Aminoglycosides are usually considered to be concentration-dependent antibiotics and to have similar pharmacodynamic and pharmacokinetic properties. To verify the equivalent activity of the aminoglycosides on a susceptible strain, we tested the killing rate of three aminoglycosides (gentamicin, tobramycin, and amikacin) on one strain of Serratia marcescens both in vitro and in vivo by using a rabbit model of left-ventricle endocarditis. Despite, similar MICs, the time-kill curve of gentamicin was consistently better than those of amikacin and tobramycin, whatever the concentration of each antibiotic used (1, 2, 4, 8, 16, or 32 mg/liter), after a 5-h incubation. The in vivo bacterial reduction in the vegetations was measured 24 h after administration of an intravenous 48-mg/kg bolus of each antibiotic or at the end of a 24-h continuous intravenous infusion of the same dose. Gentamicin was significantly more effective when administered as a bolus than when administered as a continuous infusion (2.8 +/- 0.2 versus 6.4 +/- 1.5 log10 CFU/g of vegetation, respectively; P less than 0.01), whereas amikacin was more effective as a continuous infusion than as a bolus injection (3.6 +/- 2.0 versus 7.5 +/- 1.3 log10 CFU/g of vegetation, respectively; P less than 0.01). Tobramycin was not very effective, whatever the dosage tested (approximately 6.5 to 7 log10 CFU/g). These results suggest that concentration-dependent bactericidal activities, both in vitro and in vivo, may vary greatly among aminoglycosides despite similar MICs.