Acta Orthopaedica Vol. 90 Issue 1, February 2019 by Acta

1/19 ACTA ORTHOPAEDICA

Preventing and reducing periprosthetic joint infections

FFor anaerobees

Bone cement with 2 antib tibiotics Broaad sp spectrum trum m of of act ctivity For MRSA/MRRSE

,PSODQW Ƃ[DWLRQ DW High stabilit s ty

High gh iinitial antibiotic ibi release

Revision Vol. 90, No. 1, 2019 (pp. 1–95)

PRODUCTS & SOLUTIONS YOU CAN TRUST.

Volume 90, Number 1, February 2019 www.heraeus-medical.com IORT_90_01_Cover.indd 1

ISSN 1745-3674

11-01-2019 16:16:26

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

E DITORIAL O F FICE

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

EDITOR

THE FOUNDATION BOARD OF

Anders Rydholm Lund, Sweden

THE NORDIC O RTHOPAEDIC F EDERATION AND A CTA O RTHOPAEDICA

DEPUTY EDITOR

Peter A Frandsen Odense, Denmark CO-EDITORS

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

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

WEB EDITOR

Magnus Tägil Lund, Sweden S TATISTICAL EDITOR

Jonas Ranstam Lund, Sweden P RODUCTION MANAGER

Kaj Knutson Lund, Sweden

Vol. 90, No. 1, 2019

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. Annual Institutional Subscription, Volume 89, 2018 $1,173 £725 €939 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: Håkan Pårup P.O. Box 3255, SE-103 65 Stockholm, Sweden, Tel: +46 (0)8 440 80 40. Fax: +46 (0)8 440 80 50. E-mail: hakan.parup@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.

Airfreight and mailing in the USA by agent named Air Business Ltd, c/o Worldnet Shipping Inc., 156-15, 146th Avenue, 2nd Floor, Jamaica, NY 11434, USA. Periodicals postage paid at Jamaica NY 11431. US Postmaster: Send address changes to Acta Orthopaedica, Air Business Ltd, c/o Worldnet Shipping Inc., 156-15, 146th Avenue, 2nd Floor, Jamaica, NY 11434, USA. Subscription records are maintained at Taylor & Francis Group, 4 Park Square, Milton Park, Abingdon, OX14 4RN, United Kingdom.

Copyright © 2019 The Author(s). Published by Taylor & Francis on behalf of the Nordic Orthopedic Federation. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Non-Commercial License (https://creativecommons.org/licenses/by-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. 90, No. 1, February 2019 Editorial Time to restrict the use of p-values in Acta Orthopaedica

J Ranstam

Perspective Fast-track hip and knee arthroplasty – have we reached the goal?

T W Wainwright and H Kehlet

E Ekman, A Palomäki, I Laaksonen, M Peltola, U Häkkinen, and K Mäkelä

E M Bloemheuvel, L N van Steenbergen, and B A Swierstra

J J Tolk, R P A Janssen, C A C Prinsen, M C van der Steen, S M A Bierma Zeinstra, and M Reijman

D J Stockton, L M O’Hara, N N O’Hara, K A Lefaivre, P J O’Brien, and G P Slobogean

E J A Wiegers, C A Sewalt, E Venema, N W L Schep, J A N Verhaar, H F Lingsma, and D Den Hartog S B Bruun, I Petersen, N R Kristensen, D Cronin-Fenton, and A B Pedersen J S Jacobsen, K Søballe, K Thorborg, L Bolvig, S Storgaard Jakobsen, P Hölmich, and I Mechlenburg

Hip Early postoperative mortality similar between cemented and uncemented hip arthroplasty: a register study based on Finnish national data Dual mobility cups in primary total hip arthroplasties: trend over time in use, patient characteristics, and mid-term revision in 3,038 cases in the Dutch Arthroplasty Register (2007–2016) Measurement properties of the OARSI core set of performancebased measures for hip osteoarthritis: a prospective cohort study on reliability, construct validity and responsiveness in 90 hip osteoarthritis patients High rate of reoperation and conversion to total hip arthroplasty after internal fixation of young femoral neck fractures: a population-based study of 796 patients The volume–outcome relationship for hip fractures: a systematic review and meta-analysis of 2,023,469 patients Selective serotonin reuptake inhibitor use in hip fracture patients: a Danish nationwide prevalence study Patient-reported outcome and muscle–tendon pain after periacetabular osteotomy are related: 1-year follow-up in 82 patients with hip dysplasia Knee Relationship between outcome scores and knee laxity following total knee arthroplasty: a systematic review Nonagenarians qualify for total knee arthroplasty: a report on 329 patients from the Swedish Knee Arthroplasty Register 2000–2016 Weight affects survival of primary total knee arthroplasty: study based on the Danish Knee Arthroplasty Register with 67,810 patients and a median follow-up time of 5 years Similar polyethylene wear between cemented and cementless Oxford medial UKA: a 5-year follow-up randomized controlled trial on 79 patients using radiostereometry Predictors of return to desired activity 12 months following unicompartmental and total knee arthroplasty Miscellanous Limb lengthening and deformity correction with externally controlled motorized intramedullary nails: evaluation of 50 consecutive lengthenings Suggestion for new 4.4 mm pubo-femoral distance cut-off value for hip instability in lateral position during DDH screening Correspondence High failure rate after internal fixation and beneficial outcome after arthroplasty in treatment of displaced femoral neck fractures in patients between 55 and 70 years Information to authors (see http://www.actaorthop.org/)

33 40

A Kappel, M Laursen, P T Nielsen, and A Odgaard

E A Sezgin, O Robertsson, A W-Dahl, and L Lidgren

D Gøttsche, K Gromov, P H Viborg, E V Bräuner, A B Pedersen, and A Troelsen

K Horsager, F Madsen, A Odgaard, C F Jepsen, L Rømer, P W Kristensen, B L Kaptein, K Søballe, and M Stilling

A D Harbourne, M T Sanchez-Santos, N K Arden, and S R Filbay, on behalf of the COAST Study Group

J Horn, I Hvid, S Huhnstock, A B Breen, and H Steen

H-C Husum, M B Hellfritzsch, N Hardgrib, B Møller-Madsen, and O Rahbek

M Hedström, W Ekström, A Al-Ani, and P Campenfeldt versus S Bartels, J-E Gjertsen, F Frihagen, C Rogmark, and S E Utvåg

Thank you, reviewers! Acta Orthopaedica thanks all reviewers for their work with manuscripts during year 2018. We depend on your work to keep and increase the quality of Acta. Your effort is much appreciated!

Adolfsson, Lars Evert Alt, Volker Arendt, Elizabeth A Arts, Chris Aunan, Eirik Backteman, Torsten Bandholm, Thomas Belfrage, Ola Berg, Hans Bergh, Kåre Beverland, David Birch, John Gerard Bisseling, Pepijn Björnsson Hallgren, Hanna Cecilia Bobak, Peter Bohm, Eric Richard Bolder, Stefan Bos, Pieter K Breen, Anne Berg Brismar, Harald Brix, Michael Broström, Eva W Bruzzone, Matteo Brüggemann, Anders Bråten, Martinus Burger, Koert Chantelau, Ernst Christensson, Bertil Clauss, Martin Cobb, Justin Cordero-Ampuero, José Crawford, Ross W Dale, Håvard Davis, Naomi Delaunay, Christian P Dijkstra, Sander Ding, Ming Dreher, Thomas Eastwood, Deborah M Englund, Martin Erhart, Jochen Eriksson, Karl Eskelinen, Antti Etkin, Caryn Ettema, Harmen B Felländer-Tsai, Li

Flivik, Gunnar Frihagen, Frede Furnes, Ove Gargiulo, Paolo Gehrchen, Martin Geijer, Mats Gerdhem, Paul Gillam, Marianne Hansen Gjertsen, Jan-Erik Gordon, Max Gottliebsen, Martin Graves, Stephen Ellis Gromov, Kirill Gudnason, Asgeir Hallan, Geir Hamadouche, Moussa Hansen, Torben B Haverkamp, Daniël Hedin, Hanne Hedlund, Håkan Hedman, Leif Rune Hedström, Margareta Heeg, Minne Henricson, Anders Henriksson, Marketta Hindsø, Klaus Holmberg Jørgensen, Peter Holsgaard-Larsen, Anders Horn, Joachim Houlihan-Burne, David Husted, Henrik Hägglund, Gunnar Ignatius, Anita Ilchmann, Thomas Inacio, Maria Carolina Itayem, Raed Jakobsen, Thomas Jensen, Claus Munk Johnsen, Lars Gunnar Jämsen, Esa Järvelin, Jutta Kamrad, Ilka Kanatli, Ulunay Kehlet, Henrik Keidar, Zohar Kennedy, Oran

Khan, Sameer Khalid King, Madeleine Kjærsgaard-Andersen, Per Kjeldgaard Pedersen, Line Knupp, Markus Knutson, Kaj Kopylov, Philippe Kort, Nanne Koster, Lennard Kraus, Tanja Krettek, Christian Kristensen, Pia Kjær Kuijer, Paul Kvernmo, Hebe Désirée Kärrholm, Johan Labek, Gerold Laende, Elise Lainiala, Olli Samuli Landgren, Marcus Lange, Jeppe Larsen, Morten Schultz Launonen, Antti Laursen, Mogens Lavand´homme, Patricia Lie, Stein Atle Lind, Martin Lindberg-Larsen, Martin Lindström, Maria C Little, David Lohmander, Stefan Louwerens, Jan Willem Lübbeke, Anne Maasalu, Katre Madanat, Rami Madsen, Jan Erik Maffulli, Nicola Malviya, Ajay Mann, Kenneth Marsell, Richard Martin, Hal Martinez-Carranza, Nicolas Mathijssen, Nina Mattila, Ville Mayr, Johannes McGuigan, Fiona McNally, Martin

Mertens, Fredrik Michaëlsson, Karl Mikkelsen, Kim Lyngby Miller, Freeman Millis, Michael Brian Mohaddes, Maziar Moseley, Colin F Myrseth, Lars Eldar Mäkelä, Keijo T Märtson, Aare Nagoya, Satoshi Narayan, Badri Nelson, David L Nemeth, Banne Nieboer, Daan Niemeläinen, Mika Nietosvaara, Yrjänä Nieuwenhuijse, Marc J Nijhof, Marc Nilsson, Kjell G Nordsletten, Lars Oostenbroek, Hubert Jan Overgaard, Søren Palm, Henrik Parker, Martyn Parsch, Klaus Dieter Pedersen, Alma B Pedersen, Niels Wisbech Penning, Ludo Ivar Folmer Peters, Anil Pijls, Bart G Pitkänen, Mikko Ponzer, Sari Poolman, Rudolf W

Pratt, Nicole Pronk, Yvette Pua, Yong Hao Radtke, Andreas Raeder, Johan Rahbek, Ole Ranstam, Jonas Rasmussen, Jeppe Vejlgaard Rasmussen, Sten Riad, Jacques Roberts, David Rogmark, Cecilia Rolf, Christer Rolfson, Ola Romijn, Marc G Rossvoll, Ivar Ryd, Leif Röhrl, Stephan Maximilian Røise, Olav Sakkers, Ralph Salomonsson, Björn Sandelin, Henrik Sayed-Noor, Arkan S Sayers, Adrian Schaap, Gerard Schep, Niels Schilcher, Jörg Schmidt, Andrew Sinikumpu, Jaakko Sköldenberg, Olof Smulders, Katrijn Solberg, Lene Bergendal Soriano, Alex Stark, André

Stefánsdóttir, Anna Söder, Stephan Tannast, Moritz Tarasevicius, Sarunas Terjesen, Terje Terry, Samantha Tjernström, Björn Harald Trebse, Rihard Tsagozis, Panagiotis W-Dahl, Annette Wadsten, Mats Wagner, Christof Wallander, Henrik Van der Heide, Huub J L van Hamersveld, Koen van Hooff, Miranda L van Susante, Job Wanivenhaus, Axel Varnum, Claus Weidenhielm, Lars Weiss, Rüdiger J Vendittoli, Pascal-André Verhaar, Jan A N Verheyen, Cees C P M Whitehouse, Sarah Wiig, Ola Windhager, Reinhard Witchel, Selma Vogely, H Charles Wolterbeek, Nienke von Schewelov, Thord Zijlstra, Wierd P

Acta Orthopaedica 2019; 90 (1): 1–2

Editorial

Time to restrict the use of p-values in Acta Orthopaedica

Several scientific journals have banned p-values, not because p-values are bad when used correctly, but because they are notoriously misused (Wasserstein and Lazar 2016). What is the problem with the p-value? The problem with the p-value is not the p-value itself. The problem is ignorance about statistical inference, i.e., about the principles for using empirical observations and statistical reasoning to arrive at scientifically sound conclusions. An overwhelming majority of the authors of manuscripts submitted to medical journals believe, or at least seem to believe, that the p-value is a descriptive measure of importance regarding some aspect of an analyzed dataset: the lower the p-value, the stronger the effect. The truth of the matter is, however, that the p-value measures neither effect nor importance; it measures uncertainty. P-values are developed for performing rational generalization of findings, not for describing data. Statistical hypothesis testing is performed on the basis of a statistical (null) hypothesis that specifies the properties of the population from which the studied data are collected and to which the generalization is made. The p-value is defined as the probability of drawing a random sample from this population being at least as unlikely as the observed one, given that the null hypothesis is true. For example, a p-value of 0.05 implies that the probability of drawing such a random sample is 1 in 20, or 5%, if the null hypothesis is correct. It is then a reasonable conclusion that the data and the null hypothesis are incompatible. It is a common misunderstanding that the p-value indicates the size or importance of an observed effect. This is not the case. Whether or not a p-value is clinically important depends on the importance of the null hypothesis, not on the size of the p-value. Furthermore, as an example of random variation, while a major beneficial effect of a treatment in a population of patients with a specific disease may turn out to be statistically non-significant in an observed series of consecutive patients, a less effective treatment can be statistically significant. It is also often believed that the p-value indicates something about the truth of the null hypothesis, or about the probability that the observed data have been caused by chance alone. Again, this is not the case. The p-value is simply based on the assumption that the null hypothesis is true. The problems are, however, even greater than what these misunderstandings reflect. The null hypothesis is almost

always believed to be directly related to the author’s research question, and the relation is assumed to be so unambiguous and straightforward that the null hypothesis does not need to be presented, let alone scientifically motivated. The truth is that most, if not all, medical research questions include problems with far greater complexity than can be solved using a single null hypothesis. The links between the research question and its answer need to be developed prior to the statistical analysis in the form of a study design, accounted for in the statistical analysis, and explained and motivated to the reader in the study report. This represents a major intellectual challenge and is too often ignored or inadequately performed. Instead, it is not uncommon for manuscripts to present hundreds of unstructured p-values in order to find the answer to the research question. These p-values do not all represent equally relevant null hypotheses, and already with a small number of hypothesis tests, the effects of multiplicity issues have deleterious consequences for the reliability of the tests’ outcome. Furthermore, even in the apparently simple situation of comparing 2 groups of patients with respect to the mean value of 1 specific variable, many different null hypotheses can be defined: 3 traditional null hypotheses (1 2-sided and 2 1-sided, 1 in each direction), a number of 1-sided non-inferiority null hypotheses differing with respect to their non-inferiority margin, and a number of 2-sided equivalence null hypotheses differing with respect to their equivalence margin. Just testing 1 of these null hypotheses without motivating it may, perhaps, seem objective but is in fact subjective and has the potential to mislead the reader. One consequence of the p-value misunderstandings and ignorance of statistical inference is the dichotomization of findings as either “significant” or “not significant” without any consideration of what is tested and of the risk of false positive and negative outcomes. The interpretation of these 2 categories as indicating “important” and “no difference” is wide off the mark. Whether a specific p-value can be interpreted as indicating importance depends on several things, among them the medical or biological relevance of the research question, the soundness of the study design, and the definition of the null hypothesis. It takes more than a “p < 0.05” to show scientific importance, and “p > 0.05” is not an indication of “no difference”; it reflects absence of evidence but not evidence of absence.

© 2019 The Author(s). Published by Taylor & Francis on behalf of the Nordic Orthopedic Federation. This is an Open Access article distributed under the terms of the Creative Commons Attribution-Non-Commercial License (https://creativecommons.org/licenses/by/4.0) DOI 10.1080/17453674.2018.1536526

The misunderstandings are, unfortunately, ubiquitous in medical research. It is not hard to find examples of this: 3 are presented below. Example 1. Confusing sample and population The aim of a study is described as to study a general phenomenon, e.g., “to find risk factors for hip fracture among subjects older than 65 years of age.” However, instead of generalizing the findings to all subjects over 65, the author’s conclusion is merely a brief description of what has been observed in his hospital series, referring to statistical significance as an indication of importance, e.g., “among the studied patients, smoking was a significant risk factor.” However, the null hypotheses underlying the presented p-values are all about the entire population of subjects over 65, including future patients, not about the studied hospital series. Example 2. Testing an irrelevant null hypothesis In a matched case-control study, p-values are used for evaluating the pairwise differences in matching variables between matched cases and controls. The tests are performed with the confused purpose to assess whether the observed differences of matched cases and controls are clinically important. The underlying null hypotheses, however, do not refer to the observed cases and controls but to the infinite population represented by the matched cases and controls, and for this matching is not a relevant issue. Example 3. Testing baseline imbalance after randomization P-values are wrongly used for evaluating baseline imbalance after randomization in a clinical trial. The purpose is to investigate if the randomization “was successful”, i.e., resulted in groups with the same baseline characteristics. Randomization is, however, used to prevent systematic errors in the generalization of the trial’s results (what the outcome means for patients in general), not to eliminate random imbalance among the randomized patients. Random imbalance of known prognostic factors can be avoided by stratifying the randomization on these factors. Many more examples can be presented, but the ones above should be sufficient to describe the presented phenomena. Can we just skip the p-value? Yes, and orthopedic research would benefit from it. Confidence intervals represent a superior way to present generalization uncertainty. Confidence intervals have the advantage of measuring the uncertainty of the size of an estimated effect, which p-values do not. And in contrast to when using p-values, questions regarding the clinical significance of, and empirical support for, a specific conclusion can be directly answered by the effect sizes that are included in, or excluded from, a confidence interval.

Acta Orthopaedica 2019; 90 (1): 1–2

2 simple principles To start a transition to p-value-free manuscripts, Acta Orthopaedica will enforce a policy of zero tolerance vis-à-vis p-value misconceptions. Authors who wish to publish manuscript with p-values must from now on comply with 2 principles for concluding whether or not scientifically important differences exist: 1. A statistically non-significant test is not sufficient to claim “no difference.” To show “no difference,” a smallest clinically relevant size of the difference (it might be 0) must be defined. If all clinically relevant differences are excluded from the difference’s 95% confidence interval a “no difference” conclusion is reasonable. 2. A statistically significant test is not necessarily related to a clinically important difference. The importance of the tested null hypothesis must be motivated using other arguments than the p-value, and a smallest clinically relevant difference (it might be 0) must be defined, and if the difference’s 95% confidence interval excludes all clinically irrelevant differences, a conclusion about the existence of a clinically important difference is reasonable. The required clinical definitions, motivations, and explanations should be presented in the manuscript in non-technical terms. For example (from Paavola et al. 2018), the null hypothesis and clinically relevant difference can be described: “Main outcome measure: Shoulder pain at rest ... (visual analogue scale (VAS) from 0 to 100, with 0 denoting no pain), at 24 months. The threshold for minimal clinically important difference was set at 15.” The result is presented as: “In the primary intention to treat analysis (ASD versus diagnostic arthroscopy), no clinically relevant between group differences were seen... The ... difference between groups ... in pain VAS was −5 (95% confidence interval −11 to 2) points.” The above example shows that confidence intervals are better than p-values, and that p-values are redundant when confidence intervals are presented. Arguments against replacing p-values with confidence intervals are usually motivated by a desire to be able to keep misusing p-values for showing “no difference” and “importance” without having to consider clinical relevance and without having to argue in clinical terms why an observed difference is important. The failure of such simplistic research is, however, clearly shown in the discussion on the reproducibility crisis of modern biomedical research. It is now time for a more serious approach. Jonas Ranstam Statistical editor Email: jonas.ranstam@med.lu.se Wasserstein R L, Lazar N A. The ASA’s statement on p-values: context, process, and purpose. American Statistician 2016; 70: 129-33. Paavola M, Malmivaara A, Taimela S, Kanto K, Inkinen J, Kalske J, Sinisaari I, Savolainen V, Ranstam J, Järvinen T L N. Subacromial decompression versus diagnostic arthroscopy for shoulder impingement: randomised, placebo surgery controlled clinical trial. BMJ 2018; 362: k2860.

Acta Orthopaedica 2019; 90 (1): 3–5

Perspective

Fast-track hip and knee arthroplasty – have we reached the goal?

Total hip arthroplasty (THA) and total knee arthroplasty (TKA) are common major surgical procedures that are often performed in older patients with complex comorbidities. Fast-track programs (or Enhanced Recovery after Surgery (ERAS) programs) have evolved during the past 20 years, and have been proven to reduce length of hospital stay (LOS), morbidity, and convalescence time, without an increase in readmission rates or compromising patient safety (Kehlet 2013, Khan et al. 2014, Berg et al. 2018, Petersen et al. 2018a). So far so good. However, even though patients may meet discharge criteria within 0–2 days, they have not completely recovered. In addition, despite the scientific evidence for enhanced recovery, widespread implementation is still lagging, since LOS is still around 4–6 days in many places after THA and TKA compared with 2–3 days or less in large epidemiological studies (Khan et al. 2014, Berg et al. 2018, Petersen et al. 2018a). Therefore, the challenge is to further understand the pathophysiological mechanisms of morbidity and recovery, and to optimize post-discharge functional outcomes, in order to prevent sub-acute problems turning into chronic problems. Pain during early postoperative recovery is multifactorial with several contributors been identified, e.g. postoperative inflammatory/immunological responses, sufficient pain management and organizational factors (Kehlet 2013, Gaudilliere et al. 2014). Optimal pain management is a prerequisite to enhance recovery and plenty of analgesic techniques have been described for THA and TKA, but without firm conclusions and recommendations (Karlsen et al. 2015, 2017, Soffin et al. 2018a, 2018b). Simple multimodal oral analgesia supplemented by a high dose of preoperative glucocorticoid is probably the simplest and safest option to enhance analgesia and early recovery (Kehlet and Lindberg-Larsen 2018), with use of local infiltration analgesia (LIA) in TKA but without the use of more specific peripheral nerve blocks. The latter seemingly provide better immediate postoperative pain relief, but have not led to shorter LOS, nor have they demonstrated long-lasting improved recovery. Since early ambulation is essential, the problem of early postoperative orthostatic intolerance is important, and studies have shown an undesirable shift in autonomic nervous system function towards increased parasympathetic function and loss of sympathetic stimulation, especially to the lower legs (Jans and Kehlet 2017). So far, the problem has not been solved by optimized fluid management, while some positive data using

an α-1 agonist (midodrine) needs further study (Jans and Kehlet 2017). Recently, more emphasis has been put on the role of preand postoperative anemia as risk factors for a prolonged hospital stay and re-admission after THA and TKA, clearly documenting the need for the effective preoperative diagnosis and treatment of anemia (Munoz et al. 2017). Intraoperative blood loss should be decreased by use of combined local and systemic tranexamic acid treatment, while the influence of postoperative anemia remains to be explored further especially in elderly patients and with a focus on cardiovascular complications and functional recovery. In addition, studies are required to evaluate the optimal transfusion triggers in specific highrisk patients, which unfortunately have not been included in previous THA and TKA transfusion guidelines due to lack of sufficient data (Munoz et al. 2017, 2018, Petersen et al. 2018a). Postoperative delirium and cognitive dysfunction are wellrecognized problems after surgery in general where the pathogenesis is again multifactorial and includes pain, use of opioids, sleep disturbances and inflammatory responses (Kehlet 2013). Recent large-scale data from fast-track THA and TKA have shown virtually no risk of early delirium (Petersen et al. 2017). Nevertheless, postoperative sleep disturbances (Kehlet 2013) are still present and may also be related to post-discharge functional recovery where the combination of pain, function and sleep disturbances are dominant and where future studies should clarify the relative role of each component (Krenk et al. 2010). The final challenge lies in safety as a further reduction of morbidity and mortality, which is well documented in fasttrack THA and TKA (Kehlet 2013, Khan et al. 2014, Jorgensen et al. 2016). An effort to construct a prediction risk including all conventional risk factors showed that this could be done with statistical significance, although the clinical significance was less due to the low numbers of complications, especially when separated between direct “medical” and “surgical” complications (Jorgensen et al. 2016). Part of the problem is that not only preoperative risk factors are important, but also the perioperative care regimes including avoidance of postoperative anemia. Although there is general agreement that fast-track THA and TKA decreases the risk of “medical” complications, the concern that early discharge may lead to more “surgical” complications (hip dislocation, wound problems, etc.) has not been supported by the literature (den Hartog

et al. 2013, Khan et al. 2014, Jorgensen et al. 2016, Sutton et al. 2016, Pamilo et al. 2018). Recent evidence points to patients with preoperative “psychiatric” conditions on psychopharmacological treatment – mostly with conventional antidepressants – as being at risk for prolonged LOS and more readmissions. Detailed analysis suggests that these recovery issues are related to the psychopharmacological treatment per se rather than the “psychiatric” condition, calling for future interventional studies in such patients (Gylvin et al. 2017). The risk of post THA/TKA thromboembolic complications is well known and has been a classical model in thromboembolic prophylactic studies. However, even the most recent guidelines (NICE 2018) are predominantly based on studies with prolonged LOS and thus not reflective of the benefits of early postoperative mobilization (Kjaersgaard-Andersen and Kehlet 2012). More recent prospective data from large cohort studies suggest that conventional prolonged thromboembolic prophylaxis may not be necessary within a successful enhanced recovery THA/TKA program including early mobilization (Kehlet 2013, Petersen et al. 2018b); this needs corroboration with further large-sized studies. Whilst early ambulation and reduced LOS has reduced the risk of complications, the early loss of muscle function may delay post-discharge recovery. Despite considerable efforts to improve rehabilitation, the use of preoperative exercise, conventional physiotherapy regimes, and earlier initiated and more intense postoperative strengthening regimes have all been found to have a limited effect in the “average” patient (Bandholm et al. 2018). Therefore, future work in fast-track THA and TKA must reveal which therapeutic interventions are effective, and in which patients with known preoperative indicators of likely delayed recovery such as pain status, frailty, psychological status, socioeconomic status, and unrealistic expectations of recovery. Although patient-reported outcome measures (PROMs) show improvement in the majority of fast-track THA and TKA patients, discrepancies are seen when compared with objective measures of functional performance and physical activity, both in the early and later recovery phase (Luna et al. 2017, 2018). Consequently, objective functional data are important from future studies, given the known increased healthcare costs and lower income levels of patients even after fast-track THA and TKA (Kjellberg and Kehlet 2016). Although LOS in several countries has been reported to be short (Kehlet 2013), the aim is to improve recovery to be able to go directly home and not via another institution such as nursing care facilities, rehabilitation homes etc. (Cram et al. 2018) where limited scientific information is available with regard to recovery interventions and results. The same applies to the more recent publications on outpatient THA and TKA, often from private and semi-private institutions. So far, this approach seems safe in selected patients without increasing risk of readmissions or complications (Vehmeijer et al. 2018).

Acta Orthopaedica 2019; 90 (1): 3–5

However, despite this progress, there are challenges to understand how to increase the percentage of patients recovered and discharged on the day of surgery, and for which patients that outpatient surgery may not equate to optimized care. Fasttrack protocols have been based on the concept of “first better – then faster,” so it could be that for some identified patients a planned longer stay in hospital is the best means of accelerating recovery and reducing complications, re-admissions, and morbidity. In summary, the fast-track THA and TKA approach has made major progress, but we have not achieved the final goal. Future challenges lie in: (1) preoperative prediction of highinflammatory responders, (2) further dose-finding or repeatdosing glucocorticoid or other anti-inflammatory agents in studies in hyper-inflammatory responders (Kehlet and Lindberg-Larsen 2018), (3) more focused studies on high-pain responders (preoperative opioid users, pain catastrophizers, sensitized patients, etc.) (Gilron et al. 2018), and (4) developing optimal post-discharge rehabilitation strategies. Only then we will have succeeded ino reaching the goal for the “pain and risk free” THA and TKA (Kehlet and Jorgensen 2016). Competing interests None. Thomas W Wainwright Orthopaedic Research Institute, Bournemouth University, UK Henrik Kehlet Section of Surgical Pathophysiology, Rigshospitalet, Copenhagen University, Denmark and The Lundbeck Foundation Centre for Fast-track Hip and Knee replacement, Copenhagen, Denmark. email: henrik.kehlet@regionh.dk

Bandholm T, Wainwright T W, Kehlet H. Rehabilitation strategies for optimisation of functional recovery after major joint replacement. J Exp Orthop 2018; 5(1): 44. Berg U, Bulow E, Sundberg M, Rolfson O. No increase in readmissions or adverse events after implementation of fast-track program in total hip and knee replacement at 8 Swedish hospitals: an observational before-and-after study of 14,148 total joint replacements 2011–2015. Acta Orthop 2018; 89 (5): 522–7. Cram P, Landon B E, Matelski J, Ling V, Stukel T A, Paterson J M, Gandhi R, Hawker G A, Ravi B. Utilization and short-term outcomes of primary total hip and knee arthroplasty in the United States and Canada: an analysis of New York and Ontario administrative data. Arthritis Rheumatol 2018; 70(4): 547-54. den Hartog Y M, Mathijssen N M, Vehmeijer S B. Reduced length of hospital stay after the introduction of a rapid recovery protocol for primary THA procedures. Acta Orthop 2013; 84(5): 444-7. Gaudilliere B, Fragiadakis G K, Bruggner R V, Nicolau M, Finck R, Tingle M, Silva J, Ganio E A, Yeh C G, Maloney W J, Huddleston J I, Goodman S B, Davis M M, Bendall S C, Fantl W J, Angst M S, Nolan G P. Clinical recovery from surgery correlates with single-cell immune signatures. Sci Transl Med 2014; 6(255): 255ra131.

Acta Orthopaedica 2019; 90 (1): 3–5

Gilron I, Carr D B, Desjardins P J, Kehlet H. Current methods and challenges for acute pain clinical trials. Pain Rep 2018: e647. Gylvin S H, Jorgensen C C, Fink-Jensen A, Gislason G H, Kehlet H. The role of psychiatric diagnoses for outcome after hip and knee arthroplasty. J Arthroplasty 2017; 32(12): 3611-5. Jans O, Kehlet H. Postoperative orthostatic intolerance: a common perioperative problem with few available solutions. Can J Anaesth 2017; 64(1): 10-15. Jorgensen C C, Petersen M A, Kehlet H. Preoperative prediction of potentially preventable morbidity after fast-track hip and knee arthroplasty: a detailed descriptive cohort study. BMJ Open 2016; 6(1): e009813. Karlsen A P, Geisler A, Petersen P L, Mathiesen O, Dahl J B. Postoperative pain treatment after total hip arthroplasty: a systematic review. Pain 2015; 156(1): 8-30. Karlsen A P, Wetterslev M, Hansen S E, Hansen M S, Mathiesen O, Dahl J B. Postoperative pain treatment after total knee arthroplasty: a systematic review. PLoS One 2017; 12(3): e0173107. Kehlet H. Fast-track hip and knee arthroplasty. Lancet 2013; 381(9878): 1600-2. Kehlet H, Jorgensen C C. Advancing surgical outcomes research and quality improvement within an enhanced recovery program framework. Ann Surg 2016; 264(2): 237-8. Kehlet H, Lindberg-Larsen V. High-dose glucocorticoid before hip and knee arthroplasty: to use or not to use—that’s the question. Acta Orthop 2018; 89 (5): 477-9. Khan S K, Malviya A, Muller S D, Carluke I, Partington P F, Emmerson K P, Reed M R. Reduced short-term complications and mortality following Enhanced Recovery primary hip and knee arthroplasty: results from 6,000 consecutive procedures. Acta Orthop 2014; 85(1): 26-31. Kjellberg J, Kehlet H. A nationwide analysis of socioeconomic outcomes after hip and knee replacement. Dan Med J 2016; 63(8): A5257. Kjaersgaard-Andersen P, Kehlet H. Should deep venous thrombosis prophylaxis be used in fast-track hip and knee replacement? Acta Orthop 2012; 83(2): 105-6. Krenk L, Rasmussen L S, Kehlet H. New insights into the pathophysiology of postoperative cognitive dysfunction. Acta Anaesthesiol Scand 2010; 54(8): 951-6. Luna I E, Kehlet H, Peterson B, Wede H R, Hoevsgaard S J, Aasvang E K. Early patient-reported outcomes versus objective function after total hip and knee arthroplasty: a prospective cohort study. Bone Joint J 2017; 99-B(9): 1167-75. Luna I E, Kehlet H, Wede H R, Hoevsgaard S J, Aasvang E K. Objectively measured early physical activity after total hip or knee arthroplasty. J Clin Monit Comput 2018 [Epub ahead of print].

Munoz M, Acheson A G, Auerbach M, Besser M, Habler O, Kehlet H, Liumbruno G M, Lasocki S, Meybohm P, Rao B R, Richards T, Shander A, SoOsman C, Spahn D R, Klein A A. International consensus statement on the peri-operative management of anaemia and iron deficiency. Anaesthesia 2017; 72(2): 233-47. Munoz M, Acheson A G, Bisbe E, Butcher A, Gómez-Ramírez S, Khalafallah A A, Kehlet H, Kietaibl S, Liumbruno G M, Meybohm P, Baikady R R, Shander A, So-Osman C, Spahn D R, Klein A A. An international consensus statement on the management of postoperative anaemia after major surgical procedures. Anaesthesia 2018; 73(11): 1418-31. NICE. NICE guideline [NG89]. Venous thromboembolism in over 16s: reducing the risk of hospital-acquired deep vein thrombosis or pulmonary embolism. National Institute for Health and Care Excellence; 2018. Available at: https://www.nice.org.uk/guidance/ng89/evidence/ Pamilo K J, Torkki P, Peltola M, Pesola M, Remes V, Paloneva J. Fast-tracking for total knee replacement reduces use of institutional care without compromising quality. Acta Orthop 2018; 89(2): 184-9. Petersen P B, Jorgensen C C, Kehlet H. Delirium after fast-track hip and knee arthroplasty: a cohort study of 6331 elderly patients. Acta Anaesthesiol Scand 2017; 61(7): 767-72. Petersen P B, Kehlet H, Jørgensen C C. Myocardial infarction following fasttrack total hip and knee arthroplasty—incidence, time course and risk factors: a prospective cohort study of 24,862 procedures. Acta Orthop 2018a; 89 (6): 603-9. Petersen P B, Kehlet H, Jørgensen C C. Safety of in-hospital only thromboprophylaxis after fast-track total hip and knee arthroplasty: a prospective follow-up study in 17,582 procedures. Thromb Haemost 2018b (in press). Soffin E M, Gibbons M M, Ko C Y, Kates S L, Wick E, Cannesson M, Scott M J, Wu C L. Evidence review conducted for the Agency for Healthcare Research and Quality Safety Program for Improving Surgical Care and Recovery: focus on anesthesiology for total knee arthroplasty. Anesth Analg 2018a [Epub ahead of print]. Soffin E M, Gibbons M M, Ko C Y, Kates S L, Wick E C, Cannesson M, Scott M J, Wu C L. Evidence Review Conducted for the Agency for Healthcare Research and Quality Safety Program for Improving Surgical Care and Recovery: focus on anesthesiology for total hip arthroplasty. Anesth Analg 2018b [Epub ahead of print]. Sutton J C, Antoniou J, Epure L M, Huk O L, Zukor D J, Bergeron S G. Hospital discharge within 2 days following total hip or knee arthroplasty does not increase major-complication and readmission rates. J Bone Joint Surg Am 2016; 98(17): 1419-28. Vehmeijer S B W, Husted H, Kehlet H. Outpatient total hip and knee arthroplasty: facts and challenges. Acta Orthop 2018; 89(2): 141-4.

Acta Orthopaedica 2019; 90 (1): 6–10

Early postoperative mortality similar between cemented and uncemented hip arthroplasty: a register study based on Finnish national data Elina EKMAN 1, Antton PALOMÄKI 1, Inari LAAKSONEN 1, Mikko PELTOLA 2, Unto HÄKKINEN 2, and Keijo MÄKELÄ 1 1 Department

of Orthopaedics and Traumatology, Turku University Hospital, Turku, Finland; 2 National Institute for Health and Welfare, Helsinki, Finland Correspondence: elina.ekman@tyks.fi ORCID: 0000-0003-1330-7418 Submitted 2018-06-05. Accepted 2018-11-18.

Background and purpose — Implant survival of cemented total hip arthroplasty (THA) in elderly patients is higher than that of uncemented THA. However, a higher mortality rate in patients undergoing cemented THA compared with uncemented or hybrid THA has been reported. We assessed whether cemented fixation increases peri- or early postoperative mortality compared with uncemented and hybrid THA. Patients and methods — Patients with osteoarthritis who received a primary THA in Finland between 1998 and 2013 were identified from the PERFECT database of the National Institute for Health and Welfare in Finland. Definitive data on fixation method and comorbidities were available for 62,221 THAs. Mortality adjusted for fixation method, sex, age group, and comorbidities among the cemented, uncemented, and hybrid THA was examined using logistic regression analysis. Reasons for cardiovascular death within 90 days since the index procedure were extracted from the national Causes of Death Statistics and assessed separately. Results — 1- to 2-day adjusted mortality after cemented THA was comparable to that of the uncemented THA group (OR 1.2; 95% CI 0.24–6.5). 3- to 10-day mortality in the cemented THA group was comparable to that in the uncemented THA group (OR 0.54; CI 0.26–1.1), and in the hybrid THA group (OR 0.64, CI 0.25–1.6). Pulmonary embolism or cardiovascular reasons as a cause of death were not overrepresented in the cemented THA group. Interpretation — Early peri- and postoperative mortality in the cemented THA group was similar compared with that of the hybrid and uncemented groups.

The early postoperative mortality after total hip arthroplasty (THA) is low and has been decreasing over the last few years (Aynardi et al. 2009, McMinn et al. 2012, Lalmohamed et al. 2014). 2 recent publications have indicated that 90-day mortality after primary THA performed for any indication is 0.7% (Hunt et al. 2013, Garland et al. 2015). Improvements in surgical techniques and implants, the introduction of low molecular weight heparins in the 1980s, and operative room sterility have significantly reduced mortality risks (Nurmohamed et al. 1992, Harris 2009). On the other hand, the surgery is now being performed on older patients who often have multiple comorbidities, which increase adverse outcomes (Mahomed et al. 2003, Bozic et al. 2012). Cementing has been used for decades for THA implant fixation with good implant survival rates in long term followup (Morshed et al. 2007, Mäkelä et al. 2014a). However, cemented fixation is associated with potential perioperative morbidity in the form of bone cement implantation syndrome where fat and bone marrow cause emboli during cement pressurizing into the pulmonary arteries and may lead to intra- or early postoperative hypotension, and even death of the patient (Donaldson et al. 2009). Even though survival of cemented implants in the elderly population is higher than the survival of uncemented implants, the fear of bone cement implantation syndrome is one of the reasons for the increased use of uncemented implants (Junnila et al. 2016). The leading cause of death after THA is cardiovascular (Berstock et al. 2014). We studied whether early postoperative mortality of patients treated with uncemented THA differed from that of patients treated with cemented or hybrid THA based on data from the PERFECT database maintained by the National Institute for Health and Welfare in Finland. We also assessed bone cement implantation syndrome and early cardiovascular mortality in this same population.

Acta Orthopaedica 2019; 90 (1): 6–10

Table 1. Background characteristics of patients. Values are frequency (percentage) unless otherwise stated Background characteristics

Cemented Uncemented Hybrid THA THA THA

Number of patients 23,636 Mean age 73.7 Men 8,277 (35) Hypertension 11,104 (47) Ischemic heart disease 4,234 (20) Atrial fibrillation 1,868 (7.9) Heart insufficiency 1,284 (5.4) Diabetes 2,130 (9.0) COPD and asthma 2,422 (10) Cancer 1,833 (7.8) Depression 1,522 (6.4) Parkinson’s disease 283 (1.2) Dementia 281 (1.2) Uremia 30 (0.1) Mental disorders 659 (2.8)

38,477 64.9 18,747 (49) 15,771 (41) 4,135 (11) 2,082 (5.4) 797 (2.1) 3,715 (9.7) 3,973 (10) 2,575 (6.7) 2,688 (7.0) 381 (1.0) 212 (0.6) 61 (0.2) 978 (2.5)

11,802 66.5 5,636 (48) 4,725 (40) 1,466 (12) 676 (5.7) 431 (3.7) 964 (8.2) 1,059 (9.0) 724 (6.1) 704 (6.0) 111 (0.9) 98 (0.8) 19 (0.2) 284 (2.4)

These results have not been adjusted via propensity score weighting. THA = total hip arthroplasty.

Patients and methods Study population The study population was identified from the Finnish Hospital Discharge Register (FHDR) using the 10th revision of the International Classification of Diseases (ICD-10) diagnosis codes M16.0 to M16.9, and the Finnish version of NOMESCO Classification Procedural Codes NFB30 (uncemented THA), NFB40 (hybrid THA when only the femoral stem has been cemented), or NFB50 (cemented THA). During the study period from January 1, 1998 to December 31, 2013, 73,915 patients were treated with THA for primary or secondary OA in Finland. Definitive data on fixation method and comorbidities were available for 62,221 THAs, which formed the final study population. All public and private hospitals in Finland are obliged to report all surgical procedures to the Finnish National Institute of Health and Welfare. The present study was based on the PERFECT (PERFormance, Efficiency, and Costs of Treatment Episodes) hip replacement database, which uses data from numerous registries such as the Hospital Discharge Register (maintained by the Finnish National Institute of Health and Welfare), cause of death statistics maintained by Statistics Finland, the Social Insurance Institution’s drug prescription register and drug reimbursement register, and the Finnish Arthroplasty Register. Data on comorbidities, on the use of residential care, patient ID number, provider ID number(s), age, sex, residential area codes, diagnosis, operation codes, date of admission, operation, and the date of discharge or death, whichever came first, were extracted from the PERFECT database (Peltola et al. 2011) (Table 1). The validity of the individual registries mentioned above has been studied. The Finnish Hospital Discharge Register

data have been compared with external audit data in 32 studies (Sund 2012). The coverage and positive predictive values have been over 90% in those studies. The prescription database data have been found to be in high concordance with self-reported medication (Haukka et al. 2007). To assess bone cement implantation syndrome and cardiovascular reasons separately as a cause of death, mortality reported with the associated diagnostic codes (codes I21 acute myocardial infarction, I25 ischemic heart disease, I26 pulmonary embolism, I50 heart failure, and I63 stroke in the ICD-10 classification) within 90 days since the index procedure were extracted from the national Causes of Death Statistics. The validity of the Finnish mortality statistics is reliable (Lahti and Penttilä 2003, Pajunen et al. 2005). The primary outcome used in this study was total mortality and secondary outcome cardiovascular mortality and mortality associated with pulmonary embolism. The patients were followed up for 1 year postoperatively. Statistics Mortality among the cemented, uncemented, and hybrid groups was examined using logistic regression analysis. The analysis was repeated for 365 outcomes that each described the status of the patient (alive/dead) on a certain day after the operation. In order to reduce the effects of confounding in this observational study, differences in distributions of observed covariates between the groups were adjusted: fixation method, sex, age group (< 50, 50–59, 60–69, 70–79, ≥ 80), comorbidities (Table 1), and the year of operation. In the model, treatment assignment (cemented/uncemented/hybrid) was the dependent variable and all observed background variables (Table 1) were independent variables, as the aim was to balance all observed covariates between the groups. 95% confidence intervals (CI) were calculated for adjusted mortality. Ethics, funding, and potential conflicts of interest The ethics committee of the Finnish National Institute for Health and Welfare (THL) approved the study (Dnro THL/127/5.05.00/2015). This research received no specific grant from any funding agency. The authors declare no conflicts of interest.

Results The use of cemented THA decreased in Finland during the study period, whereas the use of uncemented THA increased (Figure 1). The adjusted overall mortality or mortality associated with cardiovascular reasons or pulmonary embolism were similar between cemented THA and uncemented or hybrid THA at any of the studied time points (Figure 2). There were 9 deaths during days 1 and 2 in the cemented THA group, 4 in the uncemented THA group, and 0 in the hybrid group (Table 2). The 1- and 2-day adjusted mortality in the cemented THA

Acta Orthopaedica 2019; 90 (1): 6–10

Frequency

Relative and cumulative risk of death

5,000

5 Cemented Uncemented Hybrid

4,000

3,000

2,000

1,000

0 1998

2000

2002

2004

2006

Year

2008

2010

2012

183

273

365

Days after index operation

Figure 1. Annual numbers of cemented, uncemented, and hybrid THA in Finland during the study period.

Figure 2. Relative and cumulative risk of death in patients receiving a cemented THA compared with patients receiving an uncemented THA. No statistically significant difference in mortality was found.

Table 2. Patient mortality, raw data. Values are frequency (percentage)

Table 3. Postoperative mortality risk (OR (95% CI)) for cemented and hybrid THA compared with uncemented THA (reference)

Mortality 1–2 days 3–10 days 11–20 days 21–30 days

Cemented Uncemented Hybrid THA THA THA n = 23,636 n = 38,477 n = 11,802 9 (0.0) 45 (0.2) 35 (0.1) 22 (0.1)

4 (0.0) 23 (0.1) 14 (0.0) 4 (0.0)

0 (0.0) 6 (0.1) 6 (0.1) 3 (0.0)

30 days 111 (0.5) 45 (0.1) 15 (0.1) 90 days 228 (1.0) 93 (0.2) 29 (0.2) 180 days 389 (1.6) 155 (0.4) 52 (0.4) 365 days 712 (3.0) 254 (0.7) 115 (1.0)

group was the same as that in the uncemented THA group (OR = 1.2; CI 0.2–6.5) (Table 3). There were 45 deaths during the days 3 to 10 in the cemented THA group, 23 in the uncemented THA group, and 6 in the hybrid group (Table 2). The 3- to 10-day adjusted mortality in the cemented THA group was similar to that in the uncemented THA group (OR = 0.5; CI 0.3–1.1), and in the hybrid THA group (OR = 0.6, CI 0.3–1.6) (Table 3). Data on deaths and mortality for the follow-up periods 11 to 20 days, 21 to 30 days, 30 days, 90 days, and 365 days are presented in Tables 2 and 3. There were no deaths due to pulmonary embolism during days 1 and 2 in any of the groups (Table 4). There were 5 deaths during days 1 and 2 in the cemented THA group due to cardiovascular diseases, 4 in the uncemented THA group, and 0 in the hybrid group (Table 4). Data on cause of death for the follow-up periods 11 to 20 days, 21 to 30 days, 30 days, 90 days, and 365 days are presented in Table 4.

2 days 3–10 days 11–20 days 21–30 days

Cemented THA 1.2 (0.2–6.5) 0.5 (0.3–1.1) 0.7 (0.3–1.8) 2.8 (0.8–10.0)

Hybrid THA 0 (0.0–999.9) 0.6 (0.3–1.6) 0.9 (0.3–2.5) 1.9 (0.4–8.8)

30 days 0.8 (0.5–1.3) 0.8 (0.4–1.4) 90 days 0.8 (0.6–1.1) 0.7 (0.5–1.1) 365 days 1.2 (1.0–1.4) 1.2 (0.9–1.5) Table 4. Causes of death. Values are frequency (percentage) Cemented Uncemented Hybrid THA THA THA Cause of death n = 23,636 n = 38,477 n = 11,802 1–2 days Pulmonary embolism 0 0 0 Other cardiovascular 5 (0.02) 4 (0.01) 0 3–10 days Pulmonary embolism 1 (0.00) 2 (0.01) 0 All cardiovascular 11 (0.05) 19 (0.05) 4 (0.03) 11–20 days Pulmonary embolism 5 (0.02) 1 (0.00) 1 (0.01) All cardiovascular 15 (0.06) 9 (0.02) 5 (0.04) 21–30 days Pulmonary embolism 3 (0.01) 1 (0.00) 0 (0.00) All cardiovascular 8 (0.03) 3 (0.01) 2 (0.02) 90 days Pulmonary embolism 15 (0.06) 14 (0.04) 4 (0.03) All cardiovascular 76 (0.32) 60 (0.16) 22 (0.19) 365 days Pulmonary embolism 30 (0.13) 20 (0.05) 7 (0.06) All cardiovascular 208 (0.88) 127 (0.33) 62 (0.53) All cardiovascular: acute myocardial infarction, ischemic heart disease, pulmonary embolism, heart failure, stroke. Follow-up of the causes of death is to the end of 2013.

Acta Orthopaedica 2019; 90 (1): 6–10

Discussion Based on Finnish Registry data the adjusted early postoperative mortality after cemented THA compared with uncemented or hybrid THA was similar as regards death for any reason, death from pulmonary embolism, or death for cardiovascular reasons. However, using unadjusted data the proportion of perioperative deaths was higher in patients with cemented THA than in patients with uncemented or hybrid THA. Cementing is the gold standard for implant fixation, especially in elderly patients. In combined Nordic data, risk for revision has been both statistically and clinically significantly lower with cemented implants than with uncemented implants in patients aged 65 years or more (Mäkelä et al. 2014a, Varnum et al. 2015). Bone cement has been thought to strengthen bone from inside and therefore to decrease the risk for periprosthetic fracture, osteolysis, and loosening. Lower revision rates for cemented implants in elderly patients have been found in all major registries (Swedish Hip Arthroplasty Register 2013, AOANJRR 2016, NJR 2016). Even though superiority in implant survival of cemented THA in elderly patients, fear of bone cement implantation syndrome (BCIS) has led many surgeons towards using uncemented implant fixation (Dale et al. 2009, Fevang et al. 2010, Mäkelä et al. 2014b). BCIS is characterized by perioperative hypotension and hypoxia, and at worst cardiac arrest and death of the patient. The true incidence of cardiac arrest secondary to BCIS is unknown (Donaldson et al. 2009). In our study the 1- and 2-day adjusted mortality was similar in the cemented and uncemented THA groups. Thus, BCIS is seldom a cause of death in elective THA patients in Finland. Historically, cemented THA has been associated with greater than 3-fold higher intraoperative mortality (Coventry et al. 1974, Ereth et al. 1992, Parvizi et al. 1999), but at the end of the 1990’s a reduction in the intraoperative mortality rate has been reported (Parvizi et al. 1999) and the mortality has decreased even more during the twenty-first century (Hunt et al. 2013). A Swedish register study reported an increased adjusted risk of death during the first 14 days after surgery in patients who underwent cemented THA when compared with matched controls (HR of 1.3, 95% CI 1.11–1.44). This means 5 additional deaths per 10,000 observations. Such an increased risk of death was not found in patients with a cementless or hybrid THA. However, this risk in the cemented THA group disappeared during follow-up of 90-days (Garland et al. 2017). In our study the adjusted OR for mortality in the cemented THA group was not elevated during the first 20 postoperative days when compared with the uncemented THA group. Also, McMinn et al. (2012) reported a higher mortality rate in patients undergoing cemented THA compared with uncemented THA. However, this increase in mortality occurred gradually during 8 years after surgery and not early as would be expected if the increased mortality was caused by BCIS. We found similar adjusted mortality

regarding the use of bone cement at any time point up to 365 days postoperatively. This is in line with a study by Parvizi et al. (2001) who found no increased risk of death with cemented THA 30 days postoperatively. In a recent systematic review the overall 30-day mortality was 0.30% and 90-day mortality was 0.65% following THA. The leading cause of death was ischemic heart disease (41% of deaths) followed by cerebrovascular accidents (23%), and pulmonary embolism (12%) (Berstock et al. 2014). In our material the unadjusted mortality at 30 and 90 days for the cemented THA group was 0.5% and 1.0%, and 0.1% and 0.2% for the hybrid and uncemented groups, respectively. These differences are mainly explained by patient selection, and after adjusting for the elderly and sicker population in the cemented group the mortality was similar between the cemented and uncemented groups. The leading cause of death was cardiovascular. Parvizi et al. (1999) studied intraoperative mortality during cemented THA and found an incidence of 0.03%, the leading cause of death being pulmonary embolism. In previous studies increasing age, male sex, worse ASA score (> 3), and higher number of comorbidities have been found to increase the risk of death after THA surgery (Bozic et al. 2012, Mahomed et al. 2003, Parvizi et al. 2001, Hunt et al. 2013). In our study we attempted to account for this by adjusting the treatment groups for sex, age, and comorbidities, whereafter early overall mortality between the groups was similar at any time point. Our study has several limitations. First, we have no information regarding perioperative resuscitations because of cardiac arrest due to BCIS. Second, data on revision surgeries of the study patients were not included. Thus, we do not know whether mortality is associated for example with multiple operations. Third, the number of deaths for cardiovascular accidents or pulmonary embolism in our study was fairly small. It is possible that in a larger population some smaller differences in the mortality could be detected. Nonetheless, our material consisted of over 60,000 THAs and therefore we believe that there is no difference in clinical importance. Further the PERFECT database does not include information on patients’ socioeconomic status, which is known to affect mortality after THA (Whitehouse et al. 2014, Garland et al. 2017). Therefore some amount of residual confounding cannot be ruled out. As we modeled the mortality difference between the groups using logistic regression analysis repeatedly (356 analyses) there is a possibility of overfitting as the statistical model may contain more parameters than can be justified by the data. This can mean that the results are based on an adaptation to random variation in the sample and therefore the conclusions of our sample could not be generalized to a greater population. In summary, adjusted perioperative and short-term mortality was similar between patients treated with cemented THA and patients treated with uncemented or hybrid THA. This pertained also when cardiovascular and pulmonary embolism

mortality was studied separately. Based on our results and earlier literature, cemented THA is a safe option and should be the gold standard in the elderly patient population.

