From Clinic to Clinical Trials: Responsible AI Integration in Oncology

Dr. Paul Hanona and Dr. Arturo Loaiza-Bonilla discuss how to safely and smartly integrate AI into the clinical workflow and tap its potential to improve patient-centered care, drug development, and access to clinical trials.

TRANSCRIPT

Dr. Paul Hanona: Hello, I'm Dr. Paul Hanona, your guest host of the ASCO Daily News Podcast today. I am a medical oncologist as well as a content creator @DoctorDiscover, and I'm delighted to be joined today by Dr. Arturo Loaiza-Bonilla, the chief of hematology and oncology at St. Luke's University Health Network. Dr. Bonilla is also the co-founder and chief medical officer at Massive Bio, an AI-driven platform that matches patients with clinical trials and novel therapies. Dr. Loaiza-Bonilla will share his unique perspective on the potential of artificial intelligence to advance precision oncology, especially through clinical trials and research, and other key advancements in AI that are transforming the oncology field.

Our full disclosures are available in the transcript of the episode.

Dr. Bonilla, it's great to be speaking with you today. Thanks for being here.

Dr. Arturo Loaiza-Bonilla: Oh, thank you so much, Dr. Hanona. Paul, it's always great to have a conversation. Looking forward to a great one today.

Dr. Paul Hanona: Absolutely. Let's just jump right into it. Let's talk about the way that we see AI being embedded in our clinical workflow as oncologists. What are some practical ways to use AI?

Dr. Arturo Loaiza-Bonilla: To me, responsible AI integration in oncology is one of those that's focused on one principle to me, which is clinical purpose is first, instead of the algorithm or whatever technology we're going to be using. If we look at the best models in the world, they're really irrelevant unless we really solve a real day-to-day challenge, either when we're talking to patients in the clinic or in the infusion chair or making decision support.

Currently, what I'm doing the most is focusing on solutions that are saving us time to be more productive and spend more time with our patients. So, for example, we're using ambient AI for appropriate documentation in real time with our patients. We're leveraging certain tools to assess for potential admission or readmission of patients who have certain conditions as well. And it's all about combining the listening of physicians like ourselves who are end users, those who create those algorithms, data scientists, and patient advocates, and even regulators, before they even write any single line of code. I felt that on my own, you know, entrepreneurial aspects, but I think it's an ethos that we should all follow.

And I think that AI shouldn't be just bolted on later. We always have to look at workflows and try to look, for example, at clinical trial matching, which is something I'm very passionate about. We need to make sure that first, it's easier to access for patients, that oncologists like myself can go into the interface and be able to pull the data in real time when you really need it, and you don't get all this fatigue alerts. To me, that's the responsible way of doing so. Those are like the opportunities, right?

So, the challenge is how we can make this happen in a meaningful way – we're just not reacting to like a black box suggestion or something that we have no idea why it came up to be. So, in terms of success – and I can tell you probably two stories of things that we know we're seeing successful – we all work closely with radiation oncologists, right? So, there are now these tools, for example, of automated contouring in radiation oncology, and some of these solutions were brought up in different meetings, including the last ASCO meeting. But overall, we know that transformer-based segmentation tools; transformer is just the specific architecture of the machine learning algorithm that has been able to dramatically reduce the time for colleagues to spend allotting targets for radiation oncology. So, comparing the target versus the normal tissue, which sometimes it takes many hours, now we can optimize things over 60%, sometimes even in minutes. So, this is not just responsible, but it's also an efficiency win, it's a precision win, and we're using it to adapt even mid-course in response to tumor shrinkage.

Another success that I think is relevant is, for example, on the clinical trial matching side. We've been working on that and, you know, I don't want to preach to the choir here, but having the ability for us to structure data in real time using these tools, being able to extract information on biomarkers, and then show that multi-agentic AI is superior to what we call zero-shot or just throwing it into ChatGPT or any other algorithm, but using the same tools but just fine-tuned to the point that we can be very efficient and actually reliable to the level of almost like a research coordinator, is not just theory. Now, it can change lives because we can get patients enrolled in clinical trials and be activated in different places wherever the patient may be. I know it's like a long answer on that, but, you know, as we talk about responsible AI, that's important.

