Although composite data from separate subjects can be used to generate single-subject estimates, intersubject variation precludes rigorous ocular pharmacokinetic analysis. Therefore, a rabbit model in which sequential aqueous and vitreous humor samples were obtained following the administration of the quinolone fleroxacin was developed. Mean data from individual animals were used for pharmacokinetic analysis. Following direct intravitreal or systemic drug administration, sequential paracenteses did not alter pharmacokinetic constants or ocular penetration and were not associated with an increase in ocular protein; contamination of vitreous humor with blood was minimal (less than 0.1%). Following direct injection or intravenous administration, vitreous humor concentration-time data were best described by one- and two-compartment models, respectively. The maximum concentration and the penetration into the aqueous and vitreous humors were 1.54 and 0.5 micrograms/ml and 27 and 10%, respectively. Elimination rates from aqueous and vitreous humors and serum were similar following parenteral drug administration. Drug elimination following direct injection was rapid, and the elimination rate from the vitreous humor was not prolonged by the coadministration of probenecid. Our animal model provides a new approach to the rigorous examination of the ocular pharmacokinetics of quinolone antimicrobial agents in the eye.