2.2.5 Reproductive toxicity

(a) Effects on reproductive organs

Mice: The potential of deoxynivalenol to affect testicular morphology and testicular and epididymal sperm counts was assessed in three strains of mice: IL-6KO [B6129-IL6 (tmlKopf) (IL-6 gene-deficient)], WT [B6129F2 (wild-type to B6129-IL6 with an intact IL-6 gene)], and B6C3F₁ mice. The treated mice received deoxynivalenol at a concentration of 10 mg/kg in their diet (equivalent to 1.5 mg/kg bw per day) for 90 days. The body weight of treated animals was significantly lower than that of controls. Slight, not statistically significant changes were observed in relative testis weight and testicular spermatid counts, but no histological changes were seen. The diameter of the seminiferous tubules, the height of the seminiferous epithelium, and the number of Sertoli cell nucleoli per cross-sectioned seminiferous tubule in the treated groups were not significantly different from those of their respective untreated controls. IL-6KO and B6C3 F₁ mice had significantly lower caudal epididymal weights than controls. These changes were not due to decreased sperm counts, and the finding suggests that deoxynivalenol may have an adverse affect on the epididymides (Sprando et al., 1999).

Pigs: Diets containing uncontaminated wheat, wheat contaminated with deoxy-nivalenol at 3.7 mg/kg, or maize inoculated with Fusarium at 4.2 mg/kg was fed to groups of 12–18 23-kg male and female Yorkshire pigs (equivalent to 0, 0.14, and 0.17 mg/kg bw per day) for 7 weeks. The diets caused a 23–29% reduction in feed consumption. The weight gain of animals on the diets containing contaminated wheat or maize was 30% and 72% less than that of controls, respectively, suggesting that there were additional metabolites in the maize diet. Histological examination of the testis (seminiferous epithelium) and ovary (follicle) revealed no significant differences in sexual development attributable to the diet (Friend et al., 1986a).

(b) Multigeneration studies

The results of these studies are summarized in Table 5.

Mice: Groups of 15 weanling mice of each sex (F₀) were fed diets containing deoxynivalenol at concentrations that resulted in a dose of 0 or 2 mg/kg bw per day, and groups of seven male and 10–20 female mice received diets providing a dose of 0, 0.38, 0.75, or 1.5 mg/kg bw per day. The diets were fed continuously to the F₀ parents and their progeny for the duration of the two experiments. After 30 days, the mice were allowed to mate within their experimental groups for a maximum of three 5-day trials. Females found to have mated successfully were allowed to litter normally. The F_1a progeny of 10 dams in the control group and that receiving deoxynivalenol at 1.5 mg/kg bw per day were cross-fostered at birth, whereas the remaining F_1a progeny were reared by their natural dams. The offspring were examined up to 21 days of age and were then discarded. The F₀ mice were re-bred to produce F_1b litters, which were killed on day 19 of gestation, and the fetuses were examined for gross, visceral, and skeletal malformations. The feed and water intakes and body weights of male and female F₀ mice were reduced, as were the numbers of live pups and postnatal survivors, the postnatal body weight of F_1a progeny, the number of live fetuses, and the mean weight of F_1b fetuses. No adverse effects were found on the fertility of male and female F₀ mice, and there were no major malformations in the F_1b generation. Cross-fostering of the offspring of control dams and those at 1.5 mg/kg bw per day adversely affected postnatal survival and body weight after prenatal exposure or after combined pre- and postnatal exposure (Khera et al., 1984).

Rats: Groups of 15 male and 15 female Sprague-Dawley rats, 30 days of age, were fed diets containing deoxynivalenol (purity, 99%) to deliver a dose of 0, 0.25, 0.5, or 1 mg/kg bw per day. After 6 weeks of feeding, the rats were bred within groups, and the males were then discarded. The mated females were maintained on their respective diets throughout gestation and were killed on the last day of gestation; the fetuses were evaluated for effects on prenatal development. No reproductive effects were noted. Except for dilatation of the renal pelvis and urinary bladder, the significance of which was unclear, no other adverse effects were observed in the pups. Decreased body weight, related to decreased food consumption, was observed in dams at the two higher doses (Khera et al., 1984).

Groups of 10 male and 25 female Sprague-Dawley rats weighing 165 g were fed a diet containing purified deoxynivalenol at a concentration of 20 mg/kg (equal to about 2 mg/kg bw) for 60 and 15 days, respectively, before mating. Rats that ate the deoxynivalenol-supplemented diet throughout gestation and lactation showed no clinical signs of toxicity but had lower body weights than pair-fed rats. Only 50% of the matings between treated rats resulted in pregnancy, compared with 80% in controls fed ad libitum or pair-fed. No differences were seen in the sex ratio, survival rate, or average number and weight of litters. The weight gains of pups were comparable in all groups up to postnatal day 14, but between days 14 and 21, control male and female pups had significantly better weight gain than pups of treated dams. No treatment-related histological abnormalities were found in the testes or ovaries of treated pups (Morrissey & Vesonder, 1985).

Pigs: From the time of breeding at an average age of 178 days and a body weight of 121 kg, three groups of 12 Yorkshire gilts were offered ad libitum one of three diets containing 70% wheat and deoxynivalenol at a concentration of 0.1, 1.7, or 3.5 mg/kg from naturally contaminated wheat, equal to 0.002, 0.04, and 0.07 mg/kg bw per day. The gilts were housed and fed individually and were slaughtered on day 50–54 of gestation. The reproductive tract, oesophagus, stomach, large intestine, liver, heart, kidney, bladder, and adrenals were examined. The growth rate of gilts given the highest concentration of deoxynivalenol was significantly less (p < 0.01) than that of other gilts, probably as a result of reduced feed intake. The differences in organ weights were not significant. The fetal mortality rate, although lowest for gilts fed the highest concentration of deoxynivalenol, was not significantly different among the groups. There were significant linear trends towards lower fetal weight, decreased fetal length, and reduced osmolality of allantoic fluid with increasing concentration of deoxynivalenol which could not be attributed to a direct physiological or toxicological effect of deoxynivalenol. There was no increase in the frequency of fetal resorptions, and no gross malformations were observed, but the fetuses were not examined microscopically (Friend et al., 1983).

Groups of 6–10 Yorkshire gilts weighing 91 kg were fed restricted quantities (2 kg/day) of wheat diets containing deoxynivalenol at a concentration of 0.2, 3.8, or 6.2 mg/kg from naturally contaminated wheat, equal to 0.003, 0.06, and 0.09 mg/kg bw per day, respectively until farrowing, and unrestricted quantities of the same diets thereafter until weaning (total, 114 days). No effects on maternal weight gain or on the number and size of piglets at weaning or at time of market were observed (Friend et al., 1986b).

(c) Developmental toxicity

Mice: Deoxynivalenol dissolved in distilled water was given by oesophageal intubation to groups of 15–19 pregnant Swiss Webster mice on days 8–11 of gestation. The incidence of resorptions was 100% at 10 or 15 mg/kg bw per day and 80% at 5 mg/kg bw per day. At the lowest dose, the number of live fetuses and the average fetal weight were below those in controls. Visceral anomalies considered to be teratogenic effects were observed mainly in this group and included exencephaly (26%), syndactyly (19%), and hypoplastic cerebellum (93%). Low incidences of skeletal and visceral anomalies were found in the fetuses of dams at 1, 2.5, and 5 mg/kg bw per day. The skeletal malformations occurred in a dose-related manner and included lumbar vertebrae with fused arches or partly absent centra and absent or fused ribs attributed to retarded ossification (biologically not significant). There was no apparent maternal toxicity. No teratogenic or embryotoxic effects were seen in mice at 0.5 mg/kg bw per day (Khera et al., 1982).

Rats: The teratogenic potential of purified deoxynivalenol was studied by feeding a certified rat feed to which deoxynivalenol was added at a concentration of 0.0, 0.5, 2.0, or 5.0 mg/kg, equivalent to 0.025, 0.1, and 0.25 mg/kg bw per day ad libitum to groups of 23 female Fischer 344 rats throughout gestation. There were no overt signs of toxicity in the dams and no statistically significant differences in feed consumption in comparison with the control group. The dams receiving the two higher concentrations of deoxynivalenol tended to weigh less at term than other females, and their carcass weights were significantly lower (by 5%) than those of the control group after removal of the pups and uterus. The weights of the pups were unaffected by maternal treatment. Deoxynivalenol had no statistically significant adverse effects on the incidence of gross, skeletal, or visceral abnormalities, and neither dams nor pups showed any significant histopathological changes (Morrissey, 1984).

Rats were given an aqueous solution of deoxynivalenol by stomach tube, providing a dose of 0.2, 1, 5, or 10 mg/kg bw per day, on days 7–15 of gestation. On the basis of fetotoxic effects (skeletal abnormalities such as delayed ossification), the NOEL was 0.2 mg/kg bw per day (Tutel’ ian et al., 1991).

Rabbits: Groups of 13–15 adult female New Zealand white rabbits were fed diets containing deoxynivalenol (purity, > 98%) at a concentration of 0, 7.5, 15, 30, 60, 120, or 240 mg/kg throughout gestation, equal to 0, 0.3, 0.6, 1, 1.6, 1.8, and 2 mg/kg bw per day. An additional pair-fed control group was added for the group given 1.6 mg/kg bw per day. The incidence of fetal resorption was 100% in the females fed 1.8 or 2 mg/kg bw per day; although an increased incidence of resorption was also observed at 1 and 1.6 mg/kg bw per day, it was similar to that in the pair-fed controls. Maternal weight was decreased in the pair-fed controls and in rabbits receiving 1 or 1.6 mg/kg bw per day, with associated reductions in mean fetal weight of 7%, 7%, and 28%, respectively. No teratogenic effects were observed (Khera et al., 1986).

2.2.6 Special studies

a) Immunotoxicity

Mice: The studies in mice summarized in this section are shown in Table 6.

Groups of 12 male weanling Swiss Webster mice were given deoxynivalenol by gavage at a dose of 0.75 or 2.5 mg/kg bw per day for 5 weeks. The antibody response to sheep red blood cells was suppressed, and the lower dose decreased the weights of the spleen and thymus (Tryphonas et al., 1984).

In a follow-up study by the same group, immune function was studied in groups of 6–10 male weanling Swiss Webster mice fed purified deoxynivalenol at 0, 0.25, 0.5, or 1 mg/kg bw per day for 5 weeks. Spleen plaque-forming cells and serum antibody responses to sheep red blood cells were unaffected at any dose. At the two higher doses, deoxynivalenol induced a dose-related reduction in the time to death after a challenge with Listeria monocytogenes and increased proliferative capacity in splenic lymphocytes stimulated with phytohaemagglutinin. No effects were observed at 0.25 mg/kg bw. The authors estimated that the NOEL for immunotoxicity in mice was 0.25–0.5 mg/kg bw per day (Tryphonas et al., 1986).

After 2–3 weeks on a diet containing deoxynivalenol at 25 mg/kg, equal to 5 mg/kg bw per day, groups of five female B6C3F₁ mice showed depressed plaque-forming cell response to sheep red blood cells, delayed hypersensitivity response to keyhole limpet haemocyanin, and reduced ability to clear L. monocytogenes in comparison with pair-fed controls, whereas a diet containing 5 mg/kg, equal to 1 mg/kg bw per day, had no effect on these parameters. The effects on resistance to Listeria and delayed hypersensitivity seen after 2–3 weeks disappeared when feeding was extended to 8 weeks; however, the effects on the plaque-forming cell response were detected after both 2 and 8 weeks of ingestion of the mycotoxin. The NOEL for deoxynivalenol was 1.0 mg/kg bw per day (Pestka et al., 1987b).

Groups of eight female B6C3F₁ mice were fed AIN 76A semi-purified diets containing purified deoxynivalenol at 0, 0.5, 2, 10, or 25 mg/kg (equal to 0, 0.1, 0.4, 1, 2, and 5 mg/kg bw per day) for 6 weeks. Leukocyte counts were depressed at doses > 10 mg/kg of diet. The NOEL for this parameter was 1 mg/kg per day (Forsell et al., 1986).

Groups of 4–17 male BALB/c mice, 4–6 weeks old, were fed diets containing deoxynivalenol at a concentration of 0, 2.5, 5, 10, 20, or 50 mg/kg, equivalent to 0, 0.37, 0.75, 1.5, 3, and 7.5 mg/kg bw per day, for 1 or 2 weeks. Control animals were pair-fed at the highest dose. At concentrations > 10 mg/kg of diet, reductions were seen in the response to sheep red blood cells, splenic leukocyte responses to phytohaemagglutinin and lipopolysaccharide, and thymic responses to phytohaemag-glutinin; the weight of the thymus was reduced, with extensive atrophy. The NOEL was 5 mg/kg of diet, equivalent to 0.75 mg/kg bw per day (Robbana-Barnat et al., 1988).

