Abstract
BACKGROUND: Patients with diabetes mellitus (DM) face a significantly elevated risk of cardiovascular disease (CVD). However, existing risk stratification tools (e.g., Framingham Risk Score) perform suboptimally in diabetic populations. Traditional biomarkers (e.g., hs-CRP, NT-proBNP) and single-modality imaging parameters (e.g., coronary artery calcium score) have limitations, necessitating a multimodal approach to enhance predictive accuracy. OBJECTIVE: To develop and validate a multimodal diagnostic model integrating novel biomarkers (sST2, GDF-15) and imaging parameters (coronary artery calcium score [CACS], carotid plaque characteristics) for predicting cardiovascular complications in diabetic patients and to evaluate its clinical utility. METHODS: This single-center retrospective cohort study included 600 adults with type 2 diabetes (2015-2023) and no baseline cardiovascular disease. Laboratory biomarkers (sST2, GDF-15, HbA1c), imaging parameters (CACS, carotid plaque ulceration), and clinical variables were collected. Predictors were selected via LASSO regression, and a multimodal logistic regression model was constructed. Model performance was assessed using ROC curves (AUC), net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision curve analysis (DCA). RESULTS: The final model incorporated sST2, GDF-15, CACS, ulcerated carotid plaques, HbA1c, and SGLT2i use. In the validation cohort, the multimodal model achieved an AUC of 0.811 (95% CI: 0.73-0.83), outperforming biomarker-only (AUC = 0.774) and imaging-only models (AUC = 0.735). NRI and IDI were 0.52 (p < 0.001) and 0.09 (p < 0.001), respectively. DCA demonstrated superior clinical net benefit for the combined model across threshold probabilities (10-30%), with a net benefit of 0.32 at 20% risk threshold. Sensitivity and subgroup analyses confirmed model stability. CONCLUSION: The integration of biomarkers and imaging parameters significantly improves cardiovascular risk prediction in diabetic patients, offering enhanced clinical utility. Future multicenter prospective studies are strongly warranted to validate generalizability and cost-effectiveness.