Lung cancer detection using miRNA biomarkers among patients on lung nodule follow-up.

<bold>Background:</bold> Early detection of lung cancer is essential for better survival. Lung cancer screening with low-dose CT improves lung cancer mortality by 20%. However, the high false-positive rates necessitate the use of adjuncts. <bold>Aim:</bold> We evaluated the performance of miRNA biomarkers for lung cancer screening using machine learning tools. <bold>Methods:</bold> We performed a case-control study on 95 patients receiving LDCT for lung nodule follow-up. Expression levels of 16 miRNA biomarkers (identified through systematic review of existing literature) were obtained using reverse transcription of the RNA isolates. Random forest approach was used to rank the 16 miRNA biomarkers and multivariate logistic regression analysis was used to derive models with varying number of miRNA biomarkers. <bold>Results:</bold> Patients with malignant nodules were older, had greater smoking history and emphysema/COPD. Malignant nodules were larger, more numerous, and tended to be situated in the upper lobes, with spiculations/lobulations (Table 1). Adenocarcinoma was the most common histological subtype. All 16 biomarkers were preferentially expressed among the cases, and all but one reached statistical significance. Biomarker 5 had the best discriminative power among the 16 biomarkers (Figure 1). Our five-miRNA model (containing biomarkers 5, 3, 6, 13, 9 and 2) had the best sensitivity (100%) among all the models. <bold>Conclusion:</bold> Our five-miRNA model demonstrated high sensitivity for lung cancer diagnosis, which is essential for cancer screening. Liquid biopsy incorporating serum miRNA would be an invaluable adjunct for lung cancer screening.