Pathologic assessment of resected stage III non-small cell lung cancer after neoadjuvant chemotherapy: identification of additional prognostic factors

Abstract

Background: Non-small cell lung cancer (NSCLC) patients undergoing neoadjuvant chemotherapy (NACT) followed by surgery represent an ideal clinical setting to identify prognostic factors. To date, major pathological response (MPR) and complete pathological response (pCR) have been used as surrogates of NACT response and clinical outcome. The aim of the study was to investigate the role of additional clinico-pathological features, taking advantage of morphometry and artificial intelligence (AI). Methods: Seventy stage III NSCLC patients undergoing surgery after NACT were studied. A granular evaluation of histological parameters with morphometrical quantification of the stromal components (fibrosis/inflammation) in addition to the tumour bed analysis (2020 IASLC statement) was carried out in all cases. An AI algorithm of the different immunophenotypes was also applied on immunohistochemistry-stained whole-slide images. A ClinPATH combined score including MPR, baseline blood lymphocytes, perineural invasion, vascular invasion, proliferative index, fibrosis extension percentage and AI-quantified CD4+ cell % was tested. Results: MPR and pCR were related to disease-free survival (DFS) and overall survival (OS) but also vascular/perineural/pleural invasion and Ki-67 were useful in stratifying the study population. Concerning the tumour bed stromal components, only morphometrical quantification highlighted the prognostic role of fibrosis and inflammation, particularly when distinguishing CD4+ and FOXP3+ cells, mainly in adenocarcinomas. Interestingly, the combination of the most impactful clinico-pathological parameters in a ClinPATH combined score correlated better with DFS and OS than any individual parameter, including MPR or pCR. Conclusion: AI-based method can be used to accurately decipher the complexity of tumour bed stromal components, providing extra information for outcome prediction. The combination of different clinico-pathological features could be highly valuable in guiding therapeutic decisions and ultimately improve patient outcomes.