A composite approach that includes dropout rates when analyzing efficacy data in clinical trials of antipsychotic medications

Background: Often, outcomes in clinical trials of antipsychotic medications are examined using last observation carried forward (LOCF). One limitation of LOCF and other common approaches is that they overlook the meaning underpinning trial completion and noncompletion. Noncompletion often relates to lack of drug tolerability. Because long-term treatment is often indicated, noncompletion is an important outcome. An alternative approach is to test the composite hypothesis of the difference between (a) completion rates and (b) efficacy of complete cases. Studies to date have not applied this relatively new method. Objective: To illustrate the composite approach, we applied it to a systematic review of studies and compared the results with the reported LOCF analysis. Methods: A systematic search of the Schizophrenia Module of the Cochrane Library and Medline was conducted that identified 127 relevant randomized clinical trials of antipsychotic medications conducted since 1995. Extracted from study reports were the P values of a difference in dropout and the P value of a difference in improvement among complete cases. These P values were combined using standard approaches. Results: We identified 11 trials with 5339 participants that provided the necessary information to adequately apply the composite approach. In 2 trials, (18.2%) in which the LOCF results were not significant, the composite results were significant. Conclusions: The composite approach was more sensitive to change than LOCF and conceptually answers a more relevant question. It is likely that applying the composite approach would change how results of many trials are interpreted.