The literature on exposure to lipophilic agents such as polychlorinated biphenyls (PCBs) is conflicting, posing challenges for the interpretation of potential human health risks. Laboratory variation in quantifying PCBs may account for some of the conflicting study results. For example, for quantification purposes, blood is often used as a proxy for adipose tissue, which makes it necessary to model serum lipids when assessing health risks of PCBs. Using a simulation study, we evaluated four statistical models (unadjusted, standardized, adjusted, and two-stage) for the analysis of PCB exposure, serum lipids, and health outcome risk (breast cancer). We applied eight candidate true causal scenarios, depicted by directed acyclic graphs, to illustrate the ramifications of misspecification of underlying assumptions when interpreting results. Statistical models that deviated from underlying causal assumptions generated biased results. Lipid standardization, or the division of serum concentrations by serum lipids, was observed to be highly prone to bias. We conclude that investigators must consider biology, biologic medium (e.g., nonfasting blood samples), laboratory measurement, and other underlying modeling assumptions when devising a statistical plan for assessing health outcomes in relation to environmental exposures. Key words: causal modeling, directed acyclic graphs, organochlorines, polychlorinated biphenyls, risk estimation, serum lipids. Environ Health Perspect 113:853-857 (2005). doi:10.1289/ehp.7640 available via http://dx.doi.org/[Online 17 March 2005]

**********

Persistent lipophilic xenobiotics pose particular methodologic challenges when assessing potential human health risks. The human health effects literature on exposure to lipophilic agents such as organochlorines (OCs) is equivocal, impairing our ability to quantify risks (Calle et al. 2002; Hunter et al. 1997; Laden et al. 2001a, 2001b). For example, Wolff and colleagues (Wolff 1985; Wolff and Toniolo 1995; Wolff et al. 1993, 2000) found an increased odds ratio for breast cancer for the highest quintile of wet-weight dichlorodiphenyl-dichloroethylene (DDE) and polychlorinated biphenyls (PCBs; expressed as nanograms analyte per milliliter serum) when compared with the lowest quintile, whereas Laden et al. (2001a, 2001b) found no association when concentrations of DDE and PCBs were standardized for serum triglycerides and cholesterol. No association was reported for PCBs and risk of breast cancer when expressing concentrations either as wet weight or lipid standardization values (Helzlsouer et al. 1999).

Varying laboratory practices for expressing PCB concentrations may in part account for the equivocal findings for human health end points. Serum PCB concentrations, as with other lipophilic xenobiotics, are dependent on serum lipid concentrations (Eyster et al. 1983; Guo et al. 1987). Under certain circumstances an equilibrium is reached, and information regarding serum PCB levels and serum lipid levels may be predictive of PCB body burden (Brown and Lawton 1984). If serum lipids indeed act in this manner, higher serum lipid levels should correspond to higher serum PCB concentrations (Calvert et al. 1996). However, serum OC concentrations and lipids are affected postprandially and need to be considered in relation to quantity and timing of food consumption (Phillips et al. 1989). When it is not possible to collect adipose tissue, serum samples are frequently used. However, serum (or plasma) introduces methodologic challenges with regard to lipids when estimating health risks, particularly when nonfasting samples are used (Whitcomb et al. 2005). Collection of fasting samples can hamper the feasibility of epidemiologic research and may adversely impact study participation. Nonfasting samples require further attention to serum lipids (Brown and Lawton 1984; Brown et al. 1994; Eyster et al. 1983).

Our limited understanding of the true relation between serum and adipose tissue concentrations of lipophilic xenobiotics in relation to serum lipids and particular health outcomes makes model specifcation difficult (Calvert et al. …