The host response to microbial infection is associated with the release of inflammatory mediators. We hypothesized that the type and degree of the systemic response as reflected by levels of circulating mediators predict morbidity and mortality, according to the invasiveness of microbial infection. We prospectively studied 133 medical patients with fever and culture-proven microbial infection. For 3 days after inclusion, the circulating levels of activated complement C3a, interleukin (IL)-6, and secretory phospholipase A(2) (sPLA(2)) were determined daily. Based on results of microbiological studies performed for up to 7 days, patients were classified as having local infections (Group 1, n = 80 positive local cultures or specific stains for fungal or tuberculous infections) or bacteremia (Group 2, n = 52 plus 1 patient with malaria parasitemia). Outcome was assessed as the development of septic shock and as mortality up to 28 days after inclusion. Fifteen patients (11%) developed septic shock and overall mortality was 18% (n = 24). Bacteremia was associated with shock and shock predisposed to death. Circulating mediator levels were generally higher in Group 2 than in Group 1. Circulating levels of IL-6 and sPLA(2) were higher in patients developing septic shock and in nonsurvivors, particularly in Group 1. High C3a was particularly associated with nonsurvival in Group 2. In Group 1, the area under the curve (AUC) of the receiver operating characteristic (ROC) curve for the peak sPLA(2) for shock development was 0.79 (P < 0.05). The AUC of the ROC curve of the peak IL-6 and sPLA(2) for mortality was 0.69 and 0.68 (P < 0.05), respectively. In Group 2, the AUC of the ROC for peak C3a predicting mortality was 0.73 (P < 0.05). In conclusion, in medical patients with fever and microbial infection, the systemic inflammatory host response predicts shock and death, at an early stage, dependent on the invasiveness of microbial infection. The results suggest a differential pathogenetic role of complement activation on the one hand and release of cytokine and lipid mediators on the other in bacteremic and local microbial infections, respectively. They may partly explain the failure of strategies blocking proinflammatory cytokines or sPLA(2) in human sepsis and may extend the basis for attempts to inhibit complement activation at an early stage in patients at risk of dying from invasive microbial infections.