Objective: Our objective was to investigate population differences in the metabolic activity of cytochrome P450 (CYP) 2C9 between genotypically matched Caucasian and Japanese patients by using the unbound oral clearance of S-warfarin as an in vivo phenotypic trait measure. Methods: Ninety Japanese and 47 Caucasian patients receiving maintenance warfarin therapy were studied. Steadystate plasma unbound concentrations of S-warfarin were measured by a chiral HPLC method coupled with an ultrafiltration technique, and unbound oral clearance for S-warfarin was estimated. By combining plasma unbound concentrations of S-warfarin with the urinary excretion rates of S-7-hydroxywarfarin, the formation clearance of S-7-hydroxywarfarin was also determined. Genotyping of CYP2C9 was performed for 6 distinct alleles (CYP2C9*1, CYP2C9*2, CYP2C9*3, CYP2C9*4, CYP2C9*5, and a T/C transition in intron 2). Results: The frequency distribution of unbound oral clearance for S-warfarin obtained from Japanese patients was shifted toward higher values as compared with that in Caucasian patients. Japanese patients had lower allelic frequencies for the 5 variants than Caucasian patients. When interpopulation comparisons of CYP2C9 activity were made for genotype-matched subjects, Japanese patients with the homozygous CTP2C9*1 (wild-type) genotype (n = 85) had significantly (P < .01) greater median values for unbound oral clearance and formation clearance than Caucasian patients with the corresponding genotype (n = 26), 10.4 mL × min-1 × kg-1 versus 4.25 mL × min-1 × kg-1 and 0.015 mL × min-1 × kg-1 versus 0.010 mL × min-1 × kg-1, respectively. In addition, Japanese patients heterozygous for the CYP2C9*3 genotype (n = 4) showed a significantly (P < .05) reduced unbound oral clearance for S-warfarin, by 63%, as compared with Japanese patients possessing the homozygous CYP2C9*1 genotype. By contrast, in Caucasian patients, no significant differences were observed in this parameter between CYP2C9*1 homozygous subjects and those with heterozygous CYP2C9*2 or CYP2C9*3 genotypes. Conclusions: These findings indicate that population differences in the frequencies of known variant CYP2C9 alleles account only in part for the variability observed in in vivo CYP2C9 activity in different populations. In addition, a gene-dose effect of defective CYP2C9 alleles on the in vivo CYP2C9 activity is evident in Japanese patients but not in Caucasian patients. Further studies are required to identify currently unknown factor(s) (eg, transcriptional regulation) responsible for the large intrapopulation and interpopulation variability in CYP2C9 activity.