Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

Relationship of Perfluorooctanoic Acid Exposure to Pregnancy Outcome Based on Birth Records in the Mid-Ohio Valley

Relationship of Perfluorooctanoic Acid Exposure to Pregnancy Outcome Based on Birth Records in... Research Children’s Health Relationship of Perfluorooctanoic Acid Exposure to Pregnancy Outcome Based on Birth Records in the Mid-Ohio Valley 1 2 1 1 3 4 David A. Savitz, Cheryl R. Stein, Beth Elston, Gregory A. Wellenius, Scott M. Bartell, Hyeong-Moo Shin, 5 6 Veronica M. Vieira, and Tony Fletcher 1 2 Department of Epidemiology, Brown University, Providence, Rhode Island, USA; Department of Preventive Medicine, Mount Sinai 3 4 School of Medicine, New York, New York, USA; Program in Public Health, and School of Social Ecology, University of California, Irvine, Irvine, California, USA; Department of Environmental Health, Boston University School of Public Health, Boston, Massachusetts, USA; London School of Hygiene and Tropical Medicine, London, United Kingdom the region. The first provides comprehen- Background : Perfluorooctanoic acid (PFOA) is a potential cause of adverse pregnancy outcomes, sive coverage of the study area, but is subject but previous studies have been limited by low exposures and small study size. to some exposure misclassification. The sec - o Bjectives : Using birth certificate information, we examined the relation between estimated PFOA ond sacrifices some of the sample size, but exposure and birth outcomes in an area of West Virginia and Ohio whose drinking water was con- has improved exposure measurement. Both taminated by a chemical plant. studies rely on birth certificates for assess- Methods : Births in the study area from 1990 through 2004 were examined to generate case groups ment of health end points and use the same of stillbirth (n = 106), pregnancy-induced hypertension (n = 224), preterm birth (n = 3,613), term environmental/pharmacokinetic models to low birth weight (n = 918), term small-for-gestational-age (SGA) (n = 353), and a continuous estimate PFOA exposure. measure of birth weight among a sample of term births (n = 4,534). A 10% sample of term births ≥ 2,500 g were selected as a source of controls (n = 3,616). Historical estimates of serum PFOA Study I: PFOA and were derived from a previously developed fate and transport model. In a second study, we examined Pregnancy Outcome Based 4,547 area births linked to a survey with residential history data. on Birth Records results : In the analysis based only on birth records, we found no consistent evidence of an asso- ciation between estimated PFOA exposure and stillbirth, pregnancy-induced hypertension, preterm Methods birth, or indices of fetal growth. In the analysis of birth records linked to the survey, PFOA was The study area in the mid-Ohio Valley was unrelated to pregnancy-induced hypertension or preterm birth but showed some suggestion of an chosen to include PFOA-contaminated public association with early preterm birth. Measures of growth restriction showed weak and inconsistent associations with PFOA. water utilities in Mason and Wood counties in West Virginia and in Athens, Meigs, and c onclusions : Based on the analysis using the health survey, these results provide little support for Washington counties in Ohio, where levels an effect of PFOA exposure on most pregnancy outcomes, except for early preterm birth and pos - ranged from near U.S. averages to orders of sibly fetal growth restriction. magnitude above typical levels (Frisbee et al. k ey words : fetal growth restriction, perfluorooctanoic acid, pregnancy, pregnancy-induced hyper - 2009). We obtained computerized live birth tension, preterm birth, stillbirth. Environ Health Perspect 120:1201–1207 (2012). http://dx.doi. org/10.1289/ehp.1104752 [Online 26 March 2012] records with geographic identifiers from the Ohio and West Virginia health departments, and fetal death records from West Virginia Adverse effects of perfluorooctanoic acid the potential association between PFOA and (PFOA) and related perfluoroalkyl acids on pregnancy outcomes is warranted. Address correspondence to D.A. Savitz, Department pregnancy and development have been sug- A large population in Ohio and West of Epidemiology, Box G-S-121-2, Brown University, Providence, RI 02912 USA. Telephone: (401) 863- gested based on toxicology (Lau et al. 2007) Virginia has been exposed to PFOA predomi- 6090. Fax: (401) 863-3713. E-mail: david_savitz@ and a small but growing epidemiologic lit- nantly through industrial contamination of brown.edu erature (Steenland et al. 2010). The stron- drinking-water supplies (Frisbee et al. 2009). Supplemental Material is available online (http:// gest support from laboratory studies suggests Within this population, PFOA exposure var- dx.doi.org/10.1289/ehp.1104752). the potential for reductions in fetal growth ies considerably across time and place because The authors thank K. Steenland, the Ohio and (Lau et al. 2007). Epidemiologic findings of different levels of PFOA contamination in West Virginia Health Departments, and the West Virginia Birth Score Program for assistance. on measures of reduced growth are mixed the drinking water, depending on the year This research was funded by the C8 class action (Apelberg et al. 2007; Fei et al. 2007; Hamm and the water district, enabling accurate esti- settlement agreement [Jack W. Leach, et al. v. E.I. et al. 2010; Nolan et al. 2009; Washino et al. mation of serum levels by linking residen- du Pont de Nemours & Company (no. 01-C-608 2009), but generally consistent with a small tial histories and historical drinking-water W.Va., Wood County Circuit Court, West Virginia, decrement in birth weight associated with concentrations. The resulting range of PFOA USA)] between DuPont and plaintiffs. Funds were higher levels of serum PFOA. Duration of levels are well above background exposure administered by the Garden City Group (Melville, New York) that reports to the court. Our work and gestation has not been related to PFOA in (Steenland et al. 2009), which may allow for conclusions are independent of either party to the these same studies (Apelberg et al. 2007; a more informative assessment of the impact lawsuit. C.S. was supported by grant K01 ES019156 Fei et al. 2007; Hamm et al. 2010). More of PFOA exposure on birth outcomes. from the National Institute of Environmental Health extreme, clinically consequential end points For participants in the C8 Health Sciences (NIEHS). G.W. was supported by grant of fetal growth restriction and preterm birth Project—a survey conducted in the contami- R00-ES015774 from NIEHS. have been examined less extensively and often nated region (Frisbee et al. 2009)—we pre- The contents of this report are solely the responsi - bility of the authors and do not necessarily represent with limited statistical power; none of the viously analyzed a small number of births the official views of NIEHS or National Institutes human studies support an association (Fei using serum PFOA measurements (Stein of Health. et al. 2007; Hamm et al. 2010; Nolan et al. et al. 2009) and a much larger number using e Th authors declare they have no actual or potential 2009). Given ubiquitous exposure, sugges- modeling to assign estimated serum PFOA competing financial interests. tive toxicologic data, and limited epidemi- levels (Savitz et al. 2012). In this report, we Received 16 November 2011; accepted 26 March ologic research, continued examination of used two approaches to evaluate births in 2012. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1201 Savitz et al. only (unpublished data). We restricted the the mother estimated for the early pregnancy each imputed data set were pooled into a final analysis to pregnancies that occurred from period. The methods for estimating individual parameter estimate and variance (Rubin 1987). 1990 through 2004 because that is the period serum PFOA concentrations are described in Missing covariate data on maternal age (n = 2), when the contrasts were highest between detail elsewhere (Shin et al. 2011a, 2011b). maternal education (n = 61), parity (n = 6), exposures resulting from contaminant releases Briefly, information on plant operations and and tobacco use (n = 64) were also imputed. and background sources (Shin et al. 2011a). chemical releases was combined with environ- The primary analyses incorporate the exposure Because of the expense involved with refin - mental characteristics of the region through estimates for the street-level geocode (66%) ing address information and geocoding mater- a series of linked models to estimate air and and the imputed exposure estimate when only nal residences, which is the basis for assigning water concentrations of PFOA from 1951 ZIP code geocodes could be assigned (34%). exposure, we used case–control sampling to through 2008 (Shin et al. 2011a). Geocodes In sensitivity analyses we considered the impact enhance precision for some of the birth out- for birth residences were used to determine of restricting to births with exposure estimates comes of interest. We restricted eligibility to whether drinking water in participants’ homes based on street-level geocodes only and no singleton births and excluded the small pro- was supplied by public water sources or pri- missing covariate data. portion (3%) of nonwhite births because vate wells. When addresses could be geocoded Estimated serum PFOA concentrations there were too few to analyze separately. only at the ZIP code level, an average exposure were generated by calendar year, with the rele- Comprehensively identified case groups con - measure was calculated based on the propor- vant exposure year for each pregnancy consid- sisted of stillbirths (fetal death ≥ 20 weeks com- tion of the ZIP code population that was sup- ered to be month 3 of gestation for stillbirths pleted gestation), preterm births (< 37 weeks plied by contaminated drinking water. With and month 4 for live births. The estimated completed gestation), and term low birth these estimated environmental levels of PFOA serum PFOA level at the time of pregnancy weight births (≥ 37 weeks completed gestation based on the geocoded maternal address, indi- was skewed, so we considered serum PFOA and birth weight < 2,500 g). A 10% sample of vidual maternal serum levels for each preg- estimates as a continuous log-transformed term births ≥ 2,500 g (n = 3,616) was selected nancy were estimated using age- and sex-based measure, a continuous untransformed mea- and geocoded for assigning PFOA exposure. pharmacokinetic modeling with standard sure, and in quintiles of exposure. We aggre- To efficiently select control groups and assumptions about water intake, body weights, gated the lowest two quintiles as the reference expand the range of birth outcomes, we first and a PFOA half-life (Shin et al. 2011b). category for analysis because those quintiles constructed a stratified 10% random sample Because we had only the residence listed covered a modest range typical of U.S. serum of the entire population of live births by com- on the birth certificate for the mother—not levels and our exposure estimates are best able bining the 10% sample of all term live births a residential history—and serum levels at a to differentiate individual exposure at higher ≥ 2500 g with a 10% random sample of the given time depend on exposure history, we levels resulting from consumption of locally comprehensively identified preterm cases and assumed that she had lived in this location for contaminated water. term low birth weight cases (the case groups the previous 6 years, which would account for We adjusted all analyses for maternal age described above) (n = 4063). Pregnancy- approximately 63% of her serum concentra- (splines with 3 degrees of freedom), education induced hypertension cases, term small-for- tion given a stable exposure rate and an esti- (< 12, 12, 13–15, ≥ 16 years), parity (0, 1, gestational-age (SGA) cases, and term birth mated serum half-life of PFOA of 3.5 years ≥ 2), tobacco use (smoker, nonsmoker), expo- weight births were identified exclusively from (Olsen et al. 2007). Moving within the same sure year (splines with 4 degrees of freedom), the 10% random sample of births. Control public water supply district would have little and state of residence (Ohio, West Virginia). groups for each of the outcomes were derived or no impact on the estimated serum values, In addition, we included gestational age as noncase subsets from the 10% random but moving across water districts or especially (37, 38, 39, 40, 41, ≥ 42 completed weeks) sample of the entire population of live births. from more distant locations would lead to as a covariate in the analysis of birth weight We conducted a series of case–control