Decision trees for when to change pharmacotherapy in late-life depression: integration of pharmacogenetics, venlafaxine pharmacokinetics, and clinical predictors

Introduction We have reported that clinical variables such as episode duration, early partial response, previous antidepressant trials, age at onset, and baseline symptom severity can be used to guide whether an older person whose depression has not responded to pharmacotherapy should be switched to a different medication (Kim et al., in press). In this study, we assessed whether combining clinical variables, pharmacogenetics, and pharmacokinetic parameters would improve decision trees guiding early treatment decisions. Methods We analyzed data from 218 older participants with major depressive disorder (MDD) who had not responded after four weeks of treatment with venlafaxine XR in the Incomplete Response in Late-Life Depression: Getting to Remission (IRL-Grey) study. Participants were treated up to 12 weeks, during which venlafaxine XR was titrated up to 300 mg/day or until remission was attained. Treatment response was defined as > 50% symptom improvement. Metabolizer status was derived from Cytochrome 2D6 (CYP) genotype using existing guidelines. Venlafaxine (VEN) and o-desemethylvenlafaxine (ODV) plasma levels collected at week 12 were used for population pharmacokinetic (PK) modeling using NONMEM to obtain VEN and ODV exposures. Active moiety (AM; VEN + ODV) exposure was also calculated. Using a receiver operating characteristic (ROC) model, clinical variables, CYP metabolizer status, VEN, ODV, and AM exposures were entered as potential predictors. One decision tree minimizing false identification of future responders (false positives) and another decision tree minimizing false identification of future non-responders were created (false negatives) with five-fold cross validation. Results Not having a partial response at week 4, a longer episode duration, and a higher AM exposure were identified as predictors of non-response at week 12. Test negative predictive values of the left most terminal node of the two trees were 78.3% and 77.2%, respectively, similar to the negative predictive values of the trees that did not include the pharmacogenetics and pharmacokinetic parameters. Conclusions In conclusion, higher AM exposure was identified as a predictor of non-response. Our decision trees also included two previously identified clinical predictors of non-response: early non-response and longer episode duration. Addition of pharmacogenetics and pharmacokinetic parameters did not improve the predictive ability of decision trees using clinical variables to predict response to venlafaxine XR in older patients with MDD. This research was funded by The IRL-GRey study was supported primarily by the National Institute of Mental Health (R01 MH083660 and P30 MH90333 to University of Pittsburgh, R01 MH083648 to Washington University, and R01 MH083643 to University of Toronto). Additional funding was provided by the UPMC Endowment in Geriatric Psychiatry, the Taylor Family Institute for Innovative Psychiatric Research (at Washington University), the Washington University Institute of Clinical and Translational Sciences grant UL1 TR000448 from the National Center for Advancing Translational Sciences (NCATS), and the Campbell Family Mental Health Research Institute at the Centre for Addiction and Mental Health, Toronto. Pfizer contributed venlafaxine extended-release capsules for this study and Bristol-Myers Squibb contributed aripiprazole and matching placebo tablets for the randomized phase of the study. We have reported that clinical variables such as episode duration, early partial response, previous antidepressant trials, age at onset, and baseline symptom severity can be used to guide whether an older person whose depression has not responded to pharmacotherapy should be switched to a different medication (Kim et al., in press). In this study, we assessed whether combining clinical variables, pharmacogenetics, and pharmacokinetic parameters would improve decision trees guiding early treatment decisions. We analyzed data from 218 older participants with major depressive disorder (MDD) who had not responded after four weeks of treatment with venlafaxine XR in the Incomplete Response in Late-Life Depression: Getting to Remission (IRL-Grey) study. Participants were treated up to 12 weeks, during which venlafaxine XR was titrated up to 300 mg/day or until remission was attained. Treatment response was defined as > 50% symptom improvement. Metabolizer status was derived from Cytochrome 2D6 (CYP) genotype using existing guidelines. Venlafaxine (VEN) and o-desemethylvenlafaxine (ODV) plasma levels collected at week 12 were used for population pharmacokinetic (PK) modeling using NONMEM to obtain VEN and ODV exposures. Active moiety (AM; VEN + ODV) exposure was also calculated. Using a receiver operating characteristic (ROC) model, clinical variables, CYP metabolizer status, VEN, ODV, and AM exposures were entered as potential predictors. One decision tree minimizing false identification of future responders (false positives) and another decision tree minimizing false identification of future non-responders were created (false negatives) with five-fold cross validation. Not having a partial response at week 4, a longer episode duration, and a higher AM exposure were identified as predictors of non-response at week 12. Test negative predictive values of the left most terminal node of the two trees were 78.3% and 77.2%, respectively, similar to the negative predictive values of the trees that did not include the pharmacogenetics and pharmacokinetic parameters. In conclusion, higher AM exposure was identified as a predictor of non-response. Our decision trees also included two previously identified clinical predictors of non-response: early non-response and longer episode duration. Addition of pharmacogenetics and pharmacokinetic parameters did not improve the predictive ability of decision trees using clinical variables to predict response to venlafaxine XR in older patients with MDD.