Colorectal cancer can be effectively prevented by removal of pre-malignant polyps during colonoscopy. Large (\>=20mm) non-pedunculated colorectal polyps (LNPCPs) require careful assessment before treatment. If submucosal invasive cancer (SMI) is present, it determines whether endoscopic treatment can be curative or whether surgery is needed. Current classification systems to detect SMI are complex, require extensive training, and are underused in non-tertiary hospitals.
A simple web-based clinical decision support tool was created using well-established parameters (presence of a demarcated area, polyp size, Paris classification, location, and granularity) to identify SMI within LNPCPs. Crucially, the tool uses only standard endoscopic imaging available in most endoscopy units.
This prospective multi-centre study evaluates the accuracy of the tool during live endoscopic assessment. Endoscopists of varying experience will assess 10 large colorectal polyps using the tool, then undergo a 45-minute structured educational intervention. They will then assess a further 10 polyps using the tool. A third set of 10 assessments at 3 months evaluates durability of learning. Accuracy is compared to expert opinion.
Gastrointestinal endoscopists of varying abilities and grades
Endoscopists who have not previously encountered the clinical decision support tool Patient participants
Referred for colonoscopy with appropriate bowel preparation
Colorectal polyp >=20mm in size detected or referred for resection
Does not consent to inclusion
Does not undergo the learning intervention
No connection with endoscopy in gastroenterology
Does not consent to data collection for the study
Patients with large (≥20mm) non-pedunculated colorectal polyps (LNPCPs) detected during colonoscopy or referred for endoscopic resection. Patients with sessile serrated lesions are excluded. Endoscopists must consent and complete the learning intervention. Patients must provide written informed consent for data collection. Participation is voluntary for both groups.
Cohort
Prospective
Not listed
Accuracy of Endoscopic Assessment of the Risk of Submucosal Invasive Cancer Within LNPCPs Using the Clinical Decision Support Tool Compared to Expert Opinion During Live Endoscopy
The accuracy of each endoscopist's assessment of SMI risk using the tool, compared to expert opinion applied to a standardised video of the same procedure. Measured as sensitivity, specificity, and overall accuracy across pre-training, post-training, and 3-month durability phases.
Percentage of LNPCP Assessments in Which a Demarcated Area Is Correctly Detected Using the Clinical Decision Support Tool, Compared to Expert Opinion
Endoscopists assess presence/absence of a demarcated area on a standardised endoscopic video using the SMI clinical decision support tool. Expert opinion is the reference standard (considered 100% accurate). Accuracy = proportion of assessments concordant with the expert.
Inter-observer Agreement Among Endoscopists in Detection of a Demarcated Area Within LNPCPs, Measured by Fleiss' Kappa
Agreement among the multiple participating endoscopists rating the same standardised endoscopic videos, quantified by Fleiss' kappa (appropriate for 3+ raters; corrects for chance agreement).
Percentage of Demarcated Areas in Which the JNET Classification Is Correctly Determined, Compared to Expert Opinion
Endoscopists assign the JNET (Japan NBI Expert Team) classification to the demarcated area within each LNPCP. Expert opinion is the reference standard (considered 100% accurate). Accuracy = proportion of assessments in which the endoscopist's JNET category matches the expert reference.
Percentage of Lesions in Which Each Morphological Component (Size, Location, Granularity, Paris Classification) Is Correctly Determined, Compared to Expert Opinion
Each morphological component is assessed by the endoscopist and compared to the expert reference standard (considered 100% accurate). Accuracy = proportion of assessments concordant with the expert, reported separately for size, location, granularity, and Paris classification.
Click on the lead sponsor to view all of their trials.
