Purpose: To investigate the optimum cutoff for lymph node size to identify cases positive for perirectal lymph node (PRLN) and lateral lymph node (LPLN) metastasis of lower rectal cancer on magnetic resonance imaging (MRI).
Methods: The subjects were 449 patients who underwent preoperative MRI. Mesorectal excision was performed in all patients (combined with lateral pelvic lymph node [LN] dissection in 324) between 2004 and 2013 at 6 institutes. Cases were classified as cN positive and cN negative on the basis of the short axis of the largest LN being greater than or equal to a cutoff or less than a cutoff, respectively. PRLN and LPLN diagnoses using 5 and 10 mm cutoffs were compared with histologic diagnoses. Of the 449 patients, 55 received preoperative chemoradiotherapy. MRI was only performed after this therapy in all of these patients.
Results: For PRLNs, 5 and 10 mm cutoffs gave area under the curve (AUC) values of 0.6364 and 0.5794, respectively. The 5 mm cutoff gave a significantly higher AUC value (P = 0.0152), with an accuracy of 63.7 %, sensitivity of 72.6 %, and specificity of 54.7 %. For right LPLNs, the respective AUC values were 0.7418 and 0.6326 (P = 0.0034), and the variables (5 mm cutoff) were 77.6, 68.6, and 79.7 %. For left LPLNs, AUC values were 0.7593 and 0.6559, respectively (P = 0.0057), and the variables (5 mm cutoff) were 79.3, 70.8, and 81.0 %.
Conclusions: Identification of LN-positive cases on the basis of PRLN and LPLN sizes was superior at a short-axis 5 mm cutoff. Size-based diagnosis of LN metastasis is simple and useful, but further investigation is needed to clarify whether it is superior to diagnosis based on morphology, such as shape, border, and signal intensity.