Routine use of pharmacogenetic warfarin dosing algorithms unjustified

27 August 2010

MedWire News: The accuracy of pharmacogenetic warfarin dosing algorithms does not support the routine, prospective use of genetically-guided dosing in patients starting therapy with the anticoagulant, study findings indicate.

“Several dosing algorithms, which combine clinical and genetic parameters to predict therapeutic warfarin dose have been developed,” observe Paul Shaw (University of Massachusetts Memorial Medical Center, Worcester, USA) and colleagues.

In the present study, Shaw and team evaluated the accuracy of six pharmacogenetic warfarin dosing algorithms, and a 5-mg, fixed-dose approach in predicting therapeutic dose among 71 patients who were treated at an anticoagulation clinic.

All six algorithms contained a genetic component but differed in the weighting they assigned to different cytochrome P450 isoenzyme 2C9 and vitamin K epoxide reductase complex subunit 1 genotypes in contributing to warfarin dose.

The researchers compared the warfarin dose predicted for each patient by each of the algorithms to the actual, daily, therapeutic dose the patient was taking. They calculated the mean absolute error (MAE) and coefficient of determination (R²) of the actual dose compared with the predicted dose.

They report that the mean therapeutic warfarin dose was 6.1 mg/day in the study population. On average, all six algorithms, plus the fixed-dose approach, tended to under-predict the dose required.

Three algorithms, published by Zhu et al, Gage et al, and the International Warfarin Pharmacogenetic Consortium (IWPC), were similar in their primary accuracy endpoints with the MAE ranging from 1.7 to 1.8 mg/day and R² from 0.61 to 0.66.

The percentage of patients predicted within 1 mg/day of the actual dose was 37%, 38%, and 41% for Zhu et al, Gage et al, and the IWPC, respectively.

However, the Zhu et al algorithm severely over-predicted the warfarin dose in twice as many (14% vs 7%) patients as the Gage et al or IWPC algorithms. Over-prediction was defined as a dose two or more times higher, or 2 mg/day or more above the actual dose of warfarin used.

“The algorithms published by Gage et al and the IWPC were the two most accurate pharmacogenetically-based equations available in the medical literature in predicting therapeutic warfarin dose in our study population,” write Shaw and co-authors in the Journal of Thrombosis and Thrombolysis.

“However, the degree of accuracy demonstrated does not support the routine use of genotyping to prospectively dose all patients newly started on warfarin,” they conclude.

MedWire (www.medwire-news.md) is an independent clinical news service provided by Current Medicine Group, a trading division of Springer Healthcare Limited. © Springer Healthcare Ltd; 2010

J Thromb Thrombolysis 2010; 30: 220–225