Diagnostic accuracy of nodal enhancement pattern of rectal cancer at MRI enhanced with ultrasmall superparamagnetic iron oxide: findings in pathologically matched mesorectal lymph nodes.

Objective: The purpose of this study was to evaluate the diagnostic accuracy of the pattern of nodal enhancement at MRI enhanced with ultrasmall superparamagnetic iron oxide (USPIO) in the nodal classification of rectal cancer in pathologically matched mesorectal lymph nodes.

Subjects And Methods: Twenty-five patients with adenocarcinoma of the rectum underwent prospective evaluation with 3-mm axial T2-weighted and USPIO-enhanced T2*-weighted MRI before surgery. Mesorectal nodes visible at in vivo MRI were independently scored by two radiologists as malignant or nonmalignant according to morphologic criteria (irregular nodal contour, heterogeneous signal intensity) on T2-weighted MR images and according to USPIO enhancement pattern on T2*-weighted MR images. The sensitivity, specificity, and positive and negative predictive values of morphologic and USPIO criteria in identification of malignancy in the pathologically matched mesorectal nodes were compared by use of the McNemar test. Interobserver agreement was compared by use of kappa statistics.

Results: After surgery, radiologic-pathologic comparison of 126 mesorectal nodes (116 benign, 10 malignant) was possible. Use of morphologic criteria resulted in an average sensitivity of 65% (95% CI, 35-88%); specificity, 75% (67-83%); positive predictive value, 19% (8-34%); and negative predictive value, 96% (91-99%). Use of USPIO criteria resulted in an average sensitivity of 65% (95% CI, 35-88%); specificity, 93% (87-96%); positive predictive value, 43% (21-67%); and negative predictive value, 97% (92-99%). Use of USPIO MRI improved diagnostic specificity for both observers (p < 0.01). Interobserver agreement was fair for morphologic criteria (kappa = 0.39) but good for USPIO criteria (kappa = 0.68).

Conclusion: Use of the pattern of USPIO enhancement had higher diagnostic specificity than but the same sensitivity as morphologic findings in pathologically matched mesorectal lymph nodes.