Rheumatology 2019;58:1016–1024 RHEUMATOLOGY doi:10.1093/rheumatology/key409 Advance Access publication 3 January 2019 Original article The effect of rheumatoid arthritis on patient-reported outcomes following knee and hip replacement: evidence from routinely collected data Edward Burn 1, Christopher J. Edwards2, David W. Murray1, Alan Silman1, Cyrus Cooper1,3, Nigel K. Arden1,3, Rafael Pinedo-Villanueva1,3 and Daniel Prieto-Alhambra1,4 Abstract Objectives. To compare outcomes of total knee replacement (TKR) and total hip replacement (THR) for individuals with RA and OA. Methods. We performed a cohort study using routinely collected data. Oxford Knee Score, Oxford Hip Score, and EuroQol 5-dimension 3-level (EQ-5D-3L) questionnaires were collected before and 6 months after surgery. Multivariable regressions were used to estimate the association between diagnosis and post-operative scores after controlling for pre- operative scores and patient characteristics. Results. Study cohorts included 2070 OA and 142 RA patients for TKR and 2030 OA and 98 RA patients for THR. Following TKR, the median Oxford Knee Score was 37 [interquartile range (IQR) 29–43] for OA and 36 (27–42) for RA while the median EQ-5D-3L was 0.76 (0.69–1.00) and 0.69 (0.52–0.85), respectively. After THR, the Oxford Hip Score was 42 (IQR 36–46) for OA and 39 (30–44) for RA while the EQ-5D-3L was 0.85 (0.69–1.00) and 0.69 (0.52–1.00), respectively. The estimated effect of RA, relative to OA, on post-operative scores was 0.05 (95% CI 1.57, 1.48) for the Oxford Knee Score, 0.09 ( 0.13, 0.06) for the EQ-5D-3L following TKR, 1.35 ( 2.93, 0.22) for the Oxford Hip Score, and 0.08 ( 0.12, 0.03) for the EQ-5D-3L following THR. Conclusion. TKR and THR led to substantial improvements in joint-specific scores and overall quality of life. While diagnosis had no clinically meaningful effect on joint-specific outcomes, improvements in general quality of life were somewhat less for those with RA, which is likely due to the systemic and multijoint nature of rheumatoid disease. CLINICAL SCIENCE Key words: rheumatoid arthritis, osteoarthritis, outcome measures, quality of life, surgery Rheumatology key messages . Individuals with osteoarthritis and rheumatoid arthritis achieve substantial health gains from knee and hip replacement. . Individuals with osteoarthritis and rheumatoid arthritis get similar improvements in joint-specific outcomes. . Gains in overall quality of life are somewhat less for those with rheumatoid arthritis. Introduction clinical syndrome of failure of the joint [1], and joint re- The most common indication for total knee replacement placement is considered when non-surgical alternatives (TKR) and total hip replacement (THR) is OA. OA is the have failed. In such cases, TKR and THR typically relieve 1 pain and improve function, leading to substantial gains in Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford, 2NIHR an individual’s quality of life [2, 3]. Consequently, the pro- Wellcome Trust Clinical Research Facility, University Hospital cedures are considered cost effective, with their costs Southampton, Southampton, 3MRC Lifecourse Epidemiology Unit, Southampton University, Southampton, UK and 4GREMPAL Research justified by the expected health gains for the individuals Group, Idiap Jordi Gol and CIBERFes, Universitat Autonoma de receiving them [2, 3]. Barcelona and Instituto de Salud Carlos III, Barcelona, Spain TKR and THR are also performed on individuals with inflammatory rheumatic diseases. RA is the most Submitted 30 August 2018; accepted 1 November 2018 common of these. RA can destroy cartilage and erode Correspondence to: Rafael Pinedo-Villanueva, Botnar Research bone and the involvement of many joints is common [4]. Centre, NDORMS, University of Oxford, Oxford OX3 7HE, UK. E-mail: rafael.pinedo@ndorms.ox.ac.uk As with OA, pain, loss of function and damaged structure ! The Author(s) 2019. Published by Oxford University Press on behalf of the British Society for Rheumatology. This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited. For commercial re-use, please contact journals.permissions@oup.com RA and patient-reported outcomes are all indications for joint replacement for those with RA primary care records. Where an individual had both a [4]. However, while pain remains the foremost indication, diagnosis of both RA and OA, RA was taken as the function and structure play an increased role. Guidelines index diagnosis. Individuals could be in both the