Protein-calorie malnutrition is associated with poor prognosis in chronic liver disease, but reliable assessment is hampered by changes in body water. We prospectively evaluated the effect of fluid retention on bioelectrical impedance analysis (BIA) as a simple method for the estimation of body cell mass (BCM(BIA)) in 41 patients with cirrhosis (n = 20 with ascites; n = 21 without ascites) using total body potassium counting (BCM(TBP)) as a reference method. Arm muscle area and creatinine-derived lean body mass were compared with total body potassium data. In patients total body potassium was 24.4% lower than in controls and this loss was more severe in patients with ascites (-34.1%; P<.01). BCM(BIA) and BCM(TBP) were closely correlated in controls (r(2) =.87, P<.0001), patients without ascites (r(2) =.94, P <.0001) and patients with ascites (r(2) =.56, P<.0001). Removal of 6.2 +/- 3 L of ascites had only minor effects on BCM(BIA) (deviation of -0.18 kg/L ascites). Limits of agreement between both methods were wider in patients with ascites than in patients without (6.2 vs. 4.2 kg). In patients without ascites arm muscle area (r(2) =.64; P<.001) and lean body mass (r(2) =.55; P<.001) correlated significantly with total body potassium, but not in patients with ascites. For assessment of protein malnutrition in patients with cirrhosis, body cell mass determination by use of BIA offers a considerable advantage over other widely available but less accurate methods like anthropometry or the creatinine approach. Despite some limitations in patients with ascites, BIA is a reliable bedside tool for the determination of body cell mass in cirrhotic patients with and without ascites.