Repeated significance tests for clinical trials with a fixed number of patients and variable follow-up

Abstract

Group-sequential tests may be applied to trials in which the number of patients is fixed, a response variable is measured for each patient at successive follow-up visits, and the accumulated responses are compared across treatment groups. The standard theory is inapplicable because the increments in the accumulated responses are no longer independent. An adjustment can be made to allow for the ratio of between-patient to within-patient variance and for possible first-order autocorrelation. The method is illustrated by reference to a dental trial.