Key result
ASCVD and PREVENT models underestimate 10-year CVD risk in type 1 diabetes by >2-fold.
Why the study?
Cardiovascular disease limits quality of life and survival, and its accelerated course in type 1 diabetes may challenge standard risk prediction models.
Do traditional cardiovascular disease risk prediction models accurately predict 10-year CVD risk in patients with type 1 diabetes?
Cohort (n=1,311)
Do traditional cardiovascular disease risk prediction models accurately predict 10-year CVD risk in patients with type 1 diabetes?
Answers are grounded in this paper and cite their sources.
Traditional cardiovascular risk prediction models significantly underestimate risk in patients with type 1 diabetes, highlighting the need for tailored risk estimation incorporating enhancers like CAC and HbA1c.
Introduction and Objective: Cardiovascular disease (CVD) limits quality of life and survival and its accelerated course in type 1 diabetes (T1D) may challenge standard risk prediction models. We assessed the 10-yr CVD risk performance of the Framingham Risk Score (FRS), ASCVD Risk Estimator Plus (ASCVD), AHA PREVENT, and Steno Type 1 Risk Engine (ST1RE) prediction models over 30 yrs in a T1D population, and evaluated the role of coronary artery calcification (CAC) and HbA1c as risk enhancers. Methods: Analysis included 1311 participants followed for 30 yrs during the Epidemiology of Diabetes Interventions and Complications (EDIC) study. Model-predicted CVD risk calculated at EDIC Years 1, 10, and 20 was compared to observed CVD incidence during subsequent intervals (Years 1-10, 10-20, and 20-30); calibration and c-statistics were assessed for each period. Results: At EDIC Year 20, the 10-yr observed CVD risk in this population was 1.43 (95% CI 1.36-1.49) times higher than predicted by FRS, >2-fold higher than predicted by ASCVD and PREVENT (RR 2.69, 95% CI 2.55-2.86 and 2.40, 95% CI 2.30-2.51, respectively), and lower than predicted by ST1RE (RR 0.61-0.87 at all timepoints). ASCVD and PREVENT underpredicted risk in both sexes. All models except ST1RE underpredicted risk in females. Discrimination was modest (c-statistic 0.54-0.64) across models, with the strongest performance by PREVENT. Among individuals classified as low or moderate predicted risk, those with CAC ≥100 Agatston units had an observed risk comparable to the high-risk category; those with CAC=0 had an observed risk comparable to the low-risk category. Risk reclassification based on CAC score modestly improved discrimination (maximal c-statistic increased to 0.68). HbA1c similarly enhanced risk of CVD. Conclusion: In T1D, traditional CVD risk prediction models have limited performance. Risk enhancers improved risk classification. More accurate, tailored risk estimation is needed to identify early risk mitigation interventions in T1D. Disclosure V.R. Trapani: None. B. Braffett: None. T. Costacou: None. A. Vella: Advisory Panel; Ended; Boehringer Ingelheim International GmbH. Advisory Panel; Current; Rezolute. Research Support; Current; Dexcom, Inc. Advisory Panel; Current; Neurotronic, Amylx. G. Lorenzi: None. A.L. Brown: Advisory Panel; Current; Medtronic. Consultant; Current; ReCor Medical, Inc. R. Shah: Consultant; Current; Thyrv Therapeutics. Consultant; Ended; Novimm Therapeutics. I. Bebu: None.
No takes yet. Share an insight, caveat, or question.
Trapani et al. (2026) conducted a cohort in Type 1 Diabetes (n=1,311). Traditional CVD risk prediction models (FRS, ASCVD, PREVENT, ST1RE) vs. Observed CVD incidence was evaluated on 10-year cardiovascular disease incidence compared to model-predicted risk (RR 1.43, 95% CI 1.36-1.49). In type 1 diabetes, observed 10-year CVD risk was 1.43 times higher than predicted by FRS (95% CI 1.36-1.49) and >2-fold higher than ASCVD and PREVENT predictions.
Synapse has enriched 10 related papers. Consider them for comparative context: