Article Text
Abstract
Objectives To investigate predictors of 10-year risk of revision and 1-year risk of prosthetic joint infection (PJI) and death following total hip/total knee arthroplasty (THA/TKA) in (1) patients with rheumatoid arthritis (RA) compared with patients with osteoarthritis (OA); and (2) patients with RA treated with biological disease-modifying antirheumatic drugs (bDMARD) within 90 days preceding surgery compared with non-treated.
Methods Register-based cohort study using the Danish National Patient Register, the DANBIO rheumatology register (RA-specific confounders and treatment episodes) and the Danish Hip and Knee Arthroplasty Registers. Survival analyses were used to calculate confounder-adjusted sub-HRs (SHR) and HRs.
Results In total, 3913 patients with RA with THA/TKA were compared with 120 499 patients with OA. Patients with RA had decreased risk of revision (SHR 0.71 (0.57–0.89)), but increased risk of PJI (SHR=1.46 (1.13–1.88)) and death (HR=1.25 (1.01–1.55)). In DANBIO, 345 of 1946 patients with RA with THA/TKA had received bDMARD treatment within 90 days preceding surgery. bDMARD-treated patients did not have a statistically significant increased risk of revision (SHR=1.49 (0.65–3.40)), PJI (SHR=1.61 (0.70–3.69)) nor death (HR=0.75 (0.24–2.33)) compared with non-treated. Glucocorticoid exposure (HR=2.87 (1.12–7.34)) and increasing DAS28 (HR=1.49 (1.01–2.20)) were risk factors for mortality.
Conclusion Patients with RA had a decreased 10-year risk of revision while the risk of death and PJI was increased compared with patients with OA following THA/TKA. bDMARD exposure was not associated with statistically significant increased risk of neither PJI nor death in this study. Glucocorticoid exposure and increased disease activity were associated with an increased risk of death.
- rheumatoid arthritis
- orthopaedic surgery
- DMARDs (biologic)
- infections
- epidemiology
Statistics from Altmetric.com
Background
The introduction of biological disease-modifying antirheumatic drugs (bDMARD) has improved the treatment of patients with rheumatoid arthritis (RA), but due to their immunosuppressive actions, they could potentially increase the risk of prosthetic joint infection (PJI) following major joint surgery such as total hip arthroplasty (THA) and total knee arthroplasty (TKA).1 2 PJI is a serious complication associated with inferior outcomes in terms of pain, morbidity and mortality3; and it constitutes an economic burden in healthcare budgets with longer duration of hospital stays, extensive antibiotic treatment and prolonged recovery.4 Most,5–11 but not all studies,7 8 have found increased risk of PJI and surgical complications in patients with RA following THA and TKA but not with regard to the overall risk of revision.12 Results have been contradictory as to whether bDMARD treatment affects the risk of PJI.2 13–15
Patients with RA have increased mortality rates compared with the general population. However, with few exceptions,16 17 existing studies have reported no difference or decreased risk of short-term mortality in RA compared with other THA/TKA recipients.5–10 It is unknown whether bDMARD treatment affects the mortality risk.
The aim of this study was to estimate the risk of revision due to non-infectious causes, PJI and death among patients with RA following primary THA or TKA. We compared (1) patients with RA with patients with osteoarthritis (OA) and (2) bDMARD-treated with non-bDMARD-treated patients with RA using Danish healthcare registers. Further, we aimed at estimating the impact of glucocorticoid treatment and disease activity on the risk of revision, PJI and death among patients with RA.
Patients and methods
Study design and setting
Register-based cohort study from Denmark from 1 January 2000 to 31 December 2014. Every Danish resident receives a 10-digit personal identification number consistent throughout all national registers making register linkage possible. Study methods and results are presented in accordance with Strengthening the Reporting of Observational Studies in Epidemiology guidelines.18
Data sources
The Civil Registration System (CRS): registers dates of deaths and migrations among all Danish citizens.19
The National Patient Register (NPR): includes information on all inpatient (1977–) and outpatient (1995–) visits at Danish hospitals.20 Discharge diagnoses are registered in accordance with the International Classification of Diseases (ICD) (ICD-8: 1977–1993; ICD-10: 1994–) with up to 20 diagnoses for each discharge.
