Vol. 89, No. 4, August 2018 Guest editorial International registry collaboration and statistical approaches
Annotation Do we have an opioid crisis in Scandinavia? Time to act?
M Heilig and M Tägil
E W Paxton, M Mohaddes, I Laaksonen, M Lorimer, S E Graves, H Malchau, R S Namba, J Kärrholm, O Rolfson, and G Cafri E N Glassou, A B Pedersen, P K Aalund, S B Mosegaard, and T B Hansens K Sugand, R A Wescott, R Carrington, A Hart, and B H van Duren P H J Cnudde, S Nemes, E Bülow, A J Timperley, S L Whitehouse, J Kärrholm, and O Rolfson
Hip Meta-analysis of individual registry results enhances international registry collaboration Is gain in health-related quality of life after a total hip arthroplasty depended on the comorbidity burden? Teaching basic trauma: validating FluoroSim, a digital fluoroscopic simulator for guide-wire insertion in hip surgery Risk of further surgery on the same or opposite side and mortality after primary total hip arthroplasty: A multi-state analysis of 133,654 patients from the Swedish Hip Arthroplasty Register Patient claims in prosthetic hip infections: a comparison of nationwide incidence in Sweden and patient insurance data Outcomes following hip and knee replacement in diabetic versus nondiabetic patients and well versus poorly controlled diabetic patients: a prospective cohort study Knee Knee extensor muscle weakness and radiographic knee osteo arthritis progression: the influence of sex and malalignment Migration of all-polyethylene compared with metal-backed tibial components in cemented total knee arthroplasty: a randomized controlled trial A 2-year RSA study of the Vanguard CR total knee system: A randomized controlled trial comparing patient-specific positioning guides with conventional technique Peri-apatite coating decreases uncemented tibial component migration: long-term RSA results of a randomized controlled trial and limitations of short-term results Poor outcome after a surgically treated chondral injury on the medial femoral condyle: early evaluation with dGEMRIC and 17-year radiographic and clinical follow-up in 16 knees
374 380 386 394
P Kasina, A Enocson, V Lindgren, and L J Lapidus
E Lenguerrand, A D Beswick, M R Whitehouse, V Wylde, and A W Blom
A Dell’Isola, W Wirth, M Steultjens, F Eckstein, and A G Culvenor K T van Hamersveld, P J Marang–van de Mheen, R G H H Nelissen, and S Toksvig-Larsen
F-D Øhrn, J van Leeuwen, M Tsukanaka, and S M Röhrl
K T van Hamersveld, P J Marang–van de Mheen, R G H H Nelissen, and S Toksvig-Larsen
J Tjörnstrand, P Neuman, B Lundin, J Svensson, L E Dahlberg, and C J Tiderius
H M Rasmussen, J Svensson, M Thorning, N W Pedersen, S Overgaard, and A Holsgaard-Larsen G Hägglund, K Pettersson, T Czuba, M Persson-Bunke, and E Rodby-Bousquet J H Stouten, A T Besselaar, and M C van der Steen
J Holbeck-Brendel, and O Rahbek
Fracture healing Osteoblast precursors and inflammatory cells arrive simultaneously to sites of a trabecular-bone injury
M Bernhardsson, and P Aspenberg
Miscellaneous Osteochondral lesions of the talus: Few patients require surgery Automated detection and classification of the proximal humerus fracture by using deep learning algorithm
S G Seo, J S Kim, D-K Seo, Y K Kim, S-H Lee and H S Lee S W Chung, S S Han, J W Lee, K-S Oh, N R Kim, J P Yoon, J Y Kim, S H Moon, J Kwon, H-J Lee, Y-M Noh, and Y Kim
Children Threshold values of ankle dorsiflexion and gross motor function in 60 children with cerebral palsy: A cross-sectional study Incidence of scoliosis in cerebral palsy: A population-based study of 962 young individuals Identification and treatment of residual and relapsed idiopathic clubfoot in 88 children 5-year-old child with late discovered traumatic patellar tendon rupture—a case report
Obituary Per Aspenberg
Acta Orthopaedica 2018; 89 (4): 367
International registry collaboration and statistical approaches
International registry collaboration is often hampered by regulations preventing transferal of individual patient data between countries. In this issue of Acta, Paxton et al. (2018) report the use of meta-analysis in registry research and compare it with results based on individual-patient level. The meta-analysis approach is well known in medical scientific work, but is not well known among orthopedic surgeons (Arends et al. 2008). Using a meta-analysis approach each registry conducts analysis on its own data given a pre-specified protocol and data syntax. The risk estimates are combined in a meta-analysis. The approach was first used in international hip and knee replacement registry research in the United States. The Food and Drug Administration (FDA) funded collaboration between 6 national and regional registries of the International Consortium of Orthopaedic Registries (Sedrakyan et al. 2014, Cafri et al. 2015). Several studies were published from this collaboration reporting results on articulation and fixation of hip prostheses and stabilization of knee replacements (Sedrakyan et al. 2014). An important question is whether this approach leads to the same results and estimates as use of individual patientlevel data, which have been used in individual registry studies and the Nordic Arthroplasty Register Associations studies (Havelin et al. 2009, Robertsson et al. 2010, Johanson et al. 2017). In the study by Paxton the meta-analysis approach and individual-level data gave the same results, and adding on one additional registry to the study gave more precise estimates. The meta-analysis approach demands that a detailed protocol is prepared and a statistical syntax is made by the leading analysis center; this syntax must be used by each individual registry participating in the study. This approach is not flexible and new sub-analysis and small corrections to the protocol demand that the syntax must be redone centrally and new analysis performed. Different statistical approaches can be used such as fixed-effects and random-effects models (Arends et al. 2008, Cafri et al. 2015). Individual-level analysis is more flexible and preferred, but is often impossible since not all registries are allowed to share individual-level data even if they are anonymized. Thus, the US registry was not allowed to contribute to the analysis with individual-level data for privacy and security reasons in the Paxton study. However, the meta-analysis approach as demonstrated by Paxton gave the
same results as the individual-level analysis. This is reassuring, as it may convince more registries to contribute data to multinational studies. With the meta-analysis approach each registry has control of its own data and data-ownership issues are of less concern. The known problems of confounding of unknown variables in observational studies such as confounding by indication cannot be accounted for in either individual-level studies or meta-analysis. Both study approaches use time-to-event analysis approaches such as Kaplan–Meier and Cox analysis.
Ove Furnes The Norwegian Arthroplasty Register, Department of Orthopaedic Surgery, Haukeland University Hospital, Bergen and Department of Clinical Medicine (K1), Faculty of Medicine, University of Bergen, Norway email: firstname.lastname@example.org
Arends L R, Huninck M G M, Stijnen T. Meta-analysis of summary survival curve data. Stat Med 2008; 27: 4381-96. Cafri G, Banerjee S, Sedrakyan A, Paxton L, Furnes O, Graves S, MarinacDabic D. Meta-analysis of survival curve data using distributed health data networks: application to hip arthroplasty studies of the International Consortium of Orthopaedic Registries. Res Synth Methods 2015; 6(4): 347-56. Havelin L I, Fenstad A M, Salomonsson R, Mehnert F, Furnes O, Overgaard S, Pedersen A, Herberts P, Kärrholm J, Garrelick G. The Nordic Arthroplasty Association. A unique collaboration of 3 national hip arthroplasty registries with 280 201 THRs. Acta Orthop 2009; 80(4): 393-401. Johanson P E, Furnes O, Havelin L I, Fenstad A M, Pedersen A B, Overgaard S, Garellick G, Mäkelä K, Kärrholm J. Outcome in design-specific comparisons between highly crosslinked conventional polyethylene in total hip arthroplasty. Acta Orthop 2017; 88(4): 363-9. Paxton E W, Mohaddes M, Laaksonen I, Lorimer M, Graves S E, Malchau H, Namba R S, Kärrholm J, Rolfson O, Cafri G. Meta-analysis of individual registry results enhances international registry collaboration. Acta Orthop 2018; 89(4): 369-373. Robertsson O, Bizjajeva S, Fenstad AM, Furnes O, Lidgren L, Mehnert F, Odgaard A, Pedersen AB, Havelin LI. Knee arthroplasty in Denmark, Norway and Sweden: a pilot study from the Nordic Arthroplasty Register Association. Acta Orthop 2010; 81(1): 82-9. Sedrakyan A, Paxton E, Graves S, Love R, Marinac-Dabic D. National and international postmarket research and surveillance implementation: achievements of the International Consortium of Orthopaedic Registries Initiative. J Bone Joint Surg Am 2014; 96 (Suppl 1): 1-6.
© 2018 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 10.1080/17453674.2018.1487210
Acta Orthopaedica 2018; 89 (4): 368
Do we have an opioid crisis in Scandinavia? Time to act? In the United States, the prescription of opioids has quadrupled since 2000. This has been paralleled by a steady rise in the rate of opioid-related deaths, now reaching an estimated 65,000 American lives ended yearly by drug overdoses (Nature 2017). About 30,000 of these are due to legally prescribed drugs. Attempts to clamp down on prescriptions have done little to halt the trend, and instead in the past 2 years synthetic opioids, dominated by fentanyl and fentanyl derivatives, have accounted for an even steeper rise in overdose deaths. It is reasonable to ask whether the opioid crisis is a unique US problem, or whether we might face the same problem internationally, and in particular in Scandinavia, presently or in the years to come? The question is in part addressed by a study by Bäckryd et al. (2017). The analysis indicates that in Sweden the prescription of opioids has remained at a constant level since the turn of the century. The total prescription of opioids, measured as morphine equivalents and defined as daily doses, remained constant between 2000 and 2015. Also, the number of patients who at some period during the year were taking opioids remained constant from 2006 to 2015, as did the amount of opioids prescribed per person. The authors therefore concluded that there is no opioid crisis in Sweden similar to what the US is experiencing. This does not invalidate reports that dependence on prescription opioids is a real clinical problem. That 12% of Swedes were receiving an opioid at some period in 2015 may not represent an increase, but is still a lot. Furthermore, the study indicates a shift toward short-acting opioids, such as oxycodone and fentanyl, which are known to have a higher addictive potential. The orthopedic community should aim to minimize unnecessary opioid exposure, especially as opioids are relatively ineffective as analgesia for non-acute, non-cancer pain. In fact, in chronic pain, opioids are never the primary choice. In acute pain episodes, opioids have become increasingly popular, but even here evidence is scant (Brummett et al. 2017). In the summer of 2017, an enquiry was sent to the 6 university and 13 larger hospitals in Sweden that perform surgery on distal radius fractures. The study was an international initiative to evaluate prescription patterns after distal radius surgery, and the results were presented at a symposium on the opioid crisis held at the American Society for Surgery of the Hand in San Francisco in September 2017. All Swedish centers routinely prescribed the full dose of paracetamol. 10/15 prescribed
additional short-acting oxycodone (n = 10), tramadol (n = 1), or paracetamol + codeine (n = 1), in packets of either 14 or 28. 3 centers gave out drugs only for 3 days. 10/15 centers prescribed long-acting oxycodone in packets of either 14 or 28. Repeat prescriptions were reported as rare. The Swedish prescription pattern was found to be identical to the enquiry made among American centers. It thus appears that we do not have an ongoing full-scale opioid crisis in Sweden, at least not of the same catastrophic magnitude as the United States. However, we need to be aware of the risk that ill-indicated prescription of opioids, and in particular short-acting members of this group, may also lead patients into a lifelong opioid addiction in Sweden. A recent study from Denmark has found a considerable proportion of patients with continued or increased opioid consumption 1 year after hip or knee replacement (Jørgensen et al. 2018). It is plausible that we need to act now to stop a future crisis. Our postoperative prescription patterns match the US practice, which has obviously led to a national public-health emergency. It is naive to believe we are immune to a similar deterioration in Scandinavia.
Marcus Heilig Center for Social and Affective Neuroscience, Department of Clinical and Experimental Medicine, Linköping University, Linköping, Sweden. email: email@example.com Magnus Tägil Department of Orthopedics, Clinical Sciences, Lund University and Skåne University Hospital, Lund, Sweden. email: firstname.lastname@example.org
Bäckryd et al. Dynamiken i förskrivningen av opioider i Sverige 2000–2015: Markanta omfördelningar inom opioidgruppen, men ingen “epidemi”. Lakartidningen 2017 May 2; 114. pii: EFUE. Swedish. PubMed PMID: 28485763. Abstract in English. Brummett et al. New persistent opioid use after minor and major surgical procedures in US adults. JAMA Surg 2017; 152(6): e170504. doi:10.1001/ jamasurg.2017.0504 Jørgensen et al. Analgesic consumption trajectories in 8,975 patients 1-year after fast-track total hip or knee arthroplasty. European J Pain (in press). Nature. Editorial: Opiate deaths demand serious action. Nature 2017; 551(7682): 541–2. doi: 10.1038/d41586-017-07657-z.
© 2018 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 10.1080/17453674.2018.1465285
Acta Orthopaedica 2018; 89 (4): 369–373
Meta-analysis of individual registry results enhances international registry collaboration Elizabeth W PAXTON 1,2, Maziar MOHADDES 2,3, Inari LAAKSONEN 4, Michelle LORIMER 5, Stephen E GRAVES 6, Henrik MALCHAU 3,7, Robert S NAMBA 8, John KÄRRHOLM 2,3, Ola ROLFSON 2,3, and Guy CAFRI 1
Permanente, San Diego, CA, USA; 2 Institute of Clinical Sciences, Department of Orthopaedics, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden; 3 Swedish Hip Arthroplasty Register, Gothenburg, Sweden; 4 Turku University Hospital, Turku, Finland; 5 South Australia Health & Medical Research Institute, Adelaide, SA, Australia; 6 University of South Australia, Adelaide, SA, Australia; 7 Harvard Medical School, Boston, MA, USA; 8 Southern California Permanente Medical Group, Irvine, CA, USA Correspondence: Liz.W.Paxton@kp.org Submitted 2017-10-17. Accepted 2018-02-19.
Background and purpose — Although common in medical research, meta-analysis has not been widely adopted in registry collaborations. A meta-analytic approach in which each registry conducts a standardized analysis on its own data followed by a meta-analysis to calculate a weighted average of the estimates allows collaboration without sharing patient-level data. The value of meta-analysis as an alternative to individual patient data analysis is illustrated in this study by comparing the risk of revision of porous tantalum cups versus other uncemented cups in primary total hip arthroplasties from Sweden, Australia, and a US registry (2003–2015). Patients and methods — For both individual patient data analysis and meta-analysis approaches a Cox proportional hazard model was fit for time to revision, comparing porous tantalum (n = 23,201) with other uncemented cups (n = 128,321). Covariates included age, sex, diagnosis, head size, and stem fixation. In the meta-analysis approach, treatment effect size (i.e., Cox model hazard ratio) was calculated within each registry and a weighted average for the individual registries’ estimates was calculated. Results — Patient-level data analysis and meta-analytic approaches yielded the same results with the porous tantalum cups having a higher risk of revision than other uncemented cups (HR (95% CI) 1.6 (1.4–1.7) and HR (95% CI) 1.5 (1.4–1.7), respectively). Adding the US cohort to the meta-analysis led to greater generalizability, increased precision of the treatment effect, and similar findings (HR (95% CI) 1.6 (1.4–1.7)) with increased risk of porous tantalum cups. Interpretation — The meta-analytic technique is a viable option to address privacy, security, and data ownership concerns allowing more expansive registry collaboration, greater generalizability, and increased precision of treatment effects. ■
Orthopedic registries play a critical role in the identification of clinical best practices, outcome assessment, and device surveillance (Herberts and Malchau 1999, 2000, Graves 2010, Paxton et al. 2012, 2013). Collaborations among registries provide additional opportunities to increase statistical power, improve generalizability, and to examine variation in clinical practices and outcomes between countries (Havelin et al. 2009). Previously, registries have collaborated by sending deidentified standardized patient-level data to a centralized database and conducting statistical analyses based on the pooled individual patient-level data (Dale et al. 2012, Bergh et al. 2014, Wangen et al. 2017). Although analysis of individual patient data is an ideal approach, many registries cannot share even de-identified patient level data due to privacy, security, and data ownership regulations (Sedrakyan et al. 2014). One alternative is to collect effect sizes from similarly designed registry studies and perform a meta-analysis. Meta-analysis is a common approach used in medical research to summarize the findings of several independent studies into a single estimate of the treatment effect (Hedges and Vevea 1998, Borenstein et al. 2009). Well-designed metaanalyses can provide more precise estimates of the treatment effects of individual studies, resulting in a higher level of scientific evidence than individual clinical studies. Typically, meta-analysis consists of weighted averages of the independent study effect sizes, which can be combined using either a fixed- or random-effects model. In a fixed-effect model it is assumed that there is one true effect size common to all studies in the meta-analysis and the combined effect estimates this parameter. In a random-effect model, the effect size is assumed to vary from study to study due to study-specific differences and the combined effect is an estimate of the mean of this dis-
© 2018 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 10.1080/17453674.2018.1454383
12191 Paxton D.indd 369
tribution. The choice of which model to apply should be based on the perceived process that generates the data as well as the type of inferences desired (Hedges and Vevea 1998, Borenstein et al. 2009). While random-effect models are appealing since studies may differ (e.g., in patient characteristics, inclusion criteria, and methods) and we often want to generalize beyond data of the studies included in a meta-analysis, standardizing the design and analysis in each study can mitigate the impact of study-specific variation. Further, when there are a very small number of studies, the variation among the effects will be estimated imprecisely and this can adversely impact inferences. In these cases, the fixed-effects model is an alternative (Borenstein et al. 2009) despite having more restricted inferences than the random-effects model. Although meta-analysis of independent studies is frequently used in medical research, often there are a limited number of available studies with a comparable design/analysis that make use of orthopedic registry data (e.g., examining similar treatments, conditioning on the same covariates). As a result, alternative methods for data sharing and collaboration must be considered. One alternative approach is to use a standardized design and analysis that each registry applies on its own registry data, in turn generating model results (e.g., hazard ratios) that can be meta-analyzed across the registries. In this metaanalysis of standardized studies approach, the estimate of the effect is calculated within each registry followed by an averaging of the estimate across the registries for a combined result. Recently, a variant of meta-analysis of standardized studies was adopted in which aggregate-level survival curve data were meta-analyzed (Banerjee et al. 2014, Cafri et al. 2015). Despite the potential benefits of meta-analysis of standardized studies and prior successful implementations (Allepuz et al. 2014, Graves et al. 2014, Namba et al. 2014, Paxton et al. 2014, Sedrakyan et al. 2014), the method is not well understood or widely adopted among orthopedic registries. One way to motivate this approach is through a comparison of results obtained by meta-analysis of standardized studies and analysis of individual patient data. Therefore, the purpose of this study is to illustrate the value of meta-analysis of standardized studies as an alternative to analysis of individual patient data. The example in this study compares the risk of revision of porous tantalum cups versus other uncemented cups in primary total hip arthroplasties using data from Sweden, Australia, and a US cohort.
Patients and methods Primary total hip replacements with a porous tantalum design cup and other uncemented cups implanted between 2003 and 2015 were identified using the Australian Orthopedic Association National Joint Replacement Registry (AOANJRR), the Swedish Hip Arthroplasty Registry (SHAR), and the Kaiser Permanente Total Joint Registry. The capture rate of these
12191 Paxton D.indd 370
Acta Orthopaedica 2018; 89 (4): 369â&#x20AC;&#x201C;373
registries exceeds 95% and loss to follow-up is less than 8% over the study period. Validation and quality-control methods of these registries have been published previously (Soderman et al. 2000, Paxton et al. 2012, 2013, Australian Orthopaedic Association 2016). The study sample was restricted to metal on highly crosslinked polyethylene primary THAs. Patientlevel data were combined from AOANJRR and SHAR into a centralized database to compare the analysis of individual patient data with meta-analysis of standardized studies. Deidentified patient demographics, implant characteristics, and reasons for revisions were extracted from each registry. The US registry was prohibited from providing case-level data and therefore only provided summary-level data for the meta-analytic approach. Statistics The primary objective of this article is a comparison of analysis of individual patient data with meta-analysis of standardized studies. Ideally such a comparison is undertaken when the estimate in both approaches is the same. Beyond comparability, our proposed statistical analyses address confounding due to measured prosthesis/patient characteristics and static study characteristics (e.g., average age), as well as dependency among observations on the response due to the nesting of observations within a registry. For both individual patient data analysis and meta-analysis approaches a Cox proportional hazard model was fit for the endpoint of time to revision (for any component and reason), the treatment effect compares porous tantalum cups with other uncemented cups, and covariates to include age (continuous), sex, diagnosis, (osteoarthritis, rheumatoid arthritis, osteo-/avascular necrosis, hip dysplasia, other), head size (28, 32, > 32), and stem fixation (cemented, uncemented). Missing data were listwise deleted. There were small amounts of missing data on age (n = 27) and sex (n = 32), but more substantial missing data on whether cement was used on the stem (n = 3,438). In the meta-analysis approach a treatment effect size (i.e., hazard ratio from the Cox model) was calculated within each registry, therefore there is no dependency of observations on the response within registry and no confounding due to static study-level characteristics. To address these issues in the individual patient data approach we stratified on study (Glidden and Vittinghoff 2004, SjĂślander et al. 2013), which also leads to an estimate comparable to the one obtained from meta-analysis of standardized studies because both approaches allow for each study to have its own distinct baseline hazard. There are some alternatives to stratification that might be considered for analysis of individual patient data, but none of these provide estimates that are more comparable to the meta-analysis approach than the stratification approach adopted. One alternative is a between-within frailty model (SjĂślander et al. 2013), but this introduces more parametric assumptions (i.e., distribution of frailties and functional form of cluster effects) than stratification. Another option is use of cluster robust standard errors
Acta Orthopaedica 2018; 89 (4): 369–373
(Lee et al. 1992), but this does not address study-level confounding. Lastly, inclusion of a dummy indicator for study is possible, but this invokes a proportional hazard assumption for the cluster effect. Calculating the average treatment effect from a fixedeffect model for a meta-analysis of standardized studies is straightforward (Hedges and Olkin 1985). For each study (i = 1,2,…,k) we estimate a log hazard ratio from a Cox model, LHRi = ln(HRi). The variance of this estimate is denoted by V (LHRi) and a weight is constructed by taking the inverse of this quantity, Wi = 1 / V (LHRi). The average treatment effect across studies is then estimated using k k a weighted mean, LHR = (∑i=1 Wi LHRi) / (∑i=1 Wi). The k variance of this mean is then V (LHR) = 1 / (∑i=1 Wi) and the standard error is SE (LHR) = √V (LHR). Normal theory confidence intervals (95%) can be calculated in the conventional way: LHR ± 1.96 × SE (LHR). Point estimates and interval endpoints are exponentiated for improved interpretability (e.g., HR = exp(LHR)). A 2-tailed p-value is based on p = 2 × [1 – (Φ(| Z |))] , where p is the standard normal cumulative distribution and Z = (LHR) / SE (LHR) . Ethics, funding, and potential conflicts of interest Approval from the Institutional Review Board was obtained prior to the start of this study. IRB #5488 approved on August 27, 2009. The study was also approved by the Regional Ethical Review Board in Gothenburg, Sweden (entry number 66916). There is no funding. There are no potential conflicts of interest.
Results The porous tantalum group consisted of 2,796 from SHAR, 7,317 from the AOANJRR, and 13,088 from the US registry. Other uncemented cups consisted of 13,156 from SHAR, 70,440 from the AOANJRR, and 44,725 from the US registry. Patient, implant, and fixation, and outcomes of porous tantalum versus other uncemented cup has been reported on SHAR and AOANJRR cohorts (Laaksonen et al. 2018). Therefore, descriptive statistics and Kaplan–Meier survival focus on solely on the US cohort. The US cohort (Table 1) was similar to SHAR and AOANJRR in age, sex, diagnosis, and followup. Tables 2 and 3 display cup designs, reasons for revisions, and type of revision for the US cohort. The unadjusted survival of the cups suggested a difference among the groups (Figure). The US cohort also had similar covariate adjusted results to SHAR and AOANJRR, with a higher risk of revision for the porous tantalum group (HR = 1.6 (95% CI 1.4–1.8)). When limiting the revision endpoint to cup revisions (i.e., alone or in combination with any other components), covariate adjusted results in the US cohort also indicated a higher risk of revision for the porous tantalum group (HR = 1.4 (95% CI 1.0–1.8). The comparison of analysis of patient-level data and the meta-
12191 Paxton D.indd 371
Table 1. US cohort cup designs Cup design
Porous tantalum Continuum Trabecular metal (shell) Other uncemented Trident Pinnacle Trilogy Reflection Allofit Exceed
9,740 (17) 3,348 (5) 1,107 (2) 34,350 (59) 2,203 (4) 7,065 (13) NA NA
Table 2. US cohort descriptive data
Factor n (%) Mean age (range) Male (%) Right side (%) Diagnosis, n (%) OA RA Femoral neck fracture Dysplasia Osteonecrosis Other Follow up, years (range) Uncemented stem, n (%) Femoral head size, mm, n (%) 28 32 > 32
Porous tantalum cups
Other uncemented cups
13,088 (23) 66 (16–97) 5,447 (42) 7,139 (55)
44,725 (77) 67 (13–98) 18,369 (41) 24,412 (55)
12,000 (92) 203 (2) NA NA 249 (2) 524 (4) 112 (1) 2.8 (0–14) 12,712 (97)
40,987 (92) 656 (1) NA NA 690 (2) 2,059 (5) 333 (1) 4.6 (0–15) 41,328 (92)
430 5,403 7,255
(3) (41) (55)
6,186 16,190 22,349
(14) (36) (50)
Table 3. Reasons for revision and type of revision in US cohort a
Factor Revised Reason for revision Infection Fracture Instability Loosening Others Type of revision Cup + stem exchange Stem exchange Cup exchange Liner +/– head exchange Femoral head exchange Extraction Others
Porous tantalum cups n (%)
Other uncemented cups n (%)
76 11 118 51 118
(20) (3) (32) (14) (32)
233 47 337 108 254
(24) (5) (34) (11) (26)
13 121 47 107 8 3 75
(3) (32) (13) (29) (2) (1) (20)
47 209 154 346 32 4 187
(5) (21) (16) (35) (3) (0) (19)
and “Reason for revision” entries correspond to validated revision information. “Type of revision” is based on surgeon selfreported procedure.
Acta Orthopaedica 2018; 89 (4): 369–373
Table 4. Comparison of traditional and meta-analytic approaches Approach
HR (95% CI)
Sweden (SHAR) Australia (AOANJRR) US cohort Individual patient data (AOANJRR and SHAR) Meta-analysis (AOANJRR and SHAR) (AOANJRR, SHAR, US Cohort)
US cohort porous tantalum versus other uncemented cup survival. Number at risk:
PT cups 13,088 Other cups 44,725
7,178 3,285 1,051 31,669 22,035 14,010
analytic approaches for SHAR and AOANJRR resulted in similar findings, with porous tantalum having a higher risk of revision than other uncemented cups in the covariate-adjusted models (Table 4). When the US cohort’s data was added to Swedish and Australian data, results further indicated a higher risk of revision for porous tantalum cups versus all other uncemented cups (HR = 1.6 (95% CI 1.4–1.7). The addition of the US registry data results in greater generalizability and increased precision of estimates (Table 4).
Discussion This study has important implications for future international registry collaborations. First, similar results were obtained with analysis of individual patient data and meta-analysis of standardized studies using data from the same registries. This is because both approaches are comparable: they estimate the average cup effect by allowing each registry to have its own distinct baseline hazard. Although an analysis of individualized patient data provides more flexibility since the analyses do not need to be pre-specified, the meta-analytic approach allows each registry to control how data are analyzed and shared. Additional benefits of the meta-analytic approach are minimizing privacy and security issues to enhance international registry collaborations, which are critical for increased statistical power and generalizability in detection of implant problems, identification of variation in clinical care and outcomes, and for conducting comparative effectiveness studies. A limitation of the fixed-model approach applied in this study
12191 Paxton D.indd 372
1.45 (1.14–1.85) 1.57 (1.38–1.79) 1.60 (1.41–1.80)
0.37 0.45 0.47
0.124 0.066 0.063
0.003 < 0.001 < 0.001
1.54 (1.38–1.73) 1.57 (1.44–1.70)
< 0.001 < 0.001
is the more restricted inferences than in a random-effects model. This method also assigns weights based on individual study variance, resulting in more weight to larger registry studies. Both individual patient data analysis and meta-analysis of standardized studies are characterized by some important assumptions in their implementation in this article, among which are: (1) proportional hazards assumption for the treatment variable and covariates, (2) correct functional form for continuous variables (i.e., age effect is linear) and (3) no interactions among the explanatory variables. Although not explored in this article, alternative statistical models could be adopted that mitigate or eliminate the impact of these assumptions. For instance, a time-dependent treatment effect could be modeled if the treatment effect varied over time. In addition to contributing to registry methodological advancements, this study also has clinical implications. The AOANJRR and SHAR study reported porous tantalum cups having a higher risk of revision than other uncemented cups (Laaksonen et al. 2018). This study is the first to confirm these findings in a large US cohort. Our study also found that when focusing on cup revisions (with or without revision to other components), the porous tantalum group still had a higher risk of revision than other uncemented cups. This finding differs from a recent UK study focused on a single manufacturer as the control group whereas our study used all uncemented cups as the comparison group (Matharu et al. 2018). Differences in the study comparison groups, design, statistical analyses, and populations most likely explain the differences in findings. Although the porous tantalum cup may be effective in revision THAs or complicated primaries, the consistent findings of increased risk across 3 countries suggests the need to further investigate the use of this cup in primary THA procedures. The strengths of this study include the inclusion of high-quality data from 3 different countries allowing assessment of generalizability of the findings. Limitations of this study include the intermediate term follow-up in assessing risk of revision in porous tantalum versus other uncemented cups. However, early results seem to indicate a difference in risk of revision and should become further evaluated in longer term studies. Porous tantalum cups may also be used in more
Acta Orthopaedica 2018; 89 (4): 369–373
complex cases, which could potentially account for the difference in risk of revision. Future studies including radiographic analyses may shed light on complexity of the total hip arthroplasty within these groups. In summary, meta-analysis provides an opportunity to collaborate across registries when patient-level data sharing is not feasible. Combining data from multiple registries can enhance precision of estimated effects but is less flexible for conducting statistical analyses. While patient-level data analysis is preferable, meta-analyses provides an attractive alternative option.
EP: conception of study, interpretation of data, and manuscript preparation. MM, IL, SG, HM, RN, JK, OR: interpretation of data and manuscript preparation. ML, GC: statistical analyses, interpretation of data, and manuscript preparation.
Acta thanks Anne Lübbeke and other anonymous reviewers for help with peer review of this study.
Allepuz A, Havelin L, Barber T, Sedrakyan A, Graves S, Bordini B, Hoeffel D, Cafri G, Paxton E. Effect of femoral head size on metal-on-HXLPE hip arthroplasty outcome in a combined analysis of six national and regional registries. J Bone Joint Surg Am 2014; 96(Suppl 1): 12-18. Australian Orthopaedic Association. National Joint Replacement Registry Annual Report. Adelaide, SA, Australia: AOA; 2016. Banerjee S, Cafri G, Isaacs A J, Graves S, Paxton E, Marinac-Dabic D, Sedrakyan A. A distributed health data network analysis of survival outcomes: the International Consortium of Orthopaedic Registries perspective. J Bone Joint Surg Am 2014; 96(Suppl 1): 7-11. Bergh C, Fenstad A M, Furnes O, Garellick G, Havelin L I, Overgaard S, Pedersen A B, Mäkelä K T, Pulkkinen P, Mohaddes M, Kärrholm J. Increased risk of revision in patients with non-traumatic femoral head necrosis: 11,589 cases compared to 416,217 cases with primary osteoarthritis in the NARA database, 1995-2011. Acta Orthop 2014; 85(1): 11-17. Borenstein M, Hedges L V, Higgins J P T, Rothstein H R. Introduction to meta-analysis. Chichester: Wiley; 2009. Cafri G, Banerjee S, Sedrakyan A, Paxton L, Furnes O, Graves S, MarinacDabic D. Meta-analysis of survival curve data using distributed health data networks: application to hip arthroplasty studies of the International Consortium of Orthopaedic Registries. Res Synth Methods 2015; 6: 347-56. Dale H, Fenstad A M, Hallan G, Havelin L I, Furnes O, Overgaard S, Pedersen A B, Kärrholm J, Garellick G, Pulkkinen P, Eskelinen A, Mäkelä K, Engesæter L B. Increasing risk of prosthetic joint infection after total hip arthroplasty: 2,778 revisions due to infection after 432,168 primary THAs in the Nordic Arthroplasty Register Association (NARA). Acta Orthop 2012; 83(5): 449-58. Glidden D V, Vittinghoff E. Modelling clustered survival data from multicentre clinical trials. Stat Med 2004; 23(3): 369-88. Graves S E. The value of arthroplasty registry data. Acta Orthop 2010; 81(1): 8-9.
12191 Paxton D.indd 373
Graves S, Sedrakyan A, Baste V, Gioe T J, Namba R, Martinez Cruz O, Stea S, Paxton E, Banerjee S, Isaacs A J, Robertsson O. International comparative evaluation of knee replacement with fixed or mobile-bearing posteriorstabilized prostheses. J Bone Joint Surg Am 2014; 96(Suppl 1): 59-64. Havelin L I, Fenstad A M, Salomonsson R, Mehnert F, Furnes O, Overgaard S, Pedersen A B, Herberts P, Karrholm J, Garellick G. The Nordic Arthroplasty Register Association: a unique collaboration between 3 national hip arthroplasty registries with 280,201 THRs. Acta Orthop 2009; 80(4): 393401. Hedges L V, Olkin I. Statistical methods for meta-analysis. New York: Academic Press; 1985. Hedges L V, Vevea J L. Fixed and random-effects models in meta-analysis. Psychol Methods 1998; 3(4): 486-504. Herberts P, Malchau H. Many years of registration have improved the quality of hip arthroplasty. Lakartidningen 1999; 96(20): 2469-73, 75-6. Herberts P, Malchau H. Long-term registration has improved the quality of hip replacement: a review of the Swedish THR Register comparing 160,000 cases. Acta Orthop Scand 2000; 71(2): 111-21. Laaksonen I, Lorimer M, Gromov K, Eskelinen A, Rolfson O, Graves S, Malchau H, Mohaddes M. Trabecular metal acetabular components in primary total hip arthroplasty. Acta Orthop 2018; [Ahead of print] Lee E W, Wei L J, Amato D A. Cox-type regression analysis for large number of small groups of correlated failure time observations. In: Eds. Klein J P, Goel P K. Survival analysis: state of the art. Dordrecht: Kluwer; 1992. Matharu G S, Judge A, Murray D W, Pandit H G. Trabecular metal acetabular components reduce the risk of revision following primary total hip arthroplasty: a propensity score matched study from the National Joint Registry for England and Wales. J Arthroplasty 2018; 33(2): 447-452 Namba R S, Inacio M C S, Cafri G. Increased risk of revision for high flexion total knee replacement with thicker tibial liners. Bone Joint J 2014; 96(2): 217-23. Paxton E W, Inacio M C, Kiley M L. The Kaiser Permanente Implant Registries: effect on patient safety, quality improvement, cost effectiveness, and research opportunities. Perm J 2012; 16(2): 36-44. Paxton E W, Kiley M L, Love R, Barber T C, Funahashi T T, Inacio M C. Kaiser Permanente implant registries benefit patient safety, quality improvement, cost-effectiveness. Jt Comm J Qual Patient Saf 2013; 39(6): 246-52. Paxton E, Cafri G, Havelin L, Stea S, Pallisó F, Graves S, Hoeffel D, Sedrakyan A. Risk of revision following total hip arthroplasty: metal-on-conventional polyethylene compared with metal-on-highly cross-linked polyethylene bearing surfaces: international results from six registries. J Bone Joint Surg Am 2014; 96(Suppl 1): 19-24. Sedrakyan A, Paxton E, Graves S, Love R, Marinac-Dabic D. National and international postmarket research and surveillance implementation: achievements of the International Consortium of Orthopaedic Registries initiative. J Bone Joint Surg Am 2014; 96(Suppl 1): 1-6. Sjölander A, Lichtenstein P, Larsson H, Pawitan Y. Between-within models for survival analysis. Stat Med 2013; 32(18): 3067-76. Soderman P, Malchau H, Herberts P, Johnell O. Are the findings in the Swedish National Total Hip Arthroplasty Register valid? A comparison between the Swedish National Total Hip Arthroplasty Register, the National Discharge Register, and the National Death Register. J Arthroplasty 2000; 15(7): 884-9. Wangen H, Havelin L I, Fenstad A M, Hallan G, Furnes O, Pedersen A B, Overgaard S, Karrholm J, Garellick G, Makela K, Eskelinen A, Nordsletten L. Reverse hybrid total hip arthroplasty. Acta Orthop 2017; 88(3): 248-54.
Acta Orthopaedica 2018; 89 (4): 374–379
Is gain in health-related quality of life after a total hip arthroplasty depended on the comorbidity burden? Eva N GLASSOU 1, Alma B PEDERSEN 2, Peter K AALUND 1, Sebastian B MOSEGAARD 1, and Torben B HANSEN 1
Clinic for Hand, Hip and Knee Surgery, Regional Hospital West Jutland, Aarhus University, Holstebro; 2 Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark Correspondence: email@example.com Submitted 2017-10-04. Accepted 2018-03-06.
Background and purpose — Using patient-reported healthrelated quality of life (HRQoL), approximately 10% of patients report some degree of dissatisfaction after a total hip arthroplasty (THA). The preoperative comorbidity burden may play a role in predicting which patients may have limited benefit from a THA. Therefore, we examined whether gain in HRQoL measured with the EuroQol-5D (EQ-5D) at 3 and 12 months of follow-up depended on the comorbidity burden in THA patients Patients and methods — 1,582 THA patients treated at the Regional Hospital West Jutland from 2008 to 2013 were included. The comorbidity burden was collected from an administrative database and assessed with the Charlson Comorbidity Index (CCI). The CCI was divided into 3 levels: no comorbidity burden, low, and high comorbidity burden. HRQoL was measured using the EQ-5D preoperatively and at 3 and 12 months’ follow-up. Association between low and high comorbidity burden compared with no comorbidity burden and gain in HRQoL was analyzed with multiple linear regression. Results — All patients, regardless of comorbidity burden, gained significantly in HRQoL. A positive association between comorbidity burden and gain in HRQoL was found at 3-month follow-up for THA patients with a high comorbidity burden (coeff: 0.09 (95% CI 0.02 – 0.16)) compared with patients with no comorbidity burden. Interpretation — A comorbidity burden prior to THA does not preclude a gain in HRQoL up to 1 year after THA. ■
Using patient-reported outcome (PRO) measures, approximately 10% of total hip arthroplasty patients (THA) report some degree of dissatisfaction after surgery (Mancuso et al. 1997, Anakwe et al. 2011, Arden et al. 2011, Rolfson et al.
2011). Dissatisfaction is primarily related to unsuccessful pain relief and fulfillment of patient expectations after the THA (Anakwe et al. 2011) and a high preoperative PRO measure indicating low impact of the underlying hip disease (Arden et al. 2011). Several patient- and clinical-related factors have an impact on HRQoL. Age, sex, BMI, and socioeconomics all play a role (Singh and Lewallen 2009, Schafer et al. 2010, Gordon et al. 2013, Judge et al. 2013, Gordon et al. 2014, Greene et al. 2014, Mannion et al. 2015) as well as the preoperative pain and mobility (Berliner et al. 2016). Several studies have showed that HRQoL, pain, and satisfaction after a THA are affected by specific preoperative comorbidities (Singh and Lewallen 2013, Judge et al. 2013, Peter et al. 2015). However, using 3 different diagnosis-based comorbidity indices including CCI, Greene et al. (2015) found only a marginal association between a preoperative comorbidity burden and HRQoL in more than 22,000 THA patients registered in the Swedish Hip Arthroplasty Register from 2002 to 2007. Although PROs have been increasingly used to evaluate surgery outcome from the patient perspective, this is still not part of the prospective and nationwide data collection in the Danish Hip Arthroplasty Register (Gundtoft et al. 2016). However, at the Regional Hospital of West Jutland, covering approximately 5% (285,000 inhabitants) of the Danish population, PROs have been prospectively collected on all THA patients since 2008. To our knowledge, this Danish cohort is the largest so far including PRO data following THA and thus suitable for testing the hypothesis that HRQoL depends on the comorbidity burden in a Danish setting. The purpose of this single-center study was, therefore, to examine whether the patient-reported HRQoL at 3 and 12 months was dependent on the comorbidity burden in patients treated with a THA due to osteoarthritis (OA).
© 2018 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 10.1080/17453674.2018.1457885
12161 Glasou D.indd 374
Acta Orthopaedica 2018; 89 (4): 374–379
Patients registered in DNPR with primary THA due to OA September 2008 to December 2013 n = 1,843 Excluded (n = 261): – declined or did not answer any questionnaire, 48 – received contralateral THA during the study, 213
THA patients who accepted participationand answered at least 1 of 3 questionnaires n = 1,582 (86%) Response rates: – preoperative, 1,363 (74%) – 3 months follow-up, 1,192 (65%) – 12 months follow-up, 1,414 (77%)
Study population. 1,582 of 1,843 patients with a primary total hip arthroplasty (THA) due to osteoarthritis (OA) treated at the Regional Hospital West Jutland from September 2008 to December 2013 and registered in the Danish National Patient Register (DNPR) were included in the study.
Patients and methods Study population and setting Patients with a unilateral primary THA due to OA treated at the Regional Hospital West Jutland from September 2008 to December 2013 and registered in the Danish National Patient Register (DNPR) formed the basis of this study (Figure). Patients with revision or counter-lateral THA within the first year were excluded. All patients were assigned to a welldocumented fast-track hip arthroplasty program (Husted et al. 2010). Exposure Comorbidity was established with the CCI (Charlson et al. 1987). Based on the unique 10-digit personal identification number all citizens are assigned at birth, each procedure from the cohort was linked to the DNPR to collect information about comorbidities. Each record in the DNPR holds information about hospital treatment, surgical procedures, and discharge diagnoses (Schmidt et al. 2015). All primary and secondary diagnoses from hospitalizations and outpatient visits registered as ICD-10 codes in the DNPR over a 10-year period before the primary procedure formed the basis of the CCI calculation. The CCI score was calculated by adding the points of each disease category for each procedure. All THA procedures were then divided into 3 comorbidity burden groups based on the score: patients with no comorbidity burden, patients with a low comorbidity burden (equal score 1 and 2 in the CCI), and patients with a high comorbidity burden (equal score 3 or higher in CCI). Furthermore, to see if specific diseases diverted from the CCI index score, we classified the THA procedures according to 3 specific disease groups nested in the CCI: diabetes (type I and II diabetes and diabetes with
12161 Glasou D.indd 375
end-stage organ damage), cardiovascular diseases (myocardial infarction, congestive heart failure, peripheral vascular disease, and cerebrovascular disease) and chronic obstructive pulmonary disease (COPD). Outcome The outcome was HRQoL measured with the EQ-5D 3-level version. We defined the outcome as both the EQ-5D levels and the difference between the preoperative EQ-5D score and scores at 3 and 12 months’ follow-up. EQ-5D is a short generic questionnaire, consisting of 5 dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/ depression) which can take 1 of 3 responses (no problems, some or moderate problems, and extreme problems) (http:// www.euroqol.org). The responses are converted into a single weighted Danish index score with a minimum value at –0.594 and a maximum value at 1.0. The mean index score is 0.83 for 70–79-year-olds in the general Danish population (Sørensen et al. 2009). In preparation for contemporary and future research, patient-reported HRQoL outcomes including EQ-5D were collected from all THA patients at the Regional Hospital West Jutland from 2008 to 2013. Patients filled in paper questionnaires in relation to ambulatory visits preoperatively and at 3 and 12 months’ follow-up. Details in cohort recruitment have been described elsewhere (Larsen et al. 2010). Statistics Patient characteristics are presented as frequencies. The EQ-5D scores are presented as means. Gain in EQ-5D was calculated as the difference between the preoperative EQ-5D score and the EQ-5D score at 3 and 12 months’ follow-up. Analysis of association between comorbidity burden and gains in EQ-5D score at 3 and 12 months’ follow-up were carried out with complete-case multiple linear regressions and adjusted for age (in 5 categorical groups: < 50, 50–59, 60–69, 70–79, and ≥ 80 years), sex and type of fixation (in categories: cemented THAs, uncemented THAs, and hybrid THAs). All estimates are presented with 95% confidence intervals (CI). The gain in EQ-5D scores was tested to be normally distributed using QQ-plots. Concerning the confounders, there were only missing data in relation to type of fixation (9 observations) and we have therefore refrained from imputation of missing data. As the regression towards the mean (RTM) phenomenon may play a role when interpreting the outcome measure, we quantified the size of the RTM in relation to the 3 exposure groups according to Trochim (2006). Effect modifications from age and sex on the association between comorbidity burden and gain in EQ-5D were examined before the regression analysis. The effects were found to be homogeneous. Additionally, we tested for interaction of age and sex on the gain in EQ-5D at 3 months’ follow-up. Here we found a statistically significant but clinically irrelevant association between sex and age meaning that gain in EQ-5D was 0.003 larger per year for males than for females.
Acta Orthopaedica 2018; 89 (4): 374–379
Table 1. Patient demography
Sex Female Male Age a Age in categories 10–49 50–59 60–69 70–80 80+ Year of surgery 2008 2009 2010 2011 2012 2013 Type of fixation b Cemented THAs Uncemented THAs Hybrid THAs
All THA patients n = 1,582 n %
Comorbidity burden No Low High n = 1,129 n = 379 n = 74 n % n % n %
758 48 824 52 70 (9)
556 49 573 51 69 (9)
168 44 211 56 73 (9)
34 46 40 54 73 (8)
28 58 20 42 76 (11)
49 186 563 583 201
3 12 35 37 13
46 145 442 385 111
4 13 39 34 10
3 35 103 160 78
1 9 27 42 21
0 6 18 38 12
0 8 24 52 16
2 4 12 15 15
4 9 25 31 31
77 363 278 246 333 285
5 23 18 15 21 18
52 276 206 178 218 199
5 24 18 16 19 18
20 72 59 55 101 72
5 19 16 14 27 19
5 15 13 13 14 14
7 20 17 18 19 19
14 5 4 8 8 9
29 10 8 17 17 19
216 14 719 46 638 40
131 12 551 49 441 39
a Age as a continuous variable, mean (SD) b Numbers not equal to the total sum of THAs
67 18 146 39 164 43
Nonparticipants n = 48 n %
18 25 22 30 33 45
19 40 16 33 13 27
due to 9 missing observations.
1,582 THA patients were included (Table 1). The majority of patients (71%) had no comorbidity burden at time of surgery. 24% of the patients had a low comorbidity burden and 5% had a high comorbidity burden. In relation to non-completers (those who did not complete the questionnaires at one of the time points), there were 219 EQ-5D observations (14%) missing preoperatively. At 3 and 12 months’ follow-up, the missing EQ-5D observations accounted for 390 (25%) and 168 (11%), respectively. Due to non-completers, 1,050 and 1,227 patients, respectively, formed the basis of the analyses at 3 and 12 months of follow-up. At all 3 time points, non-completers were more often women. Preoperatively and at 3-month follow-up, non-completers were slightly older, while non-completers at 12 months of follow-up were at the same age as “completers.” Preoperatively and at 12 months of follow-up, non-completers were more often patients with a comorbidity burden. At 3 months of follow-up, the non-completers were more often patients without a comorbidity burden.
The calculation of the weighted Danish EQ-5D index scores and the analyses were performed using Stata Statistical Software, Release 12.0 (StataCorp LP, College Station, TX, USA).
Non-participants (Table 1) The 48 non-participants differed from the responders in relation to age, sex, comorbidity burden, and year of surgery.
Ethics, funding, and potential conflicts of interest Permissions were obtained from the Committee on Health Research Ethics in Central Denmark Region and from the national Danish Data Protection Agency (reference numbers: 2007-41-1197 and 2012-41-0636).No funds were received to conduct the study. The authors declare that they have no conflicts of interest.
EQ-5D scores Preoperative EQ-5D scores decreased with an increase in comorbidity burden (Table 2). At 3 months’ follow-up, the mean EQ-5D scores were approximately 0.85 irrespectively of comorbidity burden. At 12 months’ follow-up, the mean EQ-5D score for patients without a comorbidity burden reached 0.91 (SD 0.13) while the EQ-5D score had stagnated for patients with a high comorbidity burden. The gains in EQ-5D score at 3 and 12 months’ follow-up were statistically significant for all 3 comorbidity groups, but largest for
Table 2. EQ-5D scores (mean (CI)) and gain ( ) in EQ-5D between preoperative and 3 and 12 months’ follow-up in all patients and in relation to comorbidity group and disease group
Patients All No comorbidity burden Low comorbidity burden High comorbidity burden With diabetes With cardiovascular diseases With chronic obstructive pulmonary disease
12161 Glasou D.indd 376
EQ-5D scores 3 months
preoperative and 3 months
preoperative and 12 months
0.64 (0.63–0.65) 0.65 (0.64–0.67) 0.61 (0.59–0.63) 0.55 (0.49–0.61) 0.59 (0.53–0.64) 0.58 (0.55–0.62)
0.85 (0.64–0.86) 0.86 (0.85–0.87) 0.85 (0.83–0.86) 0.84 (0.79–0.88) 0.82 (0.78–0.89) 0.85 (0.83–0.88)
0.90 (0.89–0.90) 0.91 (0.90–0.91) 0.87 (0.85–0.89) 0.85 (0.81–0.89) 0.84 (0.80–0.88) 0.86 (0.84–0.88)
0.21 (0.20–0.22) 0.20 (0.19–0.22) 0.22 (0.19–0.25) 0.30 (0.22–0.38) 0.23 (0.16–0.30) 0.26 (0.21–0.30)
0.25 (0.24–0.27) 0.25 (0.24–0.26) 0.25 (0.23–0.28) 0.31 (0.23–0.38) 0.24 (0.18–0.30) 0.27 (0.23–0.31)
Acta Orthopaedica 2018; 89 (4): 374–379
Table 3. Distribution of the 5 pre-surgery EQ-5D dimensions in total and according to comorbidity group
Dimensions Mobility No problems Some problems Extreme problems Self-care No problems Some problems Extreme problems Usual activities No problems Some problems Extreme problems Pain/discomfort No problems Some problems Extreme problems Anxiety/depression No problems Some problems Extreme problems a
All THA patients n %
Comorbidity burden No Low High n % n % n %
380 26 1,083 74 3 <1
287 27 761 72 3 <1
79 22 274 78 0 –
14 23 48 77 0 –
1,090 75 356 24 7 <1
791 76 249 24 4 <1
253 73 93 27 2 <1
46 75 14 23 1 2
318 22 997 68 141 10
238 23 772 69 85 8
73 21 232 66 44 13
7 11 43 69 12 20
29 2 1,112 77 301 21
21 2 833 80 184 18
7 2 240 70 96 27
1 2 39 64 21 34
1,141 79 281 20 18 1
842 82 177 17 14 1
259 75 85 24 3 1
40 67 19 32 1 1
P-values are derived from Spearman’s rank correlation.
patients with a high comorbidity burden. The attained gains did not, however, differ statistically significantly between the 3 comorbidity groups at either 3 or 12 months’ follow-up (3 months’ follow up: p = 0.06, 12 months’ follow up: p = 0.2). For patients with a high comorbidity burden, the gain in EQ-5D after 12 months’ follow-up decreased from 0.31 to 0.27 due to RTM. For patients with no or a low comorbidity burden, the RTM phenomenon had no effect on the gain in EQ-5D after 12 months’ follow-up. In relation to the 3 specific comorbid conditions (diabetes, cardiovascular diseases, and COPD), THA patients with COPD had the lowest preoperative EQ-5D score and postoperative score after 3 and 12 months. However, the gain in EQ-5D in COPD patients was similar to the gain achieved in patients with diabetes and cardiovascular diseases (Table 2). EQ-5D dimensions For the 3 dimensions “usual activities,” pain/discomfort,” and
“anxiety/depression” there was a logic association between comorbidity burden and the severity of the problems; increased comorbidity burden gave rise to increased problems with these dimensions (Table 3). Association between comorbidity burden and gain in EQ-5D At 3 months’ follow-up, the comorbidity burden had an impact on the gain in HRQoL (CCI 1–2: coeff: 0.01 (CCI –0.02 to 0.04), CCI3+: coeff: 0.09 (CI 0.02–0.16)) compared with patients without a comorbidity burden (Table 4). After 12 months there was no difference between patients with low or high comorbidity burden compared with patients with no comorbidity burden.
Discussion < 0.001
All THA patients regardless of comorbidity burden gained in HRQoL up to 1 year of surgery. However, patients with high comorbidity burden might gain more in HRQoL within 3 months of surgery than patients without or with low comorbidity burden. The gain in HRQoL at 3 months’ follow-up for THA patients with a high comorbidity burden indicates that comorbidity does not unambiguously predict dissatisfaction after surgery. The stagnation in gain from 3 to 12 months of follow-up for patients with a comorbidity burden may, however, signify that the comorbid conditions matters in relation to HRQoL in the long run. This interpretation is emphasized by the lack of late gain in THA patients with 1 of the 3 specific comorbid diseases. Another relation is that the gain in HRQoL after 3 months of follow-up is primarily caused by the direct pain relief after surgery affecting all patients regardless of comorbidity burden and that the late gain from 3 to 12 months of follow-up is based on functional improvements for the benefit of patients without a comorbidity burden. Vogl et al. (2014) concluded after examining the effect of preoperative patient characteristics in THA patients that changes in EQ-5D were mainly explained by the preoperative score: the lower the preoperative scores, the higher change
Table 4. Associations (multiple linear regression coefficients) between comorbidity burden and gain in EQ-5D at 3 and 12 months of follow-up with 95% CI a 3 months’ follow-up Crude (CI) Adjusted a (CI) No comorbidity burden Low comorbidity burden High comorbidity burden a Adjustments
12161 Glasou D.indd 377
reference 0.02 (–0.01 to 0.05) 0.10 (0.03 to 0.16)
reference 0.01 (–0.02 to 0.04) 0.09 (0.02 to 0.16)
12 months’ follow-up Crude (CI) Adjusted a (CI) reference 0.003 (–0.02 to 0.03) 0.06 (–0.00– to 0.11)
reference –0.003 (–0.03 to 0.03) 0.05 (–0.01 to 0.11)
are made for age (in categories), sex, and type of fixation.
in the scores. This view may also be true for our population. Despite a large gain in HRQoL for patients with a comorbidity burden compared with patients without a comorbidity burden, there was no distinct association between comorbidity burden and gain in HRQoL. Our findings are also in concordance with recent findings from the UK (Loth et al. 2017). Despite methodological limitations and a limited cohort, Loth et al. (2017) reported no between-group differences in HRQoL in 251 THA patients with and without a comorbidity burden even though both groups improved substantially in the Oxford Hip Score and the Forgotten Joint Score from pre-surgery to 12 months’ follow-up. The Danish EQ-5D population norm is 0.83 for 70–79-yearolds (Sørensen et al. 2009). In a study examining factors influencing HRQoL after THA in Sweden and Denmark, Gordon et al. (2013) found that Danish patients had an EQ-5D score of 0.85 12 months postoperatively. In our study, we found an even higher EQ-5D score at 12 months of follow-up independent of comorbidity burden compared twith the Danish population norm and the earlier findings by Gordon et al. We explain this high self-reported HRQoL by the use of a well-defined fast-track program consisting of preoperative information with matching of expectations in relation to length of hospitalization, early supervised mobilization postoperatively, and selfrehabilitation after discharge (Larsen et al. 2008). A matching of expectations is shown to be of importance (Gandhi et al. 2009, Judge et al. 2011, Hawker et al. 2013). Surgical technique and type of both fixation and implant are shown to be associated with HRQoL (Lingard et al. 2009, Smith et al. 2012, Bagaric et al. 2014). By including type of fixation in the regression model and by restricting this study to THAs excluding hip resurfacing implants some of these issues are eliminated. Even though we have not restricted the population in relation to surgical technique, we consider the impact of these factors minimal as all patients were treated with the posterior approach. Compared with non-comorbid patients, patients with a comorbidity burden were older. This is a potential problem because age could be a proxy for an increased comorbidity burden. But, as there were no changes in the distribution of either comorbidity groups or age groups across the study period, we interpret the findings as an unambiguous association between comorbidity burden and HRQoL and not as an association between age and HRQoL. Additionally, we included age in the regression model. We do, however, have a potential problem with the severity of the hip disease. A late stage of OA may reduce the possibility of reaching a high level of HRQoL after a THA. We did, however, find that the EQ-5D scores after 3 months of follow-up were identical across comorbidity groups, indicating a uniform disease stage. It would have been appropriate to include disease stage in the analyses, but unfortunately information on the severity of the hip disease was not available. Our study has some limitations. The use of CCI as a measure of comorbidity may give rise to limitations. The CCI was
12161 Glasou D.indd 378
Acta Orthopaedica 2018; 89 (4): 374–379
developed to quantify the influence of comorbidity on mortality and was validated on breast cancer patients and not THA patients. Even though the index is widely used in orthopedic research it may still affect the validity. The CCI is, however, the preferred comorbidity index in Danish register research although other indices such as the Charnley classification and Elixhauser Comorbidity Index are found to be valid in relation to THA patients (Greene et al. 2015, Yurkovich et al. 2015). Another limitation in using the CCI is the omission of all psychiatric diseases except dementia. An omission of, for example, depression entails an underestimation of the found association between comorbidity and HRQoL. Additionally, the confounding may lead to differential misclassification as psychiatric diseases may affect THA patients with a high somatic comorbidity burden the most. The role of depression in relation to HRQoL after a THA is well examined in Swedish settings. Greene et al. (2016) have shown that the 10% of THA patients using antidepressants had poorer HRQoL before and after surgery and Rolfson et al. (2009) found that the preoperative anxiety/depression dimension in EQ-5D was a strong predictor for less pain relief and satisfaction 1 year after a THA. The non-responders being different from the responders in relation to age, sex, and comorbidity is a problem. The number of non-responders is, however, limited. A more serious limitation may therefore be the missing outcome data – the non-completers. Where a plausible consequence of the non-responders being more comorbid than the responders would be in favor of the association, it is more difficult to deem the result of the non-completers. The slightly larger share of comorbid non-completers at 12 months’ follow-up could change our findings from no comorbidity impact on the gain in HRQoL to an impact at this time point. For the opposite distribution at 3 months of follow-up, where the missing HRQoL data are composed of non-comorbid healthy THA patients with no need of postoperative consultations, the association we found may be weakened. Both ways, the missing EQ-5D values are missing at random and therefore we have abstained from replacing missing HRQoL data with substituted values (Little 1992, Pedersen et al. 2017). The prospective collection of PRO data on all THA patients at the Regional Hospital West Jutland from 2008 was farsighted. The large population gives a unique opportunity to study HRQoL in a Danish setting which is very much needed. Considering the prospective data collection of exposure and outcome variables, the well-described fast-track program and the patient profiles, we find that the results of this single-center study can be applied in a wider Danish context. In summary, our study demonstrates that a comorbidity burden does not preclude a gain in HRQoL after a THA. THA patients with a high comorbidity burden may after 3 months of follow-up gain the same level of HRQoL as THA patients without a comorbidity burden. Comorbid THA patients do not, however, attain the same level of HRQoL as patients without a comorbidity burden one year after the THA, but the gain
Acta Orthopaedica 2018; 89 (4): 374–379
in HRQoL after 3 months may still represent a vital difference for these patients in relation to self-independence, daily living, and outcome in general.
ENG, TBH, and ABP contributed to the conception of the study and the study design. ENG drafted the article. All authors contributed to the discussion and interpretation of the results. All authors revised the manuscript for intellectual content and approved the final version before submission. TBH contributed to the data collection. ENG, PKA, and SBM contributed to the data management.
