Background: Recent studies have suggested that (18)fluoro-deoxyglucose (FDG) positron emission tomography (PET) may be useful in the work-up of patients with fever of unknown origin (FUO), but its exact diagnostic yield has not been established. Methods: From January 1999 through December 2001, 74 (67%) of 110 prospectively collected patients who fulfilled the revised criteria of classic FUO underwent a FDG-PET scan. The diagnostic yield of this technique was assessed after diagnostic work-up and follow-up. Abnormal FDG-PET scans that pointed to the final diagnosis were categorized as helpful; all other scans were considered noncontributory. Results: A final diagnosis was established in 39 (53%) of the 74 patients who underwent FDG-PET. Fifty-three (72%) of the 74 FDG-PET scans were abnormal; 19 scans (36% of the abnormal scans or 26% of the total number of scans) were helpful. In the 39 patients with a final diagnosis, 49% of the scans were helpful. A diagnosis was established in 31 (58%) of the 53 patients with an abnormal scan and in 8 (38%) of the 21 patients with a normal scan (P=0.2). Baseline clinical variables (age and sex, as well as periodicity, duration, and height of the fever) and inflammatory parameters (erythrocyte sedimentation rate, C-reactive protein, and hemoglobin) did not predict the usefulness of FDG-PET. Conclusions: FDG-PET contributed positively to the diagnosis in a quarter of our patients with classical FUO. This number is lower than that found in previous studies. Yet, against the background of the wide array of heterogeneous disorders that make up the FUO spectrum and the low number of final diagnoses established (in only 53% of cases), the diagnostic yield of FDG-PET is encouraging. Therefore, the use of FDG-PET should be considered whenever a baseline work-up fails to reveal the cause of a prolonged, febrile illness.