This retrospective study was designed to assess the accuracy of fluorine-18 fluorodeoxyglucose positron emission tomography (FDG-PET) in diagnosing recurrence of gastric cancer. Thirty-three patients who had received surgical treatment for gastric cancer with curative intent and who had subsequently undergone FDG-PET for suspected recurrence were retrieved from the PET database. All patients were reviewed with full knowledge of prior conventional diagnostic work-up. Results were compared with a gold standard, consisting of histological confirmation or radiological and clinical follow-up. The gold standard established disease recurrence in 20/33 patients (prevalence 61%). Sensitivity and specificity of FDG-PET for the diagnosis of recurrence were 70% (14/20) and 69% (9/13), respectively. Positive and negative predictive values were 78% (14/18) and 60% (9/15), respectively. Of the six false-negative cases, all had intra-abdominal lesions (three had generalised abdominal metastases, one liver metastasis, one local recurrence and one ovarian metastasis). In the subgroup with previous signet cell differentiation of the primary tumour ( n=13, disease prevalence 62%), sensitivity was 62% (5/8) and specificity, 60% (3/5). Survival analysis for the entire patient group using Kaplan-Meier statistics yielded a longer survival in the PET-negative group (mean+/-SD, 21.9+/-19.0 months) than in the PET-positive group (mean+/-SD, 9.2+/-8.2 months) ( P=0.01). In the patient group with proven recurrence ( n=20), the mean survival for the PET-negative group was 18.5 (+/-12.5) months, as compared with 6.9 (+/-6.5) months for the PET-positive group ( P=0.05). Because of its poor sensitivity and low negative predictive value, FDG-PET is not suited for screening purposes in the follow-up of treated gastric cancer. However, FDG-PET appears to provide important additional information concerning the prognosis of recurrent gastric cancer.