Endoscopists systematically undersample patients with long-segment Barrett's esophagus: an analysis of biopsy sampling practices from a quality improvement registry

Background and Aims: Guidelines recommend systematic biopsy sampling in Barrett's esophagus (BE) to reduce sampling error. Adherence to this biopsy sampling protocol has been suggested as a quality indicator; however, estimates of adherence are not available. Using a national registry, we assessed adherence and identified predictors of adherence to biopsy sampling protocols. Methods: We analyzed data from the GI Quality Improvement Consortium Registry that included procedure indication, demographics, endoscopy, and pathology results. Patients with an indication of BE screening/surveillance or an endoscopic finding of BE were included. Adherence to the Seattle protocol was assessed by dividing BE length by number of pathology jars, with a ratio <= 2.0 with rounding down (lenient definition) or rounding up (stringent definition) for odd BE lengths considered adherent. Variables associated with adherence were assessed using generalized estimating equations to control for clustering within individual physicians. Results: Of 786,712 EGDs assessed, 58,709 (7.5%) EGDs in 53,541 patients met inclusion criteria (mean age, 61.3 years; 60.4% men; 90.2% white; mean BE length, 2.3 cm). When the lenient and stringent definitions for adherence were used, 87.8% and 82.7% of EGDs were adherent, respectively. Increasing BE length was the most significant predictor of nonadherence (odds ratio,.69; 95% confidence interval,.67-.71). Other predictors were increasing age, male gender, increasing American Society of Anesthesiologists class, and practice location. Performance of EGD by nongastroenterologist physicians was associated with nonadherence (odds ratio, .07; 95% confidence interval,.06-.10). Conclusions: Nearly 20% of endoscopies performed in BE patients were not adherent to the Seattle protocol. As BE length increases, endoscopists become less compliant with odds of nonadherence increasing by 31% with every 1-cm increase in length.