LAUSR

BACKGROUND AND OBJECTIVE Although stage I non-small cell lung carcinoma (NSCLC) typically carries a good prognosis following complete resection, early disease recurrence can occur. An accurate survival prediction model would help refine a follow-up strategy and personalize future adjuvant therapy. We developed a post-operative prediction model based on readily available clinical information for patients with stage I adenocarcinoma. METHODS We retrospectively studied the disease-free survival (DFS) of 408 patients with pathologically confirmed low-risk stage I adenocarcinoma of lung who underwent curative resection from 2013 to 2017. A tree-based method was employed to partition the cohort into subgroups with distinct DFS outcome and stepwise risk ratio. These covariates were included in multivariate analysis to build a scoring system to predict disease recurrence. The model was subsequently validated using a 2011-2012 cohort. RESULTS Non-smoker status, stage IA disease, epidermal-growth factor receptor mutants and female gender were associated with better DFS. Multivariate analysis identified smoking status, disease stage and gender as factors necessary for the scoring system and yielded 3 distinct risk groups for DFS [99.4 (95% CI 78.3-125.3), 62.9 (95% CI 48.2-82.0), 33.7 (95% CI 24.6-46.1) months, p < 0.005]. External validation yielded an area under the curve by receiver operating characteristic analysis of 0.863 (95% CI 0.755-0.972). CONCLUSION The model could categorize post-operative patients using readily available clinical information, and may help personalize a follow-up strategy and future adjuvant therapy.