A multivariate model to predict caries increment in Montreal children aged 5 years.

A study was carried out in Montreal (Canada) to predict caries development over the period of one year in primary teeth of kindergarten children (mean age 5 years 8 months +/- 4 months) living in a non-fluoridated area. The 302 children were examined at school on two occasions, one year apart. At the first examination selected predictors were collected: caries experience, salivary S. mutans and lactobacilli, buffer capacity, debris index, parents' education, fluoride consumption and family structure (one or two parents). Regression analysis was performed to select the significant factors. A total of 143 children developed new caries over the study period; the mean increment for the whole group was 2.1 dmfs. Sensitivity (Sn) and specificity (Sp) were calculated for each predictor and for the final model. The best model comprised only two factors, caries experience and lactobacillus. This could identify 81.8 per cent of children who would develop new caries during the next 12 months (Sn) and 77.4 per cent of those who would not (Sp). Among the single predictors caries experience alone reached 78.3 per cent for sensitivity and 77.4 per cent for specificity. None of the other predictors, except parents' education, was very good at predicting caries increment over one year.