An experimental Streptococcus pneumoniae pneumonia model in rabbits was used to assess the efficacy of amoxycillin, erythromycin and a new ketolide, telithromycin (HMR 3647). The MICs of amoxycillin, erythromycin and HMR 3647 for the three clinical S. pneumoniae strains used were, respectively, (mg/L): 0.01, 16 and 0.02 (strain 195); 2, 0.25 and 0.02 (strain 16089); 8, >64 and 0.02 (strain 11724). Antibiotic therapy reproduced human serum pharmacokinetics (amoxycillin 1 g iv tds or erythromycin 500 mg qds or HMR 3647 800 mg bd). Forty-eight hours of therapy with HMR 3647 and amoxycillin resulted in significant bacterial clearance in the lungs and spleen of rabbits infected by S. pneumoniae strain 195 and strain 16089 (at least 3 log(10) cfu/g decrease, P < 0.001). Erythromycin was active against only the erythromycinsusceptible strain (3 log(10) cfu/g decrease at 48 h, P < 0.001). None of the antibiotics showed significant efficacy with strain 11724. All agents produced significant bacterial clearance when time above MBC was >33%, and microbiological failure when it was <25%, whereas MIC was not correlated with microbiological outcome with HMR 3647. Our findings suggest that pharmacodynamic data integrating MBC may be predictive of microbiological success or failure with greater accuracy than with MIC. HMR 3647 produced significant bacterial clearance in both penicillin- and erythromycin-resistant pneumonia, but was less effective against the highly erythromycin-resistant S. pneumoniae strain.