Objectives Recent results from ‘ORAL Surveillance’ trial have raised concerns regarding the cardiovascular safety of tofacitinib in patients with rheumatoid arthritis (RA). We further examined this safety concern in the real-world setting.
Methods We created two cohorts of patients with RA initiating treatment with tofacitinib or tumour necrosis factor inhibitors (TNFI) using deidentified data from Optum Clinformatics (2012–2020), IBM MarketScan (2012–2018) and Medicare (parts A, B and D, 2012–2017) claims databases: (1) A ‘real-world evidence (RWE) cohort’ consisting of routine care patients and (2) A ‘randomised controlled trial (RCT)-duplicate cohort’ mimicking inclusion and exclusion criteria of the ORAL surveillance trial to calibrate results against the trial findings. Cox proportional hazards models with propensity score fine stratification weighting were used to estimate HR and 95% CIs for composite outcome of myocardial infarction and stroke and accounting for 76 potential confounders. Database-specific effect estimates were pooled using fixed effects models with inverse-variance weighting.
Results In the RWE cohort, 102 263 patients were identified of whom 12 852 (12.6%) initiated tofacitinib. The pooled weighted HR (95% CI) comparing tofacitinib with TNFI was 1.01 (0.83 to 1.23) in RWE cohort and 1.24 (0.90 to 1.69) in RCT-duplicate cohort which aligned closely with ORAL-surveillance results (HR: 1.33, 95% CI 0.91 to 1.94).
Conclusions We did not find evidence for an increased risk of cardiovascular outcomes with tofacitinib in patients with RA treated in the real-world setting; however, tofacitinib was associated with an increased risk of cardiovascular outcomes, although statistically non-significant, in patients with RA with cardiovascular risk factors.
Trial registration number NCT04772248.
- rheumatoid arthritis
- antirheumatic agents
Data availability statement
Data may be obtained from a third party and are not publicly available. The databases used in the study can be accessed through a license with MarketScan, Optum, and Medicare. No additional data are available from the authors.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
What is already known about this subject?
Recently released topline findings from ‘ORAL Surveillance’ postmarketing trial have raised concerns that tofacitinib, in comparison with tumour necrosis factor inhibitors, may increase the risk of cardiovascular disease in patients with rheumatoid arthritis.
What does this study add?
In this multidatabase, population-based study including 102 263 rheumatoid arthritis patients, tofacitinib was not associated with an increased risk of cardiovascular outcomes when compared with tumour necrosis factor inhibitors.
A numerically increased risk of cardiovascular outcomes was observed in older patients with cardiovascular risk factors or history of cardiovascular disease.
How might this impact on clinical practice or future developments?
In this study in real-world setting, tofacitinib, in comparison with tumour necrosis factor inhibitors, was not associated with increased risk of cardiovascular outcomes, although an increased risk of cardiovascular outcomes with tofacitinib cannot be ruled out in patients with cardiovascular risk factors or history of cardiovascular disease.
Rheumatoid arthritis (RA) is a chronic autoimmune disease that affects approximately 0.2% of adults worldwide.1 RA is characterised by systemic inflammation that leads to joint damage and extra-articular manifestations including cardiovascular (CV) disease (CVD) that adversely impact morbidity and mortality. Janus kinase (JAK) inhibitors (consisting of tofacitinib, baricitinib, upadacitinib) are a class of targeted synthetic disease-modifying antirheumatic drugs (DMARDs) that are increasingly used for management of patients diagnosed with moderate to severely active RA.2 3 Tofacitinib, first approved in USA in 2012, is the most commonly prescribed JAK inhibitor.2 4
Tofacitinib has been associated with improved disease control in patients with RA with similar efficacy when compared with other biological DMARDs (bDMARDs) such as adalimumab.5–7 However, recent reports have from the ‘ORAL Surveillance’ postmarketing safety trial have indicated a potential for increased risk of major adverse CV events (MACE) with tofacitinib, in comparison with tumour necrosis factor inhibitors (TNFI), among patients with RA at least 50 years of age and with at least one risk factor for CVD (HR 1.33, 95% CI 0.91 to 1.94).8–10 Thus, the aim of this study was to conduct a large population-based observational study to further examine the risk of CV outcomes with tofacitinib in patients with RA treated in routine clinical care settings.