Groups of 10 male BALB/c mice, 7 weeks of age, were given deoxynivalenol in their drinking-water at a concentration of 0, 0.2, 1, or 3 mg/L for 4 weeks, equivalent to intakes of 0, 0.024, 0.12, and 0.36 mg/kg bw per day, and resistance to Salmonella enteritidis was evaluated on day 14. These concentrations of deoxynivalenol did not cause refusal of water or feed. Deaths due to S. enteritidis infection were observed at 1 and 3 mg/L but not at 0.2 mg/L. In mice given drinking-water containing deoxynivalenol at 2 mg/L, both IgM antibody (p < 0.005) (humoral) and delayed-type hypersensitivity (p < 0.05) (cell-mediated) responses to S. enteritidis were significantly suppressed. The authors reported a LOEL, based on water intake, of 0.12 mg/kg bw per day (Sugita-Konishi et al., 1998). The Committee noted that the data on S. enteritidis infectivity were presented in a descriptive fashion without statistical analysis.

Groups of five lactating, inbred Han:NMR1 mice were given deoxynivalenol at 12.5 mg/kg bw for 1 day or at 6.25 mg/kg bw for 7 consecutive days by gavage, and their resistance to the mastitic pathogens Staphylococcus hyicus and Mycobacterium avium was examined. No suppression of the immune response was observed. Rather, both treatments enhanced resistance to S. hyicus but not to M. avium. In mice infected with S. hyicus, administration of deoxynivalenol for 1 day increased total serum IgA, whereas administration for 7 days increased IgA, IgM, and IgG (Atroshi et al., 1994). Enhanced host resistance is seen frequently when trichothecenes are administered just before challenge with a model pathogen (Bondy & Pestka, 2000).

Chickens: Groups of 10 female white Leghorn chicks, 1 day old, were fed diets containing uncontaminated wheat or naturally contaminated wheat containing deoxynivalenol at a concentration of 18 mg/kg, equivalent to 2.25 mg/kg bw per day, for 18 weeks. The contaminated diet resulted in a suppressed antibody response to Newcastle disease vaccine given at week 14. When groups of three 1-day-old broilers were fed a diet containing 50 mg/kg, equivalent to 6.25 mg/kg bw per day, a suppressed lymphocyte blastogenesis response was seen (Harvey et al., 1991).

Pigs: Groups of six to eight castrated male Yorkshire pigs (6–7 weeks old, weighing 13 kg) were given diets amended with maize naturally contaminated with deoxynivalenol at a concentration of 0, 0.95, 1.8, or 2.8 mg/kg of feed, equal to 0.08, 0.13, and 0.18 mg/kg bw per day for 28 days. The diet also contained 15-acetyldeoxynivalenol at about 25% the concentration of deoxynivalenol and zearalenone at about 4%. Antibody responses to sheep red blood cells were delayed in animals exposed to the two highest concentrations; the results for pair-fed controls indicated that this was not solely a nutritional effect. The treatments had no effect on peripheral blood mononuclear cell proliferative responses to the mitogens concanavalin A, phytohaemagglutinin, and pokeweed mitogen. At the end of the experiment, the total leukocyte count was found to be increased with increasing deoxynivalenol concentration, apparently due to increases in segmented and band neutrophil counts. No alterations were seen in monocyte and eosinophil counts. The NOEL for the most sensitive immune parameter was 0.08 mg/kg bw per day (Rotter et al., 1994b). It should be noted that castration might alter the sensitivity of pigs, as was seen in mice (Greene et al., 1994).

The effects on the immune response of diets containing naturally contaminated oats containing deoxynivalenol at a concentration of 0.6 (control), 1.8, or 4.7 mg/kg of feed (equivalent to 0.024, 0.072, and 0.2 mg/kg bw) for 9 weeks were investigated in groups of eight male and female growing Norwegian Landrace pigs. The immune response was evaluated on the basis of primary and secondary antibody titres after injection of five antigens: human serum albumin, sheep red blood cells, paratuberculosis vaccine, tetanus toxoid, and diphtheria toxoid. Tests for delayed hypersensitivity and lymphocyte stimulation were also performed. A significant, dose-dependent reduction in secondary antibody response to tetanus toxoid was observed. A slightly higher mitogen response after phytohaemagglutinin stimulation was seen in lymphocytes from animals given the two higher doses of deoxynivalenol when compared with the group given the lowest dose after 9 weeks, but this result was considered inconclusive. No other indication of dose-dependent inhibition or stimulation of immune response was found, and there was no evidence for the presence of IgA-associated nephropathy (Øvernes et al., 1997). The Committee noted that this study was limited by the absence of a toxin-free control group, and no LOEL or NOEL could be identified.

Mice: Groups of eight weanling female B6C3F₁ mice were fed AIN 76A semi-purified diet containing purified deoxynivalenol at a concentration of 0, 0.5, 2, 10, or 25 mg/kg, equal to 0, 0.1, 0.4, 2, and 5 mg/kg bw per day, for 6 weeks. The serum IgA levels were increased at doses > 0.4 mg/kg bw per day, with no effect at 0.1 mg/kg bw per day. The serum IgM level was decreased in animals at 5 mg/kg bw per day. The NOEL was 0.1 mg/kg bw per day (Forsell et al., 1986).

The same group subsequently reported that the serum IgA level could be induced maximally in female B6C3F₁ mice by feeding them a diet containing deoxynivalenol at 25 mg/kg, equal to 5 mg/kg bw per day. The effect was detectable after 4 weeks of treatment, and the level increased to 17 times the control level after 24 weeks. Concurrent decreaseswere seen in serum IgM and IgG. Comparison with diet-restricted controls showed that these effects on the Ig isotype were not due solely to reduced food intake (Pestka et al., 1989).

In a later study, the same group compared the sensitivity of groups of seven to nine male and female B6C3F₁ mice, 8–10 weeks old, and found that a dietary concentration of deoxynivalenol of at least 10 mg/kg was necessary to induce consistent, significant increases in serum IgA level in males and females at 4, 8, and 12 weeks; 2 mg/kg had no effect. This would indicate a NOEL for deoxynivalenol of 0.4 mg/kg bw per day (Greene et al., 1994).

In a 2-year study, B6C3F₁ mice were given diets containing purified deoxy-nivalenol at a concentration of 0, 1, 5, or 10 mg/kg of feed, equal to 0, 0.1, 0.5, and 1.1 mg/kg bw per day in males and 0, 0.1, 0.7, and 1.6 mg/kg bw per day in females. A linear, dose-related increase in serum IgA and IgG levels was observed in female but not male mice. The increase seen, about 1.5- fold, was much smaller than those found in studies of shorter duration from the laboratory of Pestka. The authors suggested that feeding deoxynivalenol for 2 years might have allowed for adaptation, thus masking earlier effects (Iverson et al., 1995). The Committee suggested that differences in the diets used (Purina certified feed in the last study and AIN-76A semi-purified diet in the previous studies) might also have played a role.

Pigs: In two studies (see section 2.2.2), groups of 7–11 female or castrated male growing Norwegian Landrace pigs were fed diets amended with oats naturally contaminated with deoxynivalenol at a concentration of 0, 0.7, 1.7, or 3.5 mg/kg, equal to 0.04, 0.1, and 0.2 mg/kg bw per day. No differences in serum IgA levels was detected, and no evidence for IgA-associated nephropathy was found (Bergsjø et al., 1993a).

The mechanistic basis for the increase in serum IgA induced by deoxynivalenol has been examined in detail, typically by giving a single concentration of 10–25 mg/kg of feed, equal to 2–5 mg/kg bw per day, to 8–10-week-old B6C3F₁ mice to achieve the maximal response. Peyer’s patch lymphocytes and, to a lesser extent, splenic lymphocytes isolated from female B6C3F₁ mice fed purified deoxynivalenol at 25 mg/kg of feed produced significantly more IgA than cultures derived from mice receiving diets ad libitum or restricted control diets. These results suggest that deoxynivalenol enhances differentiation to IgA secreting cells at the level of Peyer’s patches, which affects the systemic immune compartment (Pestka et al., 1989, 1990a,b; Bondy & Pestka, 1991).

With an ex-vivo approach and neutralizing antibodies, it was found that the potential for enhanced IgA production exists in mouse lymphocytes as early as 2 h and as late as 24 h after a single oral exposure to purified deoxynivalenol at 5 or 25 mg/kg bw (Yan et al., 1997). This effect may be related to an increased capacity to secrete the helper cytokines interleukin (IL)-2, IL-5, and IL-6. Both CD4⁺ and macrophage cells appear to be involved in this process. Thus, increased cytokine expression may be partly responsible for upregulation of IgA secretion in mice exposed orally to deoxynivalenol (Yan et al., 1998).

An increase in serum IgA level in female B6C3F₁ mice after ingestion of a diet containing deoxynivalenol at 25 mg/kg, equal to 5 mg/kg bw day, for 24 weeks resulted in marked deposition of IgA in the kidney mesangium, mimicking common human glomerulonephritis (Pestka et al., 1989; Dong & Pestka, 1993). This effect has since been observed in several strains of mice. These IgA deposits can persist in the kidney for at least 16 weeks after 8 weeks’ feeding of deoxynivalenol (Dong & Pestka, 1993).

Administration to groups of seven to nine male and female B6C3F₁ mice of a diet containing purified deoxynivalenol at a concentration of 2 or 10 mg/kg of feed (equal to 0.4–2 mg/kg bw per day) resulted at week 12 in a significant increase in renal mesangial IgA, in a dose-related manner, and to a greater extent in male mice. Other effects included haematuria and increased IgA immune complexes (Greene et al., 1994).

Another common feature of human IgA-associated nephropathy and the deoxynivalenol–mouse model is the involvement of polyvalent ‘natural’ IgA, which may be associated with immune complex formation and subsequent glomerulo-nephritis (Rasooly & Pestka, 1992, 1994; Rasooly et al., 1994; Yan et al., 1997).

Intermittent dietary intake by 8–10-week-old female mice of purified deoxy-nivalenol was less effective in inducing IgA-associated nephropathy than continuous exposure, perhaps due to the ability of mice to stop eating contaminated feed until it is replaced with control feed (Banotai et al., 1999a).

The presence of purified deoxynivalenol at 5 or 10 mg/kg of feed, equal to 1 or 2 mg/kg bw per day, induced IgA-associated nephropathy in murine models of systemic lupus erythematosus but did not exacerbate the manifestations of lupus (Banotai et al., 1999b).

The ability of deoxynivalenol to alter the expression of cytokines transiently is important because such effects can disrupt normal regulation of a wide variety of immune functions. Deoxynivalenol can up-regulate cytokine production in murine models in vitro and in vivo (Heller et al., 1990; Dong et al., 1994; Warner et al., 1994; Azcona-Olivera et al., 1995a,b; Ouyang et al., 1995, 1996a; Ji et al., 1998; Wong et al., 1998). The concentrations required for effects in vitro (50–1000 ng/ml) are readily attained within minutes in plasma, lymph, and other tissues of mice given 5 or 25 mg/kg bw by gavage and can last for several hours (Azcona-Olivera et al., 1995a). Thus, in-vitro approaches are suitable for exploring the mechanisms of action of deoxynivalenol in vivo.

Superinduction of cytokine gene expression by deoxynivalenol is mediated by both transcriptional and post-transcriptional mechanisms. For example, transcriptional mechanisms involving NF-B and AP-1 have been described for IL-2 in T-cell lines (Ouyang et al., 1996b; Li et al., 2000) and IL-6/TNF-alpha in cloned macrophage cell lines at deoxynivalenol concentrations of 100–250 ng/ml (Wong, 2000). With transcriptional inhibitors, superinduction of IL-2 mRNA expression by deoxynivalenol was found to be due partly to markedly increased IL-2 mRNA stability in T cells (Li et al., 1997) and IL-6/tumour necrosis factor (TNF)-alpha mRNA stability in macrophages (Wong, 2000).

The effects of deoxynivalenol on cytokine mRNA expression in groups of three male B6C3F₁ mice were investigated after a single oral dose of deoxynivalenol at 0, 0.1, 0.5, 1, 5, or 25 mg/kg bw. The abundance of cytokine mRNA in spleen and Peyer’s patches (indicators of the systemic and mucosal immune compartments, respectively) were assessed 2 h after exposure by reverse transcriptase-polymerase chain reaction in combination with hybridization analysis. At 5 and 25 mg/kg bw, deoxynivalenol significantly induced the mRNAs for the proinflammatory cytokines IL-1beta, IL-6, and TNF-alpha; the T helper 1 cytokines interferon (IFN)-gamma and IL-2; and the T helper 2 cytokines IL-4 and IL-10, whereas lower doses had no effect. IL-12p40 mRNA was also induced. but IL-12p35 mRNA was not. The effects were more pronounced in spleen than in Peyer’s patches. IL-5 and TGF-alpha mRNAs were expressed constitutively in spleen and Peyer’s patches but were not affected by deoxynivalenol. The NOEL was 1 mg/kg bw per day (Zhou et al., 1997).

