Acta Orthopaedica issue 2 / 2020 by Acta

2/20 ACTA ORTHOPAEDICA

Element of success in joint replacement

Vol. 91, No. 2, 2020 (pp. 121–220)

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.

www.heraeus-medical.com

Volume 91, Number 2, April 2020

TF-IORT200016.indd 1

07-02-2020 02:31:55

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. 91, No. 2, 2020

SUBSCRIPTION INFORMATION Acta Orthopaedica [print 1745-3674, online 1745-3682] is a peerreviewed journal, published six times a year plus supplements by Taylor & Francis on behalf of Nordic Orthopaedic Federation.

Airfreight and mailing in the USA by agent named WN Shipping USA, 156-15, 146th Avenue, 2nd Floor, Jamaica, NY 11434, USA. Periodicals postage paid at Jamaica NY 11431.

Annual Institutional Subscription, Volume 91, 2020

US Postmaster: Send address changes to Acta Orthopaedica, WN Shipping USA, 156-15, 146th Avenue, 2nd Floor, Jamaica, NY 11434, USA.

$1,291

£798

€1,035

The subscription fee purchases an online subscription. The price includes access to current content and back issues to January 1997 (if available). Printed copies of the journal are provided on request as a free supplementary service accompanying an online subscription. Supplements to the journal are also included in the subscription price. For more information, visit the journal’s website: http://www.tandfonline.com/IORT Manuscripts should be uploaded at http://www.manuscriptmanager.com/ao/ for further handling at: Acta Orthopaedica Editorial Office, Department of Orthopaedics, Lund University Hospital, SE-221 85 Lund, Sweden Correspondance concerning copyright and permissions should be sent to: Maria Montzka, Portfolio Manager – Medicine P.O. Box 3255, SE-103 65 Stockholm, Sweden, Tel: +46 (0)760 14 24 68. Fax: +46 (0)8 440 80 50. E-mail: maria.montzka@informa.com Ordering information: Please contact your local Customer Service Department to take out a subscription to the Journal: USA, Canada: Taylor & Francis, Inc., 530 Walnut Street, Suite 850, Philadelphia, PA 19106, USA. Tel: +1 800 354 1420; Fax: +1 215 207 0050. UK/ Europe/Rest of World: T&F Customer Services, Informa UK Ltd, Sheepen Place, Colchester, Essex, CO3 3LP, United Kingdom. Tel: +44 (0) 20 7017 5544; Fax: +44 (0) 20 7017 5198; Email: subscriptions@tandf.co.uk Dollar rates apply to all subscribers outside of Europe. Euro rates apply to all subscribers in Europe except the UK and Republic of Ireland. If you are unsure which applies, contact Customer Services. All subscriptions are payable in advance and all rates include postage. Journals are sent by air to the USA, Canada, Mexico, India, Japan and Australasia. Subscriptions are entered on an annual basis, i.e., January to December. Payment may be made by sterling check, US dollar check, euro check, international money order, National Giro, or credit card (Amex, Visa and Mastercard). Back issues: Taylor & Francis retains a two-year back issue stock of journals. Older volumes are held by our official stockists to whom all orders and enquiries should be addressed: Periodicals Service Company, 351 Fairview Ave., Suite 300, Hudson, New York 12534, USA. Tel: +1 518 537 4700; fax: +1 518 537 5899; e-mail: psc@ periodicals.com.

Subscription records are maintained at Taylor & Francis Group, 4 Park Square, Milton Park, Abingdon, OX14 4RN, United Kingdom.

Copyright © 2020 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-nc/3.0 . Informa UK Limited, trading as Taylor & Francis Group makes every effort to ensure the accuracy of all the information (the “Content”) contained in its publications. However, Informa UK Limited, trading as Taylor & Francis Group, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Informa UK Limited, trading as Taylor & Francis Group. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Informa UK Limited, trading as Taylor & Francis Group shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. Terms & Conditions of access and use can be found at http://www.tandfonline. com/page/terms-and-conditions Indexed/abstracted in: Allied and Complementary Medicine Library (Amed); ASCA (Automatic Subject Citation Alert); Biological Abstracts; Chemical Abstracts; Cumulative Index to Nursing and Allied Health Literature(CINAHL); Current Advances in Ecological and Environmental Sciences; Current Contents/Clinical Medicine; Current Contents/Life Sciences; Developmental Medicine and Child Neurology; Energy Research Abstracts; EMBASE/ Excerpta Medica; Faxon Finder; Focus On: Sports Science & Medicine; Health Planning and Administration; Index Medicus/MEDLINE; Index to Dental Literature; Index Veterinarius; INIS Atomindex; Medical Documentation Service; Nuclear Science Abstracts (Ceased); Periodicals Scanned and Abstracted. Life Sciences Collection; Research Alert; Science Citation Index; SciSearch; SportSearch; Uncover Veterinary Bulletin. Printed in England by Henry Ling

Acta Orthopaedica

ISSN 1745-3674

Vol. 91, No. 2, April 2020 Editorial Where do you want your drugs delivered?

121

M Ding and I Hvid

Perspective Further refinement of surgery will not necessarily improve outcome after hip fracture

123

C Rogmark

Annotion Recycling implants: a sustainable solution for musculoskeletal research

125

L Lidgren, D B Raina, M Tägil, and K E Tanner

Hydroxyapatite platform Synthetic hydroxyapatite: a recruiting platform for biologically active molecules

126

D B Raina, Y Liu, H Isaksson, M Tägil, and L Lidgren

133

S Hansson, E Bülow, A Garland, J Kärrholm, and C Rogmark

139

S Svenøy, L O Watne, I Hestnes, M Westberg, J E Madsen, and F Frihagen M H Kristoffersen, E Dybvik, O M Steihaug, T B Kristensen, L B Engesæter, A H Ranhoff, and J-E Gjertsen

Hip More hip complications after total hip arthroplasty than after hemiarthroplasty as hip fracture treatment: analysis of 5,815 matched pairs in the Swedish Hip Arthroplasty Register Results after introduction of a hip fracture care pathway: comparison with usual care Cognitive impairment influences the risk of reoperation after hip fracture surgery: results of 87,573 operations reported to the Norwegian Hip Fracture Register Perioperative, short-, and long-term mortality related to fixation in primary total hip arthroplasty: a study on 79,557 patients in the Norwegian Arthroplasty Register Early stabilization of the uncemented Symax hip stem in a 2-year RSA study Outcome of revision hip arthroplasty in patients younger than 55 years: an analysis of 1,037 revisions in the Dutch Arthroplasty Register Radiographic parameter-driven decision tree reliably predicts aseptic mechanical failure of compressive osseointegration fixation Knee Reduced survival of total knee arthroplasty after previous unicompartmental knee arthroplasty compared with previous high tibial osteotomy: a propensity-score weighted mid-term cohort study based on 2,133 observations from the Danish Knee Arthroplasty Registry The effect of fixation type on the survivorship of contemporary total knee arthroplasty in patients younger than 65 years of age: a register-based study of 115,177 knees in the Nordic Arthroplasty Register Association (NARA) 2000–2016 Ankle Better implant survival with modern ankle prosthetic designs: 1,226 total ankle prostheses followed for up to 20 years in the Swedish Ankle Registry Children Overgrowth of the lower limb after treatment of developmental dysplasia of the hip: incidence and risk factors in 101 children with a mean follow-up of 15 years Development of a risk score for scoliosis in children with cerebral palsy Gorham-Stout disease Gorham–Stout disease: good results of bisphosphonate treatment in 6 of 7 patients Deep learning Deep learning in fracture detection: a narrative review Information to authors (see http://www.actaorthop.org/)

146 152

H Dale, S Børsheim, T B Kristensen, A M Fenstad, J-E Gjertsen, G Hallan, S A Lie, and O Furnes

159

D S M G Kruijntjens, L Koster, B L Kaptein, L M C Jutten, J J Arts, and R H M ten Broeke M F L Kuijpers, G Hannink, L N van Steenbergen, and B W Schreurs

165 171

R Kagan, L Parlee, B Beckett, J B Hayden, K R Gundle, and Y-C Doung

177

A El-Galaly, P T Nielsen, A Kappel, and S L Jensen

184

M J Niemeläinen, K T Mäkelä, O Robertsson, A W-Dahl, O Furnes, A M Fenstad, A B Pedersen, H M Schrøder, A Reito, and A Eskelinen

191

A Undén, L Jehpsson, I Kamrad, Å Carlsson, A Henricson, M K Karlsson, and B E Rosengren

197

C Yoon, C H Shin, D O Kim, M S Park, W J Yoo, C Y Chung, I H Choi, and T-J Cho

203

K Pettersson, P Wagner, and E Rodby-Bousquet

209

K N Schneider, M Masthoff, G Gosheger, S Klingebiel, D Schorn, J Röder, T Vogler, M Wildgruber, and D Andreou

215

P H S Kalmet, S Sanduleanu, S Primakov, G Wu, A Jochems, T Refaee, A Ibrahim, L v. Hulst, P Lambin, and M Poeze

Acta Orthopaedica 2020; 91 (2): 121–122

121

Editorial

Where do you want your drugs delivered?

Promoting bone repair and regeneration in bone fractures by local delivery of growth factors possessing osteoinductive activity has been extensively investigated with significant advancements. Methods such as a combination of biomaterial-based scaffold and local bone active molecule delivery has been used (Raina et al. 2019b). Unfortunately, the local delivery approaches most often require surgery, and for some complex clinical fractures it is not possible to apply an invasive strategy because of access restrictions. Thus, to achieve efficient local concentrations and reduce systemic side effects, an ideal alternative method would be systemically administered targeted delivery of drugs due to ease of handling and precise spatiotemporal compatibility at fracture sites or sites of bone regeneration. However, the combination of delivery of osteopromotive molecules locally and systemically to enhance bone regeneration has been reported with limited success. Recently, Raina and co-workers demonstrate a novel method using synthetic hydroxyapatite (HAP) particles as a recruiting moiety for different drug classes administered systemically, showing that their affinity to HAP binding sites can activate the particulate material to exert a biological effect (Raina et al. 2019a). Systemically administered biomolecules (zoledronic acid, tetracycline and 18F-fluorine) all sought the HAP moiety placed in a muscle pouch. Significantly higher peri-implant bone volume and peak force were observed around an implant containing HAP particles relative to an empty implant. It was found possible to reload HAP particles on as-needed basis. The uptake of the antibiotic tetracycline was observed in the biomaterial by fluorescence microscopy, and the uptake of the radioemitter 18F-fluorine was documented in the biomaterial by positron emission tomography/computed tomography (PET/CT). Thus, the targeted accretion in locally implanted particulate HAP is achieved by systemic drug administration, loading the biomaterial which is biologically activated (Raina et al. 2019a). Importantly, several issues need to be controlled to achieve successful results. First, the targeted delivery of drugs should have a specific and high affinity to HAP binding sites, where micro- to nanoparticles of HAP at the site acting as both a carrier and a recruiting moiety for systemically administered drugs. Examples are bone-seeking drugs like bisphosphonates

for bone regeneration, HAP binding antibiotics like tetracycline for infection (Perrin 1965), and bone-seeking radioactive isotopes like 186Re for metastatic bone disease (de Klerk et al. 1992). Tetracycline can also be used for dynamic histomorphometry of bone. Second, the timing of the injection may be critical. In rat models of fracture healing, the timing between 1 and 2 weeks of a single systemic dose of zoledronic acid plays an important role in the modulation of callus properties and for HAP deposition in a fracture callus collagen network, and fits well with the fluid mechanics necessary for drug transport (Amanat et al. 2007). The current promising results using HAP as a recruiting and reloadable particulate apatite moiety to which systemically administered drugs circulating in the bloodstream could bind due to a high chemical affinity provides a novel method for treatment of various bone diseases. Apart from its promise for bone regeneration, it is also potentially applicable in the treatment of bone infections, tumors, and osteoporosis. Staphylococcus aureus for example is a causative agent of osteomyelitis and has a high affinity for bone, and can induce osteonecrosis and resorption of bone matrix (Lucke et al. 2003). A novel therapeutic approach may be by using nano- or micro-HA particles already functionalized with antibiotics, to provide extended sustained local antibiotic delivery during and after radical surgery, and later reloading HA particles on as-needed basis. A similar method can apply to metastatic bone disease by using radioactive isotopes and chemotherapeutics for treatment of locally malignant tumors or solitary metastasis (Raina et al. 2019a). Since HAP is already abundant in bone, implantation of HAP would not necessarily be required. Several bone mineral seekers are known apart from bisphosphonates, tetracycline, and fluorine—certain peptides for example (Rotman et al. 2018). The challenge of the future would be to find boneseeking molecules or nanoconstructs that can be systemically administered, can carry the appropriate drug or pro-drug, can circulate in the body without significant global effects, can bind to the bone site of interest (which might be the whole skeleton, e.g., in osteoporosis), and exert its action directly or by pro-drug activation. Certainly, much work needs to be performed before safe and efficient drug delivery systems to bone become useful on a larger scale.

© 2020 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 10.1080/17453674.2019.1707951

122

Ming Ding 1 and Ivan Hvid 2 1 Orthopaedic

Research Unit, Department of Orthopaedic Surgery and Traumatology, Odense University Hospital, and Department of Clinical Research, University of Southern Denmark 2 Section of Childrenâ&#x20AC;&#x2122;s Orthopaedics and Reconstructive Surgery, Division of Orthopaedic Surgery, Oslo University Hospital Email: Ming.ding@rsyd.dk

Acta Orthopaedica 2020; 91 (2): 121â&#x20AC;&#x201C;122

Amanat N, McDonald M, Godfrey C, Bilston L, Little D. Optimal timing of a single dose of zoledronic acid to increase strength in rat fracture repair. J Bone Miner Res 2007; 22: 867-76. de Klerk J M, van Dijk A, van het Schip A D, Zonnenberg B A, van Rijk PP. Pharmacokinetics of rhenium-186 after administration of rhenium186-HEDP to patients with bone metastases. J Nucl Med 1992; 33: 646-51. Lucke M, Schmidmaier G, Sadoni S, Wildemann B, Schiller R, Stemberger A, Haas N, Raschke M. A new model of implant-related osteomyelitis in rats. J Biomed Mater Res B Appl Biomater 2003; 67: 593-602. Perrin D D. Binding of tetracyclines to bone. Nature 1965; 208: 787-8. Raina D B, Liu Y, Isaksson H, Tagil M, Lidgren L. Synthetic hydroxyapatite: a recruiting platform for biologically active molecules. Acta Orthop 2019a; Nov 4: 1-9. doi: 10.1080/17453674.2019.1686865. Raina D B, Qayoom I, Larsson D, Zheng M H, Kumar A, Isaksson H, Lidgren L, Tagil M. Guided tissue engineering for healing of cancellous and cortical bone using a combination of biomaterial based scaffolding and local bone active molecule delivery. Biomaterials 12019b; 88: 38-49. Rotman S G, Grijpma D W, Richards R G, Moriarty T F, Egin D, Guillaume O. Drug delivery systems functionalized with bone mineral seeking agents for bone targeted therapeutics. J Control Release 2018; 268: 88-99.

Acta Orthopaedica 2020; 91 (2): 123–124

123

Perspective

Further refinement of surgery will not necessarily improve outcome after hip fracture A displaced femoral neck fracture needs urgent surgery. More than 10 years ago hip replacement was found to be beneficial over internal fixation (Rogmark and Johnell 2007), but we still debate whether it should be a hemiarthroplasty (HA) or a total hip arthroplasty (THA). In a recent issue of Acta Orthopaedica, Hansson et al. (2019) pinpoint the dilemma: THA as fracture treatment is associated with more hip complications, mainly dislocations. But our research group has previously noted fewer revision surgeries after THA compared with HA (Hansson et al. 2017). So did 2 other register studies, from Canada and UK (Ravi et al. 2019; Metcalfe et al. 2019). In contrast, Dutch register data found THA to be associated with a higher revision rate (Moerman et al. 2018). Yet another UK register paper had similar revision rates for the 2 methods (Jameson et al. 2013), but they supported the current Swedish study (Hansson et al. 2019) when describing a higher dislocation rate after THA. Can the clinical studies guide us? The HEALTH study is an ambitious international project (HEALTH Investigators 2019). This randomized trial, including 1,495 patients, cannot show any clear differences between the 2 methods. The somewhat better functional results after THA may be outweighed by slightly more complications. However, one has to question the external validity of the HEALTH study. On average, just over 2 patients per each of the 80 participating hospitals and years were recruited during the study period. So few, from an otherwise large patient group, signals a selection bias. A downside of an HA may be development of acetabular erosion. Both the short-term follow-up of 2 years and the lack of radiological follow-up mean that this condition is not covered at all by the HEALTH study. Other randomized trials on the topic confirm a lower reoperation risk but higher dislocation risk after THA, without any differences in mortality or infection. In terms of patient-reported outcome, THA leads to better results (in terms of functional scores) than HA in some, but not all, RCTs (Lewis et al. 2019). Three things may explain the contradictory results: selection bias, performance bias, or simply that the implant does not matter that much. Many of the register papers find an association between THA and lower mortality. It is not plausible that an added ace-

tabular cup per se should protect against death. It would take a substantial gain in functional outcome after THA, compared with HA, to affect general health and risk of dying. If anything, the longer surgery and higher blood loss associated with THA would increase the risk of death. Even after adjusting for comorbidity and other factors, residual confounding seems to explain the lower mortality after THA, as discussed by Hansson et al. (2019). Choice of implant is most likely influenced by the patient’s degree of frailty and physical activity, factors not available in any register. Frail and incapacitated individuals, more often selected for HA, also suffer a higher risk of dying. An indication that the surgical procedure in itself is not decisive for the risk of dying is that mortality rates after femoral neck fracture have been unchanged for the last 3 decades (Mundi et al. 2014)—a period that has seen many surgical developments taking place. A worse general condition of the typical HA patient will also interfere with the choice to perform revision surgery in case of any complication. The frailer the patient, the more reluctant she and/or the surgeon will be to revise the implant. On the other hand, we may have a lower threshold to revise an HA than a THA. In the case of dislocation or erosion, conversion of HA to THA appears relatively easy, by adding a socket. A troublesome THA, without apparent malpositioning, might more often be left in place. In any case, revision is a very blunt outcome measure after arthroplasty in the fracture population. Dislocation—more common after THA in both register and clinical studies—is a serious complication for the elderly. A second dislocation leads to a permanent loss of health-related quality-of-life (Enocson et al. 2009). The increased risk of hip complications in general, and dislocation in particular should make us think twice before widening the indications for THA as fracture treatment. Arthroplasty surgeons are more often proponents of THA. In fracture cases, the THA procedure is technically demanding on the surgeon and a high annual volume is needed for a good result. HA is considered to be more “forgiving” surgery. As hip fracture surgery may be done outside office hours and is often considered a newcomer’s task, an easy surgical technique is preferable. Hospitals may have an acute or elective profile, which may influence the result of scientific studies.

124

I strongly believe that now is the time for us as orthopedic surgeons to raise our heads from the detailed comparisons of 2 well-functioning surgical procedures and take more responsibility for the entire clinical pathway. Together with the multidisciplinary team of nurses, geriatricians, physiotherapists, GPs, and others, we have to work in an evidence-based manner to support the individual’s recovery from a hip fracture (Kammerlander et al. 2010). The aforementioned unchanged mortality is one indication that things outside surgery should be done better! Any subtle gain in function and satisfaction due to THA surgery will be lost if post-discharge rehabilitation is not structured and provided over several months. After leaving the operating theater, there is still a lot left to do! Cecilia Rogmark Department of Orthopaedics, Skåne Univerisity Hospital, Malmö, Sweden email: cecilia.rogmark@skane.se Enocson A, Pettersson H, Ponzer S, Törnkvist H, Dalén N, Tidermark J. Quality of life after dislocation of hip arthroplasty: a prospective cohort study on 319 patients with femoral neck fractures with a one-year follow-up. Qual Life Res 2009; 18(9): 1177-84. Hansson S, Nemes S, Kärrholm J, Rogmark C. Reduced risk of reoperation after treatment of femoral neck fractures with total hip arthroplasty: a matched pair analysis. Acta Orthop 2017; 88(5): 500-4.

Acta Orthopaedica 2020; 91 (2): 123–124

Hansson S, Bülow E, Garland A, Kärrholm J, Rogmark C. More hip complications after total hip arthroplasty than after hemiarthroplasty as primary hip fracture treatment: analysis of 5 815 matched pairs in the Swedish Hip Arthroplasty Register. Acta Orthop 2019; Nov 18: 1-6. [Epub ahead of print] HEALTH Investigators. Total hip arthroplasty or hemiarthroplasty for hip fracture. NEJM 2019. DOI: 10.1056/NEJMoa1906190. Jameson S S, Lees D, James P, Johnson A, Nachtsheim C, McVie J L, Rangan A, Muller S D, Reed M R. Cemented hemiarthroplasty or hip replacement for intracapsular neck of femur fracture? A comparison of 7732 matched patients using national data. Injury 2013; 44(12): 1940-44. Kammerlander C, Roth T, Friedman S M, Suhm N, Luger T J, KammerlanderKnauer U, Krappinger D, Blauth M. Ortho-geriatric service: a literature review comparing different models. Osteop Int 2010; 21(4): 637-46. Lewis D P, Wæver D, Thorninger R, Donnelly W J. Hemiarthroplasty vs total hip arthroplasty for the management of displaced neck of femur fractures: a systematic review and meta-analysis. J Arthroplasty 2019; 34(8): 1837-43. Metcalfe D, Judge A, Perry D C, Gabbe B, Zogg C K, Costa M L. Total hip arthroplasty versus hemiarthroplasty for independently mobile older adults with intracapsular hip fractures. BMC Musculoskelet Disord 2019; 20(1): 226. Moerman S, Mathijssen N M, Tuinebreijer W E, Vochteloo A J, Nelissen R G. Hemiarthroplasty and total hip arthroplasty in 30,830 patients with hip fractures: data from the Dutch Arthroplasty Register on revision and risk factors for revision. Acta Orthop 2018; 89(5): 509-14. Mundi S, Pindiprolu B, Simunovic N, Bhandari M. Similar mortality rates in hip fracture patients over the past 31 years, Acta Orthop 2014; 85(1): 54-9. Ravi B, Pincus D, Khan H, Wasserstein D, Jenkinson R, Kreder H J. Comparing complications and costs of total hip arthroplasty and hemiarthroplasty for femoral neck fractures: a propensity score-matched, population-based study. J BoneJoint Surg 2019; 101(7): 572-9. Rogmark C, Johnell O. Primary arthroplasty is better than internal fixation for displaced femoral neck fractures: a meta-analysis of 14 randomized studies with 2289 patients. Acta Orthop 2006; 77(3): 359-67.

Acta Orthopaedica 2020; 91 (2): 125

125

Annotation

Recycling implants: a sustainable solution for musculoskeletal research

About half a million people in Sweden (5%) (10.5 million inhabitants) walk around with a bone or joint implant. This could be extrapolated to 50 million only in the Americas, Europe, and the Pacific. Orthopedic surgeons are running the largest workshop for repairing humans worldwide, and, in fact, one-fifth of all people older than 75 years in Sweden have an artificial joint. In the United States, 5–10% of all hip or knee implants are being revised during a lifetime (Malchau et al. 2018). The environmental as well the financial aspects of recycling metals at revision or post-mortem has thus far not been prioritized on the orthopedic societies’ agendas. Close to 100,000 Swedes pass away annually, with 70% being cremated. To protect the environment, the Swedish Government changed its regulation in 2016, making it mandatory for the Swedish Church—which is responsible for all funerals—to arrange for recycling after cremations of all metal components including the coffin. Since 2016, all metal collected at cremation has been recycled (Figure), resulting in 60 tons of valuable metals (such as, for instance, titanium) at a net value of US$ 15 million. The recycling company, however, charges 20%, while 80% is plowed back to support societal projects handled by a large general inheritance fund (Hyckenberg 2019). If the Swedish figures were extrapolated to Europe and the United States alone, the total income from metal recycling at funeral would be around US$ 250 million annually (i.e., 700 million inhabitants, assuming 1% mortality per year, 50% cremation and if 20% had a metal device at death) (The Cremation Society 2019). In addition, recycling of extracted implants that are currently scrapped as dangerous goods should be possible if the logistics for collection could be handled properly. This could be done under the auspices of the national orthopedic societies. With an expected minimal cumulative revision rate of 5% leading to a revision arthroplasty, this recycling in Sweden should give a yearly recurring income close to US$ 500,000, which could be directed to orthopedic research. Given the ongoing “age quake” and an increasing cremation trend in the industrialized world, metal implant recycling—which has just started in Sweden—is likely to spread to other countries. Leading implant manufacturers are encouraged to start collaborating with the orthopedic societies, who should assume

Retrieved metal implants after cremation sent for recycling. (With permission from Jennie Lorentsson).

their environmental responsibility and facilitate recycling of metal implants both at revision and at post-mortem. This has to be done in a manner respectful to the deceased and his or her loved ones. In cooperation with patient organizations, we suggest that orthopedic societies put forward a motion to governments that at least part of the income from such recycling be directed to musculoskeletal research in the respective countries. Lars Lidgren 1, Deepak Bushan Raina 1, Magnus Tägil 1, and K Elizabeth Tanner 2 1 Lund

University, Faculty of Medicine, Department of Clinical Sciences Lund, Orthopedics, Lund, Sweden 2 School of Engineering and Materials Science, Queen Mary University of London, UK Corresponding author: lars.lidgren@med.lu.se

Malchau H, Garellick G, Berry D, Harris W H, Robertson O, Kärrholm J, Lewallen D, Bragdon C R, Lidgren L, Herberts P. Arthroplasty implant registries over the past five decades: Development, current, and future impact. J Orthop Res 2018; 36 (9): 2319-30. DOI: 10.1002/jor.24014. Hyckenberg A. Återvinning av metaller. (2019) https://www.svenskakyrkan. se/sollentuna/nyheter/atervinning-av-metaller. The Cremation Society, Cremation Statistics. (2019) https://www.cremation. org.uk/statistics.

© 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 10.1080/17453674.2019.1706301

126

Acta Orthopaedica 2020; 91 (2): 126–132

Synthetic hydroxyapatite: a recruiting platform for biologically active molecules Deepak Bushan RAINA 1, Yang LIU 1, Hanna ISAKSSON 1,2, Magnus TÄGIL 1, and Lars LIDGREN 1 1 Faculty

of Medicine, Department of Clinical Sciences, Orthopedics, Lund University, Lund; 2 Department of Biomedical Engineering, Lund University, Lund, Sweden Correspondence: deepak.raina@med.lu.se Submitted 2019-07-12. Accepted 2019-10-08.

Background and purpose — Targeted delivery of drugs is important to achieve efficient local concentrations and reduce systemic side effects. We hypothesized that locally implanted synthetic hydroxyapatite (HA) particles can act as a recruiting moiety for systemically administered drugs, leading to targeted drug accretion. Methods — Synthetic HA particles were implanted ectopically in a muscle pouch in rats, and the binding of systemically circulating drugs such as zoledronic acid (ZA), tetracycline and 18F-fluoride (18F) was studied. The local biological effect was verified in an implant integration model in rats, wherein a hollow implant was filled with synthetic HA particles and the animals were given systemic ZA, 2-weeks post-implantation. The effect of HA particle size on drug binding and the possibility of reloading HA particles were also evaluated in the muscle pouch. Results — The systemically administered biomolecules (ZA, tetracycline and 18F) all sought the HA moiety placed in the muscle pouch. Statistically significant higher periimplant bone volume and peak force were observed in the implant containing HA particles compared with the empty implant. After a single injection of ZA at 2 weeks, micro HA particles showed a tendency to accumulate more 14C-zoledronic acid (14C-ZA) than nano-HA particles in the muscle pouch. HA particles could be reloaded when ZA was given again at 4 weeks, showing increased 14C-ZA accretion by 73% in microparticles and 77% in nanoparticles. Interpretation — We describe a novel method of systemic drug loading resulting in targeted accretion in locally implanted particulate HA, thereby biologically activating the material.

In drug delivery, one important goal is to achieve efficient tissue concentrations in targets known for poor drug penetration. Local drug delivery can be one solution and may involve a carrier, able to act as a temporary depot to release the active biomolecules (Raina et al. 2016). The possibility of reloading such a carrier has until now not been described. Furthermore, a local delivery approach most often requires surgery. Targeted delivery of drugs by coupling them to tissue specific ligands, the so-called ligand–receptor interaction, is an example of a systemic approach to enhance drug concentration, yet the efficiency is less than 10% and it still involves complicated fabrication processes (Kirpotin et al. 2006, Bae and Park 2011). We propose implanting a recruiting and reloadable particulate apatite moiety, within the tissue of interest, to which systemically administered drugs circulating in the bloodstream could bind due to a high chemical affinity. A biomaterial in the form of particulate HA embedded in calcium sulphate (CaS) allows for in-situ setting. Based on the affinity to HA, there are antibiotics today in clinical use for bone infection that could be candidates for seeking HA as a recruiting moiety (Perrin 1965). By activating the ceramic material, it can initially exert a local antibacterial effect and later be reloaded via systemic administration. We hypothesized that particles of synthetic HA possess binding sites such as calcium, phosphate and hydroxyl groups, which when placed in a targeted tissue can act as recruiting moiety for systemically administered biomolecules. The primary aim of our study was to demonstrate whether a systemically administered bisphosphonate, zoledronic acid (ZA) with known affinity to HA, could be bound to synthetic particulate HA implanted in an ectopic location. Second, our aim was to prove a biological effect of the drug-seeking phenomenon in bone, by using a fenestrated implant containing HA particles in an orthotopic model in rats. Additionally, we present: (1) an evaluation of the effect of the HA particle size on the drug binding capacity in an ectopic implantation model;

Acta Orthopaedica 2020; 91 (2): 126–132

127

Figure 1. Overview of the experimental design used in the study.

(2) an assessment of the possibility of reloading the implanted HA particles via systemic delivery of a model drug and (3) an exploration of other binding agents, given systemically, such as an antibiotic, tetracycline, and a radioactive tracer 18F, and their ability to seek local HA.

Methods Study design The observations made in this study are based on in-vivo experiments carried out on the laboratory rat as a model biological system. The first study describes the uptake of a bisphosphonate, zoledronic acid (ZA), in a biphasic calcium sulphate (CaS)/hydroxyapatite (HA) based biomaterial. The material was implanted in an abdominal muscle pouch model, an ectopic non-osseous site, without the presence of living bone (Raina et al. 2016). To verify the biological effects, an implant integration model was used in rats. The next experiment was performed to evaluate the effect of the hydroxyapatite particle size on uptake of ZA in the ectopic muscle pouch model. The final experiments were performed to evaluate the affinity of 2 other drug classes to synthetic HA including the antibiotic tetracycline and positron-emitting radioactive tracer 18F (Figure 1). Uptake of ZA in ectopically implanted CaS/HA biomaterial pellets: hydroxyapatite as a recruiting moiety CaS/HA biomaterial pellets were prepared by mixing 1 g of CaS/HA pre-mixed powder (60% CaS and 40% HA) with 0.43 mL non-ionic radiographic contrast agent (Iohexol) using a sterile spatula. The slurry was transferred to a 1 mL graduated syringe and 40 µL of the paste was poured into each well of a sterile nylon mold (Ø = 5 mm). 12 pellets were casted and each pellet contained approximately 80 mg CaS/HA. The contents of each well were allowed to set for 30 min and the pellets were retrieved from the mold. The entire process was performed in a sterile environment under a laminar airflow bench. 6 male Sprague-Dawley (SD) rats (average weight 358 g) were used for the experiment. 4 rats received 1 pellet of

CaS/HA biomaterial each in the abdominal muscle following an established protocol described in detail by our group previously (Raina et al. 2016). The remaining 2 animals were operated the same way without receiving the CaS/HA biomaterial pellet (SHAM). After 2 weeks, all animals were given a subcutaneous injection of 14C labelled ZA (concentration: 1 mg/mL, specific radioactivity: 7.1 MBq/mL, radiochemical purity: > 95%) at a dose of 0.1 mg/kg, a standard dose in rodent experiments (Amanat et al. 2007). After a period of 24 h, all 6 animals were killed by CO2 asphyxiation. 4 pellets of CaS/HA biomaterial were retrieved from the biomaterial implanted animals. A 4-mm biopsy of the muscle (n = 2) was harvested from the SHAM operated animals from the same anatomical location. Samples were individually placed in 5 mL scintillation tubes and immersed in 2 mL of 5M HCl for 48 h at room temperature to aid in rapid decalcification and softening. All samples were homogenized to form a slurry using an ultrasound-based tissue homogenizer (1 min/sample). 0.5 mL of the slurry was mixed with 4.5 mL scintillation cocktail (Optiphase, HiSafe 2, PerkinElmer, Waltham, MA, USA), homogenously mixed and read using a scintillation counter (Wallac 1414, PerkinElmer, USA). Implant integration model: biological effect of biomodulating hydroxyapatite Medical grade polyether ether ketone (PEEK) implant (outer Ø = 3.5 mm, inner Ø = 2.1 mm, hole Ø = 1.4 mm and height = 6.3 mm) was custom made in the form of a threaded hollow cylindrical core with a conical frustum at the bottom and contained 3 equally spaced holes. A PEEK implant was used over conventional titanium to avoid metal artefacts seen using X-ray-based imaging of metallic implants. A detailed description of the chamber model is mentioned elsewhere (Raina et al. 2019). In brief, a 3.2 mm Ø hole was created in the right proximal tibia of rats just under the tibial epiphysis. The implant was placed press-fit in the metaphyseal bone and screwed with the aid of a custom-made screwdriver. 2 groups were used for evaluation of bone–implant anchorage; (1) Empty PEEK implant (control) and (2) PEEK implant filled with a CaS/HA biomaterial. The control group involving the

128

use of the empty implant has also been described earlier in our recent study and data are used for comparison only (Raina et al. 2019). 22 male SD rats were divided into 2 equally sized groups (average weight = 378 g, n = 11/group). After 2 weeks, the animals receiving implants containing CaS/HA biomaterial were injected with a single subcutaneous dose of ZA (concentration: 0.8 mg/mL, injected dose: 0.1 mg/mL). 6 weeks post-surgery, the animals were killed using CO2 asphyxiation and the harvested tibiae were evaluated for peri-implant bone formation using radiography, microcomputed tomography (micro-CT), mechanical testing, and histology. Radiography and Micro-CT imaging was performed using a NanoScan micro CT scanner (Mediso Medical Imagining System, Budapest, Hungary) to obtain images with an effective voxel size of 10 µm (X-ray voltage: 65 kV, current: 123 µA, exposure: 1300 ms). Peri-implant bone formation was measured immediately around the implant holes within the medullary canal and expressed as bone volume (BV) based on established protocols (Raina et al. 2019). Mechanical testing was performed on an Instron® (8511) biaxial load frame (Instron, Norwood, MA, USA) connected to a 250N load sensor. Samples were mounted on a custommade jig. All specimens were subjected to a pre-loading protocol for 10 s before a total pull-out was performed. A load rate of 0.5 mm/s was used for the pull-out and the force-displacement curves were used to obtain the peak force (Raina et al. 2019). Routine procedures for decalcified histology were followed. Briefly, tissue fixation was done in 4% neutral buffered formalin solution overnight following which EDTA (10% w/v) based decalcification was carried out for 5 weeks before paraffin embedding. Sections of 5 µm thickness were cut and stained with H&E by following the manufacturer’s guidelines (Thermo Fisher Scientific, Waltham, MA, USA). Role of hydroxyapatite particle size on drug accumulation and the possibility of reloading Commercially available HA particles of micrometer size (10 µm, Sigma-Aldrich, Product Number: 900203; SigmaAldrich, St Louis, MO, USA) and nanometer size (< 200 nm, Sigma-Aldrich, Product Number: 677418) were sterilized by autoclaving. For the uptake experiment, 25 mg of both particle sizes were weighed. 18 male SD rats (average weight: 326 g) were operated in the abdominal muscle pouch and each rat received loose nano-HA (nHA) particles on the right side of the abdominal midline and loose micro-HA (mHA) particles on the left side of the abdominal midline. Particles were introduced into the muscle pouch by using a sterile pipette tip as a funnel. The muscle was sutured using 2 single non-resorbable sutures, which also aided in locating the particles at harvest. At 13 days post-surgery, all animals received a single subcutaneous injection of 14C labelled ZA (concentration: 1 mg/mL, specific radioactivity: 7.1 MBq/mL, radiochemical purity: > 95%) at a dose of 0.1 mg/kg (animal weight aver-

Acta Orthopaedica 2020; 91 (2): 126–132

aged to 400 g). 1 day after the injection, 6 animals were killed and the radioactive counts were measured by following the protocol described in section 2.2. At 27 days post-surgery, 6/12 animals received a repeated subcutaneous injection of 14C labelled ZA (dose: 0.1 mg/kg, animal weight averaged to 400 g). 1 day after the second injection, all remaining 12 animals were killed to evaluate the efficacy of reloading the HA particles with ZA. The 6 animals that were injected on day 13 and killed at t = day 28 were used as controls for the reloading experiment. Evaluating the affinity of other biologically active molecules to hydroxyapatite The antibiotic tetracycline and 18F-fluoride radioactive isotope were the other 2 biologically active molecules whose ability to systemically seek and bind hydroxyapatite was evaluated. Pellets of CaS/HA biomaterial were prepared by following the same procedure as described earlier. 4 male SD rats (average weight 361 g) were used and each animal received 1 single pellet of the CaS/HA biomaterial in the abdominal muscle. After 2 weeks, 2 animals received a single subcutaneous injection of tetracycline (dose 20 mg/kg) and the animals were killed 1 day later. The pellets were harvested from the muscle pouch, carefully cleaned of all surrounding muscle/ connective tissue and formalin fixed overnight followed by routine histological preparation. Tissue sections were placed on glass slides for fluorescence microscopy. The remaining 2 animals were injected with Na-18F via the tail vein 2 weeks after CaS/HA biomaterial implantation (specific radioactivity: animal 1: 85 MBq and animal 2: 120 MBq) while maintaining isoflurane anesthesia (2% isoflurane and 1:1 mixture of O2 and N2O, flow rate: 0.4 L/min). After a waiting period of 1 h, the animals were placed in a positron emission tomography scanner (NanoPET/CT, Mediso Medical Imagining Systems, Hungary) and subjected to micro-CT to detect the anatomical location of the implanted pellet (projections: 240, scan: semi-circular, X-ray voltage: 65 kV, exposure: 500 ms, voxel size: 141 µm). Imaging setup described earlier by Mathavan et al. (2019) was used. The uptake of Na-18F in the CaS/HA biomaterial was analyzed using PET imaging and a voxel size of 0.4 mm was achieved. The CT and PET projections were overlapped to confirm the tracer uptake in the CaS/HA biomaterial. Statistics Mann–Whitney U-test was used to compare two groups. Paired data analysis was performed using Wilcoxon’s signed rank test. Data in the graphs are shown as mean ± SD. Ethics, funding, and potential conflicts of interest All animal experiments were approved by the Swedish board of agriculture (Permits: M124-14 and M79-15). The ARRIVE guidelines have been followed to provide details regarding in-vivo experiments involving the usage of laboratory ani-

Acta Orthopaedica 2020; 91 (2): 126–132

Radiograph

129

Micro-CT slice

Histology (H&E) overview

Histology magnified

Implant + empty

Implant + CaS/HA + systemic ZA

C-ZA (DPM)

20,000

BV (mm3) – micro-CT

Peak force (N) – mechanical testing

300 a

b 250

15,000

3 200

10,000

5,000

150 100 50

30 0

Muscle

CaS/HA

Figure 2. Uptake of 14C-ZA in the CaS/HA biomaterial placed in the abdominal muscle, 1 day post subcutaneous administration of 14C-ZA. Observe the difference in the scale of the y-axes for the 2 groups. Bars indicate uptake of 14C-ZA in the muscle (brown) and CaS/HA biomaterial (blue).

Implant + empty

Implant + CaS/HA + systemic ZA

Implant + empty

Implant + CaS/HA + systemic ZA

Figure 3. Biological effect of hydroxyapatite biomodulation assessed in the implant integration model in rats. Top panel shows representative radiograph, micro-CT slice, histology overview (H&E staining), and high magnification histology (H&E staining), respectively from the Implant + empty group while the middle panel shows images from the Implant + CaS/ HA + systemic ZA group. Bottom left indicates micro-CT quantification of bone volume (BV) immediately around the implant holes and bottom right shows peak pull-out force. In the histology images * indicates new bone formation around the implant. Scale bar in the overview histology images represents 1 mm and magnified images represents 100 µm. a indicates p < 0.001 and b indicates p < 0.01 (Mann–Whitney U-test). Data for the empty group are taken from an earlier study for comparison (Raina et al. 2019).

mals. VINNOVA, the Swedish agency for innovation systems (Grant number: 2017-00269), the Swedish Research Council (Vetenskapsrådet, Grant number: 2015-06717) and the Alfred Andersson foundation funded this study. LL is a board member of Bone Support AB, Sweden and Ortho Cell, Australia. MT and DBR have received options from Ortho Cell, Australia for work unrelated to this study.

Results Uptake of 14C-ZA in the CaS/HA biomaterial at an ectopic location 14C-ZA uptake was confirmed in the pellet of CaS/HA biomaterial placed in the abdominal muscle pouch using scintillation counting (Figure 2). Negligible counts were detected in the abdominal muscle (muscle: 18 DPM vs. only cocktail: 8 DPM) of the control animals that were injected with only 14C-ZA.

Implant integration model: biological effect of biomodulating hydroxyapatite Representative radiographs, micro-CT slices and the quantitative analysis of BV measured using micro-CT indicated significantly higher BV in the Implant + CaS/HA + systemic ZA group compared with the empty implant (p < 0.001) (Figure 3). Representative histology images corroborated the micro-CT results well and a higher amount of viable bone tissue around the implant was found in the systemic ZA group, compared with the empty control both at low and high magnifications. Furthermore, the higher BV fraction around the implant in the systemic ZA group also resulted in better implant osseointegration (increased pull-out peak force) as seen from the increased peak force in the pull-out testing (p = 0.008). Effect of hydroxyapatite particle size on 14C-ZA uptake A higher 14C-ZA uptake was observed in micro-sized HA

130

Acta Orthopaedica 2020; 91 (2): 126–132

C-ZA (DPM)

1,500

1,250

One injection (n = 6) Injection at Day 13 Killed at Day 14 p = 0.06

1,250

One injection (n = 6) Injection at Day 13 Killed at Day 28 p = 0.09

1,250

1,000

750

500

250

Nano-HA

Micro-HA

Nano-HA

Micro-HA

Two injections (n = 6) Injection 1 at Day 13 Injection 2 at Day 27 Killed at Day 28 p = 0.06

Nano-HA

Micro-HA

Figure 4. Timeline and uptake kinetics of 14C-ZA in micro- and nano-sized HA particles placed in the abdominal muscle pouch. Left: Uptake of 14C-ZA in the animals injected with 14C-ZA on day 13 and killed on day 14. Middle: Uptake of 14C-ZA in the animals injected with 14C-ZA on day 13 and killed on day 28. Right: Uptake of 14C-ZA in the animals injected with 14C-ZA on day 13 and day 27 and killed on day 28. Each line on the graph presents paired data from the same animal. Statistical analysis was performed using Wilcoxon’s matched pairs signed rank test.

C-ZA (DPM) – Nano-HA

1,000 800

800

600

400

200

Day 14

Day 28

Bright field

C-ZA (DPM) – Micro-HA

1,000

Fluorescence

2.5x

Day 14

Day 28

14C-ZA.

Figure 5. Reloading efficiency of HA particles with Nano- (left) and micro-sized (right) HA particles injected with 14C-ZA and its uptake 14 a after 1 or 2 subcutaneous injections of C-ZA. indicates p < 0.05 (Mann–Whitney U-test).

particles as compared with the nano-HA particles (p = 0.06) (Figure 4, left). Possibility of reloading hydroxyapatite particles There was a marked difference between the uptake of 14C-ZA after 1 and 2 subcutaneous injections of 14C-ZA, indicating the possibility of reloading HA particles on a need-be basis (Figures 4 and 5). Similar to the animals killed at day 14 (Figure 4, left), the animals that received 1 subcutaneous dose of 14C-ZA on day 13 followed by killing on day 28 (Figure 4, middle), showed an increased 14C-ZA uptake in microparticles compared with nanoparticles (p = 0.09) with a further increase after 2 subcutaneous injections (Figure 4, right). By administering 14C-ZA on 2 occasions instead of 1, a 77% increase in the average uptake of 14C-ZA was noted in the nanoparticles (p = 0.02) while micro-HA particles exhibited an increase of 73% (p = 0.02) (Figure 5).

10x

Figure 6. Uptake of tetracycline (administered systemically) in a pellet of CaS/HA biomaterial placed in the abdominal muscle pouch of rats, 24 h post-administration detected using fluorescence microscopy. Yellow box (top, right) indicates the approximate region where the high magnification (bottom, right) fluorescence image was captured. Low magnification image was captured with an exposure of 1 s while the high magnification image was captured with an exposure of 200 ms.

Uptake of antibiotic tetracycline and radioemitter 18F in CaS/HA biomaterial Fluorescence microscopy images confirmed the uptake of the antibiotic tetracycline in the CaS/HA biomaterial after subcutaneous administration of tetracycline 2 weeks post-pellet implantation and subsequent imaging 24 h later (Figure 6). Importantly, most of the fluorescent signal was seen on the periphery of the CaS/HA biomaterial with very limited fluorescence detected in the middle of the specimen.

Acta Orthopaedica 2020; 91 (2): 126â&#x20AC;&#x201C;132

2D PET/CT overlap

131

3D reconstruction

Figure 7. Uptake of 18F (administered systemically) in a pellet of CaS/ HA biomaterial placed in the abdominal muscle pouch of rats, 1 h postadministration detected using PET-CT.

Using PET-CT, the uptake of the radioemitter 18F in the CaS/ HA biomaterial ectopically placed in the abdominal muscle pouch was shown. Apart from the CaS/HA biomaterial, the uptake was also observed in other hard tissues. A substantial amount of the tracer was cleared by the kidneys and found in the urinary bladder (Figure 7).

Discussion We have shown that synthetic HA particles can be used as a recruiting moiety for different drug classes administered systemically and their affinity to HA binding sites can activate the particulate material to exert a biological effect. After systemic administration of radioactive ZA, we found an accretion in the synthetic HA particles implanted in the abdominal muscle pouch, which remained for at least 2 weeks after injection. Similar results have been observed in an orthotopic model in osteoporotic rats for up to 6 months (unpublished results). The ZA not only seeks but also modulated and activated HA, thereby enhancing bone anchorage as shown in the implant integration model. The timing of the injection may be critical, and has been established to be between 1 and 2 weeks in rat models of fracture healing (Amanat et al. 2007). This time interval theoretically fits well within the time frame reported for HA deposition in a fracture callus collagen network, and with the fluid mechanics necessary for drug transport (Andreshak et al. 1997, Amanat et al. 2007, Mathavan et al. 2019). The use of the biomodulation framework is promising also for other drug classes, apart from obvious bone-seeking drugs like bisphosphonates. Using particulate HA as a recruitment platform may also open up new treatment methods in infections and tumors. Tetracycline is commonly used for dynamic histomorphometry of bone and Perrin in 1965 confirmed that tetracycline binds to HA (Perrin 1965). Furthermore, Bernhardsson et al. (2018) recently showed uptake of a positronemitting tracer 18F in an ectopic location in rats, which contained pellets of allogenic bone or granular HA. While the

literature reveals that certain drug classes have affinity for bone or HA in general, none of the studies have looked at the relationship of physical properties of HA, such as particle size and its effect on drug-binding capacity, especially in in-vivo models. More so, our study also presents a novel concept of reloading the implanted HA particles, which opens possibilities for on-demand additional local drug targeting. Owing to a larger surface area, we hypothesized that nanoparticles in comparison with microparticles could theoretically give access to an increased number of binding sites and thus higher drug accretion. However, our findings indicate the contrary, which could be because the intra-particulate open space is larger between microparticles compared with the densely packed nanoparticles, making the binding sites more easily accessible. Another explanation could be that the biological clearance of the nanoparticles from the tissue locally is more rapid, with fewer total available particles and thus reduced binding affinity. This observation agrees with an earlier study, which compared the effect of polymer based micro- and nanoparticle delivery in the rodent peritoneum (Kohane et al. 2006). The authors could detect microparticles in the peritoneum for up to 14 days while the nanoparticles were cleared as early as 2 days with substantial uptake in the spleen and the liver. Local HA, in the form of micro- to nanoparticles at the site of infection, could act as a carrier but also as a recruiting moiety for systemically administered HA binding antibiotics. The affinity of each antibiotic would foremost depend on its chemical structure, i.e., binding capacity to calcium, phosphate, and hydroxyl groups and HA amount and size. It is compelling that some of the recommended second-level systemic antibiotics for PJI such as rifampicin, tetracycline, and daptomycin all have a chemical structure that would allow them to bind to apatite. A novel approach for a biphasic ceramic could be to have the antibiotic not only embedded in the soluble sulphate phase but also, by using millions of nano-HA particles already functionalized with antibiotics, to provide extended sustained local antibiotic delivery. The HA particles could then further be reloaded by systemic administration with the same or a different antibiotic that has high affinity for HA. Nanoparticles of HA are shown to be internalized by several cell types (Yuan et al. 2010, Zhao et al. 2018), which could give the possibility of intracellular drug delivery to eradicate bacteria known to reside within the cells. Apart from antibiotics, Lewington (2005) in a review on bone-seeking radioactive isotopes describes a group of isotopes such as 32P, 89Sr, 186Re, and 223Ra possessing high affinity to metabolically active bone. De Klerk et al. (1992) combined 186Re with hydroxyethylidene diphosphonate to form a bisphosphonate complex with an aim to achieve high bone accretion for treatment of metastatic bone disease. Even bisphosphonates have been described as carriers for various tumor-targeting radioisotopes for diagnostic purposes as well as for pain management in tumors (Palma et al. 2011). In this study, we used a model radioemitter 18F to verify accretion in

132

synthetic HA particles. Based on the literature, other clinically relevant isotopes could also be systemically administered and concentrated in a target tissue implanted with large number of nano- and micro-HA particles for theranostic treatment of locally aggressive tumors or solitary metastasis. Agents that exhibit strong affinity to hydroxyapatite can also be coupled with other therapeutic agents (Ramanlal Chaudhari et al. 2012) wherein the molecule with affinity acts as a guide and takes the drug to the target tissue. Conclusion This study recapitulates some of the early studies with systemically administered agents traced in bone and hydroxyapatite (HA). A systemically administered bisphosphonate, ZA, seeks HA acting as a recruiting moiety. A fenestrated bone-anchored PEEK implant was filled with synthetic HA microparticles and after a period of 2 weeks a bisphosphonate, zoledronic acid, was systemically administered. The synthetic HA particles acted as a ZA-recruiting biomodulated moiety and resulted in improved bone–implant anchorage, indicating a biological effect. The specific binding of locally implanted HA particles in the muscle was also verified for tetracycline and 18F. Our study also shows that the size of the HA particles plays an important role in the binding of a systemically administered drug. Micro-sized particles tend to bind more drug compared with nanoparticles. Systemic loading of synthetic HA particles can be carried out at several predetermined time intervals or on a need-be basis as long as a sufficient amount of HA particles with free binding sites are available at the target tissue. Clinical significance We present a novel concept of targeted drug delivery by providing a HA-based recruiting moiety that can attract certain classes of systemically circulating drugs and lead to their accumulation within the target tissue. A library of biomolecules that have the ability to chemically interact with HA could in future be used for targeted delivery of drugs in scenarios involving bone regeneration, infections, or tumors, and likely be implemented in other tissues.

LL, MT, DBR conceived the idea and designed the studies. MT, LL, DBR, HI, and YL carried out the experiments and data analysis. LL and DBR wrote the manuscript. All authors revised the manuscript for submission. The authors would like to thank the Medical Faculty, Lund University and Lund University Bioimaging Center (LBIC) for providing infrastructure support. Acta thanks Ming Ding for help with peer review of this study.

Acta Orthopaedica 2020; 91 (2): 126–132

Amanat N, McDonald M, Godfrey C, Bilston L, Little D. Optimal timing of a single dose of zoledronic acid to increase strength in rat fracture repair. J Bone Miner Res 2007; 22(6): 867-76. Andreshak J L, Rabin S I, Patwardhan A G, Wezeman F H. Tibial segmental defect repair: chondrogenesis and biomechanical strength modulated by basic fibroblast growth factor. Anat Rec 1997; 248(2): 198-204. Bae Y H, Park K. Targeted drug delivery to tumors: myths, reality and possibility. J Control Rel 2011; 153(3): 198-205. Bernhardsson M, Sandberg O, Ressner M, Koziorowski J, Malmquist J, Aspenberg P. Shining dead bone: cause for cautious interpretation of [(18) F]NaF PET scans. Acta Orthop 2018; 89(1): 124-7. de Klerk J M, van Dijk A, van het Schip A D, Zonnenberg B A, van Rijk P P. Pharmacokinetics of rhenium-186 after administration of rhenium186-HEDP to patients with bone metastases. J Nucl Med 1992; 33(5): 646-51. Kirpotin D B, Drummond D C, Shao Y, Shalaby M R, Hong K, Nielsen U B, Marks J D, Benz C C, Park J W. Antibody targeting of long-circulating lipidic nanoparticles does not increase tumor localization but does increase internalization in animal models. Cancer Res 2006; 66(13): 6732-40. Kohane D S, Tse J Y, Yeo Y, Padera R, Shubina M, Langer R. Biodegradable polymeric microspheres and nanospheres for drug delivery in the peritoneum. J Biomed Mater Res Part A 2006; 77(2): 351-61. Lewington V J. Bone-seeking radionuclides for therapy. J Nucl Med 2005; 46 (Suppl. 1): 8s-47s. Mathavan N, Koopman J, Raina D B, Turkiewicz A, Tägil M, Isaksson H. 18F-fluoride as a prognostic indicator of bone regeneration. Acta Biomater 2019; 90: 403-11. Palma E, Correia J D G, Campello M P C, Santos I. Bisphosphonates as radionuclide carriers for imaging or systemic therapy. Mol BioSystems 2011; 7(11): 2950-66. Perrin D D. Binding of tetracyclines to bone. Nature 1965; 208(5012): 787-8. Raina D B, Isaksson H, Hettwer W, Kumar A, Lidgren L, Tägil M. A biphasic calcium sulphate/hydroxyapatite carrier containing bone morphogenic protein-2 and zoledronic acid generates bone. Sci Rep 2016; 6: 26033. Raina D B, Larsson D, Sezgin E A, Isaksson H, Tägil M, Lidgren L. Biomodulation of an implant for enhanced bone–implant anchorage. Acta Biomater 2019; 96:619-30. Ramanlal Chaudhari K, Kumar A, Megraj Khandelwal V K, Ukawala M, Manjappa A S, Mishra A K, Monkkonen J, Ramachandra Murthy R S. Bone metastasis targeting: a novel approach to reach bone using zoledronate anchored PLGA nanoparticle as carrier system loaded with docetaxel. J Control Rel 2012; 158(3): 470-8. Yuan Y, Liu C, Qian J, Wang J, Zhang Y. Size-mediated cytotoxicity and apoptosis of hydroxyapatite nanoparticles in human hepatoma HepG2 cells. Biomaterials 2010; 31(4): 730-40. Zhao H, Wu C, Gao D, Chen S, Zhu Y, Sun J, Luo H, Yu K, Fan H, Zhang X. Antitumor effect by hydroxyapatite nanospheres: activation of mitochondria-dependent apoptosis and negative regulation of phosphatidylinositol3-kinase/protein kinase B pathway. ACS Nano 2018; 12(8): 7838-54.

Acta Orthopaedica 2020; 91 (2): 133–138

133

More hip complications after total hip arthroplasty than after hemi arthroplasty as hip fracture treatment: analysis of 5,815 matched pairs in the Swedish Hip Arthroplasty Register Susanne HANSSON 1, Erik BÜLOW 2,5, Anne GARLAND 3,4, Johan KÄRRHOLM 2,5, and Cecilia ROGMARK 1,2 1 Department

of Orthopaedics, Lund University, Skåne University Hospital, Malmö; 2 The Swedish Hip Arthroplasty Register, Registercentrum Västra Götaland, Gothenburg; 3 Department of Orthopaedics, Visby Hospital, Visby; 4 Department of Orthopaedics, Institute of Surgical Sciences, Uppsala University, Uppsala; 5 Department of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Sweden Correspondence: cecilia.rogmark@skane.se Submitted 2018-11-22. Accepted 2019-10-07.

Background and purpose — Total hip arthroplasty (THA) is increasing as treatment of displaced femoral neck fractures. Several studies compare hemiarthroplasty (HA) with THA, but results vary and few studies report on medical complications. We examined the outcome of THA and HA with a focus on medical complications, hip complications, and death. Patients and methods — Data from the Swedish Hip Arthroplasty Register on 30,953 acute hip fracture patients treated with cemented THA or HA in 2005–2011 were crossmatched with Statistics Sweden for socioeconomic data and with the National Patient Register for diagnostic codes representing medical complications within 180 days or hip complications within the study period. Propensity score matching was used to create comparable groups based on age, sex, income, level of education, marital status, Elixhauser index, and year of surgery. Logistic regression models were created for each outcome. Results — 81% were treated with HA, 73% and 71% were female (HA and THA respectively). Matching resulted in 2 groups of 5,815 patients each. THA was associated with fewer medical complications (OR = 0.83; 95% CI 0.76–0.91) and lower 1-year mortality (OR = 0.42; CI 0.38–0.48), but more hip complications (OR = 1.31; CI 1.20–1.43). Interpretation — THA as treatment of hip fracture was associated with more hip-related complications than HA. The results on mortality and medical complications are, rather, influenced by residual confounding than by the implant design per se. An expansive use of THAs for hip fracture treatment, at the expense of HAs, is not recommended based on our findings if hip complications are to be avoided.

