By Howard Wolinsky
Editor’s note: Artificial Intelligence is the topic of everyday discussion—at my house, anyway. I have covered AI in this blog and elsewhere for years.
AI is finding its place throughout medicine, from answering our queries in portals to forecasting our risks for various disease and potential benefits for treatments.
Who knows how far it can go?
Policy-makers are trying to prevent it from going too far.
One AI system passed the U.S. Medical Licensing Exam. It also has passed exams for lawyers and MBA students. Sometimes barely. But the grad with the lowest grades in med school is still called Doctor, right?
Could AI become your doctor, lawyer, and accountant in a single all-rolled-up-in-one package with a one-stop service?
AI is starting to pop up now regularly for some prostate cancer.
Check out Dr. Daniel Spratt, chair of radiation oncology at University Hospitals Seidman Cancer Center and a professor at Case Western Reserve University School of Medicine in Cleveland, Ohio, one of the leaders in AI biomarkers and prostate cancer, talking in Urology Times. He said: “The real immediate use is that this biomarker can be an adjunct, or it can be something to assist in making the shared decision making with patients [with prostate cancer].” He is using AI in deciding whether to use Androgen Deprivation Therapy (ADT) in some patients.
I know several guys who have had AI tests to guide their treatment.
(Portion of an AI test for a patient with intermediate-risk prostate cancer.)
San Francisco-based ArteraAI, the developer of multimodal artificial intelligence-based predictive and prognostic cancer tests, has started evaluating these men with localized intermediate-risk PCa,].
I arranged an email Q&A with Dr. Felix Feng, Scientific Advisor to ArteraAI.
Feng stresses below: “The ArteraAI Prostate Test is the first test that can both predict therapy benefit, particularly in the context of who should receive hormone therapy, and prognosticate long-term outcomes in localized prostate cancer. The long-term goal is to enable personalized care for patients.”
The take-home for those of us on Active Surveillance is that ArteraAI Prostate Test isn’t ready for us yet—but may be coming soon.
Feng noted: “This test does not currently address Active Surveillance – but stay tuned. ArteraAI is currently working with collaborators to develop an AI test designed specifically for patients who may be considering active surveillance.”
(Felix Feng, MD.)
Q&A on Prostate Cancer and AI with Dr. Felix Feng
The Active Surveillor: Please define AI for medical use.
Dr. Felix Feng: Many leaders across the industry agree that the future of healthcare is precision medicine. As the Scientific Advisor at ArteraAI, I believe that personalized cancer therapy requires precision artificial intelligence (AI). There are also other ways that AI can be used in healthcare - from streamlining workflows, to managing data, to helping with diagnosis and prognosis of disease. There is no one definition for AI in medical use because it has so many possibilities.
What is available with AI for prostate cancer?
From ArteraAI, the ArteraAI Prostate Test is available for patients with localized prostate cancer. The test leverages artificial intelligence to predict therapy benefits and prognosticate long-term outcomes in localized prostate cancer.
How does your AI work?
ArteraAI’s multimodal artificial intelligence (MMAI) biomarker test leverages a unique algorithm that assesses image data from a patient’s biopsy and learns from a patient's clinical data. The test does not consume any tissue. ArteraAI uses a patient's existing pathology slides and digitizes them for the AI program. The AI then combines this information to predict whether a patient will benefit from a particular therapy and estimate long-term outcomes.
Clinicians can order the ArteraAI Prostate Test when considering treatment options for a patient who is diagnosed with localized prostate cancer. When submitting an order, the clinician provides a patient’s biopsy slides to be digitized along with clinical data. The test results provide information that can help determine a patient’s prognosis and predict if they may benefit from a particular treatment.
Which patients are candidates for ArteraAI testing? Only localized cancer? Is this useful for deciding who can go on active surveillance and who needs to be treated? Does it recommend for or against biopsies? What role does the pathologist play?
The ArteraAI Prostate Test is the first test that can both predict therapy benefit, particularly in the context of who should receive hormone therapy, and prognosticate long-term outcomes in localized prostate cancer. The long-term goal is to enable personalized care for patients.
This test does not currently address active surveillance – but stay tuned. ArteraAI is currently working with collaborators to develop an AI test designed specifically for patients who may be considering active surveillance.
The ArteraAI Prostate Test does not consume tissue. ArteraAI uses a patient's existing pathology slides and digitizes them for the AI model. Using the results generated from the test, the clinician can engage in shared decision-making with their patients and feel confident that they are identifying a personalized treatment plan. With individualized predictive and prognostic results, clinicians can feel confident making treatment decisions that will likely lead to the most optimal outcomes.
As part of the multidisciplinary team of clinicians managing patients with prostate cancer, pathologists, as well as oncologists, are very much a part of a patient’s journey. A pathologist is vital to diagnosis of localized prostate cancer, and once the diagnosis is made, the ArteraAI Prostate Test can be used to help personalize therapy decisions.
What are the main benefits? In screening, detection, diagnosis, prognosis?
The ArteraAI Prostate Test is used during the prognosis stage of the cancer journey when clinicians and patients are faced with making a treatment decision. Using a patient’s existing tissue biopsy, the test is able to predict if a patient will benefit from a specific treatment, allowing for a personalized approach to cancer therapy.
