Prediction of children's deciduous teeth lead concentrations using multiple linear regression and logistic models

摘要

This paper presents the results of multiple linear regression and logistic models, using deciduous teeth lead concentrations in children aged from 4 to 14 years as a dependent variable, collected during the course of the EPLODIN Project, through a sanitary campaign promoted by the Epidemiological and Clinical Researching Unit and a private enterprise. A total of 371 data have been used for this purpose. Since data shown a positively skewed distribution and fitted to a log-normal distribution, they were normalised by logarithmic transformation for models developing. Different environmental variables, both physiological and risk ones, were used as independent variables. Multiple regression models were carried out using Backward Selection step. The R-squared values range from 0.2263 for the first model, with nine independent variables, to 0.1910 for the last model, with only two independent variables: home antiquity and tooth weight. Three different multiple logistic models were obtained. The last one included tooth weight, home antiquity and jaw. Tooth weight proved to be inversely related and belonging to the upper jaw directly related to high lead concentrations, meanwhile living at homes older than 25 years showed to be an important risk factor for high lead concentrations in children.