Hemiarthroplasty (HA) or internal fixation have been the main alternatives for treatment of displaced femoral neck fracture. Total hip arthroplasty (THA) has increased in popularity as fracture treatment (Kärrholm et al. 2018). Several studies have compared HA with THA, but the results vary. Age, activity level, health, and supposed remaining lifespan of the patient are factors influencing the choice between THA and HA in clinical practice. HAs may have a lower risk of dislocation (Burgers et al. 2012, Rogmark and Leonardsson 2016). However, since the head articulates directly against the cartilage, patients receiving HA may develope acetabular erosion (Avery et al. 2011, Wang et al. 2015). THA often entails longer surgeries (Blomfeldt et al. 2007, van den Bekerom et al. 2010). In contrast, some studies have shown THA to be associated with lower mortality (Avery et al. 2011, Hansson et al. 2017, Wang and Bhattacharyya 2017). An adverse event is defined as an unintended injury or complication resulting in temporary or permanent disability, death, or prolonged hospital stay, and is caused by the healthcare management rather than by the natural disease process (Brennan et al. 1991, Merten et al. 2015). Adverse events implicate both medical and hip complications as well as death. Studies comparing THA and HA traditionally have focused on hip function and hip complications, but fewer report on medical complications. Earlier we found function after hip fracture to be affected not only by hip complications, but also by medical complications (Hansson et al. 2015), stressing the need to include both in the comparisons between implants. We examined the difference in outcome between THA and HA with a focus on adverse events to provide support for the decision on which type of arthroplasty to use as treatment for femoral neck fractures. The outcomes studied were medical complications, hip complications, and death.

134

Acta Orthopaedica 2020; 91 (2): 133–138

Primary HA or THA for femoral neck fracture performed in 1999–2012 n = 56,259

Patients and methods We performed an observational cohort study by cross-matching data from 3 national Swedish registers: the Swedish Hip Arthroplasty Register (SHAR) (Kärrholm et al. 2018), the Swedish National Patient Register (NPR) (Ludvigsson et al. 2011), and Statistics Sweden (Statistics Sweden 2018). Each included individual was identified in all the registers through the unique personal identity number given to all Swedish residents. SHAR aims to register all hip arthroplasties in Sweden, covering all hospitals, both public and private, with a completeness of approximately 97% for emergency procedures. All types of reoperations are recorded continuously, revisions as well as any other open procedures. Closed reductions of dislocations are not recorded in the register. In 2012, the completeness of revision surgery reported to SHAR was 94% (Kärrholm et al. 2018). Hemiarthroplasties have been recorded in the register since 2005. Statistics Sweden is responsible for producing official statistics for Sweden, for example on income, education, and marital status of Swedish residents. NPR covers all hospital-based health care in Sweden from both private and public caregivers, including inpatient care, outpatient visits, and psychiatric care. Main diagnosis, secondary diagnosis, and external cause of injury are recorded as ICD-10 codes. The completeness of main diagnosis is almost 99% (Ludvigsson et al. 2011). Procedures are recorded as NOMESCO codes (NOMESCO 2011). An Elixhauser comorbidity index (Elixhauser et al. 1998) was generated from the ICD-10 codes in NPR. A dataset was created with information from all 3 registers including patients with acute hip fracture treated with THA or HA in 2005–2012. To allow for at least 1 year of followup, patients with operations in 2012 were excluded. To avoid including the same patient more than once, only the first surgery was included for patients having 2 hip fractures treated with arthroplasty within the study period. Uncemented arthroplasties were uncommon (6%) and were excluded. Due to small numbers and difficulty in matching, patients aged less than 60 years or more than 95 years were excluded (Figure 1). The presence of a large selection of ICD-10 codes and NOMESCO codes representing medical complications within 180 days after the hip fracture surgery or hip complications within the study period were noted and interpreted as the patient suffering a complication related to the hip fracture surgery (Appendix 1, see Supplementary data). In the NPR there is no information concerning laterality, which means that hiprelated complications on either side were included. An adverse event was defined as the event of any medical and/or hip complication and/or death within 180 days post-surgery. Death within 1 year was noted as a separate outcome. From Statistics Sweden, information on income, education, and marital status was gathered as these may be potential confounders. Pre-fracture characteristics were age, sex, income, education, marital status, and Elixhauser index. Because of its

Excluded (n = 25,306): – not operation in 2005–2011, 14,854 – not acute hip fracture, 4,653 – second hip fracture arthroplasty, 2,051 – double entries of same patient, 718 – uncemented stem, 1,932 – not aged 60–95 years, 979 – Elixhauser missing, 95 – marital status missing, 24 Included in analysis n = 30,953

Figure 1. Flowchart of included and excluded patients.

skewed distribution, income was transformed into a binary logarithm. The variable “age deviation” (age minus mean age) was created, allowing a more natural interpretation of the models. Elixhauser index was stratified into 4 groups for simplification (0, 1, 2, and 3+). Statistics Patients treated with HA are generally older and have more comorbidities (Kärrholm et al. 2018), which we also saw in our material (Table 1). To be able to compare the 2 groups of patients, propensity score matching was used based on all the covariates (age, sex, income, education, marital status, Elixhauser index, and year of surgery). Exact matching would have been preferred if possible but was not an option due to the curse of dimensionality. Propensity score matching was our second choice. It allowed us to improve covariate balance between the samples. There was an imbalanced sample before matching (Table 1), but propensity score matching reduced this imbalance (Table 2). Multivariable logistic regression models were created to compare THA with HA in terms of medical complications, hip complications, and 1-year mortality, with separate models for each outcome. The models included type of arthroplasty, the pre-fracture characteristics mentioned above, and year of surgery. The results are presented as odds ratios (OR), with 95% confidence intervals (CI). Ethics, funding, and potential conflicts of interest The study was approved by the Regional Ethical Review Board in Gothenburg, Sweden (ref. 271-14). This work was supported by grants from the Southern Health Care Region, Sweden. No competing interests declared.

Results 30,953 patients were included in the study. A majority were treated with HA (81%) and most of the patients were female, in both the THA and the HA group (73% and 71% respectively).

Acta Orthopaedica 2020; 91 (2): 133–138

135

Table 1. Patient demographics before matching. Values are fre quency (%) unless otherwise specified Factor

THA p-value

Sample size a 25,138 (81) 5,815 (19) Age, mean (SD) 84 (6.3) 75 (7.2) Age groups 60–69 644 (3) 1,342 (23) 70–74 1,470 (6) 1,330 (23) 75–79 3,515 (14) 1,464 (25) 80–84 7,011 (28) 1,055 (18) 85–89 12,498 (50) 624 (11) Women 17,829 (71) 4,215 (73) Men 7,309 (29) 1,601 (28) Elixhauser index, mean (SD) 1.4 (1.4) 1.1 (1.3) Income b, mean (SD) 16.88 (0.65) 16.94 (0.71) Education Primary school 15,327 (61) 2,998 (52) High school 6,795 (27) 1,908 (33) University 3,016 (12) 909 (16) Marital status, Married 7,468 (30) 2,553 (44) Unmarried 2,187 (9) 612 (11) Divorced 2,885 (12) 938 (16) Widow/widower 12,598 (50) 1,712 (30) Year of surgery) 2005 3,033 (12) 721 (12) 2006 3,384 (14) 714 (12) 2007 3,534 (14) 817 (14) 2008 3,765 (15) 862 (15) 2009 3,824 (15) 853 (15) 2010 3,775 (15) 893 (15) 2011 3,823 (15) 955 (16) a % of b log 2

< 0.001

0.02 < 0.001 < 0.001 < 0.001

< 0.001

0.08

total (n = 30,953)

On average, HA patients were 9 years older than THA patients. A higher percentage of HA patients were widowed whereas a higher proportion of THA patients were married. The THA and HA patients differed significantly (p < 0.001) on all variables except for sex and year of surgery (Table 1). After propensity score matching, 2 comparable groups of HA and THA patients were generated with 5,815 patients in each group (Table 2). One-third of the hemiarthroplasties were of unipolar design, one-third were of bipolar design, and one-third were of unknown design. The surgical approach for a majority of the patients with both HA and THA was direct lateral (Hardinge or Gammer). The mean time to death, revision, or loss of follow-up was 2.5 and 3.5 years for HA and THA respectively (Table 3). The most common medical complications were cardiovascular, pneumonia, and urinary tract infection. The most common specified hip complications were fracture surgery on femur, dislocation, and infection (Table 4). When we compared the matched populations of patients with THA with those treated with HA and adjusted for potential confounders, we found THA to be associated with fewer medical complications (OR = 0.83; CI 0.76–0.91) (Table 5, see Supplementary data) and more hip complications (OR = 1.31; CI 1.20–1.43) (Table 6, see Supplementary data). THA was also associated with a

Table 2. Patient demographics after propensity score matching. Values are frequency (%) unless otherwise specified Factor

HA THA p-value

Sample size a 5,815 (50) 5,815 (50) Age, mean (SD) 77 (6.2) 75 (7.2) < 0.001 Age groups 60–69 644 (11) 1,342 (23) 70–74 1,390 (24) 1,330 (23) 75–79 2,011 (35) 1,464 (25) 80–84 1,102 (19) 1,055 (18) 85–89 668 (11) 624 (11) Women 4,157 (71) 4,215 (72) 0.2 Men 1,658 (29) 1,600 (28) Elixhauser index 0.04 0 2,206 (38) 2,355 (41) 1 1,711 (29) 1,664 (29) 2 1,083 (19) 1,023 (18) 3+ 815 (14) 773 (13) Income b, mean (SD) 16.89 (0.70) 16.94 (0.71) 0.002 Education 0.009 Primary school 3,126 (54) 2,998 (52) High school 1,886 (32) 1,908 (33) University 803 (14) 909 (16) Marital status 0.02 Married 2,476 (43) 2,553 (44) Unmarried 570 (10) 612 (11) Divorced 908 (16) 938 (16) Widow/widower 1,861 (32) 1,712 (29) Year of surgery 0.2 2005 712 (12) 721 (12) 2006 818 (14) 714 (12) 2007 813 (14) 817 (14) 2008 850 (15) 862 (15) 2009 828 (14) 853 (15) 2010 874 (15) 893 (15) 2011 920 (16) 955 (16) Age deviation, mean (SD) 0.77 (6.2) –0.77 (7.2) < 0.001 a b

% of total (n = 11,630) log2

Table 3. Surgical data on THA and HA. Values are frequency (%) unless otherwise specified Factor

HA THA

Design of HA Unipolar 7,827 (31) Bipolar 7,507 (30) Unknown 9,804 (39) Surgical approach Hardinge 1,565 (6.2) 311 (5.3) Moore 7,333 (29) 1,497 (26) Gammer 7,000 (28) 1,890 (33) Other/missing 9,240 (37) 2,117 (36) Follow-up a, mean (SD) 2.5 (2.0) 3.5 (2.1) a

Time to death, revision or loss of follow-up

lower 1-year mortality than HA (OR = 0.42; CI 0.38–0.48) (Table 7, see Supplementary data). Income and education did not have a significant effect on outcome in any of the models,

136

Acta Orthopaedica 2020; 91 (2): 133–138

Table 4. Frequency of medical and hip complications, n (%) Medical complications Cardiovascular Pneumonia Urinary tract infection Cerebrovascular Thromboembolic Urinary retention Renal failure Stomach ulcer Pressure ulcer Hip complications Fracture surgery femur Dislocation Infection Any reoperation Prosthesis or implant extraction Wound healing problems Girdlestone, arthrodesis Other hip complication Other surgical complication

THA HA All 478 (8.2) 3,536 (14) 4,014 (13) 124 (2.1) 1,065 (4.2) 1,189 (3.8) 134 (2.3) 974 (3.9) 1,108 (3.6) 79 (1.4) 578 (2.3) 657 (2.1) 131 (2.3) 372 (1.5) 503 (1.6) 60 (1.0) 351 (1.4) 411 (1.3) 31 (0.5) 205 (0.8) 236 (0.8) 41 (0.7) 183 (0.7) 224 (0.7) 22 (0.4) 192 (0.8) 214 (0.7) 311 (5.3) 1,283 (5.1) 1,594 (5.1) 316 (5.4) 776 (3.1) 1,092 (3.5) 232 (4.0) 859 (3.4) 1,091 (3.5) 153 (2.6) 572 (2.3) 725 (2.3) 189 (3.3) 116 (2.0) 8 (0.1) 372 (6.4) 17 (0.3)

494 (2.0) 683 (2.2) 402 (1.6) 518 (1.7) 56 (0.2) 64 (0.2) 901 (3.6) 1 273 (4.1) 61 (0.2) 78 (0.3)

whereas status as widowed, divorced, or unmarried tended to be associated with a worse outcome, with some variations depending on the outcome studied.

Discussion We found THA to be associated with more hip complications than HA in hip fracture patients. This undermines the basis for the last decade’s increase in the use of THA, namely that THA results in fewer reoperations (Hopley et al. 2010, Hansson et al. 2017, Xu et al. 2018). How could the controversy between both higher hip complication rate and lower reoperation rate for THA be explained? Whether to perform secondary surgery or not is often left to the discretion of the orthopedic surgeon. In the case of dislocations, pain, and acetabular erosion in HA cases, this implant can relatively easily be converted to a THA by adding an acetabular cup. The threshold to perform revision surgery with exchange of existing implant parts might be higher, due to concerns about bone quality and patient frailty. This, we speculate, might explain the controversy. The fact that 1 register study (Jameson et al. 2013) and several randomized controlled trials (RCTs) (Baker et al. 2006, Keating et al. 2006, van den Bekerom et al. 2010, Hedbeck et al. 2011) have not found any difference in revision/reoperation rate between THA and HA may also question the superiority of THA. 1 recent register study did actually find a higher revision rate after THA compared with HA, but the THA group consisted of more young patients, more uncemented stems, and posterior approaches—all risk factors for revision (Moerman et al. 2018).

Since THA is associated with longer surgery and more blood loss (Blomfeldt et al. 2007, van den Bekerom et al. 2010), the procedure could be presumed to be more strenuous on the patients and in turn lead to more medical complications. We found the opposite: THA was associated with fewer medical complications. Neither a large register-based study (Liodakis et al. 2016) nor smaller RCTs (Baker et al. 2006, van den Bekerom et al. 2010) comparing THA and HA have previously found a significant difference in rates of medical complications. We also found THA to be associated with lower mortality. This was also found in a register-based study of 70,000 patients (Wang and Bhattacharyya 2017). However, a recent meta-analysis of RCTs (Xu et al. 2018), found no difference in mortality comparing THA with HA. The conflicting results probably reflect that observational studies suffer from selection bias rather than that the chosen type of arthroplasty affects mortality. On the other hand, the RCTs are often underpowered to detect subtle differences in mortality. Large studies enable us to find statistically significant results due to the large sample size and subsequently narrow confidence intervals. Nevertheless, the clinical significance should guide our treatment decisions. Latent diseases, non-recorded abuse and depression, and in particular the lack of data on pre-fracture function and frailty, imply residual confounding. Consequently, our findings on mortality and medical complications have to be interpreted with caution. As mentioned, THA was associated with more hip complications than HA. The meta-analysis by Xu et al. (2018) reported only on dislocation and infection. In that study, THA had a statistically significantly higher risk of dislocation but there was no difference in terms of infection. Previous RCTs have not been able to show a difference in hip complications between THA and HA (Baker et al. 2006, Hedbeck et al. 2011), probably due to smaller sample sizes. By including a large number of patients and a wide variety of diagnostic and procedural codes representing hip complications in the crosslinking with NPR, we reduced the risk of neglecting postoperative complications not reported to SHAR, for example closed reduction of dislocation. A caveat is that our lack of information on laterality means that a potential hip-related complication on the opposite side could be included, which, compared with previous studies, implies a slight overestimation of hip-related complications. These events could, however, be expected to be rare within the time frame studied and occur with about equal incidence regardless of use of a hemi- or total hip arthroplasty on the side of primary interest. The higher risk of dislocation for THA could partly explain the higher risk of hip complications for THA in our study. It should be noted that, during the first years of our study, femoral heads with a diameter of 28 mm dominated in Sweden to gradually become replaced by 32 and 36 mm heads. The cross-matched dataset did not contain information on head size. Differences in implant selection might thus be another reason for variations in results between studies and may also be one explanation for changes over time.

Acta Orthopaedica 2020; 91 (2): 133–138

To be able to compare patients treated with HA and THA, we used propensity score matching. After the matching procedure about one-third of the patients remained for analysis. The entire population of patients with THA and HA are only partly overlapping in terms of age and comorbidity. Therefore, the matching procedure aims to mimic comparison of the 2 methods in the middle group consisting of somewhat frail, sick, and functionally impaired elderly people. A majority of the patients in the fracture arthroplasty population are very frail with substantial impairment pre-fracture. Very few surgeons consider THA suitable for this group and including them would not be clinically relevant. At the other end of the spectrum are those who are healthy and active at the level of younger adults. For them hemiarthroplasty is not used, and again comparisons are not clinically relevant. Thereby we end up with a segment of assumingly comparable individuals, who are included in the ongoing clinical debate. We chose to present the results as odds ratios rather than relative risks. The most prevalent outcome was medical complications, where the largest odds ratio was around 4 (Elixhauser 3+). To approximate the relative risk by this odds ratio would overestimate the true value by close to 80%. Almost all other odds ratios are close to 1, however. To interpret these as relative risks would therefore be possible, and not misleading. The strength of our register-based study is the comparatively large sample of more than 11,000 patients matched with respect to demography, comorbidities, and socioeconomic factors. Since we included all patients between 60 and 95 years of age, irrespective of comorbidities and cognitive function, our results are applicable to almost the entire population of patients treated with hip fracture arthroplasty. RCTs on the other hand comprise much smaller samples and usually only include healthy, cognitively intact, and relatively active individuals, thus excluding most patients with hip fracture. A limitation of our study is the observational design and the inability to control for potential confounding factors not recorded in any of the registers used. An optimum comparison can only be done by randomization, where equal patient groups are studied, but with the limitations mentioned above. In addition, we do not account for patient-reported outcome, which is of paramount interest to fully understand the clinical outcome of a procedure. Meta-analyses can be used to find statistically significant and clinically relevant differences between treatment options even though the separate studies are too small to show a difference on their own. Some of the results in the recent meta-analysis by Xu et al. (2018) have already been discussed. However, Xu’s study did not report on any medical complications and the only hip complications examined were revision, dislocation, and infection. Our study includes a large number of diagnostic codes representing hip-related complications as well as medical complications and therefore gives a more complete picture of the adverse events after hip fracture surgery.

137

Finally, one cannot assume that 1 single implant type will suit all patient groups. In terms of early mortality, a register study indicated that THA in femoral neck fracture cases was comparatively safe in healthy patients less than 80 years old in comparison with those who were older and had several comorbidities (Hailer et al. 2016). The functional benefits with THA suggested by some—but not all—randomized trials (Baker et al. 2006, Keating et al. 2006, Macaulay et al. 2008, Hedbeck et al. 2011) may lie within reach for such relatively “young old” and healthy individuals, given that proper rehabilitation is provided. For the biologically aged, and that is the largest group, HA stands out as a satisfactory alternative. In conclusion, THA as treatment of hip fracture was associated with more hip-related complications than HA. This difference may partly be explained by the use of smaller heads in THA during the study period. Further studies including only contemporary implants are needed to elucidate this issue. In such studies patient-reported outcomes should preferably be included to enable studies of any trade-off between patientreported outcome and hip-related complications. The results on mortality and medical complications are influenced by residual confounding, rather than by the implant design per se. We fail to see how THA, a more strenuous operation with more local complications, should be the explanatory factor for reduced mortality and morbidity. An expansive use of total hip arthroplasties for hip fracture treatment, at the expense of hemiarthroplasties, is not recommended based on our findings if hip complications are to be avoided. Supplementary data The Appendix and Tables 5–7 are available as supplementary data in the online version of this article, http://dx.doi.org/10. 1080/17453674.2019.1690339

Conception and design of study: SH, EB, JK, CR. Acquisition of data: SH, EB, AG, JK, CR. Analysis and interpretation of data: SH, EB, AG, JK, CR. Manuscript drafting: SH, CR. Revision of the manuscript: SH, EB, AG, JK, CR. The authors express their gratitude towards past and present staff at the SHAR office in Gothenburg: Karin Davidsson, Kajsa Erikson, Karin Lindberg, Emma Nauclér, Szilard Nemes, Sandra Olausson, Karin Pettersson, Pär Werner, and Johanna Vinblad. In addition, they acknowledge the important contributions from SHAR coworkers at all Swedish hospitals and the participating Swedish hip arthroplasty patients. Acta thanks Jan-Erik Gjertsen and Lars Gunnar Johnsen for help with peer review of this study.

Avery P P, Baker R P, Walton M J, Rooker J C, Squires B, Gargan M F, Bannister G C. Total hip replacement and hemiarthroplasty in mobile, independent patients with a displaced intracapsular fracture of the femoral neck. J Bone Joint Surg Br 2011; 93–B(8): 1045-8. Baker R P, Squires B, Gargan M F, Bannister G C. Total hip arthroplasty and hemiarthroplasty in mobile, independent patients with a displaced intracapsular fracture of the femoral neck: a randomized, controlled trial. J Bone Joint Surg Am 2006; 88(12): 2583-9.

138

Blomfeldt R, Törnkvist H, Eriksson K, Söderqvist A, Ponzer S, Tidermark J. A randomised controlled trial comparing bipolar hemiarthroplasty with total hip replacement for displaced intracapsular fractures of the femoral neck in elderly patients. J Bone Joint Surg Br 2007; 89(2): 160-5. Brennan T A, Leape L L, Laird N M, Hebert L, Localio A R, Lawthers A G, Newhouse J P, Weiler P C, Hiatt H H. Incidence of adverse events and negligence in hospitalized patients. N Engl J Med 1991; 324(6): 370-6. Burgers P T P W, Van Geene A R, Van Den Bekerom M P J, Van Lieshout E M M, Blom B, Aleem I S, Bhandari M, Poolman R W. Total hip arthroplasty versus hemiarthroplasty for displaced femoral neck fractures in the healthy elderly: a meta-analysis and systematic review of randomized trials. Int Orthop 2012; 36(8): 1549-60. Elixhauser A, Steiner C, Harris D R, Coffey R M. Comorbidity measures for use with administrative data. Med Care 1998; 36(1): 8-27. Hailer N P, Garland A, Rogmark C, Garellick G, Kärrhom J, Kärrholm J. Early mortality and morbidity after total hip arthroplasty in patients with femoral neck fracture. Acta Orthop 2016; 3674: 1-7. Hansson S, Rolfson O, Åkesson K, Nemes S, Leonardsson O, Rogmark C. Complications and patient-reported outcome after hip fracture: a consecutive annual cohort study of 664 patients. Injury 2015; 46(11): 2206-11. Hansson S, Nemes S, Kärrholm J, Rogmark C. Reduced risk of reoperation after treatment of femoral neck fractures with total hip arthroplasty: a matched pair analysis. Acta Orthop 2017; 88(5): 500-4. Hedbeck C-J J, Enocson A, Lapidus G, Blomfeldt R, Törnkvist H, Ponzer S, Tidermark J. Comparison of bipolar hemiarthroplasty with total hip arthroplasty for displaced femoral neck fractures: a concise four-year follow-up of a randomized trial. J Bone Joint Surg 2011; 93(5): 445-50. Hopley C, Stengel D, Ekkernkamp A, Wich M. Primary total hip arthroplasty versus hemiarthroplasty for displaced intracapsular hip fractures in older patients: systematic review. BMJ 2010; 340: c2332. Jameson S S, Lees D, James P, Johnson A, Nachtsheim C, McVie J L, Rangan A, Muller S D, Reed M R. Cemented hemiarthroplasty or hip replacement for intracapsular neck of femur fracture? A comparison of 7732 matched patients using national data. Injury 2013; 44(12): 1940-4. Kärrholm J, Mohaddes M, Odin D, Vinblad J, Rogmark C, Rolfson O. Swedish Hip Arthroplasty Register. Annual report 2017 (in Swedish); 2018. Keating J F, Grant A, Masson M, Scott N W, Forbes J F. Randomized comparison of reduction and fixation, bipolar hemiarthroplasty, and total hip arthroplasty. J Bone Joint Surg 2006; 88(2): 249-60. Liodakis E, Antoniou J, Zukor D J, Huk O L, Epure L M, Bergeron S G. Major complications and transfusion rates after hemiarthroplasty and total hip arthroplasty for femoral neck fractures. J Arthroplasty 2016; 31(9): 2008-12.

Acta Orthopaedica 2020; 91 (2): 133–138

Ludvigsson J F, Andersson E, Ekbom A, Feychting M, Kim J-L, Reuterwall C, Heurgren M, Olausson P O. External review and validation of the Swedish national inpatient register. BMC Public Health 2011; 11(1): 450. Macaulay W, Nellans K W, Garvin K L, Iorio R, Healy W L, Rosenwasser M P. Prospective randomized clinical trial comparing hemiarthroplasty to total hip arthroplasty in the treatment of displaced femoral neck fractures: winner of the Dorr Award. J Arthroplasty 2008; 23(6 Suppl. 1): 2-8. Merten H, Johannesma PC, Lubberding S, Zegers M, Langelaan M, Jukema G N, Heetveld M J, Wagner C. High risk of adverse events in hospitalised hip fracture patients of 65 years and older: results of a retrospective record review study. BMJ Open 2015; 5(9): e006663. Moerman S, Mathijssen N M C, Tuinebreijer W E, Vochteloo A J H, Nelissen R G H H. Hemiarthroplasty and total hip arthroplasty in 30,830 patients with hip fractures: data from the Dutch Arthroplasty Register on revision and risk factors for revision. Acta Orthop 2018; 3674: 1-6. NOMESCO. NOMESCO Classification of Surgical Procedures (NCSP), version 1.16. http://norden.diva-portal.org/smash/record.jsf?pid = diva2% 3A968721&dswid = -9905. Copenhagen; 2011. Rogmark C, Leonardsson O. Hip arthroplasty for the treatment of displaced fractures of the femoral neck in elderly patients. Bone Joint J 2016; 98-B(3): 291-7. Statistics Sweden. Statistics Sweden. https://www.scb.se/; 2018. Thorngren K G. Rikshöft. Annual report 2016 (in Swedish). Lund, Sweden. https://rikshoft.se/ 2017. van den Bekerom M P J, Hilverdink E F, Sierevelt I N, Reuling E M B P, Schnater J M, Bonke H, Goslings J C, van Dijk C N, Raaymakers E L F B. A comparison of hemiarthroplasty with total hip replacement for displaced intracapsular fracture of the femoral neck: a randomised controlled multicentre trial in patients aged 70 years and over. J Bone Joint Surg Br 2010; 92(10): 1422-8. Wang Z, Bhattacharyya T. Outcomes of hemiarthroplasty and total hip arthroplasty for femoral neck fracture: a Medicare cohort study. J Orthop Trauma 2017; 31(5): 260-3. Wang F, Zhang H, Zhang Z, Ma C, Feng X. Comparison of bipolar hemiarthroplasty and total hip arthroplasty for displaced femoral neck fractures in the healthy elderly: a meta-analysis. BMC Musculoskelet Disord 2015; 16: 229. Xu D, Li X, Bi F, Ma C, Lu L, Cao J. Hemiarthroplasty compared with total hip arthroplasty for displaced fractures of femoral neck in the elderly: a systematic review and meta-analysis of fourteen randomized clinical trials. Int J Clin Exp Med 2018; 11(6): 5430-43.

Acta Orthopaedica 2020; 91 (2): 139–145

139

Results after introduction of a hip fracture care pathway: comparison with usual care Stian SVENØY 1,2, Leiv Otto WATNE 3, Ingvild HESTNES 1, Marianne WESTBERG 1, Jan Erik MADSEN 1,2, and Frede FRIHAGEN 1 1 Division of Orthopaedic Surgery, Oslo University Hospital; 2 Institute of Clinical Medicine, University of Oslo; 3 Department of Geriatric Medicine, Oslo University Hospital, Norway Correspondence: ssvenoy@hotmail.com Submitted 2019-04-30. Accepted 2019-11-18.

Background and purpose — We established a care pathway for hip fracture patients, a “Hip Fracture Unit” (HFU), aiming to provide better in-hospital care and thus improve outcome. We compared the results after introduction of the HFU with a historical control group. Patients and methods — The HFU consisted of a series of measures within the orthopedic ward, such as reducing preoperative waiting time, increased use of nerve blocks, early mobilization, and osteoporosis treatment. 276 patients admitted from May 2014 to May 2015 constituted the HFU group and 167 patients admitted from September 2009 to January 2012 constituted the historical control group. Patients were followed prospectively up to 12 months post fracture. Results — Mean preoperative waiting time was 24 hours in the HFU group and 29 hours in the control group (p = 0.003). 123 patients (47%) in the HFU were started on antiosteoporosis treatment while in hospital. “Short Physical Performance Battery” score (SPPB) was mean 5.5 in the HFU group and 3.8 in the control group at 4 months (p < 0.001), and 5.7 vs. 3.6 at 12 months (p < 0.001). The mortality rate at 4 months was 15% in both groups. No statistically significant differences were found in readmissions, complications, new nursing home admissions, in Barthel ADL index or a mental capacity test at the follow-ups. Interpretation — We found improved preoperative waiting time and better SPPB score at 4 and 12 months postoperatively after introducing the HFU.

Elderly hip fracture patients often suffer from comorbidities and the risk for complications is substantial (Haentjens et al. 2010, Smith et al. 2014, Ali and Gibbons 2017). Complication rates are known to increase with prolonged preoperative waiting time (Simunovic et al. 2010, Westberg et al. 2013, Pincus et al. 2017) and acceptable waiting times according to guidelines and national recommendations vary from 24 to 48 hours (AAOS 2014, NICE 2017). The recovery phase after surgery also inflicts a variety of challenges. Loss of function and independence is a risk. About half the hip fracture patients may not regain their pre-fracture mobility level and ability to perform daily activities, which may lead to loss of independence and result in transfer into a permanent care facility (Prestmo et al. 2015, Dyer et al. 2016). Improved perioperative care and early rehabilitation may reduce mortality, prevent loss of function, and be cost effective (Kristensen et al. 2016, NICE 2017). Comprehensive geriatric assessment and orthogeriatrics are recommended for hip fracture patients (Grigoryan et al. 2014, Wang et al. 2015, Eamer et al. 2018, Nordstrom et al. 2018). Quality improvement may also be possible using existing resources within an orthopedic department and without orthogeriatrics, but less has been written about such endeavors (Larsson et al. 2016, Haugan et al. 2017). To our knowledge, no intervention study to date has shown improved functional outcome by introducing a hip fracture care program without formal collaboration with geriatricians (Panella et al. 2018). We established a care pathway, the “Hip Fracture Unit” (HFU), in our department in May 2014, relying mainly on internal resources from the orthopedic department and without orthogeriatric intervention. The HFU was constituted of elements thought to improve the quality of care, such as reducing preoperative waiting time, preoperative femoral nerve block to reduce opiates, early mobilization, and secondary prophylaxis (Lyles et al. 2007,

140

Acta Orthopaedica 2020; 91 (2): 139–145

Table 1. Ward description Description of the orthopedic ward n Number of beds Staff order, numbers per bed Nurses Nursing assistants Physiotherapists Occupational therapists Nutritionists Social worker

52 1.2 0.06 0.07 0 0 0.02

AAOS 2014, Kristensen et al. 2016, NICE 2017, Pincus et al. 2017, Aprato et al. 2018, White et al. 2018). The main aim of this study was to find out whether implementation of an HFU was associated with improved outcome.

Patients and methods We carried out a single-center cohort study with historical controls. In both groups, the patients were prospectively registered. The patients in the HFU group were included from May 2014 to May 2015 and the patients in the control group were included from September 2009 to January 2012. Organization of the Hip Fracture Unit The HFU was established to (a) improve preoperative routines to get the patients to surgery faster, and (b) improve postoperative treatment and routines by increasing focus on early rehabilitation, preventing complications, and enhancing secondary prevention. Interdisciplinary groups were established with orthopedic surgeons, geriatricians, anesthetists, physiotherapists, and orthopedic nurses to outline the clinical routines. Checklists for nurses working on the ward were developed. A “hip fracture nurse” was in charge of the introduction of the care pathway routines in the ward. She was also responsible for educating the other nurses and worked on adherence to routines throughout the intervention, supervised by a senior orthopedic surgeon. Liaisons with pre-hospital services, emergency ward, radiology department, hospital orderlies, clinical biochemistry, and the department of anesthesia were established. A “fast track” admission was planned with direct admission to the ward via radiology, bypassing the emergency room, after conference between pre-hospital services and an orthopedic ward nurse. Regardless of whether or not the fast-track admission route was followed, prompt examination and clearance for surgery was emphasized to all personnel involved, including anesthetists and senior trauma surgeons responsible for emergency surgery prioritizing. A preoperative evaluation by an anesthetist was formalized. Physiotherapists were present in the ward every day including weekend and holidays, to ensure patient mobilization and training (Table 1). Geriatricians, nutritionists, and occupational therapists

were not part of the HFU due to lack of availability. We established a separate admission room on the ward for hip fracture patients. Access to fast track admission was 8 a.m.–8 p.m. on weekdays, and was the preferred route of admission for all hip fracture patients during opening hours when capacity allowed. Preoperative assessment included early examination by the orthopedic surgeon on call and the anesthetist, including nerve block to reduce opioid use (Table 5). The hip fracture nurse registered when nerve block was administered. No changes were made in surgical methods guidelines. Patients were aimed to be mobilized on the first postoperative day and the hip fracture nurse registered whether mobilization actually occurred. All patients were given a standard nutritional supplement drink containing a high-energy triglyceride fat emulsion, proteins, carbohydrates, and vitamins. Recommended osteoporosis treatment was cholecalciferol 100,000 IU orally and zoledronate 5 mg intravenously. When contraindications for bisphosphonates were present, the patients were started on denosumab 60 mg subcutaneously. Patients were also given a daily oral supplement with calcium (500–1,000 mg) and vitamin D (800 IU). Falls prevention was assessed individually. The HFU in the orthopedic ward was established on May 7, 2014, and patients were included for the present study during the first 12 months. Patients with high-energy trauma and patients living in other hospital regions were excluded from the present analyses (Figure 1). Historical control group Data from 167 patients from a previous trial in our hospital on hip fractures, the Oslo Orthogeriatric Trial (OOT), constituted the historical control group (Watne et al. 2014). These 167 patients were the group randomized to “usual care”, i.e., admission to the orthopedic ward. Outcome measures The outcome measures for comparisons with the historical control group were taken from post-discharge outcomes from the OOT (Watne et al. 2014). In addition, selected quality indicators were chosen to evaluate the performance of the HFU (Tables 3 and 5). Both the patients from the HFU and the patients in the control group were followed up at 4 and 12 months postoperatively. The HFU patients were seen in the outpatient clinic by an orthopedic surgeon and the control group were seen on home visits by a trained study nurse. The tests used were: the Consortium to Establish a Registry for Alzheimer’s disease (CERAD), used to measure cognitive function, where patients were asked to recall a set of 10 words presented to them visually (Welsh et al. 1994); the Short Physical Performance Battery (SPPB), combining results of gait speed, balance, and repeated chair stands, a score for mobility and function (Guralnik et al. 1994); the Barthel activities of daily life (ADL) index was used to indicate the patients’ degree of independence (Mahoney and Barthel 1965).

Acta Orthopaedica 2020; 91 (2): 139–145

Hip Fracture Unit Assessed for eligibility n = 314

141

Control group Assessed for eligibility n = 466

Excluded, not meeting inclusion criteria (n = 38): – not hip fracture, 16 – foreign citizen, 6 – from other hospital, 7 – polytrauma, 9

Excluded (n = 299): – not meeting inclusion criteria (n = 53): - not hip fracture, 20 - foreign citizen, 1 - from other hospital, 16 - polytrauma, 16 – dedclined to participate, 22 – other reasons, 59 – randomized to care in geriatric ward, 165

Other outcomes were surgical complications, readmissions, reoperations, secondary prophylaxis, and mortality. The outcomes were registered during the hospital stay and throughout the follow-up period.

Statistics No formal power calculation was performed as Included (n = 167) Included (n = 276) the number of patients available for the historical control groups was fixed. The number of patients Lost to follow-up (n = 65): Lost to follow-up (n = 46): – did not want to participate, 9 – did not want to participate, 16 included from the HFU was decided based on – appoinment not scheduled, 4 – hospitalized/too ill to approach, 4 – dead, 49 – not reached/moved, 1 the power calculations in the OOT using a com– excluded due to contralateral hip – dead, 24 fracture during inclusion period, 3 – excluded due to contralateral hip posite cognitive function outcome measure as fracture during inclusion period, 1 primary outcome measure (Watne et al. 2014) Followed 4 months (n = 121) Followed 4 months (n = 211) and the other Norwegian RCT on orthogeriatrics Incomplete outcome score: Incomplete outcome score: using the SPPB (Prestmo et al. 2015). – CERAD score missing, 18 – CERAD score missing, 5 – SPPB score missing, 13 – SPPB score missing, 2 Normality tests including Q–Q plots and the – ADL score missing, 16 – ADL score missing, 1 Kolmogorov–Smirnov test were used for evaluLost to follow-up (n = 26): Lost to follow-up (n = 35): – did not want to participate, 6 – did not want to participate, 7 ating data distribution. Variables were analyzed – dead, 19 – appoinment not scheduled, 2 by independent sample t-test, Mann–Whitney – hospitalized/too ill to approach, 11 – not reached/moved, 2 U-test and chi-square test depending on the – dead, 14 data distribution. Some of the variables were Followed 12 months (n = 95) Followed 12 months (n = 176) not normally distributed. In these cases, nonIncomplete outcome score: Incomplete outcome score: – CERAD score missing, 20 – CERAD score missing, 3 parametric tests (Mann–Whitney U) were also – SPPB score missing, 2 – SPPB score missing, 3 – ADL score missing, 8 – ADL score missing, 1 performed as sensitivity analyses producing basically the same results. Fisher’s exact test was used where some of the cell numbers were low. Multiple Patient inclusion and follow-up. a Randomized to “usual care” in the orthopedic ward in original study (165 patients were randomized to linear regression analyses were performed to investigate the admission in the geriatric ward and are not included in the present relationships between the HFU and the control group on funcb study). In addition, 11 patients at 4 months and 5 patients at 12 tional outcome; CERAD, SPPB, and ADL scores were chosen months are missing scores due to surgeons other than SS and FF seeing the patients. as dependent variables, and variables believed to influence the outcome (sex, age, ASA score, and pre-fracture residency) were chosen as independent variables. A p-value of < 0.05 was considered statistically significant. Data are presented with Table 2. Baseline characteristics of the Hip Fracture Unit (HFU) care percentages, relative risks (RR), and 95% confidence intervals pathway and the historical control group. Values are frequency (%) (CI) were appropriate. We used SPSS for Windows version 24 unless otherwise specified (IBM Corp, Armonk, NY, USA). a

HFU intervention group Baseline characteristics n = 276 Female sex Age b median (range) ASA b Living in institution Type of fracture Femoral neck Trochanteric Subtrochanteric Surgical procedure Arthroplasty Osteosynthesis Not operated Duration of surgery, min b

OOT a control group n = 167

185 (67) 129 (77) 81 (11) [80–82] 82 (10) [81–84] 84 (49–98) 85 (46–101) 2.6 (0.7) [2.5–2.7] 2.6 (0.6) [2.5–2.7] 59 (21) 51 (30) 151 (56) 113 (42) 5 (2)

98 (59) 67 (40) 2 (1)

127 (46) 149 (54) 0 (0) 71 (31) [67–74]

72 (43) 91 (55) 4 (2) 77 (43) [71–84]

a Oslo Orthogeriatric Trial. b Values are mean (SD) [95%

CI]

Ethics, registration, funding, and potential conflicts of interest The OOT was approved by the regional ethics committee (REK 2009/450). The quality register of the HFU population was approved by the Data Protection Officer (2014/12309 and 2014/1433 REK). The work was funded by the hospital. The authors declare no conflict of interest.

Results From May 2014 to May 2015, 314 patients were assessed for eligibility for the HFU, and 276 patients were included (Figure 1). There were more women in the historical control group and a larger proportion living in an institution before the injury (Table 2).

142

Acta Orthopaedica 2020; 91 (2): 139–145

Table 3. Results and complications before and after introduction of the Hip Fracture Unit Variable

HFU group n = 276

Control group n = 167

Waiting time before surgery, mean h 24 29 Operated within 48 hours, n (%) 260/271 (96) 139/163 (85) Operated within 24 hours, n (%) 157/271 (58) 83/163 (51) Started anti-osteoporosis drug treatment during stay, n (%) 123/273 (47) 1 (0.5) Zoledronic acid 95 1 Alendronate 3 0 Denosumab 25 0 Days of hospital stay, median (range) 5 (2–30) 8 (1–39) Postoperative surgical complications, n (%) 22 (8) 8 (5) Deep wound infection 9 1 Prosthetic dislocation 2 3 Fixation failure 6 3 Other a 5 1 Readmitted in 30 days, n (%) 32/271 (12) 13/165 (8) Reoperated at 4 months, n (%) 16/276 (6) 5/164 (3) at 12 months (accumulated), n (%) 18/276 (7) 9/164 (6) New admissions to a permanent care facility at 4 months, n (%) b 41/172 (24) 18/83 (22) Dead at 4 months 43 (15) 25 (15) Dead at 12 months 63 (23) 43 (26)

Relative risk and/or (CI) (1.6–7.6) 1.1 (1.0–1.2) 1.1 (0.9–1.4)

(2.2–3.7) 1.6 (0.7–3.6)

1.5 (0.8–2.7) 1.9 (0.7–5.1) 1.2 (0.5–2.5) 1.1 (0.6–1.7) 1.0 (0.6–1.6) 0.9 (0.9–1.1)

a Other

surgical complications were: 1 trochanteric avulsion. 1 peroneal nerve palsy. 1 reoperated for hematoma. 1 hemiarthroplasty had soft tissue interposed in the acetabulum. 1 reoperated for gluteus medius insufficiency. 1 had a skin laceration of the leg after a fall in the ward. b Patients who lived in a permanent care facility preoperatively were excluded.

Table 4. Functional outcome at 4 and 12 months postoperatively. Values are mean (SD) Variable a 4 months results SPPB CERAD Barthel ADL 12 months results SPPB CERAD Barthel ADL

HFU group

Control group (CI)

mean 81 years), or ASA score (mean 2.5 versus 2.4). Mean preoperative waiting time from admission to surgery start was reduced by 4.6 hours after introduction of the HFU (Table 3). The HFU group had better SPPB scores at 4 and 12 months than the control group by about 2 points (Table 4). The improvement in SPPB scores was still present after the regression analysis (Table 4). The groups had similar results regarding CERAD and ADL in the adjusted analyses. There was no improvement in readmissions, complications, or mortality after introduction of the HFU. Fast track admission in the Hip Fracture Unit 260/271 (96%) of the patients were operated within 48 hours after admission (Table 5). 55/276 (20%) patients were admitted through the fast-track pathway. 129/276 patients (47%) were admitted during fasttrack opening hours, but 74 of those 129 (57%) were still not admitted through the fast track. The reasons were: failed to alert fast-track team prehospital for 35 patients, lack of ward nurse capacity for 36 patients, and 3 patients were deemed medically unfit for direct ward admission.

Adjusted analysis b p-value B (CI)

Discussion 5.5 (4.7) 13.3 (7.6) 15.7 (5.2)

3.8 (3.4) 11.4 (7.5) 14.6 (5.5)

(0.7 to 2.6) (0.1 to 3.7) (–0.1 to 2.2)

0.03 0.9 (0.1 to 1.7) 0.5 0.5 (–0.9 to 1.9) 0.8 –0.1 (–1.1 to 0.9

We found improved functional outcome at 4 and 12 months postoperatively in the HFU group measured by SPPB compared 5.7 (4.7) 3.6 (3.3) (0.5 to 1.1) 0.02 1.0 (0.2 to 2.0) with the historical controls (see Table 4). 14.2 (8.3) 11.5 (8.4) (0.4 to 4.8) 0.8 0.2 (–1.5 to 1.8) 16.6 (4.8) 14.3 (5.6) (0.9 to 3.5) 0.1 0.9 (–0.1 to 2.0) We improved several key areas of care compared with the historical control group, a Short Physical Performance Battery (SPPB) scale 0–12, Consortium to Establish a including preoperative waiting time and Registry for Alzheimer’s disease (CERAD) scale 0–30, Barthel Activities of Daily Life (ADL) scale 0–20. osteoporosis treatment (Table 3). We had, b Linear regression on mean difference adjusted for age, gender, ASA, and pre-fracture however, low performance in some other accommodation (living in institution or home-dwelling). areas; perhaps most notably that only 1 in 5 patients were admitted through the fasttrack pathway. There was no improveComparison of the HFU group with the historical ment in readmissions or new nursing home admissions. An control group increased number (p = 0.2) of surgical complications was seen 332 patients were assessed at 4 months’ follow-up (211 patients in the HFU group) (Table 3). in the HFU group and 121 patients in the control group). 62 The relationship between preoperative delay and increased patients were lost to follow-up at 12 months (Figure 1). There morbidity and mortality is well published and has led to were no obvious differences between those lost to follow-up guidelines recommending surgery within 24–48 hours and those who were followed by sex (47/62 [76%] females (AAOS 2014, NICE 2017). A retrospective study found versus 200/271 [74%] females), age (mean 80 years versus increased 30-day mortality and more medical complications

Acta Orthopaedica 2020; 91 (2): 139–145

143

Table 5. Results for the Hip Fracture Unit (HFU) during hospital stay. Values are frequency (%) unless otherwise specified Variable Aim (%) Number of patients – Preoperative femoral nerve block – Mean (SD) preoperative waiting time, h – Operated within 48 h (100) Operated within 24 h (> 80) Mobilization started within first day after surgery (100) Urethral catheter removed before 36 hours post-surgery (> 80) Started anti-osteoporosis drug treatment during stay b (> 80) Admitted to other orthopedic wards than the HFU ward None Duration of hospital stay, median days (range) 5–7 a Total number reduced due to missing data. b For anti-osteoporosis drugs used, see Table

Results 276 155/260 (60) a 24 (12) 260/271 (96) a 157/272 (58) a 220/240 (92) a 147/250 (59) a 123/264 (47) a 62/276 (22) 5 (2–30)

when preoperative waiting time exceeded 24 hours (Pincus et al. 2017). To our knowledge, no intervention to reduce preoperative delay showing improved outcome on function, morbidity, and mortality has been published. A randomized controlled trial on fast-track admission versus traditional care pathway in 571 patients found essentially no differences between the groups (Larsson et al. 2016). In the retrospective study from Haugan et al. (2017), admission and ward routines were organized as an HFU similar to ours. They found decreased time to surgery and decreased length of stay (LOS) with fast-track admission compared with usual care. Nevertheless, no differences were observed in mortality or readmission rate. A multicenter quality improvement study from 2018, comparing a care pathway with usual admission and care, found no differences in mortality or functional outcome (Panella et al. 2018). An RCT on preoperative waiting time is being conducted and the results from that trial will probably elucidate this question (Borges et al. 2019). Reduced preoperative waiting time may reduce the LOS, but we believe that the reduced LOS in the HFU was mainly due to improved capacity in the municipal health services (Table 3) (Monkerud and Tjerbo 2016). The SPPB was used to evaluate postoperative function in the former RCT constituting our historical control group (Watne et al. 2014). The minimally meaningful change in SPPB score has been estimated to be 0.5 units (Perera et al. 2006). The between-group difference in our study suggest a clinically meaningful difference (see Table 4). In contrast, the Barthel ADL index showed no statistically significant improvement. There was a concern regarding a high number of postoperative complications in the HFU (8% vs. 5%). This was mainly driven by the difference in deep wound infections. The numbers were low and the difference was not statistically significant (p = 0.2). An increased number of complications may be

an effect of 1 or more components of our HFU, but we have failed to find an obvious link. In previous publications from our department the rate of postoperative deep infections varied from 1% to 9%, still with no apparent explanation for the differences (Frihagen et al. 2007, Westberg et al. 2013, Guren et al. 2017). The intervention in the HFU consisted of multiple smaller elements thought to improve care. It is not possible to discern which elements were beneficial, indifferent, or even harmful for the patients with hip fractures. Adherence to routines varied even though quality improvement work was done by the “hip fracture nurse” and the lead surgeon throughout the study period. None of our predefined aims for quality of care were completely reached after establishing the HFU (Table 5). Orthogeriatric care As described by Haugan et al. (2017), the organization of a clinical care pathway is based on principles from lean methodology. The key concept is standardization of routines and reducing variation in treatment (Niemeijer et al. 2013). Comprehensive geriatric assessment is established in the treatment for hip fractures in some countries, and metaanalyses of high-quality trials have reported improved outcome (Grigoryan et al. 2014, Wang et al. 2015, Eamer et al. 2018, Nordstrom et al. 2018). Introducing orthogeriatric care may, however, be difficult both due to financial and logistical issues, and to shortage of geriatricians. The formation of our HFU was inspired by improvements shown by introducing orthogeriatric care, but due to restraint on resources and logistics we were not able to establish an orthogeriatric service (Grigoryan et al. 2014, Prestmo et al. 2015, Eamer et al. 2017, Eamer et al. 2018, Nordstrom et al. 2018). Our perspective may be relevant in healthcare systems where geriatric resources are unavailable or where there is lack of willingness to pay for interdisciplinary services. Our HFU also had elements of a fast-track pathway, but the low number of patients using this pathway may imply that our HFU did not have adequate resources or a robust enough solution for the admission routine. Even with limited “opening hours” due to ward staffing we still had problems maintaining the admissions routine when it should have been available. From this experience we will in the future seek to establish routines that are intended to be available 24/7. Strength and weaknesses of the study The correlation between improvements made by introducing the HFU and the clinical scores at 4 and 12 months requires careful interpretation. Patients in the HFU were about 1.5 years younger and 9% fewer patients lived in an institution preoperatively. On the other hand, ASA scores for the 2 groups were similar. There were also more men in the HFU group, and studies have indicated that being male increases the risk of mortality and adverse events after a hip fracture (Brether-

144

ton and Parker 2015, Sathiyakumar et al. 2015). Still, there may have been systematic differences between the groups that our regression analysis failed to adjust for. Papers evaluating a quality improvement effort like ours are vulnerable to misinterpreting multifactorial improvements over time as being a result of the intervention, so any positive finding should be interpreted with caution. The patient populations were included in 2 different time periods and an improvement of care may have come about independently of our intervention, through an increased awareness of the needs of hip fracture patients. The functional outcome data were collected at home visits for the control group and at outpatient clinic visits for the HFU group. The evaluators performing the home visits in the control group were blinded to treatment group, but the orthopedic surgeons examining the HFU group were part of the HFU team. This may have led to bias. It does not, however, seem that it led to a selection of healthier patients in the HFU follow-ups, and there were still fewer patients lost to follow-up in the HFU group, although more patients had missing individual outcome scores. Some patients were mistakenly scheduled for follow-up appointment with other colleagues who were not aware of the study follow-up protocol. Thus, the clinical scores were not obtained for these patients (Figure 1). Our control group was part of an earlier RCT with wide inclusion criteria, and we believe it to be close to an unbiased population. The intervention group was also unselected, and loss to follow-up was low in both groups. The main strengths of our study are the prospective registration and the systematic follow-up, including functional outcome measures. Conclusion A low-cost HFU within the orthopedic department may be a tempting alternative when resources are limited. We failed in our attempt at fast-track admission, but reduced mean time to surgery by almost 5 hours and achieved reasonable numbers on our performance indicators. Apart from the improved SPPB during follow-up, we found limited or no effect of our HFU after discharge. A care pathway like ours may still be attempted in hospitals where quality improvement is sought. In our experience, however, some resources must be added to initiate and sustain a care pathway, and the literature supports the addition of comprehensive geriatric assessment and the formation of a strong interdisciplinary team.

All the authors were responsible for planning the study. SS performed the statistical analyses and wrote the first draft. All authors participated in the interpretation of data, and critical revision of the manuscript. We thank research nurses Elisabeth Fragaat, Tone Fredriksen, Camilla Marie Andersen, Julie Ask Ottesen, Linda Feldt, Sissel Knuts, and Elise Berg Vesterhus for assisting in data collection. Acta thanks Pia Kjær Kristensen for help with peer review of this study.

Acta Orthopaedica 2020; 91 (2): 139–145

AAOS. Management of hip fractures in the elderly; Retrieved September 2014. Available from https://www.aaos.org/research/guidelines/HipFxGuideline_rev.pdf. Ali A M, Gibbons C E. Predictors of 30-day hospital readmission after hip fracture: a systematic review. Injury 2017; 48(2): 243-52. Aprato A, Audisio A, Santoro A, Grosso E, Devivo S, Berardino M, Masse A. Fascia-iliaca compartment block vs intra-articular hip injection for preoperative pain management in intracapsular hip fractures: a blind randomized controlled trial. Injury 2018; 49(12): 2203-8. Borges F, Bhandari K M, Patel A, Avram V, Guerra-Farfan E, Sigamani A, Umer M, Tiboni M, Adili A, Neary J, Tandon V, Sancheti P K, Lawendy A, Jenkinson R, Ramokgopa M, Biccard B M, Szczeklik W C, Wang Y, Landoni G, Forget P, E Popova, Wood G, Nabi Nur A, B John, Sleczka P, Feibel R J, Balaguer-Castro M, Deheshi B, Winemaker M, de Beer J, Kolesar R, Teixidor-Serra J, TomasHernandez J, McGillion M, Shanthanna H, Moppett I, Vincent J, Pettit S, Harvey V, Gauthier L, Alvarado K, Devereaux P J. Rationale and design of the HIP fracture Accelerated surgical TreaTment And Care tracK (HIP ATTACK) Trial: a protocol for an international randomised controlled trial evaluating early surgery for hip fracture patients. BMJ Open 2019; 9(4): e028537. Bretherton C P, and Parker M J. Early surgery for patients with a fracture of the hip decreases 30-day mortality. Bone Joint J 2015; 97-B(1): 104-8. Dyer S, Crotty M M, Fairhall N, Magaziner J L, Beaupre A, Cameron I D, Sherrington C G. Fragility Fracture Network Rehabilitation Research Special Interest: a critical review of the long-term disability outcomes following hip fracture. BMC Geriatr 2016; 16: 158. Eamer G, Saravana-Bawan B, van der Westhuizen B, Chambers T, Ohinmaa A, Khadaroo R G. Economic evaluations of comprehensive geriatric assessment in surgical patients: a systematic review. J Surg Res 2017; 218: 9-17. Eamer G ,Taheri A, Chen S S, Daviduck Q, Chambers T, Shi X, Khadaroo R G. Comprehensive geriatric assessment for older people admitted to a surgical service. Cochrane Database Syst Rev 2018; 1: CD012485. Frihagen F, Nordsletten L, Madsen J E. Hemiarthroplasty or internal fixation for intracapsular displaced femoral neck fractures: randomised controlled trial. BMJ 2007; 335(7632): 1251-4. Grigoryan K, Javedan V H, Rudolph J L. Orthogeriatric care models and outcomes in hip fracture patients: a systematic review and meta-analysis. J Orthop Trauma 2014; 28(3): e49-55. Guralnik J M, Simonsick E M, Ferrucci L, Glynn R J, Berkman L F, Blazer D G, Scherr P A, Wallace R B. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol 1994; 49(2): M85-94. Guren E, Figved W, Frihagen F L, Watne O, Westberg M. Prosthetic joint infection: a devastating complication of hemiarthroplasty for hip fracture. Acta Orthop 2017; 88(4): 383-9. Haentjens P Magaziner J, Colon-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. Haugan K L, Johnsen G, Basso T, Foss O A. Mortality and readmission following hip fracture surgery: a retrospective study comparing conventional and fast-track care. BMJ Open 2017; 7(8): e015574. Kristensen P K, Thillemann T M, Soballe K, Johnsen S P. Are process performance measures associated with clinical outcomes among patients with hip fractures? A population-based cohort study. Int J Qual Health Care 2016; 28(6): 698-708. Larsson G R, Stromberg U, Rogmark C, Nilsdotter A. Prehospital fast track care for patients with hip fracture: impact on time to surgery hospital stay post-operative complications and mortality a randomised controlled trial. Injury 2016; 47(4): 881-6. Lyles K W, Colon-Emeric C S, Magaziner J S, Adachi J D, Pieper C F, Mautalen C, Hyldstrup L, Recknor C, Nordsletten L, Moore K A, Lavecchia C, Zhang J, Mesenbrink P, Hodgson P K, Abrams K, Orloff J J, Horowitz Z, Eriksen E F, Boonen S, Horizon Recurrent Fracture Trial. Zoledronic acid and clinical fractures and mortality after hip fracture. N Engl J Med 2007; 357(18): 1799-1809. Mahoney F I, Barthel D W. Functional evaluation: the Barthel Index. Md State Med J 1965; 14: 61-5.

