EXPLORING THE UTILITY OF A CVD RISK PREDICTING RETINAL AI MODEL IN IDENTIFYING CACS OVER 400 IN PATIENTS WITH AND WITHOUT DIABETES

Therapeutic Area

ASCVD/CVD Risk Assessment

Background

Dr. Noon CVD, a CVD risk prediction model based on retinal-imaging has been shown to estimate probability of coronary artery calcium (CAC). Dr.Noon risk categories (low, moderate, high) have shown performance comparable to conventional CAC risk categories (0, 1-99, ≥100) in predicting 5- year cardiovascular risk. For guiding intensive preventive strategies, CACS ≥400 is considered clinically significant. The current study explored the utility of Dr. Noon CVD to identify CACS ≥400 in patients with and without diabetes mellitus (DM).

Methods

This analysis included 1,252 participants from the CMERC-HI cohort classified into six groups based on Dr. Noon CVD risk level (low [<31], moderate [<41], high [41–51], very high [>51]) and DM status. non-DM with low (1a), moderate (2a), high (3a) and very high (4a) risk; and DM with low (1b), moderate (2b), high (3b) and very high (4b) risk. We performed ROC analysis of the Dr. Noon CVD score for predicting CACS ≥400 and compared AUCs by DM status. Prevalence of CACS ≥400 was assessed for each group and logistic regression models calculated odds ratios (ORs), adjusting for cardiovascular risk factors.

Results

CACS ≥400 was present in 189 patients (15.1%). AUC was 0.787 (non-DM) and 0.758 (DM) with no significant difference (p=0.4186). Sensitivity and specificity across Dr. Noon CVD risk categories were comparable between DM and non-DM groups. The proportion of patients with CACS ≥400 increased incrementally from Group 1a (0.6%, 1 of 155) to Group 4b (43.0%, 64 of 149). In non-DM patients, both high- and very high-risk groups showed significantly higher odds of CACS ≥400, with clear separation even between high and very high-risk (OR=4.17, p<0.001). In DM patients, only the very high-risk group showed significant elevation vs. high-risk (OR=3.81, p<0.001), while high-risk alone was not distinguishable from lower-risk groups.

Conclusions

In addition to identifying 5y-CVD risk, Dr. Noon CVD could also be used to identify patients at increased risk of having a CACS ≥400. Future studies could evaluate the effectiveness of using Dr. Noon CVD when stratified using both Dr. Noon CVD risk category and diabetes status. Further analysis could support its clinical applicability for targeted preventive interventions.