Skip to main content
NCBI home page
JAMA Network logoLink to JAMA Network
. 2024 Mar 14:e240151. Online ahead of print. doi: 10.1001/jamaophthalmol.2024.0151

Polygenic Risk Scores for Glaucoma Onset in the Ocular Hypertension Treatment Study

Rishabh K Singh 1,2, Yan Zhao 3, Tobias Elze 2, John Fingert 4, Mae Gordon 5, Michael A Kass 5, Yuyang Luo 3, Louis R Pasquale 6, Todd Scheetz 4, Ayellet V Segrè 3,7, Janey L Wiggs 3,7, Nazlee Zebardast 3,✉
PMCID: PMC10941023  PMID: 38483402

Key Points

Question

How does a polygenic risk score improve prediction of conversion from ocular hypertension to primary open-angle glaucoma (POAG)?

Finding

In this post hoc analysis of the Ocular Hypertension Treatment Study including 1009 participants, polygenic risk score was an independent risk factor for POAG that improved prediction of POAG onset.

Meaning

These findings suggest background genetic risk for POAG may help stratify patients with ocular hypertension and help clinicians and patients make decisions about early treatment.

This post hoc analysis of the Ocular Hypertension Treatment Study assesses whether a polygenic risk score improves prediction of primary open-angle glaucoma onset in patients with ocular hypertension.

Abstract

Importance

Primary open-angle glaucoma (POAG) is a highly heritable disease, with 127 identified risk loci to date. Polygenic risk score (PRS) may provide a clinically useful measure of aggregate genetic burden and improve patient risk stratification.

Objective

To assess whether a PRS improves prediction of POAG onset in patients with ocular hypertension.

Design, Setting, and Participants

This was a post hoc analysis of the Ocular Hypertension Treatment Study. Data were collected from 22 US sites with a mean (SD) follow-up of 14.0 (6.9) years. A total of 1636 participants were followed up from February 1994 to December 2008; 1077 participants were enrolled in an ancillary genetics study, of which 1009 met criteria for this analysis. PRS was calculated using summary statistics from the largest cross-ancestry POAG meta-analysis, with weights trained using 8 813 496 variants from 449 186 cross-ancestry participants in the UK Biobank. Data were analyzed from July 2022 to December 2023.

Exposures

From February 1994 to June 2002, participants were randomized to either topical intraocular pressure–lowering medication or close observation. After June 2002, both groups received medication.

Main Outcomes and Measures

Outcome measures were hazard ratios for POAG onset. Concordance index and time-dependent areas under the receiver operating characteristic curve were used to compare the predictive performance of multivariable Cox proportional hazards models.

Results

Of 1009 included participants, 562 (55.7%) were female, and the mean (SD) age was 55.9 (9.3) years. The mean (SD) PRS was significantly higher for 350 POAG converters (0.24 [0.95]) compared with 659 nonconverters (−0.12 [1.00]) (P < .001). POAG risk increased 1.36% (95% CI, 1.08-1.64) with each higher PRS decile, with conversion ranging from 9.52% (95% CI, 7.09-11.95) in the lowest PRS decile to 21.81% (95% CI, 19.37-24.25) in the highest decile. Comparison of low-risk and high-risk PRS tertiles showed a 2.0-fold increase in 20-year POAG risk for participants of European and African ancestries. In the subgroup randomized to delayed treatment, each increase in PRS decile was associated with a 0.52-year (95% CI, 0.01-1.03) decrease in age at diagnosis (P = .047). No significant linear association between PRS and age at POAG diagnosis was present in the early treatment group. Prediction models significantly improved with the addition of PRS as a covariate (C index = 0.77) compared with the Ocular Hypertension Treatment Study baseline model (C index = 0.75) (P < .001). Each 1-SD higher PRS conferred a mean hazard ratio of 1.25 (95% CI, 1.13-1.44) for POAG onset.

Conclusions and Relevance

Higher PRS was associated with increased risk for POAG in patients with ocular hypertension. The inclusion of a PRS improved the prediction of POAG onset.

Trial Registration

ClinicalTrials.gov Identifier: NCT00000125

Introduction

Glaucoma is the leading cause of irreversible blindness worldwide, with primary open-angle glaucoma (POAG) accounting for nearly 69% of global prevalence. Early identification of high-risk individuals and therapeutic intervention is necessary to prevent glaucoma-related vision loss. While POAG may present with or without elevated intraocular pressure (IOP), IOP remains the only modifiable risk factor for its development and progression. The Ocular Hypertension Treatment Study (OHTS) determined that early treatment with topical IOP-lowering therapy reduced the cumulative incidence of POAG among patients with ocular hypertension (OHTN). A prediction model including age, IOP, central corneal thickness (CCT), vertical cup-disc ratio (VCDR), and visual field pattern standard deviation (PSD) was developed for POAG risk stratification. This model was confirmed in the European Glaucoma Prevention Study and the Diagnostic Innovations in Glaucoma Study.

