Ask AI
Back Back
Can AI Help Personalize Cancer Treatment Decisions?
Activity Information Ask a Question

Activity

Progress
1
Course Completed
Activity Information

Released: September 29, 2026

John Marshall
John Marshall, MD
Catch Up On All Episodes of Oncology Unscripted

Watch Now

Main Topic: Can AI Help Personalize Cancer Treatment Decisions?

John Marshall, MD: Oncology Unscripted. Welcome back to school. It is September here in Washington. It's eerily very pretty, just like it was 25 years ago when September 11 was about to happen, and I am a little nervous about the world right now in general. My guess is some of you out there are, too, for a variety of reasons.

So, let's look at some of the science that is hopefully going to make our world a better place. And one of the things that some patient of mine today said they were more afraid of is not thermonuclear war, but AI.

And so, the first paper that I wanted to highlight was AI being able to tell us, should I give gemcitabine and Abraxane or should I give FOLFIRINOX to patients with metastatic colon cancer? And this paper came out just recently, and using an AI model, they were able to demonstrate at least favoring one choice over the other instead of what if you do what I do often, is flip a coin before I go in the room and decide which regimen I'm going to give to an individual patient.

So, this paper is probably worth looking at because here AI is making things maybe a better thing.

Now, let's look at the other end of the drug development spectrum, and this was where we have a hot new drug, adagrasib, that was given early approval in combination with cetuximab in G12C RAS, big-deal RAS, in G12C-mutated colorectal cancers.

And if you're like me, you've been using this combination in those few G12C patients. You've been eagerly hoping to get other RAS inhibitors for your other colon cancer patients. But yes, the FDA gave them approval based on a small phase II experience, but then when they did the randomized clinical trial, they actually showed that it didn't prove useful.

And the FDA withdrew that approval for G12C adagrasib plus cetuximab in colorectal cancer. A much unexpected result there. So, where we thought we had a great lead, it turned out in the randomized clinical trial not to show the kind of benefit that we were hoping for.

In a similar vein, we've taken colorectal cancer into left-sided versus right-sided, RAS wild-type versus RAS-mutated, and we've been basically seeing data that suggested that if I pick left-sided, all wild-type patients, the EGFR drugs seem to behave better in that clinical subtype.

Well, also in a current edition of the JCO, when they took that subpopulation and used 4 chemos, FOLFIRINOX and cetuximab versus bevacizumab, what they were unable to show is a clear home-run win for that subpopulation. And so, again, these biases that we bring, including me, to the table can sometimes be beat back, even though a positive study, not to the degree that was hoped for when that study was done.

And so, if we go on to precision medicine, there's a new study, smaller study, looking at using genetic testing and tumor board kind of review to pick which drugs to use, how to treat cancers of unknown primary. And basically, on one side, there are fewer cancers of unknown primary anyway because of precision medicine and are just getting better at figuring this out.

But this study, in a 3-to-1 randomization, you either got platinum-based therapy, which is sort of what the book says to do of cancer of unknown primary, versus more tailored treatment based on precision medicine. And this study was positive in that those that got the precision medicine treatment actually had an improved outcome by a few months, not through the roof, but by a few months, so that being worth it.

So, in just this past week, we've seen papers that have shown us the way using AI, using precision medicine, but also the confirmation studies that were designed to try and really demonstrate this is where we are, this is the new standard of care, pulling us back off of that knowledge ledge a bit. So, still more work to do.

Stay tuned here for Oncology Unscripted updates as new data comes forward. John Marshall, thanks for joining.

This transcript has been generated by AI and lightly edited.