Acta thanks Max Gordon and other anonymous reviewers for help with peer review of this study. Anakwe R E, Jenkins P J, Moran M. Predicting dissatisfaction after total hip arthroplasty: a study of 850 patients. J Arthroplasty 2011; 26(2): 209-13. Arden N K, Kiran A, Judge A, Biant L C, Javaid M K, Murray D W, Carr A J, Cooper C, Field R E. What is a good patient reported outcome after total hip replacement? Osteoarthritis Cartilage 2011; 19(2): 155-62. Bagaric I, Sarac H, Borovac J A, Vlak T, Bekavac J, Hebrang A. Primary total hip arthroplasty: health related quality of life outcomes. Int Orthop 2014; 38(3): 495-501. Berliner J L, Brodke D J, Chan V, SooHoo N F, Bozic K J. John Charnley Award: Preoperative patient-reported outcome measures predict clinically meaningful improvement in function after THA. Clin Orthop Relat Res 2016; 474(2): 321-9. Charlson M E, Pompei P, Ales K L, MacKenzie C R. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40(5): 373-83. Gandhi R, Davey J R, Mahomed N. Patient expectations predict greater pain relief with joint arthroplasty. J Arthroplasty 2009; 24(5): 716-21. Gordon M, Paulsen A, Overgaard S, Garellick G, Pedersen A B, Rolfson O. Factors influencing health-related quality of life after total hip replacement: a comparison of data from the Swedish and Danish hip arthroplasty registers. BMC Musculoskelet Disord 2013; 14: 316. Gordon M, Greene M, Frumento P, Rolfson O, Garellick G, Stark A. Age- and health-related quality of life after total hip replacement: decreasing gains in patients above 70 years of age. Acta Orthop 2014; 85(3): 244-9. Greene M E, Rolfson O, Nemes S, Gordon M, Malchau H, Garellick G. Education attainment is associated with patient-reported outcomes: findings from the Swedish Hip Arthroplasty Register. Clin Orthop Relat Res 2014; 472(6): 1868-76. Greene M E, Rolfson O, Gordon M, Garellick G, Nemes S. Standard comorbidity measures do not predict patient-reported outcomes 1 year after total hip arthroplasty. Clin Orthop Relat Res 2015; 473(11): 3370-9. Greene M E, Rolfson O, Gordon M, Annerbrink K, Malchau H, Garellick G. Is the use of antidepressants associated with patient-reported outcomes following total hip replacement surgery? Acta Orthop 2016; 87(5): 444-51. Gundtoft P H, Varnum C, Pedersen A B, Overgaard S. The Danish Hip Arthroplasty Register. Clin Epidemiol 2016; 8: 509-14. Hawker G A, Badley E M, Borkhoff C M, Croxford R, Davis AM, Dunn S, Gignac MA, Jaglal SB, Kreder HJ, Sale JE. Which patients are most likely to benefit from total joint arthroplasty? Arthritis Rheum 2013; 65(5): 1243-52. Husted H, Solgaard S, Hansen TB, Søballe K, Kehlet H. Care principles at four fast-track arthroplasty departments in Denmark. Dan Med Bull 2010; 57(7): A4166. Judge A, Cooper C, Arden N K, Williams S, Hobbs N, Dixon D, Gunther K P, Dreinhoefer K, Dieppe P A. Pre-operative expectation predicts: 12-month post-operative outcome among patients undergoing primary total hip replacement in European orthopaedic centres. Osteoarthritis Cartilage 2011; 19(6): 659-67.
12161 Glasou D.indd 379
Judge A, Arden N K, Batra R N, Thomas G, Beard D, Javaid M K, Cooper C, Murray D. The association of patient characteristics and surgical variables on symptoms of pain and function over 5 years following primary hipreplacement surgery: a prospective cohort study. BMJ Open 2013; 3(3): e002453. Larsen K, Sørensen O G, Hansen T B, Thomsen P B, Søballe K. Accelerated perioperative care and rehabilitation intervention for hip and knee replacement is effective: a randomized clinical trial involving 87 patients with 3 months of follow-up. Acta Orthop 2008; 79(2): 149-59. Larsen K, Hansen T B, Søballe K, Kehlet H. Patient-reported outcome after fast-track hip arthroplasty: a prospective cohort study. Health Qual Life Outcomes 2010; 8: 144. Lingard E A, Muthumayandi K, Holland J P. Comparison of patient-reported outcomes between hip resurfacing and total hip replacement. J Bone Joint Surg Br 2009; 91(12): 1550-4. Little R J A. Regression with missing X’s: A review. J Am Stat Assoc 1992; 87(420): 1227-37. Loth F L, Giesinger J M, Giesinger K, MacDonald D J, Simpson A H R W, Howie C R, Hamilton D F. Impact of comorbidities on outcome after total hip arthroplasty. J Arthroplasty 2017; 32(9): 2755-61. Mancuso C A, Salvati E A, Johanson N A, Peterson M G, Charlson ME. Patients’ expectations and satisfaction with total hip arthroplasty. J Arthroplasty 1997; 12(4): 387-96. Mannion A F, Impellizzeri F M, Naal F D, Leunig M. Women demonstrate more pain and worse function before THA but comparable results 12 months after surgery. Clin Orthop Relat Res 2015; 473(12): 3849-57. Pedersen A B, Mikkelsen E M, Cronin-Fenton D, Kristensen N R, Pham T M, Pedersen L, Petersen I. Missing data and multiple imputation in clinical epidemiological research. Clin Epidemiol 2017; 9: 157-66. Peter W F, Dekker J, Tilbury C, Tordoir R L, Verdegaal S H, Onstenk R, Benard M R, Vehmeijer S , Fiocco M, Vermeulen H M, van der Linden-van der Zwaag H M, Nelissen R G, Vliet Vlieland T P. The association between comorbidities and pain, physical function and quality of life following hip and knee arthroplasty. Rheumatol Int 2015; 35(7): 1233-41. Rolfson O, Dahlberg L E, Nilsson J A, Malchau H, Garellick G. Variables determining outcome in total hip replacement surgery. J Bone Joint Surg Br 2009; 91(2): 157-61. Rolfson O, Karrholm J, Dahlberg L E, Garellick G. Patient-reported outcomes in the Swedish Hip Arthroplasty Register: results of a nationwide prospective observational study. J Bone Joint Surg Br 2011; 93(7): 867-75. Schafer T, Krummenauer F, Mettelsiefen J, Kirschner S, Gunther K P. Social, educational, and occupational predictors of total hip replacement outcome. Osteoarthritis Cartilage 2010; 18(8): 1036-42. Schmidt M, Schmidt S A, Sandegaard J L, Ehrenstein V, Pedersen L, Sørensen H T. The Danish National Patient Registry: a review of content, data quality, and research potential. Clin Epidemiol 2015; 7: 449-90. Singh J A, Lewallen D. Age, gender, obesity, and depression are associated with patient-related pain and function outcome after revision total hip arthroplasty. Clin Rheumatol 2009; 28(12): 1419-30. Singh J A, Lewallen D G. Medical comorbidity is associated with persistent index hip pain after total hip arthroplasty. Pain Med 2013; 14(8): 1222-29. Smith A J, Wylde V, Berstock J R, Maclean A D, Blom A W. Surgical approach and patient-reported outcomes after total hip replacement. Hip Int 2012; 22(4): 355-61. Sørensen J, Davidsen M, Gudex C, Pedersen K M, Brønnum-Hansen H. Danish EQ-5D population norms. Scand J Public Health 2009; 37(5): 467-74. Trochim W M K. Web Center for Social Research Methods, 2006; http:// www.socialresearchmethods.net/kb/regrmean.php Vogl M, Wilkesmann R, Lausmann C, Hunger M, Plotz W. The impact of preoperative patient characteristics on health states after total hip replacement and related satisfaction thresholds: a cohort study. Health Qual Life Outcomes 2014; 12: 108. Yurkovich M, Avina-Zubieta J A, Thomas J, Gorenchtein M, Lacaille D. A systematic review identifies valid comorbidity indices derived from administrative health data. J Clin Epidemiol 2015; 68(1): 3-14.
Acta Orthopaedica 2018; 89 (4): 380–385
Teaching basic trauma: validating FluoroSim, a digital fluoroscopic simulator for guide-wire insertion in hip surgery Kapil SUGAND 1,2,3, Robert A WESCOTT 1, Richard CARRINGTON 3, Alister HART 1,3, and Bernard H VAN DUREN 1,3
1 Institute of Orthopaedics & Musculoskeletal Sciences, University College London, London; London; 3 Royal National Orthopaedic Hospital, Stanmore, London, UK Correspondence: Ks704@ic.ac.uk Submitted 2017-12-09. Accepted 2018-03-11.
Background and purpose — Simulation is an adjunct to surgical education. However, nothing can accurately simulate fluoroscopic procedures in orthopedic trauma. Current options for training with fluoroscopy are either intraoperative, which risks radiation, or use of expensive and unrealistic virtual reality simulators. We introduce FluoroSim, an inexpensive digital fluoroscopy simulator without the need for radiation. Patients and methods — This was a multicenter study with 26 surgeons in which everyone completed 1 attempt at inserting a guide-wire into a femoral dry bone using surgical equipment and FluoroSim. 5 objective performance metrics were recorded in real-time to assess construct validity. The surgeons were categorized based on the number of dynamic hip screws (DHS) performed: novices (< 10), intermediates (10–39) and experts (> 40). A 7-point Likert scale questionnaire assessed the face and content validity of FluoroSim. Results — Construct validity was present for 2 clinically validated metrics in DHS surgery. Experts and intermediates statistically significantly outperformed novices for tip–apex distance and for cut-out rate. Novices took the least number of radiographs. Face and content validity were also observed. Interpretation — FluoroSim discriminated between novice and intermediate or expert surgeons based on tip–apex distance and cut-out rate while demonstrating face and content validity. FluoroSim provides a useful adjunct to orthopedic training. Our findings concur with results from studies using other simulation modalities. FluoroSim can be implemented for education easily and cheaply away from theater in a safe and controlled environment. ■
Orthopedic training has declined after the introduction of the European Working Time Directive (EWTD) in 2004, leading to fewer operative training hours, a reduction from 30,000
Lab, Imperial College London, Charing Cross Hospital,
to 15,000 hours (Temple 2010). A reduction in training time, seen on both the European and North American continents, has been perceived negatively in surgical education (Zuckerman et al. 2005, Egan et al. 2012). Junior trainees are taking longer to complete operations (Wilson et al. 2010), which reduces theatre efficiency and increases economic burden. Stable extracapsular neck of femur (NOF) fractures make up a significant proportion of all hip fractures and are treated using a dynamic hip screw (DHS) (Utrilla et al. 2005). Other alternatives include compression or cannulated hip screws. The DHS implant may fail due to cut-out, reported to be between approximately 2% and 7% (Chirodian et al. 2005, Hsueh et al. 2010), predicted by the tip–apex distance (TAD) (Andruszkow et al. 2012). Fluoroscopy is used during the DHS procedure (Baratz et al. 2014), which carries radiation risks. Inexperienced trainees take more images, thus increasing radiation exposure (Khan et al. 2012). Digital imaging alternatives have been explored clinically, but are not used for training or simulation (Grutzner and Suhm 2004). Current DHS simulation options consist of virtual reality (VR) or workshop dry bones (Akhtar et al. 2015). VR DHS simulation enables trainees to learn the cognitive process of the DHS procedure with the help of digital fluoroscopy, but at the expense of not using actual equipment to practice manual dexterity. Workshop dry bone simulation develops motor skills as used in theatre; however, fluoroscopy is not used due to the radiation risks (Stirling et al. 2014). We have developed a digital fluoroscopic simulation system, FluoroSim, that produces realistic radiographs for simulation without using radiation. Both cognitive and motor skills needed for the insertion of a DHS guide-wire can be developed by giving real-time feedback through 5 objective performance metrics.
© 2018 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 10.1080/17453674.2018.1466233
12313 Sugand D.indd 380
Acta Orthopaedica 2018; 89 (4): 380–385
Figure 2. A right phantom limb produced out of a mannequin leg with an interchangeable workshop femur. In the background of the image the simulated radiograph of this construct may be seen. Figure 1. Control screen of the FluoroSim software running with the calibration femur. The software locates the colored markers and finds their center. It marks the position of the guide-wire on the camera image and, using the ATM, overlays this onto the pre-loaded radiograph. Both AP and CTL images are produced.
We aim to demonstrate whether FluoroSim can: 1) separate surgeons with different levels of surgical experience using 5 objective metrics (construct validity); 2) offer realistic steps of a guide-wire insertion into a hip (content validity); and 3) offer realistic radiographs using FluoroSim (face validity).
Materials and methods Setup FluoroSim is an augmented-reality imaging and targeting software that uses 2 Logitech c920 cameras (Logitech, Romanelsur-Morges, Switzerland) to track 2 colored markers attached to a DHS guide-wire (van Duren et al. 2018). The system is calibrated using a workshop femur (3B Scientific, Hamburg, Germany). With the guide-wire inserted into the femur, 3 points from the digital camera image are selected and matched to 3 corresponding points on a pre-loaded hip radiograph in both the anterior-posterior (AP) and cross-table lateral (CTL) plane (to produce an affine transformation matrix). Image processing algorithms locate the center of the markers on the DHS guidewire and overlay its position onto the radiograph (Figure 1). A simulation scenario was set up using the FluoroSim software run on a MacBook Pro with macOS Sierra 10.12.1 (Apple Inc., Cupertino, CA, USA) for digital imaging. A phantom limb model represented a right hip which was draped, produced from a hollow polyethylene mannequin and interchangeable workshop femurs (Figure 2). A Stryker system 4 rotary drill (Stryker, Kalamazoo, MI, USA) and a 135-degree angle guide with guide-wire were used for high-
12313 Sugand D.indd 381
fidelity immersive simulation. This equipment in total cost less than $3,000, a fraction of the price of commercial virtual reality fluoroscopic simulators. Objective metrics The FluoroSim software calculated real-time objective performance metrics including: 1) TAD (mm); 2) COR (%) according to Baumgaertner’s curve (Baumgaertner et al. 1995); 3) total procedural time (s); 4) total number of radiographs; and 5) total number of guide-wire retries. Subjective metrics All cohorts assessed the face and content validity of FluoroSim. A 7-point Likert scale questionnaire inquired as to agreement with 4 statements regarding the realistic appearance of FluoroSim and its usefulness for training. Logistics 26 surgeons from Northwick Park (London, UK), Central Middlesex (London, UK), and the Princess Alexandra Hospital (Harlow, UK) were recruited voluntarily and categorized into 3 groups based on the number of DHS procedures performed: novices (< 10), intermediates (10–39) and experts (≥ 40). Each participant received a standardized explanation of the task (Table 1) and then had 1 attempt to insert the DHS guide-wire using FluoroSim for AP or CTL views (Figure 3). The 5 objective metrics were recorded at the end of each attempt. Inclusion/exclusion criteria Inclusion criteria included having observed at least 1 DHS procedure in theatre. Exclusion criteria consisted of having attempted DHS simulation beforehand, undergraduates, and non-orthopedic trainees.
Acta Orthopaedica 2018; 89 (4): 380–385
Table 1. Checklist used to standardize the participants’ instructions Standardized instruction checklist 1. Explain the basic working of FluoroSim, highlighting the importance of not covering the tracking markers or bending the guide-wire. 2. Explain the 5 objective metrics recorded. 3. Explain the main goal of the task: To achieve optimal guide-wire placement as if they were completing a DHS procedure thus giving them an optimal TAD. 4. Highlight that time was being recorded but the focus was on achieving an optimal guide-wire placement. 5. Explain that they should indicate when they are happy with their final guide-wire placement.
Figure 3. A surgeon using FluoroSim with the phantom limb, surgical equipment and the imaging system.
Statistics The data were analyzed in SPSS (version 24.0, IBM Corp, Armonk, NY, USA). Normality was checked using histograms and Shapiro–Wilk testing at α = 5%. Objective metrics—All of the data underwent normality testing. TAD and total procedural time were normally distributed; however, the other 3 metrics were not-normally distributed. For this reason, to allow for standardized comparison between metrics, all statistical analyses used non-parametric methods. The Kruskal–Wallis test compared the distribution between all cohorts at α = 5%. Mann–Whitney U post-hoc testing was used when the Kruskall–Wallis test reached significance. Correction for multiplicity was needed due to the multiple comparisons of the groups. To correct for multiplicity, we multiplied the p-values obtained from each Mann–Whitney U comparison by 3 to maintain a consistent α cut-off value. Therefore, the corrected α cut-off value remained at p = 0.05. This is reflected in both Table 2 and Figure 4. Questionnaires—Percentages of agreement for each statement assessed content and face validity. A score of 5, 6, or 7 relating to mildly, moderately, or strongly agreeing with the statement was seen as the participant agreeing overall. Ethics, funding, and potential conflicts of interest The project outline was submitted to the Project Evaluation Panel at the Royal National Orthopaedic Hospital. Ethical approval was deemed unnecessary due to the non-clinical nature. Informed consent was gained from all participants. This project received funding from the Professor A. T. Fripp fund. B. H. van Duren is an NIHR funded clinical fellow in Trauma and Orthopedics. R. A. Wescott received funding assistance from the Goldberg Schachmann and Freda Becker
12313 Sugand D.indd 382
Table 2. Median performance of each cohort
InterNovices mediates Experts
Tip–apex distance (mm) 47 Cut-out rate (%) 55 Procedural time (s) 190 No. of radiographs (n) 16 No. of guide-wire retries (n) 0
28 4.7 206 26 1
24 2.6 222 28 2
p-value a 0.006 0.007 0.6 0.03 0.2
Memorial Fund. There were no conflicts of interest.
Results Demographics The stage of training was recorded using the number of years since medical school graduation, defined as postgraduate year (PGY): 1. Novice group (n = 8) ranged from PGY2 to PGY5 trainees; 2. Intermediate group (n = 7) ranged from PGY4 to PGY9; 3. Expert group (n = 11) ranged from PGY7 and above. Objective metrics A statistically significant difference in TAD, number of radiographs, and COR was observed between all cohorts (Table 2). The experts and the intermediates significantly outperformed the novices for TAD and COR (Table 3 and Figure 4A–B), with experts achieving the lowest scores for these metrics (Table 2). The novices used the least time, had the fewest number of guide-wire retries, and took significantly fewer radiographs compared with the experts (Table 3 and Figure 4C–E).
Acta Orthopaedica 2018; 89 (4): 380–385
Table 3. Percentage difference and (p-value) between the 3 cohorts for each objective metric a
Novices vs. intermediates
Novices vs. experts
Tip–apex distance (mm) Cut-out rate (%) Procedural time (s) No. of radiographs (n) No. of guide-wire retries (n)
40 (0.03) 92 (0.03) 8 (1.0) 39 (0.06) 100 (0.4)
48 (0.01) 13 (0.9) 95 (0.01) 44 (1.0) 14 (1.0) 7 (1.0) 43 (< 0.05) 7 (1.0) 200 (0.3) 100 (1.0)
Main findings This study showed a statistically significant difference in both TAD, COR, and number of radiographs taken between all cohorts. Both experts and intermediates outperformed novices in TAD and COR. We expected experts to be faster, use less fluoroscopy, and have fewer retries at guide-wire insertion. However, the opposite trend was observed, with novice participants using significantly fewer radiographs than experts. Face and content validity were also demonstrated.
Intermediates. vs. expert
p-value presented from Mann–Whitney U post-hoc testing.
Comparison with current literature Our study showed a statistically significant difference between novices and the other cohorts for TAD and COR (i.e., construct validity), but it was unable to differentiate between intermediate and expert surgeons. This was not unexpected, as it is harder for assessment systems to discriminate between levels of higher skill (Munz et al. 2004). BoneDoc is a computer-based VR DHS simulator that showed similar findings, being able to differentiate medical students from trainee surgeons, but not different levels of trainee surgeons (Blyth et al. 2008). Another VR DHS simulator, TraumaVision (Swemac Simulation AB, Linkoping, Sweden), demonstrated construct
Face and content validity questionnaire The questionnaire demonstrated the following: 1. 22/26 participants agreed that both the radiographs produced by FluoroSim and the phantom limb model were realistic. 2. 23/26 participants agreed that the content of the simulation would be useful to teach trainees guide-wire insertion into the hip. 3. 25/26 participants agreed that the surgical equipment used in the simulation was realistic.
Tip–apex distance (mm)
Total time of procedure (s)
Cut-out rate (%)
600 p = 0.01
p = 0.01 p = 0.03
p = 0.03
p = 0.9
p = 1.0 p = 1.0
p = 1.0
p = 1.0
Number of guide-wire retries (n)
Number of radiographs (n) 100 p = 0.05 p = 0.06
p = 0.3
8 p = 0.4
p = 1.0
p = 1.0
12313 Sugand D.indd 383
Figure 4. A series of box plots for each objective metric. The central line represents the median, the boundaries of the box represents the upper and lower quartiles respectively, and the whiskers represent the range without outliers. A significance value is presented from the adjusted Mann–Whitney U comparison.
validity. However, intermediates achieved the lowest TAD and COR, possibly due to skills decay of experts and lack of surgical experience of novices (Akhtar et al. 2015). Expert surgeons tend to experience skills decay, from no longer leading the trauma lists and reduced exposure to the DHS procedure to allow residents and fellows (intermediates) to gain more experience. Using TraumaVision again, a different research group demonstrated that senior surgeons used more guide-wire retries than junior surgeons on the DHS module (Pedersen et al. 2014). Using a VR-based drilling simulator, senior surgeons were shown to take a longer time to complete a task; however, they made fewer mistakes compared with their junior colleagues (Vankipuram et al. 2010). These results were similar in our study. Experts took more time, using more radiographs and more guide-wire retries. However, they still managed to achieve a better TAD and COR compared with novices because experienced surgeons placed more importance on the clinical predictors of DHS failure. Inexperienced novices placed less importance on obtaining the optimal TAD. Other simulation studies have used an induction period to remove the learning curve of understanding the simulation software (LeBlanc et al. 2013). However, we recorded the first attempt to achieve standardization of our participants using FluoroSim. Limitations This study failed to record the absolute number of DHS procedures completed by each participant individually. Although a cut-off of 40 procedures was selected as this is the number of procedures necessary in the UK to demonstrate competency during formal residency training, we assume 40 procedures as the point of expertise. An additional limitation was the limit of time within the study. Surgeons were asked to participate in between their daily tasks, therefore some participants had a sense of urgency to complete the task, influencing the total procedural time taken. The hand dominance of the surgeon was not accounted for but this procedure required ambidexterity. All participants completed the procedure on a right femur, using their right hand to drill and the left hand to hold the 135degree angle guide regardless of dominance. Future work Further work is needed to look at the training effect of FluoroSim and the transfer or concurrent validity of FluoroSim in comparison with similar simulators. Ideally, FluoroSim will be used instead of the C-arm in theatre to avoid the risk of radiation exposure. Conclusion FluoroSim is a useful adjunct in training guide-wire insertion into the hip. It can accurately discriminate between novices and intermediates/experts for clinically validated outcomes in DHS surgery, namely TAD and COR. This is the first of
12313 Sugand D.indd 384
Acta Orthopaedica 2018; 89 (4): 380â&#x20AC;&#x201C;385
its kind in orthopedic simulation according to current literature. FluoroSim provides an effective and affordable solution to simulate intraoperative imaging without needing radiographs.
KS proposed the study and the methodology, helped to analyze data as well as helping to write up and review the manuscript, and acting as a project lead co-supervisor. RW collected and analyzed data as well as contributed to writing the manuscript. BvD created the FluoroSim system, reviewed the paper, and acted as project lead co-supervisor. RC contributed to reviewing the paper. AH contributed in reviewing the paper as well as acting as project principal supervisor.
Acta thanks Li FellĂ¤nder-Tsai and other anonymous reviewers for help with peer review of this study.
Akhtar K, Sugand K, Sperrin M, Cobb J, Standfield N, Gupte C. Training safer orthopedic surgeons: construct validation of a virtual-reality simulator for hip fracture surgery. Acta Orthop 2015; 86(5): 616-21. Andruszkow H, Frink M, Fromke C, Matityahu A, Zeckey C, Mommsen P, et al. Tip apex distance, hip screw placement, and neck shaft angle as potential risk factors for cut-out failure of hip screws after surgical treatment of intertrochanteric fractures. Int Orthop 2012; 36(11): 2347-54. Baratz M D, Hu Y Y, Zurakowski D, Appleton P, Rodriguez E K. The primary determinants of radiation use during fixation of proximal femur fractures. Injury 2014; 45(10): 1614-19. Baumgaertner M R, Curtin S L, Lindskog D M, Keggi J M. The value of the tip-apex distance in predicting failure of fixation of peritrochanteric fractures of the hip. J Bone Joint Surg Am 1995; 77(7): 1058-64. Blyth P, Stott N S, Anderson I A. Virtual reality assessment of technical skill using the Bonedoc DHS simulator. Injury 2008; 39(10): 1127-33. Chirodian N, Arch B, Parker M J. Sliding hip screw fixation of trochanteric hip fractures: outcome of 1024 procedures. Injury 2005; 36(6): 793-800. Egan C, Elliott R, Fleming P. European Working Time Directive and the use of simulators and models in Irish orthopaedics. Irish J Med Sci 2012; 181(1): 143-6. Grutzner P A, Suhm N. Computer aided long bone fracture treatment. Injury 2004; 35: 57-64. Hsueh K K, Fang C K, Chen C M, Su Y P, Wu H F, Chiu F Y. Risk factors in cutout of sliding hip screw in intertrochanteric fractures: an evaluation of 937 patients. Int Orthop 2010; 34(8): 1273-6. Khan I A, Kamalasekaran S, Fazal M A. Risk of ionising radiation to trainee orthopaedic surgeons. Acta Orthop Belg 2012; 78(1): 106-10. LeBlanc J, Hutchison C, Hu Y P, Donnon T. A comparison of orthopaedic resident performance on surgical fixation of an ulnar fracture using virtual reality and synthetic models. J Bone Joint Surg Am 2013; 95A(9): e60(1)-e(6). Munz Y, Moorthy K, Bann S, Shah J, Ivanova S, Darzi S A. Ceiling effect in technical skills of surgical residents. Am J Surg 2004; 188(3): 294-300. Pedersen P, Palm H, Ringsted C, Konge L. Virtual-reality simulation to assess performance in hip fracture surgery. Acta Orthop 2014; 85(4): 403-7. Stirling E R B, Lewis T L, Ferran N A. Surgical skills simulation in trauma and orthopaedic training. J Orthop Surg Res 2014; 9: 126-35. Temple J. Time for training: a review of the impact of the European Working Time Directive on the quality of training. London: Department of Health; 2010. Utrilla A L, Reig J S, Munoz F M, Tufanisco C B. Trochanteric Gamma nail and compression hip screw for trochanteric fractures: a randomized, prospective, comparative study in 210 elderly patients with a new design of the Gamma nail. J Orthop Trauma 2005; 19(4): 229-33.
Acta Orthopaedica 2018; 89 (4): 380â&#x20AC;&#x201C;385
van Duren B H, Sugand K, Wescott R, Carrington R, Hart A. Augmented reality fluoroscopy simulation of the guide-wire insertion in DHS surgery: A proof of concept study. Med Eng Phys 2018; 55: 52-9. Vankipuram M, Kahol K, McLaren A, Panchanathan S. A virtual reality simulator for orthopedic basic skills: a design and validation study. J Biomed Inform 2010; 43(5): 661-8.
12313 Sugand D.indd 385
Wilson T, Sahu A, Johnson D S, Turner P G. The effect of trainee involvement on procedure and list times: a statistical analysis with discussion of current issues affecting orthopaedic training in UK. Surg J R Coll Surg E 2010; 8(1): 15-19. Zuckerman J D, Kubiak E N, Immerman I, DiCesare P. The early effects of code 405 work rules on attitudes of orthopaedic residents and attending surgeons. J Bone Joint Surg Am 2005; 87A (4): 903-8.
Acta Orthopaedica 2018; 89 (4): 386–393
Risk of further surgery on the same or opposite side and mortality after primary total hip arthroplasty: A multi-state analysis of 133,654 patients from the Swedish Hip Arthroplasty Register Peter H J CNUDDE 1–3, Szilard NEMES 1,2, Erik BÜLOW 1,2, A John TIMPERLEY 4, Sarah L WHITEHOUSE 5, Johan KÄRRHOLM 1,2, and Ola ROLFSON 1,2
1 Swedish Hip Arthroplasty Register, Centre of Registers, Gothenburg, Sweden; 2 Department of Orthopaedics, Institute of Clinical Sciences, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden; 3 Department of Orthopaedics, Hywel Dda University Healthboard, Prince Philip Hospital, Llanelli, UK; 4 Hip Unit, Princess Elizabeth Orthopaedic Centre, Royal Devon & Exeter Hospital, Exeter, UK; 5 Queensland University of Technology, Brisbane, Queensland, Australia Correspondence: firstname.lastname@example.org Submitted 2017-11-07. Accepted 2018-04-18.
Background and purpose — The hip-related timeline of patients following a total hip arthroplasty (THA) can vary. Ideally patients will live their life without need for further surgery; however, some will undergo replacement on the contralateral hip and/or reoperations. We analyzed the probability of mortality and further hip-related surgery on the same or contralateral hip. Patients and methods — We performed a multistate survival analysis on a prospectively followed cohort of 133,654 Swedish patients undergoing an elective THA between 1999 and 2012. The study used longitudinally collected information from the Swedish Hip Arthroplasty Register and administrative databases. The analysis considered the patients’ sex, age, prosthesis type, surgical approach, diagnosis, comorbidities, education, and civil status. Results — During the study period patients were twice as likely to have their contralateral hip replaced than to die. However, with passing time, probabilities converged and for a patient who only had 1 non-revised THA at 10 years, there was an equal chance of receiving a second THA and dying (24%). It was 8 times more likely that the second hip would become operated with a primary THA than that the first hip would be revised. Multivariable regression analysis reinforced the influence of age at operation, sex, diagnosis, comorbidity, and socioeconomic status influencing state transition. Interpretation — Multi-state analysis can provide a comprehensive model of further states and transition probabilities after an elective THA. Information regarding the lifetime risk for bilateral surgery, revision, and death can be of value when discussing the future possible outcomes with patients, in healthcare planning, and for the healthcare economy.
Most patients undergoing total hip arthroplasty (THA) have an uneventful and relatively pain-free future, whereas some patients will have further health-care encounters related to their hip joints. These contacts may be for revision of the ipsilateral hip. The hip might also be re-operated on for reasons not necessitating exchange or extraction of the implant or any of its parts. Some patients will need surgery on the opposite hip and may also undergo re-operation or revision surgery on the second hip. Better knowledge of patients’ hip-related timeline (HRT) may improve the understanding and expectations of patients, surgeons, and healthcare providers. Further healthcare contacts are important in the case of bundled payments, tariffs, and payments-by-results (Burwell 2015, Jubelt et al. 2017). The increasing demands and financial pressures on health systems make predictions of further contacts with healthcare providers important. Several studies have described the lifetime risk for revision and long-term mortality, but few have described the different paths (i.e., contralateral THA, revisions, and death) the patient can follow (Gillam et al. 2012, 2013, Abdel et al. 2016, Maradit Kremers et al. 2016, Sanders et al. 2017). The increased availability and quality of longitudinal data have stimulated the development of life-history models. Multi-state analysis has been advocated as a natural framework, studying transitions between different stages (Commenges 1999) and has been shown to provide a convenient framework for the handling of a wide variety of medical conditions, characterized by multiple events where longitudinal data are available (Farewell and Tom 2014). This framework allows for the combining of several possible outcomes in a single analysis and aids the depiction of the hip-related timeline that patients potentially could follow. This is in contrast with the more classical sur-
© 2018 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 10.1080/17453674.2018.1475179
12286 Cnudde D.indd 386
Acta Orthopaedica 2018; 89 (4): 386–393
Patients with THR performed between 1999 and 2012 n = 164,001 Excluded (n = 30,347): – fractures, tumors and secondary osteoarthritis, 20,210 – first THR prior to 1999, 8,546 – single stage bilateral THR, 1,202 – missing data on education, 203 – nonroutine approaches, 186 Patients included in the study n = 133,654
Figure 1. Flowchart of the patients included in this study (January 1, 1999–December 31, 2012).
vival analysis, as employed in most studies, which is only able to depict transition from one stage to another. Additionally, multi-state analysis facilitates systematic handling of the laterality problem, an inevitably characteristic of every survival analysis in the field of arthroplasty (Ranstam et al. 2011). We describe the probability for further hip-related surgery on the same or opposite side using the first primary total hip replacement on either side as the index operation and the probability of dying during the study period. We also depict the influences of known patient-related, surgery-related, and socio-economic factors using prospectively collected and linked data from a national database.
Patients and methods Prospectively collected data from the Swedish Hip Arthroplasty Register (SHAR) were obtained and analyzed for all patients who received a first recorded primary THA for elective reasons (non-acute trauma-related and no tumor surgery) using the validated and linked research database to access surgical-, patient-related, and socio-economic factors (Cnudde et al. 2016). Data were available for 133,654 patients who underwent a first THA between January 1, 1999 and December 31, 2012 (Figure 1). The choice of study period was guided by the availability of the linked dataset for this period. All patients who had received a THA prior to this period were excluded. Bilateral single-stage primaries were also excluded (1,202 cases). The selected cohort of patients was then followed until the end of the study period (December 31, 2012) or death and revisions were recorded. We defined revision as exchange or extraction of the implant or any of its parts. The Swedish healthcare system provides universal access to healthcare for its residents and each hospital contact for every individual is recorded in a centrally governed system. Government databases also hold information on socioeconomic factors of all residents. Death dates are recorded by the Tax Office and are linked on a regular basis with the SHAR as well as other governmental databases. The SHAR is part of the Quality Regis-
12286 Cnudde D.indd 387
ters in Sweden (QR) and the centralized information collection system in the Nordic Countries has been well recognized for its population-based research (Emilsson et al. 2015). Statistics Continuous variables were summarized as means (SD), categorical variables as percentages. To describe the association between ipsilateral operations, revisions, and mortality we adopted an extended irreversible disease progression illness–death model describing the movement of patients between a series of discrete states in a continuous time. This irreversible disease progression model had 5 discrete stages and describes the pathway of a patient from the first THA to the absorbing state of death as a Markov process. The patients enter the study at the time of their first elective THA surgery (State 1). They could stay in this state (unilateral THA without further intervention or death) until censoring on December 31, 2012—the end of the study period. However, the patient might advance into adjacent stages (Figure 2). If the contralateral hip has an arthroplasty the patient enters State 2, if the ipsilateral hip is revised then State 3 is applicable. A patient could reach the absorbing state, death (State 5). From State 2 the patients could advance to State 3 or State 4, which is revision of the second hip, or enter the absorbing State 5. At any given time a patient can be in a specific state and the next state to which the patient moves and the time when this transition occurs is determined by a set of transition intensities that represents instantaneous risk of moving from state i to state j, namely: qij = lim P(X(t + δt) = j⏐X(t) = i)/δt. δt→0
The transition intensities for one specified state to all others sum to 1, thus the diagonal elements of the transition intensities matrix Q that represent lingering in the specified state are given by qii = –∑i≠jqij. We defined the transition ratio as the ratio of 2 estimated intensities (e.g. qij ⁄qik). This ratio provides estimates on the likelihood of progressing to one stage or other. Statistical inference is based on approximate-normality and the δ-method. If the 95% confidence interval (CI) covers 1 we cannot reject the null-hypothesis of equal transition intensities. We assessed the association between covariates x (sex, age, clinical diagnosis at first operation, comorbidity, surgical factors, socioeconomic status) and transition intensities with the modified proportional hazards model (Marshall and Jones 1995) calculated as qij (x(t))= qij(0) exp(βijT x(t)), where exp(βij) represents the estimated hazard ratios corresponding to given covariates effect of the transition intensities from state i to state j. Separate baseline hazards and regression coefficients are estimated for each possible transition by fit-
Acta Orthopaedica 2018; 89 (4): 386–393
Table 1. Patient demographics of the cohort 1999–2012 (n = 133,654) Age, mean (SD) Sex, n (%) Male Female Diagnosis, n (%) Primary osteoarthritis Inflammatory joint disease Sequel childhood hip disorder Femoral head necrosis Elixhauser Index, mean (SD) Surgical approach, n (%) Lateral Posterior Fixation, n (%) Cemented Uncemented Hybrid Reversed hybrid Resurfacing Clinic type, n (%) University County Rural Private
68 (11) 57,058 (43) 76,596 (57) 122,568 (92) 3,199 (2.4) 3,148 (2.4) 4,735 (3.5) 0.61 (0.96) 59,355 (44) 74,299 (56) 104,560 (78) 13,500 (10) 3,336 (2.5) 9,981 (7.5) 1,666 (1.2) 14,080 (11) 44,897 (34) 55,126 (41) 19,551 (15)
ting a series of proportional hazards models (Andersen and Keiding 2002). The conditional transition probabilities matrix P(t) is calculated as etQ and its entries pij (t) are the probabilities of being in state j at time t+u given that at time t the patient is in state i. The last estimates of interest are the unconditional state occupation probabilities, i.e., at each time point the curve estimates the fraction of patients currently in that state without considering the path that led there. This estimate is based on cause-specific hazard, h(tk) = (∑i dki) ⁄nk where dki is an indicator which takes value 1 if the patient transits to state i at time t, 0 otherwise and nk gives the number of patients at risk at time t. The unconditional state occupation probabilities are given by ∑ti≤t S(ti–1) hk (ti) where S(t) is the overall survival function, which summarizes the absorbing state (or censoring). Statistical analyses were conducted with R computing environment (R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org) and the “msm” package (Jackson 2011). Ethics, funding, and potential conflict of interest Ethical review approval was obtained on April 7, 2014 from the Regional Ethical Review Board in Gothenburg, Sweden (entry number 271-14). This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. There was no support from any external organization for the submitted work and there are no financial relationships with any organizations that might have a direct interest in the submitted work in the previous 3 years.
12286 Cnudde D.indd 388
0.2% 0.7% 3% 13% 24%
0.3% 1% 4% 16% 24% UP TO: 30 days 90 days 1 year 5 years 10 years
0.04% 0.1% 0.5% 2% 3%
0.03% 0.1% 0.4% 2% 3%
0.2% 0.5% 2% 10% 20% 0.04% 0.1% 0.5% 2% 4%
Revision of first THR 0.2% 0.6% 2% 10% 16%
0.3% 1% 4% 18% 32%
0.02% 0.06% 0.3% 1% 1%
0.2% 0.4% 2% 7% 10%
Revision of second THR
0.2% 0.7% 3% 13% 25%
Figure 2. Multi-state analysis scheme and possible transitions. State 1 is the first hip replacement, state 2 is the second (contralateral) hip replacement, state 3 is the state where the first-performed hip is revised, whereas state 4 is the state where the contralateral hip is revised, and state 5 is death. The percentages stated are up to the 30- and 90-day mark as well as up to the 1-, 5-, and 10-year mark and represent the estimated transition probabilities within that given time.
Results Data were included on 160,165 primary THAs in 133,654 patients. During the study period 22,070 patients died, 26,511 patients had their contralateral hip replaced (simultaneous bilateral THAs have been excluded), 4,025 had their first replaced hip revised, and 694 their second hip. Patient demographics are presented in Table 1. The median follow-up time from the first THA until death or censoring was 5.6 years, and from the second THA until death or censoring 4.2 years (Appendix 1, see Supplementary data). Transition probabilities (Appendix 2, see Supplementary data) between the different stages varied with time (Figure 2). Likewise, so did the probability of belonging to different stages (Figure 3). During the study period patients were twice as likely (transition ratio = 2.1, CI 2.1–2.2) to have their contralateral hip replaced than to die. However, with passing time, probabilities converged and for a patient who remained in State 1 (1st THA) up to 10 years after the first hip surgery there was an equal chance of receiving a contralateral hip and dying (24%). Replacement of the contralateral hip was 7.5 times (CI 7.3–7.9) more likely than revision of the first hip. For patients who had their contralateral hip replaced, death as the next stage was 4.0 times (CI 2.4–6.6) more likely than revision of the first hip and 2.7 times (CI 1.7–4.5) more likely than revision of the second hip. For patients in state 2 (both hips sequentially replaced) the likelihood of revision of the second hip as a next step was 1.5 (CI 1.3–1.7) more likely than revision of the first hip. Viewed over the total period of observation the likelihood of revision of the first hip was,
Acta Orthopaedica 2018; 89 (4): 386–393
Figure 3. State occupation probabilities at different time points.
however, equal to the revision of the second hip (0.9, CI 0.8– 1.0). Hazard ratios of the most frequent transitions between states are provided (Table 2). They are also visualized in Forest plots (Appendix 3, see Supplementary data). Female patients were more likely to undergo surgery on both hips, but had less of risk of dying or being revised following surgery (Figure 4). Inflammatory joint disease as the indication for arthroplasty increased the likelihood of revision and dying and this influence was both age and time dependent. Operations performed for avascular necrosis of the femoral head were less likely to be performed on both sides and the risk of revision and/or dying was increased. Childhood hip diseases did not change the pattern of transitions if compared with primary osteoarthritis (OA) (Figure 5). The Elixhauser comorbidity index (compiled from the available ICD-10 codes in the year preceding surgery) had an effect on the probability of revision and an even greater effect on mortality (Figure 6). The majority of hip replacements were performed using posterior (56%) and lateral approaches (44%). We were unable
Figure 4. Effect of sex (male versus female) on transition probability from the state of 1st THR to revision of the first hip or death within 1 and 10 years from the index operation at different ages, presented with CI. A: Effect of sex on revision probability within 1 year. B: Effect of sex on revision probability within 10 years. C: Effect of sex on death probability within 1 year. D: Effect of sex on death probability within 10 years.
to identify any significant effect of surgical approach on mortality or revision in the short or longer term (Appendix 4/1, see Supplementary data). The effect of fixation on revision at 1 year showed no statistically significant difference whereas at 10 years the effect gained significance from the age of 70 onwards. The difference in mortality could be identified at 1 and 10 years (Appendix 4/2, see Supplementary data). Patients who had obtained only a lower education level were
Table 2. Hazard ratios (95% confidence intervals) and the influence of different variables on the transition between different states a Female sex State 1—State 2 State 1—State 3 State 1—State 5 State 3—State 5 Fixation: State 1—State 2 State 1—State 3 State 1—State 5 State 3—State 5
Elixhauser Comorbidity Index
1.17 (1.14–1.20) 0.71 (0.66–0.77) 0.65 (0.63–0.68) 0.68 (0.57–0.82)
0.98 (0.98–0.98) 0.99 (0.98–0.99) 1.09 (1.09–1.09) 1.08 (1.07–1.09)
0.94 (0.91–0.97) 0.98 (0.91–1.05) 1.01 (0.98–1.05) 0.83 (0.68–1.00)
1.04 (1.02–1.05) 1.18 (1.14–1.22) 1.19 (1.17–1.21) 1.25 (1.16–1.35)
0.98 (0.93–1.03) 1.15 (1.02–1.31) 0.57 (0.51–0.64) 0.87 (0.51–1.48)
0.90 (0.84–0.97) 1.08 (0.90–1.30) 0.90 (0.80–1.02) 1.25 (0.72–2.17)
1.07 (1.01–1.13) 1.34 (1.17–1.54) 0.59 (0.52–0.66) 0.78 (0.46–1.32)
0.79 (0.71–0.88) 1.60 (1.26–2.04) 0.36 (0.22–0.60) 0.48 (0.07–3.52)
Education Middle 1.09 (1.05–1.12) 1.06 (0.98–1.14) 0.86 (0.83–0.89) 0.99 (0.82–1.20)
High 1.18 (1.14–1.22) 1.08 (0.98–1.18) 0.78 (0.75–0.82) 0.83 (0.64–1.07)
1 = first THR; State 2 = second THR (contralateral); State 3 = revision of first THR (first operated side); State 4 = revision of second THR (second operated side); State 5 = death.
12286 Cnudde D.indd 389
Acta Orthopaedica 2018; 89 (4): 386â&#x20AC;&#x201C;393
Figure 5. Effect of diagnosis (indication for surgery for the first hip) on transition probability from the state of 1st THR to revision of the first hip or death within 1 and 10 years from the index operation at different ages, presented with CI. A: Effect of diagnosis on revision probability within 1 year. B: Effect of diagnosis on revision probability within 10 years. C: Effect of diagnosis on death probability within 1 year. D: Effect of diagnosis on death probability within 10 years.
Figure 6 . Effect of the Elixhauser Comorbidity Index (ECI) on transition probability from the sate of 1st THR to revision of the first hip or death within 1 and 10 years from the index operation at different ages, presented with CI. A: Effect of ECI on revision probability within 1 year. B: Effect of ECI on revision probability within 10 years. C: Effect of ECI on death probability within 1 year. D: Effect of ECI on death probability within 10 years.
less likely to undergo sequential bilateral procedures and showed an increased likelihood of dying (Appendix 4/3, see Supplementary data).
2013) evaluated the risk of subsequent contralateral THA using a multi-state analysis and found a 16% and 20% probability of receiving a contralateral hip in the Australian and Norwegian population respectively. Shao et al. (2013) described a 31% chance of receiving a contralateral THA at a mean of 18 years after original surgery. These figures are similar to what we found in Sweden. As expected the risk of revision decreases with increasing age, as a result of selection and age as a competing risk. The lifetime risk for revision has been previously studied and the age at the time of the operation has been found important (Abdel et al. 2016, Bayliss et al. 2017, Schreurs and Hannink 2017). Compared with an age-matched population, mortality remains somewhat lower in the THA population, more so in the older age groups (Cnudde et al. 2018b). The influence of diagnosis on mortality has been described previously, with patients undergoing THA for primary OA doing better than other diagnoses (Lie et al. 2000, Pedersen et al. 2011, Cnudde et al 2018b). The influence of diagnosis on revision surgery after adjusting for co-variables is obvious, with increased revision rates in the case of inflammatory arthritis or avascular necrosis of the femoral head. Bergh et al. (2014) used the Nordic Arthroplasty Register Association (NARA) database to study the effect of avascular necrosis on revision rates and found an increased revision rate in this group compared withc primary OA. This increased revision
Discussion The multi-state analysis enabled the comprehensive prediction of transition probabilities between different postoperative states and the influence of patient demographics, patient- and surgery-related factors as well as socioeconomic influences. Using longitudinally and prospectively collected data from nationwide registers we could describe part of the patientâ&#x20AC;&#x2122;s journey following the first hip replacement for elective reasons (the hip-related timeline). Multi-state analysis has the advantage of providing a better understanding of the data and a more coherent picture of the complete path instead of isolated events (Gillam et al. 2012). The number of patients with sequential bilateral THA is increasing, as also seen during our study period (Cnudde et al. 2018a). Describing the factors contributing to the contralateral operation is of interest. The probability of undergoing a further hip replacement on the contralateral side was 1 in 4. A previous register study also used this technique but in a smaller cohort and with a shorter study period Gillam et al. (2012,
12286 Cnudde D.indd 390
Acta Orthopaedica 2018; 89 (4): 386–393
rate was not confirmed in a systematic literature review by Johannson et al. (2011). Register studies and prospective studies also describe higher dislocation rates and an increased risk of periprosthetic fractures in patients with inflammatory arthritis undergoing THA (Zwartele et al. 2004, Lindahl et al. 2006, Meek et al. 2008). This will have a bearing when comparing outcomes between hospitals and surgeons with a different case-mix. However, our data do not include reoperations not necessitating exchange or extraction of the implant or any of its parts (e.g., fixation of periprosthetic fractures or reduction of dislocated implant). The importance of comorbidity on parts of the hip-related timeline, and especially on mortality, has been studied extensively. Even if the predictive power of this factor has been considered weak, the importance of comorbidity cannot be ignored (Glassou et al. 2014, 2017, Bulow et al. 2017). In view of the increasing length of survival following arthroplasty (Cnudde et al. 2018b) one has to consider that relevant comorbidity might develop and cause both morbidity (potentially leading to increased risk of infection and periprosthetic fracture) and mortality. Hunt et al. (2017) have described malignancies, cardiovascular disease, and respiratory disease as the main causes of mortality following arthroplasty surgery. In our study, patients who belonged to less educated groups were less likely to progress to a second-side operation but were equally likely to undergo revision surgery and had a higher risk of dying. The association between socioeconomic factors and increased risk of dying has been described by Whitehouse et al. (2014) in the United Kingdom and by Bennett et al. (2010) in the USA. Abdel et al. (2016) published the lifetime risk of revision using death as competing risk as well as Kaplan–Meier survivorship for a cohort of patients from their institution using the original Charnley cemented THA (DePuy International Ltd, Leeds, UK). We could identify the same effects of sex and age on both revision risk and death, using nationwide data and multiple implants as well as different fixation methods. We believe our study adds more support to the question of what will happen with time following a patient’s first replacement hip. Females are in generally more likely to receive bilateral hip replacements during their lifetime than their male counterparts and at an earlier age. We have not studied the influence of type of fixation in detail because there are too many possible confounding factors to consider within the framework of this study. However, the effect of fixation on revision at 10 years is, in our belief, quite important and strengthens the guidance that purely for revision reasons a cemented implant would be the implant of choice for patients older than 70 years. The differences in mortality rate can well be explained by patient selection as the uncemented implants in Sweden are mainly used in patients with better bone quality and better mobility. We cannot describe the effect of an anterior approach on the probability of transitions as during the study period this
12286 Cnudde D.indd 391
approach was used very rarely. Hunt et al. (2013) described improved early survival in patients operated with the posterior approach. We were unable to identify a statistically significant difference on most transitions with the exception of a minimal effect on bilateral procedures and decreased hazard ratio in transition from state 3 to 5 (revision of first hip to death). We were unable to show any statistically significant difference between approaches on mortality or revision rate. This gives further information for informed discussion on the choice of approach; the surgeon’s preference and possible differences in PROMs should, however, also be considered (Smith et al. 2012, Lindgren et al. 2014). Limitations and strengths For this analysis, patients who had had more than 1 ipsilateral revision were considered as staying in the revision group (they will remain in the same group). Patients undergoing multiple surgeries on the same hip might well have attendant possibilities for morbidity and these are not visible within the timeline. The number of patients that are undergoing multiple revisions has been limited and the complexity of adding additional states would, in our view, not be beneficial for the model. We are aware of bias in the decision-making regarding when to perform second-sided surgery/revision surgery. It is possible that some patients—despite being in need for revision or a contralateral hip operation—might not be operated upon as a result of decisions by the surgical team. We are also aware that some operations such as revisions for infections, periprosthetic fractures, and dislocations are under-reported within many of the registers (Slobogean et al. 2015). Comorbidity may well be under-reported as comorbidity records depend heavily on careful recording of comorbidity also recorded at the secondary care (Bulow et al. 2017). However, the existing measures of comorbidities have limited value in the case of revisions or death of THA patients. We believe that with an increased follow-up time our series could give a more in-depth view into what happens in the longer term. In addition, an association has been described between monoarthrodial pathology in the hip, the progression of degenerative change in the contralateral knee, and subsequent requirement for knee arthroplasty (Shakoor et al. 2014). Further studies have described associations with spinal or knee replacement surgery as degenerative or inflammatory changes are seldom limited to a single joint (Gillam et al. 2012, 2013). Our study, however, has not linked the available information with the knee or spine registries. The major strength of our study is the completeness and validity of the data within the register. Using the Swedish personal identity numbers it is unlikely we underreport mortality as every death is recorded by the tax office and subsequently in the SHAR database. A second strength is the longevity and the size of this register. This prospective nationwide program, collecting data from multiple surgeons working in multiple institutions, was set up as a quality improvement tool but the strength and validity of
the data can provide us with answers to many unsolved questions. Our results of this study contribute to a better understanding of the hip-related pathway patients are following after their initial surgery. Supplementary data Appendices 1–4 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/ 17453674.2018.1475179
Conceived the study: PC, SN, JT, OR, JK. Data collection: PC, SN. Statstical analysis: SN, EB, SW, JK. Drafting of the manuscript: PC, SN. All authors performed data analysis and editing of the manuscript. The authors would like to acknowledge the hard work of the register coordinators, Kajsa Erikson, Karin Davidsson, Karin Lindborg, and Karin Pettersson at the SHAR and the orthopedic surgeons and coordinators at all contributing hospitals for providing them with high-quality data. The authors are also very grateful to Charlotte Vitse and Daniel Odin for their contributions to the graphical representation of the multi-state analysis. Acta thanks Stein Atle Lie and Esa Jämsen for help with peer review of this study.
Abdel M P, Roth P V, Harmsen W S, Berry D J. What is the lifetime risk of revision for patients undergoing total hip arthroplasty? a 40-year observational study of patients treated with the Charnley cemented total hip arthroplasty. Bone Joint J 2016; 98-B(11): 1436–40. doi: 10.1302/0301620X.98B11.BJJ-2016-0337.R1. Andersen P K, Keiding N. Multi-state models for event history analysis. Stat Methods Med Res 2002; 11(2): 91–115. doi: 10.1191/0962280202SM276ra. Bayliss L E, Culliford D, Monk A P, Glyn-Jones S, Prieto-Alhambra D, Judge A, Cooper C, Carr A J, Arden N K, Beard D J, Price A J. The effect of patient age at intervention on risk of implant revision after total replacement of the hip or knee: a population-based cohort study. Lancet (London, England). 2017. doi: 10.1016/S0140-6736(17)30059-4. Bennett K M, Scarborough J E, Pappas T N, Kepler T B. Patient socioeconomic status is an independent predictor of operative mortality. Ann Surg 2010; 252(3): 552–7; discussion 7–8. doi: 10.1097/SLA.0b013e3181f2ac64. Bergh C, Fenstad A M, Furnes O, Garellick G, Havelin L I, Overgaard S, Pedersen A B, Makela K T, Pulkkinen P, Mohaddes M, Karrholm J. Increased risk of revision in patients with non-traumatic femoral head necrosis. Acta Orthop 2014; 85(1): 11–17. doi: 10.3109/17453674.2013.874927. Bulow E, Rolfson O, Cnudde P, Rogmark C, Garellick G, Nemes S. Comorbidity does not predict long-term mortality after total hip arthroplasty. Acta Orthop 2017: 88(5): 472-7. doi: 10.1080/17453674.2017.1341243. Burwell S M. Setting value-based payment goals: HHS efforts to improve U.S. health care. N Engl J Med 2015; 372(10): 897–9. doi: 10.1056/ NEJMp1500445. Cnudde P, Rolfson O, Nemes S, Karrholm J, Rehnberg C, Rogmark C, Timperley J, Garellick G. Linking Swedish health data registers to establish a research database and a shared decision-making tool in hip replacement. BMC Musculoskelet Disord 2016; 17(1): 414. doi: 10.1186/s12891-0161262-x. Cnudde P, Nemes S, Bulow E, Timperley J, Malchau H, Karrholm J, Garellick G, Rolfson O. Trends in hip replacements between 1999 and 2012 in Sweden. J Orthop Res 2018a; 36(1): 432-442doi: 10.1002/jor.23711.
12286 Cnudde D.indd 392
Acta Orthopaedica 2018; 89 (4): 386–393
Cnudde P, Rolfson O, Timperley AJ, Garland A, Karrholm J, Garellick G, Nemes S. Do patients live longer after THA and is the relative survival diagnosis-specific? Clin Orthop Relat Res 2018b. [Epub ahead of print] doi: 10.1007/s11999.0000000000000097. Commenges D. Multi-state models in epidemiology. Lifetime Data Anal 1999; 5(4): 315–27. Emilsson L, Lindahl B, Köster M, Lambe M, Ludvigsson J F. Review of 103 Swedish healthcare quality registries. J Intern Med 2015; 277(1): 94–136. doi: 10.1111/joim.12303. Farewell V T, Tom B D. The versatility of multi-state models for the analysis of longitudinal data with unobservable features. Lifetime Data Anal 2014; 20(1): 51–75. doi: 10.1007/s10985-012-9236-2. Gillam M H, Ryan P, Salter A, Graves S E. Multi-state models and arthroplasty histories after unilateral total hip arthroplasties: introducing the Summary Notation for Arthroplasty Histories. Acta Orthop 2012; 83(3): 220–6. doi: 10.3109/17453674.2012.684140. Gillam M H, Lie S A, Salter A, Furnes O, Graves S E, Havelin L I, Ryan P. The progression of end-stage osteoarthritis: analysis of data from the Australian and Norwegian joint replacement registries using a multistate model. Osteoarthritis Cartilage 2013; 21(3): 405–12. doi: 10.1016/j. joca.2012.12.008. Glassou E N, Pedersen A B, Hansen T B. Risk of re-admission, reoperation, and mortality within 90 days of total hip and knee arthroplasty in fast-track departments in Denmark from 2005 to 2011. Acta Orthop 2014; 85(5): 493–500. doi: 10.3109/17453674.2014.942586. Glassou E N, Pedersen A B, Hansen T B. Is decreasing mortality in total hip and knee arthroplasty patients dependent on patients’ comorbidity? Acta Orthop 2017; 88(3): 288–93. doi: 10.1080/17453674.2017.1279496. 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. doi: 10.1016/s0140-6736(13)61749-3. Hunt L P, Ben-Shlomo Y, Whitehouse M R, Porter M L, Blom A W. The main cause of death following primary total hip and knee replacement for osteoarthritis: a cohort study of 26,766 deaths following 332,734 hip replacements and 29,802 deaths following 384,291 knee replacements. J Bone Joint Surg Am 2017; 99(7): 565–75. doi: 10.2106/JBJS.16.00586. Jackson C H. Multi-State Models for Panel Data: The msm Package for R. J Stat Softw 2011; 38(8): 1-28. Johannson H R, Zywiel M G, Marker D R, Jones L C, McGrath M S, Mont M A. Osteonecrosis is not a predictor of poor outcomes in primary total hip arthroplasty: a systematic literature review. Int Orthop 2011; 35(4): 465–73. doi: 10.1007/s00264-010-0979-7. Jubelt L E, Goldfeld K S, Blecker S B, Chung W Y, Bendo J A, Bosco J A, Errico T J, Frempong-Boadu A K, Iorio R, Slover J D, Horwitz L I. Early lessons on bundled payment at an academic medical center. J Am Acad Orthop Surg 2017; 25(9): 654–63. doi: 10.5435/JAAOS-D-16-00626. Lie S A, Engesaeter L B, Havelin L I, Gjessing H K, Vollset S E. Mortality after total hip replacement: 0–10-year follow-up of 39,543 patients in the Norwegian Arthroplasty Register. Acta Orthop Scand 2000; 71(1): 19–27. doi: 10.1080/00016470052943838. Lindahl H, Malchau H, Oden A, Garellick G. Risk factors for failure after treatment of a periprosthetic fracture of the femur. J Bone Joint Surg Br 2006; 88(1): 26–30. doi: 10.1302/0301-620X.88B1.17029. Lindgren J V, Wretenberg P, Kaerrholm J, Garellick G, Rolfson O. Patient-reported outcome is influenced by surgical approach in total hip replacement. Bone Joint J 2014; 96B(5): 590–6. doi: 10.1302/0301620x.96b5.32341. Maradit Kremers H, Larson D R, Noureldin M, Schleck C D, Jiranek W A, Berry D J. Long-term mortality trends after total hip and knee arthroplasties: a population-based study. J Arthroplasty 2016; 31(6): 1163–9. doi: 10.1016/j.arth.2015.12.010. Marshall G, Jones R H. Multi-state models and diabetic retinopathy. Stat Med 1995; 14(18): 1975–83. doi: 10.1002/sim.4780141804.
