New research led by a Norwegian team has found that using artificial intelligence (AI) during colonoscopy screenings can slightly improve the detection of small growths called polyps. These are important to find because they can turn into cancer over time.
However, the study also found that while AI helps spot more polyps, it has only a very small effect on actually reducing the risk of colorectal cancer or preventing deaths.
Colorectal cancer is one of the most common cancers in Western countries. In Norway alone, around 5,000 people are diagnosed with this type of cancer every year. To help catch it early, Norway has launched a national screening program.
Right now, the country uses a test called FIT (Fecal Immunochemical Test), which checks for hidden blood in stool. If blood is found, the person is referred for a colonoscopy. However, Norway plans to switch from FIT to full colonoscopy screenings for all 55-year-olds within the next five years.
During a colonoscopy, a doctor uses a flexible tube with a camera to look inside the large intestine. The aim is to find and remove polyps before they can develop into cancer. In recent years, AI tools have been added to assist doctors during this process by helping to spot polyps that might otherwise be missed.
AI is one of the most researched technologies in medicine, especially in areas like colonoscopy. But for any new tool to be widely accepted, it must show clear health benefits. To guide doctors and health authorities, the international research group MAGIC, led by Norwegian researchers, helped create global guidelines for the use of AI in colonoscopy.
As part of this effort, they carried out a large study that reviewed the effects of AI on cancer risk, deaths, and the impact on patients and health care systems.
This study included data from 15 studies and built on a recent review of 44 clinical trials involving over 30,000 patients. It was published in BMJ Medicine and served as the foundation for the new guideline released in BMJ Rapid Recommendations on March 27.
The results showed that AI use in colonoscopy does increase the number of polyps detected by about 8%. However, when looking at the long-term risk of developing or dying from colorectal cancer, the difference was minimal.
Over a 10-year period, the cancer risk dropped from 0.82% without AI to 0.71% with AI. For people undergoing colonoscopy after a positive FIT test, the risk fell slightly from 5.82% to 5.77%.
On the other hand, detecting more polyps means more follow-up colonoscopies. The study found a 20% increase in these extra exams over 10 years. This means more medical appointments, more stress for patients, and more work for already busy health care systems.
Because of these findings, the BMJ guideline currently advises against routinely using AI in all colonoscopy screenings. Still, the authors note that the decision isn’t final — future updates could change the recommendation, especially if more research finds stronger benefits or if health economics show it’s worth the cost.
Interestingly, other major groups came to different conclusions. The American Gastroenterology Association (AGA) said there wasn’t enough evidence to make a clear recommendation either way.
Meanwhile, the European Society of Gastrointestinal Endoscopy (ESGE) supports using AI, assuming most patients would prefer the added help during screening.
These different opinions show just how complicated this issue is. It’s not just about whether AI works, but also about how it affects patients, health care workers, and the entire system. Some experts worry about the downsides, such as overdiagnosis (finding things that might never cause harm) and adding unnecessary pressure to hospitals and clinics.
Dr. Natalie Halvorsen, a physician and PhD candidate involved in the research, pointed out that while AI can help detect more polyps, the actual benefit for patients is small.
“The risk of cancer, which is most important to patients, is only minimally reduced,” she said. Professor Per Olav Vandvik, who helped lead the project, added that even though AI holds promise, tools should only be adopted if they clearly improve patient health.
The research is part of a broader EU project called OPERA, which is studying whether AI can truly lower cancer rates and deaths in the long term. The project is led by Professor Yuichi Mori at the University of Oslo.
In conclusion, the study makes it clear that while AI helps doctors spot more polyps, this does not automatically lead to a major reduction in cancer.
The modest benefit must be weighed against the extra strain it puts on patients and health care systems. At this stage, the recommendation is to proceed with caution and focus only on AI tools that show clear and proven benefits for patients.
If you care about cancer, please read studies about vitamin D and bowel cancer, and a cautionary note about peanuts and cancer.
For more health information, please see recent studies about the link between ultra-processed foods and cancer, and everyday foods to ward off pancreatic cancer.
The research findings can be found in BMJ Medicine.
Copyright © 2025 Knowridge Science Report. All rights reserved.
