Prognostic value of epicardial adipose tissue distribution as a complementary risk marker for major adverse cardiovascular events in diabetes

This study aimed to compare the predictive capabilities of BMI, WHR, Epicardial Adipose Tissue (EAT), and Visceral Adiposity Index (VAI) for MACE in individuals with diabetes mellitus. It further sought to explore whether increased EAT or visceral fat is a more significant factor for MACE than general obesity in this population. A prospective study was conducted involving 1071 individuals with diabetes mellitus (64% male, median age 58 years). EAT volume and spatial distribution were obtained by manually delineating cardiac magnetic resonance (CMR) images using CV142 software. MACE were defined as ischemic heart disease, hypertensive heart disease, heart failure, stroke, and cardiovascular disease (CVD)-related death. Baseline data, CMR parameters, and different obesity indices were compared among groups. The Kaplan–Meier (KM) curve was used to explore the relationship between different obesity indices and MACE in individuals with diabetes mellitus. The predictive performance of each index for MACE occurrence was compared using multi-model Cox multivariate regression analysis, and sensitivity analysis was performed. Over a median follow-up of 1508 days, 118 MACE occurred. The MACE and non-MACE groups differed in EAT volume, EAT distribution, and WHR. Among comorbidity subgroups, differences were observed in EAT volume, EAT distribution, BMI, WHR, and VAI. Kaplan-Meier analysis indicated that EAT distribution and WHR predicted MACE. In multivariable Cox regression, EAT location (EAT.LOC) predicted MACE in all three models and remained significant in the fully adjusted model (Model 3: HR: 1.76 → 1.85). In contrast, EAT volume, VAI, BMI, and WHR showed no predictive ability in any model. In within-model comparisons, EAT.LOC demonstrated superior predictive value (Model 3: ΔC-index = 0.016, NRI = 0.163, IDI = 0.112). Compared to traditional cardiovascular risk factor models, the model incorporating EAT.LOC also demonstrated favorable predictive value (ΔC-index = 0.013, NRI = 0.163, IDI = 0.07). Sensitivity analysis confirmed the robustness of these findings. EAT distribution, but not BMI, WHR, VAI, or EAT volume, independently predicts MACE risk in patients with diabetes mellitus. These findings indicate that EAT.LOC independently predicts MACE in diabetic patients and could be used for secondary risk stratification in this population. Nevertheless, the causal link between EAT.LOC and MACE remains to be established in future studies.