A ppropriate staging of lung cancer is critical, as it predicts prognosis and dictates treatment. Radiographic staging with CT scan and PET scan can offer clues to the extent of disease, but pathologic confi rmation of malignancy and determination of the TNM stage for non-small cell lung cancer (NSCLC) dictates the treatment choice. 1 In practice, mediastinal lymph node involvement most often differentiates those who are surgical candidates from those who are not. 2 The current methods available to adequately stage the mediastinum include mediastinoscopy, videoassisted thoracoscopy, endoscopic ultrasound (EUS), endobronchial ultrasound (EBUS), transthoracic needle aspiration, and transbronchial needle aspiration. The American College of Chest Physicians practice guidelines state that "tissue should be obtained by whatever method is easiest to perform" depending on the size and location of the lymph node, the availability of the technology, and expertise in the local facility. 1 EBUS and EUS have gained acceptance as dependable procedures to stage lung cancer with comparable accuracy to surgical methods. 1,[3][4][5] The use of ultrasound facilitates the direct visualization of the lymph node during biopsy and may offer information regarding nodal characteristics of malignant nodes.Purpose: Reliable staging of the mediastinum determines TNM classifi cation and directs therapy for non-small cell lung cancer (NSCLC). Our aim was to evaluate predictors of mediastinal lymph node metastasis in patients undergoing endobronchial ultrasound (EBUS). Methods: Patients with known or suspected lung cancer undergoing EBUS for staging were included. Lymph node radiographic characteristics on chest CT/PET scan and ultrasound characteristics of size, shape, border, echogenicity, and number were correlated with rapid on-site evaluation (ROSE) and fi nal pathology. Logistic regression (estimated with generalized estimating equations to account for correlation across nodes within patients) was used with cancer (vs normal pathology) as the outcome. ORs compare risks across groups, and testing was performed with two-sided a of 0.05. Results: Two hundred twenty-seven distinct lymph nodes (22.5% positive for malignancy) were evaluated in 100 patients. Lymph node size, by CT scan and EBUS measurements, and round and oval shape were predictive of mediastinal metastasis. Increasing size of lymph nodes on EBUS was associated with increasing malignancy risk ( P 5 .0002). When adjusted for CT scan size, hypermetabolic lymph nodes on PET scan did not predict malignancy. Echogenicity and border contour on EBUS and site of biopsy were not signifi cantly associated with cancer. In 94.8% of lymph nodes with a clear diagnosis, the ROSE of the fi rst pass correlated with subsequent passes. Conclusions: Lymph node size on CT scan and EBUS and round or oval shape by EBUS are predictors of malignancy, but no single characteristic can exclude a visualized lymph node from biopsy. Further, increasing the number of samples taken is unlikely to signifi cant...