Introduction

Brominated flame retardants, especially polybrominated diphenyl ethers (PBDEs, Figure 1), have recently been found to be increasing in human tissues in Europe (1-3), Canada (4), and the United States (5-8). The increasing presence of PBDEs in human tissue is of particular concern due to their association with endocrine disruption (9-13), reproductive/developmental toxicity including neurotoxicity (14-17), and cancer (18) in rodent studies. We recently reported elevated levels of PBDEs in human breast milk from the United States (19). Samples collected from 47 women in Dallas and Austin, TX, reported levels that were the highest in the world to date; the median level was 34 ppb, or ng/g, lipid, with some milk samples measuring as high as 419 ppb lipid; these levels were 10-100 times the levels reported in Europe. This study is continuing, and data from an additional five women (data not shown) are consistent with findings from the first 47 women's analysis. Another U.S. study, conducted by the Environmental Working Group, found PBDEs in 20 U.S. human breast milks up to 1078 ppb lipid with a median of 57.5 ppb (20). Levels in Swedish (21, 22) and German (23) breast milk from the late 1990s are much lower. While it is encouraging that Swedish human milk levels have already peaked and are declining (22), levels in most other parts of the world including Norway (2) and the United States (7) are still on the rise.

PBDEs have been used commercially as flame retardants in consumer goods, such as plastics, electronics, textiles, and foam, for several decades, but their production and use have undergone a dramatic increase starting in the 1980s (24, 25). Studies have shown that PBDEs are not only present in humans (19, 26, 27), but that body burdens are increasing (7, 28); however, the major route of exposure resulting in these body burdens is unclear. Because these lipid soluble synthetic halogenated compounds chemically and toxicologically resemble PCBs, and PCB intake into humans is almost exclusively through food of animal origin, we conducted a market basket survey of food PBDE levels. To the best of our knowledge, this is the first market basket survey of U.S. food PBDE levels. Considerable fish PBDE data exists worldwide, but there are only two previous market basket surveys of PBDEs, one in Spain (29) and one in Japan (30).

Although food from the United States is generally shipped long distances from the site of origin, we selected these three major supermarket chains in one city and sampled well-known brands of food, assuming there was a reasonable probability of these representing the foods eaten by the people in our earlier human milk study (19). We elected to measure the same 13 PBDE congeners in food as in the breast milk study (Figure 1) (BDEs 17-2,2',4; 28-2,4,4'; 47-2,2',4,4'; 66-2,3',4,4'; 77, 85-2,2',3,4,4'; 99-2,2'4,4',5; 100-2,2'4,4',6; 138-2,2'3,4,4',5'; 153-2,2',4,4',5,5'; 154-2,2'4,4',5,6'; 183-2,2',3,4,4',5',6; and 209-2,2',3,3',4,4',5,5',6,6'). The objective of this market based survey is to report food levels of PBDEs in the United States for the first time; this information will be used in a larger study to collect representative data on PBDE levels in a larger sampling of food from the United States and to calculate human PBDE exposure via dietary intake.

Figure 1 Selected PBDE congener structures and bromination patterns.

Materials and Methods

Part A: Sample Collection. Thirty food types were purchased from three major supermarket chains in Dallas, TX. One sample per food type was analyzed with the exception of two salmon fillets, one salmon steak, two catfish, and two evaporated milk samples. Food of animal origin only was purchased, with one exception, a soy infant formula. The samples were frozen at -80 C after collection and kept frozen until analyzed. The results obtained are based on sample sizes of one, with a total of 32 individual items.

Part B: PBDE Analysis. All analyses were performed following the isotope dilution method previously developed (31, 32). Twelve native standards, BDE Nos. 17, 28, 47, 66, 77, 85, 99, 100, 138, 153 154, 183, were obtained from Cambridge Isotope Laboratories (Andover, U.S.A.). One native standard, BDE 209, was from Wellington Laboratories (Guelph Canada). Six internal C¹³ labeled standards, BDE Nos. 28, 47, 99, 153, 154, and 183, were from Wellington; BDE 209 was from Cambridge. Solvents obtained from Merck (n-pentane), Promochem (cyclohexane, hexane, dichloromethane), Baker (diethyl ether), and Mallinckrodt (ethanol, toluene) were of the highest grade commercially available. Silica gel, alumina oxide, sodium sulfate, and potassium oxalate were obtained from Merck. For quality control reasons, for each block of samples (6-10 samples) a QC pool and a laboratory blank was analyzed in parallel.

