Weighted clinical-molecular prognostic score for advanced KRAS mutated lung adenocarcinoma

Background: Stage IV lung adenocarcinoma (LUAD) remains heterogeneous in prognosis despite advances in molecular profiling and targeted therapy. Identifying robust prognostic factors and integrating them into a clinically applicable risk model may support individualized treatment strategies. Methods: We conducted a retrospective analysis of 1,013 patients with stage IV LUAD treated at Evangelische Lungenklinik Berlin between 2015 and 2023. Patients who had received targeted therapies (n=169) other than Sotorasib for KRAS G12C, had unknown KRAS status (n=177), or had undergone prior tumor resection (n=24) were excluded. Clinical variables (age at diagnosis, sex, ECOG performance status, smoking history, brain metastases, and first-line therapy), immunohistochemical markers (TTF-1 status, PD-L1 status), and molecular profile ( STK11 , KEAP1 , TP53 , and KRAS subtype: G12C, non-G12C, wild-type) were assessed. The final study population (n=643) was stratified into three groups according to KRAS mutational status: KRAS G12C, non-G12C KRAS , and KRAS wild-type. Survival analyses included Kaplan–Meier estimates, Cox regression, matched-pair analysis, and Bayesian network modeling to identify prognostic dependencies. Results: A weighted risk score was constructed from baseline clinical and molecular variables for non-G12C KRAS, the most representative group. Positive points were assigned for features associated with favorable prognosis (TTF-1 positivity, TP53 mutation, PD-L1 expression, female sex), and negative points for adverse features ( STK11 and KEAP1 mutation, ECOG>1). Patients were stratified into low-, intermediate-, and high-risk groups, demonstrating statistically significant differences in overall survival (p = 0.0128), with a median OS difference exceeding 12 months between the high- and low-risk groups. Analysis of treatment distribution across risk groups, even if limited from the small population of the subgroups, suggested potential for therapy optimization guided by the score. Conclusions: We developed a model integrating clinical and molecular features that separates patients into distinct survival groups and may support personalized prognostication and first-line treatment planning, addressing a key unmet need in advanced lung cancer care. Publication History Article published online: 04 March 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany