casting the wrist in ulnar inclination in order to minimize USP dislocation has been suggested as a treatment option. Regardless of treatment, USP fracture frequently fails to unite (Abid et al. 2008, Gogna et al. 2014) but the risk factors of non-union are unclear. There are several potential complications in USP nonunion. The triangular fibrocartilage complex (TFCC) and anatomic bone congruity are the main factors contributing to the stability of the distal radioulnar joint (DRUJ) (Kazemian et al. 2011) and even minor changes in ulnar length can change the axial loads on the TFCC (Bae and Waters 2006). Growth arrest resulting from distal radius fracture appears as ulnar lengthening (ulna plus) (Schuurman et al. 2001, Waters et al. 2002). Respectively, ulnar shortening is a result of

384

Acta Orthopaedica 2019; 90 (4): 383–388

Table 1. Basic characteristics of participants (n = 139) and nonparticipants (n = 69)

All children < 16 years with distal radius fracture 1992–1999 n = 208

Factor

Participants n = 139

Isolated distal radius fracture n = 93

n = 13

n=9

Ulnar styloid process non-union 11 year after initial injury n = 22

Distal radius fracture with concomitant ulna fracture n = 46

n = 80

n = 37

Intact ulnar styloid process, control patients n = 117

Figure 1. Enrollment of the patients.

growth arrest of ulna and it may result in TFCC degeneration and rupture (Nelson et al. 1984). Thus, one of the most disabling complications after distal radius fracture is instability in DRUJ (Daneshvar et al. 2014, Gogna et al. 2014). In addition, chronic ulnar sided wrist pain (Yuan et al. 2017) and higher Disabilities of the Arm, Shoulder and Hand (DASH) scoring have been reported; nevertheless, they are still slight enough to fall outside clinical importance in short-term follow-up (Kazemian et al. 2011, Kramer et al. 2013, Wijffels et al. 2014, Mulders et al. 2018). To our knowledge, most studies concerning the clinical importance of childhood USP non-union are based on short-term outcomes (Kramer et al. 2013, Gogna et al. 2014, Mulders et al. 2018) and the understanding of long-term outcomes is insufficient (Cannata et al. 2003). This study was performed to determine the rate of USP non-union and to evaluate the clinical and radiographic recovery of USP nonunion in children with former distal radius fracture in longterm follow-up.

Patients and methods Study design This is a population-based study including all children of < 16 years of age who had suffered from distal radius fracture with or without a concomitant USP fracture and had been treated at Vaasa Central Hospital, Finland, in 1992–1999. All these patients were identified in the hospital database and they were invited to a long-term follow-up visit by letter. In cases of no-show, another letter was sent and finally the correct postal address was ensured by a nurse via a phone call. 139 patients took part out of 208 enrolled (participation 67%), at a mean of 11 years (9–15) after the injury (Figure 1). 39 out of 69 nonparticipants reported on their own initiative that they have no symptoms/no reasons to participate, while the other 30 declared no single reason (Table 1).

Participants Non-participants (n = 139) (n = 69) p-value

Male sex, n (%) 82 (60) 48 (70) Age (SD) [range] 9.6 (3.4) [3–16] 10 (3.4) [2–16] Injury, left sided, n (%) 72 (52) 36 (52) Operation rate, n (%) 44 (32) 21 (30) Immobilization, weeks (SD) [range] 3.3 (0.8) [1–5] 3.4 (0.7) [2–4] a Standardized Normal Distribution b Independent samples t-test.

0.2 a 0.5 b 1.0 a 0.8 a 0.9 b

tests.

Original radiographs and hospital charts were re-reviewed to establish the baseline characteristics of the patients, injury mechanism, clinical findings, and primary treatment. The displacement and angular deformity of radius fractures were determined. The USP fractures were classified as base and tip fractures. Outcome variables The rate of USP non-union was evaluated. Distal RUJ instability, decreased range of motion (ROM), and grip strength as compared with the uninjured side were analyzed as primary outcome variables. Measurements were made with a universal goniometer and the results were reported as mean values (SDs). Grip strength was measured using a hydraulic Jamar grip dynamometer (Sammons Preston, Bolingbrook, IL, USA). The best of 3 attempts was recorded. Both injured and uninjured sides were examined. To evaluate current symptoms, baseline functional scores in DASH questionnaires were used and the patients were asked about pain, tolerance in physical activity, and daily-life-related complaints. Current symptoms were compared between the patients with USP nonunion and patients with absence of USP non-union. In addition, possible risk factors of USP non-union were evaluated. To evaluate the radiographic outcomes, both upper extremities were examined, except in cases of graviditas. The presence of USP non-union was recognized; its displacement (mm), and possible ulnar shortening (mm) or lengthening (mm) in the injured extremity were measured. Decreased joint space, osteophytes, and subchondral bone cysts were taken as signs of early degeneration. Patients The 139 patients were on average 9.6 years (3–16) of age at the time of the injury. 60% were boys and 93% right-handed. The most usual type of injury was a fall from < 1 m (58%) and 30% fell from > 1 m. 7% resulted from traffic accidents and 4% miscellaneous injuries were found. 3% suffered from open fracture, 2 of which were Gustilo type I, and 2 of type III. The fractures were complete in 12%. Based on the primary radiographs, major (≥ 15°) angular

Acta Orthopaedica 2019; 90 (4): 383–388

385

Table 2. Clinical findings in the injured vs. uninjured extremity in patients with UPS non-union 9–15 years after distal radius fracture (n = 22). Values are mean (SD) Factor

Injured

Wrist flexion Wrist extension Deviation, ulnar Deviation, radial Grip strength, Nm

80° (5.7) 80° (10) 32° (5.7) 20° (5.7) 47 (16)

a Independent

Figure 2. USP non-union 11 years after distal radius fracture.

deformity in the radius was found in 32%, slight deformity (< 15°) in 30% and no deformity in 37%. Anterior-posterior displacement of the radius was ≥ 2 mm in 26% and 0–2 mm in 6%, while coronal plane displacement (≥ 2 mm or 0–2 mm) was found in 8% and 4%, respectively. 3% of the fractures were comminuted. USP non-union was found in 16% (22/139) of the cases, of which 9 were originally diagnosed as “distal radius fracture with concomitant ulna fracture” and 13 were isolated radius fractures; the USP fracture was invisible in the primary radiographs. According to primary and long-term radiographs, 12 of the USP fractures were at the base and 10 at the tip. Statistics Statistical analysis included a chi-square test to evaluate differences between independent groups with categorical data; Fisher’s exact test was used with small groups (< 5 cases). Independent samples t-tests were used to compare differences between continuous variables. Differences of proportions were evaluated by using the binomial standardized normal deviate (SND) test. Binary logistic regression analysis was used to evaluate the potential predictive factors concerning the risk of non-union. Odds ratios (ORs) with their 95% confidence intervals (CIs) were determined in connection with age (per year of age), sex, severity of primary injury (> 2 mm displacement primarily or > 15° of angular deformity), concomitant ulnar fracture visible in radiographs, ulnar styloid fracture type (base vs. tip), open fracture (no/yes), operative treatment of radius fracture (no/yes), junior vs. senior operating surgeon and longer vs. shorter time of immobilization (≥ 28 or < 28 days) in order to identify the factors associated with USP nonunion. SPSS version 24 software (IBM Corp, Armonk, NY, USA) and Stats Direct Ltd. 2013 version 3.1 (Sale, Cheshire, UK) were used. Less than 5% was considered to be the relevant level of statistical significance (p < 0.05).

Uninjured p-value a 83° (5.1) 80° (10) 32° (5.7) 20° (5.7) 49 (15)

0.2 1.0 1.0 1.0 0.6

samples t-test.

Ethics, funding, and potential conflicts of interest The Ethics Board of Vaasa Central Hospital approved the study in advance (§175/2008). Signed informed consent documents were obtained from all of the participants. The research was performed in compliance with the Helsinki Declaration concerning ethical principles. Grants were obtained from both public national funding (VTR-funding) and nonprofit foundations. No conflicts of interest were declared.

Results Rate of USP non-union The rate of USP non-union (Figure 2) following childhood distal forearm fracture was 16% (22/139) at mean 11 years of follow-up. 13 of the 22 fractures that failed to unite had been invisible in the primary radiographs. In the subgroup of 12 cases whose USP fracture was visible in the primary radiographs, 9 did not heal. Radiographic findings 7 cases out of the 22 with non-union vs. none in the reference group (N = 117, p < 0.001) showed ulna minus (mean 2.3 mm) when compared with the uninjured side. Respective rate of premature growth plate arrest of the radius was 4% (5/139) and the mean ulna plus was 2.1 mm. None of the non-union patients showed degenerative radiographic findings. The mean displacement of the non-united fragment was 0.99 mm (SD 0.60, range 0.0–2.3 mm). 8 patients showed fragmented USP in the radiographs. Functional findings There was no decrease in wrist movement in patients with non-union, compared with the respective movement on the uninjured side. Grip strength did not differ between the injured and uninjured wrists (Table 2). 6 patients in the entire cohort showed clinically unstable DRUJ but the rates were similar between the non-union and union groups. 5 patients with USP non-union and 18 in the reference group presented crepitation in the radiocarpal joint (p = 0.4).

386

Acta Orthopaedica 2019; 90 (4): 383–388

Table 3. Symptoms found in patients with USP non-union vs. intact USP 11.4 years after distal radius fracture with or without concomitant USP fracture. Values are frequency Factor

Non-union Union (n = 22) (n = 117)

Crepitation Visible deformity Limitation in movement Symptoms in daily life Limb-length discrepancy Subjective dissatisfaction Pain on palpation a Chi-square

p-value a

5 18 0.4 2 0 0.02 0 2 1.0 5 21 0.6 0 0 N/A 1 2 0.4 1 10 1.0

test and Fisher’s exact test.

Subjective symptoms 6/22 cases in the non-union group vs. 21/117 in the reference group reported pain during exercise (p = 0.3). No statistically significant difference was found in subjective symptoms between the patients with or without USP non-union at longterm follow-up (Table 3). Effect of fracture type, treatment, and immobilization on the risk of ulnar styloid non-union Displacement of ≥ 2 mm (OR 1.6, CI 0.52–4.9), angular deformity of > 15° (OR 1.2, CI 0.40–4.1), and Salter–Harris classification 2 or more in the radius (OR 2.8, CI 0.4–18) did not increase the risk of non-union. Further, no increased risk was found in connection with longer immobilization time (≥ 28 days) (OR 1.5, CI 0.55–4.1), lower expertise of the physician (OR 1.0, CI 0.37–2.9), or male sex (OR 1.02, CI 0.36–2.9). Both base (n = 12) and tip (n = 10) fractures were associated with USP non-union but not statistically significantly (p = 0.6). 9 of the 22 USP non-union patients showed ≥ 2 mm anterior-posterior displacement of the radius primarily. In the subgroup analysis, comparing the proportions of the base and tip factures in patients with primarily diagnosed USP fracture, both types of fracture were equally common among the patients with healed USP fracture and USP non-union.

Discussion Ulnar styloid process fracture is associated with 20% of all distal radius fractures in children and it usually fails to unite (Abid et al. 2008). In our population-based study, USP nonunion was seen in 16% of all distal radius fractures. Only onefourth of the primarily recognized USP fractures successfully progressed to ossification during the long 11-year follow-up. Such a low rate of bone union is extremely unusual in childhood fractures. The USP is unique in this regard and the high non-union risk needs to be recognized. An interesting finding was that a majority of the patients with USP non-union were primarily diagnosed with isolated distal

radius fracture but showed USP non-union in the long-term follow-up. In children, the real incidence of USP non-union can only be evaluated after ossification of the ulnar styloid, if MRI is not available. Thus, there may be under-diagnosis of acute USP fractures in young children, which may explain the previously suggested higher incidence of USP non-union in adults (Stansberry et al. 1990, Abid et al. 2008, Wijffels et al. 2014). In our study, the patients were on average 21 years of age (14–29) at the time of follow-up. In that age group, the ulnar styloid process is ossified and visible in radiographs (Gilsanz and Ratib 2005) even though physeal closure usually occurs at the age of 16–19 years (Egol et al. 2010). It is possible that the open physis of the ulna would have made radiographic evaluation more difficult in the youngest study patients. However, reference imaging of the uninjured wrist was undertaken in all cases in order to support radiographic analysis. Routine MR imaging or CT scans were not included in the study plan because they have not been reported to be superior in diagnosing USP fractures, compared with plain radiographs in patients with mature skeleton (Spence et al. 1998, Welling et al. 2008). We found that the patients with USP fracture showed excellent clinical outcomes after > 10 years of follow-up, regardless of bone healing. Some individual patients reported pain during sports and minor complaints in daily life, but the findings were similar regardless of union or non-union. Many investigators have suggested an increased risk of DRUJ instability in patients with distal radius fracture with concomitant USP fracture due to lesions in the TFCC (Kazemian et al. 2011, Daneshvar et al. 2014, Gogna et al. 2014). However, we found DRUJ instability in only a few patients and in patients with both normal USP and USP non-union. Radiographic analysis was performed at a mean of 11 years after the initial injury. The mean residual displacement of the USP was only 1.0 mm. One-third of non-union patients showed ulnar shortening, which may predispose them to later degenerative processes. Ulnar shortening has been associated with ligament disruption (Ramos-Escalona et al. 2010) and early-stage joint degeneration (Kristensen and Soballe 1987, De Smet 1994). Premature radius growth arrest occurred in 4% of our patients, which was expressed as ulnar lengthening, using the uninjured side as a reference. Such a low rate of radius growth plate arrest is in agreement with the previous literature (1% to 7%) (Buterbaugh and Palmer 1988, Cannata et al. 2003, Abzug et al. 2014). It also needs to be kept in mind that radius and ulnar length discrepancy is not a static but a dynamic condition and the relationship changes during forearm rotation and grip loading (Schuurman et al. 2001). The associated risk factors of USP non-union remain unclear. Previous studies have suggested the base type of USP fracture to be a risk factor (Abid et al. 2008, Zenke et al. 2009), which we did not find. We performed a comprehensive risk analysis on patient-related information including sex, radiographic displacement and angular deformity, severity of the radius

