Cross-Ancestry Glaucoma Genetic Risk Score in Hispanic Veterans in the Million Veteran Program

Abstract

Purpose : Risk stratification for early detection of primary open-angle glaucoma (POAG) may mitigate severe disease outcomes, including irreversible vision loss, for millions globally. Genetic risk scores (GRS), which sum allelic effects, promise clinical utility for prediction of many complex diseases but are not generalizable across different ancestral groups. Moreover, although POAG burden is higher in Hispanic (HIS) individuals, most POAG genetic studies have been reported in individuals of non-HIS European ancestry (EUR). We aimed to evaluate POAG GRS performance in HIS Veterans in the Million Veteran Program (MVP).

Methods : We calculated GRS based on 127 published POAG-associated variants in HIS (382 cases and 2965 controls) and EUR (3382 cases and 58811 controls) Veterans in the MVP. Ancestry was defined using the Harmonized Ancestry and Race/Ethnicity (HARE) algorithm. We tested for association between the 127-variant (unweighted and weighted by published cross-ancestry effect estimates) GRS and POAG via logistic regression-based analyses with unadjusted and adjusted (age, sex, and 10 sample-specific principal components) models. We compared POAG case classification across GRS deciles and evaluated GRS model performance with receiver operator characteristic (ROC) curves and area under the curve (AUC) estimates. We also performed subset analyses of HIS with <50% EUR global genetic ancestry (GGA; determined via ADMIXTURE program).

Results : Unweighted and weighted GRS were significantly associated with POAG in HIS and EUR Veterans (p<0.05). Between the top decile and bottom 90%, we observed ~3-fold higher odds of POAG case classification for unweighted (odds ratios [OR], 95% confidence intervals [CI]: HIS: 2.70 [2.03-3.56]; EUR: 2.74 [2.51-2.98]) and weighted (HIS: 3.11 [2.35-4.07]; EUR: 3.03 [2.78-3.29]) GRS. We detected no statistically significant difference between GRS performance in ROC comparisons between HIS and EUR Veterans (AUC range 0.65-0.67). In GGA subset analyses of 220 cases and 1486 controls, ROC curves were comparable to the full HIS dataset.

Conclusions : POAG studies prioritizing diverse, non-EUR groups are crucial to ensure equitable clinical introduction of GRS for POAG prediction. We confirmed that a GRS of 127 POAG risk variants performed similarly in HIS and EUR MVP Veterans, regardless of majority EUR/non-EUR GGA.

This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.