Variations between endoscopists in rates of detection of colorectal neoplasia and their impact on a regional screening program based on colonoscopy after fecal occult blood testing

Background There are few data about the performance variability among endoscopists participating to nationwide or regionwide colorectal cancer screening programs. Objective To assess the variability of neoplasia detection rates among endoscopists participating in a regional colorectal cancer screening program based on colonoscopy after biennial fecal occult blood testing (FOBT). Design Two rounds of colonoscopy were performed: round 1 took place in 2003 and 2004, and round 2 took place in 2005 and 2006. Secondary analysis of colonoscopy findings from the first 2 rounds was performed by using data drawn from all endoscopists who performed more than 30 colonoscopies in each round. Detection rates were adjusted for patient age and sex, and logistic regression analyses were conducted including these 2 variables and round number (1 or 2). Setting District of Ille-et-Vilaine in Brittany (population >900,000) between 2003 and 2007. Main Outcome Measurements The per-endoscopist adjusted rates of colonoscopies with at least 1, 2, or 3 adenomas, 1 adenoma 10 mm or larger, or a cancer. Results Among the 18 endoscopists who performed 3462 colonoscopies, the adjusted detection rates were in the following ranges: at least 1 adenoma, 25.4% to 46.8%; 2 adenomas, 5.1% to 21.7%; 3 adenomas, 2.7% to 12.4%; 1 adenoma 10 mm or larger, 14.2% to 28.0%; and cancer, 6.3% to 16.4%. Multivariate analyses showed that the endoscopist was not an independent predictor of cancer detection, but was an independent predictor of detecting adenomas, regardless of category; the R 2 of the models ranged from 6% to 13% only. Limitations Other factors known to influence colorectal neoplasia occurrence and withdrawal time could not be taken into account. Conclusions In a screening program with a high compliance rate with colonoscopy after FOBT, interendoscopist variability had no effect on cancer detection, but did influence identification of adenomas. The clinical impact of such findings merits further evaluation.