Diagnostic accuracy of computed tomography and magnetic resonance imaging obtained after neoadjuvant chemoradiotherapy in predicting the local tumor stage and circumferential resection margin status of rectal cancer

To measure the diagnostic accuracy of computed tomography (CT) and magnetic resonance imaging (MRI) obtained after neoadjuvant chemoradiotherapy (CRT) in patients with rectal cancer for a prediction of the local tumor stage and circumferential resection margin (CRM). Two independent radiologists reviewed CT and MRI obtained after neoadjuvant CRT. The accuracy of the local tumor staging and the diagnostic performance for the prediction of CRM involvement were calculated. The agreement between the measurements of the distance to potential CRM on both imaging modalities and the histopathology findings was assessed using Bland-Altman plots. 57 patients (mean age, 59.2 years; 24 females) were included. The accuracy of T and N staging were 43.9% (95% confidence interval, 30.8-57.7%) and 77.2% (64.2-87.3%) on CT and 63.2% (49.4-75.6%) and 77.2% (64.2-87.3%) on MRI for Observer 1. The accuracy of T and N staging were 54.4% (40.7-67.7%) and 77.2% (64.2-87.3%) on CT and 68.4% (54.7-80.1%) and 80.7% (68.1-90.0%) on MRI for Observer 2. Sensitivity and specificity on CRM involvement were 83.3% (43.7-97.0%) and 88.2% (76.6-94.5%) on CT and 100% (61.0-100%) and 90.2% (79.0-95.7%) on MRI for Observer 1. Sensitivity and specificity on CRM involvement were 66.7% (30.0-90.3%) and 88.2% (76.7-94.5%) on CT and 100% (61.0-100%) and 90.2% (79.0-95.7%) on MRI for Observer 2. Bland-Altman plots showed wide discrepancies between measurements of the distance to CRM on each CT and MRI and those on histopathology findings. CT and MRI showed limited performance in predicting the local tumor staging and CRM involvement in patients with neoadjuvant CRT although MRI tended to show a better performance than CT