Infant exposure to chemicals in breast milk in the United States: what we need to learn from a breast milk monitoring program.

The presence of environmental chemicals in breast milk has gained increased attention from regulatory agencies and groups advocating women's and children's health. As the published literature on chemicals in breast milk has grown, there remains a paucity of data on parameters related to infant exposure via breast-feeding, particularly those with a time-dependent nature. This information is necessary for performing exposure assessments without heavy reliance on default assumptions. Although most experts agree that, except in unusual situations, breast-feeding is the preferred nutrition, a better understanding of an infant's level of exposure to environmental chemicals is essential, particularly in the United States where information is sparse. In this paper, we review extant data on two parameters needed to conduct realistic exposure assessments for breast-fed infants: a) levels of chemicals in human milk in the United States (and trends for dioxins/furans); and b) elimination kinetics (depuration) of chemicals from the mother during breast-feeding. The limitations of the existing data restrict our ability to predict infant body burdens of these chemicals from breast-feeding. Although the data indicate a decrease in breast milk dioxin toxic equivalents over time for several countries, the results for the United States are ambiguous. Whereas available information supports the inclusion of depuration when estimating exposures from breast-feeding, the data do not support selection of a specific rate of depuration. A program of breast milk monitoring would serve to provide the information needed to assess infant exposures during breast-feeding and develop scientifically sound information on benefits and risks of breast-feeding in the United States.

Further, the Endocrine Disruptor Screening and Testing Advisory Committee recommended that the U.S. EPA screen and potentially test "representative mixtures to which large … segments of the population are exposed," including breast milk (5). Groups advocating for women's and children's health have also focused on chemicals in breast milk (6,7).
Although research has provided information on the types of chemicals likely to be found in breast milk and on the toxicologic aspects of many of these chemicals, there are few data on parameters related to infant exposure via breast-feeding, including those with a time-dependent nature. This type of information is necessary for performing exposure assessments without heavy reliance on default assumptions or on the limited databases currently available. In addition, data collected longitudinally provide information on trends in breast milk chemical levels, which indicate whether controls on sources of contaminants are effective. Without this type of information, it will continue to be difficult to provide a scientifically based and consistent message to interested parties (e.g., doctors, nurses, lactation specialists, and new mothers) on the risks and benefits of breast-feeding and to compare these to formula-feeding.
In this paper, we review the extant data on two of the parameters needed to conduct realistic exposure assessments for breast-feeding infants, the first step in risk/benefit analyses and subsequent formulation of risk/benefit messages. In particular, we focus on what is known about the levels of chemicals in human milk in the United States and the elimination kinetics (depuration) of chemicals from the mother during breast-feeding. Information on, and uncertainties associated with, other breast-feeding-related parameters have been discussed elsewhere (8).

Levels and Trends of Environmental Chemicals in Breast Milk in the United States
Chlorinated organic pesticides, polychlorinated biphenyls (PCBs), and polychlorinated dioxins and furans have been the focus of the majority of studies on environmental chemicals in breast milk. We describe the database of published studies of these chemicals in breast milk in the United States and use this database of dioxin and furan concentrations in breast milk to explore whether any trends in concentrations of environmental chemicals (from selected countries and the United States) can be discerned. The discussion on trends is limited to dioxins, which have been examined more fully than most other chemicals [with the possible exception of trichlorobis(p-chlorophenyl)ethane (DDT) and its metabolites, reviewed by Smith (9)]. Figure 1 presents data on concentrations of organochlorine pesticides in breast milk from the United States, plotted by year [pre-1986 data: Jensen and Slorach (10); post-1985 data: Kostyniak et al. (11), Jensen and Slorach (10), Mattison et al. (12)]; data for DDT and metabolites reviewed by Smith (9) have not been included. Figure 2 shows PCB data for breast milk from the United States [pre-1986 data: Jensen and Slorach (10); post-1985 data: Hong et al. (13), Kostyniak et al. (11), Mattison et al. (12)]. Data normalized by lipid level (milligrams per kilogram, lipid

Levels
The presence of environmental chemicals in breast milk has gained increased attention from regulatory agencies and groups advocating women's and children's health. As the published literature on chemicals in breast milk has grown, there remains a paucity of data on parameters related to infant exposure via breast-feeding, particularly those with a time-dependent nature. This information is necessary for performing exposure assessments without heavy reliance on default assumptions. Although most experts agree that, except in unusual situations, breast-feeding is the preferred nutrition, a better understanding of an infant's level of exposure to environmental chemicals is essential, particularly in the United States where information is sparse. In this paper, we review extant data on two parameters needed to conduct realistic exposure assessments for breast-fed infants: a) levels of chemicals in human milk in the United States (and trends for dioxins/furans); and b) elimination kinetics (depuration) of chemicals from the mother during breastfeeding. The limitations of the existing data restrict our ability to predict infant body burdens of these chemicals from breast-feeding. Although the data indicate a decrease in breast milk dioxin toxic equivalents over time for several countries, the results for the United States are ambiguous. Whereas available information supports the inclusion of depuration when estimating exposures from breast-feeding, the data do not support selection of a specific rate of depuration. A program of breast milk monitoring would serve to provide the information needed to assess infant exposures during breast-feeding and develop scientifically sound information on benefits and risks of breast-feeding in the United States.

