In one unconfirmed report, fumonisin B₁ was active in a commercial biolumines-cent bacterial assay. It increased the micronucleus frequency in primary rat hepatocytes in a non-concentration-dependent manner and caused a significant, concentration-dependent increase in the frequency of chromosomal aberrations. The Commission of the European Union (2000) considered that this report was unreliable owing to methodological limitations. Fumonisin B₁ transformed mouse embryo cells only at a concentration of 500 µg/ml and not at lower or higher concentrations.

Evidence in vitro and in vivo indicates that fumonisin B₁ can damage DNA indirectly by increasing oxidative stress (Atroshi et al., 1999; Mobio et al., 2000b). Oxidative damage was closely associated with fumonisin B₁-induced hepatotoxicity and induction of putative preneoplastic lesions in vivo, while the plasma and microsomal membranes, and to some extent the mitochondria and nuclei, appeared to be significantly affected by lipid peroxidation (Abel & Gelderblom, 1998).

Studies in primary hepatocytes showed a similar relationship between cytotoxicity and lipid peroxidation. Although lipid peroxidation was prevented by the addition of alpha-tocopherol, cytotoxicity was reduced, but not to baseline levels, suggesting that the cytotoxicity was due not only to oxidative damage but also to fumonisin B₁-induced toxic effects. Fumonisin B₁ potentiated the effect of iron on lipid peroxidation and was toxic to the liver, independently of its effect on lipid peroxidation (Lemmer et al., 1999a). The toxin induced lipid peroxidation in cell membrane preparations (Yin et al., 1998) and isolated rat liver nuclei. Nuclear membrane lipid peroxidation with concomitant DNA strand breaks was found in isolated rat liver nuclei treated with fumonisin B₁ in vitro at concentrations of 40-300 µmol/L (Sahu et al., 1998). It was suggested that the formation of hydroxy and peroxyl radicals in the close vicinity of nuclear material could induce DNA strand breaks. It is not known whether the induction of chromosomal aberrations in primary hepatocytes (Knasmüller et al., 1997) is related to oxidative damage. Oxidative damage has been reported in Vero cells (Abado-Becongnee et al., 1998) treated with fumonisin B₁ at a concentration (0.14 µmol/L) below that which inhibits protein and DNA synthesis (> 14 µmol/L). However, in mice, fumonisin B₁-induced liver toxicity occurred in the absence of evidence of increased lipid peroxidation (Riley et al., 2001). In numerous studies in vitro, fumonisin was used to protect cells from oxidant-induced cell death (for examples, see Table 3). Whether this occurs in vivo is unknown.

Although fumonisin B₁ caused oxidative damage to DNA in vitro, there is no compelling evidence to suggest that it binds covalently to DNA. Extracts of Fusarium fungi have been shown to form DNA adducts (Lu et al., 1988; Bever et al., 2001), as detected by the sensitive ³²P-postlabelling method. However, DNA adducts did not form after incubation of fumonisin B₁ with DNA in the presence or absence of microsomal protein (rat liver microsomes), and are probably formed by other mycotoxins produced by the fungus (P. Howard, personal communication). Using mild electrospray ionization mass spectrometry, Pocsfalvi et al. (2000) found no evidence of a specific interaction between fumonisin B₁ and single- or double-stranded oligonucleotides. Specific non-covalent interactions were noted with fusaproliferin and beauvercin, two mycotoxins produced by F. proliferatum and F. subglutinans.

While fumonisins are embryotoxic in vitro, there were no published data to support the conclusion that they are developmental or reproductive toxicants in farm animals or humans (WHO, 2000a). Fumonisin B₁ inhibits the biosynthesis of the glycosylphosphatidylinositol-anchored folate transporter in vitro, and folate deficiency is associated with an increased risk for neural tube defects. Inhibition of the folate transporter in vivo has not been confirmed in feeding studies. There is no evidence of neonatal toxicity in laboratory animals. Except in one study in Syrian hamsters, embryotoxicity occurred secondary to maternal toxicity. In Syrian hamsters, administration of fumonisin B₁ by gavage at a dose of 18 mg/kg bw per day on days 8 to 10 or 12 of gestation resulted in a statistically significant increase in the number of fetal deaths. The fact that there was no maternal toxicity at this dose indicates that the Syrian hamster is extremely resistant to the effects of fumonisin. In rats and mice, reproductive and developmental effects occurred at doses that were also maternally toxic, suggesting that the reproductive and developmental toxicity of fumonisin is mediated through maternal toxicity. However, in rabbits, maternal toxicity was observed at a daily dose (given by gavage in water) as low as 0.25 mg/kg bw on days 3–19 of gestation, but there was no increase in the frequency of fetal loss or of gross visceral or skeletal abnormalities at any maternal dose of fumonisin B₁ (0–1.75 mg/kg bw per day); however, slight decreases in fetal weight did occur at maternally toxic doses.

Because fumonisins are known to cause field outbreaks of equine leukoen-cephalomalacia and porcine pulmonary oedema, many studies have been undertaken to better understand the physiological and biochemical mechanisms responsible for these brain and lung diseases. The cardiovascular changes in equids are similar to those seen in pigs during onset of the syndrome (Smith et al., 1999; Constable et al., 2000a; Smith et al., 2000), and the results of studies suggest a common underlying mechanism for the two diseases involving sphingosine-mediated L-type calcium channel blockade (Constable et al., 2000b).

A thorough review of studies on these two diseases is contained in the IPCS monograph (WHO, 2000a). The following is a brief summary of that review.

2.3.1 Equine leukoencephalomalacia

Equine leukoencephalomalacia syndrome is a sporadic condition characterized by the presence of liquefactive necrotic lesions in the cerebrum. The disease appears to be unique to equids, although brain lesions have also been reported in rabbits (Bucci et al., 1996), pigs (Fazekas et al., 1998), and carp fed fumonisins (Pepeljnjak et al., 2000). The brain lesions in rabbits and pigs could not be reproduced in subsequent studies (T. Bucci, personal communication; Zomborszky et al., 2000), and the study in carp has not been repeated.

Analysis of feeds from confirmed cases of equine leukoencephalomalacia in the USA indicated that consumption of feed with a fumonisin B₁ concentration > 10 mg/kg of diet (equivalent to 0.2 mg/kg bw per day) was associated with an increased risk for developing the disease, whereas a concentration < 6 mg/kg of diet (equivalent to 0.12 mg/kg bw per day) was not. The minimum oral dose sufficient to induce equine leukoencephalomalacia appeared to be 15–22 mg/kg of diet (equivalent to 0.30 mg/kg bw per day for 150 days to 0.44 mg/kg bw per day for 241 days) in studies of naturally contaminated maize screenings or culture material (F. proliferatum) containing fumonisins. The minimum oral dose of pure fumonisin B₁ that will induce equine leukoencephalomalacia is unknown. The minimum intravenous dose of pure fumonisin B₁ that induced neurological abnormalities was 0.01–0.05 mg/kg bw per day. If the intravenous dose is assumed to represents 5% of the oral dose, the equivalent oral dose would be 0.2–1.0 mg/kg bw per day. The NOEL for cardiovascular abnormalities was 0.2 mg/kg bw per day, but the NOEL for serum biochemical abnormalities was equivalent to < 0.2 mg/kg bw per day (Constable et al., 2000b).

Equine leukoencephalomalacia has been reproduced by giving fumonisin B₁ orally or intravenously, giving naturally contaminated maize screenings, or giving diets containing F. proliferatum maize culture material with predominantly fumonisin B₁ or fumonisin B₂. In addition to the brain lesions, histopathological abnormalities are usually found in the liver, and recent studies suggest that fumonisin B₁ is also nephrotoxic in horses. Changes in serum enzymes indicative of liver damage and behavioural changes are usually preceded by increased concentrations of free sphingoid bases and cholesterol in serum or plasma. It has been hypothesized that equine leukoencephalomalacia is a result of cerebral oedema due to an inability to shut down the blood flow to the brain when the horse lowers its head to eat and drink (Constable et al., 2000b).

Evidence of disruption of sphingolipid metabolism is an early indicator of exposure of horses to fumonisins. For example, all ponies fed diets containing maize screenings naturally contaminated with fumonisins at > 22 mg/kg (primarily fumonisin B₁) had large increases in the serum concentration of free sphinganine. The increase was reversible. This and the increased sphinganine:sphingosine ratio occur before increases in serum transaminase activity and clinical signs of equine leukoencephalomalacia.