Data sources and study design
We conducted a new user, active comparator cohort study (online supplemental figure 1) using claims data from the Optum Clinformatics (November 2012–June 2020), IBM MarketScan (November 2012–December 2018), and Medicare (parts A, B and D, November 2012–December 2017) databases.11 The Optum and MarketScan claims databases capture de-identified record of over 200 million and 78 million commercially insured patients respectively in the USA. Medicare is a federal health insurance programme and provides healthcare coverage for residents of the USA aged at least 65 years and patient aged less than 65 with a disability status as ascertained by US Social Security Administration. All three data sources provide longitudinal information including patient demographics, inpatient and outpatient medical diagnoses and procedures, and prescription dispensing records. The protocol for this study was registered on ClinicalTrials.gov (NCT04772248) and reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.12 13 Requirement for patient informed consent was waived because all personal identifiers were removed from each of the datasets to ensure patient confidentiality. Signed data licence agreements were obtained for all data sources.
The base population consisted of patients initiating treatment on tofacitinib or a TNFI (infliximab, adalimumab, certolizumab pegol, etanercept and golimumab). Cohort entry date corresponded to first TNFI or tofacitinib dispensation (‘index drug’) with a minimum of 365 days of continuous enrollment in health plan prior to and including the cohort entry date. Patients required at least two diagnosis codes for RA in any setting during the 365 days baseline period (between 7 and 365 days apart).14 A previous validation study demonstrated a positive predictive value of 86% for this claims-based algorithm which combines two diagnosis codes for RA with one DMARD dispensing record.14 To ensure the inclusion of new users, we excluded TNFI users with a prescription of index TNFI and tofacitinib users with prescription of tofacitinib in the 365 days prior to cohort entry date. We also excluded patients with a prescription of tofacitinib and TNFI on cohort entry date, patients missing data on age or gender, and those with admission to nursing facility or hospice on or prior to cohort entry date. TNFI users with history of use of any JAK inhibitor or with prescriptions for multiple agents from the TNFI class on cohort entry date were also excluded. Finally, tofacitinib users with prescriptions of other approved JAK inhibitors (ie, baricitinib or upadacitinib) on or at any point prior to cohort entry date were excluded.
From this source population of patients with RA initiating treatment with tofacitinib or TNFI, we created two study cohorts. The first cohort, ‘real-world evidence (RWE)’, included all patients with RA from routine care. The RWE cohort included patients at least 18 years of age in MarketScan and Optum (≥65 in Medicare) at cohort entry date. The second cohort, ‘randomised controlled trial (RCT)-duplicate cohort’, mimicked the inclusion and exclusion criteria of the ORAL surveillance trial.9 This study population was used to calibrate our findings and ensure comparability with the ORAL Surveillance trial results.8 10 The RCT-duplicate cohort was restricted to patients at least 50 years of age (65 in Medicare) with at least one methotrexate dispensation in 6 months prior to cohort entry date. This cohort was also restricted to patients with at least one CV risk factor including history of smoking, hypertension, dyslipidaemia, diabetes mellitus, ischaemic heart disease or family history of ischaemic heart disease. Patients hospitalised with infections in the 30 days prior to cohort entry date and pregnant patients were excluded from the RCT-duplicate cohort.
Exposure and outcome definition
We used an as-treated exposure definition whereby patients were followed from treatment initiation for study outcomes until treatment discontinuation or switch, insurance disenrollment, death, or end of the study period, whichever occurred first. The primary endpoint was defined as a composite CV outcome consisting of hospitalisations for myocardial infarction (MI) or stroke. Patients were followed for CV outcomes on the day after the cohort entry date (online supplemental figure 1). Individual CV outcomes were also examined independently as secondary outcomes including MI, stroke, heart failure hospitalisation and coronary revascularisation. We also examined the risk of all-cause mortality as an additional secondary outcome. Finally, we examined the risk of herpes zoster as a positive control outcome as previous studies have established an increased risk of herpes zoster with tofacitinib.15 16
We assessed 76 potential confounders (75 in MarketScan) during the baseline covariate assessment period defined as the 365 days prior to treatment initiation (full list of covariates are included in online supplemental methods).
