Résumé : In the staging of lung cancer with positron emission tomography (PET) positive mediastinal lymph nodes, tissue sampling is required. The performance of transbronchial needle aspiration (TBNA) using linear endobronchial ultrasound (real-time EBUS-TBNA) under local anaesthesia and the value of PET for prediction of pathological results were assessed in that setting. The number of eluded surgical procedures was evaluated. All consecutive patients with suspected/proven lung cancers and FDG-PET positive mediastinal adenopathy were included. If no diagnosis was reached, further surgical sampling was required. Lymph node SUVmax (maximum standardized uptake value) was assessed in patients whose PET was performed in the leading centre. One hundred and six patients were included. The average number of TBNA samples per patient was 4.9 ± 1.1. The prevalence of lymph node metastasis was 58%. Sensitivity, accuracy and negative predictive value of EBUS-TBNA in the staging of mediastinal hot spots were 95, 97 and 91%. Patients without malignant lymph node involvement showed lower SUVmax (respective median values of 3.7 and 10.0; p < 0.0001). Surgical procedures were eluded in 56% of the patients. Real-time EBUS-TBNA should be preferred over mediastinoscopy as the first step procedure in the staging of PET mediastinal hot spots in lung cancer patients. In case of negative EBUS, surgical staging procedure should be undertaken. The addition of SUVmax cut-off may allow further refinement but needs validation. © 2008 Elsevier Ireland Ltd. All rights reserved.