A mouse model of bacteremia was used to compare the efficacies of 1.5- and 3.0-g intravenous doses of ampicillin-sulbactam. Seven strains of Escherichia coli producing various levels of TEM-1 beta-lactamase were used as the challenge isolates. These strains included six clinical isolates (MICs from 2/1 micrograms/ml [with 2 and 1 microgram/ml being the respective concentrations of ampicillin and sulbactam] to 32/16 micrograms/ml) with similar degrees of virulence in mice and a laboratory genetic transformant (E. coli AFE) which hyperproduces TEM-1 (MIC = 128/64 micrograms/ml). Human pharmacokinetics were simulated by injecting mice subcutaneously twice (1 h apart) with ampicillin-sulbactam at concentrations of 40 mg/kg of body weight (1.5 g) and 80 mg/kg (3.0 g). Against two clinical isolates for which ampicillin-sulbactam MICs were < or = 8/4 micrograms/ml, no difference was observed in either the rate or level of killing between the two doses, and both doses were 100% protective against lethal infection. Against the four clinical isolates for which ampicillin-sulbactam MICs were between 16/8 and 32/16 micrograms/ml, a slight delay in killing was noted with three of the strains. This delay was followed by a rapid 2- to 3-log drop in the level of bacteremia, and both doses of ampicillin-sulbactam were 100% protective against lethal septicemia. With strain AFE, no killing was observed with the 40-mg/kg dose compared with a 2-log killing with the 80-mg/kg dose. This difference in killing correlated with a decreased protective efficacy of the 40-mg/kg dose. These data suggest that the 1.5-g preparation of ampicillin-sulbactam is as effective as the 3.0-g dose in the treatment of experimentally induced E. coli bacteremia, as long as ampicillin-sulbactam MICs are 32/16 micrograms/ml or less.