knee- promote referral for early surgical opinion where persist- related and hip-related cohorts. ent pain, worsening joint function, progressive deformity HES APC records were then linked to these cohorts. or persistent localized synovitis have not responded opti- TKR or THR following the index diagnosis was identified mally to non-surgical treatment [5]. using the HES APC on the basis of procedure codes. Whether individuals with RA achieve equivalent gains Individuals who received bilateral surgery, where both from undergoing TKR and THR as those with OA is not the left and right knee joints were replaced simultaneously yet known. Given multiple joint involvement, systemic or staged, were excluded from the analysis. If an individ- comorbidity and polypharmacy [4], it is plausible that out- ual received a second TKR or THR following diagnosis, comes for individuals with RA may be worse than for only the first was included in the analysis. those with OA. Moreover, as it is recommended that hip The HES PROMs were then linked to the identified TKR replacement should generally precede knee replacement and THR procedures in the HES APC. Individuals were for those with RA [6], the relationship between RA and included if they had complete scores recorded for OKS/ outcomes could differ depending on the procedure. OHS and EQ-5D-3L before and after surgery. We as- The success of TKR and THR have traditionally been sessed the potential impact of this decision by comparing measured by implant survival. There is, however, an the characteristics of individuals missing and not missing increased awareness that patient-reported outcome PROM scores. measures (PROMs) provide a more appropriate indicator of effectiveness and are more closely tied to the reasons Measurement of variables for performing surgery [7]. The Oxford Knee Score (OKS) The main exposure of interest was patient diagnosis of RA and Oxford Hip Score (OHS) are joint-specific PROMs for or OA and these diagnoses were identified using CPRD knee and hip replacement, respectively, with questions on records. The date of a procedure recorded in the HES both pain and function [8, 9]. Meanwhile, the EuroQol 5- APC was used to identify the year of surgery and, given dimension 3-level (EQ-5D-3L) questionnaire measures an individual’s year of birth recorded in CPRD, age at sur- overall health-related quality of life [10]. gery was inferred. Gender was obtained from CPRD re- In this study we compared the effectiveness of TKR and cords. Diagnosis codes in the HES APC were used to THR for individuals with RA and OA in terms of both joint- identify conditions included in the Royal College of specific measures and overall health-related quality of life. Surgeons Charlson score [14]. This score provides a sum- mary measure of comorbidities. RA was omitted from cal- culations of the Charlson score, as it was considered Methods separately in the analysis. Indices of Multiple Deprivation Setting, data sources and study participants (IMD) quintiles were based on postcodes recorded in the CPRD. IMD quintiles provide a measure of socio-eco- Primary care, inpatient hospital and PROMs data from the nomic status, with 5 implying the highest level of depriv- National Health Service (NHS) in England informed this ation. BMI and smoking status were taken from the CPRD. study. The Clinical Practice Research Datalink (CPRD) The values of BMI and smoking status closest to surgery provided primary care records and demographic details. were identified and included if they were recorded in the Around 7% of the UK population is included in the CPRD, year leading up to surgery. with those included broadly representative of the general population [11]. The Hospital Episode Statistics Admitted PROMs Patient Care (HES APC) provided inpatient hospital data. The OKS and OHS are joint-specific questionnaires whose The HES APC contains data on all NHS-funded hospital scores range from 0 (worst) to 48 (best) [15]. The EQ-5D- admissions in England and includes both clinical and 3L questionnaire covers general health-related quality of demographic information [12]. The NHS provided routinely life, with questions covering five dimensions with three collected PROMs with OKS and EQ-5D-3L data from pa- possible response levels. The dimensions include mobil- tients receiving TKR and OHS and EQ-5D-3L data for ity, self-care, usual activities, pain/discomfort and anxiety/ those undergoing THR. These PROMs are completed depression. Responses are