And in terms of what keeps me up at night on this: data drift and biases, right? So, imaging protocols, all these things change, the lab switch between different vendors, or a patient has issues with new emerging data points. And health systems serve vastly different populations. So, if our models are trained in one context and deployed in another, then the output can be really inaccurate. So, the idea is to become a collaborative approach where we can use federated learning and patient-centricity so we can be much more efficient in developing those models that account for all the populations, and any retraining that is used based on data can be diverse enough that it represents all of us and we can be treated in a very good, appropriate way. So, if a clinician doesn't understand why a recommendation is made, as you probably know, you probably don't trust it, and we shouldn't expect them to. So, I think this is the next wave of the future. We need to make sure that we account for all those things.

Dr. Paul Hanona: Absolutely. And even the part about the clinical trials, I want to dive a little bit more into in a few questions. I just kind of wanted to make a quick comment. Like you said, some of the prevalent things that I see are the ambient scribes. It seems like that's really taken off in the last year, and it seems like it's improving at a pretty dramatic speed as well. I wonder how quickly that'll get adopted by the majority of physicians or practitioners in general throughout the country. And you also mentioned things with AI tools regarding helping regulators move things quicker, even the radiation oncologist, helping them in their workflow with contouring and what else they might have to do. And again, the clinical trials thing will be quite interesting to get into.

The first question I had subsequent to that is just more so when you have large datasets. And this pertains to two things: the paper that you published recently regarding different ways to use AI in the space of oncology referred to drug development, the way that we look at how we design drugs, specifically anticancer drugs, is pretty cumbersome. The steps that you have to take to design something, to make sure that one chemical will fit into the right chemical or the structure of the molecule, that takes a lot of time to tinker with. What are your thoughts on AI tools to help accelerate drug development?

Dr. Arturo Loaiza-Bonilla: Yes, that's the Holy Grail and something that I feel we should dedicate as much time and effort as possible because it relies on multimodality. It cannot be solved by just looking at patient histories. It cannot be solved by just looking at the tissue alone. It's combining all these different datasets and being able to understand the microenvironment, the patient condition and prior treatments, and how dynamic changes that we do through interventions and also exposome – the things that happen outside of the patient's own control – can be leveraged to determine like what's the best next step in terms of drugs.

So, the ones that we heard the news the most is, for example, the Nobel Prize-winning [for Chemistry awarded to Demis Hassabis and John Jumper for] AlphaFold, an AI system that predicts protein structures right? So, we solved this very interesting concept of protein folding where, in the past, it would take the history of the known universe, basically – what's called the Levinthal's paradox – to be able to just predict on amino acid structure alone or the sequence alone, the way that three-dimensionally the proteins will fold. So, with that problem being solved and the Nobel Prize being won, the next step is, "Okay, now we know how this protein is there and just by sequence, how can we really understand any new drug that can be used as a candidate and leverage all the data that has been done for many years of testing against a specific protein or a specific gene or knockouts and what not?"

So, this is the future of oncology and where we're probably seeing a lot of investments on that. The key challenge here is mostly working on the side of not just looking at pathology, but leveraging this digital pathology with whole slide imaging and identifying the microenvironment of that specific tissue. There's a number of efforts currently being done. One isn't just H&E, like hematoxylin and eosin, slides alone, but with whole imaging, now we can use expression profiles, spatial transcriptomics, and gene whole exome sequencing in the same space and use this transformer technology in a multimodality approach that we know already the slide or the pathology, but can we use that to understand, like, if I knock out this gene, how is the microenvironment going to change to see if an immunotherapy may work better, right? If we can make a microenvironment more reactive towards a cytotoxic T cell profile, for example.

So, that is the way where we're really seeing the field moving forward, using multimodality for drug discovery. So, the FDA now seems to be very eager to support those initiatives, so that's of course welcome. And now the key thing is the investment to do this in a meaningful way so we can see those candidates that we're seeing from different companies now being leveraged for rare disease, for things that are going to be almost impossible to collect enough data, and make it efficient by using these algorithms that sometimes, just with multiple masking – basically, what they do is they mask all the features and force the algorithm to find solutions based on the specific inputs or prompts we're doing. So, I'm very excited about that, and I think we're going to be seeing that in the future.

