The prognostic significance of major organ and tissue infection was examined in 909 episodes of bacteremia that were selected from 10 consecutive, randomized clinical trials of antibiotic therapy for infection in patients with cancer and neutropenia. Extensive tissue infection significantly compromised response to initial therapy (38% vs. 74%; P < .0001), ultimate outcome of infection (73% vs. 94%; P < .0001), median time to normalization of temperature (5.3 days vs. 2.5 days; P < .0001), and survival (P < .0001). Other poor prognostic factors revealed by logistic regression included shock (P < .0001) and bacteremia caused by Pseudomonas species (P = .03), Clostridium species (P = .006), or a pathogen resistant to antibiotics used for initial therapy (P < .0001). Recovery of the granulocyte count predicted a superior response (P < .0001). Although the overall mortality rate was not significantly increased when patients with bacteremia due to gram-negative organisms initially received monotherapy or when patients with bacteremia due to gram-positive organisms received delayed vancomycin therapy, these strategies increased the duration of therapy by 25%. Patients with bacteremia due to alpha-hemolytic streptococcus died more often when vancomycin was not included in the initial empirical regimen (P = .004). Because of the prognostic significance of extensive tissue or major organ infection, this factor should be considered in decisions concerning modification of therapy and use of colony-stimulating factors. The cost-effectiveness of initial monotherapy and delayed vancomycin therapy remains to be demonstrated.