Conventional CT for the prediction of an involved circumferential resection margin in primary rectal cancer

Purpose: To determine the accuracy of conventional computed tomography (CT) scan in the preoperative prediction of an involved circumferential resection margin (CRM) in primary rectal cancer. Methods: 125 patients with biopsy-proven adenocarcinoma of the rectum underwent CT of the abdomen before undergoing total mesorectal excision. Scans were scored by three observers, differing in experience. The main outcome was yes/no involvement of the CRM. Histology was taken as reference standard. Results: For the most experienced observer, observer A, sensitivity was 46.7% and specificity 92.6%. For observer B, sensitivity was 46.7% and specificity 89.5%. For the least experienced observer C, sensitivity was 43.3% and specificity 92.6%. Inter-observer variability was good between observers A and B (kappa 0.648), B and C (kappa 0.648), and intermediate between A and C (kappa 0.542). Discrepancies occurred in a total of 34 patients; 25 had a CT scan of low technical quality, 10 an anteriorly located distal tumor. Conclusion: Conventional CT scan lacks sensitivity for a clinical use in the preoperative assessment of an involved CRM in primary rectal cancer. Modern multislice spiral CT will probably resolve some of the problems of conventional CT; however, further research is needed to establish its role. Copyright (c) 2007 S. Karger AG, Basel