ATLANTA — A measure that combines a polygenic risk score, social determinants of health, and clinical features outperformed any of its components alone in predicting risk of incident coronary heart disease, researchers reported at the American College of Cardiology's annual meeting on Sunday.
By being able to better predict which patients are likely to develop incident coronary heart disease (CHD), physicians hope to intervene earlier and prevent the condition from developing. Scientists have previously developed a PRS for coronary artery disease encompassing some 6.6 million SNPs, though some other investigators have cautioned that PRS alone may only minimally improve physicians' ability to identify at-risk individuals.
At ACC, Mohammadreza Naderian, a cardiologist and member of the atherosclerosis and lipid genomics laboratory at the Mayo Clinic, said he and his colleagues suspected that using a PRS and a score encompassing social determinants of health and lifestyle factors (SDLS) could improve the accuracy of heart disease risk predictions.
"Despite their important role in disease causation, social determinants of health and polygenetic risk scores are not included in risk prediction models," Naderian said during a presentation at the meeting. He added that these factors haven't been used because their ability to improve the accuracy of heart disease risk predictions isn't clear.
Naderian and colleagues used genetic, lifestyle, and other data on more than 471,000 people from the UK Biobank without CHD at the time they were enrolled to examine three measures to predict heart disease risk: pooled cohort equations (PCE), an established clinical risk tool based on known cardiovascular risk factors like age, sex, blood pressure, and cholesterol levels; a MetaGRS PRS for CHD based on about 1.7 million SNPs; and an SDLS score based on 100 covariates falling into nine discrete domains, such as education level, housing, and income.
Within the UK Biobank cohort, the median age was 56 years, about 45 percent of participants were male, 95 percent of were of European ancestry, and 14,339 of them had a CHD event, either fatal or non-fatal, within 10 years.
Both the PRS and SDLS score correlated with clinical risk of CHD, Naderian said.
A one-standard deviation increase in the PRS was associated with 1.5-fold increase in disease risk, and individuals in the top quantiles of PRS risk had a 78 percent higher CHD risk, compared to those in the middle quantiles.
Similarly, a one-standard deviation increase in the SDLS score was also associated with a 1.5-fold increase in disease risk, and individuals in the top quantiles of SDLS risk had a 115 percent higher CHD risk, compared to those in the middle quantiles. Social and lifestyle factors contributing to increased risk included things like snoring, poor sleep, low educational attainment, and slow walking pace.
The PRS and SDLS scores, Naderian noted, were independent measures of CHD risk and, when, combined, had an additive effect. He and his colleagues bundled the PRS and SDLS score with the established PCE measure into a comprehensive model that they found had even better predictive ability than any one score alone.
In the UK Biobank cohort, the PRS score alone could classify heart disease with an AUC of 0.61, while SDLS had an AUC of 0.71 and PCE had an AUC of 0.74. The comprehensive score, meanwhile, had an AUC of 0.77.
For some members of the cohort, when researchers ran the comprehensive prediction model, adding the PRS and SDLS score to PCE, it reclassified their risk. Fifty-four percent of participants remained at low risk and 32 percent remained at high risk, but 8 percent of the cohort was reclassified from high risk to low risk and 6 percent from low risk to high.
"We found that several social determinants of health were significantly associated with the risk of incident coronary heart disease and the effect of social determinants of health on the polygenic risk score of coronary heart disease was independent and additive," Naderian said. "Joint modeling of social determinants of health and polygenic risk score increased the predictive accuracy of estimating coronary heart disease."
According to Naderian, he and his colleagues have replicated their work in the US-based All of Us cohort. The UK Biobank cohort includes individuals who are predominantly white and of European ancestry, while the All of Us cohort aims to reflect the diversity of the US population.
He noted, though, that definitions of social determinants of health varies by cohort and a unified definition is needed so that consistent data are collected.