Acta Orthopaedica 2020; 91 (2): 139â&#x20AC;&#x201C;145

Monkerud L C, Tjerbo T. The effects of the Norwegian Coordination Reform on the use of rehabilitation services: panel data analyses of service use 2010 to 2013. BMC Health Serv Res 2016; 16(a): 353. NICE. NICE guidelines for hip fracture. Available from https://www.nice.org. uk/guidance/cg124. Retrieved May 2017. Niemeijer G, Flikweert C E, Trip A, Does R J, Ahaus K T, Boot A F, Wendt K W. The usefulness of lean six sigma to the development of a clinical pathway for hip fractures. J Eval Clin Pract 2013; 19(5): 909-14. Nordstrom P K, Thorngren G, Hommel A, Ziden L, Anttila S. Effects of geriatric team rehabilitation after hip fracture: meta-analysis of randomized controlled trials. J Am Med Dir Assoc 2018; 19(10): 840-5. Panella M, Seys D, Sermeus W, Bruyneel L, Lodewijckx C, Deneckere S, Sermon A, Nijs S, Boto P, Vanhaecht K. Minimal impact of a care pathway for geriatric hip fracture patients. Injury 2018; 49(8): 1581-6. Perera S, Mody S H, Woodman R C, Studenski S A. Meaningful change and responsiveness in common physical performance measures in older adults. J Am Geriatr Soc 2006; 54(5): 743-9. Pincus D, Ravi B, Wasserstein D, A Huang, Paterson J M, Nathens A B, Kreder H J, Jenkinson R J, Wodchis W P. Association between wait time and 30-day mortality in adults undergoing hip fracture surgery. JAMA 2017; 318(20): 1994-2003. Prestmo A, Hagen G, Sletvold O, Helbostad J, Thingstad L P, Taraldsen K, Lydersen S, Halsteinli V, Saltnes T, Lamb S E, Johnsen L G, Saltvedt I. Comprehensive geriatric care for patients with hip fractures: a prospective randomised controlled trial. Lancet 2015; 385(9978): 1623-33. Sathiyakumar V, Greenberg S E, Molina C S, Thakore R V, Obremskey W T, Sethi M K. Hip fractures are risky business: an analysis of the NSQIP data. Injury 2015; 46(4): 703-8.

145

Simunovic N, Devereaux P J, Sprague S, Guyatt G H, Schemitsch E, Debeer J, Bhandari M. Effect of early surgery after hip fracture on mortality and complications: systematic review and meta-analysis. CMAJ 2010; 182(15): 1609-16. Smith T, Pelpola K, Ball M, Ong A, Myint P K. Pre-operative indicators for mortality following hip fracture surgery: a systematic review and metaanalysis. Age Ageing 2014; 43(4): 464-71. Wang H, Li C, Zhang Y, Jia Y, Zhu Y, Sun R, Li W, Liu Y. The influence of inpatient comprehensive geriatric care on elderly patients with hip fractures: a meta-analysis of randomized controlled trials. Int J Clin Exp Med 2015; 8(11): 19815-30. Watne L, O Torbergsen A C, Conroy S, Engedal K, Frihagen F, Hjorthaug G A, Juliebo V, Raeder J, Saltvedt I, Skovlund E, Wyller T B. The effect of a pre- and postoperative orthogeriatric service on cognitive function in patients with hip fracture: randomized controlled trial (Oslo Orthogeriatric Trial). BMC Med 2014; 12: 63. Welsh K A, Butters N, Mohs R C, Beekly D, Edland S, Fillenbaum G, Heyman A. The Consortium to Establish a Registry for Alzheimerâ&#x20AC;&#x2122;s Disease (CERAD), Part V: A normative study of the neuropsychological battery. Neurology 1994; 44(4): 609-14. Westberg M, Snorrason F, Frihagen F. Preoperative waiting time increased the risk of periprosthetic infection in patients with femoral neck fracture. Acta Orthop 2013; 84(2): 124-9. White S M, Altermatt F, Barry J, Ben-David B, Coburn M, Coluzzi F, Degoli M, Dillane, D, Foss N B, Gelmanas A, Griffiths R, Karpetas G, Kim J H, Kluger M, Lau P W, Matot I, McBrien M, McManus S, Montoya-Pelaez L F, Moppett I, Parker K M, Porrill O, Sanders R D, Shelton C, Sieber F, Trikha A, Xuebing X. International Fragility Fracture Network Delphi consensus statement on the principles of anaesthesia for patients with hip fracture. Anaesthesia 2018; 73(7): 863-74.

146

Acta Orthopaedica 2020; 91 (2): 146–151

Cognitive impairment influences the risk of reoperation after hip fracture surgery: results of 87,573 operations reported to the Norwegian Hip Fracture Register Målfrid Holen KRISTOFFERSEN 1,2, Eva DYBVIK 1, Ole Martin STEIHAUG 3, Torbjørn Berge KRISTENSEN 1,2, Lars Birger ENGESÆTER 1, Anette Hylen RANHOFF 4,5, and Jan-Erik GJERTSEN 1,2 1 Norwegian

Hip Fracture Register, Department of Orthopedic Surgery, Haukeland University Hospital, Bergen; 2 Department of Clinical Medicine, Faculty of Medicine, University of Bergen, Bergen; 3 Emergency Department, Haukeland University Hospital, Bergen; 4 Diakonhjemmet Hospital, Oslo; 5 Department of Clinical Sciences, Faculty of Medicine, University of Bergen, Bergen, Norway Correspondence: malfrid.holen.kristoffersen@helse-bergen.no Submitted 2019-08-20. Accepted 2019-12-09.

Background and purpose — About one-fourth of hip fracture patients have cognitive impairment. We investigated whether patients’ cognitive function affects surgical treatment, risk of reoperation, and mortality after hip fracture, based on data in the Norwegian Hip Fracture Register (NHFR). Patients and methods — This prospective cohort study included 87,573 hip fractures reported to the NHFR in 2005– 2017. Hazard rate ratios (HRRs) for risk of reoperation and mortality were calculated using Cox regression adjusted for sex, age, ASA class, fracture type, and surgical method. Results — Cognitive impairment was reported in 27% of patients. They were older (86 vs. 82 years) and had higher ASA class than non-impaired patients. There were no differences in fracture type or operation methods. Cognitively impaired patients had a lower overall reoperation rate (4.7% vs. 8.9%, HRR 0.71; 95% CI 0.66–0.76) and lower risk of reoperation after osteosynthesis (HRR 0.58; CI 0.53–0.63) than non-impaired patients. Cognitively impaired hip fracture patients had an increased reoperation risk after hemiarthroplasty (HRR 1.2; CI 1.1–1.4), mainly due to dislocations (1.5% vs. 1.0%, HRR 1.7; CI 1.3–2.1). Risk of dislocation was particularly high following the posterior approach (4.7% vs. 2.8%, HRR 1.8; CI 1.2–2.7). Further, they had a higher risk of reoperation due to periprosthetic fracture after uncemented hemiarthroplasty (HRR 1.6; CI 1.0–2.6). Cognitively impaired hip fracture patients had higher 1-year mortality than those without cognitive impairment (38% vs. 16%, HRR 2.1; CI 2.1–2.2). Interpretation — Our findings support giving cognitively impaired patients the same surgical treatment as nonimpaired patients. But since the risk of hemiprosthesis dislocation and periprosthetic fracture was higher in cognitively impaired patients, they should probably not have posterior approach surgery or uncemented implants.

In Norway, with a population of 5.2 million, about 9,000 patients are treated for a hip fracture each year (Gjertsen et al. 2008). A high proportion of hip fracture patients have cognitive impairment (Mundi et al. 2014, Mukka et al. 2017, Kristoffersen et al. 2019). Cognitive impairment is defined as a decrease in cognition beyond normal aging (Hugo and Ganguli 2014). It can be mild, it can include dementia, or it might be temporary such as in delirium (Petersen et al. 2001, Holsinger et al. 2007). Dementia is usually diagnosed according to ICD-10 criteria in Norway (Naik and Nygaard 2008), and is dependent on a history of cognitive impairment of at least 6 months’ duration in activities of daily living. Despite high prevalence of cognitive impairment among hip fracture patients, these patients are often excluded from research (Mundi et al. 2014). We investigated whether the presence of cognitive impairment affects the choice of surgical treatment for different types of hip fractures, and evaluated whether patients with cognitive impairment have a different risk of reoperation and mortality compared with cognitively fit patients.

Patients and methods Study design This is a prospective observational study based on data from the Norwegian Hip Fracture Register (NHFR). The NHFR collects data from all hospitals in Norway treating hip fractures (Gjertsen et al. 2008). Data are reported by the surgeon on a 1-page form with information on the fracture type, the operation method, and the patient, including assessment of cognitive impairment. Femoral neck fractures are classified according to the Garden classification. Trochanteric fractures are classified according to the AO/OTA classification.

Acta Orthopaedica 2020; 91 (2): 146–151

Cases in the Norwegian Hip Fracture Register 2005–2017 n = 104,980 Excluded (n = 11,060): – pathological fractures, 1,356 – patients < 65 years, 9,704 Cases eligible for inclusion n = 93,920 Excluded (n = 2,873): – total arthroplasty, 2,018 – ASA 5, 137 – other type of fracture, 718 Excluded (n = 3,474) due to missing data on: – type of fracture, 33 – type of treatment, 208 – ASA, 1,262 – cognitive status, 1,971

Figure 1. Flowchart.

Included cases n = 87,573

The surgeon evaluates patients’ cognitive function by examining their medical chart, asking them or their relatives, or using the Clock Drawing Test (Amodeo et al. 2015). Since the form is completed immediately after the operation, the information on cognitive function must be collected preoperatively. The NHFR has no data on the methods the surgeons used to obtain information on cognitive function. The question concerning cognitive impairment on the form is: “Does the patient have cognitive impairment?” Surgeons answer “Yes,” “No,” or “Uncertain.” The data on cognitive impairment reported to the NHFR have been validated against external quality databases. The positive predictive value of the data reported to the NHFR on cognitive impairment was 78% (Kristoffersen et al. 2019). The completeness of reporting of primary hip fracture operations to the NHFR has been found to be 88% for osteosynthesis and 94% for hemiarthroplasty when compared with the Norwegian Patient Register (Furnes et al. 2017). Reoperations are linked to the primary operation by the unique identification number assigned to each inhabitant in Norway. Total hip arthroplasty revisions are reported on separate operation forms to the Norwegian Arthroplasty Register and later duplicated to the files of the NHFR. It is possible to report several reasons for each reoperation, and a hierarchy of reasons was drawn up. If a deep or superficial infection was present, this was defined as the main reason for reoperation. Patient selection In the period 2005–2017, 104,980 primary hip fracture operations were reported to the NHFR. For the present study, pathological fractures and fractures in patients younger than 65 years of age were excluded (n = 11,060). Total hip arthroplasty for hip fracture was also excluded, since these operations are reported on separate forms to the Norwegian Arthro-

147

plasty Register with no information on cognitive function (n = 2,018). Further, fractures in ASA 5 patients, other fracture types than femoral neck, trochanteric or subtrochanteric fractures, operations with missing data on type of fracture, type of surgery, ASA classification, and cognitive status were excluded (n = 4,329) (Figure 1). Finally, 87,573 operations were included in the analysis. Statistics The patients were analyzed in groups according to their cognitive function: cognitively impaired, cognitively fit, and uncertain cognitive function (where the surgeon was uncertain of the patient’s cognitive function). Pearson’s chisquare test was used to compare categorical variables. Independent samples t-tests and analyses of variance (ANOVA), were used to compare the means for continuous variables. P-values < 0.05 were considered statistically significant. The Kaplan–Meier method was used to calculate time from primary surgery to reoperation. Hazard rate ratios (HRRs) are presented with 95% confidence intervals (CIs). Differences in reoperation risks between the groups were calculated using a Cox regression model with adjustments for sex, age, ASA class, fracture type, and operation method. Separate analyses were conducted for reoperations after primary osteosynthesis and those following hemiarthroplasty. Sub-analyses were performed for reoperations after hemiarthroplasty by surgical approach and fixation method. Further, the Cox regression model was used to analyze differences in mortality between the different patient groups with patients with no cognitive impairment as reference. 30-day, 90-day, and 1-year mortality were calculated with adjustments for sex, age, ASA, fracture type, and operation method. The proportional hazards assumption was fulfilled when investigated visually using log-minus-log plots. Fine and Gray analysis was also used to determine whether mortality was a competing risk in reoperation. The statistical software package IBM SPSS Statistics, version 24.0 (IBM Corp, Armonk, NY, USA) and the statistical package R, version 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria) were used for the statistical analysis. The study was performed in accordance with the REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement (Benchimol et al. 2015). Ethics, funding, and potential conflict of interest The NHFR has permission from the Norwegian Data Protection Authority to collect and store data on hip fracture patients (permission issued January 3, 2005; reference number 2004/1658-2 SVE/-). The patients signed a written, informed consent declaration, and when unable to understand or sign, their next of kin could sign the consent form on their behalf. The Norwegian Hip Fracture Register is financed by the Western Norway Regional Health Authority. No competing interests were declared.

148

Acta Orthopaedica 2020; 91 (2): 146–151

Table 1. Baseline data for patients by cognitive function. Values are frequency (%) unless otherwise specified Factor

Total

Cognitive impairment No Uncertain Yes

Total 87,573 54,859 (63) 8,985 (10) Women 62,751 (72) 39,182 (71) 6,332 (71) Mean age (SD) 83.2 (7.5) 82.0 (7.8) 84.8 (7.0) Age group 65–74 12,611 (14) 10,388 (19) 793 (8.8) 75–79 12,837 (15) 9,120 (17) 1,099 (12) 80–84 20,309 (23) 12,727 (23) 2,028 (23) 85–89 23,494 (27) 13,247 (24) 2,754 (31) ≥ 90 18,322 (21) 9,377 (17) 2,311 (26) ASA class ASA 1+2 32,293 (37) 24,298 (44) 2,485 (28) ASA 3+4 55,280 (63) 30,561 (56) 6,500 (72) Fracture type Undisplaced FNF 12,782 (15) 8,166 (15) 1,223 (14) Displaced FNF 37,006 (42) 22,978 (42) 3,780 (42) Basocervical FNF 3,112 (3.6) 1,918 (3.5) 328 (3.7) a Trochanteric A1 14,768 (17) 9,168 (17) 1,549 (17) Trochanteric A2 a 14,012 (16) 8,743 (16) 1,512 (17) Trochanteric A3 a 1,439 (1.6) 931 (1.7) 143 (1.6) Subtrochanteric 4,454 (5.1) 2,955 (5.4) 450 (5.0) Primary operation Screw osteosynthesis 16,938 (19) 10,483 (19) 1,707 (19) Hemiarthroplasty 32,667 (37) 20,522 (37) 3,284 (37) Sliding hip screw 27,161 (31) 16,956 (31) 2,827 (31) Short IM nail 7,265 (8.3) 4,529 (8.3) 815 (9.1) Long IM nail 3,542 (4.0) 2,369 (4.3) 352 (3.9) Surgical approach Anterior/anterolateral 2,495 (7.6) 1,604 (7.8) 254 (7.7) Lateral 26,401 (81) 16,596 (81) 2,680 (82) Posterior 3,286 (10) 2,008 (9,8) 308 (9.4) Other/missing data 485 (1.5) 314 (1.5) 42 (1.3) Fixation of HA Cemented 24,278 (74) 15,353 (75) 2,408 (73) Uncemented 7,851 (24) 4,854 (24) 804 (25) Missing data 538 (1.6) 315 (1.5) 72 (2.2)

23,729 (27) 17,237 (73) 85.5 (6.4) 1,430 (6.0) 2,618 (11) 5,554 (23) 7,493 (32) 6,634 (28) 5,510 (23) 18,219 (77) 3,393 (14) 10,248 (43) 866 (3.6) 4,051 (17) 3,757 (16) 365 (1.5) 1,049 (4.4) 4,748 (20) 8,861 (37) 7,378 (31) 1,921 (8.1) 821 (3.5) 637 (7.2) 7,125 (80) 970 (11) 129 (1.4) 6,517 (74) 2,193 (25) 151 (1.7)

FNF = femoral neck fracture, IM = intramedullary, HA = hemiarthroplasty. AO/OTA classification.

Table 2. Number of reoperations and risk of reoperation after hip fracture surgery by cognitive function using Cox regression model and Fine and Gray model with adjustments for age, sex, ASA classification, fracture type, and treatment Cognitive Total Reoperation impairment n n (%)

Cox regression Hazard Rate ratio (95% CI)

Fine and Gray Hazard Rate ratio (95% CI)

Total No Uncertain Yes

87,573 54,859 8,985 23,729

6,568 (7.5) 4,860 (8.9) 1 Reference 1 Reference 598 (6.7) 0.91 (0.83–0.99) 0.91 (0.84–0.99) 1,110 (4.7) 0.71 (0.66–0.76) 0.69 (0.65–0.74)

Hemiarthroplasty No Uncertain Yes

32,667 20,522 3,284 8,861

1,425 (4.4) 873 (4.3) 1 Reference 1 Reference 169 (5.1) 1.3 (1.1–1.6) 1.3 (1.1–1.6) 383 (4.3) 1.2 (1.1–1.4) 1.2 (1.0–1.3)

Osteosynthesis No Uncertain Yes

54,906 34,337 5,701 14,868

5,143 (9.4) 3,987 (11) 1 Reference 1 Reference 429 (7.5) 0.81 (0.73–0.89) 0.85 (0.77–0.94) 727 (4.9) 0.58 (0.53–0.63) 0.62 (0.57–0.67)

Results In the 87,573 hip fracture operations, 27% of the patients had been classified by the surgeon as cognitively impaired and 63% as cognitively fit. In 10% of the operations the surgeon had evaluated the patient’s cognitive function as “uncertain.” The mean follow-up time was 3.0 years (3.0–3.0). Patients with cognitive impairment had a mean follow-up time of 1.8 years (1.8–1.9), non-impaired patients 3.6 years (3.5–3.6) and “uncertain” patients 2.5 years (2.5–2.6). Baseline data There were 72% women among the patients. The patients with cognitive impairment were on average 3.5 years older and had more severe comorbidity (higher ASA score) than non-impaired patients (Table 1). Displaced femoral neck fractures (FNFs) constituted 42% of all fractures. Only small differences in the distribution of fractures and operation methods were found between the groups but, due to the large numbers, some of these small differences were statistically significant (Table 1). Surgical methods for each fracture type were not influenced by the patients’ cognitive function (Figure 2, see Supplementary data). The most common operation methods were hemiarthroplasty (37%) and osteosynthesis with a sliding hip screw (31%) (Table 1). Most hemiarthroplasties were performed with a lateral approach (81%) and three-quarters of hemiarthroplasties were cemented (Table 1). Reoperations Cox regression analysis and the Fine and Grey method showed a similar risk of reoperation (Ranstam and Robertsson 2017) (Table 2). The overall reoperation rate for all patients was 7.5% (n = 6,568) (Table 2). Patients with cognitive impairment had an overall reoperation rate of 4.7%, compared with 8.9% for cognitively fit patients (HRR 0.71; CI 0.66–0.76). Patients with “uncertain” cognitive function had a reoperation rate of 6.7% (HRR 0.91; CI 0.83–0.99). The overall reoperation rates for all patients were 4.4% after hemiarthroplasty and 9.4% after osteosynthesis. The reoperation risk for patients with cognitive impairment was

Acta Orthopaedica 2020; 91 (2): 146–151

149

slightly higher for hemiarthroplasty Table 3. Reasons for reoperation after hemiarthroplasty and osteosynthesis. Reoperations (HRR 1.2; CI 1.1–1.4) but lower for appear in the order of our hierarchy. Values are frequency (%) osteosynthesis (HRR 0.58; CI 0.53– Cognitive impairment 0.63) than for those without cogni- Factor Total No Uncertain Yes tive impairment (Table 2). There were small differences in All reoperations 6,568 (7.5) 4,860 (8.9) 598 (6.7) 1,110 (4.7) 1,425 (4.4) 873 (4.4) 169 (5.1) 383 (4.3) risk of reoperation between patients Reoperation after hemiarthroplasty Infection 672 (2.1) 416 (2.0) 81 (2.5) 175 (2.0) with and without cognitive impairPeriprosthetic fracture 151 (0.5) 90 (0.4) 17 (0.5) 44 (0.5) ment for those operated with hemiDislocation of prosthesis 395 (1.2) 206 (1.0) 55 (1.7) 134 (1.5) Loosening of hemiarthroplasty 18 (0.1) 17 (0.1) 0 (0.0) 1 (0.0) arthroplasty due to infection and a Sequelae of femoral neck fracture 31 (0.1) 24 (0.1) 2 (0.1) 5 (0.1) periprosthetic fracture. Other reason 158 (0.5) 120 (0.5) 14 (0.4) 24 (0.3) Analysis by fixation of the Reoperation after osteosynthesis 5,143 (9.4) 3,987 (12) 429 (7.5) 727 (4.9) Infection 225 (0.4) 136 (0.4) 29 (0.5) 60 (0.4) hemiprosthesis showed that patients Peri-implant fracture 363 (0.7) 247 (0.7) 34 (0.6) 82 (0.6) with cognitive impairment treated Avascular necrosis 346 (0.6) 248 (0.7) 29 (0.5) 69 (0.5) with uncemented hemiarthroplasty Osteosynthesis failure 1,541 (2.8) 1022 (3.0) 172 (3.0) 320 (2.2) Cut-out 142 (0.3) 107 (0.3) 12 (0.2) 23 (0.2) had a higher risk of reoperation for Non-union 276 (0.5) 212 (0.6) 27 (0.5) 37 (0.2) any reason (HRR 1.3; CI 1.1–1.7) Sequelae of proximal femoral fracture a 1,744 (3.2) 1,568 (4.6) 96 (1.7) 80 (0.5) and a particularly high risk due to Local pain due to osteosynthesis material 360 (0.7) 318 (0.9) 15 (0.3) 27 (0.2) Other reason 173 (0.3) 129 (0.4) 15 (0.3) 29 (0.2) periprosthetic fracture (HRR 1.6; CI 1.0–2.6), compared with patients a Reoperation with total hip arthroplasty reported to the Norwegian Arthroplasty Register. without cognitive impairment. No such differences could be found for cemented hemiarthroplasty. Further, cognitively impaired patients treated with hemiarthroplasty patients with cognitive impairment found in our study does had a higher risk of reoperation because of dislocation than not necessarily imply that these patients do better than those non-impaired patients (1.5% vs. 1.0%, HRR 1.7; CI 1.3–2.1) without cognitive impairment. Patients with cognitive impairment have been reported to (Table 3). Analysis by surgical approach showed that this risk was higher with the posterior approach (4.7% vs. 2.8%, HRR have a higher risk of poorer functional outcome after hip frac1.8; CI 1.2–2.7) and lower with the lateral approach (1.1% vs. ture incidents (Sheehan et al. 2018). Hip fracture patients with cognitive impairment are older and have comorbidities that 0.8%, HRR 1.5; CI 1.1–2.0). Few patients with cognitive impairment were reoperated increase the risk of any reoperation. It is easier for cognitively due to osteosynthesis failure and local pain (Table 3). Only fit patients to tolerate the peri- and postoperative strain and 0.5% of cognitively impaired patients treated with osteosyn- stress of revision surgery. Patients with cognitive impairment thesis had revision total hip arthroplasty, compared with 4.6% might not be offered surgical revision due to a higher risk of complications such as prosthesis dislocation and shorter life of cognitively fit patients. expectancy than in non-impaired patients. Mortality An infection is probably the most feared complication after 30-day mortality was 13% for cognitively impaired patients hip fracture surgery. In most cases, an infection leaves no and 4.6% for cognitively fit patients (HRR 2.2; CI 2.1–2.3). other options than surgical debridement. Notably, cognitive 90-day mortality was 23% for cognitively impaired patients impairment, in our study, did not seem to increase the risk and 8.5% for cognitively fit patients (HRR 2.2; CI 2.1–2.3). of reoperation due to infection. Cognitively impaired patients Finally, 1-year mortality was 38% for cognitively impaired treated with hemiarthroplasty had an increased risk of prospatients and 16% for cognitively fit patients (HRR 2.1; CI: thesis dislocation, especially when the posterior approach 2.1–2.2) (Table 4, see Supplementary data). Patients with had been used. Our results concur with those in the study by cognitive impairment had a greater overall mortality risk than Svenøy et al. (2017), who reported an 8-fold increase in risk of cognitively fit patients (HRR 2.1; CI 2.0–2.1). dislocation after the posterior approach compared with the lateral. Our results suggest that the use of the posterior approach in cognitively impaired patients should be avoided. It is well established that uncemented hemiarthroplasties Discussion have a higher risk of revision than cemented (Langslet et al. There was no difference in type of fracture or type of initial 2014, Kristensen et al. 2020). treatment among hip fracture patients in relation to cognitive In our study, cognitively impaired patients treated with uncefunction in NHFR. This supports the idea of equal treatment mented hemiarthroplasty had a higher risk of reoperation for for all hip fracture patients. The lower reoperation rate for any reason and for periprosthetic fracture than non-impaired

150

patients. No such differences were found for cemented hemiarthroplasties. Thus, uncemented hemiarthroplasties seem to yield inferior results and should not be used in cognitively impaired patients who may have a particularly high risk of recurrent falls and periprosthetic fracture. Very few patients with cognitive impairment were reoperated with a total hip arthroplasty, which may be contraindicated in these patients because of lack of compliance and increased risk of dislocation. However, the risk of dislocation can be reduced with the use of a dual-mobility cup (Jobory et al. 2019). Our study also included patients where the orthopedic surgeon had been in doubt whether the patient had cognitive impairment or not. These patients performed as an intermediate group in our analysis. One explanation could be that these patients may have had delirium, which is common in patients with hip fracture and complicates the assessment of chronic cognitive impairment and dementia. Delirium is also a risk factor for developing dementia after a hip fracture (Krogseth et al. 2011). Mortality increased 2-fold for patients with cognitive impairment, both from 30 to 90 days and from 90 days to 1 year. This finding is in line with previous studies (Söderqvist et al. 2006, Mukka et al. 2017). Our study does not include information on causes of mortality. Holvik et al. (2010) found that predictors of mortality in older hip fracture patients were admission from a nursing home, comorbidity, and frailty. All these predictors are associated with cognitively impaired patients. We have not analyzed patient-reported outcomes, and therefore have no information on how the hip fractures influenced the patients’ quality of life and how the patients performed who were not reoperated. Strengths and limitations The large number of patients in our study is an advantage and enabled us to analyze rare complications and causes of reoperation. One should, however, be careful to draw conclusions based on very small differences even if they reach statistical significance. One important limitation of the study is the accuracy of the surgeon’s assessment of cognitive function. An earlier study from the NHFR found that orthopedic surgeons identified cognitive impairment with a specificity of 90%, a sensitivity of 69%, positive predictive value of 78%, and negative predictive value of 84%, compared with information recorded in local hospital databases (Kristoffersen et al. 2019). The completeness of the reported reoperations has been found to be lower than the reporting of primary hip fracture operations in the NHFR when compared with the Norwegian Patient Register (Furnes et al. 2017). We have, however, no indication that the reporting of reoperations differs between the patient groups according to cognitive function. Accordingly, the hazard rate ratios in this study are probably reliable, but the crude number of reoperations may represent a best-case scenario and the actual number of reoperations may be higher. Follow-up time and mortality differed between the

Acta Orthopaedica 2020; 91 (2): 146–151

treatment groups. Many of the causes of reoperations, such as pain and loosening of the implant, may occur a long time after primary surgery. When comparing the treatment groups, one should therefore be aware that patients with cognitive impairment might die before the complications occur. Conclusion The results suggest that patients with cognitive impairment should be treated with the same surgical procedures as patients without cognitive impairment. However, hemiarthroplasty with uncemented stem and a posterior approach should probably be avoided in cognitively impaired patients due to the increased risk of periprosthetic fracture and dislocation. Supplementary data Figure 2 and Table 4 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1709712

MHK, JEG, and LBE planned the study. MHK wrote the manuscript. MHK and ED performed the statistical analyses. All authors contributed to the interpretation of the results, and improvement of the manuscript. The authors would like to thank all the Norwegian orthopedic surgeons who have faithfully reported their operations to the register. Acta thanks Johannes K M Fakler and Sebastian Mukka for help with peer review of this study.

Amodeo S, Mainland B J, Herrmann N, Shulman K I. The Times they are a-changin’: clock drawing and prediction of dementia. J Geriatr Psych Neur 2015; 28(2): 145-55. Benchimol E I, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, et al. The REporting of studies Conducted using Observational Routinelycollected health Data (RECORD) statement. PLoS Med 2015; 12(10): e1001885. Furnes O, Engesaeter L, Hallan G, Fjeldsgaard K, Gundersen T, Gjertsen J, et al. Annual Report, Norwegian Advisory Unit on Arthroplasty and Hip Fractures; 2017. ISBN: 978-82-91847-22-1 ISSN: 1893-8914 2017. Gjertsen J E, Engesaeter L B, Furnes O, Havelin L I, Steindal K, Vinje T, et al. The Norwegian Hip Fracture Register: experiences after the first 2 years and 15,576 reported operations. Acta Orthop 2008; 79(5): 583-93. Holsinger T, Deveau J, Boustani M, Williams J W, Jr. Does this patient have dementia? JAMA 2007; 297(21): 2391-404. Holvik K, Ranhoff A H, Martinsen M I, Solheim L F. Predictors of mortality in older hip fracture inpatients admitted to an orthogeriatric unit in Oslo, Norway. J Aging Health 2010; 22(8): 1114-31. Hugo J, Ganguli M. Dementia and cognitive impairment: epidemiology, diagnosis, and treatment. Clin Geriatr Med 2014; 30(3): 421-42. Jobory A, Kärrholm J, Overgaard S, Pedersen A B, Hallan G, Gjertsen J E, Mäkelä K, Rogmark C. Reduced revision risk for dual-mobility cup in total hip replacement due to hip fracture: a matched-pair analysis of 9,040 cases from the Nordic Arthroplasty Register Association (NARA). J Bone Joint Surg Am 2019; 101(14): 1278-85. Kristensen T, Dybvik E, Kristoffersen M, Dale H, Engesæter L B, Furnes O, Gjertsen J E. Cemented or uncemented hemiarthroplasty for femoral neck fracture? Data from the Norwegian Hip Fracture Register. Clin Orthop Relat Res 2020; 478(1): 90-100.

Acta Orthopaedica 2020; 91 (2): 146–151

Kristoffersen M H, Dybvik E, Steihaug O M, Bartz-Johannesen C A, Martinsen M I, Ranhoff A H, Gjertsen J E. Validation of orthopaedic surgeons’ assessment of cognitive function in patients with acute hip fracture. BMC Musculoskelet Disord 2019; 20(1): 268. Krogseth M, Wyller T B, Engedal K, Juliebo V. Delirium is an important predictor of incident dementia among elderly hip fracture patients. Dement Geriatr Cogn Disord 2011; 31(1): 63-70. Langslet E, Frihagen F, Opland V, Madsen J E, Nordsletten L, Figved W. Cemented versus uncemented hemiarthroplasty for displaced femoral neck fractures: 5-year followup of a randomized trial. Clin Orthop Relat Res 2014; 472(4): 1291-9. Mukka S, Knutsson B, Krupic F, Sayed-Noor A S. The influence of cognitive status on outcome and walking ability after hemiarthroplasty for femoral neck fracture: a prospective cohort study. Eur J Orthop Surg Traumatol 2017; 27: 653-8. Mundi S, Chaudhry H, Bhandari M. Systematic review on the inclusion of patients with cognitive impairment in hip fracture trials: a missed opportunity? Can J Surg 2014; 57(4): E141-5.

151

Naik M, Nygaard H A. Diagnosing dementia—ICD-10 not so bad after all: a comparison between dementia criteria according to DSM-IV and ICD-10. Int J Geriatr Psychiatry 2008; 23(3): 279-82. Petersen R C, Doody R, Kurz A, Mohs R C, Morris J C, Rabins P V, et al. Current concepts in mild cognitive impairment. Arch Neurol 2001; 58(12): 1985-92. Ranstam J, Robertsson O. The Cox model is better than the Fine and Gray model when estimating relative revision risks from arthroplasty register data. Acta Orthop 2017; 88(6): 578-80. Sheehan K J, Williamson L, Alexander J, Filliter C, Sobolev B, Guy P, et al. Prognostic factors of functional outcome after hip fracture surgery: a systematic review. Age Ageing 2018; 47(5): 661-70. Söderqvist A, Miedel R, Ponzer S, Tidermark J. The influence of cognitive function on outcome after a hip fracture. J Bone Joint Surg Am 2006; 88(10): 2115-23. Svenøy S, Westberg M, Figved W, Valland H, Brun O C, Wangen H, Madsen J E, Frihagen F. Posterior versus lateral approach for hemiarthroplasty after femoral neck fracture: Early complications in a prospective cohort of 583 patients. Injury 2017; 48(7): 1565-9.

152

Acta Orthopaedica 2020; 91 (2): 152–158

Perioperative, short-, and long-term mortality related to fixation in primary total hip arthroplasty: a study on 79,557 patients in the Norwegian Arthroplasty Register Håvard DALE 1,2, Sjur BØRSHEIM 3, Torbjørn Berge KRISTENSEN 1, Anne Marie FENSTAD 1, Jan-Erik GJERTSEN 1,2, Geir HALLAN 1,2, Stein Atle LIE 1,4, and Ove FURNES 1,2 1 The

Norwegian Arthroplasty Register, Department of Orthopaedic Surgery, Haukeland University Hospital, Bergen; 2 Department Clinical Medicine, University of Bergen, Bergen; 3 Department of Surgery, Voss Hospital, Voss; 4 Department of Clinical Dentistry, University of Bergen, Bergen, Norway Correspondence: havard.dale@helse-bergen.no Submitted 2019-04-17. Accepted 2019-10-28.

Background and purpose — There are reports on perioperative deaths in cemented total hip arthroplasty (THA), and THA revisions are associated with increased mortality. We compared perioperative (intraoperatively or within 3 days of surgery), short-term and long-term mortality after all-cemented, all-uncemented, reverse hybrid (cemented cup and uncemented stem), and hybrid (uncemented cup and cemented stem) THAs. Patients and methods — We studied THA patients in the Norwegian Arthroplasty Register from 2005 to 2018, and performed Kaplan–Meier and Cox survival analyses with time of death as end-point. Mortality was calculated for all patients, and in 3 defined risk groups: high-risk patients (age ≥ 75 years and ASA > 2), intermediate-risk patients (age ≥ 75 years or ASA > 2), low-risk patients (age < 75 years and ASA ≤ 2). We also calculated mortality in patients with THA due to a hip fracture, and in patients with commonly used, contemporary, well-documented THAs. Adjustement was made for age, sex, ASA class, indication, and year of surgery. Results — Among the 79,557 included primary THA patients, 11,693 (15%) died after 5.8 (0–14) years’ followup. Perioperative deaths were rare (30/105) and found in all fixation groups. Perioperative mortality after THA was 4/105 in low-risk patients, 34/105 in intermediate-risk patients, and 190/105 in high-risk patients. High-risk patients had 9 (CI 1.3–58) times adjusted risk of perioperative death compared with low-risk patients. All 4 modes of fixation had similar adjusted 3-day, 30-day, 90-day, 3–30 day, 30–90 day, 90-day–10-year, and 10-year mortality risk. Interpretation — Perioperative, short-term, and longterm mortality after primary THA were similar, regardless of fixation type. Perioperative deaths were rare and associated with age and comorbidity, and not type of fixation.

Perioperative (intra- or early postoperative) deaths have been reported in cemented total hip arthroplasty (THA) and hemiarthroplasty of the hip (Sierra et al. 2009, Talsnes et al. 2013, Garland et al. 2017). One reason for early mortality may be the so-called Bone Cement Implantation Syndrome (BCIS) (Donaldson et al. 2009, Olsen et al. 2014). The symptoms of BCIS are hypoxia, with or without hypotension, and/or unexpected loss of consciousness occurring at or shortly after the time of cementation, mostly in old patients with some comorbidity, and may be fatal (Olsen et al. 2014). Death is undisputedly an important adverse outcome. Thus, studying the superior mode of fixation in THA, one should also consider short-term and long-term mortality in addition to long-term revision rates. However, the complexity of several outcomes can make it difficult to conclude which fixation to choose for the individual patient scheduled for elective THA. We compared perioperative-, short-term, and long-term mortality for patients after primary, all-cemented, all-uncemented, reverse hybrid (cemented cup and uncemented stem), and hybrid (uncemented cup and cemented stem) THAs using the Norwegian Arthroplasty Register (NAR).

Patients and methods Since its inception in 1987, the NAR has registered information on primary THAs and THA revisions in Norway. Among the data collected are: the patient’s identity, date of operation, indication for THA, type of implants, method of fixation, intraoperative complications, and other surgery-related factors. In addition, patient-related factors like age, sex, and comorbidity are collected. The NAR uses the unique identification number of each Norwegian to link the primary THA to any subsequent

Acta Orthopaedica 2020; 91 (2): 152–158

153

Primary THA-patients in Norway 2005–2018 n = 81,715 Excluded Patients missing information on ASA class and indication for primary THA n = 2,158 THA-patients with complete information on covariates (n = 79,557): – cemented, 31,997 – uncemented, 21,553 – reverse hybrid, 23,052 – hybrid, 2,955

Low-risk THA-patients (n = 48,625): – cemented, 14,348 – uncemented, 16,739 – reverse hybrid, 15,915 – hybrid, 1,622

Intermediate-risk THA-patients (n = 23,565): – cemented, 12,960 – uncemented, 3,971 – reverse hybrid, 5,678 – hybrid, 956

Patients with THA due to acute or complication after hip fracture (n = 8,415 a ): – cemented, 4,339 – uncemented, 1,558 – reverse hybrid, 2,113 – hybrid, 355

High-risk THA-patients (n = 7,367): – cemented, 4,689 – uncemented, 842 – reverse hybrid, 1,459 – hybrid, 377

Patients with commonly used, contemporary, well documented THA (n = 58,777 a ): – cemented, 23,118 – uncemented, 13,847 – reverse hybrid, 20,125 – hybrid, 1,687

Figure 1. Flowchart of inclusion and exclusion of total hip arthroplasty (THA) patients. Patients in sub-groups are highlighted by green boxes. a Patients in these subgroups are also included in the 3 risk groups.

revisions, and to the National Population Register, which provides information on death or emigration. The surgeons fill in the register form immediately after surgery and mail it to the NAR, where the data are entered electronically (Havelin et al. 2000). The data are validated, with 97% completeness of reporting of primary THAs, 88% reporting of revisions, 100% coverage of Norwegian hospitals, and 100% reporting of deaths (Furnes et al. 2019). We assessed mortality after primary THAs registered with complete information on patient characteristics. The NAR has registered ASA class since 2005. Therefore, the period of inclusion and observation for the present study was from January 1, 2005 to December 31, 2018. Patients were included with their 1st primary THA only, if they had been subject to bilateral primary THA. From 2005 to 2018 the NAR contained data on 81,715 primary THA patients, 79,557 were eligible for analyses (Figure 1). Statistics We compared patients with cemented, uncemented, reverse hybrid or hybrid THAs by Kaplan–Meier (KM) survival analyses. Furthermore, we performed adjusted survival analyses using Cox regression models, adjusted for age, sex, ASA class, and indication for primary THA. Additionally, we adjusted for the year of primary surgery to minimize the effect of timedependent confounding. All patients in each fixation group were followed until date of death or emigration, or until December 31, 2018.

We ignored revisions in the analyses, and estimated adjusted hazard rate ratios, as a measure of relative risk, with 95% confidence intervals (CI), for modes of fixation and covariates. Primary outcomes were perioperative mortality (intraoperative death or death within 3 days of THA) and 10-year mortality. Secondary outcomes were 30-day, 90-day, 3–30day, 30–90-day, 90-day–10-year, and 10-year mortality. 3 risk groups were assessed: low-risk patients (age < 75 years and ASA ≤ 2), intermediate-risk patients (age ≥ 75 years or ASA > 2), high-risk patients (age ≥ 75 years and ASA > 2). In addition, we assessed 2 subgroups independently of the risk groups: (1) patients with THA due to acute or complications after hip fracture, an emerging indication for THA; (2) patients who received commonly used, contemporary, and well-documented THAs, comparable to a separate study on implant survival, since mortality is being used as a reason for choice of fixation principle (Dale et al. 2019). These two latter subgroups were also included in the risk groups. Since perioperative deaths were few, comparisons of risk factors were also performed by Fisher’s exact test with and without Bonferroni multiple comparison correction. We performed the analyses in concordance with the guidelines for statistical analyses of arthroplasty register data (Ranstam et al. 2011). The proportional hazard assumptions of the Cox survival analyses were largely fulfilled, when the smoothed Schoenfeld residuals were visually inspected (Figure 2). Competing risk models were not considered, since we ignored revision surgery. Bilaterality was not rel-

154

Acta Orthopaedica 2020; 91 (2): 152–158

Figure 2. The relationship between year postoperatively and the log relative risk (RR) of death for uncemented, reverse hybrid and hybrid THAs, compared with cemented THAs, with 95% confidence intervals. The horizontal green line shows the reference relative risk (RR = 1) of cemented THAs. It is adjusted for age, sex, ASA class, indication for primary THA, and year of primary THA in the analyses. The course is proportional if lines are parallel.

evant since patients were included only with their 1st primary THA. We considered non-overlapping 95% confidence intervals (CI) as statistically significant. The IBM SPSS 24.0 (IB Corp, Armonk, NY, USA) and R statistical software (R Centre for Statistical Computing, Vienna, Austria) packages were used for analyses. The study was performed in accordance with the RECORD and STROBE statement. Ethics, data sharing plan, funding, and potential conflicts of interests The registration of data and the study were performed confidentially on patient consent and according to Norwegian and EU data protection rules. Data may be accessible upon application to the NAR. The study was fully financed by the NAR, and no conflict of interest is declared.

Results During the study period, 11,693 (15%) of the 79,557 primary THA patients died. In general, patients who had cemented THA were older and with more comorbidity than those who had uncemented THA. Patients receiving reverse hybrid or hybrid THAs were intermediate groups concerning age and comorbidity. Patients with uncemented, reverse hybrid, and hybrid THAs had shorter median follow-up due to increased use of these fixation modes towards the end of the study period (Tables 1 and 2). Perioperative mortality Perioperative deaths were rare (30/105) and found in all modes of fixations. Adjusted 3-day mortality risk was similar after cemented, uncemented, reverse hybrid, and hybrid THAs (Table 3, see Supplementary data). The mortality on the day of surgery was 10/105.

THA patients with ASA 3 (few patients with ASA 4), age over 75 years, or THA in the course of a hip fracture had a higher risk of perioperative death after THA (Table 4). When stratified into risk groups, perioperative mortality after THA was 4/105 in low-risk patients, 34/105 in intermediate-risk patients, and 190/105 in high-risk patients. High-risk patients had nearly 9 times the risk of adjusted perioperative death after primary THA compared with low-risk patients (Table 4). We found no statistically significant difference in risk of perioperative death between the 4 modes of fixation, in either of the 3 risk groups or 2 subgroups (Table 3, see Supplementary data). That was also the finding when assessing perioperative death in the 4 fixation groups by Fisher’s exact test with Bonferroni multiple comparison correction. In Fisher’s exact test without correction, and not adjusted for patient characteristics, uncemented THA had lower perioperative mortality, compared with cemented (p = 0.03). The 24 patients who died were older and more comorbid (Table 5, see Supplementary data). Short- and long-term mortality The 10-year mortality risk and adjusted mortality was similar regardless of fixation (Table 3, Figure 3). In addition, we found similar results for all 4 fixation groups concerning 30-day, 90-day, 3–30-day, 30–90-day, 90-day–10-year, and 10-year mortality, and short- and long-term mortality did not change throughout the study period. This was also true for the finding in the “best-case” group of commonly used, contemporary, well-documented THAs.

Discussion Both the 3-day and the 10-year mortality were similar after allcemented, all-uncemented, reverse hybrid, and hybrid THA. Risk factors for perioperative death were patient-related and

Acta Orthopaedica 2020; 91 (2): 152–158

155

Table 1. Demographic risk factors by type of fixation Risk factors

Number of Dead Cemented Uncemented Reverse Hybrid THA patients patients (%) THA (%) THA (%) hybrid THA (%) THA (%) n = 79,557 n = 11,693 a n = 31,997 n = 21,553 n = 23,052 n = 2,955

Sex Male Female Age < 45 45–54 55–64 65–74 75–84 ≥ 85 ASA class 1 2 3 4 Indication for primary THA Osteoarthritis Inflammatory hip disease Acute hip fracture Complication after hip fracture childhood hip disease Osteonecrosis of the femoral head Other diagnosis a Dead

28,512 51,052

4,450 (16) 7,243 (14)

31 69

42 58

38 62

36 64

2,843 6,874 18,876 28,235 19,156 3,573

57 (2) 209 (3) 1,122 (6) 3,277 (12) 5,283 (28) 1,745 (49)

1 2 14 39 37 7

8 17 32 30 11 1

3 10 30 36 17 3

2 7 22 33 30 6

16,063 47,899 15,247 348

1,281 (8) 5,761 (12) 4,465 (29) 195 (56)

15 59 25 1

27 59 17 0.3

21 62 17 0.3

15 63 21 1

59,285 1,722 3,187

8,015 (14) 269 (16) 661 (21)

76 2 5

71 2 2

77 2 4

67 2 5

5,208 7,266

1,735 (33) 394 (5)

8 4

5 16

5 8

7 15

2,268 621

442 (19) 177 (29)

3 1

overall 15%

Table 2. Main characteristics of the included THA patients by types of fixation Reverse All Cemented Uncemented hybrid n = 79,557 n = 31,997 n = 21,553 n = 23,052 Dead at 10 years, n (%) Mean follow-up (range), years Median follow-up (IQR), years Mean age (range) Mean ASA-class At risk after 10 years

Hybrid n = 2,955

10,330 (13) 6,843 (21) 1,260 (6) 2,018 (9) 209 (7) 5.8 (0–14) 6.9 (0–14) 4.9 (0–14) 5.5 (0–14) 3.8 (0–14) 5.4 (2.6–8.8) 7.0 (3.4–10) 4.2 (1.8–7.6) 5.2 (2.9–7.7) 2.6 (1.1–4.9) 69 (11–97) 73 (22–98) 62 (11–95) 66 (14–97) 70 (17–97) 2.0 2.1 1.9 2.0 2.1 13,924 8,811 2,387 2,402 324

IQR = Inter quartile range

not related to mode of fixation. This was also the finding when assessing high-risk patients only. Traditionally, uncemented THAs have had lower implant survival than cemented THAs in register studies (Hailer et al. 2010, Mäkelä et al. 2014). This was also the case in the revision study based on most of the patients included in the present study (Dale et al. 2019). Nevertheless, the use of uncemented fixation in THA is increasing, even in old and frail patients (Troelsen et al. 2013, Mäkelä et al. 2014, Furnes et al. 2019). 1 of the explanations for this trend may be the fear of sudden perioperative deaths because of 3rd generation cementing technique (pressure cementation) and BCIS. This is a well-known complication associated with the use of bone

cement, but the incidence is low in primary THA (Donaldson et al. 2009, Sierra et al. 2009). BCIS have also been highlighted in papers on cemented hemiarthroplasty used in old and frail femoral neck fracture patients (Talsnes et al. 2013, Olsen et al. 2014, Rutter et al. 2014). Acknowledging the existence of BCIS, and that modern intensive care medicine can successfully treat the acute and potentially fatal BCIS, we considered deaths within 3 days postoperatively as potentially associated with the cementation, although other causes of death were also possible as a result of the surgical trauma. Information on the cause of death was not available in our patients. Sierra et al. (2009) reported 3/10,000 perioperative

156

Acta Orthopaedica 2020; 91 (2): 152–158

Table 4. Risk factors of perioperative death. Adjusted for age, sex, ASA class, indication for primary THA, and year of primary surgery Deaths 3-day relative within mortality Risk factor THAs 3 days risk (CI) Sex Female Male ASA class 1 2 3 4 Age group < 45 years 45–54 55–64 65–74 75–84 ≥ 85 Indication for primary THA Osteoarthritis Inflammatory hip disease Acute hip fractures Complication after hip fracture childhood hip disease Osteonecrosis of the femoral head Other diagnosis Risk class a Low Intermediate High

51,045 28,512

18 6

1 0.7 (0.3–1.9)

16,063 47,899 15,247 348

0 8 15 1

1 3.2 (1.3–7.7) 5.2 (0.6–45)

2,843 6,874 18,876 28,235 19,156 3,573

0 0 1 3 12 8

0.6 (0.1–5.6) 1 4.4 (1.2–16) 9.4 (2.4–37)

59,285 1,722 3,187

10 0 5

9.4 (3.1–28)

5,208 7,266

6 1

3.1 (1.1–8.7) 2.8 (0.4–22)

2,268 621

1 1

2.4 (0.3–19) 7.2 (0.9–59)

48,625 23,565 7,353

2 8 14

1 2.7 (0.4–16) 8.7 (1.3–58)

Figure 3. Patient survival curve for the 4 modes of fixation adjusted for age, sex, ASA class, indication for primary THA, and year of primary surgery.

a Adjusted

for age, sex, indication for primary THA, and year of primary surgery. Low = age < 75 years and ASA ≤ 2 Intermediate = age ≥ 75 years or ASA > 2 High = age ≥ 75 years and ASA > 2

deaths after cemented THA on the day of surgery, which was higher than our finding (1/10,000). Pripp et al. (2014) found that only half of the deaths on the day of cemented arthroplasty might be attributed to BCIS. A study from the Finnish Hospital Discharge register found similar 2-day mortality for patients after cemented, uncemented, and hybrid THA (Ekman et al. 2019). A Swedish study reported a small, but statistically significant increased relative risk of death in patients during the first 14 days after cemented THA, compared with uncemented THA (Garland et al. 2017). We did not find such a difference, even in the high-risk subgroup or in the group of patients with THA in the course of a hip fracture. However, we adjusted for comorbidity by ASA class, whereas Garland et al. adjusted by modified Charlson Comorbidity Index, which may have slightly different effect (Lavelle et al. 2015). The potential risk of perioperative death needs to be weighed against other outcomes, such as the risk of revision due to fracture, dislocation, and infection. Revision risk due to these causes is higher after uncemented THAs (Hailer et al. 2010, Stea et al. 2014, Dale et al. 2019). Such complications

may implicate a poorer functional outcome, increased morbidity, and long-term mortality (Lindahl et al. 2007, Gundtoft et al. 2017, Cnudde et al. 2019). In the subgroup of commonly used, contemporary, welldocumented THAs, the differences in risk of revision between the 4 modes of fixation, presented in a separate paper, were quite small (Dale et al. 2019). The exception was that uncemented THA in females over 55 years of age had a higher risk of revision due to aseptic loosening, periprosthetic fracture, and dislocation. Patients may have increased long-term mortality after such revisions (Gundtoft et al. 2017, Cnudde et al. 2019). 3-day and 10-year mortality was similar for all fixations in our study, indicating that perioperative-, short-term, or long-term mortality risk should not dictate what fixation principle to choose in primary THA, even in high-risk patients. Mortality is low in healthy old patients elected for THA (Lie et al. 2000, Jamsen et al. 2013). We found a 10-year mortality of 15–16% for all modes of fixation, and 7–8% in the low-risk class. However, Lie et al. (2010) found an excess mortality of 0.12% in the first 26 days after primary THA compared with the baseline mortality for these patients. From 70 days and onwards THA patients have been reported to have lower mortality compared with the baseline population (Lie et al. 2002). Hunt et al. (2013) reported a yearly decreasing 90-days mortality after contemporary THA in the National Joint Registry for England and Wales between 2003 and 2011, and similar results concerning 90-days mortality after all 4 modes of fixation. The decreased mortality was attributed to improved perioperative prophylaxis and treatment. We did not find a similar decrease in mortality, but several of the modifiable factors suggested, such as chemical thromboembolic prophylaxis and spinal anesthesia, were used in the vast majority of THAs in Norway during the study period. Mortality from 30 days after primary THA, regardless of mode of fixation, has been found to be lower than for controls (matched popula-

Acta Orthopaedica 2020; 91 (2): 152–158

tion without THA) (Pedersen et al. 2011, Garland et al. 2017). Either patients chosen for elective THA are healthier than the average population, or one may say that successful THA may protect the patients and result in lower mortality than the corresponding population (Cnudde et al. 2018). Probably very frail patients are not selected for elective THA. Strengths and limitations Our study is based on validated data from the NAR, with high completeness of reporting of primary THAs and deaths (Furnes et al. 2019). We therefore had the benefit of complete information on patient-related risk factors. Accordingly, we were able to adjust for important differences between the patient groups. Risk factors we included in the analyses such as age, ASA class, and indication for primary THA are associated with adverse outcomes and mortality after surgery (Belmont et al. 2014, Cnudde et al. 2018). The effect of adjustment for such risk factors on 10-year KM survival after cemented THA illustrates the importance including age and comorbidities in assessments of mortality. Perioperative deaths are rare, and it may only be possible to study these in large databases such as arthroplasty registers. We included a large number of THAs with exact survival times of the patients. Since results are from a nationwide THA population, external validity should be good. Register studies have inherent limitations (Varnum et al. 2019). A limitation in our study was the limited information on the cause of perioperative death. We therefore considered deaths within 3 days postoperatively as potentially associated with the cementation. This was an approximation, and Pripp et al. (2014) found that only half of perioperative mortalities are attributed to BCIS. This indicates that the number of perioperative deaths caused by BCIS in our study would be even lower for THAs involving bone cement. In Finland, where uncemented THA is more common, even in old and frail patients, similar perioperative mortality has been reported after cemented and uncemented THAs (Ekman et al. 2019). The follow-up is relatively short and different for the 4 fixation groups, since there has been a shift towards more use of uncemented, reverse hybrid and hybrid THAs (see Table 2) (Furnes et al. 2019). We did not find differences in perioperative, short-, or long-term mortality. Limited follow-up should therefore have a small effect, if any, on the results. Conclusion Results after contemporary primary THA may be good regardless of fixation, but uncemented THA had an increased risk of revision in a cohort of patients also included in the present study (Dale et al. 2019). Perioperative deaths, however, were associated only with patient-related risk factors, like age and comorbidity, and not surgery-related factors like fixation. Sudden death will be the most serious adverse event, but death during a course of complications is equally serious. We found that perioperative, short-term, and long-term mortality after

157

primary THA were similar, regardless of fixation mode. Use of bone cement appears to be safe in all patient groups. Perioperative deaths were associated with advanced age and comorbidity, and not type of fixation, and should therefore not guide the choice of fixation in primary THA, even in high-risk patients. Supplementary data Tables 3 and 5 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674. 2019.1701312

The authors would like to thank all Norwegian surgeons for conscientiously reporting THAs to the NAR, and the secretaries, IT analyst, and statisticians at the NAR for entering the data and preparing them for analyses. HD wrote the manuscript. HD, SAL, and AMF analyzed the data. All authors contributed to the design of the study, critical evaluation of the data and analyses, interpretation of the findings, and critical revision of the manuscript, through all stages of the study. Acta thanks Ross W Crawford and Gary Hooper for help with peer review of this study.

Belmont P J, Jr, Goodman G P, Waterman B R, Bader J O, Schoenfeld A J. Thirty-day postoperative complications and mortality following total knee arthroplasty: incidence and risk factors among a national sample of 15,321 patients. J Bone Joint Surg Am 2014; 96 (1): 20-6. Cnudde P, Rolfson O, Timperley A J, Garland A, Karrholm J, Garellick G, Nemes S. Do patients live longer after THA and is the relative survival diagnosis-specific? Clin Orthop Relat Res 2018; 476 (6): 1166-75. Cnudde P, Bulow E, Nemes S, Tyson Y, Mohaddes M, Rolfson O. Association between patient survival following reoperation after total hip replacement and the reason for reoperation: an analysis of 9,926 patients in the Swedish Hip Arthroplasty Register. Acta Orthop 2019; 90 (3): 226–30. Dale H, Børsheim S, Kristensen T B, Fenstad A M, Gjertsen J E, Hallan G, Lie S A, Furnes O. Fixation, sex, and age: highest risk of revision for uncemented stems in elderly women—data from 66,995 primary total hip arthroplasties in the Norwegian Arthroplasty Register. Acta Orthop 2019; 10.1080/17453674.2019.1682851 (Epub ahead of print). Donaldson A J, Thomson H E, Harper N J, Kenny N W. Bone cement implantation syndrome. Br J Anaesth 2009; 102 (1): 12-22. Ekman E, Palomaki A, Laaksonen I, Peltola M, Hakkinen U, Makela K. Early postoperative mortality similar between cemented and uncemented hip arthroplasty: a register study based on Finnish national data. Acta Orthop 2019; 90 (1): 6-10. Furnes O, Hallan G, Gjertsen J E, Visnes H, Gundersen T, Fenstad A M, Dybvik E, Kroken G. The Norwegian Arthroplasty Register, Annual Report; 2019. http://nrlweb.ihelse.net/eng/Rapporter/Report2019_english. pdf. Garland A, Gordon M, Garellick G, Karrholm J, Skoldenberg O, Hailer N P. Risk of early mortality after cemented compared with cementless total hip arthroplasty: a nationwide matched cohort study. Bone Joint J 2017; 99-b (1): 37-43. Gundtoft P H, Pedersen A B, Varnum C, Overgaard S. Increased mortality after prosthetic joint infection in primary THA. Clin Orthop Relat Res 2017; 475 (11): 2623-31. Hailer N P, Garellick G, Kärrholm J. Uncemented and cemented primary total hip arthroplasty in the Swedish Hip Arthroplasty Register. Acta Orthop 2010; 81 (1): 34-41.

