Statistical handling of drop‐outs in longitudinal clinical trials

This paper considers the statistical complexities that arise due to outcome related drop-outs in longitudinal clinical trials of the randomized parallel groups design with fixed assessment times and an explanatory aim. The shortcomings of currently popular methods of coping with the problem of drop-outs are discussed. It is proposed that progress can be made by applying the modern methodology that was primarily developed for sample surveys with non-response and for observational studies. A practical application using the Hamilton Rating Scale for Depression is presented.