The same authors subsequently examined the effects of repeated doses of deoxynivalenol on cytokine expression. Groups of three male B6C3F₁ mice, 8–10 weeks old, were given oral doses of purified deoxynivalenol at 0, 0.5, 2, or 5 mg/kg bw per day for 2, 4, or 7 days, and cytokine mRNAs were assessed 2 h after the last treatment in spleen and Peyer’s patches. After administration of 2 or 5 mg/kg bw per day, the relative abundance of IL-1beta, IL-6, TNF-alpha, IL-12 p35, IL-12p40, IL-2, and IL-10 mRNAs increased in a dose-related manner, whereas IFN-gamma and IL-4 mRNAs were unaffected. The NOEL was 0.5 mg/kg per day (Zhou et al., 1998).

High doses of trichothecenes promote rapid onset of leukocyte apoptosis (programmed cell death), which is manifested as immunosuppression. Flow cytometry was used to demonstrate that deoxynivalenol inhibits or enhances apoptosis in a concentration-dependent manner in T cells, B cells, and IgA⁺ cells isolated from spleen, Peyer patches, and thymus. Induction of apoptosis was dependent on the lymphocyte subset, tissue source, and glucocorticoid induction (Pestka et al., 1994).

Deoxynivalenol-induced apoptosis was observed in murine macrophage cells in vitro. These results are relevant to whole animals, since administration of trichothecenes to rodents in vivo results in apoptosis in thymus, spleen, bone marrow, and liver (Ihara et al., 1997; Shinozuka et al., 1997; Ihara et al., 1998; Miura et al., 1998; Shinozuka et al., 1998).

(b) Neurotoxicity

Groups of six male Sprague-Dawley rats weighing 180 g and 4-week-old white Leghorn cockerels were given purified deoxynivalenol at 2.5 mg/kg bw once by gavage. Whole brains collected 2, 6, 12, 24, and 48 h after treatment did not show altered concentrations of monoamine neurotransmitters or their metabolites. In a second experiment, in which brains collected 24 h after dosing were dissected into five regions, biogenic monamines were assayed by HPLC with ECD. Deoxynivalenol significantly increased the concentrations of serotonin (102–180% greater than control) and 5-hydroxyindole-3-acetic acid (27–79% greater than control) in all regions in rats, but did not significantly change the regional concentrations of noradrenaline and dopamine. In poultry, however, the treatment decreased the concentrations of noradrenaline in the hypothalamus and hippocampus and decreased those of dopamine in the pons and medulla oblongata. These results suggest that deoxynivalenol influences the metabolism of biogenic amines in brain and that there may be intraspecies differences in the central effects of this mycotoxin (Fitzpatrick et al., 1988a,b).

The effect of deoxynivalenol on brain amine concentrations was investigated in pigs, in which the toxin causes suppression of feed intake (anorexia) in susceptible animals. After administration of a single dose of deoxynivalenol at 0.25 mg/kg bw intravenously to groups of eight male castrated Yorkshire pigs, aged 10–13 weeks and weighing 15–23 kg, the concentrations of the endogenous catecholamines noradrenaline, dopamine, 3,4-dihydroxyphenyl acetic acid, and homovanillic acid and the indoleamines serotonin and 5-hydroxyindoleacetic acid were determined by HPLC with electrochemical detection in five brain regions periodically over 24 h after dosing. The effects of deoxynivalenol were specific to each transmitter, time, and brain region. The concentrations of the main transmitters (noradrenaline, dopamine, and serotonin) were statistically significantly different from those of controls in the hypothalamus, frontal cortex, and cerebellum up to 8 h after dosing. Overall, the concentration of noradrenaline in the hypothalamus was increased twofold within 1 h, decreasing thereafter, and that of dopamine in these regions was depressed; the concentration of serotonin, which increased initially(by 1 h) in the hypothalamus, dropped significantly below that of controls in both the hypothalamus and frontal cortex at 8 h. The authors considered that these were not neurochemical changes associated with chemical-induced anorexia but that the neurochemical effects of a single exposure to deoxynivalenol are due to peripheral toxicological events, such as vomiting, which can overwhelm the more subtle central feed refusal activity (Prelusky et al., 1992).

Groups of four castrated male specific pathogen-free Yorkshire pigs, aged 8–12 weeks and weighing 15–20 kg, were fitted with an indwelling catheter in the foramen magnum of the brain and given six doses of deoxynivalenol at 0 (saline), 10 (intravenous), or 30 (intragastric) µg/kg bw at 30-min intervals; the concentrations of neurotransmitters were measured in cerebrospinal fluid, as a reflection of brain activity. The cerebrospinal fluid was collected twice a day for 3 days before dosing and at 2-h intervals for 28 h and twice a day for an additional 3 days after dosing. The pigs showed no overt clinical signs of toxicosis, and the doses given did not induce emesis. The spinal fluid samples were analysed for the presence of 3-methoxy-4-hydroxyphenylethylene, 3,4-dihydroxyphenylacetic acid, 5-hydroxyindole acetic acid, homovanillyl alcohol, and homovanillic acid. Analyses for dopamine, normetanephrine, metanephrine, methoxytyramine, serotonin, and vanillic acid showed that they were present in quantities below the detection limit. A rapid, sustained increase (by about 50% in all four pigs) was found in the concentration of 5-hydroxyindole-3-acetic acid after intragastric administration, which remained elevated for up to 20 hours after oral dosing, and to a lesser extent after intravenous dosing, for up to 6 h (LOEL, 0.18 mg/kg bw after gavage and 0.06 mg/kg bw after intravenous administration). The author considered that these results indicated increased serotoninergic activity in the central nervous system and supported the theory of a link between elevated brain serotonin turnover and decreased feed intake. Alterations in dopamine metabolism were also observed. After intragastric dosing, a delayed increase in homovanillyl alcohol concentration (about 20% in three of four pigs) was seen, which suggested to the authors that the dopaminergic response was enhanced as serotoninergic activity diminished. After intravenous dosing, a rapid drop in homovanillic acid concentration (about 30% in three of four pigs) was observed, returning to normal within 36 h after dosing. The author suggested that this result indicated a direct toxic effect, possibly by inhibition of dopamine-metabolizing enzymes (Prelusky, 1993).

The effects of deoxynivalenol were investigated on gastric emptying and intestinal propulsion in groups of eight male ICR mice weighing 25–30 g and groups of eight male Wistar rats weighing 150–200 g and on gastrointestinal myoelectric activity in rats. Gastric emptying and intestinal transit were evaluated after gavage with a milk meal containing a marker (⁵¹CrO₄Na₂), and radiolabel was measured in the stomach and in 10 segments of the small intestine. The myoelectric activity of the antrum, duodenum, and jejunum was assessed by implanting electrodes for long-term electromyographic recording. Deoxynivalenol given orally at 10–1000 µg/kg bw 10 min before the test meal inhibited gastric emptying in a dose-related manner, but administration of 5 µg/kg bw intracerebrovascularly had no effect. Intestinal propulsion was reduced at the highest dose only. The inhibition of gastric emptying induced by deoxynivalenol was antagonized by ondansetron and granisetron given subcutane-ously (50 µg/kg bw) but not by ondansetron given intracerebrovascularly at 10 µg/kg bw. The anti-emetic compounds metaclopramide and domperidone at 1 mg/kg bw subcutaneously and methylsergide, ritanserin, and cisapride at 2 mg/kg bw subcutaneously did not modify the deoxynivalenol-induced inhibition of gastric emptying. In rats, gavage with a 2.5-ml milk meal increased the frequency of antral spike bursts from 1.9 ± 0.9/min in the fasted state to 4.7 ± 0.4/min, and disrupted intestinal migrating motor complexes for 85 ± 11 min. Oral administration of deoxynivalenol at 50–100 µg/kg bw 10 min before the meal did not modify the frequency of antral spike bursts but induced migrating motor complexes on the small intestine after the meal. This effect was reversed by ondansetron at 10 µg/kg bw subcutaneously. The authors concluded that, in rodents, deoxynivalenol inhibits gastric emptying by inducing intestinal migrating motor complexes through a peripheral action at the serotonin 3 receptors (Fioramonti et al., 1993).

In an investigation of the efficacy of several classes of receptor-specific antagonists in blocking the action of neurotransmitters possibly involved in the emetic effect of deoxynivalenol, groups of two castrated male specific pathogen-free Yorkshire pigs aged 6–8 weeks and weighing 15–20 kg were fitted with cannulae in the jugular vein and stomach. Each pig was used in up to four randomly assigned trials over a 4-week period, the time between trials ranging from 5 to 7 days. The ED₅₀ in pigs given deoxynivalenol (purity, > 96%) was 75 µg/kg bw when it was administered intragastrically and 20 µg/kg bw intravenously. After anti-emetic pretreatment, the pigs were given deoxynivalenol at 80 µg/kg bw intravenously or 300 µg/kg bw intragastrically, and the onset of emesis was monitored. Certain specific antagonists of the serotonin₃ receptor (ICS 205-930, BRL 43694 A) were found to prevent deoxy-nivalenol-induced vomiting, indicating that serotonin plays an important role in chemically induced emesis. The serotonin₂-receptor antagonists cyprohepta-dine and sulpiride were also moderately effective, but only at high doses. Compounds with strong anticholinergic activity were effective but apparently act directly at the emetic centre and can thus prevent emesis regardless of the cause. No effect was seen with antihistaminic and antidopaminergic anti-emetics, except those that also have considerable anticholinergic activity, or with intravenously administered chlorpromazine, which has been speculated to block specific receptors in the brain chemoreceptor trigger zone reportedly involved in initiating emesis (Prelusky & Trenholm, 1993).

Groups of four specific pathogen-free castrated male Yorkshire pigs, 8–12 weeks old and weighing 15–20 kg, were fitted with cannulae in the jugular vein and the fundic region of the stomach. The effect of deoxynivalenol (purity, > 96%) on the plasma concentrations of serotonin, its metabolite 5-hydroxyindole-3-acetic acid, and their precursor tryptophan was investigated, as a reflection of an induced peripheral serotoninergic system. Typical values for the plasma concentrations of the three compounds were established, and changes were measured for 8 h after administration of deoxynivalenol intragastrically at 30 or 300 µg/kg bw or intravenously at 10 or 100 µg/kg bw. No effect was found, even at doses sufficient to induce emesis. The author concluded that deoxynivalenol has no peripheral effect that could account for the increased serotoninergic activity reflected by altered feeding behaviour or emesis (Prelusky, 1994).

The competitive potency of deoxynivalenol against several radioactive ligands that have a high affinity for serotonin receptor subgroups, such as [¹²⁵I]lysergic acid, was investigated in a membrane receptor-binding assay in vitro. The brains were removed from 16 castrated male Yorkshire pigs, 12–14 weeks old and weighing 18–22 kg, and dissected on ice. The densities of receptor sites and the displacement profiles in 12 regions of the brain were investigated. Overall, deoxynivalenol had only minimal ability to block the serotonin ligands tested. The median concentration for inhibition of binding was at least 5 mmol/L, and a concentration of 100 mmol/L was ineffective in certain regions. In contrast, several standard serotonin antagonists had a 10³–10⁵ times greater ability than deoxynivalenol to displace binding of these ligands. As these results indicated that deoxynivalenol has only weak affinity for the serotonin receptor subtypes investigated, the author suggested that, unless relatively high concentrations of the toxin are present, its pharmacological effects in vivo may be mediated by mechanisms other than a functional interaction with serotoninergic receptors at the central level (Prelusky, 1996).

The ability of cyproheptadine, a serotonin antagonist at the serotonin₂ receptor and a known appetite stimulant, to attenuate the adverse effect of deoxynivalenol was investigated in three trials with 21-day-old male ICR mice weighing 15–18 g. Groups of 10 mice received diets containing combinations of cyproheptadine and deoxynivalenol (purity, 99%), providing doses of deoxynivalenol of 4–16 mg/kg (equivalent to 0.6–2.4 mg/kg bw per day) and of cyproheptadine of 1.2–20 mg/kg (equivalent to 0.19–3 mg/kg bw). Cyproheptadine was administered in the feed for 2 days before addition of deoxynivalenol, and the two agents were then administered concurrently for 12 days. Cyproheptadine effectively offset the reduction in feed intake caused by deoxynivalenol, but only at certain doses. At a dose of deoxy-nivalenol of 4 mg/kg of feed, the optimal dose of cyproheptadine was 1.2–2.5 mg/kg of feed; at 8 mg/kg of feed, cyproheptadine was required at 2.5 mg/kg of feed; at 12 mg/kg of feed, the required dose of cyproheptadine was 2.5–5.0 mg/kg feed; and at 16 mg/kg of feed, cyproheptadine at 5–10 mg/kg feed was required. At lower doses of cyproheptadine (­ 5 mg/kg of feed), alone or in combination with the lowest dose of deoxynivalenol tested, a modest increase in weight gain was noted, but this was not seen at higher concentrations of deoxynivalenol. The authors concluded that serotininergic mechanisms probably mediate the deoxynivalenol-induced reduction in feed intake. The finding that cyproheptadine significantly attenuated the effect of deoxynivalenol indicates the involvement of the serotonin₂ receptor in this process (Prelusky et al., 1997).

