Time trends and individual characteristics associated with polybrominated diphenyl ethers in breast milk samples 2006–2009 in Lower Saxony, Germany

International Journal of Hygiene and Environmental Health 215 (2012) 352–359

Contents lists available at SciVerse ScienceDirect

International Journal of Hygiene and Environmental Health journal homepage: www.elsevier.de/ijheh

Time trends and individual characteristics associated with polybrominated diphenyl ethers in breast milk samples 2006–2009 in Lower Saxony, Germany Michael Hoopmann a,∗ , Urs-Vito Albrecht b , Edith Gierden a , René Huppmann a , Roland Suchenwirth a a b

Governmental Institute of Public Health of Lower Saxony (Niedersächsisches Landesgesundheitsamt), Roesebeckstraße 4-6, D-30449 Hannover, Germany Institute of Legal Medicine, Hanover Medical School, Carl-Neuberg-Str. 1, D-30625 Hannover, Germany

a r t i c l e

i n f o

Article history: Received 7 April 2011 Received in revised form 23 August 2011 Accepted 31 August 2011 Keywords: Brominated flame retardants PBDE Polybrominated diphenyl ethers Breast milk

a b s t r a c t Background: Since 2006 polybrominated diphenyl ether (PBDE) concentrations have been analyzed within the scope of the breast milk project conducted by the Governmental Institute of Public Health of Lower Saxony. Objectives: Temporal trends and regional distributions of the resident population as well as the relevance of individual factors influencing PBDE concentration were to be determined. Methods: Four PBDE congeners (BDE-47, BDE-153, BDE-99, BDE-100) have been analyzed. The concentrations are fitted by linear regression models, whereby individual factors of the mother are surveyed by a standardized questionnaire. Results: A total of 2173 samples taken between 2006 and 2009 shows an estimated total PBDE mean value of 1.68 ng/g lipid weight (l.w.). In contrast to most other studies, the proportion of BDE-153 exceeds the one of BDE-47 (median: 0.51 ng/g l.w. vs. 0.31 ng/g l.w.). BMI shows a positive correlation with BDE-47 and a negative correlation with BDE-153, both statistically significant (p < 0.001). For BDE-153, other significant factors (former breast feeding periods, birth year of the mother and country of birth) reflect also dilution effects and the time of accumulation. A decreasing temporal trend is observed for BDE-47 but not for BDE-153. Conclusions: The correlation patterns, the temporal trends and the various influencing factors may reflect differences in exposure sources and/or metabolism between the major congeners BDE-47 and BDE-153. Therefore it seems to be necessary to discuss the concentrations of BDE-47 and BDE-153 separately as leading indicators instead of using a total PBDE. © 2011 Elsevier GmbH. All rights reserved.

Background The Lower Saxony breast milk program Breast milk has certain advantages as a sampling matrix: sampling is simple and noninvasive, with samples collected by the mother. It monitors body burdens in reproductive-age women and estimates in utero and nursing infant exposures, all of which is important to community health. Because the concentration of contaminants in breast milk correlates with the concentrations in the fatty tissue of the maternal organism, time-trend data from breast milk monitoring can serve as a warning system that identifies chemicals whose body burdens and human exposure are increasing (Hooper and She, 2003). Furthermore, the relevance of individual factors on the contaminant concentration can be estimated, as can

∗ Corresponding author. Tel.: +49 511 4505 342; fax: +49 511 4505 4342. E-mail address: [email protected] (M. Hoopmann). 1438-4639/$ – see front matter © 2011 Elsevier GmbH. All rights reserved. doi:10.1016/j.ijheh.2011.08.019

reference values for the analyzed contaminant concentration in breast milk. Since 1999, mothers have been able to have their milk investigated at the Governmental Institute of Public Health of Lower Saxony (NGLA). Lower Saxony is one of the 16 states of Germany which is situated in the northwest of the country and neighbours the Netherlands. The central orientation of the program is the analyticbased nursing recommendation, which informs the mother regarding the burden on her milk. The routinely detected contaminants include polychlorinated biphenyls (PCBs), pesticides and fragrance chemicals. Recent years show a continual decrease in the mean contaminant concentration in breast milk in Lower Saxony (Funcke et al., 2004; Zietz et al., 2008). Altogether, from 1999 to 2009 5976 breast milk data records from mothers living in Lower Saxony were collected, which in addition to the findings report contains information from a questionnaire regarding possible individual factors. Participants were recruited through flyers and informational material available in doctors’ offices (gynecologists or pediatricians) or on the Internet.

