Development and evaluation of a pharmacokinetic method to rapidly determine desipramine dosing requirements

A Bayesian pharmacokinetic technique was developed to estimate individual patient pharmacokinetic parameters for desipramine. The performance of the method was evaluated under two conditions; in actual patients treated for depression and in a series of simulated patients with variable clearance (CL) and volume of distribution (Vd). In actual patients, nonsteady-state serum concentrations (Cp) were obtained in the first few days of desipramine therapy. These Cp were then evaluated using the Bayesian method to estimate patient-specific CL and Vd. A second Cp was obtained during maintenance therapy. An estimate of the Cp during maintenance was made using the model-derived values of CL and Vd, and this value was compared to the measured Cp. In computer-simulated patients, a desipramine dose of 50 mg per day was used. A value for Cp after two doses was calculated using predetermined values of CL and Vd, and a first-order oral absorption model. This initial Cp was then evaluated by the Bayesian method to estimate CL and Vd. These parameters were used to estimate Cp at steady-state if the 50 mg per day dose was continued. An exact Cp at steady-state was calculated using the predetermined CL and Vd, and this value was compared to that expected using model-derived parameters. In both phases of the study the method was evaluated using prediction error analysis. In actual patients, prediction error was -12.5 ± 31.9 ng/ml, and the absolute prediction error was 22.7 ± 25.4 ng/ml. Under simulated conditions, low CL was associated with higher prediction error, but Vd did not significantly increase error. We conclude that the Bayesian technique performs well under clinical conditions. Under simulated conditions, low CL increased prediction error, but this problem was not encountered in the clinical setting.