The Antidepressant Advisor (Study 3): fMRI Study to Predict Treatment Response in Patients With Depression — ADeSS_S3  

Major Depressive Disorder Clinical Trial

Official title: The Antidepressant Advisor: A Decision Support System for UK Primary Care - a Feasibility Study: Study 3

Status Completed

Clinical Trial Summary

This prospective observational study (ADeSS-Study3) investigates candidate biomarkers prospectively predicting response to antidepressant medications and prognosis in major depressive disorder (MDD). Currently, about half of MDD patients will not respond to the first course of selective serotonin reuptake inhibitors (SSRIs), while more than 40% will also not achieve remission after a second round of another SSRI. There are functional magnetic resonance imaging (fMRI) measures in several brain regions, showing clinical potential as predictors of response and non-response to SSRIs.

The overall aim of the study is to identify the neural signatures prospectively predicting poor prognosis in MDD patients after receiving four months of treatment in UK primary care. Specifically, it looks to evaluate four fMRI measures: 1) self-blame-selective subgenual cortex and ventral striatum connectivity with the right anterior temporal lobe; 2) pregenual anterior cingulate cortex activity in response to implicit emotional facial expressions; 3) amygdala activation in response to implicit emotional facial expressions; and 4) subgenual cingulate seed-based resting state.

In addition, a more specific objective of the study is to provide the proof-of-concept for using fMRI to prospectively predict which MDD patients will not benefit from SSRI antidepressant treatments in UK primary care. The long-term translational aim is to identify such patients and provide them with alternative treatments without delay by informing a decision support system with the information provided by these candidate biomarkers.

This study is linked to the Antidepressant Advisor Trial (ADeSS-Study 1: NCT03628027), in which the feasibility is evaluated of a novel computerised decision support system for antidepressant prescribing in MDD patients in a UK primary care setting.

Clinical Trial Description

This prospective observational study focuses on identifying the neural signatures prospectively predicting poor prognosis in MDD patients after receiving four months of treatment in UK primary care.

It is expected that half of the patients will show a response to primary care treatment over the 16 weeks of the trial. This allows for comparison of baseline MRI scans of treatment responders with non-responders. The aim is to recruit at least n=12 in each group with complete data (responders vs non-responders), but if the groups are skewed, a continuous measure (i.e. the difference between final and baseline depression scores) will be used rather than to categorise into responders and non-responders. This sample size is considered sufficient for fMRI studies to run random-effect models. As there is no previous study from which effect sizes can be drawn, the study remains exploratory.

As this study aims to find candidate biomarkers predicting response to antidepressant treatment, it is important to define "response" and "antidepressant treatment". Response to treatment will be defined as a reduction of depressive symptom levels of at least 50%, as assessed by the self-rated Quick Inventory of Depressive Symptomatology (Self-report; 16-Item) from baseline to follow-up.

Due to the range of different antidepressants used in the main trial (Study 1), the primary analysis is to compare responders versus non-responders irrespective of treatment received which ensures the results are of pragmatic relevance to primary care treatment (considered as a complex multifaceted intervention for this study). The following antidepressants are likely to be prescribed in the study, either recommended by the algorithm or as initiated by general practitioners (GPs) without computerised decision support tool advice: sertraline (≥50mg; SSRI), escitalopram (≥10mg; SSRI), mirtazapine (≥30mg; alpha 2 antagonist; NaSSA; dual serotonin and norepinephrine agent), vortioxetine (≥10mg; multimodal), citalopram (≥20mg; SSRI), fluoxetine (≥20mg; SSRI), paroxetine (≥20mg; SSRI), venlafaxine (≥75mg; SSRI up to 150 mg and SNRI>150 mg), duloxetine (60mg; SNRI), therapeutic doses of tricyclics, and therapeutic doses of monoamine oxidase inhibitors.

