Assessment of fatal cardiovascular disease risk using data-driven diabetes subgroups and SCORE2-Diabetes: a prospective, observational, population-based analysis of adults with diabetes in Mexico City

Background: Cardiovascular disease (CVD) is a leading cause of diabetes-related mortality in Mexico. Although diabetes subgroups capture underlying disease heterogeneity, their association and utility for risk prediction for fatal CVD in Mexican adults remain unclear. Here, we aimed to assess the risk of fatal CVD risk and their complementary role for fatal CVD risk prediction across diabetes subgroups in Mexican adults. Methods: We included adults with diabetes from the Mexico City Prospective Study (MCPS). Participants were classified into mild obesity-related (MOD), severe insulin-deficient (SIDD), severe insulin-resistant (SIRD), and mild age-related (MARD) diabetes using a self-normalizing neural network algorithm. Fatal CVD was defined as death from ischemic heart disease or stroke (ICD-10 I20-I25, I60-I69). SCORE2-Diabetes was internally recalibrated for fatal CVD outcomes only for use within MCPS, as this does not represent a formal validation of the original composite endpoint. Cause-specific Cox proportional hazards and Fine & Gray competing risk regression models were used to estimate subgroup-specific risk, and sequential models evaluated the incremental predictive value of diabetes subgroups combined with SCORE2-Diabetes and traditional risk factors. Findings: We analyzed data from 24,943 adults living with diabetes (mean age 58 ± 12, 67% female, no ethnicity data available). Over a median follow-up of 19.3 years (IQR 12.7-20.6), 2218 fatal CVD events (8.9%) were recorded. SIDD was the most prevalent subgroup (50.0%), followed by SIRD (14.1%), MARD (17.7%), and MOD (18.3%). SIDD and MARD showed the highest adjusted risk of fatal CVD (HR 1.58 [95% CI 1.38-1.81] and 1.35 [1.13-1.60]), whereas MOD and SIRD had lower risk, even when considering competing causes of non-CVD deaths. Recalibrated SCORE2-Diabetes demonstrated adequate discrimination overall (c-statistic 0.748, 95% CI 0.734-0.762) and for most diabetes subgroups but underperformed in MARD, with internal recalibration for fatal CVD outcomes improving risk assessment. The combination of diabetes subgroups and SCORE2-Diabetes modestly improved prediction for fatal CVD outcomes. Interpretation: Diabetes subgroups show heterogeneity in fatal CVD risk in Mexican adults. SIDD and MARD identify high-risk individuals and integration subgroup classification with SCORE2-Diabetes is complementary for fatal CVD risk prediction. Funding: This research was supported by a grant provided by the Bernard Lown Scholars in Cardiovascular Health Program grant number BLSCHP-2403.