In an observational study in 19 consecutive acutely hospitalized dialysis patients, ultrafiltration (UF) volume was determined by B‐type natriuretic peptide (BNP) levels. Patients were ultrafiltrated daily until they achieved a target BNP level <500 pg/mL. The UF volumes ranged from 2 to 5 L per session. All patients were male veterans aged 68±11 years (mean ± SD), 74% were diabetic, 47% were African Americans, 58% underwent prevalent dialysis, and 53% had an arteriovenous fistula. Left ventricular ejection fraction on 2‐dimensional echocardiography was 43.8%±27.9% (n=16). The admission BNP was 2412±1479 pg/mL (range, 561–5000 pg/mL) and BNP at hospital discharge was 1245±1173 pg/mL (range, 345–5000 pg/mL) (nonparametric Wilcoxon P=.0013). Admission weight was 88.9±27.9 kg and at discharge was 78.1±25.6 kg (P=.0002). The number of antihypertensive medications taken was 3.8±2.0 at admission and 2.3±1.7 at discharge (P=.0005). The number of patients with >2 blood pressure medications decreased from 14 to 6 (Fisher exact test, P=.02). The systolic/diastolic/mean arterial blood pressure decreased from admission to discharge (153.6±43.8/80.6±21.8/102.4±27.3 to 132.1±27.9/68.9±14.6/89.9±16.5 mm Hg; P=.0222/.0139/.0329, respectively). Although all patients were volume‐overloaded at admission according to BNP criteria (>500), only 42% were identified as having heart failure. BNP‐directed UF is safe because it minimizes symptomatic hypotension, identifies occult congestive heart failure in a large number of patients, and significantly reduces blood pressure in addition to reducing body weight and number of medications used.