Abstract
Background
Patients with rheumatoid arthritis (RA) are at increased risk of infections after arthroplasty, yet risks of specific biologic medications are unknown.
Objectives
To compare the risk of post-operative infection between different biologics and to evaluate the risk associated with glucocorticoids
Design
Retrospective cohort
Setting
Medicare and Truven MarketScan® administrative data January 2006-September 2015
Patients
Adults with RA undergoing elective inpatient primary or revision total knee or hip arthroplasty with recent infusion or prescription for abatacept, adalimumab, etanercept, infliximab, rituximab, or tocilizumab before surgery.
Measurements
Propensity adjusted analyses using inverse probability weights evaluated comparative risk of 30-day hospitalized infection and 1-year prosthetic joint infection (PJI) between biologics or glucocorticoid doses. Secondary analyses evaluated non-urinary hospitalized infection and 30-day readmission.
Results
We identified 10,923 surgeries among 9,911 biologic-treated patients. Outcomes were similar in patients treated with different biologics. Compared to an 8.16% risk of hospitalized infection and 2.14% 1-year cumulative incidence of PJI with abatacept, predicted risk from propensity-weighted models ranged from 6.87% (5.30–8.90) with adalimumab to 8.90% (5.70–13.52) with rituximab for hospitalized infection and from 0.35% (0.11–1.12) with rituximab to 3.67% (1.69–7.88) with tocilizumab for PJI. Glucocorticoids were associated with a dose-dependent increase in post-operative risk for all outcomes. Compared to patients not receiving glucocorticoids, predicted risk from propensity-weighted models for >10mg/day glucocorticoids was 13.25% (9.72–17.81) vs. 6.78% for hospitalized infection and 3.83% (2.13–6.87) vs. 2.09% for 1-year cumulative incidence of PJI.
Limitations
Potential for residual confounding, small sample size for rituximab and tocilizumab
Conclusion
Risk of hospitalized infection, prosthetic joint infection, and readmission after arthroplasty was similar across biologics. In contrast, glucocorticoid use, especially > 10mg/day, was associated with greater risk of adverse outcomes.
Primary funding sources
Rheumatology Research Foundation, National Institutes of Health, and Bristol-Myers Squibb
INTRODUCTION
Biologic disease modifying anti-rheumatic drugs (bDMARDs) selectively target the immune system and are increasingly used in the treatment of rheumatoid arthritis (RA). These medications increase the risk of serious infection compared to conventional synthetic DMARDs (csDMARDs) such as methotrexate(1–4). Although bDMARDs are often discussed as a single class, side effect profiles including infection risk may differ between medications because of different mechanisms of actions or dosing. Randomized trials have infrequently compared different biologics and are not powered to assess serious infections(5). In some observational studies, infliximab and tocilizumab have been associated with greater infection risk, and abatacept and etanercept with lower risk(6–10). Studies in surgical patients are lacking.
Infection risk is particularly important in patients undergoing major surgery. Patients with RA frequently undergo orthopedic surgery, especially total hip and knee arthroplasty, and are at increased risk of post-operative complications(11,12). Joint arthroplasty can be complicated not only by immediate post-operative infections but also by prosthetic joint infections (PJI) which can occur later after surgery and carry substantial morbidity(13). Understanding the risk of post-operative infection with different immunosuppressive therapies is important to optimizing perioperative management.
We used two large administrative datasets to compare the risk of post-operative infections and readmission in patients with RA exposed to different biologic therapies before total hip or knee arthroplasty. Additionally, we evaluated the risk of post-operative infection associated with glucocorticoids.
METHODS
This retrospective cohort study evaluated patients with RA undergoing hip or knee arthroplasty using Medicare claims and Truven MarketScan® databases from January 1st, 2006 to September 30th, 2015. Medicare is a public health plan covering more than 90% of U.S. adults ≥age 65(14). Younger individuals with disabilities (e.g. RA) may also be covered. MarketScan is a U.S. database including inpatient, outpatient, and pharmacy data from large employers, health plans, and government and public organizations for >143 million individuals(15). The study was approved by the University of Pennsylvania institutional review board.
Cohort identification
Included patients were ≥18 years-old with RA, based on 2 physician ICD-9 (International Classification of Diseases, 9th edition) codes (714.xx) at least 7 days apart and use of a DMARD(16), who underwent inpatient elective primary or revision hip or knee arthroplasty from 2007-August 31st 2015 based on ICD-9 and Current Procedural Terminology codes for primary surgeries and Current Procedural Terminology codes for revisions according to validated algorithms (Appendix Table 1)(17–20). The index date was the day of surgery, occurring within the index hospitalization (study design in Appendix Figure 1).
Included patients had documented infusions or filled prescriptions for tumor necrosis factor (TNF) antagonists (infliximab, adalimumab, or etanercept), abatacept, or tocilizumab ≤8 weeks or received rituximab ≤16 weeks before surgery. We required ≥3 infusions or prescriptions in the past year (2 for rituximab) to identify patients on chronic therapy. We also required a 1-year baseline period before the index date with continuous enrollment in Medicare Parts A, B, and D or MarketScan.
We excluded patients with evidence of pre-existing infection or non-elective surgery: diagnosis or treatment for native or prosthetic joint infection in the year prior to surgery; diagnoses of femur fracture, bone or metastatic cancer from the index hospitalization; admission through the emergency department or transfer from another acute care hospital; surgery after hospital day 3; major surgery in the previous 6 months (details in Appendix Table 1). Patients with admission status “emergent” were excluded (only available in Medicare). Patients with diagnoses of inflammatory bowel disease, psoriatic arthritis, ankylosing spondylitis, HIV, or active malignancy in the past year were excluded as these could impact bDMARD use. Patients could contribute multiple surgeries if >6 months apart. To avoid double counting patients, we excluded patients in MarketScan with derived birthdate within 31 days of any patient in Medicare with the same admission date and biologic exposure (Appendix Table 2).
Outcomes
Primary outcomes were 1) hospitalized infection within 30 days, and 2) prosthetic joint infection (PJI) within 1 year after surgery. Hospitalized infections were identified based on ICD-9 diagnosis codes from any position of the discharge diagnoses (positive predictive value > 80%)(1,21), including the index hospitalization and any subsequent acute care hospitalizations with an admission date within 30 days of surgery. PJI was identified based on inpatient or outpatient physician diagnosis of PJI (ICD-9 code 996.66) within 1 year after the index hospitalization (excluding patients with PJI diagnoses from the index hospitalization, as pre-existing infection may have been the indication for surgery)(22,23). Sensitivity analyses assessed more stringent definitions of PJI: 1) requiring an inpatient diagnosis or 2) requiring an accompanying procedure code within 30 days of PJI diagnosis (i.e. arthrotomy, prosthesis removal, central venous catheter insertion, spacer, or revision surgery) (Appendix Table 1)(18).
Secondary outcomes included an alternate 30-day hospitalized infection outcome excluding urinary tract infections (which may represent more minor or incidentally discovered infections) and 30-day readmission among patients with discharge to home, home health care, acute rehabilitation, or a skilled nursing facility (excluding hospitalizations within 1 day of discharge or with primary diagnosis indicating rehabilitation)(24).
Exploratory outcomes included prolonged length of stay (a surrogate for post-operative complications)(25,26) and time to revision surgery among patients undergoing primary knee or hip arthroplasty(27). Length of stay >90th percentile by year and procedure type was considered prolonged (empirically derived as >4 days for primary surgeries 2011–2015 and >5 days for all other surgeries)(26). We also examined wound complications(28), specific infections (pneumonia, septicemia/bacteremia, urinary infection) (Appendix Table 1), and 30-day mortality (Medicare only).
