There were no significant differences in maternal age, reported smoking, and parity between cryptorchid and control boys (Tables 1 and 2). Gestational age and birth weight were slightly lower in cryptorchid Danish (but not Finnish) boys than in controls. Diabetes was more prevalent in Finnish (but not Danish) mothers of cryptorchid boys than in their controls, because we could not always find controls matched by this criterion. The date of childbirth within the study period did not differ significantly between cryptorchid boys and controls (Denmark, p = 0.327; Finland, p = 0.949).

Danish mothers were slightly older than Finnish [30.6 years; 95% confidence interval (CI), 23.6–38.8 vs. 28.7 years; 95% CI, 21.4–39.7; p < 0.011], had a lower parity (p < 0.033) and smoked more frequently (p < 0.04). The prevalence of diabetes mellitus was higher among Finnish women (p < 0.004), but BMI before pregnancy (p = 0.678) did not differ significantly between the countries (p = 0.225).

Breast milk. Seven PBDEs were measurable in all breast milk samples (Table 3). Median concentrations were higher in Danish but not Finnish cryptorchid boys than in controls, reaching statistical significance for BDEs 47, 100, 28, 66, and 154. The sum of all seven congeners was significantly higher in cryptorchid boys than in controls if both countries were analyzed together (p < 0.007) (Figure 1), also if prematurity and SGA were included in the model (p < 0.035). Similar results were obtained for PBDE expressed as nanograms per liter (data shown only for the sum of all 14 congeners; Table 3). Table 4 shows the remaining seven congeners, which were below the detection limit in a substantial number of samples, and the sum of all 14 congeners.

The concentrations of PBDE congeners in milk samples showed large variations between congeners and individuals (Figure 2). Individual congener concentrations were positively correlated with one another (r = 0.178–0.955, p < 0.0001).

There were no significant country differences in the sum of all congeners [Denmark, 3.52 ng/g (95% CI, 1.26–14.2) vs. Finland, 3.44 ng/g (95% CI, 1.25–13.94), p = 0.754] or the sum of the most prevalent seven congeners, BDEs 47, 153, 99, 100, 28, 66, and 154 [Denmark, 3.24 ng/g (95% CI, 1.26–51.1) vs. Finland, 3.23 ng/g (95% CI, 1.26–51.0), p = 0.629]. Similar results were obtained if we analyzed only milk samples from mothers of healthy boys (data not shown). The estimated daily intake of PBDEs (sum of all) for an infant at 3 months of age (median infant body weight 6.58 kg, consuming 120 mL breast milk/kg) was a median of 16 ng/kg/day (range, 6–121 ng/kg/day). Breast milk lipid (percent wet weight) differed significantly between the countries (mean ± SD = 2.99 ± 1.38 in Denmark, 4.52 ± 1.56 in Finland; p < 0.0001). The lipid content was not significantly correlated with the sum of PBDE congeners (data not shown).

The pattern of congener distribution differed between the countries. BDE-28 was significantly higher in Finland than in Denmark (p < 0.021), with a similar tendency for BDE-47 (p < 0.077), which was the most prevalent congener. BDE-153 (p < 0.0001), BDE-66 (p < 0.026), and BDE-183 (p < 0.022) were significantly higher in Denmark than in Finland, with a similar tendency for BDE-99 (p < 0.073). For congeners at very low concentrations, the percentage of measurable samples was higher in Denmark than in Finland (Table 4). There was no significant effect of breast pump use during collection of samples on the concentration of any BDE congener in the Danish breast milk samples.

In the Danish but not the Finnish samples, the date of childbirth within the cohort was significantly correlated with the concentration of BDE-154 (r = –0.346, p < 0.005), BDE-85 (r = –0.434, p < 0.0001) and BDE-75 (r = 0.376, p < 0.002) with similar tendencies for BDE-66 (r = –0.211, p = 0.092) and BDE-77 (r = –0.214, p = 0.087).

Serum LH levels correlated positively with the sum of seven PBDEs in breast milk (r = 0.218, p < 0.033) as well as with individual congeners BDE-47 (r = 0.227, p < 0.027), BDE-100 (r = 0.293, p < 0.004), and BDE-154 (r = 0.203, p < 0.048). Country-specific analyses showed significant associations between serum LH and the above listed PBDEs and their sum for Finnish milk samples, but not Danish (data not shown). No other reproductive hormones or their ratios were significantly correlated with the concentration of PBDEs.

