Do the correlations provided in this analysis demonstrate
associations strong enough to be used for risk management decisions?
If so, how?
If not, what would the panel make to enhance predictability from the two sets of data?
At this point, no. In this analysis ATSDR attempted to determine the extent of the relationship between an individual's urinary PNP level and age and environmental MP levels. Environmental data included wipe sampling data of MP for 406 households; urinary PNP data included levels for 858 participants. Environmental data were reduced to three summary variables: "kitchen composite"; arithmetic average of all samples from a household; and sampled values around the kitchen sink. Analysis consisted of linear and ordinal logistic regression of log-transformed data. Although the analyses demonstrate a general relationship between extent of environmental contamination and urinary PNP, the ability of the models to predict urinary PNP based on environmental MP contamination was poor.
The lack of a stronger association between measures of environmental contamination and urinary PNP as a biomarker of exposure is not surprising. Environmental sampling was designed to identify worst-case scenarios and may not be representative of surfaces that actually accounted for people's exposures. A major determinant of exposure is human behavior (i.e., behavioral factors account for contact with contaminated surfaces), which is not addressed in the analysis. Furthermore, sources of error in using urinary PNP as a biomarker of exposure have already been addressed. The reduction of measurement error through the use of exposure questionnaires to select appropriate environmental samples and timing of urinary PNP bioassays, use of creatinine-adjusted PNPs, use of environmental samples that more closely estimate actual exposures, and measures of high-risk behaviors temporally related to the times of collection should improve the predictive capability of future regression models.