Rheumatology 2019;58:1950–1954 RHEUMATOLOGY doi:10.1093/rheumatology/kez143 Advance Access publication 23 April 2019 Concise report Lifetime risk of knee and hip replacement following Downloaded from https://academic.oup.com/rheumatology/article/58/11/1950/5476596 by UNIVERSITAT AUTONOMA DE BARCELONA user on 22 September 2021 a diagnosis of RA: findings from a cohort of 13 961 patients from England Edward Burn 1, Christopher J. Edwards2, David W. Murray1, Alan Silman1, Cyrus Cooper1,3, Nigel K. Arden1,3, Rafael Pinedo-Villanueva1 and Daniel Prieto-Alhambra4 Abstract Objective. To estimate the lifetime risk of knee and hip replacement following a diagnosis of RA. Methods. The analysis was undertaken using routinely collected data from the English NHS. Diagnosis of RA was identified using primary care records, with knee and hip replacement observed in linked hospital records. Parametric survival models were fitted for up to 15 years of follow-up, with age, sex, Charlson comorbidity score, socioeconomic status, BMI and smoking status included as explanatory variables. A decision model was used to combine and extrapo- late survival models to estimate lifetime risk. Results. The number of individuals with a diagnosis of RA and included in the study was 13 961. Lifetime risk of knee replacement and hip replacement was estimated to be 22% (95% CI: 16, 29%) and 17% (95% CI: 11, 26%) following a diagnosis of RA for the average patient profile (non-smoking women aged 64 with no other comorbidities, BMI of 27 and in the top socioeconomic quintile). Risks were higher for younger patients. Conclusion. The lifetime risk of knee and hip replacement for individuals with a diagnosis of RA is approximately double that of the general population. These findings allow for a better understanding of long-term prognosis and healthcare resource use, and highlight the importance of timely diagnosis and effective treatment. Key words: rheumatoid arthritis, surgery, epidemiology, knee, hip CLINICAL SCIENCE Rheumatology key messages . The lifetime risk of knee and hip replacement following an RA diagnosis was not known. . Lifetime risk is around 20% for the average patient profile. . The risk for RA is approximately double that of the general population. Introduction and prevent or treat failure of structures. Knee and hip RA is characterized by persistent synovitis and systemic replacement are two of the most common surgeries for inflammation and can lead to the progressive destruction individuals with RA [1]. and secondary osteoarthritis of both small and large There have been substantial advances in management joints. Surgical procedures may help to relieve pain, main- of RA over recent decades, with numerous effective dis- tain or restore function, correct deformity or instability, ease modifying agents introduced coupled with improve- ments in both diagnostic and management strategies. 1 Nuffield Department of Orthopaedics, Rheumatology and Likely as a result, while rates of knee and hip replacement Musculoskeletal Sciences, University of Oxford, Oxford, 2NIHR have been increasing over recent years in the general Wellcome Trust Clinical Research Facility, University Hospital population [2–4], rates among those with RA have re- Southampton, 3MRC Lifecourse Epidemiology Unit, Southampton University, Southampton, UK and 4GREMPAL Research Group, Idiap mained relatively stable [5]. Jordi Gol and CIBERFes, Universitat Autonoma de Barcelona and Fifty-year-old men and women from the general popu- Instituto de Salud Carlos III, Barcelona, Spain lation in the UK have an 8 and 11% lifetime risk of knee Submitted 8 February 2019; accepted 16 March 2019 replacement and a 7 and 11% risk of hip replacement, Correspondence to: Rafael Pinedo-Villanueva, Nuffield Department of respectively [6]. Risks are lower for older individuals Orthopaedics, Rheumatology and Musculoskeletal Sciences, University of Oxford, Oxford B4495, UK. with, for example, 80-year-olds having a 3–4% lifetime E-mail: rafael.pinedo@ndorms.ox.ac.uk risk of knee and hip replacement [6]. It is not known ! 