KM designed and coordinated the study. EE collected the data and drafted the manuscript. AP helped to draft the manuscript. MP and UH calculated the statistics. All authors contributed to the interpretation of the data and results and to the preparation of the manuscript. Acta thanks Ross W Crawford for help with peer review of this study

AOANJRR. Annual Report 2016; 2016. Aynardi M, Pulido L, Parvizi J, Sharkey P F, Rothman R H. Early mortality after modern total hip arthroplasty. Clin Orthop Relat Res 2009; 467(1): 213–18. Berstock J R, Beswick A D, Lenguerrand E, Whitehouse M R, Blom A W. Mortality after total hip replacement surgery: a systematic review. Bone Joint Res 2014; 3(6): 175-82. Bozic K J, Lau E, Kurtz S, Ong K, Rubash H, Vail T P, Berry D J. Patientrelated risk factors for periprosthetic joint infection and postoperative mortality following total hip arthroplasty in Medicare patients. J Bone Joint Surg 2012; 94(9): 794-800. Coventry M B, Beckenbaugh R D, Nolan D R, Ilstrup D M. 2,012 total hip arthroplasties: a study of postoperative course and early complications. J Bone Joint Surg Am 1974; 56(2): 273-84. Dale H, Hallan G, Espehaug B, Havelin L I, Engesæter L B. Increasing risk of revision due to deep infection after hip arthroplasty. Acta Orthop 2009; 80(6): 639-45. Donaldson A J, Thomson H E, Harper N J, Kenny N W. Bone cement implantation syndrome. Br J Anaesth 2009; 102(1): 12-22. Ereth M H, Weber J G, Abel M D, Lennon R L, Lewallen D G, Ilsrtup D M, Rehder K. Cemented versus noncemented total hip arthroplasty: embolism, hemodynamics, and intrapulmonary shunting. Mayo Clin Proc 1992; 67(11): 1066-74. Fevang B-T S, Lie S A, Havelin L I, Engesaeter L B, Furnes O. Improved results of primary total hip replacement. Acta Orthop 2010; 81(6): 649-59. Garland A, Rolfson O, Garellick G, Karrholm J, Hailer N P. Early postoperative mortality after simultaneous or staged bilateral primary total hip arthroplasty: an observational register study from the Swedish Hip arthroplasty register. BMC Musculoskelet Disord 2015; 16: 77. Garland A, Gordon M, Garellick G, Kärrholm J, Sköldenberg 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. Harris W H. The first 50 years of total hip arthroplasty: lessons learned. Clin Orthop Relat Res 2009; 467(1): 28-31. Haukka J, Suvisaari J, Tuulio-Henriksson A, Lönnqvist J. High concordance between self-reported medication and official prescription database information. Eur J Clin Pharmacol 2007; 63(11): 1069-74. 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. Junnila M, Laaksonen I, Eskelinen A, Pulkkinen P, Ivar Havelin L, Furnes O, Marie Fenstad A, Pedersen A B, Overgaard S, Kärrholm J, Garellick G,

Acta Orthopaedica 2019; 90 (1): 6–10

Malchau H, Mäkelä K T. Implant survival of the most common cemented total hip devices from the Nordic Arthroplasty Register Association database. Acta Orthop 2016; 87(6): 546-53. Lahti R A, Penttilä A. Cause-of-death query in validation of death certification by expert panel: wffects on mortality statistics in Finland, 1995. Forensic Sci Int 2003; 131(2-3): 113-24. Lalmohamed A, Vestergaard P, de Boer A, Leufkens H G M, van Staa T P, de Vries F. Changes in mortality patterns following total hip or knee arthroplasty over the past two decades: a nationwide cohort study. Arthritis Rheum 2014; 66(2): 311-8. Mahomed N N, Barrett J A, Katz J N, Phillips C B, Losina E, Lew R A, Guadagnoli E, Harris W H, Poss R, Baron J A. Rates and outcomes of primary and revision total hip replacement in the United States Medicare population. J Bone Joint Surg Am 2003; 85-A(1): 27-32. McMinn D J W, Snell K I E, Daniel J, Treacy R B C, Pynsent P B, Riley R D. Mortality and implant revision rates of hip arthroplasty in patients with osteoarthritis: registry based cohort study. BMJ 2012; 344: e3319. Morshed S, Bozic K J, Ries M D, Malchau H, Colford J M. Comparison of cemented and uncemented fixation in total hip replacement: a meta-analysis. Acta Orthop 2007; 78(3): 315-26. Mäkelä K T, Matilainen M, Pulkkinen P, Fenstad A M, Havelin L, Engesaeter 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 2014a; 348(January): f7592. Mäkelä K T, Matilainen M, Pulkkinen P, Fenstad A M, Havelin L I, Engesaeter L, Furnes O, Overgaard S, Pedersen A B, Kärrholm 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 2014b; 85(2): 107-16. NJR. NJR 13th Annual Report; 2016. Nurmohamed M T, Rosendaal F R, Büller H R, Dekker E, Hommes D W, Vandenbroucke J P, Briët E. Low-molecular-weight heparin versus standard heparin in general and orthopaedic surgery: a meta-analysis. Lancet 1992; 340(8812): 152-6. Pajunen P, Koukkunen H, Ketonen M, Jerkkola T, Immonen-Räihä P, KärjäKoskenkari P, Mähönen M, Niemelä M, Kuulasmaa K, Palomäki P, Mustonen J, Lehtonen A, Arstila M, Vuorenmaa T, Lehto S, Miettinen H, Torppa J, Tuomilehto J, Kesäniemi Y A, Pyörälä K, Salomaa V. The validity of the Finnish Hospital Discharge Register and Causes of Death Register data on coronary heart disease. Eur J Prev Cardiol 2005; 12(2): 132-7. Parvizi J, Holiday A D, Ereth M H, Lewallen D G. The Frank Stinchfield Award: Sudden death during primary hip arthroplasty. Clin Orthop Relat Res 1999; (369): 39-48. Parvizi J, Johnson B G, Rowland C, Ereth M H, Lewallen D G. Thirty-day mortality after elective total hip arthroplasty. J Bone Joint Surg Am 2001; 83-A(10): 1524-8. Peltola M, Juntunen M, Häkkinen U, Rosenqvist G, Seppälä T T, Sund R. A methodological approach for register-based evaluation of cost and outcomes in health care. Ann Med 2011; 43(Suppl. 1): S4-13. Sund R. Quality of the Finnish Hospital Discharge Register: a systematic review. Scand J Public Health 2012; 40(6): 505-15. Swedish Hip Arthroplasty Register. Annual Report 2013; 2013. Varnum C, Pedersen A B, Mäkelä K, Eskelinen A, Havelin L I, Furnes O, Kärrholm J, Garellick G, Overgaard S. Increased risk of revision of cementless stemmed total hip arthroplasty with metal-on-metal bearings. Acta Orthop 2015; 86(4): 491-7. Whitehouse S L, Bolland B J R F, Howell J R, Crawford R W, Timperley A J. Mortality following hip arthroplasty: inappropriate use of National Joint Registry (NJR) data. J Arthroplasty 2014; 29(9): 1827-34.

Acta Orthopaedica 2019; 90 (1): 11–14

Dual mobility cups in primary total hip arthroplasties: trend over time in use, patient characteristics, and mid-term revision in 3,038 cases in the Dutch Arthroplasty Register (2007–2016) Esther M BLOEMHEUVEL 1, Liza N VAN STEENBERGEN 2, and Bart A SWIERSTRA 1 1 Department of Orthopedic Surgery, Sint Maartenskliniek, Nijmegen; 2 Dutch Arthroplasty Register (LROI), ’s Hertogenbosch, the Netherlands Correspondence: esther.bloemheuvel@gmail.com Submitted 2017-11-09. Accepted 2018-09-14.

Background and purpose — We noticed an increased use of dual mobility cups (DMC) in primary total hip arthroplasty (THA) despite limited knowledge of implant longevity. Therefore, we determined the trend over time and midterm cup revision rates of DMC compared with unipolar cups (UC) in primary THA. Patients and methods — All primary THA registered in the Dutch Arthroplasty Register (LROI) during 2007–2016 were included (n = 215,953) and divided into 2 groups — DMC THA (n = 3,038) and UC THA (n = 212,915). Crude competing risk and multivariable Cox regression analyses were performed with cup revision for any reason as primary endpoint. Adjustments were made for sex, age, diagnosis at primary THA, previous operation, ASA score, type of fixation, surgical approach, and femoral head size. Results — The proportion of primary DMC THA increased from 0.8% (n = 184) in 2010 to 2.6% (n = 740) in 2016. Patients who underwent DMC THA more often had a previous operation on the affected hip, a higher ASA score, and the diagnosis acute fracture or late posttraumatic status compared with the UC THA group. Overall 5-year cup revision rate was 1.5% (95% CI 1.0–2.3) for DMC and 1.4% (CI 1.3–1.4) for UC THA. Stratified analyses for patient characteristics showed no differences in cup revision rates between the 2 groups. Multivariable regression analyses showed no statistically significantly increased risk for revision for DMC THA (HR 0.9 [0.6–1.2]). Interpretation — The use of primary DMC THA increased with differences in patient characteristics. The 5-year cup revision rates for DMC THA and UC THA were comparable.

The most frequent reason for revision in the 1st year after total hip arthroplasty (THA) is dislocation (LROI annual report 2016). Dislocation of a hip prosthesis is multifactorial including femoral head diameter. Mechanical studies have shown that instability could be decreased by increasing the diameter of the femoral head. With a larger head diameter, the head– neck ratio is higher and therefore there is a lower potential for instability (Burroughs et al. 2005). Dual articulation implants were designed to increase implant stability but also to decrease polyethylene rim damage from contact between femoral neck and acetabular liner and to restore near-normal range of motion. The dual mobility cup (DMC) is a ‘cup in a cup’ and was developed in the 1970s to combine the low-friction arthroplasty principle of Charnley with the advantage of a big femoral head principle of McKee (Philippot et al. 2009). Despite concerns about increased polyethylene wear due to the large femoral head, the DMC is not only used for revisions but also in primary THA to reduce dislocations. This was shown by De Martino et al. (2017) who counted, in a review of English articles between 1974 and 2016, 12,844 primary DMC THA and 5,064 revision DMC THA. Many of these articles focused more on dislocation rates than on longevity of the implant. Also in our daily practice we noticed an increase in the use of DMC in primary THA. Therefore, we determined the trend over time and mid-term cup revision rates of DMC compared with unipolar cup (UC) in primary THA with data from the Dutch Arthroplasty Register.

Patients and methods The Dutch Arthroplasty Register (LROI) is a nationwide population-based register that includes information on arthroplasties in the Netherlands since 2007. The LROI was initiated by the Netherlands Orthopaedic Association and is well sup-

Acta Orthopaedica 2019; 90 (1): 11–14

Table 1. Types of dual mobility cup THA used in the period 2007–2016 the Netherlands (n = 3,038)

Number of DMC THA

Cumulative cup revision (%)

800

4.0 Unipolar cup THA Dual mobility cup THA

Type

Cemented Cementless

Biomet Avantage 1,904 Biomet Avantage Reload – Biomet Avantage Rev HA – Smith & Nephew Polarcup 79 Amplitude Saturne 164 Mathys SeleXys DS Cup 27 Groupe Lepine Cupule Quattro 17 Groupe Lepine Cupule HAP Press-F –

84 339 5 273 85 54 – 7

600

3.0

400

2.0

200

1.0

0 2010

2011

2012

2013

2014

2015

2016

Year of primary THA

Year after primary THA

ported by its members. This results in coverage of Figure 1. Trend in the use of dual mobil- Figure 2. Cumulative incidence of cup 100% of Dutch hospitals and a completeness of ity cup (DMC) in total hip arthroplasty revision according to type of cup (all diagnoses) in the period 2007–2016 in (THA) in the period 2010–2016 in the reporting of over 95% for primary THAs and 88% Netherlands (n = 3,038). the Netherlands (n = 215,953). THA: total hip arthroplasty. for hip revision arthroplasty (Van Steenbergen et al. 2015). The LROI database contains information on patient, pro- type of fixation, surgical approach, and diameter of the femocedure, and prosthesis characteristics registered by registrars ral head to discriminate independent risk factors for cup revifrom each hospital. For each component a product number sion arthroplasty. BMI, Charnley score, and smoking status is registered to identify the characteristics of the prosthesis. were not included as covariates, since these were only availThe vital status of all patients is obtained actively on a regular able in the LROI database since 2014. For all covariates added basis from Vektis, the national insurance database on health to the model, the proportional hazards assumption was met care in the Netherlands, which records all deaths of Dutch citi- after inspecting log-minus-log curves. zens. The LROI uses the opt-out system to require informed Reasons for revision were described according to type of consent of patients. hip arthroplasty and compared using a chi-square test to test For the present study we included all patients that under- differences between types of THA (SPSS 22.0; IBM Corp, went a primary THA in the period 2007–2016. Metal-on-metal Armonk, NY, USA). (MoM) THA (n = 6,626) and records with a missing product More than 1 reason could be chosen. P-values below 0.05 number (n = 7,017) were excluded. The remaining 215,953 were considered statistically significant. For the 95% confihips comprised 3,038 DMC THAs and 212,915 UC THAs. dence intervals (CI), we assumed that the number of observed Diagnosis was categorized as osteoarthritis (OA), acute frac- cases followed a Poisson distribution. ture, late posttraumatic, and other. Other diagnoses registered in the LROI are dysplasia, inflammatory arthritis, osteonecro- Ethics, funding, and potential conflicts of interests sis, post-Perthes, and tumor (unspecified). Cup revision was The dataset was processed in compliance with the regulations defined as a revision procedure where at least the cup was of the LROI governing research on registry data. No external exchanged or removed. Closed reduction after a dislocation funding was received. No competing interests were declared. or incision and drainage for infection were not included in the LROI. The median follow-up was 3 years (0–9). Statistics Survival time was calculated as the time from primary THA to 1st revision arthroplasty for any reason, death of the patient, or the end of the study follow-up (January 1, 2017). Cumulative crude incidence of revision was calculated using competing risk analysis, where death was considered to be a competing risk (Lacny et al. 2015, Wongworawat et al. 2015). In addition Kaplan–Meier survival analyses were performed. Multivariable Cox proportional hazard ratios were performed to compare adjusted revision rates between DMC and UC THA. Adjustments were made for sex, age at surgery, diagnosis at primary THA, previous operation, ASA score,

Results The use of DMC THA increased from 184 (0.8% of all THAs) in 2010 to 740 (2.6% of all THAs) in 2016 (Figure 1) with 8 different types of DMC used (Table 1). In the DMC THA group more patients had undergone previous surgery on the affected hip and had a higher ASA score. Furthermore the distribution of diagnoses at primary surgery was different compared with the UC THA group (Table 2). The 5-year crude cup revision rate for DMC THA was 1.5% (CI 1.0–2.3) and 1.4% (CI 1.3–1.4) for UC THA (Figure 2). Stratified analyses according to diagnosis at primary THA, previous surgery on the affected hip, and fixation of the cup

Acta Orthopaedica 2019; 90 (1): 11–14

Table 2. Patient characteristics in THA according to type of acetabular cup (n = 212,915). Values are frequency and (%) unless otherwise specified

DMC THA n = 3,038

UC THA n = 212,915

Male sex, n (%) Mean age (SD) Operations before (yes) ASA I II III–IV Fixation Cemented Hybrid (acetabulum cemented) Hybrid (femur cemented) Uncemented Diagnosis Osteoarthritis Fracture (acute) Late posttraumatic Other a Approach Anterior Anterolateral Direct lateral Posterolateral Trochanter osteotomy Other Diameter (mm) 22–28 32 36 ≥ 38

1,104 (36) 70 (13) 632 (21)

70,144 (33) 69 (11) 10,048 (5)

308 (10) 1,724 (57) 951 (31)

47,409 (22) 129,460 (61) 27,748 (13)

1,710 (56) 495 (16) 126 (4) 674 (22)

60,955 (29) 9,033 (4) 9,932 (5) 130,911 (62)

1,688 (56) 424 (14) 406 (13) 476 (16)

185,062 (87) 7,065 (3) 4,415 (2) 14,163 (7)

96 (3) 41 (1) 254 (8) 2,607 (86) 1 (0) 8 (0)

21,102 (10) 15,801 (7) 44,249 (21) 128,275 (60) 71 (0) 635 (0)

2,784 (92) – – –

66,703 (31) 93,619 (44) 4,002 (19) 1,452 (1)

Numbers do not add up to total due to missing data. DMC: dual mobility cup; UC: unipolar cup; THA: total hip arthroplasty. a Other: dysplasia, inflammatory arthritis, osteonecrosis, postPerthes, tumor (unspecified).

showed similar 5-year crude cumulative incidence of cup revisions between the DMC and UC THA groups (Table 3 and 4, see Supplementary data). The unadjusted hazard ratio for cup revision of DMC THA compared with UC THA was 1.2 (CI 0.8–1.6). Moreover, multivariable survival analyses showed a comparable risk for cup revision for DMC THA (HR 0.8 [CI 0.6–1.2]). Dislocation was the most frequently registered reason for revision in UC THA patients (0.5%), while in the DMC THA group 0.2% were revised due to dislocation. In the DMC THA group loosening of the cup, dislocation, and infection were mostly registered as reason for revision (Table 5, see Supplementary data). From the 18 DMCs that loosened 8 were cemented.

Discussion We showed that the use of primary DMC THA increased in the Netherlands, with differences in patient characteristics

between DMC and UC THA patients. The 5-year revision rates were comparable, with no differences in specific subgroups. Our study is the first register study focusing on cup survival in primary use of DMC. Our 5-year cumulative incidences of cup revision of 1.5% in DMC THA and 1.4% in UC THA are lower than the overall revision rates from the Australian Orthopaedic Association National Joint Replacement Registry (2016), which reported 4.6% revision in 2,640 primary DMC THA and 3.3% in 327,847 primary UC THA. They did not specify the type of revision (insert, femoral head, cup, stem, or all). They also performed subgroup analysis and did not find a higher revision rate in any subgroup (AOANJRR 2016). Our results differ from the Swedish Hip Arthroplasty Register (2016) where a hazard ratio of 2.4 for revision of Avantage DMC THA compared with UC THA after correction for case mix factors and after exclusion of infections was found. However, their result is also based on the overall revision rates, while our HR of 0.8 is based on cup revisions only. Risk for cup revision due to dislocations was low with primary use of a DMC. In our study 8/3,038 (0.2%) DMC THA patients had a cup revision because of a dislocation versus 1,017/215,953 (0.5%) in UC THA. Tarasevicius et al. (2017) found in the Lithuanian Arthroplasty register at 5 years a revision rate for dislocation of 4/620 (0.7%) for primary DMC THA in comparison with 52/2,170 (2.4%) in a cemented Exeter cup. Revisions of UC THA are often preceded by 1 or more closed reductions (which are not reported in arthroplasty registers), while dislocations of DMC THA, being intra-prosthetic or not, are difficult to treat by closed reduction and will more often need surgery with exchange of components (which are reported in arthroplasty registers). So revision rates for dislocation in UC and DMC do not reflect instability in the same way. (Suspicion of) infection was the second commonest reason for cup revision in the DMC THA group (10/36). In the LROI only (suspected) prosthetic joint infections as reason for revision were registered. As shown earlier, implant registries largely underscore prosthetic joint infections (Gundtoft et al. 2015) since incisions and drainages without component exchange are not included. In this respect Mukka et al. (2013) published a study of 34 hips with DMC THA with soft-tissue debridement of 3 hips due to superficial infection. Chughtai et al. (2016) reported 453 primary DMC THA with 2 septic revisions after 2 years. Differences in patient characteristics and particularly comorbidities are probably the explanation for our high amount (0.3%) of revisions due to suspected infection (Radtke et al. 2016). Furthermore, differences in hospital guidelines (early debridement in the case of wound problems), diagnosis, and treatment of implant infections could be a reason for more reported infections (Osmon et al. 2013). Comparable risk for revision rates was seen between cemented and uncemented cups. Batailler et al. (2017)

reviewed 21 studies with different cementless DMCs in primary THA with 0–8% aseptic loosenings after 2–22 years. They argued that the fixation of cementless DMC can be affected by poor bone quality. This could have been the case in patients with (post)traumatic diagnosis or other comorbidities. Strengths of our study are, first, that the LROI contains a large population-based nationwide database of primary THAs, with a completeness of nearly 100% (van Steenbergen et al. 2015, LROI 2016) and an 8-year follow-up. Second, we focused our analyses on cup revisions, since type of revision (cup, stem, insert, and/or femoral head exchange) is specified in the LROI. A limitation of this study is that in registries only limited variables are collected, correctness of data cannot be proven, and causality cannot be proven due to its observational nature. This might lead to residual confounding. Furthermore, closed dislocations are missed, since this procedure is not registered in the LROI, when no prosthesis component is added, exchanged, or removed. Dislocations for a DMC THA are almost always registered in the LROI since closed dislocation for DMC THA is most often impossible. Conversely, closed dislocation for UC THA can often be performed without surgery. This could lead to a lower revision rate in the UC THA group. The limited reliability of a diagnosis of infection has been discussed above. In summary, the use of primary DMC THA in the Netherlands increased with differences in patient characteristics in comparison with UC THA. The 5-year revision rates for DMC THA were comparable to UC THA, even after adjustment for casemix factors. However, we need to be aware of residual confounding. To determine the exact role of DMC in primary THA compared with UC, randomized controlled trials or more subgroup analyses are needed. Supplementary data Tables 3–5 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674. 2018.1542210 All authors contributed to the conception of the study, data analysis, and preparation of the manuscript. Acta thanks Ove Furnes and Per Kjærsgaard-Andersen for help with peer review of this study.

Acta Orthopaedica 2019; 90 (1): 11–14

Australian Orthopaedic Association National Joint Replacement Registry. Annual report. Adelaide: AOA; 2016; https://aoanjrr.sahmri.com/ documents/10180/275066/Hip%2C%20Knee%20%26%20Shoulder%20 Arthroplasty Batailler C, Fary C, Verdier R, Aslanian T, Caton J, Lustig S. The evolution of outcomes and indications for the dual-mobility cup: a systematic review. Int Orthop 2017(41): 645-59. Burroughs B R, Hallstrom B, Golladay G F J, Hoeffel D, Harris W H. Range of motion and stability in total hip arthroplasty with 28-, 32-, 38- and 44- mm femoral head sizes. An in vitro study. J Arthroplasty 2005: 20(1): 11-19. Chughtai M, Mistry J B, Diedrich A M, Jauregui J J, Elmallah R K, Bonutti P M, Harwin S F, Malkani A L, Kolisek F R, Mont M A. Low frequency of early complications with dual-mobility acetabular cups in cementless primary THA. Clin Orthop Relat Res 2016; 474(10): 2181-7. 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. Lacny S, Wilson T, Clement F, Roberts D J, Faris P D, Ghali W A, Marshall D A. Kaplan–Meier survival analysis overestimates the risk of revision arthroplasty: a meta-analysis. Clin Orthop Relat Res 2015; 473(11): 3431-42. LROI, annual report 2016. www.lroi-rapportage.nl De Martino I, D´Apoliti R D, Soranoglou V G, Poultsides L A, Sculco P K, Sculco T P. Dislocation following total hip arthroplasty using dual mobility acetabular components: a systematic review. Bone Joint J 2017; 99-B(1 Suppl A): 18-24. Mukka S S, Mahmood S S, Sjödén G O, Sayed-Noor A S. Dual mobility cups for preventing early hip arthroplasty dislocation in patients at risk: experience in a county hospital. Orthop Rev (Pavia) 2013; 5(2): 48-51. Osmon D R, Berbari E F, Berendt A R, Lew D, Zimmerli W, Steckelberg J M, Rao N, Hanssen A, Wilson W R. Diagnosis and management of prosthetic joint infection: clinical practice guidelines by the Infectious Diseases Society of America. Clin Infect Dis 2013; 56: 1-10 . Philippot R, Camilleri J P, Boyer B, Adam P, Farizon F. The use of a dualarticulation acetabular cup system to prevent dislocation after primary total hip arthroplasty: analysis of 384 cases at a mean follow-up of 15 years. Int Orthop 2009; 33(4): 927-32. Radtke K, Tetzlaff T, Vaske B, Ettinger M. Claassen L, Florkemeier T, Windhagen H, Lewinski G von. Arthroplast-center related retrospective analysis of risk factors for periprosthetic joint infection after primary and after revision total hip arthroplasty. Technol Health Care 2016; 14; 24(5): 721-8. Swedish Orthopaedic Hip Arhtroplasty Register. Annual Report; 2016; https://registercentrum.blob.core.windows.net/shpr/r/Annual-Report2016-B1eWEH-mHM.pdf Van Steenbergen L N, Denissen G A W, Spooren A, van Rooden S M, van Oosterhout F J, Morrenhof J W, Nelissen R G H H. More than 95% completeness of reported procedures in the population-based Dutch Arthroplasty Register. Acta Orthop 2015; 86(4): 498-505. Tarasevicius S, Smailys A, Grigaitis K, Robertsson O, Stucinskas J. Shortterm outcome after total hip arthroplasty using dual-mobility cup: report from Lithuanian Arthroplasty Register. Int Orthop 2017; 41: 595-8. Wongworawat M D, Dobbs M B, Gebhardt M C, Gioe T J, Leopold S S, Manner P A, Rimnae C M, Porcher R. Editorial: Estimating survivorship in the face of competing risk. Clin Orthop Relat Res 2015; 473: 1173-6.

Acta Orthopaedica 2019; 90 (1): 15–20

Measurement properties of the OARSI core set of performancebased measures for hip osteoarthritis: a prospective cohort study on reliability, construct validity and responsiveness in 90 hip osteo arthritis patients Jaap J TOLK 1, Rob P A JANSSEN 1, C (Sanna) A C PRINSEN 2, M (Marieke) C VAN DER STEEN 3, Sita M A BIERMA ZEINSTRA 4,5, and Max REIJMAN 1,5 1 Department of Orthopedic Surgery and Trauma, Máxima Medical Center, Eindhoven; 2 Department of Epidemiology and Biostatistics, VU University Medical Center, Amsterdam Public Health (APH) Research Institute, Amsterdam; 3 Department of Orthopedic Surgery, Catharina Hospital Eindhoven, Eindhoven; 4 Department of General Practice, Erasmus MC, University Medical Center Rotterdam, Rotterdam; 5 Department of Orthopedic Surgery, Erasmus MC, University Medical Center Rotterdam, Rotterdam, The Netherlands Correspondence: jaap.tolk@mmc.nl Submitted 2018-04-12. Accepted 2018-10-11.

Background and purpose — Improvement of physical function is one of the main treatment goals in severe hip osteoarthritis (OA) patients. The Osteoarthritis Research Society International (OARSI) has identified a core set of performance-based tests to assess the construct physical function: 30-s chair stand test (30-s CST), 4x10-meter fastpaced walk test (40 m FPWT), and a stair-climb test. Despite this recommendation, available evidence on the measurement properties is limited. We evaluated the reliability, validity, and responsiveness of these performance-based measures in patients with hip OA scheduled for total hip arthroplasty (THA). Patients and methods — Baseline and 12-month follow-up measurements were prospectively obtained in 90 end-stage hip OA patients who underwent THA. As there is no gold standard for comparison, the hypothesis testing method was used for construct validity and responsiveness analysis. A test can be assumed valid if ≥ 75% of predefined hypotheses are confirmed. A subgroup (n = 30) underwent test–retest measurements for reliability analysis. The Oxford Hip Score, Hip injury and Osteoarthritis Outcome Score— Physical Function Short Form, pain during activity score, and muscle strength were used as comparator instruments. Results — Test–retest reliability was appropriate; intraclass correlation coefficient values exceeded 0.70 for all 3 tests. None of the performance-based measures reached 75% hypothesis confirmation for the construct validity or responsiveness analysis. Interpretation — The performance-based tests have good reliability in the assessment of physical function. Construct validity and responsiveness, using patient-reported measures and muscle strength as comparator instruments, could not be confirmed. Therefore, our findings do not justify their use for clinical practice.

Improvement of physical function is one of the main treatment goals of total hip arthroplasty (THA). Physical function can be assessed using patient-reported and performance-based outcome measurement instruments (Reiman and Manske 2011). Because different domains of the construct physical function are measured, the methods are considered complementary and not competing (Stratford and Kennedy 2006, Reiman and Manske 2011, Dobson et al. 2013). 3 activities have been identified as most relevant for patients with hip OA: sit-to-stand movement, level walking, and stair negotiation (Dobson et al. 2013). Impairment on these domains is classified as “activity limitations” on the World Health Organization International Classification of Functioning, Disability and Health (ICF) (World Health Organization 2001). The Osteoarthritis Research Society International (OARSI) has identified a set of performance-based tests to assess the construct physical function (Dobson et al. 2012, 2013). The core set consists of the 30-s chair stand test (30-s CST) for assessment of sit-to-stand movement, 4x10 meter fast-paced walk test (40 m FPWT) for assessment of level walking, and a stair-climb test to assess stair negotiation (Dobson et al. 2013). The validity and responsiveness of the OARSI core set have been challenged in knee OA patients (Tolk et al. 2017), but available evidence on the measurement properties in patients with hip OA is insufficient (Dobson et al. 2012, 2013). Measurement properties of a test should be confirmed in the population in which it is to be used, but the recommendation to use the specific tests included in the OARSI core set is based on expert opinion (Dobson et al. 2012, 2013). Therefore, before further implementation of the OARSI core set for hip OA patients can be considered, additional evidence on the measurement properties of these performance measures is essential (Terwee et al. 2006, Dobson et al. 2012). We evaluated the reliability, validity, and responsiveness after THA of the

OARSI recommended performance-based measures, for measurement of physical function in patients with severe hip OA.

Patients and methods We performed a prospective cohort study of patients indicated for THA to evaluate the measurement properties of the 30-s CST, 40 m FPWT, and 10-step stair climb test (10-step SCT). The study was conducted following the COSMIN (COnsensus based Standards for the selection of health status Measurement INstruments) checklist (Mokkink et al. 2010b). The COSMIN checklist contains design requirements and preferred statistical methods for studies on measurement properties of health status measurement instruments. Patient population Patients were eligible for inclusion if they had unilateral symptomatic hip OA and were scheduled for primary THA. Patients with comorbidity leading to inability to perform the performance-based measures, insufficient knowledge of the Dutch language, and inability to visit follow-up appointments were excluded. All patients in the Máxima Medical Centre meeting these criteria, and willing to participate, signed an informed consent form. The number of patients needed for the analysis was guided by the COSMIN standards (Terwee et al. 2007, Mokkink et al. 2010b). We aimed to include ≥ 50 patients for construct validity and responsiveness analyses, and 30 patients for reliability analyses. Study procedures Patient characteristics measured at baseline were: sex, age, and BMI. The assessment of performance-based measures and comparator instruments described below was made at baseline before surgery, and 12 months after THA. The standardized testing procedures were performed by a research nurse strictly according to the manual provided by the OARSI, with a fixed order of tests (Dobson et al. 2013). Performance-based measures 30-s CST. The 30-s CST aims to quantify a patient’s performance on the activity “sit-to-stand movement” (Dobson et al. 2013). From a sitting position, the patient stands up until hips and knees are fully extended, then completely back down. This is repeated for 30 seconds and each full cycle is counted as 1 chair stand (Dobson et al. 2013). A 43-cm high, straightback chair without armrests was used. For patients with hip OA, good reliability is reported with an intraclass correlation coefficient (ICC) of 0.81 (0.63–0.91) and standard error of measurement (SEM) of 1.27 (Wright et al. 2011). No reports on construct validity are available. 40 m FPWT. The 40 m FPWT is a test for performance on the activity short-distance walking (Dobson et al. 2013). Participants are asked to walk as quickly but as safely as pos-

Acta Orthopaedica 2019; 90 (1): 15–20

sible, without running, along a 10-meter walkway for a total distance of 40 meters. Walking speed is measured in meters/ second (m/s). Use of a walking aid is allowed and recorded. Inter-rater reliability is reported to be good in patients with hip OA, with an ICC of 0.95 (0.90–0.98) and SEM of 1.0 m/s (Wright et al. 2011). There are no reports available on the construct validity. Stair climb test. The OARSI included a stair-climb test in the core set, but no specific measure is recommended (Dobson et al. 2013). We selected the 10-step stair climb test (10step SCT), as the stair in the testing area had 10 steps with a step height of 19 cm. Patients were instructed to ascend and descend the flight of stairs as quickly as possible but in a safe manner. The time needed is recorded in seconds (Dobson et al. 2013). To our knowledge, there is no evidence available on measurement properties of the 10-step stair-climb test or comparable stair-climb tests in patients with hip OA. Comparator instruments We used a combination of comparator instruments; a specification of these instruments and their measurement properties can be found in a supplementary file. For measurement of physical function 2 joint-specific PROMs were used: the Hip injury and Osteoarthritis Outcome Score—Physical Function Short Form (HOOS—PS) (Davis et al. 2009), and the Oxford Hip Score (OHS) (Dawson et al. 1996). The EuroQol 5D-3L (EQ-5D) was used as a measure of health-related quality of life (Rabin and de Charro 2001). Pain during activity was scored from 0 to 10 using a numerical rating scale (NRS pain) (RuyssenWitrand et al. 2011). At 12 months follow-up a 7-point Likert scale anchor question was scored for change in activities of daily living. Preoperatively knee extensor and hip abductor strength of the affected leg was measured using a handheld dynamometer (Holstege et al. 2011, Zeni et al. 2014). Evaluation of the measurement properties and statistics Reliability Test–retest reliability refers to the extent to which scores for patients who have not changed are the same for repeated measurement over time. For this analysis, test–retest measurements of the 3 performance-based measures were obtained in a subset of the study population. 30 minutes of rest were allowed in between, to allow for full recovery during the resting interval. Performance on the activity under study can assumed to be stable over this testing period. ICC values for absolute agreement with corresponding 95% confidence intervals (CI) were calculated using a 2-way random model with absolute agreement. The threshold for an appropriate ICC is 0.70 (Terwee et al. 2007, Prinsen et al. 2016). SEM and SDC were calculated as described by Atkinson (1998). Construct validity Construct validity refers to the degree to which the instruments under study measure the construct they aim to mea-

Acta Orthopaedica 2019; 90 (1): 15–20

Table 1. Patient characteristics. Data are mean (SD) unless otherwise stated

Consecutive OA patients scheduled for THA n = 132 Excluded (n = 42): – not willing to participate, 37 – insufficient command of Dutch language, 3 – debilitating comorbidity, 2 Included in the study n = 90

Construct validity n = 90

Reliability n = 30

Not able to perform test 40m FPWT (n = 1) 10-step SCT (n = 6)

Patients included in the analyses and lost to follow-up.

Responsiveness n = 77 Lost to follow-up (n = 13): – died, 1 – not able to visit center due to logistic reasons, 5 – not willing to participate at follow-up, 3 – unknown reason, 4 Not able to perform test 40m FPWT (n = 4) 10-step SCT (n = 4)

sure. This is the recommended method to assess validity when there is no “Gold Standard” available, as is the case for the functional domains level walking, stair negotiation, and sit-tostand movement in hip OA. Before the start of the study, an expert panel formulated hypotheses on the expected relationships of performance-based measure scores with scores on the comparative instruments (Table 3, see Supplementary data) (Mokkink et al. 2010a, de Vet et al. 2011). Direction and magnitude of the expected results were stated. The expert panel consisted of an orthopedic surgeon (RJ), orthopedic resident and PhD candidate (JT), specialist in measurement property analysis (CP), and a methodologist (MR). The hypotheses were based on the following predictions— we expected: a moderate correlation of the performance-based measures with PROMs and quadriceps strength; a stronger correlation of PROMs with pain scores than with the performance-based measures; a stronger correlation of the performance-based measures with PROMs measuring functional outcome than with a PROM measuring general health; a stronger correlation of specific questions of the PROMs regarding walking, stair negotiation, and sit-to-stand movement to their respective performance-based measure than to the total PROM score. Correlations on a convergent hypothesis were expected to be at least moderate: ≥ 0.4 or ≤ –0.4. Divergent hypotheses were expected to have a poor correlation (≥ –0.39; ≤ 0.39). Pearson or Spearman correlation coefficients were calculated, depending on normality of data distribution. Construct validity can be assumed adequate if at least 75% of the predefined hypotheses are confirmed (Terwee et al. 2007). Responsiveness Responsiveness refers to the ability of the instruments to detect change over time in the construct measured. In the absence of a gold standard, a construct approach is to be used. Hypotheses were formulated a priori by the expert panel, in a similar

Age, years Women, n BMI Hip abductor strength, N Knee extensor strength, N

Total cohort (n = 90) 69 (9.5) 61 27 (3.9) 196 (7.8) 134 (5.7)

Reliability analysis cohort (n = 30) 66 (9.4) 22 26 (2.7) 219 (7.9) 13 (4.3)

manner to the construct validity analysis (Table 5) (Terwee et al. 2007, Mokkink et al. 2010a, de Vet et al. 2011). The hypotheses were formulated according to the following criteria: the anchor question would be moderately correlated to change in the performance-based measures scores (≥ 0.4 or ≤ –0.4) and the change in PROMs would be more correlated to pain than to change in the performance-based measure scores. Pearson or Spearman correlation coefficients were calculated, depending on normality of data distribution. Adequate responsiveness can be assumed if minimally 75% of the predefined hypotheses are confirmed (Terwee et al. 2007). SPSS statistics version 24.0 was used for the analyses (IBM Corp, Armonk, NY, USA). Ethics, funding, and potential conflicts of interest The Máxima Medical Centre Medical Ethics Committee approved the study (registration code 2014-73). No funding was received for the present study. The authors declare that there are no conflicts of interest related to this article.

Results Patient characteristics In the period April to October 2015, 90 consecutive patients scheduled for arthroplasty because of hip OA were recruited (Table 1, Figure). Measurement properties Reliability analysis 30 randomly selected patients were enrolled in the test–retest study. Test–retest reliability was appropriate; ICC values exceeded 0.70 for all 3 tests (Table 2, see Supplementary data). Construct validity (hypothesis testing) None of the 3 performance-based measures reached confirmation of 75% or more of the predefined hypotheses. 4/9 were confirmed for the 30-s CST, 6/17 for the 40m FPWT, and 6/17 for the 10-step SCT (Table 3, see Supplementary data). Responsiveness The mean score on the anchor question for change in activities of daily living (7-point Likert scale) at 12-month follow-

Acta Orthopaedica 2019; 90 (1): 15–20

Table 5. Responsiveness Predefined hypotheses

30-s chair stand test 40 m fast-paced walk test 10-step stair climb test (change score) (change score) (change score) Spearman Spearman Spearman correlation Hypothesis correlation Hypothesis correlation Hypothesis coefficient confirmed coefficient confirmed coefficient a confirmed

1. Moderate correlation with anchor question (≥ 0.4) 2. Moderate correlation with change score NRS pain during activity (≤ –0.4) 3. Moderate correlation with change score HOOS-PS (≤ –0.4) 4. Moderate correlation with change OHS (≥ 0.4) 5. Correlation between change scores NRS pain and HOOS-PS is minimal 0.1 stronger than between NRS pain and performance-based test 6. Correlation between change scores NRS pain and HOOS-PS is minimal 0.1 stronger than between HOOS-PS and performance-based test 7. Correlation between changes scores NRS pain and OHS minimal 0.1 stronger than between NRS pain and performance-based test 8. Correlation between change scores NRS pain and OHS is minimal 0.1 stronger than between OHS and performance-based test Hypothesis confirmed

up was 6.2 (5.9–6.4), which represents “much improvement.” Results of the responsiveness analysis are presented in Table 5. For the 30-s CST, 4/8 of the hypothesis were confirmed, for the 40m FPWT 4/8, and for the 10-step SCT 4/8 (Table 4, see Supplementary data).

Discussion To our knowledge, this is the first thorough assessment of the measurement properties of the OARSI-recommended core set of performance-based measures in patients with severe hip OA. The reliability analysis showed excellent test–retest reliability, which is in line with previous reports (Wright et al. 2011, Dobson et al. 2017). Construct validity and responsiveness could not be confirmed. These findings are in accordance with recently published work on the OARSI core set of performance-based measures in knee OA patients (Tolk et al. 2017). All 3 performance-based measures scored poorly on the construct validity and responsiveness analysis. One of the reasons is that almost all convergent hypotheses with PROMs measuring physical function were rejected. Although both methods aim to quantify related constructs, previous research has shown that PROMs assessing physical function do not measure the exact same domain as performance-based measures (Stratford and Kennedy 2006, Reiman and Manske 2011, Dobson et al. 2013). This potentially limits the strength of the conclusions that can be drawn from the present study. For example, PROMs are known to have a higher dependency on pain scores than performance based-measures (Stratford and Kennedy 2006). When—in the absence of a gold standard—the construct

0.37

0.28

–0.18

–0.04 0.30 0.23

No No No

–0.13 0.21 0.27

No No No

0.14 –0.35 –0.26

No No No

–0.45/–0.04 Yes

–0.45/–0.13 Yes

–0.45/–0.18

Yes

–0.45/0.30 Yes

–0.45/0.21

Yes

–0.45/–0.35

Yes

–0.66/–0.04 Yes

–0.66/–0.13 Yes

–0.66/–0.18

Yes

–0.66/0.23 Yes 4/8

–0.66/0.27 Yes 4/8

–0.66/–0.26 4/8

Yes

approach is to be used, it is inherently so that there is a discrepancy between the test under study and the comparator instruments (de Vet et al. 2011). Furthermore, PROMs were not the only comparative instruments used, and hypotheses predicting a higher correlation of the performance-based measure scores with related construct compared with less related constructs were largely rejected as well. Therefore, in our opinion, the conclusion on the construct validity and responsiveness should be interpreted more broadly than only showing the known discrepancy between PROMs and these measures. As an alternative to the comparator instruments used for construct validity and responsiveness in the present study, 3-D motion analysis or inertia-based motion analysis could be used. These methods allow for a kinematic analysis in patients with hip OA, but their clinical relevance has not been defined (Kolk et al. 2014, Bolink et al. 2016). Therefore, we believe these alternative methods are not suitable for comparison purposes in a clinical perspective. The comparative instruments used in the present study were considered the most suitable instruments available. The findings on construct validity of the performance-based measures might be affected because impairment on the tested activities in daily living is not fully appreciated by merely timing the performance (Steultjens et al. 1999, Stratford and Kennedy 2006). Although others claim good face validity for the core set of performance-based measures (Dobson et al. 2013, 2017), in our view this is not straightforward. For example, standing up and sitting down in rapid sequence, as measured by the 30-s CST, is not really exemplary for standto-sit movement in daily life. Fewer repetitions on the test does not necessarily mean the quality of a sit-to-stand move-

Acta Orthopaedica 2019; 90 (1): 15–20

ment in daily living is more or less impaired. The same goes for walking speed and stair ascent, which does not directly represent more or less impairment. Merely timing the activity or counting repetitions cannot capture impairment caused by limping or joint instability, nor avoidance of an activity in daily living (Steultjens et al. 1999, Holla et al. 2014). This is a possible explanation as to why the construct validity could not be confirmed. The responsiveness analysis showed that change in pain scores was strongly correlated to change in PROM scores, but not related to performance-based measure scores. Others have presented this low correlation with pain scores as a strength of performance-based measures, claiming this makes them more “objective” (Dobson et al. 2012, 2013). In our opinion, it seems unlikely that the degree of pain during an activity would not influence performance in daily living (Holla et al. 2014). Furthermore, it has been shown that pain during activity does affect the quality of movement, and impaired quality of movement is associated with lower perceived physical function (Steultjens et al. 1999, Rosenlund et al. 2016). Although pain reduction is not related to an increase in speed on the tested activities, the quality and manner of performance might improve (Steultjens et al. 1999), and patients might no longer avoid the activities (Holla et al. 2014). These factors of physical performance are not grasped by the performancebased measures under study. The number of repetitions or speed scored on the performance-based measures might be of interest for research purposes, but tin the authors’ opinion actual change and perceived change need to be related to some degree for a test to be clinically relevant. Hypotheses in this regard were all rejected, contributing to the negative conclusion on the responsiveness of the OARSI core set of performance-based measures. The strict adherence to the methodological criteria provided by COSMIN is a strength of the present study (Mokkink et al. 2010b). Most previous reports on the measurement properties of the performance-based measures under study reported combined groups of hip and knee OA patients, resulting in heterogeneous populations (Kennedy et al. 2005, Gill and McBurney 2008, Dobson et al. 2017). The present study reports on an unselected, consecutive group of only end-stage hip OA patients. The results can therefore be considered more accurate and representative for this population. The group size for test–retest measurements was kept relatively small, to reduce the burden of repeated measurements for patients. As there is evidence from other studies showing similar results on reliability (Kennedy et al. 2005, Wright et al. 2011, Dobson et al. 2017), in our view it can be concluded that the performance-based measures under study have adequate test–retest reliability. The percentage of patients lost to follow-up for the responsiveness analysis was 14%. In our opinion, this can be considered acceptable, especially as the group of patients with incomplete data did not show systematic difference in baseline characteristics (Table 1).

In summary, the 30-s CST, 40 m FPWT, and 10-step SCT have good reliability in the assessment of the domains sit-tostand movement, walking short distances, and stair negotiation in the construct physical function. Construct validity and responsiveness, using patient-reported measures and muscle strength as comparator instruments, could not be confirmed. Therefore, the present study does not justify their use for clinical practice in patients with severe hip OA. Supplementary data Tables 2–4 and a specification of comparator instruments used are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2018.1539567

JT and MR contributed to the conception and design of the study and drafting of the article. CP provided methodological support. All authors contributed to interpretation of the data and critically revised the article. The authors would like to sincerely thank C. van Doesburg, H. Kox, D. Latijnhouwers, and M. Mariam for their work in administrative and testing procedures. Acta thanks Margareta Hedstrom and Anders Holsgaard-Larsenfor help with peer review of this study.

Atkinson G N A. Statistical methods for assessing measurement error (reliability) in variables relevant to sports medicine. Sport Med 1998; 26(4): 21-38. Bolink S A A N, Lenguerrand E, Brunton L R, Wylde V, Gooberman-Hill R, Heyligers I C, Blom A W, Grimm B. Assessment of physical function following total hip arthroplasty: inertial sensor based gait analysis is supplementary to patient-reported outcome measures. Clin Biomech 2016; 32: 171-9. Davis A M, Perruccio A V, Canizares M, Hawker G A, Roos E M, Maillefert J-F, Lohmander L S. Comparative, validity and responsiveness of the HOOS-PS and KOOS-PS to the WOMAC physical function subscale in total joint replacement for osteoarthritis. Osteoarthritis Cartilage 2009; 17(7): 843-7. Dawson J, Fitzpatrick R, Carr A, Murray D. Questionnaire on the perceptions of patients about total hip replacement. Bone Joint Surg (Br) 1996; 78(2): 185-90. de Vet H C W, Terwee C B, Mokkink L B, Knol D L. Measurement in medicine: a practical guide. Cambridge: Cambridge University Press; 2011. Dobson F, Hinman R S, Hall M, Terwee C B, Roos E M, Bennell K L. Measurement properties of performance-based measures to assess physical function in hip and knee osteoarthritis: a systematic review. Osteoarthritis Cartilage 2012; 20(12): 1548-62. Dobson F, Hinman R S, Roos E M, Abbott J H, Stratford P, Davis A M, Buchbinder R, Snyder-Mackler L, Henrotin Y, Thumboo J, Hansen P, Bennell K L. OARSI recommended performance-based tests to assess physical function in people diagnosed with hip or knee osteoarthritis. Osteoarthritis Cartilage 2013; 21(8): 1042-52. Dobson F, Hinman R S, Hall M, Marshall C J, Sayer T, Anderson C, Newcomb N, Stratford P W, Bennell K L. Reliability and measurement error of the Osteoarthritis Research Society International (OARSI) recommended performance-based tests of physical function in people with hip and knee osteoarthritis. Osteoarthritis Cartilage 2017; 6-10.

Gill S, McBurney H. Reliability of performance-based measures in people awaiting joint replacement surgery of the hip or knee. Physiother Res Int 2008; 13(3): 141-52. Holla J F M, Sanchez-Ramirez D C, van der Leeden M, Ket J C F, Roorda L D, Lems W F, Steultjens M P M, Dekker J. The avoidance model in knee and hip osteoarthritis: a systematic review of the evidence. J Behav Med 2014; 37(6): 1226-41. Holstege M S, Lindeboom R, Lucas C. Preoperative quadriceps strength as a predictor for short-term functional outcome after total hip replacement. Arch Phys Med Rehabil 2011; 92(2): 236-41. Kennedy D M, Stratford P W, Wessel J, Gollish J D, Penney D. Assessing stability and change of four performance measures: a longitudinal study evaluating outcome following total hip and knee arthroplasty. BMC Musculoskelet Disord 2005; 6: 3. Kolk S, Minten M J M, Van Bon G E A, Rijnen W H, Geurts A C H, Verdonschot N, Weerdesteyn V. Gait and gait-related activities of daily living after total hip arthroplasty: a systematic review. Clin Biomech 2014; 29(6): 705-18. Mokkink L B, Terwee C B, Knol D L, Stratford P W, Alonso J, Patrick D L, Bouter L M, de Vet H C. The COSMIN checklist for evaluating the methodological quality of studies on measurement properties: a clarification of its content. BMC Med Res Methodol 2010a; 10: 22. Mokkink L B, Terwee C B, Patrick D L, Alonso J, Stratford P W, Knol D L, Bouter L M, De Vet H C W. The COSMIN checklist for assessing the methodological quality of studies on measurement properties of health status measurement instruments: an international Delphi study. Qual Life Res 2010b; 19(4): 539-49. Prinsen C A C, Vohra S, Rose M R, Boers M, Tugwell P, Clarke M, Williamson P R, Terwee C B, Chalmers I, Glasziou P, Williamson P, Altman D, Blazeby J, Clarke M, Devane D, Gargon E, Mokkink L, Terwee C, Patrick D, Alonso J, Stratford P, Knol D, Clarke M, Schmitt J, Apfelbacher C, Spuls P, Thomas K, Simpson E, Furue M, Prinsen C, Vohra S, Rose M, King-Jones S, Ishaque S, Bhaloo Z, Boers M, Kirwan J, Wells G, Beaton D, Gossec L, D’Agostino M, Mokkink L, Terwee C, Patrick D, Alonso J, Stratford P, Knol D, Murphy M, Black N, Lamping D, McKee C, Sanderson C, Askham J, Chiarotto A, Deyo R, Terwee C, Boers M, Buchbinder R, Corbin T, Verhagen A, Vet H, Bie R, Kessels A, Boers M, Bouter L, Jones J, Hunter D, Terwee C, Bot S, Boer M, Windt D, Knol D, Dekker J, Gargon E, Gurung B, Medley N, Altman D, Blazeby J, Clarke M. How to select outcome measurement instruments for outcomes included in a “Core Outcome Set”: a practical guideline. Trials 2016; 17(1): 449. Rabin R, de Charro F. EQ-5D: a measure of health status from the EuroQol Group. Ann Med 2001; 33(5): 337-43.

Acta Orthopaedica 2019; 90 (1): 15–20

Reiman M P, Manske R C. The assessment of function: how is it measured? A clinical perspective. J Man Manip Ther 2011; 19(2): 91-9. Rosenlund S, Holsgaard-Larsen A, Overgaard S, Jensen C. The Gait Deviation Index is associated with hip muscle strength and patient-reported outcome in patients with severe hip osteoarthritis: a cross-sectional study. PLoS One 2016; 11(4): 1-13. Ruyssen-Witrand A, Fernandez-Lopez C J, Gossec L, Anract P, Courpied J P, Dougados M. Psychometric properties of the OARSI/OMERACT osteoarthritis pain and functional impairment scales: ICOAP, KOOS-PS and HOOS-PS. Clin Exp Rheumatol 2011; 29(2): 231-7. Steultjens M P, Roorda L D, Dekker J, Bijlsma J W. Responsiveness of observational and self-report methods for assessing disability in mobility in patients with osteoarthritis. Arthritis Rheum 2001; 45(15): 56-61. Steultjens M P M, Dekker J, van Baar M E, Oostendorp R B, Bijlsma J W J. Internal consistency and validity of an observational method for assessing disability in mobility in patients with osteoarthritis. Arthritis Rheum 1999; 12(1): 19-25. Stratford P W, Kennedy D M. Performance measures were necessary to obtain a complete picture of osteoarthritic patients. J Clin Epidemiol 2006; 59(2): 160-7. Terwee C B, Mokkink L B, Steultjens M P M, Dekker J. Performance-based methods for measuring the physical function of patients with osteoarthritis of the hip or knee: a systematic review of measurement properties. Rheumatology (Oxford) 2006; 45(7): 890-902. Terwee C B, Bot S D M, de Boer M R, van der Windt D A, Knol D L, Dekker J, Bouter L M, de Vet H C W. Quality criteria were proposed for measurement properties of health status questionnaires. J Clin Epidemiol 2007; 60: 34-42. Tolk J J, Janssen R P A, Prinsen C A C, Latijnhouwers D A J M, van der Steen M C, Bierma-Zeinstra S M A, Reijman M. The OARSI core set of performance-based measures for knee osteoarthritis is reliable but not valid and responsive. Knee Surgery, Sport Traumatol Arthrosc 2017; epub ahead of print. World Health Organization. International Classification of Functioning, Disability and Health. Geneva: WHO; 2001. Wright A A, Cook C E, Baxter G D, Dockerty J D, Abbott J H. A comparison of 3 methodological approaches to defining major clinically important improvement of 4 performance measures in patients with hip osteoarthritis. J Orthop Sports Phys Ther 2011; 41(5): 319-27. Zeni J, Abujaber S, Pozzi F, Raisis L. Relationship between strength, pain, and different measures of functional ability in patients with end-stage hip osteoarthritis. Arthritis Care Res (Hoboken) 2014; 66(10): 1506-12.

Acta Orthopaedica 2019; 90 (1): 21–25

High rate of reoperation and conversion to total hip arthroplasty after internal fixation of young femoral neck fractures: a population-based study of 796 patients David J STOCKTON 1, Lyndsay M O’HARA 2, Nathan N O’HARA 3, Kelly A LEFAIVRE 1, Peter J O’BRIEN 1 and Gerard P SLOBOGEAN 3 1 Department

of Orthopaedics and Clinician Investigator Program, University of British Columbia, Vancouver, British Columbia, Canada; 2 Department of Epidemiology & Public Health, University of Maryland, Baltimore, MD, USA; 3 Department of Orthopaedics, University of Maryland, R Adams Cowley Shock Trauma Center, Baltimore, MD, USA Correspondence: djstockton@telus.net Submitted 2018-07-02. Accepted 2018-11-08.

Background and purpose — Most often, the goal of non-geriatric femoral neck fracture surgery is to preserve the native hip joint. However, reoperations for painful implants, osteonecrosis, and nonunion are common. We determined the reoperation rate and time-to-reoperation following internal fixation of these fractures in a large population cohort. Patients and methods — This retrospective cohort study included patients between the ages of 18 and 50 years old who underwent internal fixation for a femoral neck fracture during 1997–2013. Patients were followed until December 2013. Primary outcomes were reoperation rate and time-to-reoperation. Time-to-event analysis was performed to estimate the rate of any reoperation and for THA specifically, while testing the dependency of time-to-reoperation on secondary variables. Results — 796 young femoral neck fracture patients were treated with internal fixation during the study period (median age 43 years, 39% women). Median follow-up was 8 years (IQR 4–13). One-third underwent at least 1 reoperation at a median 16 months after the index surgery (IQR 8–31). Half of reoperations were for implant removal, followed by conversion to total hip arthroplasty. 14% of the cohort were converted to THA. The median time to conversion was 2 years (IQR 1–4). Neither female sex nor older age had a statistically significant effect on time-to-reoperation or timeto-THA conversion. Interpretation — Following internal fixation of young femoral neck fracture, 1 in 3 patients required a reoperation, and 1 in 7 were converted to THA. These data should be considered by patients and surgeons during treatment decisionmaking.

Femoral neck fractures in non-geriatric adults are a challenge to treat successfully. They often occur from high-energy trauma and result in displaced fracture patterns (Protzman and Burkhalter 1976, Robinson et al. 1995). Reduction and internal fixation is performed for nearly all younger patients with these fractures in order to preserve the native hip joint (Tooke and Favero 1985, Ly et al. 2008). The risk of a healing complication is often high, with the most common causes being osteonecrosis (14%), nonunion (9%), and severe femoral neck shortening (13–32%) (Slobogean et al. 2015, Stockton et al. 2015, Slobogean et al. 2017). A recent meta-analysis estimated a reoperation rate of 18% following internal fixation of young femoral neck fractures; however, this rate was estimated using predominantly retrospective, short-term, case series data (Slobogean et al. 2015). Lin et al. (2014) estimated the 10-year complication-free rate for femoral neck fractures in patients age 20–40 years to be 67% in Taiwan, but the outcome was an aggregate of reoperations and medical complications. While several case series have described the short-term surgical complications of these injuries, long-term population-based studies remain lacking; it is unknown how many patients treated with internal fixation will eventually require a reoperation or conversion to total hip arthroplasty (THA). Controversy exists surrounding the role of arthroplasty in young femoral neck fractures. In these patients, THA has historically been foregone as a primary intervention due to the finite lifespan of the implant. THA is typically considered a secondary procedure for the treatment of complications, such as osteonecrosis or nonunion (Angelini et al. 2009, Pauyo et al. 2014). However, the use of THA as a secondary procedure for internal fixation is associated with a higher complication rate (McKinley et al. 2002, Mahmoud et al. 2016). Advancements in arthroplasty systems have demonstrated improved

survivorship, increasing their viability as a primary treatment for younger patients (Adelani et al. 2013). A better understanding of current rates of complications and THA conversions for younger patients with femoral neck fractures is imperative to guide the treatment choice between internal fixation and primary THA, particularly for middle-aged patients. Therefore we determined the reoperation rate (in total and specifically for conversion to THA) and time-to-reoperation for patients aged 18–50 after internal fixation of femoral neck fractures. Secondarily, we sought to determine the relationship between age, sex, and hospital volume on the rates of reoperation and conversion to THA in this patient population. We hypothesized that male sex, older age, and having the index procedure done at a low-volume center would be associated with an increased risk of reoperation.

