The Expert Committee was requested by the Codex Committee on Food Additives and Contaminants at its Thirty-second Session (Codex Alimentarius, 2000) to ‘examine exposure to aflatoxin M₁ and to conduct a quantitative risk assessment’ to compare the consequences of setting the maximum level in milk at 0.05 and 0.5 µg/kg.

Aflatoxins can be produced by three species of Aspergillus—A. flavus, A. parasiticus, and the rare A. nomius—which contaminate plants and plant products. A. flavus produces only B aflatoxins, while the other two species produce both B and G aflatoxins. Aflatoxins M₁ and M₂ are the hydroxylated metabolites of aflatoxins B₁ and B₂ and can be found in milk or milk products obtained from livestock that have ingested contaminated feed. The main sources of aflatoxins in feeds are peanut meal, maize and cottonseed meal.

Aflatoxins were evaluated by the Committee at its thirty-first, forty-sixth, and forty-ninth meetings (Annex 1, references 77, 122, and 131). At its forty-ninth meeting, the Committee considered estimates of the carcinogenic potency of aflatoxins and the potential risks associated with their intake. At that meeting, the Committee reviewed a wide range of studies conducted in animals and humans that provided qualitative and quantitative information on the hepatocarcinogenicity of aflatoxins. The Committee evaluated the potency of these contaminants, linked those potencies to estimates of intake, and discussed the potential impact of hypothetical standards on sample populations and their overall risk. In its evaluation, the Committee stated that the carcinogenic potency of aflatoxin M₁ in sensitive species is about one order of magnitude less than that of aflatoxin B₁. In particular, the Committee noted that the carcinogenic potency of aflatoxin B₁ is substantially higher in carriers of hepatitis B virus (about 0.3 cancers per year per 100 000 persons per ng/kg bw per day), as determined by the presence in serum of the hepatitis B virus surface antigen (HBsAg⁺ individuals), than in HBsAg^– individuals (about 0.01 cancers per year per 100 000 persons per ng/ kg bw per day). Populations with both a high prevalence of HBsAg⁺ and a high aflatoxin intake might benefit from reductions in aflatoxin intake. The Committee also noted that vaccination against hepatitis B virus would reduce the number of carriers of the virus, and thus reduce the potency of the aflatoxins in vaccinated populations, leading to a reduction in the risk for liver cancer.

The Committee at its forty-ninth meeting concluded that changing the hypothetical standard for aflatoxin B1 from 20 µg/kg to 10 µg/kg would not result in any observable difference in rates of liver cancer.

At its present meeting, the Committee reviewed studies published since its forty-ninth meeting, as well as other information, to elucidate further the carcinogenic potencies of aflatoxin M₁ and aflatoxin B₁ and the differences between animal species in their sensitivity to aflatoxins.

The chemical structure of aflatoxin M₁ is shown in Figure 1. Aflatoxin M₁ is the 4-hydroxy derivative of aflatoxin B₁ and is secreted in the milk of mammals that consume aflatoxin B₁. Aflatoxin M₁ (CAS No. 6795-23-9) has a relative molecular mass of 328 Da and has the molecular formula C₁₇H₁₂O₇.

2.2.1 Biotransformation

A complete review of studies of the metabolism of aflatoxins conducted up to 1997 can be found in the report of the monograph on aflatoxins published after the forty-ninth meeting of the Committee (Annex 1, reference 132) and in a book by Eaton and Groopman (1994). The papers highlighted in this section address studies of matabolic differences among species in their sensitivity to aflatoxins, comparisons of the toxicity of aflatoxin M₁ and aflatoxin B₁, studies of the differences in the carcinogenic potency of aflatoxin M₁ and aflatoxin B₁, and papers published since the last report of the Committee.

The metabolism of aflatoxin B₁ and the extent to which it binds to cell macro-molecules were compared in liver slices from humans and rats, as rats are more sensitive to the carcinogenicity of aflatoxin B₁. Liver slices were prepared from three human liver samples and incubated with [³H]aflatoxin B₁ at 0.5 µmol/L for 2 h. The rates of formation of oxidative metabolites and of specific binding to cell macromolecules showed significant interindividual variation. The rates of oxidative metabolism of aflatoxin B₁ to aflatoxin Q₁, aflatoxin P₁, and aflatoxin M₁ in the human liver samples were similar to those previously observed in rat liver slices. No aflatoxin B₁–glutathione conjugate formation was detected in the human liver samples, and there was much less specific binding of aflatoxin B₁ to cell macromolecules in the human than in the rat liver slices. For example, the level of binding between aflatoxin B₁ and DNA ranged from 3 to 26% of that in control rats. These results suggest that humans do not form as much aflatoxin B₁ 8,9-epoxide as rats, but they also suggest that humans do not have glutathione-S-transferase (GST) isozymes with high specific activity towards this epoxide. Significant individual differences in aflatoxin B₁ metabolism and binding suggest the presence of genetic and/or environmental factors that may result in large differences in susceptibility.

Aflatoxin M₁ is usually considered to be a detoxication by-product of aflatoxin B₁; it is also the metabolite present in the milk of nursing women who eat foods containing the toxin. Aflatoxin B₁ epoxide has been shown to exist as two stereoisomers—endo- and exo-epoxides—the latter being the DNA-reactive form, and a similar situation may apply to aflatoxin M₁ epoxide. In a study of the metabolism of aflatoxins M₁ and B₁ in vitro in human liver microsomes, they had a very limited capacity to catalyse epoxidation of aflatoxin M₁. The small amount of aflatoxin M₁ dihydrodiol formed from the epoxide also appeared to have a lower capacity to induce microsomal protein than did aflatoxin B₁ dihydrodiol. GST catalysed conjugation of the epoxides of both aflatoxins with glutathione; GST activity was present in mouse cytosol but not in the human liver fraction. The authors concluded that the difference between the genotoxic potency of the two toxins in vivo correlates with their mutagenicity in vitro, metabolic activation and DNA binding (Neal et al., 1998). In rats, however, activation of aflatoxin M₁ to the epoxide does not appear to be essential for its acute toxicity. Experiments in human cell lines indicated that cytochrome P450 (CYP) enzymes are involved in the cytotoxicity of aflatoxin B₁ but not of aflatoxin M₁. Studies of the toxicity of aflatoxin M₁ in human lymphoblastoid cell lines expressing or not expressing human CYP enzymes showed a direct effect in the absence of metabolic activation, in contrast to aflatoxin B₁. Aflatoxin M₁ is therefore not strictly a detoxication product of aflatoxin B₁ in biological responses in which cytotoxicity plays a significant role, such as immunotoxicity (Heinonen et al., 1996).

In studies of species sensitivity to the carcinogenicity of aflatoxin B₁, mice were resistant because they constitutively express an alpha-GST, which is strongly active against aflatoxin B₁ 8,9-epoxide, whereas rats, which do not express such a GST, were sensitive. Human hepatic alpha-class GSTs have little capacity to detoxify aflatoxin B₁ 8,9-epoxide. The nonhuman primate Macaca fascicularis showed significant constitutive hepatic GST activity towards aflatoxin B₁ 8,9-epoxide. GSTs were purified from liver tissue from this species and characterized, and GST cDNAs were cloned by reverse transcriptase-coupled polymerase chain reaction (PCR). A protein, GSHA-GST, was purified by glutathione agarose affinity chromatography, which had stronger aflatoxin B₁ 8,9-epoxide-conjugating activity than other GST-containing peaks. The GSHA-GST was shown to belong to the µ class. The authors then showed that two distinct µ-class GST cDNAs have 97% and 98% homology with the human µ-class GSTs hGSTM4 and hGSTM2, respectively. µ-Class GSTs appear to be responsible for most of the conjugating activity of aflatoxin B₁ 8,9-epoxide in the liver of M. fascicularis. None of the known human µ-class GSTs acts preferentially on the ultimate genotoxic aflatoxin B₁ metabolite exo-aflatoxin B₁ 8,9-epoxide, but large interindividual differences in the expression of GST isoforms have been shown in various tissues, and few human livers have been evaluated. The authors concluded that identification of a potential human homologue of GSHA-GST would be relevant to the design of chemointervention strategies to reduce aflatoxin B₁-induced liver cancer in highly exposed populations. Nevertheless, induction of known human GSTs with little or no activity towards the epoxide of aflatoxin B₁ might be ineffective in reducing the genotoxicity of aflatoxin B₁ (Wang et al., 2000).

The extreme sensitivity of turkeys to aflatoxin B₁ was studied by measuring microsomal activation of aflatoxin B₁ to the 8,9-epoxide, the putative toxic intermediate, cytosolic GST-mediated detoxication of aflatoxin B₁ 8,9-epoxide, and hepatic phase I and phase II enzyme activities in 3-week-old male Oorlop turkeys. Liver microsomes prepared from these turkeys activated aflatoxin B₁ in vitro with an apparent K_m of 110 µmol/L and a V_max of 1.25 nmol/mg per min. The involvement of CYP 1A2 and, to a lesser extent, 3A4 in the activation of aflatoxin B₁ was assessed with specific mammalian CYP inhibitors. The possible presence of avian orthologues of these CYPs was indicated by activity towards ethoxyresorufin and nifedipine, as well as by western immunoblotting with antibodies to human CYPs. GST-mediated conjugation of 1-chloro-2,4-dinitrobenzene and 3,4-dichloronitrobenzene was demonstrated in cytosol prepared from the turkey livers, but the rate was much lower than that observed in other species. The presence of alpha- and µ-class GSTs and another aflatoxin B₁ detoxifying enzyme, aflatoxin B₁ aldehyde reductase, was shown by western immunoblotting. Quinone reductase activity was also present in the cytosol. Furthermore, the cytosol showed no measurable GST-mediated detoxication of microsomally activated aflatoxin B₁. Thus, turkeys are deficient in the most crucial aflatoxin B₁ detoxication pathway. The authors concluded that the extreme sensitivity of this species to aflatoxin B₁ is due to a combination of efficient aflatoxin B₁ activation and deficient detoxication by phase II enzymes such as GSTs (Klein et al., 2000).

The carcinogenicity of aflatoxin B₁, aflatoxicol (aflatoxin L), aflatoxin M₁, and aflatoxicol M₁ (aflatoxin LM₁) was compared in terms of their binding to target organ DNA in rainbow trout. Tritiated compounds were synthesized, dose–response curves for DNA binding were established, and liver DNA binding indices were calculated for the four aflatoxins after a 2-week dietary intake by trout fry. The adduct levels increased linearly with dietary concentration, with relative DNA binding indices of 21, 20, 2.4, and 2.2 x 10³ (pmol/mg of DNA)/(pmol/g of diet) for aflatoxin M₁, aflatoxin L, aflatoxin M₁, and aflatoxin LM₁, respectively.

In a similar protocol, over 7200 trout fry with an average initial body weight of 1.2 g were used to establish full carcinogen dose–response curves for each aflatoxin and an estimate of the DNA binding index after a single dose. Since trout are very sensitive, < 180 µg of each aflatoxin were required. Data analysed on logit incidence versus L_n dose coordinates generated four curves, which were modelled as parallel in slope over most or all the doses studied. In this analysis, the relative tumorigenic potencies were 1.0 for aflatoxin B₁, 0.94 for aflatoxin L, 0.086 for aflatoxin M₁, and 0.041 for aflatoxin LM₁. When the data were plotted as logit incidence versus L_n adducts (effective dose received), dose–response relationships were found for all aflatoxin adducts, indicating that they are equally tumorigenic, except for aflatoxin LM₁, which was two to three times less potent. Differences in the tumorigenicity of the four aflatoxins are largely or entirely accounted for by differences in uptake and metabolism leading to DNA adduction, rather than to any inherent difference in tumour initiating potency per DNA adduct (Bailey et al., 1998).

Since most people are exposed to carcinogens in food intermittently, the effects of intermittent intake of aflatoxin B₁ on hepatic and testicular glutathione was studied in male Fischer 344 rats fed diets containing aflatoxin B₁ at a concentration of 0, 0.01, 0.04, 0.4, or 1.6 ppm at 4-week intervals up to 20 weeks. The control animals were fed an aflatoxin B₁-free NIH-31 diet. Rats eating diets containing 0.01 ppm aflatoxin B₁ did not show induction of hepatic or testicular GST activity, but intermittent intake of concentrations of 0.04–1.6 ppm significantly increased GST activity. The increase in enzyme activity was proportional to the dose and the length of intake of aflatoxin B₁ (Sahu et al., 2000).

2.2.2 Effects of oltipraz and ethoxyquin

Oltipraz is a competitive and perhaps irreversible inhibitor of CYP 1A2 and 3A4, and addition of oltipraz to rat liver microsomes or to cultured human hepatocytes blocks the oxidative metabolism of aflatoxin B₁ to its 8,9-oxide and to the hydroxylated derivative aflatoxin M₁. Inhibition of aflatoxin M₁ excretion in urine during dietary intervention with oltipraz was examined in male Fischer 344 rats before, during, and after transient intervention. The animals were housed individually in glass metabolism cages and given 25 µg of [³H]aflatoxin B₁ by gavage daily for 28 consecutive days. On days 6–16, half the rats were fed a diet supplemented with 0.075% oltipraz. Sequential 24-h urine samples were collected, and a subset was analysed for aflatoxin B₁ metabolites. Aflatoxin M₁ was the main metabolite in all samples, accounting for 2–6% of the administered dose. Its excretion was greatly reduced (by 77%) when oltipraz was added to the diet, but rapidly returned to control levels after cessation of the intervention. No such inhibition of aflatoxin M₁ excretion was seen in animals given oltipraz by gavage 24 h before dosing with aflatoxin B₁. These findings are consistent with the view that oltipraz or a short-lived metabolite inhibits CYP 1A2 in vivo (Scholl et al., 1996).