POAG is a highly heritable disease, with 127 identified common risk variants to date. While each variant individually has a small effect, they can be used in the aggregate to measure overall genetic burden with a polygenic risk score (PRS). Genotyping can be done inexpensively once per lifetime and enable the calculation of a PRS for a wide range of diseases. PRS may be calculated earlier in life, prior to the impact of environmental or phenotypic risk factors, and recalculated as knowledge of the genetic architecture for disease improves. Since the absolute score for a disease varies with the number of single-nucleotide variants included, PRS is often transformed to a z score or percentile. Prior work has shown PRS to be associated with risk and outcomes for a range of complex diseases, including coronary artery disease, obesity, atrial fibrillation, and type II diabetes. In ophthalmology, PRS have been developed for risk stratification or enhancement of prediction models in macular degeneration, diabetic retinopathy, and POAG. High POAG PRS has been previously associated with an increased risk of glaucoma, increased risk of development of severe disease, maximum recorded IOP, and younger age at diagnosis. In patients with existing POAG diagnosis, high PRS was also associated with increased rates of visual field progression and retinal nerve fiber layer thinning.

While the OHTS prediction model identified key risk factors for conversion to POAG, more than two-thirds of OHTS participants never progressed from OHTN to POAG. A prior analysis has identified a variant in TMC01 locus (rs4656461) conferring a 12% absolute increase in cumulative POAG incidence in a subgroup of participants. PRS measures a larger component of genetic burden; however, its association with glaucoma risk in patients with OHTN remains unknown. In this study, we use available genetic data from OHTS to develop a PRS and characterize whether underlying genetic risk improves POAG risk stratification beyond demographic factors and ocular biomarkers (age, IOP, CCT, VCDR, and PSD) in the OHTS prediction model.

Methods

Data Collection

This study was a secondary analysis of data from OHTS. OHTS was a randomized clinical trial assessing the safety and efficacy of early treatment with topical IOP-lowering therapy among patients with OHTN. Details of the OHTS study methods have been reported previously. This study was exempt from institutional review board approval as we used only deidentified, retrospective data. All methods adhered to the tenets of the Declaration of Helsinki for research involving human participants. The Enhancing the Quality and Transparency of Health Research reporting guidelines for post hoc analyses of randomized clinical trials were followed.

In the OHTS, 1636 participants with OHTN were recruited from 22 sites in the US. Prior to randomization, each participant was required to have Humphrey visual field (HVF) testing and optic nerve head photography without signs of glaucomatous damage. Randomization of participants to observation or treatment began in 1994. Participants were followed up with HVFs semiannually and optic nerve head photography annually until the completion of phase 1 on June 2002. After each visit, reading centers assessed serial HVFs and fundus photographs masked to chronological sequence to identify evidence of changes. The end point committee ascertained the cause of changes based on the review of clinical history. Phase 2 of OHTS began in June 2002 and continued until March 2009. During this phase, all participants, including the phase 1 observation group, received topical IOP-lowering therapy. Phase 3 began in July 2015 and continued until June 2020, assessing the cumulative incidence of POAG over 20 years. Data analysis in this study was limited to 20 years of follow-up per participant.

OHTS Genotyping and Analysis

From the 1636 original participants in OHTS, 1077 were enrolled in an ancillary genetics study, with methodology described in Scheetz et al. Quality control and imputation steps are described in the eMethods in Supplement 1. Participants were categorized based on inferred genetic ancestry (eMethods in Supplement 1). European and African ancestry were most common in OHTS participants and retained for this analysis, after removing 2 samples with mismatched self-report ancestry and inferred ancestry (n = 1012) (Figure 1; eMethods in Supplement 1). Three participants were excluded due to a lack of CCT data (n = 1009) (Figure 1). Lastly, all baseline parameters were stratified into participant-level (n = 1009) and eye-level (n = 2018) data for analysis. Associations between PRS and baseline ocular parameters and demographic characteristics were evaluated with participant-level data. Associations between PRS and time-censored POAG onset data were evaluated with eye-level data, as 257 of 483 POAG conversions (53.2%) in OHTS occurred monocularly.

Figure 1. Participant Flowchart for Ocular Hypertension Treatment Study (OHTS) Participants.

Figure 1.

Open in a new tab

CCT indicates central corneal thickness.

Generation of PRS

We constructed a POAG PRS using summary statistics from, to our knowledge, the largest cross-ancestry POAG genome-wide association study (GWAS) meta-analysis to date after exclusion of the UK Biobank cohort; summary statistics are available online. The PRS was trained using 8 813 496 genomic variants from 449 186 cross-ancestry participants in the UK Biobank using the Lassosum penalized regression framework method with the EUR.hg19 1000 Genomes Project reference panel to inform linkage disequilibrium structure (Figure 2A and B). The definition of POAG in the UK Biobank, quality control, and score computation steps are detailed in eMethods in Supplement 1.

Figure 2. Description of the 3-Stage Process for Generation of a Polygenic Risk Score (PRS) in the Ocular Hypertension Treatment Study (OHTS) Participants.

Figure 2.

Open in a new tab

A, Derivation of genome-wide association study (GWAS) meta-analysis summary statistics, including 34 179 individuals with primary open-angle glaucoma and 349 321 control individuals from 21 studies. An example Manhattan plot is shown. The dashed line represents the threshold for genome-wide significance. B, Training of the PRS in the cross-ancestry UK Biobank population, including 8 813 496 variants in 449 186 individuals. C, Calculation of the PRS in the OHTS population; 1009 participants were retained after exclusion criteria, of whom 757 had European ancestry and 252 had African ancestry.