classified as either no prob- before and 6 months after surgery, with all patients lems, some problems or extreme problems [10]. The com- undergoing TKR and THR funded by the English NHS bination of these dimensions and responses can be invited to participate [13]. Ethical approval for the study summarized as an index value, which ranges from 0.59 was granted by the CPRD Independent Scientific Advisory (worst) to 1 (best) after applying preference-based utility Committee. weights obtained from the general UK population [16]. We established knee-related and hip-related diagnosis cohorts based on clinical codes within the CPRD records. Statistical methods Individuals entered the knee-related cohort if they had an incident (newly recorded) diagnosis of RA or knee OA. Comparing patient characteristics Similarly, individuals entered the hip-related cohort if The characteristics of those with RA and OA and those they had an incident diagnosis of RA or hip OA in their with and without missing PROMs were compared. https://academic.oup.com/rheumatology 1017 Edward Burn et al. Standardized mean difference was used to assess reflect the expected change in the outcome variable whether there was balance in a characteristic, with a based on a given change in the explanatory variable, mean difference <0.1 indicating a negligible difference in while other explanatory variables are fixed. The coeffi- the mean of a covariate between groups [17]. cients for diagnosis, for example, reflect the estimated effect of RA relative to OA on post-operative score, con- Comparing observed PROMs trolling for differences in pre-operative scores and patient The OKS/OHS and EQ-5D-3L index reported by individ- characteristics. uals with OA and RA before and after TKR and THR were compared, with both the absolute change in scores and Results the percentage of potential change (PoPC) achieved calculated. The PoPC accounts for differences in baseline Study participants scores by summarizing how much an individual has A total of 2212 (2070 with OA and 142 with RA) and 2128 improved (or worsened) relative to how much they poten- (2030 with OA and 98 with RA) individuals were included in tially could have improved (or worsened) [18]. The PoPC the study cohorts for TKR and THR, respectively. The ranges between 100% (i.e. an individual has worsened study inclusion flow chart is shown in Fig. 1 and the base- to the lowest possible score) and 100% (i.e. an individual line (pre-operative) characteristics of study participants has improved to the best possible score), with a value of are summarized in Table 1. Further details on those with 0% indicating no change. missing data are provided in Supplementary Figures S2 As well as considering the EQ-5D-3L index, changes and S3 and Supplementary Table S1, available at between pre- and post-operative time points in individual Rheumatology online. dimensions and the overall profile were summarized by diagnosis. Each dimension was either better, had no Association between RA and post-operative PROMs change or was worse following surgery. Meanwhile, the overall profile was either better (i.e. at least one dimension OKS following TKR had improved, with none worse), had a mixed change (i.e. Individuals with OA improved in the OKS after TKR from a some dimensions had improved but others had wor- median of 19 [interquartile range (IQR) 14–24] to 37 sened), had no change or was worse (i.e. at least one (29–43), a change of 16 (9–23) and a PoPC realized of dimension had worsened, with none improving). 62% (35% to 81%). Meanwhile, individuals with RA im- proved in the OKS from 15 (IQR 10–22) to 36 (27–42), a Estimating the association between RA, relative to OA change of 18 (11–24) and a PoPC realized of 59% (33% to and post-operative PROMs 79%). See Table 2 for a summary of the OKS by diagnosis Ordinary least squares regressions were used to assess and Fig. 2 for the distribution of the PoPC by diagnosis. the impact of diagnosis (RA or OA) on post-operative After controlling for differences in pre-operative score and scores, controlling for both pre-operative scores and patient characteristics, the estimated effect of RA relative patient characteristics. The outcome variable was post- to OA on post-operative OKS was 0.05 ( 1.57–1.48), as operative score. Univariable models with only diagnosis shown in Table 4. and multivariable models with age, gender, Charlson score, IMD quintile, BMI, smoking status and pre-opera- EQ-5D-3L following TKR tive score also