We prospectively evaluated whether the N-terminal brain natriuretic peptide (NT-ProBNP) reduction percentage, during hospitalization for acutely decompensated heart failure (HF), has a prognostic significance in 6-month follow-up. In 120 patients consecutively admitted for acute HF to an internal medicine unit, plasma NT-ProBNP was measured on admission and at discharge. During a 6-month follow-up 52 (43.3%) patients had events: 9 (7.5%) died from cardiovascular causes, and 43 (35.8%) were readmitted for HF. In patients without events, the mean reduction percentage of NT-ProBNP was greater than in patients with events (39.5 ± 7.4 versus 26.3 ± 5.9%; P = 0.04). In receiver operating characteristic curve analysis, the mean area under the curve for NT-ProBNP reduction percentage was 0.63 (95% CI, 0.51–0.75; P = 0.04) for the composite end point (death or readmission), and 0.81 (95% CI, 0.65–0.97, P = 0.01) for cardiovascular mortality. NT-ProBNP reduction percentage less than 30% was the best cut-off for the identification of patients at risk of events. We suggest that in clinical practice the evaluation of change of NT-ProBNP levels during admission is probably more helpful than predischarge NT-ProBNP absolute value.