158

Havelin L I, Engesæter L B, Espehaug B, Furnes O, Lie S A, Vollset S E. The Norwegian Arthroplasty Register: 11 years and 73,000 arthroplasties. Acta Orthop Scand 2000; 71 (4): 337-53. Hunt L P, Ben-Shlomo Y, Clark E M, Dieppe P, Judge A, MacGregor A J, Tobias J H, Vernon K, Blom A W. 90-day mortality after 409,096 total hip replacements for osteoarthritis, from the National Joint Registry for England and Wales: a retrospective analysis. Lancet 2013; 382 (9898): 1097-104. Jamsen E, Puolakka T, Eskelinen A, Jantti P, Kalliovalkama J, Nieminen J, Valvanne J. Predictors of mortality following primary hip and knee replacement in the aged: a single-center analysis of 1,998 primary hip and knee replacements for primary osteoarthritis. Acta Orthop 2013; 84 (1): 44-53. Lavelle E A, Cheney R, Lavelle W F. Mortality prediction in a vertebral compression fracture population: the ASA Physical Status Score versus the Charlson Comorbidity Index. Int J Spine Surg 2015; 9: 63. Lie S A, Engesæter L B, Havelin L I, Gjessing H K, Vollset S E. Mortality after total hip replacement: 0–10-year follow-up of 39,543 patients in the Norwegian Arthroplasty Register. Acta Orthop Scand 2000; 71 (1): 19-27. Lie S A, Engesæter L B, Havelin L I, Furnes O, Vollset S E. Early postoperative mortality after 67,548 total hip replacements: causes of death and thromboprophylaxis in 68 hospitals in Norway from 1987 to 1999. Acta Orthop Scand 2002; 73 (4): 392-9. Lie S A, Pratt N, Ryan P, Engesæter L B, Havelin L I, Furnes O, Graves S. Duration of the increase in early postoperative mortality after elective hip and knee replacement. J Bone Joint Surg Am 2010; 92 (1): 58-63. Lindahl H, Oden A, Garellick G, Malchau H. The excess mortality due to periprosthetic femur fracture: a study from the Swedish national hip arthroplasty register. Bone 2007; 40 (5): 1294-8. Mäkelä K T, Matilainen M, Pulkkinen P, Fenstad A M, Havelin L, Engesæter L, Furnes O, Pedersen A B, Overgaard S, Kärrholm J, Malchau H, Garellick G, Ranstam J, Eskelinen A. Failure rate of cemented and uncemented total hip replacements: register study of combined Nordic database of four nations. BMJ 2014; 348: f7592. Olsen F, Kotyra M, Houltz E, Ricksten S E. Bone cement implantation syn-

Acta Orthopaedica 2020; 91 (2): 152–158

drome in cemented hemiarthroplasty for femoral neck fracture: incidence, risk factors, and effect on outcome. Br J Anaesth 2014; 113 (5): 800-6. Pedersen A B, Baron J A, Overgaard S, Johnsen S P. Short- and long-term mortality following primary total hip replacement for osteoarthritis: a Danish nationwide epidemiological study. J Bone Joint Surg Br 2011; 93 (2): 172-7. Pripp A H, Talsnes O, Reikeras O, Engesaeter L B, Dahl O E. The proportion of perioperative mortalities attributed to cemented implantation in hip fracture patients treated by hemiarthroplasty. Hip Int 2014; 24 (4): 363-8. Ranstam J, Kärrholm J, Pulkkinen P, Mäkelä K, Espehaug B, Pedersen A B, Mehnert F, Furnes O. Statistical analysis of arthroplasty data, II: Guidelines. Acta Orthop 2011; 82 (3): 258-67. Rutter P D, Panesar S S, Darzi A, Donaldson L J. What is the risk of death or severe harm due to bone cement implantation syndrome among patients undergoing hip hemiarthroplasty for fractured neck of femur? A patient safety surveillance study. BMJ Open 2014; 4 (6): e004853. Sierra R J, Timperley J A, Gie G A. Contemporary cementing technique and mortality during and after Exeter total hip arthroplasty. J Arthroplasty 2009; 24 (3): 325-32. Stea S, Comfort T, Sedrakyan A, Havelin L, Marinelli M, Barber T, Paxton E, Banerjee S, Isaacs A J, Graves S. Multinational comprehensive evaluation of the fixation method used in hip replacement: interaction with age in context. J Bone Joint Surg Am 2014; 96(Suppl 1): 42-51. Talsnes O, Vinje T, Gjertsen J E, Dahl O E, Engesaeter L B, Baste V, Pripp A H, Reikeras O. Perioperative mortality in hip fracture patients treated with cemented and uncemented hemiprosthesis: a register study of 11,210 patients. Int Orthop 2013; 37 (6): 1135-40. Troelsen A, Malchau E, Sillesen N, Malchau H. A review of current fixation use and registry outcomes in total hip arthroplasty: the uncemented paradox. Clin Orthop Relat Res 2013; 471 (7): 2052-9. Varnum C, Pedersen A B, Gundtoft P H, Overgaard S. The what, when and how of orthopaedic registers: an introduction into register-based research. EFORT Open Reviews 2019; 4 (6): 337-43.

Acta Orthopaedica 2020; 91 (2): 159–164

159

Early stabilization of the uncemented Symax hip stem in a 2-year RSA study Dennis S M G KRUIJNTJENS 1, Lennard KOSTER 2, Bart L KAPTEIN 2, Liesbeth M C JUTTEN 1, Jacobus J ARTS 1, and René H M TEN BROEKE 1 1 Department

of Orthopaedic Surgery, Research School Caphri, Maastricht University Medical Centre, Maastricht; 2 Department of Orthopaedic Surgery, RSAcore, Leiden University Medical Centre, Leiden, the Netherlands Correspondence: d.kruijntjens@gmail.com Submitted 2018-10-12. Accepted 2019-11-18.

Background and purpose — The uncemented Symax hip stem has shown early proximal ingrowth as result of the BONIT-hydroxyapatite (HA) coating and the distal DOTIZE surface treatment. We evaluated 2-year postoperative radiostereometric analysis (RSA) migration of the Symax hip stem in THA patients. We also investigated the correlation between migration at 4 weeks and clinical outcomes after 2 years. Patients and methods — Patients in a 2-year clinical follow-up single-centre RSA randomized controlled trial were randomized to 2 different cup designs. All 45 patients received a Symax hip stem. RSA migration patterns of the Symax hip stem is presented here as a single cohort. RSA examinations were performed postoperatively, but before weight-bearing, and subsequently after 1, 3, 6, 12, and 24 months. Clinical outcomes and radiographic evaluations were assessed 3, 6, 12, and 24 months postoperatively. Results — During the first 4 weeks, the Symax hip stem subsided, rotated into retroversion, and translated posteriorly, after which the migration ceased and the prosthesis stabilized. All clinical outcomes improved from preoperatively to 2 years. There was no clinically or statistically significant correlation between subsidence and retroversion at 4 weeks and clinical outcomes after 2 years. Interpretation — RSA evaluation of the uncemented Symax hip stem confirms that the design principles and coating properties lead to early stabilization of the stem, as early as 4 weeks postoperatively. There was no correlation between subsidence and retroversion at 4 weeks and clinical outcomes after 2 years. Based on the predictive potential of the RSA technique, we anticipate excellent long-term survival of this hip stem.

The uncemented Symax hip stem was developed as an optimization of the uncemented Omnifit hip stem (Stryker Orthopaedics, Kalamazoo, MI, USA) (Capello et al. 2009). The design considers the geometry of the stem, surface texture, and type and extent of the osseointegrative coating. Previous studies with histological and histomorphometric analyses on retrieved Symax hip stems have proven early bone ingrowth exclusively into the proximal part of the stem, as a result of the BONIT-hydroxyapatite (HA) coating (DOT GmbH, Rostock, Germany). A 2-year follow-up dual-energy X-ray absorptiometry (DEXA) study showed improved bone remodeling with the Symax hip stem compared with the Omnifit hip stem (ten Broeke et al. 2011, 2012, 2014). The current study is part of an RCT RSA study, assessing migration characteristics of 2 different uncemented cup designs (Trident HA and Trident Tritanium respectively; Stryker Orthopaedics, Kalamazoo, MI, USA) and the Symax hip stem with radiostereometric analysis (RSA). The focus of this study is on the migration pattern of the Symax stem in a single cohort, seen in the light of its specific optimized geometrical features, and in the light of the existing knowledge on mid-term outcome of this stem design. Several studies have shown that RSA can predict long-term prosthetic loosening based on 2-year postoperative follow-up data (Kärrholm et al. 1994, Valstar et al. 2005, Nieuwenhuijse et al. 2012). We hypothesized that the design of the Symax hip stem leads to early stabilization, by 3 months postoperatively. The primary objective of this study was to evaluate early postoperative migration of the Symax hip stem with the use of RSA. The secondary objective was to investigate whether there is a correlation between migration of the stem at 4 weeks and clinical outcomes after 2 years.

160

Figure 1. Design features of Symax hip stem, illustrating the anatomically anteverted proximal geometry, with the BONIT-HA coating; and the straight distal part with the DOTIZE surface treatment and a posterior chamfer.

Patients and methods This single-center RSA study was performed at the Maastricht University Medical Centre (MUMC), the Netherlands. It is a sub-study of an RCT RSA study comparing the uncemented Trident HA cup and the Trident Tritanium cup designs, presenting the migration results of the Symax hip stem, used in both groups, as a single cohort. The results of the acetabular cup migration will be reported in a separate manuscript. Enrolment took place between December 2011 and January 2015. Patients scheduled for uncemented primary total hip arthroplasty (THA), aged between 18 and 70 years, with BMI less or equal to 35 were eligible for this study. Exclusion criteria were bilateral hip complaints, impaired cognitive function, and use of medication or illness influencing bone metabolism (e.g., corticosteroids, bisphosphonates, osteoporosis, and metastasis). Surgical protocol The posterolateral approach was used by 2 senior hip surgeons (RtB and JG), with transosseous reattachment of capsule and external rotators in all patients. Patients received 24-hour intravenous antibiotic prophylaxis (cefuroxime), deep venous thrombosis prophylaxis with low molecular weight heparins (nadroparin) for 6 weeks, and prevention of heterotopic bone formation with NSAIDs (indometacin) for 2 weeks. Full weight-bearing under the supervision of a physiotherapist was allowed from the first postoperative day, but only after the baseline RSA imaging was completed. Implant The Symax hip stem is an uncemented design forged from Ti6Al4V alloy (CE 545074). The Symax design is an optimization of the uncemented Omnifit hip stem (Figure 1). For further details on geometry, surface treatment and coating characteristics, please refer to Szmukler-Moncler et al. (2001), Becker et al. (2004), ten Broeke et al. (2011).

Acta Orthopaedica 2020; 91 (2): 159â&#x20AC;&#x201C;164

Figure 2. The Elementary Geometrical Shapes (EGS) model used to determine the migration of the Symax hip stem.

Implanted cups were either the uncemented Trident HA (CE514410) or the uncemented Trident Tritanium (CE526088) acetabular cup, both in combination with an ultra-high molecular weight polyethylene X-3 insert (CE509982, Stryker Orthopaedics, Kalamazoo, MI, USA). RSA examination Baseline RSA examination was made 1 day after surgery, prior to loading of the operated hip. Follow-up RSA examination was made postoperatively at 1, 3, 6, 12, and 24 months. During surgery 6â&#x20AC;&#x201C;8 tantalum markers (0.8 mm diameter) were inserted in the bone surrounding the hip stem. These markers form a rigid body that is the basis for RSA calculations. The rigid body consisted of at least 3 markers with a mean error (ME) of rigid body fitting below 0.35 mm and Condition Number (CN) below 120 (ISO 16087, 2013). All RSA examinations were acquired with the patient in a supine position over a uniplanar calibration box (Medis Carbon Box nr. 013, Medis Specials bv, the Netherlands). Migration of the Symax stems was calculated using the Elementary Geometrical Shapes (EGS) hip model from Model-based RSA software (Version 4.1; RSAcore, Department of Orthopaedic Surgery, LUMC, the Netherlands) (Figure 2) as described by Kaptein et al. (2006). Translations and rotations were calculated using a coordinate system with its origin in the center of the 3D model in the baseline evaluation, and the X- and Y-axis parallel to the X- and Y-axis of the calibration box (Figure 2). Aligning the patient with the coordinate system of the calibration box allows for description of the migrations of prosthesis components using anatomical directions. Migration results from patients with a left-sided prosthesis were recalculated following the conventions as presented in the ISO standard guidelines (ISO 16087, 2013) in order to describe the migration in anatomical terms for a right-sided prosthesis. Migrations are reported as migration relative to baseline. To determine the precision of the RSA set-up, a double set of RSA examinations during the same follow-up was

Acta Orthopaedica 2020; 91 (2): 159–164

161

Table 1. Baseline characteristics of included patients Factor

Female Male (n = 22) (n = 23)

Table 2. Number of patients per follow-up moment Total (n = 45)

Mean age (range) 60 (44–70) 59 (30–70) 60 (30–70) BMI (SD) 27 (4) 27 (3) 27 (3) Osteoarthritis / avascular necrosis 22 / 0 22 / 1 44 / 1

acquired. Actual migration within the short time interval between the double examinations is expected to be 0, therefore calculated migration between these double examinations represents the measurement error. The mean error value represents the system bias, while the standard deviation (SD) is a measure for the precision of the measurements (Kaptein et al. 2006). Clinical evaluation Clinical evaluations were performed preoperatively and postoperatively at 3, 6, 12, and 24 months. The evaluated clinical outcome parameters were the Harris Hip Score (HHS), the Oxford Hip Score (OHS), the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), and the EuroQol-5D (EQ-5D). The EQ-5D index was calculated based on a Dutch value set, representative of the Dutch population with regard to age and sex. To facilitate further analysis the index was linearly rescaled on a scale of 0 to 100, with 0 the worst possible score and 100 the best possible score. Additionally, the EQ-5D VAS score was evaluated. Complications and adverse events were recorded during follow-up. Radiographic evaluation Standing anteroposterior radiographs of the pelvis and axial radiographs of the operated hips were evaluated immediately postoperatively and at 3, 6, 12, and 24 months. Radiographs were evaluated as per “adapted Gruen zones” for signs of osseointegration, such as cancellous condensation (“spotweld formation”) and formation of distal reactive lines. In the “adapted Gruen zones”, zones 1 and 7 represent the coated areas, and zones 2–3 and 5–6 respectively the lateral and medial zones, equally divided around the non-coated part of the stem (ten Broeke et al. 2012). Statistics Descriptive statistics of continuous variables were reported as means (SD) or (range) in case of skewness. Categorical variables are presented as frequencies. A paired samples t-test was used to compare migration between different follow-up moments. The Wilcoxon signed-ranks test was used as the non-parametric test to compare clinical outcome scores over time. Spearman’s rho correlation coefficients were determined to correlate Y-translation and Y-rotation at 4 weeks to clinical outcomes after 2 years. IBM SPSS for Windows 24.0.1 (IBM Corp, Armonk, NY, USA) was used for statistical data

Evaluation Baseline 4 weeks 3 months 6 months 1 year 2 years RSA analysis 35 33 32 32 31 31 Clinical 45 – 45 45 45 44 Radiographic 45 – 45 45 45 44

analysis, and p-values of < 0.05 were considered to indicate statistical significance. All results are reported with 95% confidence intervals (CI), if applicable, and clinical relevance will be discussed. Ethics, registration, funding, and potential conflicts of interests Ethical approval was obtained from the local Institutional Review Board (MEC10-1-068 NL 33832.068.10). The study was registered in the ClinicalTrial.gov database (NCT01618084). Informed consent was obtained from each patient prior to surgery. The study was conducted according to the ethical standards of the Declaration of Helsinki of 1975, as revised in 2013 in Fortaleza (Brazil), and following Good Clinical Practice (GCP) and ISO 14155 guidelines. No conflict of interest was declared and no personal funding was received. Research grant was received from Stryker Orthopaedics (Kalamazoo, MI, USA).

Results 45 patients (22 female) were included, with a mean age of 59 years (30–70), and all patients were operated because of osteoarthritis, except 1 who was operated for avascular necrosis (Table 1). RSA evaluations Due to not meeting the rigid body criteria, RSA results could not be calculated for 10 of the 45 patients. During follow-up several RSA acquisitions could not be used for RSA calculations, due to technical issues or over-projection of the bone markers by the hip stem. The number of the available migration calculations at each follow-up moment is denoted in Table 2. Mean migration of the 30 available RSA double examinations was –0.02 mm (CI –0.29 to 0.25) for X-translation, –0.01 mm (CI –0.25 to 0.23) for Y-translation, and –0.04 mm (CI –0.65 to 0.57) for Z-translation; –0.09˚ (CI –0.77 to 0.58) for X-rotation, 0.26˚ (CI –1.23 to 1.75) for Y-rotation, and –0.01˚ (CI –0.19 to 0.17) for Z-rotation. These results confirm that there is no bias in the RSA set-up. At 4 weeks mean Y-translation was –1.0 mm (CI –3.4 to 1.4), mean Y-rotation (retroversion) was 2.4˚ (CI –2.2 to 7.0), and mean Z-translation (posterior translation) was –0.4 mm

162

Acta Orthopaedica 2020; 91 (2): 159–164

Mean (SD) rotations, degree

Mean (SD) translations, mm

8 1.2

X-translation Y-translation Z-translation

0.8

X-rotation Y-rotation Z-rotation

0.4

0.0

–0.4

0 –0.8

–2

–1.2 0

Follow-up (months)

Figure 3a. Mean translation results for the Symax hip stem.

(CI –1.7 to 0.9). Migrations in the other directions were minor during follow-up (Figure 3). Subsidence was large (5.8 mm) initially in 1 patient, probably because of undersizing the stem; however, the position of this stem also stabilized after 4 weeks. This patient and the 13 patients who could not be evaluated with RSA after 2 years showed excellent clinical outcome scores at the time of their last RSA measurement as well as after 2 years. Clinical results All clinical outcomes improved clinically and statistically significantly from preoperative to 2 years (Table 3, see Supplementary data). Correlation RSA and clinical outcomes There were no clinically or statistically significant correlations between Y-translation or Y-rotation at 4 weeks and any clinical outcomes after 2 years (Table 4, see Supplementary data). Radiographic evaluations Spotweld formation was seen increasingly over time. After 2 years 35 patients showed spotweld formation in Gruen zone 2 (Figure 4, see Supplementary data). Distal reactive line formation was very rare, as it was only seen after 2 years in 3 patients in the medial and distal Gruen regions 4 till 6. Survival All stems survived the complete follow-up. Adverse events 1 patient experienced recurrent dislocations within 4 weeks after placement of the prosthesis, for which revision of the cup was necessary. This patient was not excluded from the study, since the stem was not revised. For the final follow-up at 2 years 1 patient was excluded, because she refused to participate for personal reasons.

Follow-up (months)

Figure 3b. Mean rotation results for the Symax hip stem.

Discussion The primary aim of this study was to evaluate migration of the Symax hip stem up to 2 years postoperatively. At 4 weeks postoperatively mean subsidence was 1.0 mm, mean retroversion was 2.4˚, and mean posterior translation was –0.4 mm. After 4 weeks the movement ceased and the prosthesis stabilized. These findings are in line with the design rationale and the coating properties of the Symax hip stem, in which early stabilization of the stem was expected to occur. Furthermore, these findings do confirm the results of the histological and histomorphometric analyses by ten Broeke et al. (2011) in which early bone ingrowth to the proximal part of the stem was shown. Our secondary aim was to correlate Y-translation and Y-rotation at 4 weeks to clinical outcomes at 2 years. However, no clinically or statistically significant correlations were observed (all correlation coefficients were between –0.30 and 0.30). For example, the outlier patient, who showed 5.8 mm subsidence at 4 weeks, had excellent clinical outcomes for all clinical outcome parameters at 2 years. This suggests that stabilization of the stem is more important than the absolute value of migration. Buratti et al. (2009) showed early stabilization of 85 Symax hip stems in a multicenter EBRA–FCA (Einzel-Bild-RöntgenAnalyse–femoral component analysis) study. Although subsidence increased slightly from 0.17 mm at 6 months to 0.45 mm at 2 years in their study, the threshold migration value used to define stability of a stem was 1.5 mm at 2 years. However, the accuracy of EBRA in measuring femoral stem migration is poor compared with RSA and it is less appropriate to be used as a surrogate marker for predicting long-term outcome in THA (Malak et al. 2016). Initial migration is common for many uncemented femoral stems, and is related to initial setting of the stem in the femoral canal due to the start of postoperative weight-bearing (Camp-

Acta Orthopaedica 2020; 91 (2): 159–164

bell et al. 2011, Bøe et al. 2011, von Schewelov et al. 2012, Weber et al. 2014). However, most studies perform their first follow-up RSA analysis of uncemented hip stems only after 3 months (Bøe et al. 2011, Weber et al. 2014). In a 2-year RSA and DEXA study of the fully HA-coated Corail stem (De Puy, Warsaw, IN, USA) in patients with a femoral neck fracture, von Schewelov et al. (2012) reported that during the first 3 months the Corail stem also moved mainly distally 2.7 mm, and rotated into retroversion 3.3˚, after which the position of the prosthesis stabilized. The first RSA examination in that study was performed before weight-bearing was allowed. However, a second examination was done before the patient left hospital, which is different from our study. Campbell et al. (2011) reported the 2-year RSA results of the same Corail stem, but in patients with osteoarthritis. Initial mean subsidence in this study was 0.73 mm at 6 months, and mean retroversion was 1.8˚ at 6 months. This difference in migration for the same stem indicates that initial stability is more difficult to obtain in patients with femoral neck fractures. Bøe et al. (2011) reported in their 2-year RSA and DEXA study of the Taperloc uncemented hip stem with either the BoneMaster (BM) plasma-sprayed HA coating (both Zimmer Biomet, Warsaw, IN, USA) or the electrochemically deposited HA coating, that both stems also showed initial migration during the first 3 months and then stabilized. After 2 years mean subsidence was 0.28 mm for BM and 0.25 mm for HA, and mean retroversion was 0.46˚ for BM and 0.17˚ for HA, both of which are less compared with the results in our study. However, it is unclear whether their baseline RSA examination was performed before or after patients were allowed to weight-bear. Weber et al. (2014) showed, in their randomized controlled RSA study with 5-year follow-up, a statistically significantly larger subsidence for the collarless Furlong Active stem (JRI Orthopaedics, Sheffield, UK) (0.99 mm) compared with its precursor the collar-fitted Furlong HAC stem (0.31 mm) after 3 months. Mean retroversion was 1.2˚ for the Active stem and 0.8˚ for the HAC stem after 3 months, which is substantially less compared with the Symax stem. Their baseline RSA examination was also made before weight-bearing. No other differences in migration were seen by RSA, and after 3 months the position of both stems stabilized. These findings show that there is a substantial variability in the amount of initial stem subsidence between stem designs, as was also seen for other uncemented hip stems (Ström et al. 2007). However, as mentioned earlier, it seems that the amount of initial subsidence is less important than that the position of all these stems that stabilized after 3 months, and in our study already after 1 month. So, a minor degree of early migration within the first few months, or month, can be seen as a “settling in” period, and is not a sign of inferior osseointegration. Van der Voort et al. (2015) showed in their systematic review and meta-analysis that there was no clear pattern between early migration and late aseptic revision of uncemented femoral stems. They also suggested that the

163

characteristic property of early stabilization of migration is probably a more suitable criterion than the absolute value of migration to identify safe uncemented hip stems. Comparison of the absolute value of migration of uncemented hip stems is not always straightforward as the postoperative reference RSA is sometimes made after the initiation of weight-bearing and sometimes before. Therefore, to create uniformity in RSA studies, we suggest performing baseline RSA examination before the start of weight-bearing, a first follow-up RSA examination after 1 month, and a third RSA examination after 3 months. The first phase (0–1 month) will show the initial stability due to the geometry of the stem. The second phase (1–3 months) will show secondary stability as a result of the osseointegrative capacity of the coating. Excellent clinical performance of the Symax hip stem was seen, as represented in HHS, OHS, WOMAC, and EQ-5D. These findings are in line with the good clinical performance of this implant as reported in a 1-year study, in which the Symax hip stem was compared with the predominantly diaphyseal anchored Hipstar hip stem (Stryker, Duisburg, Germany) and the straight Zweymuller (SL-Plus) hip stem (Plus Orthopedics AG, Rotkreuz, Switzerland) by Bergschmidt et al. (2010). Nevertheless, they discontinued using the Symax hip stem because of subsidence of more than 10 mm in 2 patients, and 3 intraoperative periprosthetic fractures outside the study group. We did not see these complications, nor in our previous 5-year clinical and radiographic study, in which we reported excellent clinical outcomes and radiographic signs of excellent progressive proximal fixation and favorable bone remodeling (Kruijntjens et al. 2018). The cohorts of patients of both previously mentioned studies were independent. Furthermore, a Danish single-center registry study of the Symax hip stem showed a median 6.5-year survival rate of 97.5% (95% CI 96.6–98.3%), while the overall median 6.5-year survival rate for uncemented hip stems was 95% in the same Danish registry (Edwards et al. 2018). In this registry study with follow-up of up to 10 years, 29 of 1,055 hip stems were revised. No revisions were due to aseptic loosening. A possible limitation of this study is the missing data for RSA during follow-up. The relatively high dropout rate of 10 patients out of 45 is due to the ISO criteria (ISO 16087, 2013) for stability of the markers (ME < 0.35) and for rotational reproducibility and positioning (CN < 120). Due to over-projection of markers on the medial side of the prosthesis, markers remained on the lateral side with a CN > 120. If we had used the criteria of Valstar et al. (2005), in which a CN < 150 was accepted, only 4 patients would have been excluded because of too high a condition number. Together with the patient with only 2 markers, the dropout would have been only 5 patients. However, we decided to use the stricter criteria of the ISO to decrease the error sensitivity for rotational outcomes. Patients who could not be evaluated with RSA showed excellent clinical performance on an individual basis during all follow-up evaluations.

164

In summary, the RSA analysis of the uncemented Symax hip stem confirms that the stem obtains stable fixation as early as 4 weeks postoperatively, after limited initial subsidence, retroversion, and slight posterior translation. There was no correlation between the amount of subsidence and retroversion at 4 weeks, and clinical outcomes after 2 years. The RSA, clinical, and radiographic evaluations all showed an excellently performing Symax hip stem after 2 years. Based on the predictive potential of the RSA technique, we anticipate excellent longterm survival of this hip stem. Supplementary data Tables 3–4 and Figure 4 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1709956 DK: reorganized clinical database, performed calculations and statistics, wrote and revised the manuscript; LK: organised and performed RSA analysis, wrote and revised the manuscript; BK: organised and contributed to RSA analysis, revised the manuscript; LJ: performed patient inclusion and exclusion, data entry, contributed to data management, revised the manuscript; JA: contributed to study and data management, revised the manuscript; RtB: co-designed the study, included patients, operated, performed clinical and radiographic evaluations, contributed to writing, and revised the manuscript. The authors would like to thank Jan Geurts as the second senior hip surgeon who operated on some of the patients, and to Sander van Kuijk for his help with statistical analysis. Acta thanks Stergios Lazarinis and Tatu Mäkinen for help with peer review of this study.

Becker P, Neumann H G, Nebe B, Luthen F, Rychly J. Cellular investigations on electrochemically deposited calcium phosphate composites. J Mater Sci Mater Med 2004; 15: 437-40. Bergschmidt P, Bader R, Finze S, Gankovych A, Kundt G, Mittelmeier W. Cementless total hip replacement: a prospective clinical study of the early functional and radiological outcomes of three different hip stems. Arch Orthop Trauma Surg 2010; 130: 125-33. Bøe B G, Röhrl S M, Heier T, Snorrason F, Nordsletten L. A prospective randomized study comparing electrochemically deposited hydroxyapatite and plasmasprayed hydroxyapatite on titanium stems. Acta Orthop 2011; 82: 13-19. Buratti C A, D’Arrigo C, Guido G, Lenzi F, Logroscino G D, Magliocchetti G, Mannocci C, Patella S, Patella V, Salvi V, Speranza A, Speciale D, Spinarelli A, Topa D. Assessment of the initial stability of the Symax femoral stem with EBRA-FCA: a multicentric study of 85 cases. Hip Int 2009; 19: 24-9. Campbell D, Mercer G, Nilsson K G, Wells V, Field J R, Callary S A. Early migration characteristics of a hydroxyapatite-coated femoral stem: an RSA study. Int Orthop 2011; 35: 483-8. Capello W N, D’Antonio J A, Geesink R G, Feinberg J R, Naughton M. Late remodelling around a proximally HA-coated tapered titanium femoral component. Clin Orthop Relat Res 2009; 467: 155-65.

Acta Orthopaedica 2020; 91 (2): 159–164

Edwards N M, Varnum C, Kjærsgaard-Andersen P. Up to 10-year follow-up of the Symax stem in THA: a Danish single-centre study. Hip Int 2018; 28: 375-81. ISO/TC 150 Committee (Implants for surgery S S, Bone and joint replacement). ISO 16087: 2013 (E). In: Implants for surgery: Roentgen stereophotogrammetric analysis for the assessment of migration of orthopaedic implants ISO 16087: 2013 (E). Switzerland; 2013. Kaptein B L, Valstar E R, Spoor C W, Stoel B C, Rozing P M. Model-based RSA of a femoral hip stem using surface and geometrical shape models. Clin Orthop Relat Res 2006; 448: 92-7. Kärrholm J, Borssen B, Lowenhielm G, Snorrason F. Does early micromotion of femoral stem prostheses matter? 4–7-year stereoradiographic follow-up of 84 cemented prostheses. J Bone Joint Surg (Br) 1994; 76: 912-7. Kruijntjens D S, Kjaersgaard-Andersen P, Revald P, Leonhardt J S, Arts J J, ten Broeke R H. 5-year clinical and radiographic follow-up of the uncemented Symax hip stem in an international study. J Orthop Surg Res 2018; 13: 191. Malak T T, Broomfield J A, Palmer A J, Hopewell S, Carr A, Brown C, PrietoAlhambra D, Glyn-Jones S. Surrogate markers of long-term outcome in primary total hip arthroplasty: a systematic review. Bone Joint Res 2016; 5: 206-14. Nieuwenhuijse M J, Valstar E R, Kaptein B L, Nelissen R G. Good diagnostic performance of early migration as a predictor of late aseptic loosening of acetabular cups: results from ten years of follow-up with Roentgen stereophotogrammetric analysis (RSA). J Bone Joint Surg (Am) 2012; 94: 874-80. ten Broeke R H M, Alves A, Baumann A, Arts J J C, Geesink R G T. Bone reaction to a biomimetic third-generation hydroxyapatite coating and new surface treatment for the Symax hip stem. J Bone Joint Surg (Br) 2011; 93-B: 760-8. ten Broeke R H, Hendrickx R P, Leffers P, Jutten L M, Geesink R G. Randomised trial comparing bone remodelling around two uncemented stems using modified Gruen zones. Hip Int 2012; 22: 41-9. ten Broeke R H M, Tarala M, Arts J J C, Janssen D W, Verdonschot N, Geesink R G T. Improving peri-prosthetic bone adaptation around cementless hip stems: a clinical and finite element study. Med Eng Phys 2014; 36: 345-53. Ström H, Nilsson O, Milbrink J, Mallnim H, Larsson S. The effect of early weight bearing on migration pattern of the uncemented CLS stem in total hip arthroplasty. J Arthroplasty 2007; 22: 1122-9. Szmukler-Moncler S, Perrin D, Ahossi V, Pointaire P. Evaluation of BONIT, a fully resorbable CaP coating obtained by electrochemical deposition, after 6 weeks of healing: a pilot study in the pig maxilla. Key Engineering Materials 2001; 192-195: 395-8. 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: 563-72. van der Voort P, Pijls B G, Nieuwenhuijse M J, Jasper J, Fiocco M, Plevier J W, Middeldorp S, Valstar E R, Nelissen R G. Early subsidence of shapeclosed hip arthroplasty stems is associated with late revision: a systematic review and meta-analysis of 24 RSA studies and 56 survival studies. Acta Orthop 2015; 86: 575-85. von Schewelov T, Ahlborg H, Sanzén L, Besjakov J, Carlsson A. Fixation of the fully hydroxyapatite-coated Corail stem implanted due to femoral neck fracture: 38 patients followed for 2 years with RSA and DEXA. Acta Orthop 2012; 83:153-8. Weber E, Sundberg M, Flivik G. Design modifications of the uncemented Furlong hip stem result in minor early subsidence but do not affect further stability: a randomized controlled RSA study with 5-year follow-up. Acta Orthop 2014; 85: 556-61.

Acta Orthopaedica 2020; 91 (2): 165–170

165

Outcome of revision hip arthroplasty in patients younger than 55 years: an analysis of 1,037 revisions in the Dutch Arthroplasty Register Martijn F L KUIJPERS 1, Gerjon HANNINK 2, Liza N VAN STEENBERGEN 3, and B Willem SCHREURS 1,3 1 Radboud

University Medical Center, Radboud Institute for Health Sciences, Department of Orthopaedics, Nijmegen; 2 Radboud University Medical Center, Radboud Institute for Health Sciences, Department of Operating Rooms, Nijmegen; 3 Dutch Arthroplasty Register (Landelijke Registratie Orthopedische Implantaten), ‘s Hertogenbosch, the Netherlands Correspondence: martijn.kuijpers@radboudumc.nl Submitted 2019-04-10. Accepted 2019-11-19.

Background and purpose — The increasing use of hip arthroplasties in young patients will inevitably lead to more revision procedures at younger ages, especially as the outcome of their primary procedures is inferior compared with older patients. However, data on the outcome of revision hip arthroplasty in young patients are limited. We determined the failure rates of revised hip prostheses performed in patients under 55 years using Dutch Arthroplasty Register (LROI) data. Patients and methods — All 1,037 revised hip arthroplasty procedures in patients under 55 years at the moment of revision registered in the LROI during the years 2007–2018 were included. Kaplan–Meier survival analyses were used to calculate failure rates of revised hip arthroplasties with endpoint re-revision for any reason. Competing risk analyses were used to determine the probability of re-revision for the endpoints infection, dislocation, acetabular and femoral loosening, while other reasons for revisions and death were considered as competing risks. Results — Mean follow-up of revision procedures was 3.9 years (0.1–12). 214 re-revisions were registered. The most common reason for the index revision was dislocation (20%); the most common reason for re-revision was infection (35%). The 5-year failure rate of revised hip prostheses was 22% (95% CI 19–25), and the 10-year failure rate was 28% (CI 24–33). The 10-year cumulative failure rates of index revisions with endpoint re-revision for infection was 7.8% (CI 6.1–9.7), acetabular loosening 7.0% (CI 4.1–11), dislocation 3.8% (CI 2.6–5.2), and femoral loosening 2.7% (CI 1.6–4.1). The 10-year implant failure rate of index revisions for infection was 45% (CI 37–55) with endpoint rerevision for any reason. Interpretation — Failure rate of revised hip prostheses in patients under 55 years is worrisome, especially regarding index revisions due to infection. This information facilitates realistic expectations for these young patients at the time of primary THA.

Total hip arthroplasty (THA) is used more and more in younger patients (Kurtz et al. 2009, Otten et al. 2010). Projections show that by the year 2030, more than half of all primary THA will be placed in patients younger than 65 years of age, with the biggest increase expected in patients between 45 and 54 years old (Kurtz et al. 2009). However, the outcome of primary THA in young patients is inferior compared with older patients (Walker et al. 2016, AOANJJR 2018, NJR 2018). Due to this increase in number of primary THA in young patients, and the inferior outcome, an increase in the number of revision arthroplasties is inevitable in young patients. Bayliss et al. (2017) have already shown that the lifetime risk of revision (LTRR) after THA increases with decreasing age at the time of primary surgery, with LTRR reaching almost 30% in patients between 50 and 54 years of age. Data on survivorship of revision procedures in young patients are limited. There are a few studies available that assessed the survival of revision procedures. The outcome of these studies was disappointing, with reported survival rates between 36% and 87% at 10-year follow-up (Girard et al. 2011, Adelani et al. 2014, Lee et al. 2014, Te Stroet et al. 2015, Beckmann et al. 2018). Besides this inferior outcome, most of these studies were single-center studies and had small sample sizes. In addition, previous reports focused primarily on implant design (Beckmann et al. 2018) or surgical technique (Comba et al. 2009), and there is a lack of reports focusing on the outcome of revisions in young patients using registry data. Understanding of the extent of the problem in revision arthroplasty in young patients is important, not only to reduce the number of re-revisions, but also to provide realistic expectations for this young patient group (Schreurs and Hannink 2017). Therefore, we determined the failure rate of revision hip arthroplasty performed in patients younger than 55 years of age using data from the Dutch Arthroplasty Register (LROI).

166

Acta Orthopaedica 2020; 91 (2): 165–170

Primary THA in patients younger than 55 years performed in the Netherlands 2007–2018 (primary or metastatic tumor excluded) n = 28,034 Excluded (n = 26,997): – cases without revision, 25,961 – cases with revision at age > 55, 342 – dead, 694 Indication for index revision THA in included patients younger than 55 years (n = 1,037): – dislocation, 210 – infection, 169 – acetabular loosening, 162 – femoral loosening, 162 – wear cup/liner, 3) – periprosthetic fractur, 75 – other, 414

Figure 1. Flowchart of patient selection.

Patients and methods The LROI (Dutch Arthroplasty Register) is a nationwide population-based register collecting data on arthroplasties. Initiated by the Dutch Orthopaedic Association, data collection started in 2007. The database has coverage of all Dutch hospitals, a completeness of over 95% of primary THA and 88% for revision arthroplasty (van Steenbergen et al. 2015), and 98% for both primary and revision THA in recent years (LROI 2018). Prosthesis characteristics are derived from an implant library within the LROI, which contains core characteristics of prostheses used in the Netherlands based on the article number (van Steenbergen et al. 2015). For this study, we selected all primary THA placed between January 1, 2007 and December 31, 2018 in patients younger than 55 years in the Netherlands (n = 28,034). Primary THAs performed because of a tumor (primary or metastatic) were excluded. Next, we included the subsequent revision procedures from this cohort in patients who were younger than 55 at time of their index revision procedure (n = 1,037) (Figure 1). A revision procedure was defined as an exchange of at least 1 of the components of the implant. Within the LROI, 3 revision categories are distinguished: (1) total revision—indicates a revision of the complete prosthesis, replacing both the acetabular and femoral components, (2) major partial revision— indicates a revision procedure where at least the femoral or the acetabular component is revised, and (3) minor partial revision—indicates a revision procedure where only the head and/ or the liner of the prosthesis is replaced. 2-stage revisions and Girdlestone procedures are registered at time of the definitive re-implantation of the prosthesis/components. This study was conducted and reported according to STROBE guidelines. Statistics Survival time of the implant inserted during the revision procedure was calculated as time from the index revision proce-

dure to re-revision, death of the patient or the end of study follow-up (January 1, 2019). In case of a Girdlestone procedure during the revision procedure, survival time is calculated between the re-implantation of the prosthesis (index revision) and re-revision, death of the patient, or the end of study follow-up. Kaplan–Meier survival analyses were used to estimate the survival of the implants inserted during all index revision procedures with endpoint re-revision for any reason. Results of Kaplan–Meier analyses were reported as cumulative failure rate (1 – KM) with 95% confidence intervals (CI). Next, implant survival with endpoint re-revision for any reason for the following subgroups: (1) revision category (i.e., total revision, major, and minor partial revision), and (2) reason for index revision (i.e., acetabular loosening, dislocation, and infection) were estimated using Kaplan–Meier survival analyses. Log-rank tests were used to test for differences in survival between groups. Using competing risk analyses, the probabilities of rerevision with endpoint re-revision for acetabular and femoral loosening, dislocation, and infection were determined, where death and other reasons for re-revision were considered as competing events. All analyses were performed using R version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Ethics, funding, data sharing, and potential conflict of interest Ethical approval was not applicable, as all data were received completely anonymous. This study was funded by the Van Rens Foundation, the Netherlands (VRF2017-009). The funding body had no role in the design of the study, data collection, analysis and interpretation, or in writing of the manuscript. Data are available from the LROI (Dutch Arthroplasty Registry) but restrictions apply to the availability of these data, which were used under license for the current study. The authors declare that they have no competing interests.

Results Characteristics of the study population Between January 1, 2007 and December 31, 2018, 1,037 index revisions (number of patients = 1,019) were registered in the LROI. Median age at time of revision was 49 years (18–54), and 53% were females. Other patient and implant characteristics are given in Table 1. The most common reason for the index revision was dislocation (20%), followed by infection (16%), acetabular loosening (16%), and femoral loosening (16%) (Table 2). The mean followup of the index revision procedures was 3.9 years (0.1–11.8). Of 1,037 index revisions, 21% of cases had replacement of both the acetabular and femoral component (total revision). In 53% of all index revisions, there was at least a replacement of the acetabular or femoral component (major partial revision),

Acta Orthopaedica 2020; 91 (2): 165–170

167

Table 1. Patient characteristics of 1,037 revisions including percentages in parantheses Factor Age (years) a Sex Female Male Missing ASA classification I II III–IV Missing Reason for index revision b Loosening acetabulum Loosening femur Dislocation Infection Wear cup/liner Periprosthetic fracture Other c

Index revisions (n = 1,037) 49 (18–54) 548 (53) 488 (47) 1 (0.1) 401 (39) 470 (45) 115 (11) 51 (5) 162 (16) 162 (16) 210 (20) 169 (16) 35 (4) 75 (7) 414 (40)

a b

Median (range) Total is more than 100%, as patients can have more than 1 reason for revision. c Includes periarticular ossification, symptomatic MoM, and Girdlestone procedures.

where 18% were a revision of the head and/or a replacement of the liner (minor partial revision). Of all major partial revisions, 57% involved a cup revision, where in 43% the femoral component was revised. In 8% of all index revisions, the revision category involved either a Girdlestone procedure, was reported as other, or was missing. In 3 cases, there was no re-implantation of a prosthesis after a Girdlestone procedure. Therefore, these cases were excluded from the survival analyses. There were 169 index revisions because of an infection. Of these, 44% cases had a minor partial revision (replacing only the head or liner), indicating a DAIR procedure. Furthermore, 29% of these cases were registered as a Girdlestone procedure, indicating a 2-stage revision procedure. Additionally, 22% of these cases were registered as total revision, of which 13 cases had Girdlestone as reason for revision. Therefore, these procedures can also be considered as a 2-stage revision, resulting in a total of 37% 2-stage revision procedures. The remaining 14% of total revision procedures were 1-stage revision procedures. There were 4% major partial revisions with reason given as infection; 2 cup revisions and 4 stem revisions, which were also considered as 1-stage revision procedures. Re-revision procedures 214 re-revision procedures were registered. The most common reason for re-revision was infection (35%), followed by acetabular loosening (16%) and dislocation (16%) (Table 2). Of 214 re-revision procedures, 29% had replacement of both the acetabular and femoral component (total re-revision).

Table 2. Patient characteristics of 214 re-revisions including percentages in paranthesis Factor

Re-revisions (n = 214)

Age (years) a Sex Female Male ASA classification I II III–IV Missing Reason for re-revision b Loosening acetabulum Loosening femur Dislocation Infection Wear cup/liner Periprosthetic fracture Other c a–c See

50 (19–58) 103 (48) 111 (52) 51 (24) 117 (55) 36 (17) 10 (5) 34 (16) 21 (10) 34 (16) 74 (35) 8 (4) 7 (3) 96 (45)

Table 1.

In 41% cases of all re-revisions, there was at least replacement of the acetabular or femoral component (major partial re-revision), where 18% of cases were revision of the head and/or replacement of the liner (minor partial re-revision). From all major partial re-revisions, 76% involved the cup, whereas 24% involved the femoral component. In 12% of the re-revised hips, the type of re-revision involved a Girdlestone procedure, was reported as other, or was missing. Failure rates of index revisions Using Kaplan–Meier, the 5-year implant failure rate of the 1,037 index revisions with endpoint re-revision for any reason was 22%. At 10-year follow-up, the implant failure rate was 28%. The 5- and 10-year cumulative failure rates of index revisions with endpoint re-revision for infection were 7.5% and 7.8%. For acetabular loosening, the 5- and 10-year cumulative failure rates were 3.1% and 7.0%. For dislocation, this was 3.8% and 3.8%. For femoral loosening, the 5- and 10-year cumulative failure rates were 2.3% and 2.7% (Table 3). Failure rates by revision category The 5- and 10-year failure rate for total revision procedures was 15% and 18% (CI 12.5–21.2). For major partial revisions, the 5- and 10-year failure rate was 16% and 22%. For minor partial revisions, this was 31% and 50% (Table 3). A log-rank test showed a significant difference in failure between categories of revision (p < 0.001, Figure 2). Failure rates by reason for index revision The failure rate of index revisions with reason given as infection was high. Using Kaplan–Meier analyses, the 5-year

168

Acta Orthopaedica 2020; 91 (2): 165–170

Table 3. Failure rate (%) of all index revisions, by category of revision and by reason for index revision Factor

5-year failure rate (95% CI)

10-year failure rate (95% CI)

All index revisions with endpoint re-revision for any reason 22 (19–25) 28 (24–33) dislocation 3.8 (2.6–5.2) 3.8 (2.6–5.2) infection 7.5 (5.9–9.3) 7.8 (6.1–9.6) acetabular loosening 3.1 (2.1–4.4) 7.0 (4.1–11) femoral loosening 2.3 (1.5–3.5) 2.7 (1.6–4.1) Revisions category total revision 15 (11–21) 18 (13–21) major partial revision 16 (13–20) 22 (17–27) minor partial revision 31 (24–39) 50 (32–73) Reason for index revision infection 45 (37–55) 45 (37–55) dislocation 22 (16–29) 29 (20–41) acetabular loosening 22 (16–33) 31 (21–44) femoral loosening 18 (13–26) 22 (14–34)

implant failure rate of these procedures, with endpoint rerevision for any reason, was 45%. At 10 years, the failure rate was 45%. For index revisions with reason given as dislocation, the 5-year failure rate with endpoint re-revision for any reason was 22%, and the failure rate at 10 years was 29%. For index revisions with reason given as acetabular loosening, the 5- and 10-year implant failure rate was 22% and 31%. For index revisions with reason given as femoral loosening, the 5-year failure rate with endpoint re-revision for any reason was 18%, and the failure rate at 10 years was 22% (Table 3, Figure 3). Additionally, patients who had an index revision with reason given as infection had a high cumulative failure rate for endpoint re-revision for a recurrent infection. At 5 years, more than 30% of all patients underwent a re-revision procedure with reason given as infection (cumulative failure rate 30%; CI 23–38). The competing risk analysis showed that the cumulative failure for a re-revision with recurrent reason was much lower for patients who underwent an index revision with reason given as dislocation, acetabular loosening, or femoral loosening. Only 8% of patients who underwent their index revision procedure with reason given as dislocation had a rerevision for another dislocation (cumulative failure rate 8.3%, CI 4.7–13). For acetabular loosening, this was only 5.3% (CI 2.3–10) at 5 years. For femoral loosening, the cumulative failure rate at 5 years for recurrent loosening of the femur was 5.0% (CI 2.2–9.5).

Discussion Our analysis showed a 5-year failure rate of index revision procedures with endpoint re-revision for any reason of 22% (CI 19–25), and 28% (CI 24–33) at 10-year follow-up.

Number at risk Year: 0 1 2 3 4 5 6 7 8 9 10 Major 549 449 373 314 242 196 136 85 62 34 12 Minor 187 136 106 86 66 46 29 21 13 6 4 Total 220 192 169 155 138 107 85 48 30 12 8

Figure 2. Failure rate by revision category with endpoint re-revision for any reason.

Number at risk Year: 0 1 2 3 4 5 6 7 8 9 10 Infection 166 86 63 45 36 21 15 8 6 1 1 Dislocation 210 161 133 112 83 64 46 36 29 16 7 Loosening: acetabular 161 126 103 90 72 61 51 35 24 13 5 femoral 162 129 112 93 71 54 36 22 14 5 2

Figure 3. Failure rate by reason for index revision with endpoint rerevision for any reason.

Comparison with literature Survival at 5-year follow-up was lower when compared with the available literature on young patients (Lee et al. 2014, Gromov et al. 2015, Te Stroet et al. 2015). Few papers ana-

Acta Orthopaedica 2020; 91 (2): 165–170

lyzed the mid- to long-term survival of revisions in this patient group, the number of included patients in these studies was limited, and focus was primarily on implant design. Te Stroet et al. (2015) reported a survival rate of 87% after follow-up of 10 years in a single-center study with only 34 revision procedures. Lee et al. (2014) reported a survival rate of 63% at 10-year follow-up, whereas survival at 5 years was approximately 88%. Several studies assessed the survival of revision hip arthroplasty in older patients, where reported survival varied between 81% and 83% at 5 years (Jafari et al. 2010, Ong et al. 2010) and 72% at 10 years (Lie et al. 2004). However, these results are relatively dated, and reason for revision was not reported in all studies, which makes comparison difficult. The most prevalent reason for the index revision was dislocation, whereas the most common reason for re-revision was infection. The rate of infections in the index revision procedures was 16%, which increased to 35% in all re-revision procedures. That infections are more prevalent as reason for re-revisions when compared with index revision procedures shows that management of infections plays an important role for prevention of re-revisions. Additionally, it is known that dislocation is a common complication associated with THA (Gwam et al. 2017, Seagrave et al. 2017, Rajaee et al. 2018). This was confirmed in our data, where dislocation was the most frequent reason for index revisions (16%). However, in re-revision procedures, dislocation as reason for re-revision is less pronounced when compared with infections. For prevention of re-revisions, the focus should be on treatment of infections (Berry 2017). Moreover, the survival of index revisions with reason given as infection was poor. At 5 years, almost half of all revised hips due to an infection resulted in re-revision. Furthermore, the number of re-infections was high in this group. Within 5 years, approximately 30% of all index revisions with reason given as infection underwent re-revision for a re-infection. For other reasons for revision, these numbers were much lower, with only 8% for dislocation, and 5% for both acetabular and femoral loosening. We found a substantial difference in failure rates between the different categories of revision. Approximately 70% of all index revisions were a partial revision, where in the majority of these procedures either the cup or the stem was replaced (major partial revision). Failure rate of the minor partial revisions (replacement of head and/or insert) was higher when compared with major partial revisions or total revisions, which is supported by data from the Swedish Hip Arthroplasty Register (Mohaddes et al. 2013). A possible explanation for this might be the high percentage of index revisions with reason given as infection using this method, or the exchange of heads to prevent further dislocations. Of all minor partial revisions, 40% were DAIR procedures (minor partial revisions for infection). The numbers of index revisions with reason given as infection using a total revision or a major partial revision were

169

much lower, at respectively 17% and 1%. Nevertheless, the effectiveness of the minor partial revision should be reconsidered, as survival of this revision category is lower. Limitations and strengths The completeness of revision hip arthroplasty in the Dutch Arthroplasty Register is lower compared with the completeness of primary THA, especially in the period 2007–2009, when there was no complete coverage of all Dutch hospitals. Second, there is most likely an under-registration of infections in the registry, as reoperations for infection without replacement of any of the components are not registered in the LROI (Lindgren et al. 2014, Gundtoft et al. 2015, SHAR 2015). In addition, since the outcome of index revisions due to infection is poor, information related to use of antibiotics (e.g., type of antibiotics, use of antibiotic-loaded bone cement, and adherence to guidelines on antibiotics administration) would be particularly valuable to obtain insight into this serious problem. Unfortunately, this information is not available from the Dutch Arthroplasty Register. A strength is that, compared with literature, our analysis includes a much larger number of revision procedures. Conclusion The cumulative failure rate in revision hip arthroplasty performed in patients under 55 years is worrisome. In particular, the outcome of index revisions due to infection is alarming, with a failure rate of 45% at 10-year follow-up. Moreover, within 5 years, 30% of all patients with an index revision for infection underwent a re-revision procedure with reason given as infection. Therefore, in the prevention of (re-)revisions, management of infections should play an essential role. MK, GH, LN, BS: concept and design. MK, GH, LN, BS: data analysis and interpretation. MK, GH, BS: manuscript preparation. MK, GH, LN, BS: manuscript editing. MK, GH, LN, BS: manuscript review. MK, GH, LN, BS: final approval of the version submitted. Acta thanks Keijo Mäkelä and Maziar Mohaddes for help with peer review of this study.

Adelani M A, Crook K, Barrack R L, Maloney W J, Clohisy J C. What is the prognosis of revision total hip arthroplasty in patients 55 years and younger? Clin Orthop Relat Res 2014; 472(5): 1518-25. doi: 10.1007/ s11999-013-3377-9. AOANJJR. National Joint Replacement Registry, Annual Report 2018. Available from: https://aoanjrr.sahmri.com/documents/10180/576950/ Hip%2C%20Knee%20%26%20Shoulder%20Arthroplasty. Bayliss L E, Culliford D, Monk A P, Glyn-Jones S, Prieto-Alhambra D, Judge A, Cooper C, Carr A J, Arden N K, Beard D J, Price A J. The effect of patient age at intervention on risk of implant revision after total replacement of the hip or knee: a population-based cohort study. Lancet 2017; 389(10077): 1424-30. doi: 10.1016/S0140-6736(17)30059-4. Beckmann N A, Hasler J F, Moradi B, Schlegel U J, Gotterbarm T, Streit M R. Long-term results of acetabular reconstruction using Ganz acetabular rings. J Arthroplasty 2018; 33(11): 3524-30. doi: 10.1016/j.arth.2018.06.036.

170

Berry D J. Joint registries: what can we learn in 2016? Bone Joint J 2017; 99-B(1 Suppl. A): 3-7. doi: 10.1302/0301-620X.99B1.BJJ-2016-0353.R1. Comba F, Buttaro M, Pusso R, Piccaluga F. Acetabular revision surgery with impacted bone allografts and cemented cups in patients younger than 55 years. Int Orthop 2009; 33(3): 611-6. doi: 10.1007/s00264-007-0503-x. Girard J, Glorion C, Bonnomet F, Fron D, Migaud H. Risk factors for revision of hip arthroplasties in patients younger than 30 years. Clin Orthop Relat Res 2011; 469(4): 1141-7. doi: 10.1007/s11999-010-1669-x. Gromov K, Pedersen AB, Overgaard S, Gebuhr P, Malchau H, Troelsen A. Do rerevision rates differ after first-time revision of primary THA with a cemented and cementless femoral component? Clin Orthop Relat Res 2015; 473(11): 3391-8. doi: 10.1007/s11999-015-4245-6. Gundtoft P H, Overgaard S, Schonheyder H C, Moller J K, KjaersgaardAndersen P, Pedersen A B. The “true” incidence of surgically treated deep prosthetic joint infection after 32,896 primary total hip arthroplasties: a prospective cohort study. Acta Orthop 2015; 86(3): 326-34. doi: 10.3109/17453674.2015.1011983. Gwam C U, Mistry J B, Mohamed N S, Thomas M, Bigart K C, Mont M A, Delanois R E. Current epidemiology of revision total hip arthroplasty in the United States: national inpatient sample 2009 to 2013. J Arthroplasty 2017; 32(7): 2088-92. doi: 10.1016/j.arth.2017.02.046. Jafari S M, Coyle C, Mortazavi S M, Sharkey P F, Parvizi J. Revision hip arthroplasty: infection is the most common cause of failure. Clin Orthop Relat Res 2010; 468(8): 2046-51. doi: 10.1007/s11999-010-1251-6. Kurtz S M, Lau E, Ong K, Zhao K, Kelly M, Bozic K J. Future young patient demand for primary and revision joint replacement: national projections from 2010 to 2030. Clin Orthop Relat Res 2009; 467(10): 2606-12. doi: 10.1007/s11999-009-0834-6. Lee P T, Lakstein D L, Lozano B, Safir O, Backstein J, Gross A E. Midto long-term results of revision total hip replacement in patients aged 50 years or younger. Bone Joint J 2014; 96-B(8): 1047-51. doi: 10.1302/0301620X.96B8.31587. 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. Lindgren J V, Gordon M, Wretenberg P, Karrholm J, Garellick G. Validation of reoperations due to infection in the Swedish Hip Arthroplasty Register. BMC Musculoskelet Disord 2014; 15:384. doi: 10.1186/1471-2474-15-384.

Acta Orthopaedica 2020; 91 (2): 165–170

LROI. Online LROI annual report 2018. Available from: http://www.lroirapportage.nl/media/pdf/PDF%20Online_LROI_annual_report_2018.pdf. Mohaddes M, Garellick G, Karrholm J. Method of fixation does not influence the overall risk of rerevision in first-time cup revisions. Clin Orthop Relat Res 2013; 471(12): 3922-31. doi: 10.1007/s11999-013-2872-3. NJR 15th Annual Report 2018, National Joint Registry for England, Wales, Northern Ireland and the Isle of Man. Available from: http://www.njrreports.org.uk/Portals/0/PDFdownloads/NJR%2015th%20Annual%20 Report%202018.pdf. Ong K L, Lau E, Suggs J, Kurtz S M, Manley M T. Risk of subsequent revision after primary and revision total joint arthroplasty. Clin Orthop Relat Res 2010; 468(11): 3070-6. doi: 10.1007/s11999-010-1399-0. Otten R, van Roermund P M, Picavet HS. [Trends in the number of knee and hip arthroplasties: considerably more knee and hip prostheses due to osteoarthritis in 2030]. Ned Tijdschr Geneeskd 2010; 154: A1534. Rajaee S S, Campbell J C, Mirocha J, Paiement G D. Increasing burden of total hip arthroplasty revisions in patients between 45 and 64 years of age. J Bone Joint Surg Am 2018; 100(6): 449-58. doi: 10.2106/JBJS.17.00470. Schreurs B W, Hannink G. Total joint arthroplasty in younger patients: heading for trouble? Lancet 2017; 389(10077): 1374-5. doi: 10.1016/S01406736(17)30190-3. Seagrave K G, Troelsen A, Malchau H, Husted H, Gromov K. Acetabular cup position and risk of dislocation in primary total hip arthroplasty. Acta Orthop 2017; 88(1): 10-17. doi: 10.1080/17453674.2016.1251255. SHAR. Register Annual Report 2015. Available from: https://registercentrum. blob.core.windows.net/shpr/r/Annual-Report-2015-H19dFINOW.pdf. Te Stroet M A, Rijnen W H, Gardeniers J W, van Kampen A, Schreurs B W. Satisfying outcomes scores and survivorship achieved with impaction grafting for revision THA in young patients. Clin Orthop Relat Res 2015; 473(12): 3867-75. doi: 10.1007/s11999-015-4293-y. van Steenbergen L N, Denissen G A, Spooren A, van Rooden S M, van Oosterhout F J, Morrenhof J W, Nelissen R G. More than 95% completeness of reported procedures in the population-based Dutch Arthroplasty Register. Acta Orthop 2015; 86(4): 498-505. doi: 10.3109/17453674.2015.1028307. Walker R P, Gee M, Wong F, Shah Z, George M, Bankes M J, Ajuied A. Functional outcomes of total hip arthroplasty in patients aged 30 years or less: a systematic review and meta-analysis. Hip Int 2016; 26(5): 424-31. doi: 10.5301/hipint.5000376.