Further analysis will be aimed at comparing response to another SSRI versus no response to another SSRI. Here, treatment with an SSRI will be defined as treatment with any of the earlier listed SSRIs, as well as low dose venlafaxine (up to 150mg). This includes individuals who started and subsequently stopped their SSRI or did not go up to an effective dose because of presumable side effects (intention-to-treat analysis). However, it excludes individuals whose SSRI has not been changed or whose SSRI dose has not been increased for the study period; individuals whose SSRI remained the same, but the dose increased are included. Individuals who were incompliant with their prescribed SSRI medication are included as their incompliance might be related to experiencing a poor benefit/side effect ratio (intention-to-treat analysis). Individuals who had medications with additional non-SSRI properties are excluded from this analysis (venlafaxine > 150 mg, mirtazapine > 15 mg, antipsychotics, lithium, tricyclics in therapeutic dose, monoamine oxidase inhibitors in therapeutic dose, or vortioxetine).

In order to investigate the general research aim, i.e. identifying the neural signatures prospectively predicting poor prognosis in MDD patients after receiving four months of treatment in UK primary care, the following specific neuroanatomical hypotheses are tested:

i) Moral sentiment task:

Following Lythe et al 2015, a longer version of this task predicted subsequent recurrence risk in remitted MDD patients. The authors found that patients with recurring MDD exhibited increased right superior anterior temporal cortex (RSATL) connectivity with the subgenual cortex (BA25), compared with MDD patients who remained in stable remission and a control group. In addition, the authors noted that RSATL hyper-connectivity with ventral putamen/claustrum, and the temporoparietal junction was distinctive of recurring MDD compared with stable MDD. The RSATL - temporoparietal junction hyper-connectivity did not survive a more stringent unpublished analysis applying a cluster-forming threshold of p=0.001 uncorrected as is now recommended.

Based on the inclusion criteria, it is expected that most patients in the current study will have recurring MDD. It is important to note that the study by Lythe et al 2015 was carried out in people with full remission. This suggests that their study may have been biased towards patients with an overall good prognosis for remission, as they will have responded to treatment before, and against patients with a chronic course of MDD. So, even though BA25 - ATL hyper-connectivity is associated with recurrence risk, it might also represent a marker of good prognosis for symptom remission.

Additional unpublished data collected by the Lythe et al 2015 authors was used to explore characteristics between those whose remission was attributable to serotonergic medications (n=36) and those who achieved remission through other means (n=28, e.g. psychotherapy or spontaneous remission), which was determined by the PI based on their recorded treatment history and blinded to the imaging results. Interestingly, people with a history of achieving remission using alternatives to serotonergic medications showed stronger hyper-connectivity of the right ventral striatum - RSATL compared to the others. In contrast, there was no difference between the two groups with regard to subgenual frontal - RSATL connectivity for self- versus other-blame. Notably, the groups were well-matched on demographic and clinical variables, which suggests that the lower ventral striatum - RSATL connectivity cannot be explained by other confounders and potentially is a genuine marker for serotonergic responders.

The Structured Clinical Interview for DSM-5 allows for the subtyping of MDD patients into subtypes: with anxious distress, with melancholic features (its operationalisation will be modified by requiring a lack of mood reactivity as a necessary criterion), and with atypical features. It is plausible that the prognosis varies for these MDD subtypes based on differences with regard to which part of the self-blaming bias-related neural network is more relevant. For example, MDD patients with anxious distress have been shown to have more anger towards others (captured as moderate and severe levels on the AMDP psychopathology interview) and benefit less from self-blame-related subgenual (BA25) - ATL-connectivity-reducing neurofeedback. Therefore, it is predicted that patients with anxious MDD are less likely to display self-blame-selective subgenual (BA25) - ATL hyper-connectivity.

Similarly, people with attention deficit hyperactivity disorder (ADHD), whose condition involves emotional instability which is captured as "affective lability" (moderate or severe levels) on the AMDP psychopathology interview, are predicted to show a lower likelihood of response to standard treatment in primary care with serotonergic antidepressants. It is predicted that they are more likely to experience anger towards others (captured as moderate and severe levels on the AMDP) and are therefore less likely to show self-blame-selective subgenual (BA25) - ATL hyper-connectivity.

