Published in:
01-10-2018 | Original Article
Self-Formation Assessed by Cumulative Summation Test Does Not Reach Recommended Thresholds for Optical Diagnosis of Colorectal Polyps ≤ 7 mm
Published in: Digestive Diseases and Sciences | Issue 10/2018
Login to get accessAbstract
Background and Aims
Accurate optical diagnosis of diminutive polyps would allow implementing a resect and discard strategy. We evaluated the learning curve of a single training session followed by self-education in subjects with no endoscopic experience.
Methods
Learning curves were evaluated in 38 subjects employing learning curve–cumulative summation (LC‐CUSUM) tests, with each participant attending one training session regarding narrow band imaging and optical diagnosis and then individually assessing 100 lesions, receiving feedback after each diagnosis. Diagnostic accuracy was subsequently evaluated in 180 patients with lesions ≤ 7 mm. Evaluators predicted each polyp’s histology and recommended a surveillance interval. Determinants of accuracy were explored using regression analysis.
Results
According to the LC‐CUSUM curve, 20 evaluators (52.6%) reached diagnostic competence after 57 lesions (IQR 55–76.5). During the diagnostic performance assessment, 11,666 diagnoses and 6840 follow-up recommendations were generated. Considering high confidence diagnoses, accuracy was 81.3% (80.5–82.1%), negative predictive value (NPV) for rectosigmoid adenomas 78.6% (76.4–80.6%), and sensitivity for adenomas 86.6% (85.8–87.4%). Two (5.3%) evaluators reached a ≥ 90% accuracy, 3 (7.9%) presented a NPV for rectosigmoid adenomas ≥ 90%, and 18 (47.4%) a sensitivity for adenomas ≥ 90%. Multivariable logistic regression showed high confidence and size ≥ 5 mm as the strongest predictors of accuracy. Fifteen (39.5%) evaluators recommended a correct or reduced follow-up interval in over 90% of subjects.
Conclusions
Self‐formation after a single training session did not allow most evaluators to reach the required accuracy. LC-CUSUM tests did not identify competent evaluators. Despite these results, 86.7% of follow-up intervals would have been corrected or reduced.