Prognosis of caries increment with past caries experience variables

This study aimed to select past caries experience variables as caries predictors and to compare their prognostic accuracy with the variables used in a method of caries prognosis developed in a Swiss child population. The data used for the analyses originated from an 8-year longitudinal study starting in 1980 on caries of 7.5-year-old Dutch children. Stepwise logistic regression analyses provided predictor variables. The newly introduced variables D23fi, D(123)i and D23pifi were interchangeable and the most powerful caries predictors. For the sake of uniformity and ease of application, D23fi (number of fissures of the permanent first molar with non-cavitated or cavitated caries lesions) was chosen as the first variable in the logistic regression equations. The gain in accuracy of the second and third predictor variables (number of sound primary molars and the number of buccal and lingual smooth surfaces of the permanent first molar with non-cavitated or cavitated caries) in the regression equations was limited. The D1 condition of surfaces could be omitted from the prediction models. The present forced three-predictor-regression equations for 7.5-, 9.5- and 11.5-year-old children were evaluated to assess their prognostic performance by using the area under the ROC curve as a measure of prognostic quality. For the present regression equations, the area under the ROC curve was 81-87%, which was higher compared to the Swiss regression equations for caries prognosis.