Pharmacogenomic testing for warfarin dosing: we are ready now

Warfarin, an anticoagulant, is widely used to prevent thrombosis in patients at high risk. This drug has a narrow therapeutic window, necessitating regular monitoring of the prothrombin time, as well as calculation of the international normalized ratio (INR), for maximum efficacy and toxicity avoidance. Despite this effort, there continues to be cases of underdosing, which can lead to thromboembolism, and overdosing, leading to severe bleeding. The variability of optimum warfarin dosing is linked to genetic polymorphisms that affect the pharmacokinetics (cytochrome P450 [CYP2C9]) and pharmacodynamics (vitamin K epoxide reductase complex 1 [VKORC1]) of warfarin [1]. Algorithms have been developed to optimize initial warfarin dosing based on a patient’s genotype, demographic factors and comedication records [2]. In prospective, randomized trials, use of these dosing algorithms versus standard warfarin dose initiation has reduced the time needed for patients to reach a stable INR and the duration of time that patients spend outside target INR values [3–5]. While a meta-analysis has shown that pharmacogenetics-based dosing is associated with a reduced incidence of bleeding, results have not reached statistical significance (risk ratio: 0.86; 95% CI: 0.22–2.06) owing to small numbers of enrollments, as well as adverse events [6]. Clearly, more randomized studies are needed, justifying the funding of Clarification of Optimal Anticoagulation through Genetics (COAG), a National Lung, Heart and Blood Institute (NHLBI)-sponsored trial [101], and other ongoing studies (e.g., NCT00162435, NCT00927862, NCT00904293 and NCT00654823 from ClinicalTrials.gov).

The publication of hundreds of other research papers on the pharmacogenomics of warfarin has given promise for pharmacogenomics as a new clinical discipline, and has stimulated several companies to produce US FDA-cleared assays for use in the clinical laboratory [102]. To further fuel interest, the FDA’s Center for Drug Education and Research mandated relabeling of warfarin, effective from August 2007, recommending genotyping for CYP2C9 and VKORC1 for initial dosing [7]. This action, along with other recommendations made by this committee, was designed to improve drug safety and reduce medical errors, while at the same time promoting personalized medicine. Unfortunately, this ruling backfired; it has resulted in a series of opinions, guidelines and recommendations, each suggesting that pharmacogenomic testing should not be performed owing to the lack of existing evidence of clinical benefit [8–10,103]. If the Center for Drug Education and Research had not made this recommendation, these various professional practice groups would not have been inclined to make public commentary at this time.

The latest and perhaps the most significant naysayer, at least in the USA, is the Centers for Medicare and Medicaid Services (CMS) who, in August 2009, ruled that they will not reimburse for warfarin pharmacogenomic testing unless it is part of a clinical trial [104]. This, and each of the other pronouncements, are premature and will likely unduly influence the natural course of medical science. No other new test in the history of laboratory medicine has undergone such scrutiny so early in its validation stage. In the clinical laboratory, we continue to offer and get paid for tests that are antiquated or offer duplicate information. For example, bleeding time continues to be ordered and reimbursed, despite the fact that professional groups have advocated its discontinuance over a decade ago [11]. Hundreds of millions of dollars are spent each year on testing both amylase and lipase, when the latter test is all that is needed for diagnosis of acute pancreatitis [12].

The CMS’s ruling of reimbursement denial greatly reduces the possibility of ‘self-investigations’; in other words, individual clinicians gaining experience with the warfarin pharmacogenomic tests in their own hands. Many important therapeutic and laboratory medicine discoveries have resulted from this mechanism. For example, aspirin was originally used for analgesia but, in 1950, Craven showed that this drug can be used to prevent artery thrombosis [13]. Cardiac troponin was initially FDA-cleared as a diagnostic test for acute myocardial infarction but is now widely used to also risk stratify patients for predicting future adverse cardiac events [14].

The perceived negativity towards the clinical pharmacogenomics of warfarin may also inhibit research into the discovery of new genomic associations. The current algorithms are highly predictive of patients who are warfarin-sensitive – those requiring lower warfarin dosage. Polymorphisms that are associated with warfarin resistance have not been discovered, despite research using genome-wide association studies [15]. Why some individuals require double or triple the starting dose to maintain a target degree of anticoagulation remains unknown. Finding the additional factor (genetic or nongenetic) will improve prediction algorithms and will fine-tune dosing predictions.

Teagarden suggested that warfarin pharmacogenomic testing should be adopted based on the promise of better outcomes given that there are expectations of a clinical benefit, and there do not appear to be any negative effects beyond the costs of testing itself [16]. Statistical power analyses suggest that very large numbers of subjects need to be enrolled into various clinical trials to demonstrate clinical benefit, so we may be years away from a statistical conclusion. In the meantime, patients will continue suffer adverse events from use of this drug.

Woodcock and Lesko stated that the value of pharmacogenomic testing is not the identification of the ‘average’ patient who does not go on to have an adverse event, but the ‘atypical’ patient who does [17]. Given the prospect of catastrophic outcomes, we have found that many patients are willing to pay for these services using their own discretionary funds. Denial of reimbursement for this test, especially for the elderly on a fixed income or the unemployed with limited resources, does not appear to be equitable.

An important rationale for the CMS ruling on pharmacogenomic testing is the current costs of providing such services. It must be mentioned that germ-line testing for pharmacogenomics need only be conducted once in the lifetime of the individual. The laboratory costs of pharmacogenomic testing will decrease with improvements in analytical technology. We used to think that hormone testing by radioimmunoassay required cost justification [18]. Today, the results of millions of single nucleotide polymorphisms (SNPs) from a given individual are available for a few hundred US dollars. With this explosion of interest, automated genetic analysis for individual SNPs may approach the costs of obtaining a result for thyroid-stimulating hormone today. When this does occur, cost-justification analyses that are currently marginal will be very viable [19].

Until the final evidence is known, one can question whether pharmacogenomic testing should be permitted now rather than later. In the absence of hard clinical outcomes, is the achievement of a target INR faster or for a longer duration a sufficient clinical objective? If false, the importance of INR measurements to correctly manage warfarin patients may have been overstated these last few years. Might it be possible that pharmacogenomic testing can cause a reduction in the number of dosage adjustments needed affecting the frequency of outpatient visits, or the number of needed INR measurements performed themselves? If true, pharmacogenomic testing might be ready today. In the end, the decision to test might be up to the individual or their family members. The value of a clinical laboratory test used on your mother that can produce a better starting dose for warfarin, resulting in a reduced likelihood of a subarachnoid hemorrhage? Priceless.

Financial & competing interests disclosure

Alan HB Wu is on the Medical Advisory Council of Iverson Genetics, and has received financial support from Autogenomics Inc., Osmetech Inc., and Roche Diagnostics. The author has no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.