Neurocognitive signatures predicting risk of recurrent depression
Lead Research Organisation:
King's College London
Department Name: Psychological Medicine
Abstract
Depression is a leading cause of disability, because many people who have recovered from its symptoms ("symptomatic phase") will experience recurring episodes. Most research has focused on the symptomatic phase of depression and often assumed that when people have no symptoms, they are cured. A disorder can, however, be ongoing even when patients do not experience symptoms. In this "asymptomatic" state people can be at high risk of developing symptoms in the future. Many patients take antidepressant medication over years to reduce the risk of recurrence, because our current way of predicting their risk based on number of previous episodes is very inaccurate. It is usually assumed that those treatments working to reduce symptoms of depression will also be beneficial in their future prevention, but this is largely unproven. The development of new treatments that prevent recurrence has been hampered by a lack of knowledge about psychological and brain changes that are risk factors.
In an MRC-funded study, we have identified such risk factors in recovered depression patients that predict, on an individual basis, which patient will have another episode in the next year. By adding functional MRI scans, and a novel test of being inclined to self-blame to standard measures, we achieved 83% accuracy, exceeding the recommended target for useful so-called "prognostic markers". In contrast, standard measures alone were no better than chance guessing who would develop another episode. Patients were scanned whilst they experienced self-blame, which is thought to play an important role in depression by decreasing self-worth and hope. We have demonstrated that asymptomatic patients who go on to develop depression show altered connections in the self-blame-related brain network which differed from those who remained well.
Despite these encouraging results in 50 patients, it is unknown:
1) whether we can confirm the result that functional MRI and psychological tests of self-blame predict subsequent recurrence of depression in a larger independent group
2) whether MRI is needed for predicting who will develop depression at an individual level or could be replaced by adding further psychological and hormonal measures
3) whether the brain networks found to be disrupted when blaming oneself in depression are linked to abnormal stress hormones, the only established chemical risk factor for recurrence
4) whether the disruption in brain networks when blaming oneself makes people more vulnerable to develop depression after a stressful life event measured weekly via a mobile app
To answer these questions, we propose to enrol 150 patients recovered from depression who have stopped their antidepressant medication in accordance with guidelines (as in our previous study). An initial MRI scan, cognitive tests, and stress hormones will be used to predict recurrence after one year. This will deliver much needed evidence for reproducible psychological and brain-based risk factors to 1) inform novel psychological and brain training, as well as brain stimulation treatment approaches and 2) develop a so-called "prognostic marker" to predict recurrence risk for affected individuals. This marker could be used in future clinical trials to select patients that are at high risk of recurrence. This would greatly reduce the number of patients needed for a trial and thereby reduce its cost. Further, if we can replace expensive MRI scans with cheaper measures, the prognostic marker could be studied in future trials to determine whether it helps people with depression to make decisions about continuing their antidepressant medication. After completing this project, our future goal is to facilitate the development of novel treatments more likely to remedy depression in the long-term by preventing recurrence rather than to only treat its symptoms. Our collaborator, Janssen, are actively developing markers for depression recurrence and are highly committed.
In an MRC-funded study, we have identified such risk factors in recovered depression patients that predict, on an individual basis, which patient will have another episode in the next year. By adding functional MRI scans, and a novel test of being inclined to self-blame to standard measures, we achieved 83% accuracy, exceeding the recommended target for useful so-called "prognostic markers". In contrast, standard measures alone were no better than chance guessing who would develop another episode. Patients were scanned whilst they experienced self-blame, which is thought to play an important role in depression by decreasing self-worth and hope. We have demonstrated that asymptomatic patients who go on to develop depression show altered connections in the self-blame-related brain network which differed from those who remained well.
