Abstract:
Between 1993 and 1996, a total of 452 patients were entered into a randomized trial evaluating eliprodil
(a non-competitive NMDA receptor antagonist) in patients suffering from severe head injury. The primary
efficacy analysis concerned the Glasgow Outcome Score (GOS), six months after randomization. This
outcome was classified into three ordered categories: good recovery; moderate disability, and the worst
category made up by combining severe disability, vegetative state and dead. A sample size calculation was
performed prior to the commencement of the study, using a formula which depends on the anticipated
proportions of patients in the three different outcome categories, the proportional odds assumption and on
the relationship between outcome and prognostic factors such as Glasgow Coma Score at entry. Owing to
uncertainty about the influence of prognostic factors, and about the proportion of patients in the three GOS
categories, a blinded sample size review was planned. This review was performed on the basis of the first 93
patients to respond, and this led to an increase in the sample size from 400 to 450. In this paper the pre-trial
simulations showing that the type I error rate would not be influenced and the power would be preserved
will be presented, and the implementation of the procedure will be described. ( 1998 John Wiley
& Sons, Ltd.
