In a prospective randomized comparison, 217 episodes of fever (oral temperature > 38.5 degrees C on 1, or 38.0 degrees C on 2 occasions with a minimum interval of 4 h between recordings) during neutropenia (neutrophil count < 0.5 x 10(9)/I), patients were empirically treated with trimethoprim-sulfamethoxazole plus amikacin (TMP/SMZ plus AMI) or ceftazidime. Successful antibiotic treatment was defined as eradication of all signs, symptoms and microbiological evidence of infection on the primary therapy alone. The overall success rate did not differ between the 2 treatment groups: 31/102 (30%; 21-39%, 95% confidence interval, CI) for TMP/SMZ plus AMI and 41/115 (36%; 27-44%) for ceftazidime (difference 0.06 +/- 0.13, 95% CI). The corresponding numbers for documented infections were 12/50 (24%; 12-36%) and 14/60 (23%; 12-35%), respectively (difference 0.01 +/- 0.16). One patient in the TMP/SMZ plus AMI group and 2 patients in the ceftazidime group died from Gram-negative bacteraemias within 72 h. No other early deaths were observed. Antibiotics were changed due to adverse events in 2 episodes of each treatment group. In conclusion, this study demonstrates that TMP/SMZ plus AMI combination is comparable (i.e. a difference of < 20%) to ceftazidime monotherapy with regard to efficacy and safety in haematological patients with severe neutropenia. Both regimens require frequent modifications, particularly in bacteraemic fever episodes. However, in centres with a low frequency of isolation of Pseudomonas and especially of multi-resistent Pseudomonas strains, TMP/SMZ plus AMI offers an inexpensive alternative for the empirical treatment of febrile neutropenia.