As PRS is a relative measure within a population, we transformed our PRS to be centered at 0 with an SD of 1 (z score) within the OHTS population, as recommended. Raw PRS from the UK Biobank trained weights was not used (eMethods and eAppendix 1 in Supplement 1). Within OHTS participants, mean unadjusted PRS scores were significantly higher for those of African ancestry compared with those of European ancestry (eAppendix 2 in Supplement 1), an often-observed phenomenon attributed to differences in the sizes of linkage disequilibrium blocks and variant allele frequencies among ancestral backgrounds. To correct for this, PRS values were transformed to z scores within each of those with inferred European ancestry (n = 757) and African ancestry (n = 252) (Figure 2C). An alternate set of results using raw PRS without transformation is provided in eAppendix 4 in Supplement 1 to demonstrate the robustness of our results.

Baseline Prediction Models

Survival regression analysis was performed to predict time to the event of interest: conversion of OHTN to POAG. Two multivariable Cox proportional hazards models with the following input eye-level baseline covariates were generated: model 1 included age, IOP, PSD, VCDR, CCT, and randomization status, and model 2 included model 1 covariates plus PRS. Randomization status was included in all models to control for association of delayed treatment with POAG onset in OHTS phase 1 participants. Lastly, to determine the predictive performance of each baseline covariate individually, simplified univariate models for each variable of interest were constructed, with adjustment for randomization status in each model.

Statistical Analysis

The mean normalized PRS for each participant of European and African ancestry was compared with a 2-sample t test, and differences by baseline risk tertile as defined by OHTS prediction model were compared with 1-way analysis of variance. Participants were then grouped into either deciles or low-risk, intermediate-risk, and high-risk tertiles based on PRS z scores. Differences in participant-level demographic characteristics, follow-up duration, and baseline ocular parameters were assessed with χ2 or 1-way analysis of variance. Differences in POAG onset were compared between PRS tertiles with K-sample log-rank tests. Stratification into each phase of OHTS and ancestry was performed. The association between PRS decile and POAG conversion was assessed with linear regression. The association between age at diagnosis and PRS decile was assessed with linear regression adjusted for ancestry and sex. Comparisons subject to multiple levels of stratification (by ancestry, tertile, or OHTS phase) accounted for multiplicity using false discovery rate correction.

For the prediction of 20-year POAG onset, data were divided into training and testing sets with 90% and 10% splits, respectively, and 10-fold cross validation was performed. Group k-fold methodology was applied so both eyes from the same patient appeared in the assigned training or testing sets. Concordance index values were computed over each cross validation fold to determine overall predictive performance, and the mean score is provided for each model. Cumulative and dynamic area under the receiver operating characteristic curve (AUC[t]) scores were computed to assess model stability, as sensitivity and specificity of survival regression models vary with time. AUC(t) was calculated in 0.5-year intervals from 2.25 years to 20 years. This period was chosen because randomly generated cross validation folds often exclude participants with earlier conversion. Mean AUC(t) values were calculated for OHTS phases 1, 2, and 3. The predictive performance of models 1 and 2 was compared using Wilcoxon signed rank test with mean cross-validation AUC(t) scores. Hazard ratios were calculated after input variables were scaled to fit normal distributions centered at 0 with an SD of 1. The hazard ratio represents a single SD increase in continuous variables and change of state in binary variables.

Cox proportional hazards models were implemented with scikit-survival version 0.19.0 and lifelines version 0.27.4 libraries using Python version 3.9.12 (Python Software Foundation). Significance was set at P < .05, and P values were 2-tailed.

Results

Of 1009 included participants, 562 (55.7%) were female, and the mean (SD) age was 55.9 (9.3) years. Of participants who met our inclusion criteria, 212 participants (287 eyes) developed POAG (eAppendix 3 in Supplement 1). A total of 655 participants (64.9%) who were initially enrolled in OHTS 1 were still enrolled at the start of OHTS phase 3 (eAppendix 3 in Supplement 1).

Mean (SD) PRS was significantly higher for POAG converters (0.24 [0.95]) than for nonconverters (−0.12 [1.00]) (difference, 0.36; 95% CI, 0.24-0.49; P < .001). When stratified by ancestry, mean (SD) PRS for 242 POAG converters of European descent (0.29 [0.95]) was significantly higher than 519 corresponding nonconverters (−0.13 [0.99]) (difference, 0.42; 95% CI, 0.27-0.57; P < .001). Among those with African ancestry, mean (SD) PRS was not significantly higher for 140 POAG converters (0.13 [0.93]) than for 108 nonconverters (−0.10 [1.04]) (difference, 0.23; 95% CI, −0.02 to 0.48; P = .11). There was no significant difference in mean (SD) PRS between the lowest (−0.08 [1.03]) and highest (0.18 [0.94]) baseline risk tertiles (difference, 0.26; 95% CI, −0.50 to 0.01; P = .10), as defined by the original OHTS prediction model. Among those with European ancestry, there was no significant difference in mean (SD) PRS between the lowest (−0.07 [0.98]) and highest (0.10 [1.01]) baseline risk tertiles (difference, −0.17; 95% CI, −0.51 to 0.17; P = .31). Among those with African ancestry, the difference in mean (SD) PRS between the lowest (−0.14 [1.26]) and highest (0.27 [0.85]) baseline risk tertiles was not statistically significant (difference, −0.41; 95% CI, −0.87 to 0.05; P = .12).

When PRS is aggregated into risk tertiles, there were no significant differences in baseline age, sex, IOP, or CCT (Table). There was a significant increase in VCD with increase in risk tertile (Table).