Dr. Paul Hanona: So, essentially, in a nutshell, we're saying we have the cancer, which is maybe a dandelion in a field of grass, and we want to see the grass that's surrounding the dandelion, which is the pathology slides. The problem is, to the human eye, it's almost impossible to look at every single piece of grass that's surrounding the dandelion. And so, with tools like AI, we can greatly accelerate our study of the microenvironment or the grass that's surrounding the dandelion and better tailor therapy, come up with therapy. Otherwise, like you said, to truly generate a drug, this would take years and years. We just don't have the throughput to get to answers like that unless we have something like AI to help us.

Dr. Arturo Loaiza-Bonilla: Correct.

Dr. Paul Hanona: And then, clinical trials. Now, this is an interesting conversation because if you ever look up our national guidelines as oncologists, there's always a mention of, if treatment fails, consider clinical trials. Or in the really aggressive cancers, sometimes you might just start out with clinical trials. You don't even give the standard first-line therapy because of how ineffective it is. There are a few issues with clinical trials that people might not be aware of, but the fact that the majority of patients who should be on clinical trials are never given the chance to be on clinical trials, whether that's because of proximity, right, they might live somewhere that's far from the institution, or for whatever reason, they don't qualify for the clinical trial, they don't meet the strict inclusion criteria.

But a reason you mentioned early on is that it's simply impossible for someone to be aware of every single clinical trial that's out there. And then even if you are aware of those clinical trials, to actually find the sites and put in the time could take hours. And so, how is AI going to revolutionize that? Because in my mind, it's not that we're inventing a new tool. Clinical trials have always been available. We just can't access them. So, if we have a tool that helps with access, wouldn't that be huge?

Dr. Arturo Loaiza-Bonilla: Correct. And that has been one of my passions. And for those who know me and follow me and we've spoke about it in different settings, that's something that I think we can solve. This other paradox, which is the clinical trial enrollment paradox, right? We have tens of thousands of clinical trials available with millions of patients eager to learn about trials, but we don't enroll enough and many trials close to accrual because of lack of enrollment. It is completely paradoxical and it's because of that misalignment because patients don't know where to go for trials and sites don't know what patients they can help because they haven't reached their doors yet.

So, the solution has to be patient-centric, right? We have to put the patient at the center of the equation. And that was precisely what we had been discussing during the ASCO meeting. There was an ASCO Education Session where we talked about digital prescreening hubs, where we, in a patient-centric manner, the same way we look for Uber, Instacart, any solution that you may think of that you want something that can be leveraged in real time, we can use these real-world data streams from the patient directly, from hospitals, from pathology labs, from genomics companies, to continuously screen patients who can match to the inclusion/exclusion criteria of unique trials. So, when the patient walks into the clinic, the system already knows if there's a trial and alerts the site proactively. The patient can actually also do decentralization. So, there's a number of decentralized clinical trial solutions that are using what I call the "click and mortar" approach, which is basically the patient is checking digitally and then goes to the site to activate. We can also have the click and mortar in the bidirectional way where the patient is engaged in person and then you give the solution like the ones that are being offered on things that we're doing at Massive Bio and beyond, which is having the patient to access all that information and then they make decisions and enroll when the time is right.

As I mentioned earlier, there is this concept drift where clinical trials open and close, the patient line of therapy changes, new approvals come in and out, and sites may not be available at a given time but may be later. So, having that real-time alerts using tools that are able already to extract data from summarization that we already have in different settings and doing this natural language ingestion, we can not only solve this issue with manual chart review, which is extremely cumbersome and takes forever and takes to a lot of one-time assessments with very high screen failures, to a real-time dynamic approach where the patient, as they get closer to that eligibility criteria, they get engaged. And those tools can be built to activate trials, audit trials, and make them better and accessible to patients. And something that we know is, for example, 91%-plus of Americans live close to either a pharmacy or an imaging center. So, imagine that we can potentially activate certain of those trials in those locations. So, there's a number of pharmacies, special pharmacies, Walgreens, and sometimes CVS trying to do some of those efforts.