Covariates
Covariates measured during the baseline period included demographics, comorbid conditions, an adaption of the Charlson comorbidity index(29), healthcare utilization (outpatient visits, emergency department visits, hospitalizations), previous hospitalized infection, and medication use within 90 days including non-steroidal anti-inflammatories, opioids, methotrexate, other csDMARDs, glucocorticoids, and antibiotics. Average glucocorticoid dose in the 90 days prior to surgery was calculated based on oral prescriptions for prednisone, prednisolone, and methylprednisolone, using prescribed dose in prednisone equivalents and days supply to determine each daily dose and truncating prescriptions if a new prescription was filled before the prescription end date. Disability status, skilled nursing facility residence, and quintiles of median household income based on zip code from the American Community Survey 2009–2013(30) were available in Medicare. Surgeon and hospital volume (Medicare only) were estimated among 55,812 hip or knee arthroplasties in patients with any bDMARD or methotrexate use <6 months before surgery.
Statistical analysis
Associations between pre-operative biologic exposure or glucocorticoid exposure and post-operative outcomes were assessed using logistic regression for binary outcomes and competing risk regression (Fine-Gray models) for PJI, censoring at 1 year after surgery, end of enrollment in Medicare or MarketScan, subsequent hip or knee arthroplasty, or September 30th, 2015, with death as a competing risk(31). Propensity score-derived inverse probability weights were used to balance confounders across treatment groups (see below). Predicted risk and risk differences were estimated from logistic regression models at the means of all covariates. Predicted 1-year cumulative incidence and differences in 1-year cumulative incidence were calculated using the reference cumulative incidence function and linear predictions from Fine-Gray models(32). Abatacept and no glucocorticoid use were the reference groups for the analyses of biologics and glucocorticoid use, respectively. Cause-specific hazard models with death treated as censoring provided nearly identical results.
Medicare analyses were clustered by hospital (not available in MarketScan), while MarketScan analyses were clustered by patient(33). The large number of hospitals (2,228), with 94% contributing ≤10 surgeries, prevented analyses within hospitals (Appendix Table 3). Analyses were performed separately in Medicare and MarketScan, with odds ratios (OR) and subdistribution hazard ratios (HR) combined using inverse-variance weighed fixed effects meta-analysis, assessing heterogeneity visually and using I2 values(34). Combined OR and HR were used to calculate combined-predicted risk and 1-year cumulative incidence for each exposure group, with reference risk and reference 1-year cumulative incidence based on crude results in the reference group from pooled Medicare and MarketScan data. Small sample size prevented covariate balance in rituximab and tocilizumab exposure groups in MarketScan, and so these patients were excluded from adjusted analyses, with all results coming from Medicare. Because tocilizumab was only available in the second half of the study, propensity scores were recalculated and a separate analysis was performed restricted to surgeries 2011–2015 to evaluate tocilizumab in Medicare.
Generalized propensity scores based on the probability of being in each treatment group were generated separately in each dataset using multinomial logistic regression models, including a squared term for age to account for non-linearity(35–37) (variables included in propensity score models in Appendix Tables 4–5). Propensity scores were used to create stabilized inverse probability treatment weights(36,38,39) truncated at the 1st and 99th percentile(40). The results of weighted analyses estimate the average treatment effect (ATE) among the general population of patients with RA treated with biologics – the average effect, at the population level, of moving an entire population from treatment with one biologic (here abatacept) to another biologic. Balance of covariates across treatment categories was assessed compared to the reference group abatacept with standardized mean difference ≤0.1 indicating good balance (Appendix Figures 2–4). Unbalanced covariates were added as covariates to weighted models (described in tables and figures).
Similar inverse probability weighted analyses evaluated associations between glucocorticoid dose (none, ≤5mg, 5–10mg, > 10mg) and outcomes in the same biologic-treated cohorts. Propensity score models included the same covariates as well as current and previous biologic treatment.
Sensitivity analyses
We evaluated two alternative PJI definitions as noted above. Additional sensitivity analyses included restricting to 2011–2015, to primary knee/hip replacement, or requiring biologic exposure <4 weeks before surgery (except for rituximab). Because TNF antagonists are typically first-line biologic treatment, previous biologic use could not be balanced across exposure categories and was excluded from propensity score models. Instead, a sensitivity analysis was performed restricting TNF antagonist exposed patients to those with previous biologic use. Propensity scores were recalculated for each sensitivity analysis. We assessed for different effects among methotrexate users and non-users using interaction terms with no evidence of effect modification.
All statistical tests were 2-sided, with statistical significance fixed at P = 0.050. The dataset was created with SAS 9.4 (SAS Institute) and analysis performed using Stata 15.1 (StataCorp) (commands used in Appendix Table 6). The protocol was approved by the University of Alabama at Birmingham (UAB) institutional review board.
Role of the funding source
Funding was provided by the Rheumatology Research Foundation, Patient Centered Outcomes Research Institute, and Bristol-Myers Squibb. Bristol-Myers Squibb authors made non-binding comments on the manuscript draft. Final manuscript wording and decision to submit were made solely by University of Pennsylvania and UAB investigators.
RESULTS
We identified 19,610 primary or revision hip or knee arthroplasties in Medicare and 6,446 in MarketScan among patients with RA with recent biologic exposure. Applying exclusion criteria and excluding 98 surgeries from the MarketScan cohort that were possible duplicates in the two datasets left 7,929 surgeries among 7,138 patients in Medicare and 2,994 surgeries among 2,773 patients in MarketScan (Appendix Table 2), with 951/9,911 (9.6%) patients contributing multiple surgeries.
Patient characteristics are shown in Table 1 and Appendix Tables 3–4. Mean age was 65.1, 83.0% of patients were female, and 89.5% of surgeries were primary surgeries. In addition to the bDMARD, 43.0% of patients received glucocorticoids and 45.6% received methotrexate within 90 days before surgery. Average glucocorticoid dose was >10mg/day in 423 (3.9%) with median dose 12.1mg/day (interquartile range 10.3–15.2) in these patients.