Descriptive statistics were used to summarise baseline characteristics for each study cohort. Crude incidence rates and corresponding 95% CIs were reported for each study outcome. Propensity score (PS) fine stratification weighting was used to account for measured confounders in this study (details outlined in online supplemental methods).17 Standardised differences (%) were used to assess the balance in individual covariates between two treatment groups before and after PS fine-stratification weighting.18 19 Cox proportional hazards model were used to estimate HRs and corresponding 95% CIs accounting for potential confounders using PS fine stratification weights and using an as-treated exposure definition in the primary analysis. Robust variance estimation was used to calculate 95% CI to account for weighting. We also assessed crude and weighted difference in rates and corresponding 95% CI when comparing tofacitinib with TNFI using Poisson regression. Effect estimates were pooled across three databases using fixed effects model with inverse variance weighting. We examined cumulative incidence of composite CV outcomes and the corresponding 95% CI separately for each treatment group. We used the Aetion Evidence Platform for cohort construction.20 Analyses were conducted using SAS V.9.4 (SAS Institute) and R (R Foundation for Statistical Computing, Vienna, Austria).
Secondary and sensitivity analyses
Prespecified subgroup analyses were conducted based on age (≤65 and >65), sex and baseline CVD (RWE cohorts only). In addition, we examined the risk of CV outcomes by stratifying by unique number of previous agents of bDMARDs (ie, 0 vs ≥1). Secondary analysis was also conducted by using an intention-to-treat exposure definition whereby patients were censored 365 days after initiation of treatment with tofacitinib or TNFI. We also conducted 1:1 PS matching where each patient initiating tofacitinib was matched with a patient initiating TNFI using nearest neighbour greedy matching without replacement using a calliper of 0.025 on the natural scale of the PS.21 22 Finally, sensitivity analyses were conducted by restricting the TNFI comparator group in RWE and RCT-duplicate cohorts to only adalimumab and etanercept users, the specific TNFI which were the comparator in the ORAL Surveillance trial.9 10
The Consort diagrams for construction of the RWE and RCT-duplicate cohorts are outlined in online supplemental tables 1–6. In RWE cohort, 28 568, 34 083 and 39 612 patients who met the inclusion and exclusion criteria were identified from Optum, MarketScan and Medicare, respectively, of whom 13.2%, 15.6% and 9.5% initiated treatment on tofacitinib (online supplemental table 7). The mean age, in years, comparing tofacitinib and TNFI users was 56.8 vs 54.6 in Optum, 54.7 vs 52.7 in MarketScan, and 72.1 vs 72.2 in Medicare. The majority of patients in RWE cohort were female across the three databases (77%–79%). The prevalence of CVD risk factors and previous use of comedications was slightly higher in tofacitinib users compared with TNFI users (online supplemental table 7).
In RWE cohort, 13% of patients in Optum, 10% in MarketScan, and 31% in Medicare had a history of CVD. There were no discernable differences across most markers of healthcare utilisation when comparing tofacitinib and TNFI users (online supplemental table 7). Overall, PS fine stratification achieved excellent covariate balance with standardised differences close to zero across all covariates (table 1, online supplemental table 8).
The crude incidence rates of the primary CV endpoint per 100 person-years (95% CI) for tofacitinib and TNFI users were 0.73 (0.47 to 1.09) and 0.61 (0.51 to 0.72) in Optum, 0.75 (0.52 to 1.05) and 0.52 (0.44 to 0.61) in MarketScan, and 2.14 (1.66 to 2.70) and 1.86 (1.71 to 2.02) in Medicare (table 2).