Table. Description of the Entire Study Population and Comparison of Baseline Characteristics in Polygenic Risk Score (PRS) Risk Tertiles.

Characteristic Total population (N = 1009) PRS risk tertile, mean (SD) P valuea
Low (n = 337) Intermediate (n = 323) High (n = 349)
Age, y 55.9 (9.3) 56.3 (9.3) 55.7 (9.5) 55.8 (9.2) .97
Sex, %
Female 55.7 56.1 56.0 55.0 .97
Male 44.3 43.9 44.0 45.0
Follow-up duration, y 15.2 (4.77) 15.8 (5.5) 15.0 (5.8) 14.8 (6.5) .12
IOP, mm Hg 24.94 (3.06) 24.72 (2.97) 24.89 (3.07) 25.20 (3.11) .19
VCDR 0.39 (0.20) 0.36 (0.20) 0.38 (0.20) 0.43 (0.20) <.001
CCT, μm 573.4 (38.6) 574.2 (39.6) 572.6 (39.9) 573.4 (36.4) .87
PSD, dB 1.91 (0.25) 1.91 (0.25) 1.91 (0.25) 1.91 (0.25) .97
Positive family history, % 43.9 46.0 42.9 42.7 .76
Open in a new tab

Abbreviations: CCT, central corneal thickness; IOP, intraocular pressure; PSD, visual field pattern standard deviation; VCDR, vertical cup-disc ratio.

a

P value represents false discovery rate–corrected output from 1-way analysis of variance for continuous variables or χ2 test for categorical variables between PRS risk tertiles.

There was significantly increased risk of POAG onset in the highest tertile of PRS compared with those in the lowest tertile by the end of OHTS phase 1 (3.01-fold and 5.93% absolute increase; P < .001), phase 2 (2.30-fold and 9.96% absolute increase; P < .001), and phase 3 (2.00-fold and 9.93% absolute increase; P < .001) (Figure 3A). When stratified by ancestry, there was a significant difference in POAG onset between the lowest and highest PRS tertiles among those with European ancestry by the end of OHTS phase 1 (4.37-fold and 6.50% absolute increase; P < .001), phase 2 (2.37-fold and 8.79% absolute increase; P < .001), and phase 3 (2.00-fold and 8.03% absolute increase; P < .001) (Figure 3B). Among those with African ancestry, there was a significant difference between the highest and lowest PRS tertiles by the end of OHTS phase 2 (2.14-fold and 13.29% absolute increase; P = .003) and phase 3 (2.04-fold or 16.64% absolute increase; P = .002) but not by the end of phase 1 (3.98% absolute increase; P = .18) (Figure 3C).

Figure 3. Cumulative Primary Open-Angle Glaucoma (POAG) Onset Curves Adjusted for Censoring in the Ocular Hypertension Treatment Study (OHTS) Participants.

Figure 3.

Open in a new tab

PRS indicates polygenic risk score (PRS). Shading indicates 95% CIs.

Each decile of higher PRS was associated with 2.46% (95% CI, 2.10-2.81) greater POAG conversion risk per eye, with linear regression showing an increase from 14.85% (95% CI, 11.73-17.98) conversion over 20 years in lowest decile to 36.99% (95% CI, 33.86-40.11) conversion at highest decile (eAppendix 3 in Supplement 1). When adjusted for censoring, each decile in PRS was associated with a 1.36% (95% CI, 1.08-1.64) greater POAG conversion risk, with linear regression showing an increase from 9.52% (95% CI, 7.09-11.95) 20-year conversion in the lowest decile to 21.81% (95% CI, 19.37-24.25) in the highest decile (eAppendix 3 in Supplement 1).

In multivariable linear regression with age at POAG diagnosis against PRS decile, no significant linear association was initially found (years younger age at diagnosis per higher decile: β = 0.34; 95% CI, –0.01 to 0.70; P = .06) (eAppendix 3 in Supplement 1). When stratified by delay in OHTN treatment, those with delayed treatment had a significant linear association between age at diagnosis and PRS decile (years younger age at diagnosis per higher decile: β = 0.52; 95% CI, 0.01-1.03; P = .047). Those who received early treatment had no significant association (years younger age at diagnosis per higher decile: β = 0.17; 95% CI, –0.33 to 0.67; P = .49) (eAppendix 3 Supplement 1).

The AUC(t) for multivariable Cox proportional hazards prediction models is shown in Figure 4. Average performance over each phase of OHTS and overall C index are shown in eAppendix 3 in Supplement 1. The PRS model improved mean AUC(t) in OHTS phases 1, 2, and 3 as well as overall the C index over the OHTS model (eAppendix 3 in Supplement 1). Among participants with European ancestry, the C index for model 2 was 0.78, and the C index for model 1 was 0.76 (P < .001). Among participants with African ancestry, the C index was 0.70 for model 1 and 0.71 for model 2 (P = .21).

Figure 4. Time-Dependent Area Under the Receiver Operating Characteristic Curves (AUC[t]) for Baseline Prediction Models.

Figure 4.

Open in a new tab

AUC(t) curves for baseline prediction models, calculated in 0.5-year intervals. OHTS indicates Ocular Hypertension Treatment Study; PRS, polygenic risk score.

aModel 1 covariates included age, intraocular pressure, visual field pattern standard deviation, vertical cup-disc ratio, central corneal thickness, and randomization status.

bModel 2 covariates include model 1 covariates and PRS.

