Study measures discrimination in models that predict heart failure in patients with CKD
Click Here to Manage Email Alerts
Compared with other models, N-terminal pro-brain natriuretic peptide is a low burden approach that gives the most “moderate discrimination” while predicting heart failure in patients with chronic kidney disease, data at ASN Kidney Week show.
“Patients with CKD have a roughly three times greater risk of heart failure [HF] compared with patients without CKD, both because they have a higher burden of traditional risk factors as well as CKD-specific factors that can contribute to heart failure,” Leila R. Zelnick, PhD, from the University of Washington, said. “We assessed prognostic value of cardiac biomarkers and echocardiographic (echo) variables for HF prediction compared to a published clinical equation in the Chronic Renal Insufficiency Cohort [CRIC].”
Using Cox regression analysis, Zelnick and colleagues compared the discrimination of the 11-variable Atherosclerosis Risk in Communities (ARIC) HF prediction equation to cardiac biomarkers and echo measures to estimate 10-year risk of HF hospitalization. The study included 2,146 participants (mean age, 59; 53% men; 43% were Black) from the CRIC without prior HF and with complete clinical, cardiac biomarker and echo data.
“For each model, we evaluated the discriminatory ability via the 10-fold cross-validated Harrell’s C-index and evaluated the model’s calibration both graphically and with the Grønnesby and Borgan test,” Zelnick said. “We assessed discrimination with internally valid, 10-fold cross-validated C-indices.”
Within the 6.7 years of follow-up, 268 participants experienced incident HF hospitalizations.
The ARIC HF model with clinical variables had a C-index of 0.68. Similarly, high sensitivity troponin alone (C-index, 0.69) and left ventricular mass plus left ventricular ejection fraction (C-index, 0.71) were comparable to the ARIC model. However, N-terminal pro-brain natriuretic peptide (NT-proBNP) alone had better discrimination (C-index, 0.72; P = .04).
“The ARIC heart failure prediction model for 10-year heart failure risk had modest discrimination and poor calibration in this patient population. NT-proBNP and high sensitivity troponin T together had moderate discrimination of point 0.73, which was similar to left ventricular mass and left ventricular ejection fraction alone,” Zelnick said. “Adding clinical variables further boosted performance, and models that utilize these commonly measured cardiac biomarkers may provide a low burden approach to predicting heart failure in CKD until CKD-specific heart failure prediction models can be developed.”