Table 1:
Select cohort characteristics by biologic exposure
| Abatacept N = 1764 |
Adalimumab N = 2002 |
Etanercept N = 2954 |
Infliximab N = 3391 |
Rituximab N = 423 |
Tocilizumab N = 389 |
Total N = 10923 |
|
|---|---|---|---|---|---|---|---|
| Medicare Cohort | 1369 (77.6%) | 1284 (64.1%) | 1836 (62.2%) | 2798 (82.5%) | 337 (79.7%) | 305 (78.4%) | 7929 (72.6%) |
| Female | 1517 (86.0%) | 1640 (81.9%) | 2410 (81.6%) | 2799 (82.5%) | 363 (85.8%) | 335 (86.1%) | 9064 (83.0%) |
| Age | 66.4 (10.4) | 62.1 (11.1) | 62.7 (11.0) | 68.3 (10.0) | 64.1 (11.2) | 66.1 (11.5) | 65.1 (11.0) |
| Year 2011–2015 | 1186 (67.2%) | 1096 (54.7%) | 1637 (55.4%) | 1807 (53.3%) | 263 (62.2%) | 389 (100.0%) | 6378 (58.4%) |
| Surgery Type | |||||||
| Primary knee | 1136 (64.4%) | 1298 (64.8%) | 1892 (64.0%) | 2238 (66.0%) | 260 (61.5%) | 253 (65.0%) | 7077 (64.8%) |
| Primary hip | 457 (25.9%) | 485 (24.2%) | 706 (23.9%) | 831 (24.5%) | 115 (27.2%) | 105 (27.0%) | 2699 (24.7%) |
| Revision knee | 92 (5.2%) | 118 (5.9%) | 179 (6.1%) | 212 (6.3%) | 29 (6.9%) | 21 (5.4%) | 651 (6.0%) |
| Revision hip | 79 (4.5%) | 101 (5.0%) | 177 (6.0%) | 110 (3.2%) | 19 (4.5%) | 10 (2.6%) | 496 (4.5%) |
| Glucocorticoid average dose past 90 days | |||||||
| None | 898 (50.9%) | 1170 (58.4%) | 1746 (59.1%) | 2013 (59.4%) | 201 (47.5%) | 195 (50.1%) | 6223 (57.0%) |
| ≤ 5mg | 510 (28.9%) | 531 (26.5%) | 771 (26.1%) | 842 (24.8%) | 113 (26.7%) | 110 (28.3%) | 2877 (26.3%) |
| > 5 to 10mg | 273 (15.5%) | 235 (11.7%) | 324 (11.0%) | 427 (12.6%) | 78 (18.4%) | 63 (16.2%) | 1400 (12.8%) |
| > 10mg | 83 (4.7%) | 66 (3.3%) | 113 (3.8%) | 109 (3.2%) | 31 (7.3%) | 21 (5.4%) | 423 (3.9%) |
| Current biologic course duration (years) | 1.3 [0.7–2.4] | 1.5 [0.8,2.7] | 1.7 [0.9,3.1] | 2.1 [1.2–3.9] | 0.9 [0.4,1.9] | 0.9 [0.5,1.5] | 1.6 [0.9,3.0] |
| Prior biologics | |||||||
| 0 | 770 (43.7%) | 1556 (77.7%) | 2498 (84.6%) | 2922 (86.2%) | 158 (37.4%) | 76 (19.5%) | 7980 (73.1%) |
| 1 | 712 (40.4%) | 392 (19.6%) | 367 (12.4%) | 371 (10.9%) | 155 (36.6%) | 130 (33.4%) | 2127 (19.5%) |
| ≥ 2 | 282 (16.0%) | 54 (2.7%) | 89 (3.0%) | 98 (2.9%) | 110 (26.0%) | 183 (47.0%) | 816 (7.5%) |
| Methotrexate past 90 days | 687 (38.9%) | 914 (45.7%) | 1232 (41.7%) | 1877 (55.4%) | 153 (36.2%) | 119 (30.6%) | 4982 (45.6%) |
| HCQ/SSA/LEF past 90 days | 488 (27.7%) | 464 (23.2%) | 657 (22.2%) | 657 (19.4%) | 120 (28.4%) | 73 (18.8%) | 2459 (22.5%) |
| Charlson Score | 0 [0,3] | 0 [0,2] | 0 [0,2] | 1 [0,3] | 1 [0,3] | 1 [0,3] | 1 [0,2] |
| Diabetes | 315 (17.9%) | 313 (15.6%) | 429 (14.5%) | 574 (16.9%) | 73 (17.3%) | 75 (19.3%) | 1779 (16.3%) |
| Hypertension | 944 (53.5%) | 971 (48.5%) | 1393 (47.2%) | 1841 (54.3%) | 221 (52.2%) | 223 (57.3%) | 5593 (51.2%) |
| COPD/Asthma | 246 (13.9%) | 227 (11.3%) | 384 (13.0%) | 406 (12.0%) | 73 (17.3%) | 60 (15.4%) | 1396 (12.8%) |
| Chronic kidney disease | 106 (6.0%) | 87 (4.3%) | 123 (4.2%) | 190 (5.6%) | 23 (5.4%) | 29 (7.5%) | 558 (5.1%) |
| Hospitalizations past year | |||||||
| 0 | 1321 (74.9%) | 1575 (78.7%) | 2322 (78.6%) | 2625 (77.4%) | 299 (70.7%) | 289 (74.3%) | 8431 (77.2%) |
| 1–2 | 307 (17.4%) | 284 (14.2%) | 407 (13.8%) | 579 (17.1%) | 87 (20.6%) | 70 (18.0%) | 1734 (15.9%) |
| ≥ 3 | 136 (7.7%) | 143 (7.1%) | 225 (7.6%) | 187 (5.5%) | 37 (8.7%) | 30 (7.7%) | 758 (6.9%) |
| Hospitalized infection past year | 116 (6.6%) | 107 (5.3%) | 155 (5.2%) | 183 (5.4%) | 31 (7.3%) | 28 (7.2%) | 620 (5.7%) |
| ED visits past year | |||||||
| 0 | 1162 (65.9%) | 1329 (66.4%) | 2024 (68.5%) | 2269 (66.9%) | 270 (63.8%) | 238 (61.2%) | 7292 (66.8%) |
| 1 | 384 (21.8%) | 404 (20.2%) | 533 (18.0%) | 719 (21.2%) | 97 (22.9%) | 84 (21.6%) | 2221 (20.3%) |
| 2–3 | 166 (9.4%) | 203 (10.1%) | 296 (10.0%) | 326 (9.6%) | 35 (8.3%) | 51 (13.1%) | 1077 (9.9%) |
| > 3 | 52 (2.9%) | 66 (3.3%) | 101 (3.4%) | 77 (2.3%) | 21 (5.0%) | 16 (4.1%) | 333 (3.0%) |
N (%), median [interquartile range], or mean (standard deviation). Select characteristics available in both Medicare and MarketScan are shown. Full list of characteristics included in the propensity scores in Medicare and MarketScan available in Appendix Tables 3–4. 157/1764 (8.9%) of abatacept and 14/389 (3.6%) of tocilizumab is subcutaneous. HCQ = hydroxychloroquine, SSA = sulfasalazine, LEF = leflunomide, COPD = chronic obstructive pulmonary disease
We identified hospitalized infection within 30 days after 717 (9.0%) and 140 (4.7%) surgeries in Medicare and MarketScan, respectively (Table 2). The most common infections were urinary (449, 4.1%), skin and soft tissue (123, 1.1%), and pneumonia (94, 0.9%) (Appendix Table 7). 436 (50.9%) of hospitalized infections were from the initial hospitalization during which surgery was performed. PJI occurred within 1 year after 192 (2.6% 1-year cumulative incidence) and 53 (2.0% 1-year cumulative incidence) in Medicare and MarketScan, with 157 (64%) of PJI occurring within 90 days of surgery. Secondary outcomes of non-urinary hospitalized infection occurred after 479 (6.0%) and 89 (3.0%) of surgeries, and of 30-day readmission occurred after 465/7554 (6.2%) and 68/2733 (2.5%) of surgeries in Medicare and MarketScan, respectively. Rates of primary and secondary outcomes were similar across biologic treatment groups (Table 2).
