Using artificial intelligence (AI) during screening colonoscopy could be a cost-savings strategy that could also boost the prevention of colorectal cancer (CRC) incidence and mortality, a researcher reported.
Among a simulated cohort of patients at average risk for CRC, and compared with no screening, the relative reduction of CRC incidence was 44.2% with screening colonoscopy without AI tools and 48.9% with screening colonoscopy with AI tools, for a 4.8% incremental gain. Additionally, compared with no screening, the relative reduction in CRC mortality was 48.7% for screening colonoscopy with no AI versus 52.3% for screening colonoscopy with AI, for a 3.6% incremental gain, reported Yuichi Mori, MD, of the Showa University Yokohama Northern Hospital in Japan.
AI detection tools also were associated with a cost savings of $57 per patient after lowering discounted costs from $3,400 to $3,343, which persisted in secondary colonoscopy modeling analysis, Mori said in a presentation at Digestive Disease Week (DDW). The findings were simultaneously published in the Lancet Digital Health†
“At the US population level, the implementation of AI detection during screening colonoscopy resulted in yearly additional prevention of 7,194 [CRC] cases and 2,089 related deaths, and a yearly saving of US $290 million,” Mori and colleagues wrote.
“We are able to find a cost reduction as a whole, which is very surprising because the primary use of AI increases the costs, but it can be done by the prevention effect of colorectal cancer,” Mori said at a DDW press conference.
“The use of AI for polyp detection primarily increases the costs because it can increase the detection of polyps, adenomas, and it may increase the number of polypectomies, and subsequently it can increase the number of surveillance colonoscopies,” Mori explained. “However this kind of increment can be mitigated by the benefits coming from the use of AI, namely the cancer prevention effect with the increment of the ADR [adenoma detection rates] by AI. So it is quite important how the use of AI contributes to the healthcare system in terms of cost effectiveness.”
The study had some limitations, including the fact that the authors assumed “a linear relationship between the cancer prevention effect and increased ADR, [and] there is an ongoing discussion about whether there is a threshold effect or ADR in cancer prevention.” Also, the authors “assumed the same increase of the detection rate of high-risk adenomas as low-risk adenomas under the use of AI for polyp detection, although the detection rate of advanced adenomas was not shown in [a] previous meta-analysis†
Mori and colleagues used Markov model microsimulation in a hypothetical cohort of 100,000 US patients who underwent colonoscopy screening with or without AI every 10 years, starting at age 50 and ending at age 80. Patients did not have a personal or family history of CRC, adenomas , inflammatory bowel disease, or hereditary CRC syndrome.
“Costs of AI tools and cost for downstream treatment of screening detected disease were estimated with 3% annual discount rates,” they stated.
The authors reported that, based on an assumption of 60% screening uptake, screening colonoscopy cut CRC incidence from 6.0% cases per 100,000 to 3.3% cases per 100,000, which corresponded to an absolute reduction of 2,638 cases per 100,000 and a 44.2% relative reduction versus no screening. Compared with colonoscopy without AI, the implementation of AI further reduced CRC incidence from 3.3% to 3.0% cases per 100,000 people, and CRC mortality from 1.2% to 1.1% per 100,000 people, they stated.
“This corresponds to an additional 0.3% absolute reduction (8.4% relative reduction) in [CRC] incidence and 0.1% absolute reduction (6.9% relative reduction) of [CRC] mortality, compared with colonoscopy without AI,” the authors said.
Mori’s group also found that AI further decreased costs related to CRC treatment by 8.2%, from $1,636 to $1,502 per individual, although this “was partly offset by the cost of AI implementation that increased screening costs from $1,764 to $1,841 per person (also including surveillance colonoscopies and adverse events treatment).
“I’d say use of AI during screening colonoscopy may be cost effective in the US setting,” Mori concluded.
Mori said his group has plans for a large randomized trial in Europe and Japan, with long-term follow-up of CRC incidence as the primary endpoint.
The study was funded by the European Commission and the Japan Society of Promotion of Science.
Mori disclosed relationships with Olympus and Cybernet System. Co-authors disclosed multiple relationships with industry.