Tirzepatide in Type 1 Diabetes: The Phase 2 Trial Results

Tirzepatide beat placebo for weight loss in the first published placebo-controlled trial in type 1 diabetes: adults lost 8.7 kg more than the placebo group over 12 weeks (95% CI -12.0 to -5.5), and their total daily insulin dose fell 35.1% relative to placebo^[1]. That result comes from 24 people on a dose capped at 5 mg, so it establishes a strong short-term signal and little else. The glucose findings were less clear: the HbA1c difference of -0.4% was borderline (P = 0.05), and time in range did not differ significantly (P = 0.17)^[1].

What the trial tested

Snaith and colleagues in Sydney ran a 12-week, phase 2, double-blind trial in adults with type 1 diabetes and a BMI above 30^[1]. Twelve people received tirzepatide (2.5 mg for 4 weeks, then 5.0 mg for 8 weeks) and twelve received placebo. Change in body weight was the primary end point. Overall, participants had a mean age of 41, had lived with diabetes for 23 years, weighed 101 kg and started at an HbA1c of 7.3% on 69 units of insulin a day; 54% used insulin pumps, all with automated delivery^[1].

Two features limit how far the numbers travel. The dose is one-third of the 15 mg maximum on the US labels^[13][14], so the trial says nothing about higher doses. And the injections were given weekly in a research unit, which the authors acknowledge does not reflect a home setting, while blinded investigators supported insulin adjustment^[1]. Funding came from Breakthrough T1D, an Australian government research grant, a hospital foundation and philanthropic sources, and the authors reported no relevant conflicts^[1].

What it found

The weight difference is adjusted for baseline, which is why it does not equal the simple gap between the two arms^[1]. That adjustment matters because randomization in a group this small left visible imbalances: the tirzepatide arm started heavier (104.8 vs 97.7 kg), had shorter diabetes duration (17.1 vs 29.4 years) and more often had detectable C-peptide (54.5% vs 25.0%)^[1]. The authors report that adjusting for C-peptide status, age, sex and pump use did not alter the weight result, though they did not show those data^[1].

Half of the tirzepatide group reported nausea, mostly mild and clustered around dose starts and increases^[1]. Two people left the trial, one from each arm. The tirzepatide withdrawal involved increased anxiety in someone with a prior history. A depression diagnosis also occurred in a tirzepatide participant with preexisting symptoms, and the authors thought both were exacerbated by the time demands of weekly visits^[1].

Why the insulin drop deserves the most attention

In type 1 diabetes, a falling insulin dose cuts both ways. Less insulin may help weight, but insulin is also the protection against ketoacidosis. In an exploratory look at 14 pump users, doses started falling as early as week 2, and by week 6 the between-group difference was -42% of the baseline dose (95% CI -59 to -18%)^[1]. The authors say the fall in pump users was primarily driven by the glucose response and the pump algorithm^[1]. An exploratory subgroup analysis also found similar 12-week reductions on pumps (-31%, 95% CI -48 to -8%) and on multiple daily injections (-34%, 95% CI -54 to -6%), with blinded investigators supporting dose adjustments^[1].

The clean safety record was earned under close supervision. Participants had their ketone and hypoglycemia plans reviewed before the first dose, had glucagon access confirmed, and got weekly phone calls after the first two doses^[1]. People with ketoacidosis or severe hypoglycemia in the prior 3 months, or with gastroparesis or pancreatitis, were excluded^[1]. With 24 participants over 12 weeks, zero events cannot rule out a real risk. A semaglutide crossover trial with automated insulin delivery also had no ketoacidosis, yet recorded two episodes of recurrent euglycemic ketosis without acidosis^[10], the kind of early warning that a small trial can only just catch.

How it compares with the semaglutide trials

No trial has compared the two drugs directly in type 1 diabetes, and these studies differ in length, dose and insulin delivery. Side by side, they show similar placebo-adjusted weight differences reached at different time points, with semaglutide having the firmer glucose evidence so far^[1][9][10].

