The Evaluation of Multiple Clinical Endpoints, with Application to Asthma

The situation where four primary endpoints are classified into two groups is considered, with the clinical objective to show at least one positive effect in each group. This multiple endpoint scenario may be appropriate in clinical asthma trials, with the two groups being pulmonary function variables and patient recorded outcomes. The statistical motivation for the classification is that the within-group correlation is usually higher than the between-group correlation. Three methods for the evaluation of this multiple test problem are proposed. A combination of the intersection-union principle with the Simes method leads to a powerful level-α test. Two clinical studies are used to illustrate the methods. Strategies are discussed for the incorporation of one further endpoint, that is, the premature drop-out rate due to lack of efficacy.