So, I think the sky's the limit in terms of us working together. And we've been talking with corporate groups, they're all interested in those efforts as well, to getting patients digitally enabled and then activate the same way we activate the NCTN network of the corporate groups, that are almost just-in-time. You can activate a trial the patient is eligible for and we get all these breakthroughs from the NIH and NCI, just activate it in my site within a week or so, as long as we have the understanding of the protocol. So, using clinical trial matching in a digitally enabled way and then activate in that same fashion, but not only for NCTN studies, but all the studies that we have available will be the key of the future through those prescreening hubs. So, I think now we're at this very important time where collaboration is the important part and having this silo-breaking approach with interoperability where we can leverage data from any data source and from any electronic medical records and whatnot is going to be essential for us to move forward because now we have the tools to do so with our phones, with our interests, and with the multiple clinical trials that are coming into the pipelines.

Dr. Paul Hanona: I just want to point out that the way you described the process involves several variables that practitioners often don't think about. We don't realize the 15 steps that are happening in the background. But just as a clarifier, how much time is it taking now to get one patient enrolled on a clinical trial? Is it on the order of maybe 5 to 10 hours for one patient by the time the manual chart review happens, by the time the matching happens, the calls go out, the sign-up, all this? And how much time do you think a tool that could match those trials quicker and get you enrolled quicker could save? Would it be maybe an hour instead of 15 hours? What's your thought process on that?

Dr. Arturo Loaiza-Bonilla: Yeah, exactly. So one is the matching, the other one is the enrollment, which, as you mentioned, is very important. So, it can take, from, as you said, probably between 4 days to sometimes 30 days. Sometimes that's how long it takes for all the things to be parsed out in terms of logistics and things that could be done now agentically. So, we can use agents to solve those different steps that may take multiple individuals. We can just do it as a supply chain approach where all those different steps can be done by a single agent in a simultaneous fashion and then we can get things much faster. With an AI-based solution using these frontier models and multi-agentic AI – and we presented some of this data in ASCO as well – you can do 5,000 patients in an hour, right? So, just enrolling is going to be between an hour and maximum enrollment, it could be 7 days for those 5,000 patients if it was done at scale in a multi-level approach where we have all the trials available.

Dr. Paul Hanona: No, definitely a very exciting aspect of our future as oncologists. It's one thing to have really neat, novel mechanisms of treatment, but what good is it if we can't actually get it to people who need it? I'm very much looking for the future of that.

One of the last questions I want to ask you is another prevalent way that people use AI is just simply looking up questions, right? So, traditionally, the workflow for oncologists is maybe going on national guidelines and looking up the stage of the cancer and seeing what treatments are available and then referencing the papers and looking at who was included, who wasn't included, the side effects to be aware of, and sort of coming up with a decision as to how to treat a cancer patient. But now, just in the last few years, we've had several tools become available that make getting questions easier, make getting answers easier, whether that's something like OpenAI's tools or Perplexity or Doximity or OpenEvidence or even ASCO has a Guidelines Assistant as well that is drawing from their own guidelines as to how to treat different cancers. Do you see these replacing traditional sources? Do you see them saving us a lot more time so that we can be more productive in clinic? What do you think is the role that they're going to play with patient care?

Dr. Arturo Loaiza-Bonilla: Such a relevant question, particularly at this time, because these AI-enabled query tools, they're coming left and right and becoming increasingly common in our daily workflows and things that we're doing. So, traditionally, when we go and we look for national guidelines, we try to understand the context ourselves and then we make treatment decisions accordingly. But that is a lot of a process that now AI is helping us to solve.

So, at face value, it seems like an efficiency win, but in many cases, I personally evaluate platforms as the chief of hem/onc at St. Luke's and also having led the digital engagement things through Massive Bio and trying to put things together, I can tell you this: not all tools are created equal. In cancer care, each data point can mean the difference between cure and progression, so we cannot really take a lot of shortcuts in this case or have unverified output. So, the tools are helpful, but it has to be grounded in truth, in trusted data sources, and they need to be continuously updated with, like, ASCO and NCCN and others.

