ÌÇÐÄvlog

Key Points

QuestionÌý Can polygenic risk scores be developed that assess an individual’s risk of either elevated intraocular pressure (IOP) or increased vertical cup-disc ratio (VCDR)?

FindingsÌý In this genetic association study including data on 18 071 individuals, PRSs were found to predict VCDR with and IOP with validity confirmed through independent cohorts.

MeaningÌý These findings suggest the potential clinical utility of these PRSs in glaucoma risk assessment.

ImportanceÌý Early detection of glaucoma is essential to timely monitoring and treatment, and primary open-angle glaucoma risk can be assessed by measuring intraocular pressure (IOP) or optic nerve head vertical cup-disc ratio (VCDR). Polygenic risk scores (PRSs) could provide a link between genetic effects estimated from genome-wide association studies (GWASs) and clinical applications to provide estimates of an individual’s genetic risk by combining many identified variants into a score.

ObjectiveÌý To construct IOP and VCDR PRSs with clinically relevant predictive power.

Design, Setting, and ParticipantsÌý This genetic association study evaluated the PRSs for 6959 of 51 338 individuals in the Canadian Longitudinal Study on Aging (CLSA; 2010 to 2015 with data from 11 centers in Canada) and 4960 of 5107 individuals the community-based Busselton Healthy Aging Study (BHAS; 2010 to 2015 in Busselton, Western Australia) with an artificial intelligence grading approach used to obtain precise VCDR estimates for the CLSA dataset. Data for approximately 500 000 individuals in UK Biobank from 2006 to 2010 were used to validate the power of the PRS. Data were analyzed from June to November 2023.

Main Outcomes and MeasuresÌý IOP and VCDR PRSs and phenotypic variance (R²) explained by each PRS.

ResultsÌý Participants in CLSA were aged 45 to 85 years; those in BHAS, 46 to 64 years; and those in UK Biobank, 40 to 69 years. The VCDR PRS explained 22.0% (95% CI, 20.1-23.9) and 19.7% (95% CI, 16.3-23.3) of the phenotypic variance in VCDR in CLSA and BHAS, respectively, while the IOP PRS explained 12.9% (95% CI, 11.3-14.6) and 9.6% (95% CI, 8.1-11.2) of phenotypic variance in CLSA and BHAS IOP measurements. The VCDR PRS variance explained 5.2% (95% CI, 3.6-7.1), 12.1% (95% CI, 7.5-17.5), and 14.3% (95% CI, 9.3-19.9), and the IOP PRS variance explained 2.3% (95% CI, 1.5-3.3), 3.2% (95% CI, 1.3-5.8), and 7.5% (95% CI, 6.2-8.9) (P < .001) across African, East Asian, and South Asian populations, respectively.

Conclusions and RelevanceÌý VCDR and IOP PRSs derived using a large recently published multitrait GWAS exhibited validity across independent cohorts. The findings suggest that an IOP PRS has the potential to identify individuals who may benefit from more intensive IOP-lowering treatments, which could be crucial in managing glaucoma risk more effectively. Individuals with a high VCDR PRS may be at risk of developing glaucoma even if their IOP measures fall within the normal range, suggesting that these PRSs could help in early detection and intervention, particularly among those who might otherwise be considered at low risk based on IOP alone.

Glaucoma leads to optic nerve damage, resulting in irreversible vision loss and potential blindness.¹ The estimated global prevalence of glaucoma is 3.5% (95% CI, 2.1-5.8).² Primary open-angle glaucoma (POAG) is the most common subtype in African individuals and European individuals with a global prevalence of 2.4% (95% CI, 2.0-2.8).³

Increased vertical cup-disc ratio (VCDR) and intraocular pressure (IOP) are 2 key risk factors of POAG.⁴ IOP refers to the fluid pressure of the eye.⁵ IOP is determined by the balance between the production of aqueous humor by the ciliary body and its drainage via the trabecular meshwork and uveoscleral outflow. Increased IOP can trigger optic nerve damage through mechanical compression and optic nerve ischemia, resulting in glaucoma. VCDR describes the ratio of the vertical diameter of the optic cup to the optic disc (optic nerve head).^6,7 The physiologically most relevant area on the optic disc, which is the area of the optic disc minus the cup, is called the neuroretinal rim area. As the rim of the disc contains nerve fiber, the pathological enlargement of the cup triggers the damage of optic nerves.^8,9

