An analysis performed before study completion, requiring prespecification and control of error rates to avoid biased conclusions.