included as explanatory variables were Regarding overall health-related quality of life, 75% of estimated. Restricted cubic splines were used to incorp- those with OA had a better EQ-5D-3L profile following orate non-linearity in continuous variables where the TKR, 8% had a mixed change, 10% had no change and Bayesian information criterion (BIC) was lower than that 6% were worse than before surgery. Their EQ-5D-3L of a model with a linear relationship fitted, with BIC also index improved from 0.59 (IQR 0.12–0.69) to 0.76 used to decide on how many knots to include if a non- (0.69–1.00), a mean change of 0.31 (0.07–0.59) and a linear relationship was incorporated [19]. As there were PoPC of 61% (21% to 100%). In total, 67% of those few study participants with multiple comorbidities, the with RA had a better EQ-5D-3L profile following TKR, Charlson score was dichotomized to 0 and 1+. 15% had a mixed change, 8% had no change, and 11% There were missing data for three of the explanatory were worse. Their EQ-5D-3L index improved from 0.16 variables included in the analysis: IMD, BMI and smoking (IQR 0.00–0.65) to 0.69 (0.52–0.85) for a change of status. While IMD was missing for only two participants, 0.31 (0.07–0.60) and a PoPC of 55% (11% to 71%). BMI and smoking status were missing in higher propor- Table 2 reports a summary of EQ-5D-3L indices and tions ranging from 39% to 58% among the various co- Table 3 reports the change in dimension scores by diag- horts. We used multiple imputation to account for these nosis, with further details on the dimension scores given in missing values. This was based on the assumption that Supplementary Table S2, available at Rheumatology data were missing at random, where the probability of online. While both individuals with OA and those with RA data being missing does not depend on the unobserved achieved substantial improvements in their usual activ- data, conditional on the observed data [20]. Explanatory ities, pain/discomfort and mobility dimensions, those variables and outcomes were used to impute 50 datasets. with RA were less likely to report no problems on these Pooled model coefficients and corresponding 95% CIs dimensions following surgery compared with those with were calculated using Rubin’s rules. The coefficients OA. After controlling for differences in pre-operative 1018 https://academic.oup.com/rheumatology RA and patient-reported outcomes FIG. 1 Study inclusion flow chart TABLE 1 Patient characteristics TKR THR Characteristics OA RA SMD OA RA SMD N 2070 142 2030 98 Age, years, median (IQR) 70 (64–76) 69 (62–75) 0.16 69 (63–76) 70 (61–76) 0.01 Gender, female, n (%) 1113 (54) 87 (61) 0.15 1145 (56) 75 (77) 0.44 Charlson score, n (%) 0.13 0.03 0 1466 (71) 92 (65) 1507 (74) 74 (76) 1+ 604 (29) 50 (35) 523 (26) 24 (25) BMI, median (IQR) 30 (27–35) 30 (27–35) 0.03 29 (25–32) 28 (25–31) 0.02 BMI missing, n (%) 987 (48) 73 (51) 0.07 1048 (52) 57 (58) 0.13 Smoking status, n (%) 0.28 0.21 Non-smoker 606 (29) 30 (21) 559 (28) 21 (21) Ex-smoker 576 (28) 39 (28) 496 (24) 24 (25) Current smoker 82 (4) 14 (10) 126 (6) 11 (11) Missing 806 (39) 59 (42) 849 (42) 42 (43) IMD quintile, n (%) 0.09 0.29 1 471 (23) 33 (23.2) 551 (27) 25 (26) 2 536 (26) 36 (25.4) 509 (25) 23 (24) 3 497 (24) 32 (22.5) 488 (24) 18 (18) 4 380 (18) 25 (17.6) 339 (17) 17 (17) 5 185 (9) 16 (11.3) 142 (7) 15 (15) Missing 1 (0) 0 (0) 1 (0) 0 (0) Comparison of the characteristics of individuals, split by diagnosis of OA and RA recorded prior to and following TKR and THR. SMD: standardized mean difference. https://academic.oup.com/rheumatology 1019 Edward Burn et al. TABLE 2 Joint-specific (OKS/OHS) and generic (EQ-5D-3L utility index) PROMs according to diagnosis Pre-operative, Post-operative, Change, median PoPC, median Diagnosis median (IQR) median (IQR) (IQR) (IQR) TKR OKS OA 19 (14–24) 37 (29–43) 16 (9– 23) 62 (35–81) RA 15 (10–22) 36 (27–42) 18 (11–24) 59 (33–79) EQ-5D-3L index OA 0.59 (0.12–0.69) 0.76 (0.69–1.00) 0.31 (0.07–0.59) 61 (21–100) RA 0.16 (0.00–0.65) 0.69 (0.52–0.85) 0.31 (0.07–0.60) 55 (11–71) THR OHS OA 18 (12–24) 42 (36–46) 23 (16–29) 81 (61–94) RA 12 (7–20) 39 (30–44) 24 (15–31) 73 (52–89) EQ-5D-3L index OA 0.52 (0.06–0.66) 0.85 (0.69–1.00) 0.41 (0.20–0.74) 79 (50–100) RA 0.06 ( 0.07–0.59) 0.69 (0.52–1.00) 0.47 (0.13–0.76) 70 (25–100) The OKS and OHS range from 0 to 48 while the EQ-5D-3L index ranges from 0.59 to 1, with higher scores being better for both. FIG. 2 Distribution of percentage of potential change Kernel