Table 2:
Crude frequency of outcomes by biologic and glucocorticoid exposure in Medicare and MarketScan
| N | Person-years (for PJI) | Hospitalized infection, n (%) | Prosthetic joint infection, n (incidence) | Non-urinary hospitalized infection, n (%) | 30-day readmission, n (%) | |
|---|---|---|---|---|---|---|
| Biologics | ||||||
| Medicare | ||||||
| Abatacept | 1369 | 1146 | 126 (9.2%) | 30 (2.4) | 88 (6.4%) | 92/1296 (7.1%) |
| Adalimumab | 1284 | 1066 | 108 (8.4%) | 28 (2.3) | 72 (5.6%) | 78/1225 (6.4%) |
| Etanercept | 1836 | 1562 | 173 (9.4%) | 55 (3.2) | 118 (6.4%) | 112/1767 (6.3%) |
| Infliximab | 2798 | 2434 | 249 (8.9%) | 64 (2.4) | 161 (5.8%) | 143/2673 (5.4%) |
| Rituximab | 337 | 285 | 33 (9.8%) | 3 (0.9) | 21 (6.2%) | 21/314 (6.7%) |
| Tocilizumab | 305 | 229 | 28 (9.2%) | 12 (4.3) | 19 (6.2%) | 19/279 (6.8%) |
| Total | 7929 | 6723 | 717 (9.0%) | 192 (2.6) | 479 (6.0%) | 465/7554 (6.2%) |
| MarketScan | ||||||
| Abatacept | 395 | 308 | 18 (4.6%) | 4 (1.1) | 10 (2.5%) | 8/365 (2.2%) |
| Adalimumab | 718 | 555 | 25 (3.5%) | 12 (1.8) | 16 (2.2%) | 17/669 (2.5%) |
| Etanercept | 1118 | 895 | 59 (5.3%) | 15 (1.5) | 37 (3.3%) | 24/1018 (2.4%) |
| Infliximab | 593 | 460 | 33 (5.6%) | 14 (2.7) | 21 (3.5%) | 11/527 (2.1%) |
| Rituximab | 86 | 63 | 3 (3.5%) | 4 (5.9) | 3 (3.5%) | 6/79 (7.6%) |
| Tocilizumab | 84 | 59 | 2 (2.4%) | 4 (5.4) | 2 (2.4%) | 2/75 (2.7%) |
| Total | 2994 | 2340 | 140 (4.7%) | 53 (2.0) | 89 (3.0%) | 68/2733 (2.5%) |
| Glucocorticoids | ||||||
| Medicare | ||||||
| None | 4389 | 3769 | 342 (7.8%) | 83 (2.3) | 226 (5.2%) | 212/4168 (5.1%) |
| ≤5mg/day | 2133 | 1797 | 205 (9.6%) | 54 (2.7) | 136 (6.4%) | 145/2038 (7.1%) |
| >5–10 mg/day | 1102 | 916 | 121 (11.0%) | 31 (3.0) | 84 (7.6%) | 82/1055 (7.8%) |
| >10mg/day | 305 | 241 | 49 (16.1%) | 14 (4.9) | 33 (10.8%) | 26/293 (8.9%) |
| MarketScan | ||||||
| None | 1834 | 1440 | 80 (4.4%) | 26 (1.6) | 44 (2.4%) | 35/1677 (2.1%) |
| ≤5mg/day | 744 | 591 | 38 (5.1%) | 16 (2.4) | 28 (3.8%) | 15/679 (2.2%) |
| >5–10 mg/day | 298 | 218 | 17 (5.7%) | 10 (3.9) | 14 (4.7%) | 12/269 (4.5%) |
| >10mg/day | 118 | 92 | 5 (4.2%) | 1 (0.8) | 3 (2.5%) | 6/108 (5.6%) |
Readmission analyses restricted to patients discharged to home, home health, skilled nursing facility, or inpatient rehabilitation facility (denominators shown). Glucocorticoid dose is the average glucocorticoid dose in prednisone equivalents in the 90 days prior to surgery based on prescriptions for oral glucocorticoids. PJI = prosthetic joint infection. Incidence = 1-year cumulative incidence in percent
There was no significant difference in the propensity-weighted risk of hospitalized infection, PJI, non-urinary hospitalized infection, or 30-day readmission across biologic treatment groups in stratified (Table 3, Appendix Table 8) or meta-analysis combined (Figure 1, Appendix Figure 5) analyses except for PJI with rituximab. Rituximab was associated with lower incidence of PJI vs. abatacept in Medicare [difference in 1-year cumulative incidence −2.14% (−2.44,−1.22)] (Table 3). Small sample size prevented evaluation of rituximab in MarketScan, and so rituximab results come from Medicare only, with HR 0.16(0.05,0.52) (Appendix Table 8, Appendix Figure 5) resulting in predicted 1-year cumulative incidence with rituximab of 0.35% (0.11,1.12) vs. 2.14% with abatacept (Figure 1). Results were similar in separate analyses restricted to 2011–2015 that included tocilizumab (Appendix Tables 10–12, Appendix Figures 6–7). Sensitivity analyses excluding TNF inhibitors used as first-line therapy, evaluating only primary surgeries, requiring biologic exposure within 4 weeks, or using alternative PJI definitions demonstrated similar results (Appendix Tables 13–14).
Table 3:
Predicted risk and risk differences from inverse probability weighted biologic and glucocorticoid analyses in Medicare and MarketScan
| Hospitalized infection | Prosthetic joint infection | Non-urinary hospitalized infection | 30-day readmission | |||||
|---|---|---|---|---|---|---|---|---|
| Predicted risk % (95% CI) | Risk difference % (95% CI) | Predicted 1-year cumulative incidence % (95% CI) | Difference in 1-year cumulative incidence % (95% CI) | Predicted risk % (95% CI) | Risk difference % (95% CI) | Predicted risk % (95% CI) | Risk difference % (95% CI) | |
| Biologics | ||||||||
| Medicare | ||||||||
| Abatacept (n=1369) | 9.32 (7.52, 11.12) | Reference | 2.57 (N/A) | Reference | 6.22 (4.78, 7.66) | Reference | 6.55 (5.02, 8.09) | Reference |
| Adalimumab (n=1284) | 8.16 (6.44, 9.88) | −1.16 (−3.59, 1.27) | 2.12 (1.20, 3.73) | −0.45 (−1.37, 1.16) | 5.16 (3.80, 6.51) | −1.06 (−3.03, 0.90) | 5.87 (4.48, 7.27) | −0.68 (−2.79, 1.44) |
| Etanercept (n=1836) | 9.57 (8.00, 11.14) | 0.25 (−2.12, 2.62) | 2.59 (1.62, 4.12) | 0.02 (−0.95, 1.55) | 6.70 (5.35, 8.05) | 0.48 (−1.49, 2.44) | 6.75 (5.36, 8.14) | 0.20 (−1.87, 2.26) |
| Infliximab (n=2798) | 9.02 (7.79, 10.25) | −0.30 (−2.43, 1.83) | 2.81 (1.74, 4.53) | 0.24 (−0.83, 1.96) | 5.98 (4.92, 7.04) | −0.24 (−2.03, 1.55) | 5.63 (4.59, 6.66) | −0.92 (−2.80, 0.96) |
| Rituximab (n=337) | 10.13 (6.29, 13.98) | 0.81 (−3.43, 5.06) | 0.43 (0.14, 1.35) | −2.14 (−2.44, −1.22)* | 6.22 (3.23, 9.21) | 0.00 (−3.31, 3.32) | 6.71 (3.45, 9.98) | 0.16 (−3.39, 3.71) |
| Tocilizumab (n=305)† | 7.94 (4.55, 11.34) | −0.29 (−4.18, 3.59) | 3.90 (1.72, 8.76) | 1.64 (−0.54, 6.51) | 4.64 (2.34, 6.94) | −1.49 (−4.22, 1.25) | 5.18 (2.65, 7.71) | −1.04 (−4.08, 1.99) |
| MarketScan‡ | ||||||||
| Abatacept (n=395) | 4.47 (2.18, 6.76) | Reference | 1.26 (N/A) | Reference | 2.36 (0.72, 4.01) | Reference | 1.45 (0.37, 2.52) | Reference |
| Adalimumab (n=718) | 3.11 (1.85, 4.36) | −1.36 (−3.98, 1.25) | 1.64 (0.45, 5.95) | 0.38 (−0.81, 4.69) | 2.00 (0.98, 3.02) | −0.36 (−2.30, 1.57) | 2.45 (1.20, 3.70) | 1.00 (−0.65, 2.66) |
| Etanercept (n=1118) | 5.22 (3.86, 6.57) | 0.75 (−1.92, 3.42) | 1.47 (0.41, 5.27) | 0.21 (−0.85, 4.01) | 3.22 (2.16, 4.28) | 0.86 (−1.11, 2.82) | 2.63 (1.52, 3.74) | 1.19 (−0.35, 2.73) |
