Checking distributional assumptions for pharmacokinetic summary statistics based on simulations with compartmental...

Area of Science:

• Pharmacokinetics and Drug Development
• Statistical Modeling in Pharmaceutical Sciences

Background:

• Bioequivalence (BE) studies are critical for generic drug approval, commonly using two-period, two-treatment crossover designs.
• Key pharmacokinetic parameters like area under the concentration-time curve (AUC) and maximum concentration (Cmax) are analyzed in BE studies.
• Normality assumptions for log-transformed AUC and Cmax are frequently made in BE evaluations without robust evidence.

Purpose of the Study:

• To investigate the distributional normality of log(AUC) and log(Cmax) in pharmacokinetic crossover studies.
• To assess the impact of pharmacokinetic parameter distributions and measurement error structures on the normality of these response variables.
• To evaluate deviations from normality under different compartmental models and error distributions.

Main Methods:

• Simulated concentration-time profiles using two-stage pharmacokinetic models across a range of parameters.
• Analyzed the distribution of log(AUC) and log(Cmax) from simulated data.
• Conducted sensitivity analyses to examine the effects of non-normal error distributions and various compartmental models.

Main Results:

• Simulations indicated that log(AUC) exhibits heavy tails and log(Cmax) is skewed under reasonable parameter distributions.
• Deviations from normality were observed for standardized log(AUC) and log(Cmax) when pharmacokinetic model errors or compartmental models varied.
• The findings challenge the routine assumption of normality for these key BE metrics.

Conclusions:

• The normality assumption for log(AUC) and log(Cmax) in bioequivalence studies is often violated.
• Pharmacokinetic parameters and model structures significantly influence the distribution of these variables.
• Further research and potentially alternative statistical methods are warranted for robust BE evaluation.