The Limits and Potential Future Applications of Personalized Medicine to Prevent Complex Chronic Disease

The milestone sequencing of the human genome more than a decade ago dawned a genomic revolution with the potential of transforming health. The anticipated benefits included the identification of the early biomarkers of disease, the ability to combine information on risk-conferring genes to improve risk prediction, and the discovery of new disease pathways that can serve as the target for novel therapeutics. Despite monumental breakthroughs, however, genomic science is in its infancy. A gap remains between science and patient benefit.

A growing concern is that the term “personalized medicine” (also called “predictive” or “precision” medicine) has become a proxy for DNA-centered approaches to prevention and treatment.¹ Personalized medicine is the individualizing of health care based on a person’s susceptibility to disease and response to treatment. Environmental and behavioral factors are shown to contribute more to premature death than genetic factors.² Yet their contribution to complex chronic diseases, such as diabetes mellitus, coronary arterial disease, and cancer, is overshadowed by a gene-centered view. The imbalanced emphasis on genetic research and implementation has raised caution that unrealistic expectations are being created rather than making substantive progress toward prevention—or that “it’s not just about genes, drugs, and disease.”^1,3 Biomedical research may therefore be accelerated by more comprehensive models that take into account genomic (ie, the study of genes and their interaction with the environment and other genes for expression) rather than genetic (ie, gene-disease associated) information. According to the National Institutes of Health, pioneering the measurement and tracking of social and behavioral data will be a necessary component of the Precision Medicine Initiative.⁴ The White House launched the Precision Medicine Initiative in 2015 to systematize and sustain a national effort toward leveraging advances in genomics, analysis of large data, and health information technology to further biomedical discovery.

Limitations in Applying Genetics for Primordial, Primary, and Secondary Prevention of Chronic Diseases

After the successes of genetic-based risk assessment for single-gene disorders, such as cystic fibrosis and Huntington’s disease, Mendelian concepts were applied to complex traits, including traits for chronic diseases, such as heart disease, diabetes, obesity, and cancer. Most studies on human genetic variants investigated single-nucleotide polymorphisms (SNPs), which are variants in single bases along the chromosome. By testing hundreds of thousands of SNPs in individuals for an association with disease, genome-wide association studies (GWAS) are widely used to identify SNP variants implicated in chronic disease. A GWAS catalog of all SNP-associated traits robustly implicated with common chronic diseases with real-time updates is available and maintained jointly by the National Human Genome Research Institute and the European Bioinformatics Institute.⁵

The predictive value of genetic-based risk assessment using SNPs is limited. A major challenge in GWAS is that trait-associated SNPs may not be causative variants. Small differences in DNA sequence at a specific location in a gene associated with a disease are called risk alleles. Because complex chronic diseases are associated with many risk alleles, each contributes only a small increase in risk. The median odds ratio for disease risk per copy of the risk allele is approximately 1.33 (interquartile range, 1.20-1.61).⁶ Given the high prevalence of chronic disease, most variants identified through GWAS impart an even smaller effect on the relative risk of disease. To date, although a large number of common DNA variants for heart disease, diabetes, obesity, and cancer have been revealed, these genetic variations typically collectively explain fewer than 10% of genetic susceptibility of the burden of these chronic diseases.⁷ This gap between the revealed and estimated hereditability of common disease is termed the “missing hereditability.”⁸

Although the effect of individual SNPs on complex disease phenotypes is weak, genetic risk scores can be constructed based on a large number of genotyped variants. Preliminary studies suggest that genetic risk scores may improve the accuracy of personalized risk assessments when added to traditional risk predictors, such as for heart disease. However, the ultimate clinical utility of genetic risk scores is uncertain. A promising area of investigation is the benefit of genetic risk scores on high-risk populations.⁹ Some genetic models alternatively suggest, however, that the predictive value of genetic profiling for common diseases may not become much better than the predictive value of traditional risk factors because genetic factors only partly contribute to the causal mechanisms involved in developing chronic disease.¹⁰

Another limitation of predicting risk based on genetic variants is that traits for chronic diseases are complex. Genetic variances not only account for a small percentage of heritable risk, but they also predict an even smaller percentage of overall disease risk. The population-attributable risk percentage, meaning the percentage of disease incidence that would be eliminated if the risk factor were removed, is high for nongenetic factors in common chronic diseases, often at least 80% to 90%. For example, more than 90% of diabetes mellitus¹¹ and more than 80% of coronary heart disease¹² can be prevented through 5 habits and behaviors: a healthy dietary pattern, moderate-to-vigorous physical activity, tobacco avoidance, moderating alcohol consumption, and maintaining a normal body mass index. Hence, gene-environment interactions, not accounted for in GWAS, would provide a more realistic framework for reliably predicting chronic disease than exclusively using genetic factors.

