Automated insulin delivery lowered HbA1c in insulin-treated type 2 diabetes

Design

Single-arm, multicentre pivotal trial, 13 weeks

Population

307 adults aged 19-85 with insulin-treated type 2 diabetes

Primary outcome

Change in HbA1c from baseline to 13 weeks

Effect

-0.7 percentage points (95% CI -0.9 to -0.6); time in range 57% to 73%

The twiist-T2D pivotal study enrolled 307 adults aged 19 to 85 with insulin-treated type 2 diabetes. After a baseline period on CGM, all switched to an automated insulin delivery (AID) system for 13 weeks. There was no comparison group. Published in Diabetes Care on 7 October 2026.

HbA1c fell by 0.7 percentage points, from 8.1% to 7.4%. Time in range (70 to 180 mg/dL) rose from 57% to 73%. The largest falls were in people starting at 9% or above, and benefit was seen even in those already taking a GLP-1 receptor agonist or SGLT2 inhibitor. CGM-measured hypoglycaemia stayed low, there was no severe hypoglycaemia, and one participant had diabetic ketoacidosis.

AID has become standard for type 1 diabetes; this adds to a growing body of evidence in type 2. The design limits what can be concluded: a before-and-after study cannot separate the device from the extra attention and closer follow-up of a trial.

In practice, the candidates are people on multiple daily injections whose HbA1c stays high despite adherence and add-on agents. Cost and access will decide uptake in most Indian clinics; the regulatory position of this device in India is not known.

• Consider AID for adults on intensive insulin whose HbA1c stays at 9% or above despite good adherence
• Keep GLP-1 receptor agonists and SGLT2 inhibitors going when insulin delivery is automated
• Teach ketone checks on pump failure or infusion-set problems, since ketoacidosis can still occur
• Use time in range as well as HbA1c to judge response at follow-up
• Expect the trial setting to flatter results compared with routine clinic use

Why it matters

Automated delivery is no longer only a type 1 conversation; the poorly controlled insulin user with type 2 diabetes is now a candidate.

Don't overread it

Single-arm and 13 weeks long: it shows improvement on the system, not superiority over optimised injections.

The statistics, in plain English

The 95% confidence interval for the HbA1c fall (-0.9 to -0.6 points) is narrow and well away from zero, so the improvement is real within the trial. Without a control group, though, some of it may reflect regression to the mean and the extra support every trial participant receives.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.