Document Z4O7a1ap74dErkYR6pe7YM7Xd

ehponIine.org P-1 f iZ z ^ '3 ,i4 7 ENVIRONMENTAL HEALTH PERSPECTIVES Accumulation and Clearance o f PFOA in Current and Former Residents o f an Exposed Community Ryan Seals, Scott M. Bartell, and Kyle Steenland doi: 10.1289/ehp.1002346 (available at http://dx.doi.org/) Online 22 September 2010 Health Sciences National Institutes of Health U.S. D epartm ent o f Health and Human Services Page 1 of 33 P-2 Manuscript Title: Accumulation and Clearance of PFOA in Current and Former Residents of an Exposed Community Author Names: Ryan Seals', Scott M Bartell2, Kyle Steenland' 'Department of Environmental and Occupational Health, Emory University, Atlanta, GA 2Program in Public Health and Department of Epidemiology, University of California, Irvine, CA Corresponding Author: Ryan Seals 1518 Clifton Road NE, Atlanta, GA 30322 e. rseals@hsph.harvard.edu . 404-727-5368 . 404-452-1960 f. 404-727-8744 1 Page 3 o f 33 P-3 Abstract Background. PFOA is a perfluoroalkyl acid found in over 99% of Americans. Its health effects are unknown. Prior estimates of serum half-life range from 2.3 to 3.8 years. Objectives. To assess the impact of years of residence and years since residence on serum PFOA concentration in a sample of current and former residents of six water districts in West Virginia and Ohio exposed to PFOA emissions from an industrial facility. Methods. Serum samples and questionnaires, including residential history, were collected in 2005-2006. We modeled log serum PFOA (ng/mL) for current residents as a function of years of residence in a water district, adjusted for a variety of factors. We modeled the half-life in former residents via a two-segment log-linear spline in two water districts with high exposure. Results. We modeled serum PFOA concentration in 17,516 current residents as a function of years of residence (R2=0.68). Years of residence was significantly associated with PFOA concentration (1% increase in serum PFOA per year of residence), with significant heterogeneity by water district. Half-life was estimated in two water districts comprising 1,573 total individuals. Years-since-residing in a water district was significantly associated with serum PFOA, yielding half-lives of 2.9 and 8.5 years for water districts with higher and lower exposure levels, respectively. Conclusion. Years of residence in an exposed water district was positively associated with observed serum PFOA in 2005-2006. Differences in serum clearance rate between low- and 3 Running Title: Accumulation and Clearance o f PFOA in a Community Key words: PFOA, C8, PFA, serum levels, half-life, water contamination P-4 Page 2 of 33 Acknowledgements/Grant Support/Competing Financial Interests Declaration This research is funded by the C8 Class Action Settlement Agreement (Circuit Court of Wood County, WV, USA) between DuPont and plaintiffs. Funds were administered by the Garden City Group (Melville, NY) that reports to the court. Our work and conclusions are independent of either party to the lawsuit. The authors declare they have no competing financial interests. Abbreviations: PFOA/C8 - Perfluorooctanoic acid PPARa - Peroxisome proliferator-activated receptor alpha 2 P-5 Page 4 of 33 high-exposure water districts suggest a possible concentration-dependent or time-dependent clearance process, or inadequate adjustment for background exposures. 4 Page 5 of 33 P-6 Introduction Perfluorooctanoic acid (PFOA, or C8) is a perfluoroalkyl acid used in the production of many fluoropolymers, including non-stick cookware, waterproofing, and flame retardants (Kennedy et al. 2004). Not naturally occurring, PFOA has been found in nature around the world, in multiple species, and in over 99% of serum samples obtained from the 2003-2004 National Health and Nutrition Examination Survey (Calafat et al. 2007; Houde et al. 2006). Despite recent regulatory and industrial efforts to phase out production and use by 2015, PFOA accumulates and persists in the environment, and human exposure is not expected to cease for some time. Studies in rodents suggest that PFOA may be associated with many disease outcomes, including increased hyperplasias and benign tumors of the testicles, liver, and pancreas, low birth weight, decreased immune response, and decreased cholesterol (Hines et al. 2009; Kennedy et al. 2004; Lau et al. 2007). However, the appropriateness of the animal models has been called into question because of the wide range of clearance rates observed between and within species, and because of species-specific differences in the role of the PPARa-mediated effects of PFOA (DeWitt et al. 2009; Lau et al. 2007; Rosen et al. 2009). Human studies of PFOA have been thus far largely limited to cross-sectional studies and retrospective analyses of occupational cohorts. To date no clear health effects of PFOA have been established, but studies so far are sparse. Average concentrations of 3.9 ng/mL (equivalent to parts per billion) were found in a nationally representative sample of US citizens in 2003-2004, with higher levels in males and whites (Calafat et al. 2007). Levels ranging from 100 to 5,000 ng/mL have been observed in occupational cohorts (Lau et al. 2007; Lundin et al. 2009; Olsen and Zobel 2007). 