Despite these encouraging results in 50 patients, it is unknown:
1) whether we can confirm the result that functional MRI and psychological tests of self-blame predict subsequent recurrence of depression in a larger independent group
2) whether MRI is needed for predicting who will develop depression at an individual level or could be replaced by adding further psychological and hormonal measures
3) whether the brain networks found to be disrupted when blaming oneself in depression are linked to abnormal stress hormones, the only established chemical risk factor for recurrence
4) whether the disruption in brain networks when blaming oneself makes people more vulnerable to develop depression after a stressful life event measured weekly via a mobile app
To answer these questions, we propose to enrol 150 patients recovered from depression who have stopped their antidepressant medication in accordance with guidelines (as in our previous study). An initial MRI scan, cognitive tests, and stress hormones will be used to predict recurrence after one year. This will deliver much needed evidence for reproducible psychological and brain-based risk factors to 1) inform novel psychological and brain training, as well as brain stimulation treatment approaches and 2) develop a so-called "prognostic marker" to predict recurrence risk for affected individuals. This marker could be used in future clinical trials to select patients that are at high risk of recurrence. This would greatly reduce the number of patients needed for a trial and thereby reduce its cost. Further, if we can replace expensive MRI scans with cheaper measures, the prognostic marker could be studied in future trials to determine whether it helps people with depression to make decisions about continuing their antidepressant medication. After completing this project, our future goal is to facilitate the development of novel treatments more likely to remedy depression in the long-term by preventing recurrence rather than to only treat its symptoms. Our collaborator, Janssen, are actively developing markers for depression recurrence and are highly committed.
Technical Summary
IMPORTANCE
Depression is a leading cause of disability, mainly because of recurrent major depressive (MDD) episodes. Why some patients experience recurring MDD whilst others remain well is poorly understood. Identifying the neurocognitive mechanisms of how MDD evolves from its asymptomatic precursors will enable the development of better treatments and improve long-term outcomes.
BACKGROUND
Patients with MDD show biases towards blaming themselves for failure. We have identified the associated neural network including anterior temporal and limbic forebrain regions. Combining a novel cognitive test with self-blame-related hyper-connectivity and standard clinical measures in remitted MDD prospectively predicted who will develop recurrence on an individual basis (83% cross-validated accuracy).
GAP-OF-KNOWLEDGE
Despite these encouraging results, it is unknown 1) whether the identified neurocognitive risk factors of recurrence can be replicated in a larger independent sample, 2) whether MRI could be replaced by additional psychological/biochemical measures for individual risk prediction 3) whether the neural signature is associated with the only established biochemical correlate of recurrence risk (cortisol response), and 4) how this increases vulnerability to recurrence after stressful life events measured weekly via a mobile app.
AIMS/METHODOLOGY
To answer these questions, we propose to enrol 150 medication-free patients recovered from MDD as in our previous study. Initial MRI and cortisol measures will be used to predict recurrence after 14 months.
IMPACT
This project will enhance our understanding of recurrence risk factors, which is necessary to develop better long-term treatments by: 1) informing the design of future psychological and neuromodulation interventions, and 2) contributing to developing a prognostic marker for selection of patients at high recurrence risk in trials, thereby increasing statistical power for developing preventative treatments.
Depression is a leading cause of disability, mainly because of recurrent major depressive (MDD) episodes. Why some patients experience recurring MDD whilst others remain well is poorly understood. Identifying the neurocognitive mechanisms of how MDD evolves from its asymptomatic precursors will enable the development of better treatments and improve long-term outcomes.
BACKGROUND
Patients with MDD show biases towards blaming themselves for failure. We have identified the associated neural network including anterior temporal and limbic forebrain regions. Combining a novel cognitive test with self-blame-related hyper-connectivity and standard clinical measures in remitted MDD prospectively predicted who will develop recurrence on an individual basis (83% cross-validated accuracy).
GAP-OF-KNOWLEDGE
Despite these encouraging results, it is unknown 1) whether the identified neurocognitive risk factors of recurrence can be replicated in a larger independent sample, 2) whether MRI could be replaced by additional psychological/biochemical measures for individual risk prediction 3) whether the neural signature is associated with the only established biochemical correlate of recurrence risk (cortisol response), and 4) how this increases vulnerability to recurrence after stressful life events measured weekly via a mobile app.
AIMS/METHODOLOGY
To answer these questions, we propose to enrol 150 medication-free patients recovered from MDD as in our previous study. Initial MRI and cortisol measures will be used to predict recurrence after 14 months.
IMPACT
This project will enhance our understanding of recurrence risk factors, which is necessary to develop better long-term treatments by: 1) informing the design of future psychological and neuromodulation interventions, and 2) contributing to developing a prognostic marker for selection of patients at high recurrence risk in trials, thereby increasing statistical power for developing preventative treatments.
