FAILURE OF STATISTICAL METHODS TO PROVE BIOEQUIVALENCE OF TWO MELOXICAM BIOEQUIVALENT FORMULATIONS.
II. NON-PARAMETRIC METHODS
ROXANA SANDULOVICI1, ANCA VATASESCU2, FLORIN ENACHE3, CONSTANTIN MIRCIOIU2*
1Biopharmacy & Pharmacol Res S.A., Bucharest     
2Carol Davila University of Medicine & Pharmacy, Bucharest
3Institute of Statistics, Romanian Academy, Bucharest
*corresponding author: constantin.mircioiu@yahoo.com
Download Full Article

Abstract:

Applying the “official” parametric methods to analyze the results of a clinical bioequivalence (BE) study concerning two suppository formulations containing meloxicam as active substance, the bioequivalence couldn’t be proved, tested drug appearing to have a greater bioavailability than the reference drug. Since the reference drug presented an important intervariability and the tested drug proved a greater bioavailability than the reference drug, it was considered that the products could be bioequivalent, but the official statistical test failed to prove this. Following mainly the high variability of reference drug and a distribution of plasma levels of reference drug far from normality, the failure was thought as a consequence of the application of statistical parametric tests beyond the field of their validity.
Statistical models for building non-parametric confidence intervals for the ratios of means of pharmacokinetic parameters were less restrictive that in the case of parametric analysis. In a first approximation there were neglected the sequence effects and further, both sequence and period effects. The results lead to the same failure of proving BE, like parametric methods. The conclusion was that non-parametric methods lead to the same conclusion concerning BE but are more efficient in rejecting the effects of outliers. Suspicion remains that even non-parametric methods are not efficient in correcting the bias induced by partition of data in some different classes as in the case of pharmacokinetic parameters of the reference drug.

Top