[State Changes and Stability Grading of Driver Genes in Non-small Cell Lung Cancer Based on Repeated NGS Testing].
BACKGROUND: Next-generation sequencing (NGS)-based driver gene testing has become a routine component of molecular subtyping and precision therapy for non-small cell lung cancer (NSCLC). Dynamic genomic monitoring facilitates early detection of resistance-related molecular alterations and informs timely therapeutic adjustments. However, standardized criteria for evaluating the stability of serial NGS testing are currently lacking, and the applicability of NGS using formalin-fixed paraffin-embedded (FFPE) specimens for dynamic monitoring remains poorly defined. This study aims to establish a stability grading system for driver gene status alterations based on repeated NGS testing, and to provide evidence-based support for clinical repeat biopsy strategies. METHODS: Data from 1232 patients with NSCLC who underwent two or more NGS tests on FFPE tissue specimens at Beijing Chest Hospital between June 2019 and April 2026 were collected retrospectively. Patients with an interval of ≥4 months between the initial and last tests were included to ensure the representativeness of temporal analysis, resulting in a main analysis cohort of 942 patients. The Kappa consistency test was used to evaluate the state stability of nine core driver genes [epidermal growth factor receptor (EGFR), Kirsten rat sarcoma viral oncogene homolog (KRAS), anaplastic lymphoma kinase (ALK), ROS proto-oncogene 1, receptor tyrosine kinase (ROS1), mesenchymal‑epithelial transition factor (MET), rearranged during transfection (RET), v-raf murine sarcoma viral oncogene homolog B1 (BRAF), erb‑b2 receptor tyrosine kinase 2 (ERBB2), and phosphatidylinositol‑4,5‑bisphosphate 3‑kinase catalytic subunit alpha (PIK3CA)] and to construct a five‑level grading system. Paired variant allele frequency (VAF) differences were compared using the Wilcoxon signed‑rank test. Independent influencing factors for mutation accumulation were identified by binary Logistic regression. RESULTS: The state stability of the nine genes was classified into five levels: EGFR showed high stability (Kappa=0.838), ROS1/ALK/KRAS good stability, BRAF/PIK3CA/RET moderate stability, and ERBB2 low stability, and MET showed high instability. MET exhibited the highest rate of state change (9.3%) with a raw observed agreement of 90.7%. Its Kappa value (0.172) was influenced by the low prevalence (3.7%) compression effect and should therefore be interpreted alongside the observed agreement (90.7%) and the prevalence-adjusted and bias-adjusted Kappa (PABAK). The VAF of PIK3CA increased significantly (P=0.005). T790M positivity increased from 5.8% to 10.8%, and 30 new C797S mutations were detected at the last test (13 with T790M, 17 without). The overall rate of new driver gene variants in the main cohort was 18.0%. Binary Logistic regression showed that a lower number of initial mutated genes was the only independent predictor of new variants [odds ratio (OR)=0.399, P<0.001], while sex and detection interval showed no independent association. CONCLUSIONS: A five level stability grading system for state changes of driver genes in NSCLC based on repeated NGS testing has been established. MET showed the most frequent state changes, which should be interpreted in conjunction with the prevalence effect. The VAF increase of PIK3CA is an observational finding, and its clinical significance requires further prospective validation. A lower initial mutation burden may reflect tumor clonal complexity and was associated with a higher likelihood of subsequent acquisition of new variants. FFPE based NGS is applicable for repeated testing at clinical treatment decision nodes.