(iii) Role of the chemosensitive area postrema

Conditioned aversion to the taste of saccharin was used to assess the aversive effects of deoxynivalenol in rats and to examine the putative role of the chemosensitive area postrema in the brain. In the first experiment, groups of seven adult male Long Evans rats weighing 330–370 g drank a 0.15% saccharin solution and then received an intraperitoneal injection of deoxynivalenol (purity, 98%) at 0.125 mg/kg bw or the vehicle, propylene glycol, at 0.5 ml/kg bw. In subsequent tests, the rats conditioned with deoxynivalenol had significant (p < 0.01), two- to fourfold lower absolute and relative intakes of saccharin than control rats, which showed a strong preference for the saccharin solution. In the second experiment, the area postrema of the brainstem was ablated by cauterization in groups of six adult male rats; another six rats received sham lesions. After a 10-day recovery, all rats drank the 0.15% saccharin solution for 2 days and were then given an intraperitoneal injection of deoxynivalenol at 0.125 mg/kg bw. In subsequent preference tests, the sham-operated rats showed a significant (p < 0.01) aversion to saccharin, while the area postrema-ablated rats showed a preference for the saccharin solution. The authors concluded that systemic administration of deoxynivalenol to rats after a novel taste induces conditioned taste aversions, which are mediated by the area postrema (Ossenkopp et al., 1994).

(c) In-vitro studies

The haemolytic effects of deoxynivalenol on rat erythrocytes were studied at concentrations of 130, 200, and 250 µg/ml over 11 h. Complete haemolysis was achieved at the two higher concentrations by 10 and 7 h, but the lower concentration induced insignificant haemolysis. This finding suggests that there is a threshold below which the lytic reaction does not occur. An additional test conducted in the presence of mannitol, glutathione, ascorbic acid, alpha-tocopherol, and histidine showed that these compounds inhibited the haemolytic reaction to the toxins. The authors suggested that deoxynivalenol acts on prokaryotic cells in three ways: by penetrating the phospholipid bilayer and acting at the subcellular level; by interacting with cellular membranes; and by free radical-mediated phospholipid peroxidation. More than one mechanism probably operates at the same time (Rizzo et al., 1992).

Deoxynivalenol, fusarenon-X, and nivalenol suppressed the growth of a human hepatoblastoma cell line (HuH-6KK) at 0.15 mg/L in a serum-free medium without an extracellular matrix (1 x 10⁵ cells/ml), while aflatoxins did not inhibit growth even at 5 mg/L. The medium contained insulin, transferrin, ethanolamine, and selenite. The viability and growth of cells was evaluated in a MTT [3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl-2H-tetrazolium bromide] reduction assay. In a chemiluminescence assay, the IC₅₀ for deoxynivalenol was 1.1 mg/L (Isshiki et al., 1992, 1995).

Cellular injury caused by deoxynivalenol was assessed by means of a series of enzymic and functional indexes in 4-h-old cultures of primary rat hepatocytes exposed for 24 h to deoxynivalenol at 10–2500 ng/ml. Clear evidence of cytotoxicity was obtained, as leakage of lactate dehydrogenase and alanine and aspartate amino-transferases was significantly increased (p = 0.05), while intracellular protein concentration and cell viability were significantly decreased (p < 0.01). The severity of morphological effects was dose-related, and the lesions were seen at concentra-tions > 10 ng/ml. Cytotoxic effects occurred in a dose-dependent manner, with a threshold at 50 ng/ml. Cell viability, measured as reduction of 3-(4,5-dimethylthia-zolyl-2)-2,5-diphenyltetrazolium bromide (MTT), and the intracellular protein content were affected at 10 ng/ml, with an IC₅₀ of 1200 ng/ml (Tseng, 1998).

The effects of a 2-week exposure to low concentrations of deoxynivalenol on the structural and functional characteristics of the human colonic adenocarcinoma cell lines Caco-2 and T84 were examined. Scanning electron microscopic analysis of the apical surfaces of Caco-2 cells showed reduction or abnormal formation of brush borders in the presence of deoxynivalenol at 50, 100, or 200 ng/ml. Monolayer integrity, measured as the transepithelial electrical resistance of tight junctions, was studied in Caco-2 and T84 cells cultured on permeable membranes. The integrity of Caco-2 cells was significantly reduced at all three doses of deoxynivalenol. A dose-related increase in permeability to Lucifer yellow, significant at a dose of 100 ng/ml of deoxynivalenol, was also observed in these cells. The integrity of T84 cells was significantly reduced at 100 and 200 ng/ml, and permeability to Lucifer yellow was significantly increased at 200 ng/ml. Alkaline phosphatase activity in Caco-2 cells was reduced from day 6 to day 15 of culture in the presence of 100 or 200 ng/ml of deoxynivalenol, whereas adding 50 or 100 ng/ml of deoxynivalenol for 15 or 20 days significantly decreased sucrase-isomaltase activity. The protein content was attenuated only by treatment with 200 ng/ml throughout the experiment. The results indicate that deoxynivalenol interferes with structural and functional characteristics of differentiation in enterocytes at low doses (Kasuga et al., 1998).

The human haematopoietic progenitors, granulocyte–monocyte colony-forming units (CFU-GM), were grown in culture in the presence of deoxynivalenol at 3, 90, or 300 ng/ml for up to 14 days. Complete inhibition of growth was observed at the highest dose at all times; partial inhibition was seen at 90 ng/ml, and at 3 ng/ml colonies were inhibited at day 7, with some recovery during the next 7 days. The authors proposed that the haematological lesions observed during human intoxication are due to destruction of haematopoietic progenitors such as CFU-GM (Parent-Massin et al., 1994).

Rat CFU-GM were cultured in the presence of deoxynivalenol at 3, 30, or 300 ng/ml for up to 14 days. Complete inhibition of growth was observed at the highest dose at all times. With 30 ng/ml, increased growth of colonies was observed; no toxicity was observed at 3 ng/ml (Parent-Massin & Thouvenot, 1995).

CFU-GM from human umbilical cord blood or from rat bone marrow were cultured in the presence of deoxynivalenol at 4–400 ng/ml for 14 days. Deoxynivalenol rapidly inhibited human and rat CFU-GM in a concentration-dependent manner at doses of 100–400 ng/ml. The IC₅₀ values for human CFU-GM were 12 ng/ml on days 7 and 10 and 16 ng/ml on day14; those for rat CFU-GM were 100 ng/ml on day 7, 60 ng/ml on day 10, and 64 ng/ml on day 14, the human cells being significantly more sensitive than rat cells. The authors reported that previous studies in their laboratory had shown that deoxynivalenol was about 10 times less toxic to human CFU-GM and about 100 times less toxic to rat CFU-GM than T-2 or HT-2 toxin (Lautraite et al., 1997).

As trichothecenes are known to induce haematological disorders such as neutropenia, thrombopenia, and aplastic anaemia in human and animals, the same group also studied the effect of deoxynivalenol at concentrations of 3–75 ng/ml and other trichothecenes in a model of human erythroblastic progenitors. Human cells were as sensitive to trichothecenes as human CFU-GM, except that deoxynivalenol caused only a slight decrease in proliferation at the highest dose (Rio et al., 1997).

The effects of deoxynivalenol at a concentration of 0, 30, 60, or 400 ng/ml for up to 72 h on the expression of the sequentially expressed activation markers CD69, CD25, and CD71 and on proliferation of human CD4⁺ and CD8⁺ lymphocytes provided by three healthy women were studied in culture. After 6 h at the highest concentration, less CD69 was expressed in exposed cultures than in controls, but after 24 and 48 h of exposure, an increased frequency of cells expressing CD69 was found, indicating a delay in down-regulation of CD69 expression. At 24 h, stimulation of CD25 expression was observed at doses below the IC₅₀ value, while suppression was found at higher doses. The pattern differed from that of CD69 expression, in that increased expression of CD25 did not occur after exposure to the highest concentration of the toxin and no stimulation was found after 48 h of exposure, indicating that the response was inhibited and not delayed. The effects of the toxin on CD71 expression were similar to those on CD25 expression. The authors concluded that deoxynivalenol exerts its main antiproliferative action early in the cell cycle, before or in conjunction with CD25 expression. Cell proliferation, measured by bromodeoxyuridine (BrdU) flow cytometry, was inhibited by 8%, 19% and 99% at the three concentrations of deoxynivalenol, respectively (Johannisson et al., 1999). The Committee noted that inclusion of results obtained with a concentration of about 200 ng/ml would have been informative.

Individual differences in sensitivity and the individual and combined effects of deoxynivalenol, T-2 toxin, diacetoxyscirpenol, and nivalenol on production of IgM, IgA, and IgG were studied in human lymphocytes in vitro. All four trichothecenes effectively inhibited mitogen-induced lymphocyte proliferation and immunoglobulin production in a dose-dependent manner, with limited variation in sensitivity between individuals. The IC₅₀ values for immunogloblulin production by deoxynivalenol (approximately 120 ng/ml) were similar to those in the tests for proliferation. Greater IgA synthesis than in controls was observed in cell cultures exposed to the lower doses of deoxynivalenol. Combined exposure to two of the toxins resulted mainly in additive or antagonistic effects, although synergistic effects could not be excluded. The authors concluded that the total intake of type A and B trichothecenes should be taken into account in risk assessments (Thuvander et al., 1999).

2.3.1 Pigs

The history of mycotoxicosis due to scabbed grains, which cause feed refusal and emesis, in monogastric animals was reviewed by Vesonder & Hesseltine (1980). Feed refusal by pigs and equines fed scabbed barley was reported in 1928. In 1972, maize naturally contaminated with F. graminearum resulted in feed refusal and emesis in pigs, and led to the identification of deoxynivalenol (Vesonder et al., 1973).

In 1984, complete feed refusal was observed in weanling pigs on a farm in Queensland, Australia, which was associated with the presence of F. graminearum in wheat, triticale, and barley. These grains contained deoxynivalenol at 34, 10, and < 0.1 mg/kg, respectively, and zearalenone was found at concentrations of 6.2, 2.8, and 0.1 mg/kg of feed. The feed intake of growing pigs was reduced, and young gilts were found to have red, swollen vulvas, indicating that both mycotoxins played a role in this outbreak (Moore et al., 1985).

Feed refusal by pigs was observed in 1986 in Argentina. The concentrations of deoxynivalenol were found to be 1–40 mg/kg in wheat and 1.7–8 mg/kg in formulated feed (Marpegan et al., 1988).

2.3.2 Chickens and other birds

An episode of suboptimal growth, poor feathering, and behavioural abnormalities in broilers in Scotland during the winter of 1980–81 was considered to be associated with mould-contaminated maize and wheat in the feed, from which fusaria were isolated in persistently high numbers. Four species, F. culmorum, F. tricinctum, F. nivale, and F. moniliforme, were identified. Chloroform extracts of the raw materials and of an artificial medium in which three of the Fusarium species were cultured proved toxic to cultures of a human epithelial cell line (HEp II). The mycotoxins deoxynivalenol, zearalenone, and diacetoxyscirpenol were identified by thin-layer chromatography (TLC) in some extracts, and several other areas of the chromatograms were found to be toxic to the HEp II cell system. These may have contained toxins for which standards were not available or, alternatively, previously uncharacterized fungal metabolites. The authors concluded that the toxins produced by the fusaria were major contributing factors to the symptoms in the birds (Robb et al., 1982).

In 1987, an epizootic at a wildlife research centre in Maryland, USA, caused symptoms in 80% of 300 captive whooping cranes (Grus americana) and sandhill cranes (G. canadensis) and the deaths of 15 of these birds. Gross examination revealed dehydration, atrophy of fat, renal insufficiency, and small spleens, which were considered inconclusive findings. Extensive testing resulted in isolation of Fusarium spp. from constituents of the grain-based diet, and low concentrations of T-2 toxin (1–2 mg/kg) and deoxynivalenol (0.4 mg/kg) were isolated from pelleted feed (Olsen et al., 1995).

2.4.1 Clinical observations

The acute effects of deoxynivalenol—nausea, vomiting, diarrhoea, abdominal pain, headache, dizziness, and fever—can develop within 30 min of exposure and are difficult to distinguish from gastrointestinal conditions attributed to microbes, such as the preformed emetic toxins from Bacillus cereus. No deaths attributed to deoxynivalenol have been reported in humans (Luo, 1988).

2.4.2 Epidemiological studies

Acute outbreaks of red mould toxicosis affecting humans and involving F. graminearum have been reported in China, India, and Japan. Deoxynivalenol has also been investigated for its possible role in chronic diseases such as oesophageal cancer, stomach cancer, liver cancer, and a form of osteoarthritis. Some of these studies were reviewed previously (IARC, 1993; Kuiper-Goodman, 1994; Wild & Hall, 1996).

