Close to 50,000 people are diagnosed with bowel cancer in the UK each year. A drug to treat its advanced form has recently become available on the NHS, but researchers say the majority of patients who receive it are unlikely to benefit.

Bevacizumab was approved for NHS use in December 2025. It works by slowing the growth of advanced bowel cancer, but clinical evidence indicates it is effective only in a small proportion of patients. For those it does not help, the risks remain real: the drug is associated with high blood pressure, gastrointestinal problems and blood clots. Prescribing it to patients who will not respond is, in practical terms, exposing them to harm without benefit.

Scientists at the Institute of Cancer Research in London and RCSI University of Medicine and Health Sciences in Dublin have developed an AI tool they believe can identify those patients before treatment begins. The tool, called PhenMap, short for phenotype mapping, takes a different approach to how cancers are classified. Conventionally, cancers are grouped into a limited number of broad subtypes. PhenMap works by detecting more complex patterns within tumour data, drawing on the genetic makeup of the tumour alongside clinical information including the patient's age, sex and the position of the tumour.

From those patterns, a second AI tool generates a risk score for each patient on a scale of one to 100. Patients are then assigned to one of three categories: high, moderate or low risk. The top ten per cent of scores fall into the high risk group, the bottom ten per cent into low risk, with the remainder classified as moderate. When researchers examined the clinical outcomes, they found that not one patient in the high risk group had responded to treatment with Bevacizumab and chemotherapy. The tool also identified that every patient carrying a mutation in the BRAF gene fell within the high risk category and had poor outcomes.

The identification of a set of shared characteristics among high-risk patients could serve as a biomarker. This biomarker would offer clinicians a clear basis for deciding whether to use the targeted drug Bevacizumab. According to Anguraj Sadananda, a professor in stratification and precision medicine at the ICR, "Once bowel cancer spreads to other parts of the body, there are very few treatment options available for patients." He added that it is therefore "positive that patients can now access the targeted drug Bevacizumab on the NHS." However, we know that the majority of patients won't benefit from the drug, meaning thousands of people in England could be facing unpleasant side effects unnecessarily. Until now, we haven't been able to identify these patients." He added that the AI methods used allowed researchers to detect patterns within tumour data that would otherwise be beyond human recognition.

The researchers are cautious about how far the findings can currently be applied. The results need to be validated in a larger patient cohort before the tool can be considered ready for broader clinical use. They have also noted that the underlying approach used in PhenMap may have applications beyond bowel cancer, with the potential to inform treatment decisions in other cancer types in future.

The findings were published in the journal Scientific Reports in April.

Bowel cancer is among the more treatable cancers when caught at an early stage, yet it remains a significant burden on health services. Last month, it was identified as one of 11 cancer types increasing in incidence among younger adults in England. As the patient population shifts and treatment options expand, tools that can distinguish who will and will not benefit from a given drug are becoming harder to argue against. PhenMap is not yet in clinical use, but the research suggests that more precise patient selection is achievable, and that the cost of getting it wrong is one the NHS can ill afford to keep absorbing.