Predictive Utility of Polygenic Risk Scores for Coronary Heart Disease in Three Major Racial and Ethnic Groups

Abstract

Because polygenic risk scores (PRSs) for coronary heart disease (CHD) are derived from mainly European ancestry (EA) cohorts, their validity in African ancestry (AA) and Hispanic ethnicity (HE) individuals is unclear. We investigated associations of ''restricted'' and genome-wide PRSs with CHD in three major racial and ethnic groups in the U.S. The eMERGE cohort (mean age 48 5 14 years, 58% female) included 45,645 EA, 7,597 AA, and 2,493 HE individuals. We assessed two restricted PRSs (PRS Tikkanen and PRS Tada ; 28 and 50 variants, respectively) and two genome-wide PRSs (PRS metaGRS and PRS LDPred ; 1.7 M and 6.6 M variants, respectively) derived from EA cohorts. Over a median follow-up of 11.1 years, 2,652 incident CHD events occurred. Hazard and odds ratios for the association of PRSs with CHD were similar in EA and HE cohorts but lower in AA cohorts. Genome-wide PRSs were more strongly associated with CHD than restricted PRSs were. PRS metaGRS , the best performing PRS, was associated with CHD in all three cohorts; hazard ratios (95% CI) per 1 SD increase were 1.53 (1.46-1.60), 1.53 (1.23-1.90), and 1.27 (1.13-1.43) for incident CHD in EA, HE, and AA individuals, respectively. The hazard ratios were comparable in the EA and HE cohorts (p interaction ¼ 0.77) but were significantly attenuated in AA individuals (p interaction ¼ 2.9 3 10 À3 ). These results highlight the potential clinical utility of PRSs for CHD as well as the need to assemble diverse cohorts to generate ancestry-and ethnicity PRSs.

Validation of an Integrated Risk Tool, Including Polygenic Risk Score, for Atherosclerotic Cardiovascular Disease in Multiple Ethnicities and Ancestries

“…We proceeded to investigate relative performance patterns within ethnicity (Figure 2, Supplementary Table 3) and ancestry ( Supplementary Figure 2, Supplementary Table 4) groups (groups with >50 ASCVD cases across men and women are shown, metrics for groups with fewer cases can be found in Supplementary Tables 3-4 Supplementary Figure 2b-c, Supplementary Tables 3-4). We also find that ASCVD-IRT has, across the same large ethnicity and ancestry groups, a larger NRI than that of a tool constructed in the same way as ours, but using an alternative PRS previously shown to have good cross-ancestry performance (the coronary artery disease PRS of Inouye et al 25,26 , Supplementary Figure 3).…”

Highly elevated polygenic risk scores are better predictors of myocardial infarction risk early in life than later

“…As shown by others [ 39 – 41 ], the frequency distribution of polygenic risk scores derived with weights observed in one ancestry group are generally shifted in other ancestry groups due to cumulative subtle differences in allele frequencies. For example, the metaGRS derived by Inouye et al [ 22 ] is higher by more than a standard deviation unit in each of the African-, East-Asian-, and South-Asian ancestry groups in the UKB (Fig.…”

An integrated polygenic and clinical risk tool enhances coronary artery disease prediction

“…In contrast, we have shown here that PRS prediction varies by age in men. This opens the possibility that once sufficient data are available, separate PRSs may be constructed for different groups (defined for example by age, sex, or ancestry background) which will likely further improve predictive power 29 . Even without these more sophisticated statistical approaches, the predictive power of PRSs will also increase as more population-scale data becomes available.…”