About 35 outbreaks of acute human illness were reported in China between 1961 and 1985 that were attributed to consumption of scabby wheat and mouldy maize, with at least 7818 victims. Typically, the persons became ill 5–30 min after consumption, with symptoms of nausea, vomiting, diarrhoea, abdominal pain, headache, dizziness, and fever. No deaths were reported. In an outbreak in 1984 in Xingtai County, 362 of 383 (94%) persons who ate mouldy maize became ill. Analysis by TLC of five samples associated with symptoms in this outbreak, with approximate limits of detection (LODs) of 0.1, 0.04, and 0.05 mg/kg for deoxynivalenol, T-2 toxin, and zearalenone, respectively, indicated the presence of deoxynivalenol at 3.8–93 mg/kg and zearalenone at 0.13–0.59 mg/kg in four samples; one sample contained deoxynivalenol at 0.34 mg/kg and zearalenone at 0.004 mg/kg; neither T-2 toxin nor nivalenol was found. The authors reported the presence of deoxynivalenol at a concentration of 1–40 mg/kg in scabby wheat collected from three villages and significantly higher concentrations of deoxynivalenol in wheat samples collected during the food poisoning incident than in samples not associated with the incident (Luo, 1988). The Committee noted that deoxynivalenol was probably responsible for the mycotoxicoses involving scabby wheat and mouldy maize, as zearalenone is relatively non-toxic after a single exposure.

No cases of acute illness were observed in Henan, China, in 1985 among 191 peasant families who ate scabby wheat containing deoxynivalenol at a concentration of 0.016–3.3 mg/kg (mean, 0.92 mg/kg) and nivalenol at a mean concentration of 0.13 mg/kg (both measured by GC–ECD). On the basis of a loss of deoxynivalenol during processing of about 30% and an intake of 560 g/person, the authors estimated an intake of deoxynivalenol of 380–520 µg/adult, which, for a body weight of 50 kg, would give an intake of 7.5–10 µg/kg bw. The authors indicated that the concentrations of deoxynivalenol in the following year were 0.007–0.18 mg/kg (Guo et al., 1989).

In Linxian, China, a high-risk area for oesophageal cancer (mortality rate, 132 per 100 000), no cases of acute disease were observed among oesophageal cancer patients who consumed a staple diet containing maize and maize meal containing deoxynivalenol at concentrations of 0.36–13 mg/kg (mean, 5.4 mg/kg) and nivalenol at 0.054–2.8 mg/kg (mean, 0.76 mg/kg), both analysed by TLC and HPLC, in 1985–86 (Hsia et al., 1988). The Committee noted that, as the intake of maize was not given, the intake of deoxynivalenol could not be estimated.

Studies conducted in 1989 to compare maize samples collected randomly from families with oesophageal cancer patients in Linxian with samples from families with no such cancer in Shangqiu (mortality rate from this cancer, 16 per 100 000) indicated that the mean concentration of deoxynivalenol in maize was 0.57 mg/kg (range, 0.017–3.5 mg/kg) in Linxian and 0.099 mg/kg (range, 0.011– 0.61 mg/kg) in Shangqiu. The samples were analysed by GC–ECD (Luo et al., 1990).

The natural occurrence of mycotoxins was compared in staple foods from high- and low-risk areas for oesophageal cancer in China. A total of 54 samples of maize and 40 samples of wheat intended for human consumption were collected during January and February 1995 from Linxian and Shangqui counties in Henan Province, high- and low-risk areas for oesophageal cancer, and were analysed for fumonisins and trichothecenes by GC–MS and for zearalenone by HPLC. The prevalence of trichothecenes and zearalenone in maize samples from the high-risk area was 3.7- and 11-fold higher (p < 0.01), respectively, than that in maize from the low-risk area. Significantly higher mean concentrations were found in the high-risk area than in the low-risk area, for deoxynivalenol (0.4 versus 0.05 mg/kg), 15-acetyldeoxynivalenol (0.24 versus undetected), nivalenol (0.086 versus 0.059), and zearalenone in maize and of deoxynivalenol (0.08 versus 0.04 mg/kg) and nivalenol in wheat. Fumonisins were found in 79% of maize samples from the high-risk area and 50% of samples from the low-risk area, but the concentration was similar: about 3.4 mg/kg. The authors concluded that the concentrations of trichothecenes and zearalenone correlated with the incidence of oesophageal cancer (Gao & Yoshizawa, 1997).

Fungal and mycotoxin contamination of 220 maize or wheat samples and 34 maize samples from four locations in Cixian County, an area with a high incidence of oesophageal carcinoma, was analysed in 1990–94; 26 maize samples collected from an area with a relatively low incidence of oesophageal carcinoma in Zanhuang County, with similar dietary customs, were analysed for mycotoxins in 1990. The maize and wheat were severely contaminated, F. moniliforme and Penicillium spp. being the predominant fungi in maize and these as well as Alternaria alternata and Aspergillus flavus in wheat. HPLC showed low concentrations of deoxynivalenol (0.05–0.17 mg/kg) in the four areas with a high incidence of oesophageal carcinoma and none in the control area. The authors concluded that fungal and mycotoxin contamination of foodstuffs in Cixian County is common (Zhang et al., 1998). The Committee noted that the units may be in error since the values appear to be lower than expected.

Consumption of fermented maize pancakes was associated with increased rates of death from stomach cancer in rural Linqu County in Shandong Province, China. To determine whether mycotoxins accounted for the increased risk, specimens of maize, maize meal, unfermented and fermented pancake batter, and cooked fermented pancakes from 16 households in five villages in Linqu County were obtained in 1996. The samples were analysed for aflatoxins, fumonisins, zearalenone, deoxynivalenol, and 15-acetyldeoxynivalenol, the limit of detection for the last two compounds being 0.5 mg/kg. No aflatoxin was detected, but fumonisins were detected in 6–19% of the maize products at concentrations of 0.6–7.2 mg/kg. Deoxynivalenol and 15-acetyldeoxynivalenol were detected in 58 and 17% of the raw maize specimens, respectively, and zearalenone was detected in 15% of the maize meal specimens. Analysis for deoxynivalenol showed that 33% of the raw maize samples and 21% of the cooked pancakes contained > 1 mg/kg (maximum, 2.7 mg/kg and 1.5 mg/kg, respectively). The concentrations of mycotoxins did not increase with fermentation. The authors concluded that the concentrations were similar to or lower than those found in the USA, and did not increase the risk for gastric cancer among people who consumed fermented pancakes (Groves et al., 1999). The Committee noted that differences in the rate of stomach cancer among the five villages were not reported and a control area with a low rate of stomach cancer was not included. The detection limits for the mycotoxins were high. The amount and frequency of pancakes consumed was not given, and this could be considerably greater than that in the USA.

The natural occurrence of aflatoxin B₁, fumonisins, and trichothecenes was investigated in maize samples harvested in areas of low and high risk of primary liver cancer, Penlai (Sandong) and Haimen (Jiangsu), respectively, in China, during 1993. In Haimen, 40 samples contained a mean concentration of deoxynivalenol of 0.89 mg/kg, and in Penlai, 13 kernel samples contained a mean of 0.49 mg/kg. The authors concluded that co-contamination of Chinese maize with aflatoxin B₁, fumonisins, and deoxynivalenol is common (Wang et al., 1995).

A total of 35 maize and wheat samples from several areas of high and low incidence of Kashin-Beck disease (endemic osteoarthritis deformans) in China were surveyed in 1989 for contamination with mycotoxins, including fusarochromanone, produced by F. equiseti. Trichothecenes were analysed by GC–ECD (LOD, 0.005 mg/kg; recovery, > 96%). Deoxynivalenol was found at significantly higher concentra-tions in all high-incidence areas (range, 0.005–3.9 mg/kg) than in low-incidence areas (range, 0.002–0.7 mg/kg), with mean concentrations of 0.25 versus 0.05 mg/kg, 1.0 versus 0.27 mg/kg, and 0.51 versus 0.18 mg/kg in three areas of comparison. In addition, the concentrations of 15-acetyldeoxynivalenol and 3-acetyldeoxynivalenol, while somewhat lower than those of deoxynivalenol and nivalenol (which were about 10% those of deoxynivalenol) were also significantly higher in all high-incidence areas than in low-incidence areas. Fusarochromanone was not identified, regardless of the incidence of the disease (Luo et al., 1992).

In 1987, an acute outbreak of disease, affecting about 50 000 persons in the Kashmir valley in India, was attributed to consumption of bread made from wheat damaged by rain. The wheat was reported to contain several trichothecenes, including deoxynivalenol (0.34–8.4 mg/kg in 11 of 24 samples), acetyldeoxynivalenol (0.6–2.4 mg/kg in 4 of 24 samples), nivalenol (0.03–0.1 mg/kg in 2 of 24 samples), and T-2 toxin (0.55–4 mg/kg in 3 of 24 samples) (Bhat et al., 1989a,b). Interviews with about 150 affected families revealed that only persons who had eaten wheat products had become ill, showing mainly mild gastrointestinal tract symptoms for about 2 days. The authors estimated a NOEL of 0.44 µg/kg bw per day, on the basis of the lowest concentration of deoxynivalenol found in wheat (0.34 mg/kg), an average intake of 67 g of wheat products, and a mean body weight of 52 kg (Bhat et al., 1987). The Committee noted that, as samples were not collected until 4 months after the outbreak, a clear association of specific samples with specific cases of illness or lack of illness was not established. Furthermore, many of the collected samples contained other, more toxic trichothecenes, which may have contributed to the overall disease profile Therefore, the actual NOEL for trichothecene-associated human illness, expressed as equivalents of deoxynivalenol, could not be identified from these studies.

Deoxynivalenol has a 12,13-epoxy group, three OH functions, and an alpha,beta-unsaturated keto group. Its chemical name is therefore 12,13-epoxy-3alpha,7alpha,15-trihydroxy trichothec-9-ene-8-one. Its molecular formula, C₁₅H₂₀O₆, shows that its relative molecular mass is 296.3. Deoxynivalenol crystallizes as colourless needles, with a melting-point of 151–153 °C. Its specific rotation has been determined to be [alpha]²⁰_D= +6.35. The alpha,beta unsaturated keto function results in absorption of ultra-violet (UV) radiation of short wavelength, but the UV spectrum of deoxynivalenol is not characteristic. As it is a type B trichothecene, deoxynivalenol is soluble even in water and in polar solvents such as aqueous methanol, aqueous acetonitrile, and ethyl acetate. Deoxynivalenol is stable in organic solvents (Shepherd & Gilbert, 1988), but ethyl acetate and acetonitrile are the most suitable solvents, particularly for long-term storage (Pettersson, 2000).

The 12,13-epoxy group is extremely stable to nucleophilic attack, and deoxy-nivalenol is stable at 120 °C and is not decomposed under mildly acidic conditions. The three hydroxyl groups can be derivatized (e.g. esterified), for instance before GC analysis.

Figure 2 gives the structures of the five common type B trichothecenes, deoxynivalenol, nivalenol, 3-acetyldeoxynivalenol, 15-acetyldeoxynivalenol, and fusarenon X, with the typical basic tetracyclic, sesquiterpenoid ring system.

As the 12,13-epoxy trichothecenes are closely related, physicochemical analytical methods are usually designed to determine more than one trichothecene. However, the naturally occurring trichothecenes in cereals can be divided into polar substances carrying a keto group at C₈ (type B trichothecenes) and less polar type A toxins which contain no keto function at C₈ and generally fewer free hydroxyl groups. Hence, the analytical procedures usually differ in the extraction, clean-up, and end determination steps, depending on which group of trichothecenes is to be analysed.

Several methods for trichothecenes based on TLC are included in the Official Methods of Analysis of AOAC International. HPLC and GC methods were reviewed in detail by Scott et al. (1993) and Langseth & Rundberget (1998). The analysis of mycotoxins, including trichothecenes, has also been reviewed by Cole (1986), WHO (1990), Chu (1991), Steyn et al. (1991), Gilbert (1992a), Scott (1995), Trucksess (1995), and Lawrence & Scott (2000). Various immunochemical methods, especially enzyme-linked immunosorbent assays (ELISAs), have been established for the determination of selected trichothecenes, as reviewed by Morgan (1989), Candlish (1991), and Park & Chu (1996). The method chosen depends on the instrumentation available, the required detection limit, the matrix composition, and the analyte properties.

The choice of analytical method also depends on the availability of appropriate calibrants of defined concentration. Comparative studies (Pettersson, 1998; Josephs et al., 2001) clearly show that determination of the concentration of calibrant solutions is still a serious problem in the analysis of deoxynivalenol, resulting in a coefficient of variation (CV) > 20% (Schuhmacher et al., 1996). Studies in which a common calibrant was provided to participants resulted in much better agreement for naturally contaminated samples (CV = 19%). A major task of the Standard, Measurement and Testing project of the European Union (Pettersson, 1998) was to check the purity of calibrants for trichothecenes and to develop a common procedure.

