Roupen Odabashian, Oncologist at Abbotsford Regional Hospital and Cancer Centre, shared on LinkedIn:
“A new AI system called TrialMatchAI just published its numbers in Nature Communications: for 92% of oncology patients, a relevant clinical trial showed up somewhere in its top 20 recommendations, with expert review confirming over 90% accuracy at the level of individual eligibility criteria.
On paper, that’s the trial matching problem basically solved.
But there’s an earlier study worth remembering next to it. A randomized trial of over 20,000 patients with genomically sequenced solid tumors tested what happens when oncologists actually get notified about AI matched trials for their eligible patients.
Enrollment did not go up.
Being told the match existed didn’t change what happened next.
I’ve been recruiting patients to clinical trials for years, and I’m now getting my hands dirty as a PI on one. Identifying patients was never the hard part. Everything that comes after is.
Even once you find someone eligible, you can’t just enroll them. They need to progress on a specific treatment first. Then you have to sit with them and explain what a clinical trial actually means: this is an experimental drug, we don’t know if it works, but we believe from prior literature it might work better than the standard, and I might be wrong.
If they enroll, the barriers nobody talks about show up next. The geography. The transportation. The intensive monitoring a trial requires, stacked on top of appointments they’re already struggling to make. I have patients who’ve lost their jobs from the time commitment alone, and that financial toxicity compounds with everything else a cancer diagnosis already costs them.
Matching people to trials was never the bottleneck. The bottleneck is the design of the trial itself and how hard it is to actually access, especially when what we’re asking someone to do is take an experimental drug that might or might not work, in the hope that something clicks and a few more lives get saved.
If a 92% match rate doesn’t move enrollment, what would?”
Title: TrialMatchAI: an end-to-end AI-powered clinical trial recommendation system to streamline patient-to-trial matching
Authors: Majd Abdallah, Sigve Nakken, Mikaël Georges, Mariska Bierkens, Johanna Galvis, Alexis Groppi, Slim Karkar, Lana Meiqari, Maria Alexandra Rujano, Steve Canham, Rodrigo Dienstmann, Remond Fijneman, Eivind Hovig, Gerrit Meijer, Macha Nikolski
Other articles featuring Roupen Odabashian on OncoDaily.