A stool test detects 90% of colorectal cancers with no colonoscopy needed

The second deadliest cancer in the world can now be detected with a stool sample and an algorithm. Researchers at the University of Geneva have developed a test that identifies 90% of colorectal cancer cases — without a colonoscopy, without sedation, and without a hospital visit. The method works by reading hidden signals in gut bacteria that most diagnostic tools have never been precise enough to detect.

Colorectal cancer kills over 900,000 people a year globally. When caught early, survival rates exceed 90%. The problem has always been the colonoscopy — expensive, uncomfortable, and requiring a full day of preparation that leads millions of people worldwide to simply skip screening entirely. That delay is often the difference between a treatable diagnosis and a terminal one.

The Geneva team approached the problem from a different angle entirely. Rather than looking at the colon directly, they mapped the gut microbiome at a level of detail no one had attempted before — down to the subspecies of each bacterial strain. Different subspecies of the same bacteria can behave in completely opposite ways: some promote tumor growth, others are neutral. Previous tests that looked only at the species level were missing the signal entirely.

Using machine learning trained on this subspecies-level bacterial catalogue, the team built a model that reads a standard stool sample and predicts the presence of colorectal cancer with 90% accuracy. For comparison, colonoscopy achieves 94%. Every existing non-invasive method falls short of both — the current best alternatives reach 70 to 80% detection.

The implications for public health are significant beyond the number itself. A simple, low-cost stool test that performs near colonoscopy accuracy could transform screening participation — particularly among younger adults, where colorectal cancer cases are rising for reasons that remain unexplained, and among populations where cost and access make colonoscopy effectively unavailable.

The test is not yet in clinical use. One limitation acknowledged by the researchers is that detecting precancerous lesions — advanced adenomas that have not yet become cancer — remains more difficult than detecting established tumors, a challenge shared with all existing stool-based methods. The model’s performance may also vary across genetically and geographically diverse populations, which the upcoming trial will assess.

A clinical trial in partnership with Geneva University Hospitals is now being prepared to determine which cancer stages and lesion types the method detects most effectively. If results confirm the model’s accuracy across diverse populations, the test could become a first-line screening tool — with colonoscopy reserved only for patients who test positive.