The purity of trichothecenes is often claimed to be high (> 95%), but the bound solvent or water is not always taken into account. The concentration of dissolved trichothecene calibrants should therefore be checked, e.g. by spectrophotometry. Owing to the extensive absorption below lambda = 220 nm, methanol is not suitable for UV determination, especially of type A trichothecenes, which have an absorption maximum around 205 nm in methanol. Therefore, for spectrophotometric determina-tion of both A and B trichothecenes, acetonitrile is the recommended solvent (Pettersson, 1998). Values for the molar absorptivity of B trichothecenes ranging from 4500 to 7040 are found in the literature, but the Standard, Measurement and Testing project suggested use of 6400 as the molar absorptivity for all type B trichothecenes in acetonitrile, and a spectrophotometric method for assessing the purity of solid A and B trichothecenes has been developed within the framework of the project. In addition to the UV method, HPLC separation is recommended for determination of impurities, with a final comparison of the concentrations of new and old calibrants by GC. Owing to the lack of an UV-absorbing keto-group and with the double bond between C-9 and C-10, which allow B trichothecenes to absorb at 219 nm relatively well, spectrophotometric determination of concentrations cannot be recommended for A trichothecenes.

3.2.1 Screening tests

The commonest methods for final separation and detection of mycotoxins such as trichothecenes in agricultural commodities are still based on TLC. This method is simple and economical, and, with the introduction of high-performance TLC and scanning instruments, the efficiency and precision of separation are comparable to those achieved with other types of chromatography. TLC is still commonly used for the final separation step in the determination of deoxynivalenol. Reagents such as sulfuric acid and para-anisaldehyde are required, however, to visualize the non-fluorescent, short wavelength (lambda_max = 220 nm) absorbing deoxynivalenol (Scott et al., 1970; Ueno et al., 1973). Besides the spray reagents which are specific for the 12,13-epoxy group of the trichothecenes (e.g. 4-para-nitrobenzylpyridine or nicotinamide in combination with 2-acetylpyridine), AlCl₃ is probably the most useful reagent for visualizing deoxynivalenol and other type B trichothecenes. AlCl₃ has been tested in a collaborative study for the determination of deoxynivalenol by AOAC International (Eppley et al., 1986).

ELISAs have been developed for rapid screening or quantification of tricho-thecenes in cereals, and both ELISA and TLC are used for rapid detection of deoxynivalenol. In less complex matrices, such as pure cereals, these methods can also be used for quantification of deoxynivalenol. ELISAs involve use of specific antibodies, which are derived by a series of complex procedures including immunogen synthesis, immunization of animals, isolation, and characterization of antibodies. ELISAs are selective, sensitive, rapid, and easy to use, and clean-up, if any, is minimal. Owing to the multiple properties of trichothecenes, however, production of useful antibodies against these compounds has proved difficult. A wide range of ELISAs has been developed that involve polyclonal IgG and IgY antibodies, and monoclonal antibodies have been elicited for use in direct and indirect competitive assays. Usually, acetonitrile and water, methanol and water, pure methanol, or water is used as the extraction solvent.

A rapid method has been introduced in which Mycosep is used for clean-up, with fluorimetric detection of deoxynivalenol derivatized with zirconylnitrate and ethylenediamine in methanol (Malone et al., 1998).

Trichothecenes Appendix 1 lists the performance of tests that have been developed to screen for deoxynivalenol. The most sensitive ELISA methods have been developed for 3-acetyldeoxynivalenol, with an LOD of 0.3–1 ng/g, which requires acetylation of the toxin in cereal extract before assay for deoxynivalenol and therefore results in determination of the sum of deoxynivalenol and its acetylated derivatives. ELISAs have also been developed to enable direct determination of deoxynivalenol, but with less sensitivity (LOD = 20–300 ng/g). Accurate quantification of deoxynivalenol by immunological assays is often limited because of the remarkable cross-reactivity of deoxynivalenol-related compounds (Xu et al., 1988; Park & Chu, 1996). Nevertheless, when chromatographic instruments are not available, ELISAs are an interesting alternative for determining deoxynivalenol. Typical detection limits for deoxynivalenol by TLC are 20–300 ng/g.

The one-step solid-phase extraction clean-up and subsequent fluorimetric analysis of deoxynivalenol in grains allows detection within less than 30 min with an LOD of 100 ng/g. No values have yet been published, however, for the recovery or the precision of the method.

A comparative study organized by Schuhmacher et al. (1996) showed that participating laboratories in which ELISAs were used had difficulty in determining deoxynivalenol in pure organic solvents, with a CV of 21%, for example. In addition, overestimates and matrix dependence were observed. In a study carried out 1998 (Josephs et al., 2001), laboratories in which ELISA methods were used to determine deoxynivalenol found significantly higher concentrations than those in which HPLC and GC were used. The reason for the higher values found in maize samples in one study was additional contamination of the sample with 15-acetyldeoxynivalenol at a concentration of 477 ng/g, determined by GC–ECD. This compound cannot be distinguished from deoxynivalenol with most ELISA systems.

3.2.2 Quantitative methods

The relatively polar deoxynivalenol and other type B trichothecenes are usually extracted by mechanical shaking or blending with aqueous acetonitrile or aqueous methanol. Other solvents, such as water and polyethylene glycol, chloroform and ethanol, and chloroform and methanol have also been used. Use of aqueous acetonitrile provides cleaner extracts than use of aqueous methanol. In an assessment of extraction procedures for deoxynivalenol, longer extraction times were required for naturally contaminated samples than for those that had been spiked (Trenholm et al., 1985). It is therefore recommended that extraction efficiency be evaluated with naturally contaminated samples when possible.

While clean-up is usually not required for immunoassays, physicochemical methods commonly involve extensive clean-up procedures. The main methods used in the analysis of trichothecenes are liquid–liquid partitioning, solid-phase extraction, column chromatography, immunoaffinity columns, and multifunctional clean-up columns. If necessary, interfering lipids can be removed by extracting the sample with n-hexane or another non-polar solvent before further clean-up.

Liquid–liquid partitioning is conventionally performed by shaking the sample extract with a non-miscible solvent in a separation funnel. For this type of clean-up, the aqueous phase can be supported on a column packed with solid hydrophilic matrix (ClinElut® or Extrelut®) and the organic solvent percolated down the column to elute the purified toxin.

Column chromatography can involve use of various stationary phases (silica gel, aluminium oxide, Florisil, charcoal, and C₈ or C₁₈ reversed phases) corresponding to the required polarity range of the adsorbent. The column packing material used most frequently for deoxynivalenol is a mixture of charcoal, alumina, and Celite. Modern solid-phase extraction columns are a potential alternative to the conventional column chromatographic methods. Solid-phase extraction columns are generally delivered prepacked in disposable plastic cartridges and are available with the adsorbents listed above. Gel permeation chromatography has also been used for clean-up in the analysis of deoxynivalenol.

Application of immunoaffinity columns for purification before instrumental methods has been used successfully for several mycotoxins. When the aqueous sample extract is applied to the immunoaffinity column, the analyte molecules are bound to antibodies which are linked to an organic carrier material. Before washing, the toxins can be eluted by denaturating the antibodies with pure organic solvents. However, immunoaffinity columns for most trichothecenes are not available commercially, and the columns are applicable to only one toxin. Furthermore, low recovery of deoxynivalenol was found with a commercial immunoaffinity column (Scott & Trucksess, 1997). An HPLC method involving both a charcoal–alumina–Celite column and a commercial immunoaffinity column gave reasonable recoveries, with an LOQ of 50 ng/g for cereals (Reutter, 1999).

Supercritical-fluid extraction combined with GC–ECD appears to be suitable for the determination of deoxynivalenol and related trichothecenes (Josephs et al., 1998; Krska, 1998). The heavy investment and maintenance costs for a supercritical-fluid extraction apparatus and the poor recovery of deoxynivalenol (50%) remain problems, however, which limit the applicability of these methods in routine analysis.

Another development in clean-up methods is the simple, rapid, multifunctional clean-up column MycoSep™ (Romer, 1986; Weingärtner et al., 1997). These columns consist of packing material containing various adsorbents, such as charcoal, Celite, and ion-exchange resins, which are housed in a plastic tube between filter discs, with a rubber flange at the lower end containing a porous frit and a one-way valve. When the column is inserted into a culture tube, the flange seals tight, thus forcing the extract through the packing material. The pure extract appears at the top of the plastic tube. The Mycosep® column allows sample purification within 10–30 s. A major advantage of this column is that no time-consuming rinsing steps are required, as in solid-phase extraction. In addition, nearly all the interfering substances are retained on the column, and the trichothecenes are not adsorbed onto the packing material. Mycosep® columns are among the most frequently used commercial columns for clean-up of deoxynivalenol in combination with HPLC and GC.

ELISA and TLC are used for the quantitative determination of deoxynivalenol. GC is widely used, particularly for simultaneous determination of several trichothecenes as their trimethylsilyl, pentafluoropropionyl, heptafluorobutyryl, and trifluoroacetyl derivatives, with ECD, flame ionization detection, MS, or tandem MS detection. The choice of derivatization reagent depends on the type of trichothecene to be analysed and the method of detection. The conjugated carbonyl group makes type B trichothecenes sensitive to ECD, while enhanced sensitivity of type A trichothecenes, which lack this group, is obtained when fluoroacylation is used. Trimethylsilylation allows more selective derivatization of type B trichothecenes, with less background interference than with fluoroacylation or derivatization with heptafluorobutyryl- or pentafluoropropionylimidazole (Poole, 1978). Trimethylsilyl ethers are made by derivatizing all hydroxyl groups with reagents such as N,O-bis(trimethylsilyl)acetamide, trimethylchlorosilane, and trimethylsilylimidazole. When a single compound is used, however, two peaks may arise for type B trichothecenes, owing to incomplete derivatization. Such problems can be avoided by using derivatization mixtures such as TRI-SIL TBT® and Sylon BTZ, which contain trimethylsilylimidazole (40 ± 5%), N,O-bis(trimethylsilyl)acetamide (35 ± 5%), and trimethylchlorosilane (25 ± 5%).

Several suitable HPLC methods have been published for the determination of type B trichothecenes in food and cereals. Separation is usually achieved on a C₁₈ reversed-phase column with methanol–water mixtures as the mobile phase. Use of acetonitrile (UV cut-off, 190 nm) instead of methanol (UV cut-off, 210 nm) in aqueous mixtures is preferable, however, in view of the convenient cut-off value and better transmission at the important 220-nm wavelength. Deoxynivalenol and nivalenol have been determined by HPLC with UV detection at 222 nm. Nevertheless, time-consuming clean-up is required for the determination of deoxynivalenol in complex matrices, such as food and feed. A German standard method for determination of deoxynivalenol, which is being validated in a second round trial, involves an HPLC method that includes a time-consuming three-step clean-up with liquid–liquid partitioning with petroleum ether, solid-phase extraction with a charcoal–alumina–Celite mixture, and use of an immunoaffinity column.

Analytical performance characteristics comparable to those of GC methods can be achieved by HPLC combined with pre- or post-column derivatization (Maycock & Utley, 1985). The post-column derivatization procedure that has been developed by Sano et al. (1987) involves alkaline decomposition of deoxynivalenol and nivalenol to generate formaldehyde, and subsequent reaction with acetoacetate and ammonium acetate to form a fluorescent derivative. In view of the problems associated with use of GC, this method is an interesting alternative for reliable, accurate determination of deoxynivalenol, even at the lower range of nanograms per gram. LC–MS instruments, particularly those with atmospheric pressure chemical ionization interfaces, have been used for the determination and identification of trichothecenes, including deoxynivalenol, at trace concentrations (Razzazi-Fazeli et al., 1999). Owing to the decreasing cost of suitable LC–MS instruments, this technique is also being used for the determination of mycotoxins. ECD has also been described for the determination of deoxynivalenol (Sylvia et al., 1986).

The method used in a recent comparative study (Josephs et al., 2001) reflects the techniques being used by most European laboratories involved in analysing deoxynivalenol. The trial for the determination of deoxynivalenol in cereals comprised 28 laboratories in Europe, Singapore, and the USA. Deoxynivalenol was purified from the raw extracts by solid-phase extraction on MycoSep™ or Florisil-active charcoal columns (nine laboratories). Two participants used immunoaffinity columns, and one used the Extrelut® technique in combination with solid-phase extraction on Fluorisil for clean-up. Deoxynivalenol was determined by GC–ECD (five laboratories), GC–MS (two laboratories), HPLC–UV–diode array detection (six laboratories), and HPLC–fluorescence detection (one laboratory). Four laboratories used ELISA for determination of deoxynivalenol. It should be emphasized that the ELISA methods used for determination of deoxynivalenol cannot distinguish between deoxynivalenol, 3-acetyldeoxynivalenol, 15-acetyldeoxynivalenol, and 3,15-diacetyldeoxynivalenol, since the antibodies are elicited against 3,7,15-triacetyldeoxynivalenol. The LODs of the methods used by the participants ranged from 0.30 to 110 ng/g, the recoveries varied from 60 to 100%, and the reported CVs of the methods ranged from 3 to 15%.

In a comparative study organized by Gilbert (1992a), the participants who used TLC for determination of deoxynivalenol found interference in extracts of maize. Accordingly, most of the participants changed from TLC to HPLC.