| Infliximab (n=593) | 5.14 (3.30, 6.98) | 0.67 (−2.27, 3.61) | 2.46 (0.69, 8.68) | 1.20 (−0.57, 7.42) | 3.40 (1.87, 4.94) | 1.04 (−1.21, 3.28) | 2.00 (0.68, 3.31) | 0.55 (−1.15, 2.25) |
| Rituximab (n=86)§ | N/A | N/A | N/A | N/A | N/A | N/A | N/A | N/A |
| Tocilizumab (n=84)§ | N/A | N/A | N/A | N/A | N/A | N/A | N/A | N/A |
| Glucocorticoids | ||||||||
| Medicare|| | ||||||||
| None (n=4389) | 8.10 (7.21, 8.99) | Reference | 2.34 (N/A) | Reference | 5.38 (4.66, 6.10) | Reference | 5.40 (4.66, 6.14) | Reference |
| ≤5mg/day (n=2133) | 9.44 (8.11, 10.76) | 1.33 (−0.20, 2.87) | 2.63 (1.64, 4.22) | 0.29 (−0.70, 1.88) | 6.21 (5.16, 7.27) | 0.83 (−0.40, 2.07) | 6.93 (5.85, 8.01) | 1.53 (0.21, 2.85)* |
| >5–10 mg/day (n=1102) | 10.14 (8.33, 11.95) | 2.04 (0.03, 4.04)* | 2.70 (1.55, 4.72) | 0.36 (−0.80, 2.37) | 7.08 (5.53, 8.63) | 1.70 (−0.01, 3.41) | 6.84 (5.32, 8.35) | 1.43 (−0.25, 3.12) |
| >10mg/day (n=305) | 17.23 (12.28, 22.18) | 9.12 (4.09, 14.16)* | 4.82 (2.40, 9.56) | 2.48 (0.06, 7.22)* | 12.61 (8.06, 17.17) | 7.23 (2.60, 11.86)* | 8.40 (4.86, 11.94) | 3.00 (−0.65, 6.64) |
| MarketScan¶ | ||||||||
| None (n=1834) | 4.48 (3.50, 5.45) | Reference | 1.63 (N/A) | Reference | 2.35 (1.66, 3.04) | Reference | 2.07 (1.39, 2.76) | Reference |
| ≤5mg/day (n=744) | 5.23 (3.55, 6.91) | 0.75 (−1.18, 2.69) | 2.35 (0.63, 9.24) | 0.72 (−1.00, 7.61) | 4.10 (2.57, 5.64) | 1.75 (0.07, 3.44)* | 2.21 (1.04, 3.38) | 0.14 (−1.22, 1.50) |
| >5–10 mg/day (n=298) | 7.02 (3.39, 10.64) | 2.54 (−1.22, 6.30) | 3.38 (0.82, 14.03) | 1.75 (−0.81, 12.40) | 5.79 (2.45, 9.13) | 3.44 (0.03, 6.85)* | 5.23 (2.00, 8.45) | 3.16 (−0.14, 6.45) |
| >10mg/day (n=118) | 3.61 (0.06, 7.15) | −0.87 (−4.56, 2.81) | 1.26 (0.12, 12.10) | −0.37 (−1.51, 10.47) | 2.08 (−0.54, 4.70) | −0.27 (−2.99, 2.46) | 3.82 (0.05, 7.58) | 1.75 (−2.08, 5.58) |
Predicted risk and risk differences for binary outcomes from logistic regression models with inverse probability weighting. Predicted 1-year cumulative incidence of prosthetic joint infection from competing risk regression (Fine and Gray) with death as a competing risk – no confidence interval available for the reference group, which represents the reference cumulative incidence at one year. Odds ratios and subdistribution hazards ratios from these models shown in Appendix Table 7. Imbalanced covariates added to weighted models as noted. Glucocorticoid dose is the average glucocorticoid dose in prednisone equivalents in the 90 days prior to surgery based on prescriptions for oral glucocorticoids
risk difference p < 0.05
Tocilizumab results from separate analysis restricted to 2011–2015; covariates included in 2011–2015 weighted models: methotrexate past 90 days for hospitalized infections and prosthetic joint infection; year for readmission
Covariates included in weighted models: outpatient visits
Rituximab and tocilizumab excluded from MarketScan analyses because of inadequate sample size.
Covariates included in weighted models: opioid use and number outpatient visits for all outcomes
Covariates included in weighted models: current biologic, age, age2, hydroxychloroquine/ leflunomide/ sulfasalazine use, opioid use, outpatient visits, emergency department visits, and antibiotic use past 90 days for all outcomes; number previous biologics and osteonecrosis for hospitalized infections and prosthetic joint infection; region, urban, hypertension, chronic kidney disease for readmission
Figure 1:
Predicted risk and cumulative incidence from inverse probability weighted biologic analyses in Medicare and MarketScan combined using inverse-variance weighted fixed effects meta-analysis.
Predicted risk or 1-year cumulative incidence for each exposure group were calculated using meta-analysis combined odds ratios and subdistribution hazard ratios (shown in Appendix Figure 5) and the reference risk or cumulative incidence (crude results from the reference group using pooled Medicare and MarketScan data). Confidence intervals that do not include the reference risk or cumulative incidence represent statistically significant results (p < 0.05). I2 values for heterogeneity were all <50% (see Appendix Table 9). Rituximab and tocilizumab results are from Medicare only with tocilizumab results from separate analyses restricted to 2011–2015.
Glucocorticoid exposure was associated with a dose-dependent increase in the risk of hospitalized infection, PJI, non-urinary hospitalized infection, and readmission in both datasets (Table 3, Appendix Table 8). Only 118 patients received glucocorticoids >10mg/day in MarketScan. In meta-analysis combined propensity-weighted analyses, risk of hospitalized infection was significantly greater with glucocorticoids 5–10mg/day [OR 1.32(1.06,0.64)] and >10mg/day [OR 2.10(1.48,2.98)] (Appendix Figure 8), resulting in predicted risk of 8.76% (7.16,10.66) for 5–10mg/day and 13.25% (9.72,17.81) for >10mg/day vs. 6.78% without glucocorticoids (Figure 2). Rates of PJI were numerically greater with glucocorticoids 5–10mg/day [HR 1.36(0.90,2.04)] and significantly greater with >10mg/day [HR 1.86(1.02,3.37)] (Appendix Figure 8), with predicted 1-year cumulative incidence 2.83% (1.88,4.21) for 5–10mg/day and 3.83% (2.13,6.87) for >10mg/day vs. 2.09% without glucocorticoids (Figure 2). Results were similar for non-urinary hospitalized infection and 30-day readmission, with significantly greater risk even at ≤5mg/day (Figure 2, Appendix Figure 8). Predicted risk for >10mg/day vs. risk without glucocorticoids was 9.41% (6.41,13.67) vs. 4.34% for non-urinary hospitalized infection and 6.79% (4.47,10.27) vs. 4.23% for 30-day readmission (Figure 2). Concomitant methotrexate was not associated with greater risk of any of the outcomes (Appendix Table 15).
Figure 2:
Predicted risk and cumulative incidence from inverse probability weighted glucocorticoid analyses in Medicare and MarketScan combined using inverse-variance weighted fixed effects meta-analysis.