2.3.2 Porcine pulmonary oedema

Porcine pulmonary oedema is also believed to be induced by cardiovascular dysfunction. Significant changes in oxygen consumption and in several haemodynamic parameters are seen before the onset of clinical signs in pigs fed diets containing fumonisins, suggesting that pulmonary hypertension caused by hypoxic vasoconstriction might contribute to the syndrome. Fumonisin B₁ has been shown to reduce the mechanical efficiency of the left ventricle (Constable et al., 2000a), suggesting that pulmonary oedema in pigs is due primarily to acute left-sided heart failure (Smith et al., 1999; Constable et al., 2000a; Smith et al., 2000). Hepatotoxicity is usually observed at doses lower than those that induce the pulmonary oedema. Liver lesions were induced by maize screenings providing a dose of fumonisin of 1.1 mg/kg bw per day (17 mg/kg of diet). In a study in weaned piglets, mild pulmonary oedema was induced in three of four animals fed diets that contained 10 mg/kg (equivalent to 0.4 mg/kg bw per day) of fumonisin B₁ from F. verticillioides culture material for 4 weeks (Zomborszky et al., 2000). Subtle changes in performance were reported at a concentration of fumonisin B₁ (1 mg/kg), which does not cause overt toxicity. Erratic growth was reported at 0.1 mg/kg (Rotter et al., 1996).

In 1989–90, outbreaks of this disease were reported in various parts of the USA. Maize screenings obtained from farms where pigs had died of porcine pulmonary oedema were contaminated predominantly with F. verticillioides. The concentrations of fumonisin B₁ in the suspected feeds ranged from 20 to 360 mg/kg, while those in other feeds was < 8 mg/kg (Ross et al., 1991). The clinical signs of porcine pulmonary oedema include dyspnoea, weakness, and cyanosis. At necropsy, the animals have varying degrees of interstitial and interlobular oedema, with pulmonary oedema and hydrothorax. Toxic hepatosis occurs concurrently. The concentrations of fumonisin B₁ in feed associated with this disease are usually much greater than those associated with outbreaks of equine leukoencephalomalacia. Purified fumonisin B₁ has been shown to reproduce the disease when administered intravenously. Porcine pulmonary oedema has not yet been reproduced by oral administration of pure fumonisins, although it has been induced many times with culture material containing fumonisin B₁.

In pigs fed diets prepared from naturally contaminated maize screenings, there was a dose–response relationship between the dose of fumonisin, pathological lesions in the liver, and the ratio of free sphinganine to free sphingosine in serum or liver. After 14 days, statistically significant increases in the serum ratio of free sphinganine to free sphingosine were observed at a concentration of total fumonisins in feed as low as 5 mg/kg (equivalent to 0.2 mg/kg bw per day). The concentrations of free sphingoid bases were also significantly elevated in kidney and lung at doses that induced no signs of toxicity in these organs. In pigs fed fumonisin in maize culture materials, significant effects on cardiovascular function were associated with significant increases in free sphingoid base concentrations in heart tissue. Subsequent studies showed that damage to pig alveolar endothelial cells in vivo was preceded by accumulation of free sphingoid bases in lung tissue. Increased activity of serum enzymes and increased concentrations of free sphingoid bases in serum were seen at all doses in the study of Zomborszky et al. (2000). It has been hypothesized that the cardiovascular alterations leading to acute left-sided heart failure are a consequence of sphingoid-base-induced inhibition of L-type calcium channels (Smith et al., 1999; Constable et al., 2000a; Smith et al., 2000). The minimum oral dose needed to induce porcine pulmonary oedema has not been clearly established; however, Smith et al. (1999) predicted that when the concentra-tions of free sphinganine and free sphingosine in plasma are > 2.2 and 1 µmol/L, respectively, haemodynamic changes will occur.

The association between fumonisins and other fungal toxins and risk factors for various human diseases are summarized in Table 8.

2.4.1 Oesophageal cancer

An association has been established between the occurrence of the fungus, Fusarium verticillioides (Sacc.) Nirenberg (= Fusarium moniliforme Sheldon) on maize and the incidence of oesophageal cancer in various regions of the world. Geographi-cal differences in demography, ethnic groups, genetic susceptibility, culture, economy and nutritional status all affect the rates of disease; however, some common risk factors are emerging, such as having maize as the main dietary staple and, to some extent, a low socioeconomic status. Thus, high incidences of oesophageal cancer have been associated with limited diets consisting mainly of wheat or maize and low contents of certain minerals and vitamins (Blot, 1994). Fungal contamination of maize and wheat attracted the interest of many investigators, who have characterized the toxic, mutagenic, and carcinogenic metabolites of the major fungal contaminants of these grains. The possible involvement of F. verticillioides and fumonisins in the development of oesophageal cancer in various regions of the world is evaluated on the basis of incidence rates, the presence of Fusarium spp. in maize, the presence of fumonisins and other Fusarium mycotoxins, and nutritional deficiencies and other dietary risk factors. Those aspects that were addressed by WHO (2000) are not discussed in detail.

A dramatic increase in the incidence of oesophageal cancer in the Transkei region of Eastern Cape Province, South Africa, was first described by Burrell (1957, 1962; Burrell et al., 1966). Subsequent reports were published for the periods 1955–57 and 1965–69 covering all districts of the Transkei (Rose, 1965, 1973; Rose & Fellingham, 1981) and for the periods 1981–84 (Jaskiewicz et al., 1987b) and 1985–90 (Makaula et al., 1996) focusing on cancer incidence in four selected districts: two areas of high incidence (Centane or Kentani and Butterworth in the south-west) and two areas of low incidence (Lisikisiki and Bizana in the north-east). Data for 1991–95 indicate that the age-standardized incidence rate has decreased in Butterworth to 51 per 100 000 but increased in Lusikisiki and Bizana to 37 per 100 000. Conversely, the rate in Centane has remained consistently high in males, at 56 per 100 000.

A comparative study of the incidence of Fusarium spp. in maize in regions of low (Bizana and Lusikisiki) and high (Butterworth and Centane) incidence of oesophageal cancer in the Transkei indicated that three Fusarium species (F. graminearum, F. verticillioides, and F. subglutinans) were the predominant fungal contaminants of home-grown maize during the 1976–77 season (Marasas et al., 1979). The association between the occurrence of F. verticillioides in maize and the incidence of oesophageal cancer was established in a detailed mycological study of home-grown maize in the four districts, carried out for six seasons in 1976–79, 1985–86, and 1989 (Marasas et al., 1981, 1988; Sydenham et al., 1990; Rheeder et al., 1992). The incidences of F. graminearum and F. subglutinans and the concentrations of toxins produced by these Fusarium species, i.e. deoxynivalenol, nivalenol, and zearalenone and moniliformin, respectively, were significantly higher in home-grown maize from the area of low incidence of oesophageal cancer than that of high incidence (Sydenham et al., 1990; Rheeder et al., 1992). Significantly higher concentrations of fumonisin B₁ and fumonisin B₂ were detected in visibly uncontaminated (‘healthy’) and mouldy samples of maize from the high-incidence areas than the low-incidence areas. The environmental conditions in the high-incidence area favour the colonization of maize ears and toxin production by F. verticillioides, whereas the conditions in the low-incidence area favour F. graminearum and F. subglutinans. This interaction was clear from a mycological analysis of maize harvested during 1985 (Sydenham et al., 1991), while variable results were obtained in maize samples collected during the 1976 and 1977 seasons (Marasas et al., 1979). The population in these regions is exposed not only to the known Fusarium toxins but to various other mycotoxins that may play a role in the development of cancer in humans, and specifically cancers of the oesophagus and the liver. These toxins include aflatoxin B₁, present in food and traditional beer (Van Rensburg et al., 1990), the mutagen fusarin C (Gelderblom et al., 1984), and other compounds produced by F. verticilioides (Bever et al., 2001).

People in the high-incidence area for oesophageal cancer in the Transkei were exposed not only to mycotoxin contamination of maize but also had nutritional deficiencies, such as of vitamins A, E, and B₁₂, selenium, and folate, when compared with people in the low-incidence areas (Van Helden et al., 1987; Jaskiewicz et al., 1987b, 1988a,b). In contrast to the study of Van Rensburg et al. (1983), a study by Jaskiewicz et al. (1988b) showed no difference in the plasma concentrations of zinc, copper, and magnesium in people in the low- and high-incidence areas. Deficiencies in these micronutrients and in manganese and molybdenum have been implicated as risk factors for oesophageal cancer in populations that subsist on either maize or wheat (Van Rensburg, 1985). A case–control study in Zulu men with oesophageal cancer carried out in Durban, South Africa, indicated that consumption of purchased maize meal is one of the major risk factors for development of the disease (Van Rensburg et al., 1985). Mineral deficiencies in the soil were implicated as another possible risk factor in the high-risk areas in the Transkei (Burrell et al., 1966); however, a study of the elemental content of soil and maize leaves (Rheeder et al., 1994) showed no deficiency in mineral elements and no association with the risk for oesophageal cancer.