CHILDREN'S HEALTH
Review basis) were included in Figures 1 and 2. For studies that collected data over more than 1 year, either the midpoint (for a range of more than 2 years) or the first year was plotted. Information on dioxins and furans in breast milk in the United States is shown in Figure 3.
Given the significant restrictions on manufacture, use, or release of the chemicals shown in Figures 1-3, it is unlikely that pre-1985 data are representative of current breast milk levels. Focusing on data from 1985 forward, the largest databases in the United States are for PCBs and dioxins/furans (Figures 2 and 3 Because the data describing levels of environmental chemicals in breast milk from women residing in the United States are geographically limited and from generally small populations, they cannot be considered representative of current breast milk levels of women in the United States in general.

Trends
Because of worldwide attention on dioxins/furans and their reduced release into the environment, it is expected that breast milk levels would be declining in the United States. The analysis presented here suggests that this is the case for many countries. The international databases on dioxins/furans in breast milk were used to explore the extent of breast milk dioxin data and to determine whether any trends in concentrations over time are evident (Table 1). Breast milk data were collected for the years 1970-1998 from published sources. Data were available from the following countries: Austria, Belgium, Canada, Denmark, Finland, Germany, Hungary, Japan, the Netherlands, Norway, Pakistan, Spain, Sweden, United Kingdom, United States, the former Soviet Union, Ukraine, and Yugoslavia. Extremely limited data or data for one year only were available from the following countries: Albania, China, Croatia, Czech Republic, France, India, Italy, Kazakhstan, Lithuania, New Zealand, Poland, Russia, Slovak, South Africa, and Thailand. For breast milk samples collected before 1989, we used a combination of data from a compilation by Jensen and Slorach (10) and primary literature. For breast milk samples collected after 1988, data were all from primary literature. We assembled the following information: date, country, number of donors, dioxin and furan congener concentration, toxic equivalency factors (TEF) value, percent lipid, description of sampling location/population, and reference; not all information was available for each sample. We calculated total toxic equivalents (TEQs) for dioxins and furans combined. For the purposes of this paper, "dioxin TEQs" refer to dioxin and furan TEQs combined. We used international TEF (I-TEF) values and WHO TEF values (Table 2) to calculate the TEQs  of the dioxins and furans in breast milk (our  values are based on I-TEF values).
Because sampling and analysis protocols can substantially impact the results of a breast milk sampling program and because the data assessed in this analysis derive from studies conducted with varied protocols, the comparability of study results is questionable. For example, variation in the time of breast milk sampling (including time postpartum and time of day), the age of the mother, and the number of previously breast-fed children can produce inconsistent interstudy results. In addition, most countries lack adequate numbers of breast milk samples for the data to be considered representative of the entire country. Regardless, the data assembled here represent the preponderance of published data on dioxin in breast milk.
Uncertainties in the breast milk dioxin database, in addition to those mentioned above, impact its usefulness in ascertaining trends in data over time. We describe some of the shortcomings of the reported data below.
