Intro: Welcome to Precision Pathology Podcast, featuring interviews with authors and global thought leaders who are at the forefront of technological advances in precision diagnostics and therapeutics. Your host is Dr. George Netto, the Simon Flexner Professor and Chair of Pathology and the Laboratory Medicine at the University of Pennsylvania's Perelman School of Medicine in Philadelphia. Here is Dr. Netto.

George Netto, M.D. (Host): Welcome to a new episode of P Cubed. Actually, this is the second episode and I'm honored to have our guest, Dr. Roberto Salgado. Roberto is an international expert on immuno-oncology biomarkers and developing guidelines, and that's why I wanted him to be here. He practiced in ZAS Hospital in Antwerp, Belgium. But he also worked internationally. So, he's the Honorary Research Associate at the Peter McCallum Cancer Center in Melbourne, Australia, and Francis Crick Institute in the Research Division in UK. I don't know how you do it with all these disparate locations. But clearly, you've done an amazing job.

Most importantly, for the sake of our podcast and our audience today, he's a co-chair together with Sherene Loi of the International Immuno-Oncology Biomarker Working Group. And what I love about that role is this is a working group of more than 900 pathologists over 60 countries-- amazing, Roberto. Congrats-- that they're all putting their heads together on trying to see now that immuno-oncology is very important, well, how do we develop guidelines for pathologists to do the best job on that?

I'll be remiss if I don't mention that Roberto and I meet regularly at the WHO. I had the honor to meet with him as a member of the editorial board on breast cancer. And today, the conversation will center about an editorial that him and Sherene wrote, discussing a new study that just came in cancer cell on the use of antibody drug conjugates in breast cancer. So, thank you, Roberto, for joining me. And I'm really honored you're here.

Roberto Salgado, MD: Oh, thank you very much, George. The honor is completely mine. I really love the opportunity to engage with my fellow pathologists through this medium. Thank you very much for the invitation.

Host: It's obvious. Now, you're going to be almost like being in a Hollywood movie. You're going to be all over the globe and people will see through our vodcast, they're going to see your beautiful beard and hair. I'm jealous from the hair as I don't have much remaining, I'm working on it, so...

Roberto Salgado, MD: It's lessening also from my side.

Host: It's part of maturity. So, let's go back to the topic at hand. If you start, since the study that's going to launch the conversation is on ADC, a little bit about and breast cancer, from your experience, just give us a little bit a mini, mini-symposium.

Roberto Salgado, MD: Well, indeed, it's actually a very fascinating topic for a variety of reasons. And I will try to detail them very briefly. The first one is the drug, antibody drug conjugates. How do they work? It's very simple. They use HER2 on the membrane as an anchor, as a ploy. So, they bind and then they get like a torpedo into the cancer cell. The cancer cell explodes and all the chemo gets out. That's how it works. So, the difficulty is HER2 such a biomarker? Yes or no? And all the entities that are treated with these drugs who respond very well are the separate entities, or is it just a mechanistic thing about HER2 being on the cancer cell membranes serving as a ploy and nothing else? So, that's one element. Is it a biomarker, HER2 expression, yes or no? And that's important when we discuss the ASCO-CAP guidelines on the HER2-low and the ultra-low.

If we don't consider that this is a biomarker, then the reasoning becomes a bit different. Why? Because then only quantity matters. When we tend to think about the biomarker, it means that the cancer cell is dependent on the biomarker for whatever, for its growth, its proliferation, its invasion. But now, we're talking about the biomarker is there or HER2 is there just to serve as a ploy. So now, the question comes based on a variety of prospective phase III trials called the DESTINY-04 and the DESTINY-06, which the DESTINY-04 has introduced the concept HER2-low. And the 06 DESTINY trial has introduced a concept of HER2 ultra-low.

Now, what does it mean? What does it mean? It's very simple. The lower the HER2 expression on the membranes, the more you get to the HER2 ultra-low. What does HER2 ultra-low actually mean? We know in the classical ASCO-CAP guidelines that you have HER2-0, which is actually two entities: HER2 without any expression, and HER2 with a little bit of expression, up to 10% of tumor cell. This means that one positive tumor cell, one positive tumor is called HER2 ultra-low. As pathologists, we know the more we look, probably we'll find that single positive tumor cell. And again, thinking is HER2 then a biomarker? Should we really assess HER2 as a biomarker, yes or no? We can discuss this later on during the podcast.

Now, the difficulty for daily practice is, first of all, these drugs are extremely expensive, but they work quite well, but they cause life-threatening toxicities. Certainly in the lungs, interstitial lung disease, and a variety of other complications. So, it's crucial that we find patients who may probably not respond very well to this. So, we shouldn't be giving them this treatment, because the toxicities can be extremely severe and sometimes patients die due to toxicities and not due to cancer.

Host: So, you really want to give it to people who would respond. That's why it's crucial to have that predictive. And I love the view of is this a biomarker that is helping the tumor or is just a docking station now? It sounds like we're using it now as a docking station because if it's in one cell, clearly that's not that critical anymore for the tumor. It's critical for us to target the tumor. I love that thinking.

