Influence analysis on crossover design experiment in bioequivalence studies

Increased Patient's Risk Associated with the Canadian Bioequivalence Guidance Due to Outlier Removal

BACKGROUND

The Canadian guideline on bioequivalence allows identification of outliers through for example studentized residuals, and it explicitly accepts exclusion of subject data when their studentized residuals reach a certain magnitude. The guideline also requires that the type I errors (patient's risk, chance of declaring bioequivalence for a bioinequivalent product) be maintained at the 5% level. This manuscript investigates if the outlier removal procedure increases type I errors when alpha of 5% is used for construction of the confidence interval.

METHODS

Numerical simulation is used.

RESULTS

Patient's risk may be inflated beyond 5% when the traditional 90% confidence interval is constructed. Maximum type I errors observed here are slightly above 7%. To circumvent the issue sponsors can either to abstain from outlier removal or find ways to adjust alpha levels downwards if the patient's risk is to be preserved at the 5% level.

CONCLUSION

This manuscript is the first to create awareness of a potential inflation of the type I error when 90% confidence intervals are constructed as per Canadian guidance. The issue may not just affect patients in Canada but could also affect patients in the USA since the current rules allow American patients to import a 3 month supply of drugs for personal use and because the central administration is actively pursuing plans to "(…) facilitate the importation of certain prescription drugs that are approved in Canada".