The Artera AI Model: Advancing Personalized Medicine in Prostate Cancer Treatment | Dan Spratt & Alicia Morgans Discuss NEJM Evidence Manuscript
June 29, 2023
Daniel Spratt, MD, Chair, Department of Radiation Oncology, UH Cleveland Medical Center Professor, CWRU School of Medicine, Cleveland, OH
Alicia Morgans, MD, MPH, Genitourinary Medical Oncologist, Medical Director of Survivorship Program at Dana-Farber Cancer Institute, Boston, Massachusetts
Alicia Morgans: Hi, I'm so excited to be here with Daniel Spratt, a professor and radiation oncologist who's here to talk with me today about the Artera predictive model for prostate cancer. So excited to talk with you today.
Daniel Spratt: Thank you. It's great to be here.
Alicia Morgans: Great. I'm really excited to hear a little bit more about what you and the team published in the manuscript.
Daniel Spratt: This is really exciting. And so, we've worked on this multimodal artificial intelligence model and we've talked about that there's a prognostic model that we've published and validated, but there's also a predictive model. And I think a lot of people may not know the difference in these two terms. And it's very important because a prognostic model, which is very clinically valuable, helps us identify patients at lower or higher risk of developing, let's say, recurrence or metastatic disease. A predictive biomarker, some would say is the pinnacle of personalized medicine, 'cause it tells a patient, regardless of their prognosis, will they benefit from a specific therapy or not. And so, in localized prostate cancer, we don't use any predictive biomarkers today. If you look at national guidelines, they're all prognostic biomarkers. And the Artera Prostate Test is one of those prognostic markers that has really helped us improve prognostication or risk stratification.
But for the first time in history, we've developed and validated in randomized trials a predictive biomarker to guide the use of androgen deprivation therapy. This was conducted in five facially randomized trials over 5,500 patients where four of the trials were used to train this predictive biomarker, patients that had radiation, some that had radiation with hormonal therapy, to see can we see what features both clinically, pathologically, and from this imagery data that the deep learning is assessing and extracting features from which patients had limited or greater benefit from hormone therapy. But very importantly, I can't overstate this, it was validated in the largest phase three randomized trial ever conducted in history with the longest follow-up, almost 20 years of follow up a trial that, it was called RTOG 9408, of patients that got radiotherapy alone or radiotherapy plus short-term hormonal therapy or androgen deprivation.
And once this model was locked and then validated, what we showed was that we can identify in men, especially with intermediate risk prostate cancer, about two thirds of men today that our guidelines would tell us to use hormone therapy are predicted to not benefit from hormone therapy. And that's exactly what we validated and showed. There was no benefit. The hazard ratio is almost exactly one for the development of distant metastasis. What that means is there's really no benefit to getting a hormone therapy. All the patients would experience is the potential costs and side effects. But for the one third of patients that were, we call biomarker positive, this model predicted they would benefit from hormone therapy, had an incredible benefit in terms of the reduction of distant metastasis. Its hazard ratio is less than 0.5, that means it's a greater than 50% relative reduction in them developing distant metastasis.
And so, really I think this will help us personalize the patients 'cause no one wants hormone therapy, but when you can tell a patient that they're going to drive a large benefit based on their tumor, it's something that I think is going to help us with that risk benefit personalized shared decision making. And very exciting, I'm sure you would agree, is that to be able to tell a patient, even if their prognosis is not very favorable, that just the biology of their tumor is not going to benefit from hormone therapy. I mean, we've been chasing after this for decades. And other cancers, breast cancer, they've had these biomarkers for many years and this is a real advancement for prostate cancer.
Alicia Morgans: Let's dig into that a little bit more because when you're sitting with a patient who has intermediate risk disease and you want to give them radiation, we often use the NCCN risk criteria, favorable and unfavorable, and we try to fumble our way around and say, "Well, unfavorable, you should get six months of ADT." Favorable, there's actually a lot of controversy and a lot of data that you and the team have provided to limit exposure for patients who can limit it. But I think that there's not consensus that everyone is really eliminating ADT in that setting. And so, we end up with many patients who might be exposed to six months of ADT. It sounds like this is something that is really able to, perhaps on a larger scale, give more confidence around that because it's really using this AI to help predict, as you said, a response, a benefit from ADT or not, which I as a clinician feel much more confident in.
Daniel Spratt: Yeah, absolutely. I mean, I was a resident and with my co-resident we developed the favorable versus unfavorable intermediate risk grouping that got into NCCN guidelines. And although we developed it to try to tell us which patients to receive hormone therapy or not with radiation, it's really a prognostic biomarker in the sense it's prognostic to say which men had more favorable or less favorable disease. But we know that even in unfavorable intermediate risk, many men don't benefit from hormone therapy. And there's probably men with favorable intermediate risk that do benefit. But we've tried through many iterations of different models using different clinical variables, and no one's been able to, in a randomized trial validation to ever show they can have a predictive biomarker of hormone therapy. And so, what that really enables, and I think that this validation can't be overstated, is that this isn't just something with retrospective institutional data. This is phase three randomized trials that with long-term follow up to validate that this model can actually predict which men will more or less likely benefit from hormone therapy.
And it compliments the prognostic model that Artera Prostate Test has as well, because you can use both of these in conjunction to tell you what is the risk that this patient will recur or develop mets or die from prostate cancer. But also if you were to add hormone therapy, are they likely to benefit from it or not? What's really exciting is that you can combine the information with the Artera Prostate Test for intermediate risk patients. We'll have both the prognostic model that is superior to our standard NCCN risk classification system, but you'll also have this predictive biomarker model in it as well. And so, when you have a patient in front of you, you can counsel them on their absolute risk of developing, let's say, a distant metastasis or dying from prostate cancer. But you'll also be able to tell them, just because you may have a moderate or high rate of having one of these adverse outcomes, they may not benefit as much as we think from adding hormone therapy.
What ultimately that will enable is not only personalized decision today, but just think about the very important clinical trials that we can now do to understand how do we best treat those patients.
Alicia Morgans: Absolutely. And when we're talking from a patient perspective, I think it's important for us to remember and to think about what exactly are we gaining by avoiding treatment. From your perspective, are there effects that are really important from only six months of hormonal therapy? It sounds so trivial. Does it really matter to patients in terms of their quality of life and other health effects?
Daniel Spratt: Yeah, I mean, we publish, it's called the MARCAP Consortium, and it's a meta-analysis of randomized trials that looked at the addition of hormone therapy to radiotherapy and in, we'll call them relatively unselected men, mostly intermediate and high-risk men, but there was no AI models or special biomarkers used, there's a survival benefit, there's a reduction in distant metastasis. I mean, these are meaningful things that can reduce these events by adding hormone therapy even for just four to six months.
However, hormone therapy, even short term hormone therapy can have hot flashes, sort of we call metabolic type syndrome. Guys can gain body fat, lose muscle mass, increased blood pressure, blood sugars out of control, and there's mental effects, libido effects, and it can take, some studies show that six months of hormone therapy can take up to over a year in addition to recover. So, although it's short term, it can have longer term effects. And so, yes, do we want to help men live longer? Absolutely. But this biomarker allows us to be able to tell men, you are going to probably benefit at a high probability and maybe some of those side effects, that risk benefit ratio is worth it versus some men, when you have a very limited or no probability of benefiting from hormone therapy, why on earth would we be recommending that treatment?
Alicia Morgans: And certainly why on earth for two thirds of the patients-
Daniel Spratt: Absolutely.
Alicia Morgans: ... who currently may be prescribed ADT for their intermediate risk disease. So, as you think about this advance, what would your bottom line message be?
Daniel Spratt: I think that this is really transformative. And really people need to understand that we have never had any predictive biomarkers in clinic. Every day when I see patients, we are hoping, we've been talking about this for decades, we want these tools. And this is probably the first of many of what this technology will enable us to do. And I hope that there's really strong embrace for using these tests and the further advancement of these tests to help men with prostate cancer.
Alicia Morgans: And at the end of the day, that's what it's all about. To get patients the treatments that will benefit them and to avoid overtreatment in patients who really can't see benefit in terms of disease control and certainly in terms of quality of life from something that will be ineffective. So, thank you so much for your time and your expertise.
Daniel Spratt: Thank you so much. Appreciate it.