The effects of cancer preventive agents on the metabolism of aflatoxin B was examined in non-human primates in a study designed to complement a human chemointervention trial, in which oltipraz, an antischistosomal drug approved by the Food and Drug Administration of the USA, was used to modulate the metabolism of aflatoxin B in a human population naturally exposed to this toxin in the diet. This study is discussed in section 5.3. The hepatic metabolism of aflatoxin B₁ was studied in macaque (M. nemestrina) and marmoset (Callithrix jacchus) monkeys and compared with that in humans. Thus, four adult male marmosets were used as controls, four were given oltipraz, and three received ethoxyquin. At time 0, each animal received a single dose of [³H]aflatoxin B at 100 µg/kg bw (0.5 mCi/kg) by gavage, and blood samples were drawn at 0, 2, 24, and 48 h. On days 16–28, the treated animals received the synthetic dithiolthione oltipraz at 18 mg/kg bw per day or the antioxidant ethoxyquin at 30 mg/kg bw per day in their diet, whereas the control animals received the vehicles only. On day 26, each animal received a second dose of [³H]aflatoxin B by gavage, and blood samples were drawn at 0, 2, 24, and 48 h. On day 28, the animals were killed, their livers were excised, and microsomal and cytosolic fractions were prepared. For comparison, livers from adult male macaques were obtained from the University of Washington Primate Center (USA). Twelve human liver samples were also obtained from the University of Washington, and the microsomal fractions were pooled. Microsomal oxidation of aflatoxin B₁ and GST activity were measured, DNA adducts were isolated, and [³H]aflatoxin B-derived radioactivity in DNA fractions was quantified. [³H]Aflatoxin B₁–albumin adducts in serum were determined.

The oxidative metabolism of aflatoxin B was quantitatively similar in the two monkey species and in humans. In contrast to macaques, humans and marmosets lacked aflatoxin B glutathione conjugating activity. As the metabolism of aflatoxin B in marmosets resembled that in humans more closely than that in macaques, the focus of the study was on marmosets. Both oltipraz and ethoxyquin induced aflatoxin B₁–glutathione conjugating activity in the livers of some but not all marmosets. Oltipraz inhibited CYP-mediated activation of aflatoxin B to the ultimate carcinogenic metabolite, aflatoxin B₁ 8,9-epoxide, in vitro by up to 51%, and animals that received oltipraz in vivo showed a significant reduction (average, 53%) in aflatoxin B–DNA adduct formation in comparison with control animals.

The authors interpreted these findings as indicating that oltipraz and ethoxyquin induce modest aflatoxin B glutathione conjugating activity in the livers of some marmosets, most of the activity (about 70%) being directed against the exo isomer of aflatoxin B₁ 8,9-epoxide, which is by far the most potent DNA-reactive metabolite. Other workers have demonstrated aflatoxin B–mercapturic acid in the urine of marmosets exposed to aflatoxin B₁. The hepatic GST activity towards aflatoxin B₁ 8,9-epoxide shown in this study in non-human primates was two orders of magnitude lower than that in mice, which are resistant to the carcinogenic effects of aflatoxin B₁. The authors offered two explanations for the presence of DNA adducts and the decrease in steady state of exo-aflatoxin B₁ 8,9-epoxide shown in both treated groups: GST-mediated detoxication of exo-aflatoxin B₁ 8,9-epoxide and inhibition of the CYP(s) that form it. Consistent with the results of the chemointervention trial in a human population (section 5.3), administration of oltipraz and ethoxyquin would be likely to attenuate the adverse effects of aflatoxin B in primates (Bammler et al., 2000).

In a study of the inhibition of aflatoxin M₁ production by bovine hepatocytes after intervention with oltipraz and another dithiolthione, 4-methyl-5-(2-pyrazinyl)-1,2-dithiole-3-thione, oltipraz inhibited the metabolism of aflatoxin B₁, as neither aflatoxin M₁ nor aflatoxin B dihydrodiol (the second metabolite found in bovine hepatocytes) was formed. The second dithiolthione did not significantly inhibit aflatoxin B₁ metabolism. The authors suggested that the inhibition of aflatoxin B₁ metabolism by oltipraz was due to inhibition of the activity of several CYP enzymes. Although the authors proposed that oltipraz could be administered to dairy cows that had accidentally received aflatoxin B₁ in their feed, unmetabolized aflatoxin B₁ would still reach the systemic circulation. These results obtained in vitro should be confirmed in vivo (Kuilman et al., 2000).

The roles of coumarin, benzyl isothiocyanate, and indole-3-carbinol, which are present in vegetable-enriched diets and are believed to protect against malignant disease, in regulating GST and aldo-keto reductase activity were examined in rat liver. The drugs butylated hydroxyanisole, diethyl maleate, ethoxyquin, beta-naphtho-flavone, oltipraz, phenobarbital, and trans-stilbene oxide were also investigated. In a complicated protocol, summarized briefly here, groups of three male and three female 10-week-old male and female Fischer 344 rats were given diets containing 0.75% butylated hydroxyanisole, 0.5% benzyl isothiocyanate, 0.5% coumarin, 0.5% ethoxyquin, 0.5% indole-3-carbinol, or 0.075% oltipraz for 2 weeks. Diethyl maleate at 0.5% was administered for 5 days in the food. trans-Stilbene oxide at 400 mg/kg bw was dissolved in 0.5 ml of peanut oil before daily intraperitoneal administration on 3 consecutive days, and beta-naphthoflavone at 200 mg/kg bw was dissolved in phosphate-buffered saline before daily intraperitoneal administration for 7 consecutive days. Phenobarbital was added to the drinking-water at a concentration of 0.1% for 7 days.

For the short-term intervention study of the effect of coumarin on the development of preneoplastic foci, six groups of eight 12-week-old male Fischer 344 rats were given one of the following experimental diets for 13 weeks: RM1 control maintenance diet throughout, 0.05% coumarin in RM1 diet throughout, 2 mg/kg of diet aflatoxin B₁ in RM1 diet for 6 weeks followed by RM1 control diet for 7 weeks, 2 mg/kg of diet aflatoxin B₁ in RM1 diet throughout, 0.05% coumarin in RM₁ diet for 2 weeks followed by aflatoxin B₁ at 2 mg/kg of RM1 diet containing 0.05% coumarin for 11 weeks, or 2 mg/kg of diet aflatoxin B₁ in RM1 diet for 6 weeks followed by aflatoxin B₁ in RM1 diet containing 0.05% coumarin for 7 weeks.

In a long-term intervention study with coumarin to study tumour formation, six groups of eight 12-week-old male Fischer 344 rats were given the same diets described above but were placed on diets containing coumarin and aflatoxin B₁ for 24 weeks before being transferred to a control diet from week 25 until termination of the experiment at week 50. The animals were killed with CO₂, and tissues were removed immediately. Microsomal and cytosolic fractions were prepared from fresh liver or from samples snap-frozen in liquid nitrogen.

Under these conditions, coumarin was the main inducer of aflatoxin B₁ aldehyde reductase and the aflatoxin-conjugating µ-class GST A5 subunit in rat liver, increasing the concentrations of these proteins by 25–35 times. Coumarin caused similar increases in the concentration of pi-class GST P1 subunit and NAD(P)H:quinone oxidoreductase in rat liver.

To assess the biological significance of enzyme induction by dietary coumarin, two intervention studies were performed, in which the ability of benzopyrone to inhibit aflatoxin B₁-initiated preneoplastic nodules (at 13 weeks) or aflatoxin B₁-initiated liver tumours (at 50 weeks) was investigated. Pretreatment with coumarin for 2 weeks before administration of aflatoxin B₁ and continued treatment during exposure to the carcinogen for a further 11 weeks protected the animals completely from development of hepatic preneoplastic lesions by 13 weeks. Treatment with coumarin in a longer-term dietary intervention, before and during exposure to aflatoxin B₁ for 24 weeks, resulted in significant inhibition of the number and size of tumours that developed by 50 weeks. The authors concluded that consumption of a coumarin-containing diet provides substantial protection against the initiation of hepato-carcinogenesis by aflatoxin B₁ in rats. The other phytochemicals and synthetic drugs tested in this study induced different zone- and sex-specific enzymes in the liver. The complexity of gene–environment interactions is emphasized by the fact that certain inducing agents can cause nuclear translocation of drug-metabolizing enzymes (Kelly et al., 2000).

2.3.1 Acute toxicity

The acute toxicity of aflatoxin M₁ was reviewed by van Egmond (1994) and is summarized only briefly here. In the 1960s, newly hatched ducklings were shown by several investigators to be extremely sensitive to both aflatoxin B₁ and aflatoxin M₁, with LD₅₀ values of 12–16 µg per bird. Histopathological examination showed liver lesions similar to those caused by aflatoxin B₁ and necrosis of the renal tubules. Milk naturally contaminated with aflatoxin M₁ produced fewer lesions than artificially contaminated milk, however, suggesting differences in the bioavailability of naturally and artificially occurring aflatoxin M₁. Studies on the acute toxicity of aflatoxins in 1-day-old ducklings suggest that aflatoxin M₁ and aflatoxin B₁ act by a similar mechanism in causing acute toxicity and subcellular alterations, such as changes in liver parenchymal cells, dissociation of ribosomes from the rough endoplasmic reticulum, and proliferation of the smooth endoplasmic reticulum, and that only the naturally occurring isomer of each aflatoxin is biologically active.

2.3.2 Long-term studies of toxicity and carcinogenicity

van Egmond (1994) also summarized the results of long-term studies of toxicity. In a study by Sinnhuber in 1974, rainbow trout received diets containing aflatoxin B₁ at 4 µg/kg or aflatoxin M₁ at 0, 4, 16, 32, or 64 µg/kg for 12 months and then received a control diet. Selected groups were held for 20 months to determine the effect of maturation on tumour development, and some were fed aflatoxin M₁ at 20 µg/kg of diet for 5–30 days to determine the effect of limited oral intake of this toxin. Female trout with aflatoxin M₁-induced hepatomas had a significantly higher mortality rate at maturation (16–20 months) than males. The trout receiving aflatoxin M₁ at 20 µg/kg of diet had a 3–12% incidence of hepatoma within 12 months. The author concluded that aflatoxin M₁ is a potent liver carcinogen but less potent than aflatoxin B₁.

Canton et al. in 1975 fed rainbow trout diets containing aflatoxin M₁ at 0, 5.9, or 27 µg/kg and aflatoxin B₁ at 5.8 µg/kg for 16 months. The fish were killed after 5, 9, and 12 months. Degeneration of the liver was seen in all three groups and in the control group, but no tumours or preneoplastic changes were found. At 15 months, however, the survivors fed the diet containing 5.8% aflatoxin B₁ had a 13% incidence of hepatocellular carcinoma and a 23% incidence of hyperplastic nodules, and those fed the diet with 27.3 µg/kg aflatoxin M₁ had a 2% incidence of hepatocellular carcinoma and a 6% incidence of hyperplastic nodules. The investigators concluded that differences in trout strain could have contributed to the differences between their results and those of Sinnhuber, but that aflatoxin M₁ is less carcinogenic in trout than aflatoxin B₁.

van Egmond (1994) summarized two further studies in rats. In a study in 1974, weanling Fischer rats were given 25 µg/day of synthetic aflatoxin M₁ by intubation on 5 days/week for 8 consecutive weeks. A second group of rats was given natural aflatoxin B₁ at the same concentration and under similar conditions. A control group was included. Only one rat (3%) given aflatoxin M₁ developed a hepatocellular carcinoma, whereas 28% had liver lesions (preneoplastic lesions). All rats receiving aflatoxin B₁ developed tumours, whereas controls showed no significant liver lesions. The carcinogenic potency of aflatoxin M₁ was concluded to be much lower than that of aflatoxin B₁.

In a second study, groups of Fischer rats were maintained on diets containing natural aflatoxin M₁ at 0, 0.5, 5, or 50 µg/kg and were killed between 18 and 22 months. Hepatocellular carcinomas were detected in 5% and neoplastic nodules in 15% of rats fed diets containing aflatoxin M₁ at 50 µg/kg between 19 and 20 months. No nodules or carcinomas were observed at the lower dose. Of rats fed the diet containing aflatoxin B₁ at 50 µg/kg, 95% developed hepatocellular carcinomas. Only a few rats at 50 µg/kg of aflatoxin M₁ developed intestinal carcinomas. The authors suggested that the greater polarity of aflatoxin M₁ than aflatoxin B₁ might be associated with the higher incidence of intestinal tumours. It was concluded that aflatoxin M₁ was a hepatic carcinogen, but with a potency 2–10% that of aflatoxin B₁ (Cullen et al., 1987, as described by van Egmond, 1994). This study is the one usually cited in comparisons of the carcinogenicity of aflatoxin B₁ and aflatoxin M. van Egmond (1994) concluded that the toxicity of aflatoxin M₁ is similar to or slightly lower than that of aflatoxin B₁ in rats and ducklings, and the carcinogenicity of aflatoxin M₁ is probably one to two orders of magnitude lower than that of aflatoxin B₁ (see Table 1).

2.3.3 Genotoxicity

The potency of aflatoxin B₁ and aflatoxin M₁ in inducing DNA damage and genotoxicity was tested in Drosophila melanogaster in vivo in the mei-9a mei-41D5 DNA repair test and the mwh/flr3 wing spot test, respectively. In the repair test, larval stock consisting of meiotic recombination-deficient double-mutant mei-9a mei-41D5 males and repair-proficient females was exposed to the test agent, and preferential killing of the mutant larvae was taken as evidence of DNA damage. Aflatoxin M₁ was found to be a DNA-damaging agent, with an activity about one-third that of aflatoxin B₁. In the wing spot test, in which larval flies trans-heterozygous for the somatic cell markers mwh and flr3 were treated and the wings were inspected at adulthood for spots manifesting the phenotypes of the marker, the genotoxicity of aflatoxin M₁ and aflatoxin B₁was similar. The authors concluded that aflatoxin M₁ is genotoxic in mammalian systems in vivo (Shibahara et al., 1995).

