Pooled Analysis of the Prognostic Significance of Epidermal Growth Factor Receptor (EGFR) Mutational Status in Combination with Other Driver Genomic Alterations in Stage I Resected Invasive Lung Adenocarcinoma for Recurrence-Free Survival: A Population-Based Study

• Published in: Annals of surgical oncology Vol. 32; no. 2; pp. 760 - 770
• Main Authors: Huang, Yufei, Zeng, Hui, Zhang, Guochao, Ren, Fangzhou, Yuan, Zhenlong, Ren, Jingyu, Xu, Jiaxi, Song, Zehao, Li, Wenbin, Ying, Jianming, Feng, Feiyue, Tan, Fengwei
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 01.02.2025; Springer Nature B.V
• Subjects: Adenocarcinoma; Cancer; Epidermal growth factor; Epidermal growth factor receptors; Gene deletion; Genomic analysis; Genotypes; Invasiveness; Lung cancer; Medical prognosis; Medicine; Medicine & Public Health; Mutation; Oncology; p53 Protein; Population studies; Population-based studies; Prediction models; Regression analysis; Surgery; Surgical Oncology; Survival; Survival analysis; Translational Research; Recurrence; Common driver mutations; Lung adenocarcinoma; EGFR; Machine learning
• ISSN: 1068-9265; 1534-4681; 1534-4681
• DOI: 10.1245/s10434-024-16528-7

Background The prognostic significance of epidermal growth factor receptor (EGFR) mutations in stage I invasive lung adenocarcinoma (LUAD) remains debated. Improving the lung cancer staging system requires further investigation into actionable mutations and their association with survival outcomes. Patients and Methods A total of 410 patients with stage I invasive LUAD were analyzed for their driver mutations. Survival analysis of EGFR mutations, exon 19 deletion, L858R in exon 21, and minor genotypes were stratified by clinicopathologic characteristics. Kaplan–Meier and log-rank tests were used to determine prognostic significance. Univariate and multivariate Cox proportional hazard regression models assessed variables’ impact on recurrence-free survival (RFS). Patients with further-profiled samples were divided into training and validation datasets by computer-generated random numbers. Multiple machine learning algorithms were applied to construct genomic prediction models, with C index evaluated for each. Results EGFR mutations occurred in 210 patients (51.2%). In stage I invasive LUAD, EGFR mutations strongly correlated with poor RFS ( P = 0.022), especially in never smoker ( P < 0.001), female ( P = 0.024), part-solid ( P = 0.002), and stage IA subgroups ( P = 0.020). The most frequently co-mutated gene was TP53. Moreover, patients with EGFR/TP53 co-mutations, regardless of mutant types, exhibited worse prognosis. A mutational prognostic model based on the random survival forest (RSF) algorithm achieved the highest mean C index (C index: 0.87 in training cohort versus 0.74 in validation cohort), and demonstrated strong RFS estimation performance [area under the curve (AUC):1-year, 0.87, versus 3-year, 0.92, versus 5-year, 0.92]. Conclusions EGFR mutations are robust biomarkers for RFS estimation in stage I invasive LUAD. Combining EGFR mutations with other actionable mutations enhances individualized RFS estimation.