5 P-7 Page 6 of 33 The current study population is derived from the C8 Health Project, which has been described previously (Frisbee et ai. 2009). The C8 Health Project collected data on 69,000 current and former residents of the mid-Ohio valley who had been exposed to PFOA via contaminated drinking water. The average serum PFOA in this population was 82 ng/mL, with a median of 28 ng/mL (Steeniand et al. 2009). Establishing the rate of clearance of PFOA from the body is important for retrospectively determining lifetime exposure levels and for predicting future serum concentrations. PFOA is known to persist in human serum long after exposure has ceased, and is not metabolized in the body (Kennedy et al. 2004). Current estimates of serum half-life are derived from three primary sources. A study of 26 former employees of a manufacturing facility that produced PFOA, with a mean initial serum concentration of 799 ng/mL, estimated an average half life of 3.8 years (95% Cl: 3.0-4.1), with individual half-lives ranging from 1.5 to 9.1 years based on a Five year follow up (Olsen et al. 2007), A more recent study of PFOA levels in 138 residents of a German community following implementation of charcoal filtration estimated a mean half-life of 3.26 years (range 1.03-14.67) (Brede et al. 2010). Finally, an ongoing study of 200 community residents living near a PFOA facility and part of the C8 Health Project (a subset of the same population studied here) were followed for one year and, based on multiple blood samples and a mean initial serum concentration of 180 ng/mL, exhibited a half-life of 2.3 years (95% Ci: 2.12.4) (Bartell et al. 2009). Half-life estimates based on only one year of follow-up must be considered with caution. Preliminary results indicate that a traditional exponential decay model is sufficient for describing the clearance of PFOA from the body, despite earlier indications that clearance may occur in a time-dependent fashion in animals (Tan et al. 2007). 6 Page 7 of 33 p.8 Our goal in this study was to estimate the effect of duration of residence on PFOA levels among current residents and to estimate the effect of years-since-leaving among former residents. The latter goal also involved estimating half-life. While use of cross-sectional vs. longitudinal data to estimate half-life is not optimal, it can provide useful information. 7 P-9 Page 8 of 33 Methods Data source. The C8 Health Project was conducted between August 2005 and August 2006, and collected health data from current and former residents of the study area using an extensive questionnaire and blood test, including the serum concentration of PFOA (n=69,030). A fuller description of the study has been published previously (Frisbee et al. 2009). The questionnaire, in addition to basic demographic information, included an extensive residential history beginning in 1980, and included information on water consumption source at each address (public/private, tap/bottled water). The questionnaire also queried individuals about behaviors including smoking, alcohol consumption, and vegetarianism. Study participants. We identified individuals from the C8 Health Project who had consented to further follow-up and release of identifiable data to us, and who had provided residential history during the C8 Health Project (n=48,880). As noted, our goal was to study the effect of duration of residence in a water district, and of years-since-leaving a water district, on PFOA levels measured in 2005-2006. Ideally for our purposes, water within a water district would have had a constant level of contamination over time, so that years of residence would reflect a constant exposure. In practice, PFOA emissions from the plant increased over time, peaking in the 1990s. In addition, different water districts are known to have different levels of contamination, largely due to distance from the plant (Steenland et al. 2009). We first excluded individuals who had a self-reported history of employment by DuPont because of likely high exposure levels at the chemical plant (5%). We then excluded those who had a history of residence in more than one water district (25%); ever reported a private well as 8 Page 9 of 33 p. 10 their primary source of drinking water ( 11%); or reported intermittent residence in the water district of interest (9%). These exclusions were motivated by the desire for subjects to have continuous exposure to a single source of exposure, within a single contaminated water district. Because the limit of detection for serum PFOA was 0.5 ng/mL we also excluded individuals at or below this level (2%). Finally, we excluded individuals who reported overlapping residences in their residential history (3%). After all exclusions, 19,460 subjects remained for analysis. Current residents. After the above exclusions., we identified individuals who were residing in one of the six water districts on the date of interview and testing ("current residents"; n= 17,516). The focus of the analysis of current residents was the effect of cumulative years lived in a water district. Former residents. We studied a group of former residents to determine the effect of years-since-leaving on PFOA measured in 2005-2006 and to estimate half-life. We limited our analysis of former residents to the two water districts of Little Hocking and Lubeck, because these districts are hypothesized to have higher levels of exposure and half-life could be more reliably estimated. In addition to the above criteria for current residents, among former residents we excluded individuals with less than 2 years residence in a water district (11%), and a serum PFOA concentration lower than 15 ng/mL (28%). These criteria were used to limit the analysis to individuals who (1) had enough history in the water district to build up substantial levels of PFOA, and (2) had sufficiently high baseline PFOA concentrations, such that they had not reached background levels of PFOA by the interview date. The final cohort of former residents consisted of 643 Little Hocking residents and 1,029 Lubeck residents. Statistical analysis. 