Background: The aim of this study was to evaluate the accuracy of magnetic resonance (MR) in loco-regional staging of rectal cancer by comparing the MR results with histologic findings, considered as standard reference.
Methods: Between July 2013 and March 2015, fifty-two patients, 27 (51.9%) males, age 66.75±13.77 years, with rectal cancer, were staged preoperatively with MR and proceeding straight to surgery. Two observers with experience in abdominal MR independently reviewed the images. T stage and N stage were evaluated according to the 7th edition of TNM classification. The estimate of Ln probability of malignancy (low, moderate, high) was based on nodal size, border contour and signal intensity and comparison between ADC value of the Ln's belonging to the three different classifications were performed. Statistical testing included Cohen's kappa coefficient, Mann-Whitney's, Kruskal-Wallis, χ2, Fisher's Exact Test and Receiving Operating Characteristics curve.
Results: MR correctly assessed T stage in 47/52 cases (90.4%; kw=0.89±0.06), with inter-operator concordance of k=0.81±0.08. For Ln staging, concordance between estimate of high probability malignancy and pathology was kw=0.62±0.11. ADC was significantly different for the three grades of estimated malignancy probability (P=0.0003), decreasing from 1.227±0.298×10-3 mm2/s (low) to 1.120±0.306×10-3 mm2/s (moderate) and finally to 0.818±0.168×10-3 mm2/s (high). The ROC curve procedure established the good ability of ADC to discriminate high malignancy Ln's (AUC=0.88) with cut-off at <1×10-3 mm2/s. The percentage of high malignancy Ln's in the lateral pelvic space was higher than in other sites (55.6% vs. 17.6%, P=0.0003).
Conclusions: MR is an accurate imaging method in T staging and N staging of rectal cancer: prediction of N was improved by considering dimension, morphology and signal characteristic and the ability of ADC to identify high probability malignant nodes underlines its importance in the diagnostic process.