Cluster analysis of clinical data to identify subtypes within a study population following treatment with a new pentapeptide antidepressant.

Abstract

Cluster analysis was used to evaluate the data from a placebo-controlled, double-blind clinical trial with a new pentapeptide antidepressant (INN 00835) in major depression. The objective of this paper is to examine the effect of separating the study population into homogeneous subgroups (clusters) with relatively similar response to treatment within subgroups, and significantly different response between subgroups. The list of variables for cluster analysis was selected only from the efficacy parameters investigated in the study. Three to six clusters were modelled to obtain the optimal number of clusters, based on a proportional contribution of subjects per cluster, and the maximum statistical difference between clusters. After separation, the variability of response among drug-treated subjects by cluster was attributed to plasma drug concentration. Platelet serotonin uptake, which is a putative biochemical marker of effective treatment of depression, also reproduced the same effect of separation as the initially established cluster variables.