SAFETY EVALUATION OF CERTAIN
FOOD ADDITIVES AND CONTAMINANTS

Coplanar PCBs may induce their own metabolism. In rodents exposed to relatively high doses of 2,3,4,7,8-TCDF and TCDD, a twofold induction of their metabolism was observed (Brewster & Birnbaum, 1987; Leung et al., 1990b; McKinley et al., 1993). Similarly, clear biphasic elimination of 2,3,4,7,8-PeCDF and 1,2,3,4,7,8-HxCDF was observed in Yu-cheng and Yusho patients, indicating that they had been exposed to concentrations well above background for induction of metabolism (Ryan et al., 1993a). The metabolism of TCDD was found to be substantially induced in two patients with TCDD poisoning, the half-lives being 200 and 230 days (Geusau et al., 1999).

TCDD readily crossed the placenta of pregnant Long-Evans rats given a single oral dose of TCDD at 1.2 µg/kg bw in corn oil on day 8 of gestation. The concentrations of TCDD found in the fetal compartment (fetuses plus their placentae) were 39 pg/g (0.01% of the administered dose) on day 9 of gestation, when the maternal blood concentration was 15 pg/g; 26 pg/g (0.11% of the administered dose) on day 16, with a maternal blood concentration of 18 pg/g; and 21 pg/g (0.7% of the administered dose on day 21, with a maternal blood concentration of 8 pg/g. In individual embryos, the TCDD concentrations were 40, 18 and 22 pg/g on days 9, 16 and 21. The embryo/fetal compatment may therefore be considered a nonsequestering maternal compartment (Hurst et al., 1998).

Pregnant Long Evans rats received a single oral dose of TCDD at 0.05, 0.2, 0.8, or 1 µg/kg bw in corn oil on day 15 of gestation. On day 16, these doses resulted in concentrations of 6.8, 15, 50 and 61 pg/g in the fetal compartment and 5.3, 13, 39 and 56 pg/g in single, whole fetuses, with associated maternal body burdens of 31, 97, 520 and 580 pg/g. On day 21 of gestation, the concentration of TCDD were 4.3, 14, 32 and 37 pg/g in the fetal compatment, 4.3, 15, 32 and 36 pg/g in single, whole fetuses and 27, 76, 330 and 430 pg/g in the maternal body. On day 16 of gestation, there was a good correlation between the fetal and maternal body burden and the fetal body burden and maternal blood concentration, suggesting that, at a critical time, maternal blood concentrations provide an estimate of the concentrations of dioxin in the developing fetus. On day 16, 60% of the administered dose was recovered in the dams (Hurst et al., 2000a).

Long-Evans rats were given TCDD repeatedly before (5 days/week for 13 weeks) and during gestation at a dose of 1, 10, or 30 ng/kg bw in corn oil. On day 16 of gestation, the concentration of TCDD in single fetuses was 1.4, 7.8 and 16 pg/g, and the associated maternal body burdens were 19, 120 and 300 ng/kg bw, respectively (Hurst et al., 2000b).

2.1.2 Physiologically based pharmacokinetics modelling

As described above, the toxicokinetics of dioxin-like compounds involves the complex interaction of absorption, transport via the blood, distribution in the lipid and protein fractions of the blood and organs, Ah receptor-dependent induction of hepatic CYP proteins and elimination from the body by metabolism and/or transfer to faecal lipid. These concomitant processes can be quantified by physiologically based pharmacokinetics modelling, in which the toxicokinetics of chemicals is described mathematically within a physiological context, i.e. organs connected by the bloodstream and organ-specific responses after exposure to a chemical (Figure 2).

The first physiologically based pharmacokinetics models of dioxin-like compounds (2,3,7,8-TCDF, King et al., 1983; TCDD, Leung et al., 1988) allowed for only linear kinetics, i.e. they described the accumulation of dioxin-like compound in the body as a process which depends linearly on the administered dose. However, these models could not describe the hepatic sequestration of dioxin-like compounds, in particular the binding of TCDD to induced hepatic CYP 1A2. This deficiency was overcome by introducing Ah receptor-dependent CYP induction and subsequent binding of TCDD to induced CYP 1A2 in the model. The latter model was found to describe well the non-linear kinetics of TCDD, as observed in rodents exposed to TCDD at doses that induce Ah receptor-dependent CYP proteins in the liver (Leung et al., 1990a,b; Andersen et al, 1993; Kohn et al., 1993, 1994, 1996; Andersen et al, 1997a,b; Wang et al., 1997a,b; Zeilmaker & van Eijkeren, 1997).

Physiologically based pharmacokinetics modelling has also been used to describe the kinetics of dioxin-like compounds in humans. Assuming that the body burden of dioxin-like compounds is composed mainly of the amounts in the liver and adipose tissue, Carrier et al. (1995a,b) modelled the non-linear kinetics of dibenzo-para-dioxins and dibenzofurans resulting from their preferential accumulation in human liver. The model was used to describe the kinetics of 2,3,4,7,8-PeCDF in Yu-Cheng patients (see section 2.3.2). van der Molen (1998) developed a generic physiologically based pharmacokinetics model to describe the accumulation of dibenzo-para-dioxins and dibenzofurans in the blood and maternal milk of persons who had been exposed to background concentrations of these compounds. Kreuzer et al. (1997) and Pollitt (1999) used this type of modelling to evaluate the effect of exposure to these compounds in mother’s milk on the long-term body burden. In both cases, the exposure was found to have only a limited effect on the long-term body burden. Zeilmaker and van Eijkeren (1997; 1998) and Zeilmaker et al. (1999) used a slightly different approach: a physiologically based pharmacokinetics model of TCDD in rodents was scaled to humans. Figures 3, 4 and 5 show typical simulations made with this model of the accumulation of TCDD in the human body as expected after a variety of exposure scenarios.

2.1.3 Biochemical effects

Cells exposed to chemicals may respond by increasing the activity of metabolizing enzymes, in particular phase I and phase II enzymes (enzyme induction). Although this mechanism can lead to the removal of chemicals with deletorious effects, enzyme induction also has clear disadvantages. As the induced enzymes often have broad substrate specificity, increased activity may increase the metabolism of chemicals other than the inducing compound. In particular, exposure to persistent chemicals like PCDDs, PCDFs and PCBs can lead to sustained, unwanted changes in chemical metabolism. Examples of the latter effect are the increased metabolism of thyroid hormones after induction of UDP-glucuronosyl transferase (UGT1) activity (see section on Thyroid hormones) and increased estrogen metabolism in the liver (see section on TGF-alpha/EGF pathway).

The following working model prevails for enzyme induction by TCDD (Whitlock et al., 1997). After TCDD has diffused into the cell, it binds to the intracellular Ah receptor protein, which is maintained in its ligand inactivated state by complexation with heat shock protein (hsp)-90. After binding of TCDD, the TCDD–Ah receptor complex dissociates from the hsp-90 protein. This complex than translocates to the nucleus, where it combines with the Ah receptor nuclear translocator (Arnt) to a transcription factor, which may bind to a specific DNA enhancer site, the so-called xenobiotic responsive element. Concurrently, transcription factors bind to gene-specific promotor sites, thereby increasing gene transcription. In this way, TCDD may modulate the transcription of CYP 1A1 (Kedderis et al., 1991; Tritscher et al., 1992; DeVito et al., 1996), CYP 1A2 (DeVito et al., 1996), CYP 1B1, NAD(P)H: quinone oxidoreductase, gluthathione A-transferase Ya subunit and UGT (Whitlock et al., 1997). Increased concentrations of protein may manifest as increased enzyme activities, the activity of ethoxyresorufin O-deethylase (EROD) relating mainly to CYP 1A1 and that of acetanilide-4-hydroxylase and methoxyresorufin O-demethylase mainly to CYP 1A2 (De Jongh et al., 1995; DeVito et al., 1996).

TCDD efficiently induced CYP 1A1 and CYP 1A in rats, in which constitutive and inducible expression of CYP 1A2 is observed only in the liver. After administration of a single dose of TCDD ranging from 1 to 3000 ng/kg bw per day, < 50-fold induction of hepatic EROD activity was observed. Statistically significant induction was observed even at 3 ng/kg bw per day (Abraham et al., 1988). In concordance with this result, single doses of 0.1–1 ng of TCDD led to significant induction of CYP 1A1 mRNA in rat liver. A good correlation was found between hepatic CYP mRNA and EROD activity (r² = 0.93) (van den Heuvel et al., 1994).

Similarly, CYP 1A1 and CYP 1A2 were induced by 23- and 5-fold, respectively, in rats given repeated doses of 3.5–125 ng of TCDD (Tritscher et al., 1992). As shown in Figure 6, whereas the hepatic TCDD concentration increased linearly as a function of the administered dose, the induced protein concentrations increased non-linearly as a function of the hepatic concentration of TCDD, until a maximum was reached.

TCDD-induced EROD activity is not limited to the liver. When B6C3F₁ mice were given TCDD orally at a dose of 1.5–150 ng/kg bw per day on 5 days per week for 13 weeks, the activities of both enzymes were induced in the lungs and the skin at the lowest dose, being 30 times higher than the basal hepatic EROD activity in the lungs and 140 times higher in the skin. The dose–response characteristics of EROD induction were similar in the two organs. The lowest dose also significantly increased the phosphorylation of Cdc2 cyclin-dependent kinase, a protein associated with the G₂ to M phase transition of cells, in the liver but not in skin (DeVito et al., 1994).

Whereas induction of acetanilide 4-hydroxylase is limited to the liver (DeVito et al., 1994), a single dose of TCDD at 0.1, 1, or 10 µg/kg bw to mice induced EROD activity in a dose-dependent fashion in liver, lungs and skin, the EROD activity in the lungs and skin being 6% and 0.6% of the corresponding hepatic activity (Diliberto et al., 1995). Similar observations were made in rats, in which dose-dependent induction of EROD activity and the amount of CYP 1A1 was observed in the liver, lungs and kidneys of rats given a single dose of 0.1, 1, or 10 µg/kg bw. As in mice, the induced EROD activity in the lungs and kidneys represented only a small fraction of the corresponding hepatic activity (lungs, 0.8%; kidneys, 2.5%). In all three organs, a strong correlation was found between induced EROD activity and the amount of CYP 1A1 protein. Similar observations were made for hepatic methoxyresorufin O-demethylase activity, with a dose-dependent increase in activity in the liver, which correlated well with induced CYP 1A2 protein levels. As expected, hardly any CYP 1A2 protein was observed in the lungs or kidneys (Santastefano et al., 1996).

CYP 1A1 and CYP 1A2 can also be induced by compounds other than TCDD. Administration of 2,3,4,7,8-PeCDF at single a dose of 300 µg/kg bw to C57BL/6N and 129/Sv mice caused marked induction of EROD and acetanilide 4-hydroxylase activity in the liver and of EROD activity in the lungs. Similar effects were not found after administration of 2,2´,4,4´,5,5´-HxCB at a dose of 36 mg/kg bw (Diliberto et al., 1999). Hepatic EROD and acetanilide 4-hydroxylase activities were induced in B6C3F₁ mice treated orally on 5 days per week for 4 weeks with TCDD at 0.15 µg/kg bw per day, with 2,3,7,8-TCDF at 1.5 µg/kg bw per day, with OCDF at 150 µg/kg bw per day, with 3,3´,4,4´-TCB at 15 mg/kg bw per day, with 2,3,4´,4,4´,5-HxCB at 30 µg/kg bw per day or with 3,3´,4,4´,5-PeCB at 1.5 µg/kg bw per day. No induction was found with 1,2,3,7,8-PeCDF at 9 µg/kg bw per day or with the PCBs 2,3,3´,4,4´-PeCB at 3 mg/kg bw per day, 2,3,3´,4,4´,5´-HxCB at 300 µg/kg bw per day, or 3,3´,4,4´,5,5´-HxCB at 3 µg/kg bw per day. Although the absolute activity of EROD in the liver was 15-fold higher than that in the lungs, similar dose–response relationships were found in these organs. In skin, increased EROD activity was found only with TCDD, OCDF and the PCBs 3,3´,4,4´-TCB and 2,3,3´,4,4´,5-HxCB (DeVito et al., 1993). Administration of OCDD on 5 days/week for 13 weeks at a dose of 50 µg/kg led to significant induction of EROD in the livers of Fischer 344 rats (Couture et al., 1988).

Significant EROD induction was found in the liver and skin of B6C3F₁ mice given 2,3,7,8-TCDF at 1500 ng/kg bw for 4 and 13 weeks (DeVito & Birnbaum, 1995).

After administration of a single oral dose of 2,2´,4,4´,5,5´-HxCB at 91 mg/kg bw to C57BL/6J mice, no EROD induction was observed in the liver, but induction was observed with the PCBs 2,3,3´,4,4´,5-HxCB at 17 mg/kg bw and 3,3´,4,4´,5,5´-HxCB at 2.1 mg/kg bw. The induction was potentiated by concomitant administration of 2,3,3´,4,4´,5-HxCB and 2,2´,4,4´,5,5´-HxCB, whereas no such potentiation was found with 3,3´,4,4´,5,5´-HxCB, 1,2,3,7,8-PeCDD, 1,2,3,6,7,8-HxCDD, or 2,3,4,7,8-PeCDF (De Jongh et al., 1993). A single dose of 1000 µmol of 2,2´,4,4´,5,5´-HxCB doubled hepatic EROD and acetanilide 4-hydroxylase activity in C57BL/6J mice (De Jongh et al., 1995). Similarly, B6C3F₁ mice given 2,2´,4,4´,5,5´-HxCB at a single dose of 360 mg/kg bw had a 2.5-fold increase in hepatic EROD and MROD and 17-fold increase in that of pentoxyresorufin-O-depentylase (van Birgelen et al., 1996c).

Dietary administration of TCDD (14–1024 ng/kg bw per day), 3,3´,4,4´,5-PeCB (0.47–10.1 µg/kg bw per day), or 2,3,3´,4,4´,5-HxCB (81–729 µg/kg bw per day) to Sprague-Dawley rats clearly induced EROD activity. No such induction was observed with 2,2´,4,4´,5,5´-HxCB at 0.7–6 mg/kg bw per day (van Birgelen, 1995b). The 95% CIs for the NOELs were 0.35–0.89 for TCDD-induced EROD activity and 0.55–3.8 ng/kg bw per day for that of acetanilide 4-hydroxylase (van Birgelen et al., 1995a).

Oral administration of 3,3´,4,4´,5-PeCB to B6C3F₁ mice at a dose of 0.015–1.5 µg/kg bw per day induced EROD and acetanilide 4-hydroxylase activity in the liver and EROD activity in the skin and lungs (LOEL, 0.045–1.5 µg/kg bw per day). The activities of liver-specific enzymes were induced by 2,3,3´,4,4´,5-HxCB at 450 µg/kg bw per day, and the activity of EROD in skin and lung was induced by doses > 1500 µg/kg bw per day. 3,3´,4,4´,5,5´-HxCB at doses up to 3 µg/kg bw per day did not induce enzyme activity. The LOEL for the induction of hepatic EROD activity was 3900 µg/kg bw per day with the PCB 2,3,3´,4,4´-PeCB and 300 µg/kg bw per day with 2,3´,4,4´,5-PeCB, whereas that with TCDD was 0.0015 µg/kg bw per day (DeVito et al., 2000).

Induction of EROD by PCDDs and PCDFs in primary hepatocytes has been found to be a sensitive end-point. Half-maximum EROD activity was achieved by incubating the cells with 16 pmol/l of TCDD, 90 pmol/l of 1,2,3,7,8-PeCDD, 184 pmol/l of 1,2,3,4,7,8-HeCDD, 329 pmol/l of 1,2,3,7,8,9-HeCDD, 441 pmol/l of 1,2,3,6,7,8-HeCDD, 702 pmol/l of 1,2,3,4,6,7,8-HpCDD, or 3859 pmol/l of OCDD (Schrenk et al., 1991). TCDD also induced EROD activity in primary human hepatocytes, although with very different dose–response characteristics. Half-maximum EROD activity was observed in cells exposed to ~ 100 pmol of TCDD (Schrenk et al., 1995).

Like CYPs 1A1 and 1A2, CYP 1B1 is induced by TCDD. When C57BL/6J and DBA/2J mice were given single intraperitoneal doses of TCDD ranging from 0.001 to 50 µg/kg bw, dose-dependent accumulation of CYP 1B1 mRNA was observed in the liver. C57BL/6J mice showed a significant increase in mRNA at doses as low as 0.1 µg/kg bw, and maximum induction (200 times background) was found at 10 µg/kg bw. A much steeper dose–response curve was observed with CYP 1A1 mRNA than with CYP 1B1 mRNA (increase from 0.01 µg/kg; maximum induction at 1 µg/kg). The estimated effective doses at which half the maximum inducibility of CYP 1A1 and CYP 1B1 was observed (ED₅₀) were 0.08 and 1.3 µg/kg bw. Similarly, in DBA/2J mice, significant induction of CYP 1A1 and 1B1 mRNA was found at doses of 1 and 10 µg/kg bw, respectively. Again, the dose–response curve of CYP 1A1 induction was much steeper than that for CYP 1B1, the ED₅₀ values for CYP 1A1 and CYP 1B1 induction being 1.5 and 3.4 µg/kg bw, respectively. The differences in susceptibility of C57BL/6J and DBA/2J mice are due to several mutations in the Ahr^d allele of the Ah receptor, resulting in lower ligand binding affinity: C57BL/6J mice carry the Ahr^b-1 allele, conferring relative high susceptibility to TCDD, and DBA/2J mice carry the Ahr^d allele, conferring relatively low susceptibility to TCDD (Abel et al., 1996).

In rat liver, the induction of UGT1, PAI2 and transforming growth factor (TGF)-alpha mRNA clearly deviated from that of CYP 1A1 mRNA. Although the dose required to increase UGT1 mRNA (1 µg/kg) was much higher than that required to induce CYP 1A1 mRNA (1 ng/kg), doses of TCDD up to 100 µg/kg did not induce PAI2 or TGF-alpha mRNA (Van den Heuvel et al., 1994).

Epidermal growth factor (EGF) is a plasma membrane receptor which, after binding a specific ligand, functions as a signal tranducer regulating cellular proliferation. This effect is mediated by internalization of the ligand–receptor complex and then phosphorylation of intracellular targets by tyrosine kinase. The effects of TCDD on hepatic EGF and their relationship to hepatic carcinogenesis are as follows. TCDD is a known inducer of liver tumours in female, but not male, rats (Kociba et al., 1978). It also induces proliferation of hepatocytes and preneoplastic foci in intact, but not ovariectomized, female rats (Lucier et al., 1991), indicating an important role of estrogens in hepatocarcinogenesis. Furthermore, treatment with TCDD results in a dose-dependent decrease in the number of EGF binding sites in the liver (Lucier et al., 1991; Kohn et al., 1993, short-term exposure to 3.5–125 ng/kg bw per day), indicating internalization of the receptor after TCDD binding (Kohn et al., 1993) and induction of TGF-alpha, a ligand of the EGF receptor.

The interactions of TCDD, EGF, TGF-alpha and inducible CYP enzymes in the liver are shown schematically in Figure 7. The hepatic TCDD–Ah receptor induces not only CYP 1A1 and 1A2 but also expression of the TGF-alpha gene and/or post-transcriptional or post-translational TGF-alpha modifications. This expression may be enhanced by the estrogen receptor–estrogen complex. As a consequence, more TGF-alpha is secreted into the interstitial space in the liver, where it may combine with EGF on the liver cell membrane. The TGF-alpha–EGF receptor complex may then be internalized to exert its biochemical signalling function. The TCDD–Ah receptor complex inhibits synthesis of the estrogen receptor. Finally, estrogen metabolism (estradiol-2-hydroxylase activity) is catalysed by CYP 1A2.

PCDDs, PCDFs and PCBs may affect plasma thyroid hormone levels in one of three ways:

(1) By induction of hepatic microsomal enzymes, resulting in accelerated metabolism and excretion in the bile (Bastomsky, 1977). The stimulation of thyroid metabolism may be compensated by increased amounts of thyroid-stimulating hormone (TSH) in the blood. When sustained, such compensation may result in chronic stimulation of the thyroid and, ultimatally, thyroid cancer (Kohn et al., 1996; Whitlock et al., 1997).

The main pathway is glucuronidation of thyroxine (3,5,3´,5´-tetra-iodothyronine, T4) by UGT1, which catalyses the formation of T4 glucuronides, which are excreted in the bile. At least two forms of glucuronosyltransferase contribute to the activity of T4 UGT: UGT 1A1 (known to be induced by 3-methylcholantrene) and UGT 1A2 (known to be induced by phenobarbital).

Decreased plasma T4 levels were found after exposure of rats to phenobarbital, 3-methylchlolanthrene or the PCB mixture Aroclor 1254. This decrease correlated well with increased T4 UGT activity (Barter & Klaassen, 1992). Furthermore, PCBs and phenobarbital caused increased biliary clearance of T4 (Bastomsky, 1974; McClain et al., 1989; Beetstra et al., 1991). Short-term dietary intake of TCDD (14–1000 ng/kg bw per day) by rats lowered the concentrations of total T4 and free T4 in plasma at doses > 47 ng/kg bw per day. No effect was found on total triiodothyronine (T3) (van Birgelen et al., 1995b). Furthermore, short-term dietary intake of TCDD (14–1000 ng/kg bw per day), 3,3´,4,4´,5-PeCB (0.47–10 µg/kg bw per day) or 2,3,3´,4,4´,5-HxCB (81–730 µg/kg bw per day) resulted in dose-dependen increases in the activity of hepatic UGT 1A1, T4 UGT and CYP 1A1 and concomitant decreases in free and total T4 in plasma. Total T4 in plasma and hepatic T4 UGT activity were negativily correlated, whereas UGT 1A1 and CYP 1A1 activities were positivily correlated. This suggests involvement of the Ah receptor in inducing hepatic T4 glucuroni-dation and, consequently, in modulating thyroid hormone metabolism (van Birgelen et al., 1994a,b, 1995a).

(2) By binding of hydroxy metabolites of PCBs to transthyretin, thereby affecting the binding of T4 (and retinol) to transthyretin, its major transport protein in blood (Brouwer & van den Berg, 1986)

(3) By a direct effect of PCBs on the functioning of the thyroid gland. For example, Aroclor 1254 inhibits proteolysis of thyroglobulin, the protein responsible for release of thyroid hormones from the gland (Collins & Capen, 1980). Furthermore, exposure to TCDD results in an increase in serum TSH concentration and, after long-term exposure, to an increased volume of thyroid follicular cells, increased thyroid weight and thyroid hyperplasia (Bastomsky, 1977; Andrae & Greim, 1992; Hill et al., 1989).

In rats, short-term dietary intake of TCDD (14–1000 ng/kg bw per day) reduced the hepatic concentrations of retinol and retinylpalmitate at the lowest dose tested. The concentration of plasma retinol was increased concomitantly. The mechanism behind this effect consists of induction of CYP enzymes, which oxidize retinol and reduce the activity of acyl coenzyme A:retinyl acyltransferase and lecithin:retinol acyltransferase. Both enzymes participate in the esterification of retinol. In a short-term study in Sprague-Dawley rats treated in the diet, the NOEL of 2,3,3´,4,4´,5-HxCB for depletion of hepatic retinoids was 81 µg/kg bw per day (van Birgelen et al., 1994a). In a similar study, the LOEL of 3,3´,4,4´,5-PeCB for this effect was 0.47 µg/kg bw per day (van Birgelen et al., 1994b).

Disturbed biosynthesis of haem may lead to porphyria, a condition in which precursors of haem accumulate in blood and are hence excreted in the urine.

Short-term intake by rats of diets containing TCDD (14–1000 ng/kg bw per day), 3,3´,4,4´,5-PeCB (0.47–10 µg/kg bw per day), 2,3,3´,4,4´,5-HxCB (81–730 µg/kg bw per day) or 2,2´,4,4´,5,5´-HxCB (0.7–6 mg/kg bw per day) maximally induced a twofold increase in hepatic porphyrin concentration (lowest effect doses: TCDD, 47 ng/kg bw per day; 3,3´,4,4´,5-PeCB, 3.2 µg/kg bw per day; 2,3,3´,4,4´,5-HxCB, 360 mg/kg bw per day). Concomitant administration of TCDD and 3,3´,4,4´,5-PeCB or 2,3,3´,4,4´,5-HxCB resulted in an additional twofold increase in porphyrin concentration. When these compounds were administered with a non-inducing dose of TCDD (33 ng/kg bw per day), however, an 800-fold increase in accumulation of porphyrins was seen, in particular uroporphyrin III and heptacarboxylic porphyrin, in the liver, indicating a strong synergistic action. Furthermore, induced acetanilide 4-hydroxylase activity correlated well with induction of accumulation of porphyrins in the liver. The mechanism behind these effects consisted of CYP 1A2-mediated oxidation of uroporphyrinogen III to uroporphyrin III and induction of the activity of delta-aminolaevulinic acid synthetase, the rate-limiting enzyme in haem biosynthesis, by 2,2´,4,4´,5,5´-HxCB (van Birgelen et al., 1996a).