9 p. 11 Page 10 of 33 General models without exposure terms of interest. For all analyses, based on normalizing residuals for skewed data and in accordance with prior published results (Steenland et al. 2009), we modeled the natural logarithm of PFOA as measured in 2005-2006 as the outcome of a linear set of predictors. Variables considered as potential covariates were: sex, age, race (white vs. non-white), BMI, growing one's own vegetables, vegetarianism, alcohol consumption, current and former smoking, regular exercise, and use of bottled water as primary source of drinking water. These variables were all measured in 2005/2006, and had been used in prior analyses of PFOA levels (Steenland et al. 2009). Age, BMI, and date of interview were categorized as previously (Steenland et al. 2009). For the analysis of current residents, with six water districts, using a backward selection process with a cutoff of 0.10 and without including duration of residence (our principal variable of interest) we created models individually for each of the six water districts, to determine which covariates would be included in final models. The backward selection process iteratively fit models, dropping the least significant covariate at each step until all were significant at the cutoff level of 0.10. In analyses with the six water districts combined, we added an indicator variable for water district to the model, which allowed us to determine the relative importance of residence in a particular water district, as well as the effect of having resided in that district. This process was repeated for the analysis of former residents which was restricted to two water districts. Analyses for duration of exposure The goal of the first analysis was to estimate the relationship between duration of exposure to public water within a district and the measured serum PFOA level in 2005 or 2006. 10 Page 11 of 33 p. 12 For this analysis we considered only individuals residing in the six water districts on the date of interview and testing (current residents). Cumulative years in the water district was analyzed as both a continuous and categorical variable. ln(PFOA2oo5) = a + 0 - CUM YEARS + 8 X [1] where CUM YEARS represents the number of years lived in the water district, and 8 and X are parameter and covariate vectors. Analyses by years since leaving (half-life analysis) A second analysis was performed to estimate the half-life in former residents only (restricted to two water districts), via analyzing the relationship between the number of years since living in the water district and the measured serum PFOA level in 2005 or 2006, using the following model: ln(PFOA2005 - 5) = a + p r YEARS SINCE + j}2 CUM YEARS +8 X [2] where YEARS SINCE represents the number of years elapsed since residence in the water district, CUM YEARS is the number of years lived in the water district, and 8 and X are parameter and \ covariate vectors. This analysis was restricted to two water districts which had the highest levels, so as to avoid as much as possible the problem of background levels affecting our estimation of the elimination parameter (p). In this analysis, we subtracted background levels (5 ng/ml) from all subjects, and required that all subjects had at least 15 ng/ml PFOA in 2005. Although performed on a single cross-sectional measurement of serum PFOA, rather than the more traditional longitudinal analysis of repeated measurements, the analysis by years since leaving can provide an approximation of the clearance rate of PFOA,. The number of years 11 p. 13 Page 12 of 33 elapsed since living in the water district was analyzed as both a continuous and categorical variable. The half-life of a serum concentration describes the number of years required for the concentration required to reach one-half of the baseline level. Elimination of a substance from the circulatory system is usually described by a logarithmic process, where the concentration of the substance at time t (Ct) is related to the baseline concentration (Co) by the time-dependent term e x\ where X is a positive decay constant, i.e. C,= Coetx". This is called a first order elimination in which the rate of elimination is constant and does not depend on initial concentration. To obtain the half-life {tin), we seek the time required such that C, is V%of Co , or {tin) such that l/2*Co = Coe('h>. Rearranging, we have: tl/2 = - ln(l/2) / X. [3] The slope of the line describing the relationship between YEARS and ln(PFOA) is p, which is equal to X in equation 3 above (also see below), and hence can be used to solve for the estimated number of years that would be required for the PFOA to fall by half. Since by model (2) we have predicted pFOA=ea epiYEARSSINCEefix, and, given that some change in YEARS SINCE will cut predicted PFOA in half, we have V4=p*iVEARSS[NCEl YEARSS1NCE2\ sjnce the intercept and covariate terms cancel out. The change in YEARS SINCE is then the half-life, and taking the log of the last expression we regain model (3) and have ln(0.5) = /? t)>2 [4] where /? is equivalent to X. 12 Page 13 of 33 p. 14 Graphing a scatter plot of log PFOA by YEARS SINCE, we found an apparent non-linear relationship using a LOESS non-parametric curve. We then modeled the relationship using a two-segment linear spline (Steenland and Deddens 2004). The spline is included in the model through the addition of a time-dependent variable that is zero prior to the knot and increases after the knot. Below is an example, with a knot at YEARS SINCE=4\ ln(PFOA2005 - 5 ) - a + fit YEARS SINCE + /?2 max[0,(YEARS SINCE - 4)] + CUM YEARS + 5 X [51 The expression max[0,(KA/?S SINCE - 4)] evaluates to 0 when YEARS SINCE < 4 and equals (YEARS SINCE - 4) when YEARS SINCE > 4. The slope of the regression line is therefore Pi prior to the knot, and (P1+P2) after the knot. We chose the knot based first on visual inspection of the relationship between years elapsed and ln(PFOA), to determine the likely region of interest, followed by an iterative procedure where we picked the knot with the best model likelihood.. We used an F-test to test the increase in the goodness-of-fit in the spline model over the