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359

PBDE Since 2006, the Lower Saxony breast milk study has routinely detected polybrominated diphenyl ethers (PBDEs). These bromine-containing organic compounds are high production volume chemicals, which were widely used as flame retardants for electronic equipment, carpeting, and polyurethane foam in furniture and other products. The chemical structure of PBDE is very similar to that of PCB. However, in PBDE the two phenyl groups are connected to each other by an oxygen bridge. Because of the different number and position of bromine substituents at the two phenyl rings, theoretically a total of 209 congeners divided into ten homologous groups of mono- to decabromodiphenyl ethers are possible (Birnbaum and Cohon Hubal, 2006). Variant technical mixtures of PBDE have been produced, especially pentabromodiphenyl (PeBDE), octabromodiphenyl (OBDE) and decabromodiphenyl ether (DBDE). The congeners of these mixtures are classified according to the number of bromine atoms in their congeners: BDE-47 is a dibromodiphenyl ether, BDE-99 and BDE-100 are pentabromodiphenyl ethers and BDE-153 a hexabromodiphenyl ether (Darnerud et al., 2001; Birnbaum and Cohon Hubal, 2006). Sanders et al. (2005) discuss dioxin-like contaminants in these commercial mixtures. PDBEs are relatively stable in the environment, with a potential for bioaccumulation in the food chain, which increases with the degree of bromination. Because of their lipophilic properties, PDBEs can be absorbed by humans through burdens in foodstuffs or the air and accumulated in fatty tissue. For congener profiles observed in the environment, both PBDE exposure and pharmacokinetics play a role (Birnbaum and Cohon Hubal, 2006). Accordingly, in the environment BDE-209 partially metabolizes to less brominated congeners (Costa et al., 2008; WDEH, 2006). In the general population, several potential routes of PBDE exposure have been identified (Herbstman et al., 2007): (a) direct contact with treated products; (b) dietary, as these lipophilic compounds are stored in the fat of meats, fish and dairy products, (c) ingestion and to a lesser degree, inhalation of PBDE-laden house dust. The estimated half-lives of the PBDE congeners in humans are quite high: BDE-47 1.8–3.0 years, BDE-99 2.9–5.4 years, BDE-153 6.5–11.7 years (Herbstman et al., 2007). None of the commercial mixtures are genotoxic (Birnbaum and Cohon Hubal, 2006). In general, PeBDE is more toxic than OBDE, whereas DBDE is essentially non-toxic to invertebrates. However, there is growing concern for developmental and endocrine effects (Birnbaum and Cohon Hubal, 2006) Worldwide concern regarding PDBE bioaccumulation and toxicity led to an incremental ban on the chemical. In Sweden, its production and use were prohibited in 1999. In 2003, PeBDE and OBDE were prohibited throughout Europe. This ban was implemented in Germany by the Chemicals Prohibition Ordinance in 2004. In 2004, production of PeBDE und OBDE was also halted in the USA (Birnbaum and Cohon Hubal, 2006). At the fourth Conference of the Parties at the Stockholm Convention in May 2009, PBDE and OBDE were added to the list of banned substances. Since May 2010, their manufacture and use have been banned worldwide. Due to reduced production and use, at least in Europe, environmental concentrations of these substances have been partially decreasing (Lind et al., 2003; Fangström et al., 2005a). Population-based studies of PBDE in breast milk The North American studies of PBDE concentrations in the environment or humans clearly indicate the highest concentrations in the world. This also affects the proven PBDE content in breast milk, especially of BDE-47 (Schecter et al., 2003; Ryan et al., 2006;

353

Johnson-Restrepo et al., 2007; She et al., 2007; Daniels et al., 2010). In samples taken in the northwest Pacific region in 2003, at a peak value for total PBDE of 321 ng/g l.w., the detected concentrations of PBDE were in some cases higher than that of PCB (She et al., 2007). The predominant exposure source for these concentrations in the human body, evidently higher in comparison to European or Asian studies, was primarily attributed to ingestion of house dust (Lorber, 2008). Yet also within Europe, the flame retardants were used to different degrees or their distribution was restricted at different times, so various concentrations in separate European nations exist at various measurement time points. In Table 1, individual study results are displayed in synoptic opposition to indicate both regional as well as temporal differences. Only publications that were based on more than 30 breast milk samples (partly analyzed as mixed samples) were considered. (A 2003 Norwegian study with 38 samples gave no information about the individual congeners (total PBDE: 2.96 ng/g l.w.) and so is excluded from the table (Polder et al., 2004).) The four congeners BDE-47, BDE-99, BDE-100 and BDE-153 are presented separately. Because distribution of BDE-209 with the corresponding technical mixture decabromodiphenyl was restricted much later, studies of its concentration in breast milk are lacking. Still, higher concentrations of BDE-209 have already been detected in human blood than milk, as indicated in corresponding comparative studies (Kawashiro et al., 2008; Inoue et al., 2006). There is no indication of a total PBDE in Table 1 because over the individual studies, a different number of congeners were included in totaling. Most notably, the comparison for milk fat-related main congeners BDE-46 and BDE-153 is of major interest. Certain studies have analyzed individual samples as mixed samples (Fangström et al., 2005a; Ingelido et al., 2007), which meant conclusions could only be drawn regarding the mean PBDE concentrations, and not their median values. A special case is the national survey in China, in which in 12 provinces in each part of a rural and urban area pooled mild samples corresponding to 30–65 participants were analyzed. This allows the detection of regional differences. For the pool samples, BDE-47 has a range from 0.21 to 0.73 ng/g l.w., and BDE-153 has a range from 0.10 to 1.14 ng/g l.w. Another important criteria for comparison of the studies is whether the study is restricted to mothers nursing for the first time [“primaparae”] (Lind et al., 2003; Baumann et al., 2003; She et al., 2007). In addition to the temporal or regional effects discussed above, the different concentrations recognized in Table 1 may also be due to varying distributions of test person characteristics (e.g. nutritional habits, age, occupational exposure, weight, number of previous nursing periods) and/or the respective milk sample. Also the form of recruiting can be relevant for interpretation. Thus, previous nursing duration (sampling after birth) in studies which acquired breast milk samples through cooperation with gynecologic and maternity clinics (e.g. Kazda et al., 2004; Baumann et al., 2003) is on average shorter than that of studies using other approaches. Other inclusion criteria or random sample characteristics affect the time point of sample acquisition or individual nutritional characteristics. In the English study (Kalantzi et al., 2004), for the Lancaster region alone, a sample time point within the first month after start of nursing was specified as inclusion criterion. A German study with particular relevance to the evaluation of our results (Vieth et al., 2005) had the primary goal of facilitating comparison of PDBE concentrations in vegetarians to those with an omnivorous diet. Thus, vegetarians are overrepresented here in comparison to their actual proportion in the population. The mixed samples in the Italian studies (Ingelido et al., 2007) were combined according to the mother’s fish consumption.

