Using ¹⁸F-FDG PET/CT-derived body composition features to predict lymphovascular invasion in non-small cell lung cancer.

Abstract

Lymphovascular invasion (LVI) in non-small cell lung cancer (NSCLC) is a critical prognostic marker linked to higher risks of metastasis and recurrence. This study aimed to develop a non-invasive predictive model using body composition features from ¹⁸F-FDG PET/CT imaging to assess LVI risk in early-stage NSCLC patients. We retrospectively analyzed 248 patients, including 153 from Vienna (training cohort) and 95 from Budapest (validation cohort). Preoperative ¹⁸F-FDG PET/CT scans were used to assess tumor metabolic parameters, including standardized uptake values (SUVmax, SUVmean), metabolic tumor volume (MTV), and total lesion glycolysis (TLG), as well as body composition features, including visceral, subcutaneous, and intermuscular adipose tissue, skeletal muscle at L1-L5. LASSO regression identified key body composition features, and a logistic regression-based nomogram was constructed and validated through ROC analysis, calibration, decision curve analysis, and survival analysis. LVI was present in 66/153 (43.1%) of Vienna and 39/95 (41.1%) of Budapest patients. The nomogram, developed using the Vienna training cohort, incorporating MTV, N stage, and body composition achieved an AUC of 0.839 and 0.790 in the Budapest validation cohort. Statistical tests confirmed that the nomogram significantly outperformed models based on either clinical (p = 7.92e-06) or imaging variables alone (p = 0.0474). Furthermore, LVI predicted by the nomogram was associated with significantly poorer 3-year recurrence-free and 5-year survival. Integrating body composition with clinical and tumor metabolic features from PET/CT enables preoperative prediction of LVI in NSCLC, supporting improved risk stratification.