As POAG is one of the leading causes of irreversible blindness worldwide,^1,10 early and accurate diagnosis is crucial for effective management. Standard treatment is lowering IOP through the use of eye drops, laser therapy, or surgery, slowing glaucomatous optic neuropathy progression. However, due to the lack of noticeable symptoms during the early stages of POAG, patients frequently receive a diagnosis at advanced stages, where the vision loss incurred cannot be fixed with existing treatments. Research suggests that many individuals with POAG remain undiagnosed, emphasizing the stealthy progression of this condition and the challenges associated with early detection.^11,12 Previous studies have suggested polygenic risk scores (PRSs) have the potential to make early screening of POAG more feasible.¹³ While the POAG-specific PRS was shown to be useful in a range of contexts,¹³ here we sought to derive PRSs that will provide predictions for IOP and VCDR, 2 of the most important risk factors for glaucoma risk.

Family studies have revealed that IOP and VCDR have heritability estimates of 40% to 70% and 48% to 57%, respectively.^14,15 Given the high genetic correlation between IOP and VCDR and glaucoma, studies have shown that PRSs developed based on IOP and VCDR loci are effective for POAG risk stratification.^13,16,17 A recent genome-wide association study (GWAS) incorporating POAG, IOP, and VCDR data reported more than 300 POAG loci.¹⁸ Here we derived new PRSs for IOP or VCDR and validated them in 2 independent datasets: the Canadian Longitudinal Study on Aging (CLSA) and the Busselton Healthy Aging Study (BHAS). The application of a PRS for IOP could be instrumental in identifying individuals who might benefit from more aggressive IOP-lowering treatments, while individuals with a high PRS for VCDR might be at risk of developing glaucoma even if their IOP remains within the normal range.

All cohorts used in this study were approved by local research ethics boards. Written informed consent was obtained from all individual participants included in the original study. No compensation was provided to participants. No additional approval or consent were required for this analysis owing to the deidentified nature of the data. The study methods followed the World Medical Association Declaration of Helsinki ethical standards for medical research¹⁹ and the Strengthening the Reporting of Genetic Association Studies () reporting guideline.

In the study by Han et al,¹⁸ a multitrait analysis of a GWAS (MTAG) was applied to jointly analyze POAG, VCDR, and IOP. One attractive property of the MTAG model used to combine POAG, IOP, and VCDR is that the software can produce effect size estimates that are specific to each of the input traits, with the precision of the estimates optimized because of the correlation between POAG, VCDR, and IOP (pairwise correlations: POAG-IOP, 0.71; POAG-VCDR, 0.50; and IOP-VCDR, 0.22).¹³ Here, we used the single-nucleotide variant (SNV) weights for VCDR and IOP generated by the MTAG model in Han et al¹⁸ to derive a PRS. Additional details are provided in the eAppendix in Supplement 1. The overall design of this study is outlined in Figure 1.

The CLSA is a longitudinal study (2010 to 2015 with data from 11 centers in Canada) of 51 338 Canadians aged 45 to 85 years at enrollment. The analysis by Han et al¹⁸ was based on CLSA release version 2.0 data (n = 19 669); we validated the PRS in 6959 samples only presented in CLSA release version 3.0 data (n = 26 622) to avoid sample overlap. The VCDR and vertical disc diameter (VDD) were assessed using retinal fundus photography and the artificial intelligence (AI) models trained in UK Biobank.¹⁷ To account for the confounding effect of optic disc size, VCDR gradings were adjusted for VDD.¹⁷ The AI-graded VCDR and VDD values were averaged between the left and right eyes. A total of 5890 participants of European ancestry with AI-labeled VCDR and genetic data were included.

Corneal-compensated IOP measurements in CLSA were available for both baseline and follow-up data on both eyes. In keeping with previous work, IOP values were averaged across multiple measurements.²⁰ For patients receiving topical glaucoma medication in the CLSA baseline medication record, we estimated pretreatment IOP by dividing by 0.7, as in previous studies.^20-22 We then removed IOP outliers of less than 5 mm Hg or greater than 60 mm Hg.¹⁶ People with missing IOP measurements were excluded.