Alcohol consumption and tobacco smoking are widely regarded as major factors in the development of oesophageal cancer in developed countries (Blot, 1994). In the endemic areas in the Transkei, a number of the patients did not use tobacco or consume alcohol, hence ruling out these factors as the sole causative agents (Rose, 1973; Sammon, 1992). In Zulu men, however, smoking was identified as a risk factor, while alcohol consumption had no appreciable effect (Van Rensburg et al., 1985).

A survey in Henan, Hebei, and Shanxi, three provinces in northern China, showed that the highest mortality rates from oesophageal cancer were those in two counties, Yancheng and Hebei, with crude rates of 135 and 140 per 100 000, respectively (Yang, 1980). Data for 12 years (1959–70) in a cancer registry in Linxian County showed an average incidence rate of 108 per 100 000, while the lowest rates were recorded in Hunyan County (1.4 per 100 000) and Tatong County (2.8 per 100 000). The male:female ratio ranged from 1.44:1 to 2.63:1.

The fungal contamination of 1121 food samples of wheat, maize, dried sweet potato, rice, and soya bean obtained from five counties of high incidence and three of low incidence for oesophageal cancer in Henan Province indicated that, in addition to Penicillium and Aspergillus species, the incidence of F. verticilioides, was significantly higher in the high-incidence area (Zhen, 1984). An increased risk was associated with a high intake of wheat and maize, further emphasizing the importance of performing detailed studies on the fungal contamination of these food commodities (Li et al., 1989).

Analysis of 31 maize samples collected from households in Linxian and Cixian counties showed high concentrations of fumonisin B₁ in both mouldy and healthy samples; low concentrations of aflatoxin B₁ and various type A and B trichothecenes were also found in the mouldy samples (Chu & Li, 1994). In studies on maize (Yoshizawa et al., 1994) and on wheat and maize (Gao & Yoshizawa, 1997), the frequency of fumonisin contamination was higher in Linxian County, while the mean concentrations of fumonisin B mycotoxins were similar to those in Shangqiu County, a low-incidence area of oesophageal cancer in Henan Province. The samples were also frequently co-contaminated with trichothecenes, consisting mainly of deoxy-nivalenol and to some extent nivalenol. The concentrations of these mycotoxins were significantly higher in Linxian, and zearalenone was detected only in the maize samples from this County.

Aflatoxin B₁ and fumonisin B contamination was measured in 246 samples of healthy maize kernels collected from villages in the high-risk counties of Cixian, Linxian, and Anyang and the low-risk counties of Fanxian and Yanqui, during 1995–96. Once again, the frequency of fumonisin contamination was higher in the high-incidence areas (65%) than in the low-incidence areas (28%). Although no clear relationship could be detected between fumonisin contamination and oesophageal cancer incidence, people in the high-incidence area were exposed to higher mean concentrations of fumonisin B in maize than those in the low-incidence areas (Zhang et al., 1997). Significantly greater contamination with deoxynivalenol, nivalenol, and zearalenone was found in the main staple foods (maize and wheat) in the high-incidence county of Linxian than in the low-incidence area, Shangui (Luo et al., 1990).

A nutritionally inadequate diet appears to play an important role in the incidence of oesophageal cancer and the development of the precancerous dysplastic state. The diet in Linxian consisted mainly of maize, wheat, millet, rice, and some seasonal vegetables. Up to 80% of the calorie intake was obtained from grains, while consumption of fruit and vegetables was low, resulting in deficiencies in vitamins A and C during certain times of the year. Detailed analyses of food and drinking-water indicated an inverse correlation between the rate of mortality from oesophageal cancer and the mineral content of the diet, including molybdenum, manganese, and zinc. Analyses of hair, serum, and urine indicated that the molybdenum content was significantly lower in men from Linxian than from the low-incidence areas of Yuxian and Xinyangxian. Similar studies carried out in oesophageal cancer patients in Henan indicated lower zinc concentrations than in normal subjects. Analysis of resected oesophagi from patients in Linxian showed significantly lower molybdenum concentrations in cancerous tissue. The consumption of pickled vegetables heavily infested with various Aspergillus and Fusarium spp. and mouldy food is common in the high-incidence areas, whereas the intake in the areas of lower incidence is less common (Yang, 1980).

Li et al. (1989) found no association with the intake of pickled vegetables, however, while the intake of wheat and maize was associated with an increased risk and that of fresh vegetables and fruit with a decreased risk for oesophageal cancer. No association was found with alcohol intake, while smoking was a mild risk factor. The presence of nitrosamine precursors in drinking-water and food and the formation of nitrosamines in the stomach have been associated with the development of oesophageal cancer (Yang, 1980). Various nitrosamines (N-nitrosodimethylamine, N-nitrosodiethylamine, and N-nitrosomethylbenzylamine) were formed in maize bread inoculated with F. verticillioides in the presence of low concentrations of sodium nitrate, and secondary nitrosamines were detected with some Aspergillus and Penicillium species (Li et al., 1979). These findings have not, however, been confirmed. Maize meal infested with F verticillioides induced tumours in the oesophagus and stomach of rats and mice (Li et al., 1980, 1982).

The incidence of oesophageal cancer varied by up to 30-fold in men and sixfold in women along a 300-mile stretch of the Caspian littoral of Iran. Gonbad, an eastern region of Mazandaran Province, was regarded as one of the world’s ‘hotspots’ for oesophageal cancer in 1960–61, with truncated age-standardized rates of 206 per 100 000 for men and 262 per 100 000 for women (Kmet & Mahboubi, 1972; Hormozdiari et al., 1975). The incidence was higher among women than men in the high-incidence area. Lower incidences were recorded in the western part of the Caspian littoral.

A survey of the mycoflora of maize kernels collected from seed production centres in Sari, Moghan, and Karaj indicated that isolates of Aspergillus, Fusarium, and Penicillium were dominant (Bujari & Ershad, 1993). These include the fumonisin-producing species F. verticillioides and F. proliferatum. Shephard et al. (2000) reported that maize in Mazandaran Province contained higher concentrations of fumonisins than maize in Isfahan Province to the south of Mazandaran, a low-incidence area for oesophageal cancer. This report of the occurrence of fumonisins in maize in a high-risk area for oesophageal cancer in Iran must be followed up by studies of the maize consumption pattern in this region, the fungal contamination, and the presence of other Fusarium mycotoxins in maize. Furthermore, the samples were intended for animal consumption.

Heavy fungal contamination was also found in wheat stored in underground pits in Iran, the most prevalent fungus being Alternaria alternata (Kmet & Mahboubi, 1972). The toxins associated with this fungus have a similar structure to fumonisins and have similar biological effects in plants and cultured mammalian cells (van der Westhuizen et al., 1998). It is not known whether they also mimic the characteristic biological effects of fumonisins in animals. It has been reported that A. alternata can produce fumonisins (Chen et al., 1992).

In a detailed study of the distribution of exogenous factors related to the differences in the incidence of oesophageal cancer in the Caspian littoral, no single causative agent could be identified. Bread and sheep and goat milk or yoghurt were the main dietary staples in high-incidence areas, while rice was the mainly staple in low-incidence regions (Hormozdiari et al., 1975). A low intake of vitamins A and C, riboflavin, animal protein, fresh vegetables and fruit and a high intake of bread and tea were common in the high-incidence areas. The micronutrient deficiencies are typical of those among persons of low socioeconomic status (Joint Iran–ARC Study Group, 1977; Cook-Mozaffari et al., 1979). Food samples from both the low- and the high-incidence areas contained low concentrations of aflatoxins, polycyclic aromatic hydrocarbons, and nitrosamines (Hormozdiari et al., 1975). Alcohol and tobacco smoking were reported to have no role, as intake was negligible and the women abstained from both alcohol and nass, a mixture of opium, lime, and ash favoured by some male inhabitants. Studies performed by the Joint Iran–IARC Study Group (1977) and Ghadirian et al. (1985) provided some support for the hypothesis that pyrolysate products of opium are involved in the etiology of oesophageal cancer. Subsequent studies by Kmet & Mahboubi (1972), Ghadirian (1987), and O’Neil et al. (1980) implicated deficiencies of micronutrients such as iron, manganese, copper, and zinc, thermal irritation from hot tea, contamination of bread with silica fibre, and consumption by women of a diet consisting of sour promegranate seeds, black pepper, and garlic.

