Predictive Accuracy of Optical Evaluation for High-Risk Serrated Lesions and the Impact of a Brief Educational Intervention.

Background Real-time characterization of advanced lesions during colonoscopy informs the therapeutic approach. Sessile serrated lesions with dysplasia are a subgroup of advanced lesions that are difficult to recognize and accurately characterize. The ability of expert endoscopists to recognize these lesions is unknown; moreover, the impact of a simple training module on subsequent performance is of interest. Methods Seventy high-quality images were selected from a database of large sessile and flat polyps. Histopathology was used as the gold standard. The images were presented in randomized order to expert endoscopists and trainees. Participants were asked to predict the underlying histology. Subsequently, an educational module was delivered, and participants were asked to repeat the test using the same images in a different randomized order. Results Thirty-seven participants were included: 16 trainees and 21 experts. Baseline predictive accuracy for SSL-D was only 24.0% across the group; 18.3% for trainees and 28.2% for experts. This improved significantly, regardless of experience level, after the educational module to 69.7% ( p < 0.001) for the whole cohort; 67.5% for trainees and 71.4% for experts. Before training, SSL-D was incorrectly identified as tubular adenomas or tubulovillous adenomas in 40% of cases. Posttraining, this reduced significantly to 7.6% ( p < 0.001) for all participants. For experts, this decreased to 7.0% ( p < 0.001) and trainees to 8.3% ( p < 0.001). Conclusion SSL-D lesions are underdiagnosed by both expert and trainee endoscopists. A brief educational intervention can significantly improve the predictive accuracy and performance of experts and trainees.