March 1, 2013 (Vol. 33, No. 5)
Ramy A. Arnaout, M.D., D.Phil.
Genomic research is widely expected to transform patient care, but progress has been slower than was first expected.1,2 For critics, delays represent broken promises and suggest that genomic research funding might be better spent elsewhere.3-5 For proponents, the pace simply underscores the complexity of the relationship between genetics and disease and argues, if anything, for more funding.6,7
Until recently, there has been little hard data available for informing these opinions. This has started to change. Today it is possible to bring the tools of predictive modeling to bear on at least one area of medicine—prescription drugs (Figure)—to gain insight into when genomic research can be expected to have a major impact on public health, and how much this research can be expected to cost.
Despite being the mainstay of modern medicine, prescription drugs often fail to work as intended. For example, of 30 million Americans who take aspirin to prevent stroke or heart disease, a quarter experience treatment failure in the form of aspirin resistance.8,9 Of 60 million who take statins to lower cholesterol, three million experience muscle pain, elevated liver enzymes, or rhabdomyolysis as side effects that can lead to nonadherence and thereby to heart disease related to hypercholesterolemia.9,10
Treatment failure and side effects lead to adverse drug events, defined as injuries that result from drug-related medical intervention.11 Adverse drug events affect one in every five outpatients and one in 11 inpatients12 and cost $80 billion a year in the United States alone.13 Thus, lowering rates of treatment failure and side effects—what we collectively term drug-related adverse outcomes—would have a major impact on both public health and healthcare spending.6,14-18
Evidence suggests that genomics can contribute materially to this goal. Of the 80% of outpatient and nearly half of inpatient adverse drug events that are currently considered nonpreventable12, a majority are now thought to result from genomic variation.19,20
For example, many patients on warfarin, an anticoagulant used by 30 million Americans, experience bleeding or clotting due to variability in dosing requirements.21,22 Two-thirds of this variability can be explained by single-nucleotide polymorphisms (SNPs) in a set of six genes.19
In HIV-positive patients receiving the reverse-transcriptase inhibitor abacavir, genetic screening for variants in genes of the major histocompatibility complex almost completely avoids potentially life-threatening hypersensitivity reactions.23
Despite understandable excitement,23,24 so far only a few associations between germline genetic variants and adverse outcomes have proven clinically useful. However, a number of associations are in advanced clinical investigation. These include VKORC1 and CYP2C9 and warfarin-induced bleeding, HLA-B and abacavir-induced hypersensitivity, CYP2C19 and cardiac events due to clopidogrel resistance, HLA-B and carbamazepine-induced Stevens-Johnson Syndrome, SLC01B1 and statin-induced myopathy, and COMT and return to smoking following use of the nicotine patch.20, 26-28
For these associations, the incidence of the adverse outcome, the frequency of the associated variant in the population, and the percent attributable risk of each variant for its associated adverse outcome are all known.
As a result, one can set aside opinions about the value of investing in genomic medicine and instead ask objective questions. For example, by factoring in the time and money it took to discover and validate known associations, one can estimate how much more time and money it will take to discover and validate pharmacogenomic guidelines for a more comprehensive set of associations—specifically, enough to cut the rate of drug-related adverse outcomes in half.
Today, predictive modeling is commonplace for understanding everything from presidential elections to the weather.29-32 Predictive modeling in pharmacogenomics suggests that cutting adverse outcomes by as much as half will require a research investment of about $6 billion and take 20 years.33 Numbers like these allow comparisons. For example, $6 billion annualized over 20 years represents 4% of the U.S. NIH’s 2011 budget,34 2% of the pharmaceutical industry’s 2011 research budget,35 and 0.2% of 2009 payments made by U.S. private health insurers.36 These numbers are also comparable to the price-tag and timeline for developing just four to five new drugs from scratch.37
Numbers also help set expectations. For example, a 20-year timeline is consistent with forecasts by the National Human Genome Research Institute that genomics will produce some advances in the science of medicine before 2020 but much more thereafter.6-8 But for the specific goal of developing pharmacogenomic guidelines that can halve drug-related adverse outcomes, forecasting predicts a specific completion date (2032) and a requirement that most of the investment—around $3 billion at $400–500 million per year—will have to come over the next 5–6 years, before most guidelines appear. Thus the evidence supports patience, for now.