Predicted risk or 1-year cumulative incidence for each exposure group were calculated using meta-analysis combined odds ratios and subdistribution hazard ratios (shown in Appendix Figure 8) and the reference risk or cumulative incidence (crude results from the reference group using pooled Medicare and MarketScan data). Confidence intervals that do not include the reference risk or cumulative incidence represent statistically significant results (p < 0.05). I2 values for heterogeneity were all <75% (see Appendix Table 9). Glucocorticoid dose is the average glucocorticoid dose in prednisone equivalents in the 90 days prior to surgery based on prescriptions for oral glucocorticoids.
We found no consistent differences in prolonged length of stay or time to revision surgery across the biologic or glucocorticoid exposure groups (Appendix Table 16).
DISCUSSION
In this study of patients with RA undergoing elective hip or knee arthroplasty, we found that rates of serious post-operative infection and 30-day readmission were similar between biologics. We found a consistent dose-dependent association between glucocorticoids and adverse post-operative outcomes, however, with increased risk even at modest doses of glucocorticoids, suggesting that limiting glucocorticoids should be a focus of perioperative management.
Because infections after surgery, while serious, are relatively rare, large studies are needed to evaluate post-operative risk. After elective joint replacement, rates of pneumonia are approximately 1%, all infections 3–5%, and PJI approximately 0.5–1% (18,20,41), with greater rates of infection reported in patients with RA (12,41,42). We found that rates of infection were greater in Medicare vs. MarketScan, as expected because Medicare includes older and disabled patients.
Rates of 30-day readmission and hospitalized infection after surgery were similar across the biologics studied, regardless of whether or not urinary infections were included in the outcome. While it is not possible to rule out differences in post-operative risk between these therapies, any differences are likely to be small, especially compared to other risk factors for infection.
Results from this study may appear at odds with some studies suggesting a modestly greater risk of infection with infliximab compared to other biologics in the non-operative setting(6–8), yet several factors may explain this apparent discrepancy. First, many other factors contribute to infection risk, including surgical experience and technique, post-operative care, perioperative antibiotic use, and patient characteristics(18,20,42), which may dwarf smaller differences between biologics. Secondly, we specifically evaluated patients who had received multiple infusions or prescriptions to identify patients on stable chronic therapy, as is typical of patients undergoing elective joint replacement. Studies suggest that the risk of infection is greatest in the initial months after biologic initiation(6,43,44), and so differences in infection risk between biologics may not be present among patients on chronic therapy with RA presumably under relatively good control.
We also evaluated PJI within 1 year, as these infections may occur months after surgery. We found no significant difference in the rates of PJI across most of the biologics studied. Because of the small number of patients receiving rituximab and tocilizumab, we were unable to precisely assess PJI risk with these therapies. Rituximab was associated with a significantly lower risk of infection in Medicare, but these results were based on only three outcomes and crude MarketScan results were in the opposite direction. Rates of PJI were greater in tocilizumab treated patients in both datasets, although results were not statistically significant and might be affected by residual confounding. This observation deserves further evaluation given some studies suggesting greater infection risk with tocilizumab (9,10). Notably, tocilizumab was not associated with greater risks of hospitalized infection or readmission.
In contrast, we found that glucocorticoids were strongly associated with a dose-dependent increase in the risk of post-operative infection and readmission, as has been demonstrated in the non-operative setting(3,45,46) and suggested in smaller perioperative studies(47–49). The risk of hospitalized infection and readmission was significantly greater even in patients treated with an average of 5–10mg/day of glucocorticoids, and the risk of PJI was significantly greater at doses >10mg/day. Results were similar in both Medicare and MarketScan, although in MarketScan risk of adverse outcomes was greater with doses 5–10mg/day but not >10mg/day, likely due to the small number of patients receiving >10mg/day in MarketScan leading to imprecise results. Of note, patients receiving methotrexate with their biologic were not at greater risk of post-operative infection, although we were not able to determine whether methotrexate was stopped before surgery.
Clinicians may be focused on the risk of infection with biologic DMARDs, and indeed recent guidelines recommend holding bDMARDs for one dosing interval before joint replacement surgery(50). These guidelines also recommend that glucocorticoids >20mg/day be avoided. The results of this study suggest that post-operative risk may be increased even with lower doses of glucocorticoids (5–10mg/day). As we averaged glucocorticoid dose over 90 days, we cannot distinguish between patients receiving a constant low dose of glucocorticoids over 90 days and patients receiving higher doses of glucocorticoids in the weeks before surgery or in the immediate post-operative period, perhaps because biologic therapy has been interrupted. Recent glucocorticoid exposure likely has a stronger affect on post-operative outcomes, as has been found in the non-operative setting(51). Cumulative glucocorticoid exposure also contributes to risk, however, and the strong dose-dependent relationship between glucocorticoids and outcomes suggests that minimizing glucocorticoid use before surgery is a critical part of perioperative management.
We did not find strong associations between glucocorticoids with our exploratory outcomes of prolonged length of stay or time to revision surgery. Prolonged length of stay has been used as a surrogate for post-operative complications(25), although recent data suggests that it is heavily dependent on patient characteristics and hospital practice and may not perform well in assessing complications after surgery(26,52).
Even with careful balancing of measured confounders with propensity scores residual confounding by indication is possible. To limit this bias we only evaluated biologic treated patients, as patients not receiving biologics are inherently quite different. Additionally, we performed sensitivity analyses excluding patients receiving a TNF inhibitor as first-line therapy, since these patients may have less refractory disease or fewer comorbidities than patients treated with a non-TNF biologic(53). Although direct measures of disease activity were not available, all patients were well enough to undergo elective joint replacement surgery and we included surrogates of disease activity such as non-steroidal, opioid, and glucocorticoid use and measures of healthcare utilization. Outcomes were not confirmed with medical record review, although hospitalized infection definitions have been validated in other settings, 30-day readmission results were similar, and infection rates were similar to expected rates(41,54). Additionally, we were able to demonstrate strong associations between glucocorticoids and these outcomes. We were unable to determine if patients stopped non-infusion biologics before surgery, although we found similar results requiring an infusion or prescription within 4 weeks. Moreover, we have previously shown that interruptions of infliximab before surgery are not associated with significant differences in infection risk(37). Estimating glucocorticoid dose with prescription data is imprecise and we could not capture short-term dose changes, yet we still found strong associations between glucocorticoid dose and outcomes. Numbers were not sufficient in Medicare to conduct within-hospital analyses, and hospital information was not available in MarketScan, but we did include measures of hospital and surgeon volume in Medicare analyses.
Strengths of this study include the large cohort of biologic treated patients from two separate administrative databases, the careful balancing of measured confounders, and the evaluation of multiple post-operative outcomes.
In conclusion, patients with RA treated with different biologic DMARDs before surgery have similar rates of post-operative infection and readmission after total hip and knee arthroplasty. Glucocorticoid use is strongly associated with post-operative infection risk even at modest doses. Minimizing glucocorticoid exposure before surgery should be a primary focus of perioperative medication management.
Supplementary Material
Acknowledgments
Funding/Grant sources: Michael George is supported by the Rheumatology Research Foundation Scientist Development Award and the National Institute of Arthritis and Musculoskeletal and Skin Diseases 1K23AR073931-01. Jeffrey Curtis is supported by the Patient Centered Outcomes Research Institute PPRND-1507-32163. This study also received funding from Bristol-Myers Squibb.
Footnotes
Reproducible Research Statement:
Protocol: posted as a data supplement at Annals website
Statistical Code: Available to interested readers by contacting Dr. George at michael.george@uphs.upenn.edu
Data: Data use is governed by data use agreements. Medicare data is available through the Centers for Medicare & Medicaid Services. MarketScan data is available to license through Truven Health.