Acta Orthopaedica 2019; 90 (4): 383–388

fracture according to Salter–Harris classification, treatment method, and the length of immobilization time to evaluate the risk of USP non-union. None of these was associated with a greater risk of USP non-union in this long-term study. The participation rate of 67% is a weakness of this study. It is probable that some of the non-participants had fewer symptoms than participants and were thus not interested in the follow-up examination. We assume that there were fewer symptomatic patients and non-unions among the non-participants, while a majority of non-participants reported full recovery on their own initiative, when they declined their participation. Nevertheless, it sounds reasonable that the rate of USP nonunion could have been smaller with full participation. It is also possible that some of the patients did not participate for reasons unrelated to the analyzed outcome variables in this study (Dunn et al. 2004, Galea and Tracy 2007). Such cases are considered to be missing completely at random, thus having no effect on the results (Kristman et al. 2004). Considering the relatively great time span between the primary injury in childhood and the follow-up visit, the participation rate was satisfactory, according to the literature, and a 60–80% participation rate has been recommended in epidemiologic long-term follow-up studies by many (Kristman et al. 2004, Galea and Tracy 2007, Fewtrell et al. 2008). Further, age, sex, and treatment were similar among participants and non-participants, which strengthens the findings. As another limitation, the primary data were based on those in hospital registers and not all interesting particulars were available. Further, we found only 22 patients with USP nonunion and a larger study would be necessary to confirm possible causal links between the primary injury, its treatment and the risk of USP non-union. In addition, complete DASH data were not utilized, although the baseline scores were collected and analyzed. However, longer follow-up is warranted to make conclusions concerning morbidity in later adulthood. In summary, although USP non-union was relatively common after a childhood distal radius fracture, the long-term clinical results were good. Ulnar shortening was found in one in three, however, and may predispose an individual to degenerative processes later in life.

LK analyzed the data and wrote the manuscript, SV analyzed all radiographs and critically reviewed the manuscript, WS contributed to the study design and data collection and critically reviewed the manuscript, JJS initiated the study, contributed to study design, carried out the follow-up visits and clinical examinations of the patients, and critically reviewed the manuscript. The authors would like to thank the Vaasa Foundation of Physicians, Finska Läkaresällskapet, the Emil Aaltonen Foundation, the Finnish Foundation of Pediatric Research and the Alma and K. A. Snellman Foundation for supporting the study.

387

Abid A, Accadbled F, Kany J, de Gauzy J S, Darodes P, Cahuzac J P. Ulnar styloid fracture in children: a retrospective study of 46 cases. J Pediatr Orthop B 2008; 17(1): 15-19. Abzug J M, Little K, Kozin S H. Physeal arrest of the distal radius. J Am Acad Orthop Surg 2014; 22(6): 381-9. Bae D S, Waters P M. Pediatric distal radius fractures and triangular fibrocartilage complex injuries. Hand Clin 2006; 22(1): 43-53. Buterbaugh G A, Palmer A K. Fractures and dislocations of the distal radioulnar joint. Hand Clin 1988; 4(3): 361-75. Cannata G, De Maio F, Mancini F, Ippolito E. Physeal fractures of the distal radius and ulna: long-term prognosis. J Orthop Trauma 2003; 17(3): 180. Chen Y, Zheng X, Shi H, Wangyang Y, Yuan H, Xie X, Li D, Wang C, Qiu X. Will the untreated ulnar styloid fracture influence the outcome of unstable distal radial fracture treated with external fixation when the distal radioulnar joint is stable. BMC Musculoskelet Disord 2013; 14: 186. Daneshvar P, Chan R, MacDermid J, Grewal R. The effects of ulnar styloid fractures on patients sustaining distal radius fractures. J Hand Surg Am 2014; 39(10): 1915-20. de Putter C E, van Beeck E F, Looman C W N, Toet H, Hovius S E R, Selles R W. Trends in wrist fractures in children and adolescents, 1997–2009. J Hand Surg Am 2011; 36(11): 1815.e2. De Smet L. Ulnar variance: facts and fiction review article. Acta Orthop Belg 1994; 60(1): 1-9. Dunn K M, Jordan K, Lacey R J, Shapley M, Jinks C. Patterns of consent in epidemiologic research: evidence from over 25,000 responders. Am J Epidemiol 2004; 159(11): 1087-94. Egol K, Koval K, Zuckerman J. Radius and ulna shaft. In: Handbook of fractures, 4th ed. Philadelphia: Lippincott Williams & Wilkins 2010; p 257-68. Fewtrell M S, Kennedy K, Singhal A, Martin R M, Ness A, Hadders-Algra M, Koletzko B, Lucas A. How much loss to follow-up is acceptable in long-term randomised trials and prospective studies? Arch Dis Child 2008; 93(6): 458-61. Galea S, Tracy M. Participation rates in epidemiologic studies. Ann Epidemiol 2007; 17(9): 643-53. Gilsanz V, Ratib O. Hand bone age: a digital atlas of skeletal maturity. Berlin/ Heidelberg: Springer Science & Business Media; 2005. Gogna P, Selhi H S, Mohindra M, Singla R, Thora A, Yamin M. Ulnar styloid fracture in distal radius fractures managed with volar locking plates: to fix or not? J Hand Microsurg 2014; 6(2): 53-8. Kazemian G H, Bakhshi H, Lilley M, Emami Tehrani Moghaddam M, Omidian M M, Safdari F, Mohammadpour I. DRUJ instability after distal radius fracture: a comparison between cases with and without ulnar styloid fracture. Int J Surg 2011; 9(8): 648-51. Khosla S, Melton L J, Dekutoski M B, Achenbach S J, Oberg A L, Riggs B L. Incidence of childhood distal forearm fractures over 30 years: a populationbased study. JAMA 2003; 290(11): 1479-85. Kramer S, Meyer H, O’Loughlin P F, Vaske B, Krettek C, Gaulke R. The incidence of ulnocarpal complaints after distal radial fracture in relation to the fracture of the ulnar styloid. J Hand Surg Eur 2013; 38(7): 710-17. Kristensen S S, Soballe K. Kienbock’s disease: the influence of arthrosis on ulnar variance measurements. J Hand Surg Br 1987; 12(3): 301-5. Kristman V, Manno M, Cote P. Loss to follow-up in cohort studies: how much is too much? Eur J Epidemiol 2004; 19(8): 751-60. Logan A J, Lindau T R. The management of distal ulnar fractures in adults: a review of the literature and recommendations for treatment. Strategies Trauma Limb Reconstr 2008; 3(2): 49-56. Mulders M A M, Fuhri Snethlage L J, de Muinck Keizer, Robert-Jan O, Goslings J C, Schep N W L. Functional outcomes of distal radius fractures with and without ulnar styloid fractures: a meta-analysis. J Hand Surg Eur 2018; 43(2): 150-7. Nelson O A, Buchanan J R, Harrison C S. Distal ulnar growth arrest. J Hand Surg Am 1984; 9(2): 164-70.

388

Ramos-Escalona J, Garcia-Bordes L, Martinez-Galarza P, Yunta-Gallo A. Ulnar variance and scaphoid fracture. J Hand Surg Eur 2010; 35(3): 195-7. Schuurman A H, Maas M, Dijkstra P F, Kauer J M. Assessment of ulnar variance: a radiological investigation in a Dutch population. Skeletal Radiol 2001; 30(11): 633-8. Souer J S, Ring D, Matschke S, Audige L, Marent-Huber M, Jupiter J B. Effect of an unrepaired fracture of the ulnar styloid base on outcome after plate-and-screw fixation of a distal radial fracture. J Bone Joint Surg Am 2009; 91(4): 830-8. Spence L D, Savenor A, Nwachuku I, Tilsley J, Eustace S. MRI of fractures of the distal radius: comparison with conventional radiographs. Skeletal Radiol 1998; 27(5): 244-9. Stansberry S D, Swischuk L E, Swischuk J L, Midgett T A. Significance of ulnar styloid fractures in childhood. Pediatr Emerg Care 1990; 6(2): 99-103. Waters P M, Bae D S, Montgomery K D. Surgical management of posttraumatic distal radial growth arrest in adolescents. J Pediatr Orthop 2002; 22(6): 717-24.

Acta Orthopaedica 2019; 90 (4): 383â&#x20AC;&#x201C;388

Welling R D, Jacobson J A, Jamadar D A, Chong S, Caoili E M, Jebson P J. MDCT and radiography of wrist fractures: radiographic sensitivity and fracture patterns. AJR Am J Roentgenol 2008; 190(1): 10-16. Wijffels M M E, Keizer J, Buijze G A, Zenke Y, Krijnen P, Schep N W L, Schipper I B. Ulnar styloid process nonunion and outcome in patients with a distal radius fracture: a meta-analysis of comparative clinical trials. Injury 2014; 45(12): 1889-95. Yuan C, Zhang H, Liu H, Gu J. Does concomitant ulnar styloid fracture and distal radius fracture portend poorer outcomes? A meta-analysis of comparative studies. Injury 2017; 48(11): 2575-81. Zenke Y, Sakai A, Oshige T, Moritani S, Nakamura T. The effect of an associated ulnar styloid fracture on the outcome after fixation of a fracture of the distal radius. J Bone Joint Surg Br 2009; 91(1): 102-7. Zoetsch S, Kraus T, Weinberg A M, Heidari N, Lindtner R A, Singer G. Fracture of the ulnar styloid process negatively influences the outcome of paediatric fractures of the distal radius. Acta Orthop Belg 2013; 79 (1): 48-53.

Acta Orthopaedica 2019; 90 (4): 389–393

389

Revision of trapeziometacarpal arthroplasty: risk factors, procedures and outcomes Simo MATTILA and Eero WARIS

Department of Hand Surgery, Helsinki University Central Hospital, Finland Correspondence: mattilsimo@gmail.com Submitted 2018-07-08. Accepted 2019-02-25.

Background and purpose — Revision surgery after trapeziometacarpal arthroplasty is sometimes required. Varying revision rates and outcomes have been reported in rather small patient series. Data on risk factors for revision surgery, on the final outcome of revision, and possible factors affecting the outcome of revision are also limited. We evaluated these factors in 50 patients. Patients and methods — From 1,142 trapeziometacarpal arthroplasties performed during a 10-year period, 50 patients with 65 revision surgeries were retrospectively identified and invited to participate in a follow-up study involving subjective, objective, and radiologic evaluation. The revision rate, risk factors for revision, and factors affecting the outcome of revision were analyzed. Results — The revision rate was 5%. Scaphometacarpal impingement was the most common reason for revision surgery. Patient age ≤ 55 years was a risk factor with a revision rate of 9% in this age group, whereas an operation on both thumbs during the follow-up period was a negative risk factor for revision surgery. There was no difference in revision risk between ligament reconstruction and tendon interposition with or without a bone tunnel. 9 patients had multiple revision procedures and their final outcome did not differ significantly from patients revised only once. Most of the patients felt subjectively that they had benefited from revision surgery and the subjective outcome measures (QuickDash and pain VAS) and the Conolly score were in the same range as previously described for revision trapeziometacarpal arthroplasty. Interpretation — Age ≤ 55 years is a risk factor for revision surgery. The type of primary surgery does not affect the risk of revision surgery and multiple revision procedures do not result in worse outcomes than cases revised only once. Mechanical pain caused by contact between the metacarpal and scaphoid is the most common indication for revision surgery. In general, patients seem to benefit from revision surgery for trapeziometacarpal osteoarthritis.

Osteoarthritis (OA) of the trapeziometacarpal (TMC) joint is a common degenerative disorder frequently treated surgically. Partial or complete trapeziectomy alone or combined with ligament reconstruction and/or tendon interposition (LRTI) are the most commonly used surgical methods, but also various implant procedures have been described (Muermans and Coenen 1998, Richard et al. 2014). In general, surgery for TMC OA is well tolerated with few complications and high patient satisfaction (Wajon et al. 2015). However, some patients may present with persistent or recurrent symptoms such as pain and hand dysfunction. In these cases, revision surgery is considered (Cooney et al. 2006, Megerle et al. 2011). Studies have reported various revision techniques including autologous or alloplastic interposition, ligament reconstruction, suspension, conversion to fusion, and re-excision arthroplasty with varying results (Conolly and Rath 1993, Cooney et al. 2006, Megerle et al. 2011, Papatheodorou et al. 2017, Renfree and Dell 2002, Wilkens et al. 2017). Risk factors for revision surgery have been analyzed in only 2 previous studies (Cooney et al. 2006, Wilkens et al. 2017). So far, no factors have been identified to affect the outcome of patients after revision surgery. We evaluated the incidence of failed TMC arthroplasty resulting in revision procedures, searched for risk factors for revision surgery, searched for factors affecting the results of revision surgery, and analyzed the final subjective and objective outcomes of revised patients. To our knowledge, this is the largest patient cohort to date on revision TMC arthroplasties.