Acta Orthopaedica 2020; 91 (2): 171–176

171

Radiographic parameter-driven decision tree reliably predicts aseptic mechanical failure of compressive osseointegration fixation Ryland KAGAN 1, Lindsay PARLEE 1, Brooke BECKETT 2, James B HAYDEN 1, Kenneth R GUNDLE 1,3, and Yee-Cheen DOUNG 1 1 Department

of Orthopaedics and Rehabilitation, Oregon Health & Science University, Portland, OR; 2 Department of Diagnostic Radiology, Oregon Health & Science University, Portland, OR; 3 Operative Care Division, Portland Veterans Administration Medical Center, Portland, OR, USA Correspondence: Kagan@ohsu.edu Submitted 2019-08-23. Accepted 2020-01-03.

Background and purpose — Compressive osseointegration fixation is an alternative to intramedullary fixation for endoprosthetic reconstruction. Mechanical failure of compressive osseointegration presents differently on radiographs than stemmed implants, therefore we aimed to develop a reliable radiographic method to determine stable integration. Patients and methods — 8 reviewers evaluated 11 radiographic parameters from 29 patients twice, 2 months apart. Interclass correlation coefficients (ICCs) were used to assess test–retest and inter-rater reliability. We constructed a fast and frugal decision tree using radiographic parameters with substantial test–retest agreement, and then tested using radiographs from a new cohort of 49 patients. The model’s predictions were compared with clinical outcomes and a confusion matrix was generated. Results — 6 of 8 reviewers had non-significant intra-rater ICCs for ≥ one parameter; all inter-rater ICCs were highly reliable (p < 0.001). Change in length between the top of the spindle sleeve and bottom of the anchor plug (ICC 0.98), bone cortex hypertrophy (ICC 0.86), and bone pin hypertrophy (ICC 0.81) were used to create the decision tree. The sensitivity and specificity of the training cohort were 100% (95% CI 52–100) and 87% (CI 74–94) respectively. The decision tree demonstrated 100% (CI 40–100) sensitivity and 89% (CI 75–96) specificity with the test cohort. Interpretation — A stable spindle length and at least 3 cortices with bone hypertrophy at the implant interface predicts stable osseointegration; failure is predicted in the absence of bone hypertrophy at the implant interface if the pin sites show hypertrophy. Thus, our decision tree can guide clinicians as they follow patients with compressive osseo integration implants.

Compressive osseointegration fixation offers an alternative to cemented or non-cemented intramedullary stems for endoprosthetic reconstruction. This technology creates a stable, high-pressure bone–implant interface that theoretically avoids stress shielding (Frost 1994, Kramer et al. 2008, Bini et al. 2000). The continuous force at the bone–implant interface creates ingrowth into the porous surface of the component, resulting in stable integration. A spring system within the Compress (Zimmer Biomet, Warsaw, IN, USA) immediately applies high compressive forces to the bone–implant interface (Figure 1). This technology has the potential to decrease the rate of aseptic mechanical failure (loosening), allow for stable short-segment fixation, and preserve bone stock if revision is

Figure 1. Compress components with distal femoral adaptor.

172

required (Bini et al. 2000, Calvert et al. 2014, Monument et al. 2015). Loosening of compressive osseointegration technology is thought to occur sooner than with traditional intramedullary stems, which fail later due to stress shielding (Pedtke et al. 2012, Kagan et al. 2017). Due to these differences, radiographic parameters used to assess other forms of endoprosthetic reconstruction may be insufficient for identifying and predicting aseptic mechanical failure of compressive osseointegration fixation. Bone hypertrophy visible at the bone–implant interface is thought to represent stable integration and thus it is a radiographic parameter often used to evaluate compressive osseointegration technology (Kramer et al. 2008, Pedtke et al. 2012, Healey et al. 2013, Zimel et al. 2016). Radiographic findings indicative of loosening include progressive gross decrease in the spindle height (Pedtke et al. 2012), deformation of the implant that suggests bending or breaking of the device (Healey et al. 2013), and bony atrophy at the bone–implant interface (Healey et al. 2013. However, there is no suitable methodology for evaluating aseptic mechanical failure of this technology (Pedtke et al. 2012, Healey et al. 2013). Lacking validated gold standards, quantitative radiographically based tools such as the Radiographic Union Score for Tibial (RUST) fractures (Whelan et al. 2010) or the Radiographic Union Score for Hip (RUSH) score (Chiavaras et al. 2013) have been developed. These tools show how systematically evaluating radiographic parameters improves reliability and reproducibility. Fast and frugal decision trees (FFTs) are a classification heuristic that provides dichotomous choices in series (Phillips et al. 2017). This model has been used to stratify patient risk for acute myocardial injury (Green and Mehr 1997). If reliable radiographic parameters suggestive of compressive osseointegrative failure are used as decision points, an FFT may be a suitable model for categorizing patient risk of aseptic mechanical failure. To evaluate whether there has been stable integration of compressive osseointegration technology we asked: What is a reliable radiographic method for determining stable bone– implant fixation? What radiographic parameters best show failure of fixation? Can an FFT be used to classify radiographs into stable and failed fixation categories?

Patients and methods In this 2-phase cohort study, separate cohorts of patient radiographs were evaluated to develop and validate (i.e., train and test) a model to predict aseptic mechanical failure of the Compress in the lower extremity. Training cohort Radiographs from patients who received Compress implants between 2006 and 2014 were reviewed. During this time, sur-

Acta Orthopaedica 2020; 91 (2): 171–176

TRAINING COHORT Patients identified n = 109 Compress n = 132

TEST COHORT Patients identified n = 82 Compress n = 87

Patients selected for inclusion n = 29

Patients selected for inclusion n = 49

AP and lateral radiographs reviewed – 6-week postoperative radiographs – 1-year postoperative radiographs

AP and lateral radiographs re-reviewed ≥ 2 months later – 6-week postoperative radiographs – 1-year postoperative radiographs

Figure 2. Training and test cohort protocol.

geons at one center implanted 132 lower extremity Compress devices in 109 patients. Indications for compressive osseointegration ﬁxation use included reconstruction of the proximal femur, distal femur, and proximal tibia where there was massive bone loss necessitating endoprosthetic reconstruction. Patients were considered for this technology if they had previous failed arthroplasty, fracture nonunion, malunion, or required a reconstruction after an oncologic resection. Older age was not a contraindication for use. Study inclusion criteria were a minimum clinical and radiographic follow-up of 2 years. The compressive force used on each patient was determined based on the cortical thickness of the bone. The spindle size and shape were determined, based on the individual patient anatomy, at the time of surgery by 1 of the 2 senior authors (YCD, JBH). The diameter chosen was always larger than the largest diameter of the bone so that there is overhang for potential bony hypertrophy at the bone–implant interface. Antirotation pins were not routinely used, and a preference for 800 pounds per square inch was given whenever there was a sufficient amount of remaining cortical bone. The spindle surface type (hydroxyapatite or porous titanium) was determined by the availability of the implants. Following surgery, all patients were instructed to follow a strict touchdown weight-bearing protocol for 6 weeks, followed by progression to weight-bearing as tolerated. To identify reliable parameters consistent with aseptic mechanical failure, radiographs of 29 patients from this cohort were evaluated (Figure 2). 3 of these patients were known loosenings who went on to revision of the implant. 8 reviewers were asked to assess the 29 sets of patient postoperative radiographs. Each set included anterior-posterior and lateral radiographs taken within 6 weeks post-operation and taken at approximately 1-year postoperatively. Each set of patient radiographs was assessed twice by all reviewers, with the second assessment done no less than 2 months after the first was completed. The reviewers had varying levels of medical education and included 2 orthopedic oncology attending faculty (JBH, YCD), 1 orthopedic resident, 2 musculoskeletal

Acta Orthopaedica 2020; 91 (2): 171–176

173

Unchanged length between the spindle sleeve and the anchor plug? YES

Hypertrophy in more than 3 bone cortices? YES Correct rejection 30 Miss 0 YES Correct rejection 15 Miss 0

False alarm Hit

0 2

NO Hypertrophy ≤ 25% at the bone pin interface? NO False alarm Hit

7 4

Figure 4. Training cohort FFT.

Figure 3. Radiographic parameters used to construct the fast and frugal tree model. A. Immediate postoperative radiograph of distal femoral reconstruction showing the initial length between the top of the spindle sleeve and bottom of anchor plug (Spindle). B. 1-year postoperative radiograph of the same patient as in A, depicting the loss of length of the Spindle. C. Immediate postoperative radiograph of distal femoral reconstruction. D. 1-year postoperative radiograph of the same patient as in Figure 3C, depicting hypertrophy at both the bone–implant interface (BoneInterface) and at the bone–pin interface (BonePins).

radiology attending faculty, and 3 medical students. In both phases of the study reviewers were blinded to patient name, demographics, and outcomes of the implants. All radiographs were reviewed and scoring was done on our institutional digital PACS system (Agfa Healthcare IMPAX Version 6.5.5.3020; Agfa Healthcare, Mortsel, Belgium). For a power analysis with 8 raters, and an expected ICC of at least 0.7, selecting 29 subjects provided 90% confidence that the ICC would fall within 0.3 of the reported value (Saito et al. 2006). The radiographic parameters assessed were varus/valgus alignment (Coronal), flexion/extension alignment (FlexEx), evidence of bone hypertrophy at the implant (BoneInterface) and at the pins (BonePins), evidence of bone osteolysis (OsteolysisInterface), evidence of intramedullary remodeling (IMRemodel), number of cortices with bone hypertrophy (BoneCortices), difference in bone width (DeltaBone), and distance between the top of the spindle sleeve and the bottom of the anchor plug (Spindle) (Figure 3). In addition, scores

similar to RUST and modified RUST were assessed. Scores for each of the 4 cortices were assigned 1 (no change), 2 (hypertrophy with osteolysis), or 3 (hypertrophy without osteolysis). Each cortical score was summed; thus, total scores of bone hypertrophy ranged from 4 to 12 per radiograph. After radiographs were scored, intraclass correlation coefficients (ICCs) were calculated to assess test–retest and interrater reliability. The ICCs were used to determine which parameters to use in constructing the FFT. The final parameters used were: (1) Spindle; (2) BoneCortices; and (3) BonePins (Figure 4). The FFT model was trained to predict aseptic mechanical failure based on the 58 radiographs (2 from each of 29 patients) from this training cohort. Test cohort It is important to test such models separately, to avoid overfitting. Therefore, in addition to the training cohort, we used a second separate cohort to test the FFT. The test cohort consisted of patients who received lower extremity Compress implants between March 2013 and November 2017. During this time surgeons at one center treated 82 patients with 87 Compress implants for lower extremity reconstructions. Indications, inclusion/exclusion criteria, follow-up, surgical specifications, implant location, and postoperative protocol remained unchanged from the training cohort. From the 82 patients, anterior-posterior and lateral radiographs from 49 patients were randomly chosen to be reviewed. The reviewers included 2 orthopedic oncology attendings (JBH, YCD) and 1 orthopedic arthroplasty attending (RK) from the same institution. They each reviewed 49 patient radiographs a single time. Each set included anterior-posterior and lateral radiographs taken within 6 weeks postoperatively and taken at approximately 1-year postoperatively (Figure 2). Each of the three reviewers evaluated the radiographs using the three parameters identified in phase I of the study: (1) Spindle; (2) BoneCortices; (3) BonePins. The scored radiographs were evaluated by the FFT model (Figure 5). Sensitivity, specificity, and balanced accuracy (BACC) (Phillips et al.

174

Acta Orthopaedica 2020; 91 (2): 171–176

Table 1. Training cohort: test–retest intra-rater reliability Parameter a

Tester 1 Tester 2 Tester 3 Tester 4 Tester 5 Tester 6 Tester 7 Tester 8 ICC p-value ICC p-value ICC p-value ICC p-value ICC p-value ICC p-value ICC p-value ICC p-value

Spindle Coronal FlexEx BoneInterface OsteolysisInterface BonePins IMRemodel BoneCortices SBH MSBH DeltaBone

0.98 < 0.001 0.89 < 0.001 0.93 < 0.001 0.93 < 0.001 0.92 < 0.001 0.82 < 0.001 0.33 0.04 0.25 0.09 0.87 < 0.001 0.74 < 0.001 0.35 0.03 0.03 0.5 0.36 0.03 0.62 < 0.001 0.91 < 0.001 0.88 < 0.001 0.36 0.02 0.29 0.06 0.55 0.001 0.52 0.001 0.82 < 0.001 0.48 < 0.001 0.49 0.003 0.14 0.2 0.31 0.05 0.48 0.003 0.69 < 0.001 0.99 < 0.001 0.31 0.05 0.77 < 0.001 0.59 < 0.001 0.78 < 0.001 0.59 < 0.001 0.79 < 0.001 0.67 < 0.001 0.69 < 0.001 0.24 0.10 0.41 0.01 0.57 < 0.001 0.57 < 0.001 0.39 0.02 0.40 0.01 0.86 < 0.001 0.89 < 0.001 0.92 < 0.001 0.86 < 0.001 0.68 < 0.001 0.73 < 0.001 0.74 < 0.001 0.9 < 0.001 0.92 < 0.001 0.92 < 0.001 0.30 0.06 0.64 < 0.001 0.79 < 0.001 0.89 < 0.001 0.94 < 0.001 0.92 < 0.001 0.28 0.07 0.19 0.2 0.87 < 0.001 0.85 < 0.001 0.99 < 0.001 0.54 < 0.001 0.86 < 0.001 0.88 < 0.001

0.86 < 0.001 0.45 0.006 0.22 0.1 0.17 0.2 0.22 0.1 0.67 < 0.001 0.69 < 0.001 0.42 0.01 0.99 < 0.001 0.32 0.04 0.81 < 0.001 0.65 < 0.001 0.61 < 0.001 0.71 < 0.001 0.63 < 0.001 0.73 < 0.001 0.47 0.004 0.54 0.001 0.47 0.004 0.42 0.01 0.75 < 0.001 0.52 0.002

a Spindle: distance

between the top of the spindle sleeve and the bottom of anchor plug; Coronal: varus/valgus alignment; FlexEx: flexion/extension alignment; BoneInterface: evidence of bone hypertrophy at implant interface; OsteolysisInterface: evidence of bone osteolysis; BonePins: evidence of bone hypertrophy at anchor plug pins; IMRemodel: evidence of intramedullary remodeling; BoneCortices: number of cortices with bone hypertrophy; SBH: score of bone hypertrophy; MSBH: modified score of bone hypertrophy; DeltaBone: difference in bone width.

2017) were calculated. Statistical analyses were completed in R version 3.5.0 (R Core Team, 2018; R Foundation for Statistical Computing, Vienna, Austria) within RStudio version 1.1.453, including the FFTree package (Phillips et al. 2017). Ethics, funding, and potential conflicts of interest The Oregon Health & Science University institutional review board approved this study and waived requirement for written informed consent. There was no external funding provided for this study. JBH receives royalties from Zimmer Biomet for a product unrelated to this study. No other authors report a conflict of interest.

Table 2. Training cohort: inter-rater reliability Parameter a

ICC p-value

Spindle Coronal FlexEx BoneInterface OsteolysisInterface BonePins IMRemodel BoneCortices SBH MSBH DeltaBone

0.81 0.34 0.46 0.54 0.53 0.65 0.32 0.68 0.58 0.52 0.76

a See

< 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 < 0.001

footnote Table 1.

Results When comparing the test–retest reliability of individual reviewers, the intraclass correlation (ICC) p-value was not always significant. 6 of 8 reviewers had an intra-rater ICC that was not statistically significant for 1 or more parameters (Table 1). 6 of 11 parameters had 1 or more reviewers with non-statistically significant intra-rater reliability. Coronal, FlexEx, and MSBH were parameters where 2 or more reviewers had non-statistically significant intra-rater reliabilities (Table 1). Some parameters are less reproducible by raters, and thus would not be useful if incorporated into a clinical model for predicting compressive osseointegration fixation aseptic mechanical failure. While individual reviewers may have had non-statistically significant intra-rater ICCs for 1 or more parameters, all of the inter-rater ICCs were highly reliable (p < 0.001) (Table 2). Both the intra-rater and inter-rater reliability were highest for change in distance between the top of the spindle sleeve and the bottom of the anchor plug (Spindle, ICC = 0.98; ICC = 0.81) and bone cortices with hypertrophy (BoneCortices, ICC

= 0.86; ICC = 0.68). Since these parameters had the highest intra- and inter-rater ICC and were statistically significant, they were used as decision points in the FFT (Figure 3). A sensitivity analysis was completed while only including staff orthopedic surgeons and radiologists, without any substantial differences in the results. When the radiograph scores were interpreted by the FFT they went through a series of 3 yes or no questions to predict whether aseptic mechanical failure had occurred or not. FFTs are designed with balance in mind. In order to have a balanced FFT there must be a balanced number of nodes representing negative and positive outcome, with the final node being a neutral outcome. The negative and positive outcome parameters elected were Spindle and BoneCortices, with BonePins as the neutral node. Any time a data point (radiograph) reached a node and the finding was present, a decision was made and the data point exited the FFT. Those data points that did not exit continued to the next node on the FFT. Predicted failures exited on the

Acta Orthopaedica 2020; 91 (2): 171–176

175

Unchanged length between the spindle sleeve and the anchor plug? YES

Hypertrophy in more than 3 bone cortices? YES Correct rejection 21 Miss 0 YES Correct rejection 19 Miss 0

False alarm Hit

0 2

for the test cohort was determined to be 100% (CI 40–100) sensitive and 89% (CI 75–96) specific, BACC 94 (Figure 5). The model was then applied to the subset of patients with distal femoral reconstructions. We identified 24 distal femoral reconstructions, which included 4 failures. The FFT had a 100% sensitivity with 3 false positives, for an 85% specificity.

NO Hypertrophy ≤ 25% at the bone pin interface?

Discussion

NO False alarm Hit

5 2

Figure 5. Test cohort FFT.

right side of the table while those not predicted to fail exited to the left. Of the 58 radiographs in the training cohort, 2 demonstrated change in the Spindle parameter and thus exited to the right. Of the remaining 56 radiographs that continued to the next question, 30 were found to have hypertrophy in more than 3 bone cortices (BoneCortices) and exited to the left. The 26 remaining data points continued to the final neutral question, where 15 exited left because they had no pin hypertrophy and 11 exited right due to pin hypertrophy; of these, 7 were false positives and 4 were true failures (Figure 3). The predictions made by the FFT were then compared with the clinical outcomes associated with each patient. Of the 13 radiographs that were predicted to fail, 6 were true failures. All of the radiographs that were predicted not to fail did not end up failing. Thus, the sensitivity of the FFT was 100% (95% CI 52–100) and the specificity 87% (CI 74–94), BACC 93. In the training cohort, failure was always indicated when there was an observable change in Spindle parameter although this measurement alone was not sensitive. Furthermore, whenever there was hypertrophy visible at the bone–implant interface, failure did not occur. Finally, while hypertrophy at the bone pins was predictive of failure, failure did not always result when this was observed on the radiograph. The same FFT created for the training cohort was used to evaluate the test cohort. 49 radiographs, each scored by 3 reviewers, were run through the FFT with similar outcomes to the training cohort. 2 radiographs demonstrated changes in the Spindle parameter and were predicted to fail, so they exited right. 47 radiographs continued to the second question, and of those 21 demonstrated hypertrophy in 3 or more bone cortices and thus were predicted not to fail and exited to the left. 26 radiographs reached the final question before exiting the FFT. 19 of those did not demonstrate hypertrophy at the bone pins and thus exited to the left and were not expected to fail. Meanwhile, 7 displayed bone pin hypertrophy and exited to the right, because they were expected to fail. In comparison with clinical outcomes, of the 9 radiographs that were expected to fail, only 4 did. Additionally, all of 40 radiographs anticipated not to fail, did not fail. Similar to the training cohort, the FFT

Compressive osseointegration is an alternative to intramedullary fixation for endoprosthetic reconstruction, with modes of failure that present differently on radiographs. We found that the test–retest reliability of individual reviewers was not always statistically significant, illustrating that some parameters are less reproducible by raters. While individual reviewers may have had non-statistically significant intra-rater ICCs for 1 or more parameters, all of the inter-rater ICCs were highly reliable (p < 0.001). The intra-rater and inter-rater reliability were highest for the Spindle parameter (ICC 0.98; ICC 0.81) and Bone Cortices with Hypertrophy (ICC 0.86; ICC 0.68) and these parameters were used as decision points in the FFT. The sensitivity and specificity of the FFT was 100% and 87%, respectively, for the training cohort and 100% and 89%, respectively, for the test cohort. This model may be helpful for ruling in, and even more helpful for ruling out, loosening. We found the test–retest reliability of individual reviewers’ p-value was not always significant. Given that the reviewers had different levels of training, it is reasonable to suspect that reviewers with the least training were less capable of reproducing the same results when reading the same radiograph for a second time. It is possible that reviewers 5, 6, and 7 had the least training given that reviewer 6 had non-significant intrarater ICCs for 4 parameters, and reviewer 5 and 7 each had non-significant intra-rater ICCs for 2. In the development of the RUST score the authors (Whelan et al. 2010) also included reviewers with varying levels of education, from trainee (resident) physicians, to community orthopedic surgeons and fellowship-trained surgeons. They found a trend towards improved reliability for the fellowship-trained traumatologists compared with those with less training. While the individual reviewers may have had non-statistically significant intra-rater ICCs for 1 or more parameters, all inter-rater ICCs were highly reliable suggesting that once individuals become proficient at reading radiographs, they will be able to reproduce their findings. It is likely that statistical significance was achieved for all 11 inter-rater ICC parameters since 6 of the 8 reviewers were either faculty level or in their postgraduate medical education. Pedtke et al. (2012) suggested 5 separate radiographic parameters suggestive of aseptic failure of osseointegration that were evaluated by the treating surgeons; however, interobserver variability was not assessed. 1 of the suggested parameters was a progressive gross decrease in the distance between

176

the anchor plug base and the top of the spindle sleeve. We found that both the intra-rater and inter-rater reliability were high for this radiographic parameter we called “Spindle.” Pedtke et al. along with multiple other authors (Kramer et al. 2008, Healey et al. 2013, Zimel et al. 2016) have also suggested that bone hypertrophy at the bone–implant interface, what we called “Bone Cortices with Hypertrophy,” was suggestive of stable osseointegration. We also found this to have high intra-rater and inter-rater reliability. Since these parameters had the highest intra- and inter-rater ICCs and were statistically significant, they were used as decision points as we developed our decision tree. Our model indicates that if the spindle is stable (has not changed height), and at least 3 cortices have bone hypertrophy, then stable osseointegration has occurred. However, if the spindle is stable, but there are less than 3 cortices with hypertrophy, then hypertrophy around the pins can be considered. Based on our model, bone pin hypertrophy was predictive of failure but failure did not always result when this was present on radiographs. It is possible that this may only be a clue as to inadequate or delayed bone healing at the implant–bone interface. Our study has a number of limitations. First, there currently is no gold standard with which to compare our decision tree. However, our results support the use of the FFT to create a standardized protocol for future investigations of compressive osseointegration and we hope our analysis is verified or modified in the future. Second, we found 100% sensitivity but only 87% specificity in the training cohort, and 89% specificity in the test cohort. This suggests that this decision tree is likely more helpful as a supplemental tool that clinicians may use for ruling out loosening. Third, we did not have patientreported outcomes or pain scores to correlate with the decision tree; these findings may also be associated with loosening and future investigations may consider including these clinical findings. Fourth, our training cohort was only 29 patients and there can be limitations due to this relatively small sample size. The study was powered to evaluate inter-rater reliability in the training cohort. While a strength of the study is the separate testing cohort, ultimately the sample size is limited by the relative infrequency of the scenarios for which these implants are used. Finally, this is 1 center’s experience, with surgeons and radiologists who have experience evaluating compressive osseointegration fixation. The results may not be generalizable to other centers, and the external validity of this model would benefit from evaluation in additional cohorts. This study takes a combination of radiographic parameters in a systematic approach to create a decision tree that may be utilized by clinicians evaluating compressive osseointegration fixation. Our decision tree showed high sensitivity and slightly lower specificity suggesting that this model may help clinicians rule out aseptic mechanical failures. Future studies

Acta Orthopaedica 2020; 91 (2): 171–176

should be performed to potentially improve on our decision tree by utilizing clinical outcomes or advanced cross-sectional imaging studies. Additionally, as there is currently no gold standard to evaluate compressive osseointegration, we hope this is simply a first step and a supplemental tool for clinicians and is improved upon in the future. RK, LP, and KG wrote the manuscript. KG analyzed the data. All authors contributed to the design of the study, critical evaluation of the data and analyses, interpretation of the findings, and critical revision of the manuscript, through all stages of the study. Acta thanks Richard O’Donnell and Yan Li for help with peer review of this study.

Bini S A, Johnston J O, Martin D L. Compliant prestress fixation in tumor prostheses: interface retrieval data. Orthopedics 2000; 23(7): 707-11. Calvert G T, Cummings J E, Bowles A J, Jones K B, Wurtz L D, Randall R L. A dual-center review of compressive osseointegration for fixation of massive endoprosthetics: 2- to 9-year followup. Clin Orthop Relat Res 2014; 472(3): 822-9. Chiavaras M M, Bains S, Choudur H, Parasu N, Jacobson J, Ayeni O, Petrisor B, Chakravertty R, Sprague S, Bhandari M. The Radiographic Union Score for Hip (RUSH): the use of a checklist to evaluate hip fracture healing improves agreement between radiologists and orthopedic surgeons. Skeletal Radiol 2013; 42(8): 1079-88. Frost H M. Wolff’s Law and bone’s structural adaptations to mechanical usage: an overview for clinicians. Angle Orthod 1994; 64(3): 175-88. Green L, Mehr D R. What alters physicians’ decisions to admit to the coronary care unit? J Fam Pract 1997; 45(3): 219-26. Healey J H, Morris C D, Athanasian E A, Boland P J. Compress knee arthroplasty has 80% 10-year survivorship and novel forms of bone failure. Clin Orthop Relat Res 2013; 471(3): 774-83. Kagan R, Adams J, Schulman C, Laursen R, Espana K, Yoo J, Doung Y C, Hayden J. What factors are associated with failure of compressive osseointegration fixation? Clin Orthop Relat Res 2017; 475(3): 698-704. Kramer M J, Tanner B J, Horvai A E, O’Donnell R J. Compressive osseointegration promotes viable bone at the endoprosthetic interface: retrieval study of Compress implants. Int Orthop 2008; 32(5): 567-71. Monument M J, Bernthal N M, Bowles A J, Jones K B, Randall R L. What are the 5-year survivorship outcomes of compressive endoprosthetic osseointegration fixation of the femur? Clin Orthop Relat Res 2015; 473(3): 883-90. Pedtke A C, Wustrack R L, Fang A S, Grimer R J, O’Donnell R J. Aseptic failure: how does the Compress((R)) implant compare to cemented stems? Clin Orthop Relat Res 2012; 470(3): 735-42. Phillips, N D, Neth, H, Woike J K, Gaissmaier W. A toolbox to create, visualize, and evaluate fast-and-frugal decision trees. Judgment and Decision Making 2017; 12(4): 344-68. Saito Y, Sozu T, Hamada C, Yoshimura I. Effective number of subjects and number of raters for inter-rater reliability studies. Stat Med 2006; 25: 1547-60. Whelan D B, Bhandari M, Stephen D, Kreder H, McKee MD, Zdero R, Schemitsch E H. Development of the radiographic union score for tibial fractures for the assessment of tibial fracture healing after intramedullary fixation. J Trauma 2010; 68(3): 629-32. Zimel M N, Farfalli G L, Zindman A M, Riedel E R, Morris C D, Boland P J, Healey J H. Revision distal femoral arthroplasty with the Compress((R)) prosthesis has a low rate of mechanical failure at 10 years. Clin Orthop Relat Res 2016; 474(2): 528-36.

Acta Orthopaedica 2020; 91 (2): 177–183

177

Reduced survival of total knee arthroplasty after previous unicompartmental knee arthroplasty compared with previous high tibial osteotomy: a propensity-score weighted mid-term cohort study based on 2,133 observations from the Danish Knee Arthroplasty Registry Anders EL-GALALY 1,2, Poul T NIELSEN 1, Andreas KAPPEL 1,2, and Steen L JENSEN 1,2 1 Orthopaedic Research Unit, Aalborg University Hospital, Aalborg; 2 Department of Clinical Medicine, Aalborg University, Aalborg, Denmark Correspondence: galaly@me.com Submitted 2019-10-22. Accepted 2019-12-03.

Background and purpose — Both medial unicompartmental knee arthroplasties (UKA) and high tibial osteotomies (HTO) are reliable treatments for isolated medial knee osteoarthritis. However, both may with time need conversion to a total knee arthroplasty (TKA). We conducted the largest nationwide registry comparison of the survival of TKA following UKA with TKA following HTO. Patients and methods — From the Danish Knee Arthroplasty Registry, aseptic conversions to TKA from UKA and TKA converted from HTO within the period of 1997–2018 were retrieved. The Kaplan–Meier method and the Cox proportional hazards regression were used to estimate the survival and hazard ratio (HR) for revision, considering confounding by indication utilizing propensity-score based inverse probability of treatment weighting (PS-IPTW). Results — PS-IPTW yielded a well-balanced pseudocohort (standard mean difference (SMD) < 0.1 for all covariates, except implant supplementation) of 963.8 TKAs following UKA and 1139.1 TKAs following HTO. The survival of TKA following UKA was significantly less than that of TKA following HTO with a 5-year estimated survival of 0.88 (95% confidence interval (CI) 0.85–0.90) and 0.94 (CI 0.93–0.96), respectively. The differences in survival corresponded to an implant-supplementation adjusted HR of 2.7 (CI 2.4–3.1) for TKA following UKA compared with TKA following HTO. Interpretation — Previous UKA more than doubled the revision risk of a subsequent TKA compared with previous HTO. This potential risk should be considered in the shared treatment decision of patients who are candidates for both UKA and HTO.

In isolated osteoarthritis of the medial knee compartment, both medial unicompartmental knee arthroplasties (UKA) and high tibial osteotomies (HTO) are solutions with reliable clinical outcomes (Cao et al. 2018). The survival of UKA is secondary to that of total knee arthroplasties (TKA) with a recent meta-analysis reporting 15-year survival of 76% and 93%, respectively (Evans et al. 2019). The long-term survival of HTO seems inferior to both UKA and TKA with a declining survival from 75% at 10 years to 55% at 15 years (van Wulfften Palthe et al. 2018). When UKA or HTO fail, conversion to TKA is a common solution (Lee et al. 2019). Nationwide registry studies have investigated the survival of either TKA following UKA or TKA following HTO compared with primary or revision TKA. They have reported an increased risk of revision in TKA following UKA (Robertsson and W-Dahl 2015, Leta et al. 2016, Lewis et al. 2018, El-Galaly et al. 2019) while no consensus regarding the influence of HTO on the survival of a subsequent TKA has been reached (Niinimäki et al. 2014, Badawy et al. 2015, Robertsson and W-Dahl 2015, El-Galaly et al. 2018). However, a direct comparison of the survival estimates from these studies is prone to confounding by indication due to a range of unadjusted baseline characteristics associated with the survival of TKA, such as implant constraints and hospital volume of arthroplasty surgeries (Jasper et al. 2016). This concern is further encouraged by a recent single-center study reporting similar short-term survival of TKA following UKA and TKA following HTO (Lim et al. 2017). Based on the Danish Knee Arthroplasty Registry (DKR), our study compares the survival of TKA converted from UKA with TKA converted from HTO with consideration for confounding by indication utilizing propensity-score based inverse probability of treatment weighting (PS-IPTW) (Inacio et al. 2015).

178

Acta Orthopaedica 2020; 91 (2): 177–183

Covariates At the time of TKA surgery, several patients and surgical characteristics are registered in the DKR (Table 1). Comorbidity is depicted by Charnley class sorted Medial unicompartmental Total knee arthroplasty knee arthroplasty (UKA) (TKA) preceded by high into class A (unilateral arthritis), class B1 (bilateral n = 11,514 tibial osteotomy (HTO) arthritis), class B2 (arthroplasty in the opposite knee), n = 1,163 and class C (other condition affecting walking capacExcluded (n = 10,526): Excluded due to other ity) (Bjorgul et al. 2010). Knee function at the time prior surgeries (n = 7): – unrevised UKA, 10,368 – revised due to infection, 50 – ACL reconstruction, 4 of surgery is registered using American Knee Society – osteosynthesis, 2 – revised to UKA, 106 Score (KKS) divided into clinical and functional sub– TKA before UKA, 2 – other, 1 scores, both ranging from 0 to 100 (Insall et al. 1989). TKA following HTO TKA following UKA Level of implant constraint is divided into cruciate n = 1,156 n = 988 retaining (CR), posterior stabilized (PS), constraint Excluded (n = 1) Excluded (n = 10) condylar (CCR), and hinged. Perioperative suppleRevised before Revised before index surgery index surgery mentation (stems, augments, or cones) and perioperative complications (e.g. fractures, rupture of the TKA following UKA TKA following UKA n = 1,155 n = 978 patellar tendon or ligament injuries) are registered. We defined the hospital volume of arthroplasties as Figure 1. Flowchart depicting the formation of the study cohort. the mean annual volume during the study period and divided it into 4 groups (< 100, 100–249, 250–449, > 449). Year of surgery was classified into 2 periods (1997–2007 and Patients and methods 2008–2018). For TKAs following UKAs, we retrieved type of Data source UKA bearings and indications of conversion to TKA. Since 1997, the DKR has prospectively collected information on Danish knee arthroplasties through standardized forms Outcome completed by surgeons. Since 2007, the registration of arthro- The outcome was TKA revision of any indication with revision plasties has been mandatory for all hospitals in Denmark lead- defined in accordance with the DKR as removal, exchange, or ing to a registry completeness above 90% for primary arthro- addition of an implant. The indications for TKA revision have plasties and 80% for revision arthroplasties (Danish Knee recently been thoroughly evaluated in both groups, and thus Arthroplasty Registry 2019). The DKR is reported suitable are not presented in this study (El-Galaly et al. 2018, 2019). for epidemiological studies and is crosslinked with the Danish Civil Registration System (DCRS) which contains vital and Missing values emigration status for all Danish citizens (Pedersen et al. 2012, Missing values existed in height (n = 1,214), weight (n = 67), Schmidt et al. 2014). Mandatory registration and linkage to KSS clinical sub-score (n = 55), KSS functional sub-score (n = the DCRS enable complete follow-up in a population-based 39), Charnley class (n = 12), fixation (n = 11), duration of surgery cohort (Schmidt et al. 2019). (n = 8), and patella replacement (n = 3). Missing values in height were deemed too high for meaningful imputation and discarded. Study cohort The remaining missing values were estimated by multiple impuIn the registry, each patient is identified by a unique code, and tation with chain equation (MICE), generating 5 datasets under the side of surgery is denoted. Therefore, each knee can be the assumption of missing at random (Azur et al. 2011). considered a unique observation, which has been reported to provide unbiased results in large arthroplasty studies (Rob- Statistics ertsson and Ranstam 2003). All UKAs indicated by osteoar- PS-IPTW thritis in knees without prior surgery from January 1, 1997 This study is subjected to confounding due to the non-random until December 31, 2018 were retrieved. We identified later assignment of prior UKA or HTO. Therefore, PS-IPTW was revisions conducted on the same knee and excluded revisions utilized to account for confounding by indication. PS were to UKA and conversions from UKA to TKA due to infections estimated with logistic regression and applied by IPTW with (Figure 1). In the same timeframe, we retrieved all TKAs indi- stabilized weights aiming to estimate the average effect of cated by osteoarthritis in knees previously treated with HTO. treatment (Austin 2014). Based on the considerations depicted The validity of the registration of a previous HTO was evalu- in the directed acyclic graph (Williams et al. 2018) (Figure ated in a recent study and confirmed in 96% of the cases (El- 2, see Supplementary data), the following covariates were Galaly et al. 2018). However, HTO is not divided in open- and included in the model: sex, age (quantiles), weight (quantiles), closed-wedge osteotomies, and thus the methods are consid- KSSs, Charnley class, level of constraint, patella resurfacing, fixation, hospitals annual arthroplasty volume, and period of ered as one in the DKR. Danish Knee Arthroplasty Registry Knee arthroplasty for primary osteoarthritis January 1, 1997 to December 31, 2018

Acta Orthopaedica 2020; 91 (2): 177–183

179

Table 1. Baseline characteristics for original cohort and the PS-IPTW cohort at the time of conversion to TKA Patient characteristics

Original cohort PS-IPTW cohort TKA TKA TKA TKA following following following following UKA HTO SMD UKA HTO SMD

Observations 978 1,155 963.8 1139.1 Male sex , n (%) 324 (33) 657 (57) 0.24 419.3 (44) 529.4 (47) Mean age (range) 66 (34–95) 63 (32–90) 0.35 64 (34–95) 64 (32–90) Mean weight, kg (range) 82 (30–183) 84 (30–200) 0.04 89 (30–183) 88 (30–200) Charnley class, n (%) 0.41 A 531 (54) 449 (39) 450.0 (47) 523.0 (46) B1 181 (18) 414 (36) 254.0 (26) 327.6 (29) B2 211 (22) 244 (21) 215.4 (22) 239.5 (21) C 55 (6) 48 (4) 44.3 (5) 49.0 (4) Knee Society Clinical score 0.22 mean (range) 41 (0–99) 35 (0–99) 35 (0–99) 36 (0–99) Knee Society Functional score 0.28 mean (range) 45 (0–100) 52 (0–100) 50 (0–100) 51 (0–100) Surgical characteristics, n (%) Level of constraint 0.62 Cruciate retaining 574 (59) 977 (85) 686.6 (71) 840.9 (74) Posterior stabilized 241 (25) 130 (11) 183.9 (19) 195.8 (17) Constrained condylar 161 (16) 44 (4) 90.9 (10) 99.3 (9) Hinged 2 (< 1) 4 (< 1) 2.3 (< 1) 3.1 (< 1) Fixation 0.49 Cemented 895 (91) 848 (73) 806.2 (84) 922.7 (81) Hybrid 68 (7) 228 (20) 115.7 (12) 160.0 (14) Uncemented 15 (2) 79 (7) 41.9 (4) 56.4 (5) Patella resurfacing 904 (78) 862 (88) 0.10 822.4 (85) 948.5 (83) Supplementation 0.80 Stem 271 (28) 27 (2) 216.5 (23) 51.6 (5) Augment 75 (8) 2 (< 1) 50.2 (5) 2.9 (< 1) Cone 59 (6) 1 (< 1) 31.5 (3) 6.1 (< 1) Annual arthroplasty volume 0.35 < 100 80 (8) 176 (15) 148.9 (15) 143.8 (13) 100–249 218 (22) 321 (28) 217.1 (23) 277.9 (24) 250–449 373 (38) 276 (24) 300.4 (31) 359.0 (31) > 449 307 (32) 382 (33) 297.4 (31) 358.4 (32) Period of surgery 0.38 1997–2007 144 (15) 604 (52) 333.3 (35) 405.9 (36) 2008–2018 834 (85) 551 (48) 630.5 (65) 733.2 (64)

0.03 0.06 0.05 0.06

0.02 0.06

0.06

0.07

0.02 0.58

0.09

0.01

SMD: Standardized mean difference.

surgery. Implant supplementation was rare in TKA following HTO and therefore omitted from the PS estimation to avoid overweighting rare observations. The balance of the baseline characteristics was evaluated graphically and by standardized mean differences (SMD) with an SMD of 0 indicating perfect balance and SMD < 0.1 deemed an acceptable balance between the groups (Austin 2009). Survival analyses The Kaplan–Meier method was used to estimate the survival with revision as the primary endpoint. Unrevised knees were censored by death, emigration, or end of study period at December 31, 2018. The risk of revision was estimated by Cox regression with robust variance estimator to account for dependencies in the PS-IPTW cohort. The assumption of proportional hazards was evaluated by Schoenfeld’s plots

and Schoenfeld’s residual test. Implant supplementation was included as covariate in the Cox regression to account for remaining imbalance following the PS-IPTW. E-value The robustness of the estimated hazard ratios (HR) was evaluated by calculating their E-values, which estimates the magnitude of association unmeasured confounders must have with both the exposure and outcome to negate the observed HRs (Van Der Weele and Ding 2017). Significance Means are presented with absolute range, medians with interquartile range (IQR), and SMD are calculated to assess balance between the groups. Estimates from the imputed datasets were combined by Rubin’s rule (White et al. 2011), and all

180

Acta Orthopaedica 2020; 91 (2): 177–183

Table 2. Baseline characteristics for original cohort and PS-IPTW cohort at the time of conversion to TKA. Values are counts/weighted counts (%)

Prior UKA surgery Original cohort PS-IPTW cohort

Indications for conversion Aseptic loosening 271 (28) 285.4 (29) Unexplained pain 262 (27) 274.7 (29) Progression of arthritis 243 (25) 207.2 (22) Instability 77 (8) 69.3 (7) Other 65 (6) 68.7 (7) Unknown 44 (4) 45.4 (5) Wear 16 (2) 13.0 (1) Bearing Mobile 823 (84) 759.0 (79) Fixed 155 (16) 204.8 (21)

estimates are presented with 95% confidence interval (CI) to address their significance (Ranstam 2019). Statistical programs Data were sorted in STATA 15 (StataCorp, College Station, TX, USA) and all analyses were conducted in R© Version 3.5.1 (R Foundation for Statistical Computing, Vienna, Austria). Ethics, funding, and potential conflict of interests The study was approved by the Danish Data Protection Agency (entry number: 2008-58-0028) and financed by the Orthopaedic Research Unit at Aalborg University Hospital. No conflict of interest is present among the authors.

Results Original cohort Characteristics The baseline covariates differed in sex, age, Charnley class, KKSs, level of constraint, patella resurfacing, fixation, implant supplementation, annual arthroplasty volume, and period of surgery as depicted in Table 1. The median duration of sur-

gery in TKA following UKA was 90 minutes (IQR 75–115) compared with 80 minutes (IQR 65–100) in TKA following HTO (SMD = 0.35). In 12 of the TKAs following UKAs perioperative complications (8 fractures, 4 other) were registered compared with 18 registered complications (5 fractures, 5 ligament/tendon rupture, 8 other) in TKAs following HTOs (SMD = 0.11). Survival Of the 978 TKAs following UKAs, 121 (12%) were revised within the study period and 93 (10%) were censored due to either death or emigration. In comparison, 101 (9%) and 234 (20%) of the 1,155 TKAs following HTOs were revised or censored, respectively. The median follow-up in TKA following UKA was 4.7 years (IQR 1.9–7.7) compared with 9.3 years (IQR 5.0–13) for TKA following HTO (SMD = 0.87). The 1st, 5th, and 10th year survival estimates were 0.97 (CI 0.96–0.98), 0.88 (CI 0.86–0.91), and 0. 82 (CI 0.78–0.85) for TKA following UKA compared with 0.98 (CI 0.97–0.99), 0.95 (CI 0.93–0.96), and 0.92 (CI 0.90–0.94) for TKA following HTO, which corresponds to an HR of 2.3 (CI 2.1–2.6) associated with TKA following UKA (Table 3). PS-IPTW cohort Characteristics Following PS-IPTW, all covariates included in the estimation of the PS were well balanced between TKAs following UKAs and TKAs following HTOs (Table 1 and Figure 3, see Supplementary data). However, the difference in implant supplementation was still unbalanced following PS-IPTW (SMD = 0.58). Table 2 depicts the distribution of indication of UKA conversion and type of UKA-bearing in TKA following UKA, which was clinically comparable before and after PS-IPTW. The imbalance in duration of surgery was unchanged by PSIPTW with a median duration of 90 minutes (IQR 75–120) in TKA following UKA and 80 minutes in TKA following HTO (IQR 66–100) (SMD = 0.35). PS-IPTW did not balance the difference in perioperative complications with registered complications in 7.7 of the TKAs following UKAs and 24.1 in TKAs following HTOs (SMD = 0.13).

Table 3. Survival estimates, hazard ratios (HR), and E-value for the original cohort and PS-IPTW cohort Follow-up Revision Survival estimates n median (IQR) n (%) 1-year (CI) 5-year (CI) 10-year (CI) Original cohort TKA following UKA 978 TKA following HTO 1155 PS-IPTW cohort TKA following UKA 963.8 TKA following HTO 1139.1 a

4.7 (1.9–7.7) 9.3 (5.0–13.4)

121 (12) 0.97 (0.96–0.98) 0.88 (0.86–0.91) 101 (9) 0.98 (0.97–0.99) 0.95 (0.93–0.96)

5.5 (2.1–9.3) 169.1 (17) 0.96 (0.95–0.97) 0.88 (0.85–0.90) 7.8 (4.2–11.2) 89.4 (8) 0.98 (0.97–0.99) 0.94 (0.93–0.96)

adjusted for differences in implant supplementation.

0.82 (0.78–0.85) 0.92 (0.90–0.94)

Hazard ratio E-value estimates estimates (CI) (lower CI) 2.3 (2.1–2.6) 4.1 (3.5) Ref. Ref.

0.75 (0.71–0.79) 2.7 a (2.4–3.1) 4.9 (4.3) 0.92 (0.90–0.94) Ref. Ref.

Acta Orthopaedica 2020; 91 (2): 177–183

181

Figure 4. Kaplan–Meier survival estimates for the PS-IPTW cohort with confidence interval and weighted knees at risk. Knees at risk: Years sincs TKA: TKA following HTO TKA following UKA

1,139 964

804 517

390 203

122 61

Survival Of the 963.8 TKAs following UKAs, 169.1 (17%) were revised within the study period and 117.5 (12%) were censored due to death or emigration. Similarly, 89.4 (8%) and 187.8 (16%) of the 1139.1 TKAs following HTOs were revised or censored, respectively. This corresponded to a significantly inferior survival of TKA following UKA compared with TKA following HTO (Figure 4 and Table 3), with an implant-supplementation adjusted HR of 2.7 (CI 2.4–3.1) associated with TKA following UKA (Table 3).

Discussion Our study showed that in a cohort from the Danish Knee Arthroplasty Registry, with well-balanced baseline covariates, the survival of TKA following UKA was lower than the survival of TKA following HTO. More specifically, the risk of revision more than doubled when TKA was preceded by UKA compared with HTO. During recent years, UKA has gained popularity while the use of HTO has decreased, indicating a trend towards treating patients with UKA instead of HTO (Niinimäki et al. 2012, Henkel et al. 2019). While both UKAs and HTOs relieve pain, the procedures differ. UKAs replace the diseased compartment with an implant and thus preserve the mechanical axis of the knee. In contrast, HTOs shift the mechanical axis laterally, unloading the diseased compartment while increasing the load on the lateral compartment. Therefore, a subsequent conver-

sion from HTO to TKA due to progression of osteoarthritis might be expected, whereas progression of arthritis or implant failure are considered an adverse event in UKA surgery. To our knowledge, this study constitutes the largest direct comparison of TKA following UKA with TKA following HTO based on a nationwide registry (Pearse et al. 2012) and expands upon a range of recent nationwide registry studies comparing the survival of either TKA following UKA or TKA following HTO with primary TKA (Niinimäki et al. 2014, Badawy et al. 2015, Robertsson and W-Dahl 2015, Leta et al. 2016, El-Galaly et al. 2018, 2019, Lewis et al. 2018). Due to inconsistent adjustment for confounding, a direct comparison of the results in the current literature might be affected by residual confounding, which could result in acceptance of a false causal relationship (Kyriacou and Lewis 2016). Our study compliments current literature by directly comparing TKA following UKA with TKA following HTO while expanding the statistical adjustment for confounding using PS-IPTW. The PS is the probability of an observation receiving a treatment given a set of baseline covariates and, thus, dependence on the PS creates balance in the included covariates between the groups. Dependence on the PS can be achieved by matching, weighting, adjusting, or stratifying (Austin 2011). We used IPTW and, thus, weighted the observation based on their inverse probability of treatment (i.e., PS) to create a pseudo-cohort with comparable baseline characteristics between the groups. As depicted in Figure 3 (see Supplementary data), this approach eliminated imbalances in a range of baseline covariates, and thus diminished the influence of the confounders presented in Figure 2 (see Supplementary data) except implant supplementation, which was including in the Cox regression. In this pseudo-cohort, TKA following UKA was associated with a 2.7-fold increase in the risk of revision compared with TKA following HTO. Limitations The study has some limitations. First, nationwide registries are prone to misclassifications. However, as the data are collected prospectively by the surgeon on a standardized form, the misclassifications are assumed to be non-differential and thus bias the results towards no difference between the groups. Second, even though the PS-IPTW successfully balances a wide range of covariates, residual confounding is unavoidable in non-randomized studies. We calculated the E-value for the presented HRs to elucidate which magnitude unmeasured confounders must have to negate the presented HRs (Table 3). The E-value indicated that unmeasured confounders must be associated with both TKA following UKA (exposure) and subsequent revisions (outcome) by a ratio of at least 4.3 (lower CI) to move the HR’s CI to include 1. In comparison, diabetes has recently been associated with a risk ratio of revision at 1.3 (CI 1.02–1.6) in a large retrospective study of both TKAs and total hip arthroplasties (Maradit Kremers et al. 2017). Therefore, the presented HRs seemed robust for residual confounding. Third, the completeness in the DKR has increased

182

from 69% of primary arthroplasties in 1997 to above 91% since 2008, with a similar evolution in revision arthroplasties with a completeness from 54% in 1997 to above 87% since 2008 (Danish Knee Arthroplasty Registry 2019). The overall completeness of TKA following HTO might be less than the overall completeness of TKA following UKA, as more HTOs were converted before 2008. This imbalance might have overestimated the risk of revision associated with TKA following UKA compared with TKA following HTO. We included the period of surgery in the PS estimation to contain the bias induced by the difference in completeness. Conclusion In this propensity-score weighted cohort study, TKA following UKA was associated with a more than 2-fold increased risk of revision compared with TKA following HTO. This potential risk emphasized that UKA should be considered a definitive treatment in line with TKA rather than a temporary treatment to postpone TKA. Supplementary data Figures 2 and 3 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674. 2019.1709711

The authors thank the Danish knee surgeons for a thorough registration of their procedures, the steering committee of the Danish Knee Arthroplasty Registry for their goodwill in relation to data acquisition, and Angélica Meleñdez-Muñoz for her linguistic contribution to the manuscript. All authors contributed to the study design. AEG received and analyzed the data with supervision from all authors. AEG wrote the initial draft, which was revised and accepted by all authors. Acta thanks Leif Ryd and Annette W-Dahl for help with peer review of this study.

Austin P C. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med 2009; 28(25): 3083-107. Austin P C. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res 2011; 46(3): 399-424. Austin P C. The use of propensity score methods with survival or time-toevent outcomes: reporting measures of effect similar to those used in randomized experiments. Stat Med 2014; 33(7): 1242-58. Azur M J, Stuart E A, Frangakis C, Leaf P J. Multiple imputation by chained equations: what is it and how does it work? Int J Methods Psychiatr Res 2011; 20(1): 40-9. Badawy M, Fenstad A M, Indrekvam K, Havelin L I, Furnes O. The risk of revision in total knee arthroplasty is not affected by previous high tibial osteotomy: a 15-year follow-up of 32,476 total knee arthroplasties in the Norwegian Arthroplasty Register 2015; 86(6): 734-9. Bjorgul K, Novicoff W M, Saleh K J. Evaluating comorbidities in total hip and knee arthroplasty: available instruments. J Orthop Traumatol 2010; 11(4): 203-9.

Acta Orthopaedica 2020; 91 (2): 177–183

Cao Z, Mai X, Wang J, Feng E, Huang Y. Unicompartmental knee arthroplasty vs high tibial osteotomy for knee osteoarthritis: a systematic review and meta-analysis. J Arthroplasty 2018; 33: 952-9. Danish Knee Arthroplasty Registry. Danish Knee Arthroplasty Registry— Annual Report 2019; 2019. El-Galaly A, Nielsen P T, Jensen S L, Kappel A. Prior high tibial osteotomy does not affect the survival of total knee arthroplasties: results from the Danish Knee Arthroplasty Registry. J Arthroplasty 2018; 33(7): 2131-5.e1. El-Galaly A, Kappel A, Nielsen P T, Jensen S L. Revision risk for total knee arthroplasty converted from medial unicompartmental knee arthroplasty. Comparison with primary and revision arthroplasties, Based on mid-term results from the Danish Knee Arthroplasty Registry. J Bone Joint Surg 2019; 101(22): 1999-2006. Evans J T, Walker R W, Evans J P, Blom A W, Sayers A, Whitehouse M R. How long does a knee replacement last? A systematic review and metaanalysis of case series and national registry reports with more than 15 years of follow-up. Lancet 2019; 393(10172): 655-63. Henkel C, Mikkelsen M, Pedersen A B, Rasmussen L E, Gromov K, Price A, Troelsen A. 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. Acta Orthop 2019; 90(4): 354-9 . Inacio M C S, Chen Y, Paxton E W, Namba R S, Kurtz S M, Cafri G. Statistics in brief: an introduction to the use of propensity scores. Clin Orthop Relat Res 2015; 473(8): 2722-6. 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. Jasper L L, Jones C A, Mollins J, Pohar S L, Beaupre L A. Risk factors for revision of total knee arthroplasty: a scoping review. BMC Musculoskelet Disord 2016; 17(1): 1-9. Kyriacou D N, Lewis R J. Confounding by indication in clinical research. JAMA 2016; 316(17): 1818-19. Lee Y, Kim H, Mok S, Lee O-S. Similar outcome, but different surgical requirement in conversion total knee arthroplasty following high tibial osteotomy and unicompartmental knee arthroplasty: a meta-analysis. J Knee Surg 2019; 32(7): 686-700. Leta T H, Lygre S H L, Skredderstuen A, Hallan G, Gjertsen J E, Rokne B, Furnes O. Outcomes of unicompartmental knee arthroplasty after aseptic revision to total knee arthroplasty: a comparative study of 768 TKAs and 578 UKAs revised to TKAs from the Norwegian Arthroplasty Register (1994 to 2011). J Bone Joint Surg Am 2016; 98(6): 431-40. Lewis P L, Davidson D C, Graves S E, De Steiger R N, Donnelly W, Cuthbert A. Unicompartmental knee arthroplasty revision to TKA: are tibial stems and augments associated with improved survivorship? Clin Orthop Relat Res 2018; 476(4): 854-62. Lim J B T, Chong H C, Pang H N, Tay K J D, Chia S L, Lo N N, Yeo SJ, Lim J B T, Joint B. Revision total knee arthroplasty for failed high tibial osteotomy and unicompartmental knee arthroplasty have similar patientreported outcome measures in a two-year follow-up study. Bone Joint J 2017; 99-B(10): 1329-34 . Maradit Kremers H, Schleck C D, Lewallen E A, Larson D R, Van Wijnen A J, Lewallen D G. Diabetes mellitus and hyperglycemia and the risk of aseptic loosening in total joint arthroplasty. J Arthroplasty 2017; 32(9): S251-3. Niinimäki T T, Eskelinen A, Ohtonen P, Junnila M, Leppilahti J. Incidence of osteotomies around the knee for the treatment of knee osteoarthritis: a 22-year population-based study. Int Orthop 2012; 36(7): 1399-402. Niinimäki T, Eskelinen A, Ohtonen P, Puhto A-P, Mann B S, Leppilahti J. Total knee arthroplasty after high tibial osteotomy: a registry-based casecontrol study of 1,036 knees. Arch Orthop Trauma Surg 2014; 134(1): 73-7. Pearse A J, Hooper G J, Rothwell A G, Frampton C. Osteotomy and unicompartmental knee arthroplasty converted to total knee arthroplasty: data from the New Zealand Joint Registry. J Arthroplasty 2012; 27(10): 1827-31. 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.

