• Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, PA, USA.
• Cancer Epidemiology and Prevention Program, UPMC Hillman Cancer Center, Pittsburgh, PA, USA.
• Department of Radiology, School of Medicine, University of Pittsburgh, Pittsburgh, PA, USA. jip13@pitt.edu.

OBJECTIVES: To investigate if body composition can serve as a biomarker for assessing the risk of developing lung cancer.

MATERIALS AND METHODS: We conducted a retrospective study using low-dose computed tomography (LDCT) scans from the Pittsburgh lung screening study (PLuSS) (n = 3635, 22 follow-up years) and the NLST-ACRIN (n = 16,360, 8 follow-up years) cohort. Five types of body tissues, including subcutaneous adipose tissue (SAT), visceral adipose tissue (VAT), intramuscular adipose tissue (IMAT), skeletal muscle (SM), and bone, were automatically segmented by our previously developed algorithms. Volume and density metrics were computed. Cause-specific Cox proportional hazards models were utilized to assess hazard ratios (HRs). Time-dependent area under the receiver operating characteristic curve (AUC-ROC) was used to evaluate model performance. The cumulative incidence function was estimated for different risk groups.

RESULTS: The final composite models were formed by age (HR = 1.30 (95% CI: 1.17-1.43)), current smoking status (HR = 1.85 (1.49-2.28)), bone volume (HR = 1.38 (1.25-1.52)), bone density (HR = 0.80 (0.71-0.89)), SM density (HR = 0.62 (0.58-0.66)), IMAT ratio (HR = 0.65 (0.58-0.73)), and SAT volume (HR = 0.76 (0.67-0.87)). The model trained on the PLuSS cohort achieved a mean AUC of 0.77 (0.75-0.79) over 21 years and 0.71 (0.68-0.74) over the first 7 years for lung cancer prediction. External validation on the NLST cohort yielded AUC values ranging from 0.63 to 0.66 over a 7-year follow-up period. The model trained on a combined dataset of PLuSS and NLST achieved a mean AUC of 0.71 (0.7-0.72) over 21 years.

CONCLUSION: Three-dimensional body composition metrics assessed through LDCT are a significant predictor of lung cancer risk.

KEY POINTS: Question Is body composition a biomarker for lung cancer risk assessment? Findings Body composition metrics derived from low-dose CT scans, including volumes and densities of fat, bone, and muscle, are strong predictors of lung cancer risk. Clinical relevance Lung cancer risk stratification can be improved by body composition features, providing guidance for personalized lung cancer screening strategies.