Multi-Institutional Integration of Circulating miRNAs and Protein Tumor Markers for Early Lung Cancer Detection

Early detection of lung cancer is essential for improving survival rates. However, previously reported diagnostic biomarkers based on circulating microRNAs (cfmiRNAs) are often hindered by single-institutional biases and a lack of reproducibility. This study aimed to identify and verify a robust cfmiRNA-based diagnostic signature integrated with protein tumor markers using a large-scale, multi-institutional cohort.

A multi-institutional observational study was conducted across 11 hospitals in Taiwan between 2024 and 2025. Plasma samples were collected from 752 participants, including 255 preoperative patients with lung cancer and 497 healthy controls. To mitigate site-specific effects and confounding factors, candidate cfmiRNAs were identified through stability selection with Monte Carlo sampling. A multiomics classification model was developed by integrating the miRNA panel with established protein markers (CEA, CA19-9, SCC, and Cyfra21-1).

A 15-miRNA diagnostic biosignature, functionally linked to tumor progression and cellular stress-related pathways, was identified. The final integrated multiomics model achieved an area under the ROC curve of 84% (95% confidence interval: 76-91) in an independent verification data set. At a fixed specificity of 80%, the cancer detection model reported a sensitivity of 73%. The multi-institutional approach provided a more generalized assessment of biomarker performance than localized cohorts.

Our findings demonstrate the potential of multiomics integration in overcoming institution-specific biases and provide the necessary groundwork for lung cancer detection in an Asian population.