ASCO Daily News How AI Can Improve Patient Identification and Recruitment for Clinical Trials
Dr. Shaalan Beg and Dr. Arturo Loaiza-Bonilla discuss the potential of artificial intelligence to assist with patient recruitment and clinical trial matching using real-world data and next-generation sequencing results.
TRANSCRIPT
Dr. Shaalan Beg: Hello, and welcome to the ASCO Daily News Podcast. I'm Dr. Shaalan Beg, your guest host for the podcast today. I'm an adjunct associate professor at UT Southwestern's Simmons Comprehensive Cancer Center in Dallas and senior advisor for clinical research at the National Cancer Institute. On today's episode, we will be discussing the promise of artificial intelligence to improve patient recruitment in clinical trials and advanced clinical research. Joining me for this discussion is Dr. Arturo Loaiza-Bonilla, the medical director of oncology research at Capital Health in Philadelphia. He's also the co-founder and chief medical officer at Massive Bio, an AI-driven platform that matches patients with clinical trials and novel therapies.
Our full disclosures are available in the transcript of this episode.
Arturo, it's great to have you on the podcast today.
Dr. Arturo Loaiza-Bonilla: Thanks so much, Shaalan. It's great to be here and talking to you today.
Dr. Shaalan Beg: So we're all familiar with the limitations and inefficiencies in patient recruitment for clinical trials, but there are exciting new technologies that are addressing these challenges. Your group developed a first-in-class, AI-enabled matching system that's designed to automate and expedite processes using real-world data and integrating next-generation sequencing results into the algorithm. You presented work at the ASCO Annual Meeting this year where you showed the benefits of AI and NGS in clinical trial matching and you reported about a twofold increase in potential patient eligibility for trials. Can you tell us more about this study?
Dr. Arturo Loaiza-Bonilla: Absolutely. And this is just part of the work that we have seen over the last several years, trying to overcome challenges that are coming because of all these, as you mentioned, inefficiencies and limitations, particularly in the manual patient trial matching. This is very time consuming, as all of us know; many of those in the audience as well experience it on a daily basis, and it's resource intensive. It takes specialized folks who are able to understand the nuances in oncology, and it takes, on average, even for the most experienced research coordinator or principal investigator oncologist, 25 minutes per trial. Not only on top of that, but in compound there's a lack of comprehensive genomic testing, NGS, and that complicates the process in terms of inability to know what patients are eligible for, and it can delay also the process even further.
So, to address those issues, we at Massive Bio are working with other institutions, and we're part of this … called the Precision Cancer Consortium, which is a combination of 7 of the top 20 top pharma companies in oncology, and we got them together. And let's say, okay, the only way to show something that is going to work at scale is people have to remove their silos and barriers and work as a collaborative approach. If we're going to be able to get folks tested more often and in more patients, assess for clinical trials, at least as an option, we need to understand further the data. And after a bunch of efforts that happened, and you're also seeing those efforts in CancerX and other things that we're working on together, but what we realize here is using an AI-enabled matching system to basically automate and expedite the process using what we call real-world data, which is basically data from patients that are actually currently being treated, and integrating any NGS results and comparing that to what we can potentially do manually. The idea was to do multi-trial matching, because if we do it for one study, yeah, it will be interesting, but it will not show the potential applicability in the real world.
So with all that background, the tool itself, just to give you the punchline of it, was proven highly effective in terms of efficiency. We were able to increase the number of potential matches, and not only that, but reducing the time to the matching. So basically, instead of spending 25 minutes, it could be done in a matter of seconds. And when you compound all that across multiple clinical trials, in this case, it was several sponsors coming together, we were able to reduce the manual effort of seeing patients and testing for clinical trials to basically 1 hour when it would have otherwise taken a ridiculous amount of time. And it was quantified as 19,500 hours of manual work, compared to 1 hour done by the system to uniquely match a cohort of about 5,600 patients that came into the platform. And this was across 23 trials. Now imagine if we can do it for the 14,000 clinical trials currently in clinicaltrials.gov.
So for us, this kind of was an eye-opening situation that if we can increase not only the efficiency but find even more trials by integrating comprehensive genomic testing, which in this case was a twofold increase in eligibility for clinical trials, that gives us not only the opportunity for optimized processes using AI but also a call to action that there is still a lot of under-genotyping. And I know American Cancer Society and ASCO and many others are working hard on getting that into fruition, but we need to have systems that remind us that certain patients are not tested yet and that can improve not only real patients, but the R&D and the process of innovation in the future.
Dr. Shaalan Beg: Yeah, it's always an important reminder that even some of the highest impact IT solutions or AI solutions are most effective if they can be integrated into our normal clinical processes and into the normal workflow that we have in our clinics to help clinicians do their work quickly and more efficiently.