Therefore, the following is hypothesised:

1. a) Self-blame-selective BA25 (using the Lythe et al 2015 results as a cluster-based ROI)

• ATL hyper-connectivity (using the same right superior ATL seed as in Lythe et al 2015) will be associated with better prognosis after receiving four months of primary care treatment.

b) Moreover, a lack of self-blame-selective hyper-connectivity will be associated with patients with anxious distress, who have a poorer prognosis and are more frequently encountered in treatment-resistant and chronic MDD samples. This type of sample is expected to be recruited for this study, whereas they are less likely to be recruited in remitted MDD studies such as the one published by Lythe et al 2015.

2. Self-blame-selective ventral striatum - ATL hyper-connectivity will be associated with poorer prognosis after receiving four months of primary care treatment. This is based on the fact that primary care mostly relies on SSRIs and SSRI response, which was retrospectively associated with a lack of self-blame-selective ventral striatum-ATL hyper-connectivity.

3. Lower self-blame-selective ventral striatum - ATL connectivity will be associated with better response to serotonergic medications. This is based on the comparison of those whose remission was attributable to serotonergic medications and those who achieved remission through other means (unpublished secondary data analysis of Lythe et al 2015 data).

ii) Implicit facial emotions task:

A) Following Godlewska et al 2018, the study team will probe the reproducibility of their findings regarding pregenual anterior cingulate cortex (pgACC) activity in response to implicit facial emotions as a predictor of response to treatment (six-weeks of treatment with escitalopram). The authors showed increased activity in the pgACC to sad versus happy faces in responders compared to non-responders.

Therefore, the hypothesis is that decreased activity in the pgACC ROI (obtained from the authors) to implicit sad versus happy faces is prospectively associated with poor prognosis after receiving four months of treatment.

B) In addition, this task allows to investigate amygdala activation as a potential predictor of response to treatment. Williams et al 2015 reported that responders to antidepressants (escitalopram (mean = 11.3 mg, sd = 6.7 mg), sertraline (mean = 59.3 mg, sd = 27.0 mg) and venlafaxine (mean = 92.3, sd = 32.2)) showed lower activation of the bilateral amygdala (Automated Anatomical Labelling (AAL) atlas ROI) for happy and left amygdala for sad faces compared to non-responders at baseline. Moreover, response to SSRIs was associated with lower activation of the amygdala for sad faces, and non-responders to venlafaxine displayed higher activation for sad faces. This finding is corroborated by a meta-analysis of functional predictors of treatment response, which linked increased activation in the right amygdala to an increased likelihood of poor response.

Therefore, the hypothesis is that increased activation of the amygdala for implicit sad and happy faces prospectively predicts poor prognosis after receiving four months of treatment, and the study team predicts that this increased amygdala response is more pronounced in patients with anxious distress given the link between increased arousal and anxiety, as well as significance and arousal signals being linked to amygdala activation in response to visual stimuli.

iii) Resting state fMRI:

Following Dunlop et al 2017, the study team will probe the reproducibility of their findings regarding subgenual cingulate cortex (SCC) connectivity as a predictor of response to treatment. The authors identified resting state functional connectivity between the SCC and three regions as a potential predictor: the left dorsal midbrain, the ventrolateral prefrontal cortex (VLPFC, Brodmann Area (BA) 47) and the left ventromedial frontopolar cortex (VMPFC, BA10). Absent and/or negative functional connectivity with the SCC was associated with remission following SSRI/SNRI treatment, whereas positive functional connectivity with the SCC was linked with treatment failure.

Therefore, the hypothesis is that positive functional connectivity between the SCC seed used by Dunlop - and the three ROIs to be obtained from the authors: VLPFC, SCC - VMPFC and SCC - dorsal midbrain prospectively predict poor prognosis after receiving four months of treatment. ;

Related Conditions & MeSH terms

• Depression
• Depressive Disorder
• Depressive Disorder, Major
• Major Depressive Disorder

• NCT number: NCT04342299
• Study type: Observational
• Source: King's College London
• Status: Completed
• Start date: August 1, 2018
• Completion date: November 18, 2021