BHAS is a community-based cohort study based in the City of Busselton, Western Australia (2010 to 2015).²³ A total of 4960 individuals aged 46 to 64 years in European ancestry groups were selected for PRS validation. When matching genotype and phenotype data, there were 1588 individuals used for VCDR PRS validation and 4839 individuals for IOP PRS validation. A high-resolution macula-centered fundus photograph was taken using a digital retinal camera (Canon CR-1). The VCDR and VDD of participants were then manually graded by an experienced observer for a subset of these participants. IOP was measured using a rebound tonometer (ICare Finland Oy). The IOP, VCDR and VDD values were averaged between left and right eyes.

UK Biobank is a large-scale United Kingdom database containing genetic and phenotypic data from approximately 500 000 participants between the ages of 40 to 69 years at recruitment (2006 to 2010). The genotype data underwent quality control and imputation procedures as previously described.²⁴ For phenotypic analysis, an AI model, previously detailed by Han et al,¹⁷ graded the VCDR and VDD using images of both eyes from 2 visits. We included 2285 African individuals, 573 East Asian individuals, and 2524 South Asian individuals with VCDR estimations for assessing the VCDR PRS. For IOP analysis, individuals with extreme (below 5 mm Hg or above 60 mm Hg) corneal-compensated IOP measurements (Data-Field 5254 and 5262) were excluded. We averaged the VCDR and IOP measurements between the left and right eyes. The remaining IOP measurements were averaged between the left and right eyes for 4304 African individuals, 916 East Asian individuals, and 5087 South Asian individuals, facilitating the validation of the IOP PRS.

We derived the PRSs using popular PRS approaches, including PLINK version 1.90b6.8,²⁵ megaPRS version 5.2,²⁶ PRS–continuous shrinkage (CS) November 3, 2022, release,²⁷ and SBayesRC version 0.1.3²⁸ and compared phenotypic variance (R²) explained by each PRS. Both VCDR and IOP results suggested the PRS calculated by SBayesRC weights had the highest R² (Table 1). Thus our main analyses were based on the PRS generated through SBayesRC.²⁸ We used a linkage disequilibrium reference calculated from the European UK Biobank ancestry and functional annotation information for approximately 7 million SNVs from baseline model 2.2.²⁹ We computed PRSs using SBayesRC weights in the CLSA, BHAS, and UK Biobank cohorts using PLINK version 1.96b. Each PRS was normalized to mean 0, variance 1. As a comparison, we calculated the PRS for individuals in CLSA and BHAS using published lead VCDR and IOP SNVs.^16,17 These PRSs were calculated by PLINK version 1.96b.

For VCDR and VDD, we used the rank-based inverse normal transformation (Rankit) method to normalize the data. We also tested the VCDR adjusted by VDD to account for the potential influence of disc size. The residuals from linear regression (the differences between the observed VCDR values and the predicted VCDR by the linear model based on VDD) represent the VDD-adjusted VCDR. We applied linear regression to compute the phenotypic variance explained (R²) by PRS. The baseline model includes sex, age, and top 10 principal components. The variance explained by the PRS (incremental R²) was computed through R²_{±Ê¸é³§â€‰+ b²¹²õ±ð±ô¾±²Ô±ð} − R²_baseline. Data were analyzed from June to November 2023.

Figure 2 and eTable 1 in Supplement 1 outline the characteristics of the study sample, encompassing VCDR and IOP measurements. The VCDR PRS, generated by SBayesRC, was validated in 5890 samples from the CLSA dataset. Using linear regression, we examined the association between PRS and VCDR and IOP. The baseline model, which included age, sex, and 10 principal components, accounted for 2.5% (95% CI, 1.8-3.4) of the variance in VDD-adjusted VCDR and 3.8% (95% CI, 2.9-4.8) of the variance in unadjusted VCDR. To evaluate the improved VCDR variance explained by the new PRS, we compared the new PRS against a PRS constructed using recently published 231 lead VCDR SNVs.¹⁷ The PRS explained 22.0% (95% CI, 20.1-23.9) of the phenotypic variance of VDD-adjusted VCDR in CLSA and 21.9% (95% CI, 20.1-23.8) of unadjusted VCDR within the CLSA dataset (Table 2). In comparison, the VCDR PRS based on the published lead VCDR SNVs¹⁷ explained 11.6% (95% CI, 10.1-13.2) of VCDR variance and 11.1% variance of VDD-adjusted VCDR. Hence the new VCDR PRS nearly doubled the phenotypic variance explained. A significant association between all VCDR PRS and VCDR or VDD-adjusted VCDR measurements was observed (Table 2).