Planned Impact
The aim of this proposal is to validate risk factors of recurrence to pursue our future goal of facilitating the development of novel treatments to remedy depression in the long-term. Our impact objectives are:
1) Identify reproducible risk factors at the group level to inform the design of novel treatments to be investigated in further trial grant applications:
a) Psychological: For example, replicating self-blame-related action tendencies such as feeling-like-hiding as risk factors, would provide a strong rationale for probing novel interventions such as imagery re-scripting approaches to tackle these (Holmes et al, 2007; Brewin et al, 2009).
b) fMRI: Identify which parts of the ATL-frontolimbic network are reproducibly predictive of recurrence, informing novel fMRI neurofeedback and neurostimulation protocols. For example, in a recent proof-of-concept trial of ATL-subgenual neurofeedback to tackle self-blame (ISRCTN10526888), we found highest response rates in non-anxious MDD; data validating other brain regions would allow optimised multiregional neurofeedback designs to benefit people with anxious MDD which is generally more treatment-resistant.
2) Provide individual predictions to develop a prognostic marker:
a) If fMRI can be replaced by cheaper measures whilst retaining accuracy, these could comprise a prognostic marker for enriching long-term clinical trials with patients at high-recurrence risk and for clinical decision support systems for prophylactic treatments. The latter application would require trials in patients who have recovered from depression but have not yet decided whether to stop their medication and to test the benefits of informing their decision with prognostic marker information vs. providing mock decision support in the control arm.
or b) If fMRI cannot be replaced in this project, we would apply for funding to study replacing fMRI measures with cheaper technologies, e.g. near infra-red spectroscopy (fNIRS) that is now portable but cannot reach deep brain regions, yet could measure ATL activation and connectivity with the frontopolar cortex which were strong predictors in our previous model.
This aligns with Janssen's goals, who have prioritised identifying a marker of recurrence risk in mood disorders. This is because the marker's presence as inclusion criterion could be used to enrich clinical trial samples with MDD patients at high risk of recurrence, thereby increasing the statistical power for detecting the benefits of new treatments in decreasing recurrence risk in longer-term trials. So far, trials have usually probed prevention of recurrence by continuing the same antidepressant that helped patients recover from symptoms vs. switching them to placebo. This design assumed that treatments which help patients to recover are also best for keeping them stable. This classical paradigm has, however, shifted with Janssen's licence for intranasal Esketamine which has rapid but transient antidepressant effects. So, novel treatments will need to be developed for those patients who have only responded to these rapid-onset antidepressants to maintain their response. Our prognostic marker could be further validated in this population for example and then be used for enriching long-term trials of novel maintenance treatments. Such trial designs, where recovered MDD patients were enrolled to investigate preventative treatments were used for Mindfulness-Based Cognitive Therapy, now NICE-recommended for remitted MDD at recurrence risk (>2 previous episodes).
BENEFICIARIES OF OUR RESEARCH
By pursuing these objectives, this project will maximise the likelihood of benefits to people with depression and their families, the National Health Service, and the pharmaceutical industry, such as our collaborator Janssen, as well as the e-health industry, such as EMIS PLC with whom we are already collaborating on a decision support system for MDD treatment (ClinicalTrials.gov:NCT03628027).
1) Identify reproducible risk factors at the group level to inform the design of novel treatments to be investigated in further trial grant applications:
a) Psychological: For example, replicating self-blame-related action tendencies such as feeling-like-hiding as risk factors, would provide a strong rationale for probing novel interventions such as imagery re-scripting approaches to tackle these (Holmes et al, 2007; Brewin et al, 2009).
b) fMRI: Identify which parts of the ATL-frontolimbic network are reproducibly predictive of recurrence, informing novel fMRI neurofeedback and neurostimulation protocols. For example, in a recent proof-of-concept trial of ATL-subgenual neurofeedback to tackle self-blame (ISRCTN10526888), we found highest response rates in non-anxious MDD; data validating other brain regions would allow optimised multiregional neurofeedback designs to benefit people with anxious MDD which is generally more treatment-resistant.
2) Provide individual predictions to develop a prognostic marker:
a) If fMRI can be replaced by cheaper measures whilst retaining accuracy, these could comprise a prognostic marker for enriching long-term clinical trials with patients at high-recurrence risk and for clinical decision support systems for prophylactic treatments. The latter application would require trials in patients who have recovered from depression but have not yet decided whether to stop their medication and to test the benefits of informing their decision with prognostic marker information vs. providing mock decision support in the control arm.
or b) If fMRI cannot be replaced in this project, we would apply for funding to study replacing fMRI measures with cheaper technologies, e.g. near infra-red spectroscopy (fNIRS) that is now portable but cannot reach deep brain regions, yet could measure ATL activation and connectivity with the frontopolar cortex which were strong predictors in our previous model.
