Subgroup Analysis and Covariate Adjustment in Randomized Clinical Trials of Traumatic Brain Injury: A Systematic Review

OBJECTIVE: Few randomized clinical trials (RCTs) in the field of traumatic brain injury (TBI) have shown a significant treatment benefit. We critically reviewed the use of two types of secondary analyses, covariate adjustment and subgroup analysis, which are common in TBI trials. METHODS: We performed a systematic review of therapeutic phase III RCTs, including adult patients with acute, moderate-to-severe TBI. Glasgow Outcome Scale (GOS) at > or =3 months as outcome, and > or =50 patients per arm were required. We compared the actual reporting of covariate adjustment and subgroup analyses with the Consolidated Standards of Reporting Trials (CONSORT) recommendations. Likewise, we reviewed six protocols of large multicenter RCTs and compared planned and reported subgroups. RESULTS: We identified 18 RCTs (n = 6439). Sixteen trials used GOS at 6 months as outcome. Five RCTs reported covariate adjustment. The number of covariates was limited (< or =5), most frequently including age. Many covariates were outcome predictors. Four RCTs reported only adjusted treatment effects as the main efficacy parameter. Eleven RCTs reported subgroup analyses. Several subgroup factors (< or =7, mainly outcome predictors) and outcomes (< or =4) were included. The highest total number of subgroups was 15, and only three RCTs completely pre-specified subgroups. Notably, 10 of 11 RCTs performed inappropriate separate subgroup analyses. Of 11 RCTs, 5 gave subgroups the same emphasis as the overall effect. Reported subgroup analyses were insufficiently described and clearly differed from those planned in the protocol. CONCLUSION: The reported covariate adjustment and subgroup analyses from TBI trials had several methodological shortcomings. Appropriate performance and reporting of covariate adjustment and subgroup analysis should be considerably improved in future TBI trials because interpretation of treatment benefits may be misleading otherwise.