Polygenic Risk Scores Do Not Improve Prediction of Heart Disease

By Will Boggs MD

February 20, 2020

NEW YORK (Reuters Health) - Polygenic risk scores do not meaningfully improve the prediction of coronary-artery disease (CAD), compared with clinical risk scores, according to a pair of new studies in JAMA.

"Despite the fact that we used sophisticated methods to develop a polygenic risk score with more than 1 million genetic variants, the incremental predictive value of the polygenic score over and above the currently used pooled cohorts equation was limited," Dr. Ioanna Tzoulaki of Imperial College London told Reuters Health by email.

Several genetic variants and single-nucleotide polymorphisms have been associated with CAD, and a number of polygenic risk scores for CAD prediction have been developed. Whether these scores provide added value for predicting CAD risk remains unclear.

In one study, Dr. Tzoulaki and colleagues used data from the UK Biobank to evaluate the potential of a polygenic risk score to improve risk prediction for CAD over and above pooled cohort equations used for risk stratification in U.S. clinical practice and the QRISK3 used in U.K. clinical practice.

Addition of the polygenic risk score for CAD showed a statistically significant improvement in discrimination, compared with pooled cohort equations, but the improvement in C statistic was only 0.02.

When the polygenic risk score for CAD was added to the pooled cohort equations model, predicted risk changed by less than 1% for 79.5% of participants and by 5% or more for only 1.1% of participants.

The net reclassification improvement (NRI) was 4.4% for cases and -0.4% for non-cases.

Similarly, the incremental value of the polygenic risk score for CAD over and above QRISK3 was statistically significant, but it amounted to a C statistic improvement of just 0.015.

"Genetic information is getting cheaper and cheaper, and polygenic risk scores are constantly improved," Dr. Tzoulaki said. "As this information can be measured at birth, with a single measurement and remains unchanged throughout the life course, it represents an attractive predictor of future health."

"However," she concluded, "the use of genetic testing in clinical care still needs evaluation in order to find those situations where genetic testing can help risk assessment over and above current means. For heart disease, conventional testing with age, sex, smoking status, cholesterol, blood pressure, and type 2 diabetes is still the most accurate method to assess future risk of heart disease and recommend preventive treatment with statins."

In the other study, Dr. Jonathan D. Mosley of Vanderbilt University School of Medicine in Nashville, Tennessee and colleagues used data from the Atherosclerosis Risk in Communities (ARIC) and Multi-Ethnic Study of Atherosclerosis (MESA) studies to evaluate the performance of a polygenic risk score for prediction of incident coronary-heart-disease events.

In both ARIC and MESA, the polygenic risk score was significantly associated with incident coronary heart disease.

The C statistics associated with the polygenic risk score were 0.549 for the ARIC cohort and 0.587 for the MESA cohort, and both improved with the addition of age and sex (to 0.669 and 0.672, respectively).

The addition of the polygenic risk score to the pooled equations predictor, however, did not significantly change the C statistic in either cohort.

Moreover, the polygenic risk score did not significantly improve calibration or classification accuracy compared with the conventional predictors.

"For the average, middle-aged patient, a polygenic risk score does not improve risk estimation for cardiovascular disease in middle-aged adults," Dr. Mosley told Reuters Health by email. "Measuring polygenic risk is not superior to the current clinical practice of measuring risk in terms of modifiable factors, such as smoking, blood pressure, and cholesterol levels, and does not provide additional information about the optimal treatment and prevention strategies for a patient."

"While the polygenic risk score did not appear to be useful for clinical risk prediction in this study, we believe that continued research into understanding how our genetics influences our disease risk will continue to be important for developing new and improved treatment and prevention strategies," he said.

Dr. Sadiya Sana Khan of Northwestern University Feinberg School of Medicine, in Chicago, who co-authored a linked editorial, told Reuters Health by email, "There has been a lot of excitement about the potential use of polygenic risk scores to enhance or improve precision in our ability to predict who may get disease, but we are finding that this may not be the path forward to improve outcomes for our patients."

"With the rapid growth of direct-to-consumer genetic testing, I am routinely asked by my patients to interpret results from 23andme and other companies," she said. "However, the evidence is now clear - there is no clinical utility or added value for risk prediction from current versions of the polygenic risk scores for cardiovascular disease."

"I would like to emphasize that these studies are focused on common genetic variants, and rare genetic variants that are linked with either high cholesterol or cardiomyopathies are still an important part of clinical care in cardiovascular medicine," Dr. Khan said.

"I would urge us to not 'throw the baby out with the bath water' as skepticism for utility of polygenic risk scores grows," he added. "The impact of cascade screening for autosomal dominant conditions, such as familial hyperlipidemia or dilated cardiomyopathy, can help identify individuals at significantly increased risk of cardiovascular disease early who may still be asymptomatic and ensure targeted preventive treatment in those at highest risk."

SOURCE: https://bit.ly/2V5Uznl, https://bit.ly/3244gEc and https://bit.ly/38ztBbH JAMA, February 18, 2020.