354

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359

Table 1 Mean PBDE concentration in breast milk: overview to studies of PDBE in human milk (in ng/g l.w.). Land/region

Sampling year

Sweden/Stockholm Sweden/Stockholm Sweden/Stockholm Sweden/Stockholm Sweden/Uppsala Netherlands Italy/Venice Italy/Rome Germany/Berlin England/London, Lancaster Czech Republic/Olomounc Norway Japan Taiwan China USA/Canada – Pacific Nw. USA/Massachusetts USA/Texas USA/North Carolina

1980 1988 1996 2004 1996–99 1998 1998–2000 2000–2001 2001–2004 2001–2003 2003 2003–2005 2005 2007–2008 2007 2003 2004 2002 2004–2006

a b c d

n 116a 20a 20a 20a 93 103 29a 10a 89 52 103 393 89 32 1237d 40 38 47 303

BDE-47 0.12 0.57 2.0 0.93 2.35 1.53 1.03 1.9 0.91 3c 0.86 1.7 0.41c 0.58 0.43 50.1 40.7 40.8 28b

BDE-99 0.06 0.26 0.74 0.26 0.62 0.53 0.40 0.97 0.38 0.9c 0.28 0.49 0.08c 0.19 0.08 10.27 11.8 14.0 5b

BDE-100 0.044 0.16 0.43 0.29 0.38 0.37 0.21 0.48 0.26 0.6c 0.17 0.40 0.14c 0.19 0.09 11.7 6.91 8.2 5b

BDE-153 0.037 0.24 0.53 0.92 0.60 1.03 0.48 0.47 0.59 1.4c 0.19 0.56 0.34c 0.96 0.49 16.3 5.15 5.3 6b

Literature source Fangström et al. (2005a) Fangström et al. (2005a) Fangström et al. (2005a) Fangström et al. (2005a) Lind et al. (2003) Baumann et al. (2003) Ingelido et al. (2007) Ingelido et al. (2007) Vieth et al. (2005) Kalantzi et al. (2004) Kazda et al. (2004) Thomsen et al. (2010) Inoue et al. (2006) Koh et al. (2010) Zhang et al. (2011) She et al. (2007) Johnson-Restrepo et al. (2007) Schecter et al. (2003) Daniels et al. (2010)

Analyzed as pooled sample. Substitute entry of median instead of arithmetic mean. Substitute entry of geometric mean instead of arithmetic mean. Analyzed in 24 pooled samples.

Whether the inclusion characteristics or form of participant recruiting actually have a significant influence on the average PBDE content, however, remains questionable. Thus, for example, a correlation between PBDE in blood serum and nursing duration which was not seen for PBDE in breast milk could be indicated by a 2005 Japanese study (Inoue et al., 2006) whose 89 milk samples (human milk and blood serum) came from four study regions (Hokkaido, Miyagi, Gifu and Hyogo). Here, the PBDE concentration was comparable to European studies and so was clearly less than that indicated by North American studies (median total PBDE 1.39). Also, as in the English or Italian studies (Kalantzi et al., 2004; Ingelido et al., 2007), region emerges as a decisive cause variable in Europe. Methods PBDE determination A method which facilitated quantification of the proportionally dominant congeners by low resolution mass spectrometry (HRGCLRMS) was established for routine operation. The methods used by the NGLA required first isolating the polybrominated diphenyl ether and then quantifying it by GC–MS. Here, negative chemical isolation was used (Hoopmann et al., 2009). In contrast to this quantification method using low resolution mass spectrometry, other studies primarily used high resolution mass spectrometry to measure polybrominated diphenyl ether (cf. Vieth et al., 2005; Ingelido et al., 2007). After the method was developed, a total of 20 samples were sent to a competent laboratory (Eurofins) in 2005 and 2008 for comparative determination with high resolution mass spectrometry. This indicated a good agreement between the two measurement processes (Hoopmann et al., 2009). The total PBDE of this routine operation is estimated as the sum of the four congeners multiplied by the coefficient 1.24, which was derivated from the comparative analyses mentioned above. Statistics Linear models were used to assess whether individual characteristics were associated with the log-logarithmic of the PBDE

levels (BDE-47, BDE-153, total PBDE) and to test the differences in effect estimates for each category of each covariate. These including maternal birth year (continuous), sampling year (2006, 2007, 2008, 2009), duration of nursing (≤8 weeks postpartum; >8 weeks; >12 weeks), previous nursing periods (none; one; two and more), maternal origin (“German” vs. “non-German”), current maternal BMI (normal, ≤25.0 kg/m2 ; overweight, >25.0 kg/m2 ; obese, >30.0 kg/m2 ) and a dummy indicator variable for Lower Saxony resident (less than/at least 10 years). Because PBDEs are stored in fat and during pre-pregnancy and therefore BMI reflects a woman’s long-term fat stores, the use of pre-pregnancy BMI instead of current BMI seems preferable, but – as shown by Daniels et al. (2010) – differences in the effect estimates are marginal. The potential existence of interactions between the listed factors was also tested. In addition, mothers born in Germany between 1968 and 1981 and nursing for the first time were considered as a homogeneous subgroup in corresponding multivariate linear models: All statistical analyses were conducted with the statistical package SYSSTAT© , Version 11.

Results Description In all, there are 2173 PBDE determinations from the years 2006 to 2009, wherein the highest PDBE finding, whose corresponding PBDE total was more than twice as high as the next lowest, was interpreted as an outlier and excluded from the following results display. This concerns a sample from 2008, from a mother of German origin nursing for the first time whose total PBDE was estimated at 55.05 ng/g l.w. Here, the congeners BDE-153 (36.22 ng/g l.w.) and BDE-100 (5.49 ng/g l.w.) were conspicuous. For the remaining data, median, range and the empirical 95% quantile of the four specified PBDE congeners are displayed in Table 2 along with the extrapolated PBDE total concentrations. A total of 1018 congener concentrations could not be precisely specified because they were below the quantification limit [BDE-47: 11; BDE-99: 428; BDE-100: 572; BDE-153:7]: These concentrations were uniformly assigned the value 0.03 ng/g l.w., which

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359

355

Table 2 Overview of the calculated PDBE content (in ng/g l.w.) according to sampling year: entire data pool: mean, median( 0.5 ), 0.95 quantile ( 0.95 ), range. Sample year (observations)

