[May results from randomized clinical trials be generalized? An example from psychopharmacology]

Exemplified by a randomized trial on antimanic treatment, this paper addresses the question whether selection of patients for drug trials may limit the applicability of study results from the randomized patients to a wider population. During a two-year period, all consecutively admitted patients from a defined catchment area were screened for fulfilling the inclusion criteria concerning age, diagnosis and severity of illness. The subsequently excluded subgroups of patients were compared with the randomized patients by means of multivariate data analysis. One hundred and sixty four patients met the inclusion criteria. However, after exclusion for various reasons, only 27 (17%) patients remained for randomization. The randomized patients and the excluded patients differed substantially. The generalizability of trial results is limited. Reports of randomized drug trials should carefully describe the screening procedure for inclusion, and, when possible, present relevant comparisons between the randomized patients and the various subgroups of patients meeting the exclusion criteria.