The use of mechanistic DM-PK-PD modelling to assess the power of pharmacogenetic studies -: CYP2C9 and warfarin as an example

Aim To assess the power of in vivo studies needed to discern the effect of genotype on pharmacokinetics (PK) and pharmacodynamics (PD) using CYP2C9 and (S)-warfarin as an example. Methods Information on the in vitro metabolism of (S)-warfarin and genetic variation in CYP2C9 was incorporated into a mechanistic population-based PK-PD model. The influence of study design on the ability to detect significant differences in PK (AUC(0-12 h)) and PD (AUEC(0-12 h) INR) between CYP2C9 genotypes was investigated. Results A study size of 90 (based on the natural abundance of genotypes and uniform dosage) was required to achieve 80% power to discriminate the PK of (S)-warfarin between wild type (*1/*1) and the combination of all other genotypes. About 250 subjects were needed to detect a difference in anticoagulant response. The power to detect differences between specific genotypes was much lower. Analysis of experimental comparisons of the PK or PD between wild-type and other individual genotypes indicated that only 21% of cases (20 of 95 comparisons within 11 PD and four PK-PD studies) reported statistically significant differences. This was similar to the percentage expected from our simulations (20%, chi(2) test, P = 0.80). Simulations of studies enriched with specific genotypes indicated that only three and five subjects were required to detect differences in PK and PD between wild type and the *3/*3 genotype, respectively. Conclusions The utilization of prior information (including in vivo enzymology) in clinical trial simulations can guide the design of subsequent in vivo studies of the impact of genetic polymorphisms, and may help to avoid costly, inconclusive outcomes.