References
- 1.Curtis JR, Patkar N, Xie A, Martin C, Allison JJ, Saag M, et al. Risk of serious bacterial infections among rheumatoid arthritis patients exposed to tumor necrosis factor alpha antagonists. Arthritis Rheum. 2007. April;56(4):1125–33. [DOI] [PubMed] [Google Scholar]
- 2.Bongartz T, Sutton AJ, Sweeting MJ, Buchan I, Matteson EL, Montori V. Anti-TNF antibody therapy in rheumatoid arthritis and the risk of serious infections and malignancies: systematic review and meta-analysis of rare harmful effects in randomized controlled trials. JAMA. 2006. May 17;295(19):2275–85. [DOI] [PubMed] [Google Scholar]
- 3.Grijalva CG, Chen L, Delzell E, Baddley JW, Beukelman T, Winthrop KL, et al. Initiation of tumor necrosis factor-α antagonists and the risk of hospitalization for infection in patients with autoimmune diseases. JAMA. 2011. December 7;306(21):2331–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Singh JA, Cameron C, Noorbaloochi S, Cullis T, Tucker M, Christensen R, et al. Risk of serious infection in biological treatment of patients with rheumatoid arthritis: a systematic review and meta-analysis. Lancet. 2015. July 18;386(9990):258–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Schiff M, Weinblatt ME, Valente R, van der Heijde D, Citera G, Elegbe A, et al. Head-to-head comparison of subcutaneous abatacept versus adalimumab for rheumatoid arthritis: two-year efficacy and safety findings from AMPLE trial. Ann Rheum Dis. 2014. January;73(1):86–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Yun H, Xie F, Delzell E, Levitan EB, Chen L, Lewis JD, et al. Comparative Risk of Hospitalized Infection Associated With Biologic Agents in Rheumatoid Arthritis Patients Enrolled in Medicare. Arthritis & Rheumatology (Hoboken, NJ). 2016. January;68(1):56–66. [DOI] [PubMed] [Google Scholar]
- 7.Yun H, Xie F, Delzell E, Chen L, Levitan EB, Lewis JD, et al. Risk of hospitalised infection in rheumatoid arthritis patients receiving biologics following a previous infection while on treatment with anti-TNF therapy. Ann Rheum Dis. 2015. June;74(6):1065–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Curtis JR, Yang S, Patkar NM, Chen L, Singh JA, Cannon GW, et al. Risk of hospitalized bacterial infections associated with biologic treatment among US veterans with rheumatoid arthritis. Arthritis Care Res (Hoboken). 2014. July;66(7):990–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Rutherford AI, Subesinghe S, Hyrich KL, Galloway JB. Serious infection across biologic-treated patients with rheumatoid arthritis: results from the British Society for Rheumatolgy Biologics Register for Rheumatoid Arthritis. Ann Rheum Dis. 2018. March 28; [DOI] [PubMed] [Google Scholar]
- 10.Sakai R, Cho S-K, Nanki T, Watanabe K, Yamazaki H, Tanaka M, et al. Head-to-head comparison of the safety of tocilizumab and tumor necrosis factor inhibitors in rheumatoid arthritis patients (RA) in clinical practice: results from the registry of Japanese RA patients on biologics for long-term safety (REAL) registry. Arthritis Res Ther. 2015. March 23;17:74. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Shourt CA, Crowson CS, Gabriel SE, Matteson EL. Orthopedic Surgery Among Patients with Rheumatoid Arthritis 1980–2007: A Population-based Study Focused on Surgery Rates, Sex, and Mortality. The Journal of Rheumatology. 2012. March 1;39(3):481–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ravi B, Croxford R, Hollands S, Paterson JM, Bogoch E, Kreder H, et al. Increased Risk of Complications Following Total Joint Arthroplasty in Patients With Rheumatoid Arthritis: Increased Risk of Specific Complications in TJA Recipients With RA. Arthritis & Rheumatology. 2014. February;66(2):254–63. [DOI] [PubMed] [Google Scholar]
- 13.Horan TC, Gaynes RP, Martone WJ, Jarvis WR, Emori TG. CDC definitions of nosocomial surgical site infections, 1992: a modification of CDC definitions of surgical wound infections. Infect Control Hosp Epidemiol. 1992. October;13(10):606–8. [PubMed] [Google Scholar]
- 14.Medicare C for, Baltimore MS 7500 SB, Usa M. SummaryMedicareMedicaid [Internet]. 2015. [cited 2015 May 1]. Available from: http://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/MedicareProgramRatesStats/SummaryMedicareMedicaid.html
- 15.Truvenhealth MarketScan Databases [Internet]. Available from: http://truvenhealth.com/Markets/Life-Sciences/Products/Data-Tools/MarketScan-Databases
- 16.Kim SY, Servi A, Polinski JM, Mogun H, Weinblatt ME, Katz JN, et al. Validation of rheumatoid arthritis diagnoses in health care utilization data. Arthritis Res Ther. 2011. Feb 23;13(1):R32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Mahomed NN, Barrett JA, Katz JN, Phillips CB, Losina E, Lew RA, et al. Rates and outcomes of primary and revision total hip replacement in the United States medicare population. J Bone Joint Surg Am. 2003. January;85-A(1):27–32. [DOI] [PubMed] [Google Scholar]
- 18.Katz JN, Barrett J, Mahomed NN, Baron JA, Wright RJ, Losina E. Association between hospital and surgeon procedure volume and the outcomes of total knee replacement. J Bone Joint Surg Am. 2004. September;86-A(9):1909–16. [DOI] [PubMed] [Google Scholar]
- 19.Mahomed NN, Barrett J, Katz JN, Baron JA, Wright J, Losina E. Epidemiology of total knee replacement in the United States Medicare population. J Bone Joint Surg Am. 2005. June;87(6):1222–8. [DOI] [PubMed] [Google Scholar]
- 20.Katz JN, Losina E, Barrett J, Phillips CB, Mahomed NN, Lew RA, et al. Association between hospital and surgeon procedure volume and outcomes of total hip replacement in the United States medicare population. J Bone Joint Surg Am. 2001. November;83-A(11):1622–9. [DOI] [PubMed] [Google Scholar]
- 21.Schneeweiss S, Setoguchi S, Weinblatt ME, Katz JN, Avorn J, Sax PE, et al. Anti-tumor necrosis factor alpha therapy and the risk of serious bacterial infections in elderly patients with rheumatoid arthritis. Arthritis Rheum. 2007. June;56(6):1754–64. [DOI] [PubMed] [Google Scholar]
- 22.Kurtz SM, Ong KL, Lau E, Bozic KJ, Berry D, Parvizi J. Prosthetic joint infection risk after TKA in the Medicare population. Clin Orthop Relat Res. 2010. January;468(1):52–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Bozic KJ, Lau E, Kurtz S, Ong K, Rubash H, Vail TP, et al. Patient-Related Risk Factors for Periprosthetic Joint Infection and Postoperative Mortality Following Total Hip Arthroplasty in Medicare Patients. The Journal of Bone and Joint Surgery (American) [Internet]. 2012. May 2 [cited 2015 Mar 6];94(9). Available from: http://jbjs.org/cgi/doi/10.2106/JBJS.K.00072 [DOI] [PubMed] [Google Scholar]
- 24.Centers for Medicare & Medicaid Services (CMS). 2016. Procedure-Specific Measures Updates and Specifications Report Hospital-Level 30-Day Risk-Standardized Readmission Measures. Elective Primary Total Hip Arthroplasty (THA) and/or Total Knee Arthroplasty (TKA) – Version 5.0 [Internet]. 2016 [cited 2018 Jan 9]. Available from: https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/HospitalQualityInits/Measure-Methodology.html
- 25.Fry DE, Pine M, Locke D, Pine G. Prediction models of Medicare 90-day postdischarge deaths, readmissions, and costs in bowel operations. Am J Surg. 2015. March;209(3):509–14. [DOI] [PubMed] [Google Scholar]
