Robust prognostic biomarkers and a risk prediction model specific to EGFR-mutated lung adenocarcinoma patients via multicohort meta-analysis

Background: Epidermal growth factor receptor (EGFR) mutation is a key therapeutic target and clinical predictive biomarker for lung adenocarcinoma. However, few prognostic biomarkers and risk prediction models for EGFR-mutated patients exist due to small sample sizes and heterogeneous cancer mechanisms across patient population. We aimed to develop reliable prognostic markers and a risk prediction model specific to EGFR-mutated patients through integrative analysis of multicohort data. Methods: We analyzed transcriptome profiles from 1495 patients across six public datasets, dividing them into EGFR-mutated and non-EGFR-mutated groups. We examined hazard ratios and concordance indices using the univariate Cox proportional hazards model to clarify different prognostic marker characteristics between each group. We then identified significant prognostic genes for each group for subsequent analyses. Throughout the marker derivation process, we applied a multicohort meta-analysis to draw robust conclusions. We performed functional enrichment analyses to verify functional specificity of the markers for each patient group. Furthermore, we established a risk scoring model through optimal selection of significant genes for EGFR-mutated patients and compared its performance with the models for other groups and previously proposed markers. Results: Compared to other groups, significant prognostic genes identified in the EGFR-mutated patients demonstrated lower hazard ratios, higher concordance indices (p < 2.2E-16), and greater functional homogeneity. Our proposed 32-gene prognostic model for EGFR-mutated patients outperformed other models in cross-validation (0.810 vs. 0.641) and independent cohort validation (0.711 vs. 0.660) in risk prediction using the concordance index. The identified markers were significantly associated with functions related to the G2M checkpoint and E2F targets, which aligns with known EGFR-mutated patient-derived evidences. Conclusion: We identified prognostic biomarkers specific to EGFR-mutated patients and developed a gene expression-based risk scoring system for further risk stratification. Our findings may contribute to developing more refined patient-specific treatment strategies and improving survival rates.