Messy Data in PK Bioequivalence Studies
*Charles Bon, Biostudy Solutions, LLC  Lindsey A. Katz, Biostudy Solutions, LLC 

Keywords: Bioequivalence, Pharmacokinetics, Replicated Crossover Study, Highly Variable Drugs, Reference-Replicated, Fully-Replicated

The basic statistical methods for evaluation of Bioequivalence using pharmacokinetic data will be reviewed with emphasis on replicated design studies. The use of 3-period, two-treatment, reference-replicated crossover studies has become routine in bioequivalence testing of highly variable drugs. The 4-period, two-treatment, fully-replicated design is a close second, with some recent FDA guidance documents requiring this design. The FDA has given industry SAS code in some guidance documents for replicated design BE evaluations. Under ideal circumstances, this makes the statistical analyses fairly straight forward. Unfortunately, such ideal circumstances seldom occur in real studies and the Biostatistician is often faced with messy data due to it being unbalanced, study conduct inconsistent with the planned design, or convergence problems. Several examples of messy PK data will be discussed, including situations involving subject dropouts and inestimable PK parameters, and having multiple, instead of single, dosing groups when Scaled Average BE evaluation was required. The discussion will cover the changes made to standard statistical analysis methods that enabled bioequivalence assessments in these messy data situations. The results of simulations evaluating the validity of these changes will also be presented.