WHO FOOD ADDITIVES SERIES: 52

First draft prepared by

Professor R. Kroes
Institute for Risk Assessment Sciences, Utrecht University, Soest, Netherlands

and

Dr P. Verger
National Institute for Agricultural Research, SAFE Consortium on Food Safety, Brussels, Belgium

Annatto extracts have been used for over two centuries as a food colour, especially in cheese, and various types are now used in a wide range of food products. Annatto extracts are obtained from the outer layer of the seeds of the tropical tree Bixa orellana. The principal pigment in annatto extract is cis-bixin, which is contained in the resinous coating of the seed itself. Processing primarily entails the removal of the pigment by abrasion of the seeds in an appropriate suspending agent. Traditionally, water or vegetable oil is used for this purpose, although solvent extraction is also employed to produce annatto extracts with an higher pigment content. Microcrystalline bixin products of 80–97% purity have been developed in response to the need for more concentrated annatto extracts.

Annatto extracts were evaluated by the Committee at its thirteenth, eighteenth and twenty-sixth meetings (Annex 1, reference 19, 35, 59–61).

At its eighteenth meeting, the Committee considered the results of long-term and short-term tests in experimental animals fed an annatto extract containing 0.2–2.6% pigment expressed as bixin. A long-term study in rats provided the basis for evaluation; the noobservedeffect level (NOEL) in this study was 0.5% in the diet, the highest dose tested, equivalent to 250 mg/kg bw. A temporary acceptable daily intake (ADI) for annatto extract was established at 0–1.25 mg/kg bw.

The Committee re-evaluated annatto extract at its twenty-sixth meeting, when the results of the metabolic studies that had been requested became available. Studies of mutagenicity, additional long-term (1year) studies in rats, and observations of the effects of annatto extract in humans were also considered. No evidence was found for the accumulation of annatto pigments in the tissues of rats fed at low concentrations (20–220 mg/kg bw per day of annatto extracts containing up to 2.3% bixin/norbixin mixture) for 1 year. Studies in both rats and humans showed that although annatto pigments are absorbed from the intestine into the blood, clearance from the plasma is rapid.

The NOEL in the original long-term study in rats was determined as 0.5% in the diet, equivalent to 250 mg/kg bw, and the ADI was established at 0–0.065 mg/kg bw of annatto extract expressed as bixin. In this re-evaluation, the Committee considered the highest concentration of bixin in the material tested (i.e. 2.6%) and established an ADI on the basis of the content of bixin.

B. orellanais grown in many countries and various procedures are used to produce annatto extracts from the seeds for commercial use. The following extracts were considered for evaluation.

Annatto extract (solvent-extracted bixin): Annatto B¹

The seeds are extracted with solvent to dissolve the pigment. The extract is filtered to remove insoluble material. Subsequent processing involves removal of fats and waxes, solvent removal, crystallization and drying.

Annatto extract (solvent-extracted norbixin): Annatto C

The seeds are extracted with solvent to dissolve the pigment. The extract is filtered to remove insoluble material. Subsequent processing involves removal of fats and waxes, removal of the solvent, crystallization and drying. Aqueous alkali is added to the resultant powder, which is heated to hydrolyse the pigment and then cooled. The aqueous solution is filtered, and acidified to precipitate the norbixin. The precipitate is filtered, washed, dried and milled to give a granular powder.

Annatto extract (oil-processed bixin suspension): Annatto D

The seeds are abraded in hot vegetable oil to remove the pigment.

Annatto extract (aqueous-processed bixin): Annatto E

The seeds are abraded in cold aqueous alkali (potassium or sodium hydroxide) to remove the pigment. The resulting suspension is acidified to precipitate the bixin. The precipitate is filtered, washed, dried and milled to give a granular powder.

Annatto extract (alkali-processed norbixin): Annatto F

The seeds are abraded in cold aqueous alkali (potassium or sodium hydroxide) to remove the pigment. Additional alkali is added to the resultant suspension, which is heated to dissolve the pigment and then cooled. Fats and waxes are removed. The aqueous solution is filtered, and acidified to precipitate the norbixin. The precipitate is filtered, washed, dried and milled to give a granular powder.

Annatto extract (alkali-processed norbixin, not acidprecipitated): Annatto G

The seeds are abraded in cold aqueous alkali (potassium or sodium hydroxide) to remove the pigment. Additional alkali is added to the resultant suspension, which is heated to dissolve the pigment, and then cooled. Fats and waxes are removed. The aqueous solution is filtered, and dried. Potassium carbonate may be added.

Bixin and norbixin, the main pigments contributing to the colour of annatto extracts, are present at different concentrations in different commercial preparations. These preparations are traded between primary processors of annatto seeds and the final vendors of colour products to the food companies. They are, however, too concentrated to add directly to foods and require dilution with carriers such as vegetable oil (with emulsifiers), propylene glycol or alkali. The non-pigment fractions of the concentrated extracts are not well characterized.

At its present meeting, the Committee evaluated new studies involving several concentrated preparations containing bixin and norbixin. The new studies consist primarily of 28-day and 90-day studies, disposition studies, studies of effects on microsomal oxidation enzymes and studies of genotoxicity with these formulations, one study of developmental toxicity and data on the potential allergenicity of annatto extract.

The pigment content, expressed as bixin and norbixin, of the extracts considered for evaluation is as follows:

New toxicological data were made available for four of these extracts: annatto B, C, E and F.

2.1.1 Absorption, distribution, metabolism and excretion

Preliminary data from studies in rats were reviewed previously by the Committee at its twenty-sixth meeting (Annex 1, references 59, 60). Since then, additional and new information regarding the absorption of bixin and its action on metabolic enzymes has become available and is described below.

(a) Published studies in animals

Rats

In studies in rats, Philp (1981) used the same three preparations of bixin as those administered by van Esch et al. (1959) in long-term feeding tests.

These were:

• OSB, a solution containing 0.1% annatto (0.22% bixin) and thermal degradation products, in vegetable oil;
• R10, a suspension containing 0.02% annatto (mainly bixin, 1.84%), in vegetable oil;
• WSA, a watersoluble preparation containing 0.1% annatto (mainly norbixin, 0.27%).

