Perfluorinated Chemicals and Fetal Growth: A Study within the Danish National Birth Cohort — TEI-tagged view

Chunyuan FeipersName, Joseph K. McLaughlinpersName, Robert E. TaronepersName, Jørn OlsenpersName (2007date). “Perfluorinated Chemicals and Fetal Growth: A Study within the Danish National Birth Cohort.” Environmental Health Perspectivestitle·j.

Object of study — maternal plasma PFOSterm·key and PFOAterm in early pregnancy and infant birth weightterm·mth and gestation length

Research question — whether maternal PFOSterm·key and PFOAterm levels are associated with birth weightterm·mth and length of gestation in humans

Contribution — the authors' nationwide cohort evidence of an inverse association between maternal PFOAterm and birth weightterm·mth

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Perfluorinated Chemicals and Fetal Growthterm·key: A Study within the Danish National Birth Cohortterm·key Abstract Perfluorooctanesulfonate (PFOSterm·key) and perfluorooctanoate (PFOAterm) are man-made, persistent organic pollutants widely spread throughout the environment and human populations. They have been found to interfere with fetal growthterm·key in some animal models, but whether a similar effect is seen in humans is uncertain.

We investigated the association between plasma levels of PFOSterm·key and PFOAterm in pregnant women and their infants’ birth weightterm·mth and length of gestation.

We randomly selected 1,400 women and their infants from the Danish National Birth Cohortterm·key among those who completed all four computer-assisted telephone interviews, provided the first blood samples between gestational weeks 4 and 14, and who gave birth to a single live-born child without congenital malformation. PFOSterm·key and PFOAterm were measured by high performance liquid chromatography–tandem mass spectrometer.

PFOSterm·key and PFOAterm levels in maternal plasma were on average 35.3 and 5.6 ng/mL, respectively. Only PFOAterm levels were inversely associated with birth weightterm·mth (adjusted β = −10.63 g; 95% confidence interval, −20.79 to −0.47 g). Neither maternal PFOSterm·key nor PFOAterm levels were consistently associated with the risk for preterm birthterm·dom or low birth weightterm·dom. We observed no adverse effects for maternal PFOSterm·key or PFOAterm levels on small for gestational ageterm·dom.

Our nationwide cohort data suggest an inverse association between maternal plasma PFOAterm levels and birth weightterm·mth. Because of widespread exposure to these chemicals, our findings may be of potential public health concern.

Perfluorooctanesulfonate (PFOSterm·key) and perfluorooctanoate (PFOAterm) are persistent organic pollutants (POPsterm·dom), which are produced synthetically or from the metabolism of other perfluorinated chemicals (PFCs) [ Butenhoff et al. 2006 ; U.S. Environmental Protection Agency (EPA) 2006 ]. These compounds are widely used as industrial surfactants and emulsifiers and in numerous consumer products. Nonstick pans, carpets, furniture, household cleaners, shampoos, shoes, clothing, and convenience food packaging are some of the products that can contain PFCs ( Butenhoff et al. 2006 ; U.S. EPA 2006 ).

PFOAterm and PFOSterm·key are slowly metabolized and have half-lives in humans of around 4 years and 5 years, respectively ( OlsenpersName et al. 2007 ). They are absorbed through oral intake or inhaled and, to a lesser extent, through dermal exposure. They have been detected in the sera of occupationally and nonoccupationally exposed human populations in various countries ( Calafat et al. 2007 ; Kannan et al. 2004 ; Karrman et al. 2006 ; OlsenpersName et al. 2003 ; Yeung et al. 2006 ), including pregnant women and their newborn babies ( Apelberg et al. 2007a ; Inoue et al. 2004 ; Jin et al. 2004 ; Karrman et al. 2007 ; Midasch et al. 2007 ; So et al. 2006 ; Tittlemier et al. 2004 ).