Similarly, the IOP PRS explained 12.9% (95% CI, 11.3-14.6) of the phenotypic variance in 5754 samples within the CLSA dataset (Table 2), which showed improvement in predictive capability compared with a PRS generated using published 101 lead IOP SNVs (phenotypic variance explained, 4.5%; 95% CI, 3.5-5.6).¹⁶ The baseline model accounted for 3.1% (95% CI, 2.3-4.0) of the phenotypic variance of IOP measurements. Analyzing decile groups of VCDR and IOP PRSs revealed a consistent upward trend, indicating that higher PRS values correlated with elevated mean VCDR or IOP (Figure 3).

Furthermore, we conducted PRS validation in 1588 samples within the BHAS dataset. Characteristics of the individuals in BHAS are listed in eTable 2 in Supplement 1. Our baseline model, adjusting for age, sex, and 10 principal components, explained 0.98% (95% CI, 0.3-2.2) of the variance in VDD-adjusted VCDR and 0.79% (95% CI, 0.2-1.9) of unadjusted VCDR. The VCDR PRS explained 19.7% (95% CI, 16.3-23.3) of the phenotypic variance in VDD-adjusted VCDR and 18.3% (95% CI, 15.0-21.8) in unadjusted VCDR (Table 2). The PRS constructed using 2021 lead VCDR SNVs¹⁷ explained 9.13% (95% CI, 6.6-12.0) in VCDR and 9.23% (95% CI, 6.7-12.1) in VDD-adjusted VCDR.

The IOP PRS demonstrated its predictive capacity by explaining 9.6% (95% CI, 8.1-11.2) of the phenotypic variance in IOP, based on 4839 samples from the BHAS dataset (Table 2). The baseline model accounted for 0.43% (95% CI, 0.1-0.9) of the phenotypic variance of IOP measurements. We observed significant improvement when comparing the new IOP PRS to IOP PRS using published IOP SNVs in the same BHAS samples (variance explained, 3.4%; 95% CI, 2.4-4.4).¹⁶ We grouped BHAS samples into 10 deciles. Figure 3 reflects the rising VCDR and IOP values within each decile of BHAS samples, suggesting a positive correlation between genetic risk and VCDR and IOP measurements in the BHAS cohort.

We further evaluated the VCDR or IOP PRS in African, East Asian, and South Asian samples within the UK Biobank dataset. We observed an association between the VCDR PRS and VCDR in East Asian (R², 12.1%; P < .001) and South Asian (R², 14.3%; P < .001) populations. The VCDR PRS variance explained 5.2% (95% CI, 3.6-7.1), 12.1% (95% CI, 7.5-17.5), and 14.3% (95% CI, 9.3-19.9), and the IOP PRS variance explained 2.3% (95% CI, 1.5-3.3), 3.2% (95% CI, 1.3-5.8), and 7.5% (95% CI, 6.2-8.9) (P < .001) across African, East Asian, and South Asian populations, respectively (Table 2). The variance explained in people of African ancestry in UK Biobank was one-fourth that seen in European individuals. For IOP, we observed similar reductions in prediction accuracy for ancestry groups with increasing genetic distance from Europe (South Asian prediction > East Asian prediction > African prediction) (Table 2). These reductions in predictive accuracy for non-European populations are broadly in keeping with the results seen for other complex traits.³⁰ Nevertheless, while the explained variance may be lower, the VCDR and IOP PRSs still offered some degree of predictive capability across all populations.

In this genetic association study, we constructed PRSs that provided predictions for IOP and VCDR, 2 key glaucoma risk factors. We validated these PRSs in independent participants from the CLSA dataset, followed by further confirmation in BHAS. The validation results showed associations between the VCDR and IOP PRSs and the corresponding VCDR and IOP measurements. These PRS demonstrated their potential value in stratifying individuals based on their VCDR or IOP values; for example, comparing the top and reference PRS deciles (fifth decile) led to approximate 0.15-unit and 2.5–mm Hg increases in VCDR and IOP, respectively (Figure 3). These effect sizes were substantial; in comparison, the moderate penetrance 368X variant in MYOC increased IOP by approximately 2 mm Hg.³¹ This underscores the clinical potential of VCDR and IOP PRSs in inferring how a person’s glaucoma risk may become manifest (via elevated VCDR, IOP, or both).

