Sample Size Determination from A Pilot Bioequivalence study for A Future Pivotal Bioequivalence Study: A SAS Procedure

The purpose of research study develops an automated computer statistical procedure to determine sample size from a pilot bioequivalence study for a future pivotal bioequivalence study. The sample size is critical in a pivotal bioavailability/ bioequivalence study, and was often estimated from a pilot study. A comprehensive review of current literature reveals that the software for sample size determination using data from a pilot study has not been reported (D. Inouye, Bull. Ecol. Soc. Am., 1997). A series computer code (SAS) which contains four macro-driven programs has been developed to perform the following steps: 1) The statistical analysis of a two-period crossover pilot bioequivalence study according to the FDA guidance ‘Statistical procedures for bioequivalence studies using a standard twotreatment crossover design’ and the paper by Sauter et al (Int. J. Clin. Pharmacol. Ther. Toxicol. 30:23356(1992); 2) Consider alpha=0.05 and the bioequivalence range of 90% confidence intervals (80%, 125%), the sample size was determined for a future pivotal BA/BE study according to the table generated by Diletti et al (int. J. Clin. Pharmacol. Ther. Toxicol. 29:1-8 (1991); 3) Using the estimated ratio of geometric means (μΑ/μΒ), MSS, MSE from step 1) and the sample size n from step 2), a new set of data were generated based on the error model described by Jones and Kenward (Design and Analysis of Cross-Over Trials, pp16-88, Chapman &Hall, New York, 1989); 4) The same statistical analysis procedure in step 1) was performed on the new data set to confirm whether so calculated sample size would warrant that the confidence intervals of ratio of geometric means falls within 80% 125%. The results of this research study showed that the sample size was determined from the developed procedure readily. The results also showed that such a sample size warranted that the confidence intervals of ratio of geometric means falls within 80% 125% in a simulated future pivotal bioequivalence trial. It is concluded that the applications of this computer procedure with real data from pilot bioequivalence studies were successfully demonstrated. The outcomes of this study has significant implications in clinical trials.