(d) Performance characteristics

Trichothecenes Appendix 2 lists the performance characteristics of quantitative methods for trichothecenes. Figure 3 shows a typical chromatogram obtained after separation and detection with GC–ECD after trimethylslylation of five B trichothecenes. The method thus allows simultaneous quantification of several B trichothecenes. The typical detection limits of quantitative methods for the determination of deoxynivalenol in cereals are 100–1600 ng/g (HPLC–UV), 6–40 ng/g (HPLC–MS), 20 ng/g with HPLC–fluorescence detection) 20–50 ng/g and even lower with GC–ECD, and down to approximately 5 ng/g with GC–MS with heptafluorobutyryl and pentafluoropropionyl derivatives. Typical recoveries were 70–110%. The results of the comparative studies (Schuhmacher et al., 1996; Pettersson, 1998; Josephs et al., 2001) indicate that the analysis of the most relevant Fusarium mycotoxin, deoxynivalenol, is still not satisfactory. This is clear in the comparison of determinations of deoxynivalenol within the framework of the Standard, Measurement and Testing project organized by Pettersson (1998) involving 20 participants from 12 countries, in which no common calibrant was provided and for which the CV was 60%. The values for a naturally contaminated wheat sample ranged from 185 to 1100 ng/g, after rejection of outliers. In general, the results of the study showed relatively wide variation in both repeatability and reproducibility for all trichothecenes analysed. The Community Bureau of Reference comparative studies (Pettersson, 1998) and studies for the certification of deoxynivalenol in wheat and maize were different (Gilbert, 1992a,b), since mainly HPLC methods were used.

The main objective of the project funded by the Commission of the European Union is to improve the analysis of trichothecenes with GC methods and to obtain better agreement among laboratories. The main methodological problems in the project coordinated by Pettersson (1998) were:

Although a standardized GC method for deoxynivalenol and other trichothecenes is needed, these problems and sources of variation must be addressed. Otherwise, the method will not be robust enough and will result in wide variations in reproducibility and repeatability.

Generation of meaningful survey data requires collection of representative samples from carefully selected batches of food which, in turn, are representative of clearly defined locations (e.g. country, region within a country). For general aspects of the sampling protocols see Trichothecenes Appendix 6.

The variability of measurements of deoxynivalenol in barley and wheat has been studied by Freese et al. (2000) and Whitaker et al. (2000), respectively. Freese et al. collected 225-kg bulk samples from six batches of barley and riffle-divided each sample into 16 samples of 0.1 kg, 16 samples of 0.8 kg, and 16 samples of 7 kg. The samples were comminuted in a Romer mill, and 50-g portions were taken from each sample for determination of deoxynivalenol. An evaluation of the analytical results indicated that the variation associated with sample preparation and analysis was of greater significance than sampling variance for all sample sizes and that the variation was not significantly reduced by increasing the sample size. In a further experiment, 10 samples of about 2.5 kg were taken from each of 10 truck loads of barley, by sampling methods prescribed by the Grain Inspection, Packers and Stockyards Administration (1995). Each 2.5-kg sample was comminuted in a Romer mill, and two 50-g portions were taken from each sample. A single determination of deoxynivalenol was made on the extract from the first portion, whereas duplicate determinations were performed on the second portion, by an ELISA procedure. In this instance, an evaluation of the variance associated with the sampling, sample preparation, and analytical steps indicated an approximately equal contribution from each step. It was concluded that sample sizes of 100–200 g were adequate, assuming that they were obtained by riffle division of a large bulk sample. However, it was also concluded that batches might be stratified to different degrees, depending on the amount of mixing during handling, and that stratification could have a significant effect on sampling variance.

Whitaker et al. (2000) adopted a similar approach in studying the variances in sampling, sample preparation, and analysis during determination of deoxynivalenol in wheat. A 20-kg bulk sample was taken from each of 24 commercial batches, and each sample was riffle-divided into 32 samples weighing 0.45 kg (1lb). Each 0.45-kg sample was finely comminuted in a Romer mill, which was set to produce a representative, comminuted 25-g portion automatically. Deoxynivalenol was determined in each of 768 (24 x 32) 25-g portions by the Romer FluoroQuant™ fluorimetric procedure, and the analytical data were used to determine total variance (sampling, sample preparation, and analysis). Twenty comminuted samples, with a wide range of concentrations of deoxynivalenol, were selected from the residual 768 comminuted samples, and eight 25-g portions were taken from each sample by riffle division. The Romer method was then used to determine the deoxynivalenol concentration in four replicate extracts prepared from each 25-g portion. The combined variance for sample preparation and analysis and the analytical variance alone were estimated by the SAS procedures (Statistical Analysis System Institute, Inc., 1997). The total CV varied from 260 (batch concentration, 0.02 mg/kg deoxynivalenol) to 7.9 (14 mg/kg). For a batch concentration of deoxynivalenol of 5.0 mg/kg, the CVs associated with sampling, sample preparation, and analysis were 6.3, 10, and 6.3%, respectively. The sampling variance was specific to a 0.45-kg sample, the sample preparation variance to a Romer mill and a 25-g analytical sample, and the analytical variance to the Romer FluoroQuant method. The total variance was 13%. The low variance associated with the sampling step (relative to those of other mycotoxins and other commodities) is due partly to the small kernel count of wheat (about 30 kernels per gram), which is about 10 times higher than that of shelled maize and 30 times higher than that of shelled groundnuts.

During preharvest of crops with Fusarium head blight, the fungi kill some developing seeds, resulting in shrunken, shrivelled kernels of low weight. In general, at concentrations up to 1 mg/kg mycotoxins such as deoxynivalenol are typically found near the surface of the kernel, whereas at high concentrations they may be more evenly distributed (Charmley & Prelusky, 1994). Many of the shrivelled kernels can be removed by the use of gravity separators, which separate particles on the basis of differences in specific gravity, size, shape, and surface texture. The concentrations of deoxynivalenol after storage were always lower in samples from which the infected kernels were removed before storage than in samples that contained infected kernels (Wilcke et al., 1999). The disappearance of deoxynivalenol during cleaning of grain, such as removal of infected kernels and washing, has been reported to be up to 74% (Charmley & Prelusky, 1994); however, the degree of decontamination varies among studies. Washing barley and maize three times in distilled water reduced the deoxynivalenol concentration by 65–69%. Using 1 mol/L sodium carbonate solution for the first wash reduced the concentration by 72–74%. Soaking barley, maize, or wheat in a 0.1 mol/L sodium carbonate solution for 24–72 h caused a 42–100% reduction in toxin concentration.

Natural degradation of deoxynivalenol has been observed in cereal grain both in the field and during storage. Several explanations have been put forward for how the concentrations of mycotoxins are decreased in a natural system. The concentra-tion of free mycotoxin in natural ecosystems may be the result of concurrent synthesis, transport, conjugation, release from bound forms, and degradation by the plant or by other microbes (Karlovski, 1999).

The effectiveness of milling practices in reducing trichothecene concentrations in flour fractions from that in whole grain has been reviewed (Patey & Gilbert, 1989; Scott, 1991; Charmley & Prelusky, 1994). Milling usually resulted in a higher concentration of deoxynivalenol in bran, shorts, and feed flour and lower concentra-tions in straight-grade flour. The distribution of deoxynivalenol in the various milling fractions of wheat depends to a large extent on the degree of fungal penetration of the endosperm (Nowicki et al., 1988). In other words, milling grain in which the deoxynivalenol contamination is located predominantly at the surface of the kernel would result in flour with a low deoxynivalenol concentration. The ability of fungi to penetrate the kernel appears to vary among different varieties of wheat. Dry milling of maize containing deoxynivalenol gave a higher concentration in germ meal than wet milling, after which the highest concentrations were in steep liquor and gluten.

The trichothecenes are stable at 120 °C, moderately stable at 180 °C, and decompose within 30–40 min at 210 °C (Kamimura, 1989). Neira et al. (1997) showed an average reduction of 44% in the deoxynivalenol concentration in dough and in the final baked products. Approximately half of the reduction occurred after the dough fermentation step.

In a study of the effect of high-temperature and high-pressure processing of foods spiked with deoxynivalenol, no significant reduction was found after processing in extruded maize grits, extruded dry dog food, or autoclaved moist dog food. Autoclaved cream-style maize showed a reduction of only 12% (Wolf-Hall et al., 1999).

During cooking of noodles and spaghetti, trichothecenes may leach into the boiling water to a considerable extent (Scott, 1991).

Trichothecenes are not stable in the presence of alkali. Only 18–28% of deoxynivalenol in maize was retained during tortilla fabrication, in which the maize was first boiled in calcium hydroxide solution (Abbas et al., 1988).

Deoxynivalenol was not metabolized by yeast strains of technological relevance (Böswald et al., 1995).

Microbiological transformation could be used to transform deoxynivalenol to less toxic metabolites such as the de-epoxy metabolite (Yoshizawa et al., 1986; Figure 4). However, only mixed cultures have been used in the microbiological methods, whereas a defined isolate is a requirement for feed additives.

Several studies were conducted to investigate the ability of microorganisms isolated from rumen fluid and soil to degrade or biotransform deoxynivalenol under anaerobic or aerobic conditions in liquid culture (King et al., 1984a). He et al. (1992) detected transformation of 50% of deoxynivalenol to the de-epoxy metabolite in the first culture of soil microorganisms but no further transformation in later subcultures. Anaerobic degradation of deoxynivalenol has been studied extensively (Binder et al., 1998), and rumenal bacteria capable of transforming deoxynivalenol and 3-acetyldeoxynivalenol to a the de-epoxy metabolite have been isolated. The rumenal contents of a fistulated cow were used as a natural habitat for the enrichment and isolation of anaerobic bacteria (Binder et al., 2000). The active bacterium is a gram-positive, non-spore-forming, strictly anaerobic, irregular rod belonging to the genus Eubacterium. The efficiency of this bacterium as a feed additive, commercially available as such in e.g. South America, has been tested in feeding trials with pigs and chickens. Microbes in rumenal fluid had no effect on the concentration of deoxynivalenol (Kiessling et al., 1984).

Mouldy maize was detoxified microbially by incubation with the contents of the large intestine of chickens, and the response of young pigs fed this detoxified deoxynivalenol-contaminated feed was evaluated. A significant amount of the deoxynivalenol in the maize had been transformed to the de-epoxy metabolite. Microbial inocula from rumenal fluid, soil and the contents of the large intestines of chickens and of pigs were tested for their ability to transform deoxynivalenol in vitro. Microorganisms in the chicken intestinal contents completely transformed pure deoxynivalenol, and this activity was retained through six serial subcultures. No alteration of the toxin was detected after incubation with pig intestinal contents, whereas 35% of the deoxynivalenol was metabolized in the original culture of rumenal fluid and 50% was metabolized by the soil sample. About 50% of the deoxynivalenol in mouldy maize in culture medium was transformed by microorganisms from chicken intestine (He et al., 1993).

A mixed microbial culture capable of metabolizing deoxynivalenol was obtained from soil samples by an enrichment culture procedure. A bacterium isolated from the enrichment culture completely removed exogenously applied deoxynivalenol from the culture medium after incubation for 1 day (Shima et al., 1997). The main metabolite was identified as 3-keto-4-deoxynivalenol. This compound had remarkably less (one-tenth) immunosuppressive toxicity than deoxynivalenol, indicating that the 3-OH group on deoxynivalenol is likely to be involved in its immunosuppressive effects.

The concentration of deoxynivalenol remained stable when deoxynivalenol-containing wort was fermented with strains of Saccharomyces cerevisae for 7–9 days (Scott et al., 1992).

In a study of the transmission of deoxynivalenol from naturally contaminated barley and wheat malts into beer, the deoxynivalenol content increased markedly during mash production. The increase was suggested to be due to enzymatic activity in the malt. Deoxynivalenol was stable after the wort was boiled for 90 min. Fermentation did not affect the initial toxin content of the wort. The concentration in finished beer was similar to that in the respective malt (Niessen & Donhauser, 1993).

Argentina, Brazil, Canada, China, Finland, Germany, Italy, Norway, Sweden, the United Kingdom, Uruguay, and the USA submitted data on contamination of grains and food products with deoxynivalenol to FAO/WHO. Other data on contamination of food by these toxins were taken from the published literature for 1990 through early 2000. Previous data were reviewed in Environmental Health Criteria 105 (WHO, 1990). As the information available was not complete in many cases, most of the authors were contacted; many of them sent details on sampling and analytical methods and some also sent unpublished data.

The data on the natural occurrence of deoxynivalenol are summarized in Appendix A. The criteria for accepting data on food contamination included the date of the report between 1990 and 2000, use of adequate analytical methods for this toxin, and random collection of samples. Details of the sampling procedures are given in Appendix A and in Trichothecenes Appendix 6. The data were analysed on a case-by-case basis. When no information was provided about sampling or analytical methods, the data where not included.

The Appendix includes the primary reference (P), the reference for sampling (S), and the references for the analytical method (A) given by the authors. Mean values were calculated when the mean of positive samples was given or values for ‘not detected’ were given as half the detection limit. The mean values in the Appendix therefore indicate the mean of all samples, with those below the limit of detection taken as 0, with some exceptions, as shown.

