A method is proposed for the prediction of the costs of a selective caries prevention programme; that is one where preventive treatment is given only to individuals who have been classified as at high risk of caries. A person so identified may be correctly classified as "high risk" or misclassified as "low risk". Similarly, a person at low risk may be correctly classified or misclassified as high risk. Therefore costs have to be calculated for each of the four situations, and expected frequencies of all four possible classifications have to be taken into account. After the identification of a logistic regression model which permits the prediction of whether a child will experience a high caries increment, sensitivity (SN) and specificity (SP) can be calculated for a set of different decision rules (critical scores, cutpoints or cut-off points Pcp). An example is presented in detail to demonstrate how the expected costs etc. for a specified prevention programme can be calculated without additional effort. The plotting of SN, SP and measures of costs as functions of Pcp permits the anticipation of expected effects of a caries prevention programme and helps to choose the appropriate cutpoint. Such plots may be helpful when comparing different caries prevention programmes and when deciding whether such a programme should be started.