This aligns with Janssen's goals, who have prioritised identifying a marker of recurrence risk in mood disorders. This is because the marker's presence as inclusion criterion could be used to enrich clinical trial samples with MDD patients at high risk of recurrence, thereby increasing the statistical power for detecting the benefits of new treatments in decreasing recurrence risk in longer-term trials. So far, trials have usually probed prevention of recurrence by continuing the same antidepressant that helped patients recover from symptoms vs. switching them to placebo. This design assumed that treatments which help patients to recover are also best for keeping them stable. This classical paradigm has, however, shifted with Janssen's licence for intranasal Esketamine which has rapid but transient antidepressant effects. So, novel treatments will need to be developed for those patients who have only responded to these rapid-onset antidepressants to maintain their response. Our prognostic marker could be further validated in this population for example and then be used for enriching long-term trials of novel maintenance treatments. Such trial designs, where recovered MDD patients were enrolled to investigate preventative treatments were used for Mindfulness-Based Cognitive Therapy, now NICE-recommended for remitted MDD at recurrence risk (>2 previous episodes).
BENEFICIARIES OF OUR RESEARCH
By pursuing these objectives, this project will maximise the likelihood of benefits to people with depression and their families, the National Health Service, and the pharmaceutical industry, such as our collaborator Janssen, as well as the e-health industry, such as EMIS PLC with whom we are already collaborating on a decision support system for MDD treatment (ClinicalTrials.gov:NCT03628027).
Organisations
- King's College London (Lead Research Organisation, Project Partner)
- University of Sussex (Collaboration)
- Radboud University Nijmegen Medical Center (Collaboration)
- Technical University of Dresden (Collaboration)
- Scients Institute, USA (Collaboration)
- Universidade Federal do ABC (Project Partner)
- FC Donders Centre for Cognitive Neuroima (Project Partner)
- D'OR Institute f Research and Education (Project Partner)
Publications
Fennema D
(2023)
Self-blame-selective hyper-connectivity between anterior temporal and subgenual cortices predicts prognosis in major depressive disorder
in NeuroImage: Clinical
Rutherford S
(2022)
Charting brain growth and aging at high spatial precision.
in eLife
Description | MRC-DTP at King's College London |
Amount | £70,000 (GBP) |
Organisation | Medical Research Council (MRC) |
Sector | Public |
Country | United Kingdom |
Start | 10/2021 |
End | 03/2025 |
Description | NIHR Maudsley Biomedical Research Centre for Mental Health |
Amount | £41,000,000 (GBP) |
Funding ID | Matthew Hotopf CI - £41 Million NIHR contribution overall, Roland Zahn deputy lead for the Mood Disorders theme, receiving over £1 Million |
Organisation | National Institute for Health Research |
Sector | Public |
Country | United Kingdom |
Start | 12/2022 |
End | 11/2027 |
Description | Unravelling neural states and traits in recurrent depressive disorder |
Amount | £214,000 (GBP) |
Organisation | Scients Institute, USA |
Sector | Charity/Non Profit |
Country | United States |
Start | 07/2022 |
End | 06/2025 |
Title | fMRI biomarker of recurrence risk in major depressive disorder |
Description | This project has attained its aims of identifying the first rigorously investigated fMRI biomarker of recurrence risk in major depressive disorder. |
Type Of Material | Technology assay or reagent |
Year Produced | 2021 |
Provided To Others? | No |
Impact | We have published the final results of our study on the novel biomarker and will make it available to other labs on publication. We have already used this biomarker in a proof-of-concept clinical trial. |
URL | https://github.com/AndrewLawrence/dCVnet |
Description | Co-supervisor and PI of PhD students as part of 10 year German Research Funded International Graduate School |
Organisation | Technical University of Dresden |
Country | Germany |
Sector | Academic/University |
PI Contribution | I am a co-PI of two PhD project themes which will run over the next years as a joint international PhD programme with Dresden University funded by the German Research Foundation with over 4 Million Euros. |
Collaborator Contribution | They will primarily supervise the PhD students and the funding will be held by Dresden. |
Impact | not yet |
Start Year | 2022 |
Description | Collaboration with Drs Marquand and Ruhe on machine learning analysis of imaging data |
Organisation | Radboud University Nijmegen Medical Center |
Country | Netherlands |
Sector | Academic/University |
PI Contribution | We have shared anonymised data from our MRC-funded project for further machine learning analyses linked to a PhD studentship and an ERC-grant |
Collaborator Contribution | The partners provide one PhD student and carry out all the analyses |
Impact | No outputs so far |
Start Year | 2021 |
Description | External PhD supervisor University of Sussex |
