AI-derived body composition analysis reveals muscle volume and metformin-associated adipose effects and the obesity paradox in non-small cell lung cancer

作者：Xinan Wang, Akinori Hata, Noriaki Wada, Jiyeon Song, Taiki Fukuda, Yusei Nakamura, Mizuki Nishino, Yi Li, Mark L. Schiebler, David C. Christiani, Hiroto Hatabu · 发表于：EBioMedicine · 年份：2025 · DOI：10.1016/j.ebiom.2025.105995 · 被引用次数：4 · 研究领域：Nutrition and Health in Aging、Cancer Risks and Factors、Metabolism, Diabetes, and Cancer

Background Body composition is emerging as a prognostic factor in cancer, with metformin showing potential antitumour effects in patients with obesity and non-small cell lung cancer (NSCLC). We aimed to determine which specific body composition components drive survival differences in different patient subgroups and explore underlying molecular mechanisms. Methods In this retrospective cohort study, we analysed 1275 patients with confirmed NSCLC diagnosis and available chest CT imaging for body composition analysis at the Massachusetts General Hospital. Six body composition parameters (muscle density and volume, visceral adipose tissue [VAT], subcutaneous adipose tissue [SAT], VAT/SAT ratio, and intramuscular adipose tissue [IMAT]) were quantified using validated CNN-based segmentation algorithms from baseline CT scans. The primary outcome was overall survival (OS), assessed from diagnosis until death or last follow-up. Associations with OS were analysed using multivariable Cox proportional hazards model, followed by subgroup analyses by sex, BMI, diabetes, and metformin use. Circulating cytokine levels (n = 93) and oncogenic driver mutations (n = 646) from the imaging cohort and comprehensive tumour mutation profiles from an independent Memorial Sloan Kettering Cancer Center cohort (n = 5363) were analysed for mechanistic insights. Findings Muscle volume demonstrated a dose–response relationship with OS (T2 vs. T1: HR=0.78, 95% CI: 0.65–0.94; T3 vs. T1: HR=0.71, 95% CI: 0.57...