We applied our VCDR predictions to 2 cohorts, with better prediction performance seen in CLSA than in BHAS. A key difference between CLSA and BHAS was that CLSA used an AI-derived phenotype, whereas BHAS used a manually graded VCDR phenotype. Previous work¹⁷ has shown that AI-derived VCDR phenotypes are more accurate than manually graded phenotypes, and this is likely to explain some of the improvement in VCDR PRS performance in the CLSA cohort. The VCDR estimates from BHAS and CLSA also showed a slightly different distribution of values (Figure 2) for the 2 cohorts; this is partly explained by the AI measure used and partly due to the CLSA cohort being older.

VCDR and IOP have been well established as crucial endophenotypes for glaucoma in previous research.³² Our findings indicate that VCDR and IOP PRSs could potentially predict patients’ lifetime risk of high VCDR or IOP. While the translation of IOP and VCDR PRSs still needs more research, in theory, the stratification of risk through a VCDR or IOP PRS can be performed even before the clinical manifestation of the disease. A previous study³³ has reported that VCDR-related loci were associated with normal-tension glaucoma, thus preclinical risk assessment by a VCDR PRS proves valuable in identifying individuals with high risk of normal-tension glaucoma, making the PRS a potentially important tool for screening purposes. Additionally, previous studies have indicated associations between IOP-related PRS and key glaucoma indicators. Specifically, an earlier IOP PRS³⁴ has been linked to parameters such as mean deviation in automated static perimetry, suggesting its importance in understanding both focal and diffuse visual field loss. The earlier IOP PRS was also reported to be associated with age at the diagnosis of glaucoma as an indicator of the progression of POAG.^13,35 Our PRSs appeared to be more accurate than the PRS possible from previous work^15,16 because it was based on a larger VCDR and IOP GWAS. These associations further highlight the potentially multifaceted utility of PRSs in understanding and managing glaucoma risk.

This study has several strengths. First, we used the summary statistics from the study by Han et al,³⁶ which applied the MTAG approach to perform a large-scale multitrait analysis of GWAS. MTAG leveraged the power of genetic correlations across input traits that boosted the statistical power of GWAS, offering more accurate estimations of SNV effect sizes and improved P values. Previous studies^13,37 have shown that the MTAG-derived PRS also demonstrated elevated predictive capability. Second, we generated SNV weights via SBayesRC that incorporate functional genomic annotations.²⁸ The traditional C + T method selects SNVs for inclusion in the analysis by applying inclusion criteria based on GWAS association P value thresholds, resulting in SNVs with small effect sizes being dropped and reducing the predictive capability. Other recent PRS methods (eg, PRS-CS²⁷ and megaPRS²⁶) rely on estimating the modified effects of all common SNVs from the GWAS summary statistics by linkage disequilibrium or per-SNV heritability, but these methods usually only consider a limited set of SNVs (approximately 1 million) from the GWAS summary statistics. SBayesRC imputes SNVs, and the full set of imputed SNVs is used for the PRS computation. Moreover, imputed SNVs are also incorporated with functional genomic annotation to obtain a better identification of causal effect size, which leads to increased predictive accuracy.

In the present study, we have focused on deriving a new PRS for IOP and a new PRS for VCDR. Previous work has used a multitrait model to derive a PRS for glaucoma risk (rather than for the glaucoma risk factors considered here).¹³ The PRSs for IOP and VCDR appear to complement a PRS for glaucoma risk because they allow inference as to whether a person’s glaucoma risk may be driven by pressure, nerve head damage, or both. In some clinical applications, current IOP and VCDR values may be available for a patient but a person’s current IOP or VCDR may not reflect their future values for these key glaucoma risk parameters. Our potentially improved genetics-based predictions, applicable anytime from birth, may provide reasonably accurate predictions; for example, our baseline + PRS model for VCDR predicted 25.8% of the variance in VCDR.

