Bioelectric impedance analysis (BIA) was carried out in 55 cirrhotic patients (32 with and 23 without ascites) and 20 healthy controls with the aim of evaluating its usefulness in cirrhosis. Furthermore, in 13 of the ascitic patients BIA was performed immediately before and 24 hours after paracentesis. No differences were observed among the three groups in either resistance or the values of total body water (TBW), lean body mass (LBM) and fatty body mass (FBM) estimated by BIA. On the other hand, reactance and the phase angle were lower in cirrhotic patients. Likewise, the percentage of body cell mass (BCM) was lower in ascitic patients than in the non ascitic patients and the controls (37.7 +/- 1.2 vs 42.3 +/- 1.2 vs 45.5 +/- 1.6%; p = 0.0003). Mid-arm muscle circumference was correlated with TBW and LBM in all the groups and with only BCM in the controls and those without ascites. In general, the precision of the regression lines was lower in the ascitic patients than in the non ascitic patients and controls. TBW decreased following paracentesis (28.3 +/- 1.7 vs 30.3 +/- 2.0 kg; p = 0.008) but this decrease did not correspond with the volume of ascitic fluid obtained in any case. The mean BCM did not change with paracentesis although individual differences ranged between -8.0 and +11.2% (CI 95%). It is concluded that BIA is not a reliable method for nutrition evaluation in cirrhotic patients with ascites in clinical practice and that the anthropometric parameters are still preferable in these cases.