Utilizing AI Precision Medicine to Best Choose Prostate Cancer Therapy

A new AI precision medicine test is being used to help clinicians better identify a personalized treatment plan to treat patients with prostate cancer, as Dr. Nicholas James told CURE.

A new artificial intelligence (AI) precision medicine test is being used to help clinicians better identify a personalized treatment plan to treat patients with prostate cancer, according to Dr. Nicholas James, a professor of prostate and bladder cancer research.

To further discuss the topic of what this test is and how it is being utilized, he sat down for an interview with CURE. He expanded on the importance of knowing which men with high-risk, non-metastatic prostate cancer will truly benefit from intensified treatment in another interview with CURE.

James also serves as a group leader in Prostate and Bladder Cancer Research, at The Institute of Cancer Research and The Royal Marsden Hospital, located in London, England.

CURE: How does the ArteraAI Prostate Test use biopsy images to help doctors decide if prostate cancer patients need short- or long-term hormone therapy, and is it available yet?

James: Previously, Artera had acquired similar digital images from trials in the United States. These trials involved men with early-stage prostate cancer and compared different durations of hormone therapy combined with radiotherapy. The question Artera was trying to answer was which men with this earlier disease would benefit from hormone therapy and whether they needed a short or long course of treatment.

The software works by being fed many these digitized slides. We then ask the AI to find features in the slides that predict who will do well and who will do poorly, specifically, who will benefit from the extra hormone therapy. The AI learns which features are associated with beneficial outcomes, such as responding well to hormone therapy.

After the initial training, we run the same software over a new set of samples that the computer hasn't seen and ask it to predict the outcomes again. We can repeat these loops as many times as we like to improve the machine learning and make it better at finding adverse features. This process had previously been carried out with men who had mostly curable prostate cancer.

The test is now approved for use in both the United States and the UK as an adjunct to the early diagnostic workup to help clinicians make decisions. I haven't actually used it in this context yet because it was only approved a few weeks ago in the UK, but it has been approved for longer in the US.

Can this AI test show which men with advanced prostate cancer might benefit from adding a second hormone therapy, and how do you tell who needs it most?

We then took this fully developed test and applied it to a completely different set of patients. These patients all had advanced disease and were all receiving hormone therapy, so we predicted they should have high scores on this test. We knew they all needed hormone therapy, and sure enough, they all did have high scores. The drug we added in the trial, abiraterone, is also a form of hormone therapy. We hypothesized that this test might still work to identify patients who would benefit from a second hormone therapy. We found that all these patients had higher scores compared to the patients the test had been developed on. Of course, there are little high scores and very, very high scores— we get a range of scores.

We performed various analyses to determine if we could establish a cutoff point in these test results that separates patients who benefit from those who don't. In essence, we found that the quarter of patients with the very highest scores were overwhelmingly the ones likely to be dying of prostate cancer, while the three-quarters of patients with the lowest scores were mostly doing well with the standard treatment.

How could the ArteraAI Prostate Test help men with prostate cancer avoid unnecessary treatment, side effects, and costs while still identifying those who need extra therapy?

When we looked at the patients on the experimental arm of the trial who received the extra drug, the test still worked. We were therefore able to identify the quarter of patients who needed the extra treatment, separately from the 75% of patients who probably didn't. This is a very attractive result because it would mean that if you ran the test on a patient's slide, you could say, “You're actually probably going to do fine. You don't really need this extra treatment.” This would result in a one-off fee for a test that saves two years of drug costs.

From a healthcare system point of view, it is attractive, but from a patient's point of view, the extra drug involves extra monitoring and extra side effects. It increases your risk of hypertension and diabetes, which you clearly don't want unless the drug is reducing your risk of death from prostate cancer.

This is a very exciting result for us because it's allowing us to tailor patients' treatment, potentially more accurately than just saying, “You fit into a broad bucket of patients that had the general features, so therefore, we treat everyone,” which is what we currently do.

Transcript has been edited for clarity and conciseness

Reference

1. “ArteraAI Prostate Test: For Patients,” by Artera. https://artera.ai/for-patients

For more news on cancer updates, research and education, don’t forget to subscribe to CURE®’s newsletters here.