In a 4week study, groups of four male and four female Wistar rats were fed diets containing 0% or 5% of OSB, R10 or WSA. Animals received either: (1) diet containing annatto extract during the first 2 weeks and normal diet for the second 2 weeks; or (2) normal diet for 2 weeks followed by the diet containing annatto extract for 2 weeks. In the animals that were killed immediately after receiving annatto extract for 2 weeks, measurable amounts of yellow pigment were observed in the blood, but in animals that were killed 2 weeks after treatment with annatto extract had stopped, only trace amounts were detected. Yellow pigment was also found in the adipose tissue of animals treated with OSB and R10, but not in animals receiving WSA. Chromatographic analysis of these pigments confirmed that they were not major annatto pigments (bixin or norbixin). There was also a clear difference in the degree of discoloration of the fat in animals killed immediately after cessation of treatment, compared with that in animals killed 2 weeks after treatment had stopped, indicating that clearance of the pigment was rapid. Faeces were collected during the second week of treatment and analysed for pigment content. About 20% of the administered dose of OSB and WSA and about 55% of R10 was recovered unchanged from the faeces.

In a second series of studies, five groups of four male and four female Colworth Wistar rats were given large single doses of undiluted OSB (2 ml/kg bw), R10 (2 ml/kg bw) or WSA (10 ml/kg bw) by gavage. Controls received either water (10 ml/kg bw) or sunflower oil (2 ml/kg bw). Blood samples taken by cardiac puncture 3 h and 24 h after dosing were analysed for pigment content (Table 1).

Table 1. Plasma concentration of pigment in rats given a single dose of annatto extract

From Philp (1981)

The results indicate that the absorption and clearance of the water-soluble preparation, WSA, is good, while the extracts in vegetable oil are less well absorbed and are metabolized and excreted more slowly. The low plasma concentrations of R10 are in agreement with the previous finding that high concentrations of unabsorbed pigment are present in the faeces of animals treated with R10 (Philp, 1981).

(b) New unpublished studies

Rats

A study in rats treated with single and repeated oral doses of annatto extracts B, E or F was designed to provide pharmacokinetic data on the absorption, distribution and excretion of bixin and norbixin. The annatto extracts used were from the same batches as used in the 90-day studies of toxicity, described below. Analysis of the annatto preparations showed that annatto B (solventextracted) had a pigment content of 92%, of which 97% was cis-bixin and 1.7% cis-norbixin; annatto E (aqueousprocessed) had a pigment content of 26%, of which 90% was cis-bixin and 4% cis-norbixin; Annatto F (alkali processed) had a pigment content of 38% of which 87% was cis-norbixin and no cis-bixin. Authentic standard samples of purified bixin and norbixin were also used. Sprague Dawley rats (108 males and 108 females) strain Crl:CD(SD)IGS BR VAF/plus, with a body-weight range of 180–240 g, were housed either in groups of three in polypropylene cages, or individually in metabolism cages. Food (RM1 diet, SDS UK Ltd) was withheld overnight, and returned 2 h after dosing; water was available ad libitum throughout the study. Annatto B and annatto E were suspended in corn oil and annatto F was suspended in 0.5% carboxymethyl cellulose in sterile water. Rats were given annatto B, E or F by gavage in a single oral dose of 100 mg/kg bw or 1000 mg/kg bw and killed by exsanguination (three males and three females per dose and per extract) 2, 4, 8, 12 and 24 h after dosing.

Pooled plasma samples (from animals of the same sex, treated with the same annatto extract, at the same dose, and killed at the same time; n = 72) were analysed for bixin and norbixin (cis and trans isomers of both compounds) using a validated assay. Pharmacokinetic analysis was performed for the major components.

In addition, urine was collected before and after dosing from animals killed 24 h after dosing, and analysed for bixin and norbixin using the above methodology validated for use in urine. Bone marrow and livers were excised from rats killed at 4 h and 12 h and analysed for bixin and norbixin using the plasma assay method. All carcasses and faeces collected from rats killed at 24 h were frozen and stored in case they should be required for analysis at a future date.

After dosing orally with annatto B at 100 or 1000 mg/kg bw, the following chemical species were present in the plasma of male and female rats: 9’cisbixin, trans-bixin (retention time, 26 min), 9’cis-norbixin, trans-norbixin, di cis-norbixin and a norbixin isomer with retention time, 6.8 min. After dosing with annatto E (at both concentrations), the above isomers plus an additional trans-bixin species with retention time of 267 min were detected in the plasma of both sexes. In contrast, after administration of annatto F (at both doses), only the norbixin isomers (9’cis-norbixin, trans-norbixin, di cis-norbixin and the isomer with a retention time 6.8 minutes) were present in the plasma of male and female rats.

Plasma concentrations of 9’cis-norbixin were higher than those of 9’cis-bixin after the administration of annatto B, E or F, despite the fact that annatto B and E contain >90% 9’cis-bixin. The T_max for the major component in plasma for each extract in males and females at both doses was 2–4 h; by 12 h, only trace amounts of bixin remained, although concentrations of norbixin were still measurable at 24 h.

When the oral dose was raised from 100 mg/kg bw to 1000 mg/kg bw, the plasma concentrations of the major components of annatto B (9’cis-bixin), annatto E (9’cis-bixin) and annatto F (9’cis-norbixin) were increased, but not by 10-fold. The peak concentrations of bixin and norbixin in plasma are shown in Table 2.

Table 2. Pharmacokinetic parameters in Sprague Dawley rats treated orally with annatto extracts B, E or F

The uptake of bixin isomers into liver and bone marrow was examined; the isomers present in these tissues were consistent with those seen in plasma at 4 h and 12 h for each extract, but the assay method was not validated for these tissues, so no quantitative conclusions can be drawn (Bowman Research, 2002a).

Humans

Levy et al. (1997) developed a technique using reversed-phase high-performance liquid chromatography (HPLC) to determine concentrations of bixin and norbixin in human plasma, to a sensitivity of 5 µg/l. After a normal breakfast, seven male and female volunteers each ingested a single dose of 1 ml of a commercial annatto food colour containing 16 mg of cis-bixin and approximately 0.5 mg of cis-norbixin in soya bean oil, followed by a glass of milk. Blood samples were taken 0, 2, 4, 6 and 8 h after ingestion; in some subjects, additional samples were taken after 24 and 48 h. No control of food intake was made after 6 h (Levy et al., 1997). The average values and range of concentrations of bixin and norbixin in the plasma of the subjects are shown in Table 3.

Table 3. Plasma concentrations of bixin and norbixin in volunteers given a single dose of commercial annatto food colour

From Levy et al. (1997)

2.1.2 Biotransformation

No new information available.