Patients and methods We performed a retrospective chart review to search for all arthroplasty procedures on the thumb trapeziometacarpal (TMC) joint performed during a 10-year period from January 2003 to December 2013 at the single hand surgical unit of Helsinki University Hospital, Finland. The indications for the primary procedures were pain related to primary (Eaton–

390

Acta Orthopaedica 2019; 90 (4): 389–393

Table 1. Patient demographics Patients and procedures No. of patients Sex (male/female) Mean age Primary procedures Bilateral procedures Revisions No. of patients Mean age Revision procedures Primary revision procedures Cases with multiple revisions Follow-up Patients available Mean follow-up, months

930 121/809 61 (34–92) 1,142 212 50 (45 female) 57 (43–80) 65 52 9 38 43 (8–132)

Glickel stages 1–4) (n = 1,133) or posttraumatic (n = 9) osteoarthritis of the carpometacarpal (CMC-1) joint. Patients with rheumatoid arthritis and patients having had CMC-1 or scaphotrapeziotrapezoidal fusion were excluded from the study. Arthroplasties performed with implants were also excluded due to the small number of patients (n = 32). A consecutive series of 930 patients with 1,142 primary TMC arthroplasties was identified (Table 1). The following data regarding the primary procedures were collected: age, sex, operated side, arthroplasty performed on both sides, surgeon experience (resident/senior hand surgeon), postoperative immobilization time, type of surgery, simultaneous surgery on the metacarpophalangeal (MCP) joint, other simultaneous surgeries, and postoperative complications. The patients were divided into 3 groups based on the primary procedure: (1) trapeziectomy and LRTI with the abductor pollicis longus (APL) tendon either through a bone tunnel in the base of the metacarpal (LRTI + bone tunnel group) (Kaarela and Raatikainen 1999) or (2) LRTI with APL without a bone tunnel (a slip of the APL tendon weaved between the remaining APL tendon and flexor carpi radialis (FCR) tendon) (LRTI group) as described by Ceruso et al. (1991). The 3rd group consisted of simple trapeziectomy, partial trapeziectomy with interposition of palmaris longus tendon, LRTI with FCR (Weilby 1988), LRTI with extensor pollicis brevis, or total trapeziectomy and tendon interposition with the palmaris longus tendon without ligament reconstruction (Dell et al. 1978). From the medical records, 50 patients who had undergone revision surgery were identified and invited for a follow-up visit. The indication for revision surgery, the number of revision procedures performed for each hand, and the techniques of the revision procedures were determined from the medical records. Pre-revision radiographs were available for 42 hands. They were reviewed for the minimum distance between the base of the first metacarpal and scaphoid (scaphometacarpal space) seen in the posteroanterior view, residual bone fragments in the operative area, and MCP-joint hyperextension.

38 of the 50 patients attended the follow-up visit. The mean time from revision to follow-up was 43 months (8–132). Subjective assessment was performed with the Quick Disabilities of the Hand Shoulder and Arm score (QuickDASH), patient evaluation measure (PEM), and the visual analog score for pain (pain VAS). Objective assessment included grip strength with the Jamar Hand Dynamometer (Saehan Corporation, Seoul, South Korea), key and tip pinch strength with the pinch gauge, the ability to flatten hand measurement, thumb palmar and radial abduction, and thumb MCP and interphalangeal joint range of motion. Furthermore, the outcome of revision surgery was assessed with the Conolly–Rath score (Conolly and Rath 1993) and finally the patients were asked to assess subjectively whether or not they had benefited from the revision surgery. Posteroanterior, oblique, and lateral radiographs were taken of the operated hands. From these radiographs, the scaphometacarpal space was measured and the radiographs were evaluated for MCP-joint hyperextension and residual bone fragments in the operative area. Statistics Risk factors for revision surgery were analyzed with a conditional mixed model. There were 212/930 patients who had had surgery on both hands. Therefore, we used patients as random effects when estimating risk factors for revision surgery. Age was categorized according to quartiles since the impact of age on revision was nonlinear. The type of surgery was categorized into LRTI + bone channel, LRTI without bone channel, and other (all other primary surgery techniques) and immobilization time postoperatively was dichotomized at 0–4 weeks or 5–8 weeks. Variables were entered into the multivariable conditional mixed model one by one if their p-value in the univariable model was < 0.3. A variable was left in the final model if its p < 0.05 or the change in the Pseudo-Likelihood function was significant compared with the previous model. The results of the conditional mixed model are presented as odds ratio (OR) with a 95% confidence interval (CI). 2-tailed p-values are presented. The odds ratios can be interpreted as relative risks due to the small incidence of revision surgery (5%). For simple correlations, Spearman’s correlation coefficient (rho) was calculated. Continuous data were analyzed with Student’s t-test or Welch’s t-test, the latter if Levene’s test showed unequal variances. Paired categorical data were analyzed with McNemar’s test. SPSS for windows (IBM Corp. Released 2018. IBM SPSS Statistics for Windows, Version 25.0; IBM Corp, Armonk, NY, USA) and SAS (version 9.4; SAS Institute Inc, Cary, NC, USA) were used for analyses. Ethics, funding, and potential conflicts of interest The institutional review board and ethical committee of Helsinki University Hospital, Finland approved the study, Dnro 6/13/03/02/2013. This research was funded by the University

Acta Orthopaedica 2019; 90 (4): 389–393

391

Table 2. Primary procedures and revisions Primary procedures

Number of Number of primary surgeries revisions (%)

Conventional arthroplasty LRTI with APL tendon 679 LRTI with APL tendon + bone tunnel 418 Others LRTI with FRC tendon 2 LRTI with EPB tendon 2 Interposition with palmaris longus tendon without ligament reconstruction 5 Simple trapeziectomy 29 Partial trapeziectomy 7 Total number of procedures 1,142 Total number of patients 930

32 (4.7%) 16 (3.8%) 4a 0 0 0 2 2 52 50

revision rate not calculated due to small total number of operated patients.

of Helsinki and the Department of Hand Surgery, Helsinki University Hospital, the Finnish Medical foundation, and Vappu Uuspää foundation. The authors report no conflicts of interest.

Results Altogether 50 patients (52 thumbs) had revision surgery (revision rate 4.6 %). The revision rate was 4.7% for the LRTI group and 3.8% for the LRTI + bone tunnel group. In group 3 there were 4 revisions in 45 cases (Table 2). The total number of revisions performed was 65 with 9 patients having had multiple revision procedures for the same thumb (range 2–5 procedures) (Table 3, see Supplementary data). The mean time from primary surgery to the first revision procedure was 23 months (8–92). The indication for revision surgery was pain in all cases, which was typically related to hand use and could be provoked at the outpatient clinic by loading the thumb axially towards the scaphoid, medially towards the trapezoid, or laterally stretching the joint capsule. In 30 cases, the main reason for pain as judged by the treating physician was subsidence of the metacarpal against the scaphoid or trapezoid bones (Figure 1) or contact of the metacarpal with remnants of the trapezium (Figure 2) in the resection cavity. In 13 cases, instability of the base of the metacarpal associated with MCP-joint hyperextension was the reason for the pain. In 2 cases (both hands of the same patient) the pain was caused by carpal instability. Pain related to tendon irritation or tenosynovitis was the reason for pain in 3 cases with tenderness of the APL tendon in the operative area or the distal FCR tendon related to irritation by the APL sling. Pain related to radial sensory nerve irritation or neuroma in 4 cases was provoked by touch or pressure with a positive Tinel sign.

Figure 1. Subsidence of the metacarpal against the scaphoid 3 years after LRTI with APL.

Figure 2. A remnant of the trapezium in the resection cavity after LRTI with APL.

Several techniques were used in revision cases (Table 4, see Supplementary data). For cases of metacarpal subsidence (Figure 2) the most common technique was interposition of a strip of fascia lata, suspension, and tendon interposition. In cases of instability of the thumb base and MCP-joint hyperextension, suspension arthroplasty and MCP-joint fusion were the most common procedures. Tenosynovitis of the FCR tendon caused by the APL sling was treated with FCR tenotomy or release of the sling. Neuromas were treated by release from scar tissue. No neuroma resections or nerve reconstructions were performed. Patient age ≤ 55 years was a risk factor for revision surgery (9% revision rate) compared with age groups 56–60 years (OR 0.4, p = 0.02) (4% revision rate), 61–65 years (OR 0.1, p < 0.001) (1.2 % revision rate) and > 66 years (OR 0.40, p = 0.02)(4.6% revision rate). There was a negative risk for revision surgery in patients operated on both thumbs at some point during the follow-up period (OR 0.4, p = 0.02). Furthermore, revision surgery was a rare event on the second operated hand (2 cases out of 202) if the operation on the first operated hand was successful. There was no statistically significant difference between the revision risk of LRTI with APL compared with LRTI with APL + bone tunnel (OR 0.8, p = 0.5) (Table 5). Comparison of age, sex, failed primary revision procedure, or the radiographic data obtained before revision surgery (scaphometacarpal space and MCP-joint hyperextension) with the outcome variables QuickDash, PEM, pain VAS, and key pinch showed that key pinch strength was statistically significantly higher in cases with a scaphometacarpal space ≤ 1mm (Table 6, see Supplementary data). A comparison of the radiographic data (scaphometacarpal space and MCP-joint hyperextension) between pre-revision and final follow-up showed fewer cases with a completely lost scaphometacarpal space (0–1mm) in the final follow-up radiographs (6/36) compared with pre-revision radiographs (11/42), but the difference was not statistically

392

Table 5. Risk factors for revision surgery

Acta Orthopaedica 2019; 90 (4): 389–393

2006, Megerle et al. 2011). Regarding the subjective outcomes the mean QuickDASH score in our study of Univariable conditional Multivariable conditional mixed model mixed model 38 and the mean Pain VAS score of Risk factor Odds ratio (95% CI) p-value Odds ratio (95% CI) p-value 42 are in the same range as previously reported for revised patients Surgery on both thumbs (negative risk) 49 (0.25–0.96) 0.04 0.42 (0.21–0.87) 0.02 after TMC arthroplasty (Megerle Age, years et al. 2011, Sadhu et al. 2016). Our ≤ 55 1.0 1.0 results and the literature show that 56–60 0.43 (0.21–0.88) 0.02 0.43 (0.21–0.87) 0.02 61–65 0.12 (0.04–0.42) < 0.001 0.12 (0.04–0.41) < 0.001 the majority of patients seem to ben> 66 0.40 (0.19–0.87) 0.02 0.34 (0.15–0.76) 0.01 efit from revision surgery but the a Sex 1.4 (0.55–3.8) 0.5 a final outcome may still be worse than Operated side 0.87 (0.50–1.5) 0.6 a Surgeon experience 1.04 (0.58–1.9) 0.9 that of non-revised patients (Sadhu et Surgical method al. 2016). The results of cases revised a LRTI APL 1.0 multiple times were not worse than LRTI APL + bone tunnel 0.81 (0.43–1.5) 0.5 Others 2.0 (0.64–6.1) 0.2 those revised only once. Therefore, it a Complications of primary surgery 0.83 (0.25–2.8) 0.8 seems that it is beneficial to operate a Immobilization time after surgery 1.6 (0.82–3.3) 0.2 on these patients several times if necSimultaneous surgery on the a metacarpophalangeal joint 0.70 (0.09–5.4) 0.7 essary to achieve a decent outcome. a Other simultaneous surgeries 0.60 (0.21–1.7) 0.3 Age ≤ 55 years was found to be a Odds ratio not calculated because no significance in univariable model. a risk factor for revision with 9% of patients in this age group having revision surgery. In patients who significant (p = 0.5). The number of cases with MCP-joint have had a successful arthroplasty on the first thumb, the incihyperextension in radiographs did not change significantly (p dence of revision surgery on the contralateral second oper= 0.3). The mean pain VAS score was 40 mm (0–100), the ated thumb was low (1%). Risk factors for revision surgery mean DASH score 37 (2–73), and mean grip strength 23 kg have been analyzed in only 2 studies (Cooney et al. 2006, (5–48). According to the Connolly–Rath score, there were 4 Wilkens et al. 2017). Patient age, type of primary surgery, good, 27 fair, and 5 poor results. Regarding patient satisfac- and surgeon experience were identified as risk factors for tion, 31 of 34 patients felt subjectively that they had benefited revision by Wilkens et al. but Cooney et al. did not identify from revision surgery. See Table 7 (Supplementary data) for any significant risk factors. Our study showed a similar result regarding age. This is probably related to the higher physiall the results of revision surgery. cal demands of younger patients, leading to more mechanical problems. The type of primary surgery in our study was not a risk factor for revision probably because the trapezium Discussion was completely removed in both of the main primary proceIn our study, the revision rate for TMC arthroplasty was 4.6 dures groups (LRTI and LRTI + bone tunnel). In the study %, close to that of previous studies (2.6–4.0 %) (Cooney et al. of Wilkens et al. the primary procedures included partial tra2006, Megerle et al. 2011, Wilkens et al. 2017). Although gen- peziectomies and implant procedures, which are generally at erally only 1 revision is required (Cooney et al. 2006, Megerle higher risk for revision (Muermans and Coenen 1998, Richard et al. 2011), repeat revision procedures are not uncommon and et al. 2014, Wajon et al. 2015). In patients having an operation an average of 5 procedures were required in 1 study (Renfree on both hands, the reason for the low incidence of revision and Dell 2002). In our study, 9 of 50 patients had more than procedures on the second operated hand could be related to 1 revision procedure. Pain caused by metacarpal subsidence ligament laxity affecting joint stability and possibly increasing or instability seems to be the reason for revision surgery in the risk for metacarpal subsidence, which is involved in many almost all cases in the literature (Cooney et al. 2006, Megerle revision cases (Megerle et al. 2011, Papatheodorou et al. 2017, et al. 2011, Papatheodorou et al. 2017, Wilkens et al. 2017), Wilkens et al. 2017). Also factors such as patient confidence in the procedure being a good choice and that there is a good which is similar to our findings. Both good (Cooney et al. 2006) and poor (Megerle et al. indication for surgery might explain this difference. A large variety of procedures have been used for revision of 2011) results have been reported after revision TMC arthroplasty. The majority of our patients had a Conolly score TMC arthroplasty (Cooney et al. 2006, Megerle et al. 2011, of fair and said they benefited from the revision surgery. Papatheodorou et al. 2017, Wilkens et al. 2017). No previous However, there were 5 poor results in 38 patients, which is studies have been able to identify any correlation between outapproximately in line with previous studies (Cooney et al. come data and the type of revision procedure performed. This

