Impact of an artificial intelligence–based management model on psychological and behavioral outcomes in breast cancer patients undergoing home CDK4/6 inhibitor therapy: A multicenter randomized controlled trial.

Abstract

633 Background: Patients receiving CDK4/6 inhibitors (CDK4/6i) often face significant psychological and physical challenges. This study evaluated the multi-dimensional impact of an AI-based management model (Group A), a customized active follow-up model (Group B), and conventional care (Group C) on the psychological status, quality of life (QoL), and biological indicators of patients with HR+/HER2- breast cancer and their primary caregivers . Methods: In this randomized controlled trial (2024–2025), 140 patient-caregiver dyads were enrolled and randomized into Group A (n=47), Group B (n=45), or Group C (n=48) . Longitudinal assessments were conducted using the Self-Rating Anxiety/Depression Scales (SAS/SDS), EORTC QLQ-C30, and PRO-CTCAE . Statistical analyses included mixed-design ANOVA for longitudinal changes, multiple correspondence analysis (MCA) for caregiver coping patterns, and Pearson correlation for psychological and serological indicators . Results: Baseline characteristics were balanced across groups (P > 0.05) . Mixed-design ANOVA revealed significant time-by-group interaction effects for anxiety (F=5.961, P < 0.001) and depression (F=4.931, P < 0.001) . During follow-up, Groups A and B showed significantly better psychological stability and depression relief compared to Group C; Group C’s psychological status deteriorated over treatment cycles, peaking at T3 (16 weeks), while Groups A and B remained stable or improved (P < 0.001) . Regarding QoL, Groups A and B significantly outperformed Group C in cognitive, emotional, social functioning, and global health (P < 0.05), with Group A showing superior cognitive function scores . MCA identified a strong clustering between the AI-based model and "positive coping strategies" among caregivers, whereas Group C was associated with "severe psychological stress" . Pearson analysis showed SAS/SDS scores positively correlated with ALT (r=0.178, P < 0.05) and glucose (r=0.221, P < 0.05) . Furthermore, Group A achieved a higher capture rate of adverse events via AI monitoring, effectively mitigating high-grade toxicity risks . Conclusions: AI-based management and customized follow-up significantly improve psychological outcomes and QoL in breast cancer patients on CDK4/6i. The correlation between psychological stress and physiological markers (ALT, Glucose) highlights the necessity of a "bio-psycho-social" approach in oncology management. These findings support integrating AI models into precision survivorship care .