The purpose of the study was to determine if binding of the drugs to the sampling equipment during microdialysis would influence the results for carbamazepine, phenytoin and phenobarbital. In vitro experiments with microdialysis catheters and separate parts of catheters were performed to estimate the degree of drug binding to the dialysis equipment. A mathematical model to calculate drug binding and recovery is proposed. In vivo protein unbound carbamazepine concentrations in subcutaneous extracellular fluid at different flow rates (6 patients), unbound carbamazepine (1 patient) and unbound phenobarbital (I patient) in subdural cerebrospinal fluid and subcutaneous extracellular fluid were estimated and the in vivo data were compared to the in vitro results and data generated by the mathematical model. Binding to the soft outlet polyurethane tubing was extensive and variable for phenytoin, which precluded in vivo testing, but limited and more predictable for carbamazepine and phenobarbital. None of the three compounds bound to the hard internaltubing. Phenytoin and phenobarbital did not bind to the dialysis membrane, while a small degree of binding may be present for carbamazepine. In vivo estimates of carbamazepine protein unbound subcutaneous extracellular concentrations by microdialysis, adjusted for binding to the plastic tubing, were 81% of protein unbound plasma concentrations. In single case studies, subdural cerebrospinal fluid and subcutaneous extracellular levels of carbamazepine and phenobarbital were similar and when corrected for binding to the plastic tubings they were also close to protein unbound plasma concentrations. Microdialysis can be used for reliable estimations of protein unbound carbamazepine and possibly phenobarbital concentrations when drug binding to the plastic tubing is considered. Reliable estimation of unbound phenytoin is not possible at present.