Short-term intake of TCDD (0.15–450 ng/kg bw per day), 1,2,3,7,8-PeCDD (90–9000 ng/kg bw per day), 2,3,7,8-tetrabromodibenzo-para-dioxin (30–3000 ng/kg bw per day), 2,3,7,8-TCDF (15–1500 ng/kg bw per day), 1,2,3,7,8-PeCDF, 90–9000 ng/kg bw per day), 2,3,4,7,8-PeCDF, 9–900 ng/kg bw per day), OCDF (1.5–150 µg/kg bw per day), 3,3´,4,4´,5-PeCB (0.3 and 15 µg/kg bw per day), 2,3,3´,4,4´-PCB (390–39 000 µg/kg bw per day), 2,3´,4,4´,5-PeCB (3000–30 000 µg/kg bw per day) or 2,3,3´,4,4´,5-HxCB (45–4500 µg/kg bw per day) by mice resulted in hepatic porphyria in all cases (van Birgelen et al., 1996b). The relative potencies of these compounds to induce hepatic porphyrin accumulation were TCDD > PeCDD = tetrabromodibenzo-para-dioxin = 4-Pe-CDF ^a TCDF > 3,3´,4,4´,5-PeCB ^a 1-PeCDF > OCDF > 2,3,3´,4,4´,5-HxCB > 2,3´,4,4´,5-PeCB ^a 2,3,3´,4,4´-PeCB. Again, hepatic CYP 1A2 enzyme activity and total hepatic porphyrin accumulation correlated well.

The Ah receptor appears to play an important role in dioxin-like porphyria, as hepatic porphyria is associated with the inducibility of Ah hydroxylase after administration of TCDD to Ah-responsive C57Bl/6 and Ah-nonresponsive DBA mice (Jones & Sweeney, 1980). Furthermore, non-coplanar PCBs, such as 2,2´,4,4´,5,5´-HxCB and 2,2´,3´,4,4´,5,5´-HpCB do not induce porphyria (Stonard & Greig, 1976; Van Birgelen et al., 1996b; Koss et al., 1993). Finally, hepatic porphyrin accumulation and Ah-receptor dependent benzo[a]pyrene hydroxylation (a well known Ah-receptor response) are increased by 2,2´,3,3´,4,4´-HxCB and 2,2´,3,4,4´,5´-HxCB (Stonard & Greig, 1976).

2.1.4 Toxic equivalency factors

Exposure to PCDDs, PCDFs and coplanar PCBs actually consists of exposure to a mixture of congeners with toxic effects (dermal toxicity, immunotoxicity, carcinogenicity, reproductive and developmental toxicity, disruption of endocrine functions) similar to those of TCDD, the most toxic congener of this class of compounds. This finding led to the development of the concept of toxic equivalency factors (TEFs). In this concept, the toxic potency of a congener, i.e. its toxicity as described in all studies in vivo and in vitro, is expressed relative to that of a reference compound, in this case TCDD (for which the TEF is arbitrarily set at 1).

Application of the TEF concept rests on the following assumptions (van den Berg et al, 1998), noting that the TEF concept does not apply to toxicity that is not mediated by the Ah receptor and does not take into account modulation of Ah receptor-dependent responses by non-Ah-receptor ligands:

• PCDDs, PCDFs and (some) PCBs have a common mechanism of action. In this case, the mechanism consists of binding to the Ah receptor. This binding is considered to be the first, necessary though not sufficient, step in the toxicity of dioxins.
• Congeners exert their combined (dose or concentration) effect additively; i.e. in a mixture of congeners, each congener exerts its toxicity independently of the other congeners in the mixture.

Derivation of congener-specific TEFs is based on evaluation and interpretation of the results of all available studies of toxicity on the basis of expert judgement. The first consultations on TEF concluded that the concept can be applied to PCDDs and PCDFs (international TEFs; NATO/CCMS, 1988) and non-ortho and mono-ortho PCBs (Ahlborg et al., 1994). The TEF concept was re-evaluated in 1998 (van den Berg et al., 1998, 2000).

In attributing a TEF to a compound, the following criteria were used:

• The compound shows a structural relationship to PCDDs and PCDFs.
• The compound binds to the Ah receptor
• The compound elicits Ah receptor-mediated biochemical and toxic effects.
• The compound is persistent and accumulates in the food chain.

Furthermore, TEFs in mammals were derived mainly from studies conducted in vivo, in preference to data generated in vitro and/or quantitative structure–activity relationships. Studies of toxicity were placed in the following order of priority: long-term > short-term > acute. Similarly, Ah receptor-dependent toxic end-points were given priority over biochemical end-points such as enzyme induction. TEFs for mammals were considered to apply to humans too.

Currently, there is consensus on the TEFs for mammals listed in Table 1 (van den Berg et al., 1998, 2000; Scientific Committee on Food, 2000). When combined with data on residues in matrices such as tissue, soil and water, TEFs allow determination of the toxic equivalents concentration of the residue. For a particular residue containing a mixture of i PCDDs, j PCDFs and k PCBs, the toxic equivalence is calculated according to the following equation:

In using the TEF concept, it should be kept in mind that non-Ah receptor-mediated toxicity (decreased dopamine concentration, effects on retinoid and thyroid hormone concentrations and estrogen receptor binding), shown by some PCBs, is not covered. Similarly, the TEF concept does not apply to halogenated compounds other than PCDDs, PCDFs and PCBs which show Ah receptor-dependent toxicity (brominated and chloro or bromo analogues of PCDDs, PCDFs, naphthalenes, diphenyl ethers, diphenyl toluenes, phenoxyanisoles, biphenyl anisoles, xanthenes, xanthones, anthracenes, fluorenes, dihydroanthracenes, biphenyl methanes, phenylxylyl-ethanes, dibenzothiophenes, quaterphenyls, quaterphenyl ethers and biphenylenes). Furthermore, non-additivity was found with mixtures of PCDDs, PCDFs and PCBs. For example, antagonistic effects of non-coplanar PCBs have been described on Ah receptor-dependent effects like induction of EROD and fetal cleft palate in mice. However, synergistic interactions between PCBs and PCDDs/PCDFs have also been reported for effects such as those on CYP 1A1 and thyroid hormone concentrations (van den Berg et al., 1998). Whereas the currently agreed TEFs are based on toxicological evaluations of dose–response relationships between external exposure, i.e. the levels of intake of congeners, and toxicity in organs, future TEFs will be based on the dose–response relationship between internal exposure, i.e. the actual concentrations of congeners in tissues, and toxicity in organs (van den Berg et al., 1998).

2.2.1 Acute toxicity

The acute toxicity of TCDD and related dioxins and furans substituted in at least the 2, 3, 7 and 8 positions can vary widely between and among species (Table 3). For example, in guinea-pigs, an LD₅₀ of 0.6 µg/kg bw was recorded after oral administration, as compared with an LD₅₀ of > 5000 µg/kg bw in Syrian hamsters. Explanations for this variation include differences in the Ah receptor, such as size, transformation and binding to the dioxin response element, pharmacokinetics (metabolic capacity, tissue distribution) and body fat content (Geyer et al., 1990; van den Berg et al., 1994; Pohjanvirta et al., 1998). While data on acute toxicity were available for various commercial PCB mixtures (LD₅₀ values usually > 100 mg/kg bw), few data were available on the acute toxicity of the individual coplanar PCB congeners in mammals. In Ah-responsive rodent species, it is thought that lethality correlates to Ah receptor binding affinity.

One of the commoner symptoms associated with dioxin-induced death is generalized delayed wasting syndrome, characterized by inhibition of gluconeo-genesis, reduced feed intake and loss of body weight. Although some differences exist between species, other toxic responses observed after single doses of dioxins include haemorrhage in a number of organs, thymic atrophy, reduced bone-marrow cellularity and loss of body fat and lean muscle mass. For example, in groups of five or six golden Syrian hamsters of each sex given a single oral or intraperitoneal dose of TCDD at 0, 5, 25, 100, 250, 500, 750, 1000, 2000 or 3000 µg/kg bw, deaths occurred at doses > 500 µg/kg bw, oral administration generally being more toxic. At the highest dose, 33% of males treated intraperitoneally and 80% of those treated orally died. The consistent pathological findings included gradual loss of adipose and muscle tissue (wasting syndrome), thymic atrophy and gastrointestinal lesions. The estimated LD₅₀ values were > 3000 µg/kg bw for intraperitoneal administration and 1200 µg/kg bw for oral dosing (Olson et al., 1980).

Groups of 11 male C57BL/6J mice that were responsive or sensitive to TCDD (Ahb/b) were given a single oral dose of TCDD at 0, 5, 100, 200, 300 or 400 µg/kg bw, and a congenic non-responsive strain (Ahd/d) was given a single oral dose of 0, 400, 800, 1600, 2400 or 3200 µg/kg bw. All mice were observed for 35 days before necropsy. Significant mortality occurred in groups of the Ahb/b strain given doses > 200 µg/kg bw, 100% of animals at the two higher doses dying by about 21 days after dosing; an LD₅₀ value of 160 µg/kg bw was estimated by probit analysis. Conversely, few deaths were seen in the Ahd/d strain, only 33% of those at the high dose dying; an LD₅₀ value of 3400 µg/kg bw was estimated. Decreased body-weight gain, increased liver weight and thymic and splenic atrophy were observed in both strains of mice (Birnbaum et al., 1990).

Groups of 30–60 female Sprague Dawley rats were given 1,2,3,4,6,7,8-HpCB at a total dose of 0, 2.5, 2.8, 3.1, 3.4, 3.8, 4.1, 5 or 10 mg/kg bw by gavage (four doses over 2 days) and then observed for deaths. Many deaths due to lethal wasting, haemorrhage and/or anaemia were observed at the four higher doses, 83–100% of the animals dying within 25 days after dosing. None of the animals at the lowest dose died, while 8.3%, 31% and 66% of those at the next higher doses, respectively, died by day 44 after dosing. The author noted that 30% (9/30) of animals at the lowest dose died from squamous-cell carcinoma of the lungs (Rozman, 1999).

2.2.2 Short-term studies of toxicity

The toxicity of TCDD and related coplanar chemicals in short-term studies is characterized, depending on the route and species, by similar biological and toxicological effects.

In studies of the porphyrinogenic potential of various chlorinated dioxins, furans and coplanar PCBs, groups of five female B6C3F₁ mice were treated on 5 days per week for 13 weeks by gavage with concentrations related by TEFs to doses of TCDD of 0, 0.15, 0.45, 1.5, 4.5, 15, 45, 150 or 450 ng/kg bw per day: 1,2,3,7,8-PeCDD, TEF, 0.05; 2,3,7,8-tetrabromodibenzo-para-dioxin, TEF, 0.15; 2,3,7,8-TCDF, TEF, 0.3; 1,2,3,7,8-PeCDF, TEF, 0.05; 2,3,4,7,8-PeCDF, TEF, 0.5; OCDF, TEF, 0.003; 2,3,3´,4,4´-PeCB, TEF, 0.00001; 2,3´,4,4´,5-PeCB, TEF, 5 x 10^–6; 2,3,3´,4,4´,5-HxCB, TEF, 0.0001; 3,3´,4,4´,5-PeCB, TEF, 0.03. Analysis of hepatic tissue for highly carboxylated porphyrins and CYP 1A1 and 1A2 induction indicated that the binding affinity of the congeners to the Ah receptor in vivo was related to CYP 1A2 induction, which was in turn correlated to hepatic porphyrin accumulation. The LOELs associated with significant increases in total hepatic porphyrin content were: 15 ng/kg bw per day for TCDF; 45 ng/kg bw per day for TCDD; 300 ng/kg bw per day for PeCDD, 2,3,4,7,8-PeCDF and 3,3´,4,4´,5-PeCB, 900 ng/kg bw per day for tetrabromodibenzo-para-dioxin and 1,2,3,7,8-PeCDF; 45 µg/kg bw per day for OCDF; 1500 µg/kg bw per day for 2,3,3´,4,4´,5-HxCB; 7500 µg/kg bw per day for 2,3´,4,4´,5-PeCB; and 13 000 µg/kg bw per day for 2,3,3´,4,4´-PeCB. Induction of hepatic porphyrins by the mono-ortho-substituted PCBs was greater than that estimated from the TEFs assigned to them, which was considered to be related in part to non-Ah receptor-dependent induction of CYP 2B1 and delta-aminolaevulinic acid synthetase (van Birgelen et al., 1996b).

Groups of 12 Sprague-Dawley rats of each sex given TCDD at 0, 0.001, 0.01, 0.1 or 1 µg/kg bw per day by gavage on 5 days/week for 13 weeks had decreased organ and body weights, haematological effects and deaths at the two higher doses. The deaths occurred only at the highest dose, four females dying during the 13 weeks of treatment and two males and two females dying between 14 and 49 days during the 13-week post-dosing observation period. Minor increases in relative liver weight (5–8%) observed in animals at 0.01 µg/kg bw per day were considered to be an adaptive response, as no corresponding histopathological changes were seen. The NOEL was thus 0.01 µg/kg bw per day (equivalent to 0.007 µg/kg bw per day when averaged over the 13 weeks), which resulted in TCDD concentrations in the liver of 2.6–3.7 µg/kg (Kociba et al., 1976).

Groups of six male and six female rats of the same strain were fed diets containing a variety of penta- and hexachlorinated dioxins and dibenzofurans, separately and in a mixture, for 13 weeks: TCDD at 0.2, 2 or 20 µg/kg bw per day; 2,3,4,7,8-PeCDF, 1,2,3,7,8-PeCDF and 1,2,3,6,7,8-HxCD at 2, 20 or 200 µg/kg bw per day; or a mixture of TCDD at 0.2 or 2 µg/kg bw per day, 1,2,3,7,8-PeCDD at 1 or 10 µg/kg bw per day, 2,3,4,7,8-PeCDF at 2 or 20 µg/kg bw per day and 1,2,3,6,7,8-HxCD at 1 or 10 µg/kg bw per day. On the basis of decreases in body weight, histological changes in the liver and thymus and deaths, the relative toxicity of each congener was seen to be related to its binding affinity to the Ah receptor, while the toxicity of the mixture concurred with an additivity concept (TCDD > 2,3,4,7,8-PeCDF > 1,2,3,6,7,8-HxCDF > 1,2,3,7,8-PeCDF > 1,2,3,4,8-PeCDF) (Poiger et al., 1989b).

Male and female Fischer 344 rats were given oral doses of TCDD designed to generate a liver concentration of 0.03, 30 or 150 ng/g. After an initial loading dose of 0.005, 2.5 or 12 µg/kg bw, three maintenance doses of 0.0009, 0.60 or 3.5 µg/kg bw were given every fourth day, and the animals were killed at various times up to day 14. The terminal body weights of males and females at the highest dose were significantly reduced by approximately 11%, while the relative liver weights were increased on average by 30% and 43% at the two higher doses, respectively. Induction of hepatic CYP 1A1 (all doses) and CYP 1A2 (two higher doses) was determined by northern blot hybridization, and dose-dependent induction of a human TCDD-responsive CYP gene was detected in the livers of rats of each sex. Induction of this gene has not been associated with various treatments designed to induce hepatocellular proliferation (Fox et al., 1993).

Two weeks after receiving a protocol designed to initiate preneoplastic lesions (N-nitrosomorpholine at 80 mg/l of drinking-water for 25 days), female Wistar rats, were given subcutaneous injections of various dioxins every 2 weeks for 13 weeks, at approximate daily doses as follows: TCDD, 0, 2, 20 or 200 ng/kg bw; 1,2,3,4,6,7,8-HpCDD, 0, 0.2, 2 or 20 µg/kg bw; or a defined mixture of 49 dioxin congeners, 0, 0.2, 2 or 20 µg/kg bw. At the end of treatment, the body weights of animals at the highest doses of all three treatments were decreased by 5–7%, and the relative liver weights of animals were increased at the two higher doses of TCDD and the highest doses of the other compounds. Hepatic EROD activity was significantly increased in all treated groups when compared with controls, with similar induction at the low, intermediate and high doses. Linear regression analysis of hepatic EROD activity and measured toxic equivalents (ng/g liver) revealed slight differences in the slope of the regression lines (m = 0.87, 0.76, 0.67 for TCDD, HpCDD and the dioxin mixture, respectively; all r₂ = 0.92). When the tumour promoting ability of the three treatments was assessed (relative focal volume of ATPase-deficient preneoplastic liver tissue), similar results were obtained after modelling the toxic equivalents for liver with TCDD and the dioxin mixture; however, the response to the latter was about twofold lower at the highest dose. The authors concluded that TEFs based on enzyme induction in vitro provide only an approximation of the tumour promoting ability of dioxin congeners and gave an overestimate of the response to the HpCDD (Schrenk et al., 1994).

Groups of six male and six female Sprague-Dawley rats were treated by gavage with total doses of 1,2,3,6,7,8-HxCDF of 0, 18, 220, 1300, 4000 or 6000 µg/kg bw for females and 0, 31, 370, 2200, 6700 or 10 000 µg/kg bw for males over 13 weeks; the total dose was divided into four daily loading doses, each comprising about 13% of the total dose, and was followed by six maintenance doses given every 2 weeks, each equivalent to 7.8% of the total dose. Treatment induced a variety of biochemical and toxic effects similar to those seen with TCDD in a group of rats given a total dose of 42 µg/kg bw for females and 70 µg/kg bw for males by the same dosing regimen. On the basis of the measured end-points (deaths, hepatic EROD induction, decreased plasma T4 concentration, haematological indices), the TEF for this congener was estimated to be 0.007, in close agreement with the WHO-assigned TEF of 0.01 (Viluksela et al., 1997).

In an experiment of a similar design, groups of 20 rats of the same strain were given total toxic equivalents of 0, 0.14, 1.7, 10, 31 or 47 µg/kg bw for females and 0, 0.22, 2.6, 16, 47 or 70 µg/kg bw for males, with contributions of similar toxic equivalents from TCDD (TEF, 1), 1,2,3,7,8-PeCDD (TEF, 0.2), 1,2,3,4,7,8-HeCDD (TEF, 0.05) and 1,2,3,4,6,7,8-HeCDD (TEF, 0.007). The same regimen of four daily loading doses (each comprising about 10% of the total toxic equivalents) followed by six maintenance doses (each comprising about 10% of the total toxic equivalents) every 2 weeks for 13 weeks. Whereas there was a significant, dose-dependent increase in hepatic EROD activity even at the lowest toxic equivalents (0.14 and 0.22 µg/kg bw for females and males, respectively), most of the additional biochemical end-points (decreased hepatic phosphoenolpyruvate carboxykinase activity, serum glucose and serum total T4) were affected only by total toxic equivalents > 10 µg/kg bw. Overall, the effects seen with the toxic equivalent mixture, TCDD or 1,2,3,6,7,8-HxCDD alone were comparable (mortality rate, growth reduction, hepatic enzyme induction, haematological effects), providing support for the TEF and additivity concept for chlorinated dioxins (Viluksela et al., 1998a,b).

In an experiment designed to identify the lowest effective doses and body burdens of TCDD in rats, groups of eight female Sprague-Dawley rats were fed diets formulated to deliver TCDD at a dose of 0, 14, 26, 47, 320 or 1000 ng/kg bw per day for 13 weeks. The most sensitive effects, seen at the lowest dose, included hepatic CYP 1A1 and 1A2 induction and significant decreases in thymus weights and hepatic retinol concentration. At the higher doses, differences in liver, kidney and spleen weights and decreases in plasma T3 and free T4 concentrations were seen. The estimated NOEL (by sigmoidal curve fitting) for hepatic EROD induction was 0.35 ng/kg bw per day, which corresponds to a liver TCDD content of 0.037 ng/g (van Birgelen et al., 1995a,b).

Thyroid hormone status was assessed in rats in a short-term assay for tumour promotion. After initiation with N-nitrosodietheylamine at 70 days of age, female Sprague-Dawley rats were treated with TCDD by gavage every 2 weeks for 30 weeks at doses designed to deliver 0, 0.1, 0.35, 1, 3.5, 11, 36 or 125 ng/kg bw per day. The rats were necropsied 1 week after the last TCDD dose, and serum samples were analysed for T3, T4 and TSH. T4 concentrations were significantly reduced at doses of TCDD > 11 ng/kg bw per day in the initiated rats and > 36 ng/kg bw per day in the uninitiated animals (maximum reduction, 42%). While there was no effect on T3 concentration, that of serum TSH was increased by about 2.5-fold in uninitiated rats at the highest dose (3.3 ng/ml, with 1.3 ng/ml in controls). Histological changes in the thyroid included diffuse follicular hyperplasia; in rats at 3.5 ng/kg bw per day, the ratio of parenchymal to follicular area was significantly increased. Hepatic CYP 1A1 and UGT1 mRNA levels were increased at doses > 0.35 ng/kg bw per day and > 3.5 ng/kg bw per day, respectively (Sewall et al., 1995).

Groups of 10 Hartley guinea-pigs of each sex were given diets containing TCDD at a concentration of 0, 2, 10, 76 or 430 ng/kg of diet for 13 weeks (equal to 0, 0.12, 0.61, 4.9 and 26 ng/kg bw per day for males and 0, 0.14, 0.68, 4.9 and 31 ng/kg bw per day for females). The effects induced were similar to those in rats, including deaths at the highest dose. The NOEL for changes in organ and body weights and clinical effects was 0.61 ng/kg bw per day, confirming the greater sensitivity of this species to TCDD (DeCaprio et al., 1989).

When eight female rhesus monkeys were given a diet containing TCDD at a concentration of 500 pg/g for 9 months, dermatological effects were seen by 3 months and changes in haemoglobin and erythrocyte volume fraction by 6 months, which persisted and increased in severity up to the end of treatment. One animal died after 7 months on the diet, and four further deaths occurred within 2 months after removal from the diet, after total TCDD intakes estimated to be 19 and 12 µg/kg bw per day (Allen et al., 1977). Similar effects were seen when three male rhesus monkeys were fed diets containing 2,3,7,8-TCDF at 5 or 50 µg/g for up to 6 months, except that the animals that survived to the end of treatment tended to recover quickly after being removed from the diets (McNulty et al., 1981).

2.2.3 Long-term studies of toxicity and carcinogenicity

Previous WHO expert groups have evaluated the carcinogenicity of PCDDs, PCDFs and PCBs (IARC 1987, 1997; van Leeuwen & Younes, 2000). Most of the long-term experiments designed to determine the toxicity of dioxin and coplanar chemicals were conducted with various rodent species or non-human primates. The carcinogenic effects assessed in long-term studies are summarized in Table 4.

Groups of 45 male Swiss mice were given TCDD by gavage at a dose of 0, 0.007, 0.7 or 7 µg/kg bw per week, equal to 0.001, 0.1 and 1 µg/kg bw per day, for up to 1 year. Dose-dependent increases in the incidence of both ulcerative skin lesions and amyloidosis were observed at the two lower doses and a decreased lifespan in mice at the highest dose. An increased incidence of liver tumours (hepatomas and hepatocellular carcinomas) was observed at the intermediate dose, but a similar increase at the highest dose was not significant (Tóth et al., 1979).

In a study of the carcinogenicity of TCDD in two rodent species, groups of 50 Osborne-Mendel rats and 50 B6C3F₁ mice of each sex were given TCDD by gavage twice a week for 104 weeks at a dose of 0, 0.01, 0.05 or 0.5 µg/kg bw per week for rats and male mice (equal to 0, 0.0014, 0.0071 and 0.071µg/kg bw per day) and 0, 0.04, 0.2 or 2 µg/kg bw per week for female mice (equal to 0, 0.006, 0.03 and 0.3 µg/kg bw per day). Decreased body-weight gain was seen in male and female rats at the highest dose, and an increased incidence of hepatic lesions described as ‘toxic hepatitis’ was seen in both species at the highest dose. Increased incidences of follicular-cell adenomas or carcinomas of the thyroid were found in male rats at the two higher doses (16% and 22%, respectively), with a non-significant increase (13%) in female rats. Furthermore, 24% of female rats at the highest dose had neoplastic nodules in the liver and 6% had hepatocellular carcinoma. The incidence of hepatocellular carcinoma was also increased in male (34%) and female (13%) mice at the highest dose. Females at this dose had an increased incidence of follicular-cell adenomas of the thyroid (11%) and an increased incidence (with a dose-related trend) of histiocytic lymphomas in the haematopoietic system (National Toxicology Program, 1982).