BDE-47

BDE-99

BDE-100

BDE-153

Total PBDE

2006–2009 (n = 2173)

Mean  0.5  0.95 Range

0.50 0.31 1.30 0.03–15.49

0.15 0.09 0.40 0.03–6.07

0.12 0.09 0.33 0.03–2.05

0.58 0.51 1.18 0.03–6.10

1.68 1.30 3.55 0.15–27.49

2006 (n = 507)

Mean  0.5  0.95 Range

0.60 0.37 1.60 0.03–11.42

0.17 0.11 0.53 0.03–2.65

0.14 0.10 0.37 0.03–1.60

0.64 0.55 1.29 0.03–6.10

1.92 1.50 4.57 0.15–20.02

2007 (n = 571)

Mean  0.5  0.95 Range

0.52 0.33 1.29 0.03–9.25

0.14 0.09 0.40 0.03–3.03

0.11 0.08 0.31 0.03–0.94

0.58 0.51 1.22 0.03–2.37

1.67 1.31 3.53 0.33–17.17

2008 (n = 652)

Mean  0.5  0.95 Range

0.47 0.29 1.28 0.03–15.49

0.15 0.10 0.39 0.03–3.65

0.13 0.10 0.34 0.03–1.96

0.59 0.52 1.17 0.03–3.78

1.67 1.31 3.35 0.38–27.47

2009 (n = 443)

Mean  0.5  0.95 Range

0.40 0.24 0.92 0.03–13.31

0.12 0.08 0.27 0.03–6.07

0.11 0.08 0.27 0.03–2.05

0.51 0.47 0.96 0.03–4.24

1.42 1.12 2.77 0.15–27.49

on average corresponds to approximately half the determination limit, which is dependent on the individual fat content of the mother. The concentrations over the 4 years vary most starkly for BDE47. Here, a strict negative trend seems to exist for mean, median, and 0.95 quantile. BDE-47 and BDE-153 dominate the congeners. BDE-153 can even be detected in stronger concentrations, which has not been indicated in any other major European study to date (c.f. Table 1). In international comparison, the estimated total PBDE concentration of 1.8 ng/g l.w. is very low. The values for BDE-47, BDE-99 and BDE-100 are – if the Swedish result from the first years of sampling since 1980 (Fangström et al., 2005a) is disregarded – in part clearly below the results from other European countries and the North American studies. Only recent studies from Asia (Japan, Taiwan, China) evince a similarly low level (cf. Table 1). Comparison with the earlier German study (Vieth et al., 2005) indicates the BDE-153 concentrations are nearly the same, while in the older study the concentrations of the remaining three

congeners are noticeably higher. The relationship of the mean BDE47 to that of BDE-99, as in nearly all of the studies listed in Table 1, is about three to one. The same statistics as in Table 2 are displayed in Table 3 for the subgroup of mothers of German origin nursing for the first time, birth cohort 1968–1981. The descriptive analysis for this subgroup yielded minimally higher values for the measures of location. Multivariate analysis – identification of influencing factors As described, multivariate regression models were compiled to allow the identification of influencing factors under control of the remaining regressors. Table 4 displays the distribution of the descriptors used in the following linear models. In the overview, Table 5 indicates which factors have a statistically significant correlation with total PBDE, BDE-47 or BDE-153 (logarithmized concentration value) in the multivariate analysis. The corresponding p-value (e.g. hypothesis “the regression coefficients of the different categories of the factor are identical”)

Table 3 Overview of the detected PBDE content (in ng/g l.w.) according to sampling year. First time nursing mothers of German origin, birth cohort 1968–1981. Sample year (observations)

BDE-47

BDE-99

BDE-100

BDE-153

Total PBDE

2006–2009 (n = 903)

Mean  0.5  0.95 Range

0.54 0.32 1.44 0.03–15.49

0.15 0.10 0.40 0.03–3.65

0.14 0.10 0.39 0.03–1.96

0.63 0.55 1.21 0.03–6.10

1.80 1.39 3.81 0.15–27.47

2006 (n = 220)

Mean  0.5  0.95 Range

0.70 0.39 1.79 0.03–11.42

0.19 0.11 0.64 0.03–2.34

0.16 0.11 0.53 0.03–1.60

0.69 0.61 1.30 0.03–6.10

2.15 1.55 5.98 0.15–20.02

2007 (n = 223)

Mean  0.5  0.95 Range

0.52 0.36 1.27 0.03–8.99

0.14 0.09 0.37 0.03–2.90

0.12 0.09 0.39 0.03–0.93

0.62 0.55 1.25 0.03–2.37

1.78 1.44 3.99 0.45–16.40

2008 (n = 274)

Mean  0.5  0.95 Range

0.54 0.30 1.50 0.07–15.49

0.16 0.10 0.39 0.03–3.65

0.15 0.10 0.43 0.03–1.96

0.62 0.55 1.17 0.17–1.98

1.83 1.39 3.98 0.51–27.47

2009 (n = 186)

Mean  0.5  0.95 Range

0.37 0.24 1.00 0.03–3.80

0.11 0.08 0.27 0.03–3.20

0.11 0.08 0.27 0.03–1.82

0.57 0.52 1.02 0.03–4.24

1.43 1.18 2.77 0.15–12.23

356

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359

Table 4 Distribution of the variables used as descriptors in the linear models.

