[The retrospective view of controlled clinical trials: is the subsequent formation of subgroups allowed? (author's transl)].

Additional data appearing in clinical trials can be evaluated only outside the protocol, i.e. retrospectively. An exclusion of patients before randomization is feasible, a withdrawal of patients after randomization is not allowed, but if unavoidable a "central blind review" should be introduced. No loss of patients should occur after the closing of the study. Differences in prognostic factors between groups of patients can be overcome by retrospective stratification. If there are different outcomes, a subsequent significance test at the 1% level is considered as well as a free interpretation.