Evaluation of clinical dental variables to build classifiers to predict celiac disease

The aim of this study was to evaluate the use of salivary variables to build statistical models for predicting celiac disease in symptomatic children.

The study group consisted of 52 children with celiac disease diagnosed by bowel biopsy, grade III or IV (4 to 12 years old, both sexes) and 23 healthy children as a control group. A logistic regression model was applied to evaluate an individual's belonging to one group or another. The performance of the model was evaluated by the value of area under the ROC curve. The salivary variables included in the model were the concentration of total proteins, calcium, Ca/P molar ratio, buffer capacity and salivary flow.

The total proteins (p = 0.0016) and Ca/P molar ratio (p = 0.0237) variables were significantly associated with the celiac condition. The value of the area under the ROC curve, estimated from the probabilities of the logistic model, showed that salivary component values allow the celiac condition of patients to be predicted with 85% accuracy (p <0.0001).

Logistic discriminant analysis built with salivary variables shows that these are good for predicting this eating pathology with 85% accuracy.