GastroBot: Artificial Intelligence Applied to Bowel Preparation

Colonic Adenoma, Colonic Disease, Colonic Neoplasms, Colonic Polyp

bowel preparation, colonoscopy quality

It is estimated that about 20% of colonoscopies have inadequate preparation. (5) This is associated with lengthy procedures and less detection of adenomas, reduces the screening intervals, and increases the costs and risks of complications. Several strategies have been proposed to improve the quality of bowel preparation. Mobile healthcare Apps have been developed to increase adherence to bowel preparation agents, improving the quality of bowel preparation. However, adherence to mobile healthcare Apps is also a quality criterion and a pending problem to solve with this new technology. GastroBot is a new technology based on artificial intelligence that allows, through a software bot, to carry out a personalized follow-up of the patient's bowel cleansing, advising the patient to overcome contingencies that arise with the preparation, which in other circumstances could lead to the failure of it. The primary aim of this study is to determine the improvement in bowel preparation after GastroBot assistance compared with the traditional explanation. As a secondary aim, this study also pursues to determine adenoma and polyp detection rates (ADR and PDR, respectively), bowel preparation agents' tolerance, and GastroBot functionality.

Background Colorectal cancer (CRC) is the third most frequent tumor, the most frequent gastrointestinal tumor, and the second cause of cancer-related death. (1) In more than 80-90% of cases, CRC has a precursor lesion, an adenomatous polyp or adenoma, slowly progressing towards CRC. Colonoscopy is considered the gold standard in its prevention since it allows the detection and treatment of its initial form. (2) Considering this, several colonoscopy quality indicators have been described, such as cecal intubation rate, withdrawal time, and adenoma/polyp detection rate (ADR); the last is the most important indicator correlating with CRC risk. (3) Therefore, focusing on improving the ADR is mandatory to reduce the incidence of CRC. Many techniques have been described for this purpose, like improving endoscopists' education and training, split-dosing bowel preparations, withdrawal time \>9 minutes and right colon second view, high-definition white light endoscopy, Endocuff vision, G-EYE scope or Artificial Intelligence. (2, 4) However, all these techniques have in common the need for optimal visualization of the intestinal mucosa, which depends on bowel cleansing. (3,4) Problem It is estimated that about 20% of colonoscopies have inadequate preparation. (5) This is associated with lengthy procedures and less detection of adenomas, reduces the screening intervals, and increases the costs and risks of complications. This causes frustration for the patient and physician with medico-legal conflicts. (6) The ideal cleansing method must be safe, well-tolerated, and effective. However, none of the current options fulfills these characteristics. The main cause of inappropriate cleansing (80% of cases) is a failure to adequately follow preparation instructions and mostly because of intolerance to the oral solution. (7,8) Several strategies have been proposed to improve the quality of bowel preparation. As in other fields, mobile healthcare Apps have been developed to increase adherence to bowel preparation agents, improving quality bowel preparation. However, adherence to mobile healthcare Apps is also a quality criterion and a pending problem to solve with this new technology. Also, as with any mobile App, mobile healthcare Apps must be compatible with specific devices. GastroBot is a new technology based on artificial intelligence that allows, through a software bot, to carry out a personalized follow-up of the patient's bowel cleansing, advising the patient to overcome contingencies that arise with the preparation, which in other circumstances could lead to the failure of it. Aim The primary aim of this study is to determine the improvement in bowel preparation after GastroBot assistance compared with the traditional explanation. As a secondary aim, this study also pursues to determine adenoma and polyp detection rates (ADR and PDR, respectively), bowel preparation agents' tolerance, and GastroBot functionality. MATERIALS AND METHODS Study design Study type. The following is a cross-section simple-blind and single-center controlled randomized trial. Two groups will be established: the GastroBot-assisted bowel preparation (GB-group) and the conventional-assisted bowel preparation (C-group) group. Setting. It will be performed in consecutive patients with bowel preparation agents indication before undergoing a colonoscopy with cecal intubation at the Instituto de Gastroenterología y Endoscopía de Avanzada (IGEA), Hospital de la Asociación Médica (HAM) Dr. Felipe Glasman Bahía Blanca, Buenos Aires province, Argentina. The study protocol and consent form have been approved by the Institutional Review Board (IRB) and will be conducted according to the declaration of Helsinki. Patients will sign an informed consent. Intervention A clinical coordinator will be responsible for patients' randomization. Patients from both study groups will receive the same type of preparation with polyethylene glycol in split dose, establishing the intake time according to three-time segments (8-11 am, 11-2 pm, 2-4 pm). The C-group will receive the instructions in writing without prior personalized advice. The GB group will receive the instructions through the WhatsApp application, guided by the software bot with multiple and personalized alternative instructions according to results. The endoscopist will perform the endoscopy by assessing primary and secondary endpoints, blinded to the patient's study group. Sample size Considering the proportion of insufficient BBPS (\<6) among the App-group (7.7%) vs. controls (16.9%) described by Walter B et al. (2021), a sample size of 194 cases per study group was estimated to determine a two-sided difference on BBPS between GB-group vs. C-group with an 80% statistical power. Statistical analysis Baseline characteristics will be compared between the case and control group using Chi-square o Fisher test for categorical variables and Mann-Whitney U or Student's t-test for continuous variables. A P value of less than 0.05 will be considered statistically significant. All the statistical analysis will be performed using the latest version of the statistical program R (R Foundation for Statistical Computing; Vienna, Austria). Limitations The protocol will be performed in only one center and by six endoscopists. It is a simple blind study. The patients will know they are using (or not) a novel instrument to increase bowel preparation quality.

No related papers found yet.

Argentina