Background Although the performance of target-controlled infusion (TCI) have been studied extensively, the accuracy and safety of a TCI system that targets the effect site remains to be demonstrated. This study was to investigate the relations of TCl of propofol to its concentrations in cerebral spinal fluid (CSF), the effect-site concentrations and bispectral index (BIS). Methods Twelve mongrel dogs were used for investigations. The target effect-site concentration was set at 3 μg/ml and the infusion was lasted for 15 minutes. CSF and blood samples were then collected and propofol concentrations were determined by using high performance liquid chromatography with fluorescence detection. BIS and hemodynamic data were monitored continuously. Results The predicted plasma concentrations were generally overestimated. Median performance error (MDPE) and absolute median performance error (MDAPE) were -10.0% and 29.9% respectively. Propofol CSF concentrations were much lower than its effect-site concentrations. Changes in BIS were consistent with propofol concentrations in CSF, both of which changed direction at 5 minutes while the effect-site concentrations relatively lagged behind. Better correlation (r2=0.9195) was found between BIS and CSF concentrations, when compared with that between BIS and effect-site concentrations (r2=O.554). Conclusion With 1% enflurane inhaled, the inconsistency of drug effect to the effect-site concentrations may result from inaccuracy of pharmacokinetic parameters. CSF may show effect-site concentrations more accurately than plasma when using target effect-site concentration infusion.Concentrations of propofol in cerebral spinal fluid: target-controlled infusion