Artificial intelligence can now identify which patients with advanced bowel cancer are likely to benefit from a targeted drug recently made available on the NHS, a development that could spare many from enduring serious side effects for no clinical gain.
The drug, bevacizumab, was approved for use on the NHS in England and Wales in December for previously untreated metastatic colorectal cancer. While it can slow tumour growth, it is effective for only a subset of patients and carries risks including high blood pressure, gastrointestinal problems, and blood clots. With nearly 10,000 cases of advanced bowel cancer diagnosed in England each year and limited treatment options, the need to predict who will respond is acute.
Pinpointing the patients who will benefit
Scientists from The Institute of Cancer Research (ICR) in London and the RCSI University of Medicine and Health Sciences in Dublin have developed an AI method to solve this problem. Their research, published in Scientific Reports, studied 117 European patients treated with bevacizumab and chemotherapy.
At the core of their work is an AI tool developed at the ICR called PhenMap, short for phenotype mapping. Professor Anguraj Sadanandam, who leads the Systems and Precision Cancer Medicine team at the ICR, explained its power. Unlike older methods that group cancers into a small number of broad subtypes, PhenMap integrates vast and complex datasets to uncover subtle, previously invisible biological signals.
It weaves together detailed information on the genetic makeup of a patient’s tumour with key clinical data such as their age, gender, and the side of the body the tumour is located on. By analysing these interconnected factors, the AI can identify intricate patterns relevant to how a patient will respond to bevacizumab, effectively placing patients on a refined scale rather than in crude categories.
Building on PhenMap’s analysis, a subsequent AI tool generates a specific risk score, categorising patients into ‘high’, ‘moderate’, or ‘low’ risk groups for dying after the combination treatment. Critically, in the study, none of the patients placed in the ‘high risk’ group responded to bevacizumab at all. The AI also identified that patients with a mutation in the BRAF gene consistently fell into this high-risk group, facing a significantly higher risk of death.
The drug’s limits and the rising need
The promise of this predictive tool is underscored by the specific limitations of bevacizumab. The drug, a VEGF inhibitor, works by depriving tumours of proteins needed to grow. Its wider adoption by the NHS was facilitated by the availability of lower-cost biosimilar versions. According to trial data cited by researchers, adding bevacizumab to frontline chemotherapy extends the time before cancer progression from 8 to 9.4 months and improves overall survival from 19.9 to 21.3 months, a modest but valuable gain for those who respond.
“Once bowel cancer spreads to other parts of the body, there are very few treatment options available for patients,” said Professor Sadanandam. “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.”
The need for smarter treatment is amplified by a concerning trend: a rise in bowel cancer cases among young adults. Between the early 1990s and 2018, diagnoses in adults aged 25 to 49 in the UK increased by 22%, with factors like diet, obesity, and lifestyle under investigation.
A model for the future of cancer treatment
The researchers stress that their findings must be validated in larger patient cohorts to ensure they are applicable to all. However, the approach signals a shift towards highly personalised care. The pattern of features identified by the AI could serve as a biomarker, helping clinicians direct bevacizumab only to those likely to benefit.
Professor Kristian Helin, Chief Executive of the ICR, highlighted the broader potential. “AI has revolutionised cancer research – by enabling us to rapidly analyse large, complex datasets and predict how patients will respond to treatment,” he said. He added that the method has the potential to be explored for many cancer types and for predicting responses to other targeted therapies.
The work, funded by EU Horizon 2020, Research Ireland, the Ian Harty Charitable Trust, and the ICR, exemplifies how AI is being leveraged not just for discovery but for making existing treatments smarter and kinder. The ICR has a noted track record in drug discovery and, through tools like PhenMap, is now applying AI to refine exactly how those drugs are used for each individual patient.