Acta Orthopaedica 2020; 91 (2): 177–183

Ranstam J. Time to restrict the use of p-values in Acta Orthopaedica. Acta Orthop 2019; 90(1): 1-2. Robertsson O, Ranstam J. No bias of ignored bilaterality when analysing the revision risk of knee prostheses: analysis of a population based sample of 44,590 patients with 55,298 knee prostheses from the national Swedish Knee Arthroplasty Register. BMC Musculoskelet Disord 2003; 4: 1. Robertsson O, W-Dahl A. The risk of revision After TKA is affected by previous HTO or UKA. Clin Orthop Relat Res 2015; 473(1): 90-3. Schmidt M, Pedersen L, Sorensen H T. The Danish Civil Registration System as a tool in epidemiology. Eur J Epidemiol 2014; 29(8): 541-9. Schmidt M, Schmidt S A J, Adelborg K, Sundbøll J, Laugesen K, Ehrenstein V, Sørensen HT. The Danish health care system and epidemiological

183

research: from health care contacts to database records. Clin Epidemiol 2019; 11: 563-91. Van Der Weele T J, Ding P. Sensitivity analysis in observational research: introducing the E-value. Ann Intern Med 2017; 167(4): 268-74. White I R, Royston P, Wood A M. Multiple imputation using chained equations: issues and guidance for practice. Stat Med 2011; 30(4): 377-99. Williams T C, Bach C C, Matthiesen N B, Henriksen T B, Gagliardi L. Directed acyclic graphs: a tool for causal studies in paediatrics. Pediatr Res 2018; 84(4): 487-93. van Wulfften Palthe A F Y, Clement N D, Temmerman O P P, Burger B J. Survival and functional outcome of high tibial osteotomy for medial knee osteoarthritis: a 10–20-year cohort study. Eur J Orthop Surg Traumatol 2018; 28(7): 1381-9.

184

Acta Orthopaedica 2020; 91 (2): 184–190

The effect of fixation type on the survivorship of contemporary total knee arthroplasty in patients younger than 65 years of age: a registerbased study of 115,177 knees in the Nordic Arthroplasty Register Association (NARA) 2000–2016 Mika J NIEMELÄINEN 1, Keijo T MÄKELÄ 2,3, Otto ROBERTSSON 4, Annette W-DAHL 4, Ove FURNES 5,6, Anne M FENSTAD 5, Alma B PEDERSEN 7, Henrik M SCHRØDER 8, Aleksi REITO 1, and Antti ESKELINEN 1,2 1 Coxa

Hospital for Joint Replacement, and Faculty of Medicine and Health Technologies, University of Tampere, Tampere, Finland; 2 Finnish Arthroplasty Register, National Institute for Health and Welfare, Helsinki, Finland; 3 Department of Orthopaedics and Traumatology, Turku University Hospital, Turku, Finland; 4 The Swedish Knee Arthroplasty Register, Department of Orthopedics, Skane University Hospital, Lund, Sweden; Department of Clinical Sciences, Orthopedics, Lund University, Sweden; 5 The Norwegian Arthroplasty Register, Department of Orthopaedic Surgery, Haukeland University Hospital, Bergen, Norway; 6 Department of Clinical Medicine, University of Bergen, Haukeland University Hospital, Bergen, Norway; 7 Department of Clinical Epidemiology, Aarhus University Hospital. Denmark and Danish Knee Arthroplasty Registry; 8 Department of Orthopaedic Surgery, Naestved Hospital, Denmark Correspondence: mika.niemelainen@coxa.fi Submitted 2019-07-17. Accepted 2019-12-03.

Background and purpose — Cemented fixation is regarded as the gold standard in total knee arthroplasty (TKA). Among working-age patients, there has been controversy regarding the optimal fixation method in TKA. To address this issue, we conducted a register-based study to assess the survivorship of cemented, uncemented, hybrid, and inverse hybrid TKAs in patients aged < 65 years. Patients and methods — We used the Nordic Arthroplasty Register Association data of 115,177 unconstrained TKAs performed for patients aged < 65 years with primary knee osteoarthritis over 2000–2016. Kaplan–Meier (KM) survival analysis with 95% confidence intervals (CI) and Cox multiple-regression model with adjustment for age, sex, and nation were used to compare fixation methods in relation to revision for any reason. Results — The 10-year KM survivorship of cemented TKAs was 93.6% (95% CI 93.4–93.8), uncemented 91.2% (CI 90.1–92.2), hybrid 93.0% (Cl 92.2–93.8), and inverse hybrid 96.0% (CI 94.1–98.1). In the Cox model, hybrid TKA showed decreased risk of revision after 6 years’ follow-up compared with the reference group (cemented) (hazard ratio [HR] 0.5 [CI 0.4–0.8]), while uncemented TKAs showed increased risk of revision both < 1 year (HR 1.4 [1.1–1.7]) and > 6 years’ (HR 1.3 [1.0–1.7]) follow-up compared to the reference. Interpretation — Both cemented and hybrid TKAs had 10-year survival rates exceeding 92–>93% in patients aged < 65 years. Cemented TKA, however, was used in the vast majority (89%) of the operations in the current study. As it performs reliably in the hands of many, it still deserves the status of gold standard for TKA in working-age patients.

Previous studies reported both highest increase in incidence of TKAs and also highest risk for revision in patients younger than 65 years of age (Julin et al. 2010, Carr et al. 2012, Leskinen et al. 2012, Meehan et al. 2014, Nemes et al. 2015, Niemelainen et al. 2017). This has increased the interest in finding a more durable fixation method for TKA. A previous systematic review did not report any differences in survival or functional outcome between cemented and uncemented TKAs in patients aged 60 years or less (Franceschetti et al. 2017). A meta-analysis without age limit showed better survival rates with cemented TKAs when all studies were combined, but in randomized studies survivals were equivocal (Gandhi et al. 2009). Uncemented fixation in TKA has offered outcomes comparable with cemented TKA in a few studies, but higher costs of uncemented components have favored cemented TKA still as gold standard (Dalury 2016, Miller et al. 2018, Zhou et al. 2018). A previous study applying radiostereometric analysis (RSA) showed that early migration seen with uncemented tibial components settled until 2 years whereas cemented ones continued to migrate (Wilson et al. 2012, Henricson and Nilsson 2016). So far, the use of uncemented TKAs has been limited. Previous studies have reported an increased risk for aseptic loosening of the tibial component in patients treated with uncemented TKA (Bassett 1998, Duffy et al. 1998, Berger et al. 2001a, Goldberg and Kraay 2004, Carlsson et al. 2005), but due to evolvement of designs and materials uncemented fixation has become an interesting choice, especially for younger patients with good bone quality (Hu et al. 2017). Trabecular metal (TM) has showed promising results in both register and clinical studies (Niemelainen et al. 2014, Henricson et al. 2013,

Acta Orthopaedica 2020; 91 (2): 184–190

All knee arthroplasties in the NARA database 2000–2016 n = 550,570 Excluded (n = 435,393): – patients ≥ 65 years, 366,151 – other than primary OA, 24,771 – UKAs, 17,764 – PS knees, 9,666 – performed before 2000, 6,631 – operated in 2017, 4,258 – other types of knees, 2,119 – degree of constrain unknown, 1,371 – revision knees, 1,306 – PFAs, 756 – type of implant unknown, 298 – fixation unknown, 149 – fully stabilized, 67 – patella unknown, 46 – other partial, 37 – sex unknown, 3 Included primary TKAs (n = 115,177): – cemented, 102,170 – uncemented, 6,132 – hybrids, 6,329 – inverse hybrids, 546

Figure 1. Flow chart.

Pulido et al. 2015). Although differences have been observed between different fixation concepts in terms of revision rates, functional outcomes have been equivalent irrespective of the fixation method (Gandhi et al. 2009, Gao et al. 2009, Demey et al. 2011, Arnold et al. 2013). The optimal fixation method in TKA still remains controversial for these younger patients. We assessed survivorships of 4 different fixation methods (cemented, uncemented, hybrid, and inverse hybrid) in patients younger than 65 years of age based on the Nordic Arthroplasty Register Association (NARA) database.

Patients and methods We included all uni- or bilateral unconstrained primary TKAs that had been implanted in patients aged less than 65 years for primary OA over 2000–2016 (Figure 1). Previous reports have shown that the effect of including bilateral cases in studies of hip and knee joint prosthesis survival is negligible (Robertsson and Ranstam 2003, Lie et al. 2004). The Swedish Knee Arthroplasty Register (SKAR), the Danish Knee Arthroplasty Register (DKR), the Norwegian Arthroplasty Register (NAR), and the Finnish Arthroplasty Register (FAR) participated in the study. The Nordic Arthroplasty Register Association (NARA) compiles data on 4 Nordic countries that have similar healthcare organizations and comparable patient characteristics (Robertsson et al. 2010). A NARA minimal dataset was created to contain data that all 4 registers could deliver (NARA report 2016). The NARA dataset includes 20 different main variables and in total 90 variables. All registers use individual-based registration of operations. Selection and transformation of the respective data sets and de-identification

185

of the patients, which included the deletion of personal identity numbers, were performed within each national register. The anonymous data were then merged into a common database. Data were treated with full confidentiality, according to the rules of the respective countries. The quality of data in the Nordic registers is high, including both 100% coverage and the following completeness: SKAR 97%, DKR 97%, NAR 97%, FAR 96% (NARA report 2016) (Espehaug et al. 2006). The fixation of TKAs was divided into 4 groups: (1) cemented, (2) uncemented, (3) hybrid (uncemented femur with cemented tibia), and (4) inverse hybrid (cemented femur with uncemented tibia). Statistics We assessed the descriptive statistics of the patients included. The inclusion time period was 2000–2016. We used Kaplan– Meier (KM) survival analysis to assess implant survival probability (with respective 95% confidence interval [CI]) of the TKA fixation at 10 and 15 years. The results in tables and figures were not shown when less than 40 knees were at risk. Outcome was defined as removal, addition, or exchange of at least one of the components, including polyethylene insert exchanges of modular tibial components, for any reason. We used Cox regression analysis to estimate hazard ratios associated with implant survival. Covariates included in the analysis were fixation type, sex, country, and age. Age was included as a continuous variable whereas the others were categorical. Correlation of scaled Schoenfeld residuals with time was examined to investigate violation of proportional hazard (PH) assumption. Log–log survival curves were also inspected visually to see if assumption was met. We detected multiple violations of PH assumption. In order to deal with PH violation, we used time-dependent coefficients (fixation, age, sex, and nation) using step function. Based on the log–log curves cut-offs were set as follows: 1, 3, and 6 years. We did stratified analyses based on age and implant brand group and similar time axis division was made according to log–log curves and residual testing. For the time dependent coefficients the data were broken into time-dependent parts according to the time intervals used in the time axis division. For each final analysis the PH test investigating Schoenfeld residuals was performed. Statistical analyses were performed using R 3.5.2, survival package (R Foundation for Statistical Computing, Vienna, Austria). Ethics, data sharing, funding, and potential conflicts of interest Ethical approval for the study was obtained through the ethical approval process of each national registry: the Ethics Board of Lund University (LU20-02) (Sweden), the National Institute of Health and Welfare (Dnro THL/1743/5.05.00/2014) (Finland), the Norwegian Data Inspectorate (ref 24.1.2017: 16/01622-3/CDG) (Norway) and the Danish Data protection agency (1-16-02-54-17) (Denmark).

186

Acta Orthopaedica 2020; 91 (2): 184–190

Table 1. Demographic data

Count 10,000

Uncemented

Fixation concept Inverse hybrid Hybrid

9,000

Cemented

No of TKAs (%) 6,132 (5.3) 546 (0.5) 6,329 (5.5) 102,170 (88.7) Mean age, years (SD) 57 (5.6) 57 (5.4) 58 (5.2) 59 (4.9) Men, % 44 42 41 40 Country, n of TKAs (%) Finland 900 (2.5) 350 (1.0) 146 (0.4) 34,406 (96) Norway 1,191 (8.7) 10 (0.1) 1,981 (14) 10,565 (77) Sweden 2,284 (5.0) 128 (0.3) 74 (0.2) 43,268 (95) Denmark 1,757 (8.8) 58 (0.3) 4,128 (21) 13,931 (70)

8,000

Cementented Uncemented Inverse hybrid Hybrid

7,000 6,000 5,000 4,000 3,000 2,000 1,000 0

2000 2002 2004 2006 2008 2010 2012 2014 2016

Year of primary operation

Table 4. Unadjusted Kaplan–Meier (KM) 10- and 15-year survival rates (%) with 95% confidence intervals (CI) for uncemented, inverse hybrid, hybrid, and cemented TKA

Figure 2. Number of operations.

Type of No. of 10-year K–M survivorship 15-year K–M survivorship fixation knees revisions n at risk rate (CI) n at risk rate (CI) Uncemented 6,132 363 Inverse hybrid 546 16 Hybrid 6,329 330 Cemented 102,170 5 ,040

915 66 1,349 24,954

91.2 (90.1–92.2) 96.0 (94.1–98.1) 93.0 (92.2–93.8) 93.6 (93.4–93.8)

214 – 239 4,259

88.7 (87.0–90.4) – 91.4 (90.2–92.6) 91.3 (91.0–91.7)

No funding was received. Authors did not have any conflicts of interest. Data sharing is not possible.

Results The mean follow-up time standard deviation (SD) was 6.4 (4.3) years for cemented TKA, 4.7 (3.4) years for uncemented TKA, 6.0 (4.3) years for hybrid TKA, and 6.1 (3.2) years for inverse hybrid TKA. There were slight differences in the proportion of men between the fixation groups, ranging from 40% in the cemented to 44% in the uncemented group (Table 1). TKA models varied between countries without a common trend and the most commonly used TKA models in the participating countries are given in Table 2 (see Supplementary data). Nexgen, PFC, and Triathlon were the most commonly used models within the fixation concepts (Table 3, see Supplementary data). The number of TKAs performed annually grew substantially over 2000–2009, and remained rather stable after that; cemented fixation was used in the vast majority of TKAs over the whole study period (Figure 2). Altogether, cemented fixation was used in 89% of all TKAs, and uncemented in 5.3%, hybrid in 5.5%, and inverse hybrid in 0.5%, respectively. The patella was resurfaced in 24,487 TKAs (21%) and uncemented patellar buttons were used in only 151 (0.1%) TKAs. In the subgroup of Nexgen TKAs, the patella was resurfaced in 5,821 (22%) TKAs, and an uncemented patellar button was used only in 2 knees (both of them in the cemented Nexgen group).

Figure 3. Unadjusted Kaplan–Meier cumulative risk of revision by fixation type in patients < 65 years of age.

At 15 years, KM-based survival rates were: cemented 91.3% (Cl 91.0–91.7), hybrid 91.4% (CI 90.2–92.6), uncemented 88.7% (CI 87.0–90.4). For inverse hybrid only 10-year survival was available (96.0% [CI 94.1–98.1]) (Table 4, Figure 3). In the Cox regression analysis, uncemented fixation showed an increased risk of revision compared with the reference group (cemented TKA) both during the first postoperative year and also after 6 years of follow-up. Hybrid fixation was associated with a decreased risk of revision compared with the cemented fixation after 6 years of follow-up. The risk of revision was similar between the inverse hybrid and the reference group (Table 5). Because of the age dependence of TKA survivorship, the additional Cox regression analyses were conducted for 2 different age groups: 55–64 years of age (Table 6) and less than 55 years of age (Table 7). In patients aged 55–64 years, risk of revision with uncemented TKAs was increased in comparison with the cemented reference group during the first 3 years of follow-up and after that similar compared with reference. Hybrid TKAs still showed a decreased risk of revision after 6 years of follow-up, a finding that was already seen

Acta Orthopaedica 2020; 91 (2): 184–190

187

Table 5. Cox regression with time-dependent coefficients (all patients aged < 65 years included, cemented TKA as reference) Type of fixation

Follow-up (years)

Uncemented < 1 1–3 3–6 > 6 Inverse hybrid < 1 1–3 3–6 > 6 Hybrid < 1 1–3 3–6 > 6 Cemented

Table 7. Cox regression with timedependent coefficients in patients aged < 55 years

Hazard ratio (95% CI)

Type of fixation Uncemented Inverse hybrid Hybrid Cemented

1.38 (1.13–1.70) 1.14 (0.97–1.35) 0.95 (0.72–1.25) 1.32 (1.00–1.73) 0.29 (0.07–1.16) 0.67 (0.34–1.35) 0.91 (0.38–2.19) 0.54 (0.13–2.15) 1.11 (0.88–1.39) 0.94 (0.78–1.12) 1.07 (0.82–1.40) 0.54 (0.38–0.78) 1.0 Reference

Follow-up (years)

Uncemented < 1.5 1.5–3 3–6 > 6 Inverse hybrid < 1.5 1.5–3 3–6 > 6 Hybrid < 1.5 1.5–3 3–6 > 6 Cemented

Hazard ratio (95% CI) 1.37 (1.13–1.67) 1.31 (1.01–1.69) 0.86 (0.59–1.24) 1.32 (0.96–1.83) 0.44 (0.14–1.37) 0.65 (0.21–2.02) 0.88 (0.28–2.75) 0.49 (0.07–3.48) 1.15 (0.94–1.41) 0.90 (0.68–1.20) 1.14 (0.85–1.53) 0.55 (0.37–0.83) 1.0 Reference

Discussion

in the whole study cohort (Table 5). In patients aged less than 55 years, revision risks were similar between fixation methods (Table 7). Differences between age, sex, and country were the other covariates in the Cox regression analysis and their results are listed in Table 8 (see Supplementary data).

We found that both cemented and hybrid TKAs showed 10-year survival rates exceeding 92–>93% in patients aged < 65 years. Even though hybrid/inverse hybrid versions of the well-performing contemporary TKA designs provided younger patients with a good mid-term outcome in our study, they were still used in a limited number of patients. And especially in the inverse hybrid group, one single TKA design with a very good track record comprised the vast majority of the whole group. It is thus safe to conclude that cemented TKA still fulfils the most important task of a TKA: it works very

Table 10. Unadjusted Kaplan–Meier 7- and 10-year survival rates with 95% confidence intervals for uncemented, inverse hybrid, hybrid, and cemented TKA in the Nexgen subgroup Type of No. of 7-year K–M survivorship fixation knees revisions n at risk rate (CI)

1.10 (0.91–1.32 0.62 (0.29–1.29 0.83 (0.67–1.04) 1.0 Reference

The inverse hybrid group mainly comprised Nexgen TKAs (95% of the knees) (Table 3, see Supplementary data), and approximately more than 80% of the inverse hybrid Nexgen TKAs used TM monoblock tibial components (an estimate from national registers’ data). Because of the obvious risk for selection bias, we conducted an additional sensitivity analysis to diminish bias between groups. For this analysis, we included only patients operated on with Nexgen TKAs (Table 9, see Supplementary data). In this sensitivity analysis, survival rates of different fixations were in descending order: the inverse hybrid 96.6% (CI 94.7–98.5), cemented 95.8% (CI 95.5–96.1), uncemented 93.2% (CI 91.9–94.6), and hybrid 92.0% (CI 90.4–93.7) at 7 years (Table 10). In the Cox analysis of the Nexgen subgroup, increased risk of revision was found for uncemented and hybrid TKAs compared with cemented TKAs, and for inverse hybrid TKAs the risk of revision was comparable to cemented TKAs (Table 11).

Table 6. Cox regression with time-dependent coefficients in patients aged 55–65 years Type of fixation

Hazard ratio (95% CI)

10-year K–M survivorship n at risk rate (CI)

Uncemented 2,311 114 238 93.2 (91.9–94.6) – – Inverse hybrid 497 13 185 96.6 (94.7–98.5) 55 96.6 (94.7–98.5) Hybrid 1,629 91 155 92.0 (90.4–93.7) – – Cemented 27,934 901 8,477 95.8 (95.5–96.1) 3,691 94.9 (94.6–95.3)

Table 11. Cox regression with timedependent coefficients in patients aged < 65 years in the Nexgen subgroup Type of fixation Uncemented Inverse hybrid Hybrid Cemented

Hazard ratio (95% CI) 1.37 (1.12–1.67) 0.59 (0.34–1.03) 1.47 (1.16–1.87) 1.0 Reference

188

reliably in the hands of many. Also, cemented TKA should still be considered as the gold standard in TKA of all OA patients irrespective of their age. We acknowledge certain strengths and limitations in our study. The major strength of our study is the unique collaboration of 4 national registers in the creation of a multinational database comprising a high number of patients. This NARA database enables international comparisons to reveal possible differences in trends and outcomes of TKA. To our knowledge, this is the first multi-national, register-based study comparing the outcomes of all 4 fixation methods in TKA. There are also a few obvious limitations in our study. First, there were clearly fewer patients in the alternative fixation groups as compared with the cemented reference group (Figure 2). There are potential sources of selection bias in our data. Other concepts than cemented TKAs may have been done in higher volume units, and there may have been less preoperative bone loss or less severe deformity. On the other hand, uncemented components may have been used in patients with higher demands and also there may have been concerns about cemented fixation during operation. If the choice of fixation had been constant at hospital level in our study population, this might lower this risk of bias. Further, especially inverse hybrid fixation, but also hybrid fixation to some extent, had another obvious advantage over cemented fixation in our study setting. That is the monoblock uncemented tibial component, since wash-out procedures for infection in such knees (without exchange of any component) have not been regarded as revisions in the NARA data. Thus, due to a small number of patients and also the possibility of some missing infection revisions, the results of inverse hybrids should be interpreted with caution. Further, Nexgen TKAs comprised 91% of the inverse hybrid group. This implant has been reported to have 97–99% 10-year survival rate in previous studies (Kim et al. 2012, Niemelainen et al. 2014, Robertsson et al. 2020). Further, in Finland Nexgen inverse hybrid TKAs (with TM tibial component) have been performed in only 3 hospitals, 1 of which is a high-volume specialized center (Niemelainen et al. 2014). In the hybrid group, 3 TKA designs with a very good track record (PFC, Nexgen, Profix) comprised 76% of all TKAs. The second limitation is that, due to the nature of the NARA dataset, we had a limited number of covariates for analysis and also we did not have exact information on whether some of the uncemented implants were hydroxide apatite coated or not. On the hip side, HA coating does not seem to provide any added value in terms of improved survival rates (Hailer et al. 2015, Lazarinis et al. 2017), thus it most probably does not cause any bias to these TKA results. In our study, the vast majority of TKAs performed for younger patients in the 4 Nordic reporting countries were still cemented, and very small changes, if any, were observed in the fixation methods used over the study period (Figure 2). The same trend in general has also been reported from other national registers: the annual report 2017 of the National Joint Registry

Acta Orthopaedica 2020; 91 (2): 184–190

for England, Wales and Northern Ireland (NJR) reported that the proportion of all cemented TKA implants increased from 82% in 2003 to 87% in 2016 (NJR annual report 2017). During the same time period uncemented implants decreased from 6.7% to 2.0% and hybrid implants from 2.8% to 0.4%. The same increasing trend of using cemented implants was seen in the Australian Joint Registry (AOANJRR annual report 2017). In our study, the proportion of cemented TKAs decreased only slightly from 96% in 2000 to 91% in 2016, and a simultaneous small increase in usage of uncemented TKAs was seen (from 2.5% to 6.5%, respectively). In our study, both cemented and hybrid TKAs had up to 15-year survival rates exceeding 91% in patients aged < 65 years. Hybrid TKAs showed decreased risk of revision in comparison with cemented TKA after 6 years of follow-up. Inverse hybrid TKAs showed 96% survivorship at 10 years. Uncemented TKAs had the worst 10-year survival rate (91%). These findings are in line with previous literature. In a Finnish register-based study, uncemented TKAs had 1.4 times elevated adjusted hazard ratios (HR) for revision for any reason compared with cemented TKAs (Julin et al. 2010). In the AOANJRR annual report in 2017, the cumulative 10-year revision probability of minimally stabilized TKA was 4.5 (4.3–4.6) with cemented TKA, 6.1 (5.9–6.3) with uncemented TKA, and 4.6 (4.4–4.7) with hybrid TKAs. In the New Zealand Joint Register annual report in 2017, the revision rate with patient 55–64 years old was the highest with an uncemented implant: 0.84 (CI 0.67–1.05)/100 component-years compared with 0.62 (CI 0.58–0.66)/100 component-years with cemented implants and 0.61 (CI 0.47–0.77)/100 component-years with hybrid implants. To our knowledge, this study is the first to compare the survivorships of all 4 different fixation concepts in TKA. In theory, younger patients might benefit from biologic fixation, i.e., bone ingrowth into uncemented implants. A metaanalysis (Gandhi et al. 2009) based on 5 RCTs and 10 observational studies, with different mean ages of patients and with a minimum follow-up of 2 years, found improved survival for cemented compared with uncemented implants when revision for aseptic loosening was used as an endpoint. Another systematic review and meta-analysis (Voigt and Mosier 2011) compared hydroxyapatite-coated, porous coated, and cemented tibial components. Evidence of more stable fixation after 2 years with hydroxyapatite-coated components compared with porous-coated and cemented implants was found, but revision rates at 10 year follow-up were similar. In an RCT no revision rates and survival were similar between the cemented and uncemented TKAs with mean follow-up of 15 years (Baker et al. 2007). In a systematic review of 11 RCTs to identify whether there was an association between fixation method and clinical outcome, it was found that short- and long-term outcomes were not influenced by fixation type (Arnold et al. 2013). In previous studies, early failures of uncemented TKAs were mainly caused by aseptic loosening of the patellar button and the tibial component (Collins et al. 1991, Bassett 1998,

Acta Orthopaedica 2020; 91 (2): 184–190

Duffy et al. 1998, Berger et al. 2001b, Barrack et al. 2004, Goldberg and Kraay 2004, Carlsson et al. 2005). Uncemented fixation has been associated with a high failure rate due to inadequate bone ingrowth in TKAs (Lombardi et al. 2007). As stated earlier, Nexgen comprised 95% of the TKAs in the inverse hybrid group, and 87% of these Nexgen TKAs had been used with TM tibial components, which are known to have good results (Niemelainen et al. 2014). We tried to tackle the obvious possibility of selection bias by conducting a sensitivity analysis including only patients with Nexgen TKAs (Tables 10 and 11). In that analysis, it appeared that there was no statistically significant difference in mid-term survival rates or Cox-adjusted revision risks between inverse hybrid and cemented Nexgen TKAs. Further, hybrid and uncemented fixation showed an increased risk for revision in this Nexgen subgroup. Thus, the more expensive uncemented/hybrid/ inverse hybrid versions did not provide these younger patients with any advantage over cemented fixation in the 10-year follow-up of Nexgen TKAs. To conclude, cemented TKA still deserves the status of gold standard in TKA irrespective of the patients’ age. In addition to age, the optimal fixation method in younger patients may also be influenced by patients’ other characteristics such as level of activity, anatomy, or bone quality. Even though hybrid/inverse hybrid versions of the well-performing contemporary TKA designs provided younger patients with a good mid-term outcome in our study, these results do not support systematic use of these more expensive components in TKA for younger patients. Supplementary data Tables 2, 3, 8, and 9 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2019.1710373

Study design: MN, AE. Analysis of data and statistics: AR, AE, MN. Review and interpretation of the results: MN, KM, OR, AW-D, OF, AF, AP, HS, AR, AE. Revision and approval of the final manuscript: MN, KM, OR, AW-D, OF, AF, AP, HS, AR, AE. Acta thanks Alexander Liddle and Ola Rolfson for help with peer review of this study.

AOANJRR annual report 2017. Available from https://aoanjrr.sahmri.com/ annual-reports-2017. Arnold J B, Walters J L, Solomon L B, Thewlis D. Does the method of component fixation influence clinical outcomes after total knee replacement? A systematic literature review. J Arthroplasty 2013; 28(5): 740-6. Baker P N, Khaw F M, Kirk L M, Esler C N, Gregg P J. A randomised controlled trial of cemented versus cementless press-fit condylar total knee replacement: 15-year survival analysis. J Bone Joint Surg Br 2007; 89(12): 1608-14. Barrack R L, Nakamura S J, Hopkins S G, Rosenzweig S. Winner of the 2003 James A. Rand young investigator’s award: Early failure of cementless mobile-bearing total knee arthroplasty. J Arthroplasty 2004; 19(7 Suppl. 2): 101-6.

189

Bassett R W. Results of 1,000 performance knees: cementless versus cemented fixation. J Arthroplasty 1998; 13(4): 409-13. Berger R A, Lyon J H, Jacobs J J, Barden R M, Berkson E M, Sheinkop M B, Rosenberg A G, Galante J O. Problems with cementless total knee arthroplasty at 11 years followup. Clin Orthop Relat Res 2001a; (392): 196-207. Berger R A, Rosenberg A G, Barden R M, Sheinkop M B, Jacobs J J, Galante J O. Long-term followup of the Miller-Galante total knee replacement. Clin Orthop Relat Res 2001b; (388): 58-67. Carlsson A, Bjorkman A, Besjakov J, Onsten I. Cemented tibial component fixation performs better than cementless fixation: a randomized radiostereometric study comparing porous-coated, hydroxyapatite-coated and cemented tibial components over 5 years. Acta Orthop 2005; 76(3): 362-9. Carr A J, Robertsson O, Graves S, Price A J, Arden N K, Judge A, Beard D J. Knee replacement. Lancet 2012; 379(9823): 1331-40. Collins D N, Heim S A, Nelson C L, Smith P 3rd. Porous-coated anatomic total knee arthroplasty: a prospective analysis comparing cemented and cementless fixation. Clin Orthop Relat Res 1991; (267): 128-36. Dalury D F. Cementless total knee arthroplasty: current concepts review. Bone Joint J 2016; 98-B(7): 867-73. Demey G, Servien E, Lustig S, Ait Si Selmi T, Neyret P. Cemented versus uncemented femoral components in total knee arthroplasty. Knee Surg Sports Traumatol Arthrosc 2011; 19(7): 1053-9. Duffy G P, Berry D J, Rand J A. Cement versus cementless fixation in total knee arthroplasty. Clin Orthop Relat Res 1998; (356): 66-72. Espehaug B, Furnes O, Havelin L I, Engesaeter L B, Vollset S E, Kindseth O. Registration completeness in the Norwegian Arthroplasty Register. Acta Orthop 2006; 77(1): 49-56. Franceschetti E, Torre G, Palumbo A, Papalia R, Karlsson J, Ayeni O R, Samuelsson K, Franceschi F. No difference between cemented and cementless total knee arthroplasty in young patients: a review of the evidence. Knee Surg Sports Traumatol Arthrosc 2017; 25(6): 1749-56. Gandhi R, Tsvetkov D, Davey J R, Mahomed N N. Survival and clinical function of cemented and uncemented prostheses in total knee replacement: a meta-analysis. J Bone Joint Surg Br 2009; 91(7): 889-95. Gao F, Henricson A, Nilsson K G. Cemented versus uncemented fixation of the femoral component of the NexGen CR total knee replacement in patients younger than 60 years: a prospective randomised controlled RSA study. Knee 2009; 16(3): 200-6. Goldberg V M, Kraay M. The outcome of the cementless tibial component: a minimum 14-year clinical evaluation. Clin Orthop Relat Res 2004; (428): 214-20. Hailer N P, Lazarinis S, Makela K T, Eskelinen A, Fenstad A M, Hallan G, Havelin L, Overgaard S, Pedersen A B, Mehnert F, Karrholm J. Hydroxyapatite coating does not improve uncemented stem survival after total hip arthroplasty! Acta Orthop 2015; 86(1): 18-25. Henricson A, Nilsson K G. Trabecular metal tibial knee component still stable at 10 years. Acta Orthop 2016; 87(5): 504-10. Henricson A, Rosmark D, Nilsson K G. Trabecular metal tibia still stable at 5 years: An RSA study of 36 patients aged less than 60 years. Acta Orthop 2013; 84(4): 398-405. Hu B, Chen Y, Zhu H, Wu H, Yan S. Cementless porous tantalum monoblock tibia vs cemented modular tibia in primary total knee arthroplasty: a metaanalysis. J Arthroplasty 2017; 32(2): 666-74. Julin J, Jamsen E, Puolakka T, Konttinen Y T, Moilanen T. Younger age increases the risk of early prosthesis failure following primary total knee replacement for osteoarthritis: a follow-up study of 32,019 total knee replacements in the Finnish Arthroplasty Register. Acta Orthop 2010; 81(4): 413-9. Kim Y H, Park J W, Kim J S. High-flexion total knee arthroplasty: survivorship and prevalence of osteolysis. Results after a minimum of ten years of follow-up. J Bone Joint Surg Am 2012; 94(15): 1378-84. Lazarinis S, Makela K T, Eskelinen A, Havelin L, Hallan G, Overgaard S, Pedersen A B, Karrholm J, Hailer N P. Does hydroxyapatite coating of uncemented cups improve long-term survival? An analysis of 28,605 primary total hip arthroplasty procedures from the Nordic Arthroplasty Register Association (NARA). Osteoarthritis Cartilage 2017; 25(12): 1980-7.

190

Leskinen J, Eskelinen A, Huhtala H, Paavolainen P, Remes V. The incidence of knee arthroplasty for primary osteoarthritis grows rapidly among baby boomers: a population-based study in Finland. Arthritis Rheum 2012; 64(2): 423-8. Lie S A, Engesaeter L B, Havelin L I, Gjessing H K, Vollset S E. Dependency issues in survival analyses of 55,782 primary hip replacements from 47,355 patients. Stat Med 2004; 23(20): 3227-40. Lombardi A V Jr, Berasi C C, Berend K R. Evolution of tibial fixation in total knee arthroplasty. J Arthroplasty 2007; 22(4 Suppl. 1): 25-9. Meehan J P, Danielsen B, Kim S H, Jamali A A, White R H. Younger age is associated with a higher risk of early periprosthetic joint infection and aseptic mechanical failure after total knee arthroplasty. J Bone Joint Surg Am 2014; 96(7): 529-35. Miller A J, Stimac J D, Smith L S, Feher A W, Yakkanti M R, Malkani A L. Results of cemented vs cementless primary total knee arthroplasty using the same implant design. J Arthroplasty 2018; 33(4): 1089-93. NARA report 2016. Available from http://www.nordicarthroplasty.org. Nemes S, Rolfson O, W-Dahl A, Garellick G, Sundberg M, Karrholm J, Robertsson O. Historical view and future demand for knee arthroplasty in Sweden. Acta Orthop 2015; 86(4): 426-31. Niemelainen M, Skytta E T, Remes V, Makela K, Eskelinen A. Total knee arthroplasty with an uncemented trabecular metal tibial component: A registry-based analysis. J Arthroplasty 2014; 29(1): 57-60. Niemelainen M J, Makela K T, Robertsson O, W-Dahl A, Furnes O, Fenstad A M, Pedersen A B, Schroder H M, Huhtala H, Eskelinen A. Different incidences of knee arthroplasty in the Nordic countries. Acta Orthop 2017; 88(2): 173-8.

Acta Orthopaedica 2020; 91 (2): 184â&#x20AC;&#x201C;190

NJR annual report 2017. Available from https://reports.njrcentre.org.uk. Pulido L, Abdel M P, Lewallen D G, Stuart M J, Sanchez-Sotelo J, Hanssen A D, Pagnano M W. The mark Coventry award: Trabecular metal tibial components were durable and reliable in primary total knee arthroplasty: a randomized clinical trial. Clin Orthop Relat Res 2015; 473(1): 34-42. Robertsson O, Ranstam J. No bias of ignored bilaterality when analysing the revision risk of knee prostheses: analysis of a population based sample of 44,590 patients with 55,298 knee prostheses from the national Swedish Knee Arthroplasty Register. BMC Musculoskelet Disord 2003; 4: 1. Robertsson O, Bizjajeva S, Fenstad A M, Furnes O, Lidgren L, Mehnert F, Odgaard A, Pedersen A B, Havelin L I. Knee arthroplasty in Denmark, Norway and Sweden: a pilot study from the Nordic Arthroplasty Register Association. Acta Orthop 2010; 81(1): 82-9. Robertsson O, Sundberg M, Sezgin E A, Lidgren L, W-Dahl A. Higher risk of loosening for a four-pegged TKA tibial baseplate than for a stemmed one: a register-based study. Clin Orthop Relat Res 2020; 478(1): 58-65. Voigt J D, Mosier M. Hydroxyapatite (HA) coating appears to be of benefit for implant durability of tibial components in primary total knee arthroplasty. Acta Orthop 2011; 82(4): 448-59. Wilson D A, Richardson G, Hennigar A W, Dunbar M J. Continued stabilization of trabecular metal tibial monoblock total knee arthroplasty components at 5 years measured with radiostereometric analysis. Acta Orthop 2012; 83(1): 36-40. Zhou K, Yu H, Li J, Wang H, Zhou Z, Pei F. No difference in implant survivorship and clinical outcomes between full-cementless and full-cemented fixation in primary total knee arthroplasty: a systematic review and metaanalysis. Int J Surg 2018; 53: 312-19.

Acta Orthopaedica 2020; 91 (2): 191–196

191

Better implant survival with modern ankle prosthetic designs: 1,226 total ankle prostheses followed for up to 20 years in the Swedish Ankle Registry Alexandra UNDÉN 1,2, Lars JEHPSSON 2,3, Ilka KAMRAD 2,3, Åke CARLSSON 2,3, Anders HENRICSON 4, Magnus K KARLSSON 2,3, and Björn E ROSENGREN 2,3 1 Department

of Radiology, Skåne University Hospital, Malmö; 2 Department of Clinical Sciences Malmo (IKVM), Lund University; 3 Department of Orthopedics, Skåne University Hospital, Malmö; 4 Department of Orthopedics, Falu Central Hospital and Center of Clinical Research Dalarna, Falun, Sweden Correspondence: alexandra.unden@med.lu.se Submitted 2019-05-31. Accepted 2019-11-12.

Background and purpose — We have previously reported on the prosthetic survival of total ankle replacements (TAR) in Sweden performed between 1993 and 2010. Few other reports have been published on 5- and 10-year survival rates. Furthermore, there is a lack of long-term outcome data on modern prosthetic designs. Therefore, we compared early and current prosthetic designs after a mean 7-year follow-up. Patients and methods — On December 31, 2016, 1,230 primary TARs had been reported to the Swedish Ankle Registry. We analyzed prosthetic survival, using exchange or permanent extraction of components as endpoint for 1,226 protheses with mean follow-up of 7 years (0–24). Differences between current (Hintegra, Mobility, CCI, Rebalance, and TM Ankle) and early prosthetic designs (STAR, BP, and AES) were examined by log rank test. Results — 267/1,226 prostheses (22%) had been revised by December 31, 2016. We found an overall prosthetic survival rate at 5 years of 0.85 (95% CI 0.83–0.87), at 10 years 0.74 (CI 0.70–0.77), at 15 years 0.63 (CI 0.58–0.67), and at 20 years 0.58 (CI 0.52–0.65). For early prosthetic designs the 5- and 10-year survival rates were 0.81 (CI 0.78–0.84) and 0.69 (CI 0.64-0.73) respectively, while the corresponding rates for current designs were 0.88 (CI 0.85– 0.91) and 0.84 (CI 0.79–0.88). Current prosthetic designs had better survival (log rank test p < 0.001). Interpretation — Our results point to a positive time trend of prosthetic survival in Sweden; use of current prosthetic designs was associated with better prosthetic survival. Improved designs and instrumentation, more experienced surgeons, and improved patient selection may all have contributed to the better outcome.

Uncemented, 3-component total ankle replacements (TAR) have shown promising but somewhat varying results in medium- and long-term reports (Wood and Deakin 2003, Fevang et al. 2007, Skytta et al. 2010, Bonnin et al. 2011, Henricson et al. 2011b, Mann et al. 2011, Tomlinson and Harrison 2012, Barg et al. 2013, Zaidi et al. 2013, Henricson and Carlsson 2015, Kerkhoff et al. 2016, Frigg et al. 2017, Palanca et al. 2018, Clough et al. 2019). Some evidence points to better results with modern prosthetic designs (Barg et al. 2015, Koivu et al. 2017a, Clough et al. 2019). National registries give better insight into current real-world results, including more patients and different surgeons, than data from single surgeons or institutions. 5-year survival rates of between 0.78 and 0.89 have been reported by national registries from Finland, New Zealand, Norway, and Sweden (Fevang et al. 2007, Henricson et al. 2007, Hosman et al. 2007, Skytta et al. 2010, Henricson et al. 2011b). Results beyond 5 years are uncertain. Here we present medium and longer-term follow-up (up to 20 years) of TARs reported to the Swedish Ankle Registry (http:// www.swedankle.se), and compare early and current designs.

Patients and methods Since 1993, Swedish hospitals performing TARs have reported information on date of index surgery and any revision surgery, including data on the patient and the procedure, to the Swedish Ankle Registry. The current procedure-based coverage and completeness are both estimated at close to 100%. Until December 31, 2016, 1,230 primary TARs (all uncemented, 3-component designs) had been registered in 1,132 patients with mean annual numbers of 51 (6–87). The mean

192

Acta Orthopaedica 2020; 91 (2): 191–196

follow-up time was 7 years (0–24). 4 cases lost to follow-up were not included in the survival analyses. Early prosthetic designs Current prosthetic designs 546 of the remaining 1,226 cases we Year STAR a BP b AES c Hintegra d Mobility e CCI f Rebalance g TM Ankle h Total refer to as “early prosthetic designs” (STAR, BP, and AES) as these 1993 6 6 1994 13 13 designs have not been implanted in 1995 11 11 Sweden since 2008; the other 680 1996 8 8 we refer to as “current prosthetic 1997 24 24 1998 34 34 designs” (Hintegra, Mobility, CCI, 1999 25 25 Rebalance, and TM Ankle) (Table 2000 45 1 46 1). Revision was defined as removal 2001 48 48 2002 39 21 3 10 73 or exchange of 1 or more of the pros2003 25 23 17 14 79 thetic components with the exception 2004 18 29 16 4 67 of incidental exchange (exchange 2005 18 13 23 2 9 65 2006 8 11 21 6 21 67 of the polyethylene insert during 2007 1 7 18 20 46 a secondary procedure undertaken 2008 4 17 32 16 69 because of a different indication) of 2009 31 42 73 2010 41 22 63 the polyethylene insert (Henricson 2011 44 22 21 87 et al. 2011a). We chose to analyze 2012 23 31 27 81 the number of prostheses rather than 2013 32 12 34 78 2014 15 4 37 5 61 the number of patients (including 96 2015 3 1 2 36 11 53 bilateral cases), in line with our pre2016 9 23 21 53 vious study (Henricson et al. 2011b) Total 323 109 115 48 269 151 178 37 1,230 as this approach has been found to a Scandinavian Total Ankle Replacement (Waldemar LINK GmBH, Hamburg, Germany) have a negligible effect on the surb Buechel–Pappas (Wright Cremascoli, Toulon, France) vival estimates (Ranstam and Robc Ankle Evolutive System (Biomet, Valence, France) d Hintegra (Newdeal SA, Lyon, France) ertsson 2010). e Mobility (DePuy International, Leeds, UK) Since 1993, TAR has been perf CCI–Ceramic Coated Implant (Wright Medical Technology, Arlington) formed in 25 hospitals in Sweden g Rebalance (Biomet, Bridgend, UK) h Trabecular Metal Total Ankle (Zimmer inc, Warsaw, Indiana, USA) by 43 surgeons. In the total cohort, 60% were women and the mean age at primary TAR was 60 years (18– Table 2. Data on distribution of diagnoses, sex and age per 88). The most common primary diagnoses leading to surgery prosthetic design group were posttraumatic arthritis in one-third of patients and rheumatoid arthritis in one-third (Table 2). Table 1. Distribution of TARs implanted in Sweden during 1993–2016 by year of implantation and prosthetic design

Design group Diagnosis n (%)

women Age (%) mean (SD) [range]

All prosthetic designs Posttraumatic arthritis 443 (36) 53 Rheumatoid arthritis 401 (32) 81 Osteoarthritis 291 (24) 46 Other a 95 (8) 40 All diagnoses 1,230 60 Current prosthetic designs Posttraumatic arthritis 268 (39) 52 Rheumatoid arthritis 182 (27) 87 Osteoarthritis 155 (23) 45 Other a 78 (11) 37 All diagnoses 683 58 Early prosthetic designs Posttraumatic arthritis 175 (32) 54 Rheumatoid arthritis 219 (40) 77 Osteoarthritis 136 (25) 48 Other a 17 (3) 53 All diagnoses 547 62 a

60 (12) [25–86] 56 (14) [18–85] 64 (11) [30–88] 59 (12) [28–76] 60 (13) [18–88] 62 (12) [30–85] 57 (14) [18–83] 67 (10) [37–88] 60 (11) [34–76] 62 (12) [18–88] 56 (12) [25–86] 56 (14) [21–85] 61 (10) [30–84] 53 (15) [28–74] 57 (13) [21–86]

Including hemophilia, hemochromatosis and psoriatic arthritis

Statistics To visualize differences in prosthetic survival rate, we used Kaplan–Meier estimator. Differences between current and early prosthetic designs were examined by log rank test. Data are reported as numbers and proportions (%), mean (SD), or median (range). We considered a probability of less than 5% as statistically significant and used 95% confidence intervals (CI) within parentheses to describe uncertainty. Even though our study may not be considered sample based we have chosen to present measure of uncertainty in order to facilitate generalization to probable future outcomes in Sweden and to other similar populations. We used SPSS version 23 (IBM Corp, Armonk, NY, USA). Ethics, data sharing, funding, and potential conflicts of interest The study protocol was approved by the Ethical Review Board

Acta Orthopaedica 2020; 91 (2): 191–196

193

Table 3a. Information on reasons for revision for early prosthetic designs Reasons STAR for revision n = 323 Talus and/or tibia loosening 65 Technical error 17 a Instability 1 Infection 16 Intractable pain 10 PE failure 22 Painful varus 2 Painful valgus Fracture/dislocation 3 Other Total number of revisions (%) 136 (42)

BP n = 109

AES n = 115

Total (%) n = 547

7 2 3 1 1 2 2 1 3

16 3 4 1 2 2 4 2 5

88 (45) 19 (10) 7 (4) 21 (11) 12 (6) 26 (13) 6 (3) 5 (3) 8 (4) 5 (3)

22 (20)

39 (34) 197 (36)

Proportion not revised – all designs 1.0

0.8

0.6

a The

inferior results of the STAR prosthesis is documented in previous reports from the Swedish Ankle Registry (Henricson and Carlsson 2015). The high frequency of “technical errors” may partly be explained by suboptimal instrumentation, limited experience of surgeons, and technical demands (Henricson et al. 2011b).

0.4

0.2

0.0

Years since surgery

Figure 1. Estimated cumulative prosthetic survival for all 1,226 TARs. Number of patients still at risk of experiencing the primary endpoint and prosthetic survival with 95% CI per 5-year period are indicated in the life table Years since surgery:

All prostheses (n = 1,226) Number at risk 707 325 75 5 TAR survival 0.85 0.74 0.63 0.58 CI 0.83–0.87 0.70–0.77 0.58–0.67 0.52–0.65

Table 3b. Information on reasons for revision for current prosthetic designs Reasons Hintegra Mobility CCI Rebalance TM ankle for revision n = 48 n = 269 n = 151 n = 178 n = 37 Talus and/or tibia loosening 4 8 18 5 Technical error 2 1 1 Instability 1 4 1 1 Infection 1 1 1 Intractable pain 6 4 1 PE failure 1 Painful varus 1 2 1 Painful valgus 1 1 Fracture/Lux 1 Other 2 1 Total number of revisions (%) 9 (19) 25 (9) 28 (19) 9 (5) 0 (0)

of Lund University (Dnr 2014/448). The study was conducted in accordance with the Helsinki Protocol. The registration of data and the study was performed confidentially on patient consent and according to Swedish and EU data protection rules. Data may be accessible upon application to the registry. This work was supported by grants from ALF and FoUU of Region Skåne, Greta Koch, Herman Järnhardt, Maggie Stevens, Skåne University Hospital foundations, and the Swedish Association of Local Authorities and Regions. The funders had no influence on the design of the study, the collection, analysis, and interpretation of data, on writing the manuscript, or on any other part of the study. The authors declare no conflict of interest.

Total (%) n = 683

Results

Of the 1,230 prostheses implanted since 1993, 22% had been revised by December 31, 2016. The most common reason for revision 35 (49) was loosening of the tibial and/or the talar 4 (6) 7 (10) component, responsible for about half of 3 (4) the revisions in both the early design group 11 (15) and current design group. PE (polyethylene) 2 (3) 4 (6) insert failure (13%) was the second most 2 (3) common reason for revision in the early 1 (1) design group, whereas this was only reported 2 (3) once in the current design group (Table 3). 71 (10) We found an overall TAR survival at 5 years of 0.85 (95% CI 0.83–0.87), at 10 years 0.74 (CI 0.70–0.77), at 15 years 0.63 (CI 0.58–0.67) and at 20 years 0.58 (CI 0.52–0.65) (Figure 1). For early and current designs 10-year TAR survival was 0.69 (CI 0.64–0.73) and 0.84 (CI 0.79–0.88) respectively (Figure 2). Log rank test revealed a statistically significant difference in TAR survival between early and current designs in favor of current designs (p < 0.001) (Figure 2). Analyses by specific prosthetic design revealed a 5-year TAR survival rate of 0.89 (CI 0.82–0.97) for the currently most implanted design in Sweden, Rebalance. At the time of writing, no revision of TM Ankle design had been reported to the registry (Figures 3 and 4). We found similar TAR survival rates (ranging from 0.57 to 0.69 at 15 years) for different diagnoses (Figure 5, see Supplementary data).

194

Acta Orthopaedica 2020; 91 (2): 191–196

Proportion not revised

1.0

0.8

0.6

0.4

0.2

BP AES STAR

Current designs Early designs 0.0

0.0

Years since surgery

Figure 2. Estimated cumulative prosthetic survival for early and current designs. Number of patients still at risk of experiencing the primary endpoint and prosthetic survival with 95% CI per 5-year period are indicated in the life table Years since surgery:

Early designs (n = 546) Number at risk 420 287 75 5 TAR survival 0.81 0.69 0.58 0.54 CI 0.78–0.84 0.64–0.74 0.53–0.63 0.47–0.60 Current designs (n = 680) Number at risk 287 38 TAR survival 0.88 0.84 CI 0.85–0.91 0.79–0.88

Discussion Using the Swedish Ankle Registry, we have demonstrated that survival of current TAR prostheses is higher than for earlier designs. In previous reports from our national registry the 5-year TAR survival rates were 0.78 for procedures made from 1993 to 2005 (Henricson et al. 2007) and 0.81 for 1993 to 2010 (Henricson et al. 2011b) respectively. In the current report we found the corresponding survival to be 0.85, indicating an improvement over time. This notion is further supported by 10-year TAR survival of 0.74 in the current study (Table 4, see Supplementary data) and 0.69 in the previous (Henricson et al. 2011b) (procedures undertaken from 1993 to 2010). In Sweden today, fewer units and surgeons perform TAR compared with earlier, resulting in larger volumes per surgeon. This and the presumed growing experience of these surgeons associated with time, may have contributed to the improving results. Reports from other national registries, published between 2007 and 2017, have presented 5-year survival rates between 0.78 and 0.89 and 10-year rates between 0.69 and 0.83, thus comparable to our findings, although the definition of revision is not always identical to our report (Table 4, see Supplementary data). In comparison with nationwide registries, results from single-center specialized units are often better (Labek et al. 2011), but have seldom been reproduced in countries where national registry data are available. Several such studies report 15-year survival rates between 0.64 and 0.76 (Table 4). This may be due to larger volumes of TARs per surgeon and also

Figure 3. Estimated cumulative prosthetic survival for early designs. Number of patients still at risk of experiencing the primary endpoint and prosthetic survival with 95% CI per 5-year period are indicated in the life table. Years since surgery: STAR (n = 322) Number at risk TAR survival CI BP (n = 109) Number at risk TAR survival CI AES (n = 115) Number at risk TAR survival CI

240 177 74 1 0.79 0.65 0.54 0.50 0.74–0.83 0.59–0.70 0.48–0.60 0.43–0.57 88 67 1 0.83 0.80 0.75 0.76–0.90 0.72–0.88 0.63–0.87 92 43 0.82 0.64 0.75–0.89 0.55–0.74

Proportion not revised 1.0

0.8

0.6

0.4 TM Ankle Mobility Rebalance CCI Hintegra

0.2

0.0

Years since surgery

Figure 4. Estimated cumulative prosthetic survival for current designs. Number of patients still at risk of experiencing the primary endpoint and prosthetic survival with 95% CI per 5-year period are indicated in the life table. Years since surgery: Hintegra (n = 46) Mobility (n = 269) CCI (n = 151) Rebalance (n = 177) TM Ankle (n = 37)

Number at risk TAR survival (CI) Number at risk TAR survival (CI) Number at risk TAR survival (CI) Number at risk TAR survival (CI) Number at risk

23 19 0.77 (0.63–0.91) 0.73 (0.58–0.88) 171 19 0.91 (0.87–0.95) 0.88 (0.83–0.93) 76 0.82 (0.75–0.88) 17 0.89 (0.82–0.97) –

case mix, and supports the notion that TAR surgery takes a long time to master (Henricson et al. 2014, Barg et al. 2015).

Acta Orthopaedica 2020; 91 (2): 191–196

Follow-up of 15 years or more has to our knowledge not previously been published by any other national registry. We found TAR survival rates at 15 years of 0.63 and at 20 years of 0.58. It is perhaps encouraging that long-term survival beyond 15 years is good with little change in survival between then and 20 years, although this inference is drawn from few patients. If we compare our 15-year TAR survival rates with 0.80 after knee arthroplasty and 0.88 after hip arthroplasty (Koskinen et al. 2008, Makela et al. 2014), the results are still, as previously noted, clearly inferior (Labek et al. 2011). The general trend seems to be an increase in the use of TAR for treatment of ankle arthritis (Zaidi et al. 2013, Rybalko et al. 2018), although we did not find this to be true in Sweden. During 2016, 53 total ankle replacements were performed in Sweden (10 million inhabitants, 9 million inhabitants ≥ 15 years) indicating 0.6 replacements per 105 inhabitants over the age of 15, a decrease compared with 1/105 during 2011 when the incidence had plateaued (Henricson et al. 2011b). The low numbers in Sweden may partly be explained by logistical factors and staffing shortage in operating units and local procurement of prosthetic designs during recent years. A tendency of fewer patients with RA requiring TAR has also been noted during recent years, perhaps representing benefits of better non-surgical treatment. For the early and current prosthetic design groups, 10-year TAR survival rate was 0.69 (CI 0.64–0.73) and 0.84 (CI 0.79– 0.88) respectively (Figure 2). This may in part be referred to better prosthetic designs including instrumentation, but also to increasing surgeon experience, more careful patient selection, and improved healthcare. Concerning separate prosthetic designs, we found that the currently most implanted models in Sweden (TM Ankle and Rebalance) show promising short-term results. Harris et al. (2014) studied 220 Rebalance prostheses, and similarly found encouraging early results. Some of the early models, such as AES, were withdrawn due to higher than expected complication rates (Di Iorio et al. 2017, Koivu et al. 2017b). Robust long-term follow-up results are important. The most common reason for revision was loosening of the tibial and/or the talar component, accounting for about half of the revisions in both the early design group and current design group. Loosening has also been found to be the most common complication in several other studies from national registries as well as specialized/high-volume units (Barg et al. 2015). Polyethylene failure was the second most common complication in the early design group, whereas this complication was reported only once in the current design group. This may indicate improvement in implant design or manufacturing but may also partly be the result of longer follow-up of earlier designs. Further in-detail examination is necessary, preferably in a collaboration between several national registries. In Sweden, available options of prosthetic designs are dictated by local procurement arrangements, which are based

195

on local review of scientific reports and reports from the industry. This underlines the importance of future independent real-world studies from national registries. The strength of national registry data is the real-world representation, i.e., data from several different surgeons and units, thus giving a picture of actual results in contrast to reports from single units or surgeons. Furthermore, this study has to our knowledge the hitherto largest cohort and follow-up time from a national ankle registry. Age, sex, and primary diagnosis seem overall to be similar to what has been reported in the literature. The generalizability of our results may be limited. Data are derived from the specific setting of the Swedish Ankle Registry and results must be interpreted with care. However, results may still be applicable to similar populations (Table 2). Another limitation of registry data is the uncertainty as to whether reporting is accurate and complete. We are confident that the reporting of primary TARs and revisions to the Swedish national ankle registry are accurate as the data are continuously compared with official data from the Swedish National Patient Registry by personal identity number and because the community of surgeons performing TARs is small. Furthermore, prosthetic failure was estimated from the date of the revision surgery, not from the date when failure was established. In addition, only revisions are reported, not failed implants that have not been revised. Given the observational study design, we cannot draw any causal inferences. In conclusion, use of current prosthetic designs was associated with better TAR survival. This may in part reflect better prosthetic designs and instrumentation, but also increasing surgeon experience and better patient selection, as well as improved healthcare. However, TAR has a long way to go to approach survival rates for hip and knee arthroplasty. Supplementary data Figure 5 and Table 4 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2019.1709312

AU, LJ, IK, ÅC, AH, MK, and BR designed the study; AU and ÅC collected data; AU, LJ, and BR interpreted data and performed statistical analyses; AU and BR wrote the first version; all authors together finalized the manuscript. Acta thanks Helka Koivu and Dawson Muir for help with peer review of this study.

Barg A, Zwicky L, Knupp M, Henninger H B, Hintermann B. HINTEGRA total ankle replacement: survivorship analysis in 684 patients. J Bone Joint Surg Am 2013; 95(13): 1175-83. Barg A, Wimmer M D, Wiewiorski M, Wirtz D C, Pagenstert G I, Valderrabano V. Total ankle replacement. Dtsch Arztebl Int 2015; 112(11): 177-84. Bonnin M, Gaudot F, Laurent J R, Ellis S, Colombier J A, Judet T. The Salto total ankle arthroplasty: survivorship and analysis of failures at 7 to 11 years. Clin Orthop Relat Res 2011; 469(1): 225-36.

