twiist automated insulin delivery in type 2 diabetes: 13-week single-arm trial

The study

This trial tested the twiist automated insulin delivery (AID) system in adults with insulin-treated type 2 diabetes. It was a single-arm, multicentre trial. All 307 participants used the system, and none were assigned to a comparison group. They were aged 19-85 years. Baseline continuous glucose monitoring (CGM) data were collected first. Then participants used twiist for 13 weeks.

The primary outcome was the change in HbA1c from baseline to 13 weeks. The authors also looked at CGM measures, safety events, user satisfaction and sleep quality.

What they found

Mean HbA1c fell by 0.7% (95% CI -0.9 to -0.6). It went from 8.1% ± 1.5% at baseline to 7.4% ± 0.9% after 13 weeks. The authors report P < 0.001 for both non-inferiority and superiority. The abstract does not report the margin or the comparator behind these tests.

The effect was greatest in participants with baseline HbA1c of 9.0% or higher. A substantial fall in HbA1c was also seen when AID was added to a GLP-1 receptor agonist or an SGLT2 inhibitor that participants had used before the study and continued during it.

CGM time in range (70-180 mg/dL) rose from 57% ± 26% to 73% ± 17%. The mean change was a rise of 16% (95% CI 14-18; P < 0.001). The authors say all other multiplicity-controlled CGM outcomes reflecting hyperglycaemia also improved.

CGM-measured hypoglycaemia was low at baseline and stayed low during follow-up. Diabetic ketoacidosis occurred in one participant. No severe hypoglycaemia events occurred. Participants were more satisfied with twiist than with their earlier insulin delivery method (P < 0.001), and sleep quality improved (P < 0.001).

How much weight to give it

The authors conclude that AID with twiist was safe and associated with substantially improved glycaemia. That is a fair reading of a before-and-after comparison, but the design limits what can be said.

• There was no control group. Without one, changes in HbA1c and time in range cannot be attributed to twiist alone, and the abstract does not say how other influences were handled.
• Follow-up was 13 weeks. Longer-term control, safety and persistence of use are unknown.
• HbA1c and CGM measures are surrogate outcomes. The abstract reports no outcomes such as complications, hospital admissions or mortality.
• Diabetic ketoacidosis occurred in one of 307 participants; the full paper should be checked for how it was assessed. No severe hypoglycaemia occurred.
• The abstract does not report funding, dropouts, details of participants' prestudy insulin regimens, or cost.

This is a single-arm pivotal trial. It does not show that the system caused the improvements. It does not show that the system caused the improvements.

For your clinic

These results suggest that AID may be worth considering in future for some adults with insulin-treated type 2 diabetes, especially those with high HbA1c. HbA1c also fell substantially in participants already taking a GLP-1 receptor agonist or SGLT2 inhibitor. This does not support changing practice on its own.

Before acting on the findings, read the full paper. Check who was enrolled, how many completed the trial, what training and support were given, and how ketoacidosis was assessed. Local guidelines and the practical issues of access, cost and follow-up support also matter. The abstract gives no information on these.

Whether any patient is suitable for AID remains an individual clinical decision.