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 Lifetime risk of knee and hip replacement in RA whether the improvements in the management of RA have Estimating lifetime risk led to this risk being reduced to that of the general popu- Analyses were undertaken separately for risk of knee and lation, or whether there remains an increased risk of knee hip replacement. Parametric survival models were esti- and hip replacement for individuals with RA. mated for cause-specific risks of knee or hip replacement Downloaded from https://academic.oup.com/rheumatology/article/58/11/1950/5476596 by UNIVERSITAT AUTONOMA DE BARCELONA user on 22 September 2021 Understanding the lifetime risk of knee and hip replace- and death following diagnosis of RA. Age, sex, Charlson ment following a diagnosis of RA would help provide patients score, IMD quintile, BMI and smoking status were and clinicians with an indication of long-term prognosis and included as explanatory variables in the models. There future health care utilization. Therefore, in this study we esti- was evidence of non-proportionality in age for risks of mated the lifetime risks of knee and hip replacement follow- knee and hip replacement and so models were estimated ing a diagnosis of RA and the effect of patient characteristics separately for age groups based on quartiles of the distri- at time of diagnosis on this risk. bution and with continuous age included as an explana- tory variable within each of the stratified models. Methods A range of distributions were considered for the parametric survival models, with distributions chosen on the basis of their Study participants and data collection fit to the observed data and, if required, the plausibility of their Patient-level data from the English NHS were used. Primary extrapolation. Under an exponential distribution the hazard is care records were extracted from practices within the expected to be constant over time, while under Weibull and Clinical Practice Research Datalink (CPRD), which contains log-normal distributions the hazard is expected to be either medical records and demographic details for around 7% of declining or increasing as time passes. Meanwhile, a spline- the UK population and is broadly representative of the gen- based distribution allows for more flexible shapes, for example eral population [7]. These data were linked to hospital re- where the hazard first decreases but later increases. Knots cords from Hospital Episode Statistics Admitted Patient (inflection points) for spline models were fitted at 33 and 67% Care, which has data on all admissions to English NHS hos- quantiles of log time. When incorporating continuous explana- pitals and admissions to independent sector hospitals paid tory factors (age and BMI), non-linearity was accounted for for by the NHS [8], and mortality data from the Office for using quadratic terms if the Akaike information criterion was National Statistics. The linked dataset covered 1997–2014. reduced compared with if they were specified as linear. Ethical approval for the study was granted by the CPRD To estimate lifetime risk, the survival models were com- Independent Scientific Advisory Committee. bined using a state-based Markov model. Risk of knee and Individuals were included if they had an incident diag- hip replacement was extrapolated over the remaining life- nosis of RA recorded in CPRD during the study period. RA times. Risk of mortality was based on the relevant paramet- was identified using an established diagnostic algorithm ric models for the first 15 years following diagnosis, after with high validity for CPRD, with >80% sensitivity and which they were assumed to revert to estimates based on specificity [6]. Individuals were excluded from the study age- and sex-specific lifetables for the UK. Models were run if they had a history of knee or hip replacement prior to for individuals with average characteristics (median for con- diagnosis, although no minimum prior observation time tinuous variables and mode for categorical ones). To esti- was required for inclusion into the study. A sensitivity ana- mate the partial effect of explanatory factors on lifetime risk, lysis was performed with study participants excluded if models were re-run with participant profiles varying in the they had less than a year of observation time prior to explanatory factor of interest, while holding other character- diagnosis. Individuals were also excluded if they were istics constant at their average. younger than 40 years of age at time of diagnosis as there was insufficient follow-up to plausibly estimate life- Results time risk for them. The following patient characteristics at time of diagno- The number of individuals diagnosed with RA and sis were extracted: age, sex, Charlson score, index of included in the analysis was 13 961. A study flowchart is multiple deprivation (IMD), BMI and smoking status. provided in Supplementary Fig. S1, available at Charlson scores, which provide a measure of comorbidity, Rheumatology online and patient characteristics at time were categorized as 0 or 1+ as relatively few study par- of diagnosis are given in Table 1. ticipants had multiple