Clinic-radiologic predictors of pathological characteristics in pure ground-glass nodules: Development and validation of a predictive nomogram.

Abstract

The objective of this study was to explore the relationship between the clinic-radiologic characteristics of pure ground-glass nodules (pGGNs) and key pathologic features of lung adenocarcinoma, including tumor invasiveness, proliferative activity, driver oncogenic mutations, and the tumor immune microenvironment. A total of 1070 surgically resected pGGNs were retrospectively analyzed and categorized pathologically into 224 precursor glandular lesions (PGLs), 600 minimally invasive adenocarcinomas (MIAs), and 246 invasive adenocarcinomas (IAs). Receiver operating characteristic curves and area under the curve were used to assess the diagnostic performance of various size and density parameters. After adjusting for clinical covariates, computed tomography (CT) features were compared across the different pathologic subgroups. The relationships between clinic-radiologic characteristics and the Ki-67 proliferation index, EGFR mutation status, and tertiary lymphoid structures (TLS) were examined. Finally, a clinic-radiologic nomogram was developed and externally validated to enable accurate preoperative prediction of IA presenting as pGGNs. Receiver operating characteristic analysis revealed that density-related parameters of pGGNs had strong discriminatory power for both MIA and IA, with the lung/max CT value showing the greatest diagnostic performance, followed by the max CT value. The optimal cut-off values for differentiating MIA from PGL were 1.62 (lung/max CT value), -548 HU (max CT value), and 8.6 mm (max diameter), whereas for distinguishing IA from MIA, the respective thresholds were 1.89, -458 HU, and 11.5 mm. After adjusting for clinical covariates, CT features such as size, density, shape, border, and voxel heterogeneity remained significantly different across the PGL, MIA, and IA groups. Furthermore, a positive correlation was found between max CT value and pathologic indicators, including Ki-67 proliferation index, EGFR mutation, and TLS. A clinic-radiologic nomogram that incorporated age, max diameter, lung/max CT value, and shape demonstrated excellent discriminatory ability for IA (area under the curve, 0.883-0.920), along with good calibration and clinical utility. Size and density, as key radiologic features in the assessment of pGGNs, proved to be reliable predictors of pathologic invasiveness, proliferative activity, EGFR mutation status, and TLS levels. A clinic-radiologic nomogram for noninvasive prediction of IA offered a valuable tool to guide surveillance strategies and therapeutic decisions for patients with pGGNs.