Psilocybin therapy for treatment resistant depression: prediction of clinical outcome by natural language processing

A machine learning model that analyzes language from therapy sessions can predict which patients with treatment-resistant depression will respond to psilocybin therapy. Researchers used a zero-shot classifier based on the BART large language model to measure sentiment (valence and arousal) in transcripts of therapist-patient conversations one day after COMP360 psilocybin administration. These sentiment scores, combined with the Emotional Breakthrough Index and treatment arm, were fed into multinomial logistic regression models. The models predicted responder status at week 3 and through week 12 with 85% and 88% accuracy, respectively, and AUC values of 88% and 85%. This approach could enable early identification of patients needing alternative treatments.

Study at a glance

Abstract

Abstract Rationale Therapeutic administration of psychedelics has shown significant potential in historical accounts and recent clinical trials in the treatment of depression and other mood disorders. A recent randomized double-blind phase-IIb study demonstrated the safety and efficacy of COMP360, COMPASS Pathways’ proprietary synthetic formulation of psilocybin, in participants with treatment-resistant depression. Objective While the phase-IIb results are promising, the treatment works for a portion of the population and early prediction of outcome is a key objective as it would allow early identification of those likely to require alternative treatment. Methods Transcripts were made from audio recordings of the psychological support session between participant and therapist 1 day post COMP360 administration. A zero-shot machine learning classifier based on the BART large language model was used to compute two-dimensional sentiment (valence and arousal) for the participant and therapist from the transcript. These scores, combined with the Emotional Breakthrough Index (EBI) and treatment arm were used to predict treatment outcome as measured by MADRS scores. (Code and data are available at https://github.com/compasspathways/Sentiment2D .) Results Two multinomial logistic regression models were fit to predict responder status at week 3 and through week 12. Cross-validation of these models resulted in 85% and 88% accuracy and AUC values of 88% and 85%. Conclusions A machine learning algorithm using NLP and EBI accurately predicts long-term patient response, allowing rapid prognostication of personalized response to psilocybin treatment and insight into therapeutic model optimization. Further research is required to understand if language data from earlier stages in the therapeutic process hold similar predictive power.