Antimicrobial efficacy in orthopedic device infections is diminished because of bacterial biofilms which express tolerance to antibiotics. Recently, the use of high doses of levofloxacin with rifampin has been recommended for staphylococcal infections. In the present study, we evaluated the efficacy of levofloxacin at doses of 50 mg/kg/day and 100 mg/kg/day (mimicking the usual and high human doses of 500 mg/day and 750 to 1,000 mg/day, respectively) and compared it to that of to linezolid, cloxacillin, vancomycin, and rifampin in a rat tissue cage model of experimental foreign-body infection by Staphylococcus aureus. The antimicrobial efficacy in vitro (by MIC, minimum bactericidal concentration, and kill curves) for logarithmic- and stationary-phase bacteria was compared with the in vivo efficacy. In vitro bactericidal activity at clinically relevant concentrations was reached by all drugs except rifampin and linezolid in the log-phase studies but only by levofloxacin in the stationary-phase studies. The bacterial count decreases from in vivo tissue cage fluids (means) for levofloxacin at 50 and 100 mg/kg/day, rifampin, cloxacillin, vancomycin, linezolid, and controls, respectively, were: -1.24, -2.26, -2.1, -1.56, -1.47, -1.15, and 0.33 (all groups versus controls, P < 0.05). Levofloxacin at 100 mg/kg/day (area under the concentration-time curve/MIC ratio, 234) was the most active therapy (P = 0.03 versus linezolid). Overall, in vivo efficacy was better predicted by stationary-phase studies, in which it reached a high correlation coefficient even if the rifampin group was excluded (r = 0.96; P < 0.05). Our results, including in vitro studies with nongrowing bacteria, pharmacodynamic parameters, and antimicrobial efficacy in experimental infection, provide good evidence to support the use of levofloxacin at high doses (750 to 1,000 mg/day), as recently recommended for treating patients with orthopedic prosthesis infections.