There are several limitations to this study, which may affect the interpretation of the results. One important limitation of this work is that the PRSs were derived from European ancestry samples and became steadily less predictive (Table 2) with increased genetic distance from European ancestry. While the genetics-based predictions we derived here were least accurate in African individuals, a future PRS in African individuals may be more important than in other groups because overall IOP levels are systematically high in African individuals.¹⁷ In keeping with this, POAG prevalence is elevated in groups such as African individuals.³⁸ Future studies should endeavor to develop more accurate genetics-based risk prediction models that work across a range of ancestral groups; larger GWAS sample sizes in non-European groups will be critical for improving prediction accuracy. Additionally, it is important to note that, to date, these findings have not been demonstrated to impact clinically relevant outcomes in glaucoma when applied in practice. Further research is necessary to explore the potential clinical implications and applications, as well as to determine whether these results can be translated into meaningful improvements in patient care and disease management.

In conclusion, we derived new IOP and VCDR PRSs based on the latest GWASs for the respective traits. Relative to what was possible with previously available data, we showed potential improvements in accuracy for predicting VCDR or IOP. The PRSs for IOP and VCDR may complement existing and upcoming PRSs assessing a person’s risk of developing glaucoma.

Accepted for Publication: September 23, 2024.

Published Online: November 21, 2024. doi:10.1001/jamaophthalmol.2024.4856

Corresponding Author: Weixiong He, MsC, Queensland Institute of Medical Research Berghofer Medical Research Institute, 300 Herston Rd, Herston, Brisbane, QLD 4006, Australia (weixiong.he@qimrberghofer.edu.au).

Author Contributions: Mr He and Dr MacGregor had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: He, Gharahkhani, Craig, Mackey, MacGregor.

Acquisition, analysis, or interpretation of data: He, Lee, Diaz Torres, Han, Hunter, Balaratnasingam, Craig, Hewitt, MacGregor.

Drafting of the manuscript: He, MacGregor.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: He, Diaz Torres, Han, MacGregor.

Obtained funding: Gharahkhani, Craig, Hewitt, Mackey, MacGregor.

Administrative, technical, or material support: Lee, Hunter, Balaratnasingam, Mackey, MacGregor.

Supervision: Gharahkhani, Craig, Hewitt, Mackey, MacGregor.

Conflict of Interest Disclosures: Prof Craig and Drs Hewitt and MacGregor are cofounders of and hold stock in Seonix; in addition, Prof Craig has a patent for AU 2019/290035 A1 pending. No other disclosures were reported.

Funding/Support: This study was funded by the Australian National Health and Medical Research Council (Program Grant 1150144 and Centre of Research Excellence1116360 [Prof Craig and Drs Hewitt, Mackey, and MacGregor] and fellowship and investigator grant funding [Prof Craig and Drs Gharahkhani, Hewitt, Mackey, and MacGregor]), the Western Australia Department of Health (Dr Lee), and the BrightFocus Foundation (Dr Gharahkhani).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The opinions expressed in this manuscript are the authors’ own and do not reflect the views of the Canadian Longitudinal Study on Aging or any affiliated institution.

Data Sharing Statement: See Supplement 2.

Additional Information: This project used data from UK Biobank under application number 25331. This research was made possible using the data and biospecimens collected by the Canadian Longitudinal Study on Aging (CLSA). Funding for the Canadian Longitudinal Study on Aging (CLSA) is provided by the Government of Canada through the Canadian Institutes of Health Research (CIHR) under grant reference LSA 94473 and the Canada Foundation for Innovation, as well as the following provinces: Newfoundland, Nova Scotia, Quebec, Ontario, Manitoba, Alberta, and British Columbia. This research has been conducted using the CLSA Baseline Comprehensive Dataset version 4.0, Follow-Up 1 Comprehensive Dataset version 1.0, Genome-Wide Genetic Data Version 3.0, and the retinal image data under application ID 190225. The CLSA is led by Parminder Raina, PhD, McMaster University; Christina Wolfson, PhD, McGill University; and Susan Kirkland, PhD, Dalhousie University. The Busselton Healthy Aging Study is supported by grants from the Government of Western Australia (Department of Jobs, Tourism, Science and Innovation and Department of Health), the Commonwealth Government (Department of Health), the City of Busselton, and from private donations to the Busselton Population Medical Research Institute, along with infrastructure support from the Western Australian Country Health Service.