Bootstrap Calibrated Tests for Average Bioequivalence and Scaled Average Bioequivalence

For the assessment of average bioequivalence, the two one-sided t-test (TOST) is widely used. The test is known to be very conservative for large values of the variance, suggesting that the TOST is likely to have type I error probabilities close to the nominal level if the test is implemented using a significance level larger than the nominal level. In this chapter, a bootstrap calibration is proposed to estimate the required significance level. Numerical results show that a bootstrap calibrated TOST provides better type I error probabilities, and hence increased power. The test may be an attractive alternative to the scaled average bioequivalence test, especially since the latter cannot be implemented using data from a 22 crossover design, whereas such data are adequate for carrying out the bootstrap calibrated TOST. The bootstrap calibrated TOST is easy to understand and implement, unlike some of the theoretical improvements of the TOST available in the literature. The bootstrap calibrated TOST can be adopted for the assessment of multivariate average bioequivalence. Finally, the bootstrap calibration methodology is applied for obtaining an accurate test procedure for the assessment of scaled average bioequivalence for highly variable drugs with data generated using a four-period crossover design.