A 1-SD higher PRS conferred a hazard ratio of 1.25 (95% CI, 1.15-1.37) for conversion to POAG in model 2. Simplified models to assess predictive performance of each baseline covariate demonstrated that VCDR was the best predictor overall based on C index (eAppendix 3 in Supplement 1).

Discussion

We show that a high POAG PRS, based on, to our knowledge, the largest GWAS meta-analysis to date, was associated with increased risk of POAG conversion among patients with OHTN. This association was independent of most phenotypic baseline risk factors, genetically determined ancestry, and phase of OHTS follow-up period. We also found that early IOP-lowering treatment may delay clinically detectable disease onset in those with high cumulative genetic risk. Lastly, the inclusion of a PRS improved multivariable risk stratification models.

PRS was a strong predictor of POAG risk in our study population with OHTN, with absolute risk increasing from 9.52% (95% CI, 7.09-11.95) to 21.81% (95% CI, 19.37-24.25) from the lowest PRS decile to the highest and significantly increased 20-year POAG incidence in participants of European and African descent. When PRS was grouped into risk tertiles, the incidence of POAG tripled in those with early conversion (OHTS phase 1) and doubled in those with late conversion (phase 3). These increases in incidence are comparable with the 12% absolute increase previously attributed to TMCO1 locus risk alleles in OHTS.

When stratified by ancestry, this association held for those with European ancestry over all phases of OHTS and those with African ancestry over OHTS phases 2 and 3. Comparison of the lowest and highest PRS risk tertiles in those with African ancestry lacked precision in OHTS phase 1 due to low POAG incidence. The decreased risk attributable to PRS tertile in later phases of OHTS was most likely due to increasing contributions of age or cumulative environmental burden. While the increase in absolute risk appeared lower than the 6-fold to 45-fold increases seen in other studies, it is important to note that our study includes only patients with OHTN with healthy optic nerves at baseline but increased risk of conversion compared with those with normal IOP. Furthermore, low-risk patients may have undiscovered protective variants not captured by our PRS.

Analysis of collinearity between PRS and other baseline covariates only showed a significant association with mean (SD) VCDR, with a modest increase from 0.36 (0.20) to 0.43 (0.20) from the lowest to highest PRS tertile (Table). This suggests that genetic variants influencing VCDR may also play a role in the conversion of OHTN to POAG. While prior work has shown certain risk variants to be associated with IOP, we did not find a statistically significant difference in IOP between the lowest to highest PRS tertiles (Table). It is possible that IOP association effects are limited by selection bias for high IOP. A clinically important finding is the loss of significance of the association between PRS and age at diagnosis for participants receiving early treatment. This suggests that early treatment may mitigate high genetic burden for POAG. Based on the regression slope, early treatment may delay clinically detectable disease onset by up to 5.2 years when comparing the lowest PRS decile with the highest. Other studies have similarly shown associations between POAG PRS and age at diagnosis.

Importantly, Cox proportional hazards prediction models demonstrated improvement in the prediction of POAG onset with the addition of PRS. A 1-SD higher PRS increased the 20-year risk for POAG by 25%. This is consistent with our findings of relative independence of PRS from other baseline covariates. PRS added a relatively constant improvement in prediction over time, with AUC(t) improvements in all phases of OHTS, most likely because it measures immutable background risk. The absolute increase in performance of prediction models (C index of 0.75 to 0.77) may be smaller than expected. This may be due to selection bias in this population, as OHTN alone confers increased risk of POAG, and the aggregate phenotype represented by VCD, CCT, and PSD may share associations with PRS not represented by univariate collinearity analysis alone. As PRS presents a constant level of lifetime risk, its predictive power is most likely higher at earlier clinical stages and younger age prior to development of phenotypic markers. Furthermore, the absence of newer phenotypic baseline information, such as structural optical coherence tomography, may be limiting potential improvement in predictive models as early signs of glaucomatous change may have been present in patients defined as healthy at randomization. Future studies incorporating modern imaging modalities should consider replication of our findings with PRS in newer risk stratification models.

Simplified predictive models showing prediction from each baseline covariate alone demonstrated that VCDR was the strongest predictor of POAG conversion over the entire tested follow-up period. PRS was a stable predictor, with a small decrease in AUC(t) in later phases of OHTS. CCT, baseline IOP, and PSD were most predictive of early POAG onset and showed decreased performance over time. Age was a stable predictor that gained relative importance over time, becoming the second most important predictor during phase 3. This is consistent with our findings showing decreased risk stratification by PRS tertile during later phases of OHTS.

A remaining challenge in the incorporation of PRS in clinical care is susceptibility to bias from inadequate representation of individual patients in study populations. We found differences in the OHTS PRS distribution compared with the study population in the UK Biobank (eAppendix 1 in Supplement 1). Training of the PRS in the UK Biobank population may also limit applicability to OHTS participants, as the training population mostly includes those with European ancestry and may be exposed to different environmental factors. Several of our findings lost significance in the subset of participants with African ancestry, including the addition of PRS to the OHTS prediction model. While these comparisons may be underpowered due to low sample size, the PRS may also have limitations. Most participants in prior POAG GWAS are of European descent, and thus prior GWAS may be underpowered for identification of risk alleles associated with other ancestries. While moderate to high correlation in risk loci effect sizes has been shown between those with European, African, and Asian ancestries, new ancestry-specific variants may be discovered with GWAS targeting populations for ancestry-specific PRS development. Moreover, PRS may only represent a portion of genetically determined risk for disease, as it only includes single-nucleotide variant–based heritability with a frequency of greater than 1% in the study population. As published POAG PRS improve, a patient’s risk based on prior genotyping can be recalculated, allowing for ease of future clinical use.