exposure misclassification. among term births. analyses: a) stillbirths (n = 106) compared with We were able to geocode the maternal For the binary outcomes, we used logistic West Virginia live births (n = 1,844) because address listed on the birth certificate to the regression to estimate the odds ratio (OR) and stillbirth information was available only from street level for 66% of births and to the ZIP 95% confidence interval (CI) associated with West Virginia; b) pregnancy-induced hyper- code level for the remaining 34%, with lit- PFOA exposure, using the log-transformed tension cases (n = 224) compared with births tle variation in these proportions by case or value to generate ORs for an interquartile without pregnancy-induced hypertension control status. We found a high correlation shift from the 25th to 75th percentile, and the (n = 3,828); c) preterm birth cases < 37 weeks between the exposure estimate using street untransformed variable used to generate ORs (n = 3,613) and < 32 weeks (n = 491) cor- level geocode and ZIP code averages (r = 0.84), per 100-ng/mL shift and for quintiles (com- responding to all preterm and very preterm indicating modest loss of information from paring the 3rd, 4th, and 5th with the aggre- births, respectively, compared with term births lack of street level assignment. To make opti- gated 1st and 2nd quintiles). Both crude and (n = 3,695), with assignment based on clinical mal use of the available data, we used the ZIP adjusted effect estimates were generated. We assessment of gestational age; d) term low birth code average exposure and other information used multiple linear regression to estimate the weight cases (n = 918) compared with term from the birth certificate (maternal educa- association between PFOA and birth weight, births ≥ 2,500 g (n = 3,616); and e) term SGA tion, maternal age, parity, tobacco use, expo- using the same measures of estimated PFOA cases [births < 10th percentile by gestational sure year, residential state) to impute a new exposure described for the categorical out- age and sex (Oken et al. 2003)] (n = 353) exposure estimate for the pregnancies with ZIP comes. All statistical analyses were conducted compared with term appropriate-for-gesta- code geocodes only (n = 2,780). We used mul- using SAS version 9.2 (SAS Institute Inc., tional-age births [births between the 10th and tiple imputation to generate 20 replications to Cary, NC) and R version 2.12.1 (R Project 90th percentiles, inclusive, by gestational age obtain accurate information on the variability for Statistical Computing, Vienna, Austria). and sex (Oken et al. 2003)] (n = 2,990). In in the imputed values (Graham et al. 2007). Results addition, we analyzed a continuous measure of Markov Chain Monte Carlo imputation was birth weight among term births (n = 4,534). used with 200 burn-in iterations and 100 itera- Births were nearly evenly divided by calendar The exposure metric used in the analy- tions between each imputation. The parameter period and by state (Table 1). Estimated expo- sis is modeled serum PFOA concentration of estimates and standard errors determined from sure to PFOA tended to be higher for births | | 1202 volume 120 number 8 August 2012 • Environmental Health Perspectives PFOA and pregnancy outcome in birth records that were more recent, among older mothers, on the untransformed measure (OR = 1.20 PFOA was not related to risk of term more highly educated mothers, nonsmokers, per 100 ng/mL; 95% CI: 0.86, 1.68), the lack low birth weight, with some suggestion of and West Virginia residents (Table 1). of gradient across quintiles does not support a reduced risk of SGA with higher exposure Stillbirth was unrelated to PFOA expo- an association. Pregnancy-induced hyperten- but no gradient across quintiles (Table 4). sure based on either continuous or categorical sion was unrelated to PFOA exposure regard- Birth weight among term births was largely exposure measures (Table 2). Although there less of metric (Table 2). Preterm birth was not unrelated to estimated PFOA (Table 4). was some suggestion of an increased risk based associated with PFOA (Table 3). There was a suggestion of a modest decrease Table 1. Study I: PFOA and pregnancy outcome based on birth records: characteristics of study population by pregnancy outcome, Mid-Ohio Valley, 1990–2004 [n (%)]. Estimated PFOA among Maternal Preterm all births (ng/mL) characteristic Stillbirth PIH < 37 weeks Term LBW SGA All births median (25th, 75th percentiles) Total 106 224 3,613 918 353 8,253 7.7 (4.9, 17.2) Exposure year 1990–1994 28 (26) 51 (23) 1,068 (30) 295 (32) 117 (33) 2,685 (33) 7.0 (4.5, 17.7) 1995–1999 49 (46) 84 (38) 1,170 (32) 312 (34) 124 (35) 2,776 (34) 7.9 (5.0, 16.4) 2000–2004 29 (27) 89 (40) 1,375 (38) 311 (34) 112 (32) 2,792 (34) 8.2 (5.0, 17.6) State of residence WV 88 (83) 111 (50) 1,470 (41) 401 (44) 157 (44) 3,610 (44) 10.2 (5.3, 19.8) OH 18 (17) 113 (50) 2,143 (59) 517 (56) 196 (56) 4,643 (56) 6.3 (4.6, 13.4) Age (years) < 20 18 (17) 29 (13) 602 (17) 186 (20) 75 (21) 1,330 (16) 6.7 (4.7, 13.5) 20–24 39 (37) 67 (30) 1,187 (33) 321 (35) 138 (39) 2,743 (33) 7.2 (4.9, 16.4) 25–29 23 (22) 69 (31) 949 (26) 209 (23) 77 (22) 2,185 (26) 8.4 (4.9, 18.9) 30–34 17 (16) 44 (20) 572 (16) 128 (14) 46 (13) 1,346 (16) 8.5 (4.9, 19.8) ≥ 35 9 (8) 15 (7) 303 (8) 74 (8) 17 (5) 649 (8) 8.4 (4.9, 19.7) Education (years) < 12 21 (20) 24 (11) 772 (21) 273 (30) 113 (32) 1,701 (21) 6.9 (4.8, 14.3) 12 51 (48) 94 (42) 1,518 (42) 389 (42) 134 (38) 3,435 (42) 7.5 (4.9, 17.3) 13–15 24 (23) 67 (30) 865 (24) 167 (18) 68 (19) 1,996 (24) 8.3 (4.9, 19.8) ≥ 16 10 (9) 39 (17) 458 (13) 89 (10) 38 (11) 1,121 (14) 8.3 (5.0, 17.2) Parity 0 49 (46) 137 (61) 1,710 (47) 457 (50) 178 (50) 3,801 (46) 7.6 (4.9, 16.7) 1 29 (27) 46 (21) 1,104 (31) 271 (30) 102 (29) 2,668 (32) 7.7 (4.9, 17.1) ≥ 2 28 (26) 41 (18) 799 (22) 190 (21) 73 (21) 1,784 (22) 7.6 (4.8, 18.2) Smoking status Smoker 27 (25) 32 (14) 1,039 (29) 430 (47) 167 (47) 2,364 (29) 7.0 (4.8, 15.1) Nonsmoker 79 (75) 192 (86) 2,574 (71) 488 (53) 186 (53) 5,889 (71) 8.0 (4.9, 18.1) Abbreviations: PIH, Pregnancy-induced hypertension; LBW, low birth weight. Some Ohio residents delivered stillbirths in West Virginia. Table 2. Study I: PFOA and pregnancy outcome based on birth records: association of PFOA with stillbirth and pregnancy-induced hypertension, Mid-Ohio Valley, 1990–2004. Stillbirth Pregnancy-induced hypertension a b b Live births Cases Crude Adjusted Live births Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) (n) OR OR (95% CI) IQR(lnPFOA) increase 1,844 106 0.92 1.00 (0.76, 1.32) 3,828 224 1.06 1.02 (0.86, 1.21) 100-ng/mL increase 1,844 106 1.01 1.20 (0.86, 1.68) 3,828 224 1.06 1.06 (0.86, 1.31) < 40th percentile (1.0 to < 6.1 ng/mL ) 504 35 1.0 1.0 1,527 82 1.0 1.0 40th to < 60th percentile (6.1 to < 10.2 ng/mL) 330 15 0.9 0.9 (0.4, 2.0) 764 48 1.2 1.0 (0.7, 1.6) 60th to < 80th percentile (10.2 to < 21.0 ng/mL) 487 31 1.1 1.0 (0.5, 1.7) 759 51 1.1 1.0 (0.6, 1.5) ≥ 80th percentile (21.0 to 717.6 ng/mL) 523 25 0.7 0.8 (0.5, 1.5) 778 43 1.1 1.0 (0.7, 1.5) a b c West Virginia only. Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.50]. The category boundaries are from the first imputed data set. Table 3. Study I: PFOA and pregnancy outcome based on birth records: association of PFOA with preterm birth, Mid-Ohio Valley, 1990–2004. < 37 weeks gestation < 32 weeks gestation a a Term births Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) OR OR (95% CI) IQR(lnPFOA) increase 3,695 3,613 0.99 1.02 (0.96, 1.08) 491 0.95 0.98 (0.86, 1.11) 100-ng/mL increase 3,695 3,613 1.01 1.02 (0.94, 1.10) 491 0.88 0.90 (0.74, 1.10) < 40th percentile (1.0 to < 6.1 ng/mL ) 1,460 1,467 1.0 1.0 197 1.0 1.0 40th to < 60th percentile (6.1 to < 10.2 ng/mL) 755 720 1.0 1.0 (0.8, 1.1) 108 1.0 1.0 (0.7, 1.4) 60th to < 80th percentile (10.2 to < 21.0 ng/mL) 733 714 1.0 1.0 (0.9, 1.2) 95 1.0 1.1 (0.8, 1.4) ≥ 80th percentile (21.0 to 717.6 ng/mL) 747 712 1.0 1.0 (0.9, 1.2) 91 0.9 1.0 (0.7, 1.3) a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.50]. The category boundaries are from the first imputed data set. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1203 Savitz et al. in birth weight with higher exposure based state of residence + mother’s address + mother’s Pregnancy-induced hypertension showed on continuous exposure measures {somewhat age + month and year of infant’s birth + infant essentially no association with estimated serum stronger for male births [see Supplemental sex. This process yielded 3,323 matched births. PFOA based on the uncalibrated estimates Material, Table 1 (http://dx.doi.org/10.1289/ We extended the matching through the West (Table 6). With Bayesian calibration, the log- ehp.1104752)]}, but no monotonic gradi- Virginia Birth Score project, a program of the transformed and categorical measures sug- ent across quintiles. In the sensitivity analysis West Virginia Department of Health (2012) gested a positive association, with an increased restricted to the 66% of births with street-level to identify infants at risk of developmental risk primarily in the 4th quintile. Traditional geocodes and complete covariate informa- problems. Using the mother’s Social Security calibration showed elevated risk in the 3rd and tion (see Supplemental Material, Tables 2–4), number as well as other birth certificate items, 4th quintiles, but the continuous exposure results were largely unchanged except for gen- Birth Score was able to identify an additional measures did not indicate an association. erating weak positive associations with both 1,224 matched births through 2004, yielding a Preterm birth < 37 weeks was weakly pregnancy-induced hypertension and late total of 4,547 birth certificates linked to the C8 related to PFOA levels based on continuous preterm births. Health Project for analysis. measures in the uncalibrated analyses, but We used the same environmental/ with no gradient across quintiles (Table 7). Study II: PFOA and pharmacokinetic models from study I to gen- Bayesian calibration yielded a similar pattern Pregnancy Outcome Based erate annual exposure estimates, except that and traditional calibration yielded no asso- on Birth Records Linked to the the link to the C8 Health Project enabled us ciation. Preterm birth < 32 weeks, with only to incorporate lifetime residential history, thus 40 cases, showed some suggestion of an associ- C8 Health Project reducing exposure misclassification (Shin et al. ation without calibration, enhanced somewhat Methods 2011a, 2011b). We examined the model-based with traditional calibration, and markedly The C8 Health Project enrolled > 69,000 resi- exposure estimates, a Bayesian time-dependent strengthened with Bayesian calibration. dents of the mid-Ohio Valley who were part of calibration in which the measured 2005–2006 Term low birth weight was largely unre- a class action lawsuit, administered a question- serum concentration was used for updat- lated to PFOA based on exposure measures naire eliciting health and residential history ing estimates (Savitz et al. 2012), and a tra- without calibration, except for an isolated information, and collected blood to measure ditional calibration method assuming that increased risk in the 4th quintile (Table 8). serum levels of perfluoroalkyl acids and clini - a higher-than-expected 2005–2006 serum With both calibration methods, the log- cal health markers. Geocoding had already concentration reflects an entire lifetime of transformed continuous measure showed a been attempted for every residential address in higher-than-expected exposure. As in study I, stronger association than the untransformed the C8 Health Project (Shin et al. 2011b). we considered a log-transformed continuous continuous measure. There were also irregular We sought to link births from 1990–2004 measure, an untransformed continuous mea- findings across quintiles, but none suggestive of birth certificates in a 13-county region in West sure, and quintiles, aggregating the lowest two a monotonic gradient. Term SGA was weakly Virginia (Cabell, Jackson, Mason, Pleasants, quintiles as the referent. Pregnancy outcomes associated with increasing PFOA levels based Putnam, Ritchie, Wirt, and Wood counties) and covariates were defined as described for on continuous and categorical exposures across and Ohio (Athens, Gallie, Meigs, Morgan, and study I. We used generalized estimating all exposure estimates (Table 8). The continu - Washington counties) to self-reported births equations with an exchangeable correlation ous birth weight analysis yielded some support in the C8 Health Project. This 13-county structure to account for correlation across a for an association (Table 8), with a predicted region encompasses the 5 counties included in woman’s multiple pregnancies. decrement of 