Roberto Salgado, MD: Absolutely. Yeah. And it goes even further, meaning we all know as pathologists that practically every breast cancer has HER2.

Host: Wow. It's just how low do you want go?

Roberto Salgado, MD: So, how low do you want to go? So, that's the major concept. Now, the question is there has been a tremendous variety of research ongoing, and none of them has found one suitable predictive factor. And here comes the trick and here comes the trial. It's a phase II. We can discuss about the phase II trials. Is that sufficient to change practice? But phase II trials can be very informative to inform us on the biology.

So, the FASCINATE trial is a phase II trial in HER2-positive disease. Now, let's forget a little bit for the moment the HER2-low and ultra-low story. Let's consider antibody drug conjugate in HER2-positive disease. That's a concept. And then, we will combine both of them together. A Chinese group from Fudan Cancer University, it's a tremendously powerful and very influential group, they have developed a phase II trial with a single agent, so without chemo and an antibody drug conjugate, just the antibody drug conjugate. They gave it to 60 patients with HER2-positive disease. And what they saw, that's the manuscript that they published in Cancer Cell a few months ago.

Host: I just want to interject. The lead author is Ding Ma, and this was published probably last month or the month before in Cancer Cell with the editorial that is authored by you and Sherene.

Roberto Salgado, MD: So, what did they find? And that's very interesting. They found and told the results immediately and then, how did they come to these results? Basically, they saw, first of all, that in the HER2-positive hormone receptor-negative population, the immune system is actually a very powerful predictor for response. That's one. In the hormone receptor-positive HER2-positive population, the immune system wasn't actually that important. What was important there is let's call it the architecture of the cancer, how the cancer looks like. And we know when we look through an H&E, we see a variety of patterns. But we never considered formally that those patterns that we see may tell us something about will the drug respond or will the drug act on the cancer? Yes or no? So, this group found that in hormone receptor-positive disease, the closer the cancer cells are packed, the less the drug works.

Host: The ones that are expressing HER2, right? Cells expressing HER2.

Roberto Salgado, MD: All cells are expressing HER2, HER2-positive hormone receptor-positive. The closer the cancer cells are packed, the more solid the growth pattern is, the less the drug works

Host: Wow.

Roberto Salgado, MD: And the less packed, the less closer, so the more disaggregated, the more the better the drug works

Host: So, pattern does matter?

Roberto Salgado, MD: Absolutely.

Host: And you can do that with AI.

Roberto Salgado, MD: Exactly. And that's the point, because I can score the TILs with a microscope. You can do it with AI. But architecture is something that we see with a microscope, but don't ask me to quantify it. Don't ask me to put a number on it. And don't ask certainly me as a pathologist to give it a number and integrate it in a predictive manner because what have these guys done? These guys have analog. They have combined the clinical variables, the pathological variables. And then, they did some special profiling, which was validated using DNA and RNA sequencing data and combined the spatial profiling with the immunohistochemistry and the spatial profiling on an H&E, they put it all together in what they call a predictive model. And that model was actually very predictive in predicting which are the patients that will respond to antibody drug conjugates, meaning they saw that the stronger the immune system. And you can measure the immune system on an H&E just looking through the microscope or using TILs, or you can measure it with more complicated immune metrics, what they have done, finding always the same thing.

But architecture is something that we can't measure and that's an important lesson. When I read the manuscript, because I was one of the reviewers, I was wondering would I be able to quantify a pattern for clinical use? And my answer, to be honest, is no. I can score the TILs very easily. I can do a CD-8. You can do genomics for immune system. Practically, assessing the immune system is important, whatever way you do it. But assessing a pattern, for that, you need AI.

What they did, they combined everything together. Now, what they are now doing as a next step is validating this predictive model in the subsequent trial. And that's important how we want to progress because a phase II trial is informative and that's it. So, conceptually, it doesn't change clinical practice. So, it doesn't mean that we have to start immediately implementing this predictive model in our daily practice.

Host: So, validation is critical, and you just beautifully described how this study took. And by the way, the antibody drug conjugate name is HRA 1811. So, it's just an ADC that hone on HER2. And, again, I want to emphasize this was HER2-positive in this trial. But you mentioned this multimodal approach to AI models and multimodal meaning H&E, looking at lymphocytes that are stained, looking at HER2 cells, tumor cells that are stained, all this with AI, but also the clinical, the pathologic and the DNA and RNA data. This is the power of AI. And our piece, the pathologist is important and we want to get to a point where can we start having an algorithm to provide our piece, right? And say this is dense, this is architecturally heterogeneous, and architecture whether it's dense versus spread out of HER2-positive cells matter, and can you put a number on it maybe with an AI model? That's very exciting, and it's really germane to why this podcast, I always want to focus on studies like that, that bring the future power of AI, of genomics. And this combined both. It's a good point now, you know, besides the study. And I refer you, for the audience, to the figure in the editorial. Beautiful figure simplifying, describing exactly what Roberto shared with you about the study from our point of view as pathologists. Can you say a few words about the HER2-low and ultra-low, the ADC landscape and anything from our side that we need to know as pathologists or in any solid tumor, since I know you do other than breast?