Acta Orthopaedica 2018; 89 (4): 386–393
Meek R M, Allan D B, McPhillips G, Kerr L, Howie C R. Late dislocation after total hip arthroplasty. Clin Med Res 2008; 6(1): 17–23. doi: 10.3121/ cmr.2008.770. Pedersen A B, Baron J A, Overgaard S, Johnsen S P. Short- and long-term mortality following primary total hip replacement for osteoarthritis: a Danish nationwide epidemiological study. J Bone Joint Surg Br 2011; 93B(2): 172–7. doi: 10.1302/0301-620X.93B2.25629. Ranstam J, Karrholm J, Pulkkinen P, Makela K, Espehaug B, Pedersen A B, Mehnert F, Furnes O, group Ns. Statistical analysis of arthroplasty data, II: Guidelines. Acta Orthop 2011; 82(3): 258–67. doi: 10.3109/17453674.2011.588863. Sanders T L, Maradit Kremers H, Schleck C D, Larson D R, Berry D J. Subsequent total joint arthroplasty after primary total knee or hip arthroplasty: a 40-year population-based study. J Bone Joint Surg Am 2017; 99(5): 396– 401. doi: 10.2106/JBJS.16.00499. Schreurs B W, Hannink G. Total joint arthroplasty in younger patients: heading for trouble? Lancet 2017; 389(10077): 1374-5. doi: 10.1016/S01406736(17)30190-3. Shakoor N, Foucher K C, Wimmer M A, Mikolaitis-Preuss R A, Fogg L F, Block J A. Asymmetries and relationships between dynamic loading, muscle strength, and proprioceptive acuity at the knees in symptomatic unilateral hip osteoarthritis. Arthritis Res Ther 2014; 16(6): 455. doi: 10.1186/ s13075-014-0455-7.
12286 Cnudde D.indd 393
Shao Y, Zhang C, Charron K D, Macdonald S J, McCalden R W, Bourne R B. The fate of the remaining knee(s) or hip(s) in osteoarthritic patients undergoing a primary TKA or THA. J Arthroplasty 2013; 28(10): 1842–5. doi: 10.1016/j.arth.2012.10.008. Slobogean G P, Giannoudis P V, Frihagen F, Forte M L, Morshed S, Bhandari M. Bigger data, bigger problems. J Orthop Trauma 2015; 29(Suppl 12): S43–6. doi: 10.1097/BOT.0000000000000463. Smith A J, Wylde V, Berstock J R, Maclean A D, Blom A W. Surgical approach and patient-reported outcomes after total hip replacement. Hip Int 2012; 22(4): 355–61. doi: 10.5301/HIP.2012.9455. Whitehouse S L, Bolland B J, 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. doi: 10.1016/j. arth.2014.04.022. Zwartele R E, Brand R, Doets H C. Increased risk of dislocation after primary total hip arthroplasty in inflammatory arthritis: a prospective observational study of 410 hips. Acta Orthop Scand 2004; 75(6): 684–90.
Acta Orthopaedica 2018; 89 (4): 394–398
Patient claims in prosthetic hip infections: a comparison of nationwide incidence in Sweden and patient insurance data Piotr KASINA 1, Anders ENOCSON 1, 2, Viktor LINDGREN 2, and Lasse J LAPIDUS 1
1 Department of Clinical Science and Education, Södersjukhuset, Karolinska Institutet, Stockholm; 2 Department of Molecular Medicine and Surgery, Karolinska Institutet, Stockholm, Sweden Correspondence: email@example.com Submitted 2017-09-11. Accepted 2018-04-12.
Background and purpose — Patients in Sweden are insured against avoidable patient injuries. Prosthetic joint infections (PJIs) resulting from intraoperative contamination are regarded as compensable by the Swedish public insurance system. According to the Patient Injury Act, healthcare personnel must inform patients about any injury resulting from treatment and the possibility of filing a claim. To analyze any under-reporting of claims and their outcome, we investigated patients’ claims of PJI in a nationwide setting Patients and methods — The national cohort of PJI after primary total hip replacement, initially operated between 2005 and 2008, was established through cross-matching of registers and review of individual medical records. We analyzed 441 PJIs and the number of filed patients’ claims, with regards to incidence, outcome, and any national, sex-linked or socioeconomic differences. Results — We identified 329/441 (75%) patients with PJIs as non-claimants. 96% of the filed claims were accepted. 64 (57%) of claimants sustained permanent disability. 2 factors were found to statistically significantly reduce the odds of filing claims: patient’s age above 73 years and fracture as indication for surgery. There were no significant national, sex-linked, or socioeconomic differences. Interpretation — The incidence of patients’ claims of PJI is low but claims are usually accepted when filed. Healthcare personnel should increase their knowledge of the Patient Injury Act to inform patients about possibilities of eligible compensation.
Everyone treated in Sweden’s publicly financed healthcare is insured against injury resulting from avoidable patient injuries. Prosthetic joint infections (PJIs) are serious complications and may lead to severe consequences for those patients affected. PJIs after total hip replacement (THR) surgery, resulting from intraoperative contamination and not from hematogenous spread, are considered as compensable injuries by the Swedish no-fault insurance system. During the last decade over 16,500 primary THRs were performed in Sweden each year (Karrholm et al. 2016). A study of primary THRs operated between 2005 and 2008 showed that the incidence of PJIs in Sweden, up to 2 years after surgery, is 0.9% (Lindgren et al. 2014). The Swedish Patient Safety Act obligates healthcare personnel to inform patients about the possibility of filing a malpractice claim (SR-PSA 2017). Knowledge of the act varies among healthcare personnel (Espersson and Hellbacher 2016). Therefore there may be an under-reporting of patient injuries and consequently patients are not compensated to the extent intended by both the healthcare and legal system. The aim of this study was to determine the proportion of filed claims and the outcome of these claims (accepted, rejected, and disability). We also analyzed any presence of national, socioeconomic, age, and sex-linked differences in patient claims. The patient insurance scheme in Sweden Patients’ claims are handled by the national patient insurance company Landstingens Ömsesidiga Försäkringsbolag (LÖF), founded in 1975 and co-owned by the 21 Swedish county councils (public healthcare). LÖF handles all medical professions according to the Patient Injury Act (SR-PIA
© 2018 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 10.1080/17453674.2018.1477708
12082 Kasina D.indd 394
Acta Orthopaedica 2018; 89 (4): 397–398
2017) and regardless of the actual healthcare provider being a public, county council owned unit or a private division, delivering healthcare after procurement. The numbers of claims are increasing annually and in 2016 LÖF handled 16,000 claims with almost one-third of reimbursed claims being injuries related to orthopedic procedures (LÖF 2017). Moreover, THA is the procedure associated with the highest numbers of claims (Pukk-Härenstam et al. 2008). According to conditions of insurance, 6 types of injuries are covered, resulting from: treatment injury, technical damage, inferior diagnostics, infection, patient accidents in care, and medication-related injury. Patients report their injuries free of charge by completing a simple form, which must be filed within 3 years of injury detection but can under certain circumstances be accepted up to 10 years after treatment. LÖF then obtains full medical records before review of claims by specialists with expertise within the medical field concerned. In a case where the event was not avoidable and no causative relation or inferiority between given treatment and outcome is observed, LÖF can reject a claim. If the opposite is observed, LÖF compensates for the prolonged recovery time or awards payouts for sustained permanent disability, in accordance with Insurance Sweden consensus tables (IS 2014). The economic compensation is non-tort, reimbursing income loss, unreimbursed medical costs and also includes the possibility to recompense for pain and suffering caused by the injury. It is also blame-free for practitioners and no records are shared with the regulatory authorities. Therefore it is neither punitive damage compensation nor a sanctioning tool of healthcare providers but supplements the extensive coverage offered by Swedish social and medical care systems.
Patients and methods Data collection This study is based on a previously described population (Lindgren et al. 2014) regarding the national incidence of PJI after primary THR in Sweden. All patients who had undergone a primary THR, between July 1, 2005 December and 31, 2008, were initially retrieved from the Swedish Hip Arthroplasty Register (SHAR), n = 49,219. Their Swedish personal identification number was used to match patients with the Swedish Prescribed Drug Register for continuous outpatient antibiotic medication for at least 4 weeks after surgery. The observation time was set to 2 years postoperatively to include only early and delayed PJIs (Zimmerli and Ochsner 2003). This protocol allowed for inclusion of the compensable PJIs, since intraoperative contamination occurs within the first 2 years. Simultaneously the late appearing PJIs caused by hematogenous spread were excluded. No additional selection was made to exclude any possible hematogenously spread infections within the first 2 postoperative years. Antibiotic treatments with indications other then PJI were excluded. A questionnaire for each
12082 Kasina D.indd 395
of the identified infected 2,217 patients, with long postoperative antibiotic consumption, was sent to the operating units. This verified treatment for PJI and the case-specific diagnostic criteria of PJI, according to the definition by the Workgroup of the Musculoskeletal Infection Society (Parvizi et al. 2011). 99% of all questionnaires were returned and Lindgren et al. (2014) concluded a final number of 443 treated PJIs. Consequently, we could identify 441 PJIs in LISA (the national agency Statistics Sweden’s longitudinal integration database for health insurance and labor market studies). This enabled matching on level of education as a socioeconomic factor. Finally, we compared all PJIs with LÖF’s database in November 2016 for patients’ claims and outcome of claims review. The timeframe was regarded as sufficient both for patients to file claims after delayed infections in case of complicated and prolonged PJI treatment and also for LÖF to review and conclude its decision. Study variables For each patient, we recorded sex, age at primary THR operation, educational level, treating hospital, and the indication for surgery. SHAR divides these indications into 8 groups. Primary osteoarthritis and hip fracture are the 2 most common indications, with apparent different patient characteristics. Consequently we analyzed these 2 groups separately. The other 6 indications (inflammatory joint disease, sequel to pediatric disease, idiopathic femoral head necrosis, secondary osteoarthritis after trauma, malignancy, and other secondary osteoarthritis) each consist of fewer patients and are all mainly operated in elective settings. They were therefore merged together into 1 group, called “other”. Highest achieved educational level was also grouped, from initial 7 levels (elementary school < 9 years, elementary school > 9 years, 3 years’ high school, postgraduate < 3 years, postgraduate > 3 years, doctoral education, and unknown) to 4 levels: elementary school, high school, postgraduate, and unknown. To examine any national differences, each operating unit was classified according to its location by provision of care (the 21 Swedish counties) and separately by order: university hospital, referral county hospital, local hospital, and private. Filed patient claims at LÖF were recorded and their decisions were grouped into 6 outcomes: rejected, prolonged recovery (< 3 months and > 3 months), and permanent disability (1–15%, 16–30%, and > 30%). Statistics The Mann–Whitney U test was used for the age variable in independent groups and Pearson’s chi-square test was used for nominal variables. Logistic regression analysis was performed on patients’ characteristics to evaluate factors associated with insurance claims. We used both a univariable and multivariable model. The former includes age, sex, diagnosis, and level of education, and the latter model consists of age and diag-
Acta Orthopaedica 2018; 89 (4): 394–398
Table 1. Baseline data for all patients included in relation to claim of injury at LÖF. Values are number (percentage) unless otherwise stated All n = 441 Age, mean (SD) distribution < 50 51–60 61–70 71–80 > 80 Sex female male Diagnosis primary osteoarthritis hip fracture other Education elementary school high school postgraduate unknown Hospitals university county local private
No claim Claim (25%) n = 329 n = 112 p-value
15 (3) 66 (15) 116 (26) 167 (38) 77 (18)
10 (67) 45 (68) 78 (67) 126 (76) 70 (91)
5 (33) 21 (32) 38 (33) 41 (25) 7 (9)
222 (50) 219 (50)
158 (71) 171 (78)
64 (29) 48 (22)
328 (74) 67 (15) 46 (10)
236 (72) 58 (87) 35 (76)
92 (28) 9 (13) 11 (24)
205 (47) 167 (38) 66 (15) 3 (1)
161 (79) 119 (71) 46 (79) 3 (100)
44 (22) 48 (29) 20 (30) 0 (0)
55 (13) 203 (46) 151 (34) 32 (7)
46 (84) 158 (78) 104 (69) 21 (66)
9 (16) 45 (22) 47 (31) 11 (34)
SD = standard deviation.
nosis. These 2 factors were verified as statistically significant in the univariable model and are simultaneously important in a clinical setting. Associations are presented as odds ratios (ORs) and risk ratios (RRs) with 95% confidence intervals (CIs). We used IBM SPSS Statistics software (ver. 23) (IBM Corp, Armonk, NY, USA).
Ethics, funding, and potential conflicts of interest The study was conducted in accordance with the Declaration of Helsinki and the protocol was approved by the Ethical Review Board in Gothenburg, Sweden (622-16). We received support from LÖF, in the form of salary support for PK. LL has previously been involved in LÖF’s expert reviews but is not currently committed. There are no other conflicts of interest.
Results 329 (75%) of the 441 patients with PJIs did not file a claim for injury with LÖF. Of those 112 that did, 108 (96%) were accepted as eligible for compensation (Table 1). Patients’ age above the median of 72 years (OR = 0.4, CI 0.3–0.7, p = < 0.01) and fracture diagnosis (OR = 0.4, CI 0.2–0.9, p = 0.02) were the only significant factors associated with not filing a claim for injury in a univariate logistic regression model (Table 2). There was also a tendency toward higher rates of claims among female patients. When adjusted for age, sex, diagnosis, and level of education in the later multivariate model, they remained significant, with similar OR, CI, and p-values. The variation in claims between counties ranged between 0% and 50%. Patients from smaller hospitals filed claims more often, with the highest rates from the private hospitals (34%) and the lowest from university hospitals (16%) (p = 0.06). We could not observe any statistically significant correlations between age of patients and claim rates from the different types of hospitals. Private hospitals operated only on patients with osteoarthritis (except 1 patient with femoral head necrosis). 44/112 (39%) claimants were compensated for prolonged recovery time and 64 (57%) for permanent disability, due to
Table 2. Logistic regression to evaluate factors associated with insurance claim
Age < 73 years 73 years Sex male female Diagnosis primary osteoarthritis hip fracture other Level of education c elementary school secondary school university
Claim made n (%)
Univariable OR a (95%CI)
Multivariable OR a (95%CI)
RR b (95%CI) p-value
74 (33) 38 (18)
1 (reference) 0.4 (0.3–0.7)
1 (reference) 0.4 (0.3–0.7)
48 (22) 64 (29)
1 (reference) 1.4 (0.9–2.2)
328 67 46
92 (28) 9 (13) 11 (24)
1 (reference) 0.4 (0.2–0.8) 0.8 (0.4–1.7)
1 (reference) 0.4 (0.2–0.8) 0.7 (0.3–1.4)
0.5 (0.3–0.8) 0.8 (0.4–1.3)
205 167 66
44 (22) 48 (29) 20 (30)
1 (reference) 1.5 (0.9–2.4) 1.6 (0.9–3.0)
a OR = odds ratio. b RR = risk ratio. c 3 missing values.
12082 Kasina D.indd 396
Acta Orthopaedica 2018; 89 (4): 397–398
Number of patients
Age groups < 50 51–60 61–70 71–80 >80
Recovery Recovery Disability <3 months >3 months 1–15%
Disability > 30%
LÖF’s claims grouped by decisions (n) in the 112 claimed cases with distribution of age groups within.
PJI (Figure). There was an equal distribution of age groups within the 3 largest decision groups: recovery < 3 months, recovery > 3 months, and disability 1–15%.
Discussion Our main finding was the high incidence of non-claimants and that almost all filed claims were accepted for compensation by LÖF. This relation indicates on the one hand that those few patients who are claiming an injury from PJI are well informed about eligibility for compensation, but on the other hand it also suggests insufficiency of information from healthcare personnel to the large number of patients who do not make a claim. It is tempting to explain this by deficient knowledge of the Patient Injury Act among the personnel. An explanation may be a scarce awareness of legal obligations concerning patients’ information. Another is a possible incorrect impression of the system not being blame-free for practitioners, with a following unwillingness to report on colleagues if informing patients about LÖF. Third, in cases of fast and relatively complicationfree recovery, healthcare personnel may not inform patients based on their own judgment, an assumption of certain cases not being severe enough to generate compensation. There may also be other reasons contributing to not filing a claim; several of these have been discussed in previous publications (Sager et al. 1990, Studdert et al. 2000, Bismark et al. 2006, Järvelin et al. 2012, Zengerink et al. 2016). Although intraoperative contamination and hematogenously spread infections are not specified in our cohort, the probability of receiving compensation for PJI is high. We believe hematogenous spread was limited by the postoperative observation time and most PJIs are a result of IC (Zimmerli and
12082 Kasina D.indd 397
Ochsner 2003, Azzam et al. 2010, Lee et al. 2010, Blomfeldt et al. 2015, Sendi et al. 2017). Additionally, 59% of claimants suffered a permanent disability, which also has a great impact on compensation. Age above 73 years and fracture diagnosis were the 2 significant factors associated with lower rates of filed claims. Higher rates of non-claimants in the elderly correspond to previous findings (Sager et al. 1990, Studdert et al. 2000, Bismark et al. 2006, Järvelin et al. 2012, Zengerink et al. 2016). To our knowledge, specific claim rates of THR after fracture have not been studied previously. This finding suggests that healthcare personnel may be less prone to inform elderly patients, with possibly higher comorbidity, about the insurance. Increased age and fracture diagnosis, possibly due to associated comorbidities, are associated with poorer outcome after treatment of PJI (Azzam et al. 2010, Blomfeldt et al. 2015). Our cohort included only 9 claimants who suffered a fracture and did not allow for any further analysis. Age was not observed to affect LÖF’s decisions in our study. Since the overall incidence of claims was low, it is not possible to draw any conclusions concerning outcome after PJI in this population. A possible explanation for our equal distribution of claim outcomes is that the more frail patients never filed a claim. Our finding of a trend among women to be more prone to filing a claim is supported by the earlier studies of claims after THR (Järvelin et al. 2012, Zengerink et al. 2016) and claims in the general population (Bismark et al. 2006). Another trend is the higher rate of claimants among patients operated in private hospitals. Their population of elective patients, commonly healthier and more aware of their rights and entitlements, may explain this. The private units may also have a better dialogue with their patients, which could also be a contributive factor. Swedish compensation is substantially lower than seen in Anglo-American tort systems but simpler for the claimants and with higher overall appeal success rates (50 vs. 30%) (Kachalia et al. 2008, Pukk-Härenstam et al. 2008). This financial difference can be explained partly by the lack of a punitive component and partly by the existence of other forms of social insurance. The general support of the Swedish medical and social system may also diminish the economic importance of economic compensation. Additionally, elderly, injured, and often frail patients suffering from a PJI may refrain from filing claims due to perception of the process as difficult and long lasting. The fact of a free-of-charge, 2-page form and the relatively short time to decision (70% within 8 months) (Kachalia et al. 2008) is not obvious to the general population. Finally, patients may not realize the blame-free nature of the insurance. Some may refrain from claiming for their injury, as they are not willing to blame their doctor or department. A limitation of our study is the inability to measure any rate of patient information about LÖF. We cannot conclude that non-claimants were not informed, nor can we assess the efficiency of delivery or quality of given information. Another
limitation is the inability to measure any correlation between clinical outcome of PJI treatment and the likelihood of claiming injury. Additionally, we were not able to exclude hematogenously spread infections in our cohort beyond the limitation of the 2-year postoperative observation period. Therefore the presence of any hematogenously spread infections among non-claimants cannot be assessed. However, there were no such infections among the claimants and none of the 4 cases were rejected due to the route of contamination. Nonetheless, we strongly suspect that most patients may not be aware that they have sustained an injury that is recognized by the healthcare system and possibly compensated by an insurance. It is therefore important to clearly inform patients suffering from PJI about their legal right to file a claim with LÖF and provide assistance when needed. In summary our study shows that only every fourth PJI after THA is claimed as a patient injury. Simultaneously almost all claims were reimbursed by the Swedish national patient insurance, indicating that patients who file claims are informed about their rights. The low claim rate suggests insufficient patient information and is of concern from both a legal and a professional aspect.
All authors contributed to the conception and design of the study, analysis of data and its interpretation, as well as revision of the manuscript. PK and AE performed the statistical analysis. Acta thanks Pieter K Bos and Jutta Järvelin for help with peer review of this study.
Azzam K A, Seeley M, Ghanem E, Austin M S, Purtill J J, Parvizi J. Irrigation and debridement in the management of prosthetic joint infection: traditional indications revisited. J Arthroplasty 2010; 25(7): 1022-7. Bismark M, Brennan T A, Davis P B, Studdert D M. Claiming behaviour in a no-fault system of medical injury: a descriptive analysis of claimants and non-claimants. Med J Aust 2006; 185(4): 203-7. Blomfeldt R, Kasina P, Ottosson C, Enocson A, Lapidus L J. Prosthetic joint infection following hip fracture and degenerative hip disorder: a cohort study of three thousand, eight hundred and seven consecutive hip arthroplasties with a minimum follow-up of five years. Int Orthop 2015; 39(11): 2091-6. Espersson C, Hellbacher U. Patientskadelagen – en kommentar m. m. Vulkan, Riga 2016.
12082 Kasina D.indd 398
Acta Orthopaedica 2018; 89 (4): 394–398
IS. Insurance Sweden. Medicinsk Invaliditet –skador 2013. Insurance Sweden 2014. Available from: http://www.svenskforsakring.se/globalassets/medicinska-tabellverk/medicinska-tabellverk/medicinsk_invaliditet_skador_ rev_jan2014.pdf. Järvelin J, Häkkinen U, Rosenqvist G, Remes V. Factors predisposing to claims and compensations for patient injuries following total hip and knee arthroplasty. Acta Orthop 2012; 83(2): 190-6. Kachalia A B, Mello M M, Brennan T A, Studdert D M. Beyond negligence: avoidability and medical injury compensation. Soc Sci Med 2008; 66(2): 387-402. Karrholm J, Lindahl H, Malchau H, Mohaddes M, Rogmark C, Rolfson O. Swedish Hip Arthroplasty Register: Annual Report 2015. Gothenburg, Sweden 2016. Available from: https://registercentrum.blob.core.windows. net/shpr/r/Annual-Report-2015-H19dFINOW.pdf Lee J, Kang C I, Lee J H, Joung M, Moon S, Wi Y M, Chung D R, Ha C W, Song J H, Peck K R. Risk factors for treatment failure in patients with prosthetic joint infections. J Hosp Infect 2010; 75(4): 273-6. Lindgren V, Gordon M, Wretenberg P, Kärrholm J, Garellick G. Deep infection after total hip replacement: a method for national incidence surveillance. Infect Control Hosp Epidemiol 2014; 35(12): 1491-6. LÖF. Injury Report 2016. Anmälningar till LÖF 2016. Available from: http:// lof.se/wp-content/uploads/Statistik-2016-Hela-Sverige.pdf. Landstingens Ömsesidiga Försäkringsbolag; 2017. Parvizi J, Zmistowski B, Berbari E F, Bauer T W, Springer B D, Della Valle C J, Garvin K L, Mont M A, Wongworawat M, Zalavras C G. New definition for periprosthetic joint infection: from the Workgroup of the Musculoskeletal Infection Society. Clin Orthop Rel Res 2011; 469(11): 2992-4. Pukk-Härenstam K, Ask J, Brommels M, Thor J, Penaloza R V, Gaffney F A. Analysis of 23 364 patient-generated, physician reviewed malpractice claims from a non-tort, blamefree, national patient insurance system: lessons learned from Sweden. Qual Saf Health Care 2008; 17(4): 259-63. Sager M, Voeks S, Drinka P, Langer E, Grimstad P. Do the elderly sue physicians? Arch Intern Med 1990; 150(5): 1091-3. Sendi P, Lotscher PO, Kessler B, Graber P, Zimmerli W, Clauss M. Debridement and implant retention in the management of hip periprosthetic joint infection. Bone Joint J 2017; 99-B(3): 330-6. Studdert D M, Thomas E J, Burstin H R, Zbar B I, Orav E J, Brennan T A. Negligent care and malpractice claiming behavior in Utah and Colorado. Med Care 2000; 38(3): 250-60. SR-PIA. Sveriges Riksdag. Svensk författningssamling, Patientskadelag(1996:799) [The Swedish Parliament, Patient Injury Act], 2017:43. Available from: http://www.riksdagen.se/sv/dokument-lagar/dokument/ svensk-forfattningssamling/patientskadelag-1996799_sfs-1996-799. SR-PSA. Sveriges Riksdag. Svensk författningssamling, Patientsäkerhetslag(2010:659). [The Swedish Parliament, Patient Safety Act], 2017:786. Available from: http://www.riksdagen.se/sv/dokument-lagar/dokument/ svensk-forfattningssamling/patientsakerhetslag-2010659_sfs-2010-659. Zimmerli W, Oshsner P E. Management of infection associated with prosthetic joints. Infection 2003; 31(2): 99-108. Zengerink I, Reijman M, Mathijssen N M, Eikens-Jansen M P, Bos P K. Hip arthroplasty malpractice claims in the Netherlands: closed claim study 2000–2012. J Arthroplasty 2016; 31(9): 1890-3.
Acta Orthopaedica 2018; 89 (4): 399–405
Outcomes following hip and knee replacement in diabetic versus nondiabetic patients and well versus poorly controlled diabetic patients: a prospective cohort study Erik LENGUERRAND 1, Andrew D BESWICK 1, Michael R WHITEHOUSE 1,2, Vikki WYLDE 1,2, and Ashley W BLOM 1,2
1 Musculoskeletal Research Unit, Translational Health Sciences, Bristol Medical School, University of Bristol, Bristol; 2 National Institute for Health Research, Bristol Biomedical Research Centre, University Hospitals Bristol NHS Foundation Trust and University of Bristol, UK Correspondence: firstname.lastname@example.org Submitted Accepted
Background and purpose — The impact of diabetes and glycemic control before joint replacement on clinical and patient-reported outcomes is unclear. We compared pain, function, complications, and length of hospital stay in diabetic and nondiabetic patients receiving primary total hip (THR) or knee replacement (TKR) and compared these outcomes in patients with poorly controlled versus well-controlled diabetes. Patients and methods — We conducted a prospective cohort study of patients undergoing primary THR (n = 300) or TKR (n = 287) for osteoarthritis. Self-reported diabetes and glycemic control (HbA1c ≤ or >7%) extracted from medical notes were used. Adjusted comparisons were performed with generalized linear models including body mass index (BMI) and comorbidities. Results — Diabetes prevalence was 11% (THR 8%; TKR 14%). Diabetic patients were more likely to have a higher BMI and greater number of comorbidities. The median length of hospital stay was 1 day longer in diabetic patients (p = 0.004), but this attenuated after adjustments for BMI and comorbidities (p = 0.3). Inpatient pain was greater for diabetic patients but attenuated following adjustment. The 12-month postoperative WOMAC subscales were similar by diabetes status following adjustment. There was little evidence of difference in outcomes according to glycemic control. Interpretation — The associations between diabetes and worse postoperative outcomes in patients undergoing THR or TKR for osteoarthritis appear to be predominantly due to associated obesity and comorbidities. In diabetic patients there is little evidence of association between postoperative outcome and preoperative glycemic control. The underlying mechanisms and causal pathways of obesity, diabetes, and multimorbidity that lead to worse outcomes after joint replacement are not well known.
Multimorbidity is common among people undergoing joint replacement, with approximately 70% of patients reporting at least 1 condition additional to osteoarthritis. Prevalent comorbid conditions include degenerative disc disease, osteoporosis, visual and hearing impairment, anxiety (Blom et al. 2016). In the USA, about 14% of people with osteoarthritis have diabetes (Bhattacharyya et al. 2002) and in a large UK cohort, approximately 6% and 10% of patients undergoing THR and TKR respectively reported having diabetes (Arden et al. 2017). After joint replacement surgery, higher rates of complications such as pneumonia, stroke, bleeding (Bolognesi et al. 2008) and prosthetic joint infection (Kunutsor et al. 2016) are experienced by diabetic patients. Some studies report that diabetes does not affect pain and function after joint replacement as measured by patient-reported outcome measures (Cushnaghan et al. 2009, Clement et al. 2013), but at least 1 study has shown transient worse functional outcome in diabetic patients after TKR, which resolved over time (Robertson et al. 2012). The importance of tight glycemic control before surgery in diabetic patients is contentious. A USA study of nearly 1 million patients undergoing hip or knee replacement reported that the risk of postoperative complications was greater in the 3.6% of patients with poorly controlled diabetes based on criteria including self-monitoring, treatment and HbA1c > 7% (Marchant et al. 2009). However, a more recent systematic review of cohort studies showed that an HbA1c > 7% before surgery, reflecting poorly controlled diabetes, was not consistently associated with postoperative morbidity or mortality (Rollins et al. 2016). The aims of this study were 2-fold. First, we compared complications, pain, function, and length of stay in diabetic and nondiabetic patients receiving a primary THR or TKR. Second, we compared these outcomes between patients with poorly controlled diabetes and well-controlled diabetes.
© 2018 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 10.1080/17453674.2018.1473327
12298 Lenguerrand D.indd 399
Acta Orthopaedica 2018; 89 (4): 399–405
Site of surgery
Patients and methods The reporting of this cohort study follows STROBE guidelines. Study population We performed a secondary analysis of data from participants in the Arthroplasty Pain Experience (APEX) trials (Wylde et al. 2015). Between 2009 and 2012, 322 patients receiving primary THR and 316 patients receiving primary TKR in the UK were recruited into 2 single–center randomized controlled trials that investigated the effectiveness of local anesthetic wound infiltration in reducing chronic pain after joint replacement. Patient inclusion criteria were a primary unilateral THR or TKR for osteoarthritis. Exclusion criteria included inability to provide informed consent or complete questionnaires, and medical comorbidity precluding use of spinal anesthesia, regional blocks, and strong analgesics postoperatively. The study sample included the APEX participants who underwent surgery and reported their diabetic status; patients who withdrew prior to surgery or during the inpatient stay were excluded. Details of the sample size calculation for the original trial have been published previously (Wylde et al. 2015). Patient characteristics Information on age, sex, and BMI were extracted from participants’ medical records by a research nurse. Self-reported comorbidities were collected preoperatively with the 18-item Functional Co-morbidity Index (FCI) (Groll et al. 2005). Psychological distress was assessed with the Hospital Anxiety and Depression Scale (Zigmond and Snaith 1983). Health-related quality was measured using the EQ-5D (Williams and Kind 1992). Participants completed the 24-item Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (Bellamy et al. 1988). This tool has 3 dimensions: pain, stiffness, and physical function. Separate subscale scores were calculated and transformed to a 0–100 scale (worst to best). Treatment allocation in the original trial was also considered. Diabetic status Preoperative diabetic status was assessed as part of the FCI, which asks patients whether they have been diagnosed with diabetes. The FCI does not differentiate between type 1 and type 2 diabetes. During the preoperative clinic assessment a research nurse screened the medication usage of all patients for diabetes medication. Relevant reported drugs are listed in Table 1 (see Supplementary data). Diabetic patients were classified in two groups depending on their glycemic control as determined by preoperative HbA1c with a cut-off of > 7%(53 mmol/mol) (American Diabetes Association 2017).
12298 Lenguerrand D.indd 400
Figure 1. Directed acyclic graph depicting the potential associations underlying the effect of the main exposures on the outcomes. U: unmeasured dietary habits and sedentary lifestyle factors. Exposure: diabetes mellitus or HbA1c status. Outcome: patient reported, clinical or complication outcomes.
Postoperative outcomes Outcomes collected during the inpatient stay included the length of hospital stay and acute postsurgical pain. Participants were asked to “indicate the intensity of your present hip/ knee pain” at rest and on movement for 3 days post-surgery using a visual analogue scale from 0 to 10 cm (best to worst). The mean VAS scores at rest and on movement across the 3 postoperative days were calculated. At 3 and 12 months post-operation a research nurse extracted details of postoperative complications from the medical notes. These included: infection; dislocation; deep vein thrombosis; pulmonary embolism; nerve damage; hospital readmission related to the index operation; and further surgery on the index joint. The number of inpatient hospital stays, day case, and outpatient visits related to the index joint were also collected at the 12-month follow-up. Participants were sent a postal questionnaire at 3, 6, and 12 months postoperatively, which included the WOMAC questionnaire. The WOMAC pain subscale measured at 12 months postoperatively was the primary outcome of the APEX trials. Statistics THR and TKR patients were analyzed together to ensure a sufficient sample of diabetic patients with good and poor glycemic control. Patient characteristics and outcomes are reported by diabetic status and glycemic control (HbA1c level ≤ or > 7%) using frequencies and percentages for categorical variables and means (SD) or medians with percentiles (25th, 75th) for continuous variables, depending on their distribution. We first investigated the unadjusted relationship between diabetes status or glycemic control (exposure factors) and the different outcomes. Multivariable regression models were then fitted to condition these associations on age, sex, site of surgery, BMI, and number of comorbidities and therefore determine whether the exposure factors had an independent and direct relationship with the studied outcomes. These covariates were imbalanced between exposure levels (Tables 2 and 5, see Supplementary data) and have evidence of association
Acta Orthopaedica 2018; 89 (4): 399–405
with the outcomes in the literature (Judge et al. 2010, 2012) and in univariable models. As shown in the directed acyclic graph (Figure 1), none of them were expected to be colliders but were required in the models to “block open-paths”, i.e., to control for confounding bias or “non-causal structural associations” (Shrier and Platt 2008). The first multivariable models included adjustment for the non-modifiable factors age, sex, site of surgery, and trial intervention. The variable “trial intervention” was forced as it is related to the design of the primary study for which the analyzed data were originally collected. BMI and comorbidity profile—(partially) “modifiable factors” in the perioperative period with non-surgical or pharmaceutical interventions—were finally added to these first multivariable models to identify their specific impact on the exposure–outcome associations. To account for repeated measures of the studied outcomes, we used linear mixed models for unadjusted and adjusted analyses. The model residuals were not normally distributed (checked with normal quantile–quantile plot) due to outcome distribution heterogeneity across the postoperative period: the scores were normally distributed until 3 months postoperatively but substantially skewed thereafter, in particular at 12 months when numerous patients experienced no or minor symptoms. Each outcome assessment was therefore separately investigated. Unadjusted investigations were performed with unpaired Student t-tests or Mann–Whitney tests for continuous variables and chi-square or Fisher’s exact test for categorical variables. For the adjusted investigations, multivariable linear regression models were used. In the absence of appropriate Box–Cox transformation to use the linear regression model in agreement with its assumptions, outcomes were categorized and analyzed with an ordered logistic regression model. Length of hospital stay was modelled as an ordered categorical variable, ≤ 3, 4, 5, and > 5 days (respectively 20%, 29%, 22%, and 29% of the sample). The WOMAC scores were transformed using previously published threshold definitions (severe [0–50], moderate [51–75], mild [76–99], no  symptoms) (Wylde et al. 2015). The proportionality assumption was investigated with the Brant test and none of the presented models violated this assumption. Multiple imputation by chained equations (MICE) was also performed to generate 20 imputations sets and investigate bias induced by missing data on the estimations of the completecases analyses (White et al. 2011): patients with high BMI or larger number of comorbidities had more missing item information for the postoperative WOMAC scores. The imputation model included all variables used in the adjusted model, variables presented in Table 1, and all repeated measurements of the studied outcomes. Estimates were combined using Rubin’s rules. Analyses were conducted in Stata (Stata Statistical Software: Release 14.1—StataCorp LLC, College Station, TX, USA).
12298 Lenguerrand D.indd 401
Ethics, registration, funding, and potential conflicts of interest The original trials were registered in the Clinical Trial Registry (ISRCTN96095682) and had received ethics approval (NHS-REC South West 09/H0504/94). All participants provided informed written consent. This article presents independent research funded by the NIHR under its Programme Grants for Applied Research program (RP-PG-0407-10070). The authors acknowledge the support of the NIHR, through the Comprehensive Clinical Research Network. This study was also supported by the NIHR Biomedical Research Centre at University Hospitals Bristol NHS Foundation Trust and the University of Bristol. The views expressed in this publication are those of the authors and not necessarily those of the NHS, the National Institute for Health Research or the Department of Health.There are no conflicts of interest to be reported by the authors.
Results Participants A total of 304 participants undergoing THR and 290 participants undergoing TKR had complete preoperative information and remained in the APEX trials at the time of their discharge from hospital. Seven patients with no details of their preoperative diabetic status were excluded. Of the remaining 587 patients, 29 (24 nondiabetic and 5 diabetic patients) withdrew from the study or died during the follow-up period (Table 2). The trial intervention to which patients had been allocated was balanced between diabetic and nondiabetic patients. The mean age of participants was 68 years (SD 10) and 55% were female. The prevalence of diabetes in the overall sample was 11%, higher in the TKR group than in the THR group (14% vs. 8%). Diabetic patients were more likely to have a higher BMI (mean (SD): 34 (6) vs. 31 (6) for nondiabetic patients) and more comorbidities (mean (SD): 5 (2) vs. 3 (2) comorbidities). All patients with a medication for diabetes had correctly reported their diabetes in the FCI. Of the 64 diabetic patients, 38 were on medication for diabetes prior to their surgery. All treated patients were on metformin tablets; 6 had also been prescribed insulin injection and 1 had been prescribed exenatide injection. Postoperative outcomes by diabetic status A comparison of postoperative outcomes by preoperative diabetic status is provided in Table 3 (see Supplementary data). Diabetic patients had a longer hospital stay than those without diabetes (5 vs. 4 days, p = 0.004). This difference was not explained by age, sex, or the type of surgery, but was not apparent when BMI and the comorbidity profile were accounted for. During the first 3 days of their inpatient stay, diabetic patients reported a higher mean pain level on movement and at rest,
Acta Orthopaedica 2018; 89 (4): 399–405
Table 2. Sample characteristics by diabetic status at preoperative assessment. Values are number (percentage) unless otherwise stated Total sample n = 587 Trial intervention Injection 294 (50) Standard care 293 (50) Primary total joint replacement Hip 300 (51) Knee 287 (49) Age, mean (SD) 68 (10) Sex Female 325 (55) Male 262 (45) BMI, mean (SD) 31 (6) 30 303 (52) > 30 284 (48) EQ VAS a, n 553 mean (SD) 66 (20) EQ-5D Score b, n 555 mean (SD) 0.42 (0.32) FCI c (Number of comorbidities) <2 192 (33) 2 168 (29) 3 115 (20) >3 112 (19) HADS d Total, n 567 mean (SD) 13 (7) WOMAC e Pain, n 587 mean (SD) 42 (17) WOMAC Function, n 558 mean (SD) 44 (18) WOMAC Stiffness, n 548 mean (SD) 44 (22) WOMAC Total, n 537 mean (SD) 44 (17)
Nondiabetic patients n = 523
Diabetic patients n = 64
263 (50) 260 (50)
31 (48) 33 (52)
276 (53) 247 (47) 68 (10)
24 (37) 40 (63) 70 (8)
295 (56) 228 (44) 31 (6) 282 (54) 241 (46) 492 66 (20) 495 0.43 (0.32)
30 (47) 34 (53) 34 (6) 21 (33) 43 (67) 61 64 (19) 60 0.37 (0.31)
178 (34) 156 (30) 102 (19) 87 (17) 505 13 (7) 523 43 (18) 499 44 (18) 487 43 (22) 479 44 (17)
14 (22) 12 (19) 13 (20) 25 (39) 62 13 (6) 64 41 (15) 59 43 (16) 61 44 (18) 58 44 (14)
a EQ visual analogue scale (0–100, worst to best health state). b EQ-5D 3L descriptive system. c Modified Functional Co-morbidity Index without BMI and diabetes
diagnoses (categorized sum of the 16 remaining diagnoses). Anxiety and Depression Scale with anxiety and depression scores combined (0–42, best to worst distress state). e Western Ontario and McMaster Universities Osteoarthritis Index (0–100 worst to best pain, function, stiffness or total score). d Hospital
but differences were not apparent when adjustments were included for BMI and the comorbidity profile. No evidence of differences in postoperative complications were observed at either the 3- or 12-month time points. The number of inpatient stays, and day case and outpatient visits related to the operated joint did not differ between diabetic and nondiabetic patients. WOMAC pain, function, and stiffness at the 3-month and 6-month assessments did not differ between diabetic and nondiabetic patients. At 12 months, diabetic patients reported worse pain, function, and stiffness, but these differences were partially attenuated when the linear regression model included adjustments for age, sex, site of surgery, and trial intervention and not evident with further adjustments for BMI and number
12298 Lenguerrand D.indd 402
of comorbidities. Similar results were found in the imputed analyses (Table 4, see Supplementary data). Postoperative outcomes by HbA1c level for patients with diabetes A comparison of postoperative outcomes by HbA1c level for patients with diabetes is provided in Table 6 (see Supplementary data). 25 of the 64 diabetic patients had a preoperative HbA1c level > 7%. Their mean HbA1c was 8.5% (SD 1.6) compared with 6.2% (SD = 0.6) for patients with HbA1c ≤ 7% (Table 6, see Supplementary data). Patients with HbA1c > 7% had greater preoperative stiffness and a lower quality of life, but were less likely to have more than 3 comorbidities. They were also more likely to be on medication for their diabetes. Comparison of outcomes in diabetic patients by HbA1c level revealed little difference between groups. The median length of hospitalization did not differ between HbA1c groups. Inpatient postoperative VAS pain scores were also comparable. Those with a preoperative HbA1c ≤ 7% maintained lower levels of postoperative HbA1c compared with those who had higher preoperative HbA1c level, but no evidence of pre–postoperative change in the level of HbA1c was found within each group (paired t-test: p = 0.7 within the HbA1c ≤ 7% group; p = 0.2 within the HbA1c > 7% group). After adjustments for confounding factors, diabetic patients with HbA1c >7% reported more pain on the WOMAC scale than those with HbA1c < 7% at 6 months but not 3 months or 12 months postoperatively. Little evidence of difference in the WOMAC stiffness scores was found between the 2 HbA1c groups at the 3-month and 6-month assessments, but at 12 months postoperatively patients with HbA1c > 7% reported a higher level of stiffness. This difference remained significant in the adjusted comparisons and after adjusting further for preoperative level of WOMAC stiffness (p < 0.0001). WOMAC function and total WOMAC scores were similar between HbA1c groups. The median numbers of post-surgery complications, inpatient stay, day case or outpatient visits were also comparable between the HbA1c ≤ 7% and HbA1c > 7% groups. Similar results were found in the imputed analyses (Table 7, see Supplementary data).
Discussion In this study of 587 patients receiving THR or TKR, unadjusted analysis revealed that patients with diabetes had longer hospital stays, more severe acute postoperative pain, and worse patient-reported outcomes at 12 months after surgery than nondiabetic patients. However, when the analysis included adjustment for comorbidities and BMI, these differences in outcomes were no longer apparent. This adds to the existing knowledge by showing that diabetes per se may not
Acta Orthopaedica 2018; 89 (4): 399â&#x20AC;&#x201C;405
be related to poorer outcomes after joint replacement, independent of obesity and multimorbidity. This study has limitations that should be acknowledged when interpreting the results. THR and TKR patients were analyzed together to ensure a sufficient sample in which to investigate disparities in outcome by diabetes and glycemic status. Although the multivariable analyses were adjusted for surgical site, including both TKR and THR patients in the analysis may have confounded the results because of differences between these 2 groups, such as a higher prevalence of diabetes in patients with knee osteoarthritis and poor outcomes after TKR. Investigation of the associations between diabetes/glycemic status and outcomes was limited to patient characteristics collected in the original trial and based on an observational study design approach. Residuals and unmeasured confounding cannot be ruled out and the few independent associations observed in this analysis may be explained by other unobserved factors. This limitation does not affect the associations that were explained by the observed confounders accounted for in the adjusted models. The sample size of this study was initially determined by a power calculation used to address the primary research aim of the original trial. As such, the comparisons by diabetic status performed in this study should be interpreted with care and acknowledgement that a larger sample and/or different study design may be required to draw definitive, causal conclusions. Comparisons by HbA1c status were certainly underpowered and therefore should be considered as purely exploratory. A larger sample, such as that provided by a national arthroplasty registry linked to national primary, secondary, and tertiary care use registries, would be required to investigate the effect of diabetes on rare complications such as infection (Lenguerrand et al. 2017); however glycemic status, pain, or function outcomes are not collected on all patients recorded in these large datasets. Very few studies have investigated the role of diabetes and HbA1c on both clinical complications and longitudinal patient-reported outcomes, especially outside the USA. Our analysis adds to the existing literature, and highlights the need for further research, particularly in a European context. Diabetic status was identified using the patient-reported FCI tool. This self-reported status was cross-checked with the preoperative list of medications and all patients with a medication for diabetes had correctly reported their diabetes in the FCI. Although all patients underwent routine preoperative blood glucose measurement, it is possible that some patients in our comparator group may have had undiagnosed and untreated diabetes. Type 1 and 2 diabetes cannot be differentiated in the FCI and were thus analyzed together. Type 1 diabetes has a greater negative impact on postoperative complications than Type 2 diabetes, and therefore it may have been more appropriate to have excluded or analyzed these patients separately (Viens et al. 2012). However, most treated patients in this study were on metformin tablets only, suggesting that this lack of differentiation by diabetic type is likely to have little impact
12298 Lenguerrand D.indd 403
on our results, which are more generalizable to those patients undergoing lower joint arthroplasty with Type 2 diabetes, the most frequent type among adults. No information on the duration and severity of the disease were available. Also, no information was available on the number of diabetic patients who had their surgery delayed until their glycemia was controlled/ treated, nor about the length of any delay. At the time of the study it was not unit local policy to delay surgery for poor glycemic control. Finally, this work is an ad-hoc, un-prespecified secondary analysis of data initially collected for a randomized trial. The selection criteria (Wylde et al. 2015) used to identify and recruit patients into the original trial were not related to diabetes and are not likely to have introduced any selection bias towards a particular group of diabetic/nondiabetic patients. The patients randomized had also been found to be representative of patients receiving hip and knee replacement in England and Wales. Moreover, the multivariable analyses included an adjustment for intervention allocation. The risks of selection bias associated with the secondary use of randomized trial data and â&#x20AC;&#x153;interventionâ&#x20AC;? bias are low in this analysis. Extended postoperative hospitalization for patients with diabetes has been reported previously in orthopedic populations (Marchant et al. 2009, Lovecchio et al. 2014, Winemaker et al. 2015), and diabetic patients are more likely to have a non-routine discharge (Bolognesi et al. 2008, Marchant et al. 2009). However, most of these studies did not adjust for BMI and key comorbidities, 2 sets of factors strongly associated with diabetes. Our results imply that extended hospitalization is driven by clinical characteristics commonly encountered in diabetic patients rather than by diabetes in isolation. Indeed, within our sample of patients without diabetes, multimorbidity and obesity were also associated with prolonged hospitalization. Efforts to decrease length of stay should thus target those with obesity and multimorbidity, irrespective of the presence of diabetes. In terms of complications, our analysis found that rates of postoperative complications, hospital readmissions, and daycase or outpatient visits were comparable between patients with and without diabetes. This may be due to the low rate of postoperative complications after joint replacement meaning that our study was underpowered to detect differences despite having nearly 600 patients. Previous research that has evaluated the relationship between diabetic status and functional outcomes after joint replacement has reported conflicting results (Gandhi et al. 2010, Singh and Lewallen 2013). We found that patients with diabetes had worse functional results, but not when obesity and multimorbidity were accounted for in the models. In patients with diabetes, functional outcomes and length of stay were comparable between those with good and poor glycemic control. Some large observational studies have found a detrimental effect of uncontrolled glycemic level on these outcomes (Marchant et al. 2009, Jamsen et al. 2010), but our findings are in agreement with a systematic review published on
this topic, which concluded that elevated preoperative HbA1c was not definitively associated with increased postoperative morbidity or mortality in patients with diabetes (Rollins et al. 2016). No, or weak, evidence of an association between diabetes, HbA1c level, and all-cause rehospitalization has previously been reported (Adams et al. 2013). Concerns have been raised as to suboptimal management of diabetes in surgical patients in the UK, particularly with reference to optimization of blood glucose control (Howieson et al. 2014). Our work suggests that in the joint replacement setting optimal preoperative glycemic control may not be critical in determining outcome. This gives further credence to our finding that diabetes is not the major driver of extended hospitalization in diabetic patients undergoing joint replacement. Our findings regarding pain are mixed, with evidence of increased pain in those with worse glycemic control at earlier time points, but no differences in the longer term. In our analysis that combined patients undergoing THR and TKR, we found no association between the presence of diabetes and postoperative stiffness. In contrast other authors have reported increased stiffness in diabetic patients undergoing TKR (Bawa et al. 2013); evidence for THR is scarce. As we included patients undergoing THR this may explain the difference with those studies. However, we did find that patients with poorly controlled diabetes (72% of them had undergone TKR) had a higher level of stiffness than those with well-controlled diabetes. This may be due to increased scar formation in hyperglycemic patients or difficulty for those with poorly controlled diabetes in complying with their postoperative rehabilitation leading to stiffness. In summary, we found that the associations between diabetes and worse postoperative outcomes were due to obesity and comorbidities. In diabetic patients there is little evidence of associations between preoperative glycemic control and postoperative outcomes. Further research is required to understand the underlying mechanisms and causal pathways of obesity, diabetes, and multimorbidity that lead to worse outcomes after joint replacement. Evaluation of care packages to optimize the management of obesity and multimorbidity in diabetic patients is needed to determine whether these could improve outcomes following joint replacement. Supplementary data Tables 1 and 3–7 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/ 17453674.2018.1473327
Study conception and design: EL, ADB, MRW, VW, AWB. Data acquisition: ADB, VW, AWB. Data analysis: EL. Data interpretation: EL, ADB, MRW, VW, AWB. Drafting of manuscript: EL, ADB, MRW, VW, AWB. Acta thanks Per Kjærsgaard-Andersen and Kjell G Nilsson reviewers for help with peer review of this study.
12298 Lenguerrand D.indd 404
Acta Orthopaedica 2018; 89 (4): 399–405
Adams A L, Paxton E W, Wang J Q, et al. Surgical outcomes of total knee replacement according to diabetes status and glycemic control, 2001 to 2009. J Bone Joint Surg Am 2013; 95(6): 481-7. American Diabetes Association. Standards of Medical Care in Diabetes. Diabetes Care 2017; 40(Suppl 1): S1-S134. Arden N, Altman D, Beard D, et al. Lower limb arthroplasty: can we produce a tool to predict outcome and failure, and is it cost-effective? An epidemiological study. Southampton, UK: NIHR Journals Library; 2017 June. Bawa H S, Wera G D, Kraay M J, et al. Predictors of range of motion in patients undergoing manipulation after TKA. Clin Orth Relat Res 2013; 471(1): 258-63. Bellamy N, Buchanan W W, Goldsmith C H, et al. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol 1988; 15(12): 1833-40. Bhattacharyya T, Iorio R, Healy W L. Rate of and risk factors for acute inpatient mortality after orthopaedic surgery. J Bone Joint Surg Am 2002; 84-A(4): 562-72. Blom A W, Artz N, Beswick A D, et al. Improving patients’ experience and outcome of total joint replacement: the RESTORE programme. Southampton (UK): NIHR Journals Library; 2016, August. Bolognesi M P, Marchant M H Jr, Viens N A, et al. The impact of diabetes on perioperative patient outcomes after total hip and total knee arthroplasty in the United States. J Arthroplasty 2008; 23(6 Suppl 1): 92-8. Clement N D, Jenkins P J, MacDonald D, et al. Socioeconomic status affects the Oxford knee score and short-form 12 score following total knee replacement. Bone Joint J 2013; 95-B(1): 52-8. Cushnaghan J, Bennett J, Reading I, et al. Long-term outcome following total knee arthroplasty: a controlled longitudinal study. Ann Rheum Dis 2009; 68(5): 642-7. Gandhi R, Razak F, Davey J R, et al. Metabolic syndrome and the functional outcomes of hip and knee arthroplasty. J Rheumatol 2010; 37(9): 1917-22. Groll D L, To T, Bombardier C, et al. The development of a comorbidity index with physical function as the outcome. J Clin Epidemiol 2005; 58(6): 595-602. Howieson A J, Brunswicker A, Dhatariya K. A retrospective review of the assessment of current perioperative management of diabetes in patients undergoing knee replacement surgery. JRSM Open 2014; 5(2): 2042533313515864. Jamsen E, Nevalainen P, Kalliovalkama J, et al. Preoperative hyperglycemia predicts infected total knee replacement. Eur J Intern Med 2010; 21(3): 196-201. Judge A, Cooper C, Williams S, et al. Patient-reported outcomes one year after primary hip replacement in a European collaborative cohort. Arthritis Care Res (Hoboken). 2010; 62(4): 480-8. Judge A, Arden N K, Cooper C, et al. Predictors of outcomes of total knee replacement surgery. Rheumatology (Oxford). 2012; 51(10): 1804-13. Kunutsor S K, Whitehouse M R, Blom A W, et al. Patient-related risk factors for periprosthetic joint infection after total joint arthroplasty: a systematic review and meta-analysis. PLoS One 2016; 11(3): e0150866. Lenguerrand E, Whitehouse M R, Beswick A D, et al. Revision for prosthetic joint infection following hip arthroplasty: evidence from the National Joint Registry. Bone Joint Res 2017; 6(6): 391-8. Lovecchio F, Beal M, Kwasny M, et al. Do patients with insulin-dependent and noninsulin-dependent diabetes have different risks for complications after arthroplasty? Clin Orth Relat Res 2014; 472(11): 3570-5. Marchant M H Jr, Viens N A, Cook C, et al. The impact of glycemic control and diabetes mellitus on perioperative outcomes after total joint arthroplasty. J Bone Joint Surg Am 2009; 91(7): 1621-9. Robertson F, Geddes J, Ridley D, et al. Patients with Type 2 diabetes mellitus have a worse functional outcome post knee arthroplasty: a matched cohort study. Knee 2012; 19(4): 286-9. Rollins K E, Varadhan K K, Dhatariya K, et al. Systematic review of the impact of HbA1c on outcomes following surgery in patients with diabetes mellitus. Clin Nutr (Edinburgh, Scotland) 2016; 35(2): 308-16.
Acta Orthopaedica 2018; 89 (4): 399â&#x20AC;&#x201C;405
Singh J A, Lewallen D G. Diabetes: a risk factor for poor functional outcome after total knee arthroplasty. PLoS One 2013; 8(11): e78991. Shrier I, Platt R W. Reducing bias through directed acyclic graphs. BMC Med Res Methodol 2008; 8: 70. Viens NA, Hug KT, Marchant MH, et al. Role of diabetes type in perioperative outcomes after hip and knee arthroplasty in the United States. J Surg Orthop Adv 2012;21(4):253-60. White I R, Royston P, Wood A M. Multiple imputation using chained equations: issues and guidance for practice. Stat Med 2011; 30(4): 377-99. Williams A, Kind P. The present state of play about QALYs. In: (Hopkins A, ed) Measure of the quality of life: the uses to which they may be put. London: RCP Publications; 1992.
12298 Lenguerrand D.indd 405
Winemaker M, Petruccelli D, Kabali C, et al. Not all total joint replacement patients are created equal: preoperative factors and length of stay in hospital. Can J Surg 2015; 58(3): 160-6. Wylde V, Lenguerrand E, Gooberman-Hill R, et al. Effect of local anaesthetic infiltration on chronic postsurgical pain after total hip and knee replacement: the APEX randomised controlled trials. Pain 2015; 156(6): 1161-70. Zigmond A S, Snaith R P. The hospital anxiety and depression scale. Acta Psychiatr Scand 1983; 67(6): 361-70.
Acta Orthopaedica 2018; 89 (4): 406–411
Knee extensor muscle weakness and radiographic knee osteoarthritis progression The influence of sex and malalignment Andrea DELL’ISOLA 1, Wolfgang WIRTH 2, Martijn STEULTJENS 1, Felix ECKSTEIN 2, and Adam G CULVENOR 2,3
of Applied Health Research/School of Health and Life Sciences, Glasgow Caledonian University, Glasgow, Scotland; 2 Institute of Anatomy, Paracelsus Medical University Salzburg and Nuremburg, Salzburg, Austria; 3 La Trobe Sport and Exercise Medicine Research Centre, School of Allied Health, La Trobe University, Bundoora, Australia Correspondence: email@example.com Submitted 2017-12-14. Accepted 2018-03-04.
Background and purpose — Knee extensor (KE) muscle weakness is a modifiable feature commonly observed in individuals with knee osteoarthritis (KOA) and constitutes a potential target for patient-specific interventions. Therefore, in this study, we explored whether KE weakness is associated with radiographic (medial and/or lateral) KOA progression and how this relationship differs depending on frontal plane knee alignment and sex. Patients and methods — We studied 3,075 knees (1,961 participants, 58% female) from the Osteoarthritis Initiative with radiographic Kellgren–Lawrence grade 1–3. Peak KE torque (Nm/kg) was assessed at baseline, and progression defined as fixed-location joint space width loss (> 0.7mm) in medial and lateral tibiofemoral compartments from baseline to 4-year follow-up. Kneebased generalized estimating equations, stratified by alignment (malaligned vs. neutral), estimated the relative risk (RR) of progression for those in the lowest (and middle) vs. highest KE torque group (split by tertiles). Secondary analyses explored whether this relationship was compartmental- or sex-specific. Results — Being in the lowest (or middle) compared with the highest torque group increased the risk of progression in neutrally aligned knees (relative risk [RR] 1.2 [95% CI 1.0–1.4]; and 1.2 [CI 1.0–1.4], respectively), but not after adjusting for age, sex, BMI, pain, and radiographic severity. In secondary analyses, women with neutral alignment in the lowest compared with the highest torque group had significantly increased risk of lateral compartment progression independent of age, BMI, disease severity, and pain (RR 1.3 [CI 1.0–1.8]). No association was observed between KE torque and KOA progression in men, irrespective of alignment. Interpretation — These results identify a potentially important clinical phenotype: KE weakness may be a more important risk factor for radiographic KOA progression in women without knee malalignment. ■
Knee extensor (KE) muscle weakness is a modifiable feature commonly observed in individuals with knee osteoarthritis (KOA) and a risk factor for incident radiographic KOA (Øiestad et al. 2015). In terms of disease progression, the relationship between KE weakness and radiographic KOA is less clear, perhaps due to an apparent sex-specific effect whereby women (but not men) with muscle weakness have an increased risk of KOA progression (Culvenor et al. 2017b). Frontal plane knee alignment, an independent risk factor for KOA, has been reported to represent a potential confounder in the sex-specific relationship between KE weakness and KOA progression (Sharma et al. 2003). Contradictory to the concept of KE weakness increasing the risk of KOA progression, Sharma et al. (2003), in a study of 171 participants (328 knees), observed that greater KE torque increased the risk of radiographic KOA progression in individuals with established KOA and knee malalignment. These data provide preliminary evidence that the relationship between KE torque and disease progression depends on the local mechanical environment, where malalignment may determine how the medial and lateral tibiofemoral joint responds to muscle force. However, contradictory longitudinal MRI data in 265 older adults with KOA (mean age 67 years) subsequently showed that KE weakness did not influence medial and lateral compartment cartilage loss in either aligned or malaligned knees, but without accounting for previously observed differences in men and women (Amin et al. 2009). Insights into the alignment- and sex-specific impact of KE weakness on medial and lateral KOA progression may be of value for developing personalized treatment strategies. Therefore, we determined whether the relationship between lower KE torque and risk of radiographic disease progression depends on knee alignment in a large cohort of > 3,000 knees,
© 2018 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 10.1080/17453674.2018.1464314
12408 DELL’ISOLA D.indd 406
Acta Orthopaedica 2018; 89 (4): 406–411
and whether this relationship is compartment- and sex-specific. Based on previous data, we hypothesized that alignment modifies the relationship between KE torque and risk of radiographic progression and that women with KE weakness are at higher risk of KOA progression compared with men.