22 g for an interquartile shift study I. We considered a match to be accurate in exposure (95% CI: –46, 2.1 g), and decre- Results based on consistency between the birth cer- ments of 25 g (95% CI: –64, 13 g) and 33 g tificate and C8 Health Project information for As in study I, PFOA exposures were higher in (95% CI: –73, 6.5 g) in the fourth and fifth one of three combinations: a) mother’s county the more recent time period, among older and quintiles of exposure based on the uncalibrated and state of residence + mother’s date of birth more highly educated mothers, and among exposure estimates. This pattern was similar for + month and year of infant’s birth + infant’s nonsmokers; however, in study II expo- the Bayesian calibrated estimates and somewhat sex; b) mother’s county and state of residence sures were higher for Ohio births than West reduced with traditional calibration. No sex + mother’s age + mother’s first name, middle Virginia births, presumably reflecting the dif - differences were apparent [see Supplemental initial, last name + month and year of infant’s ferent geographic scope of the two studies Material, Table 5 (http://dx.doi.org/10.1289/ birth + infant sex; or c) mother’s county and (Table 5). ehp.1104752)]. Table 4. Study I: PFOA and pregnancy outcome based on birth records: association of PFOA with indicators of fetal growth, Mid-Ohio Valley, 1990–2004. Term low birth weight Term SGA Change in term birth weight (g) a a a Term births Cases Crude Adjusted Term AGA Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA ≥ 2,500 g (n) (n) OR OR (95% CI) (n) (n) OR OR (95% CI) (n) difference difference (95% CI) IQR(lnPFOA) increase 3,616 918 0.96 1.02 (0.92, 1.13) 2,990 353 0.85 0.91 (0.78, 1.06) 4,534 3.20 –10.72 (–32.26, 10.82) 100-ng/mL increase 3,616 918 0.91 1.00 (0.86, 1.15) 2,990 353 0.78 0.86 (0.67, 1.11) 4,534 6.44 –14.80 (–43.28, 13.68) < 40th percentile 1,420 378 1.0 1.0 1,172 148 1.0 1.0 1,798 0 0 (referent) (1.0 to < 6.1 ng/mL) 40th to < 60th percentile 741 176 0.9 0.9 (0.7, 1.2) 600 81 1.0 1.0 (0.7, 1.4) 917 21.3 22.8 (–32.9, 78.5) (6.1 to < 10.2 ng/mL) 60th to < 80th percentile 717 188 1.0 1.0 (0.8, 1.3) 596 69 1.0 1.0 (0.7, 1.5) 905 –9.6 2.3 (–50.3, 54.8) (10.2 to < 21.0 ng/mL) ≥ 80th percentile 738 176 0.9 1.0 (0.8, 1.3) 622 55 0.7 0.8 (0.6, 1.2) 914 10.9 –9.5 (–58.4, 39.4) (21.0 to 717.6 ng/mL) AGA, appropriate for gestational age. a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence, gestational age (Term birth weight analysis only). Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.50]. The category boundaries are from the first imputed data set. | | 1204 volume 120 number 8 August 2012 • Environmental Health Perspectives PFOA and pregnancy outcome in birth records Discussion essentially no association between PFOA and inconsistent definitions and inaccurate report- In study I, PFOA and Pregnancy Outcome pregnancy-induced hypertension, whereas ing. The C8 Health Project addresses self- Based on Birth Records, we found little indi- study II provided modest support, primarily reported “preeclampsia” (Savitz et al. 2012; cation of an association between PFOA and with calibrated exposure estimates. Among Stein et al. 2009) and birth certificates code stillbirth, pregnancy-induced hypertension, all pregnancy outcome measures, pregnancy- “pregnancy-induced hypertension,” although preterm birth, or indicators of fetal growth. induced hypertension is the most vulnerable to there is likely to be misclassification in both These analyses had the advantage of relative freedom from selection bias and large study Table 5. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: characteristics of the study population, Mid-Ohio Valley, 1990–2004 [n (%)]. size, but exposure was estimated based only on the residential address reported on the birth Estimated PFOA among Maternal Preterm all births (ng/mL) certificate rather than a lifetime residential his - characteristic PIH < 37 weeks Term LBW SGA All births [median (25th, 75th percentiles)] tory. Study II, PFOA and Pregnancy Outcome Total 250 405 99 362 4,547 13.4 (5.6, 61.2) Based on Birth Records Linked to the C8 Exposure year Health Project, yielded sporadic positive asso- 1990–1994 53 (21) 77 (19) 28 (28) 105 (29) 1,296 (29) 7.2 (4.6, 42.8) ciations but little overall support for an associa- 1995–1999 110 (44) 142 (35) 31 (31) 121 (33) 1,674 (37) 13.7 (5.9, 65.7) tion with pregnancy-induced hypertension or 2000–2004 87 (35) 186 (46) 40 (40) 136 (38) 1,577 (35) 18.3 (7.6, 63.2) preterm birth< 37 weeks, modest support for State of residence an association with indicators of fetal growth West Virginia 150 (60) 227 (56) 64 (65) 213 (59) 2,735 (60) 10.4 (5.3, 44.6) restriction, and strong but imprecise indications Ohio 100 (40) 178 (44) 35 (35) 149 (41) 1,812 (40) 23.9 (6.6, 81.2) of an association with preterm birth < 32 weeks. Age (years) < 20 32 (13) 51 (13) 19 (19) 66 (18) 541 (12) 8.8 (5.0, 33.7) These analyses incorporated better estimates of 20–24 82 (33) 124 (31) 37 (37) 149 (41) 1,514 (33) 10.8 (5.3, 46.0) exposure by using lifetime residential history, 25–29 77 (31) 118 (29) 23 (23) 81 (22) 1,320 (29) 15.6 (5.8, 73.1) but a smaller sample size and the potential for 30–34 46 (18) 79 (20) 18 (18) 51 (14) 855 (19) 19.0 (6.3, 74.6) selection bias from incomplete matching of ≥ 35 13 (5) 33 (8) 2 (2) 15 (4) 317 (7) 23.0 (6.3, 106.1) birth certificates to the C8 Health Project. Education (years) Study I and study II differ in several ways, < 12 26 (10) 65 (16) 24 (24) 86 (24) 605 (13) 8.6 (5.1, 23.1) including comprehensiveness and precision, but 12 109 (44) 176 (43) 46 (46) 157 (43) 1,970 (43) 11.3 (5.2, 50.3) exposure assignment is probably most influen - 13–15 67 (27) 110 (27) 17 (17) 83 (23) 1,309 (29) 18.7 (6.3, 79.6) ≥ 16 48 (19) 54 (13) 12 (12) 36 (10) 663 (15) 22.4 (6.3, 84.2) tial: Study II had residential history information Parity for reconstructing exposure, and study I did 0 167 (67) 189 (47) 52 (53) 189 (52) 2,007 (44) 11.9 (5.5, 52.8) not. Among the participants in both studies 1 50 (20) 143 (35) 24 (24) 108 (30) 1,652 (36) 15.6 (5.8, 70.0) (n = 626), the Spearman rank order correlation ≥ 2 33 (13) 73 (18) 23 (23) 65 (18) 888 (20) 13.2 (5.4, 61.5) coefficient between the exposure estimate based Smoking status on reported lifetime residential history versus Smoker 20 (8) 96 (24) 40 (40) 154 (43) 870 (19) 10.2 (5.2, 32.4) just the maternal residence on the birth cer- Nonsmoker 230 (92) 309 (76) 59 (60) 208 (57) 3,677 (81) 14.7 (5.7, 68.1) tificate was 0.64, reflecting a modest correlation Abbreviations: PIH: pregnancy-induced hypertension; LBW, low birth weight. and a clear loss of accuracy in study I. There is some overlap between the births Table 6. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: association of PFOA with pregnancy-induced hypertension, Mid-Ohio Valley, 1990–2004. included in the two studies given the shared geographic scope. More than 70% of preterm Live births Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) and term low birth weight cases in study II were also included in study I because of the Uncalibrated comprehensive assessment of these case groups IQR(lnPFOA) increase 3,905 250 1.08 1.05 (0.85, 1.31) 100-ng/mL increase 3,905 250 0.98 1.01 (0.91, 1.12) in study I. Fewer than 8% of all study I births, < 40th percentile (3.9 to < 8.9 ng/mL) 1,632 96 1.0 1.0 however, were included in study II. 40th to < 60th percentile (8.9 to < 21.8 ng/mL) 778 52 1.1 1.0 (0.7, 1.4) Integrating these findings into the litera - 60th to < 80th percentile (21.8 to 83.3 ng/mL) 736 52 1.2 1.1 (0.7, 1.5) ture, there are three previous studies in this ≥ 80th percentile (83.3 to 921.3 ng/mL) 759 50 1.1 1.1 (0.8, 1.5) geographic area (Nolan et al. 2009; Savitz et al. Bayesian calibration 2012; Stein et al. 2009) as well as other studies IQR(lnPFOA) increase 3,905 250 1.11 1.13 (0.92, 1.37) of PFOA and birth outcomes (Apelberg et al. 100-ng/mL increase 3,905 250 0.92 0.97 (0.85, 1.11) < 40th percentile (3.9 to < 8.9 ng/mL) 1,635 88 1.0 1.0 2007; Fei et al. 2007; Hamm et al. 2010). The 40 to < 60th percentile (8.9 to < 19.6 ng/mL) 757 46 1.1 1.0 (0.7, 1.4) study by Nolan et al. (2009) addressed births 60 to < 80th percentile (19.6 to 53.1 ng/mL) 735 67 1.7 1.5 (1.1, 2.1) in 2003–2005 in Washington County, Ohio, ≥ 80th percentile (53.1 to 1897.0 ng/mL) 778 49 1.2 1.2 (0.8, 1.7) comparing birth weight and preterm birth risk Traditional calibration among residents of ZIP codes with varying IQR(lnPFOA) increase 3,905 250 1.01 1.03 (0.87, 1.22) levels of PFOA in the water. 100-ng/mL increase 3,905 250 0.83 0.89 (0.75, 1.06) Stillbirth was examined in our previ- < 40th percentile (0.05 to < 11.4 ng/mL) 1,628 82 1.0 1.0 ous analysis of C8 Health Project partici- 40th to < 60th percentile (11.4 to < 21.0 ng/mL) 759 63 1.6 1.5 (1.1, 2.2) 60th to < 80th percentile (21.0 to 49.0 ng/mL) 729 62 1.7 1.5 (1.1, 2.2) pants (Savitz et al. 2012) and neither Savitz ≥ 80th percentile (49.0 to 2468.4 ng/mL) 789 43 1.1 1.1 (0.8, 1.7) et al. (2012) nor the present study suggested a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates rep- an association with PFOA. Preeclampsia resent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels was weakly associated with PFOA exposure [IQR(lnPFOA) = 2.39]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th in other analyses of this population (Savitz d percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.92]. Effect estimates represent the change in outcome for a et al. 2012; Stein et al. 2009). Study I showed shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.61]. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1205 Savitz et al. data sources. Pregnancy-induced hypertension (DiGiuseppe et al. 2002; Lydon-Rochelle et al. was hampered by small numbers of cases refers to elevated blood pressure with or with- 2005). These largely negative findings should and yielded little evidence of an association. out protein in the urine; preeclampsia spe- be tempered by recognition of substantial Previous reports on preterm birth in this study cifically requires proteinuria. When compared outcome misclassification. area, whether based on serum measures in with medical records, birth certificates yield Evaluation of preterm birth in previous recent pregnancies (Stein et al. 2009), histori- sensitivity and positive predictive values < 50% studies (Fei et al. 2007; Hamm et al. 2010) cal exposure estimates among larger numbers Table 7. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: association of PFOA with preterm birth, Mid-Ohio Valley, 1990–2004. < 37 weeks gestation < 32 weeks gestation a a Term births Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) OR OR (95% CI) Uncalibrated IQR(lnPFOA) increase 4,142 405 1.13 1.09 (0.91, 1.32) 40 1.16 1.29 (0.74, 2.23) 100-ng/mL increase 4,142 405 1.06 1.09 (1.00, 1.18) 40 1.00 1.10 (0.86, 1.40) < 40th percentile (3.9 to < 8.9 ng/mL) 1,669 150 1.0 1.0 13 1.0 1.0 40th to < 60th percentile (8.9 to < 21.8 ng/mL) 810 99 1.4 1.2 (0.9, 1.5) 12 1.9 1.5 (0.7, 3.5) 60th to < 80th percentile (21.8 to 83.3 ng/mL) 841 68 0.9 0.8 (0.6, 1.1) 7 1.1 0.9 (0.4, 2.4) ≥ 80th percentile (83.3 to 921.3 ng/mL) 822 88 1.2 1.2 (0.9, 1.6) 8 1.2 1.4 (0.5, 3.6) Bayesian calibration IQR(lnPFOA) increase 4,142 405 1.09 1.10 (0.92, 1.31) 40 1.33 1.67 (1.03, 2.70) 100-ng/mL increase 4,142 405 1.06 1.12 (1.02, 1.23) 40 1.06 1.25 (1.02, 1.52) < 40th percentile (3.9 to < 8.9 ng/mL) 1,666 153 1.0 1.0 10 1.0 1.0 40th to < 60th percentile (8.9 to < 19.6 ng/mL) 820 89 1.2 1.0 (0.8, 1.3) 9 1.8 1.5 (0.6, 4.1) 60th to < 80th percentile (19.6 to 53.1 ng/mL) 827 82 1.1 0.9 (0.7, 1.3) 12 2.4 2.3 (1.0, 5.5) ≥ 80th percentile (53.1 to 1897.0 ng/mL) 829 81 1.1 1.0 (0.8, 1.4) 9 1.8 2.3 (0.9, 5.7) Traditional calibration IQR(lnPFOA) increase 4,142 405 0.98 0.95 (0.82, 1.11) 40 1.17 1.33 (0.86, 2.07) 100-ng/mL increase 4,142 405 0.99 1.03 (0.94, 1.13) 40 1.01 1.13 (0.97, 1.32) < 40th percentile (0.05 to < 11.4 ng/mL) 1,649 170 1.0 1.0 13 1.0 1.0 40th to < 60th percentile (11.4 to < 21.0 ng/mL) 831 78 0.9 0.8 (0.6, 1.1) 7 1.0 1.0 (0.4, 2.6) 60th to < 80th percentile (21.0 to 49.0 ng/mL) 825 84 1.0 0.9 (0.7, 1.1) 14 2.1 2.1 (1.0, 4.5) ≥ 80th percentile (49.0 to 2468.4 ng/mL) 837 73 0.9 0.9 (0.6, 1.2) 6 0.9 1.2 (0.4, 3.1) a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 2.39]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.92]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.61]. Table 8. