Curtailed binomial sampling procedures for clinical trials with paired data.

The use of curtailed sampling procedures in designing clinical trials with a dichotomous response and paired data is reviewed. Tables of expected sample size savings and a simple to evaluate asymptotic formula for the expected sample size savings are presented where the sample size savings is calculated relative to the fixed sample size test. Comparisons are made among fixed sample size, curtailed, and repeated significance hypothesis testing procedures. In view of the substantial savings in sample size available under certain conditions and the accuracy of the approximating formula, it is recommended that curtailed procedures be given serious consideration in the design of clinical trials.