2.3.4 Special studies

The tree shrew (Tupaia belangeri chinensis) is unique in that it can be infected with human hepatitis B virus (HBV) and is susceptible to aflatoxin B₁-induced liver cancer; a synergistic interaction between HBV and aflatoxin B₁ for liver cancer has been observed. In studies in which the tree shrew model was used to evaluate experimental chemoprevention strategies for populations at high risk for liver cancer, two groups of tree shrews were fed milk containing aflatoxin B₁ at a concentration providing a dose of 400 µg/kg bw per day for 4 weeks. One week before administration of aflatoxin B₁, one group also received oltipraz at 0.5 mmol/kg bw per day orally for 5 weeks. Samples of 1 ml of blood and 24-h urine were obtained from each animal at weekly intervals. Aflatoxin–albumin adducts in serum were identified by radioimmunological assay, and aflatoxin–N7-guanine adducts in urine were measured by high-performance liquid chromatography (HPLC). The concentration of aflatoxin–albumin adducts increased rapidly over 2 weeks, to reach a plateau at 20 pmol/mg of protein, and decreased after cessation of exposure to aflatoxin B₁. Oltipraz significantly attenuated the overall burden of aflatoxin–albumin adducts throughout exposure, with a median reduction of 80%. As measured in a single cross-sectional analysis at the end of treatment with aflatoxin B₁, oltipraz decreased the urinary aflatoxin–N7-guanine content by 93%. The authors concluded that oltipraz reduces risk biomarkers for aflatoxin B₁ in the tree shrew, as it does in rodents and humans, and established a rationale to evaluate cancer chemoprevention by oltipraz in tree shrews infected with human HBV and exposed to aflatoxin B₁.

The authors recalled that reductions of comparable magnitude in both aflatoxin–albumin adducts in serum and aflatoxin–N7-guanine adducts in urine were found in rats pretreated with oltipraz and exposed to aflatoxin B₁. Tree shrews appear to be less susceptible to hepatocarcinogenesis than rats. The tree shrew model is useful and may allow determination of whether agents such as oltipraz sustain their chemopreventive effect against aflatoxin in the presence of chronic infection with HBV (Li et al., 2000).

The effects of methyl deficiency and dietary restriction on hepatic-cell proliferation and telomerase activity were studied in 5-week-old male Fischer 344 rats pretreated with aflatoxin B₁ at 25 µg/rat per day by gavage on 5 days/week for 3 weeks or given solvent (100 µg of 75% dimethyl sulfoxide). The rats were then separated into groups fed a methyl-sufficient or -deficient diet ad libitum or with dietary restriction. When the rats were 15, 20, and 32 weeks of age, hepatic-cell proliferation, telomerase activity, and the number of GST-placental form (P)-positive foci were determined. Dietary restriction reduced hepatic-cell proliferation, while the methyl-deficient diet and aflatoxin B₁ pretreatment increased cell proliferation. Telomerase activity was decreased by dietary restriction and increased by the methyl-deficient diet and aflatoxin B₁ pretreatment. The same trend was observed for GST-P⁺ foci in aflatoxin B₁-pretreated rats: methyl deficiency increased the number of foci, and dietary restriction decreased the number. These results are consistent with a role of telomerase in hepatocarcinogenesis, although the origin of the cells giving rise to the increase in telomerase activity was not determined (Chou et al., 2000).

In a study of the effect of ascorbic acid on the toxicity of aflatoxin B₁, young guinea-pigs were either fed diets containing 0 or 25 mg/day of ascorbic acid or were given 300 mg/day by gavage for 21 days and the LD₅₀ dose of aflatoxin B₁ on day 22. Seven of 10 animals fed no ascorbic acid died within 73 h of administration of aflatoxin B₁, and their livers showed massive regional necrosis and multilobular degeneration. None of the animals given 25 mg/day ascorbic acid died, but their livers showed changes similar to those seen in the group that received no ascorbic acid. The activities of serum alanine and aspartate aminotransferases were elevated.

None of the animals given 300 mg/day of ascorbic acid died or had pathological changes in the liver, and their alanine and aspartate aminotransferase activities were unaffected. Production of aflatoxin M₁ by liver microsomes tended to be higher than that in the other two groups. Three animals receiving 300 mg/day of ascorbic acid were given a second intraperitoneal LD₅₀ dose of aflatoxin B₁ 1 month after the first. One animal died, and the livers of all animals showed centrilobular degeneration and moderate necrosis in scattered hepatocytes. Hepatic microsomal CYP and cytosolic GST activities and aflatoxin M₁ production were drastically reduced, and the activities of alanine and aspartate aminotransferase were increased. The results indicate that intake of 300 mg of ascorbic acid virtually protected the animals from the acute toxicity of aflatoxin B₁ given by gavage but not when administered as a second dose intraperitoneally (Netke et al., 1997).

In a study of the effects of carotenoids on the initiation of liver carcinogenesis by aflatoxin B₁, male weanling rats were fed beta-carotene, beta-apo-8´-carotenal, cantha-xanthin, astaxanthin, or lycopene at 300 mg/kg of diet; an excess of vitamin A (21 000 retinol equivalents per kg of diet); or 3-methylcholanthrene at 6 x 20 mg/kg bw intraperitoneally before and during treatment with aflatoxin B₁ at 2 x 1 mg/kg bw. The rats were then treated with 2-acetylaminofluorene and partial hepatectomy, and GST-P⁺ liver foci were detected and quantified. Aflatoxin B₁-induced hepatic DNA damage was evaluated as single-strand breaks and binding of [³H]aflatoxin B₁ to liver DNA and plasma albumin in vivo. Modulation of aflatoxin B₁ metabolism by carotenoids or by 3-methylcholanthrene was investigated by incubation in vitro of [¹⁴C]aflatoxin B₁ with liver microsomes from rats that had been fed carotenoids or treated with 3-methylcholanthrene; the metabolites formed were analysed by HPLC.

Neither lycopene nor an excess of vitamin A had any effect, but beta-carotene, beta-apo-8´-carotenal, astaxanthin, and canthaxanthin decreased the metabolism of aflatoxin B₁ to aflatoxin M₁, a less genotoxic metabolite. The authors concluded that these carotenoids exert their protective effect by deviating aflatoxin B₁ metabolism towards detoxication pathways. In contrast, beta-carotene did not protect hepatic DNA from aflatoxin B₁-induced alterations and affected the metabolism of aflatoxin B₁ to only a minor degree. Its protective effect against the initiation of liver preneoplastic foci by aflatoxin B₁ appears to be mediated by other mechanisms (Gradelet et al., 1998).

The hepatotoxicity of aflatoxin B₁ is augmented by bacterial endotoxin lipopoly-saccharide in rats. At intraperitoneal doses > 1 mg/kg bw, aflatoxin B₁ caused pronounced injury to the periportal regions of the liver. Male Sprague-Dawley rats were given aflatoxin B₁ at 1 mg/kg bw or the vehicle, 0.5% dimethyl sulfoxide and saline, and then Escherichia coli lipopolysaccharide (7.4 x 10⁶ enzyme units per kg) or its saline vehicle 4 h later. Liver injury was assessed 6, 12, 24, 48, 72, or 96 h after administration of aflatoxin B₁. Histological examination of liver sections and measurements of alanine and aspartate aminotransferase activity in serum were used to evaluate hepatic parenchymal-cell injury. Biliary-tract alterations were evaluated as increased concentration of serum bile acids and activities of gamma-glutamyl-transferase, alkaline phosphatase, and 5’-nucleotidase in serum.

No or little injury was seen in rats treated with aflatoxin B₁ or lipopolysaccharide alone, but hepatic parenchymal-cell injury was pronounced by 24 h in the group treated with aflatoxin B₁ and lipopolysaccharide, returning to control values by 72 h. The injury began in the periportal region and spread mid-zonally with time. Changes in serum markers indicative of biliary-tract alterations were evident by 12 h, but the values had returned to control levels by 72 h. The nature of the hepatic lesions suggested that lipopolysaccharide potentiated the effects of aflatoxin B₁ on both parenchymal and bile-duct epithelial cells.

The authors suggested that the results of this study might partly explain the severity of human cases of acute aflatoxicosis. In addition, persons with hepatitis who have an inflammatory response may be predisposed to the carcinogenic effects of aflatoxin B₁, as the results of this study suggest that inflammation accompanied by hepatic parenchymal-cell hyperplasia might contribute epigenetically to aflatoxin B-induced carcinogenesis by promoting tumour formation (Barton et al., 2000).

Aflatoxin M₁, like other aflatoxins, is produced by fungi that grow naturally on plants in the field or on stored feeds. Aflatoxins are among the most toxic of the known mycotoxins and have been implicated in the deaths of humans and animals that have consumed mouldy food . While the liver is the target organ for aflatoxicosis, aflatoxins are also found in other animal tissues and products, such as meat, milk, and eggs. As mature animals modify and eliminate toxins effectively, however, the main concern is long-term intake of low concentrations of these toxins, which can lead to cancer and immunosuppression. Although intake of low doses of aflatoxins may not cause death or tissue damage, it may severely affect the cost-effectiveness of animal production.

Sensitivity to aflatoxins varies from one species to another, and, within the same species, the severity of toxicity depends on dose, duration of intake, age, and breed of the animals, and their dietary protein content. The results of toxicological studies in domestic animals are given in Table 2.

In general, ingestion of aflatoxin results in a variety of clinical signs which depend on the amount consumed and the species and age of the animal. Aflatoxin may make an animal more susceptible to infectious diseases by impairing its immune system or potentiating a bacterial infection. Symptoms of secondary infection may obscure the symptoms of aflatoxicosis. Intake of aflatoxins during gestation may affect offspring as well as adults (Miller & Wilson, 1994).

The Food and Drug Administration (USA) set a tolerance limit of 20 µg/kg for aflatoxins in maize and 0.5 µg/kg of aflatoxin M₁ in milk. The latter can be achieved by consuming a diet contaminated with < 30 µg/kg (Shane, 1994). The carry-over of aflatoxin from animal feed to milk and tissue is discussed in section 8.1.

Liver cancer has been related to dietary intake of aflatoxins. The most recent epidemiological studies tend to indicate that individuals who are carriers of persistent viral infection with HBV and who are exposed to aflatoxin in their diets are at increased risk for progression to liver cancer as compared with HBV carriers who are not exposed to aflatoxins. No similar interaction has been reported with chronic infection with hepatitis C virus (HCV). The epidemiological studies to date have focused on aflatoxin B₁; ingestion of aflatoxin M₁ with milk and milk products has not been directly related to liver cancer. Some of the epidemiological observations that implicate aflatoxins in the etiology of liver cancer derive from observations of unusual clusters of disease or unexplained trends in incidence. Additional indications are provided by the results of case–control and cohort studies based on adequate techniques and comprehensive evaluation of the risk factors in the etiology of liver cancer.

Increasing trends have been reported in the rates of hospitalization, incidence, and mortality attributable to liver cancer in the black and white populations of both sexes in the USA. The age-specific curves showed a shift towards liver cancer among persons aged 40–60. The authors discuss in detail and quite convincingly the factors that interfere in analyses of time trends for liver cancer, including the widespread introduction of new diagnostic means (ultrasound, a-fetoprotein analysis), improved registration practices (histological confirmation and coding), and the quality of analysis (trend and birth–cohort analyses). The authors attribute the trend to intravenous drug abuse in the relevant generations, HCV being the predominant causative factor (El Serag & Mason, 1999).

HCV was also shown to be related to the increased rate of death from liver cancer in Japan after a vaccination campaign against tuberculosis under non-sterile conditions (Okuda, 1991).

Also in Japan, a report on trends in death from liver cancer showed that alcohol was the agent primarily responsible. Birth cohort analyses showed small effects. The effect of alcohol may have been overestimated because the incidence of liver cancer was used as a surrogate measure of alcohol consumption by women, who are assumed to have low consumption of alcohol. Other surrogate measures used were the incidence of oesophageal cancer and the mortality rate from cirrhosis (Makimoto & Higuchi, 1999).

The role of chronic infection with HBV and HCV in the etiology of liver cancer is well established. Several epidemiological studies have examined the association between seropositivity for HBsAg and the risk for liver cancer. The risk estimates ranged from 3 to 30 in case–control studies and from 5.3 to 148 in cohort studies (IARC, 1994). A meta-analysis of studies published before 1998 gave an estimated relative risk of 17 for persons with antibodies to HCV who are HBsAg^– (Donato et al., 1998). Table 3 shows current estimates of the attributable fractions for the main risk factors associated with liver cancer. Worldwide, 52% of liver cancer cases (230 000) have been attributed to chronic HBV infection, with 19 000 in developed countries and 210 000 in developing countries. The fraction of liver cancer cases attributable to HCV infection is 110 000 (25% of the world total), with 17 000 cases in developed countries and 93 000 in developing countries.

The presence of HBV DNA or HCV RNA serum and liver tumour tissue from patients with liver cancer, mostly in European countries, who were seronegative for antibodies to both viruses, was investigated in a collaborative multicentre study. Of the specimens, 33% contained HBV DNA and 7% contained HCV RNA (Brechot et al., 1998). The results have been confirmed. The trend suggests that, in countries where HBV is common, the presence of HBV DNA among HBsAg^– patients with liver cancer is higher than the 33% found in Europe. These findings reinforce the strong relationship between HBV and HCV viral infections and liver cancer and suggest that the attributable fractions shown in Table 3 may be underestimates.

2.5.1 Correlation studies

In a review of several studies in China (some of which were evaluated by the Committeeat its forty-ninth meeting), in which aflatoxin–albumin adducts were measured, a correlation was found between death from primary liver cancer and aflatoxin B₁–albumin in serum from persons in Fusui but not in those from Shanghai. In Fusui County, primary liver cancer was correlated to intake of aflatoxin B₁ but not aflatoxin, M₁ and the decreasing trend in the aflatoxin–albumin adducts over time in Fusui were attributed to improved agricultural practices (Yu et al., 1998).

