Skip to main content

Together we are beating cancer

Donate now
  • Science & Technology

An AI test to help guide bowel cancer treatment

by Sadaf Shafaghmotlagh , Tim Gunn | Analysis

27 January 2025

0 comments 0 comments

A microscope image showing immune cells in bowel cancer.
Immune cells (in darker purple) moving into a bowel tumour. Photo by Nephron, licensed under CC BY 3.0.

Cancer is a complex and challenging disease to treat, but one of the most powerful tools for stopping it lies within us: our immune system. 

This fact is particularly important in bowel cancer (colorectal cancer). Scientists have found that bowel tumours with larger numbers of immune cells in specific areas have a lower risk of returning after surgery. It’s as if these tumours are hemmed in by the body’s defences, so it’s harder for individual cancer cells to escape the surgeon’s knife. 

Soon, doctors could have a new way of working with those defences, making the treatment decisions that come after surgery much easier. An expert team led by one of our researchers has shown that an AI test can analyse important CD3 immune cells in stage 2 bowel tumours and identify the people who most need chemotherapy as a secondary shield to stop their cancer coming back. 

The study, which was published in the Journal of Clinical Oncology last year, was conducted by the National Pathology Imaging Co-operative (NPIC) through a grant from Innovate UK. 

“This has the potential to be the most important test patients with early-stage bowel cancer ask for,” says Dr Christopher Williams, the lead researcher and a Cancer Research UK Clinical Trials Research Fellow at the University of Leeds’ School of Medicine. 

Williams, who specialises in treating people with bowel cancer, is speaking from experience. 

“It’s one of the most difficult conversations that we have in the clinic: should you or shouldn’t you have chemotherapy after your surgery,” he says. “And that conversation is most difficult in stage 2 bowel cancer, where your chances of benefit, we know, are lesser.” 

Picking the right path for stage 2 bowel cancer

Thanks to our QUASAR trial, which ran from the 1990s into the 2000s, we know that giving people with bowel cancer chemotherapy after surgery helps more of them stay cancer-free for longer. 

Chemotherapy after surgery is also known as ‘adjuvant chemotherapy’, after the Latin for ‘to help’. The name reflects the idea that this chemotherapy is designed to help surgeons clear out cancer.

But stage 2 bowel cancers are in a difficult grey area for doctors like Williams. They’re not as far advanced as (late stage) stage 3 bowel cancers, which have already shown clear signs of spreading, meaning there’s a higher chance some cancer cells might be left behind after surgery. On the other side, they’re not as localised as stage 1 bowel cancers, which haven’t grown into the deeper layers of the bowel and very rarely need adjuvant treatments.  

Dr Christopher Williams
Dr Christopher Williams

So, when he’s talking to patients who’ve had their stage 2 bowel tumours surgically removed, Williams explains that they could be in one of three groups. Most of them will be in group one, which means they have no cancer cells left after surgery and so nothing to gain from further chemotherapy. The rest could be in group two, which means they’ll receive chemotherapy that will kill their remaining cancer cells, or group three, with some remaining cancer cells that chemotherapy won’t be able to get rid of. 

Those groups seem clearly defined, but there’s still an underlying problem. “Today, I have no way of knowing which one of those three groups the patient will be in when I’m speaking to them, or when I’m delivering the chemotherapy,” says Williams. 

Data driven decisions 

All this means some people who receive chemotherapy after surgery for stage 2 bowel cancer will be putting their lives on hold and dealing with potentially difficult side effects for no reason at all. Because of the limitations of our current tools, we just don’t know who these people might be. 

“That’s a really hard thing for an individual patient to absorb,” says Williams. “So, anything that helps us define the risk more precisely is really useful.”  

The CD3 test that Williams has been studying, which was developed with Roche Diagnostics, does exactly that. 

Using data from the QUASAR trial, Williams’ team has shown that the new AI-powered test can sort people into three groups based on their actual risk of cancer recurrence. It works by assessing and comparing the density of CD3 (and CD8) T cells in different parts of the tumour and generating a ‘CD3 Score’. The study found that those whose CD3 Score put them in the highest risk group had a three times higher risk of their cancer returning in the five years after surgery than those in the lowest risk group. 

