NT-proBNP outperforms ESC/EASD risk model for 10-year fatal CVD risk prediction in T2DM cohort
Performance of the recommended ESC/EASD cardiovascular risk stratification model in comparison to SCORE and NT-proBNP as a single biomarker for risk prediction in type 2 diabetes mellitus.
The recent ESC/EASD guidelines recommend, for the first time, the use of a CVD risk stratification model for a more personalized approach of treatment in patients with prediabetes and diabetes . The ESC/EASD risk model stratifies patients with diabetes into three different risk categories based on 10-year risk estimates for fatal CVD. However, the predictive performance of this model has not been verified in individuals with diabetes yet.
The Systematic COronary Risk Evaluation (SCORE) model is used to estimate the 10-year risk of fatal CVD in the general population and suggested to be used for a rough CVD risk assessment in individuals with diabetes . Yet the predictive performance has not been demonstrated in patients with long-standing diabetes. Also, the NT-proBNP biomarker has prognostic value for CVD outcomes in patients with diabetes, but is not recommended in the new guidelines to use for CVD risk estimations in these patients [1,3-6]. This study evaluated the prognostic performance of the ESC/EASD risk stratification model compared with the SCORE estimation model and the single NT-proBNP biomarker for 10-year fatal CVD risk in patients with T2DM.
Patients with T2DM were included in a prospective registry from December 2005 through January 2010. Data from patients (n=1690) from 4 diabetes outpatient clinics were analyzed (including measurement of baseline eGFR, NT-proBNP levels, and urine albumin/creatine ratio [UACR]). For the ESC/EASD cardiovascular risk categories of 10-year risk of CVD death, patients were categorized in 3 risk groups: moderate (<5%) risk, high (5-10%) risk, and very high (>10%) risk. Using the SCORE risk chart for 10-year fatal CVD risk based on age, sex, smoking status, total cholesterol, and SBP, patients were categorized as <5%, 5-10%, and >10% risk groups. The SCORE risk estimates were multiplied by 2 for men and 4 for women to account for increased CV risk in T2DM. Patients were categorized based on tertiles of NT-proBNP (1st tertile: 59 pg/mL [IQR 59-59], 2nd tertile: 122 pg/mL [IQR 90-156], 3rd tertile: 376 pg/mL [IQR 267-648]) or as two groups with a cut off of 125 pg/mL. The primary endpoint was CVD death at 10 years. Additional secondary outcomes were all-cause death at 10 years, and 5-year hospitalization for CVD or all-causes.
- The ESC/EASD risk model classified 25 (1.5%) patients as moderate, 252 (14.9%) as high, and 1125 (66.6%) as very high risk for 10-year fatal CVD. 288 Patients were unclassifiable based on the ESC/EASD criteria. The SCORE model estimated 654 (39%) patients with a <5% risk, 525 (31%) patients between 5% and 10% and 511 (30%) with a >10% risk. 871 Patients had normal NT-proBNP (<125 pg/ml) levels vs. 819 with elevated levels.
- The ESC/EASD risk model did not accurately predict the primary endpoint of 10-year CVD death, but was significantly associated with the secondary outcome of 10-year all-cause death (<5% risk vs. 5%-10% risk P=0.046 and <5% risk vs. >10% risk P=0.031). In contrast, the SCORE risk estimation model and NT-proBNP were significantly associated with both 10-year CVD death and all-cause death (all P<0.001).
- Receiver operating characteristic (ROC) curves for the risk models showed superiority of NT-proBNP over the ESC/EASD stratification model for 10-year CVD death and all-cause death in patients with T2DM (C-index CVD death: 0.80 vs. 0.53, respectively, P<0.001 and C-index all-cause death: 0.73 vs. 0.52, respectively, P<0.001). The results were similar when NT-proBNP was used as a categorical variable based on tertiles. Also the SCORE estimated risk model outperformed the ESC/EASD risk model for both fatal outcomes (C-index CVD death: 0.64 vs. 0.53, respectively, P=0.001 and C-index all-cause death: 0.66 vs. 0.52, respectively, P<0.001).
- Discriminative performance of the three risk models in patients without CVD (n=1379) and patients without CVD between 40 and 64 years (n=707) showed superiority of NT-proBNP over the ESC/EASD model for 10-year CVD death and all-cause death (all P<0.001). SCORE performed best in diabetic patients without CVD aged 40-64 years. The ESC/EASD model gave the lowest C-indices for CVD deaths and all-cause deaths in both subgroups.
- The discriminatory performance of NT-proBNP for hospitalization for all causes and unplanned CVD events at follow-up of 5 years was superior to the ESC/EASD model (C-index CVD hospitalization: 0.74 vs. 0.54, respectively, P<0.001 and C-index all-cause hospitalization: 0.62 vs. 0.55, respectively, P<0.001). Also, the SCORE estimated risk model had a significantly better performance compared to the ESC/EASD model (C-index CVD hospitalization: 0.62 vs. 0.54, respectively, P=0.003 and C -index all-cause hospitalization: 0.59 vs. 0.55, respectively, P=0.04).
This study showed that the ESC/EASD risk stratification model was less reliable in risk predictions for fatal CVD or all-cause death in patients with T2DM than the SCORE risk estimation and single NT-proBNP assessment, with NT-proBNP as most robust risk estimator. Both NT-proBNP and SCORE were associated with 10-year fatal CVD and all-cause death and 5-year all-cause hospitalization in patients with T2DM, while the ESC/EASD model was associated only with the secondary outcomes 10-year all-cause death and 5-year hospitalization due to all-causes.