Insulin immunization in animal models induces T-h e l p e r (Th) 2-like antibody subclass responses to insulin and other -cell antigens. The aim of this study was to d e t e rmine whether exposure to insulin in humans resulted in a similar subclass bias of the humoral immune response. Levels of IgG subclass antibodies to insulin (IAs), GAD, and IA-2 were measured before and after treatment w i t h insulin in the following groups of patients: 29 patients with newly diagnosed type 1 diabetes treated with intravenous and/or subcutaneous insulin; 10 newly diagnosed patients randomized to cyclosporin A (CsA) or placebo plus subcutaneous insulin for 12 months; and 14 islet cell antibody-positive relatives receiving either intravenous and subcutaneous insulin prophylaxis or no treatment. At the onset of diabetes, the major subclass distributions of insulin autoantibodies (IAAs) were IgG1 and, to a lesser extent, IgG4. After insulin treatment in the 29 new-onset patients, IAs were initially of the IgG1 subclass. IgG4-IAs appeared later, but at 12 months, they were at higher levels than IgG1-IAs in 11 patients. Responses were higher in children compared with adults and were higher in subjects with IAAs (P < 0.001). Insulin prophylaxis in relatives showed a similar profile, with a decline in levels of IgG1-IAs after cessation of daily subcutaneous insulin. Patients treated with CsA took longer to develop IAs and showed suppressed levels of IgG4-IAs; however, their levels of high-titer IgG1-IAs persistently rebounded after completion of CsA therapy. Despite the presence of IgG4-IAs in most insulin-treated patients and relatives, a shift to IgG4-anti-GAD or IgG4-IA-2 was not found for up to 3 years after the initiation of insulin t h e r a p y. While our findings need to be correlated with T-cell cytokine responses, we suggest that the strong IgG4-IA response in insulin-treated patients is consistent with an enhancement of Th2 immunity, but there is no evidence of subsequent spreading of potentially Th2-associated IgG4 responses to other autoantigens. D i a b e t e s 4 9 :9 1 8-925, 2000