Methods Participants Participants were selected from the Osteoarthritis Initiative (OAI, http://www.oai.ucsf.edu), an ongoing multicenter cohort study in the United States designed to identify biomarkers and risk factors associated with KOA incidence and progression. The OAI includes 4,796 participants, aged 45–79 years, with, or at risk of, symptomatic KOA. For the current study, we included knees with radiographic Kellgren–Lawrence (KL) grade 1–3 at baseline (central reading release 0.7). Knees without any radiographic evidence of OA (KL 0) were excluded, as the current analysis focuses specifically on disease progression. Knees with end-stage disease (KL 4) were excluded due to having a limited capacity to progress. Eligible knees were those with KE strength, alignment (radiographic femur–tibia angle [FTA]) at baseline, and compartment-specific joint space width (JSW) measures recorded at baseline and 4-year follow-up (Neumann et al. 2009). From the entire 4,796 OAI participants (9,592 knees), a total of 3,178 participants (5,187 knees) had KL grade 1–3. Of these, 3,075 knees from 1,961 participants (58% female) were included in the current analysis (Table 1) (Additional File 2). Evaluation of knee extensor muscle strength Peak isometric KE strength was measured in N at baseline using the “Good Strength Chair” (Metitur Oy, Jyvaskyla, Finland) as described previously (Culvenor et al. 2016). Torque per body weight (Nm/kg) was calculated using the lever arm length recorded at the strength assessment. In the absence of previously defined thresholds, we used sex-specific tertiles of torque per body weight to create three equal-sized groups based on sex-specific KE torque (lowest, middle, highest group). Radiographic disease progression Medial and lateral radiographic JSW was measured at baseline and follow-up with customized software at fixed locations, based on a range from X = 0% to X = 100% of the distal femur mediolateral width (central release 0.6) (Ornetti et al. 2009). For the current analysis, baseline fixed-location measures in the center of the medial (X = 22.5%) and lateral compartment (X = 80%) were used, as these were shown to display high sensitivity to change in KOA (Wirth et al. 2013). Radiographic progression was defined as a JSW reduction of ≥ 0.7mm. This cut-off has been shown to provide the best predictive value for detecting progression at 3-year follow-up and was further validated using the Bland–Altman method
12408 DELL’ISOLA D.indd 407
to calculate the smallest detectable change (Bruyere et al. 2005, Ornetti et al. 2009). In these studies, a minimal JSW (mJSW) reduction of ≥ 0.7 mm in OAI control participants showed a minimal probability (≤ 10%) of change due to measurement error over a 12-month period (Ornetti et al. 2009). In the absence of a threshold for fixed location measures, we used the 0.7-mm threshold previously determined for mJSW; hence subjects with a reduction of JSW ≥ 0.7 mm in either the medial (X = 22.5%) or lateral (X = 80%) compartment were defined as progressors. Frontal plane knee alignment Frontal plane knee alignment was assessed at baseline using the FTA, as previously described (release 0.6) (IranpourBoroujeni et al. 2014) (Additional File 1), which was more frequently recorded in the OAI database than the full-limb hip–knee–ankle (HKA) angle. More negative values indicate greater varus alignment. The FTA method has high intra- and inter-reader reproducibility (ICC ≥ 0.98) (Iranpour-Boroujeni et al. 2014) and is highly correlated with the HKA angle, having similar ability in predicting 2-year tibiofemoral cartilage loss (Iranpour-Boroujeni et al. 2014). For the current study, the HKA angle was derived from the FTA using sexspecific conversion formulae (Iranpour-Boroujeni et al. 2014). Alignment categories were valgus (HKA ≥ +2°), varus (HKA ≤ –2°), and neutral (HKA < ±2° from zero). Statistics To adjust for correlations between limbs within each subject and to assess the risk of radiographic progression in subjects with lower KE torque, we used Poisson regression models with generalized estimating equations. An independent working correlation structure was used. To explore the effect of lowest (or middle) vs. highest KE torque (i.e., lowest vs. highest; middle vs. highest) on the risk of progression, we estimated the relative risk (RR) and 95% confidence intervals (CI) of KOA progression. For the primary analysis, KOA progression was defined as occurring in either the medial or lateral tibiofemoral compartment and risk of progression was stratified by knee alignment (neutral vs. malaligned [i.e., either varus or valgus]) as per Sharma et al. (2003). Analyses were adjusted for confounders, selected using clinical reasoning and literature review, and portrayed in direct acyclic graphs to minimize over-adjustment and collider stratification bias. Analyses were adjusted for age, sex, BMI, KL grade, and baseline pain using kneespecific Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain score (0–100). Older age, sex, and knee-related pain increase risk of both KE weakness and KOA progression (Hunter 2009), while BMI and baseline KL grade directly influence KOA progression as well as being on a causal pathway via muscle weakness (i.e., BMI can influence KL grade which increases pain, leading to muscle weakness). We also adjusted for alignment (continuous variable to
Acta Orthopaedica 2018; 89 (4): 406–411
Table 1. Baseline characteristics of the sample a All n = 1,961 Knees b Age, years BMI, kg/m2 Height, m Weight, kg WOMAC, 0–100 pain physical function stiffness KL grade b 1 2 3 Knee extensor strength (N) Knee extensor torque per body weight, Nm/kg Frontal plane alignment, Alignment b Neutral Varus Valgus
Women n = 1,131
Table 2. Risk of radiographic progression by sex-specific quadriceps torque group, stratified by knee alignment Men n = 830
3,075 (100) 1,825 (59.3) 1,250 (40.7) 62.2 (8.8) 62.4 (8.7) 61.9 (9) 29.4 (4.7) 29.5 (5.1) 29.4 (4) 1.68 (0.09) 1.63 (0.06) 1.76 (0.06) 83.7 (16) 78.0 (14.5) 92.0 (14.3) 12.9 (16.9) 13.1 (16.0) 20.4 (20.3)
14.2 (17.9) 16.0 (17.1) 22.1 (21.1)
616 (20) 367 (20) 1,650 (54) 1,029 (56) 809 (26) 429 (24) 339 (129) 278 (93)
11.0 (15.0) 10.6 (13.9) 17.7 (18.7) 249 (20) 621 (50) 380 (30) 428 (24)
1.2 (0.4) –1.1 (2.5)
1.1 (0.4) –0.2 (2.3)
1.5 (0.5) –2.5 (2.2)
1,609 (52) 1,159 (38) 307 (10)
1,183 (65) 374 (20) 268 (15)
426 (34) 785 (63) 39 (3)
BMI, body mass index; KL, Kellgren–Lawrence; WOMAC, Western Ontario and McMaster Universities Osteoarthritis Index. a Except where indicated otherwise, values are mean (SD). b Values are n (% of category)
account for differences within each alignment category) as per Amin et al. (2009). For secondary analyses, we calculated the compartmentspecific risk of progression stratified by sex and knee alignment categories (neutral, varus, valgus). Knees without JSW reduction in either compartment were considered as non-progressors and used as controls. Analyses were adjusted as per primary analysis (excluding sex) and performed using SPSS (v22.0; IBM Corp, Armonk, NY, USA). p < 0.05 was considered statistically significant. Ethics, funding, and potential conflict of interest Ethical approval and informed consent were obtained as part of the original OAI participant recruitment and data collection process. No specific ethical approval was therefore required for the current study. The datasets analyzed during the current study are available in the Osteoarthritis Initiative (OAI) repository (https://oai. epi-ucsf.org/datarelease/). The OAI is a public–private partnership comprising 5 contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institutes of Health. Funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation; GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health. This research has received funding from the European Union Seventh Framework Programme (FP7-PEO-
12408 DELL’ISOLA D.indd 408
Sex-specific Progressors torque group n (%) Crude RR (CI) Neutral alignment Low Middle High Malalignment b Low Middle High
Adjusted a RR (CI)
226 (33) 241 (36) 209 (31)
1.16 (1.01–1.36) 1.18 (1.02–1.37) 1.00
1.03 (0.78–1.40) 1.12 (0.96–1.30) 1.00
253 (33) 256 (34) 246 (33)
1.13 (0.99–1.28) 1.05 (0.93–1.20) 1.00
1.00 (0.87–1.14) 0.99 (0.86–1.12) 1.00
RR: relative risk; CI: 95% confidence interval. for baseline Kellgren–Lawrence grade, baseline WOMAC pain, age, body mass index, sex, and alignment. b Malalignment defined as 2° varus or valgus malalignment. a Adjusted
PLE-2013-ITN; KNEEMO) under grant agreement number 607510. AGC is a recipient of a National Health and Medical Research Council (NHMRC) of Australia Early Career Fellowship (Neil Hamilton Fairley Clinical Fellowship No. 1121173). The sponsors were not involved in the design and conduct of this particular study, in the analysis and interpretation of the data, and in the preparation, review, or approval of the manuscript. Wolfgang Wirth has part-time employment with Chondrometrics GmbH and is a co-owner of Chondrometrics GmbH, a company providing MRI analysis services to academic researchers and to industry. Felix Eckstein is CEO of Chondrometrics GmbH; he has provided consulting services to Merck Serono, Samumed, and Bioclinica/Synarc, has prepared educational sessions for Medtronic, and has received research support from Pfizer, Eli Lilly, Merck Serono, Novartis, Stryker, Abbvie, Kolon, Synarc, Ampio, BICL, Orthotrophix, and Tissue Gene.
Results Women had 35% lower absolute muscle force (N) and 27% lower torque per body weight (Nm/kg) than men (Table 1). Approximately half of all knees had neutral alignment (52%), while 38% and 10% had varus and valgus malalignment, respectively (Table 1). In the primary analysis, 1,431 knees (47%) displayed radiographic progression. 670 knees (22%) and 521 knees (17%) displayed medial and lateral progression, respectively, while 240 knees (7.8%) displayed progression in both compartments, and 1,644 (54%) no progression in either compartment. Being in the lowest or middle KE torque group increased the risk of KOA progression in either compartment for neutrally aligned knees in the unadjusted analysis (lowest KE torque group RR: 1.2 [1.0–1.4]; middle KE torque group RR: 1.2 [1.0–1.4]) but not after adjustment (Table 2). In malaligned knees, KE torque
Acta Orthopaedica 2018; 89 (4): 406–411
Table 3. Risk of radiographic progression by sex-specific extensor torque, stratified by sex and knee alignment
Progressors (%) a
Women Neutral alignment Low Middle High Valgus alignment c Low Middle High Varus alignment c Low Middle High Men Neutral alignment Low Middle High Valgus alignment c Low Middle High Varus alignment c Low Middle High
Medial compartment Crude RR Adjusted b RR
Lateral compartment Progressors (%) a Crude RR Adjusted b RR
92 (33) 102 (36) 86 (31)
1.23 (0.93–1.62) 1.23 (0.94–1.60) 1.00
0.97 (0.71–1.32) 1.09 (0.83–1.44) 1.00
92 (37) 87 (34) 73 (29)
1.45 (1.08–1.94) 1.23 (0.92–1.65) 1.00
1.34 (1.04–1.84) 1.18 (0.87–1.60) 1.00
17 (32) 21 (40) 15 (28)
0.89 (0.46–1.72) 1.12 (0.62–2.01) 1.00
0.69 (0.33–1.44) 0.99 (0.55–1.79) 1.00
48 (37) 43 (33) 38 (29)
0.99 (0.73–1.35) 0.90 (0.67–1.24) 1.00
0.85 (0.61, 1.17) 0.87 (0.63–1.20) 1.00
50 (36) 44 (32) 44 (32)
1.31 (0.94–1.84) 1.18 (0.83–1.67) 1.00
1.18 (0.83–1.68) 1.18 (0.84–1.66) 1.00
20 (29) 28 (40) 21 (30)
1.10 (0.59–2.07) 1.57 (0.93–2.65) 1.00
0.92 (0.46–1.84) 1.45 (0.84–2.52) 1.00
44 (35) 42 (34) 39 (31)
1.02 (0.70–1.49) 1.08 (0.75–1.54) 1.00
0.97 (0.63–1.49) 1.08 (0.75–1.55) 1.00
48 (36) 45 (34) 39 (30)
1.11 (0.76–1.62) 1.15 (0.79–1.68) 1.00
1.07 (0.71–1.62) 1.12 (0.77–1.64) 1.00
NA NA NA
NA NA NA
NA NA NA
NA NA NA
1.26 (0.96–1.58) 1.06 (0.84–1.34) 1.00
1.04 (0.82–1.32) 0.92 (0.72–1.16) 1.00
0.90 (0.62–1.31) 0.97 (0.68–1.36) 1.00
0.88 (0.60–1.30) 0.94 (0.65–1.34) 1.00
1 (12) 6 (76) 1 (12) 105 (34) 100 (33) 101 (33)
5 (33) 7 (46) 3 (20) 46 (28) 56 (34) 62 (38)
RR: relative risk; CI: 95% confidence interval. a Percentage indicates progressors in the alignment category. b Adjusted for baseline Kellgren–Lawrence grade, baseline WOMAC pain, age, c Varus and valgus alignment defined as a variation 2° in either direction.
was not significantly associated with radiographic progression (Table 2). In secondary analyses, female knees without malalignment with the lowest KE torque had increased risk of radiographic progression in the lateral compartment compared with knees with the highest KE torque, both before and after adjustment (RRadj: 1.4 [1.0–1.8]). In contrast, KE torque did not significantly increase the risk of radiographic medial or lateral compartment progression in female knees with valgus or varus malalignment. Knee extensor torque also did not show any significant effect on the likelihood of KOA progression in men, regardless of the compartment and alignment (Table 3). However, due to the very low number of men with valgus malalignment (n = 39), it was not possible to perform analyses in male knees with valgus.
Discussion This is the first study to investigate the impact of KE weakness on the compartment- and sex-specific risk of radiographic KOA progression in the context of variation in frontal plane knee alignment. Our results from > 3,000 knees suggest that lower KE torque is generally not associated with sub-
12408 DELL’ISOLA D.indd 409
body mass index, and alignment.
sequent tibiofemoral radiographic progression, particularly when accounting for age, sex, BMI, radiographic severity, and pain. However, compartment- and sex-specific analyses revealed that lower KE torque was associated with lateral compartment progression in women with neutrally aligned knees (before and after adjustment). These results identify a potentially important clinical phenotype and highlight that KE muscle weakness may be an important risk factor for disease progression specifically in women without knee malalignment. The findings of our primary analysis in neutral and malaligned knees do not confirm previous data from a smaller number of knees (n = 228) in the Mechanical Factors in Arthritis of the Knee (MAK) cohort that found KE weakness had a significant protective effect on KOA progression in malaligned knees (Sharma et al. 2003). In contrast, our results showed effect sizes of similar magnitude between neutral and malaligned knees, and that higher KE strength tended to be more (but not statistically significant) protective of KOA progression, as recently observed in a systematic review (Culvenor et al. 2017b). Our results extend this recent systematic review and meta-analysis that did not observe a significant association between KE weakness and tibiofemoral structural deterioration (Culvenor et al. 2017b) by accounting for varia-
tions in knee alignment and evaluating the sex-specific effect on medial and lateral compartment progression. In stratifying by knee alignment categories in men and women, we observed that KE weakness increased the risk of (lateral) KOA progression in women with neutral knee alignment (before and after adjustment). That women, but not men, appear to be at increased risk of lateral tibiofemoral joint progression when KE weakness is present may be explained by women having a lower absolute strength capacity, and thereby potentially being closer to a threshold below which the risk of OA progression increases (Culvenor et al. 2017b). Moreover, KOA has a higher prevalence in women where biochemical differences (i.e., hormonal) are thought to play a role in the development and progression of disease, and the ability of KE muscle fibers to generate force diminishes with greater BMI in women, but not in men (Culvenor et al. 2017a). Knee malalignment represents a well-established risk factor for KOA progression (Felson and Kim 2007), and, as hypothesized, the local mechanical environment appears to influence the relationship between KE torque and KOA progression demonstrated in previous work (Culvenor et al. 2016, Øiestad et al. 2015). It appears that once malalignment is present, the lack of KE torque has little effect on the risk of KOA progression. This is supported by the observation that 52% of all knees with malalignment displayed radiographic progression whereas only 42% did in the neutral group. In neutral aligned knees, in contrast, muscle weakness appears to play a role in subsequent progression, particularly in women, and therefore optimizing muscle impairments provides a potential avenue to help modify the risk of progression in women with neutral knee alignment. Our findings somewhat contrast those of Amin et al. (2009) who observed no influence of sex-specific tertiles of KE torque on MRI-assessed medial or lateral cartilage lesion progression in knees without varus malalignment (i.e., < 5° varus). The inclusion of valgus knees in the “neutral aligned” group, the lack of stratification by sex, the different definition of progression (semi-quantitative MRI-based cartilage lesion scores), and the shorter observation period (15 to 30 months’ follow-up) may explain these differences and the lack of sensitivity in this previous study (Amin et al. 2009). Our findings of lateral compartment radiographic progression in female neutrally aligned knees may be the result of the differences in gait kinematics between sexes. Women with normal alignment have a larger knee abduction angle, hip adduction, and internal rotation during gait compared with men with normal alignment (Phinyomark et al. 2016). This specific kinematic pattern is thought to move the internal knee load towards the lateral compartment and has previously been associated with lateral KOA (Weidow et al. 2006). Women with muscle weakness therefore potentially have a higher risk of KOA progression in the lateral compartment due to the absence of muscle stability driving abnormal biomechanical load. Studies estimating knee internal contact forces are necessary in order to confirm this hypothesis.
12408 DELL’ISOLA D.indd 410
Acta Orthopaedica 2018; 89 (4): 406–411
Limitations of our study include the estimation of mechanical alignment from the FTA. However, a strong correlation between FTA and HKA suggests that similar results would occur irrespective of alignment assessment approach (Iranpour-Boroujeni et al. 2014). Despite the large study sample of OAI participants, we were unable to complete evaluations of the influence of KE weakness in men with valgus alignment due to the small number of men with knee valgus. The relatively small number of progressors in some other strata limits generalizability and means interpretation should be made with some caution. Second, to determine disease progression in both medial and lateral compartment we used the JSW threshold of 0.7 mm. This cut-off has been determined from test– retest measurements of medial compartment mJSW (Ornetti et al. 2009). However, since the mJSW can only be measured in the medial (and not the lateral) compartment, and because no cut-off has been established for fixed-location JSW measures, we decided to apply a cut-off of 0.7mm for fixed-location measures in the current study. Overall, generalizability of the results may be limited by cut-offs used to determine strata (i.e., muscle torque tertiles, alignment, and radiographic disease progression). In addition, considering the slow rate of progression that characterizes KOA, a longer follow-up period (> 4 years) should be considered to improve the estimates of the risk of OA radiographic progression associated with KE deficit. Finally, it is important to acknowledge that, in our secondary analysis, we did not account for multiple tests, but the analyses stratified by sex and compartment support the trends seen in the primary analyses. Targeting the right patient with the right treatment constitutes a priority in KOA care (Dell’Isola et al. 2016). Optimizing muscle impairments could provide a potential avenue to help modify the risk of progression in women with neutral knee alignment. Exercise therapy, including muscle strengthening and neuromuscular exercises, is the first-line treatment for patients with KOA (Fernandes et al. 2013, National Institute of Clinical Excellence 2014). International guidelines indicate that these interventions need to be specifically tailored to the individual (Nelson et al. 2014). Results of our study suggest that including exercises aimed to increase KE torque may be particularly beneficial for structural outcomes in females with neutral alignment, in addition to optimizing functional capacity and reducing symptoms in all patients with KOA (Lange et al. 2008). In summary, in the tibiofemoral joint of men and women as a whole, lower KE torque is generally not associated with KOA progression over 4 years, particularly after adjustment for other risk factors. However, in unravelling this relationship further, this study has identified an important subset of women (without malalignment), in which KE weakness was associated with (lateral) tibiofemoral progression. No relationship between KE weakness and compartment-specific progression was observed in men. Optimizing muscle impairments may help modify the risk of progression in women with neutral knee alignment.
Acta Orthopaedica 2018; 89 (4): 406–411
Supplementary data Additional Files 1 and 2 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/ 17453674.2018.1464314
All persons designated as authors qualify for authorship. Each author participated in the work and made substantial contributions to the manuscript.
The OAI is a public–private partnership comprising five contracts (N01-AR2-2258, N01-AR-2-2259, N01-AR-2-2260, N01-AR-2-2261, and N01-AR-22262) funded by the National Institutes of health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. This article was prepared using an OAI public use dataset and does not necessarily reflect the opinions or views of the OAI investigators or the NIH. The authors would like to kindly acknowledge “KNEEMO—Prevention and personalized treatments in knee osteoarthritis: an Initial Training Network” for the training.
Acta thanks Eva W Broström and other anonymous reviewers for help with peer review of this study.
Amin S, Baker K, Niu J, Clancy M, Goggins J, Guermazi A, Grigoryan M, Hunter D J, Felson D T. Quadriceps strength and the risk of cartilage loss and symptom progression in knee osteoarthritis. Arthritis Rheum 2009; 60(1): 189-98. Bruyere O, Richy F, Reginster J-Y. Three year joint space narrowing predicts long term incidence of knee surgery in patients with osteoarthritis: an eight year prospective follow up study. Ann Rheum Dis 2005; 64(12): 1727-30. Culvenor A, Wirth W, Ruhdorfer A, Eckstein F. Thigh muscle strength predicts knee replacement risk independent of radiographic disease and pain in women: data from the Osteoarthritis Initiative. Arthritis Rheumatol 2016; 68(5): 1145-55. Culvenor A G, Felson D T, Niu J, Wirth W, Sattler M, Dannhauer T, Eckstein F. Thigh muscle specific strength and the risk of incident knee osteoarthritis: the influence of sex and greater body mass index. Arthritis Care Res (Hoboken). 2017a; 69(8): 1266-70. Culvenor A G, Ruhdorfer A, Juhl C, Eckstein F, Øiestad B E. Knee extensor strength and risk of structural, symptomatic, and functional decline in knee osteoarthritis: a systematic review and meta-analysis. Arthritis Care Res (Hoboken). 2017b; 69(5): 649-58. Dell’Isola A, Allan R, Smith S L, Marreiros S S P, Steultjens M. Identification of clinical phenotypes in knee osteoarthritis: a systematic review of the literature. BMC Musculoskelet Disord 2016;17(1): 425.
12408 DELL’ISOLA D.indd 411
Felson D T, Kim Y-J. The futility of current approaches to chondroprotection. Arthritis Rheum 2007; 56(5): 1378-83. Fernandes L, Hagen K B, Bijlsma J W J, Andreassen O, Christensen P, Conaghan P G, Doherty M, Geenen R, Hammond A, Kjeken I, Lohmander L S, Lund H, Mallen C D, Nava T, Oliver S, Pavelka K, Pitsillidou I, da Silva J A, de la Torre J, Zanoli G, Vliet Vlieland T P M, European League Against Rheumatism (EULAR). EULAR recommendations for the non-pharmacological core management of hip and knee osteoarthritis. Ann Rheum Dis 2013; 72(7): 1125-35. Hunter D J. Risk stratification for knee osteoarthritis progression: a narrative review. Osteoarthritis Cartilage 2009; 17(11): 1402-7. Iranpour-Boroujeni T, Li J, Lynch J A, Nevitt M, Duryea J. A new method to measure anatomic knee alignment for large studies of OA: data from the Osteoarthritis Initiative. Osteoarthritis Cartilage 2014; 22(10): 1668-74. Lange A K, Vanwanseele B, Fiatarone singh M A. Strength training for treatment of osteoarthritis of the knee: a systematic review. Arthritis Rheum 2008; 59(10): 1488-94. National Institute of Clinical Excellence. Osteoarthritis: care and management. Clinical guideline; 2014. Nelson A E, Allen K D, Golightly Y M, Goode A P, Jordan J M. A systematic review of recommendations and guidelines for the management of osteoarthritis: the chronic osteoarthritis management initiative of the U.S. bone and joint initiative. Semin Arthritis Rheum 2014; 43(6): 701-12. Neumann G, Hunter D, Nevitt M, Chibnik L B, Kwoh K, Chen H, Harris T, Satterfield S, Duryea J, Health ABC Study. Location specific radiographic joint space width for osteoarthritis progression. Osteoarthritis Cartilage 2009; 17(6): 761-5. Øiestad B E, Juhl C B, Eitzen I, Thorlund J B. Knee extensor muscle weakness is a risk factor for development of knee osteoarthritis: a systematic review and meta-analysis. Osteoarthritis Cartilage 2015; 23 (2): 171-7. Ornetti P, Brandt K, Hellio-Le Graverand M-P, Hochberg M, Hunter D J, Kloppenburg M, Lane N, Maillefert J-F, Mazzuca S A, Spector T, UtardWlerick G, Vignon E, Dougados M. OARSI-OMERACT definition of relevant radiological progression in hip/knee osteoarthritis. Osteoarthritis Cartilage 2009; 17(7): 856-63. Phinyomark A, Osis S T, Hettinga B A, Kobsar D, Ferber R. Gender differences in gait kinematics for patients with knee osteoarthritis. BMC Musculoskelet Disord 2016; 17(1): 157. Sharma L, Dunlop D D, Cahue S, Song J, Hayes K W. Quadriceps strength and osteoarthritis progression in malaligned and lax knees. Ann Intern Med 2003; 138(8): 613-19. Weidow J, Tranberg R, Saari T, Kärrholm J. Hip and knee joint rotations differ between patients with medial and lateral knee osteoarthritis: gait analysis of 30 patients and 15 controls. J Orthop Res 2006; 24(9): 1890-9. Wirth W, Duryea J, Hellio Le Graverand M-P, John M R, Nevitt M, Buck R J, Eckstein F, OA Initiative Investigators Group. Direct comparison of fixed flexion, radiography and MRI in knee osteoarthritis: responsiveness data from the Osteoarthritis Initiative. Osteoarthritis Cartilage 2013; 21(1): 117-25.
Acta Orthopaedica 2018; 89 (4): 412–417
Migration of all-polyethylene compared with metal-backed tibial components in cemented total knee arthroplasty A randomized controlled trial Koen T VAN HAMERSVELD 1, Perla J MARANG-VAN DE MHEEN 2, Rob G H H NELISSEN 1, and Sören TOKSVIG-LARSEN 3
1 Department 3 Department
of Orthopaedics, Leiden University Medical Center, Leiden; 2 Medical Decision Making, Leiden University Medical Center, Leiden; of Orthopaedics, Hässleholm Hospital, Hässleholm, Sweden and Department of Clinical Sciences, Lund University, Lund, Sweden Correspondence: firstname.lastname@example.org Submitted 2017-11-27. Accepted 2018-03-16.
Background and purpose — With a rapidly increasing population in need of total knee arthroplasty (TKA), there is renewed interest in cost-saving all-polyethylene designs. Differences between metal-backed and all-polyethylene designs in initial component migration assessed by radiostereometric analysis (RSA), a proven predictor for late aseptic loosening, have been scantily reported. The purpose of this study was to compare implant migration and clinical outcomes of all-polyethylene tibial components versus metal-backed trays of similar geometrical shape. Patients and methods — In this randomized controlled trial, 59 patients received a cemented Triathlon condylar-stabilizing implant (Stryker, Mahwah, NJ, USA) with either an all-polyethylene (n = 29) or a metal-backed tibial component (n = 30). RSA measurements and clinical scores (the Knee Society Score, Forgotten Joint Score, and Knee Osteoarthritis and Injury Outcome Score) were evaluated at baseline and postoperatively at 3, 12, and 24 months. A linear mixed-effects model was used to analyze the repeated measurements. Results — A statistically significant difference in mean migration after 2 years was found in favor of the all-polyethylene group, with a mean maximum total point motion of 0.61 mm (95% CI 0.49–0.74) versus 0.81 mm (95% CI 0.68–0.96) for the cemented group (p = 0.03). However, this difference was smaller and not statistically significant after post hoc adjustment for surgeon effect. Both groups showed comparable improvements on all clinical outcome scores over time. Interpretation — The Triathlon all-polyethylene tibial component showed less migration, suggesting a lower risk of late loosening as compared with its metal-backed counterpart. However, the found surgeon effect warrants further investigation. ■
Metal-backed tibial components in total knee arthroplasty (TKA) have primarily been used since their introduction in the late 1970s, as clinical results were superior to the first generation of all-polyethylene tibial components (Gioe and Maheshwari 2010). With a rapidly increasing population in need of knee arthroplasty, the associated healthcare costs are expected to rise exponentially (Kurtz et al. 2007). This triggered renewed interest in all-polyethylene designs as manufacturing such implants costs 20% to 50% less (Gioe and Maheshwari 2010). Meta-analyses comparing modern all-polyethylene and metal-backed tibial components show equivalent results in terms of risk for revision and clinical scores, yet all-polyethylene designs are still rarely used (Voigt and Mosier 2011, Nouta et al. 2012, Voss et al. 2016). Given that first-generation all-polyethylene designs often failed secondary to aseptic loosening, many surgeons today are reluctant to use all-polyethylene components (Voss et al. 2016). More evidence is thus needed on the fixation of today’s all-polyethylene designs, preferably by radiostereometric analysis (RSA). None of the few RSA studies published to date has shown superiority of metal-backed designs over allpolyethylene designs (Adalberth et al. 1999, 2000, 2001, Norgren et al. 2004, Hyldahl et al. 2005a, b, Muller et al. 2006). Moreover, Hyldahl et al. (2005a) found lower initial migration in AGC all-polyethylene components (Biomet, Warsaw, IN, USA). They hypothesized that these—to some degree elastic—components may partly absorb eccentric forces, while the more rigid metal-backed design is thought to transform asymmetric load throughout the entire component, inducing adverse tensile forces. With further improvements in implant design and quality of materials over the past decades, the clinical performance of
© 2018 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 10.1080/17453674.2018.1464317
12364 HAMERSVELD D.indd 412
Acta Orthopaedica 2018; 89 (4): 412–417
either design could nowadays well outperform the other. We therefore conducted a randomized controlled trial in which we compared implant migration and clinical performance of a relatively new all-polyethylene tibial component with a similarly designed metal-backed tray of the Triathlon total knee prosthesis (Stryker, Mahwah, NJ, USA). The femoral component of this prosthesis is designed to rotate about a single axis during flexion, which should provide ligament isometry and a larger contact area throughout the range of motion (Wolterbeek et al. 2012). Any remaining peripheral peak stresses that could compromise implant fixation might be better absorbed by the more elastic all-polyethylene design. Based on this theory, we hypothesized the all-polyethylene design to show less implant migration as compared with its metal-backed counterpart.
Patients and methods This randomized controlled trial was conducted in Hässleholm Hospital, Sweden. All consecutive patients with primary osteoarthritis scheduled to undergo TKA between June 2014 and November 2014 were asked to participate. The main exclusion criterion was when regular postoperative visits for RSA and clinical evaluations were considered impractical, due to, for example, long travel time. A computer-generated randomization list was created by the study monitor (1:1 ratio with a block size of 20). Opening the sequentially numbered, opaque, sealed envelopes only on the day of surgery ensured concealment of treatment allocation. Patients remained blinded throughout follow-up, which was not the case for surgeons and observers performing clinical follow-up due to the marked difference in radiographic appearance between implant designs. Prosthesis and surgical technique Surgeries were performed by 2 experienced surgeons using standardized techniques according to the Triathlon knee system surgical protocol. All patients received condylar-stabilizing (i.e., with a deep-dished polyethylene insert) cruciate-retaining Triathlon total knee prostheses indicated for cemented fixation, with either modular metal-backed tibial components using highly cross-linked polyethylene inserts or monoblock all-polyethylene tibial components of similar geometrical shape made from conventional N2/Vac ultra-high molecular weight polyethylene. Both surgeons used a standard midline incision and medial parapatellar arthrotomy, preserved the posterior cruciate ligament and used pulsatile lavage prior to applying SmartSet GHV bone cement (DePuy CMW, Blackpool, UK) with the tibial keel uncemented in all procedures. No tourniquet was used and patellae were not resurfaced. For RSA purposes, 8 tantalum markers were inserted into the proximal tibial metaphysis and 5 markers were inserted (proximally) in the polyethylene insert at standardized posi-
12364 HAMERSVELD D.indd 413
tions (0.8 mm diameter; RSA Biomedical, Umeå, Sweden). Postoperatively, low molecular heparin (enoxaparin intramuscular 40 mg/day) was prescribed for 10 days and patients were stimulated to mobilize with immediate full weight-bearing. Follow-up Preoperatively, the Knee Society Score (KSS), Knee injury and Osteoarthritis Outcome Score (KOOS), and hip–knee– ankle angle (HKA) measurements (with varus < 180°) were assessed. Postoperative evaluations including RSA radiographs were performed on the first day after surgery. Subsequent RSA and clinical examinations including the KSS, KOOS, and the Forgotten Joint Score (FJS) were scheduled at 3 months, 1 year, and 2 years after surgery. The FJS questionnaire is a relatively new outcome measurement with increased discriminatory power in especially well-performing patients (i.e., able to detect small differences between good, very good, and excellent patients) (Behrend et al. 2012, Thomsen et al. 2016). HKA measurements were repeated at 3 months’ follow-up. Radiostereometric analysis To ensure similar measurement techniques between the radiolucent all-polyethylene design and the metal-backed design, marker-based RSA analysis was performed using the tantalum markers inserted at standardized positions in both designs. RSA radiographs were made in supine position with the knee in a calibration cage (Cage 10, RSA Biomedical, Umeå, Sweden) and analyzed using MB-RSA software version 4 (RSAcore, LUMC, Leiden, the Netherlands). The precision of the RSA setup was determined by taking “double examinations” at the 1-year follow-up and, as no actual migration is expected within the few minutes of time between examinations, is expressed as the upper limits of the 95% confidence interval (CI) around zero motion (ISO 16087:2013(E) 2013). Levene’s test for equality of variances was applied to test for differences in precision between modular (metal-backed) and monoblock (all-polyethylene) components. Positive directions along and about the orthogonal axes are according to the right-hand screw rule (Valstar et al. 2005). Migration was described as translation of the geometric center of the prosthesis markers and rotation about the geometric center of gravity. The maximum total point motion (MTPM), which is the length of the translation vector of the marker or virtual marker in a rigid body that has the greatest migration, was used as the primary outcome measure (ISO 16087:2013(E) 2013). The direct postoperative RSA examination served as the reference for the migration measurements. Besides migration on a group level, the number of individual components showing “continuous migration,” defined by Ryd et al. (1995) as an increase in MTPM of 0.2 mm or more in the second postoperative year indicating an increased risk for aseptic loosening, are also reported. Marker stability and scatter values were within the limits of RSA guidelines (Valstar et al. 2005).
Acta Orthopaedica 2018; 89 (4): 412–417
Randomized consecutive eligible patients (n = 60) Excluded (n = 1) withdrew preoperatively after randomization
Enrollment Allocation Allocated to all-polyethylene TKA (n = 29): – received all-polyethylene TKA, 29
Allocated to metal-backed TKA (n = 30): – received metal-backed TKA, 30
Follow-up Lost to follow-up (n = 0): – revised, 0 – died, 0
Lost to follow-up (n = 0): – revised, 0 – died, 0 Analysis
RSA radiographs analyzed: – postoperatively, 29 – at 3 months, 28 a – at 1 year, 28 a – at 2 years, 29
RSA radiographs analyzed: – postoperatively, 30 – at 3 months, 29 a – at 1 year, 30 – at 2 years, 28 b
Table 1. Baseline demographic characteristics
Outcome Age, mean (SD) BMI, mean (SD) Female sex, n Ahlbäck’s classification, n II III IV HKA preoperative, n Varus (< 177°) Neutral (177–183°) Valgus (> 183°) HKA postoperative, n Varus (< 177°) Neutral (177–183°) Valgus (> 183°) Surgeon 1, n performed Surgeon 2, n performed
All-polyethylene (n = 29)
Metal-backed (n = 30)
69 (5.5) 28 (4.2) 22
68 (5.6) 29 (3.0) 13
6 21 2
10 19 1
22 5 2
25 3 2
4 19 6 20 9
7 22 1 14 16
HKA: hip–knee–ankle angle.
Figure 1. CONSORT flow diagram. TKA = total knee arthroplasty. a Missed follow-up; b Technical reasons, clinical follow-up only.
Sample size Earlier RSA studies using the Triathlon total knee prosthesis have shown measurement errors of less than 0.25 mm (Molt et al. 2016). With an alpha of 0.05 and power of 80%, 17 patients were needed to detect a mean difference larger than 0.25 mm. To account for loss to follow-up, 30 patients were randomized to each group. Statistics All outcome measurements were analyzed according to the intention-to-treat principle using a linear mixed-effects model. This method accounts for the correlation of the repeated measurements in patients and deals effectively with missing values (Ranstam et al. 2012). Treatment, time, and the interaction of time with treatment were modeled as fixed factors, patients were included as a random factor and a compound symmetry covariance structure was assumed. MTPM was log-transformed during statistical modeling to obtain a normal distribution, computed as log10(MTPM+1). Additionally, we conducted a post hoc sensitivity analysis to determine the effect of possible confounders on treatment by adding any baseline characteristic that was by chance not evenly distributed between groups as variables to the model, as well as their interaction with time. To analyze differences in mean migration along and about each orthogonal axis, only absolute values were used (as calculating the resultant of positive and negative displacement vectors requires all vectors to act on the same prosthesis) (Derbyshire et al. 2009). These outcome parameters were also log-transformed in a similar manner to MTPM to obtain normal distribution. Significance was set at p < 0.05 (IBM SPSS Statistics 23.0; IBM Corp, Armonk, NY, USA).
12364 HAMERSVELD D.indd 414
Ethics, registration, funding, and potential conflicts of interest The trial was performed in compliance with the Declaration of Helsinki and Good Clinical Practice guidelines. The study was approved by the Regional Ethical Review Board in Lund prior to enrollment (entry no. 2013/434) and registered at isrctn. com (ID: ISRCTN04081530). Informed consent was obtained from all patients. Reporting of the trial was in accordance with the CONSORT statement. Stryker provided funds in support of the costs associated with RSA radiographs and extra clinical follow-up examinations. The sponsor did not take any part in the design, conduct, analysis, and interpretations stated in the final manuscript.
Results 60 patients were randomized of whom 1 patient withdrew from the study prior to surgery. This patient was not replaced, resulting in 29 patients receiving the allocated all-polyethylene components and 30 patients receiving allocated metalbacked components (Figure 1). At 2-year follow-up, the RSA images of 2 patients with metal-backed components could not be analyzed for technical reasons (1 stereo image had too few reference cage markers and 1 stereo image did not match). Both patients had low migration up to 1 year (MTPM < 0.3 mm) and at 2-year follow-up no signs of loosening on conventional radiographs and good clinical scores. Due to chance, more females were randomized to the all-polyethylene group and surgeries were not evenly distributed between the two surgeons (Table 1). Other than that, groups were comparable at baseline.
Acta Orthopaedica 2018; 89 (4): 412–417
Table 2. RSA migration analysis of mean maximum total point motion (logMTPM values are back-transformed in original scale in millimeters), as provided by the mixed-effects model
MTPM (mm) 1.0 Metal-backed All-polyethylene
All-polyethylene mean (95% CI)
Metal-backed mean (95% CI)
3 months 1 year 2 years
0.47 (0.36–0.59) 0.57 (0.46–0.69) 0.61 (0.49–0.74)
0.48 (0.38–0.60) 0.69 (0.57–0.82) 0.81 (0.68–0.96)
Radiostereometric analysis The precision of the RSA setup was determined by making double examinations in 48 patients (of which 22 patients had metal-backed components) at one-year follow-up. The precision (expressed as the CI around zero motion) of transverse, longitudinal, and sagittal axis translation was 0.09 mm, 0.13 mm, and 0.11 mm, respectively; and of transverse, longitudinal, and sagittal rotation 0.15°, 0.12°, and 0.11°, respectively. There were no differences in precision between groups (p > 0.15 for all translations and rotations). The results of the primary outcome MTPM showed a higher mean MTPM of 0.81 mm (CI 0.68–0.96) for the metal-backed group versus 0.61 mm (CI 0.49–0.74) for the all-polyethylene group after 2 years’ follow-up (p = 0.03, Table 2). In both groups, 4 prostheses showed continuous migration in the second postoperative year, ranging from 0.2 mm up to 1.5 mm (Figure 2). Most components showing continuous migration still had MTPM values < 1.5 mm at 2-year follow-up (Figure 2). The other RSA parameters revealed similar translations and rotations between groups at 2-year follow-up except for sagittal translation; the mean translation in the all-polyethylene group was 0.25 mm (CI 0.17–0.34) versus 0.43 mm (CI 0.34–0.52) for the metal-backed group (p = 0.006) (Table 3, see Supplementary data). In the post hoc sensitivity analysis (adjusting for a possible effect of the unevenly distributed covariates sex and surgeon), a statistically significant surgeon effect was found on migration; the mean logMTPM difference between surgeons at 2-year follow-up was 0.13 (CI 0.09–0.17, p < 0.001); sex had no statistically significant effect on migration (Table 4, see Supplementary data). Although all-polyethylene components showed on average less migration in both surgeon groups, the difference with metal-backed components was, in contrast with the primary analysis, not statistically significant anymore when adjusting for the surgeon effect (p = 0.2) (Figure 3 and Table 4, see Supplementary data). Clinical results and adverse events The KSS score and all patient-reported outcome scores (KOOS and FJS) showed comparable improvements over time between groups (Table 5, see Supplementary data). Several adverse events occurred (all in patients of the metalbacked group, except for the last patient described below).
12364 HAMERSVELD D.indd 415
Years after operation MTPM (mm) Metal-backed All-polyethylene ‘Continuous’ metal-backed ‘Continuous’ all-polyethylene
Years after operation Figure 2. RSA analysis results of maximum total point motion (MTPM). Top: mean and 95% confidence interval for the groups; bottom: mean and 95% confidence interval for the same groups excluding 8 individual components showing continuous migration of > 0.2 mm in the second postoperative year. These individual components are illustrated as 4 dashed blue lines (metal-backed) and 4 dashed red lines (all-polyethylene).
1 patient suffered from peroneal nerve dysfunction directly postoperatively, which partially resolved. 2 venous thromboembolisms occurred within 3 months (1 deep-vein thrombosis and 1 pulmonary embolism) requiring temporary pharmacologic treatment. 1 patient experienced persistent anterior knee pain with patellar maltracking for which a medial patellofemoral ligament reconstruction was performed 14 months after the primary surgery (all components remained in situ). The patient continued to participate in the study showing moderate clinical scores at 2-year follow-up. Lastly, 1 patient (a 67-yearold female with an all-polyethylene component) sustained a supracondylar femur fracture of the ipsilateral leg following a fall accident 15 months after the primary surgery. She was initially treated using a lateral distal femoral locking plate, but this was converted to an intramedullary nail due to plate failure after 2 months. At 2 years’ follow-up, the patient and her knee functioned well with excellent clinical scores, no signs of loosening of the femoral component and a stable tibial component migration pattern similar to the group average.
Discussion The results of the primary outcome of this study confirm our hypothesis that all-polyethylene components show statistically significantly lower migration after 2 years of follow-up compared with metal-backed trays of similar geometrical shape. However, smaller, non-significant differences were found after adjustment for surgeon effect in the post hoc analysis. As high initial migration is predictive for late aseptic loosening (Ryd et al. 1995, Pijls et al. 2012), our results suggests that by using a Triathlon all-polyethylene tibial component the risk of late loosening is at least comparable with, if not less than, that of its metal-backed counterpart. Whereas the first-generation all-polyethylene TKA designs often failed due to loosening, our findings support a growing body of evidence that modern all-polyethylene designs are performing at least equally as well as metal-backed TKA designs (Voigt and Mosier 2011, Nouta et al. 2012, Voss et al. 2016). Previous RSA studies have shown all-polyethylene designs of various manufacturers to have comparable implant migration to its metal-backed counterpart (Adalberth et al. 1999, 2000, 2001, Norgren et al. 2004, Hyldahl et al. 2005a, b, Muller et al. 2006). Depending on the cementing technique, Hyldahl et al. (2005a, b) found comparable or lower migration of all-polyethylene components owing to the “teeter-totter” effect (i.e., tensile forces on the opposite side of the implant upon peripheral compressive loading). This adverse effect on migration was found to be greater when the tibial stem of the more rigid metal-backed tray was not cemented. As the tibial components in our study were only horizontally cemented, this could explain the higher migration of the metal-backed components in our study too. Although there is a strong association between high initial migration and late loosening, it remains unclear how to optimally define “high” migration when comparing the performance of different implants (Henricson and Nilsson 2016). The found difference in mean MTPM suggests superiority of the all-polyethylene components over the metal-backed components. On the other hand, 4 components showed continuous migration in the second postoperative year in both groups, thus the number of individual components considered at risk for loosening is equal between groups. Furthermore, in the sensitivity analysis (adjusting for surgeon effect), results within each surgeon group appeared to be still in favor of the all-polyethylene components, but the differences were smaller and not significant anymore. The found surgeon effect highlights that, even today with all of the instrumentation available to promote standardization of surgical procedures, meticulous performance of each surgical step can improve the outcome, at least on a subclinical level. The results of the sensitivity analysis should, however, be regarded with caution due to multiple testing and an insufficient sample size for stratification by surgeon. It would be of interest if future RSA studies further explore this surgeon
12364 HAMERSVELD D.indd 416
Acta Orthopaedica 2018; 89 (4): 412–417
effect by randomizing patients to 2 or more surgeons using identical implants. Most RSA studies have used maximum total point motion as the primary outcome to predict the occurrence of aseptic loosening (Grewal et al. 1992, Ryd et al. 1995, Pijls et al. 2012). Recently, however, Gudnason et al. (2017) advocated the use of other RSA parameters as the main predictor for loosening as MTPM has its limitations. One of the limitations is that one cannot infer the direction of migration of the MTPM values alone, resulting in uncertainty concerning the failure mechanism. But as motion implies a biological effect, which is expected to be greatest at the point of maximum motion (Valstar et al. 2005), merely expressing migration in fixed directions (e.g., anterior/posterior tilt) would in our opinion underestimate this effect in combined directions (e.g., subsidence into the medial-posterior tibial plateau with internal rotation). Another limitation of MTPM is that any movement between the polyethylene insert and the metal tray influences MTPM in marker-based RSA if polyethylene markers are used to represent the tibial component. Although improved locking mechanisms of modern fixed-bearing designs should prevent the insert from moving with respect to the metal tray, one should be aware of this phenomenon as previous studies have shown such movements to occur in older fixed-bearing designs, resulting in unreliable RSA measurements in the transverse plane (Rao et al. 2002, Nilsson et al. 2003, Hansson et al. 2005). It is therefore possible that the found difference is partly caused by movements between the modular components of the metalbacked design, rather than actual migration of the metal tray. One way to overcome this potential problem is to use modelbased RSA measurements, but since all-polyethylene components are radiolucent, model-based RSA was only a possibility in the metal-backed trial arm. Given the known differences in precision between marker-based and model-based analysis (Kaptein et al. 2007), the current study was set up to use only marker-based RSA in both arms, rather than using different RSA methods in each arm. Furthermore, double examinations showed comparable precision between designs in all directions, indicating that the modular insert is most likely securely fixed within the tray. The influence of such movements on MTPM is therefore expected to be negligibly small. In summary, a statistically significantly lower mean migration after 2 years was found in favor of the Triathlon allpolyethylene design, which may put patients at lower risk of aseptic loosening as compared with its metal-backed counterpart. However, smaller, non-significant differences in migration were found after adjustment for surgeon effect in the post hoc analysis. This unexpected surgeon effect warrants further investigation. Supplementary data Tables 3, 4, and 5 and Figure 3 are available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2018.1464317
Acta Orthopaedica 2018; 89 (4): 412–417
The study was designed and coordinated by STL. Data collection was performed by KH. Statistical analysis was done by KH and PM. KH, PM, RN, and STL interpreted the data and wrote the initial draft manuscript. All authors critically revised and approved the manuscript.
Acta thanks Stephan Maximilian Röhrl and other anonymous reviewers for help with peer review of this study.
Adalberth G, Nilsson K G, Bystrom S, Kolstad K, Mallmin H, Milbrink J. Stability assessment of a moderately conforming all-polyethylene tibial component in total knee arthroplasty: a prospective RSA study with. 2 years of follow-up of the Kinemax Plus design. Am J Knee Surg 1999; 12(4): 233-40. Adalberth G, Nilsson K G, Bystrom S, Kolstad K, Milbrink J. Low-conforming all-polyethylene tibial component not inferior to metal-backed component in cemented total knee arthroplasty: prospective, randomized radiostereometric analysis study of the AGC total knee prosthesis. J Arthroplasty 2000; 15(6): 783-92. Adalberth G, Nilsson K G, Bystrom S, Kolstad K, Milbrink J. All-polyethylene versus metal-backed and stemmed tibial components in cemented total knee arthroplasty: a prospective, randomised RSA study. J Bone Joint Surg Br 2001; 83(6): 825-31. 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. Derbyshire B, Prescott R J, Porter M L. Notes on the use and interpretation of radiostereometric analysis. Acta Orthop 2009; 80(1): 124-30. Gioe T J, Maheshwari A V. The all-polyethylene tibial component in primary total knee arthroplasty. J Bone Joint Surg Am 2010; 92(2): 478-87. Grewal R, Rimmer M G, Freeman M A. Early migration of prostheses related to long-term survivorship: comparison of tibial components in knee replacement. J Bone Joint Surg Br 1992; 74(2): 239-42. Gudnason A, Adalberth G, Nilsson K G, Hailer N P. Tibial component rotation around the transverse axis measured by radiostereometry predicts aseptic loosening better than maximal total point motion. Acta Orthop 2017; 88(3): 282-7. Hansson U, Toksvig-Larsen S, Jorn L P, Ryd L. Mobile vs. fixed meniscal bearing in total knee replacement: a randomised radiostereometric study. Knee 2005; 12(6): 414-18. Henricson A, Nilsson K G. Trabecular metal tibial knee component still stable at 10 years. Acta Orthop 2016; 87(5): 504-10. Hyldahl H, Regner L, Carlsson L, Karrholm J, Weidenhielm L. All-polyethylene vs. metal-backed tibial component in total knee arthroplasty—a randomized RSA study comparing early fixation of horizontally and completely cemented tibial components, part 1: Horizontally cemented components: AP better fixated than MB. Acta Orthop 2005a; 76(6): 769-77. Hyldahl H, Regner L, Carlsson L, Karrholm J, Weidenhielm L. All-polyethylene vs. metal-backed tibial component in total knee arthroplasty—a randomized RSA study comparing early fixation of horizontally and completely cemented tibial components, part 2: Completely cemented components: MB not superior to AP components. Acta Orthop 2005b;76(6):77884. ISO 16087:2013(E). Implants for surgery: Roentgen stereophotogrammetric analysis for the assessment of migration of orthopaedic implants. Geneva, Switzerland: International Organization for Standardization; 2013.
12364 HAMERSVELD D.indd 417
Kaptein B L, Valstar E R, Stoel B C, Reiber H C, Nelissen R G. Clinical validation of model-based RSA for a total knee prosthesis. Clin Orthop Relat Res 2007; 464: 205-9. 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(4): 780-5. Molt M, Ryd L, Toksvig-Larsen S. A randomized RSA study concentrating especially on continuous migration. Acta Orthop 2016; 87(3): 262-7. Muller S D, Deehan D J, Holland J P, Outterside S E, Kirk L M, Gregg P J, McCaskie A W. Should we reconsider all-polyethylene tibial implants in total knee replacement? J Bone Joint Surg Br 2006; 88(12): 1596-602. Nilsson K G, Henricson A, Dalen T. In vivo determination of modular tibial insert micromotion. Trans Orthop Res Soc 2003; 28: 1402. Norgren B, Dalen T, Nilsson K G. All-poly tibial component better than metal-backed: a randomized RSA study. Knee 2004; 11(3): 189-96. Nouta K A, Verra W C, Pijls B G, Schoones J W, Nelissen R G. All-polyethylene tibial components are equal to metal-backed components: systematic review and meta-regression. Clin Orthop Relat Res 2012; 470(12): 354959. Pijls B G, Valstar E R, Nouta K A, Plevier J W, Fiocco M, Middeldorp S, Nelissen R G. Early migration of tibial components is associated with late revision: a systematic review and meta-analysis of 21,000 knee arthroplasties. Acta Orthop 2012; 83(6): 614-24. Ranstam J, Turkiewicz A, Boonen S, Van Meirhaeghe J, Bastian L, Wardlaw D. Alternative analyses for handling incomplete follow-up in the intentionto-treat analysis: the randomized controlled trial of balloon kyphoplasty versus non-surgical care for vertebral compression fracture (FREE). BMC Med Res Methodol 2012; 12:35. Rao A R, Engh G A, Collier M B, Lounici S. Tibial interface wear in retrieved total knee components and correlations with modular insert motion. J Bone Joint Surg Am 2002; 84-a(10): 1849-55. Ryd L, Albrektsson B E, Carlsson L, Dansgard F, Herberts P, Lindstrand A, Regner L, Toksvig-Larsen S. Roentgen stereophotogrammetric analysis as a predictor of mechanical loosening of knee prostheses. J Bone Joint Surg Br 1995; 77(3): 377-83. Thomsen M G, Latifi R, Kallemose T, Barfod K W, Husted H, Troelsen A. Good validity and reliability of the forgotten joint score in evaluating the outcome of total knee arthroplasty. Acta Orthop 2016; 87(3): 280-5. Valstar E R, Gill R, Ryd L, Flivik G, Borlin N, Karrholm J. Guidelines for standardization of radiostereometry (RSA) of implants. Acta Orthop 2005; 76(4): 563-72. Voigt J, Mosier M. Cemented all-polyethylene and metal-backed polyethylene tibial components used for primary total knee arthroplasty: a systematic review of the literature and meta-analysis of randomized controlled trials involving 1798 primary total knee implants. J Bone Joint Surg Am 2011; 93(19): 1790-8. Voss B, El-Othmani M M, Schnur A K, Botchway A, Mihalko W M, Saleh K J. A meta-analysis comparing all-polyethylene tibial component to metalbacked tibial component in total knee arthroplasty: assessing survivorship and functional outcomes. J Arthroplasty 2016; 31(11): 2628-36. Wolterbeek N, Garling E H, Mertens B J, Nelissen R G, Valstar E R. Kinematics and early migration in single-radius mobile- and fixed-bearing total knee prostheses. Clin Biomech (Bristol, Avon) 2012; 27(4): 398-402.
Acta Orthopaedica 2018; 89 (4): 418–424
A 2-year RSA study of the Vanguard CR total knee system: A randomized controlled trial comparing patient-specific positioning guides with conventional technique Frank-David ØHRN 1,2, Justin VAN LEEUWEN 3,5, Masako TSUKANAKA 4, and Stephan M RÖHRL 4,5
Hospital, Møre and Romsdal Health Trust, Kristiansund; 2 NTNU Norwegian University of Science and Technology; 3 Department of Orthopaedic Surgery, Betanien Hospital, Skien; 4 Division of Orthopaedic Surgery, Oslo University Hospital, Oslo; 5 Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Norway Correspondence: email@example.com Submitted 2017-12-18. Accepted 2018-04-04.
Background and purpose — There is some concern regarding the revision rate of the Vanguard CR TKA in 1 registry, and the literature is ambiguous about the efficacy of patient-specific positioning guides (PSPGs). The objective of this study was to investigate the stability of the cemented Vanguard CR Total Knee using 2 different surgical techniques. Our hypothesis was that there is no difference in migration when implanting the Vanguard CR with either PSPGs or conventional technique. We hereby present a randomized controlled trial of 2-year follow-up with radiostereometric analysis (RSA). Patients and methods — 40 TKAs were performed between 2011 and 2013 with either PSPGs or the conventional technique and 22 of these were investigated with RSA. Results — The PSPG (8 knees) and the conventional (14 knees) groups had a mean maximum total point motion (MTPM) (95% CI) of 0.83 (0.48–1.18) vs. 0.70 (0.43–0.97) mm, 1.03 (0.60–1.43) vs. 0.86 (0.53–1.19), and 1.46 (1.07– 1.85) vs. 0.80 (0.52–1.43) at 3, 12, and 24 months respectively (p = 0.1). 5 implants had either an MTPM > 1.6 mm at 12 months and/or a migration of more than 0.2 mm between 1- and 2-year follow-ups. 2 of these also had a peripheral subsidence of more than 0.6 mm at 2 years. Interpretation — 5 implants (3 in the PSPG group) were found to be at risk of later aseptic loosening. The PSPG group continuously migrated between 12 and 24 months. The conventional group had an initial high migration between postoperative and 3 months, but seemed more stable after 1 year. Although the difference was not statistically significant, we think the migration in the PSPG group is of some concern.
Not all patients are satisfied after total knee arthroplasty (TKA); in several studies up to 25% of patients have persistent pain and dysfunction (Baker et al. 2007, Beswick et al. 2012, Howells et al. 2016). Many revisions are caused because of aseptic loosening of the implant. Younger patients undergoing TKA (Kurtz et al. 2009, Ravi et al. 2012) show a higher revision rate (Civinini et al. 2017). Thus, patient dissatisfaction, aseptic loosening, and demographic changes are good reasons to try to improve prosthesis designs and surgical precision. At the same time, all changes in clinical practice or choice of implant should follow the principle of stepwise introduction (Malchau 2000, Nelissen et al. 2011, Pijls and Nelissen 2016). The Vanguard Cruciate Retaining (CR) Total Knee (Vanguard Complete Knee System, Zimmer Biomet Inc., Warsaw, IN, USA) was introduced in 2003. In some registries (AOANJRR , NJR) the prosthesis has showed promising results. Yet in another (SKAR), the Vanguard CR had a significantly higher relative risk of revision compared with other implants. The implant can be inserted by conventional surgical technique or with patient-specific positioning guides (PSPGs). PSPGs are customized and manufactured from preoperative CT or MRI data to improve postoperative alignment (van Leeuwen et al. 2015). The literature is still ambiguous regarding the efficacy of PSPGs (Boonen et al. 2012, Nunley et al. 2012, An et al. 2017). Altered surgical technique or alignment might influence the early stability of the implants. The hypothesis of this study was that the cemented Vanguard CR TKA is a stable implant using PSPGs. Therefore we investigated the stability of the cemented Vanguard CR Total Knee using 2 different surgical techniques.
© 2018 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 10.1080/17453674.2018.1470866
12421 +ÿhrn D.indd 418
Acta Orthopaedica 2018; 89 (4): 418–424
Multicenter study, randomized, n = 109
Patients and methods This study was part of a randomized controlled multicentre trial (RCT) in Oslo and Skien, Norway, which compared clinical and radiological but no radiostereometric analysis (RSA) results of the PSPG technique (Signature Personalized Patient Care System; Zimmer Biomet) with the conventional technique for TKA. The exclusion criteria were published in that study (van Leeuwen et al. 2018). All surgeries and investigations in the RSA cohort were performed at Ullevål Hospital, Oslo, Norway. 40 patients participated in the RSA study at Ullevål Hospital, but only 22 were included in the RSA analyses (Figure 1). These were operated between December 2011 and December 2013. A Vanguard Cruciate Retaining (CR) TKA was performed in all patients (Cemented Vanguard Complete Knee System; Zimmer Biomet Inc., Warsaw, IN): 8 with the PSPG technique and 14 with the conventional surgical method. Design This study was designed as a single blinded RCT of patients receiving TKA for symptomatic osteoarthritis of the knee.
Excluded (n = 15) see van Leeuwen et al. 2018 for details Multicenter study, included, n = 84 RSA patients at Ullevål Hospital, n = 40 Allocation Allocated to PSPG TKA (n = 18) Excluded (n = 10): – withdrew consent/ not operated, 4 – MRI artefacts, 1 – no RSA beads or inadequate RSA pictures, 5
Excluded (n = 8): – withdrew consent, 3 – no RSA beads or inadequate RSA pictures, 5 Analyses
Initial RSA (n = 7) Excluded due to high CN (n = 1)
Initial RSA (n = 14)
3-months RSA (n = 7) Excluded due to high CN (n = 1)
3-months RSA (n = 14)
12-months RSA (n = 6) Excluded (n = 2): – high CN, 1 – did not attend, 1
12-months RSA (n = 14)
24-months RSA (n = 7) Excluded due to high CN (n = 1)
24-months RSA (n = 13) Excluded due to high RBE (n = 1)
Figure 1. Flow chart.
Patients The participants were assigned to either the conventional or PSPG technique according to the protocol of the multicentre RCT with block randomization obtained by variable block sizes (van Leeuwen et al. 2018). In the original RCT the sample size was calculated for the frontal mechanical axis and the secondary outcome KOOS score. That study was terminated when the total number of patients was sufficient according to the primary outcome measure, hence the suboptimal number of patients in the RSA study (Figure 1). The surgeries were performed by 2 experienced surgeons. 7 patients withdrew their consent or were not operated for various reasons after randomization, 1 had MRI artefacts that precluded manufacturing of PSPGs, and 10 did not have beads inserted or had inadequate RSA pictures, thus only 22 had RSA in this study. At 2 of the time points there were only 20 patients included in these analyses. 1 patient had a high condition number (CN) and was excluded at all time points, 1 did not show up at 12 months, and 1 had high rigid body error (RBE) at 24 months.
