[en] Purpose: Mediastinoscopy is routinely carried out on the majority of nonmetastatic, non-small-cell lung cancer (NSCLC) patients in our institution. We used the results of mediastinoscopy from a Stage III NSCLC cohort to assess the reliability of positron emission tomography (PET) scans at identifying involved mediastinal lymph nodes (MLN) when used during radiotherapy planning. Methods and Materials: Mediastinoscopy was the gold standard. Characteristics of PET were calculated for nodal sensitivity. To compare the impact on contouring, theoretical nodal targets (NTs) containing involved MLNs were generated using PET and mediastinoscopy. We determined whether the NT derived from PET (NT-P) was equivalent to, greater than, or less than that seen with the mediastinoscopy (NT-M). Results: Data for 122 patients with Stage III NSCLC, treated between 2000 and 2004, were analyzed. After exclusions, 87 patients with Stage III disease by mediastinoscopy were analyzed. Overall PET sensitivity was 61% and positive predictive value was 94%. Of the 87 patients, 33 (38%) had no abnormal MLN findings by PET. Of 36 Stage IIIA cancer patients, 18 (50%) had NT-P equivalent to NT-M, 10 (28%) had smaller NT-Ps, and 8 (22%) had larger NT-Ps compared with NT-Ms. Of 18 Stage IIIB cancer patients, NTs were equivalent in 6 (34%); in 1 patient (5%) NT-P was larger than the corresponding NT-M, and in 11 (61%) smaller than the corresponding NT-M. Conclusions: In this study PET had modest sensitivity to detect MLN involvement and underestimated the extent of involved nodes for target definition. The role of PET in mediastinal contouring needs to be evaluated prospectively and ideally correlated with a pathology standard