Considerable effort has been made to identify bio-behavioral markers—neural, behavioral, and clinical variables—that will predict whether a depressed person will respond or fail to respond to commonly prescribed antidepressant drug therapies. The problem is not that such variables and correlations have not been found. It is that none so far has stood up to subsequent independent replication and validation efforts in a manner sufficiently robust to affect how the drugs are prescribed.

The response rate to a first antidepressant prescription is estimated at 30% to 50%. Many depressed people try a number of different SSRIs, in trial-and-error fashion, hoping to find one that both reduces symptoms and has few or no significant side effects. A team of researchers that includes several BBRF Scientific Council members and grantees has been involved in recent years in an effort aimed at doubling the number of people who respond to a first antidepressant prescription, with “response” defined as a reduction in symptoms of 50% or more within a specified time from initiation of treatment, typically one or several months.

In a new paper appearing in Nature Mental Health, researchers led by Diego A. Pizzagalli, Ph.D., who is affiliated with McLean Hospital, Harvard Medical School, and the University of California, Irvine, tested a set of predictors of response to two widely prescribed antidepressants, sertraline and bupropion. They developed the predictors, which included biological (including MRI-based) and behavioral markers, from data generated in a large, completed clinical trial (called EMBARC). In their new paper, they report results of a test of these predictors in a new trial (called SMART-D) involving a group of as yet untreated depressed patients who were randomly assigned to one or the other of the medicines.

This approach addresses the problem of reproducibility. The practical question about any biological or behavioral predictors of treatment is whether their use would help doctors determine which medicine to give a new patient. Does knowing that a patient carries a specific biomarker predicting a positive response to a given treatment correspond with better response rates than the 30%-50% seen in the “trial-and-error” approach with commonly prescribed antidepressants?

Dr. Pizzagalli, a member of the BBRF Scientific Council, is a 2017 BBRF Distinguished Investigator and 2008 Independent Investigator. Also on the team were Kerry J. Ressler, M.D., Ph.D., BBRF Scientific Council, 2017 BBRF Distinguished Investigator, and 2005 and 2002 Young Investigator; Gordana Vitaliano, M.D., 2010 and 2007 BBRF Young Investigator; and Brain P. Brennan, M.D., 2012 and 2008 BBRF Young Investigator. Co-first authors of the paper were Drs. Peter Zhukovsky, Manuel Kuhn, and Lauren R. Borchers.

The EMBARC trial (Establishing Moderators and Biosignatures of Antidepressant Response in Clinical Care) was a landmark, multi-center clinical trial focused on precision medicine for Major Depressive Disorder (MDD). It sought to discover biosignatures to predict if specific depressed patients would respond well to a particular antidepressant. 296 participants were randomly assigned to receive either sertraline (Zoloft) or placebo. In the more recent SMART-D trial, 47 never-treated patients with major depression were randomized to receive either sertraline or bupropion (Wellbutrin).

The medicines assigned to participants in SMART-D differed in type. Sertraline is an SSRI medicine which is thought to exert antidepressant effects by increasing serotonin neurotransmission, while bupropion is NDRI medicine thought to exert therapeutic effects by increasing neurotransmission of dopamine and norepinephrine. Despite their different mechanisms of action, the two medicines both act on “monoamine” neurotransmitters and have similar efficacy rates, as meta-analyses have indicated.

The team developed a series of biological and behavioral markers based on EMBARC data that they used to predict response of participants in the SMART-D trial to sertraline and bupropion. The markers included data on each patient for resting-state fMRI connectivity (in a circuit connecting the nucleus accumbens and the rostral anterior cingulate cortex—rACC); indicators of reward learning and sensitivity and cognitive control, clinical variables (depression severity and the trait of neuroticism), and employment status (a demographic indicator).

The EMBARC results had revealed that higher depression severity, higher neuroticism, older age, less impairment in cognitive control, and being employed specifically predicted a positive response to sertraline compared with placebo. Data related to the trial also was collected for those who did not respond to sertraline after 8 weeks but might respond to bupropion. Here, the key markers suggesting a positive response were: higher resting-state functional connectivity between the nucleus accumbens and the rACC; and in behavioral tests, a stronger response bias to a more frequently rewarded stimulus and higher reward sensitivity.

The SMART-D participants were over 70% female and were mostly in their 20s and 30s. Each completed fMRI and clinical and cognitive assessments, and within 3 days the team ascribed to each a biomarker status, reflecting the team’s modeling of whether the individual would respond well to sertraline, bupropion, both medicines, or neither medicine. Then each participant was randomly assigned to receive one of the two drugs. The results of treatment in each patient would, by definition, either be consistent or inconsistent with what the team, using their biomarker status, predicted their response would be.

Based on the biomarker-based models, 14 SMART-D participants were expected to respond well to both drugs; 17 to respond well to bupropion but not to sertraline; 10 to sertraline but not to bupropion; and 7 to neither drug.

Participants who were biomarker-negative for both drugs showed significantly worse depression symptom trajectories over the 8 weeks of the trial, compared with those who were biomarker-positive for at least one of the drugs or for both drugs (regardless of which one they received).

Specifically, participants with positive biomarkers for both drugs collectively had a 71.4% response rate, compared with a 42.9% response rate among those with no positive biomarkers. The latter figure is within the range of expected response rates estimated by experts to the “trial-and-error” approach to prescribing an initial antidepressant. The rate among those who had positive markers for both drugs thus was 66% higher—not a doubling of the trail-and-error rate, but a large increase that would translate into important information for prescribing doctors, assuming they knew their patients’ biomarker status.

For trial participants with positive markers for one medicine (either one), the response rate was 65.4%, which was 52% higher than the response rate among those with no positive biomarkers.

The fact that trial participants with positive biomarkers for both drugs collectively had a 71.4% response rate and those with positive markers for one of the drugs had a 65.4% response rate suggests the possibility that the two drugs tested in this case, despite differences in their mechanism of action, affect common pathways important in the antidepressant response. If this is ultimately proven to be true in subsequent validating research, then biomarkers predicting that a particular patient will respond to a drug engaging those pathways could help lay foundations for precision psychiatry in depression.