OBJECTIVES: To evaluate, in daily practice, the benefits of drotrecogin alfa (DA) in the treatment of severe septic patients with multiple organ failure and optimum intensive care support. METHODS: In this prospective, observational pre-post study, the clinicians were free to include any patient meeting DA's inclusion criteria before and after DA's marketing. An optimal propensity score matching technique was used to reduce recruitment bias. Survival was modeled using a Cox proportional hazards model with a shared frailty term to account for the clustering of patients within the intensive care units. The number of bleeding events measured DA's safety. RESULTS: Respectively 509 and 587 patients were included in the before and after groups. There is strong evidence of recruitment bias: patients in the after group are younger, more frequently ventilated, have less comorbidities but more organ failures. After propensity score matching, 840 patients were retained in the analysis, with a better balance between the groups. The use of a frailty model improves significantly the variance explained by the survival model, showing a non-negligible cluster effect. When considering the whole sample of patients, without adjustments, survival is improved in the after (i.e. with DA) group (p=2.5%), with a hazard ratio (HR) of 0.805. In the matched sample, there are no significant survival differences (HR=0.900, p=35.0%). However, after stratifying by the LODS severity score quartiles, significance is reached (HR=0.795, p=4.8%). In the matched sample, a negative binomial model best described bleeding events. In this model, patients in the after group have a higher mean of bleeding events (p=2.0%). CONCLUSION: This observational study confirms DA's clinical trial results in the real practice setting. However, the use of the propensity score cannot replace randomization to assure perfect balance for all patient characteristics, measured and unmeasured. The results should therefore be considered with caution.