Accuracy of MRI for nodal restaging in rectal cancer: a retrospective study of 166 cases.

Abstract

Assessing metastatic mesorectal nodal involvement is a challenge in rectal cancer, especially in the post chemoradiation setting. We aim to assess the accuracy of MRI for nodal restaging and the validity of SAR criteria (≥ 5mm size being metastatic). This was an IRB-approved retrospective study of 166 patients with locally advanced rectal cancers, operated after neoadjuvant treatment. Two dedicated oncoradiologists reviewed the 166 post-chemoradiation presurgical MRIs in consensus. Nodal size and morphology (shape, margins, and signal intensity) were noted. The most accurate cut-off for size for predicting positive pN status was determined using the Youden index. MRI understaged 30/166 (18%) and overstaged 40/166 (24%) patients using the SAR criteria. The most accurate cut-off for node size was 5.5mm, with a sensitivity of 75%, specificity of 60.2%, PPV of 40.7%, NPV of 86.9% (95% CI:78-92.5%), accuracy of 64.2%, and area under the curve (AUC) 0.657 (95% CI-0.524-0.79). Morphological characteristics were not significant to determine involvement, with positive nodes including 42% of round and 31% of oval nodes, 40% of heterogeneous and 45% of homogeneous nodes, and 31% irregularly marginated and 46% nodes with regular margins beingpositiveon pathology. MRI was accurate in predicting pathology for mucinous nodes in 9/29 (31%) cases. Seven cases which were yN2 on MRI and yN0 on pathology demonstrated mucinous changes on MRI and had acellular mucin on histopathology. MRI has good negative predictive value, poor positive predictive value and moderate accuracy in nodal restaging. The cut-off of 5.5mm demonstrated in our study is close to the SAR cut-off of 5mm in the post-treatment setting. MRI accuracy is lower in patients with mucinous nodes.