The credibility of subgroup analyses reported in stroke trials is low: A systematic review

Background: Subgroup analyses are widely used to evaluate the heterogeneity of treatment effects in randomized clinical trials. However, there is a limited investigation of the quality of prespecified and reported subgroup analyses in stroke trials. This study evaluated the credibility of subgroup analyses in stroke trials. Methods and analysis: We searched Medline/PubMed, Embase, the Cochrane Central Register of Controlled Trials, and the Web of Science from inception to 24 March 2021. Three reviewers screened, extracted, and analyzed the data from the publications. Primary publications of stroke trials that reported at least one subgroup effect and had published corresponding study protocols were included. The Instrument for Assessing the Credibility of Effect Modification Analyses (ICEMAN) was used to examine the quality of the subgroup effects reported, with each subgroup effect assigned a credibility rating ranging from very low to high. Subgroup effects with two or more “definitely no” responses received a low credibility rating. The risk of bias was assessed using the Cochrane Risk-of-Bias tool for randomized trials version 2. Results: Seventy-four articles met the inclusion criteria and reported a combined total of 647 subgroup effects. The median sample size was 1264 (interquartile range (IQR): 380–3876), and the median number of subgroups prespecified in the protocol was 6 (IQR: 2–10). Sixty-one (82%) studies used the univariate test of interaction. Of the total 647 subgroup effects reported in these studies, 319 (49%) were reported in acute stroke trials, while 423 (65%) had low credibility. Conclusion: The quality of subgroup analysis reporting in stroke trials remains poor. More effort is needed to train trialists on the best methods for designing and performing subgroup analyses, and how to report the results.