Groups of 10 male Sprague-Dawley rats were maintained on diets containing TCDD at a concentration of 0, 0.001, 0.005, 0.05, 0.5, 1, 5, 50, 500 or 1000 ng/kg for up to 78 weeks, equal to weekly doses of 0, 0.0003, 0.001, 0.01, 0.1, 0.4, 2, 24, 240 and 500 µg/kg bw or daily doses of 0.00004, 0.00014, 0.0014, 0.014, 0.057, 0.29, 3.4, 34 and 71 µg/ kg bw. All animals at the five higher doses died, the time to 100% mortality ranging from week 3 for animals at 24 µg/kg bw per week to week 31 for animals at 0.4 µg/kg bw per week. Further deaths (40–50% of animals) occurred at 0.001, 0.01 and 0.1 µg/kg bw per week before the end of the study. Neoplasms were found at multiple sites in 57% of animals at doses > 0.0001 µg/kg bw per week. Among those reported were ear-duct carcinoma, renal adenocarcinoma, skin angiosarcoma, Leydig-cell adenoma, fibrosarcoma, squamous-cell carcinoma of the skin and lung, glioblastoma, astrocytoma, cholangiocarcinoma and lymphocytic leukaemia. No tumours were found at the lowest dose (Van Miller et al., 1977). The Committee noted that the small number of animals used and the high mortality rates limit interpretation of this study.

Groups of 50 Sprague-Dawley rats of each sex (86 rats of each sex as vehicle controls) were fed diets formulated with TCDD to provide a dose of 0, 0.001, 0.01 or 0.1 µg/kg bw per day for 2 years. Body weight and food consumption were measured throughout the study; haematological examinations and urinary analyses were performed on eight rats of each sex per group after 3, 12 and 23 months. Serum samples were collected twice during the study. The biochemical and histopathological examinations were extensive. Increased incidences of hepatocellular carcinoma, squamous-cell carcinomas of the lung, hard palate and tongue were observed at the highest dose. TCDD not only affected the incidence rates of cancer but had additional toxicological effects, particularly at the highest dose, which included increased mortality (females only), decreased body-weight gain, splenic and thymic atrophy, hepatic degeneration and necrosis. On the basis of increased urinary excretion of porphyrins and delta-aminolaevulinic acid and hyperplastic nodules in the liver in females at 0.01 µg/kg bw per day, the NOEL was 1 ng/kg bw per day, which, at termination, resulted in a concentration of TCDD of 540 ng/kg in fat and liver (Kociba et al., 1978b).

As part of a study of reproductive toxicity, groups of female rhesus monkeys were given diets containing TCDD at 5 or 25 pg/g diet for 3.5–4.0 years, providing doses of 0.15 and 0.67 ng/kg bw per day. Animals at the higher dose showed marginal signs of toxicity (Bowman et al., 1989; see section 2.2.5 for details).

2.2.4 Genotoxicity

Previous WHO expert groups have evaluated the genotoxicity of PCDDs, PCDFs and PCBs (IARC, 1987, 1997; van Leeuwen & Younes, 2000). Several short–term assays for genotoxicity with TCDD covering various end-points gave primarily negative results. Furthermore, TCDD did not bind covalently to mouse liver DNA.

TCDD did not induce mutations in Salmonella typhimurium strain TA98, TA100, TA1535, TA1537 or TA1538 with or without the addition of an exogenous metabolic activation system (Geiger & Neal, 1981; Mortelmans et al., 1984). Assays for Tk^+/– mutation in mouse lymphoma L5178Y cells gave variable results, the outcome depending on the mutant selection protocol used, methotrexate or thymidine selection leading to a positive response and ouabain or arabinose C selection leading to a negative response. Thioguanine selection resulted in a weakly positive response (Rogers et al., 1982; McGregor et al., 1991).

Sister chromatid exchange and micronuclei were found in human lymphocytes treated with TCDD, in the absence or presence of alpha-naphthoflavone (Nagayama et al., 1993, 1994).

In assays for cell transformation in C3H10T1/2 mouse and rat tracheal epithelial cells, TCDD increased the formation of foci in cells initiated with N-methyl-N-nitro-N-nitrosoguanidine (Abernethy et al., 1985; Tanaka et al., 1989). TCDD also transformed Ad12-SV40-immortalized cells but not primary human epidermal keratinocytes, as revealed by growth in soft agar, and increased foci formation, cell density and the carcinogenic response in nude mice. The transformed cells caused a 100% incidence of squamous-cell carcinomas when injected into nude mice and 0% in control mice (Yang et al., 1992).

TCDD did not induce unscheduled DNA synthesis in normal mammary epithelial cells (Eldridge et al., 1992).

OCDD did not induce mutations in S. typhimurium (Zeiger et al., 1988), and 1,2,3,6,7,8-HeCDD and 1,2,3,7,8,9-HeCDD did not transform C3H10T1/2 mouse cells (Abernethy & Boreiko, 1987).

TCDD did not bind to mouse liver DNA (Turteltaub et al., 1990), but it induced DNA single-strand breaks in the liver and in peritoneal lavage cells in rats (Wahba et al., 1988, 1989; Alsharif et al., 1994).

When administered concomitantly with 12-O-tetradecanoylphorbol 13-acetate, TCDD increased the cell transforming capacity of peritoneal macrophages in mice in a dose-dependent manner (Massa et al., 1992).

TCDD enhanced alpha-naphthoflavone-induced sister chromatid exchange frequency in cultured rat lymphocytes (Lundgren et al., 1986). It did not induce sister chromatid exchange, micronuclei or chromosomal aberrations in mouse bone marrow (Meyne et al., 1985) or in lymphocytes of persons who had been exposed to high concentrations of TCDD (Reggiani, 1980; Tenchini et al., 1983; Zober et al., 1993).

TCDD increased the mutagenic and recombinogenic activity of N-ethyl-N-nitrosourea in the mouse spot test by twofold (Fahrig, 1993).

In lacI transgenic rats, TCDD increased neither the mutation frequency nor the mutation spectrum in the liver (Thornton et al., 2001). Similarly, TCDD did change the spontaneous spectrum of H-ras codon 61 point mutations in mouse liver, nor did it affect the mutation spectrum of H-ras mutations in hepatocellular adenomas and carcinomas after treatment of mice with vinyl carbamate (Watson et al., 1995).

2.2.5 Reproductive toxicity

In a three-generation study of reproductive toxicity, male and female Sprague-Dawley rats (16 males and 32 females in the control and high-dose groups; 10 males and 20 females at the low and intermediate doses) were maintained on diets containing TCDD designed to deliver a dose of 0, 0.001, 0.01 or 0.1 µg/kg bw per day. After 90 days on diet, F₀ rats were bred to produce the F_1a generation and then again, 33 days after weaning, to produce the F_1b generation. The F_1b and F₂ litters were mated when the animals were about 130 days of age to produce the F₂ and F₃ generations, respectively. Fertility, litter sizes and neonatal survival were severely decreased for animals at the highest dose at the F₀ matings and in ensuing generations at the intermediate dose. Although slight effects were seen on pup survival and renal morphology at the low dose, they did not occur consistently across all generations. The NOEL was 0.001 µg/kg bw per day (Murray et al., 1979). The steady-state body burdens of TCDD of the rats at the two lower doses were estimated to have been 0.029 µg/kg bw and 0.29 µg/kg bw, respectively (Peterson et al., 1993).

Groups of 30 pregnant Wistar rats were given a single dose of 3,3´,4,4´,5,5´-HxCB at 0, 0.2, 0.6 or 1.8 mg/kg bw by gavage on day 1 of gestation, and their pups were followed through to breeding about 1 year later. In the F₁ generation, fecundity and fertility were decreased at the highest dose, only 2/16 pairs resulting in a pregnancy, in comparison with 15/17 and 11/17 at 0.2 and 0.6 mg/kg bw, respectively. After a second breeding, in which 14 treated males and seven treated females at 1.8 mg/kg bw were mated to control animals, fertility was again reduced, with no pregnancies and possible effects on unspecified aspects of male reproductive behaviour (Smits-van Prooije et al., 1993).

Groups of 10–12 female B6C3F₁ mice were treated by gavage with various dioxin, furan and PCB congeners once every 3 weeks for five dosing periods to assess the survival of autotransplanted endometrial lesions. The doses used were 0, 1, 3 and 10 µg/kg bw for TCDD; 0, 3 and 30 mg/kg bw for 2,2´,4,4´,5,5´-HxCB; 0, 100, 300 and 1000 µg/kg bw for 3,3´,4,4´,5-PeCB; 0, 10, 30 and 100 µg/kg bw for 2,3,4,7,8-PCDF; and 0, 2 and 20 mg/kg bw for 1,3,6,8-TCDD. Endometrial tissue was transplanted during the second dosing period. Hepatic EROD activity was induced in a dose-dependent manner at the two highest doses of TCDD and at all doses of 3,3´,4,4´,5-PeCB and PCDF. The diameter of the endometrial lesions was significantly increased only at 1 and 3 µg/kg bw of TCDD and 100 µg/kg bw of PCDF. Whereas the weights of the lesions appeared to decrease with increasing dose of TCDD, the weights were still higher than those of controls. No significant effects were seen on ovarian or uterine horn weights, but the thymus weights were decreased and the liver weights increased at the two higher doses of PCDF. Examination of ovarian tissue showed the absence of active corpora lutea and increased numbers of regressive corpora lutea in mice at 10 µg/kg bw TCDD and 100 and 1000 µg/kg bw 3,3´,4,4´,5-PeCB (Johnson et al., 1997). In a study of similar experimental design with TCDD only, a dose of 10 µg/kg bw significantly increased the diameter of the endometriotic site in female Sprague-Dawley rats, and doses of 3 and 10 µg/kg bw decreased the ovarian weight by 15% and 27%, respectively (Cummings et al., 1996).

Groups of 25 female Wistar rats were given an initial subcutaneous loading dose of [¹⁴C]TCDD at 0, 25, 60 or 300 ng/kg bw, followed by weekly maintenance doses of 0, 5, 12 or 60 ng/kg bw, beginning 2 weeks before the beginning of mating and continually through mating, gestation and lactation. The size of the maintenance doses was based on a reported elimination half-life of 3 weeks for adult rats. The dams were killed after weaning, and 20 male pups per group were either assessed at 70 or 170 days of age for sexual development (sex organ weights, sperm analysis) or bred to untreated females on postnatal day 170. The only reproductive or pregnancy outcome that was significantly affected was the pregnancy index (16% reduction from that of controls) and only at the highest dose (300 ng/kg bw loading plus 60 ng/kg bw maintenance). The numbers of sperm from the cauda epididymis and daily sperm production were decreased (the latter by up to 50%) and the epididymal sperm transit rates and per cent abnormal sperm increased at all doses on postnatal day 170. No effects were seen on reproductive performance (mating, pregnancy, fertility). At the highest dose, the serum testosterone concentration was decreased at adulthood, and permanent changes in the testicular tubuli were found, including pyknotic nuclei and the occurrence of cell debris in the lumen. Mounting and intromission latencies were significantly increased at the lowest and highest doses. The LOEL was 25 ng/kg bw loading plus 5 ng/kg bw maintenance, equivalent to 0.8 ng/kg bw per day, which, by the end of weaning corresponded to 0.24 ng/g liver in male pups (Faqi et al., 1998).

Groups of 5–10 immature female Sprague-Dawley rats were given TCDD at a single dose of 0, 0.3, 1, 3, 10, 30 or 60 µg/kg bw by gavage, and ovarian follicular maturation was induced 24 h later by a single intraperitoneal dose of 20 IU of equine chorionic gonadotropin. The rats were killed 24, 48, 56 and 72 h later (study termination), and body weight, ovarian weights, ovulation status and serum prolactin, luteinizing hormone and follicle-stimulating hormone were measured. In a separate experiment, rats were hypophysectomized before dosing with TCDD at 0, 3, 10 or 60 µg/kg bw, equine chorionic gonadotropin (same schedule) and a single dose of 10 µg bovine luteinizing hormone. Reduced body-weight gain and ovarian weights were seen in rats at doses > 10 µg/kg bw, and ovarian weights were reduced by about 50% in rats at the high dose at termination. Ovulation, as defined by the number of ova per rat, was also significantly reduced at doses > 10 µg/kg bw (maximum reduction, about 80% at 30 and 60 µg/kg bw). The serum concentrations of estradiol were significantly increased over control values at 10 and 60 µg/kg bw 48–56 h after administration of equine chorionic gonadotropin, the concentrations in rats at the high dose being more than four times higher than that of controls at termination. Treatment at 10 µg/kg bw also resulted in a 13–15-fold increase in the serum concentrations of luteinizing and follicle-stimulating hormone and suppression of the preovulatory surge induced in control animals 54 h after administration of equine chorionic gonadotropin. Similar decreases in body and ovarian weight and inhibition of ovulation were seen in the hypophysectomized animals treated with TCDD at 10 and 60 µg/kg bw (Li et al., 1995).

Experiments in which rhesus monkeys where fed diets containing TCDD at 50 or 500 pg/g for 7 months resulted in reproductive failure: only two of eight monkeys at 50 pg/g and one of eight at 500 pg/g carried their infants to full term. It was estimated that the monkeys at 50 pg/g group had consumed 1.8 µg of TCDD during the 7 months before the initiation of breeding, equivalent to 1.5 ng/kg bw per day (Allen et al., 1979).

In relevant studies involving nonhuman primates, groups of eight female rhesus monkeys were fed diets containing TCDD at 0, 5 or 25 ng/kg for 3.5 or 4.0 years, respectively. After 7 months of treatment, the monkeys were bred to untreated males, produced the F₁ generation and were bred again after 27 months on diet. While reproduction was not affected at 5 ng/kg, only 20% of the combined breedings of animals at 25 ng/kg diet produced viable offspring, while an average of 75–80% of breedings of controls and animals at 5 ng/kg succeeded (Bowman et al., 1989). The concentration of TCDD detected in mesenteric fat samples from weaned F₁ infants of mothers at 5 ng/kg of diet at 5 months of age was 380 ± 140 pg/g tissue. The total amount of TCDD ingested by the mothers was estimated to have been 370 ng after 16.2 months on diet (Schantz et al., 1992). The body burdens of the maternal animals were estimated to be 25–42 ng/kg bw.

While there was no indication of maternal toxicity at the time, analysis of the monkeys 10 years after termination of the study revealed a dose-dependent increase in the frequency and severity of endometriosis, and three monkeys at the higher dose died due to severe peritoneal endometriosis. Endometrial lesions were detected in 33% of control monkeys and 71% and 86% of those at 5 and 25 ng/kg of diet, respectively (Rier et al., 1993). Moderate-to severe disease was seen only in TCDD-treated animals (43% and 71% at 5 and 25 ng/kg of diet, respectively). As there was a high concentration of coplanar PCBs in the blood of monkeys in which endometriosis was originally diagnosed, there may have been an unknown source of exposure to PCBs that caused the reported lesions (Rier et al., 2001a). The average dietary concentration of PCBs was reported to be 7.8 µg/kg (Schantz et al., 1992), which, on the basis of the values for TCDD intake, would represent an average intake of 0.2 µg/kg bw per day.

Groups of five or six female cynomolgus monkeys were given TCDD-containing gelatin capsules 5 days/week for 12 months after surgical auto-implantation of endometrial strips at multiple abdominal sites. The average doses of TCDD delivered were 0, 0.71, 3.6 and 18 ng/kg bw per day. Laparoscopic examinations were conducted on all monkeys at 1, 3, 6 and 12 months (study termination). At necrospy, significantly more endometrial strips were found to have survived in monkeys at the two higher doses than in controls (27% and 33% vs. 16%, respectively), while the size of the implants was increased only at the highest dose (maximum and minimum length and width measured). The surviving endometrial strips in animals at the lowest dose actually regressed in size. The serum concentrations of interleukin (IL)-6 were significantly decreased, while those of IL-6 soluble receptor were increased in monkeys at the highest dose at termination, with a 2.5-fold change for both cytokines (Yang et al., 2000).

Dioxins, and specifically TCDD, induce a distinct series of developmental effects, including fetal mortality, structural malformations and postnatal functional alterations, in a variety of species at doses below those associated with maternal toxicity. These effects are thought to be due in part to interactions with the Ah receptor and its related transcriptional factor, ARNT, the expression of which appears to be involved in aspects of normal embryonic development (Kozak et al., 1997; Abbott et al., 1999).

In most species tested to date, TCDD can induce significant embryolethality (early or late resorptions, abortions, stillbirths), which is usually associated with indications of maternal toxicity. The timing of dosing and the age of the embryo or fetus have been shown to be major determinants of TCDD-induced prenatal mortality. Characteristic, sensitive indications of the teratogenicity of TCDD in responsive strains of mice include induction of cleft palate and hydronephrosis at doses not associated with maternal toxicity. Various strains of pregnant mice treated with TCDD at doses of 1–10 µg/kg bw per day during days 6–15 of gestation (organogenesis) produced litters with these effects in a dose-dependent fashion (Neubert et al., 1973; for review, see Environmental Protection Agency, 2000a, Part II, Chapter 5). While cleft palate in the absence of fetal or maternal toxicity usually occurs only in mice, common effects in rats and hamsters include renal malformations, oedema and gastrointestinal haemorrhage. The ability of most coplanar chemicals to induce frank developmental effects is related to their binding affinity to the Ah receptor.

Female C57Bl/6N mice were given a single oral dose of TCDD at 0, 6, 12, 15 or 18 µg/kg bw on day 10 of gestation or 0, 6, 9, 12 or 15 µg/kg bw on day 12 and then killed before parturition on day 18. For comparison, mice were treated with retinoic acid at 0, 10, 20, 30, 40, 50 or 60 mg/kg bw on day 10 of gestation or at 0, 40, 80, 100, 120, 160 or 200 mg/kg bw on day 12. Groups of mice were also treated with various combinations of TCDD (at 0, 3, 6, 12 or 15 µg/kg bw) and retinoic acid (0, 5, 10, 20, 30, 40 or 80 mg/kg bw) on day 10 or 12 of gestation. Decreased maternal body-weight gain was seen only in mice treated with TCDD on day 10 of gestation, whereas the relative liver weight was increased at all doses of TCDD. There was a trend towards a decrease in maternal body-weight gain in mice treated with retinoic acid on day 10 and increased relative liver weight after treatment on both days 10 and 12. The increase in relative liver weight seen in mice treated with both chemicals was similar to that seen with TCDD alone. No effects were seen on the mortality rate or body weight of fetuses of TCDD-treated mice, while the body weights of fetuses of mice at the two higher doses of retinoic acid given on day 12 of gestation were reduced by 6–10%. All treated groups had 6–20 litters. Hydronephrosis was seen in all groups treated with TCDD, with a dose-dependent increase in severity. The median effective dose (ED₅₀) for this lesion was estimated to be 3.9 µg/kg bw for both times of treatment. The incidence of cleft palate was also increased by TCDD at the three higher doses at both times of treatment, with approximate ED₅₀ values of 12 µg/kg bw for treatment on day 12 and 16 µg/kg bw for treatment on day 10. Major defects observed in the mice treated with retinoic acid consisted of skeletal abnormalities and cleft palate, with ED₅₀ values for the latter of 42 mg/kg bw for treatment on day 10 and 190 mg/kg bw for treatment on day 12. Combined treatment with the two chemicals had no effect on either TCDD-induced hydronephrosis or retinoic acid-induced skeletal anomalies, but they had an interactive effect on cleft palate induction. For example, at doses of retinoic acid that induced no cleft palate (5 and 10 mg/kg bw), concomitant administration of TCDD at 12 µg/kg bw increased the incidence by approximately 2.4- and 4.2-fold, respectively (21% per litter as compared with 50% and 88%) (Birnbaum et al., 1989).

Groups of 7–10 pregnant C57Bl/6 mice were given a single oral dose of 3,3´,4,4´,5-PeCB at 0, 130, 260 or 520 µg/kg bw on day 10 of gestation and were killed on day 17. Significant increases in relative liver weights were seen in maternal animals at all doses, but there were no effects on litter size, litter weight or fetal mortality. Dose-dependent increases in the incidence of both hydronephrosis and cleft palate were seen, with ED₅₀ values of 160 µg/kg bw and 360 µg/kg bw, respectively (Mayura et al., 1993).

Various structural end-points associated with the developing urogenital system of rodents have been shown to be extremely sensitive to perturbation by TCDD and coplanar chemicals. In males in particular, indices of delayed puberty (rats and hamsters), reductions in prostate growth and development (rats and mice), decreased epididymal and cauda epididymal weights (rats), decreased testicular and epididymal sperm numbers (rats, mice, hamsters) and decreased numbers of ejaculated sperm (rats and hamsters) have been observed (Roman & Peterson, 1998). In certain strains of rat (Holtzman and Long Evans), TCDD given as single doses as low as 0.05–0.064 µg/kg by gavage on day 15 of gestation has been associated with reduced ventral prostate weights and reduced epididymal and ejaculated sperm counts (Mably et al., 1992a,b,c; Gray et al., 1995, 1997a,b).

Pregnant Holtzman rats were given a single oral dose of TCDD at 0 or 1.0 µg/kg bw on day 15 of gestation, and the litters were collected at various times up to day 21 or the pups were collected after parturition up to postnatal day 5. Additional maternal rats were treated with a single oral dose of 0, 0.064, 0.16, 0.4 or 1 µg/kg bw on day 15 of gestation, and the postnatal physical and sexual development of the offspring was followed up to postnatal day 120. While no maternal toxicity was seen, fetal survival was decreased by 6–15% and neonatal body weight was decreased by 7–22% at the highest dose. When the food intake of male offspring was assessed after lactation, it was found to be decreased at the two higher doses, by 13–21% from that on postnatal day 28–56. This contributed to a decrease in body weight of 8–19% in the same groups up to postnatal day 70 (post-pubertal). After this time and until sexual maturity (postnatal day 120), no significant differences were seen for either parameter.

With regard to physical development, eye opening was accelerated by approximately 1 day in animals at the highest dose when compared with controls, while testis descent (androgen-dependent) was delayed in a dose-dependent manner at doses > 0.16 µg/kg bw (maximum delay, 2 days). A decreased relative anogenital distance were seen on day 4 in male pups of dams at doses > 0.16 µg/kg bw. Treatment at 1 µg/kg bw also caused a significant decrease (53%) in the post-parturition testosterone surge normally seen in male pups within 2–4 h after birth. Although there was a trend towards decreased plasma testosterone and dihydrotestosterone concentrations in male offspring from postnatal day 32–120, none of the changes was significant. The plasma concentration of luteinizing hormone in males at the highest dose was significantly reduced by about 95% on postnatal day 32. The weights of the seminal vesicle and ventral prostate were reduced in male offspring of dams at doses > 0.16 µg/kg bw throughout sexual development, but at the highest dose only the latter effect persisted at sexual maturity (postnatal day 120) (Mably et al, 1992a).

Male offspring from litters in the above study, standardized on postnatal day 1 to five pups of each sex per litter from dams treated with graded doses of TCDD on day 15 of gestation, were also assessed for sexual behaviour on postnatal days 56–63, 70–77 and 112–119. The latencies to mounting and to ejaculation were significantly increased in pups of dams at doses > 0.16 µg/kg bw at all times. While the number of intromissions required before ejaculation was not significantly affected, the number of mounts required before ejaculation was slightly increased in all groups during testing on postnatal days 112–119. The copulatory rate (mounts plus intromissions per minute) was significantly decreased in a dose-dependent manner from that of controls at all times (maximum reduction, about 53%) in male offspring of dams at the three higher doses. Feminine sexual behaviour (lordosis) was also significantly increased in castrated male offspring of dams at doses > 0.16 µg/kg bw, while a feminized pattern of luteinizing hormone surge after progesterone treatment was seen in castrated males after treatment of dams at the two higher doses. Hepatic EROD activity, when determined on postnatal day 120, was not significantly increased at any dose (Mably et al., 1992b).

Male pups isolated after lactation were assessed for various aspects of the development of the reproductive system on postnatal day 32, 39, 63 and 120 (juvenile, pubertal, postpubertal and sexually mature, respectively). Although slight decreases in paired testis weights were observed at all doses except 0.16 µg/kg bw on postnatal day 32, no significant difference was apparent by postnatal day 120. The weight of the right epididymis was decreased in a dose-dependent manner at a maternal dose as low as 0.064 µg/kg bw on postnatal days 49 and 120 and 0.16 µg/kg bw on postnatal days 32 and 63, with a 22–30% decrease at the highest dose. The weights of the cauda epididymis were lower than those of controls at all doses after puberty (postnatal days 63 and 120). Daily sperm production and cauda epididymal sperm numbers were also decreased at all doses on postnatal days 63 and 120, those of pups of dams at the highest dose being as little as 74% and 44% of control values on postnatal day 120. No significant effects were seen on cauda epididymal sperm motility or morphology. The plasma concentration of follicle-stimulating hormone was slightly reduced only on postnatal day 32 at all maternal doses except 0.16 µg/kg bw. When the male offspring were mated with control females on postnatal day 70, no effects were observed on fertility, gestation index, litter size or pup survival (Mably et al., 1992c).

