Structured population data were available for 4 of 4 supporting studies. Diabetes type was heterogeneous: 2 studies with type 2 diabetes, 1 studies with type 1 diabetes. 3 studies described adult participants.

1 relevant study has unresolved authoritative design evidence and does not contribute to design-specific or RoB-specific claims.

Upstream consistency scoring indicates mixed findings across the evidence.

Evidence uses heterogeneous comparison contexts: active_comparator, usual_care.

1 supporting study has high, serious, critical, unclear, or unavailable risk-of-bias information.

4 of 4 studies with duration data have follow-up of 12 weeks or less.

1 study contains conflicting p-value and significance fields; those significance claims are suppressed.

Not extracted from the authoritative source.

MDI therapy

Clinician-reference record is usable with limitations in: sample_size.

Patients treated with the Android‐HCL therapy for type 1 diabetes.

HCL therapy

The intraday glucose fluctuation (especially after lunch, before sleep, and during sleep) of the Android‐HCL system was significantly superior to that of the sequential multi‐injection insulin therapy.

Patients treated with multiple injections of insulin.

Compared with MDI, HCL had a higher overall TIR ( p < 0.05), lower TAR, TBR, and average blood glucose.

Authoritative quality diagnostic: p value significance conflict.

Artificial pancreas may improve glucose control in type 1 diabetes

Source evidence could not be resolved sufficiently for this field.

Conventional insulin pump therapy

Clinician-reference record is usable with limitations in: design, risk_of_bias.

Participants received treatment with a dual-hormone artificial pancreas system.

Dual-hormone artificial pancreas

The time In target, time between 4.0–10.0 mmol/L, time above 8.0 mmol/L, time above 10.0 mmol/L, and plasma glucose concentrations are all nearly identical in both studies for the three arms.

Participants received treatment with a single-hormone artificial pancreas system.

Single-hormone artificial pancreas

Participants received standard care with a conventional insulin pump.

Authoritative quality diagnostic: source evidence unresolved.

Cloud platform improves algorithm testing for type 1 diabetes

Core clinician-reference fields are available or not applicable.

Participants received the open-source hybrid closed-loop AID system.

Closed-loop group

Participants were randomly assigned to the open-source hybrid closed-loop automated insulin delivery (AID) group (closed-loop group) or the control group.

Participants received conventional insulin pump therapy.

In the control group, participants received conventional insulin pump therapy with the same devices.

Authoritative quality diagnostic: invalid or missing unit.

Authoritative quality diagnostic: invalid or missing value.

Authoritative quality diagnostic: sample size arm total mismatch.

Open-source AID improves glycemic control in type 2 diabetes during surgery

Insulin dose adjustment without vildagliptin

Patients receiving vildagliptin treatment demonstrated significant reductions in HbA1c and GA levels.

Vildagliptin Group

Patients receiving vildagliptin treatment demonstrated significant reductions in HbA1c and GA levels (P<0.001 and P=0.009, respectively).

Patients who only had their insulin doses adjusted without adding vildagliptin.

The control group consisted of patients who only had their insulin doses adjusted without adding vildagliptin.

Insulin therapy × Time in range

Available supporting studies

Artificial pancreas may improve glucose control in type 1 diabetes

Cloud platform improves algorithm testing for type 1 diabetes

Open-source AID improves glycemic control in type 2 diabetes during surgery

Vildagliptin improves glycemic control in type 2 diabetes patients on insulin