density distributions with points showing observed data and solid vertical lines denoting the 25th percentile, 50th percentile (median) and 75th percentile from the observed data. score and patient characteristics, the estimated effect of (52% to 89%). See Table 2 for a summary of the OHS by RA relative to OA on the post-operative EQ-5D-3L index diagnosis and Fig. 1 for the distribution of the PoPC by was 0.09 (95% CI 0.13, 0.06), as detailed in Table 4. diagnosis. After controlling for differences in pre-operative score and patient characteristics, the estimated effect of OHS following THR RA relative to OA on post-operative OHS was 1.35 (95% Following THR, individuals with OA had an improved OHS CI 2.93, 0.22) (Table 4). from 18 (IQR 12–24) to 42 (36–46), a change of 23 (16–29) and a PoPC of 81% (61% to 94%). Meanwhile, individuals EQ-5D-3L following THR with RA had an improved OHS from 12 (IQR 7–20) to 39 Of the study participants with OA, 86% had a better EQ- (30–44) for a change of 24 (15–31) and a PoPC of 73% 5D-3L profile following THR, 6% had a mixed change, 5% 1020 https://academic.oup.com/rheumatology RA and patient-reported outcomes TABLE 3 Change in EQ-5D-3L dimension scores by diagnosis TKR THR Characteristics OA, n (%) RA, n (%) OA, n (%) RA, n (%) Dimensions Usual activities Better 977 (47) 56 (39) 1282 (63) 55 (56) No change 1019 (49) 80 (56) 712 (35) 39 (40) Worse 74 (4) 6 (4) 36 (2) 4 (4) Anxiety/ depression Better 480 (23) 38 (27) 638 (31) 43 (44) No change 1460 (71) 86 (61) 1318 (65) 49 (50) Worse 130 (6) 18 (13) 74 (4) 6 (6) Pain/ discomfort Better 1242 (60) 88 (62) 1540 (76) 72 (73) No change 789 (38) 49 (35) 471 (23) 23 (23) Worse 39 (2) 5 (4) 19 (1) 3 (3) Mobility Better 993 (48) 44 (31) 1190 (59) 35 (36) No change 1046 (51) 96 (68) 824 (41) 61 (62) Worse 31 (1) 2 (1) 16 (1) 2 (2) Self-care Better 385 (19) 29 (20) 821 (40) 40 (41) No change 1567 (76) 100 (70) 1146 (56) 55 (56) Worse 118 (6) 13 (9) 63 (3) 3 (3) Overall profile Better 1561 (75) 95 (67) 1738 (86) 78 (80) Mixed change 161 (8) 21 (15) 115 (6) 10 (10) No change 215 (10) 11 (8) 110 (5) 4 (4) Worse 133 (6) 15 (11) 67 (3) 6 (6) Study participants’ pre- and post-operative EQ-5D-3L profiles were compared. For each dimension, an individual either had a better score post-operatively, no change or a worse score than what they reported pre-operatively. An individual’s overall profile was either better (i.e. at least one dimension had improved, with none worse), had a mixed change (i.e. some dimen- sions had improved but others had worsened), had no change or was worse (i.e. at least one dimension had worsened, with none improving). TABLE 4 Estimated effects of RA, relative to OA, on post-operative PROMs TKR THR Measure Univariable model Multivariable model Univariable model Multivariable model OKS/ OHS 1.65 ( 3.27, 0.03) 0.05 ( 1.57, 1.48) 3.48 ( 5.16, 1.81) 1.35 ( 2.93, 0.22) EQ-5D-3L index 0.13 ( 0.17, 0.09) 0.09 ( 0.13, 0.06) 0.14 ( 0.19, 0.09) 0.08 ( 0.12, 0.03) Data presented as coefficients with 95% CIs. Dependent variables were post-operative scores. In addition to diagnosis (RA or OA), multivariable models also included age at surgery, gender (male or female), Charlson score, IMD quintile, pre-operative score for the corresponding dependent variable, BMI and smoking status as explanatory variables. had no change and 3% were worse after surgery. Their 0.47 (0.13–0.76) and a PoPC of 70% (25% to 100%). EQ-5D-3L index improved from 0.52 (IQR 0.06–0.66) to Table 2 shows a summary of EQ-5D-3L indices and 0.85 (0.69–1.00) for a change of 0.41 (0.20–0.74) and a Table 3 the change in dimension scores by diagnosis, PoPC of 79% (50% to 100%). A total of 80% of those with further details on the dimension scores given in with RA had a better EQ-5D-3L profile following TKR, Supplementary Table S2, available at Rheumatology 10% had a mixed change, 4% had no change and 6% online. As with TKR, although both those with OA and were worse. Their EQ-5D-3L index improved from 0.06 RA achieved substantial improvements in the usual activ- (IQR 0.07–0.59) to 0.69 (0.52–1.00) for a change of ities, pain/discomfort and mobility dimensions, those with https://academic.oup.com/rheumatology 1021 Edward Burn et al. RA were less likely to report no problems on these 315 with OA and found RA to be associated with reduced dimensions following surgery compared with those with improvements in knee pain following TKR relative to OA OA. After controlling for differences in pre-operative [23]. However, another study, based on 1991 individuals score and patient characteristics, the estimated effect of of whom 45 had RA at a large orthopaedic centre in RA relative to OA on the post-operative EQ-5D-3L index England, found RA to be associated with significantly was 0.08 (95% CI 0.12, 0.03), as summarized in greater improvements in knee pain compared with OA Table 4. [24]. Meanwhile, in line with our findings, another study of 4456 participants (136 with RA) undergoing TKR at Association between other explanatory factors and one hospital found those with RA had similar pain and post-operative PROMs function scores following surgery despite having worse The estimated effects of other explanatory variables from pre-operative scores [25]. Another study compared out- the ordinary least squares regressions are provided in comes following THR for 62 individuals with OA and 35 Supplementary Table A3, available at Rheumatology with RA [26] and found that the gains in joint-specific online, and the partial effect at the mean of diagnosis and scores appeared to be slightly less for those with RA com- pre-operative score, age and BMI are plotted in pared with those with OA, although, given the small Supplementary Figures S7–S12, available at Rheumatology sample size, potential confounders could not be online. Pre-operative scores had the largest impact on post- considered. operative scores, with these expected to increase as pre- As with joint-specific outcomes, few studies have been operative scores improved. A non-linear relationship was performed comparing overall quality of life following TKR estimated for the OKS, OHS and EQ-5D-3L index following and THR for OA and RA. The aforementioned study by THR, with the size of the increase in post-operative score Borstlap et al. [26] found that those with RA achieved a generally reduced as pre-operative scores reached higher smaller gain in overall quality of life following THR than levels. Older age, male gender, being in a more deprived those with OA. Another study compared outcomes for IMD quintile, comorbidities, higher BMI and being a smoker 130 TKR (76 OA and 54 RA) and 205 THR (164 OA and were, in general, associated with worse post-operative 41 RA) patients in one English hospital [27], finding that scores, although in many cases these differences were not those with OA generally achieved greater improvements statistically significant. than those with RA. These findings are in line with those from our study. Discussion Key findings Study strengths and limitations Individuals with OA and RA had substantial improvements While previous studies have typically been based on in the OKS, OHS and EQ-5D-3L index following TKR and single centres, this study was informed by PROMs rou- THR. The absolute change in scores (the journey) indi- tinely collected across the English NHS. These data were cated clinically meaningful improvements in health [21, linked to individuals’ hospital and primary care records. 22]. The PoPC (the proportion of journey towards perfect While there is often concern about the accuracy of coding health completed) achieved indicated that, on average, when using routinely collected data, RA was identified patients reached >50% of their potential improvement. from primary care records using validated diagnostic clin- Indeed, those with OA reached >80% of their potential ical codes [28]. As well as providing generalizable find- improvement in the OHS following THR. ings, the sample size and variables collected allowed for While those with RA had an absolute change in scores the control of potentially key confounding variables. similar or better than those with OA, the PoPC for those Moreover, the modelling approach used in this study did with RA was generally lower, particularly for the EQ-5D-3L not categorize continuous variables and incorporated index. This was likely because individuals with RA had non-linearity where merited. worse pre-operative scores and hence greater potential In this study we assessed PROMs for joint-specific out- for improvement. After accounting for the difference in comes and overall quality of life. However, these outcome baseline scores and patient characteristics, diagnosis of measures were only recorded at one point in time follow- RA relative to OA had little effect on post-operative joint- ing surgery. Consequently, it was not possible to consider specific scores. However, RA was associated with a re- whether the effect of diagnosis varies with time