Exposure of rodent dams during pregnancy to both PFOSterm·key and PFOAterm has been associated with decrements in birth weightterm·mth and length of gestation at birth ( Butenhoff et al. 2004b ; Case et al. 2001 ; Grasty et al. 2003 ; Lau et al. 2004 , 2006 ; Luebker et al. 2005a , 2005b ; Thibodeaux et al. 2003 ; Wolf et al. 2007 ). Although it is unknown how PFOSterm·key and PFOAterm may impair fetal growthterm·key, several mechanisms have been suggested, including a disturbance of lipid metabolism ( Kennedy et al. 2004 ; Luebker et al. 2005b ), a reduction of food and water intake, or altered hormones levels ( Betts 2007 ; Lau et al. 2004 ; Luebker et al. 2005b ; Thibodeaux et al. 2003 ), but a direct fetotoxic effect has also been suggested ( Betts 2007 ; Henderson and Smith 2007 ; Hinderliter et al. 2005 ; Luebker et al. 2005a , 2005b ). No effect of PFOSterm·key exposure on fetal growthterm·key in humans was found in a study based on self-reports of birth weightterm·mth and occupational histories ( Grice et al. 2007 ) or in a study of maternal and cord blood samples from 15 children ( Inoue et al. 2004 ). However, a recent study reported levels of both PFOSterm·key and PFOAterm to be inversely associated with birth weightterm·mth, newborn head circumference, crown–heel length, and ponderal index, but the study was based on cord blood samples taken after delivery ( Apelberg et al. 2007b ). No large-scale study based on prospectively collected biomarkers at the time of fetal growthterm·key has been published to date.

In this study we examined the relation between maternal plasma PFOSterm·key and PFOAterm levels during pregnancy and birth weightterm·mth as well as length of gestation using data from the Danish National Birth Cohortterm·key ( OlsenpersName et al. 2001 ).

Methods Study population The Danish National Birth Cohortterm·key (DNBC) includes data on 91,827 pregnant women from March 1996 to November 2002 ( OlsenpersName et al. 2001 ). Women were recruited nationwide (a population of 5.4 million people) by their general practitioners (GPs) early in pregnancy. About half of all GPs in the country participated, and about 60% of the pregnant women accepted the invitation to participate. After providing written informed consent, women took part in four computer-assisted telephone interviews at approximately 12 and 30 weeks of gestation, and at approximately 6 and 18 months after birth. A food frequency questionnaire was filled out at home approximately at gestational week 25. Moreover, blood was drawn twice from the women, once in the first and once in the second trimester. An umbilical cord blood sample was obtained shortly after birth. Each blood sample was sent by ordinary mail to the State Serum Institute in Copenhagen, Denmark, and then separated and stored in freezers at −20°C or in liquid nitrogen. Blood samples were thus transported at outside temperatures for 4–48 hr, but most of the samples arrived within 28 hr.

Among all participants who gave birth to a single live-born child without a reported congenital malformation ( n = 87,752), who provided the first blood sample between gestational weeks 4 and 14 ( n = 80,678), and who had responded to all four telephone interviews ( n = 43,045), we randomly selected 1,400 mothers. Of the 1,400 selected, 1,102 had a second blood sample, and 200 were randomly selected among these. Of these 200, 146 had a cord blood sample from the child, and 50 were randomly selected among these. We used the second maternal and cord blood samples to examine whether PFOSterm·key and PFOAterm levels were stable during pregnancy and whether the mother–offspring correlation in blood levels was high. Sample sizes were determined in large part by the capacity of the laboratory to perform PFOSterm·key and PFOAterm assays and by what would be acceptable for the Steering Committee that gives permission for use of cohort data.

Information on potential confounders—such as infant sex, maternal age, parity, socio-occupational status, prepregnancy body mass index (BMI), and smoking during the pregnancy—was collected by highly structured questionnaires (available at http://www.bsmb.dk ) ( OlsenpersName et al. 2001 ). Socio-occupational status was based on education and current job titles. Women with a higher education (4 years beyond secondary school education) or in management level jobs were classified as “high” social status; women with middle-range training and skilled workers were classified as “middle”; and unskilled workers and unemployed were classified as “low.” Birth weightterm·mth and gestational age at birth were obtained from the National Hospital Discharge Register at the National Board of Health in Denmark. The assessment of gestational age was done by the midwives based on either the first day of the last menstrual period (LMP) (conditioned on a regular bleeding pattern during the preceding 6 months, and no use of oral contraceptives during the 3 months preceding pregnancy) or from ultrasound examination done before 24 weeks of gestation ( Albertsen et al. 2004 ). If estimates differed, ultrasound measures replaced the LMP estimates. In most cases, gestational age was based on early ultrasound measures. If data were missing or out of the expected range ( n = 10), we used the expected date of delivery provided by the pregnant women at the second interview (after their ultrasound examinations). Low birth weightterm·dom (LBW) was defined as birth weightterm·mth < 2,500 g. Preterm birthterm·dom was defined as the birth of an infant before 37 completed weeks of gestation (259 days), post-term birth as after 42 completed weeks of gestation (294 days), and term birth as between 37 and 42 completed weeks of gestation. Small for gestational ageterm·dom (SGA) was defined as an infant with birth weightterm·mth below the 10th percentile at a specific gestational age in weeks, based on all singleton live births of same sex and parity who fulfilled our sampling criteria.

