INTERIM ANALYSES IN CLINICAL TRIALS

 Dr. David Bristol, Schering-Plough Research Institute


Interim analyses can be used to terminate a comparative
clinical trial if outstanding between-treatment results
are observed during the duration of the trial. However,
in most situations, such results are "unlikely". Repeated
between-treatment comparisons provides multiple opportunities
to reject the null hypothesis; this multiplicity must be
incorporated to maintain the overall Type I Error rate. 
Examination of the data prior to the planned completion can 
provide useful information regarding other aspects of the
trial, such as validity of assumptions and within-treatment
assessments. Do such examinations influence the Type I Error
rate? Interim analyses in equivalence studies and in studies
with multiple between-treatment comparisons introduce
additional complexities. Techniques for adjusting for interim
analyses, with focus on "flexible" designs in complicated
trials, will be discussed.