Improving the screening ability of neuron-specific enolase on small cell lung cancer

• Published in: Lung cancer (Amsterdam, Netherlands) Vol. 199; p. 108078
• Main Authors: Wu, Yixian, Tang, Yingdan, Huang, Wen, Zhu, Chen, Ju, Huanyu, Wu, Juan, Zhang, Qun, Zhao, Yang, Kong, Hui
• Format: Journal Article
• Language: English
• Published: Ireland Elsevier B.V 01.01.2025
• Subjects: Adult; Aged; Biomarkers, Tumor - blood; Correction; Early Detection of Cancer - methods; Female; Humans; Lung Neoplasms - diagnosis; Machine learning; Male; Middle Aged; Neuron-specific enolase; Phosphopyruvate Hydratase - blood; Phosphopyruvate Hydratase - metabolism; ROC Curve; Sensitivity and Specificity; Small cell lung cancer; Small Cell Lung Carcinoma - diagnosis; Small cell lung cancer; Correction; Machine learning; Neuron-specific enolase
• ISSN: 0169-5002; 1872-8332; 1872-8332
• DOI: 10.1016/j.lungcan.2024.108078

[Display omitted] •In this work, a total of 112,562 healthy individuals and 105 SCLC patients were recruited.•We developed an NSE correction model for metabolic factors, using residual correction and machine learning techniques.•Compared to uncorrected NSE, NSEcorrected significantly increased the specificity and reduced the false positive rate.•We developed an R-based Shiny application to calculate NSEcorrected and made it publicly available online. Neuron-specific enolase (NSE) is one of the most common biomarkers of small cell lung cancer (SCLC) and is widely used in lung cancer screening. But its specificity is affected by many factors. Using residual correction and machine learning, corrected NSE and its reference range were constructed based on metabolic factors and smoking history affecting NSE in the training set of 48,009 healthy individuals recruited from the First Affiliated Hospital of Nanjing Medical University. External validation including additional 64,553 healthy subjects and 105 SCLC patients were enrolled to evaluate the efficacy of NSEcorrected for SCLC screening. The reference range of NSEcorrected could significantly improve the specificity of NSE for SCLC and reduce false positives. In the external validation set, NSEcorrected increased the specificity from 85.71 % to 97.09 %(P < 0.0001), and reduced the false positive rate from 14.26 % to 2.91 %(P < 0.0001). ROC curve, calibration curve and decision analysis curve also showed that NSEcorrected had better screening performance. The calculation of NSEcorrected was converted into an online R-based app for more convenient use. NSEcorrected can improve the screening effect of SCLC, reduce the false positive rate, and is more suitable for large population screening and optimize the allocation of lung cancer resources.