Aims/hypothesis. To analyse the incidence of Type I (insulin-dependent) diabetes mellitus in the 0-34 years age group in Sweden 1983-1998. Methods. Incidence and cumulative incidence per 100 000 and Poisson regression analysis of age-period effects was carried out using 11 751 cases from two nation-wide prospective registers. Results. Incidence (95%-CI) was 21.4 (20.8-21.9) in men and 17.1 (16.6-17.5) in women between 0 and 34 years of age. In boys aged 0-14 and girls aged 0-12 years the incidence increased over time, but it tended to decrease at older age groups, especially in men. Average cumulative incidence at 35 years was 748 in men and 598 in women. Cumulative incidence in men was rather stable during four 4-year periods (736, 732, 762, 756), while in women it varied more (592, 542, 617, 631). In males aged 0-34 years, the incidence did not vary between the 4-year periods (p=0.63), but time changes among the 3-year age groups differed (p<0.001). In females the incidence between the periods varied (p<0.001), being lower in 1987-1990 compared to 1983-1986, but time changes in the age groups did not differ (p=0.08). For both sexes median age at diagnosis was higher in 1983-1986 than in 1995-1998 (p<0.001) (15.0 and 12.5 years in males; 11.9 and 10.4 in females, respectively). Conclusion/interpretation. During a 16-year period the incidence of Type I diabetes did not increase in the 0-34 years age group in Sweden, while median age at diagnosis decreased. A shift to younger age at diagnosis seems to explain the increasing incidence of childhood Type I diabetes.