Date of sampling. In many cases, the actual year that breast milk sampling was conducted was not provided. In these instances, we used the year of publication for the sake of consistency. However, this clearly biases the time frame of sampling (which was likely to have occurred from 1 to several years before publication) and increases uncertainty in the time-trend analysis. This is particularly important because the preponderance of the data span approximately 15 years; uncertainty regarding the sampling year can clearly impact the results.
In some cases, sampling occurred over a period of more than 1 year. For the purposes of this analysis, if the sampling time frame was 2 years, we used the earlier reported year. If the sampling time frame was > 2 years, we used the midpoint in time.
Congener concentration measurements. In some cases, study authors did not report data for each specific congener, but rather provided summations of certain congeners, particularly for the 2,3,7,8-pentachlorinated dibenzofurans. For these congeners, we used the more conservative TEF of 0.5. Nonreported congeners were considered to have a value of zero. For data reported as "nondetect," and for which detection limits were provided, we used the detection limit as the concentration value.      Concentration (ppt TEQ, lipid) 1973 1975 1977 1987 1989 Year 1991 1985 1983 1981 1979 Measure of central tendency. Because the dioxin breast milk concentrations are not necessarily normally distributed, the median, geometric mean, or other statistic might be the preferred measure of central tendency (9). In addition, it is not clear that central tendency is of primary interest; for example, frequencies of extreme (high or low) concentrations may be more important. However, the arithmetic mean was the most commonly reported measure; thus, for consistency, we used arithmetic means in this analysis.
Sources of variability. Most samples represent a different number of donors (due to pooling). In addition, a certain amount of laboratory variability is associated with the analytical results of each sample. Thus, the samples have different inherent variability; therefore, care is required in interpreting apparent trends.
The results of this analysis indicate an international decline in concentration of dioxin in breast milk over time (Figure 4). High levels of dioxin TEQs in breast milk from the early 1970s are from mothers residing in areas in Vietnam that had been sprayed with Agent Orange, a defoliant contaminated with dioxins, during the Vietnam war. If we focus on the data from the 1980s and 1990s, it is more difficult to discern a trend in breast milk dioxin levels. This is likely due, in part, to the general paucity of data and the uncertainties in the database described above. By examining the data from each country individually, a clearer picture emerges. Breast milk dioxin data from several countries seem to suggest a decline in levels over time (including Austria, Belgium, Denmark, Finland, Germany, Hungary, Japan, the Netherlands, Norway, Pakistan, Sweden (42), the United Kingdom, Ukraine, Vietnam, and Yugoslavia); Figure 5 shows  (42) observed a decline in dioxin and furan levels in breast milk from mothers in the Stockholm region from 1972 to 1985. Between 1985 and 1989, however, they reported that this trend did not continue. The European Union data for Sweden reported for the early 1990s also suggested a plateau in the dioxin levels (see data for Sweden in Table 1).] Germany, with its rich database, seems to provide the most compelling evidence for a decline in breast milk dioxin levels over time (43,44). For example, Fürst and Wilmers (43), in their analysis of approximately 1,000 breast milk samples from North Rhine-Westphalia, reported that dioxin TEQ levels decreased from 34 ppt (lipid based) in 1989 to 14.2 ppt in 1996, about a 60% decline. The data for Japan (30), Norway, and the United Kingdom also provide convincing evidence for a decrease in breast milk dioxin levels. Iida et al. (30) reported a slight decline in PCDD TEQs in breast milk from 1994 to 1996; their data are not aggregated (i.e., dioxin and furan data were not combined) and their assessment did not include dioxin data from the 1980s. The data from Canada, the Czech Republic, France, Spain, and the United States are more ambiguous, whereas those from Italy and Lithuania suggest an increase in dioxin levels in breast milk. Craan and Haines (45) summarized Canadian breast milk data collected over 25 years by Health Canada, including data for dioxin, and reported the following decline in dioxin TEQs (ppt, lipid basis): 24.7 ppt for 1981-1982; 15.6 ppt for 1986-1987; and 14.5 ppt for 1992. For Canada (Figure 3), the decline indicated by the data reported by Craan and Haines (45) is obscured by additional province-specific data reported by Liem et al. (17). A smoothed fit (least squares) through the data for the Western European countries also suggests an overall decline ( Figure 6). In contrast, Figure 3, which includes dioxin/furan data for the United States, reveals the limit of our knowledge of what constitutes a "representative" level of dioxins/furans in U.S. breast milk and whether levels in the United States are declining.
