Systematic reviewers neglect bias that results from trials stopped early for benefit

Abstract

Objective: To examine how authors of systematic reviews that include randomized clinical trials (RCTs) that are stopped early for benefit (truncated RCTs-tRCTs) address the potential for overestimation of treatment effects and to determine the weight of the tRCTs on pooled results.

Study design and setting: We searched the Cochrane Library and MEDLINE and evaluated systematic reviews that include at least one tRCT. We documented approaches that authors used to address potential overestimates of treatment effect introduced by including tRCTs. We assessed the impact of tRCTs in meta-analyses on the outcomes that led to their early termination.

Results: Of 96 systematic reviews that included at least one tRCT, 44 (46%) included >1 tRCT, 68 (71%) did not mention truncation at all, and 2 (2%) documented early stopping for benefit as a criterion for methodological quality. Of 47 meta-analyses in which authors reported, or we could calculate the contribution of the tRCTs to the pooled result, the tRCTs contributed more than 40% of the weight in 16/47 (34%).

Conclusion: Most systematic reviews and meta-analyses including tRCTs fail to consider the possible overestimates of effect that may result from early stopping for benefit. We recommend safeguards that address this possibility.