Tenuous dose-response correlations for common disease states: case study of cholesterol and perfluorooctanoate/sulfonate (PFOA/PFOS) in the C8 Health Project

Abstract

Persistent organic chemicals, such as perfluorooctanoic acid (PFOA), perfluorooctanesulfonate (PFOS), dioxins, and polychlorinated biphenyls, pose investigative challenges because they are found in virtually everyone (there is no unexposed control group). To overcome this problem, outcome data in some studies are sorted by chemical dose level and findings in low-end dose groups are compared to sequential higher dose groups. An example is the C8 Health Project that evaluated serum PFOA/PFOS (C8) and total cholesterol among 46,294 West Virginia residents who lived, worked, or went to school for at least 1 year in a C8 contaminated drinking-water district and were over age 18 in 2005–2006. The risk for high total cholesterol (>240 mg/dL) measured via odds ratios (ORs) in logistic regression models showed sequential OR increases with PFOA quartile, in comparison to the lowest quartile (OR = 1.00), that were each significantly elevated (OR = 1.21, 1.33, and 1.40, respectively), but age, sex, and body mass index were stronger correlates. Importantly, the magnitude of cholesterol increase was small (12 mg/dL from lowest to highest exposure deciles) and comparison to similar statistics for the general U.S. population showed the C8 cohort had lower rates of high cholesterol. This suggests that inadvertent selection bias may have affected the lowest exposure quartile (control group), making tenuous the dose-response relationship between PFOA/PFOS and risk of high cholesterol. This case illustrates the substantial difficulties in assigning toxicological importance to statistical comparisons for common disease states that utilize subgroups with low exposures as an effective control group.

Keywords::

• perfluoroalkyl
• epidemiology
• obesity
• body mass index
• causation
• confounding
• selection bias