Members of the fluoroquinolone class are being actively evaluated for inclusion in tuberculosis chemotherapy regimens, and we sought to determine the best in vitro and pharmacodynamic predictors of in vivo efficacy in mice. MICs for Mycobacterium tuberculosis H37Rv were 0.1 mg/liter (sparfloxacin [SPX]) and 0.5 mg/liter (moxifloxacin [MXF], ciprofloxacin [CIP], and ofloxacin [OFX]). The unbound fraction in the presence of murine serum was concentration dependent for MXF, OFX, SPX, and CIP. In vitro time-kill studies revealed a time-dependent effect, with the CFU reduction on day 7 similar for all four drugs. However, with a J774A.1 murine macrophage tuberculosis infection model, CIP was ineffective at up to 32x MIC. In addition, MXF, OFX, and SPX exhibited less activity than had been seen in the in vitro time-kill study. After demonstrating that the area under the concentration-time curve (AUC) and maximum concentration of drug in plasma were proportional to the dose in vivo, dose fractionation studies with total oral doses of 37.5 to 19,200 mg/kg of body weight (MXF), 225 to 115,200 mg/kg (OFX), 30 to 50,000 mg/kg (SPX), and 38 to 100,000 mg/kg (CIP) were performed with a murine aerosol infection model. MXF was the most efficacious agent (3.0+/-0.2 log10 CFU/lung reduction), followed by SPX (1.4+/-0.1) and OFX (1.5+/-0.1). CIP showed no effect. The ratio of the AUC to the MIC was the pharmacodynamic parameter that best described the in vivo efficacy. In summary, a lack of intracellular killing predicted the lack of in vivo activity of CIP. The in vivo rank order for maximal efficacy of the three active fluoroquinolones was not clearly predicted by the in vitro assays, however.