The differential diagnosis between Type 1 (insulin-dependent) and Type 2 (non-insulin-dependent) diabetes is complicated since no specific markers are available for either disease. In this study, 244 consecutive patients were diagnosed with diabetes mellitus during two years in Malmö (230,000 inhabitants), corresponding to an incidence rate of 53.100,000(-1).year-1. Age, body mass index, HbA1c, C-peptide, and levels of islet cell antibodies were determined at the clinical onset, and related to the classification at diagnosis and at follow-up (n = 233) after a median time of 31 (range 1-49) months. After diagnosis, 42 of 244 (17%) were started on insulin while 202 of 244 (83%) were not. Islet cell antibodies were present in 25 of 42 (60%), and in 18 of 183 (10%), respectively. In the non-insulin treated group, patients with islet cell antibodies had lower body mass index (p less than 0.001), higher HbA1c (p less than 0.004), and lower C-peptide (p less than 0.001) than patients without. At follow-up, 11 of 18 (61%) islet cell positive patients were changed to insulin treatment, as were six other patients. Insulin was discontinued in five initially insulin-treated but islet cell antibody negative patients. The sensitivity, specificity and predictive value for insulin treatment at follow-up were for islet cell antibody positivity; 72%, 96% and 84%, respectively, and for low C-peptide value; 60%, 96%, and 80%, respectively. Islet cell antibodies and low C-peptide at diagnosis of diabetes mellitus are concluded to be useful markers to predict insulin dependence.