2.1.3 Effects on enzymes

In view of the increase in absolute and/or relative liver weight observed in the 90-day studies of toxicity with annatto extract, it was decided to assay samples of the liver for CYP (cytochrome P450) enzymes, to determine whether induction of CYP enymes might account for the increased liver weight. At the termination of the 90-day studies of toxicity with annatto extracts B, E, and F conducted and reported later, liver samples were taken from 10 males and 10 females in the groups with the highest numbers of animals (Bowman Research, 2002b). Microsomes prepared from these samples were frozen, transported, and analysed for the expression of CYP apoproteins using highly specialized Western blotting techniques by which levels of specific apoproteins can be assessed (Boobis, 2002).

Total CYP content and the levels of several of the CYP apoproteins were sexually dimorphic. In all cases, the data were consistent with those reported in previous publications. The annatto extracts did not cause an increase in total CYP content, with the possible exception of annatto E when administered at a high dose to females; total CYP content in females receiving annatto E at a dose of 1000 mg/kg bw was 1.42 nmol/mg microsomal protein compared with the control value of 0.96 nmol/mg microsomal protein.

Annatto B and E induced expression of CYP1A2, but not of CYP1A1, suggesting that induction was via an aryl hydrocarbon-independent mechanism. None of the extracts induced CYP2B1 or CYP2B2, and hence none acted via a phenobarbital-like mechanism. None of the extracts induced CYP2E1 expression. CYP3A1 was induced by two to threefold in some groups, whilst CYP3A2 expression was not affected consistently. In comparison with the positive control (pregnanolone 16alpha-acarbonitrile, PCN), any PCN-type induction by the annatto extracts was extremely modest.

All three annatto extracts induced the expression of CYP4A (primarily CYP4A2 and CYP4A3), annatto F having the most potent effect. There was an appreciable difference between the sexes, females exhibiting a much weaker response, consistent with published data. Hepatic microsomal CYP4A was constitutively expressed in both male and female rats, levels being four to fivefold higher in males. Annatto F caused a dose-dependent increase in the expression of CYP4A in males. There was no increase in the expression of CYP4A in the females receiving low doses of annatto F, but in the females receiving high doses, there was a five to six-fold increase in expression. Similarly, annatto B caused a dose-dependent increase in the expression of CYP4A in male rats, whilst any increase in females was very modest. The effects of annatto E were less consistent, with the expression of CYP4A in males increasing by approximately three-fold, apparent only in the group receiving the higher dose. In females, the response was also not as consistent, but increased expression of CYP4A was observed in groups receiving both the low dose (approximately three-fold) and the high dose (approximately two-fold). The positive control, clofibric acid, produced the anticipated increase in the expression of CYP4A. The increases in expression of CYP4A in male rats after treatment with high doses of the annatto extracts, particularly annatto F, were similar to those observed after treatment with clofibric acid (Boobis, 2002).

Bixin has also been investigated for its effect on drug-metabolizing enzymes. A variety of carotenoids were administered to male Wistar rats at a concentration of 300 mg/kg of diet over a period of 16 days and the activities of a range of enzymes in the liver, lung, kidney and small intestine were assessed. Control animals received diet only or the enzyme-inducing agent, 3-methylcholanthrene. CYP enzyme activity, glutathione-S-transferase and reduced glutathione status, and carotenoid uptake into the tissues were all assessed. Bixin was shown to induce expression of the CYP1A1 iso-enzyme in the liver, lung and kidney and, to a lesser extent, the CYP2B1/2 enzyme in the liver. None of the carotenoids had detectable effects on intestinal enzymes or on glutathione status (Jewel & O’Brien, 1999).

2.1.4 Pharmacological and other biochemical studies

The pharmacological activity of extracts of the root of Bixa orellana has been considered in previous evaluations by the Committee. Since then, a few observations have been made concerning the effects of annatto preparations obtained from the seeds of Bixa orellana on the physiological and biochemical functioning of the body after administration to animals.

A suspension of the seeds of Bixa orellana in oil, known as "bush tea", is used in the West Indies as a folk remedy for the treatment of diabetes mellitus. Preliminary studies in dogs (Morrison & West, 1982) demonstrated that bush tea had significant hypoglycaemic activity. To investigate the mechanism of this effect, the residue of chloroform-extracted sun-dried seeds of Bixa orellana was dissolved in either oil (peanut or olive) or 95% ethanol and administered at doses equivalent to 1 g and 2 g of annatto (100–200 mg/kg bw) by stomach tube to 10 anaesthetized mongrel dogs of each sex. Six experiments were performed on each animal, to determine the effect on a 3-h oral glucose tolerance test. Both types of extract (oil or ethanol) produced a mild suppression of the rate of increase of plasma glucose at the lower dose, and at the same time depressed plasma insulin concentrations, but not significantly. At the higher dose, although the oil extract produced hypoglycaemia, the ethanol extract caused a hyperglycaemic effect (Morrison & West, 1985).

In a subsequent study, the ethanol extract previously shown to cause hyperglycaemia was dried to produce a reddish-brown crystalline solid that was redissolved as a neutral aqueous solution. This was then fed to male and female mongrel dogs each day for a period of 14 days, so that each dog received the equivalent of 2 g of the crystalline solid (125–222 mg/kg bw per day). Oral glucose tolerance tests performed immediately and at weekly intervals, confirmed the hyperglycaemic action of this extract of annatto, which persisted for 4 weeks after treatment had ended. Concomitant administration of riboflavin (3 mg per day) to some animals greatly reduced the hyperglycaemic effect. Examination by electron microscope of the liver and pancreas of treated animals revealed disarrangement of internal mitochondrial structure, fusion of mitochondria and formation of a large number of residual bodies. The apparent stacking of the hepatic mitochondrial cristae was also a striking feature. Dogs treated additionally with riboflavin exhibited only minimal mitochondrial changes (Morrison et al., 1987).

In a further study in anaesthetized mongrel dogs, an annatto extract (the residue of chloroform-extracted sun-dried seeds of Bixa orellana dissolved in 95% ethanol) purified by column chromatography was monitored for biological activity using the oral glucose tolerance test. This revealed that the active constituent responsible for the hyperglycaemic action was the methyl ester trans-bixin (relative molecular mass, 394). As little as 0.8 g of the purified methyl ester transbixin, equal to 60 mg/kg bw per day, produced a sustained hyperglycaemia, similar to that seen with 15 g of the crude extract. The purified material also caused mitochondrial damage in the liver and pancreas but, as with the crude extract, pre-treatment of the dogs with riboflavin a few days before the trans-bixin was administered resulted in little or no damage (Morrison et al., 1991).