Maternal birth year Mean Range Sampling year 2006 2007 2008 2009 Previous nursing periods None One Two or more BMI To 25 >25–30 (overweight) >30 (obese) Maternal origin Native country Germany Non-German origin Previous nursing duration Up to 8 weeks >8–12 weeks Over 12 weeks Lower Saxony resident Less than 10 years At least 10 years

Entirety

Subgroup

1975.4 1960–1990

1975.7 1968–1981

23.3% 26.3% 30.0% 20.4%

24.4% 24.7% 30.3% 20.6%

55.1% 33.7% 11.3%

100%

60.5% 28.2% 11.3%

62.6% 26.3% 11.1%

89.2% 10.8%

100%

25.6% 19.0% 55.4%

23.4% 18.8% 57.7%

31.2% 68.8%

32.3% 67.7%

Table 6 Direction and statistical significance of factors: first time nursing mothers of German origin only. Factor

is entered as soon as it is at least 0.1: o = p-value ≤ 0.1, * = pvalue ≤ 0.05, ** = p-value ≤ 0.001, *** = p-value ≤ 0.001. In this way, the continuous variable mother’s birth year (1 DF) was tested for a linear correlation between birth year and logarithmized PBDE concentration. For the remaining variables, whether the PBDE concentrations differ between the categories of the factor was tested. The direction of the investigated influencing factors for BDE153 resemble those for the classical organic chlorine compounds (Hoopmann et al., 2009): previous nursing period, non-German maternal origin, relative maternal weight (BMI) and birth year display a clear statistical negative correlation with the BDE-15 [pvalue < 0.001]. In contrast, the models for the BDE-47 exhibit partially different regressors: birth cohort and a non-German origin seem to have no effect on BDE-47. Compared to BDE-153, BMI and the indicator “Lower Saxony resident for at least 10 years” exhibit an opposite correlation direction. These opposing effects of the congeners for BMI yield no indication of correlation between total PBDE and BMI. No consistent effect on PBDE content could be identified for the previous nursing duration of the current nursing period. The test of whether interaction between the factors exists delivers a statistically significant interaction between maternal origin and duration of their residence in Lower Saxony for BDE-153 [see Table 5, column 4]. While the length of time the mother has resided in Lower Saxony is irrelevant for mothers of German origin, there

Model for: ln(PBDE)

Maternal birth year Sampling year (2006–2008) ↓*** Nursing time since birth Maternal BMI (current) Lower Saxony resident at least 10 years 0.051 Goodness-of-fit R2

ln(BDE-47)

ln(BDE-153)

↓***

↓** **

↑***

↓***

0.081

0.104

are differences between mothers of non-German origin regarding the length of time they have resided in Lower Saxony. The same dominant cause variables emerge when the subgroup, first time nursing mothers of German origin, is analyzed (Table 6). Because the models for first time nursing mothers of German origin are based on less than half of the total observations, the statistical significance of the exhibited effects is reduced despite comparable regression coefficients. These regression coefficients are listed individually for Table 6. To supplement Tables 5 and 6, which only describe the significance of the factors, a potential significance of the individual categories in contrast to the respective reference categories is given (Table 7). In comparison to the models for the entire data report or for the subgroup and except for the variable “Lower Saxony resident for at least 10 years”, the estimators for the regression coefficient are strongly similar. Presumably, this is explained by the comparatively large number of participants of non-German origin residing in Lower Saxony for less than 10 years, so that interactions occur here. As well, a stronger decline in BDE-47 is noted when the subgroup is observed. In all, the models present a high statistical significance, i.e. the selected factors contribute strongly to explaining the variability of the PBDE concentrations, even when the coefficients of determination R2 as measures of goodness-of-fit of the linear models are not excessively high. Discussion of results Mean concentration in NLGA program Three main results are: • In comparison to other (older) studies, the proven concentrations of PBDE in the Lower Saxony breast milk program are relatively low. • The mean PBDE-153 concentration is higher than that of BDE-47. • Concentrations for BDE-47, BDE-99 and BDE-100 decrease over the four sampling years, but not to the same degree as BDE-153.

Table 5 Direction and statistical significance of the applied regressors: entire model. Factor (degrees of freedom)

ln(PBDE)

Maternal birth year Sampling year (2006–2009) (3 df) Duration of nursing (2 df) Previous nursing periods (2 df) Non-German maternal origin Maternal BMI (current) (2 df) Lower Saxony resident at least 10 years Interactiona Goodness-of-fit: R2

↓o ↓***

a

ln(BDE-47)

ln(BDE-153)

ln(BDE-153)

↓***

↓*** ***

↓*** *** ↓o ↓*** ↓*** ↓*** ↑*** ↑** 0.198

↓*** ↓*** *

↓*

0.070

0.061

Multiplier effect between non-German origin and less than 10 years in Lower Saxony.

↑***

↓*** ↓*** ↓*** ↑** 0.192

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359

357

Table 7 Regression coefficients of the models for BDE-47 and BDE-153. Factor

Constant Maternal birth year Sampling year (reference: 2006) Nursing duration (reference: to 8 weeks) Previous nursing periods (reference: none) Non-German maternal origin Maternal BMI (reference: max 25) Lower Saxony resident at least 10 years

Category

2007 2008 2009 8–12 weeks Over 12 weeks One Two and more 25–30 30 and more

Given the exposure reduction resulting from the pertinent substance regulations (production and distribution), this observed 4 year trend of the congeners is plausible. The decline of BDE-47 (along with its associated congeners), BDE-99 and BDE-100 could already be seen in the 1980–2004 Stockholm analysis of stored mixed samples. In particular, the concentration peak for BDE-47 occurred in Stockholm during the mid-90s, while the concentration of BDE-153 continued to grow until 2000 without a clear rise or fall thereafter (Fangström et al., 2005a). For BDE-47 the mixed sample concentration from 2004 was 0.93 ng/g l.w. and for BDE-153, 0.92 ng/g l.w. In relation to each other, this roughly corresponds to the ratio of BDE-47/BDE-153 in the NLGA measurements. Because it must be assumed that due to earlier national restrictions on production and distribution of PBDE in Sweden concentrations decreased earlier than those in Germany, it is plausible that the results regarding the congener pattern in breast milk are noted in Germany with a corresponding time lag. Thus, given this multiyear time interval, the comparatively low concentration in the NLGA study is thoroughly consistent with the previous German study (Vieth et al., 2005) where an evident dominance of BDE-47 is indeed noted, yet the BDE-153 concentration is already practically identical. Therefore the situation already described and hypothetically formulated by Fangström et al. (2005a,b) does exist: congeners associated with the long-banned technical mixtures PBDE and OBDE gradually disappear from the environment while the concentration of BDE-153 remains comparatively constant. In Germany, by now this process has reached the stage where BDE-153 has become the dominant congener, at least in breast milk, while the concentration of the other congeners has dropped off. In contrast, previous studies from North America and Europe – with the exclusion of a small study from the Faroe Islands (Fangström et al., 2005b) consistently displayed a markedly higher concentration of PBDE-47 compared to PBDE-153. On the contrary, some recent studies from Asia listed in Table 1 (Koh et al., 2010; Zhang et al., 2011) show the general dominance of BDE-153 in breast milk. Although there is a differential partitioning of PBDE congeners in milk and blood, several recent studies on blood samples also show the tendency towards an increased dominance of BDE-153 compared to BDE-47 (e.g. Lin et al., 2011). It can be expected that future studies, at least in Asia and Europe, will also detect this current dominance. Identified factors One of the greatest strengths of the study program was that its scope allowed the identification of even low effects of factors dependant on individual behavior (nursing duration, nursing period, maternal age, weight and origin) and that subgroups can be analyzed. Perhaps the most pertinent effects were detected