196

Clough T, Bodo K, Majeed H, Davenport J, Karski M. Survivorship and long-term outcome of a consecutive series of 200 Scandinavian Total Ankle Replacement (STAR) implants. Bone Joint J 2019; 101-B(1): 47-54. Di Iorio A, Viste A, Fessy M H, Besse J L. The AES total ankle arthroplasty: analysis of failures and survivorship at ten years. Int Orthop 2017; 41(12): 2525-33. Fevang B T, Lie S A, Havelin L I, Brun J G, Skredderstuen A, Furnes O. 257 ankle arthroplasties performed in Norway between 1994 and 2005. Acta Orthop 2007; 78(5): 575-83. Frigg A, Germann U, Huber M, Horisberger M. Survival of the Scandinavian total ankle replacement (STAR): results of ten to nineteen years follow-up. Int Orthop 2017; 41(10): 2075-82. Harris N, Henricson A, Rydholm U, Knutson K, Popelka S. The early multicenter results of the rebalance total ankle replacement. Orthopaedic Proceedings 2014; 96-B(SUPP_17): 2-. Henricson A, Carlsson Å. Survival analysis of the single- and double-coated STAR ankle up to 20 years: long-term follow-up of 324 cases from the Swedish Ankle Registry. Foot Ankle Int 2015; 36(10): 1156-60. Henricson A, Skoog A, Carlsson Å. The Swedish Ankle Arthroplasty Register: an analysis of 531 arthroplasties between 1993 and 2005. Acta Orthop 2007; 78(5): 569-74. Henricson A, Carlsson Å, Rydholm U. What is a revision of total ankle replacement? Foot Ankle Surg 2011a; 17(3): 99-102. Henricson A, Nilsson J A, Carlsson Å. 10-year survival of total ankle arthroplasties: a report on 780 cases from the Swedish Ankle Register. Acta Orthop 2011b; 82(6): 655-9. Henricson A, Cöster M, Carlsson Å. The Swedish National Ankle Registry. Fuß & Sprunggelenk 2014; 12(2): 65-9. Hosman A H, Mason R B, Hobbs T, Rothwell A G. A New Zealand national joint registry review of 202 total ankle replacements followed for up to 6 years. Acta Orthop 2007; 78(5): 584-91. NZ Joint Registry. Annual report 2017. Available from: http://www.nzoa.org. nz. Accessed May 1, 2019. Swedish Ankle Registry. Annual report 2017. Available from: http://www. swedankle.se. Accessed May 1, 2019. Kerkhoff Y R, Kosse N M, Metsaars W P, Louwerens J W. Long-term functional and radiographic outcome of a mobile bearing ankle prosthesis. Foot Ankle Int 2016; 37(12): 1292-302.

Acta Orthopaedica 2020; 91 (2): 191–196

Koivu H, Kohonen I, Mattila K, Loyttyniemi E, Tiusanen H. Long-term results of Scandinavian Total Ankle Replacement. Foot Ankle Int 2017a; 38(7): 723-31. Koivu H, Kohonen I, Mattila K, Loyttyniemi E, Tiusanen H. Medium to long-term results of 130 Ankle Evolutive System total ankle replacement: inferior survival due to peri-implant osteolysis. Foot Ankle Surg 2017b; 23(2): 108-15. Koskinen E, Eskelinen A, Paavolainen P, Pulkkinen P, Remes V. Comparison of survival and cost-effectiveness between unicondylar arthroplasty and total knee arthroplasty in patients with primary osteoarthritis: a follow-up study of 50,493 knee replacements from the Finnish Arthroplasty Register. Acta Orthop 2008; 79(4): 499-507. Labek G, Thaler M, Janda W, Agreiter M, Stockl B. Revision rates after total joint replacement: cumulative results from worldwide joint register datasets. J Bone Joint Surg Br 2011; 93(3): 293-7. Makela 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. Mann J A, Mann R A, Horton E. STAR ankle: long-term results. Foot Ankle Int 2011; 32(5): S473-84. Palanca A, Mann R A, Mann J A, Haskell A. Scandinavian Total Ankle Replacement: 15-year follow-up. Foot Ankle Int 2018; 39(2): 135-42. Ranstam J, Robertsson O. Statistical analysis of arthroplasty register data. Acta Orthop 2010; 81(1): 10-4. Rybalko D, Schwarzman G, Moretti V. Recent national trends and outcomes in total ankle arthroplasty in the United States. J Foot Ankle Surg 2018; 57(6): 1092-5. Skytta E T, Koivu H, Eskelinen A, Ikavalko M, Paavolainen P, Remes V. Total ankle replacement: a population-based study of 515 cases from the Finnish Arthroplasty Register. Acta Orthop 2010; 81(1): 114-8. Tomlinson M, Harrison M. The New Zealand Joint Registry: report of 11-year data for ankle arthroplasty. Foot Ankle Clin 2012; 17(4): 719-23. Wood P L, Deakin S. Total ankle replacement: the results in 200 ankles. J Bone Joint Surg Br 2003; 85(3): 334-41. Zaidi R, Cro S, Gurusamy K, Siva N, Macgregor A, Henricson A, Goldberg A. The outcome of total ankle replacement: a systematic review and metaanalysis. Bone Joint J 2013; 95-B(11): 1500-7.

Acta Orthopaedica 2020; 91 (2): 197–202

197

Overgrowth of the lower limb after treatment of developmental dysplasia of the hip: incidence and risk factors in 101 children with a mean follow-up of 15 years Chan YOON 1, Chang Ho SHIN 2, Dong Ook KIM 2, Moon Seok PARK 3, Won Joon YOO 2, Chin Youb CHUNG 3, In Ho CHOI 2, and Tae-Joon CHO 2 1 Department of Orthopaedic Surgery, Seoul Bumin Hospital, Seoul; 2 Division of Pediatric Orthopaedics, Seoul National University Children’s Hospital, Seoul; 3 Department of Orthopaedic Surgery, Seoul National University Bundang Hospital, Seongnam, Gyeonggi, Republic of Korea Correspondence: CHS: drshinch@gmail.com Submitted 2019-04-08. Accepted 2019-10-15.

Background and purpose — There are few studies on overgrowth of the affected limb after treatment of developmental dysplasia of the hip (DDH). We investigated the incidence of overgrowth and its risk factors in DDH patients. Patients and methods — 101 patients were included in this study. Overgrowth was defined by 2 criteria: when the height of the femoral head of the affected side was higher than that of the contralateral side by more than 10 mm, or by more than 15 mm. The potential risk factors of distinct overgrowth were retrospectively examined using multivariable analysis. Results — When overgrowth was defined as femoral head height difference (FHHD) > 10 mm, its incidence was 44%, and only femoral osteotomy was identified as a significant risk factor with a relative risk (RR) of 1.6 (95% confidence interval [CI] 1.0–2.5). When overgrowth was defined as FHHD > 15 mm, its incidence was 23%, and femoral osteotomy was identified as the only significant risk factor with an RR of 2.3 (CI 1.2–4.5). Overgrowth developed more frequently in patients who underwent femoral osteotomy at the age of 2 to 4 years (87%) than in the others (46%) (p = 0.04). Interpretation — Overgrowth of the affected limb is common in DDH patients. Patients who underwent femoral osteotomy, especially at the age of 2 to 4 years, may require careful follow-up because of the substantial risk for overgrowth.

Leg length discrepancy (LLD) sometimes occurs during the treatment of developmental dysplasia of the hip (DDH) (Kalamchi and MacEwen 1980, Porat et al. 1994, Zadeh et al. 2000, Inan et al. 2008). LLD may manifest as shortening of the affected limb from proximal femoral growth disturbance, or as overgrowth of the affected limb. Most previous studies focused on shortening due to proximal femoral growth disturbance (Kalamchi and MacEwen 1980, Porat et al. 1994, Inan et al. 2008). To our knowledge, only 1 study reported the incidence of overgrowth of the affected limb in patients with DDH (Zadeh et al. 2000). In that study, all hips that showed overgrowth of the affected limb by more than 15 mm had had a femoral osteotomy in conjunction with anterolateral open reduction. Femoral osteotomy is performed to facilitate reduction, to correct excessive femoral anteversion, and to redirect the femoral head toward the acetabular center with intent to improve the stability of reduction, which is the primary stimulus for acetabular remodeling (Smith et al. 1963). However, femoral osteotomy also may risk overgrowth as a femoral shaft fracture (Staheli 1967, Zadeh et al. 2000). Overgrowth and consequent LLD results in hip adduction and decrease of lateral center–edge angle on the long limb side in the weight-bearing position. This may lead to excessive load on the growth plate between the acetabular cartilage and the ilium and can consequently compromise normal acetabular development, resulting in the so-called “long-leg dysplasia” (Ponseti 1978, Zadeh et al. 2000). We assessed the incidence and risk factors of LLD by overgrowth in patients who had been treated for DDH.

198

Figure 1. Measurement of femoral head height difference (FHHD) on standing anteroposterior radiograph of the hip.

Patients and methods This was a retrospective cohort study. Medical records and serial radiographs of patients with DDH who were treated between April 1982 and December 2004 were reviewed. Inclusion criteria were dislocated-type DDH with unilateral involvement, which had not received any prior treatment before being referred to our hospital. Of 196 consecutive patients meeting these criteria, the following patients were excluded: patients who were not followed up until skeletal maturity (n = 68); patients associated with neuromuscular disease (n = 10); 1 patient with other congenital anomaly; and 4 patients who had medical conditions affecting leg length, such as septic arthritis of the hip. We also excluded 10 hips that presented after 5 years of age and 2 hips with type III osteonecrosis according to the criteria by Bucholz-Ogden (Roposch et al. 2012). Hips with type I or II osteonecrosis were included in the study. No hips had type IV osteonecrosis. Based on these criteria, 101 patients (101 hips) were enrolled in the study. LLD was determined on standing anteroposterior radiographs of the hip by measuring the femoral head height difference (FHHD) at skeletal maturity or at the time of intervention for overgrowth (Figure 1) (Friberg 1983). LLD was recorded as a positive value when the affected side was longer than the unaffected side. Distinct overgrowth was determined to be present with 2 criteria: FHHD > 10 mm or FHHD > 15 mm. It has been reported that LLD > 10 mm results in a significant mediolateral shift in the center of pressure toward the longer leg (Mahar et al. 1985, Gurney 2002). Demographic data, initial severity of DDH, reduction method, osteotomy site, and deformity of proximal femur were considered candidate risk factors for overgrowth. Relative risk (RR) with 95% confidence interval (CI) was calcu-

Acta Orthopaedica 2020; 91 (2): 197–202

Figure 2. Measurement of center–head distance discrepancy (CHDD) on anteroposterior radiograph of the hip. The CHDD was defined as the difference in the center–head distance between the DDH side and the normal side, and expressed as a percentage of the normal side measurement.

lated and multivariable analysis was performed, respectively for the two definition of overgrowth. In order to evaluate initial severity, hips were graded according to the Tönnis classification (Tönnis et al. 1987), and the acetabular index (AI) was measured at the time of reduction. To evaluate deformity of proximal femur, osteonecrosis was classified according to Bucholz–Ogden criteria (Roposch et al. 2012), and the widest diameter of the femoral head was measured at skeletal maturity or just before the intervention for overgrowth. Coxa magna was recorded when the femoral head diameter of the affected side was larger by 10% than that of the unaffected side (Young et al. 2014). Skeletal maturity was determined based on the closure of the proximal femoral growth plate and triradiate cartilage. Some patients had undergone repeated multiple osteotomies, and others had undergone both femoral and pelvic osteotomy, making the definition of “age at osteotomy” ambiguous. In turn, we did not include age at osteotomy in the multivariable analysis. In the subgroup of patients who underwent a single femoral or pelvic osteotomy, the association between age at osteotomy and development of distinct overgrowth was analyzed. Hip radiograph around 3 years of age (2 ~ 4 years) was available in 42 of the 44 patients who did not undergo any osteotomy. The association between the AI and center–head distance discrepancy (CHDD) (Chen et al. 1994) around 3 years of age and development of overgrowth was analyzed in these patients (Figure 2). They were not measured in the osteotomy group because osteotomy was performed before 3 years of age in some patients and osteotomy could change those parameters. All radiographs were reviewed by the 2 authors. To determine intra-observer reliability, measurements were made by the first author (CY) on 2 different days, 4 weeks apart. To determine the inter-observer reliability, the same measure-

Acta Orthopaedica 2020; 91 (2): 197–202

199

Table 1. Treatment modalities applied to patients Treatment modalities

Number of hips (N = 101)

Closed reduction (CR) 34 CR with femoral osteotomy 1 Medial open reduction (OR) 10 Anterolateral OR a 33 with femoral osteotomy 5 b with pelvic osteotomy 14 with femoral and pelvic osteotomies 4 Osteotomy for residual dysplasia Femoral osteotomy c 10 Pelvic osteotomy d 10 Femoral and pelvic osteotomies e 13 femoral pelvic

Table 2. Femoral head height difference (FHHD) in the patients with distinct overgrowth. Values are number of hips Mean age (SD) at treatment 14 (6) months 36 months 13 (5) months 17 (9) months 19 (3) months 27 (15) months 28 (10) months 3.3 (2.0) years 4.8 (2.8) years 4.2 (2.6) years 4.5 (2.7) years

a 2 hips had been redislocated after CR. b 1 hip had been redislocated after CR and 1 hip after anterolateral OR. c There were patients who had repeated femoral osteotomies (twice,

n = 2; 3 times, n = 1) before skeletal maturity. patient had had pelvic osteotomy twice before skeletal maturity. e There were patients who had repeated femoral osteotomies (twice, n = 1; 3 times, n = 1) or repeated pelvic osteotomies (twice, n = 1) before skeletal maturity. CR = closed reduction under general anesthesia; OR = open reduction. d1

ments were made by another author (DOK) after a consensusbuilding session to define the radiographic measurements. Intra-observer and inter-observer reliability were evaluated by intraclass correlation coefficients (ICCs), which were calculated assuming absolute agreement and a single measurement with a 2-way random-effects model (see Appendix). Statistics A sample size of 68 participants was required to detect a difference of 14% between groups in the incidence rate of LLD over 15 mm, using a 2-sided Z-test of the difference between proportions with a power of 80% at a level of significance of p < 0.05. This 14% difference represents the difference between a 20% LLD incidence rate in the DDH group and 6% rate in the normal population (Knutson 2005). Continuous data were statistically analyzed using the independent Student t-test or Mann–Whitney U-test after the Kolmogorov–Smirnov normality test, and categorical data were analyzed using the chi-square test or Fisher’s exact test. Associations between risk factors and the development of distinct overgrowth were assessed using a log-binomial model to calculate adjusted RR and CI. Since the incidence of overgrowth was more than 10% in this study, we used the log-binomial model instead of logistic regression analysis to avoid overestimating the risk (McNutt et al. 2003). Univariable analysis was performed initially to assess baseline differences between patients with and without distinct overgrowth. Next, variable selection for multivariable analysis was based on a causal path diagram that was created using the directed acyclic graph (DAG) (Shrier and Platt 2008). Covariates in DAG were

FHHD, mm > 20 > 15 > 10 a FHHD b FHHD

Intervention for overgrowth a (n = 24)

No intervention for overgrowth b (n = 20)

3 14 24

2 9 20

Total 5 23 44

was measured at intervention. was measured at skeletal maturity.

selected based on previous literature and hypothesized relationship (Kalamchi and MacEwen 1980, Tönnis et al. 1987, Zadeh et al. 2000, Hefti 2007). A receiver operating characteristic (ROC) curve was applied to determine cut-off values for the AI and CHDD at the age of 3 years, which distinguished between the cases with and without distinct overgrowth in the non-osteotomy group. P-values of < 0.05 were considered statistically significant. Ethics, funding, and potential conflicts of interest This study was approved by the institutional ethics committee (H-1711-013-895) and was performed in accordance with the Declaration of Helsinki. No funding was received and there are no competing interests declared.

Results There were 91 female and 10 male patients. 59 hips were leftside hips. Preoperatively, 76 hips were Tönnis grade II, 16 hips grade III, and 9 hips grade IV. Various treatment modalities had been used (Table 1). Pre-reduction skin traction was used in 16 patients. The mean period of traction was 8 days (3–26). The mean age at the latest follow-up was 17 years (12–29), and the follow-up duration averaged 15 years (8–26). 5 hips had type I, and 29 hips had type II osteonecrosis. FHHD was more than 10 mm in 44 patients (95% CI 35–53) and more than 15 mm in 23 patients (CI 16–29) (Table 2). 24 patients underwent intervention for LLD. 16 patients had epiphysiodesis in the distal femur at a mean age of 11.6 years (10.7–12.6), and 8 patients had femoral shortening combined with varization osteotomy at a mean age of 7.9 years (3.4–12.5). Their mean FHHD was 13 mm (10–19) at surgical intervention and 1 mm (–15 to 13) at skeletal maturity. In the univariable analysis, anterolateral OR and femoral osteotomy were significant risk factors in both definitions of distinct overgrowth (Table 3, see Supplementary data). On the basis of the DAG and our univariable analysis, the following variables were included in the relevant multivariable analysis: age at reduction, initial severity, reduction method, and femoral osteotomy.

200

Acta Orthopaedica 2020; 91 (2): 197–202

Table 4. Multivariable analysis of risk factors for development of overgrowth of the affected limb in overall patients

Discussion

Little has been reported on the incidence and risk factors of LLD by overgrowth in patients with DDH. In the current study, more than 40% of Age at reduction 1.0 (1.0–1.0) 0.8 1.0 (1.0–1.1) 0.6 Initial severity patients treated by closed reduction (CR) or open Tőnnis grade ≥ III 1.0 (0.6–1.5) 0.8 0.8 (0.4–1.7) 0.6 reduction (OR) had LLD exceeding 10 mm. LLD AI at reduction 1.0 (1.0–1.0) 0.6 0.9 (0.9–1.0) 0.1 of 10 mm may not have a considerable influence Reduction method Anterolateral OR 1.6 (1.0–2.8) 0.08 2.4 (1.0–5.9) 0.06 on normal hips (Song et al. 1997). However, in Femoral osteotomy 1.6 (1.0–2.5) a 0.03 2.3 (1.2–4.5) a 0.02 patients with DDH, a small amount of overgrowth a Statistically significant. might compromise development of the acetabulum, FHHD = femoral head height difference; RR = relative risk; which is already dysplastic, by increased mechaniCI = confidence interval; AI = acetabular index; OR = open reduction. cal compression of the growth plate of the acetabular cartilage complex and Hueter–Volkmann law (Ponseti 1978, Stokes 2002). It can break the balWhen distinct overgrowth was defined as FHHD > 10 mm ance between the growth of the acetabular and triradiate cartior FHHD > 15 mm, only femoral osteotomy was found to be lages, which is important for normal acetabular development a significant risk factor with a RR of 1.6 (CI, 1.0–2.5) or a RR to occur as the pelvis enlarges (Ponseti 1978). of 2.3 (CI, 1.2–4.5), respectively, according to multivariable In our study cohort, overgrowth > 10 mm was observed in analysis (Table 4). 44% of patients, and > 15 mm in 23%. This incidence is much Demographic and clinical characteristics of femoral oste- higher than that of a healthy cohort of 600 military recruits, otomy and non-femoral osteotomy groups were comparable 4% of whom had an LLD of more than 15 mm (Hellsing except for the proportion of hips with coxa magna (Table 5, 1988). Our results are similar to a previous study reporting an see Supplementary data). Of 33 patients in the femoral oste- incidence of 17% of overgrowth more than 15 mm, and recurotomy group, 31 patients underwent femoral varization dero- rence of hip dysplasia in 5 of 12 hips with an increase in leg tational osteotomy, and 2 patients underwent femoral derota- length (Zadeh et al. 2000). tional osteotomy. Neck shaft angle of the affected side was We found femoral osteotomy to be an independent risk 153° (SD 7°) preoperatively, 135° (10°) immediately after factor for overgrowth after adjusting for other risk factors. femoral osteotomy, and 135° (6°) at skeletal maturity or at Similar to our results, Zadeh et al. (2000) reported that all the the time of intervention for overgrowth. The neck shaft angle hips that showed overgrowth after OR for DDH had underof the contralateral side was 151° (10°), 152° (8°), and 135° gone femoral osteotomy. Geometrically, proximal femoral (6°), respectively. Distinct overgrowth developed more fre- varus osteotomy shortens the effective length of the femur quently in the femoral osteotomy group than in the non-femo- (Suda et al. 1995). However, we found that the affected leg ral osteotomy group (Table 5, see Supplementary data). In 28 showed overgrowth after femoral varus osteotomy and eventupatients who underwent a single femoral osteotomy, distinct ally became longer than the unaffected leg at skeletal maturity overgrowth developed much more frequently in patients who or at the time of intervention for overgrowth (Figure 3). This underwent femoral osteotomy at the age of 2 to 4 years (13/15) justifies the intentional shortening of the effective femur length than those who underwent femoral osteotomy before the age using the medial closing-wedge technique of varus osteotomy, of 2 years (2/5) or after the age of 4 years (4/8) (p = 0.04). and further shortening by trapezoidal wedge resection may be In the non-osteotomy group, the CHDD at the age of 3 years considered. This overgrowth phenomenon may share the same was significantly larger in the overgrowth group than in the pathogenic mechanism with overgrowth after femoral shaft no-overgrowth group when distinct overgrowth was defined as fracture (Staheli 1967, Shapiro 1981, Corry and Nicol 1995). FHHD > 10 mm (p = 0.005), while it was not when it was Many studies reported that it occurs mainly in children over 2 defined as FHHD > 15 mm (Table 6, see Supplementary data). years of age (Staheli 1967, Corry and Nicol 1995). In accorThe AI at the age of 3 years was not significantly different dance with results of these studies, overgrowth in our DDH between the overgrowth and no-overgrowth groups in both cohort occurred more often when femoral osteotomy was definitions of distinct overgrowth. An ROC curve showed the performed at the age of 2 to 4 years. In contrast, Suda et al. optimal cutoff value for distinct overgrowth (FHHD > 10 mm) (1995) reported no difference in femoral length between the to be a CHDD of 7%, with 77% sensitivity and 76% specific- affected and unaffected sides at skeletal maturity after femoral ity (area under the curve = 0.8, CI 0.6–0.9; p = 0.009). The varus osteotomy in DDH patients. However, they evaluated incidence of distinct overgrowth (FHHD > 10 mm) was higher LLD in only 45% of the 42 subjects due to the unavailability in patients with a CHDD of > 7% (10/17) than patients with a of radiographs, and their mean age at femoral osteotomy was CHDD of ≤ 7% (3/25) in the non-osteotomy group (p < 0.002). 4.7 years, which was older than the most vulnerable age for Risk factors

FHHD > 10 mm RR (95% CI) p-value

FHHD > 15 mm RR (95% CI) p-value

Acta Orthopaedica 2020; 91 (2): 197â&#x20AC;&#x201C;202

201

but showed borderline significance in multivariable analysis after adjusting for other variables, such as performance of osteotomy. In our study, occurrence of type II osteonecrosis was not associated with overgrowth. It could be partly because the deformity in type II osteonecrosis is caput valgum rather than coxa valga and the center of rotation is close to the top of the femoral head (Shin et al. 2017). Moreover, severe type II osteonecrosis shortens the femoral neck, which may compensate for the lengthening effect of the proximal femoral valgus. Our study has several limitations. First, LLD measured by iliac crest height difference better reflects pelvic tilt and its influence on the spine compared with LLD measured by FHHD. Figure 3. An example of overgrowth of the affected limb after treatment of DDH. A girl underwent However, we had no choice but to anterolateral open reduction at age 1.5 years (A) and femoral osteotomy at age 5 years. At 2 months post-osteotomy FHHD was not distinct (B). However, FHHD became +14 mm at age 8.5 measure FHHD because this study years (C) and +19 mm at age 11.5 years (D) resulting in pelvic tilt. She underwent percutaneous was a retrospective study and the iliac epiphysiodesis using transphyseal screws (E) and eventually had a level pelvis at age 15 years crest was not covered in many radio(F). graphs; this may be the reason why pelvic osteotomy was not a significant overgrowth in our study. This finding suggests that the risk risk factor for overgrowth in this study. In addition, wholeof overgrowth should be considered when performing femoral leg radiographs were not available in many cases, which was also due to the retrospective design of this study. Therefore, osteotomy, especially at the age of 2 to 4 years. In the non-osteotomy group, large CHDD around 3 years of although we hypothesize that LLD was attributable to femoage was associated with FHHD > 10 mm. Although it failed ral overgrowth rather than tibial overgrowth in most cases, we to show a statistically significant association with FHHD > could not prove it. Second, there may be a selection bias in 15 mm may be due to type II error, hips with overgrowth had estimating the incidence of overgrowth in our DDH cohort. larger mean CHDD than hips with no overgrowth around 3 Those who had risk factors for overgrowth, such as femoral years of age. It is difficult to speculate its pathogenic mecha- osteotomy and large CHDD, tended to be more compliant nism. A study on adult hip dysplasia showed that two-thirds in terms of clinical visits compared with those who showed of patients who did not undergo any surgery during childhood uneventful hip joint development. By the same token, those had an affected leg longer than the unaffected leg by more treated by a Pavlik harness were not followed up until skeletal than 5 mm (Metcalfe et al. 2005). Altered mechanical loading maturity and were not included in this study. Despite these limitations, we conclude that overgrowth of on the proximal femur by lateral subluxation, which appeared as large CHDD, might affect leg length through the Hueterâ&#x20AC;&#x201C; the affected limb is a commonly encountered problem after Volkmann law (Stokes 2002). We could not exclude the possi- DDH treatment. DDH patients who had undergone femoral bility that LLD persisted in early childhood before measuring osteotomy, especially between the ages of 2 to 4 years, and CHDD because standing hip radiographs could not be taken those who have a large CHDD around 3 years of age, require in early childhood and whole-leg radiograph was not routinely careful follow-up for LLD development because it may jeoptaken during follow-up. ardize normal acetabular development. Further studies are Before the commencement of this study, we had an impres- warranted to prove the association between overgrowth of the sion that anterolateral OR is an independent risk factor for affected leg and recurrence of hip dysplasia. overgrowth. In a previous study, which did not adjust confounding variables, all hips that showed overgrowth under- Supplementary data went femoral osteotomy in conjunction with anterolateral Appendix and Tables 3, 5, and 6 are available as supplemenOR (Zadeh et al. 2000). In our study, anterolateral OR was tary data in the online version of this article, http://dx.doi.org/ a statistically significant risk factor in univariable analysis 10.1080/17453674.2019.1688485

202

CY, DOK: data curation, data analysis, and writing of the original draft. CHS, TJC: conceptualization, methodology, and revision of the paper. MSP: methodology and data analysis. WJY, CYC, IHC: resources, supervision, and validation. Acta thanks Klaus Dieter Parsch and Terje Terjesen for help with peer review of this study.

Chen H, Kuo K N, Lubicky J P. Prognosticating factors in acetabular development following reduction of developmental dysplasia of the hip. J Pediatr Orthop 1994; 14(1): 3-8. Corry I, Nicol R. Limb length after fracture of the femoral shaft in children. J Pediatr Orthop 1995; 15(2): 217-19. Friberg O. Clinical symptoms and biomechanics of lumbar spine and hip joint in leg length inequality. Spine 1983; 8(6): 643-51. Gurney B. Leg length discrepancy. Gait Posture 2002; 15(2): 195-206. Hefti F. Pediatric orthopedics in practice. New York: Springer Science & Business Media; 2007. Hellsing A L. Leg length inequality: a prospective study of young men during their military service. Ups J Med Sci 1988; 93(3): 245-53. Inan M, Chan G, Bowen J R. The correction of leg-length discrepancy after treatment in developmental dysplasia of the hip by using a percutaneous epiphysiodesis. J Pediatr Orthop B 2008; 17(1): 43-6. Kalamchi A, MacEwen G. Avascular necrosis following treatment of congenital dislocation of the hip. J Bone Joint Surg Am 1980; 62(6): 876-88. Knutson G A. Anatomic and functional leg-length inequality: a review and recommendation for clinical decision-making, Part I: Anatomic leg-length inequality: prevalence, magnitude, effects and clinical significance. Chiropr Osteopat 2005; 13(1): 11. Mahar R, Kirby R, MacLeod D. Simulated leg-length discrepancy: its effect on mean center-of-pressure position and postural sway. Arch Phys Med Rehabil 1985; 66(12): 822-4. McNutt L A, Wu C, Xue X, Hafner J P. Estimating the relative risk in cohort studies and clinical trials of common outcomes. Am J Epidemiol 2003; 157(10): 940-3. Metcalfe J E, Banaszkiewicz P, Kapoor B, Richardson J, Jones C W, Kuiper J. Unexpected long femur in adults with acetabular dysplasia. Acta Orthop Belg 2005; 71(4): 424.

Acta Orthopaedica 2020; 91 (2): 197–202

Ponseti I V. Growth and development of the acetabulum in the normal child: anatomical, histological, and roentgenographic studies. J Bone Joint Surg Am 1978; 60(5): 575-85. Porat S, Robin G, Howard C. Cure of the limp in children with congenital dislocation of the hip and ischaemic necrosis: fifteen cases treated by trochanteric transfer and contralateral epiphysiodesis. J Bone Joint Surg Br 1994; 76(3): 463-7. Roposch A, Wedge J H, Riedl G. Reliability of Bucholz and Ogden classification for osteonecrosis secondary to developmental dysplasia of the hip. Clin Orthop Relat Res 2012; 470(12): 3499-505. Shapiro F. Fractures of the femoral shaft in children: the overgrowth phenomenon. Acta Orthop Scand 1981; 52(6): 649-55. Shin C H, Hong W K, Lee D J, Yoo W J, Choi I H, Cho T J. Percutaneous medial hemi-epiphysiodesis using a transphyseal screw for caput valgum associated with developmental dysplasia of the hip. BMC Musculoskelet Disord 2017; 18(1): 451. Shrier I, Platt R W. Reducing bias through directed acyclic graphs. BMC Med Res Methodol 2008; 8: 70. Smith W S, Coleman C R, Olix M L, Slager R F. Etiology of congenital dislocation of the hip: an experimental approach to the problem using young dogs. J Bone Joint Surg Am 1963; 45(3): 491-500. Song K M, Halliday S E, Little D G. The effect of limb-length discrepancy on gait. J Bone Joint Surg Am 1997; 79(11): 1690-8. Staheli L T. Femoral and tibial growth following femoral shaft fracture in childhood. Clin Orthop Relat Res 1967; (55): 159-64. Stokes I. Mechanical effects on skeletal growth. J Musculoskelet Neuronal Interact 2002; 2(3): 277-80. Suda H, Hattori T, Iwata H. Varus derotation osteotomy for persistent dysplasia in congenital dislocation of the hip: proximal femoral growth and alignment changes in the leg. J Bone Joint Surg Br 1995; 77(5): 756-61. Tönnis D, Legal H, Graf R. Congenital dysplasia and dislocation of the hip in children and adults. Berlin: Springer-Verlag; 1987. Young E Y, Gebhart J J, Bajwa N, Cooperman D R, Ahn N U. Femoral head asymmetry and coxa magna: anatomic study. J Pediatr Orthop 2014; 34(4): 415-20. Zadeh H, Catterall A, Hashemi-Nejad A, Perry R. Test of stability as an aid to decide the need for osteotomy in association with open reduction in developmental dysplasia of the hip. J Bone Joint Surg Br 2000; 82(1): 17-27.

Acta Orthopaedica 2020; 91 (2): 203–208

203

Development of a risk score for scoliosis in children with cerebral palsy Katina PETTERSSON 1,2, Philippe WAGNER 2, and Elisabet RODBY-BOUSQUET 1,2 1 Department of Clinical Sciences, Lund University, Orthopedics, Lund, Sweden; 2 Centre for Clinical Research, Region Västmanland—Uppsala University, Västerås, Sweden Correspondence: katina.pettersson@med.lu.se Submitted 2019-04-09. Accepted 2019-12-09.

Background and purpose — Children and young adults with cerebral palsy (CP) have an increased risk of developing scoliosis, with a prevalence ranging from 11% to 29%. Information on risk factors for the emergence and progression of scoliosis is inconclusive. This study aimed to develop a risk score based on 5-year-old children with CP to predict the risk of scoliosis before the age of 16. Patients and methods — This prospective registry study included 654 children with CP in Sweden born in 2000 to 2003 and registered with the Swedish CP follow-up program (CPUP) at the age of 5 years, including all Gross Motor Function Classification System (GMFCS) levels. 92 children developed a scoliosis before the age of 16 years. Univariable and multivariable logistic regressions were used to analyze 8 potential predictors for scoliosis: GMFCS, sex, spastic subtype, epilepsy, hip surgery, migration percentage, and limited hip or knee extension. Results — 4 predictors for scoliosis remained significant after analyses: female sex, GMFCS levels IV and V, epilepsy, and limited knee extension, and a risk score was constructed based on these factors. The predictive ability of the risk score was high, with an area under the receiver operating characteristics curve of 0.87 (95% CI 0.84–0.91). Interpretation — The risk score shows high discriminatory ability for differentiating between individuals at high and low risk for development of scoliosis before the age of 16. It may be useful when considering interventions to prevent or predict severe scoliosis in young children with CP.

Children and adults with cerebral palsy (CP) have a high risk of developing scoliosis (Saito et al. 1998, Hägglund et al. 2018a). The risk for neuromuscular scoliosis also increases with age (Persson-Bunke et al. 2012, Hägglund et al. 2018a) and early onset of scoliosis has been identified as a predictor for severe scoliosis (Saito et al. 1998, Gu et al. 2011, PerssonBunke et al. 2012, Yoshida et al. 2018). To date, information on the risk factors for emergence and progression of scoliosis in children with CP is inconclusive (Loeters et al. 2010). The Gross Motor Function Classification System (GMFCS) levels III, IV, and V have been identified as risk factors for scoliosis in children with CP (Loeters et al. 2010, Hägglund et al. 2018a). New findings suggest that girls with CP have a higher risk than boys of developing scoliosis (Bertoncelli et al. 2017, Hägglund et al. 2018a, Pettersson et al. 2019), and that epilepsy is another independent risk factor (Bertoncelli et al. 2017). Lateral displacement of the hips, hip dislocations, and previous hip surgery are sometimes associated with neuromuscular scoliosis (Persson-Bunke et al. 2006, Bertoncelli et al. 2017, Hägglund et al. 2018b), whereas successful hip surveillance leading to a reduced number of dislocated hips results in a lower proportion of scoliosis (Hägglund et al. 2014). In addition, limited hip or knee extension is highly associated with scoliosis, windswept hips, and postural asymmetries in adults with CP (Rodby-Bousquet et al. 2013, Ágústsson et al. 2018). We developed a risk score to predict the individual risk for a 5-year-old child with CP to develop a severe scoliosis before the age of 16 years. This individual risk score can prompt clinicians to initiate and implement preventive interventions and strategies at an early stage and hopefully reduce the risk of scoliosis.

204

Patients and methods This prospective registry study was based on data from the combined Swedish Cerebral Palsy Follow up-program and National quality registry (CPUP) (Hägglund et al. 2014, Alriksson-Schmidt et al. 2017). We included all children with CP in Sweden born in 2000 to 2003 who were reported to the registry at 2 time points: at 5 years and before the age of 16 years. Over 95% of all children with CP in Sweden participate in CPUP (Westbom et al. 2007). Children are assessed clinically twice a year until they are 6 years of age, and then once a year (Alriksson-Schmidt et al. 2014). CP diagnosis and neurological subtype are classified from the age of 4 years. Exclusion and inclusion criteria are consistent with those of the Surveillance of Cerebral Palsy network in Europe (SCPE) (2000). Scoliosis was defined as either having: (1) a radiographically measured Cobb angle of at least 40° (Persson-Bunke et al. 2012, Hägglund et al. 2018a); (2) a spinal fusion because of scoliosis; or (3) a severe scoliosis at clinical examination (Persson-Bunke et al. 2015) before the age of 16 years. In CPUP, clinical assessment of the spine is used as a screening tool to identify children in need of further radiographic examination. The clinical examination is performed with the child in a sitting position, both upright and forward-bending. A pronounced curve preventing the child from attaining an upright position without external support is rated as severe scoliosis and treated as scoliosis in this study. This standardized clinical spinal assessment has high interrater reliability, sensitivity, specificity, and criterion-related validity compared with a radiographically measured Cobb angle (Persson-Bunke et al. 2015). Mild or moderate curves do not exceed 25 degrees of Cobb angle (Persson-Bunke et al. 2015), and in this study were treated as having no scoliosis. Timepoint 1 used data from the assessment performed closest to the 5th birthday of each child, and timepoint 2 used the latest assessment before 16 years of age. Based on previous findings the following 8 variables were analyzed as potential predictors of scoliosis: GMFCS levels IV and V, female sex, spastic subtype, epilepsy, hip surgery, migration percentage (MP) > 40%, and limited hip or knee extension. Gross motor function was classified using the expanded and revised version of the GMFCS, levels I to V (Palisano et al. 2008). We grouped and used GMFCS levels I to III (higher motor function) as the reference category to compare each GMFCS level, IV and V (lower motor function). Male was used as the reference category for sex (Bertoncelli et al. 2017, Hägglund et al. 2018a, Pettersson et al. 2019). Epilepsy was reported as Yes or No, with no epilepsy as the reference category (Bertoncelli et al. 2017). Neurological subtype was classified as spastic CP (spastic unilateral and bilateral CP), versus non-spastic CP (ataxic, dyskinetic, mixed type), which was used as the reference category (Bertoncelli et al. 2017, 2018). All types of hip surgery (including femur osteotomy, pelvic osteotomy, and adductor

Acta Orthopaedica 2020; 91 (2): 203–208

psoas tenotomy) were grouped as hip surgery, with no surgery used as the reference category (Bertoncelli et al. 2017). Lateral displacement/migration of the hip joint was measured using MP (Reimers 1980). The value for the worst side was used. We defined lateral migration as an MP > 40% (Hägglund et al. 2007, Hermanson et al. 2015), using MP ≤ 40% as the reference category. Passive range of motion (ROM) for hip and knee extension was measured by goniometer in a standardized position (www.cpup.se), and the value for the worst side was used for all analyses. Data were dichotomized into either full hip or knee extension or limited hip or knee extension (–5° or less), and the former was used as the reference category. Statistics We analyzed the risk of developing scoliosis after 5 years of age and before the age of 16, using predictors measured at the age of 5 years. The odds ratio (OR) with 95% confidence interval (CI) for scoliosis was calculated using logistic regression for the following variables: GMFCS, sex, spastic CP, epilepsy, hip surgery, MP, passive hip extension, and knee extension. The first step in the analysis was to calculate ORs for each variable using univariable logistic regression. The next step was multivariable logistic regression analysis using a stepwise backward elimination process, whereby 1 explanatory variable (that with the highest nonsignificant p-value) at a time was removed from the regression model. This step was repeated until only significant variables remained in the model. A risk score was constructed using the remaining variables as independent, significant predictors of scoliosis. A p-value of less than 0.05 was considered statistically significant. This risk score was then evaluated using the area under the receiver operating characteristics (ROC) curve (AUC). The AUC can be interpreted as the probability that a randomly selected child with scoliosis has a higher predicted risk of severe scoliosis before the age of 16 than a randomly chosen individual without scoliosis. An AUC value of 1 is considered perfect and a value of 0.5 no better than chance (Steyerberg 2019). For these statistical analyses, IBM SPSS Statistics v24 (IBM Corp, Armonk, NY, USA) was used. The risk score AUC was additionally validated using 10-fold cross validation. The risk score development process was also validated using a different predictor selection approach, L1-penalized logistic regression (Tibshirani 1996, Zou and Hastie 2005). The latter analyses were performed using R (R Core Team 2019, R Foundation for Statistical Computing, Vienna, Austria). Ethics, funding, and potential conflicts of interest The study was approved by the Medical Research Ethics Committee at Lund University (383/2007, 443-99), and permission was obtained to extract data from the CPUP registry. The study was funded by the Norrbacka-Eugenia Foundation, Region Västmanland, Promobilia, Stiftelsen för bistånd åt rörelsehindrade i Skåne and Forte. The funding sources had no decision-

Acta Orthopaedica 2020; 91 (2): 203–208

205

Table 1. Demographic distribution for included variables. Values are frequency (%) Variable/category a

GMFCS I II III IV V Sex Male Female CP b subtype Spastic unilateral Spastic bilateral Ataxic Dyskinetic Mixed type Missing Birthyear 2000 2001 2002 2003 Epilepsy, yes Hip surgery c Worst side MP d 41–100% Limited hip extension ≤ –5º Limited knee extension ≤ –5º

Results

n = 654 264 (40) 117 (18) 53 (8.1) 115 (18) 105 (16) 372 (57) 282 (43) 268 (41) 216 (33) 41 (6.3) 93 (14) 22 (3.4) 14 (2.1) 113 (17) 144 (22) 191 (29) 206 (32) 222 (34) 53 (8.1) 45 (6.9) 35 (5.4) 91 (14)

a GMFCS = Gross Motor Function Classification System. b CP = cerebral palsy. c Femur-, pelvic osteotomy, adductor psoas tenotomy. d MP = migration percentage.

making role or influence on the study design, data collection, data analysis, data interpretation, or writing of the report. The authors declare that they have no conflicts of interest.

The risk score was based on data of all 654 children with CP born in Sweden from 2000 to 2003, with data reported closest to 5 years and before 16 years of age (Table 1). Their ages ranged from 4.0 to 5.9 years. We identified 92/654 (14%) individuals who met our criteria for severe scoliosis before the age of 16. Of these children, 59 had undergone spinal fusion for scoliosis, a further 6 had a reported Cobb angle of at least 40°, and 27 had severe scoliosis identified on clinical examination. Of the initial 8 possible predictors at the age of 5 years, 4 remained significantly associated with the development of severe scoliosis before the age of 16 years: female sex, GMFCS levels IV and V, epilepsy, and limited knee extension (Table 2). For the population frequencies of the 8 variables included, see Table 1. The sensitivity and (1 – specificity) of the risk score are shown in Figure 1. The sensitivity is the proportion of children with scoliosis correctly predicted to develop severe scoliosis before the age of 16 years. (1 – specificity indicates the proportion of those without scoliosis who were correctly predicted not to develop scoliosis before 16 years of age.) The discriminatory accuracy of the risk score was high, with an AUC = 0.87 (CI 0.84–0.91), indicating a strong ability to differentiate between high- and low-risk individuals. The AUC remained high after cross-validation, AUC = 0.866. The equation for calculating the risk score was determined to be: Risk Score = –4.219 + 0.655·sex + 2.288·GMFCSIV + 3.366·GMFCSV + 0.622·epilepsy + 0.614·limited knee extension. Sex is a dichotomous indicator variable that took a value of 1 for females and 0 for males. GMFCS IV is also a dichotomous indicator variable that took a value of 1 when the individual had GMFCS level IV, and 0 otherwise. GMFCS V is a

Table 2. Odds ratios (OR) with 95% confidence intervals (CI) for possible predictors of scoliosis measured at the age of 5 (steps 1 and 2) a Step 1 Univariable analysis Risk factors OR (95% CI) GMFCS I–III GMFCS IV GMFCS V Female Epilepsy Knee extension ≤ –5° Hip extension ≤ –5° Spastic CP Hip surgery MP > 40% a

Step 2 Step 3 Step 4 Multivariable analysis, Multivariable analysis, L1-penalized first regression last regression regression OR (95% CI) OR (95% CI) OR

Ref. 12 (5.6–25) 41 (20–83) 1.7 (1.1–2.6) 3.9 (2.5–6.2) 4.3 (2.6–7.1) 3.5 (1.7–7.3) 0.4 (0.3–0.7) 7.8 (4.3–14) 5.9 (3.1–11)

Ref. Ref. 10 (4.2–24) 9.9 (4.6–21) 25 (9.5–65) 29 (14–61) 1.9 (1.1–3.4) 1.9 (1.1–3.3) 1.6 (0.9–2.9) 1.9 (1.1–3.2) 1.9 (1.0–3.7) 1.9 (1.0–3.4) 1.2 (0.5–3.0) 1.3 (0.7–2.4) 1.3 (0.6–2.9) 1.3 (0.6–3.0)

8.4 21 1.8 1.6 1.8 1.2 1.2 1.3 1.3

Step 3 shows the significant factors remaining after the stepwise procedure, which predict the development of scoliosis before the age of 16 years for children with cerebral palsy (CP). Step 4 shows the ORs from the sensitivity analysis using L1-penalized regression. Spastic CP includes uni- and bilateral spasticity. For CP, GMFCS, Hip surgery, and MP, see Table 1

206

Acta Orthopaedica 2020; 91 (2): 203–208

Table 3. Risk of developing scoliosis before the age of 16, corresponding to each risk score level

Sensitivity/Specificity 1.0

Risk score

0.8

< –2.2 –2.2 to –1.4 –1.3 to –0.85 –0.85 to –0.41 –0.41 to 0 0 to 0.41 0.41 to 0.85 0.85 to 1.4 1.4 to 2.2 > 2.2

0.6

0.4

0.2 Sensitivity Specificity

Risk of scoliosis (%) 0 to 10 10 to 20 20 to 30 30 to 40 40 to 50 50 to 60 60 to 70 70 to 80 80 to 90 90 to 100

0 0

0.2

0.4

0.6

0.8

1.0

Predicted risk of scoliosis

Figure 1. Graph showing the proportion of children with scoliosis correctly predicted to develop scoliosis before the age of 16 years (the sensitivity) and the proportion of children without scoliosis correctly predicted not to develop scoliosis before the age of 16 years (the specificity) for choice of a cutoff to indicate a high-risk individual.

corresponding indicator. Epilepsy took a value of 1 when epilepsy was present and 0 when it was not. Limited knee extension corresponded to the value on the worst side when both sides were measured and took the value of 1 when the limited knee extension was –5° or less, and 0 when the individual had full knee extension. The risk score can be translated into the risk of developing scoliosis using Table 3. As an example, a female child at GMFCS level V, with epilepsy and limited ROM for knee extension, will have a risk score of 1.04 and a 70% to 80% risk of developing severe scoliosis before the age of 16. The sensitivity analysis results are presented in Table 2 in the far-right column. AUC for the resulting risk score was marginally worse than the original risk score, with a crossvalidated AUC of 0.85.

Discussion We developed a risk score based on the following risk factors assessed at the age of 5 years: female sex, GMFCS levels IV and V, epilepsy, and having limited knee extension in 654 children with CP in Sweden born in 2000 to 2003. These were identified as independent predictors for the development of scoliosis before the age of 16 years. The AUC of the resulting risk score was 0.87 (CI 0.84–0.91), indicating a high accuracy in differentiating between high- and low-risk individuals. The AUC remained at this level after cross-validation, showing that its high value was not due to overfitting, and may generalize to other populations. However, true external validity is yet to be verified in additional CP populations in future studies (Steyerberg and Harrell 2016). The sensitivity analysis further showed that our method had satisfactory performance in terms of selecting suitable predictors, at least compared with

another popular approach, L1-penalized logistic regression (Ranstam and Cook 2018). To our knowledge, this is the first study creating a risk score for development of severe scoliosis based on predictors identified in 5-year old children with CP. Scoliosis is usually defined as a radiographically measured lateral spinal curvature of at least 10° (Cobb 1948, Roberts and Tsirikos 2016), and a Cobb angle of ≥ 40° has been suggested as a cut-off for severe scoliosis when considering surgical interventions (Saito et al. 1998, Persson-Bunke et al. 2012, Hägglund et al. 2018a). But at present there are no internationally agreed criteria for the recommendation of spine surgery (Toovey et al. 2017). Scoliosis can also be identified at clinical examination. Clinically defined moderate or severe scoliosis show a sensitivity of 75% and a specificity of 96% compared with radiographic Cobb angle (Persson-Bunke et al. 2015). Even though scoliosis usually develops after the age of 8 years (Persson-Bunke et al. 2012), some children may start to develop scoliosis from the age of 5 years (Hägglund et al. 2018a). The most important and rapid growth spurt in children generally occurs from 11 to 14 years of age (Negrini et al. 2018). We therefore decided to identify potential variables at the age of 5 years to predict the risk for development of scoliosis before the age of 16 years. Our results are consistent with previous findings (PerssonBunke et al. 2012, Hägglund et al. 2018a) in identifying GMFCS as a strong predictor of scoliosis, with an OR of 9.9 for GMFCS level IV up to 29 for GMFCS level V. When only GMFCS was included in the ROC analysis, the AUC was 0.85 (CI 0.81–0.89). We included children at all GMFCS levels, thereby allowing comparisons between those with higher gross motor function (GMFCS levels I to III) and those with lower gross motor function (GMFCS levels IV and V). Notably, those at GMFCS levels I and II have a low incidence of scoliosis (Hägglund et al. 2018a). We found an increased risk for scoliosis in girls, with an OR of 1.9 for girls compared with boys, which confirms findings from recent studies (Bertoncelli et al. 2017, Hägglund et al. 2018a, Pettersson et al. 2019). Our study also confirmed the previous observation that epilepsy is a predictor of scoliosis (Bertoncelli et al. 2017), even after

Acta Orthopaedica 2020; 91 (2): 203–208

adjustment for other variables (Table 2). In supine lying, limited knee or hip extension may cause the legs to tilt to one side, forcing the pelvis and trunk into rotation. This deviation could be reinforced by time and gravity and is likely one of the reasons why limited knee and hip extension increases the probability for scoliosis and windswept hips in adults with CP (Ágústsson et al. 2018). With previous findings in mind (Rodby-Bousquet et al. 2013, Ágústsson et al. 2018, Cloodt et al. 2018), it is not surprising that limited knee extension contributes to the risk of developing scoliosis. Limited hip and knee extension may co-occur (Ágústsson et al. 2018), which might explain why limited hip extension did not remain significant in the multivariate regression analyses. Because knee contracture is one of the most common types of contracture (Rodby-Bousquet et al. 2013, Ágústsson et al. 2018, Cloodt et al. 2018), this reinforces the importance of monitoring range of motion and posture from an early age, and promptly addressing contractures and postural asymmetries. The ROC curve (Figure 1) shows consecutive cutoffs for the predicted risk of scoliosis, and shows the possible cutoffs for classifying individuals as high- or low-risk (Steyerberg 2019). For a clinician to make a reasonable assessment of the risk of scoliosis, it would be advisable to base the decision regarding cutoffs on the nature of the treatment option being considered (Hermanson et al. 2015). For minor interventions like spinal orthosis, one should try to capture as many children as possible so as not to risk missing someone (high sensitivity). However, for major interventions like surgery, it is important to select a more conservative cutoff that includes only high-risk individuals, to prevent unnecessary interventions that in turn can have a negative impact on quality of life (high specificity). In general, spinal surgery is not performed before severe scoliosis is already present. However, our risk score may identify children who need close clinical and radiographic surveillance of their spine. Most of the children included in this study have been followed regularly and received early interventions such as hip surgery in accordance with CPUP’s protocol and guidelines, meaning that by using a preventive follow-up program (that might include early corrective interventions), it can give protective effects of both hips and spines (Elkamil et al. 2011, Hägglund et al. 2014). This way of implementing early interventions may differ from other countries without follow-up programs. The chosen definition of limited knee extension with a cutoff value of –5° or more affected the observed frequency of knee contracture. A higher cutoff value, for example –10°, would have halved the prevalence from 14% to 7%. There might be other predictors that were not identified or included in this study. A strength of the study is that the risk score is based on the total population of children with CP who have been followed in a standardized way. Scoliosis, once established, is usually a lifelong condition for individuals with CP (Hägglund et al. 2018a), in contrast to idiopathic scoliosis where the risk of further progression

207

is much lower after spinal growth is complete (Negrini et al. 2018). It has been shown that young adults with CP at GMFCS levels IV and V have a 50% increased risk of moderate or severe scoliosis by the age of 18. By 20 years of age, 75% of adults at GMFCS level V have a Cobb angle ≥ 40° (PerssonBunke et al. 2012, Hägglund et al. 2018a). This further reinforces the importance of early identification of children with CP at risk of developing scoliosis. Based on this population of children with CP, we conclude that the presence of the following 4 predictors at the age of 5 years increases the probability of severe scoliosis before the age of 16 years: female sex, GMFCS levels IV and V, epilepsy, and having limited knee extension. The risk score has a high AUC and discriminatory accuracy for differentiating between children with CP and it is hoped will provide clinicians with early insight regarding a child’s risk of developing scoliosis and allow them to differentiate between high-risk and low-risk individuals.

Study design: KP, ERB, PW. Data collection: KP, ERB. Data analysis: KP, ERB, PW. Manuscript preparation: KP, ERB. The authors would like to thank Magnus Tägil, Lund University, and Piotr Michno, Linköping University, for their help with constructive ideas in the beginning of this project. Acta thanks Thomas Andersen and Sebastian I Wolf for help with peer review of this study.

(SCPE) SoCPiE. Surveillance of cerebral palsy in Europe: a collaboration of cerebral palsy surveys and registers. Dev Med Child Neurol 2000; 42(12): 816-24. Ágústsson A, Sveinsson T, Pope P, Rodby-Bousquet E. Preferred posture in lying and its association with scoliosis and windswept hips in adults with cerebral palsy. Disabil Rehabil 2018: 1-5. doi: 10.1080/09638288.2018.1492032. Alriksson-Schmidt A, Hagglund G, Rodby-Bousquet E, Westbom L. Followup of individuals with cerebral palsy through the transition years and description of adult life: the Swedish experience. J Pediatr Rehabil Med 2014; 7(1): 53-61. doi: 10.3233/PRM-140273. Alriksson-Schmidt A I, Arner M, Westbom L, Krumlinde-Sundholm L, Nordmark E, Rodby-Bousquet E, Hagglund G. A combined surveillance program and quality register improves management of childhood disability. Disabil Rehabil 2017; 39(8): 830-6. doi: 10.3109/09638288.2016.1161843. Bertoncelli C M, Solla F, Loughenbury P R, Tsirikos A I, Bertoncelli D, Rampal V. Risk factors for developing scoliosis in cerebral palsy: a crosssectional descriptive study. J Child Neurol 2017; 32(7): 657-62. doi: 10.1177/0883073817701047. Bertoncelli C M, Bertoncelli D, Elbaum L, Latalski M, Altamura P, Musoff C, Rampal V, Solla F. Validation of a clinical prediction model for the development of neuromuscular scoliosis: a multinational study. Pediatr Neurol 2018; 79: 14-20. doi: 10.1016/j.pediatrneurol.2017.10.019. Cloodt E, Rosenblad A, Rodby-Bousquet E. Demographic and modifiable factors associated with knee contracture in children with cerebral palsy. Dev Med Child Neurol 2018; 60(4): 391-6. doi: 10.1111/dmcn.13659. Cobb J. Outline for the study of scoliosis. In Instructional course lectures. American Academy of Orthopedic Surgeons; 1948.

