Local anaesthetics (LA's), administrated via intrathecal or peridural routes, are used for regional anaesthesia and for regional control of major pain. Microdialysis should be a useful method to study the disposition of LA's in cerebrospinal fluid (CSF) because of the difficulty in sampling CSF. One of the main problems with microdialysis analysis is the necessity of a reliable in vivo calibration method which allows accurate in vivo concentrations to be established. In the current, retrodialysis was applied to calibrate the microdialysis probes. Several flow rates were tested in vitro to select an optimal in vivo flow rate (1μl/min). In vitro and in vivo relative recovery (RR) and relative loss (RL) were determined in a series of LA's. The influence of the concentration of LA's on RR and RL was studied. A significant variability in RR and RL among different probes within a batch was established (RR of bupivacaine ranging from 49 to 65 %). The ratio between RL of internal standard and RL of solute of interest, used to correct the dialysate concentrations, decreased in vivo but was constant. Following in vivo implantation in rabbit CSF during 4 hours, the probes were tested again in vitro and did not show any variation in RR. After epidural administration, the concentrations of bupivacaine in CSF were highly variable (CSF mean peak concentrations: 350μg/ml; ranging from 97 to 630 μg/ml). The concentrations were much higher than plasma mean peak concentrations (1.4 μg/ml; ranging from 0.7 to 2.5 μg/ml). Microdialysis should be an interesting tool to better understand the biological fate of local anaesthetics administrated via spinal routes.