Total

Subgroup

BDE-47

BDE-153

BDE-47

BDE-153

−0.960 0.000 −0.162*** −0.288*** −0.467*** −0.052 0.030 −0.082 −0.101o −0.054 0.063o 0.282*** −0.038

26,755*** −0.014*** −0.040 −0.004 −0.162*** −0.044 −0.050* −0.065** −0.166*** −0.458*** −0.214*** −0.352*** 0.062**

−7.188 0.003 −0.233** −0.327*** −0.560*** 0.030 0.123o

30.726*** −0.014** −0.065 −0.042 −0.158** −0.054 −0.072o

0.098 0.347*** −0.054

−0.200*** −0.383*** 0.004

for BMI, because both main congeners have an opposite direction of association, so no correlation with the total BDE can be supported. To the authors’ knowledge, this negative correlation for BDE-153 has not been previously demonstrated. Indications for a negative correlation between highly brominated congeners and BMI are noted in the data from Koh et al. (2010) but not specifically for BDE-153. A similar concentration decrease along with rising BMI, yet is also found in our data for PCB (Zietz et al., 2008). Because BDE-153 correlates more strongly with the individual PCB congeners than with the remaining three PBDE congeners (Hoopmann et al., 2009), the negative correlation between BDE-153 and BMI is plausible. The literature also contains little evidence of a positive correlation for BDE-47: Koh et al. (2010) found no correlation for PentaBDE and therefore, for the dominant BDE-47. Still, these studies included few mothers who were strongly overweight. On the other hand, Daniels et al. (2010) indicate a similar rise as in our study after a BMI of 30. Apparently an effect of relative body weight on BDE-47 concentration can first be established after a threshold value. This study relied on current BMI, while other studies used the pre-pregnancy body mass index (Koh et al., 2010). The completeness of the information regarding current BMI is, however, higher in this study. As well, the results show no significant dependence on selected BMI (Daniels et al., 2010); current BMI is a good substitute for pre-pregnant BMI. As additional factor on contaminant concentration apart from BMI itself, the relative difference of prevs. post-pregnancy BMI could nonetheless be included in complementary analyses. No consistent effect could be shown for nursing duration either. Yet, longitudinal studies investigating PBDE content in breast milk at various points during nursing detected no relevant influence of nursing duration, in complete agreement with other studies. Hooper et al. (2007) analyzed time series data of 18 individual sample test series, which indicate body burdens of PBDEs are lowered by lactation, but only slowly. Other individual maternal characteristics had a much more obvious influence on PBDE content. To prove the decline over nursing duration, the authors therefore chose a regression model with random intercept (per mother) and fixed effects over the mother for the duration of nursing. The authors therefore also conclude that the PBDE content for a second nursed child is not necessarily less than for the first child. For their 303 samples, Daniels et al. (2010) could identify no influence of previous nursing time. They do indicate that nursing duration basically consisted of at least 3 months, so the effect at the beginning of nursing was not recorded. There was no statistically different change for the 83 mothers for whom two samples were present. A significant rise was only noted for BDE-153 (pvalue 0.005). Whether this effect is caused by nursing or by an

358

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359

individual behavior that changed after the first months of nursing was not discussed by Daniels et al. (2010). Statistical support for an influence of previous nursing duration would require longitudinal studies on a large study collective that also covers the first months of nursing. The estimated effects of the variable “Lower Saxony resident for at least 10 years” are not robust in a statistical sense and cannot be directly interpreted as a regional effect. In addition, no difference could be identified when residential sub-regions of lower Saxony as an alternative descriptor was used instead (data not shown). The effect seems to result more from the occurrence of yet another differentiation within the group according to the length of stay for the group of mothers of non-German origin, and so a global rather than a small regional effect contributes to significance. Results such as those indicated by Ingelido et al. (2007) or Kalantzi et al. (2004), both of which compare a major urban congested area with a control region, could not be supported by this study. Nonetheless, no individual factors could be considered during these both analyses. Thus, explicitly regional effects may be caused by strongly varying individual factors between the results regions. As in Thomsen et al. (2010), the goodness-of-fit is considerably better for BDE than for BDE-47. The lower value for the coefficient of determination for PBDE in comparison with the corresponding models for PCB may be an indication that additional factors exist. But another reason may be that the relative importance of the error parameter, which also contains measurement uncertainty and rounding errors, is much larger for PBDE. Ultimately, the concentrations for PBDE continue to remain markedly below those of PCB so that the relative measurement error, which cannot be explained by the model approach, is assumed to be higher. Forecast In order to be able to both describe contaminant behavior in the environment and in particular, to provide participant statements about individual body burden in comparison to the classical organochloride compounds of current environmental contamination, a modified PDBE analytic which enabled a cost-effective method to substantiate the relative congeners needed to be established for the routine diagnostics. In future it will also be necessary to adapt determination methods for contaminants in breast milk that have recently entered the discussion to the framework parameters of the Lower Saxony human milk program. Potential laboratory or method effects could not be completely excluded, yet at most contribute only minimally to the explanation of the congener distribution cluster or the concentration development over time, but still do not suffice as explanation of which factors are identified to be correlated with PBDE in breast milk. One of the greatest strengths of the study program was that its scope allowed limited effects of potential cause variables (nursing duration, nursing period, maternal age, weight and origin) to be identified. In particular, limited temporal or regional trends could also be covered. One side effect of the program is derivation of a reliable population-related reference value for the investigated contaminants (Hoopmann, 2010). As a whole, the identified cause variables for BDE-153 are already similar to those of the classical organic chloride compounds and plausible through bioaccumulation even if they do less to explain the variability in the individual concentration values than for PCB: the effect of nursing duration can be seen as describing a dilution effect, birth cohort or age the accumulation duration and relative body weight (BMI) the fat reservoir, while origin can be seen as an indicator for the internationally varying distribution of chemical flame retardants. However, the design of the program does not permit individual nutritional or occupational factors to be identified. In

