Colorectal cancer detection on routine abdomen/pelvis CT

Purpose

To assess the performance of routine abdomen/pelvis CT in detecting colorectal cancer.

Methods

A retrospective dataset consisting of 204 routine abdomen/pelvis CTs was compiled based on review of our institution's pathology/imaging archives. 64 cases had pathologically confirmed colorectal cancer, while 140 cases had no colonic neoplasms on colonoscopy. Two radiologists blinded to the patient history and imaging indication reviewed the cases independently. They scored each case for any colonic abnormality which could represent colorectal cancer. The imaging abnormality location, morphology, presence of regional lymphadenopathy, and degree of diagnostic confidence were recorded.

Results

Reviewers demonstrated a sensitivity of 64–77% and specificity of 83–86% for detection of colorectal cancer. Detected tumors were significantly larger than missed cancers, with mean maximum diameter of 5.2–5.7 cm for detected cancer (n = 81) and 2.8–2.9 cm for missed cancer (n = 25) (reviewer 1 p < 0.001; reviewer 2 p = 0.01). Positive predictive value (PPV) strongly correlated with diagnostic confidence, with 100% of pooled cases (39/39) reported as 5 (Almost Certainly Neoplastic) positive for cancer compared to 53% of cases (17/32) reported as 3 (Could be Benign or Neoplastic). The presence of regional lymphadenopathy was a strong predictive feature, with 94% of pooled cases with identified lymphadenopathy (48/51) positive for cancer compared to 51% (42/82) without lymphadenopathy (p < 0.001).

Conclusion

While colorectal cancers larger than 3 cm were detected with relatively high specificity, abdomen/pelvis CT performed poorly with smaller tumors. Radiologist-reported diagnostic confidence was strongly associated with diagnostic accuracy. Regional lymphadenopathy is a highly specific finding, although insensitive.