Background: Early-life glycaemic control disturbances can negatively affect brain development and plasticity. This study aimed to assess the efficacy and safety of automated insulin delivery (AID) with the MiniMed 780G system in children with type 1 diabetes aged 2–6 years requiring at least 6 units of insulin a day.

Methods: In this open-label, randomised crossover trial, we enrolled children from 12 hospitals in Finland, Italy, Slovenia, and the UK. Participants underwent a 2-week run-in phase in which the MiniMed 780G system was used in manual mode with the suspend before low feature enabled (manual+SBL), followed by a 26-week study phase. For the study phase, participants were randomly assigned to a sequence comprising a 12-week auto mode, 2-week washout, and 12-week manual+SBL mode (AM-MM sequence), or the reverse order sequence (MM-AM). The primary endpoint was the between-treatment difference for auto mode versus manual+SBL mode in the percentage of time in range (TIR, 70–180 mg/dL), expressed as the least-squares mean difference adjusted for sequence effect and run-in TIR, and assessed for non-inferiority (margin 7·5 percentage points). Secondary endpoints were the adjusted between-treatment difference in mean HbA1c at the end of each 12-week period assessed for non-inferiority (margin 0·4 percentage points), and the adjusted between-treatment difference in mean TIR and HbA1c assessed for superiority. The primary and secondary analyses were by intention to treat. Safety was assessed in all participants who signed informed consent. The trial was registered with ClinicalTrials.gov, NCT05574062, and is completed.

Findings: Recruitment took place from March 24 to Sept 21, 2023. 98 participants began the study phase (AM-MM: n=50, with two withdrawals during the study phase; MM-AM: n=48). Overall mean age was 4·7 years (SD 1·2); 48 (49%) of 98 participants were female and 50 (51%) were male. Mean TIR was 58·1% (SD 14·3) in the run-in phase, 68·3% (6·9) in auto mode, and 58·3% (12·5) in manual+SBL mode (adjusted between-treatment difference 9·9 percentage points [95% CI 8·0 to 11·7]; non-inferiority met for the primary endpoint with superiority for auto mode). Mean HbA1c was 7·53% (SD 0·96; 58·8 mmol/mol [SD 10·5]) in the run-in phase, 7·00% (0·53; 53·0 mmol/mol [5·8]) in auto mode, and 7·61% (0·91; 59·7 mmol/mol [9·9]) in manual+SBL mode (between-treatment difference: –0·61 percentage points [95% CI –0·76 to –0·46; non-inferiority met with superiority for auto mode). There were nine serious adverse events (five during auto mode, two during manual+SBL mode, one during run-in, and one during washout; all deemed to be unrelated to the study device or procedure), including one serious adverse event of diabetic ketoacidosis during auto mode. No severe hypoglycaemia events were reported.

Interpretation: In children aged 2–6 years with type 1 diabetes, the TIR with the MiniMed 780G system in auto mode was non-inferior to that in manual+SBL mode. The significantly improved TIR and HbA1c in favour of auto mode could help prevent diabetes-related complications. Additionally, the observed safety profile in auto mode was acceptable.

Funding: Medtronic.