Superior Warfarin Dosing Using Pharmacogenomics in a Multi-Ethnic Patient Population.

Purpose: Coumadin (Warfarin) is known for its high inter-individual dose requirements and narrow therapeutic index. Common SNPs in the CYP 2C9 and haplotypes in the VKORC-1 have been linked to lower warfarin dose requirements. Our initial study was designed to compare the current pharmacy protocol for warfarin dosing to a multi-variant pharmacogenomic (PGx) dosing algorithm including CYP2C9 and VKORC1 genotype, patient height, weight, age, current INR, target INR, primary indication, and interfering medications.

Method: The strategy developed with our Pharmacy and therapeutics committee has been to genotype all in-patients put on sodium heparin for prospective warfarin sensitivity genotyping without altering current dosing protocols. Patients were genotyped for CYP 2C9 *1, *2 and *3 and for the VKORC-1 (-1639) SNP. The genotype and dose recommendations were reported from our clinical lab. Inclusion criteria were genotype, 4 or more doses of warfarin, concurrent INR and the initial dose selection per the current pharmacy protocol. The final patient therapeutic dose was compared to Pharmacy’s initial dose and to our PGx predicted initial dose using a scatter plot analysis with a trend line fitted to our results.

Results: Of the initial 75 patients, 30 were not transitioned to warfarin while in the hospital and thus were excluded from the study analysis. 25 of the remaining 45 patients met all study inclusion criteria. The pharmacy dosing strategy when compared to the final therapeutic dose showed a correlation of 0.068 (1.4% R² value). However, our pharmacogenomic dosing strategy showed a correlation of 0.44 (44.5% R² value). Per the pharmacy protocol 72% of patients (18/25) were given initial doses 2 or more mg above or below the final therapeutic dose. The pharmacogenomic algorithm predicted the final therapeutic dose within 2 mg for 76% of patients tested. Overdosing by 2 mg or more occurred in 52% (13/25) with the current dosing method. Use of the Pharmacogenomic algorithm would have resulted in only 8% (2/25) of patients being overdosed by more than 2 mg.

Conclusion: The combination of demographic and genetic information accounted for 44% of the inter-patient variation in warfarin dosing. This demonstrates a significant improvement over the current method which accounted for only 6.8% of inter-patient warfarin dosing variation. Under the current dosing strategy only 24% of patients receive initial doses within 2 mg of their final therapeutic dose by 2 mg or less compared to 76% using the PGx algorithm. With the current dosing, 52% of patients were overdosed by 2 mg or more compared to 8% using the PGx algorithm. The PGx algorithm would also have reduced the number of overdosed patients above 2 mg by 85%. Comparison of the genotype, daily dose and INR reveals that patients harboring 2 or more variant alleles have more fluctuations in INR and require more dose adjustments than patients with 0 or 1. By identifying these patients and initiating therapy with more accurate dosing we should reduce the fluctuations in INR and number of dose adjustments. Our data suggests adopting PGx algorithm warfarin dosing will lead to a reduction in dose adjustments, a reduction in time to therapeutic INR, and identify difficult to dose patients. Future open label prospective studies are needed to demonstrate and quantify the clinical effectiveness of this predictive algorithm.