In the primary analysis, the pooled weighted HR (95% CI) for CV outcomes when comparing tofacitinib with TNFI was 1.01 (0.83 to 1.23) with weighted rate difference (95% CI) corresponding to 0.02 (−0.19 to 0.23) CV events per 100 person-years (figure 1 and online supplemental table 9). Correspondingly, there was no differences in cumulative incidence of composite CV outcomes when comparing tofacitinib with TNFI in any of the three databases (online supplemental figure 2). Among tofacitinib users, the median (IQR) months to CV events was 6.9 (2.9–16.3) in Optum, 5.1 (2.0–12.3) in MarketScan and 6.0 (2.0–13.4) in Medicare. Among TNFI users, the median (IQR) months to CV events was 7.5 (2.7–17.5) in Optum, 6.1 (2.5–11.7) in MarketScan and 6.3 (2.4–15.1) in Medicare.
In subgroup analyses, the pooled weighted HR (95% CI) was 1.27 (0.95 to 1.70) and 0.81 (0.61 to 1.07) among patients with and without history of CVD respectively (figure 2 and online supplemental table 9). The pooled weighted HR (95% CI) among patients ≤65 years of age was 1.00 (0.66 to 1.50) and 1.05 (0.84 to 1.33) for patient aged more than 65 years. No association was observed across other subgroups including among males (pooled weighted HR 1.05, 95% CI 0.70 to 1.56), females (pooled weighted HR 0.97, 95% CI 0.77 to 1.23), patients with previous use of bDMARDs (pooled weighted HR 1.06, 95% CI 0.79 to 1.40), and patients without previous use of bDMARDs (pooled weighted HR 1.02, 95% CI 0.77 to 1.35). Consistent results were observed across other sensitivity and secondary analyses including PS matching, intention-to-treat exposure definition, and restriction of the TNFI comparator to adalimumab and etanercept users (online supplemental table 9).
For individual CV outcomes, the pooled weighted HR (95% CI) was 1.04 (0.82 to 1.33) for MI, 0.93 (0.66 to 1.31) for stroke, 1.07 (0.79 to 1.46) for heart failure hospitalisation and 1.04 (0.78 to 1.40) for coronary revascularisation (online supplemental table 10) when comparing tofacitinib users with TNFI users. The pooled weighted HR (95% CI) was 1.20 (0.98 to 1.46) for all-cause mortality. For the positive control outcome, we successfully replicated the known association between tofacitinib and risk of herpes zoster (pooled weighted HR 1.98, 95% CI 1.78 to 2.19).
In the RCT-duplicate cohort, 6878, 8071 and 20 121 patients were identified from Optum, MarketScan and Medicare, respectively, of whom 11.6%, 14.3% and 7.7% initiated treatment with tofacitinib (online supplemental table 11). Overall, PS fine stratification weighting achieved excellent covariate balance in this study population with standardised differences close to zero for all covariates (online supplemental table 12).
The crude incidence rates of the primary CV endpoint per 100 person-years (95% CI) for tofacitinib and TNFI users were 1.33 (0.64 to 2.45) and 0.94 (0.71 to 1.23) in Optum, 1.22 (0.65 to 2.08) and 0.80 (0.60 to 1.04) in MarketScan, and 2.39 (1.64 to 3.38) and 1.78 (1.58 to 2.00) in Medicare (table 2). In the primary analysis, the pooled weighted HR (95% CI) for primary CV outcome was 1.24 (0.90 to 1.69) corresponding to a pooled weighted rate difference (95% CI) of 0.28 (−0.24 to 0.80) CV events per 100 person-years when comparing tofacitinib users with TNFI users (figure 1 and online supplemental table 13). The cumulative incidence of CV outcomes was similar when comparing with tofacitinib with TNFI users in Optum and MarketScan but was slightly higher among tofacitinib users in Medicare, although with wide CIs for these analyses (online supplemental figure 3). The median months (IQR) to CV events among tofacitinib users was 6.9 (2.6–14.2) in Optum, 6.0 (2.8–11.8) in MarketScan and 5.2 (1.7–12.4) in Medicare. Among TNFI users, the median (IQR) months to CV events was 6.9 (3.0–11.0) in Optum, 7.1 (2.8–12.5) in MarketScan and 6.8 (2.2–15.8) in Medicare. In sensitivity analysis restricting comparator to adalimumab and etanercept (online supplemental table 13), the pooled weighted HR (95%) CI for primary CV outcome was 1.32 (0.94 to 1.86).
