Network structure of depression, anxiety and stress in patients with non-Hodgkin's lymphoma receiving chemotherapy: a multicentre cross-sectional study.

Abstract

To estimate the item-level conditional association network of depression, anxiety and stress symptoms in patients with non-Hodgkin's lymphoma (NHL) during a prespecified chemotherapy intercycle period and to identify symptoms with comparatively high expected influence, bridge expected influence and predictability. Multicentre cross-sectional study. Oncology departments of four tertiary hospitals in Jiangsu Province, China, from October 2024 to June 2025. A total of 798 adults with pathologically confirmed NHL were assessed during the prespecified chemotherapy intercycle period. The 21 items of the Depression Anxiety Stress Scales-21 (DASS-21) were modelled as nodes in a regularised partial-correlation network. Primary network measures were edge weights, expected influence and bridge expected influence; secondary measures were node predictability and bootstrap indices of accuracy and stability. The strongest conditional association was between difficulty initiating tasks (D2) and lack of enthusiasm (D5; edge weight=0.282). Irritability (S7; expected influence=1.106), depression/dejection (D4; 1.065) and nervous energy (S3; 1.051) had the highest expected influence and also the highest bridge expected influence (0.972, 0.803 and 0.720, respectively). Trembling (A3; predictability=0.561), irritability (S7; 0.481) and intolerance of hindrance (S6; 0.476) had the highest predictability. Correlation-stability coefficients for expected influence and bridge expected influence were both 0.67. Irritability, depression/dejection and nervous energy showed comparatively high signed connectivity within and across the prespecified DASS-21 domains, while the D2-D5 pair had the strongest conditional association. These cross-sectional findings are descriptive and hypothesis-generating and do not establish temporal, causal or treatment effects.