DANBIO: established in 2000 as a national register for use and efficacy of bDMARDs in treatment of inflammatory arthritis (coverage >91%), and more than 25 000 patients with RA are registered. 21 Information includes date of diagnosis, antirheumatic treatment series, Health Assessment Questionnaire Disability Index (HAQ-DI), Disease Activity Score using the 28-joint count (DAS28) and C-reactive protein (CRP).
Danish Hip Arthroplasty Register (DHR): is a nationwide register used for mandatory registration of all primary and revision THA surgeries established in 1995.22 The coverage for primary THA surgery is 97%. Only PJIs treated with surgery are registered in DHR.23
Danish Knee Arthroplasty Register (DKR): established in 1997, DKR is a nationwide database in which orthopaedic surgeons register all primary and revision TKA surgeries.24 In 2010, the completeness of registration was estimated to 97% for both primary and revision surgeries.
Study population
Patients with RA. Patients with RA were identified in DANBIO and NPR. Until 2006, only patients treated with bDMARDs were mandatorily registered in DANBIO. To include more bDMARD-naive patients, we identified patients registered in NPR with a diagnosis of RA at an inpatient or outpatient facility specialised in rheumatology or general internal medicine.25 Identified patients with RA were subsequently linked with DHR and DKR to identify those with primary THA/TKA during 2000–2014. Only surgeries performed after RA diagnosis were included; and patients only contributed with their first primary THA/TKA in case of multiple (eg, bilateral THA/TKA). Patients were excluded if, according to DHR/DKR, THA/TKA was performed for reasons other than RA or OA sequelae (eg, fracture, avascular necrosis etc.).
Patients with OA. In DHR and DKR, we identified patients with a first primary THA/TKA due to OA and no history of RA.
bDMARD treatment in patients with RA. Patients with RA registered in DANBIO were used for these analyses to obtain information on treatment episodes of DMARD and glucocorticoids as well as disease activity and severity markers. For each patient, information on start and stop dates for treatment episodes was obtained and categorised according to drug types (bDMARD, conventional synthetic DMARD (csDMARD) and glucocorticoid).21 bDMARD, csDMARD or glucocorticoid exposure was defined as treatment within 90 days preceding surgery. bDMARD-treated patients were compared with patients registered in DANBIO who had not been exposed to bDMARDs during this 90-day window (non-bDMARD treated).
Outcomes
Primary outcomes were 10-year risk of revision for non-infectious reasons, 1-year risk of PJI treated with or without surgery23 and 1-year risk of death following primary THA/TKA. Revision and PJIs treated surgically were captured in DHR and DKR where the cause of revision surgery is preoperatively registered by the surgeon. The positive predictive value (PPV) of PJI in DHR is 66%.26 We further captured PJIs not registered in DHR/DKR using the NPR (ICD-10: T84.5, infection and inflammatory reaction due to internal joint prosthesis; PPV=85%).27 Mortality data were obtained by linkage to CRS. Secondary outcomes were 10-year risk of PJI and death.
Follow-up
Follow-up started at date of primary THA/TKA. In analyses of revision, follow-up ended at date of revision, PJI, death, emigration, at 10 years of follow-up or the end of 2014, whichever came first. In PJI analyses, follow-up ended at date of PJI, death, emigration, end of 2014 or at 1 year of follow-up, whichever came first. In mortality analyses, follow-up ended at date of death, emigration, end of 2014 or at 1 year of follow-up, whichever came first.
Confounders
Demographics
All multivariable analyses were adjusted for age at surgery, sex, calendar period and duration of surgery.
Comorbidities
Through linkage with NPR, we obtained information on diagnoses of chronic obstructive pulmonary disease (COPD), diabetes mellitus, ischaemic heart disease and hospitalisation due to infection with a look-back window of 20 years (see online supplementary table 1 for details).
Supplementary file 1
RA-specific confounders
We adjusted for seropositive RA (DANBIO-reported positive IgM-rheumatoid factor and/or anti-citrullinated protein antibodies; or an ICD-10 code of seropositive RA (M05.8/M05.9) in NPR). For DANBIO patients, registration of DAS28-CRP and HAQ-DI within 1 year prior to surgery was obtained.
Statistical analyses
Demographic and descriptive data are presented as means and SDs. Groups were compared by independent t-test and χ2 test as appropriate. For each exposure group and outcome, the number of events, person-years at risk, crude incidence rates and rate ratios with 95% CIs were calculated.