correspondence with this approach – substantial number of samples with information regarding the donor mother restricted to a few items – the limitation to only four PBDE congeners is appropriate. An even further limitation to the main congeners BDE-47 and BDE-153 is conceivable here.

Acknowledgements The study was funded by the Federal State of Lower Saxony. This study would not have been possible without the daily effort of the NLGA environmental medical laboratory. Therefore, the authors would like to extend their most sincere gratitude to the current colleagues Gabriele Brünner-Kühn, Angela Prien, Kristin Richter, Christa Steffens, Bianca Sterenberg and Birgit Stöteknuel.

References Baumann, B., Hijman, W., van Beuzekom, S., Hoogerbrugge, R., Houweling, D., Zeilmaker, M., 2003. PBDEs in human milk from the Dutch 1998 monitoring programme. Organohalogen Compd. 61, 187–190. Birnbaum, S.B., Cohon Hubal, E.A., 2006. Polybrominated diphenyl ethers: a case study for using biomonitoring data to address risk assessment questions. Environ. Health Perspect. 114, 1770–1775. Costa, L.G., Giordano, G., Tagliaferri, S., Caglieri, A., Mutti, A., 2008. Polybrominated diphenyl ether (PBDE) flame retardants: environmental contamination, human body burden and potential adverse health effects. Acta Biomed. 79, 172–183. Daniels, J.L., Pan, I.-J., Jones, R., Anderson, S., Patterson Jr., D.G., Needham, L.L., Sjödin, A., 2010. Individual characteristics associated with PBDE levels in U.S. human milk samples. Environ. Health Perspect. 118, 155–160. Darnerud, P.O., Eriksen, G.S., Jóhannesson, T., Larsen, P.B., Viluksela, M., 2001. Polybrominated diphenyl ethers: occurrence, dietary exposure, and toxicology. Environ. Health Perspect. 109 (Suppl 1), 49–68. Fangström, B., Strid, A., Bergman, Å., 2005. Temporal trends of brominated flame retardents in milk from Stockholm mothers, 1980–2004. Report; Stockholm. Available from: http://www.imm.ki.se/Datavard/PDF/mj%C3%B6lk poolade NV%20rapport%202005%20modersmjolk.pdf (accessed 06.01.11). Fangström, B., Strid, A., Granjean, P., Weihe, P., Bergman, Å., 2005b. A retrospective study of PBDEs and PCBs in human milk from the Faroe Islands. Environ. Health 4, 12. Funcke, M., Gierden, E., Hoopmann, M., Heidrich, S., Jorritsma, U., Windorfer, A., 2004. Breast 23 Milk Analyses 1999–2003. Available from: http://www.nlga. niedersachsen.de/live/live.php?navigation id=6618&article id=19364& psm 1 and=20 (accessed 06.01.11). Herbstman, J.B., Sjödin, A., Apelberg, B.J., Witter, F.R., Patterson Jr., D.G., Halden, R.U., Jones, R.S., Park, A., Zhang, Y., Heidler, J., Needham, L., Goldman, l.L.R., 2007. Determinants of prenatal exposure to polychlorinated biphenyls (PCBs) and polybrominated diphenyl ethers (PBDEs) in an urban population. Environ. Health Perspect. 115, 1794–1800. Hooper, K., She, J., 2003. Lessons from the polybrominated diphenyl ethers (PBDEs): precautionary principle, primary prevention, and the value of communitybased body-burden monitoring using breast milk. Environ. Health Perspect. 111, 109–114. Hooper, K., She, J., Sharp, M., Chow, J., Jewell, N., Gephart, R., Holden, A., 2007. Depuration of polybrominated diphenyl ethers (PBDEs) and polychlorinated biphenyl (PCBs) in breast milk from California first-time mothers (Primiparae). Environ. Health Perspect. 115, 1271–1275. Hoopmann, M., 2010. Ableitung von Referenzwerten für PBDE in Frauenmilch. Available from: http://www.nlga.niedersachsen.de/live/live.php?navigation id=27085&article id=19364& psmand=20 (accessed 06.01.11). Hoopmann, M., Huppmann, R., Albrecht, U.-V., Gierden, E., Suchenwirth, R., 2009. Polybromierte Diphenylether (PBDE) als neue Substanzklasse im niedersächsischen Muttermilchprojekt. Umweltmed Forsch Prax 14, 183–194. Ingelido, A.M., Ballard, T., Dellatte, E., di Domenico, A., Ferri, F., Fulgenzi, A.R., Herrmann, T., Iacovella, N., Miniero, R., Päpke, O., Porpora, M.G., De Felipe, E., 2007. Polychlorinated biphenyls (PCBs) and polybrominated diphenyl ethers (PBDEs) in milk from Italian women living in Rome and Venice. Chemosphere 67, S301–S306. Inoue, K., Harada, K., Takenaka, K., Uehara, S., Kono, M., Shimizu, T., Takasuga, T., Senthilkumar, K., Yamashita, F., Koizumi, A., 2006. Levels and concentration ratios in polychlorinated biphenyls and polybrominated diphenyl ethers in serum and breast milk in Japanese mothers. Environ. Health Perspect. 114, 1179–1185, Extended results table under: http://www.ehponline.org/docs/ 2006/9032/suppl.pdf (accessed 06.01.11). Johnson-Restrepo, B., Addink, R., Wong, C., Arcaro, K., Kannan, K., 2007. Polybrominated diphenyl ethers and organochlorine pesticides in human breast milk from Massachusetts, USA. Environ. Monit. 9, 1205–1212. Kalantzi, O.I., Martin, F.L., Thomas, G.O., Alcock, R.E., Tang, H.R., Drury, S.C., Carmichael, P.L., Nicholson, J.K., Jones, K.C., 2004. Different levels of