since sur- duction in the post-operative EQ-5D-3L index of 0.09 and gery. While this would be of interest, our analysis can be 0.08 for TKR and THR, respectively. expected to have captured the effect of diagnosis on the peak gain from surgery. Previous studies have found that Study findings in context the full gains from undergoing TKR and THR have typically A relatively small number of studies have compared joint- been realized by 6 months following surgery [29]. specific PROMs of people with RA and OA following TKR Only those individuals with both pre- and post-opera- and THR, and the findings of the studies that have are tive PROMs were included in the analysis. Bias may have somewhat mixed. One study, with study participants re- arisen if those who did not complete PROMs were sys- cruited from the practices of rheumatologists, compared tematically different from those who did. However, a com- outcomes following TKR of 834 individuals with RA and parison of the patient characteristics with and without 1022 https://academic.oup.com/rheumatology RA and patient-reported outcomes PROMs recorded found these groups to be broadly financial support from Amgen, all outside the submitted similar. work. The other authors declared no conflicts of interest. Lastly, in this study we provide a broad comparison of PROMs for those with a diagnosis of OA or RA. However, individuals with these diagnoses are not homogeneous Supplementary data and patient experiences are likely to vary depending on disease-specific factors, such as disease severity and Supplementary data are available at Rheumatology online. treatment history. Analysis of such factors was beyond the scope of this study, with few disease-specific factors References recorded in routinely collected data. Thorough analysis of the impact of disease-specific factors would require pro- 1 Cooper C, Adachi JD, Bardin T et al. How to define re- spectively collected data. sponders in osteoarthritis. Curr Med Res Opin 2013;29:719–29. Study implications 2 Pivec R, Johnson AJ, Mears SC, Mont MA. Hip arthro- plasty. Lancet 2012;380:1768–77. Joint-specific outcomes appear to be similar for those 3 Carr AJ, Robertsson O, Graves S et al. Knee replacement. with OA and RA following TKR and THR, but improve- Lancet 2012;379:1331–40. ments in overall quality of life following surgery appear 4 Wilkinson JM, Stanley D, Getty CJM. Surgical manage- to be somewhat less for those with RA. This is likely be- ment of the rheumatoid patient. Curr Orthop cause the health of individuals with RA is being limited due 2004;18:357–70. to other issues beyond the targeted joint, with RA being a systemic inflammatory disease that involves multiple 5 National Institute for Health and Clinical Excellence. Rheumatoid arthritis in adults: management. NG100. London: joints and typically necessitates polypharmacy. National Institute for Health and Clinical Excellence, 2009. Consequently, while TKR and THR are likely addressing the main health problem of most patients with OA, it is 6 Clement ND, Breusch SJ, Biant LC. Lower limb joint re- placement in rheumatoid arthritis. J Orthop Surg Res only addressing one of a number of health problems for 2012;7:27. many patients with RA, thus limiting the potential improve- ment in overall health from ‘fixing’ the targeted joint. 7 Price AJ, Longino D, Rees J et al. Are pain and function better measures of outcome than revision rates after TKR Indeed, it has previously been observed that although in the younger patient? Knee 2010;17:196–9. TKR and THR improve disease activity scores for individ- uals with RA, they do not generally result in disease ac- 8 Harris KK, Price AJ, Beard DJ et al. Can pain and function be distinguished in the Oxford Hip Score in a meaningful tivity reaching remission [30]. way?: an exploratory and confirmatory factor analysis. Regardless of the diagnosis, we found that TKR and Bone Joint Res 2014;3:305–9. THR typically lead to substantial and clinically meaningful improvements in both joint-specific measures and overall 9 Harris K, Dawson J, Doll H et al. Can pain and function be distinguished in the Oxford Knee Score in a meaningful health-related quality of life. While the procedures should way? An exploratory and confirmatory factor analysis. not be seen as a cure-all, particularly for those with RA, Qual Life Res 2013;22:2561–8. they offer a valuable treatment option for those individuals 10 Devlin NJ, Parkin D, Browne J. Patient-reported outcome with appropriate