Aminoglycoside bactericidal activity during the first 24 h of treatment probably is a determining parameter in the prognosis of severe gram-negative infections in immunocompromised patients. To identify the predictive factors involved in the definition of the best therapeutic regimen for Enterobacter cloacae and Serratia marcescens infections, we studied different gentamicin, tobramycin, and amikacin regimens by using an experimental model of rabbit endocarditis. Two factors appear to play an important role in predicting in vivo efficacy: (i) the level of in vivo bactericidal activity, which can differ widely from one aminoglycoside to another for the same bacterial strain and from one strain to another of the same species, and (ii) the critical serum drug concentration (CSC, in milligrams per liter), defined as the lowest serum antibiotic concentration capable of producing a significant CFU reduction (P less than 0.05) in endocarditis vegetations 24 h after the beginning of a continuous infusion. Stepwise regression analysis showed that for gentamicin and S. marcescens, the area under the concentration-time curve above the CSC and then the time above the CSC are the determining parameters for efficacy (R = 0.69; F = 13.5; P = 0.001), whereas for amikacin and S. marcescens, the time above the CSC and then the area under the concentration-time curve above the CSC predict efficacy (R = 0.74; F = 24.0; P = 0.0001). The lowest CSC is that of amikacin (about 8 mg/liter); those of gentamicin and tobramycin are about 15 mg/liter. In severe S. marcescens infections, intermittent amikacin dosing offers excellent bactericidal activity within the first 24 h.