Objectives. (i) To validate caries risk profiles assessed with a computer program against actual caries development in preschool children, (ii) to study the possible impact of a preventive program on the risk profiles, and (iii) to compare the individual risk profiles longitudinally. Material and methods. Caries risk was assessed in 125 two-year-old children invited to participate in a 2-year caries-preventive trial with xylitol tablets. At 7 years of age, 103 were available for follow-up, 48 from the former intervention group and 55 from the control group. At baseline and after 5 years, 7 variables associated with caries were collected through clinical examinations and questionnaires, and scored and computed with a risk assessment program (Cariogram). Results. Children assessed as having a "low chance (0-20%) of avoiding caries" had significantly higher caries at 7 years of age compared to children with a lower risk in the control group (p<0.05) but not in the intervention group. Overall predictive accuracy and precision, however, were moderate in both groups. Less than half of the children remained in the same risk category at both ages, despite a largely unchanged consumption pattern of sugar. The majority of the children who changed category displayed a lowered risk at 7 years. The intervention program seemed to impair the predictive abilities of Cariogram. Conclusion. A modified Cariogram applied on preschool children was not particularly useful in identifying high caries risk patients in a low-caries community.