Multifactorial modelling for caries prediction in Jordanian university students

To construct a prediction model for caries experience in Jordanian university students using a number of explanatory risk factors as predictors.

Data on salivary flow rate, buffering capacity, streptococci and lactobacilli counts, plaque accumulation, oral hygiene and between meal sugar intakes were tested as predictors of clinically and radiographically registered DMFS: Methods of analysis included correlation, then multiple regression, and finally dichotomisation of the DMFS data and application of discriminant analysis and logistic regression. The latter analyses were conducted in order to predict in which caries risk group an individual belonged rather than predicting (as with regression) their actual caries status. Two dichotomisation schemes were investigated; dichotomisation at the mean and at the 75th percentile.

The University of Jordan.

A random sample of 180 university students (77 male and 103 female).

Relationships were expressed as correlation coefficients, R2, and sensitivity, specificity and predictive values of the predictors, and also their validity and efficiency.

The highest correlation coefficient achieved was 0.43 (P < 0.0001) between sugar-containing snack intakes and DMFS. The predicted power of the fitted multiple regression model was low R2 = 0.38). Logistic regression with the DMFS data dichotomised at the 75th percentile indicated that the fitted caries model correctly identified 76 per cent of the subjects. Sensitivity and specificity values of the predictive battery were 80 per cent and 75 per cent respectively.

The multifactorial aetiology of caries remains unclear and requires further research. In the meantime, well-documented preventive measures should be implemented for this and similar populations.