Groups of 12 pregnant Long Evans rats were given TCDD at a single dose of 0, 0.05, 0.2 or 0.8 µg/kg bw by gavage on day 15 of gestation, and maternal and pup development and viability were followed until postnatal day 63. Some male pups were retained for assessment of reproductive organ weights and sperm counts at 15 months of age. Both pup survival and body-weight gain by the end of lactation were significantly decreased (15–17%) in the group at the highest dose, and eye opening was accelerated by 1 day at all doses in comparison with controls. On postnatal days 49 and 63, most of the developmental affects in male offspring occurred at the highest dose and included decreased numbers of epididymal sperm (postnatal day 49 and 63) and decreased weights of the cauda epididyma (postnatal day 63), ventral prostate and seminal vesicle (postnatal day 49). The onset of puberty was delayed by about 2 and 4 days at the two higher doses, respectively. In males at 15 months of age, significant reductions in the weight of the glans penis (8%) and the numbers of epididymal and cauda epididymal sperm (11–17%) were seen at the two higher doses. The number of ejaculated sperm was decreased (47%) only at the highest dose. However, when the results were combined with those of a previous study (Gray et al., 1995), animals at both 0.05 and 0.2 mg/kg bw also showed a significant decrease (25–26%) (Gray et al., 1997a).

These effects of low doses of dioxin and coplanar chemicals on the developing male reproductive system are usually not paralleled by decreased reproductive capacity or androgen status. Also, in general, the effects of TCDD and related coplanar chemicals on the urogenital tract during fetal and neonatal development of rodents, although dependent on the timing and age of the fetus, can be achieved by low doses in utero alone.

Not only male reproductive tract development but also various aspects of male sexual behaviour can be affected by perinatal exposure to low doses of TCDD and coplanar PCBs. In the experiments by Mably et al., assessment of the sexual behaviour of male Holtzman rats on postnatal days 60, 75 and 115 showed increased latency for intromission, mounting and ejaculation in offspring of dams given doses of TCDD as low as 0.064 and 0.16 µg/kg bw, respectively. Similarly, in Long Evans rats (Gray et al., 1995), partial demasculinization of male sexual behaviour, as determined by an increased total number of mounts, mounts with intromissions before ejaculation and latency before ejaculation, was observed. Increased numbers of mounts with intromissions before ejaculation were also observed in male Wistar rats after maternal exposure to 3,3´,4,4´,5-PeCB at a dose of 10 µg/kg bw (Faqi et al., 1998). While demasculinization of sexual behaviour was reproducible in three strains of rat, it was not seen in hamster offspring. Feminization of male sexual behaviour, as assessed by increased intensity and duration of lordosis, was also observed in male Holtzman rats at maternal doses of TCDD as low as 0.16 µg/kg bw (Mably et al., 1992c) but not in Long Evans rats. Alteration of defeminization of sexual behaviour by TCDD has been shown to require exposure during lactation or after parturition.

In rats and hamsters, single maternal doses as low as 0.2 and 2 µg/kg bw on days 15 and 11.5 of gestation, respectively, also increased the frequency of female pups with cleft phallus and vaginal threads, the latter condition being specific to coplanar chemicals (Gray et al., 1997b; Wolf et al., 1999). However, in ICR mice, a single maternal dose of TCDD at < 60 µg/kg on day 14 of gestation was not associated with genital malformations in female offspring (Theobald & Peterson, 1997).

Female offspring from the study of Gray et al. (1997a) were isolated after weaning and followed for sexual development until postnatal day 70. Additional females were mated to control males on postnatal day 100 for assessment of fertility. On postnatal day 70, the age of onset of vaginal opening was delayed in pups of dams at 0.8 µg/kg bw by approximately 2 days when compared with controls, and pups of dams at 0.2 and 0.8 µg/kg bw had decreased distance between urethral and vaginal openings (by 40% at the highest dose). After mating on postnatal day 100, no significant differences were seen in treated animals with respect to fertility, but the time to pregnancy of offspring of dams at the highest dose was increased by three estrous cycles when compared with controls (4.1 days and 14.4 days, respectively). Significantly increased numbers of female offspring at 0.2 and 0.8 µg/kg bw had urogenital anomalies (vaginal thread, cleft phallus). For example, 92% of females at the highest dose and 2.5% of controls had permanent vaginal threads. A cross-fostering experiment with the group at 1 µg/kg bw indicated that exposure in utero was necessary for the induction of vaginal and external genitalia anomalies. Cleft phallus and changes in the urethral opening position have also been observed in rats given other estrogen-like chemicals (diethylstilbestrol, RU 2858, estradiol) (Gray et al., 1997b).

An increase in the severity of urogenital malformations was seen in groups of eight pregnant Long Evans rats given TCDD at 1 µg/kg bw on day 15 of gestation as compared with day 8 of gestation. After parturition, the litters were standardized to four pups of each sex, and the female offspring were monitored for sexual development, estrous cycle and fertility. Although no maternal toxicity or effect on litter size were noted, the body weights of pups were significantly reduced (average, 17%) from postnatal day 3 throughout the study in the group treated on day 15 of gestation but only on postnatal day 3 for the group treated on day 8; pup survival was decreased by 11% in the group treated on day 15. Both treatments resulted in a variety of malformations in the external genitalia of female offspring, including complete or partial clefting of the phallus and the presence of a vaginal thread; the severity and/or frequency were usually greater in offspring treated on day 15 of gestation. For example, a vaginal thread was present in 79% of progeny treated on day 15 and 14% (nonsignificant) of those treated on day 8. When estrus was assessed throughout the reproductive lifespan, from postnatal day 125, more offspring treated on day 8 (47%) were in constant estrus at about 1 year of age than either controls (16%) or those treated on day 15 (16%). This effect was thought to be related to the decreased fertility rate of female progeny treated on day 8 when they were entered into a continuous breeding phase (five successive litters). Whereas the fertility rates of females treated on day 15 were similar to those of controls up to the fourth litter (but significantly reduced by the fifth litter), the fertility rate of females treated on day 8 was reduced by the second litter and throughout the duration of the breeding phase. Although overall fertility and fecundity in the first litter produced by females treated on day 15 were not affected, the high prevalence of vaginal threads caused greater difficulty in mating with control males, as illustrated by the increased number of mounts without intromission and increased ejaculation latency. Similar genital malformations (83% of offspring with vaginal threads) and suppressed body weight were seen in female offspring of Holtzman rats treated with TCDD at 1 µg/kg bw on day 15 of gestation, except that neonatal survival was decreased by 50% by postnatal day 16. Owing to the low pup survival, reproduction was not assessed in Holtzman rats, although the ovarian weights were reduced by about 47% in comparison with controls at necropsy (Gray & Ostby, 1995).

Pregnant Syrian hamsters, about 4 months old, were given a single dose of TCDD at 0 (n = 9) or 2 µg/kg bw (n = 10) on day 11.5 of gestation by gavage. The dose reflected the higher LD₅₀ seen with this species than with rats (1200 µg/kg bw and 20 µg/kg bw, respectively). After parturition and lactation, the female offspring were monitored for various developmental landmarks until maturity (postnatal day 152–166). Although TCDD treatment had no effect on the body-weight gain or litter size of dams, pup body weights were reduced by up to 30% by the end of lactation, and the reduction persisted until postnatal day 140. Eye opening was accelerated by about 1 day in the pups treated with TCDD, while vaginal opening was delayed (100% of control pups by postnatal day 13 and 100% of TCDD-treated pups by postnatal day 19). After mating of the female offspring on postnatal day 152–166, a variety of parameters, including fertility, weaning index, number of implants, litter size and pup survival, were significantly reduced in the treated animals. TCDD treatment also increased the frequency of F₁ female offspring with mild hypospadias or cleft phallus (0.78% in controls, 92% with TCDD) (Wolf et al., 1999).

Pregnant ICR mice (CD-derived) were treated by gavage on day 14 of gestation with TCDD at a single oral dose of 0, 15, 30 or 60 µg/kg bw (n = 17, 16, 19 and 14, respectively). The litters were standardized on postnatal day 1 to four of each sex, and development was assessed after interim sacrifices on postnatal days 23 (males only), 44, 65, 113–114, 128 (males only) and 142 (females only). Duration of gestation, prenatal mortality, litter size and dam weight were not affected by TCDD, and no effect was seen on postnatal pup mortality or pup weight gain throughout the experiment. Assessment of male reproductive organs during development indicated no effects on anogenital distance, onset of puberty, serum testosterone concentration, growth of the testis, epididymis, dorsal prostate or seminal vesicle, while the weights of the ventral prostate and coagulating gland (anterior prostate) were decreased at most doses from postnatal day 44 and continuing throughout the assessment (36% and 27% decrease at the highest dose, respectively, on postnatal day 114 or 128). While the epididymal sperm counts of mice at 30 and 60 µg/kg bw were reduced by about 30% on postnatal day 65, no effect was seen on postnatal day 114 or 128, and there were no effects on daily sperm production. The weight of the pituitary gland was reduced in male offspring only at all doses but not in a dose-dependent manner (maximum reduction, 42–52% at the lowest dose). TCDD also accelerated the average time to eye opening by about 1 day and decreased thymus weights (not dose-dependent) on postnatal day 44 in male but not female offspring. No urogenital tract malformations were seen in female offspring, and the time to vaginal opening was not affected. The uterine weights of mice both in and out of estrus were significantly reduced by about 34% at the highest dose only. Ovarian weights were not affected at any time (Theobald & Peterson, 1997).

In pregnant Long Evans rats, single doses of TCDD at 0.05–1 µg/kg bw per day on day 15 of gestation, which are associated with a variety of genital malformations in male and female pups, resulted in fetal concentrations on day 16 that were 0.51–0.17% of the maternal dose (13 pg/kg at 0.2 µg/kg dose) (Hurst et al., 2000b). Similar fetal concentrations of TCDD have been reported after exposure of adult animals to 10 ng/kg bw per day for 90 days before breeding. For comparison, the single dose of 0.05 µg/kg on day 15 of gestation resulted in a maternal body burden on day 21 of 27 ng/kg bw (Hurst et al., 1998a).

In a preliminary study, pregnant Wistar rats were treated on day 15 of gestation with a single oral dose of 3,3´,4,4´-TCB at 100 µg/kg bw (n = 17), 3,3´,4,4´,5-PeCB at 10 µg/kg bw (n = 23) or solvent (n = 20), and male pups were assessed for developmental end-points during lactation, on postnatal day 65 (puberty) and on postnatal day 140 (adult). At birth, the offspring of females given 3,3´,4,4´-TCB were slightly heavier than controls, the weight difference increasing to about 15% by the end of lactation; however, by postnatal days 65 and 140, the male pups were 8–12% lighter than controls. Whereas pups of dams given 3,3´,4,4´,5-PeCB weighed slightly less than controls at birth, no further significant effects on body weight were observed during the remainder of the experiment. 3,3´,4,4´-TCB treatment also induced slight increases in the weights of pup testis and epididymis and in daily sperm production, but the weight of the seminal vesicle and serum testosterone concentration (only on postnatal day 140) were reduced. The effects observed in pups of dams given 3,3´,4,4´,5-PeCB included a reduced ratio of anogenital distance:body length (postnatal day 22), a delay of 8 days in vaginal opening, slight reductions in ventral prostate weights and an increased number of mounts with intromission when compared with either controls or dams given 3,3´,4,4´-TCB. The serum testosterone concentrations on postnatal day 140 were also reduced in male pups of dams given 3,3´,4,4´,5-PeCB by up to 64%, but reproductive performance was not affected in either group treated with PCBs (Faqi et al., 1998).

In a study designed to determine the most sensitive effect of TCDD on the rodent male reproductive system, groups of six pregnant Holtzman rats were given a single oral dose of TCDD at 0, 12.5, 50, 200 or 800 ng/kg bw on day 15 of gestation, and the male offspring were assessed for reproductive development on postnatal days 49 and 120. TCDD had no effect on the body-weight gain of dams during gestation or on the weight of male pups during the assessment period. Maternal treatment with TCDD also had no effect on the relative testis or epididymal weights, daily sperm production, cauda epididymal sperm reserve or serum luteinizing hormone, follicle-stimulating hormone or testosterone concentrations of male offspring. At postnatal day 120, the maternal dose of 800 ng/kg bw resulted in TCDD concentrations of 24 pg/g of adipose tissue and 0.49 pg/g of testis in male offspring. Although the anogenital distance was not affected on postnatal day 49, it was significantly reduced at doses > 50 ng/kg bw on postnatal day 120 (maximum decrease, about 18%). In addition, on postnatal day 120 the weight of the ventral prostate was significantly reduced in pups of dams at 200 and 800 ng/kg bw, by 27% and 39%, respectively. Semi-quantitative reverse transcription polymerase chain analysis indicated that the levels of androgen receptor mRNA in this organ were significantly reduced in all treated groups on postnatal day 49 but not on postnatal day 120. Administration of TCDD at any dose resulted in a dose-dependent increase in 5alpha-reductase type 2 mRNA and a decrease in androgen receptor mRNA in the ventral prostate of rats killed on day 49 but not in those killed on day 120, with no adverse sequelae at the lowest dose of 12.5 ng/kg bw. Although the authors postulated that the decreased size of the ventral prostate was due to the decrease in androgen receptor mRNA, previous results (Gray et al., 1995) with Long Evans rats indicated that maternal exposure to TCDD at doses up to 0.8 µg/kg bw did not affect the number of androgen receptors in the ventral prostate of male offspring (Ohsako et al., 2001).

2.2.6 Special studies

There is currently limited experimental evidence for an Ah receptor-based mechanism of action for dioxin-induced neurotoxicity. Central nervous system functioning and neurobehaviour have been assessed in some animal models.

Pregnant Wistar rats were treated by subcutaneous injection with a loading and maintenance dosing regime designed to approximate human perinatal exposure. After an initial dose of TCDD at 0, 1 or 0.3 µg/kg bw on day 19 of gestation, the animals were given a weekly maintenance dose of 0, 0.4 or 0.12 µg/kg bw throughout lactation, and the weaned pups were given the same doses until postnatal week 11. Offspring of dams at the highest dose had significantly reduced body weight (by about 15%) throughout lactation, but not after postnatal day 31. After a slight initial decrease in body weight on postnatal day 7, the body weights of pups of dams on the low-dose regimen were comparable to those of controls. Greater percentages of the TCDD-treated offspring than controls acquired a righting reflex on postnatal day 5 or 6, but they showed a delayed ability to remain on a rotating rod. The activity of 3-month-old female offspring was significantly reduced after treatment with TCDD in comparison with controls after an amphetamine challenge. The authors concluded that TCDD may induce neurobehavioural changes (Thiel et al., 1994).

Groups of three viable rhesus offspring, born after groups of eight maternal animals had ingested a diet containing TCDD at 5 ng/kg bw for 16.2 months, were tested for aspects of peer-group interactive behaviour at 8.6 months of age. Monkeys treated with TCDD showed an increased overall level of behavioural arousal, including increased self-directed behaviour, when compared with controls. No significant relationship was observed between any of the behavioural parameters and the concentration of TCDD in the body fat of the infants. Similar effects were not seen when a second cohort of offspring obtained after the maternal animals had been on control diet for 16 months were tested. At that time, the mean concentration of TCDD in the adipose tissue of the offspring was about 50% less (188 pg/g) than that of the first cohort (Schantz et al., 1992).

A further group of infant rhesus monkeys (one female, four males) was obtained after the maternal animals had been on the TCDD-containing diet for 36.3 months. The total TCDD consumption of the monkeys was estimated to have been 880 ng, and the concentrations of TCDD in mesenteric fat obtained when the offspring were 5 months of age were similar to that in the first cohort (320 pg/g). Cognitive testing of the monkeys in both groups was conducted when they were about 14 months of age. Monkeys exposed perinatally to TCDD took a significantly longer time to learn the first reversal of a shape discrimination–reversal problem than controls (47 versus 27 trials). No significant effect of TCDD was seen for subsequent reversals (up to eight). In addition, no significant effect was seen on learning of spatial or colour reversal problems or in the ability to respond to a delayed spatial alternation test. The performance of offspring on a shape discrimination–reversal problem was not correlated to the concentration of TCDD in their adipose tissue. The authors reported that similar cognitive effects were observed in rhesus infants that had been exposed during development to lead (Schantz & Bowman, 1989).

TCDD-induced immunosuppression has been observed in several experimental animal species, including rodents, guinea-pigs, rabbits and non-human primates (reviewed by Kerkvliet & Burleson, 1994; Holsapple, 1995). Several tissues and cells of the immune system are targeted by TCDD (Kerkvliet & Burleson, 1994).

TCDD-induced thymic atrophy has been observed after perinatal exposure in all species examined (Holladay et al., 1991; Blaylock et al., 1992; Gehrs & Smialowicz, 1997; Gehrs et al., 1997; Gehrs & Smialowicz, 1999). TCDD-induced thymic atrophy is probably of less clinical significance in adult animals, as no correlation has been established between effects on the thymus and functional immune suppression. The reported acute ED₅₀ values for thymic atrophy in adult animals vary considerably among species: 26 µg/kg bw in Sprague-Dawley rats, 0.8 µg/kg bw in Hartley guinea-pigs, 280 µg/kg bw in C57Bl/6 mice and 48 µg/kg bw in Syrian hamsters (Hanberg et al., 1989). Several hypotheses have been proposed to explain the mechanism by which TCDD induces thymic atrophy. Apoptosis induced by TCDD and changes in molecules encoded by the major histocompatibiity complex (MHC) may contribute to the thymic atrophy seen after exposure to TCDD (McConkey et al., 1988; Gerschenson & Rotello, 1992). Rhile et al. (1996) investigated the role of Fas (CD95), an important molecule involved in the induction of apoptosis, on TCDD-mediated immunotoxicity in mice bearing a homozygous lpr mutation, which leads to failure of expression of Fas. TCDD administered orally at 0.0, 0.1, 1 or 5 µg/kg bw for 11 days was less toxic to thymocytes from C57Bl/6 lp/lpr mice (Ah-responsive, Fas^–) than to C57Bl/6 ^+/+ mice (Ah-responsive, Fas⁺). Similar results were obtained when T-cell responsiveness was tested with an antigenic challenge with conalbumin, suggesting that TCDD acts at a time when T cells are in the process of differentiating in response to antigenic challenge, as TCDD had no effect on naïve or resting T cells that had not been challenged with an antigenic stimulus. When mice that differed only at the MHC locus were compared for immunotoxic effects of TCDD, it was observed that B10.D2 (Ah-responsive, H-2d) mice were more sensitive to TCDD-induced thymic atrophy and altered peripheral T-cell function than B10 mice (Ah-responsive, H-2b). Thymic atrophy in all TCDD-sensitive strains was accompanied by depletion in T cell subsets (CD4⁺, CD4⁺ CD8⁺, CD4^– CD8^– and CD8⁺). On the basis of these results, the authors suggested that the observed immunotoxic effects of TCDD may be mediated partly through the presence of the Fas molecule in activated T cells and that MHC phenotype may also play an important role (Rhile et al. 1996).

TCDD affects both cellular and humoral immunity. Its effects on T-cell function are characterized by changes in several end-points, including delayed-type hypersensitivity (DTH) responses, rejection of skin allografts, generation of cytotoxic T lymphocytes and lymphoproliferative responses of lymphocytes to mitogens and specific antigens in vitro. The effects on T-cell function occur at concentrations three orders of magnitude lower than those that affect thymus cellularity. Species differ in their sensitivity to TCDD, guinea-pigs being more sensitive than rodents. For example, the DTH response to tuberculin was decreased in guinea-pigs after eight weekly doses of 40 ng/kg bw TCDD, whereas in rats the DTH response to tuberculin was unaffected by six weekly doses of 5 µg/kg bw (Vos et al., 1973).

Differences in DTH response have also been observed in the same species challenged with different antigens. In C57Bl/6 mice given TCDD as four weekly intraperitoneal doses of 0, 0.4, 4 or 40 µg/kg bw, the DTH response to sheep red blood cells (SRBC), footpad swelling, and to oxazolone, ear thickness, was measured. TCDD at doses of 4 and 40 µg/kg bw suppressed the DTH response to oxazolone, whereas the DTH response to SRBC was not significantly affected by any of the doses tested (Clark et al., 1981).

TCDD-induced effects on the cytotoxic T lymphocyte response of spleen cells have been found less consistently. Clark et al. (1981) treated C57Bl/6 mice with TCDD as four weekly intraperitoneal doses of 0, 0.4, 4 or 40 µg/kg bw and examined the generation of cytotoxic T lymphocytes. Generation was inhibited at all doses. Mechanistic studies by the same group indicated that generation of allospecific cytotoxic T cells in C57Bl/6 mice was significantly impaired after exposure to doses as low as 4 ng/kg bw (Clark et al., 1981). On the basis of mechanistic studies in vitro, the authors concluded that the cellular basis for suppression of the cytotoxic T lymphocyte response was induction of T-suppressor cells in the thymus which act selectively against the cytotoxic T lymphocyte response (Clark et al., 1983). Dooley et al. (1990) found, however, that the number of T-suppressor cells in splenocytes of B6C3F₁ mice treated with TCDD by gavage at 1 µg/kg bw for 5 days was not increased. The mechanism by which TCDD affects the cytotoxic T lymphocyte response therefore remains to be elucidated.

Immunosuppression, indicated by suppression of the anti-SRBC humoral immune response, has been observed by several investigators (reviewed by Kerkvliet & Burleson, 1994). The dose–response relationships for the immunosuppressive effects of TCDD, 2,3,4,7,8-PeCDF, 1,2,3,7,9-PeCDF, 2,3,7,8-TCDF and 1.3.6.8-TCDF on the splenic plaque-forming cell response to SRBC in C57Bl/6 mice, measured as the ED₅₀, were 2.4 nmol/kg bw for TCDD, 3 nmol/kg bw for 2,3,4,7,8-PeCDF, 14 nmol/kg bw for 2,3,7,8-TCDF, 710 nmol/kg bw for 1,2,3,7,9-PeCDF and 36 000 nmol/kg bw for 1,3,6,8-TCDF (Davis & Safe, 1988). Mice of different strains had different susceptibility to aryl hydrocarbon hydroxylase induction and different levels and/or affinity for a specific cytosolic binding protein; they also differed in their response to SRBC. Specifically, TCDD at a dose as low as 1.2 µg/kg bw significantly inhibited the plaque-forming cell response to SRBC in C57Bl/6 and C3H/HeN mice, which are representative of more susceptible strains, while a dose of at least 6 µg/kg bw was required for partial suppression in DBA/2 and AKR strains. Exposure for up to 8 weeks did not increase the sensitivity of DBA/2 mice to treatment (Vecchi et al., 1983). The plaque-forming cell and immunoglobulin M responses of C57Bl/6 and DBA/2 mice to the T cell-independent antigen trinitrophenyl-lipopolysaccharide also varied with different congeners, with ED₅₀ values in C57Bl/6 mice of 2.8 and 1.6 µg/kg bw for TCDD, 11 and 14 µg/kg bw for 3,3´,4,4´,5-PeCB and 25 and 20 µg/kg bw for 3,3´,4,4´,5,5´-HxCB, and values in the less Ah-responsive DBA/2 mice for the same congeners of 8.5 and 10 µg/kg bw, 61 and 69 µg/kg bw and 73 and 71 µg/kg bw, respectively (Harper et al., 1994).

Differences among species in the primary response to antigens have also been observed. B6C3F₁ mice and Fischer 344 rats were given a single intraperitoneal injection of TCDD in corn oil at a dose of 1, 3, 10 or 30 µg/kg bw on 7 days before immunization with SRBC for induction of a humoral response. The plaque-forming cell response in mice was significantly suppressed, with an ED₅₀ value of 0.7 µg/kg bw, while the response in rats was unaffected at doses up to 30 µg/kg bw. The apparent inability of TCDD to suppress humoral immunity in rats was unrelated to induction of hepatic CYP 1A1 and CYP 1A2. Furthermore, in rats but not in mice, the splenic CD4^– CD8⁺ T-cell sub-population was reduced in a dose-related manner, and this reduction was accompanied by a dose-related increase in immunoglobulin M⁺ splenocytes (Smialowicz et al., 1994). In a similar experiment with the same doses of TCDD, the response of B6C3F₁ mice to the T cell-independent, B cell-dependent antigen trinitrophenyl-lipopolysaccharide was reduced at 10 and 30 µg/kg bw dose, and a similar reduction was seen in rats at 30 µg/kg bw, with no significant changes in splenic lymphocyte subsets. The mechanism responsible for the observed differences between the two species in their response to T cell-dependent and T cell-independent antigens remains to be elucidated (Smialowicz et al., 1996).

Differences in the response to various antigens after exposure to TCDD have also been observed within the same species. Antibody responses to three antigens, SRBC, DNP-Ficoll and trinitrophenyl-lipopolysaccharide,were compared in C57Bl/6 mice after treatment with 1,2,3,4,6,7,8-HpCDD. The median inhibitory doses were 53, 130 and 520 µg/kg for the three antigens, respectively, indicating that higher doses of TCDD are required to suppress the T cell-independent, B cell-dependent humoral immune response (Kerkvliet & Brauner, 1987).