This hyperglycaemic activity has been confirmed in rats, but an hypoglycaemic effect has been reported in a study in mice. The authors showed that when rats were given norbixin in their drinking-water for 21 days, hyperglycaemia ranging from 26.9–52.6% (at a dose of 8.5 and 74 mg/kg bw, respectively) above control values (p < 0.01) was observed. On the other hand, in mice given norbixin, hypoglycaemia ranging from 14.4–21.5% (at doses of 0.8 and 66 mg/kg bw, respectively) below control values was observed. The pancreatic beta-cells were functional, as indicated by the presence of hyperinsulinaemia in rats and hypoinsulinaemia in mice treated in this way (Fernandes et al., 2002).

Together with 16 other Guatemalan medicinal plants, Bixa orellana was investigated for ability to prevent platelet aggregation. Bixa orellana, and a number of traditional remedies, inhibited the thrombin-induced aggregation of washed human platelets (Villar et al., 1997).

A number of food colours, including bixin, have been shown to inhibit the production in vitro of immunoglobulin E (IgE) by rat spleen lymphocytes, at doses of 10 and 1 mmol/l. Variable effects were seen on the production of immunoglobulin G (IgG) and immunoglobulin M (IgM), depending on the dose, but water-soluble colourings enhanced the production of IgM at concentrations as low as 1 mmol/l. The authors concluded that these food colours might regulate the production of immunoglobulins (Kuramoto et al., 1996).

Bixin, together with other carotenoids such as beta-carotene, lutein and canthaxanthin, has been shown to suppress the respiratory burst induced by paramethoxyamphetamine (PMA) in rat peritoneal macrophages (Zhao et al., 1998). The action appears to be associated with the ability of carotenoids to scavenge superoxide, and the authors suggested a protective role for carotenoids in vivo to protect host cells from the harmful effects of oxygen metabolites (see section 2.2.4). Earlier work showed that bixin binds to the non-polar regions of mitochondria thought to be associated with high-energy states (Inada et al., 1971). Furthermore, bixin acts as an inhibitor of the ATP-forming process (state 3) associated with mitochondrial respiration (Hirose et al., 1972).

The effects of annatto on cellular enzymes and metabolites in the rat have also been investigated as part of a wider investigation of synthetic and natural food colours, in an attempt to provide a simple diagnostic test for toxicological damage. Four groups of eight male albino rats were used for the part of the study concerning annatto. Group 1 served as a control, group 2 was given annatto in a single dose of 800 mg/kg of diet, equivalent to approximately 80 mg/kg bw, and groups 3 and 4 were given 400 mg/kg of diet, equivalent to approximately 40 mg/kg bw, for 3 weeks. Animals in groups 2 and 3 were killed by decapitation 24 h after the last dose, while the animals in group 4 were killed 1 week after the cessation of treatment. Brain, liver and kidneys were removed from all animals, and cytoplasm and mitochondria were prepared for measurement of the activities of glucose-6-phosphate dehydrogenase (G-6-PD) and 6-phosphogluconate dehydrogenase (6-PGD). These enzymes were selected because of their key role in DNA and RNA synthesis. Annatto produced a minimal rise in the activity of these enzymes, mainly in the liver, and consistently less than that produced by synthetic food colours such as tartrazine, carmoisine and sunset yellow. In group 4, in which treatment had been stopped for a week before killing, the enzyme activities had returned to normal. It was noted that these effects could be a result of metabolism of the colour within the body (Hamama et al., 1987).

In a second study of identical design in male rats, annatto produced a slight increase in ATP and decreases in ADP and AMP in the brain, liver and kidney. As previously, these effects were much smaller than those seen with synthetic food colours, and normal values rapidly returned after treatment was stopped (Hamama et al., 1991).

2.2.1 Acute toxicity

Studies of acute toxicity in rats and mice were reported previously (Annex 1, references 35, 60). More recently, Germano et al. (1997) showed good cutaneous tolerability (as shown by histological examination of skin and hair) of extracts of Bixa orellana, after single or repeated application to the skin of rabbits, although these studies are not relevant for oral food use. The acute toxicity of annatto when administered orally is low.

2.2.2 Short-term studies of toxicity

Studies of toxicity in mice, rats, dogs and pigs have been previously reported; the results for all four species were discussed in the previous evaluations by the Committee. Relevant results are summarized below, together with the results of new studies.

(a) Previously considered studies

Mouse

In a recent study, pure norbixin or annatto extract containing 50% norbixin was given for 21 days at a dose of 0.8, 7.6, 66 or 274 mg/kg bw. Norbixin induced an increase in plasma alanine amino-transferase (ALT) while both norbixin and annatto induced a decrease in plasma total protein and globulins (p < 0.05). No signs of toxicity were detected by histopathological analysis of the liver, and no enhancement in DNA strand breakage was detected in liver or kidney from mice treated with annatto pigments, as evaluated by the comet assay. Hypoglycaemia was induced in the mice as discussed in section 2.1.4 (Fernandes et al., 2002).

Rats

Five groups of six male and six female Wistar rats received diets containing three annatto preparations, either singly or in combination, as follows:

The rats were killed during weeks 48–52 of treatment and were fed with control diet during the 24 h before death. Liver, kidneys and abdominal adipose tissue were removed at postmortem, weighed, examined (gross appearance only) and subjected to an extraction procedure to identify the presence of annatto pigments. Body weights and food intakes were similar in all treatment groups and not significantly different from those of the control group. The appearances and weights of organs and tissues showed that the animals were in good health, with no differences between the groups. Analysis of the extracts of liver, kidney, adipose tissue and carcass failed to reveal the presence of annatto pigments (Philp, 1981).

In more recent studies, a similar absence of toxicity was observed when rats were given drinking-water containing either annatto extract consisting of 50% norbixin (at a dose of 0.8, 7.5 or 68 mg/kg bw per day) or pure norbixin (at a dose of 0.8, 8.5 or 74 mg/kg bw per day) was for 21 days. No toxic effects on plasma clinical chemistry parameters were detected. A significant hyperglycaemic effect was reported (in contrast to the hypoglycaemia reported in mice); this is discussed in section 2.1.4 (Fernandes et al., 2002).

(b) New unpublished short-term studies of toxicity in rats

(i) 28-day study of toxicity with annatto extracts B, D, E, F and G

In a 28-day study of palatability, groups of five males and five females were fed diets containing varying amounts of annatto extract. Body weights were recorded twice weekly and food intakes were measured weekly. After 28 days of treatment, the animals were killed and subjected to full postmortem examination, major organs were weighed and tissues were fixed for histological examination.

