AI-derived body composition analysis reveals muscle volume and metformin-associated adipose effects and the obesity paradox in non-small cell lung 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: 0.57-0.88) and the effect was intensified in patients with a BMI ≥25 kg/m2 (T2 vs. T1: HR = 0.64, 95% CI: 0.49-0.85; T3 vs. T1: HR = 0.56, 95% CI: 0.41-0.77) and with diabetes (T2 vs. T1: HR = 0.61, 95% CI: 0.39-0.97; T3 vs. T1: HR = 0.50, 95% CI: 0.30-0.84). Higher SAT volume (T2 vs. T1: HR = 0.48, 95% CI: 0.26-0.88; T3 vs. T1: HR = 0.42, 95% CI: 0.21-0.83) and IMAT (T2 vs. T1: HR = 0.52, 95% CI: 0.28-0.97; T3 vs. T1: HR = 0.42, 95% CI: 0.22-0.82) predicted prolonged OS specifically in patients who have used metformin. Genomic analysis revealed that KRAS mutation was significantly enriched in patients with higher SAT, VAT and BMI ≥25 kg/m2 while EGFR mutation was enriched in patients with lower SAT, IMAT and BMI <25 kg/m2. INTERPRETATION: Higher muscle volume consistently predicts improved survival in patients with NSCLC, with enhanced effects in patients with elevated BMI or diabetes. Increased adipose tissue components benefit only metformin users, suggesting drug-specific metabolic interactions. Our work demonstrates the clinical utility of detailed body composition and suggests opportunities for personalised treatment approaches based on comprehensive body composition profiles rather than BMI alone. FUNDING: This work was supported by grants from the National Cancer Institute of the National Institutes of Health: U01CA209414 to X.W. and D.C.C., 1K99CA297010 to X.W.