In summary, for many of the countries for which dioxin TEQs in breast milk have been reported, the data indicate a decrease in breast milk dioxin TEQs over time. For countries with ambiguous results, including the United States, it is possible that an improved database (e.g., greater number of samples collected over time from a broader geographic area with appropriate sampling and analysis protocols) might reveal similar future reductions in breast milk dioxin TEQs.
Generally speaking, extremely limited data on organic environmental chemicals in breast milk in the Unites States have been reported in the published literature. In fact, as reported by Hooper (46), "…more is known about the breast milk contamination and body burdens of the mother, infant, and child living in Ukraine or Kazakhstan than, for example, about similar groups living in California." Further, the limited data available in the United States do not provide information on chemicals that are only now beginning to receive attention [e.g., polybrominated biphenyl ethers (47)]. Although we can draw inferences from breast milk data from other countries, the paucity of breast milk data for the United States limits the confidence in our ability to assess infants' exposures, risks, and benefits from breastfeeding, to compare these risks and benefits to formula feeding, and to reach conclusions about the effectiveness of contaminant source controls.

Depuration of Environmental Chemicals from Breast Milk during Lactation
The typical procedure for estimating intakes of environmental chemicals by nursing infants involves selecting a daily volume of breast milk consumed (typically approximately 800 mL/day) and multiplying this value by an empirical or modeled concentration of a chemical in breast milk. The advantage to this approach is the simplicity of the computation. However, the limitations are clear-these estimates do not account for variability in exposure, and more importantly, there is no recognition that a woman's stores of lipophilic chemicals in adipose tissue and breast milk are depleted over the duration of lactation. In fact, the depuration of chemicals due to breast-feeding is a critical, yet poorly characterized, parameter in evaluating infant exposure to chemicals in breast milk (8). It is not clear which factors may influence elimination kinetics; for example, it is likely that some of the following would influence depuration: initial chemical concentration, age of the mother, parity, volume of milk consumed by infant, supplementation with formula or solid foods, and properties of individual chemicals.
A few previous efforts to model concentrations of lipophilic compounds in breast milk have incorporated the depuration process by estimating the decline in chemical concentration over the duration of breastfeeding. For example, LaKind et al. (8) incorporated depuration rates of 30, 50, or 70% over 6 months for 2,3,7,8-TCDD. Patandin et al. (48) used a 20% decrease in dioxin/PCB body burden of the breast-feeding mother to calculate a weekly decrease of 1.7% in dioxin/PCB concentration in breast milk (modeled as o ∫ T e -0.017t dt). Sullivan et al. (49) modeled the decrease in dioxin in breast milk as first order elimination. Kreuzer et al. (50) reported a good correlation between modeled and published values of TCDD in mother's milk by assuming an approximately 70% decline in the levels of TCDD in milk after 6 months of daily breast-feeding. In this review of the published literature on depuration of environmental chemicals in breast milk, we describe the uncertainties associated with the available information. Our focus is on lipophilic environmental chemicals, and each section below describes the database for a particular chemical or group of chemicals (studies are described in chronological order).

Dioxins/Furans
Fürst et al. (51) collected milk samples from one mother for 1 year after the birth of her second child and analyzed the samples for dioxins and furans. The mother provided breast milk samples during week 1, week 5, weeks 10-13, and weeks 52-60 postpartum. The analytical results are presented in Figures 7 (dioxins) and 8 (furans). OctaCDD is reduced by approximately 50% between the first and fifth weeks; the other congeners decline by 15-30% (51). Fürst et al. (51) cautiously conclude that a strong mobilization of … PCDDs and PCDFs takes place within the first few weeks after delivery.
In addition, 168 women provided breast milk samples and information on the period of lactation when the sample was collected. On average, the levels of dioxins and furans in mothers breast-feeding their second child (74 samples) were 20-30% lower than mothers breast-feeding their first child (79 samples). Generally, Fürst et al. (51) found the greatest decline for organochlorines, including PCBs and pesticides, during the transition from colostrum to ripe human milk (the authors did not provide data for these chemicals).
