Imaging biomarkers related to tumor-associated macrophage in immunotherapy treatment planning for non-small cell lung cancer

Background: Recent studies have highlighted the critical role of lymph nodes in the anti-tumor immune response. To identify noninvasive biomarkers for selecting optimal candidates for immunotherapy, we hypothesized that quantitative information extracted from lymph nodes via computed tomography (CT) radiomics analysis could provide supplementary predictive value. We further investigated the relationship between the radiomics model and tumor-associated macrophages (TAMs) to elucidate the biological basis of the model. Methods: We analyzed 1,070 primary and lymph node lesions from 212 patients with non-small cell lung cancer (NSCLC) receiving anti-programmed cell death 1 (PD-1) therapy. Pre-treatment contrast-enhanced CT images were divided into training, test, and external test cohorts to derive radiomic signatures from the whole tumor, peritumoral regions, and lymph nodes. These signatures were then evaluated for their association with durable clinical benefit (DCB), objective response, pseudoprogression, progression-free survival (PFS), and overall survival (OS). Finally, the biological relevance of the radiomic signatures was assessed using immunohistochemical data and multivariate linear regression. Results: The established radiomic signature achieved sensitivities of 84.48%, 85%, and 92.31% in identifying patients with DCB across the three cohorts, with nearly consistent specificities. It accurately identified 28.57% of programmed cell death ligand 1 (PD-L1)-negative patients who derived clinical benefit from immunotherapy. Additionally, the signature distinguished partial/complete response (PR/CR) with area under the receiver operating characteristic curve (AUC) values of 0.67, 0.72, and 0.71 across the cohorts. It also predicted PFS (C-indices: 0.68, 0.66, and 0.64) and OS (C-indices: 0.61, 0.65, and 0.62), respectively. Furthermore, the signature demonstrated an AUC of 0.80 in differentiating pseudoprogression from hyperprogression, aiding treatment decision-making. Notably, it was independently associated with M2-like TAMs (P=0.002). Conclusions: By integrating lymph node-related quantitative information, we developed a radiomics signature linked to treatment response and clinical outcomes. This signature not only enhances the identification of optimal immunotherapy candidates with higher sensitivity prior to treatment initiation but also supports treatment strategy adjustments following therapy onset.