Has the FDA approved ArteraAI? It is available commercially for about $4,000. Does Medicare cover it? Insurance?
The ArteraAI Prostate Test is not currently approved by the FDA; it is currently offered as a Laboratory Developed Test.
[Note: The Food and Drug Administration defines a laboratory developed test (LDT) as a type of in vitro diagnostic test that is designed, manufactured and used within a single laboratory. LDTs can be used to measure or detect a wide variety of analytes (substances such as proteins, chemical compounds like glucose or cholesterol, or DNA), in a sample taken from a human body.]
The test is clinically available through their CLIA-certified laboratory and can be ordered directly from ArteraAI. ArteraAI is not yet in network with any insurers, but ArteraAI will bill a patient's insurance or Medicare. Out-of-pocket costs to patients may depend not only on the coverage status of the test under a patient's health plan but also on the individual's specific plan, including deductible and out-of-network benefits.
What studies have been published? How well does your system fare vs. a pathologist? Can you cite some figures and links to research?
ArteraAI’s test has been validated using multiple Phase 3 randomized trials with up to 15-year follow-up. These data were used to develop and validate prognostic and predictive AI models. Once a patient is diagnosed with localized prostate cancer by a pathologist, the clinician and patient are faced with multiple treatment options. The ArteraAI Prostate Test supports personalized treatment decisions with individualized prognostic and predictive results.
Our study published in Nature Digital Medicine shows that, compared to the standard risk-stratification tool—risk groups developed by the National Comprehensive Cancer Network (NCCN)—our models have superior discriminatory performance across all endpoints, ranging from 9.2% to 14.6% relative improvement in a held-out validation set.
As you may know, prostate cancer disproportionately impacts African-American men, with this population making up 16.3% of prostate cancer patients but only an average of 9.4% of clinical trial participants. Recent data demonstrate that ArteraAI’s Prostate Test has similar performance in both African American and non-African American prostate cancer patients.
There is also the recent data published in the New England Journal of Medicine Evidence, which validated the first-ever predictive AI biomarker of androgen deprivation therapy (ADT) benefit in prostate cancer. This study used novel deep learning methodology and histopathology image data from more than 5,000 patients across five Phase 3 randomized trials, and will advance the ability to create a personalized approach for the treatment of localized prostate cancer.
Does ArteraAI only focus on digital images of prostate biopsy slides? Will you eventually factor in biomarkers, PSAs, PSA Density, MRI data, biopsies, genomic tests, etc.?
The ArteraAI Prostate Test uses image data from the patient's previous biopsy and their clinical data to help predict a patient’s prognosis and their likelihood to benefit from a particular treatment. These clinical variables include age, PSA, Gleason score, and T-stage.
What are the downsides to AI?
While not necessarily downsides, it is important to note that AI is still new, and it is something that is rapidly growing and improving. In order for AI to be integrated into healthcare, the technology needs to be clinically validated. AI tools need to be backed by research and proven to be useful for both the patient and clinician. AI models should be developed using robust datasets and be rigorously validated through research.
The data used to validate these tools should also be representative of the patient population affected by this disease. Historically, clinical trials have underrepresented minority populations. In the U.S., prostate cancer disproportionately impacts African American men, and yet this group is historically underrepresented in clinical trials. As mentioned, the ArteraAI Prostate Test has been shown to have similar performance in both African American and non-African American prostate cancer patients. It was critical that ArteraAI ensured this population was included in model development.
Overall, clinicians need to feel like they understand and can trust AI. This trust will need to be earned over time.
Are there tests that ArteraAI can replace?
The ArteraAI Prostate Test is not meant to replace any tests. The test is meant to add to the tools that clinicians have to provide personalized cancer therapy to patients. The test supports shared decision-making for clinicians and patients, as it provides results that are specific to the individual patient. With predictive and prognostic data, clinicians may feel informed and confident with their treatment decisions.
Who are your competitors?
The ArteraAI Prostate Test is the first of its kind to provide both predictive and prognostic results for patients with localized prostate cancer.
Dr. Felix Feng, MD, a radiation oncologist, is a Scientific Advisor at ArteraAI and a Professor of Radiation Oncology, Urology, and Medicine at UCSF. The physician-scientist is also the Associate Director for Translational Clinical Medicine at the UCSF Helen Diller Comprehensive Cancer Center. He has spent his career as both a physician and a scientist trying to personalize cancer therapy through the use of biomarkers, which is what led me to ArteraAI. He is a leader in a number of national and international phase III trials for prostate cancer patients.
Thanks for sharing, Richard.
I used AI to summarize something I wrote.
It had a strange reference. AI made it up. The author never wrote for the journal.
My wife told me of a case where AI gave a reference for a fake book. Then, it wrote the book to justify the reference.
It's useful. But be careful when using AI.
Howard
I just used AI to help me draft an appeal with my insurance company for NGS for metastatic hormone sensitive prostate cancer. It was short on citation to authority and contained a lot of generic text, nevertheless it provided a good start for the appeal. The bottom line is AI is a garbage in garbage out process. And the user has to know what went in to know if what came out. Is useful.