Oscillatory and recurrent drivers of symptom networks in head and neck cancer radiotherapy: a longitudinal cross-lagged panel network study.

Abstract

To determine the dynamic complexity of symptom networks in patients with head and neck cancer during radiotherapy to inform precise symptom management. Symptom severity was assessed using the Chinese version of the MD Anderson Symptom Inventory-Head & Neck module at four time points(T1, 1 week before radiotherapy; T2, 2-5 sessions; T3, 16-20 sessions; T4, 30-33 sessions). Longitudinal predictive relationships among the 20 symptoms were analyzed using a cross-lagged panel network (CLPN), and the accuracy and stability of the network were evaluated using nonparametric bootstrap methods. As a directed network derived from longitudinal data, the directionality within the CLPN was also examined using out-expected influence (out-EI) and in-expected influence (in-EI). Among 366 patients analyzed, symptom incidence and severity increased significantly during radiotherapy, peaking at T3. Dry mouth, taste disturbance, and loss of appetite were the most prevalent and severe symptoms. Oropharyngeal pain exhibited an oscillatory driver pattern: negative (out-EI = -2.48) → positive (out-EI = 2.30) → negative (out-EI = -2.21). Loss of appetite displayed a recurrent driver (out-EI = 1.65)-latent (out-EI = 0.35)-redriver (out-EI = 1.20) trajectory. Drowsiness transitioned from a mid-radiotherapy driver (out-EI = 1.43) to a late-phase receiver (in-EI = 1.61). The symptom network demonstrated good stability from T2 to T4. This study presents the first longitudinal network analysis of symptom dynamics during radiotherapy for HNC. The findings lay the groundwork for stage-specific interventions to optimize symptom management and improve patient quality of life.