linear model. Outliers with absolute studentized residuals greater than 3 were discarded (0.5%). Prior work has suggested that half-life estimates after truncation to account for near background levels can introduce bias (Michalek et al. 1998). We performed a sensitivity analysis using various truncation values (serum PFOA concentrations below the truncation value were discarded) to assess the robustness of our half-life estimates, retaining the same models and knot locations as in the initial analysis with truncation at 15 ng/mL. In all analyses, 5 ng/mL was subtracted after truncation and before regression. The method for ascertaining the serum concentration of PFOA has been described previously (Frisbee et al. 2009). This study was approved by IRBs at all C8 Science Panel 13 p. 15 Page 14 of 33 institutions, and all applicable requirements for human research were met. All participants gave written informed consent to participate in the C8 Health project; consent procedures have been described previously (Frisbee et al. 2009). All analyses were performed using SAS v9.l (Cary, NC). Images were generated using PASW vl7.0 (Chicago, IL). 14 Page 15 of 33 p. 16 Results Median levels of PFOA for current and former residents are shown in Table 1. In residents still residing in the six water districts at the time of the interview, differences in serum PFOA levels were apparent across water districts, sex, use of bottled water, growing own vegetables, smoking history, and date of testing, similar to results reported previously for the entire C8 Health Study cohort (Steenland et al. 2009). The subset of 1,672 former residents had higher PFOA levels than the current residents because this subset was limited to residents of Little Hocking and Lubeck, the two highest-exposed water districts. Current Residents. Figure 1 displays the relationship between cumulative years of residence in the six water districts and the natural logarithm of serum PFOA (ng/mL), in individuals reporting residence in one of the six water districts on the date of interview in 20052006 (current residents). The positive slope is significant at the p<0.001 level for all six districts. The effect of cumulative years is reasonably linear with ln(PFOA). The results of the full model after backward selection, with an indicator variable for water district, are shown in Table 2. The R-squared for the full model was 0.68. Water district residence explained the majority of the variance (partial R-squared), with residence in Little Hocking alone accounting for 39.4%. After residence, cumulative years of residence explained 1.5% of the variance. Previously observed associations were also replicated: higher levels in males, a U-shaped relationship with age, higher levels in current vs. never smokers, and higher levels in those who grow their own vegetables (Steenland et al. 2009). The average increase in PFOA levels for each year of residence in a water district was 1.2% (95% Cl: 1.1-1.4%). However, because exposure levels are known to be different between 15 p. 17 Page 16 of 33 water districts, and because median serum PFOA levels differed so greatly by water district, we fit models for each water district separately to yield district-specific effects of cumulative residence (Table 3; F-test for six interaction terms significant at pcO.OOl). As expected, districts with the highest exposure levels display the largest relationship between years of residential history and serum PFOA. In Pomeroy and Mason County, the districts with the lowest exposures as measured in current residents, the effect of years of residence was least. Former residents. Using a two-segment linear spline regression with the same variables as above, we obtained estimates for the effect of years elapsed since residence on ln(PFOA) for the two segments of the spline curve. Based on visual inspection of a LOESS curve and goodness-of-fit statistics comparing various possible knots, we chose four years as the knot for Little Hocking and nine years as the knot for Lubeck. Figures 2a and 2b show the plots and fitted lines for the two water districts. An F-test for reduction in model error after moving from 1 segment (standard linear regression) to 2 segments was significant at pcO.OOl for both water districts, and overall model fit (R-square) increased by 4% in Little Hocking and 3% in Lubeck after inclusion of the spline. A three-segment linear spline was not a significantly better fit to the data in either water district. The estimated half-lives and percent change in PFOA by year for the two line segments in each water district are shown in Table 4. Because the half-lives are calculated from the slope, they represent the half-life that would result if the instantaneous rate of clearance were to continue indefinitely. The shallower line after the knot likely reflects either the gradual slowing of PFOA clearance over time (Little Hocking) and/or the decline of low exposures to near background (15 ppb) in the case of Lubeck. For example, an individual from Little Hocking with an initial serum PFOA of 55 ng/mL would have a concentration of 21 ng/mL after four years 16 Page 17 of 33 p. 18 (21% reduction per year), and a concentration of 15 ng/mL four years later (8% reduction per year). The resultant half-lives for Little Hocking were 2.9 and 10.1 years for the two spline segments, while for Lubeck they were 8.5 years in the initial spline segment and undefined for the second segment (the estimated parameter was 0.0021, or approximately 0, indicating no further decrease in PFOA levels over time). Our estimated half-lives were sensitive to the truncation cutpoint we used, below which subjects were excluded on the basis that they were near background levels. Table 5 displays various half-life estimates for the first spline segment in Little Hocking and