糖心vlog

We read with interest 3 studies examining the potential clinical utility of polygenic risk scores (PRSs) for primary open-angle glaucoma (POAG).^1-3 A PRS is a quantitative metric, often incorporating information from thousands or even millions of genetic variants, which summarizes an individual鈥檚 genetic risk of a condition into 1 number. Glaucoma is one of the most heritable of all common human diseases. Recent very large genome-wide association studies (GWAS) have led to a step change in our understanding of glaucoma鈥檚 complex genetic causes. These GWAS were enabled by huge cohort studies, such as the UK Biobank,⁴ and a strong global collaborative community (the International Glaucoma Genetics Consortium).⁵ These GWAS inform the genetic variant weightings that underpin PRSs. The derived PRSs reflect the outcomes measured in the original GWAS. Therefore, to date, POAG PRSs primarily reflect the risk of POAG in the general population rather than more specific clinical parameters, such as disease severity or risk of progression. A high PRS helps to identify prevalent POAG in independent studies.⁶ Therefore, a potential clinical translation of these PRSs could be to enable targeted glaucoma screening. While a PRS alone is, of course, not diagnostic for glaucoma, it is a potential prescreening test enriching the proportion of POAG and thus increasing the positive predictive value of subsequent screening tests, such as intraocular pressure (IOP) and optic disc imaging.

Most POAG-related GWAS to date have studied people of European ancestry. It is unclear how well PRSs predict POAG in non-European populations. If a PRS approach could enable targeted population screening, it is important that the approach works well for people of all ancestries. We therefore welcome the study by Chang-Wolf and colleagues,¹ which examined the performance of POAG PRSs in multiple studies of individuals with African ancestry; the results are generally promising, with a high PRS effectively enriching for POAG in all cohorts, despite the weighting of the PRSs being largely based on European studies. While the POAG odds ratios comparing the top quintile to bottom quintile of PRS were smaller for studies involving individuals with African heritage than for European heritage, these results could also be influenced by the number of cases in the bottom quintile, which may be less clinically relevant. And while the increase in area under the receiver operating characteristic curve (AUROC) achieved by adding the PRS to baseline factors was lower in the studies with participants of African ancestry, the final AUROC was actually higher for most of these studies than the European studies. Overall, these results support the utility of the PRS approach for POAG enrichment in people of African ancestry, and it is hopeful that there will be further improvements with larger GWAS in non-European populations and with modern PRS methods that consider cross-ancestry genetic architecture variations. However, population-based (rather than just case-control) studies in non-European participants are now needed so that the performance of a PRS at a population level can be evaluated.

He and colleagues² examined the ability of PRS derived specifically for IOP or vertical cup-disc ratio (VCDR) to predict actual clinical measurements of IOP or VCDR in 2 independent cohort studies. They found that these PRSs could predict a substantial amount of VCDR (20%-22%) and IOP (10%-13%). It is perhaps unsurprising that a PRS can predict a one-off measurement of VCDR better than IOP, given how physiologically variable IOP is compared to VCDR. Despite a superior predictive performance for VCDR, it is genetic determinants of IOP that are most strongly associated with POAG and form the strongest basis for increasing the power of POAG GWAS and PRS derivation using multitrait methodology.⁶ The authors posit reasonable clinical utility hypotheses in relation to their results. They suggest that the IOP PRS may better predict patients who need more intensive IOP-lowering therapy and that the VCDR PRS may predict patients who develop glaucoma despite lower IOP. Clearly, these hypotheses need to be tested in longitudinal datasets (ideally prospectively) before these PRSs can be implemented in clinical care. Outside ophthalmology, PRSs for disease onset or prevalence are often not strongly associated with disease prognosis or response to treatment, and it may be necessary to derive new PRSs on the basis of longitudinal clinical data in the future.

A 2016 study by Scheetz and colleagues⁷ demonstrated that risk variants in the TMCO1 gene can predict conversion to POAG in the Ocular Hypertension Treatment Study (OHTS) with surprising strength. It was therefore exciting to see a study examining the PRS approach to prediction of POAG conversion in OHTS.³ Sekimitsu and colleagues³ demonstrated that a PRS externally trained to predict POAG prevalence was able to predict disease-free status at 20-year follow-up; a binary PRS cutoff lower than the 48th percentile indicated a 1.5-fold higher chance of being disease free. This strength of association, while potentially clinically meaningful, is perhaps not much stronger than for the TMCO1 risk allele alone,⁷ and future replication studies will help determine the most consistently effective and generalizable approach to genetic prediction. As previously discussed, risk factors for the onset of a disease are often different to those for the progression of disease, and therefore a PRS based on disease prevalence may not be the best approach for predicting progression. It is encouraging that this study included participants of African ancestry in addition to participants of European ancestry, again supporting the concept that future personalized care approaches using PRS can be implemented equitably. A particularly interesting finding in the Sekimitsu and colleagues study³ is that the PRS effect was mainly evident in participants at the highest baseline risk of conversion using the well-established OHTS risk factors (age, baseline IOP, central corneal thickness, pattern standard deviation, and VCDR). If these higher-risk patients are usually treated anyway, could the potential utility of the PRS be to decide whom among those with high baseline risk not to treat, and is the PRS association strong enough to warrant withholding treatment in these patients? Again, future replication studies will help clarify the potential use case for PRS in ocular hypertension.

As research evidence grows for the utility of polygenic prediction tools, we need to plan how best to implement these approaches into clinical care. Treating ophthalmic specialists may choose to request genetic testing specifically for the care of a patient鈥檚 glaucoma, but this may be less cost-effective than leveraging genetic data that are more readily available for all patients as part of a drive from a health care system to genotype all their users. Widespread genotyping is already established in Iceland and Finland and is on the way in the UK, with the Our Future Health study recruiting 5 million participants (over 10% of UK adults). The genetic data in these studies can inform not only a personalized approach to detection and treatment of glaucoma, but will also inform the management of a wide range of other health conditions. Given the growing evidence, it seems likely that people at risk of vision loss due to glaucoma will be some of the earliest beneficiaries of a new era of genomics-enabled personalized care.

Corresponding Author: Anthony P. Khawaja, PhD, University College London Institute of Ophthalmology, 11-43 Bath St, London EC1V 9EL, United Kingdom (anthony.khawaja@ucl.ac.uk).

Published Online: November 21, 2024. doi:10.1001/jamaophthalmol.2024.5103

Conflict of Interest Disclosures: Dr Khawaja reported personal fees from AbbVie, Google Health, Heidelberg Engineering, Novartis, Reichert, Santen, Thea, and Topcon. Dr Foster reported personal fees from AbbVie, AlphaSights, GLG, Google Health, Guidepoint, PwC, and Santen and grants from Alcon Research, Fight for Sight (London), and The Desmond Foundation. No other disclosures were reported.