A bivariate confidence interval approach to calculate drug interaction and bioequivalence study power

Sample size for drug‐drug interaction (DDI) and bioequivalence (BE) studies is usually calculated based on AUC and Cmax as independent parameters. The default acceptance regions using the confidence interval (CI) approach are typically predefined dependent on variability of each parameter. Clinical DDI experience shows that changes in AUC and Cmax are highly correlated regardless of effect size. This may impact the power to establish no‐effect (equivalence) for both parameters. Purpose. To evaluate the probability (P) of declaring no‐effect for select variability, effect and sample size scenarios, assuming a correlation (r) range from 0.3 to 0.9. Methods. P of different scenarios to declare no‐effect was calculated using univariate and bivariate CI approaches for simulated datasets using SAS ® . Both approaches were applied to actual clinical data for comparison. Results. P appears to decrease with increasing r regardless of sample size. When no‐effect was present, the P of meeting the 0.8–1.25 limits for declaring no‐effect for both parameters was lower for the bivariate approach. When an effect was present, P for declaring no‐effect was higher for the univariate approach. Conclusions. The bivariate CI approach with a consistent 0.75–1.33 acceptance region is recommended for evaluating DDI and BE studies for both high and low variability drugs, especially when equivalence is expected. The relationship between P and r is dependent on relative variance between Cmax and AUC. Clinical Pharmacology & Therapeutics (2004) 75 , P91–P91; doi: 10.1016/j.clpt.2003.11.349