Overall, in this large population-based study, tofacitinib in comparison with TNFI was not associated with risk of composite CV outcome in patients with RA treated in real-world settings (pooled weighted HR 1.01, 95% CI 0.83 to 1.23). Results from the RCT-duplicate cohort were consistent with those reported from the ORAL surveillance trial (pooled weighted HR 1.24, 95% CI 0.90 to 1.69 vs trial: 1.33, 95% CI 0.91 to 1.94).8 10 An increased risk of CV outcomes was also observed among RWE patients with history of CVD (pooled weighted HR 1.27, 95% CI 0.95 to 1.70) but not those without history of CVD (pooled weighted HR 0.81, 95% CI 0.61 to 1.07).
The findings from previous studies examining the association between tofacitinib and CV outcomes have been discordant. Recent reports from the ORAL Surveillance trial have indicated that both 5 mg and 10 mg two times per day dose of tofacitinib, in comparison with TNFI, were associated with increased risk of MACE (HR 1.24, 95% CI 0.81 to 1.91 and HR 1.43, 95% CI 0.94 to 2.18, respectively).8 10 This trial consisted 4362 patients who were at least 50 years of age, with one at least one risk factor for CVD, and with a background of treatment with methotrexate.9 10 Results from a recent meta-analysis of RCTs, excluding ORAL surveillance, were inconclusive for the association between tofacitinib and CV risk in patients with RA (OR 1.29, 95% CI 0.40 to 4.13) or chronic plaque psoriasis (OR 3.61, 95% CI 0.71 to 18.43) due to low event rates.23 In contrast, among patients enrolled in the CORRONA RA registry in the USA, tofacitinib, in comparison with bDMARDs (including TNFI and non-TNFI biologics), was not associated with an increased risk of MACE which was defined as MI, stroke, transient ischaemic attack or CV death (HR 0.61, 95% CI 0.34 to 1.06).24 Our results suggest that the association between tofacitinib and CV outcomes may be modified by baseline CV risk. In patients from our RWE cohort who had no underlying CV risk factors or history (87% of patients in Optum, 90% in MarketScan and 69% in Medicare), we noted no detectable impact of tofacitinib treatment on risk of adverse CV outcomes. However, among patients with CV risk factors or history, estimates consistently suggested a potentially elevated risk. We recommend continuing research to better understand risk–benefit trade-offs of this important treatment option in a wide range of patients with RA.
Overall, there is no known direct mechanism that would explain a detrimental effect of tofacitinib on risk of CV outcomes. In phase II and III trials, both doses of tofacitinib (5 mg or 10 mg) either as monotherapy or in combination with non-bDMARDs were associated with 15% to 20% increase in low-density lipoprotein cholesterol (LDL-C) and high-density lipoprotein cholesterol (HDL-C levels) when comparing 4 weeks after treatment initiation with baseline. The changes in LDL-C and HDL-C levels persisted 12 months after treatment initiation.25 However, the levels of the ratio of total-cholesterol: HDL-C and LDL-C: HDL-C levels, which are more reliable predictors of CV events, did not change.25 In another study with 46 patients with RA, tofacitinib was not associated with changes in carotid intima–media thickness when comparing 54 weeks after treatment initiation with baseline (1.09±0.69 and 1.08±0.78 mm, respectively).26 Additional mechanistic studies will be required in light of potential increased risk of CV outcomes associated with tofacitinib in ORAL Surveillance trial.
