Objective: Little information is available regarding current practice in continuous renal replacement therapy (CRRT) for the treatment of acute renal failure (ARF) and the possible clinical effect of practice variation. Design: Prospective observational study. Setting: A total of 54 intensive care units (ICUs) in 23 countries. Patients and participants: A cohort of 1006 ICU patients treated with CRRT for ARF. Interventions: Collection of demographic, clinical and outcome data. Measurements and results: All patients except one were treated with venovenous circuits, most commonly as venovenous hemofiltration (52.8%). Approximately one-third received CRRT without anticoagulation (33.1%). Among patients who received anticoagulation, unfractionated heparin (UFH) was the most common choice (42.9%), followed by sodium citrate (9.9%), nafamostat mesilate (6.1%), and low-molecular-weight heparin (LMWH; 4.4%). Hypotension related to CRRT occurred in 19% of patients and arrhythmias in 4.3%. Bleeding complications occurred in 3.3% of patients. Treatment with LMWH was associated with a higher incidence of bleeding complications (11.4%) compared to UFH (2.3%, p = 0.0083) and citrate (2.0%, p = 0.029). The median dose of CRRT was 20.4 ml/kg/h. Only 11.7% of patients received a dose of > 35 ml/kg/h. Most (85.5%) survivors recovered to dialysis independence at hospital discharge. Hospital mortality was 63.8%. Multivariable analysis showed that no CRRT-related variables (mode, filter material, drug for anticoagulation, and prescribed dose) predicted hospital mortality. Conclusions: This study supports the notion that, worldwide, CRRT practice is quite variable and not aligned with best evidence. © 2007 Springer-Verlag.