Urgent treatment of serious infections with broad spectrum antibiotics usually starts before bacteriological evidence of the infective organism(s) becomes available. In this study 471 patients with a clinical diagnosis of sepsis were treated empirically with ceftazidime (CAZ) monotherapy (249 patients), or with an aminoglycoside+(ureido)penicillin combination (AG+PEN) (222 patients) to establish clinical outcome and bacteriological response. Up to 72 h post-treatment 94.5% of patients in the CAZ group and 93.8% in the AG+PEN group were treated successfully (treatment difference 0.7%, P < 0.01, 95% confidence interval -3.8%, 5.2%); 2-4 weeks after treatment neither regimen proved clinically superior. The differences in bacteriological response up to 72 h, and at 2-4 weeks after treatment, were 5.6% and 12.4% in favour of CAZ, however, these were not statistically significant. Overall, 56 patients reported 72 adverse events in the CAZ group, compared with 33 patients reporting 41 adverse events in the AG+PEN group. Deaths, 40 on CAZ and 21 on AG+PEN, were mainly related to their underlying condition. The two regimens were shown to be clinically equivalent in seriously ill patients treated empirically.