This research was approved by the UCLA Office for Protection of Research Subjects (Reference No. 06-08-023-01) and the Danish Data Protection Agency (Reference No. J.Nr 2006-41-6324).

Exposure assessment We measured plasma concentrations of PFOSterm·key and PFOAterm using high performance liquid chromatography–tandem mass spectrometry in the 3M Toxicology Laboratory ( Ehresman et al. 2007 ). The laboratory was blinded to birth outcomes and types of blood drawn (first or second maternal blood, or cord blood). Stable labeled analogs of PFOSterm·key ( 18 O 2 PFOSterm·key) and PFOAterm ( 13 C 2 PFOAterm) were used in all extracted procedures. All extractions were performed using solid phase extraction techniques and based on 100 μL of plasma. Details of the analytical methods used are available elsewhere ( Ehresman et al. 2007 ).

Quality control materials, a single lot of newborn calf serum, were analyzed together with the samples to ensure the accuracy and reliability of the data. Previous evaluation comparing serum and plasma samples has shown no significant difference when using an extracted serum spiked standard curve ( Ehresman et al. 2007 ). Two levels of spiked calf serum controls were extracted using 30 individual solid phase extractions to establish “within”-run means and standard deviations for each separate level of control. These levels were calculated at 15 ng/mL and 44 ng/mL for PFOAterm and 10 ng/mL and 30 ng/mL for PFOSterm·key. A total of 35 batch runs was necessary to complete all sample analyses. The coefficient of variations for the between-batch spiked control values were 3.5% and 3.2% for PFOAterm, respectively, and 2.8% and 2.5% for PFOSterm·key, respectively. Blind samples were also selected and relabeled for repeated analysis. There were strong correlations between the repeated blind analyses and the original results for PFOSterm·key ( r = 0.993, n = 100) and PFOAterm ( r = 0.987, n = 100). Only two PFOAterm values and no PFOSterm·key values were reported at the lower limit of quantitation (LLOQ) of 1.0 ng/mL. Values below the LLOQ ( n = 2) were assigned a value of half the LLOQ. PFOSterm·key and PFOAterm measurements were determined on the residual plasma component from 15 frozen whole blood samples (12 first maternal blood samples, 1 second blood sample, and 2 cord blood samples). These were multiplied by a factor of two to make them comparable to measurements made on plasma ( Ehresman et al. 2007 ).

Statistical analysis We used analyses of variance and linear regression to assess the associations between birth weightterm·mth and length of gestation and maternal plasma PFOSterm·key and PFOAterm levels. We used logistic regression to estimate odds ratio (OR) and 95% confidence interval (CI) for low birth weightterm·dom, SGA, and preterm birthterm·dom. The PFOSterm·key and PFOAterm levels were entered into the analyses as both continuous and categorical variables (< 25th, 25th–< 50th, 50th– < 75th, ≥ 75th). Plasma levels below the 25th percentile were used as the reference group when using categorical variables. Both log-transformed and untransformed blood PFC concentrations were tested in the statistical models, and the findings were similar. For this article, we used only the results of untransformed concentrations. We also conducted regression diagnostics for influential points using Cook’s distance, but no substantial changes in estimates were observed before or after the influential points were excluded, and our results are therefore all based on the whole data set.

