Development and internal validation of a serum biomarker signature for non-small cell lung cancer: an exploratory case–control study

Abstract

Blood-based biomarkers are increasingly explored for their ability to distinguish non-small cell lung cancer (NSCLC) from non-malignant conditions. Single markers, however, often show limited performance, which has led to growing interest in combining multiple biological features. Aim. To develop a multivariable diagnostic model based on serum biomarkers and assess its discriminatory performance in an exploratory TRIPOD Type 1a study. Materials and methods. This study represents a secondary analysis of a previously published case–control cohort, focusing on multivariable model development rather than single biomarker assessment. A total of 275 participants were enrolled in this retrospective case–control study supplemented by prospectively collected samples, including 85 patients with histologically confirmed NSCLC and 190 controls. Serum concentrations of 20 biomarkers representing tumor-associated, inflammatory, and metabolic pathways were analyzed. Univariable logistic regression was performed, followed by multivariable modeling including variables with p < 0.10. Model discrimination was evaluated using receiver operating characteristic analysis and the area under the curve (AUC), and internal validation was performed by bootstrap resampling on complete-case data. Results. Three independent variables were retained in the final multivariable model: human epididymis protein 4 (HE4), apolipoprotein A4 (ApoA4), and age. The final model showed high apparent discrimination within the study dataset (AUC = 0.975; 95% CI: 0.958–0.992). Bootstrap internal validation demonstrated minimal optimism, with an optimism-corrected AUC of 0.972. HE4 and age were positively associated with NSCLC, whereas ApoA4 showed an inverse association. In the age-matched sensitivity analysis, HE4 and ApoA4 retained essentially unchanged associations while age lost independent significance, supporting a biomarker-driven rather than age-driven contribution to model discrimination. The multivariable model demonstrated improved discrimination compared with individual biomarkers. Conclusion. This exploratory model derivation study identified a candidate serum biomarker combination associated with NSCLC. The findings should be considered hypothesis-generating. Independent validation in clinically representative populations is required before any clinical interpretation or application.