The mutation induced by aflatoxin B₁ in exon 3 of the human HPRT gene in B lymphoblasts is a GC to TA transversion at base 209, occurring in 17% of aflatoxin B₁-induced mutants. In an analysis of the HPRT mutation frequency in an area with heavy intake of aflatoxin B₁, the residents of Qidong County, China, were studied to determine the combined contributions of aflatoxins and other risk factors to the high incidence rate of liver cancer in the region. The study cohort comprised 42 men and 65 women aged 40–65. Blood samples were analysed for mean aflatoxin B₁ in albumin, HPRT mutations by a T-cell clonal assay, HBsAg status, serum alanine aminotransferase activity, leukocyte count, haemoglobin concentration, and platelet count. Subjects were categorized as having a low or a high intake of aflatoxin B₁ and were dichotomized around the population mean of aflatoxin–albumin adducts.

A major assumption in this study was that an individual’s aflatoxin B₁ content was representative of his or her intake of aflatoxin throughout life, even though aflatoxin B₁–albumin adducts indicate recent intake. The amounts of aflatoxin B₁ measured were comparable with those in previous year-long studies in this population. The typical mutation frequency in the HPRT gene in normal, healthy adults is 5–8 x 10^–6 per cell, whereas the frequency in this population was 26 x 10^–6. Thus, the population was exposed to environmental agents that damage DNA. The authors concluded that the aflatoxin-induced DNA damage in T lymphocytes, assessed as the validated marker, albumin adducts, led to an increased mutation frequency, reflected as the increase in HPRT gene mutations (Wang et al., 1999a).

The limitation of the study is that HBsAg seropositivity was presumed to indicate the presence of HBV; however, epidemiological studies that rely on HBsAg status instead of detection of HBV DNA (Brechot et al., 1998; Bosch et al., 1999) systematically underestimate the risk due to HBV. Furthermore, HCV status was not measured.

2.5.2 Case–control and cohort studies

Studies on intake of aflatoxins and liver cancer published in 1997–2000 incorporated biomarkers of intake of aflatoxin in order to compare series of cases and controls. Some of these studies had the advantage of being nested in cohort studies, thus including data from biological specimens collected some time before the occurrence of liver cancer.

A small case–control study in the Sudan showed a relationship between grain storage conditions and the aflatoxin contamination of peanuts, and some association between storage conditions and the occurrence of liver cancer (Omer et al., 1998).

A 10-year follow-up study for hepatocellular carcinoma in Qidong, China, was reported in which 145 carriers of HBV provided eight monthly urine samples which were tested for aflatoxin M₁ by a sensitive assay (3.6 ng/L). At the beginning of the study, 54% of the subjects had aflatoxin M₁ in their urine; 22 subsequently developed hepatocellular carcinoma. The predictors of liver cancer among HBV carriers were found to be the presence of aflatoxin M₁ in urine, antibodies to HCV, and a family history of hepatocellular carcinoma. The estimated relative risk associated with aflatoxin M₁ was 3.6 (95% confidence interval, 1.3–9.9) . The authors compared this estimates with that for a cohort in Shanghai with exposure to both HBV and aflatoxins (relative risk, 59) and found no statistically significant difference, mainly because of the small number of cases in both studies. All four patients with hepatocellular carcinoma who had aflatoxin M₁ in their urine and who were tested for mutations in the P53 oncogeneshowed the missense mutation in codon 249. The authors concluded that aflatoxin is a substantial risk factor for progression to hepatocellular carcinoma among carriers of HBV. The (unstable) estimated attributable fraction was 0.55 (0.09–0.94) (Sun et al., 1999).

In a prospective follow-up study of 737 HBV carriers and 699 with no HBV, aflatoxin–albumin adducts were measured in 30 HBsAg⁺ patients with liver cancer and 150 controls (HBV status unclear). A significantly larger proportion of patients had adducts (odds ratio [OR], 3.5), and they had a significantly higher overall level of adducts than controls. The authors concluded that aflatoxins are a significant co-factor with HBV in the induction of primary liver cancer (Lu et al., 1998).

Forty-three patients with hepatocellular carcinoma in Taiwan were compared with 86 matched controls for the urinary concentration of aflatoxin metabolites and aflatoxin B₁–albumin adducts in specimens taken in 1988–92 and for GST activity. All but one of the patients were HBsAg⁺ and the other had antibodies to HCV. The levels of biomarkers of aflatoxin in urine differed somewhat between cases and controls. A trend to increased risk with urinary aflatoxin M₁ was reported, but the numbers were too small to ensure proper power. A high risk was found for persons with detectable levels of aflatoxin–albumin adducts and aflatoxin B₁–N7 guanine adducts (OR, 10; 1.6–61). The authors concluded that aflatoxin is a significant factor for primary liver cancer in HBV carriers and that there may be an interaction between the GSTM1 genotype and intake of aflatoxin B₁ (Yu et al., 1997a).

A small case–control study was conducted among black Africans to determine the effect of iron overload and other environmental factors on the risk for hepato-cellular carcinoma. The OR associated with HBV seropositivity in 24 cases and 48 hospital controls was 33 (7.2–150) for HCV infection, 6.4 (0.30–130) for alcohol consumption, 2.0 (0.50–8.2) for iron overload, and 11 (1.5–77) for aflatoxin–albumin adducts. There was no association with hepatocellular carcinoma (median prevalence of adducts, 7.3 in cases and 22 in controls; OR not reported) (Mandishona et al., 1998).

Thus, three studies from Asia reported increased risks for persons infected with HBV and with either aflatoxin M₁ metabolites in urine or aflatoxin–albumin adducts in serum. The ORs reported ranged from 3 to 10. The evidence is not entirely consistent, and the study from southern Africa found no significant association berween the presence of aflatoxin–albumin adducts in serum and liver cancer.

Other cohort studies gave conflicting results with regard to the role of aflatoxin in the etiology of liver cancer. A follow-up study was conducted to estimate the risk for primary liver cancer among male HBV carriers in areas with different intakes of aflatoxin: in Senegal, China, and persons of Asian origin resident in the USA. The cohorts were selected to examine why the estimated risk for liver cancer in some areas of China was significantly higher (two- to threefold) than that on the west coast of Africa. The prevalence of HBsAg was only moderately higher in Senegal (20% versus 16%), and the expected intake of aflatoxin was higher in the African setting. In an analysis of differences in host response to the viral infection in these populations, viral replication (HBV DNA detected by Southern blotting) was 25–30% in HBsAg carriers of all age groups in China, whereas in Senegal a strong decay in HBV DNA rates was seen with age, from 14% in persons aged 20–29 to 3% in those aged 30–49 and undetectable in persons > 50. Asian–American HBsAg carriers also showed a strong decline in HBV DNA with age, from 37% in the 20–29 age group to 5% in those over 50. Prolonged viral replication at a high titre correlated with several parameters of liver damage and may be a determinant in the high rate of mother-to-child transmission of HBV and in the incidence of liver cancer in Chinese populations .

In the first report of this study, the authors showed that the risk of an HBV carrier for progression to hepatocellular carcinoma was lower in Senegal (high risk for exposure to both HBV and aflatoxin) than in China (high risk for exposure to HBV but lower risk for intake of aflatoxin). This was contrary to expectations if a strong interaction between HBV and aflatoxins is the central determinant of liver cancer in these areas. In fact, active DNA replication throughout life seems to explain the higher progression rate in China and may be a consequence of the high rate of mother-to-child transmission of HBV and the high incidence of liver cancer in Chinese populations (Evans et al., 1998).

2.5.3 Metabolic epidemiological studies

The metabolism of aflatoxins is not yet fully understood. It has been hypothesized that metabolic polymorphisms of the genes that regulate the metabolism of aflatoxins could explain the established interspecies differences in susceptibility to aflatoxin-induced carcinogenicity and the largely hypothetical differences in susceptibility among human groups. Some epidemiological studies have addressed the environmental factors that may modulate the natural history of aflatoxins and biomarkers of aflatoxins under various conditions of exposure.

The determinants of aflatoxin–albumin adducts in blood were investigated in 357 persons, including 181 chronic carriers of HBV, in The Gambia. Several environmental factors (season, place of residence, HBV status) and aspects of the metabolism of aflatoxin (the GST genotypes M1, T1, P1 and epoxide hydrolase) were recorded, and the ratio of 6beta-hydroxycortisol:cortisol as a marker of CYP 3A4 activity was measured in urine. The major determinants of the amounts of aflatoxin in blood were place of residence and season. The mean adduct levels were higher in persons without HBV infection and the GSTM1 null genotype. The authors concluded that environmental factors leading to food contamination are better determinants of intake than metabolic measures and are more amenable to intervention (Wild et al., 2000).

Understanding the natural history of the biomarkers used in epidemiological studies is an absolute requirement for proper interpretation of much of the available literature. In a study in Taiwan, aflatoxin B₁–N7-guanine adducts were measured in urine as a function of the hormonal and nutritional parameters that may affect aflatoxin adduct formation. In a cross-sectional study of 42 male HBV carriers and 43 HBV-free men, adduct formation was detected in 42%. Significant determinants of adduct formation were HBV status, with higher levels in HBsAg carriers, and plasma concentrations of cholesterol, alpha-tocopherol and alpha- and beta-carotene. The associations were significant and dose-dependent. Lycopene concentrations were inversely related to adduct formation. This study is significant in that it indicates some of the environmental and host determinants of adduct formation for use in etiological studies (Yu et al., 1997b).

If HBV status is directly related to adduct formation, case–control studies will systematically show that the presence of adducts in urine is a risk factor for hepatocellular carcinoma (Sohn et al., 2000). Nutritional determinants of adduct formation, if validated, should be treated as confounders in epidemiological studies in which urinary adducts are used as a biomarker. The roles of folate and other nutrients in the natural history of HBV infection and liver cancer have been confirmed in a number of studies (e.g. McGlynn et al., 1999), and the role of nutrients in the mutagenicity and carcinogenicity of aflatoxins is corroborated by evidence from experiments in rats (Soni et al., 1997) and in studies summarized in section 2.3.4.

GST expression was found to be inversely related to the HBV status of patients with normal livers, suggesting that viral replication decreases the ability of liver cells to detoxify liver carcinogens such as aflatoxin efficiently. Tissue from liver tumours had less GST activity, and subjects with the null GSTM1 genotype had less GST alpha- and µ-isoenzymes, with some overexpression of pi. These results suggest that GST expression should be treated as a confounder in epidemiological studies in which urinary adducts are used as a biomarker (Zhou et al., 1997).

In a comparison of the sensitivity to mutagens of 28 cases of hepatocellular carcinoma and 110 controls, on the basis of the count of chromatid breaks induced by bleomycin or benzo[a]pyrene diol epoxide, the OR was 36. The tests show defects in predisposition to chromosome breakage or the capacity to repair chromatid breaks or both. The results suggest that individual susceptibility can be important in determining the outcome in exposed individuals. No epidemiological studies have been reported in which host factors determined by these methods were adjusted for (Wu et al., 1998).

2.5.4 Intervention studies with vaccination against hepatitis B virus

A study in Taiwan evaluated by the Committee at its forty-ninth meeting showed that, in an area hyperendemic for infection and with moderate-to-high intake of aflatoxins, immunization against HBV had reduced the rate of HBV carriage in 6-year-old children from about 10% in 1981–86 to 0.8–0.9% in the period 1990–94. A more recent report noted that 15–20% of the population of Taiwan were estimated to be HBV carriers in the early 1980s. A programme of mass vaccination against HBV was launched in 1982, and, since 1986, all newborns and, progressively, preschool children, primary-school children, adolescents, young adults, and others have also been vaccinated. The coverage of newborns is over 90%, and 79% of pregnant women are screened for HBsAg. The proportion of babies born to highly infectious mothers who also became carriers decreased from 86–96% to 12–14%. The average annual incidence of hepatocellular carcinoma in children aged 6–14 decreased significantly from 0.7 per 100 000 in 1981–86 to 0.36 per 100 000 in 1990–94; and the annual incidence of hepatocellular carcinoma in children aged 6–9 declined from 0.52 per 100 000 for those born in 1974–84 to 0.13 per 100 000 in those born in 1986–88. Thus, the mass vaccination programme has been highly effective in controlling chronic HBV infection and in preventing liver cancer in Taiwan. If a vaccination coverage rate of 90% of all newborns against HBV can be maintained, the carriage rate in Taiwan can be expected to decline to 0.1% by 2010. The cost of the programme has been about US$ 100 million (Huang & Lin, 2000).

Recent reports from Taiwan and from other areas where massive HBV vaccination campaigns have been conducted have shown the presence of HBV mutants in the surface gene which induces chronic carriage among immunized children. The prognosis of these infected children is uncertain, but the observation should be considered in evaluating the occurrence of liver cancer in HBV-vaccinated populations (Hsu et al., 1999).

A study of vaccination against HBV was reported from the Republic of Korea, where the prevalence of HBV infection is among the highest in the world. In this prospective cohort study, 370 285 men over the age of 30 who were clinically free of liver disease and had not been vaccinated against HBV at the time of enrolment were included. About 5% of the cohort were HBsAg⁺, 78 094 had antibodies to the HBV surface marker, and 273 277 were negative for both. About 13% of the men had been vaccinated against HBV in 1985. Cases of liver cancer were ascertained by record linkage and from medical records covering 1986–89. A multivariate log-linear model was used to test for statistical significance and to estimate relative risks. The follow-up period represented 1 404 566 person-years (average, 3 years and 10 months), and 302 cases were ascertained, to give an overall incidence rate of liver cancer of 22 per 100 000 person-years. The relative risk for primary liver cancer was18% (95% confidence interval, 14–23) among chronically infected men, 0.34 (0.19–0.60) among unvaccinated infected men, and 0.58 (0.31–1.1) in the vaccinated group. The study suggests that vaccination against HBV, even in adulthood, reduces the risk for liver cancer (Lee et al., 1998).

