OBJECTIVE: Despite improved resuscitation and sepsis care, acute renal failure (ARF) remains common in critically ill surgical patients. New methods of renal replacement therapy (RRT) are being used in surgical intensive care units (SICUs), including high-flux hemodialysis (HD) and continuous RRT (CRRT). RRT is being used increasingly early in the course of ARF, but data are scant to suggest that mortality is improved. Consequently, we determined whether outcomes were improved with CRRT in SICU patients, and hypothesized that CRRT lowers mortality for patients with ARF. METHODS: Patients who developed ARF (acute increase in serum creatinine concentration >or=2.4 mg/dL) in the SICU from 1993 to 2004 were identified. Data collected prospectively included year of admission, age, gender, Acute Physiology and Chronic Health Evaluation (APACHE) III score, cumulative multiple organ dysfunction score and its individual components, cumulative nonrenal organ dysfunction score, and need for RRT. Patients were stratified January 1994 to January 2001 (pre-CRRT) and February 2001 to December 2004 (post-CRRT). The primary endpoint was mortality. RESULTS: Among 8,505 SICU patients, 530 (6.2%) developed ARF. Three hundred and eleven patients were treated pre-CRRT and 219 thereafter. Female patients comprised 35% of ARF patients. The mean age was 69 years +/- 2 years, and the mean APACHE III score was 81 +/- 1 point for ARF patients. HD was performed in 15.6% of ARF patients before 2001 and 5.5% of ARF patients in 2001 and thereafter. CRRT was performed in 20.1% of ARF patients in 2001 and thereafter. Overall mortality for ARF patients was 45% (APACHE III normative predicted mortality: 55%) with no difference over time (pre-CRRT = 46.3%, post-CRRT = 45.2%, p = 0.86). Patients who required RRT had a mean APACHE III score of 91 +/- 1 point, with 61% mortality (predicted mortality: 67%), with no difference over time. Independent predictors of mortality overall and for ARF patients included age and the magnitude of renal, cardiovascular, hepatic, and neurologic dysfunction. In comparison with CRRT, HD was associated with a decreased risk of death. CONCLUSIONS: Despite more frequent RRT and the use of CRRT, the mortality of ARF in critically ill surgical patients remains high because of nonrenal organ dysfunction. Considering that ARF-related mortality was decreased by intermittent HD, and that intermittent RRT is less costly, patients who need RRT should be treated preferentially with HD.