Univariable and multivariable Fine-Gray competing risks regression analyses were used to calculate sub-HRs (SHR) for revision and PJI, whereas Cox proportional hazard models were used to calculate HRs for death. Time since surgery was used as underlying time scale in all models. The proportional hazard assumption was checked by tests based on Schoenfeld residuals. Different confounder adjustment approaches were applied in multivariable analyses as described below.
Analyses of patients with RA versus patients with OA
(1) A basic model containing age at surgery (1-year increments), sex, calendar time of surgery (2000–2002, 2003–2005, 2006–2008, 2009–2011, 2012–2014) and duration of surgery. (2) A model that further included adjustment for comorbidities.
Analyses of bDMARD-treated compared with non-bDMARD-treated patients with RA
First, a basic model with bDMARD treatment (yes/no) as the key exposure variable and adjusted for age at surgery, sex and calendar period of surgery (2000–2009 vs 2010–2014) was created. Due to few events and our main interest being impact of bDMARD treatment, we used the change-of-estimate approach to decide on confounders to be included in the final model.28 Covariates changing the SHR/HR with ≥10% for the bDMARD exposure variable when added to the basic model were tested in a stepwise forward-selection manner. Tested covariates are listed in the confounder section above. We tested for (and found no) effect modification between bDMARD and glucocorticoid exposure, and csDMARD and glucocorticoid exposure on each of the outcomes.
Multiple imputation
Due to missing values for DAS28-CRP and HAQ, multiple imputation by chained equations (MICE) was carried out using the MI suite in Stata29 (see online supplementary files).
In all tests, P values <0.05 were considered statistically significant. Statistical analyses were performed using Stata (V.13.1, StataCorp, Texas, USA) and R V.3.3.0.30
Sensitivity analyses
Sensitivity analyses included adjustment for comorbidities using the Charlson Comorbidity Index (CCI) (see online supplementary table 2).31 Also, to distinguish a potential spike in mortality caused by surgery from the generally increased mortality associated with RA, we calculated crude mortality rates, and adjusted HRs for years 0–1 and 1–10 after surgery.
In exploratory analyses, we investigated if there was a dose–response effect of glucocorticoid exposure on risks of PJI and death comparing users treated with doses ≤7.5 mg and doses >7.5 mg, respectively with non-users.32 Similarly, we compared patients who received intra-articular/intramuscular glucocorticoid injections prior to surgery with glucocorticoid non-users.
Results
Patients with RA compared with patients with OA
Following register linkage, 3913 patients with RA and primary THA/TKA were available for comparison with 120 499 patients with OA (patient flow chart in online supplementary figure 1). Patients with RA were more likely to be female, younger at time of surgery and suffer from comorbidities (table 1).
Revision
Of 3913 patients with RA, 81 had surgical revision for reasons other than PJI within 10 years of primary THA/TKA resulting in a crude rate ratio of 0.82 (95% CI 0.66 to 1.03) for revision (table 2). A decreased risk of revision among patients with RA remained in the final multivariable model (table 3).
Prosthetic joint infection
Within a year of surgery, 63 patients with RA had a PJI, and were at increased risk of PJI in univariable and multivariable analyses with a final adjusted SHR of 1.46 (95% CI 1.13 to 1.88) (tables 2 and 3). Other risk factors for PJI were male sex, increasing duration of surgery, a history of hospitalisation due to infection, COPD and diabetes mellitus (table 3).
Death
During the first year, 86 patients with RA died and RA was associated with increased mortality risk in univariable and multivariable analyses (HR in final adjusted model 1.25; 95% CI 1.01 to 1.55) (table 2). Other risk factors were comorbidities and increasing duration of surgery, whereas women had a decreased risk of death.
bDMARD-treated compared with non-bDMARD-treated patients with RA
In DANBIO, 1946 patients with RA and primary THA/TKA were identified. Of these, 345 had received bDMARD treatment within 90 days preceding surgery (table 4). On average, bDMARD-treated patients were younger at time of surgery, had higher DAS28-CRP and HAQ-DI, and a higher proportion was seropositive and treated with concomitant csDMARD and glucocorticoids (table 4).
