RG 5187: Towards precision psychotherapy for non-respondent patients: From signatures to predictions to clinical utility

Although cognitive-behavioral therapy (CBT) is a first-line treatment for internalizing disorders, a substantial proportion of patients fails to benefit - with severe consequences for patients and costs for societies. Precision mental health can help to identify patients at risk for non-response (NR) already prior to treatment initialization. The paucity of standard clinical features that allow for single-case predictions serves as an impetus to search for additional layers of NR. The work pro-gram of this Research Unit (RU) will foster the development of precision psychotherapy by i) in-vestigating clinical and bio-behavioral signatures of NR to improve our understanding of this phenomenon, ii) applying state-of-the-art machine learning technology for single-case predic-tions, and iii) validating these for clinical utility in an ecologically valid treatment setting, bring-ing together four major academic outpatient clinics in Berlin. Our effort will thus pave the way for a priori patient stratification to intensified or augmented treatments in a putative second funding period. To achieve this, we will set up a prospective-longitudinal multicenter observational study on n = 500 patients with internalizing disorders (specific phobia, social anxiety disorder, panic disorder, agoraphobia, generalized anxiety disorder, obsessive-compulsive disorder, post-traumatic stress disorder, unipolar depressive disorders) who will be deeply phenotyped prior to CBT using hypotheses-based clinical, e-mental health, psychophysiological and neuroimaging measures. Assessment batteries and treatment documentation will be harmonized across cen-ters. Predictive analytics will be provided by our methods platform, including computer vision algo-rithms such as convolutional neural networks, multiple kernel and transfer learning and an infra-structural basis (hard- and software, data management plans, high-performance computing). The RU aims to significantly improve the field by 1) setting up a multilevel and -method assessment battery to search for the best predictors, combinations thereof, and cost-efficient proxies, 2) in-vestigating bio-behavioral signatures of emotion regulation as a putative key mechanism of CBT, 3) applying a transdiagnostic focus on NR signatures, 4) within one comprehensive sample that exerts a high degree of ecological validity, thus fostering translation to clinical practice with diverse patient characteristics. These goals can only be achieved by concerted ac-tion of experts in the fields of clinical psychology, psychotherapy, e-mental health, psychophysiol-ogy, cognitive neuroscience, and neuroinformatics. We will maximize synergies with large-scale consortia (UK Biobank, ENIGMA, CRC-TRR 58, BMBF psychotherapy initiative, PING, KODAP). This RU will make substantial progress in answering the question how we can better under-stand the phenomenon of NR, identify and address this vulnerable and cost-intensive group of NR patients.