Background: In mouse models of diabetes, prophylactic administration of insulin reduced incidence of the disease. We investigated whether administration of nasal insulin decreased the incidence of type 1 diabetes, in children with HLA genotypes and autoantibodies increasing the risk of the disease.
Methods: At three university hospitals in Turku, Oulu, and Tampere (Finland), we analysed cord blood samples of 116 720 consecutively born infants, and 3430 of their siblings, for the HLA-DQB1 susceptibility alleles for type 1 diabetes. 17 397 infants and 1613 siblings had increased genetic risk, of whom 11 225 and 1574, respectively, consented to screening of diabetes-associated autoantibodies at every 3-12 months. In a double-blind trial, we randomly assigned 224 infants and 40 siblings positive for two or more autoantibodies, in consecutive samples, to receive short-acting human insulin (1 unit/kg; n=115 and n=22) or placebo (n=109 and n=18) once a day intranasally. We used a restricted randomisation, stratified by site, with permuted blocks of size two. Primary endpoint was diagnosis of diabetes. Analysis was by intention to treat. The study was terminated early because insulin had no beneficial effect. This study is registered with ClinicalTrials.gov, number NCT00223613.
Findings: Median duration of the intervention was 1.8 years (range 0-9.7). Diabetes was diagnosed in 49 index children randomised to receive insulin, and in 47 randomised to placebo (hazard ratio [HR] 1.14; 95% CI 0.73-1.77). 42 and 38 of these children, respectively, continued treatment until diagnosis, with yearly rates of diabetes onset of 16.8% (95% CI 11.7-21.9) and 15.3% (10.5-20.2). Seven siblings were diagnosed with diabetes in the insulin group, versus six in the placebo group (HR 1.93; 0.56-6.77). In all randomised children, diabetes was diagnosed in 56 in the insulin group, and 53 in the placebo group (HR 0.98; 0.67-1.43, p=0.91).
Interpretation: In children with HLA-conferred susceptibility to diabetes, administration of nasal insulin, started soon after detection of autoantibodies, could not be shown to prevent or delay type 1 diabetes.