Design considerations for handling dropouts in anti-depressant drug trials

Abstract

BackgroundIn clinical trials, statistical analysis often requires certain assumptions about missing data for a valid statistical inference. If the dropout rate is high, a wrong assumption about the missing data may compromise the validity of statistical inferences. PurposeTo mitigate the high dropout rates commonly observed in psychiatry clinical trials, we consider two design approaches for short-term controlled trials submitted in support of marketing applications for drug products for the major depressive disorder (MDD) indication: (1) shortening the trial duration and (2) treating time to treatment discontinuation as an alternative primary efficacy endpoint. MethodsSubject-level efficacy data from 45 trials for drugs approved for an MDD indication between 1997 and 2014 were collected. We analyzed change from baseline in Hamilton Depression Rating Scale (HAMD-17) total score using the mixed model repeated measures approach. We compared the least squares means and the 95% confidence intervals of the treatment effect among three different trial durations, 4, 6, and 8weeks. We considered two definitions of discontinuation: (i) all-cause discontinuation, (ii) discontinuation due to lack of efficacy. We compared the two-sided log-rank p-values with the p-values from the protocol-specified primary analysis. ConclusionsOur findings suggest that MDD trials in the acute setting may be shortened to 6weeks provided that the treatment difference between drug and placebo on HAMD-17 total score reaches approximately 2units at Week 6. However, our exploratory analyses of available data do not support the use of time to treatment discontinuation as an alternative primary efficacy endpoint.