Evaluating impacts of human physiological variability on distributions of internal dose metrics
On this page:
The chemical risk assessment process often involves dosimetric calculations that translate applied doses into internal dose metrics (DMs). Because of physiological variability in humans, it may be useful to treat DMs as distributions rather than scalars. In this work, we sought to understand general features of DM distributions and tested the hypothesis that DMs are lognormally distributed. To accomplish this, we used a physiologically based pharmacokinetic (PBPK) model for chloroform to calculate internal DMs. Applying Monte Carlo (MC) methods to account for variability in PBPK model parameters, we generated samples of DMs to analyze their distributional shapes. Next, we performed a global sensitivity analysis to determine which human parameters have the greatest influence on DMs and generated DM distributions using only the parameters meeting a minimal influence criterion. Using the Shapiro-Wilk normality test we found that some DM distributions are lognormal in shape, but this depended on the distributions chosen to represent parameter variability. Also, the DM distributions tend toward lognormality when more parameters are used. Including only the most influential parameters in the MC analysis yielded comparable DM distributions to those produced using the full MC analysis. Information we generated about DM distributions and techniques demonstrated for identifying influential parameters can be used with other PBPK models to better inform uncertainty in dosimetry.