12421 +ÿhrn D.indd 419
Allocated to conventional TKA (n = 22)
Clinical evaluation—KOOS (n = 32) Excluded (n = 8): – withdrew consent/not operated, 7 – MRI artefacts, 1
Intervention We used a standard midline incision and medial parapatellar capsulotomy in all patients. A tourniquet was used in all cases. For details of the operative procedure see van Leeuwen (2018). Both surgical techniques (conventional and PSPG) for this implant were well established in the department prior to the inclusion of RSA patients, so we assumed that there was no learning curve. During surgery 6 to 8 1.0 mm tantalum markers (RSA Biomedical, Umeå, Sweden) were inserted in the tibia. All patients followed the same standardized postoperative rehabilitation protocol. Evaluation Implant migration was evaluated using RSA. The first examination took place within a week postoperatively, then after 3, 12, and 24 months. They were all performed in the supine position by the same radiographers at each time point. We used calibration cage number 43 (RSA Biomedical, Umeå, Sweden) and ceiling mounted X-ray tubes (Proteus XR/A, GE Healthcare and Canon Triathlon T3).
Acta Orthopaedica 2018; 89 (4): 418–424
Table 1. Baseline characteristics of conventional vs. PSPG patients. Values are mean (SD) (range) unless otherwise specified Factor Number of patients Left/right, n Men/women ratio, n Age BMI Body weight (kg) Operation time (min) Postoperative HKA (°)
Conventional 14 10/4 7/7 65 (7.9) (53–77) 28 (4.2) (22–36) 84 (18) (60–105) 121 (37) (68–228) 181 (4.8) (172–188)
PSPG 8 3/5 2/6 60 (5.1) (50–68) 30 (4.8) (22–35) 87 (12) (70–100) 111 (8.9) (95–125) 178 (5.7) (171–186)
Figure 2. Fictive points of the tibial implant (the posterior fictive point is hidden behind the stem).
MB-RSA 3.40 (RSAcore, Leiden, The Netherlands) software was used for the migration analysis. The migration was described both as segment motion of all 6 degrees of freedom (translations and rotations) and maximum total point motion (MTPM), the latter being primary outcome. In addition, we analyzed the point motion from fictive points added to the computer aided design (CAD) model of the tibial component. We had the following 7 fictive points: stem tip, anterior, posterior, posteromedial, posterolateral, medial, and lateral (Figure 2). All these points were reported for X, Y, and Z translations. As we performed double examinations, the 2 RSA pictures were run against all the others, making a total of 4 motions for each patient at each time point. The average of these 4-point motions represented the motion of the individual implant at each time point. The movements of the 13 left knees were converted to right knees for stability analysis (Valstar et al. 2005). Our upper limit for CN was 100, and for RBE 0.50 mm. 1 patient exceeded 0.5 mm RBE at 2 years and was excluded from the RSA examination at this time point. The rest had RBE of less than 0.35mm. Precision was assessed by double RSA examinations of all patients at all time points, and reported as absolute mean difference of double examinations ± 1.96 x standard deviation (SD). For clinical assessment we used the Knee injury and Osteoarthritis Outcome Score (KOOS) (Roos et al. 1998). All complications were registered. Statistics We used linear mixed models to evaluate differences in MTPM, translations, rotations, and point motions within groups (PSPG vs. conventional) over the entire follow-up period and to control for repeated measurements. The fixed effects were time, group, and time-by-group interaction. The model included a random slope. As the KOOS scores were not normally distributed, the Wilcoxon signed-rank test was used to evaluate difference from preoperative to 2 years. Further, patients were divided into high- and low-risk group according their migration data (Ryd et al. 1995, Pijls et al. 2012). To estimate differences in KOOS scores between these 2 inde-
12421 +ÿhrn D.indd 420
pendent groups we used the Mann–Whitney U test. Fisher’s exact test was used to detect associations between categorical independent variables. The results are reported as means or proportions with 95% confidence intervals (CI), if not stated otherwise. All statistical calculations were performed using the IBM SPSS Statistics version 23 (IBM Corp, Armonk, NY, USA). Ethics, registration, funding, and potential conflicts of interest The study was approved by the Regional Committee for Medical and Health Research Ethics, West-Norway (REC West, approval number 2010/2056) and the institutional review board at Oslo University Hospital (2011/7613), and registered at clinicaltrials.gov (NCT01696552). All patients were included with written consent. No financial funding from companies has been received for this study, and the authors declare that there are no conflicts of interest.
Results Demographics See Table 1 for baseline characteristics. RSA The mean MTPM, and relevant point motions, translations, and rotations are shown in Tables 2 and 3 (see Supplementary data) and Figures 3–6. The PSPG group had an increasing migration pattern compared with the conventional group. The results from the linear mixed model analysis showed a statistically significant change in MTPM within (p < 0.001), but not between the 2 groups after 2 years (p = 0.1) (Table 4). Generally in the point motions analysis, we found a larger subsidence in the PSPG than the conventional group, but no statistical significance could be found (Table 3, see Supplementary data, Figure 4). On an individual basis 4 implants had more than 1.6 mm migration at 12 months, but 1 of these was excluded due to an RBE > 0.5 mm at 2 years (Figure 5). 4 implants had more
Acta Orthopaedica 2018; 89 (4): 418–424
Y-axis point motion (mm)
0.2 Tip Conv
0.1 1.6 mm threshold PSPG
Tip PSPG Anterior Conv
Conventional Precision 95% CI
0.5 mm threshold
Posterior PSPG Posteromedial Conv Posteromedial PSPG Posterolateral Conv Posterolateral PSPG
Lateral Conv Lateral PSPG
Figure 3. Mean MTPM over time for the whole cohort and for the PSPG and conventional groups with thresholds (Pijls et al. 2012).
Months after index operation
Months after index operation
Figure 4. Y (axial, lift-off, subsidence) point motions stratified in PSPG (dashed lines) vs. conventional.
X and Z rotation (°) / Y translation (mm)
Individual MTPM (mm)
X rotation Conv
X rotation PSPG
Z rotation Conv
Z rotation PSPG
Y translation Conv
Y translation PSPG
Months after index operation
Figure 5. Individual time profiles of MTPM in the two subgroups (n = 21). Conventional marked with blue lines, PSPGs with green lines.
Months after index operation
Figure 6. X and Z rotation in degrees and Y translation in mms (PSPG vs. conventional).
than 0.2 mm migration between 1- and 2-year follow-ups. Of the 3 remaining patients with MTPM > 1.6 mm at 1 year, 2 had migration of more than 0.2 mm between 1 and 2 years, hence 5 patients had either >1.6 mm at 1 year, or > 0.2 mm migration between 1 and 2 years. 2 of these patients also met the criteria for distal or proximal peripheral translation. No other implants met these criteria. None of the implants met the criteria of transversal rotation (Gudnason et al. 2017). In 3 of these 5 high-risk patients the PSPG method had been used. The precisions of our RSA examinations were the following: 0.31 mm for MTPM (95% CI 0.00–0.74), 0.01 mm for X translation (95% CI –0.12 to 0.14), 0.01 mm for Y translation (95% CI –0.07 to 0.08), 0.03mm for Z translation (95% CI
12421 +ÿhrn D.indd 421
–0.25 to 0.32), 0.05° for X rotation (95% CI –0.28 to 0.37), 0.04° for Y rotation (95% CI –0.65 to 0.72), and 0.00° for Z rotation (95% CI –0.14 to 0.15). Clinical results We found a statistically significant improvement of all the KOOS subscales from preoperative through 2 years in the whole cohort. We could not see any difference in clinical performance for implants with migration at risk or PSPG and conventional groups (Mann–Whitney U test) (Figure 7, see Supplementary data). Neither could we demonstrate any other subgroup to explain the inferior stability of the high-risk group (Table 5, see Supplementary data).
Acta Orthopaedica 2018; 89 (4): 418–424
Table 4. Results of the linear mixed model analysis of MTPM at 2 years after randomization into PSPGs and conventional subgroups Parameter Intercept Time Randomization Time x Randomization
95 % CI
–0.13 0.30 –0.18 0.16
0.2 < 0.001 0.3 0.1
–0.32 to 0.07 0.18 to 0.42 –0.50 to 0.14 –0.04 to 0.36
Among the patients with high-risk migration we found 1 with inferior clinical scoring. This person was in the conventional group and had a postoperative hematoma with no need for further surgery. Postoperative radiographs showed an HKA-angle of 172° (varus). Complications 5 complications occurred: 1 deep hematoma that was evacuated, 1 stiff knee requiring mobilization under anaesthesia, 2 superficial hematomas, and 1 superficial infection. The latter was in the PSPG group, the others in the conventional group. None of the complications required reoperation. Except for the aforementioned high-risk patient, they had good clinical scores after 2 years.
Discussion Our main finding was that the implants in the PSPG group had continuous migration between 12 and 24 months. The implants in the conventional group showed migration between postoperatively and 3 months; thereafter the mean migration abated and the implant stabilized. The difference in MTPM and subsidence between the two groups, however, was not statistically significant. Several studies have discussed threshold levels for increased risk of aseptic loosening. Ryd et al. (1995) showed that the process of loosening probably starts directly after the operation and that a migration of more than 0.2 mm after 1 year gives a high risk of revision. Pijls et al. (2012) showed that more than 1.6 mm migration at 12 months gives an “unacceptable” risk of later revision. Gudnason et al. (2017) recently suggested a transversal rotation of more than 0.8°, or peripheral distal or proximal translation of more than 0.6 mm or 0.9 mm respectively at 2 years as a threshold. 5 of the 22 implants in our study met 1 or more of these criteria for high-risk implants. We could not identify any other factor than surgical technique that could explain why these knees performed worse than the rest, such as obesity, age, postoperative valgus or varus. However, due to the limited numbers of observations, emphasis should be given to the estimated values rather than p-values. All the high-risk patients, except for 1, also performed well clinically, with no symptoms of early loosening after 2 years. It is important to stress that although implants for individual
12421 +ÿhrn D.indd 422
patients met the criteria for “unacceptable risk” according to Pijls et al. (2012), it does not mean that the specific implant is loose. The meta-analysis of Pijls focused on mean MTPM with 12 months’ observation time, mainly because not all studies reported 2-year results, or migration in all degrees of freedom. The discussion concerning which criteria to follow therefore continues. In our study, as many as 5 of 22 patients were at risk of later loosening based on several studies (Ryd et al. 1995, Pijls et al. 2012), but following the recent study by Gudnason et al. (2017), only 2 implants were at risk of aseptic loosening, 1 in each group. Also, in that study the authors concluded that MTPM after 1 and 2 years is inferior to transversal rotation, peripheral subsidence, and lift-off in predicting late aseptic loosening. A limitation of our study is the sample size. Ideally, we would have liked between 25 and 30 participants in each group, yet several other RSA studies have suboptimal sample sizes for various reasons (Hansson et al. 2005, Molt and Toksvig-Larsen 2014, Henricson and Nilsson 2016, Meinardi et al. 2016). RSA research is costly and tedious work, and it is not always possible to recruit enough patients. As the cohort was part of a larger RCT assessing clinical and radiological outcome of 2 different surgical methods, the RSA study was not powered as an RCT. This may be a possible reason why we could not find a statistically significant difference in MTPM between the 2 groups. In addition, in the larger RCT, a statistically significant difference was found in the position of the tibia in the frontal and sagittal planes (van Leeuwen et al. 2018). As the surgical techniques were well established in the department, we assume there was no learning curve. Longer follow-up of both groups is needed, especially with a focus on the continuous migration of the PSPG group. One strength of our study is that we used fictive points in our RSA model. We could therefore show with which pattern the implant was migrating. Many studies include only MTPM and the segmental micromotions, their absolute values are often smaller than the peripheral point motions, and they do not tell us exactly how the implant migrates. Thus we could also evaluate the implant with respect to Gudnason’s data (Gudnason et al. 2017). The long-term results of the implant we used diverge in the literature. Some registries show excellent results after 5- and 10-year follow-up (AOANJRR , NJR), the latter with only a few hundred patients reaching 10 years, and with no information regarding surgical technique. Several clinical studies also show excellent results (Kievit et al. 2014, Schroer et al. 2014, Faris et al. 2015, Emerson et al. 2016, Flament et al. 2016). To our knowledge, there is only 1 other study that has assessed the Cemented Vanguard CR with RSA (Schotanus et al. 2017). They found a mean MTPM for this implant of 0.7 mm at 12 months and 0.8 mm after 24 months. Although slightly lower migration than our data suggest, it leaves the implant in the same risk category according to Pijls et al.(2012). However, they did not use PSPGs. Another register found the implant to
Acta Orthopaedica 2018; 89 (4): 418–424
perform worse compared with other implants (SKAR). This effect is not present when a patellar button is implanted during primary surgery. As long-term data are still lacking, especially on the Signature System (PSPG), and only 1 RSA study shows early follow-up data on the implant, our study adds knowledge for users of this implant. In summary we found that the cemented Vanguard CR had a higher initial mean migration than expected at 12 months, but from 12–24 months the conventional group stabilized. The PSPG group also had continuous migration at this point. None of the implants in our study rotated more than recommended, and only 2 implants had a total peripheral subsidence above that recommended, 1 in each group. Although the PSPG group did not have a statistically different MTPM from the conventional group, we think that the findings of the migration pattern of this technique are of some concern and call for longer follow-up. Supplementary data Tables 2, 3, and 5 and Figure 7 are available in the online version of this article, http://dx.doi.org/ 10.1080/17453674. 2018.1470866
FDØ analysed parts of the data and wrote the manuscript. MT analysed parts of the data and critically reviewed the manuscript. JvL and SMR designed the study and critically reviewed the manuscript. The authors would like to thank the radiographers Mona Risdal, Silje Clausen, and Alexis Hinojosa, statistician Cathrine Brunborg and the study coordinators Marte Traae Magnusson and Anette Simonsen for their contributions to the study. Acta thanks Bart G Pijls and Nikolaj Sebastian Winther for help with peer review of this study.
An V V, Sivakumar B S, Phan K, Levy Y D, Bruce W J. Accuracy of MRIbased vs. CT-based patient-specific instrumentation in total knee arthroplasty: A meta-analysis. J Orthop Sci 2017; 22(1): 116–20. doi: 10.1016/j. jos.2016.10.007. AOANJRR. Australian Orthopaedic Association National Joint Replacement Registry. Annual Report 2017 p 193. https://aoanjrrsahmricom/annualreports-2017. Baker P N, van der Meulen J H, Lewsey J, Gregg P J. The role of pain and function in determining patient satisfaction after total knee replacement. Data from the National Joint Registry for England and Wales. J Bone Joint Surg Br 2007; 89(7): 893–900. doi: 10.1302/0301-620x.89b7.19091. Beswick A D, Wylde V, Gooberman-Hill R, Blom A, Dieppe P. What proportion of patients report long-term pain after total hip or knee replacement for osteoarthritis? A systematic review of prospective studies in unselected patients. BMJ Open 2012; 2(1): e000435. doi: 10.1136/bmjopen-2011-000435. Boonen B, Schotanus M G, Kort N P. Preliminary experience with the patientspecific templating total knee arthroplasty. Acta Orthop 2012; 83(4): 387– 93. doi: 10.3109/17453674.2012.711700.
12421 +ÿhrn D.indd 423
Civinini R, Carulli C, Matassi F, Lepri A C, Sirleo L, Innocenti M. The survival of total knee arthroplasty: current data from registries on tribology: review article. HSS J 2017; 13(1): 28–31. doi: 10.1007/s11420-016-95139. Emerson R H Jr, Barrington J W, Olugbode S A, Alnachoukati O K. A comparison of 2 tibial inserts of different constraint for cruciate-retaining primary total knee arthroplasty: an additional tool for balancing the posterior cruciate ligament. J Arthroplasty 2016; 31(2): 425–8. doi: 10.1016/j. arth.2015.09.032. Faris P M, Ritter M A, Davis K E, Priscu H M. Ten-year outcome comparison of the anatomical graduated component and Vanguard total knee arthroplasty systems. J Arthroplasty 2015; 30(10): 1733–5. doi: 10.1016/j. arth.2015.04.042. Flament E M, Berend K R, Hurst J M, Morris M J, Adams J B, Lombardi A V, Jr. Early experience with vitamin E antioxidant-infused highly cross-linked polyethylene inserts in primary total knee arthroplasty. Surg Technol Int 2016; Xxix: 334–40. Gudnason A, Adalberth G, Nilsson K G, Hailer N P. Tibial component rotation around the transverse axis measured by radiostereometry predicts aseptic loosening better than maximal total point motion. Acta Orthop 2017: 88(3): 282-7. doi: 10.1080/17453674.2017.1297001. Hansson U, Toksvig-Larsen S, Jorn L P, Ryd L. Mobile vs. fixed meniscal bearing in total knee replacement: a randomised radiostereometric study. Knee 2005; 12(6): 414–18. doi: 10.1016/j.knee.2004.12.002. Henricson A, Nilsson K G. Trabecular metal tibial knee component still stable at 10 years. Acta Orthop 2016; 87(5): 504–10. doi: 10.1080/17453674.2016.1205169. Howells N, Murray J, Wylde V, Dieppe P, Blom A. Persistent pain after knee replacement: do factors associated with pain vary with degree of patient dissatisfaction? Osteoarthritis Cartilage 2016; 24(12): 2061–8. doi: 10.1016/j.joca.2016.07.012. Kievit A J, Schafroth M U, Blankevoort L, Sierevelt I N, van Dijk C N, van Geenen R C. Early experience with the Vanguard complete total knee system: 2–7 years of follow-up and risk factors for revision. J Arthroplasty 2014; 29(2): 348–54. doi: 10.1016/j.arth.2013.05.018. Kurtz S M, Lau E, Ong K, Zhao K, Kelly M, Bozic K J. Future young patient demand for primary and revision joint replacement: national projections from 2010 to 2030. Clin Orthop Relat Res 2009; 467(10): 2606–12. doi: 10.1007/s11999-009-0834-6. Malchau H. Introducing new technology: a stepwise algorithm. Spine (Phila Pa 1976). 2000; 25(3): 285. Meinardi J E, Valstar E R, Van Der Voort P, Kaptein B L, Fiocco M, Nelissen R G. Palacos compared to Palamed bone cement in total hip replacement: a randomized controlled trial. Acta Orthop 2016; 87(5): 473–8. doi: 10.1080/17453674.2016.1199146. Molt M, Toksvig-Larsen S. Similar early migration when comparing CR and PS in Triathlon TKA: a prospective randomised RSA trial. Knee 2014; 21(5): 949–54. doi: 10.1016/j.knee.2014.05.012. Nelissen R G, Pijls B G, Karrholm J, Malchau H, Nieuwenhuijse M J, Valstar E R. RSA and registries: the quest for phased introduction of new implants. J Bone Joint Surg Am 2011; 93 (Suppl 3): 62–5. doi: 10.2106/jbjs.k.00907. NJR. National Joint Registry for England, Wales, Northern Ireland and the Isle of Man. 14th Annual Report 2017, p. 137. Available from: http://www. njrreports.org.uk/. Nunley R M, Ellison B S, Zhu J, Ruh E L, Howell S M, Barrack R L. Do patient-specific guides improve coronal alignment in total knee arthroplasty? Clin Orthop Relat Res 2012; 470(3): 895–902. doi: 10.1007/ s11999-011-2222-2. Pijls B G, Nelissen R G. The era of phased introduction of new implants. Bone Joint Res 2016; 5(6): 215–17. doi: 10.1302/2046-3758.56.2000653. Pijls B G, Valstar E R, Nouta K A, Plevier J W, Fiocco M, Middeldorp S, Nelissen R G. Early migration of tibial components is associated with late revision: a systematic review and meta-analysis of 21,000 knee arthroplasties. Acta Orthop 2012; 83(6): 614–24. doi: 10.3109/17453674.2012.747052.
Ravi B, Croxford R, Reichmann W M, Losina E, Katz J N, Hawker G A. The changing demographics of total joint arthroplasty recipients in the United States and Ontario from 2001 to 2007. Best Pract Res Clin Rheumatol 2012; 26(5): 637–47. doi: 10.1016/j.berh.2012.07.014. 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. doi: 10.2519/jospt.19220.127.116.11. Ryd L, Albrektsson B E, Carlsson L, Dansgard F, Herberts P, Lindstrand A, Regner L, Toksvig-Larsen S. Roentgen stereophotogrammetric analysis as a predictor of mechanical loosening of knee prostheses. J Bone Joint Surg Br 1995; 77(3): 377–83. Schotanus M G, Pilot P, Kaptein B L, Draijer W F, Tilman P B, Vos R, Kort N P. No difference in terms of radiostereometric analysis between fixed- and mobile-bearing total knee arthroplasty: a randomized, single-blind, controlled trial. Knee Surg Sports Traumatol Arthrosc 2017; 25(9): 2878–85. doi: 10.1007/s00167-016-4138-6.
12421 +ÿhrn D.indd 424
Acta Orthopaedica 2018; 89 (4): 418–424
Schroer W C, Stormont D M, Pietrzak W S. Seven-year survivorship and functional outcomes of the high-flexion Vanguard complete knee system. J Arthroplasty 2014; 29(1): 61–5. doi: 10.1016/j.arth.2013.04.018. SKAR. The Swedish Knee Arthroplasty Register—Annual Report 2016, p. 46, http://www.myknee.se/pdf/SVK_2016_Eng_1.0.pdf. Valstar E R, Gill R, Ryd L, Flivik G, Borlin N, Karrholm J. Guidelines for standardization of radiostereometry (RSA) of implants. Acta Orthop 2005; 76(4): 563–72. doi: 10.1080/17453670510041574. van Leeuwen J A, Grogaard B, Nordsletten L, Rohrl S M. Comparison of planned and achieved implant position in total knee arthroplasty with patient-specific positioning guides. Acta Orthop 2015; 86(2): 201–7. doi: 10.3109/17453674.2014.985154. van Leeuwen J, Snorrason F, Rohrl S M. No radiological and clinical advantages with patient-specific positioning guides in total knee replacement. Acta Orthop 2018; 89(1): 89–94. doi: 10.1080/17453674.2017.1393732.
Acta Orthopaedica 2018; 89 (4): 425–430
Peri-apatite coating decreases uncemented tibial component migration: long-term RSA results of a randomized controlled trial and limitations of short-term results Koen T VAN HAMERSVELD 1, Perla J MARANG-VAN DE MHEEN 2, Rob G H H NELISSEN 1, and Sören TOKSVIG-LARSEN 3
1 Department of Orthopaedics, 2 Medical Decision Making, Leiden University Medical Center, Leiden, The Netherlands; 3 Department of Orthopaedics, Hässleholm Hospital, Hässleholm, Sweden and Department of Clinical Sciences, Lund University, Lund, Sweden Correspondence: firstname.lastname@example.org Submitted 2017-12-22. Accepted 2018-04-04.
Background and purpose — Biological fixation of uncemented knee prostheses can be improved by applying hydroxyapatite coating around the porous surface via a solution deposition technique called Peri-Apatite (PA). The 2-year results of a randomized controlled trial, evaluating the effect of PA, revealed several components with continuous migration in the second postoperative year, particularly in the uncoated group. To evaluate whether absence of early stabilization is diagnostic of loosening, we now present long-term follow-up results. Patients and methods — 60 patients were randomized to PA-coated or uncoated (porous only) total knee arthroplasty of which 58 were evaluated with radiostereometric analysis (RSA) performed at baseline, at 3 months postoperatively and at 1, 2, 5, 7, and 10 years. A linear mixed-effects model was used to analyze the repeated measurements. Results — PA-coated components had a statistically significantly lower mean migration at 10 years of 0.94 mm (95% CI 0.72–1.2) compared with the uncoated group showing a mean migration of 1.72 mm (95% CI 1.4–2.1). Continuous migration in the second postoperative year was seen in 7 uncoated components and in 1 PA-coated component. All of these implants stabilized after 2 years except for 2 uncoated components. Interpretation — Peri-apatite enhances stabilization of uncemented components. The number of components that stabilized after 2 years emphasizes the importance of longer follow-up to determine full stabilization and risk of loosening in uncemented components with biphasic migration profiles.
Early migration of tibial components, which can be accurately measured with radiostereometric analysis (RSA), has been shown to predict future aseptic loosening (Ryd et al. 1995, Pijls et al. 2012b). Uncemented components typically display a biphasic migration pattern with high initial migration before stabilization (Pijls et al. 2012a, Wilson et al. 2012, Henricson and Nilsson 2016), while cemented components are initially more stable as the cement provides instant fixation, yet continuous bone resorption at the cement–bone interface may result in continuous migration (Nilsson et al. 2006, van Hamersveld et al. 2017). Given the importance of stabilization in the first months after implantation, one method to improve bone ingrowth after uncemented total knee arthroplasty (TKA) is the application of osteoconductive hydroxyapatite (HA) coatings (Nelissen et al. 1998, Carlsson et al. 2005). Most HA coatings are plasma sprayed onto the porous beaded implant surface area. Plasma spraying is a “line of sight” technique and therefore only able to coat the substrate surface (Hansson et al. 2008). Contrarily, Peri-Apatite HA (PA) (Stryker, Mahwah, NJ, USA)is an alternative technique to deposit HA from an aqueous solution at room temperature, thereby increasing the coverage of HA onto the 3D beaded implant surface (Serekian 2004). However, without the effect of high temperatures up to 20,000 ºC associated with plasma spraying, the HA remains pure and 100% crystalline, while a lower crystallinity has been shown to improve the bioactivity and resorption profile of HA (Overgaard et al. 1999, Serekian 2004). In addition, the adhesion of the relatively thin PA layer (of 20 µm compared with 50–75 µm for most HA coatings) is fragile when touching the coated metal during implantation and, like any HA coating, might delaminate or release particles over time (Bloebaum et al. 1994, Morscher et
© 2018 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 10.1080/17453674.2018.1469223
12444 Hamersveld D.indd 425
al. 1998). Only a few randomized RSA studies have assessed the short-term (2-year follow-up) effect of PA on uncemented tibial component migration (van der Linde et al. 2006, Hansson et al. 2008, Therbo et al. 2008, Molt and Toksvig-Larsen 2014). All trials concluded that the PA coating appears to improve stabilization up to 2 years after implantation. However, no studies have examined long-term migration profiles of PA-coated tibial components. It is therefore unknown whether the found short-term effect on component fixation is sustained over time. Furthermore, in the short-term report of the current study (Molt and Toksvig-Larsen 2014), a number of both uncoated and PA-coated components showed continuous migration in the second postoperative year. It is unclear whether this leads to future aseptic loosening or if this high initial migration is merely part of a migration pattern typical for uncemented components. We therefore now report 10-year follow-up results of this double-blinded, randomized controlled trial comparing implant migration measured with RSA and clinical results of PA-coated with uncoated uncemented TKAs.
Patients and methods Study design Full details of the design and patient selection of this randomized controlled trial have been described previously (Molt and Toksvig-Larsen 2014). In short, all consecutive patients scheduled to undergo TKA due to primary osteoarthritis from July 2007 until February 2008 in Hässleholm Hospital (Sweden) were asked to participate. 60 patients were randomized in a 1:1 ratio. Patients received either “PA-coated” (applied on both the femoral and tibial component) or “uncoated” components of an otherwise identical (fully) uncemented cruciate retaining Triathlon total knee prosthesis (Stryker, Mahwah, NJ, USA). The porous undersurface (in both versions) consisted of cobalt-chromium sintered beads with a porosity of 35% and mean pore size of 425 µm. Highly cross-linked polyethylene inserts were used in all cases. At all follow-up points, the Knee Society Score (KSS) and the Knee injury and Osteoarthritis Outcome Score (KOOS) were obtained. Both patients and observers performing clinical follow-up and RSA measurements remained blinded to the allocated group during the entire follow-up period. Radiostereometric analysis RSA radiographs were made on the first day after surgery when weight bearing was achieved. Subsequent examinations were performed after 3 months, 1 year, 2, 5, 7, and 10 years. RSA radiographs were performed in supine position with the knee in a calibration cage (Cage 10, RSA Biomedical, Umeå, Sweden). RSA measurements were analyzed using UmRSA software (v6.0, RSA, Biomedical, Umeå, Sweden). Positive directions along and about the orthogonal axes are
12444 Hamersveld D.indd 426
Acta Orthopaedica 2018; 89 (4): 425–430
according to RSA guidelines (Valstar et al. 2005). Migration was described as translation of the geometric center of the prosthesis markers and rotation of the rigid body defined by the prosthesis markers about this geometric center of gravity. The length of the translation vector of the marker or virtual marker in a rigid body that has the greatest migration, i.e., the maximum total point motion (MTPM), was used as the primary outcome measure (ISO 16087:2013(E) 2013). The first postoperative RSA examination served as the reference for the migration measurements. Individual components with “continuous migration,” defined by Ryd et al. (1995) as an increase in MTPM of 0.2 mm or more in the second postoperative year, were classified as “loose.” This threshold was set at 0.1 mm per year after 2-year follow-up according to the modified continuous migration criterion (Ryd et al. 1995). Consequently, implants classified in the second postoperative year as loose were considered stabilized if the migration was less than 0.1 mm/year between 2-year and final follow-up (Wilson et al. 2012, Molt et al. 2016). The precision of the local RSA set-up after the 2-year follow-up period, specified as the 95% confidence interval (CI) around zero motion, and measured with 15 double examinations (ISO 16087:2013(E) 2013), was 0.10 mm, 0.10 mm, and 0.09 mm for transverse, longitudinal, and sagittal translations; 0.20°, 0.20°, and 0.24° for transverse, longitudinal, and sagittal rotations, respectively. The mean error of rigid body fitting of the RSA markers was below 0.35 mm and the upper limit for the condition number was set at 120, complying with the suggested limits of the RSA guidelines (ISO 16087:2013(E) 2013). The mean condition number was 40 (CI 37–42) and 51 (CI 49–54) for the implant and tibial markers, respectively. Statistics Given the high accuracy of RSA measurements, only 17 patients were needed in each group to detect a decrease in migration from 1.0 to 0.5 ± 0.5 mm with 80% power and alpha set at 0.05, as described previously (Molt and ToksvigLarsen 2014). 30 patients were randomized to each group to account for possible dropouts. The original primary outcome reported by Molt and Toksvig-Larsen (2014) was a difference in migration (MTPM) after 2 years of follow-up. For this long-term outcome report, the primary outcome was a difference in MTPM after 10 years of follow-up as registered at ClinicalTrials.gov (ID: NCT03198533). Data were analyzed according to the intention-to-treat principle. A linear mixed-effects model was used for all repeated measurements to effectively deal with missing values within patients during follow-up. As MTPM is always a positive vector, normal distribution was only obtained after log-transformation (logMTPM), computed as log10(MTPM+1). Differences in mean progression of logMTPM between groups were modeled as a function of time and the interaction of time with treatment. A random-intercepts term was used and remaining variability was modelled with a heterogeneous autoregressive order 1
Acta Orthopaedica 2018; 89 (4): 425–430
Table 1. Baseline demographic characteristics. Values are mean (SD) unless otherwise specified
Randomized (n = 60) Allocation
Age Body mass index Female sex (n) Previous knee surgery (n) none joint debridement meniscectomy other Ahlbäck’s grade (n) II III IV ASA classification (n) I II III Hip–knee–ankle angle preoperative postoperative
Uncoated (n = 29)
PA-coated (n = 29)
67 (6.8) 30 (4.3) 16
65 (8.1) 30 (4.9) 17
22 1 5 1
25 1 2 1
12 15 2
6 22 1
8 20 1
6 21 2
175 (5.0) 179 (2.8)
176 (6.2) 179 (3.2)
Allocated to uncoated TKA (n = 30): – received allocated treatment, 29 – received cemented TKA due to an intraoperative fracture, 1
Follow-up Lost to follow-up (n = 12): – revised, 3 a – died after 1, (2x) 2 and 7 years, 4 – withdrew after 1, 5 and (3x) 7 years, 5 RSA analyzed: Postop: 29 3 months: 28 1 year: 25 2 years: 25 5 years: 16 7 years: 20 10 years: 17
Lost to follow-up (n = 13): – died after 1 and (2x) 5 years, 3 – withdrew after (3x) 1, (3x) 2, (2x) 5 and 7 years, 9 – technical error after 2 years, 1 b Analysis
RSA analyzed: Postop: 29 3 months: 29 1 year: 28 2 years: 25 5 years: 16 7 years: 17 10 years: 16
Figure 1. CONSORT flow diagram. TKA = total knee arthroplasty. a revised after 3 months (early infection), 1 year (late infection), and 10 years (mechanical failure). b clinical follow-up only, see text.
covariance structure. Secondary outcomes (RSA translations and rotations, flexion, extension, KSS, and KOOS scores) were analyzed with a similar mixed-effects model. Differences in mean migration along and about each orthogonal axis were calculated using log-transformed absolute values (as the resultant of positive and negative displacement vectors requires all vectors to act on the same prosthesis) (Derbyshire et al. 2009). Given the non-normal distribution of knee extension and the KSS knee score (not resulting in a normal distribution after a log transformation), a comparable generalized estimating equations (GEE) approach was used to correct the standard errors via the sandwich estimator. Post hoc testing was performed to estimate between-group differences in MTPM using 3 months, 1 year, and 2 years as the reference. IBM SPSS Statistics 24.0 (IBM Corp, Armonk, NY, USA) was used for all outcome measures; a p-value < 0.05 was considered significant. Ethics, registration, funding, and potential conflicts of interest The trial was performed in compliance with the Declaration of Helsinki and Good Clinical Practice guidelines. This trial was approved by the local ethics committee prior to enrollment (entry no. 445/2005) and registered at ClinicalTrials. gov (new ID: NCT03198533, originally registered in 2007 as a sub-study of NCT00436982). Informed consent was obtained from all patients. Stryker provided funds in support of the costs associated with RSA radiographs and extra clinical follow-up examinations. The sponsor did not take any part in the design, conduct, analysis, and interpretations stated in the final manuscript.
12444 Hamersveld D.indd 427
Allocated to PA-coated TKA (n = 30): – received allocated treatment, 29 – excluded day of surgery due to cerebral infarct, 1
Results 60 patients were randomized, of which 1 patient in each group was excluded on the day of surgery. Baseline characteristics were similar (Table 1). During follow-up, 3 knees were revised (2 infections and 1 loosening, see adverse events), 7 patients died, 14 patients refused further follow-up due to the burden of coming to the clinic at high age or moving out of the region, and 2 patients could not be analyzed reliably for technical reasons (Figure 1). Of the 2 cases with unreliable measurements, 1 had insufficient bone markers available causing high condition numbers (up to 216) after 1 year; reversed RSA migration results showed stable minor translations, and this patient had no knee complaints and no signs of loosening on conventional radiographs. The other case had unreliable measurements after 5 years (condition number of 135) due to over-projection of the femoral component and this component was revised after 10 years for mechanical failure (see below). RSA migration measurements PA-coated components stabilized earlier as compared with uncoated components, resulting in a lower mean migration at 10 years: 0.94 mm (CI 0.72–1.2) for the PA-coated group and 1.7 mm (CI 1.4–2.1) for the uncoated group (p < 0.001). Over time, differences in migration between groups were seen in almost any direction (Table 2). Most of the difference in migration was already seen at 1 year, as the PA-coated components stabilized within the first 3 months while the uncoated components stabilized after 1 year of follow-up (Figure 2). Post hoc analysis showed that when using different baselines,
Acta Orthopaedica 2018; 89 (4): 425–430
Table 2. RSA migration measurements in absolute mm or degrees (95% CI) (log-transformed values are back-transformed in the original scale) 1 year Translations (mm): Transverse 0.4 (0.28–0.49) Longitudinal 0.5 (0.42–0.67) Sagittal 0.5 (0.42–0.68) Rotations (º): Transverse 1.1 (0.79–1.37) Longitudinal 0.7 (0.50–0.82) Sagittal 0.6 (0.49–0.83) MTPM (mm) 1.5 (1.24–1.88) a p-values
0.3 (0.21–0.40) 0.3 (0.18–0.39) 0.3 (0.17–0.38)
0.4 (0.33–0.54) 0.5 (0.41–0.66) 0.7 (0.52–0.80)
0.3 (0.19–0.38) 0.3 (0.17–0.38) 0.2 (0.13–0.34)
0.4 (0.30–0.54) 0.5 (0.41–0.69) 0.7 (0.55–0.87)
0.4 (0.24–0.48) 0.3 (0.17–0.41) 0.3 (0.17–0.42)
0.2 < 0.001 < 0.001
0.6 (0.39–0.83) 0.3 (0.17–0.42) 0.4 (0.30–0.59) 0.9 (0.71–1.18)
1.2 (0.92–1.55) 0.8 (0.67–1.04) 0.8 (0.66–1.04) 1.7 (1.35–2.01)
0.6 (0.35–0.80) 0.3 (0.18–0.44) 0.5 (0.36–0.67) 0.9 (0.70–1.17)
1.3 (0.94–1.64) 1.0 (0.78–1.21) 0.8 (0.57–0.97) 1.7 (1.41–2.08)
0.6 (0.33–0.83) 0.3 (0.14–0.43) 0.5 (0.30–0.65) 0.9 (0.72–1.19)
< 0.001 < 0.001 0.004 < 0.001
stated in this column indicate testing the between-group mean differences with time over the entire postoperative follow-up period.
MTPM (mm) 2.5
Years after index operation MTPM (mm) 9 8
Uncoated PA-coated Loose (uncoated) Stabilized (uncoated) Stabililzed (PA-coated) Revised (uncoated)
7 6 5 4 3 2 1 0 0
Years after index operation
Figure 2. Maximum total point motion (back-transformed in the original scale in mm) during 10 years of follow-up: (top) the mean and 95% CI for the groups and (bottom) the mean and 95% CI for the groups and separate lines for the components showing continuous migration in the second postoperative year (in green the stabilized components after 2 years, in dashed brown the components failing to stabilize after 2 years and suspected for aseptic loosening, and in solid brown the revised component).
12444 Hamersveld D.indd 428
no statistically significant between-group mean differences were seen from 1 year onwards (p = 0.1) and from 2 years onwards (p = 0.7) (Table 3, see Supplementary data). Between 1 and 2 years of follow-up, 7 uncoated components showed more than 0.2 mm MTPM and were suspected for loosening, compared with 1 in the PA-coated group. 5 of the 7 uncoated components stabilized, while 2 did not: 1 (clinically still asymptomatic patient) showed continuous migration of 0.14 mm/year up to 10-year follow-up (Figure 3, see Supplementary data) and 1 showed continuous migration of 0.11 mm/year up to 7-year follow-up who, despite having progressive complaints, refused to visit for 10-year follow-up (Figure 4, see Supplementary data). 1 uncoated component that was initially classified as loose was lost to follow-up but showed full stabilization at final (5-year) follow-up. 1 uncoated component was revised after 10 years as the patient had increasing pain and instability due to mechanical failure (see below). The PA-coated component initially classified as loose was stabilized at 5-year follow-up. None of the PA-coated components classified as stable showed continuous migration at any follow-up measurement beyond 2 years. Clinical results and adverse events There were no statistically significant between-group differences with respect to improvement in knee flexion, extension, both KSS scores, and 4 of 5 KOOS subscales. The KOOS subscale quality of life improved equally between groups up to 5-year follow-up (p = 1.0), but substantially decreased in the PA-coated group between 5 and 10 years, resulting in a between-group mean difference after 10-year follow-up (p = 0.02) (Table 4, see Supplementary data). 3 patients (all with uncoated components) underwent revision surgery; the first due to an early prosthetic joint infection (at 3 months), the second due to a late infection (at 1 year) and the third (at 10 years) due to mechanical failure (complaints of pain and instability, posteromedial wear of the insert, and tibial component loosening was found during revision surgery) (Figure 5, see Supplementary data). 1 patient (randomized to
Acta Orthopaedica 2018; 89 (4): 425–430
the uncoated group) received a cemented implant due to an intraoperative fissure of the proximal tibia and was excluded. 1 patient (randomized to the PA-coated group) was transferred on the day of surgery to another hospital to receive appropriate treatment after a cerebral infarct and was also excluded.
Discussion Our results show that the short-term effect of Peri-Apatite™ on uncemented tibial component migration is sustained over time, resulting in less mean migration and absence of components with continuous migration after 10 years. As shown in other long-term RSA studies, stabilization of uncemented tibial components can be achieved despite high initial migration (Pijls et al. 2012a, Henricson and Nilsson 2016). In the present long-term study, 6 individual components stabilized even after 2 years. Only 2 uncoated components migrated continuously throughout follow-up. Given that most prostheses stabilized within 2 years, the mean migration from 1 year onwards was not statistically significantly different between groups as confirmed in the post hoc analysis. Both “excessive” initial migration in the first year (of more than 0.5 mm for a group of patients) and continuous migration after 1 year (0.2 mm in the second postoperative year for an individual patient) are associated with, and frequently used as predictors for, aseptic loosening (Ryd et al. 1995, Pijls et al. 2012b). These studies, however, combined prostheses that rely on primary fixation (cemented and uncemented with screws) and those that rely on secondary biological fixation (uncemented) to evaluate the migration thresholds for prostheses suspected for loosening. Several studies have shown that the typical migration pattern of an uncemented component differs from that of a primary fixated component, especially during the first 2 years (Nilsson et al. 2006, Dunbar et al. 2009, Pijls et al. 2012a, Wilson et al. 2012, Henricson and Nilsson 2016, van Hamersveld et al. 2017). We therefore question whether the current migration thresholds are justified for uncemented prostheses, especially for designs without biological mediators (e.g., hydroxyapatite or highly porous metal) to enhance bone ingrowth, and can be used to classify such implants being loose in RSA studies with only 2 years of follow-up. In our study, 1 TKA was revised at 10-year follow-up due to progressive pain and function impairment due to mechanical failure. Posteromedial polyethylene wear and tibial component loosening was found during revision surgery (Figure 5, see Supplementary data). This patient was not flagged as “loose” through RSA measurements as MTPM values were stable up to 5 years of follow-up but further follow-up measurements were unreliable due to high condition numbers (solid red line in Figure 2). The exact failure mechanism is unknown. Causal factors of posteromedial failure include overloading the medial compartment and malalignment of the femoral component, increasing posteromedial peak contact
12444 Hamersveld D.indd 429
stresses (Morra et al. 2003). Some authors have reported that by cross-linking the polyethylene the fatigue crack propagation resistance is decreased, especially in TKA (Bradford et al. 2004, Ries 2005). However, later reports of fatigue failure are rare and mainly limited to tibial post fractures in posterior-stabilized knees, suggesting this mechanism is unlikely to account for failure in our patient (Jung et al. 2008, Yu et al. 2016). Although all other subscales of the KOOS score were similar between uncoated and PA-coated components, we did observe a statistically significant difference in the quality of life subscale after 10 years of follow-up. Similar to the occurrence of both the infection cases and the revised case due to mechanical failure (which could all have occurred in either group), the statistical difference in quality of life is most likely a spurious finding and not related to the implant type. Nevertheless, we continue to monitor these patients to observe whether any adverse effect of the given treatment occurs. Several limitations can be noted. First, a high number of patients were lost to follow-up. Consequently, only 16 patients were available for analysis in the PA-coated group at 10-year follow-up. However, results from the linear mixed-effects model are based on all measurements, not only on remaining patients at final follow-up. Furthermore, as most implants of the lost patients appeared to have stabilized, it is unlikely that the observed results would substantially differ from those presented if patients had continued follow-up. Results of the secondary clinical outcomes should, however, be regarded as exploratory due to the limited sample size and the lower accuracy and precision of these outcome measurements. Second, it remains unknown why 6 components stabilized while 2 did not. Logically, the magnitude of component migration plays a role in preventing the onset of a prosthesis-settling phase. However, other (baseline) factors that may predict high risk patients cannot be found without performing “one-variableat-a-time” subgroup analyses, which are likely both underpowered and produce false-positive results due to multiple comparisons (Kent et al. 2016). We therefore refrained from performing such subgroup analyses. Third, a strict intentionto-treat analysis requires all randomized patients to be analyzed, which was not the case for the 2 excluded patients on the day of surgery. These 2 patients were excluded from further follow-up measurements at the time, hence no data were available for analysis. Furthermore, not receiving the studied intervention can be a legitimate reason for patient exclusion without risking bias, even in an intention-to-treat trial (Fergusson et al. 2002). In summary, the typical biphasic migration pattern of uncemented implants was seen in both the uncoated group and the PA-coated group, but the latter showed statistically significantly less mean migration and absence of components with continuous migration at 10-year follow-up. When evaluating uncemented prostheses, especially those without biological mediators to enhance bone ingrowth, the initial migration
phase is longer than in cemented components and can last over 2 years. With such prostheses, short-term RSA cut-off values to determine the risk of failure seem of limited value. Evaluation should thus be based on longer follow-up data and include mean migration results as well as individual component migration results. Supplementary data Tables 3 and 4 and Figures 3–5 are available in the online version of this article, http://dx.doi.org/ 10.1080/17453674. 2018.1469223
The study was designed by STL. Surgeries were performed by STL and 2 other colleagues. Data collection and RSA analysis were performed by KH. Statistical analysis was done by KH and PM. KH, PM, RN, and STL interpreted the data and wrote the initial draft manuscript. KH, PM, RN, and STL critically revised and approved the manuscript.
Acta thanks Anders Henricson and Leif Ryd for help with peer review of this study.
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. Bradford L, Baker D, Ries M D, Pruitt L A. Fatigue crack propagation resistance of highly crosslinked polyethylene. Clin Orthop Relat Res 2004; (429): 68-72. Carlsson A, Bjorkman A, Besjakov J, Onsten I. Cemented tibial component fixation performs better than cementless fixation: a randomized radiostereometric study comparing porous-coated, hydroxyapatite-coated and cemented tibial components over 5 years. Acta Orthop 2005; 76(3): 362-9. Derbyshire B, Prescott R J, Porter M L. Notes on the use and interpretation of radiostereometric analysis. Acta Orthop 2009; 80(1): 124-30. Dunbar M J, Wilson D A, Hennigar A W, Amirault J D, Gross M, Reardon G P. Fixation of a trabecular metal knee arthroplasty component: a prospective randomized study. J Bone Joint Surg Am 2009; 91(7): 1578-86. Fergusson D, Aaron S D, Guyatt G, Hebert P. Post-randomisation exclusions: the intention to treat principle and excluding patients from analysis. BMJ 2002; 325(7365): 652-4. Hansson U, Ryd L, Toksvig-Larsen S. A randomised RSA study of PeriApatite HA coating of a total knee prosthesis. Knee 2008; 15(3): 211-16. Henricson A, Nilsson K G. Trabecular metal tibial knee component still stable at 10 years. Acta Orthop 2016: 87(5): 504-10. ISO 16087:2013(E). Implants for surgery: roentgen stereophotogrammetric analysis for the assessment of migration of orthopaedic implants. Geneva: International Organization for Standardization; 2013. Jung K A, Lee S C, Hwang S H, Kim S M. Fracture of a second-generation highly cross-linked UHMWPE tibial post in a posterior-stabilized Scorpio knee system. Orthopedics 2008; 31(11): 1137. Kent D M, Nelson J, Dahabreh I J, Rothwell P M, Altman D G, Hayward R A. Risk and treatment effect heterogeneity: re-analysis of individual participant data from 32 large clinical trials. Int J Epidemiol 2016; 45(6): 2075-88.
12444 Hamersveld D.indd 430
Acta Orthopaedica 2018; 89 (4): 425–430
Molt M, Toksvig-Larsen S. Peri-Apatite™ enhances prosthetic fixation in TKA: a prospective randomised RSA atudy. J Arthritis 2014; 3(3): 134. Molt M, Ryd L, Toksvig-Larsen S. A randomized RSA study concentrating especially on continuous migration. Acta Orthop 2016; 87(3): 26-7. Morra E A, Postak P D, Plaxton N A, Greenwald A S. The effects of external torque on polyethylene tibial insert damage patterns. Clin Orthop Relat Res 2003; (410): 90-100. 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 Br 1998; 80(2): 267-72. Nelissen R G, Valstar E R, Rozing P M. The effect of hydroxyapatite on the micromotion of total knee prostheses: a prospective, randomized, doubleblind study. J Bone Joint Surg Am 1998; 80(11): 1665-72. Nilsson K G, Henricson A, Norgren B, Dalen T. Uncemented HA-coated implant is the optimum fixation for TKA in the young patient. Clin Orthop Relat Res 2006; 448: 129-39. Overgaard S, Bromose U, Lind M, Bunger C, Soballe K. The influence of crystallinity of the hydroxyapatite coating on the fixation of implants: mechanical and histomorphometric results. J Bone Joint Surg Br 1999; 81(4): 725-31. Pijls B G, Valstar E R, Kaptein B L, Fiocco M, Nelissen R G. The beneficial effect of hydroxyapatite lasts: a randomized radiostereometric trial comparing hydroxyapatite-coated, uncoated, and cemented tibial components for up to 16 years. Acta Orthop 2012a; 83(2): 135-41. Pijls B G, Valstar E R, Nouta K A, Plevier J W, Fiocco M, Middeldorp S, et al. Early migration of tibial components is associated with late revision: a systematic review and meta-analysis of 21,000 knee arthroplasties. Acta Orthop 2012b; 83(6): 614-24. Ries M D. Highly cross-linked polyethylene: the debate is over—in opposition. J Arthroplasty 2005; 20(4 Suppl. 2): 59-62. Ryd L, Albrektsson B E, Carlsson L, Dansgard F, Herberts P, Lindstrand A, et al. Roentgen stereophotogrammetric analysis as a predictor of mechanical loosening of knee prostheses. J Bone Joint Surg Br 1995; 77(3): 377-83. Serekian P. Hydroxyapatite: from plasma spray to electrochemical deposition. In: (Epinette J, Manley MT, eds.) Fifteen years of clinical experience with hydroxyapatite coatings in joint arthroplasty. Dordrecht: Springer; 2004. p. 29-33. Therbo M, Lund B, Jensen K E, Schroder H M. Effect of bioactive coating of the tibial component on migration pattern in uncemented total knee arthroplasty: a randomized RSA study of 14 knees presented according to new RSA guidelines. J Orthop Traumatol 2008; 9(2): 63-7. Valstar E R, Gill R, Ryd L, Flivik G, Borlin N, Karrholm J. Guidelines for standardization of radiostereometry (RSA) of implants. Acta Orthop 2005; 76(4): 563-72. van der Linde M J, Garling E H, Valstar E R, Tonino A J, Nelissen R G. Periapatite may not improve micromotion of knee prostheses in rheumatoid arthritis. Clin Orthop Relat Res 2006; 448: 122-8. van Hamersveld K T, Marang-van de Mheen P J, Tsonaka R, Valstar E R, Toksvig-Larsen S. Fixation and clinical outcome of uncemented peri-apatite-coated versus cemented total knee arthroplasty: five-year follow-up of a randomised controlled trial using radiostereometric analysis (RSA). Bone Joint J 2017; 99-B(11): 1467-76. Wilson D A, Richardson G, Hennigar A W, Dunbar M J. Continued stabilization of trabecular metal tibial monoblock total knee arthroplasty components at 5 years measured with radiostereometric analysis. Acta Orthop 2012; 83(1): 36-40. Yu B F, Yang G J, Wang W L, Zhang L, Lin X P. Cross-linked versus conventional polyethylene for total knee arthroplasty: a meta-analysis. J Orthop Surg Res 2016; 11: 39.
Acta Orthopaedica 2018; 89 (4): 431–436
Poor outcome after a surgically treated chondral injury on the medial femoral condyle: early evaluation with dGEMRIC and 17-year radiographic and clinical follow-up in 16 knees Jon TJÖRNSTRAND 1, Paul NEUMAN 2, Björn LUNDIN 3, Jonas SVENSSON 4, Leif E DAHLBERG 1, and Carl Johan TIDERIUS 1
1 Department 3 Department
of Orthopaedics, Clinical Sciences, Lund, Lund University; 2 Department of Orthopaedics, Clinical Sciences, Malmö, Lund University; of Radiology, Clinical Sciences, Lund, Lund University; 4 Department of Medical Radiographic Physics, Clinical Sciences, Malmö, Lund University, Sweden Correspondence: email@example.com Submitted 2017-12-21. Accepted 2018-04-04.
Background and purpose — The optimal treatment for traumatic cartilage injuries remains unknown. Contrastenhanced MRI of cartilage (dGEMRIC) evaluates cartilage quality and a low dGEMRIC index may predict radiographic osteoarthritis (OA). The purpose of this study was (a) to explore the results 17 years after surgical treatment of an isolated cartilage knee injury and (b) to evaluate the predictive value of dGEMRIC. Patients and methods — 16 knees with an isolated traumatic cartilage injury of the medial femoral condyle had cartilage repair surgery either by microfracture or autologous cartilage implantation. dGEMRIC of the injured knee was performed 2 years after surgery and radiographic examinations were performed 17 years after the operation. Results — Radiographic OA was present in 12 of 16 knees. Irrespective of surgical method, the dGEMRIC index was lower in repair tissue compared with adjacent cartilage in the medial compartment, 237 ms vs. 312 ms (p < 0.001), which in turn had lower value than in the non-injured lateral cartilage, 312 ms vs. 354 ms (p < 0.008). The dGEMRIC index in the cartilage adjacent to the repair tissue correlated negatively with radiographic osteophyte score, r = –0.75 (p = 0.03). Interpretation — A traumatic cartilage injury is associated with a high prevalence of OA after 17 years. The low dGEMRIC index in the repair tissue 2 years postoperatively indicates fibrocartilage of low quality. The negative correlation between the dGEMRIC index in the adjacent cartilage and future OA suggests that the quality of the surrounding cartilage influences outcome after cartilage repair surgery.
Cartilage injuries, with or without complicating ligamentous/ meniscal injury, often occur after a twisting/compression trauma during sports activities. Cartilage has a limited healing potential with a complex structure with low chondrocyte density and avascularity. The best treatment for chondral defects remains controversial, despite decades of efforts (Hunziker et al. 2015). Microfracture (MFX) and autologous chondrocyte implantation (ACI) are the most commonly used techniques. In MFX, mesenchymal stem cells are recruited from the bone marrow by drilling or punching multiple holes in the subchondral bone plate of the cartilage lesion. First described in 1959 and having subsequently evolved, this is presently the most used technique (Pridie 1959, Steadman et al. 2001). ACI is a more technically demanding procedure, introduced in 1994 (Brittberg et al. 1994). This first-generation ACI includes two surgical procedures with arthroscopic harvesting of cartilage at the first operation for in vitro cultivation of chondrocytes. At the second operation, 3 weeks later, the expanded chondrocytes are injected under a periosteum flap that is sutured over the cartilage defect. For both the MFX and the ACI technique, good 2–8 year results have been reported in younger patients with a welldefined traumatic injury. No clear difference has been observed between these treatments regarding failure rate or clinical outcome (PROMS) in randomized controlled trials (RCT) (Kraeutler et al. 2018). However, in a longer perspective, > 10 years, several studies have shown that approximately half of the patients have developed radiographic osteoarthritis (OA), after both MFX and ACI treatment (Gobbi et al. 2014, Martinčič et al. 2014, Knutsen et al. 2016).
© 2018 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 10.1080/17453674.2018.1481304
12423 TJO RNSTRAND D.indd 431
OA is a slowly developing degenerative disease on the timescale of 10 to 20 years. In the very early stages of the disease, molecular and cellular processes decrease cartilage quality but symptoms or radiographic changes may not yet be present. Delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) is a non-invasive method to assess and monitor such early degenerative changes in vivo. The method is based on the principle that a negatively charged contrast agent distributes into articular cartilage in an inverse relationship to the concentration of negatively charged glucosaminoglycans (GAG). In several clinical studies, the method has demonstrated a predictive potential for subsequent OA development, both in the hip (Cunningham et al. 2006, Palmer et al. 2017) and the knee (Owman et al. 2008, 2014). The aims of the present study were to: (a) study the radiographic progression to OA 17 years after surgical treatment of a traumatic chondral injury on the medial femoral condyle, and (b) to evaluate if dGEMRIC has a predictive value in terms of future OA development.
Acta Orthopaedica 2018; 89 (4): 431–436
Patient data, dGEMRIC index at 27 months, and radiographic outcome at 17 years A
1 2 3
M M M
36 36 37
ACI ACI ACI
150 200 300
346 367 334
287 362 315
259 213 253
37 45 36 36 31 37 30 34 39 40 38 38 30 37.2 (4.7)
MFX MFX ACI ACI ACI MFX ACI MFX MFX ACI ACI MFX ACI
300 400 225 600 600 200 200 250 150 150 150 180 120 261 (151)
343 311 330 381 269 446 385 376 382 271 347 327 453 354 (51)
245 243 370 306 292 306 309 281 297 332 293 325 428 312 (46)
241 218 249 210 241 276 250 231 216 263 239 207 235 237 (21)
4 F 5 M 6 F 7 M 8 M 9 F 10 F 11 M 12 F 13 F 14 M 15 M 16 M Mean (SD)
HTO at 3 years, presently scheduled for TKA No radiographic OA, Clinical OA < 2 years conversion to mosaic, HTO at 16 years, presently considering TKA Radiographic OA, clinical OA Radiographic OA, clinical OA UKA at 6 years Radiographic OA, clinical OA TKA at 14 years UKA at 7 years No radiographic OA, clinical OA Radiographic OA at 9 years Radiographic OA, clinical OA TKA at 13 years No radiographic OA, no OA symptoms Radiographic OA at 4 years No radiographic OA, no OA symptoms
A Studied knee number B Sex C Age at index operation D Cartilage repair procedure E Size of lesion, mm2 F dGEMRIC index lateral at 27 months G dGEMRIC index medial (adjacent to repair) at 27 months H dGEMRIC index repair tissue at 27 months I Outcome: radiographic OA, OA surgery, or clinical OA by KOOS score Radiographic OA was defined using the OARSI score and clinical OA defined using the Knee Osteoarthritis Outcome Score (KOOS). 1 patient had bilateral operations (knees number 12 and 13). 2 patients (knees number 11 and 15) did not participate in the 17-year radiographic followup but had radiographs recorded 4 and 9 years postoperatively. Knees that had undergone OA surgery (high tibial osteotomy (HTO), unicompartmental knee arthroplasty (UKA) or total knee arthroplasty (TKA)) were dichotomized as OA diagnosis but excluded from analysis of outcome measures and radiographic change.
Patients and methods Patients Between 1997 and 2000, 16 knees in 15 patients were treated surgically due to a symptomatic isolated traumatic cartilage injury on the medial femoral condyle. Patients had no symptoms before the injury of the affected knee. The patients were initially included in an RCT of ACI vs. MFX that was designed for a larger number of patients. Due to logistical challenges and lack of patients, the study was not completed and the preliminary results have not been published. Exclusion criteria were: radiographic evidence of OA or evidence of cartilage degeneration at arthroscopy; patients with concomitant disease, injury, or malalignment. The group included 10 men and 5 women (1 woman had bilateral injuries) with a median age at index operation of 37 years (30–47). The mean size of the traumatic chondral injury was 261 mm2 (120–600) (Table). Patients were randomized to either MFX or ACI treatment. However, 2 patients in each group had to be reoperated within
12423 TJO RNSTRAND D.indd 432
the first 2 years (Figure 1). This resulted in 9 knees finally treated with ACI and 6 knees treated with MFX drilling. One patient initially treated with MFX was reoperated with mosaicoplasty as a salvage procedure. In that patient, cartilagebone plugs were harvested from unloaded joint regions and implanted to the injury site. The medical records, including surgical reports and archived radiographs, were studied in all patients until the 17-year follow-up (Figure 3 and Table). dGEMRIC values were compared (Figure 4) to a cohort of 19 asymptomatic individuals (mean age 24 years) that previously had been investigated with an identical MRI protocol (Tiderius et al. 2001). Surgical procedures At the first operation, the cartilage injury was verified with arthroscopy and the treatment randomized. Knees randomized to ACI had cartilage harvested from unloaded cartilage on the upper medial femoral condyle (Brittberg et al. 1994). The donor site was remote from the cartilage lesion. The second surgery, 2–3 weeks later, was a mini-arthrotomy with debridement of the cartilage lesion to stable edges in all patients. The
Acta Orthopaedica 2018; 89 (4): 431–436
Randomized to ACI n = 10
Randomized to MFX n=6 Graft failure Re-repair to ACI Graft failure Re-repair to MFX
Last operation ACI dGEMRIC 30 months n=9
No radiographic OA n=4
Graft failure Re-repair to mosaicoplasty
Last operation MFX dGEMRIC 24 months n=6
Radiographic OA n=6
Last operation mosaic dGEMRIC 12 months n=1
Figure 2. Illustration of how the regions of interest (ROIs) for dGEMRIC were drawn. All ROIs included full-thickness cartilage. In the lateral compartment (A), the ROI was drawn from the center of the tibial plateau to the rear insertion of the meniscus (red), according to a standardized protocol (Tiderius et al. 2004b). In the medial compartment (B), one ROI included the repair tissue (blue) and one ROI (yellow) the remaining weight-bearing cartilage to the rear insertion of the meniscus.