In a study on fecal elimination of dioxins and furans in a 3-month-old breast-fed infant, four samples of breast milk were collected from the mother (details on collection, such as sampling time, were not provided) (52). A general decline in levels of some of the congeners of dioxins and furans over time are observable in Figures 9 and 10.
Hori (53) provided minimal information on levels of dioxins and furans in breast milk lipid from one mother 4-26 weeks after delivery. No information on collection of breast milk samples was provided. PCDDs (TEQs, lipid basis) decreased from 29 ppt at 6 weeks to 21 ppt at 26 weeks, and PCDFs decreased from 18 ppt at 6 weeks to 12 ppt at 26 weeks ( Figure 11).
Schecter et al. (54) describe the results of a study of dioxins and furans in the breast milk of a somewhat less typical mother who breast-fed one child for 16 months and then breast-fed twins for over 2 years. The first breast milk sample was collected in February 1992, after the mother had nursed her first child for about 1 year. The second sample was collected in March 1993 (approximately 3 months after the birth of the twins), and the last in September 1995. [Schecter et al. (54) also provided data from March 1993 to December 1993; however, these are averages of 10 samples over that time period.] From March 1993 to September 1995, the total PCDDs, PCDFs, and PCDDs/PCDFs in milk (TEQs, lipid basis) decreased by 70%, 66%, and 69%, respectively (Table 3). Schecter et al. (55) postulated that the increase in dioxins and furans between December 1994 and September 1995 may have been caused by a decrease in breastfeeding and by a decreased intake of dioxins from food.
Abraham et al. (56,57) studied the intake and fecal elimination of chemicals, including dioxins and furans, in infants. Two samples

Children's Health • Infant exposure to chemicals in breast milk
Environmental Health Perspectives • VOLUME 109 | NUMBER 1 | January 2001 Table 2. I-TEFs and the more recent WHO-TEFs for dioxins and furans (41).

I-TEF
WHO-TEF     of mother's milk (at least 100 mL each) were obtained by pumping empty the whole breast. Reported results (Table 4) indicate that for the dioxins, octaCDD and heptaCDD appear to increase over the course of 5 months, whereas the concentrations of the remaining congeners stay relatively constant (56). In contrast, the level of octaCDF appears to decrease while the other furan congeners remain essentially unchanged. Abraham et al. (57) studied two breastfed infants (as well as one formula-fed infant, which is not discussed here). Mother's milk (two samples during each sampling period) was obtained by emptying the entire milk content in the breast by pump. The levels measured in the diet of two infants are not shown here because, after the first month's measurements, the level of dioxins and furans reported in their diets included those measured in foods other than breast milk, including vegetables and rice pudding prepared with cow's milk.
In reviewing the above dioxin/furan studies, it is clear that information reported to date is not sufficient to confidently derive depuration rates for dioxins and furans or to make generalizations about the factors which might influence elimination kinetics (Table  5). For example, the limited information on breast milk sample collection methodologies does not permit an evaluation as to whether representative samples were obtained. Further, there was only one woman included in each study, with little information on such factors as age and parity, and not all studies examined depuration immediately postpartum [for certain studies (52,56), analysis of elimination kinetics was not the intent of the research].

PCBs and Polybrominated Biphenyls
To assess whether levels of polybrominated biphenyls changed in breast milk over time, Brilliant et al. (58) studied one woman over 3 months, but they provided no information on sampling methodology for this individual. The authors noted day-to-day variations but no trend in concentrations.
A study to examine long-term excretion of PCBs in mother's milk was conducted with a woman who was occupationally exposed to PCBs (Kanechlor 300 and 500) through work in a capacitor factory (59,60). Before giving birth, the subject underwent 2 years of fasting treatment for PCB intoxication. The authors reported an approximately 76% decrease in PCB levels in milk 16 months after delivery and described a half-life of 8 months for PCBs in breast milk (breast milk was used for study purposes only) (Figure 12).
Hofvander et al (61) collected breast milk samples from 18 mothers at 3 months postpartum and from 23 other mothers at 6 months postpartum. The mean levels of PCBs in the 3-and 6-month groups were comparable. It is difficult to interpret the results of this study because breast milk from two separate groups of women was sampled and only mean values were provided.
