Abstract
In recent years, the therapeutic landscape of non-small cell lung cancer (NSCLC) has been transformed by immune checkpoint inhibitors (ICIs), which have led - in some patients - to unprecedented survival expectancy. Nevertheless, identifying patients most likely to benefit from ICI remains a major challenge. While PD-L1 expression and tumor mutation burden (TMB) represent established predictive biomarkers, their predictive ability still needs to be improved, which underscores the need for identifying additional (bio)markers for treatment selection. Recent research has highlighted multiple emerging biomarkers, including genomic alterations (eg, KEAP1, STK11, SMARCA4), markers of metabolic pathway dysregulation (IDO, adenosine axis), tumor-infiltrating lymphocytes, and blood-based biomarkers (eg soluble markers of inflammation, germline HLA diversity, and circulating tumor DNA). Host-related determinants, such as the history of tobacco exposure and the body mass index, further contribute to immunotherapy outcomes. In addition, artificial intelligence (AI) and machine learning (ML) approaches are enabling integration of multidimensional data, leading to predictive scoring systems which have outperformed conventional biomarkers in certain settings. This review synthesizes current evidence on established and emerging predictive biomarkers in NSCLC, highlighting the potential of combining biological, host, and computational features to inform precision immunotherapy strategies.