Quantitative analysis of a novel DNA hypermethylation panel using bronchial specimen for lung cancer diagnosis.

ABSTRACT

Non-diagnostic findings are common in transbronchial lung biopsy (TBLB) and endobronchial ultrasound-guided transbronchial lung biopsy (EBUS-TBLB). One of the challenges is to improve the detection of lung cancer using these techniques. To address this issue, we utilized an 850 K methylation chip to identify methylation sites that distinguish malignant from benign lung nodules. Our study found that a combination of HOXA7, SHOX2 and SCT methylation analysis has the best diagnostic yield in bronchial washing (sensitivity 74.1%; AUC 0.851) and brushing samples (sensitivity 86.1%; AUC 0.915). We developed a kit comprising these three genes and validated it in 329 unique bronchial washing samples, 397 unique brushing samples and 179 unique patients with both washing and brushing samples. The panel's accuracy in lung cancer diagnosis was 86.9%, 91.2% and 95% in bronchial washing, brushing and washing + brushing samples, respectively. When combined with cytology, rapid on-site evaluation (ROSE), and histology, the panel's sensitivity in lung cancer diagnosis was 90.8% and 95.8% in bronchial washing and brushing samples, respectively, and 100% in washing + brushing samples. Our findings suggest that quantitative analysis of the three-gene panel can improve the diagnosis of lung cancer using bronchoscopy.