In regions with a high prevalence of granulomatous diseases, benign inflammatory fluorine-18 fluorodeoxyglucose (FDG) uptake in the mediastinum is frequently observed even in healthy subjects. We examined parameters of mediastinal FDG uptake to determine whether they can differentiate malignancy from benign lesions. Seventy patients with non-thoracic tumours who had mediastinal uptake on FDG positron emission tomography (PET) were included (33 males, 37 females; age 57.5 +/- 16.9 years; 168 lymph nodes). Determination of metastasis was confirmed by biopsy or computed tomography (CT) follow-up over 12 months (metastasis, 29; benign lesions, 41). No significant difference between the metastasis group and the benign group was found in terms of residual disease in the primary site (48% vs 46%), lung invasion (29% vs 20%), number of sites of uptake (2.3 vs 2.4), smoking history (30.3% vs 46.3%) or bilateral uptake (52% vs 54%). Maximal standardised uptake values (SUVs) in the mediastinal metastasis group were higher (4.9 +/- 1.8) than those in the benign group (2.5 +/- 0.9) (P < 0.05). Using 3.4 as a cut-off value for maximal SUV, a sensitivity of 86% and a specificity of 85% were achieved (AUC = 0.917). Maximal SUV showed better predictive value than lymph node size measured on chest CT (P < 0.05). In 8 of 51 normal subjects who underwent FDG PET as a routine check-up, mediastinal FDG uptake was observed. Maximal SUV in normal subjects was 2.5 +/- 0.8, which was similar to that in the benign group. In conclusion, maximal SUV was identified as a significant parameter for determining whether mediastinal FDG uptake represents malignant metastasis. When maximal SUV exceeded 3.4, the metastasis rate was high regardless of lymph node size.