Project Abstract
Because not all patients with major depressive disorder (MDD) respond to standard treatments, alternative
therapies such as transcranial magnetic stimulation (TMS), ketamine, and vagus nerve stimulation (VNS) have
been introduced. These alternative therapies require more provider-intensive monitoring and thus are typically
only recommended for those with “treatment resistant” depression (TRD). However, establishing treatment
resistance clinically requires evidence of multiple treatment failures, and this is a costly and lengthy
process that burdens patients. If we could identify patients prone to multiple treatment failures of first-line
therapies, we could refer to alternative treatments sooner. The most relevant component of the task of
stratifying patients to alternative treatments is to separate partial response (PR) from true treatment
failure (TF) in the patients who do not meet response criteria. The current proposal will use trajectory
classification methods to empirically identify a group of patients with consistent nonresponse to a range of
first-line treatments (SSRIs, SNRIs, CBT), more accurately representing true treatment failure (TF). We will
use the method in both combined clinical trial data (PReDICT, iSPOT-D, STAR*D) as well as treatment as
usual (TAU) data (MARS). Then to demonstrate that these patients are more likely to be treatment resistant,
we will demonstrate that those with TF are less likely to respond to switch/augmentation of first-line treatments
than partial responders (PR) for the trial data and examine an association with a measure of TRD collected
for the observational data. If there is a link between initial TF and TRD, this group could then be targeted for
earlier referral to alternative treatments. The study will then demonstrate an analytic framework for future
modeling of TF by estimating effect size for a previously derived genetic biomarker using proper weighting
methods. The current study will thus allow us to bridge the gap from well-controlled clinical trial data to
future studies of more heterogeneous electronic health record (EHR) data and pave the way for possible real-
world changes in recommendations for treatment resistant patients.