We adjusted for risk factors that could influence fetal growthterm·key or length of gestation at birth. These factors included maternal age, parity, socio-occupational status, prepregnancy BMI, smoking during pregnancy, infant sex, and gestational week at blood drawing. For birth weightterm·mth, we also adjusted for gestational age at birth, measured in days and entered as linear and quadratic terms into the model to capture the reduced weight gain during the last weeks. All covariates but gestational age at birth and gestational week at blood drawing were introduced into models as indicator variables. Stratified analyses by parity, prepregnancy BMI, term of birth (preterm, term, and post-term) and sex were also performed to evaluate the consistency of the associations across strata.

Results Demographic and other characteristics of the study participants are presented in Table 1 . Almost half of the women were having their first babies, and the average age at delivery was about 30 years. Almost 25% smoked in early pregnancy, but one-third of these reported quitting smoking during pregnancy. Almost 30% had a prepregnancy BMI of ≥ 25 kg/m 2 . The mean birth weightterm·mth was 3,623 g, and the incidence of LBW was 1.7%; the mean gestational age at birth was 280 days, and the incidence of preterm birthterm·dom and post-term birth was 3.8% and 9.9%, respectively ( Table 2 ). On average, birth weightterm·mth increased with maternal age, parity, and prepregnancy BMI and was lower among smokers ( Table 1 ).

All the first maternal blood samples had PFOSterm·key levels above the LLOQ, and only one had a PFOAterm level below the LLOQ. The plasma concentrations varied substantially, with a range of 6.4–106.7 ng/mL for PFOSterm·key values, and less than the LLOQ to 41.5 ng/mL for PFOAterm values. PFOSterm·key and PFOAterm levels decreased with increasing parity. The mean concentration of PFOAterm was highest in the age group of < 25 years (6.2 ng/mL), and lowest in the age group of ≥ 35 years (5.1 ng/mL), but after adjustment for parity the difference was reduced to 0.4 ng/mL. High levels were also observed in overweight and obese women ( Table 1 ).

Figure 1 shows scatterplots for the first maternal blood PFOSterm·key and PFOAterm levels and birth weightterm·mth. The adjusted association between PFOSterm·key levels and birth weightterm·mth was close to null, but birth weightterm·mth was inversely associated with plasma levels of PFOAterm in the first blood sample ( Table 3 ). The unadjusted regression coefficient was −20.52 (95% CI, −31.49 to −9.56) ( Figure 1 ), but adjustment for potential confounders reduced the estimate to −10.63 (95% CI, −20.79 to −0.47) ( Table 3 ). Parity was the covariate that most markedly changed the regression coefficient. PFOAterm levels were correlated with PFOSterm·key levels ( r = 0.62), and after adjustment for PFOSterm·key levels in the first maternal blood, the regression coefficient of PFOAterm became slightly stronger (β = −14.85; 95% CI, −27.67 to −2.02). Analyses stratified by term of birth (preterm, term, post-term) revealed that the effects of PFOAterm on birth weightterm·mth were more pronounced in preterm and post-term babies ( Table 3 ), but did not differ markedly by sex or parity. PFOAterm levels were significantly inversely associated with birth weightterm·mth only among normal-weight women.

Infants born to mothers with PFOAterm levels in the three highest quartiles had an adjusted average birth weightterm·mth of 96, 98, and 105 g lower than infants born to mothers in the lowest quartile, respectively ( Table 4 ). No statistically significant association was observed between either PFOSterm·key or PFOAterm and length of gestation ( Table 4 ). Elevated risk estimates for preterm birthterm·dom were observed at all levels of PFOSterm·key and PFOAterm, but only the ORs for the third quartile of PFOSterm·key and the second quartile of PFOAterm were statistically significant ( Table 5 ). We did not observe any statistically significant association between LBW or SGA and maternal PFOSterm·key or PFOAterm levels ( Table 5 ).

There was a high degree of correlation for PFOSterm·key or PFOAterm levels between the first and second maternal blood samples ( r = 0.87 for PFOSterm·key and 0.88 for PFOAterm) ( Figure 2 ), but the average concentrations were lower in the second blood samples. PFOSterm·key and PFOAterm concentrations in cord and maternal blood were also correlated ( r = 0.72 for PFOSterm·key and 0.84 for PFOAterm if two outliers were not included) ( Figure 3 ), but cord levels were significantly lower than maternal levels. The mean ratios between first maternal PFOSterm·key and PFOAterm and cord concentrations were, respectively, 3.40 and 1.83, but decreased between second maternal and cord concentrations to 2.96 and 1.46.