A report on the results of vaccination programmes in China and The Gambia showed that the vaccine must be given as early as possible in life: vaccination within 48 h of birth reduced carriage by at least 70%. In China, some 40% of carriers of HBV were infected by perinatal transmission from their mothers. The effectiveness of vaccination has been reported to be 70% in some areas of China and as much as 90% in others. In Africa, introduction of the vaccine into the routine programme for infant vaccination reduced the carriage rate by 94%. In The Gambia, protection was shown to be maintained up to the age of 9 years, which is well past the age at which the risk of becoming a carrier is high; thus, these children effectively have lifelong protection against HBV-associated liver cancer (Wild & Hall, 2000).

Programmes to reduce the burden of liver cancer in developing countries should therefore give priority to HBV vaccination and to the prevention of HCV contamination. This implies reinforcing the control of blood and blood products and the use of sterile medical equipment. HBV carriers may benefit from reductions in intake of aflatoxins in their diets, and this may also offer some protection to HCV carriers. However, a reduction in intake of aflatoxin B₁ or a reduction in the concentration of aflatoxin M₁ in milk or milk products is unlikely to result in an observable reduction in the rate of liver cancer in most developed countries. In these populations, alcohol consumption may account for most cases of liver cancer without viral markers.

2.5.5 Intervention studies with oltipraz

In spite of the effectiveness of hepatitis B vaccination in preventing chronic HBsAg carrier status in unexposed newborns, infants, and adults, a substantial number of persons (some 300 million worldwide) are HBsAg carriers. No effective treatment has been developed for these persons (Torresi & Locarnini, 2000).

Epidemiological studies suggest that dietary intake by chronic HBV carriers of aflatoxins may increase the rate of progression to hepatocellular carcinoma (Qian et al., 1994). Thus, it has been suggested that oltipraz, a drug that modifies the metabolism of aflatoxin and has a number of other biological properties (reviewed by Kensler et al., 1999) might be used as a chemopreventive agent. The experimental basis for this proposal is the demonstration of remarkable anticancer activity against aflatoxin B₁-induced hepatocarcinogenesis in rats (see section 2.3.4). Oltipraz has also been evaluated as a chemopreventive agent for cancers of the colon, liver, bladder, and skin (reviewed by Kensler & Helzlsouer, 1985; Kensler et al., 1999).

In rats, continuous administration of oltipraz significantly reduced the formation of aflatoxin–albumin adducts and the occurrence of liver neoplasms (Kensler et al., 1997). Oltipraz affected the life cycle of HBV in vitro by blocking transcription in 2.2.15 cells, resulting in dose-related inhibition of HBV replication, perhaps mediated through induction of wild-type p53 (Chi et al., 1998).

Oltipraz has been tested in in phase I/II trials in China. The results of pilot studies for these trials showed reasonably good compliance with the regimen and mild toxicity, with no observed interaction with the HBV status of the individual (Jacobson et al., 1997). The results of another pilot trial showed that low daily doses of oltipraz induced phase-2 conjugation of aflatoxin, as measured by an increase in the urinary excretion of aflatoxin mercapturic acid, with no reduction in the concentration of aflatoxin M₁. Intermittent high doses of oltipraz decreased the phase-1 metabolism of aflatoxin, leading to a significant reduction in excretion of aflatoxin M₁, indicating that the metabolic pathways of aflatoxins in humans can be strongly modified by oltipraz (Wang et al., 1999b). The long-term effects of this chemopreventive treatment remain to be established.

2.5.6 Other studies

Various biomarkers were studied in 23 cases of liver cancer in the USA. HBV markers including HBV DNA were found in 13 cases, HCV antibodies in sera in 5/22, overexpression of P53 in tissue in 5/23, and mutations in codon 249 in 0/5. Surprisingly, aflatoxin B₁–DNA adducts were found in liver tumour tissue in 3/19 cases and aflatoxin B₁–lysine adducts in sera in 5/5, none of which had concurrent overexpression of P53 or mutations at codon 249. Few cases were available for each test, and only the abstract has been published; however, the presence of markers of aflatoxin in cases of liver cancer in the USA is interesting because intake of aflatoxin in that country is expected to be low. HBV or HCV infection would also be expected to be infrequent, depending on the subpopulation sampled (Hoque et al., 1999).

Aflatoxin–albumin adducts were also identified in serum from 104 volunteers in the United Kingdom. There was no direct correlation with a particular food (Turner et al., 1998).

2.5.7 Studies of P53 as a marker of intake of aflatoxin

Mutations of P53 are a relevant marker in the molecular epidemiology of liver cancer, as some 20% of cases show mutations of this oncogene. Moreover, a mutation at the third base of codon 249 (a GC to TA transversion leading to a change from arginine to serine) has been described in geographical correlation studies of intake of aflatoxin B₁.

The IARC and other databases on p53 clearly describe the ‘hot spot’ at 249 as the predominant mutation in liver cancer. The IARC database is heavily biased by publication and reporting selection, but 35–40% of cases of liver cancer in areas where there is high intake of aflatoxins show the presence of the 249 mutation, whereas there is a much lower prevalence (0–2%) in areas where there is low intake (Soussi et al., 2000).

The most recent reviews of ‘molecular fingerprints’ of carcinogens appear to converge in accepting a few for which some specificity can be claimed: GC to TA transversions in lung cancer associated with smoking; GC to TA transversions in codon 249 in liver cancer associated with aflatoxin B₁; and CC:GG to TT:AA transversions in skin cancer associated with exposure to ultra-violet light (Hainaut & Vahakangas, 1997; Wang & Groopman, 1999).

Since the last evaluation by the Committee, additional studies have shown that a mutation in codon 249 of P53 is found regularly in a proportion of cases of liver cancer in certain countries and not in others. Several studies showed that these mutations are poorly correlated with another marker of aflatoxin intake, the presence of aflatoxin B₁ adducts, in hepatocellular carcinoma tissue (Hsie et al., 1995; Soini et al., 1996; Lunn et al., 1997). Experimental data have also shown that the mutation can be induced in hepatocytes exposed to aflatoxin B₁; aflatoxin metabolites bind to the third base in codon 249; and 249 ser p53 expression inhibits apoptosis and p53-mediated transcription and enhances liver cell growth in vitro (reviewed by Hussain & Harris, 2000).

Most of the studies reported below involved few cases, and even fewer cases with P53 mutations, and many of the comparisons and ORs calculated from them are therefore quite unstable. Many of the reports focus on mutations and not on the full spectrum of genetic alterations in P53 that characterizes hepatocellular carcinoma. Most of these mutations have been identified in individuals who are also HBV carriers.

Of 21 cases of hepatocellular carcinoma in India, three had mutations in P53 (two in 249 GT and one in a 250 CT transition). In investigations for HBV status, 59% of the cases were shown to have HBV DNA by dot blotting, 90% to have HBV DNA by polymer chain reaction, and 71% to be HBsAg⁺ by enzyme-linked immunosorbent assay (ELISA). The report indicated that intake of aflatoxin was common in that part of India (Katiyar et al., 2000).

In a report on 24 cases of liver cancer in Shanghai and Qidong, China, all specimens had integrated HBV DNA, and 63% had the null GSTM1 genotype; 95% had alterations in P53: 12 had mutations in P53, and 13 had overexpression. Loss of heterozygosity at 4q was found in 50%, at 1p in 46%, at 16q in 42%, and at 13q in 38%. Mutation at codon 249 was found in seven cases from Qidong (all those with a P53 mutation) and in three of five from Shanghai (Rashid et al., 1999).

Of 30 cases of hepatocellular carcinoma from Guangxi, China, an area of high risk for HBV and exposure to aflatoxin, 90% were HBsAg⁺, and 43% showed P53 expression and a linear response with the stage of tumour (Qin et al., 1997).

Seven of 21 samples of tissue from patients with hepatocellular carcinoma in Tongan, China, had point mutations at codon 249 resulting in a G to T transversion. Only one of the patients was HBV-negative (method not stated). The authors also reported another case of HBV-negative hepatocellular carcinoma with a mutation at codon 249 of P53 (Yang et al., 1997).

In a study in Taiwan of 110 cases of liver cancer and 37 controls, HBV status was assessed by assay for HBsAg, intake of aflatoxin by aflatoxin B₁ adducts in liver tissue, P53 status by immunohistochemistry, and DNA mutations by single-stranded conformation polymorphism and sequencing. The main findings were elevated risks associated with HBsAg seropositivity (OR, 8.4) and aflatoxin B₁ adducts (OR, 3.9), with an OR of 68 for both. P53 mutations were found in 29% of cases, and mutations at codon 249 in 13%. The presence of aflatoxin B₁ adducts in liver tissue was related to the presence of P53 protein and DNA mutations (borderline significance). Mutations in codon 249 were found only in HBsAg⁺ subjects, suggesting that HBV is involved in the selection of these mutations. Because liver tissue was required for these comparisons, the controls were patients with liver or biliary-tract conditions, including hepatic metastases from other primary cancers (Lunn et al., 1997).

In a small correlation study, the P53 mutation patterns in 31 cases of hepato-cellular carcinoma from northern and southern Jiang-Su Province in China were compared. Mutations in codon 249 were found in 9/16 cases in the area with high intake of aflatoxin and 1/15 in the area with lower intake (Shimizu et al., 1999).

In a study in Spain, 120 paraffin blocks from hepatocellular carcinoma cases were studied by single-stranded conformation polymorphism and sequencing techniques. No mutation was found in P53, although P53 overexpression was found in 14 cases. The authors concluded that P53 mutations are not common in hepatocellular carcinoma induction or promotion in Spain (Boix-Ferrero et al., 1999).

An investigation was conducted of circulating DNA for P53 codon 249 mutations in a series of 53 hepatocellular carcinoma cases in The Gambia and in 13 patients with liver cirrhosis. There were 53 controls and a second control group of 60 French patients with a variety of liver conditions. The relevant mutation was found in 19 cases of hepatocellular carcinoma, two patients with cirrhosis and three controls. The OR for hepatocellular carcinoma was 16 (3–90). None of the patients in France had the 249 mutation (Kirk et al., 2000).

In a series of 62 hepatocellular carcinoma cases in Taiwan, P53 mutations were investigated by single-stranded conformation polymorphism and sequencing; 37 of the patients were HBsAg⁺ and 25 HBsAg^–. Twenty patients had mutations, which were widely distributed along exons 5–8. Four patients, all HBsAg carriers, had a G to T mutation at codon 249 (Sheu, 1997).

A cluster of mutations was found at position 220 in P53 in patients with genetic haemochromatosis and liver cancer. Mutations in codon 249 exon 7 A/T were observed in one case. Although anecdotal, these observations suggest that a mutation at this locus may also be acquired in other contexts (Vautier et al., 1999).

A recent, unpublished meta-analysis on the relationship between HBV, aflatoxin, and mutation at codon 249 of P53 showed that the geographical relationship between intake of aflatoxins (broadly classified into three levels) and P53 mutations (any spot) was strongly correlated. The correlation was due almost entirely to the G to T mutation, but a significant (albeit unstable) trend with mutations at other codons remained. There was no indication that the presence of the 249 mutation varied with HBV status, although one study showed a significant interaction (Stern et al., 2001, personal communication).

Table 4 summarizes the results of studies on P53 mutations in cases of liver cancer. Intake of aflatoxins is expressed crudely in relation to the geographical source of the specimens, thus ignoring local and individual variation. Although P53 mutations occur at both the hot spot (249 G to T) and other spots, mutations at codon 249 predominated (92% versus 2%) in relation to the geographical classification. The presence of the mutation was affected only moderately by the concurrent presence of HBV.

There is growing consensus that the mutational spectra of P53, and in particular codon 249, is a relevant biomarker of intake of aflatoxin B₁ in relation to hepatocellular carcinoma. In order to make full use of this biomarker, the natural history of the mutation should be characterized, and its relationship to the dose of aflatoxin B₁ and to better validated biomarkers, such as aflatoxin B₁–albumin adducts in urine, should be established.

In epidemiological studies, the finding of this mutation in specimens of hepatocellular carcinoma can be interpreted as reflecting the involvement of aflatoxin B₁, but its absence, particularly in cancers in people living in areas of heavy intake of aflatoxin B₁ should be interpreted with caution, as aflatoxin B₁ might induce hepatocellular carcinoma by mechanisms other than DNA damage.

Early detection of aflatoxin M₁ and removal of small lots of contaminated milk can prevent contamination of much larger volumes. Screening methods are particularly useful if they can be carried out quickly, easily, and economically. They should allow detection of concentrations of aflatoxin M₁ in milk as low as those detected by the ultimate quantitative methods (see section 3.2). Theoretically, screening methods should never give false-negative results.

The screening tests used for aflatoxin M₁ in milk and milk products are usually immunochemical. For aflatoxin M₁, both radioimmunoassays and enzyme immunoassays have been developed (Frémy & Chu, 1989), although enzyme immunoassays are used more often.

The protocol for radioimmunoassay of aflatoxin M₁ usually includes simultaneous incubation of a test solution containing an unknown amount of aflatoxin M₁ (or a standard solution of a known amount of aflatoxin M₁ in phosphate buffer) with a constant amount of labelled aflatoxin and its specific antibody. Free and bound labelled aflatoxin are then separated, and the radiolabel on aflatoxin is determined. The aflatoxin M₁ concentration of the test solution is determined by comparing the results to a standard curve, which is established by plotting the ratio of bound and the initial (total) amount of labelled aflatoxin multiplied by 100 (% binding) versus the concentration of aflatoxin M₁ standard (Chu, 1984). Radioimmunoassay has found little use in routine investigations of aflatoxin M₁ in milk. A recent exception is an investigation of samples of milk in Thailand (Saitanu, 1997). There have been no collaborative studies on radioimmunoassays for aflatoxin M₁.

Most enzyme immunoassays for aflatoxin M₁ are heterogeneous, with separation of the immunocomplex and the unreacted material. One of the commonest heterogeneous enzyme immunoassays, the ELISA, is generally used to determine aflatoxin M₁. Several direct competitive ELISAs for aflatoxin M₁ are available commercially. The 96-well microtitre plate assay is most commonly used for quantitative measurements. ELISAs for aflatoxin M₁ are usually designed for rapid screening. For legal purposes, positive results in an ELISA require confirmation by an accepted reference method.

