Many cancer patients become granulocytopenic as a result of therapy and, as such, are likely to have fever during neutropenic episodes. Approximately 20 percent of these episodes have an associated gram-negative rod bacteremia; these infections occur among the most profoundly granulocytopenic patients and are associated with the highest mortality. Most infections are caused by one of three organisms: Escherichia coli, Pseudomonas aeruginosa, or Klebsiella pneumoniae. The standard approach to therapy has been the empiric utilization of an antibiotic combination, most often an aminoglycoside with either an anti-Pseudomonas penicillin or a cephalosporin. In patients for whom concern about aminoglycoside-associated nephro- or ototoxicity is high, a double beta-lactam combination has been considered. Also, with the introduction of increasingly active, exceptionally broad-spectrum antimicrobials, empiric therapy with single agents has been considered. Beta-lactam/aminoglycoside combinations more often than not are synergistic, although antagonism can be detected on occasion. Some double beta-lactam combinations demonstrate antagonism, whereas in other cases, synergism, or at least partial synergism, can be observed. Antibiotic combinations can be evaluated through in vitro models, such as the capillary model system, or through animal models designed to mimic the neutropenic state with gram-negative bacteremia, to determine potential agents or combinations of agents for this patient population. These preclinical approaches have suggested that some agents may prove effective as monotherapy and, indeed, have been comparable in activity to some of the standard antibiotic combinations. However, clinical trials have had insufficient numbers of particularly high-risk patients with profound, persistent granulocytopenia and gram-negative rod bacteremia to be able to assess their usefulness in such patients. In general, it still appears to be advantageous to use combinations such as those used in the most recent European Organization for Research on Treatment of Cancer antimicrobial trial, which compared azlocillin/amikacin with ceftazidime/amikacin. In order to reduce aminoglycoside toxicity, patients were randomly assigned to receive amikacin either for a short period or for the entire length of therapy. The study should help to determine whether it is possible to maintain the advantages of two-drug combinations while reducing the disadvantages of prolonged aminoglycoside therapy.