Main Topic In this episode of Oncology Unscripted, Dr Marshall examines the evolving role of blood-based cancer screening and highlights new research showing how dogs may help detect early cancers from breath samples. Alternative Data Centers, Medicine, and AI Guilt? In MedBuzz, Dr Marshall discusses the emergence of betting on clinical trial outcomes, the rise of AI in medicine, the environmental impact of data centers, and whether growing reliance on artificial intelligence should leave clinicians with a sense of AI guilt.
Main Topic
In this episode of Oncology Unscripted, Dr Marshall examines the evolving role of blood-based cancer screening and highlights new research showing how dogs may help detect early cancers from breath samples.
Alternative Data Centers, Medicine, and AI Guilt?
In MedBuzz, Dr Marshall discusses the emergence of betting on clinical trial outcomes, the rise of AI in medicine, the environmental impact of data centers, and whether growing reliance on artificial intelligence should leave clinicians with a sense of AI guilt.
When Dogs Detect Cancer
[00:00:06] John Marshall, MD: John Marshall, Oncology Unscripted. Happy summer. It is a hot summer here in Washington. We did have that thunderstorm last night—some tornadoes not too far away—but we made it. We needed the water. I hope nobody got hurt. I don't think so. I've been thinking a little bit lately about, as we get older, what the proper dress is to go to work in, and I'm still a guy that kind of wears a tie every now and then. I wear a tie on patient days, particularly. During the pandemic, if you're like me, a lot of people stopped wearing ties. That whole infectious disease thing about ties—if I'm wearing a white coat anyway, I'm not really sure about it. I was just in a meeting, a global sort of drug development meeting, and there were only two other men wearing ties in the meeting. And I kind of brought it up, and they all were razzing each other. The younger ones without ties were razzing the older ones with ties. And then I'm not even sure this tie actually works with this shirt, but you know what? What the heck, at my age and on Oncology Unscripted, I can pretty much wear what I want.
So, welcome to our July episode of Oncology Unscripted.
A lot has been going on since ASCO, but I wanna drill down on our main topic, as it is something that really is starting to shift how we're thinking about screening for cancer. Now, you all know, and I know, that there are only a few cancers that we typically screen for, right? They are breast and colon, and now lung if you're a smoker, as well as skin cancers and cervical cancer.
But what screening was supposed to be was inexpensive and noninvasive, with high sensitivity and high specificity.
But as you know, there's been an increasing shift and emphasis on screening for cancers in general. And so, the latest one that just came out is actually pretty controversial because we've been talking a lot about precision medicine and being able to find tumor/cancer DNA in the blood.
We've been doing this as a predictor of recurrence in patients with known cancer. But as you also know, there has been increasing work in patients who are well, who are just going in to be seen by a doctor and get a checkup, and who are getting blood tests using blood screens for cancer. And very recently, there was an approval for a blood test looking for colon cancer.
And when this came out, the American Cancer Society actually put out a position that said, “We support this.” You have to be very careful with the language because what happened on the news that night was that they said, “Blood testing is now approved for screening for colon cancer.” That's not really what it says.
What it says is, if the patient refuses a colonoscopy, if the patient refuses stool testing, then you can default to blood testing. And remember, colonoscopies are designed around the biology of polyps, so you only need to do them every so often, okay, whereas stool tests and blood tests need to be done much more regularly. And I don't think that message gets out to people at all.
One-and-done is not really what you need for blood testing or stool testing. And so, it really sent a very mixed message that we in the colon cancer community were a bit uncomfortable with.
But it's gonna go even further because now all of those cancers that we're not screening for automatically—you could go in, in theory, and get tested to see whether you have bile duct cancer or pancreas cancer or some other cancer that's out there.
And so, we're clearly going to see an increasing trend of this. We're clearly seeing that those companies that do blood testing for cancer patients are gonna move to the earlier screening, uh, place. And maybe, just maybe, one day we'll do blood tests—oh, you have cancer—and we'll treat it without having to do scans and surgery. Fingers crossed that that's where it evolves.
But then, with all of this background, my favorite article of the month is in this journal called The Journal of Clinical Oncology. You remember that. This is a paper journal. And it caught my eye immediately 'cause it starts off with canine olfaction. Now, I don't know if you're a dog lover or not; if you're a cat person, maybe you have fish.
I'm a dog person, and this says, “Canine olfaction combined with Bayesian modeling,” my favorite kind of statistics, Bayesian modeling, “for multicancer detection from breath samples.” It's a phase 2 study done in India, where they don't have the resources to be doing blood testing on people. And this paper makes it into the Journal of Clinical Oncology because, guess what?
The dogs were able to do it. They got breath samples; they were trained first; they got breath samples; and, with pretty high statistical positivity and sensitivity, they were able to find patients who had early cancers and those who did not. And so, all I could think about is: how much money are we spending on going to get this blood test when, in fact, maybe, you know, your daughter's dog—that's my favorite grand-dog.
I don't have any grandchildren; I have a grand-dog. But maybe it's your daughter's dog that's gonna come up and go, “You know what? Maybe you should get checked because I'm a little worried about you.” But I just loved this paper, and I commend it to you.
But I do think we, as the medical community, need to be responsible overseers of this message: yes, this blood test is out there; yes, it is for patients who refuse any other test because some test is probably better than no test.