The standardized mortality rate for oesophageal cancer among men in Pordenone Province in the Friuli–Venezia Giullia region of northeastern Italy was 17 per 100 000 (Franceschi et al., 1990). Fumonsin-producing Fusarium species were shown to be present in maize produced in northern Italy (Logrieco et al., 1995). One study showed the presence of fumonisin B in 20 samples of polenta at concentrations of 0.15–3.8 mg/kg (Pascale et al., 1995).

Two studies showed a correlation between consumption of maize, particularly polenta, and the incidence of cancers of the upper digestive track and oesophagus in the Fruili–Venezia Giulia region (Rossi et al., 1982; Franceschi et al., 1990). In both studies, deficiencies of several micronutrients in refined maize, including riboflavin and niacin, were implicated, especially in conjunction with alcohol consumption, which may aggravate the nutritional deficiency induced by maize-based diets.

A retrospective study in Kenya of 667 cases of oesophageal cancer from the major hospitals and the Kenya Cancer Registry during the period 1968–75 showed a male:female ratio of 8:1. About 45% of the cases were from western Kenya and 45% from the centre, which are considered to be high-incidence areas as compared with the coastal and northern regions and the Rift Valley (Gatei et al., 1978).

A survey of fungal contamination of maize showed that F. verticillioides was the most frequent contaminant in maize kernels from western and central Kenya (Macdonald & Chapman, 1996), and the finding was confirmed on screening the fungal contamination of 150 maize kernel samples collected in the tropical highlands of western Kenya. Chemical analyses of 197 samples collected in this region showed little fumonisin contamination: 46% of the samples contained concentrations above the limit of detection (100 µg/kg), 5% contained > 1 mg/kg, and a few samples (of poor quality) contained fumonisin B₁ at 3.6–12 mg/kg (Kedera et al., 1999). A low concentration (0.06 mg/kg) of total fumonisins was also detected in only one of seven samples collected in western Kenya (Van der Westhuizen et al., 1999).

Consumption of alcohol does not explain the geographical and ethnic variations, as dietary patterns and tribal customs in Kenya vary considerably (Gatei et al., 1978). No significant correlation was observed between the presence of nitrosamine-like compounds and the incidence of oesophageal cancer. The spatial distribution corresponds to the annual rainfall patterns, with high-incidence areas in regions of heavy rainfall.

The population-based cancer registry in Zimbabwe showed cancer incidence rates similar to those in other countries of sub-Saharan Africa, with high rates of cancers of the liver, prostate, cervix, and oesophagus (Bassett et al., 1995). The age-standardized incidence rates of oesophageal cancer among men were 30 per 100 000 in Harare in 1990–92 and 59 per 100 000 Bulawayo, further to the south, in 1963–72, and the rates in females were about 8 per 100 000 in both locations. Only three maize-based samples were analysed for fumonisins; one sample, a breakfast cereal intended for human consumption, contained fumonisin B at up to 4.9 mg/kg (Sydenham et al., 1993).

A survey of the mortality rates from oesophageal cancer in the USA between 1950 and 1969 indicated a clustering of cases among African–Americans in a narrow region along the southeastern coast of the Atlantic (Fraumeni & Blot, 1977; O’Brien et al., 1982; Brown et al., 1988). A small area of the Sea Islands and the coastal mainland in the vicinity of Charleston, South Carolina, was found to have a particularly high rate of death from this cancer. In Charleston County, the death rate among black males was about 170 per 100 000, and that among black females was 12 per 100 000, which is significantly higher than the State rate of 6.2 per 100 000.

Seven maize-based human foods purchased in retail outlets in Charleston in 1989 contained mean concentrations of fumonisin B₁ and fumonisin B₂ of 0.64 and 0.18 mg/kg, respectively (Sydenham et al., 1991).

Mortality rates in the USA tended to be inversely proportional to the socioeconomic indices of income and education (Fraumeni & Blot, 1977). The increased risk among black men in coastal South Carolina was associated with use of tobacco and alcohol, including ‘moonshine’ distilled from fermented maize meal and a low intake of fresh fruits (Brown et al., 1988).

Cancer of the oesophagus is the seventh most important cause of death from cancer among males in Brazil. The southern regions of Santa Catarina, Paranà, and Rio Grande do Sul have the highest incidence rate for oesophageal cancer in the country, 18 per 100 000 (Instituto Nacional do Cancer, 1989). The incidence is three to four times higher in men than in women, and marked differences in incidence are found in small geographical areas and as a function of time.

The south and western regions of Santa Catarina State, where there is a high volume of maize production and heavy consumption of maize by-products (especially polenta in rural areas), also have the highest incidence of oesophageal cancer. F. verticillioides was the predominant fungal contaminant in feed samples associated with mycotoxicoses in animals in Paranà (Sydenham et al., 1992a), and F. verticillioides and A. flavus were the most prevalent fungal contaminants in maize samples during the 1990–91 crop year in various regions of Paranà and the tropical regions, including Mate Grosso do Sul and Goias (Hirooka et al., 1996). A survey of the mycoflora in postharvested and stored maize in São Paulo indicated that Fusarium, Penicillium, and Aspergillus spp. were most common (Pozzi et al., 1995).

Fumonisin B₁ and B₂ concentrations < 38 and 12 mg/kg, respectively, were found in feed samples associated with outbreaks of mycotoxicoses in the State of Paranà (Sydenham et al., 1992a). Concentrations of up to 10 mg/kg of the two toxins were found in samples collected in the tropical regions of Mate Grosso do Sul, with lower concentrations in Paranà: fumonisin B₁ at 4 mg/kg and fumonisin B₂ at 3 mg/kg (Hirooka et al., 1996). Maize and maize-based food samples from Paranà and the southeast (São Paulo) contained fumonisin B₁ and fumonisin B₂ at concentrations up to 12 mg/kg and 10 mg/kg, respectively (Scaff & Scussel, 1999a). In São Paulo, maize cultivars grown in various regions contained up to 7 mg/kg of fumonisin B₁ and 2 mg/kg of fumonisin B₂ (Camargos et al., 2000). A high concentration of fumonisin B₁ (5 mg/kg) was found in maize meal samples from markets and supermarkets in Campiñas, São Paulo (Machinski & Valente Soares, 2000). Fumonisin B₁ was detected at up to 32 mg/kg in maize intended for human consumption collected in the western part of Santa Catarina (Hermans et al., 2000).

No information is available about the nutritional status of the populations of the regions of high incidence of oesophageal cancer in Brazil. Smoking and alcohol drinking have been implicated, although the prevalence of these habits does not differ from those in regions with lower incidence rates (Scaff & Scussel, 1999b). The consumption of hot beverages such as maté and chimarrao has also been implicated as a possible risk factor for oesophageal cancer (Victoria et al., 1987). An epidemiolgical study in Rio Grande do Sul indicated that smoking, alcohol and maté drinking, farm work, and having a father with cancer were more frequent in cases of oesophageal cancer (Dietz et al. (1998). Other studies suggested that consumption of maize, especially polenta, is a risk factor for this cancer in rural areas in southern and western Santa Catarina (Scaff & Scussel, 1999b).

2.4.2 Liver cancer

The role of fumonisins in the causation of liver cancer was evaluated in Haimen (Jiangsu County) and Penlai (Shandong Province), China, the mortality rate in Haimen (52–65 per 100 000) being about fourfold higher than that in Penlai. In a 3-year survey of 240 maize samples, a 10–50-fold higher fumonisin B content was found in Haimen. Most of the samples also contained aflatoxin B₁, but there was no significant difference between the two regions. However, deoxynivalenol was detected mainly in maize samples in Haimen. The authors suggested that a synergistic interaction between fumonisins, aflatoxins, and trichothecenes and the algal toxins, microcystins (Ueno et al., 1996), might contribute to the development of liver cancer (Ueno et al., 1997).

In Transkei, South Africa, the intake of aflatoxin B₁ correlated with the incidence of liver cancer (Van Rensburg et al., 1990), the rates varying between 3.8 per 100 000 in Butterworth and 7.7 per 100 000 in Kentani (Jaskiewicz et al., 1987b) and 2.4 per 100 000 in Kentani and 13 per 100 000 in Lusikisiki (Makaula et al., 1996), with no apparent association with the concentration of fumonisins in maize.

