Warfarin (a.k.a. Coumadin, Jantoven, Marevan, or Waran) is the most widely-prescribed blood-thinning agent on the market. It’s also (in the words of Howard McLeod) a “terrible drug” – it has a very narrow therapeutic window, meaning that the minimal useful dose and the maximal safe dose are very close together. (The effects of over-dosing on warfarin – reduced blood clotting – are so severe that the drug is also used as a highly effective rat poison.)
To further complicate things, the dosage of the drug that is both effective and safe differs widely between individuals, and is known to be altered both by environmental factors (e.g. smoking) and by genetics. This variation means that determining the right dosage for a patient is a fairly hit-and-miss affair, and patients often have to be tested at several different doses before a stable dose is found.
At the current time there are two genetic variants known to explain a fairly large chunk of this variation, one in a gene involved in drug metabolism (CYP2C9) and another in a gene involved in the clotting response (VKORC1). A paper in this week’s New England Journal of Medicine looked at whether the use of genetic information from these two variants could improve the ability of clinicians to predict the optimal dosage for patients.
The short answer is: yes. The authors used data from 4,043 patients to determine the optimal parameters for their prediction algorithm, and then tested the results on a validation set of 1,009 subjects. They found that the addition of genetic information significantly increased the accuracy of dose predictions, with the improvements being seen mainly in patients at the extremes (i.e. those who required higher or lower doses of the drug than the population average).
The clinical benefits of this testing are clear: incorporating genetic data means that more patients will be placed on the correct dosage of warfarin from the outset, rather than having to go through multiple (potentially dangerous) trial runs at different doses before settling on a stable figure. Further studies on the genetic variants associated with warfarin response, as well as gene-environment interactions, should improve predictive performance even further.
The authors have made their prediction algorithm available as a nifty online calculator.
I noted in my post last week on routine whole-genome sequencing of newborns that the real clinical benefits of widespread genetic screening will come first in pharmacogenomics – these results provide a neat demonstration of this process in action.