The effects of TCDD on host resistance to various infectious agents, including bacteria, viruses and parasites, has been well documented. Four-week-old male C57B1/6J (J67) mice received TCDD by gavage at a dose of 0.5–20 µg/kg bw once a week for 4 weeks and were challenged 2 days after the fourth dose with either Salmonella bern or Herpesvirus suis. Increased mortality in response to S. bern was seen at 1 µg/kg bw and a shorter latency before death at 5 µg/kg bw. None of the doses compromised the ability of mice to combat infection by H. virus suis (Thigpen et al., 1975). Compromised resistance to Salmonella was also shown in mice given feed containing TCDD at 50 or 100 ng/kg for 8 weeks (Hinsdill et al., 1980).

Increased parasitic infestation has been reported after administration of TCDD. A single dose of 5 or 10 µg/kg bw administered by gavage increased the susceptibility of 6–8-week-old female B6C3F₁ mice to Plasmodium yoelii 17 XNL, which causes malaria (Tucker et al., 1986). A dose of 10 or 30 µg/kg bw administered intraperitoneally 7 days before infection of B6C3F₁ mice with Trichinella spiralis resulted in delayed onset of parasite elimination, while a dose of 1 µg/kg bw suppressed the proliferative response of splenocyte and mesenteric lymph node cells stimulated with T. spiralis antigen (Luebke et al., 1994). In contrast, the same doses of TCDD given to adult Fischer 344 rats 7 days before infection with T. spiralis had no effect on the rate of elimination of the parasite or the number of encysted larvae in muscle, and the dose of 30 µg/kg bw enhanced the proliferative response of lymphocytes after stimulation with the parasite antigen (Luebke et al., 1995).

The ability of hosts to combat viral infection was compromised at doses of TCDD lower than those that affected resistance to bacteria and parasites. Mice given an intraperitoneal injection of TCDD at 0.04, 0.4 or 4 µg/kg bw once a week for 4 weeks and challenged 7–22 days later with Herpes simplex type II strain 33 virus had a significantly enhanced death rate from viral infection (Clark et al., 1981). In female B6C3F₁ mice, 6–8 weeks of age, given TCDD at 10, 1 or 0.1 µg/kg bw intraperitoneally and challenged with A/Taiwan/1/64 (H2N2) virus 7–10 days later, the LOEL for decreased resistance to virus was 0.1 µg/kg bw, the lowest dose tested (House et al., 1990).

An even lower LOEL of 0.01 µg/kg bw was seen for enhanced mortality from influenza due to A/Hong Kong/8/68 (H3N2) virus. Eight-week-old female B6C3F₁ mice were given a single dose of TCDD at 0.1, 0.05 or 0.01 µg/kg bw in corn oil by gavage. Seven days later, the mice were infected intranasally with a concentration of Influenza A/Hong Kong/8/68 (H3N2) virus calculated to result in 30% mortality in mice untreated with TCDD, and the animals were observed for 22 days. The mortality rates of mice given TCDD was statistically significantly higher than that of controls, but the rate at doses of 0.001 and 0.005 µg/kg bw was not significantly different from that of controsl. TCDD did not alter the replication or clearance of the virus in these mice (Burleson et al., 1996).

Young animals are reported to be highly sensitive to prenatal or neonatal exposure to TCDD. Pregnant B6 or B6C3F₁ mice were given TCDD oraly at a dose of 1, 2, 5 or 15 µg/kg bw on days –7, 0, +7 and +14 relative to parturition. The LOEL for an increased incidence of PYB6 tumours was 1 µg/kg bw, the lowest dose tested. The LOAEL for increased allograft rejection time was 2 µg/kg bw and that for decreased mortality from L. monocytogenes infection, T-cell blastogenesis, thymus and spleen weights, bone-marrow cellularity and bone-marrow colony forming units was 5 µg/kg bw. The LOEL for lipopolysaccharide blastogenesis and anti-SRBC serum titres was > 15 µg/kg bw (Vos & Moore, 1974; Luster et al., 1980).

Pregnant Swiss Webster mice were fed diets containing TCDD at 1, 2.5 or 5 µg/kg for 7 weeks before and after parturition. The LOEL for increased mortality of offspring due to endotoxin was the lowest dose, that for decreased thymus weight was 2.5 µg/kg, and that for decreased plaque-forming cell response to SRBC and DTH responses was 5 µg/kg. The LOEL for increased mortality due to L. monocytogenes, T and B cell blastogenesis and anti-SRBC serum titres was > 5 µg/kg of diet (Thomas & Hinsdill, 1979).

When pregnant Fischer 344 rats were given TCDD at 1 or 5 µg/kg bw per day orally on days –3, 0, +7 and +14 relative to parturition, the LOEL for increased allograft rejection time, for decreased T cell blastogenesis, DTH response and body and thymus weight and for increased L. monocytogenes-induced mortality was 5 µg/kg bw per day (Vos & Moore, 1974; Faith & Moore, 1977).

A single dose of TCDD at 0.1, 0.3, 1 or 3 µg/kg bw given by gavage to pregnant Fischer 344 rats on day 14 of gestation resulted in suppression of the DTH response to bovine serum albumin in both male and female pups. The suppression persisted for up to 19 months in the male offspring of dams given 3 µg/kg bw. Exposure during both gestation and lactation was required for suppression of the DTH response (Gehrs et al., 1997; Gehrs & Smialowicz, 1999).

Further experiments with TCDD at 0, 0.1, 0.3 or 1 µg/kg bw indicated that doses as low as 0.1 µg/kg bw given to dams on day 14 of gestation could suppress the DTH response in males up to 14 months of age, while a maternal dose of 0.3 µg/kg bw was necessary to cause suppression that continued to 14 months in female offspring. The DTH response to a second antigen, keyhole lympet haemocyanin, was also reduced in the 4-month-old male offspring of dams dosed at 3 µg/kg bw on day 14 (Gehrs & Smialowicz, 1999). Thus, doses of TCDD as low as 0.1 µg/kg bw administered orally to dams on day 14 of gestation caused significant, long-lasting effects on the DTH response of offspring, males being more sensitive to treatment than females.

Groups of six male Wistar rats were maintained on diets containing TCDD at 0, 0.34 or 5 µg/kg for 30 or 180 days. The TCDD-containing diets were given on 5 days/week and the control diet on 2 days/week. The total intake of TCDD was about 3 µg/kg bw or 100 ng/kg bw per day for 30 days or 17 ng/kg bw per day for 180 days. All rats were killed 3 days after the last feeding. No significant effects were seen on body-weight gain or relative spleen or liver weight, while the relative thymus weight was reduced by about 27% in rats treated for 30 days. The mitogenic proliferative response of spleen cells to lipopolysaccharide in vitro was enhanced in the group treated for 30 days to lipopolysaccharide, and the response to concanavalin A stimulation was slightly decreased in the group treated for 180 days, in comparison with controls. Production of IL-1 by spleen macrophages in vitro was significantly decreased in both groups, by 38% in those treated for 30 days and by 52% in those treated for 180 days. Stimulation of IL-2 production by spleen cells was also reduced, but only in the group treated for 180 days. Both groups showed decreased conconavalin A-induced expression of IL-2 receptors by splenic T-cells in vitro (Badesha et al., 1995).

In summary, immunotoxic effects of TCDD have been observed in several species at multiple targets in the immune system. The severity of TCDD-induced immunotoxic effects varies among species and depends largely on the end-point investigated. On the basis of the above review, the LOEL for suppressed DTH responses in offspring of perinatally exposed Fischer 344 rats was 0.1 µg/kg bw, and the LOEL for increased susceptibility of adult B6C3F₁ mice to H2N2 virus was 0.01 µg/kg bw, with a NOEL of 0.005 µg/kg bw.

The effects on the thyroid observed in experimental animals after exposure to dioxins or coplanar chemicals usually involve decreases in free and total T4 in serum, with no compensatory effects on TSH or T3 (Brucker-Davis, 1998). Dose-dependent decreases in plasma T4 concentration were observed in groups of eight female Sprague-Dawley rats given diets supplemented with TCDD at a concentration of 0, 0.2, 0.4, 0.7, 5 or 20 µg/kg for 13 weeks, with estimated daily intakes of 0, 14, 26, 47, 320 and 1000 ng/kg bw per day. The concentration of total T4 in plasma was significantly reduced at doses > 47 ng/kg bw per day (van Birgelen et al., 1995b). Similar effects were observed in groups of eight female rats of the same strain given diets containing 3,3´,4,4´,5-PeCB at a concentration of 0, 7, 50 or 180 µg/kg for 13 weeks, providing daily intakes of 0, 0.47, 3.2 and 10 µg/kg bw per day. The lowest dose of 3,3´,4,4´,5-PeCB that induced a significant decrease in free or total T4 in plasma was 3.2 µg/kg bw per day. The LOEL was 0.047 µg/kg bw per day (van Birgelen et al., 1995a).

Thyroid hormone status was assessed in rats from a short-term assay for tumour promotion. After initiation with N-nitrosodiethylamine at 70 days of age, female Sprague-Dawley rats were treated by gavage every 2 weeks for 30 weeks with TCDD at doses designed to deliver 0, 0.1, 0.35, 1, 3.5, 11, 36 or 120 ng/kg bw per day. The rats were necropsied 1 week after the last dose of TCDD, and serum samples were analysed for T3, T4 and TSH. The concentration of T4 was significantly reduced at doses of TCDD > 11 ng/kg bw per day in the initiated rats and > 36 ng/kg bw per day in the uninitiated rats; the maximum reduction was by 42%. While there was no effect on T3 concentration, that of TSH in serum was increased approximately 2.5-fold in uninitiated rats at the highest dose when compared with controls; 3.3 ng/ml and 1.3 ng/ml, respectively. Histological changes in the thyroid, including diffuse follicular hyperplasia, were also seen in TCDD-treated rats, and, at doses > 3.5 ng/kg bw per day, the ratio of parenchymal to follicular area was significantly increased. Hepatic CYP 1A1 and UGT1 mRNA levels were increased at doses > 0.35 ng/kg bw per day and 3.5 ng/kg bw per day, respectively (Sewall et al., 1995).

Whereas other related polyhalogenated aromatic hydrocarbons, including some coplanar PCBs, can interact directly with the thyroid gland and/or thyroid hormone transport mechanisms, dioxins and furans appear to function primarily through hormone metabolism (Brouwer et al., 1998). Significant positive correlations have been observed between TCDD-induced decreases in plasma T4 concentration and concomitant induction of hepatic UGT1, indicating that TCDD functions predominantly at an extrathyroidal level (Schuur et al., 1997). Further evidence for a mechanism of action that does not directly involve the thyroid gland was obtained in the long-term bioassay in which male and female Sprague-Dawley rats were given TCDD at doses of 0.001–0.1 µg/kg/bw per day for up to 2 years, with no increase in the incidence of non-neoplastic lesions in the thyroid, parathyroid, adrenal or pituitary glands (Kociba et al., 1978b).

On the basis of TEFs (in parentheses), a mixture comprising TCDD (1), 1,2,3,7,8-PeCDD (3.3), 2,3,4,7,8-PeCDF (17), 3,3´,4,4´,5-PeCB (61), 2,3´,4,4´,5-PeCB (12 800) and 2,3,3´,4,4´,5-HxCB (1888) was more potent than TCDD in reducing the total plasma concentration of T4 in groups of 10–12 female Sprague-Dawley rats after subcutaneous injection of a toxic equivalent of 1 µg/kg bw per week (van der Plas et al., 2001). TCDD at 1 µg/kg bw per week reduced the total plasma T4 concentration by 38%, while a toxic equivalent of the mixture reduced it by 54%. Indications of reduced T4 protein binding (decreased ratio of total:free T4) suggested competitive displacement of T4 from the major rodent thyroid hormone transport protein, transthyretin, possibly due to the presence of hydroxylated PCB metabolites from the mixture. Various hydroxylated metabolites of PCBs have been shown to competitively displace T4 binding to human transthyretin but not to T4-binding globulin (Lans et al., 1994).

Groups of six or seven male Sprague-Dawley rats treated by gavage with TCDD at a dose of 0, 0.02, 0.23, 1.2, 3.5, 7 or 12 µg/kg bw per week for 10 weeks showed dose-dependent decreases in total serum T4 at doses > 0.23 µg/kg bw per week. Four of seven rats at the highest dose died, and decreased body-weight gain was seen at the three higher doses. After a 6-week recovery period, at which time it was estimated that about 75% of the administered TCDD would have been eliminated, the total T4 concentration in serum continued to be reduced in groups with a cumulative total dose of TCDD > 35 µg/kg bw (Li & Rozman, 1995).

Groups of 10–14 pregnant Sprague-Dawley rats were given 3,3´,4,4´-TCB at 2 or 8 mg/kg bw per day, 3,3´,4,4´,5-PeCB at 0.25 or 1 µg/kg bw per day, TCDD at 0.025 or 0.1 µg/kg bw per day or solvent orally on days 10–16 of gestation. Treatment had no effect on the weight gain of dams during gestation, on litter size or on pup survival or growth during lactation. When selected pups were necropsied on postnatal day 21, the absolute thymus weights were decreased only at the higher doses of 3,3´,4,4´-TCB and TCDD, by 15% and 23%, respectively, while the liver weights were significantly increased at both doses of 3,3´,4,4´,5-PeCB, by 13% and 17%, respectively. Perinatal treatment had no effect on the thyroid hormone status (T3, T4, TSH) of male pups, and only the serum T4 concentration was decreased in female pups, by 15–20% at the higher doses of 3,3´,4,4´-TCB and TCDD. Hepatic UGT1 and EROD activity were increased by all treatments except the lower dose of 3,3´,4,4´-TCB; UGT1 activity was increased by up to threefold and EROD activity by up to 24-fold with the higher doses of 3,3´,4,4´,5-PeCB and TCDD. No sex-specific differences were seen with respect to enzyme induction (Seo et al., 1995).

Characteristic symptoms of vitamin A deficiency have been observed in experimental animals and wildlife species after exposure to TCDD and various coplanar chemicals. In rats given single oral doses of TCDD at 1–10 µg/kg bw, the hepatic stores of retinol were reduced and those in serum, kidney, urine and faeces were increased (excretion). The mechanism of action appears to be persistent impairment of the ability of liver stellate cells to store vitamin A due to inhibition of lecithin:retinol acyltransferase. Treatment of male Sprague-Dawley rats with a single oral dose of 10 µg/kg bw resulted in a significant decrease in the activity of this enzyme in the non-parenchymal liver cell fraction 7 days later (Nelson et al., 1996). The initial kinetic effects after exposure to TCDD involve increased mobilization of vitamin A from hepatic storage sites, probably by retinal ester hydrolysis (Kelley et al., 2000).

Guinea-pigs, rats, Ah-responsive and Ah-unresponsive strains of mice (C57Bl/6 and D/A/2, respectively) and hamsters were given a single intraperitoneal dose of TCDD representing 9–67% of their respective LD₅₀ value (0.5–400 µg/kg bw) and killed at various times up to 112 days after treatment. Hepatic and pulmonary vitamin A stores were decreased in all species except C57Bl/6 mice (liver only), the degree of change being related to obvious signs of toxicity (lethality, decreased growth rate, liver hypertrophy, thymic involution) (Hawkinston et al., 1979).

Male Sprague-Dawley rats were given a single oral dose of TCDD at 0, 0.1, 1, 10, 30, 100 or 300 nmol/kg bw (approximately 0, 0.032, 0.32, 3.2, 9.6, 32 and 96 µg/kg bw), and three rats at each dose were killed 12 days later for analysis of retinol, retinal palmitate, retinal palmitate hydrolase and acyl coenzyme A:retinol acyltrans-ferase activity in kidney and liver. Additional control groups consisted of rats maintained on a vitamin A-free diet for 12 or 26 days. The relative liver weights were significantly increased in a dose-dependent manner in rats at doses > 0.32 µg/kg bw, except those at the highest dose, which had an approximately 50% loss of body weight. The relative kidney weights were significantly increased only at the lowest dose. Rats fed the vitamin A-free diet for 12 days had increased relative liver weight similar to that observed in rats given the three lower doses of TCDD (10–22%). The concentrations of hepatic retinal palmitate were decreased at all doses of TCDD when compared with controls, with a maximum reduction of 75% in the group at the highest dose. The concentrations of hepatic retinal palmitate levels were also reduced in the rats fed the vitamin A-free diet for 12 days to an extent similar to that of rats given the three lower doses of TCDD (about 40%). The concentrations of retinol and retinal palmitate in kidney were increased at doses of TCDD of 3.2, 9.6 and 32 µg/kg bw and in the rats fed the vitamin A-free diet for 26 days. The activity of acyl coenzyme A:retinol acyltransferase in kidney was increased only at 9.6 and 32 µg/kg bw and in rats fed the vitamin A-free diet for 26 days (Jurek et al., 1990).

Groups of six male and six female Sprague-Dawley rats were fed diets containing various concentrations of dioxins and furans for 13 weeks, and hepatic retinol concentrations were determined at sacrifice. The estimated median effective dietary concentrations of TCDD for reducing hepatic retinol were 0.5 µg/kg of diet for females and 0.8 µg/kg of diet for males, equivalent to 50 and 80 ng/kg bw per day, respectively, while the values for the next most potent congener, 1,2,3,7,8-PeCDD, were 0.9 µg/kg and 1.2 µg/kg of diet, respectively. The ranking of potency was TCDD > 1,2,3,7,8-PeCDD > 2,3,4,7,8-PeCDF > 1,2,3,6,7,8-HxCDF > 1,2,3,7,8-PeCDF > OCDF > OCDD > 1,2,3,4,8-PeCDF. The estimated relative potency factors for hepatic vitamin A depletion corresponded to the values for indices of toxicity in short-term tests (thymic atrophy, body-weight reduction, liver lesions). Dietary intake of a variety of chlorinated dioxins and furans for 13 weeks thus reduced hepatic vitamin A concentrations in a dose-dependent manner (Fattore et al., 2000).

As retinol is transported in blood bound to a complex of retinol binding protein and transthyretin, destabilization of this complex could reduce the binding ability of retinol, with subsequent elimination by renal glomerular filtration. In the study of van der Plas et al. (2001), a mixture of coplanar PCBs caused a decrease in hepatic retinal palmitate concentration similar to that induced by an equivalent dose of TCDD alone. However, only the mixture reduced plasma retinol concentrations, presumably by the generation of hydroxylated PCB metabolites and effects on retinol binding protein–transthyretin binding.

When groups of six male weanling Sprague-Dawley rats were given a single intraperitoneal injection of 3,3´,4,4´-TCB or 3,3´,4,4´,5-PeCB at a dose of 150 µmol/kg and killed 7 days later, significant decreases in the concentrations of hepatic retinyl palmitate and serum retinol were observed. Both PCBs significantly increased the relative liver weights and decreased the relative thymus weights, while 3,3´,4,4´,5-PeCB induced a 2.7-fold increase in hepatic lipid accumulation, when compared with controls. 3,3´,4,4´,5-PeCB-treated rats also lost approximately 20 g of body weight. 3,3´,4,4´-TCB treatment caused a significant increase (approximately 10-fold) in the renal stores of retinyl palmitate over that of controls, and both PCBs decreased the serum retinol binding protein concentration (Chen et al., 1992a). Various hydroxylated metabolites of coplanar PCBs have been shown to bind with high affinity to transthyretin, resulting in inhibition of T4 binding to this protein and disruption of the transthyretin–retinol binding protein complex (Brouwer et al., 1998).

The epidemiological studies on dioxins include studies of the exposure of workers producing phenoxy herbicide and chlorophenols, studies of the population exposed in the industrial accident in Seveso, Italy, studies of persons exposed during application of herbicides and particularly cohorts of personnel in the US Air Force in Viet Nam, commercial applicators and case–control studies of exposure of communities.

The most informative epidemiological studies are those of the population of Seveso who were accidentally exposed to dioxins in 1976, those of workers producing chlorophenols and chlorophenoxy herbicides contaminated with dioxins and US Air Force applicators. Considerable effort has been made to characterize the exposure of these populations to TCDD, which was 10–1000 times higher than that of the general population. The populations included in these studies are shown in Table 5.

2.3.1 Cancer

This section is based largely on a review of the carcinogenicity of dioxins by a working group convened by the IARC (1997) and a WHO evaluation for a tolerable daily intake conducted in 1998 (Kogevinas, 2000).

The incidence of and mortality from cancer were investigated in the population of Seveso that was exposed during the industrial accident in 1976. The contaminated area was subdivided into three zones (zone A, zone B and zone R), according to the average concentration of TCDD measured in soil samples. The concentrations in serum in 1976 in 19 non-randomly selected persons from zone A ranged from 830 pg/g to 56 000 pg/g, and the geometric mean serum concentrations measured 20 years after the accident (Landi et al., 1997) were 53 pg/g in seven randomly selected residents of zone A (the most heavily contaminated zone), 11 pg/g in 55 persons in zone B (the second most heavily contaminated zone) and 4.9 pg/g in 59 persons in the uncontaminated area (zone non-ABR). The populations of 11 municipalities surrounding the contaminated area were used as a reference (zone R).

The mortality rate during 1976–91 and cancer incidence rates for the period 1977–86 have been reported (Bertazzi et al., 1989, 1993, 1997), and follow-up of the exposed population for deaths was extended to 1996 (Bertazzi et al., 2001). The rate of death from cancer in general was not increased over the period of observation. However, 15 years after the accident, the mortality rate from cancer was increased in men among the 804 inhabitants of zone A and the 5941 inhabitants of zone B, with a rate ratio of 1.3 and a 95% confidence interval (CI) of 1.0–1.7. Increases were also seen in the rate of death from rectal cancer (rate ratio, 2.4; 95% CI, 1.2–4.6) and lung cancer (rate ratio, 1.3; 95% CI, 1.0–1.7), but with no definite pattern of latency. Lymphohaematopoietic neoplasms occurred at an excess rate in males and females (rate ratio, 1.7; 95% CI, 1.2–2.5). The mortality rate from Hodgkin disease was high during the first 10-year observation period (rate ratio, 4.9; 95% CI, 1.5–16), whereas the greatest increases in the rates of death from non-Hodgkin lymphoma (rate ratio, 2.8; 95% CI, 1.1–7.0) and myeloid leukaemia (rate ratio, 3.8; 95% CI, 1.2–12) were seen after 15 years. No cases of soft-tissue sarcoma were observed in zones A and B, although 0.8 was expected.

The study populations considered overlap in many instances. Figure 8 shows the relationships among the various publications considered relevant to this evaluation.

In an accident at a BASF plant producing trichlorophenol in Ludwigshafen, Germany, in 1953, a total of 247 employees (243 men, 4 women) were identified as having been involved directly or in subsequent clean-up, repair or maintenance activities (Zober et al., 1990; Ott & Zober, 1996). Measurements of TCDD in serum were available for 138 of these persons (Ott et al., 1993). The geometric mean concentration was 15 pg/g (range, < 1–550 pg/g). The cumulative dose of TCDD at the time of exposure was calculated from a model for each person. The mean concentrations were 38 pg/g for a subgroup with no chloracne, 420 pg/g for a subgroup with moderate chloracne and 1000 pg/g for a group with severe chloracne. The cohort was observed for deaths and cancer incidence until 1992, and the rate of mortality from cancer was found to increase with increasing exposure (Figure 9). Mortality rates from cancers of the digestive system and respiratory tract also tended to increase with increasing exposure. Similar results were obtained for cancer incidence. A joint analysis of dose of TCDD and cigarette smoking showed a relationship between the dose and the occurrence of all cancers combined among current cigarette smokers but not among non-smokers.

Becher et al. (1996) reported on the mortality rates among 2479 male workers employed in four German plants involved in the production of phenoxy herbicides and chlorophenols or who were likely to have been in contact with these substances and with their contaminants, PCDDs (often including TCDD) and PCDFs. The study did not include the persons involved in the BASF accident. It included workers from a Boehringer-Ingelheim plant in Hamburg, for whom results have been reported in several publications (Manz et al., 1991; Nagel et al., 1994; Flesch-Janys et al., 1995, 1996). The concentrations of TCDD were measured in the blood of workers in three of the four plants (Kogevinas et al., 1997). The mortality rate was increased in the whole cohort from all cancers combined (138 deaths; standardized mortality ratio [SMR], 1.2; 95% CI, 1.0–1.4), cancer of the oral cavity and pharynx (nine deaths; SMR, 3.0; 95% CI, 1.4–5.6), lung cancer (47 deaths; SMR, 1.4; 95% CI, 1.1–1.9), lymphatic and haematopoietic neoplasms (13 deaths; SMR, 1.7; 95% CI, 0.9–2.9) and non-Hodgkin lymphoma (six deaths; SMR, 3.3; 95% CI, 1.2–7.1).