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: association of PFOA with indicators of fetal growth, Mid- Ohio Valley, 1990–2004. Term low birth weight Term SGA Change in term birth weight (g) Term births Term, a a a ≥ 2,500 g Cases Crude Adjusted AGA Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) (n) OR OR (95% CI) (n) difference difference (95% CI) Uncalibrated IQR(lnPFOA) increase 4,043 99 0.87 1.04 (0.75,1.44) 3,375 362 1.02 1.18 (0.97, 1.43) 4,142 0.97 –21.89 (–45.91, 2.13) 100-ng/mL increase 4,043 99 0.93 1.00 (0.82,1.21) 3,375 362 1.01 1.07 (0.98, 1.17) 4,142 4.27 –9.14 (–20.30, 2.02) < 40th percentile (3.9 to < 8.9 ng/mL) 1,629 40 1.0 1.0 1,356 144 1.0 1.0 1,669 0 0 (referent) 40th to < 60th percentile (8.9 to < 21.8 ng/mL) 791 19 0.9 0.9 (0.5, 1.7) 659 72 1.0 1.0 (0.7, 1.4) 810 –19.9 –3.8 (–40.4, 32.8) 60th to < 80th percentile (21.8 to 83.3 ng/mL) 814 27 1.4 1.6 (1.0, 2.8) 689 76 1.0 1.1 (0.8, 1.6) 841 –30.4 –25.4 (–63.7, 12.9) ≥ 80th percentile (83.3 to 921.3 ng/mL) 809 13 0.7 0.9 (0.5, 1.7) 671 70 1.0 1.3 (0.9, 1.7) 822 4.9 –33.3 (–73.1, 6.5) Bayesian calibration IQR(lnPFOA) increase 4,043 99 0.97 1.16 (0.86,1.58) 3,375 362 1.03 1.19 (1.00, 1.43) 4,142 7.78 –21.51 (–43.62, 0.61) 100-ng/mL increase 4,043 99 0.93 1.04 (0.85,1.27) 3,375 362 0.98 1.06 (0.97, 1.16) 4,142 3.73 –18.55 (–31.31, –5.80) < 40th percentile (3.9 to < 8.9 ng/mL) 1,624 42 1.0 1.0 1,358 148 1.0 1.0 1,666 0 0 (referent) 40th to < 60th percentile (8.9 to < 19.6 ng/mL) 803 17 0.8 0.8 (0.4, 1.5) 676 68 0.9 0.9 (0.7, 1.3) 820 –15.8 10.8 (–24.9, 46.5) 60th to < 80th percentile (19.6 to 53.1 ng/mL) 803 24 1.1 1.3 (0.7, 2.2) 664 74 1.0 1.1 (0.8, 1.5) 827 –9.0 –11.0 (–49.8, 27.8) ≥ 80th percentile (53.1 to 1897.0 ng/mL) 813 16 0.8 1.0 (0.6, 1.9) 677 72 1.0 1.3 (0.9, 1.7) 829 8.7 –32.3 (–71.5, 6.8) Traditional calibration IQR(lnPFOA) increase 4,043 99 1.16 1.33 (1.04,1.69) 3,375 362 1.05 1.17 (1.00, 1.36) 4,142 2.45 –16.90 (–34.89, 1.08) 100-ng/mL increase 4,043 99 1.00 1.07 (0.96,1.18) 3,375 362 1.03 1.08 (1.01, 1.16) 4,142 4.54 –12.76 (–26.08, 0.57) < 40th percentile (0.05 to < 11.4 ng/mL) 1,614 35 1.0 1.0 1,351 147 1.0 1.0 1,649 0 0 (referent) 40th to < 60th percentile (11.4 to < 21.0 ng/mL) 817 14 0.8 0.8 (0.4, 1.5) 686 70 0.9 1.0 (0.7, 1.3) 831 –9.2 4.2 (–31.2, 39.6) 60th to < 80th percentile (21.0 to 49.0 ng/mL) 793 32 1.8 2.2 (1.3, 3.6) 659 72 1.0 1.1 (0.8, 1.5) 825 1.1 1.8 (–37.7, 41.4) ≥ 80th percentile (49.0 to 2468.4 ng/mL) 819 18 1.1 1.4 (0.8, 2.5) 679 73 1.0 1.2 (0.9, 1.7) 837 22.7 –21.2 (–59.6, 17.2) AGA, appropriate for gestational age. a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence, gestational age (term birth weight analysis only). Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 2.39]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.92]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.6]. | | 1206 volume 120 number 8 August 2012 • Environmental Health Perspectives PFOA and pregnancy outcome in birth records of C8 Health Project pregnancies (Savitz et al. and excretion of PFOA, and may therefore Refe Rences 2012), or geographic analyses of exposed ZIP share some physiologic determinants of birth Apelberg BJ, Witter FR, Herbstman JB, Calafat AM, Halden RU, codes (Nolan et al. 2009) are consistent in weight (Longnecker 2006). In contrast, expo- Needham LL, et al. 2007. Cord serum concentrations of showing an absence of association. Study I sure assignments in the present study (except perfluorooctane sulfonate (PFOS) and perfluorooctano- likewise showed no evidence of an associa- for the calibrated estimates in study II) were ate (PFOA) in relation to weight and size at birth. Environ Health Perspect 115:1670–1676. tion, but the findings in study II are more based on distinctive sources of environmental DiGiuseppe DL, Aron DC, Ranbom L, Harper DL, Rosenthal GE. ambiguous. Preterm birth < 37 weeks had at contamination and were unaffected by physi - 2002. Reliability of birth certificate data: a multi-hospital most modest suggestions of an association, ologic differences in PFOA metabolism. Third, comparison to medical records information. Matern Child Health J 6(3):169–179. whereas preterm birth < 32 weeks generated the physiologic or clinical relevance of these Fei C, McLaughlin JK, Tarone RE, Olsen J. 2007. Perfluorinated large and imprecise elevations in risk, mark- small shifts in birth weight is uncertain, pos- chemicals and fetal growth: a study within the Danish edly strengthened by Bayesian calibration to sibly serving as a marker of subtle toxicity, but National Birth Cohort. Environ Health Perspect 115:1677–1682. Frisbee SJ, Brooks AP Jr, Maher A, Flensborg P, Arnold S, estimate exposure. Although the size of the not of direct importance to health. Not only Fletcher T, et al. 2009. The C8 health project: design, methods, associations warrant attention, the role of ran- are the implications of low birth weight subject and participants. Environ Health Perspect 117:1873–1882. dom error and instability across exposure mea- to varying interpretations (Wilcox 2001), but Graham JW, Olchowski AE, Gilreath TD. 2007. How many impu- tations are really needed? Some practical clarifications of sures must also be taken into account. shifts within the normal range, which drives multiple imputation theory. Prev Sci 8(3):206–213. Previous analyses of reduced birth weight the association because most births are within Hamm MP, Cherry NM, Chan E, Martin JW, Burstyn I. 2010. have focused largely on continuous birth the normal range, are even more questionable Maternal exposure to perfluorinated acids and fetal weight measures rather than on low birth as markers of adverse impacts on health. growth. J Expo Sci Environ Epidemiol 20(7):589–597. Lau C, Anitole K, Hodes C, Lai D, Pfahles-Hutchens A, Seed J. weight or SGA. Studies of other populations The limitations in these data warrant com - 2007. Perfluoroalkyl acids: a review of monitoring and toxico- have had limited ability to precisely examine ment, with most issues applicable to both stud- logical findings. Toxicol Sci 99(2):366–394. extreme levels of birth weight and have found ies. Birth certificates are a useful, if imperfect, Longnecker MP. 2006. Pharmacokinetic variability and the miracle of modern analytical chemistry. Epidemiology no support for an association with PFOA (Fei source of data on pregnancy outcome: Birth 17(4):350–351. et al. 2007; Hamm et al. 2010). Two previous weight is generally accurate, more so than ges- L y d o n - R o c h e l l e M T , H o l t V L , C a r d e n a s V , N e l s o n J C , studies provided support for an association tational age (DiGiuseppe et al. 2002), with Easterling TR, Gardella C, et al. 2005. The reporting of pre- existing maternal medical conditions and complications of between continuous measures of PFOA and pregnancy-induced hypertension the most fal- pregnancy on birth certificates and in hospital discharge birth weight (Apelberg et al. 2007; Fei et al. lible outcome. Information on potential con- data. Am J Obstet Gynecol 193(1):125–134. 2007), and two smaller studies provided less founders was limited to basic demographic Nolan LA, Nolan JM, Shofer FS, Rodway NV, Emmett EA. 2009. The relationship between birth weight, gestational age precise statistical indications of modest reduc- information and smoking, though the spatial and perfluorooctanoic acid (PFOA)-contaminated public tions in birth weight with increasing levels of manner in which the exposure was distributed drinking water. Reprod Toxicol 27(3–4):231–238. PFOA (Hamm et al. 2010; Washino et al. makes strong confounding by socioeconomic Oken E, Kleinman KP, Rich-Edwards J, Gillman MW. 2003. A nearly continuous measure of birth weight for gestational 2009). Previous analyses in the Mid-Ohio or behavioral factors unlikely. Even the expo- age using a United States national reference. BMC Pediatr Valley showed little indication of an effect of sure estimates based on full residential histo- 3:6; doi:10.1186/1471-2431-3-6 [Online 8 July 2003]. PFOA on fetal growth (Nolan et al. 2009; ries in study II are subject to error, given the Olsen GW, Burris JM, Ehresman DJ, Froehlich JW, Seacat AM, Savitz et al. 2012; Stein et al. 2009), but stud- inability to predict with precision the levels Butenhoff JL, et al. 2007. Half-life of serum elimination of perfluorooctanesulfonate, perfluorohexanesulfonate, and ies based on just the C8 Health Project (Savitz of contaminants in drinking water at a given perfluorooctanoate in retired fluorochemical production et al. 2012; Stein et al. 2009) were unable to place and time, individual variability in water workers. Environ Health Perspect 115:1298–1305. examine a continuous measure of birth weight. use and PFOA pharmacokinetics, and inac- Rubin DB. 1987. Multiple Imputation for Nonresponse in Surveys. New York:Wiley. In study I, there was some indication of decre- curacies in dates and locations of self-reported Savitz DA, Stein CR, Bartell SM, Elston E, Gong J, Shin HM, ments in birth weight associated with increas- residences. The magnitude of these errors is et al. 2012. Perflourooctane acid exposure and pregnancy ing PFOA levels, particularly for male births in exacerbated in study I because of the absence of outcome in a highly exposed community. Epidemiology 23(3):386–392. the categorical analysis. In study II, all growth longitudinal information on residential history. S h i n H M , V i e i r a V M , R y a n P B , D e t w i l e r R , S a n d e r s B , indicators—term low birth weight, term SGA, Overall, we found little support for PFOA Steenland K, et al. 2011a. Environmental fate and trans- and a continuous measure of birth weight— having adverse effects on pregnancy-induced port modeling for perfluorooctanoic acid emitted from the Washington Works facility in West Virginia. Environ Sci showed small, inconsistent suggestions of a hypertension or preterm birth in the aggre- Technol 45(4):1435–1442. possible association, which is comparable with gate; modest and inconsistent evidence of Shin HM, Vieira VM, Ryan PB, Steenland K, Bartell SM. 2011b. the previous reports. reduced fetal growth; and large, imprecise Retrospective exposure estimation and predicted versus observed serum perfluorooctanoic acid concentrations There are a number of issues to consider indications of an association with early pre- for participants in the C8 Health Project. Environ Health with regard to possible decrements in birth term birth. The irregularity across exposure Perspect 119:1760–1765. weight associated with PFOA exposure. First, indices and susceptibility to inconsistent Steenland K, Fletcher T, Savitz DA. 2010. Epidemiologic evi- the studies of typical levels of PFOA (Apelberg variation based on calibration tempers the dence on the health effects of perfluorooctanoic acid (PFOA). Environ Health Perspect 118:1100–1108. et al. 2007; Fei et al. 2007) yield associations suggestions of possible causal effects on these Steenland K, Jin C, MacNeil J, Lally C, Ducatman A, Vieira V, that are similar to our effect sizes even though pregnancy outcomes. The potential associa - et al. 2009. Predictors of PFOA levels in a community we examined a much greater range of exposure. tion of PFOA with small decrements in birth surrounding a chemical plant. Environ Health Perspect 117:1083–1088. If there is a causal effect in which the impact weight is of unknown significance, but is Stein CR, Savitz DA, Dougan M. 2009. Serum levels of perfluo - is linear at both very low and very high expo- perhaps the most readily amenable to more rooctanoic acid and perfluorooctane sulfonate and preg - sures, we would have expected the larger abso- refined studies for continued investigation, nancy outcome. Am J Epidemiol 170(7):837–846. Washino N, Saijo Y, Sasaki S, Kato S, Ban S, Konishi K, et al. lute exposure contrasts in the present study with studies of preterm birth < 32 weeks, a 2009. Correlations between prenatal exposure to perfluori- to yield stronger associations than those from rare outcome, requiring nested case–control nated chemicals and reduced fetal growth. Environ Health studies of a narrower exposure range. Second, studies based on stored serum for exposure Perspect 117:660–667. West Virginia Department of Health. 2012. Birth Score. the previous studies examining typical serum assessment. If an impact on birth weight is Available: http://www.wvdhhr.org/birthscore/ [accessed levels of PFOA during pregnancy were sus- causal, which is yet to be established, it reflects 6 July 2012]. ceptible to distortion by relying on exposure a biological process that is worth elucidating Wilcox AJ. 2001. On the importance—and the unimportance—of birthweight. Int J Epidemiol 30(6):1233–1241. measures that reflect the uptake, metabolism, even if not of direct health concern. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1207 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Environmental Health Perspectives Unpaywall

Relationship of Perfluorooctanoic Acid Exposure to Pregnancy Outcome Based on Birth Records in the Mid-Ohio Valley

Download PDF
7 pages

Loading next page...