It was concluded that annatto B was well tolerated in the diet at concentrations of up to 50 000 ppm; no dose-related absolute or relative organ weight changes were observed, but macroscopic examination revealed orange coloration of the gastrointestinal tract. At the highest dose, orange coloration of the mesentery was observed in a few animals.

Annatto D was well tolerated in the diet at concentrations of up to 11 000 ppm. Increases in concentration up to 22 000 ppm were less well accepted and there was a suggestion that gradual increase in dosage was better tolerated than abrupt introduction of a high dose. An increase in relative liver weights was recorded for all treated males compared with their respective controls, but only in females treated with up to 22 000 ppm. Microscopic examination of the livers of rats in the control group and of males treated with annatto D at up to 16 500 ppm revealed minor periacinar hepatocyte hypertrophy in 4/5 treated males. Macroscopic examination again revealed orange coloration of the tongue and gastrointestinal tract in the majority of the treated animals. In addition, the mesentery tissue was stained orange in several of the treated animals.

Annatto E at concentrations of up to and including 40 000 ppm was found to be acceptable to the animals. However, increased relative liver weights were seen in all treated animals. Microscopic examination of the livers revealed diffuse and periacinar hepatocyte hypertrophy in rats treated with 20 000–30 000 ppm or 20 000–40 000 ppm over the period of study. Again, orange staining of the tongue and gastrointestinal tract was reported for all animals receiving annatto E. The majority of the treated animals had orange staining of the mesentery.

Treatment with annatto F resulted in orange coloration of the tongue and gastrointestinal tract in the majority of the animals and of the mesentery in a few animals. The most significant finding in this study was the increased relative liver weight seen in all groups treated with annatto F (at concentrations ranging from 7000 to 20 000 ppm).

Animals treated with annatto G showed similar orange coloration of the tongue and gastrointestinal tract to that in animals treated with annatto F. Increased liver weights, accompanied by periacinar to diffuse hepatocytic hypertrophy, were noticed at all concentrations (7000–30 000 ppm) and concentrations of >15 000 ppm were less well tolerated (Huntingdon Life Sciences Ltd, 2000a, 2000b, 2001a–c).

(ii) 90-day study of toxicity with annatto extracts B, C, E and F

The four materials tested comprised two preparations of bixin (annatto B and annatto E) and of norbixin (annatto C and annatto F), one with an high pigment content and one with a lower pigment content. Preparations of annatto B, E and F were investigated in the same laboratory in compliance with good laboratory practice (GLP), including quality assurance audit, and the study designs met the requirements of Japan, the United States Food & Drug Administration and the European Union. In addition, a functional observational battery of tests was performed on animals fed with annatto B and F. As the sample of annatto B used in the 28-day study contained high levels of residual solvent, a new sample of annatto B was obtained from the manufacturers. The samples of annatto E and annatto F used in the 90-day studies were from the same stock as that used in the 28-day studies.

Annatto B

Four groups of 20 male and 20 female rats (Crl: CD (SD) IGS BR VAF/Plus), aged 5 weeks received diets containing annatto B at a concentration of 0, 5000, 16 000 or 50 000 ppm for 90 days. Group mean achieved actual dosages over period of treatment were 407, 1311 or 4201 mg/kg bw per day for males in each group, respectively, and 449, 1446 or 4507 mg/kg bw per day for the equivalent groups of females. Since annatto B contained 92% bixin, these achieved doses of annatto B correspond to doses of bixin of 374, 1206 or 3865 mg/kg bw for males and 413, 1330 or 4146 mg/kg bw for females.

There was no overall significant effect on body-weight gain in either the males or females. Neither was there any effect of treatment on food consumption (which was within 3% of control values overall) or on food conversion efficiency. The functional observational battery tests revealed no treatment-related effects on behavioural, motor or neurological activity. Ophthalmoscopic examinations revealed no abnormalities.

Haematological investigations revealed no findings that were clearly related to treatment. Raised lymphocyte counts (9.28 ± 1.64 compared with 7.30 ± 1.72 × 10⁹/l in controls) were noted for males receiving annatto B at 50 000 ppm; however, although the change was statistically significant, the change was minor, was not seen in females, and no dose–response relationship was observed.

When compared with the controls, statistically but not biologically significant elevations were seen in plasma concentrations of phosphorus (2.17 compared with 1.97 mmol/l) in males at 50 000 ppm, in plasma concentrations of potassium in females at 16 000 or 50 000 ppm, and in plasma concentrations of glucose in females at 50 000 ppm.

The activities of alanine and aspartate amino-transferase were significantly raised in 2/10 and 3/10 females receiving annatto B at a concentration of 16 000 or 50 000 ppm, respectively. This effect was not seen in males. These findings were not accompanied by histopathological evidence of liver damage and were considered by the author to be of only minor importance in the evaluation of the toxicity of annatto B.

Minor statistically significant variations in albumin:globulin ratios were observed, but these were not dose-related, and were not considered to be toxicologically significant.

Significantly raised concentrations of protein were recorded in urine samples obtained from males receiving annatto B at a concentration of 50 000 ppm. Microscopic examination of the urine sediment revealed the presence of "crystals with a red sediment on top" for one female at 5000 ppm and two males at 16 000 ppm. In another female at 5000 ppm and in three males and four females at 50 000 ppm, patches of red staining were seen on the microscope slides. These presumably reflect the dark colour of the annatto extract.

Analysis of the data on organ weights revealed a slight but significant increase in relative (but not absolute) liver weights for males (9% increase) and females (6% increase) receiving annatto B at a concentration of 50 000 ppm, when compared with controls. These increases were not associated with any histopathological findings. No other effects on absolute or relative organ weights were observed.

Microscopic examination revealed no findings that were related to treatment with annatto B.

The NOEL for annatto B was 16 ppm, equivalent to an overall achieved dosage of 1311 mg/kg bw per day (1206 mg/kg bw per day expressed as bixin) in the males, and 1446 mg/kg bw per day (1330 mg/kg bw per day expressed as bixin) in the females (Huntingdon Life Sciences Ltd, 2002a).

Annatto E

Four groups of the same strain of rats received diets containing annatto E at a concentration of 0, 3000, 10 000 or 30 000 ppm for 90 days. Group mean achieved actual dosages over the period of treatment were 224, 734 or 2204 mg/kg bw per day for males receiving 3000, 10 000 or 30 000 ppm, respectively, and 238, 801 or 2398 mg/kg bw per day for the corresponding groups of females. Since annatto E contains 26% bixin, the corresponding doses of bixin were 58, 191 and 573 mg/kg bw for males and 62, 208 and 623 mg/kg bw for females.

