Influence of study characteristics on harm estimates from randomised controlled trials in patients with inflammatory arthritis receiving biological or synthetic antirheumatic drugs: a meta-epidemiological study

Objective: To examine the association between study characteristics and the harms reported in randomised controlled trials (RCTs) on biological and targeted synthetic disease-modifying antirheumatic drugs (b/tsDMARDs) in patients with inflammatory arthritis (IA). Methods: We searched MEDLINE for all Cochrane reviews and for systematic reviews published since April 2015. RCTs were eligible if they included patients with IA receiving b/tsDMARD, compared with any comparator arm. Harms were evaluated based on number of withdrawals due to adverse events (WDdtAEs), total withdrawals (WDs), serious adverse events (SAEs) and deaths. Data were extracted for 48 trial/patient characteristics and meta-regression analyses were performed to relate the relative risk ratio (RRR) of harms to the trial characteristics. Results: A total of 284 trials (from 245 reviews) with 97 607 patients were included, contributing 490 comparisons for the primary analysis. Overall, the relative risk of WDdtAEs was lower when trials used active comparators (RRR, 0.74 (95% CI 0.58 to 0.94)) and higher when requiring raised inflammatory markers at enrolment (RRR, 1.25 (1.01 to 1.55)). Our meta-regression analyses suggested that trials with eligibility criteria for minimum tender/swollen joint count and maximum disease duration decreased the risk of WDs, while previous b/tsDMARDs use at the time of enrolment increased the risk of SAEs. Conclusions: Most study characteristics do not affect the reported harm measures. However, a trend was observed where trials selecting patients with higher baseline disease activity found a higher risk ratio of WDdtAEs and SAEs, but also a lower risk of WDs, compared with trials not selecting patients with a high disease activity. PROSPERO registration number: CRD42020171124.