Integrating Pharmacogenomics in Practice: One GIFT at a Time or a Package Deal?

Testing for genetic variants related to warfarin metabolism has been the poster child for pharmacogenomics in clinical practice for over a decade. A recent publication of the Genetic Informatics Trial (GIFT) provides additional evidence for the clinical benefits of incorporating pharmacogenomics testing into the management of warfarin initiation. In this multicenter study of older individuals receiving warfarin for deep vein thrombosis (DVT) prophylaxis after elective orthopedic surgery, incorporating variant testing in an algorithm for selecting initial dosing improved a composite outcome (major bleeding, biochemical measures , venous thromboembolism, or death) vs standard of care. Notable features included large size, number of variants incorporated in the testing, testing prior to drug initiation, and the composite outcome powered to detect important health outcomes.

To date, the evidence supporting the use of pharmacogenomics testing for warfarin has been inconsistent, with larger studies suggesting no clear health benefit. A number of small studies suggest benefits in indirect metrics of warfarin management. As a result, the GIFT publication has deservedly received attention from the public, clinicians, and scientists. The study provides support for guidelines recommending consideration of pharmacogenomics testing prior to initiating warfarin and is consistent with existing FDA labeling of the drug.

Nevertheless, the absolute benefit derived from warfarin pharmacogenomics testing in this study was relatively small, a 3.9% absolute risk reduction yielding a number needed to test of 26 persons to see one less individual event. Further, the event most likely to be averted would be a biochemical change and not a major bleeding event or death. At a very practical level, this raises an important question: in health systems already doing a good job of managing DVT risk in post-operative joint replacement patients, is the expense of incorporating one-off, point of care testing into existing clinical workflows a justified use of scarce resources? Reasonable people could and likely will differ in their responses to this question. It is far from clear if the GIFT study will lead to widespread adoption of warfarin pharmacogenomics testing.

More generally, the study illustrates what seems to be an emerging conundrum of incorporating pharmacogenomics and other genomic data into the routine care of individuals. For any individual and most single clinical applications of genomics (risk assessment, pharmacogenomics, carrier testing), genomic data will likely be of little additional clinical value. For example, most individuals harbor no remarkable variants conferring risk relevant to the clinical application, or they never experience an event where the genomic variant becomes relevant (i.e., choosing not to have biological children, never developing a cancer for which a targeted therapy exists, or never developing a condition for which a drug dosed on pharmacogenomics would be useful). However, in aggregate across all emerging evidence-based genomic applications, the likelihood of any individual deriving at least moderate health benefits from having genomic information readily available is reasonably high, even today. For a small subset of individuals, genomic data (or a lack of genomic data) could have life or death consequences. This is certainly true for individuals affected by selected genetic conditions, such as the (1/200) individuals with familial hyperlipidemia or the 1/250 individuals that carry variants associated with hereditary breast and ovarian cancer, or Lynch syndrome. These conditions are part of the CDC Tier 1 list of genomic applications for which evidence based guidelines exist to prevent cancer or heart disease. Lack of knowledge of genetic status for these disorders may result in substantial, but preventable, morbidity and mortality, if undetected and untreated into a late age.

Given this emerging picture demonstrating that genomic data is an important potential contributor to managing a patient’s health over a life-span, building a one at a time genomic testing platform for single applications seems highly inefficient. If warfarin pharmacogenomics data were readily available in the electronic health record (EHR) when individuals were prescribed warfarin, there seems little doubt that the data would be used routinely. But, if an entire system needs to be developed at the point of care for obtaining and using warfarin pharmacogenomics data at the time of a decision to treat, it may never become part of routine care.

It is time for a coordinated effort to ensure that all health record systems be able to incorporate structured genomic information in a way that can be queried, updated, analyzed and shared. The NIH funded IGNITE network has identified the inability to adapt EHRs to manage genomic data as the most important barriers to utilizing genomic data to improve patient management in health care systems. Remedying these barriers across all U.S. health care institutions with EHRs will be an incredible challenge. Considerable groundwork has been laid over the last decade by the groups such as the Personalized Healthcare Workgroup of the American Health Information Community, and a broad group of government and non-government collaborators interested in this topic. Developing electronic health records that can manage genomic data is crucial for the future U.S. health care system in order to ensure broad and equitable access to pharmacogenomics and other evidence-based genomic interventions.