The objective of this study was to compare the evolution of patients with acute renal failure (ARF) treated conservatively or with different dialytic techniques in an intensive care unit (ICU). From June 1992 to November 1994, 1087 consecutive patients were admitted in our ICU. Two hundred and twenty of these presented with ARF, and were divided into three groups; group I (control group): 156 patients with ARF who did not receive substitutive techniques; group II: 21 patients under intermittent hemodialysis (IHD) or peritoneal dialysis (PD); group III: 43 patients under continuous hemodiafiltration (CHDF). The studied variables were age, etiology of renal failure, requirement of dialysis, type of dialysis, length of ICU and hospital stay, and renal function outcome. APACHE II and SAPS scores were recorded on admission and analyzed for hospital mortality. Chi-square test and the analysis of variance were used for the statistical analysis. Results are presented as mean +/- SD. A p value below 0.05 was considered statistically significant. Although etiology of ARF was multifactorial, we found a high frequency of ARF due to sepsis (56.8%), hypoperfusion (58.7%), and acute tubular necrosis (62.5%). Sepsis and heart failure were clinical conditions associated to a greater mortality. We did not find any statistical difference between the two dialyzed groups for all the studied variables, nor between the three groups regarding APACHE II and hospital stay. Significant differences were found between dialyzed and non-dialyzed patients respect to age, group I: 64.1 +/- 13.6, group II: 56.4 +/- 19.7, and group III: 56.0 +/- 14.1 (p < 0.001), creatinine peak serum levels, group I: 260 +/- 130, group II: 494 +/- 209, and group III: 441 +/- 170 mumol/L (p < 0.0001), and mortality, group I: 46.9%, group II: 66.7%, and group III: 76.2% (p < 0.002). SAPS score showed differences between the control group and the CHDF group 13.9 +/- 4.8 and 16.4 +/- 5.4 (p < 0.007), respectively. The use of dialytic techniques in critically ill ARF patients is associated with greater mortality. Prognostic indexes on admission did not correctly classify our patients with ARF. Continuous hemodiafiltration does not involve greater mortality or length of stay as compared to conventional dialysis.