The aim of the present work was to assess the forecasting efficiency of methods of dosage individualization for carbamazepine according to mean population pharmacokinetic parameters (method 1) and from information relating to one (method 2) or more (method 3) measured steady-state serum drug levels in 344 epileptic patients. All the individuals studied were outpatients on a multiple dosage regimen with carbamazepine. Carbamazepine serum levels were analyzed by an enzyme multiplied immunoassay technique (EMIT). The accuracy and precision of methods 1, 2 and 3 were judged by the mean prediction error (m.e.) and the root mean squared error (m.s.e.) that had values of -1.1 +/- 3.5 and 2.8 +/- 2.3; 0.3 +/- 2.0 and 1.4 +/- 1.5 and -0.15 +/- 1.9 and 1.4 +/- 1.2 for methods 1, 2 and 3, respectively. From the statistical analysis of the experimental data and those predicted by the three methods studied, it may be inferred that the availability of data referring to the drug serum levels contributes decisively to the establishment of a correct dosage regimen.