Acta Orthopaedica 2019; 90 (4): 389–393

is probably related to the large number of different procedures in use for revision surgery, which leads to small subgroups of procedures that limit the statistical power of the analysis. This same problem applied to our study, making statistical analysis of the correlation between the type of revision procedure and final outcome unreliable. Analysis of radiographs taken at final follow-up showed that the scaphometacarpal space was lost completely in 6/36 cases. This may be because none of the surgical methods used for revision in our study are able to adequately address the problem of scaphometacarpal impingement. One of the strengths of this study was the analysis of a large number of primary TMC arthroplasties performed for OA. This provided considerable statistical power to the analysis of risk factors for revision. However, because revision surgery is a rare event, there still was a limited number of revised patients, which made the statistical analysis of several potential factors affecting the final outcome unreliable. Also, a limitation of this study is the retrospective design, which introduces a risk of bias. In our hand surgical unit, trapeziectomy with and rarely without ligament reconstruction and interposition was the method of choice during the study period. However, in some selected cases for high-demand patients, who could be at high risk for revision, alternative methods such as implant arthroplasties were performed, which were excluded from the analysis due to the small number of cases. In summary, patients ≤ 55 years are at greater risk for revision than older age groups. Revision surgery on the second operated hand after successful surgery on the first hand is rare. Repeat revision procedures are sometimes required but the outcome does not differ from patients undergoing only one revision. A bone tunnel (LRTI + bone tunnel) to stabilize the thumb base does not reduce the risk of revision surgery compared with other surgical methods used in this study. Supplementary data Tables 3, 4, 6, and 7 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1599253

393

SM composed the manuscript, gathered the data, and met the patients at the follow-up visit. EW contributed substantially to the design of the study, the composition of the manuscript, and planning of the data acquisition. The authors wish to thank statistician Pasi Ohtonen, MSc, for consultation with the statistical analysis in this study. Acta thanks Philippe Kopylov for help with peer review of this study.

Ceruso M, Innocenti M, Angeloni R, Lauri G, Bufalini C. L’Artrosi del primo raggio. Riv Chir Mano 1991; 28: 67-75. Conolly W B, Rath S. Revision procedures for complications of surgery for osteoarthritis of the carpometacarpal joint of the thumb. J Hand Surg Eur 1993; 18B: 533-9. Cooney W P, Leddy T P, Larson D R. Revision of thumb trapeziometacarpal arthroplasty. J Hand Surg Am 2006; 31: 219-27. Dell P, Brushart T, Smith R. Treatment of trapeziometacarpal arthritis: results of resection arthroplasty. J Hand Surg 1978; 3: 243-9. Kaarela O, Raatikainen T. Abductor pollicis longus tendon interposition arthroplasty for carpometacarpal osteoarthritis of the thumb. J Hand Surg Am 1999; 24: 469-75. Megerle K, Grouls S, Germann S. Revision surgery after trapeziometacarpal arthroplasty: Arch Orthop Trauma Surg 2011; 131: 205-10. Muermans S, Coenen L. Interpositional arthroplasty with Gore-Tex, Marlex or tendon for osteoarthritis of the trapeziometacarpal joint: a retrospective comparative study. J Hand Surg Eur 1998; 23: 64-8. Papatheodorou L K, Winston J D, Bielicka D L, Rogozinski B J, Lourie G M, Sotereanos D G. Revision of the failed thumb carpometacarpal arthroplasty. J Hand Surg Am 2017; 42: 1032.e1-1032.e7. Renfree K J, Dell P C. Functional outcome following salvage of failed trapeziometacarpal joint arthroplasty. J Hand Surg 2002; 27: 96-100. Richard M, Lunich J, Correll G. The use of the Artelon cmc spacer for osteoarthritis of the basal joint of the thumb. J Hand Ther 2014; 27: 122-6. Sadhu A, Calfee R P, Guthrie A, Wall L B. Revision ligament reconstruction tendon interposition for trapeziometacarpal arthritis: a case control investigation. J Hand Surg Am 2016; 41: 1114-21. Wajon A, Vinycomb T, Carr E, Edmunds I, Ada L. Surgery for thumb (trapeziometacarpal joint) osteoarthritis. Cochrane Database Syst Rev 2015; 23: CD004631. Weilby A. Tendon interposition arthroplasty of the first carpo-metacarpal joint. J Hand Surg 1988; 13B: 421-5. Wilkens S C, Zichao X, Mellema J J, Ring D, Neal C. Unplanned reoperation after trapeziometacarpal arthroplasty: rate, reasons and risk factors: Hand 2017; 12: 446-52.

394

Acta Orthopaedica 2019; 90 (4): 394–400

Artificial intelligence detection of distal radius fractures: a comparison between the convolutional neural network and professional assessments Kaifeng GAN 1,2, Dingli XU 2, Yimu LIN 3, Yandong SHEN 1,2, Ting ZHANG 1, Keqi HU 1, Ke ZHOU 1, Mingguang BI 1, Lingxiao PAN 1, Wei WU 4, and Yunpeng LIU 5 1 Department of Orthopaedics, Ningbo Medical Center, Lihuili Hospital, Ningbo, 315000, China; 2 School of Medicine, Ningbo University, Ningbo, 315000, China; 3 Department of Orthopaedics, Second Affiliated Hospital of Wenzhou Medical University, Wenzhou, 325027, China; 4 Department of Orthopaedics, Second Hospital of Ningbo, Ningbo, 315000, China; 5 Faculty of Electronics & Computer, Zhejiang Wanli University, Ningbo, 315000, China

Correspondence: liuypjoy@163.com Submitted 2018-12-11. Accepted 2019-03-09.

Background and purpose — Artificial intelligence has rapidly become a powerful method in image analysis with the use of convolutional neural networks (CNNs). We assessed the ability of a CNN, with a fast object detection algorithm previously identifying the regions of interest, to detect distal radius fractures (DRFs) on anterior–posterior (AP) wrist radiographs. Patients and methods — 2,340 AP wrist radiographs from 2,340 patients were enrolled in this study. We trained the CNN to analyze wrist radiographs in the dataset. Feasibility of the object detection algorithm was evaluated by intersection of the union (IOU). The diagnostic performance of the network was measured by area under the receiver operating characteristics curve (AUC), accuracy, sensitivity, specificity, and Youden Index; the results were compared with those of medical professional groups. Results — The object detection model achieved a high average IOU, and none of the IOUs had a value less than 0.5. The AUC of the CNN for this test was 0.96. The network had better performance in distinguishing images with DRFs from normal images compared with a group of radiologists in terms of the accuracy, sensitivity, specificity, and Youden Index. The network presented a similar diagnostic performance to that of the orthopedists in terms of these variables. Interpretation — The network exhibited a diagnostic ability similar to that of the orthopedists and a performance superior to that of the radiologists in distinguishing AP wrist radiographs with DRFs from normal images under limited conditions. Further studies are required to determine the feasibility of applying our method as an auxiliary in clinical practice under extended conditions.

Conventional radiographs remain the primary diagnostic approach to detect distal radius fractures (DRFs) (Mauffrey et al. 2018, Waever et al. 2018). Non-orthopedic surgeons or young radiologists at emergency departments, where urgent decision-making is often required, are usually the first doctors to assess radiographs. Therefore, an accurate and efficient assistant technology in fracture detection is of interest. Artificial intelligence (AI) is achieving remarkable progress in image interpretation (He et al. 2015, Russakovsky et al. 2015). Since 2012, deep learning, a branch of AI, has rapidly become a powerful method in image analysis with the use of convolutional neural networks (CNNs), which are well suited for analyzing image features (Russakovsky et al. 2015, Lakhani and Sundaram 2017). There are increasing numbers of experimental trials that apply deep learning in medical image analysis in certain fields, including the automated analysis of pulmonary tuberculosis (Lakhani and Sundaram 2017), lung nodules (Hua et al. 2015, Nishio et al. 2018), retinopathy (Ting et al. 2017), gastric cancer (Wang et al. 2018), and dermatological diseases (Li and Shen 2018, Yap et al. 2018, Fujisawa et al. 2019). In the field of traumatic orthopedics, a few studies (Olczak et al. 2017, Chung et al. 2018, Kim and MacKinnon 2018, Urakawa et al. 2019) investigated the experimental applications of deep learning to detect fractures on plain radiographs; all the CNNs adopted showed excellent performance, and some (Chung et al. 2018, Urakawa et al. 2019) had abilities superior to that of humans. To further validate the feasibility of AI as an automatic diagnostic model, we first evaluated the ability of a CNN, with a fast object detection algorithm previously identifying the regions of interest, to detect DRFs on AP wrist radiographs. Second, the diagnostic performances of CNNs were compared with those of radiologists and orthopedists.

Acta Orthopaedica 2019; 90 (4): 394â&#x20AC;&#x201C;400

395

Materials and methods Design of study With the dataset, a fast object detection algorithm based on deep learning was first trained to identify the distal radiuses on AP wrist radiographs as the regions of interest (ROIs). Second, we adopted this fast object detection algorithm, of which the feasibility had been verified by a validation process, to automatically annotate the ROIs on AP wrist radiographs in the training dataset and test dataset. The ROIs were extracted as images, with which a diagnostic CNN model was then trained and tested in detecting the DRFs. The diagnostic performances in terms of accuracies, sensitivities, specificities, and Youden Index were finally compared among the diagnostic CNN model, radiologists, and orthopedists. Dataset 2 senior orthopedists with more than 10 years of orthopedic professional experience retrospectively reviewed 2,359 plain wrist radiographs with diagnostic reports from 2,359 adult patients (the inclusion and exclusion criteria for this study are given in the Supplementary data) who underwent radiological examinations at the Medical Center of Ningbo City, Lihuili Hospital, of the Ningbo University School of Medicine, between January 2010 and September 2017 to confirm that each case had an accurate diagnosis (with DRFs or without DRFs). A consensus was achieved in consultation with a third senior orthopedist with 22 years of orthopedic professional experience. For cases in which all 3 orthopedists did not agree, the corresponding wrist CT images were reviewed; CTs were available in most of these cases and a consensus on each case was reached after discussion. 19 controversial cases without CT exams were excluded from the study. 2,340 AP wrist radiographs (1,491 DRF cases and 849 normal wrists) from 2,340 adult patients were ultimately included in the final dataset. Data preparation Each plain AP wrist radiograph, originally stored as a Digital Imaging and Communications in Medicine (DICOM) file, was exported as a Joint Photographic Experts Group (JPEG) file with a matrix size of 600 x 800 pixels from the Picture Archiving and Communication System (PACS) by using eWorld Viewer (TomTaw Tech, Ningbo, China). For further analyses, 1,491 images with DRFs and 849 images without DRFs (randomized with the Research Randomizer program, http://www. randomizer.org) were randomly divided into an original training dataset of 2,040 images (1,341 images with DRFs and 699 images without DRFs) and a test dataset of 300 images (150 images with DRFs and 150 images without DRFs).

Figure 1. A wrist radiograph was manually annotated with a red rectangle as the ground truth bound and automatically annotated with a blue rectangle as the candidate bound. The red rectangle and blue rectangle represent edges of the region of interest (ROI) detected by the orthopedists and edges of the ROI detected by Faster R-CNN, respectively.

Training the CNN models The detailed experimental environment is described in the Appendix (see Supplementary data). Training the Faster R-CNN (Region-based CNN) Faster R-CNN (Ren et al. 2017) technology, one of the fast object detection algorithms based on deep learning, has excellent performance in locating the regions of interest (ROIs) on graphics. In this study, we trained and tested Faster R-CNN as an auxiliary algorithm to the diagnostic CNN model. The detailed training procedure of the Faster R-CNN model is shown in the Appendix. Validation of the Faster R-CNN A regression analysis (Mitra et al. 2018) was used to assess the training process of Faster R-CNN. The mean square error (MSE) (Kumar et al. 2018) was calculated to measure the loss of Faster R-CNN in the automatic annotation of the ROI. The test dataset of 300 images, including 150 images with fractures and 150 images without fractures, was used to evaluate the capacity of the trained Faster R-CNN model in automatic annotation of the ROI on images. First, using LabelImg (https://github.com/tzutalin/labelImg), 2 orthopedists with more than 5 years of orthopedic professional experience annotated each imageâ&#x20AC;&#x2122;s ROI as the ground truth bounds (GTBs), in which the whole distal radius was definitely encased. Then, a candidate boundary (CB) on each image with a GTB was annotated as the automatically detected ROI by the trained Faster R-CNN (Figure 1). The matrix sizes of the identified ROIs ranged from 207 to 223 pixels in width and from 208 to 231 pixels in height, respectively. The Intersection of the Union (IOU) (Mitra et al. 2018) was calculated, as illustrated

396

Acta Orthopaedica 2019; 90 (4): 394â&#x20AC;&#x201C;400

Figure 2. The formula with which the Intersection of the Union (IOU) was calculated.

