Can Artificial Intelligence Reduce Consumption of Standard High Volume Bowel Preparation Regimen Among Older Population, Without Compromising the Quality of Colonoscopy? An International Multi-centre Randomized Controlled Trial.
Can Artificial Intelligence Reduce Consumption of Standard High Volume Bowel Preparation Regimen Among Older Population, Without Compromising the Quality of Colonoscopy? An International Multi-centre Randomized Controlled Trial.
The goal of this study is to assess the clinical and cost-effectiveness of the AI bowel preparation evaluation system in reducing the consumption volume of the standard high-volume PEG regimen before colonoscopy among the older population.
Researchers will compare bowel preparation adequacy and other colonoscopy quality matrices, and patient satisfaction between patients using the AI system and those using Standard Practice.
Colonoscopy is the gold standard for screening colorectal cancer (CRC), which is the third most common cancer and the second leading cause of cancer-related mortality worldwide. Polypectomy during screening colonoscopy reduces both the long-term incidence and mortality of CRC because the removal of adenomatous polyps, the precursors of CRC, prevents the development of CRC by interrupting the adenoma-carcinoma sequence. As the older population grows, the demand for colonoscopy is increasing rapidly because of indications for treatment, diagnosis and surveillance, as well as the wide global implementation of organized CRC screening program for individuals ≥50 years.
Bowel preparation plays a crucial role in colonoscopy because it directly affects two important quality indicators that contribute to procedural accuracy: cecal intubation rate (CIR) and adenoma detection rate (ADR). Adequate bowel preparation is essential for complete visualization of the colonic mucosa and the detection of colorectal lesions. In contrast, inadequate bowel preparation (IBP) is associated with a lower CIR, longer procedural time, increased risk of complications, higher adenoma miss rate, and increased healthcare costs owing to the need for earlier repeat colonoscopy. These negative consequences place a significant burden on both patients and the healthcare system. IBP is a common issue worldwide, with rates reported to range from 11% to 28% in the general population undergoing colonoscopy and up to 50% in older individuals.
Enhanced bowel preparation instructions are recommended16. Effective strategies include providing additional instructions, visual aids, cartoons, booklets, videos, phone calls, short message services, smartphone applications and mobile messenger.
Over the past decade, there has been an exponential increase in the computational power, reduced data storage costs, improved algorithmic sophistication, and an increased availability of electronic health data. Artificial intelligence (AI) has been widely adopted in various healthcare settings, particularly for colonoscopy17.
A recent meta-analysis reported that brown liquid rectal effluent is one of the most significant risk factors for IBP, increasing the odds by more than 4.5-fold.18. Some newly developed convolutional neural network (CNN)-based AI models, trained using thousands of rectal effluent images, have been validated in randomized controlled trials (RCTs) as effective tools for guiding bowel preparation before colonoscopy19,20. Unlike enhanced instructions that focus on patient education, these AI models allow patients to predict the adequacy of bowel preparation before colonoscopy by analyzing rectal effluent images.
In one study, the AI model demonstrated comparable performance in predicting bowel preparation adequacy to standard practice (SP) of self-evaluation using written instructions with photographic examples (AI: 90.7% vs SP: 91.5%, p=0.976), while patients in AI group achieved a higher mean Boston Bowel Preparation Scale (BBPS) score (7.32±1.4 vs 7.16±1.46, p=0.044)19. In another study, the AI model achieved a higher bowel preparation adequacy rate (88.54% vs 65.59%, p <0.001) and a higher mean BBPS score (6.74±1.25 vs 5.97±1.81, p <0.001) than those in the SP 20. These AI models were developed as mobile apps or websites accessible via smartphones. In HK, the overall smartphone ownership rate has rapidly increased to 93% recently: 73% among individuals aged ≥65 years versus 99% among those aged 45-64 years21. The social media participation rate was 83% across age groups and 78% among those aged ≥45 years22. The AI bowel preparation evaluation system only requires patients to take and upload photographs of their rectal effluent, with the evaluation results provided immediately on the same page, making it simpler to use than social media. This indicates the feasibility of using a smartphone-based AI bowel preparation evaluation system, even in the older population. Caregivers can assist older individuals with low digital literacy with AI evaluations. For those who live alone, have limited mobility, or have advanced medical conditions, the usual practice is to admit them for inpatient colonoscopy. This means that ward nurses can perform AI evaluations.
Although two studies reported the effectiveness of AI in improving the bowel preparation quality, there are several important gaps in the existing research. First, one study had a relatively small sample size of approximately 500 patients20, and both studies were conducted in a single country, limiting the generalizability of their findings. Second, the baseline ADR in the SP group in both studies did not meet the international quality benchmark of 25%23,24. In one study, the baseline bowel preparation adequacy in the SP group was only 65.6%, and the mean BBPS score was 5.97, both of which were below the recommended standard23,25. Third, the same study allowed for a 1 L remedial dose on top of the split 3 L PEG regimen if the rectal effluent was deemed IBP by the patient's or AI evaluation. As a result, more patients in the AI group consumed a total of 4 L PEG compared to those in the SP group20. It is uncertain that if the increase in bowel preparation adequacy rate in the AI group was due to an increased PEG dosage or AI use. Fourth, although both studies reported that AI could enhance bowel preparation adequacy, they failed to demonstrate its efficacy in enhancing adenoma detection during colonoscopy. Finally, most of the patients recruited in these two studies did not have high-risk factors for IBP as they were generally young, healthy, free of advanced medical conditions, and had undergone elective outpatient colonoscopies19,20.
Although there are different bowel preparation regimens, high-volume PEG is the most commonly used, particularly for the older population, because of its high safety and efficacy in achieving adequate bowel preparation quality16,26,27. Older population is also at a higher risk of developing IBP18. One reason for this is that they tend to dislike and poorly tolerate high-volume PEG regimens with unpleasant taste9,27. Thus, there is a clear need to improve the acceptance and tolerability of high-volume PEG regimens in the older population.
To the best of our knowledge, no studies have explored the feasibility and effectiveness of AI bowel preparation evaluation in reducing the consumption of high-volume PEG while maintaining bowel preparation adequacy among older populations who are at higher risk of IBP18 and require standard high-volume PEG for safety reasons16,26,27. Furthermore, the cost-effectiveness of AI bowel preparation evaluation before colonoscopy has not yet been investigated. Therefore, we aimed to conduct a large-scale, international, multi-centre randomized controlled trial (RCT) to assess the clinical and cost-effectiveness of an AI bowel preparation evaluation system in reducing the consumption volume of the standard high-volume PEG regimen among older population, without compromising the bowel preparation adequacy and other colonoscopy quality matrices.
Inclusion Criteria:
Exclusion Criteria:
felixsia@cuhk.edu.hk852-26370428