The administration of annatto E produced a number of treatment-related changes, particularly in the groups receiving the highest dose. At the intermediate dose, a reduced body-weight gain in the absence of any effect on food conversion efficiency in the females was most likely to have been caused by an effect of the palatability of the diet on food intake.

Increased liver weights seen in many of the treated animals, with the exception of females at 3000 ppm, were associated with slight adaptive centrilobular hepatocyte hypertrophy; no clear signs of liver pathology were noticed.

Thyroid follicular cell hypertrophy was present in some animals receiving annatto E at a concentration of 30 000 ppm.

Raised plasma concentrations of creatinine were seen in all females treated with annatto E and in males at 30 000 ppm. A slight increase in the weight of the kidney relative to body weight was seen in the females at the highest concentration. Raised plasma concentrations of phosphorus seen in males and females at the highest concentration of annatto E may indicate a reduction in the rate of glomerular filtration.

The no-observed-adverse-effect level (NOAEL) for both sexes can be considered to be the intermediate concentration of 10 000 ppm in the diet. This is equivalent to an intake of annatto E of 734 mg/kg bw per day for males and 801 mg/kg bw per day for females. As annatto E contains 26% bixin, these NOAELs correspond to a bixin intake of 191 and 201 mg/kg bw per day in males and females respectively (Huntingdon Life Sciences Ltd, 2002b).

Annatto F

Four groups of the same strain of rats received diets containing annatto F at a concentration of 0, 1000, 3000 or 9000 ppm for 90 days. Group mean achieved actual dosages were 79, 240 or 753 mg/kg bw per day for the groups of males receiving 1000, 3000 or 9000 ppm, respectively, and 86, 275 or 816 mg/kg bw per day for the equivalent groups of females. Since annatto F contains 38.4% norbixin, these dosages correspond to an intake of norbixin of 30, 92 or 289 mg/kg bw for males and 33, 106 or 313 mg/kg bw for females.

The administration of annatto F did not result in any treatment-related deaths. Transitory reductions in body weight and food consumption were noticed in males receiving the highest dose of annatto F in the first weeks of the experiment. The functional observational battery tests revealed no evidence of behavioural, motoror neurotoxicity, and ophthalmoscopic examination did not indicate the presence of any treatment-related lesions.

Slightly (but statistically significant) low erythrocyte volume fractions, concentrations of haemoglobin, red blood cell counts and mean cell volumes were apparent for females at 3000 or 9000 ppm, suggesting slight anaemia caused by annatto F.

In the animals receiving annatto F at a concentration of 9000 ppm, significantly increased activity of alkaline phosphatase was reported in males, with activity of alanine amino-transferase also being increased in two of these males. Concentrations of urea, creatinine, glucose, triglyceride and albumin were raised in males and females at 9000 ppm. Concentrations of total protein were raised in females. Reduced concentrations of alpha-1-globulin and beta-globulin were noticed in males, but concentrations of alpha-1-globulin were raised in females. The albumin:globulin ratio in males was increased. At 3000 ppm, total protein and concentrations of albumin were significantly raised in females, whereas the concentration of beta-globulin and the albumin:globulin ratio were increased in males. The plasma obtained from all treated animals had a definite yellow coloration, which was presumably norbixin.

A very marked increase in absolute and relative liver weights were seen in males treated with annatto F at a concentration of 9000 ppm and in females at 3000 and 9000 ppm. In females at 1000 ppm, only slight, although statistically significant, increases were noted. No changes were found microscopically, except in animals receiving a concentration of 1000 ppm, in which these changes were associated with centrilobular hepatocyte hypertrophy. The hypertrophy observed after 90 days of treatment was no greater than that observed after 28 days of treatment in the study on annatto F reported above.

The weight of the kidney was slightly increased in males and females at 9000 ppm and in males at 3000 ppm. Plasma concentrations of creatinine were raised in animals receiving annatto F at a concentration of 3000 or 9000 ppm, and concentrations of urea were raised in males and females at 9000 ppm.

The NOAEL for annatto F was 1000 ppm, this corresponds to to an overall achieved dosage of 79 mg/kg bw per day for males and 86 mg/kg bw per day for females, corresponding to an intake of norbixin of 30 mg/kg bw for males and 33 mg/kg bw for females (SanEi F.F.I Inc., 2002a).

(iii) Twoweek study to electronmicroscopically characterize liver cell hypertrophy caused by annatto C

Male Crj: CD (SD) IGS rats received diets containing annatto C at a concentration of 0 (one animal) and 9000 ppm (three animals) for 2 weeks. Absolute and relative liver weights were increased and mild hypertrophy and mild necrosis of hepatocytes were noticed on histopathological examination of rats treated with annatto C. Electron microscopic examination revealed an abundance of mitochrondia in the cytoplasm of hepatocytes of the treated animals. It was concluded that the hypertrophy of hepatocytes was linked to the increase of cytoplasmic mitochondria (SanEi F.F.I Inc., 2002c).

(iv) 90-day study of toxicity with repeated doses of annatto C

A 90-day study of toxicity with repeated doses of annatto C was carried out in Japan in compliance with GLP and the Guidelines For Designation of Food Additives and For Revision of Standards For Use of Food Additives of the Japanese Ministry of Health and Welfare. Four groups of 10 male and 10 female (Crj: CD (SD) IGS) rats were given diets containing annatto C at a concentration of 0, 1000, 3000 or 9000 ppm for 90 days. The overall intakes of annatto C were 69, 204 or 598 mg/kg bw for males and 76, 242 or 735 mg/kg bw for females, corresponding to an intake of norbixin of 63, 187 or 548 mg/kg bw per day for males, and 70, 222 or 673 mg/kg bw per day for females in the groups receiving low, intermediate and high doses respectively.

There was no effect on body weight or food intake, and no changes in haematology parameters. Changes in clinical chemistry parameters were observed in animals at the highest dose and, to a minor degree, at the intermediate dose. Increased absolute and relative liver weights were observed in animals of each sex in the groups receiving the two higher doses, and this was accompanied by hepatocyte hypertrophy in animals in these groups and focal necrosis in one male and one female in the groups receiving the highest dose. A small increase in kidney weight in each sex at the highest dose was not accompanied by any histopathological changes and may reflect a physiological response to the extra metabolic load imposed by the annatto extract. The lowest concentration of 1000 ppm, equivalent to an intake of annatto C of 69 and 76 mg/kg bw per day, or an intake of norbixin of 63 and 70 mg/kg bw, for males and females, respectively, can be considered to be the NOEL (SanEi F.F.I Inc., 2002a).