DNA-based disease risk estimates will translate into disease prevention only if they are acted upon. An additional limitation and critical question is whether knowledge of genetic risk information motivates behavior change. For example, does communicating to adults that they have an increased genetic risk of developing diabetes motivate physical activity and dietary change? Available evidence does not support the expectation that results from DNA-based testing for risk-conferring gene variants for common complex diseases motivates long-term behavior change. In a meta-analysis of 18 controlled trials involving adults in which one group received personalized DNA-based estimates of disease risk for conditions in which risk could be reduced by behavior change, no significant effect was seen on smoking cessation, physical activity, diet, alcohol use, or other behaviors.¹³ The results of this study are consistent with a study of more than 2000 participants in which direct-to-consumer genome-wide profiling did not result in any measurable short-term changes in psychological health, dietary fat intake, or exercise behavior.¹⁴ These studies suggest that genetic susceptibility in isolation may not increase behavior motivation.

Potential Applications of Genetic-Based Knowledge for Personalized Chronic Disease Prevention

Although the tension between personalized medicine for the benefit of individuals and public health interventions that equitably benefit the population is inherent, the success of personalized medicine for chronic disease prevention may ultimately require a population-based approach. Advancing wellness rather than the treatment of disease through the study of dynamic biologic interactions at a system level combined with bioinformatics is termed “P4 medicine.” Its 4 components include predictive, preventive, personalized, and participatory medicine. As argued by Khoury et al,¹⁵ P4 medicine will have its greatest success by integrating the “population perspective”—that is, the “fifth P”—into each of the other 4 components. Precision public health, guided by genetic and gene-environment interactions, may move us closer toward fulfilling the promise of personalized medicine for the greatest benefit of individuals and the population.

Environmental population-level prevention measures are shown to be more cost-effective and equitable than efforts directed to individuals.^16,17 They may also be more efficacious. Despite knowledge about the importance of reducing the prevalence of modifiable factors (eg, smoking, poor diet, and physical inactivity), which the Centers for Disease Control and Prevention suggests may prevent at least one-third of deaths,¹⁸ attempts to change these behaviors among individuals are often met with resistance. Interventions that target nonconscious processes have been proposed to be more effective in changing behavior than those that engage conscious deliberative processes.¹⁹ Designing choices in ways that affect decision making is termed “choice architecture.” In workplace environments, designs that reduce obstacles toward healthful choices may be promising for promoting beneficial eating behavior.²⁰

Although an anticipated goal of personalized medicine is individualized diagnosis and treatment recommendations, the most individualized conclusions may rely on the process of placing an individual in a subgroup that shares similar clinical features from among a large, diverse study population. State-of-the-art molecular profiling corresponding to precise and comprehensive observable phenotypic disease variations, referred to as “deep phenotyping,” will expand clinical classifications. Yet it will not eliminate the need for the fundamental raison d’être of classification, which is to enable clinical diagnosis.²¹ Similarly for treatment, as in diagnosis, genomic information will increase the number of subgroups of a disease and potentially the number of corresponding treatments. However, personalizing the management of each subgroup will be limited by the need for sufficient evidence supporting the utility and safety of the treatment for the subgroup.

Despite the clear interdependence of personalized medicine and population health, their greatest synergistic benefit may be in translational applications that enable “precision prevention,”³ or tailoring preventive interventions based on disease susceptibility rather than applying them universally. For example, taking into account genetic risk and, hence, gene-environment interaction strengthens the association between traffic-related exposure to air pollution and incident childhood asthma.²² Protecting the health of those most susceptible may guide the threshold of air pollution for population-level policy interventions.

Large population-based studies that incorporate genetics may improve the understanding of disease and identification of disease targets and can be a way to provide evidence that comes close to the strength of a randomized trial. For example, if a genetic variant alters the biologic effect of an environmental risk factor, its association with a disease can strongly support a causal role for that environmental factor. Because most people are not aware of the gene variants, or polymorphisms, that they carry, using polymorphisms with biologic effect as a proxy for an environmental exposure has been referred to as “Mendelian randomization.”²³ One example is a variant of apolipoprotein B found in familial defective apolipoprotein B-100, an autosomal dominant genetic disorder of lipid metabolism associated with hyperlipidemia and elevated risk for atherosclerosis.²⁴ Establishing the association between familial defective apolipoprotein B-100 with elevated cholesterol level and coronary heart disease risk strengthens causal evidence that elevated cholesterol is a modifiable risk factor for coronary heart disease. Although apolipoprotein B has no implication for screening because of its small population-attributable risk (familial defective apolipoprotein B-100 accounts for only a small percentage of coronary heart disease), it can play an important role in understanding disease etiology and identifying targets of disease prevention.

As a complement to population-level interventions, genetic risk profiling may also enable health promotion and disease prevention programs to be more efficiently directed at subgroups of the population than when directed without the inclusion of genetic information. For example, even though lung cancer prevention efforts are substantially more effective through non–genetic-based strategies (eg, bans, taxes, and government-sponsored antismoking campaign) than through genetic-based strategies, genetic testing could identify those at higher risk of lung cancer to motivate cessation, or identify candidates for intense cessation programs.²⁵

Conclusions

The brave new world enabled by genomic science will undoubtedly change the future of medicine. Chronic diseases, however, are caused by multiple, interrelated complex pathways and environmental influences, leading to an inherent limitation in the role of a gene-centered approach to individualizing prevention. Genomic interventions that take into account environmental interactions will be needed to unlock the full potential of genomic medicine for chronic disease prevention. Integration of personalized medicine with population-based interventions and studies may overcome some of the current limitations of personalized medicine and ultimately lead to the most widespread translational preventive benefits from genomics.