In order to overcome the problems of interdosage fluctuations of body fluid concentrations of carbamazepine, a slow-release formulation has been developed. In an open, controlled, within-patient study, the diurnal plasma concentrations of carbamazepine and its 10,11-epoxide were measured in 25 epileptic children first treated with conventional carbamazepine tablets (Tegretol) and then with the Tegretol slow-release preparation. The diurnal plasma concentration curves during treatment with the slow-release formulation showed significantly less variation over 24 hours than during treatment with the ordinary preparation, as measured by the fluctuation index. Mean concentration values also differed significantly, which is explained by a somewhat reduced bioavailability (22% less) of the slow-release formulation. There were no differences in efficacy and tolerability between the two formulations, but there was a clear-cut reduction of reported side effects, especially tiredness, on treatment with the slow-release formulation. For that reason, the slow-release formulation should be a major advantage in treating children with epilepsy, in order to avoid interference with cognitive functions. In 12 children, simultaneous measurements of the concentration of carbamazepine and its epoxide in saliva were made and compared with the plasma values. As expected, the concentration curves corresponded, indicating that saliva sampling is an appropriate alternative for monitoring the concentration of carbamazepine. All children remained on the slow-release preparation after the trial and were followed up for 12 months or more.