Menu Back to Poster-Presentations-Details

W 47: Testing for Bioequivalence in Higher-Order Crossover Designs: Two-at-a-Time Principle Versus Pooled ANOVA

Poster Presenter

      Pina D'Angelo

      • Senior Director, Scientific Affairs
      • Novum Pharmaceutical Research Services
        United States


The objective of this research is to determine which method of statistical analysis is more appropriate to conclude bioequivalence in higher-order crossover studies.


Empirical data from three-way crossover bioequivalence studies were examined using 1) the two-at-a-time principle and 2) a pooled approach using one ANOVA. Simulations will also be performed to control and compare the type I and II errors resulting from both methods of analyses.


Statistical results for the pharmacokinetic parameters AUC and Cmax were generated using the MIXED procedure in SAS©. The Test/Reference ratios and their associated 90% confidence intervals were estimated using 1) the two-at-a-time principle, where each Test formulation was compared to the Reference formulation in two separate incomplete block design ANOVAs, and 2) the pooled ANOVA where data from all three products were analyzed using one ANOVA. Analysis of empirical data showed that one method can pass bioequivalence criteria whereas the other method can fail bioequivalence criteria. This is mainly as result of 1) artificial increase in sample size using a pooled ANOVA approach to compare each of two Test formulations to a Reference, and 2) non-homogeneity of variances and different point estimates across formulations.


The method of statistical analysis of higher-order crossover studies can affect the bioequivalence conclusions of each formulation being tested. Empirical results show that using a two-at-a-time principle or a pooled ANOVA approach may produce different bioequivalence conclusions for the targeted comparison owing to the influence of the other formulations on those results. Simulations where type I and II errors are controlled will show which method of analysis produces less bias and more accurate bioequivalence conclusions.