Leo P. Sugrue, MD, PhD¹; Rahul S. Desikan, MD, PhD^1,2

Author AffiliationsArticle Information

JAMA. Published online April 8, 2019. doi:10.1001/jama.2019.3893

Mendelian disorders and monogenic traits result from combinations of variants in 1 or a few genes that have a large effect on the propensity for developing a certain disease or characteristic. In contrast, complex traits, such as eye color or cardiovascular disease, are determined by variations occurring in many genes that have smaller effect sizes and act over long periods of time, often in concert with environmental factors. The cumulative risk derived from aggregating contributions of the many DNA variants associated with a complex trait or disease is referred to as a polygenic risk score (also known as a genetic risk score). This JAMA Genomics and Precision Health article explains polygenic risk scores as determinants of an individual’s inherited risk for complex disease.How It Works

Polygenic risk scores reflect a mathematical aggregate of risk conferred by many DNA variants to estimate the likelihood of a specific outcome, such as disease onset in an individual.^1,2 The scores are the output of statistical models developed using data from large genome-wide association studies (GWAS). Models used to derive polygenic scores can vary in 3 important respects: the number of genetic variants considered, the specific type of statistical model used to combine risks associated with individual variations, and the ability of the score to generalize to the entire population. There are no universally agreed upon standards for developing polygenic risk score models; this can result in many different approaches for predicting risk in the same condition.

The number of DNA variants contributing to a polygenic score can range from dozens to thousands. The statistical model should reflect the underlying disease process. For example, in modeling a disease such as late-onset Alzheimer disease, polygenic scores must consider that disease risk varies with age rather than remaining constant throughout life. Additionally, polygenic scores should accurately reflect disease risk in the general population. For example, GWAS tend to only consider individuals with disease (cases) and individuals without disease (controls), while in reality, many conditions exist not as dichotomous disease vs no-disease states but on a continuous spectrum of disease (eg, from glucose intolerance to prediabetes to type 2 diabetes). To be clinically useful, GWAS-based polygenic scores should estimate an individual’s absolute risk for disease along this continuum and not just relative risk compared with a particular control group.¹ Furthermore, most GWAS have been conducted in populations of European descent, meaning that polygenic scores derived from these data may not generalize to the many individuals who do not share this specific ancestry.²Important Care Considerations

For an individual with a particular combination of genetic risk factors, polygenic risk scores estimate the probability of developing disease over time. Importantly, polygenic risk, like any prediction, is not deterministic, meaning that nongenetic factors influencing disease risk, such as lifestyle changes or therapeutic interventions, may prevent or modify the trajectory of disease predicted by polygenic risk. Conversely, a low polygenic risk score does not mean a disease cannot occur.

Commercial products offering polygenic risk scores directly to consumers are proliferating, most without the rigorous validation or oversight that would be required in clinical applications of risk prediction. Nevertheless, accurate, generalizable, polygenic risk scores have clear potential to inform clinical practice (Figure). For example, people with the highest polygenic risk scores for coronary heart disease are at greatest risk for myocardial infarction and may benefit from statin therapy.² Polygenic scores can also affect disease-screening strategies. For example, the US Preventive Services Task Force recommends against population-wide disease screening for prostate-specific antigen (PSA), arguing that the risk from false positives outweighs the benefits.³ Recently, however, a polygenic score has been shown to predict risk for aggressive prostate cancer, suggesting that genetic variation could be used to identify individuals who would benefit from PSA screening.⁴ A similar approach has been used to stratify women by breast cancer risk and provide a personalized recommendation about when to initiate screening mammography.⁵Figure.

Steps for Calculating Polygenic Risk Score

Polygenic scores have the potential to inform many dimensions of medical practice including disease risk stratification, screening strategies, diagnosis, therapeutic management, biomarker evaluation, and life planning.

It is critical to recognize that polygenic scores are not equivalent to diagnostic tests. Polygenic scores measure risk for developing disease, not whether an individual has or does not have disease. This distinction raises some subtle but important methodological considerations about how polygenic scores are derived and evaluated. First, most GWAS define individuals with a particular disease based on prevalence of disease in a population; however, polygenic scores derived from these data are used in an attempt to predict disease incidence not prevalence. Second, for related reasons, the appropriate metric to use in evaluating the performance of polygenic risk scores is a matter of some debate. Many investigators use diagnostic metrics, such as area under the curve or positive predictive value and negative predictive value, to quantify performance based on a score’s ability to discriminate between groups with disease vs groups without disease, but new methods are under development.^1,4,6Value

The value of polygenic risk scores cannot be assessed until the clinical utility has been established through research studies that evaluate their use in clinical practice and therapeutic trials. Companies are already offering these tests directly to consumers at costs ranging from one to a few hundred dollars. As whole-exome sequencing and whole-genome sequencing become less expensive and more widely available, it should be possible to compute any polygenic score from an individual’s genomic data at a low cost.Evidence Base

Polygenic risk scores for common complex disease have yet to become part of routine clinical care or be included in the practice guidelines of any major medical organization. Before polygenic scores can be translated into clinical practice they will need to be extensively validated in clinical and population-based cohorts for their ability to predict meaningful outcomes that can be modified with intervention.Bottom Line

Polygenic risk scores for common complex disease will become part of clinical care in the near future. That said, genetic susceptibility for complex conditions should not be viewed in isolation but be considered along with lifestyle and environmental factors in multivariate evaluation of disease risk.Section Editor: W. Gregory Feero, MD, PhD, Associate Editor, JAMA.

Corresponding Author: Leo P. Sugrue, MD, PhD, Department of Radiology and Biomedical Imaging, University California, San Francisco, 505 Parnassus Ave, L352, San Francisco, CA 94143 ([email protected]).

Published Online: April 8, 2019. doi:10.1001/jama.2019.3893

Conflict of Interest Disclosures: None reported.References1.Torkamani  A, Wineinger  NE, Topol  EJ.  The personal and clinical utility of polygenic risk scores.  Nat Rev Genet. 2018;19(9):581-590. doi:10.1038/s41576-018-0018-xPubMedGoogle ScholarCrossref2.Knowles  JW, Ashley  EA.  Cardiovascular disease: the rise of the genetic risk score.  PLoS Med. 2018;15(3):e1002546. doi:10.1371/journal.pmed.1002546PubMedGoogle ScholarCrossref3.US Preventive Services Task Force; Grossman  DC, Curry  SJ, Owens  DK,  et al.  Screening for prostate cancer: US Preventive Services Task Force recommendation statement.  JAMA. 2018;319(18):1901-1913. doi:10.1001/jama.2018.3710

ArticlePubMedGoogle ScholarCrossref4.Seibert TM, Fan CC, Wang Y, et al; PRACTICAL Consortium. Polygenic hazard score to guide screening for aggressive prostate cancer: development and validation in large scale cohorts. BMJ. 2018;360:j5757. doi:10.1136/bmj.j5757PubMedGoogle ScholarCrossref5.Maas P, Barrdahl M, Joshi AD, et al. Breast cancer risk from modifiable and nonmodifiable risk factors among white women in the United States. JAMA Oncol. 2016;2(10):1295-1302. doi:10.1001/jamaoncol.2016.1025
ArticlePubMedGoogle ScholarCrossref6.Tan CH, Fan CC, Mormino EC, et al; Alzheimer’s Disease Neuroimaging Initiative. Polygenic hazard score: an enrichment marker for Alzheimer’s associated amyloid and tau deposition. Acta Neuropathol. 2018;135(1):85-93. doi:10.1007/s00401-017-1789-4PubMedGoogle ScholarCrossref