Patient dropout has been one of major obstacles in efficacy evaluation in longitudinal clinical trials. In some psychiatric therapeutic areas, patient dropout rate frequently reaches 40% or higher. Several statistical methods have been developed for analyzing such data under various assumptions (MCAR, MAR, shared parameter model assumption, assumption that one can correctly model dropout mechanism, etc.) However, it's hard to evaluate if these assumptions are violated in a given study. So the reliability of these methods in analyzing clinical trial data remains unknown and questionable. With the FDA database of all schizophrenia trials, we will explore patient dropout patterns and characteristics. The results will shed some lights on which dropout model assumptions are more plausible and when and how these assumptions are likely violated.