Artificial intelligence (AI) detection tools in screening colonoscopy can lower costs and aid in preventing colorectal cancer (CRC) incidence and mortality, according to study results published in The Lancet.
For the analysis, researchers used a Markov model to simulate performing a colonoscopy with and without AI for CRC screening. A total of 100,000 individuals in a hypothetical cohort were assigned average risk for CRC. The hypothetical study participants had no family or personal history of CRC, inflammatory bowel disease, adenomas, or hereditary CRC syndrome. Primary analysis examined colonoscopy with vs without AI to aid in polyp detection every 10 years between ages 50 and 80 years, with follow-up until 100 years. Secondary analyses included a model of once-in-life screening colonoscopy at age 65 years.
The simplified current guideline was used to guide postpolypectomy surveillance. Estimations of the costs of AI tools and downstream treatment of screening any detected disease were made using a 3% annual discount rate. The primary outcomes were the comparative incremental effect of colonoscopy with vs without the aid of AI on the incidence and mortality rate of CRC, as well as the cost-effectiveness of AI-assisted screening.
According to researchers, compared with no screening, AI-aided screening was associated with a relative reduction of 48.9% in incidence of CRC, and screening without AI with a relative reduction of 44.2% (4.8% incremental gain). Similarly, AI-aided screening was associated with a 52.3% relative reduction in CRC mortality compared with no screening, and screening without AI with a relative reduction of 48.7% (1.3% absolute reduction in mortality). The use of AI detection tools reduced CRC treatment-related cost per screened individual by $57 (cost reduced from $3400 to $3343). Secondary analyses on once-in-life colonoscopy yielded similar results. Over the simulated US population, the use of AI in CRC screening resulted in an annual savings of $290 million, an annual additional prevention of 7194 cases of CRC, and a prevention of 2089 mortalities per year.
Study limitations included uncertainties due to necessary assumptions: the assumption of a linear relationship between cancer prevention effects and higher adenoma detection rates, the assumption of the same increase in detection rate of high- and low-risk adenomas, a difference between estimated deaths and actual deaths due to CRC, and uncertainty regarding whether AI’s benefits in clinical trials will also be observed in clinical practice.
The study researchers concluded that “implementation of AI detection tools in screening colonoscopy is a cost-saving strategy to further prevent [CRC] incidence and mortality.”
Disclosure: Some study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.
Areia M, Mori Y, Correale L, et al. Cost-effectiveness of artificial intelligence for screening colonoscopy: a modelling study. Lancet. Published online April 13, 2022. doi: 10.1016/S2589-7500(22)00042-5