Page 106 - Biomarkers for risk stratification and guidance in heart failure

Serial NT-proBNP measurements during and early after HF admission.
Multivariate Cox proportional hazard regression analysis was performed with the use of all covariates associated with outcome, except renal function and NT-proBNP concentration, to assess the clinical model. Variables were added in a stepwise fashion with P < .05 and P < .1 as the cutoffs for entry or retention, respectively. After assessment of the clinical model, renal function (eGFR < 30 mL/ min, eGFR 30-60 mL/min, or eGFR >60 mL/min) was added to form the reference model. To assess the independent prognostic value of NT-proBNP concentration on admission, at discharge, inpatient change, early outpatient change, and NT- proBNP concentration at 1 month after hospital discharge, these parameters were added to the reference model in a stepwise fashion to form the final NT- proBNP model.
Model accuracy and discrimination were evaluated for both mortality and the
combined end point of HF readmission or mortality within 1 year of follow-up by
(i) c-statistic, a measure of the area under the receiver operating characteristic
curve (AUC) and (ii) integrated discrimination improvement (IDI). Calculations
were done with the use of IBM SPSS Statistics 21.0 (IBM, Armonk, New York) and
Medcalc 13.3.3.0 (Medcalc Software, Ostend, Belgium). 5
Composite NT-proBNP Score
Independent predictive NT-proBNP parameters were used to form the composite NT-proBNP score. This was done by giving each independent prognostic NT- proBNP parameter 1 point. To assess the prognostic impact of the composite NT-proBNP score, 90-, 180, and 365-day mortality and the combined end point of HF readmission or mortality were calculated for every composite NT-proBNP score category. Furthermore, Kaplan-Meier survival curves were assessed and compared with the use of the log-rank test.
Finally the value of relatively small early outpatient changes in NT-proBNP (ie, decrease vs increase in NT-proBNP concentration of <30%) was assessed in a multivariate manner.
RESULTS
Patient Characteristics
In 309 out of 345 patients included in the PRIMA study, NT-proBNP levels at admission, discharge, and 1 month after hospital discharge were available. Patient
105