Controlling type I error in the reference-scaled bioequivalence evaluation of highly variable drugs

Area of Science:

• Pharmacokinetics and Drug Development
• Biostatistics
• Regulatory Science

Background:

• Reference-scaled average bioequivalence (RSABE) methods are used for highly variable drugs.
• Existing RSABE methods exhibit type I error control issues at the transition point between constant and scaled bioequivalence limits.
• These issues stem from the maximum probability of type I error occurring at this specific switching variability value.

Purpose of the Study:

• To explore adjustments to the European Medicines Agency (EMA) and U.S. Food and Drug Administration (FDA) regulatory RSABE approaches.
• To evaluate a potential improvement to the original EMA method, termed HoweEMA.
• To develop statistically correct RSABE procedures with controlled type I error rates.

Main Methods:

• Linear scaling of bioequivalence limits based on reference formulation within-subject variability.
• Analysis of type I error probability across different variability values.
• Adjustment of significance levels to ensure type I error remains below nominal levels.
• Evaluation of adjusted EMA, FDA, and HoweEMA methods.

Main Results:

• The proposed adjustments result in RSABE methods that are completely correct regarding type I error probability.
• The probability of type I error is controlled below the nominal significance level across all variability values.
• Potential power loss associated with these adjustments is generally small.
• Power loss becomes negligible at moderately large sample sizes, which are common in real-world studies.

Conclusions:

• Adjusted RSABE methods provide statistically sound bioequivalence assessments for highly variable drugs.
• These corrected methods address the type I error control problems of existing RSABE approaches.
• The adjusted methods offer a practical solution with minimal power reduction, ensuring reliable regulatory decisions.