To determine whether a systematic review of the literature could identify changes in the mortality of septic shock over time. A review of all relevant papers from 1958 to August 1997, identified through a MEDLINE search and from the bibliographies of articles identified. The search identified 131 studies (99 prospective and 32 retrospective) involving a total of 10,694 patients. The patients' mean age was 57 yrs with no change over time. The overall mortality rate in the 131 studies was 49.7%. There was an overall significant trend of decreased mortality over the period studied (r=.49, p < .05). The mortality rate in those patients with bacteremia as an entry criterion was greater than that rate in patients whose entry criterion was sepsis without definite bacteremia (52.1% vs. 49.1%; chi2=6.1 and p< .05). The site of infection altered noticeably over the years. Chest-related infections increased over time, with Gram-negative infections becoming proportionately less common. If all other organisms and mixed infections are included with the Gram-positives, the result is more dramatic, with these organisms being causative in just 10% of infections between 1958 and 1979 but in 31% of infections between 1980 and 1997. The present review showed a slight reduction in mortality from septic shock over the years, although this result should be approached with caution. The heterogeneity of the articles and absence of a severity score for most of the studies limited our analysis. Furthermore, there was an increasing prevalence of Gram-positive causative organisms, and a change of the predominant origin of sepsis from the abdomen to the chest.