OBJECTIVES: This study investigated the correlation between body mass index (BMI), bioelectrical impedance analysis (BIA)-derived specific body compositions, and the obesity paradox, with a focus on sex differences in non-smallcell lung cancer (NSCLC). METHODS: This study is a multicenter prospective cohort. Association with survival was determined using Cox proportional hazard regression analysis. The mediating effect of systemic inflammation on the association between body composition and survival rate was determined using mediation analysis. RESULTS: Fat mass had a strong positive correlation with BMI, but a weak positive correlation with other BIA-derived body compositions. Obese, overweight, and normal weight patients had significantly longer median survival than patients with BMI <
18.5 ([23.2 versus 18.5 versus 17.5 versus 15.8] months, P <
0.002). Increased BMI and BIA-derived body composition indicators were associated with decreased all-cause mortality. Multivariable-adjusted analysis demonstrated that BMI, fat mass (FM), fat-free mass, muscle mass, bone mass, protein mass, total body water, extracellular water, and intracellular water were independent protective factors affecting the prognosis of patients with NSCLC. A difference in body compositions performance in the prognostic assessment of the different sexes was observed. Fat-related body compositions exhibited a significant obesity paradox in females compared to in males. Systemic inflammation played a crucial role in the relationship between body compositions and disease prognosis, with a mediating effect on body fat in males (31%) and females (19.7%). CONCLUSIONS: The obesity paradox exists among patients with NSCLC and is not influenced by specific body compositions. In addition, this paradox is notably more common in female patients. TRIAL REGISTRATION: Registration number: ChiCTR1800020329.