Establishing mediastinal spread of non-small cell lung cancer (N2-3 disease) precludes surgery and worsens prognosis; whether PET-CT imaging can improve overall accuracy or safely prevent mediastinoscopies is still unknown. Fischer et al re-heat the data from their 2009 NEJM randomized trial in Denmark, with EUS-FNA and mediastinoscopy on 189 NSCLC patients, in which they concluded that the half who first got PET-CTs had fewer futile thoracotomies. Here, they report: 1) PET-CT improved diagnostic accuracy, mainly in peripheral tumors, with the gold standard of consensus N staging; 2) Size of lymph nodes on imaging was extremely important: For mediastinal LNs >=10 mm, PET had a 15% false-negative rate, but for normal-sized LNs, the rate was only 4%. They recommend A) PET-CT for all new diagnoses of NSCLC; B) straight-to-resection for people with normal-sized mediastinal LNs that are PET-negative; C) enlarged LNs need invasive sampling regardless of PET uptake, as do PET-positive LNs. Accuracy of mediastinal sampling techniques (mediastinoscopy / EUS / EBUS FNA) vary by local availability and expertise. Thorax 2011;66:294-300.