This study has strengths and limitations. First, we conducted a large multidatabase population-based studies with more than 102 000 patients, a sample size larger than the ORAL Surveillance trial and CORRONA RA registry study in the USA.8 10 24 Second, we comprehensively assessed the risk of individual CV outcomes including MI, stroke, heart failure and coronary revascularisation. Third, we employed a new user design active comparator design to control for confounding by disease severity and circumvent prevalent-user bias.11 Fourth, we calibrated our results using the RCT-duplicate cohort to assess the validity of the study and ensure that are results were comparable with those of the ORAL surveillance postmarketing trial.8 10 Fifth, we found an increased risk of herpes zoster infection which was included as a positive control outcome, consistent with previous studies.15 16 Finally, the study protocol was preregistered prior to conducting our study.12 Our study has some limitations. First, residual confounding by factors not captured in administrative claims including RA activity is possible. However, we used an active comparator group (ie, TNF inhibitors) and adjusted for 76 confounders (75 in MarketScan) including multiple variables that may serve as proxies for RA disease severity. Reassuringly, a recent study using the CORRONA RA registry, a prospective disease-based registry inclusive of 50 605 patients with RA across 177 private and academic practices in the USA, demonstrated that patients with RA who are treated with tofacitinib are comparable to patients treated with bDMARDs in regards to RA-related variables including clinical disease activity index.24 In addition, we observed approximately equal distribution in time to CV events throughout follow-up after treatment initiation among tofacitinib and TNFI users in RWE and RCT-duplicate cohorts. Second, exposure misclassification is possible due to incomplete adherence to study drugs. To minimise exposure misclassification, we implemented an as-treated exposure definition where the occurrence of study outcomes was assessed while patients were on treatment. Finally, we could not assess the risk of CV outcomes with newer JAK inhibitors including baricitinib and upadacitinib, and thus, additional studies will be required to examine the risk of CV outcomes with these newer agents.
Overall, in this multidatabase population-based study, we did not find evidence for an increased risk of CV outcomes with tofacitinib, in comparison with TNFI, among patients with RA treated in the real-world setting. However, concordant with results from ORAL Surveillance safety trial, tofacitinib, in comparison with TNFI, was associated with an elevated risk of CV outcomes, though statistically non-significant, in patients with RA with CV risk factors or a history of CVD. Thus, an elevated risk of CV outcomes cannot be ruled out in patients with CV risk factors or history of CVD.
Data availability statement
Data may be obtained from a third party and are not publicly available. The databases used in the study can be accessed through a license with MarketScan, Optum, and Medicare. No additional data are available from the authors.
Patient consent for publication
The study protocol received approval (IRB #2011P002580-207) from the Institutional Review Board of Brigham and Women’s Hospital (Boston, Massachusetts).
RJD has received research grants to the Brigham and Women’s Hospital from Bayer, Novartis and Vertex for unrelated projects. SCK has received research grants to the Brigham and Women’s Hospital from Roche/Genentech, Pfizer, Bristol-Myers Squibb, Roche, and AbbVie for unrelated studies. All other authors have no conflict of interests to disclose.
Handling editor Josef S Smolen
Contributors RJD had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: RJD, SCK and FK-K. Analysis of data: RJD, FK-K and SBL. Interpretation of data and drafting of the manuscript: all authors. RJD acts as guarantor and accepts full responsibility for the work and/or the conduct of the study, had access to the data and controlled the decision to publish.
Funding This study was funded by internal sources of the Division of Pharmacoepidemiology and Pharmacoeconomics, Brigham and Women’s Hospital & Harvard Medical School, Boston, MA, USA. SCK is supported by the National Institutes of Health (NIH) grant - K24AR078959. FK-K is supported by a postdoctoral fellowship from Fonds de recherche du Québec-Santé (FRQS). This work was previously presented at American College of Rheumatology Convergence 2021 Meeting (Abstract 1939): FK-K, SCL, HL, SBL and RJD. Risk of Cardiovascular Outcomes in Patients Treated with Tofacitinib: First Results from the Safety of TofAcitinib in Routine Care Patients with Rheumatoid Arthritis (STAR-RA) Study. 2021.
Competing interests RJD has received research grants to the Brigham and Women’s Hospital from Bayer, Novartis, and Vertex for unrelated projects. SCK has received research grants to the Brigham and Women’s Hospital from Roche/Genentech, Pfizer, Bristol-Myers Squibb, Roche, and AbbVie for unrelated studies. All other authors have no conflict of interests to disclose.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.