Artificial intelligence developed by researchers at UCL has revealed a promising combination of cancer treatments that could help patients with locally advanced rectal cancer live longer.
A new study, published in eBioMedicine, showed that combining the cancer treatment irinotecan with standard chemoradiotherapy improves survival against advanced rectal cancer - which has begun to spread to other nearby tissues - in those with a high concentration of cancerous cells in their tumours.
The effectiveness of this combination was revealed using a specially-developed AI trained by the research team to distinguish patients by their cancer cell density on tumour samples taken at the time of diagnosis.
For patients who had a high tumour cell density in their biopsy samples prior to starting treatment, adding irinotecan reduced the risk of cancer recurrence by about 43% and reduced the risk of death by about 50%* compared with those who received a typical treatment - chemotherapy using the drug capecitabine combined with radiation therapy. Those with low concentrations of cancer cells showed no difference.
Irinotecan is a type of chemotherapy that is often used to treat advanced bowel cancer. Though long thought to be a promising treatment for advanced rectal cancer when combined with standard chemoradiotherapy, previous research hadn't seen an improvement in patient outcomes.
AI was instrumental in revealing the combination's effectiveness by accurately and efficiently separating the patients into high and low concentration groups than had been possible before. The team developed the free online tool Octopath, where clinicians can upload biopsy slides to be analysed.
Previously, manually identifying high concentrations of cancer cells in patients has been a time-consuming and impractical process for large-scale studies like this which featured 414 samples. By developing their AI, it allowed the researchers to scan standard images of biopsies, count millions of cells and sort patients into categories of high and low concentrations in a fraction of the time, ultimately revealing which patients benefited from the treatment.
The research is part of the analysis of the ARISTOTLE trial, a large clinical study that enrolled patients from 75 UK hospitals, which investigated whether adding irinotecan to standard chemoradiotherapy yields better results.
Lead author Dr Zhuoyan Shen (UCL Medical Physics & Biomedical Engineering) said: "While the original trial showed little benefit from adding irinotecan, by using artificial intelligence we found that we could distinguish patients who actually benefitted from those who did not. This demonstrates how AI can reveal tumour biology that is difficult to measure consistently by conventional means, and hopefully can be used to reveal additional insights that could lead to future treatments."
Predicting treatment outcomes
Colorectal cancer, which affects the final few inches of the large intestine, is the fourth most fatal cancer in the UK, and its advanced stages have a high rate of recurrence, necessitating more intensive treatments.
Combining irinotecan with capecitabine has shown promise to boost their effectiveness against cancer, but it is also likely to intensify the serious and sometimes debilitating side effects of the drugs, such as diarrhoea or low white blood cell counts. Because of the additional side effects of these combined medications, doctors will not prescribe the more intensive treatment unless there's a high-level of evidence that the patient would benefit from it.
By correctly identifying the concentrations of cancer cells, the AI can tell which patients would likely benefit from the treatment before starting it, sparing those who wouldn't benefit from unnecessary side effects. Automating the process using AI allows the technique to be scaled up so more patients suffering from rectal cancer could receive the treatment.
Senior author Professor Maria Hawkins (UCL Medical Physics & Biomedical Engineering and clinician working at UCLH) said: "Intensifying already taxing treatments puts additional strain on patients suffering from cancer. Clinicians need reliable ways to identify who is most likely to benefit before such treatments begin so potential side effects are avoided. Our findings show that doctors assisted by AI can pinpoint which patients will likely benefit from the more intensive treatment before it begins."
The long-term results of the full ARISTOTLE trial are being published this week in Lancet Oncology, finding that the addition of irinotecan did not improve outcomes for all patients generally. This larger scale clinical trial is led by Professor David Sebag-Montefiore from the University of Leeds and co-ordinated by the UCL Clinical Trial Centre.
Co-author Professor David Sebag-Montefiore of the University of Leeds said: "This exciting AI study demonstrates the crucial importance of understanding the difference in biological behaviour of rectal cancer, and we are so very grateful to the patients who participated in the ARISTOTLE trial and agreed for their biopsy samples to be used for research."
AI training
The researchers trained their AI on large-scale open-source datasets and applied it to detailed microscopic images of 414 rectal cancer biopsy slides from patients enrolled in the phase III ARISTOTLE trial. The analysis classified 188 of them as having high concentrations of cancerous tumour cells, while 226 had low concentrations.
The AI was able to recognise the tumour from the surrounding tissue, and identify which cells were cancerous, and which were healthy.
Though promising, the researchers added that additional independent verification and clinical studies will be needed before the AI can be used for clinical decision making, or the treatment combination is available for the patients widely.
This study was funded by Cancer Research UK and the UKRI Future Leaders Fellowship.
* This refers to overall survival over the five-year follow up period.
