Background: Time in range (TIR) goals are rarely met in children with type 1 diabetes, except at the cost of increased hypoglycaemia episodes. Our objective was to evaluate the safety and efficiency of the Diabeloop DBL4K (Diabeloop, Grenoble, France) hybrid closed-loop system in prepubescent children. Methods: We did a multicentre, open-label, randomised, controlled, non-inferiority, two-session crossover study in the paediatric endocrinology departments of three university hospitals in France and Belgium. Eligible participants were aged 6-12 years with type 1 diabetes for at least 1 year, glycated haemoglobin A1C 9% (75 mmol/mol) or less, and insulin pump treatment for at least 3 months. Participants were randomly assigned (1:1) to a closed-loop device or sensor-augmented pump (open loop) therapy. Randomisation was by a permuted block randomisation scheme, using an interactive web-based response system, and was stratified on centre (block size 6). The assessed closed-loop device, the Diabeloop for Kids DBL4K hybrid closed-loop system, is an automated blood glucose regulation system composed of a handset, insulin pump, and continuous glucose monitor. The open-loop system is defined as a sensor-augmented pump therapy composed of the usual insulin pump used by the patient and a continuous glucose monitor. A 72-h in-patient period was followed by a 6-week home phase. After a 1-week washout period, the participants crossed over to the other device. The primary outcome, assessed in the intention-to-treat population, was the mean proportion of time spent in hypoglycaemia (3·9 mmol/L [<70 mg/dL]) during the hospital phase, with a non-inferiority margin of -2·5% (absolute value). Safety was assessed in the intention-to-treat population on a per-protocol basis. This study was registered with ClinicalTrials.gov, NCT03671915. Findings: Between May 6 and Dec 23, 2019, we included 21 participants (closed loop then open loop, n=10; open loop then closed loop, n=11). The proportion of time spent in hypoglycaemia was significantly lower with the closed-loop system than the open-loop system in both groups (2·04% [95% CI 0·44 to 3·64] vs 7·06% [5·46 to 8·66]; non-inferiority one-sided p<0·0001). No severe ketoacidosis, nor severe hyoglycaemic events or fatal adverse events occurred. All 25 adverse events (18 with the closed-loop system, seven with the open-loop system) were related to the treatment. Interpretation: The closed-loop Diabeloop system decreased hypoglycaemic episodes and provided good metabolic control in prepubescent children with type 1 diabetes, under real-life conditions. This finding supports the safe use of closed-loop technology in this paediatric population. Funding: Diabeloop. Translation: For the French translation of the abstract see Supplementary Materials section.