Canned Milk/Liquid Baby Food/Milk Powder Samples. Before extraction the mixture of 7 internal BDE standards was added to the sample (100 pg/sample for each congener except for BDE 209, where 10 000 pg/sample was added). 15-25 mL of liquid samples was extracted three times with 15 mL of pentane, after adding 2 mL of saturated potassium oxalate solution, 20 mL of ethanol, and 10 mL of ether. The extract was washed with water and dried over sodium sulfate. Gravimetric lipid determination was performed after solvent evaporation.

Fish/Meat/Cooked Egg/Cheese/Ice Cream/Sausage Samples. A total of 5-200 g of tissue was homogenized and mixed with sodium sulfate. Before column extraction a mixture of 7 internal BDE standards was added to the sample (100 pg/sample for each congener except for BDE 209, where 10 000 pg/sample was added). For column extraction a mixture of cyclohexane and dichloromethane was applied. The extract was washed with water and dried over sodium sulfate. Gravimetric lipid determination was performed after solvent evaporation.

Butter/Margarine Samples. A 2 g sample was liquefied and dissolved in pentane. Before further treatment a mixture of 7 internal BDE standards was added to the sample (100 pg/sample for each congener except to BDE 209, where 10 000 pg/sample was added). The dissolved sample was dried over a sodium sulfate column. After solvent evaporation gravimetric lipid determination was performed.

Clean Up. Clean up of all lipid extracts was performed by acid treated and activated silica gel and alumina oxide column. The final extract was reduced in volume by a stream of nitrogen. The final volume was 50 L containing C¹³ labeled BDE 139 for recovery standard.

The measurements were performed using high-resolution gas chromatography/high-resolution mass spectrometry (HRGC/HRMS, HP 5890 coupled with VG Autospec) at RP = 10 000 using a DB 5 (30 m, 0.25 mm i.d., 0.1 m film) column for gas chromatographic separation. The two most abundant masses were used for measurement (M⁺ for Tri- and Tetra-BDE and M-2BR⁺ for Penta- to Deca-BDE). The identification of BDEs was based on retention time and isotope ratio. The quantification was performed by using a five point calibration curve.

Reduction of solvents and control of blank data is an important step in quality control when analyzing PBDEs at ultratrace levels. Solvents and reagents were tested before the laboratory procedures. All glassware was rinsed with solvents prior to use. Silica gel and sodium sulfate were prewashed. Rotary evaporators were not used to reduce the risk of contamination. No plastic equipment was used. Quantification was only done if sample data was at least twice the blank value. In the case when the level was below the detection limit, limit of detection was noted.

Results and Discussion

Findings. This study is the first to report the levels of PBDEs in U.S. foods in a market basket survey. Samples were collected from three separately owned Dallas, TX, large-chain supermarkets in 2003 and were analyzed for the presence and concentration of 13 PBDEs. We found a wide variation in PBDE congener profiles and sum of the congener concentrations across the food groups sampled. The results are presented on a wet weight basis (fat content provided) and grouped into these categories: fish (Figure 2), meat (Figure 3), dairy products (Figure 4), and miscellaneous (Figure 5) that includes soy infant formula, nonfat milk, livers, and chicken eggs. Thirty-two samples were individually analyzed.

	Figure 2 PBDE congener levels and profiles of individual samples of fish products purchased in the United States (pg/g or ppt, ww).
	Figure 3 PBDE congener levels and profiles of individual samples of meat products purchased in the United States (pg/g or ppt, ww).
	Figure 4 PBDE congener levels and profiles of individual samples of dairy products purchased in the United States (pg/g or ppt, ww).
	Figure 5 PBDE congener levels and profiles of individual miscellaneous products purchased in the United States (pg/g or ppt, ww).

The results of this study found that fish have the highest overall PBDE levels (median of 1725 ppt, range 8.5-3078 ppt, wet wt), followed by meat products (median of 283 ppt, range ND (0.2)-1373 ppt), and dairy products (median of 31.5 ppt, range 0.9-679 ppt). Two liver samples were analyzed individually; the chicken liver (data not shown) had the second highest PBDE level in this study (2835 ppt), whereas the level in calf liver was 115 ppt. In chicken eggs (n = 6, one analysis) the total PBDE level was 73.7 ppt.

