Despite known limitations of positron emission tomography (PET) for mediastinal staging of non-small cell lung cancer (NSCLC), radiation treatment fields are generally based on PET-identified disease extent. However, no studies have examined the accuracy of FDG-PET/CT on a per-node basis in patients being considered for curative-intent radiotherapy in NSCLC.In a prospective trial, patients with NSCLC being considered for definitive thoracic radiotherapy (± systemic chemotherapy) underwent minimally invasive systematic mediastinal evaluation with endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) following noninvasive staging with integrated PET-CT.Thirty patients underwent EBUS-TBNA, with TBNA performed from a mean 2.5 lymph node (LN) stations per patient (median 3, range 1-5). Discordant findings between PET-CT and EBUS-TBNA were observed in 10 patients (33%, 95% CI 19%-51%). PET-occult LN metastases were demonstrated by EBUS in 4 patients, whereas a lesser extent of mediastinal involvement, compared with FDG-PET, was demonstrated by EBUS in 6 patients, including 2 patients downstaged from cN3 to pN2. LNs upstaged by EBUS were significantly smaller than nodes downstaged by EBUS, 7.5 mm (range 7-9) versus 12 mm (range 6-21), P = 0.005.A significant proportion of patients considered for definitive radiotherapy (+/-chemotherapy) undergoing systematic mediastinal evaluation with EBUS-TBNA in this study have an extent of mediastinal NSCLC involvement discordant with that indicated by PET-CT. Systematic EBUS-TBNA may aid in defining the extent of mediastinal involvement in NSCLC patients undergoing radiotherapy. Systematic EBUS-TBNA has the potential to contribute significantly to radiotherapy planning and delivery, by either identifying occult nodal metastases, or demonstrating FDG-avid LNs to be disease-free.