Diagnostic Accuracy and Reliability of CT-based Node-RADS for Esophageal Cancer

Background/aim: The recently published Node-Reporting and Data System (Node-RADS) can aid the characterization of lymph nodes in cross-sectional imaging. This study investigated the Node-RADS system in computed tomography (CT) to characterize lymph nodes in esophageal cancer.

Patients and methods: Overall, 126 patients (15 female, 11.9%) with a mean age of 62.1±10.4 years comprised the patient sample. All patients underwent resection with curative intent and the lymph nodes were histopathologically analyzed during clinical routine. For every patient, the locoregional lymph nodes were scored in accordance with the Node-RADS classification. For statistical analysis, receiver-operating characteristics (ROC) with area under the curve (AUC) were used to test for diagnostic accuracy; inter-reader variability was assessed with Cohen's kappa.

Results: Overall, 54 patients were nodal positive (42.9%), 72 patients were nodal negative (57.1%). Inter-reader agreement was substantial for the overall Node-RADS scoring ([Formula: see text]=0.65, p<0.001). ROC curve analysis for lymph node discrimination (N0 versus N1-3) showed an AUC of 0.69 (95% confidence interval=0.59-0.79). A threshold score of more than 2 resulted in a sensitivity of 0.77 and a specificity of 0.55 for correctly predicting nodal positivity. Node-RADS 1 category had a malignancy rate of 30%, Node-RADS 2 of 14%, Node-RADS 3 of 81%, Node-RADS 4 of 90.1% and Node-RADS 5 of 86.5%.

Conclusion: The Node-RADS score on staging CT is associated with the malignancy rate of lymph nodes in patients with EC with only moderate diagnostic accuracy. The inter-reader variability is moderate, which could pose difficulties for translation into clinical routine.