Can you talk about how, over the last few years, the availability of NGS data in our electronic medical record (EMR) has evolved and whether that's evolving for the better? And what are some next steps in terms of making that data available at EMR so that such solutions can then pull that data out and do clinical trial matching?
Dr. Arturo Loaiza-Bonilla: Yes. So one of the things that we have seen over the last couple of years is because of the applicability of the 21st Century Cures Act, there is less "information blocking," which is patients not being able to access their information in real time. Now, with the appearance of health exchanges, with patient-centric approaches, which is something that many innovators, including ours, are trying to apply, it's really becoming more relevant. So it's not only helping us to find the patients when they really need to get tested, but also is giving us the opportunity to put those patients into the right treatment pathway when found.
Something that's still a challenge and I think we can work by being more collaborative once again – is my dream – is having these pre-screening hubs where no matter where you are in your cancer journey, you just go into that funnel and then are able to see, "Okay, you are in the second-line setting for non-small cell lung cancer, EGFR-mutated. Now, do you have a meta amplification, then you go for this study or this trial. Oh, you haven't been tested yet. You should get tested. You're a pancreas cancer patient who is KRAS wild type; well, there is a significant chance that you may have a biomarker because that's where most patients are enriched for."
So having that opportunity to at scale, just for the whole country, to get those patients access to that information, I think is crucial for the future of oncology. And I think you working at the NCI, more than most, know how the impact of that can help for those underrepresented patients to get more access to better treatment options and whatnot. And we can activate clinical trials as well in new models, decentralized models, adjusting time models, all those things can be leveraged by using biomarker testing in real time. Identification when the patient really needs a trial option or a medication option, because the data is telling us when to activate that in real time.
Dr. Shaalan Beg: And identifying the patient for a potential clinical trial is one challenge. In oncology, given a lot of our trials, we are looking to enroll people at a specific time in their disease journey. So we call it first-line or second-line or third-line, becomes the next challenge. So just knowing someone has mutation number 1, 2, or 3 isn't enough to say they would be eligible for a second-line BRAF X, Y and Z mutation at a given trial. I've heard you talk a lot about this last-mile navigation for people once you've identified that they may be a soft match for a clinical trial. Can you talk about what you've seen in the ecosystem being developed on how AI is helping both clinics and patients navigate this last mile from the time they're identified for a clinical trial to the time they actually receive cycle 1, day 1?
Dr. Arturo Loaiza-Bonilla: Yeah, absolutely. And that is such a critical point because, as you know, we have helped tons of patients getting trial options in thousands of cases. But even my own patients, I give them a report for trial options and they're like, "Okay, I still need help." And we have been talking with ASCO, with the American Cancer Society, and many other very good teams, and what we see as an opportunity in technology here is leveraging those cancer journeys to know when the patient really has the opportunity to enroll in a trial, because this is a very dynamic environment. Not only the patient's condition changes because their cancer progresses, the hemoglobin changes, the cancer moves from one place to the other, and there's nuances in between, but also new medications are coming up, studies open and close, sites open and close.
So having this information as a hub, as what we call a command center, is the key to make this happen. And we can use the same tools that we use for Uber or for Instacart or whichever thing you want to do; it's already the same concept. When you need groceries, you don't need groceries every day. But Amazon gives you a ding that's like, "Well, I think you may be running out of milk," because they already know how often you buy it, or just having the data behind the scenes of how typically these, in this case, patient journeys, may manifest based on the biomarker. So let's say a smoldering multiple myeloma is not the same across. One patient with biomarkers that make them very high risk, the risk of progressing to a multiple myeloma, first-line treatment-eligible patient is going to be much different than someone who has better risk cytogenetics. So using that tool to optimize the cancer journeys of those patients and being able to notify them in real time of new trial options, and also knowing when the patient really has that disease progression so there's a time of activation for trial matching again, the same way you get a credit score for buying a house, then you know exactly what options are in front of you at that very moment. And that is the last-mile component, which is going to be key.
What we have seen that we feel is important to invest on, and we have invested heavily on it, is that until the patient doesn't sign the consent form for the clinical trial, that patient is completely unknown to most people. The site doesn't know them because they haven't been there, and they may be there, but they don't know about the options sometimes. But no one's going to invest in getting that patient to the finish line. There's a lot of support for patients on trials, but not before they enroll on trials. And we feel that this is a big opportunity to really exponentially grow the chances of patients enrolling in trials if we support them all the way from the very time they get diagnosed with cancer in any setting. And we can help that patient on a very unique journey to find the trial options using technology. So it's very feasible. We see it once again in many other equally complex tasks, so why not do it in oncology when we have all the bonafides across wanting to do this.
Dr. Shaalan Beg: Can you give examples of where you are seeing it done outside of oncology that's a model that one can replicate?