One successful use of a direct competitive ELISA, in which the results were confirmed by a validated HPLC method, involved ELISA for the determination of aflatoxin M₁ in pasteurized milk, infant formula, powdered milk, and yoghurt (Kim et al., 2000). A valuable addition was comparison of the ELISA with the validated HPLC method of Ferguson-Foos & Warren (1984; see also section 3.2). Kim et al. concluded that the results for aflatoxin M₁ obtained with ELISA were similar to those obtained by HPLC. In a study in which ELISA and HPLC with immunoaffinity purification were compared, the latter technique was found to be superior (Biancardi, 1997). An interesting development is a portable field test involving a patented, membrane-based flow-through enzyme immunoassay (Sibanda et al., 1999), which can be carried out on farms. The kit comprises a nylon membrane spotted with anti-mouse antibodies, a plastic snap-fit device, absorbent cotton wool, mouse monoclonal antibodies against aflatoxin M₁, and aflatoxin B₁–horse radish peroxidase conjugate. Clean-up is done on an immunoaffinity column (see section 3.2). No collaborative studies have been carried out, at least not under the auspices of international organizations, of ELISAs for aflatoxin M₁.

Techniques other than immunochemical procedures can, in principle, be used for rapid analysis of milk and milk products for aflatoxin M₁. One such technique involves electrochemical flow injection monitoring on filter-supported bilayer lipid membranes (Andreou & Nikolelis, 1998). The method has not been formally validated.

It has been recommended that extensive collaborative studies be conducted to validate the performance characteristics of immunoassays, including their reproducibility and repeatability, accuracy, sensitivity, limits of detection, specificity, and selectivity (Frémy & Chu, 1989). This recommendation was echoed in 2000 (van Egmond, 2000), as no such studies have been carried out by AOAC International, the organization responsible for collaborative studies of methods of analysis (International Union of Pure and Applied Chemistry, 1989). The International Dairy Federation (1999) has produced a guideline document that outlines the parameters necessary for the evaluation and validation of competitive enzyme immunoassays for quantitative determination of aflatoxin M₁ in milk and milk products.

Quantitative analytical methods for aflatoxin M₁ usually follow the general pattern for mycotoxin assays, i.e. extraction, clean-up, concentration, separation, detection, and quantification. A homogeneous sample of aflatoxin M₁ in milk is easily obtained, because the toxin is distributed evenly in fluid milk. The initial problem encountered in analysing milk is the extraction step. Because milk is a complex natural product, aflatoxin M₁ is not easily extracted and purified for final assay, owing (partly) to adsorption of aflatoxin M₁ to casein proteins. A process is required to separate aflatoxin M₁ from milk easily, efficiently, and economically. Once purified extracts are obtained, the concentration of aflatoxin M₁ can be determined in one of several ways. Most quantitative methods involve thin-layer chromatography (TLC) or HPLC. Aflatoxin M₁ is a weakly polar component and is extractable with solvents such as methanol, acetone, chloroform, or combinations of these solvents with water. In practice, the choice of solvent depends on the clean-up and the separation procedure.

The quantitative methods that have been developed and validated for aflatoxin M₁ in milk and milk products were originally designed to analyse milk powder. Milk was spray-dried or lyophilized to preserve its shelf life and to reduce sample bulk. Various mixtures of methanol and water (Masri et al., 1968, 1969a; Fehr et al., 1971), acetone and water and acetone, chloroform, and water (Purchase & Steyn, 1967) were used to extract aflatoxin M₁ from milk powder.

In the first effective method for the determination of aflatoxin M₁ in fluid milk, methanol and water were used as the extraction solvents (Jacobson et al., 1971). This method was modified by others (McKinney, 1972; Stubblefield & Shannon, 1974a), leading to a method suitable for collaborative studies. The method of Stubblefield & Shannon (1974a) involved extraction with acetone and water, precipitation with lead acetate solution to deproteinize the milk, and a defatting step with hexane. TLC with fluorescence detection was used for ultimate separation, detection, and quantification. The collaborative study was successful (Stubblefield & Shannon, 1974b), and the method became an official method for aflatoxin M₁ of the AOAC and IUPAC ( AOAC official method 974.17, surplussed in 1993; see Table 5 for performance characteristics).

In another method, liquid milk was partitioned with chloroform in a separating funnel and cleaned-up over a small silica gel column. Final separation was by TLC with fluorescence detection, and aflatoxin M₁ spots were quantified by visual or densitometric estimation (Stubblefield, 1979). This method was further modified to allow determination of aflatoxin M₁ in cheese, in which two-dimensional TLC was used to improve separation of the aflatoxin M₁ spots from the background. The method was evaluated in an AOAC/IUPAC collaborative study (Stubblefield et al., 1980) and became an official AOAC method for aflatoxin M₁ in milk and cheese (AOAC official method 980.21; see Table 5 for performance characteristics).

With advances in HPLC methods in the 1980s, laboratories moved away from TLC to HPLC determination. In addition, factory-prepared solid-phase extraction columns were introduced for the purification of milk extracts. A method that successfully combined these two developments was that of Ferguson-Foos & Warren (1984), originally developed for normal-phase HPLC. The method was modified for reversed-phase HPLC, with the preparation of trifluoroacetic acid derivatives of aflatoxins M₁ and M₂, and evaluated in an AOAC collaborative study (Stubblefield & Kwolek, 1986). The method became AOAC official method 986.16. for aflatoxin M₁ in fluid milk (see Table 5 for performance characteristics).

A more recent advance in quantitative extraction of aflatoxin M1 and subsequent clean-up is use of immunoaffinity cartridges. These columns are composed of monoclonal antibodies specific for aflatoxin M₁, which are immobilized on Sepharose® and packed into small cartridges (see Figure 2). The first published method for aflatoxin M₁ with immunoaffinity columns was that of Mortimer et al. (1987).

A milk sample containing aflatoxin M₁ is loaded onto the affinity gel column, and the antigen aflatoxin M₁ is selectively complexed by the specific antibodies on the solid support into an antibody–antigen complex. The column is then washed with water to remove all other matrix components of the sample. Aflatoxin M₁ is eluted from the column with a small volume of pure acetonitrile, and the eluate is concentrated and analysed by HPLC coupled with fluorescence detection. The method can be applied to whole milk, skimmed milk, and low-fat milk.

Modifications of the immunoaffinity-based methods for aflatoxin M₁ were subsequently published and studied collaboratively under the auspicies of the International Dairy Federation (Tuinstra et al., 1993) and AOAC International (Dragacci et al., 2001) by groups of mainly European laboratories that could determine aflatoxin M₁ in milk at concentrations < 0.05 µg/L. The collaborative study of Tuinstra et al. led to International Dairy Federation Standard 171 (see Table 5 for performance characteristics). Another collaborative study resulted in approval as AOAC method 2000.08 (Dragacci et al., 2001; see Table 5 for performance characteristics).

The combination of immunoaffinity clean-up and liquid chromatography offers the best means for efficient purification and precise determination of low concen-trations of aflatoxin M₁. The approach is widely used in some parts of the world where low limits for aflatoxin M₁ in milk are in force, e.g. the European Union, where the method is successfully practised by the National Reference Laboratories for Milk and Milk Products (Dragacci et al., 2001). The method may, however, be too expensive for routine use in developing countries. An interesting lower-cost alternative is a method that combines immunoaffinity clean-up with TLC and a computer-based, low-cost densitometer. This method is being validated for detection of low concentrations of aflatoxin M₁ in a formal collaborative study.

Ensuring that regulations on limits for aflatoxin M₁ in milk and milk products are met requires validated methods of analysis. Several organizations (AOAC International, IUPAC, the International Dairy Federation, and the European Standardization Committee [Comité Européen de Normalisation]) have tested methods of analysis for aflatoxin M₁ collaboratively in order to establish their performance characteristics. The practical characteristics include: cost of performance, time required, and level of training needed. The scientific characteristics include: accuracy, precision, specificity, and lower limit of detection. Although all these characteristics are important, the most important from the regulatory point of view are accuracy and interlaboratory variation (reproducibility).

Few collaboratively studied methods of analysis for aflatoxin M₁ have been published in the scientific literature. Nearly all are based on TLC and HPLC. The performance characteristics of the methods described above, derived from collaborative studies, are summarized in Table 5.

Criteria for acceptance of collaboratively derived performance characteristics of methods for the determination of aflatoxin M₁ were not formalized until 1999 in countries that had regulations with respect to aflatoxin M₁. In 1999, European Union legislation for aflatoxin M₁ came into force. The Directive (98/53/EC; Commission of the European Union, 1998) includes a section stating specific requirements for the methods of analysis used to determine concentrations of aflatoxins in certain foodstuffs, including aflatoxin M₁. These requirements are based on report 13505 (Comité Européen de Normalisation, 1999). The recommended recovery of aflatoxin M₁ present in milk is 60–120% at a concentration of 0.01–0.05 µg/L and 70–110% at a concentration > 0.05 µg/L. The recommended precision of the relative standard deviation for reproducibility for all concentrations of aflatoxin M₁ in milk is that derived from the Horwitz (1989) equation, i.e. RSD_R = 2 (1–0.5 logC), where RSD_R is the relative standard deviation calculated from results generated under reproducible conditions [S_R/x], and C is the concentration expressed as powers of 10 (e.g. 1 mg/kg (ppm) = 10^–6). The maximum permitted value is twice the RSDR. The detection limits of the methods used are not stated, as the precision values are given at the concentrations of interest.

Most collaborative studies on aflatoxin M₁ were carried out under the auspices of AOAC International, which did not require the reporting of recovery values or the limit of detection or determination in collaborative studies until 2000. In addition, reporting of HORRAT values was not mandatory, and no criteria were established for HORRAT values until 2000 (AOAC International, 2000). Thus, except for the collaborative study of Dragacci et al. (2001), there were no criteria to verify the performance of a method for aflatoxin M₁ in collaborative studies.

The availability of collaboratively studied ‘official’ methods of analysis for aflatoxin M₁, with acceptable performance characteristics, is no guarantee of accurate results. Check sample programmes for aflatoxins, including aflatoxin M₁ in milk, organized by IARC (Friesen & Garren, 1982) have shown that there can be wide variation in results. In compliance with the principles of analytical quality assurance, measurements of the mycotoxin by different laboratories should be reliable and comparable. A quality assurance programme includes, when possible, use of (certified) reference materials and many other elements. Certified reference materials are stable, homogeneous products containing certified amounts of the analyte(s) of interest. Such materials for aflatoxin M₁ have been developed in the past with the coordination of the European Union’s Community Bureau of Reference, now known as the Standards, Measurements and Testing Programme. Several full-cream milk powders certified for their aflatoxin M₁ content and an aflatoxin M₁ calibrating solution are available worldwide from the European Union Joint Research Centre, Institute for Reference Materials and Measurements in Geel, Belgium. Their characteristics are shown in Table 6 (Boenke, 1997). The supplies of certified milk powder reference materials are nearly exhausted, and the Institute is planning to produce and certify new batches in 2001.

Another increasingly important quality assurance component is proficiency testing. The European Union’s Community Reference Laboratory for Milk and Milk Products, in Paris, France, conducted proficiency tests for national reference laboratories in 1996 (Dragacci et al., 1996) and 1998 (Grosso et al., 1999). The organizers concluded that, considering the very low concentrations of aflatoxin M₁ in the distributed samples, the network of national reference laboratories had shown good analytical competency for the determination of aflatoxin M₁ in milk at the permitted level in the European Union (0.05 µg/kg). It is foreseen that proficiency tests for aflatoxin M₁ will be expanded to include national reference laboratories that are charged with detecting mycotoxins. Laboratories that do not belong to the network can take part in the food analysis performance assessment scheme based in the United Kingdom but with worldwide participation. The scheme has organized rounds of testing for aflatoxin M₁ since 1999, and samples are sent out every 3–4 months (Ministry of Agriculture, Fisheries and Food, 1999, 2000a,b). Another possibility is participation in the laboratory proficiency programme of the American Oil Chemists (2000).

Many methods of analysis have become available for the determination of aflatoxin M₁ in milk and milk products, both for screening and for quantitative estimates. Most were developed for the analysis of milk and milk products, but they can be used for other dairy products, with minor modifications. The limits of determination have decreased over the years, while the precision of the methods has improved, as demonstrated in formal collaborative studies of performance characteristics. With modern methods of analysis, aflatoxin M₁ can be determined at concentrations well below 0.05 µg/kg of milk. The combination of immunoaffinity columns and liquid chromatography offers the best means for efficient clean-up and precise determination of low concentrations of aflatoxin M₁. Cheaper alternatives based on immunoaffinity clean-up with TLC and computer-based densitometry have been developed and are being validated.

A value of 0.05 µg/kg of milk is currently the legal limit in those countries that have the most stringent regulations for aflatoxin M₁. This value was proposed by the Codex Committee on Food Additives and Contaminants in 1992 (Codex Alimentarius, 1992), but it was still at the proposal stage in 2000, as several countries had reserved their positions (Codex Alimentarius, 2000). The availability of collaboratively studied methods (which usually involve TLC and HPLC) is, however, no guarantee of accurate results. The reliability and comparability of analytical results can be significantly improved by making use of reference materials and by taking part in proficiency studies. Certified milk powder reference materials for determination of aflatoxin M₁ are available, and proficiency studies with worldwide participation are organized.

Two effective methods for controlling aflatoxin M₁ in the food supply are to sample dairy feed for aflatoxin B₁ or to sample the milk directly for aflatoxin M₁. The performance of sampling plans for alfatoxin in granular feed products such as shelled maize (Park et al., 2000; Johansson et al., 2000a,b,c) and cottonseed (Whitaker et al., 1976) has been evaluated, but there has been little evaluation of sampling plans to detect aflatoxin M₁ in milk. It might be difficult to design an effective programme to control aflatoxin in granular feed, particularly at low concentrations, because of its heterogeneous distribution in these commodities, which results in wide sampling variation.