One salmon fillet had the highest total concentration of PBDEs (3078 ppt) of all food specimens, whereas tilapia fillets had the lowest level of total PBDEs (8.5 ppt) in fish. The range of PBDE concentrations in fish purchased in U.S. supermarkets was significantly higher than meat samples (beef, pork, turkey, and duck): Concentrations in fish range from 8.5 to 3078 ppt, whereas PBDEs in meat products range from 40 to 1378 ppt. Dairy products appear to have even lower levels of PBDEs, ranging from 0.9 ppt in evaporated milk to 679 ppt in cheese. Nonfat milk was excluded in the calculation of dairy products because all fat had been removed, likely resulting in the removal of the fat soluble PBDEs.

Comparison to Others. Although the majority of studies consistently report BDE congeners with relatively high levels of 47 and 99, which can provide a basis for comparison, the inconsistency in congener measurements makes comparisons with other studies more difficult with regard to total PBDE concentration and PBDE congener profiles. We elected to analyze 13 PBDE congeners in this study: BDEs 17, 28, 47, 66, 77, 85, 99, 100, 138, 153, 154, 183, and 209, which are also the congeners previously found in human breast milk samples from women residing in the same area in which the food samples were collected.

In market basket fish, when compared to U.S. wildlife fish studies, our median total PBDE concentration of 1731 ppt (wet wt) is lower than those reported in some wildlife fish studies to date. Concentrations as high as 1250 ppb (wet weight) were reported in mountain whitefish from the Spokane River (33), 80 ppb (wet wt) in salmon from Lake Michigan (34), 65 ppb (wet wt) in Hadley Lake, Indiana (35), up to 1140 ppb (wet wt) in Virginia (36), and some European wildlife fish (37-39), while having a slightly wider range when compared to the market based fish in Spain (29) and Japan (30), and wild salmon study of Scotland and Belgium (40). Fish in the Spanish and Japanese studies may not be so common in the American diet; however, comparisons between market-based samples and wildlife are of some interest.

Compared to total PBDE concentrations in fish from market basket surveys in other countries, U.S. levels appear to be much higher. Bocio et al. (29) reported an upper concentration value of 340 ppt; which is much lower than our high end sample (3078 ppt). A similar situation is also seen when meat and dairy product concentrations are compared. In our study, the median concentration in meat was 283 ppt (maximum of 1373 ppt), versus an upper concentration of 109 ppt in the Spanish study (29) and 63 ppt in the Japanese study (30). PBDE levels in U.S. dairy products (high value of 679 ppt with a median of 99 ppt) were also much higher than in Spain (high value of 47.9 ppt).

Congener Profiles. In fish, the majority of PBDE congener profiles from our study were dominated by BDE 47, which accounted for 40 to 70% of total measured PBDEs. This pattern is similar to previous findings (33-36). The two catfish were exceptions to this norm; BDE 99 was the predominating congener found in one catfish fillet, while BDE 209 (1269 ppt) accounted for slightly over 50% of all PBDEs in the freshwater catfish sample. Congener profiles appear different from those found in some commercial mixtures (25). In carp collected by Rice et al. (41) from the Des Plaines River in Illinois (PBDE = 12.48 ppb wet weight), BDEs 181 and 183 dominated the profile, followed by BDEs 190, 154, and 153. They also reported that carp taken the Detroit River in Michigan had PBDE congener profiles dominated by BDE 47. These differing congener profiles may be explained by metabolism as carp have shown to selectively metabolize certain PBDE congeners as well as break down BDE 209 to lower congeners (42, 43). The difference in congener profiles could also be due to differential exposure; either scenario could be occurring in the catfish samples from our study.

In meat samples, BDE 99 was the dominant congener, followed by BDE 47 and other tetra- and penta-BDEs. It is interesting to note that in calf and also in chicken liver, BDE 209 was the most prevalent and third most prevalent congener detected, respectively. Furthermore, BDE 209 was the dominant congener found in calf liver, soy instant formula, cheese, and margarine. Because BDE 209 is not commonly analyzed, we are unable to compare our findings to other studies. However, the results of this study suggest that BDE 209 may be a major contributor in some food supplies. Congeners 153 and 154 are present in several food samples in ratios that differ somewhat from ratios found in the commercial mixtures (25).