Figure 3. A typical example of the augmentation on 1 image from the annotated training dataset during the training course of Inception-v4. Augmentation procedure

Augmented dataset (2,040 x 3 images)

Training dataset Manually annotating by labelImg

Test dataset with manual annotations

Annotated dataset with coordinates

Trained Faster R-CNN

Faster R-CNN

Test dataset (300 images)

Augmentation procedure

Inception-v4

Automatically annotated and extracted dataset (300 images)

Figure 4. Flow diagram of the training and test courses of Faster R-CNN (shown in a green) and Inception-v4 (shown in a red).

in Figure 2, to statistically evaluate the trained Faster R-CNN, with a value greater than 0.5 indicating success in detecting the ROI on an image. Training the diagnostic CNN model We used Inception-v4 (Szegedy et al. 2017) as the diagnostic model, which has achieved state-of-the-art results in recent image classification contests. In this study, only the imagesâ&#x20AC;&#x2122; ROIs automatically annotated by Faster R-CNN were used as the recognition targets; after the ROI was extracted, the rest region on each initial image was discarded as unnecessary interference factors and noises. Since the areas where a DRF would occur were focused on, the Inception-v4 modelâ&#x20AC;&#x2122;s training process of distinguishing images with fractures from normal images was much faster and more accurate than analyzing the entire image.

Result of the Faster R-CNN test Training dataset (2,040 images)

Automatically annotated and extracted dataset (2,040 images)

Augmented dataset (2,040 x 3 images)

Trained Inception-v4

Results of the Inception-v4 test

First, each initial image in the original training dataset, including 1,341 images with DRFs and 699 images without DRFs, was automatically annotated by the trained Faster R-CNN. The result of the annotations on 2,040 images was reviewed by 2 orthopedists, and each distal radius region was then confirmed to be appropriately encased in the bounds. The ROIs extracted from all the annotated images were resized to 200 x 200 pixels, and stored as JPEG files, which were then augmented via the same technique as that used in the training of Faster R-CNN (Figure 3). Finally, there were 6,120 images in the data pool as the final training dataset for the Inception-v4 model, including 4,023 images with DRFs and 2,097 images without DRFs; 15% of the dataset was randomly selected into the validation dataset. The summary of the training course is illustrated in Figure 4. The detailed training procedure of the Inception-v4 model is shown in the Appendix.

Acta Orthopaedica 2019; 90 (4): 394–400

Evaluation of the diagnostic performance of Inception-v4 First, each initial image in the original test dataset, including 150 images with DRFs and 150 images without DRFs, was automatically annotated by the trained Faster R-CNN. The result of the annotations on 300 images was reviewed by 2 orthopedists, and each distal radius region was then confirmed to be appropriately encased in the bounds. The ROIs extracted from the 300 annotated images were all resized to 200 x 200 pixels, and stored as JPEG files, consisting of a new test dataset. The final analysis of the trained Inception-v4 model was performed using the new test dataset of 300 images to inspect its ability to discern images with fractures from the normal images. Each image was analyzed using the trained Inceptionv4 model, which resulted in a score representing the likelihood that the image would be classified as “with a DRF” or “without a DRF.” This score had a continuous value between 0 and 1. The receiver operating characteristic (ROC) curve was generated using a Python script (https://www.python.org), and the AUC was determined. Evaluation of the performance of the medical professionals We set up a group of radiologists and a group of orthopedists to compare their results with those of the CNN to evaluate its diagnostic performance. The groups consisted of 3 radiologists who had at least 3 years of radiological professional experience and had passed the intermediate certificate exams and 3 orthopedists (none of whom participated in the validating process of review) with more than 5 years of orthopedic professional experience. The detailed procedure is described in the Appendix. Each image in the new test dataset was diagnosed as either “with a DRF” or “without a DRF.” In situations where disagreements arose in the same group regarding the diagnoses, the final decisions were made by a majority vote. Comparison of the results of Inception-v4 and those of the medical professionals After the ROC curve of Inception-v4 had been generated, the diagnostic cut-off at a threshold designed to maximize the Youden Index was set, and sensitivity, specificity, and accuracy of the Inception-v4 model were then calculated and statistically compared with these of the human groups. Statistics The SPSS software (version 22.0, IBM Corp, Armonk, NY, USA) was used to perform all statistical analyses. The demographic characteristics of all patients enrolled in this study are presented as mean (95% confidence intervals (CIs)) for age and count (percentage) for sex. P-values were derived from 1-way analysis of variance for age and chi-square tests for sex. The significance level was set at p < 0.05. The IOU of Faster R-CNN and AUC of Inception-v4 were calculated and described in terms of the means and CIs. CIs of the distributions for the 4 kinds of outcomes and for the

397

Table 1. Demographic data of the whole dataset with 2,340 patients enrolled in this study Factor

Patients Patients with DRFs without DRFs Total Comparison (n = 1,491) (n = 849) (n = 2,340) (p-value)

Age, mean (CI) 48 (48–49) 48 (47–49) 48 (48–49) 0.4 Sex, n (%) Male 833 (56) 533 (63) 1,366 (58) < 0.01 Female 658 (44) 316 (37) 974 (42) DRFs = distal radius fractures. CI = 95% confidence interval. P-values were derived from 1-way analysis of variance for age and chi-square tests for sex.

differences in the outcomes between the CNN model and each human group were computed via bootstrapping with 10,000 bootstraps. Comparisons between the CNN model and each human group were performed using a 1-way analysis of variance, followed by Dunnett’s test for multiple comparison with the significance level set at p < 0.05. Ethics, funding, and potential conflicts of interest The Ningbo Lihuili Hospital Ethics Committee approved the study (LH2018-039). Financial support for the study was from the Ningbo Natural Science Fund (No.2018A610164). All authors declare no conflicts of interest.

Results Demographic data of the included patients All of the patient radiographs (1,366 men and 974 women) were kept anonymous throughout this study. The patients’ mean age at the time they took the radiographs was 48 years (20–88). No statistically significant difference was found in age (p = 0.4) between the group of patients with fractures and group of patients without fractures, but there was significant difference in sex between the 2 groups (p < 0.01) (Table 1). Performance of Faster R-CNN The learning courses of Faster R-CNN in the final training and validation datasets are shown in the Appendix. In the test dataset, the average IOU value of Faster R-CNN was 0.87 (CI 0.86–0.87), and none of the IOU values was less than 0.5. 2 orthopedists reviewed each annotated CB, which was confirmed by encasing the whole distal radius on each image. Performance of the Inception-v4 model The learning courses of Inception-v4 in the final training and validation datasets are shown in the Appendix. The ROC curve for the test output of Inception-v4 is plotted in Figure 5, and the AUC was 0.96 (CI 0.94–0.99). At the optimal cut-off point, the value of the threshold was 0.64.

398

Acta Orthopaedica 2019; 90 (4): 394–400

Table 3. Performance differences in the outcomes between Inception-v4 and each human group Differences between Inception-v4 and orthopedists radiologists Factor Difference a (CI) Difference b (CI) Accuracy (%) Sensitivity (%) Specificity (%) Youden Index

–1 (–5 to 3) –3 (–9 to 4) 1 (–5 to 7) –0.01 (–0.09 to 0.06)

9 (3–15) 9 (1–16) 9 (3–16) 0.18 (8–27)

CI = 95% confidence interval. = (mean of the outcome of Inception-v4) – (mean of the outcome of orthopedists). b Difference = (mean of the outcome of Inception-v4) – (mean of the outcome of radiologists). a Difference

Figure 5. The receiver operating characteristic (ROC) curve for the test output of the Inception-v4 model. The dots on the plot represent the sensitivity and 1-specificity of the human groups (the blue dot represents the orthopedists’ group; the red dot represents the radiologists’ group). The sensitivity/1-specificity dot of the radiologists’ group lies below the ROC curve of the Inception-v4 model, and the sensitivity/1specificity dot of the orthopedists’ group lies above the ROC curve of the Inception-v4 model.

Table 2. Diagnostic performance of the model and human groups Factor

Inception-v4

Orthopedists

Radiologists

F-value p-value

Accuracy (%) [CI] 279/300 (93) [90–96] 281/300 (94) [91–96] 252/300 (84) [80–88] a 10.19 < 0.001 Sensitivity (%) [CI] 135/150 (90) [85–95] 139/150 (93) [89–97] 122/150 (81) [75–87] a 5.07 0.007 Specificity (%) [CI] 144/150 (96) [93–99] 142/150 (95) [91–98] 130/150 (87) [81–92] a 4.82 0.009 Youden Index (CI) 0.86 (0.80–0.91) 0.87 (0.82–0.93) 0.68 (0.61–0.75) a 11.62 < 0.001 CI = 95% confidence interval. a Statistically significant in a comparison of Inception-v4 and each human group (results from Dunnett’s test).

Comparison between the Inception-v4 model and human performance The model showed a superior capacity compared with the radiologists’ group to distinguish images with DRFs from normal images in terms of accuracy, sensitivity, specificity, and Youden Index. The CNN model presented a similar diagnostic capability to that of the orthopedists in terms of the outcomes (Tables 2 and 3).

Discussion In our study, both deep learning models demonstrated an excellent ability to recognize image traits in wrist radiographs. The trained Faster R-CNN, which had a 100% success rate in automatically annotating the ROIs on images from the test dataset, acted as a valid auxiliary algorithm to the Inceptionv4 model, which was trained to distinguish images with DRFs from normal images. The Inception-v4 model exhibited a similar diagnostic capability to that of the orthopedists and superior performance to that of the radiologists.

Previous studies investigating the feasibility of applying CNNs to detect fractures on radiographs showed promising results, consistent with those of our study. Kim and MacKinnon (2018) trained Inception-v3 to recognize wrist fractures on lateral wrist radiographs; their results showed that the value of AUC was 0.954 and the maximized values of the sensitivity and specificity were 0.9 and 0.88, respectively. Olczak et al. (2017) performed a study in which a Visual Geometry Group 16-layer (VGG_16) network was trained to detect fractures on hand, wrist, and ankle radiographs with an accuracy of 83%, similar to the performance of the radiologists (who had an accuracy of 82%). Chung et al. (2018) evaluated the ability of the Residual Network (ResNet) model to detect and classify proximal humerus fractures using shoulder radiographs. The CNN showed superior top-1 accuracy, an accuracy of 96%, which was greater than that of the orthopedists (93%). Urakawa et al. (2019) conducted a study in which they compared the capacities of the VGG_16 network and orthopedic surgeons in detecting intertrochanteric fractures on radiographs, revealing the diagnostic performance of the CNN; the CNN had an accuracy of 96%, which exceeded that of ortho-

Acta Orthopaedica 2019; 90 (4): 394–400

Figure 6. The same wrist with a DRF in the anterior–posterior view radiograph (a) and in the lateral view radiograph (b). The hidden DRF in the anterior–posterior view was apparent in the lateral view (the fracture is shown by the red arrow).

pedic surgeons, who had an accuracy of 92%. All the previous studies mentioned prepared the images that were used in the training datasets and test datasets by manually cropping them into certain matrix sizes before the images were input into the CNNs. Since the images were uniform and had concentrated matrix sizes, the ROIs on the images were recognized faster and more accurately by the deep learning models, thereby remarkably improving the efficiency of the CNNs in the learning and test procedures. We employed and trained the Faster R-CNN model to automatically annotate the ROIs on images as a reliable substitution for manual cropping, which resulted in a low processing time and decreased bias (Urakawa et al. 2019). There is a great potential for the Inception-v4 model to be combined with Faster R-CNN to detect DRFs in clinical practice, where wrist radiographs with both ROIs and irrelevant regions are presented. We plotted the ROC curve for the test output; at the optimal cut-off point, Inception-v4 showed a sensitivity of 90% (135/150), which is much lower than the specificity (96%). In such conditions, some wrist AP radiographs with DRFs appeared to be misdiagnosed as normal images by the model, resulting in a delay in essential treatment for injured patients. After reviewing the 15 images with fractures, whose predicted values by the Inception-v4 model were less than the threshold (0.64), we found that 5 of them displayed an absence of apparent fracture traits (e.g., fracture lines or fracture fragment displacement). However, such traits were visible on the lateral radiographs corresponding to the AP images, as shown in Figure 6. The ensemble of model analyses using AP and lateral radiographs has the potential to enhance the sensitivity in fracture detection. In the total enrolled dataset, there was a statistically significant difference between the group of patients with DRFs and the group of patients without DRFs in sex (Table 1). This difference may affect the results of training and testing the CNN models due to difference in anatomical traits in distal radius between the male and the female (Oppermann et al. 2015). But

399

we cannot declare to what extent the effect of gender difference would be on the results in this study. There are several limitations in our study. First, the original sample size in our dataset was small. However, we did not increase the original sample size by obtaining new wrist radiographs from other medical centers to maintain uniformity in the image quality. This small sample size might restrict the improvement of the CNN’s performance in the training and test procedures. Data augmentation was used to address the sample size issue, since it can reduce over-fitting and improve performance (Wong et al. 2016). Second, the assessment of the diagnostic performance of the deep learning models was based on anterior–posterior wrist radiographs, so the procedure may not represent a practical scenario. Typically, at least 2 wrist radiographs (an anterior–posterior image and a lateral image) are obtained by the reader to review. We will investigate whether the performance of the CNN would improve when the dataset consists of anterior–posterior wrist radiographs and matched lateral radiographs in our next planned project. Finally, we trained Inception-v4 to simply distinguish images with DRFs from normal images. The deep learning algorithm could accurately classify proximal humerus fractures based on Neer’s classification on shoulder radiographs (Chung et al. 2018), so as part of our next project we will train the CNN model to classify DRFs based on 1 particular fracture classification system. In summary, the network exhibited a similar diagnostic capability to that of the orthopedists and a superior performance to that of the radiologists in distinguishing AP wrist radiographs with DRFs from normal radiographs under limited conditions. Further studies are required to determine the feasibility of applying the diagnostic network with the object detection algorithm as an auxiliary in clinical practice under extended conditions. Supplementary data The inclusion and exclusion criteria for this study and the Appendix are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674. 2019.1600125

Conception and design of this study: KG, YL, and DX. Collection and preparation of the datasets: KG, DX, YL, YS, TZ, KH, and KZ. Training and testing of models: YL, DX, MB, YS, TZ, and KH. Analysis and interpretation of the results and data: KG, YL, LP, and WW. Acta thanks Martin Gerdin Wärnberg, Max Gordon and Takaaki Urakawa for help with peer review of this study. Chung S W, Han S S, Lee J W, Oh K S, Kim N R, Yoon J P, Kim J Y, Moon S H, Kwon J, Lee H J, Noh Y M, Kim Y. Automated detection and classification of the proximal humerus fracture by using deep learning algorithm. Acta Orthop 2018; 89(4): 468-73.