Revision
Nine revisions occurred among the 345 bDMARD-treated compared with 28 among 1601 non-bDMARD-treated patients (table 5). In the basic model, the SHR for long-term revision was 1.69 (95% CI 0.79 to 3.61) among bDMARD-treated compared with non-bDMARD-treated patients (table 6). Following the change-of-estimate procedure, glucocorticoid was added to the final model resulting in an SHR of 1.49 (95% CI 0.65 to 3.40).
Prosthetic joint infection
Only nine PJI cases were observed in the first year following surgery among bDMARD-treated patients (table 5). Following the change-in-estimate procedure, glucocorticoid exposure and DAS28-CRP were added to the model resulting in an SHR of 1.61 (0.70–3.69) for PJI among bDMARD-treated patients (table 6).
Death
Five deaths occurred among 329 bDMARD-treated patients during the first year following surgery resulting in an HR of 1.44 (95% CI 0.53 to 3.88) in the basic model compared with non-bDMARD-treated patients (tables 5 and 6). This estimate, however, was reduced to 0.75 (95% CI 0.24 to 2.33) when glucocorticoid treatment was added to the model (table 6).
Sensitivity analyses
Adjusting for CCI did not change the results in analyses of RA compared with OA nor bDMARD-treated compared with non-bDMARD-treated patients (online supplementary tables 3 and 4). Analysing 10-year risk of PJI and death in patients with RA compared with OA increased the risk estimates (PJI: SHR=1.84 (1.55–2.18); death: HR=1.58 (1.47–1.69)) (online supplementary table 5). Mortality rates among patients with RA in years 0–1 compared with years 1–10 did not reveal increased mortality in the first year compared with later years (HR 1.26 (1.01–1.56) in year 0–1 vs 1.62 (1.51–1.75) in years 1–10) (online supplementary table 6). Ten-year risk of PJI and death did not differ from 1-year risk among bDMARD-treated compared with non-bDMARD-treated patients (online supplementary table 7). For glucocorticoids, there was a dose–response relation on the risk of PJI and death with highest risk estimates observed among users treated with doses >7.5 mg (PJI: SHR=3.21, 95% CI 1.07 to 9.67; death: HR=4.16, 95% CI 1.32 to 13.08), but users treated with doses ≤7.5 mg also had increased risk compared with non-users (PJI: SHR=1.21, 95% CI 0.35 to 4.15; death: HR=3.32, 95% CI 1.06 to 10.41) (online supplementary table 8). Intra-articular/intramuscular glucocorticoid injections within 90 days prior to surgery were associated with increased risk of death (HR=10.80, 95% CI 2.93 to 39.80) (online supplementary table 9).
Discussion
In this nationwide register-based cohort study, we investigated the risk of long-term revision due to non-infectious causes and 1-year risks of PJI and death among patients with RA following THA/TKA: our main findings were increased risks of PJI and death, but a decreased long-term risk of revision in patients with RA compared with patients with OA. bDMARD-treated patients with RA were not at increased risk of PJI nor death. Glucocorticoid treatment proved a significant risk factor for PJI and death with clear dose–response relations.
One previous study and a meta-analysis of two studies have also reported a decreased risk of revision among patients with RA.8 12 Possible explanations for a reduced risk of revision are decreased wear of the implant due to lower physical activity in patients with RA compared with patients with OA, and surgeons possibly being less prone to perform revisions on patients with RA due to comorbidities and fragile bone stocks.33 34 bDMARD treatment did not have an impact on the risk of revision, and we are not aware of other studies that have investigated this.
