Abstract Background/Introduction The SCORE2-Diabetes risk score was developed to predict the risk of cardiovascular events (CV-death, non-fatal MI and non-fatal stroke) in patients with T2DM without known cardiovascular disease. Current ESC guidelines recommend the use of this risk score to tailor treatment strategies in this population. Purpose We sought to determine whether the SCORE2-Diabetes risk score was able to appropriately stratify an international cohort of patients from recent cardiovascular clinical outcomes trials. Furthermore, we sought to determine the accuracy and precision of the model including in patients with and without known cardiovascular disease. Methods Patients from 4 randomized cardiovascular outcome trials (DEVOTE, LEADER, PIONEER-6, SUSTAIN-6) were included (n=23,464). Mean follow-up was 2.6 years and outcomes were adjudicated. The SCORE2-Diabetes risk score was calculated using the uncalibrated version which does not account for risk-region. Patients were stratified into low (<5%), moderate (5-<10%), high (10-<20%), and very high risk (≥20%) groups. The SCORE2-Diabetes risk predication model was designed to predict 10-year risk; however, given the mean follow-up from the cohort, outcomes were reported at 36 months. The Brier score, which evaluates estimated risk, and c-index, which evaluates the ability to discriminate between participants, were used to evaluate the precision of the risk score. We further compared the risk score with an estimated risk of MACE calculated using a cause specific Cox model that accounts for competing risk. Results Most patients in the cohort did not have prior myocardial infarction or stroke (57.4%, n=13,467). The SCORE2-Diabetes risk score was able to stratify patients into low, moderate, high, and very high risk cohorts (Figure 1). Observed cardiovascular events at 36 months were higher than the SCORE2-Diabetes risk score predicted (Table 1). The risk score, when included in a regression model, independently predicted cardiovascular death, myocardial infarction, or stroke (HRadj 1.03, p<0.001) and non-cardiovascular death (HRadj 1.05, p<0.001). Including the score in models with only age and sex improved the discrimination and performance of the model. The c-index for the risk score in the overall cohort was 0.62 and was 0.66 in the cohort without known cardiovascular disease. When using the Brier score to evaluate the utility of the model, the Brier score of the SCORE2-D model was 0.08 which is no different than a null model which assumes a baseline risk of 10% in all patients. Conclusion In a cohort of patients that was older, more geographically diverse, and at higher risk of cardiovascular events, the SCORE2-D risk score appropriately stratified patients into categories of risk. However, the overall risk prediction underestimated the risk of cardiovascular events and the discrimination of the risk score was modest.Figure:CV events over timeTable:CV events by risk group
Read full abstract