Magnetic resonance imaging in rectal cancer downstaged using neoadjuvant chemoradiation: accuracy of prediction of tumour stage and circumferential resection margin status

Abstract

The aim was to examine the accuracy of magnetic resonance imaging (MRI) in predicting circumferential resection margin (CRM) involvement, T- and N-stage in patients with locally advanced carcinoma of the rectum, who had undergone long-course downstaging chemoradiation (CRT). Patients with rectal cancer were selected for long-course downstaging CRT if their tumour was considered to threaten (< or = 1 mm) or involve the CRM on MRI. Eighty such patients had a repeat MRI at a median of 6 weeks post-CRT followed by surgical excision soon thereafter. The findings on the post-CRT MRI were compared with histological examination of the surgical specimen. For CRM involvement, post-CRT restaging MRI had an accuracy of 81% (65/80) a sensitivity of 54% (7/13), a specificity of 87% (58/67), a positive predictive value of 44% (7/16) and a negative predictive value of 91% (58/64). Accuracy for T- and N-staging was 43% (34/80) and 78% (62/80), respectively. 38% of T-stages were overstaged and 20% understaged. 4% of N-stages were overstaged and 19% understaged. The 13 patients with histological positive CRM had worse clinical outcomes than the 67 patients with negative CRM in terms of disease-free survival (relative risk of reduced DFS 4.6, P = 0.001) and overall survival (relative risk of death 3.6, P = 0.016). Magnetic resonance imaging has good specificity and negative predictive value for predicting an uninvolved CRM post downstaging CRT in locally advanced rectal cancer although sensitivity and positive predictive value for an involved CRM were unsatisfactory. The shortcomings of MRI stem from poor differentiation of viable tumour from posttreatment changes and inability to identify small nodal and tumour deposits. Clinical correlates in this group of patients have confirmed the importance of achieving a clear CRM at surgery.