comorbidities. IMD, a measure of Over observed follow-up, 672 individuals had a knee socioeconomic status, was extracted with individuals replacement and 493 had a hip replacement following a grouped by quintile (where quintile 5 represented the diagnosis of RA. The 15-year cumulative incidence of most deprived). Smoking status (classified as non- knee and hip replacement was 9.69% (95% CI: 8.75, smoker, ex-smoker or current smoker) and BMI were 10.73%) and 7.69% (95% CI: 6.80, 8.71%), respectively. identified using records from within the year prior to diag- The lifetime risk of knee replacement and hip replacement nosis, with the record closest to time of diagnosis used. was 22% (95% CI: 16, 29%) and 17% (95% CI: 11, 26%), Study participants were followed for up to 15 years after respectively, for the average patient profile (age: 64 years; diagnosis. Knee and hip replacement were identified using sex: female; Charlson score: 0; IMD quintile: 1; BMI: procedure codes within Hospital Episode Statistics 27 kg/m2; smoking status: non-smoker). Admitted Patient Care. Mortality was identified using The cumulative incidence stratified for the different explana- Office for National Statistics records. tory factors is summarized in detail in Supplementary Table S1, https://academic.oup.com/rheumatology 1951 Edward Burn et al. TABLE 1 Study participant characteristics at time of RA having comorbidities at time of diagnosis was associated diagnosis with a lower lifetime risk of knee and hip replacement (see Supplementary Fig. S10, available at Rheumatology online), with differences mainly due to the effect of comorbidities on Characteristic (n = 13 961) Downloaded from https://academic.oup.com/rheumatology/article/58/11/1950/5476596 by UNIVERSITAT AUTONOMA DE BARCELONA user on 22 September 2021 risk of mortality (see Supplementary Tables S2 and S3, avail- Age, median (IQR), years 64.0 (54.0–73.0) able at Rheumatology online). Point estimates also sug- Age, n (%) gested that a higher BMI at diagnosis was associated with 540 to <54 3216 (23.0) increased lifetime risks, particularly for knee replacement 554 to <64 3720 (26.6) (see Supplementary Fig. S12, available at Rheumatology 564 to <73 3212 (23.0) online), driven by an estimated effect on cause-specific 573 to 499 3813 (27.3) risks for surgery (see Supplementary Tables S2 and S3, Sex, n (%) available at Rheumatology online). Women 9468 (67.8) Men 4493 (32.2) Requiring a year of observation time led to 9% of study Charlson score, n (%) participants being excluded. This had relatively little 0 12 399 (88.8) impact on estimated lifetime risk for individuals with an 1+ 1562 (11.2) average patient profile or older (see Supplementary Fig. BMI, median (IQR), kg/m2 27.1 (23.9–31.4) S13, available at Rheumatology online). For younger indi- BMI group, n (%) viduals, however, lifetime risk was estimated to be lower Normal or underweight 1701 (12.2) with fewer knee and hip replacements within the youngest Obese class I 988 (7.1) age group. Obese class II 431 (3.1) Obese class III 225 (1.6) Overweight 1788 (12.8) Discussion Missing 8828 (63.2) IMD, n (%) Lifetime risk of knee replacement and hip replacement 1 (least deprived) 3132 (22.4) was estimated to be 22% (95% CI: 16, 29%) and 17% 2 3118 (22.3) (95% CI: 11, 26%), respectively, following a diagnosis of 3 2988 (21.4) RA for the average patient profile. Risks were found to be 4 2700 (19.3) higher for younger patients, but were similar by gender, 5 (most deprived) 2001 (14.3) socioeconomic status and smoking status. Point esti- Missing 22 (0.2) mates implied that comorbidities and a higher BMI at Smoking status, n (%) time of diagnosis may increase lifetime risks. Non-smoker 2465 (17.7) The lifetime risks of knee and hip replacement for 50- Ex-smoker 2452 (17.6) year-olds from the general UK population has previously Current smoker 1642 (11.8) Missing 7402 (53.0) been estimated respectively as 11 and 12% for women and 8 and 7% for men [7], with primary osteoarthritis being the most common indication for surgery. Based IMD: index of multiple deprivation; IQR: interquartile range. on our findings, despite the improvement in the diagnosis and treatment of the disease, RA is still associated with a lifetime risk of both knee and hip replacement approxi- mately double that of the general population. A number available at Rheumatology online. Timing to knee and hip of studies have described rates of knee and hip replace- replacement differed across the age stratified models. In ment for individuals with RA [8–15], and our findings of general, risks were closer to being constant over time for lifetime