Experimental evidence for synergistic interactions between aflatoxin B₁ and fumonisin B₁ (Gelderblom et al., 1999a; Carlson et al., 2001) and between aflatoxin B₁ and nivalenol (Ueno et al., 1992) in inducing hepatic cancer in rats was reported.

2.4.3 Neural tube defects

In South Africa. high incidence rates of neural tube defects were recorded in Mpumulanga Province (3.6 per 1000) and Umzimkulu District (3.8 per 1000) in the Transkei (Venter et al., 1995; Ncayiyana, 1986). The rate in rural Transkei is approximately 5–10 times higher than that for blacks in Cape Town (Cornell et al., 1983). High rates of neural tube defects (5.7 per 1000) were also recorded in Hebei Province, China (Moore et al., 1997), and in northeastern USA between 1920 and 1949, peaking between 1929 and 1932 (2.3–4.3 per 1000). The latter epidemic seemed to correspond to two major socioeconomic events: the great depression and alcohol prohibition, although there are some discrepancies (Machon & Yen, 1971). A high rate of neural tube defects (2.7 per 1000) was also recorded in the lower Rio Grande valley in southern Texas, USA, among the offspring of women who had conceived during 1990–91 (Hendricks, 1999). An association between the epizootics of equine leukoencephalomalacia and porcine pulmonary oedema that occurred late in 1989 in the USA has been postulated (Ross et al., 1991). The concentrations of fumonisins in maize-based foods were high in both the Transkei and Hebei Province (Sydenham et al., 1990; Chu & Li, 1994), and maize-meal foods obtained in the USA during 1990–91 also had a relatively high concentration (1.2 mg/kg) of fumonisin B mycotoxins (Sydenham et al., 1991).

Interference with folate metabolism has been related to the development of neural tube defects, as supplementation with folate decreased the risk (Missmer et al., 2000). The blockage of folate uptake, a critical requirement during organogenesis (Lucock et al., 1998), by fumonisins was also implicated in the induction of neural tube defects (Stevens & Tang, 1997; Hendricks, 1999).

2.4.4 Acute mycotoxicosis

Consumption of rain-damaged, mouldy sorghum and maize by the inhabitants of 27 villages in the Deccan Plateau in southern India resulted in an episode of human mycotoxicosis in 1995. Diarrhoea was reproduced in 1-day-old cockerels fed contaminated grain from the affected households. An epidemiological survey in India indicated that consumption of unleavened bread prepared from mouldy sorghum or maize resulted in a gastrointestinal disease characterized by abdominal pain, borborymi, and diarrhoea. The victims were of low socioeconomic status and ate the mouldy grains mainly because of lack of access to other foods, such as rice.

The dominant mycoflora in the sorghum were Aspergillus, Fusarium, and Alternaria spp., while the first two fungal species were dominant in maize. Fumonisin B₁ was the most common mycotoxin in both sorghum and maize samples, and a relatively high concentration of aflatoxin B₁ was also detected in the maize. Fumonisin B₁ concentrations up to 8.5 mg/kg were associated with diarrhoea in laying hens and 1-day-old cockerels, and addition of fumonisin B₁ at 8 or 16 mg/kg of normal diet induced a similar response (Shetty & Bhat, 1997). The control of mycotoxins in human foods in India, especially as related to intake of aflatoxins and fumonisins, has been reviewed (Vasanthi & Bhat, 1998).

The chemical structures of fumonisin B₁, fumonisin B₂, fumonisin B₃, and fumonisin B₄ are given in Figure 1. Fumonisin B₁ is the diester of propane-1,2,3-tricarboxylic acid and 2S-amino-12S,16R-dimethyl-3S,5R,10R,14S,15R-pentahy-droxyeicosane in which the C-14 and C-15 hydroxy groups are esterified with the terminal carboxy group of propane-1,2,3-tricarboxylic acid (Bezuidenhout et al., 1988; Hoye et al., 1994). Fumonisin B₂ is the C-10 deoxy analogue of fumonisin B₁ in which the corresponding stereogenic centres on the eicosane backbone have the same configuration (Bezuidenhout et al., 1988; Harmange et al., 1994). The full stereochemical configurations of fumonisin B₃ and fumonisin B₄ are unknown, although the amino terminal of fumonisin B₃ has the same absolute configuration as that of fumonisin B₁ (i.e. 2S, 3S) (Hartl & Humpf, 1998). The configuration of the chiral centre on the tricarboxylic acid moieties has been determined by three groups, but with conflicting results. In the initial publication of the stereochemistry of fumonisin B₁, the S configuration was assigned to this centre (Shier et al., 1995), but subsequently both fumonisin B₁ and fumonisin B₂ were reported to have an R configuration (Boyle & Kishi, 1995a,b). A later study with a synthetic, optically active gamma-lactone related to the tricarboxylic acid correlated this with the same lactone in fumonisin B₁ and confirmed the R configuration for the chiral site on the side-chains (Edwards et al., 1999).

The most abundant fumonisins in naturally contaminated maize and maize-based products are fumonisins B₁ and B₂. Hence, analytical methods (and surveys) have been developed mainly for these two toxins. In general, methods developed for fumonisins B₁ and B₂ have been found to be valid for fumonisin B₃ as well (Sydenham et al., 1992b), although use of these methods is limited by problems in the supply of analytical standards for fumonisin B₃. No specific analytical methods have been developed for fumonisin B₄, and little is known about its natural occurrence.

Screening tests for fumonisins are based either on thin-layer chromatography (TLC) separation after appropriate clean-up of maize extracts or on commercially available enzyme-linked immunosorbent assays (ELISAs). Other immunologically based methods, such as dipstick (Schneider et al., 1995) and biosensor methods (Thompson & Maragos, 1996; Maragos, 1997; Mullett et al., 1998), have been described but have not found general use. Immunoaffinity columns have been designed to purify extracts before high-performance liquid chromatography (HPLC) separation and quantification of fumonisins B₁, B₂, and B₃ analogues, and have also been used in a direct fluorimetric method for rapid determination of ‘total fumonisin’ (Duncan et al., 1998).

The TLC and other chromatographic methods for fumonisins have been reviewed (Shephard, 1998). The reversed-phase technique developed by Rottinghaus et al. (1992) has been used in surveys of contamination of maize with fumonisins (Shelby et al., 1994a). When combined with an efficient extract clean-up procedure based on use of immunoaffinity columns and detection by densitometry, TLC can be considered quantitative (Preis & Vargas, 2000).

The performance characteristics of screening tests for fumonisins in maize, based on interlaboratory collaborative studies, have not been reported in the literature. However, in-house comparisons between HPLC methods and the various screening tests have been described. The TLC method of Rottinghaus et al. (1992) has been compared with HPLC over a contamination range of fumonisin B₁ of 1–250 mg/kg (correlation coefficient, r = 0.953; p < 0.0005; Schaafsma et al., 1998). The results obtained with a fibre-optic immunosensor in a direct competitive monoclonal antibody format with a fumonisin B₁–fluorescein isothiocyanate conjugate compare favourably with those obtained with HPLC (Maragos, 1997).

The commercial availability of ELISA methods has made them popular for screening for fumonisin contamination. Although the antibodies used in ELISAs are raised against fumonisin B₁, they generally have significant (but lower) cross-reactivity with fumonisins B₂ and B₃. The performance of ELISAs is generally assessed by comparison with HPLC determination of fumonisins and has been found to depend on the antibody used (Pestka et al., 1994; Usleber et al., 1994; Sydenham et al., 1996a,b; Kulisek & Hazebroek, 2000). The correlation between the results of HPLC and ELISA for naturally contaminated samples has been reported to vary from 0.51 (p < 0.05; Pestka et al., 1994) to 0.97 (p < 0.001; Sydenham et al., 1996a). However, such comparisons have generally shown an overall trend for the concentrations of ‘total fumonisins’ with ELISA to be greater than those determined in the same samples by HPLC.

Quantitative analytical methods for fumonisins have been reviewed (Sydenham & Shephard, 1996; Wilson et al., 1998; Shephard, 1998). Almost all these methods involve HPLC separation of fumonisins B₁, B₂, and B₃ after solvent extraction from maize or maize-based food matrices and purification by solid-phase extraction. In most surveys and studies on the natural occurrence of fumonisins involved use of variations or minor modifications of a few basic analytical methods.

