JAMA Studies: Genetic Tests for Heart Disease Don't Have Much Predictive Power

February 18, 2020

Two studies in JAMA produce meh results for polygenic risk scores

Genetic test results don’t add much, if anything, to the risk factor predictions about who will develop cardiovascular disease, according to two studies published in this week’s JAMA.

The study results, while far from the final word, may dent further dent the reputation of genetic testing and give ammunition to skeptics who believe it’s not ready for clinical use. Payers and providers taking on financial risk may have another reason to tap the brakes.

“The findings of both these articles lend further support to the lack of meaningful improvement in risk stratification for CAD [coronary artery disease] in different populations of middle-aged individuals of European ancestry when genome-wide risk scores are added to pooled cohort equations,” says an accompanying editorial, which acknowledged that the utility of the tests in a younger or more diverse population remains an open question.

One of the studies used a polygenic risk score developed from a case-control study about 16,000 CAD cases with matched controls. The researchers, most of whom are at the Imperial College London’s School of Public Health,  then applied it to a 350,000 individuals from the UK Biobank, 6,272 of whom had a heart attack or some other CAD event during a median follow-up period of eight years. When the polygenic risk score for CAD was used, the predicted risk changes by less than 1% for nearly 80% of the participants. At a risk threshold of 7.5%, 526 of the 6,272 (8.4%) were correctly reclassified to the higher risk category but 240 (4%) were incorrectly moved to a lower risk category.  Among the 346,388 noncases, more individuals were incorrectly moved up to a high-risk category than correctly moved down to a lower one  (6,723 vs. 5,284) when the polygenic risk score was used.

Lead author Joshua Elliott and his colleagues characterized the number of people meaningfully changing risk category as “relatively small” and noted the “worse reclassification” among the noncases.

The other study used data from the 4,847 adults in the Atherosclerosis Risk in Communities  study and 2,390 people participating in the Multi-Ethnic Study of Atherosclerosis. The research team, led by Jonathan D. Mosley at Vanderbilt, tested how the addition of a polygenic risk affected prediction of CHD events (heart attacks, silent infarctions, revascularization procedures) over a 10-year period. They found that the testing did not “significantly improve classification accuracy” in either study and, furthermore, among those who developed CHD the reclassification were incorrect about 80% of the time.

Mosley and his colleagues noted that their findings are in keeping with the frequent mismatch between statistical association and predictive performance for risk biomarkers. They noted that the odds ratio associated with being in the top 5% of the polygenic risk score (about 4) is similar to other biomarkers like C-reactive protein and homocysteine that “have been shown to have similarly modest predictive utility.”