In advanced head and neck cancer, an organ-sparing approach comprising radiation therapy combined with intra-arterial chemotherapy has become an important technique. However, the high incidence of residual masses after therapy remains a problem. In this study, we prospectively evaluated the use of 2-[18F]fluoro-2-deoxy-D-glucose positron emission tomography (FDG-PET) delayed imaging for the detection of recurrence of head and neck cancer after radio-chemotherapy, and compared the FDG-PET results with those of magnetic resonance imaging (MRI) or computed tomography (CT). Forty-three lesions from 36 patients with head and neck cancer suspected to represent recurrence after radio-chemotherapy (median interval from therapy, 4 months) were studied. PET was performed at 2 h after FDG injection, and evaluated. The results were compared to those of contrast studies with MRI or CT performed within 2 weeks of the PET study, and to histological diagnosis (in all patients suspected of having recurrence) or clinical diagnosis. The lesion-based sensitivity (visual interpretation) and negative predictive value of FDG-PET (88% and 91%, respectively) were higher than those of MRI/CT (75% and 67% respectively). The specificity, accuracy and positive predictive value of FDG-PET (78%, 81% and 70%, respectively) were significantly ( P<0.05) higher than those of MRI/CT (30%, 47% and 39% respectively). Three of six patients with false positive findings had post-therapy inflammation. Receiver operating characteristic (ROC) analysis showed that retrospective evaluation with the standardised uptake ratio yielded the best results (sensitivity 87.5%, specificity 81.5%), followed by visual interpretation and then the tumour/neck muscle ratio. An FDG-PET delayed imaging protocol yielded significantly better results for the detection of recurrence of head and neck cancer after radio-chemotherapy than MRI/CT. Because of the high negative predictive value of FDG-PET (91.3%), if PET is negative, further invasive procedures may be unnecessary.