Pitfalls in diagnosing colon cancer on abdominal CT

To assess the frequency of undetected colon cancer on conventional abdominal CT and to evaluate the imaging features that are characteristic of those cancers.

The present study included consecutive patients diagnosed with colorectal cancer at colonoscopy (2006-2015) who also underwent abdominal computed tomography (CT) performed for various reasons within a year prior to the colonoscopy. The frequency of undetected lesions was evaluated for the original CT interpretations ("original readers"). Two radiologists ("study readers"), blinded to the tumour location, independently performed interpretations oriented for colon cancer detection. The study readers analysed the imaging features of detected tumours (tumour shape, length, maximal wall thickness, free fluid, fat stranding, vascular engorgement, stenosis, and lymphadenopathy). Imaging features of the cancers undetected by the original readers were evaluated.

The study included 127 patients. The original readers' frequency of undetected cancer was 25/127 (19.7%). Each study reader could not identify the cancer in 8/127 (6.3%) patients. Imaging features associated with undetected cancers by the original readers included the absence of fat stranding (p=0.007, p=0.003), absence of vascular engorgement (p<0.0001, p<0.0001) and absence of lymphadenopathy (p=0.005, p=0.004). Undetected tumours were shorter than those detected (original reader: 33.2±11.9 versus 51.4±18.2 mm; study reader: 32.5±9.6 versus 61.3±23.4 mm; p<0.001).

Colon cancer is undetected in 20% of abdominal CT examinations in patients subsequently proven to have colon cancer at colonoscopy. The absence of fat stranding, vascular engorgement, or lymphadenopathy, and an average tumour length of 3.3 cm are contributing factors for failure of detection. Radiologists' training should emphasis these findings as it may improve cancer detection, and clinicians should be aware of the limitations of abdominal CT.