We investigated whether a relationship between risk of death and treatment effect could explain the disparate results between the preclinical and clinical sepsis trials of antiinflammatory agents over the last decade. A metaregression analysis of cited preclinical studies showed that the treatment effects of these agents were highly dependent on risk of death (p = 0.0001) and that animals were studied at significantly higher control mortality rates than humans (median [25th-75th quartile], 88% [79-96%] versus 39% [32-42%], p = 0.0001). An analysis of the clinical trials showed that antiinflammatory agents were also significantly more efficacious in septic patients with higher risk of death (p = 0.002) and were harmful in those with low risk. To test this relationship prospectively, we studied antiinflammatory agents in models employing differing doses of bacterial challenge to produce the full range of risk of death. We found that the efficacy of these agents, although very beneficial at high control mortality rates, was much reduced (p = 0.0001) and similar to those in human trials at moderate control mortality rates (i.e., 30 to 40%). The efficacy of antiinflammatory agents during sepsis is dependent on the risk of death, an observation that explains the apparent contradiction between preclinical and clinical trial results. Accounting for this relationship may be necessary for the safe and effective development of antiinflammatory therapies for sepsis.