Disease-modifying antirheumatic drugs (DMARDs) for rheumatoid arthritis after failure of biologic or targeted synthetic therapy: a systematic review and network meta-analysis.

Rationale

After inadequate response to first-line biologic or targeted synthetic (b/ts) disease-modifying antirheumatic drug (DMARD) therapy in adults with rheumatoid arthritis, there are numerous alternative DMARD options, and current understanding of their comparative benefits and harms is limited.

Objectives

The aim of this living systematic review and network meta-analysis was to compare the benefits and harms of DMARDs after failure of biologic or targeted synthetic DMARDs in adults with rheumatoid arthritis. SEARCH

Methods

We searched CENTRAL, MEDLINE, Embase, and two trial registries (ClinicalTrials.gov and the WHO ICTRP) from inception until 28 November 2025, with no restrictions on language or date of publication. ELIGIBILITY CRITERIA We included randomised controlled trials (RCTs) of adults aged 18 years or older diagnosed with rheumatoid arthritis according to 1958, 1987, or 2010 classification criteria who previously demonstrated inadequate response to a b/ts DMARD. Eligible interventions included conventional synthetic DMARDs (methotrexate, antimalarials, sulfasalazine, leflunomide, ciclosporin, and azathioprine), biologic DMARDs (adalimumab, certolizumab, etanercept, golimumab, infliximab, abatacept, rituximab, tocilizumab, sarilumab, and anakinra), and targeted synthetic DMARDs (tofacitinib, baricitinib, and upadacitinib). OUTCOMES Our critical outcomes were American College of Rheumatology 50% (ACR50) response, withdrawals due to adverse events, radiographic progression, Disease Activity Score 28 (DAS28) remission, pain as measured by visual analogue scale, function as measured by the Health Assessment Questionnaire (HAQ), and serious adverse events. Important outcomes included ACR20, ACR70, serious infections, fatigue, and quality of life. RISK OF BIAS We used Cochrane's RoB 1 tool to assess risk of bias in the included studies. SYNTHESIS

Methods

We first screened the records using an approach that combined machine learning and crowdsourcing to identify probable RCTs. We then reviewed the records identified as RCTs for eligibility and simultaneously classified them to the appropriate Population, Intervention, Comparator, and Outcome (PICO) question(s). Two review authors then extracted relevant data from the included studies in duplicate and independently, with any disagreements resolved by a third review author. A Bayesian random-effects network meta-analysis was conducted using a semi-informative prior probability distribution. We assessed the certainty of evidence for each outcome using the GRADE approach. INCLUDED STUDIES We included 19 unique studies (4779 participants) in the review, all of which were parallel-design RCTs. Eleven trials were placebo controlled; two trials had an inactive comparator arm; and six trials had an active comparator arm. The trials were performed in a well-established rheumatoid arthritis population, with the median baseline disease duration ranging from 6.4 to 14 years, median age of participants ranging from 49 to 58 years, and median baseline disease activity (DAS28) ranging from 4.87 to 6.79. SYNTHESIS OF

Results

ACR50 response We found moderate-/high-certainty evidence that using a tumour necrosis factor (TNF) inhibitor not previously tried, interleukin-6 (IL-6) inhibitors, abatacept, rituximab, and Janus kinase (JAK) inhibitors was more effective than placebo: TNF inhibitor not previously tried (odds ratio (OR) 6.04, 95% credible interval (CrI) 2.49 to 16.3; high-certainty evidence), sarilumab (OR 3.11, 95% CrI 1.25 to 7.76; high-certainty evidence), tocilizumab 4 mg/kg intravenous (OR 5.31, 95% CrI 2.09 to 12.09; high-certainty evidence), tocilizumab 8 mg/kg intravenous (OR 10.03, 95% CrI 3.65 to 31.27; high-certainty evidence), subcutaneous abatacept (OR 4.31, 95% CrI 0.97 to 18.28; moderate-certainty evidence), intravenous abatacept (OR 4.57, 95% CrI 2.21 to 10.18; high-certainty evidence), rituximab (OR 5.50, 95% CrI 2.31 to 13.12; high-certainty evidence), upadacitinib (OR 3.93, 95% CrI 1.53 to 10.32; high-certainty evidence), tofacitinib (OR 3.93, 95% CrI 1.53 to 10.32; high-certainty evidence), baricitinib 2 mg (OR 2.00, 95% CrI 0.77 to 5.23; moderate-certainty evidence), baricitinib 4 mg (OR 2.79, 95% CrI 1.10 to 7.22; high-certainty evidence). With an assumed risk for placebo of 78 out of 1000 patients, the expected effects for the active drugs ranged from 143 (baricitinib 2 mg) to 455 (tocilizumab 8 mg/kg). Withdrawals due to adverse events For most interventions, there were sparse data with low-certainty evidence, except for TNF inhibitor not previously tried (risk ratio (RR) 0.32, 95% CrI 0.07 to 1.1; moderate-certainty evidence), which is probably less harmful than placebo, and sarilumab (RR 1.98, 95% CrI 0.57 to 7.19; moderate-certainty evidence), which is probably more harmful than placebo. Low-certainty evidence suggests that intravenous abatacept (RR 0.92, 95% CrI 0.34 to 2.79), upadacitinib (RR 0.41, 95% CrI 0.08 to 1.83), and baricitinib 2 mg (RR 0.93, 95% CrI 0.22 to 4.01) may be less harmful than placebo. Low-certainty evidence suggests that tocilizumab 4 mg/kg (RR 1.39, 95% CrI 0.39 to 5.28), tocilizumab 8 mg/kg (RR 1.49, 95% CrI 0.51 to 4.81), subcutaneous abatacept (RR 3.11, 95% CrI 0.05 to 249.6), rituximab (RR 2.38, 95% CrI 0.44 to 23.31), tofacitinib (RR 1.37, 95% CrI 0.35 to 5.58), and baricitinib 4 mg (RR 1.43, 95% CrI 0.38 to 5.81) may be more harmful than placebo. Data were insufficient to perform a network meta-analysis for radiographic progression. For DAS28 and the HAQ, there was mostly moderate-/high-certainty evidence of a benefit, with some exceptions for comparisons with indirect evidence only that was of low or very low certainty. For the other efficacy outcomes, data were sparse with wide credible intervals, and the certainty of evidence was typically low. AUTHORS'

Conclusions

We found high-certainty evidence that nine therapies and moderate-certainty evidence that two therapies provide a clinically important benefit in improving disease activity compared to placebo for people with rheumatoid arthritis after failure of b/ts DMARD therapy. There was significant uncertainty surrounding treatment-related harms, with the evidence having been downgraded for serious or extremely serious imprecision. Pair-wise comparisons showed no significant differences among therapies, although the certainty of evidence was low. The lack of clarity regarding safety and comparative efficacy suggests that treatment decisions should be guided by individual patient characteristics and preferences.

Funding

This work was supported by grants from the Canadian Institutes for Health Research (CIHR) [Funding Reference Numbers (FRN) 178375 and 180324] and the National Health and Medical Research Council (NHMRC) Cochrane Collaboration. This research was supported by Arthritis Society Canada (Doctoral Studentship TGP-23-0211). REGISTRATION This study was outlined in a Cochrane protocol (CD013562; DOI 10.1002/14651858.CD013562).