ePWV as an AI-applicable Predictor as a Scalable Risk Factor for Large-Scale Glaucoma Screening: Evidence from a National Chinese Cohort

Yupeng XuQiaoyun GongMingjie ZhuChuming LuHaiyan Wang^*

• Shanghai General Hospital, Shanghai, China

Background: To examine the association between estimated pulse wave velocity (ePWV), a marker of arterial stiffness, and glaucoma incidence in Chinese cohort, highlight ePWV's artificial intelligence intelligence(AI) potential for screening, highlighting ePWV's potential as a scalable population-level risk factor for glaucoma screening Materials and methods: Data were obtained from the China Health and Retirement Longitudinal Study (CHARLS). A total of 11,968 adults aged ≥45 years without glaucoma at baseline (2011-2018) were followed for up to 7 years. ePWV was calculated from age and blood pressure and divided into quartiles. Cox proportional hazards models assessed risk, while restricted cubic spline and two-piecewise models explored dose–response patterns. Subgroup analyses tested effect modification by demographic and lifestyle factors. Results: During a 7-year follow-up, participants in the highest ePWV quartile ( ≥ 10.58 m/s) had a higher risk of glaucoma compared with the lowest quartile (<8.01 m/s) (HR[hazard ratio] 1.39, 95% CI[confidence interval] 1.00-1.93). Each 1 m/s increase in ePWV was associated with a 7% higher glaucoma risk (HR 1.07, 95% CI 1.01-1.13). The dose response relationship was linear without evidence of a threshold. Associations were consistent across most subgroups, with a stronger effect among non-drinkers (P for interaction = 0.047)), indicating that the ePWV-glaucoma association was significantly stronger in non-drinkers. Sensitivity analyses showed that ePWV was a stronger predictor of glaucoma than age or blood pressure alone. Conclusion: Higher ePWV independently links to greater glaucoma risk in middle-aged and older Chinese. This observed association indicates that ePWV provides incremental predictive value beyond traditional demographic or clinical factors. Building on this characteristic, incorporating ePWV into future artificial intelligence (AI)-enabled glaucoma screening models may potentially contribute to improving risk-stratification accuracy and facilitating early identification of high-risk individuals. As a quantifiable, accessible parameter with high AI modeling potential, it enhances algorithm performance and improves early high-risk detection when integrated into multimodal AI glaucoma screening models.