Can an App Dose Insulin Like Your Endocrinologist? New Research Says: Pretty Much

If you've ever wished your endocrinologist could ride along in your pocket — checking your numbers, nudging your doses, catching the patterns you'd miss — researchers at Oregon Health & Science University just moved that future a lot closer.
Their team built an AI tool that pairs with a smartphone app to guide glucose adjustments for people with type 1 diabetes. And in a pilot study, the results were striking: over four weeks with 16 patients, expert endocrinologists judged the AI's recommendations safe more than 99% of the time. Even better, the AI's guidance lined up with what the experts themselves would have recommended nearly 68% of the time.
Here's the part that caught our attention: the team didn't train the AI on thousands of patient records. Lead study author Nichole Tyler, an MD-PhD candidate, explains they built it using only a mathematical simulator. "[Y]et, when the algorithm was validated on real-world data from people with type 1 diabetes at OHSU, it generated recommendations that were highly similar to recommendations from endocrinologists," Tyler notes (via HealthExec).
That's a big deal, because as senior author Peter Jacobs, PhD, points out, most similar efforts never make it this far: "Very few have shown a statistically relevant outcome," Jacobs says, "and most do not compare algorithm recommendations with those of a physician." This one did both — and it's the first of its kind to be tested clinically.
And it's not an isolated result. In a six-month randomized trial presented recently, AI-generated insulin titration produced glycemic outcomes comparable to those of experienced endocrinologists — with clinicians accepting the AI's recommendations up to 97% of the time. A separate JAMA-published study found an AI insulin-dosing system noninferior to senior physicians for hospitalized patients with type 2 diabetes.
Let's be clear about what this doesn't mean: nobody's suggesting your endocrinologist is headed for the unemployment line. What it does mean is genuinely exciting — especially for the millions of people who can't easily get to a specialist. An AI that thinks like an endocrinologist could mean better guidance between appointments, faster dose adjustments, and expert-level support for people in rural or underserved areas.
The OHSU team's full findings are published in Nature Metabolism, and they're free to read. We'll be watching where this goes next — because a future where everyone with diabetes has an expert in their pocket is a future worth rooting for.
Sources:
HealthExec, "AI helps diabetes patients manage glucose levels" (quotes from Nichole Tyler and Peter Jacobs via OHSU news coverage cited therein) — https://healthexec.com/topics/precision-medicine/ai-helps-diabetes-patients-manage-glucose-levels
AJMC, "AI Reshaping Diabetes Care From Nutrition to Clinical Decisions" — https://www.ajmc.com/view/ai-reshaping-diabetes-care-from-nutrition-to-clinical-decisions
Endocrinology Advisor, "AI-Assisted Insulin Titration System Noninferior to Physicians" (JAMA study) — https://www.endocrinologyadvisor.com/news/ai-assisted-insulin-titration-system-noninferior-to-physicians/



Comments