External Validation of SCORE2-Diabetes in Hong Kong Across Age and Chronic Kidney Disease Groups: A Regionally Representative Cohort Study

The 2023 ESC Guidelines recommend SCORE2-Diabetes for predicting 10-year cardiovascular disease (CVD) risk and guiding personalised risk-stratified interventions in people with diabetes. This study aimed to validate SCORE2-Diabetes in Hong Kong and to assess its generalizability across age groups and renal function subgroups.

This study utilized data derived from the Hong Kong RAMP-DM cohort (2009–2011), comprising patients with type 2 diabetes mellitus (T2DM) aged 40–89 years who had no prior history of CVD. The predictive performance of the original, uncalibrated SCORE2-Diabetes model and four region-specific recalibrated versions for 10-year CVD risk was assessed. Cardiovascular endpoints were defined in strict conformity with the SCORE2-Diabetes criteria, encompassing non-fatal myocardial infarction (MI), non-fatal ischaemic stroke (IS), and fatal CVD. We used the C-index to assess the discrimination performance of SCORE2-Diabetes.

Of 108,726 Hong Kong patients with T2DM (mean follow-up: 8.6 years), 14,644 (13.5%) experienced CVD events, and 13,970 (12.8%) died of non-CVD causes. The SCORE2-Diabetes model yielded a C-statistic of 0.704 (95% CI: 0.698–0.710) in men and 0.734 (95% CI: 0.728–0.739) in women for predicting 10-year CVD. Though Hong Kong is classified as a low CVD risk region, the low-risk region recalibrated model underestimated the 10-year CVD risk by 11.6% in women and 17.3% in men in Hong Kong. The uncalibrated model exhibited optimal calibration, overestimating risk by only 0.6% in women and underestimating by 0.1% in men. The uncalibrated model categorized 35.89% of men and 25.60% of women as high-risk, and 22.37% of men and 21.45% of women as very-high-risk. For patients aged 40–69 years and those with CKD stages G1–G2, the model’s C-statistic ranged from 0.619 to 0.694 but declined below 0.6 in older adults and patients with advanced CKD.

Following optimization of scaling factors, SCORE2-Diabetes exhibits accurate CVD risk prediction in the Hong Kong population. However, its discrimination and calibration ability differs across age strata and CKD stages. Notably, additional clinical judgment is warranted in older adults and patients with advanced CKD. Our findings emphasize the imperative of validating and recalibrating SCORE2-Diabetes before routine clinical implementation.