Strengths and Limitations

The strengths of our study include those of the prospective OHTS dataset; 20 years of data from close follow-up in a diverse sample, with validation of POAG onset by reading centers and an end point committee. Therefore, our results are less likely to be biased by delay in onset and formal diagnosis in asymptomatic stages. In addition, PRS were generated from a very large sample with, to our knowledge, the greatest number of alleles identified to date.

Limitations include potential misclassification of POAG in the UK Biobank based on International Classification of Diseases code or self-report. This is most likely to result in suboptimal training of the PRS and understate how well PRS predicts POAG. Furthermore, the relatively small size of the OHTS cohort, low proportion of participants with confirmed POAG, and loss to regular follow-up after 15 years likely limits the statistical power of our stratified comparisons and the predictive performance of our model. Furthermore, while our cross-validation methodology improves internal validity, validation of a PRS from the OHTS population in external study populations would improve generalizability of our results.

Conclusions

In conclusion, high polygenic risk for POAG was associated with increased risk for and earlier development of POAG in patients with OHTN, independent of most ocular phenotypic risk factors, patient ancestry, and duration of follow-up up to 20 years.

Supplement 1.

eMethods.

eAppendix 1. Comparison of Ocular Hypertension Treatment Study (OHTS) and UK Biobank (UKBB) Polygenic Risk Score (PRS)

eAppendix 2. Polygenic Risk Score (PRS) Distributions (Raw and After Transformation to z Scores)

eAppendix 3. Proportion Developing POAG by Decile and Associations With Age at Diagnosis

eAppendix 4. Results With Raw Polygenic Risk Score (PRS) (No Transformation Within Inferred Ancestries)

eReferences.

jamaophthalmol-e240151-s001.pdf (1.4MB, pdf)
Supplement 2.

Data Sharing Statement

jamaophthalmol-e240151-s002.pdf (16.2KB, pdf)

References

  • 1.Tham YC, Li X, Wong TY, Quigley HA, Aung T, Cheng CY. Global prevalence of glaucoma and projections of glaucoma burden through 2040: a systematic review and meta-analysis. Ophthalmology. 2014;121(11):2081-2090. doi: 10.1016/j.ophtha.2014.05.013 [DOI] [PubMed] [Google Scholar]
  • 2.Stein JD, Khawaja AP, Weizer JS. Glaucoma in adults—screening, diagnosis, and management: a review. JAMA. 2021;325(2):164-174. doi: 10.1001/jama.2020.21899 [DOI] [PubMed] [Google Scholar]
  • 3.Leske MC, Heijl A, Hussein M, Bengtsson B, Hyman L, Komaroff E; Early Manifest Glaucoma Trial Group . Factors for glaucoma progression and the effect of treatment: the Early Manifest Glaucoma trial. Arch Ophthalmol. 2003;121(1):48-56. doi: 10.1001/archopht.121.1.48 [DOI] [PubMed] [Google Scholar]
  • 4.Kass MA, Heuer DK, Higginbotham EJ, et al. The Ocular Hypertension Treatment Study: a randomized trial determines that topical ocular hypotensive medication delays or prevents the onset of primary open-angle glaucoma. Arch Ophthalmol. 2002;120(6):701-713. doi: 10.1001/archopht.120.6.701 [DOI] [PubMed] [Google Scholar]
  • 5.Lichter PR, Musch DC, Gillespie BW, et al. ; CIGTS Study Group . Interim clinical outcomes in the Collaborative Initial Glaucoma Treatment Study comparing initial treatment randomized to medications or surgery. Ophthalmology. 2001;108(11):1943-1953. doi: 10.1016/S0161-6420(01)00873-9 [DOI] [PubMed] [Google Scholar]
  • 6.The AGIS Investigators . The Advanced Glaucoma Intervention Study (AGIS): 7. the relationship between control of intraocular pressure and visual field deterioration. Am J Ophthalmol. 2000;130(4):429-440. doi: 10.1016/S0002-9394(00)00538-9 [DOI] [PubMed] [Google Scholar]
  • 7.Collaborative Normal-Tension Glaucoma Study Group . The effectiveness of intraocular pressure reduction in the treatment of normal-tension glaucoma. Am J Ophthalmol. 1998;126(4):498-505. doi: 10.1016/S0002-9394(98)00272-4 [DOI] [PubMed] [Google Scholar]
  • 8.Gordon MO, Torri V, Miglior S, et al. ; Ocular Hypertension Treatment Study Group; European Glaucoma Prevention Study Group . Validated prediction model for the development of primary open-angle glaucoma in individuals with ocular hypertension. Ophthalmology. 2007;114(1):10-19. doi: 10.1016/j.ophtha.2006.08.031 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Wiggs JL, Pasquale LR. Genetics of glaucoma. Hum Mol Genet. 2017;26(R1):R21-R27. doi: 10.1093/hmg/ddx184 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Gharahkhani P, Jorgenson E, Hysi P, et al. ; NEIGHBORHOOD Consortium; ANZRAG Consortium; Biobank Japan Project; FinnGen Study; UK Biobank Eye and Vision Consortium; GIGA Study Group; 23 and Me Research Team . Genome-wide meta-analysis identifies 127 open-angle glaucoma loci with consistent effect across ancestries. Nat Commun. 2021;12(1):1258. doi: 10.1038/s41467-020-20851-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Wray NR, Lin T, Austin J, et al. From basic science to clinical application of polygenic risk scores: a primer. JAMA Psychiatry. 2021;78(1):101-109. doi: 10.1001/jamapsychiatry.2020.3049 [DOI] [PubMed] [Google Scholar]
  • 12.Adeyemo A, Balaconis MK, Darnes DR, et al. ; Polygenic Risk Score Task Force of the International Common Disease Alliance . Responsible use of polygenic risk scores in the clinic: potential benefits, risks and gaps. Nat Med. 2021;27(11):1876-1884. doi: 10.1038/s41591-021-01549-6 [DOI] [PubMed] [Google Scholar]
  • 13.Collister JA, Liu X, Clifton L. Calculating polygenic risk scores (PRS) in UK Biobank: a practical guide for epidemiologists. Front Genet. 2022;13:818574. doi: 10.3389/fgene.2022.818574 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Osterman MD, Kinzy TG, Bailey JNC. Polygenic risk scores. Curr Protoc. 2021;1(5):e126. doi: 10.1002/cpz1.126 [DOI] [PubMed] [Google Scholar]
  • 15.Dudbridge F. Power and predictive accuracy of polygenic risk scores. PLoS Genet. 2013;9(3):e1003348. doi: 10.1371/journal.pgen.1003348 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Torkamani A, Wineinger NE, Topol EJ. The personal and clinical utility of polygenic risk scores. Nat Rev Genet. 2018;19(9):581-590. doi: 10.1038/s41576-018-0018-x [DOI] [PubMed] [Google Scholar]
  • 17.Khera AV, Chaffin M, Aragam KG, et al. Genome-wide polygenic scores for common diseases identify individuals with risk equivalent to monogenic mutations. Nat Genet. 2018;50(9):1219-1224. doi: 10.1038/s41588-018-0183-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Khera AV, Chaffin M, Wade KH, et al. Polygenic prediction of weight and obesity trajectories from birth to adulthood. Cell. 2019;177(3):587-596.e9. doi: 10.1016/j.cell.2019.03.028 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Fan BJ, Bailey JC, Igo RP Jr, et al. Association of a primary open-angle glaucoma genetic risk score with earlier age at diagnosis. JAMA Ophthalmol. 2019;137(10):1190-1194. doi: 10.1001/jamaophthalmol.2019.3109 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Craig JE, Han X, Qassim A, et al. ; NEIGHBORHOOD Consortium; UK Biobank Eye and Vision Consortium . Multitrait analysis of glaucoma identifies new risk loci and enables polygenic prediction of disease susceptibility and progression. Nat Genet. 2020;52(2):160-166. doi: 10.1038/s41588-019-0556-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Li Y, Mitchell W, Elze T, Zebardast N. Association between diabetes, diabetic retinopathy, and glaucoma. Curr Diab Rep. 2021;21(10):38. doi: 10.1007/s11892-021-01404-5 [DOI] [PubMed] [Google Scholar]
  • 22.Ding Y, Liu Y, Yan Q, et al. ; AREDS2 Research Group . Bivariate analysis of age-related macular degeneration progression using genetic risk scores. Genetics. 2017;206(1):119-133. doi: 10.1534/genetics.116.196998 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Forrest IS, Chaudhary K, Paranjpe I, et al. Genome-wide polygenic risk score for retinopathy of type 2 diabetes. Hum Mol Genet. 2021;30(10):952-960. doi: 10.1093/hmg/ddab067 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Ajana S, Cougnard-Grégoire A, Colijn JM, et al. ; EYE-RISK Consortium . Predicting progression to advanced age-related macular degeneration from clinical, genetic, and lifestyle factors using machine learning. Ophthalmology. 2021;128(4):587-597. doi: 10.1016/j.ophtha.2020.08.031 [DOI] [PubMed] [Google Scholar]
