Robustness of sample size re-estimation with interim binary data for double-blind clinical trials

A central question that arises in clinical trials is, how many additional observations, if any, are needed beyond those originally planned? Earlier contributions considered continuous response that assumes a normal distribution. Some treat the portion up to the interim stage as an 'internal pilot study' and thus require unblinding. In this paper, for a certain randomized design with interim binary data for double blind clinical trials, we obtain closed-form expressions for the effective level of significance and operating power at the specified alternative. Also considered is the totally unblinded case. Numerical examples indicate a significant increase in the level of significance in the case of the blinded design which is contrary to what Shih and Zhao (Statist. Med. 16(17) (1997) 1913) have claimed. However, the type I error probability is under control in the unblinded case.