Comparative Effectiveness of Methotrexate-Based Disease-Modifying Antirheumatic Drug Combinations for Rheumatoid Arthritis



Journal Title

Journal ISSN

Volume Title



Objective: The study objectives were [1] to examine factors associated with adding a conventional synthetic (cs) disease-modifying antirheumatic drug (DMARD), tumor necrosis factor inhibitor (TNFi) biological (b) DMARD, non-TNFi bDMARD, or targeted synthetic (ts) DMARD for patients with rheumatoid arthritis (RA) initiating methotrexate (MTX); [2] to compare the effectiveness of MTX-based disease-modifying antirheumatic drugs (DMARDs) combinations, including csDMARDs plus MTX, TNFi plus MTX, non-TNFi biologics plus MTX or tsDMARDs plus MTX in RA; [3] to examine the risk of cardiovascular (CV) events associated with different types of DMARDs exposure in RA, focusing on individual DMARDs and their combinations with MTX. Methods: This study was performed using IBM MarketScan commercial claims (2012 and 2014). In aim 1, a retrospective cohort included patients with RA (International Classification of Diseases, Ninth Revision-Clinical Modification [ICD9-CM] code (714.0x-714.4x and 714.81) identified using inpatient/outpatient medical diagnosis files and initiating at least 1 MTX between July 2012 and Dec 2013 (the first MTX claim was the index date); this required at least 6-mon pre-index period to identify new users of MTX, and with 1-year post-index period through 2014 used for follow-up. Given study focus on first addition of a DMARD prescription, patients who received any prescription for DMARDs of interests during the pre-index period till 30-day post-index period were excluded. Further, in order to examine treatment addition of a DMARD, combination therapy cohorts involving the addition of a DMARD to MTX were created, defined as having ≥1 DMARD prescription during the post-index period as well as having an MTX refill within 30 days of the DMARD prescription. Eligible patients were then followed for one year post-index period to examine the type of addition of a csDMARD, TNFi, non-TNFi or tsDMARD. The DMARD prescriptions, including oral and self-injected biologics were identified using National Drug Codes (NDCs) from outpatient pharmacy claims, and physician-administrated infusion/injectable biologics were identified using the Healthcare Common Procedural Coding System (HCPCS) in medical claims. Multivariable logistic regression was used to investigate factors associated with adding a DMARD, adjusting for covariates within the context of the Anderson healthcare utilization model, using the addition of a csDMARD as a reference. In aim 2, study cohort included RA patients aged 18 years who were newly prescribed a DMARD prescription (a csDMARD, TNFi, non-TNFi, or tsDMARDs) between July 2012-Dec 2013 (the first prescription fill date was the index date); to implement new user design, patients were initiators of DMARDs because no prescription claims of these DMARDs in the 6-mo pre-index period. All patients had continuous enrollment for 6-mon pre-index till 12-mon post-index period. To focus on patient population with RA-specific DMARD initiation, study excluded patients with inflammatory disorders during the 6-mon interval of the index date. The combination therapy cohorts were patients initiating a DMARD in combination with MTX, defined as having an MTX in the 30-d post-index and 30 d of overlapping use of the MTX prescription and the index DMARD prescription; according to the first DMARD prescription dispensed or infusion received, drug exposure was classified into 4 groups: (1) csDMARD + MTX; (2) TNFi + MTX; (3) Non-TNFi + MTX; (4) tsDMARD + MTX. According to ABM, baseline covariates, including demographics and clinical variables (Elixhauser score, claims-based RA severity score (CIRAS), RA-related comorbidities and comedications, healthcare utilization markers) identified in the 6-month baseline period were conceptualized into predisposing, enabling and need factors. The outcome measured was the claims based effectiveness algorithm, which classified the therapy as effective if all six criteria were satisfied during the 12-month post-index period. These criteria measured changes to the index therapy, including high adherence, no switching/addition of b/tsDMARDs, no addition of new csDMARDs, no increase in dose or frequency of the DMARD compared to the index claim, no >2 glucocorticoid injection or no new/increased oral glucocorticoid use. Propensity score, measured as the probability of being assigned in 1 treatment initiation group conditional upon baseline covariates were estimated using the Gradient boosted model (GBM). Multivariable logistic regression adjusting for propensity score and inverse probability of treatment weighting derived