In the decade 1980–90, deoxynivalenol occurred commonly, particularly in wheat and maize, at concentrations usually < 1000 µg/kg (WHO, 1990). This value was therefore chosen for comparison with concentrations reported in the decade 1990–2000. A concentration of 100 µg/kg was used to compare the concentration of deoxynivalenol with those of T-2 and HT-2 toxins.

The analytical methods used to derive the data included in Appendix A were GC–ECD, GC–MS, TLC, HPLC, and ELISA, and the details are described in Trichothecenes Appendix 5. Methods for ELISAs were included in Trichothecenes Appendix 5 only when the paper provided information on recovery or when the extraction solvent was not methanol–water. In approximately half of the reports, deoxynivalenol was quantified by GC with ECD or MS, and in one-fourth of the reports by TLC. The recovery of the toxin in various substrates sometimes affected the data on occurrence more than the sensitivity of the method. When the extraction and clean-up steps are effective, low limits of detection or quantification can be obtained even with low-cost methods such as TLC.

Some authors defined LODs and LOQs in different ways. The LOD was sometimes expressed as two or three times the noise over background, and sometimes as the smallest quantity of the standard that could be detected. The LOQ was sometimes defined as five or six times the noise (measured in standard solutions), without taking into account recovery in the matrix, while other investigators included the data on recovery in their estimation of the LOQ. Investigators should report the method by which they calculated the LOD and LOQ, in order to standardize the procedure. Intake can best be estimated by considering studies in which the recovery was > 60% the LOQ. Analytical methods, particularly with regard to recovery, in different matrixes require constant improvement.

No specific sampling method has been reported for these toxins. Those used are described in Trichothecenes Appendix 6.

Some interesting studies were not included in Appendix A, as they could not be used to estimate the total dietary intake of the toxin. These include reports of the occurrence of the toxin in areas where disease was found, estimates of variations in the concentration of deoxynivalenol in different lots, studies on genetic variation, samples of grains with scab symptoms, and comparisons of the presence of deoxy-nivalenol in mouldy and healthy grains or in different parts of the plant. In other cases, the results reported were insufficient to be entered in Appendix A, as for example when deoxynivalenol was expressed as total type B trichothecenes obtained after hydrolysis (Perkowski et al., 1990; Luo et al., 1992; Okoye, 1993; Chu & Gy, 1994; Trigo-Stockli et al., 1995; Menna et al., 1997; Perkowski et al., 1997; Li et al., 1999; Perkowski, 1999; Sohn et al., 1999; Wetter et al., 1999; Chelkowski et al., 2000; Freese et al., 2000).

In one study, beans were found to be contaminated by trichothecenes (deoxy-nivalenol, diacetoxyscirpenol, and T-2 toxin). As the samples were not collected at random, the data were not included in Appendix A. However, the quantities of toxins detected were substantial and warrant investigation in other legumes (Tseng et al., 1995).

Deoxynivalenol was also detected in soya beans, with discolouration damage, and in some soya bean products from a pilot plant. The concentrations ranged from not detectable to 490 µg/kg in whole soya beans, from < 10 µg/kg to 420 µg/kg in hulls, from < 5 µg/kg to 600 µg/kg in meal, and from not detectable to 30 µg/kg in crude unrefined oil (Jacobsen et al., 1995). In Appendix A, only five samples of soya beans were found to have no detectable deoxynivalenol.

In the United Kingdom, samples of traditional foods were reported to be contaminated with deoxynivalenol at concentrations similar to those found in wheat (Patel et al., 1996).

As shown in Appendix A, deoxynivalenol was a frequent contaminant of cereal grains such as wheat (11 444 samples, 57% contaminated), maize (5349 samples, 40% contaminated), oats (834 samples, 68% contaminated), barley (1662 samples, 59% contaminated), rye (295 samples, 49% contaminated), and rice (154 samples, 27% contaminated). It has also been detected in some wheat and maize products, such as wheat flour, bread, breakfast cereals, noodles, baby and infant foods, and cooked pancakes. It has been reported in barley products and beer: more than 50% of 321 beer samples analysed were contaminated with deoxynivalenol (Niessen et al., 1993; Scott et al., 1993; Moltó et al., 2000).

In some countries, such as Germany and Italy, the presence of deoxynivalenol was reported in grains grown organically and by conventional techniques.

Although a large amount of data was summarized (23 380 samples), information on some products in various parts of the world is still lacking.

Some authors presented data on distribution in the form of frequency tables or histograms (for example Müller et al., 1997a,b). Fontán et al. (1994) showed that the Fisher–Tippett distribution could be adjusted adequately for the data on the 1990–92 wheat harvests in Argentina. Some individual data were almost discrete, owing to rounding, and could not be used for the distribution analysis. For this purpose, only data on deoxynivalenol contamination of wheat were available.

Figure 5 presents Q–Q plots of the log-transformed data for the 1994 and 1997 harvests in Argentina, the 1999 harvest in Italy, and the 1998 harvest in the United Kingdom. Similar patterns were found for the harvest in Finland in 1998 (data not shown). For the analysis, only values greater than 20 µg/kg (LOD) for the United Kingdom, 50 µg/kg (LOD) for Italy, and 48 µg/kg (LOQ) for Argentina were considered. In all cases, goodness-of-fit tests were applied. The p values obtained (> 0.5) indicated that the log-normal distribution provided an adequate model for these data.

The distribution function for processed products such as bread could be different from that adjusted for raw cereal, as contamination tends to be more homogeneous during processing. The adjusted distribution function could therefore belong to other distribution families (Schollenberger et al., 2000a).

Most of the information on contamination with deoxynivalenol has been reported for wheat (11 000 samples) and maize (6000 samples). These large data sets allow determinations of annual variation in concentrations. For instance, wheat harvested in Germany, Norway and the Russian Federation in various years showed similar annual variations in contamination (Appendix A). Data from Canada obtained in consecutive years showed that the average contamination in hard wheat was lower than that in soft wheat but that the annual variation in the two products was similar. Figure 6 shows the annual variation in the average contamination of wheat and the frequency of contamination in Argentina from a survey conducted in the main production area between 1985 and 1999.

The annual variation in the average concentration of deoxynivalenol in maize in Canada and New Zealand in 1987–94 is shown in Figure 7.

Although there were annual variations, the mean concentrations of deoxynivalenol in maize in Argentina, Canada, and Uruguay over the past 5 years were < 120 µg/kg (Appendix A). Oats and barley also showed annual variations in contamination with deoxynivalenol (Appendix A). These are clearly observed when the frequency of contamination is high, which has been the case for barley for some years. These results emphasize the need for regular screening for deoxynivalenol in cereal crops. The annual variation observed could be due mainly to heavy rainfall between flowering and harvest, which facilitates F. graminearum infection and accumulation of deoxynivalenol (Trigo-Stockli et al., 1998).

The dietary intake of deoxynivalenol was assessed according to the recommen-dations of a FAO/WHO workshop on methods for assessing exposure to contaminants and toxins, which was held in Geneva in June 2000 (WHO, 2000). The workshop recommended that the median concentration be given when data on individual samples were available, whereas a mean should be given when only pooled or aggregated data were available. In the case of commodities that contribute significantly to intake, distribution curves should be generated to allow risk managers to determine the effects on dietary intake of different maximum levels.

The workshop further recommended that international estimates of dietary intake should be calculated by multiplying the mean or median concentration by the values for consumption of the commodity in the five GEMS/Food regional diets (WHO, 1998). The diets (African, European (which includes Australia, Canada, New Zealand, and the USA), Far Eastern, Latin American, and Middle Eastern) were established on the basis of information on food balance sheets compiled by FAO. As such information is available for most countries, the data are comparable across countries and regions of the world. The regional diets represent the average availability of food commodities per capita rather than actual food consumption.

The report of the workshop noted that national intake estimates should also be reported when available, as they may provide information about intake by specific population subgroups or extreme consumers, which cannot be derived from GEMS/Food regional diets.

For this assessment, concentrations of deoxynivalenol in food commodities and in some processed foods were reported to FAO/WHO or were obtained from the literature. The quality and reporting of the data are discussed in the previous section. Since the dietary intakes were based on the GEMS/Food regional diets, which include information on consumption of raw or minimally processed foods, concentrations of deoxynivalenol in processed foods were not used to estimate dietary intake.

Information was available on the concentrations of 10 commodities: barley, maize, popcorn, oats, rice, rye, sorghum, triticale, wheat, and other cereals. Data were received from 28 countries, representing four of the five GEMS/Food regional diets (Table 7); no data were reported for the Middle Eastern diet. Of the 10 commodities for which data were available for the intake assessment, data on barley, maize, and wheat predominated, with limited reports on popcorn, oats, rice, rye, sorghum, triticale, and other cereals (Table 8).

Most of the data available for this evaluation were pooled; that is, each data point represented the mean concentration in a number of individual samples. In calculating the mean values, samples in which the concentration was below the LOQ or LOD were assumed to have a value of zero. The maximum analytical value and the number of samples with concentrations below the LOD or LOQ were also reported for each data point. A total of 375 data points (mean values) representing about 23 000 individual samples were included in the intake assessment (Table 8). Of those 375 data points, 243 were reported from countries represented by the GEMS/Food European diet. The remaining 132 data points represented intake of the nine commodities for the other regional diets.

For each commodity, the data were sorted according to the country groupings of the GEMS/Food regional diets. The number of data points reported, the number of individual samples represented, the highest maximum analytical value reported, the proportion of samples with concentrations below the LOD or LOQ, and the average of all mean values are summarized in Table 9. For each commodity, the mean of all data, weighted by sample size, is also reported.

The concentrations of deoxynivalenol used in estimating dietary intakes are summarized by commodity and region in Table 9.

Barley: Data on the concentrations of deoxynivalenol in barley were received from 11 countries. Of the 1778 samples analysed, 41% had concentrations below the LOD or LOQ. The unweighted means by region ranged from 130 µg/kg in the Far Eastern diet to 860 µg/kg in the European diet. The weighted mean for all samples combined was 720 µg/kg, and the maximum analytical value reported was 34 000 µg/kg.

Maize: Seventeen countries reported data on a total of 5719 samples of maize. Of these, 52% contained concentrations below the LOD or LOQ. The unweighted means across regional diets ranged from 66 µg/kg in the Latin American to 640 µg/kg in the European diet. The weighted mean of all samples combined was 180 µg/kg, and the maximum analytical value reported was 19 000 µg/kg.

Popcorn: Only two countries (Argentina and USA) reported data on popcorn. Of the 50 samples, 92% had concentrations below the LOQ. The weighted mean for all samples was 310 µg/kg, and the maximum analytical value reported was 4500 µg/kg.

Oats: Seven countries representing only two regional diets submitted data on a total of 834 samples of oats. Most of the data came from countries with the European diet; only one record representing six samples (all with concentrations below the LOQ) was reported for the Latin American diet. Thirty-two percent of all samples had concentrations below the LOQ. The maximum analytical value reported was 2600 µg/kg.

Rice: Only four countries representing two regional diets submitted data on rice, representing a total of 203 samples. Overall, 80% of the samples had concentrations below the LOD or LOQ. The maximum analytical value reported was 9500 µg/kg. The weighted mean concentration in all samples was 150 µg/kg.

Rye: Four countries, all with the European regional diet, submitted data on a total of 295 samples of rye. Just over half (51%) of the samples had concentrations below the LOQ. The maximum analytical value reported was 1300 µg/kg, and the unweighted mean was 39 µg/kg; the weighted mean for all samples was 65 µg/kg.

Sorghum: Only one country reported data on 15 samples of sorghum. Deoxy-nivalenol was not detected in any of the samples.

Triticale: Only one data point representing 10 samples was reported for triticale. The mean value was 92 µg/kg; the maximum analytical value was 200 µg/kg.

Wheat: Eighteen countries with three regional diets reported data on 14 200 samples of wheat. Overall, 38% of the samples had concentrations below the LOD or LOQ. A maximum analytical value of 30 000 µg/kg was reported for two sample sets. The unweighted means ranged from 310 µg/kg for the European diet to 560 µg/kg for the Far Eastern. The weighted mean for all samples was 390 µg/kg.

Other cereals: Only two countries submitted data on a total of 254 samples of cereal grains other than those specified above. Overall, 55% of the samples had concentrations below the LOD or LOQ. The maximum analytical value reported was 2300 µg/kg. Both the unweighted and the weighted means for all samples were 46 µg/kg.

The average intakes of deoxynivalenol were calculated by multiplying the weighted mean concentration of each commodity by the corresponding amount of each commodity consumed in each of the five GEMS/Food regional diets (Table 10). As limited data were available on the concentrations of all commodities in diets other than the European one, a mean concentration for each commodity could not be derived for each region.

Intakes were estimated per person per day and converted to intake per kilogram of body weight per day, assuming a body weight of 60 kg, as recommended by FAO/WHO for international intake assessments (WHO, 1985). The results are reported separately for each GEMS/Food regional diet in Tables 11–15. The intakes from the different diets are compared in Table 16.