1. Home>
2. Specialties>
3. Oncology and Hematology>
4. Summary and Comment

Pharmacogenetics-Based Warfarin Dosing

A genotype-guided approach improved dose prediction for patients who required either high or low doses of warfarin.

Warfarin is an effective and inexpensive anticoagulant, but inappropriate dosing causes serious adverse effects. Whereas roughly half of all patients who receive a fixed dose of warfarin (5 mg daily) have prothrombin times in the therapeutic range (international normalized ratio [INR], 2–3), this dose is too high or too low for many others and can result in fatal bleeding or thrombosis. To improve dosing accuracy, adjustments are often made for clinical variables that affect warfarin’s action and metabolism, such as age, weight, race, and concomitant medications. However, recent findings have shown that one third of the variation in ability to tolerate warfarin is a consequence of mutations in two genes, cytochrome P450 (CYP2C9) and vitamin K epoxide reductase (VKORC1; Hum Genet 2007; 121:23).

To evaluate the accuracy of a novel warfarin-dosing algorithm that incorporates both pharmacogenetic and clinical data, researchers conducted an industry-supported, international, retrospective study involving more than 5000 patients with target INRs of 2 to 3 (>65% were >60 years of age, consistent with the population that receives warfarin). One third of patients were homozygous for the VKORC1 A/A mutation, and roughly 20% had CYP2C9 mutations. The investigators assigned 80% of patients to an algorithm-derivation cohort and the remaining 20% to a validation cohort. Three warfarin-dosing algorithms were tested, based on clinical data only, fixed dosing (5 mg daily), or pharmacogenetic plus clinical data.

The pharmacogenetics-based algorithm yielded dose estimates that were significantly closer to actual therapeutic doses needed by patients in the derivation cohort than did either the clinical algorithm or the fixed-dose algorithm (mean absolute errors in mg per week, 8.3 vs. 10.0 or 13.3). Of note, among the 33.9% of patients in either cohort who required small doses (<21 mg weekly), significantly fewer patients who were treated using the pharmacogenetics-based algorithm would exceed therapeutic doses than would those who were treated using the clinical algorithm or the fixed-dose algorithm (59.7% vs. 74.8% or 100%; P<0.001 for both comparisons). Similarly, among the 12.4% of patients in either cohort who required high doses (>49 mg weekly), significantly more patients who were treated using the pharmacogenetics-based algorithm would receive therapeutic doses than would those who were treated using the clinical algorithm or the fixed-dose algorithm (32.8% vs. 13.3% or 0%; P<0.001 for both comparisons). However, among the 53.8% of patients in either cohort who required intermediate doses (21–49 mg weekly) dosing accuracy was similar regardless of which algorithm was used.

Comment: These results show that using clinical data to determine warfarin dosing is more accurate than using fixed dosing and that using pharmacogenetic plus clinical data is more accurate than using clinical data alone. The benefit of the phamacogenetics approach was confined to the 46.2% of patients who had high- or low-dosing requirements and thus were at greatest risk for bleeding or thrombosis from overdosing or underdosing. Although the advantages of using pharmacogenetics to guide warfarin dosing are clearly demonstrated by these findings, the usefulness of this approach is currently limited by difficulties in obtaining patient genotypes within the first few days of warfarin administration. Indeed, a recent study showed that genotype-guided warfarin dosing is not currently cost-effective (JW Oncol Hematol Feb 10 2009).

— David Green, MD, PhD

Published in Journal Watch Oncology and Hematology February 18, 2009

Citation(s):

Klein TE et al. for the International Warfarin Pharmacogenetics Consortium. Estimation of the warfarin dose with clinical and pharmacogenetic data. N Engl J Med 2009 Feb 19; 360:753.

• Original article (Subscription may be required)
• Medline abstract (Free)