Algorithms for evaluating reference scaled average bioequivalence: power, bias, and consumer risk.

Area of Science:

• Pharmacokinetics and Biopharmaceutics
• Statistical Methods in Drug Development
• Regulatory Science

Background:

• Evaluating bioequivalence for highly variable drug products is complex, with current regulatory methods (e.g., FDA's Hyslop, EMA's expanding limits) having limitations.
• Existing algorithms may result in lower statistical power or higher consumer risk than desired.
• Previous 'Exact' methods for bioequivalence had limited applicability.

Purpose of the Study:

• To propose and evaluate two modified Exact methods for reference scaled average bioequivalence (RSABE).
• To simplify computation and broaden the applicability of Exact methods to any study design.
• To compare the performance of the modified Exact methods against current regulatory approaches.

Main Methods:

• Simulated 3-period and 4-period bioequivalence studies were used for evaluation.
• Four algorithms were compared: Hyslop's (FDA), average bioequivalence with expanding limits (EMA), and two modified Exact methods.
• Performance was assessed based on statistical power and consumer risk at small sample sizes.

Main Results:

• The modified Exact methods demonstrated substantially higher statistical power than Hyslop's algorithm at small sample sizes.
• The Exact methods exhibited lower consumer risk compared to the EMA's approach.
• Higher than 5% consumer risk was observed only with unbalanced designs or additional regulatory demands, similar to Hyslop's algorithm.

Conclusions:

• The improved Exact algorithms offer a favorable alternative to current RSABE procedures.
• These methods are based on bias correction, acknowledging that scaled difference statistics are measured with bias.
• The findings necessitate a revision of the statistical theory and methods for RSABE, particularly for pilot study evaluations.