Cardiovascular risk assessment in adults with type 1 diabetes: A cross-sectional comparison of two prediction models

Accurate cardiovascular risk assessment is essential in type 1 diabetes (T1D), but agreement between disease-specific prediction models remains insufficiently characterized. This retrospective, single-center, cross-sectional study compared 10-year cardiovascular risk estimates from the Steno Type 1 Risk Engine (ST1RE) and the Scottish/Swedish risk prediction model in 100 adults aged 20-35 years with T1D and no previous atherosclerotic cardiovascular disease or major cardiovascular event. Continuous and categorical agreement were assessed using Bland-Altman analysis and linearly weighted Cohen's κ, respectively, with identical thresholds applied for risk classification. Risk estimates were strongly correlated (Spearman's ρ = 0.849, p < 0.001), with no significant paired difference on the Wilcoxon signed-rank test (p = 0.893). However, categorical agreement was fair, with a weighted κ of 0.334 and a 95% confidence interval (CI) of 0.078-0.590. The mean difference (ST1RE - Scottish/Swedish) was -1.46 percentage points (95% CI: -2.48 to -0.44), with 95% limits of agreement from -11.54 to 8.62 percentage points. Significant proportional bias indicated progressively higher Scottish/Swedish estimates as mean predicted risk increased (β = 0.91, p < 0.001). Despite strong correlation, individual-level discrepancies and proportional bias suggest limited interchangeability of these models in this cohort. Larger prospective studies assessing cardiovascular outcomes are needed to establish their comparative predictive accuracy and clinical utility.