208

Elkamil A I, Andersen G L, Hägglund G, Lamvik T, Skranes J, Vik T. Prevalence of hip dislocation among children with cerebral palsy in regions with and without a surveillance programme: a cross sectional study in Sweden and Norway. BMC Musculoskelet Disord 2011; 12(1): 284. doi: 10.1186/1471-2474-12-284. Fender D, Baker A D L. Spinal disorders in childhood II: spinal deformity. Surgery (Oxford) 2011; 29(4): 175-80. doi: 10.1016/j.mpsur.2011.01.009. Gu Y, Shelton J E, Ketchum J M, Cifu D X, Palmer D, Sparkman A, JermerGu M K, Mendigorin M. Natural history of scoliosis in nonambulatory spastic tetraplegic cerebral palsy. PM R 2011; 3(1): 27-32. doi: 10.1016/j. pmrj.2010.09.015. Hermanson M, Hägglund G, Riad J, Rodby-Bousquet E, Wagner P. Prediction of hip displacement in children with cerebral palsy: development of the CPUP hip score. Bone Joint J 2015; 97-B(10): 1441-4. doi: 10.1302/0301620X.97B10.35978. Hägglund G, Lauge-Pedersen H, Wagner P. Characteristics of children with hip displacement in cerebral palsy. BMC Musculoskelet Disord 2007; 8:101. doi: 1471-2474-8-101 [pii] 10.1186/1471-2474-8-101. Hägglund G, Alriksson-Schmidt A, Lauge-Pedersen H, Rodby-Bousquet E, Wagner P, Westbom L. Prevention of dislocation of the hip in children with cerebral palsy: 20-year results of a population-based prevention programme. Bone Joint J 2014; 96-B(11): 1546-52. doi: 10.1302/0301620X.96B11.34385. Hägglund G, Pettersson K, Czuba T, Persson-Bunke M, Rodby-Bousquet E. Incidence of scoliosis in cerebral palsy. Acta Orthop 2018a; 89(4): 443-7. doi: 10.1080/17453674.2018. Hägglund G, Goldring M, Hermanson M, Rodby-Bousquet E. Pelvic obliquity and measurement of hip displacement in children with cerebral palsy. Acta Orthop 2018b; 89(6): 652-5. doi: 10.1080/17453674.2018.1519104. Loeters M J, Maathuis C G, Hadders-Algra M. Risk factors for emergence and progression of scoliosis in children with severe cerebral palsy: a systematic review. Dev Med Child Neurol 2010; 52(7): 605-11. doi: 10.1111/j.14698749.2010.03617.x. Negrini S, Donzelli S, Aulisa A G, Czaprowski D, Schreiber S, de Mauroy J C, Diers H, Grivas T B, Knott P, Kotwicki T, Lebel A, Marti C, Maruyama T, O’Brien J, Price N, Parent E, Rigo M, Romano M, Stikeleather L, Wynne J, Zaina F. 2016 SOSORT guidelines: orthopaedic and rehabilitation treatment of idiopathic scoliosis during growth. Scoliosis Spinal Disord 2018; 13(1): 3. doi: 10.1186/s13013-017-0145-8. Palisano R J, Rosenbaum P, Bartlett D, Livingston M H. Content validity of the expanded and revised Gross Motor Function Classification System. Dev Med Child Neurol 2008; 50(10): 744-50. doi: 10.1111/j.14698749.2008.03089.x. Persson-Bunke M, Hagglund G, Lauge-Pedersen H. Windswept hip deformity in children with cerebral palsy. J Pediatr Orthop B 2006; 15(5): 335-8. Persson-Bunke M, Hagglund G, Lauge-Pedersen H, Wagner P, West-

Acta Orthopaedica 2020; 91 (2): 203–208

bom L. Scoliosis in a total population of children with cerebral palsy. Spine (Phila Pa 1976) 2012; 37(12): E708-13. doi: 10.1097/ BRS.0b013e318246a962. Persson-Bunke M, Czuba T, Hägglund G, Rodby-Bousquet E. Psychometric evaluation of spinal assessment methods to screen for scoliosis in children and adolescents with cerebral palsy. BMC Musculoskelet Disord 2015; 16(1): 351. doi: 10.1186/s12891-015-0801-1. Pettersson K, Rodby-Bousquet E. Prevalence and goal attainment with spinal orthoses for children with cerebral palsy. J Pediatr Rehabil Med 2019; 12(2): 197-203. doi: 10.3233/PRM-180596. R Core Team. R: A language and environment for statistical computing. Vienna: R Foundation for Statistical computing; 2019. Ranstam J, Cook J A. LASSO regression. Br J Surg 2018; 105(10): 1348-. doi: 10.1002/bjs.10895. Reimers J. The stability of the hip in children. A radiological study of the results of muscle surgery in cerebral palsy. Acta Orthop Scand (Suppl.) 1980; 184: 1-100. Roberts S B, Tsirikos A I. Factors influencing the evaluation and management of neuromuscular scoliosis: a review of the literature. J Back Musculoskelet Rehabil 2016; 29(4): 613-23. doi: 10.3233/bmr-160675. Rodby-Bousquet E, Czuba T, Hägglund G, Westbom L. Postural asymmetries in young adults with cerebral palsy. Dev Med Child Neurol 2013; 55(11): 1009-15. doi: 10.1111/dmcn.12199. Saito N, Ebara S, Ohotsuka K, Kumeta H, Takaoka K. Natural history of scoliosis in spastic cerebral palsy. Lancet 1998; 351(9117): 1687-92. doi: 10.1016/S0140-6736(98)01302-6. Steyerberg E W. Clinical prediction models: a practical approach to development, validation, and updating. Springer Nature: [S.l.]; 2019. Steyerberg E W, Harrell F E, Jr. Prediction models need appropriate internal, internal–external, and external validation. J Clin Epidemiol 2016; 69:2457. doi: 10.1016/j.jclinepi.2015.04.005. Tibshirani R. Regression shrinkage and selection via the lasso. J Roy Stat Soc: Series B (Methodological) 1996; 58(1): 267-88. Toovey R, Harvey A, Johnson M, Baker L, Williams K. Outcomes after scoliosis surgery for children with cerebral palsy: a systematic review. Dev Med Child Neurol 2017; 59(7): 690-8. doi: 10.1111/dmcn.13412. Westbom L, Hagglund G, Nordmark E. Cerebral palsy in a total population of 4–11 year olds in southern Sweden. Prevalence and distribution according to different CP classification systems. BMC Pediatr 2007; 7:41. doi: 10.1186/1471-2431-7-41. Yoshida K, Kajiura I, Suzuki T, Kawabata H. Natural history of scoliosis in cerebral palsy and risk factors for progression of scoliosis. J Orthop Sci 2018; 23(4): 649-52. doi: 10.1016/j.jos.2018.03.009. Zou H, Hastie T. Regularization and variable selection via the elastic net. J Roy Stat Soc: series B (statistical methodology) 2005; 67(2): 301-20.

Acta Orthopaedica 2020; 91 (2): 209–214

209

Gorham–Stout disease: good results of bisphosphonate treatment in 6 of 7 patients Kristian Nikolaus SCHNEIDER 1, Max MASTHOFF 2, Georg GOSHEGER 1, Sebastian KLINGEBIEL 1, Dominik SCHORN 1, Julian RÖDER 1, Tim VOGLER 1, Moritz WILDGRUBER 2, and Dimosthenis ANDREOU 1 1 Department

of Orthopaedics and Tumor Orthopaedics, University Hospital of Münster, Germany; 2 Institute of Clinical Radiology, University Hospital of Münster, Germany Correspondence: kristian.schneider.mh@gmail.com Submitted 2019-09-05. Accepted 2019-12-16.

Background and purpose — Gorham–Stout disease (GSD) is a rare mono- or polyostotic condition characterized by idiopathic intraosseous proliferation of angiomatous structures resulting in progressive destruction and resorption of bone. Little is known about the course of disease and no previous study has evaluated patients’ quality of life (QoL). Patients and methods — This is a retrospective analysis of 7 consecutive patients (5 males) with a median age at diagnosis of 14 years and a median follow-up of 7 years who were diagnosed with GSD in our department between 1995 and 2018. Data regarding clinical, radiographic, and histopathological features, and treatment, as well as sequelae and their subsequent therapy, were obtained. QoL was assessed by Musculoskeletal Tumor Society Score (MSTS), Toronto Extremity Salvage Score (TESS), and Reintegration to Normal Living (RNL) Index. Results — 3 patients had a monoostotic and 4 patients a polyostotic disease. Besides a diagnostic biopsy, 4 of the 7 patients had to undergo 8 surgeries to treat evolving sequelae. Using an off-label therapy with bisphosphonates in 6 patients, a stable disease state was achieved in 5 patients after a median of 20 months. The median MSTS, TESS, and RNL Index at last follow-up was between 87% and 79%. Interpretation — Due to its rare occurrence, diagnosis and treatment of GSD remain challenging. Off-label treatment with bisphosphonates appears to lead to a stable disease state in the majority of patients. QoL varies depending on the individual manifestations but good to excellent results can be achieved even in complex polyostotic cases with a history of possibly life-threatening sequelae.

Gorham–Stout disease (GSD) is a rare mono- or polyostotic condition characterized by the idiopathic intraosseous proliferation of angiomatous structures, resulting in progressive destruction and resorption of bone (Rauh and Gross 1997, Boyer et al. 2005, Dellinger et al. 2014). Gorham and Stout (1955) specified the condition concluding that “the progressive osteolysis is always associated with an angiomatosis of blood and sometimes of lymphatic vessels, which seemingly are responsible for it.” Today, around 300 cases of GSD have been described. Clinical manifestations depend on the affected site as well as on evolving sequelae, like bone deformity, spontaneous fractures, pericardial effusion, chyloperitoneum, or chylothorax due to leaks in the lymphatic vessels network or thoracic duct invasion (Patrick 1976, Tie et al. 1994, Ludwig et al. 2016). The diagnosis of GSD can be difficult and laboratory findings are usually normal (Liu et al. 2016). Local radiographs may initially demonstrate unspecific patchy radiolucencies (Adams et al. 2016, Ramaroli et al. 2019), while progressive bone osteolysis can be observed later on (Kuriyama et al. 2010; Dellinger et al. 2014). The final diagnosis is based on histopathological examination of a biopsy specimen of the affected bones (Dellinger et al. 2014, Zanelli et al. 2020). There have been isolated reports in the literature describing cases where an osteolysis stopped progressing after several to many years (Boyer et al. 2005). Treatment options include systemic treatment with bisphosphonates, sirolimus and interferon alpha-2b (IFNa-2b), radiation, and local surgery but no gold standard has yet emerged (Hammer et al. 2005, Heyd et al. 2011, Li et al. 2018, Mo et al. 2018, Ramaroli et al. 2019). We evaluated the long-term disease course of patients with GSD focusing on clinical disease features, treatment including bisphosphonates, and sequelae, as well as patients’ QoL using a combination of standardized scoring systems (Tunn et al. 2008).

210

Acta Orthopaedica 2020; 91 (2): 209–214

Table 1. Clinical data of 7 patients with GSD

Patient Sex

Age diagnosed

Follow-up

1 2 3 4 5 6

Male Female Male Male Female Male

5y7m 29y1m 21y2m 13y8m 11y7m 42y7m

7y3m 9y3m 2y3m 21y10m 8m 6y10m

Male

12y4m

4y7m

Initial symptoms Pain, swelling Pain Pain, swelling Pain Pain, swelling Pain, shoulder impingement Pain, weight loss

Involved bones Skull, thoracic and lumbar vertebrae, ribs, RL pelvis, RL femur, RL humerus R ilium RL femur, L pelvis, R humerus, thoracic and lumbar vertebrae Thoracic and lumbar vertebrae, sternum, RL pelvis, R humerus L tibia L scapula Cervical vertebrae

Figure 1. Dilated thin-walled lymphatic and vascular vessels within rarefied lamellar bone.

Patients and methods

Results

7 consecutive patients (5 males) who were diagnosed with GSD in our department between 1995 and 2018 were included in this study. All patients were symptomatic and/or had documented disease progression. Anonymized pertinent data were obtained from patients’ records. Patients’ quality of life at last follow-up was determined using a combination of standardized scoring systems, as proposed by Tunn et al. (2008): Musculoskeletal Tumor Society Score (MSTS), Toronto Extremity Salvage Score (TESS), and Reintegration to Normal Living (RNL) Index. The median age at diagnosis was 14 years (5–42) and the median follow-up amounted to 7 years (1–22). The median time from first symptoms to final diagnosis was 27 months (3–60). All patients complained of local pain without history of previous trauma, combined with local swelling in 3 cases, restricted range of movement in 1, case and weight loss in another case. 4 patients had a polyostotic and 3 patients a monoostotic disease (Table 1). The diagnosis was histopathologically confirmed in all patients. Biopsies showed dilated thin-walled lymphatic and vascular vessels within rarefied lamellar bone (Figure 1).

Local surgical treatment was required in 4 patients (Table 2). 1 patient underwent plate fixation after a femoral fracture (Figure 2), while another patient required occipitocervical fusion due to a progressive atlantoaxial instability (Figure 3). A stabilizing spinal instrumentation (T6–L4) was necessary in a third patient due to progressive kyphosis. The same patient required further surgical treatment 4 years later due to progressive osteolysis of T11 and L5 leading to cerebrospinal fluid leaks. Another revision surgery was necessary 5 years later following a rod fracture due to material fatigue. The remaining patient suffered from recurrent erysipelas in his left leg, which was severely deformed from the disease (Figure 4). During an erysipelas bout he developed a life-threatening sepsis, which rendered a knee disarticulation necessary. 2 patients developed a chylothorax 4 and 14 years after initial diagnosis, respectively (Figure 5). Surgical treatment with pleurodesis followed by pleurectomy was required in one of these patients due to progressive respiratory distress, while the other patient remains asymptomatic with a stable chylothorax for over 5 years. No patient underwent radiation treatment. 6 patients received bisphosphonate therapy, which was contraindicated in the remaining patient due to a severe

Acta Orthopaedica 2020; 91 (2): 209–214

211

Table 2. Clinical data of 7 patients with GSD Patient A B 1

Medication

Chylothorax

3m 1m 12m

08/11—today: zoledronic acid 2 mg every 4 weeks until 03/18, Plate osteosynthesis since 03/18: every 6 weeks due to pathologic 02/10—today: cholecalciferol 1–2 x 1000 IE daily distal femoral 02/10—today: Ca 2 x 500 mg for 10d after zoledronic acid IV fracture L (5y5m) 03/17—07/17 sirolimus 0.5 mg x 2 daily –> discontinued due to recurrent aphthous ulcerations 07/15—09/17 propranolol 1 mg per kg of body weight –> discontinued due to bradycardia

2y 18m 20m

12/10—06/11 zoledronic acid 4 mg every 4 weeks –> discontinued due to incompliance

5y – – Due to a severe nephropathy, bisphosphonate treatment is currently contraindicated in this patient

L knee disarticulation – due to recurrent erysipelas and sepsis (0m)

3y 13y10m 3y6m

09/09—09/10 peginterferon alpha2b 50µg every week –> discontinued due to hyperthyreosis 09/10—06/12 sirolimus 1 mg daily –> discontinued due to progressive chylothorax 01/13—today: sirolimus 1–2 mg daily 02/10—today: zoledronic acid 4 mg every 4 weeks 02/10—today: cholecalciferol 1–2 x 1000 IE daily 02/10—today: Ca 2 x 500mg for 10d after zoledronic acid IV

Spinal instrumentation (T6–L4) (12y7m) Augmented closure of cerebrospinal fluid leaks T11 and L5 (16y3m) Revision surgery after after rod fracture in (22y)

–

02/18—today: zoledronic acid 2 mg every 4 weeks

–

01/12—today: alendronic acid 70 mg every week

–

2y 1m 20m 03/14—today: zoledronic acid 4mg every 4 weeks

–

R 4y after GSD diagnosis

–

L 14y after GSD diagnosis a

Occipitocervical fusion (O–C2) due to atlantoaxial instability (8m) –

A. Time from first symptoms until diagnosis. B. Time from diagnosis until start of medication. C. Time of medication until stable disease. D. Required surgery (time after initial diagnosis) a Chylothorax surgery: pleurodesis 10/2011, pleurectomy 03/2015

Table 3. Median values of QoL scores of 7 patients with GSD

MSTS

Male sex Female sex < 18 years old > 18 years old Monoostotic disease Polyostotic disease No bisphosphonate treatment Bisphosphonate treatment

TESS

RNL index

87 93 85 83 85 77.5 93 93 85 77 84.5 77 87 87 79 80 87.5 80 23 43 39 90 90 82

MSTS – Musculoskeletal Tumor Society Score TESS – Toronto Extremity Salvage Score RNL – Reintegration to Normal Living

nephropathy (Table 2). No adverse effects were observed under bisphosphonate treatment. Stable disease (defined as no radiographic progression of bone resorption) was achieved in 5 patients after a median of 20 months (8–42). 1 patient has been under bisphosphonate treatment for only 2 months with no radiographic follow-up as yet. 2 patients with chylothorax underwent an additional treatment with sirolimus and 1

patient an initial IFNa-2b therapy that had to be discontinued due to hyperthyreosis after 12 months of treatment (Table 2). The median MSTS at last follow-up amounted to 87% (23– 97); the median TESS was 87% (43–97) while the median RNL Index amounted to 79% (39–88). Although the small number of cases in our cohort precluded the use of statistical analyses, we observed that younger patients tended to achieve better score results than patients aged 18 years and older, while male patients fared slightly better than female patients. On the other hand, we found that patients with a monoostotic and polyostotic disease had similar scores. The 2 patients who developed a chylothorax had a somewhat lower MSTS but a similar TESS and RNL Index compared with the patients without chylothorax. Good to excellent results in all 3 QoL scores were achieved in the 6 patients who received bisphosphonates whilst poor results were obtained in the patient who was not able to undergo bisphosphonate treatment (Table 3).

Discussion GSD is a rare disease and severe symptoms have been described at presentation, such as bone deformities, patho-

212

Acta Orthopaedica 2020; 91 (2): 209–214

b Figure 4. Patient no 5, 12-year-old girl with severely deformed left lower extremity (a) with proliferations of angiomatous structures within bone, muscle, and subcutaneous tissue (b).

Figure 2. Patient no 1, 11-year-old boy with pathologic left distal femoral fracture with wide, dense metaphyseal bands corresponding to continuous 5-year bisphosphonate therapy (a); fracture consolidation 6 weeks postoperatively (b).

Figure 5. Patient no 1, boy at age 9 year with asymptomatic right chylothorax 4 years after initial diagnosis of GSD.

Figure 3. Patient no 7, 13-year-old boy with progressive atlantoaxial instability with multiple GSD lesions in C1–C7 (a); 6 weeks after occipitocervical fusion (b).

logic fractures, or neurological deficits due to vertebral osteolysis (Kulenkampff et al. 1990). However, most patients appear to initially develop unspecific symptoms such as pain, regional swelling, or a restricted range of motion (Möller et al. 1999, Patel 2005, Dellinger et al. 2014, Agyeman et al. 2017). We have observed similar initial symptoms in our patients as well (Table 1). As a result, medical consultation, imaging studies and diagnosis may be delayed, with only 2 of our patients being diagnosed within 3 months after developing complaints.

Acta Orthopaedica 2020; 91 (2): 209–214

Due to disease progression and local symptoms, 6 of our 7 patients underwent bisphosphonate treatment, and 5 of them achieved radiologically stable disease, while treatment was only recently started in the remaining patient. Several case reports have demonstrated the clinical benefit from bisphosphonates (Mignogna et al. 2005, Yerganyan et al. 2015, Brance 2017, Ramaroli et al. 2019). Hagberg et al. (1997) reported the first successful treatment of GSD with bisphosphonates. Following an unsuccessful radiation treatment, they initiated a bisphosphonate treatment with clodronic acid in combination with IFNa-2b in a 19-year-old male patient who suffered from progressive, polyostotic disease with a chylothorax, and achieved a “rapid improvement” of the patient’s general condition. Hammer et al. (2005) later reported that bisphosphonate monotherapy in GSD resulted in an “immediate clinical improvement” and radiologically stable disease within two years in a 45-year-old woman. Despite the activity and clinical benefits in GSD, various possible adverse effects of bisphosphonate treatment have been described. Besides gastrointestinal intolerance, flu-like symptoms (after intravenous application) and the development of osteonecrosis of the jaw (ONJ) have been described (Reid 2011, Wessel et al. 2008). Despite the fact that we have administered a frequent bisphosphonate regime to the majority of patients and the average treatment duration amounted to over 4 years, none of our patients developed an ONJ or other relevant adverse effects in long-term follow-up. Informed consent to off-label usages is required as bisphosphonates are not approved by the U.S. Food and Drug Administration (FDA) or the European Medicines Agency (EMA) for treatment of GSD (Dellinger et al. 2014). Another treatment option for GSD patients reported in the literature is sirolimus, an mTOR inhibitor that acts as a downregulator of cellular proliferation and angiogenesis (García et al. 2016). García et al. (2016) described a successful sirolimus treatment in a 43-year-old female patient with GSD of her left hemithorax accompanied by a left-sided chylothorax. Despite achieving remission within 4 weeks, treatment had to be discontinued due to metrorrhagia. We initiated sirolimus treatment additionally to bisphosphonate treatment in our 2 patients who developed a chylothorax. However, 1 patient developed recurrent aphthous ulcerations that required discontinuation of the treatment, while the other patient’s chylothorax progressed under treatment and required surgery. In the first patient, the chylothorax remains asymptomatic and stable, so that the sirolimus treatment was not resumed. Due to its antiangiogenic effect, treatment with IFNa-2b is also an effective treatment option in GSD. As with sirolimus, IFNa-2b is frequently used as combination therapy with bisphosphonates (Hagberg et al. 1997, Kuriyama et al. 2010, Ramaroli et al. 2019). We initiated IFNa-2b monotherapy in our very first patient. After reports of the effect of bisphosphonate treatment in GSD, we later opted for combination therapy with bisphosphonates and IFNa-2b. Due to the development of

213

a severe hyperthyreosis, the IFNa-2b treatment had to be discontinued. As stable disease was achieved under bisphosphonate monotherapy, the IFNa-2b treatment was not resumed. Regarding local treatment options, several authors have reported on the effectiveness of radiotherapy in GSD (Kulenkampff et al. 1990, Dunbar et al. 1993, Rauh and Gross 1997, Heyd et al. 2011, Yerganyan et al. 2015). Considering the relatively young age of GSD patients at diagnosis as well as the late effects of radiotherapy, including particularly the risk for secondary malignancies, we preferred to avoid radiotherapy in our patients. Surgical treatment may be required in GSD patients both for the affected bones and for possible sequelae (Patel 2005). In the long bones, most authors recommend resection followed by reconstruction or endoprosthetic replacement in weightbearing bones (Turra et al. 1990, Chan et al. 2016, Ellati et al. 2016, Liu et al. 2018). Spinal stabilization surgery may be required in patients with spinal involvement and actual or impending neurological deficits. 2 of our patients required stabilization procedures, with one of them undergoing 2 further surgeries due to disease progression with development of cerebrospinal fluid leaks and material fatigue. Surgery might also be required in patients who develop a chylothorax—a potentially life-threatening sequela of GSD with a reported incidence of about 17% and a mortality of up to 34% (Patrick 1976, Tie et al. 1994, Ludwig et al. 2016). A novel finding of our study concerns the QoL of affected patients. To our knowledge, no study has previously evaluated this aspect of GSD. Good to excellent MSTS, TESS, and RNL Index scores were achieved in all patients undergoing bisphosphonate treatment. Indeed, the only patient with very poor results in our cohort was the one with a contraindication to medical treatment due to a severe nephropathy, providing further evidence to the observation that systematic treatment leads to a rapid relief of symptoms (Hagberg et al. 1997). Another somewhat surprising finding was that the patientreported TESS and RNL Index were similar in patients with mono- and polyostotic disease, as well as in patients with and without sequelae (Table 3). We also found that younger patients tended to have higher QoL scores compared with older patients. This fact has also been observed by other authors and could be attributed to better adaptation to the disease by younger patients who grow up with it and cope better with the subsequent impairments and disabilities (Tunn et al. 2008, Heaver et al. 2016). The low number of patients in our cohort is, naturally, a limitation of our study. However, among the approximately 300 cases reported, up to now, in the literature, our cohort represents one of the largest series treated at a single institution, reflecting the low incidence of GSD and the challenges of accruing large patient numbers. In conclusion, due to its rare occurrence, diagnosis and treatment of GSD remain challenging. Off-label treatment

214

with bisphosphonates appears to lead to a stable disease state in the majority of patients. Ethics, funding, and potential conflicts of interest The study was approved by our local ethics committee EthikKommission, Ärztekammer Westfalen-Lippe. Reference number: 2018–617–f–S. The authors received no specific funding for this work. The authors declare that they have no competing interests.

KS, DA, MW, and MM designed the study and collected the data. KS, SK, DS, JR, TV, MW, and DA were responsible for data management, data analysis, and preparation of figures. KS and DA wrote the manuscript. KS, GG, and DA helped with data analysis and with editing of the manuscript. The authors acknowledge support from the Open Access Publication Fund of the University of Münster/Germany. Acta thanks Jendrik Hardes and Ulrich Exner for help with peer review of this study.

Adams D M, Trenor C C, Hammill A M, Vinks A A, Patel M N, Chaudry G, et al. Efficacy and safety of sirolimus in the treatment of complicated vascular anomalies. Pediatrics 2016; 137(2): e20153257-7. Agyeman K, Pretell-Mazzini J, Subhawong T, Kerr DA, Jose J. Gorham disease. Am J Orthop 2017; 46(6): E458-62. Boyer P, Bourgeois P, Boyer O, Catonné Y, Saillant G. Massive Gorham– Stout syndrome of the pelvis. Clin Rheumatol 2005; 24(5): 551-5. Brance M L. Two cases of Gorham–Stout disease with good response to zoledronic acid treatment. CCMBM 2017; 14(2): 250-4. Chan C K, Mohamed R-M, Azlina A A, Azhar M M. Multicentric disappearing bone disease treated with arthroplasty. Malays Orthop J 2016; 10(3): 42-5. Dellinger M T, Garg N, Olsen B R. Viewpoints on vessels and vanishing bones in Gorham–Stout disease. Bone 2014; 63(C): 47-52. Dunbar S F, Rosenberg A, Mankin H, Rosenthal D, Suit H D. Gorham’s massive osteolysis: the role of radiation therapy and a review of the literature. Int J Radiat Oncol Biol Phys 1993; 26(3): 491-7. Ellati R, Attili A, Haddad H, Al-Hussaini M, Shehadeh A. Novel approach of treating Gorham–Stout disease in the humerus: case report and review of literature. Eur Rev Med Pharmacol Sci 2016; 20(3): 426-32. García V, Alonso-Claudio G, Gómez-Hernández M-T, Chamorro A-J. Case report: Sirolimus on Gorham–Stout disease. Colomb Med Universidad del Valle; 2016; 47(4): 213-16. Gorham L W, Stout A P. Massive osteolysis (acute spontaneous absorption of bone, phantom bone, disappearing bone): its relation to hemangiomatosis. J Bone Joint Surg Am 1955; 37-A(5): 985-1004. Hagberg H, Lamberg K, Aström G. Alpha-2b interferon and oral clodronate for Gorham’s disease. Lancet 1997; 350(9094): 1822-3. Hammer F, Kenn W, Wesselmann U, Hofbauer L C, Delling G, Allolio B, et al. Gorham–Stout disease: stabilization during bisphosphonate treatment. J Bone Miner Res 2005; 20(2): 350-3. Heaver C, Isaacson A, Gregory J J, Cribb G, Cool P. Patient factors affect-

Acta Orthopaedica 2020; 91 (2): 209–214

ing the Toronto extremity salvage score following limb salvage surgery for bone and soft tissue tumors. J Surg Oncol 2016;113(7): 804-10. Heyd R, Rabeneck D, Dörnenburg O, Tselis N, Zamboglou N. Gorham–Stout syndrome of the pelvic girdle treated by radiation therapy: a case report. Strahlenther Onkol 2011; 187(2): 140-3. Kulenkampff H A, Richter G M, Hasse W E, Adler C P. Massive pelvic osteolysis in the Gorham–Stout syndrome. Int Orthop 1990; 14(4): 361-6. Kuriyama D K, McElligott S C, Glaser D W, Thompson K S. Treatment of Gorham–Stout disease with zoledronic acid and interferon-α: a case report and literature review. J Pediatr Hematol Oncol 2010; 32(8): 579-84. Li M-H, Zhang H-Q, Lu Y-J, Gao P, Huang H, Hu Y-C, et al. Successful management of Gorham–Stout disease in scapula and ribs: a case report and literature review. Orthop Surg 2018; 10(3): 276-80. Liu Y, Zhong D-R, Zhou P-R, Lv F, Ma D-D, Xia W-B, et al. Gorham–Stout disease: radiological, histological, and clinical features of 12 cases and review of literature. Clin Rheumatol 2016; 35(3): 813-23. Liu S, Zhou X, Song A, Kong X, Wang Y, Liu Y. Successful treatment of Gorham–Stout syndrome in the spine by vertebroplasty with cement augmentation. Medicine 2018; 97(29): e11555-6. Ludwig K F, Slone T, Cederberg K B, Silva A T, Dellinger M. A new case and review of chylothorax in generalized lymphatic anomaly and Gorham– Stout disease. Lymphology 2016; 49(2): 73-84. Mignogna M D, Fedele S, Russo Lo L, Ciccarelli R. Treatment of Gorham’s disease with zoledronic acid. Oral Oncol 2005; 41(7): 747-50. Mo A Z, Trenor C C, Hedequist D J. Sirolimus therapy as perioperative treatment of Gorham–Stout disease in the thoracic spine: a case report. JBJS Case Connect 2018; 8(3): e70. Möller G, Priemel M, Amling M, Werner M, Kuhlmey A S, Delling G. The Gorham–Stout syndrome (Gorham’s massive osteolysis). J Bone Joint Surg Br 1999; 81-B(3): 501-6. Patel D V. Gorham’s disease or massive osteolysis. Clin Med Res 2005; 3(2): 65-74. Patrick J H. Massive osteolysis complicated by chylothorax successfully treated by pleurodesis. J Bone Joint Surg Br 1976; 58-B(3): 347-9. Ramaroli D A, Cavarzere P, Cheli M, Provenzi M, Barillari M, Rodella G, et al. A child with early-onset Gorham–Stout disease complicated by chylothorax: near-complete regression of bone lesions with interferon and bisphosphonate treatment. Horm Res Paediatr 2019; 10: 1-5. Rauh G, Gross M. Disappearing bone disease (Gorham–Stout disease): report of a case with a follow-up of 48 years. Eur J Med Res 1997; 2(10): 425-7. Reid I R. Bisphosphonates in the treatment of osteoporosis: a review of their contribution and controversies. Skeletal Radiol 2011; 40(9): 1191-6. Tie M L, Poland G A, Rosenow E C. Chylothorax in Gorham’s syndrome: a common complication of a rare disease. Chest 1994; 105(1): 208-13. Tunn P U, Pomraenke D, Goerling U, Hohenberger P. Functional outcome after endoprosthetic limb-salvage therapy of primary bone tumours: a comparative analysis using the MSTS score, the TESS and the RNL index. Int Orthop 2008; 32(5): 619-25. Turra S, Gigante C, Scapinelli R. A 20-year follow-up study of a case of surgically treated massive osteolysis. Clin Orthop Relat Res 1990; (250): 297-302. Wessel J H, Dodson T B, Zavras A I. Zoledronate, smoking, and obesity are strong risk factors for osteonecrosis of the jaw: a case-control study. J Oral Maxillofac Surg 2008; 66(4): 625-31. Yerganyan V V, Body J J, De Saint Aubain N, Gebhart M. Gorham–Stout disease of the proximal fibula treated with radiotherapy and zoledronic acid. J Bone Oncol 2015; 4(2): 42-6. Zanelli M, Zizzo M, Martino G, Bisagni A, De Marco L, Ascani S. Bone marrow biopsy disclosing a rare osteolytic disorder: Gorham–Stout syndrome. Int J Surg Pathol 2020; 28(1): 76-7 .

Acta Orthopaedica 2020; 91 (2): 215–220

215

Deep learning in fracture detection: a narrative review Pishtiwan H S KALMET 1*, Sebastian SANDULEANU 2*, Sergey PRIMAKOV 2, Guangyao WU 2, Arthur JOCHEMS 2, Turkey REFAEE 2, Abdalla IBRAHIM 2, Luca v. HULST 1, Philippe LAMBIN 2, and Martijn POEZE 1,3

* Shared first authorship 1 Maastricht University Medical Center+, Department of Trauma Surgery, Maastricht; 2 The D-Lab: Decision Support for Precision Medicine, GROW— School for Oncology and Developmental Biology, Maastricht University Medical Center+, Maastricht; 3 Nutrim School for Nutrition, Toxicology and Metabolism, Maastricht University, Maastricht, The Netherlands Correspondence: pishtiwan.kalmet@mumc.nl Submitted 2019-08-15. Accepted 2019-12-09.

Abstract — Artificial intelligence (AI) is a general term that implies the use of a computer to model intelligent behav ior with minimal human intervention. AI, particularly deep learning, has recently made substantial strides in perception tasks allowing machines to better represent and interpret complex data. Deep learning is a subset of AI represented by the combination of artificial neuron layers. In the last years, deep learning has gained great momentum. In the field of orthopaedics and traumatology, some studies have been done using deep learning to detect fractures in radiographs. Deep learning studies to detect and classify fractures on computed tomography (CT) scans are even more limited. In this nar rative review, we provide a brief overview of deep learning technology: we (1) describe the ways in which deep learning until now has been applied to fracture detection on radio graphs and CT examinations; (2) discuss what value deep learning offers to this field; and finally (3) comment on future directions of this technology.

Artificial intelligence

Machine learning Deep learning

Figure 1. Visualization of Artificial Intelligence sub-family.

The demands for radiology services, e.g., magnetic resonance imaging (MRI), computed tomography (CT), and radiographs, have increased dramatically in recent years (Kim and Mac Kinnon 2018). In the United Kingdom, the number of CT exam inations increased by 33% between 2013 and 2016 (Faculty of Clinical Radiology, Clinical Radiology UK workforce census 2016 report 2016). In the Netherlands, more than 1.7 million CT examinations were carried out in all hospitals (National Institute for Health and Environment 2016). This demand will increase substantially in the coming years resulting in a con siderable strain on the workforce. On the other hand, there is a shortage of radiologists due to a lag in recruitment and the large number of radiologists approaching retirement. Further more, analyzing medical images can often be a difficult and time-consuming process. Artificial intelligence (AI) has the potential to address these issues (Kim and MacKinnon 2018). AI is a general term that implies the use of a computer to model intelligent behavior with minimal human intervention (Hamet and Tremblay 2017). Furthermore, AI, particularly deep learning, has recently made substantial strides in the per ception of imaging data allowing machines to better represent and interpret complex data (Hosny et al. 2018). Deep learning is a subset of AI represented by the combina tion of artificial neuron layers. Each layer contains a number of units, where every unit is a simplified representation of a neuron cell, inspired by its structure in the human brain (McCulloch and Pitts 1943). Today, deep learning algorithms are able to match and even surpass humans in task-specific applications (Mnih et al. 2015, Moravčík et al. 2017). Deep learning has transformed the field of information technology by unlocking large-scale, data-driven solutions to what once were time-intensive problems.

216

Acta Orthopaedica 2020; 91 (2): 215–220

Input A1

Neuron 1

Output

w1 A2 .. .

w2 wN

z = ∑ ai wi +b

Aout = g(z)

i=1

Figure 2. Visualization of artificial neuron model. Where A1–AN are the inputs, W1–WN are the weights for the input connections to neuron, b is the bias value, z is the output from the neuron.

In the last years, deep learning has gained great momen tum (Adams et al. 2019). Recent studies have shown that deep learning has the ability to perform complex interpretation at the level of healthcare specialists (Gulshan et al. 2016, Esteva et al. 2017, Lakhani and Sundaram 2017, Lee et al. 2017, Olczak et al. 2017, Ting et al. 2017, Tang et al. 2018). In the field of orthopaedic traumatology, a number of studies have been done using deep learning in radiographs to detect frac tures (Brett et al. 2009, Olczak et al. 2017, Chung et al. 2018, Kim and MacKinnon 2018, Lindsey et al. 2018, Adams et al. 2019, Urakawa et al. 2019). However, studies performing deep learning in fractures on CT scans are scarce (Tomita et al. 2018). In this narrative review, we provide a brief overview of deep learning technology; (2) describe the ways in which deep learning has been applied to fracture detection on radiographs and CT examinations thus far; (3) discuss what value deep learning offers to this field; and finally (4) comment on future directions of this technology. Artificial intelligence technology Deep Learning (DL) is a family of methods, which is part of a broad Machine-learning field and an even broader Artificial Intelligence field (Figure 1). These algorithms are unified by the idea of learning from data instead of following explicitly specified instructions. This level of abstraction makes Deep Learning algorithms applicable to solve a variety of problems in a number of quantitative fields (LeCun et al. 2015). Deep Learning has showed outstanding performance for solving semantic image processing tasks. Cireşan et al. (2012) demonstrated that DL can outperform humans by a factor of 2 in traffic sign recognition. Tompson et al. (2014) have shown that DL has significantly outperformed existing stateof-the-art techniques for human pose estimation. Chen et al. (2015) assessed DL potential in autonomous driving applica tion. ImageNet (Russakovsky et al. 2015) demonstrated that DL can be successfully applied to a variety of image-specific tasks and gain state-of the-art performance. After the DL suc cess in the computer vision field, the medical imaging field started to adopt these methods for solving its own problems

such as, e.g., medical image classification (Gao et al. 2017, Yang et al. 2018, Tran et al. 2019), medical image segmenta tion (Cha et al. 2016, Dou et al. 2017, Roth et al. 2018) and noise reduction (Chen et al. 2017, Wolterink et al. 2017). Due to the high abstractness of DL algorithms, there is no need to change methodology when moving from a problem in one field to another field. Moreover, by using this so-called trans fer learning approach, DL algorithms are able to benefit from previous successes even if the model was solving a different problem (Yang et al. 2018). The essential DL layer is composed of a number of neu rons that to a certain extent mimic the activity of a neuron cell (Figure 2). Every neuron in the layer has its own weight w for each input connection and the bias value b, where each weight w represents the strength for the particular connection, and the bias value b allows us to shift the activation function along with the weighted sum of the inputs to the neuron, control ling the value at which the activation function will trigger. In other words, each weight w defines how much influence the corresponding input will have on the neuron output and bias b, allowing the model to better fit the data. In order to create the output for the neuron and introduce non-linearity to the neuron decision, one of the activation functions, g, is applied to the neuron output z. Expanding this interaction logic for the rest of the neurons, we get the DL layer. The layer where all possible connections between input nodes and output nodes are introduced is called the “Dense layer.” In order to learn more complex features and prevent overfitting, the too close fitting of the model to a lim ited set of data points in the training dataset, another type of layers was introduced such as the “Convolution layer,” “Pool ing layer” and “Dropout layer.” Given the DL model built from such layers and the representative dataset describing the prob lem we can solve the weights optimization task by using one of the optimization algorithms, e.g., Gradient Descent (GD). GD is used to find a minimum of the cost function by itera tively moving in the direction of steepest descent. It is used due to computational limitations we meet trying to solve the optimization task analytically. The cost function quantifies the error between predicted and the ground truth labels. By cal culating the derivative of the error with respect to each neural network weight we obtain the individual gradients, which are subsequently used to update the weights for all correspond ing neuron connections. The described procedure represents 1 cycle of the neural network (NN) training process. During the model training process every image from the training data set contributes to the weights optimization. Thereby, the DL model learns to solve the problem directly from data. Finding and classification of fractures on radiographs and CT images with high sensitivity and specificity can be assisted or even replaced by the automated DL system with high accu racy. Given a few thousand images we can address several problems with DL. Using such models as VGG16 (Simonyan and Zisisserman 2015), Inception V3 (Szegedy et al. 2015),

Acta Orthopaedica 2020; 91 (2): 215–220

and Xception (Chollet 2016), we can classify the images, for example to detect whether there is a fracture, or even differen tiate between fracture types. Given the bounding box annota tions or labels for the regions of interest, we can train such models as ResNet (He et al. 2016), U-net (Ronneberger et al. 2015), Mask-RCNN (He et al. 2017), Faster-RCNN (Ren et al. 2015) for the fracture detection and segmentation problem. The mentioned DL architectures have been widely used in the DL community and have demonstrated their efficiency in solving such tasks (Ruhan et al. 2017, Li et al. 2018, Couteaux et al. 2019, Li et al. 2019, Lian et al. 2019, Zhu et al. 2019). Applications of AI in fracture detection A number of studies have demonstrated the application of deep learning in fracture detection (Brett et al. 2009, Olczak et al. 2017, Chung et al. 2018, Kim and MacKinnon 2018, Lindsey et al. 2018, Tomita et al. 2018, Adams et al. 2019, Urakawa et al. 2019). In a retrospective study by Kim and Mac Kinnon (2018), they aimed to identify the extent to which transfer learning from deep convolutional neural net works (CNNs), pre-trained on non-medical images, can be used for automated fracture detection on plain wrist radio graphs. Authors used the inception V3 CNN (Szegedy et al. 2015), which was originally trained on non-radiological images for the IMageNet Large Visual Recognition Chal lenge (Russakovsky et al. 2015). They used a training data set of 1,389 radiographs (manually labeled) to re-train the top layer of the inception V3 network for the binary classification problem. They achieved an AUC of 0.95 on the test dataset (139 radiographs). This demonstrated that a CNN model that has been pre-trained on non-medical images can be success fully applied to the problem of fracture detection on plain radiographs. Specificity and sensitivity reached 0.90 and 0.88 respectively. This level of accuracy surpasses previous com putational methods for automated fracture analysis such as segmentation, edge detection, feature extraction (such stud ies reported sensitivities and specificities in the range of 80–85%). Although this study provides proof of concept, a number of limitations remain. A small discrepancy was found between the training accuracy and the validation accuracy at the end of the training process. This was likely to reflect overfitting. There are several strategies that can be used to minimize overfitting. One strategy would be to use automated segmentation of the most appropriate region of interest; the pixels outside of the region of interest would be cropped from the image so that irrelevant features would not influence the training process. Another strategy to minimize overfitting would be the introduction of advanced augmentation tech niques. In addition (too small < [1000:10000]) study popula tion size is often a limiting factor in machine learning field. A large sample corresponds to a more accurate reflection of a true population (Lindsey et al. 2018). A similar study by Chung et al. (2018) aimed to evalu ate the ability of deep learning to detect and classify proxi

217

mal humerus fractures using plain AP shoulder radiographs. Results of the CNN network were compared with the assess ment of specialists (general physicians, orthopedic surgeons, and radiologists). Their total dataset consisted of 1,891 plain AP radiographs and they used a pre-trained ResNet-152 model, which was fine-tuned to their proximal humerus frac ture datasets. The trained CNN showed high performance in distinguishing normal shoulders from proximal humerus fractures. In addition, promising results were found for clas sifying fracture type based on plain AP shoulder radiographs. The CNN exhibited superior performance to that of general physicians and general orthopedic surgeons, and similar per formance to that of shoulder specialized orthopedic surgeons. This indicates the possibility of automated diagnosis and clas sification of proximal humerus fractures and other fractures or orthopaedic diseases diagnosed accurately using plain radio graphs (Chung et al. 2018). The retrospective study by Tomita et al. (2018) aimed to eval uate the ability of deep learning to detect osteoporotic vertebral fractures (OVF) on CT scans and developed a machine learning approach, fully powered by a deep neural network framework, to automatically detect OVFs on CT scans. For their OVF detection system, they used a system that consisted 2 major components: (1) a CNN-based feature extraction module; and (2) an RNN module to aggregate the extracted features and make the final diagnosis. For the processing and extraction of features from CT scans they used a deep residual network (ResNet) (He et al. 2016). Their training dataset consisted of 1,168 CT scans; their validation set consisted of 135 CT scans and their test set consisted of 129 CT scans. The performance of their proposed system on an independent test set matched the level performance of practicing radiologists in both accu racy and F1 (mean of precision and recall) score (Tomita et al. 2018). This automatic detection system has the potential to reduce the time and the manual burden on radiologists of OVF screening, as well as reducing false-negative errors arising in asymptomatic early stage vertebral fracture diagnoses (Tomita et al. 2018). A summary of clinical studies involving computeraided fracture detection is given in the Table. Value of deep learning in radiology/orthopedic traumatology As seen from the examples of deep learning in radiology described above, there are potential benefits to the develop ment and integration of deep learning systems in everyday practice, in fracture detection as well as fracture characteriza tion tasks (Figure 3). In general, using deep learning as an adjunct to standard practices within radiology has the poten tial to improve the speed and accuracy of diagnostic testing while decreasing workforce due to offloading human radiolo gists from time-intensive tasks. Alongside that, deep learning systems are subject to some of the pitfalls of human-based diagnosis such as inter- and intra-observer variance. Deep learning, applied in academic research settings, can at least

218

Acta Orthopaedica 2020; 91 (2): 215–220

Figure 3. Deep learning aided workflow in fracture detection

Summary of clinical studies involving computer-aided fracture detection Reference

Region of interest Modality

Conclusion

Olczak et al. 2017 Wrist/Hand/ Radiographs This study supports the use of orthopaedic radiographs Ankle of artificial intelligence, which can perform at a human level Kim et al. 2018 Wrist Radiographs The AUC scores for this test were comparable tostate-of- the-art providing proof of concept for transfer learning from CNNs in fracture detection on plain radiographs Chung et al. 2018 Proximal Radiographs The use of artificial intelligence can accurately detect and humerus classify proximal humerus fractures on plain shoulder AP radiographs Heimer et al. 2018 Skull CT Classification based on the existence of skull fractures on CMIPs with deep learning is feasible Lindsey et al. 2018 Wrist Radiographs Deep learning methods are a mechanism by which senior medical specialists can deliver their expertise to generalists on the front lines of medicine, thereby providing substantial improvements to patient care Tomita et al. 2018 Pelvis CT The proposed system will assist and improve OVF diagnosis in clinical settings by pre-screening routine CT examinations and flagging suspicious cases prior to review by radiologists Pranata et al. 2019 Calcaneus CT The feasibility using deep CNN and SURF for computeraided classification and detection of the location of calcaneus fractures in CT images Adams et al. 2019 Pelvis Radiographs As impressive as recognising fractures is for a DCNN, similar learning can be achieved by top-performing medically naïve humans with less than 1 hour of perceptual training

Performance (metric) 0.83 (accuracy) 0.95 (AUC) 0.90 (sensitivity) 0.88 (specificity) Detection: 0.96 (accuracy) 1 (AUC) 0.99 (sensitivity) 0.97 (specificity) Classification: 0.65–0.86 (accuracy) 0.90–0.98 (AUC) 0.88–0.97 (sensitivity) 0.83–0.94 (specificity) 0.97 (AUC) 0.91 (sensitivity) 0.88 (specificity) 0.97 (AUC) on Test set1 0.98 (AUC) on Test set2

0.89 (accuracy) 0.91 (F1 score) 0.98 (accuracy)

0.91 (accuracy) 0.98 (AUC)

Abbreviations: CT = computed tomography; AUC = area under curve; CNN = convolutional neural network; AP = plain anteroposterior; CMIP = curved maximum intensity projections; OVF = Osteoporotic vertebral fractures; SURF = speeded-up robust features; DCNN = deep convolutional neural networks.

Acta Orthopaedica 2020; 91 (2): 215–220

match and sometimes exceed human performance in fracture detection and classification on plain radiographs and CT scans. Combining deep learning with a radiomics approach Radiomics is a method that extracts large amount of predefined quantitative features from medical images beyond the level of detail accessible to the human eye. Deep learning learns from the entire image, whereas radiomics characterizes only the region of interest of a particular disease. Therefore, it is our opinion that deep learning and radiomics provide com plementary imaging biomarkers. Furthermore, as radiomics is more stable in the face of smaller datasets, it is desirable to include these features in models to hedge against the possible overfitting of deep learning networks. Future directions The inclusion of artificial intelligence in decision support systems has been debated for decades (Kahn 1994). As appli cations of artificial intelligence in radiology/orthopedic trau matology will increase there are several areas of interest that we believe will hold significant value in the future (Brink et al. 2017). There is a consensus that inclusion of AI in radi ology/image-based disciplines would enhance diagnostic accuracy (Recht and Bryan 2017). However, there is also a consensus that such tools need to be carefully investigated and interpreted, before integration into clinical decisionsupport systems. A future challenge to address will be the radiologists–AI relationship. Jha and Topol (2016) suggested that AI can be used for redundant pattern-recognition tasks, while radiolo gists focus on cognitively challenging tasks. At large, radiolo gists would need to have a basic understanding of AI and AIbased tools; however, these tools would not replace radiolo gists’ work, and their role would not be limited to interpreting AI findings. Rather, AI tools can be used as a complementary tool to confirm/validate radiologists’ doubts and decisions (Liew 2018). Further research regarding radiologists–AI rela tionship is needed in order to properly integrate these disci plines, including research on how to train radiologists to use AI tools and interpret their results. AI systems must continue to expand their library of clinical applications. As seen in this review, there are several promis ing studies that demonstrate how AI can improve our perfor mance on clinical tasks such as fracture detection on radio graphs and CT scan, including fracture classifications and treatment decision support. Conflict of interest The authors declare that they have no conflict of interest. All authors have participated in this article. All authors have read and have approved the final version of the manuscript. There is no funding source. Acta thanks Max Gordon, Seppo Koskinen, and Rajkumar Saini for help with peer review of this study.

219

Adams M, Chen W, Holcdorf D, McCusker M W, Howe P D, Gaillard F. Computer vs human: deep learning versus perceptual training for the detection of neck of femur fractures. J Med Imaging Radiat Oncol 2019; 63: 27-32. Brett A, Miller C G, Hayes C W, Krasnow J, Ozanian T, Abrams K, Block J E, van Kuijk C. Development of a clinical workflow tool to enhance the detection of vertebral fractures: accuracy and precision evaluation. Spine 2009; 34: 2437-43. Brink J A, Arenson R L, Grist T M, Lewin J S, Enzmann D. Bits and bytes: the future of radiology lies in informatics and information technology. Eur Radiol 2017; 27: 3647-3651. Cha K H, Hadjiiski L, Samala R K, Chan H P, Caoili E M, Cohan R H. Uri nary bladder segmentation in CT urography using deep-learning convolu tional neural network and level sets. Med Phys 2016; 43: 1882. Chen C, Seff A, Kornhauser A, Xiao J. Deepdriving: learning affordance for direct perception in autonomous driving. Conference: IEEE International Conference on Computer Vision (ICCV); 2015. Chen H, Zhang Y, Kalra M K, Lin F, Chen Y, Liao P, Zhou J, Wang G. Lowdose CT with a residual encoder-decoder convolutional neural network. IEEE Trans Med Imaging 2017; 36: 2524-35. Chollet F. Xception: deep learning with depthwise separable convolutions. PDF, arxiv.org [cs.CV]; 2016. 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 classifica tion of the proximal humerus fracture by using deep learning algorithm. Acta Orthop 2018; 89: 468-73. Cireşan D, Meier U, Masci J, Schmidhuber J. Multi-column deep neural net work for traffic sign classification. Neural Netw 2012; 32: 333-8. Couteaux V, Si-Mohamed S, Nempont O, Lefevre T, Popoff A, Pizaine G, Villain N, Bloch I, Cotten A, Boussel L. Automatic knee meniscus tear detection and orientation classification with Mask-RCNN. Diagn Interv Imaging 2019; 100: 235-42. Dou Q, Yu L, Chen H, Jin Y, Yang X, Qin J, Heng P A. 3D deeply supervised network for automated segmentation of volumetric medical images. Med Image Anal 2017; 41: 40-54. Esteva A, Kuprel B, Novoa R A, Ko J, Swetter S M, Blau H M, Thrun S. Der matologist-level classification of skin cancer with deep neural networks. Nature 2017; 542: 115–118. Faculty of Clinical Radiology, Clinical Radiology UK workforce census 2016 report; 2016. Available at: http://www.rcr.ac.uk. Gao X W, Hui R, Tian Z. Classification of CT brain images based on deep learning networks. Comput Methods Programs Biomed 2017; 138: 49-56. Gulshan V, Peng L, Coram M, Stumpe M C, Wu D, Narayanaswamy A, Venu gopalan S, Widner K, Madams T, Cuadros J, Kim R, Raman R, Nelson P C, Mega J L, Webster D R. Development and validation of a deep learn ing algorithm for detection of diabetic retinopathy in retinal fundus photo graphs. JAMA 2016; 316: 2402-10. Hamet P, Tremblay J. Artificial intelligence in medicine. Metabolism 2017; 69S: S36-S40. He K, Zhang X, Ren S, Sun J. Deep Residual Learning for Image Recogni tion. arxiv.org [cs.CV]; 2016. He K, Gkioxari G, Dollár P, Girshick R. Mask R-CNN. arxiv.org [cs.CV]; 2017. Heimer J, Thali M J, Ebert L. Classification based on the presence of skull fractures on curved maximum intensity skull projections by means of deep learning. J Forensic Radiol Imaging 2018; 14: 16-20. Hosny A, Parmar C, Quackenbush J Schwartz L H, Aerts H J W L. Artificial intelligence in radiology. Nat Rev Cancer 2018; 18: 500-10. Jha S, Topol E J. Adapting to artificial intelligence: radiologists and patholo gists as information specialists. JAMA 2016; 316: 2353-4. Kahn C E. Artificial intelligence in radiology: decision support systems. Radiographics 1994; 14: 849-61. Kim D H, MacKinnon T. Artificial intelligence in fracture detection: transfer learning from deep convolutional neural networks. Clin Radiol 2018; 73: 439-45.

220

Lakhani P, Sundaram B. Deep learning at chest radiography: automated clas sification of pulmonary tuberculosis by using convolutional neural net works. Radiology 2017; 284: 574-82. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature 2015; 521: 436-44. Lee J G, Jun S, Cho Y W, Lee H, Kim G B, Seo J B, Kim N. Deep learning in medical imaging: general overview. Korean J Radiol 2017; 18: 570-84. Li X, Chen H, Qi X, Dou Q, Fu C W, Heng P A. H-Dense UNet: hybrid densely connected UNet for liver and tumor segmentation from ct volumes. IEEE Trans Med Imaging 2018; 37: 2663-74. Li R, Zeng X, Sigmund S E, Lin R, Zhou B, Liu C, Wang K, Jiang R, Freyberg Z, Lv H, Xu M. Automatic localization and identification of mitochondria in cellular electron cryo-tomography using faster-RCNN. BMC Bioinfor matics 2019; 20: 132. Lian S, Li L, Lian G, Xiao X, Luo Z, Li S. A global and local enhanced residual U-Net for accurate retinal vessel segmentation. IEEE/ACM Trans Comput Biol Bioinform 2019.doi: 10.1109/TCBB.2019.2917188. [Epub ahead of print] Liew C. The future of radiology augmented with artificial intelligence: a strat egy for success. Eur J Radiol 2018; 102: 152-6. Lindsey R, Daluiski A, Chopra S, Lachapelle A, Mozer M, Sicular S, Hanel D, Gardner M, Gupta A, Hotchkiss R, Potter H. Deep neural network improves fracture detection by clinicians. Proc Natl Acad Sci USA 2018; 115: 11591-6. McCulloch W S, Pitts W H. A logical calculus of the ideas immanent in ner vous activity. Bulletin of Mathematical Biophysics 1943; 5:115-33. Mnih V, Kavukcuoglu K, Silver D, Rusu A A, Veness J, Bellemare M G, Graves A, Riedmiller M, Fidjeland A K, Ostrovski G, Petersen S, Beattie C, Sadik A, Antonoglou I, King H, Kumaran D, Wierstra D, Legg S, Has sabis D. Human-level control through deep reinforcement learning. Nature 2015; 518: 529-33. Moravčík M, Schmid M, Burch N, Lisý V, Morrill D, Bard N, Davis T, Waugh K, Johanson M, Bowling M. DeepStack: expert-level artificial intelligence in heads-up no-limit poker. Science 2017; 356: 508-13. National Institute for Health and Environment (Rijksinstituut voor volksge zondheid en milieu [RIVM]); 2016. Available at: https: //www.rivm.nl/ medische-stralingstoepassingen/trends-en-stand-van-zaken/diagnostiek/ computer-tomografie/trends-in-aantal-ct-onderzoeken. Olczak J, Fahlberg N, Maki A, Razavian A S, Jilert A, Stark A, Sköldenberg O, Gordon M. Artificial intelligence for analyzing orthopedic trauma radio graphs. Acta Orthop 2017; 88: 581-6. Pranata Y D, Wang K C, Wang J C, Idram I, Lai J Y, Liu J W, Hsieh I H. Deep learning and SURF for automated classification and detection of calcaneus fractures in CT images. Comput Methods Programs Biomed 2019; 171: 27-37. Recht M, Bryan R N. Artificial intelligence: threat or boon to radiologists? J Am Coll Radiol 2017; 14: 1476-80. Ren S, He K, Girshick R, Sun J. Faster R-CNN: towards real-time object detection with region proposal networks. arxiv.org [cs.CV]; 2015. Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for bio medical image segmentation. Lecture Notes in Computer Science 2015; 234-41. Available from http: //dx.doi.org/10.1007/978-3-319-24574-4 28.

Acta Orthopaedica 2020; 91 (2): 215–220

Roth H R, Oda H, Zhou X, Shimizu N, Yang Y, Hayashi Y, Oda M, Fujiwara M, Misawa K, Mori K. An application of cascaded 3D fully convolutional networks for medical image segmentation. Comput Med Imaging Graph 2018; 66: 90-9. Ruhan S, Owens W, Wiegand R, Studin M, Capoferri D, Barooha K, Greaux A, Rattray R, Hutton A, Cintineo J, Chaudhary V. Intervertebral disc detec tion in X-ray images using faster R-CNN. Conf Proc IEEE Eng Med Biol Soc 2017; 564-7. Russakovsky O, Deng J, Su H. ImageNet Large Scale Visual Recognition Challenge. IJCV paper | bibtex | paper content on arxiv | attribute annota tions; 2015. Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition. Published as a conference paper at ICLR; 2015. Szegedy C, Vanhoucke V, Loffe S. Rethinking the Inception Architecture for Computer Vision. arxiv.org [cs.CV]; 2015. Tang A, Tam R, Cadrin-Chênevert A, Guest W, Chong J, Barfett J, Chepelev L, Cairns R, Mitchell J R, Cicero M D, Poudrette M G, Jaremko J L, Rein hold C, Gallix B, Gray B, Geis R; Canadian Association of Radiologists (CAR) Artificial Intelligence Working Group. Canadian Association of Radiologists white paper on artificial intelligence in radiology. Can Assoc Radiol J 2018; 69: 120-35. 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 diabe tes. JAMA 2017; 318: 2211-23. Tomita N, Cheung YY, Hassanpour S. Deep neural networks for automatic detection of osteoporotic vertebral fractures on CT scans. Comput Biol Med 2018; 98: 8-15. Tompson J, Jain A, LeCun Y, Bregler C. Joint training of a convolutional network and a graphical model for human pose estimation. Advances in Neural Information Processing Systems 2014; 27: 1799–1807. Tran G S, Nghiem T P, Nguyen V T, Luong C M, Burie J C. Improving accu racy of lung nodule classification using deep learning with focal loss. J Healthc Eng 2019; 5156416. Urakawa T, Tanaka Y, Goto S, Matsuzawa H, Watanabe K, Endo N. Detecting intertrochanteric hip fractures with orthopedist-level accu racy using a deep convolutional neural network. Skeletal Radiol 2019; 48: 239-44. Wolterink J M, Leiner T, Viergever M A, Isgum I. Generative adversarial net works for noise reduction in low-dose CT. IEEE Trans Med Imaging 2017; 36: 2536-45. Yang Y, Yan L F, Zhang X, Han Y, Nan H Y, Hu Y C, Hu B, Yan S L, Zhang J, Cheng D L, Ge X W, Cui G B, Zhao D, Wang W. Glioma grading on con ventional MR images: a deep learning study with transfer learning. Front Neurosci 2018; 12: 804. Zhu H, Shi F, Wang L, Hung S C, Chen M H, Wang S, Lin W, Shen D. Dilated dense U-Net for infant hippocampus subfield segmentation. Front Neuro inform 2019; 13: 30.

2/20 ACTA ORTHOPAEDICA

Element of success in joint replacement

Vol. 91, No. 2, 2020 (pp. 121–220)

www.heraeus-medical.com

Volume 91, Number 2, April 2020

TF-IORT200016.indd 1

07-02-2020 02:31:55