Placentas. The average levels of 14 PBDE congeners per gram fat in placentas were lower than in breast milk (Table 5), and more samples were nondetectable. Therefore, the sum of five congeners was used instead of the seven congeners used for milk. This had a minor influence on the total, because the sums of the five and the 14 congeners were very close.

The distribution of congeners resembled the distribution in breast milk, with BDE-47 and BDE-153 constituting the main fraction of PBDEs. There was no significant country difference for the sum of all 14 congeners (p = 0.198) or the sum of five (BDE-47, BDE-153, BDE-99, BDE-100, and BDE-28; p = 0.192). This was also true when only placentas from healthy boys were analyzed (data not shown). The concentrations of the five most prevalent congeners were positively correlated with each other (r = 0.171–0.827, p < 0.0001). Placenta lipid content (percent wet weight) differed significantly between the countries (mean ± SD = 1.09 ± 0.17 in Denmark, 1.21 ± 0.13 in Finland; p < 0.0001). Placenta lipid content was not significantly correlated with the sum of PBDE congeners (data not shown).

The date of childbirth within the cohort was significantly correlated in the Danish placentas with the concentration of BDE-66 (r = –0.246, p < 0.001), and in the Finnish placentas with BDE-85 (r = 0.239, p < 0.01) and BDE-153 (r = 0.275, p < 0.003).

There was no significant difference in the placenta concentration of the five most prevalent PBDEs between cryptorchid boys and healthy boys in Denmark (p = 0.10–0.976) or Finland (p = 0.09–0.835) or for their sum (p = 0.312 and p = 0.128, respectively). The results remained nonsignificant if prematurity and SGA were included in the model. There were no correlations between placental PBDEs and serum reproductive hormones in 3-month-old infants.

Paired samples. The median concentrations (nanograms per gram fat) of the 14 congeners in placenta were lower than the concentration found in the paired breast milk samples (n = 86) (Table 6), but there were significant correlations between the measurements, except for BDE-85 and BDE-138.

The sum of all congeners in milk was 3.39 (95% CI, 1.43–48.2) for boys with cryptorchidism and 3.15 (95% CI, 1.07–24.9) for controls (p = 0.228), and in placenta 1.22 (95% CI, 0.64–9.32) for boys with cryptorchidism and 1.17 (95% CI, 0.49–5.46) for controls (p = 0.871). Infant reproductive hormones at 3 months of age (27 Danish, 35 Finnish boys) were not correlated to PBDE concentrations in placenta or milk in this data subset.

Discussion

To our knowledge, this is the first study describing an association between congenital cryptorchidism in humans and exposure to PBDEs. An association was found for the sum of seven PBDEs (BDEs 47, 153, 99, 100, 28, 66, 154) in breast milk as well as for the individual congeners BDEs 47, 100, 28, 66, and 154. Concentrations of BDEs 47, 100, and 154 were also positively correlated with increasing serum LH values. This suggested that a higher gonadotropin drive was necessary to ensure normal testosterone production by the Leydig cells (Suomi et al. 2006), and thus a subtle primary testicular disfunction.

In this study we assessed infant exposure by measuring the concentration of PBDEs in breast milk, which reflects the accumulated body burden of the mother (Inoue et al. 2006; Jensen and Slorach 1991; Waliszewski et al. 2001). It also is a proxy for prenatal fetal exposure because PBDEs, especially the lower brominated compounds, can cross the placenta (Bi et al. 2006; Mazdai et al. 2003) and are transferred to breast milk during lactation (Darnerud and Risberg 2006). However, when we assesssed exposure by measuring PBDE in placenta, the results did not support the above findings, despite the fact that 86 mother–child pairs were represented with both milk and placenta samples. At present, it is not clear why this is the case.

