Background: Multi-institutional studies of endobronchial-ultrasound transbronchial needle aspiration (EBUS-TBNA) for mediastinal staging in lung cancer are scarce. It is unclear if the high diagnostic performance of EBUS-TBNA reported by experts' guidelines can be generally achieved. Methods: This is a retrospective study performed in five tertiary referral centers of thoracic surgery in Italy, to assess the EBUS-TBNA diagnostic performance in patients with non-small cell lung cancer (NSCLC). Patient inclusion criteria were: both genders; >18 years old; with suspect/confirmed NSCLC; undergoing EBUS-TBNA for mediastinal node enlargement at computed tomography (size >1 cm, ≤3 cm) and/or pathological uptake at positron emission tomography. Altogether we included 485 patients [male, 366; female, 119; median age, 68 years (IQR, 61-74 years)] undergoing mediastinal staging between January 2011 and July 2016.All EBUS-TBNAs were performed by experienced bronchoscopists, without pre-defined quality standards.Depending on usual practice in each center, EBUS-TBNA was done under conscious sedation, with 21-or 22-Gauge (G) needle, and specimen preparation was cell-block, or cytology slides, or core-tissue. Sampling was classified inadequate in absence of lymphocytes, or when sample was insufficient. We analyzed the EBUS-TBNA procedural steps likely to influence the rate of adequate samplings (diagnostic yield).Results: EBUS-TBNA sensitivity, negative predictive value (NPV) and accuracy respectively were 90%, 78% and 93% in the whole cohort. At multivariate analysis, use of 21-G needle was associated with better diagnostic yield (P<0.001). Center and specimen processing technique were not independent factors affecting EBUS-TBNA diagnostic yield. Conclusions: In this multicentric study, EBUS-TBNA was a highly sensitive and accurate method for NSCLC mediastinal node staging. Results indicate better performance of EBUS-TBNA with 21-G needle, and suggest that specimen processing technique could be chosen according to the local practice preference.