A nonsignificant result here does not show that tirzepatide lacks a time-in-range effect. The point estimate of 8.4 points is close to the ADJUST-T1D figure, but the confidence interval runs from -4.0 to 20.9, and Snaith’s authors say their trial was likely underpowered for glycemic outcomes^[1]. It also enrolled people across pumps and injections rather than only automated delivery users, the setting most relevant to the GLP-1 and insulin pump question. ADJUST-T1D recorded two severe hypoglycemia events in each arm and no ketoacidosis^[9]. Its post hoc analysis found a 22.6% fall in total daily insulin, mainly from bolus doses, with early effects largely independent of weight loss^[11], a pattern that resembles the rapid tirzepatide response.

What the observational cohorts add, and where they fall short

Retrospective studies reach further in time and dose than the trial can, but they compare people whose clinicians chose to prescribe tirzepatide with people whose clinicians did not. They show association only.

• Sheffield, UK. 142 adults against 50 matched controls over a median of 365 days: mean reductions versus controls of 13.4 kg in weight, 0.6% in HbA1c and 29.9 units in daily insulin, with 89.4% still on the drug at 1 year and no between-group difference in severe hypoglycemia, ketoacidosis or hospitalization^[4].
• Colorado. 61 adults against 54 matched controls: time in range was 7.4% higher at 12 months, and 50.8% vs 25.9% met composite CGM targets, with time below range similar and no severe hypoglycemia or ketoacidosis in either group^[6].
• Sydney. 23 propensity-matched pairs over about 28 to 31 weeks: weight change of -10.01% vs +0.69%, and insulin change of -21.82 vs +5.62 units a day, while other CGM, blood pressure and lab outcomes did not differ^[5].
• Pooled. A meta-analysis of six observational studies (248 adults) estimated -0.61% HbA1c, -9.9 kg and -23.73 units a day at 6 months, while warning that patient cohorts overlap^[3]. A systematic review covering one RCT and seven observational studies rated certainty as low or very low^[2].

Two safety threads remain unresolved. In a survey of 230 adults at one Colorado center, symptomatic hypoglycemia was reported more often on tirzepatide than semaglutide (29% vs 13.4%, P < 0.001), while gastrointestinal events did not differ^[7]. That self-reported finding sits awkwardly beside CGM data from overlapping authors showing no excess time below range against non-users^[6]. And in a chart review of eye outcomes, new retinopathy developed in 25.6% of tirzepatide users and 29.2% of controls, but it was more frequent across both groups in people whose HbA1c fell quickly (39.3% vs 17.1%, P = 0.048)^[8]. The numbers there are small, with 63 people free of eye disease at baseline.

Where this leaves tirzepatide in type 1 diabetes

The August 2026 Mounjaro label covers glycemic control and cardiovascular risk reduction in type 2 diabetes^[13], and the Zepbound label covers obesity and obstructive sleep apnea^[14]. Neither mentions type 1 diabetes, and researchers describe adjunctive use in this group as off-label^[7]. The Zepbound label does tell prescribers to “consider reducing the dose of concomitantly administered insulin” when starting^[14], but the Snaith insulin data show how large and fast that change can be in type 1 diabetes^[1].

Larger trials are under way, and the two phase 3 studies list primary completion for November 2026^[15][16]. Sponsored by Eli Lilly, SURPASS-T1D-1 is listed as a phase 3 placebo-controlled study with four tirzepatide dose arms, an estimated 905 participants, and HbA1c change at week 40 as its primary outcome^[15]. SURPASS-T1D-2 is a long-term phase 3 study with an estimated 465 participants and about 20 months of participation^[16]. A single-center trial in Kuwait plans to randomize 60 patients to usual care with or without the maximum tolerated dose^[12]. Those results, more than the current data, will decide whether the HbA1c signal is real and how often ketosis and hypoglycemia occur at scale.

Questions to raise with a prescriber

• How will my basal and bolus doses change in the first weeks, and who reviews them? In the trial, the fall in insulin dose appeared within 2 weeks in pump users^[1].
• What is my plan for ketone testing and sick days, particularly during nausea or reduced eating? Trial participants reviewed ketone risk before starting^[1].
• Should I have a retinal exam before starting? One chart review linked rapid HbA1c decline to new retinopathy^[8], the Mounjaro label says patients with a history of diabetic retinopathy should be monitored^[13], and the trial excluded people with proliferative retinopathy or macular edema^[1].
• Is the goal weight, glucose or insulin dose? The randomized evidence is much stronger for weight than for glucose control^[1][2].