What question did this study set out to answer?

This research aims to identify demographic, clinical, and imaging predictors of EGFR-positive lung adenocarcinoma among individuals who have never smoked.

May 29, 2026

Predictors of EGFR-mutant primary lung adenocarcinoma.

Key Points

This research aims to identify demographic, clinical, and imaging predictors of EGFR-positive lung adenocarcinoma among individuals who have never smoked.
Data from Weill Cornell’s Lung Cancer Database were used, integrating genomic and clinical data.
LASSO-penalized logistic regression was applied for predictor identification within a cohort of 1145 adults with surgically resected lung adenocarcinoma.
Model performance was assessed using cross-validated AUC and classification accuracy.
EGFR-positive adenocarcinomas were more likely to present as subsolid nodules (39.3% vs 21.7%, p<0.001).
Distinct characteristics predicting EGFR positivity included female sex, Asian ethnicity, and high SSN density.
The model achieved good discrimination with an AUC of 0.77, indicating effective prediction of EGFR mutation status.

Abstract

10591 Background: The proportion of lung cancer cases in individuals who have never smoked (INS) is increasing. Most of these cases are adenocarcinoma (AC) harboring Epidermal Growth Factor Receptor (EGFR) mutations. As predictors of AC and mutational status (EGFR positive EGFRp vs EGFR negative EGFRn) are incompletely elucidated, we sought to identify demographic, clinical, and imaging predictors of EGFRp AC. Methods: Data were obtained from Weill Cornell’s Lung Cancer Database, integrating EGFR mutation status derived from genomic testing, validated natural language processing of unstructured clinical data, and individual and area-level predictors from a large New York City health system. We included patients with early stage, surgically resected primary lung AC, allowing for uniform, robust pathologic and genomic characterization. Lesions were identified on computed tomography (CT) prior to diagnosis The primary outcome was EGFR positivity. LASSO-penalized logistic regression with 10-fold cross-validation was used to identify key predictors in high-dimensional, correlated data. Model performance was evaluated using cross-validated AUC and classification accuracy. Results: Among 1145 adults with primary lung AC (312 subsolid nodules SSNs and 833 solid nodules; mean age 69.9 SD, 9.6), 58.1% were female and 75.4% had a history of smoking. EGFRp ACs (vs EGFRn) were more likely to present as SSNs (39.3% vs 21.7%, p1 SSN on CT, former smoking status, family history of LC, and higher exposure to ambient air pollutants (PM2.5, NO2, O3). The model demonstrated good discrimination (AUC 0.77, 0.77-0.78). In the SSN subgroup, new LASSO-selected features positively associated with EGFR positivity included right middle lobe location and lobulated margins on CT and SO2 exposure. Cystic morphology, reticulation, and a new or growing solid component on CT were inversely associated with EGFR positivity. Conclusions: In a robust genomically linked and clinically, radiologically, and environmentally enriched dataset, we identified distinct demographic characteristics (female sex, Asian race, family history of LC), CT features (SSN density, multiplicity, lobulated margins), and environmental exposures (PM2.5, NO2, O3, SO2) that may predict EGFR positivity. Moreover, three-fold fewer adults with EGFRp AC met USPSTF LCS criteria. These findings support integrating imaging and non-tobacco risk factors to inform lung AC risk stratification, genomic testing, and more inclusive screening strategies.

Mark Helpful

Bookmark

Relay