indications. measures in the NHS: new methods for analysing and reporting EQ-5D data. Health Econ 2010;19:886–905. Acknowledgments 11 Herrett E, Gallagher AM, Bhaskaran K et al. Data resource profile: Clinical Practice Research Datalink (CPRD). Int J The authors would like to thank Susan Thwaite (National Epidemiol 2015;44:827–36. Rheumatoid Arthritis Society) for her role as the patient 12 Herbert A, Wijlaars L, Zylbersztejn A, Cromwell D, Hardelid and public representative and her role on the study steer- P. Data resource profile: hospital episode statistics ing committee. This article presents independent research admitted patient care (HES APC). Int J Epidemiol funded by the National Institute for Health Research 2017;46:1093–3i. (NIHR). The views expressed are those of the authors 13 Secondary Care Analysis (PROMs) NHS Digital. Patient and not necessarily those of the NHS, the NIHR or the Reported Outcome Measures (PROMs) in England: A Department of Health. guide to PROMs methodology. Leeds: 2017. Funding: This work was supported by an NIHR Clinician 14 Armitage JN, Van Der Meulen JH. Identifying co-morbidity in surgical patients using administrative data with the Scientist award (CS-2013-13-012) received by DPA. This Royal College of Surgeons Charlson score. Br J Surg work was supported by the NIHR Biomedical Research 2010;97:772–81. Centre, Oxford. 15 Murray DW, Fitzpatrick R, Rogers K et al. The use of the Disclosure statement: NKA has received personal fees from Oxford hip and knee scores. J Bone Jt Surg Br Freshfields Bruckhaus Deringer, Bioventus, Flexion, Merck 2007;89:1010–4. and Regeneron, all outside the submitted work. DPA reports 16 Dolan P. Modeling valuation for EuroQol health states. grants from Amgen, Servier and UCB Biopharma and non- Med Care 1997;35:1095–108. https://academic.oup.com/rheumatology 1023 Edward Burn et al. 17 Austin PC. An introduction to propensity score methods 25 Goodman SM, Johnson B, Zhang M et al. Patients with for reducing the effects of confounding in observational rheumatoid arthritis have similar excellent outcomes after studies. Multivar Behav Res 2011;46:399–424. total knee replacement compared with patients with 18 Kiran A, Hunter DJ, Judge A et al. A novel methodo- osteoarthritis. J Rheumatol 2016;43:46–53. logical approach for measuring symptomatic change fol- 26 Borstlap M, Zant JL, Van Soesbergen M, Van der Korst lowing total joint arthroplasty. J Arthroplasty JK. Effects of total hip replacement on quality of life in 2014;29:2140–5. patients with osteoarthritis and in patients with rheumatoid 19 Harrell FE. Regression modeling strategies. Cham, arthritis. Clin Rheumatol 1994;13:45–50. Switzerland: Springer International, 2015. 27 Kirwan JR, Currey HLF, Freeman MAR, Snow S, Young 20 White IR, Royston P, Wood AM. Multiple imputation using PJ. Overall long-term impact of total hip and knee joint chained equations: issues and guidance for practice. Stat replacement surgery on patients with osteoarthritis and Med 2011;30:377–99. rheumatoid arthritis. Rheumatology 1994;33:357–60. 21 Beard DJ, Harris K, Dawson J et al. Meaningful changes 28 Hawley S, Delmestri A, Judge A et al. Total hip and knee for the Oxford hip and knee scores after joint replacement replacement among incident osteoarthritis and rheuma- surgery. J Clin Epidemiol 2015;68:73–9. toid arthritis patients within the UK Clinical Practice Research Datalink (CPRD) compared to Hospital Episode 22 Walters SJ, Brazier JE. Comparison of the minimally im- Statistics (HES): a validation study. Pharmacoepidemiol portant difference for two health state utility measures: Drug Saf 2016;25:251. EQ-5D and SF-6D. Qual Life Res 2005;14:1523–32. 29 Schilling C, Dowsey MM, Clarke PM, Choong PF. Using 23 Dusad A, Pedro S, Mikuls TR et al. Impact of total knee patient-reported outcomes for economic evaluation: get- arthroplasty as assessed using patient-reported pain and ting the timing right. Value Health 2016;19:945–50. health-related quality of life indices: rheumatoid arthritis versus osteoarthritis. Arthritis Rheumatol 2015;67:2503–11. 30 Momohara S, Inoue E, Ikari K et al. Efficacy of total joint arthroplasty in patients with established rheumatoid arth- 24 Judge A, Arden NK, Cooper C et al. Predictors of out- ritis: improved longitudinal effects on disease activity but comes of total knee replacement surgery. Rheumatol not on health-related quality of life. Mod Rheumatol 2012;51:1804–13. 2011;21:476–81. 1024 https://academic.oup.com/rheumatology