PURPOSE: To evaluate the diagnostic accuracy of positron emission tomography (PET) with administration of 2-deoxy-2-[fluorine-18]fluoro-D- glucose (FDG) relative to that of magnetic resonance (MR) imaging and/or computed tomography (CT) in recurrent head and neck cancers. MATERIALS AND METHODS: Twelve adult patients (mean age, 63 years) with previously treated head and neck cancers and clinical suspicion of recurrence underwent FDG PET and MR imaging and/or CT. All images were blindly and independently interpreted without histopathologic findings (obtained within 1 week of imaging). The level of confidence in image interpretation was graded by using a five-point rating system (0 = definitely no recurrence to 4 = definite recurrence). RESULTS: Recurrence was confirmed in eight patients. With a rating of 4 as a positive finding, FDG PET yielded a sensitivity and specificity of 88% (seven of eight) and 100% (four of four), respectively; MR imaging and/or CT, 25% (two of eight) and 75% (three of four), respectively. Receiver-operating characteristic analysis showed significantly better diagnostic accuracy with FDG PET than with MR imaging and/or CT (area under curve = 0.96 vs 0.55, P < .03). CONCLUSION: These data indicate that PET metabolic imaging, as compared with anatomic methods, has improved diagnostic accuracy for recurrent head and neck cancer.