In Germany a TCI-system for propofol (Disoprifusor-TCI) has been commercially available since spring 1997. We investigated the prediction error and precision of this TCI system as part of a multicentre study. Bias, precision, blood concentrations and dosage of propofol were compared with patients receiving propofol via a manually controlled infusion device.After approval by the local Ethics Committee and written informed consent, 21 patients of ASA-classification I to III scheduled for major abdominal surgery received either a target controlled infusion (group T, Disoprifusor-TCI) or a manually controlled infusion (group M) of propofol. The propofol plasma concentrations were measured by HPLC. The prediction error for each measurement, the median prediction error (MDPE) or bias, the median absolute prediction error (MDAPE) or precision and the divergence (change of the prediction error over infusion time) were calculated for both groups.For all patients in group T (n = 12) the bias of the TCI system was 6.7% and the precision 27.5%. For 70% of all measured plasma concentrations the absolute prediction error was < or = 37%. The divergence was -5.4% per hour. For all patients in group M (n = 9) the bias was 44.2% and the precision 50%. The mean amount of propofol infused per kilogram body weight and hour was significant higher in T (9.0 +/- 1.2 mg/kg/h) than in M (6.6 +/- 1.2 mg/kg/h, p < 0.005).With a precision of 27.5% the investigated TCI system (Diprifusor-TCI) showed an acceptable inaccuracy, as for TCI-systems a median prediction error of +/- 30% has to be expected due to the inherent variability of pharmacokinetic parameters. Further studies will be necessary to find out whether the investigated TCI system for propofol may offer substantial advantages.