Model-Dependent Cardiovascular Risk Stratification in Obese Populations: Multicenter Study in Türkiye

Background: Cardiovascular risk prediction models are essential for preventive cardiology; however, most were developed in populations with limited representation of individuals with severe obesity. Given the high prevalence of obesity in Türkiye and its strong association with cardiometabolic disorders, uncertainty exists regarding the applicability and agreement of contemporary cardiovascular risk estimation tools in this population. This study aimed to compare cardiovascular risk estimates derived from SCORE2, SCORE2-DM, and AHA PREVENT models across geographical regions in a large obese cohort from Türkiye, and to formally evaluate inter-model agreement and confounder-adjusted regional differences. Methods: This multicenter retrospective analysis included 6,378 obese individuals recruited from seven geographical regions of Türkiye. Demographic characteristics, comorbidities, anthropometric measurements, and laboratory data were obtained from standardized electronic medical records. Ten-year cardiovascular risk was estimated, rather than prospectively predicted, using SCORE2, SCORE2-DM, and AHA PREVENT equations. Regional comparisons were performed using the Kruskal–Wallis test and chi-square analysis. Inter-model agreement was assessed using Spearman correlation, intraclass correlation coefficients (ICC), Bland–Altman analysis, and linear-weighted Cohen’s kappa for guideline-based risk strata. Multivariable linear regression was used to evaluate regional differences after adjustment for age, sex, smoking, diabetes, hypertension, dyslipidemia, statin use, and body mass index. Statistical significance was set at p< 0.05. Results: Substantial regional heterogeneity was observed in smoking prevalence, cardiometabolic comorbidities, and medication use. Cardiovascular risk estimates differed according to the selected prediction model: SCORE2 identified the highest median risk in Mediterranean and Southeastern Anatolia, SCORE2-DM demonstrated relatively homogeneous risk estimates across regions, and PREVENT showed greater regional discrimination, identifying the highest risk in the Aegean region and the lowest risk in Eastern and Central Anatolia. SCORE2 and PREVENT showed strong rank-order agreement (Spearman ρ =0.95) but only modest categorical agreement (weighted κ=0.31). SCORE2-DM systematically estimated 10-year risk approximately 6 percentage points higher than PREVENT (mean difference 6.37%, 95% LoA 5.70 to 7.04). Regional differences in risk estimates remained statistically significant after multivariable adjustment for all three models (all p≤ 0.009). Conclusion: Cardiovascular risk estimation in obese individuals is highly sensitive to the choice of prediction model and regional population characteristics. Although the three contemporary models rank patients similarly, their absolute risk estimates and categorical risk classifications can diverge substantially, particularly in obese patients with diabetes. Because this analysis is based on model-derived estimates rather than observed cardiovascular outcomes, our findings should be interpreted as descriptive risk estimation rather than validated risk prediction. These findings highlight the importance of cautious interpretation of risk scores and emphasize the need for outcome-based, population-specific validation of cardiovascular risk prediction tools in high-risk obese populations.