 
/lp/unpaywall/relationship-of-perfluorooctanoic-acid-exposure-to-pregnancy-outcome-ZKMEuKdEaJ

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Unpaywall
ISSN
0091-6765
DOI
10.1289/ehp.1104752
Publisher site
See Article on Publisher Site

Abstract

Research Children’s Health Relationship of Perfluorooctanoic Acid Exposure to Pregnancy Outcome Based on Birth Records in the Mid-Ohio Valley 1 2 1 1 3 4 David A. Savitz, Cheryl R. Stein, Beth Elston, Gregory A. Wellenius, Scott M. Bartell, Hyeong-Moo Shin, 5 6 Veronica M. Vieira, and Tony Fletcher 1 2 Department of Epidemiology, Brown University, Providence, Rhode Island, USA; Department of Preventive Medicine, Mount Sinai 3 4 School of Medicine, New York, New York, USA; Program in Public Health, and School of Social Ecology, University of California, Irvine, Irvine, California, USA; Department of Environmental Health, Boston University School of Public Health, Boston, Massachusetts, USA; London School of Hygiene and Tropical Medicine, London, United Kingdom the region. The first provides comprehen- Background : Perfluorooctanoic acid (PFOA) is a potential cause of adverse pregnancy outcomes, sive coverage of the study area, but is subject but previous studies have been limited by low exposures and small study size. to some exposure misclassification. The sec - o Bjectives : Using birth certificate information, we examined the relation between estimated PFOA ond sacrifices some of the sample size, but exposure and birth outcomes in an area of West Virginia and Ohio whose drinking water was con- has improved exposure measurement. Both taminated by a chemical plant. studies rely on birth certificates for assess- Methods : Births in the study area from 1990 through 2004 were examined to generate case groups ment of health end points and use the same of stillbirth (n = 106), pregnancy-induced hypertension (n = 224), preterm birth (n = 3,613), term environmental/pharmacokinetic models to low birth weight (n = 918), term small-for-gestational-age (SGA) (n = 353), and a continuous estimate PFOA exposure. measure of birth weight among a sample of term births (n = 4,534). A 10% sample of term births ≥ 2,500 g were selected as a source of controls (n = 3,616). Historical estimates of serum PFOA Study I: PFOA and were derived from a previously developed fate and transport model. In a second study, we examined Pregnancy Outcome Based 4,547 area births linked to a survey with residential history data. on Birth Records results : In the analysis based only on birth records, we found no consistent evidence of an asso- ciation between estimated PFOA exposure and stillbirth, pregnancy-induced hypertension, preterm Methods birth, or indices of fetal growth. In the analysis of birth records linked to the survey, PFOA was The study area in the mid-Ohio Valley was unrelated to pregnancy-induced hypertension or preterm birth but showed some suggestion of an chosen to include PFOA-contaminated public association with early preterm birth. Measures of growth restriction showed weak and inconsistent associations with PFOA. water utilities in Mason and Wood counties in West Virginia and in Athens, Meigs, and c onclusions : Based on the analysis using the health survey, these results provide little support for Washington counties in Ohio, where levels an effect of PFOA exposure on most pregnancy outcomes, except for early preterm birth and pos - ranged from near U.S. averages to orders of sibly fetal growth restriction. magnitude above typical levels (Frisbee et al. k ey words : fetal growth restriction, perfluorooctanoic acid, pregnancy, pregnancy-induced hyper - 2009). We obtained computerized live birth tension, preterm birth, stillbirth. Environ Health Perspect 120:1201–1207 (2012). http://dx.doi. org/10.1289/ehp.1104752 [Online 26 March 2012] records with geographic identifiers from the Ohio and West Virginia health departments, and fetal death records from West Virginia Adverse effects of perfluorooctanoic acid the potential association between PFOA and (PFOA) and related perfluoroalkyl acids on pregnancy outcomes is warranted. Address correspondence to D.A. Savitz, Department pregnancy and development have been sug- A large population in Ohio and West of Epidemiology, Box G-S-121-2, Brown University, Providence, RI 02912 USA. Telephone: (401) 863- gested based on toxicology (Lau et al. 2007) Virginia has been exposed to PFOA predomi- 6090. Fax: (401) 863-3713. E-mail: david_savitz@ and a small but growing epidemiologic lit- nantly through industrial contamination of brown.edu erature (Steenland et al. 2010). The stron- drinking-water supplies (Frisbee et al. 2009). Supplemental Material is available online (http:// gest support from laboratory studies suggests Within this population, PFOA exposure var- dx.doi.org/10.1289/ehp.1104752). the potential for reductions in fetal growth ies considerably across time and place because The authors thank K. Steenland, the Ohio and (Lau et al. 2007). Epidemiologic findings of different levels of PFOA contamination in West Virginia Health Departments, and the West Virginia Birth Score Program for assistance. on measures of reduced growth are mixed the drinking water, depending on the year This research was funded by the C8 class action (Apelberg et al. 2007; Fei et al. 2007; Hamm and the water district, enabling accurate esti- settlement agreement [Jack W. Leach, et al. v. E.I. et al. 2010; Nolan et al. 2009; Washino et al. mation of serum levels by linking residen- du Pont de Nemours & Company (no. 01-C-608 2009), but generally consistent with a small tial histories and historical drinking-water W.Va., Wood County Circuit Court, West Virginia, decrement in birth weight associated with concentrations. The resulting range of PFOA USA)] between DuPont and plaintiffs. Funds were higher levels of serum PFOA. Duration of levels are well above background exposure administered by the Garden City Group (Melville, New York) that reports to the court. Our work and gestation has not been related to PFOA in (Steenland et al. 2009), which may allow for conclusions are independent of either party to the these same studies (Apelberg et al. 2007; a more informative assessment of the impact lawsuit. C.S. was supported by grant K01 ES019156 Fei et al. 2007; Hamm et al. 2010). More of PFOA exposure on birth outcomes. from the National Institute of Environmental Health extreme, clinically consequential end points For participants in the C8 Health Sciences (NIEHS). G.W. was supported by grant of fetal growth restriction and preterm birth Project—a survey conducted in the contami- R00-ES015774 from NIEHS. have been examined less extensively and often nated region (Frisbee et al. 2009)—we pre- The contents of this report are solely the responsi - bility of the authors and do not necessarily represent with limited statistical power; none of the viously analyzed a small number of births the official views of NIEHS or National Institutes human studies support an association (Fei using serum PFOA measurements (Stein of Health. et al. 2007; Hamm et al. 2010; Nolan et al. et al. 2009) and a much larger number using e Th authors declare they have no actual or potential 2009). Given ubiquitous exposure, sugges- modeling to assign estimated serum PFOA competing financial interests. tive toxicologic data, and limited epidemi- levels (Savitz et al. 2012). In this report, we Received 16 November 2011; accepted 26 March ologic research, continued examination of used two approaches to evaluate births in 2012. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1201 Savitz et al. only (unpublished data). We restricted the the mother estimated for the early pregnancy each imputed data set were pooled into a final analysis to pregnancies that occurred from period. The methods for estimating individual parameter estimate and variance (Rubin 1987). 1990 through 2004 because that is the period serum PFOA concentrations are described in Missing covariate data on maternal age (n = 2), when the contrasts were highest between detail elsewhere (Shin et al. 2011a, 2011b). maternal education (n = 61), parity (n = 6), exposures resulting from contaminant releases Briefly, information on plant operations and and tobacco use (n = 64) were also imputed. and background sources (Shin et al. 2011a). chemical releases was combined with environ- The primary analyses incorporate the exposure Because of the expense involved with refin - mental characteristics of the region through estimates for the street-level geocode (66%) ing address information and geocoding mater- a series of linked models to estimate air and and the imputed exposure estimate when only nal residences, which is the basis for assigning water concentrations of PFOA from 1951 ZIP code geocodes could be assigned (34%). exposure, we used case–control sampling to through 2008 (Shin et al. 2011a). Geocodes In sensitivity analyses we considered the impact enhance precision for some of the birth out- for birth residences were used to determine of restricting to births with exposure estimates comes of interest. We restricted eligibility to whether drinking water in participants’ homes based on street-level geocodes only and no singleton births and excluded the small pro- was supplied by public water sources or pri- missing covariate data. portion (3%) of nonwhite births because vate wells. When addresses could be geocoded Estimated serum PFOA concentrations there were too few to analyze separately. only at the ZIP code level, an average exposure were generated by calendar year, with the rele- Comprehensively identified case groups con - measure was calculated based on the propor- vant exposure year for each pregnancy consid- sisted of stillbirths (fetal death ≥ 20 weeks com- tion of the ZIP code population that was sup- ered to be month 3 of gestation for stillbirths pleted gestation), preterm births (< 37 weeks plied by contaminated drinking water. With and month 4 for live births. The estimated completed gestation), and term low birth these estimated environmental levels of PFOA serum PFOA level at the time of pregnancy weight births (≥ 37 weeks completed gestation based on the geocoded maternal address, indi- was skewed, so we considered serum PFOA and birth weight < 2,500 g). A 10% sample of vidual maternal serum levels for each preg- estimates as a continuous log-transformed term births ≥ 2,500 g (n = 3,616) was selected nancy were estimated using age- and sex-based measure, a continuous untransformed mea- and geocoded for assigning PFOA exposure. pharmacokinetic modeling with standard sure, and in quintiles of exposure. We aggre- To efficiently select control groups and assumptions about water intake, body weights, gated the lowest two quintiles as the reference expand the range of birth outcomes, we first and a PFOA half-life (Shin et al. 2011b). category for analysis because those quintiles constructed a stratified 10% random sample Because we had only the residence listed covered a modest range typical of U.S. serum of the entire population of live births by com- on the birth certificate for the mother—not levels and our exposure estimates are best able bining the 10% sample of all term live births a residential history—and serum levels at a to differentiate individual exposure at higher ≥ 2500 g with a 10% random sample of the given time depend on exposure history, we levels resulting from consumption of locally comprehensively identified preterm cases and assumed that she had lived in this location for contaminated water. term low birth weight cases (the case groups the previous 6 years, which would account for We adjusted all analyses for maternal age described above) (n = 4063). Pregnancy- approximately 63% of her serum concentra- (splines with 3 degrees of freedom), education induced hypertension cases, term small-for- tion given a stable exposure rate and an esti- (< 12, 12, 13–15, ≥ 16 years), parity (0, 1, gestational-age (SGA) cases, and term birth mated serum half-life of PFOA of 3.5 years ≥ 2), tobacco use (smoker, nonsmoker), expo- weight births were identified exclusively from (Olsen et al. 2007). Moving within the same sure year (splines with 4 degrees of freedom), the 10% random sample of births. Control public water supply district would have little and state of residence (Ohio, West Virginia). groups for each of the outcomes were derived or no impact on the estimated serum values, In addition, we included gestational age as noncase subsets from the 10% random but moving across water districts or especially (37, 38, 39, 40, 41, ≥ 42 completed weeks) sample of the entire population of live births. from more distant locations would lead to as a covariate in the analysis of birth weight We conducted a series of case–control exposure misclassification. among term