Work

Efficacy of a large language model-driven application for emotion management in breast cancer patients (EMBRACE): protocol for a multicentre randomised controlled trial

Year: 2026

Type: article

Abstract: Introduction In breast cancer, psychological distress predicts poorer quality of life, adherence and clinical outcomes. Most digital psychosocial interventions rely on static content and infrequent monitoring, missing patients’ evolving emotional states. Therefore, a key clinical challenge is the timely identification of these emotional fluctuations and the tailoring of interventions accordingly. We have developed a large language model (LLM)-enabled emotion management system, whose core methodological innovation lies in leveraging patient-generated text from social media platforms as the primary analytical source to deliver guideline-aligned personalised supportive interventions for patients. The study aims to establish whether this intervention improves health-related quality of life (HRQoL), psychological outcomes, symptom burden while reducing unplanned hospitalisations in patients with breast cancer. Methods and analysis This prospective stratified block randomised controlled trial targets a total sample of 440 eligible patients with breast cancer with 1:1 allocation, recruited from six tertiary hospitals stratified by regional economic levels. Eligibility criteria include histologically confirmed primary breast cancer, smartphone access and regular social media use; those with cognitive impairment, concurrent malignancies or active suicidal ideation are excluded. The intervention group receives usual care plus an LLM-based emotion management system that analyses patient text to generate personalised emotion labels and match symptom keywords, then provides tailored evidence-based interventions. The control group receives usual care. Assessments are conducted at baseline, 1, 3 and 6 months. The primary outcome is HRQoL assessed using the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire-Core 30 (EORTC QLQ-C30) summary score, with the primary endpoint defined as the 6-month assessment after randomisation; secondary outcomes cover individual HRQoL domains of the EORTC QLQ-C30 to examine effects on function and symptom status, anxiety (Generalised Anxiety Disorder-7), depression (Patient Health Questionnaire-9), symptom burden (Memorial Symptom Assessment Scale), rate of unplanned hospitalisation and time to first unplanned hospitalisation. Ethics and dissemination This trial is conducted in accordance with the principles of the Declaration of Helsinki and the International Council for Harmonisation Guidelines for Good Clinical Practice. Ethical approval was obtained from the ethics committees of 6 hospitals. Written informed consent was collected from all participants before enrolment. Participants can withdraw freely at any time without prejudice to routine clinical care. Trial results will be submitted for publication in peer-reviewed journals. Trial registration number NCT07452055.

Source: BMJ Open

DOI: 10.1136/bmjopen-2026-126107