Joint HTO replacement Major joint surgery n=6
Figure 1. Flow-chart of treatment and follow-up of all 16 knees (in 15 patients).
knees randomized to ACI had a periosteal flap sutured over the lesion and sealed by fibrin glue, under which the in-vitro expanded chondrocytes were injected (Brittberg et al. 1994). The knees randomized to MFX had the lesion drilled with a ∅2 mm drill, hole centers spaced 6 mm apart. The protocol for postoperative management of ACI described by Brittberg et al. (1994) was followed for both groups consisting of 6 weeks of unloading followed by 6 weeks of progressive weight-bearing and detailed supervised physiotherapy. dGEMRIC The dGEMRIC investigations were performed on average 2 years after the cartilage repair procedure using a standard 1.5 T MRI system with a dedicated knee coil (Magnetom Vision; Siemens Medical Solutions, Erlangen, Germany). Gd-DTPA2(Magnevist®, Bayer Schering Pharma AG, Berlin, Germany) at 0.3 mmol/kg body weight dosage was injected intravenously. To optimize the distribution of Gd-DTPA2- into the cartilage, the patients exercised by walking up and down stairs for 10 min, starting 5 min after the injection. Post-contrast MR imaging was performed 2 hours after the injection according to a standardized protocol (Tiderius et al. 2001). Two sagittal slices covering the central parts of the weight-bearing lateral and medial femoral cartilage respectively were acquired using sets of 6 turbo inversion recovery images with different inversion times (TI = 50, 100, 200, 400, 800, and 1600 ms), from which the T1 relaxation time was subsequently calculated. Other imaging parameters were: TR = 3000 ms, TE = 15 ms, turbofactor 7, field of view (FOV) 120×120 mm2, matrix = 256×256, slice thickness = 3 mm. Regions of interest (ROIs) were drawn (Figure 2) in the weight-bearing central parts of the lateral and medial femoral
12423 TJO RNSTRAND D.indd 433
cartilage, respectively. In the medial compartment, 1 ROI covered the area of cartilage repair tissue and 1 ROI was drawn in adjacent, non-injured weight-bearing cartilage. In the lateral compartment, the ROI was drawn between the center of the tibia plateau to the rear insertion of the meniscus according to a previously validated protocol (Tiderius et al. 2004b). Results are presented as mean T1 ms of each ROI (the dGEMRIC index). Radiography At median 17 years (15–19) after the cartilage repair surgery, standardized weight-bearing radiographs in 20° flexion of both knees were obtained. Blinded to clinical presentation and to the type of cartilage repair procedure, assessment of the ante-posterior radiographs was performed by 2 of the authors independently: an orthopedic surgeon specialized in joint replacement (JT) and a senior radiologist specialized in skeletal radiology (BL). In cases of discrepancy the images were reassessed by the 2 investigators together and consensus was reached. The OARSI atlas (Altman and Gold 2007) was used for the medial and lateral compartments respectively, grading radiographic change on a 4-point scale for joint space narrowing (JSN) (0–3, 0 = no evidence of JSN) and marginal osteophytes of femoral and tibial condyles (0–3 each, 0 = no bony change). Dichotomization for diagnosis of radiographic OA was defined as any of the following criteria fulfilled in either of the 2 tibiofemoral compartments: JSN ≥ 2, osteophyte score ≥ 2, or JSN grade 1 in combination with osteophyte grade 1 in the same compartment. This definition (Englund et al. 2003) approximates grade 2 knee OA based on the Kellgren–Lawrence scale. 2 patients could not partake in the radiographic examination. However, both these patients had previous visits to an orthopedic surgeon due to knee pain respectively 4 and 9 years postoperatively. Weightbearing radiographs from these visits demonstrated OA by the above criteria.
Acta Orthopaedica 2018; 89 (4): 431–436
Number of knees without OA
dGEMRIC index (T1Gd ms)
dGEMRIC index (T1 Gd ms)
16 500 14
p = 0.008
300 p < 0.001
2 0 0
Years from cartilage repair surgery to OA
Figure 3. Temporal assessment of OA after surgical repair of a chondral injury on the medial femoral condyle. OA development was defined as either high tibial osteotomy, arthroplasty, or radiographic OA. Time points of OA diagnosis are from surgery date (HTO or joint replacement) or date of radiographic OA either due to radiographic evaluation of clinical symptoms in the intermediate time or by radiographs at the 17-year follow-up. The initial number of knees (n = 16) without OA already starts to decrease two years after surgery. At the end of the study period (17 years), only 4 knees lack radiographic OA.
Lateral non-injured cartilage
Medial cartilage adjacent to injury
Figure 4. dGEMRIC index of the 3 investigated ROIs in each knee (n = 16): the lateral femoral condyle (354 ms SD 51), the medial cartilage adjacent to the cartilage lesion (312 ms, SD 46) and the repair tissue (237 ms, SD 20). The dGEMRIC index was lower in repair tissue vs. adjacent cartilage in the medial femoral cartilage (p < 0.001). The dGEMRIC index was higher in the uninjured lateral femoral cartilage than in the medial cartilage adjacent to the cartilage lesion (p < 0.008). Horizontal black bars are mean values. For comparison, the red solid (mean) and dashed lines (SD) represent the dGEMRIC index in healthy volunteers previously investigated with an identical protocol by our group (Tiderius et al. 2001).
PROMS Patient-related outcome measures (PROMS) were completed at the 17-year follow-up by self-administered pen on paper forms for VAS, Lysholm, and KOOS. The algorithm based on the KOOS score described by Englund et al. (2003) was used to dichotomize for clinical OA. Patients who had been treated for OA with osteotomy or arthroplasty were excluded from PROMS analysis. Statistics SPSS 25 Statistics for Windows (IBM Corp, Armonk, NY, USA) and SigmaPlot 11.0 (Systat Software, San Jose, CA) was used for statistical analysis. Despite the fact that data from 1 bilateral operation are not independent, we included both knees in that patient for analysis. After testing for normal distribution (Shapiro–Wilk) and equal variance (Levene’s mean test), the Student t-test was used for continuous variables. A paired test was used for regional measurements in the same knee; a non-paired test was used in all other instances, and 2-tailed distribution was assumed in all tests. Spearman’s Rho was used for correlation of ordinal data and continuous variables. Fisher’s exact test was used to compare the distribution in cases of 2 dichotomous variables. The statistical power was low due to the few patients eligible for this study. Ethics, funding, and potential conflicts of interest The study was approved by the ERB at Lund University (Etik-
12423 TJO RNSTRAND D.indd 434
Cartilage repair tissue
Figure 5. The dGEMRIC index of medial adjacent cartilage (blue) correlated negatively with radiographic osteophyte score at 17 years, r = –0.75 (p = 0.03). A similar trend, but not statistically significant, was found in the lateral compartment (green), r = –0.60 (p = 0.1). Note that the 6 knees that had already had surgery for OA (HTO or arthroplasty) were excluded from this correlation analysis as were the 2 knees that had only early follow-up radiographs.
prövningsnämnden #EPN:2014/752, LU#73-96 and LU#65100), the Radiation Protection Committee (Strålskyddskommiten #SSFo2014-050), and the Image Research Committee (BOF053). The study was supported by grants from the Regional Research Council of Region Skåne, Governmental Funding of Clinical Research within National Health Service (ALF), Erik and Angelica Sparres forskningsstiftelse and the Johan and Greta Kock Foundation. No conflicts of interest declared.
Results OA, radiographic, and symptomatic All 16 knees could be assessed for radiographic OA (rOA) after mean 17 years follow-up (Figure 1 and Table). 6 knees had received OA surgery; 2 had high tibial osteotomy, 2 had unicompartmental medial knee arthroplasty and 2 had total knee arthroplasty. 6 knees had radiographic OA based on OARSI scores, and 4 knees had no radiographic OA. Thus 12 of 16 knees had failed by progressing to OA (Figure 3). The KOOS score was indicative of OA (Englund et el. 2003) in all but 2 knees with a normal age-matched KOOS score and no radiographic OA (Table). All 4 knees that needed a second cartilage repair procedure progressed to OA. All MFX-treated knees and 6 of 10 ACI knees developed radiographic OA (p = 0.2).
Acta Orthopaedica 2018; 89 (4): 431–436
dGEMRIC All knees were examined with dGEMRIC at median 27 (12– 57) months after the cartilage repair procedure. The 4 early graft failure reoperations were performed 12–15 months prior to the dGEMRIC examination. The mean dGEMRIC index (T1Gd in ms) differed between the 3 ROIs as illustrated in Figure 4. The mean dGEMRIC index was 33% lower in the repair tissue compared with the non-affected lateral femoral cartilage, 237 vs. 354 ms (p < 0.001). In addition, the dGEMRIC index in the cartilage adjacent to the lesion was lower than in the non-affected lateral femoral cartilage, 311 ms (SD 58) vs. 354 ms (SD 51) (p = 0.008). The cartilage adjacent to the repair tissue was higher after ACI compared with MFX surgery with borderline significance: 331 ms (SD 47) vs. 283 ms (SD 33) (p = 0.05). The dGEMRIC index in the medial cartilage adjacent to the lesion correlated negatively with the radiographic osteophyte score in the medial compartment at the 17-year follow-up, r = –0.75 (p = 0.03). A similar trend, although not statistically significant, was found in the lateral compartment, r = –0.60 (p = 0.1) (Figure 5). JSN did not seem to correlate in the medial compartment r = –0.25 (p = 0.5); in the lateral compartment there was no correlation as all knees had zero JSN. There was a trend towards higher dGEMRIC index in the adjacent cartilage of the 4 knees that did not develop radiographic OA compared with OA knees, 348 (SD 61) vs. 300 (SD 35) (p = 0.07). BMI, sex, and size of injury were similar between knees that developed radiographic OA and those that did not. Non-OA patients were on average 4 years younger than patients who developed radiographic OA (p = 0.1).
Discussion The main finding of this study is that most (12/16) of patients that had surgical treatment for a traumatic chondral injury on the medial femoral condyle had developed radiographic OA 17 years after the surgery. In addition, 2 of the 4 patients with no radiographic OA had clinical OA according to the KOOS score. For comparison, the prevalence of radiographic knee OA in the general population of similar age (54–65 years) is between 10% and 23% (Felson et al. 1987, Turkiewicz et al. 2015). The goal of cartilage repair is twofold: to alleviate symptoms and to avoid future OA. Despite decades of research there is no consensus regarding the optimal treatment for traumatic cartilage injuries. In a short- and mid-term perspective, several studies of MFX report good results with a large proportion of patients returning to high levels of activity (Mithoefer et al. 2009, Erggelet and Vavken 2016). However, 10–15 years after the operation, results deteriorate with half of patients having radiographic OA (Gobbi et al. 2014, Knutsen et al. 2016). Theoretically, the reason for failure might be explained by the
12423 TJO RNSTRAND D.indd 435
fact that the repair cartilage after MFX operation lacks collagen II and does not show the zonal organization of hyaline cartilage (Mithoefer et al. 2009, Erggelet and Vavken 2016). By contrast, the ACI technique was designed to yield repair tissue with a hyaline-like structure that potentially also has mechanical properties that resembles healthy cartilage. A recent review of 9 ACI studies (Pareek et al. 2016) with 9–13 years’ followup reported on average 81% successful results, defined by no diagnosed graft failures and good or excellent clinical results. However, the only study that presented radiographic followup (Martinčič et al. 2014) at 10 years postoperatively found OA in half of the cases, i.e., similar to that reported for MFX. An RCT of ACI vs. MFX with a 15-year follow-up of 78 knees (Knutsen et al. 2016) had one-third failures and half of the remaining knees had radiographic OA with no difference between the treatment groups. The small numbers of patients in our study, in combination with the crossover that occurred, hampers a relevant comparison between MFX and ACI treatments. However, it should be pointed out that the 4 patients that did not end up with radiographic OA were all treated with ACI. In addition, dGEMRIC indicated better cartilage quality in the cartilage adjacent to the repair tissue in ACI compared with MFX patients. The evaluation of cartilage quality a few years after the cartilage repair procedure is a major strength of our study. Obviously, the assessment of cartilage status is equally relevant for MFX and ACI cases. dGEMRIC is a validated in-vivo technique to estimate cartilage quality, in particular the glucosaminoglycan content. We found a low dGEMRIC index in the repair tissue both after MFX and ACI indicating fibrocartilage with low GAG content. Several previous studies have shown that a low dGEMRIC index is associated with an increased risk of future radiographic OA, both in the hip (Cunningham et al. 2006, Palmer et al. 2017) and in the knee (Owman et al. 2008, 2014). For example, in middle-aged patients (mean age 50 years) with superficial cartilage fibrillation on the femoral cartilage, a low dGEMRIC index (circa 300 ms) was associated with radiographic OA in two-thirds of patients 6 years after the dGEMRIC investigation (Owman et al. 2008). dGEMRIC as a prognostic tool was suggested also in our study; we found a correlation between a low dGEMRIC index in the cartilage adjacent to the repair tissue and the future prevalence of radiographic OA. This may indicate that the surrounding cartilage should be evaluated at the time of cartilage repair surgery. In support, a clinical MFX study found that visual mild degeneration of surrounding cartilage at the primary operation had a worse outcome at 10–14 years’ followup than patients with normal-appearing cartilage (Solheim et al. 2016). Importantly, in our study, we do not know if low cartilage quality was present already at the time of surgery, or if it developed between surgery and the dGEMRIC investigation, approximately 2 years later. Also, the non-injured lateral cartilage demonstrated lower dGEMRIC values (Figure 4) than previously observed in
healthy volunteers (Tiderius et al. 2001) investigated with an identical protocol. This finding may reflect that cartilage degeneration in the medial compartment affects the whole joint, with GAG loss also in the lateral femoral condyle. Other possible explanations for this finding are reduced loading during rehabilitation as the dGEMRIC index is known to respond to changes in activity level (Roos and Dahlberg 2005), correlate to level of activity (Tiderius et al. 2004a) and correlate to thigh muscle strength (Ericsson et al. 2009). A strength of our study is that all included patients could be assessed regarding OA development and that the study had a strict inclusion criterion: an isolated traumatic chondral injury only on the medial femoral condyle. The main limitation is the small number of patients, resulting in low statistical power, especially regarding the comparison between the 2 surgical methods, MFX and ACI. Furthermore, the value of PROMS was limited because several patients had major joint surgery between the cartilage repair surgery and the 17-year follow-up. In summary, we found a high prevalence of OA at followup 17 years after cartilage repair. There was no evidence of hyaline-like cartilage 2 years after ACI, as demonstrated with a low dGEMRIC index. The negative correlation between the dGEMRIC index in the adjacent cartilage and future OA indicates that dGEMRIC can predict future radiographic OA and that the quality of the surrounding cartilage influences the outcome after cartilage repair surgery. The authors would like to thank Håkan Lövkvist at the Department of Medical Statistics and Epidemiology at Lund University for statistical advice. Study design: JT, LD and CT. Analysis of radiographs: JT and BL. Analysis of MR images: JT, JS and CT. Collection of clinical data: JT. Writing of manuscript: JT, PN, JS, LD and CT. Revision of manuscript: JT and CT. Acta thanks Martin Lind and other anonymous reviewers for help with peer review of this study. Altman R D, Gold G E. Atlas of individual radiographic features in osteoarthritis, revised. Osteoarthritis Cartilage 2007; 15 (Suppl A): A1-56. Brittberg M, Lindahl A, Nilsson A, Ohlsson C, Isaksson O, Peterson L. Treatment of deep cartilage defects in the knee with autologous chondrocyte transplantation. N Engl J Med 1994; 331(14): 889-95. Cunningham T, Jessel R, Zurakowski D, Millis M B, Kim Y J. Delayed gadolinium-enhanced magnetic resonance imaging of cartilage to predict early failure of Bernese periacetabular osteotomy for hip dysplasia. J Bone Joint Surg Am 2006; 88(7): 1540-8. Englund M, Roos E M, Lohmander L S. Impact of type of meniscal tear on radiographic and symptomatic knee osteoarthritis: a sixteen-year followup of meniscectomy with matched controls. Arthritis Rheum 2003; 48(8): 2178-87. Erggelet C, Vavken P. Microfracture for the treatment of cartilage defects in the knee joint: a golden standard? J Clin Orthop Trauma 2016; 7(3): 14552. Ericsson Y B, Tjornstrand J, Tiderius C J, Dahlberg L E. Relationship between cartilage glycosaminoglycan content (assessed with dGEMRIC) and OA risk factors in meniscectomized patients. Osteoarthritis Cartilage 2009; 17(5): 565-70.
12423 TJO RNSTRAND D.indd 436
Acta Orthopaedica 2018; 89 (4): 431–436
Felson D T, Naimark A, Anderson J, Kazis L, Castelli W, Meenan R F. The prevalence of knee osteoarthritis in the elderly: the Framingham Osteoarthritis Study. Arthritis Rheum 1987; 30(8): 914-18. Gobbi A, Karnatzikos G, Kumar A. Long-term results after microfracture treatment for full-thickness knee chondral lesions in athletes. Knee Surg Sports Traumatol Arthrosc 2014; 22(9): 1986-96. Hunziker E B, Lippuner K, Keel M J, Shintani N. An educational review of cartilage repair: precepts & practice—myths & misconceptions—progress & prospects. Osteoarthritis Cartilage 2015; 23(3): 334-50. Knutsen G, Drogset J O, Engebretsen L, Grontvedt T, Ludvigsen T C, Loken S, Solheim E, Strand T, Johansen O. A randomized multicenter trial comparing autologous chondrocyte implantation with microfracture: long-term follow-up at 14 to 15 years. J Bone Joint Surg Am 2016; 98(16): 1332-9. Kraeutler M J, Belk J W, Purcell J M, McCarty E C. Microfracture versus autologous chondrocyte implantation for articular cartilage lesions in the knee: a systematic review of 5-year outcomes. Am J Sports Med 2018; 46(4): 995-9. Martinčič D, Radosavljevič D, Drobnič M. Ten-year clinical and radiographic outcomes after autologous chondrocyte implantation of femoral condyles. Knee Surg Sports Traumatol Arthrosc 2014; 22(6): 1277-83. Mithoefer K, McAdams T, Williams R J, Kreuz P C, Mandelbaum B R. Clinical efficacy of the microfracture technique for articular cartilage repair in the knee. Am J Sports Med 2009; 37(10): 2053-63. Owman H, Tiderius C J, Neuman P, Nyquist F, Dahlberg L E. Association between findings on delayed gadolinium-enhanced magnetic resonance imaging of cartilage and future knee osteoarthritis. Arthritis Rheum 2008; 58(6): 1727-30. Owman H, Ericsson Y B, Englund M, Tiderius C J, Tjornstrand J, Roos E M, Dahlberg L E. Association between delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) and joint space narrowing and osteophytes: a cohort study in patients with partial meniscectomy with 11 years of follow-up. Osteoarthritis Cartilage 2014; 22(10): 1537-41. Palmer A, Fernquest S, Rombach I, Park D, Pollard T, Broomfield J, Bangerter N, Carr A, Glyn-Jones S. Diagnostic and prognostic value of delayed gadolinium enhanced magnetic resonance imaging of cartilage (dGEMRIC) in early osteoarthritis of the hip. Osteoarthritis Cartilage 2017; 25(9): 146877. Pareek A, Carey J L, Reardon P J, Peterson L, Stuart M J, Krych A J. Longterm outcomes after autologous chondrocyte implantation: a systematic review at mean follow-up of 11.4 years. Cartilage 2016; 7(4): 298-308. Pridie K H. A method of resurfacing osteoarthritic knee joint. J Bone Joint Surg Br 1959; 41-B(3): 618-19. Roos E M, Dahlberg L. Positive effects of moderate exercise on glycosaminoglycan content in knee cartilage: a four-month, randomized, controlled trial in patients at risk of osteoarthritis. Arthritis Rheum 2005; 52(11): 3507-14. Solheim E, Hegna J, Inderhaug E, Oyen J, Harlem T, Strand T. Results at 10–14 years after microfracture treatment of articular cartilage defects in the knee. Knee Surg Sports Traumatol Arthrosc 2016; 24(5): 1587-93. Steadman J R, Rodkey W G, Rodrigo J J. Microfracture: surgical technique and rehabilitation to treat chondral defects. Clin Orthop Relat Res 2001; (391): S362-S9. Tiderius C J, Olsson L E, de Verdier H, Leander P, Ekberg O, Dahlberg L. Gd-DTPA2--enhanced MRI of femoral knee cartilage: a dose-response study in healthy volunteers. Magn Reson Med 2001; 46(6): 1067-71. Tiderius C J, Svensson J, Leander P, Ola T, Dahlberg L. dGEMRIC (delayed gadolinium-enhanced MRI of cartilage) indicates adaptive capacity of human knee cartilage. Magn Reson Med 2004a; 51(2): 286-90. Tiderius CJ, Tjornstrand J, Akeson P, Sodersten K, Dahlberg L, Leander P. Delayed gadolinium-enhanced MRI of cartilage (dGEMRIC): intra- and interobserver variability in standardized drawing of regions of interest. Acta Radiol 2004b; 45(6): 628-34. Turkiewicz A, Gerhardsson de Verdier M, Engstrom G, Nilsson P M, Mellstrom C, Lohmander L S, Englund M. Prevalence of knee pain and knee OA in southern Sweden and the proportion that seeks medical care. Rheumatology (Oxford). 2015; 54(5): 827-35.
Acta Orthopaedica 2018; 89 (4): 437–442
Threshold values of ankle dorsiflexion and gross motor function in 60 children with cerebral palsy A cross-sectional study Helle M RASMUSSEN 1, 2, Joachim SVENSSON 2, Maria THORNING 1, Niels W PEDERSEN 1,2, Søren OVERGAARD 1,2, and Anders HOLSGAARD-LARSEN 1,2
of Orthopedic Surgery and Traumatology, Odense University Hospital; 2 Department of Clinical Research, University of Southern Denmark, Odense, Denmark Correspondence: firstname.lastname@example.org Submitted 2017-09-22. Accepted 2018-02-22.
Background and purpose — Threshold values defining 3 categories of passive range of motion are used in the Cerebral Palsy follow-Up Program to guide clinical decisions. The aim of this study was to investigate the threshold values by testing the hypothesis that passive range of motion in ankle dorsiflexion is associated with gross motor function and that function differs between the groups of participants in each category. Patients and methods — We analyzed data from 60 ambulatory children (aged 5–9 years) with spastic cerebral palsy. Outcomes were passive range of motion in ankle dorsiflexion with flexed and extended knee and gross motor function (Gait Deviation Index, Gait Variable Score of the ankle, peak dorsiflexion during gait, 1-minute walk, Gross Motor Function Measure, the Pediatric Quality of Life Inventory Cerebral Palsy Module, and Pediatric Outcomes Data Collection Instrument). Results — Significant (p < 0.05) and moderate correlations were documented for range of motion versus Gait Variable Score of the ankle (r = –0.37 and r = –0.37) and range of motion versus peak dorsiflexion (r = 0.49 and r = 0.55). Differences between the groups formed by the categories were shown for Gait Variable Score of the ankle and peak dorsiflexion (p < 0.05). No other significant correlations or differences between the categories were observed. Interpretation — The results suggest that threshold values for ankle dorsiflexion used in the Cerebral Palsy follow-Up Program are of limited clinical value in assessing overall gross motor function, but may be used to identify deviations in ankle-specific gait function. ■
Muscle contractures and joint deformities are common manifestations of cerebral palsy (CP) (Nordmark et al. 2009). A surveillance program, entitled the Cerebral Palsy follow-Up Program (CPUP), is used to ensure early identification and treatment of deterioration (Rasmussen et al. 2016, AlrikssonSchmidt et al. 2017). As part of the evaluation, the CPUP uses threshold values inspired by traffic light signals on passive range of motion (ROM). The ROM is classified on the basis of threshold values into 3 categories with the following interpretation: “green” means “clear” and that no indication of deterioration is noted, “yellow” indicates that vigilant observation, modified treatment, or initiation of treatment is necessary, and “red” indicates “alert” and that treatment is urgently needed, assuming no contraindications are present (Alriksson-Schmidt et al. 2017). For children on the Gross Motor Function Classification System (GMFCS) levels I–III, the threshold values have been set to ensure that the patient has enough ROM to perform ankle dorsiflexion (DF) in the stance and swing phases of walking (CPOP 2017). The categories are used as easy-tounderstand interpretations of the measurement of ROM and are used to guide decisions about future examinations and interventions (Hagglund and Wagner 2011). A study investigating the association between the 3 categories of ROM in DF and motor function measured with the Functional Mobility Scale found a significant association between the categories (chi-square association, rΦ = –0.27, p = 0.01) in young adults with CP (Brantmark et al. 2015). Despite the use of the threshold values for clinical evaluation in several countries, how these thresholds were identified has never been fully explained and, to our knowledge, they are not evidence-based. Furthermore, their potential relationships
© 2018 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 10.1080/17453674.2018.1456749
12123 Rasmussen D.indd 437
with measures of gross motor function in children with CP have never been established. Thus, the aim of the study was to investigate the threshold values used by CPUP by testing the hypothesis that ROM in DF is associated with gross motor function and that gross motor function differs between the groups of participants in each category. Gross motor function is measured by various methods describing the overall gross motor capacity, the ankle-specific gait capacity, and the use of gross motor skills in everyday life.
Patients and methods We performed a cross-sectional study based on data from the baseline assessment in a randomized controlled trial (Rasmussen et al. 2015b)/NCT02160457. The reporting of the current study conforms to recommendations by the STROBE panel. Participants and setting Patients registered in the Danish version of the CPUP in the Region of Southern Denmark and the North Denmark Region were screened for eligibility and invited to participate. Eligibility criteria have been described in detail previously (Rasmussen et al. 2015b). In brief, eligible participants were children aged 5 to 8 years, diagnosed with spastic CP at GMFCS levels I or II. Children were not eligible if they had received earlier interventions in the form of orthopedic surgery in the previous 52 weeks or injection with botulinum toxin in the 12 weeks prior to the baseline assessment (exclusion criteria). Furthermore, children were excluded if they were unable to demonstrate sufficient cooperation and cognitive understanding to participate in the examination, if they relocated to another region during the trial, or if their parents could not speak and understand Danish. Participants were recruited and data collected from June 2014 until July 2016. Questionnaires were mailed to the parents prior to the examination at the Motion Analysis Laboratory at Odense University Hospital. 6 experienced physiotherapists, who performed each examination in pairs, were involved in data collection. Measurements The baseline assessment methodology has previously been described in detail (Rasmussen et al. 2015b). In short, the assessment consisted of: patient characteristics, a thorough physical examination, 3-D instrumented gait analysis, functional tests of walking and gross motor capacity, and patientreported outcomes concerning the use of gross motor skills in everyday life (see below). Patient characteristics were described using the following measures: sex, age (years), height (meters), weight (kilograms), CP subtype (unilateral or bilateral spastic CP), and classification according to the GMFCS.
12123 Rasmussen D.indd 438
Acta Orthopaedica 2018; 89 (4): 437–442
Table 1. Patient characteristics Girls/boys 21/39 Age (years, months), mean (SD) 6y 10m (1y 3m) Height (meters), mean (SD) 1.23 (0.1) Weight (kg), mean (SD) 24 (6.8) Cerebral palsy subtype and function: CP spastic subtype, unilateral/bilateral 43/17 GMFCS levels I/II 42/18 Passive range of motion: Dorsiflexion with knee 90°, mean (SD) 21 (11) Dorsiflexion with knee 0°, mean (SD) 14 (10) Ankle ROM categories Dorsiflexion (knee 90°), participants, n Red: range 10° 14 Yellow: range 10°–20° 6 Green: range 20° 40 Dorsiflexion (knee 0°), participants, n Red: range 0° 14 Yellow: range 0°–10° 5 Green: range 10° 49 Gait summary measures and peak dorsiflexion: GDI score, mean (SD) 76 (13) GVS ankle, median (IRQ) 7.9 (6.6) Peak dorsiflexion (degrees), mean (SD) 13 (6.5) Gross motor capacity and performance: 64 (11) 1-min walk test (meters), mean (SD) a GMFM, mean (SD) 82 (8.4) Pedsql movement and balance, mean (SD) 75 (19) PODCI transfer and basic mobility, median (IRQ) 93 (15) a Data are available for only 57 participants. Abbreviations: CP: cerebral palsy; GDI: Gait Deviation Index; GMFCS: Gross Motor Function Classification System; GMFM: Gross Motor Function Measure; GVS: Gait Variable Score; IQR: inter-quartile range; Pedsql: the Pediatric Quality of Life Inventory Cerebral Palsy Module; PODCI: Pediatric Outcomes Data Collection Instrument; ROM: range of motion; SD: standard deviation.
Classification according to the 3 categories was obtained from measurements of maximal DF with flexed knee (DF (knee 90°)) and extended knee (DF (knee 0°)) performed by 2 examiners with a goniometer according to the CPUP protocol (2017). The starting positions were supine, hip and knee in 90° of flexion when measuring DF (knee 90°), or with the hip and knee extended when measuring DF (knee 0°). While the hind foot was maintained in neutral to avoid calcaneal valgus or varus, the fixed arm of the goniometer was placed parallel to the front of the tibia and the moving arm at the lateral side of the foot (Nordmark et al. 2009). The threshold values for DF used by the CPUP are outlined in Table 1. The method used for data collection with 3-D instrumented gait analysis has previously been described (Rasmussen et al. 2015b). Briefly, we collected 10 walking trials at a self-selected speed using an 8-camera motion capture system (Vicon MX03, Oxford, UK) and the Plug-in-Gait model (Davis et al. 1991). Vicon Nexus software (version 1.7.1-1.8.5) and Vicon Polygon software (version 3.5.2-4.3) were used for data processing to define gait cycles for 10 trials from each participant. Subsequently, 5 trials with a consistent velocity (±15%) were selected and used for the calculation of the Gait Deviation
Acta Orthopaedica 2018; 89 (4): 437–442
Index (GDI), Gait Variable Score (GVS) of the ankle, and the maximal active DF in the stance phase (Schwartz and Rozumalski 2008, Baker et al. 2009). The GDI and GVS of the ankle are gait summary measures of overall gait function and ankle joint kinematics, providing information on the deviation from a reference group. For reference, we used our own dataset of 30 typically developing children (Rasmussen et al. 2015a). A reliability study performed in our laboratory has documented excellent intra-rater reliability and acceptable agreement for the GDI and fair to good intra-rater reliability and acceptable agreement for the GVS across 2 repeated sessions in children with CP (Rasmussen et al. 2015a). Gross motor capacity was assessed using the 1-minute walk test (McDowell et al. 2009) and selected items from the 66-item Gross Motor Function Measure (Russell et al. 2013). The use of gross motor skills in everyday life was assessed using a Danish version of the subscale movement and balance of the Pediatric Quality of Life Inventory Cerebral Palsy Module (Varni et al. 2006) and a Danish version of the subscale of transfer and basic mobility of the Pediatric Outcomes Data Collection Instrument (Daltroy et al. 1998). Statistics The current study is based on a sample of children with CP who volunteered to participate in a randomized controlled trial. The sample size calculation for the original study was based on a between-group change score of 7.9 points on the primary outcome measure: the GDI (Rasmussen et al. 2015b). Descriptive statistics were calculated for sex, CP subtype, and classification according to the GMFCS. In the statistical analysis, the median scores of the GDI and GVS of the ankle from 5 trials were used. In the analysis of ROM, the GDI, GVS of the ankle, and Peak dorsiflexion we used data from the affected side of patients with unilateral CP, and for participants with bilateral CP we used the most affected side. The most affected side was determined as the leg with the highest number of measurements in the red and/or yellow categories in DF and in cases without differences in categories, the side with the lowest GDI. The statistical distribution of data was investigated using normal probability plots and the Shapiro– Wilk test. The GVS of the ankle (p < 0.001) and Pediatric Outcomes Data Collection Instrument transfer and basic mobility scores (p < 0.001) were not normally distributed. Scatterplots with fitted values of the outcome measures were prepared to provide an overview of the data. Correlations between ROM and the outcome variables were investigated with Pearson correlation coefficients, except for the GVS of the ankle and Pediatric Outcomes Data Collection Instrument transfer and basic mobility scores, where Spearman’s rank correlation coefficients were used. The correlation coefficients were interpreted according to Dancey and Reidy (2011). Differences in the normally distributed outcome variables in the 3 ROM categories were investigated with one-way ANOVA. The GVS of the ankle and Pediatric Outcomes Data Collec-
12123 Rasmussen D.indd 439
tion Instrument transfer and basic mobility scores were both assessed with the Kruskal–Wallis test and, if relevant, pairwise comparisons with the Wilcoxon rank-sum test (MannWhitney). Statistical analyses were performed using Stata/IC 14.2 for Mac (StataCorp, College Station, TX, USA). The significance level for all statistical results was p < 0.05. Ethics, registration, funding, and potential conflicts of interest The study complies with the principles of the Declaration of Helsinki. Study approval was obtained from the Committee for Medical Research Ethics in the Region of Southern Denmark (S-20120162) and the Danish Data Protection Agency (200858-0035). Informed consent to participate was achieved. Trial registration: ClinicalTrials.gov NCT02160457. The study was funded by the University of Southern Denmark; Odense University Hospital Research grants, the Region of Southern Denmark Research and PhD grants, the Physiotherapy Praxis Foundation, Ludvig and Sara Elsass Foundation, and the Danish Physiotherapy Research Fund. The funding organizations had no involvement in the study design; in collection, analysis, and interpretation of the data; in the writing of the report; or in the decision to submit the article for publication. All the authors of this manuscript declare that they have no conflict of interest related to the current study.
Results 160 patients were invited to participate in the randomized controlled trial. Of these, 48 patients did not answer, 36 were not interested in further information and 16 were not eligible after screening. Consequently, this cross-sectional study was based on 60 participants with spastic CP at GMFCS levels I and II (21 girls; average age 6 years and 10 months (SD 1 year 3 months), 43 diagnosed with unilateral CP) (Table 1). Statistically significant moderate correlations were observed between the GVS of the ankle and DF with flexed knee (r = –0.37, [95% CI –0.57 to –0.13], p < 0.05) and extended knee (r = –0.37, [CI –0.57 to –0.13], p < 0.05) and peak dorsiflexion and DF with flexed knee (r = 0.49, [CI 0.26–0.67], p < 0.001) and extended knee (r = 0.55, [CI 0.35–0.71], p < 0.001) (Table 2). There were statistically significant differences in the GVS of the ankle and peak dorsiflexion between the 3 groups of participants based on the categories with flexed and extended knee (Table 2). For DF with flexed knee, the median GVSs of the ankle for the red and green categories were 14° and 7.6°; the distributions in the 2 groups differed significantly ((z-score, p-value), z = –2.63, p = 0.009) and with extended knee, the median GVS of the ankle for the red and green categories were 17° and 7.6°; the distributions in the 2 groups
Acta Orthopaedica 2018; 89 (4): 437–442
Table 2. Correlation coefficients, coefficient of determination, mean/median scores of the outcome measures in the groups formed by their ankle ROM, and results of the ANOVA/Kruskal–Wallis test. Values for ankle ROM are mean (SD) unless otherwise stated
r Dorsiflexion (knee 90°), n GDI score 0.12 Peak dorsiflexion 0.49 1-min walk test a 0.11 GMFM 0.09 Pedsql –0.01 GVS ankle b –0.37 PODCI b 0.01 Dorsiflexion (knee 0°), n GDI score 0.05 Peak dorsiflexion 0.55 1-min walk test a 0.17 GMFM 0.09 Pedsql 0.06 GVS ankle b –0.37 PODCI b 0.04
Correlation 95% CI
[–0.14 to 0.36] [0.26 to 0.67] [–0.15 to 0.35] [–0.17 to 0.34] [–0.26 to 0.24] [–0.57 to –0.13] [–0.24 to 0.26]
0.4 < 0.001 0.4 0.5 0.96 < 0.05 > 0.05
[–0.21 to 0.30] [0.35 to 0.71] [–0.09 to 0.41] [–0.17 to 0.34] [–0.20 to 0.31] [–0.57 to –0.13] [–0.22 to 0.29]
0.7 < 0.001 0.2 0.5 0.7 < 0.05 > 0.05
Red 14 75 8.2 76 83 78 13 95 6 78 4.9 73 83 70 17 85
(10) (5.8) (18) (8.3) (22) (11) (17) (7.6) (4.3) (29) (8.3) (16) (5.2) (30)
Ankle ROM Yellow 6 78 15 78 82 73 7.2 92 5 71 6.6 80 82 84 16 93
(21) (7.0) (15) (11) (19) (3.7) (31) (18) (5.6) (14) (11) (20) (13) (17)
Green 40 77 14 80 82 74 7.6 92 49 77 14 80 82 74 7.6 94
(12) (6.0) (13) (8.2) (19) (4.5) (14)
0.9 0.007 0.7 1.0 0.7 0.03 0.7
(13) (5.7) (12) (8.2) (20) (4.2) (14)
0.6 < 0.001 0.5 1.0 0.5 0.02 0.8
Data available for 57 participants (data are missing for 3 participants in the green category). Ankle ROM values are median (IQR) Abbreviations: See Table 1.
differed significantly (z = –2.43, p = 0.02). For peak dorsiflexion, we observed a difference in red versus green and red versus yellow ROM categories with flexed knee ((mean (95% CI) –9.6° (–14 to –4.7) and –7.9° (–13 to –2.6), respectively) and between red versus green and yellow versus green ROM categories with extended knee (–9.6° (–15.4 to –3.8) and –7.9° (–14 to –1.5), respectively). No statistically significant groupmean differences were observed between the participants classified into each of the ROM categories of DF on the variables of GDI, 1-minute walk, Gross Motor Function Measure, the Pediatric Quality of Life Inventory Cerebral Palsy Module, and Pediatric Outcomes Data Collection Instrument transfer and basic mobility scores.
Discussion This study aimed to investigate the threshold values of ROM in DF used by the CPUP. We hypothesized that DF and gross motor function would be associated and that there would be differences in gross motor function between the 3 groups based on the categories. We found moderate correlations between DF and deviations in ankle movement during gait in children with CP at GMFCS levels I and II. Furthermore, we found differences between scores of the specific gait function in the ankle joint (GVS of the ankle and peak dorsiflexion) and the 3 ROM categories but no association or differences were observed for overall measures of gross motor capacity (GDI, 1-minute walk, and Gross Motor Function Measure) or the use of gross motor skills in everyday life. Thus, our
12123 Rasmussen D.indd 440
hypotheses were only partly confirmed and the results suggest that threshold values of DF in the CPUP are of limited clinical value in assessing overall gross motor capacity and the use of gross motor skills in everyday life, but may be used to identify deviations in ankle-specific gait function. Detection of deviations in ankle-specific gait function might be useful in the identification of distal deterioration before it might progress to more proximal involvement. Our findings accord with the relationship between changes in passive ROM and gait function from a research project investigating the effects of gastrocnemius fascia lengthening in 19 children with CP (mean age 8 years) on DF by goniometry and gait function by gait summary measures (Galli et al. 2005). That study reported improvements in DF with flexed knee (from 4.3° before to 8.6°after surgery) and extended knee (from –4.3° before to 9.4° after surgery) with accompanying improvements in overall gait function (GDI from 70 before to 83 after surgery) (GVS of the ankle from 22° before to 12°after surgery). These findings suggest that improvements in DF entail improvement in gait function at the joint level (GVS of the ankle) and, to some degree, in overall gait function (GDI). The moderate correlations we found suggest that factors other than ROM may explain the majority of the observed variation. This is supported by a study comparing clinical examination (passive ROM, spasticity, strength, and selective motor control) with 3-D instrumented gait analysis in children with CP (Desloovere et al. 2006). That study found a fair to moderate correlation between the clinical examination and data from the 3-D instrumented gait analysis and concluded
Acta Orthopaedica 2018; 89 (4): 437–442
that both types of data provide important information on the problems faced by children with CP. The categories used in the CPUP are set to ensure that the patient has enough ROM to perform adequate DF in walking (CPOP 2017). Our results support the complex interaction between different dimensions of function, as proposed by the International Classification of Functioning, Disability and Health (WHO 2007). However, it might be important to investigate the categories for other important issues of spastic gait, such as energy expenditure or the risk of developing deformities (Hagglund et al. 2005, Nordmark et al. 2009). To ensure valid data from the 3-D instrumented gait analysis we decided not to include children at GMFCS level III. However, to promote a representative sample of young children with spastic CP at GMFCS levels I and II, our inclusion and exclusion criteria were kept open. Thus, the study sample is not representative of the total population of children with CP and only a few participants experienced ankle ROM affected above red threshold values, which can be caused by the fact that reduced ROM usually first arises later in life (Nordmark et al. 2009). Thus, there remains a need to investigate the ROM categories in older children, with possibly more reduced ROM. Furthermore, children on higher levels of the GMFCS and ROM categories for the other joints of the lower extremities should also be investigated. Finally, it would be important to investigate whether certain subgroups (i.e. CP subtype, weight, or age) behave differently according to ROM thresholds but due to a relatively small study sample this was not possible. A design limitation of this cross-sectional study is that data collection was performed only during one session and therefore no conclusions regarding causality can be inferred. Furthermore, the strength of our results is limited by the small sample size. In addition, measurement errors of 10–15° of goniometric measurements of ROM have been reported (Nordmark et al. 2009). A study using the Generalizability Theory has shown a measurement error in DF of 9° in between-day measurements, when performed by 3 physiotherapists (McDowell et al. 2000). Due to an inclusion period of 25 months and changes in staff, 6 physiotherapists were involved in data collection. All physiotherapists underwent training in the research protocol. However, we did not investigate the consistency of their measurements, therefore this may have increased the variability of the measurements and, thus, must be seen as a limitation of the study. In summary, our study found that DF is associated with ankle-specific measures of gross motor function (GVS of the ankle and peak dorsiflexion) and that the mean scores of the ankle-specific measures are different in the 3 groups based on the categories. In contrast to our hypothesis, we did not find an important relationship between DF and the 3 related categories to overall measures of gross motor capacity and the use of gross motor skills in everyday life in young children at GMFCS levels I and II.
12123 Rasmussen D.indd 441
The implications of our findings suggest that the current threshold values of DF used in the CPUP are of limited clinical value for assessing overall gross motor function, but may be used to identify isolated deviation of ankle function during gait. As a consequence, other measures that are more related to gait function should be considered in the identification of children at risk of functional decline and who may benefit from interventions.
All the authors participated in the concept and design of the study. HMR and AHL were involved in drafting the manuscript. JS, NWP, MT, and SO revised the first draft and commented on and revised the subsequent draft.
The authors would like to thank all participants in the project; physiotherapists Lotte Jensen, Rasmus Sørensen, Line Kiilerich, and Christina Fonvig for helping with data collection and Suzanne Capell for English-language proofreading of the manuscript.
Acta thanks Jacques Riad and other anonymous reviewers for help with peer review of this study.
Alriksson-Schmidt A I, Arner M, Westbom L, Krumlinde-Sundholm L, Nordmark E, Rodby-Bousquet E, Hagglund G. A combined surveillance program and quality register improves management of childhood disability. Disabil Rehabil 2017; 39(8): 830-6. Baker R, McGinley J L, Schwartz M H, Beynon S, Rozumalski A, Graham H K, Tirosh O. The Gait Profile Score and Movement Analysis Profile. Gait Posture 2009; 30(3): 265-9. Brantmark A, Westbom L, Nordmark E. Mobility and joint range of motion in adults with cerebral palsy: a population-based study. Eur J Physiother 2015; 17(4): 192-9. Cerebral Parese Oppfølgingsprogram (CPOP): Alarmverdier for passive bevegeutslag 2014. Available from: http://oslo-universitetssykehus.no/ seksjon-avdeling/Documents/CPOP alarmverdier uex.pdf: Oslo Universitetssjukhus; 2017. Cerebral Parese Uppfölgnings Program (CPUP): Manual till CPUP - Fysioterapeuter - Pappersformulär 2017-01-01. Available from: http://cpup.se/ wp-content/uploads/2017/01/FT-manual-2017.pdf; 2017. Daltroy L H, Liang M H, Fossel A H, Goldberg M J. The POSNA pediatric musculoskeletal functional health questionnaire: report on reliability, validity, and sensitivity to change. Pediatric Outcomes Instrument Development Group. Pediatric Orthopaedic Society of North America. J Pediatr Orthop 1998; 18(5): 561-71. Dancey C P, Reidy J. Statistics without maths for psychology. 5th ed. Harlow: Prentice Hall; 2011. Davis R B, Õunpuu S, Tyburski D, Tyburski D, Gage J G. A gait analysis data collection and reduction technique. Hum Moc Sci 1991; 10(5): 575-87. Desloovere K, Molenaers G, Feys H, Huenaerts C, Callewaert B, de Walle P V. Do dynamic and static clinical measurements correlate with gait analysis parameters in children with cerebral palsy? Gait Posture 2006; 24(3): 302-13. Galli M, Cimolin V, Crivellini M, Albertini G. Gait analysis before and after gastrocnemius fascia lengthening in children with cerebral palsy. J Appl Biomater Biomech 2005; 3(2): 98-105. Hagglund G, Andersson S, Duppe H, Lauge-Pedersen H, Nordmark E, Westbom L. Prevention of severe contractures might replace multilevel surgery in cerebral palsy: results of a population-based health care programme and new techniques to reduce spasticity. J Pediatr Orthop B 2005; 14(4): 26973.
Hagglund G, Wagner P. Spasticity of the gastrosoleus muscle is related to the development of reduced passive dorsiflexion of the ankle in children with cerebral palsy: a registry analysis of 2,796 examinations in 355 children. Acta Orthop 2011; 82(6): 744-8. McDowell B C, Hewitt V, Nurse A, Weston T, Baker R. The variability of goniometric measurements in ambulatory children with spastic cerebral palsy. Gait Posture 2000; 12(2): 114-21. McDowell B C, Humphreys L, Kerr C, Stevenson M. Test-retest reliability of a 1-min walk test in children with bilateral spastic cerebral palsy (BSCP). Gait Posture 2009; 29(2): 267-9. Nordmark E, Hagglund G, Lauge-Pedersen H, Wagner P, Westbom L. Development of lower limb range of motion from early childhood to adolescence in cerebral palsy: a population-based study. BMC Med 2009; 7: 65. Rasmussen H M, Nielsen D B, Pedersen N W, Overgaard S, Holsgaard-Larsen A. Gait Deviation Index, Gait Profile Score and Gait Variable Score in children with spastic cerebral palsy: intra-rater reliability and agreement across two repeated sessions. Gait Posture 2015a; 42(2): 133-7. Rasmussen H M, Pedersen N W, Overgaard S, Hansen L K, Dunkhase-Heinl U, Petkov Y, Engell V, Baker R, Holsgaard-Larsen A. The use of instrumented gait analysis for individually tailored interdisciplinary interventions in children with cerebral palsy: a randomised controlled trial protocol. BMC Pediatr 2015b; 15(1): 202.
12123 Rasmussen D.indd 442
Acta Orthopaedica 2018; 89 (4): 437â&#x20AC;&#x201C;442
Rasmussen H M, Nordbye-Nielsen K, Moller-Madsen B, Johansen M, Ellitsgaard N, Pedersen CR, Rackauskaite G, Engberg H, Pedersen N W. The Danish Cerebral Palsy Follow-up Program. Clin Epidemiol 2016; 8: 45760. Russell D J, Wright M, Rosenbaum P L, Avery L M. Gross motor function measure (GMFM-66 & GMFM-88) Userâ&#x20AC;&#x2122;s manual 2nd ed. London: MacKeith Press; 2013. Schwartz M H, Rozumalski A. The Gait Deviation Index: a new comprehensive index of gait pathology. Gait Posture 2008; 28(3): 351-7. Varni J W, Burwinkle T M, Berrin S J, Sherman S A, Artavia K, Malcarne V L, Chambers H G. The PedsQL in pediatric cerebral palsy: reliability, validity, and sensitivity of the Generic Core Scales and Cerebral Palsy Module. Dev Med Child Neurol 2006; 48(6): 442-9. World Health Organization. International classification of functioning, disability and health: children & youth version: ICF-CY. Geneva: WHO; 2007.
Acta Orthopaedica 2018; 89 (4): 443–447
Incidence of scoliosis in cerebral palsy A population-based study of 962 young individuals Gunnar HÄGGLUND 1,2, Katina PETTERSSON 1,3, Tomasz CZUBA 4, Måns PERSSON-BUNKE 1,2, and Elisabet RODBY-BOUSQUET 1,3
University, Department of Clinical Sciences, Lund, Orthopedics; 2 Department of Orthopaedics, Skane University Hospital, Lund; 3 Centre for Clinical Research, Uppsala University, Region Västmanland, Västerås; 4 National Competence Center for Quality Registers, University Hospital, Lund, Sweden Correspondence: email@example.com Submitted 2017-12-06. Accepted 2018-02-23.
Background and purpose — Surveillance of scoliosis in individuals with cerebral palsy (CP) is important for ensuring timely diagnosis and identification of curve progression. We analyzed the incidence of scoliosis in relation to age, sex, and gross motor function in a population-based cohort of individuals with CP. Patients and methods — This was a prospective register study of all 1,025 individuals born 1990–2012 in southern Sweden (1.4 million inhabitants) in the Swedish surveillance program for CP, which included > 95% of the total population of people with CP in the area. Annual clinical examinations and radiographic measurement of the Cobb angle of those with a moderate or severe scoliosis were registered. We determined the incidence of scoliosis related to age, sex, and the Gross Motor Function Classification System (GMFCS) level. Results — The inclusion criteria were fulfilled by 962 individuals. The number of people (140/962) with scoliosis increased up to 20–25 years of age. The incidence of scoliosis was related to age and GMFCS level. In individuals at the lowest level of gross motor function (GMFCS V) scoliosis was seen in 10/131 before 5 years of age and at the age of 20 years 75% of these individuals had a Cobb angle > 40°. No one in the highest level of motor function (GMFCS I) developed a Cobb angle > 40° Interpretation — Surveillance programs for scoliosis in CP should be based on age and GMFCS level and should be initiated at a young age and continued into adulthood. ■
Scoliosis is common in non-ambulatory individuals with cerebral palsy (CP) (Saito et al. 1998, Koop 2009). Spasticity, muscle weakness, and incomplete muscle control contribute to impaired trunk control and the development of spinal deformity (Imrie and Yaszay 2010, Tsirikos 2010). Severe scoliosis may cause additional motor dysfunction, sitting and transfer problems, compromised pulmonary function, and pain with
reduced quality of life (Kotwicki and Jozwiak 2008, Fender and Baker 2011). The reported incidence of scoliosis in people with CP varies because studies have used different definitions of scoliosis, age groups, and distribution of gross motor function. Most studies report an incidence of 20–25% (Balmer and MacEwen 1970, Persson-Bunke et al. 2012). The risk is higher in people with total body involvement, for whom an incidence of 64% has been reported (Madigan and Wallace 1981). The overall goal for management of a spinal deformity in patients with CP is to maintain or improve functional abilities, seating, positioning, and quality of life. Close surveillance and evaluation are key components for identifying curve progression and improving or maintaining overall function. This can be done through seating adaptations (Holmes et al. 2003), spinal orthoses (Terjesen et al. 2000, Rutz and Brunner 2008) and surgical treatment (Toovey et al. 2017). Increased knowledge of the incidence of scoliosis in an unselected group of people with CP is of value for predicting future risk of scoliosis, and identifying critical ages for surveillance. CPUP has been a Swedish CP surveillance program for over 2 decades, and was designated as a National Quality Registry in 2005. The registry has a coverage rate of over 95%, and thereby represents almost all children with CP in Sweden (Westbom et al. 2007). CPUP aims to prevent hip dislocations, contractures, and deformities in children with CP (Hägglund et al. 2014, Alriksson-Schmidt et al. 2017). In 2012 an analysis of the incidence of scoliosis in southern Sweden based on data from 1995 through 2008 was presented, and represented 666 children aged 0–18 years (Persson-Bunke et al. 2012). The aim of this study was to further analyze the incidence and prevalence of scoliosis related to gross motor function, sex, and age in the same area with the cohort expanded with longitudinal data between 2008 and 2016.
© 2018 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 10.1080/17453674.2018.1450091
12373 Ha¦êgglund D.indd 443
Acta Orthopaedica 2018; 89 (4): 443–447
Table 1. Classification of scoliosis on clinical examination in CPUP No scoliosis Mild scoliosis Moderate scoliosis Severe scoliosis
A discrete curve visible only during forward bending An obvious curve visible during both extended and forward bending A pronounced curve preventing the child from attaining an upright position without external support
Patients and methods In CPUP, children are enrolled at the earliest suspicion of CP, and the CP diagnosis is determined by a neuropediatrician after the age of 4 years. CPUP includes a continuous standardized follow-up of gross motor function, hand function, clinical findings, and treatment. The children are examined by their local physiotherapist following a standardized method twice a year until 6 years of age and then once a year (AlrikssonSchmidt et al. 2017). In CPUP, children with CP are also followed regularly into adulthood. Gross motor function is classified by the child’s physiotherapist according to the Gross Motor Function Classification System (GMFCS), a five-level system based on self-initiated movement, where level I represents the highest level of function and level V the lowest (Palisano et al. 2008). The clinical examination includes an assessment of the spine in a sitting position, upright and with forward bending. Any spinal deviation is graded as mild, moderate or severe scoliosis according to guidelines outlined in a manual (http://www.cpup.se, Table 1). This standardized clinical spinal assessment has shown high interrater reliability, sensitivity, specificity, and criterion-related validity compared with radiographic Cobb angle measurement (Persson-Bunke et al. 2015). In this study, scoliosis graded as moderate or severe was our event of interest. In CPUP, children 8 years and younger with a fixed scoliosis and children older than 8 years with moderate or severe scoliosis at clinical examination are examined radiographically with anteroposterior and lateral views of the entire spine. The radiographic examination is done in the standing or sitting position if possible and in the supine position if not. The curve magnitude is measured as the Cobb angle. Further radiographic examinations are scheduled based on the progression of the Cobb angle, age, and GMFCS level. In children with a Cobb angle > 40°, surgical treatment is considered (Saito et al. 1998). In this prospective cohort study, all 1,025 children and young adults with CP in the area born between January1, 1990 and December 31, 2012, and participating in CPUP were included at baseline. The study area of southern Sweden comprises 1.4 million inhabitants. 10 children who died and 5 who moved out of the area before the age of 5 years were excluded. People with CP who moved into the area after 5 years of age (n = 58) were also excluded. After 5 years of age, 44 children died
12373 Ha¦êgglund D.indd 444
Table 2. Sex and GMFCS distribution
Sex Male Female Total
236 157 393
114 76 190
GMFCS level III IV 45 50 95
75 60 135
87 62 149
557 405 962
(median age 13 (6–25) years) and 37 moved out of the area (median age 9 (6–24) years). They were included in the analysis during the time they were alive and lived in the area. The clinical and radiographic measurements performed from July 1, 1995, until February 3, 2017, were used for the analyses. Statistics The period prevalence of children with scoliosis was calculated as the number of children with scoliosis during the study period July 1, 1995–February 3, 2017 related to the total population of children with CP in the registry during the same period. Kaplan–Meier analysis was used to identify the age at diagnosis of moderate or severe scoliosis stratified by sex and GMFCS level, and the age at diagnosis of scoliosis with a Cobb angle > 40° stratified by GMFCS. The numbers at risk are presented in 5-year intervals. Cox regression analysis was used to compare the risk of scoliosis in different age groups, GMFCS levels, and in males and females for both criteria. The model fulfilled the proportional hazard assumption. Chi-square tests were used to compare differences in frequency between males and females. The analyses were performed using Stata (IC v.13, StataCorp LP, College Station. TX, USA). Ethics, funding, and potential conflicts of interest The study was approved by the Medical Research Ethics Committee at Lund University (LU-433-99). The study was funded by Stiftelsen för bistånd åt rörelsehindrade i Skåne and by the Norrbacka-Eugenia Foundation. The authors declare no conflict of interest.
Results There were 962 individuals (557 males, 405 females) who fulfilled the inclusion criteria. The sex distribution related to GMFCS level is presented in Table 2. During the follow-up period, 140 of the 962 individuals (15%) developed moderate or severe scoliosis based on the latest clinical examination. The scoliosis was graded as moderate in 48 cases (26 males, 22 females) and severe for 92 (50 males, 42 females). Moderate or severe scoliosis was documented in 14% (76/557) of the males and in 16% (64/405) of the females. Spinal fusion was performed in 50 (54%) of the cases, all classified as
Acta Orthopaedica 2018; 89 (4): 443–447
Figure 1. Percentage of individuals at GMFCS I–V with moderate or severe scoliosis at clinical examination and with a Cobb angle exceeding 20° or 40° at radiographic examination. Individuals with Cobb > 40° are also included in the presentation of Cobb > 20°.
having severe scoliosis; 26/50 males (52%) and 24/42 females (57%)). The mean age at surgery was 14 (6–22) years and the median preoperative Cobb angle was 72° (40°–115°). The GMFCS levels of the individuals operated were levels III (n = 2), IV (n = 15), and V (n = 53). Radiographic examination was registered for 128 of the 140 persons with moderate or severe scoliosis. The Cobb angle was < 20° in 27 cases, of which all were graded as moderate on clinical examination. In 14 cases the Cobb angle was 20°– 39°. Half of them were rated as moderate and half as severe on clinical examination. In the remaining 87 cases the Cobb angle was > 40° (median 60° (40°–100°); 80 were classified as having severe scoliosis on clinical examination. Of the 12 children not examined radiographically, 6 had moderate and 6 had severe scoliosis. Their GMFCS levels were I (n = 2), II (n = 1), III (n = 1), IV (n = 2), and V (n = 6). The children in GMFCS level V were considered to be in too poor health to receive scoliosis surgery. The reason for not being referred for radiographic examination in the 6 children in GMFCS I–IV was not reported. The overall frequency and severity of scoliosis increased with GMFCS level, from 0–1% in GMFCS level I to 42–55% in GMFCS level V based on clinical or radiographic examination (Figure 1). Kaplan–Meier survival estimates based on the results of the clinical examination showed that scoliosis was seen in younger ages in children with higher GMFCS levels (Figure 2). The incidence of scoliosis increased with age and GMFCS level. At 10 years of age about 1% of the children at GMFCS I–II, 5% at GMFCS III, 10% at GMFCS IV, and 30% at GMFCS V had a moderate or severe scoliosis. At 20 years of age the corresponding percentages were 5%, 30%, 45%, and 80% respectively. Survival estimations based on the results of radiographic examination and surgery showed a similar pattern (Figure 3). At 10 years of age 2% of children at GMFCS
12373 Ha¦êgglund D.indd 445
Figure 2. Survival function with 95% confidence interval showing the risk of having a moderate or severe scoliosis diagnosed at different GMFCS levels and ages. Numbers at risk at inclusion and at 5-year intervals reported.