In line with previous studies, we found an increased risk of PJI among patients with RA.7 10 11 35–37 In a study from Canada, Ravi et al reported an HR of 1.47 for the 2-year risk of PJI among patients with RA compared with patients with OA following TKA, similar to our 1-year risk estimate (HR 1.46).7 We investigated the 1-year risk of PJI as most PJIs occur during this period.23However, in the analysis of 10-year risk of PJI, the risk estimates increased further (SHR 1.84). Scandinavian studies have reported increased risk of late, haematogenous PJIs among patients with RA compared with other THA/TKA recipients, which might explain the increase in long-term risk estimates in our study.11 38 39 We did not find a statistically increased risk of PJI among patients with RA treated with bDMARDs. A meta-analysis from 2014 reported an increased risk of surgical site infection with perioperative exposure to tumour necrosis factor-alpha inhibitor treatment.2 However, the meta-analysis was limited by the heterogeneity of the included studies and use of unadjusted estimates. Our results indicate that ongoing inflammation (high DAS28) and use of glucocorticoids may constitute even greater risk factors for PJI than bDMARD treatment. In accordance with our results, a recent study by George et al reported an increased risk of infection and PJI with perioperative glucocorticoid treatment (HR 2.70 for >10 mg/day compared with non-users) in relation to THA/TKA among 4288 patients receiving infliximab treatment for various inflammatory diseases.15 We report even higher risk estimates with an SHR of 2.31 for PJI within the first year for glucocorticoid users (any dose); and a SHR of 4.88 for doses >7.5 mg compared with non-users. Crowson et al and Zink et al have published risk scores for serious infections in patients with RA that support our findings: glucocorticoid exposure increases the risk of serious infections in a dose-dependent fashion, whereas bDMARDs do not have the same impact.32 40 Recently published US guidelines on the perioperative management of antirheumatic medication in patients with RA undergoing THA/TKA recommend withholding bDMARDs before and after surgery, whereas glucocorticoids can be continued in daily doses but should not be given in ‘stress-doses’.41
Our finding of an increased 1-year mortality in patients with RA versus patients with OA is in agreement with an Australian study.17 The increased mortality rates in years 1–10 compared with year 0–1 after surgery in our study suggest disease-specific rather than surgery-induced mortality. This does, however, not explain why other studies have reported similar or even decreased short-term (<90 days) risk of death among patients with RA compared with other THA/TKA recipients.7 8 36 42–44 bDMARD exposure was not associated with increased 1-year mortality risk in our study, whereas glucocorticoid treatment proved a risk factor in a dose-dependent manner, which to our knowledge is a novel finding.
Our study has several strengths of which the most important is the nationwide set-up utilising registers with no loss to follow-up. We adjusted for several important confounders in our analysis of the impact of bDMARD treatment, including disease-specific factors such as DAS28-CRP and glucocorticoid treatment.
However, our study also has some important limitations: the number of events in bDMARD analyses was low thereby decreasing the statistical power. Also, we cannot exclude some degree of misclassification in DMARD and glucocorticoid treatment episodes: currently, there is no validation study of medication data from DANBIO, which is also the reason we did not investigate the risk with preoperative pausing versus continued bDMARD treatment. Furthermore, PJI can be difficult to diagnose, and this carries over to the (lack of) registration in arthroplasty and healthcare registers with sometimes less than optimal PPVs, although we have no reason to believe this problem should be particularly biased in one patient group over another. Lastly, we cannot rule out that the increased risk of death and PJIs among patients with RA and particularly glucocorticoid-treated patients with RA could be due to residual confounding although we believe the risk of this being the case is minimal.
In conclusion, patients with RA had increased 1-year risks of PJI and death and decreased risk of long-term revision compared with patients with OA following THA/TKA. bDMARD treatment within 90 days preceding surgery was not associated with statistically significant increased risk of revision, PJI nor death. On the contrary, glucocorticoid treatment was a strong risk factor for 1-year mortality and increased DAS28 was independently associated with 1-year PJI risk.
Acknowledgments
We would like to thank all the departments of rheumatology in Denmark for reporting to the DANBIO Register, and equally so all departments of orthopaedic surgery for their devotion to registering in DHR and DKR.
References
Footnotes
Handling editor Josef S Smolen
Contributors RLC and LD had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: RLC, LD, AO, LEK and SO. Acquisition, analysis and interpretation of data: all authors. Drafting of the manuscript: RLC. Critical revision of the manuscript for important intellectual content: all authors. Statistical analysis: RLC and LD. Study supervision: LD, AO, LEK and SO.
Funding The study was funded by The Danish Rheumatism Association, The Bjarne Jensen Foundation, the Oak Foundation and the Danish Council for Independent Research.
Competing interests PH has received speaking fees from Celgene and UCB outside the present work. LEK has received fees for speaking and/or consultancy from Pfizer, AbbVie, Amgen, UCB, Celgene, BMS, Biogen, Sanofi, MSD, Novartis, Eli Lilly and Janssen Pharmaceuticals. LD has received speaking fees from MSD and UCB outside the present work.
Ethics approval According to Danish legislation, publication of data from registers and databases does not require patient consent or ethics approval. Approval was given by the Danish Data Protection Agency (GEH-2014-043, I-Suite: 03166).
Provenance and peer review Not commissioned; externally peer reviewed.