Sample Size Re-estimation In Two-stage Adaptive Designs With Multiple Endpoints

We propose a general method for two-stage sample size re-estimation in randomized comparative studies with multiple endpoints. The new method employs intersection-union tests to obtain an asymptoticα-level test and re-estimates the sample size using updated information of the nuisance parameters from an internal pilot study. The method is easy to implement and also modified for blinded and unbalanced cases as they arise frequently in practice. Numerical studies show that the proposed method provides good protection for type-Ⅰerror and achieves the desired power. The expected overall sample size appears to be close to the sample size required in the corresponding single stage design assuming the nuisance parameters are known.