Colorectal surveillance interval assignment based on in vivo prediction of polyp histology: impact of endoscopic quality improvement program.

BACKGROUND: Validated training methods are needed before in vivo optical diagnosis can be used to guide colorectal cancer surveillance interval assignments.

OBJECTIVES: To assess the impact of endoscopist training on accuracy of surveillance intervals based on in vivo optical diagnosis.

DESIGN: Prospective, randomized study.

SETTING: Tertiary referral center.

PATIENTS: A total of 1231 patients with colorectal polyps (before training, n = 577; after training, n = 654).

INTERVENTIONS: Endoscopists were randomized to either participate or not participate in two 1-hour training sessions. Endoscopists' in vivo pathology predictions were recorded for all resected polyps. The accuracy of in vivo prediction-guided surveillance intervals were compared with pathology-guided intervals as the criterion standard. Surveillance interval accuracy before and after training were analyzed.

MAIN OUTCOME MEASUREMENTS: Surveillance interval accuracy based on in vivo polyp predictions with pathology-guided intervals as criterion standard before and after a training intervention.

RESULTS: Optically predicted surveillance intervals for the trained group of endoscopists had an overall accuracy of 84% (242/287) before training and 82% (261/317) after training. There was little evidence to suggest that training had a meaningful impact on the accuracy of optically predicted surveillance intervals (odds ratio 1.20; 95% CI, 0.76-1.89, P = .44).

LIMITATIONS: Limitations include small sample size, lack of endoscopist prediction confidence levels, and optional use of narrow-band imaging.

CONCLUSIONS: Surveillance interval accuracy did not meet published thresholds (>90%) to use in vivo optical diagnosis without pathologic confirmation. Larger studies, focused specifically on the impact of training, are needed to validate methods to improve polyp diagnosis and surveillance predication accuracy.