50P Radiomic features of both primary lung nodules and lymph node metastases on chest CT associated with progression-free survival in advanced non-small cell lung cancer patients

Metastatic nodules, particularly lymph node (LN) metastases, share biological and histological characteristics with primary lung nodules. However, most radiomics studies on lung cancer have focused only on primary nodules, and little is known about the prognostic value of radiomic features from metastatic nodules. We hypothesize that LN contains independent prognostic value, and a combination of both primary lung nodules and metastatic LN nodules on chest CT can predict patient outcomes more accurately. De-identified chest CT scans from 79 advanced NSCLC patients receiving first-line immunotherapy at Memorial Sloan Kettering Cancer Center (MSKCC) were analyzed. Radiologists annotated primary lung nodules and LN metastases, extracting textural radiomic features. The study's primary endpoint was progression-free survival (PFS). Three models, utilizing a cross-validation approach, were developed based on a) primary nodules, b) metastatic lymph nodes, and c) a combination of both. The least absolute shrinkage and selection operator (LASSO) Cox regression built the radiomic signature for PFS, yielding a radiomic risk score (RRS). High- and low-risk groups, defined by median RRS, were compared using a statistical Student t test for hazard ratios across different models. The RRS, computed from the top six selected features with corresponding coefficients, demonstrated a significant association with PFS (HR = 2.11, 95% CI: [1.54 – 3.43], P = 0.002) when derived from both primary nodules and LN metastases. In contrast, radiomic features from either primary nodules or LN metastases alone exhibited lower HRs for PFS (HR=1.95 vs HR=1.37). Notably, a statistically significant difference in HRs was observed between the LN+PN model and LN alone, while no difference was found in HRs between LN+PN and PN alone. The study findings suggest that combining radiomic features from both primary lung nodules and lymph node metastases provides complementary information that improves the prediction of progression-free survival in advanced non-small cell lung cancer patients.