Insulin + pramlintide (amylin) closed loop

Most artificial-pancreas work uses one hormone — insulin — to push glucose down. This approach adds a second hormone, pramlintide, a lab-made copy of amylin, which the OHSU program page describes as produced in the pancreas when you eat or drink and lost in T1D when pancreatic cells are destroyed (same-cell co-secretion detail per review scope, not re-verified verbatim in the extracts read).² Pramlintide does not raise glucose the way glucagon does; instead it slows stomach emptying per the OHSU page (glucagon-surge/fullness details per review scope, not re-verified verbatim in the extracts read), so the meal-time spike is smaller and slower.² In a closed loop, that smoother rise is easier for the insulin algorithm to chase — the stated goal is steadier glucose with less reliance on perfect carb counting (editorial synthesis).³

What the trials show. A McGill randomized crossover trial found that a rapid-insulin-plus-pramlintide artificial pancreas raised time-in-range from 74% to 84% versus insulin alone per the trial abstract (full-report detail not re-verified in this review), driven mostly by better daytime control (63% to 78% per abstract), without any increase in hypoglycemia per abstract event counts.³ A follow-on McGill trial showed the pramlintide system could replace carb counting with a simple meal announcement while keeping glucose control as good as carb-counting alone per the trial abstracts (full-report detail not re-verified).⁴⁵ A fully-closed-loop version with no meal input held a high time-in-range and produced fewer low-glucose events than the insulin-only loop per the abstract (one third vs over half of participants had a low).⁶ The recurring cost is gut side effects — mild-to-moderate nausea or bloating in a meaningful minority per trial abstracts (full safety tables not re-verified).³⁴

The 2026 test did not establish noninferiority. The published ADA conference abstract reports 26 completers across three 21-day periods: time in range was 71.1% with pramlintide 8 µg/U, 71.2% with 10 µg/U and 72.1% with carbohydrate-counting hybrid control. Neither comparison met the prespecified 5-percentage-point margin (p=0.0787 and p=0.0584). Insulin doses were lower, but gastrointestinal events occurred only with pramlintide. No severe hypoglycemia or ketoacidosis occurred. These are conference-abstract results; a full report is needed to assess missing data, withdrawals and subgroup claims.¹

Glucagon vs amylin. This is the key contrast with glucagon-based dual-hormone systems. Glucagon actively raises glucose, giving the loop a brake against lows, but it is described as chemically unstable in a pump in unopened sources (not re-verified in the extracts read). Pramlintide doesn't defend against lows directly — it makes the whole post-meal curve gentler, so the loop has less to fight.²

The formulation challenge. Pramlintide and insulin cannot share a reservoir per the Symlin do-not-mix instruction: pramlintide is formulated at an acidic pH, insulin near neutral, and the two aggregate and lose activity when mixed (pH/aggregation chemistry per unopened full texts, not re-verified verbatim), which is why Symlin must be a separate injection.⁷⁷ Today's research systems therefore use two pumps or two pods — one for each hormone (setup detail per unopened sources, not re-verified verbatim).³ A major engineering effort aims to fix this: Stanford researchers built a stable insulin-pramlintide co-formulation that resisted aggregation for over 16 hours per the paper (stability figure not re-verified verbatim in the extracts read),⁷ and the company Adocia has tested a fixed co-formulation (ADO09, "M1Pram") in people, improving post-meal glucose and time-in-range with less mealtime insulin per the trial title/abstract.⁸ A single stable mixture is what would let this run from one pump.

That commercial path has now stalled. On 3 July 2026 Adocia announced that Sanofi's three-year exclusive right to negotiate a partnership on M1Pram — an option Sanofi paid for in 2023 but never exercised — had expired, and that Adocia, keeping global rights, has put M1Pram development on hold given its current priorities, while saying it will keep evaluating the path forward.⁹ Adocia pointed to a landscape reshaped by the rise of GLP-1 therapies per the release (no formulation-failure statement in the release excerpts read). On hold is not the same as abandoned — but Adocia retains the program but has not announced a restart in the reviewed update.

Maturity and access. The drug itself is mature — pramlintide was FDA-approved as Symlin in 2005 as a subcutaneous adjunct to mealtime insulin in T1D, and carries a boxed warning for severe insulin-induced hypoglycemia.¹⁰ But the automated closed-loop co-delivery is investigational: there is no approved insulin-pramlintide artificial pancreas, and it is reachable today only through studies (for example OHSU's "iPancreas" program, testing insulin-plus-pramlintide on an Omnipod platform with a Dexcom sensor).¹¹

What remains uncertain. M1Pram’s commercial development remains on hold per Adocia. A separate McGill tirzepatide trial does not establish that investigators have abandoned amylin (analysis; trial detail per unopened sources). Registry listings and older recruitment pages do not establish current site availability. Pramlintide’s historical approval also does not establish current Symlin supply (supply detail per unopened sources); see the standalone drug record.