Discussion To our knowledge, this is the first nationwide population-based study on birth weightterm·mth and length of gestation with prospectively collected data on levels of perfluorinated chemicals in maternal sera drawn early in pregnancy. We found only maternal plasma PFOAterm levels to be inversely associated with birth weightterm·mth. Risk estimates for preterm birthterm·dom did not increase monotonically across quartiles of PFOSterm·key or PFOAterm exposure. For all four outcomes evaluated, the outcome in the lowest exposure quartile differed from outcomes in the upper three quartiles, but there was no evidence of a dose–response association above the 25th percentile.

Maternal plasma levels in the present study (mean concentrations: PFOSterm·key, 35.3 ng/mL; PFOAterm, 5.6 ng/mL) are similar to most levels reported for U.S. populations, but much lower than those found in fluorochemical workers or residents living near the fluoropolymer production facilities ( Emmett et al. 2006a , 2006b ; OlsenpersName et al. 2003 ). We found PFOSterm·key and PFOAterm levels higher than those seen among pregnant women in other countries ( Inoue et al. 2004 ; Karrman et al. 2007 ; Midasch et al. 2007 ), but that may reflect blood drawing at different gestational week, or use of different assay laboratory technologies.

This study has several strengths. We used state-of-the-art laboratory facilities, rigorous blinding, and hospital records for obtaining data on birth weightterm·mth and gestational age. We found a high correlation between the concentrations of PFOSterm·key and PFOAterm taken over pregnancy time, but the levels declined over pregnancy time, possibly related to blood volume expansion and decreased serum albumin concentration during pregnancy ( Swanson and King 1983 ; Thibodeaux et al. 2003 ), changes in PFC disposition and pharmacokinetics during pregnancy ( Frederiksen 2001 ), or placental transfer of PFCs to the fetus ( Apelberg et al. 2007a ; Inoue et al. 2004 ; Midasch et al. 2007 ; Tittlemier et al. 2004 ). We also found a high correlation between maternal and cord blood concentrations. In this study we used only one measurement of PFOSterm·key and PFOAterm, and it is unknown how well this marker reflects an etiologically relevant dose, which may be the average dose over the entire pregnancy duration, a concentration at a different time in pregnancy, or the concentration in fetal blood. These concentrations do vary with gestational age, but whether this is important for a potential fetotoxic effect is not known.

Our nationwide study has the advantage of being selected from a well-described cohort of > 90,000 pregnant women and their infants ( OlsenpersName et al. 2001 ). Selection bias related to PFOAterm and PFOSterm·key levels or birth weightterm·mth is unlikely, because participants were enrolled early in pregnancy and had no information on any of these variables at that time. Information on gestational age and birth weightterm·mth came from the Danish Hospital Discharge Register and was based on standard clinic procedures using ultrasound based estimates and electronic weights. We also benefited from extensive data on potential confounders. Besides the covariates in our final model, we evaluated possible confounding by alcohol drinking, fish intake, protein, fat, carbohydrate, and energy intake. Data were missing on dietary covariates for around 20% of births, but the results remained materially the same after these adjustments.

Adjusting for parity and prepregnancy BMI substantially changed the crude measures of association. For instance, adjustment for only parity in the regression model led to attenuation of the regression coefficient between PFOAterm and birth weightterm·mth from −20.5 to −7.2. We did find substantially higher PFOAterm levels in nulliparous women than in multiparous women, which has been shown before in cord blood ( Apelberg et al. 2007a ), perhaps related to fetal uptake during pregnancy and excretion during lactation.