Lubeck after various truncation values were applied. Little Hocking half-life estimates ranged from 2.5 to 3.0 years, while in Lubeck estimates ranged from 5.9 to 10.3 years. At all truncation values the half-life in Little Hocking was lower than in Lubeck, with larger discrepancies at higher truncation values. Because former Lubeck residents had lower serum PFOA concentrations, more individuals were discarded from Lubeck at all truncation values, with the discrepancy larger at higher values. Our estimated half-lives were also sensitive to the amount we subtracted off of our PFOA levels, a subtraction designed to eliminate background levels in estimating half-life. To test the robustness of our estimate to this subtraction, we performed a similar analysis that eliminated all individuals below 15 ng/mL in 2005, and subtracted 15, rather than 5, from their measured value. Results were similar to our original results: the estimated half-life for the first four years of clearance in former residents of Little Hocking was 3.0 years, while for Lubeck (for the first nine years) it was 9.4 years. PFOA clearance appears to be sex-dependent in rats (with a much longer half-life in males), but not in monkeys (Lau et al. 2007). In our data for Little Hocking, males were associated with a faster rate of clearance (p=0.02), but only in the first four years. Annual 17 p. 19 Page 18 of 33 reduction in serum PFOA was 27% in males vs. 18% in females. The effect was non-significant after four years. However, prior longitudinal analyses of 200 residents in Little Hocking and Lubeck found no sex differences in half-life (Bartell et al. 2009). We did not observe sex differences in former residents of Lubeck. 18 Page 19 of 33 p. 20 Discussion We found a significant positive association between years of residence in an exposed water district and serum PFOA, with an average of l % increase per year of residence. Lower levels of serum PFOA in former vs. current residents residence has been demonstrated in this cohort previously (Steenland et al. 2009), but this analysis now demonstrates a significant trend within current residents (those still residing in exposed water districts in 2005-2006 based on their prior residential history). We also found a more substantial relationship between PFOA and years of residence in water districts closer to the industrial facility, as expected. After water district, years of residence accounted for the greatest variance in the fitted model. These findings provide preliminary justification for possible use of residential history as a proxy for prior exposure in epidemiologic studies. In former residents the main finding from our analysis was that the use of a two-segment spline increased the model fit and better approximated the observed relationship than a simple linear model. In both water districts, an apparent nonlinear relationship resulted in a significantly lower clearance rate after the knot of either 4 or 9 years. If our assumptions are correct this implies that a simple first order elimination model may not hold, and that the rate of elimination may be concentration-dependent or time-dependent. We feel that the results suggest both a concentration- and time-dependent relationship because the time factor is the same for both Little Hocking and Lubeck (years since former residence), but exposure was lower in Lubeck. However, the apparent time-dependent relationship could also be due to the concentration decrease over time. It is interesting that the rate of decay (slope) of the second linear segment for Little Hocking is similar to the rate of decay for the first segment for Lubeck, at similar concentration levels. In our cohort, former residents of Little Hocking had PFOA levels roughly 19 p. 21 Page 20 of 33 twice as high former residents of Lubeck. If serum clearance were concentration-independent the equation describing the relationship between PFOA and years living in the district and years elapsed since living in district would be the same in Little Hocking and Lubeck. Furthermore, within each water district the decay in In(PFOA) would be linear rather than exhibiting a lower slope at lower concentrations. As in prior studies of this population, we observed decreasing serum concentrations across dates of testing (Steenland et al. 2009). This may be due to behavior modification as the putative health effects of PFOA became publicized, both in increased bottled water usage and decreased tap water consumption. We observed a slight increase in reported bottled water usage over the testing period, and Little Hocking was offering free bottled water to individuals. Additionally, it is plausible that those who tested earliest were those who lived closer to the industrial facility and in more highly exposed water districts. However, adjusting for date of testing did not significantly alter any of our parameters of interest. Prior studies in humans have found no difference in clearance rates between men and women, but animal studies have suggested that females may be more effective clearers of PFOA (Bartell et al. 2009; Brede et al. 2010; Lau et al. 2007). Harada et al. demonstrated in moderately exposed city-dwellers that renal excretion rates in both males and females were negligibly small, but that female clearance may be age-dependent (Harada et al. 2005). In our cohort we observed lower PFOA levels in females, an observation consistent with prior studies in this cohort and others (Calafat et al. 2007; Steenland et al. 2009). However, we observed a significantly faster clearance rate in men in the initial years of former Little Hocking residents. This calls into question the assumption that lower levels in females are due to faster rates of clearance, but we cannot rule out that the apparent sex effect is due to concentration. 