Plasma Long Non-coding RNAs for Molecular Diagnosis of Colon Cancer: A Validated Multivariable Model

Circulating long non-coding RNAs (lncRNAs) are emerging as minimally invasive biomarkers for cancer detection; however, their clinical utility remains insufficiently validated. We evaluated the diagnostic performance of plasma lncRNAs and developed and externally validated multivariable models for detecting non-metastatic colon cancer.

Plasma lncRNA expression was quantified using reverse transcription quantitative PCR (RT-qPCR) in a training cohort of patients with histopathologically confirmed non-metastatic colon cancer and healthy controls. Multivariable logistic regression models integrating molecular and demographic variables were developed in the training cohort and externally validated in an independent cohort.

Nine candidate lncRNAs were evaluated, of which five (CCAT2, HOTAIR, LINC00909, PVT1, and UCA1) were consistently detectable in plasma and significantly upregulated in patients with colon cancer. A multivariable model combining UCA1, HOTAIR, and PVT1 with age and sex achieved an area under the ROC curve (AUC) of 0.97 in the training cohort and of 0.87 in the validation cohort. Internal bootstrap validation showed limited optimism (optimism-corrected AUC, 0.95), and calibration analysis demonstrated good agreement between predicted and observed probabilities. At a sensitivity-oriented threshold, the model reached 90% sensitivity and 69% specificity. A simplified model based on UCA1 retained good discrimination but with reduced overall performance.

Plasma lncRNA-based multivariable models show promise as minimally invasive tools for the early detection of colon cancer and highlight the importance of external validation in biomarker development.