We measured zidovudine concentrations in blood, muscle, and brain extracellular fluid (ECF) by microdialysis and in serum ultrafiltrate and cerebrospinal fluid (CSF) samples during a continuous intravenous infusion (15 mg/kg/h) and after bolus dosing (50-80 mg/kg over 15 min) in nonhuman primates to determine whether CSF drug penetration is a valid surrogate for blood-brain barrier penetration. Recovery was estimated in vivo by zero net flux for the continuous infusion and retrodialysis for the bolus dosing. In vivo recovery was tissue-dependent and was lower in brain than in blood or muscle. Mean (+/-S.D.) steady-state blood, muscle, and brain zidovudine concentrations by microdialysis were 112 +/- 63.8, 105 +/- 51.1, and 13.8 +/- 10.4 microM, respectively; and steady-state serum ultrafiltrate and CSF concentrations were 81.2 +/- 40.2 and 14.1 +/- 8.0 microM, respectively. Brain ECF penetration (microdialysis brain/blood ratio) and CSF penetration (standard sampling CSF/serum ratio) at steady state were 0.13 +/- 0.06 and 0.17 +/- 0.02, respectively. With bolus dosing the mean (+/-S.D.) zidovudine area under concentration-time curve (AUC) normalized to a dose of 80 mg/kg was 577 +/- 103 microM. h in blood, 528 +/- 202 microM. h in muscle, and 108 +/- 74 microM. h in brain (brain/blood ratio of 0.18 +/- 0.10) by microdialysis. Serum ultrafiltrate AUC was 446 +/- 72 microM. h and the CSF AUC was 123 +/- 4.7 microM. h (CSF/serum ratio of 0.28 +/- 0.06). In conclusion, recovery was tissue-dependent. CSF and brain ECF zidovudine concentrations were comparable at steady state, and the corresponding AUCs were comparable after bolus injection. Thus, zidovudine penetration in brain ECF and CSF in nonhuman primates is limited to a similar extent, presumably by active transport, as in other species.