Statistical strategies for randomized controlled clinical trials to detect differential onset of action

Designing a randomized controlled clinical trial to detect differential onset of action of treatments for mood and anxiety disorders poses several challenges. The definition of onset of action must embrace both the degree of symptom reduction and the duration of reduction over time. Perhaps the most difficult aspect of the definition is the classification of subjects with symptom reduction that is not sustained. The scheduling of assessments in such a trial is critical. Unlike standard RCTs for mood or anxiety disorders, weekly or biweekly assessments could prove to be insensitive to group differences. The data analytic procedure must focus on change over time in order to detect differential onset, yet account for problems of multiplicity and attrition. Repeated statistical tests at each successive assessment time might appear to be intuitively appealing, but they require multiplicity adjustments. The data analytic approach must also be flexible enough to incorporate available data from subjects who fail to complete some assessments. The applicability of mixed-effects models and the cure model to challenges in the study of onset of action is described.