20 Page 21 of 33 p. 22 This study has three major limitations. The first is the cross-sectional nature of the analysis. Particularly in the estimation of half-life, this limited our ability to draw inferences from the analysis. Although cross-sectional half-life estimation has been used in an analogous setting for urinary bisphenol A after fasting (Stahlhut et al. 2009), traditional half-life studies follow individuals over time, allowing researchers to compare serum concentration at any point in time to the initial concentration. Cross-sectional analyses must rely on model-based estimation of the initial concentrations instead of directly observed values. Our regression model included years of residence in the contaminated water district, sex, age, growing own vegetables, smoking, and consuming bottled water. We relied on recall via questionnaire to develop prior residential history. In addition to missing and incomplete data (gaps in residential history, which led to the exclusion of some subjects from the analysis), there is the possibility that individuals misreported their water district and/or years of residence. The second major limitation is the implied assumption that exposure was uniform within a water district, both between individuals and over time, which we know to be false. Although we excluded individuals who were employed by DuPont or who reported private well use to limit the heterogeneity of the population, individual exposure was undoubtedly varied based on geographical location, individual behavior, and other uncontrollable factors. Also, we know that PFOA emissions from the plant were not constant over time and peaked in the late 1990s, but we were unable to account for this without quantitative estimates of annual water system concentrations. Further studies of this population will make use of advanced exposure models that account for both individual and temporal variations in exposure. A third major limitation of our analysis is the potential bias introduced by the exclusion of participants with serum levels below 15 ng/mL. Truncation below a fixed concentration 21 p. 23 Page 22 of 33 threshold is known to introduce bias in half-life estimates for longitudinal data (Michalek et al. 1998), and is likely to have a similar effect in cross-sectional analyses. Although restricting the analysis to individuals with PFOA serum concentrations below 15 ng/mL avoids one type of bias (overestimation of half-lives among participants whose PFOA serum concentrations are no longer in decline by the time of the serum sample), it is likely to introduce another type of bias resulting in overestimation of half-lives, because excluded participants are likely to have shorter half-lives on average than retained participants. Our sensitivity analysis using different truncation values resulted in a smaller range of values for the more highly exposed residents of Little Hocking, while the half-life in former Lubeck residents was more sensitive to the truncation value. Notably, Lubeck residents tended to have lower concentrations, so truncation at all values resulted in more individuals discarded from the Lubeck analysis, with a progressively larger difference at higher truncation values. A minor limitation of this study was the inability to differentiate between variable exposure levels and accumulation due to constant exposure. However, because emission levels and predicted water concentrations were known to be variable over the study period, peaking in the late 1990s, we feel that some of the annual increase as shown by the significance of years of residence is likely due to increasing exposures, rather than approach to a steady-state (Paustenbach et al. 2007). Further work will be done with exposure estimates that vary by year and location of residence. These results suggest that the half-life for PFOA lies between the previously reported estimates of 2.3 and 3.8 years for more highly exposed individuals, but that serum clearance of PFOA may be concentration-dependent. 22 Page 23 of 33 p. 24 References Bartell SM, Calafat AM, Lyu C, Kato K, Ryan PB, Steenland K. 2009. Rate of Decline in Serum PFOA Concentrations after Granular Activated Carbon Filtration at Two Public Water Systems in Ohio and West Virginia. Environ Health Perspect 118(2):222-228, Brede E, Wilhelm M, Gen T, Mller J, Rauchfuss K, Kraft M, et al. 2010. Two-year follow-up biomonitoring pilot study of residents'and controls' PFC plasma levels after PFOA reduction in public water system in Arnsberg, Germany. International Journal of Hygiene and Environmental Health 213:217-223. Calafat A, Wong L, Kuklenyik Z, Reidy J, Needham L. 2007. Polyfluoroalkyl chemicals in the US population: data from the National Health and Nutrition Examination Survey (NHANES) 2003-2004 and comparisons with NHANES 1999-2000. Environmental Health Perspectives 115(11): 1596-1602. DeWitt J, Shnyra A, Badr M, Loveless S, Hoban D, Frame S, et al. 2009. Immunotoxicity of perfluorooctanoic acid and perfluorooctane sulfonate and the role of peroxisome proliferator-activated receptor alpha. Critical reviews in toxicology 39(l):76-94. Frisbee S, Brooks Jr A, Maher A, Flensborg P, Arnold S, Fletcher T, et al. 2009. The C8 Health Project: design, methods, and participants. Environ Health Perspect 117(12): 1873-1882. Harada K, Inoue K, Morikawa A, Yoshinaga T, Saito N, Koizumi A. 2005. Renal clearance of perfluorooctane sulfonate and perfluorooctanoate in humans and their species-specific excretion. Environmental Research 99(2):253-26l. Hines E, White S, Stanko J, Gibbs-Floumoy E, Lau C, Fenton S. 2009. Phenotypic dichotomy following developmental exposure to perfluorooctanoic acid (PFOA) in female CD-I mice: Low doses induce elevated serum leptin and insulin, and overweight in mid-life. Molecular and Cellular Endocrinology 304:97-105. Houde M, Martin