births. analyses: a) stillbirths (n = 106) compared with We were able to geocode the maternal For the binary outcomes, we used logistic West Virginia live births (n = 1,844) because address listed on the birth certificate to the regression to estimate the odds ratio (OR) and stillbirth information was available only from street level for 66% of births and to the ZIP 95% confidence interval (CI) associated with West Virginia; b) pregnancy-induced hyper- code level for the remaining 34%, with lit- PFOA exposure, using the log-transformed tension cases (n = 224) compared with births tle variation in these proportions by case or value to generate ORs for an interquartile without pregnancy-induced hypertension control status. We found a high correlation shift from the 25th to 75th percentile, and the (n = 3,828); c) preterm birth cases < 37 weeks between the exposure estimate using street untransformed variable used to generate ORs (n = 3,613) and < 32 weeks (n = 491) cor- level geocode and ZIP code averages (r = 0.84), per 100-ng/mL shift and for quintiles (com- responding to all preterm and very preterm indicating modest loss of information from paring the 3rd, 4th, and 5th with the aggre- births, respectively, compared with term births lack of street level assignment. To make opti- gated 1st and 2nd quintiles). Both crude and (n = 3,695), with assignment based on clinical mal use of the available data, we used the ZIP adjusted effect estimates were generated. We assessment of gestational age; d) term low birth code average exposure and other information used multiple linear regression to estimate the weight cases (n = 918) compared with term from the birth certificate (maternal educa- association between PFOA and birth weight, births ≥ 2,500 g (n = 3,616); and e) term SGA tion, maternal age, parity, tobacco use, expo- using the same measures of estimated PFOA cases [births < 10th percentile by gestational sure year, residential state) to impute a new exposure described for the categorical out- age and sex (Oken et al. 2003)] (n = 353) exposure estimate for the pregnancies with ZIP comes. All statistical analyses were conducted compared with term appropriate-for-gesta- code geocodes only (n = 2,780). We used mul- using SAS version 9.2 (SAS Institute Inc., tional-age births [births between the 10th and tiple imputation to generate 20 replications to Cary, NC) and R version 2.12.1 (R Project 90th percentiles, inclusive, by gestational age obtain accurate information on the variability for Statistical Computing, Vienna, Austria). and sex (Oken et al. 2003)] (n = 2,990). In in the imputed values (Graham et al. 2007). Results addition, we analyzed a continuous measure of Markov Chain Monte Carlo imputation was birth weight among term births (n = 4,534). used with 200 burn-in iterations and 100 itera- Births were nearly evenly divided by calendar The exposure metric used in the analy- tions between each imputation. The parameter period and by state (Table 1). Estimated expo- sis is modeled serum PFOA concentration of estimates and standard errors determined from sure to PFOA tended to be higher for births | | 1202 volume 120 number 8 August 2012 • Environmental Health Perspectives PFOA and pregnancy outcome in birth records that were more recent, among older mothers, on the untransformed measure (OR = 1.20 PFOA was not related to risk of term more highly educated mothers, nonsmokers, per 100 ng/mL; 95% CI: 0.86, 1.68), the lack low birth weight, with some suggestion of and West Virginia residents (Table 1). of gradient across quintiles does not support a reduced risk of SGA with higher exposure Stillbirth was unrelated to PFOA expo- an association. Pregnancy-induced hyperten- but no gradient across quintiles (Table 4). sure based on either continuous or categorical sion was unrelated to PFOA exposure regard- Birth weight among term births was largely exposure measures (Table 2). Although there less of metric (Table 2). Preterm birth was not unrelated to estimated PFOA (Table 4). was some suggestion of an increased risk based associated with PFOA (Table 3). There was a suggestion of a modest decrease Table 1. Study I: PFOA and pregnancy outcome based on birth records: characteristics of study population by pregnancy outcome, Mid-Ohio Valley, 1990–2004 [n (%)]. Estimated PFOA among Maternal Preterm all births (ng/mL) characteristic Stillbirth PIH < 37 weeks Term LBW SGA All births median (25th, 75th percentiles) Total 106 224 3,613 918 353 8,253 7.7 (4.9, 17.2) Exposure year 1990–1994 28 (26) 51 (23) 1,068 (30) 295 (32) 117 (33) 2,685 (33) 7.0 (4.5, 17.7) 1995–1999 49 (46) 84 (38) 1,170 (32) 312 (34) 124 (35) 2,776 (34) 7.9 (5.0, 16.4) 2000–2004 29 (27) 89 (40) 1,375 (38) 311 (34) 112 (32) 2,792 (34) 8.2 (5.0, 17.6) State of residence WV 88 (83) 111 (50) 1,470 (41) 401 (44) 157 (44) 3,610 (44) 10.2 (5.3, 19.8) OH 18 (17) 113 (50) 2,143 (59) 517 (56) 196 (56) 4,643 (56) 6.3 (4.6, 13.4) Age (years) < 20 18 (17) 29 (13) 602 (17) 186 (20) 75 (21) 1,330 (16) 6.7 (4.7, 13.5) 20–24 39 (37) 67 (30) 1,187 (33) 321 (35) 138 (39) 2,743 (33) 7.2 (4.9, 16.4) 25–29 23 (22) 69 (31) 949 (26) 209 (23) 77 (22) 2,185 (26) 8.4 (4.9, 18.9) 30–34 17 (16) 44 (20) 572 (16) 128 (14) 46 (13) 1,346 (16) 8.5 (4.9, 19.8) ≥ 35 9 (8) 15 (7) 303 (8) 74 (8) 17 (5) 649 (8) 8.4 (4.9, 19.7) Education (years) < 12 21 (20) 24 (11) 772 (21) 273 (30) 113 (32) 1,701 (21) 6.9 (4.8, 14.3) 12 51 (48) 94 (42) 1,518 (42) 389 (42) 134 (38) 3,435 (42) 7.5 (4.9, 17.3) 13–15 24 (23) 67 (30) 865 (24) 167 (18) 68 (19) 1,996 (24) 8.3 (4.9, 19.8) ≥ 16 10 (9) 39 (17) 458 (13) 89 (10) 38 (11) 1,121 (14) 8.3 (5.0, 17.2) Parity 0 49 (46) 137 (61) 1,710 (47) 457 (50) 178 (50) 3,801 (46) 7.6 (4.9, 16.7) 1 29 (27) 46 (21) 1,104 (31) 271 (30) 102 (29) 2,668 (32) 7.7 (4.9, 17.1) ≥ 2 28 (26) 41 (18) 799 (22) 190 (21) 73 (21) 1,784 (22) 7.6 (4.8, 18.2) Smoking status Smoker 27 (25) 32 (14) 1,039 (29) 430 (47) 167 (47) 2,364 (29) 7.0 (4.8, 15.1) Nonsmoker 79 (75) 192 (86) 2,574 (71) 488 (53) 186 (53) 5,889 (71) 8.0 (4.9, 18.1) Abbreviations: PIH, Pregnancy-induced hypertension; LBW, low birth weight. Some Ohio residents delivered stillbirths in West Virginia. Table 2. Study I: PFOA and pregnancy outcome based on birth records: association of PFOA with stillbirth and pregnancy-induced hypertension, Mid-Ohio Valley, 1990–2004. Stillbirth Pregnancy-induced hypertension a b b Live births Cases Crude Adjusted Live births Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) (n) OR OR (95% CI) IQR(lnPFOA) increase 1,844 106 0.92 1.00 (0.76, 1.32) 3,828 224 1.06 1.02 (0.86, 1.21) 100-ng/mL increase 1,844 106 1.01 1.20 (0.86, 1.68) 3,828 224 1.06 1.06 (0.86, 1.31) < 40th percentile (1.0 to < 6.1 ng/mL ) 504 35 1.0 1.0 1,527 82 1.0 1.0 40th to < 60th percentile (6.1 to < 10.2 ng/mL) 330 15 0.9 0.9 (0.4, 2.0) 764 48 1.2 1.0 (0.7, 1.6) 60th to < 80th percentile (10.2 to < 21.0 ng/mL) 487 31 1.1 1.0 (0.5, 1.7) 759 51 1.1 1.0 (0.6, 1.5) ≥ 80th percentile (21.0 to 717.6 ng/mL) 523 25 0.7 0.8 (0.5, 1.5) 778 43 1.1 1.0 (0.7, 1.5) a b c West Virginia only. Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.50]. The category boundaries are from the first imputed data set. Table 3. Study I: PFOA and pregnancy outcome based on birth records: association of PFOA with preterm birth, Mid-Ohio Valley, 1990–2004. < 37 weeks gestation < 32 weeks gestation a a Term births Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) OR OR (95% CI) IQR(lnPFOA) increase 3,695 3,613 0.99 1.02 (0.96, 1.08) 491 0.95 0.98 (0.86, 1.11) 100-ng/mL increase 3,695 3,613 1.01 1.02 (0.94, 1.10) 491 0.88 0.90 (0.74, 1.10) < 40th percentile (1.0 to < 6.1 ng/mL ) 1,460 1,467 1.0 1.0 197 1.0 1.0 40th to < 60th percentile (6.1 to < 10.2 ng/mL) 755 720 1.0 1.0 (0.8, 1.1) 108 1.0 1.0 (0.7, 1.4) 60th to < 80th percentile (10.2 to < 21.0 ng/mL) 733 714 1.0 1.0 (0.9, 1.2) 95 1.0 1.1 (0.8, 1.4) ≥ 80th percentile (21.0 to 717.6 ng/mL) 747 712 1.0 1.0 (0.9, 1.2) 91 0.9 1.0 (0.7, 1.3) a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.50]. The category boundaries are from the first imputed data set. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1203 Savitz et al. in birth weight with higher exposure based state of residence + mother’s address + mother’s Pregnancy-induced hypertension showed on continuous exposure measures {somewhat age + month and year of infant’s birth + infant essentially no association with estimated serum stronger for male births [see Supplemental sex. This process yielded 3,323 matched births. PFOA based on the uncalibrated estimates Material, Table 1 (http://dx.doi.org/10.1289/ We extended the matching through the West (Table 6). With Bayesian calibration, the log- ehp.1104752)]}, but no monotonic gradi- Virginia Birth Score project, a program of the transformed and categorical measures sug- ent across quintiles. In the sensitivity analysis West Virginia Department of Health (2012) gested a positive association, with an increased restricted to the 66% of births with street-level to identify infants at risk of developmental risk primarily in the 4th quintile. Traditional geocodes and complete covariate informa- problems. Using the mother’s Social Security calibration showed elevated risk in the 3rd and tion (see Supplemental Material, Tables 2–4), number as well as other birth certificate items, 4th quintiles, but the continuous exposure results were largely unchanged except for gen- Birth Score was able to identify an additional measures did not indicate an association. erating weak positive associations with both 1,224 matched births through 2004, yielding a Preterm birth < 37 weeks was weakly pregnancy-induced hypertension and late total of 4,547 birth certificates linked to the C8 related to PFOA levels based on continuous preterm births. Health Project for analysis. measures in the uncalibrated analyses, but We used the same environmental/ with no gradient across quintiles (Table 7). Study II: PFOA and pharmacokinetic models from study I to gen- Bayesian calibration yielded a similar pattern Pregnancy Outcome Based erate annual exposure estimates, except that and traditional calibration yielded no asso- on Birth Records Linked to the the link to the C8 Health Project enabled us ciation. Preterm birth < 32 weeks, with only to incorporate lifetime residential history, thus 40 cases, showed some suggestion of an associ- C8 Health Project reducing exposure misclassification (Shin et al. ation without calibration, enhanced somewhat Methods 2011a, 2011b). We examined the model-based with traditional calibration, and markedly The C8 Health Project enrolled > 69,000 resi- exposure estimates, a Bayesian time-dependent strengthened with Bayesian calibration. dents of the mid-Ohio Valley who were part of calibration in which the measured 2005–2006 Term low birth weight was largely unre- a class action lawsuit, administered a question- serum concentration was used for updat- lated to PFOA based on exposure measures naire eliciting health and residential history ing estimates (Savitz et al. 2012), and a tra- without calibration, except for an isolated information, and collected blood to measure ditional calibration method assuming that increased risk in the 4th quintile (Table 8). serum levels of perfluoroalkyl acids and clini - a higher-than-expected 2005–2006 serum With both calibration methods, the log- cal health markers. Geocoding had already concentration reflects an entire lifetime of transformed continuous measure showed a been attempted for every residential address in higher-than-expected exposure. As in study I, stronger association than the untransformed the C8 Health Project (Shin et al. 2011b). we considered a log-transformed continuous continuous measure. There were also irregular We sought to link births from 1990–2004 measure, an untransformed continuous mea- findings across quintiles, but none suggestive of birth certificates in a 13-county region in West sure, and quintiles, aggregating the lowest two a monotonic gradient. Term SGA was weakly Virginia (Cabell, Jackson, Mason, Pleasants, quintiles as the referent. Pregnancy outcomes associated with increasing PFOA levels based Putnam, Ritchie, Wirt, and Wood counties) and covariates were defined as described for on continuous and categorical exposures across and Ohio (Athens, Gallie, Meigs, Morgan, and study I. We used generalized estimating all exposure estimates (Table 8). The continu - Washington counties) to self-reported births equations with an exchangeable correlation ous birth weight analysis yielded some support in the C8 Health Project. This 13-county structure to account for correlation across a for an association (Table 8), with a predicted region encompasses the 5 counties included in woman’s multiple pregnancies. decrement of 22 g for an interquartile shift study I. We considered