LY3938577: glucose-sensitive insulin in type 1 diabetes
Eli Lilly's glucose-sensitive insulin candidate in healthy participants and in people with type 1 diabetes. The design is the most revealing detail: the T1D arms retain mealtime insulin lispro alongside LY3938577, which implies Lilly is developing a glucose-responsive *basal* insulin — a chemical safety floor against lows — rather than a single insulin that replaces both basal and bolus. Recruiting as of August 2026.
Primary endpoints
- Part A: number of participants with one or more adverse events and serious adverse events
- Parts B and D: number of participants with one or more adverse events and serious adverse events
- Part A: number of participants with clinically significant changes in vital signs
Results so far
None posted. Started 15 May 2024, still recruiting as of August 2026, with an estimated completion of September 2026 and a target enrolment of 118. The primary endpoints are safety, not glucose response, so even a full readout would not settle whether the molecule senses glucose usefully in people.
The full picture
Lilly's entry into human testing of a glucose-sensitive insulin, and the study whose design tells you more than its (unpublished) results will.
Read the comparator arms. The trial includes insulin degludec — a basal insulin — and insulin lispro, a mealtime insulin, in the type 1 diabetes parts. Retaining prandial lispro alongside the investigational molecule means LY3938577 is not being asked to cover meals. The implied product is a glucose-responsive basal insulin: something that reduces its own action as glucose falls overnight and between meals, giving a chemical floor under hypoglycemia, while the user keeps bolusing normally.
That is a much less ambitious product than "smart insulin" usually implies in public coverage — and a much more achievable one. It is not a total-replacement insulin, which would have to cover fasting, meals, activity and day-to-day sensitivity changes with no separate bolus.
What the endpoints will and will not answer. The primary endpoints are adverse events and vital signs — this is a safety and tolerability study. A clean readout would establish that the molecule can be given to people with T1D without harm. It would not establish that its potency actually tracks glucose, which is the claim the whole class rests on and the one MK-2640 failed.