Warfarin anticoagulation is complicated by the highly variable inter-individual response. Approximately 50% of the dose variability arises from clinical factors and variants in two genes, CYP2C9 (*2 and *3 variants) and VKORC1 -1173 C > T. We tested variants in five additional genes (EPHX1, PROC, APOE, CYP4F2, CALU and a new variant in VKORC1 in an attempt to further reduce the variability in predicted stable warfarin dose. Consecutive consenting outpatients requiring anticoagulation on stable warfarin dose (target INR 2-3) were genotyped; the association of SNP genotypes with stable warfarin dose was evaluated using the test of linear contrasts in analysis of variance (ANOVA). Study participants were 71 ± 13 years, 53% female, 85 ± 23 kg, body mass index 29 ± 7 kg/m(2). Genotypes were in Hardy-Weinberg equilibrium with the exception of VKORC1 -1639. Weekly stable dosages were 31.7 ± 13.9 mg/week; median: 30 mg/week, range: 11-70 mg/week. Significant associations with dose were seen for VKORC1 -1639 (P < 0.001), CYP2C9*2 (P = 0.005) and *3 (P = 0.003), the CYP4F2 SNP (P-trend = 0.00037), and VKORC1 3730 (p-trend = 0.042). In linear regression, age, sex, weight, and CYP2C9 *2 and *3 and VKORC1-1639 genotype explained 42% of variance. The addition of CYP4F2 genotype to the regression model increased the degree of variance explained to 47%. Addition of VKORC1 SNP -1639 to a model eliminated the association of VKORC1 3730 with warfarin dose (P-trend = 0.74), but -1639 remained highly significant. No impact on dose was observed for the other tested genetic variants.