Patients and methods This was a retrospective cohort study. We used British Columbia (BC) administrative health-care data collected and linked by Population Data BC, a multi-hospital data and education resource facilitating interdisciplinary research on the determinants of human health, well-being, and development of the citizens of BC. The main data sources were Medical Services Plan (MSP) Payment Information Files that capture data on medically necessary services provided by physicians (i.e., licensed orthopedic surgeons) to individuals covered by MSP, the province’s universal insurance program (British Columbia Ministry of Health 2016a); the Discharge Abstracts Database (DAD), which contains demographic, administrative, and clinical data for all patients discharged from acute-care hospitals in BC (Canadian Institute for Health Information 2016); and Consolidation Files that contain basic demographic information including geo-coding that indicates location of residence (British Columbia Ministry of Health 2016b). These databases, which provide information from 1985 onwards, are held securely in linked, de-identified form at Population Data BC (www.popdata.bc.ca). We included all patients aged 18–50 from 1997 to 2013 who sustained femoral neck fractures (AO/OTA Type 31B) with subsequent osteosynthesis (MSP code 55751 for closed reduction internal fixation [CRIF] and 55755 for open reduction internal fixation [ORIF]) (Province of BC 2018). Patients who concomitantly experienced femoral shaft fractures (MSP codes 55782, 55783, and 55785) were also included; however, those that had pelvic or acetabular fracture (MSP code 55741, 55745, or 55746) were excluded due to the increased risk conferred to the viability of the native hip joint. Prior to delivering the final cohort for analysis, the Data Stewards at Population Data BC also excluded patients if they died or if they moved out of province after their index surgery (identified by the Province of Patient code from the DAD). All patients were followed until December 31, 2013.

Acta Orthopaedica 2019; 90 (1): 21–25

Age and sex variables were also obtained from the DAD. Age was analyzed as a continuous variable and as a dichotomized categorical variable, with an age of 45 years old chosen as the separating value. 45 years old is the lowest suggested age that has been recommended in the literature for consideration of treatment by primary arthroplasty (Swart et al. 2017). Index treatment was provided at 54 unique hospitals, with each hospital treating between 1 and 59 patients during the study period. The 3 hospitals that treated more than 40 index fractures during the study period were coded as “high-volume hospitals.” Study outcomes The primary outcomes were rate of reoperation and rate of conversion to THA. Rate was estimated using person-years as the denominator, such that patients with an index surgery early in the study period would contribute proportionally more to the denominator relative to those patients with more recent index surgeries (Szklo and Nieto 2014). Reoperation was defined as any of any of the following procedures undertaken after the index surgery: implant removal (MSP code 55415 or 55420), proximal femur osteotomy (55603), bone grafting (MSP code 55651), non-union fixation (MSP code 55633), hip hemiarthroplasty (55662), and THA (MSP code 55663). Secondary outcomes included time-to-any-reoperation and time-to-THA. Statistics The Pearson chi-square test was used to compare patient demographic and hospital variables by outcome status. A descriptive analysis of reoperation type was also conducted. A Kaplan–Meier analysis was performed to estimate the rate of failure of the index procedure for any reoperation and for THA specifically at 1-, 2-, 5-, and 10- year intervals as well as a mean time-to-reoperation. A Cox proportional-hazards regression model was used to study the dependency of timeto-reoperation and time-to-THA on patient and hospital variables. All statistical analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC, USA) and SPSS, Version 22 (IBM Corp, Armonk, NY, USA). Ethics, funding, and potential conflicts of interest The study was approved by the Clinical Research Ethics Board at the University of British Columbia (H14-03413). No external sources of funding were utilized. The fees for accessing the administrative databases held by Population Data BC were waived via a Student Waiver. None of the authors have any conflicts of interest. All inferences, opinions, and conclusions drawn in this study are those of the authors, and do not reflect the opinions or policies of the Data Stewards at Population Data BC.

Results The final cohort consisted of 796 patients aged 18–50 years who underwent internal fixation for femoral neck fracture

Acta Orthopaedica 2019; 90 (1): 21–25

Table 1. Demographic data (n = 796) Variable

Table 3. Cox proportional hazards model for risk (HR) and 95% confidence interval (CI) of reoperation and risk of THA

Reoperation No reoperation p-value

No. of index cases 235 561 NA Age, mean (SD) 40 (8) 41 (9) 0.7 Male sex, n (%) 138 (59) 346 (62) 0.4 Follow-up years, mean (SD) 9 (5) 8 (5) < 0.01 Hospital volume, n (%) a < 40 cases 187 (80) 451 (80) 0.9 ≥ 40 cases 48 (20) 110 (20) a Total

number of young femoral neck fracture fixation cases performed from 1997 to 2013.

Table 2. Types of reoperation (n = 351 a)

Variable Risk for reoperation: Age (45–50 years) Female sex Index surgery at high-volume center Risk for conversion to THA: Age (45–50 years) Female sex Index surgery at high-volume center

HR (CI)

p-value

0.8 (0.6–1.1) 1.2 (0.9–1.5) 0.9 (0.7–1.3)

0.1 0.3 0.6

1.2 (0.8–1.9) 1.2 (0.8–1.8) 0.7 (0.4–1.2)

0.3 0.5 0.2

Estimated survival probability 1.0

Type of reoperation Implant removal Total hip arthroplasty Nonunion fixation Revision CRIF and/ ORIF Hip hemiarthroplasty Bone grafting Osteotomy

n (%) 192 (55) 102 (29) 18 (5) 18 (5) 9 (3) 9 (3) 3 (1)

0.8

0.6

0.4

This total includes all reoperations, not just the first reoperation.

between 1997 and 2013. The number of included patients ranged between 40 and 60 cases per year. The median age of the cohort was 43 years (IQR 35–48) and 61% were male (Table 1). Median follow-up time was 8 years (IQR 4–13). There were 351 reoperations performed during the study period among 235 unique patients (30%) who required at least 1 reoperation. The most common reoperation was implant removal (n = 192, 55%), followed by conversion to THA (n = 102, 29%). 52 patients (7%) in the cohort required 2 or more reoperations within the study period. Among these 52 patients, the median number of reoperations was 2 (IQR 2–4) (Table 2). The median time to the first reoperation was 16 months (IQR 8–31). The Kaplan–Meier survival curve (Figure) shows that the 1-, 2-, 5-, and 10-year reoperation rates were 12%, 21%, 30%, and 34% respectively. Reoperation rates were similar whether stratified by age or hospital volume (Table 3). Our results remained statistically insignificant whether age was analyzed as a continuous or categorical variable. For conversion to THA specifically, the 1-, 2-, 5-, and 10-year reoperation rates were 3%, 6%, 10%, and 14% respectively. For those patients that required conversion to THA, the median time-toTHA was 27 months (IQR 12–50). Data on comorbidities, socioeconomic status, and time-tofixation were incomplete and were not included in the multivariable model. In the final model, neither age, sex, nor hospital volume was associated with a higher risk of reoperation, or with a higher risk of conversion to THA (Table 3).

0.2

0.0

Years after index operation

Kaplan–Meier survival curve for reoperation.

Discussion This study used population-based health data to estimate the long-term outcomes of young femoral neck fractures. This represents the largest North American cohort of such fractures with long-term outcomes to date. A 10-year reoperation rate of 34% and a 10-year rate of conversion to THA of 14% suggests a substantial opportunity to improve the treatment of femoral neck fractures in the non-geriatric population. Our results are comparable to previous literature, which predominantly consists of single-center case series. The largest series, by Huang et al. (2010), followed 146 patients for at least 2 years post-surgery (mean 5 years). Overall, 33 patients (23%) were converted to a prosthetic replacement, most due to osteonecrosis. Duckworth et al. (2011) followed a series of 122 patients with displaced young femoral neck fractures and found that 29% required a reoperation and 22% conversion to arthroplasty at a mean of 11 months (0.5–39). Haidukewych et al. (2004) found an 18% rate of conversion to arthroplasty at a mean of 7 years (3 months–15 years). Using populationbased, linked health data, our long-term reoperation rate was higher than in previous reports; however, our rate of conversion to THA was lower than many studies, even those with

shorter follow-up. While the reason for this discrepancy is unclear, it is important to acknowledge that most case series focus on displaced femoral neck fractures whereas our population-based cohort included patients with less severe fracture patterns. Scant population-based data exist for young femoral neck fractures. Samuel et al. (2016) performed a retrospective cohort study of 1,361 patients using the National Trauma Data Bank. They advocated for expedited care of these fractures after finding that delayed treatment (after 24 hours) resulted in higher odds of in-hospital complications. Longterm outcomes, however, were not examined. Lin et al. (2014) conducted a population-based study in Taiwan using their National Health Insurance Research Database. Of 2,905 femoral neck fractures in patients aged 20–40 years in their cohort, 67% were complication-free at 10 years. The 1-, 2-, 5-, and 10-year complication rates are similar to our reported findings at 13%, 20%, 25%, and 27%. However, direct comparisons with Lin et al. are limited given their use of a composite endpoint that included death, readmission within 90 days of index surgery, and reoperation along with their exclusion of study participants older than 40 years of age. Similarly, they did not observe an association between sex and the time to the first complication. Our findings expand on the previous literature by contributing an estimate of the median time-to-reoperation (16 months). To our knowledge, no other study has quantified the trajectory of reoperations for young femoral neck fractures. With a median time-to-reoperation of 16 months and the initially steep Kaplan–Meier survival curve seen in Figure 1, it is evident that most reoperations occur within the first 5 years of the index surgery. In our cohort, the most common type of reoperation was implant removal (55%) followed by THA (29%), nonunion fixation (5%), and revision internal fixation (5%). Previous studies have suggested that the most common reasons for reoperation following internal fixation of young femoral neck fracture are osteonecrosis and nonunion (Haidukewych et al. 2004, Pauyo et al. 2014); however, our results indicate that implant removal accounts for most reoperations. Implant removal does often precede more extensive operations such as THA. While the Medical Services Plan billing code for the major operation is reliably coded, the code for the more minor procedure accompanying it is not. While this represents an unfortunate reality of population-level databases, it does not account for all cases of implant removal. The fact remains that a large number of reoperations were for implant removal in isolation, which may be consistent with the issue of femoral neck shortening suspected to be associated with either an altered abductor moment arm or symptomatic implants (Slobogean et al. 2017). Implant removal notwithstanding, hip arthroplasty (THA and hemiarthroplasty) accounted for 32% of reoperations while hip-preserving procedures (nonunion fixation, revision

Acta Orthopaedica 2019; 90 (1): 21–25

fixation, bone grafting, and osteotomy) accounted for 14%. An arthroplasty procedure typically denotes an unsalvageable joint, most likely due to advanced osteonecrosis or nonunion. With regard to nonunion, some surgeons in British Columbia may treat difficult cases with arthroplasty; however, in many cases the first option for an aseptic nonunion is a hip-preserving procedure. Revision fixation using a valgus-producing trochanteric osteotomy is a recommended treatment for nonunion in high Pauwels angle fractures (Deakin et al. 2015). Our results suggest that this technique may actually be quite rare (0.8%); however, we suspect that surgeons tended to bill this procedure as a “nonunion fixation” instead of an “osteotomy.” Nonunion after fracture of the young femoral neck is unique in that the treatment may involve an osteotomy, a fact that the surgical billing codes do not account for. In this instance, it is likely that surgeons chose to bill using a code that represented the indication for the procedure. Recent evidence supporting primary total arthroplasty for difficult periarticular fractures with high complication rates has emerged for certain proximal humerus, distal femur, and elbow fractures (McKee et al. 2009, Chen et al. 2017, Sebastia-Forcada et al. 2017). There is increasing controversy over exactly what patient- and fracture-related characteristics should lead the treating surgeon to consider THA as the primary treatment for young femoral neck fractures. Converting a failed fixation to THA represents an additional burden to the individual and the health system, and does not fully restore function (Zielinski et al. 2014). In a Markov economic decision analysis, primary THA was found to be the most costeffective intervention for young femoral neck fracture for healthy patients > 54 years old, > 47 years old for those with mild comorbidities, and > 44 years old for those with multiple comorbidities (Swart et al. 2017). That analysis used an estimated rate for failed ORIF converted to THA for patients < 50 years old of 13%, drawn from a systematic literature review. Our observed 14% arthroplasty conversion rate supports their model parameters and may even lower the transition ages at which THA becomes the more cost-effective primary intervention. However, our analysis did not find that patients age 45–50 years old had a higher probability of needing a reoperation. Discerning the appropriate primary intervention for these fractures will likely take into account multiple factors that remain to be clearly defined. Our secondary analyses were unable to support our hypothesized associations between key variables and the risk of reoperation or conversion to THA. Neither age, sex, nor hospital volume were associated with these outcomes. The most significant limitation of this large administrative database study is our inability to measure the fracture severity or fixation quality. Additionally, the socioeconomic, provider volume, and time-to-initial fixation data were incomplete and therefore had to be excluded from the multivariable analyses. Excluding patients that died or moved out of province may have slightly biased our estimates if these patients were more

Acta Orthopaedica 2019; 90 (1): 21–25

prone to reoperation. Furthermore, while our analyses focused on the important outcome of reoperation, we are unable to comment on the functional outcome of the cohort, which could be associated with hospital or patient variables. Regardless, these limitations are mitigated by the unique strength of using a linked, population-based database from a single-payer health system, which ensures that reoperations that occurred several years after the index fixation or at another hospital were accurately captured. In summary, our study expands the understanding of young femoral neck fracture outcomes by estimating short-, medium-, and long-term reoperation rates. 1 in 3 patients will require a reoperation, while 1 in 7 will be converted to THA. These values are useful prognostic information for both surgeons and patients, and they provide further impetus for research to determine what fracture- and patient-related characteristics should be used to discern the optimal treatment for this difficult fracture.

DJS: study design, data collection, data analysis, writing of the draft paper, and revision of the paper. LMO, NNO, GPS: study design, data analysis, and revision of the paper. KAF, PJO: study design and revision of the paper. Acta thanks Hans Berg and Ville Mattila for help with peer review of this study.

Adelani M A, Keeney J A, Palisch A, Fowler S A, Clohisy J C. Has total hip arthroplasty in patients 30 years or younger improved? A systematic review. Clin Orthop Relat Res 2013; 471(8): 2595-601. Angelini M, McKee M D, Waddell J P, Haidukewych G, Schemitsch E H. Salvage of failed hip fracture fixation. J Orthop Trauma 2009; 23(6): 471-8. British Columbia Ministry of Health. Consolidation File (MSP Registration & Premium Billing). V2. Population Data BC. Data extract. MOH; 2016a. http: //www.popdata.bc.ca/data. British Columbia Ministry of Health. Medical Services Plan (MSP) Payment Information File. V2. Population Data BC. Data extract. MOH; 2016b. http: //www.popdata.bc.ca/data. Canadian Institute for Health Information. Discharge Abstract Database (Hospital Separations). V2. Population Data BC. Data extract. MOH; 2016. http: //www.popdata.bc.ca/data. Chen F, Li R, Lall A, Schwechter E M. Primary total knee arthroplasty for distal femur fractures: a systematic review of indications, implants, techniques, and results. Am J Orthop (Belle Mead NJ) 2017; 46(3): e163-71. Deakin D E, Guy P, O’Brien P J, Blachut P A, Lefaivre K A. Managing failed fixation: valgus osteotomy for femoral neck nonunion. Injury 2015; 46(3): 492-6. Duckworth A D, Bennet S J, Aderinto J, Keating J F. Fixation of intracapsular fractures of the femoral neck in young patients: risk factors for failure. J Bone Joint Surg Br 2011; 93(6): 811-16. Haidukewych G J, Rothwell W S, Jacofsky D J, Torchia M E, Berry D J. Operative treatment of femoral neck fractures in patients between the ages of fifteen and fifty years. J Bone Joint Surg Am 2004; 86-A(8): 1711-16. Huang H K, Su Y P, Chen C M, Chiu F Y, Liu C L. Displaced femoral neck fractures in young adults treated with closed reduction and internal fixation. Orthopedics 2010; 33(12): 873.

Lin J C, Wu C C, Lo C, Liang W M, Cheng C F, Wang C B, Chang Y J, Wu H C, Leu T H. Mortality and complications of hip fracture in young adults: a nationwide population-based cohort study. BMC Musculoskelet Disord 2014; 15: 362. Ly T V, Swiontkowski M F. Treatment of femoral neck fractures in young adults. J Bone Joint Surg Am 2008; 90(10): 2254-66. Mahmoud S S, Pearse E O, Smith T O, Hing C B. Outcomes of total hip arthroplasty, as a salvage procedure, following failed internal fixation of intracapsular fractures of the femoral neck: a systematic review and metaanalysis. Bone Joint J 2016; 98-b(4): 452-60. McKee M D, Veillette C J, Hall J A, Schemitsch E H, Wild L M, McCormack R, Perey B, Goetz T, Zomar M, Moon K, Mandel S, Petit S, Guy P, Leung I. A multicenter, prospective, randomized, controlled trial of open reduction–internal fixation versus total elbow arthroplasty for displaced intraarticular distal humeral fractures in elderly patients. J Shoulder Elbow Surg 2009; 18(1): 3-12. McKinley J C, Robinson C M. Treatment of displaced intracapsular hip fractures with total hip arthroplasty: comparison of primary arthroplasty with early salvage arthroplasty after failed internal fixation. J Bone Joint Surg Am 2002; 84-A(11): 2010-15. Pauyo T, Drager J, Albers A, Harvey E J. Management of femoral neck fractures in the young patient: a critical analysis review. World J Orthop 2014; 5(3): 204-17. Protzman R R, Burkhalter W E. Femoral neck fractures in young adults. J Bone Joint Surg Am 1976; 58(5): 689-95. Province of BC. MSC Payment Schedule: gov.bc.ca; 2018. Available from: https: //www2.gov.bc.ca/gov/content/health/practitioner-professionalresources/msp/physicians/payment-schedules/msc-payment-schedule. Robinson C M, Court-Brown C M, McQueen M M, Christie J. Hip fractures in adults younger than 50 years of age: epidemiology and results. Clin Orthop Relat Res 1995; (312): 238-46. Samuel A M, Russo G S, Lukasiewicz A M, Webb M L, Bohl D D, Basques B A, Grauer J N. Surgical treatment of femoral neck fractures after 24 hours in patients between the ages of 18 and 49 is associated with poor inpatient outcomes: an analysis of 1361 patients in the National Trauma Data Bank. J Orthop Trauma 2016; 30(2): 89-94. Sebastia-Forcada E, Lizaur-Utrilla A, Cebrian-Gomez R, Miralles-Munoz F A, Lopez-Prats F A. Outcomes of reverse total shoulder arthroplasty for proximal humeral fractures: primary arthroplasty versus secondary arthroplasty after failed proximal humeral locking plate fixation. J Orthop Trauma 2017; 31(8): e236-40. Slobogean G P, Sprague S A, Scott T, Bhandari M. Complications following young femoral neck fractures. Injury 2015; 46(3): 484-91. Slobogean G P, Stockton D J, Zeng B F, Wang D, Ma B, Pollak A N. Femoral neck shortening in adult patients under the age of 55 years is associated with worse functional outcomes: analysis of the prospective multi-center study of hip fracture outcomes in China (SHOC). Injury 2017; 48(8): 183742. Stockton D J, Lefaivre K A, Deakin D E, Osterhoff G, Yamada A, Broekhuyse H M, O’Brien P J, Slobogean G P. Incidence, magnitude, and predictors of shortening in young femoral neck fractures. J Orthop Trauma 2015; 29(9): e293-8. Swart E, Roulette P, Leas D, Bozic K J, Karunakar M. ORIF or arthroplasty for displaced femoral neck fractures in patients younger than 65 years old: an economic decision analysis. J Bone Joint Surg Am 2017; 99(1): 65-75. Szklo M, Nieto F J. Epidemiology: beyond the basics. 3rd ed. Burlington, MA: Jones & Bartlett Learning; 2014. Tooke S M, Favero K J. Femoral neck fractures in skeletally mature patients, fifty years old or less. J Bone Joint Surg Am 1985; 67(8): 1255-60. Zielinski S M, Keijsers N L, Praet S F, Heetveld M J, Bhandari M, Wilssens J P, Patka P, Van Lieshout E M. Functional outcome after successful internal fixation versus salvage arthroplasty of patients with a femoral neck fracture. J Orthop Trauma 2014; 28(12): e273-80.

Acta Orthopaedica 2019; 90 (1): 26–32

The volume–outcome relationship for hip fractures: a systematic review and meta-analysis of 2,023,469 patients Eveline J A WIEGERS 1, Charlie A SEWALT 1, Esmee VENEMA 1,2, Niels W L SCHEP 3, Jan A N VERHAAR 4, Hester F LINGSMA 1, and Dennis DEN HARTOG 5 1 Department 3 Department

of Public Health, Erasmus University Medical Center, Rotterdam; 2 Department of Neurology, Erasmus University Medical Center, Rotterdam; of Surgery, Maasstad Hospital, Rotterdam; 4 Department of Orthopaedics, Erasmus University Medical Center, Rotterdam; 5 Department of Surgery-Traumatology, Erasmus University Medical Center, Rotterdam, The Netherlands Correspondence: e.wiegers@erasmusmc.nl Submitted 2018-07-03. Accepted 2018-10-09.

Background and purpose — It has been hypothesized that hospitals and surgeons with high caseloads of hip fracture patients have better outcomes, but empirical studies have reported contradictory results. This systematic review and meta-analysis evaluates the volume–outcome relationship among patients with hip fracture patients. Methods — A search of different databases was performed up to February 2018. Selection of relevant studies, data extraction, and critical appraisal of the methodological quality was performed by 2 independent reviewers. A random-effects meta-analysis using studies with comparative cut-offs was performed to estimate the effect of hospital and surgeon volume on outcome, defined as in-hospital mortality and postoperative complications. Results — 24 studies comprising 2,023,469 patients were included. Overall, the quality was reasonable. 11 studies reported better health outcomes in high-volume centers and 2 studies reported better health outcomes in low-volume centers. In the meta-analysis of 11 studies there was a statistically non-significant association between higher hospital volume and both lower in-hospital mortality (adjusted odds ratio (aOR) 0.87, 95% confidence interval (CI) 0.73–1.04) and fewer postoperative complications (aOR 0.87, CI 0.75– 1.02). Four studies on surgeon volume were included in the meta-analysis and showed a minor association between higher surgeon volume and in-hospital mortality (aOR 0.92, CI 0.76–1.12). Interpretation — This systematic review and metaanalysis did not find an evident effect of hospital or surgeon volume on health outcomes. Future research without volume cut-offs is needed to examine whether a true volume–outcome relationship exists.

Several studies have shown a relationship between higher surgeon or hospital volume and better health outcomes in different areas of orthopedics, such as elective hip or knee arthroplasty (Shervin et al. 2007) and in the operative treatment of scoliosis (Vitale et al. 2005). There are several explanations for the existence of the volume–outcome relationship in surgical procedures. First, hospital staff and surgeons in particular develop more skills if they treat more patients with the same procedure. Second, hospitals with better health outcomes receive more referrals and thus increase their volume. However, studies concerning the volume–outcome relationship in hip fractures differ greatly in study design, patient population, and outcomes, which makes their results difficult to interpret. A previous systematic review of the volume–outcome relationship in orthopedic procedures found a slight association between higher volume and mortality and postoperative complications for hip fracture patients (Malik et al. 2018). However, this study included only 12 studies and did not perform a meta-analysis to estimate the size of this effect. Centralization of trauma care is of broad and current interest, which increases the volume of trauma departments in many countries. Therefore, the aim of this systematic review and meta-analysis is to evaluate and quantify the relationship between surgeon and hospital volume of hip fracture patients and health outcomes.

Methods For this systematic review and meta-analysis, we have used the Preferred Reporting Items for Systematic Reviews (PRISMA) guidelines and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.

Acta Orthopaedica 2019; 90 (1): 26â&#x20AC;&#x201C;32

Literature search The databases Embase.com (Medline and Embase), Web of Science, Cochrane Central and Google Scholar were searched until January 30, 2018 to identify published studies that examined the association between the volume of hip fractures and health outcomes. No time restrictions were set. The search strategy contained text words for hip fractures, hospital volume, and health outcomes and was developed by an experienced librarian (see Supplementary data). To identify additional relevant articles, reference lists of the included studies were searched. No additional databases or registries were searched. Inclusion and exclusion criteria All observational cohort or cross-sectional studies that provided original data on the relationship between the volume (either hospital volume or surgeon volume) of hip fractures and health outcomes (e.g., mortality, postoperative complications, length of stay, or readmission) were eligible for inclusion in this systematic review. Studies that examined and compared levels of trauma centers were also included when numbers of patients per hospital were reported. Studies on elective arthroplasty were excluded from this systematic review. Only English-language articles and articles that were available as full text were taken into account. Conference abstracts and book chapters were excluded from our search. References of included articles were screened for potentially eligible articles. In the meta-analyses we included articles that reported adjusted odds ratios (ORs), hazard ratios (HRs), or risk ratios (RRs) regarding the outcomes for in-hospital mortality and postoperative complications, because those outcomes were mentioned in more than 3 articles. Data screening and extraction 2 reviewers (CS and EW) independently screened titles and abstracts to identify potentially eligible articles. Full-text reports of such articles were retrieved and 2 reviewers (CS and EW) independently screened these full-text articles and identified eligible articles. Any disagreement was resolved through discussion or, if necessary, a third review author (HL) was consulted. The PRISMA flowchart was used to provide an overview of the data-screening process. Study characteristics (authors, study number, publication year, study design, study period, country, data source), patient characteristics (inclusion and exclusion criteria, sample size), definition of volume (unit of measurement, continuous or categorical variable with corresponding thresholds), patient outcomes, and key findings (unadjusted and adjusted estimates) were extracted from the studies by 2 independent reviewers (CS and EW). If we had data with multiple overlapping publications, we decided to include all relevant articles in our systematic review and only include the study with the highest quality in

our meta-analysis. In case of unreported or unclear data, we attempted to contact the corresponding author for clarification. Quality assessment The methodological quality of the eligible studies was independently investigated by 2 reviewers (CS and EW). Assessment of quality and generalizability of the studies was based on the key domains for observational studies (Sanderson et al. 2007). These key domains were subjectively ranked by the 2 reviewers as low or high. If any queries arose, the third reviewer (HL) was consulted. To assess potential selection bias, we examined whether inclusion and exclusion criteria were clearly described and population based. Furthermore, we assessed the methodology for measuring exposure and outcome variables. Potential design-specific sources of bias and methods to control for confounding were also examined, since these key domains are considered important when evaluating validity. For example, to check the possibility of information bias, we assessed whether studies clearly reported their cut-off of volume-groups. Data analysis To assess the potential role of publication bias, a funnel plot was made. Different measures of relative risk (ORs, HRs, RRs) were considered to be equivalent since the outcomes observed could be stated to be rare. Summary estimates of the relationship between either hospital or surgeon volume and mortality or postoperative complications and their corresponding 95% confidence intervals (CI) and 95% prediction intervals (PI) were calculated with inverse variance weighted random-effects meta-analyses to account for expected heterogeneity. Effect estimates for in-hospital mortality and postoperative complications were pooled separately for hospital and surgeon volume. Studies using 30-day mortality as outcome instead of in-hospital mortality were included in the pooled effect estimates of in-hospital mortality. For the meta-analyses, different cut-offs of number of hip fracture patients per year were examined. The number of patients that was most often used to distinguish low-volume centers from high-volume centers was used to decide which studies could be included in our meta-analysis: 170 patients per year for hospital volume and 35 patients per year for surgeon volume. ORs were provided with low-volume centers set as the reference group. Effect sizes were converted when the highest-volume group was used as a reference by taking 1/ OR. When adjusted effect estimates were unobtainable, we decided to include raw data. Statistical heterogeneity was assessed using the Cochran Q test quantified by the IÂ˛ statistic. In order to interpret the odds ratios, the baseline mortality and complication rates were calculated, weighted for the number of included patients in the meta-analysis studies. The meta-analysis was conducted with Review Manager 5.3 (https://community.cochrane.org/help/tools-and-software/ revman-5).

Acta Orthopaedica 2019; 90 (1): 26–32

Records identified through database searching n = 9,181

Additional records identified through other sources n=3

Records after duplicates removed n = 5,366 Records excluded after screening n = 5,164 Full-text articles assessed for eligibility n = 202 Excluded (n = 178): – no hip fracture population, 113 – not about volume outcome relationship, 42 – no observational cohort study, 16 – different outcome measure, 7 Studies included in qualitative synthesis n = 24 Studies included in quantitative synthesis n = 11

Figure 1. Flowchart of study selection.

Funding and potential conflicts of interest No funding sources were used for this study. No conflicts of interest were declared.

Results Included studies The literature search identified 9,181 articles and one relevant article was identified when searching the reference lists. After removing the duplicates, 5,366 articles were screened on title and abstract (Figure 1). The remaining 202 articles were assessed full-text for eligibility, which resulted in including 24 articles in the systematic review (Flood et al. 1984, Riley and Lubitz 1985, Hughes et al. 1988, Hamilton and Hamilton 1997, Hamilton and Ho 1998, Lavernia 1998, Franzo et al. 2005, Shah et al. 2005, Genuario et al. 2008, Browne et al. 2009, Forte et al. 2010, Sund 2010, Castronuovo et al. 2011, Takahashi et al. 2011, Kristensen et al. 2014, Hentschker and Mennicken 2015, van Laarhoven et al. 2015, Elkassabany et al. 2016, Guida et al. 2016, Maceroli et al. 2016, Metcalfe et al. 2016, Nimptsch and Mansky 2017, Okike et al. 2017, Treskes et al. 2017, see Table 1, Supplementary data. References with number(s) in the following refer to numbers in Table 1). 1 author (Treskes) was contacted to provide information about the number of patients in different study groups (Treskes et al. 2017). Study characteristics Of the 24 studies included, 23 studies were observational

cohort studies (Table 1, Supplementary data). 21 studies (1, 3–7, 9–17, 19–24) were retrospective cohort studies, 2 studies (2, 18) were prospective cohort studies and 1 study (8) was a cross-sectional study. 12 studies (1, 3–5, 7, 12, 14–16, 18–20) were conducted in the United States, 9 studies (2, 6, 8, 11, 13, 17, 21, 23, 24) were conducted in Europe, 2 studies (9, 10) were conducted in Canada, and 1 study (22) was conducted in Japan. Mortality was used as outcome in 23 studies (1–22, 24). In 14 studies, mortality was defined as in-hospital mortality (1, 4–7, 9–12, 14, 16, 17, 20, 24), in 7 studies as 30-day mortality (2, 3, 5, 8, 13, 15, 18), and in 2 studies as 60-day mortality (Riley and Lubitz 1985, Forte et al. 2010). 9 studies used complication rates as outcome measure (1, 7, 14–16, 18, 20, 23, 24). Typical complications assessed included postoperative infections and reoperation rates. Other outcomes reported were time to surgery, length of stay (LOS), and readmission. 21 studies evaluated hospital volume (1–13, 15–18, 20–23), 7 studies evaluated surgeon volume (1, 5, 14, 18, 19, 20, 23), and 1 study compared a level I hospital with a level II hospital (24). The cut-off used to separate high-volume centers or surgeons from low-volume centers or surgeons was highly variable. For example, 1 study used a cut-off of > 400 hip fracture patients per year (Franzo et al. 2005), while another study used ≥ 62 patients per year as a high-volume center (Shah et al. 2005). Quality assessment All of the 24 included studies were population-based, and about half of the studies were nationwide (Table 2, see Supplementary data). The majority of the studies reported the number of patients per volume group, while the cut-off of the volume groups was reported in 18 studies. Adjusted odds ratios were reported in 16 studies, but crude odds ratios were reported in 3 studies. In 16 studies, adjustments for comorbidity and patients’ demographic characteristics were made. There was a limited impact of loss to follow-up in all studies. As shown in Table 2, studies that were included in our meta-analysis were considered of higher quality than studies that were excluded from our meta-analysis. These studies more often reported cut-off of volume groups and reported their results with adjustments for patient characteristics and comorbidities. The funnel plot indicated no publication bias was present (Figures 2 and 3). Hospital volume and mortality 20 studies assessed the relationship between hospital volume and mortality, either in-hospital or 30-day (1, 2, 4–18, 20, 21, 24). 10 studies did not find any association between hospital volume and in-hospital mortality (1, 2, 4, 8, 10, 16–18, 20, 21). 8 studies observed that high-volume centers had a statistically significantly lower mortality rate compared with low volume centers (5, 7, 9, 11, 12, 14, 15, 24). There were 2 studies which reported that higher hospital volume was asso-

Acta Orthopaedica 2019; 90 (1): 26–32

SE(log[OR])

0.2

0.1

0.4

0.2

0.6

0.3

0.8

0.4 Adjusted mortality Complications

1.0

0.1

0.2

0.5

0.1

0.2

0.5

Figure 2. Funnel plot hospital volume. SE = standard error.

Figure 3. Funnel plot surgeon volume. SE = standard error.

ciated with a higher mortality rate compared with low hospital volume (6, 13). 8 studies provided adjusted odds ratios with the corresponding cut-off of 170 patients with hip fractures per year and could be included in the meta-analysis (1, 2, 3, 11, 15, 16, 18, 23). The meta-analysis showed a trend towards higher hospital volume in terms of in-hospital mortality, although it was not statistically significant (OR 0.87, CI 0.73–1.04, 95% PI 0.51–1.49). Between-study heterogeneity was large (I² = 82%) (Figure 4). The weighted baseline in-hospital mortality risk in low-volume centers for studies included in the meta-analysis was 8%.

Hospital volume and complications 8 studies evaluated the relationship between hospital volume and complications (1, 7, 15, 16, 18, 20, 23, 24). 5 studies found significantly more complications in low-volume centers compared with high volume centers (1, 15, 20, 23, 24). 3 studies described no differences in complication rates between low- and high-volume centers (7, 16, 18). 5 studies provided adjusted odds ratios for postoperative complications (1, 7, 16, 18, 23) (Figure 4). There was a statistically non-significant relationship between higher hospital volume and postoperative complications with reasonable heterogeneity (OR 0.87, CI 0.75–1.02, I² = 65%, 95% PI 0.52–1.46). The weighted baseline complication rate in low-volume centers for studies included in the meta-analysis was 16%.

Hospital volume and length of stay 8 studies assessed the relationship between hospital volume and hospital length of stay (7, 9, 10, 12, 13, 16, 20, 22). 5 studies found a significant relationship between hospital volume and length of stay (7, 13, 16, 20, 22). 4 studies reported that low-volume centers were associated with a longer length of stay (7, 16, 20, 22), while one study (13) observed that highvolume centers had a longer length of stay compared with low-volume centers. No meta-analysis regarding hospital volume and length of stay could be performed, since we did not include enough studies that provided adjusted ORs.

Figure 4. Comparisons of high- and low-volume hospitals. SE = standard error, df = degrees of freedom, and IV = inverse variance.

Acta Orthopaedica 2019; 90 (1): 26–32

Figure 5. Comparisons of high- and low-volume surgeons. SE = standard error, df = degrees of freedom, and IV = inverse variance.

Surgeon volume and mortality 2 studies observed a significant relationship between surgeon volume and in-hospital mortality (1, 20). Low surgeon volume was associated with a significant higher mortality rate compared with high surgeon volume in both studies. 3 studies did not find a significant relationship between surgeon volume and in-hospital mortality (5, 14, 19). The only meta-analysis regarding surgeon volume was on mortality. A cut-off of 15 patients per surgeon per year was used with low volume as reference group. 4 studies provided adjusted odds ratios regarding surgeon volume and in-hospital mortality (Figure 5) (1, 5, 20, 23). There was no significant relationship between surgical volume and in-hospital mortality with moderate heterogeneity (OR 0.92, CI 0.76–1.12, I² = 61%, 95% PI 0.44–1.94, p = 0.4). The weighted baseline in-hospital mortality rate for low-volume surgeons for studies included in the meta-analysis was 3.0%. Surgeon volume and complications 5 studies looked at the relationship between surgeon volume and complications (1, 14, 18, 20, 23). 4 studies found no statistically significant relationship between higher surgeon volume and postoperative infections and morbidity (1, 18, 20, 23). 1 study reported a statistically significant relationship between higher surgeon volume and complications comparing high surgeon volume to low surgeon volume (14). Surgeon volume and length of stay 3 studies evaluated the relationship between surgeon volume and length of stay. All of these studies showed that high surgeon volume was significantly associated with a lower length of stay (1, 14, 20).

Discussion This study included 24 studies that evaluated the volume– outcome relationship for hip fractures. There was no consistent effect of the impact of hospital and surgeon volume in terms of health outcomes. The quality of the included studies was reasonable. Nearly all studies were population based, reported the total number of severely injured patients, had limited impact of loss of follow-up, and reported crude ORs or

mortality percentages. However, not all studies reported their cut-off in volume groups clearly or presented adjusted ORs. The methodological approaches of assessing the relationship between hospital and/or surgeon volume and outcomes greatly varied between studies. Therefore, we applied strict inclusion criteria for the meta-analysis to reduce heterogeneity. Our meta-analysis suggested that in-hospital mortality and postoperative complications are lower in high-volume hospitals (defined as > 170 cases per year), although this relationship was not statistically significant. Also, there was no association between surgeon volume (high volume defined as > 35 cases per year) and in-hospital mortality. Definition of volume The definition of either hospital or surgeon volume is heterogeneous and in most studies an arbitrarily chosen cut-off was used. In 1 study, 62 annual cases were considered as high hospital volume (Shah et al. 2005), while this number was considered as low hospital volume in another study (Kristensen et al. 2014). For this reason, only a few studies where comparable cut-offs were used could be included in the meta-analysis. To evaluate whether the lack of a clear relationship is a real effect, it is important not to use arbitrarily selected volumes but to treat volume as a continuous variable. This makes it possible to identify whether a true volume–outcome relationship exists and decreases the information loss due to categorizing (MacCallum et al. 2002). Case-mix differences It is known that large case-mix differences exist between highand low-volume centers. For example, Level I trauma centers with a high volume of patients admit patients who have more complex fractures and higher comorbidity then Level II trauma centers (Cudnik et al. 2009). These differences may bias outcomes as patients are not randomly distributed across hospitals. Some studies clearly reported confounders used in adjustments, but most of them were insufficient to completely adjust for the case-mix differences. For instance, demographic factors and comorbidity were not always taken into account because this information was unavailable. All studies that we included in our meta-analysis did adjust for patient demographic characteristics and comorbidity.

Acta Orthopaedica 2019; 90 (1): 26–32

Underlying mechanism volume–outcome relationship It is important to investigate which underlying mechanism drives the volume–outcome relationship. The “practice makes perfect theory,” which implies that repetition of a certain procedure is associated with a better outcome, might be a good explanation for studies that report high-volume hospitals as performing better. However, this “practice makes perfect theory” cannot be fully responsible for reported volume–outcome relationships. Several potential factors of processes of care might be behind the possible effect of volume. For example, some low-volume hospitals have longer delays from admission to surgery, which might contribute to an increased mortality rate (Forte et al. 2010). In addition, it has been suggested that high-volume hospitals might have more developed pathways, processes like a reduced delay to the operating theater, and better access to alternative forms of anesthesia (Metcalfe et al. 2016). High-volume hospitals may have more established screening systems that elevate the complication rates in higher volume centers, but improve patient outcomes in the long term (Genuario et al. 2008). Furthermore, in high-volume centers physicians could often more easily make an appeal to senior doctors or physicians specialized in orthogeriatrics, which might prevent the development of complications such as postoperative delirium. On the other hand, mechanisms that are in favor of low-volume hospitals are proposed as well. As indicated by a previous study, the effect of volume decreased: those with 350 hip fracture patients or more had higher mortality rates. First, high-volume hospitals less frequently use guidelines on recommended processes of care, which might be a key mediator at high-volume hospitals. Second, it is speculated that hip fracture patients experience less attention from nurses in high-volume hospitals since they would have to “compete” with more complicated orthopedic surgery patients, although evidence for this mechanism does not exist (Kristensen et al. 2014). The surgeon volume relationship is expected to be influenced by several factors. The selection of the appropriate procedure and intraoperative technique might differ due to local standards or guidelines. Furthermore, surgical outcome may also be influenced by preoperative planning and postoperative care (Browne et al. 2009), which highly depends on the workflow in a hospital. Most studies did only focus either on the number of cases per year per surgeon or the number of cases per year per hospital without accounting for the overall experience of the individual surgeon. In order to gain a better understanding of the interaction among experience, surgeon, and hospital volume, future research should account for surgeon experience. Since there are also studies that did not find a positive or a negative association between volume and patient outcomes, we expect other factors that are not related to volume to play a role in the outcome of hip fracture patients as well. Improvements in surgical and anesthetic techniques or an extended

application of thromboembolism and hospital infection prevention protocols might be alternative explanations for reductions of mortality risk and postoperative complications (Franzo et al. 2005). Most studies included only patients who underwent arthroplasties (both hemi- and total hip). However, some studies also included patients who underwent open reduction or internal fixation or both. This might possibly have influenced the results, since high-volume centers or surgeons might more easily choose to do more complicated surgery. Further research is needed to assess the importance of these factors compared with hospital and or surgeon volume. Limitations In addition to the bias caused by case-mix differences, our study could also suffer from publication bias. Studies showing no volume–outcome relationship, or an inverse volume–outcome relationship might be unpublished. But, as the funnel plot shows no correlation between effect size and their standard error, publication bias seems unlikely. Furthermore, mortality, complications, and length of stay might not be sensitive enough to detect outcome differences after hip fractures. Other quality indicators such as operation time or quality of life might be more influenced by hospital or surgeon volume than hard outcomes such as mortality. All included studies in our meta-analysis used mortality as outcome; however, variability in the time-point of mortality and volume cut-offs is a likely source of heterogeneity and a potential limitation of our study. Nevertheless, substantial systematic differences are unlikely to be present since different time-points of mortality could still indicate the effect of volume on outcomes. Furthermore, we used a random effects model to account for between-study heterogeneity. In summary, this systematic review and meta-analysis did not find an evident effect of either hospital or surgeon volume on different health outcomes. Studies examining the volume– outcome relationship in patients with hip fracture appear to be heterogeneous, specifically the cut-offs that are used. Future research without volume cut-offs is needed to examine whether a true volume–outcome relationship exists, to assess the best cut-off for high volume and determine which processes of care are important in the care of patients with hip fracture. Supplementary data Tables 1–2 and the search strategy are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2018.1545383

The authors would like to thank Wichor Bramer, biomedical information specialist from Erasmus Medical Center, for his help with the search strategy. Also, they would like to thank Kaij Treskes (PhD student, Academic Medical Center Amsterdam) for providing data for the meta-analysis.

EW/CS: literature search, study selection, data extraction, data analysis, drafting the manuscript—contributed equally. EV/HL: study selection, dataanalysis, critical revision of the manuscript. NS/JV: critical revision of the manuscript.DDH: study selection, critical revision of the manuscript. Acta thanks Michael Brix and Pia Kjær Kristensen for help with peer review of this study.

Browne J A, Pietrobon R, Olson S A. Hip fracture outcomes: does surgeon or hospital volume really matter? J Trauma 2009; 66(3): 809-14. doi: 10.1097/ TA.0b013e31816166bb. Castronuovo E, Pezzotti P, Franzo A, Di Lallo D, Guasticchi G. Early and late mortality in elderly patients after hip fracture: a cohort study using administrative health databases in the Lazio region, Italy. BMC Geriatr 2011; 11: 37. doi: 10.1186/1471-2318-11-37. Cudnik M T, Newgard C D, Sayre M R, Steinherg S M. Level I versus Level II trauma centers: an outcomes-based assessment. J Trauma 2009; 66(5): 1321-6. doi: 10.1097/TA.0b013e3181929e2b. Elkassabany N M, Passarella M, Mehta S, Liu J, Neuman M D. Hospital characteristics, inpatient processes of care, and readmissions of older adults with hip fractures. J Am Geriatr Soc 2016; 64(8): 1656-61. doi: 10.1111/jgs.14256. Flood A B, Scott W R, Ewy W. Does practice make perfect? Part I: The relation between hospital volume and outcomes for selected diagnostic categories. Med Care 1984; 22(2): 98-114. Forte M L, Virnig B A, Swiontkowski M F, Bhandari M, Feldman R, Eberly L E, Kane R L. Ninety-day mortality after intertrochanteric hip fracture: does provider volume matter? J Bone Joint Surg 2010; 92(4): 799-806. doi: 10.2106/jbjs.h.01204. Franzo A, Francescutti C, Simon G. Risk factors correlated with post-operative mortality for hip fracture surgery in the elderly: a population-based approach. Eur J Epidemiol 2005; 20(12): 985-91. doi: 10.1007/s10654005-4280-9. Genuario J, Koval K J, Cantu R V, Spratt K F. Does hospital surgical volume affect in-hospital outcomes in surgically treated pelvic and acetabular fractures? Bull NYU Hosp Jt Dis 2008; 66(4): 282-9. Guida P, Iacoviello M, Passantino A, Scrutinio D. Intra-hospital correlations among 30-day mortality rates in 18 different clinical and surgical settings. Int J Qual Health Care 2016; 28(6): 793-801. doi: 10.1093/intqhc/mzw112. Hamilton B H, Hamilton V H. Estimating surgical volume. Outcome relationships applying survival models: accounting for frailty and hospital fixed effects. Health Econ 1997; 6(4): 383-95. doi: 10.1002/(sici)10991050(199707)6:4<383::aid-hec278>3.0.co; 2-l. Hamilton B H, Ho V. Does practice make perfect? Examining the relationship between hospital surgical volume and outcomes for hip fracture patients in Quebec. Med Care 1998; 36(6): 892-903. Hentschker C, Mennicken R. The volume–outcome relationship and minimum volume standards: empirical evidence for Germany. Health Econ 2015; 24(6): 644-58. doi: 10.1002/hec.3051. Hughes R G, Garnick D W, Luft H S, McPhee S J, Hunt S S. Hospital volume and patient outcomes: the case of hip fracture patients. Med Care 1988; 26(11): 1057-67. Kristensen P K, Thillemann T M, Johnsen S P. Is bigger always better? A nationwide study of hip fracture unit volume, 30-day mortality, quality of

Acta Orthopaedica 2019; 90 (1): 26–32

in-hospital care, and length of hospital stay. Med Care 2014; 52(12): 10239. doi: 10.1097/mlr.0000000000000234. Lavernia C J. Hemiarthroplasty in hip fracture care: effects of surgical volume on short-term outcome. J Arthroplasty 1998; 13(7): 774-8. doi: 10.1016/ s0883-5403(98)90029-8. MacCallum R C, Zhang S, Preacher K J, Rucker D D. On the practice of dichotomization of quantitative variables. Psychol Methods 2002; 7(1): 19-40. Maceroli M, Nikkel L E, Mahmood B, Ketz J P, Qiu X, Ciminelli J, Messing S, Elfar J C. Total hip arthroplasty for femoral neck fractures: improved outcomes with higher hospital volumes. J Orthop Trauma 2016; 30(11): 597-604. doi: 10.1097/bot.0000000000000662. Malik A T, Panni U Y, Masri B A, Noordin S. The impact of surgeon volume and hospital volume on postoperative mortality and morbidity after hip fractures: A systematic review. Int J Surg 2018; 54(Pt B): 316-27. doi: 10.1016/j.ijsu.2017.10.072. Metcalfe D, Salim A, Olufajo O, Gabbe B, Zogg C, Harris M B, Perry D C, Costa M L. Hospital case volume and outcomes for proximal femoral fractures in the USA: an observational study. BMJ Open 2016; 6(4): e010743. Nimptsch U, Mansky T. Hospital volume and mortality for 25 types of inpatient treatment in German hospitals: observational study using complete national data from 2009 to 2014. BMJ Open 2017; 7(9): e016184. doi: 10.1136/bmjopen-2017-016184. Okike K, Chan P H, Paxton E W. Effect of surgeon and hospital volume on morbidity and mortality after hip fracture. J Bone Joint Surg Am 2017; 99(18): 1547-53. doi: 10.2106/JBJS.16.01133. Riley G, Lubitz J. Outcomes of surgery among the Medicare aged: surgical volume and mortality. Health Care Financ Rev 1985; 7(1): 37-47. Sanderson S, Tatt I D, Higgins J P. Tools for assessing quality and susceptibility to bias in observational studies in epidemiology: a systematic review and annotated bibliography. Int J Epidemiol 2007; 36(3): 666-76. Shah S N, Wainess R M, Karunakar M A. Hemiarthroplasty for femoral neck fracture in the elderly: surgeon and hospital volume-related outcomes. J Arthroplasty 2005; 20(4): 503-8. doi: 10.1016/j.arth.2004.03.025. Shervin N, Rubash H E, Katz J N. Orthopaedic procedure volume and patient outcomes: a systematic literature review. Clin Orthop Relat Res 2007; (457): 35-41. doi: 10.1097/BLO.0b013e3180375514. Sund R. Modeling the volume–effectiveness relationship in the case of hip fracture treatment in Finland. BMC Health Serv Res 2010; 10:238. doi: 10.1186/1472-6963-10-238. Takahashi C, Fushimi K, Matsuda S. Factors associated with a protracted hospital stay after hip fracture surgery in Japan. Geriatr Gerontol Int 2011; 11(4): 474-81. doi: 10.1111/j.1447-0594.2011.00711.x. Treskes K, Voeten S C, Tol M C J M, Zuidema W P, Vermeulen J, Goslings J C, Schep N W L. Trauma surgery by general surgeons: still an option for proximal femoral fractures? Injury 2017; 48(2): 339-44. doi: 10.1016/j. injury.2016.11.020. van Laarhoven J J, van Lammeren G W, Houwert R M, van Laarhoven C J, Hietbrink F, Leenen L P, Verleisdonk E J. Isolated hip fracture care in an inclusive trauma system: a trauma system wide evaluation. Injury 2015; 46(6): 1042-6. doi: 10.1016/j.injury.2015.02.015. Vitale M A, Arons R R, Hyman J E, Skaggs D L, Roye D P, Vitale M G. The contribution of hospital volume, payer status, and other factors on the surgical outcomes of scoliosis patients: a review of 3,606 cases in the State of California. J Pediatr Orthop 2005; 25(3): 393-9.

Acta Orthopaedica 2019; 90 (1): 33–39

Selective serotonin reuptake inhibitor use in hip fracture patients: a Danish nationwide prevalence study Stine B BRUUN 1, Irene PETERSEN 1,2, Nickolaj R KRISTENSEN 1, Deirdre CRONIN-FENTON 1, and Alma B PEDERSEN 1 1 Department

of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark; 2 Department of Primary Care and Population Health, University College London, UK Correspondence: stinebb@post.au.dk Submitted 2018-07-27. Accepted 2018-10-16.

Background and purpose — Selective serotonin reuptake inhibitors (SSRIs) are often used in the elderly and are associated with adverse effects. Therefore, it is important to ascertain the prevalence of SSRI use in fragile and surgerytreated hip fracture patients. Methods — This population-based prevalence study included Danish hip fracture patients aged ≥ 65 years operated in 2006–2016 (n = 68,607) and matched individuals from the background population (n = 343,020). Using Poisson regression, prevalence risk ratios (PRRs) with 95% confidence intervals (CIs) were estimated to compare hip fracture patients with the general population, and to estimate the association between hip fracture patient characteristics and SSRI prescriptions. Results — The prevalence of SSRI use among hip fracture patients was 23% compared with 12% in the general population. During 2006–2016, the prevalence decreased from 26% to 18% among hip fracture patients and from 13% to 10% in the general population. Factors associated with SSRI use in hip fracture patients were age 75–84 years (PRR 1.18, CI 1.13–1.23), age ≥ 85 years (PRR 1.17, CI 1.11–1.22), female sex (PRR 1.13, CI 1.09–1.17), unmarried status (PRR 1.15, CI 1.11–1.19), living in a residential institution (PRR 2.30, CI 2.19–2.40), Charlson comorbidity index (CCI) score 1–2 (PRR 1.50, CI 1.45–1.55), CCI score 3+ (PRR 1.62, CI 1.55–1.69), and several medications. Interpretation — The prevalence of SSRI use was high among hip fracture patients compared with the general population. Our data stress the importance of continued clinical awareness of frailty, comorbidity, and polypharmacy of hip fracture patients and the potentially adverse drug effects of SSRI treatment.

Hip fracture affects approximately 6,500 elderly aged 65 years or older every year in Denmark (5.7 million inhabitants) (Centre for Clinical Epidemiology and Biostatistics North 2017, Statistics Denmark 2017). Most hip fractures in the elderly result from falls; only 5% of hip fracture patients have no fall history (Parker and Johansen 2006). Falls in the elderly are associated with several risk factors including gait and balance impairment, frailty, disability, comorbid conditions, and polypharmacy (Vieira et al. 2016). Polypharmacy is frequent in the elderly and often includes use of anticoagulants, statins, antipsychotics, and antidepressants (Maher et al. 2014). The most commonly prescribed antidepressant drugs are selective serotonin reuptake inhibitors (SSRIs) (Mars et al. 2017). SSRI use is associated with an increased risk of falls resulting in an emergency department visit or hospitalization (Macri et al. 2017). SSRI use is further associated with an increased risk of hip fracture (Souverein et al. 2016), which may be due to bone loss, hyponatremia, or hemodynamic disturbances (Bakken et al. 2013). SSRI use may be more prevalent in hip fracture patients compared with the general population (Hung et al. 2017). However, there are a limited number of studies on this. Furthermore, factors associated with SSRI use in hip fracture patients are not identified. One study in individuals aged 65 years or older without hip fracture found that SSRI users were more likely to be women and aged 75–94 years (Marengoni et al. 2016). SSRI use was strongly associated with use of other drugs such as other psychiatric drugs, statins, antihypertensives, and anticoagulants (Marengoni et al. 2016). Another study found an association between hypertension, dementia, depression, and polypharmacy and potentially inappropriate medication use including SSRI use in individuals aged 65 years or older receiving health care services (Alhmoud et al. 2015). Some of these associated factors may contribute to the

increased risk of hip fracture in SSRI users. However, no previous studies have investigated this. We therefore assessed the prevalence of SSRI use in elderly hip fracture patients before the fracture compared with the prevalence of SSRI use in the general population. Furthermore, we aimed to identify factors associated with SSRI use in hip fracture patients before occurrence of the fracture.

