Mesorectal fascia invasion after neoadjuvant chemotherapy and radiation therapy for locally advanced rectal cancer: accuracy of MR imaging for prediction

PURPOSE: To retrospectively assess sensitivity and specificity of magnetic resonance (MR) imaging after chemotherapy and radiation therapy for predicting tumor invasion of the mesorectal fascia (MRF) in locally advanced primary rectal cancer, by using results of histologic examination and surgery as the reference standard, and to determine morphologic MR imaging criteria for MRF invasion. MATERIALS AND METHODS: The Ethical Committee of University Hospital Maastricht approved this study and waived informed consent. Two observers independently scored postchemoradiation MR images in 64 patients with rectal cancer (38 male [mean age, 60 years] and 26 female [mean age, 64 years] patients) for MRF tumor invasion with a confidence level scoring system defined by subjective criteria. In a subsequent consensus reading session, morphologic MR criteria for invasion were defined by comparing morphologic changes with histologic findings. These criteria were evaluated and compared with the subjective criteria by comparing areas under the receiver operating characteristic curves (AUCs). RESULTS: AUCs of postchemoradiation MR imaging for predicting MRF tumor invasion were 0.81 and 0.82 for observers 1 and 2, respectively. The following four types of morphologic tissue patterns at MR imaging were associated with whether or not MRF invasion was present at histologic examination: (a) development of fat pad larger than 2 mm (seen in no quadrants with and in four quadrants without invasion), (b) development or persistence of spiculations (seen in no quadrants with and in 22 quadrants without invasion), (c) development of diffuse hypointense "fibrotic" tissue (seen in 21 quadrants with and in 32 quadrants without invasion), and (d) persistence of diffuse iso- or hyperintense tissue (seen in 19 quadrants with and in two quadrants without invasion). AUC of postchemoradiation MR imaging for predicting MRF invasion on the basis of morphologic criteria was 0.80. There was no significant difference between the performance of subjective and morphologic criteria (P = .73-.76). CONCLUSION: Postchemoradiation MR imaging findings have moderate accuracy for predicting tumor invasion of the MRF related to the limitation in differentiating between diffuse "fibrotic" tissue with and that without small tumor foci. Specific other types of morphologic patterns at MR imaging can highly predict a tumor-free or invaded MRF.