COLO-DETECT: Can an Artificial Intelligence Device Increase Detection of Polyps During Colonoscopy?

COLO-DETECT: A Randomised Controlled Trial of Lesion Detection at Colonoscopy Using the GI Genius Artificial Intelligence Platform

COLO-DETECT is a clinical trial to evaluate whether an Artificial Intelligence device ("GI Genius", manufactured by Medtronic) can identify more polyps (pre-cancerous growths of the bowel lining) during colonoscopy (large bowel camera test) than during colonoscopy without it.

Colorectal cancer is common, affecting 1 in 15 men and 1 in 18 women in the UK in their lifetime. Many colorectal cancers develop from polyps via the adenoma-carcinoma sequence: there is a pre-cancerous stage (adenoma) during which it is possible to remove the polyp and therefore prevent it from progressing to colorectal cancer. The gold standard tool for doing this is colonoscopy. However, colonoscopy does not pick up all polyps, particularly flat polyps. Missed polyps can result in colorectal cancer, so it is imperative to detect and remove as many polyps as possible. Many different interventions have been introduced to improve polyp detection, the most recent of which is artificial intelligence devices. GI Genius is an artificial intelligence device which integrates with existing colonoscopy equipment and analyses the video feed from the colonoscope camera in real time. Any areas that may represent an abnormality are then highlighted (without any lag) within a green box, alerting the colonoscopist to its presence. The potential abnormality can then be assessed more closely by the colonoscopist to decide whether it needs to be removed or not. COLO-DETECT is a 2-arm, prospective, randomised controlled trial to assess whether GI Genius is able to detect more polyps (specifically, adenomas) during colonoscopy than standard colonoscopy without GI Genius. The primary outcome will be the mean number of adenomas per procedure (MAP) and the key secondary outcome will be the proportion of colonoscopies in which one or more adenomas is detected (Adenoma Detection Rate - ADR). These are both important quality markers for colonoscopy; the study will be powered to detect a clinically meaningful difference in ADR, which will by default detect a meaningful difference in MAP as the sample size required for ADR is larger. In addition to measuring the effect of GI Genius on polyp detection, COLO-DETECT will provide a health economics analysis concerning the use of GI Genius, perform long-term passive follow-up to examine for future outcomes related to colorectal polyps and colorectal cancer, and perform additional nested studies (subject to ethical approval) that examine the effect upon users (for example through a visual scanning study) and their experience of using the GI Genius.

Colonic Polyp, Colorectal Polyp, Colorectal Adenoma, Colorectal Adenomatous Polyp, Colorectal SSA, Sessile Serrated Adenoma, Sessile Colonic Polyp

The patient and colonoscopist (care provider) cannot be blinded to the allocation. The investigator and statistician will remain blinded to the allocation until analysis has been conducted.

In the GGC arm, participants will undergo colonoscopy as per standard care for the unit where they are having their procedure, except that at some point prior to commencing withdrawal of the colonoscope, a member of the endoscopy staff will turn on the GI Genius machine. This will remain operational from the time it is switched on until the end of the procedure.

In the SC arm, participants will undergo colonoscopy as per standard care for the unit where they are having their procedure.

Participants will undergo diagnostic colonoscopy, which will be identical to the normal standard of care at the unit where they are undergoing their procedure, except that GI Genius will be turned on during the procedure.

Diagnostic colonoscopy will be performed as per the standard of care for the unit where the patient is having their procedure.

Number of adenomas per participant detected at colonoscopy as indicated by the Mean Number of Adenomas per Procedure (MAP)

The number of adenomas identified during each colonoscopy will be summed and divided by the total number of colonoscopies performed. MAP is usually expressed as a number to one decimal place (e.g. 1.2).

Proportion of participants in whom at least one adenoma is detected at colonoscopy, as indicated by the Adenoma Detection Rate (ADR)

Whether or not at least one adenoma is detected at colonoscopy will be determined for each participant. The number of colonoscopies where one or more adenomas is identified will be divided by the total number of colonoscopies to give the ADR. ADR is usually expressed as a percentage.

Number of adenomas per participant detected at colonoscopy in the 'screening' participant population, as indicated by MAP for that participant population.

