Programmes aiming at prediction and prevention of insulin-dependent diabetes mellitus (IDDM), a multifactorial autoimmune disease, have been launched or are in the planning phase in several countries. We hypothesized that the costs of finding the correct target subjects for preventive interventions are likely to vary markedly according to the prediction strategy chosen. Average direct costs accruing in the Finnish IDDM Prediction and Prevention Project (DIPP) were analysed from the health care provider's viewpoint. The genetically targeted strategy included costs of assessing genetic IDDM susceptibility followed by measurement of marker(s) of islet autoimmunity in the susceptibility restricted population at 3 to 6-month intervals. In the pure immunological strategy markers of autoimmunity were repeatedly analysed in the entire population. The data were finally exposed to sensitivity analysis. The genetically targeted prediction strategy is cost-saving in the first year if autoimmune markers are analysed as frequently as under the DIPP project, and in all circumstances later. The 10-year direct costs per child are US$ 245 (present value $ 217, 5% discount rate) if the genetically targeted approach is used and $ 733 (present value $ 619) if the pure immunological strategy is chosen. In sensitivity analysis the 10-year costs (present value) per child of the genetically targeted strategy and of the pure immunological strategy varied from $ 152 to $ 241 and from $ 430 to $ 788, respectively. The genetically targeted IDDM prediction strategy is remarkably cost-saving as compared with the pure immunological strategy mainly because fewer subjects will need retesting during the follow-up.