Multiple outcomes and analyses in clinical trials create challenges for interpretation and research synthesis

Objective: To identify variations in outcomes and results across reports of randomized clinical trials (RCTs).

Study design and setting: Eligible RCTs examined gabapentin for neuropathic pain and quetiapine for bipolar depression, reported in public (e.g., journal articles) and nonpublic (e.g., clinical study reports) sources by 2015. We prespecified outcome domains. From each source, we collected "outcomes" (i.e., domain, measure, metric, method of aggregation, and time point); "treatment effect" (i.e., outcome plus the methods of analysis [e.g., how missing data were handled]); and results (i.e., numerical contrasts of treatment and comparison groups). We assessed whether results included sufficient information for meta-analysis.

Results: We found 21 gabapentin (68 public, 6 nonpublic reports) and seven quetiapine RCTs (46 public, 4 nonpublic reports). For four (gabapentin) and seven (quetiapine) prespecified outcome domains, RCTs reported 214 and 81 outcomes by varying four elements. RCTs assessed 605 and 188 treatment effects by varying the analysis of those outcomes. RCTs reported 1,230 and 661 meta-analyzable results, 305 (25%) and 109 (16%) in public reports.

Conclusion: RCTs included hundreds of outcomes and results; a small proportion were in public reports. Trialists and meta-analysts may cherry-pick what they report from multiple sources of RCT information.