Roberto Salgado, MD: Absolutely. The first element is how should we measure HER2? So, the DESTINY-04 has introduced HER2-low in the public domain. The DESTINY-06 trial introduced a concept of HER2 ultra-low. So, what we don't know based on these two trials is whether no expression, zero, by the pathologist predicts benefit. This is being tested in an additional prospective trial called the DESTINY-15 trial, which will include patients with score zero, no expression at all. And then, we will know. And here comes a thing, we have been trained, all of us, by the concept of clinical utility and clinical validity. So, you need prospective randomization.

And now, let's think critically as pathologists. And here comes a nuance about how we translate clinical trial findings to daily practice. The DESTINY-06, the analysis for HER2 ultra-low-- and that particular trial was a magnificent trial, by the way-- it was an exploratory subgroup analysis, and it wasn't an integral biomarker in the trial. Now, think again. It's a research objective. It's an exploratory biomarker. And it was not entailed to the trial. And yet, people jumped at it because it's a drug. You can give a drug and you can find patients who respond. So, imagine, that's what I tell my peers is imagine you have an academic group that does a prospective phase III trial, and they do an exploratory analysis with a particular biomarker. Let's call it TILs or whatever. And then, they go to the FDA. They say, "We want this biomarker to be approved as a biomarker. We want the assay to be approved as a combined diagnostic, and we want the drug to be approved."

 The classical comment also by academics would be, "But you haven't done a prospective randomized clinical trial. Now, here comes big pharma going to the FDA with the same thing, and they get a combined diagnostic. They get a drug approved. And now, here comes the nuance, Daiichi is completely right. Why? Because the biology is the docking station. So, it's completely rational to treat patients with HER2 ultra-low with this drug. Because what the drug does, it binds HER2 and it gets into the cancer cell. So, it's perfect.

Host: For the drug, and that's the only thing we need for the drug. But we'll see if the zero complete zero, no HER2 expression at all will work. But then, how are you going to argue for that in terms of using docking, if it's not on the surface of cells? It could be that it is in a way, but it's not being labeled. But we'll deal with that on another podcast, on that trial. You said it's DESTINY-15, right?

Roberto Salgado, MD: DESTINY-15. And you know what would solve a lot of things is coming back to the AI spectrum in the digital pathology space. What I firmly believe based on evidence is that we need clinical trials to validate these assays. So, what we see in the literature in all journals is a lot of wonderful science based on large retrospective data sets or publicly available datasetS who are very informative to the community, but they are not driving the change in daily practice. What is driving the change in daily practice is the validation clinically in large, randomized clinical trials that have been finalized to give you a very concrete example.

We have just finalized with Friends of Cancer Research a comparison of up to 10 different HER2 assays using AI. And we did see variability, the same variability that we see between pathologists reading AI. And we see the same artifacts causing this variability. So, nothing that surprising to be honest.

The question now is does it matter? Does it matter if you have one AI assay who scores a zero and the other one who scores one plus? Does it matter when you have one AI assay who scores ultra-low and the other one scores HER2-low? The only way in which we will know is to test these assays in trials that have been finalized. And industry, and I work a lot with industry and I love them really, they're extremely reluctant to use their samples for these type of comparisons for a variety of reasons. FDA is not willing to do them, and they're scared for the label, at this and that, and perfectly understandable. But we need to go beyond these doubts.

So as a community, and I will use this podcast to spread the message, we need to get clinical trials available for validation of all those wonderful digital AI assays that we are developing. Otherwise, we will not be able to bring that level of evidence to the community.

Host: Wonderful. This has been very enjoyable. It sounds like we need another episode to talk further about AI and immuno-oncology with you, and I'm sure the opportunity will come. I'm really grateful for the great conversation. And I wish you and the working group the best.

Roberto Salgado, MD: Thank you very much. So, thank you very much for the invitation, and I would love to continue to engage with you and all your activities. Thank you very much.

Host: Thank you. It's such an honor. For our audience, I hope you enjoyed it. Please listen to us. And like I just mentioned, now we have a vodcast. So on YouTube, you can see the episode and see the successful pathologist Roberto Salgado on video too. And until next episode.

Outro: We would like to extend our sincere gratitude to today's guests for their valuable contributions and thoughtful perspectives. We also wish to thank our dedicated production team and the technical team at DoctorPodcasting. Special thanks to Jennifer Bepler for her dedication and skill in coordinating our episodes.

Please note the opinions expressed by our speakers are their own and do not necessarily represent the Perelman School of Medicine or the University of Pennsylvania. Thank you for listening, and we look forward to welcoming you back for future episodes.