It has been postulated that treatment with nicotinamide may prevent or delay the onset of insulin dependent diabetes mellitus. We report the findings of a population based diabetes prevention trial which tests this hypothesis. 33,658 school children aged 5-7.9 years were randomly selected (by school) from a total population of 81,993 of such children in the Auckland (New Zealand) region. They were offered testing for islet cell antibodies. 20,195 (60%) consented to testing. Of these 185 had islet cell antibodies and met the criteria for treatment with nicotinamide. 173 received this treatment. The study population has an average follow up time of 7.1 years. The diabetes incidence of the untested controls was: 16.07 (12.4-20.5 95% CI) /100,000 person years at risk; in the group who were tested and treated when deemed appropriate: 7.14 (3.1-14.1 95% CI); and in the group offered testing but who did not consent ("refusers'): 18.48 (10.1-31.0 95% CI). The tested group had a rate of diabetes of 41% (20-85 95% CI) of the other groups combined after an age adjustment, which is significant (p = 0.008). The tested group combined with the "refuser' group (i.e. "intention to treat') also has a lower incidence than the control group (p = 0.12). Nicotinamide has a protective effect against the development of insulin dependent diabetes in this setting but the size of the effect has a wide confidence interval. Further follow up may define the magnitude of the protective effect within narrower limits.