Impact/Purpose:

The outstanding issues are a mix of computational and theoretical statistical problems, and are best investigated through exploring specific models and datasets. Thus, we need to identify PBPK/PD models for chemicals of interest to the Agency, identify the outstanding theoretical statistical issues that need to be solved, and set up a computing platform to handle the large computing loads this exploration will entail. Finally, theoretical statistical exploration needs to precede hand-in-hand with practical PBPK/PD model development, ultimately leading to the publication of a practical framework for applying statistical methodology to PBPK/PD models.

Description:

PBPK/PD models are large dynamic models that predict tissue concentration and biological effects of a toxicant before PBPK/PD models can be used in risk assessments in the arena of toxicological hypothesis testing, models allow the consequences of alternative mechanistic hypotheses to be predicted in highly non-linear systems. Whether for quantifying the uncertainty of the extrapolation of a dose metric or determining which hypothesis is better supported by the evidence, and by how much, it is critical that the uncertainties inherent in such modeling be properly quantified, and established principles of statistical inference be brought to bear.

However, such quantification is far from simple. Biological systems models generally involve large numbers of parameters whose values are known with varying degrees of uncertainty. Some parameters (for example, tissue-specific fractions of cardiac output and fractions of body mass) can reasonably be thought of as varying among individuals of a population. Other parameter values have been estimated in in vitro systems, or may have been calculated from properties of the chemicals involved (for example, partition coefficients in a physiologically-based pharmacokinetic model). It is typically the case that the values of parameters in a biological systems model will be derived from the results of several different experiments, of very heterogeneous designs, and often conducted in different laboratories at different times. Over the last decade or so a small number of statistical investigators have begun to explore the issues involved in applying statistical methods to such systems. While there are examples of the application of statistical methodology to PBPK/PD models in the literature, there is as yet no coherent approach to this activity that takes into account the special issues that arise in PBPK/PD modeling. The NCCT is establishing a program to further develop the statistical methodology to support the rigorous quantification of both the uncertainty of parameter estimates and of model predictions, with the aim of providing a framework for statistical applications in PBPK/PD modeling.

Record Details:

Record Type:PROJECT

Projected Completion Date:09/30/2008

OMB Category:Other

Record ID: 149143

Show Additional Record Data

Project Information:

Progress :Investigators in NHEERL, NERL, and CIIT Centers for Health Research have been approached for appropriate models and data sets, and several models have been identified for further work. A graduate student in biostatistics and his advisor have been recruited to work on outstanding theoretical issues and apply solutions to models and data sets of interest to the Agency. The literature surrounding this topic has been surveyed, and a literature review is in progress.

Approach :The outstanding issues are a mix of computational and theoretical statistical problems, and are best investigated through exploring specific models and datasets. Thus, we need to identify PBPK/PD models for chemicals of interest to the Agency, identify the outstanding theoretical statistical issues that need to be solved, and set up a computing platform to handle the large computing loads this exploration will entail. Finally, theoretical statistical exploration needs to precede hand-in-hand with practical PBPK/PD model development, ultimately leading to the publication of a practical framework for applying statistical methodology to PBPK/PD models.

Relevance :PBPK/PD models are of increasing interest in the Agency for use in extrapolation from animals to humans, across routes of exposure, and for integrating acute effects over discrete episodic exposures. Better characterization of methods for estimating parameters and quantifying uncertainties in such models predictions will remove one major impediment to their more general application. This will allow replacement of default uncertainty factors with transparent mechanism-based statements of scale and uncertainty, in turn decreasing the subjectivity and increasing the transparency of environmental health risk assessments.

Project IDs:

ID Code :IIID-1

Project type :Partner Specific