Several reports have been published on workers at a chemical plant operated by Boehringer-Ingelheim in Hamburg, Germany, that produced herbicides heavily contaminated with TCDD and other PCDDs and PCDFs (Manz et al., 1991; Nagel et al., 1994; Flesch-Janys et al., 1995). An outbreak of chloracne in 1954 led to a halt in the production of trichlorophenol and 2,4,5-trichlorphenoxyacetic acid. In 1957, production was resumed, with a new process that reduced the formation of TCDD. The vital status of all persons who had been permanent employees at the plant for at least 3 months between 1 January 1952 and 31 December 1984 (1184 men, 399 women) was investigated through to 1989 (Manz et al., 1991). The causes of death were ascertained from medical records or, when medical records were not available, from death certificates. The concentrations of TCDD were measured in samples of serum or adipose tissue from workers in various production departments, mainly after the plant had closed in 1984.

The rate of mortality from all cancers combined among men was increased in comparison with the rate for the general population (93 deaths; SMR, 1.2; 95% CI, 1.0–1.5). The increase was greatest for men who had started work at the plant before 1954, who had the heaviest exposure to TCDD on the basis of the history of the plant and subsequent serum measurements, and who had remained employed at the plant for many years. The mortality rates of female workers were further investigated by Nagel et al. (1994). The rate of death from breast cancer was higher than expected (10 cases; SMR, 2.4; 95% CI, 1.1–4.4) and increased with duration of employment.

The mortality rates of 1189 male workers were investigated through 1992 (Flesch-Janys et al., 1995, 1996). The concentrations of various PCDD and PCDF congeners were measured in adipose tissue or whole blood from 190 workers, and the values for each worker at the end of exposure were calculated on the basis of a one-compartment first-order kinetics model. The mean estimated concentration of TCDD for the whole cohort was 140 ng/kg of lipid (median, 38 ng/kg). In some departments, workers had been exposed to other carcinogens, such as dimethyl sulfate and benzene. The mortality rate from all cancers was increased in all categories of exposure to TCDD and showed a significant trend (p = 0.01) with increasing intensity of exposure (Figure 10). No data were reported on deaths from cancers at specific sites.

Mortality rates among workers employed between 1955 and 1986 in the synthesis and formulation of phenoxy herbicides and chlorophenols in The Netherlands were examined (Bueno de Mesquita et al., 1993). In one of the plants, where the main compound produced was 2,4,5-trichlorphenoxyacetic acid and its derivatives, an accident in 1963, which caused a release of PCDDs, including TCDD. In a second factory, (4-chloro-ortho-toloxy)acetic acid, 4-chloro-2-methyl phenoxypropionic acid and 2,4-D were produced. The study involved 2074 men working in manufacture at the two plants (963 exposed to phenoxy herbicides, 1111 not exposed); in addition, 145 workers who had probably been exposed to TCDD during the industrial accident and clean-up were examined. The study was later extended to 1991 and enlarged (2298 persons, including 191 women) (Hooiveld et al., 1998). Exposure was assessed from job histories and on modelled concentrations of TCDD in serum, measured in 1993 in a subset of 31 exposed (mean concentration, 53 ng/kg of lipid; range, 1.9–190 ng/kg) and 16 unexposed (mean concentration, 8 ng/kg of lipid) persons. Fourteen persons who had been exposed during the accident in 1963 had the highest mean concentration (96 ng/kg of lipid; range, 16–190 ng/kg). In this factory, the mortality rates from all causes (139 deaths; SMR, 1.3; 95% CI, 1.1–1.5) and all cancers (51 deaths; SMR, 1.5; 95% CI, 1.1–1.9) were significantly increased. Excess numbers of deaths from cancers of the urinary bladder (four deaths; SMR, 3.7; 95% CI, 1.0–9.5) and kidney (four deaths; SMR, 4.1; 95% CI, 1.1–10) and from non-Hodgkin lymphoma (three deaths; SMR, 3.8; 95% CI, 0.8– 11) were observed. When the workers were subdivided into three categories of low, medium and high exposure according to the serum concentrations of TCDD predicted from a model, the relative risks for death from all causes, all cancers and lung cancer were significantly increased for workers with medium and high exposure and were highest for the group with the heaviest exposure (Figure 11).

A study of workers in 12 plants in the USA was designed by the National Institute for Occupational Safety and Health (Fingerhut et al., 1990, 1991; Steenland et al., 1999). The cohort comprised 5172 men and included most workers in the USA likely to have been exposed to TCDD during manufacture principally of trichlorophenol and 2,4,5-trichlorphenoxyacetic acid. The average concentration of TCDD in serum from 253 workers at two plants in 1987 was 230 ng/kg of lipid, whereas that in 79 unexposed workers was 7 ng/kg. The concentration increased to 420 ng/kg in serum from 119 workers who had been exposed for more than 1 year. Extrapolation to the date at which these workers had been employed, assuming a half-life of 7.1 years, indicated a mean serum concentration at that time of 2000 ng/kg of lipid (highest concentration, 32 000 ng/kg). In the latest update, of mortality rates through 1993 (Steenland et al., 1999), the cohort was restricted to 3538 workers from eight plants for whom a detailed occupational history was available. An exposure matrix was used to estimate the degree of exposure to TCDD in the whole cohort. The SMR for all cancers combined was 1.1 (95% CI, 1.0–1.2). A statistically significant, positive, linear trend in risk for all cancers combined and for lung cancer was found with increasing exposure (Figure 12). The SMR for all cancers combined for the group with the heaviest exposure was 1.6 (95% CI, 1.2–1.8). The excess cancer risk was limited to the workers with the heaviest exposure, which was likely to have been 100–1000 times higher than that experienced by the general population and similar to the doses of TCDD used in studies in experimental animals.

An international cohort of workers exposed to phenoxyacetic acid herbicides and chlorophenols was set up by the IARC (Saracci et al., 1991). The cohort comprised 16 863 men and 1527 women who had been employed in production or spraying, distributed among 20 cohorts in 10 countries. The cohort was updated and expanded with the data of Fingerhut et al. (1991) and Becher et al. (1996) (Kogevinas et al., 1997). The length of follow-up differed by plant, but workers at most of the European plants were followed through 1991–92, and those in the USA through 1987. Current concentrations of TCDD were measured in 574 workers in 10 plants in seven countries ranged from 3 to 390 ng/kg of lipid. This study represents the largest cohort of TCDD-exposed workers and includes groups with heavy exposure to TCDD and cohorts with little or very little exposure. Among the workers who had been exposed to TCDD or higher chlorinated PCDDs, the mortality rate was increased for soft-tissue sarcoma (six deaths; SMR, 2.0; 95% CI, 0.8–4.4), and the rates of mortality from all cancers combined (710 deaths; SMR, 1.1; 95% CI, 1.0–1.2), non-Hodgkin lymphoma (24 deaths; SMR, 1.3; 0.9–2.1) and lung cancer (225 deaths; SMR, 1.1; 95% CI, 1.0–1.3) were slightly elevated. The risks for all cancers combined and for soft-tissue sarcomas and lymphomas increased with time since first exposure. In a direct comparison of persons exposed to higher chlorinated PCDDs and those exposed to lower chlorinated PCDDs or none, the rate ratio for all cancers combined was 1.3 (95% CI, 1.0–1.8). Increased risks were found for breast cancer in both women and men, endometrial cancer and testicular cancer (Table 6). The increased mortality rate from breast cancer was confined to female workers in the Boehringer plant in Germany (nine deaths; SMR, 2.8; 95% CI, 1.3–5.4). Two of three deaths from endometrial cancer similarly occurred among women who had worked in this plant. An excess risk was also seen for cancer of ‘other endocrine organs’, both deaths being from tumours of the suprarenal glands.

Two nested case–control studies of soft-tissue sarcoma (11 incident cases, 55 controls) and non-Hodgkin lymphoma (32 incident cases, 158 controls) were conducted by Kogevinas et al. (1995) within the IARC cohort. A panel of three industrial hygienists estimated the exposure of the workers to 21 chemicals or mixtures, and a cumulative exposure score was calculated for each person and chemical, on the basis of estimated intensity and duration of exposure (in years). An excess risk for soft-tissue sarcoma was associated with exposure to any phenoxy acetic acid herbicide (odds ratio, 10; 95% CI, 1.2–91). Soft-tissue sarcoma was also associated with exposure to TCDD (odds ratio, 5.2; 95% CI, 0.9–32). The odds ratio for non-Hodgkin lymphoma and exposure to TCDD was 1.9 (95% CI, 0.7– 5.1). A monotonic increase in risk was observed with cumulative exposure to TCDD (Figure 13).

This review does not cover studies of phenoxy acetic acid herbicide applicators, including US Air Force personnel who applied Agent Orange, numerous community based case–control studies on soft-tissue sarcoma, malignant lymphoma and other neoplasms, and studies of persons exposed to unspecified combinations of pesticides and herbicides or to herbicides not contaminated by PCDDs (e.g. Axelson et al., 1980; Hardell et al., 1981; Blair et al., 1983; Coggon et al., 1986; Michalek et al., 1990; Coggon et al., 1991; Lynge, 1993). The available information indicates that, in these studies, the concentrations of TCDD were lower and, in most cases, considerably lower than that to which industrial workers and the population of Seveso were exposed. In one Swedish case–control study (Nygren et al., 1986), the mean serum concentration was 2 pg/g in 13 exposed persons and 3 pg/g in 18 unexposed persons. In a study of the most heavily exposed commercial sprayers in New Zealand (Smith et al., 1992), the concentration of TCDD was near the background level in serum of sprayers with 5–10 years of experience. A higher mean concentration of TCDD (53 pg/g) was found for sprayers who had first been employed before 1960 and with a mean duration of spraying 2,4,5-trichlorphenoxyacetic acid of 16 years. Lower values were observed among commercial sprayers in Australia (Johnson et al., 1992) and in US Air Force personnel who had sprayed phenoxyacetic acid herbicides in Viet Nam (Ketchum et al., 1999), the latter being near background.

The findings of studies of cancer risk among persons evaluated in community-based studies and of applicators are contradictory (Figure 14). The large discrepancies are probably due to misclassification of exposure, as most of the persons classified as exposed in these studies probably had TCDD burdens very similar to or only slightly higher than those of persons classified as unexposed.

One case–control study on liver cancer was conducted in Hanoi, Viet Nam. An increased risk for hepatocellular carcinoma (OR, 8.8; 95% CI, 1.9–4.1) was observed among persons who had been in military service in southern Viet Nam for 10 years or more during the time of spraying of Agent Orange by the US Air Force (Cordier et al., 1993).

The incidence of cancer among persons aged 0–19 years in Seveso, Italy, was analysed separately (Pesatori et al., 1993). Given the small number of cases, the three contaminated zones and the two sexes were grouped. Seventeen cases were observed (relative risk, 1.2; 95% CI, 0.7–2.1). Two cases of ovarian cancer were found, with none expected. Two thyroid gland cancers among girls gave a relative risk of 4.6 (95% CI, 0.6–33). The incidence of lymphatic and haematopoietic neoplasms was increased (nine cases; relative risk, 1.6; 95% CI, 0.7–3.4), and particularly those of Hodgkin lymphoma (three cases; RR, 2.0; 95% CI, 0.5–7.6) and myeloid leukaemia (three cases; relative risk, 2.7; 95% CI, 0.7–11).

Low excess risks of the order of 50% were found for all neoplasms combined, in all studies of industrial cohorts in which exposure had been assessed adequately (Table 7). These excess risks were highly statistically significant, and any effect of chance can be excluded. The risk tended to be higher for workers with the heaviest exposure. Increased risks with time since first exposure were observed in those studies in which latency was evaluated (Kogevinas et al., 1997; Steenland et al., 1999). The risks for certain cancers were increased in some studies (lymphomas, multiple myeloma, soft-tissue sarcoma, lung cancer, liver cancer, breast cancer, testicular cancer, endometrial cancer), but, overall, the results were not consistent among studies, and no particular cancer appears to predominate.

In examining the findings on cancer risk in the most informative epidemiological studies, a number of issues should be noted. The excess risks observed were highly statistically significant, and any effect of chance could be excluded. In addition, the studies were prospective and were well conducted. Nevertheless, the results must be evaluated with caution, given that the overall risks are not very high and that the strongest evidence is for industrial populations. Furthermore, there are very few precedents of carcinogens that affect the risk for all cancers with no clear excess for any specific cancer (IARC, 1997). Finally, the strongest evidence comes from studies of persons whose exposure was two to three orders of magnitude higher than that of the general population. In order to extrapolate to the general population, models must be used, with the assumption of similar effects at high and low doses. At present, the real dilemma is not whether dioxins are or are not carcinogenic but, rather, the size of the risk associated with the very low exposure of the general population.

2.3.2 Effects other than cancer

Exposure to dioxins has been associated with a variety of adverse effects (Table 8), including chloracne and related skin lesions, meibomian gland dysfunction, peripheral neuropathy, developmental deficits and coughing and other symptoms of respiratory irritation. Most of the responses observed are biologically plausible, in the sense that experimental evidence also indicated such effects (Birnbaum & Tuomisto, 2000). Most of the available epidemiological studies, however, focused on mortality from cancer and were not designed to evaluate morbidity, e.g. neuropsychological effects, or transient effects such as changes in reproductive hormones. Epidemiological evidence exists for only few of these effects and, in contrast to the experimental evidence, is inconsistent for most effects other than cancer. The evidence in humans is currently conclusive only for dermatological effects and temporary increases in liver enzyme concentrations, and there is increasing evidence for an association with cardiovascular disease.

While there was clear evidence that dioxins and coplanar chemicals can cause a variety of adverse reproductive effects in experimental animals, the evidence for human populations was limited, except for the two episodes of rice oil poisoning in South-East Asia. Accidental consumption of rice oil contaminated with heat-degraded PCBs occurred in two separate episodes, in Japan in 1968 and in Taiwan (Province of China) in 1979. These incidents are referred to as ‘Yusho’ and ‘Yu-cheng’, which mean ‘oil poisoning’ in Japanese and Chinese, respectively. Large numbers of persons were exposed. In the Yusho incident, the average intake of the rice oil contaminated with PCBs, PCDFs and polychlorinated quaterphenyls was estimated to have been 150 ng/kg bw per day (as toxic equivalents), which is five orders of magnitude higher than the reported average background intake in several countries. Analysis of the rice oils in question showed that other chlorinated compounds, including polychlorinated terphenyls, naphthalenes, quarterphenyls and furans, were present, in addition to the PCBs Chen et al., 1985; Masuda et al., 1985).

A number of attempts have been made to determine whether US Air Force personnel involved in spraying TCDD-contaminated herbicides during the Viet Nam War had higher rates of adverse reproductive outcomes (for a review, see Environmental Protection Agency, 2000a). While there have been no strong indications of decreased fertility rates or an increased risk for adverse birth outcomes, such as spontaneous abortion, stillbirths or intrauterine growth retardation, some results suggest higher rates of birth defects, both total and specific (Centers for Disease Control, 1988; Michalek et al., 1998a). A number of the studies lack consistency and may have involved reporting bias or exposure misclassification (Wolfe et al., 1995).

When all birth outcomes (total, 2900) of women from the affected areas in Seveso in 1977–82 were reviewed, no significant increase in the frequency of total or specific birth defects was observed (Mastroiacovo et al., 1988). The authors noted that most of the participants were from outside the two zones with the heaviest TCDD contamination.

Both US Air Force personnel involved in spraying in Viet Nam and workers in the USA employed in factories producing TCDD-contaminated herbicides (chloro-phenoxyacetic acid derivatives) have undergone assessment for circulating gonadotropins and sperm characteristics. In the former group, current serum concentrations of testosterone, follicle-stimulating hormone and luteinizing hormone and testicular abnormalities, sperm count, sperm abnormalities and testicular volume were not associated with current or back-extrapolated serum TCDD concentrations (Henriksen et al., 1996). In 1987, the median serum TCDD concentration in this group was 13 pg/g of lipid (up to 620 pg/g) (Roegner et al., 1991). However, in the more heavily exposed group of herbicide production workers, linear regression analysis indicated that serum TCDD concentrations > 20 pg/g of lipid (upper quartile, 244–3400 pg/g) were positively associated with high serum concentrations of luteinizing and follicle-stimulating hormones and negatively associated with a low serum concentration of testosterone (Egeland et al., 1994).

In addition to the variety of developmental abnormalities observed in children born to mothers who consumed poisoned rice oil (Yusho and Yu-Cheng; see below), about twice as many women affected in the Yu-Cheng incident reported abnormal menstrual flow 14 years after poisoning when compared with neighbourhood control women (Yu et al., 2000). More Yu-Cheng-affected women also reported having had a stillbirth after 1979 (4.2%, with 1.7% in controls), although the overall fertility rates appeared to be aimilar.

Since the initial report that long-term exposure to low doses of TCDD was associated with endometriosis in non-human primates (Rier et al., 1993), attempts have been made to investigate the situation in humans. It has been pointed out (Bois & Eskenazi, 1994) that the level of exposure to dioxin of women in Seveso was comparable to the dose that caused endometriosis in rhesus monkeys. A retrospective study is under way to examine these effects in women in Seveso. Among 79 women attending an infertility clinic in Jerusalem, detectable concentrations of TCDD were found in the blood of eight of 44 with endometriosis (0.6–1.2 pg/g) and only one of 35 with mechanical infertility (0.4 pg/g) (Mayani et al., 1997). In Belgium, endometriosis was diagnosed by laparoscopy in 42 of 101 women who were undergoing treatment for infertility in 1995–98. Coplanar compounds were found in the blood (toxic equivalents, > 100 pg/g of serum lipid) of seven of these 42 women and only two patients who were infertile for mechanical reasons (Pauwels et al., 1999). After that initial report, an adjusted odds ratio for high serum toxic equivalents (presumably > 100 pg/g of lipid) and risk for endometriosis of 4.6 was reported (i.e. 4.6 times as many women with endometriosis would have high serum toxic equivalent values as determined by the CALUX bioassay in vitro) (Pauwels et al., 2000).

Further evidence for a link between TCDD and endometriosis is provided by experimental studies. Groups of eight ovariectomized nude mice injected intraperitoneally with human proliferative phase endometrial tissue which had been treated in vitro with 10^–8 mol/l of estradiol and 10^–6 mol/l of TCDD for 24 h developed more than twice as many lesions as tissue treated with 10^–8 mol/l of estradiol alone (42 and 20, respectively). Furthermore, treatment of human endometrial tissue with TCDD completely blocked the antiproliferative effect seen when the tissue was pretreated with 5 x 10^–7 mol/l of progesterone and estradiol (Bruner-Tran et al., 1999).

An initial report from Seveso indicated that the sex ratio of children born to parents who had lived in zone A, with the highest level of TCDD contamination, was altered in favour of females (Mocarelli et al., 1996). Of the births that occurred from 9 months after the accident to 7 years later, about twice as many children were female (26 males, 48 females). In nine highly exposed families (maternal and paternal serum TCDD concentrations > 100 pg/g of lipid), no male children were born (0 males, 9 females). During 1985–94, the sex ratio apparently reverted to normal (60 males, 64 females). In a follow-up study (Mocarelli et al., 2000), evidence was reported that a paternal serum concentration of TCDD > 80 pg/g of lipid and a body burden of 16–20 ng/kg bw was a significant predictor of a female birth. The most extreme modification of the sex ratio was observed among fathers who had been exposed when they were < 19 years of age (sex ratio, 0.38; 95% CI, 0.30–0.47). A biological explanation of this phenomenon has not yet been provided. General decreases in the normal male:female sex ratio of 1.06:1.0 have been observed in a number of industrialized countries since 1950 but not to the extent temporarily observed in Seveso (Davis et al., 1998).

Comparable analyses in other populations exposed to dioxins, PCBs or PCDFs (in studies of oil poisoning and herbicide spraying in Viet Nam) have not provided support for these findings (Michalek et al., 1998b; Rogan et al., 1999). However, the exposure of these populations to TCDD was considerably lower than that in Seveso. Of 137 births between 1978–85 to women who were identified from the Yu-Cheng registry, approximately equal numbers of males and females (68 and 69) were born (Rogan et al., 1999). Although women who had estimated body burdens of toxic equivalents of 2–3 µg/kg (up to two orders of magnitude higher than that of the background population) delivered children with a variety of developmental and neurological abnormalities, the heavy maternal (and presumably paternal) exposure to PCBs and coplanar contaminants appears to have had a limited effect on the sex ratio of their children.

Neurodevelopmental effects in children have been examined mainly in relation to exposure to mixtures of organochlorinated compounds, predominantly PCBs (Ribas et al 2001).

The most characteristic symptoms observed at birth in the infants of mothers certified as Yusho patients was a gray–dark-brown skin and gingival pigmentation, which usually resolved within 2–5 months. Other clinical features included natal teeth, abnormal skull calcification, ‘rocker-bottom’ heels and intrauterine growth retardation (Yamashita & Hayashi, 1985). Although limited follow-up has been reported of infants exposed in this incident perinatally, one group of 13 children has been described as being apathetic and hypotonic, with lower IQs, 7 years after the event (Harada, 1976).

In the larger Yu-Cheng episode of poisoning, children born up to 6 years after the outbreak (1979–85) weighed on average 0.5 kg less than age-matched neighbourhood controls. While the weight difference had resolved by the time the children were 6–13 years of age, they continued to be shorter (3.1 cm) than the control group (Guo et al., 1995). Behavioural and developmental assessment of this cohort has shown them to be hyperactive, with lower mean IQs and delayed cognitive and psychomotor development. Recent assessment of boys affected by the episode indicated they have short penises and alterations in spermatozoa, 12 youths aged 16–18 years having reduced sperm motility and increased abnormal sperm morphology when compared with 23 controls (Guo et al., 2000). A variety of adverse developmental deficits seen in the Yu-Cheng children were associated with maternal body burdens of ­ 2 µg/kg toxic equivalents, which were about two orders of magnitude higher than those of control populations.

The children exposed in the Yu-Cheng incident were assessed continuously from the age of 6 months to 15 years with one of four tools for measuring cognitive development (Lai et al., 2001). In comparison with a control group matched for age, residence and socio-economic status, Yu-Cheng children scored lower on a psychomotor development index at 2 years of age. At 4–5 years of age, the Yu-Cheng children had lower IQs (2–7-point difference on the Stanford-Binet test) than the controls, and they had lower scores on tests designed to evaluate general intelligence factors at 7, 8, 11 and 12 years of age. When the children reached 13 years of age, there was no significant difference in the test scores from those of the control group.

In a study of 418 mother–infant pairs enrolled in 1990–92, 209 infants were to be breastfed for at least 6 weeks and 209 infants were to be fed exclusively with formula. The median total toxic equivalents in the breast milk was 63 ng/kg of lipid (range, 25–160 ng/kg). The physical and neurological developmental of the infants was assessed from 2 weeks after birth to the present (42 months completed). Previous findings had indicated that prenatal exposure to dioxins and PCBs was associated with lower birth weight and reduced infant growth up to the age of 3 months, subtle alterations in thyroid hormone concentrations and immunological parameters and poorer neurological condition at 18 months of age (Cuijpers et al., 1997; Patandin et al., 1998, 1999; Boersma et al., 2000). The childrens’ intellectual functioning was assessed with the Dutch version of the Kaufman Assessment Battery for Children up to 42 months of age. Although higher PCB concentrations in maternal plasma (> 3.0 µg/l) were associated with lower cognitive scores in the entire cohort and in the formula-fed group, no effects on performance were seen of prenatal or lactational exposure (Patandin et al., 1999). All the scores for cognitive ability were within the range of those of the normal population. The immunological status of the children was also assessed at 42 months of age by administering a health questionnaire, measuring plasma antibodies and conducting lymphocyte marker analyses (Weisglas-Kuperus et al., 2000). The contributions of mono-ortho (14 ng/kg of lipid) and planar PCBs (15 ng/kg of lipid) toxic equivalents in breast milk were associated with a 10–17% higher incidence of recurrent (> six episodes) middle-ear infections, while the toxic equivalent of dioxin (36 ng/kg of lipid) was associated with a slightly (6%) higher prevalence of coughing, chest congestion or phlegm lasting for 10 or more days. There was no association between antibody titres, leukocyte counts and immunological markers and dioxin and/or PCB toxic equivalents.

In a study designed to investigate the causes of enamel defects in teeth, children born in Finland in 1987 underwent dental examinations at the age of 6–7 years and were classified according to the extent of changes in mineralization in their teeth. Although the duration of breastfeeding and breast milk toxic equivalents at 4 weeks of age (mean, 49 pg/g of lipid) were not associated with the changes in teeth mineralization, exposure to dioxin or furan toxic equivalents was associated with these changes (Alaluusua et al., 1996, 1999). In a related study, 202 children in a PCB contaminated area of Slovenia underwent a dental examination to assess developmental defects of the enamel (Jan & Vrbi, 2000). The tooth dentine of these children had more than five times more PCB than that of children in a control group (38 and 7 ng/g, respectively). On the basis of an analysis of maternal serum and milk samples and representative foods from the same region, the children were estimated to have been exposed to toxic equivalents at a rate of 39 pg/kg bw per day. About twice as many exposed children as controls were found to have permanent teeth with enamel defects on the labial surface (22% and 13%, respectively).

