The aim of this study was to evaluate the prognostic value of BNP in elderly patients hospitalised for acute diastolic cardiac failure. 108 consecutive subjects were included, aged at least 70 years old, hospitalised for isolated acute diastolic cardiac failure. All of them had a left ventricular ejection fraction > or = 50% and evidence of diastolic dysfunction on echocardiography performed shortly after admission. The plasma BNP concentration measured in the emergency department on admission was >100 pg/ml in all of the patients except five. It was positively correlated with age (R = 0.29, p = 0.002), with the plasma creatinine level (R = 0.37, p < 0.0001) and the plasma urea level (R = 0.41, p < 0.0001). On univariate analysis, compared to the patients who survived, the 20 patients who died before discharge were significantly older (88.6 versus 84.4 years, p = 0.01), and were more often residents of a care home (60 versus 31%. p = 0.02), had a lower systolic blood pressure on admission (127 +/- 33 versus 154 +/- 30 mm Hg), a higher plasma urea level (16.8 +/- 12 versus 8.9 +/- 5 mmol/l, p = 0.002) and a higher BNP (median = 1290 pg/ml, interquartile range: 721, 3026 pg/ml versus 430 pg/ml, interquartile range: 243, 886 pg/ml). On multivariate analysis, the only factors that remained significantly associated with mortality were the BNP levels (p = 0.005) and the systolic blood pressure (p = 0.01). The negative predictive value of a BNP level < 631 pg/ml (median) for death was 94% (95% confidence interval: 91 to 97%). We conclude that BNP does have an independent prognostic value for in-hospital death in elderly subjects with acute diastolic cardiac failure.