Accuracy of preoperative magnetic resonance imaging in prediction of pathological stage and circumferential resection margin in rectal cancer

Abstract

To evaluate the accuracy of preoperative magnetic resonance imaging (MRI) in prediction of pathological staging and involvement of circumferential resection margin (CRM) in rectal cancer. Fifty-three patients undergoing total mesorectal excision for biopsy-proven rectal cancer were assessed prospectively using high-resolution MRI for tumour (T) and mesorectal nodal (N) staging as well as CRM status using the depth of tumour spread, tumour node metastasis and CRM involvement. Preoperative MRI assessment of these prognostic factors was compared with the histopathological findings in carefully matched whole-mount sections of the specimen. MRI correctly staged the tumor in 41 patients, understaged in 8, and overstaged in 4. The accuracy of T stage was 77.4% (41/53). There was ageneric correlation between pathologic and MRI tumor staging (Kappa = 0.602, P < 0.001). Node status was correctly staged in 37 patients, overstaged in 10, and understaged in 6. The accuracy of node staging was 69.8% (37/53), sensitivity was 75% (18/24), and specificity was 65.5% (19/29). The correlation between pathologic and MRI node staging was poor (Kappa = 0.399, P = 0.003). The CRM status was correctly reported in 51 patients, overstaged in 1, and understaged in 1. The accuracy of CRM status was 96.2% (51/53), sensitivity was 80% (1/5), and specificity was 97.9% (47/48). There was a good correlation between pathologic and MRI CRM involvement (Kappa = 0.779, P < 0.001). Preoperative MRI provides poor predictive data as to subsequent pathologic tumor and mesorectal node stage, but does produce reliable prediction of clear CRM.