risks are broadly in line with what would be ex- younger patients whereas risks peaked shortly after diag- pected based on these studies. nosis for older patients (see Supplementary Figs S2–S9, While being older has been associated with an available at Rheumatology online, for details). The effect of increased risk of orthopaedic surgery for people with RA age on risk of knee and hip replacement appears to be [10], those aged over 80 have been seen to have lower risk non-linear with increasing age first associated with an of knee and hip replacement compared with younger pa- increased risk for younger patients but then a falling risk tients [9]. Patients over 80 who underwent surgery also for older patients, while risk of mortality increased as age had their procedure more quickly following diagnosis increased (see Supplementary Tables S2 and S3, avail- than younger patients [9]. A similar relationship between able at Rheumatology online). The effect of age on mor- age and risk for surgery was seen in this study, with non- tality, and so time at risk, appeared to drive differences in proportionality in hazards (with older patients more likely lifetime risk with younger age at diagnosis associated with to have surgery shortly after diagnosis) and age having a a higher lifetime risk (see Fig. 1). non-linear association with risks of knee and hip replace- Lifetime risk estimates were similar by gender (see ment (with risks first increasing and then decreasing as Fig. 1), socioeconomic status (see Supplementary Fig. age at diagnosis increased). S11, available at Rheumatology online) and smoking The strength of this study include that it was informed status (see Supplementary Fig. S13, available at by real world, routinely collected data with a large and Rheumatology online). Point estimates implied that representative study cohort. Fitting parametric survival 1952 https://academic.oup.com/rheumatology Lifetime risk of knee and hip replacement in RA FIG. 1 Partial effect of age and sex on lifetime risk of knee replacement and hip replacement Downloaded from https://academic.oup.com/rheumatology/article/58/11/1950/5476596 by UNIVERSITAT AUTONOMA DE BARCELONA user on 22 September 2021 Point estimates with 95% CI. HR: hip replacement; KR: knee replacement. models allowed lifetime risk to be estimated and the can also be expected to have an important effect on life- impact of patient characteristics on this risk to be as- time risk, with immediate initiation of treatment with dis- sessed, but the required extrapolation leads to sizeable ease-modifying antirheumatic drugs found to reduce risk uncertainty particularly for younger patients (as reflected of surgery when compared with a delayed start [15]. in the wide CIs and the impact of requiring study partici- In summary, an individual with an average profile at pants to have a year of observation time prior to diagno- diagnosis of RA has a slightly lower than 1 in 5 chance sis). The estimates from this study were based on hospital of having a knee replacement and a slightly higher than 1 procedures funded by the NHS and so are likely to have in 5 chance of hip replacement over his or her remaining missed some of the surgeries done in the private sector, lifetime. This is approximately double the risk of the gen- leading to lifetime risk being somewhat underestimated. eral population. Risks are higher for those at a younger While individuals were excluded if they had a record of age when diagnosed. knee or hip replacement prior to diagnosis, 9% of study participants had less than a year of prior observation time Acknowledgements and 39% had less than 5 years, hence limiting our ability to definitely exclude those with a previous knee or hip D.P.A. is funded by a National Institute for Health replacement. Another limitation of the study is that while Research Clinician Scientist award (CS-2013–13-012). the effects of patient characteristics identified at time of The authors would like to thank Miss Susan Thwaite diagnosis were considered, factors such as BMI and (National Rheumatoid Arthritis Society) for her role as comorbidities may change over time and so may have a the patient and public representative and her role in the time-varying effect on risks. In addition, there are a study steering committee. E.B., C.J.E., D.W.M., A.S., number of other factors not collected in the data we C.C., N.K.A., D.P.A. and R.P.V. all made substantial con- used that are likely to influence risks. In particular, disease tributions to conception and design of the study. E.B., severity at time of diagnosis can be expected to have a R.P.V. and D.P.A. undertook