Methods Setting and design The study is a nationwide cohort study using prospectively collected data from Danish medical registries. The health care system in Denmark is tax-funded, and all 5.7 million citizens have equal access to health care services (Statistics Denmark 2017). Data sources The Danish Multidisciplinary Hip Fracture Registry (Sørensen et al. 2009) contains detailed pre- and postoperative clinical data on all patients aged 65 years or older with first-time hip fracture admitted to any orthopedic department in Denmark since 2004. These data were linked to data from the Danish Civil Registration System (Schmidt et al. 2014). The Danish Civil Registration System was established in 1968 and holds data on date of birth, vital status, and migration on all individuals in Denmark. Every citizen has a unique civil personal registration number, which allows for individual-level linkage across all Danish registries. First, data were linked to the Danish National Patient Registry (Schmidt et al. 2015), which contains data on civil personal registration number, hospital admission and discharge diagnosis codes, and diagnostic and surgical procedure codes from all Danish somatic hospitals since 1977. Diagnoses were coded using the International Classification of Diseases Eighth Revision (ICD-8) until the end of 1993 and Tenth Revision (ICD-10) thereafter. Furthermore, data were linked to the Danish National Database of Reimbursed Prescriptions (Johannesdottir et al. 2012), which tracks reimbursed medicine dispensing at all community pharmacies and hospital-based outpatient pharmacies in Denmark since 2004. Prescription information on individuals living in nursing homes is included in the database as they receive medication from community pharmacies. The database holds data on civil personal registration number, Anatomical Therapeutic Chemical code, redemption date, item quantity, pack size, defined daily dose, dose form, and generic substitution done at pharmacy. Study population Hip fracture patients aged 65 years or older undergoing surgical treatment for hip fracture between 1 January 2006 and 31 December 2016 were identified in the Danish Multidisciplinary Hip Fracture Registry. Using the Civil Registration

Acta Orthopaedica 2019; 90 (1): 33–39

System, each hip fracture patient was frequency matched with up to 5 individuals from the general population on age and sex at the time of surgery (index date). Variables Information on age and sex of all participants was obtained from the Danish Civil Registration System. 3 age categories were created: 65–74 years, 75–84 years, and ≥ 85 years. Comorbidities of all participants were identified using the Danish National Patient Registry. Overall comorbidity was summarized according to the Charlson Comorbidity Index (CCI) score. The CCI was categorized as low (score 0), medium (score 1–2), and high (score ≥ 3) comorbidity score. The following medications were assessed from the Danish National Database of Reimbursed Prescriptions: SSRIs, nonsteroid anti-inflammatory drugs (NSAIDs), corticosteroids, anticoagulants, statins, non-SSRI antidepressants, and antipsychotics. Use of each medication including SSRIs was defined as follows: users redeemed at least 1 prescription within 1 year and non-users redeemed no prescriptions within 365 days before hip fracture surgery date (index date). Marital status of hip fracture patients was obtained from the Danish Civil Registration System. Housing, operation year, and BMI information of hip fracture patients were assessed from the Danish Multidisciplinary Hip Fracture Registry. 4 categories comprising housing information were created: own accommodation, residential institution, homeless, and unknown. Likewise, 5 categories based on BMI values were created: underweight (BMI < 18.5), normal weight (BMI ≥ 18.5 < 25), overweight (BMI ≥ 25 < 30), obese (BMI ≥ 30), and unknown. All codes defining study variables are available in Tables 1–3 in the Supplementary data. Statistics The annual prevalence of SSRI prescription redemption both overall and stratified by generic type was calculated in both hip fracture patients and the general population sample. SSRI use in hip fracture patients was compared with SSRI use in the general population sample using Poisson regression analyses and stratifying by age, sex, CCI, and other medication use. The model was evaluated for effect modification by age and sex. Crude and adjusted prevalence risk ratios (PRR) of SSRI use in hip fracture patients were estimated according to patient characteristics using Poisson regression analysis for comparison. The regression analyses were adjusted for age and sex. All statistical analyses were performed using Stata 14 for Windows (Stata Corp, College Station, TX, USA). Ethics, funding, and potential conflicts of interest Ethical approval was not required. As this study did not involve contact with patients or an intervention, it was not necessary to obtain permission from the Danish Scientific Ethical Committee. The study was approved by the Danish Data Protection Agency (record number: 1-16-02-467-15). This work

Acta Orthopaedica 2019; 90 (1): 33–39

Table 4. Participant characteristics according to selective serotonin reuptake inhibitor (SSRI) use 2006–2016. Values are frequency (%) Variable

All participants

Hip fracture patients SSRI users Non-users

General population SSRI users Non-users

Total 411,627 (100) 16,081 (23) 52,526 (77) 41,191 (12) 301,829 (88) Median age 83 (64–108) 84 (65–105) 83 (65–108) 86 (64–106) 83 (64–108) Age 65–74 79,667 (19) 2,705 (20) 10,585 (80) 4,690 (7) 61,687 (93) 75–84 157,030 (38) 6,331 (24) 19,797 (76) 14,569 (11) 116,333 (89) ≥ 85 174,930 (43) 7,045 (24) 22,144 (76) 21,932 (15) 123,809 (85) Sex Male 118,706 (29) 4,219 (21) 15,567 (79) 8,288 (8) 90,632 (92) Female 292,921 (71) 11,862 (24) 36,959 (76) 32,903 (13) 211,197 (87) Operation year 2006 37,681 (9) 1,618 (26) 4,663 (74) 4,070 (13) 27,330 (87) 2007 38,406 (9) 1,687 (26) 4,714 (74) 4,102 (13) 27,903 (87) 2008 41,028 (10) 1,753 (26) 5,085 (74) 4,370 (13) 29,820 (87) 2009 37,351 (9) 1,599 (26) 4,627 (74) 3,979 (13) 27,146 (87) 2010 39,024 (10) 1,684 (26) 4,820 (74) 4,183 (13) 28,337 (87) 2011 39,012 (9) 1,591 (24) 4,911 (76) 4,163 (13) 28,347 (87) 2012 37,230 (9) 1,442 (23) 4,763 (77) 3,815 (12) 27,210 (88) 2013 37,255 (9) 1,320 (21) 4,890 (79) 3,583 (12) 27,462 (88) 2014 36,462 (9) 1,209 (20) 4,868 (80) 3,203 (11) 27,182 (89) 2015 36,180 (9) 1,206 (20) 4,824 (80) 3,118 (10) 27,032 (90) 2016 31,998 (8) 972 (18) 4,361 (82) 2,605 (10) 24,060 (90) CCI score Low (0) 255,954 (62) 6,470 (19) 27,979 (81) 20,459 (9) 201,046 (91) Medium (1–2) 121,213 (30) 6,917 (28) 17,999 (72) 15,894 (17) 80,403 (83) High (3+) 34,460 (8) 2,694 (29) 6,548 (71) 4,838 (19) 20,380 (81) Other medication NSAIDs 85,307 (21) 3,559 (24) 11,033 (76) 8,677 (12) 62,038 (88) Corticosteroids 34,400 (8) 1,884 (27) 5,214 (73) 4,224 (15) 23,078 (85) Anticoagulants 184,999 (45) 9,077 (27) 24,463 (73) 22,373 (15) 129,086 (85) Statins 111,799 (27) 4,565 (26) 13,040 (74) 11,711 (12) 82,483 (88) Non-SSRI antidepressants 39,550 (10) 3,798 (38) 6,117 (62) 9,683 (33) 19,952 (67) Antipsychotics 20,579 (6) 2,683 (42) 3,669 (58) 5,123 (36) 9,104 (64) CCI: Charlson Comorbidity Index

was supported by the Independent Research Fund Denmark (grant number 6120-00034). There are no conflicts of interest to declare.

Results We identified 68,607 first-time hip fracture patients aged 65 years or older operated between 2006 and 2016 and 343,020 age- and sex-matched individuals from the general population. 16,081 (23%) hip fracture patients were SSRI users compared with 41,191 (12%) in the general population sample. Most participants were women (71%) with a median age of 83 (64–108) years at the index date (Table 4). Overall, 62% of the participants had a low CCI score, 29% had a medium CCI score, and 8% had a high CCI score, but the proportion of SSRI users increased with increasing CCI score in both hip fracture patients and the general population sample. Use of other medication was common among all participants and NSAIDs, anticoagulants, and statins were prescribed to more

than 20% (Table 4). Almost half of the hip fracture patients lived in their own accommodation (31,076, 45%) whereas 7,105 (10%) lived in a residential institution, where the proportion of SSRI users was larger (41%) compared with patients living in their own accommodation (18%). Most hip fracture patients were normal weight (31,635, 46), while 17,640 (26%) were overweight or obese. 30,404 (44%) hip fracture patients were missing housing data and 13,329 (19%) were missing BMI data (Table 5, see Supplementary data).

Prevalence of SSRI use Between 2006 and 2016, the overall prevalence of SSRI use decreased from 26% to 18% in hip fracture patients and from 13% to 10% in the general population sample (Table 4 and Figure). Overall, the prevalence of SSRI use in both populations was stable from 2006 to 2011, after which the prevalence decreased until the end of the study period. Citalopram was the most frequently redeemed SSRI and decreased from 18% to 11% in hip fracture patients and from 9% to 6% in the general population sample during the study period. The overall prevalence of SSRI prescriptions decreased except for sertraline prescriptions, which increased from 2% to 4% in hip fracture patients and from 1% to 2% in the general population sample between 2006 and 2016. Table 6 shows the PRRs of SSRI use in hip fracture patients compared with SSRI use in the general population sample. Overall, hip fracture patients had a higher prevalence of SSRI use compared with the general population. When stratifying by age, hip fracture patients aged 65–74 years had a PRR of 2.88 (CI 2.75–3.02), patients aged 75–84 years had a PRR of 2.18 (CI 2.11–2.24), and patients aged 85 years or older had a PRR of 1.60 (CI 1.56–1.65) compared with the general population of the same age. Male hip fracture patients were more than twice as likely to have SSRI treatment compared with the general population. The prevalence of SSRI treatment for female hip fracture patients was also nearly doubled compared with the general population. Hip fracture patients had a higher prevalence of SSRI use than the general population in all CCI groups and when stratifying on other medication use. However, the PRR was only 1.20 (CI 1.16–1.25) when stratifying on non-SSRI antidepressant use and 1.19 (CI 1.14–1.25) when stratifying on antipsychotic use comparing hip fracture patients with the general population.

Acta Orthopaedica 2019; 90 (1): 33–39

hip fracture patients were also associated with increased use of SSRIs compared with low Fluoxetin CCI score. Considering the individual comorCitalopram bidities, hip fracture patients with dementia 25 25 Paroxetin were twice as likely to receive SSRI treatment Sertralin Fluvoxamin 20 20 compared with hip fracture patients without Escitalopram dementia. Other comorbidities indicating 15 15 higher probability of SSRI use among hip fracture patients were myocardial infarction, 10 10 congestive heart failure, cerebrovascular disease, chronic pulmonary disease, ulcer dis5 5 ease, liver disease, diabetes, hemiplegia, and 0 0 renal disease. Hip fracture patients using other 2006 2008 2010 2012 2014 2016 2006 2008 2010 2012 2014 2016 medication such as NSAIDs, corticosteroids, Year Year anticoagulants, and statins had a higher probAnnual prevalence of selective serotonin reuptake inhibitor use in hip fracture patients (left panel) and in the general population (right panel), 2006–2016, overall and stratified ability of redeeming SSRIs compared with by generic type. non-users of other medication. Hip fracture patients using non-SSRI antidepressants and antipsychotics were more likely to receive SSRI treatment Table 6. Selective serotonin reuptake inhibitor (SSRI) use in hip compared with non-users (Table 7). fracture patients compared with SSRI use in the general population Prevalence (%)

Prevalence (%)

Total

sample

Variable

Unadjusted PRR (95% CI)

Adjusted PRR a (95% CI)

Discussion

The prevalence of SSRI use among hip fracture patients was Total 2.0 (1.9–2.0) 2.0 (1.9–2.0) approximately double the prevalence in the general populaAge 65–74 2.9 (2.8–3.0) 2.9 (2.8–3.0) tion sample during 2006 to 2016, irrespective of age, sex, and 75–84 2.2 (2.1–2.2) 2.2 (2.1–2.2) CCI score at index date. SSRI prevalence was higher among ≥ 85 1.6 (1.6–1.7) 1.6 (1.6–1.7) younger hip fracture patients compared with elderly patients. Sex Male 2.5 (2.5–2.6) 2.5 (2.5–2.6) Citalopram was the most frequently prescribed SSRI in both Female 1.8 (1.8–1.8) 1.8 (1.8–1.8) hip fracture patients and the general population sample. SevCCI score eral factors including age 75 years and above, female sex, Low (0) 2.0 (2.0–2.1) 1.9 (1.9–2.0) Medium (1–2) 1.7 (1.6–1.7) 1.7 (1.7–1.8) unmarried status, living in a residential institution, comorbidHigh (3+) 1.5 (1.5–1.6) 1.5 (1.5–1.6) ity, and use of other medication were associated with higher Other medication SSRI use among hip fracture patients. NSAIDs 2.0 (1.9–2.1) 2.0 (1.9–2.1) Corticosteroids 1.7 (1.6–1.8) 1.8 (1.7–1.9) Strengths of this study include the large nationwide study Anticoagulants 1.8 (1.8–1.9) 1.9 (1.8–1.9) population, enabling us both to study the use of SSRIs in hip Statins 2.1 (2.0–2.2) 2.1 (2.0–2.2) fracture patients and to compare this with use in the genNon-SSRI antidepressants 1.2 (1.1–1.2) 1.2 (1.2–1.3) Antipsychotics 1.2 (1.1–1.2) 1.2 (1.1–1.3) eral Danish population. The data originate from registries a Adjusted

for age and sex. CCI: Charlson Comorbidity Index

Factors associated with SSRI use in hip fracture patients Hip fracture patients aged 75–84 years and 85 years or more were more likely to receive SSRI treatment than hip fracture patients aged 65–74 years. In hip fracture patients, SSRI use was higher among women than among men. Likewise, SSRI use was higher among unmarried hip fracture patients than among married. Living in a residential institution markedly increased the likelihood of receiving SSRI treatment as a hip fracture patient. Medium CCI score and high CCI score in

based on a tax-supported and uniformly organized health care system, reducing the risk of recall bias. SSRIs are only available by prescription, which ensures accurate data on the redemption of the medication in the study population. The population of Danish hip fracture patients is similar regarding age and sex to the populations used in other studies on hip fractures in the elderly (Lawrence et al. 2002, Roche et al. 2005). However, the study presents some limitations. First, redemption of the prescribed medication does not necessarily mean that patients were compliant and took the medication. However, prescription redemption is considered a good measure of medication intake even in the presence of misclassification (Schneeweiss and Avorn 2005). Furthermore, patients redeeming 2 prescriptions are more likely to take their medication and a sensitivity analysis in a previous study

Acta Orthopaedica 2019; 90 (1): 33–39

the Danish National Patient Registry, which holds only data on patients treated in the hospital sector. Hence, hip fracture patients may be more likely to have registered comorbidities because they are admitted to hospital. This may lead to either Variable Unadjusted PRR Adjusted PRR a non-differential misclassification or residual confounding. (95% CI) (95% CI) Third, the comorbidity information did not include data on Age psychiatric diagnoses. However, we included other medica65–74 1.0 1.0 tion such as NSAIDs, corticosteroids, anticoagulants, statins, 75–84 1.2 (1.1–1.3) 1.2 (1.1–1.2) non-SSRI antidepressants, and antipsychotics, which may ≥ 85 1.2 (1.1–1.2) 1.2 (1.1–1.2) Sex help to estimate the association between psychiatric diagnoMale 1.0 1.0 ses and diagnoses treated in general practice only and SSRI Female 1.1 (1.1–1.2) 1.1 (1.1–1.2) use in hip fracture patients. Fourth, we did not have informaMarital status Married 1.0 1.0 tion on smoking or alcohol use, which may be risk factors for Unmarried 1.2 (1.2–1.2) 1.2 (1.1–1.2) several comorbidities and SSRI use. However, Souverein et Housing al. (2016) found that additional adjustment for these variables Own accommodation 1.0 1.0 Homeless 0.5 (0.1–2.0) 0.5 (0.1–2.1) had limited impact on effect estimates examining the associaResidential institution 2.3 (2.2–2.4) 2.3 (2.2–2.4) tion between SSRI use and the risk of hip fracture. The CCI Operation year included chronic obstructive pulmonary disease, which may 2006 1.0 1.0 2007 1.0 (1.0–1.1) 1.0 (1.0–1.1) function as a surrogate measure of smoking. Fifth, housing 2008 1.0 (0.9–1.1) 1.0 (0.9–1.1) and BMI data were incomplete, but previous studies using 2009 1.0 (0.9–1.1) 1.0 (0.9–1.1) multiple imputation of BMI on hip fracture data showed no 2010 1.0 (0.9–1.1) 1.0 (0.9–1.1) 2011 1.0 (0.9–1.0) 1.0 (0.9–1.0) impact on the estimates after adjustment for BMI (Pedersen 2012 0.9 (0.8–1.0) 0.9 (0.8–1.0) et al. 2017). 2013 0.8 (0.8–0.9) 0.8 (0.8–0.9) Overall, the prevalence of SSRI use was doubled in hip 2014 0.8 (0.7–0.8) 0.8 (0.7–0.8) 2015 0.8 (0.7–0.8) 0.8 (0.7–0.8) fracture patients compared with the general population 2016 0.7 (0.7–0.8) 0.7 (0.7–0.8) sample. A study conducted in the UK reported a prevalence of Body mass index SSRI use of 6% in hip fracture patients and 3% in the general < 18.5: Underweight 1.1 (1.0–1.1) 1.0 (1.0–1.1) ≥ 18.5 < 25: Normal weight 1.0 1.0 population (Hubbard et al. 2003). However, the study was ≥ 25: Overweight or obese 1.0 (1.0–1.0) 1.0 (1.0–1.1) conducted between 1987 and 1999, when SSRIs were generCharlson comorbidity index ally less used (Mars et al. 2017). Nonetheless, the prevalence Low (0) 1.0 1.0 Medium (1–2) 1.5 (1.4–1.5) 1.5 (1.5–1.6) ratio between hip fracture patients and the general population High (3+) 1.6 (1.5–1.6) 1.6 (1.6–1.7) is similar to our results. The decrease in prevalence between Comorbidity 2006 and 2016 was reflected in both hip fracture patients Myocardial infarction 1.1 (1.0–1.2) 1.1 (1.0–1.2) Congestive heart failure 1.1 (1 .1–1.2) 1.1 (1.1–1.2) and the general population sample. The observed decrease in Peripheral vascular disease 1.1 (1.1–1.2) 1.2 (1.1–1.2) SSRI prescribing among hip fracture patients aged 65 years Cerebrovascular disease 1.6 (1.5–1.6) 1.6 (1.5–1.7) and older may be explained by an increased clinical awareDementia 2.0 (1.9–2.1) 2.0 (1.9–2.1) Chronic pulmonary disease 1.3 (1.2–1.3) 1.3 (1.2–1.3) ness of the adverse effects related to SSRIs, which have been Connective tissue disease 1.0 (0.9–1.1) 1.0 (0.9–1.1) subject to debate during recent years (Topiwala et al. 2014). Ulcer disease 1.3 (1.3–1.4) 1.3 (1.3–1.4) Our finding that citalopram was the most frequent subtype of Liver disease 1.1 (1.0–1.3) 1.2 (1.0–1.3) Diabetes type 1 and 2 1.1 (1.0–1.1) 1.1 (1.1–1.2) prescribed SSRIs is consistent with results obtained from the Hemiplegia 1.6 (1.2–2.0) 1.6 (1.3–2.1) general UK population (Lockhart and Guthrie 2011, McCrea Moderate to severe renal disease 1.1 (1.1–1.2) 1.2 (1.1–1.3) et al. 2016). Cancer 1.0 (1.0–1.1) 1.0 (1.0–1.1) Other medication The prevalence of SSRI use increased with increasing age NSAIDs 1.1 (1.0–1.1) 1.1 (1.0–1.1) and CCI score, and women had a higher prevalence of SSRI Corticosteroids 1.2 (1.1–1.2) 1.2 (1.1–1.2) use than men in both hip fracture patients and the general Anticoagulants 1.4 (1.3–1.4) 1.4 (1.3–1.4) Statins 1.2 (1.1–1.2) 1.2 (1.1–1.2) population sample. However, when comparing hip fracture Non-SSRI antidepressants 1.8 (1.8–1.9) 1.8 (1.8–1.9) patients with the general population sample the relative prevAntipsychotics 2.0 (1.9–2.0) 1.0 (1.9–2.1) alence increases with decreasing age and CCI score and is a Adjusted for age and sex. larger in men than in women. This indicates that SSRI use may be more prevalent in male hip fracture patients aged 65–74 years with low CCI score than in female hip fracture showed no difference between hip fracture patients redeem- patients aged 85 years or older with high CCI score. Bakken ing only 1 prescription and those redeeming 2 (Bruun et al. et al. (2013) found a more pronounced excess risk of hip 2018). Second, comorbidity information was obtained from fracture among men exposed to SSRIs than among exposed Table 7. Prevalence risk ratios (PRRs) of selective serotonin reuptake inhibitor (SSRI) use according to hip fracture patients’ characteristics

women. This indicates that exposure to SSRIs may affect the risk of hip fracture relatively more in men than in women. Our results support this finding as the relative prevalence of SSRI use in hip fracture patients compared with the general population is higher in men than in women. A Swedish study found that high serum serotonin and SSRI use predicts a lower bone mineral density and an increased risk of hip fracture in elderly men (Kristjansdottir et al. 2018). This may explain the higher proportion of male SSRI users among elderly hip fracture patients compared with the general population. A study by Zhou et al. (2018) further supports this as it shows no difference between female SSRI users and non-users when comparing bone mineral density values. However, further investigations are needed to assess whether the effect of SSRIs is more pronounced in men than in women. Our results further suggest that the younger age group and lower CCI score group of hip fracture patients have a relatively higher prevalence of SSRI use than the older age group and higher CCI score group. Nevertheless, there may be a risk of bias as we only had data on comorbidities treated in the hospital. Hip fracture patients may be more likely to have these comorbidities registered as they were admitted to hospital than the general population sample. An Italian study investigated factors associated with SSRI use among the elderly aged 65 years or older. The study found an association between SSRI use and female sex, age 75–94 years, and other medication such as psychiatric drugs, statins, and antihypertensives (Marengoni et al. 2016), which is similar to our findings in hip fracture patients. However, we also found an association with unmarried status, living in a residential institution, and several comorbidities. These lifestyle factors and conditions suggests that hip fracture patients receiving SSRI treatment are more fragile than nonusers. We expect any potential bias to be non-differential as we do not expect registration of the associated factors to differ between hip fracture patients using SSRIs and nonusers. Our results suggest that hip fracture patients using SSRIs may have a more complicated course of treatment than hip fracture patients not treated with SSRIs. Though several factors were associated with SSRI use among elderly hip fracture patients, further studies are needed in order to uncover whether these patients are more prone to complications after surgery. In summary, this nationwide cohort study based on prospectively collected data from population-based Danish registries found a high prevalence of SSRI use of almost one-quarter among elderly hip fracture patients versus one-tenth in the general population sample. Factors associated with SSRI use in hip fracture patients were age 75 years and above, female sex, unmarried status, living in a residential institution, CCI score 1–2 and ≥ 3, and use of other medication. Our data stress the importance of continued clinical awareness of frailty among hip fracture patients and potentially monitoring any evidence of adverse drug effects of SSRI treatment.

Acta Orthopaedica 2019; 90 (1): 33–39

Supplementary data Supplementary Tables 1–3 and 5 are available in the online version of this article, http://dx.doi.org/ 10.1080/17453674. 2018.1543842 The authors would like to thank the staff of the hospital departments caring for patients with hip fracture for their continuous effort and contribution to acquisition of data in the Danish Multidisciplinary Hip Fracture Registry. ABP conceived the study. SBB, ABP, DCF, and IP designed the study. SBB and ABP undertook data application. NRK undertook the initial data management with input from SBB, ABP, DCF, and IP. SBB conducted the statistical analyses and drafted the first manuscript, with ABP, DCF, and IP contributing to subsequent iterations. All authors provided critical input on the manuscript. Acta thanks Jan-Erik Gjertsen and Karl Michaëlsson for help with peer review of this study. Alhmoud E, Khalifa S, Bahi AA. Prevalence and predictors of potentially inappropriate medications among home care elderly patients in Qatar. Int J Clin Pharm 2015; 37(5): 815-21. Bakken MS, Engeland A, Engesaeter L B, Ranhoff A H, Hunskaar S, Ruths S. Increased risk of hip fracture among older people using antidepressant drugs: data from the Norwegian Prescription Database and the Norwegian Hip Fracture Registry. Age Ageing 2013; 42(4): 514-20. Bruun S B, Petersen I, Kristensen N R, Cronin-Fenton D, Pedersen A B. Selective serotonin reuptake inhibitor use and mortality, postoperative complications, and quality of care in hip fracture patients: a Danish nationwide cohort study. Clin Epidemiol 2018; 10: 1053-71. Centre for Clinical Epidemiology and Biostatistics North. The Danish Multidisciplinary Hip Fracture Register—National Annual Report 2017, The Danish Clinical Registries. Hubbard R, Farrington P, Smith C, Smeeth L , Tattersfield A. Exposure to tricyclic and selective serotonin reuptake inhibitor antidepressants and the risk of hip fracture. Am J Epidemiol 2003; 158(1): 77-84. Hung S C, Lin C H, Hung H C, Lin C L, Lai S W. Use of selective serotonin reuptake inhibitors and risk of hip fracture in the elderly: a case-control study in Taiwan. J Am Med Dir Assoc 2017; 18(4): 350-4. Johannesdottir S A, Horvath-Puho E, Ehrenstein V, Schmidt M, Pedersen L, Sorensen H T. Existing data sources for clinical epidemiology: the Danish National Database of Reimbursed Prescriptions. Clin Epidemiol 2012; 4: 303-13. Kristjansdottir H L, Lewerin C, Lerner U H, Waern E, Johansson H, Sundh D, Karlsson M, Cummings S R, Zetterberg H, Lorentzon M, Ohlsson C, Mellstrom D. High serum serotonin predicts increased risk for hip fracture and nonvertebral osteoporotic fractures: the MrOS Sweden Study. J Bone Miner Res 2018; 33(9): 1560-7. Lawrence V A, Hilsenbeck S G, Noveck H, Poses R M, Carson J L. Medical complications and outcomes after hip fracture repair. Arch Intern Med 2002; 162(18): 2053-7. Lockhart P, Guthrie B. Trends in primary care antidepressant prescribing 1995–2007: a longitudinal population database analysis. Br J Gen Pract 2011; 61(590): e565-72. Macri J C, Iaboni A, Kirkham J G, Maxwell C, Gill S S, Vasudev A, Whitehead M, Seitz D P. Association between antidepressants and fall-related injuries among long-term care residents. Am J Geriatr Psychiatry 2017; 25(12): 1326-36. Maher R L, Hanlon J, Hajjar E R. Clinical consequences of polypharmacy in elderly. Expert Opin Drug Saf 2014; 13(1): 57-65. Marengoni A, Onder G, Degli Esposti L, Russo P, Sangiorgi D, Buda S, Fini M, Marchionni N, Bonassi S, Mammarella F, Marrocco W, Pozzi G, Palmer

Acta Orthopaedica 2019; 90 (1): 33–39

K, Monaco A, Pecorelli S, Pani L. Adherence to selective serotonin and serotonin-norepinephrine reuptake inhibitor prescriptions affects overall medication adherence in older persons: evidence from the Italian Nationwide OsMed Health-DB Database. J Clin Psychiatry 2016; 77(12): 1712-18. Mars B, Heron J, Kessler D, Davies N M, Martin R M, Thomas K H, Gunnell D. Influences on antidepressant prescribing trends in the UK: 1995–2011. Soc Psychiatry Psychiatr Epidemiol 2017; 52(2): 193-200. McCrea R L, Sammon C J, Nazareth I, Petersen I. Initiation and duration of selective serotonin reuptake inhibitor prescribing over time: UK cohort study. Br J Psychiatry 2016; 209(5): 421-6. Parker M, Johansen A. Hip fracture. BMJ 2006; 333(7557): 27-30. Pedersen A B, Mikkelsen E M, Cronin-Fenton D, Kristensen N R, Pham T M, L Pedersen, Petersen I. Missing data and multiple imputation in clinical epidemiological research. Clin Epidemiol 2017; 9: 157-66. Roche J J, Wenn R T, Sahota O, Moran C G. Effect of comorbidities and postoperative complications on mortality after hip fracture in elderly people: prospective observational cohort study. BMJ 2005; 331(7529): 1374. 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, Sandegaard J L, Ehrenstein V, Pedersen L, Sorensen H T. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7: 449-90.

Schneeweiss S, Avorn J. A review of uses of health care utilization databases for epidemiologic research on therapeutics. J Clin Epidemiol 2005; 58(4): 323-37. Souverein P C, Abbing-Karahagopian V, Martin E, Huerta C, Abajo F, Leufkens H G M, Candore G, Alvarez Y, Slattery J, Miret M, Requena G, Gil M J, Groenwold R H H, Reynolds R, Schlienger R G, Logie J W, Groot M C H, Klungel O H, Staa T P, Egberts T C G, De Bruin M L, Gardarsdottir H. Understanding inconsistency in the results from observational pharmacoepidemiological studies: the case of antidepressant use and risk of hip/femur fractures. Pharmacoepidemiol Drug Saf 2016; 25(S1): 88-102. Statistics Denmark Statistical Yearbook 2017. Copenhagen, Statistics Denmark; 2017. Sørensen H T, Christensen T, Schlosser H K, Pedersen L. Use of Medical Databases in Clinical Epidemiology. Aarhus, SUN-TRYK, Aarhus Universitet; 2009. Topiwala A, Chouliaras L, Ebmeier K P. Prescribing selective serotonin reuptake inhibitors in older age. Maturitas 2014; 77(2): 118-23. Vieira E R, Palmer R C, Chaves P H. Prevention of falls in older people living in the community. BMJ 2016; 353: i1419. Zhou C, Fang L, Chen Y, Zhong J, Wang H, Xie P. Effect of selective serotonin reuptake inhibitors on bone mineral density: a systematic review and meta-analysis. Osteoporos Int 2018; 29(6): 1243-51.

Acta Orthopaedica 2019; 90 (1): 40–45

Patient-reported outcome and muscle–tendon pain after periacetabular osteotomy are related: 1-year follow-up in 82 patients with hip dysplasia Julie Sandell JACOBSEN 1,2, Kjeld SØBALLE 3,4, Kristian THORBORG 5, Lars BOLVIG 6, Stig STORGAARD JAKOBSEN 3, Per HÖLMICH 5, and Inger MECHLENBURG 3,4,7 1 Department

of Physiotherapy and Department of Research in Rehabilitation and Health Promotion, Faculty of Health Science, VIA University College, Aarhus, Denmark; 2 Department of Physiotherapy and Occupational Therapy, Aarhus University Hospital, Denmark; 3 Department of Orthopaedic Surgery, Aarhus University Hospital, Denmark; 4 Department of Clinical Medicine, Aarhus University, Aarhus, Denmark; 5 Sports Orthopaedic Research CenterCopenhagen (SORC-C), Department of Orthopaedic Surgery, Copenhagen University Hospital, Amager-Hvidovre, Denmark; 6 Department of Radiology, Aarhus University Hospital, Denmark; 7 Department of Public Health, Aarhus University, Aarhus, Denmark Correspondence: jsaj@via.dk Submitted 2018-08-08. Accepted 2018-11-08.

Background and purpose — Larger prospective studies investigating periacetabular osteotomy (PAO) with patientreported outcome measures developed for young patients are lacking. We investigated changes in patient-reported outcome (PRO), changes in muscle–tendon pain, and any associations between them from before to 1 year after PAO. Patients and methods — Outcome after PAO was investigated in 82 patients. PRO was investigated with the Copenhagen Hip and Groin Outcome Score (HAGOS). Muscle– tendon pain in the hip and groin region was identified with standardized clinical tests, and any associations between them were analyzed with multivariable linear regressions. Results — HAGOS subscales improved statistically significantly from before to 1 year after PAO with effect sizes ranging from medium to very large (0.66–1.37). Muscle– tendon pain in the hip and groin region showed a large decrease in prevalence from 74% (95% CI 64–83) before PAO to 35% (95% CI 25–47) 1 year after PAO. Statistically significant associations were observed between changes in HAGOS and change in the sum of muscle–tendon pain, ranging from –4.7 (95% CI –8.4 to –1.0) to –8.2 (95% CI –13 to –3.3) HAGOS points per extra painful entity across all subscales from before to 1 year after PAO. Interpretation — Patients with hip dysplasia experience medium to very large improvements in PRO 1 year after PAO, associated with decreased muscle–tendon pain. The understanding of hip dysplasia as solely a joint disease should be reconsidered since muscle–tendon pain seems to play an important role in relation to the outcome after PAO.

Traditionally, hip dysplasia is considered a joint disease with insufficient coverage of the femoral head, which is related to early painful degenerative changes (Mechlenburg 2008, Ross et al. 2011). Pain and physical function can be improved by periacetabular osteotomy (PAO) (Hartig-Andreasen et al. 2012, Lerch et al. 2017), a well-established surgical treatment of symptomatic hip dysplasia in young patients. We have previously challenged this traditional understanding as we reported a high prevalence of muscle–tendon pain in young patients with hip dysplasia, which negatively affected patientreported outcome (PRO) (Jacobsen et al. 2018b). PAO is commonly investigated with patient-reported outcome measures (PROMs) developed for older patients with osteoarthritis (Hartig-Andreasen et al. 2012, Clohisy et al. 2017). In young patients, however, a high score on a PROM designed for older patients does not necessarily indicate an acceptable health status, as more strenuous activities such as sports and recreational activities are not evaluated. Only a few previous studies have investigated PRO after PAO with PROMs developed for young and physically active patients (Jacobsen et al. 2014, Khan et al. 2017). In these studies, outcomes were investigated with the Non Arthritic Hip Score (NAHS) and the Copenhagen Hip and Groin Outcome Score (HAGOS). The results of the studies showed medium to very large effect sizes. The NAHS, however, was developed for and validated in patients with hip osteoarthritis (Khan et al. 2017), while the study using HAGOS was fairly small and focused on objective gait characteristics rather than PRO (Jacobsen et al. 2014). Therefore, it is warranted to investigate the outcome of the PAO with PROMs designed for young and active patients, including the possible negative effect of muscle–tendon pain. We investigated changes in PRO, changes in muscle–tendon pain, and any associations between them from before to 1 year after PAO.

Acta Orthopaedica 2019; 90 (1): 40–45

Patients and methods Patients from the Department of Orthopaedics at Aarhus University Hospital in Denmark were prospectively included from May 2014 to August 2015. They were part of a study population from a previous study (Jacobsen et al. 2018b) and were included if Wiberg’s centre-edge (CE) angle was < 25° (Wiberg 1939), if they had had groin pain for at least 3 months, and if they were scheduled for PAO. Furthermore, only patients < 45 years, with BMI < 30, normal range of motion (minimum 110° of hip flexion), and with Tönnis’s osteoarthritis grade < 2 (Tönnis 1987) were operated on. Patients with comorbidities and previous surgical interventions affecting their hip function were excluded. Further details on study design are reported in our previous studies on the same study population, reporting prevalence of muscle–tendon pain and structural abnormalities before PAO (Jacobsen et al. 2018a, 2018b). Before PAO, patient characteristics including age, sex, and duration of pain were recorded based on standardized questions. A single rater measured the CE angle, the Tönnis acetabular index (AI) angle (Tönnis 1987), and the Tönnis osteoarthritis grade on standing anteroposterior radiographs. Information on comorbidities and previous treatments was extracted from hospitals charts. Pain was recorded using the FABER (Flexion/Abduction/External Rotation) test and the FADDIR (Flexion/Adduction/Internal Rotation) test (Troelsen et al. 2009, Martin et al. 2010). Additionally, a standardized test was used to record occurrence of internal snapping hip (Tibor and Sekiya 2008). Periacetabular osteotomy The minimally invasive transsartorial approach for PAO was performed by 2 experienced orthopedic surgeons via 3 separate osteotomies (Troelsen et al. 2008). In short, an approximately 7 cm incision was made alongside the sartorius muscle beginning at the anterior superior iliac spine. The sartorius muscle was divided parallel with the direction of its fibers. The medial part of the split muscle was retracted medially together with the iliopsoas muscle, and this was followed by osteotomies. The patients were presented with a standardized post-surgery rehabilitation program on the ward, and discharged after approximately 2 days of hospitalization. Partial weight-bearing was allowed in the first 6–8 weeks. Moreover, all patients were offered an individual-based rehabilitation program of 2 weekly training sessions starting 6 weeks after PAO and lasting generally for 2–4 months. Patient-reported outcome measure The HAGOS was completed before and 1 year after PAO by all patients prior to assessment of muscle–tendon pain. HAGOS consists of 6 separate subscales covering pain, symptoms, physical function in daily living (ADL), physical function in sports and recreation (sports/recreation), participation

restriction (PA), and quality of life (QOL) (Thorborg et al. 2011). HAGOS is tailored to reflect physically active young and middle-aged patients with long-standing hip and/or groin pain. HAGOS measures the patient’s hip-related disability during the past week on a scale from 0 to 100 points, where 100 points indicates best possible outcome. Pain at rest was measured before and 1 year after PAO on a numeric rating scale (NRS) from 0 (no pain) to 10 (unbearable pain). Muscle–tendon pain We assessed muscle–tendon pain before and 1 year after PAO with standardized clinical examinations originally proposed by Hölmich et al. (2004) and modified by the Doha consensus statement (Weir et al. 2015). The standardized examination covers a number of pain-provoking tests including anatomical palpation, resistance testing, and passive muscle stretch in specific anatomical regions, named clinical entities (Weir et al. 2015). In our study, we added identification of pain in the hamstrings and abductors following the same principle as the other clinical entities, since these structures were considered important in our study. Moreover, pain related to the rectus abdominis (Hölmich et al. 2004) was the focus in this study as an alternative to inguinal-related pain defined in the Doha consensus since most patients were women. Muscle–tendon pain was assessed in 5 anatomical entities: the iliopsoas, the abductors, the adductors, the hamstrings, and the rectus abdominis. Muscle–tendon pain was investigated within each entity and as the sum of positive clinical entities for each patient, ranging from 0 to 5. The outcome of each entity test was dichotomous (pain yes or no). Sample size considerations The aim of this study was to investigate 1-year outcome of the PAO using HAGOS. Moderate effect sizes were considered relevant, which would require a sample size of 85 patients to show an effect size of 0.5 from before to 1 year after PAO. This sample size is based on a power of 90% and a significance level of 5%. Statistics Parametric continuous data were reported as means with standard deviations (SD) if normally distributed, otherwise reported as medians with interquartile ranges (IQR). Histograms and probability plots were used to test for normality. Categorical data were reported as numbers of events and percentage with a 95% confidence interval (CI). Changes in each HAGOS subscale measured before and 1 year after PAO were considered important and were tested with paired t-tests. Similarly, changes in each muscle–tendon pain entity from before to 1 year after PAO were considered equally important. These were tested with the McNemar test. Effect sizes of HAGOS and muscle–tendon pain were calculated from the paired t-test as Cohen’s d on the formula:

Acta Orthopaedica 2019; 90 (1): 40–45

Table 2. Patient-reported outcome of PAO in 82 patients with hip dysplasia

Patients scheduled for PAO with informed consent n = 100 Lost to 1-year follow-up (n = 18): – postponed surgery, 7 – time and transportation, 4 – serious disease unrelated to PAO, 3 – injuries unrelated to PAO, 2 – non-union of the pubic bone, 1 – emigrated, 1 Paired analysis n = 82

Flowchart of the study process. 100 consecutive patients with symptomatic hip dysplasia scheduled for periacetabular osteotomy (PAO) were included from Department of Orthopedic Surgery, Aarhus University Hospital, Denmark.

1 year Before PAO after PAO Difference Effect size Outcome score mean (SD) mean (SD) mean (CI) Cohen’s d p-value HAGOS Pain Symptoms ADL Sport/recreation PA Quality of life NRS Pain (IQR)

Age, years a BMI a Men, n Duration of pain, years b Bilateral hip dysplasia, n CE angle (°) a AI angle (°) a Tönnis’s osteoarthritis grade 1, n Positive FADIR test, n Positive FABER test, n Positive internal snapping hip test, n

Before PAO 30 (9) 23 (3) 11 3 (1–6) 74 17 (5) 14 (5) 3 70 62 25

76 (17) 26 (22–30) 67 (19) 19 (15–23) 81 (18) 27 (22–31) 64 (23) 25 (20–31) 44 (33) 21 (14–28) 57 (25) 28 (22–33) 0 (0–2)

1.37 0.99 1.25 1.02 0.66 1.11 0.74

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

Abbreviations: HAGOS = Copenhagen Hip and Groin Outcome Score (0–100 points), ADL = physical function in daily living, PA = preferred physical activity participation, NRS = numeric rating scale, IQR = interquartile range.

Table 1. Characteristics of 82 consecutive patients with hip dysplasia Patient characteristics

50 (17) 48 (17) 55 (22) 39 (21) 23 (24) 29 (14) 3 (2–5)

1 year after PAO 31 (9) 24 (3) – – – 30 (5) 3 (4) 5 55 47 16

Characteristics are presented as numbers or a mean values (standard deviation), b median values (interquartile range) Abbreviations: CE = centre-edge, AI = Tönnis’s acetabular index, FADIR = Flexion/Adduction/Internal Rotation, FABER = Flexion/Abduction/External Rotation.

t statistic/√(n), and from McNemar test as Cohen’s w on the formula: w statistic/√(n). Finally, crude and adjusted multivariable linear regression analyses were performed to assess associations between changes in HAGOS (pain, symptoms, ADL, sport/recreation, PA, and QOL) and change in the sum of muscle–tendon pain entities (i.e., the sum of positive clinical entities for each patient) from before to 1 year after PAO. Changes in each HAGOS subscale were the dependent variables, and the sum of muscle–tendon pain entities for each patient was the independent variable. Potential co-variates were identified using causal diagrams for observational research, based on knowledge from previous studies (Greenland et al. 1999). Co-variates included in the analysis were CE angles measured before and 1 year after PAO (continuous), age (continuous), and sex (dichotomous). Crude and adjusted β-coefficients were estimated and the assumptions

(independent observations, linear association, constant variance of residuals, and normal distribution of residuals) of the regression models were met. The β-coefficients refer to the slope of the regression line, indicating a decrease in changed PRO per unit increase in muscle–tendon pain. The STATA 14.0 (StataCorp, College Station, TX, USA) software package was used for data analysis, and results were considered statistically significant if p < 0.05. Ethics, registration, funding, and potential conflicts of interest This study was conducted and reported in accordance with the WMA declaration of Helsinki and the STROBE statement. All patients gave informed consent to participate, and ethical approval was obtained from the Central Denmark Region Committee on Biomedical Research Ethics (5/2014). The Danish Data Protection Agency (1-16-02-47-14) authorized the handling of personal data, and the protocol was registered at ClinicalTrials.gov (20140401PAO). This study was funded by the Danish Rheumatism Association, the Aase and Ejnar Danielsen Fund, and the Fund of Family Kjaersgaard, Sunds. The authors declare that they have no potential conflicts of interest.

Results Informed consent was obtained from 100 consecutive patients before PAO. 18 patients were lost to follow-up (Figure), leaving 82 patients for this study (Table 1). We found no statistically significant differences in any of the listed patient characteristics between the 18 patients lost to follow-up and the analyzed patients (data not shown). Patient-reported outcome We found statistically significant increases in all HAGOS subscales from before to 1 year after PAO (Table 2). The effect

Acta Orthopaedica 2019; 90 (1): 40–45

Table 3. Muscle–tendon pain in 82 patients with hip dysplasia Clinical entities Iliopsoas-related pain Abductor-related pain Adductor-related pain Hamstrings-related pain Rectus-abdominis-related pain Patients with minimum 1 positive clinical entity

Before PAO 1 year after PAO % (CI) % (CI)

Difference % points (CI)

Effect size Cohen’s w p-value

54 (42–65) 37 (26–48) 12 (6–21) 6 (2–14) 4 (0–10)

22 (14–32) 15 (8–24) 7 (3–15) 1 (0–7) 0 (0–0)

–32 (–46 to –17) –22 (–36 to –8) –5 (–16 to 6) –5 (–12 to 2) –4 (–9 to 2)

–1.96 –1.12 –0.11 –0.29 –0.33

< 0.001 0.002 0.5 0.2 0.3

74 (64–83)

35 (25 to 47) –39 (–54 to –24)

–2.46

< 0.001

Table 4. Associations between change in HAGOS (0–100 points) and change in muscle–tendon pain (n = 82)

Discussion

We investigated changes in PRO, changes in muscle–tendon pain, and any associations between them from before to 1 year after PAO. Our patients Pain –4.7 (–8.5 to –0.9) 0.02 –4.7 (–8.4 to –1.0) 0.01 reported clinically relevant improvements in all Symptoms –4.8 (–8.6 to –1.0) 0.02 –4.7 (–8.6 to –0.9) 0.02 HAGOS subscales with effect sizes ranging from ADL –6.1 (–10 to –1.9) 0.005 –6.2 (–10 to –2.1) 0.004 medium to very large. All improvements were Sport/recreation –5.9 (–11 to –1.0) 0.02 –6.0 (–11 to –0.9) 0.02 PA –1.2 (–7.9 to 5.5) 0.7 –1.2 (–7.9 to 5.6) 0.7 higher than the minimally important change (MIC) Quality of life –8.2 (–13 to –3.3) 0.001 –8.2 (–13 to –3.3) 0.001 of the HAGOS (Thomeé et al. 2014). Neverthea Adjusted for CE angles measured before and 1 year after PAO, age, and sex. less, 1 year after PAO, one-third of the patients Abbreviations: See Table 2 still experienced muscle–tendon pain, most commonly affecting the iliopsoas and the hip abductors. Furthermore, our regression analyses showed that sizes ranged from 0.66 to 1.37, corresponding to medium to change in the sum of muscle–tendon pain was associated with very large effect sizes. Moreover, 26/82 patients reported a changes in PRO of the PAO. Only 2 previous studies have investigated the outcome of sport/recreation score of 0–50 points 1 year after PAO, 14/82 patients reported a sport/recreation score of >50–70 points, PAO with PROMs tailored to reflect activity levels of young while 42/82 patients reported a sport/recreation score >70 and active patients. Jacobsen et al. (2014) reported PRO after points 1 year after PAO. Similarly, NRS pain decreased sta- PAO, applying HAGOS, and found medium to very large tistically significantly, corresponding to a medium effect size improvements corresponding to effect sizes from 0.78 to 1.75 in a study population of 29 patients with hip dysplasia. of 0.74. Similarly, a consecutive study on 168 patients with hip dysMuscle–tendon pain plasia reported very large patient-reported improvement corThe analysis of the individual entities showed a significant responding to an effect size of 1.5 measured with the NAHS decrease in iliopsoas-related pain and abductor-related pain (Khan et al. 2017). The HAGOS subscale scores, we found, from before to 1 year after PAO, whereas the decrease in ranged from 44 to 81 points, which is in accordance with the the other 3 entities was not statistically significant (Table 3). study by Jacobsen et al. (2014), reporting scores of 50 to 90 Moreover, the proportion of patients with minimum 1 positive HAGOS points 1 year after PAO in another series of patients. muscle–tendon pain entity decreased by 39 percentage points The association between change in HAGOS PA and change in the sum of muscle–tendon pain entities was not statisti(CI 24–54) 1 year after PAO. cally significant. However, the lack of association between Associations between changes in HAGOS and change in HAGOS PA and change in muscle–tendon pain change in muscle–tendon pain can be explained since HAGOS PA measures patients’ selfApart from HAGOS PA, both crude and adjusted analyses perceived ability to participate in a preferred physical activshowed statistically significant negative linear association ity (Thorborg et al. 2011), which is not related to specific between changes in HAGOS and change in the sum of muscle– functions and activities like the other HAGOS subscales. tendon pain entities, ranging from –4.7 to –8.2 HAGOS points Therefore, it is not surprising that perceived participatory per extra painful entity across all subscales from before to 1 capacity is not reflected in change in muscle–tendon pain year after PAO (Table 4). and vice versa. Crude Adjusted a HAGOS β coefficient (CI) p-value β coefficient (CI) p-value

We did not include a healthy control group; however, an age- and sex-matched healthy control group was included in a previous study, investigating the outcome 1 year after PAO using HAGOS (Jacobsen et al. 2014). All healthy subjects reported a PRO of 84–100 HAGOS sport/recreation points, which was different from the HAGOS sport/recreation score measured 1 year after PAO in the patients with hip dysplasia (Jacobsen et al. 2014). In this study, half of the patients reported a HAGOS score ≤ 70 sport/recreation points 1 year after PAO. This indicates that patients still experience a low functional level, associated with muscle–tendon pain. Our patients showed a clinically relevant reduction in pain related to the iliopsoas and the hip abductors. This decrease may be a result of both surgical reorientation and the intensive individual-based rehabilitation program. The effect of muscle–tendon pain on changes in HAGOS score was 5–8 HAGOS points for every change in number of painful clinical entities across all HAGOS subscales, which is in line with the MIC of HAGOS (Thomeé et al. 2014). This suggests that resolving just 1 painful clinical entity after PAO has a clinically relevant impact on the outcome 1 year after PAO. Muscle–tendon affection in patients with hip dysplasia has only been reported in 1 previous study, which reported muscle–tendon affection among one-fifth of the patients identified by hip arthroscopy (Domb et al. 2014). However, associations between muscle–tendon affection and patient-reported outcomes were not investigated. Methodological considerations and limitations As reported in our previous study on the same study population, we found low inter-rater reliability of the muscle–tendon pain assessment, indicating a high random variation in our estimates (Jacobsen et al. 2018b). The potentially negative impact of this is small because we included a large study population. The associations between changes in HAGOS and change in the sum of muscle–tendon pain were mainly driven by patients who either reduced many painful entities or increased the number of painful entities 1 year after PAO. The same analysis without these patients would have resulted in smaller associations or even lack of significant associations between HAGOS and muscle–tendon pain. Summary Patients with hip dysplasia experience medium to very large improvements in PRO 1 year after PAO, associated with decreased muscle–tendon pain. The understanding of hip dysplasia as solely a joint disease should be reconsidered as muscle–tendon pain seems to play an important role in relation to the outcome after PAO. Future intervention studies ought to investigate the optimal treatment for patients with hip dysplasia and coexisting muscle–tendon pain.

Acta Orthopaedica 2019; 90 (1): 40–45

All authors were involved in the planning and writing of the manuscript. IM, KT, PH, and LB gave supervision on design and methods. KS and SSJ included and operated on all patients. JSJ examined patients, while SSJ made all radiologic evaluations. JSJ coordinated all elements, wrote the first description of the study, did all analyses, and wrote the first draft of the manuscript. The authors would like to thank Charlotte Møller Sørensen and Gitte Hjørnholm Madsen for their invaluable assistance carrying out the clinical examinations. Acta thanks Minne Heeg and Yvette Pronk for help with peer review of this study.

Clohisy J C, Ackerman J, Baca G, Baty J, Beaulé P E, Kim Y-J, Millis M B, Podeszwa D A, Schoenecker P L, Sierra R J, Sink E L, Sucato D J, Trousdale R T, Zaltz I. Patient-reported outcomes of periacetabular osteotomy from the prospective ANCHOR cohort study. J Bone Joint Surg Am 2017; 99(1): 33-41. Domb B G, Lareau J M, Baydoun H, Botser I, Millis M B, Yen Y-M. Is intraarticular pathology common in patients with hip dysplasia undergoing periacetabular osteotomy? Clin Orthop Relat Res 2014; 472(2): 674-80. Greenland S, Pearl J, Robins J M. Causal diagrams for epidemiologic research. Epidemiology 1999; 10(1): 37-48. Hartig-Andreasen C, Troelsen A, Thillemann T M, Søballe K. What factors predict failure 4 to 12 years after periacetabular osteotomy? Clin Orthop Relat Res 2012; 470(11): 2978-87. Hölmich P, Hölmich L R, Bjerg A M. Clinical examination of athletes with groin pain: an intraobserver and interobserver reliability study. Br J Sports Med 2004; 38(4): 446-51. Jacobsen J S, Bolvig L, Hölmich P, Thorborg K, Jakobsen S S, Søballe K, Mechlenburg I. Muscle–tendon-related abnormalities detected by ultrasonography are common in symptomatic hip dysplasia. Arch Orthop Trauma Surg 2018a; 138(8): 1059-67. Jacobsen J S, Hölmich P, Thorborg K, Bolvig L, Jakobsen S S, Søballe K, Mechlenburg I. Muscle–tendon-related pain in 100 patients with hip dysplasia: prevalence and associations with self-reported hip disability and muscle strength. J Hip Preserv Surg 2018b; 5(1): 39-46. Jacobsen J S, Nielsen D B, Sørensen H, Søballe K, Mechlenburg I. Joint kinematics and kinetics during walking and running in 32 patients with hip dysplasia 1 year after periacetabular osteotomy. Acta Orthop 2014; 85(6): 592-9. Khan O H, Malviya A, Subramanian P, Agolley D, Witt J D. Minimally invasive periacetabular osteotomy using a modified Smith–Petersen approach. Bone Joint J 2017; 99-B(1): 22-8. Lerch T D, Steppacher S D, Liechti E F, Tannast M, Siebenrock K A. Onethird of hips after periacetabular osteotomy survive 30 years with good clinical results, no progression of arthritis, or conversion to THA. Clin Orthop Relat Res 2017; 475(4): 1154-68. Martin H D, Kelly B T, Leunig M, Philippon M J, Clohisy J C, Martin R L, Sekiya J K, Pietrobon R, Mohtadi N G, Sampson T G, Safran M R. The pattern and technique in the clinical evaluation of the adult hip: the common physical examination tests of hip specialists. Arthroscopy 2010; 26(2): 161-72. Mechlenburg I. Evaluation of Bernese periacetabular osteotomy. Acta Orthop 2008; 79(Suppl 329): 1-43. Ross J R, Zaltz I, Nepple J J, Schoenecker P L, Clohisy J C. Arthroscopic disease classification and interventions as an adjunct in the treatment of acetabular dysplasia. Am J Sports Med 2011; 39(Suppl): 72S-8S.

Acta Orthopaedica 2019; 90 (1): 40–45

Thomeé R, Jónasson P, Thorborg K, Sansone M, Ahldén M, Thomeé C, Karlsson J, Baranto A. Cross-cultural adaptation to Swedish and validation of the Copenhagen Hip and Groin Outcome Score (HAGOS) for pain, symptoms and physical function in patients with hip and groin disability due to femoro-acetabular impingement. Knee Surg Sports Traumatol Arthrosc 2014; 22(4): 835-42. Thorborg K, Hölmich P, Christensen R, Petersen J, Roos E M. The Copenhagen Hip and Groin Outcome Score (HAGOS): development and validation according to the COSMIN checklist. Br J Sports Med 2011; 45(6): 478-91. Tibor L M, Sekiya J K. Differential diagnosis of pain around the hip joint. Arthrosc J Arthrosc Relat Surg 2008; 24(12): 1407-21. Tönnis D. Congenital dysplasia and dislocation of the hip in children and adults. Berlin Heidelberg, New York: Springer; 1987.

Troelsen A, Elmengaard B, Søballe K. A new minimally invasive transsartorial approach for periacetabular osteotomy. J Bone Joint Surg Am 2008; 90(3): 493-8. Troelsen A, Mechlenburg I, Gelineck J, Bolvig L, Jacobsen S, Søballe K. What is the role of clinical tests and ultrasound in acetabular labral tear diagnostics? Acta Orthop 2009; 80(3): 314-8. Weir A, Brukner P, Delahunt E, Ekstrand J, Griffin D, Khan K M, Lovell G, Meyers W C, Muschaweck U, Robinson P, De Vos R-J, Vuckovic Z. Doha agreement meeting on terminology and definitions in groin pain in athletes. Br J Sport Med 2015; 49: 768-74. Wiberg G. Studies on dysplastic acetabula and congenital subluxation of the hip joint. Acta Orthop Scand Suppl 1939; 58: 1-132.

Acta Orthopaedica 2019; 90 (1): 46–52

Relationship between outcome scores and knee laxity following total knee arthroplasty: a systematic review Andreas KAPPEL 1, Mogens LAURSEN 1, Poul T NIELSEN 1, and Anders ODGAARD 2 1 Department

of Orthopedic Surgery/Clinical Institute, Aalborg University Hospital, Aalborg, Denmark; 2 Department of Orthopedic Surgery, Copenhagen University Hospital Herlev-Gentofte, Hellerup, Denmark Correspondence: andreas.kappel@rn.dk Submitted 2018-08-17. Accepted 2018-11-06.

Background and purpose — Instability following primary total knee arthroplasty (TKA) is, according to all national registries, one of the major failure mechanisms leading to revision surgery. However, the range of soft-tissue laxity that favors both pain relief and optimal knee function following TKA remains unclear. We reviewed current evidence on the relationship between instrumented knee laxity measured postoperatively and outcome scores following primary TKA. Patients and methods — We conducted a systematic search of PubMed, Embase, and Cochrane databases to identify relevant studies, which were cross-referenced using Web of Science. Results — 14 eligible studies were identified; all were methodologically similar. Both sagittal and coronal laxity measurement were reported; 6 studies reported on measurement in both extension and flexion. In knee extension from 0° to 30° none of 11 studies could establish statistically significant association between laxity and outcome scores. In flexion from 60° to 90° 6 of 9 studies found statistically significant association. Favorable results were reported for posterior cruciate retaining (CR) knees with sagittal laxity between 5 and 10 mm at 75–80° and for knees with medial coronal laxity below 4° in 80–90° of flexion. Interpretation — In order to improve outcome following TKA careful measuring and adjusting of ligament laxity intraoperatively seems important. Future studies using newer outcome scores supplemented by performance-based scores may complement current evidence.

Modifiable surgical factors influence the outcome of TKA procedures. Implant alignment, soft tissue balancing, and choice of implant constraint is dependent on preoperative anatomical conditions, surgical technique, and the experience, preference, and thoroughness of the surgeon. Implant alignment is known to affect both revision rate (Ritter et al. 2011, Gromov et al. 2014, Kim et al. 2014, Lee et al. 2018) and outcome (Longstaff et al. 2009, Huang et al. 2012, Gromov et al. 2014), but the influence of soft tissue laxity is not as well described and has not previously been the subject of systematic review. In TKA surgery, knee laxity is evaluated both intraoperatively and at follow-up. Different surgical techniques to obtain optimal soft tissue balance have been described (Babazadeh et al. 2009, Mihalko et al. 2009), and numerous tools, such as computer-assisted surgery, trial insert sensors, tensioners, spreaders, spatulas and spacer blocks, have been developed to assist the surgeon quantify intraoperative laxity. However, intraoperative evaluation of soft tissue laxity is still challenging and can among other factors be influenced by the position of the patella, muscular tension, and the external load on the knee. Furthermore, the laxity measured with trial implants might change following implantation of the final implant (Nodzo et al. 2017) and the ligament tension might change following surgery. In most cases soft-tissue balance is based on subjective assessment, and therefore depends on the individual surgeon’s experience and preferences. Clinical evaluation of knee laxity at follow-up has low levels of intra- and inter-observer reliability (Liow et al. 2000), and the clinical evaluation might also be biased by patient complaints. Methods for instrumented laxity measurements are available but not incorporated in clinical practice on a large scale, and a gold standard on instrumented laxity measure-

Acta Orthopaedica 2019; 90 (1): 46–52

ment following TKA surgery has not yet been established. The range of soft-tissue laxity that favors both pain relief and optimal knee function following TKA remains unclear. The objective of this systematic review is to clarify evidence regarding the relationship between objectively quantifiable soft tissue laxity at follow-up and outcome scores in primary TKA.