The concentration of all PBDE congeners per gram fat in placenta was considerably lower than in milk, and more congeners were nondetectable. These differences did not have any remarkable influence on analytical errors, because the amount of fat used for analysis was similar for placentas and milk samples. In theory, placenta should be a better proxy for fetal exposure than breast milk, because it represents the direct route of chemical transfer during pregnancy. The number of placentas analyzed in this study was larger than the number of milk samples, and should thus better represent the pool of cases and controls. We could not establish any obvious selection bias of milk donors. On the other hand, due to higher concentrations, milk analyses were somewhat more reliable toward the lower end of concentration. This should not be important, because the main group differences were seen at the higher end of PBDE concentrations. It is currently unknown whether PBDEs accumulate in placenta, but our paired samples indicated that this may not be the case. We found positive correlations between measurements in placenta and breast milk, but three to four times lower absolute concentrations in placenta. Placenta concentrations may resemble measurements in single blood samples, and thus reflect the situation at delivery but not the long-term exposure.

There is some controversy as to whether the placental transport of PBDEs may differ between lower and higher brominated PBDEs. An American study reported a strong correlation of lipid-adjusted BDE concentrations in maternal serum at term and cord blood (Mazdai et al. 2003), whereas studies from countries with a generally lower exposure level found weaker correlations and higher values in maternal than in fetal samples (Bi et al. 2006). Segregation from serum into breast milk samples did not always follow a 1:1 pattern. The congener distribution appeared to be similar for lower brominated BDEs (Bi et al. 2006; Mazdai et al. 2003), whereas higher brominated BDEs such as BDE-209 were 10 times higher in serum than in breast milk (Inoue et al. 2006). Thus, the relative distribution of congeners between placenta and breast milk may also depend on the fat composition of these two matrices causing different solubility. We found significantly higher fat concentrations in the Finnish than in the Danish samples, which may reflect dietary habits in the two countries, because both long-term and short-term diet as well as the nutritional status may influence content and composition of breast milk lipids (Ortiz-Olaya et al.1996).

Current knowledge about human reproductive health consequences after exposure to PBDEs is very limited (Agency for Toxic Substances and Disease Registry 2004). A study from Taiwan showed an association between the sum of 12 PBDE congeners in breast milk, 10 of which were the same as in our study, and lower birth weight and length (Chao et al. 2007). Recently, a Swedish case–control study found higher values of PBDEs (sum of BDEs 47, 153, and 99) in blood samples from mothers of young men with testicular cancer than in age-matched controls (Hardell et al. 2006). However, maternal PBDE exposure was assessed 15–25 years after the critical time period—that is, at the time of the cancer diagnosis—which considerably weakens the possibility to establish a causal link between exposure and outcome. Also in this study (Hardell et al. 2006), exposures to other persistent chemicals occurred simultaneously, and it cannot be determined how these substances interact in their effect on reproductive development. Testicular cancer is the most severe clinical symptom of the testicular dysgenesis syndrome (TDS), which also encompasses impaired semen quality, congenital cryptorchidism, and hypospadias (Skakkebæk et al. 2001). TDS may be caused by genetic, hormonal, or environmental factors (Sharpe 2006; Skakkebæk et al. 2001). Prenatal exposure to PBDEs may have an adverse effect on testicular growth and differentiation in utero.

In animal studies, penta-brominated PBDEs showed antiandrogenic activity (Stoker et al. 2005). A peripubertal single-dose exposure to a commercial mixture of penta-BDEs delayed the onset of puberty in male and female rats (Stoker et al. 2004) and suppressed the growth of the seminal vesicles and ventral prostate. In adult rats, a single-dose exposure significantly increased LH concentration in serum (Stoker et al. 2005). Our observation of an association between PBDE levels in breast milk and serum LH in infants at 3 months of age is in line with these animal data. Gestational exposure of rats to BDE-99 caused a shortening of the anogenital distance in male and female rats, reduction in primordial and secondary ovarian follicles, a lower sperm count, and lower estradiol and testosterone levels in adulthood (Kuriyama et al. 2005; Lilienthal et al. 2006). In vitro assays showed a competitive androgen receptor binding for BDEs 47, 99, and 100. BDEs 47, 71, and 100 inhibited dihydrotestosterone-induced transcriptional activity (Stoker et al. 2005). Because testicular descent is highly androgen-dependent (Toppari 2003), the adverse effect of PBDE on testicular descent could be caused by their antiandrogenic properties.