Mes and Lau (62) examined the change in PCB levels in the milk of one woman during the course of lactation. They reported that despite fluctuations, the PCB congener content remained relatively constant in the milk during lactation, except for those congeners with six and seven chlorine atoms in the molecule. Mes and Lau (62) reported a statistically significant increase in the hexachlorobiphenyl content of the breast milk.
Mes et al. (63) sampled breast milk from 16 women during eight intervals of a 98-day lactation period. The milk samples were collected over a 24-hr period. Data were reported on a whole milk basis as averages of all samples collected at a given time during lactation; we used lipid levels to convert the whole milk values to lipid-corrected values ( Table 6). Even after lipid correction, there is no obvious trend in these data.
Rogan and colleagues (64,65) studied breast milk from the mothers of 856 children and reported a decline in the PCB levels, on average, by about 20% after 6 months ( Table 7). The authors did not describe breast milk sampling procedures. Forty-three percent of the women were primiparous, and the median time for breast-feeding was 29 weeks.
Fooken and Butte (66)      It was difficult to discern any common trend among these women except for an increase in PCBs from the colostrum samples to the first milk sample. However, the authors reported that PCBs showed an increase of 6% per month. Hori (53) provided minimal information on the levels of PCBs in breast milk lipid from one mother 4-26 weeks after delivery. No information on collection of breast milk samples was provided. Coplanar PCBs (TEQs, lipid basis) decreased from 50 ppt at 6 weeks to 32 ppt at 26 weeks (Figure 13).
In a study on the intake and fecal elimination of chemicals in infants, Abraham et al. (56) reported depuration data on three PCBs. Two samples of a mother's milk (at least 100 mL each) were obtained by pumping empty the whole breast. The authors reported increases in concentrations of PCB 138 and PCB 180, but there was no obvious trend in the data for PCB 153 ( Table 8).
The research by Schecter et al. (55) on a mother breast-feeding twins was described in "Dioxins/Furans." Schecter et al. (55) also analyzed breast milk samples for PCB congeners. Concentrations of total PCBs are shown in Table 9 (the concentrations of individual PCB congeners detected in breast milk lipid decreased from 52% to 95% over the study duration).
Kostyniak et al. (11) analyzed breast milk samples from lactating female members and spouses of male members of the New York State Angler Cohort. The samples were analyzed for 77 PCB congeners and several pesticides. Approximately half of the population was primiparous, and the parity of the remaining women was ≥ 2. Breast milk samples were collected after the second morning feeding (hindmilk was collected). The study was not longitudinal-in other words, the authors did not analyze concentrations of PCBs in breast milk over time for individual women. However, they performed Spearman rank correlations for the total months of lactation (over a lifetime) and the PCB concentrations in breast milk fat for all 98 study participants and reported negative correlation. For primiparous women, Kostyniak et al. (11) reported a significant negative correlation for total PCBs and five PCB congeners.
Information reported to date on depuration of PCBs is not sufficient to confidently derive depuration rates for this group of chemicals or to make generalizations about the factors that might influence elimination    kinetics (Table 10). As with the dioxin/furan studies, limited information on breast milk sample collection methodologies does not permit an evaluation as to whether representative samples were obtained. Pooling of samples, small sample sizes, and minimal data on such factors as age and parity further limit our ability to quantify depuration.

Chlorinated Organic Pesticides
Curley and Kimbrough (68) analyzed breast milk samples from five women in one of the first explorations of organochlorine concentrations in breast milk at various times during lactation and provided mean concentrations. The mean total DDT concentrations increased during lactation; this was not considered statistically significant because of a large individual variation (68). Bakken and Seip (69) analyzed colostrum and breast milk from three women for hexachlorobenzene (HCB), benzene hexachloride (BHC), and total DDT for up to 9-16 weeks postpartum. Wide fluctuations were seen; in one woman, BHC increased more than 4 times over the course of 4 days, from 8.6 to 40.8 ppb. The authors generally found the highest concentrations in colostrum, with declining values at later sampling times. Bakken and Seip (69) did not indicate whether breast milk was sampled in a way that would account for diurnal variations or for variability in lipid content (results were on a whole milk basis).
