OHSU closed-loop crossover: insulin alone vs insulin plus pramlintide

A healthy pancreas releases two hormones at every meal from the same cells: insulin, and amylin. People with type 1 diabetes lose both. Every artificial pancreas on the market replaces only the first. The question this trial asks is narrow and fair: if you take one closed-loop controller and run it twice — once on insulin alone, once on insulin plus pramlintide, the lab-made copy of amylin — how much flatter is the meal?¹

What was tested. The same model-predictive-control algorithm, in two modes. A Dexcom G6 sent a glucose reading every five minutes to a smartphone running the controller; the controller drove one Omnipod for insulin and a second Omnipod for pramlintide, because the two drugs cannot share a reservoir. It also took activity data from a Polar M600 watch. Each participant did the study twice, in randomized order — one visit on insulin only, one on insulin plus pramlintide — and for the three days before the pramlintide visit they dosed pramlintide before meals to get used to it.¹

Who, and for how long. 33 adults with type 1 diabetes at OHSU in Portland, aged 18 to 70, sponsored by OHSU with the NIDDK. Recruitment ran from July 2024 and the study completed in January 2025.¹

This is an acute study, and that matters. Each "visit" was 12.5 hours in a clinic room, with breakfast and lunch provided. That is the correct design for the question being asked — you want to watch the post-meal curve under controlled conditions, with the same food, and iAUC over six hours is the cleanest way to see whether amylin blunted the spike. But it is not evidence about living with the system. It says nothing about three weeks of two pods, about nausea on a Tuesday afternoon at work, about overnight, or about whether people would stick with it. For that, the field has McGill's longer free-living trials — and those came back flat in June 2026.

What it measured. Two co-primary endpoints, both over the six hours after the first meal: the incremental area under the glucose curve (how much glucose piled up above where it started — the size of the spike) and percent time between 70 and 180 mg/dL (3.9-10.0 mmol/L). Secondary measures went after the honest costs and benefits of adding a second hormone: how much insulin and how much pramlintide the loop actually delivered, nausea scored on the Baxter Retching Faces scale, minutes of nausea reported, hypoglycemia episodes, and rescue carbohydrates eaten.¹

What came back. Results are posted on the registry, but no peer-reviewed publication of them has been located, so the headline figures summarized in the Registry primary outcomes section below should be read as registry-posted numbers, not peer-reviewed findings.¹ One detail is worth surfacing, because it is the investigators' own description of their data: so few participants had any sensor glucose below 70 mg/dL (3.9 mmol/L) that the low-glucose outcomes could not sensibly be reported as a percentage of time, and were instead reported as a count of how many people had any time below that threshold at all.¹ In an acute post-meal setting, that is reassuring about the immediate hypoglycemia risk of adding amylin — and it is a long way from proving anything about hypoglycemia over weeks.

Why it matters. Pramlintide may change post-meal glucose handling, but an acute clinic comparison cannot establish long-term safety, treatment burden or freedom from meal announcements. The registered outcomes should be interpreted within its short observation window.

Registry primary outcomes

The posted first-meal analysis included 31 participants per arm. Incremental glucose AUC favored insulin/pramlintide in the registered mixed model (adjusted difference −5.9 in (mg/dL·min)/1000, 95% CI −10.8 to −1.1; p=0.017). The other primary outcome, first-meal time in range, did not differ significantly (adjusted 6.3 percentage points, 95% CI −3.7 to 16.4; p=0.218). These controlled clinic outcomes do not establish sustained free-living benefit.¹