Records identified through database searching n = 4,685

Additional records identified through other sources n = 12

Records screened after duplicates removed n = 3,228 Records excluded n = 3,193 Full-text articles assessed for eligibility n = 34

Methods We searched the PubMed, Embase, and Cochrane databases for combinations of search words describing knee arthroplasty, soft tissue laxity, and outcome to identify papers reporting on the relationship between knee laxity and outcome following primary TKA. Only studies reporting association of instrumented laxity measurement following primary TKA and outcome scores were included. Studies comparing laxity of specific implants were included only if laxity measurements were analyzed with respect to outcome scores. Studies reporting manual, non-instrumented laxity measurements were not included. Range of motion (ROM) was not considered an outcome score. For the full search history, see Supplementary data. The search was performed in June 2017 and updated in January 2018. Additional papers were added based on references; for all the included studies cross-references were identified and reviewed using Web of Science. AK conducted the primary review of the search results and the identification of full-text articles for assessment. All authors participated in the assessment of full-text articles and a minimum of 2 authors would agree on inclusion of studies in the analysis; disagreement was solved by consensus. The corresponding author independently completed data extraction according to the study protocol and data were verified by a second reviewer; disagreement was solved by consensus. Quality of the included studies was assessed using the Methodological Index for Non Randomized Studies (MINORS). Registration, funding, and potential conflicts of interest The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was followed when performing this review. The review protocol was published on the PROSPERO database in July 2017, with registration number CRD42017069779 (https://www.crd.york.ac.uk/ PROSPERO/). The protocol was updated in April 2018. The authors’ institutions funded the study. No conflict of interest declared.

Full-text articles excluded (n = 20): – descriptive, no analysis of correlation, 5 – comparison of techniques, no analysis of correlation, 4 – comparison of implants, no analysis of correlation, 8 – manual laxity measurements, 1 – only intraoperative laxity measurement, 1 – study design, 1 Studies included in qualitative synthesis n = 14

Flowchart demonstrating the PRISMA technique used to evaluate the studies

Results After removal of duplicates 3,228 studies were screened based on title, abstract, and in some cases full text. 34 full-text articles were assessed for eligibility and 14 articles fulfilled the criteria (Yamakado et al. 2003, Kuster et al. 2004, Ishii et al. 2005, Jones et al. 2006, Seon et al. 2007, Van Hal et al. 2007, Seon et al. 2010, Schuster et al. 2011, Seah et al. 2012, Nakahara et al. 2015, Oh et al. 2015, Graff et al. 2016, Tsukiyama et al. 2017, Matsumoto et al. 2017) (Figure). All eligible studies were cohort studies and methodologically quite similar with follow-up examination on a cohort of uncomplicated and non-revised primary total knee arthroplasties with measurement of laxity and outcome. Level of evidence according to “The Oxford 2011 Levels of Evidence” (http:// www.cebm.net/index.aspx?o=5653) was level III or below. The studies obtained MINORS scores from 9 to 13; the maximum score is 16. Most commonly low scores were obtained for criteriona 2 (Inclusion of consecutive patients), criterion 5 (Unbiased assessment) and criterion 8 (Prospective calculation of the study size) (Table 1, see Supplementary data). The number of patients/knees ranged from 15/21 to 112/127. Mean time from surgery to follow-up was 1 year to 7 years and the mean age of the patients was 68 years to 76 years. Sex distribution varied from 43% to 95% women. Regarding surgical technique, gap-balancing dominated but 5 studies did not report details regarding technique. Intended mechanical alignment was specified in only 3 studies, all aiming for neutral mechanical alignment. All combinations of posterior cruciate

Acta Orthopaedica 2019; 90 (1): 46–52

Table 2. Baseline characteristics of the included studies Author

Number of Follow-up Mean age Women Surgical Navigation Constrained knees/patient years (range) years (range) % technique a % articulation b (%)

Matsumoto Tsukiyama Graff Nakahara Oh Seah Schuster Seon Seon van Hal Jones Ishii Kuster Yamakado a GB b CR

110/81 50/41 24/24 94/68 61/61 100/100 127/112 55/55 42/42 51/49 97/88 77/71 44/22 21/15

4.4 (1.1–11.5) 4.8 (2.0–13.8) 2.3 (1.0–4.8) 4.6 (1.1–11.0) 2.2 (1.0–5.0) 2 (–) 3.9 (0.8–5.0) 2.8 (2.0–4.3) 1 (–) 4.6 (4.1– 5.4) 7 (5.4–9.9) 6.4 (5.2–9.4) 4.5 (2–7) 7.1 (4–8)

76 (26–91) 73 (59–82) 69 (54–80) 73 (50–86) 68 (59–82) 67 (50–83) 71 (50–89) 68 (55–81) – 73 (59–87) 70 (–) 77 (–) 69 (32–82) 68 (58–78)

85 76 46 85 79 68 71 84 95 76 43 86 55 80

GB – – MR GB GB GB GB GB GB – – – –

18 – 42 26 0 0 0 100 100 0 0 0 0 0

PS-FB = 23, PS-MB = 77 PS-FB = 100 CR-FB = 100 PS-FB = 100 CR-FB = 100 CR-FB = 100 CR-FB = 75, CR-MB = 25 CR-MB = 100 CR-MB = 100 CR-FB = 100 CR-FB = 100 CR = 69, PS = 31 FB = 16, MB = 84 CR-FB = 100

= gap balancing, MR = measured resection, – = not specified = posterior cruciate retaining, PS = posterior cruciate sacrificing, FB = fixed bearing, MB = mobile bearing.

retaining, posterior cruciate substituting, fixed bearing, and mobile bearing were used in the studies (Table 2). Laxity measurements were performed in both the sagittal and the coronal plane, and in angulations from full extension to 90° of flexion; 7 studies reported on more than 1 condition for the measurements. Statistically significant results were not found for the 12 measurements obtained with a flexion angle between 0° and 30°, but for the 10 measurementd performed between 60° and 90°, 6 showed significant results. None of the included studies measured laxity in the range between 30° and 60° (Table 3). Statistical analysis was carried out either by calculating a correlation coefficient between laxity and outcome or by stratification upon laxity followed by group comparison. The correlation coefficient was calculated in 9 studies but found to be significant in only 1. Stratification was used in 7 studies, and significant results obtained in 5. The 2 studies that did not find statistically significant correlation obtained significant results following stratification (Table 3). Sagittal laxity measurements were done using an arthrometer in 8 studies (KT-1000 in 3 studies, KT-2000 in 2 studies [Genourob, Laval, France], Rolimeter in 2 studies [Aircast, Summit, NJ, USA], and KS measure arthrometer in 1 study [Sigmax Medical, Tokyo, Japan]) and stress radiography with the Telos device in 2 studies [Telos Arzt- und Krankenhausbedarf GmbH, Hungen, Germany]. The method resembles the drawer test and the result was measured as a distance in mm. Sagittal laxity measurements performed in the range from 60° to 90° of flexion were found to associate with outcome in 4 of 7 studies. Statistically significant correlation was found in the study by Matsumoto et al. (2017) who found correlation between laxity at 60° and 1 KOOS sub-score, i.e., KOOS pain; no correlation to laxity at 90° was found. 4 studies analyzed the results following stratification. Seon et al. (2010)

measured laxity using stress radiographs at 90° and found that stable knees with laxity below 10 mm obtained better WOMAC scores. Seah et al. (2012) and Jones et al. (2006) performed the laxity measurements under equal conditions, with KT-1000 at 75–80°, and used the same stratification of the results, and both studies reported statistically significantly better outcomes for the group with laxity in the range from 5 to 10 mm. Both studies included only CR implants. Schuster et al. (2011) performed the measurements with the Rolimeter at 90° of flexion and used different limits for stratification, but did not find any significant association. Coronal laxity was quantified using stress radiography, where the knee is opened in the coronal plane by applying manual pressure or a specified force to the medial or lateral side of the knee and the opening angle between the femoral and tibial components was measured from the radiographs. The Telos device was used in 3 studies, manual stress in 2 studies and a spring scale was used to quantify stress in 2 studies. In the 2 studies reporting measurement in angulation from 80° to 90° significant results were obtained. Tsukiyama et al. (2017) and Oh et al. (2015) reported on stress radiography in flexion using the epicondylar view with coronal stress applied and measured by a spring scale. Following stratification, significant association was obtained in both studies. Tsukiyama et al. stratified the cases into a tight group with opening angle below or equal to 3° and a loose group with larger opening angle. The medial and lateral opening angle was analyzed separately, and statistically significantly better outcome scores were obtained for the medially tight group. Oh et al. used quite another stratification. Balanced knees with numerical difference between medial and lateral opening angle being equal to or below 3° obtained the best scores. Following further stratification of the balanced knees into grades of total laxity defined as the sum of medial and lateral opening angle, statistically significantly

Acta Orthopaedica 2019; 90 (1): 46–52

Table 3. Methods and results of the included studies Author

Soft-tissue laxity measurement • Significant results

Anatomical plane, degree of flexion and mean (SD) of measurement

Outcome scores

Statistical method to compare laxity and outcome score

Matsumoto

Spearman rank correlation

Tsukiyama

Stratification based on laxity Wilcoxon rank-sum test Pearson correlation coefficient

KS Measure Arthrometer, Sagittal 30°: 4.5 (2.2) mm KSS, KOOS mean of 3 measurements Sagittal 60°: 3.6 (1.9) mm Sagittal 90°: 3.0 (1.9) mm • Inverse correlation between 1 of 6 KOOS sub-scores (KOOS-pain) and laxity at 60° Stress radiographs: Coronal extension: 2011 KS Telos, 150 N in extension Varus stress: 4.0° (2.5°) epicondylar view, Valgus stress: 4.0° (2.4°) 50 N in flexion Coronal 80°: Varus stress: 6.2° (4.4°) Valgus stress: 3.9° (2.6°) • 4 of 6 2011 KS sub-scores better in knees medially tight in flexion

Graff

KT-1000, 89 N, Sagittal 20°: 3.8 (2.0) mm mean of 3 measurements • No correlation

Nakahara

Stress radiographs: Telos, 150 N • No correlation

OKS, KOOS, KSS, SF12

Coronal 10°: New KSS Varus stress: 5.9° (2.7°) Valgus stress: 5.0° (1.6°)

Stress radiographs: Coronal 90°: KSS, WOMAC epicondylar view, 50 N Varus stress: 4.7° (2.4°) Valgus stress: 4.1° (2.1°) • KSS-f and WOMAC better in balanced group. • In the balanced group KSS and WOMAC better for grade II laxity

Seah

KT-1000, 89 N, sum of Sagittal 75°: not reported anterior and posterior stress, mean of 3 measurements • Intermediate laxity group better OKS

Schuster

Rolimeter, sum of anterior and posterior stress, mean of 3 measurements • No differences between groups

Sagittal 25°: 4.6 (2.1) mm Sagittal 90°: 4.9 (2.2) mm

Pearson correlation coefficient

Pearson correlation coefficient Stratification based on laxity T-test (balanced vs. unbalanced) and Kruskal–Wallis analysis (subgroups of laxity in the balanced group)

KSS, OKS, SF-36

Stratification based on laxity One-way ANOVA

KSS, VAS Pain, VAS satisfaction

Stratification based on laxity Kruskal–Wallis analysis

Seon 2010

Stress radiographs: Sagittal 90°: 8.3 mm HSS, WOMAC Telos, 89 N, sum of anterior and posterior stress • Stable group significantly better WOMAC function

Stratification based on laxity Mann–Whitney U-test Pearson correlation coefficient

Seon 2007

Sagittal 90°: 7.1 (4.1) mm Coronal extension: Varus stress: 4.4° (2.2°) Valgus stress: 3.5° (1.4°)

m-HSS

Pearson correlation coefficient

Van Hal Rolimeter • No correlation

Sagittal 30°: 2.8 (1.1) mm

KSS

Spearman rank correlation

WOMAC, KSS, SF12

Stratification based on laxity Duncan test

HSS

Spearman rank correlation

Stress radiographs: Telos, 150 N, sagittal difference between anterior and posterior stress • No correlation

Jones

KT1000, 89 N, sum of anterior Sagittal 30°: 7.3 (4.0) mm and posterior translation, Sagittal 75–80°: 4.6 (3.1) mm mean of 3 measurements • Intermediate laxity group better KSS than the large laxity group. Ishii KT-2000, anterior force 133N, Sagittal 30°: CR: 5.8 (2.9) mm, posterior force 89N, sum of PS: 5.3 (3.2) mm anterior and posterior stress, Sagittal 75°: CR: 4.8 (2.3) mm, mean of 3 measurements PS: 3.4 (1.5) mm • No correlation Kuster

Manual stress radiographs Coronal 30°: m-HSS, Stratification based on laxity Varus stress: 4.3° (1.9°) preferred knee T-test and chi-square Valgus stress: 4.0° (2.1°) • No significance, 11 bilateral cases with a knee in each laxity group, significantly preferred laxed knee over tight knee

Yamakado

KT2000, 133N, and coronal manual stress radiographs • No correlation

Sagittal 30°: 9.1 (1.1) mm m-KSS Coronal extension: Varus stress: 6.2° (0.9°) Valgus stress: 4.3° (0.5°)

Pearson correlation coefficient and multiple regression

better scores were demonstrated for the group with total laxity from 6° to 10°. In both studies mean lateral laxity (varus stress) exceeded mean medial laxity (valgus stress) (Table 3). The most frequently used outcome score was the Knee Society Score (KSS), used in 6 studies, followed by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), used in 3 studies. 13 different outcome scores were used and most papers reported the use of more than one score (Table 3).

Discussion This systematic review deals with outcome scores and quantified measurements of soft tissue laxity following primary TKA. Any statistically significant influence of laxity in extension to limited flexion on outcome scores could not be established in the reviewed studies, but Aunan et al. (2015) found that intraoperative quantification of coronal laxity in extension correlated to KOOS when stratifying for postoperative mechanical alignment of the limb. None of the reviewed studies considered mechanical alignment in the analysis, and this result should be the subject of further research. In 6 of 9 studies measuring laxity in flexion a significant association was found, and in the studies analyzing stratified data only one study did not find significance. Hence, a correlation between outcome score and laxity in flexion must be assumed. Convincing results were found with both sagittal arthrometer measurements and coronal stress radiography. The 2 studies reporting on coronal stress radiography in flexion (Oh et al. 2015, Tsukiyama et al. 2017), found comparable mean opening angles (Table 3). In both studies the mean lateral opening angle, the result of varus stress, exceeded the mean medial opening angle, and it may cautiously be concluded that the medial opening angle should not exceed 4°. Measurement of laxity was not performed between 30° and 60° of flexion in the studies, consequently this review does not add any clarification to the discussion regarding mid-flexion instability (Vince 2016). The included studies are all cohort studies and none are methodologically errorless, which is reflected in the MINORS scores. 2 studies that did not find statistically significant results must be assumed to be under-powered as the number of knees analyzed is quite low compared with all other studies and compared with the studies showing significant results (see Table 1). Multiple testing, with more than 1 method for laxity measurement, more than 1 outcome score or sub-score, and more than 1 statistical test, which is used in some studies, introduces the risk of false positive results (see Table 3). The methods used to measure sagittal laxity have been validated in the non-arthroplasty knee (Lefevre et al. 2014) and 2 reports of validation following TKA were found (Matsuda et al. 1999, Mochizuki et al. 2017). 1 method for stress radiography to assess coronal laxity measurement following TKA is validated but was not used in the studies included in the review

Acta Orthopaedica 2019; 90 (1): 46–52

(Stähelin et al. 2003). The methods used are to the best of our knowledge not yet validated. 1 study reported validation by double measurements of 4 patients, which in this particular study equals 8 knees and 16 radiographs; the results from the double measurements are reported to lie within 1°, but the results are not described statistically (Kuster et al. 2004). Reading of angulation from coronal stress radiography is validated (Nakahara et al. 2015, Hatayama et al. 2017). The outcome measures used in the included studies do not differ from general studies regarding TKA surgery (Theodoulou et al. 2016), with the KSS representing the most frequently used measure. However, the outcome measures used may not reveal subtle differences between knees within the range of normal surgical variation, which differ only mildly in stability. The ideal outcome scores to reveal functional differences caused by variations in soft tissue laxity should allow discrimination between patients who only undertake activities of daily life and those who perform high-demand activities like sports. The KSS is known to have a high ceiling effect, and may not reveal such differences (Na et al. 2012, Jenny et al. 2014, Aunan et al. 2016). It could be argued that the outcome measures used in some studies have not been sufficiently validated, and many of these have been surpassed by more modern outcome measures that are solely patient-reported (Behrend et al. 2012, Dawson et al. 2014). Performance-based outcome measurements of TKA patients are known to reveal functional differences that are not reflected in the outcome scores (Witvrouw et al. 2002, Stevens-Lapsley et al. 2011, Bolink et al. 2015, Naili et al. 2017). The Osteoarthritis Research Society International (OARSI) recommends that performance-based outcome measures be included to complement PROMs in future osteoarthritis research (Dobson et al. 2013). This might be of special relevance in studies investigating the impact of laxity. To what extent knee laxity changes following surgery is debated. Changes in coronal laxity immediately following TKA implantation, stress relaxation, is described by Bellemans et al. (2006) who with the aid of computer navigation reported increased mediolateral laxity, by on average 1 mm. Matsumoto et al. (2012) measured intraoperative coronal laxity using an off-set tensor device and found correlation to 5-year follow-up stress radiography measurement in extension; however, in flexion correlation was found only for CR knees. The course of sagittal laxity is described by Mizu-uchi et al. (2006) and Schuster et al. (2011); both evaluated sagittal laxity continuously up to 5 years following cruciate retaining TKA, and significant changes in mean laxity were not detected. Regarding the course of outcome scores following TKA, a recent review reported no different in outcome scores between 12 months’ and 24 months’ follow-up when using KSS or WOMAC (Ramkumar et al. 2018). The mean followup period in the studies we reviewed ranged from 1 year to 7 years and seems appropriate, but the large range that is present in some of the studies might introduce bias.

Acta Orthopaedica 2019; 90 (1): 46–52

Differences in soft tissue laxity between specific implants and concepts of constraint might blur this review as conflicting results are reported regarding both the influence of mobile- versus fixed-bearing (Luring et al. 2006, Schuster et al. 2011, Matsumoto et al. 2017) and constraint where implant conformity may affect laxity (Ishii et al. 2005, Matsumoto et al. 2014, Yoshihara et al. 2016, Song et al. 2017, Wautier and Thienpont 2017). Further haze might occur due to differences in surgical technique, where gap-balancing and measured resection represent 2 different approaches to implant positioning and soft-tissue balancing, which is shown to have impact on the laxity (Lüring et al. 2009, Pang et al. 2011, Matsumoto et al. 2014, Clement et al. 2017). We found no studies investigating to what extent preoperative anatomical conditions are reflected in the postoperative measurements of laxity. However, the preoperative mechanical axis has been reported not to correlate with intraoperative measurement of laxity at the end of surgery (Aunan et al. 2015). Preoperative and intraoperative factors not accounted for, such as mechanical alignment, severity of osteoarthritis, component alignment, and component rotation, might also introduce bias. Tsukiyama et al. (2017) challenge the surgical gold standard of rectangular joint gaps in flexion and extension, as it was found that only medial coronal laxity, in opposition to lateral coronal laxity, in flexion influences outcome. This finding is in line with recommendations from some authors (Bellemans et al. 2006, Aunan et al. 2015, Risitano and Indelli 2017). The range of soft-tissue laxity that favors both pain relief and optimal knee function following total knee arthroplasty (TKA) still needs clarification. Future studies using validated instruments should address the methodological issues of the reviewed studies, and might benefit from including performance-based outcome measurements. The combined impact of mechanical alignment and laxity on outcome should be investigated. However, this systematic review confirm that surgeons should measure and adjust ligament laxity intraoperatively in order to improve outcome following TKA. Supplementary data Table 1 and full search history is available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2018.1554400

AK: Primary draft of study protocol, database search and identification of full-text articles for assessment and primary draft of work. AO, PTN, MBL, AK: Revision and approval of study protocol, assessment of full-text articles, analysis and interpretation of data, revision and approval of work. Professional and kind help from the university librarians Jette Frost Jepsen and Pernille Skou Gaardsted was highly appreciated. Acta thanks Eirik Aunan and Kirill Gromov for help with peer review of this study.

Aunan E, Kibsgård T J, Diep L M, Röhrl S M. Intraoperative ligament laxity influences functional outcome 1 year after total knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2015; 23(6): 1684-92. Aunan E, Næss G, Clarke-Jenssen J, Sandvik L, Kibsgard T J. Patellar resurfacing in total knee arthroplasty: functional outcome differs with different outcome scores. Acta Orthop 2016; 87(2): 158-64. Babazadeh S, Stoney J D, Lim K, Choong P F M. The relevance of ligament balancing in total knee arthroplasty: how important is it? A systematic review of the literature. Orthop Rev (Pavia) 2009; 1(e26): 70-8. Behrend H, Giesinger K, Giesinger J M, Kuster M S. The “forgotten joint” as the ultimate goal in joint arthroplasty: validation of a new patient-reported outcome measure. J Arthroplasty 2012; 27(3): 430-6.e1. Bellemans J, D’Hooghe P, Vandenneucker H, Van Damme G, Victor J. Soft tissue balance in total knee arthroplasty: does stress relaxation occur perioperatively? Clin Orthop Relat Res 2006; 452: 49-52. Bolink S A A N, Grimm B, Heyligers I C. Patient-reported outcome measures versus inertial performance-based outcome measures: a prospective study in patients undergoing primary total knee arthroplasty. Knee 2015; 22(6): 618-23. Clement N D, Makaram N, Bell J, Tiemessen C H, Mehdi S A, Livingston S J. Columbus® computer navigated total knee arthroplasty: gap balancing versus measured resection. Knee 2017; 24(6): 1442-7. Dawson J, Beard D J, McKibbin H, Harris K, Jenkinson C, Price A J. Development of a patient-reported outcome measure of activity and participation (the OKSAPQ) to supplement the Oxford Knee Score. Bone Joint J 2014; 96 B(3): 332-8. Dobson F, Hinman R S, Roos E M, Abbott J H, Stratford P, Davis A M, Buchbinder R, Snyder-Mackler L, Henrotin Y, Thumboo J, Hansen P, Bennell K L. OARSI recommended performance-based tests to assess physical function in people diagnosed with hip or knee osteoarthritis. Osteoarthritis Cartilage 2013; 21(8): 1042-52. Graff C, Hohmann E, Bryant A L, Tetsworth K. Subjective and objective outcome measures after total knee replacement: is there a correlation? ANZ J Surg 2016; 86(11): 921-5. Gromov K, Korchi M, Thomsen M G, Husted H, Troelsen A. What is the optimal alignment of the tibial and femoral components in knee arthroplasty? Acta Orthop 2014; 85(5): 480-7. Hatayama K, Terauchi M, Saito K, Higuchi H. Does residual varus alignment cause increasing varus laxity at a minimum of five years after total knee arthroplasty? J Arthroplasty 2017; 32(6): 1808-13. Huang N F R, Dowsey M M, Ee E, Stoney J D, Babazadeh S, Choong P F. Coronal alignment correlates with outcome after total knee arthroplasty: five-year follow-up of a randomized controlled trial. J Arthroplasty 2012; 27(9): 1737-41. Ishii Y, Matsuda Y, Ishii R, Sakata S, Omori G. Sagittal laxity in vivo after total knee arthroplasty. Arch Orthop Trauma Surg 2005; 125(4): 249-53. Jenny J Y, Louis P, Diesinger Y. High activity arthroplasty score has a lower ceiling effect than standard scores after knee arthroplasty. J Arthroplasty 2014; 29(4): 719-21. Jones D P G, Locke C, Pennington J, Theis J-C. The effect of sagittal laxity on function after posterior cruciate-retaining total knee replacement. J Arthroplasty 2006; 21(5): 719-23. Kim Y H, Park J W, Kim J S, Park S D. The relationship between the survival of total knee arthroplasty and postoperative coronal, sagittal and rotational alignment of knee prosthesis. Int Orthop 2014; 38(2): 379-85. Kuster M S, Bitschnau B, Votruba T. Influence of collateral ligament laxity on patient satisfaction after total knee arthroplasty: a comparative bilateral study. Arch Orthop Trauma Surg 2004; 124: 415-7. Lee B, Cho H, Bin S, Kim J, Jo B. Femoral component varus malposition is associated with tibial aseptic loosening after TKA. Clin Orthop Relat Res 2018; 476:400-7. Lefevre N, Bohu Y, Naouri J F, Klouche S, Herman S. Validity of GNRB® arthrometer compared to TelosTM in the assessment of partial anterior cruciate ligament tears. Knee Surgery, Sport Traumatol Arthrosc 2014; 22(2): 285-90.

Liow R Y, Walker K, Wajid M A, Bedi G, Lennox C M. The reliability of the American Knee Society Score. Acta Orthop 2000; 71(6): 603-8. Longstaff L M, Sloan K, Stamp N, Scaddan M, Beaver R. Good alignment after total knee arthroplasty leads to faster rehabilitation and better function. J Arthroplasty 2009; 24(4): 570-8. Luring C, Bathis H, Oczipka F, Trepte C, Lufen H, Perlick L, Grifka J. Twoyear follow-up on joint stability and muscular function comparing rotating versus fixed bearing TKR. Knee Surgery, Sport Traumatol Arthrosc 2006; 14(7): 605-11. Lüring C, Oczipka F, Perlick L, Tingart M, Grifka J, Bäthis H. Two year follow-up comparing computer assisted versus freehand TKR on joint stability, muscular function and patients satisfaction. Knee Surgery, Sport Traumatol Arthrosc 2009; 17(3): 228-32. Matsuda S, Miura H, Nagamine R, Urabe K, Matsunobu T, Iwamoto Y. Knee stability in posterior cruciate ligament retaining total knee arthroplasty. Clin Orthop Relat Res 1999; (366): 169-73. Matsumoto T, Muratsu H, Kubo S, Matsushita T, Kurosaka M, Kuroda R. Intraoperative soft tissue balance reflects minimum 5-year midterm outcomes in cruciate-retaining and posterior-stabilized total knee arthroplasty. J Arthroplasty 2012; 27(9): 1723-30. Matsumoto T, Muratsu H, Kawakami Y, Takayama K, Ishida K, Matsushita T, Akisue T, Nishida K, Kuroda R, Kurosaka M. Soft-tissue balancing in total knee arthroplasty: cruciate-retaining versus posterior-stabilised, and measured-resection versus gap technique. Int Orthop 2014; 38(3): 531-7. Matsumoto K, Ogawa H, Yoshioka H, Akiyama H. Postoperative anteroposterior laxity influences subjective outcome after total knee arthroplasty. J Arthroplasty 2017; 32(6): 1845-9. Mihalko W M, Saleh K J, Krackow K A, Whiteside L A. Soft-tissue balancing during total knee arthroplasty in the varus knee. J Am Acad Orthop Surg 2009; 17(12): 766-74. Mizu-uchi H, Matsuda S, Miura H, Nabeyama R, Okazaki K, Iwamoto Y. Anteroposterior stability in posterior cruciate ligament-retaining total knee arthroplasty. J Arthroplasty 2006; 21(4): 592-8. Mochizuki T, Tanifuji O, Sato T, Hijikata H, Koga H, Watanabe S, Higano Y, Ariumi A, Murayama T, Yamagiwa H, Endo N. Association between anteroposterior laxity in mid-range flexion and subjective healing of instability after total knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2017; 25(11): 3543-8. Na S E, Ha C W, Lee C H. A new high-flexion knee scoring system to eliminate the ceiling effect. Clin Orthop Relat Res 2012; 470(2): 584-93. Naili J E, Iversen M D, Esbjörnsson A-C, Hedström M, Schwartz M H, Häger C K, Broström E W. Deficits in functional performance and gait one year after total knee arthroplasty despite improved self-reported function. Knee Surgery, Sport Traumatol Arthrosc 2017; 25(11): 3378-3386. Nakahara H, Okazaki K, Hamai S, Okamoto S, Kuwashima U, Higaki H, Iwamoto Y. Does knee stability in the coronal plane in extension affect function and outcome after total knee arthroplasty? Knee Surgery, Sport Traumatol Arthrosc 2015; 23(6): 1693-8. Nodzo S R, Franceschini V, Gonzalez Della Valle A. Intraoperative loadsensing variability during cemented, posterior-stabilized total knee arthroplasty. J Arthroplasty 2017; 32(1): 66-70. Oh C-S, Song E-K, Seon J K, Ahn Y S. The effect of flexion balance on functional outcomes in cruciate-retaining total knee arthroplasty. Arch Orthop Trauma Surg 2015; (135): 401-6. Pang H N, Yeo S J, Chong H C, Chin P L, Ong J, Lo N N. Computer-assisted gap balancing technique improves outcome in total knee arthroplasty, compared with conventional measured resection technique. Knee Surgery, Sport Traumatol Arthrosc 2011; 19(9): 1496-503.

Acta Orthopaedica 2019; 90 (1): 46–52

Ramkumar P N, Navarro S M, Haeberle H S, Ng M, Piuzzi N S, Spindler K P. No difference in outcomes 12 and 24 months after lower extremity total joint arthroplasty: a systematic review and meta-analysis. J Arthroplasty 2018; 33(7): 2322-9. Risitano S, Indelli P F. Is “symmetric” gap balancing still the gold standard in primary total knee arthroplasty? Ann Transl Med 2017; 5(16): 325. Ritter M A, Davis K E, Meding J B, Pierson J L, Berend M E, Malinzak R A. The effect of alignment and bmi on failure of total knee replacement. J Bone Joint Surg Am 2011; 93(17): 1588-96. Schuster A J, von Roll A L, Pfluger D, Wyss T. Anteroposterior stability after posterior cruciate-retaining total knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2011; 19(7): 1113-20. Seah R B, Pang H N, Lo N N, Chong H C, Chin P L, Chia S L, Yeo S J. Evaluation of the relationship between anteroposterior translation of a posterior cruciate ligament-retaining total knee replacement and functional outcome. J Bone Joint Surg Br 2012; 94-B(10): 1362-5. Seon J K, Song E K, Yoon T R, Bae B H, Park S J, Cho S G. In vivo stability of total knee arthroplasty using a navigation system. Int Orthop 2007; 31(1): 45-8. Seon J K, Park S J, Yoon T R, Lee K B, Moon E S, Song E K. The effect of anteroposterior laxity on the range of movement and knee function following a cruciate-retaining total knee replacement. J Bone Joint Surg Br 2010; 92(8): 1090-5. Song E K, Lim H A, Joo S D, Kim S K, Lee K B, Seon J K. Total knee arthroplasty using ultra-congruent inserts can provide similar stability and function compared with cruciate-retaining total knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2017; 25(11): 3530-5. Stevens-Lapsley J E, Schenkman M L, Dayton M R. Comparison of selfreported knee injury and osteoarthritis outcome score to performance measures in patients after total knee arthroplasty. PM R 2011; 3(6): 541-9. Stähelin T, Kessler O, Pfirrmann C, Jacob H A C, Romero J. Fluoroscopically assisted stress radiography for varus–valgus stability assessment in flexion after total knee arthroplasty. J Arthroplasty 2003; 18(4): 513-5. Theodoulou A, Bramwell D, Spiteri A, Kim S W, Krishnan J. The use of scoring systems in knee arthroplasty: a systematic review of the literature. J Arthroplasty 2016; 31(10): 2364-2370.e8. Tsukiyama H, Kuriyama S, Kobayashi M, Nakamura S, Furu M, Ito H, Matsuda S. Medial rather than lateral knee instability correlates with inferior patient satisfaction and knee function after total knee arthroplasty. Knee 2017; 24(6): 1478-84. Van Hal C T H, Van Hellemondt G G, Wymenga A B, Jacobs W C H. The anterior–posterior laxity after total knee arthroplasty inserted with a ligament tensor. Knee Surgery, Sport Traumatol Arthrosc 2007; 15(8): 1019-22. Vince K. Mid-flexion instability after total knee arthroplasty: woolly thinking or a real concern? Bone Joint J 2016; 98-B(1): 84-8. Wautier D, Thienpont E. Changes in anteroposterior stability and proprioception after different types of knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2017; 25(6): 1792-800. Witvrouw E, Victor J, Bellemans J, Rock B, Van Lummel R, Van Der Slikke R, Verdonk R. A correlation study of objective functionality and WOMAC in total knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2002; 10(6): 347-51. Yamakado K, Kitaoka K, Yamada H, Hashiba K, Nakamura R, Tomita K. Influence of stability on range of motion after cruciate-retaining TKA. Arch Orthop Trauma Surg 2003; 123(1): 1-4. Yoshihara Y, Arai Y, Nakagawa S, Inoue H, Ueshima K, Fujiwara H, Oda R, Taniguchi D, Kubo T. Assessing coronal laxity in extension and flexion at a minimum of 10 years after primary total knee arthroplasty. Knee Surgery, Sport Traumatol Arthrosc 2016; 24(8): 2512-6.

Acta Orthopaedica 2019; 90 (1): 53–59

Nonagenarians qualify for total knee arthroplasty: a report on 329 patients from the Swedish Knee Arthroplasty Register 2000–2016 Erdem A SEZGIN 1, Otto ROBERTSSON 2,3, Annette W-DAHL 2,3, and Lars LIDGREN 2,3 1 Gazi

University, Faculty of Medicine, Department of Orthopedics and Traumatology, Ankara, Turkey; 2 Lund University, Faculty of Medicine, Department of Clinical Sciences Lund, Orthopedics, Lund, Sweden; 3 The Swedish Knee Arthroplasty Register, Lund, Sweden Correspondence: lars.lidgren@med.lu.se Submitted 2018-06-10. Accepted 2018-09-05.

Background and purpose — The nonagenarian (those aged 90 years and older) population is expected to double in the next 20 years. This demographic age quake may have a significant impact on the incidence of total knee arthroplasty (TKA), although current literature provides limited data. We examined death and revision rates, patient-reported outcomes (PROs) and bias on patient selection of nonagenarian patients operated on with TKA for osteoarthritis (OA) between 2000 and 2016. Patients and methods — The Swedish national knee arthroplasty register was used to identify 329 nonagenarians (mean age, 92 years). Each patient was followed-up until death or the end of 2017. PRO data of 22 of these patients were compared with 65- to 74-year-old patients operated in 2015, from the same register. Results — 5 patients (1.5%) died within 90 days and 23 (7%) patients died within 365 days after TKA. 8 patients (2.4%) developed knee complications that needed revision. For patients followed for 5 and 10 years, more than 50% and 10%, respectively, lived without being revised. The patients had statistically significant improvements in PROs, not significantly different from the younger SKAR cohort. However, the material is small and this statistical finding does not preclude that there may be clinically relevant differences. TKA incidence was different amongst the 21 counties in the country (range, 0–5.1/10,000). Interpretation — Our study suggests that nonagenarians with knee OA qualify for TKA, having similar outcomes to younger patients. The data presented may help surgeons and patients assessing the risks and outcome associated with the procedure.

The nonagenarian population in Sweden was 95,300 (0.94%) in 2017 and is expected to rise to 199,900 (1.7%) in 2040 (Statistics Sweden). This demographic age quake has a substantial impact on the prevalence of symptomatic osteoarthritis (OA) (Nemes et al. 2015). Amongst OA of weight-bearing joints, knee OA has the most pronounced correlation with age. It affects one-fourth of the reported 56- to 84-year-old individuals in a population study from southern Sweden (Turkiewicz et al. 2014). The demand for primary total knee arthroplasty (TKA) is on the rise all round the world and this trend is expected to continue in the future (Koh et al. 2013, Culliford et al. 2015, Kurtz et al. 2014, Nemes et al. 2015, Inacio et al. 2017, Niemelainen et al. 2017). In studies reporting the outcome of the oldest knee arthroplasty patients, the cut-off has often been 85 years and older (Laskin 1999, Biau et al. 2006, Easterlin et al. 2013, Williams et al. 2013). With a focus on nonagenarians, we have found only 4 studies, all with low patient numbers (12–42) which include data from the 1970s until early 2000s (Belmar et al. 1999, Joshi and Gill 2002, Pagano et al. 2004, Karrupiah et al. 2008) (Table 1). In this study we evaluated the nonagenarians in Sweden operated on with TKA for OA between 2000 and 2016 on 5 aspects: (1) the demographics, (2) early/total death rates (3) revision rate and reasons for revision, (4) patient-reported outcome (PRO), (5) difference in TKA incidence among the counties in Sweden.

Patients and methods Patients aged 90 years and older who had undergone primary TKA for OA between January 1, 2000 to December 31, 2016 were identified in the Swedish Knee Arthroplasty Register (SKAR).

Acta Orthopaedica 2019; 90 (1): 53–59

Table 1. Summary of demographics and outcome data of current study and relevant literature Study characteristics

Present study Karrupiah et al. Pagano et al. Joshi and Gill

Belmar et al.

Year of publication 2008 2004 2002 1999 Timespan 2000–2016 1990–2006 1970–1997 1976–1999 1983–1997 Country of data Sweden UK USA USA USA Demographics Number of patients (knees) 329 (359) 42 (42) 34 (44) 18 (20) 12 (15) Age mean (range) 92 (90–101) 90 (90–91) 92 (90–102) 91(90–93) 92 (90–96) Female sex (%) 69 N/A 68 72 75 Months of follow-up (range) 50 (0.2–145) 90 48 62 (1–152) N/A (6–30) Diagnosis of OA (%) 100 100 91 100 80 Outcomes Death rate 90 days, n (%) 5 (1.5) 0 (45 days) 1 (3) 1 (5.6) 0 Death rate 1st year, n (%) 23 (7) 6 (14.3) N/A 2 (11.1) N/A Mean survival after TKA First revision, n (%) 8 (2.4) N/A 0 1 (5.6) 0 Mean length of stay (days) 6.2 11 13.5 10.1 15 PROs KOOS KSS Pain KSS Pain KSS Pain KSS Pain Pain 47–82 25–81 30–86 45–95 2–49 ADL 51–76 QoL 28–75 EQ-VAS WOMAC KSS Function KSS Function KSS Function 71–76 62–41 29–38 28–53 26–33 VAS satisfaction Satisfaction 21/22 33/34 ADL= activities of daily life function, QoL = quality of life, N/A = not available.

Initiated in 1975 by the Swedish Orthopedic Association, the SKAR is the world’s first national arthroplasty register. The SKAR has high completeness and correctness of data (SKAR 2017). Patient demographics including age at surgery, date of surgery, sex, and county of residence were identified from the SKAR as well as data for outcome measures including current age or date of death, revisions, and reasons for these procedures. In the SKAR, revisions are defined as a new operation in a previously resurfaced knee in which 1 or more of the components are exchanged, removed, or added (including arthrodesis and amputation) (SKAR 2017). All the identified patients were followed-up until death, or until December 31, 2017. The proportion of death within 90 days and 1 year after the TKA was assessed. Data from the Swedish life tables was used to compare 1-year death rates and cumulative death rates of the TKA patients included with the national average of the same age group in 2016 (Human Mortality Database). Revisions, reasons for, and time of revisions and consequent operations (if these occurred) were compiled and revision rate was calculated. We identified PROs in patients 90 aged years and older operated on with TKA for OA from the SKAR PROM database in addition to the aforementioned data. We included patients operated between 2009 and 2015, but excluded the second knee if both knees had an arthroplasty within 1 year and the left knee if simultaneous surgery was performed (1 patient). The PROs of this sub-cohort of patients were then compared with those of the 65- to 74-year-old primary TKA

patients operated on for OA during 2015 (SKAR 2017). The SKAR PROM project started in 2009 with patients undergoing knee arthroplasty in southern Sweden. Since then, hospitals from other regions have entered the PROM project. In 2015 approximately 28% of all TKAs for OA were included in the project that had both preoperative and 1-year postoperative PRO data and the response rate was 72% in 2015 for the 65to 74-year-old cohort (SKAR 2017). As PROs, we used the Knee injury and Osteoarthritis Outcome Score (KOOS), the EuroQol-visual analogue scale (EQ-VAS) and the visual analogue scale for patient satisfaction (VAS–satisfaction) (Roos et al. 1998, EuroQol Group 2015). Considering the difference in outcome between the groups, ≥ 8 points in the KOOS and ≥ 15 mm in EQ-VAS was considered a clinically relevant difference for statistically significant results. In addition, the KOOS was converted to the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) to be able to classify each patient as Osteoarthritis Research Society International– Outcome Measures in Rheumatology (OMERACT–OARSI) responder or not at 1 year based on a combination of absolute and relative changes in WOMAC pain, function, and total scores (Pham et al. 2004). The outcome at 1 year was dichotomized into ‘responders’ and ‘non-responders’ according to these criteria (Pham et al. 2004). Patient satisfaction with the arthroplasty surgery 1 year postoperatively was evaluated using a 0–100 scale (VAS) in which 0 indicates the highest imag¬inable satisfaction and 100 the worst imaginable satisfaction. The satisfaction (VAS) score was categorized into 5 groups: very satisfied (0–20), satisfied (21–40), moderately

Acta Orthopaedica 2019; 90 (1): 53–59

SKAR 2000–2016 Knee arthroplasty in patients aged ≥ 90 years n = 408 knees (377 patients) Excluded Not TKA and/ not OA n = 49 knees (48 patients) TKA for OA 2000–2016 in patients aged ≥ 90 years n = 359 knees (329 patients) TKA for OA 2009–2015 in patients aged ≥ 90 years in PROM cohort n = 42 knees (41 patients)

SKAR 2015, TKA for OA in patients aged 65–74 years in PROM cohort n = 1,431 patients

Excluded (n = 20 knees): – bilateral surgery during follow-up, 1 knee – dead, 8 patients – missing PROMs, 11 patients Available for analysis n = 22 knees/patients

Excluded Missing PROMs n = 396 patients

Available for analysis n = 1,035 patients

Figure 1. Flow diagram on the selection of patients.

satisfied (41–60), unsatisfied (61–80), and very unsatisfied (81–100). In routine practice, PROs are collected 2 to 6 weeks prior to surgery at the outpatient visit. 1 year postoperatively, the same questionnaire is mailed to the patients together with the VAS–satisfaction questionnaire. Data for the length of stay (LoS) were only accessible for the region of Skane. Considering that LoS changed over the years, the data from only the last 10 years (January 1, 2007– December 31, 2016) were used for analysis to better reflect modern trends. LoS data were then compared between nonagenarians and the rest of the TKA population from the same period. To compare patient selection bias between counties in Sweden, we used population data and number of TKAs for OA in each county to calculate incidence ratios. Statistics Life-table or actuarial analysis was used to access the cumulative death rate. Student’s 2-sample t-test was used to compare the mean age at surgery between sexes for nonagenarians. Welch’s t-test was used for comparisons of the KOOS and the EQ-VAS between the 2 age cohorts considering unequal variances and unequal sample sizes with assumption of normal distribution. The normality of the PRO data was controlled using the 2-sample Kolmogorov–Smirnov test for equality of distribution functions and the distribution was found to be normally distributed for all PRO variables except for the preoperative KOOS symptoms. For analysis of proportions of OMERACT–OARSI responder and satisfaction with the surgery, a chi-square test was used for comparisons. The cumula-

Figure 2. Death rate within 365 days after primary surgery.

Figure 3. Cumulative death rate.

tive revision rate was calculated using the STATA Life-table method using one-month intervals. Statistical analyses were carried out using STATA (Stata Statistical Software: Release 15; StataCorp LLC, College Station, TX, USA). Ethics, funding, and potential conflicts of interest The data gathering from the Swedish Knee Arthroplasty Register was approved by the Ethics Board of Lund University (LU20-02). The authors received no specific funding for this work. No conﬂicts of interest were declared.

Results Among 377 patients (408 knees) aged 90 years and older who had undergone knee arthroplasty, 329 patients with 359 primary TKAs inserted for OA were included in our study (Figure 1). 31 patients had bilateral TKAs. 227 (69%) patients were women; the mean age at surgery (first knee in the case of bilateral knees) was 92 years (90–101) and this was similar between women and men. The mean length of follow-up was 50 months (0.2–145). 5 patients (1.5%) died in the first 90 days postoperatively and 23 patients (7%) died within 365 days after the TKA surgery (Figure 2). Up to December 31, 2017, 214 patients had died, leaving 115 alive. According to the cumulative death rate calculations, more than 50% of the patients had lived more than 5 years with their TKA and more than 10% had lived more than 10 years after the primary TKA (Figure 3). 2 patients who had bilateral TKAs had the implants inserted simultaneously. One was 92 years and lived for 1 year and 8 months. The other was 95 years and lived for 4 years and 11 months. The oldest patient operated was 101 years old when he received his second TKA (the first being replaced 12 years earlier). He lived for another 1 year and 4 months. None of the arthroplasties in these 3 patients were revised. 8 patients (2.4%) developed complications that needed revision. The revisions were mostly early (from 1 week to 5

Acta Orthopaedica 2019; 90 (1): 53–59

Table 2. The 8 revisions and following surgeries after the first revision Time after operation Indication First revision 1 2 3 4 5 6

7 8

Consequent problems following revisions— time after first revision

Months until death

1 month Knee luxation with TKA Arthrodesis after 2 weeks 125 rupture of quadriceps 1 month Infection Exchange plastic insert, DAIR – 46 5 months Patella fracture Osteosynthesis, exchange 1. Skin necrosis—wound 98 tibial component and plastic revision after 2 weeks insert 2. Amputation after 5 weeks 1 month Patellar tendon rupture Exchange plastic insert, DAIR – (Alive at 54) and suspected infection 3 weeks Infection Exchange plasticinsert, DAIR – (Alive at 58) 3 weeks Infection Exchange plasticinsert, DAIR 1. Luxation of femorotibial 46 components—exchange insert after 2 months 2. Fracture—osteosynthesis after 2.5 months 1 week Infection Exchange plastic insert, DAIR – 33 3 months Patella luxation Addition of patellar component – 46

DAIR: debridement, antibiotics, irrigation and retention.

Table 3. Patient-reported outcome preoperatively and 1-year postoperatively in nonagenarians and the SKAR PROM cohort of 65- to 74-year-olds operated with TKA for OA in 2015

Preoperatively Nonagenarian 65–74 year n = 22 n = 1,035 p-value Difference

KOOS–Pain, mean (CI) 47 (38–56) 41 (40–42) 0.2 6 (–3 to 15) KOOS–Symptoms, mean (CI) 61 (51–71) 47 (46–48) <0.01 14 (4 to 24) KOOS–ADL, mean (CI) 51 (41–60) 46 (45–47) 0.4 4 (–5 to 14) KOOS-Sports/Rec, mean (CI) 17 (6–27) 13 (12–14) 0.5 4 (–7 to 15) KOOS– QoL, mean (CI) 28 (19–36) 23 (22–24) 0.3 5 (–4 to 13) EQ-VAS, mean (CI) 71 (63–79) 67 (66–69) 0.3 4 (–4 to 12) Satisfaction, n (%) OMERACT-OARSI responder, n (%)

1-year postoperatively Nonagenarian 65–74 year n = 22 n = 1,035 p-value Difference 82 (71–92) 87 (82–93) 76 (67–85) 35 (21–47) 75 (64–85) 76 (68–85) 21 17

82 (81–83) 79 (78–80) 80 (79–82) 42 (41–44) 67 (66–69) 78 (76–79) 878 (84.8) 925 (89.4)

0.9 –1 (–11 to 10) 0.005 9 (3 to 14) 0.3 –5 (–14 to 4) 0.3 –7 (–22 to 7) 0.2 –8 (–3 to 18) 0.7 2 (–10 to 7) 0.5 0.07

CI = 95% confidence interval, KOOS = Knee injury and Osteoarthritis Outcome Score, ADL = activity of daily life function, Sport/Rec = Sport and recreation function, QoL = quality of life, VAS = visual analogue scale, OMERACT–OARSI = Osteoarthritis Research Society International–Outcome Measures in Rheumatology.

months). The cause was infection/suspected infection in 5/8 revisions and they were treated with debridement, antibiotics, irrigation, retention (DAIR), and exchange of the insert. 3 patients required additional surgery, including 1 arthrodesis and 1 amputation. None of the patients died due to their revision surgery (Table 2). Both preoperative and 1-year postoperative PRO data were available for 22 of the nonagenarian patients (Figure 1). When PRO scores of the sub-cohort were compared with those of the 65- to 74-year-old primary TKA patients operated on for OA during 2015, the results were similar preoperatively and 1-year postoperatively in all PROs, except for the KOOS symptoms subscale. On the KOOS symptoms subscale, the nonagenarian group reported statistically and clinically significantly fewer symptoms, both preoperatively and 1-year postoperatively,

than the younger group (Table 3). The proportion of OMERACT–OARSI responders was 89% in the younger group and three-fourths of the nonagenarians (17/22 patients) were classified as responders. Furthermore, 85% of the younger group and 21/22 of nonagenarian patients were very satisfied or satisfied with the surgery (Table 3). The incidence of primary TKA for OA in nonagenarians was different amongst counties (range, 0–5.1/10,000 inhabitants) (Table 4 and Figure 4). The mean LoS for nonagenarians was somewhat longer (6.2, CI 5.1–7.4) than for the rest of the TKA for OA population (4.1, CI 4.1–4.2) in the region of Skane between January 1, 2007, and December 31, 2016. In addition, there was a noticeable change in the mean LoS for the TKA population during these 10 years (6.2 to 2.8 days).

Acta Orthopaedica 2019; 90 (1): 53–59

Table 4. Incidence ratio calculated for each county, using the number of TKAs and inhabitants in each county

4.0

1.2

2.7

1.6

2.6

2.4 4.5 5.1

0.8 2.4

1.8

3.4

2.0 4.0

2.9 3.3

1.6

0 2.4

0.8

Nonagenarian Nonagenarian TKA TKAs TKAs/ Nonagenarian incidence/ County 2000–1016 year inhabitants 10,000 Stockholm 89 5.2 15,286 Uppsala 19 1.1 2,490 Sörmland 8 0.5 2,404 Östergötland 26 1.5 3,784 Jönköping 16 1 3,244 Kronoberg 0 0 1,888 Kalmar 7 0.4 2,502 Gotland 3 0.2 568 Blekinge 2 0.1 1,489 Skåne 45 2.7 11,119 Halland 15 0.9 2,696 Västra Götaland 42 2.5 13,617 Värmland 23 1.4 2,641 Örebro 11 0.7 2,655 Västmanland 3 0.2 2,250 Dalarna 11 0.6 2,698 Gävleborg 11 0.4 2,524 Västernorrland 6 0.4 2,157 Jämtland 6 0.4 1,302 Västerbotten 4 0.2 1,907 Norrbotten 12 0.7 1,783 Total 359 21

3.4 4.5 2.0 4.0 2.9 0 1.6 3.1 0.8 2.4 3.3 1.8 5.1 2.4 0.8 2.4 2.6 1.6 2.7 1.2 4.0

3.1

Figure 4. Incidence ratio calculated for each county, using the number of TKAs and inhabitants in each county.

Discussion Improvements in general health, medical services, and social welfare have caused a substantial increase in the nonagenarian population. This will undoubtedly result in a higher number of patients with advanced knee OA. Our results on death rates, revision rates, and PROs suggest that nonagenarians with knee OA who qualify for TKA have good outcomes as compared with the other age groups having TKA. However, the material is small and there may be differences. We found varying incidences (0–5/100,000) of TKA in nonagenarians in different counties in Sweden, which indicates a bias in patient selection. Our findings concerning death rates, revision rates, and PROs are mostly in line with others (Belmar et al. 1999, Joshi and Gill 2002, Pagano et al. 2004, Karrupiah et al. 2008) (Table 1). Complications are one of the key factors to be considered when making the decision for TKA in nonagenarian patients. Perioperative complication rates after TKA have been reported as being as high as 11 of 15 TKAs and as low as 5 of 20 TKAs in the nonagenarian population (Belmar et al. 1999, Joshi and Gill 2002). D’Apuzzo et al. (2014) identi-

fied more than 58,000 nonagenarians in the USA Nationwide Inpatient Sample (NIS) who had had TKA or THA surgery for various diagnoses and found that age affected the rate of early medical complications after total joint arthroplasty. Although complication rates seem to be higher in nonagenarians than in younger patients undergoing TKA, it is important to point out that almost all of these complications were reported to be typically transient, being resolved in the early postoperative days without long-term sequelae. It should also be kept in mind that the definition of complications, follow-up duration (days versus years), and inclusion of historic data varies substantially amongst studies, which results in various rates of reported complications (Belmar et al. 1999, Joshi and Gill 2002, Pagano et al. 2004, Alfonso et al. 2007, Karrupiah et al. 2008, Petruccelli et al. 2012, Kennedy et al. 2013, D’Apuzzo et al. 2014, Kuperman et al. 2016). Death rate may be considered as an indicator of safety; however, its use in elective surgery such as TKA in very old patients is limited as these patients naturally have a shorter life expectancy, irrespective of the surgery. Average life expectancy of a 90-year-old Swedish person was 4.2 years in 2016 (Human Mortality Database). In this case, the long-term death rate would be less important compared with patients’ ability to be active and independent in their remaining years. Because of this, early and 1-year postoperative death rates are better indicators for safety. That being said, nonagenarians who are offered or demand a primary TKA are more likely to have better health than their counterparts. As previously shown

(Joshi and Gill 2002, Alfonso et al. 2007), it is no surprise that the average life expectancy of nonagenarians who had TKA surpasses the national average of their counterparts. Our cohort with a mean age of 92 years had lower 1-year death rate than the average 92-year-old individual in Sweden in 2016 (7% and 20%) (Human Mortality Database). Another important point to note is that cumulative death rates of the patients in our study at both 5 years and 10 years after the operation were noticeably lower than their 92-year-old counterparts (43% and 73%, 85% and 97% respectively) (Figure 3). This shows that more than half of the nonagenarian patients can expect to live more than 5 years with their implant. Our findings on death rates were in line with the literature (Belmar et al. 1999, Joshi and Gill 2002, Pagano et al. 2004, Karrupiah et al. 2008) (Table 1). Similar death rates were found in reports studying both TKA and THA in nonagenarians (Alfonso et al. 2007, Petruccelli et al. 2012). Petruccelli et al. (2012) showed in a literature review consisting of 105 total arthroplasties in nonagenarians that the perioperative death rate was 2%, and 1-year postoperative death rate was 8%, while D’Apuzzo et al. (2014) showed that the US cohort of 58,000 nonagenarians operated on with total arthroplasty had a substantially higher in-hospital death rate (3%) than younger patients (45–89 years, 0.2%) even after adjusting for comorbidities. The higher rate may be the result of their material including several indications for arthroplasty such as hip fractures and rheumatoid arthritis. Revision rates for primary TKA in the nonagenarians have not been thoroughly analyzed in the literature, probably due to a limited number of patients included (Belmar et al. 1999, Joshi and Gill 2002, Pagano et al. 2004, Karrupiah et al. 2008) (Table 1). In our study, 8 of 329 (2.4%) patients had revision surgery, all within the first 5 months after the primary TKA. Their first revisions probably did not affect these patients’ death rate, as they lived for another 2–10 years after the first revision. Our study showed a comparable outcome regarding pain to the other nonagenarian studies reporting patient/surgeonreported outcomes, while those studies reported less improvement in function (Belmar et al. 1999, Joshi and Gill 2002, Pagano et al. 2004, Karrupiah et al. 2008) (see Table 1). Alfonso et al. (2007) suggested the reason for lower functional improvement might be other conditions such as systemic disease, poor balance, and impaired vision rather than the inefficiency of TKA. However, these studies, except for the study by Karrupiah et al. (2008), probably reported the surgeon’s opinions with the earlier KSS rather than the patients’ experiences. It is also important to point out that the nonagenarians in our study reported fewer symptoms in the KOOS both preoperatively and 1-year postoperatively than the younger age group; it could be speculated that nonagenarians may have had more disabling/pronounced symptoms from other parts of the body than those from the knee or, owing to the bias in patient selection, older patients may have had a more positive attitude

Acta Orthopaedica 2019; 90 (1): 53–59

in life and thus experienced less pain and fewer symptoms than the average patient (Thompson et al. 2018). Satisfaction with the surgery was high (21 of 22) and most of the patients were classified as OMERACT–OARSI responders (17 of 22). However, TKA surgery hardly improved the EQ-VAS in both nonagenarian and 65- to 74-year-old group, preoperatively to 1-year postoperatively according to our definition of clinical relevance. The reason may be that TKA surgery is not expected to improve the health in this group of patients. The variation in the regional incidence of TKA in Sweden indicates possible discrimination against nonagenarians amongst orthopedic surgeons in different counties. In addition, if general practitioners lack belief in knee arthroplasty in nonagenarians these patients are not referred to the orthopedic surgeon, which could contribute to the patient selection bias. This has rarely been discussed in the literature. In a prospective cohort study by Hamel et al. (2008), 174 patients with severe radiographic knee and hip OA were identified. During 1-year follow-up they reported 123 of the patients did not undergo TKA. Half of these patients reported that surgery had never been suggested, and the rest had been hesitant to have surgical treatment. Patients who were not operated were reported to be older than the rest of the cohort, indicating age discrimination. Our data on this topic may help both the surgeon and the patient to make an informed decision. The burden on health resources can also be debated while deciding on elective surgery in the elderly population. Fewer years of expected life may seem like a limiting factor for the cost-effectiveness of the surgery. Our data from the last 10 years showed that nonagenarian patients’ LoS was 2 days longer compared with those who were younger, which is associated with higher costs per TKA. In addition, there was a noticeable change in the mean LoS for the TKA cohort in this study during these 10 years (6.2 to 2.8 days), which reflects the change in healthcare trends. In Sweden, the average cost of TKA for 90- to 94-year-old patients was US$ 7,810 in 2016 (Swedish Association of Local Authorities and Regions 2017). This amount should be compared with the possible economic burden caused by keeping a healthy, active patient away from TKA. Pain, loss of strength, and lower bone density are few of the problems that can diminish the ability of nonagenarians to live independently. Karrupiah et al. (2008) have assumed that nonagenarian patients with severe OA would have required nursing home admission without the surgery, as 15% of all patient admissions to nursing homes are due to immobility secondary to arthritis (Guccione et al. 1989). In countries where the government finances social care; delaying nursing home placement may reduce the burden on health resources (Karrupiah et al. 2008). This might be suggested to be the case for Sweden because the average annual cost per person in a nursing home is US$ 90,561 (Swedish Association of Local Authorities and Regions 2018). We acknowledge limitations of this study. Clinical information (i.e., comorbidities, amount of blood loss, medical com-

Acta Orthopaedica 2019; 90 (1): 53–59

plications) was not available in the SKAR, which precluded analysis of risks associated with death. Another limitation is that PRO data have been collected since 2009 in the SKAR and the data collection is not compulsory, which reduced the number of patients for PRO analysis. In summary, our study emphasizes that TKA is a safe procedure with good outcome in nonagenarians with OA. Our findings may help surgeons and patients assessing the risks and outcome associated with the procedure.

