Early Symptom Improvement as a Predictor of Antidepressant Response in Children and Adolescents Diagnosed with Depression: Translating Evidence from Randomized Controlled Trials to Community Practice

Abstract

Statement of the Problem: A common problem among children and adolescents diagnosed with depression who receive care in community settings is that antidepressant regimen changes such as psychotropic augmentation may occur soon after starting treatment. This raises the question as to whether such changes are implemented among youth who would otherwise respond to the antidepressant. Thus, the overarching objectives of this dissertation were to 1) distinguish early in treatment children and adolescents who are likely to respond, and 2) empirically evaluate the association between predicted response and psychotropic augmentation or switching in real world settings. Summary of Methods: Using randomized clinical trial (RCT) data, this research applied a Bayesian approach to predict the likelihood of initial (12 week) and sustained (18 week) response to treatment as a function of early changes in depressive symptoms (i.e. mood, somatic, subjective and behavioral) and other demographic and clinical factors. An innovative application of combined sample multiple imputations (CSMI) was used to estimate the 12-week predicted probability of response among commercially insured adolescents who received care in real-world settings. Each adolescent received a probability of treatment response, which was then used to compare the odds of psychotropic augmentation or switch. Results: Early changes in mood and somatic symptoms within the first six weeks of treatment are primary predictors of initial (at 12 weeks) and sustained (at 18 weeks) response to an antidepressant. Baseline depression severity is an important prognostic factor for initial response, and additional, though minimal improvement, in somatic symptoms from weeks 6 to 12 is indicative of sustained response. In a highly selected cohort of adolescents receiving care in community settings, an augmentation or switch occurred similarly among adolescents with a high versus low likelihood of responding to fluoxetine. Conclusion: The results suggest that other factors beyond expected antidepressant response (or lack thereof) might influence current treatment practices. Our findings have clinical and public health implications that support measurement-based care in pediatric depression. Our application of CSMI highlights several key areas of consideration for future pharmacoepidemiologic research aimed at translating RCT evidence to real world data to better understand clinical practices patterns.