De Bellini et al. (70) analyzed human milk for organochlorine chemicals from 13 women over 30 days. They found increases in p,p´-DDT and p,p´-DDE and decreases in heptachlor epoxide, hexachlorocyclohexane (HCH), and dieldrin (63,70). Yakushiji et al. (59) examined long-term excretion of PCBs in mother's milk. They also examined p,p´-DDE, but provided no data. However, the authors described a halflife of 8 months for p,p´-DDE in breast milk.
Krauthacker et al. (71) determined concentrations of DDT and metabolites from 34 breast milk samples collected 3-5 days postpartum and from 37 samples obtained at later times (up to 55 weeks postpartum). They provided no information on specific sampling methodology, other than that breast milk was manually expressed. Concentrations were given as means on a whole milk basis. According to the authors, the ranges of concentrations were large for samples collected over the 55-week-period and overlapped completely. Krauthacker et al. (71) concluded that the concentration of p,p´-DDE at the beginning of lactation was not significantly different from that from later lactation periods.
Hofvander et al. (61) collected breast milk samples from 18 mothers at 3 months postpartum and from 23 other mothers at 6 months postpartum. The mean levels of organochlorine compounds (DDT/metabolites, HCB, HCH, and dieldrin) in the 3and 6-month groups were comparable. The results of this study cannot be used to draw conclusions about depuration because breast milk from two separate groups of women were sampled, introducing considerable uncertainty.
Andersen and Orbék (72) studied organochlorine levels in human breast milk in Denmark; although data were not provided, the authors noted that the content of HCB in milk fat declined slowly with the time of postpartum sampling but that there was no similar decline in levels of DDE, DDT, dieldrin, or PCBs.
Mes et al. (63) sampled breast milk from 16 women during eight intervals of a 98-day lactation period. The milk samples were collected over a 24-hr period at different times during each feeding, and if possible, from alternating breasts. The authors reported the following conclusions: a) a general downward trend in residue concentrations in breast milk was interrupted by sporadic increases; b) most residues showed a statistically nonsignificant increase in residue levels during the first 30 days; and c) during lactation, a statistically      significant decrease was observed for HCB, oxychlordane, transnonachlor, β-HCH, p,p´-DDE, and p,p´-DDT.
Rogan and and colleagues (64,65) studied breast milk from mothers of 865 children and reported a decline in the levels of DDE, on average, by about 20% after 6 months (Table 11). Breast milk sampling procedures were not described. Klein et al. (73) studied the elimination kinetics of several organochlorine compounds from day 2 to day 10 of breast-feeding (30 volunteers). DDT was below the level of detection in all samples. The authors noted a rapid decrease in the DDE concentration over time; the other chemicals, with the exception of heptachlor, showed a linear decrease over the study duration ( Figure 14).
Fooken and Butte (66) collected breast milk samples from five women and examined variations in organochlorine residue levels (HCH, HCB, p,p´-DDT, and p,p´-DDE) during lactation. Month-mix samples were composed of breast milk samples that were collected weekly. The authors found no changes in residue level over time, and there were no observable trends in the fluctuations.
Galetin-Smith et al. (67) examined the levels of p,p´-DDT, o,p´-DDE, and p,p´-DDE in colostrum and milk samples from seven women. No information was provided on collection methodology. The authors reported a 3%/month decrease in levels of p,p´-DDE during lactation, but noted that this result was only marginally statistically significant because individual variation was pronounced. DDT increased 3%/month. Greater variability would be anticipated in these results because they were reported on a whole milk, rather than a lipid, basis.
Abraham et al. (56), in their study of the intake and fecal elimination of chemicals in infants, reported depuration data on HCB. The authors obtained two samples of mother's milk (at least 100 mL each) at 1 month and 5 months postpartum by pumping empty the whole breast. The HCB concentrations decreased by approximately 8% over 5 months.
Schecter et al. (55) analyzed DDE and HCB in the breast milk from a mother nursing twins. The authors reported a 92% decrease in HCB in breast milk lipid over approximately 30 months of lactation; DDE in breast milk lipid declined by 81% during the same time period.