The study was conceived by LL. OR, AWD, and EAS performed the analyses and LL and EAS wrote the initial draft. All the authors contributed to the interpretation of the data and to revision of the manuscript. The authors would like to thank Sofia Löfvendal for contribution to the discussion on health economics. Acta thanks Ove Furnes and Sten Rasmussen for help with peer review of this study.

Alfonso D T, Howell R D, Strauss E J, et al. Total hip and knee arthroplasty in nonagenarians. J Arthroplasty 2007; 22 (6): 807. Belmar C J, Barth P, Lonner J H, et al. Total knee arthroplasty in patients 90 years of age and older. J Arthroplasty 1999; 14 (8): 911. Biau D, Mullins M M, Judet T, Piriou P. Is anyone too old for a total knee replacement. Clin Orthop Relat Res 2006; 448 (448): 180-4. Culliford D, Maskell J, Judge A, Cooper C, Prieto-Alhambra D, Arden N K. Future projections of total hip and knee arthroplasty in the UK: results from the UK clinical practice research datalink. Osteoarthritis Cartilage 2015; 23: 594-600. D’Apuzzo M R, Pao A W, Novicoff W M, Browne J A. Age as an independent risk factor for postoperative morbidity and mortality after total joint arthroplasty in patients 90 years of age or older. J Arthroplasty 2014; 29 (3): 477-80. Easterlin M C, Chang D G, Talamini M, Chang D C. Older age increases short-term surgical complications after primary knee arthroplasty. Clin Orthop Relat Res 2013; 471 (8): 2611-20. EuroQol Group. EQ-5D. EuroQol Group website. http://www.euroqol.org; 2015. Guccione A A, Meenan R F, Anderson J J. Arthritis in nursing home residents: a validation of its prevalence and examination of its impact on institutionalization and functional status. Arthritis Rheum 1989; 32: 1546-53. Hamel M B, Toth M, Legedza A, Rosen M P. Joint replacement surgery in elderly patients with severe osteoarthritis of the hip or knee: decision making, postoperative recovery, and clinical outcomes. Arch Intern Med 2008; 168 (13): 1430-40. Human Mortality Database. University of California, Berkeley (USA), and Max Planck Institute for Demographic Research (Germany). Available at www.mortality.org or www.humanmortality.de (data downloaded on 01.04.2018). Inacio M C S, Paxton E W, Graves S E, Namba R S, Nemes S. Projected increase in total knee arthroplasty in the United States: an alternative projection model. Osteoarthritis Cartilage 2017; 25 (11): 1797-1803. Joshi A B, Gill G. Total knee arthroplasty in nonagenarians. J Arthroplasty 2002; 17 (6): 681.

Karrupiah S V, Banaszkiewicz P A, Ledingham W M. The mortality, morbidity and cost benefits of elective total knee arthroplasty in the nonagenarian population. Int Orthop 2008; 32 (3): 339-43. Kennedy J W, Johnston L, Cochrane L, Boscainos P J. Total knee arthroplasty in the elderly: does age affect pain, function or complications? Clin Orthop Relat Res 2013; 471: 1964-9. Koh I J, Kim T K, Chang C B, Cho H J, In Y. Trends in use of total knee arthroplasty in Korea from 2001 to 2010. Clin Orthop Rel Res 2013; 471 (5): 1441-50. Kuperman E F, Schweizer M, Joy P, Gu X, Fang M M. The effects of advanced age on primary total knee arthroplasty: a meta-analysis and systematic review. BMC Geriatrics 2016; 16: 41. Kurtz S M, Ong K L, Lau E, Bozic K J. Impact of the economic downturn on total joint replacement demand in the United States. J Bone Joint Surg (Am) 2014; 96-A: 624-30. Laskin R S. Total knee replacement in patients older than 85 years. Clin Orthop Relat Res 1999; 367: 43-9. Nemes S, Roflson 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 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. Pagano M W, McLamb L A, Trousdale R T. Total knee arthroplasty for patients 90 years of age and older. Clin Orthop Relat Res 2004; (418): 179. Petruccelli D, Rahman W A, de Beer J, Winemaker M. Clinical outcomes of primary total joint arthroplasty among nonagenarian patients. J Arthroplasty 2012; 27 (9): 1599-1603. Pham T, van der Heijde D, Altman R D, Anderson J J, Bellamy N, Hochberg M, Simon L, Strand V, Woodworth T, Dougados M. OMERACT–OARSI initiative: osteoarthritis Research Society International set of responder criteria for osteoarthritis clinical trials revisited. Osteoarthritis Cartilage 2004; 12 (5): 389-99. Roos E M, Roos H P, Lohmander L S, Ekdahl C, Beynnon B D. Knee Injury and Osteoarthritis Outcome Score (KOOS): development of a self-administered outcome measure. J Orthop Sports Phys Ther 1998; 28(2): 88-96. SKAR. Swedish Knee Arthroplasty Register Annual Report 2017. 2017; ISBN 978-91-8807-15-4. Statistics Sweden. The future population of Sweden 2018–2070. Demographic reports 2018:1. Swedish Association of Local Authorities and Regions. The Swedish Cases Costing Database (KPP) 2017. skl.se/ekonomijuridikstatistik/statistik/ kostnadperpatientkpp/kppdatabas.1079 (data downloaded on 25.05.2018). Swedish Association of Local Authorities and Regions. Öppna jämförelser 2017 Vård och omsorg om äldre: jämförelser mellan kommuner och län. 2018; ISBN 978-91-7875-674-2. Thompson K A, Bulls H W, Sibille K T, Bartley E J, Glover T L, Terry E L, Vaughn I A, Cardoso J S, Sotolongo A, Staud R, Hughes L B, Edberg J C, Redden D T, Bradley L A, Goodin B R, Fillingim R B. Optimism and psychological resilience are beneficially associated with measures of clinical and experimental pain in adults with or at risk for knee osteoarthritis. Clin J Pain 2018; [published online ahead of print Jul 21]. doi: 10.1097/AJP. 0000000000000642. Turkiewicz A, Gerhardsson de Verdier M, Engström G, Nilsson P M, Mellström C, Lohmander S, Englund M. Prevalence of knee pain and knee OA in southern Sweden and the proportion that seeks medical care. Rheumatology 2014; 54: 827-35. Williams D P, Price A J, Beard D J, Hadfield S G, Arden N K, Murray D W, Field R E. The effects on patient-reported outcome measures in total knee replacements. Bone Joint J 2013; 95-B: 38-44.

Acta Orthopaedica 2019; 90 (1): 60–66

Weight affects survival of primary total knee arthroplasty: study based on the Danish Knee Arthroplasty Register with 67,810 patients and a median follow-up time of 5 years David GØTTSCHE 1, Kirill GROMOV 1, Petra H VIBORG 2, Elvira V BRÄUNER 3, Alma B PEDERSEN 4, and Anders TROELSEN 1 1 Department

of Orthopedic Surgery, Copenhagen University Hospital Hvidovre, Clinical Orthopaedic Research Hvidovre (CORH), Denmark; 2 The Danish Clinical Registries (RKKP); 3 Department of Growth and Reproduction, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark; 4 Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark Correspondence: dgmadsen@gmail.com Submitted 2018-07-06. Accepted 2018-09-26

Background and purpose — Obesity is a rising issue worldwide and growing evidence supports poor outcome amongst obese patients following total knee arthroplasty (TKA). Using nationwide registries we investigated the association between bodyweight and risk of revision of primary TKA. Patients and methods — All primary TKA performed during 1997–2015, weight at time of primary TKA and subsequent TKA revisions were identified in the Danish Knee Arthroplasty Register (DKR). Data on comorbidities and a priori selected confounding variables were collected from nationwide registries. The association between weight and 1st time TKA revision was calculated as both crude and adjusted hazard ratios (aHR) with 95% confidence intervals (CI) using Cox regression. Results — Of 67,810 identified primary TKAs, 4.8% were revised within a median follow-up time of 5.4 years. No association between weight and risk of any revision in patients aged 18–54 and 55–70 years was found. Increased risk of any revision was seen in patients > 70 years, 80–89 kg (aHR = 1.5, CI 1.2–1.8), 90–99 kg (aHR = 1.7, CI 1.3–2.1) and patients > 99 kg (aHR = 1.6, CI 1.3–2.1), as well as those weighing 45–60 kg (aHR = 1.4, CI 1.1–1.9) compared with same aged patients weighing 70–79 kg. Interpretation — We found a complex association between weight and knee arthroplasty survival. There was an increased risk of any revision in patients older than 70 years of age weighing < 60 kg and > 80 kg. Patients aged 18–55 years weighing 60–69 kg had a lower risk of revision compared with all other weight groups, whereas weight was not found to affect risk of any revision in patients aged 55–70 years.

Obesity and associated healthcare issues comprise a rising problem worldwide. Several patient-related factors, including obesity, have been reported to be associated with poor outcome following total knee arthroplasty (TKA). Obese TKA patients have an increased risk of infection and deep vein thrombosis (Si et al. 2015, Electricwala et al. 2017), as well as increased load on the prosthesis–bone junction, leading to increased risk of bone or ligament insufficiency and migration (Astephen Wilson et al. 2010). Increased risk for revision following primary TKA has been reported in patients who are overweight (BMI > 25) (Morrison 1970, Griffin et al. 1998, Foran et al. 2004a, 2004b, Ward et al. 2015, Zingg et al. 2016). However, patients can have high weight even if BMI is normal. Therefore, it seems justifiable to use weight instead of BMI to assess the stress load on the prosthesis. No previous studies have assessed the relationship between patient weight and revision of the primary TKA in a broad spectrum of TKA patients. Previous studies are limited by small sample size, inclusion of only patients with severe obesity compared with normal BMI, single-center study, and a high number of patients lost to follow-up or a lack of information regarding these patients. Thus, little is known about whether specific weight indices can be used as cutoffs to determine which patients are at increased risk for having a poor postoperative outcome. We hypothesized that risk of revision of primary TKA increased with increasing weight. Therefore, we conducted a population-based follow-up study to examine the association between weight at the time of primary TKA and risk of any revision as well as risk of revision due to aseptic loosening or infection.

Acta Orthopaedica 2019; 90 (1): 60–66

Patients and methods Study design and settings We conducted this study in Denmark (5.7 million inhabitants) using prospectively collected data from population-based medical and administrative registers. The Danish National Health Service provides tax-supported healthcare to all Danish residents. The study was performed in accordance with the Reporting of studies Conducted using Observational Routinely collected health Data (RECORD) statement (Benchimol et al. 2015). Data sources The Danish Knee Arthroplasty Register (DKR) is a national clinical quality database which includes records of all primary TKAs and TKA revisions. The DKR started in January 1, 1997 and collects pre-, peri-, and postoperative data. The DKR is certified by and operates according to requirements of the Danish National Board of Health. Using DKR, we identified our study population, including all patients registered with a primary TKA in DKR in the period January 1, 1997 until end of follow-up on November 19, 2015. From DKR we also obtained information on weight, perioperative complications, type of fixation, indication for primary TKA, and date and indication for revision surgery. The data completeness in DKR was 70% and 56% for primary TKA and revision TKA in 1997, rising to 99% and 96% in 2015 (Knæalloplastikregister 2017). Patients registered with more than 1 primary TKA on the same knee, more than 1 primary TKA without information on laterality, revision of TKA prior to primary TKA on the same knee, missing value for weight, weight recorded as either < 45 or > 200 kg, or registration error of date were excluded from analyses. In addition, the study population was restricted to patients aged at least 18 years of age at the time of primary TKA. Patients with inverse hybrid fixation techniques (cemented femur and cementless tibia) were also excluded from analyses as this method is rare and the registration is most likely due to error (Figure 1). Excluded patients were similar to the included patients with respect to age and sex distribution at the time of primary TKA or indication for primary TKA (data not shown). Data from DKR were linked to the Danish Civil Registration System (CRS), to obtain information on vital status (active, date of death/emigration). The CRS was established in 1968. All Danish residents (native or immigrated) are registered and assigned a unique personal identification number (CPR number) in the CRS encoding age, sex, and date of birth. The CRS is updated daily. The CPR number allows unambiguous linking between all the Danish medical and administrative registries (Pedersen 2011, Schmidt et al. 2014). Further, data were linked to the Danish National Patient Register (NPR), which contains nationwide clinical data on

Primary total knee arthroplasty (TKA) 01.01.1997 to 19.11.2015 n = 72,491 Excluded (n = 4,681): – > 1 primary TKA on same knee, or > 1 primary TKA without side registration, 619 – revision before date of primary TKA, or other date registration error, 112 – missing or incomplete registration of weight, or weight < 45 or > 200 kg, 3,516 – age < 18 years, 2 – inverse hybrid used on primary TKA, 432 Study population n = 67,810

Figure 1. Inclusion and exclusion of patients in the study population. Patients with inverse hybrid fixation techniques were excluded from analyses as this method is rare and the registration is most likely due to error.

inpatients admitted to Danish hospitals since 1977, and outpatients since 1994 (Lynge et al. 2011, Schmidt et al. 2015). Each record in the NPR contains the patient’s CPR number and information on treatment, surgery, dates of admission, primary discharge diagnosis, and up to 20 secondary discharge diagnoses. Diagnoses are classified according to the International Classification of Diseases (ICD). The 8th edition (ICD-8) was used from 1977 to 1993 and the 10th edition (ICD-10) thereafter. We obtained information on complete hospitalization history of each patient from the NPR to assess comorbidity at the time of primary TKA. We computed Charlson’s Comorbidity Index (CCI) at the time of primary TKA and translated into corresponding ICD-8 and ICD-10 codes, similar to the approach by Deyo et al. (1992) (see Supplementary data). We classified patients into 3 levels according to the degree of comorbidity: index low (0 points), corresponding to patients with no previous recorded disease categories implemented in CCI; index medium (1–2 points); and index high (≥ 3 points). Exposure Weight at the time of primary TKA was the exposure in our study and patients were categorized into 1 of the following 6 intervals (45–60, 60–69, 70–79, 80–89, 90–99, and 99–200 kg). Outcome The outcome was time to revision following primary TKA, defined as a new surgical intervention involving partial or complete removal of the implant. We analyzed revision due to (1) all-cause, (2) aseptic loosening, or (3) infection. Statistics Follow-up started on the date of the primary TKA, and ended on the date of the 1st revision, death, emigration, disappearance, or end of follow-up on November 19, 2015, whichever came first. Cumulative incidence function (CIF) was used to estimate cumulative implant failure probabilities as a result of: (1) all-

cause, (2) aseptic loosening, or (3) infection including death or emigration as a competing risk. A Cox proportional hazards analysis was used to assess the risk of revision by computing cause-specific hazard ratios (HR) for the revision with 2-sided 95% confidence interval (CI). All data were analyzed with and without adjustment for a priori selected variables based on potential confounding effects extracted at the time of primary TKA including sex, age (18–55, 55–70, > 70 years); CCI (low, medium, high); indication for primary TKA (primary arthrosis, secondary arthrosis (defined as arthrosis with a known cause, e.g. sequelae after meniscectomy), fracture (tibia, femur, patella), arthritis, or other); fixation technique (cemented, hybrid, cementless); and perioperative complications (yes, no). We estimated the HRs for revision comparing different weight intervals with the 70–79 kg weight interval (reference). Since there was an interaction between age and weight, the aHR were calculated separately for each of the 3 age groups. P-values of < 0.05 were considered to be statistically significant. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc, Cary, NC, USA). Ethics, funding and potential conflicts of interest The study was approved by the Danish Data Protection Agency (DDPA): approval number 2012-58-0004. By Danish law, ethical approval and informed consent are not required for entirely register-based studies not involving contact with study participants. Financial support was received from PROCRIN (Program for Clinical Research Infrastructure), a grant of 2 months’ salary for statistical and epidemiological support (PV and EB). No conflicts of interest to declare.

Results Among 72,491 patients recorded in the DKR, 67,810 were included in final analyses. Median follow-up time was 5.4 years. A total of 3,270 patients sustained all-cause revision (4.8%), of which 1,109 (34%) were due to aseptic loosening, 173 (5.3%) due to infection and 1,988 (61%) due to other causes including pain without loosening, instability, secondary insertion of patella component, replacement of polyethylene, second part of 2-stage revision, progression of arthrosis and other causes (Table 1). The median age at time of primary TKA was 68 years and men represented 39% of the study population. Fixation of primary TKA was mainly cemented (77%), with hybrid (femur cementless and tibia cemented) and cementless techniques representing 16% and 7.7% respectively. A high comorbidity (CCI > 2) was found in 18,099 patients (27%) at the time of their primary TKA (Table 1). Association between weight and revision of primary TKA Patients weighing more than 90 kg had the highest cumulative incidence of all-cause revision during follow-up,

Acta Orthopaedica 2019; 90 (1): 60–66

whereas patients who weighed 60–69 kg had the lowest cumulative incidence of revision (Figure 2). The cumulative incidences for the different weight groups were shown to be significantly different by utilization of the non-parametric Gray’s test (< 0.001). After 10 years, the CIF were 0.05 (standard error 0.005), 0.04 (0.003), 0.05 (0.002), 0.06 (0.002), 0.06 (0.003), and 0.08 (0.003) for the weight groups 45–60 kg, 60–69 kg, 70–79 kg, 80–89 kg, 90–99 kg, and 99–200 kg, respectively. After approximately 17 years, the accuracy declines due to the low number of patients with such long follow-up (Figure 2). Crude HRs for all-cause revision were 1.2 (CI 1.1–1.3) and 1.4 (CI 1.3–1.6) for weight groups 90–99 kg and 99–200 kg, and 0.8 (CI 0.7–0.9) for weight group 60–69 kg compared with patients in weight group 70–79 kg (Table 2). In the adjusted analyses in which sex, weight, and age were adjusted for, a statistically significant interaction between age and weight groups was observed, thus these variables could not be considered separately (data not shown). In the youngest age group (18–55 years), patients in the weight group 60–69 kg had a lower risk of revision (HR = 0.7, CI 0.5–1.0) compared with the reference group (70–79 kg). Higher weight in the youngest patients was not associated with increased risk of all-cause revision (Table 2). In the age group > 70 years, the weight groups 45–60 kg, 80–89 kg, 90–99 kg, and 90–200 kg had a 40–68% increased risk of all-cause revision compared with the weight group 70–79 kg. Association between weight and revision due to aseptic loosening We found reduced risk of revision due to aseptic loosening (aHR = 0.7, CI 0.5–0.8) for the weight group 60–69 kg compared with the group 70–79 kg. There was no association between other weight groups and revision due to aseptic loosening when adjusting for all available confounders (Table 3). The cumulative incidences for the different weight groups were likewise shown to be significantly different by utilization of the non-parametric Gray’s test (< 0.001). After 10 years, the CIF were 0.02 (0.003), 0.01 (0.002), 0.02 (0.001), 0.02 (0.002), 0.02 (0.002), and 0.03 (0.002) for the weight groups 45–60 kg, 60–69 kg, 70–79 kg, 80–89 kg, 90–99 kg, and 99–200 kg, respectively. As for all-cause revision, the accuracy declines after approximately 17 years due to low number of patients (Figure 3). Association between weight and revision due to infection The cumulative incidences for the risk of revision due to infection between weight groups were not statistically significantly different (data not shown). However, it should be noted that few patients were revised due to infection (173), with only 4 patients in the lowest weight group. Revision due to infection was not found to be related to weight groups (Table 4).

Acta Orthopaedica 2019; 90 (1): 60–66

Table 1. Demography across weight groups Parameter Value

No. of patients 45–-60 60–69

Weight (kg) 70–79 80–89

90–99

99–200

p-value a

Subjects b 67,810 3,074 (4.5) 9,336 (13.8) 16,336 (24.1) 16,538 (24.4) 10,585 (15.6) 11,941 (17.6) Sex b Women 41,506 2,947 (4.3) 8,182 (12.1) 11,492 (16.9) 9,072 (13.4) 4,767 (7.0) 5,046 (7.4) < 0.001 Men 26,304 127 (0.2) 1,154 (1.7) 4,844 (7.1) 7,466 (11.0) 5,818 (8.6) 6,895 (10.2) Age at primary TKA (overall median age: 68) median, years 67,810 74 72 70 68 66 64 < 0.001 Charlson’s Comorbidity Index (CCI) c Low (CCI 0) 38,355 1,535 (49.9) 5,255 (56.3) 9,349 (57.2) 9,502 (57.5) 6,090 (57.5) 6,624 (55.5) Medium (CCI 1–2) 11,356 471 (15.3) 1,519 (16.3) 2,758 (16.9) 2,786 (16.8) 1,785 (16.9) 2,037 (17.1) High (CCI > 2) 18,099 1,068 (34.7) 2,562 (27.4) 4,229 (25.9) 4,250 (25.7) 2,710 (25.6) 3,280 (27.5) Indication for primary TKA c, d Primary arthrosis 55,683 2,266 (74.2) 7,536 (81.3) 13,599 (84.1) 13,662 (83.4) 8,707 (83.0) 9,913 (83.7) Secondary arthrosis 7,220 257 (8.4) 850 (9.2) 1,519 (9.4) 1,867 (11.4) 1,303 (12.4) 1,424 (12.0) Sequelae after fracture 1,538 157 (5.1) 267 (2.9) 384 (2.4) 316 (1.9) 199 (1.9) 216 (1.8) Arthritis 2,182 337 (11.0) 506 (5.5) 552 (3.4) 413 (2.5) 184 (1.8) 190 (1.6) Other e 595 39 (1.3) 108 (1.2) 121 (0.8) 134 (0.8) 97 (0.9) 96 (0.8) Fixation technique of primary TKA c Cemented 2,465 (80.2) 7,237 (77.5) 12,307 (75.3) 12,766 (77.2) 8,058 (76.1) 9,125 (76.4) Uncemented 215 (7.0) 697 (7.5) 1,346 (8.2) 1,275 (7.7) 812 (7.7) 906 (7.6) Hybrid 394 (12.8) 1,402 (15.0) 2,683 (16.4) 2,497 (15.1) 1,715 (16.2) 1,910 (16.0) Revision TKA (rTKA) c 3,270 136 (4.4) 337 (4.0) 751 (4.6) 807 (4.9) 530 (5.0) 709 (5.9) Perioperative complications c 589 35 (1.1) 105 (1.1) 136 (0.8) 130 (0.8) 77 (0.7) 106 (0.9) Time to revision (overall median time to revision: 1.88 years) median, years 3,270 1.86 1.94 1.85 1.95 1.89 1.78 0.9 Indication for rTKA c Aseptic loosening 1,109 43 (31.6) 98 (29.1) 276 (36.8) 272 (33.7) 177 (33.4) 243 (34.3) Infection 173 4 (2.9) 19 (5.6) 37 (4.9) 42 (5.2) 34 (6.4) 37 (5.2) Other e 1,988 89 (65.4) 220 (65.3) 438 (58.3) 493 (61.1) 319 (60.2) 429 (60.5) Deaths and emigration (overall 18.6%) during follow-up c 12,632 906 (29.5) 2,153 (23.1) 3,407 (20.9) 2,879 (17.4) 1,532 (14.5) 1,755 (14.7) Follow-up time (years), median Non-revised 64,540 5.4 5.7 5.8 5.6 5.2 5.0 Revision—all cause 3,270 1.9 1.9 1.9 2.0 1.9 1.8 Revision—aseptic loosening 1,109 3.6 2.8 3.0 3.0 2.6 2.9 Revision—infection 173 3.4 1.9 1.6 0.6 1.3 0.8 a For

categorical variables the chi-squared test was used and for continuous variables the Kruskal–Wallis test was used. P-value < 0.05 indicates a significant difference between values in same row. b Frequency and % of total c Frequency and % of patients in each weight group d The total number of included patients is 67,810; due to missing data on indication for primary TKA only 67,218 patients are listed. e Other indications for revision TKA include: pain without loosening, instability, secondary insertion of patellar component, replacement of polyethylene, 2nd part of 2-stage revision, progression of arthrosis and other causes.

Table 2. Crude and adjusted hazard ratio (HR) with 95% confidence interval (CI) for revision (all causes) according to different weight and age groups. Values are number of subjects (95% CI) HR

No. of Weight (kg) patients 45–-60 60–69 70–79 (ref.) 80–89

Crude HR Adjusted HR c 18–55 years 55–70 years > 70 years

67,810

0.98 (0.82–1.20)

7,151 32,218 27,849

0.91 (0.61–1.36) 0.73 (0.54–0.98) b 0.86 (0.63–1.18) 0.84 (0.70–1.00) 1.40 (1.05–1.87) b 0.90 (0.71–1.13)

a P-value for linear trend b Significant, p < 0.05 c Adjusted for sex, comorbidities,

0.79 (0.69–0.89) b

1 1.09 (0.99–1.21)

90–99

99–200

p-value a

1.17 (1.05–1.31) b 1.40 (1.27–1.55) b < 0.001

1 0.92 (0.74–1.15) 0.89 (0.70–1.13) 0.96 (0.78–1.18) < 0.001 1 0.90 (0.78–1.04) 0.89 (0.76–1.04) 1.07 (0.93–1.23) 1 1.48 (1.22–1.79) b 1.68 (1.34–2.11) b 1.60 (1.25–2.05) b

perioperative complications, years after primary TKA, type of fixation, and indication for primary TKA. Due to missing values in data, the total number of patients in the adjusted calculations is 67,218.

Acta Orthopaedica 2019; 90 (1): 60–66

CIF curves for any revision

CIF curves for aseptic loosening

45–59 kg 70–79 kg 90–99 kg

0.125

0.07

60–69 kg 80–89 kg 100–200 kg

0.06

0.100 0.05 0.075

0.04

0.03

0.050

0.02 0.025 0.01 0.000

0.00 0

Years after index operation

Figure 2. Cumulative all-cause incidence curves for revision of primary TKA according to weight groups.

Discussion In this nationwide cohort study, we assessed the effects of weight on the risk of revision following primary TKA. We found an association between weight above 80 kg and increased risk of all-cause revision in patients aged over 70 years at the time of primary TKA. Further, patients over 70 years who weighed 45–60 kg were at increased risk of allcause revision compared with patients weighing 70–79 kg. There was no indication of dose response. Our findings imply that high weight for aged patients is an important risk factor for revision of primary TKA. In the crude analysis and in the cumulative all-cause incidence we found that patients weighing 60–69 kg had the lowest risk of revision. Whether obesity is associated with higher revision rate is debatable. Bordini et al. (2009) found no different rates of survival in obese patients compared with normal-weight patients at 5 year follow- up. However, the total number of TKAs in that study was only 9,735 with 186 revisions and a revision rate of 1.9 %. The number of morbidly obese patients having revision was only 4 and only 59 obese patients had revision; it is therefore not unlikely that no difference were found as some subgroups had a low number of patients included. Several researchers suggest, however, that obesity is associated with increased risk of revision (Si et al. 2015, Ward et al. 2015, Werner et al. 2015, Zingg et al. 2016, Electricwala et al. 2017). Obesity is a risk factor for primary arthrosis and primary arthrosis is the main indication for primary TKA; it is therefore expected to find lower age among the heaviest patients. Shah et al. (2017) found a significantly higher occurrence of obesity and morbid obesity among patients < 65.We could not show that increased weight was associated with increased risk of revision in younger patients, but patients weighing 60–69

Years after index operation

Figure 3. Cumulative incidence curves for revision of primary TKA due to aseptic loosening for weight groups. For color codes, see Figure 2

kg had a lower risk of revision. These results may be due to the lower number of patients aged 18–55 years. We cannot explain why patients > 70 years with high weight are at higher risk of revision than younger patients of the same weight. Revisions due to aseptic loosening represented 34% of the revisions and were thereby the single largest indication for revision TKA among our patients. The CIF curves show, similar to crude HR, that only patients weighing 60–69 kg and 99–200 kg were at statistically significantly different risk than the reference group. Even though aseptic loosening is the commonest indication for revision, it still only represents 1,109 patients, which might not be enough to show a statistically significant difference between weight groups. Electricwala et al. (2017) found aseptic loosening to be the 3rd highest reason for revision after infection and instability. They also found that not only obese patients but also overweight patients had a higher risk of revision. In other studies, Zingg et al. (2016) and Ward et al. (2015) similarly found a higher risk of revision with rising BMI, but did not separate their results into reasons for revision. Revisions due to infection disclosed a completely different pattern in which no statistically significant difference in risk of revision for different weight groups was observed, which backs the notion that infection occurs randomly across risk groups. This might be due to the low occurrence of infection leading to revision; only 173 patients were revised due to infection or 0.26% of patients. TKA is a safe procedure and the overall revision rate in our data was 4.8% with a median follow-up time of 5.4 years. Obesity is correlated with a number of risk factors and comorbidities affecting major surgery like TKA. It is strongly advisable for overweight patients to lose weight prior to TKA surgery. Both our results regarding stress load on the prosthesis in

Acta Orthopaedica 2019; 90 (1): 60–66

relation to patient weight and other comparable studies show increased risk of revision in overweight patients. Among our patients older than 70 years the revision risk was statistically increased even when adjusted for a number of a priori chosen confounders. Strengths and limitations The completeness of DKR is high, especially since 2007, but some patients have been lost during registration. A validation of the Danish Hip Arthroplasty Register regarding reporting of infections to the register has been published. Several data sources including NPR were used, and showed an estimated “true” incidence of surgically treated infections to be 40% higher than reported by national arthroplasty registries alone (Gundtoft et al. 2015). To our knowledge no similar study has been undertaken with data from the DKR, but we may expect some degree of under-reporting of infections to the DKR as well. When assessing our results, we adjusted for a number of a priori chosen confounders such as sex, age, comorbidities, peroperative complications, years after primary surgery, type of fixation, and indication for primary TKA. But residual confounders may occur because, for example, we lack data on comorbidities recorded at general practitioners, and on psychiatric comorbidities, as well as severity of some comorbidities included in the CCI. Other risk factors like alcohol and smoking were not available in our dataset, causing unmeasured confounding. Likewise it was not possible to investigate the correlation between BMI and revision rate of primary TKA, since registration of patient height first began nationwide in 2011. Most studies use BMI to graduate obesity, but with disadvantages, since this does not account for the absolute weight load on the prosthesis. Percentage body fat has been shown to be superior to BMI but those data were not available in our registries (Ledford et al. 2014). Summary and perspectives We found an increased risk of any revision following primary TKA in patients older than 70 years of age weighing < 60 kg and > 80 kg. Patients aged 18–55 years weighing 60–69 kg had a lower risk of revision compared with all other weight groups, whereas weight did not affect risk of any revision in patients 55–70 years. Data sharing If approved by the DDPA access to raw data will be granted on request. Access to protocol and programming code will be granted on request by contact with the corresponding author. Supplementary data Charlson Comorbidity Index (CCI) and Tables 3 and 4 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2018.1540091

All authors contributed significantly to this paper. DG, KG, and AT jointly initiated the study and wrote the protocol. EB modified the protocol and applied for approval by the DDPA. DG wrote the main paper. PV performed the statistics and wrote the statistics rapport. All authors performed critical revision of the paper. Acta thanks Martin Lindberg-Larsen and Maziar Mohaddes for help with peer review of this study.

Astephen Wilson J L, Wilson D A, Dunbar M J, Deluzio K J. Preoperative gait patterns and BMI are associated with tibial component migration. Acta Orthop 2010; 81(4): 478-86. doi: 10.3109/17453674.2010.501741. Benchimol E I, Smeeth L, Guttmann A, Harron K, Moher D, Petersen I, Sorensen H T, von Elm E, Langan S M, Committee R W. The REporting of studies Conducted using Observational Routinely-collected health Data (RECORD) statement. PLoS Med 2015; 12(10): e1001885. doi: 10.1371/ journal.pmed.1001885. Bordini B, Stea S, Cremonini S, Viceconti M, De Palma R, Toni A. Relationship between obesity and early failure of total knee prostheses. BMC Musculoskelet Disord 2009; 10: 29. doi: 10.1186/1471-2474-10-29. Deyo R A, Cherkin D C, Ciol M A. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992; 45(6): 613-19. Electricwala A J, Jethanandani R G, Narkbunnam R, Huddleston J I, 3rd, Maloney W J, Goodman S B, Amanatullah D F. Elevated body mass index is associated with early total knee revision for infection. J Arthroplasty 2017; 32(1): 252-5. doi: 10.1016/j.arth.2016.05.071. Foran J R, Mont M A, Etienne G, Jones L C, Hungerford D S. The outcome of total knee arthroplasty in obese patients. J Bone Joint Surg Am 2004a; 86-A(8): 1609-15. Foran J R, Mont M A, Rajadhyaksha A D, Jones L C, Etienne G, Hungerford D S. Total knee arthroplasty in obese patients: a comparison with a matched control group. J Arthroplasty 2004b; 19(7): 817-24. Griffin F M, Scuderi G R, Insall J N, Colizza W. Total knee arthroplasty in patients who were obese with 10 years followup. Clin Orthop Relat Res 1998; (356): 28-33. 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. Knæalloplastikregister D. Dansk Knæalloplastikregister, Årsrapport 2017; 2017. Ledford C K, Ruberte Thiele R A, Appleton J S, Jr, Butler R J, Wellman S S, Attarian D E, Queen R M, Bolognesi M P. Percent body fat more associated with perioperative risks after total joint arthroplasty than body mass index. J Arthroplasty 2014; 29(9 Suppl): 150-4. doi: 10.1016/j.arth.2013.12.036. Lynge E, Sandegaard J L, Rebolj M. The Danish National Patient Register. Scand J Public Health 2011; 39(7 Suppl): 30-3. doi: 10.1177/1403494811401482. Morrison J B. The mechanics of the knee joint in relation to normal walking. J Biomech 1970; 3(1): 51-61. Pedersen C B. The Danish Civil Registration System. Scand J Public Health 2011; 39(7 Suppl): 22-5. doi: 10.1177/1403494810387965. 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. doi: 10.1007/s10654-014-9930-3. Schmidt M, Schmidt S A, Sandegaard J L, Ehrenstein V, Pedersen L, Sorensen H T. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7:449-90. doi: 10.2147/ CLEP.S91125. Shah S H, Schwartz B E, Schwartz A R, Goldberg B A, Chmell S J. Total knee arthroplasty in the younger patient. J Knee Surg 2017; 30(6): 555-9. doi: 10.1055/s-0036-1593619.

Si H B, Zeng Y, Shen B, Yang J, Zhou Z K, Kang P D, Pei F X. The influence of body mass index on the outcomes of primary total knee arthroplasty. Knee Surg Sports Traumatol Arthrosc 2015; 23(6): 1824-32. doi: 10.1007/ s00167-014-3301-1. Ward D T, Metz L N, Horst P K, Kim H T, Kuo A C. Complications of morbid obesity in total joint arthroplasty: risk stratification based on BMI. J Arthroplasty 2015; 30(9 Suppl): 42-6. doi: 10.1016/j.arth.2015.03.045.

Acta Orthopaedica 2019; 90 (1): 60â&#x20AC;&#x201C;66

Werner B C, Evans C L, Carothers J T, Browne J A. Primary total knee arthroplasty in super-obese patients: dramatically higher postoperative complication rates even compared to revision surgery. J Arthroplasty 2015; 30(5): 849-53. doi: 10.1016/j.arth.2014.12.016. Zingg M, Miozzari H H, Fritschy D, Hoffmeyer P, Lubbeke A. Influence of body mass index on revision rates after primary total knee arthroplasty. Int Orthop 2016; 40(4): 723-9. doi: 10.1007/s00264-015-3031-0.

Acta Orthopaedica 2019; 90 (1): 67–73

Similar polyethylene wear between cemented and cementless Oxford medial UKA: a 5-year follow-up randomized controlled trial on 79 patients using radiostereometry Kristian HORSAGER 1, Frank MADSEN 2, Anders ODGAARD 3, Claus Fink JEPSEN 2, Lone RØMER 4, Per Wagner KRISTENSEN 5, Bart L KAPTEIN 6, Kjeld SØBALLE 1, and Maiken STILLING 1 1 Department

of Clinical Institute, Aarhus University, Aarhus, Denmark; 2 Department of Orthopedics, Aarhus University Hospital, Aarhus, Denmark; 3 Department of Orthopaedics, Herlev-Gentofte Hospital, Copenhagen University Hospital, Hellerup, Denmark; 4 Department of Radiology, Aarhus University Hospital, Aarhus, Denmark; 5 Department of Orthopedics, Vejle Hospital, Vejle, Denmark; 6 Department of Orthopedic Surgery, Biomechanics and Imaging Group, Leiden University Medical Center, Leiden, Netherlands Correspondence: Kristian.horsager@gmail.com Submitted 2018-05-08. Accepted 2018-10-09.

Background and purpose — Hydroxyapatite (HA)coated implants have been associated with high polyethylene wear in hip arthroplasties. HA coating as a promoter of wear in knee arthroplasties has not been investigated. We compared the wear-rate of the polyethylene bearing for cemented and cementless HA-coated Oxford medial unicondylar knee arthroplasties (UKA). Secondarily, we investigated whether wear-rates were influenced by overhang or impingement of the bearing. Patients and methods — 80 patients (mean age 64 years), treatment-blinded, were randomized to 1 of 3 Oxford medial UKA versions: cemented with double-pegged or single-pegged femoral component or cementless HA-coated with double-pegged femoral component (ratios 1:1:1). We compared wear between the cemented (n = 55) and cementless group (n = 25) (ratio 2:1). Wear, impingement, and overhang were quantified between surgery and 5-year follow-up using radiostereometry. Clinical outcome was evaluated with the Oxford Knee Score. Results — The mean wear-rate for patients without bearing overhang was 0.04 mm/year (95% CI 0.02–0.07) for the cemented group and 0.05 mm/year (CI 0.02–0.08) for the cementless group. The mean difference in wear was 0.008 mm/year (CI –0.04 to 0.03). No impingement was identified. Half of the patients had medial bearing overhang, mean 2.5 mm (1–5). Wear increased by 0.014 mm/year for each mm increment in overhang. The mean Oxford Knee Score was 39 for the cementless group and 38 for the cemented group at the 5-year follow-up. Interpretation — The wear-rates were similar for the 2 fixation methods, which supports further use of the cementless Oxford medial UKA. However, a caveat is a relatively large 95% CI of the mean difference in wear-rate. Component size and position is important as half of the patients presented with an additional increase in wear-rate due to medial bearing overhang.

Polyethylene (PE) wear is an important factor for failure of knee arthroplasties (Lonner et al. 1999, Røkkum et al. 1999, Sharkey et al. 2002). The failure mechanism is particleinduced osteolysis and subsequent late aseptic loosening or accelerated catastrophic wear (Kadoya et al. 1998, Harris 2001, Naudie et al. 2007). The Swedish Knee Arthroplasty Register from 2016 reported that 11% of all unicondylar knee arthroplasty (UKA) revisions were performed due to PE wear (Sundberg et al. 2016). The Oxford medial UKA is designed to reduce PE wear with a fully congruent mobile bearing dynamically linked between a spherical femoral component and a flat tibial surface (O’Connor and Goodfellow 1996). The PE bearing has previously been found with very low wear-rates of less than 0.03 mm/year, if the bearing moves freely with no impingement against surrounding structures (Psychoyios et al. 1998, Kendrick et al. 2010a, 2010b). Traditionally, the Oxford medial UKA has been inserted by use of bone cement. In 2004, a cementless hydroxyapatite (HA)-coated design was introduced to improve fixation properties, eliminate cementing errors, and reduce duration of surgery (Tai and Cross 2006, Pandit et al. 2009, Liddle et al. 2013, Kendrick et al. 2015). However, HA coating has been associated with high wear-rates and unacceptable revision rates in several studies of total hip arthroplasties (THA) (Hallan et al. 2006, Kim et al. 2006, Gottliebsen et al. 2012). Revision studies have found HA particles embedded in the articulating surface of the PE liner and argue that HA is the most probable cause of third-body wear (Bloebaum et al. 1994, 1997, Morscher et al. 1998, Røkkum and Reigstad 1998). To our knowledge, HA coating as a promoter of PE wear in knee arthroplasties has not been investigated. Radiostereometry (RSA) and recent developments of advanced model-based software allow for the measurement of in-vivo PE wear in the Oxford medial UKA with high accuracy (Van IJsseldijk et al. 2011, 2014).

Acta Orthopaedica 2019; 90 (1): 67–73

Enrollment

Table 2. Patient demographics at baseline

Assessed for eligibility n = 163 Excluded (n = 83): – not meeting inclusion criteria, 34 – declined to participate, 40 – inclusion ended, all operated, 7 – other reasons, 2 Randomized n = 80 Allocation

Allocated to cemented double–peg design n = 26

Allocated to cemented single–peg design n = 29

Cemented Cementless (n = 55) (n = 25)

Men : women, n Right : left, n Age, years a Weight, kg a Surgeons involved, n

30 : 25 32 : 23 63 (9) [47–81] 87 (13) [67–121] 7

a mean

18 : 7 11 : 14 65 (10) [38–81] 88 (14) [61–110] 5

(SD) [range]

Allocated to cementless double–peg design n = 25

gibility and 83 were excluded (Figure 1). All patients received a phase 3-alpha Oxford medial UKA with ArCom ultra-high molecular weight polyethyleneFollow-up bearing (ZimmerBiomet, Warsaw, IN, USA). The Lost to follow-up (n = 4): Lost to follow-up (n = 2): implant was inserted with bone cement (Refobacin – died after 1–2 years follow-up, 3 – died after 2 years follow-up, 1 – left study due to sickness, 1 – absent from 5 year follow-up, 1 Bone Cement R, ZimmerBiomet, Warsaw, IN, USA) Discontinued intervention (n = 2): Discontinued intervention (n = 0) or press-fit cementless HA-coated fixation during – revision due to bearing dislocation, 1 2009–2011. The cementless Oxford UKA used in this – revision due to loosening, 1 study (2nd generation) had 0.75 mm titanium plasma Analysis spray and was coated with 55 μm HA. Analyzed (n = 23) Analyzed (n = 48) Patients were randomized in blocks of 12 with Excluded (n = 0): Excluded (n = 1): sequentially numbered envelopes that were opened – excluded from RSA because of a size XS tibial compent, 1 during surgery. Randomization was performed 3-armed with equal distribution, as the cemented Figure 1. Consort flow chart. All available data were used in the statistical analysis. group could receive a double-pegged or singleA complete dataset was collected for 48 patients in the cemented group and for 23 patients in the cementless group. Demographic data at baseline were compapegged femoral component. The cementless femoral rable between the 2 groups (Table 2). All patients were followed with RSA, concomponent was only double-pegged. This allowed a ventional radiographs, and Oxford Knee Score over the 5-year follow-up period. 3-way comparison of the femoral-component migration, which is a primary endpoint of the RCT (unpublished) in which the present study is nested. This The primary endpoint of this study was a comparison of PE is not of interest with respect to PE wear and therefore we wear between cemented and cementless Oxford medial UKA, considered the randomization 2-armed with a 2:1 ratio for the which is a secondary objective of the randomized controlled cemented (n = 55) and cementless group (n = 25). trial (RCT). We hypothesized that there would be no difference in the wear-rate. Secondarily, we investigated whether Sample size wear-rates were influenced by bearing overhang or impinge- Sample size and power were not calculated for PE wear. The sample size was based on the primary migration outcome ment of the bearing against the vertical wall. (unpublished data). A detailed description of all sub-study purposes is given at ClinicalTrials.gov (NCT00679120). Received allocated intervention n = 55

Received allocated intervention n = 25

Patients and methods Participants This patient- and observer-blinded multicenter RCT was carried out at 2 hospitals in Denmark (Aarhus University Hospital and Vejle Hospital). Patients with knee pain and isolated medial compartment osteoarthritis in telos-stress radiographs and otherwise suitable for treatment with Oxford medial UKA were assessed for eligibility to participate. The surgeons enrolled all patients. Inclusion criteria were medial compartment osteoarthritis, patients > 18 years old, and informed consent. Exclusion criteria are outlined in Table 1 (see Supplementary data). 163 consecutive patients were evaluated for eli-

RSA examinations RSA examinations were taken immediately postoperatively, and at 6, 12, 24, and 60 months of follow-up. Double examinations for assessment of precision were conducted on all patients at the 6 months follow-up in accordance with the ISO 2013 RSA standards (Table 3, see Supplementary data). During the initial study period (2009–2014), examinations were performed with the Arcoma system (Arco Ceil model 0070-S, Växjö, Sweden). After January 2014, a direct digital dedicated stereo X-ray system was used (AdoraRSAsuite, NRT, Aarhus, Denmark). For both RSA systems, the X-ray tubes were positioned horizontally with an angle of

Acta Orthopaedica 2019; 90 (1): 67–73

mJSW

Figure 2. This figure outlines the set-up of the RSA examinations. To standardize the set-up, a rectangular foam support was applied between the leg and calibration-box. Patients were also asked to move their center of gravity over the operated leg to make sure that the prosthetic knee joint was loaded.

40° between tubes, and a source-to-image-detector distance of 160 cm. The uniplanar calibration-box (CarbonBox 14, Medis Medical Imaging Systems, Leiden, Netherlands) was equipped with 2 digital image detectors: the Adora system used wireless CXDI-70C detectors (Canon, Tokyo, Japan) and the Arcoma system used FCR Profect CS detectors (Fujifilm, Vedbaek, Denmark). During the examinations, patients were standing and facing the calibration box, with the knee loaded and slightly flexed (10–20°) (Figure 2). The resolution of the RSA images was 203 pixels per inch. Model-based RSA analysis The stereoradiographs were analyzed with Model-based RSA (Version 4.01, RSAcore, Leiden, Netherlands) (Kaptein et al. 2003). Wear was evaluated from the minimal joint space width (mJSW). The mJSW is an indirect measure of the bearing thickness, as it defines the shortest distance between the spherical femoral component perpendicular to the flat tibial component (Figure 3) (Van IJsseldijk et al. 2011, Horsager et al. 2018). The relative difference in the measured mJSW over the 5-year follow-up represents the combined upper- and lower-surface PE wear. Bearing overhang and impingement against the vertical wall were approximated using the femorotibial contact point, as the bearing is not visible on the stereoradiographs. The femorotibial contact point was defined from the projected mJSW line and reflects the center position of the bearing (Figure 3). This is allowed because of the spherical design of the femoral component and the fully congruent bearing. Overhang was noted when the distance between the femorotibial contact point and the most medial, anterior, or posterior edge of the tibial component was less than the corresponding dimensions of the bearing. Impingement against the vertical wall was noted when the mediolateral distance between the femorotibial con-

mJSW

Figure 3. The frontal view to the left outlines the measurement of the minimal joint space width (mJSW). The dotted line represents the sagittal cross-sectional view presented to the right. The mJSW reflects the bearing thickness and the projected dotted line on the tibial component represents the femorotibial contact point. This allows the estimation of PE wear and the position of the bearing. Overhang is seen when the bearing exceeds the outline of the tibial plateau. Impingement can be identified if the bearing slides against the vertical wall (see frontal view).

tact point and the vertical wall was less than half the width of the bearing (Horsager et al. 2018). The measurements assume that the bearing is kept parallel with the vertical wall, and that no rotation occurs. Bearing overhang and impingement were only used for further analysis if this exceeded the precision for the corresponding contact-point location (Table 3, see Supplementary data). For each patient, the measured maximal bearing overhang at all follow-up times was used in the statistical analysis. If impingement was measured, it would indicate an error-full mJSW, as it would imply lift-off. Any mJSW measurements with impingement would be excluded in the statistical analysis. Clinical outcome Clinical outcome was evaluated using the Oxford Knee Score (OKS). OKS scores for baseline, 5-year follow-up, and the difference between baseline and 5-year follow-up (D OKS) are presented (Table 4, see Supplementary data). Statistics A linear mixed model for repeated measurements was used to evaluate the wear-rate based on the mJSW measurements from baseline to 5-year follow-up using a linear relationship. 2 models were computed: (1) a crude model including the fixedeffects of fixation type (cemented vs. cementless), bearing overhang and their appropriate interaction with time, and (2) an “adjusting” model adding the fixed effects of sex, baseline weight, and D OKS. Included interaction terms with time were justified by visual inspection of the plotted model residuals against the interaction variable. Interaction terms for “fixation type and time” and “medial bearing overhang and time” were included. Patient ID was included as a random factor and the development over time was identified by the random slope. Model diagnostics

Acta Orthopaedica 2019; 90 (1): 67–73

Mean PE wear (mm) 0.5

Patients

25 Cemented

Cemented Cementless

Cementless

0.4

0.3

0.2

0.1

0.0

Follow−up (years) Figure 4. The linear wear-rate of the cemented and cementless Oxford UKA with 95% CI from the linear mixed model.

1−2

2–3

3–4

4–5

5–6

1−2

2–3

3–4

4–5

5–6

Medial overhang (mm) Posterior overhang (mm) Figure 5. Distribution plots for the extent of bearing overhang. Medial overhang is shown to the left and posterior overhang is shown to the right.

were validated by visual inspection of residuals and fitted values. Gaussian distributions were checked using QQ plots. P-values were derived from the model output, which allowed a test of the main null hypothesis and sub-hypothesis. Sex, weight, and clinical outcome were not tested for significance. The effect of the adjusting variables was evaluated from the change in wear-rate between the crude model and adjusting model. A 2-sample Satterthwaite t-test was used to compare the D OKS. The 95% confidence interval (CI) was calculated for all wear-rate measurements and all OKS scores. The statistics were performed in collaboration with the Biostatistical Advisory Service at Aarhus University, Denmark. P-values < 0.05 were regarded as statistically significant. Graphs and analysis were generated using Stata 14.0 (StataCorp LP, College Station, TX, USA). Ethics, registration, funding, and potential conflict of interest Approvals were obtained from the local ethics committee (M-20070258; d. 15/01/2008) and Data Protection Agency (2008-41-2104; d. 28/03/2008) and the study was conducted in agreement with the Helsinki II declaration. The study was registered at ClinicalTrials.gov (NCT00679120). The study was financially supported by Biomet Inc. The authors have no conflict of interest to declare.

mal effect on the wear-rate, as the adjusting model produced equivalent results: no change was seen for the cementless design and the wear-rate of the cemented design increased to 0.05 mm/year (CI 0.02–0.07). There was no statistically significant difference between the cemented and cementless fixation method (p = 0.6) (Figure 4). An extreme patient outlier was identified from the residuals, and a sensitivity analysis excluding the outlier from the model generated the same results. Also, a sub-analysis showed equal wear-rates for the single- and double-pegged design for the cemented Oxford medial UKA. Bearing overhang ranging from 1 to 5 mm was seen in 41 patients on the medial side, mean 2.5 mm (1–5 mm) and in 14 patients for the posterior edge of the tibial component, mean 2 mm (1–5 mm) (Figures 5 and 6). None of the patients had anterior bearing overhang or impinged the bearing against the vertical wall. The effect of medial bearing overhang was statistically significant, and the wear-rate increased by 0.014 mm/year (CI 0.004–0.025) for each mm increment of overhang (p = 0.01), resulting in wear-rates ranging from 0.064 to 0.12 mm/year. This did not change after adjusting for weight, sex, and clinical outcome. The effect of posterior overhang on wear-rate was not included in the model, as there was no sign of interaction with time in the plotted residuals. The D OKS scores were similar between the cemented and cementless group (Table 4, see Supplementary data).

Results The mean wear-rate was 0.04 mm/year (CI 0.02–0.07) for the cemented design and 0.05 mm/year (CI 0.02–0.08) for the cementless design, if no bearing overhang or impingement was measured (crude model). The mean difference in wear-rate was –0.008 mm/year (CI –0.04 to 0.03). Sex, weight, and clinical outcome (adjusting variables) had mini-

Discussion To our knowledge, this is the first study to compare PE wear for cemented and HA-coated cementless knee arthroplasties. The concern of HA coating as a promoter of PE wear is mostly derived from studies of THAs (Bloebaum et al. 1994, 1997,

Acta Orthopaedica 2019; 90 (1): 67–73

Overhang (%)

Case 1

Overhang (%)

Case 2

Mean position + 95%PI

Overhang (%)

1.0

0.5

Bearing Tibia

0.0

Posterior

1.0

0.5

Bearing 0.0

Posterior

Tibia Anterior

Bearing 0.0

Posterior

Tibia Anterior

Figure 6. This figure visualizes a “top view” of the PE bearing and tibial component of the Oxford Medial UKA. The figures are based on the assumption that the bearing is kept parallel with respect to the vertical wall. Extreme cases of bearing overhang are visualized in Case 1 and 2. The mean bearing position for all patients with 95% prediction interval is visualized in the right panel. Case 1 represents the patient with 5 mm medial overhang and case 2 represents the patient with 5 mm posterior bearing overhang. The graphs are computed using the femorotibial contact point and the exact ratio between the size of the bearing and tibial component.