2.2.3 Long-term studies of toxicity and carcinogenicity

Three studies in mice and three studies in rats were previously submitted for evaluation. No new studies have become available. No separate studies of carcinogenicity, other than those submitted for previous evaluations, have been conducted.

2.2.4 Genotoxicity

Studies previously reported showed that annatto preparations to have no effect in three standard tests for mutagenicity (Lück & Rickerl, 1960; HavelandSmith, 1981). Studies reported since then are described in Table 4.

Table 4. Results of assays for genotoxicity with annatto extracts

No mutagenic activity was reported in silk worms treated with watersoluble annatto (Kawachi et al., 1980, 1981). Using extracts similar to those used in the test for dominant-lethal mutations reported above, Aranez & Rubio (1996) showed that chloroform extracts and petroleum extracts of the ripe seeds of Bixa orellana were genotoxic in dividing onion cells as shown by a significantly lower mitotic index value when compared with controls treated with dimethylsulfoxide (DMSO). On the other hand, the petroleum ether extract made from the residue of chloroform extraction did not cause more mitotic aberrations, although in the test for dominant-lethal mutations it was more genotoxic. A test for genotoxicity has also been carried out in Drosophila (Banzon & Aranez, 1984). Treatment with crude chloroform extract and the petroleum preparation produced significant increases in the number of flies with wing abnormalities, whereas the residue from the petroleum preparation did not.

In summary, the majority of data from tests with annatto extracts in bacteria have not shown evidence of genotoxic activity in vitro. Some weak positive findings at doses at which precipation was observed have been noticed, but only in the absence of metabolic activation. There were no significant responses in the presence of metabolic activation. The genotoxic activity would appear, therefore, not to be caused by bixin or norbixin, but by a noncoloured component of these samples, the concentration of which continues to increase in the culture medium at doses at which the colour component(s) precipitate. Indeed, if the normal high dose limitation had been exercized in the tests for mutagenicity in bacteria, then no mutagenic activity would have been found; doses well in excess of those at which precipitation first occurred were necessary in order to produce a significant response.

The results of tests for genotoxicity in mammalian cells in vitro are of a different complexity that is not so readily open to explanation. The closest concordance of the results of these tests is with annatto F, which was very weakly active in mouse lymphoma cells and inactive in Chinese hamster ovary cells. For annatto E and annatto F, however, there is a disparity in that the mutagenic activity of these extracts was expressed only in the absence of metabolic activation from S9 mix in mouse lymphoma cells, whereas in Chinese hamster ovary cells annatto E was clastogenic only in the presence of metabolic activation, but annatto F gave negative results at toxic doses. In addition, annatto B, which gave negative results in the mouse lymphoma cells, was clastogenic and induced endoreduplication in the absence of metabolic activation, again at toxic doses. While no persuasive explanation is forthcoming to account for these differences, it is possible that the weakness and the variability of the effects may play a role.

Since the effects observed in the tests in mammalian cells in vitro occurred at much lower exposures than those which produced mutations in bacteria, it cannot be presumed that the same moiety of the test material is involved in assays in mammalian cells versus assays in bacteria. Such incongruity can frequently result when a complex mixture is tested. The experimental data may indicate that the activities observed in the assay in mouse lymphoma cells were not caused by cis-bixin.

The only studies which show positive results in vivo can be criticized both from the point of view of methodology and for the fact that the extracts tested were made using solvents which are not permitted for the preparation of annatto extracts for food use. Recent unpublished data indicate that in the test for formation of micronuclei in mouse bone marrow, none of the annatto samples B, E or F demonstrated any potential to cause genetic damage.

2.2.5 Reproductive toxicity

(a) Multigeneration studies

Three-generation studies in rats were previously reported to the Committee.

(b) Developmental toxicity

In a study of developmental toxicity in rats, an annatto extract with a bixin content of 28% was administered by gavage to groups of 16–27 female Wistar rats at a dose of 0, 31.2, 62.5, 125, 250 or 500 mg/kg bw per day on days 6–15 of gestation. The annatto extract was neither maternally toxic or embryotoxic. Caesarean sections were performed on day 21; implantations, living and dead fetuses and resorptions were recorded. Fetuses were weighed and examined for externally visible abnormalities. One-third of the fetuses in each litter were examined for visceral abnormalities using a microsectioning technique. The remaining fetuses were cleared and stained with alizarin red S for skeletal evaluation. No adverse effects of the annatto extract on the dams were noted; there was no increase in embryolethality and no reduction in fetal body weight. Furthermore, the annatto extract did not induce any increase in the incidence of externally visible visceral, or skeletal abnormalities in the exposed offspring. The NOAEL for annatto extract in this study was 500 mg/kg bw per day, the highest dose tested, equivalent to an intake of bixin of 140 mg/kg bw per day (Paumgartten et al., 2002).

2.2.6 Special studies

(a) Cancer promotion

In an unpublished bioassay for mediumterm liver carcinogenesis in rats, diets containing annatto C (87.3% norbixin) at a concentration of 0, 300, 1000 or 3000 ppm (0, 20, 66 or 200 mg/kg bw, respectively) to male F344/DuCrj rats that had been treated previously with a single dose of Nnitrosodiethylamine (DEN) of 200 mg/kg bw, given intraperitoneally. As a positive control, phenobarbital sodium was given at a concentration of 500 ppm in the diet. Three groups of animals given diets containing annatto C at a concentration of 0, or 3000 ppm, or phenobarbital sodium at 500 ppm were not treated with DEN and served as controls. All groups were given the treated diet for 6 weeks. Partial hepatectomy was performed during week 3. The experiment was ended after 8 weeks. Body weights and food and water consumption in animals treated with annatto C did not differ from those of controls. Absolute and relative liver weights were significantly increased in the animals receiving annatto C at 1000 and 3000 ppm. Treatment with annatto C did not increase the quantitative values measured for liver cell foci observed to be positive for glutathione S-transferase placental form (GSTP) after DEN initiation, in contrast to treatment with phenobarbital sodium in the group of positive controls. These results indicate that annatto C lacks modifying potential for liver carcinogenesis in this test system (SanEi F.F.I Inc., 2002d).

