Objectives: Critical illness increases the tendency to both coagulation and bleeding, complicating anticoagulation for continuous renal replacement therapy (CRRT). We analyzed strategies for anticoagulation in CRRT concerning implementation, efficacy and safety to provide evidence-based recommendations for clinical practice. Methods: We carried out a systematic review of the literature published before June 2005. Studies were rated at five levels to create recommendation grades from A to E, A being the highest. Grades are labeled with minus if the study design was limited by size or comparability of groups. Data extracted were those on implementation, efficacy (circuit survival), safety (bleeding) and monitoring of anticoagulation. Results: Due to the quality of the studies recommendation grades are low. If bleeding risk is not increased, unfractionated heparin (activated partial thromboplastin time, APTT, 1-1.4 times normal) or low molecular weight heparin (anti-Xa 0.25-0.35 IU/1) are recommended (grade E). If facilities are adequate, regional anticoagulation with citrate may be preferred (grade C). If bleeding risk is increased, anticoagulation with citrate is recommended (grade D-). CRRT without anticoagulation can be considered when coagulopathy is present (grade D-). If clotting tendency is increased predilution or the addition of prostaglandins to heparin may be helpful (grade C-). Conclusion: Anticoagulation for CRRT must be tailored to patient characteristics and local facilities. The implementation of regional anticoagulation with citrate is worthwhile to reduce bleeding risk. Future trials should be randomized and should have sufficient power and well defined endpoints to compensate for the complexity of critical illness-related pro- and anticoagulant forces. An international consensus to define clinical endpoints is advocated.