Comparison of warfarin pharmacogenetic dosing algorithms in a racially diverse large cohort

Aims: Multiple warfarin pharmacogenetic dosing algorithms have been reported to date. However, there is only limited information available on the performance of the algorithms that can be used with the results of a US FDA-cleared warfarin pharmacogenetic test. We compared the performance of warfarin pharmacogenetic dosing algorithms in a large racially diverse cohort. Materials & methods: Warfarin pharmacogenetic dosing algorithms were identified using the PubMed database. Patient information from the International Warfarin Pharmacogenetics Consortium database was used to predict therapeutic warfarin doses according to each algorithm. By using bootstrapping analysis, the performance of algorithms was tested by comparing the mean absolute error and mean percentage of patients whose predicted dose fell within 20% of actual dose (percentage within 20%) in the entire cohort, and by race and therapeutic dose range. Results: A total of 13 algorithms and 1940 patients were included in the study. Overall, all the algorithms had similar performances (mean absolute error: 10.3 mg/week and mean percentage within 20%-41.4%). However, algorithms derived from racially mixed populations tended to perform better than those derived from single race populations. Mixed population algorithms had the lowest mean absolute error and the highest percentage within 20% across the racial groups. Most algorithms performed better in the intermediate-dose range (between 21 and 49 mg/week) than in the low (<= 21 mg/week) or high(>= 49 mg/week) range. Conclusion: Published warfarin pharmacogenetic algorithms performed similarly, although mixed population algorithms tended to perform better than race-specific algorithms.