Dr. Arturo Loaiza-Bonilla: I mean, oncology is the toughest use case to crack. You have experiences with DCTs in the past and all that. So the big opportunities are for patients, for example, in psychiatry, when they need certain counseling and help. We see that also in medical devices, when people have diabetes and they really need a device specifically for that unique situation, or also for patients with cardiovascular risk that they can in real time get access to novel therapeutics. And that's how they have been able to enroll so quickly. And all these GLP-1 inhibitors, all those models are really almost completely decentralized nowadays in something that we can extrapolate for oncology once we have aligned the ecosystem to make it see them. This is something that we can really revolutionize care while we manage all the complex variables that typically come with oncology uses.
Dr. Shaalan Beg: I would imagine while you translate those learnings from outside of oncology into oncology, a lot of those processes will be human and AI combination activities. And as you learn more and more, the human component becomes a smaller fraction, and the technology and the AI becomes more of a component. Are you seeing a similar transition in the clinical trial matching space as well?
Dr. Arturo Loaiza-Bonilla: Yes. So that's why people say humans are going to be replaced. They're not. Patients still want to see a human face that they recognize, they trust. Even family members of mine want to hear from me, even if they are in the top place in the world. What we can change with technology are those things that are typically just friction points. In this case, information gathering, collecting records, getting the data structured in a way that we can use it for matching effectively, knowing in real time when the patient progresses, so we can really give them the chances of knowing what's available in real time. And collecting the information from all these other stakeholders. Like, is the site open? Is the budget approved for that place? Is the insurance allowing the specific … do they have e-consent? Those things can be fully automated because they're just burdensome. They're not helping anyone. And we can really make it decentralized for e-consent, for just getting a screening. They don't need to be screened at the site for something that they're going to receive standard of care. We can really change that, and that's something that we're seeing in the space that is changing, and hopefully we can translate it fully in oncology once we are getting the word out. And I think this is a good opportunity to do so.
Dr. Shaalan Beg: You talked about your dream scenario for clinical trial matching. When you think about your dream scenario as a practicing oncologist, what are the AI tools that you are most excited about making their way into the clinic, either wishful thinking or practically?
Dr. Arturo Loaiza-Bonilla: I typically get feedback from all over the place on doing this, and I also have my own thoughts. But I always come to this for a reason. We all became physicians and oncologists because we like being physicians. We like to talk to patients. We want to spend the time. I tell folks in my clinic, I will see a thousand patients all the time as long as I don't have to do notes, as long as I don't have to place orders. But of course, they will have to hire 1,000 people ancillary to do all the stuff that we do.
If we can go back and spend all that time that we use on alert fatigue, on clicking, on gathering things, fighting insurance, and really helping align those incentives with clinical trials and biomarker testing and really making it a mankind or a humankind situation where we're all in this really together to solve the problem, which is cancer, that will be my dream come true. So I don't have to do anything that is clerical, that is not really helping me, but I want to use that AI to liberate me from that and also use the data that is generated for better insights.
I think that I know my subject of expertise, but there's so many things happening all the time that it is hard to keep up, no matter how smart you are. If the tool can give me insights that I didn't even know, then leverage that as a CME or a board certification, that would be a dream come true. Of course, I'm just dreaming here, but it's feasible. Many of these ideas, as I mentioned, they're not new. The key thing is getting them done. The innovative part is getting stuff done, because I'm sure there's a gazillion people who have the same ideas as I did, but they just don't know whom to talk to or who is going to make it happen in reality. And that's my call to action to people: Let's work together and make this happen.
Dr. Shaalan Beg: Well, Arturo, thanks a lot for sharing your insights with us today on the ASCO Daily News Podcast.
Dr. Arturo Loaiza-Bonilla: Well, thank you so much for the time and looking forward to having more exchanges and conversations and seeing everyone in the field.
Dr. Shaalan Beg: And thank you to our listeners for your time today. You'll find a link to the studies discussed today in the transcript of this episode. And if you value the insights that you hear on the 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.
Find out more about today's speakers:
Follow ASCO on social media:
Disclosures:
Dr. Arturo Loaiza-Bonilla:
Leadership: Massive Bio
Stock and Other Ownership Interests: Massive Bio
Consulting or Advisory Role: Massive Bio, Bayer, PSI, BrightInsight, Cardinal Health, Pfizer, Eisai, AstraZeneca, Regeneron, Verily, Medscape
Speakers' Bureau: Guardant Health, Bayer, Amgen, Ipsen, AstraZeneca/Daiichi Sankyo, Natera
Dr. Shaalan Beg:
Consulting or Advisory Role: Ispen, Cancer Commons, Foundation Medicine, Genmab/Seagen
Speakers' Bureau: Sirtex
Research Funding (An Immediate Family Member): ImmuneSensor Therapeutics
Research Funding (Institution): Bristol-Myers Squibb, Tolero Pharmaceuticals, Delfi Diagnostics, Merck, Merck Serono, AstraZeneca/MedImmune