So, the reason why the ASCO Guidelines Assistant, for instance, works is because it builds on all these recommendations, is assessed by end users like ourselves. So, that kind of verification is critical, right? We're entering a phase where even the source material may be AI-generated. So, the role of human expert validation is really actually more important, not less important. You know, generalist LLMs, even when fine-tuned, they may not be enough. You can pull a few API calls from PubMed, etc., but what we need now is specialized, context-aware, agentic tools that can interpret multimodal and real-time clinical inputs. So, something that we are continuing to check on and very relevant to have entities and bodies like ASCO looking into this so they can help us to be really efficient and really help our patients.

Dr. Paul Hanona: Dr. Bonilla, what do you want to leave the listener with in terms of the future direction of AI, things that we should be cautious about, and things that we should be optimistic about?

Dr. Arturo Loaiza-Bonilla: Looking 5 years ahead, I think there's enormous promise. As you know, I'm an AI enthusiast, but always, there's a few priorities that I think – 3 of them, I think – we need to tackle head-on. First is algorithmic equity. So, most AI tools today are trained on data from academic medical centers but not necessarily from community practices or underrepresented populations, particularly when you're looking at radiology, pathology, and what not. So, those blind spots, they need to be filled, and we can eliminate a lot of disparities in cancer care. So, those frameworks to incentivize while keeping the data sharing using federated models and things that we can optimize is key.

The second one is the governance on the lifecycle. So, you know, AI is not really static. So, unlike a drug that is approved and it just, you know, works always, AI changes. So, we need to make sure that we have tools that are able to retrain and recall when things degrade or models drift. So, we need to use up-to-date AI for clinical practice, so we are going to be in constant revalidation and make it really easy to do.

And lastly, the human-AI interface. You know, clinicians don't need more noise or we don't need more black boxes. We need decision support that is clear, that we can interpret, and that is actionable. "Why are you using this? Why did we choose this drug? Why this dose? Why now?" So, all these things are going to help us and that allows us to trace evidence with a single click. So, I always call it back to the Moravec's paradox where we say, you know, evolution gave us so much energy to discern in the sensory-neural and dexterity. That's what we're going to be taking care of patients. We can use AI to really be a force to help us to be better clinicians and not to really replace us. So, if we get this right and we decide for transparency with trust, inclusion, etc., it will never replace any of our work, which is so important, as much as we want, we can actually take care of patients and be personalized, timely, and equitable. So, all those things are what get me excited every single day about these conversations on AI.

Dr. Paul Hanona: All great thoughts, Dr. Bonilla. I'm very excited to see how this field evolves. I'm excited to see how oncologists really come to this field. I think with technology, there's always a bit of a lag in adopting it, but I think if we jump on board and grow with it, we can do amazing things for the field of oncology in general. Thank you for the advancements that you've made in your own career in the field of AI and oncology and just ultimately with the hopeful outcomes of improving patient care, especially cancer patients.

Dr. Arturo Loaiza-Bonilla: Thank you so much, Dr. Hanona.

Dr. Paul Hanona: Thanks to our listeners for your time today. If you value the insights that you hear on ASCO Daily News Podcast, please take a moment to rate, review, and subscribe wherever you get your podcasts.

Disclaimer:

The purpose of this podcast is to educate and to inform. This is not a substitute for professional medical care and is not intended for use in the diagnosis or treatment of individual conditions. Guests on this podcast express their own opinions, experience, and conclusions. Guest statements on the podcast do not express the opinions of ASCO. The mention of any product, service, organization, activity, or therapy should not be construed as an ASCO endorsement.

More on today's speakers:

Dr. Arturo Loaiza-Bonilla

@DrBonillaOnc

Dr. Paul Hanona

@DoctorDiscover on YouTube

Follow ASCO on social media:

@ASCO on Twitter

ASCO on Facebook

ASCO on LinkedIn

ASCO on BlueSky

Disclosures:

Paul Hanona: No relationships to disclose.

Dr. Arturo-Loaiza-Bonilla:

Leadership: Massive Bio

Stock & Other Ownership Interests: Massive Bio

Consulting or Advisory Role: Massive Bio, Bayer, PSI, BrightInsight, CardinalHealth, Pfizer, AstraZeneca, Medscape

Speakers' Bureau: Guardant Health, Ipsen, AstraZeneca/Daiichi Sankyo, Natera