Prognostic stratification using baseline objective clinical information in patients with stage IV lung cancer receiving best supportive care alone as initial treatment: A multicenter retrospective study (WJOG20824L).

12054 Background: Despite expanded systemic therapy for lung cancer, best supportive care (BSC) alone remains the initial treatment for some patients, often due to advanced age, poor performance status (PS), or comorbidities. However, large-scale data on prognostic heterogeneity and risk stratification in this population are limited. We aimed to stratify overall survival (OS) in stage IV lung cancer patients initially managed with BSC alone using baseline clinical information. Methods: We analyzed a multicenter database from WJOG15121L including 6,751 patients with stage IV lung cancer diagnosed between January 2016 and December 2019, identifying those who received BSC alone as initial treatment. Sixty-two baseline variables (demographics, comorbidities, genomic alterations, and laboratory data) were assessed. The primary analysis was conducted in complete cases (BSC complete cohort), with multiple imputation (MI) in the entire cohort as sensitivity analysis. A boosting-based model was used to derive a prognostic risk score for OS and to stratify patients into low-, medium-, and high-risk groups based on the risk score distribution. Survival outcomes were summarized using Kaplan–Meier methods with log-rank tests. Cox regression with Lasso regularization (Cox-Lasso) was also performed as an alternative approach to assess robustness. Results: A total of 923 patients received BSC alone as initial treatment; 266 comprised the BSC complete cohort. The BSC population was elderly (mean age: 75.8 years) with a high proportion of poor PS (ECOG PS ≥2: 74.8%). Kaplan–Meier–based risk stratification separated OS (median OS: 277 vs 74 vs 25 days for low-, medium-, and high-risk groups, respectively). The Kaplan–Meier–estimated 90-day OS rates were clearly separated across the three groups (75.4% vs 29.8% vs 0%; log-rank p < 0.001). Baseline characteristics differed by risk group, including PS and metastatic burden (e.g., liver metastasis). Risk stratification was largely driven by routinely available laboratory markers (e.g., LDH, neutrophil count, CRP, ALP, platelet count). Findings were consistent in the MI-based sensitivity analysis and were broadly supported by the Cox-Lasso approach. Conclusions: Among patients with stage IV lung cancer initially managed with BSC alone, outcomes are heterogeneous and can be stratified into clinically meaningful risk groups using baseline clinical information. Kaplan–Meier summaries (median OS and 90-day OS) demonstrate marked differences across risk strata, and the dominant contribution of routine laboratory variables highlights the potential practicality of this approach in real-world settings. Further validation and comparative analyses with patients receiving systemic therapy who share similar risk profiles are ongoing.