In addition, BDEs 47, 100, 75, and 51, particularly their hydroxylated metabolites, are weakly estrogenic, and BDEs 153, 166, and 190 are antiestrogenic in vitro (Legler and Brouwer 2003; Meerts et al. 2001). Exposure of female rats to BDE-99 led to the formation of abundant vesicles and vacuolization of the ovary (Talsness et al. 2005), which showed signs of compromised steroidogenesis. Prenatal female exposure to BDE-99 in rats affected estrogen target genes in the uterus (Ceccatelli et al. 2006). Thus, the delicate balance between androgens and estrogens in the fetus may become altered by PBDE exposure. In mice, the metabolism of BDE-47 was highly dependent on the developmental stage of the animal, being slowest in pups (Staskal et al. 2006). Whether this also plays a role for human fetal development is currently unknown. Most exposure doses used in animal studies are several orders of magnitude higher than the levels of PBDEs found in breast milk in our study. However, there is emerging evidence that also low-dose exposure to, for example, BDE-99 close to levels found in human adipose tissue may have an adverse effect on the reproductive health of the offspring (Anonymous 2005; Kuriyama et al. 2005).

The distribution pattern of BDE congeners in breast milk corresponded to the distribution of BDE congeners in commercially available mixtures of PBDEs (Alaee et al. 2006; Darnerud et al. 2001; Law et al. 2006; Lilienthal et al. 2006). The absolute concentrations found in our study are within the same order of magnitude as reported from other Nordic and European countries (Jaraczewska et al. 2006; Kalantzi et al. 2004; Lind et al. 2003; Strandman et al. 2000) as well as China (Bi et al. 2006) and Japan (Eslami et al. 2006; Inoue et al. 2006). There are, however, significant geographic differences between and within these countries, which point toward differences in general contamination levels. In our study, the total amount of PBDEs did not differ between Finnish and Danish milk or placenta samples, but the pattern of congener distribution varied, which indicated different sources and timing of exposure. PBDE levels reported from American studies of breast milk appear to be considerably higher (Betts 2002). In contrast, the previous exponential increase in penta-PBDEs in Swedish breast milk samples since the 1970s has reversed since the late 1990s (Law et al. 2006; Meironyte et al. 1999; Sjödin et al. 2003), when penta-BDEs were gradually phased out. The collection of breast milk samples in our study covered a 5-year period, and we found a negative correlation between the level of PBDEs in breast milk and the infant date of birth. This is in line with the expected decline in the use of penta-PBDEs in the two regions. The high variability in PBDE concentrations among individual mothers has been described also for other human matrices (McDonald 2002), and may reflect variability in both exposure and metabolism. Deca-BDE can be converted through sunlight exposure into lower brominated BDEs, which are more readily absorbed from the intestine and bioaccumulate due to their longer half-life than deca-BDE (Watanabe and Tatsukawa 1987). However, it is as yet unknown how much this process contributes to human environmental exposure to lower brominated PBDEs.

Exposure to PBDEs cannot explain the observed geographic difference in the prevalence of cryptorchidism between Denmark and Finland. Breast milk contains significant levels of other persistent and nonpersistent chemical compounds (Damgaard et al. 2006; Main et al. 2006a), which can affect perinatal testicular development. Mothers with high levels of PBDE exposure also may be exposed to high levels of other persistent chemicals. Thus, the combined exposure to multiple environmental factors may cause the association between congenital cryptorchidism and PBDE concentration in breast milk (Koppe et al. 2006). In addition, other lifestyle factors and genetic susceptibility may play a role (Damgaard et al. 2007; Sharpe 2006; Virtanen et al. 2006). In our total study population, the geographic difference in the prevalence of cryptorchidism was observed mainly for mild forms of undescended testes, which had a high degree of spontaneous postnatal descent (Boisen et al. 2004). This pattern was also seen in this subpopulation, in which PBDEs was analyzed. However, also mild and transient forms of cryptorchidism are associated with a subtle impairment of primary testicular function (Suomi et al. 2006).

In conclusion, we used two different biological matrices in this study to assess infant perinatal exposure. Breast milk, but not placenta, showed an association with congenital cryptorchidism. There are valid arguments for either matrix, and risk assessment will require more scrutiny. The association between PBDE contamination levels in breast milk and congenital cryptorchidism is still of concern because exposure to PBDEs is considerable in some areas.