An 11-year prospective study was carried out in 180 non-diabetic patients with organ-specific autoimmune diseases to evaluate islet cell antibodies in predicting Type 1 (insulin-dependent) diabetes mellitus. Islet cell antibodies were characterised according to titres, persistence, complement-fixing ability, and pattern. During follow-up, 14 of 46 patients with islet cell antibodies persistently greater than 5 Juvenile Diabetes Foundation Units (JDF-U) (30.4%), none of 23 with islet cell antibodies between 2.5 and 5 JDF-U or fluctuating, and 3 of 109 without islet cell antibodies (2.7%), developed diabetes. The cumulative risk of developing diabetes was 70%, 0%, and 4%, respectively. All the patients who developed diabetes were females. Eight progressed to insulin-dependence acutely, four showed a transient period of non-insulin-dependence, while two were still insulin-free. No difference was found in titres of islet cell antibodies for the risk of diabetes. Complement-fixing islet cell antibodies enhanced the cumulative risk for the disease in patients with conventional islet cell antibodies at low-middle (> or = 2.5-40 JDF-U), but not at high (> or = 80 JDF-U) titres. Forty-two patients with islet cell antibodies were investigated for the whole or the selective pattern. In the presence of the whole pattern the cumulative risk for diabetes rose to 100%, while with the selective pattern it declined to 34%. The whole pattern was found in 83% of patients who developed Type 1 diabetes acutely. In patients with organ-specific autoimmune diseases, the whole islet cell antibody pattern greatly enhances the prediction for diabetes.