The first quantitative HPLC method for determining fumonisins B₁ and B₂ in naturally contaminated maize involved extraction with methanol:water (3:1), clean-up on strong anion exchange solid-phase extraction cartridges, and quantification by reversed-phase HPLC after precolumn derivatization with the fluorigenic reagent, ortho-phthaldialdehyde (Shephard et al., 1990). The reproducibility of this method when used for naturally contaminated maize was studied collaboratively under the sponsorship of the Commission on Food Chemistry of IUPAC (Thiel et al., 1993). The method was subsequently improved and extended to include the determination of fumonisin B₃ (Sydenham et al., 1992b). A further collaborative study of its accuracy and reproducibility was conducted under the auspices of both the Commission on Food Chemistry of IUPAC and AOAC International and resulted in adoption of the method by the latter (Sydenham et al., 1996c). It has now been accepted as official AOAC method 995.15 for the determination of fumonisins B₁, B₂, and B₃ in maize (Trucksess, 2000). The performance characteristics in the international collaborative trials are given in Tables 9–12.

Table 10. Results of collaborative study by 12 laboratories of the accuracy and reproducibility of the method of Sydenham et al. (1992a) for the determination of fumonisins B₁, B₂, and B₃ in spiked and naturally contaminated maize

Table 11. Results of comparative study in 24 European laboratories of a maize sample containing fumonisin B₁ at approximately 2000 µg/kg and fumonisin B₂ at 1000 µg/kg prepared by spiking a naturally contaminated, ground maize sample

Table 12. Results of collaborative study by 23 laboratories of the reproducibility of high-performance liquid chromatography with immunoaffinity column clean-up for determining fumonisins B₁ and B₂ in maize and cornflakes

Modifications to the above method led to the development of other analytical methods. Acetonitrile:water (1:1) was formulated as an alternative extraction solvent, while reversed-phase C₁₈ solid-phase extraction cartridges have been used to purify extracts (Bennett & Richard, 1994; Rice et al., 1995). Although these cartridges yield less pure extracts, they are necessary for determination of the hydrolysis products of fumonisins and for cases in which strong anion exchange columns give poor recovery (Shephard, 1998). The introduction of immunoaffinity columns containing antibodies reactive towards fumonisins has greatly improved the clean-up step in analytical methods (Ware et al., 1994; Duncan et al., 1998). Most researchers report ed quantification by means of precolumn derivatization with ortho-phthaldialdehyde, despite its limited stability. Use of naphthalene-2,3-dicarbox-aldeyde has been proposed as an alternative (Bennet & Richard, 1994). Recent advances in analytical instrumentation have resulted in the introduction of bench-top liquid chromatography–mass spectrometers, which are more sensitive and specific for the detection and quantification of fumonisins (Shephard, 1998).

The official AOAC International method for fumonisins in maize has also been used for maize-based foods (Stack & Eppley, 1992), although problems have been reported in the recovery of fumonisins from certain matrices (Scott & Lawrence, 1994). In order to improve the recovery from these matrices, other extraction solvents have been investigated, while retaining the solid-phase extraction clean-up steps and ortho-phthaldialdehyde derivatization for HPLC quantification. The alternative solvents include methanol:borate buffer (pH 9.2; 3:1), methanol:water:hydrochloric acid (2 or 5 N; 3:1:0.3), methanol:0.1 mol/L HCl (3:1); acetonitrile:methanol:water (1:1:2), and acetonitrile:phosphate buffer (0.1 mol/L; pH 3.0; 1:1) (Scott & Lawrence, 1994; Zoller et al., 1994; Scott & Lawrence, 1996; Scott et al., 1999; Solfrizzo et al., 2000; De Girolamo et al., 2001).

As part of a programme to develop analytical methods for mycotoxins in foods at concentrations of interest for future legislation for the European Union, methods for fumonisins have been investigated within the Measurements and Testing Programme and the Community Bureau of Reference. A comparative study was conducted by 24 European laboratories on a contaminated maize sample containing fumonisin B₁ at approximately 2000 µg/kg and fumonisin B₂ at 1000 µg/kg (Visconti & Boenke, 1995; Visconti et al., 1996a). The results are summarized in Table 11.

Further improvements have been made in the analytical methods for fumonisins in maize, cornflakes, maize muffins, extruded maize, and infant formula, resulting in adoption of acetonitrile:methanol:water (1:1:2) as the extraction solvent and use of immunoaffinity columns for clean-up (Visconti et al., 1999; Solfrizzo et al., 2000). These extraction and clean-up techniques and HPLC quantification of preformed ortho-phthaldialdehyde derivatives of fumonisins B₁ and B₂ were used in a collaborative study of maize and cornflakes, which resulted in the adoption of the method by AOAC International (Visconti et al., 2001). The performance characteristics achieved in the collaborative study are shown in Table 12.

A collaborative study for determination of total fumonisins with a commercially available ELISA kit is under consideration by the methods programme of AOAC International (2000).

Whitaker and co-workers have studied the sampling variance associated with the testing of shelled maize for fumonisin (Whitaker et al., 1998). A bulk sample of about 45 kg (100 lbs) was taken from each of 24 batches of shelled maize which had been harvested from 24 fields in North Carolina, USA. Each bulk sample was riffle-divided into 32 1.1-kg test samples, and these were comminuted in a Romer mill. A nested design was used to determine the variation associated with sampling, sample preparation, and analysis. Briefly, 10 batches with a wide range of fumonisin concentrations were selected. From each batch, 10 comminuted test samples were taken randomly, and two 25-g portions were taken from each by riffle division. Finally, the concentrations of fumonisins B₁, B₂, and B₃ were determined by AOAC official method 995.15 (Sydenham et al., 1996c). At a batch contamination concentration of 2 mg/kg, the coefficient of variation associated with sampling (1.1-kg sample) was 17%, that associated with sample preparation (Romer mill and 25-g analytical portion) was 9.1%, and that for analysis was 9.7%. These values were independent of the fumonisin type (B₁, B₂, B₃, or total). The coefficient of variation associated with the total test procedure (sampling, sample preparation, and analysis) was 45%, which was of the same order of magnitude as that for measuring aflatoxin in shelled maize by a similar test procedure.

The proposed sampling plans are shown in Table 13. Some general precepts are:

Fumonisins are primarily produced by the fungi F. verticillioides (Sacc.) Nirenberg and F. proliferatum (Matsushima) Nirenberg, which are major pathogens of maize (Zea mays L.) around the world. The commodities most often contaminated with fumonisins are therefore maize and maize-based foods. Data on contamination of maize by fumonisins was submitted to the Committee by Argentina, Brazil, Canada, China, Denmark, Sweden, the United Kingdom, and the USA. Further data were derived from the literature published between 1995 and early 2000. Studies of the natural occurrence of fumonisins conducted before 1995 were reviewed by Shephard et al. (1996b).

Data on the occurrence of fumonisins B₁, B₂, and B₃ are presented in Appendices A, B, and C, respectively. The information is often incomplete: most authors reported the limit of detection (LOD), rather than the limit of quantification (LOQ), of the analytical method; the mean value in samples is frequently given, and in these cases (or when its use implied), the data were recalculated so that all data represent the means of samples, those below the limit of detection being taken as zero. When possible, the numbers of samples containing > 1000 and > 2000 µg/kg are also given. The former level represents the legislative limit for fumonisins B₁ and B₂ in Switzerland (FAO, 1997). The references in the appendices show the primary reference, which usually includes details on sampling and some information on analytical method; further references to methods are recorded as ‘A =’. Few references were given for sampling methods.

Little is known about the natural occurrence of fumonisin B₄. It is produced by strains of F. verticillioides, generally at lower concentrations than fumonisin B₁, B₂, or B₃ (Abbas et al., 1992; Plattner et al., 1996). Fumonisin B₄ was identified in 23 of 44 mouldy maize samples in the Republic of Korea (mean, 1600 µg/kg; range, 80–11 000 µg/kg) at concentrations lower than those of fumonisins B₁, B₂, and B₃ in these samples (Seo & Lee, 1999).

Some surveys have been undertaken of contamination of food commodities other than maize products. Analysis of 41 local and imported beers in Canada by HPLC showed a low incidence of contamination, with only four samples containing > 2 ng/ml fumonisin B₁ (maximum, 59 ng/ml) (Scott & Lawrence, 1995). Of these, three contained fumonisin B₂ at a concentration > 2 ng/ml (maximum, 12 ng/ml). Of the rest, seven contained 1 ng/ml or traces of fumonisin B₁. A survey of 46 imported beer samples in Canada by ELISA showed that 11 samples contained > 1 ng/ml (maximum, 25 ng/ml), and a further 11 contained 1 ng/ml or traces > 0.15 ng/ml (Scott et al., 1997). A similar survey of 32 Spanish beer samples by ELISA showed 14 positive samples (LOD, 3 ng/ml) with a range of 4.8–86 ng/ml (Torres et al., 1998). Another survey by HPLC of 29 domestic and imported beers in the USA showed 25 positive samples (> 0.3 ng/ml), with 17 samples containing > 1 ng/ml (maximum, 13 ng/ml) for fumonisins B₁ plus B₂ (Hlywka & Bullerman, 1999).

