AI Can Help Cut Colorectal Cancer Death Rates, Reduce Costs

A study published in The Lancet Digital Health indicates that the use of artificial intelligence (AI) during screening colonoscopies has the potential to prevent colorectal cancer (CRC) incidence, reduce mortality, and increase cost-effectiveness across the care continuum.

The Centers for Disease Control and Prevention (CDC) reports that CRC is the third most common cancer, not counting some types of skin cancer, and the third leading cause of cancer-related death in the US. Colonoscopies to screen for colorectal polyps are the standard method for CRC prevention and early detection.

AI has been shown to help reduce the miss rate of polyp detection in research settings, which may decrease CRC incidence rates if properly integrated into the clinical setting. The study aimed to investigate the effect of the implementation of AI detection tools during screening colonoscopies on CRC incidence and mortality, while also evaluating the cost-effectiveness of those tools.

The researchers conducted a microsimulation of using screening colonoscopies with and without AI for a hypothetical group of 100,000 people between the ages of 50 and 100 who are at average risk for colon cancer. The microsimulation was then subject to multiple analyses with various parameters.

The primary analysis compared screening colonoscopies with and without AI every ten years from ages 50 to 80, with follow-up until age 100. The analysis assumed that 60 percent of this hypothetical population received a screening. The secondary analysis modelled once-in-life screening colonoscopy at age 65 in adults aged 50 to 79 who are at average risk for colorectal cancer. These analyses were compared with data resulting from no screening.

When compared with no screening, the primary analysis found a relative reduction in CRC incidence of 44.2 percent for screening colonoscopies without AI versus 48.9 percent for colonoscopies with AI. CRC mortality followed a similar trend, with a 48.7 percent relative reduction for screenings without AI versus 52.3 percent relative reduction for those with AI when compared with no screening. The secondary analyses yielded similar results.

To measure cost effectiveness, data on the direct and indirect costs of the AI tools and treatments for screening-detected disease were obtained from recently published literature. The use of AI detection tools during screening colonoscopies decreased costs from $3,400 to $3,343 per screened individual, representing savings of $57 per person.

When the researchers projected these outcomes onto a simulation of the US population, based on 2008 US census data and assuming a 60 percent screening uptake, they found that AI implementation during screening colonoscopies had the potential to prevent 7,194 CRC cases and 2,089 related deaths. These reductions translated to savings of $290 million per year.

The study did not consider individuals at high risk for CRC, but research suggests that machine learning (ML) can help flag high-risk patients and allow care teams to schedule screening colonoscopies. Of the individuals flagged through the ML approach, 68 percent were scheduled for a screening, and 70 percent of those screened had a significant finding. These data could indicate that AI implementation has the potential for improving high-risk patient outcomes as well, but more research is needed.