An experimental rat pneumonia model using two amoxicillin-susceptible (MICs, < or =0.015 and 2 microg/ml) and two non-amoxicillin-susceptible (MIC, 4 microg/ml) Streptococcus pneumoniae strains was developed for testing the efficacy of amoxicillin administered to simulate human serum kinetics after treatment with amoxicillin-clavulanate (2,000 and 125 mg, respectively, twice a day, for 2.5 days). The end points for efficacy were reductions in bacterial loads in the lungs and reductions in levels of pulmonary damage. For the amoxicillin-susceptible strains (serotypes 23F and 14), a decrease greater than 4.5 log(10) CFU/pair of lungs was obtained, and the time for which the serum antibiotic concentration (SAC) was higher than the MIC (T(S)(A)(C)(>)(MIC)) was greater than 60% of the dosing interval. For non-amoxicillin-susceptible strains, the decrease in bacterial load was 1.34 to 1.75 log(10) CFU/pair of lungs, with a T(S)(A)(C)(>)(MIC) of 46.7% of the dosing interval. An in vitro study showed that serotype 9V non-amoxicillin-susceptible strains behaved as tolerant-like to concentrations similar to those in the in vivo model. The high and maintained SACs (T(S)(A)(C)(>)(MIC), >46% for all strains) significantly diminished lung injury (affected area of the lung and lung weight), compared to that in controls, by all strains, regardless of the MIC, bactericidal behavior in in vitro killing curves, or the serotype of the infecting strain. These results show the importance of host therapeutic end points in the evaluation of antibiotic efficacy. The antibiotic was more efficacious, for one nonsusceptible strain tested, when the treatment was started early (1 h postinoculation [p.i.]) than when treatment was delayed (24 h p.i.).