Timing of Virtual Reality Relaxation Training for Patients With Breast Cancer Undergoing Chemotherapy: Randomized Comparative Study With Multimodal Physiological Monitoring and Explainable Machine Learning (Preprint)

Résumé fourni par la source

BACKGROUND Chemotherapy for breast cancer is frequently accompanied by psychological distress and treatment-related symptoms that may be associated with autonomic, respiratory, electrodermal, and neural changes. Virtual reality (VR) relaxation can provide immersive and guided breathing, but the optimal timing of VR delivery relative to chemotherapy has not been well characterized using objective multimodal physiology. OBJECTIVE This study aimed to compare the short-term effects of VR-based mindfulness breathing administered before versus after chemotherapy in patients with breast cancer and to identify physiological features associated with pre- to post-intervention state changes using explainable machine learning. METHODS A total of 78 female patients with stage I-III breast cancer scheduled for chemotherapy were randomly assigned to a before-chemotherapy group (BC group; n=39) or an after-chemotherapy group (AC group; n=39). Participants completed a 12-minute immersive VR relaxation intervention with guided mindfulness breathing. Electroencephalography (EEG), electrocardiography(EKG)-derived heart rate variability (HRV), respiratory effort, and galvanic skin response (GSR) were collected during baseline and postintervention periods. Patient-reported stress and symptom burden were assessed using the Newly Diagnosed Breast Cancer Stress Scale-Revised and the Memorial Symptom Assessment Scale-Short Form. Within-group changes were analyzed using paired t tests and Cohen d. Machine learning models were trained to distinguish pre- and post-intervention physiological states, and SHapley Additive exPlanations (SHAP) were used to identify influential features. RESULTS The mean age was 54.95 (SD 7.14) years in the BC group and 54.00 (SD 10.34) years in the AC group. Both groups showed postintervention physiological changes, but the AC group showed broader and more consistent modulation across modalities. EEG alpha and beta features changed significantly in both groups, with larger and more widespread effects in the AC group (eg, FC3 alpha: d=1.064, P<.001; FCz alpha: d=0.780, P=.001). HRV changes were observed in both groups, but the AC group showed significant changes in more indices, including RMSSD, SDSD, LF relative power, HF relative power, HF log power, FFT ratio, SD1, and SD ratio. Respiratory variability changed significantly only in the AC group, including RRV_ApEn (d=0.737, P=.002). The machine learning classification was more accurate in the AC group, particularly using fused multimodal features, and SHAP highlighted HRV LF relative power, occipital alpha activity, GSR mini skin conductance responses, and respiratory median breath-to-breath intervals as influential features. CONCLUSIONS A single 12-minute VR relaxation session was associated with measurable physiological changes in patients with breast cancer undergoing chemotherapy. Changes were more extensive after chemotherapy than before chemotherapy, suggesting that postchemotherapy delivery may be a particularly responsive window for VR-based relaxation support. These findings are exploratory and should be confirmed in adequately powered trials with complete reporting of patient-reported outcomes, adverse events, clinical symptom changes, and longitudinal effects. CLINICALTRIAL This study was registered as NCT06541587 on ClinicalTrials.gov.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.