- 26.Krell RW, Girotti ME, Dimick JB. Extended length of stay after surgery: complications, inefficient practice, or sick patients? JAMA Surg. 2014. August;149(8):815–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Ong KL, Lau E, Suggs J, Kurtz SM, Manley MT. Risk of Subsequent Revision after Primary and Revision Total Joint Arthroplasty. Clin Orthop Relat Res. 2010. November;468(11):3070–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Van Arendonk KJ, Tymitz KM, Gearhart SL, Stem M, Lidor AO. Outcomes and costs of elective surgery for diverticular disease: a comparison with other diseases requiring colectomy. JAMA Surg. 2013. April;148(4):316–21. [DOI] [PubMed] [Google Scholar]
- 29.Gagne JJ, Glynn RJ, Avorn J, Levin R, Schneeweiss S. A combined comorbidity score predicted mortality in elderly patients better than existing scores. J Clin Epidemiol. 2011. July;64(7):749–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.U.S. Census Bureau, 2009–2013 5-Year American Community Survey [Internet]. [cited 2015 Oct 3]. Available from: https://www.census.gov/programs-surveys/acs
- 31.Fine JP, Gray RJ. A Proportional Hazards Model for the Subdistribution of a Competing Risk. Journal of the American Statistical Association. 1999. June 1;94(446):496–509. [Google Scholar]
- 32.Austin PC, Fine JP. Practical recommendations for reporting Fine-Gray model analyses for competing risk data. Stat Med. 2017. November 30;36(27):4391–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Cameron AC, Miller DL. A Practitioner’s Guide to Cluster-Robust Inference. J Human Resources. 2015. March 31;50(2):317–72. [Google Scholar]
- 34.Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003. September 6;327(7414):557–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Spreeuwenberg MD, Bartak A, Croon MA, Hagenaars JA, Busschbach JJV, Andrea H, et al. The multiple propensity score as control for bias in the comparison of more than two treatment arms: an introduction from a case study in mental health. Med Care. 2010. February;48(2):166–74. [DOI] [PubMed] [Google Scholar]
- 36.Feng P, Zhou X-H, Zou Q-M, Fan M-Y, Li X-S. Generalized propensity score for estimating the average treatment effect of multiple treatments. Stat Med. 2012. March 30;31(7):681–97. [DOI] [PubMed] [Google Scholar]
- 37.George MD, Baker JF, Yenchih Hsu J, Wu Q, Xie F, Chen L, et al. Perioperative timing of infliximab and the risk of serious infection after elective hip and knee arthroplasty. Arthritis Care Res (Hoboken). 2017. January 27; [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.McCaffrey DF, Griffin BA, Almirall D, Slaughter ME, Ramchand R, Burgette LF. A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med. 2013. August 30;32(19):3388–414. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Austin PC, Schuster T. The performance of different propensity score methods for estimating absolute effects of treatments on survival outcomes: A simulation study. Stat Methods Med Res. 2014. February 3; [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Cole SR, Hernán MA. Constructing inverse probability weights for marginal structural models. Am J Epidemiol. 2008. September 15;168(6):656–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Stundner O, Chiu Y-L, Sun X, Goodman SM, Russell LA, Calloway JJ, et al. Perioperative outcomes in patients with rheumatoid versus osteoarthritis for total hip arthroplasty: a population-based study. Clin Exp Rheumatol. 2013. December;31(6):889–95. [PubMed] [Google Scholar]
- 42.Ravi B, Croxford R, Austin PC, Hollands S, Paterson JM, Bogoch E, et al. Increased Surgeon Experience With Rheumatoid Arthritis Reduces the Risk of Complications Following Total Joint Arthroplasty: Surgical Experience Performing TJA in RA is Protective. Arthritis & Rheumatology. 2014. March;66(3):488–96. [DOI] [PubMed] [Google Scholar]
- 43.Strangfeld A, Eveslage M, Schneider M, Bergerhausen HJ, Klopsch T, Zink A, et al. Treatment benefit or survival of the fittest: what drives the time-dependent decrease in serious infection rates under TNF inhibition and what does this imply for the individual patient? Ann Rheum Dis. 2011. November;70(11):1914–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Dixon WG, Symmons DPM, Lunt M, Watson KD, Hyrich KL, Silman AJ. Serious infection following anti–tumor necrosis factor α therapy in patients with rheumatoid arthritis: Lessons from interpreting data from observational studies. Arthritis & Rheumatism. 2007. September 1;56(9):2896–904. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Listing J, Gerhold K, Zink A. The risk of infections associated with rheumatoid arthritis, with its comorbidity and treatment. Rheumatology (Oxford). 2013. January;52(1):53–61. [DOI] [PubMed] [Google Scholar]
- 46.Dixon WG, Suissa S, Hudson M. The association between systemic glucocorticoid therapy and the risk of infection in patients with rheumatoid arthritis: systematic review and meta-analyses. Arthritis Res Ther. 2011. August 31;13(4):R139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Somayaji R, Barnabe C, Martin L. Risk factors for infection following total joint arthroplasty in rheumatoid arthritis. Open Rheumatol J. 2013. November 29;7:119–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Michaud K, Fehringer EV, Garvin K, O’Dell JR, Mikuls TR. Rheumatoid arthritis patients are not at increased risk for 30-day cardiovascular events, infections, or mortality after total joint arthroplasty. Arthritis Res Ther. 2013;15(6):R195. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Gilson M, Gossec L, Mariette X, Gherissi D, Guyot M-H, Berthelot J-M, et al. Risk factors for total joint arthroplasty infection in patients receiving tumor necrosis factor α-blockers: a case-control study. Arthritis Research & Therapy. 2010;12(4):R145. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Goodman SM, Springer B, Guyatt G, Abdel MP, Dasa V, George M, et al. 2017 American College of Rheumatology/American Association of Hip and Knee Surgeons Guideline for the Perioperative Management of Antirheumatic Medication in Patients With Rheumatic Diseases Undergoing Elective Total Hip or Total Knee Arthroplasty. Arthritis & Rheumatology (Hoboken, NJ). 2017. August;69(8):1538–51. [DOI] [PubMed] [Google Scholar]
- 51.Dixon WG, Abrahamowicz M, Beauchamp M-E, Ray DW, Bernatsky S, Suissa S, et al. Immediate and delayed impact of oral glucocorticoid therapy on risk of serious infection in older patients with rheumatoid arthritis: a nested case-control analysis. Ann Rheum Dis. 2012. July;71(7):1128–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Lyman S, Fields KG, Nocon AA, Ricciardi BF, Boettner F. Prolonged Length of Stay Is Not an Acceptable Alternative to Coded Complications in Assessing Hospital Quality in Elective Joint Arthroplasty. J Arthroplasty. 2015. November;30(11):1863–7. [DOI] [PubMed] [Google Scholar]
- 53.Frisell T, Baecklund E, Bengtsson K, Di Giuseppe D, Forsblad-d’Elia H, Askling J, et al. Patient characteristics influence the choice of biological drug in RA, and will make non-TNFi biologics appear more harmful than TNFi biologics. Ann Rheum Dis. 2017. December 13; [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54.Jämsen E, Varonen M, Huhtala H, Lehto MUK, Lumio J, Konttinen YT, et al. Incidence of Prosthetic Joint Infections After Primary Knee Arthroplasty. The Journal of Arthroplasty. 2010. January;25(1):87–92. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.