Unlike the more lipophilic POPsterm·dom, PFOSterm·key and PFOAterm do not typically accumulate in the lipid tissues because of their oil-repellent properties ( Martin et al. 2003 ). Because these compounds are retained primarily in blood and liver, PFOSterm·key and PFOAterm would not be expected to be associated with BMI. Our data did, however, demonstrate higher plasma levels of PFOSterm·key and PFOAterm among overweight or obese women. In a study among 3M male employees ( OlsenpersName et al. 1998 ), BMI was also slightly but not significantly higher in the highest PFOAterm category. In stratified analyses, PFOAterm levels were only significantly associated with birth weightterm·mth in normal-weight women, suggesting effect-measure modificationterm·mth by prepregnancy BMI of the association between PFOAterm and birth weightterm·mth.

PFOSterm·key and PFOAterm can cross the placental barrier, but cord concentrations are usually lower than those observed in maternal serum or plasma ( Inoue et al. 2004 ; Midasch et al. 2007 ; Tittlemier et al. 2004 ). In a recent study based on 11 plasma samples of mothers and the 11 corresponding cord plasma samples of neonates ( Midasch et al. 2007 ), PFOSterm·key cord concentrations were significantly lower than the maternal values, but 9 of 11 neonates had higher PFOAterm plasma level compared with those of their mothers. We found lower mean concentrations of PFOAterm and PFOSterm·key in cord blood than in maternal blood, but PFOAterm levels in cord blood were much closer than PFOSterm·key to maternal values, which may indicate that PFOAterm crosses the placental barrier more easily than PFOSterm·key. If so, that may partly explain why we observe only an inverse association between PFOAterm and birth weightterm·mth, although animal studies, as well as one cord blood study in humans ( Betts 2007 ), indicate an effect of both chemicals.

The association between PFOAterm and birth weightterm·mth may or may not have public health importance, even if it is causal. LBW has been associated with higher mortality and morbidity that is not restricted to early years of life ( Glucksman and Hanson 2006 ), although this may reflect underlying causes of fetal growthterm·key impairment rather than LBW itself. Furthermore, the observed association may not be causal. We did not find a monotone dose–response relation, but our results for birth weightterm·mth, risk of preterm delivery, LBW, and SGA are consistent with a threshold effect. Chance may play a role in some of our findings, and it is also possible that there are factors associated with both lowest PFOAterm levels and high birth weightterm·mth that were not taken into account in our statistical analyses.

The incidence of LBW and preterm birthterm·dom in our study population is relatively low due to the selection criteria used for sampling the sub-cohort. This should be taken into consideration when trying to extrapolate our results to other populations. We used clinical recordings of gestational age that were often ultrasound-based. Ultrasound-based estimates are biased if the exposure of interest reduces very early fetal growthterm·key. This may explain to some extent the higher incidence of preterm birthterm·dom we see among the exposed, but analysis of gestational age by exposure category did not provide evidence for such an effect, and such a bias would not explain differences in mean birth weightterm·mth.

We observed an inverse association between PFOAterm levels and birth weightterm·mth, although the maternal concentrations were orders of magnitude lower than those at which developmental effects have been observed in the animal toxicity studies. A risk characterization of PFOAterm shows average sera concentrations associated with the lower 95% confidence limit of the benchmark dose for 10% response levels for various developmental outcomes were at least 29,000 ng/mL in rat ( Butenhoff et al. 2004a ). The approximate maternal sera concentrations at which a significant reduction of fetal body weight occurred in mice were much higher, at 175,000 ng/mL ( Lau et al. 2006 ). Contrary to the relatively rapid rates of elimination reported in laboratory animals, PFOAterm is eliminated slowly in humans ( Kennedy et al. 2004 ). Therefore, it is hard to make a comparison with animal data, and the necessary toxicokinetic data in humans are lacking. Still, the adverse effect on birth weightterm·mth observed at low levels of exposure in our study combined with ubiquitous PFOAterm contamination in human blood may be a cause for concern.

In conclusion, our nationwide cohort study suggests an inverse association between maternal levels of PFOAterm and birth weightterm·mth. Although PFOSterm·key and PFOAterm have been phased out by some manufacturers in the United States and in some European countries, they are still produced elsewhere, and other similar compounds are believed to break down to PFOSterm·key or PFOAterm in the environment ( Organisation for Economic Co-operation and Development 2002 ; U.S. EPA 2006 ). If these compounds reduce fetal growthterm·key, continued exposure will be of public health concern, and other potential reproductive effects should be studied.