A MODELING APPROACH TO THE THERAPEUTIC EFFECT OF PAROXETINE FOR DEPRESSION WITH A FOCUS ON PHARMACOKINETICS- PHARMACODYNAMICS RELATIONSHIP

Abstract Background The mainstay of depression treatment is pharmacotherapy with antidepressants such as selective serotonin reuptake inhibitors (SSRIs), but large individual differences exist in their blood drug concentrations and efficacy, making it difficult to predict their therapeutic response. Aims & Objectives In this study, we attempted to construct a model focusing on the pharmacokinetic (PK)-pharmacodynamic (PD) relationship to examine the relationship between each patient factor affecting the therapeutic effect of paroxetine (PAX). Methods Blood PAX concentrations were measured in 179 PAX-treated patients with depression (age: 41.5 ± 14.2 years) at weekly intervals from the start of treatment until the 6th week, and the PK parameters were estimated for each patient by partially modifying the results of our previous population PK analysis. Next, 86 patients (age: 46.3 ± 13.3 years) for whom data from the Montogomery-Asberg Depression Scale (MADRS), a severity of depression, were available among the subjects in the PK analysis and decision tree, a machine learning algorithm, and nonlinear mixed-effects (NLME) model were constructed to predict each patient's response to treatment. We then compared their prediction accuracy. The patient factors such as calculated PK parameters, PAX dosage, age, Cloninger's Temperament Characteristics (TCI), and genotype were used as covariates. Results In the decision tree model, patients could be classified by MADRS at the second week after the start of treatment, age, weight, daylight hours, and blood PAX concentration in the first week after treatment initiation, with or without remission at the 4th week after starting PAX treatment. On the other hand, the Emax model was employed in the NLME model. Since the blood PAX concentration at each time point was not incorporated into the model, we assessed the cumulative area under the blood concentration-time curve of the blood PAX concentration from the starting point of PAX treatment as the total PAX exposure, considering the delayed onset of PAX effect. The results showed that total PAX exposure up to the first week after the start of treatment and TCI self-directedness affected the maximum improvement rate, while MADRS after 1 week of PAX treatment affected the duration of treatment to obtain 50% of the maximum improvement rate. Both validity and predictivity for the NLME model were good, exceeding the predictive performance of the decision tree. Discussion & Conclusion The present study showed for the first time that the total exposure dose from the start of PAX administration is related to the presence or absence of therapeutic effects. The findings of this study indicates that calculating total exposure from blood drug concentrations may have represented the various effects that occur over time from immediately after PAX administration to the onset of effect. The results of this study suggest that predicting treatment responsiveness early after SSRIs administration may allow for optimal individualized dosing design and therapeutic drug selection, which should lead to optimized treatment and improved quality of life in depressed patients.