Development and temporal validation of five year retinal arteriosclerosis risk prediction in high risk adults from an East China cohort.

Structural variations in retinal vessels predict clinical conditions related to vascular health. This study aims to describe the incidence and distribution of retinal arteriosclerosis and to develop risk prediction models for high-risk populations. Data were obtained from a retrospective health examination cohort at Hua Dong Sanatorium, China. The primary outcome was retinal arteriosclerosis. Candidate risk predictors were selected based on biological plausibility and potential predictive ability. Risk prediction models were developed using the Cox proportional hazards model. Predictive performance was evaluated through discrimination and calibration accuracy.Internal validation was performed using bootstrap resampling (100 iterations) to estimate optimism and to report optimism-corrected discrimination (C-index) and calibration (slope). In the derivation cohort, 10,323 individuals were diagnosed with retinal arteriosclerosis over 288,525 person-years (incidence density: 35.8 per 1,000 person-years). Males over 40 and females over 45 were identified as relatively high-risk groups. Risk predictors included age, body mass index, waist circumference, hip circumference, systolic blood pressure, diastolic blood pressure, smoking, hypertension, diabetes, high-density lipoprotein, and serum creatinine. The C-index of the risk prediction models was approximately 0.8 in the validation set. These models are accessible to potential users via an online application. The risk prediction models may facilitate early intervention by modifying risk factors, thereby reducing the incidence of retinal arteriosclerosis in high-risk populations.