OBJECTIVE To evaluate plasma population pharmacokinetics and penetration into cerebrospinal fluid (CSF) by indinavir (IDV) in HIV-infected individuals receiving IDV, zidovudine and lamivudine. METHODS Plasma population pharmacokinetic analysis was performed on 805 IDV plasma values from 171 patients, using a non-linear mixed-effects modeling approach. CSF data from 19 patients were analyzed using an individual approach. RESULTS Mean individual Bayesian estimates for oral clearance (CL) and volume of distribution (V) by the final model that incorporated interoccasion variability were 0.75 l/h per kg [coefficient of variation (CV) 54.8%] and 1.74 l/kg (CV 82.7%), respectively. Mean model-predicted plasma IDV level at 8 h, maximal level, area under the plasma level-time curve up to 8 h and plasma half-life were 0.42 micromol/l (CV 57.5%), 9.51 micromol/l (CV 47.3%), 29.56 micromol/l x h (CV 46.9%) and 1.50 h (CV 20.9%), respectively. The mean IDV CSF level was 0.11 micromol/l (CV 49.7%) and the mean CSF:plasma concentration ratio was 0.017. CONCLUSIONS Population estimates of pharmacokinetic parameters of IDV and its CSF penetration were in excellent agreement with previously reported data from individual analyses. Intraindividual interoccasion variability of IDV pharmacokinetics was estimated to be of similar order of magnitude to its interindividual variability, which may affect response to long-term antiretroviral therapy involving IDV. CSF levels of IDV exceeded its in vitro 95% inhibitory concentration of HIV replication. Given that CSF is virtually free of protein, viral suppression in the central nervous system should be achievable with an IDV-containing regimen.