Figure 3. Survival function with 95% confidence interval showing the risk of having a scoliosis with Cobb angle > 40° diagnosed at different GMFCS levels and ages. Numbers at risk at inclusion and at 5-year intervals reported. For GMFCS color codes see Figure 2.
Figure 4. Survival function with 95% confidence interval showing the risk of having a moderate or severe scoliosis in males and females respectively. Numbers at risk at inclusion and at 5-year intervals reported.
Acta Orthopaedica 2018; 89 (4): 443–447
Table 3. Cox regression analysis of the hazard ratio (HR) for developing clinically moderate or severe scoliosis in relation to GMFCS level and sex HR GMFCS level III vs. I/II IV vs. I/II V vs. I/II Females vs. males
8 15 53 1.4
95% CI 4–17 9–30 28–100 1–2
p-value < 0.001 < 0.001 < 0.001 0.04
Table 4. Cox regression analysis of the hazard ratio (HR) for developing scoliosis with Cobb angle 40° in relation to GMFCS level a and sex HR GMFCS level IV vs. III V vs. III Females vs. males
2.3 10 1.4
95% CI 0.96–5.9 4.5–24 0.88–2.1
p-value 0.04 < 0.001 0.1
*No child in GMFCS I or II had Cobb 40°.
III, 5% at GMFCS IV, and 20% at GMFCS V had a Cobb angle exceeding 40°. At 20 years of age the corresponding percentages were 8%, 35%, and 75% respectively. No child in GMFCS I–II developed scoliosis with Cobb angle > 40°. The survival estimation based on sex showed equal development for males and females (Figure 4). In the Cox regression analysis, a high GMFCS level indicated a high risk of scoliosis (Tables 3 and 4).
Discussion The main findings in this study were that higher GFMCS level was a significant risk factor for the development of scoliosis, that scoliosis occurred at younger ages in individuals classified at a higher GMFCS level, and that the incidence of scoliosis continued to increase up to the age of 20–25 years. A strength of the study is that the data are based on > 95% of the total population of people with CP in the area followed in a standardized way in CPUP, the Swedish national CP surveillance program. In the Kaplan–Meyer analysis based on clinical examination first occurrence of moderate or severe scoliosis was considered an event of interest. A study of the reliability and validity of the CPUP grading system of scoliosis showed that most children with a mild scoliosis had a Cobb angle of only 5°–15° (Persson-Bunke et al. 2015). In the analysis based on radiographic examination a Cobb angle > 40° was used as the cut-off. The rationale for this is that a Cobb angle greater than 40° has been found to predict significant progression of the
12373 Ha¦êgglund D.indd 446
magnitude of the curve, and is therefore an indication for considering surgery (Saito et al. 1998, Gu et al. 2011). The cohort comprised 58% males, a distribution that is consistent with the overall boy/girl ratio reported for CP (Westbom et al. 2007). The proportion of males in GMFCS I–II was 60% and in levels III–V 55%. The incidence of scoliosis and the proportion operated with spinal fusion was slightly higher in females. Scoliosis was strongly related to the child’s GMFCS level (see Figure 1). For children at GMFCS level I or II the incidence was low and similar to that of adolescent idiopathic scoliosis in typically developed children (Willner and Udén 1982). A previous study from the same area that included 666 children followed between 1999 and 2008 showed a 50% risk for moderate or severe scoliosis in children at GMFCS level IV or V at 18 years of age (Persson-Bunke et al. 2012). In that study the size of the sample did not allow separate analyses for children at GMFCS levels IV and V. The larger cohort and longer follow-up time in the present study made this separation possible and showed that the risk increased significantly from level IV to V (see Figures 2 and 3). A systematic review by Loeters et al. (2010) found 4 studies that reported a relationship between the severity of CP and scoliosis. The study by Persson-Bunke et al. (2012) showed no relationship between CP subtype and scoliosis, which is why this variable was not included in the present study. Advances in surgical technique in recent decades have provided methods to treat scoliosis effectively while also reducing complication rates (Cloake and Gardner 2016). There are several arguments for early detection and treatment of scoliosis with increasing curve magnitude. Preoperative curve flexibility is an important predictor of the degree of curve correction obtained at surgery (Beckmann et al. 2016). A greater angle of scoliosis is associated with a higher risk of complications (Lipton et al. 1999, Master et al. 2011). Scoliosis in CP often leads to a pelvic obliquity, which can increase the risk of hip dislocation on the high side. It also results in internal hip rotation and reduced range of hip flexion on the high side (Ágústsson et al. 2017), which may result in sitting problems postoperatively. One aim of CPUP is to prevent hip dislocation and severe contractures. Scoliosis is sometimes preceded by hip dislocation or windswept hip deformity, which can cause a pelvic obliquity and initiate scoliosis (Letts et al. 1984, Hägglund et al. 2016). CPUP has substantially reduced the prevalence of children with hip dislocation and windswept hip deformity (Hägglund et al. 2014, 2016). Consequently, the prevalence of scoliosis is probably higher than in the present study in areas without hip surveillance programs. Scoliosis was seen at younger ages in the children at GMFCS level V. Surgery with complete spinal fusion in young children can cause problems by impairing growth of the spine and thorax, which has effects on respiratory development and function (Vitale et al. 2008). If the curve is flexible, it is pos-
Acta Orthopaedica 2018; 89 (4): 443–447
sible to reduce the progression rate and delay the need for surgery by spinal bracing (Olafsson et al. 1999, Terjesen et al. 2000). Spinal brace treatment to delay surgery was rare in the study area. The exact number of children using spinal orthoses to prevent curve progression could not be obtained from the reported data. More recent techniques allow the use of “growing rods” to provide spinal support and correct deformity while allowing for growth (McElroy et al. 2012). Not all individuals with moderate or severe scoliosis on clinical examination were examined radiographically. Some of the radiographic examinations were done in the lying position, which might have underestimated the curve magnitude in children with a flexible curve. The numbers at risk in the Kaplan–Meier analysis were low at some GMFCS levels at 20 and 25 years of age. In summary, the incidence of scoliosis in individuals with CP was strongly related to their GMFCS level. At 20 years of age about 75% of those at GMFCS level V had a Cobb angle > 40°. Surveillance programs should be based on age and GMFCS level and must start at young ages and continue into adulthood.
Study design: GH, KP, TC, MPB, ERB. Data collection: GH, MPB. Data analysis: GH, KP, TC, MPB, ERB. Manuscript preparation: GH, KP, TC, MPB, ERB.
Acta thanks Thomas Andersen and other anonymous reviewers for help with peer review of this study.
Ágústsson A, Sveinsson Þ, Rodby-Bousquet E. The effect of asymmetrical limited hip flexion on seating posture, scoliosis and windswept hip distortion. Res Developmental Disabil 2017; 71: 18-23. Alriksson-Schmidt A I, Arner M, Westbom L, Krumlinde-Sundholm L, Nordmark E, Rodby-Bousquet E, Hägglund G. A combined surveillance program and quality register improves management of childhood disability. Disabil Rehabil 2017; 39(8): 830-6. Balmer G A, MacEwen G D. The incidence and treatment of scoliosis in cerebral palsy. J Bone Joint Surg Br 1970; 52(1): 134-7. Beckmann K, Lange T, Gosheger G, Bövingloh A S, Borowski M, Bullmann V, Liljenqvist U, Schulte T L. Surgical correction of scoliosis in patients with severe cerebral palsy. Eur Spine J 2016; 25(2): 506-16. Cloake T, Gardner A. The management of scoliosis in children with cerebral palsy: a review. J Spine Surg 2016; 2(4): 299-309. Fender D, Baker A D L. Spinal disorders in childhood, II: Spinal deformity. Surgery 2011; 29(4): 175-80. Gu Y, Shelton J E, Ketchum J M, Cifu D X, Palmer D, Sparkman A, JermerGu M K, Mendigorin M. Natural history of scoliosis in nonambulatory spastic tetraplegic cerebral palsy. PM R 2011; 3(1): 27-32. Hägglund G, Alriksson-Schmidt A, Lauge-Pedersen H, Rodby-Bousquet E, Wagner P, Westbom L. Prevention of dislocation of the hip in children with cerebral palsy: 20-year results of a population-based prevention programme. Bone Joint J 2014; 96-B(11): 1546-52. Hägglund G, Lauge-Pedersen H, Persson-Bunke M, Rodby Bousquet E. Windswept hip deformity in children with cerebral palsy: a populationbased prospective follow-up 2015. J Child Orthop 2016; 10(4): 275-9.
12373 Ha¦êgglund D.indd 447
Holmes K J, Michael S M, Thorpe S L, Solomonidis S E. Management of scoliosis with special seating for the non-ambulant spastic cerebral palsy population: a biomechanical study. Clin Biomech 2003; 18(6): 480-7. Imrie M N, Yaszay B. Management of spinal deformity in cerebral palsy. Orthop Clin North Am 2010; 41(4): 531-47. Koop S E. Scoliosis in cerebral palsy. Dev Med Child Neurol 2009; 51(Suppl 4): 92-8. Kotwicki T, Jozwiak M. Conservative management of neuromuscular scoliosis: personal experience and review of literature. Disabil Rehabil 2008; 30(10): 792-8. Letts M, Shapiro L, Mulder K, Klassen O. The windblown hip syndrome in total body cerebral palsy. J Pediatr Orthop 1984; 4: 55-62. Lipton G E, Miller F, Dabney K W, Altiok H, Bachrach S J. Factors predicting postoperative complications following spinal fusions in children with cerebral palsy. J Spinal Disord 1999; 12(3): 197-205. Loeters M J, Maathuis C G, Hadders-Algra M. Risk factors for emergence and progression of scoliosis in children with severe cerebral palsy: a systematic review. Dev Med Child Neurol 2010; 52(7): 605-11. Madigan R R, Wallace S L. Scoliosis in the institutionalized cerebral palsy population. Spine 1981; 6(6): 583-90. Master D L, Son-Hing J P, Poe-Kochert C, Armstrong D G, Thompson G H. Risk factors for major complications after surgery for neuromuscular scoliosis. Spine 2011; 36(7): 564-71. McElroy M J, Sponseller P D, Dattilo J R, Thompson G H, Akbarnia B A, Shah S A, Snyder B D. Growing rods for the treatment of scoliosis in children with cerebral palsy: a critical assessment Spine 2012; 37(24): E150410. Olafsson Y, Saraste H, Al-Dabbagh Z. Brace treatment in neuromuscular spine deformity. J Pediatr Orthop 1999; 19(3): 376-9. Palisano R J, Rosenbaum P, Bartlett D, Livingston M H. Content validity of the expanded and revised Gross Motor Function Classification System. Dev Med Child Neurol 2008; 50(10): 744-50. Persson-Bunke M, Hägglund G, Lauge-Pedersen H, Wagner P, Westbom L. Scoliosis in a total population of children with cerebral palsy. Spine 2012; 37(12): E708-13. Persson-Bunke M, Czuba T, Hagglund G, Rodby-Bousquet E. Psychometric evaluation of spinal assessment methods to screen for scoliosis in children and adolescents with cerebral palsy. BMC Musculoskelet Disord 2015; 16: 351. Rutz E, Brunner R. Short lumbar brace for treatment of neuromuscular scoliosis. Med Orthop Tech 2008; 3: 71-74. Saito N, Ebara S, Ohotsuka K, Kumeta H, Takaoka K. Natural history of scoliosis in spastic cerebral palsy. Lancet 1998; 351(9117): 1687-92. Terjesen T, Lange J E, Steen H. Treatment of scoliosis with spinal bracing in quadriplegic cerebral palsy. Dev Med Child Neurol 2000; 42(7): 448-54. Toovey R, Harvey A, Johnson M, Baker L, Williams K. Outcomes after scoliosis surgery for children with cerebral palsy: a systematic review. Dev Med Child Neurol 2017; 59(7): 690-8. Tsirikos A. Development and treatment of spinal deformity in patients with cerebral palsy. Indian J Orthop 2010; 44(2): 148-58. Westbom L, Hägglund G, Nordmark E. Cerebral palsy in a total population of 4-11 year olds in southern Sweden. Prevalence and distribution according to different CP classification systems. BMC Pediatr 2007; 7: 41. Willner S, Udén A. A prospective prevalence study of scoliosis in southern Sweden. Acta Orthop Scand 1982; 53(2): 233-7. Vitale M G, Matsumoto H, Bye M R, Gomez J A, Booker W A, Hyman J E, Roye D P Jr. A retrospective cohort study of pulmonary function, radiographic measures, and quality of life in children with congenital scoliosis: an evaluation of patient outcomes after early spinal fusion. Spine 2008; 33(11): 1242-9.
Acta Orthopaedica 2018; 89 (4): 448–453
Identification and treatment of residual and relapsed idiopathic clubfoot in 88 children Jurre H STOUTEN 1,2, Arnold T BESSELAAR 1,2, and M C (Marieke) VAN DER STEEN 1
1 Department of Orthopaedic Surgery, Catharina Hospital Eindhoven, Eindhoven, The 2 Orthopaedic Center Máxima, Máxima Medical Center, Eindhoven, The Netherlands
Correspondence: firstname.lastname@example.org Submitted 2017-09-04. Accepted 2018-03-26.
Background and purpose — The Ponseti treatment is successful in idiopathic clubfoot. However, approximately 11–48% of all clubfeet maintain residual deformities or relapse. Early treatment, which possibly reduces the necessity for additional surgery, requires early identification of these problematic clubfeet. We identify deformities of residual/relapsed clubfeet and the treatments applied to tackle these deformities in a large tertiary clubfoot treatment center. Patients and methods — Retrospective chart review of patients who visited our clinic between 2012 and 2015 focused on demographics, deformities of the residual/ relapsed clubfoot, and applied treatment. Residual deformities were defined as deformities that were never fully corrected and needed additional treatment. We defined relapse as any deformity of the clubfoot reoccurring, after initial successful treatment, with necessity for additional treatment. Results — We identified 33 patients with residual and 55 patients with relapsed clubfeet. In both groups decreased dorsal flexion and adduction were the most often registered deformities. Furthermore, often equinus/decreased dorsiflexion, active supination, and varus occurred. In more than half, typical profiles of combined deformities were found. Relapses occurred at all stages of treatment and followup; half of the residual or relapsed clubfeet were identified before the end of the bracing period. In half of the patients, additional treatment consisted of the Ponseti treatment, one– quarter also required adaptation of the brace protocol, and one–quarter needed additional surgery. The Ponseti treatment was mainly reapplied if feet presented with relapses or residues until the age of 5. Interpretation — Practitioners should especially be aware of equinus/decreased dorsiflexion, adduction, and active supination as a sign of a residual or relapsed clubfoot. Due to the heterogeneous profiles of these clubfeet, treatment strategy should be based on a step-by step approach including recasting, bracing, and if necessary surgical intervention.
The Ponseti treatment has shown to be a very successful treatment for idiopathic clubfoot (Ponseti et al. 1992, Morcuende et al. 2004). Unfortunately, approximately 11–48% of treated clubfeet show problems during follow up (Ponseti 2002, Morcuende et al. 2004, Hosseinzadeh et al. 2017). Some feet do not fully correct, while others have a tendency to relapse (Ponseti 2002). In the literature, the distinction between these problematic clubfeet is not always clear (Hosseinzadeh et al. 2017). In the current study we define residual deformities as deformities that underwent primary treatment but were never fully corrected and need additional treatment (Radler and Mindler 2015). A relapse is defined as any feature of the clubfoot reoccurring after initial successful treatment, which needs additional treatment (Laaveg and Ponseti 1980). The pathology of residual or relapsed clubfeet is still unknown. It is clear that inappropriate bracing leads to relapses (Morcuende et al. 2004, Dobbs et al. 2004). But non-compliance with the bracing protocol does not explain all relapses. Any deformity of the initial clubfoot can be present in a residual or relapsed clubfoot. Furthermore, toe deformities, stiffness, or articular incongruence might be present (Uglow and Kurup 2010, Parsa et al. 2014). The treatment of residual and relapsed clubfeet involves many challenges that were, in the past, frequently tackled by means of extensive surgical interventions (Radler and Mindler 2015). Nowadays, opinions have shifted and the additional treatment takes on a more reserved or nonoperative approach. A step-by-step approach is important and based on the original Ponseti treatment, involves repeated casting, a proper bracing period, and only if necessary surgical intervention (Dietz 2006, Jowett et al. 2011, Radler and Mindler 2015). The Ponseti treatment includes the possibility to expand the foot correction with tibialis anterior tendon transfer to redress the common problem of adduction and supination in residual and relapsed clubfeet (Parsa et al. 2014, Radler and Mindler 2015). Furthermore, treatment should be specific to the
© 2018 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 10.1080/17453674.2018.1478570
12063 Stouten D.indd 448
Acta Orthopaedica 2018; 89 (4): 448–453
pathoanatomy of the deformity and functional needs of the patients should be taken into account during treatment planning (Radler and Mindler 2015). Applied to residual, relapsed, neglected, and complex clubfoot, the Ponseti treatment has shown positive results with respect to pain, functionality, and cosmesis (Dietz 2006, Lourenço and Morcuende 2007, Radler and Mindler 2015, Matar et al. 2016). Early identification of residues and relapses allows for early treatment and therefore may diminish the necessity for major surgical interventions (Ponseti 2002, Dietz 2006). However, due to the variable manner in which a residual or relapsed clubfoot may occur, they may be difficult to identify at an early stage. Therefore, the aim of this study is to gain insight into the deformities of residual and relapsed clubfeet and the applied treatment at our clubfoot treatment center. Being aware which deformities occur most frequently and at what stage of the treatment aids in determining the optimal treatment and timing of this.
Patients and methods We performed a retrospective dossier study of clubfoot patients treated by one orthopedic surgeon specialized in clubfoot pathology (ATB) between 2012 and 2015. Potential study participants were identified by means of the Dutch diagnosis and treatment code (DBC) for clubfoot. Patients were included in the study if they had idiopathic clubfoot and underwent treatment for a residual or relapsed deformity of their clubfoot at our tertiary institute. Patients were excluded if they did not have idiopathic clubfoot but rather a syndromic, neurogenic, or positional clubfoot. Data on demographics, clubfoot deformities, the primary treatment, and additional treatment were gathered from the electronic patient files. Known clubfoot deformities were recorded: adduction, equinus, varus, and cavus. Since decreased dorsiflexion and active supination are early signs of relapsing clubfeet and as such require treatment to prevent further problems, we documented these as well. Active supination is caused by suboptimal function of the tibialis anterior (TA) muscle. In residual or relapsed deformities, the shape of the foot results in relative medialization of the insertion of the TA. This leads to an over-supination in the early swing phase but also by landing on the lateral border of the foot. This mechanism can be studied during walking but also when sitting on a bench with the lower legs free. If the patient is asked to raise the foot, it can clearly be seen if the foot is dorsiflexed neutrally or with a supination component. Because equinus and decreased dorsiflexion are strictly related, and both mark a deformity in the sagittal plane of the ankle, we decided to combine the equinus and decreased dorsiflexion (EqDD) and treated them as a single entity. The data on the deformities were gathered retrospectively and were cumulatively gathered per individual. Because of
12063 Stouten D.indd 449
From Catharina Hospital Eindhoven n = 146
From Maxima Medical Center Veldhoven n = 167
Identiﬁed clubfoot patients n = 313 Excluded patients (n = 225): – non–idiopathic clubfoot, 51 – no problematic clubfoot a, 139 – insufficient medical data, 15 – no additional treatment in our center, 20 Patients with problematic a idiopathic clubfoot eligible for data assessment n = 88 Relapse: 55 patients (73 feet) Residue: 33 patients (49 feet)
Figure 1. Flowchart of patient selection. a Residual and relapsed clubfeet.
the retrospective character of this study, the moment at which specific deformities occurred could not be determined. Given these facts, the deformities in a single patient formed a profile of the deformities that occurred over time. Primary treatment was defined as the initial treatment of any kind that was performed with the intention to fully correct the primary clubfoot. Treatment for residual or relapsed clubfoot that had been performed outside our own clinic was also recorded. Additional treatment comprised the treatment applied for a residual or relapsed clubfoot at our center. For the different treatment stages, we registered the date of the initial corrective casts, usage of braces, and any surgical treatment. We composed 3 treatment groups that differ from each other in extensiveness of the additional treatment necessary to treat the residual or relapsed clubfoot. In the first group (extended Ponseti protocol) additional treatment consisted of additional treatment following the Ponseti protocol. This could entail a second casting phase, renewed Achilles tendon tenotomy (re-ATT), and/or a tibialis anterior tendon transfer (TATT) (Ponseti 2002). In the second group (brace adaptation), the treatment protocol of the Ponseti group was combined with adaptation of the bracing phase. In this group bracing was prolonged according to age or adjusted by the use of another type of brace. In those—often older—children who did not tolerate the standard foot abduction brace with a bar, an abduction dorsiflexion mechanism brace was used as an alternative. This brace consists of an alternative abduction, endorotation, dorsiflexion mechanism. It is constructed without a bar between both feet and therefore is usable unilaterally. The third group (additive surgery) entailed patients who received additional extra- and/or intra-articular surgical treatment that is not part of the aforementioned extended Ponseti protocol.
Acta Orthopaedica 2018; 89 (4): 448–453
Table 1. Descriptive data on both groups for patients and feet Residual
Table 2. Proportions of profiles of deformities in the relapse and residue group. Values are number of feet
Patients, n 33 Age at identification in months mean (range) 41 (3–187) Follow up since identification in months mean (range) 45 (5–120) Male/female ratio (n) 4.5/1.0 (27/6) Unilateral:bilateral ratio (n) 1.1/1.0 (17/16) Feet, n 49 Pirani at start of primary treatment median (IQR) a 5.8 (4.8–6.0) Ponseti treatment used in primary treatment, n yes 26 no 19 missing 4 Number of casts used in Ponseti median (IQR) b 8.0 (6.5–15.0) Deformity, n equinus 20 decreased dorsiflexion 22 adduction 20 active supination 15 cavus 2 varus 13 Group of additional treatment, n Ponseti protocol 32 brace adaptation 8 additional surgery 9 a Data b Data
59 (3–182) 45 (3–250) 2.0/1.0 (36/19) 1.1/1.0 (29/26) 73 5.5 (4.5–6.0) 49 16 8 5.0 (4.0–6.0) 12 35 34 21 3 13
Single deformity EqDD a active supination adduction cavus EqDD involved EqDD + active supination EqDD + adduction EqDD + varus EqDD + active supination + varus EqDD + active supination + adduction EqDD + adduction + cavus EqDD + varus + adduction EqDD + active supination + varus + adduction Other combinations active supination + adduction active supination + varus active supination + varus + adduction varus + adduction adduction + cavus a EqDD
17 14 1 1 1 16 2 5 0 1 2 0 4 2 10 3 4 0 2 1
31 22 2 7 0 20 2 4 1 0 7 1 3 2 13 4 3 1 3 2
= equinus/decreased dorsiflexion
27 23 23
only available of 22/10 feet. only available of 37/17 feet.
Ethics, funding, and potential conflicts of interest Ethics approval was obtained from the local medical ethical committee (niet-WMO 2016-23). No funding was obtained for this study. No conflicts of interest declared.
Results Patient selection Initially, we identified 416 patients by means of the diagnosis/treatment code (DBC) for clubfoot. First, 103 patients were excluded from the database because these consisted of mothers expecting a child with clubfoot who were counselled before delivery by the orthopedic surgeon or patients who had been labelled incorrectly as clubfoot patients, leaving 313 clubfoot patients. Ultimately, 88 patients with 122 residual or relapsed clubfeet were identified and included in the following analyses (Figure 1). Patient characteristics The residual group comprised 33 patients (49 feet) and 55 patients (73 feet) had relapsed deformities. A substantial part (52/88) of the population was referred to our clinic from other centers. This explains the relatively high percentage of residual and relapsed feet in our population. As expected, the mean
12063 Stouten D.indd 450
age at identification of the relapse was considerably higher than the age at identification of residual deformities (respectively 4.9 (0–15) years and 3.4 (0–16) years). The mean follow-up since identification of the residual and relapsed clubfoot was similar in both groups; in the residual group this was 3.8 (0.4–10) years, and in the relapse group 3.7 (0.3–21) years. (Table 1) The male-to-female ratio in the residual group was 4.5:1 and in the relapse group 2:1. Unilateral and bilateral clubfoot occurred frequently in the relapse group, as it did in the residual group (1.1:1.0). It should be noted that not all patients with bilateral clubfeet in the relapse group had a bilateral relapse. In the relapse group 8 patients with a bilateral clubfoot had a unilateral relapse. Often Pirani scores and number of casts are used to give an idea of the primary severity of the clubfoot. Unfortunately, data on these two particular variables, especially of referred patients, were not complete (Table 1). Based on the available data, clubfeet in the residual group had a Pirani score at the start of the primary treatment of 5.8 with an interquartile range (IQR) of 4.8–6.0. For the relapse group the median Pirani score was 5.5 with an IQR of 4.5–6.0. In the residual group the number of primarily treated patients according to the Ponseti method was 26 of 49 patients and a median of 8.0 casts were used (IQR = 6.5–15) (Table 1). In the relapse group 49 of the 73 patients were treated with the Ponseti method as primary treatment and a median number of 5.0 casts was used to achieve initial correction (IQR = 4.0–6.0).
Acta Orthopaedica 2018; 89 (4): 448–453
Number of cases
Number of cases
Number of cases
EqDD Other single deformity EqDD involved Other proﬁles
Additive surgery Brace adaptation Ponseti protocol
Age at identiﬁcation
Age at identiﬁcation
Re-ATT TPTT OEA OIA Re-ATT TPTT OEA OIA Re-ATT TPTT OEA OIA TATT PED PMR TATT PED PMR TATT PED PMR
Figure 2. Age at identification of the residual and relapsed clubfoot compared to the group of profiles of deformities in the clubfoot. Blue bars show the feet with solitary equinus and/or decreased dorsiflexion and green bars are those with other solitary deformities. Red bars indicate feet with combined profiles that contain EqDD and purple bars show feet with other combined profiles. Plain tones show patients that were referred to our clinic.
Figure 3. Age at identification of residual and relapsed clubfoot compared with their treatment group. Blue bars represent patients that had satisfying results with the Ponseti protocol. For patients displayed by green bars the Ponseti protocol was not sufficient and adaptation to a bracing protocol was needed to get good results. The red group contains those patients in which the previous 2 treatment options did not suffice and additive surgery was needed. Plain toned bars contain patients that were referred to our clinic.
Deformities (Table 2) Table 1 shows the occurrence of the different deformities. Portraying these deformities as single entities does shed some light on the deformities that appear in residual and relapsed clubfeet, but it does not grasp the full complexity of the clubfeet. In 26 of the residual clubfeet and 33 of the relapses, a profile with multiple deformities evolved over time. In total, 13 profiles with multiple deformities were identified in addition to 4 deformities occurring as a single entity (Table 2). EqDD occurred in 30 of the residual clubfeet and 42 of the relapsed clubfeet. In the majority of the profiles, EqDD played a role. Furthermore, EqDD is the most prevalent single deformity that occurs in residual and relapsed clubfoot. Adduction as a single entity, however, occurs in a rather large proportion in the relapse group as well. Adduction was present only once as a single entity in the residual group. Profiles in which active supination and adduction played a role were abundant as well (28 in the residue group and 41 in the relapse group). Varus was always combined with other deformities (Table 2). As a means of examining the relation between profiles of deformities and the age at which the residual and relapsed clubfeet were initially identified, we plotted these variables against each other (Figure 2). As stated in the methods section, it should be noted that not all feet had sufficient data available to determine the age at identification (20/122 missing). An important proportion of the solitary EqDD deformity (27 of 36) occurs in the first year of life (Figure 3). When patients become older the amount of deformities occurring as a single entity diminishes. In our cohort 50% of the residual and relapsed
12063 Stouten D.indd 451
Figure 4. Age of patients at the moment of surgical intervention. Blue color marks the surgical treatments that are part of the Ponseti protocol. Green portrays extra-articular (EA) treatments that are not part of the Ponseti protocol. The red bars show the intra-articular (IA) treatments. Re-ATT: renewed Achilles tendon tenotomy, TATT: tibialis anterior tendon transfer, TPTT: tibialis posterior tendon transfer, PED: partial epiphysiodesis of the ventral distal tibia, OEA: other extra-articular surgery, PMR: posteromedial release, OIA: other intra-articular surgery.
clubfeet were identified before the end of the bracing period (before the age of 4). Furthermore, a peak is seen at the age of 5, 1 year after the end of the bracing period. After this peak, the amount of new residual and relapsed clubfeet decreases even further. Additionally, the proportion of clubfeet where single deformities play a role seems to decrease with age and these are seldom seen at the age of 6 and older. Combined deformities are detected more often as residual or relapsed deformities at a higher age. Figure 3 also distinguishes our own (striped) from referred patients (plain color). In the first two years of life patients are often referred with clubfeet that display EqDD. Referred patients often demonstrate a more complex profile as they are diagnosed later in life. Treatment groups 3 treatment groups were composed (see Methods section). The extended Ponseti protocol was sufficient in 32 out of 49 of the residue group and 27 out of the 73 clubfeet in the relapse group. An extra brace adaptation was sufficient in 8 of the residual clubfeet and in 9 of these feet additional surgery was needed (Table 1). In the relapse group 23 of the feet were treated with brace adaptation and 23 of the patients needed additional surgical interventions that were not a part of the Ponseti protocol. The Ponseti protocol was mainly used when feet presented with residues or relapses until the age of 5 (Figure 2). Brace adaptations increase up to the age of 6, but by the age of 9 brace adaptation was not used in any of the cases. The additional surgery group was not influenced by age.
Surgical interventions In the 3 previously depicted treatment groups, ATT and TATT were considered part of the extended Ponseti protocol, whereas any other surgery was not part of the extended Ponseti protocol and was classified as additional surgery. Surgical treatment mostly consisted of extra articular procedures, of which re-ATT was applied 67 times out of 135 surgeries. Until the age of 4, re-ATT was almost exclusively the surgical treatment of choice. After that, TATT was used more often as well as other extra articular procedures (Figure 4). The additional surgeries which are not part of the Ponseti protocol (see Figure 4, in green and red) were mostly preserved for older children in whom other treatment did not prove to be effective. We subdivided these additional surgeries into extra-articular treatments (green) and intra-articular treatments (red). The partial epiphisiodesis (PED) of the anterior segment of the distal end of the tibia was the most predominantly used extra-articular treatment that was not part of the Ponseti protocol. Intra-articular surgeries (see Figure 4, in red) were used in older children as well and consisted of posteromedial releases, closing wedge osteotomies, a triple arthrodesis, a fascia plantaris release, and the use of external fixators (in medial column lengthening and metatarsal osteotomies).
Discussion We describe a cohort of residual and relapsed clubfeet treated at a tertiary clubfoot treatment center. By identifying the profiles of deformities occurring in these problematic clubfeet and describing the treatment performed we depict the strength of the Ponseti treatment used in these patients but also show the necessity for additional surgical interventions in some patients. Our population showed high resemblance to the normal clubfoot population in terms of affected foot and sex (Werler et al. 2013). The majority of the included patients had already initially been treated with the Ponseti method, nowadays the preferred initial treatment for idiopathic clubfeet in the Netherlands (Besselaar et al. 2017). The residual group showed a high number of casts used (median 8 casts), while the initial Pirani score was not higher. This suggests that these clubfeet already showed difficulties during the initial correction. Zhao et al. (2016) also showed that difficulty in correcting the initial deformity was predictive for a relapse. Furthermore, Ponseti et al. (2006) noted a specific group of complex clubfeet that are difficult to treat and require a modified Ponseti treatment. All known clubfoot deformities could be present in residual and relapsed clubfeet. As was pointed out by Ponseti in 2002, equinus/decreased dorsiflexion is the most frequent reoccurring deformity, as we found as well. We also found many different profiles of deformities, though deformities in children younger than 4 years old often occurred solitarily.
12063 Stouten D.indd 452
Acta Orthopaedica 2018; 89 (4): 448â&#x20AC;&#x201C;453
The majority of relapses occur in the first 2 to 3 years of life and rarely after the age of 5 (Dietz 2006, Ponseti et al. 2006). However, we also saw an increase in cases at the end of the bracing period. This difference might be related to ours being a tertiary clubfoot treatment center. As a consequence, patients may be older at presentation than they would be in other centers. However, the peak of residuals and relapses around the age of 5 might also point toward the important role of the foot adduction brace in preventing relapses. It should, however, be noted that difficulties during bracing often are a sign of a residual or relapsed foot causing a fitting problem as a result of decreased dorsiflexion (Dietz 2006). In our study, however, in children over the age of 5 relapses often occurred in more complex deformity profiles. Deformities, age, and treatment are all associated with each other, as age and deformities determine the suitable treatment. Treatment using the Ponseti protocol is particularly prevalent in the first 4 years of life, while brace adaptations and additional surgery are commoner in older children; we found the same in our cohort. This is in line with the step-by-step approach described by Radler and Mindler (2015) and Jowett et al. (2011). Early identification seems to be essential in preventing the need for additional surgical interventions, which have been associated with less positive outcomes in pain, functionality, and cosmesis (Radler and Mindler 2015). In our retrospective cohort no objective scores on pain, functionality, and cosmesis were available. Of course, the retrospective nature of our study also imposes several other limitations. The most important of these are incomplete data and difficulty distinguishing between residual and relapsed clubfeet, especially in referred cases. We defined a relapse in line with the Iowa group. Radler and Mindler (2015) suggest that the differentiation has minimal effect on further treatment. Perhaps a classification of residual and relapsed clubfeet based on severity might be more useful. However, it is difficult to compare the severity of the different profiles of deformities. Bhaskar and Patni (2013) identified 5 relapse patterns based on the involvement of dorsiflexion, adduction, or supination and whether the deformity was either dynamic or fixed. As we found many profiles of deformities, regularly also including cavus and varus, we felt that the Bhaskar classification was not capable of embracing the complexity of both residual and relapsed clubfeet. It should be noted that as we are a tertiary center, we generally treat the more severe cases and therefore the profiles of deformities might be more complex than in a standard clubfoot population. In summary, our study showed that relapses occur at all stages of treatment and follow-up. All deformities of the initial clubfeet can (re)occur in residual and relapsed clubfeet and often a combination of deformities is seen. Practitioners should especially be aware of EqDD, adduction, and active supination as signs of a relapse. Due to the heterogeneous nature of residual and relapsed clubfeet, the treatment strategy should be based on a step-by-step approach including
Acta Orthopaedica 2018; 89 (4): 448–453
recasting, bracing, and if necessary surgical intervention. In the majority of our cases, especially if identified in an early stage, treatment according to Ponseti was sufficient to treat the residual and relapsed clubfeet. Identifying residues or relapses at an early stage could prevent the need for additional surgery.
JHS: Data collection, analysis and, interpretation. Drafting the article. MCS: Conception or design of the work. Data interpretation. Drafting the article. ATB: Conception or design of the work. Data interpretation. Critical revision of the article. Acta thanks Klaus Dieter Parsch and other anonymous reviewers for help with peer review of this study.
Besselaar A T, Sakkers R J B, Schuppers H A, Witbreuk M M E H, Zeegers E V C M, Visser J D, Boekestijn R A, Margés S D, Van der Steen M C M, Burger K N J. Guideline on the diagnosis and treatment of primary idiopathic clubfoot. Acta Orthop 2017; 88(3): 305-9. Bhaskar A, Patni P. Classification of relapse pattern in clubfoot treated with Ponseti technique. Indian J Orthop 2013; 47(4): 370-6. Dietz F R. Treatment of a recurrent clubfoot deformity after initial correction with the Ponseti technique. Instr Course Lect 2006; 55: 625-9. Dobbs M B, Rudzki J R, Purcell D B, Walton T, Porter K R, Gurnett C A. Factors predictive of outcome after use of the Ponseti method for the treatment of idiopathic clubfeet. J Bone Joint Surg Am 2004; 86-A(1): 22-7. Hosseinzadeh P, Kiebzak G M, Dolan L, Zionts L E, Morcuende J. Management of clubfoot relapses with the Ponseti method: results of a survey of the POSNA members. J Pediatr Orthop 2017; Feb 7. doi: 10.1097/ BPO.0000000000000953. [Epub ahead of print]
12063 Stouten D.indd 453
Jowett C R, Morcuende J A, Ramachandran M. Management of congenital talipes equinovarus using the Ponseti method: a systematic review. J Bone Joint Surg Br 2011; 93(9): 1160-4. Laaveg S J, Ponseti I V. Long-term results of treatment of congenital club foot. J Bone Joint Surg Am 1980; 62(1): 23-31. Lourenço A F, Morcuende J A. Correction of neglected idiopathic club foot by the Ponseti method. Bone Joint J 2007; 89-B(3): 378-81. Matar H E, Bierne P, Bruce C E, Garg N K. Treatment of complex idiopathic clubfoot using the modified Ponseti method: up to 11 years follow-up. J Pediatr Orthop B 2016; 26(2): 137-42. Morcuende J A, Dolan L A, Dietz F R, Ponseti I V. Radical reduction in the rate of extensive corrective surgery for clubfoot using the Ponseti method. Pediatrics 2004; 113(2): 376-80. Parsa A, Moghadam M H, Jamshidi M H . Relapsing and residual clubfoot deformities after the application of the Ponseti method: a contemporary review. Arch Bone Jt Surg 2014; 2(1): 7-10. Ponseti I. Current concept review: treatment of congenital club foot. J Bone Joint Surg 1992; 74A(3): 448. Ponseti I V. Relapsing clubfoot: causes, prevention and treatment. Iowa Orthop J 2002; 22: 55-6. Ponseti I V, Zhivkov M, Davis N, Sinclair M, Dobbs M B, Morcuende J A. Treatment of the complex idiopathic clubfoot. Clin Orthop Relat Res 2006; 451: 171-6. Radler C, Mindler G T. Treatment of severe recurrent clubfoot. Foot Ankle Clin 2015; 20(4): 563-86. Uglow M G, Kurup H V. Residual clubfoot in children. Foot Ankle Clin 2010; 15(2): 245-64. Werler M M, Yazdy M M, Mitchell A A, Meyer R E, Druschel C M, Anderka M, et al. Descriptive epidemiology of idiopathic clubfoot. Am J Med Genet A 2013; 161A(7): 1569-78. Zhao D, Li H, Kuo K N, Yang X, Wu Z, Liu J, Zhu J. Prognosticating factors of relapse in clubfoot management by Ponseti method. Pediatr Orthop B 2016; Sep 22. Epub
Acta Orthopaedica 2018; 89 (4): 454–456
5-year-old child with late discovered traumatic patellar tendon rupture—a case report Jesper HOLBECK-BRENDEL 1,2, and Ole RAHBEK 1,2 1 Department of Orthopaedic Surgery, Aarhus University Hospital; 2 Orthopaedic Research Laboratory, Department of Orthopaedic Surgery, Aarhus University Hospital, Aarhus, Denmark Correspondence: email@example.com Submitted 2017-11-06. Accepted 2018-03-06.
A 5-year-old otherwise healthy boy was referred to our outpatient clinic with a high-riding patella in the left knee. The patient’s mother had 2 months earlier discovered a small round lump on the patient’s left thigh proximal to where the patella should be. The boy had complained about pain in both legs and the mother discovered the lump during comforting. The boy had had a traumatic incident 5 months earlier, when he fell down from a climbing frame and landed on a flexed left knee. A small discoloration was briefly observed around the knee. The parents noticed no dysfunction and the boy did not complain of pain. He recovered quickly and therefore no further actions were taken. The boy had no history or clinical signs of collagen disease such as Ehlers-Danlos or osteogenesis imperfecta. He walked normally. Inspection of the left knee revealed a patella alta. There was no effusion or other signs of injury. The boy could extend the left knee against force, but the strength was markedly reduced compared with the right side. He could elevate
Figure 1. The left and right knee prior to operation.
his left leg with the knee fully extended. The range of motion was normal. Palpation of the left knee joint showed that the patella sat high but it was possible to manipulate it distally to near normal position (–2 cm) compared with the right knee. The ligamentum patellae could not be palpated. The boy did not complain of pain during the examination. Radiograph of the left knee joint showed marked patella alta (Figure 1). MRI showed signs of injury of the patellar tendon (Figure 2). The MRI was suboptimal due to movement artefacts. Dynamic ultrasound imaging showed a thin, elongated ligamentum patellae, which could not be clearly identified either distally or proximally. There was minimal atrophy of the quadriceps muscle without fatty degeneration. The quadriceps tendon was intact. Furthermore, there were signs of intact medial and lateral patella retinacula. Surgery with repair of the patellar tendon was proposed due to the reduced extension force. It was performed under general anesthesia. The patellar tendon was elongated and shred-
Figure 2. T2 weighted MRI image of the left knee. The patella is proximally located. The ligamentum patella is seen with normal thickness and signal intensity in the most proximal part, but is ruptured distally.
Figure 3. The left knee 1½ years after the operation.
© 2018 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 10.1080/17453674.2018.1467846
12281 HOLBECK D.indd 454
Acta Orthopaedica 2018; 89 (4): 454–456
ded, particularly in the distal part were a complete rupture was seen. The proximal two-thirds of the tendon was macroscopically normal and was surgically exposed. The distal part was resected. The lateral and medial retinacula were intact. The periosteum on the tibial tuberosity was surgically exposed and 2 periosteal flaps were made. The patella could be repositioned to a normal position compared with the right side. 2 5.0 Mitek anchors (DePuy Synthes Sports Medicine (Mitek), Raynham, MA, USA) with screw threads were placed in the proximal epiphysis of the tibia with fluoroscopic guidance. The tendon was pulled down and sutured with Vicryl 0 (Ethicon Inc., Somerville, NJ, UA) to the Mitek anchors with appropriate tension compared with the right side. We confirmed the correct placement of the patella central in the femoral groove with fluoroscopic guidance. The strength of the repair was tested by passive knee flexion, which was possible up to 30 degrees when tension increased. A DonJoy brace (DJO Global, Vista, CA, USA) was placed allowing 0–20 degrees of flexion to protect against hyperflexion trauma. Mobilization with full loading on the operated leg was allowed immediately after surgery, but the boy was instructed to refrain from other physical activities. The boy did not complain of pain during the postoperative period. The DonJoy brace was set to allow 0–80 degrees flexion after 6 weeks. The boy could fully extend the operated knee. Radiographs obtained at 10 weeks postoperatively showed that the patella was positioned just proximal to the physis of the distal femur and still more proximal compared with the right side. The anchors were in place. The boy had no physical limitations and no complaints 1½ years after the operation. Full flexion and extension of the knee joint was possible. The quadriceps strength was normalized. The patella was located 1.5–2 cm proximal on the operated side compared with the right side and moved normally into the femoral groove at full flexion. Radiographs indicated that the patella was in the same position as it was 10 weeks after surgery (Figure 3).
Our 5-year-old, otherwise healthy boy functioned normally with a completely ruptured patellar tendon in the left knee joint. The boy did not present the typical signs of traumatic patellar tendon rupture besides a small discoloration around the knee joint shortly after injury, patella alta and markedly reduced left knee extension strength compared with the right side. It is remarkable that the boy could function so well in the period after the injury. We believe that the intact lateral and medial retinacula may have helped with function. To our knowledge this is the first case with a traumatic patellar tendon rupture at such a young age and furthermore with the injury being 7 months old before diagnosis. We think that a traumatic lesion is the only plausible explanation as no trochlear dysplasia or fatty atrophy of the quadriceps was present, arguing against a congenital or long-lasting abnormality. In addition, patella alta was first noted by the parents after a relevant trauma and not before. Lastly the boy did not have clinical signs of collagen disease and there was no history of this in the family. Thus a pathological tendon elongation seems unlikely (ElGuindy et al. 2011). Muratli et al. (2005) present a remotely similar case but their patient had bilateral traumatic patellar tendon rupture with the right side being of 1 month’s duration, and the patient was 9 years old. The patient showed patella alta, a depression between the patella and tibial tubercle, and unlike our case marked atrophy of the quadriceps muscle and no ability to actively extend the injured knee. We suspect that the boy’s patella alta, unchanged during the 6.5-year follow-up, was due to stretching of the ligament and one could argue that augmentation of the patellar tendon rupture with tendon allograft could have prevented this. Furthermore, it could be argued that the repair was fixed too proximally and therefore placed the patella too proximally. However, knee motion and muscle strength became normal.
JHB did the research and wrote the manuscript. OR treated the patient and revised the manuscript.
Discussion Traumatic rupture of the patellar tendon is extremely rare in preadolescent children (Felt and Arora 2017). It represents 0.6% of the musculoskeletal tendinous injuries in the general population with peak age incidence around 50 years for males and 70 for females (Clayton and Court-Brown 2008). Considerable force equivalent to 18 times body weight is required to rupture a human adult patellar tendon and as such is very rare in childhood (Pires e Albuquerque et al. 2015). Ruptures of the tendons in children or adolescents are rare; fractures are more common (Williams et al. 2015, Yousef and Rosenfeld 2017). This usually occurs distal to the patellar tendon by an avulsion fracture of the tibial tuberosity or proximally as a patellar sleeve fracture (Berg 1995). Optimal treatment of acute isolated patellar tendon rupture is prompt surgical repair.
12281 HOLBECK D.indd 455
Informed consent for publication of this case study was obtained from the boy’s parents.
Acta thanks Johannes Mayr and Yrjänä Nietosvaara for help with peer review of this study.
Berg E E. Bipolar infrapatellar tendon rupture. J Pediatr Orthop 1995; 15(3): 302-3. Clayton R A E, Court-Brown C M. The epidemiology of musculoskeletal tendinous and ligamentous injuries. Injury 2008; 39(12): 1338-44. ElGuindy A, Lustig S, Servien E, Fary C, Weppe F, Demey G, Neyret P. Treatment of chronic disruption of the patellar tendon in osteogensis imperfecta with allograft reconstruction. Knee 2011; 18(2): 121-4. Felt J, Arora R. Patellar tendon rupture in a pre-adolescent male. J Pediatr 2017; 180: 283-283.e1
Muratli H H, Celebi L, Hapa O, Biçimoğlu A. Bilateral patellar tendon rupture in a child: a case report. Knee Surg Sports Traumatol Arthrosc 2005; 13(8): 677-82. Pires e Albuquerque R S, de Araújo GC, Labronici P J, Gameiro V S. Patellar ligament rupture in an adolescent. BMJ Case Rep 2015; 2015. doi: 10.1136/bcr-2014-208070.
12281 HOLBECK D.indd 456
Acta Orthopaedica 2018; 89 (4): 454–456
Williams D, Kahane S, Chou D, Vemulapalli K. Bilateral proximal tibial sleeve fractures in a child: a case report. Arch Trauma Res 2015; 23;4(3): e27898. Yousef M A A, Rosenfeld S. Acute traumatic rupture of the patellar tendon in pediatric population: case series and review of the literature. Injury 2017; 48(11): 2515-21.
Acta Orthopaedica 2018; 89 (4): 457–461
Osteoblast precursors and inflammatory cells arrive simultaneously to sites of a trabecular-bone injury Magnus BERNHARDSSON, and Per ASPENBERG
Orthopedics, Department of Clinical and Experimental Medicine, Faculty of Health Sciences, Linköping University, Linköping, Sweden. Correspondence: firstname.lastname@example.org Submitted 2018-01-31. Accepted 2018-05-08.
Background and purpose — Fracture healing in the shaft is usually described as a sequence of events, starting with inflammation, which triggers mesenchymal tissue formation in successive steps. Most clinical fractures engage cancellous bone. We here describe fracture healing in cancellous bone, focusing on the timing of inflammatory and mesenchymal cell type appearance at the site of injury Material and methods — Rats received a proximal tibial drill hole. A subgroup received clodronate-containing liposomes before or after surgery. The tibiae were analyzed with micro-CT and immunohistochemistry 1 to 7 days after injury. Results — Granulocytes (myeloperoxidase) appeared in moderate numbers within the hole at day 1 and then gradually disappeared. Macrophage expression (CD68) was seen on day 1, increased until day 3, and then decreased. Mesenchymal cells (vimentin) had already accumulated in the periphery of the hole on day 1. Mesenchymal cells dominated in the entire lesion on day 3, now producing extracellular matrix. A modest number of preosteoblasts (RUNX2) were seen on day 1 and peaked on day 4. Osteoid was seen on day 4 in the traumatized region, with a distinct border to the uninjured surrounding marrow. Clodronate liposomes given before the injury reduced the volume of bone formation at day 7, but no reduction in macrophage numbers could be detected. Interpretation — The typical sequence of events in shaft fractures was not seen. Mesenchymal cells appeared simultaneously with granulocyte and macrophage arrival. Clodronate liposomes, known to reduce macrophage numbers, seemed to be associated with the delineation of the volume of tissue to be replaced by bone.
Most fracture healing studies in animal models concern cortical bone in shafts. However, most fractures in patients occur in cancellous bone in the metaphysis, such as the distal radius or in the vertebrae. A growing body of evidence suggests that there are important differences between the healing processes in cortical and cancellous bone. Shaft fracture healing relies on the recruitment of cells from external sources, such as the periosteum, surrounding soft tissues and blood circulation, and proceeds by forming a large, expanding callus (Kumagai et al. 2008). In contrast, the healing of cancellous bone, which is rich in mesenchymal stem cells, is strictly localized within the confinements of the injured region (Bernhardsson et al. 2015). John Charnley observed this phenomenon during the 1950s when working with knee arthrodeses. The cancellous bone formation in the traumatized regions rarely extended more than a couple of millimeters, meaning that just a small gap between Charnley’s resection surfaces could jeopardize healing (Charnley and Baker 1952). The complex interactions between bone and immune cells are incompletely understood, but macrophages seem to play a major role in bone healing. The depletion of macrophages with clodronate liposomes impairs the healing of cancellous bone, both mechanically and morphometrically (Sandberg et al. 2017). With this descriptive study we explored the timing of cell arrival at the site of cancellous bone injury and the effect of attempted monocyte depletion, to get an idea of the possible interaction between inflammation and new bone formation.
© 2018 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 10.1080/17453674.2018.1481682
12548 Bernhardsson D.indd 457
Material and methods Experimental overview 2 experiments were conducted for this paper. The first monitored the healing of drill holes in cancellous bone during the first 5 days. 30 male SOPF Sprague-Dawley rats, 10 weeks old (Janvier, Saint-Berthevin Cedex, France), weighing 444 g (SD 23) received bilateral drill holes in their proximal tibiae. The rats were randomized to be killed either 1, 2, 3, 4, or 5 days after surgery, 6 animals each day. The tibiae were harvested and prepared for histology and immunohistochemistry. The second experiment monitored the effect of macrophage depletion in healing of drill holes in cancellous bone. 36 male SOPF Sprague-Dawley rats, 10 weeks old (Janvier, SaintBerthevin Cedex, France), weighing 503 g (SD 25) received a drill hole in their right proximal tibiae. 24 of the animals received a single injection in the tail vein of 20 mg/kg of clodronate liposomes (5 mg/mL; Liposoma BV, Amsterdam, Netherlands) either 24 hours prior to surgery or 24 hours after surgery. The animals were killed on day 3 or 7 after surgery (n = 6). The tibiae were harvested, analyzed with micro-CT and prepared for histology and immunohistochemistry. Surgical procedure The surgery was performed under isoflurane anesthesia and aseptic conditions. A 1.2 mm drill hole was made by hand, using a 18G syringe needle, about 4 mm below the growth plate in the antero-medial surface of the proximal tibia. The skin was then sutured, and analgesia was given every 8–12 hours for the following 48 hours. This procedure has been described in greater detail before (Bernhardsson et al. 2015). Micro-CT The clodronate-treated tibiae were analyzed with micro-CT (Skyscan 1174, v. 2; Bruker, Aarteselaar, Belgium). In a 180° scan, a pixel size of 11.2 µm, aluminum filter 0.5 mm, rotation step of 0.4° and frame averaging of 3, and energy settings of 50 kV and 800 µA were used to acquire radiographic images. NRecon (Skyscan, v. 18.104.22.168; Aarteselaar, Belgium) was used to reconstruct the images and correct them for ring artifacts and beam hardening. Within the former drill holes, a volume of interest (VOI) was defined as a cylinder with a diameter of 1.2 mm and 1.5 mm in length into the bone marrow cavity, starting from the endosteal side. 2 hydroxyapatite standards of known density (0.25 and 0.75 g/cm3) were used to calibrate the bone mineral density. The total bone volume (BV/TV) of the VOIs was analyzed in CTAn (Skyscan, v. 1.10; Aarteselaar). Histology The tibiae were fixed in 4% paraformaldehyde for 24 hours before they were put in 10% EDTA for 10 days for decalcification. The demineralized tibiae were dehydrated in a series of
12548 Bernhardsson D.indd 458
Acta Orthopaedica 2018; 89 (4): 457–461
increasing concentration of ethanol and embedded in paraffin for sectioning. The tibiae were sectioned longitudinally, perpendicular to the drill holes, in 4 µm sections and stained with hematoxylin and eosin. Immunohistochemistry The tibiae were prepared for immunohistochemical staining of granulocytes (myeloperoxidase, MPO), macrophages (CD68), mesenchymal cells (vimentin) and preosteoblasts (runt-related transcription factor 2, RUNX2). The paraffin embedded tibiae were sectioned longitudinally in 4 µm sections. Care was taken to make the sections perpendicular to the direction of the drill hole. Sections were dewaxed and rehydrated in a series of decreasing concentration of ethanol. Heat mediated antigen retrieval was performed by incubating the slides in Tris-EDTA buffer, pH 9, for 60 min. Blocking of endogenous peroxidase in 3% hydrogen peroxide, permeabilization with 0.25% Triton X-100 and blocking with Protein Block, Serum-free (X0909, DAKO, Santa Clara, CA, USA) was performed before the sections where incubated with primary antibodies for 80 min (overnight for RUNX2) in room temperature. The following antibodies were used: anti-MPO (1:200, ab9535, Abcam, Cambridge, UK), anti-CD68 (1:250, ab125212, Abcam, Cambridge, UK), anti-vimentin (1:1000, ab92547, Abcam, Cambridge, UK), and anti-RUNX2 (1:100, ab23981, Abcam, Cambridge, UK). The sections were rinsed in TBS and then incubated with biotin-labeled goat anti-rabbit secondary antibody (1:200, E0432, DAKO, Santa Clara, CA, USA). After rinsing, the sections were incubated with Vectastain Elite ABC Kit (PK-6100, Vector Laboratories, Burlingame, CA, USA) according to the manufacturer’s instructions. Staining was visualized with Vector® VIP Peroxidase Substrate Kit (SK-4600, Vector Laboratories, Burlingame, CA, USA) and counterstained with methyl green. The quantification of stained MPO, CD68, vimentin, and RUNX2 positive cells was conducted using a light microscope, equipped with a color camera. Images were taken with 12.5X magnification, 1 image inside and 1 outside of the traumatized region for each specimen. The aim was to pick sites that were homogenous, without any trabeculae or artifacts interfering. The examiner was blinded regarding time point. Positively stained cells were quantified in the images using image analysis software (ImageJ; https://imagej.nih.gov/ij/ index.html). Statistics This was a descriptive study, and no hypothesis was specified in advance. We therefore refrained from formulating and testing hypotheses in retrospect, and thus present 95% confidence intervals (CI) for the differences between group means, based on t-distributions, but no p-values. Ethics, funding, and potential conflicts of interest All procedures were approved by the Research Ethics Board in
Acta Orthopaedica 2018; 89 (4): 457–461
Figure 1. Immunohistochemistry image (vimentin) of drill hole, 2 days after trauma. Spindle-shaped mesenchymal cells could be seen in the periphery of the lesion (blue marking), forming a circle at the interface.
Linköping, Sweden, in accordance with the Swedish Animal Welfare Act (1988:534) and EU-Directive 2010/63/EU. The registration ID is 49-15. This study was supported by the Swedish Research Council (2031-47-5), AFA insurance company, EU 159 7th framework program (FP7/2007-2013, grant 279239) and a specific grant from Linköping 160 University. No conflicts of interest were declared.
Results Exclusions 2 animals in the –24 h clodronate liposome group died postoperatively for unknown reasons. Cell arrival in the drill hole Day 1 — The drill hole was easily detected as a circular area with mainly necrotic material, erythrocytes, and scattered cells. Granulocytes (myeloperoxidase, MPO) and macrophages (CD68) were present in moderate numbers. Mesenchymal cells (vimentin-positive cells with a morphology spanning from spindle-shaped cells to osteocytes) and preosteoblasts (RUNX2; mainly a subcategory of vimentin-labeled cells) were all present in moderate numbers. Mesenchymal cells, mostly with a spindle-shaped morphology, were particularly present in the periphery of the lesion, while some cells with round morphology could be seen in the center. Day 2 — Granulocytes had decreased in numbers, while macrophages showed a slight increase. Mesenchymal cells and preosteoblasts (a minority of the mesenchymal cells) both showed more than a 2-fold increase from day 1, and now had infiltrated further into the lesion. The mesenchymal cells in the periphery of the lesion now appeared as a circle around it (Figure 1). Day 3 — Granulocytes kept on decreasing. Macrophages had increased in numbers 3-fold compared with day 2. Mesenchymal cells could now be seen in the whole traumatized region, still with mainly spindle-shaped morphology, but now producing extracellular matrix. Many of them were also
12548 Bernhardsson D.indd 459
labeled as preosteoblasts, which were seen in the entire lesion and increased in numbers. Day 4 — Granulocytes were now very few, and macrophages had decreased in numbers. Most mesenchymal cells were preosteoblasts. They had increased in numbers and now occupied the whole region. Formation of osteoid tissue could now be seen within the borders of the traumatized region. Day 5 — Granulocytes were almost absent. Macrophages were few. Vimentin-positive cells with a more osteoblastic (cuboidal) or osteocytic (round) morphology could now be seen in the region. Many of these cells were no longer RUNX2-positive (Figure 2). Outside the drill hole No greater alteration in numbers of cells could be seen over time in the intact marrow, outside of the traumatized region, for any of the cell populations (Figure 5, see Supplementary data). Effect of clodronate liposome treatment Micro-CT — The drill holes were filled with new woven bone after 7 days in the controls. Clodronate liposomes given 24 hours before trauma reduced bone volume (BV/TV) by 33% (CI: 14 to 79) compared with controls, but clodronate liposomes administered 24 hours after trauma had no such effect (Figures 3–4).
Discussion We characterized the cell composition within the spatial confinement of a cancellous bone injury. Mesenchymal cells and inflammatory cells arrived simultaneously at the injury site within the first 24 hours. Clodronate liposomes administered before trauma reduced the regenerated bone volume after 7 days. The simultaneous arrival of mesenchymal and inflammatory cells is an important difference from the healing pattern of shaft fractures. In the literature, shaft fracture healing is
Acta Orthopaedica 2018; 89 (4): 457â&#x20AC;&#x201C;461
Figure 3. Micro-CT analysis of drill holes in proximal tibia, 7 days after trauma. Clodronate liposomes given 24 hours prior to trauma reduced the bone formation (BV/TV) by 33% (CI 14â&#x20AC;&#x201C;79) compared with controls. However, this effect was absent when clodronate liposomes were administered 24 hours after trauma.
Figure 2. Quantification of cell populations in drill holes in proximal tibia. Granulocytes (A; myeloperoxidase, MPO). Macrophages (B; CD68). Mesenchymal cells (C; vimentin). Preosteoblasts (D; RUNX2).
