Observational StudyPrediction of Antidepressant Treatment Response Using Machine Learning Classification Analysis

Despite significant advances in pharmacological treatment, the global burden of depression is increasing worldwide. The major challenge in antidepressant treatment is the clinicians' inability to predict the variability in individual response to the treatment. The development of biomarkers to predict treatment outcomes would enable clinician to find the right medication for a particular patient at the early stage of the treatment and thus could reduce prolonged suffering and ineffective protracted treatment. Brain imaging studies that examined brain predictors of treatment response based on group comparisons have limited value in classifying individuals as responders or non-responders. Machine learning classification techniques such as the support vector machine (SVM) method have proven useful in the classification of individual brain image observations into distinct groups or classes. However, studies that have applied the SVM method to structural and functional magnetic resonance scans (fMRI) involved small sample sizes and were confounded by placebo responses. Furthermore, a recent meta-analysis of clinical trials and EEG studies have shown that early clinical responses and brain changes at the early phase of antidepressant treatment may predict later clinical outcomes suggesting that neural markers measured in the early phase of antidepressant treatment may improve predictive accuracy. However, there is no fMRI study to date that has examined the predictive accuracy of data obtained in early phase of the treatment. We have preliminary fMRI data relating to early treatment response that form the basis of this proposed study.

The main objective of this study is to use machine learning method to examine the predictive value (sensitivity, specificity, accuracy) of resting state and emotional task-related fMRI data collected at pre-treatment baseline (week 0) and in the early phase of antidepressant treatment (week 2) in the classification of remitters (< 10 MADRS scores after 12 weeks of treatment) and non-remitters in patients with major depressive disorder (MDD). A secondary objective is to determine which data set (week 0 or week 2) gives the best predictive value.