The number of adenomas identified during each colonoscopy within the 'screening' participant population will be summed and divided by the total number of colonoscopies in that participant population. The MAP for the 'screening' participant population within each study arm will be compared

Number of adenomas per participant detected at colonoscopy in the 'symptomatic' participant population, as indicated by MAP for that participant population

The number of adenomas identified during each colonoscopy within the 'symptomatic' participant population will be summed and divided by the total number of colonoscopies in that participant population to calculate MAP. The MAP for the 'symptomatic' participant population within each study arm will be compared

Proportion of participants in the 'screening' participant population in whom at least one adenoma is detected at colonoscopy, as indicated by ADR for that participant population

Whether or not at least one adenoma is detected at colonoscopy will be determined for each participant within the 'screening' participant population. The number of colonoscopies where one or more adenomas is identified will be divided by the total number of colonoscopies in that participant population to calculate ADR. The ADR for the 'screening' participant population within each study arm will be compared

Proportion of participants in the 'symptomatic' participant population in whom at least one adenoma is detected at colonoscopy, as indicated by ADR for that participant population

Whether or not at least one adenoma is detected at colonoscopy will be determined for each participant within the 'symptomatic' participant population. The number of colonoscopies where one or more adenomas is identified will be divided by the total number of colonoscopies in that participant population to calculate ADR. The ADR for the 'symptomatic' participant population within each study arm will be compared

Number of polyps per participant detected at colonoscopy, as indicated by the Mean number of Polyps per Procedure (MPP)

The total number of polyps detected during each colonoscopy will be summed, and divided by the total number of colonoscopies, to calculate MPP. MPP is usually expressed as a number to one decimal place.

Number of polyps per participant detected at colonoscopy in the 'screening' participant population, as indicated by the Mean number of Polyps per Procedure (MPP)

The total number of polyps detected during colonoscopy for each participant within the 'screening' participant population. will be summed, and divided by the total number of colonoscopies in that participant population, to calculate MPP. MPP is usually expressed as a number to one decimal place.

Number of polyps per participant detected at colonoscopy in the 'symptomatic' participant population,as indicated by the Mean number of Polyps per Procedure (MPP)

The total number of polyps detected during colonoscopy for each participant within the 'symptomatic' participant population will be summed, and divided by the total number of colonoscopies in that participant population, to calculate MPP. MPP is usually expressed as a number to one decimal place.

Proportion of participants in whom at least one polyp is detected at colonoscopy, as indicated by Polyp Detection Rate (PDR)

Whether or not at least one polyp is detected at colonoscopy will be determined for each participant. The number of colonoscopies where one or more polyps is detected will be divided by the total number of colonoscopies in that participant population to calculate PDR, which is normally expressed as a percentage.

Proportion of participants in the 'screening' participant population in whom at least one polyp is detected at colonoscopy, as indicated by Polyp Detection Rate (PDR)

Whether or not at least one polyp is detected at colonoscopy will be determined for each participant within the 'screening' participant population. The number of colonoscopies where one or more polyps is identified will be divided by the total number of colonoscopies in that participant population to calculate PDR, which is normally expressed as a percentage.

Proportion of participants in the 'symptomatic' participant population in whom at least one polyp is detected at colonoscopy, as indicated by Polyp Detection Rate (PDR)

Whether or not at least one polyp is detected at colonoscopy will be determined for each participant within the 'symptomatic' participant population. The number of colonoscopies where one or more polyps is identified will be divided by the total number of colonoscopies in that participant population to calculate PDR, which is normally expressed as a percentage.

The location, size, and morphology of the polyps identified (and histology if retrieved) in each study arm will be compared. This will also be analysed for both the screening and symptomatic participant populations in each study arm.

Assessed over duration of colonoscopy procedure and at time of 14 day post-colonoscopy review (once histology is known)

The number of colonoscopies in each study arm in which one or more SSPs is identified, divided by the total number of colonoscopies in each arm. This will also be analysed for both the screening and symptomatic participant populations in each study arm.

The number of CRCs detected in each study arm divided by the total number of colonoscopies in each arm. This will include polyps removed and later found to cancerous on histology and lesions felt to be cancerous at the time of colonoscopy. This will also be analysed for both the screening and symptomatic participant populations in each study arm.

The number of AAs detected in each study arm divided by the total number of colonoscopies in each arm. This will also be analysed for both the screening and symptomatic participant populations in each study arm.

Caecal intubation rate (the proportion of colonoscopies in which the colonoscope reaches the furthest extent of the colon) will be compared between the study arms to assess for non-inferiority

Insertion time to caecum (time taken to reach the furthest point of the large bowel) will be compared between the study arms to assess for non-inferiority

Total time required to perform the colonoscopy will be compared between the study arms to assess for non-inferiority

Total withdrawal time (time taken to remove the colonoscope from the furthest point of the colon) in the absence of any polyps will be compared between the study arms to assess for non-inferiority

Colonoscopist-assessed patient comfort scores will be compared between the study arms to assess for non-inferiority

Nurse-assessed patient comfort scores will be compared between the study arms to assess for non-inferiority

A validated Patient-Reported Experience Measure (Newcastle ENDOPREM) will be used to compare patient experience of colonoscopy between study arms

The EuroQoL EQ-5D-5L (validated quality of life questionnaire) will be used to compare patient-reported health-related quality of life, between study arms

The need for further colonoscopy for each participant is determined by the findings at the index colonoscopy, according to national guidelines on polyp surveillance. This may differ between study arms if more polyps are identified in one arm.

