Abstract TH955: Framingham as Benchmark: Convergence of a Literature-Based Modular Model for Ten-Year Coronary Heart Disease Risk

Background: The Framingham Risk Score (FRS) is a foundational tool for estimating ten-year coronary heart disease (CHD) risk and is widely used in preventive cardiology. However, the FRS’s fixed coefficients limit adaptability to changing demographics and evolving epidemiology. Flexible literature-based methods may preserve alignment with FRS while allowing recalibration from contemporary evidence. Objectives: To determine whether a modular literature-based risk model can (1) replicate FRS estimates when supplied with identical inputs, and (2) maintain comparable performance when populated with contemporary epidemiological inputs. Methods: A synthetic cohort of 10,000 adults (48% male; ages 30–74) was simulated to reflect Framingham characteristics. Synthetic data ensured precise control of covariates and enabled a direct one to one comparison of calculators without bias from missingness, competing outcomes, or unmeasured factors. Risk factors included age, sex, blood pressure, cholesterol (total or LDL), HDL, diabetes, and smoking. Two conditions were tested: (1) FRS aligned using original covariate definitions and published coefficients and (2) contemporary using updated relative risks and population prevalences from recent peer reviewed research. Agreement with FRS was assessed using Pearson and Spearman correlations, intraclass correlation coefficient (ICC), Bland–Altman analysis, and ANCOVA adjusted for age and sex. Results: With FRS inputs, convergence was high (total cholesterol: r=.95, ρ=.94, and ICC=.83; LDL: r=.94, ρ=.92, ICC=.81). With contemporary inputs, convergence remained strong (total cholesterol: r=.87, ρ=.86, ICC=.81; LDL: r=.85, ρ=.86, ICC=.80 for LDL). Limits of agreement were narrower among women (±5%) than men (±6%) and modestly wider with advancing age (<50 years ±2%, >70 years ±5%). Conclusions: A modular literature-based framework can faithfully reproduce FRS predictions under shared assumptions while flexibly incorporating updated epidemiological data. This approach offers a transparent and updatable alternative to fixed coefficient risk tools and may support more responsive preventive cardiology. Prospective validation against clinical outcomes is recommended.