Diagnostic accuracy of SUVmax in predicting malignancy of supraclavicular lymph nodes from primary oesophageal cancer

To determine the diagnostic accuracy and optimum cut-off value of SUVmax on PET to predict malignancy of supraclavicular lymph nodes (SCLNs) in patients with oesophageal carcinoma.All diagnosed cases of oesophageal cancer were retrospectively reviewed (2010-2016). Patients that had a confirmed diagnosis of oesophageal cancer with avid SCLNs on staging PET were included in the study. 33 SCLNs that subsequently underwent ultrasound guided biopsy for staging were analysed. The maximum uptake values (SUVmax) of the SCLNs and primary tumours were measured. A receiver operating characteristic (ROC) analysis was performed to determine the optimum cut off of SUVmax in predicting malignancy.24/33 PET-detected SCLNs were malignant. ROC analysis identified the best nodal SUVmax cut-off to be 3.0. The diagnostic accuracy of PET was 76.0 % (sensitivity = 78.9 %, specificity = 66.6 %). For SCLNs with SUVmax > 3.0, PET showed a positive predictor value of 88.2 %; for SCLNs < 3.0, PET showed a negative predictor value of 50 %. The median SUVmax of pathologically negative and positive nodes were 2.8 (range 1.8-6.0) and 5.3 (range 1.9-13.4). The median primary tumour SUVmax was 13.8 (range 3.7-30.0). The SUVmax of metastatic lymph nodes were significant higher than those of benign lesions (p < 0.05).Our study revealed an accuracy rate of 76 % for PET detected SCLNs in patients with oesophageal carcinoma. For SCLNs with SUVmax > 3.0, PET had a high PPV (88 %), which can minimize the need for further diagnostic tests.