New data on the use of polygenic risk scores for predicting certain forms of cardiovascular disease raise questions about the validity of genetic scores in real-world settings.
Despite the advent of new technologies, new data is suggesting the use of traditional risk factors may still be the most effective predictor of cardiovascular disease.
A pair of studies from the University of Texas Southwestern Medical Center and Imperial College London found the addition of polygenic risk score may not be as beneficial as previously indicated for predicting certain conditions including coronary artery disease (CAD) and incident coronary heart disease.
“The idea that genetics may also be important for predicting common diseases has been a source of excitement over the past several years. But as an everyday clinical tool for predicting cardiovascular risk, human genetics isn’t there yet,” said Thomas Wang, MD, lead investigator of the study from UT Southwestern, in a statement. “We should not lose sight of traditional risk factors for assessing risk of cardiovascular disease, counseling about that risk, and strategizing on reducing it.”
Spurred by the increasing popularity of polygenic risk scores for disease prediction, two separate groups of investigators conducted studies examining the use of these scores versus traditional risk factors or clinical risk scores for predicting certain forms of cardiovascular. In the study conducted by Wang and colleagues, the condition of interest was coronary heart disease in a population of more than 7000 while the UK investigators from Imperial College London focused on CAD in a cohort of more than 365,000.
In Wang’s study, investigators used data from the Atherosclerosis Risk in Communities (ARIC) and Multi-Ethnic Study of Atherosclerosis (MESA) studies. Briefly, ARIC participants were between the ages of 45 and 64 years old followed from 1986-2015 and MESA participants were between the ages of 45 and 84 years old and were followed from 2000 through 2015—however, only participants with European ancestry were included in the current analysis as the existing polygenic risk score was derived from mostly people of European ancestry.
The primary outcome of the retrospective cohort study was the prediction of 10-year incident coronary heart disease events, which included myocardial infarctions, fatal coronary events, silent infarctions, revascularization procedures, or resuscitated cardiac arrest, assessed using measures of discriminative accuracy, calibration, and net reclassification improvement. Of note, discriminative accuracy was assessed using C statistic, calibration was defined as a comparison of the observed versus expected event probabilities, and reclassification improvement used a one-year risk threshold of 7.5%.
Results of the analyses indicated polygenic risk score was significantly associated with ten-year coronary heart disease incidence in both the ARIC and MESA study cohorts—with investigators noting hazard ratios per SD increment of 1.24 (95% CI, 1.15-1.34) and 1.38 (95% CI, 1.21-1.58), respectively. When examining the addition of polygenic risk score to the pooled cohort equations, results indicated the addition did not significantly increase the C statistic in the ARIC (change in C statistic, −0.001; 95% CI, −0.009 to 0.006) or MESA (0.021; 95% CI, −0.0004 to 0.043) cohorts. Likewise, when examining 10-year risk, adding polygenic risk score to the pooled cohort equations did not results in significant improvement in either cohort (ARIC, NRI: 0.018, 95% CI, −0.012 to 0.036) or (MESA, NRI: 0.001, 95% CI, −0.038 to 0.076).
In the Imperial College London study, which was led by Ioanna Tzoulaki, PhD, of the Department of Epidemiology and Biostatistics at the ICL School of Public Health, investigators used the data from UK Biobank participants including 15,947 participants with prevalent CAD. For their comparison, CAD cases were age and sex frequency-matched with controls from within the same population database and after creating a separate cohort of 352,660.
Similarly to the study from Wang and colleagues, this analysis had a primary outcome of prediction of CAD assessed via discrimination, calibration, and reclassification using a risk threshold of 7.5%. Unlike the analysis by Wang et al., the median follow-up time in the current study was 8 years.
Upon analysis, 6722 incident CAD events occurred in the 352,660 participant cohort. Results indicated the C statistic for CAD discrimination for polygenic risk score, pooled cohort equations, and both combined was 0.61 (95% CI, 0.60-0.62), 0.76 (95% CI, 0.75-0.77), and 0.78 (95% CI, 0.77-0.79), respectively. Investigators noted a change in C statistic of 0.02 (95% CI, 0.01 to 0.03) when comparing pooled cohort equations and the use of both polygenic risk score and pooled cohort equations.
Overall, the addition of polygenic risk score to pooled cohort equations resulted in net reclassification for 4.4%(95% CI, 3.5%-5.3%) of cases of -0.4% (95% CI, −0.5% to −0.4%) of noncases—this correlates to an overall net reclassification improvement of 4.0% (95% CI, 3.1%-4.9%).
While multiple limitations were noted by investigators of both studies, investigators from both concluded the use of genetic information and polygenic risk scores may not improve prediction in the populations examined and the concept requires more research before implementation.
These studies, “Predictive Accuracy of a Polygenic Risk Score Compared With a Clinical Risk Score for Incident Coronary Heart Disease,” and “Predictive Accuracy of a Polygenic Risk Score—Enhanced Prediction Model vs a Clinical Risk Score for Coronary Artery Disease,” were published in JAMA.