Two images from the research paper showing microscope images of bowel cancer tissue samples. At the bottom right of each image, the AI test has highlighted immune cells.
This figure from Williams' research paper show the AI test identifying CD3 and CD8 T cells at the edge of a bowel cancer tumour.

Williams explains that this information can be used to generate a ‘number needed to treat’, showing the chances that adjuvant chemotherapy could play a role in keeping a particular patient cancer-free. “It’s quite easy to understand and talk about in clinic with patients,” he says. “People’s perception of risk is very different, and that’s where figures like this can be really useful.” 

In fact, the number needed to treat is a big part of why Williams thinks the CD3 Score could be so important. It gives doctors and patients the specific information they need to talk through their options and decide whether to use chemotherapy after surgery on a case-by-case basis.  

To put it another way, thanks to the information provided by the test, more of the people who are least likely to need chemotherapy after surgery should be able to stay cancer free while avoiding the side effects of extra treatment.  

The science behind a CD3 Score 

The CD3 test works by converting two samples from a tumour into digital images, which are then analysed for immune cells by a specially trained AI.  

Williams believes we already have most of the evidence we need to bring the test to doctors and patients across the NHS. The size and detail of the QUASAR trial dataset (based, like Williams, in Leeds) meant the team could divide it into separate segments for defining the test and then validating whether it worked, effectively rolling two projects into one. 

“The test has been rigorously evaluated and found to be reliable when testing was repeated in a second set of patients,” explains Williams. “The evidence from our trial provides a clear rationale for the test’s adoption by the NHS and our hope is that it will be made available to patients as soon as possible.” 

Unlike some of the other AIs that are beginning to capture people’s attention, the CD3 Score is also easy to monitor and understand. It isn’t a ‘black box’ that makes decisions in ways we can’t track.  

Williams points out that this makes it “auditable”: regulators and doctors can continually check on the test’s AI to make sure it’s working correctly, just like any other high or low-tech tool. “The AI will label the cells it regards as positive, positive,” he says. “It’s strongly supervised, so we’re able to audit and check what it’s doing.” 

Longer, better lives after bowel cancer

So, the CD3 Score could be a way to use AI to ease the stresses of a delicate part of cancer care, but there’s still a bit more work to do. Future studies will need to apply the team’s findings to more recent data and calculate a more up-to-date set of numbers needed to treat, as the QUASAR trial doesn’t reflect more recent improvements in bowel cancer treatment and survival.  

That’s a limitation for this study, but it helps highlight some very good news. Overall, our stats show that advances like the use of chemotherapy after surgery pushed up the proportion of people in the UK surviving bowel cancer for 10 years or more from 2 in 10 in the 1970s to almost 6 in 10 in the 2010s.  

Now, innovative tools and technologies like the CD3 test mean we can focus on quality of life alongside survival. They can also help the health system direct vital time, money and resources to the people who are most in need of treatment. 

And, according to Williams, using the test for bowel cancer is only the beginning.  

“In some cancers, the relationship with the immune system is stronger than others, and the CD3 test is a way of quantifying how strong that relationship is. So, it could have relevance in other tumour types too.”

Behind the research

Dr Christopher Williams earned his PhD in biomarkers for bowel cancer at the University of Leeds, which has a history as one of the UK’s most important centres for bowel cancer research.  

Along with the QUASAR trial, the only study in the world that could be used to train and validate an AI like the one used for the CD3 Score, Leeds is also home to our FOxTROT set of bowel cancer chemotherapy trials. 

Williams worked on the CD3 Score study after being awarded one of our Clinical Trial Fellowships, which supports doctors with an interest in cancer clinical trials. 

“The trials fellowship has been a fantastic opportunity to develop skills, particularly in my interest, which is late phase clinical trials in colorectal cancer,” says Williams. 

“It has set me up for a future career in delivering my own clinical trials as an independent Chief Investigator, which is where I want to be.” 

Sadaf and Tim

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.

Tell us what you think

Leave a Reply

Your email address will not be published. Required fields are marked *

Read our comment policy.