Organisation | University of Sussex |
Department | Brighton and Sussex Medical School |
Country | United Kingdom |
Sector | Academic/University |
PI Contribution | Based on our previous work on fMRI neurofeedback in depression and predicting risk of depression, I have been asked to act as an external supervisor for a PhD student superived by Drs Stone and Colasanti. Data from previous work has been shared using a data sharing agreement. |
Collaborator Contribution | They have provided funding for a full-time PhD student and research costs. |
Impact | not yet |
Start Year | 2021 |
Description | Memorandum of Understanding with Scients Institute USA |
Organisation | Scients Institute, USA |
Country | United States |
Sector | Charity/Non Profit |
PI Contribution | I have formed a partnership between King's College London and Scients Institute USA which led to a memorandum of understanding. |
Collaborator Contribution | Scients Institute which is a charitable foundation in the USA has funded research costs for one of my PhD students, Suqian Duan, to launch an innovative project using virtual reality to assess the phenomenology of depression |
Impact | No outputs have arisen yet |
Start Year | 2020 |
Title | Multimodal risk prediction modal including functional MRI for recurrence risk in major depressive disorder |
Description | https://doi.org/10.1016/j.bpsc.2021.06.010 describes the prediction model and associated software code. We have received funding from the MRC for external validation and further development as part of the NESPRED project |
Type | Diagnostic Tool - Imaging |
Current Stage Of Development | Initial development |
Year Development Stage Completed | 2022 |
Development Status | Under active development/distribution |
Impact | First imaging-based prediction tool for recurrence risk in depression achieving individual predictive values in clinically relevant range >80% |
URL | https://doi.org/10.1016/j.bpsc.2021.06.010 |
Title | dCVnet - software for clinical prediction |
Description | dCVnet is a package for R- which allows doubly nested cross-validated regularised regression using elastic net. The software is designed for clinicians who are unable to code and allows prediction of clinical outcomes in small samples using statistical learning methods. |
Type Of Technology | Software |
Year Produced | 2022 |
Open Source License? | Yes |
Impact | The software has allowed us to build a robust prediction model for recurrence risk in depression which we are currently replicating. |
URL | https://github.com/AndrewLawrence/dCVnet |
Description | "Brain functions and mental health - the example of depression". Medication in Mental Health conference, 24.11.2021 |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | National |
Primary Audience | Patients, carers and/or patient groups |
Results and Impact | Online Talk for a conference on Medication in Mental Health organised by a carer charity and directed at people with lived experience and their families |
Year(s) Of Engagement Activity | 2021 |
URL | https://mmhuk.com/ |
Description | Freely available MOOC on Depression for European Psychiatric Association |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | In 2020, I have developed and filmed an online course on Understanding and Treating Depression for the European Psychiatric Association with Profs Young and Cleare which has been completed by a large group of psychiatrists and also entailed my group's research on self-blaming biases in depression. This course was Janssen-sponsored. |
Year(s) Of Engagement Activity | 2020 |
URL | https://elearning.europsy.net/enrol/synopsis/index.php?id=8 |
Description | Inspire the mind Blog Article https://medium.com/inspire-the-mind/self-blame-in-depression-957fc1b4bd09 |
Form Of Engagement Activity | Engagement focused website, blog or social media channel |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Public/other audiences |
Results and Impact | Blog post about my research |
Year(s) Of Engagement Activity | 2022 |
URL | https://medium.com/inspire-the-mind/self-blame-in-depression-957fc1b4bd09 |
Description | Youtube video citing our work on Psych2Go |
Form Of Engagement Activity | Engagement focused website, blog or social media channel |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Media (as a channel to the public) |
Results and Impact | A YouTube video (https://youtu.be/foduemIqFGM) on Psych2Go cited our work (Zahn et al., 2015) and explains "self-blame" and related feelings as a key symptoms of depression in Oct 2021, which has been watched over 300,000 times in the first two months already. Although "self-blame" has been used in the psychological literature before, we have coined "self-blaming feelings" as an umbrella term to capture a variety of feelings which together we observed in over 80% of patients with depression and which were not restricted to guilt, but also included self-disgust/contempt/loathing, shame and self-directed anger (Zahn et al., 2015). We had developed an addendum to a German psychopathology interview to capture blame-related feelings. |
Year(s) Of Engagement Activity | 2021 |
URL | https://youtu.be/foduemIqFGM |