In dairy samples, we noted more variation in congener profiles than either fish or meat. Cheese has the highest total PBDE level at 679 ppt, with BDE 209 predominating. In butter, goat milk, and ice cream samples, it appears both BDE 47 and 99 have the highest, although similar levels. The lipid percent for each item analyzed was as follows: Figure 2: tilapia, 1%; shrimp, 0.6%; rainbow trout, 4.2%; catfish fillet, 5.3%; salmon, 10.3%; shark, 0.4%; salmon fillet 1, 8%; catfish, 11.1%; salmon fillet 2, 13.9%. Figure 3: bacon 1, 43.4%; pork, 8.9%; ground beef, 13.6%; bacon 2, 35.3%; chicken breast, 4.9%; ground turkey, 11.1%; duck, 75.1%; wieners, 32.9%; pork sausage, 23.7%. Figure 4: margarine, 83.3%; evaporated milk 1, 6.6%; evaporated milk 2, 6.3%; milk formula, 3.4%; lowfat yogurt, 1.3%; ice cream, 19.9%; goat milk, 6.7%; butter, 78.3%; cheese, 39.2%. Figure 5: nonfat milk, 0%; soy formula, 3.2%; chicken eggs, 11.5%; calf liver, 6.4%; chicken liver, 13.1%.

Conclusion

These findings are preliminary and will be updated in the future with analyses of new samples. The results of this study show that food of animal origin in the U.S. is contaminated with PBDEs. The levels in U.S. food are higher than levels found in the Spanish (29) and Japanese (30) studies, the only two PBDE market basket studies published to date. Although the relationship between human levels and dietary intake is not well characterized, the results of these findings suggest that dietary intake can be a significant contributor to the high body burden reflected in tissue levels found in U.S. residents, although other routes of exposure may also contribute. BDE 47 predominates in human milk and blood, similar to most of the fish based on this and other previous findings (33-40). In addition, this study reports BDE 209, which we have shown is a major contributor in several food samples. Our study also included 13 congeners, more than previous studies. Because of a lack of consensus regarding which PBDE congener to analyze, total PBDE levels cited here and some comparisons to our study results do not include the same congeners.

Fish appear to be the most studied food or wildlife thus far for PBDEs. On the basis of congener profiles, all studies to date reported that one tetraBDE (congener 47) and two pentaBDEs (congeners 99, 100) are the dominant congeners in fish. In both Europe and the United States, based on some published results to date, sometimes wild fish analyzed can have significantly higher PBDE levels than levels typically observed in some farm raised fish (35-42). In meat and chicken eggs, congener 99 is the most dominant PBDE. In most of our samples, total PBDE levels in U.S. food are high compared to the two other market basket studies (29, 30). Our fish PBDE levels are lower than that found in some previous U.S. wildlife fish studies (33-36) but are higher compared to European and Japanese market basket studies (29, 30).

Levels of PBDEs in our pork and beef samples are also higher than reported in the other market basket studies (29, 30) and tend to be dominated by pentaBDEs. In our meat products, chicken has overall PBDE level at 282 ppt wet wt or 5800 ppt lipid based, which is within the range of a previous study (44). More samples and studies are needed to assess representative levels and patterns of PBDE contaminations in meat samples. Since U.S. mother's milk has the highest PBDE levels compared to the human milk of other countries (19), and most food items in this market basket survey were found to contain higher PBDE levels than the other countries (30, 31), the results here demonstrate a need for understanding the exposure of Americans to PBDEs via food. If, as is anticipated, Deca-BDE becomes the only commercial PBDE manufactured in the United States beginning in 2005, we would expect to see a change in future PBDE patterns.

Acknowledgment

Funding for this study was generously provided by the CS Fund, Warsh-Mott Legacy, the Albert Kundstadter Family Foundation, the Samuel Rubin Foundation, and the EPA/UNC Toxicology Research Program Training Agreement; NHEERL-DESE Cooperative Training in Environmental Sci ences Research, EPA CT826513, with the Curriculum in Toxicology, University of North Carolina at Chapel Hill, which is gratefully acknowledged. We thank Miriam Jacobs for her thoughtful suggestions. The information in this document has been subjected to review by the National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, and approved for publication. Approval does not signify that the contents reflect the views of the Agency, nor does mention of trade names or com mercial products constitute endorsement or recommenda tion for use. The research presented in this document was funded in part by the U.S. Environmental Protection Agency.