Bioelectrical impedance analysis-derived body composition metrics as prognostic biomarkers in lung cancer

Background: Body composition changes are common in lung cancer and relate to poor outcomes, but their treatment-related changes and prognostic value are not well defined. This study serially evaluated body composition metrics in patients with lung cancer receiving palliative treatment. Methods: In this study, we prospectively and serially evaluated body composition metrics in patients with lung cancer receiving palliative treatment, using bioelectrical impedance analyzer (InBody 770 and S10, InBody Co., Seoul, Korea). The patients were scheduled for a minimum of 12 weeks of treatment, and the analysis was conducted every 3 weeks, in addition to collecting clinical data. The optimal thresholds of the death and progression prediction model were identified based on area under curve (AUC) and F1 score. Results: A total of 141 patients were enrolled in the study. The median follow-up duration was 21.3 months (range, 13.2-24.9 months). Nearly half of the cohort was aged ≥65 years (N=75, 53.2%). The majority of patients had non-small cell lung cancer (N=127, 90.1%), with a smaller subset having small cell lung cancer (N=14, 9.9%). Among the enrolled patients, 92 (65.2%) received palliative cytotoxic chemotherapy (CTx group), and 49 (34.8%) received palliative tyrosine kinase inhibitor therapy (TKI group). From baseline to the end of treatment, CTx group showed a decrease in phase angle (from 4.7 ± 0.78 to 4.48 ± 0.71; p < 0.001) and an increase in extracellular water (ECW) ratio (from 0.392 ± 0.008 to 0.395 ± 0.008; p < 0.001). Whereas no significant changes of metrics were observed in TKI group. For death prediction using time-point-specific body composition metrics, the CTx group showed the highest performance for events occurring within 1 year, whereas the TKI group performed best for events within 2 years. For progression prediction, the CTx group performed best for 6-month events, while the TKI group showed the best performance for 18-month events. In the CTx group, the prediction models achieved AUCs of 0.899 for death and 0.831 for progression, with corresponding F1 scores of 0.708 and 0.790. In the TKI group, the AUCs were 0.961 for death and 0.829 for progression, with F1 scores of 0.640 and 0.817, respectively. Shapley Additive exPlanations (SHAP) analysis identified percent body fat (PBF) and skeletal muscle index (SMI) as the most influential predictors for both outcomes. Conclusions: Serial body composition metrics assessments captured treatment-related body composition changes in lung cancer, and models based on these metrics showed prognostic performance. These findings support the use of longitudinal body composition monitoring in personalized patient management.