In the course of a multicentre clinical trial evaluating two antibacterial regimens for the empiric treatment of suspected infection in febrile neutropenic cancer patients, a suboptimal response was noted among recipients of antibacterial prophylaxis with trimethoprim/sulphamethoxazole (TMP/SMX). Multivariate analysis identified TMP/SMX prophylaxis as a predictor of poor outcome independent of other variables such as classification of infection, marrow recovery, neutrophil count at first fever, indwelling central venous catheter use, and underlying disease. This effect appeared to be restricted to recipients of tobramycin plus ticarcillin (TT). TMP/SMX suppresses potentially pathogenic aerobic Gram-negative bacilli and allows colonization and subsequent infection by Gram-positive microorganisms against which TT-like regimens have limited activity. Recognition of this phenomenon may permit a more appropriate selection of antibacterial agents for the therapy of suspected infection in the neutropenic patient.