Information on depuration of organochlorine pesticides is not sufficient to confidently derive depuration rates for this group of chemicals or make generalizations Table 10. Synopsis of study data provided on parameters that potentially influence elimination kinetics of PCBs and/or chlorinated organic pesticides (COPs) (presented in chronological order). about the factors that might influence elimination kinetics (Table 10). An additional complication involves comparing different classes of chemicals. As stated above, limited information on breast milk sample collection methodologies does not permit an evaluation as to whether representative samples were obtained. Pooling of samples, small sample sizes, and minimal data on factors such as age and parity further limit our ability to quantify depuration.

Study
In summary, several factors could potentially influence reported depuration rates. These include the number of previous children nursed, initial body burden of the mother, diet, sampling methodology, amount of lipid in breast milk, and the amount of milk consumed by the infant. There are, at present, insufficient existing data to explore whether these factors play a role in rates of depuration. Without this type of information, the discrepancies in the reported rates of depuration cannot be resolved. Thus, the available information supports the inclusion of depuration when estimating infant exposure to environmental chemicals from breast milk, but the data do not support the selection of a specific rate of depuration.

Conclusions
Environmental chemicals in human milk have been studied since the 1950s, when the pesticide DDT was first detected in breast milk (1). These studies are the main source of information with which to estimate health benefits and risks to an infant who is breastfed rather than formula-fed. Each of these studies has strengths and weaknesses; taken individually, many provide snapshots of concentrations of environmental chemicals in the breast milk of a small population at one time and place. It is difficult to make widely applicable statements about levels of environmental chemicals in breast milk from these studies because of a lack of consistent sampling methodologies and reporting of the results.
Although most experts in the fields of pediatric health and lactation agree that, except in unusual situations, breast-feeding is the preferred nutrition for infants, a better understanding of an infant's level of exposure to environmental chemicals is essential, particularly in the United States where there is relatively little information. Considering both the levels of chemicals in breast milk of women residing in the United States and the kinetics of elimination of those chemicals during lactation, existing data are extremely limited. Shortcomings of published studies include inconsistent sampling and analysis protocols, incomplete reporting of sampling methods, nonrepresentative sampling (geographic, parity, age), duration of sampling, limited number of study participants, and the number and types of chemicals analyzed.
These limitations restrict our ability to predict infant body burdens, particularly during the early days and weeks of lactation. A carefully planned and executed program of breast milk sampling and analysis would serve to provide the information needed to assess infant exposures during breast-feeding and to provide consistent and scientifically sound information on benefits and risks of breast-feeding in the United States.
Increased sampling of breast milk is necessary to provide a better basis for characterizing the levels of chemicals in breast milk; therefore, a program should be initiated in the United States to sample and analyze breast milk. This type of program would provide information on current levels of environmental chemicals in breast milk and enable the development of a scientifically based and consistent message to interested parties (e.g., doctors, nurses, lactation specialists, and new mothers) on the risks and benefits of breast-feeding.
The objectives and goals of a breast milk monitoring program for women in the United States are as follows: • Information should be obtained on women from diverse geographic regions of the United States and from different socioeconomic and demographic backgrounds. For example, the United States could be divided into four compartments: Northeast, Southeast, Northwest, and Southwest. Samples should be collected from both rural and urban locations. • Previous studies should be extended by testing for an increased number of environmental chemicals in breast milk. In addition to the chemicals discussed in this paper, analytes should include certain heavy metals as well as other chemicals with significant lipid solubility and long biological half-life. • Longitudinal information should be obtained during the course of lactation so that the decrease in concentration of the chemical over time can be assessed. Lactating women should be enrolled in the study on a longitudinal basis, donating samples on a monthly basis (or more frequently in the first 2 months) and then every 2-3 months if lactation continues. Recruitment of participants may be aided by lactation consultants. • Harmonization of sampling and analysis protocols should be promoted to improve the comparability of the results. Studies should include harmonized sampling and analysis protocols, such as protocols for collecting breast milk samples, gathering information on study participants relevant to the study (e.g., mother's smoking status, age, parity, dietary information, occupational exposure information, infant dietary supplementation), reporting of breast milk data, and reporting of methodologic information. VOLUME 109 | NUMBER 1 | January 2001 • Environmental Health Perspectives