Importance: Type 1 diabetes requires major lifestyle changes and carries increased morbidity and mortality. Prevention or delay of diabetes would have major clinical effect.

Objective: To determine whether oral insulin delays onset of type 1 diabetes in autoantibody-positive relatives of patients with type 1 diabetes.

Design, setting, and participants: Between March 2, 2007, and December 21, 2015, relatives with at least 2 autoantibodies, including insulin autoantibodies and normal glucose tolerance, were enrolled in Canada, the United States, Australia, New Zealand, the United Kingdom, Italy, Sweden, Finland, and Germany. The main study group (n = 389) had first-phase insulin release on an intravenous glucose tolerance test that was higher than the threshold. The 55 patients in the secondary stratum 1 had an identical antibody profile as the main study group except they had first-phase insulin release that was lower than the threshold. Secondary strata 2 (n = 114) and strata 3 (n = 3) had different autoantibody profiles and first-phase insulin release threshold combinations. Follow-up continued through December 31, 2016.

Interventions: Randomization to receive 7.5 mg/d of oral insulin (n = 283) or placebo (n = 277), including participants in the main study group who received oral insulin (n = 203) or placebo (n = 186).

Main outcome and measures: The primary outcome was time to diabetes in the main study group. Significance was based on a 1-sided threshold of .05, and 1-sided 95% CIs are reported.

Results: Of a total of 560 randomized participants (median enrollment age, 8.2 years; interquartile range [IQR], 5.7-12.1 years; 170 boys [60%]; 90.7% white non-Hispanic; 57.6% with a sibling with type 1 diabetes), 550 completed the trial including 389 participants (median age, 8.4 years; 245 boys [63%]), 382 (96%) in the main study group. During a median follow-up of 2.7 years (IQR, 1.5-4.6 years) in the main study group, diabetes was diagnosed in 58 participants (28.5%) in the oral insulin group and 62 (33%) in the placebo group. Time to diabetes was not significantly different between the 2 groups (hazard ratio [HR], 0.87; 95% CI, 0-1.2; P = .21). In secondary stratum 1 (n = 55), diabetes was diagnosed in 13 participants (48.1%) in the oral insulin group and in 19 participants (70.3%) in the placebo group. The time to diabetes was significantly longer with oral insulin (HR, 0.45; 95% CI, 0-0.82; P = .006). The HR for time to diabetes for the between-group comparisons for the 116 participants in the other secondary stratum was 1.03 (95% CI, 0-2.11; P = .53) and for the entire cohort of 560 participants was 0.83 (95% CI, 0-1.07; P = .11), which were not significantly different. The most common adverse event was infection (n = 254), with 134 events in the oral insulin group and 120 events in the placebo group, but no significant study-related adverse events occurred.

Conclusions and relevance: Among autoantibody-positive relatives of patients with type 1 diabetes, oral insulin at a dose of 7.5 mg/d, compared with placebo, did not delay or prevent the development of type 1 diabetes over 2.7 years. These findings do not support oral insulin as used in this study for diabetes prevention.

Trial registration: clinicaltrials.gov Identifier: NCT00419562.