Robust determination of the concentration-time profile of anti-infective agents in certain specialized compartments is often limited by the inability to obtain more than a single sample from such a site in any one subject. Vitreous humor and cerebrospinal fluid are obvious examples for which the determination of concentrations of anti-infective agents is limited. Advances in pharmacodynamics have pointed out the importance of understanding the profiles of drugs in the plasma and in specialized compartments in order to dose the drugs to obtain the best patient outcomes. Advances in population pharmacokinetic modeling hold the promise of allowing proper estimation of drug penetration into the vitreous (or other specialized compartment) with only a single vitreous sample, in conjunction with plasma sampling. We have developed a rabbit model which allows multiple samples of vitreous to be obtained without breaking down the blood-vitreous barrier. We have employed this model to test the hypothesis that robust estimates of vitreous penetration by the fluoroquinolone ciprofloxacin can be obtained from a traditional intensive plasma sampling set plus a single vitreous sample. We studied 33 rabbits which were receiving 40 mg of ciprofloxacin per kg of body weight intravenously as short infusions and from which multiple plasma and vitreous samples were obtained and assayed for ciprofloxacin content by high-performance liquid chromatography. Data were analyzed by the iterative two-stage population modeling technique (IT2S), employing the iterative two-stage program of Forrest et al. (Antimicrob. Agents Chemother. 37:1065-1072, 1993). Two data sets were analyzed: all plasma and vitreous samples versus all plasma samples and the initially obtained single vitreous sample. The pharmacokinetic parameter values identified were used to calculate the percent vitreous penetration as the ratio of the area under the concentration-time curve for the vitreous to that for the plasma. The values identified, 4% penetration for the full data set versus 3% penetration for the single vitreous sample data set, and their corresponding estimates were not statistically significantly different. We conclude that population modeling holds promise for the analysis of penetration of antimicrobiol agents into specialized spaces from which only single samples can be obtained, particularly for patients with whom robust plasma sampling can be performed.