Why the study?
There is an overabundance of cardiovascular disease risk-prediction models applicable to patients with Type 2 diabetes, but most still require external validation.
Do cardiovascular risk scores accurately predict incident cardiovascular disease events in a Spanish cohort of patients with Type 2 diabetes and no prior cardiovascular disease?
Do cardiovascular risk scores accurately predict incident cardiovascular disease events in a Spanish cohort of patients with Type 2 diabetes and no prior cardiovascular disease?
Models derived for or adapted to diabetes patients, or incorporating diabetes-related metabolic measures, demonstrated better performance in predicting CVD risk in a Spanish T2D cohort, with SCORE2-diabetes and ADVANCE showing optimal calibration and simplicity.
Supports diabetes-specific scores for Spanish T2D risk stratification; extends validation but leaves generalizability open.
AIMS: There is an overabundance of cardiovascular disease (CVD) risk-prediction models applicable to patients with Type 2 diabetes (T2D), but most of them still require external validation. Our aim was to assess the performance of 18 CVD risk scores in a Spanish cohort of patients with T2D. METHODS AND RESULTS: The CARdiovascular Risk in patients with DIAbetes in Navarra (CARDIANA) cohort, which includes 20 793 individuals with T2D and no history of CVD, was used to externally validate 13 models developed in patients with T2D [Action in Diabetes and Vascular Disease (ADVANCE), Atherosclerosis Risk in Communities, Basque Country Prospective Complications and Mortality Study risk engine, Cardiovascular Healthy Study, Diabetes Cohort Study, DIAL2, DIAL2-extended, Fremantle, Kaasenbrood, Swedish National Diabetes Register (NDR), PREDICT1-diabetes, SCORE2-diabetes, and Wan] and 5 models developed in the general population (ASCVD, PREVENT-basic, PREVENT-full, QRISK2, and SCORE2). Harrell's C-statistic and calibration plots were used as measures of discrimination and calibration, respectively. There were 991 incident CVD events within 5 years of follow-up, resulting in a cumulative incidence of 5.0% (95% confidence interval 4.7-5.3). Discrimination ability was moderate for all the models, with SCORE2-diabetes, NDR, PREDICT1-diabetes, PREVENT-full, Wan, ADVANCE, and both DIAL2 models showing the highest C-index values. All models showed good calibration, although most of them required recalibration, with the exception of ADVANCE-, DIAL2-, and SCORE2-related models. CONCLUSION: In our context, models derived for or adapted to diabetes patients, as well as models derived in the general population but incorporating diabetes-related metabolic measures (such as Hb1Ac) as predictors, demonstrated better performance than the others. DIAL2, DIAL2-extended, SCORE2-diabetes, and ADVANCE showed optimal calibration even without recalibration, which implies greater applicability, especially for SCORE2-diabetes and ADVANCE because of their simplicity.
No takes yet. Share an insight, caveat, or question.
Enguita‐Germán et al. (2025) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: