Background: Although plasma concentrations of brain natriuretic peptides (BNP) increase in hemodialysis (HD) patients as well as patients with cardiovascular diseases (CD), the clinical significance of BNP in HD patients has yet to be elucidated. In this study, we investigated the pathophysiological significance of BNP in HD patients. Methods: Plasma BNP concentrations were measured in 164 HD patients after HD and 14 healthy volunteers. In 12 patients without CD, BNP was also measured before HD. Multiple regression analysis was performed to determine the important factors causing increased plasma BNP concentrations. Cardiac mortality was monitored for 36 months after baseline analysis, and the prognostic role of BNP was examined by Cox proportional hazards regression analysis. Results: Plasma BNP concentrations of HD patients without CD decreased significantly during HD session (124.5 ± 90.7 vs. 91.4 ± 67.6 pg/ml, mean ± SD, p = 0.004), but were still significantly higher than those of the healthy subjects (9.7 ± 9.2 pg/ml, p = 0.0002). Plasma BNP concentrations of patients with CD were significantly higher than of those without CD (579.6 ± 564.3 vs. 204.0 ± 241.5 pg/ml, p < 0.0001). Plasma BNP concentrations were also significantly higher in diabetes mellitus (DM) patients than in non-DM patients (514.1 ± 585.4 vs. 296.0 ± 347.0 pg/ml, p = 0.0031). Multiple regression analysis showed that left ventricular mass index (LVMI), CD and DM were independent factors for the elevated BNP (R2 = 0.303, p < 0.0001). During a 36-month follow-up period, cardiac death occurred in 13 patients. Kaplan-Meier survival estimates of patients from varying plasma BNP quartiles (<200, 200–450, 450–700 and >700 pg/ml) differed between the four groups (p < 0.0001). The group with the highest BNP level (>700 pg/ml) had the lowest survival. When compared with patients with BNP <200, the hazard ratios for cardiac death of patients with BNP of 200–450, 450–700 and >700 pg/ml were 2.3 [95% confidence interval (CI) 0.14–36.7], 18.7 (1.9–183.4) and 51.9 (6.5–416.3), respectively. The univariate Cox proportional hazards model showed that BNP, left ventricular ejection fraction, LVMI, age, DM, serum albumin and C-reactive protein (CRP) were significantly associated with the risk of cardiac mortality. By stepwise multivariate Cox proportional hazards analysis, only BNP, LVMI and CRP remained powerful independent predictors of cardiac death. The relative risk ratios were 1.002 (95% CI 1.001–1.002) for BNP, 2.192 (1.532–3.135) for CRP and 1.027 (1.013–1.042) for LVMI. Conclusion: High plasma BNP concentrations in HD patients were associated with volume overload, left ventricular hypertrophy, CD and DM. Plasma BNP concentration may be a useful parameter for assessing the risk of cardiac death in HD patients by providing prognostic information independently of other variables previously reported.