Other Important Methodological Issues Related to the Measurement of Lead
When making plans for the measurement of lead (or any toxicant, for that matter), one must consider the method detection limit (MDL) for the proposed sample analysis. The MDL refers to the level of precision for a particular laboratory’s analytic method—and these levels can vary greatly. For example, analysis of postnatal lead in the Oswego Children’s Study (Gump, Reihman, Bendinskas, Morgan, Dumas, Palmer, et al., 2005) relied on abstracting lead levels from primary care charts in New York State (NYS) as a consequence of the state’s mandated lead testing for one- to two-year-old children. However, for this sort of routine testing, finger/toe sticks are frequently used, and the subsequent lead analytic methods are designed for low-cost/high-throughput analyses with relatively high MDLs (typically 4.0 or 5.0 pg/dL). The high MDLs offered by these methods may be adequate if the investigator is considering the effects of high lead exposure; however, recent research has identified variables that are significantly associated with lead in cohorts in which nearly the entire sample has BLLs below 4 pg/dL. For example, our recent work documenting an association between lead and vascular responses to acute psychological stress (Gump, MacKenzie, Bendinskas, Morgan, Dumas, Palmer, et al., 2011) required a venous blood draw coupled with an analytic method having an MDL of 0.34 pg/dL.
Presence of Other Toxicants and Nonessential Metals
Exposure to nonessential metals rarely occurs as a single metal exposure, particularly when considering low-level exposures. Similarly, exposure to nonessential metals frequently occurs in the context of exposure to other toxicants (e.g., pesticides). In addition to simply using these exposures and potential confounds as covariates, a number of investigators have suggested a need to consider the effect of “metal mixtures" (Agay-Shay, Martinez, Valvi, Garcia-Esteban, Basagana, Robinson, et al., 2015; Dorea, 2014), as well as specific interactions between different toxicants (Stewart, Reihman, Lonky, Darvill, & Pagano, 2003). However, there are a number of methodological and analytic issues to consider when addressing these more complex models. First, it is necessary to consider sample collection issues. There are legal and ethical limitations with respect to the quantity of biological samples (e.g., blood draw limits) as well as potential budgetary limits to analysis of multiple toxicants (e.g., measurement of multiple organic compounds—such as perfluorochemicals—could quickly exceed $1000 per participant). With these considerations in mind, investigators should carefully consider the relevant literature and study hypotheses when selecting the appropriate toxicants to monitor.
Second, an analysis of multiple toxicants requires a large enough sample size to provide the power to test an interaction. Third, and related to the prior consideration of power, testing interactions between three or more toxicants (“mixtures”) within a general linear model requires a sample size that is unrealistic in many situations. New analytic methods are currently being developed to specifically test these higher-order interactions as well as potential nonlinear associations without requiring excessively large samples (Carlin, Rider, Woychik, & Birnbaum, 2013; Forns, Mandal, Iszatt, Polder, Thomsen, Lyche, et al., 2016; Kelley, Banker, Goodrich, Dolinoy, Burant, Domino, et al., 2019). Such interactions among toxicants are important to understand, as toxicant associations may vary across studies (producing inconsistencies) as a result of differences in levels of other toxicants across these study populations.
