Accuracy of computed tomography in determining resectability for locally advanced primary or recurrent colorectal cancers

AIM: To determine the accuracy of computed axial tomography (CT) in determining tumor resectability in patients with locally advanced primary (T4) or locally recurrent colorectal cancer. METHODS: Computed tomography scans of 84 patients with 'resectable' locally advanced primary rectal cancer (n = 12) or recurrent colorectal cancer (n = 72) were compared with the operative findings to assess the accuracy of abdominal and pelvic CT in determining extent of disease and resectability. RESULTS: At surgery, disease was confined to the pelvis in 63 patients, the abdomen in 7, and involved both the pelvis and abdomen in 14. Computed tomography correctly identified these anatomic sites of tumor in 87% of patients, with 89% and 80% accuracies for pelvic and abdominal disease respectively. Tumor resection was performed in 71 patients (85%), but was not in 13 patients because of locally unresectable disease in 8 and metastatic disease in 5. The accuracy of predicting tumor related operability was 85%. With regard to adjacent organ resection, CT was accurate in determining the need for sacrectomy or hysterectomy, but overestimated the need for urinary organ resection. Based on histological examination of resection margins, CT correctly staged (n = 45) or overstaged (n = 9) 54 patients (64%) and understaged the remaining 30. The ability of CT to preoperatively predict a locally advanced tumor after preoperative radiation therapy as not being fixed was 30%, fixed but resectable 75%, and fixed but not resectable 25%. CONCLUSIONS: Computed tomography is generally reliable at identifying disease as being confined to one region, and for predicting the need for adjacent organ resection. It is less discriminating for predicting local tumor resectability.