Percentage Visualized Mucosa as a Marker for the Quality of Colonoscopy

Summary

Participation Requirements

Description

PROJECT OBJECTIVES AND DESIGN 1.1 Hypothesis Primary hypothesis: The percentage of mucosa visualized during the withdrawal of a colonoscopy with reference to the position in the colon by means of the scope-guide is an appropriate parameter for assessment of the quality of colonoscopy and correlates ...

PROJECT OBJECTIVES AND DESIGN 1.1 Hypothesis Primary hypothesis: The percentage of mucosa visualized during the withdrawal of a colonoscopy with reference to the position in the colon by means of the scope-guide is an appropriate parameter for assessment of the quality of colonoscopy and correlates with the probability of the detection of adenomas. Secondary hypothesis: The investigators hypothesize that the percentage of visualized mucosa differs according patient specific parameters (e.g. gender, age, BMI, bowel preparation), procedure specific parameters (e.g. patient position, device, buscopan, fentanyl, withdrawal time) and investigator specific parameters (professional experience as resident and consultant). STUDY DESIGN AND PROCEDURES 2.1 Study Design: The investigators will perform a non-interventional, single center exploratory study. Only total colonoscopies will be included. All of the examinations will be performed by board-certified specialists or fellows in training, supervised by the former. The investigators will enroll only participants who will have an explicit indication, in other words who will get a colonoscopy anyway. Therefore, colonoscopies will be conducted for a wide variety of indications including, but not limited, to CRC screening, CRC/ adenoma surveillance, abdominal symptoms such as pain, irregular defecation (diarrhea or constipation), gastrointestinal occult or overt bleeding, family history of polyps or CRC, prior colonic resection, hereditary polyposis syndromes and inflammatory bowel diseases (IBD). Since the study does not pose any additional risk to the participants (the only thing which changes for the patient is that the endoscopic images from colonoscopy and scope guide will be filmed), the enrollment can be done during consultation hour, at the ward or shortly before a colonoscopy. 2.2 Recruitment, screening and informed consent procedure: Patients will be recruited by the project team in the Department of Visceral Surgery and Medicine at the Inselspital. The patient will be provided with information regarding the study. If the patient shows interest, the investigator will explain the study protocol, answer questions and check inclusion and exclusion criteria. For this study, no compensatory fee will be paid. 2.3 Study procedures: If the patient agrees to participate in the study and has signed the informed consent the colonoscopy will be performed in the usual setting and using general safety measures. The endoscopist will start filming all the phases of the colonoscopy and the position of the scope-guide simultaneously. Both films will be uploaded to the local server and then cut and analyzed with the help of iMovie. The only study procedure specific for the study is the filming and no other study specific measures will be taken. Basic parameters of patient history and colonoscopy procedure will be recorded (compare CRF). As always, any findings of the colonoscopy will be reported to the participant/ patient after the procedure. 2.4 Withdrawal and discontinuation: If an individual withdraws informed consent after the colonoscopy or during the analysis of the data, he/she will be withdrawn from the project. All videos will be anonymized. In case of withdrawal, patient data and videos will not be considered for study purposes and will be deleted completely. STATISTICS AND METHODOLOGY 3.1 Evaluation of Videos: All videos will be anonymized. Then the percentage of visible mucosa and lumen during withdrawal will be determined in each video. The scope-guide will allow to correlate the images with the position of the colonoscope within the colon. The investigators will record the amount of time spend in colon segments such as the cecum, ascending colon, right flexure, right transverse colon, left transverse colon, left flexure, descending colon, sigma and rectum. For each of these segments, the investigators will record the amount of time with good mucosa visibility and poor mucosa visibility. In addition to manual analysis, computer-aided analysis and artificial intelligence methods will be developed to assist the evaluation of the videos. The development of computer-aided methods consists of two phases. In the first phase, frames in the video will be extracted as individual images. Experienced physicians will be assigned to annotate the images. The position of the scope-guide will be extracted for the annotation. Then a classifier based on deep neural network will be trained on these annotated images. The percentage of visible mucosa and lumen during withdrawal will be determined in each video based on the classification results of each individual frame. The classification according to the scope-guide will allow to correlate the images with the position of the colonoscope within the colon. The amount of time spending in colon segments such as the cecum, ascending colon, right flexure, right transverse colon, left transverse colon, left flexure, descending colon, sigma and rectum will be determined based on deep learning prediction results. Each frame will be classified to good mucosa visibility and poor mucosa visibility and the amount of time with good mucosa visibility and poor mucosa visibility of each of the segments will be determined accordingly. In the second phase, videos will be analysed directly with deep learning methods and the consistency between the frames will be explored during the analysis to improve the accuracy. The investigators will also analyse advancements of the endoscope from the anus to the cecum. To this end, the time until each of the segments is reached will be recorded as well as all loops which have formed within the colon. 3.2 Statistical analysis plan 3.2.1. Statistical evaluation of the primary endpoint For evaluation of the primary endpoint (adenoma detection rate as a function of mucosa visibility) the investigators will rank all investigations regarding mucosa visibility and will divide the study population in 50% individuals with the best mucosa visibility and compare this number to the 50% of individuals with the worst mucosa visibility. They will compare the fraction of individuals with at least 1 detected adenoma to individuals without a detected adenoma (Fisher exact test). They will subsequently use multivariate logistic regression to correct for confounders (age, gender, family history of CRC, personal history of adenomas, advanced adenomas, carcinomas, BMI and bowel preparation). 3.2.2. Power analysis Our study is exploratory and no proper power analysis is possible. To provide an estimation the investigators assume that the effect of poor vs. good mucosa visibility will be as good as the effect of one of a number of technical devices recently tested in clinical studies to improve adenoma detection during colonoscopy. Such new technical devices (NTDs) include mucosa caps attached to the tips of the endoscope which flatten and stretch the mucosa or lenses to improve the field and angle of view. In a recent meta-analysis, NTDs increased the odds of adenoma detection by 35% (odds ratio 1.35) 25. In a screening colonoscopy, an adenoma detection rate of at least 20% would be expected 26 (i.e. 50 individuals with at least one adenoma among 250 individuals). In individuals with optimal mucosa visibility, this number would be 35% higher (i.e. 68 in 250 individuals). According to the power analysis using G* Power 3.1 27, 419 individuals would be needed to detect such a difference (i.e. 20% vs. 27% adenoma detection) with a power of 80% at an alpha level of 5%. The Investigators are aiming to include 500 individuals to account for an estimated 15% of investigations which could not be evaluated due to incomplete examinations or technical problems. 3.2.3 Statistical analysis of the secondary endpoints For the secondary outcomes, observed effects and interactions will be statistically substantiated by parametric and non-parametric tests and/ or multivariate analyses considering confounders as appropriate. A p-value <0.05 will be considered significant. 3.3 Handling of missing data: The nature of this exploratory study allows for some tolerance regarding missing data. Patients with incomplete or missing data/videos due to technical problems will be excluded from the study.

Tracking Information