Investigation of milk and beef for contamination by fumonisins failed to raise any concern. Fumonisins were detected in beef muscle only after continuous exposure to highly contaminated feed (Smith & Thakur, 1996). In a survey of 165 milk samples in the USA, only one was found to be contaminated with fumonisin B₁ at a concentration of 1.3 ng/ml by gas chromatography–mass spectrometry and > 5 ng/ml by HPLC (Maragos & Richard, 1994). Studies in which cows were given feed contaminated with culture material equivalent to 75 mg/kg of feed or pure fumonisin B₁ at a dose equivalent to contamination of feed with 125 mg/kg showed no detectable carry-over of fumonisins into milk (LOD, approximately 5 ng/ml) (Scott et al., 1994; Richard et al., 1996). Re-analysis of these milk samples by ELISA (LOD, 0.5 ng/ml) also showed no carry-over into the milk of treated cows (Prelusky et al., 1996a). Similar studies in lactating sows given a diet containing fumonisin B₁ at 100 mg/kg for 14 days showed no toxin in the milk at a LOD of 30 µg/kg (Becker et al., 1995). Pigs fed diets containing the toxin at 2–3 mg/kg also did not show accumulation of fumonisin residues in muscle, although residues did accumulate in kidneys and liver while the contaminated feed was being consumed (Prelusky et al., 1994, 1996a,b). Replacement of the contaminated feed by clean feed resulted in a rapid decline in the concentrations. Studies in laying hens also showed no fumonisin residues in most tissue samples or in eggs (Vudathala et al., 1994; Prelusky et al., 1996a).

Contamination of rice with fumonisins was reported in the USA in samples collected from fields known to have plants with symptoms of Fusarium sheath rot (Abbas et al., 1998). Fumonisins B₁, B₂, and B₃ were detected (LOD, 500 µg/kg by HPLC) in 8 of 20 samples at concentrations < 5600 µg/kg. The appendices give the results of determinations in rice samples in Argentina, in which four of five imported samples showed contamination at maximum concentrations of 229 µg/kg fumonisin B₁, 126 µg/kg fumonisin B₂, and 16 µg/kg fumonisin B₃. A further 11 samples of locally produced rice showed no fumonisin contamination (LOQ, 8 µg/kg), and five rice samples included in a survey of foods in the United Kingdom also contained no detectable fumonisin (LOD, 10 µg/kg) (Patel et al., 1997). Fumonisin B₁ has been reported in rice in China (Appendix A). It has also been found on Fusarium-infected mouldy navy, adzuki, and mung beans (Tseng et al., 1995; Tseng & Tu, 1997) and in F. proliferatum-infected asparagus plants (Logrieco et al., 1998). In a survey of 35 sorghum syrup samples in the USA, one contained fumonisin B₁ (at 120 µg/kg; LOQ, 100 µg/kg; LOD, 10 µg/kg) (Trucksess et al., 2000). A single analysis of a sorghum meal sample in Burundi showed a very high concentration (28 200 µg/kg) of fumonisin B₁, but five sorghum samples were not contaminated (LOD not given) (Munimbazi & Bullerman, 1996). In an analysis of 20 sorghum and sorghum meal samples in Botswana, three contained fumonisin B₁ at a concentration > 20 µg/kg (mean, 43 µg/kg; maximum, 60 µg/kg; LOD, 20 µg/kg) (Siame et al., 1998). A further two samples of sorghum meal in Botswana contained traces of fumonisin B₁ (20 µg/kg) (LOD, < 20 µg/kg) (Doko et al., 1996), and five samples collected from control households during an epidemiological study in India contained fumonisin B₁ at a concentration of 70–360 µg/kg (Bhat et al., 1997). In an extensive study of 43 normal sorghum samples in India, only two had concentrations above the LOD of 25 µg/kg (range, 150–510 µg/kg) (Shetty & Bhat, 1997). A survey of the effects of storage on sorghum in Brazil over 1 year showed mean concentrations of 110–150 µg/kg (Da Silva et al., 2000).

When numerical data on the distribution of fumonisins in harvested maize and maize-based products were available in the surveyed literature, the numbers are given in Appendix A. These data include the combined results from surveys in the USA on a wide range of maize-based products sampled between 1990 and 1994 (Pohland, 1996). A number of authors presented distributions in the form of histograms, from which it was not always possible to extract unambiguous, relevant numerical information.

Limited data were available on annual variation in fumonisin concentrations in maize harvested in consecutive years, although it is clear that considerable variation can occur. The most extensive data on annual variation was collected for maize sampled in Iowa, USA, between 1988 and 1996 (Appendices A–C; Murphy et al., 1993; Rice & Ross, 1994; P.A. Murphy, personal communication). During the initial years of the survey (1988–91), the average concentration of fumonisin B₁ was > 2000 µg/kg each year, whereas in 1992–96 the concentration was significantly lower, with a mean < 450 µg/kg. A 5-year survey (1989–93) was conducted in South Africa on fumonisin concentrations in white and yellow maize. The concentrations of fumonisin B₁ in white maize over the 5 years ranged from 320 to 570 µg/kg; those in yellow maize ranged from 170 to 190 µg/kg during the first 4 years and rose to 680 µg/kg in the last year (Shephard et al., 1996b). Contamination of maize in Croatia with fumonisins B₁and B₂ over two consecutive seasons (1996 and 1997) has also been reported (Appendix A; Jurjevic et al., 1999). The data for Brazil drawn from the GEMS/Food programme is poor (Appendices A and B). Similar data are available for Argentina over a number of years (Appendices A–C), although the geographical origin of the samples included in the survey varied slightly.

The annual variation in total fumonisin concentrations over the 5-year period 1994–98 in various maize-based food products in the USA is shown in Appendix A, with data on contamination of maize meal (degerminated, partially degerminated, and whole grain) from the 1997 and 1998 US crops.

The stability of mycotoxins during food processing is affected by many factors, including: the moisture of the product, the toxin concentration and its location, the presence of additives, and the type of food matrix (Scott, 1984). These factors should therefore be considered with respect to each type of processing (e.g. milling, roasting, canning, oil extraction) when estimating the fate of the mycotoxin. After decompositon or destruction of mycotoxins, the resulting compound(s) should be identified to ensure that undesirable materials (with harmful biological effects) do not contaminate the food.

Dry milling of maize is a physical process by which the components of the grain are separated. The primary products derived from the process are grits, bran, germ, meal, and flour (Alexander, 1987). Since fumonisins are not expected to be destroyed during this process, their distribution has been evaluated in the fractions obtained from commercially and experimentally dry-milled maize. After commercial dry milling, fumonisins were found in flaking grits, flour, germ, and bran. The pattern of distribution after experimental dry milling varied slightly for different types of maize, but in general the concentrations were lower in grits and higher in germ, bran, and fines. As all the fractions found to contain fumonisins are used in the production of animal or human food, maize- based ingredients should be monitored for the presence of fumonisins (Katta et al., 1997).

Wet milling is used to obtain products such as germ (for oil and feed preparation), fibres and gluten (used in animal feed), and starch (for e.g. syrup). Analysis of the fractions obtained from laboratory-scale wet milling of maize contaminated with fumonisins revealed that the starch fractions did not contain measurable amounts of fumonisin B₁; 22% of the toxin was found in the steeping and process water, while the fumonisin content of the other fractions diminished in the order gluten > fibre > germ and represented 10–40% of the concentration found in the original maize. The presence of fumonisin in gluten, fibres, and germ could be a hazard to animal health, since these products are used as feed ingredients (Bennett et al., 1996).

As maize contaminated with fumonisin B₁ has a low density, 86% of the toxin can be removed in the buoyant fraction after treatment with saturated sodium chloride solution (Shetty & Bhat, 1999). Steeping maize in water, which is a step in wet milling, was also effective in reducing the concentrations of fumonisins B₁ and B₂ in maize, but sulfur dioxide delayed toxin extraction from the contaminated grain (Canela et al., 1996). Nevertheless, steeping maize in 0.2% sulfur dioxide solution containing fumonisin B₁ significantly decreased the amount of toxin in the solution, indicating that maize may contain fumonisin-binding constituents (Pujol et al., 1999).