a match to be accurate in exposure (95% CI: –46, 2.1 g), and decre- Results based on consistency between the birth cer- ments of 25 g (95% CI: –64, 13 g) and 33 g tificate and C8 Health Project information for As in study I, PFOA exposures were higher in (95% CI: –73, 6.5 g) in the fourth and fifth one of three combinations: a) mother’s county the more recent time period, among older and quintiles of exposure based on the uncalibrated and state of residence + mother’s date of birth more highly educated mothers, and among exposure estimates. This pattern was similar for + month and year of infant’s birth + infant’s nonsmokers; however, in study II expo- the Bayesian calibrated estimates and somewhat sex; b) mother’s county and state of residence sures were higher for Ohio births than West reduced with traditional calibration. No sex + mother’s age + mother’s first name, middle Virginia births, presumably reflecting the dif - differences were apparent [see Supplemental initial, last name + month and year of infant’s ferent geographic scope of the two studies Material, Table 5 (http://dx.doi.org/10.1289/ birth + infant sex; or c) mother’s county and (Table 5). ehp.1104752)]. Table 4. Study I: PFOA and pregnancy outcome based on birth records: association of PFOA with indicators of fetal growth, Mid-Ohio Valley, 1990–2004. Term low birth weight Term SGA Change in term birth weight (g) a a a Term births Cases Crude Adjusted Term AGA Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA ≥ 2,500 g (n) (n) OR OR (95% CI) (n) (n) OR OR (95% CI) (n) difference difference (95% CI) IQR(lnPFOA) increase 3,616 918 0.96 1.02 (0.92, 1.13) 2,990 353 0.85 0.91 (0.78, 1.06) 4,534 3.20 –10.72 (–32.26, 10.82) 100-ng/mL increase 3,616 918 0.91 1.00 (0.86, 1.15) 2,990 353 0.78 0.86 (0.67, 1.11) 4,534 6.44 –14.80 (–43.28, 13.68) < 40th percentile 1,420 378 1.0 1.0 1,172 148 1.0 1.0 1,798 0 0 (referent) (1.0 to < 6.1 ng/mL) 40th to < 60th percentile 741 176 0.9 0.9 (0.7, 1.2) 600 81 1.0 1.0 (0.7, 1.4) 917 21.3 22.8 (–32.9, 78.5) (6.1 to < 10.2 ng/mL) 60th to < 80th percentile 717 188 1.0 1.0 (0.8, 1.3) 596 69 1.0 1.0 (0.7, 1.5) 905 –9.6 2.3 (–50.3, 54.8) (10.2 to < 21.0 ng/mL) ≥ 80th percentile 738 176 0.9 1.0 (0.8, 1.3) 622 55 0.7 0.8 (0.6, 1.2) 914 10.9 –9.5 (–58.4, 39.4) (21.0 to 717.6 ng/mL) AGA, appropriate for gestational age. a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence, gestational age (Term birth weight analysis only). Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.50]. The category boundaries are from the first imputed data set. | | 1204 volume 120 number 8 August 2012 • Environmental Health Perspectives PFOA and pregnancy outcome in birth records Discussion essentially no association between PFOA and inconsistent definitions and inaccurate report- In study I, PFOA and Pregnancy Outcome pregnancy-induced hypertension, whereas ing. The C8 Health Project addresses self- Based on Birth Records, we found little indi- study II provided modest support, primarily reported “preeclampsia” (Savitz et al. 2012; cation of an association between PFOA and with calibrated exposure estimates. Among Stein et al. 2009) and birth certificates code stillbirth, pregnancy-induced hypertension, all pregnancy outcome measures, pregnancy- “pregnancy-induced hypertension,” although preterm birth, or indicators of fetal growth. induced hypertension is the most vulnerable to there is likely to be misclassification in both These analyses had the advantage of relative freedom from selection bias and large study Table 5. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: characteristics of the study population, Mid-Ohio Valley, 1990–2004 [n (%)]. size, but exposure was estimated based only on the residential address reported on the birth Estimated PFOA among Maternal Preterm all births (ng/mL) certificate rather than a lifetime residential his - characteristic PIH < 37 weeks Term LBW SGA All births [median (25th, 75th percentiles)] tory. Study II, PFOA and Pregnancy Outcome Total 250 405 99 362 4,547 13.4 (5.6, 61.2) Based on Birth Records Linked to the C8 Exposure year Health Project, yielded sporadic positive asso- 1990–1994 53 (21) 77 (19) 28 (28) 105 (29) 1,296 (29) 7.2 (4.6, 42.8) ciations but little overall support for an associa- 1995–1999 110 (44) 142 (35) 31 (31) 121 (33) 1,674 (37) 13.7 (5.9, 65.7) tion with pregnancy-induced hypertension or 2000–2004 87 (35) 186 (46) 40 (40) 136 (38) 1,577 (35) 18.3 (7.6, 63.2) preterm birth< 37 weeks, modest support for State of residence an association with indicators of fetal growth West Virginia 150 (60) 227 (56) 64 (65) 213 (59) 2,735 (60) 10.4 (5.3, 44.6) restriction, and strong but imprecise indications Ohio 100 (40) 178 (44) 35 (35) 149 (41) 1,812 (40) 23.9 (6.6, 81.2) of an association with preterm birth < 32 weeks. Age (years) < 20 32 (13) 51 (13) 19 (19) 66 (18) 541 (12) 8.8 (5.0, 33.7) These analyses incorporated better estimates of 20–24 82 (33) 124 (31) 37 (37) 149 (41) 1,514 (33) 10.8 (5.3, 46.0) exposure by using lifetime residential history, 25–29 77 (31) 118 (29) 23 (23) 81 (22) 1,320 (29) 15.6 (5.8, 73.1) but a smaller sample size and the potential for 30–34 46 (18) 79 (20) 18 (18) 51 (14) 855 (19) 19.0 (6.3, 74.6) selection bias from incomplete matching of ≥ 35 13 (5) 33 (8) 2 (2) 15 (4) 317 (7) 23.0 (6.3, 106.1) birth certificates to the C8 Health Project. Education (years) Study I and study II differ in several ways, < 12 26 (10) 65 (16) 24 (24) 86 (24) 605 (13) 8.6 (5.1, 23.1) including comprehensiveness and precision, but 12 109 (44) 176 (43) 46 (46) 157 (43) 1,970 (43) 11.3 (5.2, 50.3) exposure assignment is probably most influen - 13–15 67 (27) 110 (27) 17 (17) 83 (23) 1,309 (29) 18.7 (6.3, 79.6) ≥ 16 48 (19) 54 (13) 12 (12) 36 (10) 663 (15) 22.4 (6.3, 84.2) tial: Study II had residential history information Parity for reconstructing exposure, and study I did 0 167 (67) 189 (47) 52 (53) 189 (52) 2,007 (44) 11.9 (5.5, 52.8) not. Among the participants in both studies 1 50 (20) 143 (35) 24 (24) 108 (30) 1,652 (36) 15.6 (5.8, 70.0) (n = 626), the Spearman rank order correlation ≥ 2 33 (13) 73 (18) 23 (23) 65 (18) 888 (20) 13.2 (5.4, 61.5) coefficient between the exposure estimate based Smoking status on reported lifetime residential history versus Smoker 20 (8) 96 (24) 40 (40) 154 (43) 870 (19) 10.2 (5.2, 32.4) just the maternal residence on the birth cer- Nonsmoker 230 (92) 309 (76) 59 (60) 208 (57) 3,677 (81) 14.7 (5.7, 68.1) tificate was 0.64, reflecting a modest correlation Abbreviations: PIH: pregnancy-induced hypertension; LBW, low birth weight. and a clear loss of accuracy in study I. There is some overlap between the births Table 6. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: association of PFOA with pregnancy-induced hypertension, Mid-Ohio Valley, 1990–2004. included in the two studies given the shared geographic scope. More than 70% of preterm Live births Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) and term low birth weight cases in study II were also included in study I because of the Uncalibrated comprehensive assessment of these case groups IQR(lnPFOA) increase 3,905 250 1.08 1.05 (0.85, 1.31) 100-ng/mL increase 3,905 250 0.98 1.01 (0.91, 1.12) in study I. Fewer than 8% of all study I births, < 40th percentile (3.9 to < 8.9 ng/mL) 1,632 96 1.0 1.0 however, were included in study II. 40th to < 60th percentile (8.9 to < 21.8 ng/mL) 778 52 1.1 1.0 (0.7, 1.4) Integrating these findings into the litera - 60th to < 80th percentile (21.8 to 83.3 ng/mL) 736 52 1.2 1.1 (0.7, 1.5) ture, there are three previous studies in this ≥ 80th percentile (83.3 to 921.3 ng/mL) 759 50 1.1 1.1 (0.8, 1.5) geographic area (Nolan et al. 2009; Savitz et al. Bayesian calibration 2012; Stein et al. 2009) as well as other studies IQR(lnPFOA) increase 3,905 250 1.11 1.13 (0.92, 1.37) of PFOA and birth outcomes (Apelberg et al. 100-ng/mL increase 3,905 250 0.92 0.97 (0.85, 1.11) < 40th percentile (3.9 to < 8.9 ng/mL) 1,635 88 1.0 1.0 2007; Fei et al. 2007; Hamm et al. 2010). The 40 to < 60th percentile (8.9 to < 19.6 ng/mL) 757 46 1.1 1.0 (0.7, 1.4) study by Nolan et al. (2009) addressed births 60 to < 80th percentile (19.6 to 53.1 ng/mL) 735 67 1.7 1.5 (1.1, 2.1) in 2003–2005 in Washington County, Ohio, ≥ 80th percentile (53.1 to 1897.0 ng/mL) 778 49 1.2 1.2 (0.8, 1.7) comparing birth weight and preterm birth risk Traditional calibration among residents of ZIP codes with varying IQR(lnPFOA) increase 3,905 250 1.01 1.03 (0.87, 1.22) levels of PFOA in the water. 100-ng/mL increase 3,905 250 0.83 0.89 (0.75, 1.06) Stillbirth was examined in our previ- < 40th percentile (0.05 to < 11.4 ng/mL) 1,628 82 1.0 1.0 ous analysis of C8 Health Project partici- 40th to < 60th percentile (11.4 to < 21.0 ng/mL) 759 63 1.6 1.5 (1.1, 2.2) 60th to < 80th percentile (21.0 to 49.0 ng/mL) 729 62 1.7 1.5 (1.1, 2.2) pants (Savitz et al. 2012) and neither Savitz ≥ 80th percentile (49.0 to 2468.4 ng/mL) 789 43 1.1 1.1 (0.8, 1.7) et al. (2012) nor the present study suggested a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates rep- an association with PFOA. Preeclampsia resent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels was weakly associated with PFOA exposure [IQR(lnPFOA) = 2.39]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th in other analyses of this population (Savitz d percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.92]. Effect estimates represent the change in outcome for a et al. 2012; Stein et al. 2009). Study I showed shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.61]. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1205 Savitz et al. data sources. Pregnancy-induced hypertension (DiGiuseppe et al. 2002; Lydon-Rochelle et al. was hampered by small numbers of cases refers to elevated blood pressure with or with- 2005). These largely negative findings should and yielded little evidence of an association. out protein in the urine; preeclampsia spe- be tempered by recognition of substantial Previous reports on preterm birth in this study cifically requires proteinuria. When compared outcome misclassification. area, whether based on serum measures in with medical records, birth certificates yield Evaluation of preterm birth in previous recent pregnancies (Stein et al. 2009), histori- sensitivity and positive predictive values < 50% studies (Fei et al. 2007; Hamm et al. 2010) cal exposure estimates among larger numbers Table 7. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: association of PFOA with preterm birth, Mid-Ohio Valley, 1990–2004. < 37 weeks gestation < 32 weeks gestation a a Term births Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) OR OR (95% CI) Uncalibrated IQR(lnPFOA) increase 4,142 405 1.13 1.09 (0.91, 1.32) 40 1.16 1.29 (0.74, 2.23) 100-ng/mL increase 4,142 405 1.06 1.09 (1.00, 1.18) 40 1.00 1.10 (0.86, 1.40) < 40th percentile (3.9 to < 8.9 ng/mL) 1,669 150 1.0 1.0 13 1.0 1.0 40th to < 60th percentile (8.9 to < 21.8 ng/mL) 810 99 1.4 1.2 (0.9, 1.5) 12 1.9 1.5 (0.7, 3.5) 60th to < 80th percentile (21.8 to 83.3 ng/mL) 841 68 0.9 0.8 (0.6, 1.1) 7 1.1 0.9 (0.4, 2.4) ≥ 80th percentile (83.3 to 921.3 ng/mL) 822 88 1.2 1.2 (0.9, 1.6) 8 1.2 1.4 (0.5, 3.6) Bayesian calibration IQR(lnPFOA) increase 4,142 405 1.09 1.10 (0.92, 1.31) 40 1.33 1.67 (1.03, 2.70) 100-ng/mL increase 4,142 405 1.06 1.12 (1.02, 1.23) 40 1.06 1.25 (1.02, 1.52) < 40th percentile (3.9 to < 8.9 ng/mL) 1,666 153 1.0 1.0 10 1.0 1.0 40th to < 60th percentile (8.9 to < 19.6 ng/mL) 820 89 1.2 1.0 (0.8, 1.3) 9 1.8 1.5 (0.6, 4.1) 60th to < 80th percentile (19.6 to 53.1 ng/mL) 827 82 1.1 0.9 (0.7, 1.3) 12 2.4 2.3 (1.0, 5.5) ≥ 80th percentile (53.1 to 1897.0 ng/mL) 829 81 1.1 1.0 (0.8, 1.4) 9 1.8 2.3 (0.9, 5.7) Traditional calibration IQR(lnPFOA) increase 4,142 405 0.98 0.95 (0.82, 1.11) 40 1.17 1.33 (0.86, 2.07) 100-ng/mL increase 4,142 405 0.99 1.03 (0.94, 1.13) 40 1.01 1.13 (0.97, 1.32) < 40th percentile (0.05 to < 11.4 ng/mL) 1,649 170 1.0 1.0 13 1.0 1.0 40th to < 60th percentile (11.4 to < 21.0 ng/mL) 831 78 0.9 0.8 (0.6, 1.1) 7 1.0 1.0 (0.4, 2.6) 60th to < 80th percentile (21.0 to 49.0 ng/mL) 825 84 1.0 0.9 (0.7, 1.1) 14 2.1 2.1 (1.0, 4.5) ≥ 80th percentile (49.0 to 2468.4 ng/mL) 837 73 0.9 0.9 (0.6, 1.2) 6 0.9 1.2 (0.4, 3.1) a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 2.39]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.92]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.61]. Table 8. Study II: PFOA and pregnancy outcome based on birth records linked to the C8 Health Project: association of PFOA with indicators of fetal growth, Mid- Ohio Valley, 1990–2004. Term low birth weight Term SGA Change in term birth weight (g) Term births Term, a a a ≥ 2,500 g Cases Crude Adjusted AGA Cases Crude Adjusted Cases Crude Adjusted Estimated PFOA (n) (n) OR OR (95% CI) (n) (n) OR OR (95% CI) (n) difference difference (95% CI) Uncalibrated IQR(lnPFOA) increase 4,043 99 0.87 1.04 (0.75,1.44) 3,375 362 1.02 1.18 (0.97, 1.43) 4,142 0.97 –21.89 (–45.91, 2.13) 100-ng/mL increase 4,043 99 0.93 1.00 (0.82,1.21) 3,375 362 1.01 1.07 (0.98, 1.17) 4,142 4.27 –9.14 (–20.30, 2.02) < 40th percentile (3.9 to < 8.9 ng/mL) 1,629 40 1.0 1.0 1,356 144 1.0 1.0 1,669 0 0 (referent) 40th to < 60th percentile (8.9 to < 21.8 ng/mL) 791 19 0.9 0.9 (0.5, 1.7) 659 72 1.0 1.0 (0.7, 1.4) 810 –19.9 –3.8 (–40.4, 32.8) 60th to < 80th percentile (21.8 to 83.3 ng/mL) 814 27 1.4 1.6 (1.0, 2.8) 689 76 1.0 1.1 (0.8, 1.6) 841 –30.4 –25.4 (–63.7, 12.9) ≥ 80th percentile (83.3 to 921.3 ng/mL) 809 13 0.7 0.9 (0.5, 1.7) 671 70 1.0 1.3 (0.9, 1.7) 822 4.9 –33.3 (–73.1, 6.5) Bayesian calibration IQR(lnPFOA) increase 4,043 99 0.97 1.16 (0.86,1.58) 3,375 362 1.03 1.19 (1.00, 1.43) 4,142 7.78 –21.51 (–43.62, 0.61) 100-ng/mL increase 4,043 99 0.93 1.04 (0.85,1.27) 3,375 362 0.98 1.06 (0.97, 1.16) 4,142 3.73 –18.55 (–31.31, –5.80) < 40th percentile (3.9 to < 8.9 ng/mL) 1,624 42 1.0 1.0 1,358 148 1.0 1.0 1,666 0 0 (referent) 40th to < 60th percentile (8.9 to < 19.6 ng/mL) 803 17 0.8 0.8 (0.4, 1.5) 676 68 0.9 0.9 (0.7, 1.3) 820 –15.8 10.8 (–24.9, 46.5) 60th to < 80th percentile (19.6 to 53.1 ng/mL) 803 24 1.1 1.3 (0.7, 2.2) 664 74 1.0 1.1 (0.8, 1.5) 827 –9.0 –11.0 (–49.8, 27.8) ≥ 80th percentile (53.1 to 1897.0 ng/mL) 813 16 0.8 1.0 (0.6, 1.9) 677 72 1.0 1.3 (0.9, 1.7) 829 8.7 –32.3 (–71.5, 6.8) Traditional calibration IQR(lnPFOA) increase 4,043 99 1.16 1.33 (1.04,1.69) 3,375 362 1.05 1.17 (1.00, 1.36) 4,142 2.45 –16.90 (–34.89, 1.08) 100-ng/mL increase 4,043 99 1.00 1.07 (0.96,1.18) 3,375 362 1.03 1.08 (1.01, 1.16) 4,142 4.54 –12.76 (–26.08, 0.57) < 40th percentile (0.05 to < 11.4 ng/mL) 1,614 35 1.0 1.0 1,351 147 1.0 1.0 1,649 0 0 (referent) 40th to < 60th percentile (11.4 to < 21.0 ng/mL) 817 14 0.8 0.8 (0.4, 1.5) 686 70 0.9 1.0 (0.7, 1.3) 831 –9.2 4.2 (–31.2, 39.6) 60th to < 80th percentile (21.0 to 49.0 ng/mL) 793 32 1.8 2.2 (1.3, 3.6) 659 72 1.0 1.1 (0.8, 1.5) 825 1.1 1.8 (–37.7, 41.4) ≥ 80th percentile (49.0 to 2468.4 ng/mL) 819 18 1.1 1.4 (0.8, 2.5) 679 73 1.0 1.2 (0.9, 1.7) 837 22.7 –21.2 (–59.6, 17.2) AGA, appropriate for gestational age. a b Adjusted for maternal age, education, parity, smoking status, exposure year, state of residence, gestational age (term birth weight analysis only). Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 2.39]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.92]. Effect estimates represent the change in outcome for a shift from the 25th percentile to the 75th percentile in estimated PFOA serum levels [IQR(lnPFOA) = 1.6]. | | 1206 volume 120 number 8 August 2012 • Environmental Health Perspectives PFOA and pregnancy outcome in birth records of C8 Health Project pregnancies (Savitz et al. and excretion of PFOA, and may therefore Refe Rences 2012), or geographic analyses of exposed ZIP share some physiologic determinants of birth Apelberg BJ, Witter FR, Herbstman JB, Calafat AM, Halden RU, codes (Nolan et al. 2009) are consistent in weight (Longnecker 2006). In contrast, expo- Needham LL, et al. 2007. Cord serum concentrations of showing an absence of association. Study I sure assignments in the present study (except perfluorooctane sulfonate (PFOS) and perfluorooctano- likewise showed no evidence of an associa- for the calibrated estimates in study II) were ate (PFOA) in relation to weight and size at birth. Environ Health Perspect 115:1670–1676. tion, but the findings in study II are more based on distinctive sources of environmental DiGiuseppe DL, Aron DC, Ranbom L, Harper DL, Rosenthal GE. ambiguous. Preterm birth < 37 weeks had at contamination and were unaffected by physi - 2002. Reliability of birth certificate data: a multi-hospital most modest suggestions of an association, ologic differences in PFOA metabolism. Third, comparison to medical records information. Matern Child Health J 6(3):169–179. whereas preterm birth < 32 weeks generated the physiologic or clinical relevance of these Fei C, McLaughlin JK, Tarone RE, Olsen J. 2007. Perfluorinated large and imprecise elevations in risk, mark- small shifts in birth weight is uncertain, pos- chemicals and fetal growth: a study within the Danish edly strengthened by Bayesian calibration to sibly serving as a marker of subtle toxicity, but National Birth Cohort. Environ Health Perspect 115:1677–1682. Frisbee SJ, Brooks AP Jr, Maher A, Flensborg P, Arnold S, estimate exposure. Although the size of the not of direct importance to health. Not only Fletcher T, et al. 2009. The C8 health project: design, methods, associations warrant attention, the role of ran- are the implications of low birth weight subject and participants. Environ Health Perspect 117:1873–1882. dom error and instability across exposure mea- to varying interpretations (Wilcox 2001), but Graham JW, Olchowski AE, Gilreath TD. 2007. How many impu- tations are really needed? Some practical clarifications of sures must also be taken into account. shifts within the normal range, which drives multiple imputation theory. Prev Sci 8(3):206–213. Previous analyses of reduced birth weight the association because most births are within Hamm MP, Cherry NM, Chan E, Martin JW, Burstyn I. 2010. have focused largely on continuous birth the normal range, are even more questionable Maternal exposure to perfluorinated acids and fetal weight measures rather than on low birth as markers of adverse impacts on health. growth. J Expo Sci Environ Epidemiol 20(7):589–597. Lau C, Anitole K, Hodes C, Lai D, Pfahles-Hutchens A, Seed J. weight or SGA. Studies of other populations The limitations in these data warrant com - 2007. Perfluoroalkyl acids: a review of monitoring and toxico- have had limited ability to precisely examine ment, with most issues applicable to both stud- logical findings. Toxicol Sci 99(2):366–394. extreme levels of birth weight and have found ies. Birth certificates are a useful, if imperfect, Longnecker MP. 2006. Pharmacokinetic variability and the miracle of modern analytical chemistry. Epidemiology no support for an association with PFOA (Fei source of data on pregnancy outcome: Birth 17(4):350–351. et al. 2007; Hamm et al. 2010). Two previous weight is generally accurate, more so than ges- L y d o n - R o c h e l l e M T , H o l t V L , C a r d e n a s V , N e l s o n J C , studies provided support for an association tational age (DiGiuseppe et al. 2002), with Easterling TR, Gardella C, et al. 2005. The reporting of pre- existing maternal medical conditions and complications of between continuous measures of PFOA and pregnancy-induced hypertension the most fal- pregnancy on birth certificates and in hospital discharge birth weight (Apelberg et al. 2007; Fei et al. lible outcome. Information on potential con- data. Am J Obstet Gynecol 193(1):125–134. 2007), and two smaller studies provided less founders was limited to basic demographic Nolan LA, Nolan JM, Shofer FS, Rodway NV, Emmett EA. 2009. The relationship between birth weight, gestational age precise statistical indications of modest reduc- information and smoking, though the spatial and perfluorooctanoic acid (PFOA)-contaminated public tions in birth weight with increasing levels of manner in which the exposure was distributed drinking water. Reprod Toxicol 27(3–4):231–238. PFOA (Hamm et al. 2010; Washino et al. makes strong confounding by socioeconomic Oken E, Kleinman KP, Rich-Edwards J, Gillman MW. 2003. A nearly continuous measure of birth weight for gestational 2009). Previous analyses in the Mid-Ohio or behavioral factors unlikely. Even the expo- age using a United States national reference. BMC Pediatr Valley showed little indication of an effect of sure estimates based on full residential histo- 3:6; doi:10.1186/1471-2431-3-6 [Online 8 July 2003]. PFOA on fetal growth (Nolan et al. 2009; ries in study II are subject to error, given the Olsen GW, Burris JM, Ehresman DJ, Froehlich JW, Seacat AM, Savitz et al. 2012; Stein et al. 2009), but stud- inability to predict with precision the levels Butenhoff JL, et al. 2007. Half-life of serum elimination of perfluorooctanesulfonate, perfluorohexanesulfonate, and ies based on just the C8 Health Project (Savitz of contaminants in drinking water at a given perfluorooctanoate in retired fluorochemical production et al. 2012; Stein et al. 2009) were unable to place and time, individual variability in water workers. Environ Health Perspect 115:1298–1305. examine a continuous measure of birth weight. use and PFOA pharmacokinetics, and inac- Rubin DB. 1987. Multiple Imputation for Nonresponse in Surveys. New York:Wiley. In study I, there was some indication of decre- curacies in dates and locations of self-reported Savitz DA, Stein CR, Bartell SM, Elston E, Gong J, Shin HM, ments in birth weight associated with increas- residences. The magnitude of these errors is et al. 2012. Perflourooctane acid exposure and pregnancy ing PFOA levels, particularly for male births in exacerbated in study I because of the absence of outcome in a highly exposed community. Epidemiology 23(3):386–392. the categorical analysis. In study II, all growth longitudinal information on residential history. S h i n H M , V i e i r a V M , R y a n P B , D e t w i l e r R , S a n d e r s B , indicators—term low birth weight, term SGA, Overall, we found little support for PFOA Steenland K, et al. 2011a. Environmental fate and trans- and a continuous measure of birth weight— having adverse effects on pregnancy-induced port modeling for perfluorooctanoic acid emitted from the Washington Works facility in West Virginia. Environ Sci showed small, inconsistent suggestions of a hypertension or preterm birth in the aggre- Technol 45(4):1435–1442. possible association, which is comparable with gate; modest and inconsistent evidence of Shin HM, Vieira VM, Ryan PB, Steenland K, Bartell SM. 2011b. the previous reports. reduced fetal growth; and large, imprecise Retrospective exposure estimation and predicted versus observed serum perfluorooctanoic acid concentrations There are a number of issues to consider indications of an association with early pre- for participants in the C8 Health Project. Environ Health with regard to possible decrements in birth term birth. The irregularity across exposure Perspect 119:1760–1765. weight associated with PFOA exposure. First, indices and susceptibility to inconsistent Steenland K, Fletcher T, Savitz DA. 2010. Epidemiologic evi- the studies of typical levels of PFOA (Apelberg variation based on calibration tempers the dence on the health effects of perfluorooctanoic acid (PFOA). Environ Health Perspect 118:1100–1108. et al. 2007; Fei et al. 2007) yield associations suggestions of possible causal effects on these Steenland K, Jin C, MacNeil J, Lally C, Ducatman A, Vieira V, that are similar to our effect sizes even though pregnancy outcomes. The potential associa - et al. 2009. Predictors of PFOA levels in a community we examined a much greater range of exposure. tion of PFOA with small decrements in birth surrounding a chemical plant. Environ Health Perspect 117:1083–1088. If there is a causal effect in which the impact weight is of unknown significance, but is Stein CR, Savitz DA, Dougan M. 2009. Serum levels of perfluo - is linear at both very low and very high expo- perhaps the most readily amenable to more rooctanoic acid and perfluorooctane sulfonate and preg - sures, we would have expected the larger abso- refined studies for continued investigation, nancy outcome. Am J Epidemiol 170(7):837–846. Washino N, Saijo Y, Sasaki S, Kato S, Ban S, Konishi K, et al. lute exposure contrasts in the present study with studies of preterm birth < 32 weeks, a 2009. Correlations between prenatal exposure to perfluori- to yield stronger associations than those from rare outcome, requiring nested case–control nated chemicals and reduced fetal growth. Environ Health studies of a narrower exposure range. Second, studies based on stored serum for exposure Perspect 117:660–667. West Virginia Department of Health. 2012. Birth Score. the previous studies examining typical serum assessment. If an impact on birth weight is Available: http://www.wvdhhr.org/birthscore/ [accessed levels of PFOA during pregnancy were sus- causal, which is yet to be established, it reflects 6 July 2012]. ceptible to distortion by relying on exposure a biological process that is worth elucidating Wilcox AJ. 2001. On the importance—and the unimportance—of birthweight. Int J Epidemiol 30(6):1233–1241. measures that reflect the uptake, metabolism, even if not of direct health concern. | | Environmental Health Perspectives • volume 120 number 8 August 2012 1207

Journal

Environmental Health Perspectives – Unpaywall

Published: Aug 1, 2012

There are no references for this article.