Accuracy of visual prediction of pathology of colorectal polyps: how accurate are we?

Background: Small and diminutive polyps seen at colonoscopy could be left unresected because of a balance of risks between perceived low malignant potential and complications of removing the polyp. This relies on a high accuracy in prediction of the pathology of the polyp. This study was undertaken to determine if experienced endoscopists could determine the histological types of polyps found at colonoscopy.

Methods: We performed a prospective study at Christchurch Hospital from December 2010 to December 2012. Gastroenterologists and surgeons were asked to fill out a questionnaire at time of colonic polypectomy, predicting pathology of polyp excised. Data were collected on indication for colonoscopy, location, size of polyp and use of narrow band imaging. Predictive accuracy was then calculated.

Results: A total of 527 polyps were excised from 258 patients. Ninety-one per cent were less than 1 cm. Predictive accuracy of hyperplastic polyps was 47%. Of the 53% wrong predictions, 82% were adenomas. Predictive accuracy of adenoma was 42%, with 66% of incorrect predictions being hyperplastic polyps. Accuracy was lowest in the right colon, with only 25% of predicted hyperplastic polyps actually being hyperplastic. A patient's age, sex, polyp size and site did not affect prediction accuracy. Overall accuracy of visual prediction of pathology of all colorectal polyps was 45%. Narrow band imaging was used too infrequently to determine how useful it was.

Conclusion: Pathology of colorectal polyps cannot be accurately predicted visually alone. All polyps visualized at colonoscopy should be excised for ex vivo histology, regardless of size, location or predicted pathology.