Prediction of imipramine serum levels in enuretic children by a Bayesian method: comparison with two other conventional dosing methods.

The aim of the present study was to characterize the kinetic behavior of imipramine (IMI) and desipramine in enuretic children and to evaluate the performance of different methods for dosage prediction based on individual and/or population data. The study was carried out in 135 enuretic children (93 boys) ranging in age between 5 and 13 years undergoing treatment with IMI in variable single doses (25-75 mg/day) administered at night. Sampling time was one-half the dosage interval at steady state. The number of data available for each patient varied (1-4) and was essentially limited by clinical criteria. Pharmacokinetic calculations were performed using a simple proportional relationship (method 1) and a multiple nonlinear regression program (MULTI 2 BAYES) with two different options: using the ordinary least-squares method (method 2) and the least-squares method based on the Bayesian algorithm (method 3). The results obtained point to a coefficient of variation for the level/dose ratio of the drug (58%) that is significantly lower than that of the metabolite (101.4%). The forecasting capacity of method 1 is deficient both regarding accuracy [mean prediction error (MPE) = -5.48 +/- 69.15] and precision (root mean squared error = 46.42 +/- 51.39). The standard deviation of the MPE (69) makes the method unacceptable from the clinical point of view. The more information that is available concerning the serum levels, the greater are the accuracy and precision of methods (2 and 3). With the Bayesian method, less information on drug serum levels is needed to achieve clinically acceptable predictions.