via generalized boosted models was conducted to compare the algorithm-based effectiveness outcome by these combination groups. In aim 3, in the nested case-control study, the base cohort included adults with RA aged 18 years who newly prescribed a DMARD between Jan 2013 and Dec 2014 (cohort entry date) and were free of CV history in the 1-year before the cohort entry date. All patients had continuous enrollment during 1-year before the cohort entry date. Cases were those who developed any CV event in the composite measure of myocardial infarction, stroke, heart failure, or coronary revascularization procedure since cohort entry date and incident density procedure were used to randomly selected ten controls. Information on DMARDs exposure was measured in 90 days preceding the event date, stratified as combination therapy with MTX or individual DMARD use. Factors related to indication of DMARD use or CV risk including demographics, Elixhauser score, CIRAS, RA comorbidities and comedications, and healthcare ulitization markers were measured in the 1-year before the cohort entry. Conditional logistic regression was used to examine the adjusted odds ratio for the risk of CV with DMARD use, adjusting for demographics and clinical variables. All statistical analysis was conducted using SAS 9.4. Results: In aim 1, after the inclusion/exclusion criteria, 8,350 RA patients were newly prescribed an MTX during the identification period; of them, 32% (n=2,665) initiated a DMARD in a 1-year post-index period. Overall, 35% (n=945) started the MTX-based DMARDs combination regimen (mean [SD] 49 [10] years; 746 [78.9%] women). The majority [58% (n=550)] added the TNFi to MTX, followed by 42% (n=352), who added csDMARDs to MTX; only 15% (n=40) added the non-TNFi to MTX, and very few (2%, 3) added the tsDMARDs to MTX. Addition with, non-TNFi biologics or tsDMARDs were not considered for the analysis. Enabling factors of Preferred Provider Organization (odds ratio (OR) [95% confidence interval (CI)], 1.40 [1.03-1.89]), and need factors of chronic pulmonary disease (1.73 [1.02-2.96]), liver disease (4.11 [1.46-11.61]), Elixhauser index (0.93 [0.88-0.98]) were associated with adding a TNFi. In aim 2, 3,174 RA patients were initiating the MTX-based DMARD combination (mean [SD] 50 [10] years; 2539 [80%] women). There were 1,568 (49%) initiated the csDMARD plus MTX, 1,343 (42%) initiated the TNFi plus MTX, 240 (8%) initiated the non-TNFi biologics plus MTX, and 23 (1%) started the tsDMARD plus MTX. The analysis did not include the tsDMARD combination group because of the small sample size. The effectiveness rate was 9.95% of patients in the csDMARD+ MTX group, 20.83% in the non-TNFi biologics + MTX group, and 20.48% in the TNFi+MTX group. In the multivariable logistic analysis with IPTW, the csDMARD combination group was significantly less effective than the TNFi combination group (aOR: 0.422 [95% CI: 0.341-0.524]), whereas there was no difference between non-TNFi biologic combination group compared to the TNFi combination group in the algorithm effectiveness (aOR: 1.063 [95% CI: 0.680-1.662]) (Table 3). In aim 3 of the baseline cohort, including patients with early RA starting a DMARD prescription, 270 cases with incident CV events and 2,700 matched controls (mean [SD] 54[8] years; 2257 [76%] women) were included. Using the MTX as the reference, the use of DMARDs in combination therapy with MTX showed no risk of CV, including TNFi plus MTX (adjusted odds ratio [aOR] 1.306; 95% confidence interval (CI) 0.578-2.952); non-TNFi plus MTX (aOR 0.844; 95% CI 0.157-4.547) or another non-MTX csDMARDs plus MTX (aOR 0.859; 95% CI 0.375-1.971). Also, individual DMARDs therapy, including TNFi (aOR 1.379; 95% CI 0.627-3.032), non-TNFi (aOR 0.925; 95% CI 0.299-2.859), or another non-MTX csDMARDs (aOR 1.552; 95% CI 0.766-3.144) had a similar risk of CV than MTX monotherapy. Conclusions: The research found that enabling factor (insurance type) and need factors (Elixhauser score, coexisting liver disease and pulmonary diseases) were associated with adding a TNFi vs. adding a csDMARD in RA patients after starting a MTX. Using the claim-based algorithm, the combinations of non-TNFi plus MTX was similarly effective against the TNFi plus MTX in achieving the algorithm, but the csDMARD plus MTX was less effective than the TNFi plus MTX. In addition, the comparative safety analyses found that the MTX combination treatment with DMARDs and DMARDs monotherapy had similar risk of CV, as compared to MTX monotherapy in patients with RA.



Rheumatoid Arthritis, Disease-modifying Antirheumatic Drugs, Biological Disease-modifying Antirheumatic Drugs, Conventional Synthetic Disease-modifying Antirheumatic Drugs, Comparative Effectiveness, Cardiovascular Risk