Just a moment, the page is loading...

A Data-Driven Statistical/Machine Learning Analysis for Extracting Clinically Actionable Predictions of Antidepressant Treatment Outcomes with Individual Depressive Symptoms

A Data-Driven Statistical/Machine Learning Analysis for Extracting Clinically Actionable Predictions of Antidepressant Treatment Outcomes with Individual Depressive Symptoms

William V. Bobo

Mayo Clinic, Dept. of Psychiatry and Psychology

12 October 2018

Major depressive disorder (MDD) affects over 15 million people in the U.S., and is the leading cause of disability worldwide. MDD can generally be managed with appropriate pharmacotherapy. However, most antidepressants used to treat MDD require several weeks to take effect, and multiple therapeutic trials of antidepressants typically needed to achieve remission. Therefore, there is a critical need to develop clinically actionable prediction models which not only predict eventual clinical outcomes (remission, response, or non-response) at the earliest point in time after treatment initiation, but also provide quantitative prognostic evidence of the predicted outcome. Such evidence-based predictions would be of immense value to clinicians, as it would provide an objective means of triaging patients to an alternative drug at a relatively early stage of antidepressant treatment based on a high likelihood of eventual non-response - a marked advance over the current “wait-and-see” approach used in managing antidepressant treatment. This has stimulated interest in the use of statistical/machine learning approaches to achieve accurate prediction of antidepressant treatment outcomes using large and complex arrays of individual patient-level data. The clinical utility of existing statistical/machine learning-based predictive approaches, which have only used baseline symptoms and sociodemographic factors as predictor variables, is limited because of: 1) the high degree of heterogeneity of individual depressive symptoms, 2) the weak predictive effects of sociodemographic variables--individually or in aggregate, and 3) poor capability of depression rating scales for predicting antidepressant treatment outcomes. In addressing these limitations, our goal is to develop a statistical/machine learning tool to identify a set of individual depressive symptoms (core symptoms) which demonstrate longitudinal homogeneity in their response to antidepressants in such a way that their early change (e.g., at 4 weeks in an 8-week treatment period) are highly prognostic and predictive of eventual treatment outcomes. In a preliminary study, our novel machine learning workflow comprising probabilistic graphs identified core symptoms from the full rating scales (QIDS-C and HAM-D) using data from 3 large 8-week clinical SSRI trials (STAR*D, PGRN-AMPS and ISPC) of citalopram/escitalopram treatment in over 1,400 MDD patients. Supervised learning methods predicted sex-specific non-response and remission at 8 weeks with AUC 0.62-0.94 using the baseline severity of core symptoms and their associated changes at 4 weeks. The prognostic capabilities of the core symptoms identified using our approach replicated across all 3 clinical trials. Building upon the replicating findings from our preliminary study, the central hypothesis of this proposal is that our machine learning approach will identify a similar set of core depressive symptoms whose baseline and early changes (e.g., at 4 weeks) can be used to accurately predict longitudinal responses to SSRIs, SNRIs, and other antidepressants over 8-12 weeks of treatment.

[{ "PostingID": 1649, "Title": "LILLY-F1J-US-HMFA", "Description": "Duloxetine Versus Placebo in the Long-Term Treatment of Patients With Late-Life Major Depression" },{ "PostingID": 1956, "Title": "LILLY-F1J-US-HMFS", "Description": "Duloxetine Versus Placebo in Patients With Major Depressive Disorder (MDD): Assessment of Energy and Vitality in MDD" },{ "PostingID": 2412, "Title": "LILLY-F1J-MC-HMBV", "Description": "Duloxetine Versus Placebo in the Treatment of Elderly Patients With Major Depressive Disorder" },{ "PostingID": 2602, "Title": "*BI-1208.24", "Description": "Study to Assess Clinical Response of Duloxetine in Patients Hospitalized for Severe Depression" },{ "PostingID": 3090, "Title": "LILLY-F1J-US-HMGR", "Description": "A Phase 4, 8-week, Double-blind, Randomized, Placebo-controlled Study Evaluating the Efficacy of Duloxetine 60 mg Once Daily in Outpatients with Major Depressive Disorder and Associated Painful Physical Symptoms" },{ "PostingID": 3092, "Title": "LILLY-F1J-US-HMGU", "Description": "Duloxetine Versus Placebo in the Acute Treatment of Patients With Major Depressive Disorder and Associated Painful Physical Symptoms" }]

Statistical Analysis Plan