M. Hoopmann et al. / International Journal of Hygiene and Environmental Health 215 (2012) 352–359 polybrominated diphenyl ethers (PBDEs) and chlorinated compounds in breast milk from two U.K. regions. Environ. Health Perspect. 112, 1085–1091. Kawashiro, Y., Fukata, H., Omori-Inoue, M., Kubonoya, K., Jotaki, T., Takigami, H., Sakai, S.-I., Mori, C., 2008. Perinatal exposure to brominated flame retardants and polychlorinated biphenyls in Japan. Endocr. J. 55, 1071–1084. Kazda, R., Hajslová, J., Poustka, J., Cajka, T., 2004. Determination of polybrominated diphenyl ethers in human milk samples in the Czech Republic: comparative study of negative chemical ionisation mass spectrometry and time-of-flight high-resolution mass spectrometry. Anal. Chim. Acta 520, 237–243. Koh, T.W., Chen, S.C.C., Chang-Chien, G.P., Lin, D.Y., Chen, F.A., Chao, H.R., 2010. Breast-milk levels of polybrominated diphenyl ether flame retardants in relation to woman’s age and pre-pregnant body mass index. Int. J. Hyg. Environ. Health 213, 59–65. Lin, S.-M., Chen, F.-A., Huang, Y.-F., Hsing, L.-L., Chen, L.-L., Wu, L.-S., Liu, T.-S., ChangChien, G.-P., Chen, K.-C., Chao, H.-R., 2011. Negative associations between PBDE levels and thyroid hormones in cord blood. Int. J. Hyg. Environ. Health 214, 115–120. Lind, Y., Darnerud, P.A., Atuma, S., Aune, M., Becker, W., Bjerselius, R., Cnattingius, S., Glynn, A., 2003. Polybrominated diphenyl ethers in breast milk from Uppsala County, Sweden. Environ. Res. 93, 186–194. Lorber, M., 2008. Exposure of Americans to polybrominated diphenyl ethers. JESEE 18, 2–19. Polder, A., Thomsen, C., Becher, G., Skaare, J.U., Løken, K., Eggesbø, M., 2004. The Norwegian Human Milk Study – HUMIS. Variations in levels of chlorinated pesticides, PCBs and PBDEs in Norwegian breast milk. Organohalogen Compd. 66, 2448–2454. Ryan, J.J., Wainman, B.C., Schecter, A., Moisey, J., Kosarac, I., Sun, W.F., 2006. Trends of the brominated flame retardants, PBDESs and HBCD, in human milk from North America. Organohalogen Compd. 68, 778–781.

359

Sanders, J.M., Burka, L.T., Smith, C.S., Black, W., James, R., Cunningham, M.L., 2005. Differential expression of CYP1A, 2B, and 3A genes in the F344 rat following exposure to a polybrominated diphenyl ether mixture or individual components. Toxicol. Sci. 88, 127–133. Schecter, A., Pavuk, M., Päpke, O., Ryan, J.J., Birnbaum, L., Rosen, R., 2003. Polybrominated diphenyl ethers (PBDEs) in U.S. mothers’ milk. Environ. Health Perspect. 111, 1723–1729. She, J., Holden, A., Sharp, M., Tanner, M., Williams-Derry, C., Hooper, K., 2007. Polybrominated diphenyl ethers (PBDEs) and polychlorinated biphenyls (PCBs) in breast milk from the Pacific Northwest. Chemosphere 67, S307–S317. Thomsen, C., Stigum, H., Frøshaug, M., Broadwell, S.L., Eggesbø, M., 2010. Determinants of brominated flame retardents in breast milk from a large scale Norwegian study. Environ. Int. 36, 68–74. Vieth, B., Rüdiger, T., Ostermann, B., Mielke, H., 2005. Residues of flame retardants in breast milk from Germany with specific regard to polybrominated diphenyl ethers (PBDEs). Bundesinstitut für Risikobewertung (BfR), Berlin. Available from: http://www.bfr.bund.de/cm/245/residues of flame retardants in breast milk from germany final report.pdf (accessed 06.01.11). WDEH (Washington State Depts. of Ecology and Health), 2006. Washington State Polybrominated Diphenyl Ether (PBDE) Chemical Action Plan: Final Plan. Washington. Available from: http://www.ecy.wa.gov/pubs/0507048.pdf (accessed 06.01.11). Zhang, L., Li, J., Zhao, Y., Li, X., Yang, X., Wen, S., Cai, Z., Wu, Y., 2011. A national survey of polybrominated diphenyl ethers (PBDEs) and indicator polychlorinated biphenyls (PCBs) in Chinese mothers’ milk. Chemosphere 84, 625–633. Zietz, B.P., Hoopmann, M., Funcke, M., Huppmann, R., Suchenwirth, R., Gierden, E., 2008. Long-term biomonitoring of polychlorinated biphenyls and organochlorine pesticides in human milk from mothers living in northern Germany. Int. J. Hyg. Environ. Health 211, 624–638.