Accuracy of MRI in the detection of pathological complete response post neoadjuvant therapy in locally advanced colorectal cancer.

e15670 Background: Pathological Complete Response (pCR) conventionally denotes the absence of malignant cells in rectal specimens resected from patients who have undergone Neoadjuvant Therapy. Patients showing pCR show better prognosis and Radical Surgery can be avoided in these patients. Assessment of pCR status post Neoadjuvant Therapy aids in tailoring management through rectal sparing approaches. Several techniques are used to assess pCR status such as Digital Rectal Exam, Rectoscopy, CT and MRI scans. Absence of residual tumor or tumor regression observed on post NACT MRI scans are likely parameters which indicate Complete Response and can be utilized in the non invasive diagnosis of pCR status. This study aims to evaluate the accuracy of MRI in predicting pathological complete response after neoadjuvant therapy. Methods: A total of 23 patients who underwent Neoadjuvant Therapy and MRI post NACT, followed by surgical resection in a single centre were retrospectively evaluated. MRI was performed on a 1.5 T machine using the institutional MRI Rectal Protocol. Intravenous contrast was not in accordance with the institutional protocol. Protocols included were large FOV and small FOV (field of view) , T1 weighted and T2 weighted axial oblique coronal T2W and axial diffusion weighted images (DWI). The images were retrospectively reviewed by a single experienced radiologist. Based on the MRI imaging, patients were divided into complete responders and non complete responders based on tumor regression grade (mrTRG) as assessed on MRI. Pathological tumor regression grade (pTRG) was considered the gold standard and mrTRG was compared with it. Interobserver agreement between mrTRG and pTRG was assessed by Cohens kappa. Results: 4 out of 23 patients had a complete response as assessed on histopathology. MRI correctly identified 1 complete responder. Accuracy was calculated to be 0.78. Sensitivity, specificity, PPV and NPV for detection of pathological complete response were 0.25, 0.89, 0.33, 0.85 respectively. Interobserver agreement between MRI and histopathology was poor (kappa = 0.161, p > 0.05). Conclusions: MRI shows high Accuracy and Specificity in the identification of patients with Complete Response, yet it is associated with poor sensitivity. This implies that MRI can be used to predict non responders with high reliability whereas it is not dependable to determine the status of complete responders.