[en] Purpose: To assess the pattern of local failure using ¹⁸F-fluorodeoxyglucose (FDG)-positron emission tomography (PET) scans after radiotherapy (RT) in non-small-cell lung cancer (NSCLC) patients treated with definitive RT whose gross tumor volumes (GTVs) were defined with the aid of pre-RT PET data. Method and Materials: The data from 26 patients treated with involved-field RT who had local failure and a post-RT PET scan were analyzed. The patterns of failure were visually scored and defined as follows: (1) within the GTV/planning target volume (PTV); (2) within the GTV, PTV, and outward; (3) within the PTV and outward; and (4) outside the PTV. Local failure was also evaluated as originating from nodal areas vs. the primary tumor. Results: We analyzed 34 lesions. All 26 patients had recurrence originating from their primary tumor. Of the 34 lesions, 8 (24%) were in nodal areas, 5 of which (63%) were marginal or geographic misses compared with only 1 (4%) of the 26 primary recurrences (p = 0.001). Of the eight primary tumors that had received a dose of <60 Gy, six (75%) had failure within the GTV and two (25%) at the GTV margin. At doses of ≥60 Gy, 6 (33%) of 18 had failure within the GTV and 11 (61%) at the GTV margin, and 1 (6%) was a marginal miss (p < 0.05). Conclusion: At lower doses, the pattern of recurrences was mostly within the GTV, suggesting that the dose might have been a factor for tumor control. At greater doses, the treatment failures were mostly at the margin of the GTV. This suggests that visual incorporation of PET data for GTV delineation might be inadequate, and more sophisticated approaches of PET registration should be evaluated