Robert Lordo

Battelle

Columbus, Ohio

Statistical Approach to Estimating the Effects of Environmental-Lead Exposure On Children's Blood-Lead Concentration

On January 5, 2001, the U.S. Environmental Protection Agency (EPA) issued new standards for levels of lead in dust and soil to be considered when evaluating the presence and magnitude of lead-based paint hazards in older homes containing children. To provide a scientific foundation for setting these regulatory standards, EPA asked Battelle's Statistics and Data Analysis Systems department to quantify the health risks to young children from exposures to lead-based paint hazards, lead-contaminated dust, and lead-contaminated soil in the nation's housing, and to develop a statistical approach to estimating how these risks may be reduced following promulgation of the lead standards. In this presentation, selected topics will be discussed on the statistical approaches that were developed in support of both of these objectives.

One topic that will be addressed in this presentation is the following: In estimating children's health risks associated with lead exposure, it was necessary to characterize the dose-response relationship between environmental-lead exposures and children's blood-lead concentration. As a first step in this process, the most recent, nationally representative, and otherwise "best available" data for characterizing both environmental-lead and blood-lead levels were found. However, EPA did not have an existing model which would address all objectives on risk estimation given the available environmental-lead data. Therefore, it was necessary to develop an additional predictive model, which was done using data from a previous lead exposure study. While this model was based on standard linear regression techniques, a number of statistical issues were encountered and needed to be resolved during the model's formulation, such as the need to adjust the model parameters to reflect the type of data to be used as input to the model, as this differed from the type of data used to fit the model. Such issues and how they were resolved will be addressed.

Back to Colloquium Series