Surrogate outcomes are associated with low methodological quality of studies of rheumatoid arthritis treated with antitumour necrosis factor agents: a systematic review.

Background: Surrogate endpoints may be used as substitutes for, but often do not predict clinically relevant events. Objective To assess the methodological quality of articles that present their conclusions based on clinically relevant or surrogate outcomes in a systematic review of randomised trials and cohort studies of patients with rheumatoid arthritis treated with antitumour necrosis factor (TNF) agents.
Methods: PubMed, Embase and Cochrane databases were searched. The Jadad score, the percentage of Consolidated Standards Of Reporting Trials (CONSORT) statement items adequately reported and levels-of-evidence (Center for Evidence-based Medicine, Oxford) were used in a descriptive synthesis.
Results: Among 88 articles appraised, 27 had surrogate endpoints, mainly radiographic, and 44 were duplicate publications; 74% of articles with surrogate and 39% of articles with clinical endpoints (p=0.006). Fewer articles with surrogate endpoints represented a high level of evidence (Level 1b, 33% vs 62%, p=0.037) and the mean percentage of CONSORT statement items met was also lower for articles with surrogate endpoints (62.5 vs 70.7, p=0.026). Although fewer articles with surrogate endpoints were randomised trials (63% vs 74%, p=0.307) and articles with surrogate endpoints had lower Jadad scores (3.0 vs 3.2, p=0.538), these differences were not statistically significant.
Conclusion: Studies of anti-TNF agents that report surrogate outcomes are of lesser methodological quality. As such, inclusion of such studies in evidence syntheses may bias results.