This article has been edited to clarify that both the Standard and Enhanced PRS Sets outperformed previously published PRS sets.
NEW YORK – Polygenic risk scores (PRS) for coronary artery disease (CAD) and breast cancer may reflect levels of risk similar to that of rare monogenic mutation carriers, which could argue for the inclusion of PRS in wider screening programs.
Researchers from the UK-based firm Genomics Plc evaluated PRS for 28 diseases and 25 quantitative traits using data from the UK Biobank (UKB) Polygenic Risk Score Release. They evaluated these scores' performance compared to a set of previously released PRS, also from the UKB, and for their possible clinical utility.
The company said the similarity in risk between PRS and pathogenic mutations in certain cases advances the argument for including PRS in screening programs. The firm published its results, along with a benchmarking software tool for directly comparing the performance of different PRSs for the same disease or trait, in the journal PLOS One.
Overall both the Standard and Enhanced PRS Sets outperformed previously published PRS sets, in terms of correlations between PRS and disease occurrence and between PRS and the effect size of quantitative traits such as body mass index, total triglycerides, and resting heart rate
"Basically, what the paper shows is that those scores are more powerful than any of the others that have been developed," said Peter Donnelly, CEO and cofounder of Genomics Plc.
The strongest examples of equivalent risks in the study came from CAD and breast cancer, which the researchers assessed by evaluating odds ratios and area under the curve as performance metrics. The study found that PRS indicating a high disease risk reflects a risk level similar to that of people with known single-gene pathogenetic mutations.
But people with high risk due to the accumulation of common, lower-risk variants that are captured by PRS outnumber people who are carriers of rare high-risk variants, "often massively so," Donnelly said, suggesting that adding PRS to some screening programs might significantly improve diagnosis rates.
For instance, the researchers examined individuals in the UKB who were not taking statins and found the top 8 percent of individuals with high CAD PRS had a similar level of CAD risk as people carrying genetic variants associated with familial hypercholesterolemia (FH), a known risk factor for developing CAD. The study estimated that there were between 18 and 29 times the number of CAD events, depending on age, among the individuals within that top 8 percent as compared to FH carriers, suggesting that PRS can identify more at-risk people.
Donnelly noted that while FH mutations are currently included in many CAD screening programs, PRS are not.
In Germany, for example, the southern state of Bavaria initiated an FH carrier screening program in 2021 as part of its broader DigiMed program to advance predictive, preventive, personalized, and participatory public health measures by combining datasets of patients diagnosed with atherosclerotic diseases, such as coronary heart disease and stroke, or with genetic risk factors. The FH screening program, called VRONI, has since spread throughout the country and across Europe.
"It seems impossible to argue that you should go out of your way to look out for one group of people because they happen to have a genetic change that makes [disease] more [likely] and then ignore another group of people who are at exactly the same level of risk because of their genetics," he said.
Similarly, individuals in the top 0.3 percent of PRS for breast cancer in the study showed equivalent lifetime risks as carriers of pathogenic BRCA1 and BRCA2 mutations.
Jonathan Mosley, a professor of medicine at the Vanderbilt University's Center for Precision Medicine, who was not involved in the study but whose research frequently involves PRS, cautioned that one known issue in using the UKB to benchmark PRS is that it consists entirely of observational data and the outcomes are not gold standard assessments for any given disease, as opposed to a cohort study where complete outcomes are actively ascertained at frequent intervals.
"In the end," he said via email, "it would be important to benchmark the PRS against a number of different study populations, as populations vary with respect to environmental exposures and other compositional features."
Mosley also pointed out that the study focuses largely on association metrics such as odds ratios and cumulative incidence plots and less on metrics related to discrimination, such as concordance statistics, or on reclassification statistics, which might better help understand any potential clinical utility of PRS.
Further, with respect to the findings regarding CAD and breast cancer, Mosley said that direct comparisons were complicated by the "fundamental difference" of distinct biological mechanisms measured via PRS versus pathogenic mutations.
"The risk associated with FH variants is linked to a specific underlying disease mechanism, which can be targeted, while the PRS does not point to a targetable disease mechanism," Mosley said. "So the utility of the risk estimates from PRS versus pathogenic variants is not necessarily comparable. If a person carries a high-risk FH variant, you will want to direct therapy towards lowering their cholesterol levels; a PRS does not provide a specific therapeutic target."
Donnelly acknowledged that PRS currently tend not to shed light on specific biological mechanisms but said they can still function as good predictors of disease by aggregating many tiny effects from different biological processes. Finding PRS that capture specific aspects of biology is a "big hope" in the field, he said.
"It's certainly something we look at in some of the work we do with drug companies," he said, adding that the hope is to leverage PRS to identify which individuals will be most likely to respond to specific therapies.
Still, Genomics' findings, particularly with respect to FH mutation carriers, are in line with findings from other groups. Last year, genomics software company Allelica published a set of ancestry-specific PRS for CAD, in which the company also identified high-PRS individuals with equivalent risk as FH mutation carriers, even when their LDL-C levels were not elevated.
"These individuals with high PRS escape traditional risk assessments and walk around thinking their LDL-C levels are fine, while in reality they are at extremely high risk [of CAD]," Allelica CEO Giordano Bottà said in an email.
The PRS published in Genomics' study, Bottà said, "confirm what the field already knew: We can stratify disease based on common variants, and in some cases, the risk conferred is equivalent to monogenic mutations."
Bottà did caution, however, that PRS must be evaluated on a case-by-case basis, as there are nuances to consider in calculating them, and while they might provide lifesaving results in certain disease contexts, they could prove "completely useless" in others.
Despite the ongoing debate regarding the clinical utility of PRS, they are finding their way into various aspects of the overall molecular diagnostics industry, from diagnostic and screening assays to pharmacogenetic research and clinical trial design.
Myriad Genetics, for example, added its RiskScore PRS test to its MyRisk hereditary cancer risk test and has forged numerous partnerships aimed at screening women for breast cancer risk, including with Onsite Women's Health and SimonMed Imaging.
Genomics Plc itself has been studying the clinical utility of PRS as well as the feasibility of including them in routine clinical care. Earlier this year, the company published a study that found high levels of acceptance and understanding of PRS among both physicians and patients in the primary care setting.
The study addressed the concern that the complexity of PRS would make using them confusing and difficult for both patients and physicians. However, results showed that, given a measure of education concerning what PRS are and how they work, both patients and practitioners reported feeling confident in using them.
"I think we can allay a lot of the fears that some people have had about putting [PRS] into practice," Donnelly said.
Another perennial issue in the PRS field is obtaining scores that work well in populations of non-European ancestry, as these are generally underrepresented in genome-wide association studies. As in past studies, all the PRS in the current PLOS One study performed best in people of European genetic heritage and declined for other populations, with some of the biggest declines seen in those of African ancestry.
Genomics Plc continues to seek new partnerships across the life sciences spectrum in its pursuit of nudging PRS into routine clinical and pharmaceutical practice, and Donnelly said that the company hopes to be able to announce several new collaborations within the coming six months or so.
"Going forward, we certainly want to be very focused on continuing to be at the cutting edge of having the most powerful tools," he said.