400

Fujisawa Y, Otomo Y, Ogata Y, Nakamura Y, Fujita R, Ishitsuka Y, Watanabe R, Okiyama N, Ohara K, Fujimoto M. Deep learning-based, computeraided classifier developed with a small dataset of clinical images surpasses board-certified dermatologists in skin tumor diagnosis. Br J Dermatol 2019; 180(2): 373-81. He K, Zhang X, Ren S, Sun J. Delving deep into rectifiers: surpassing humanlevel performance on ImageNet classification. 2015 IEEE International Conference on Computer Vision; 2015. Hua K L, Hsu C H, Hidayati S C, Cheng W H, Chen Y J. Computer-aided classification of lung nodules on computed tomography images via deep learning technique. Onco Targets Ther 2015; 8: 2015-22. Kim D H, MacKinnon T. Artificial intelligence in fracture detection: transfer learning from deep convolutional neural networks. Clin Radiol 2018; 73(5): 439-45. Kumar S N, Lenin Fred A, Sebastin Varghese P. Compression of CT images using contextual vector quantization with simulated annealing for telemedicine application. J Med Syst 2018; 42(11): 218. Lakhani P, Sundaram B. Deep learning at chest radiography: automated classification of pulmonary tuberculosis by using convolutional neural networks. Radiology 2017; 284(2): 574-82. Li Y, Shen L. Skin lesion analysis towards melanoma detection using deep learning network. Sensors (Basel) 2018; 18(2). Mauffrey C, Stacey S, York P J, Ziran B H, Archdeacon M T. Radiographic evaluation of acetabular fractures: review and update on methodology. J Am Acad Orthop Surg 2018; 26(3): 83-93. Mitra A, Banerjee P S, Roy S, Roy S, Setua S K. The region of interest localization for glaucoma analysis from retinal fundus image using deep learning. Comput Methods Programs Biomed 2018; 165: 25-35. Nishio M, Sugiyama O, Yakami M, Ueno S, Kubo T, Kuroda T, Togashi K. Computer-aided diagnosis of lung nodule classification between benign nodule, primary lung cancer, and metastatic lung cancer at different image size using deep convolutional neural network with transfer learning. PloS One 2018; 13(7): e0200721. Olczak J, Fahlberg N, Maki A, Razavian A S, Jilert A, Stark A, Skoldenberg O, Gordon M. Artificial intelligence for analyzing orthopedic trauma radiographs. Acta Orthop 2017; 88(6): 581-86.

Acta Orthopaedica 2019; 90 (4): 394–400

Oppermann J, Bredow J, Beyer F, Neiss W F, Spies C K, Eysel P, Dargel J, Wacker M. Distal radius: anatomical morphometric gender characteristics. Do anatomical preshaped plates pay attention on it? Arch Orthop Trauma Surg 2015; 135 (1): 133–39. Ren S, He K, Girshick R, Sun J. Faster R-CNN: towards real-time object detection with region proposal networks. IEEE Transactions on Pattern Analysis and Machine Intelligence 2017; 39(6): 1137-49. Russakovsky O, Deng J, Su H, krause J, Satheesh S, Ma S, Huang Z, Karpathy A, Khosla A, Bernstein M, Berg A C, Li F F. ImageNet large scale visual recognition challenge. Int J Comput Vis 2015; 115(3): 211-52. Szegedy C, Ioffe S, Vanhoucke V, Alemi A A. Inception-v4, InceptionResNet and the impact of residual connections on learning. Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence (AAAI-17); 2017. Ting D S W, Cheung C Y, Lim G, Tan G S W, Quang N D, Gan A, Hamzah H, Garcia-Franco R, San Yeo I Y, Lee S Y, Wong E Y M, Sabanayagam C, Baskaran M, Ibrahim F, Tan N C, Finkelstein E A, Lamoureux E L, Wong I Y, Bressler N M, Sivaprasad S, Varma R, Jonas J B, He M G, Cheng C Y, Cheung G C M, Aung T, Hsu W, Lee M L, Wong T Y. Development and validation of a deep learning system for diabetic retinopathy and related eye diseases using retinal images from multiethnic populations with diabetes. JAMA 2017; 318(22): 2211-23. Urakawa T, Tanaka Y, Goto S, Matsuzawa H, Watanabe K, Endo N. Detecting intertrochanteric hip fractures with orthopedist-level accuracy using a deep convolutional neural network. Skeletal Radiol 2019; 48(2): 239-44. Waever D, Madsen M L, Rolfing J H D, Borris L C, Henriksen M, Nagel L L, Thorninger R. Distal radius fractures are difficult to classify. Injury 2018; 49(Suppl .1): S29-S32. Wang Z, Meng Q, Wang S, Li Z, Bai Y, Wang D. Deep learning-based endoscopic image recognition for detection of early gastric cancer: a Chinese perspective. Gastrointest Endosc 2018; 88(1): 198-99. Wong S C, Gatt A, Stamatescu V, McDonnell M D. Understanding data augmentation for classiﬁcation: when to warp? International Conference on Digital Image Computing: Techniques and Applications; 2016. Yap J, Yolland W, Tschandl P. Multimodal skin lesion classification using deep learning. Exp Dermatol 2018; 27(11): 1261-7.

Acta Orthopaedica 2019; 90 (4): 401–405

401

Pain in fibrous dysplasia: relationship with anatomical and clinical features Bas C J MAJOOR 1, Eva TRAUNMUELLER 2, Werner MAURER-ERTL 2, Natasha M APPELMAN-DIJKSTRA 3, Andrea FINK 2, Bernadette LIEGL 4, Neveen A T HAMDY 3, P D Sander DIJKSTRA 1, and Andreas LEITHNER 2 1 Department

of Orthopaedic Surgery, Leiden University Medical Center, Netherlands; 2 Department of Orthopaedics and Trauma, Medical University of Graz, Austria; 3 Department of Medicine, Division of Endocrinology & Centre for Bone Quality, Leiden University Medical Center, Netherlands; 4 Institute of Pathology, Medical University of Graz, Austria Correspondence: b.c.j.majoor@lumc.nl Submitted 2017-12-22. Accepted 2019-01-28.

Background and purpose — Fibrous dysplasia (FD) is a rare bone disorder associated with pain, deformities, and pathological fractures. The pathophysiological mechanism of FD-related pain remains ill-understood. We evaluated the degree of pain and the potential contributory factors in 2 patient cohorts from Austria and the Netherlands. Patients and methods — 197 patients (16–85 years) with FD (Graz n = 105, Leiden n = 92) completed a survey concerning the presence and severity of pain at their FD site. Sex, age, type of FD, and localization of FD lesions were examined for a relationship with the presence and severity of pain. Results — Of 197 patients from the combined cohort (61% female, mean age 49 (SD 16) years, 76% monostotic) who completed the questionnaires, 91 (46%) reported pain at sites of FD lesions. Severity of pain was higher in patients with lesions of the lower extremities and ribs compared with upper extremity or craniofacial lesions. Severe subtypes of FD (polyostotic/McCune–Albright syndrome) were more often associated with pain, often severe. Interpretation — Our data suggest that almost 50% of patients with FD report pain at FD sites, thus representing a major clinical manifestation of the disorder, importantly also in patients with monostotic lesions. Lesions in lower extremities and ribs were more painful.

Fibrous dysplasia (FD) is a congenital, non-inherited rare bone disorder, characterized by replacement of healthy bone by fibrous tissue limited to one bone (monostotic FD) or extending to multiple bones (polyostotic FD). Bony lesions may be single, asymptomatic, and accidentally detected at routine radiological examination, but may also be present in multiple skeletal sites and responsible for a wide range of clinical symptoms, predominantly bone pain, bone deformities, and pathological fractures (Harris et al. 1962). In severe cases skeletal manifestations may also be associated with extraskeletal manifestations in the form endocrinopathies or café-aulait skin patches in McCune–Albright syndrome or intramuscular myxomas in Mazabraud’s syndrome. In a previous study we have shown that pain is a major determinant of impaired quality of life in patients with FD (Majoor et al. 2017c). It has also been shown that FD pain is negatively age-related, suggesting that FD lesions may undergo agerelated changes with less prevalence and less severity of pain, as a patient gets older (Kelly et al. 2008 , Robinson et al. 2016). This notion is further supported by studies reporting lower fracture rates, denser and more sclerotic changes on plain radiography of FD lesions, and fewer characteristic histologic features of FD such as fibrotic changes and ill-woven bone texture in older patients (Leet et al. 2004, Kuznetsov et al. 2008). Despite this tendency for FD to become more quiescent as a patient ages, pain has also been reported to increase over time in some patients, possibly due to secondary arthritic changes in adjacent joints (Kelly et al. 2008). Management of pain remains problematic, as its underlying mechanism is unknown. We examined the prevalence and severity of pain in a combined cohort of 197 patients with FD from 2 specialized bone centers in Austria and the Netherlands. A further aim of the study was to examine the relationship between a

© 2019 The Author(s). Published by Taylor & Francis on behalf of the Nordic Orthopedic Federation. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Non-Commercial License (https://creativecommons.org/licenses/by/4.0) DOI 10.1080/17453674.2019.1608117

402

number of clinical and demographic factors and the presence and severity of pain.

Patients and methods Study design This study addresses the prevalence and severity of pain in FD and was conducted using a cross-sectional study design, including all patients with an established diagnosis of FD seen at the Medical University of Graz (MUG) between 1984 and 2016 and at the Leiden University Medical Center (LUMC) between 2012 and 2015 as identified from the respective centers’ hospital registries. The diagnosis of PFD was established on the basis of clinical and radiological features, with occasional histological and genetic confirmation. All identified patients were invited to take part in the study either by means of an interview (Graz cohort) or by completing a validated questionnaire (Leiden cohort). Additional demographic, clinical, and radiologic data were retrieved from the patients’ medical records and the 2 cohorts were combined into 1 large single cohort before analysis of data. Patients and methods 146 patients who were evaluated and treated at the MUG between 1984 and 2016 were approached by phone for an interview on the presence of pain on the basis of the validated Pain Numeric Rating Scale (PNRS), a standardized 11-step pain score validated for use in the assessment of pain in clinical trials (Hartrick et al. 2003). 138 patients who were seen at the outpatient clinic of the LUMC over a period of 3 years before the start of the study were invited by mail to complete the Brief Pain Inventory (BPI) questionnaire as previously described (Majoor et al. 2017c). Patients who did not respond to the questionnaires by mail were contacted by phone, with a maximum 2 attempts in case of no answer. Of 146 patients from the MUG cohort who were contacted by phone, 105 (72%) agreed to be interviewed by phone and of the 138 patients from the LUMC cohort who were invited to take part in the study by mail, 92 (67%) returned a completed BPI, resulting in a combined cohort of 197 patients in whom data on pain was available for analysis. Collected data included data on the presence of absence of pain (yes/no) and when present, the severity of current pain on a scale ranging from 0 to 10 with 0 indicating “no pain” and 10 indicating “the worst possible pain imaginable.” Data on sex, age, and type of FD (monostotic/polyostotic/ McCune–Albright syndrome/Mazabraud syndrome) were retrieved from the patients’ medical records at the respective medical centers. Data were also retrieved on the localizations of FD lesions including the craniofacial region, upper extremity, lower extremity including the pelvis, ribs, and spine. The extent of skeletal disease, as measured by the skeletal burden

Acta Orthopaedica 2019; 90 (4): 401–405

score (Collins et al. 2005), was available only in one center and therefore not included in the analysis. Statistics Statistical analysis was performed using SPSS Statistics 23.0 (IBM Corp, Armonk, NY, USA). Results are presented as mean (SD) or as median (intermediate range) and in the case of categorical data as percentages. Difference in pain between FD localizations (e.g., craniofacial, upper extremity, lower extremity, ribs, and spine) was assessed using the ANOVA test. Only monostotic patients were included in this sub-analysis in order to evaluate a potential difference in pain symptoms between different FD localizations. 5 patients with monostotic disease of the spine were excluded from this analysis due to low numbers. Other potential risk factors (e.g., sex, age, type of FD) were analyzed using logistic regression analysis (results presented as OR, p-value) for the presence of pain (yes/no), and with linear regression analysis (results presented as B-coefficient per degree of pain on VAS, p-value) for the extent of current pain on a scale from 0 to 10. Both analyses were primarily performed using univariable analysis followed by a multivariable analysis. Ethics, funding, and potential conflicts of interest Ethical approval was obtained from the Medical Ethics Committee of both centers participating in the study. This research was funded by a research grant from the Bontius Foundation of the Leiden University Medical Center for research into Fibrous Dysplasia. The authors have declared that they have no conflict of interest.

