In an effort to improve the risk-benefit profile of therapies in clinical care, precision medicine seeks to identify and make use of factors, often genetic variants or biomarkers, that influence or predict the response to treatment. The Precision Medicine Initiative defines precision medicine as “an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment, and lifestyle for each person.”1 Across the United States, departments, institutes, and centers have created new groups focused on this approach. Precision medicine has acquired strong vested interests, including industry and academic medical centers with committed researchers, spin-off companies, and occasionally, some enthusiastic clinicians. A favored theory can be a powerful force.
The principle underlying precision medicine is that the same treatment may not be effective for all patients. If, for instance, only 20% of patients benefit from a treatment, why also prescribe that same treatment for the other 80% of patients? Or if a subgroup of patients will experience toxic effects of an intervention without any benefit, why not avoid that treatment in these patients? With such an approach, both the number needed to benefit and the number needed to harm can be improved. While these considerations are sensible, the underlying assumptions of the precision-medicine model may not be obvious: they include the existence of meaningful subgroups, the ability to identify them, and the availability of tailored treatments.
A timely example of the precision-medicine model is hemophilia B, a bleeding disorder caused by a genetic mutation associated with low levels of factor IX. Traditionally, treatment involves replacement of exogenous factor IX. A report involving 10 patients with hemophilia who received gene therapy with a high specific-activity factor IX variant demonstrated that gene transfer largely eliminated the need for prophylaxis, bleeding events, and factor use for a year.2 This novel treatment, which targeted the deficiency in hemophilia B, successfully advances the frontier of precision medicine with gene therapy.
This precision-medicine example differs from the current standard in preventive medicine. Preventive medicine therapies and interventions target a risk factor such as high blood pressure to reduce the risk of cardiovascular events, and the number needed to treat to prevent an event may be 100 or more. Broad exposure without immediate benefit means the treatment must be safe, especially for treatments and interventions that involve large segments of the population. Moreover, the relative risk reduction associated with antihypertensive treatment, for example, tends to be the same across phenotypic subgroups. Targeting high-risk groups, including those defined by genetic risk scores, decreases the number needed to treat and, therefore, represents cost-effective medicine rather than precision medicine. Despite intense investigation for decades, no known procedure or biomarker makes it possible to select the subgroup patients for treatment, such as those with hypertension, whose cardiovascular event will be prevented. The fact that many individuals may need to be treated to prevent 1 event simply reflects the biological fact that the preventive medicine model of drug treatment is, at this stage, probabilistic, although future research may eventually reveal a targetable subgroup.
Hemophilia and hypertension represent 2 different models of treatment. The first is a disease with a deterministic etiology, and in these types of conditions, targeting the genetic variant, the infectious agent, or the precise biological deficit or derangement is often extremely effective. In contrast, in conditions such as hypertension and hyperlipidemia, high levels of the risk factor increase the probability of a cardiovascular event. Many current research efforts on risk factors seek to identify a biological subgroup that derives large benefit from a specific therapy, a research effort that may or may not be successful. To demonstrate that one subgroup responds differently from others, a formal test for interaction, which requires large sample sizes for adequate power, is necessary. To change practice in a meaningful way, evidence of clinical validity and utility is also necessary.
Perhaps the most successful area of precision medicine is cancer treatment. Tumors tend to arise from genetic variants, often somatic variants, in a limited set of genes; and several new treatments target these variants successfully. Examples include imatinib for Philadelphia chromosome–positive leukemia, gefitinib for non–small cell lung cancer, vemurafenib for malignant melanoma, and immune checkpoint inhibitors for tumors with mutation repair deficiencies. In each instance, a subgroup of patients, within a tumor type or across anatomical lesions, has a common biology such that a genetic variant or variants define susceptibility to a specific therapy.
More recently, studies have evaluated treatments in a variety of tumors that all share the same genetic variant. In these “basket studies,” the treatment response seems, however, to differ among anatomical tumors. Not only the mutation but also tumor type and other factors influence the success of treatment. Even among the most responsive tumors types, not every patient responds. As a result, the deterministic cancer drivers are also affected by stochastic elements, such as other drivers or mutations. The goal of precise selection of responsive patients is still under investigation. In an effort to separate drivers from passengers for a tailored therapy, early basket studies need to be followed up with phase 3 studies that demonstrate improved survival or progression-free survival. Without sufficient evidence, the danger is the use of mutation testing and costly off-label drugs that are likely to offer at once little or no chance of benefit and, at the same time, unnecessary exposure to serious adverse events.
Although the progress in cancer biology has been substantial, translating this knowledge into long-term health benefits remains difficult. Tumors are heterogeneous, pathways are adaptable and plastic, and new somatic mutations confer resistance. For lung cancer, for instance, epidermal growth factor receptor mutations occur in 10% to 20% of white patients, and the median improvement in survival associated with gefitinib is only 3 months.3 In contrast, smoking cessation increases mean life expectancy by 7 to 8 years among adults 35 years of age and by 1.4 to 2.7 years among adults 65 years of age.4
Compared with cancer, cardiovascular medicine has been more resistant to the precision-medicine effort. In a review of US Food and Drug Administration drug labels,5 fewer cardiovascular drugs (n = 15) than oncology drugs (n = 54) were associated with pharmacogenetic biomarkers, and the evidence of clinical utility was much less common for cardiology (13%) than oncology (57%). For the treatment of blood pressure and cholesterol, a few genetic variants have been shown to have small effects on these surrogate end points, but there are no demonstrated effects on the prevention of events such as myocardial infarction or stroke. For warfarin treatment, genetic variants influence time in anticoagulation control as the primary outcome, but coordinated care in anticoagulation clinics is probably more effective than genetic testing in improving health outcomes. The new generation of anticoagulants may well make the pharmacogenetics findings for warfarin only of historical interest.
The large-scale genetic studies of cardiovascular conditions, which have reliably identified scores of genetic variants, have little to do with precision medicine. The usual goals of these studies are neither prediction nor the identification of drug-gene interactions; rather, the goal is greater understanding of human population biology. These studies are primarily discovery efforts designed “to highlight causal pathways that can be modulated for therapeutic gain.”6 The PCSK9 (proprotein convertase subtilisin/kexin type 9) inhibitors are one example of human genetic studies that identify novel drug targets.
Precision medicine tends to medicalize conditions and risk factors, often within a gene-centric framework, and intense efforts to identify and target subgroups may ignore or overlook the mismatch between the burden and causes of disease in the United States. What is needed are public health efforts to increase prevention of cancer and cardiovascular disease and to reduce social and racial disparities as well as inequalities in education, income, and opportunities.7 For medical treatments, one-size-fits-all treatments that are known to be effective will continue to work well until or unless convincing evidence becomes available that targeting subgroups actually improves health outcomes. These 2 medical-treatment models, which may both apply to the same disease across time, reflect in part differences in the underlying biology and in part differences in the state of scientific and clinical knowledge.