Sepsis--the most common cause of death in hospitalized patients--affects over 18 million people worldwide and has an expected 1% increase of incidence per year. Recent clinical trials indicate that therapeutic approaches effective in diseases with similar pathogenesis have a modest effect against sepsis. Although the reason for this failure remains controversial, recent studies provide new insights and promising experimental strategies. We propose that the current definition of sepsis is too broad and encompasses heterogeneous groups of patients suffering similar, but different, syndromes that are historically grouped under the general diagnosis of sepsis. Future clinical trials might define patient populations and therapeutic strategies according to the profile of expression of cytokines.