the statistical analysis. substantial impact on the risks, with baseline functional E.B., R.P.V. and D.P.A. drafted the manuscript with disability previously found to be a strong predictor for C.J.E., D.W.M., A.S., C.C. and N.K.A. revising it for im- undergoing surgery [10]. Treatment following diagnosis portant intellectual content. All authors read and approved https://academic.oup.com/rheumatology 1953 Edward Burn et al. the final manuscript. CPRD data with Hospital Episode England and Wales: an interrupted time-series analysis. Statistics linkage were provided under a licence that Semin Arthritis Rheum 2018;47:605–10. does not permit sharing. Data are obtainable from CPRD 6 Thomas SL, Edwards CJ, Smeeth L et al. How accurate subject to a full application. are diagnoses for rheumatoid arthritis and juvenile idio- Downloaded from https://academic.oup.com/rheumatology/article/58/11/1950/5476596 by UNIVERSITAT AUTONOMA DE BARCELONA user on 22 September 2021 pathic arthritis in the general practice research database? Funding: This article presents independent research Arthritis Care Res 2008;59:1314–21. funded by the National Institute for Health Research 7 Culliford DJ, Maskell J, Kiran A et al. The lifetime risk of (NIHR). The views expressed are those of the authors total hip and knee arthroplasty: results from the UK gen- and not necessarily those of the NHS, the NIHR or the eral practice research database. Osteoarthritis Cartilage Department of Health. This work was supported by the 2012;20:519–24. NIHR Biomedical Research Centre, Oxford. 8 Shourt CA, Crowson CS, Gabriel SE et al. Utilization of Disclosure statement: N.K.A. has received personal fees from orthopedic surgery among patients with rheumatoid arth- Freshfields Bruckhaus Deringer, Bioventus, Flexion, Merck, ritis 1980–2007: a population-based study focused on and Regeneron, all outside the submitted work. D.P.A. re- surgery rates, sex, and mortality. J Rheumatol ports grants from Amgen, Servier and UCB Biopharma, and 2012;39:481–5. non-financial support from Amgen, all outside the submitted 9 Waljee J, Zhong L, Baser O et al. The incidence of upper work. D.W.M. reports grants and personal fees from Zimmer and lower extremity surgery for rheumatoid arthritis Biomet. In addition, D.W.M. has various patents related to among Medicare beneficiaries. J Bone Joint Surg Am Unicompartmental Knee Replacement (Zimmer Biomet) with 2015;97:403–10. royalties paid, all outside the submitted work. 10 Gwinnutt JM, Symmons DPM, MacGregor AJ et al. Predictors of and outcomes following orthopaedic joint surgery in patients with early rheumatoid arthritis followed Supplementary data for 20 years. Rheumatology 2017;56:1510–7. 11 Wolfe F, Zwillich SH. The long-term outcomes of Supplementary data are available at Rheumatology online. rheumatoid arthritis: a 23-year prospective, longitudinal study of total joint replacement and its predictors in 1,600 patients with rheumatoid arthritis. Arthritis Rheum References 1998;41:1072–82. 1 James D. Orthopaedic intervention in early rheumatoid 12 Da Silva E, Doran MF, Crowson CS et al. Declining use of arthritis. Occurrence and predictive factors in an inception orthopedic surgery in patients with rheumatoid arthritis? cohort of 1064 patients followed for 5 years. Results of a long-term, population-based assessment. Rheumatology 2003;43:369–76. Arthritis Rheum 2003;49:216–20. 2 Culliford DJ, Maskell J, Beard DJ et al. Temporal trends in 13 Massardo L, Gabriel SE, Crowson CS et al. A population hip and knee replacement in the United Kingdom. J Bone based assessment of the use of orthopedic surgery in Joint Surg Br 2010;92–B:130–5. patients with rheumatoid arthritis. J Rheumatol 3 Cram P, Lu X, Kates SL et al. Total knee arthroplasty 2002;29:52–6. volume, utilization, and outcomes among Medicare ben- 14 Nikiphorou E, Carpenter L, Morris S et al. Hand and foot eficiaries, 1991–2010. JAMA 2012;308:1227. surgery rates in rheumatoid arthritis have declined from 4 Cram P, Lu X, Callaghan JJ et al. Long-term trends in hip 1986 to 2011, but large-joint replacement rates remain arthroplasty use and volume. J Arthroplasty unchanged: results from two UK inception cohorts. 2012;27:278–85.e2. Arthritis Rheumatol 2014;66:1081–9. 5 Hawley S, Cordtz R, Dreyer L et al. Association between 15 Verstappen SMM, Hoes JN, Ter Borg EJ et al. Joint sur- NICE guidance on biologic therapies with rates of hip and gery in the Utrecht Rheumatoid Arthritis Cohort: the effect knee replacement among rheumatoid arthritis patients in of treatment strategy. Ann Rheum Dis 2006;65:1506–11. 1954 https://academic.oup.com/rheumatology