Chloracne is the best recognized dermal effect of exposure to TCDD. Chloracne was reported in at least a few of the workers involved in the reported accidents at trichlorophenol production facilities. Chloracne has also been reported among workers involved in daily production of TCDD-contaminated products (Suskind & Hertzberg, 1984), such as phenoxyacetic acid herbicide workers, and in three laboratory technicians exposed to pure TCDD. Chloracne was diagnosed in at least 193 (0.6%) Seveso residents, mostly children, exposed during the accident (Table 9; Bertazzi et al., 1998). It is not clear whether the children were more susceptible or whether they had heavier exposure to TCDD. Of the populations with known exposure to TCDD, US Air Force personnel in Viet Nam and persons in Missouri, USA, who were accidentally exposed were not reported to have chloracne, but the latter were examined 10 years after exposure.

Chloracne persisted in a few workers in plants in Germany and the USA many years after exposure. In Seveso residents, the condition disappeared after discontinuation of exposure, despite very high initial serum concentrations of TCDD, ranging from 820 to 56 000 pg/g, measured within 1 year of the accident.

Although chloracne is associated with the intensity of exposure to TCDD, there is no direct correlation between exposure and the occurrence of chloracne. Persons with severe chloracne are often found, on a group level, to have had heavier average exposure than those with moderate or no chloracne. This can clearly be seen in the analysis of TCCD concentrations of BASF workers exposed during the accident (Zober et al., 1997). However, an absence of chloracne is not equivalent to an absence of exposure, and the presence of chloracne is not necessarily equivalent to very heavy exposure.

Various other dermal effects other than chloracne have been described among persons exposed to TCDD, including red, irritated eyes, conjunctivitis and blepharitis, eyelid cysts, hyperpigmentation and hirsutism. These findings have been reported less frequently than chloracne.

Increased liver size was reported in workers in two trichlorophenol production plants, one in the former Czechoslovakia and the other in the USA. Temporary hepatomegaly was reported in five of 22 Seveso residents with severe chloracne (Reggiani, 1980), but hepatomegaly has not been observed in other populations of workers or in US Air Force personnel in Viet Nam.

lncreased activity of gamma-glutamyl transferase was observed in children in Seveso shortly after the accident, but this declined during the 5 subsequent years (Mocarelli et al., 1986). Persistently high gamma-glutamyl transferase activity has been observed in trichlorophenol production workers in various plants, and significantly high activity was observed in the US Air Force personnel (Roegner et al., 1991).

Increased serum activity of alanine and aspartate aminotrasferases is probably a transient effect of exposure. No increases were found in studies of persons 10–30 years after exposure. Clinical evidence of liver disease was not found in any of the studies in which increases were identified.

Increased concentrations of D-glucaric acid were found in adults and children in Seveso in 1976 (Ideo et al., 1985). By 1981, however, the concentrations were within the normal range. No increases were found in US Air Force personnel in Viet Nam (Roegner et al., 1991).

The metabolism of porphyrins was examined in two studies. In one study of trichlorophenol production workers (Bleiberg et al., 1964), porphyria cutanea tarda was reported in 11 of 29 persons with chloracne. In a study of industrial cohorts in the USA, no association was found between exposure to TCDD and the prevalence of porphyria cutanea tarda (Calvert et al., 1994).

Thyroid function was associated with exposure to dioxin in some cohorts of production workers and in US Air Force personnel in Viet Nam, but most of the results were not statistically significant or consistent among studies (Sweeney & Mocarelli, 2000). In a study in a trichlorophenol production factory, no significant difference in T4 and T4-binding globulin concentrations was found between exposed and unexposed workers (Suskind & Hertzberg, 1984). In workers exposed in the BASF accident, the concentrations of TSH, T4 and T4-binding globulin were within normal levels; the concentrations of the latter two were positively correlated with that of TCDD (Ott et al., 1994). In the industrial cohort in the USA, no significant differences were found between exposed and unexposed persons, although the free T4 index and T4 concentration were increased in trichlorophenol production workers (Calvert et al., 1999). US Air Force personnel spraying herbicides in Viet Nam showed a slight, nonsignificant increase in TSH concentrations which was associated with TCDD concentration (Grubbs et al., 1995).

Higher mean glucose concentrations were found among TCDD-exposed than among unexposed persons in the industrial cohort in the USA (Calvert et al., 1999), in US Air Force personnel in Viet Nam (Henriksen et al., 1997) and in workers exposed at the BASF facility at the time of the study, but not when the concentration was estimated at the time of last exposure (Ott et al., 1994). No association was found in workers in Nitro, West Virginia, USA (Suskind & Hertzberg, 1984). Among US Air Force personnel exposed in Viet Nam (Henriksen et al., 1997), the prevalence of diabetes and use of oral medication to control diabetes increased with increasing exposure to TCDD, while the time to onset of diabetes decreased with the intensity of exposure to dioxin (Figure 15). Among trichlorophenol production workers in the USA, the prevalence of diabetes was not associated with serum concentrations of TCDD. Men with very high TCDD concentrations (> 1500 pg/g), however, tended to have a high prevalence of diabetes (Calvert et al., 1999). In the industrial cohort in the USA (Steenland et al., 1999), any mention of diabetes on the death certificate was negatively associated with exposure. An increased rate of mortality from diabetes was seen among women who had been in zone A or B in Seveso (rate ratio, 24; 95% CI, 1.2–4.6), but no excess was observed among men (Pesatori et al., 1998).

In the BASF accident cohort and the industrial cohort in the USA, no relationship was found between the concentration of total cholesterol or high-density and low-density lipoproteins and increasing serum concentrations of TCDD (Ott et al., 1994; Calvert et al., 1996). Similar results were found for Seveso residents (Mocarelli et al., 1986; Assennato et al., 1989). In contrast, a positive relationship between serum TCDD and total cholesterol concentration was found among US Air Force personnel in Viet Nam (Roegner et al., 1991). This association was less strong in a later follow-up study (Grubbs et al., 1995).

No or very small differences in concentrations of triglycerides were found in TCDD-exposed workers in the BASF accident cohort and the industrial cohort in the USA when compared with unexposed persons (Ott et al., 1994; Calvert et al., 1996), with similar results in Seveso (Mocarelli et al., 1986; Assennato et al., 1989). TCDD concentrations were consistently associated with triglyceride concentrations in US Air Force veterans of Viet Nam (Roegner et al., 1991; Grubbs et al., 1995).

Exposure to dioxins has been associated with an excess risk for heart disease. Excess mortality from ischaemic heart disease was found in several industrial cohorts (Table 10) and in zone A at Seveso. The study of non-flying US Air Force personnel gave mainly negative results, but there was an excess risk for personnel with the heaviest estimated exposure to TCDD. No excess risk was observed in an analysis of cardiovascular morbidity in a subset of the US industrial cohort. An increased risk with increasing exposure was observed in the Dutch production cohort (Figure 11) and the Boehringer cohort (Figure 10). In the US industrial cohort, the rate of mortality from heart disease showed a weak increasing trend with heavier exposure.

Numerous case reports have been made of an association between short- or long-term exposure to TCDD and headache, insomnia, nervousness, irritability, depression, anxiety, loss of libido and encepalopathy. There have been some reports of persistent symptoms. Effects were described in US Air Force personnel in Viet Nam (abnormal coordination), workers at the plant in Seveso (polyneuropathy in lower limbs) and in Seveso residents (neuropathy) (Pocchiari et al., 1979; Filippini et al., 1981; Roegner et al., 1991). No association was found between exposure to TCDD and depression in the US industrial cohort or in the US Air Force personnel (Roegner et al., 1991; Alderfer et al., 1992).

The results of epidemiological studies on most health outcomes other than cancer are still inconsistent. The evidence is currently conclusive only for dermatological effects and temporary increases in liver enzyme activity, and there is increasing evidence for as association with cardiovascular disease, particularly among the most heavily exposed workers. Experimental data indicate that endocrine and reproductive effects should be among the most sensitive effects in both animals and humans. Only a few of these effects have been evaluated in epidemiological studies.

PCDDs and PCDFs are tricyclic aromatic compounds with similar physical and chemical properties, and the two classes are structurally similar. In simplified terms, congeners of both classes are called ‘dioxins’, although they are differentiated chemically as dioxins and furans. The PCDDs comprise 75 individual compounds and the PCDFs comprise 135 compounds. Figure 16 shows the chemical structures of these classes. Of the 210 theoretically possible congeners, a subset of 17 has chlorine substitutions at the 2, 3, 7 and 8 positions. The most toxicologically potent dioxin is TCDD (Figure 17). Table 1 lists all 17 congeners with chlorine substitution in the 2,3,7,8 position, and all these congeners are included in this assessment.

The most important properties of PCDDs and PCDFs are their chemical stability and high solubility in fat and organic solvents, with low solubility in water. Because of these properties, PCDDs and PCDFs are primarily associated with particulate and organic matter in environmental samples, and they accumulate in the food chain.

PCBs have been manufactured as technical products, which are liquids with variable viscosity. These liquids are mixtures of dozens of different compounds (‘congeners’) depending on the degree of chlorination.

A total of 209 congeners are theoretically possible. Following IUPAC rules, Ballschmiter & Zell (1980) developed systematic numbering of all congeners, which is now used worldwide for identification of a specific congener. Four octachloro-biphenyls were numbered incorrectly and corrected subsequently (Schulte & Malisch, 1983; Deutsche Forschungsgemeinschaft, 1988).

The physical and chemical properties of each congener vary according to the degree and position of chlorine substitution. Figure 18 shows the chemical structure of PCBs.

From a chemical and toxicological point of view, PCBs can be divided into three groups: with no chlorine substitution in the ortho position (non-ortho PCBs, coplanar PCBs), with one chlorine in the ortho position (mono-ortho PCBs) and di-ortho-substituted congeners. A WHO consultation (van Leeuwen & Younes, 2000) included 12 of the 209 PCBs in the set of coplanar compounds (see Table 1). The PCBs considered to have coplanar properties are those with either one or no chlorine substitution in the ortho position. The toxicity of these congeners is different from that of other PCBs. Figure 19 shows the planar structure of 3,3´,4,4´-TCB as an example of a non-ortho PCB, which is coplanar, as both rings are in the same plane. Figure 20 shows the twisted structure of di-ortho-substituted PCBs.

The mono-ortho PCB 2,4,4´-trichlorobiphenyl and the di-ortho PCBs 2,2´,5,5´-tetrachlorobiphenyl, 2,2´,4,5,5´-pentachlorobiphenyl, 2,2´, 3,4,4´,5´-hexachloro-biphenyl, 2,2´,4,4´,5,5´-hexachlorobiphenyl and 2,2´,3,4,4´,5,5´-heptachlorbiphenyl (and also sometimes the mono-ortho PCB 2,3´,4,4´,5-PeCB) are called ‘marker PCBs’ and are used as decisive congeners in determining tolerances for PCBs in various matrices in some countries. They represent the subsets of low- and high-chlorinated PCBs, as they are indicative congeners that can be determined readily. However, coplanar PCB content cannot be derived by determination of these marker PCBs. A variety of PCB products was available with an unknown range of contamination with coplanar PCBs. Except in the case of accidents, food samples are contaminated with PCBs from different sources as a result of various transport and bioaccumulation processes. For these reasons, information on the occurrence of PCBs in general (e.g. estimatates of total PCBs or of marker PCBs) is not sufficient to describe exposure to the coplanar PCBs. Similarly, it is not known whether regulations for total PCBs adequately address coplanar PCBs.

As with PCDDs and PCDFs, the most important properties of PCBs are their chemical stability and high lipophilicity. Because of these properties, PCBs are also primarily associated with particulate and organic matter in environmental samples and accumulate in the food chain.

3.2.1 Congeners to be determined

PCDDs and PCDFs are usually found as complex mixtures of variable composition in various matrices. Their identification and quantification require highly sophisticated analysis, as the 17 toxic congeners with 2,3,7,8-chlorine substitution must be separated from the less toxic congeners. Usually, PCDDs and PCDFs are determined by capillary gas chromatography (GC) with mass spectrometry (MS).

In the past, analyses of PCBs focused mainly on determination of total PCBs or the marker congeners 2,4,4´-trichlorobiphenyl, 2,2´,5,5´-TCB, 2,2´,4,5,5´-PeCB, 2,2´,3,4,4´,5´-HxCB, 2,2´,4,4´,5,5´-HeCB and 2,2´,3,4,4´,5,5´-HpCB, which are the predominant PCB congeners in humans and foods of animal origin. However, these PCB congeners appear to have relatively little toxicity. On the basis of the available toxicological information, the non-ortho PCBs 3,3´,4,4´-TCB, 3,4,4´,5-TCB, 3,3´,4,4´,5-PeCB and 3,3´,4,4´,5,5´-HxCB and the mono-ortho congeners 2,3,3´,4,4´-PeCB, 2,3,4,4´,5-PeCB, 2,3´,4,4´,5-PeCB, 2´,3,4,4´,5-PeCB, 2,3,3´,4,4´,5-HxCB, 2,3,3´,4,4´,5´-HxCB, 2,3´,4,4´,5,5´-HxCB and 2,3,3´,4,4´,5,5´-HpCB were assigned TEFs by a WHO expert group in 1997 and are therefore analysed to determine their PCB toxic equivalent. Data on these coplanar PCB congeners are still scarce. As non-ortho PCBs occur in a much lower range of concentrations, mono-ortho and non-ortho PCBs must be determined separately. Reliable determinations of non-ortho PCBs in food have been performed by high-resolution MS, as demonstrated in interlaboratory studies.

3.2.2 Availability of official methods

In contrast to methods for the determination of food additives, residues (pesticides, veterinary drugs) and many contaminants, there are no standardized or harmonized methods (‘official methods’) for the determination of dioxins or coplanar PCBs in food. Reliable results can be obtained in the absence of official methods if the method used is shown to fit the purpose and to fulfill analytical quality criteria developed in other fields of residue analysis.

3.2.3 Upper- and lower-bound concentrations

For comparison of analytical results with regulatory limits and, in general, with results from other laboratories, the limit of detection (LOD; lowest limit for qualitative identification) and/or the limit of quantification (LOQ) must be taken into account. For analysis of PCDDs and PCDFs, all 17 congeners with 2,3,7,8-substitution must be determined. To calculate toxic equivalents, the results for each congener are multiplied by the specific TEF and then added up. In most cases, the concentrations of a few of the 17 congeners are below the LOD and/or the LOQ. The situation can become critical if many congeners cannot be determined or if the toxicologically important congeners are not found.

Various approaches are used to address this problem:

• calculate the contribution of each undetected congener to the toxic equivalents as zero (lower-bound concentrations);
• calculate the contribution of each undetected congener to the toxic equivalents as the LOD or LOQ (upper-bound concentration);
• calculate the contribution of each undetected congener to the toxic equivalents as half the LOD or LOQ;
• replace an undetected congener in a data set by the minimum of the usual contribution to the toxic equivalents and the LOD;
• make multiple imputations, with censoring of data.

The LOD or LOQ can become a critical factor in decisions based on analytical results if many congeners are not determined or if congeners representing the higher TEFs are not found. This is because the value of ‘undetected’ congeners is needed in order to estimate the overall toxic equivalents. For example, TCDD and 1,2,3,7,8-PCDD have a TEF of 1, while other, more prevalent congeners have TEFs of 0.1 or 0.01. If analysis of a sample results in detection of large quantities of TCDD or 1,2,3,7,8-PCDD, the resulting toxic equivalents will be affected by the method used to estimate the values. Some laboratories calculate the contribution of undetected congeners to toxic equivalents as zero, whereas others use the full LOD or the full LOQ to estimate toxic equivalents. Thus, estimates of dioxin content may have low or high bias. The effect of these biases is obscured by the reporting of a single value for toxic equivalents.

An example of the effect of using replacement values for undetected congeners in samples is shown in Table 11. It is clear that the method used to estimate a value for undetected congeners can dramatically affect the reported toxic equivalents value for a food. If 0 is used, low estimates of dioxin content may result, owing either to truly low concentrations in the sample or to high LODs/LOQs that did not allow quantification of the congeners with higher TEFs. For example, Table 11 shows that the imputation method has little effect on the estimated toxic equivalents for beef Stroganoff when high-resolution MS is used. However, with ion trap MS, with which the LOD is 5–10 times higher, use of 0 to equate to nondetection results in an artificially low estimate of toxic equivalents when compared with that achieved with high-resolution MS.

Table 11 illustrates the importance of using analytical methods with low LODs when making decisions about tolerances. Even with the more sensitive high-resolution MS, however, the estimate of toxic equivalents for corn chips in Table 11 was strongly influenced by the choice of replacement value for undetected compounds, with a 183-fold difference when zero and the LOQ are used. Thus, it is important not only to be able to use high-resolution MS for determination but also to use an appropriate amount of sample for extraction and clean-up to obtain a sufficient concentration in the final sample for use in GC–MS. The explanations, conclusions and recommendations in sections 3.2.4 and 3.2.5 below indicate the analytical requirements to avoid using methods with insufficient sensitivity.

To ensure that a finding of low concentrations of dioxin is really the result of low concentrations in the sample, the concept of tolerances as upper-bound concentrations has been developed in some countries and for some uses of estimates of concentrations. This concept requires use of the full LOD or LOQ instead of zero for undetectable substances: Upper-bound concentrations are calculated on the basis of the assumption that all the values for various congeners that are below the LOD/LOQ are equal to the LOD/LOQ.

When the LOQs are high relative to the decision criteria for congeners, therefore adding significantly to the estimated toxic equivalents, use of the upper-bound LOQ can result in artificially high toxic equivalents. This should be considered in defining background contamination, monitoring tolerances or estimating intake. In methods with insufficient sensitivity, the difference between lower-bound and upper-bound concentrations may be 10–100 or, in extreme cases, even higher. For example, if the sensitivity of a method is inappropriate for monitoring a tolerance, use of the concept of upper-bound LOD leads to estimates of toxic equivalents that are ‘false-positive’ results. This is a clear indication that a more sensitive method is needed. In particular, use of low-resolution MS in analysing food or samples of low weight or quantity (for a quick, easy analysis) can result in relatively high values for dioxin content as the upper-bound LOD. This bias cannot be seen in reported toxic equivalents, unless results for individual congeners are available. Thus, in defining background contamination or evaluating exposure, published data must be reviewed critically to eliminate relatively high values that are the result simply of inadequate LODs.

For setting and monitoring tolerances on the basis of toxic equivalents, the closeness of the LOQ to the appropriate tolerance must be evaluated as part of a decision to accept or reject a food. High LODs relative to the tolerance should lead to rejection of an analytical result for a sample on the basis of poor quality assurance, and consequently poor reliability of the estimate of toxic equivalents. Therefore, some governments may choose to apply upper-bound estimates of toxic equivalents, with a preference for the upper-bound LOQ rather than the upper-bound LOD, as a screening method in order to remove questionable foods from the marketplace. In the absence of these steps, there is a risk that foods in which a maximum level of a toxicant is exceeded will reach consumers. It is the responsibility of laboratories to achieve the required sensitivity in order to avoid unnecessary rejection of analytical results or of foods.

In risk assessment, use of upper-bound concentrations leads to overestimes of intake and use of lower-bound concentrations to underestimates. For this purpose, imputation of half the LOD for undetected congeners yields an acceptable estimate of both toxic equivalents and its associated standard deviation of uncertainty (Hoogerbrugge & Liem, 2000).

The Committee therefore recommended that, in future, laboratories report their results in relation to the lower-bound, upper-bound and half the LOD. In that way, all the necessary information is available for interpretation of the results according to specific requirements. As a minimum, it must be clear from a report which concept was applied.

3.2.4 Required sensitivity and analytical approaches

Whereas many environmental samples (such as soil and sewage sludge) can be analysed by low-resolution MS, feed, food and human milk or tissue samples should be analysed for ‘ultra-trace’ concentrations (usually, 0.1–1 pg/g of lipid in milk and meat and in eggs of caged chickens, as toxic equivalents; mean, 10 pg/g of lipid in wild and farmed freshwater fish or > 100 pg/g of lipid in cases of (highly) elevated concentrations, as toxic equivalents; 0.1–0.5 pg/g dry matter for food of vegetable origin, as WHO toxic equivalents). As the fat content of foods of animal origin varies widely, a wide range of values for PCDD/PCDF is calculated on a fresh weight basis. Therefore, lipid-adjusted toxic equivalents are used for dioxins in animal foods, in order to provide a uniform basis for setting tolerances. For fish, dioxin content should be reported on the basis of both the fat content and on fresh weight, in view of the extremely wide range of fat contents of various kinds of fish.

For reliable analysis of food samples with contamination in the range of that of the normal background, high-resolution MS has been shown to have the required sensitivity and specificity. In collaborative studies for the determination of PCDDs/PCDFs in various foods, use of high-resolution MS was successful (Chemisches Landes- und Staatliches Veterinäruntersuchungsamt Münster, 1996; Malisch et al., 1996, 1997, 2000b; Lindström et al., 2000). This was also the basis for selecting reference laboratories for WHO-coordinated studies of exposure to dioxins in human milk (see section 6.4.2).

The required specificity can also be provided by tandem MS (MS/MS). While MS/MS with sector or quadrupole instruments requires a series of mass analysers in space, ion traps require one mass analyser to perform MS/MS in time. As the techniques for MS/MS depend on the type of mass analyser used, a variety of instruments has been designed. The advantage of ion-trapping MS/MS systems is their low price. Nevertheless, their sensitivity is considerably lower than that of high-resolution MS instruments. The LOQ for ion-trapping MS/MS systems for TCDD (signal:noise, 3:1) can be assumed to be in the range 100–300 fg, whereas that of modern high-resolution MS instruments is about 3 fg. The reduced sensitivity can be compensated to a certain degree by using much larger samples for extraction and clean-up. However, use of 10-fold (or more) larger samples causes problems with regard to the availability of sample material and the analytical procedure. Therefore, ion-trapping MS/MS could be used as screening method for selecting high concentrations, like bioassays. However, unlike bioassays, this screening method makes it possible to see patterns of congeners.

Several reviews (Cooke et al., 2000; Hilscherova et al., 2000; Hoogenboom et al., 2000; Behnisch et al., 2001a,b) and guidelines (reviewed by Cooke et al., 2000; Behnisch et al., 2001a) have been published on new techniques for measuring dioxin toxic equivalents. Bioassays such as CALUX (chemical-activated luciferase gene expression), Ah immunoassays, EROD bioassays and enzyme immunoassays have been developed for rapid screening of various matrices (food, sediments, soil, fly ash). So far, only CALUX has been used for food. In contrast to MS methods, these methods allow biological interpretations. While GC–MS is the most powerful method for identifying and quantifying congeners and for congener-specific pattern recognition, it does not provide a direct measure of total dioxin-like toxicity (toxic equivalents) for all congeners in a matrix that act through the Ah receptor pathway. The TEF concept is used to transform the GC–MS database into results relevant to toxicity, i.e. the individual concentrations of those congeners that have been assigned a TEF is multiplied by the TEF, and these are added to give the total toxic equivalents. The bioassays themselves provide an indication of the total toxic equivalents of dioxin-like activity present in a certain matrix, including possible interactive (synergistic or antagonistic) effects of all the coplanar congeners in a complex mixture. However, the bioassays in their present form cannot discriminate between different classes and/or congeners of PCBs, PCDDs and PCDFs, partly depending on the clean-up procedure, and therefore cannot reveal the pattern of congeners.

A ‘toxicity identification evaluation’ approach (Environmental Protection Agency, 1992) has been proposed, to give a ‘real world picture’ of which compounds are most responsible for the dioxin-like activity of the sample, on the basis of differences in polarity in fractionation and selection of stable and unstable compounds. This approach may allow detection of novel coplanar compounds auch as bromodioxins or combinations of bromo–chlor dioxins. It is a means for identifying patterns in different compound classes rather than single congeners, but it requires further developments and is not considered useful for screening for PCDD/Fs and coplanar PCBs.

Hoogenboom et al. (2000, 2001) reported that the CALUX has been validated in their laboratory for milk fat and citrus pulp pellets to monitor adherence to limits of 5 pg/g of lipid and 500 pg/kg of product, respectively, as toxic equivalents. The results for other animal and plant fats were comparable to those for milk fat, and those for feed ingredients (with the exception of kaolinitic clay) were similar to the results for citrus pulp. The CALUX method is sensitive not only to dioxins but also to other Ah-receptor agonists, possibly resulting in ‘false-positive’ results. This potential problem was partly overcome by use of selective clean-up procedures (such as acid silica) and long exposure of the cells (24 h). However, as the possibility of ‘false-negative’ results appears to be negligible, the method was considered ideal for screening samples with no dioxins and samples suspected of containing dioxins and requiring further investigation by the GC–MS reference method.