As the distribution of aflatoxin M₁ in liquid milk can be expected to be reasonably homogeneous, sampling of liquid milk for aflatoxin M₁ will be more accurate than sampling of granular feed products. Most of the uncertainty in estimates of aflatoxin M₁ in milk is probably associated with the analytical procedure.

The European Union, the South American Common Market (MERCOSUR), and the USA have designed plans for sampling aflatoxin M₁. A Directive of the Commission of the European Union (1998) and a Decision (Commission of the European Union, 1991) specify that a minimum of 9.5 kg (or L) should be collected from a batch of milk mixed by manual or mechanical means and should be composed of at least five increments. The batch is accepted if the concentration of aflatoxin M₁ does not exceed the permitted limit. In the USA, the Food & Drug Administration (1996) stipulates that samples should consist of at least 10 lbs (4.5 kg) of milk, composed of no fewer than 10 randomly selected portions.

In the absence of information on the efficacy of sampling plans for the determination of aflatoxin M₁ in milk, it is recommended that the European model, in which a 500-g sample composed of five 100-g portions of milk is taken from a batch, be used for the minimum sample size and sample selection method.

As aflatoxin M₁ occurs frequently in milk, two questions can be raised. What happens to the aflatoxin M₁ when contaminated milk is processed normally in the dairy industry? What can be done to reduce or destroy aflatoxin M₁ in contaminated milk and milk products? The many investigations performed to resolve these questions were reviewed in detail by Yousef & Marth (1989). The main studies and conclusions are summarized below.

Treatments that are common in the dairy industry can be separated into two distinct processes: those that do not involve separation of milk components, such as heat treatment, low-temperature storage, and yoghurt preparation; and processes that involve separation of milk components, such as concentration, drying, and cheese and butter production.

The stability of aflatoxin M₁ during heat processing, such as pasteurization (Alcroft & Carnaghan, 1962) and heating milk directly on a fire for 3–4 h (Patel et al., 1981), has been studied. Although the results of the studies are not consistent, most indicate that such heat treatments do not change the amount of aflatoxin M₁ in these products appreciably. Studies of the stability of aflatoxin M₁ in milk during cool or frozen storage gave variable results (Yousef & Marth, 1989), but storage of frozen contaminated milk and other dairy products for a few months did not appear to affect the aflatoxin M₁ content. The manufacture of cultured dairy products, such as kefir and yoghurt, also did not significantly decrease the aflatoxin M₁ content (Wiseman & Marth, 1983).

The results of several investigations have been published in which the effects of removal of water on aflatoxin M₁ content were studied. The processes involved both heat- (spray or roller drying) and freeze-drying. The studies were reviewed by Yousef & Marth (1989). Large losses of aflatoxin M₁ were reported in some studies, whereas in others concentrating milk did not affect its aflatoxin M₁ content substantially. The few studies that addressed partitioning of aflatoxin M₁ during cream and butter processing confirmed that a small proportion of aflatoxin M₁ is carried over to cream and a yet smaller proportion to butter. No loss of aflatoxin M₁ occurred as the remainder was found in skim milk and buttermilk, respectively.

The manufacture of cheese involves several processes. Aflatoxin M₁ does not appear to be degraded in the first phase, conversion of milk into pressed curd, as the amount in whey and curd is approximately the same as in the original milk (Yousef & Marth, 1989). Aflatoxin M₁ seemed rather to occur predominantly with casein, so that cheese curd contained a higher concentration than whey. The association of aflatoxin M₁ with casein is also manifested in a higher concentration in cheese than in the milk from which the cheese is made. Yousef & Marth (1989) expressed the concentration of aflatoxin M₁ in milk divided by the concentration of aflatoxin M₁ in cheese as the enrichment factor. On the basis of several studies, these researchers concluded that the enrichment factor is 2.5–3.3 in many soft cheeses and 3.9–5.8 in hard cheeses. During the second phase of cheese manufacture, ripening, some discrepancies were found in the stability of aflatoxin M₁, but, in general, it did not appear to be degraded during ripening.

The observation that the processes described above do not generally lead to loss of aflatoxin M₁ is of considerable practical importance. Several possibilities for eliminating or inactivating aflatoxin M₁ in milk, involving chemical and physical treatment, have been investigated. The chemicals that have been studied for their ability to degrade aflatoxin M₁ are limited to those that are permitted as food additives: sulfites, bisulfites, and hydrogen peroxide (Applebaum & Marth, 1982a,b). When raw milk naturally contaminated with aflatoxin M₁ was treated with 0.4% potassium bisulfite at 25 °C for 5 h, the concentration decreased by 45%. A higher concentration of bisulfite was less effective. Aflatoxin M₁ in naturally contaminated milk was not affected by the presence of 1% hydrogen peroxide at 30 °C for 30 min, but addition of hydrogen peroxide at a concentration of 0.05–0.1% with lactoperoxidase reduced the amount by 50%.

Physical processes that have been explored to remove aflatoxin M₁ from milk include adsorption and radiation. Five per cent bentonite in milk adsorbed 89% of aflatoxin M₁ (Applebaum & Marth, 1982a). In a study of the effects of ultra-violet radiation with and without hydrogen peroxide, the concentration of aflatoxin M₁ was reduced by 3.6–100%, depending on the length of time the milk was exposed to radiation, the volume of treated milk, the presence of hydrogen peroxide, and other aspects of the experimental design (Yousef & Marth, 1985).

The chemical and physical treatments described are not readily applicable in the dairy industry, at least at present, as little is known about the biological safety, or the nutritional value of the treated products. Moreover, the costs of the processes may be considerable and prohibitive for large-scale application. If aflatoxin M₁ cannot be destroyed or removed readily, it can be excluded from milk only by eliminating aflatoxin B₁ from the diet of animals.

Data on contamination of milk and various milk products with aflatoxin M₁ were submitted to FAO by Argentina, Canada, the European Union, Norway, the Dubai Municipality of the United Arab Emirates, and the USA. Although not formally submitted, unpublished data were also available from an Indonesian research institute. Other data were taken from literature published between 1985 and early 2000 and from a review by Galvano et al. (1996a). In addition to published data, information was presented by the Philippines, the Republic of Korea, and Thailand. Data on natural occurrence before the mid-1980s and mid-1990s were surveyed by van Egmond (1989) and Galvano et al. (1996a), respectively. Although the available data are incomplete, they are presented in Appendix A.

None of the samples examined in Argentina in 1999 (Centro de Referencia de Micología, 1999) and Uruguay in 1993–95 (Pineiro et al., 1996) contained aflatoxins at concentrations > 0.05 µg/kg, and the maximum concentration was 0.03 µg/kg. Data for Brazil were taken from Martins & Martins (1986), de Sylos et al. (1996), Correa et al. (1997), Oliveira et al. (1997), Souza et al. (1999), and Prado et al. (1999, 2000). None of the milk samples was contaminated with aflatoxin M₁ at > 0.5 µg/kg. Four of 52 milk samples in 1992 contained > 0.05 µg/kg, with a maximum of 0.37 µg/kg (de Sylos et al., 1996), and 35 of 54 samples taken in 1992–93 contained 0.006–0.077 µg/kg (Prado et al., 1999, 2000). Of the samples of milk powder, 2% (6/300) of those taken in 1992–93 contained > 0.5 µg/kg, with a maximum of 1.0 µg/kg (Oliveira et al., 1997).

The data submitted by Health Canada for 1994, although limited, indicated that none of the processed milk samples contained > 0.5 µg/kg, and the mean was < 0.063 µg/kg. Data for 1997–98 (not shown in Appendix A), in which the number of samples was not stated, showed no aflatoxin M₁ in any of six types of milk at the limit of detection (LOD; 0.015 µg/kg).

Extensive data collected between 1995 and early 2000 were submitted by the Center for Food Safety and Applied Nutrition (Food & Drug Administration, 2000) in Washington DC, USA. This submission contained some industry-generated data, for 1998–2000 from southwestern USA (n = 5801), the midwest (n = 438), and southeastern USA (n = 13 093), and data from three Food & Drug Administration (FDA) regional laboratories: laboratory A (n = 277), laboratory B (n = 380), and laboratory 27 (n = 4225). No data were provided from northern USA because aflatoxin M₁ is not a problem under the climatic conditions that prevail.

In the surveys by FDA laboratory 27 for 1995–2000, 185 (4.6%) of 4000 raw milk samples were contaminated with aflatoxin M₁ at a concentration > 0.05 µg/kg and five (0.13%) at > 0.5 µg/kg, only in 1996. Of the finished milk samples during the same period, 10 (4.4 %) of 225 samples contained 0.05–< 0.5 µg/kg. In midwestern and southwestern USA after 1998, 10 (2.3%) of 438 and 14 (0.11%) of 13 093 samples, respectively, contained > 0.5 µg/kg. The incidence of aflatoxin M₁ contamination was higher in southwestern (industry data) and southern states (FDA laboratory B) than in other areas. In 1998–2000, milk samples contaminated with > 0.05 and > 0.5 µg/kg were found in 21% (1239/5801) and 0.78% (45/5801) of southwestern states and in 40% (153/380) and 18% (68/380) of southern states, respectively. The numbers of samples for southwestern and southern states were provided within two distributions (0.00–0.5 µg/kg and > 0.5 µg/kg), with no information on mean or maximum values.

In Asia, high incidences and levels of aflatoxin M₁ contamination were found in Indonesia, the Philippines, and Thailand. Unpublished data for Indonesia in 1990–93 and 1999 were obtained from the Research Institute for Veterinary Science (2000) in Bogor. Of 342 milk samples, 199 samples (58%) contained aflatoxin M₁ (limit of quantification [LOQ], 0.1 µg/kg), and 73 (21%) contained > 0.5 µg/kg, with mean values of 0.31–5.4 µg/kg and maximum values of 2.0–23 µg/kg.

Data for Thailand in 1990–93 and 1995–96 were obtained from the Department of Medical Sciences, Ministry of Public Health (Boriboon & Suprasert, 1994) and the literature (Saitanu, 1997). Of 310 liquid milk samples, more than 261 (> 84%) were contaminated with aflatoxin M₁ at aconcentrations > 0.05 µg/kg, and 58 samples (19%) contained > 0.5 µg/kg, with a maximum of 6.6 µg/kg. In the Philippines, data from the Bureau of Animal Industry, Department of Agriculture, for 1997 (Begino, 1998) indicated that 88% and 18% of 91 milk samples were contaminated with aflatoxin M₁ at > 0.05 µg/kg and > 0.5 µg/kg, respectively.

The data for the Republic of Korea in 1995 and 1997 were taken from the literature (Shon et al., 1996; Kim et al., 2000). Of 134 liquid milk samples, 50 (37%) contained aflatoxin M₁ at a concentration > 0.05 µg/kg, with a maximum of 0.28 µg/kg. In the United Arab Emirates, data for 1998–99 from the Dubai Central Laboratory, Dubai Municipality, showed that 33 (56%) of 59 liquid milk samples (including 15 imported products) were contaminated with > 0.05 µg/kg, with a maximum of 0.31 µg/kg.

The report from the Commission of the European Union (1999) provided data on aflatoxin M₁ concentrations in milk samples (total, 7573) in 1999 from 10 Member States: Austria (20 samples), Belgium (192), Finland (296), France (234), Germany (6537), Ireland (62), the Netherlands (30), Portugal (96), Sweden (11) and the United Kingdom (95). Of these, 7259 samples (96%) contained aflatoxin M₁ at concentrations below the the LOQ or LOD of the method (0.001–0.03 µg/kg); 314 samples (4.2%) contained up to 0.05 µg/kg, and none of samples contained > 0.05 µg/kg.

The Commission of the European Union (1989–95) also provided the SCOOP report, which gives the concentrations of aflatoxin M₁ in milk analysed between 1984 and 1995 in eight Member States (Austria, Belgium, Denmark, Finland, France, Italy, Spain, and the United Kingdom). Of the 8791 milk samples (including crude and heat-treated milk), 34% contained concentrations below the LOQ (0.01 µg/kg)/LOD (0.001–0.01 µg/kg), 44% had concentrations up to 0.05 µg/kg, 0.33% (23 French and six British samples) had 0.051–0.1 µg/kg, and three samples from the United Kingdom had > 0.1 µg/kg, with a maximum of 0.22 µg/kg.

Additional information on a total of 18 945 samples taken in 1994–99 and 2000 was provided by the Commission of the European Union. Of 1048 samples taken in Finland in 1995–98, only one contained < 0.05 µg/kg and the others contained < 0.005 µg/kg. All 251 samples taken in France in 1998 contained < 0.05 µg/kg, and four had > 0.03 µg/kg). In 17 181 samples taken in Germany in 1996–98 and 2000, the maximum concentration was 0.033 µg/kg. All 168 samples of reconstituted milk taken in the Netherlands in 1998 contained < 0.01 µg/kg. The maximum concentration in 42 samples taken in Spain in 1998 was 0.027 µg/kg, and all 255 samples taken in the United Kingdom in 1994–95 had < 0.05 µg/kg.

Data submitted by the Norwegian Food Control Authority showed that 51 (94%) of 54 samples contained aflatoxin M₁ at a mean value of 0.0014 µg/kg and a maximum of 0.009 µg/kg. Data for Cyprus were derived from the literature (Ioannou-Kakouri et al., 1999). Of 112 samples, including 10 imported products, 11 were contaminated with aflatoxin M₁ at a level of 0.01–0.04 µg/kg.

Appendix B shows the distribution of aflatoxin M₁ contamination in various types of milk and milk products. Table 7 gives data calculated from the surveys of FDA laboratory B (1998–2000, 380 milk samples from southern USA), FDA laboratory 27 (1995–2000, 4225 raw and finished milk samples), and the Ministry of Public Health, Thailand (1990–93, 60 raw and pasteurized milk samples; Boriboon & Suprasert, 1994).