The availability of numbers shifts the debate to practical issues. These include deciding whether the clinical benefit of developing guidelines is worth $6 billion, and if so what party or parties—for example government (via taxation), pharmaceutical companies, or health insurers—should pay; finding measures of progress during the pump-priming years; and ensuring that downstream processes for implementing guidelines be ready by the time the bulk of the guidelines start appearing.
Predictive modeling can inform policy choices by pointing out bottlenecks, for example discovering and confirming candidate associations between genetic variants and adverse outcomes. Modeling suggests that this process could be sped up by comprehensively mapping genomic variation vs. incidence of adverse outcomes for ~1,500 people of representative ethnicities taking each of the 40–50 most-used prescription drugs.33 Such a “50,000 Pharmacogenomes Project” would be a nontrivial undertaking. But in the spirit of the 1,000 Genomes Project, UK10K, and the Million Veteran Program, it would represent a disruptive improvement that could save time and money. Finally, modeling helps set research priorities—in this case better understanding the extent to which, and the ways in which, genomic variation influences adverse outcomes.33
Genomics is maturing. As the fruits of research come within reach of clinical medicine, charting the road ahead will become easier. The numbers in this article are not the last word, but they do start a conversation and illustrate what is possible. Similar analyses covering cancer, infectious disease, and heritable conditions will soon be possible (Figure).
Being specific about goals, costs, and benefits helps advance the debate about future directions and helps focus attention on making choices and meeting the challenges of translating research into better patient care. In this way, it helps keep the promise of genomic medicine bright, free from the tarnish of false expectations.
Ramy A. Arnaout, M.D., D.Phil. (firstname.lastname@example.org), is in the Department of Pathology, Beth Israel Deaconess Medical Center; Division of Clinical Informatics, Department of Medicine, Beth Israel Deaconess Medical Center; and Harvard Medical School.
1 N. Wade, “A Decade Later, Genetic Map Yields Few New Cures,” The New York Times, June 13 2010, pp. A1.
2 A. Pollack, “Awaiting the Genome Payoff,” The New York Times, June 15 2010.
3 “The Genome, 10 Years Later,” The New York Times, June 21 2010, pp. A28.
4 J. P. Evans, E. M. Meslin, T. M. Marteau, T. Caulfield, Genomics. Deflating the genomic bubble. Science 331, 861 (2011).
5 S. S. Hall, Revolution postponed. Sci Am 303, 60 (2010).
6 E. D. Green, M. S. Guyer, Charting a course for genomic medicine from base pairs to bedside. Nature 470, 204 (2011).
7 M. S. Boguski, R. Arnaout, C. Hill, Customized care 2020: how medical sequencing and network biology will enable personalized medicine. F1000 Biology Reports 1, 73 (2009).
8 G. Kasotakis, Pipinos, II, T. G. Lynch, Current evidence and clinical implications of aspirin resistance. J Vasc Surg 50, 1500 (2009).
9 J. J. Verschuren et al., A systematic review on pharmacogenetics in cardiovascular disease: is it ready for clinical application? Eur Heart J, (Jul 30, 2011).
10 J. C. LaRosa et al., Intensive lipid lowering with atorvastatin in patients with stable coronary disease. N Engl J Med 352, 1425 (2005).
11 D. W. Bates et al., Incidence of adverse drug events and potential adverse drug events. Implications for prevention. ADE Prevention Study Group. JAMA 274, 29 (Jul 5, 1995).
12 S. V. Tache, A. Sonnichsen, D. M. Ashcroft, Prevalence of adverse drug events in ambulatory care: a systematic review. Ann Pharmacother 45, 977 (Jul, 2011).
13 D. W. Bates et al., The costs of adverse drug events in hospitalized patients. Adverse Drug Events Prevention Study Group. JAMA 277, 307 (1997).