Morscher et al. 1998, Røkkum and Reigstad 1998, Stilling et al. 2009, Gottliebsen et al. 2012). Our results contradict this expectation, as the cemented and cementless Oxford medial UKA presented similar wear-rates of 0.04–0.05 mm/year. Another finding was the increase in wear-rate ranging from 0.064 to 0.12 mm/year for half of the patients due to medial bearing overhang. HA may only cause third-body wear if it enters the joint. This is hypothesized to happen during insertion, from chemical dissolution and from mechanical abrasion of the coating due to loosening of the implant and lack of initial stability (Morscher et al. 1998, Røkkum and Reigstad 1998, Røkkum et al. 1999, Duffy et al. 2004). Full osseointegration is believed to limit the dissociation of HA-particles to the joint and protect the fixation-interface from wear debris—a phenomenon called “Sealing-effect” (Kadoya et al. 1998, Morscher et al. 1998, Rahbek et al. 2005). The HA-coated parts of the Phase 3 Oxford medial UKA were completely covered by bone and the fixation interface stabilizes within the first 6 months (Kendrick et al. 2015). These are optimal conditions for a successful sealing effect. It may be speculated that only little force is applied to the HA coating during implantation of an Oxford UKA. This should limit the risk of initial delamination of the HA coating compared with the forces generated for the insertion of a press-fit cup. Moreover, the fully congruent mobile bearing causes compressive loads and limits shear forces, which may help reduce debonding of the HA coating before full osseointegration (Müller and Patsalis 1997, Horsager et al. 2017, 2018). Overall, the Oxford medial UKA exhibit several features that in theory reduce the dissociation of HA particles to the joint, which thus limits the potential of third-body wear. This could explain similar wear-rates for both fixation methods. However, the 95% CI of the mean difference in wear-rate was relatively large and cannot exclude all clinically relevant differences in wear-rates, as no threshold value for wear has been established. The mean wear-rate for the cemented Oxford UKA (with phase 1 and 2 bearings) is well documented and ranges from

0.026 mm/year to 0.07 mm/year (Argenson and O’Connor 1992, Psychoyios et al. 1998, Price et al. 2005, Kendrick et al. 2010a, 2010b). This is comparable to our results (phase 3 bearings). However, Kendrick et al. (2010b) and Psychoyios et al. (1998) found that retrieved bearings with no sign of abnormal macroscopic wear or impingement had very low wearrates of 0.01 mm/year. It was concluded that well-functioning bearings, which move freely with no impingement against surrounding structures, such as osteophytes or the implant itself, should achieve wear-rates < 0.03 mm/year (Kendrick et al. 2010b). To account for this, we approximated whether the bearings were “well-functioning,” by measuring if the bearings impinged against the vertical wall or exceeded the outline of the tibial plateau (bearing overhang). Bearing overhang does not necessarily reflect a poorly functioning bearing. Yet it does lower the surface contact area and increase the risk of edge-loading and impingement (Figure 6). The measured wear-rate of 0.05 mm/year for the cementless design and 0.04 mm/year for the cemented design (no overhang or impingement) represents a “well-functioning” bearing. This is somewhat higher than the expected wear-rate of < 0.03 mm/year. However, it does not statistically violate the statement given the 95% confidence interval. None of the implants showed impingement. This could be due to the standing RSA examinations, as impingement could be induced by knee motion (Horsager et al. 2018). Medial bearing overhang ranging from 1 to 5 mm was measured in half of the patients. We found an additional increase in wear-rate of 0.014 mm/year for each mm increment in overhang, which leads to a relatively high wear-rate ranging from 0.064 to 0.12 mm/year (Figure 5). This is in line with Kendrick et al. (2010b) and Psychoyios et al. (1998), who found wear-rates > 0.1 mm/year, if the retrieved bearings had signs of impingement. The measured posterior bearing overhang could be expected to have the same effect, although this was not evident. To our knowledge, this is the first study to report the presence, extent, and effect of bearing overhang. The finding outlines the importance of component alignment and careful surgical technique.

The use of the femorotibial contact point as an indirect measure of impingement and bearing overhang raises some limitations. The method assumes that the bearing is kept in the center of the femorotibial contact point. Rotation of the bearing and impingement from osteophytes or soft tissue is undetectable. Still, this is probably the best we can achieve as the bearing is not visible on the stereoradiographs. The measurements were further limited from the standing RSA examinations, as the mobile bearing moves throughout motion (Horsager et al. 2018). Some concern has been addressed with the risk of overestimation of wear with shorter follow-up periods, as the PE bearing is stated to creep in the first 6 months (Glyn-Jones et al. 2008, Kendrick et al. 2010b). We did not identify initial creep of the bearing, as the residuals for the wear measurements did not conflict with the linear model. This does not imply that creep does not occur—only that it is probably insignificant. The proportion of loss to follow-up was equal for the cemented and cementless group and below the critical limit of 20%, which minimizes the risk of bias (Dettori 2011). The bias should be further minimized from the linear mixed model, as it handles missing data effectively and allowed the use of all available data (Krueger and Tian 2004). Sample size and power were not calculated, since wear of the PE bearing was a secondary effect parameter and therefore the results can be a type II error. Our study is also susceptible to multiplicity issues, as several effect parameters have been studied in the RCT. However, this is the largest PE wear study performed on the Oxford medial UKA (Argenson and O’Connor 1992, Psychoyios et al. 1998; Price et al. 2005, Kendrick et al. 2010a, 2010b). If a difference exists in PE wear between wellfunctioning cemented and cementless Oxford medial UKAs, it would be small and most likely clinically insignificant. The patient’s weight, sex, and clinical outcome did not seem to influence the wear-rate. This may indicate that obese patients or a poor clinical outcome are not associated with high wear-rates in the UKA. We believe our study provides solid data for the wear-rate of clinically successful Oxford medial UKAs, due to the precise loaded state-of-the-art in-vivo RSA measurements and sufficient follow-up time of 5 years. The results may support the continued use of the cementless HA-coated Oxford medial UKA and indicate that the most significant contributing factor for PE wear is based on the kinematics and position of the PE bearing. In addition, although the Oxford medial UKA design is ingenious, it is technical demanding and may require more strict alignment for optimal conditions. Furthermore, our results indicate that the wear-rate of the Oxford medial UKA is generally higher than expected and may approach linear wear-rates measured for non-congruous fixed-bearing UKAs, which have been reported as 0.15 mm/ year for the St George Sled UKA (Ashraf et al. 2004). Volumetric wear could be comparable due to the fully congruent design of the Oxford medial UKA (Burton et al. 2012).

Acta Orthopaedica 2019; 90 (1): 67–73

In summary, we found similar wear-rates for the cemented and HA-coated cementless Oxford medial UKA in a randomized study design, which supports continued use of the cementless Oxford medial UKA. However, a caveat is a relatively large 95% CI of the mean difference in wear-rate. Half of the patients presented with an additional increase in wearrate due to medial bearing overhang, which emphasizes the importance of careful component positioning. Overall, the measured wear-rate was low, but slightly higher than expected for clinically successful Oxford medial UKA—especially if medial bearing overhang was present. Supplementary data Tables 1, 3, and 4 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/ 17453674.2018.1543757 KS, MS, AO, FM, and PWK designed the study. FM, AO, CFJ, PWK, LR, and MS collected the data. KH, MS, and BLK analyzed the data and performed the statistics. KH, BLK, MS, FK, PWK, and AO helped with data interpretation. KH wrote the initial manuscript draft. All authors revised and approved the final manuscript. The authors thank orthopedic surgeons Dan Blohm, Bjørn Gottlieb, and Anders Lamberg for surgeries on some of the patients. They thank Rikke Mørup, Lone Løvgren Andersen, and Henriette Appel for their valued assistance in patient management and data collection. ZimmerBiomet, the Danish Rheumatoid Association, and the A.P. Møller Foundation kindly provided financial support for data analyses. Acta thanks Stephan Maximilian Röhrl and Leif Ryd for help with peer review of this study. Argenson J N, O’Connor J J. Polyethylene wear in meniscal knee replacement: a one- to nine-year retrieval analysis of the Oxford knee. J Bone Joint Surg Br 1992; 74(2): 228-32. Ashraf T, Newman J H, Desai V V, Beard D, Nevelos J E. Polyethylene wear in a non-congruous unicompartmental knee replacement: a retrieval analysis. Knee 2004; 11: 177-81. Bloebaum R D, Beeks D, Dorr L D, Savory C G, DuPont J A, Hofmann A A. Complications with hydroxyapatite particulate separation in total hip arthroplasty. Clin Orthop Relat Res 1994; (298): 19-26. Bloebaum R, Zou L, Bachus K, Shea K, Hofmann A, Dunn H. Analysis of particles in acetabular components from patients with osteolysis. Clin Orthop Relat Res 1997; (338): 109-18. Burton A, Williams S, Brockett C L, Fisher J. In vitro comparison of fixed- and mobile meniscal-bearing unicondylar knee arthroplasties: effect of design, kinematics, and condylar liftoff. J Arthroplasty 2012; 27(8): 1452-9. Dettori J R. Loss to follow-up. Evid Based Spine Care J 2011; 2(1): 7-10. Duffy P, Sher J L, Partington P F. Premature wear and osteolysis in an HA-coated, uncemented total hip arthroplasty. J Bone Joint Surg Br 2004; 86B: 34-8. Glyn-Jones S, McLardy-Smith P, Gill H S, Murray D W. The creep and wear of highly cross-linked polyethylene. J Bone Joint Surg Br 2008; 90-B: 556-61. Gottliebsen M, Rahbek O, Ottosen P F, Søballe K, Stilling M. Superior 11-year survival but higher polyethylene wear of hydroxyapatite-coated Mallory-Head cups. Hip Int 2012; 22(1): 35-40. Hallan G, Lie S A, Havelin L I. High wear rates and extensive osteolysis in 3 types of uncemented total hip arthroplasty: a review of the PCA, the Harris Galante and the profile/tri-lock plus arthroplasties with a minimum of 12 years median follow-up in 96 hips. Acta Orthop 2006; 77(4): 575-84.

Acta Orthopaedica 2019; 90 (1): 67–73

Harris W H. Wear and periprosthetic osteolysis: the problem. Clin Orthop Relat Res 2001; (393): 66-70. Horsager K, Kaptein B L, Rømer L, Jørgensen P B, Stilling M. Dynamic RSA for the evaluation of inducible micromotion of Oxford UKA during step-up and step-down motion. Acta Orthop 2017; 88(3): 275-81. Horsager K, Kaptein B L, Jørgensen P B, Jepsen C F, Stilling M. Oxford medial unicompartmental knees display contact-loss during step-cycle motion and bicycle motion: a dynamic radiostereometric study. J Orthop Res 2018; 36(1): 357-64. ISO. Implants for surgery—Roentgen stereophotogrammetric analysis for the assessment of migration of orthopedic implants; 2013. Kadoya Y, Kobayashi A, Ohashi H. Wear and osteolysis in total joint replacements. Acta Orthop 1998; (Suppl 278): 1-16. Kaptein B L, Valstar E R, Stoel B C, Rozing P M, Reiber J H C. A new modelbased RSA method validated using CAD models and models from reversed engineering. J Biomech 2003; 36(6): 873-82. Kendrick B J L, Longino D, Pandit H, Svard U, Gill H S, Dodd C A F, Murray D W, Price A J. Polyethylene wear in Oxford unicompartmental knee replacement: a retrieval study of 47 bearings. J Bone Joint Surg Br 2010a; 92-B(3): 367-73. Kendrick B J L, Simpson D J, Kaptein B L, Gill H S, Murray D W, Price A J. Polyethylene wear of mobile-bearing unicompartmental knee replacement at 20 years. J Bone Jt Surg Br 2010b; 93-B: 470-5. Kendrick B J L, Kaptein B L, Valstar E R, Gill H S, Jackson W F M, Dodd C A F, et al. Cemented versus cementless Oxford unicompartmental knee arthroplasty using radiostereometric analysis: a randomised controlled trial. Bone Joint J 2015; 97-B(2): 185-91. Kim S-Y, Kim D-H, Kim Y-G, Oh C-W, Ihn J-C. Early failure of hemispheric hydroxyapatite-coated acetabular cups. Clin Orthop Relat Res 2006; (446): 233-8. Krueger C, Tian L. A comparison of the general linear mixed model and repeated measures ANOVA using a dataset with multiple missing data points. Biol Res Nurs 2004; 6(2): 151-7. Liddle A D, Pandit H, O’Brien S, Doran E, Penny I D, Hooper G J, Burn P J, Dodd C A F, Beverland D E, Maxwell A R, Murray D W. Cementless fixation in Oxford unicompartmental knee replacement: a multicentre study of 1000 knees. Bone Joint J 2013; 95-B(2): 181-7. Lonner J, Siliski J, Richard D. Prodromes of failure in total knee arthroplasty. J Arthroplasty 1999; 14(4): 488-92. Morscher E W, Hefti A, Aebi U. Severe osteolysis after third-body wear due to hydroxyapatite particles from acetabular cup coating. J Bone Joint Surg 1998; 80-B(2): 267-72. Müller R T, Patsalis T. Shear and tensile strength of hydroxyapatite coating under loading conditions: an experimental study in dogs. Arch Orthop Trauma Surg 1997; 116: 334-7.

Naudie D D R, Ammeen D J, Engh G A, Rorabeck C H. Wear and osteolysis around total knee arthroplasty. J Am Acad Orthop Surg 2007; 15(1): 53-64. O’Connor J J, Goodfellow J W. Theory and practice of meniscal knee replacement: designing against wear. J Eng Med 1996; 210(3): 217. Pandit H, Jenkins C, Beard D J, Gallagher J, Price A J, Dodd C A F, Goodfellow J W, Murray D W. Cementless Oxford unicompartmental knee replacement shows reduced radiolucency at one year. J Bone Joint Surg Br 2009; 91(2): 185-9. Price A J, Short A, Kellett C, Beard D, Gill H, Pandit H, Dodd C A F, Murray D W. Ten-year in vivo wear measurement of a fully congruent mobile bearing unicompartmental knee arthroplasty. J Bone Joint Surg Br 2005; 87-B(11): 1493-7. Psychoyios V, Crawford R W, O’Connor J J, Murray D W. Wear of congruent meniscal bearings in unicompartmental knee arthroplasty. J Bone Joint Surg Br 1998; 80-B: 976-82. Rahbek O, Kold S, Bendix K, Overgaard S, Søballe K. Superior sealing effect of hydroxyapatite in porous-coated implants: experimental studies on the migration of polyethylene particles around stable and unstable implants in dogs. Acta Orthop 2005; 76(3): 375-85. Røkkum M, Reigstad A. Polyethylene wear with an entirely HA-coated total hip replacement. Acta Orthop Scand 1998; 69(3): 253-8. Røkkum M, Brandt M, Bye K, Hetland K R, Waage S, Reigstad A. Polyethylene wear, osteolysis and acetabular loosening with an HA-coated hip prosthesis: a follow-up of 94 consecutive arthroplasties. J Bone Joint Surg Br 1999; 81(4): 582-9. Sharkey P F, Hozack W J, Rothman R H, Shastri S, Jacoby S M. Why are total knee arthroplasties failing today? Clin Orthop Relat Res 2002; (404): 7-13. Stilling M, Rahbek O, Søballe K. Inferior survival of hydroxyapatite versus titanium-coated cups at 15 years. Clin Orthop Relat Res 2009; 467(11): 2872-9. Stilling M, Søballe K. clinicaltrials.gov/ct2/show/NCT00679120?term = NCT00679120&rank = 1 [Internet]. [cited 2017 Sep 14]. Available from: http://www.clinicaltrials.gov Sundberg M, Lidgren L, W-Dahl A, Robertsson O. Swedish Knee Arthroplasty Register Annual Report, 2016. Available from: http://www.myknee.se Tai C C, Cross M J. Five- to 12-year follow-up of a hydroxyapatite-coated, cementless total knee replacement in young, active patients. J Bone Joint Surg Br 2006; 88(9): 1158-63. Van IJsseldijk E A, Valstar E R, Stoel B C, Nelissen R G H H, Reiber J H C, Kaptein B L. The robustness and accuracy of in vivo linear wear measurements for knee prostheses based on model-based RSA. J Biomech 2011; 44(15): 2724-7. Van Ijsseldijk E A, Harman M K, Luetzner J, Valstar E R, Stoel B C, Nelissen R G H H, et al. Validation of a model-based measurement of the minimum insert thickness of knee prostheses. Bone Joint Res 2014; 3(10): 289-96.

Acta Orthopaedica 2019; 90 (1): 74–80

Predictors of return to desired activity 12 months following unicompartmental and total knee arthroplasty Alexander D HARBOURNE 1, Maria T SANCHEZ-SANTOS 2, Nigel K ARDEN 2,3, and Stephanie R FILBAY 2, on behalf of the COASt Study Group 1 Nuffield

Department of Orthopaedics, Rheumatology & Musculoskeletal Sciences, University of Oxford, Botnar Research Centre, Oxford, UK; 2 Arthritis Research UK Centre for Sport, Exercise and Osteoarthritis, Nuffield Department of Orthopaedics, Rheumatology & Musculoskeletal Sciences, University of Oxford, Botnar Research Centre, Oxford, UK; 3 MRC Lifecourse Epidemiology Unit, University of Southampton, UK Correspondence: stephanie.filbay@uq.net.au Submitted 2018-06-01. Accepted 2018-09-28.

Background and purpose — 1 in 5 patients are dissatisfied following unicompartmental or total knee arthroplasty (UKA or TKA). This may be partly explained by failing to return to desired activity post-arthroplasty. To facilitate return to desired activity, a greater understanding of predictors of return to desired activity in UKA and TKA patients is needed. We compared rates of return to desired activity 12 months following UKA vs. TKA, and identified and compared predictors of return to desired activity 12 months following UKA vs. TKA. Patients and methods — Patients were prospectively recruited from 2 hospitals prior to undergoing UKA or primary TKA. Patients reported preoperatively the activity/ activities that were limited due to their knee that they wished to return to after arthroplasty. At 12-months postoperatively, patients reported whether they had returned to these activities (‘return to desired activity’). Preoperative predictors evaluated were age, sex, BMI, education, comorbidities, pain expectations, Oxford Knee Score (OKS), UCLA Activity Score, and EQ-5D. Generalized linear models assessed the relationship between potential predictors and return-todesired-activity. Results — The response rate of all patients eligible for 12-month follow-up was 74%. TKA patients (n = 575) were older (mean (SD) 70 (9) vs. 67 (10)) with a greater BMI (31 (6) vs. 30 (5)) than patients undergoing UKA (n = 420). 75% of UKA and 59% of TKA patients returned to desired activity. TKA patients had a greater risk of non-return to desired activity than patients undergoing UKA (risk ratio (95% CI) 1.5 (1.2–1.8)). Predictors of non-return to desired activity following UKA were worse OKS (0.96 (0.93–0.99)), higher BMI (1.04 (1.01–1.08)), and worse expectations (1.9 (1.2– 2.8)). Predictors of non-return to desired activity following TKA were worse EQ-5D (0.53 (0.33–0.85)) and worse OKS (0.98 (0.96–1.0)).

Interpretation — UKA patients were more likely to return to desired activity than TKA patients. Predictors of return to desired activity differed following UKA and TKA. Optimizing selection of arthroplasty procedure based on patient characteristics and targeting predictors of poor outcome may facilitate return to desired activity with potential to enhance postoperative satisfaction.

As many as 20% of patients are dissatisfied with unicompartmental knee arthroplasty (UKA) (Von Keudell et al. 2014, Kim et al. 2017) and total knee arthroplasty (TKA) outcomes (Scott et al. 2010). This presents a significant concern considering that the number of both UKA and TKA procedures is rising in Europe and North America (Kurtz et al. 2007, Leskinen et al. 2012). 1 in 5 patients who receive TKA have isolated unicompartmental OA that could be treated by either procedure (Arno et al. 2011). Selecting the most appropriate knee arthroplasty procedure in line with a patient’s preoperative characteristics has the potential to improve postoperative satisfaction, function, and participation in desired activities. Selection of one surgical technique over the other is frequently based on traditional criteria, including that UKA patients should be over 60 years of age at the time of operation, not obese, not extremely physically active, have minimal knee pain at rest, and an adequate range of motion (Kozinn and Scott 1989, National Imaging Associates Inc. 2015). This may be due to a scarcity of literature that directly compares predictors of outcome following UKA and TKA procedures. It is pertinent to assess postoperative outcomes that are of relevance and importance to patients. Patients who undergo UKA have a greater likelihood of returning to sport compared

Acta Orthopaedica 2019; 90 (1): 74–80

with TKA patients (Witjes et al. 2016). However, patients undergoing these procedures may desire participation in activities with contrasting demands, and it is not clear if a similar relationship exists with regards to returning to an individual’s desired activity. Improved function in desired activities that were limited preoperatively is expected by as many as 72% of knee arthroplasty patients (Nilsdotter et al. 2009). Failure to meet expectations regarding participation in desired activities is a key determinant of postoperative dissatisfaction following knee arthroplasty (Bourne et al. 2010, Scott et al. 2012). Identifying patient characteristics that predict return to desired activity following UKA and TKA may inform targeted preoperative interventions with potential to improve postoperative patient satisfaction. Therefore, we compared rates of return to desired activity 12 months following UKA vs. TKA, and identified and compared predictors of return to desired activity 12 months following UKA vs. TKA.

Patients and methods The Clinical Outcomes in Arthroplasty Study (COASt) The data for this analysis were collected through the Clinical Outcomes in Arthroplasty Study (COASt), a prospective, longitudinal cohort study based at 2 UK hospitals, Nuffield Orthopaedic Centre (NOC), Oxford and Southampton University Hospital NHS Foundation Trust (UHS) (Arden et al. 2017). Knee arthroplasty procedures were performed from 2010 to 2016. Patient-reported outcomes were collected at baseline (pre-operation) and at 12 months following knee arthroplasty. Recruitment and procedure Patients were recruited from waiting lists for hip or knee arthroplasty from both hospitals. To be eligible for COASt, patients were required to (i) be over 18 years of age at the time of recruitment, (ii) be competent and willing to consent to partake in the study, and (iii) not show signs of any severe neurological disorder. A recruitment pack was sent to potentially eligible patients. Contact was subsequently made 2 weeks later to ascertain eligibility and a verbal desire to participate. If both were met, a research appointment was arranged to obtain baseline measurements and written consent. Baseline data were collected through questionnaires and a physical examination performed by a research nurse, physiotherapist, or podiatrist. Postoperatively, patients received follow-up questionnaires that were completed by post or during an appointment. To be considered eligible for the current study, patients had to have undergone a UKA or primary TKA, and completed the 12-month follow-up questionnaire by post. Patients who had a hip arthroplasty, a revision TKA, or more than 1 arthroplasty on their knees were ineligible. As of May 2017, 1,491 individuals had undergone UKA or primary TKA. Patients who died (n =

Patients who underwent primary UKA or TKA n = 1,491 Excluded (n = 60): – died before 12-month follow-up, 36 – underwent a second UKA or TKA within the follow-up period, 24 Patients eligible for 12-month follow-up n = 1,431 Excluded Did not complete 12-month follow-up n = 372 Patients who participated in 12-month follow-up n = 1,059 Excluded (n = 64): – completed only telephone follow-up at 12 months but not by post, 38 – did not complete the return-to-desiredactivity item at baseline and/or 12 month follow-up, 26 Patients included in the analysis n = 995

Participant recruitment flow chart. UKA: unicompartmental knee arthroplasty; TKA: total knee arthroplasty.

36) or had a second UKA or TKA on either knee (n = 24) within the follow-up period were excluded (Figure). 1,431 patients were eligible to complete the 12-month follow-up questionnaire, of which 372 did not respond (response rate = 74%). In addition, patients who completed the follow-up administered over the telephone (n = 38) and those who had missing data on the return to desired activity variable at baseline and/or 12-month follow-up (n = 26) were excluded. Data from 995 patients were included in this study (Figure). Indications for knee arthroplasty (not reported for 15% of participants) were OA (81%), rheumatoid arthritis (2%) and other reasons (2%). Outcome—return to desired activity Patients were asked at baseline: “What activity/activities does your knee stop or limit you from doing, that you wish to return to after your operation?” Patients were then given the option to report between 1 and 3 activities. Then at 12 months post-operation patients were asked: “Have you been able to return to the activity (or activities) that your knee stopped you from doing 12 months ago? (Yes/No).” Since not returning to desired activity (answering “No” to the above question) was the more infrequent outcome, predictors were assessed in relation to non-return to desired activity. Potential predictor variables Predictors used in this study were chosen following our review of the literature and clinical reasoning. To minimize

over-adjustment and collider stratification biases, the selected variables were incorporated into directed acyclic graphs (DAGs; useful to depict assumed relationships between variables). They were performed in reference to a 6-step process to achieve unbiased estimates (Shrier and Platt 2008). The variables used in the DAGs are described below. Patient demographics Patient demographics included as potential predictors in this analysis were: the age at operation; sex; BMI; highest education level attained (dichotomized to “did not complete General Certificate of Secondary Education (GCSE) or above” and “completed GCSE or above”). The GCSE is a qualification in England, Wales, and Northern Ireland taken in a specific subject by school students typically aged 14–16 years. Baseline Oxford Knee Score (OKS) The OKS is a 12-item questionnaire designed to assess knee pain and function, with each item scored between 0 (worst outcome) and 4 (best outcome) (Dawson et al. 1998). Items are summated to give a total OKS score (possible range of 0–48) (Murray et al. 2007). A lower baseline OKS score suggests more pain and worse preoperative knee function. The OKS has adequate test–retest reliability, good sensitivity and responsiveness to change, and is valid for use in UKA and TKA populations (Collins et al. 2011, Jenny and Diesinger 2012). Mean substitutions were carried out in line with guidelines to replace missing items in OKS data (Murray et al. 2007). University of California, Los Angeles Activity Score (UCLAAS) The UCLA-AS is a patient-reported measure that classifies a patient’s activity into 1 of 10 levels, where level 1 represents the lowest activity level (inactive and dependent on others) and level 10 represents the highest activity level (regular participation in contact sports). It is valid for clinical assessment of routine activity following knee arthroplasty (Zahiri et al. 1998, Fisher et al. 2006). EuroQol-5D-3L (EQ-5D) The EQ-5D is a self-reported measure of general health status. It comprises 5 questions relating to a patient’s mobility, selfcare, usual activities, pain/discomfort, and anxiety/depression. Each has 3 levels of response and an overall 5-part profile is calculated using weightings outlined in EuroQol Group guidelines (EuroQol Group 1990). Index values (between –1 and 1) were then generated using value sets from the United Kingdom (Ramos-Goni and Rivero-Arias 2011). It is sufficiently valid and reliable for use in patients with knee OA (Fransen and Edmonds 1999). Comorbidities Patients were asked to give “yes/no” responses to whether they had ever had any of the following comorbidities: osteo-

Acta Orthopaedica 2019; 90 (1): 74–80

porosis, gout, hypertension, stroke, heart attack, heart failure, high cholesterol, diabetes, and renal, bowel, lung, or liver problems. The number of comorbidities for each patient was assessed and divided into 2 categories for the analysis (no comorbidities vs. ≥ 1 comorbidity). Postoperative pain expectation Patients responded to the following question: “Overall, how much do you expect that pain in your knee will interfere with your life 12 months after surgery?” by selecting an option from a 5-point Likert scale (Not at all, Mildly, Moderately, Severely, Extremely). Due to few patients expecting severe/extreme (n = 6, 0.8%) or moderate (n = 53, 7.2%) pain, this variable was dichotomized to “not at all” vs. “mildly to extremely”. Statistics To determine differences between patient characteristics in UKA and TKA groups, unequal t-tests, chi-squared tests or Mann–Whitney U-tests were performed, as appropriate. Linearity of the effect between continuous variables with return to desired activity was assessed using fractional polynomials. This is a flexible parametric method that models the relationship between exposure and outcome, enabling findings to be interpreted by readers without expertise in statistics (Royston et al. 1999). Interactions between procedure type and all other potential predictors were evaluated. Missing data were assumed to be missing at random. Multiple imputation using chained equations was performed and 50 datasets were generated on the total dataset (including statistically significant (p < 0.05) interactions) and on both UKA and TKA subgroups, separately. To assess whether the imputed data accurately reflected the raw data, distribution and associations between imputed variables and outcome were compared between the raw and imputed data. Since similar results were observed between raw and imputed datasets, we considered that imputation was appropriate. Generalized linear models (GLM) using a log link, Poisson family, and robust error variances (Zou 2004) were applied to assess the association of potential predictors on outcome. Risk ratios (RR) and the corresponding 95% confidence intervals (CI) were calculated. All analyses were completed using Stata/IC 14.1 (StataCorp, College Station TX, USA). Ethics, funding, and potential conflict of interest Ethics approval for COASt was obtained: Oxford REC A (Ethics Reference: 10/H0604/91). This article presents independent research commissioned by the National Institute for Health Research (NIHR) under its Programme Grants for Applied Research funding scheme (RP-PG-0407-10064). Prof Arden and Dr Filbay are supported by the Arthritis Research UK Centre for Sport, Exercise and Osteoarthritis (grant reference 21595). The views of the author(s) expressed in this article do not necessarily represent those of the National Health Service, the NIHR, or the Department

Acta Orthopaedica 2019; 90 (1): 74–80

Table 1. Patient characteristics at baseline and 12 months post-arthroplasty

All patients (n = 995)

Missing data

UKA (n = 420)

TKA (n = 575)

p-value a

Baseline characteristics Age at operation b 69 (9) 0 (0%) 67 (10) 70 (9) < 0.001 Female sex c 549 (55%) 0 (0%) 220 (52%) 329 (57%) 0.1 BMI b 30 (5) 5 (0.5%) 30 (5) 31 (6) 0.006 Education level c 159 (16%) Did not complete GCSE or above 385 (39%) 144 (42%) 241 (49%) 0.04 Completed GCSE or above 451 (45%) 201 (58%) 250 (51%) Baseline EQ-5D d 0.62 (0.16–0.69) 106 (11%) 0.62 (0.16–0.69) 0.59 (0.16–0.69) 0.05 Postoperative pain expectation c,e 273 (27%) No pain expected 389 (39%) 172 (56%) 217 (52%) 0.3 Mild to extreme pain expected 333 (34%) 133 (44%) 200 (48%) Baseline UCLA Activity Score d 4 (3–5) 283 (28%) 4 (3–5) 4 (3–5) 0.1 No reported comorbidities c ,f 140 (14%) 245 (25%) 75 (25%) 65 (15%) < 0.001 Baseline OKS b 20.4 (7.7) 79 (8%) 21.6 (7.6) 19.5 (7.7) < 0.001 12 months post-surgery Non-return to desired activity c,g 341 (34%) 0 (0%) 105 (25%) 236 (41%) <0.001 a

p-values for differences between UKA and TKA groups, assessed using unequal t-tests, chi-squared tests or Mann– Whitney U-tests as appropriate. b Mean (SD) c Number (%) d Median (IQR) e Postoperative pain expectation: assessed preoperatively using a 5-point Likert scale (Not at all, Mildly, Moderately, Severely, Extremely) in response to the question: “Overall, how much do you expect that pain in your knee will interfere with your life 12 months after surgery?” f No reported comorbidities: the number of patients who had not been diagnosed with any of the following comorbidities (compared with a diagnosis of 1 or more of these comorbidities): osteoporosis, gout, hypertension, stroke, heart attack, heart failure, high cholesterol, diabetes, renal, bowel, lung, and liver problems. g Non-return to desired activity: Proportion of patients that responded “No” to the following question at 12-month follow-up: “Have you been able to return to the activity (or activities) that your knee stopped you from doing 12 months ago?” UKA: unicompartmental knee arthroplasty; TKA: Total knee arthroplasty; BMI: body mass index; GCSE: General Certificate of Secondary Education; EQ-5D: EuroQol 5 dimensions questionnaire; UCLA: University of California, Los Angeles; OKS: Oxford Knee Score.

of Health. NKA is a consultant for Freshfields Bruckhaus Deringer and has received honorariums from Bioventus, Flexion, Regeneron, and grants from Bioiberica and Merck outside the submitted work. ADH, MTSS, and SRF declare no conflict of interest.

Results Patient characteristics (Table 1) Patients (55% women) were a mean 69 (SD 9) years old at the time of surgery and had a mean BMI of 30 (SD 5). 86% of patients reported ≥ 1 comorbidity. 34% of all patients did not return to desired activity 12 months after knee surgery. UKA was performed on 42% (n = 420), and TKA on 58% (n = 575) of patients. Patients who had TKA were older (70 (9) vs. 67 (10), p < 0.001), had a greater mean BMI (31 (6) vs. 30 (5), p = 0.006) and a greater percentage did not complete GCSE or above (49% vs. 42%, p = 0.04) compared with UKA patients. Patients who had UKA reported better baseline median EQ-5D scores (0.62, IQR (0.16–0.69) vs. 0.59 (0.16– 0.69), p = 0.05), better baseline mean OKS scores (22 (7.6) vs. 20 (7.7), p < 0.001), and a greater percentage reported no

comorbidities (25% vs. 15%, p < 0.001) compared with TKA patients. Return to desired activity 12 months following UKA vs. TKA (Table 1) The percentage of patients that did not return to desired activity 12 months following surgery was greater following TKA than UKA (41% vs. 25%, p < 0.001). The most common activities that patients wished to return to were similar between UKA and TKA patients; the 4 most common activities were walking (UKA 58%; TKA 57%), gardening (14%; 16%), cycling (8%; 9%), and swimming (9%; 7%). Predictors of return to desired activity 12 months following knee arthroplasty (Table 2) A 1-unit greater baseline OKS score was associated with a 3% lower risk of not returning to desired activity (RR 0.97 (0.95– 0.99)). Patients who expected some degree of pain interference with life 12 months post-arthroplasty had a 1.3 (1.1–1.7) times greater risk of not returning to desired activity compared with patients who expected no pain interference. TKA was associated with a higher risk of not returning to desired activity compared with UKA (1.5 (1.2–1.8)).

Acta Orthopaedica 2019; 90 (1): 74–80

Table 2. Multivariable analysis reporting risk ratios for non-return to desired activity following a generalized linear model for all, UKA, and TKA patients Predictors Age Sex a BMI Baseline OKS Baseline UCLA-AS Baseline EQ-5D Score Pain expectation b Education c Comorbidities d Procedure e

All patients (n = 995) RR (95% CI) p-value 1.00 (0.99–1.01) 1.15 (0.96–1.38) 1.00 (0.98–1.01) 0.97 (0.95–0.99) 1.02 (0.95–1.10) 0.71 (0.47–1.07) 1.34 (1.09–1.65) 0.94 (0.78–1.13) 1.04 (0.98–1.11) 1.47 (1.21–1.78)

UKA (n = 420) RR (95% CI) p-value

TKA (n = 575) RR (95% CI) p-value

1.0 0.99 (0.97–1.01) 0.2 1.01 (1.00–1.02) 0.09 0.1 1.43 (0.99–2.05) 0.06 1.04 (0.85–1.27) 0.7 0.7 1.04 (1.01–1.08) 0.006 0.99 (0.97–1.00) 0.1 0.001 0.96 (0.93–0.99) 0.02 0.98 (0.96–1.00) 0.04 0.6 1.07 (0.94–1.23) 0.3 1.00 (0.92–1.09) 1.0 0.1 0.91 (0.43–1.92) 0.8 0.53 (0.33–0.85) 0.008 0.005 1.86 (1.24–2.78) 0.003 1.15 (0.92–1.45) 0.2 0.5 1.08 (0.76–1.52) 0.7 0.95 (0.76–1.18) 0.6 0.2 1.08 (0.95–1.22) 0.3 1.03 (0.95–1.11) 0.5 < 0.001

All variables were included in these multivariable analyses, except for “Procedure” in the UKA and TKA analyses. Outcome coded as Returned-to-desired-activity=0, and Did-not-return-to-desired-activity=1 a Female = 1 (compared with Male = 0). b Pain expectation: “Some = 1” (mildly, moderately, severe or extremely) compared with “none = 0” (not at all) (preoperative response to the following question: “Overall, how much do you expect that pain in your knee will interfere with your life 12 months after surgery?”). c “Completed GCSE or above” = 1 (compared with “did not complete GCSE or above” = 0). d ≥ 1 comorbidity = 1 (compared with no comorbidities = 0). e TKA = 1 (compared with UKA = 0). For abbreviations, see Table1

Predictors of return to desired activity 12 months following UKA vs. TKA (Table 2) UKA — For every 1-unit greater BMI, patients who underwent a UKA had a 4% greater risk of not returning to desired activity (RR 1.04 (1.01–1.08)). UKA patients who expected some degree of postoperative pain interference had a 1.9 (1.2–2.8) times greater risk of not returning to desired activity compared with UKA patients who expected no postoperative pain interference. With every 1-unit better baseline OKS, the risk of not returning to desired activity was 4% lower (0.96 (0.93–0.99)) following UKA. Baseline EQ-5D did not predict return to desired activity following UKA. TKA — A 1-unit better baseline OKS corresponded to a 2% lower risk of not returning to desired activity following TKA (0.98 (0.96–1.00)). Better EQ-5D values before undergoing TKA were associated with a lower risk of not returning to desired activity (0.53 (0.33–0.85)). BMI, expectations, and sex did not predict return to desired activity following TKA.

Discussion A greater proportion of UKA patients returned to desired activity 12 months after arthroplasty, compared with TKA patients. For both UKA and TKA, the most common desired activities patients wished to return to were walking, gardening, cycling, and swimming. TKA patients had a 1.5 times greater risk of not returning to desired activity compared with UKA patients. Similarities and differences were found in predictors of nonreturn to desired activity between UKA and TKA patients. Better baseline OKS predicted better outcome following both

UKA and TKA. Higher BMI and worse expectations only predicted non-return to desired activity after UKA. On the other hand, worse baseline EQ-5D scores predicted non-return to desired activity following TKA, but not following UKA. Age, preoperative activity level, education level, and comorbidities were not associated with return to desired activity. We found that, irrespective of arthroplasty procedure, patients who had less preoperative knee pain and better function were more likely to return to desired activity. Less preoperative knee pain and better function have also been found to be associated with postoperative satisfaction and better OKS after both UKA and TKA (Munk et al. 2011, Judge et al. 2012, Sanchez-Santos et al. 2018). Preoperative exercise is one strategy that may be effective in reducing knee pain and improving function prior to knee arthroplasty (Wallis and Taylor 2011). Patients on the waiting list for knee arthroplasty who report a large degree of knee impairment may benefit from targeted management to improve knee pain and function. We found that a greater BMI was associated with not returning to desired activity after UKA. This is in line with Williams et al. (2012) who reported an association between greater preoperative BMI and worse activity outcomes 12 months after UKA. On the other hand, other studies found BMI did not predict revision surgery, postoperative knee pain, function, or satisfaction following UKA (Liddle et al. 2014, Burnett et al. 2014, Zuiderbaan et al. 2016). Thus, the relationship between higher preoperative BMI and post-UKA activity limitations may be explained by factors other than the clinical status of the knee (such as motivation, deconditioning, knee confidence), although this was not specifically explored in this study. Weight management in obese patients awaiting UKA

Acta Orthopaedica 2019; 90 (1): 74â&#x20AC;&#x201C;80

may improve a patientâ&#x20AC;&#x2122;s ability to return to desired activity. This may be a valuable area for future research, considering there is insufficient evidence to determine the effectiveness of short-term non-pharmacological, non-surgical weight management interventions on patient outcomes following knee arthroplasty (Lui et al. 2015). A previous study in the COASt cohort performed an indepth analysis of the relationship between pain expectations and outcome (Filbay et al. 2018). Patients who expected moderate to extreme pain interference had greater odds of being dissatisfied and not achieving a meaningful improvement on the OKS compared with those who expected no pain interference, irrespective of arthroplasty procedure (Filbay et al. 2018). However, the odds of a poor outcome (not returning to desired activity, postoperative dissatisfaction, not achieving minimally important change in OKS) in patients expecting moderate to severe postoperative pain were higher following UKA compared with TKA (Filbay et al. 2018). Further research is needed to explore and compare patient expectations between UKA and TKA procedures. Considering expectations are potentially modifiable, targeted education for preoperative patients undergoing UKA with poor expectations has potential to improve postoperative outcome (Mancuso et al. 2008, McDonald et al. 2014). Concordant with our findings, there is support for baseline EQ-5D as a predictor of TKA outcomes (Judge et al. 2012). Worse preoperative general health and the presence of anxiety or depression (assessed in the EQ-5D measure) have been found to predict less improvement in OKS following TKA (Baker et al. 2012, Hanusch et al. 2014). However, there is a need to further investigate the relationship between baseline EQ-5D scores and postoperative outcome in UKA populations. By comparing predictors of return to desired activity following UKA and TKA, our results both support and refute elements of traditional selection criteria. Our findings substantiate Kozinn and Scott (1989) who recommended that patients with a BMI below the obese category, and those with less preoperative knee pain, may be most likely to have a favorable outcome following UKA. However, in contrast to Kozinn and Scott, we found no evidence to suggest selection should be based on patient age or preoperative activity level. Other studies have also reported no association between age and other postoperative outcomes (revision rate, knee pain, stiffness or function) (Burnett et al. 2014, Lewis et al. 2015, Zuiderbaan et al. 2016, Alattas et al. 2017). Notably, traditional guidelines were not designed with reference to return to desired activity. Considering that more recent recommendations are similar to the Kozinn and Scott framework (National Imaging Associates Inc. 2015, Quinn et al. 2017), patient selection criteria for UKA vs. TKA procedures should be updated based on current evidence, with a greater focus on patient-centered outcomes. . In addition to informing treatment selection for patients with unicompartmental knee OA, our findings highlight patients at risk of poor outcome who may benefit from targeted pre-

operative intervention (e.g. preoperative rehabilitation for patients with severe knee pain and poor function, weight-loss strategies for overweight patients, and education for patients expecting a poor surgical outcome). Strengths of this study include the large sample size, comprehensive baseline data enabling control of confounders (a common limitation in previous studies), and the large number of patients who had UKA. The most common activities that UKA and TKA patients desired to return to were comparable. However, further analyses in UKA and TKA sub-groups of patients who expect to return to higher or lower levels of activity following surgery may be a fruitful area for future research. Due to limitations inherent to non-randomized studies, preoperatively UKA patients were younger, had better OKS, lower BMI, and better EQ-5D scores than TKA patients. Since these differences were accounted for in multivariable analysis, they are unlikely to explain the observed between-group differences in return to desired activity. Patients lost to follow-up represent a potential for bias and reduce the representativeness of the sample to all patients having knee arthroplasty. We did not have information regarding which knee compartments were affected by arthritis. It is possible this was related to the observed differences in return to desired activity between procedures. In summary, we found that TKA patients were less likely to return to desired activity than UKA patients. UKA patients with a high BMI, worse preoperative pain/function, and patients with worse pain expectations had a greater risk of not returning to desired activity, compared with other UKA patients. TKA patients with worse preoperative pain/function and those with a worse preoperative health status were at greater risk of not returning to desired activity compared with other TKA patients. This information may assist in identifying patients who may benefit from targeted preoperative intervention to improve surgical prognosis. All authors conceived and designed this analysis. ADH wrote the first draft of the manuscript. NKA contributed to all phases of design, obtaining funding and data collection for the COASt cohort. ADH, MTSS, and SRF analyzed and interpreted the data, edited and revised the manuscript. The authors would like to thank the individuals who participated in the COASt Study, the National Institute for Health Research (NIHR), and the Oxford Medical Sciences Division for their funding support to the study, and the rest of the COASt team for their time and dedication: D Altman, D Beard, A Carr, C Cooper, D Culliford, T Griffin, K Javaid, J Latham, D Murray, R Pinedo-Villanueva, A Price, and D Prieto-Alhambra. Acta thanks Nanne Kort and Paul Kuijer for help withÂ peer review of this study. Alattas S A, Smith T, Bhatti M, Wilson-Nunn D, Donell S. Greater pre-operative anxiety, pain and poorer function predict a worse outcome of a total knee arthroplasty. Knee Surg Sports Traumatol Arthrosc 2017; 25(11): 3403-10. Arden N, Altman D, Beard D, Carr A, Clarke N, Collins G, Cooper C, Culliford D, Delmestri A, Garden S. Lower limb arthroplasty: Can we produce a tool to predict outcome and failure, and is it cost-effective? An epidemiological study. Programme Grants for Applied Research 2017; 5 (12): 1-246.

Arno S, Maffei D, Walker P S, Schwarzkopf R, Desai P, Steiner G C. Retrospective analysis of total knee arthroplasty cases for visual, histological, and clinical eligibility of unicompartmental knee arthroplasties. J Arthroplasty 2011; 26: 1396-403. Baker P N, Deehan D J, Lees D, Jameson S, Avery P J, Gregg P J, Reed M R. The effect of surgical factors on early patient-reported outcome measures (PROMS) following total knee replacement. J Bone Joint Surg Br 2012; 94: 1058-66. Bourne R B, Chesworth B M, Davis A M, Mahomed N N, Charron K D. Patient satisfaction after total knee arthroplasty: who is satisfied and who is not? Clin Orthop Relat Res 2010; 468: 57-63. Burnett R S, Nair R, Hall C A, Jacks D A, Pugh L, McAllister M M. Results of the Oxford phase 3 mobile bearing medial unicompartmental knee arthroplasty from an independent center: 467 knees at a mean 6-year follow-up. Analysis of predictors of failure. J Arthroplasty 2014; 29: 193-200. Collins N J, Misra D, Felson D T, Crossley K M, Roos E M. Measures of knee function: International Knee Documentation Committee (IKDC) subjective knee evaluation form, Knee Injury and Osteoarthritis Outcome Score (KOOS), Knee Injury and Osteoarthritis Outcome Score Physical function Short form (KOOS-PS), Knee Outcome Survey Activities of Daily Living scale (KOS-ADL), Lysholm Knee Scoring Scale, Oxford Knee Score (OKS), Western Ontario and McMaster universities osteoarthritis index (WOMAC), Activity Rating Scale (ARS), and Tegner Activity Score (TAS). Arthritis Care Res (Hoboken) 2011; 63(Suppl 11): 208. Dawson J, Fitzpatrick R, Murray D, Carr A. Questionnaire on the perceptions of patients about total knee replacement. J Bone Joint Surg Br 1998; 80: 63-9. EuroQol Group. EuroQol: a new facility for the measurement of healthrelated quality of life. Health Policy 1990; 16: 199-208. Filbay S R, Judge A, Delmestri A, Arden N K, COASt Study Group. Evaluating patients’ expectations from a novel patient-centered perspective predicts knee arthroplasty outcome. J Arthroplasty 2018; 33(7): 2146-52 . Fisher N, Agarwal M, Reuben S F, Johnson D S, Turner P G. Sporting and physical activity following Oxford medial unicompartmental knee arthroplasty. Knee 2006; 13: 296-300. Fransen M, Edmonds J. Reliability and validity of the EuroQol in patients with osteoarthritis of the knee. Rheumatology (Oxford) 1999; 38: 807-13. Hanusch B C, O’Connor D B, Ions P, Scott A, Gregg P J. Effects of psychological distress and perceptions of illness on recovery from total knee replacement. Bone Joint J 2014; 96: 210-16. Jenny J Y, Diesinger Y. The Oxford Knee Score: compared performance before and after knee replacement. Orthop Traumatol Surg Res 2012; 98: 409-12. Judge A, Arden N K, Cooper C, Kassim Javaid M, Carr A J, Field R E, Dieppe P A. Predictors of outcomes of total knee replacement surgery. Rheumatology (Oford) 2012; 51: 1804-13. Kim M S, Koh I J, Choi Y J, Lee J Y, In Y. Differences in patient-reported outcomes between unicompartmental and total knee arthroplasties: a propensity score-matched analysis. J Arthroplasty 2017; 32: 1453-9. Kozinn S C, Scott R. Unicondylar knee arthroplasty. J Bone Joint Surg Am 1989; 71: 145-50. Kurtz S, Ong K, Lau E, Mowat F, Halpern M. Projections of primary and revision hip and knee arthroplasty in the United States from 2005 to 2030. J Bone Joint Surg Am 2007; 89: 780-5. 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: 423-8. Lewis G N, Rice D A, McNair P J, Kluger M. Predictors of persistent pain after total knee arthroplasty: a systematic review and meta-analysis. Br J Anaesth 2015; 114: 551-61. Liddle A D, Judge A, Pandit H, Murray D W. Determinants of revision and functional outcome following unicompartmental knee replacement. Osteoarthritis Cartilage 2014; 22: 1241-50.

Acta Orthopaedica 2019; 90 (1): 74–80

Lui M, Jones C A, Westby M D. Effect of non-surgical, non-pharmacological weight loss interventions in patients who are obese prior to hip and knee arthroplasty surgery: a rapid review. Syst Rev 2015; 4: 121. Mancuso C A, Graziano S, Briskie L M, Peterson M G, Pellicci P M, Salvati E A, Sculco T P. Randomized trials to modify patients’ preoperative expectations of hip and knee arthroplasties. Clin Orthop Rel Res 2008; 466: 424-31. McDonald S, Page M J, Beringer K, Wasiak J, Sprowson A. Preoperative education for hip or knee replacement. Cochrane Library 2014. New York: Wiley for the Cochrane Collaboration; 2015. Munk S, Odgaard A, Madsen F, Dalsgaard J, Jorn L P, Langhoff O, Jepsen C F, Hansen T B. Preoperative lateral subluxation of the patella is a predictor of poor early outcome of Oxford phase-III medial unicompartmental knee arthroplasty. Acta Orthop 2011; 82: 582-8. Murray D W, Fitzpatrick R, Rogers K, Pandit H, Beard D J, Carr A J, Dawson J. The use of the Oxford Hip and Knee Scores. J Bone Joint Surg Br 2007; 89: 1010-4. National Imaging Associates Inc. Clinical guidelines: knee arthroplasty. 2015; 2018. Phoenix, AZ: National Imaging Associates Inc. Nilsdotter A K, Toksvig-Larsen S, Roos E M. Knee arthroplasty: are patients’ expectations fulfilled? A prospective study of pain and function in 102 patients with 5-year follow-up. Acta Orthop 2009; 80: 55-61. Quinn R H, Murray J, Pezold R, Members of the Writing and Voting Panels of the AUC on Surgical Management of Osteoarthritis of the Knee. The American Academy Of Orthopaedic Surgeons appropriate use criteria for surgical management of osteoarthritis of the knee. J Bone Joint Surg Am 2017; 99: 697-9. Ramos-Goni J M, Rivero-Arias O. Eq5d: A command to calculate index values for the EQ-5D quality-of-life instrument. Stata Journal 2011; 11 (1): 120. Royston P, Ambler G, Sauerbrei W. The use of fractional polynomials to model continuous risk variables in epidemiology. Int J Epidemiol 1999; 28: 964-74. Sanchez-Santos M T, Garriga C, Judge A, Batra R N, Price A J, Liddle A D, Javaid M K, Cooper C, Murray D W, Arden N K. Development and validation of a clinical prediction model for patient-reported pain and function after primary total knee replacement surgery. Sci Rep 2018; 8: 3381. Scott C, Howie C R, MacDonald D, Biant L C. Predicting dissatisfaction following total knee replacement. 2010; 92: 1253-8. Scott C E, Bugler K E, Clement N D, MacDonald D, Howie C R, Biant L C. Patient expectations of arthroplasty of the hip and knee. J Bone Joint Surg Br 2012; 94: 974-81. Shrier I, Platt R W. Reducing bias through directed acyclic graphs. BMC Med Res Methodol 2008; 8: 70. Von Keudell A, Sodha S, Collins J, Minas T, Fitz W, Gomoll A H. Patient satisfaction after primary total and unicompartmental knee arthroplasty: an age-dependent analysis. Knee 2014; 21: 180-4. Wallis J A, Taylor N F. Pre-operative interventions (non-surgical and nonpharmacological) for patients with hip or knee osteoarthritis awaiting joint replacement surgery: a systematic review and meta-analysis. Osteoarthritis Cartilage 2011; 19: 1381-95. Williams D H, Greidanus N V, Masri B A, Duncan C P, Garbuz D S. Predictors of participation in sports after hip and knee arthroplasty. Clin Orthop Relat Res 2012; 470: 555-61. Witjes S, Gouttebarge V, Kuijer P P, van Geenen R C, Poolman R W, Kerkhoffs G M. Return to sports and physical activity after total and unicondylar knee arthroplasty: a systematic review and meta-analysis. Sports Med 2016; 46: 269-92. Zahiri C A, Schmalzried T P, Szuszczewicz E S, Amstutz H C. Assessing activity in joint replacement patients. J Arthroplasty 1998; 13: 890-5. Zou G. A modified Poisson regression approach to prospective studies with binary data. Am J Epidemiol 2004; 159: 702-6. Zuiderbaan H A, van der List, J P, Chawla H, Khamaisy S, Thein R, Pearle A D. Predictors of subjective outcome after medial unicompartmental knee arthroplasty. J Arthroplasty 2016; 31: 1453-8.

Acta Orthopaedica 2019; 90 (1): 81–87

Limb lengthening and deformity correction with externally controlled motorized intramedullary nails: evaluation of 50 consecutive lengthenings Joachim HORN 1, Ivan HVID 1, Stefan HUHNSTOCK 1, Anne B BREEN 1, and Harald STEEN 2 1 Section

of Children’s Orthopaedics and Reconstructive Surgery, Division of Orthopaedic Surgery, Oslo University Hospital; 2 Biomechanics Lab, Division of Orthopaedic Surgery, Oslo University Hospital, Norway Correspondence: hornjoachim@gmail.com Submitted 2018-06-25. Accepted 2018-09-11.

Background and purpose — Limb lengthening with an intramedullary motorized nail is a relatively new method. We investigated if lengthening nails are reliable constructs for limb lengthening and deformity correction in the femur and the tibia. Patients and methods — 50 lengthenings (34 Precice and 16 Fitbone devices) in 47 patients (mean age 23 years [11–61]) with ≥ 12 months follow-up are included in this study. 30 lengthenings were done due to congenital and 20 because of posttraumatic deformity (21 antegrade femora, 23 retrograde femora, 6 tibiae). Initial deformities included a mean shortening of 42 mm (25–90). In 15 patients, simultaneous axial correction was done using the retrograde nailing technique. Results — The planned amount of lengthening was achieved in all but 2 patients. 5 patients who underwent simultaneous axial correction showed minor residual deformity; unintentionally induced minor deformities were found in the frontal and sagittal plane. The consolidation index was 1.2 months/cm (0.6–2.5) in the femur and 2.5 months/ cm (1.6–4.0) in the tibia. 2 femoral fractures occurred in retrograde femoral lengthenings after consolidation due to substantial trauma. There were 8 complications, all of which were correctable by surgery, with no permanent sequelae. Interpretation — Controlled acute axial correction of angular deformities and limb lengthening can be achieved by a motorized intramedullary nail. A thorough preoperative planning and intraoperative control of alignment are required to avoid residual and unintentionally induced deformity. In the femur relatively fast consolidation could be observed, whereas healing was slower in the tibia.

Distraction osteogenesis by use of an external fixator is a wellestablished method. To overcome problems associated with the use of external fixation, several techniques that allow early removal of the frame have been developed, including lengthening over a nail (Bost and Larsen 1956, Paley et al. 1997), lengthening and then nailing (Faber et al. 1991, Rozbruch et al. 2008), and lengthening and then plating (Harbacheuski et al. 2012). Further progress has been made by the development of mechanical (Guichet 1999, Cole et al. 2001) and externally controlled motorized intramedullary lengthening devices like the Fitbone nail (Betz et al. 1990, Baumgart et al. 1997) and the Precice nail (Kirane et al. 2014, Schiedel et al. 2014, Paley 2015). Lengthening with a fully implantable motorized and remotecontrolled intramedullary nail is a relatively new method, and only a limited number of reports exist evaluating these lengthening devices. Hence we evaluated our first 50 consecutive cases of limb lengthening with motorized nails in terms of: (1) achieved lengthening and alignment, (2) unintentionally induced deformity, (3) healing of the regenerate and (4) complications. Furthermore, we compared the antegrade nailing technique in the femur with the retrograde nailing technique with respect to these outcome measures.

Patients and methods Patients 50 lengthenings in 47 patients (24 men) with a follow-up of at least 12 months after consolidation of the regenerate are included in this study. 15 of these 50 lengthenings have been previously published (Horn et al. 2015) but are also included in the current paper with additional outcome measures added.

