The efficacies of cefepime and ceftazidime in an experimental Escherichia coli infection in granulocytopenic mice were related to their in vitro activites and their pharmacokinetic profiles. Cefepime had a higher intrinsic activity in vitro than ceftazidime, and it had a different pharmacokinetic profile, resulting in higher peak concentrations in plasma and a longer elimination half-life. To predict the antibacterial efficacy in vivo on the basis of in vitro activity and pharmacokinetics, we applied a mathematical model in which the in vitro effect is expressed as the difference in growth rate between control cultures and cultures grown in the presence of the antibiotic (E(R)), whereas the in vivo effect is given by the difference in the number of CFU between controls and antibiotic-treated animals (E(N)). The integral of E(R) over time, called E(R)t, was calculated by using in vivo concentrations. A significant linear relationship was found between E(N) and E(R)t for different doses at various times up to 4 h after administration, although the slope of this relationship was slightly but significantly less for cefepime (0.44) than for ceftazidime (0.59).