The clinical utility of a glaucoma polygenic risk score in four European-ancestry cohorts

Abstract

Purpose : Glaucoma is the leading cause of irreversible blindness worldwide. Given the lack of curative glaucoma, early screening and intervention is essential. Determining genetic predisposition and risk for glaucoma may facilitate screening efforts. In this prospective cohort study, we evaluate the utility of a polygenetic risk score (PRS) for identification of high-risk subgroups in 4 European-ancestry cohorts: the Rotterdam Study (RS), the U.S. based Nurses’ Health Study (NHS), NHS2, and the Health Professionals Follow-up Study.

Methods : We applied Lassosum, a penalized regression framework, to imputed genotype and phenotype data for 449,186 participants in the UK Biobank to estimate PRS weights, using genome-wide association study summary statistics from a large cross-ancestry meta-analysis of primary open angle glaucoma (POAG). A total of 144,020 SNPs with non-zero beta’s were identified. We calculated a standardized PRS for each participant and estimated the hazard ratio (HR) for POAG and this PRS using multivariable Cox proportional hazard models, correcting for age, sex, intraocular pressure (IOP), several non-ocular risk factors, and family history of glaucoma. A pooled HR was calculated using a generic inverse variance random effects model. Additionally, several prediction models for POAG were made using logistic regression, the accuracy of which were tested by calculating the area under the receiver operating characteristic curve (AUC).

Results : In total, 1,160 cases and 48,966 controls were included in our analyses. The pooled hazard ratio for one standard deviation increase in PRS was 2.28 (95% CI, 1.40, 3.70; P < 0.001). In the RS the AUC (95% CI) for a model including baseline age, sex, IOP and family history was 0.61 (0.57, 0.66). Notably, a model including the PRS alone achieved an AUC of 0.77 (0.75, 0.82), and a model with both the PRS and the previously mentioned covariates achieved an AUC of 0.79 (0.75, 0.82). In the US-cohorts, the AUC for a model using the above mentioned covariates was 0.77 (0.75, 0.79), and adding the PRS to the model increased the AUC to 0.79 (0.77, 0.80).

Conclusions : Including a glaucoma-specific PRS to a predictive model increases the accuracy of glaucoma prediction up to 80%. Given the insidious onset and relatively low prevalence of glaucoma, the use of genetic data to identify high-risk individuals may help with more targeted screening efforts.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.