Results Patient characteristics (Table 1) Women predominated (120 women vs. 77 men). The majority of the patients (98%) were adults and all patients were skeletally mature (range 16–85 years). Mean age at the time of pain assessment was 49 years (SD 16), and mean overall followup was 16 years (SD 11). The majority of patients (76%) had monostotic FD, 20% had polyostotic FD, 5% had McCune– Albright syndrome, and 3% had Mazabraud’s syndrome. In the whole cohort, the lower extremity was the most common localization of FD (52%), followed by the craniofacial region (26%), ribs (19%), upper extremity (16%), and spine (11%). Data on medical and surgical treatment were not used in the analysis of factors potentially affecting presence or severity of pain, due to the heterogeneity in agents, doses, schedules, and duration of use of these agents in this combined FD cohort. Differences between the Dutch and Austrian FD cohorts (Table 1) Patients from the LUMC cohort were younger compared with patients from the MUG cohort. The LUMC cohort also

Acta Orthopaedica 2019; 90 (4): 401–405

403

Table 1. Cohort characteristics and differences between the two cohorts. Data pertaining to localization of FD lesions included the multiple lesions of polyostotic patients

Austrian

Dutch

P-value

Total

Invited patients 146 138 284 Included patients 105 92 197 response rate (%) 71.9 66.7 69.4 Women/men 60/45 60/32 0.2 120/77 Age a 51.5 (16.4) 46.1 (15.3) 0.02 49.0 (16.1) Follow-up a 15.1 (7.8) 16.7 (14.1) 0.3 15.8 (11.3) Type of FD, n (%) Monostotic 90 (86) 58 (63) < 0.001 149 (76) Polyostotic 12 () 26 (28) < 0.001 39 (20) McCune-Albright 1 (1) 8 (9) < 0.001 9 (5) Mazabraud’s syndrome, n (%) 1 (1) 5 (5) 0.1 6 (3) Localization of FD, n (%) Craniofacial 29 (28) 22 (24) 0.6 51 (26) Upper extremity 24 (23) 8 (9) 0.1 32 (16) Lower extremity 63 (60) 40 (45) 0.004 103 (52) Ribs 27 (26) 11 (3) 0.003 38 (19) Spine 11 (11) 11 (13) 0.9 22 (11) a Mean (SD) years Craniofacial site (n = 38) Pain present in 10 patients Mean pain score: 0.79 Rib site (n = 18) Pain present in 7 patients Mean pain score: 1.72

Upper extremity site (n = 19) Pain present in 5 patients Mean pain score: 0.89

Lower extremity site (n = 69) Pain present in 34 patients Mean pain score: 2.01

Figure 1. Presence and severity of pain in monostotic FD sites (n = 144). The presence of pain (yes/no) and the mean pain scores in the group of monostotic patients is highest in patients with lesions of the lower extremity and lowest in patients with craniofacial FD. Table 2. Factors that may contribute to pain in a combined cohort of 197 patients with FD. Possible risk factors for the presence pain were analyzed with logistic regression analysis (odds ratio, OR) and for the severity of pain with linear regression analysis (B-coefficient (B) per 1 degree of pain on VAS score) Pain (yes/no) (OR) Univariable analysis Multivariable analysis Severity of pain (B) Univariable analysis Multivariable analysis

Sex OR/B p-value

Age Type of FD OR/B p-value OR/B p-value

0.57 0.06 0.56 0.06

0.99 0.99

0.5 0.6

0.67 0.08 0.65 0.08

0.003 0.8 0.000 1.0

0.36 0.001 0.36 0.001 1.07 0.002 1.05 0.002

included more patients with polyostotic FD and McCune–Albright syndrome than the MUG cohort. Other demographic or clinical features were similar between the 2 cohorts. Differences between responders and nonresponders Distribution of age, sex, and type of FD was similar between responders and non-responders from the MUG cohort. Within the LUMC cohort, type of FD and age distribution were similar between responders and non-responders. However, there was a greater proportion of women who completed the questionnaire in the LUMC cohort. Prevalence and severity of pain Of the 197 survey responders, 46% reported having pain at the site of their FD lesions (Figure). In the group of patients reporting having pain at the time of the survey, the median pain score was 4 (1–9). A higher proportion of patients with monostotic lesions of the lower extremity (49%) or ribs (39%) reported having pain (yes/no) compared with patients who had monostotic lesions of the upper extremity (26%) or craniofacial region (26%), although these differences in prevalence of pain were not statistically significant (p = 0.09). In contrast, there was a statistically significant difference (p = 0.05) in severity of pain between monostotic lesions depending on their localization. Pain was thus reported to be most severe in patients with monostotic lesions of the lower extremity, mean 2.0 (SD 2.7), followed by ribs, mean 1.7 (SD 2.5), upper extremity mean 0.9 (SD 2.0), and lastly craniofacial lesions, mean 0.8 (SD 1.7). Potential risk factors for pain in fibrous dysplasia (Table 2) Univariable regression analysis showed that only the more severe types of FD were predictive for both the presence (OR 0.36, p = 0.001) and severity of pain (B-coefficient 1.07, p = 0.002). There was no relationship observed between sex or age and the presence or severity of reported pain. After correction for age and sex, the severe type of FD remained a statistically significant risk factor for the presence and severity of pain in multivariable analysis (respectively OR 0.36, p = 0.001 and B-coefficient 1.0, p = 0.002). This difference is clinically relevant, as patients with polyostotic FD are at risk of scoring 1 degree higher on VAS compared with patients with monostotic disease, as are patients with McCune–Albright of scoring 1 degree higher compared with polyostotic patients.

404

Discussion This survey on the prevalence and severity of pain in patients with FD from a large combined cohort from Austria and the Netherlands highlights the importance of this clinical manifestation of FD, with nearly half of the patients reporting having pain at the site of their FD lesions with a mean score of 4 out of a maximum of 10. In patients with monostotic lesions, pain was more often present and was more severe when the FD lesions were localized in the lower extremities or ribs compared with lesions localized in the craniofacial region or upper extremities. Severity of type of FD was predictive for the presence of pain and for its severity as expressed by higher pain scores. An 81% prevalence of pain has been reported among adult patients with FD, although a lesser prevalence of 49%, similar to our prevalence, was reported in a more recent study (Kelly et al. 2008, Benhamou et al. 2016). This discrepancy in findings might be explained by the composition of our combined cohort in which a high proportion of patients had monostotic FD (76%). 2 studies addressing Quality of Life (QoL) in FD have shown that patients with FD have lower scores in the bodily pain domain of the SF-36 compared with the general population (Kelly et al. 2005, Majoor et al. 2017c). We hypothesized that patients with limited monostotic FD would report less pain than those with more extensive polyostotic disease, or those with the more severe McCune–Albright syndrome. Similarly to previous reports, we found a prognostic value of type of FD (polyostotic/McCune–Albright syndrome) as regards quality of life and severity of pain. Monostotic patients with lesions of the lower extremities demonstrated the highest prevalence and severity of pain compared with lesions localized elsewhere in the skeleton as also reported by Kelly et al. (2008). Weight-bearing forces acting on the lower extremities combined with the poor quality of FD bone may result in deformities and pathological fractures, and thus a higher prevalence and severity of pain, particularly in the case of the femur (Nakashima et al. 1984, Leet et al. 2004, DiCaprio and Enneking 2005, Benhamou et al. 2016). FD lesions in bones of the lower extremities are also less likely to benefit fully from surgical interventions compared with lesions of non-weight bearing bones (DiCaprio and Enneking 2005, Majoor et al. 2017a). Although severe pain has been described in patients with craniofacial FD, we found a low prevalence and severity of pain in patients with craniofacial FD as reported by others (Chao and Katznelson 2008, Kelly et al. 2008, Makitie et al. 2008). Extensive skeletal involvement of FD, as measured by the Skeletal Burden Score, has been reported to be associated with high circulating levels of bone turnover markers and of FGF-23 and with impaired QoL, including pain, in a number of studies (Collins et al. 2005, Kelly et al. 2008, Majoor et al. 2017b). This suggests that the extent of FD lesions would be

Acta Orthopaedica 2019; 90 (4): 401–405

a major cause for the presence and severity of pain in patients with demonstrated high skeletal burden scores as seen in polyostotic FD and McCune–Albright syndrome. However, Majoor et al. (2017c) found no association between skeletal burden scores and pain. Other factors than extent of FD lesions, such as anatomical location and age-related changes, may also be responsible for the prevalence and severity of pain in patients with FD, also in those with the milder monostotic types, particularly of the lower extremities (Kelly et al. 2008). Indeed, we found that localization of FD lesions was a main determinant of the presence and severity of pain. Whereas the precise mechanism of pain in FD remains elusive, extra-skeletal factors may also play a role, as observed in patients with McCune–Albright syndrome, who report more pain than patients with polyostotic FD but with no extra-skeletal manifestations of FD. Neurogenic involvement may also play a role as alluded to by Chapurlat et al. (2012). These results imply that small, monostotic lesions may be associated with severe pain depending on their anatomical location. For example, a patient with a small monostotic lesion of the proximal femur may experience more pain than a patient with extensive disease of the humerus. The potential contributory role of extra-skeletal factors in the pathogenesis of pain in FD may help explain non-response or poor response to treatment with bone-modifying agents such as bisphosphonates. In our patients, age did not appear to influence the prevalence and severity of pain. However, Kelly et al. (2008) showed a higher prevalence and severity of pain in adults than in children with FD. The difference in prevalence and severity of pain before and after growth is completed suggests a contributing role for factors associated with growth in the pathophysiology of pain in FD. Women are generally believed to experience more pain than men (Berkley 1997). Similar to the report by Kelly et al. (2008), we found no sex difference in the severity of pain, although female FD patients in our cohort did report a higher prevalence of pain. Strengths and limitations One of the main strengths of our study is the inclusion of patients with different types of FD from 2 relatively large cohorts from different countries in whom data on the prevalence of pain and its severity were specifically and individually collected by interview or questionnaire. A further strength of the study is the predominance of patients with monostotic FD, compared with earlier studies including a lesser proportion of these patients, which allowed us to determine the relevance of anatomical localization of isolated FD lesions in the prevalence and severity of pain. Our study has also a number of limitations, including the use of 2 different, albeit comparable questionnaires, to calculate pain scores and the single time-point measurement of pain, as opposed to repeated measurements, which might have

Acta Orthopaedica 2019; 90 (4): 401–405

allowed us to demonstrate a time pattern for the pain. Because we included patients from two specialized bone-tumor centers, there might be a selection within our cohort towards the more severely affected FD patients and thus the patients with more pain. The hypothesis that the majority of monostotic FD patients have asymptomatic disease and will therefore never consult a physician, in combination with the possibility that painful patients are more accessible to participate in a study, might enhance this effect. However, due to the rarity of FD, this disorder is often, if not always, treated in specialized centers. Our cohort represents one of the largest cohorts of patients with FD ever reported, and more specifically the general population of FD patients that will seek counsel and/or that need treatment. Lastly, we did not include data on previous surgery and treatment with bisphosphonates in our analysis of pain in these patients. Conclusion We found that although the more severe types of FD (polyostotic/McCune–Albright syndrome) are predictive for presence and severity of pain, these are also determined by the localization of the lesions in the weight-bearing lower extremity and the rib lesions in patients with monostotic FD. These results have clinical implications in the management of patients with FD, as they highlight the fact that small, monostotic lesions may be the source of severe pain depending on their anatomical location.

BCJM and NMA-D were involved in acquisition, analysis, and interpretation of the data and in drafting the manuscript. ET, WM-E, AF, NAT-H, BL, PDSD, and AL were involved in acquisition and interpretation of the data and in drafting the manuscript. Acta thanks Johnny Keller and Aare Märtson for help with peer review of this study.

Benhamou J, Gensburger D, Messiaen C, Chapurlat R. Prognostic factors from an epidemiologic evaluation of fibrous dysplasia of bone in a modern cohort: the FRANCEDYS study. J Bone Miner Res 2016; 31: 2167-72. Berkley K J. Sex differences in pain. Behav Brain Sci 1997; 20(3): 371-80.

