Using fMRI and machine learning to predict symptom improvement following cognitive behavioural therapy for psychosis.
|
| Abstract |
Cognitive behavioural therapy for psychosis (CBTp) involves helping patients to understand and reframe threatening appraisals of their psychotic experiences to reduce distress and increase functioning. Whilst CBTp is effective for many, it is not effective for all patients and the factors predicting a good outcome remain poorly understood. Machine learning is a powerful approach that allows new predictors to be identified in a data-driven way, which can inform understanding of the mechanisms underlying therapeutic interventions, and ultimately make predictions about symptom improvement at the individual patient level. Thirty-eight patients with a diagnosis of schizophrenia completed a social affect task during functional MRI. Multivariate pattern analysis assessed whether treatment response in those receiving CBTp (n = 22) could be predicted by pre-therapy neural responses to facial affect that was either threat-related (ambiguous 'neutral' faces perceived as threatening in psychosis, in addition to angry and fearful faces) or prosocial (happy faces). The models predicted improvement in psychotic (r = 0.63, p = 0.003) and affective (r = 0.31, p = 0.05) symptoms following CBTp, but not in the treatment-as-usual group (n = 16). Psychotic symptom improvement was predicted by neural responses to threat-related affect across sensorimotor and frontal-limbic regions, whereas affective symptom improvement was predicted by neural responses to fearful faces only as well as prosocial affect across sensorimotor and frontal regions. These findings suggest that CBTp most likely improves psychotic and affective symptoms in those endorsing more threatening appraisals and mood-congruent processing biases, respectively, which are explored and reframed as part of the therapy. This study improves our understanding of the neurobiology of treatment response and provides a foundation that will hopefully lead to greater precision and tailoring of the interventions offered to patients.
|
|---|---|
| Authors | Eva Tolmeijer, Veena Kumari, Emmanuelle Peters, Steven C R Williams, Liam Mason |
| Journal | NeuroImage. Clinical (Neuroimage Clin) Vol. 20 Pg. 1053-1061 ( 2018) ISSN: 2213-1582 [Electronic] Netherlands |
| PMID | 30343250 (Publication Type: Journal Article, Research Support, Non-U.S. Gov't) |
| Copyright | Copyright © 2018 The Authors. Published by Elsevier Inc. All rights reserved. |
| Topics |
|
Join CureHunter, for free Research Interface BASIC access!
Realize the full power of the drug-disease research graph!