Acta Orthopaedica 2019; 90 (1): 81–87

Table 1. Etiology of LLD and treatment methods Retrograde Antegrade Total femoral femoral Tibial number nail nail nail Etiology n = 50 n = 23 n = 21 n = 6 Posttraumatic Congenital femoral deficiency/ fibula hemimelia Hypoplasia Idiopathic Pes equino varus Achondroplasia Léri–Weill syndrome Short stature Beta thalassemia Hip dysplasia/Perthes Multiple osteochondroma Polio Amniotic band syndrome Metaphyseal dysplasia Osteopathia striata Cerebral paresis

13 11 1 1 10 5 4 1 5 1 4 3 2 1 3 1 2 2 2 2 2 2 2 2 1 1 2 2 1 1 1 1 1 1 1 1 1 1 1 1

The patients’ mean age at operation was 23 years (11– 61). The leg length discrepancy (LLD) was caused by various etiologies (Table 1). In 3 patients consecutive lengthening of both femora was performed due to short stature below 2 SD from average adult height (diagnoses: Léri– Weill dyschondrosteosis, achondroplasia, neonatal growth restraint due to prematurity). Initial deformities included shortening in all patients with a mean of 41 mm (25–88). 23 patients received retrograde femoral nails (RFN), 21 antegrade femoral nails (AFN), and 6 tibia nails (TN). None of the patients had been previously lengthened in the respective segment. In 15 procedures, simultaneous axial correction was done using the RFN. 8 of these patients had initial valgus deformity with a mean lateral (positive) mechanical axis deviation (MAD) of 21 mm (4–50), 5 patients had a varus deformity with a mean medial (negative) MAD of –31 mm (–14 to –58) and 2 patients had a femoral procurvatum deformity of 12° and 26°, respectively. In the remaining 35 procedures, isolated lengthening was performed (21 AFN, 8 RFN, and 6 TN). Mean follow-up after consolidation of the regenerate was 28 months (12–72). Implants In 34 lengthenings we used the Precice lengthening nail (NuVasive Inc, San Diego, CA, USA) and in 16 lengthenings the Fitbone device (Wittenstein intens GmbH, Igersheim, Germany), both fully implantable motorized lengthening nails. Preoperative assessment and surgical technique All patients were assessed preoperatively with physical examination and calibrated radiographs, including long standing radiographs.

Of 44 femoral lengthenings 21 were done with an AFN (all Precice) and 23 with an RFN (16 Fitbone, 7 Precice). All AFN had a 10º proximal bend and we used a trochanteric entry point for insertion. In 15 RFN acute axial correction was performed when the nail was inserted. All 6 tibial lengthenings were done by use of TN with a 10º bend (all Precice). For the retrograde technique preoperative planning was done based on the reverse planning method (Baumgart 2009). A 3 mm diameter threaded Steinmann pin was inserted into the proximal and distal fragment to control rotation, followed by performing multiple drill holes at the intended osteotomy site to vent the canal during reaming. This was done to reduce the risk for fat embolism and to potentially enhance the healing of the regenerate by accumulation of reamer debris at the osteotomy site. In the retrograde technique reaming was done in a preplanned way using rigid, straight reamers, and the use of blocking screws to guide the reamers intraoperatively and to maintain the acutely performed deformity correction. In antegrade femoral and in tibial nailing flexible reamers over a ball-tipped guide wire were used. After reaming, osteotomies were completed in a percutaneous manner (10 mm skin incision) with fan-shaped drilling using a 4-mm diameter drill bit and a 10 mm osteotome. Prophylactic antibiotics were given every 90th minute during surgery. Postoperative care and follow-up Weight-bearing up to 20 kg was permitted immediately. Dalteparin was given 6 hours postsurgery (2500 units) and once a day for the first 7 days after surgery (2500–5000 units depending on body weight). Lengthening in the femur was initiated 7 days after surgery with a distraction rate of 1 mm/day (3 times 0.33 mm/day) and in the tibia 10–14 days after surgery with a rate of 0.66 mm/day (2 times 0.33 mm). Patients were followed with outpatient visits every 2nd week during lengthening and every 6th week during the consolidation phase. They had physical therapy 3 times per week besides daily home exercises with the main emphasis on active and passive extension of the knee. Patients with congenitally short femur and insufficient cruciate ligaments were obliged to use a knee orthosis that kept the knee in full extension for 22 hours per day during the distraction phase and for the first 2 months after completed lengthening. Full weight-bearing was permitted when radiographs showed at least 3 consolidated cortices of the regenerate. Outcome parameters The outcome parameters were: (1) achieved lengthening and alignment, (2) unintentionally induced malalignment, (3) healing of the regenerate and (4) complications. Long standing radiographs were obtained from all patients preoperatively, and after consolidation of the regenerate. Deformity analysis was done based on the malalignment test and malorientation test as described by Paley (2002).

Acta Orthopaedica 2019; 90 (1): 81–87

Analysis of alignment included evalu- Table 2. Outcome measures (mean and range values) of achieved lengthening and alignation of any unintentionally induced ment, including unintentionally induced malalignment malalignment, which might occur due Preoperative Postoperative Change p-value to lengthening along the anatomical axis in the femur resulting in a shift Achieved lengthening in all procedures (n = 50) of the mechanical axis (MA) to lateral Lengthening (mm) Intended Achieved 41 (25–80) 40 (25–65) (Burghardt et al. 2012) or due to uninAchieved alignment in cases with simultaneous axial correction (n = 15) tended angulation of bone fragments MAD (mm) 23 (–58 to 26) 6 (–23 to 24) 17 (4 to 58) 0.01 (Muthusamy et al. 2016). For this purmLDFA (°) 85 (73 to 102) 88 (80 to 91) 5 (2 to 13) < 0.01 Unintentionally induced malalignment in cases of isolated lengthening (n = 35) pose, mechanical axis deviation (MAD) Frontal plane alignment femur and mechanical lateral distal femoral MAD (mm) (antegrade angle (mLDFA) values were measured and retrograde nail) 1 (–20 to 43) 0 (–17 to 32) 3 (0 to 11) 0.9 MNSA (°) (antegrade nail) 124 (106 to 138) 122 (107 to 135) –3 (–9 to 3) 0.008 in all isolated femoral lengthenings mLDFA (°) (retrograde nail) 90 (85 to 95) 90 (87 to 94) 2 (0 to 4) 0.6 preoperatively and after lengthenSagittal plane alignment femur ing, as well as medial neck shaft angle Antegrade femoral nail (°) 7 (0 to 11) 5 (0 to 10) –2 (–9 to 4) 0.02 Retrograde femoral nail (°) 6 (0 to 16) 4 (0 to 12) –2 (–9 to 7) 0.04 (MNSA), which was measured in all Frontal and sagittal plane tibia antegrade nails. Angulation in the sagMPTA (°) 87 (85 to 89) 88 (86 to 89) 1 (1 to 2) 0.2 ittal plane in the femur was measured PPTA (°) 79 (75 to 81) 76 (75 to 81) 3 (0 to 9) 0.1 on standard lateral femur radiographs. MAD: mechanical axis deviation, medial MAD (–), lateral MAD (+); In the tibia medial proximal tibia angle mLDFA: mechanical lateral distal femoral angle; (MPTA) and posterior proximal tibia MNSA: medial neck shaft angle; MPTA: medial proximal tibia angle; angle (PPTA) were measured preopPPTA posterior proximal tibia angle; eratively and after consolidation of the sagittal plane: recurvatum(–), procurvatum (+). regenerate. Furthermore, the precision of the MAD analyses was calculated by evaluating repeated measurements of the healthy lower AFN was 20 years (11–52) and 27 years (15–61) in RFN group. The age distribution in these groups was similar. Patients who extremity in 28 of the individuals included in this study. Radiographs of the bone segment under lengthening were received TN were 21 years (16–30). All patients who received evaluated with respect to the bone regenerate and the amount either RFN or TN were skeletally mature. of achieved lengthening. Consolidation index was defined as the time from the osteotomy to radiographic consolidation Surgical procedure divided by the achieved lengthening distance in centimeters. There were no intraoperative complications, and none of the Complications were graded into problems, obstacles, and patients had any significant blood loss intra- or postoperatively. Mean duration of all surgeries was 159 min (65–330). sequelae according to Paley (1990). Mean duration of surgery for the AFN was 112 min (65–163) Statistics and 187 min (120–330) for the RFN (p = 0.005). In TN the Statistical analyses were done based on Student’s t-test and the mean surgical time was 170 min (147–184). paired t-test. P-values less than 0.05 were considered statistically significant. In order to comply with requirements for inde- Achieved lengthening and alignment pendent observations, the 2nd lengthening in 3 patients with A mean lengthening of 40 mm (25–65) was achieved (Table 2). The planned amount of lengthening was accomplished in all but bilateral lengthening was excluded from all statistical analysis. 2 patients in whom femoral lengthening was terminated due to Ethics, funding, and potential conflicts of interest knee pain and loss of some knee extension when 40 and 50 mm The study was approved as a quality control study by the of lengthening had been achieved. No loss of length was found local office of privacy protection and information safety (Ref. during further follow-up in any patient. MAD (medial deviation 2016/11785). No funding was received and there are no com- = negative, lateral deviation = positive) in all procedures with peting interests declared simultaneous axial correction was significantly reduced and changed from mean 23 mm (–58 to 26) preoperatively to 6 mm (–23 to 24) postoperatively (p = 0.01). Preoperative mLDFA changed from mean 85° (73–102) to 88° (80–91°) (p < 0.01). 5 Results patients who underwent simultaneous axial correction showed Patients some residual deformity. 4 of these patients had a varus deforThe patient’s mean age at surgery was 23 years (11–61) (20 mity with a mean value of 17 mm (12–23) MAD; 1 patient had patients were ≤ 20 years of age). Mean age in the group with residual valgus deformity with 24 mm MAD.

Acta Orthopaedica 2019; 90 (1): 81â&#x20AC;&#x201C;87

Figure 1. 20-year-old woman with CFD and fibular hemimelia. Initial shortening and valgus in the femur and valgus in the tibia. Femoral valgus and shortening were corrected with the retrograde nail technique; some remaining knee valgus due to deformity in the tibia. However, acceptable mechanical axis and currently no further surgery required. High-riding trochanter, but no Trendelenburg gait.

Figure 2. 11-year-old girl with cerebral paresis and hemiplegia on her left side. Initial leg length discrepancy was 3 cm, where the hemiplegic side was the longer extremity. She had serious gait problems and gait analysis with shoe augmentation of 4 cm on the right side showed significant improvement of gait parameters. The girl was lengthened 4 cm in her right femur, overcorrecting her by 1 cm, which was advantageous with respect to her left-sided hemiplegia and future growth. Long standing radiographs demonstrate that no shift of MA axis was observed.

Figure 3. 30-year-old man with 35 mm of tibial shortening due to traumatic injury to the proximal tibial growth plate in childhood. Furthermore, the patient had a symptomatic high-riding fibular head (a). Initial PPTA was slightly below normal (b). The patient was lengthened with a tibia PreciceÂŽ nail. The fibula was osteomized, but transfixation was done only between the tibia and fibula distally to the osteotomy. This resulted in some lengthening of the fibula as well as an intended distalization of the fibular head (c). However, PPTA increased slightly from preoperatively, which was not intended (d). Delayed healing of the regenerate anteriorly (d). However, there was solid callus on 3 cortices (c, d).

Acta Orthopaedica 2019; 90 (1): 81–87

Figure 4. A 12-year-old girl with achondroplasia. She underwent consecutive 50 mm lengthenings of both femurs with the shortest available Precice nail (16.5 cm). Radiographs preoperatively (a), when lengthening was completed (b) and after consolidation (c). The nail has only one locking option proximally and distally to allow for 50 mm lengthening, despite the shortness of the nail. The patient’s femurs had not been lengthened earlier. Lengthening indices for both femurs were 0.6 months/cm, the fastest healing of all procedures included in the current study.

Unintentionally induced malalignment Frontal plane femur A change of MA in isolated femoral lengthening (n = 29) was not intended. Mean MAD preoperatively was 1 mm (–20 to 43) and postoperatively MAD was 1 mm (–17 to 32). The change of MAD from preoperatively to when lengthening was completed was on average 3 mm (0–11), a finding that was not statistically significant (p = 0.9) (Table 2). Change of MNSA would allow for quantifying unintended malalignment in the proximal femur when using antegrade femoral lengthening nails. In the group with AFN preoperative MNSA was mean 124° (106–138) and postoperatively 122° (107–135); this difference is statistically significant (p = 0.008) but not clinically relevant. In 8 patients a varization of the proximal fragment of ≥ 5° was observed, resulting in a shift of the mechanical axis 10 mm to the medial side in 1 patient. In those patients who underwent isolated lengthening (n = 8) or lengthening combined with sagittal plane deformity correction (n = 2) by means of a RFN, changes in mLDFA would allow for conclusions regarding unintentionally induced frontal plane malalignment. However, it has to be noted that isolated lengthening along the mechanical axis would normally lower the mLDFA. Preoperative mLDFA was mean 89° (85–95) and postoperatively 90° (86–94).

Figure 5. A 24-year-old woman who was lengthened 30 mm for idiopathic LLD with a retrograde lengthening nail. After consolidation of the regenerate she fell from a bicycle, sustaining a pertrochanteric femoral fracture (a). A 16-year-old girl with CFD, who underwent 40 mm of lengthening and correction of a valgus deformity with a retrograde lengthening nail. After consolidation of the regenerate she fell 2 m in a waterfall, sustaining a femoral fracture and breakage of the nail at the level of a locking bolt (b).

Sagittal plane femur In the sagittal plane recurvatum (–) and procurvatum (+) of fragments were measured, showing a mean change of –2° (–9 to 4) for the AFN, and –2° (–9 to 7) for RFN. This means a tendency towards dorsiflexion of the fragments (Table 2), which has to be considered a result of reduction of the natural femoral procurvatum by insertion of a straight nail in both antegrade and retrograde femoral lengthening. The 2 patients with preoperative procurvatum deformity were excluded from this calculation. 1 patient developed an unintended procurvatum deformity of 7° during the course of lengthening. Frontal and sagittal plane tibia In the tibia, no change of MTPA could be observed in the 6 patients included in the current study; however, 3 patients showed reduction in PPTA (3°, 3°, and 9°) (Figure 3). Healing of the regenerate All but 1 lengthening consolidated without further interventions. In the femur the lengthening index was mean 1.2 months/cm (0.6–2.5), whereas the lengthening index for the AFN was 1.1 months/cm (0.6–2.0) and 1.3 months/cm (0.8–

2.5) for the RFN (Figure 4). Differences in lengthening indices between the AFN and RFN were statistically significant (p = 0.03). However, in 15 patients receiving RFN an axial correction was performed when the nail was inserted. Exclusion of these patients and comparison of isolated lengthening procedures with the RFN and AFN did not show any statistically significant difference in lengthening indices between the two techniques (p = 0.2). The lengthening index in the tibia was 2.5 months/cm (1.6–4.0). Mean lengthening indices in the femur were for children (≤ 18 years) 0.9 months/cm (0.6–1.3), and 1.4 months/cm (0.8–2.5) for adults. This difference was statistically significant (p = 0.005). Precision of MA analysis As a routine, long standing anterior-posterior radiographs of all patients were obtained before surgery and after healing of the regenerate. These radiographs were taken of both lower extremities simultaneously. In those 28 patients with a healthy contralateral extremity and available radiographs from 2 different time points, a calculation of the precision of our long standing radiograph measurements could be done. MAD at the first measurement was mean 1 mm (–20 to 16) and –1 mm (–19 to 19) at the second measurement. The mean change of MAD between the 2 measurements was 3 mm (0–7). Precision of the MA measurements on our long standing radiographs was therefore assumed to be ± 3 mm. Complications 8 complications occurred that could be solved by surgery without sequelae, and were therefore graded as obstacles according to Paley (1990): in 1 patient (amniotic band syndrome, 40 mm tibial lengthening) autologous bone grafting was required to achieve healing, which occurred within 3 months after that procedure. In 1 patient, an AFN had to be exchanged due to failure of the lengthening mechanism (Precice nail), which was considered to be caused by too extensive hammering during insertion of the nail. 3 patients had to be revised due to migration of locking screws, and 1 patient due to insufficient connection of the receiver in a Fitbone nail. 2 patients sustained femoral fractures due to adequate trauma (fall from bicycle and an accident in a waterfall) 2 and 3 months after consolidation of the regenerate and initiation of full weightbearing (Figure 5). Both patients were lengthened with RFN, which had not been removed at the time of fracture. Both fractures were treated with open reduction and osteosynthesis with locking compression plates. There were no postoperative infections, and no other problems, obstacles, or complications occurred.

Discussion Controlled axial correction and lengthening could be achieved with an externally controlled motorized lengthening nail, which confirms findings by other authors (Krieg et al. 2011,

Acta Orthopaedica 2019; 90 (1): 81–87

Al-Sayyad 2012, Fragomen and Rozbruch 2017). Residual minor deformity was present in one-third of our cases where acute axial corrections were performed during nail insertion. To avoid this, thorough preoperative planning and intraoperative control of alignment are required. Furthermore, unintentionally induced minor frontal plane deformities occurred in one-third of the antegrade femoral lengthenings, resulting in a substantial shift of the MA in 1 patient. It is essential to insert the nail at the very tip of the trochanter to reduce the risk of varization of fragments, and blocking screws might be used to control alignment (Baumgart 2009, Muthusamy et al. 2016, Hammouda et al. 2017). Trochanteric entry can be used in femoral nailing in the skeletally immature patient (MacNeil et al. 2011, Hammouda et al. 2017). We used trochanteric entry point in all patients with an AFN. A fossa piriformis entry point might be less frequently associated with varization of fragments and can be safely used in adult patients (Kim et al. 2016). Unintentionally induced minor deformity in the sagittal plane in femoral lengthening could be observed in most patients and is caused by the fact that a straight nail is inserted into a naturally flexed femur. Adverse clinical effects on range of motion and function have not been observed. In cases of isolated lengthening no shift of mechanical axis to the lateral could be observed, although lengthening with nails is performed along the anatomical axis of the femur. This might be due to 2 reasons: (1) the amount of lengthening was not great enough to result in a measurable shift of MA, detectable on long standing radiographs, considering the accuracy of this measurement of ±3 mm; (2) in more than one-third of the procedures some unintended varization of the proximal fragment in antegrade nailing was induced, which may have compensated for the effect of lengthening along the anatomical axis to some degree. Few tibial lengthenings were included in our study. Most frequently, unintended procurvatum and valgus deformity can occur with tibial lengthenings (Muthusamy et al. 2016). However, we only observed minor unintended procurvatum deformity of no clinical significance in our study. All femurs healed within an appropriate time, and most of them showed fast healing, which corresponds to findings by other authors (Krieg et al. 2011, Kucukkaya et al. 2015, Karakoyun et al. 2016, Laubscher et al. 2016). Healing in the femur seems faster, when the patient’s age is ≤ 18 years, which might have importance for the timing of reconstructive procedures. In the tibia slower healing was observed. However, the number of tibial lengthenings was low and all patients who underwent tibial lengthening had an underlying diagnosis associated with reduced bone healing potential. The number of complications was low (8/50), although additional surgery was required in these patients. However, the current study analyses all our first 50 consecutive cases and the method certainly has a learning curve allowing for reduction of the number of complications with increasing experience. 2 patients sustained a femoral fracture due to trauma after consolidation of the regenerate and return to full weightbearing. Besides the occurrence of these 2 femoral fractures we

Acta Orthopaedica 2019; 90 (1): 81â&#x20AC;&#x201C;87

found similar outcome measures between the antegrade and retrograde nailing technique in the femur and there are no other reports comparing these 2 lengthening techniques. Nevertheless, we prefer the antegrade approach in the femur when isolated lengthening is performed. A strength of our study is that we report all consecutive lengthenings with externally controlled intramedullary nails that have been performed at our department. In summary, both femoral and tibial motorized lengthening nails are reliable implants for limb lengthening, with a low number of complications. Preoperative planning, intraoperative control, and adequate postoperative care and follow-up are essential in order to achieve a good result.

JH, SH, IH, and AB performed the surgeries and examined the patients at follow-up. JH wrote the manuscript. HS and IH revised the manuscript. Acta thanks John Gerard Birch and Hubert Jan Oostenbroek for help withÂ peer review of this study.

Al-Sayyad M J. Lower limb lengthening and deformity correction using the Fitbone motorized nail system in the adolescent patient. J Pediatr Orthop B 2012; 21 (2): 131-6. Baumgart R. The reverse planning method for lengthening of the lower limb using a straight intramedullary nail with or without deformity correction: a new method. Oper Orthop Traumatol 2009; 21 (2): 221-33. Baumgart R, Betz A, Schweiberer L. A fully implantable motorized intramedullary nail for limb lengthening and bone transport. Clin Orthop Relat Res 1997 (343): 135-43. Betz A, Baumgart R, Schweiberer L. [First fully implantable intramedullary system for callus distractionâ&#x20AC;&#x201D;intramedullary nail with programmable drive for leg lengthening and segment displacement. Principles and initial clinical results]. Chirurg 1990; 61 (8): 605-9. Bost F C, Larsen L J. Experiences with lengthening of the femur over an intramedullary rod. J Bone Joint Surg Am 1956; 38-A (3): 567-84. Burghardt R D, Paley D, Specht S C, Herzenberg J E. The effect on mechanical axis deviation of femoral lengthening with an intramedullary telescopic nail. J Bone Joint Surg Br 2012; 94 (9): 1241-5. Cole J D, Justin D, Kasparis T, DeVlught D, Knobloch C. The intramedullary skeletal kinetic distractor (ISKD): first clinical results of a new intramedullary nail for lengthening of the femur and tibia. Injury 2001; 32 (Suppl 4): SD129-39. Faber F W, Keessen W, van Roermund P M. Complications of leg lengthening: 46 procedures in 28 patients. Acta Orthop Scand 1991; 62 (4): 327-32. Fragomen A T, Rozbruch S R. Retrograde magnetic internal lengthening nail for acute femoral deformity correction and limb lengthening. Expert Rev Med Devices 2017; 14 (10): 811-20.

Guichet J M. [Leg lengthening and correction of deformity using the femoral Albizzia nail]. Orthopade 1999; 28 (12): 1066-77. Hammouda A I, Jauregui J J, Gesheff M G, Standard S C, Herzenberg J E. Trochanteric entry for femoral lengthening nails in children: is it safe? J Pediatr Orthop 2017; 37 (4): 258-64. Harbacheuski R, Fragomen A T, Rozbruch S R. Does lengthening and then plating (LAP) shorten duration of external fixation? Clin Orthop Relat Res 2012; 470 (6): 1771-81. Horn J, Grimsrud O, Dagsgard A H, Huhnstock S, Steen H. Femoral lengthening with a motorized intramedullary nail. Acta Orthop 2015; 86 (2): 248-56. Karakoyun O, Sokucu S, Erol M F, Kucukkaya M, Kabukcuoglu Y S. Use of a magnetic bone nail for lengthening of the femur and tibia. J Orthop Surg (Hong Kong) 2016; 24 (3): 374-8. Kim J W, Oh J K, Byun Y S, Shon O J, Park J H, Oh H K, et al. Incidence of avascular necrosis of the femoral head after intramedullary nailing of femoral shaft fractures: a multicenter retrospective analysis of 542 cases. Medicine (Baltimore) 2016; 95 (5): e2728. Kirane Y M, Fragomen A T, Rozbruch S R. Precision of the PRECICE internal bone lengthening nail. Clin Orthop Relat Res 2014; 472 (12): 3869-78. Krieg A H, Lenze U, Speth B M, Hasler C C. Intramedullary leg lengthening with a motorized nail. Acta Orthop 2011; 82 (3): 344-50. Kucukkaya M, Karakoyun O, Sokucu S, Soydan R. Femoral lengthening and deformity correction using the Fitbone motorized lengthening nail. J Orthop Sci 2015; 20 (1): 149-54. Laubscher M, Mitchell C, Timms A, Goodier D, Calder P. Outcomes following femoral lengthening: an initial comparison of the Precice intramedullary lengthening nail and the LRS external fixator monorail system. Bone Joint J 2016; 98-b (10): 1382-8. MacNeil J A, Francis A, El-Hawary R. A systematic review of rigid, locked, intramedullary nail insertion sites and avascular necrosis of the femoral head in the skeletally immature. J Pediatr Orthop 2011; 31 (4): 377-80. Muthusamy S, Rozbruch S R, Fragomen A T. The use of blocking screws with internal lengthening nail and reverse rule of thumb for blocking screws in limb lengthening and deformity correction surgery. Strategies Trauma Limb Reconstr 2016; 11 (3): 199-205. Paley D. Problems, obstacles, and complications of limb lengthening by the Ilizarov technique. Clin Orthop Relat Res 1990 (250): 81-104. Paley D. Principles of deformity correction. Berlin, Heidelberg, New York: Springer; 2002. Paley D. PRECICE intramedullary limb lengthening system. Expert Rev Med Devices 2015; 12 (3): 231-49. Paley D, Herzenberg J E, Paremain G, Bhave A. Femoral lengthening over an intramedullary nail: a matched-case comparison with Ilizarov femoral lengthening. J Bone Joint Surg Am 1997; 79 (10): 1464-80. Rozbruch S R, Kleinman D, Fragomen A T, Ilizarov S. Limb lengthening and then insertion of an intramedullary nail: a case-matched comparison. Clin Orthop Relat Res 2008; 466 (12): 2923-32. Schiedel F M, Vogt B, Tretow H L, Schuhknecht B, Gosheger G, Horter M J, et al. How precise is the PRECICE compared to the ISKD in intramedullary limb lengthening? Reliability and safety in 26 procedures. Acta Orthop 2014; 85 (3): 293-8.

Acta Orthopaedica 2019; 90 (1): 88–93

Suggestion for new 4.4 mm pubo-femoral distance cut-off value for hip instability in lateral position during DDH screening Hans-Christen HUSUM 1, Michel B HELLFRITZSCH 2, Nina HARDGRIB 3, Bjarne MØLLER-MADSEN 1, and Ole RAHBEK 1 1 Department 3 Department

of Children’s Orthopaedics, Aarhus University Hospital, Aarhus, 2 Department of Radiology, Aarhus University Hospital, Aarhus, of Paediatrics, Aarhus University Hospital, Aarhus, Denmark Correspondence: hchusum@hotmail.com Submitted 2017-12-08. Accepted 2018-11-06.

Background and purpose — Current selective screening algorithms for developmental dysplasia of the hip (DDH) are insufficient. Universal screening programs have been proposed but so far have been deemed too expensive and time consuming. The pubo-femoral distance may solve this problem as a quick, low-cost, highly sensitive, and specific sonographic measurement for DDH, but this has only been validated in the supine position. Therefore we validated pubo-femoral distance (PFD) in the lateral position as an indicator for instability of the hip. Methods — All participants had undergone ultrasonographic diagnostics using the modified Graf technique. In addition, PFD measurements in lateral position were performed. Results were compared between 25 infants who had been treated for DDH because of dysplastic appearance on ultrasound combined with clinical instability and a control group consisting of 100 untreated infants screened for DDH. Sensitivity, specificity, and cut-off points were determined using Receiver operating characteristics (ROC) analysis. Results — We found a mean PFD of 6.8 mm (6.2–7.4) in the treated group with a control group PFD of 3.4 mm (3.3–3.6) (p < 0.005). A PFD value above a threshold of 4.4 mm yielded a sensitivity of 100% and a specificity of 93% for detecting unstable DDH. Interpretation — PFD measured in lateral position was statistically significantly increased in hips of children treated for DDH with Denis Browne hip brace compared with healthy children with unaffected stable hips. Furthermore, the PFD measurement had a high level of sensitivity and specificity at a cut-off value of 4.4 mm. A cut-off value of 6.00 mm has previously been reported as the gold standard in supine position. We suggest that 4.4 mm is used in lateral position.

Universal ultrasound screening decreases the rate of late diagnosis and surgical interventions, and is cost-effective (Thaler et al. 2011) as well as useful in detecting developmental dysplasia of the hip (DDH) in children with no apparent clinical symptoms or risk factors (Marks et al. 1994). In Denmark universal ultrasound screening is not implemented. Infants are routinely screened for hip instability by a midwife after birth, and at a 5-week routine follow-up at their general practitioner. The screening is clinical and is based primarily on the Barlow and Ortolani maneuvers. If there are positive clinical findings, the child is referred to the next level of the screening program, which is an ultrasound by an experienced radiologist for further diagnostics as clinical examination alone is not considered sufficient (Rosenberg et al. 1998, Roposch et al. 2006). Children with established risk factors for the development of DDH (breech presentation, 1 degree relative to DDH, oligohydramnios, congenital deformities), bypass the general screening program, and are referred to ultrasound examination directly by the midwife. These infants are screened for DDH with a combination of clinical examination of the hip instability and ultrasound using a modified Graf technique (Graf 1983). However, these gold standard ultrasound examinations require a skilled radiologist in order to obtain a correct diagnosis of DDH (Hell 2008, Omeroğlu et al. 2001). Treguier et al. (2013) have implemented the pubo-femoral distance (PFD) measured in supine position, as developed by Couture et al. (2011), in a screening program in France. This is a simple measurement to detect hip instability and identify at-risk hips where early intervention is warranted. The PFD method does not require an experienced radiologist; it can be used by radiologists or technicians with limited experience in ultrasound. This could potentially reduce

Acta Orthopaedica 2019; 90 (1): 88–93

Patients treated with DB brace n = 159 Excluded for missing PFD data n = 135 Patients n = 24 Patients with unilateral DDH n = 12

Patients with bilateral DDH n = 12 Excluded for missing data n=4 Affected hips n = 20

Affected hips n = 12 Affected hips n = 32

Figure 1. Recruitment process: flowchart of cases with hip instability.

Figure 2. The ultrasound examination with the child in lateral position.

the need for skilled radiologists to be involved in the early stage of screening and facilitate implementation of a universal ultrasound screening program. The PFD method has been demonstrated by Teixeira et al. (2015) to be highly sensitive and specific in the diagnosis of DDH at a cut-off point of 4.6 mm with the infant in lateral position. In contrast Treguier et al. (2013) use a cut-off point of 6.0 mm when measuring the infant in supine position and the hip flexed in adduction. The cut-off point for PFD may therefore be dependent on position of the patient and the ultrasonic plane as the reported thresholds are conflicting. Furthermore, the method has not to our knowledge been validated in a Danish population. We validated PFD as a tool for early diagnosis of DDH in Danish children by evaluating the PFD measurements of children diagnosed with, and treated for, unstable DDH on the basis of Graf’s method combined with clinical stability testing. The sensitivity and specificity of the method is examined, and an optimal cut-off point for the PFD measurement in lateral position in the diagnosis of instable DDH is determined.

referred from hospitals with different ultrasound examination algorithms and therefore no recorded PFD values (n = 3). A total of 24 patients (23 girls) were included, 12 with unilateral DDH and 12 with bilateral DDH, in all 32 affected hips (Figure 1) (4 hips in the bilateral group were missing PFD values). Mean age at the time of referral to initial ultrasound screening was 25 days. Radiological charts were reviewed for assessing the Graf measurements, bone rim percentage (BRP), and PFD values.

Methods

Ultrasound examination The ultrasound examinations were performed by experienced musculoskeletal radiologists from radiological departments in the region including our own in-house radiologists. The examination was done in the presence of the parents with a highfrequency transducer (Hitachi model: L74M 13-5; Hitachi Europe, Maidenhead, UK). According to our standard scanning protocol, a modification of the Tréguier method was used, as the child was positioned on the side as per international guidelines (AIUM 2009) (Figure 2). The PFD measurements were performed after obtaining the measurements according to the Graf method. The infant was positioned lying on its side in a specific cradle with the hip flexed in adduction, for determination of pubo-femoral distance (Tréguier et al. 2013).

Cases We identified 159 patients examined sonographically for DDH and subsequently treated with a Denis Browne (DB) hip brace for unilateral or bilateral unstable DDH at Aarhus University Hospital from February 2013 to September 2016. All treatments with the DB hip brace are coded and registered systematically with the treatment code BLPD10-13 in the hospital system. The identified patients have been validated against the radiological PACS reports to ensure the correct diagnosis. The total number of patients treated was 41. Patients were excluded if PFD measurements were missing (n =17). This was due to dislocation of the hip making the PFD immeasurably high (n = 1) or patients who had been

Controls The healthy control group consists of all patients referred for an ultrasound examination at the same department in the period January 2017 to March 2017, whose ultrasound examination was deemed normal using the modified Graf technique. No patients were excluded from the control group. 100 patients had both hips examined sonographically with a normal result, in the period January 2017 to March 2017 in our department, totaling 200 hips examined, of which 95 were female and 103 were male, and mean age at referral was 39 days. PFD values were available on 198 hips, making up our control group.

Acta Orthopaedica 2019; 90 (1): 88–93

PDF (mm) 12 10 8 6 4 2 0

Figure 3. Ultrasound image of normal hip at 4 weeks. All anatomical landmarks present. The horizontal line extending from the iliac wing crosses the perpendicular femoral head diameter and defines the upper limit for d which is used to calculate BRP = d/D. The PFD is marked here by the double-headed arrow.

Figure 4. Ultrasound image of abnormal hip, girl aged 4 weeks, all anatomical landmarks present. PFD estimated at 5.8 mm as indicated by the 2 + symbols.

The transducer was placed in a rectangular position, and thereafter slid from ventral to dorsal to visualize the bony acetabulum and rotation of the transducer to visualize the straight ilium. No inclination of the transducer is permitted as this results in bowing of the ilium and loss of hip anatomy. The quality criteria required were 2 cartilaginous landmarks consisting of the cartilaginous femoral head and the triangular hyperechoic fibrocartilaginous rim, and 3 bony landmarks consisting of the horizontal iliac wing, the bony acetabular roof at its greatest depth, and the pubic bone (Figure 3). Hip stability was evaluated by Barlow equivalent provocation test, with the infant in the lateral position. Alpha angles, beta angles, BRP, and PFD measurements were taken for each hip. The BRP measurement was determined by the ratio d/D, where d = the portion of the diameter of the femoral head covered by the acetabular bone, below the horizontal line extending from the iliac bone, and D = the (entire) femoral head diameter. PFD was measured between the medial margin of the epiphysis and the pubic bone according to the technique used by Treguier et al. (2013). If the ultrasound detected structural abnormalities and/or hip instability was suspected, the patient would be referred to a pediatric orthopedic department for clinical testing. If DDH was confirmed, the patient would be treated with a DB brace for a minimum of 6 weeks with subsequent follow-up with repeated ultrasound examination.

Control

Treated

Figure 5. Box and whiskers plot of PFD values of controls plotted against hip with instability. The boxes represent the interquartile range. Whiskers represent the range of all values. The red line within the boxes is median value.

Statistics Data were tested for normality by Shapiro–Wilk test and a paired t-test was used for comparison between groups. A p-value < 0.05 was considered statistically significant. Mean values and 95% confidence intervals (CI) are given. Receiver operating characteristic (ROC) curve analysis was done using Medcalc from MedCalc Software (https://www.medcalc.org/) according to the methodology of Delong et al. (1988) in order to evaluate specificity, sensitivity and determine an optimal cut-off point for PFD. Ethics, funding, and potential conflicts of interest The Danish Data Protection Agency and the Danish Health Authority approved this single institution case-control study. The authors received no funding for this study and declare no conflicts of interest.

Results We found a mean PFD of 6.8 mm (CI 6.2–7.4) in the treated group compared with 3.4 mm (CI 3.3–3.6) in the controls. When compared with the control group (Table 1, Figures 4 and 5) there was a statistically significant difference for all tested parameters (alpha angle, BRP, and PFD). The difference between the PFD value of the right and left hip (PFDΔ) was analyzed in both unilaterally and bilaterally

Table 1. Alpha angles, bone rim percentage (BRP) and pubofemoral distance (PFD) of the treated and control group. Values are mean (standard deviation) [95% CI] Group

Alpha angle (°)

p-value

Controls (n = 198) Treated (n = 32)

70 (17) [68–72] < 0.001 55 (5.3) [53–57]

BRP (%)

p-value

65 (1.0) [65–66] < 0.001 50 (9.7) [47–53]

PFD (mm)

p-value

3.4 (0.96) [3.3–3.6] < 0.001 6.8 (1.7) [6.2–7.4]

Acta Orthopaedica 2019; 90 (1): 88–93

Table 2. Difference in PFD between each patient’s two hips in the treated and control group. Values are mean (mm) (standard deviation) [95% CI]

Sensitivity 100 80

PFDΔ p-value

Normal (n = 98) Unilateral DDH (n = 12) Bilateral DDH (n = 8)

0.47 (0.38) [0.40–0.55] 2.6 (1.3) [1.9–3.3] 2.2 (2.3) [0.7–3.7]

< 0.001 < 0.001

P-value = 0.6 between unilateral DDH vs. bilateral DDH.

60 40 20 0

affected patients (Table 2). We found a mean PFDΔ of 2.6 mm and 2.2 mm respectively. There was no statistically significant difference between PFDΔ of the unilaterally and bilaterally affected (p = 0.6), but a significant difference (p < 0.05) was found when compared with the control group. The sensitivity was 100% and specificity 93% at a cut-off value of 4.4 mm (Figure 6).

Discussion The current literature on PFD measurement is sparse, even though the method has the potential to be a low-budget universal screening tool. However, the cut-off level for instability is not clear and needs to be established. Our study suggests that the cut-off value for intervention should be set lower at 4.4 mm to reach 100% sensitivity compared with 6.0 mm as previously reported. These findings may be explained by population characteristics (numbers included, hip morphology, age), different thresholds for treatment, or the fact that we used lateral position as opposed to supine as previously reported (Tréguier et al. 2013). Our findings support the findings of Teixeira et al. (2015) that a lower threshold is needed when a lateral position with adducted flexed hip is used. They found a cut-off value of 4.6 mm in their study. We found PFD measured in lateral position to be statistically significantly increased in children with hip instability, when compared with a group of infants with normal ultrasound examinations. If diagnosis was to be made on the basis of PFD alone, all our patients treated for DDH with a DB brace would have been diagnosed and treated correctly with only 7% receiving a false-positive diagnosis. PFD measurements have previously been shown to be very useful for assessing DDH (Couture et al. 2011, Salut et al. 2011, Tréguier et al. 2013). Treguier et al. (2013) diagnosed DDH among 1-month old patients and found sensitivity and specificity values of 100% and 80% respectively, with a cut-off point of 6 mm in supine flexed position. We chose to scan the infant in lateral position as all infants had Graf measurements performed as per international guidelines (AIUM 2009); this allowed both tests to be performed simultaneously and more easily.

100

100-Specificity Figure 6. ROC graph illustrating the sensitivity and specificity for PFD in diagnosing DDH. Cut-off point was 4.4 mm. AUC = 0.99 (0.97–1.0), p < 0.001. Sensitivity was 100% and specificity was 93%.

The PFD method requires little experience by the sonographic examiner to correctly and quickly classify the patients needing intervention (Tréguier et al. 2013, Teixeira et al. 2015). The inter-observer reproducibility was tested by Treguier et al. (2013) by analyzing the level of agreement between experienced and inexperienced operators in determining the PFD and BRP classification. They found agreement in the operators’ classification in 92% of the cases for PFD and only 41% of the cases of BRP classification (Tréguier et al. 2013). The inexperienced operator used in that study was a resident radiologist in the early phase of his training, but the data may be extrapolated to other examiners such as midwives or general practitioners who have received a minimum of training in the procedure. The simplicity and the reproducibility of the method have since been confirmed and measurements have been demonstrated to be independent of hip positions (flexion or neutral) (Teixeira et al. 2015). In contrast the Graf method can be complex and difficult to perform by inexperienced sonographers and radiologists, and requires extensive training to master (Omeroğlu et al. 2001, Hell 2008). Rosendahl et al. (1995) compared 2 observers with 5 and 2 years of radiological experience and found a low level of agreement when evaluating each one’s image acquisition. There was a higher level of agreement when evaluating the same static images, which could indicate that the disagreement lies in the technique and obtainment of the sonographic images rather than the interpretation itself. This is supported by the distinct influence of probe positioning on the measurement results as Graf initially found himself (Graf 1983), as well as acceptable inter-observer agreement for the Graf angle measurements when performed using static images previously captured and reviewed by an experienced radiologist (Teixeira et al. 2015). The PFD 6.0 mm threshold as diagnostic for DDH is suggested by Tréguier et al. (2015), and also a 1.5 mm difference between the 2 hips was suggested as a threshold in their paper. In our study, the differences in mean PFDΔ of affected

hips when compared with the control group were found to be statistically significantly increased; however, between the hips of respectively unilaterally and bilaterally affected patients PFDΔwas found to be non-significant. This finding does not support the recommendations by Treguier et al. (2015), but this can most likely be explained by the low number of patients involved in our study and may be regarded as a type 2 error. Therefore, no firm conclusions can be made on this matter based on our study. We recommend that PFDΔ above 1.5 mm is kept as a diagnostic criterion in the lateral position until larger data-sets are available. Our study is limited by the fact that the final impact on the rate of late diagnosis of DDH has not yet been assessed, and by its small sample population size, though we were able to demonstrate a statistically significant PFD increase in lateral position at the time of diagnosis in patients treated for DDH even with our limited number of included subjects. In several cases, PFD measurements were not supplied in referrals from local hospitals, due to different sonographic examination standards and algorithms, and in some cases this was not performed by in-house radiologists due to dislocation of the examined hip, thus reducing the amount of data available for analysis. Dislocation of the hip makes the PFD immeasurably high and it would therefore exceed our proposed cut-off point, as well as producing a positive clinical examination, and the patient would therefore be referred for specialized ultrasound on both grounds. Since this method was introduced to the screening program of 2 million people in the French population by Tréguier et al. (2013), they have reported a 75% reduction in the rate of late diagnosis. In order to completely eradicate late DDH diagnoses, they propose a universal screening program for girls, as they have found that 70% of infants put into traction for DDH are girls, and this itself is a risk factor for DDH. This finding is supported by data from other groups (Marks et al. 1994, Paton et al. 2005, Vane et al. 2005) and our study where 23/25 of treated patients were girls. Clinical examination alone is insufficient in diagnosing DDH in infants as mild acetabular dysplasia and instability are common, and even selective ultrasound screening has been found to be inadequate when using current screening criteria (Sink et al. 2014). This prompts the need for a more generalized screening program for DDH, which would require a sensitive, easily learned and low-cost method of examination. We believe the PFD measurement fulfills these criteria for the reasons listed above and that it could become a future referral criterion for referral for the specialized Graf US examination. A primary concern of the application of universal screening of female infants is the risk of generating a high number of false-positives. This, combined with the fact that the PFD measurement in some cases is seen to be increased above the diagnostic threshold due to thickening of the pubic bone cartilage (Tréguier et al. 2013) in otherwise clinically normal and

Acta Orthopaedica 2019; 90 (1): 88–93

stable hips in healthy infants, is a concern. However, since the implementation of the screening program in their female population, Treguier et al. has not seen a significant increase in consultation or splintage rates. In order to further evaluate PFD measurements as a referral criterion in the screening algorithm for DDH in Denmark in the future, a study with a prospective design, and consistent use of the measurements in all referred patients—both primary referrals and secondary referrals from other hospitals— is necessary. In summary, PFD in lateral position was shown to be significantly increased in hips of children treated for DDH with DB brace compared with healthy children with unaffected stable hips. A cut-off value of 6.0 mm has previously been reported as optimal in supine position. We suggest that 4.4 mm is used in lateral position. PFD may be a feasible low-cost tool for universal screening of infants and could be a supplement to the current highly specialized examination of risk groups in Denmark. HH, OR, and MH planned the paper. HH drafted the manuscript and performed statistical analyses. All authors revised and approved the final manuscript Acta thanks Deborah M Eastwood and Ola Wiig for help with peer review of this study.

AIUM. Practice guideline for the performance of an ultrasound examination for detection and assessment of developmental dysplasia of the hip. J Ultrasound Med 2009; 28: 114-19. Couture A, Baud C, Prodhomme O, Saquintaah M, Veyrac C. Ultrasound of the neonatal hip: initial evaluation and follow-up. J Radiol 2011; 92(2): 142-65. Delong E R, Delong D M, Clarke-Pearson D L. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 837–45. Graf R. New possibilities for the diagnosis of congenital hip joint dislocation by ultrasonography. J Pediatr Orthop 1983 3(3): 354-9. Hell A. Inter- and intraobserver reliability in Graf’s sonographic hip examination. Z Orthop Unfall 2008; 146(5): 624-9. Marks D S, Clegg J, al-Chalabi A N. Routine ultrasound screening for neonatal hip instability. Can it abolish late-presenting congenital dislocation of the hip? J Bone Joint Surg Br 1994; 76(4): 534-8. Omeroğlu H, Biçimoğlu A, Koparal S, Seber S. Assessment of variations in the measurement of hip ultrasonography by the Graf method in developmental dysplasia of the hip. J Pediatr Orthop B 2001; 10(2): 89-95. Paton R W, Hinduja K, Thomas C D, Hinduja K. The significance of at-risk factors in ultrasound surveillance of developmental dysplasia of the hip: a ten-year prospective study. J Bone Joint Surg [Br] 2005; 87(Ci): 1264-6. Roposch A, Moreau N M, Uleryk E, Doria A S. Developmental dysplasia of the hip: quality of reporting of diagnostic accuracy for US. Radiology 2006; 241(3): 854-60. Rosendahl K, Aslaksen A, Lie R T, Markestad T. Reliability of ultrasound in the earlydiagnosis of developmental dysplasia of the hip. Pediatr Radiol 1995; 25(3): 219-24. Rosenberg N, Bialik V, Norman D, Blazer S. The importance of combined clinical and sonographic examination of instability of the neonatal hip. Int Orthop 1998; 22(3): 185-8.

Acta Orthopaedica 2019; 90 (1): 88–93

Salut C, Moriau D, Pascaud E, Layré B, Peyrou P, Maubon A. [Initial results from an ultrasound screening program for the detection of developmental dysplasia of the hip in girls]. J Radiol 2011; 92(10): 920-9. Sink E L, Ricciardi B F, Torre K D, Price C T. Selective ultrasound screening is inadequate to identify patients who present with symptomatic adult acetabular dysplasia. J Child Orthop 2014; 8(6): 451-5. Teixeira S R, Dalto V F, Maranho D A, Zoghbi-Neto O S, Volpon J B, Nogueira-Barbosa M H. Comparison between Graf method and pubo-femoral distance in neutral and flexion positions to diagnose developmental dysplasia of the hip. Eur J Radiol 2015; 84(2): 301-6.

Thaler M, Biedermann R, Lair J, Krismer M, Landauer F, Landauer F. Costeffectiveness of universal ultrasound screening compared with clinical examination alone in the diagnosis and treatment of neonatal hip dysplasia in Austria. J Bone Joint Surg Br 2011; 93-B: 1126-30. Tréguier C, Chapuis M, Branger B, Bruneau B, Grellier A, Chouklati K, Proisy M, Darnault P, Violas P, Pladys P, Gandon Y. Pubo-femoral distance: an easy sonographic screening test to avoid late diagnosis of developmental dysplasia of the hip. Eur Radiol 2013; 23(3): 836-44. Vane A G S, Gwynne Jones D P, Dunbar J D, Theis J-C. The diagnosis and management of neonatal hip instability: results of a clinical and targeted ultrasound screening program. J Pediatr Orthop 2005; 25(3): 292-5.

Acta Orthopaedica 2019; 90 (1): 94–95

Correspondence

High failure rate after internal fixation and beneficial outcome after arthroplasty in treatment of displaced femoral neck fractures in patients between 55 and 70 years Sir,—It was with great interest we read the article “High failure rate after internal fixation and beneficial outcome after arthroplasty in treatment of displaced femoral neck fractures in patients between 55 and 70 years: An observational study of 2,713 patients reported to the Norwegian Hip Fracture Register” by Bartels et al. (2018). The article raised a lot of questions and we would like to make some comments regarding the conclusions drawn by the authors. The authors concluded that younger patients operated on with an arthroplasty had a better health-related quality of life (EQ-5D), less pain, and better patient satisfaction compared with those patients operated on with internal fixation (IF). First, in our opinion the follow-up was too short for studying surgical complications after an arthroplasty. The reason why an arthroplasty has not been the treatment of choice in younger age is the risk for long-term complications. It is therefore unclear why the follow-up was as short as one year. Second, since this is a register study and not a RCT study it is a biased situation and there seem to be a lot of uncontrolled confounders, as always in register studies. Already in the choice of surgical treatment there is selection bias, i.e. the IF group consisted of considerably more men compared with the arthroplasty groups. This finding is not discussed at all. In the light of experience we know that younger men with hip fractures often have other considerable comorbidities and high alcohol consumption, factors that may have influenced the PROM data. Furthermore there were only 549 out of 2,713 patients (20%) that answered the patient questionnaires (PROM). This leads to a problem with the external validity and the generalizability of the findings and is only mentioned briefly by the authors. The authors have chosen to analyze patients with healed fractures together with those who had reoperation (IF group). This might have led to false low mean values (PROM), and further to the conclusion drawn by the authors, of a beneficial outcome of an arthroplasty compared with the IF group. The majority of the patients who did not need reoperation (70%), and therefore would be of greater interest to discuss, would probably have a higher EQ-5D index score if the groups were analyzed separately. This has been found in an earlier study (Campenfeldt et al. 2017). It is clear that those patients who need reoperation after IF suffer, but it is still unclear what the clinical outcomes are for the absolute majority of the IF

group—those whose fractures heal. This is a relevant scientific question that still needs to be answered and this is not addressed at all in the present study. Furthermore, the differences found in VAS and EQ-5D after 12 months between groups were low and the clinical relevance is questionable and consequently also the conclusion drawn by the authors. Margareta HEDSTRÖM 1,2, Wilhelmina EKSTRÖM 3, Amer AL-ANI 4 and Pierre CAMPENFELDT 1,5 1 Department

of Clinical Science, Intervention and Technology (CLINTEC), Karolinska Institutet; 2 Department of Orthopaedics, Karolinska University Hospital Huddinge, Stockholm, Sweden; 3 Department of Molecular Medicine and Surgery, Karolinska Institutet and Department of Orthopaedics, Karolinska University Hospital Solna, Stockholm, Sweden; 4 Orthopaedic Clinic, Vällingby-Läkarhus, Sweden; 5 Defence Inspector for Medicine and Environmental Health, Swedish Armed Forces, Sweden. Correspondence: margareta.hedstrom@karolinska.se

Sir,—We thank Dr Hedström and colleagues for the comments on our recently published article. The results of our study are not controversial; they actually confirm the results from randomized controlled studies, albeit in a slightly younger population. In RCTs on healthy individuals aged over 60 years, both Keating et al. (2006) and Frihagen et al. (2007) found better results after arthroplasty than after internal fixation. The strength of the current study is to confirm this in a national population, a finding with high external validity. The limitations are, as pointed out by Hedström and colleagues, the difficulties in obtaining patient-reported outcome from elderly patients. However, the Norwegian Hip Fracture Register is unique in its ambitions to achieve truly national reporting of PROM. The risk of residual confounding is addressed in the discussion. This is the downside of register studies. Nevertheless, national register studies are needed as a valuable complement to RCTs, thanks to their generalizability and large study population.

Acta Orthopaedica 2019; 90 (1): 94–95

capsular femoral neck fractures may benefit from treatment with arthroplasty is scientifically sound. The acceptance of such a suggestion is reflected by THA already being more common as primary treatment for displaced femoral neck fractures than IF in Sweden when patients reach around 58 years old (Swedish Fracture Register, Annual Report 2017). Stefan BARTELS 1, Jan-Erik GJERTSEN 2,3, Frede FRIHAGEN 4, Cecilia ROGMARK 5, and Stein Erik UTVÅG 1,6 1 Department

Figure 2. Adjusted survival of implants for the different treatment groups with major reoperations as endpoint, distributed by primary treatment method. Cox regression analyses with adjustments for age, sex, and ASA classification (Bartels et al. (2018)).

Hedström and colleagues suggest that a larger share of men with alcohol abuse in the IF group would lower the group’s mean PROM results. A pragmatic answer would be that these individuals are more often non-responders than responders. We do agree that reoperation after THA can occur late. Chammout et al. (2012) followed patients with a mean age of 68 years for 17 years. 4 of 43 THA patients had reoperations between 4 and 9 years later. On the other hand, 20 of 57 IF patients had conversion to THA within 2 years, i.e. they were exposed to secondary THA with a known higher risk of revisions than a primary THA. As a final result after IF, 2 had Girdlestone procedures, 2 deep infections, and 1 was revised due to aseptic loosening. Thus starting with IF is apparently no guarantee to avoid revision of arthroplasties. In our study, the patients had 1-year follow-up for PROM data and up to 8 years’ follow-up for implant-related problems. As seen in Figure 2 from our article, no major reoperations occurred later than 3 years postoperatively after hemiarthroplasty or later than 5 years postoperatively after total hip arthroplasty. This is in line with another study reporting no major reoperations between 2 and 5–7 years postoperatively in a medium follow-up of cemented hemiarthroplasties (Støen et al. 2014). In addition, patients with healed fractures do not gain a functional advantage or better relief from pain than those with uncomplicated replacements (Keating et al. 2006, Frihagen et al. 2007, Gjertsen et al. 2010, Leonardsson et al. 2010). Finally, to focus on only those with healed fractures after fixation is not meaningful from a patient point of view, as patients cannot choose whether to have successful healing or not. In summary, with all respect for the caveats mentioned by Hedström and colleagues, we still think our conclusion that patients between 55 and 70 years of age with displaced intra-

of Orthopaedic Surgery, Akershus University Hospital, Lørenskog, Norway; 2 Norwegian Hip Fracture Register, Department of Orthopaedic Surgery, Haukeland University Hospital, Bergen; 3 Department of Clinical Sciences, University of Bergen, Bergen, Norway; 4 Department of Orthopaedic Surgery, Oslo University Hospital, Oslo, Norway; 5 Department of Orthopaedics, Skåne University Hospital, Lund University, Malmö, Sweden; 6 Institute of Clinical Medicine, University of Oslo, Norway. Correspondence: stba@ahus.no

Bartels S, Gjertsen J E, Frihagen F, Rogmark C, Utvåg S E. High failure rate after internal fixation and beneficial outcome after arthroplasty in treatment of displaced femoral neck fractures in patients between 55 and 70 years. Acta Orthop 2018; 89(1): 53-8. Campenfeldt P, Hedström M, Ekström W, Al-Ani A. Good functional outcome but not regained health related quality of life in the majority of 20–69 years old patients with femoral neck fracture treated with internal fixation: a prospective 2-year follow-up study of 182 patients. Injury 2017; 48: 2744–53. Chammout G K, Mukka S S, Carlsson T, Neander G F, Stark A W, Skoldenberg O G. Total hip replacement versus open reduction and internal fixation of displaced femoral neck fractures: a randomized long-term follow-up study. J Bone Joint Surg Am 2012; 94(21): 1921-8. 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. Gjertsen J E, Vinje T, Engesaeter L B, et al. Internal screw fixation compared with bipolar hemiarthroplasty for treatment of displaced femoral neck fractures in elderly patients. J Bone Joint Surg Am 2010; 92: 619-28. Keating J F, Grant A, Masson M, Scott N W, Forbes JF . Randomized comparison of reduction and fixation, bipolar hemiarthroplasty, and total hip arthroplasty: treatment of displaced intracapsular hip fractures in healthy older patients. J Bone Joint Surg Am 2006; 88(2): 249-60. Leonardsson O, Sernbo I, Carlsson A, Akesson K, Rogmark C. Long-term follow-up of replacement compared with internal fixation for displaced femoral neck fractures: results at ten years in a randomised study of 450 patients. J Bone Joint Surg Br 2010; 92(3): 406-12. Støen R Ø, Lofthus C M, Nordsletten L, Madsen J E, Frihagen F. Randomized trial of hemiarthroplasty versus internal fixation for femoral neck fractures: no differences at 6 years. Clin Orthop Relat Res 2014; 472: 360-7. Swedish Fracture Register Annual Report; 2017. https://stratum.registercentrum.se/#!page?id=1525

1/19 ACTA ORTHOPAEDICA

Preventing and reducing periprosthetic joint infections

FFor anaerobees

Bone cement with 2 antib tibiotics Broaad sp spectrum trum m of of act ctivity For MRSA/MRRSE

,PSODQW Ƃ[DWLRQ DW High stabilit s ty

High gh iinitial antibiotic ibi release

Revision Vol. 90, No. 1, 2019 (pp. 1–95)

PRODUCTS & SOLUTIONS YOU CAN TRUST.

Volume 90, Number 1, February 2019 www.heraeus-medical.com IORT_90_01_Cover.indd 1

ISSN 1745-3674

11-01-2019 16:16:26