Early Individualization of Tricyclic Antidepressant Dosing Using a Bayesian Pharmacokinetic Model

Abstract

Existing methods to prospectively dose tricyclic antidepressants (TCAs) require either specific test doses, precisely timed serum sampling, or both. We prospectively tested a new pharmacokinetic model that allows flexible dosing and sampling to determine maintenance requirements in patients receiving TCAs. Thirty-four patients entered the study. Drug concentrations were measured on the third day after starting TCA therapy. These values were analyzed using a Bayesian pharmacokinetic model to determine drug clearance and volume of distribution. This information was then used to predict the serum concentration resulting from a maintenance dose chosen by the psychiatrist. In phase I (n=17), patients received imipramine without specific starting doses. Phase II (n=17) was performed to provide a preliminary evaluation of the method in the usual clinical environment. In this phase, patients received either amitriptyline, imipramine, desipramine, doxepin (75 mg on day 1, 100 mg on day 2), or nortriptyline (50 mg on day 1,75 mg on day 2). Lower doses were allowed if clinically indicated. The predictability of future serum concentrations was then compared between the two phases. The mean prediction errors (model bias) in phases I and II were −15.5 ± 27.3 and −12.3 ± 21.8 ng/mL and were not different (p>0.05). The absolute prediction errors (model precision) were 18.5 ± 25.1 and 18.8 ± 16.0 ng/mL and were not different (p>0.05). Two slow metabolizers were identified (clearance <0.10 L/kg/h). This new method allows the determination of maintenance dose requirements early in therapy without standard test doses or specifically timed serum sampling. An additional benefit is early identification of slow metabolizers, preventing potential drug toxicity.