Diagnostic performance of MRI for prediction of candidates for local excision of rectal cancer (ypT0-1N0) after neoadjuvant chemoradiation therapy

Purpose: To evaluate the diagnostic performance of rectal MRI in predicting candidates for local excision (LE; ypT0-1N0) after neoadjuvant chemoradiation therapy (CRT) in patients with rectal cancer. Materials and Methods: The institutional review board approved our retrospective study and waived informed consent. Inclusion criteria were as follows: patients with pathologically confirmed mid to lower rectal cancer (cT3NxM0 before neoadjuvant CRT) who underwent neoadjuvant CRT and had MRI performed at 3T before and after neoadjuvant CRT. A total of 168 patients met the study criteria between 2011 and 2012. Two observers independently assessed tumor and nodal stages on MR images obtained after neoadjuvant CRT. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy for identifying a candidate for LE (ypT0-1N0) were calculated. Interobserver agreement was assessed with kappa value. Predictive factors for ypT0-1N0 were evaluated by logistic regression models. Results: MRI had relatively high accuracy, specificity, and NPV (85.9%, 93.8%, and 88.9% for observer 1 and 85.3%, 96.1%, and 86.6% for observer 2), moderate PPV (71.4% and 76.2%), and relatively low sensitivity (57.1% and 45.7%) for predicting ypT0-1N0. The interobserver agreement was fair (kappa value = 0.593). Carcinoembryonic antigen levels after neoadjuvant CRT and the maximal extramural depth of tumor spread were significant predictors of ypT0-1N0 (P = 0.037 and 0.017, respectively). Conclusion: MRI after neoadjuvant CRT can predict a candidate for LE (ypT0-1N0) after neoadjuvant CRT with moderate PPV and relatively high NPV. J. Magn. Reson. Imaging 2016;44:471–477.