405

Chao K, Katznelson L. Use of high-dose oral bisphosphonate therapy for symptomatic fibrous dysplasia of the skull. J Neurosurg 2008; 109(5): 88992. Chapurlat R D, Gensburger D, Jimenez-Andrade J M, Ghilardi J R, Kelly M, Mantyh P. Pathophysiology and medical treatment of pain in fibrous dysplasia of bone. Orphanet J Rare Dis 2012; 24: 7(Suppl 1): S3. Collins M T, Kushner H, Reynolds J C, Chebli C, Kelly M H, Gupta A, et al. An instrument to measure skeletal burden and predict functional outcome in fibrous dysplasia of bone. J Bone Min Res 2005; 20(2): 219-26. DiCaprio M R, Enneking W F. Fibrous dysplasia: pathophysiology, evaluation, and treatment. J Bone Joint Surg Am 2005; 87(8): 1848-64. Harris W H, Dudley H R, Barry R J. The natural history of fibrous dysplasia: an orthopaedic, pathological, and roentgenographic study. J Bone Joint Surg Am 1962; 44-A: 207-33. Hartrick C T, Kovan J P, Shapiro S. The numeric rating scale for clinical pain measurement: a ratio measure? Pain Pract 2003; 3(4): 310-16. Kelly M H, Brillante B, Kushner H, Gehron Robey P, Collins M T. Physical function is impaired but quality of life preserved in patients with fibrous dysplasia of bone. Bone 2005; 37(3): 388-94. Kelly M H, Brillante B, Collins M T. Pain in fibrous dysplasia of bone: agerelated changes and the anatomical distribution of skeletal lesions. Osteoporos Int 2008; 19(1): 57-63. Kuznetsov S A, Cherman N, Riminucci M, Collins M T, Robey P G, Bianco P. Age-dependent demise of GNAS-mutated skeletal stem cells and “normalization” of fibrous dysplasia of bone. J Bone Min Res 2008; 23(11): 1731-40. Leet A I, Chebli C, Kushner H, Chen C C, Kelly M H, Brillante B A, et al. Fracture incidence in polyostotic fibrous dysplasia and the McCune– Albright syndrome. J Bone Min Res 2004; 19(4): 571-7. Majoor B C, Peeters-Boef M J, van de Sande M A, Appelman-Dijkstra N M, Hamdy N A, Dijkstra P D. What is the role of allogeneic cortical strut grafts in the treatment of fibrous dysplasia of the proximal femur? Clin Orthop Relat Res 2017a; 475(3): 786–95. Majoor B C, Appelman-Dijkstra N M, Fiocco M, van de Sande M A, Dijkstra P D, Hamdy N A. Outcome of long-term bisphosphonate therapy in McCune-Albright syndrome and polyostotic fibrous dysplasia. J Bone Min Res 2017b; 32(2): 264-76. Majoor B C J, Andela C D, Bruggemann J, van de Sande M A J, Kaptein A A, Hamdy N A T, et al. Determinants of impaired quality of life in patients with fibrous dysplasia. Orphanet J Rare Dis 2017c; 12(1): 80. Makitie A A, Tornwall J, Makitie O. Bisphosphonate treatment in craniofacial fibrous dysplasia: a case report and review of the literature. Clin Rheumatol 2008; 27(6): 809-12. Nakashima Y, Kotoura Y, Nagashima T, Yamamuro T, Hamashima Y. Monostotic fibrous dysplasia in the femoral neck: a clinicopathologic study. Clin Orthop Relat Res 1984; (191): 242-8. Robinson C, Collins M T, Boyce A M. Fibrous dysplasia/McCune-Albright syndrome: clinical and translational perspectives. Curr Osteoporos Rep 2016; 14(5): 178-86.

406

Acta Orthopaedica 2019; 90 (4): 406–409

Technical note

The development of an online implant manufacturer application: a knowledge-sharing platform for the Swedish Hip Arthroplasty Register Johanna VINBLAD 1,2, Daniel ODIN 1, Johan KÄRRHOLM 1,2, and Ola ROLFSON 1,2 1 The Swedish Hip Arthroplasty Register, Centre of Registers Västra Götaland; 2 Department of Orthopedics, Institute of Clinical Science, Sahlgrenska Academy, University of Gothenburg, Sweden Correspondence: johanna.vinblad@registercentrum.se Submitted 2018-11-09. Accepted 2019-03-23.

We describe the development of an application for assessing implant performance based on data in the Swedish Hip Arthroplasty Register (SHAR) designed for implant manufacturers.

Technique

outcome measures program preoperatively, and 1, 6, and 10 years after surgery. Patient-reported outcome measures are all maintained in the principal database. Lastly, the register contains a separate surgical environment database with hospitallevel aggregate information. For the development of the manufacturer application, we included variables related to diagnosis, component features, reason for surgery, and surgery outcome in terms of revisions. Patient-reported outcome measures and surgical environment variables were not included.

Data source SHAR (founded in 1979) is a national quality register covering all orthopedic units performing hip replacement surgery in Sweden. The completeness of registrations during the last 10 years varies between 97–98% for total hip replacements, Work process 95–97% for hemi-arthroplasties, and 91–95% for revision The project delivery model can be divided into 7 phases procedures when linking data to the Swedish national patient (Figure). Planning, development, and test were carried out in register (Kärrholm et al. 2018). The register currently covers iterative cycles to optimize product outcome. more than 360,000 hip procedures, 1,100,000 registered items and 6,100 unique components. AdminisIdea • Project leader assigned. trative staff at the units report surgi• Identification of key competences and stakeholders (registry directors, orthopedic surgeons, implant industry representatives, cal variables for primary procedures statiticians, and system developer). • Assembly of project team. and reoperations. Medical records • Identification of demands related to content, function, and output. copies covering admission, surgi• Defining scope. cal procedure, and discharge for all patients undergoing reoperations are Planning Development Test • Defining variables and • Short daily development • Internal tests and revision sent to the register for further extracnomenclature within the project. team meetings. of application. • Defining work packages. • Biweekly or monthly project • End-user tests. tion of data. 57 variables are col• Assessment of technical feasibility. team meetings. lected regarding the primary surgery and 99 variables for reoperations. Launch The register also retains a separate • Free trial period for implant manufactures. component database with variables describing attributes of the implant. Closure of development project The component database contains 141 variables. Additionally, the Initiation of continuous improvement project register comprises 54 variables collected through the patient-reported The project delivery model. Lorem ipsum

Lorem ipsum

Acta Orthopaedica 2019; 90 (4): 406–409

Platform specification The design of the application is based on the IT platform new Stratum and scripts for the R free software environment (https://www.r-project.org/) for statistical computing and graphics. Stratum is a technical platform for describing, collecting, and presenting data from quality registers in healthcare developed by the Centre of Registers, Västra Götaland. The platform provides a register and its users with a range of features for continuous quality improvement and follow-up, including advanced form management, statistics engine (in R), and visualization support (user interface components from ExtJS, developed by Sencha [https://www.sencha.com/]). Validation process The validation was done in several steps. The back-end code, which was written in R, was reviewed by 2 statisticians. Repeated quality assurance tests were performed. Comparisons were carried out between raw data and R-script outcome as well as between outcome of the R-scripts and outcome in the interface. Finally, the interface was tested by several persons not involved in the development of the project to find inconsistencies. Access/login requirements User access to data is restricted. The application displays only data on implants for the specific company, which the user represents. An electronic personal identification system widely used in Sweden, Mobile BankID, is used to confirm identity and securely log in to the application. Restriction of data usage In order to access the application, a contract must be in place between the industry and the register. This agreement restricts the use of data. The analyses performed may be used within the company for internal use, for regulatory purposes, and for marketing purposes as reported to regulatory agencies. The SHAR charges an annual subscription fee for the service based on number of implants registered during the previous year in addition to a fixed basic rate.

Product description Using previous experience from industry collaboration in combination with discussions between industry and registry as well as in-house orthopedic expertise, key areas were identified. These were consolidated into 4 modules in the application: volume, revised implants, implant survival, and market share. The 4 modules will be further discussed below, and a summary of the choices and filters can be accessed in Tables 1–4, see Supplementary data. Volume Broken down by hospitals, the volume module displays number of implants registered based on catalogue number

407

for a selected period. There are several data-filtering options available. Type of surgery is one of them, allowing for filtering with regards to primary or revision surgery. Other filtering opportunities are type of prosthesis and group of implant. The date range is variable between 1999 and current date with 28 days being the minimum number of days that you can choose (Table 1, see Supplementary data). Revisions per implant The 2nd module presents number of inserted and revised implants on article number level in Sweden. The filtering option “Type of surgery” allows the user to explore revisions after primary surgery and/or all re-revisions after revision surgery. Search criteria allow for selection of either total or hemi-arthroplasty as well as implant family as previously defined by the register. It is also possible to select type of revision; all 1st-time revisions, 1st stem revision, 1st cup revision, and 1st revision of other kind. For example, if 1st stem revision is selected, then any revision that is not a stem revision will be disregarded. In this module, there is also a possibility to focus on cause for revision. The returned revision data will be divided into revision occurring 0–90 days, 91 days to 2 years, and more than 2 years after the surgery. The date range may be set between any time point from 1999 to current date and the shortest date range is 1 year (Table 2, see Supplementary data). Implant survival The 3rd module displays Kaplan–Meier and cumulative incidence survival graphs for stems and/or cup families based on 1st revision after primary surgery. The survival graphs can be tailored for specific needs by choosing patient population based on diagnosis, type of prosthesis, type of revision, and cause for revision (Table 3, see Supplementary data). In this module, the stems and cups are grouped, since there are usually not a sufficient number of observations related to a specific article number (e.g., a specific cup design with a specific size) to perform a robust analysis. On the other hand, if a specific cup design is selected for analysis all article numbers included (e.g., sizes) can easily be extracted. It is possible to choose 1 or more implant families to be analyzed in 1 group. A specified group of company stems can be combined with a corresponding group of cups without restrictions. It is also possible to combine a specified group of stems with all company cups and vice versa and compare combinations of stems and cups from one’s own company with aggregated data on hip prostheses from all other suppliers. A number of revision outcomes can be defined such as all 1st-time revisions, 1st stem revision, 1st cup revision, and other types of 1st-time revisions. The Kaplan–Meier analyses provide estimates of the probability of a selected implant or implant group being revised at given time points. Competing risk-based cumulative incidence, on the other hand, will visualize the proportion of implants that have been revised and the proportion of patients who have died at given time points. The date range may be set between any time

408

point from 1999 to current date and the shortest date range is 1 year. The primary surgeries performed in the chosen time interval are followed up until current date or until the number of hips at risk is below 50. In addition, 95% confidence intervals are visualized in the graphs. Market share The last module in the application addresses market share for a selected type of implant (i.e. cup, stem, head, liner, and distal plug) from the company. It is also possible to filter with regards to type of prosthesis (i.e., total or hemi arthroplasty) and fixation. In addition to market share, manufacturers may also access the total number of registered implants in Sweden as compared with all implants used for the specific company (Table 4, see Supplementary data). Outputs All aggregated results are presented in tables and graphs may be downloaded to excel.

Discussion The overall aim with quality registers is to sustain and improve healthcare for patients. The Swedish Hip Arthroplasty Register is working with a wide group of stakeholders in order to ensure delivery of high-quality healthcare. Strong collaboration between the registry and the industry is paramount. Early detection of implants with substandard performance is important for the industry, the healthcare system and patients. There are values in the form of saving patient suffering as well as a high economic value in detecting failing implants early on. Several ongoing international initiatives aim to monitor and assess implants survival, such as the Orthopaedic Data Evaluation Panel (ODEP, http://www.odep.org.uk/), Beyond Compliance (http://www.beyondcompliance.org.uk/), and Arthroplasty Watch (http://www.arthroplastywatch.com/). The consequences of using an evidence-based system for rating of implants in the UK are demonstrated by Ng Man Sun et al. (2013). That paper highlights that healthcare providers in the UK follow recommendations based on clinical evidence regarding choice of implants. The Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) has previously reported a method of detecting prostheses with a higher than expected rate of revision, so-called “outlier” prostheses (de Steiger et al. 2013). The coexistence of several robust systems for detecting failing implants should increase the likelihood of as early detection as possible. Limitations Early detection of failing implants or product development may require different statistical robustness of data. In some

Acta Orthopaedica 2019; 90 (4): 406–409

cases, one would wish to access data as early as possible to look at trends and for other activities, and the data must be solid and robust. When discussing this, patient data security is also an important factor. We have included time-range restrictions and minimal number of hips at risk to allow for any analysis in order to ensure patient confidentiality as well as data robustness. In any real-time search application it is important to keep in mind the register validation process. At the beginning of a new year the data going into the annual report for the previous year are validated in several steps to ensure high quality. During the year, administrative staff at the hospital carry out registrations. The time between surgery and registration will to a certain extent differ between hospitals, which means that there is an inherent uncertainty in analyses based on aggregated data collected during the last weeks. When the annual report has been published for a specific year, the data can be regarded as well validated up to this specific year. Until then the data available from the latest calendar year should be considered as preliminary data. When comparing data generated in the application one should consider that patient population characteristics, for example age, sex, and diagnosis, might differ between data sets and this could potentially influence the results. In conclusion, the sharing of data between register and manufacturer comes with a responsibility. The manufacturers must be aware of limitations and take the responsibility when presenting the data. There is a critical balance between early access to data with the intention to alert regarding failing implants, and delivering robust high-quality data. All relevant stakeholders must be aware of this and use the data appropriately. This is partly addressed in the contract as data can only be used for marketing purposes as reported to the regulatory agency. To summarize, a well-established collaboration between the registry and the industry is not only beneficial for industry but also for the register, the orthopedic profession and not least the patient. Poorly performing implants can indeed be identified without involvement of the industry, but we think that this application will increase their interest in this process. We hope that our newly developed application will stimulate a collaboration to find the true background behind substandard implant performance, which may or may not be related to the properties of the prosthesis or the implant part of interest itself. The registry will continue to build on the application and will continue with yearly meetings with the industry in order to share knowledge and develop the collaborations further. Supplementary data Tables 1–4 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674. 2019.1608094

Acta Orthopaedica 2019; 90 (4): 406–409

OR, JK, and JV conceived the project. DO developed scripts and performed testing of statistical methods. All authors were involved in the internal testing. JV and OR drafted the manuscript. Acta thanks Geke Denissen and Stephen Ellis Graves for help with peer review of this study.

409

de Steiger R N, Miller L N, Davidson D C, Ryan P, Graves S E. Joint registry approach for identification of outlier prostheses. Acta Orthop 2013; 84(4): 348-52. Kärrholm J, Mohaddes M, Odin D, Vinblad J, Rogmark C, Rolfson O. Svenska Höftprotesregistret, Årsrapport 2017. 2018. Ng Man Sun S, Gillott E, Bhamra J, Briggs T. Implant use for primary hip and knee arthroplasty: are we getting it right first time? J Arthroplasty 2013; 28(6): 908-12.

4/19 ACTA ORTHOPAEDICA

Element of success in joint replacement

www.heraeus-medical.com

Vol. 90, No. 4, 2019 (pp. 297–409)

Volume 90, Number 4, August 2019