BACKGROUND: Warfarin sodium (warfarin) is commonly prescribed in surgical practice. Warfarin use is complicated by an unpredictable dose response that may be due in part to genetically determined differences in metabolic capacity. To better understand the interaction between genotype and response to warfarin therapy, we determined the frequency and functional effects of polymorphisms of the predominant cytochrome P450 subfamily responsible for warfarin metabolism (eg, CYP2C9) in an ethnically defined U.S. patient population. DESIGN: Patients requiring chronic anticoagulation with warfarin sodium (warfarin) were recruited over an 11-month period (June 1999 through May 2000) from the inpatient and outpatient divisions of a tertiary care medical center in this prospective observational study. Clinical and demographic information was collected and CYP2C9 genotype was determined. RESULTS: One hundred fifty-three patients receiving warfarin therapy for at least four weeks and comprising two ethnic groups [33 African Americans (22%) and 120 Caucasians (78%)] were genotyped. The mean weekly warfarin dose (± SEM) for all patients [36.9 (± 1.5) mg] was not influenced by gender [85 males (56%), 68 females (44%)] or ethnicity (p>0.05 for both), but was significantly affected by age (p = 0.006 for weight adjusted warfarin dose). The frequencies of CYP polymorphisms were as follows: 2C9*2 (24/153) 15.7%, 2C9*3 (23/153) 15.0%. There were no gender differences in polymorphism frequency (CYP2C9*2 frequency = (13/85) 15.3% in males, (12/68) 17.6% in females, p= 0.74; CYP2C9*3 frequency = (15/85) 17.6% in males and (8/68) 11.8% in females, p = 0.38). CYP polymorphisms were much less common in African Americans than Caucasians [(5/33) 15.2% versus (47/120) 39.2%, respectively p = 0.05)]. Patients with CYP polymorphisms (2C9*2, 2C9*3) had significantly lower warfarin doses compared to patients with wild-type genotypes [30.6 (±2.5) mg versus 40.1 (± 1.7) mg, p = 0.0021]. Stepwise backward regression analysis suggested a moderate ability to predict warfarin dose based on CYP genotype (r2 = 0.26), p < 0.01). CONCLUSIONS: CYP2C9 polymorphisms are common, associated with significant reductions in warfarin dose, and partly account for interpatient variability in warfarin sensitivity. As interactions between genetic factors and other variables that influence warfarin effect are more completely understood, CYP analysis may prove a useful adjunct for increasing the safety and efficacy of this agent.