Published online Jun 26, 2026. doi: 10.4252/wjsc.119550
Revised: April 24, 2026
Accepted: May 27, 2026
Published online: June 26, 2026
Processing time: 91 Days and 23.4 Hours
An urgent clinical need exists to stratify postoperative prognosis in patients with non-small cell lung cancer (NSCLC). However, the prognostic value of the cancer stem cell (CSC) markers CD133, aldehyde dehydrogenase 1A1 (ALDH1A1), and SRY-box transcription factor 2 (SOX2) remains incompletely characterized. These markers are better viewed as complementary biomarkers for postoperative risk enrichment than as replacements for tumor-node-metastasis (TNM) staging or molecular classification.
To investigate expression levels of CD133, ALDH1A1 and SOX2, markers related to CSC, in NSCLC tissues for postoperative survival and prognosis.
A total of 200 patients with pathologically confirmed NSCLC who underwent radical resection at Jiangxi Provincial People’s Hospital (The First Affiliated Hospital of Nanchang Medical College) between January 2023 and December 2025 were included retrospectively. Expressions of CD133, ALDH1A1, and SOX2 in tumor tissues was detected by immunohistochemistry, and the tumors were divided into high- and low-expression groups according to the immunoreactive score. All patients were followed up until December 2025 to record their overall survival (OS) and disease-free survival (DFS). The t-test or χ2 test was used for comparison between groups. The Kaplan-Meier method and log-rank test were used for survival analysis. Univariate and multivariate analyses were performed using a Cox proportional risk model, and a prognostic model including CSC markers was constructed to evaluate the C-index and area under the curve of the receiver operating characteristic curve for 2-year OS. The immunoreactive score cut-off selection was based on prior literature and cohort distribution rather than on receiver operating characteristic derivation. DFS was analyzed as a conventional composite endpoint and model discrimination was internally corrected using bootstrap resampling.
Among 200 patients, the high expression rates of CD133, ALDH1A1, and SOX2 were 44.00% (88/200), 53.00% (106/200), and 40.00% (80/200), respectively. The median follow-up period was 23.50 months (interquartile range: 14.20-31.60 months), resulting in 54 deaths (27.00%) and 70 recurrences/metastases (35.00%). After adjusting for age, gender, smoking history, histological type, differentiation degree, TNM stage, and adjuvant therapy, high expression of CD133 [hazard ratio (HR) = 1.450, 95% confidence interval (CI): 1.033-2.037, P = 0.032], high expression of ALDH1A1 (HR = 1.380, 95%CI: 1.001-1.902, P = 0.049), and TNM stage III (HR = 1.980, 95%CI: 1.235-3.173, P = 0.004 compared to stage I) were independent adverse prognostic factors for OS. Patients with CSC score ≥ 2 had significantly shorter OS (P = 0.004), and this association remained significant in the multivariate model (HR = 1.550, 95%CI: 1.078-2.228, P = 0.018). After adding the CSC score, the predicted area under the curve value for 2-year OS increased from 0.675 to 0.785.
High expressions of CD133 and ALDH1A1 in NSCLC suggests a worse survival outcome. Nonetheless, the CSC score should be interpreted together with the TNM stage, histological background, and molecular features rather than used in isolation for clinical decision-making.
Core Tip: Cancer stem cells (CSCs) are key drivers of tumor recurrence, metastasis, and therapeutic resistance in non-small cell lung cancer. This retrospective study systematically evaluated the prognostic significance of the CSC-related markers CD133, aldehyde dehydrogenase 1A1, and SRY-box transcription factor 2 using immunohistochemistry in surgically resected non-small cell lung cancer tissues. We further developed a composite CSC score integrating multiple stemness markers, which significantly improved the prognostic stratification beyond conventional clinicopathological factors. Its potential clinical value lies in complementary postoperative risk enrichment, closer surveillance planning, and future integration with molecular subtyping, rather than the replacement of existing decision frameworks.