Development of an oral reference dose for the perfluorinated compound GenX

Abstract

Ammonium 2,3,3,3-tetrafluoro-2-(heptafluoropropoxy)-propanoate, also known as GenX, is a processing aid used in the manufacture of fluoropolymers. GenX is one of several chemistries developed as an alternative to long-chain poly-fluoroalkyl substances, which tend to have long clearance half-lives and are environmentally persistent. Unlike poly-fluoroalkyl substances, GenX has more rapid clearance, but has been detected in US and international water sources. There are currently no federal drinking water standards for GenX in the USA; therefore, we developed a non-cancer oral reference dose (RfD) for GenX based on available repeated dose studies. The review of the available data indicate that GenX is unlikely to be genotoxic. A combination of traditional frequentist benchmark dose models and Bayesian benchmark dose models were used derive relevant points of departure from mammalian toxicity studies. In addition, deterministic and probabilistic RfD values were developed using available tools and regulatory guidance. The two approaches resulted in a narrow range of RfD values for liver lesions observed in a 2-year bioassay in rats (0.01-0.02 mg/kg/day). The probabilistic approach resulted in the lower, i.e., more conservative RfD. The probabilistic RfD of 0.01 mg/kg/day results in a maximum contaminant level goal of 70 ppb. It is anticipated that these values, along with the hazard identification and dose-response modeling described herein, should be informative for risk assessors and regulators interested in setting health-protective drinking water guideline values for GenX.

Keywords: GenX; benchmark dose (BMD) modeling; per- and polyfluoroalkyl substances (PFAS); reference dose (RfD); risk assessment.

Figure 1

Comparison of 6‐carbon perfluoroether carboxylic acid GenX to the 8‐carbon perfluoroalkyl carboxylic acid, PFOA [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 2

Effects of GenX on male rat red blood cell parameters. Top row: dose‐response for RBC, Hb and percentage hematocrit in 90‐day rat study. Bottom row: dose‐response for RBC, Hb and percentage hematocrit at 3, 6 and 12 months in the 2‐year rat bioassay. Data adapted from Haas (2009) and Craig (2013). Hb, hemoglobin; RBC, red blood cells

Figure 3

Comparison of maternal effects with developmental effects in mice. A, Maternal bodyweight on GD18. B, Maternal bodyweight LD21. C, Maternal bodyweight LD21 attributed to liver weight increases. D, Maternal food intake from LD1‐21. E, Female pup weight at PND1, PND21 and PND 40. F, Male pup weight at PND1, PND21 and PND40. Data represent mean ± SD. *Statistical significance relative to control (Dunnett's test, P < .01). Data adapted from Edwards (2010b). GD, gestational day; LD, lactational day; PND, postnatal day

Figure 4

Effects of GenX on β‐oxidation enzyme activity. Data represent hepatic peroxisome β‐oxidation activity toward [¹⁴C]palmitoyl coenzyme A in male (left) and female (right) mice (top) and rats (bottom). *Statistical significance relative to control (Dunnett's test, P < .01). Data adapted from Haas (2008a, b)

Figure 5

BMD modeling (frequentist models). A, Cystic focal degeneration in male rats of 2‐year bioassay (fit P = .58). B, Fetal weight in rats (exponential model; fit P = .62). Data adapted from Caverly Rae et al. (2015) and Edwards (2010a). BMD, benchmark dose; BMDL, benchmark dose and their corresponding 95% lower confidence limit values [Colour figure can be viewed at wileyonlinelibrary.com]

Figure 6

pRfD derivation. Bayesian model average BMD₁₀ and BMDL₁₀ values for cystic focal degeneration in male rats served as input for generating an animal distribution for 10% extra risk (AD₁₀) of liver lesions (steps 1 and 2). Note: a frequentist BMD model plot from BMDS v2.7 is shown here for diagrammatic purposes only. Default adjustment factor distributions for allometric scaling factors (AF_BW) and additional interspecies extrapolation uncertainty (AF_TK/TD) were used to estimate equivalent human distributions of doses associated with 10% extra risk (HD₁₀ ^50%) of liver lesion in the typical human (step 3). Default adjustment factor distributions for human variability between the median and most‐sensitive human (defined here as the difference between the first and 50th percentiles) (AF_H) were used to estimate a distribution for doses associated with 10% extra risk (HD₁₀ ^50%) of liver lesion in the most‐sensitive 1% of individuals (HD₁₀ ^1%) (step 4). The fifth percentile of this distribution was considered as an interim pRfD, which was then further adjusted by a threefold UF_D. Note: the default adjustment factor distributions are those described in Chiu et al. (2018). BMD, benchmark dose; BMDL, benchmark dose and their corresponding 95% lower confidence limit values; pRfD, probabilistic reference dose [Colour figure can be viewed at wileyonlinelibrary.com]