But you can't rely on a negative to say, “Oh, it was negative; I don't have cancer,” like you can with some of these other, more traditional tests. So, just keep a nose out for the latest news in precision medicine and its use for screening; maybe one day we won't do all these blood tests, and we'll just let our pups do the sniffing.
John Marshall, Oncology Unscripted.
Would You Bet on a Clinical Trial?
[00:06:33] John Marshall, MD: John Marshall, Oncology Unscripted. Lots of stuff going on out there in our business world today. Our favorite thing that our crackpot staff found that's just new is that you can actually bet on the outcome of a clinical trial.
Are you worried about all of the promotion that's going on with these gambling apps that they're pushing to our kids and to young people—“We'll give you a free $100 with your first bet”—and that sort of thing? And you know that, if you're successful at your phone app betting, what they do is they actually stop you.
They tell you, you can't go there anymore because all they really want is people who aren't very good at it on their betting apps. And now you can bet on the outcome of a clinical trial. And I was thinking to myself, but isn't this a little like, you know, stock investment? Oh, this company just opened up a phase 3 study in colon cancer.
I'm gonna bet on that by investing in their stock and then see what happens. But, as was pointed out by the crackpot staff that we have—crackerjack staff, not crackpot staff, crackerjack staff—you know, there are rules around stock investment and the like, whereas you could just bet on a clinical trial outcome. So, be cautious there. What works in phase 2 doesn't always work in phase 3, so it's a pretty risky gamble that's out there.
But what I really wanted to focus on that's new out there is data centers. One thing that our entire country seems to be in agreement on is that we don't want them near us.
We've been coached that they might, on the positive side, produce jobs, but on the negative side, their energy sucks, and they need a lot of water and all of this stuff. And here in Washington, of course, we get a lot of “co-promotion yes, co-promotion no,” about whether we should be doing it or not doing it. But I was thinking about how, here in Washington, there are a ton of data centers out by Dulles, and then I use AI.
Do you use AI? I don't know if we should confess that out loud, but I use it to look medical stuff up. My Google search engine sometimes gives me that AI option up there. A great website called Open Evidence is basically a big AI-generating, re-reporting system that I find very, very useful.
And we've talked on this program before about whether we trust it or not and how to interpret what we get from it. But I'm suffering a little bit lately from AI guilt because, in order to have AI, I need data centers. It's like, in order to have a car, I need gas, and so I got to go drill for gas somewhere.
Is it the same issue? Every time I use AI, am I using energy? Am I creating more of the problem of needing more and more data centers and, therefore, consuming more and more energy? Look, I have a plug-in hybrid, for goodness' sake. I don't even really want to use gas. So, every time I go on, I'm thinking to myself, “Am I doing something a bit irresponsible?”
But in conjunction with that, we have everything from the FDA approving a new blood test looking for Alzheimer disease. Well, how did we figure that out? We figured that out from databases that then were analyzed using machine-learning AI to come up with the ability to measure some of these targets and some of these tests.
The University of California in San Diego has, in fact, put AI into practice. They've created—and I'll read this to make sure and get it right—the new Institute for Applied Health Intelligence at UCSD, and it's basically designed to connect AI research to clinical practice. So, we're gonna lean into it more and more.
Singapore, the nation, has put in a whole bunch of money, 2.5 billion, into preventive precision medicine, which is a major AI data kind of utilization to try and figure out who's at risk and who's not, so that you can overall save money and be more effective. And the NCI just announced that it reached a major milestone in precision medicine, with three-quarters of a million participants now in the integrated genomic and electronic health record database.
So, everyone is sure that collecting these huge amounts of data is gonna turn into some sort of return on investment. We're living it day in and day out with our AI utilization in every field. Let's face it: even our church is using AI to try and help communicate with members of our congregation, and I'm sure, in your life, you're using it, big and small, in your day-to-day life.
But what we need to make sure of, I think, is that we're also checking in on this data center issue. Is this going to be an environmental issue? Is this really a good thing, or is it ultimately somehow going to be used in a bad way for us or have a negative impact on our overall world? So, I would be curious to see what you guys think about AI guilt.
Do you have it or not? Or you don't care—you're just leaning in?
And I wanna close with one positive that I'm very proud of: at ASCO this year, we launched a new initiative that we're calling Kleos, K-L-E-O-S. if you look up Kleos in Greek, it actually is: what do you think about those who came before? Sort of that hero thing. But I think about it as all those patients who I've taken care of over the years who didn't make it, who lost their battle to cancer, and how they've influenced me going forward. So, they're my heroes in many ways.
But this clinical trial we are now launching across colorectal cancer—yes, using intelligence and AI and databases, so we're gonna collect those, but also beginning to intervene in an individualized way in earlier lines of therapy with colorectal cancer—is sort of disrupting the current clinical research profile and pattern so that we can drop costs; speed, uh, the go/no-go signals that everybody is looking for; and reduce the cost, uh, of drug development so more people can take advantage of our innovation and all that we are learning.
And I'm very pleased to be, uh, helping to lead that, uh, throughout the United States, and I know many of you out there are participating and are aware of that. So, I wanna thank you for your ongoing support and energy. But it, too, is dependent on AI and the collection of data.
So, share with me: is AI good or bad?
Are you having a little AI guilt like I'm having every time I look something up and the machine whirls a little bit, and I'm thinking, “Am I using energy out there at Dulles?” I don't know the right answer, but I'm probably not gonna stop right now 'cause this is Oncology Unscripted, and I'm John Marshall.
Thanks for joining.
This transcript was generated by AI and lightly edited.