The results of a comparison of simultaneous analysis of 19 samples (eight fats, three feeds, four eggs, four milks) by CALUX and GC–MS were presented by van Overmeire et al. (2000). A strong correlation was found from a double-log graph in the range 0.1–10 000 pg/g of lipid as toxic equivalents, although the double-log presentation reveals some differences in the numerical values. According to the authors, the observed correlation indicates that analyses by CALUX are predictive of the results of GC–MS, and the method can be used as a rapid, sensitive screen for the dioxin content of feed and food samples. Concentrations of 5 pg/g of lipid, as toxic equivalents, were concordant with the two methods. A tolerance of 5 pg/g lipid, as WHO toxic equivalents (only for PCDD/PCDF), was set by Belgium during the recent incident of dioxin contamination for milk, eggs and meat samples, in order to exclude highly contaminated samples from being marketed.

The usual background concentration of contamination of food of animal origin (except fish from European waters) is 0.1–2 pg/g of lipid, as toxic equivalents; most samples contain < 1 pg/g of lipid, as WHO toxic equivalents (only for PCDD/PCDF). If maximum concentrations are set, they will be based on concentrations above the average background. Therefore, a method should be developed to allow screening for elevated concentrations (e.g. above a fixed maximum or action level) or as a routine or reference method to determine the content of PCDD/PCDF and/or coplanar PCBs in the range of the usual background contamination or target values. This is true for all kinds of methods of detection, whether GC–MS or bioassays.

3.2.5 Conclusions and recommendations

These examples show that the principles common to all analytical methods are valid for dioxins as well. First, the purpose of a method should be defined clearly, including the matrix to be analysed, the content to be determined reliably, possible limitations and whether it is to be used for quick screening or reliable determination (routine or reference method). Then, it should be shown that the method is suitable for the purpose, by validation and demonstration of basic minimal statistical requirements. Last but not least, the applicability of the method should be proven in collaborative studies.

As mentioned above, there are no official methods for the determination of PCDDs/PCDFs or coplanar PCBs in food. Each laboratory has its own method, although it must demonstrate that the method is suitable for the purpose. Nevertheless, the criteria for accepting methods for determining dioxin in food must be harmonized. Given the wide variety of methods and the unequal quality assurance and control across laboratories, harmonization is needed to allow free trade if maximum levels are developed. These principles should be valid for both GC–MS methods and for bioassays. Joint papers by 13 authors in nine different international institutions contain a discussion of these general considerations and of GC–MS methods (Malisch et al., 2001), and papers by 11 authors in 10 institutions address the bioassays (Behnisch et al., 2001c).

Legislative measures for dioxins in food can comprise ‘maximum limits’, ‘action levels’ and ‘target levels’. Maximum limits can be set at a strict but feasible concentration in order to allow elimination of products contaminated at an unaccep-tably high level. Action levels can be set to allow monitoring of increased concentrations. Target levels could be set at a concentration that would result in an ultimate dietary intake below the lower range recommended by WHO. Some governments have begun to discuss such levels, the ratio of maximum limits to action levels being about 1.5 and that of maximum levels to target levels being about 5.

If legislative measures are based on these three concepts, analysis for certification must allow relaible determination of the dioxin content in the range of maximum levels and action levels. For evaluation of exposure and time trends, analysis must be oriented to the target levels.

General statistical parameters have been established for the analysis of other residues which might provide orientation. For example, to allow certification (in national or international commerce) relative to maximum levels, laboratories should be able to meet certain basic requirements, such as:

• demonstration that the performance of a method is in the range of the maximum level, e.g. 0.5, 1 and 2 times the maximum level, with an acceptable coefficient of variation for repeated analyses in the range of interest;
• a LOD representing at least one-fifth of the maximum level, to ensure that acceptable coefficients of variations are met in the range of the maximum level;
• continuous use of blank controls and spiking experiments or analysis of control samples (preferably, if available, certified reference material) as internal quality assurance measures;
• a result based on toxic equivalents that is accurate (closeness of the mean of repeated analyses to the ‘true’ value, determined in comparison with reference methods) within about ± 20% of the maximum limits or action levels;
• successful participation in interlaboratory studies to prove competence in specific analyses; and
• whenever possible, accreditation of laboratories that supply analytical data by a recognized body, to ensure that they are applying analytical quality assurance. For example, laboratories could be accredited as following ISO 17025, supplemented by standard operating procedures and monitored by quality control managers following the principles of OECD for Good Laboratory Practice.

Studies from laboratories in which high-resolution GC with high-resolution MS are used have shown that all 17 PCDD/PCDF congeners with 2,3,7,8-substitution can be determined reliably, even at concentrations < 1 pg/g of lipid, as WHO toxic equivalents (only PCDD/PCDF included). However, successful participation in inter-calibration studies for e.g. soil or sewage samples does not necessarily prove competence in the analysis of food samples, with their lower range of contamination. Therefore, continuous participation in interlaboratory studies for determination of dioxins and coplanar PCBs in the relevant food matrices is mandatory.

As long as no target values are fixed, the following requirements should be met for reliable determination of concentrations in the range of the usual background contamination:

• For food of vegetable origin in general, a LOQ of approximately 0.1 pg/g dry matter, as WHO toxic equivalents, is appropriate to allow reliable differentiation between samples with high dioxin concentrations and background contamination (note that this sensitivity requirement should not be confused with the sensitivity needed for enforcement of maximum levels, in which case it is defined as a fraction of the maximum level). However, to follow time trends of background contamination in those matrices, the LOQ should be at least a factor of 10 lower. For products of terrestrial origin and for fish and fish products, a LOQ of about 0.2 pg/g of lipid, as WHO toxic equivalents (upper-bound LOQ) would be a ‘marginally acceptable’ LOQ and appropriate for differentiating between samples with high dioxin concentrations and background contamination. As the usual background contamination, e.g. in milk products and pork meat, in some countries is 0.1–0.3 pg/g of lipid as WHO toxic equivalents, a fully acceptable LOQ for exact determination and time trends should be < 0.1 pg/g of lipid. These values should result from the congener patterns found usually in food.
• For GC–MS methods, the difference between the upper-bound LOQ and the lower-bound LOQ should not exceed 10–20% for food of animal origin contaminated with dioxin at a concentration of about 1 pg/g of lipid, as WHO toxic equivalents (only PCDD/PCDF included). This requirement should be met for products such as butter, beef, cheese and non-defatted milk products, and similar requirements on a fresh weight basis can be derived for products such as skinned fish fillets with a low fat content.
• The accuracy of a result based on toxic equivalents (closeness of the mean of repeated analyses to the ‘true’ value determined in comparison to reference methods) should be within ± 30% of the target values.

• Recovery must be monitored by adding [¹³C]-2,3,7,8-chlorine-substituted internal PCDD/PCDF standards. At least one of these congeners must be added for each tetra- to octa-chlorinated homologue group, with a clear preference for use of all 17 [¹³C]-2,3,7,8-chlorine-substituted internal standards. Relative response factors should be determined for those congeners for which no ¹³C-labelled analogue is added.
• For vegetables, addition of the internal standards before extraction is mandatory. For foods of animal origin, the internal standards can be added either before extraction or after fat extraction, if complete extraction of fat can be demonstrated.
• A recovery standard must be added before GC–MS analysis.
• The recoveries of the internal standards should be 50–120%.
• PCDDs/PCDFs should be separated from interfering chlorinated compounds such as chlorinated diphenyl ethers.
• GC baseline separation of isomers should be sufficient (< 25% peak-to-peak between 1,2,3,4,7,8-HxCDF and 1,2,3,6,7,8-HxCDF).
• Identification should be performed according to the principles of EPA method 1613 (revision B).

Before biological or chemical analyses are begun, the relevant quality control criteria should be well defined. The characteristics of these criteria will vary, depending on the analytical approach being used. Three analytical approaches can be used in bioassays:

(1) The first is a screening approach, in which the response of samples is compared with that of a reference sample at the action limit. Samples with a response less than that of the reference are considered to be negative, while those with a response higher than that of the reference are considered suspect. The requirements may be less strict than those for a quantitative method:

• A blank and reference standard should be included in each test series, which are extracted and tested at the same time as the test sample, under identical conditions. The reference sample must show a clearly greater response than the blank.
• Extra reference samples at 0.5 and 2 times the maximum level should be included to demonstrate proper performance of the test in the range of interest for monitoring adeherence to the maximum levels.
• In testing other matrices, the suitability of the reference standards should be demonstrated, preferably by including samples shown by GC–MS to contain dioxin at a concentration similar to that of the reference sample or a blank spiked at this concentration.
• As no internal standards can be used, tests for repeatability are important to obtain information about the standard variation within one test series. The coefficient of variation should be < 20%.
• For bioassays, the target compounds, possible interferences and how low the concentration in the reagent blanks used should be clearly stated.

(2) The second is a quantitative approach, which requires a standard dilution series, duplicate or triplicate clean-up and measurement, and blank and recovery controls. The result should be expressed as toxic equivalents (pg/g), thereby assuming that the compounds responsible for the signal correspond to the TEF principle. This can be done by using TCDD (or a standard mixture of dioxins and furans) to produce a calibration curve for calculating the WHO toxic equivalents in the extract and thus in the sample. This is subsequently corrected for the WHO toxic equivalents calculated for a blank sample (which may include impurities from the solvents and chemicals used) and recovery (calculated from the WHO toxic equivalents in a reference sample around the limit of the residue). It should be noted that part of the apparent loss to recovery may be due to differences between TEF values in the bioassays (relative potency) and the official TEF values set by WHO.

(3) The third design involves use of bioassays for various toxicological end-points, e.g. Ah receptor or antibodies. In this case, a wide range of concentrations should be used to evaluate a full median effective concentration (EC₅₀) and to obtain toxic equivalent values from various measures (EC₅₀, EC₁₀ or lowest data point closest to the minimal LOQ).

• Sampling, extraction and clean-up procedures and general validation should be done according to the guidelines for chemical analysis (e.g. EPA method 8290) or specified for individual bioassays (e.g. EPA method 4025 or 4425).
• A standard reference material containing PCBs, dioxins/furans and probably polycyclic aromatic hydrocarbons should be included in every test.
• Interference in the test method (e.g. tributyl-zinc) should be defined as extensively as possible.
• The laboratory should be approved according to an ISO norm (e.g. 9001 or others) and/or Good Laboratory Practice.
• Standard operating procedures must be established for extraction, clean-up, use of blank samples, bioassay performance, data reporting and data handling.
• Charts should be maintained to record the long-term stability of the bioassay response for TCDD (e.g. induction factor, EC₅₀ value). If the response in a test is greater than two standard deviations from the long-term mean, the results may be invalid and the samples should be re-tested.
• Validation in inter- and intra-laboratory studies must be performed.
• An instrumental control chart should be maintained, and the sensitivity and linearity of the machine should be tested at least monthly with a standard.
• Routine quality control procedures associated with bioassays include the analysis of samples of standards, samples and spiked, unspiked and solvent blanks.

• Information should be provided on the numbers of false-positive and false-negative results for a large set of samples below and above the maximum levels if they are fixed, otherwise, in comparison with the dioxin content as determined by GC–MS. Less than 5% actual false-negative rates could be accepted (EPA method 4025). The acceptable rate of false-positive results is more difficult to determine, as a positive result may be due to a true Ah-receptor agonist that does not belong to classes of target compounds (in most cases, dioxins and coplanar PCBs). However, under normal circumstances, the overall rate of positive samples should be low enough to make use of a screening tool advantageous.
• Positive results should be confirmed by high-resolution GC–high-resolution MS. Samples within a wide range of toxic equivalents should be confirmed (approximately 2–10% of the total samples). Information on correspondence should be made available.
• Standard testing procedures for use on environmental samples should first meet the criteria of national organizations (e.g. EPA method 4425 or 4025; ASTM Guide E1853-98M or APHA standard method 8070) or international standards (Cooke et al., 2000; Behnisch et al., 2001a).

• Every test run in a bioassay must be accompanied by a series of reference concentrations of TCDD or a dioxin/furan mixture for a full dose–response curve with a R₂ > 0.95). However, for screening purposes, an expanded low concentration curve could be used for analysing samples with low concentrations.
• A TCDD reference concentration (about three times the minimal LOD) on a quality control sheet should be used for comparison of the outcome of the bioassay over a constant time, or EC₅₀, maximum response or induction factor. Alternatively, the relative response of a reference sample in comparison with the TCDD calibration curve could be used, as the response of cells may depend on many factors.
• Quality control charts for each type of reference material should be maintained and checked to ensure that the outcome is in accordance with the stated guidelines. The results of screening should correlate by > 95% with the reference method used.
• Particularly for quantitative calculations, the induction of the dilution of sample used must be on the linear portion of the response curve. Samples that give responses above that portion must be diluted and re-tested. Therefore, at least three dilutions should be tested at one time.
• The per cent standard deviation should not be > 10% in triplicate determinations for each sample dilution and should not be > 20% in three independent experiments.
• The LOQ could be set as three times the standard deviation of the solvent blank or of the background response. Another approach is to apply a response that is clearly greater than background (induction factor, five times the solvent blank) to the equation of the day from the calibration curve.
• The final toxic equivalents could be measured from the difference in EC₅₀, EC₂₅ or EC_10–15 between TCDD and the sample or from a fixed effect (for review, see Cooke et al., 2000).

The standard quality criteria requirements for kit-based bioassays could include the following (see e.g. EPA method 4025 or 4035):

• Follow the manufacturer’s instructions for sample preparation and analysis.
• Do not use test kits that are past their expiration date.
• Do not use materials or components designed for use with other kits.
• Use the test kits at the specified storage and operating temperatures.
• An acceptable LOD in immunoassays is 10 times the standard deviation of the blank divided by the slope. The comparable limit for an immunoassay machine would be 10 times the standard deviation of that of a sample with undetectable dioxin/furan concentrations.
• Reference standards should be used in tests by the production facility and/or the user to ensure that the responsiveness to the standard is within an acceptable range.

Essentially, kits consist of screening assays, and many of the same quality criteria guidelines suggested for screening assays should be followed.

The next level of screening should involve more detailed characterization. A positive sample (with toxic equivalents) should be characterized by chemical analysis, involving fractionation of the sample and identification of the responsible coplanar compounds and then analysed in a bioassay. The correspondence should be reported, as discussed above.

Samples for which the results of bioassays for toxic equivalents cannot be explained by chemical characterization should be studied further in vitro or in vivo by other techniques.

There are no specific guidelines for sampling protocols for food samples to be analysed for their dioxin content. Therefore, basic rules for sampling for organic contaminants or pesticides should be followed. The primary requirement is a representative, homogeneous laboratory sample with no secondary contamination.

3.3.1 Personnel

A qualified, authorized person should perform sampling.

3.3.2 Representative sample

Samples must be representative of the lots or sublots from which they are taken. Compliance with maximum levels or action levels should be established on the basis of the concentrations determined in the laboratory sample.

Lots are identifiable quantities of food delivered at one time and determined by the official to have common characteristics, such as origin, variety, type of packaging, packer, consigner or markings. In the case of fish, they should be of comparable size. Sublots are designated parts of a large lot to which the sampling method is applied. Each sublot must be physically separate and identifiable. An incremental sample is a quantity of material taken from a single place in a lot or sublot. As far as possible, incremental samples should be taken at various places distributed throughout the lot or sublot. An aggregate sample is the combined total of all the incremental samples taken from the lot or sublot. It should be at least 1 kg, unless impractical. A laboratory sample for the purposes of enforcement, trade and refereeing should be taken from the homogenized aggregate sample, unless this conflicts with States’ regulations on sampling. The sample used to ensure enforcement should be large enough to allow at least duplicate analysis.

3.3.3 Packaging, transport and storage of aggregate and laboratory samples

Each aggregate and laboratory sample should be placed in a clean, inert container offering adequate protection from contamination, loss of analytes by adsorption to the internal wall of the container or damage in transit. Glassware offers good protection from contamination and can be cleaned easily. Polyethylene and polypropylene containers also provide protection against damage during transit. Containers made from halogenated substances (such as polyvinylchloride) are not considered suitable for this purpose. Although dioxins are chemically stable, samples must be stored and transported in such a way that the food sample does not deteriorate. In particular, the fat content should not be changed, for example by microbiological or enzymatic processes, as the content of the compounds in food of animal origin is generally calculated on a fat basis.

3.3.4 Human milk samples

WHO has recommended addition of K₂Cr₂O₇ tablets to human milk samples during collection of portions and for transport in the third round of studies of exposure, if freezing of the portions cannot be guaranteed. This helps to avoid microbiological deterioration of the samples. If the portions can be frozen immediately after collection and the collected portions can be shipped in a frozen state, addition of K₂Cr₂O₇ tablets is unnecessary.

3.3.5 Sealing and labelling

Each sample taken for official use should be sealed at the place of sampling and identified following States’ regulations. A record must be kept of each sampling, permitting each lot to be identified unambiguously and giving the date and place of sampling, together with any additional information likely to be of assistance to the analyst.

3.3.6 Edible parts

For determination of dioxins in food, only the edible parts are analysed. Vegetables should be washed with water to separate them from adhering soil.

As dioxins are chemically stable, lipophilic substances, no changes in dioxin content with processing would be expected. However, studies have been performed to identify possible changes during processing or preparation of food.

At the end of the 1980s, transfer of PCDDs/PCDFs from paper into milk was found in milk samples packaged in cardboard containers. Large amounts of 2,3,7,8-TCDF and the presence of 1,2,7,7-TCDF and other less important TCDFs gave a characteristic pattern of contamination. The migration of these substances from paper through polyethylene foil into milk was clearly demonstrated. Additionally, carry-over of PCDDs/PCDFs and alkylated chlorodibenzofurans from coffee filter paper, from bleaching of pulp and paper with chlorine, into brewed coffee has been detected. Measures were taken to change the process, and significant reductions in the dioxin content were achieved.

Smoking of meat or fish samples can increase the dioxin and furan content depending on the smoking conditions (Körner & Hagenmaier, 1991; Mayer, 1998; Mayer & Jahr, 1998).

Information on the levels of PCDDs, PCDFs and coplanar PCBs in food relate primarily to uncooked food, although possible changes in PCDD/PCDF congener content and toxic equivalents after cooking have been reported. Broiling of hamburger samples resulted in an approximately 50% decrease in the total toxic equivalents (wet weight) per hamburger, but the decrease appeared to be due solely to the decrease in wet weight associated with loss of water and loss of PCDD/sPCDFs with the fat (Schecter et al., 1996).

In further studies, it was shown that the total toxic equivalents (PCDDs/PCDFs and PCBs) in hamburger, bacon and catfish decreased by an average of 50% as a result of broiling. However, the concentration remained the same in hamburger, increased by 84% in bacon and decreased by 34% in catfish (Schecter et al., 1997). On average, the total measured concentration (pg/kg whole weight) increased by 14% in hamburger and by 29% in bacon and decreased by 33% in catfish (Schecter et al., 1998b).

In a study of the effect of pan-frying beef patties, the PCDD/PCDF concentration was reduced by 40–50% by cooking. Most of the reduction was accounted for by the amount in fat liberated from the patties during cooking. There was nevertheless an overall deficit of 6–14% for each congener (Petroske et al., 1997, 1998). The authors attributed the losses to volatility, degradation and processing, but errors due to analytical imprecision or loss of fat were not considered. The general physical characteristics of these compounds would suggest that volatility and degradation are unlikely reasons for loss during cooking or broiling. Nevertheless, pan frying of ground beef significantly reduces the amounts of PCDD/PCDFs as consumed, provided that the fats and juices are not eaten.

After beef patties were cooked by several methods, each PCDD/PCDF congenr could be completely accounted for on a mass balance basis (within the experimental variability of the method). There was no indication of either loss or formation of PCDDs/PCDFs during the cooking process (Thorpe et al., 1999).

No de-novo synthesis of dioxins was observed after deep frying of scallops of pork covered with egg and crumbs either with or without salt and pepper. The frying temperature was high (180 °C). These results were confirmed by analysis of fats used for deep frying from a large hotel. Only a balance between the dioxin content of the raw scallops and the fat was observed (Schwind & Hecht, 2000).

It is therefore unlikely that PCDDs/PCDFs are formed or lost during usual cooking processes. Changes in dioxin content can be seen on a fresh weight basis owing to changes in fat and water content. As a whole, the dioxin content on a mass balance basis is expected to be constant in meat, fat and juices; however, changes can occur between these phases.

Data were submitted by six countries (Belgium, Canada, Japan, New Zealand, Poland and the USA) and by the Commission of the European Union (2000a) (see Table 12).

In all countries in which substantial numbers of samples have been analysed over time, the concentrations of dioxin in food were decreasing up to the end of the 1990s. In several countries, this decrease was slowed, or even partly reversed in some food categories, due to contamination of feed. In addition, at the end of the 1990s, the measurements initiated at the beginning of the decade did not show the same strong reduction. For the international intake assessment reported here, only data collected after 1995 were considered. These represent 13 589 food products pooled in 1256 samples for PCDDs/PCDFs and 2636 food products pooled in 796 samples for PCBs.

There were large differences in the amount, detail and quality of the data from the participating countries. In particular, the results appeared to have been obtained largely without adequate harmonization of analytical procedures and/or intercalib-ration among the laboratories in different countries. This ultimately affects the comparability of the results. In addition, some of the data may be overestimates of the dioxin concentration, owing to lack of sufficient sensitivity for determination in the analytical laboratories where the assays were performed.

Most data were submitted for PCDDs/PCDFs and PCBs, as a sum of the congeners, weighted by the WHO TEFs. In some cases, an upper-bound approach was used with the LOQ in calculating toxic equivalents for those congeners that were not quantified. In other cases, the concentration of the contaminant may have been underestimated, when a lower-bound approach, with a zero value for those congeners that were not quantified, was used. As the Committee did not have access to the original analytical results, all the concentrations were expressed as the sum of congeners.

As it was impossible to identify analytical results for targeted samples, all the data were considered representative of total contamination of foods. Some data were not used because the mean level of contamination was not quantified.

Belgium: Belgium submitted information about the occurrence of dioxins in pooled milk and cheese samples, using the GEMS/Food electronic submission format. Data on 49 individual samples analysed in March, August and October 2000 were included.

Canada: Canada submitted the results of a total diet study conducted between 1992 and 1995 on dioxins and PCBs in 220 composite foods collected in five cities. All the results submitted were for individual samples. Data from older studies conducted between 1980 and 1989 were also submitted but were excluded from the international intake assessment.

Japan: Japan provided results on the occurrence of PCDDs/PCDFs and PCBs in fish and fish products, meat and meat products, various fruits and vegetables and milk. All the results provided were for individual samples on both a fresh weight and lipid basis, allowing estimation of the distribution of contamination.

New Zealand: New Zealand submitted the results of analyses for 53 foods purchased in 1997 in five cities. The foods were grouped into 22 composites and analysed to determine the concentrations of PCDDs/PCDFs and 23 PCB congeners including non-ortho PCBs.

Poland: Poland submitted the results of determinations of PCDD/PCDF content in Polish products. As there were not enough results to represent the eastern European region, they were not used in the international intake assessment.

USA: A draft report from the EPA was submitted, which contained a large amount of information collected in the USA and other countries. Additional information was provided by the Food and Drug Administration.

Commission of the European Union: A report of Scientific Cooperation (SCOOP) Task Force 3.2.5, entitled Assessment of Dietary Intake of Dioxins and Related PCBs by the Population of European Union, Member States, was submitted (Commission of the EuropeanUnion, 2000a). Ten countries (Belgium, Denmark, Finland, France, Germany, Italy, The Netherlands, Norway, Sweden and the United Kingdom) provided data on the occurrence of PCDDs/PCDFs and coplanar PCBs in food products and human milk. Samples were obtained nationally from rural and industrial sites in the period 1982–99.

The SCOOP database comprises information reviewed previously (IARC, 1997; AEA Technology (1999), but it also contains more recent material resulting from studies conducted up to the end of 1999. In addition, it contains information on coplanar PCBs that was not taken into consideration in the other reports.

For most countries, there were not enough individual data to allow generation of a full curve for the distribution of concentrations, as they were submitted in an aggregated format. As agreed at the FAO/WHO workshop on exposure assessment to contaminants (WHO, 2000), aggregated data were weighted as a function of the number of initial samples. Each result was multiplied by the number of individual samples in the original survey, and the sum of the products was then divided by the total number of individual samples to obtain a weighted mean of the contamination of foods by PCDDs/PCDFs and PCBs.

National data were therefore aggregated by region when sufficient data were available (western Europe, North America, Oceania, Far East; Table 13). The data were not sufficient to permit a realistic estimate of the distribution of contaminants for the rest of the world.

In a second step, a log-normal distribution of contaminants in foods was assumed, and the distribution was modelled from the weighted mean and a geometric standard deviation of 3.0 derived from the data on concentrations, for six broad groups of foods: meat and meat products, eggs, fish and fish products, milk and milk products, vegetable products and fats and oils. The geometric standard deviation was determined after log-transformation of the data by the following formula:

where N is the total number of individual samples, n_i is the number of individual samples in a pooled result x, x_i is the national pooled result, and x is the mean of the national pooled results.

The percentiles of these distribution curves were determined, and the median values (50th percentiles) are presented in Table 13.