RE: “INVITED COMMENTARY: EXPOSURE BIOMARKERS INDICATE MORE THAN JUST EXPOSURE”

The commentary by Savitz and Wellenius (1) on the article by Sagiv et al. (2) illustrates the importance of understanding determinants of chemical exposure biomarkers and empirically testing how they influence relationships between chemical exposures and human health. A key point in this understanding is the relative contribution of physiological and nonphysiological factors to between-person variation in chemical exposure biomarkers. I believe that Savitz and Wellenius underestimate the nonphysiological variation of several chemical exposure biomarkers in their statement that “many toxicants, including PFAS, phthalates, and fire retardants, are ubiquitous in our environment, and levels are not likely to be notably different across homes, product selection, or other factors” (1). I argue below that the relative contribution of nonphysiological sources to between-person variation in chemical exposure biomarkers is as great as or greater than physiological ones for some chemicals, especially phthalates. Moreover, the relative contributions of nonphysiological and physiological sources of between-person variation differ across the life span.

In the case of phthalates, there is considerable evidence that nonphysiological factors result in substantial between-person differences in urinary concentrations of phthalate metabolites. My colleagues and I previously reported that urinary monoethyl phthalate concentrations were 2.5-fold higher among women who used cologne than among those who did not (geometric mean = 111 ng/mL vs. 42 ng/mL) (3). In addition, Rudel et al. (4) found that urinary levels of di-2-ethylhexyl phthalate metabolites decreased by more than 50% during a dietary intervention that minimized the use of plastic packaging in food preparation and storage. Moreover, other nonpersistent chemicals, like parabens, also have substantial between-person variation related to the use of personal care products, and this has been demonstrated in both experimental and observational studies (3, 5). Between-person variation in phthalate exposure may also arise from the types or brands of products used. Koo et al. (6) previously reported that phthalate diester levels in different brands of perfume, nail polish, hair products, and deodorant could range from nondetectable to over 12 parts per thousand. The magnitude of these variations in phthalate exposure and phthalate biomarkers is quite large, especially compared with the relatively weak associations between physiological factors and chemical exposure biomarkers (2, 7).

A second important point to consider is that the relative contribution of physiological and nonphysiological factors to between-person variation in chemical exposure biomarkers depends on the timing of development when exposure assessment is conducted. For instance, breastfed infants have considerably higher levels of perfluoroalkyl substances (PFAS) than nonbreastfed infants (8, 9). In one study, duration of breastfeeding explained the largest amount of variation in children’s serum PFAS concentrations, as compared with other exposure sources (9). Breastfeeding duration is also a predictor of levels of other persistent pollutants, including polybrominated diphenyl ether flame retardants (10). Thus, for persistent chemical exposure biomarkers, including PFAS, the proportion of between-person variation related to nonphysiological factors will be greater than the variation due to physiological factors during infancy or childhood.

Epidemiologists must carefully consider the multiple sources of between-person variation in chemical exposure biomarkers, while also appreciating that the relative contribution of these different sources varies by chemical exposures and across developmental life stages. Thus, broad and sweeping generalizations about how physiological factors influence biomarkers of chemical exposure need to be accompanied by appropriate caveats. Otherwise, such broad generalizations could diminish the potential value of chemical biomarkers, when in fact there are many cases where they can validly and reliably distinguish interindividual differences in true exposure.