The results of the four EORTC trials conducted over the past 15 years suggest: (1) early empiric therapy with broad spectrum antibiotics directed against gram-negative bacterial bacteremia (GNBB) is a reasonable approach in febrile granulocytopenic patients (GCP); (2) the level and dynamics of the granulocyte count are extremely important in determining the outcome of bacteremia; severely and/or persistently neutropenic patients are the true tests of antibiotic efficacy and they benefit from antimicrobial synergism; (3) mortality from GNBB in GCP is not related directly to a given empiric antimicrobial regimen which may 'buy time' and allow appropriate therapeutic alterations; (4) only microbiologically documented infections and especially bacteremias are useful to compare responses to antimicrobial regimens; (5) the response rate of GNBB is clearly influenced by the susceptibility of the causative pathogen to the beta-lactam component of the empiric regimen and emergence of resistance to some antibiotics (cephalothin, carbenicillin, ticarcillin, azlocillin) has rendered some combinations less effective. The combination of an anti-Pseudomonas beta-lactam plus an aminoglycoside is recommended as the 'standard' for empiric therapy in febrile GCP; (6) gram-positive pathogens have become a common cause of bacteremia in GCP and although the response rate to empiric regimens may be marginal, the associated mortality is low. A general conclusion from these trials is that studies of the management of infection in GCP should include sufficient numbers of eligible patients to allow for evaluation of bacteremic patients at highest risk of death. The need for large collaborative studies stems directly from these considerations.