A major aim in type 1 diabetes research has been to prevent or slow down the autoimmune loss of pancreatic β-cell function. Trials of immune modulation with ciclosporin, an anti-CD3 monoclonal antibody, rituximab, and abatacept have shown limited success but these interventions are all potentially toxic. Animal experiments have suggested that beneficial immune tolerance might be induced by using β-cell antigens. Now (Lancet 2011;378:319–27; see also Comment, ibid: 291–2) North American researchers have tried immunising with glutamic acid decarboxylase(GAD), a recognised target antigen in type 1 diabetes, without success. They enrolled patients with type 1 diabetes within 100 days of diagnosis. Initially patients aged 16–45 were enrolled but later children aged 3–15 years were included and the median age of the whole cohort was 14.5 years. A total of 145 patients were randomised to three groups: subcutaneous injections at 0, 4 and 12 weeks of GAD plus alum adjuvant, GAD for the first two injections and alum alone for the third, or all three injections of alum alone. The primary endpoint (the geometric mean area under the curve of serum C-peptide at 2 h during a 4-h mixed meal tolerance test at 1 year) did not differ significantly between the three groups. Similarly there were no group differences for HbA_{1c …}