Colorectal cancer prevention.

Abstract

Introduction: To develop a regression model that can reliably predict the recommended time interval for surveillance colonoscopy, possibly eliminating the need for pathology. Methods: We retrospectively analyzed 625 patients who had colonoscopy with polyps in the year 2013 at a university hospital. Exclusion criteria included patients with prior colon cancer or colorectal surgery, IBD, colon cancer syndrome, polyps not retrieved, or a lesion only biopsied. Binary logistic regression was performed using Statistical Package for the Social Sciences v.21. The binary dependent variable was the surveillance interval recommendation of 3 years or other than 3 years. The reference value was the 2012 AGA guidelines. In this study, we treated screening and surveillance together and hyperplastic polyps beyond the splenic flexure as adenomas, as there is evidence that pathologist may misclassify SSA as hyperplastic. Independent variables were defined as follows: number of polyps (1, 2, 3, or > 4), size of the polyps (any polyp >10mm or not), and location of the polyps (0, 1, 2, or >3 polyps proximally). Results: Two thousand two hundred three colonoscopies were reviewed. Six hundred twenty-five exams met the study criteria. Of those, 429 were for screening and 196 for surveillance. Polyp size, number, and location were all significant factors in determining the interval. The nagelkerke R square is 0.803, and the omnibus tests of the coefficients were all <0.000 (Table 1). Sensitivity and specificity for the model is 92.2% and 94.9%, respectively. Area under the ROC curve was 0.93. For the 3-year interval, 3.5% of patients got a false positive result, and 2.4% got a false negative result. Two patients with pathology of rectal carcinoid and a rectal leiomyoma were assigned “not 3 years.” False positives occurred mainly in patients with multiple hyperplastic polyps (15 patients) or large benign polyps (5 patients). False negatives occurred in 3 patients with tubulovillous adenoma less than 10 mm and in 15 patients with between 3-5 small polyps.Table 1: Coefficients of RegressionConclusion: This regression model may allow gastroenterologists to reliably determine surveillance interval at the time of procedure, saving time, cost, and possibly improving adherence and decreasing anxiety. Furthermore, the consistency of pathology has been challenged for diagnosis of sessile serrated adenomas in the proximal colon, thus putting the utility of pathology in question. Though validation is needed, this model may be used to discard resected polyps without pathology evaluation when the results are unlikely to change clinical behavior.