Multi-ancestry polygenic risk scores were associated with substantially higher odds of type 2 diabetes across diverse ancestry groups and may improve risk stratification beyond traditional clinical factors, according to findings.
Researchers developed and validated single-ancestry and multi-ancestry polygenic risk scores (PRSs) using genome-wide association study (GWAS) data from European, African or African American, Admixed American, South Asian, and East Asian ancestry groups. The overall analysis included 409,959 participants with type 2 diabetes and nearly 2 million controls.
The GWAS component included 359,819 participants with type 2 diabetes and 1.8 million controls from 125 type 2 diabetes GWAS data sets. Researchers used 10,992 participants with type 2 diabetes and 31,792 controls for model training and validated the scores in an independent cohort comprising 39,148 participants with type 2 diabetes and 125,824 controls, with at least 4 validation cohorts per ancestry group.
Compared with single-ancestry PRSs, multi-ancestry PRSs showed larger effect sizes across all ancestry groups. Per standard deviation increase in PRS, the odds of type 2 diabetes ranged from 1.73 times the odds in African or African American groups to 2.82 times the odds in Admixed American groups.
Associations were strongest at the extremes of the PRS distribution. Compared with participants in the interquartile range, those in the 97.5th percentile had 3.43 times the odds of type 2 diabetes in African or African American groups, 7.47 times the odds in Admixed American groups, 6.62 times the odds in East Asian groups, 6.25 times the odds in European groups, and 4.50 times the odds in South Asian groups.
Predictive performance varied substantially by ancestry. Single-ancestry PRSs showed the strongest discrimination in European and East Asian populations, with incremental area under the curve (AUC) values ranging from 0.07 to 0.14 in European cohorts and 0.02 to 0.16 in East Asian cohorts. Performance remained lower in African or African American, Admixed American, and South Asian populations, where incremental AUC values ranged from 0.02 to 0.06.
The investigators noted that this disparity likely reflects the continued overrepresentation of European ancestry populations in GWAS data sets. European GWAS cohorts were at least 12 times larger than cohorts representing African or African American, Admixed American, and South Asian populations.
The study also benchmarked the models against 55 previously published type 2 diabetes PRSs from the Polygenic Score Catalog. The multi-ancestry PRSs demonstrated significantly better predictive performance in African or African American, Admixed American, and European populations, though not in East Asian or South Asian populations.
In longitudinal analyses restricted to African or African American, Admixed American, and European cohorts, 2,074 of 17,920 participants developed incident type 2 diabetes over a median follow-up of 7.4 years, while 15,846 remained disease-free. Compared with the lowest PRS tertile, the highest tertile was associated with higher hazards of incident type 2 diabetes across all 3 ancestry groups.
The PRS remained independently associated with incident diabetes following adjustment for a 9-component clinical risk score that included age, sex, parental history of type 2 diabetes, body mass index, systolic blood pressure, high-density lipoprotein cholesterol, total cholesterol, triglycerides, and random glucose concentration.
Adding PRS information modestly improved concordance in Admixed American and European populations. The researchers also found that participants with normal glucose concentrations but in the highest PRS tertile had diabetes-free survival risks similar to those of participants in the lowest PRS tertile who had random glucose concentrations of at least 140 mg/dL, a threshold consistent with prediabetes.
Risk increased substantially when elevated glucose concentrations and high PRS values occurred together. Compared with participants in the lowest PRS tertile who had random glucose concentrations below 140 mg/dL, those with both high PRS values and elevated glucose concentrations had hazard ratios for incident type 2 diabetes of 5.35 in African or African American populations, 13.82 in Admixed American populations, and 22.43 in European populations.
In secondary analyses using All of Us electronic health record data from African or African American, Admixed American, and European populations, higher PRSs were associated with earlier onset of type 2 diabetes by up to 2.3 years.
Analyses of diabetes-related complications were restricted to African or African American, Admixed American, and European groups because these populations had sufficient representation in the All of Us cohort. Higher PRS values were associated with increased odds of diabetic retinopathy, diabetic nephropathy, proliferative diabetic retinopathy, and end-stage diabetic nephropathy across all 3 groups. An association with coronary artery disease was observed only in the Admixed American group.
The investigators cautioned that several limitations remain. Although the multi-ancestry PRSs incorporated more diverse GWAS data than prior models, single-nucleotide polymorphism effect estimates remained heavily influenced by large European cohorts. The authors also noted that the use of discrete ancestry categories incompletely captures genetic heterogeneity, particularly among admixed populations.
In addition, the study evaluated predictive performance rather than clinical implementation, and whether PRS-guided care improves patient outcomes remains unknown. Ongoing randomized trials, including studies evaluating PRS disclosure in primary care settings alongside conventional risk assessment, are assessing how these tools may be incorporated into routine care.
“These validated, publicly available PRSs can improve risk stratification for type 2 diabetes onset and complications across diverse ancestries,” wrote lead study author Alicia Huerta-Chagoya, PhD, of the Broad Institute of MIT and Harvard, Massachusetts General Hospital Center for Genomic Medicine, and Massachusetts General Hospital Diabetes Unit, and colleagues.
Disclosures: The researchers reported no competing interests.