Clinical And Radiological Predictors of Pathological Response to Chemo-Immunotherapy in Surgically Resected Non-Small Cell Lung Cancer.
Authors
Affiliations (5)
Affiliations (5)
- Department of Cardiac, Thoracic Vascular Science and Public Health, Padova, Italy. Electronic address: [email protected].
- Department of Cardiac, Thoracic Vascular Science and Public Health, Padova, Italy.
- Unit of Biostatistics, Epidemiology and Public Health, University of Padova, Italy.
- Unit of Advanced Clinical and Translational Imaging, Department of Cardiac, Thoracic Vascular Sciences and Public Health - DCTV, University of Padova, Italy.
- Medical Oncology 2, Veneto Institute of Oncology IOV IRCCS, Padova, Italy.
Abstract
Chemo-immunotherapy has shifted the treatment paradigm for advanced non-small cell lung cancer (aNSCLC), improving complete pathological response (CPR) and radical resection rates. However, preoperative parameters to predict pathological response (PR) and tailor surgical strategy are lacking. This study aimed to develop an integrated clinico-radiological model to predict PR to neoadjuvant chemo-immunotherapy. Pre-treatment contrast chest-CT radiomic features were integrated with clinical variables and PET-CT metabolic parameters. PR was classified as CPR, major pathological response (MPR), or non-response (NR). A multistep machine learning approach using penalized multinomial regression and a random forest (RF) classifier was developed and assessed via internal cross-validation and multiclass AUC. We retrospectively evaluated 83 patients with aNSCLC undergoing neoadjuvant chemo-immunotherapy and surgery between 2017 and 2025 (18 excluded due to lack of contrast CT). Most patients (74%) were stage IIIA-IIIB; lobectomy was performed in 80%. Pathological examination revealed CPR in 41.5%, MPR in 26.2%, and NR in 32.3% of cases. The integrated clinical-radiomic RF model achieved good discrimination among PR categories, yielding a macro-averaged mean AUC of 0.82 (95% CI: 0.72-0.97). Radiomic texture features reflecting tumor heterogeneity emerged as the strongest predictors of PR. Preoperative prediction of PR after neoadjuvant chemo-immunotherapy is feasible using routinely available CT scans and clinical data. This integrated model provides a practical framework for surgical stratification, directly impacting patient selection and operative planning in aNSCLC.