  • 25.Qassim A, Souzeau E, Siggs OM, et al. An intraocular pressure polygenic risk score stratifies multiple primary open-angle glaucoma parameters including treatment intensity. Ophthalmology. 2020;127(7):901-907. doi: 10.1016/j.ophtha.2019.12.025 [DOI] [PubMed] [Google Scholar]
  • 26.Siggs OM, Qassim A, Han X, et al. Association of high polygenic risk with visual field worsening despite treatment in early primary open-angle glaucoma. JAMA Ophthalmol. 2022;141(1):73-77. doi: 10.1001/jamaophthalmol.2022.4688 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Scheetz TE, Faga B, Ortega L, et al. Glaucoma risk alleles in the Ocular Hypertension Treatment Study. Ophthalmology. 2016;123(12):2527-2536. doi: 10.1016/j.ophtha.2016.08.036 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Kass MA, Heuer DK, Higginbotham EJ, et al. ; Ocular Hypertension Study Group . Assessment of cumulative incidence and severity of primary open-angle glaucoma among participants in the Ocular Hypertension Treatment Study after 20 years of follow-up. JAMA Ophthalmol. 2021;139(5):1-9. doi: 10.1001/jamaophthalmol.2021.0341 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Kass MA, Gordon MO, Gao F, et al. ; Ocular Hypertension Treatment Study Group . Delaying treatment of ocular hypertension: the Ocular Hypertension Treatment Study. Arch Ophthalmol. 2010;128(3):276-287. doi: 10.1001/archophthalmol.2010.20 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Mak TSH, Porsch RM, Choi SW, Zhou X, Sham PC. Polygenic scores via penalized regression on summary statistics. Genet Epidemiol. 2017;41(6):469-480. doi: 10.1002/gepi.22050 [DOI] [PubMed] [Google Scholar]
  • 31.Ayellet Segrè Lab . Data. Accessed December 21, 2022. https://www.asegrelab.org/data
  • 32.Sekimitsu S, Xiang D, Smith SL, et al. ; UK Biobank Eye and Vision Consortium . Deep ocular phenotyping across primary open-angle glaucoma genetic burden. JAMA Ophthalmol. 2023;141(9):891-899. doi: 10.1001/jamaophthalmol.2023.3645 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Patel AP, Wang M, Ruan Y, et al. ; Genes & Health Research Team; the Million Veteran Program . A multi-ancestry polygenic risk score improves risk prediction for coronary artery disease. Nat Med. 2023;29(7):1793-1803. doi: 10.1038/s41591-023-02429-x [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Bycroft C, Freeman C, Petkova D, et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203-209. doi: 10.1038/s41586-018-0579-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Lewis CM, Vassos E. Prospects for using risk scores in polygenic medicine. Genome Med. 2017;9(1):96. doi: 10.1186/s13073-017-0489-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Khera AV, Chaffin M, Zekavat SM, et al. Whole-genome sequencing to characterize monogenic and polygenic contributions in patients hospitalized with early-onset myocardial infarction. Circulation. 2019;139(13):1593-1602. doi: 10.1161/CIRCULATIONAHA.118.035658 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Duncan L, Shen H, Gelaye B, et al. Analysis of polygenic risk score usage and performance in diverse human populations. Nat Commun. 2019;10(1):3328. doi: 10.1038/s41467-019-11112-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Kamarudin AN, Cox T, Kolamunnage-Dona R. Time-dependent ROC curve analysis in medical research: current methods and applications. BMC Med Res Methodol. 2017;17(1):53. doi: 10.1186/s12874-017-0332-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Heagerty PJ, Lumley T, Pepe MS. Time-dependent ROC curves for censored survival data and a diagnostic marker. Biometrics. 2000;56(2):337-344. doi: 10.1111/j.0006-341X.2000.00337.x [DOI] [PubMed] [Google Scholar]
  • 40.Lambert J, Chevret S. Summary measure of discrimination in survival models based on cumulative/dynamic time-dependent ROC curves. Stat Methods Med Res. 2016;25(5):2088-2102. doi: 10.1177/0962280213515571 [DOI] [PubMed] [Google Scholar]
  • 41.Pölsterl S. scikit-survival: A library for time-to-event analysis built on top of scikit-learn. J Mach Learn Res. 2020;21(212):1-6.34305477 [Google Scholar]
  • 42.Davidson-Pilon C. lifelines: Survival analysis in Python. J Open Source Softw. 2019;4(40):1317. doi: 10.21105/joss.01317 [DOI] [Google Scholar]
  • 43.Waksmunski AR, Kinzy TG, Cruz LA, et al. ; VA Million Veteran Program . Glaucoma genetic risk scores in the Million Veteran Program. Ophthalmology. 2022;129(11):1263-1274. doi: 10.1016/j.ophtha.2022.06.012 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Hsiao YJ, Chuang HK, Chi SC, et al. Genome-wide polygenic risk score for predicting high risk glaucoma individuals of Han Chinese ancestry. J Pers Med. 2021;11(11):1169. doi: 10.3390/jpm11111169 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Qassim A, Mullany S, Awadalla MS, et al. A polygenic risk score predicts intraocular pressure readings outside office hours and early morning spikes as measured by home tonometry. Ophthalmol Glaucoma. 2021;4(4):411-420. doi: 10.1016/j.ogla.2020.12.002 [DOI] [PubMed] [Google Scholar]
  • 46.Manolio TA, Collins FS, Cox NJ, et al. Finding the missing heritability of complex diseases. Nature. 2009;461(7265):747-753. doi: 10.1038/nature08494 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Karlsson IK, Escott-Price V, Gatz M, et al. Measuring heritable contributions to Alzheimer’s disease: polygenic risk score analysis with twins. Brain Commun. 2022;4(1):fcab308. doi: 10.1093/braincomms/fcab308 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplement 1.

eMethods.

eAppendix 1. Comparison of Ocular Hypertension Treatment Study (OHTS) and UK Biobank (UKBB) Polygenic Risk Score (PRS)

eAppendix 2. Polygenic Risk Score (PRS) Distributions (Raw and After Transformation to z Scores)

eAppendix 3. Proportion Developing POAG by Decile and Associations With Age at Diagnosis

eAppendix 4. Results With Raw Polygenic Risk Score (PRS) (No Transformation Within Inferred Ancestries)

eReferences.

jamaophthalmol-e240151-s001.pdf (1.4MB, pdf)
Supplement 2.

Data Sharing Statement

jamaophthalmol-e240151-s002.pdf (16.2KB, pdf)

Articles from JAMA Ophthalmology are provided here courtesy of American Medical Association

RESOURCES