14 W. G. Feero, A. E. Guttmacher, F. S. Collins, Genomic medicine–an updated primer. N Engl J Med 362, 2001 (2010).
15 D. A. Mrazek, C. Lerman, Facilitating clinical implementation of pharmacogenomics. JAMA 306, 304 (Jul 20, 2011).
16 S. W. Paugh et al., Cancer Pharmacogenomics. Clin Pharmacol Ther 90, 461 (Jul 27, 2011).
17 R. A. Wilke, M. E. Dolan, Genetics and variable drug response. JAMA 306, 306 (Jul 20, 2011).
18 M. V. Holmes, J. P. Casas, A. D. Hingorani, Putting genomics into practice. BMJ 343, d4953 (2011).
19 M. Wadelius et al., Association of warfarin dose with genes involved in its action and metabolism. Hum Genet 121, 23 (2007).
20 L. Wang, H. L. McLeod, R. M. Weinshilboum, Genomics and drug response. N Engl J Med 364, 1144 (2011).
21 A. L. Gullov, B. G. Koefoed, P. Petersen, Bleeding during warfarin and aspirin therapy in patients with atrial fibrillation: the AFASAK 2 study. Atrial Fibrillation Aspirin and Anticoagulation. Arch Intern Med 159, 1322 (1999).
22 P. Smith, H. Arnesen, I. Holme, The effect of warfarin on mortality and reinfarction after myocardial infarction. N Engl J Med 323, 147 (1990).
23 S. Mallal et al., HLA-B*5701 screening for hypersensitivity to abacavir. N Engl J Med 358, 568 (Feb 7, 2008).
24 G. S. Ginsburg, H. F. Willard, Genomic and personalized medicine: foundations and applications. Transl Res 154, 277 (Dec, 2009).
25 Institute of Medicine (IOM), “The value of genetic and genomic technologies: Workshop summary.” (The National Academies Press, Washington, D.C., 2010).
26 M. V. Relling, T. E. Klein, CPIC: Clinical Pharmacogenetics Implementation Consortium of the Pharmacogenomics Research Network. Clin Pharmacol Ther 89, 464 (Mar, 2011).
27 J. A. Johnson et al., Clinical Pharmacogenetics Implementation Consortium Guidelines for CYP2C9 and VKORC1 Genotypes and Warfarin Dosing. Clin Pharmacol Ther, (Sep 7, 2011).
28 Materials and methods are available as supplementary material on Science Online.
29 E. Pfanner, in International Herald Tribune Rendezvous New York Times, Ed. (2012).
30 J. Manyika et al., “Big data: The next frontier for innovation, competition, and productivity” (McKinsey Global Institute, 2011).
31 P. A. Duin, P. V. D. Duin, E. Masini, Eds., Knowing Tomorrow?: How Science Deals With the Future, (Eburon Publishers, Delft, 2009).
32 A. L. Porter et al., Technology futures analysis: Toward integration of the field and new methods. Technological Forecasting and Social Change 71, 287 (Mar, 2004).
33 R. Arnaout, T. P. Buck, P. Roulette, V. P. Sukhatme, Predicting the Cost and Pace of Pharmacogenomic Advances: An Evidence-Based Study. Clin Chem, (Dec 10, 2012).
34 HHS, “HHS Budget in Brief” (Department of Health and Human Resources, 2012).
35 The Pharmaceutical Research and Manufacturers of America (PhRMA). (2012), vol. 2012, pp. Information about and from the pharmaceutical industry’s largest trade association.
36 Centers for Medicare and Medicaid Services. (Centers for Medicare and Medicaid Services, 2012), vol. 2012, pp. Summarizes national health expenditure data for 2009. .
37 J. A. Vernon, J. H. Golec, J. A. Dimasi, Drug development costs when financial risk is measured using the Fama-French three-factor model. Health Econ 19, 1002 (2010).
38 E. Green, Biomedical research. The genome project: what will it do as a teenager? Interview by Jocelyn Kaiser. Science 331, 660 (Feb 11, 2011).