Figure 4. Micro-CT images of drill holes in proximal tibia, 7 days after trauma. Clodronate liposomes administered 24 hours prior to trauma impaired the bone formation in the marrow compartment, but not when given 24 hours after trauma.
described as a sequence of events with overlapping phases, where inflammatory cells are the first to arrive at the injured site before mesenchymal cells are recruited and initiate healing (Kumagai et al. 2008, Ono and Takayanagi 2017). Recruitment of mesenchymal cells from distant sources seems to be required, since mesenchymal stem cells are few at the site of a cortical shaft fracture. If the inflammatory response is attenu-
12548 Bernhardsson D.indd 460
ated, using NSAIDs, the healing of shaft fractures is impaired (Sandberg and Aspenberg 2015). There is, therefore, a need of recruitment signals from inflammatory cells that have arrived first at the injured site. However, mesenchymal cell activation is not necessarily dependent on inflammatory cells. Mesenchymal cells can respond directly to stimuli associated with injury, such as damage-associated molecular patterns (DAMPs) and mechanical stimuli (Lotfi et al. 2011, Pistoia and Raffaghello 2011, Delaine-Smith and Reilly 2012). In contrast to the cortical shaft, cancellous bone is rich in mesenchymal cells with a stronger osteogenic potential. Therefore, a more efficient healing response can be expected in cancellous bone compared with shaft fractures (Siclari et al. 2013). Stability of fractures is a factor worth mentioning, and since our drill hole model is of a more stable nature, could we expect different behavior if we introduced instability to some degree to the injury? We cannot see how the addition of instability in a cancellous bone injury would affect the timing of inflammatory or mesenchymal cell arrival. However, a change in cell composition and tissue formation could be expected. Probably there would be more chondrocytes initially at the site, producing cartilaginous tissue, securing a primary stable structure before bone formation (Claes et al. 2012). The new bone in the former drill hole had the shape of a cylinder with the same radius as the hole. Clodronate liposomes led to a volume reduction in the new bone cylinder, so that the entire lesion was not replaced with bone, even though the bone that was formed looked qualitatively the same on CT and histology. We have previously seen a reduction in new-
Acta Orthopaedica 2018; 89 (4): 457–461
formed bone volume due to clodronate liposomes in a screw fixation model in mice, where the reduction in bone volume led to a poorer fixation of the screw and a lower pull-out force (Sandberg et al. 2017). We then speculated from our results that macrophages may have a role in inducing mesenchymal proliferation. However, similar to this study, we could not see any local effect of clodronate on either macrophages or mesenchymal cell numbers in the lesion or surrounding intact tissue, either on day 1 or on day 3 after surgery. Prior to this study we thought that clodronate liposomes depleted macrophages generally in the circulation and tissues, including the bone marrow. However, in our case, where only 1 injection of clodronate liposomes was administered, it seems that circulating monocytes are primarily depleted (Sunderkötter et al. 2004). The liposomes cannot penetrate the vascular endothelium, and therefore a longer treatment period with repeated injections is needed for successful depletion of resident macrophages in tissues. Other evidence shows that tissue-resident macrophages are segregated from circulating monocytes and can repopulate by themselves through local proliferation (Hashimoto et al. 2013). Some subpopulations of macrophages might even increase in numbers upon monocyte depletion (Côté et al. 2013). It is an enigma how the mesenchymal cells manage to restrict bone formation only to the drill hole, with such a sharp demarcation to the surrounding tissue. We now speculate that circulating monocytes, arriving in the hematoma from ruptured blood vessels into the traumatized region, deposit molecules in the hematoma like a “scent blueprint,” which determines the shape of the volume to be filled with new bone several days later. A disturbed “scent blueprint” might explain why monocyte depletion only at the time of trauma leads to reduced new bone volume several days later. The hematoma is known to be important for fracture healing, and early hematoma ablation in shafts fractures leads to poor healing (Grundnes and Reikerås 1993). Our data suggest that monocytes in the hematoma are also important for osteogenesis in cancellous bone. A limitation of this study is the lack of data on a corresponding drill hole injury in cortical bone, as a control, to compare the influx and timing of inflammatory and osteoprogenitor cells. Other limitations are the lack of data which confirms monocyte depletion, and that we have done this study in rats. Most of the earlier publications concerning monocyte/macrophage depletion have been conducted in mice, and we do not know how, or if, clodronate liposomes might have a different effect between species, even though our data correspond with earlier studies. In summary, mesenchymal and inflammatory cells appear to be activated simultaneously upon trauma in cancellous bone. This is different from the sequential events in shaft fracture healing. Early monocyte depletion reduces late bone volume, suggesting that the presence of monocytes in the hematoma during the first days is crucial.
12548 Bernhardsson D.indd 461
Supplementary data Figure 5 is available as supplementary data in the online version of this article, http://dx.doi.org/ 10.1080/17453674.2018. 1481682
Both authors were involved in designing the study, analyzing the data, preparing and approving the submitted manuscript. MB conducted the experiments and acquired the data. Acta thanks Olav Reikerås and other anonymous reviewers for help with peer review of this study.
Bernhardsson M, Sandberg O, Aspenberg, P. Experimental models for cancellous bone healing in the rat. Acta Orthop 2015; 86(6): 745-50. Charnley J, Baker S L. Compression arthrodesis of the knee: a clinical and histological study. J Bone Joint Surg Br 1952; 34-B(2): 187-99. Claes L, Recknagel S, Ignatius A. Fracture healing under healthy and inflammatory conditions. Nat Rev Rheumatol 2012; 8(3): 133-43. Côté C H, Bouchard P, van Rooijen N, Marsolais D, Duchesne E. Monocyte depletion increases local proliferation of macrophage subsets after skeletal muscle injury. BMC Musculoskelet Disord 2013; 14: 359. Delaine-Smith R M, Reilly G C. Mesenchymal stem cell responses to mechanical stimuli. Muscles Ligaments Tendons J 2012; 2(3): 169-80. Grundnes O, Reikerås O. The importance of the hematoma for fracture healing in rats. Acta Orthop Scand 1993; 64(3): 340-2. Hashimoto D, Chow A, Noizat C, Teo P, Beasley M B, Leboeuf M, Becker C D, See P, Price J, Lucas D, Greter M, Mortha A, Boye, S W, Forsberg E C, Tanaka M, van Rooijen N, García-Sastre A, Stanley E R, Ginhoux F, Frenette P S, Merad M. Tissue-resident macrophages self-maintain locally throughout adult life with minimal contribution from circulating monocytes. Immunity 2013; 38(4): 792-804. Kumagai K, Vasanji A, Drazba J A, Butler R S, Muschler G F. Circulating cells with osteogenic potential are physiologically mobilized into the fracture healing site in the parabiotic mice model. J Orthop Res 2008; 26(2): 165-75. Lotfi R, Eisenbacher J, Solgi G, Fuchs K, Yildiz T, Nienhaus C, Rojewski M T, Schrezenmeier H. Human mesenchymal stem cells respond to native but not oxidized damage associated molecular pattern molecules from necrotic (tumor) material. Eur J Immunol 2011; 41(7): 2021-8. Ono T, Takayanagi H. Osteoimmunology in bone fracture healing. Curr Osteoporos Rep 2017; 15(4): 367-75. Pistoia V, Raffaghello L. Damage-associated molecular patterns (DAMPs) and mesenchymal stem cells: a matter of attraction and excitement. Eur J Immunol 2011; 41(7): 1828-31. Sandberg O, Aspenberg P. Different effects of indomethacin on healing of shaft and metaphyseal fractures. Acta Orthop 2015; 86(2): 243-7. Sandberg O H, Tätting L, Bernhardsson M E, Aspenberg P. Temporal role of macrophages in cancellous bone healing. Bone 2017; 101: 129-33. Siclari V A, Zhu J, Akiyama K, Liu F, Zhang X, Chandra A, Nah H D, Shi S, Qin L. Mesenchymal progenitors residing close to the bone surface are functionally distinct from those in the central bone marrow. Bone 2013; 53(2): 575-86. Sunderkötter C, Nikolic T, Dillon M J, Van Rooijen N, Stehling M, Drevets D A, Leenen P J. Subpopulations of mouse blood monocytes differ in maturation stage and inflammatory response. J Immunol 2004; 172(7): 4410-17.
Acta Orthopaedica 2018; 89 (4): 462–467
Osteochondral lesions of the talus Few patients require surgery Sang Gyo SEO 1, Jin Soo KIM 1, Dong-Kyo SEO 2, You Keun KIM 1, Sang-Hoon LEE 3, and Ho Seong LEE 1
of Orthopedic Surgery, Asan Medical Center, College of Medicine, University of Ulsan, Seoul; 2 Department of Orthopedic Surgery, Gangneung Asan Hospital, College of Medicine, University of Ulsan, Gangneung; 3 Department of Radiology, Asan Medical Center, College of Medicine, University of Ulsan, Seoul, Republic of Korea Correspondence: HSL: email@example.com Submitted 2017-11-28. Accepted 2018-02-26.
Background and purpose — The frequency of progression of osteoarthritis and persistence of symptoms in untreated osteochondral lesion of the talus (OCL) is not well known. We report the outcome of a nonoperative treatment for symptomatic OCL. Patients and methods — This study included 142 patients with OCLs from 2003 to 2013. The patients did not undergo immobilization and had no restrictions of physical activities. The mean follow-up time was 6 (3–10) years. Initial MRI and CT confirmed OCL and showed lesion size, location, and stage of the lesion. Progression of osteoarthritis was evaluated by standing radiographs. In 83 patients, CT was performed at the final follow-up for analyses of the lesion size. We surveyed patients for limitations of sports activity, and Visual Analogue Scales (VAS), AOFAS, and SF-36 were assessed. Results — No patients had progression of osteoarthritis. The lesion size as determined by CT did not change in 69/83 patients, decreased in 5, and increased in 9. The mean VAS score of the 142 patients decreased from 3.8 to 0.9 (p < 0.001), the mean AOFAS ankle–hindfoot score increased from 86 to 93 (p < 0.001), and the mean SF-36 score increased from 52 to 72 (p < 0.001). Only 9 patients reported limitations of sports activity. The size and location of the lesion did not correlate with any of the outcome scores. Interpretation — Nonoperative treatment can be considered a good option for patients with OCL. ■
Osteochondral lesions of the talus (OCL) occur in the articular cartilage and subchondral bone of the talus and are commonly associated with ankle injuries, such as sprains and fractures (Bruns 1997, van Dijk et al. 2010). Treatment of OCL may be operative or nonoperative. With recent advances in arthroscopic techniques, the number of operations for OCL
is increasing (van Dijk et al. 1997, Best et al. 2015, Werner et al. 2015). One indication for operative treatment of OCL is bone fragments within the joint, causing symptoms such as locking. However, other indications remain controversial (Zengerink et al. 2010, Badekas et al. 2013, Hannon et al. 2014). Operative treatment of OCL is costly, and complications such as iatrogenic nerve and articular cartilage injury may occur (Barber et al. 1990, Ferkel et al. 1996, Ferkel et al. 2001, Young et al. 2011, Deng et al. 2012, Vega et al. 2016). In addition, patient satisfaction rates following operative treatment vary between 54% and 89% (Verhagen et al. 2003, Zengerink et al. 2010). Although several studies have reported on nonoperative treatment of OCL, these were limited by small numbers of patients, short follow-up periods, or inclusion of asymptomatic participants (McCullough and Venugopal 1979, Bauer et al. 1987, Pettine and Morrey 1987, Shearer et al. 2002, Elias et al. 2006, Klammer et al. 2015). In addition, the progress of OCL is still unclear after nonoperative treatment. In this study, we report the outcome of patients who were diagnosed with symptomatic OCL but were not treated.
Patients and methods Patient demographics Diagnosis of OCL was initially confirmed for all patients by MRI or CT. Inclusion criteria were as follows: (1) consecutive patients with symptomatic OCL who visited the orthopedic outpatient clinic between 2003 and 2013; (2) patients with a follow-up period of ≥ 3 years; and (3) patients who had received nonoperative treatment. Patients with advanced osteoarthritis (over Kellgren-Lawrence Grade 2) (Kellgren
© 2018 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 10.1080/17453674.2018.1460777
12365 Seo D.indd 462
Acta Orthopaedica 2018; 89 (4): 462–467
Osteochondral lesion of the talus n = 244 Non-operative treatment n = 217 Follow-up (> 3 years) (n = 142): – outcome questionnaire, 142 – plain radiography, 142 – subset with CT, 83
Operative treatment n = 27
Lost to follow-up but had a telephone survey (n = 75): – responders, 56 – non-responders, 19
Figure 1. Flowchart of study participants.
and Lawrence 1957) at the time of diagnosis and those undergoing surgery during the same period were excluded. 244 patients were diagnosed with OCL, among whom 217 (89%) and 27 (11%) received nonoperative and operative treatments, respectively. Among the 27 patients who were surgically treated, 17 had large lesions, severe symptoms, and advanced lesion stage at the time of diagnosis and 10 underwent surgery because of symptomatic deterioration (8 patients) and increased lesion size (2 patients) during the follow-up period. Radiographic follow-up was possible in 142 of 217 patients who received nonoperative treatment (Figure 1). Protocol for nonoperative treatment Patients receiving nonoperative treatment for OCL did not undergo immobilization or restriction of sports activities. From the initial visit, we recommended unlimited daily activity and prescribed the use of NSAIDs as needed for intermittent pain. We explained the natural course of OCL and recommended nonoperative treatment to the patients. Briefly, the nonoperative treatment was composed of “skillful neglect” along with tolerable levels of exercise. Operative treatment was selectively performed in patients whose pain persisted or those who requested surgery for different reasons; some patients wanted to undergo surgery for insurance or military problems. Staging and radiographic evaluation Standing radiographs of the ankle joint (anteroposterior, lateral, and mortise) were evaluated based on the readings of the musculoskeletal radiologist. Osteoarthritis progression (over Kellgren-Lawrence grade 2) (Kellgren and Lawrence 1957) was assessed using plain weight-bearing radiography. The stage of OCL was determined using the Berndt and Harty staging system (Berndt and Harty 1959) with plain weight-bearing radiography and the Ferkel and Sgaglione staging system (Ferkel et al. 1990) with CT. Subchondral cyst and bone-marrow edema were identified using initial MRI. A 9-zone system was used to categorize the location of the lesion, and the lesion size was defined as width, length, and depth. Follow-up CT was used to assess changes in the lesion and cyst size in 83 patients. CT was paid from research funding, and patients who agree with it participated.
12365 Seo D.indd 463
Clinical evaluation Baseline data were recorded at the initial assessment, including history of trauma, symptoms related to instability, and limitation of preferred sports activity. Ankle instability, range of motion, and heel alignment were evaluated by physical examination. VAS (Visual Analogue Scales), AOFAS (Kitaoka el al. 1994), and SF-36 scores (Turner-Bowker et al. 2002) were compared between the initial and final follow-up visits. We investigated the VAS score and limitation of sports activity for the 75 patients lost to follow-up by telephone. 19 did not respond to the telephone survey (Figure 1). Statistics VAS, AOFAS ankle–hindfoot, and SF-36 scores at the initial and final follow-up visits were compared using paired t-tests. The Pearson’s correlation coefficient (r) was used for evaluation of correlations between predisposing factors (sex, age, height, weight, BMI, stage and size of the lesion, and accompanying injury) and the outcome of questionnaires. Interclass correlation coefficients ranged from −1 to +1, with +1 indicating a perfectly positive correlation and −1 indicating a perfectly negative correlation. A value of 0 indicated no correlation. 3 orthopedic surgeons in our author group with 9, 7, and 5 years of experience, respectively, assessed the intraclass correlation coefficient (ICC) of inter-observer reliability for the sizes and stages of the lesions. Prior sample-size estimation and a statistical power analysis indicated the need for assessment of minimum 36 ankle radiographs with CT and MRI to ensure precision in our data. In addition, validation of the differences in lesion size observed with CT and MRI was confirmed by ICC. An ICC value of 1 indicated perfect reliability, and an ICC greater than 0.8 indicated excellent reliability (Donner and Klar 2000). A value of p < 0.05 was considered significant. Statistical analyses were performed using IBM SPSS (Statistics for Windows, Version 21.0; IBM Corp, Armonk, NY, USA). Ethics, funding, and potential conflicts of interest This retrospective study was approved by the Institutional Review Board of Asan Medical Center, which is a tertiary referral hospital. The study has received research support funding from the Asan Institute for Life Sciences, Asan Medical Center, Korea (2016-0634). The authors declared no potential conflicts of interest.
Results Demographics (Table 1) Of the 217 patients who received nonoperative treatment, 142 patients (82 males) underwent plain weight-bearing radiography at the final follow-up and were evaluated for clinical outcomes and progression of osteoarthritis.
Acta Orthopaedica 2018; 89 (4): 462–467
Table 1. Demographic data of the 142 subjects presented as mean values (SD) Factor
Sex: male/female Age (year) Height (cm) Weight (kg) Body mass index Follow-up period (year)
82/60 47 (15) 164 (11) 64 (10) 23 (2.8) 5.7 (2.1)
83 (49 males) of these patients also underwent CT at the final follow-up and were evaluated for changes in lesion size. Stage and location of lesion at the initial diagnosis The number of patients with stages I, II, III, and IV lesions was 33, 67, 18, and 24, respectively. Medial lesions were present in 126 patients (89%) and lateral lesions in 16 (11) (Table 2, see Supplemantary data). Initially, the mean width, length, and depth measurements of OCL were 6.9 (1.7–13.1) mm, 9.4 (1.9–19.2) mm, and 5.4 (1.0–15.5) mm, respectively.
Figure 2. A 29-year-old man with OCL. (A) Initial standing radiograph and (C) initial MRI showed medial OCL. (D) CT at 1-year follow-up, (B) standing radiograph and (E) CT at 9-year follow-up showed no change in lesion size. The AOFAS ankle– hindfoot score improved from 92 to 100.
Follow-up radiologic evaluation The weight-bearing radiographs did not demonstrate progression of ankle osteoarthritis in any of the 142 patients.
Of the 83 patients who underwent follow-up CT (see Figure 1), 69 patients had no change in lesion size (Figure 2), 5 had decreased lesion size (Figure 3), and 9 had increased lesion size (Figures 4 and 5). Radiographic measurements of the lesion size showed excellent interobserver reliability (ICC, 0.901–0.933).
Figure 3. A 53-year-old man with OCL. (A) Initial standing radiograph and (C) CT. (B) Standing radiograph and (D) CT at 7-year follow-up. The lesion size decreased, and the AOFAS ankle–hindfoot score improved from 73 to 100.
Figure 4. A 48-year-old man with OCL. (A) Initial standing radiograph and (C) MRI. (B) Standing radiograph and (D) CT at 3.8-year follow-up. Although the lesion size increased, the AOFAS ankle–hindfoot score improved from 71 to 80.
12365 Seo D.indd 464
Acta Orthopaedica 2018; 89 (4): 462–467
Figure 5. A 59-year-old man with OCL. (A) Initial MRI and (B) CT at 4-year follow-up. Although we recommended surgery to the patient at the first visit, the patient refused the operation because the symptoms were tolerable. The lesion size was increased on the CT at 4-year follow-up. However, the patient did not have worse symptoms and still did not want surgery.
than those in male patients (p = 0.01). Moreover, a negative correlation was observed between age and SF-36 scores (p< 0.001). However, there was no significant correlation between AOFAS ankle– hindfoot scores and sex (p = 0.2) or age (p = 0.05). Height, weight, and BMI had no influence on the results of the outcome questionnaires. In addition, the location and size of the lesion at the initial visit did not affect outcome scores. There were also no clinical correlations between outcome scores and associated injury (bone-marrow edema and bone cyst). At the final follow-up, the higher the stage, the higher the VAS (p = 0.02) and the lower the AOFAS ankle–hindfoot scores (p = 0.04) (Table 3, see Supplemantary data).
Initial depth of OCLs was larger (p = 0.02) in the increasedsize group compared with that in the group with no change in lesion size, but there was no significant difference in the initial width (p = 0.09) or length (p = 0.1) of OCL between the 2 groups. In the group with decreased lesion size, there was no significantly difference in the initial width (p = 0.2), length (p = 0.2), or depth (p = 0.6) of OCL in the decreased-size group compared with the no-change group. All patients in the increased-size group had medially located lesions. Among these, 6 patients were in stage II, 3 in stage I, and none in stage III or IV. Age, sex, and BMI did not correlate with changes in lesion size.
Telephone evaluation for radiologic follow-up lost patients Among 56 of the 75 patients lost to radiological follow-up but who responded to the telephone survey, 50 were still undergoing nonoperative treatment. None of the patients reported limitation of their preferred sports activity, and the mean VAS score was 1.0 (0–3). Among the 56 patients who responded, 6 had undergone operative treatment at another hospital. The mean follow-up period of the telephone survey was 5.0 (3.5– 9.4) years.
Clinical evaluation A history of trauma was detected in 116 of the 142 patients. The most common reported injury was ankle sprain (90 patients), followed by ankle fracture (13 patients) and contusion (11 patients). 23 patients had symptoms of ankle instability.
Numerous studies have reported outcomes for operative treatment of OCL (Kumai et al. 1999, Taranow et al. 1999, Murawski and Kennedy 2013) but few have reported on results of nonoperative treatment and natural history, and the numbers of patients in these studies are low (McCullough and Venugopal 1979, Bauer et al. 1987, Elias et al. 2006). Patient satisfaction with nonoperative treatment was high and the lesions showed little worsening on radiological imaging. Klammer et al. (2015) observed 48 patients with OCL using only plain radiography and VAS scores for at least 2 years (max x years) and reported that 41 had no symptoms or had a reduction in VAS scores to ≤ 3.0. Bauer et al. (1987) followed 30 patients with OCL for an average of 21 years and reported that the most of the patients’ symptoms improved, with only 2 patients developing ankle arthrosis and 1 with worsening of symptoms. Other studies on nonoperative treatments for OCL have also reported relatively favorable results (McCullough and Venugopal 1979, Pettine and Morrey 1987, Shearer et al. 2002, Elias et al. 2006). Because of concerns of possible progression to osteoarthritis or increase in lesion size, many patients diagnosed with OCL are recommended for operative treatment, even when their symptoms are not severe. In fact, among the 142 patients included in this study, 104 had been recommended operative
Outcome questionnaire The mean VAS score decreased from 3.8 (1–8) at the initial visit to 0.9 (0–4) at the final follow-up (p < 0.001). The mean AOFAS ankle–hindfoot score improved from 86 (41–93) at the initial visit to 93 (65–100) at the final follow-up (p < 0.001). The mean SF-36 score increased from 52 (30–90) to 71 (37–97) (p < 0.001). Only 9 patients reported limitation of their preferred sports activity. Even in the increased-size group, the mean VAS score decreased from 4.7 (1–8) at the initial visit to 0.8 at the final follow-up, the mean AOFAS ankle–hindfoot score improved from 82 (65–96) to 93 (71–100), and the mean SF-36 score increased from 55 (36–87) to 77 (52–96). Correlation between outcome questionnaire and predisposing factors In female patients, SF-36 scores at final follow-up were lower
12365 Seo D.indd 465
treatment at another hospital or clinic. We found no changes in talar tilt or progression to osteoarthritis in our patients with 10 years of follow-up. In addition, only 9 of 83 patients who underwent follow-up CT (see Figure 1) showed an increase in lesion size (mean follow-up, 5 years); in the others, lesion size shrank or remained the same, similar to the study by Klammer et al. (2015). The 9 patients with increased lesion size had a favorable clinical outcome, and it is noteworthy that age, sex, and BMI were not associated with an increase in lesion size. It is also clinically meaningful that the increase in lesion size mainly occurred with large cystic lesions (initial mean depth: 8.3 mm in the increased lesion-size group and 5.4 mm in all patients) but not at stages III and IV. This means that lesions in which bony fragment displacement has already occurred will have little change in size in the future, even at higher OCL stages. Therefore, it is necessary to carefully evaluate changes in stage II cystic lesion size. Although there were no differences in initial symptoms or causes of OCL in the increased-size group, it is necessary to consider the possibility that cystic lesions of increasing size may be in a different category because of the fact that lesions in the increased-size group were limited to stages I and II. This is supported by the fact that degenerated fibrous tissue with a focal osteocartilaginous component was found on histological examination of the patients who underwent surgery because of the increase in lesion size. Our study included 89% medial and 11% lateral lesions, which is a higher rate of medial lesions than in previous studies. Bernt and Harty (1959) reported a rate of lateral lesions as high as 43%, and Klammer et al. (2015) have reported a 34% rate of lateral lesions. In our study, all patients in the increased-size group had medial lesions. However, it is possible that more patients with medial lesions were included during patient selection because they had mild symptoms, and this should be taken into account when analyzing the results of our study. Only the stage of the lesion was found to be a factor affecting the outcome of nonoperative treatment. This is similar to previous studies, which also concluded that the higher the stage of the lesion, the longer the duration of symptoms. However, in our study of 142 patients, 42 had lesions that were at stages III and IV and most of these patients had good clinical outcomes. We found no change in the size or progression of osteoarthritis, even at stages III and IV. Therefore, except in cases of intra-articular loose bodies that may cause locking in the ankle joint, we also consider nonoperative treatment to be indicated for patients with stages III and IV lesions. The 10 patients who underwent surgery during the followup period were operated on an average of 8 months after the final follow-up because of aggravation of symptoms (8 patients) without increased lesion size, or increased lesion size (2 patients). Our study has some limitations. First, there was a selection bias; the study included only patients selected for non-
12365 Seo D.indd 466
Acta Orthopaedica 2018; 89 (4): 462â&#x20AC;&#x201C;467
operative treatment. Therefore, a comparison of outcomes with operative treatment was not made. In addition, because our hospital is a tertiary hospital, there is a selection bias in this regard. Second, our study had a retrospective design. In the future, randomized controlled trials of patients receiving operative and nonoperative treatment for symptomatic OCLs will be helpful to establish clinical practice guidelines. Third, our study was the result of a mid-term follow-up. Long-term follow-up for changes in lesion size and progression of osteoarthritis will be required as well. In summary, nonoperative treatment of patients with OCL had a satisfactory clinical outcome after a mean follow-up of 6 years. Radiologically, no progression to degenerative osteoarthritis was observed. Nonoperative treatment can be considered a good option for patients with OCL. Supplementary data Tables 2 and 3 are available as supplementary data in the online version of this article, http://dx.doi.org/10.1080/17453674. 2018.1460777
SGS developed the methodology, performed the analysis, and wrote the manuscript. JSK and DKS developed the methodology and performed the analysis. YKK collected the data. HSL designed the study, performed the analysis, and wrote the manuscript. Acta thanks Marc G. Romijn and other anonymous reviewers for help with peer review of this study.
http://opendata.hira.or.kr/home.do. Badekas T, Takvorian M, Souras N. Treatment principles for osteochondral lesions in foot and ankle. Int Orthop 2013; 37(9): 1697-706. doi: 10.1007/ s00264-013-2076-1. Barber F A, Click J, Britt B T. Complications of ankle arthroscopy. Foot Ankle 1990; 10(5): 263-6. Bauer M, Jonsson K, Linden B. Osteochondritis dissecans of the ankle. A 20-year follow-up study. J Bone Joint Surg Br 1987; 69(1): 93-6. Berndt A L, Harty M. Transchondral fractures (osteochondritis dissecans) of the talus. J Bone Joint Surg Am 1959; 41-A: 988-1020. Best M J, Buller L T, Miranda A. United States national trends in ankle arthroscopy: analysis of the National Survey of Ambulatory Surgery and National Hospital Discharge Survey. Foot Ankle Spec 2015; 8(4): 266-72. doi: 10.1177/1938640014560166. Bruns J. [Osteochondrosis dissecans]. Orthopade 1997; 26(6): 573-84. doi: 10.1007/PL00003414. Deng D F, Hamilton G A, Lee M, Rush S, Ford L A, Patel S. Complications associated with foot and ankle arthroscopy. J Foot Ankle Surg 2012; 51(3): 281-4. doi: 10.1053/j.jfas.2011.11.011. Donner A, Klar N. Design and analysis of cluster randomization trials in health research. London: Edward Arnold; 2000. Elias I, Jung J W, Raikin S M, Schweitzer M W, Carrino J A, Morrison W B. Osteochondral lesions of the talus: change in MRI findings over time in talar lesions without operative intervention and implications for staging systems. Foot Ankle Int 2006; 27(3): 157-66. doi: 10.1177/107110070602700301. Ferkel R D, Sgaglione N A, DelPizzo W, et al. Arthroscopic treatment of osteochondral lesions of the talus: long-term results. Orthop Trans. 1990; 14: 172-3.
Acta Orthopaedica 2018; 89 (4): 462–467
Ferkel R D, Heath D D, Guhl J F. Neurological complications of ankle arthroscopy. Arthroscopy 1996; 12(2): 200-8. Ferkel R D, Small H N, Gittins J E. Complications in foot and ankle arthroscopy. Clin Orthop Relat Res 2001; (391): 89-104. Hannon C P, Smyth N A, Murawski C D, Savage-Elliott I, Deyer T W, Calder J D, Kennedy J G. Osteochondral lesions of the talus: aspects of current management. Bone Joint J 2014; 96-B(2): 164-71. doi: 10.1302/0301620X.96B2.31637. Kellgren J H, Lawrence J S. Radiological assessment of osteoarthrosis. Ann Rheum Dis 1957; 16: 494-502. Kitaoka, H B, Alexander I J, Adelaar R S, Nunley J A, Myerson M S. Clinical rating systems for the ankle-hindfoot, midfoot, hallux and lesser toes. Foot Ankle Int 1994; 15(7): 349-53. Klammer G, Maquieira G J, Spahn S, Vigfusson V, Zanetti M, Espinosa N. Natural history of nonoperatively treated osteochondral lesions of the talus. Foot Ankle Int 2015; 36(1): 24-31. doi: 10.1177/1071100714552480. Kumai T, Takakura Y, Higashiyama I, Tamai S. Arthroscopic drilling for the treatment of osteochondral lesions of the talus. J Bone Joint Surg Am 1999; 81(9): 1229-35. McCullough C J, Venugopal V. Osteochondritis dissecans of the talus: the natural history. Clin Orthop Relat Res 1979(144): 264-8. Murawski C D, Kennedy J G. Operative treatment of osteochondral lesions of the talus. J Bone Joint Surg Am 2013; 95(11): 1045-54. doi: 10.2106/ JBJS.L.00773. Pettine K A, Morrey B F. Osteochondral fractures of the talus: a long-term follow-up. J Bone Joint Surg Br 1987; 69(1): 89-92. Shearer C, Loomer R, Clement D. Nonoperatively managed stage 5 osteochondral talar lesions. Foot Ankle Int 2002; 23(7): 651-4. doi: 10.1177/107110070202300712.
12365 Seo D.indd 467
Taranow W S, Bisignani G A, Towers J D, Conti S F. Retrograde drilling of osteochondral lesions of the medial talar dome. Foot Ankle Int 1999; 20(8): 474-80. Turner-Bowker D M, Bartley B J, Ware J E. SF-36 health survey and “SF” bibliography. 3rd ed. Lincoln, RI: QualityMetric Inc; 2002. van Dijk C N, Tol J L, Verheyen C C. A prospective study of prognostic factors concerning the outcome of arthroscopic surgery for anterior ankle impingement. Am J Sports Med 1997; 25(6): 737-45. doi: 10.1177/036354659702500603. van Dijk C N, Reilingh M L, Zengerink M, van Bergen C J. Osteochondral defects in the ankle: why painful? Knee Surg Sports Traumatol Arthrosc 2010; 18(5): 570-80. doi: 10.1007/s00167-010-1064-x. Vega J, Golano P, Pena F. Iatrogenic articular cartilage injuries during ankle arthroscopy. Knee Surg Sports Traumatol Arthrosc 2016; 24(4): 1304-10. doi: 10.1007/s00167-014-3237-5. Verhagen R A, Struijs P A, Bossuyt P M, van Dijk C N. Systematic review of treatment strategies for osteochondral defects of the talar dome. Foot Ankle Clin 2003; 8(2): 233-42, viii-ix. Werner C, Burrus M T, Park J S, Perumal V, Gwathmey F W. Trends in ankle arthroscopy and its use in the management of pathologic conditions of the lateral ankle in the United States: a national database study. Arthroscopy 2015; 31(7): 1330-7. doi: 10.1016/j.arthro.2015.01.020. Young B H, Flanigan R M, DiGiovanni B F. Complications of ankle arthroscopy utilizing a contemporary noninvasive distraction technique. J Bone Joint Surg Am 2011; 93(10): 963-8. doi: 10.2106/JBJS.I.00977. Zengerink M, Struijs P A, Tol J L, van Dijk C N. Treatment of osteochondral lesions of the talus: a systematic review. Knee Surg Sports Traumatol Arthrosc 2010; 18(2): 238-46. doi: 10.1007/s00167-009-0942-6.
Acta Orthopaedica 2018; 89 (4): 468–473
Automated detection and classification of the proximal humerus fracture by using deep learning algorithm Seok Won CHUNG 1, Seung Seog HAN 2, Ji Whan LEE 1, Kyung-Soo OH 1, Na Ra KIM 3, Jong Pil YOON 4, Joon Yub KIM 5, Sung Hoon MOON 6, Jieun KWON 7, Hyo-Jin LEE 8, Young-Min NOH 9, and Youngjun KIM 10
1 Department of Orthopaedic Surgery and 3 Department of Radiology, Konkuk University School of Medicine, Seoul; 2 Department of Dermatology, I-dermatology clinic, Seoul ; 4 Department of Orthopaedic Surgery, Kyungpook National University College of Medicine, Daegu, Korea; 5 Department of Orthopaedic Surgery, Myungji Hospital, Goyang; 6 Department of Orthopaedic Surgery, Kangwon National University College of Medicine, Chuncheon, Korea; 7 Department of Othopaedic Surgery, National Police Hospital, Seoul; 8 Department of Orthopaedic Surgery, Catholic University College of Medicine, Seoul, St Mary’s Hospital, Seoul, Korea; 9 Department of Orthopaedic Surgery, Dong-A University College of Medicine, Pusan; 10 Center for Bionics, Korea Institute of Science and Technology, Seoul, Korea Correspondence: firstname.lastname@example.org Submitted 2018-01-07. Accepted 2018-02-21.
Background and purpose — We aimed to evaluate the ability of artificial intelligence (a deep learning algorithm) to detect and classify proximal humerus fractures using plain anteroposterior shoulder radiographs. Patients and methods — 1,891 images (1 image per person) of normal shoulders (n = 515) and 4 proximal humerus fracture types (greater tuberosity, 346; surgical neck, 514; 3-part, 269; 4-part, 247) classified by 3 specialists were evaluated. We trained a deep convolutional neural network (CNN) after augmentation of a training dataset. The ability of the CNN, as measured by top-1 accuracy, area under receiver operating characteristics curve (AUC), sensitivity/specificity, and Youden index, in comparison with humans (28 general physicians, 11 general orthopedists, and 19 orthopedists specialized in the shoulder) to detect and classify proximal humerus fractures was evaluated. Results — The CNN showed a high performance of 96% top-1 accuracy, 1.00 AUC, 0.99/0.97 sensitivity/specificity, and 0.97 Youden index for distinguishing normal shoulders from proximal humerus fractures. In addition, the CNN showed promising results with 65–86% top-1 accuracy, 0.90–0.98 AUC, 0.88/0.83–0.97/0.94 sensitivity/specificity, and 0.71–0.90 Youden index for classifying fracture type. When compared with the human groups, the CNN showed superior performance to that of general physicians and orthopedists, similar performance to orthopedists specialized in the shoulder, and the superior performance of the CNN was more marked in complex 3- and 4-part fractures. Interpretation — The use of artificial intelligence can accurately detect and classify proximal humerus fractures on plain shoulder AP radiographs. Further studies are necessary to determine the feasibility of applying artificial intelligence in the clinic and whether its use could improve care and outcomes compared with current orthopedic assessments. ■
Proximal humerus fractures are primarily diagnosed using plain radiographs, and the fracture type is determined according to its anatomical location as well as fragmentation and displacement levels. However, since non-orthopedic surgeons or insufficiently experienced orthopedic surgeons are frequently the first doctors to assess fractures, it is not unusual for proximal humerus fractures to be misdiagnosed. In addition, even an experienced orthopedic surgeon can misdiagnose the fracture type due to variable presentation (Mora Guix et al. 2009, Foroohar et al. 2011). Thus, a more efficient and accurate manner of diagnosing and classifying fracture type is of interest. Deep learning is a branch of artificial intelligence that uses a cascade of many layers of nonlinear processing units to extract features and create transformations and is based on the learning of multiple levels of features or representations of the data (Wang and Summers 2012, Bengio et al. 2013, LeCun et al. 2015). Deep learning comprises a neural network with multiple hidden layers that enhance image recognition accuracy, thereby increasing its versatility for capturing representative features (Shin et al. 2013). Since 2012, deep learning has rapidly become the cutting-edge method of enhancing performance in medial image analysis with the use of convolutional neural networks (CNN), which are well suited for analyzing images, and has led to a decrease in the classification error rate from about 25% in 2011 to 3.6% in 2015 (Russakovsky et al. 2015, Lakhani et al. 2017). With such success in identifying and classifying images using a deep learning algorithm, there has been interest in applying deep learning to medical image analysis in several fields, including the detection of skin cancer (Esteva et al. 2017), diabetic retinopathy (Gulshan et al. 2016), mammographic lesions (Kooi et al. 2017), and lung nodules (Hua
© 2018 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 10.1080/17453674.2018.1453714
12479 Kim D.indd 468
Acta Orthopaedica 2018; 89 (4): 468–473
Figure 1. Each shoulder anteroposterior radiograph was manually cropped into a square in which the humeral head and neck are centered such that they comprise approximately 50% of the square image as illustrated above. Images were then resized to 256 × 256 pixels. Examples of normal and each fracture type: (A) normal, (B) greater tuberosity fracture, (C) surgical neck fracture, (D) 3-part fracture, and (E) 4-part fracture.
et al. 2015). However, in the field of orthopedic surgery and traumatology, trials are very scarce despite its importance to public health. To our knowledge, only one study (Olczak et al. 2017) has applied deep learning to fracture orthopedics, and reported promising outcomes of deep learning in identifying fracture, laterality, type of view, and body part. Thus, we aimed to evaluate the diagnostic accuracy of the deep learning algorithm with deep CNN for detecting and classifying proximal humerus fractures using plain anteroposterior (AP) shoulder radiographs. We then compared the results with those of humans.
Patients and methods Dataset 1,891 plain shoulder AP radiographs (1,376 proximal humerus fracture cases and 515 normal shoulders) from 1,891 patients (591 men, 1,300 women; 1,083 from Konkuk University Medical Center, 209 from Kyungpook National University Hospital, 165 from Myungji Hospital, 203 from Kangwon National University Hospital, 41 from the National Police Hospital, 25 from Seoul Saint Mary’s Hospital, and 165 from Wonkwang University Sanbon Hospital) were used as the total dataset in this study. We used only 1 image per person to decrease the overperformance of deep learning by the inclusion of a very similar image of the same patient in each test and training set. The mean age of patients was 65 (24–90) years. Fracture classification To evaluate the performance of fracture classification, we classified the proximal humerus fractures into 4 types based on Neer’s classification, which is the most commonly used classification for the proximal humerus fracture: greater tuberosity, surgical neck, 3-part, and 4-part (Neer et al. 1970). A greater tuberosity fracture was defined as 1 displaced fragment of the greater tuberosity component, and A surgical neck fracture as 1 displaced fragment of the surgical neck component. A 3-part fracture was defined as 2 displaced fragments, while a 4-part fracture was defined as having 3 or more displaced
12479 Kim D.indd 469
fragments from the proximal humerus. In cases of proximal humerus fractures combined with shoulder dislocation (fracture dislocation type), we used the images after reduction and classified them. Each plain shoulder AP radiograph was manually cropped into a square in which the humeral head and neck were centered and constituted approximately 50% of the square image, resized to 256 × 256 pixels, and stored as a JPEG file (Figure 1). Fracture classification was performed by 2 shoulder orthopedic specialists with 14 and 17 years of experience (SWC and KSO) and 1 radiologist with expertise in musculoskeletal diseases and 15 years of experience (NRK). For cases in which the 3 specialists could not agree, the corresponding CT images were checked (CTs were available for all fractures that failed consensus) and then re-discussed. If consensus still could not be achieved even after the evaluation of the CT image(s), the images were excluded from the dataset (n = 21). 346 cases were ultimately classified as greater tuberosity fractures, 514 cases as surgical neck fractures, 269 cases as 3-part fractures, and 247 cases as 4-part fractures. In addition, 515 cases without proximal humerus fractures were classified as the normal group to evaluate the ability of the CNN to distinguish between normal and fractured shoulders (Figure 1). Training of the deep CNN and framework We used and trained the deep CNN using the training dataset and validated it using the test dataset. The dataset of the 1,891 images was divided into 10 partitions without overlapping images. Among the 10 partitions, 1 partition was used as a test dataset, while all other images were used as training datasets. Thus, for the 10 parts (1 partition = test dataset, the other 9 partitions and remnants = training dataset) 10 experiments were performed, after augmenting the training dataset. The entire training process was then repeated 3 times to adjust for possible deviations in the results. We ran Caffe 9 (http://caffe. berkeleyvision.org/) on Ubuntu 16.04 (https://www.ubuntu. com/download/desktop) with NVIDIA GTX 1070 (CUDA 8.0 and cuDNN 5.1) (https://developer.nvidia.com/cuda-zone and https://developer.nvidia.com/cudnn) and used the open source pre-trained Microsoft ResNet-152 (https://github.com/kaim-
ingHe/deep-residual-networks) as a deep CNN model, and further fine-tuned the pre-trained ResNet model to our proximal humerus fracture datasets (fine-tuning = training using our datasets). The detailed process of the deep CNN training process is shown in the Appendix (see Supplementary data). Evaluation of the deep CNN algorithm After training the deep CNN, we computed the top-1 accuracy. The deep CNN has to answer top-1 (the one with highest probability) to compute the top-1 accuracy, which is the conventional accuracy for the deep CNN answer (top-1) being exactly the expected answer, among 5 choices of normal, greater tuberosity fracture, surgical neck fracture, 3-part fracture, and 4-part fracture. The deep CNN had to find whatever differences it could to make up criteria and define the groups. Then, algorithm performance was measured using the area under the receiver operating curve (AUC) generated by plotting sensitivity versus 1-specificity, which reported the best sensitivity and specificity that maximizes the sum of sensitivity and specificity. In addition, the Youden index (sensitivity + specificity – 1) was calculated. The performance for discerning fractures from normal shoulders and for classifying fractures (the ability to define a certain fracture group (4 fracture group) after excluding normal shoulders from the test set) was evaluated using each value. For fracture type classification, performance was measured only in the fracture images after excluding the normal shoulder images to evaluate the actual performance of fracture classification, thus avoiding the possibility of overfitting of the deep CNN by the inclusion of normal cases that are relatively easy to discern. Evaluation of the diagnostic performance of human readers To compare the performance in diagnosing and classifying the proximal humerus fracture between the CNN and human readers, we provided each reader with the same information as the CNN. The readers consisted of 3 groups of general physicians (n = 28), general orthopedists (n = 11), and orthopedists specialized in shoulders (n = 19). The orthopedic surgeons mainly composed the human readers, as generally an orthopedic surgeon both classifies the fracture on radiographs and takes the decision to operate or not, and then performs surgeries. The 10 parts (181 images each) were converted into 10 image sheets (181 images each) containing the proximal humerus images without explanations (Figure 2, see Supplementary data). Each reader then received 3 image sheets that were randomly selected using a randomization program (http://www. randomizer.org) and were requested to provide the most probable diagnosis of each image (543 (3 × 181) images) in the form of 1 to 5 (1, normal; 2, greater tuberosity fracture; 3, surgical neck fracture; 4, 3-part fracture; 5, 4-part fracture). We calculated the top-1 accuracy, AUC, sensitivity/specificity, and Youden index for each group of human readers as with the CNN and then compared the values.
12479 Kim D.indd 470
Acta Orthopaedica 2018; 89 (4): 468–473
Statistics All statistical analyses were performed using SPSS 15.0 (SPSS, Inc., Chicago, IL, USA). The receiver operating characteristic curves were generated using a Python script, and each AUC was determined. Descriptive statistics were used to report each value of the top-1 accuracy, AUC, sensitivity/ specificity, and Youden index, which was described as a mean and a 95% confidence interval (CI). Comparisons between the CNN and each human group were performed using a one-way analysis of variance, followed by Bonferroni post hoc analysis for multiple comparison with the significance level set at p < 0.05. Ethics, funding, and potential conflicts of interest The study protocol was approved by the local ethics committee (IRB no. KUH1060143) with a waiver of informed consent. This work was supported by Konkuk University in 2017. All authors declare no conflict of interest.
Results Deep learning CNN performance The top-1 accuracy of the deep learning CNN model in distinguishing between normal and proximal humerus fractured shoulders exhibited more than 95% accuracy (96%, CI 94–97%). Among the proximal humerus fracture cases, the top-1 accuracy of the CNN model for distinguishing each fracture type from the other fracture types was 86% (CI 83–88%) for greater tuberosity fractures, 80% (CI 77–83%) for surgical neck fractures, 65% (CI 59–71%) for 3-part fracture, and 75% (CI 71–79%) for 4-part fractures. The distribution of mispredicted cases in the CNN model is described in Table 1. The deep learning CNN exhibited excellent diagnostic performance with an AUC of 0.996 (CI 0.995–0.998) for discerning normal cases from fracture cases. The CNN accurately classified proximal humerus fractures with an AUC of 0.98 (CI 0.98–0.99) for greater tuberosity fractures, 0.94 (CI 0.93–0.94) for surgical neck fractures, 0.90 (CI 0.89–0.92) for 3-part fractures, and 0.94 (CI 0.93–0.94) for 4-part fractures. At the optimal cutoff point, the mean sensitivity/specificity in the CNN model were 0.99/0.97, 0.97/0.94, 0.90/0.85, 0.88/0.83, and 0.93/0.85 for normal versus all, greater tuberosity, surgical neck, 3-part, and 4-part fractures, respectively. The mean Youden index of each group in the CNN model was as follows: normal, 0.97 (CI 0.96–0.97); greater tuberosity fracture, 0.90 (CI 0.88–0.92); surgical neck fracture, 0.75 (CI 0.73–0.77); 3-part fracture, 0.71 (CI 0.68–0.74); and 4-part fracture, 0.78 (CI 0.77–0.80). Comparison between CNN and human reader performance (Tables 2 and 3 and Figure 3) The CNN showed superior results in diagnosing proximal humerus fractures compared with every human group,
Acta Orthopaedica 2018; 89 (4): 468–473
Table 1. Mispredicted cases in the convolutional neural network model. Values are n (%)
Normal (n = 1,500) a Greater tuberosity fracture (n = 990) Surgical neck fracture (n = 1,500) Three-part fracture (n = 750) Four-part fracture (n = 690) a 50
Types mispredicted as Greater Surgical Threetuberosity neck part fracture fracture fracture 47 (3)
37 (4) 16 (1) 0 (0) 2 (0)
19 (1) 39 (5) 1 (0)
19 (1) 30 (3) 135 (18) 98 (14)
1 (0) 68 (7) 115 (8)
Fourpart fracture 0 (0) 5 (1) 148 (10) 88 (12)
in each partition x three repetitions x 10 partitions
Table 2. Diagnostic accuracy for differentiating proximal humerus fractures from normal shoulders among the CNN and human groups. Values are mean (CI)
CNN Top-1 accuracy (%) Sensitivity Specificity Youden index
96 (94–97) 0.99 (0.99–1.00) 0.97 (0.97–0.98) 0.97 (0.96–0.97)
General physician (80–90) a
85 0.82 (0.78–0.87) a 0.94 (0.93–0.96) a 0.77 (0.72–0.82) a
93 (90–96) 0.93 (0.89–0.97) 0.97 (0.96–0.98) 0.90 (0.87–0.94)
Orthopedists specialized in shoulder
93 (87–99) 0.96 (0.95–0.98) 0.98 (0.96–1.00) 0.94 (0.92–0.96)
< 0.001 < 0.001 0.002 < 0.001
CNN, convolutional neural network Youden index was calculated as [sensitivity + specificity – 1]. a Statistically significant in a comparison of CNN and each human group (results from a Bonferroni post hoc analysis)
although the comparison with the general orthopedist and shoulder orthopedist groups did not reach statistical significance (Table 2). In addition, the CNN showed the highest performance for classifying proximal humerus fracture types among all fracture types except for greater tuberosity fractures, despite several comparisons with the shoulder orthopedist group not showing statistical significance (Table 3, see Supplementary data). The diagnostic superiority of the CNN compared with the human groups was more marked in 3- and 4-part fractures (Table 3). The CNN was superior to a general physician or general orthopedist on comparing the diagnostic performance of CNN and each human group by overall distribution of the sensitivity/specificity point per person on a receiver operating characteristic curve of the CNN (Figure 3).
Discussion In this study, we demonstrate the very high performance of deep learning CNN in distinguishing normal shoulders from proximal humerus fractures. We additionally show promising results for classifying fracture type based on plain shoulder AP radiographs, with the deep learning CNN exhibiting superior performance to that of general physicians and general
12479 Kim D.indd 471
orthopedists and similar performance to that of the shoulder orthopedists. This indicates the possibility of automated diagnosis and classification of proximal humerus fractures and other fractures or orthopedic diseases diagnosed accurately using plain radiographs. As additional proximal humerus fractures would further enhance the diagnostic performance of the CNN, we think that the deep learning CNN may outperform even the shoulder orthopedists as data accumulate. Moreover, we found higher performance of CNN, especially in more complex type fractures such as 3- or 4-part fractures, compared with humans, which suggests the superiority of CNN for classifying fractures with various fracture shapes based on plain radiographs because humans have greater difficulty, especially classifying complex fractures, but CNN performs relatively well. Since the number of images for the CNN training was smaller for 3- and 4-part fractures, the results seem more promising. With more training cases of 3and 4-part fractures, the diagnostic performance of CNN for detecting and classifying complex fractures would improve. The higher performance of CNN for detecting and classifying proximal humerus fractures, especially complex fractures, may in part come from the fact that machine does not suffer from decreases in concentration and is consistent when presented with the same input data (i.e., the CNN will make the same prediction on a specific image every time) unlike
Acta Orthopaedica 2018; 89 (4): 468â&#x20AC;&#x201C;473
Figure 3. The diagnostic performance between the CNN and each human group was compared using the receiver operating characteristics curves of the CNN and the sensitivityâ&#x20AC;&#x201C;specificity distribution of each human group to differentiate normal shoulders from proximal humerus fractures (A) and to classify each fracture type: (B) greater tuberosity fracture, (C) surgical neck fracture, (D) 3-part fracture, and (E) 4-part fracture. CNN = convolutional neural network; AUC = area under curve of the receiver operating characteristics curve. The representative receiver operating characteristics curve of the CNN was selected as the curve with the closest AUC value to the average AUC. The dots on the plots represent the sensitivity and specificity of each group (yellow, general physicians; green, general orthopedists; red, orthopedists specialized in the shoulder). All AUCs for the normal shoulder and each fracture type were over 90%. The CNN achieved superior performance at least to a general physician (yellow dot) or to a general orthopedist (green dot), most of whose sensitivity/specificity point lay below the receiver operating characteristic curve of the CNN.
humans, who are likely to make an error after a distorted previous experience in fracture classification (humans seem to have a tendency to guess right more often in a typical case but have difficulty when the fracture configuration is a less familiar shape) and through limited concentration. The machine can potentially be trained with an incredible amount of training samples, vastly more than any orthopedist will experience in his/her lifetime, which results in an incomparable possibility of deep learning CNN. In addition, the diagnostic accuracy of CNN for classifying greater tuberosity fractures was the highest and that of 3-part fracture was the lowest. Greater tuberosity fractures exhibited a distinctive fracture line in the anatomical site of the greater tuberosity with a low variance in the fracture shape among greater tuberosity fractures, whereas all other fracture types in this study have fracture lines in the surgical neck site. We think this anatomical characteristic of the greater tuberosity fracture makes the detection of this fracture type easier with a low error rate. Conversely, the 3-part fracture has a shape between that of a surgical neck fracture and a 4-part fracture. Thus, the
12479 Kim D.indd 472
CNN seems to confuse more severe 3-part fracture cases with more displacement and angulation with 4-part fractures, while less severe 3-part fracture cases with less displacement and angulation, especially in the greater tuberosity fragments, are confused with surgical neck fractures. This automated system for detecting and classifying proximal humerus fractures has potential benefits, such as increased accuracy, consistent interpretation, efficiency, near-instantaneous reporting of results, reproducibility, and decreased barriers to access. Since a deep CNN algorithm can have multiple operating points, its sensitivity and specificity can be tuned to match the requirements of specific clinical settings, such as high sensitivity for a screening setting if necessary. With additional data, deep learning will facilitate diagnosis. Furthermore, we believe that the clinical application of deep learning for detection and classification can be expanded to other orthopedic diseases that use radiographs for diagnosis. Our study has several limitations. First, even though the Neer classification is the most commonly used tool for proximal humerus fracture classification, it has only fair to moderate
Acta Orthopaedica 2018; 89 (4): 468–473
reliability, and there is no gold standard for proximal humerus fracture classification. Development of a more reliable classification system for proximal humerus fracture could enhance the reliability in classification of the deep learning algorithm. However, the promising result of this study in detecting and classifying proximal humerus fracture by using a deep learning algorithm does not mean that it can be used immediately in clinical practice. This study was not to guide treatment. This study only has the significance that we showed the possibility of the future use of this deep learning algorithm even in the field of orthopedic surgery or traumatology. CNNs that consistently classify fractures could be a giant leap forward. Second, we evaluated the diagnostic performance of CNN based on a cropped single shoulder AP radiograph to keep this project simple, which may not actually reflect a clinically relevant scenario because a fracture evaluation would involve at least 2 radiographs under review. However, the evaluations based on various shoulder radiographs or CT images may enhance the diagnostic performance of CNN as well. Finally, the images were down-sampled to 256 × 256 pixels before they were fed into the network because of the sheer number of parameters inherent to the networks. The diagnostic accuracy may be improved using higher-resolution images. More development on the memory of graphics processing units would allow larger matrix sizes without increasing the training time. In addition, the lossy JPEG compression may influence the image quality. It may be better to use non-lossy compression such as PNG or TIFF. In conclusion, the use of artificial intelligence can accurately detect and classify proximal humerus fractures on plain shoulder AP radiographs. Further studies are necessary to determine the feasibility of applying artificial intelligence in the clinic and whether its use could improve care and outcomes compared with current orthopedic assessments. Supplementary data Figure 2, Table 3 and the Appendix are available as supplementary data in the online version of this article, http://dx.doi. org/ 10.1080/17453674.2018.1453714
Acta thanks Max Gordon and other anonymous reviewers for help with peer review of this study.
Bengio Y, Courville A, Vincent P. Representation learning: a review and new perspectives. IEEE Trans Pattern Anal Mach Intell 2013; 35(8): 1798-828. Esteva A, Kuprel B, Novoa R A, Ko J, Swetter S M, Blau H M, Thrun S. Dermatologist-level classification of skin cancer with deep neural networks. Nature 2017; 542(7639): 115-18. Foroohar A, Tosti R, Richmond J M, Gaughan J P, Ilyas A M. Classification and treatment of proximal humerus fractures: inter-observer reliability and agreement across imaging modalities and experience. J Orthop Surg Res 2011; 6: 38. Gulshan V, Peng L, Coram M, Stumpe M C, Wu D, Narayanaswamy A, Venugopalan S, Widner K, Madams T, Cuadros J, Kim R, Raman R, Nelson P C, Mega J L, Webster D R. Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 2016; 316(22): 2402-10. Hua K L, Hsu C H, Hidayati S C, Cheng W H, Chen Y J. Computer-aided classification of lung nodules on computed tomography images via deep learning technique. Onco Targets Ther 2015; 8: 2015-22. Kooi T, Litjens G, van Ginneken B, Gubern-Merida A, Sanchez C I, Mann R, den Heeten A, Karssemeijer N. Large scale deep learning for computer aided detection of mammographic lesions. Med Image Anal 2017; 35: 30312. Lakhani P, Sundaram B. Deep learning at chest radiography: automated classification of pulmonary tuberculosis by using convolutional neural networks. Radiology 2017; 284(2): 574-82. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature 2015; 521(7553): 43644. Mora Guix J M, Pedros J S, Serrano A C. Updated classification system for proximal humeral fractures. Clin Med Res 2009; 7(1-2): 32-44. Neer C S, 2nd. Displaced proximal humeral fractures, I: Classification and evaluation. J Bone Joint Surg Am 1970; 52(6): 1077-89. Olczak J, Fahlberg N, Maki A, Razavian A S, Jilert A, Stark A, Sköldenberg O, Gordon M. Artificial intelligence for analyzing orthopedic trauma radiographs. Acta Orthop 2017; 88(6): 581-6. Russakovsky O, Deng J, Su H, Krause J, Satheesh S, Ma M, Huang Z, Karpathy A, Khosla A, Bernstein M, Berg A C, Fei-Fei L. ImageNet large scale visual recognition challenge. Int J Comput Vis 2015; 115(3): 211-52. Shin H C, Orton M R, Collins D J, Doran S J, Leach M O. Stacked autoencoders for unsupervised feature learning and multiple organ detection in a pilot study using 4D patient data. IEEE Trans Pattern Anal Mach Intell 2013; 35(8): 1930-43. Wang S, Summers R M. Machine learning and radiology. Med Image Anal 2012; 16(5): 933-51.
Supervision: SWC, YK. Conception and design: SWC, SSH, JWL, K-SO, NRK, YK. Acquisition of data: SWC, JWL, K-SO, JPY, JYK, SHM, JK, H-JL, Y-MN. Analysis and interpretation of data: SWC, SSH, NRK, JPY, JYK, SHM, JK, H-JL, Y-MN, YK.
12479 Kim D.indd 473
Acta Orthopaedica 2018; 89 (4): 474
Per Aspenberg, 1949–2018
Professor Per Aspenberg has sadly passed away. He was for many years one of the most influential personalities in the Swedish orthopedic community and combined a brilliant intellect with a constant curiosity and innovative ability. He began his career as an orthopedic surgeon in Västerås but was from early on interested in research, which attracted him to the academic environment at Lund University, where he soon became a well-respected scientist with a primary focus on bone healing and metabolism. After having distinguished himself as a successful researcher with an established international reputation he was in 2001 appointed Professor of Orthopedics at Linköping University, his home town. Here he continued his work in the field of bone and tendon healing by taking charge of the local research laboratory, which he revitalized. He was the principal mentor of almost 30 PhD projects, several of which have had a significant impact on orthopedic science. Aspenberg’s research spanned wide fields, among them bone metabolism, epidemiology, and fracture treatment. He invented the “Bone chamber,” successfully used for numerous bone metabolism experiments. He could show in animal and clinical studies that PTH advances fracture healing. He was early in identifying the substantially increased risk of atypical fractures associated with bisphosphonate use but could also show that bisphosphonates could be used to decrease the risk of loosening of joint prostheses. One of his hobbyhorses was to question the benefit of some fracture surgery. He applied an American questionnaire used to identify unsuitable pilots to show that the propensity to do (unnecessary) surgery increased with what was identified by the questionnaire as macho features (not good in pilots and perhaps not in surgeons…). His never-ending enthusiasm for science and ability to inspire his colleagues and coworkers was always sincerely appreciated and he had the aptitude of conveying a positive
atmosphere to whatever gathering in the department. Among other achievements, his CV comprises over 300 publications in medical journals (1 in this issue of Acta, pp 457–461) and although the majority of his substantial scientific production was based on laboratory work, he took part in many clinical studies and was constantly involved and knowledgeable in everyday clinical practice. Being open-minded and interested in all fields of orthopedics he was always the perfect discussion partner. Aspenberg had been co-editor of Acta Orthopaedica since 2005 and helped many authors to further improve already good manuscripts. He was just as good at explaining to authors why their manuscripts should not be published, but without offending them. He enjoyed music and art and was a talented musician as well as a keen craftsman in wood. The annual meeting of the Swedish Orthopaedic Society will truly miss his ingenious wooden sculptures that were always displayed at the session of “Art in Orthopaedics.” He was a sailor, and often alone in the boat (see https://youtu.be/voAqdCrSnu8). Even in the last days of his life he stayed engaged in research and only a few weeks before his demise he published an article questioning the role of the placebo effect, typical of his commitment to challenge established concepts that he felt had been misinterpreted or misused. The orthopedic community has lost one of its most colorful, influential and important individuals and he is sincerely missed. Lars Adolfsson and all Editors of Acta Orthopaedica
© 2018 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 10.1080/17453674.2018.1491747