The effects of heating on the stability of fumonisins during various food processing procedures have been studied. The temperatures required to achieve reductions > 50% in the concentration of fumonisin B₁ and/or fumonisin B₂ varied according to the process: for example, 190 °C for dry or moist maize meal (Scott & Lawrence, 1994) or frying maize chips (Jackson et al., 1997); extrusion temperatures for batter made from maize flour (Pineiro et al., 1999); 218 °C in roasting maize meal (Castelo et al., 1998a), and 160 °C for extrusion cooking of maize grits (Katta et al., 1999). The results indicate that fumonisins are fairly stable to heat and that significant removal occurs only during processes that reach temperatures > 150 °C. At each temperature tested in a specific food-processing system, the stability of the fumonisin was time-dependent (Jackson et al., 1996) and was strongly affected by other factors, some of which are mentioned above. The efficiency with which fumonisins are extracted from heated foods may decrease with time (Bordson et al., 1993), probably because of binding to the food matrix (Scott & Lawrence, 1994). As a consequence, controlled recovery experiments must be conducted for each type of processing system to ensure that the most appropriate methods of detection are used (Scott & Lawrence, 1994). Methods are needed to differentiate between binding and removal of fumonisins (Bullerman & Tsai, 1994).

Populations who consume large amounts of maize and maize products (e.g. in countries in South and Central America) are at high risk of exposure to fumonisins. The technique for preparing staple foods in these countries, such as tortillas, involves alkaline cooking and heating (nixtamalization), during which hydrolysed fumonisins are formed. Hydrolysed fumonisin B has been detected in commercial masa and tortilla chips, probably due to formation during nixtamalization (Murphy et al., 1996). In studies of the fate of fumonisins during processing and toxicological studies of the products, nixtamalized F. moniliforme culture (Voss et al., 1996a) and a maize-based diet containing F. proliferatum remained toxic after processing (Hendrich et al., 1993). Although the main toxic product formed from fumonisin during nixtama-lization is hydrolysed fumonisin, other products may be formed (Hendrich et al., 1993).

Treatment of fumonisin–contaminated, ground maize with calcium hydroxide resulted in transfer of the aminopentol moiety and, possibly, also the tricarballylic acid moiety to the aqueous fraction; these could easily be separated. Maize kernels from which the pericarp had been partially removed contained significantly greater amounts of fumonisin B₁ and its aminopentol moiety, indicating that this treatment would reduce the concentration of fumonisins in maize (Sydenham et al., 1995). Use of a modified nixtamalization procedure on maize contaminated with fumonisin B₁ resulted in a 100% reduction in the fumonisin content and removed the mutagenic potential of the maize extracts (Park et al., 1996).

Although it was expected that the fumonisin in nixtamalized maize products would be in the hydrolysed form, the content of that form in masa and tortillas in Mexico was lower than the content of fumonisins B₁ and B₂. Several explanations were proposed, including incomplete nixtamalization, a high initial concentration of fumonisin, and only partial removal of the pericarp, in which most of the fumonisin is located (Dombrink-Kurtzman & Dvorak, 1999).

Better control of fumonisins can thus be achieved by monitoring the degree of pericarp removal and analysing samples from each stage of production. Nixtamalization alone does not ensure complete detoxication of fumonisin-contaminated maize, and the process should be carefully controlled to ensure at least a significant reduction in fumonisins and other toxic products in the processed material.

The fermentation of naturally contaminated maize to produce ethanol led to only limited degradation of fumonisin B₁, most of which was recovered in distillers’ dried grain, thin stillage, and distillers’ soluble fractions (Bothast et al., 1992). Fumonisins have also been found in beer, indicating that the toxins persist under the conditions (temperature, pH) prevailing during the brewing process (Scott & Lawrence, 1995; Scott et al., 1997; Hlywka & Bullerman, 1999).

Surveys of the frequency of occurrence and concentrations in maize and maize products throughout the world are summarized in Appendices A–C. Over 60% of the surveyed products contained detectable amounts of fumonisins. Unfortunately, most of the reports did not contain sufficient information to allow determination of the representativeness of the data, i.e. they lacked information on sampling methods and involved insufficient numbers of samples to allow statistical evaluation of the fumonisin contamination of the diet.

The highest concentrations of fumonisins are found in visibly damaged or mouldy maize. The data show that the concentrations and incidence of contamination vary considerably in relation to the commodity tested and the source. The highest frequency was recorded in maize feeds, followed by ground maize products such as flour, grits, polenta, semolina, and gluten, maize kernels, and miscellaneous maize foods. The list of commercial retail foods that may be contaminated with fumonisins includes maize flour, grits, polenta, semolina, snacks, cornflakes, sweet, canned, and frozen maize, extruded maize, maize bread, maize-extruded bread, biscuits, cereals, chips, pastes, starch, infant foods, gruel, purée, noodles, popcorn, porridge, tortillas, tortilla chips, masa, popped maize, soup, tacos, and tostadas. Generally, processed maize foods have lower concentrations and a lower frequency of contamination than untreated foods. These differences may be the result of dilution of maize in food commodities or may depend on differences in maize cultivars or in the quality requirements for different destinations. Additionally, fumonisins are water-soluble, and processes that involve washing or water treatment may result in their partial or complete removal from the final food product.

In this evaluation, only the consumption of contaminated maize or maize-containing food products was considered, as the contributions of other commodities to the intake of fumonisins are too low and too variable to affect overall long-term exposure significantly.

Dietary intake of fumonisins was assessed in accordance with the recommen-dations of an FAO/WHO workshop on methods for assessing exposure to contaminants and toxins (WHO, 2000b). That workshop recommended that the best available data on concentrations in foods be used to estimate intake. For commodities that contribute significantly to intake, distribution curves should be generated to provide options to governments and national and international regulatory agencies for use in risk management. The Workshop further recommended that international estimates of dietary intake be generated by multiplying mean or median concentrations by the values for consumption of a commodity in the five GEMS/Food regional diets (WHO, 1998), which were derived from food balance sheets compiled by FAO. Since these sheets are available for most countries, they provide a set of data that are comparable across countries and regions of the world (WHO, 2000b). The five regions represented by the diets are Africa, Europe (which includes Australia, Canada, New Zealand, and the USA), the Far East, Latin America, and the Middle East. The regional diets represent the average per-capita availability of food commodities rather than actual food consumption, and data on availability generally result in an overestimate of consumption by about 15% (WHO, 1998). The workshop noted that, if available, national intake estimates should also be reported, as they may provide information about the intake of specific population subgroups or consumers of large amounts, which cannot be derived from GEMS/Food regional diets.

As the toxic effects in humans attributed to consumption of fumonisins in food are primarily of a long-term nature, this assessment is concerned solely with long-term intake.

Nine countries, Argentina, Brazil, Canada, China, Denmark, Sweden, the United Kingdom, the United States, and Uruguay, submitted information on the concentra-tions of fumonisins in maize and maize-derived foods. Fumonisins were detected in over 60% of all food products tested. The rate of detection was much lower in sound maize than in mouldy maize, and processed maize-containing foods generally contained lower concentrations of fumonisins than maize grain, flour, or grits.

A frequency distribution of the concentrations of fumonisins in maize was derived from available data in 1997 and published as part of an assessment of human intake of fumonisins in the Netherlands (de Nijs et al., 1998a). All maize consumed in the Netherlands is imported, and most was from Europe, South America, and the USA, although some was imported from Asia and Africa. As the concentrations of fumonisins and the incidences of fumonisin contamination reflected those found in the submitted data, for the purposes of this analysis, they were taken as representative of the maize available in trade throughout the world. Analysis of data available since 1997 showed little change in the patterns of incidence and concentration of fumonisins in maize and maize-based foods.

The concentrations of fumonisins were shown by the least-squares method to be distributed log-normally. The arithmetic mean concentration in the 349 samples used in the distribution was 1.36 mg/kg, and this distribution was combined with appropriate food intakes for assessing intake.

Data on food intake from the GEMS/Food regional diets (Table 14; WHO, 1998) were combined with the distribution of fumonisin concentrations in maize in commerce described above to yield estimates of fumonisin intake. A commercially available software product, @Risk (Palisade Corp.) was used with Microsoft Excel 97 to derive the distributions. Each simulation was run with Latin Hypercube sampling and 25 000 iterations.