J, Letcher R, Solomon K, Muir D. 2006. Biological monitoring of polyfluoroalkyl substances: a review. Environ Sei Technol 40(11):3463-3473. Kennedy G, Butenhoff J, Olsen G, OConnor J, Seacat A, Perkins R, et al. 2004. The toxicology of perfluorooctanoate. Critical reviews in toxicology 34(4):351-384. Lau C, Anitole K, Hodes C, Lai D, Pfahles-Hutchens A, Seed J. 2007. Perfluoroalkyl acids: a review of monitoring and toxicological findings. Toxicological Sciences 99(2):366-394. Lundin J, Alexander B, Olsen G, Church T. 2009. Ammonium Perfluorooctanoate Production and Occupational Mortality. Epidemiology 20(6):921. Michalek J, Tripathi R, Kulkarni P, Gupta P, Selvavel K. 1998. Correction for bias introduced by truncation in pharmacokinetic studies of environmental contaminants. Environmetrics 9(2): 165-174. Olsen G, Burris J, Ehresman D, Froehlich J, Seacat A, Butenhoff J, et al. 2007. Half-life of serum elimination of perfluorooctanesulfonate, perfluorohexanesulfonate, and perfluorooctanoate in retired fluorochemical production workers. Environmental Health Perspectives 115(9): 1298. Olsen G, Zobel L. 2007. Assessment of lipid, hepatic, and thyroid parameters with serum perfluorooctanoate (PFOA) concentrations in fluorochemical production workers. International Archives of Occupational and Environmental Health 81(2):231-246. Paustenbach D, Panko J, Scott P, Unice K. 2007. A methodology for estimating human exposure to perfluorooctanoic acid (PFOA): A retrospective exposure assessment of a community (1951-2003). Journal of Toxicology and Environmental Health Part A 70(l):28-57. 23 p. 25 Page 24 of 33 Rosen M, Lau C, Corton J. 2009. Does Exposure to Perfluoroalkyl Acids Present a Risk to Human Health? Toxicological Sciences 111(1): 1-3. Stahlhut R, Welshons W, Swan S. 2009. Bisphenol A data in NHANES suggest longer than expected half-life, substantial nonfood exposure, or both. Environmental Health Perspectives 117(5):784-789. Steenland K, Deddens JA. 2004. A practical guide to dose-response analyses and risk assessment in occupational epidemiology. Epidemiology 15(l):63-70. Steenland K, Jin C, MacNeil J, Lally C, Ducatman A, Vieira V, et al. 2009. Predictors of PFOA levels in a community surrounding a chemical plant. Environ Health Perspect 117(7): 1083-1088. Tan Y-M, Clewell lii HJ, Andersen ME. 2008. Time dependencies in perfluorooctylacids disposition in rat and monkeys: A kinetic analysis. Toxicology Letters 177(l):38-47. 24 Page 25 of 3S p. 26 Tables Table 1. Median PFOA (ng/mL) in 2005-2006 (N) for current and former residents. Variable Current Former11 Variable Current Former Total 33.0(18,068) 36.5(1,672) Drink Bottled Water Yes 55.5 (532) 77.7 (70) Water District No 32.9(17,031) 36.2(1,518) Belpre 31.0(1,999) -- Little Hocking 241.0 (3,154) 60.6 (643) Grow Own Vegetables Lubeck 69.4(3,131) 31.0(1,029) Yes 38.3 (4,885) 39.4 (228) Mason County 12.4 (5,052) No 31.5(13,183) 35.9(1,444) Pomeroy 11.8 (640) -- Tuppers Plains 36.4 (4,092) Vegetarian Yes 37.1 (168) 38.0(13) Sex No 33.0(17,900) 36.5 (1,659) F 31.0 (9,330) 35.1 (872) M 34.9 (8,738) 37.3 (800) Currently Consume Alcohol Yes 37.1 (6,108) 35.8 (845) Race No 30.6(11,216) 36.8 (784) White 32.9(17,579) 36.5(1,637) non-White 35.2 (489) 35.1 (35) Smoker Current 28.1 (3,473) 35.5(353) BMI Former 38.0 (3,968) 37.7 (364) <24 32.3 (6,328) 37.1 (575) Never 33.1 (10,627) 36.6 (949) 24-26 35.8 (3,454) 37.5 (354) 27-29 34.8 (3,099) 36.9(315) Date of Testing11 >30 31.1(5,187) 33.6(428) First two months 59.7 (2,068) 39.2(134) Second two months 51.8(2,377) 37.8 (218) Regular Exercise Third two months 34.0 (5,389) 34.8 (514) Yes 36.8 (5,794) 36.6 (645) Fourth two months 28.7 (4,537) 35.3(481) No 31.2(12,274) 36.4(1,027) Fifth two months 21.5(2,431) 40.1 (147) Last two months 17.0(1,266) 38.0(178) "Former residents limited to individuals in Litt e Hocking and Lubeck with >2 years residence and >15 ng/mL PFOA. bDates of testing: 8/1/05-9/30/05, 10/1/05-11/30/05, 12/31/05-1/31/06,2/1/06-3/31/06,4/1/06- 5/31/06,6/1/06-8/31/06 25 p. 27 Page 26 of 33 Table 2. Multivariate linear regression results* (R2=0.68), current residents (n= 17,516b). Variable Predicted change in PFOA (% from referent) Log change in PFOA 95% Cl Cumulative years of residence 1% 0.012 0.011 0.014 Sex, Female -12% -0.133 -0.153 -0.112 Age <20 Referent -- ---- 20-29 -23% -0.261 -0.302 -0.220 30-39 -12% -0.126 -0.169 -0.083 40-49 -1% -0.006 -0.045 0.033 50-59 4% 0.042 0.004 0.080 60-69 18% 0.167 0.126 0.208 >70 27% 0.236 0.192 0.279 Grow vegetables 11% 0.106 0.083 0.129 Smoking Never Referent -- ---- Current 12% 0 117 0.088 0.146 Former 1% 0.012 -0.016 0.039 Bottled water -26% -0.301 -0.361 -0.241 Water district Tuppers Plains Referent - ---- Belpre -12% -0.129 -0.166 -0.091 Little Hocking 495% 1.783 1.750 1.815 Lubeck 82% 0.600 0.566. . 0.634 Mason County -64% -1.018 -1.047 -0.989 Pomeroy -67% -1.102 -1.161 -1.043 "Model also adjusted for date of visit b552 individuals missing covariate data (Bottled water=505, Smoking=66). Variance (%) in In(PFOA) (partial R2) 1.5% 0.9% -- 0.9% 0.2% 0.0% 0.0% 0.4% 0.6% 0.4% -- 0.4% 0.0% 0.5% -- 0.3% 39.4% 6.5% 21.1% 7.1% 26 Page 27 of 33 p. 28 Table 3. Effect of years of residence on serum PFOA by water district, with district-specific model fit. Water District Tuppers Plains Belpre Little Hocking Lubeck Mason County Pomeroy N Model RJ 3,986 1,940 3,054 3,044 4,885 607 0.18 0.10 0.10 0.23 0.08 0.12 Variance (%) in In(PFOA) (partial R2) explained by years of residence 3.2% 0.6% 0.8% 3.6% 0.6% 0.5% % Change in Predicted PFOA by Year of Residence 1.7% 0.7% 1.2% 1.9% 0.6% 0.5% 95% Cl 1.4% 0.3% 0.7% 1.6% 0.4% -0.1% 2.0% 1.1% 1.7% 2.3% 0.9% 1.1% 27 p. 29 Page 28 of 33 Table 4. Multivariate linear regression results, former residents of Little Hocking (n=602a) and Ltibeck (n=971b). Variable Estimated Half-Life (years) % Change in PFOA