Insulin deficiency due to autoimmune destruction of pancreatic beta cells (insulin is an autoantigen) is responsible for insulin-dependent diabetes mellitus. Since 1923, substitutive administration of insulin has been used to treat the disease. Surprisingly, initial usage of insulin is associated with partial resumption of insulin secretion in most patients. This phenomenon is intensified by aggressive insulin therapy. When observed at a late phase of destruction, it has been interpreted as an immunomodulatory effect of insulin which is presumed to act either by masking the target of effector cells in the autoimmune reaction (beta cells at rest because of glycaemic normalisation would expose fewer antigens) or by direct action on autoreactive T lymphocytes (which are rich in insulin receptors). There could also be a direct beneficial effect on anti-apoptotic or pro-regenerative beta cells. Efficient prevention of diabetes has been achieved by administration of parenteral insulin to non-obese diabetic (NOD) mice. Certain sequences of the B chain appear to be responsible for this effect, which seems to be immunomediated. Some preliminary data from the groups of G. Eisenbarth and N. MacLaren have suggested that this effect could be obtained in man by administering small doses of subcutaneous insulin to prediabetic patients. Two trials have been under way since 1994: DPT1 (a non-randomised trial concerning children and adults at high risk) in the United States, and EPP-SCIT (a randomised trial concerning children at very high risk) in Europe. Another approach has also been attempted in diabetes as well as other diseases with an organ-specific autoimmune reaction (SEP, PR) i.e. oral administration of an antigen present at the reaction site. A positive effect has been shown by the group of H. Weiner in the NOD mouse in which islet infiltration was reduced and diabetes prevented by "oral tolerisation" with insulin. Oral insulin is easy to use in therapeutic studies and is currently being administered in two trials: DPT1 (prediabetic children and adults at moderate risk) in the United States, and DIOR (recently diabetic children and adults) in France. However, recent experimental data suggest that oral administration of an antigen in certain artificial circumstances can trigger an autoimmune reaction in the animal, which would indicate that due caution should be exercised in trials involving prediabetic patients.