[The value of CT for the diagnosis of rectal cancer recurrence. A ROC study and a retrospective analysis].

To assess the ability of CT to detect recurrent tumor growth after colorectal surgery for rectal carcinoma independently of clinical bias and patient selection, a receiver operator characteristic (ROC) analysis was performed. A set of 100 CT scans, 50 scans of 35 patients with proven recurrent disease, and 50 normal postoperatives of 37 patients were read by four observers without knowledge of the present clinical status. The four ROC curves had an area under the curve (AUC) ranging from 0.69 to 0.78 (SD from 0.046 to 0.053). The slightly raised AUC indicates the poor diagnostic power of CT in discriminating between postoperative changes and recurrence. In a retrospective analysis of the CT scans of the 72 patients CT had an overall accuracy of 83%. CT was unable to detect small anastomotic recurrence. The morphological characteristics of recurrence were not specific, but an irregular globular mass with or without inhomogeneity or asymmetry strongly suggests recurrent tumor and stresses the need for further evaluation, e.g., CT-guided biopsy.