Some colonoscopists partake in the national Bowel Cancer Screening Programme (BCSP) and some do not. MAP will be analysed by colonoscopist status within each study arm.

Some colonoscopists partake in the national Bowel Cancer Screening Programme (BCSP) and some do not. ADR will be analysed by colonoscopist status within each study arm.

Change in number of adenomas detected per colonoscopy, for each colonoscopist, over the course of the study, as indicated by MAP

MAP for the first 20 percent of participants will be compared to MAP for the last 20 percent of participants scoped by each participating colonoscopist, to assess for change over the course of the study.

Change in proportion of participants in whom at least one adenoma is detected during colonoscopy, for each colonoscopist, over the course of the study, as indicated by ADR.

ADR for the first 20 percent of participants will be compared to ADR for the last 20 percent of participants scoped by each participating colonoscopist, to assess for change over the course of the study.

Change in number of adenomas detected per participant, for each participating colonoscopist, from pre-study to intra-study (SC arm only)

MAP may vary from baseline, even in the control arm due to a contamination or learning effect; comparing baseline values to those during the study assesses for this effect.

Proportion of participants in whom at least one adenoma is detected during colonoscopy, for each participating colonoscopist, from pre-study to intra-study (SC arm only)

ADR may vary from baseline, even in the control arm due to a contamination or learning effect; comparing baseline values to those during the study assesses for this effect.

Equipment, staff, histology, unplanned admission, and other related costs will be calculated and used to determine cost-effectiveness of GGC versus SC.

Costs associated with each participant's procedure and care will be calculated at time of 14-day review

Number of adenomas per participant detected at colonoscopy, amongst colonoscopists not participating in the study, as indicated by MAP

MAP values over the duration of the study, for colonoscopists not participating in the study but performing colonoscopy at study sites, will assist with baseline comparisons. These data are reported by endoscopy units as part of the normal endoscopy quality assurance programme.

Proportion of participants in whom at least one adenoma is detected at colonoscopy, by colonoscopists not participating in the study, as indicated by ADR

ADR values over the duration of the study, for colonoscopists not participating in the study but performing colonoscopy at study sites, will assist with baseline comparisons. These data are reported by endoscopy units as part of the normal endoscopy quality assurance programme.

Inclusion Criteria: Able to give informed consent Patients attending for colonoscopy Through standard National Health Service (NHS) care (most commonly due to iron deficiency anaemia, altered bowel habit, weight loss, rectal bleeding, positive FIT (faecal immunohistochemical test) based on symptoms, those referred on basis of family history, abnormal cross- sectional imaging, polyp surveillance or post CRC surveillance) Through Bowel Cancer Screening Programme (FIT positive, surveillance) Colonoscopy to be performed by colonoscopist trained to perform GGC as part of the study Exclusion Criteria: Absolute contraindications to colonoscopy Patients lacking capacity to give informed consent Confirmed or expected pregnancy Established or suspected large bowel obstruction or pseudo-obstruction Known presence of colorectal cancer or polyposis syndromes Known colonic strictures (meaning that the colonoscopy maybe incomplete) Known active colitis (ulcerative colitis, Crohn's colitis, diverticulitis, infective colitis) Inflammatory Bowel Disease (IBD) surveillance procedures Patients who are on clopidogrel, warfarin, or other antiplatelet agents or anticoagulants who have not stopped this for the procedure (as polyps cannot be removed and thus histology cannot be confirmed) Patients who are attending for a planned therapeutic procedure or assessment of a known lesion Patients referred with polyps identified on Bowel Scope procedure

Seager A, Sharp L, Hampton JS, Neilson LJ, Lee TJW, Brand A, Evans R, Vale L, Whelpton J, Rees CJ. Trial protocol for COLO-DETECT: A randomized controlled trial of lesion detection comparing colonoscopy assisted by the GI Genius artificial intelligence endoscopy module with standard colonoscopy. Colorectal Dis. 2022 Oct;24(10):1227-1237. doi: 10.1111/codi.16219. Epub 2022 Jun 28.

COLO-SPEED Platform website,containing details of the COLO-SPEED network, studies linked to it and recent articles related to colorectal cancer research.