A Bayesian approach, in which the estimates of the population pharmacokinetic parameters of lidocaine reported by Vozeh et al. [1984 a and b] and one measurement of serum lidocaine concentration as feedback information were utilized, was prospectively evaluated in 20 Japanese patients. The concentration of lidocaine in serum sampled from each patient at 1.5 to 3 hours after the commencement of i.v. lidocaine therapy was measured and the most likely pharmacokinetic parameters of that patient were predicted by the Bayesian program for microcomputer. Having the predicted parameters, we got the concentrations of lidocaine at 12 and 24 hours (C12 and C24) after starting the therapy within the recommended therapeutic range (2-5 micrograms/ml) by adjusting the individual dosage. As a result, the values of C12 and C24 remained within the therapeutic range in all patients and in all except one, respectively. The mean % prediction error (measured minus predicted concentration standardized by measured one) and its 95% confidence interval were 1.8% and -11.4 to 14.7% for C12, and -2.3% and -11.9 to 7.4% for C24, showing that the prediction of both C12 and C24 was not biased. The correlation coefficients between the measured and predicted concentrations were 0.574 for C12 and 0.721 for C24 (p < 0.01). From these findings, the present Bayesian method implemented with the estimates of pharmacokinetic parameters in a Swiss patient population is useful for sufficiently accurate and precise assessment of individual dosage requirement within a few hours after starting lidocaine therapy.