ARVO Annual Meeting Abstract, June 2024
연구 초록
Purpose : This Reti-CVD successfully predicted CVD events in individuals with no history of CVD. The present study aims to evaluate the predictive ability of retinal images and deep learning for all-cause mortality in patients with a baseline diagnosis of CVD, using the UK Biobank dataset.
Methods : Our study incorporated participants with a baseline CVD diagnosis in the UK Biobank (n=4908). We calculated Reti-CVD scores and divided these into three risk categories based on tertiles among those patients. Survival analysis was performed on the UK Biobank’s longitudinal data to evaluate the performance of Reti-CVD in predicting all-cause mortality. We used Cox proportional-hazards models to assess Reti-CVD’s predictive capability for all-cause mortality, and hazard ratios (HRs) were computed.
Results : In this longitudinal study, of the 4908 CVD patients at baseline, 121 (2.4%) experienced all-cause mortality. There were 11, 36, and 73 cases of all cause mortality in the 1st, 2nd, and 3rd Reti-CVD-based tertiles, respectively. We observed a total of 47,312 person-years over a 10-year follow-up period (median 9.88 years). After adjustment for baseline hypertension, diabetes, and smoking, Reti-CVD showed a significant association with the incidence of all-cause mortality (HR=2.19, 95% CI, 1.36-3.52, for the 3rd tertile versus 1st tertile as reference). In a gender specific subgroup analysis, Reti-CVD was significantly associated with the incidence of all-cause mortality in males (HR=2.38, 95% CI, 1.44-3.93, for the 3rd tertile versus 1st tertile as reference), but this association was not significant in females (HR=1.94, 95% CI, 0.41-9.12, for the 3rd tertile versus 1st tertile as reference).
Conclusions : Reti-CVD enabled further risk stratification in patients with cardiovascular disease. However, these findings differed by gender. While significant risk stratification was possible in males, it was not meaningful in females.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.