Comparison of parametric and bootstrap method in bioequivalence test.

Area of Science:

• Pharmacokinetics and Drug Development
• Biostatistics
• Regulatory Science

Background:

• Bioequivalence (BE) testing relies on parametric 90% confidence intervals (CIs) for AUC and Cmax ratios.
• The validity of parametric CIs hinges on the assumption of normally distributed formulation effects in log-transformed data.

Purpose of the Study:

• To compare parametric CIs with nonparametric bootstrap CIs for bioequivalence testing.
• To evaluate the impact of non-normally distributed data on the reliability of parametric CIs.

Main Methods:

• Utilized 3 archived datasets for AUC and Cmax values.
• Performed 1,000 bootstrap resamples for each dataset to generate nonparametric CIs.
• Compared bootstrap nonparametric 90% CIs with original parametric 90% CIs using SAS software.

Main Results:

• Parametric and nonparametric CIs showed slight differences.
• While most resampled data approximated normal distribution, 2 of 3 log (AUC) datasets deviated significantly.
• Bias-corrected and accelerated (BCa) CIs shifted outside parametric CIs in non-normally distributed log (AUC) data.

Conclusions:

• The standard 80-125% rule, based on parametric CIs, may be unreliable when formulation effects are not normally distributed.
• Nonparametric confidence intervals offer a more robust alternative for bioequivalence assessment in cases of non-normal data distribution.