by Year 95% Cl Little Hocking Years Elapsed, <4 2.9 -21.4% -26.1% Years Elapsed, >4 10.1 -7.6% -18.1% Years of Residence -- 1.9% 0.8% Lubeck Years Elapsed, <9 8.5 -7.8% -9.1% Years Elapsed, >9 n.a.c 0.2% -3.3% Years of Residence - 2.5% 1.8% "Models also adjusted for sex, age, growing own vegetables, smoking, and consuming bottled water. bFinal analysis numbers due to missingness in smoking history and consumption of bottled water. cParameter (0-002) yields a positive half-life not significantly greater than zero. -16.5% 6.4% 3.0% -6.5% 3.8% 3.1% 28 Page 29 o f 33 p. 30 Table 5. Sensitivity analysis for half-life after various truncation cut points of serum PFOA. Value (ng/mL) 20 15 10 5 Half-life, years (95% CI)a Little Hocking Lubeck 3.0 (2.4-4.0) 10.3 (8.713.1) 2.9 (2.3-3.8) 8.5(7.1-10.1) 2.5 (2.0-3.3) 6.6 (5.8-7.8) 2.7 (2.1-3.9) 5.9 (5.1-7.1) "Models also adjusted for sex, age, growing own vegetables, smoking, and consuming bottled water 29 p. 31 Page 30 of 33 Figure Legends Figure 1. Plots of natural logarithm of PFOA (ng/mL) by cumulative years of residence in a water district, current residents, LOESS regression. Figure 2a. Predicted decay of serum PFOA concentration based on half-lives estimated from former Little Hocking residents in discrete segments of less than 4 and greater than 4 years since living in Little Hocking (solid line; adjusted for covariates), and LOESS regression (dashed line; unadjusted for covariates). Figure 2b. Predicted decay of serum PFOA concentration based on half-lives estimated from former Lubeck residents in discrete segments of less than 9 and greater than 9 years since living in Lubeck (solid line; adjusted for covariates), and LOESS regression (dashed line; unadjusted for covariates). 30 Page 31 of 33 Pom eroy Mason County Belpre < OUa- 2- Tuppers Plains Little Hocking ^ i 1 I i I T - i i i i i i r *T-- i-- i-- i-- i-- r C 5 1C 15 20 25 3C 0 -3 1C 15 20 25 30 C. 5 1C 13 20 25 30 Years of Residence 164x130mm (300 x 300 DPI) p. 33 Page 32 of 33 165x 134mm (300 x 300 DPI) In(PFOA) Page 33 of 33 Years Elapsed 168x135mm (300 x 300 DPI) p. 35 0 9 / 3 0 / 2 0 1 0 1 0 : 3 7 FAX p. 36 1^1005 Status report Patterns o f age o f puberty am ong children in the M id-O hio V alley in relation to Perflu orooctanoic Acid (PFOA) and Ferfluorooctane Sulfonate (PFO S) The C8 Science Panel (Tany Fletcher, K yle Steenland, D avid Savitz) Sept 30 2010 This status report summarizes the findings o f a statistical analysis o f the relationship between levels o f periluorooctanoic add (PFOA, also called C 8), and perfluorooctane sulfonate (PFOS) measured in the blood serum o f the children who participated in the C8 Health Project, and puberty. A full report o f these findings w ill be submitted to a peer-reviewed scientific journal. Introduction. It has been suggested that som e polyfiuoroalkyl compounds (PFCs) may alter animal sexual maturation. The aim o f this study was to examine fire relationship between levels o f two PFCs - PFOA (or C8) and PFOS - with puberty based on sex hormone levels and selfreported onset o f menstruation. We used data from file CR Health Project supplemented with detailed date o f birth available from those who consented to be in the Science Panel studies. M ethods. Among the young participants aged 8-18 at the time o f the C8 Health Project survey (20052006), w e examined data for 3076 boys and 2931 girls, all residents for at least a year in the six water districts which had been contaminated with PFOA. They were classified as having reached puberty at file tim e o f interview based on either sex hormone Mood levels (testosterone >50 ng/dL or free testosterone >5 pg/mL for boys, and estradiol >20 pg/mL for girls), or having reported that they had started menarche (periods). Statistical models estimated the chance o f reaching puberty in relation to PFOA, and PFOS levels, while controlling for other potential explanatory factors. From these m odels, w e could also estimate the average age o f reaching puberty for children with different exposure levels, and present the difference (earlier or later) in days o f reaching puberty between different exposure groups. We divided the population into four equal groups (quartiles) by exposure to PFOS and PFOA. Results. 0 9 /3 0 /2 0 1 0 1 0 :3 7 FAX p. 37 006 The mid-point o f PFOA and PFOS serum levels were 26 and 20 ng/mL in boys, and 20 and 18 ng/mL in girls. For boys, there was a clear relationship o f reduced odds o f having readied puberty with increasing PFOS (delay o f 190 days between the highest and low est quartile), but not PFOA. For girls, higher exposure to either PFOA or PFOS was associated with reduced odds o f having reached puberty. The highest PFOA group had an average age o f puberty 130 days later than the low est exposure group, and for PFOS, the delay was estimated as 138 days comparing the highest and low est exposure group. These results are consistent in directum and magnitude with one published study which suggested delayed puberty (also measured as self-reported menarche) in relation to PFOS exposure in girls, and in contrast to another study which repotted younger puberty (measured as breast maturation) in girls in relation to PFOA exposure. C onclusions. Delays o f puberty have been observed in this population correlated with PFOS in boys and PFOA and PFOS exposure in girls. Caution is needed in interpreting these results, due to the feet that Mood PFC levels and puberty status based on sex hormone levels were determined at the same time, and menarche was self-reported. For exam ple, it may be that growth changes associated with puberty lead to changes in PFOA and PFOS blood levels, rather ban these compounds having any effect on age at puberty. Further work is planned to investigate patterns o f puberty by age in relation to exposure prior to puberty.