Limited information was available on annual variations in the concentration of aflatoxin M₁ in milk. The most recent data are derived from the survey submitted by FDA Laboratory 27 for 1995–2000 (Table 8). The incidence of contamination with aflatoxin M₁ at a concentration > 0.05 µg/kg varied from 1.9% to 3.4% in raw milk and from 0.0% to 3.8% in finished milk in a 6-year survey, except in 1996, when these concentrations were found in 25% of raw milk samples and 19% of finished milk samples. A concentration > 0.5 µg/kg was found in 1.3% of raw milk samples only in 1996.

In data for Germany in 1997–2000 submitted by the Commission of the European Union, 0.11–0.27% of samples were contaminated at > 0.005 µg/kg and 0.06–0.14% at > 0.01 µg/kg in 1997–98 and 2000, and 4.3% at 0.005–0.05 µg/kg and 0.45% at 0.01–0.05 µg/kg in 1999 (Appendix A).

Extensive data on annual variation is contained in the report of a 15-year survey (1978–92) of aflatoxin M₁ in milk and bulk raw milk in France (Table 9; Dragacci & Frémy, 1993). For 1978, 6246 milk samples were classified as containing < 0.05 µg/kg, 0.05–0.5 µg/kg, and > 0.5 µg/kg. According to this classification, two periods of contamination occurred in France, 1978–84 and the winters of 1988–91; highly contaminated samples, some containing up to 5 µg/kg, were found during the winters of 1978–83. The seasonal trend in milk contamination was attributed to the fact that cows receive less concentrated feed in summer, when they are grazing. Few contaminated samples were found between 1984 and 1988 and after 1991.

Dietary intake of aflatoxin M₁ was estimated from data on the concentrations of aflatoxin M₁ in milk submitted to FAO/WHO, from selected reports in the literature, and from data on milk consumption in the GEMS/Food Regional Diets (WHO, 1998).

Eight data sets providing information on aflatoxin M₁ concentrations in milk were submitted for consideration by the Committee. Six of these were in the submissions from Argentina, Canada, the European Union, Norway, the United Arab Emirates, and the USA, and these provided the results of analyses for aflatoxin M₁ in milk from government or university institutions. Experimental data were available from an Indonesian research institute. The other two submissions were summaries of published papers from laboratories in Brazil and Uruguay. The eighth data set was in the form of summaries of published papers on the concentrations of aflatoxin M₁ in milk.

The data are discussed below and summarized in Appendix B, with separate entries for aflatoxin M₁ in different samples of milk. The data in Appendix B for which the number of samples and the mean value were available are summarized in Appendix C by GEMS/Food regional diet.

7.1.1 Canada

Data for 1993–94, 1994–95, 1995–96, and 1997–98 were submitted by Health Canada (2000). The detection limit was 0.015 µg/kg, and the analytical method included immunoaffinity clean-up columns followed by liquid chromatography coupled with fluorimetric detection. In data for 1997–98, the mean concentrations and ranges of aflatoxin M₁ in nine types of milk were all < 0.015 µg/kg (Table 10).

7.1.2 Argentina

Data on the aflatoxin M₁ content of milk in Argentina were submitted by the Universidad National de Lujan (2000). Concentrations of 0.0, 0.4, 0.6, 0.6, 0.6, and 0.9 µg/L (roughly equal to µg/kg) were found in six samples of fluid milk from Buenos Aires that were analysed in 1998. In response to a request about the analytical method and its LOQ/LOD, the method was identified as ‘interchemistry’, which is similar to the AOAC method but with a different clean-up procedure, allowing for an LOQ of 0.026 µg/L. It was also indicated that the milk had come from cows with problematic milk production which had been given a feed not typically used in Argentina (< 12% of milk production obtained from cows given this type of feed). Hence, the aflatoxin M₁ concentrations in the six milk samples may not be typical of those in milk in this country.

7.1.3 USA

Data on aflatoxin M₁ in milk in the USA were submitted by the FDA (2000), including some data provided by industry for 1998–2000 from the southwest (n = 5801), midwest (n = 438), and southeast (n = 13 093). The data from the southeast were provided as the number of samples within two distributions (0.00–0.5 ng/kg and > 0.5 ng/kg), with no information on mean or maximum values. The submission also included data from three FDA regional laboratories: Laboratory A provided data from southeastern states for 1999–2000 (n = 199) and for 1997–2000 (n = 78) within distributions of 0.00–0.5 ng/kg and > 0.5 ng/kg with no mean or maximum values. Laboratory B provided data from southern states for 1998 (n = 163), 1999 (n = 167), and 2000 (n = 49). Laboratory 27 provided two sets of data for 1995–2000, one set for raw milk (n = 3942) and one for finished milk (n = 273). As the data from laboratories B and 27 were provided as individual numbers, the mean values could be calculated, and values < 0.05 µg/kg and < 0.5 µg/kg, maximum values, and values at the 90th percentile of consumption were determined.

Most of the values (72–100%) were 0–0.5 µg/kg (see Appendix B). The values for the southern states were higher than those for other regions.

7.1.4 Norway

Data were submitted by the Norwegian Food Control Authority, with clarification of the mean values for all samples and for positive samples, the analytical method, and individual values for all samples. Aflatoxin M₁ was determined in 54 samples of milk from 49 dairies in four areas of Norway in 1998. The LOD of the analytical method was 0.0001 µg/kg. Aflatoxin M₁ was found in 51 (94%) of the samples, giving a mean value of 0.0014 µg/kg for all samples, a mean of 0.0014 µg/kg for positive samples, and a median value of 0.0009 µg/kg. The highest value was 0.009 µg/kg.

7.1.5 United Arab Emirates

Data on the concentrations of aflatoxin M₁ in eight types of milk product in 1998 and 1999 were submitted by the Dubai Municipality, United Arab Emirates (Dubai Central Laboratory, 2000). It was assumed that the milk described as from ‘local dairies’ was either raw or whole. The data are summarized in Table 11.

Although the original report indicated that the LOD was 0.005 ng/kg and the LOQ was 0.01 ng/kg, the Committee assumed that this was an error and that the values should be in µg/kg. ‘< LOQ’ in Table 11 refers to the total number of samples less the number of samples designated as ‘positive’ in the report, as it was assumed that ‘positive’ values were those above the LOQ. These data are presented in Appendix B, but are not included in Appendix C because an overall mean for the samples was not provided and it could not be calculated because individual values were lacking.

7.1.6 European Union

The Commission of the European Union (1999) provided a summary of the concentrations of aflatoxin M₁ detected in 7573 milk samples in 1999 from 10 Member States: Austria, Belgium, Finland, France, Germany, Ireland, the Netherlands, Portugal, Sweden and the United Kingdom. Concentrations of aflatoxin M₁ below the LOD/LOQ (which differed among the countries) were found in 7259 (96%) of the samples. The concentrations in all 314 samples in which aflatoxin M₁ was detected (4.2%) were < 0.05 µg/kg. Mean values were not reported.

The Commission of the European Union (1989–95) also provided a Scientific Cooperation (SCOOP) report, containing data for milk analysed between 1989 and 1995 in nine Member States. Of the 8791 samples, 3338 (38%) contained concentrations of aflatoxin M₁ below the LOQ/LOD; 1017 samples (12%) contained concentrations < 0.05 µg/kg; six samples (0.07%) contained 0.05–0.1 µg/kg; and three samples (0.03%) contained > 0.1 µg/kg. The nine samples containing > 0.05 µg/kg were from the United Kingdom.

7.1.7 Indonesia

Appendix A gives unpublished data from the Research Institute for Veterinary Science in Bogor, Indonesia, on concentrations of aflatoxin M₁ in milk. The data include the mean and maximum concentrations of aflatoxin M₁ in 342 samples of milk collected during 1990–93 and 1999. The LOQ of the analytical method was 0.1 µg/kg. The concentration was below the LOQ in 143 samples, > 0.05 µg/kg in 199, and > 0.5 µg/kg in 73.

7.1.8 Summary of published data from Brazil and Uruguay

A submission from Brazil (Ministerio da Saude, 2000) provided a summary of data on the concentrations of aflatoxin M₁ in samples of domestic cow milk from five references (Sabino et al., 1989; de Sylos et al., 1996; Souza et al., 1999; Prado et al., 1999, 2000). A second submission from Brazil (Rodriquez-Amaya, 2000) was a paper prepared for the Tenth International IUPAC Symposium on Mycotoxins and Phycotoxins, which summarized data on concentrations of aflatoxin M₁ in milk and milk products in Brazil (de Sylos et al., 1996; Correa et al., 1997; Oliveira et al., 1997) and Uruguay (Pineiro et al., 1996). The information from the eight papers (converted to µg/kg) is summarized in Table 12.

7.1.9 Additional data from the literature

Additional information about concentrations of aflatoxin M₁ in milk was sought in the literature for guidance on appropriate levels for estimating daily intake and for information for geographic areas not covered by the submitted data. In particular, there were no submissions from Africa and few from the Far East or Middle East. Pertinent data (converted to µg/kg, assuming that 1 L of milk weighs 1.02 kg) are shown in Table 13. Other data were evaluated but not included in Table 13, for several reasons. Data reported by Lemieszek-Chodorowska (1974) for Poland and by Brewington & Weihrauch (1970) for the USA were considered to be too old; Fukal & Brezina (1991) did not provide mean values for their data from the former Czechoslovakia; the data from Cyprus (Ioannou-Kakouri et al., 1999) were of doubtful quality because of a low per cent recovery; Smith et al. (1994) in the USA reported data for only nine samples of goat milk, which is not commonly consumed in the USA; and the data of Galvano et al. (1998) for Italy probably duplicated data submitted by the Commission of the European Union. Other reports that did not contain pertinent data were those of Masri et al. (1969b), Grant & Carlson (1971), Purchase et al. (1972), Jung & Hanssen (1974), Stoloff et al. (1975), Kiermeier & Mashaley (1977), Nikov et al. (1991), el-Nezami et al. (1995), Saad et al. (1995), and Navas et al. (2000).

Additional data from the literature, summarized in Appendix A, were considered. There is already considerable overlap in the data in Appendices A and B, and there may be some duplication, because the Commission of the European Union provided summaries of data from individual Member States without identifying the sources of the data within each country. The criteria for selecting data from Appendix A for addition to Appendix B were that the data were not already in Appendix B; they were relatively recent (1990 or later); they related to fluid milk or yoghurt, rather than dry milk or other dairy products; and the number of samples analysed and the mean concentration of aflatoxin M₁ in the samples was given. The data from Appendix A added to Appendix B were those from Africa (El-Gohary, 1996), France (Dragacci & Frémy, 1993), the Philippines (Begino, 1998), Poland (Domagala et al., 1997), the Republic of Korea (Shon et al., 1996), Spain (Macho et al., 1992; Jalon et al., 1994), and Thailand (Boriboon & Suprasert, 1994; Saitanu, 1997). The data from France (Dragacci & Frémy, 1993) and Spain (Macho et al., 1992; Jalon et al., 1994) were assumed to be included in the summary from the Commission of the European Union and are therefore included in Appendix B but are not summarized in Appendix C, to avoid duplication.

7.1.10 Summary of data on aflatoxin M₁ in milk

The data in Appendix B are summarized in Appendix C for the five regional diets by aggregating milk types and/or periods from the same source or reference. The European regional diets are derived from data submitted from Canada, the Commission of the European Union and nine of its Member States, Norway, Poland, and the USA. Latin American diets are derived from data from Argentina, Brazil, and Uruguay. Far Eastern diets are represented by data from India, Indonesia, the Philippines, the Republic of Korea, and Thailand. Middle Eastern diets are represented by data from Greece and the United Arab Emirates, and African diets are represented by data from Egypt.

For the European regional diet, the mean value of aflatoxin M₁ in milk ranged from about 0.004 to 0.40 µg/kg on the basis of data from the Commission of the European Union, Norway, and the USA. The calculated weighted mean was 0.023 µg/kg.

For the Latin American regional diet, the mean value ranged from 0 to 0.52 µg/kg. The sources were six published papers (Sabino et al., 1989; de Sylos et al., 1996; Correa et al., 1997; Oliveira et al., 1997; Prado et al., 1999; Souza et al., 1999) and a submission of laboratory data from Argentina. The weighted mean was calculated to be 0.022 µg/kg.

For the Far Eastern regional diet, the sources were five published papers (Boriboon & Suprasert, 1994; Rajan et al., 1995; Shon et al., 1996; Begino, 1998; Kim et al., 2000) and a submission from a laboratory in Indonesia. The weighted mean was calculated to be 0.36 µg/kg. The values from Indonesia were higher than those from other countries in both the Far Eastern region and other regions.

For the Middle Eastern regional diet, the sources were two published papers (Karaioannoglou et al., 1989; Markaki & Melissari, 1997). The weighted mean was 0.005 µg/kg.

Only one source, a published paper from Egypt (El-Gohary, 1996), was available for the African regional diet. This source provided an average value for aflatoxin M₁ in powdered milk of 0.015 µg/kg, which would be equivalent to 0.0018 µg/kg in fluid milk, as 30 g of dried milk yields 244 g of fluid milk (the ratio of dried to fluid milk is 1:8.2, and 0.015/8.2 = 0.0018 µg/kg).

In the European, Latin American, Middle Eastern, and African diets, the weighted mean value for aflatoxin M₁ in milk was below the proposed maximum levels of 0.05 µg/kg and 0.5 µg/kg. In the Far Eastern regional diet, the weighted mean value for aflatoxin M₁ in milk (0.36 µg/kg) was greater than the proposed maximum level of 0.05 µg/kg but below 0.5 µg/kg.

7.2.1 Milk consumption

The GEMS/Food regional diets (WHO, 1998) provide the dietary intakes of food commodities in five geographical areas. Table 14 shows the intake of all milk and milk products for the five regions. The major food class in which aflatoxin M₁ was identified was milk, the term ‘milk’ being assumed to include the mammalian milks (buffalo, camel, cow, goat, and sheep) listed in the GEMS/Food regional diets and not cheese, butter, or other dairy products derived from milk. Because the GEMS/Food database does not include separate data on the consumption of fermented milks, it was assumed that consumption of yoghurt and other fermented milks was subsumed within the figures for milk.