(b) Antigenotoxic activity

A number of studies have been carried out to look at the possible anti-mutagenic action of carotenoids, including bixin, and their protective role in preventing cell damage induced by radiation or chemicals. For example, some studies have shown that bixin can provide protection against chromosomal damage induced by irradiation in rats (Thresiamma et al., 1996, 1997, 1998) and can also inhibit the mutagenic activity of hydrogen peroxide (H₂O₂) in Salmonella typhimurium strain TA102 (Kovary et al., 2001). In a model designed to investigate the antigenotoxic potential of norbixin, Kovary and colleagues (2001) also exposed Balb/c fibroblasts to H₂O₂ and evaluated DNA strand breakage. Prior incubation with norbixin showed firstly that norbixin alone was not genotoxic, secondly, that in concentrations of up to 50 mmol/l, norbixin reduced oxidative DNA damage in a manner that was inversely related to concentration and, finally, at higher concentrations it enhanced the damage induced by H₂O₂.

Finally, annatto extract (30% bixin) was studied for its possible mutagenicity and antimutagenicity in a test for formation of micronuclei in mice (Alves de Lima et al., 2003).

No increase in frequencies of micronucleated polychromatic erythrocytes was seen when mice were given diets containing annatto extract at a concentration of 1330, 5330 or 10 670 ppm (90, 360 and 470 mg/kg bw respectively) for 7 days. When annatto extract was given for 7 days at these concentrations, together with cyclophosphamide on day 7, no interference with the mutagenic action of cyclophosphamide was seen at the two lower concentrations. At the highest dose, however, the frequency of micronucleated polychromatic erythrocytes was increased. The authors suggested that at the highest dose, annatto extract might enhance the formation of the active metabolite of cyclophosphamide.

Annatto has been implicated as a cause of allergic reactions, these reactions generally taking the form of angioedema, urticaria or eczema, although one case of anaphylaxis has been documented. In well-controlled trials, the incidence of allergic reactions to annatto is small. Since annatto has been given in combination with other food colourings, it is not always possible to know whether the observed effects are caused by annatto or by one of the other ingredients or colours in the material used for the challenge.

For example, when 56 patients who had previously suffered from chronic urticaria and angioedema were given a gelatin capsule containing 25 µl of an extract of annatto (the amount of annatto contained in 25 g of butter), 27% of the patients revealed symptoms of urticaria and angioedema. This trial however was not double-blind, did not include placebo controls, and some patients had symptoms on entering the study (Mikkelsen et al., 1978).

In a study of 330 patients (121 men, 209 women) suffering from urticaria and angioedema, various potentially allergenic substances were administered in an attempt to diagnose the cause of these conditions. There were no placebo controls and the study was not double-blind. One hundred and twelve patients received annatto extract (5 mg or 10 mg); of these, 10% showed a positive reaction, 14% showed an uncertain reaction, while the remainder showed negative reactions (Juhlin, 1981).

In controlled studies in children, the incidence of adverse reactions to preparations containing annatto was much lower than previously described in adults, or was nonexistent. A study in Denmark in which 271 children (98 controls and 173 with atopic symptoms) underwent open oral challenges to a variety of food additives prepared in a lemonade solution, 17 children had positive reactions. The additives included preservatives, natural colours, synthetic colours, flavourings and acids. Twelve of the 17 children then underwent double-blind placebo-controlled challenges to the additives prepared in gelatine capsules. Of the 12, five reacted to the synthetic food colourings and one reacted positively to citric acid, but none reacted positively to the natural food colourings capsule which contained turmeric (2.5 mg/100 ml), annatto (1.6 mg/100 ml), beta-carotene (6.0 mg/100 ml), canthaxanthine (1.0 mg/100 ml) and beet colouring (5.5 mg/100 ml) (Fuglsang et al., 1993).

In a further study by the same author, 335 children underwent open oral challenges; 23 had positive reactions and of these 16 underwent double-blind placebo-controlled challenges in their own homes. Two patients experienced positive reactions to the natural food colouring (atopic dermatitis in one and urticarial symptoms in the other) (Fuglsang et al., 1994).

Two double-blind placebo-controlled studies of the effects of oral challenges have also been reported in adults. In the first of these, 132 subjects underwent high or low dose challenges with a range of food additives, administered (either alone or in combination) in opaque gelatin capsules tinted with iron oxide and titanium dioxide. Placebo capsules contained lactose powder. Annatto was administered at doses of 1 mg or 10 mg. Eighty-one patients completed the study. It was concluded that the prevalence of reactions to annatto in the population was estimated to be between 0.01% (lower limit) and 0.07% (upper limit) with a confidence interval of 95% (Young et al., 1987).

In the second double-blind placebo-controlled study of the effects of oral challenges in adults, 101 patients (25 men and 76 women) suffering from eczema were studied. After a standard elimination diet, the subjects received capsules containing a placebo control or one of four different mixtures of five food additives from the list below:

• Sodium benzoate;
• Sodium propionate;
• Sorbic acid;
• A mixture of food colourings (10 mg annatto, 10 mg erythrosine, 10 mg Ponceau 4R, 10 mg tartrazine, 10 mg patent blue V, 10 mg sunset yellow, 10 mg betanine, 10 mg curcumin, 10 mg quinoline;
• Placebo.

Twenty-five patients reacted to the food colourings while 76 did not. However, since 16 patients reacted to the placebo, reaction to the food colourings was not a statistically significant effect. Of the patients who reacted to the first oral challenge, only one-third reacted when challenged again (Vien et al., 1987).

An anaphylactic reaction attributed to annatto was reported in a 62-year-old man who had consumed Fiber One® cereal for the first time. This cereal product contains wheat bran, corn bran, aspartame, corn syrup, vitamins A, C, D, B₆, B₁₂, thiamine and annatto extract. Within minutes of eating this product, the patient developed symptoms characteristic of anaphylactic shock (generalized pruritis, generalized urticaria, angioedema of the eyes and lips, undetectable blood pressure and loss of consciousness). Skin prick tests (1:10 000) to milk, corn and wheat and annatto extract were all negative, but a positive reaction to 1:1000 annatto extract was obtained and the patient’s serum was positive for the presence of annatto-specific IgE. The IgE was shown to recognize a protein (relative molecular mass, 50) in the annatto extract, thought by the authors to be a contaminant from the pericarp extraction process. Annatto extract is widely present in food, and although it has been implicated in this case, an anaphylactic reaction of this type must be regarded as a very rare event (Nish et al., 1991).

In summary, annatto extracts have sporadically been associated with hypersensitivity reactions such as urticaria, eczema and angioedema. However, when investigated in well-designed double-blind trials, the incidence of reactions to annatto alone is very small.

Annatto is usually marketed as an extract of the annatto seed, which contains the active agents bixin and norbixin in amounts that can vary between <1% and >10%.

Throughout this monograph, all data on annatto, annatto extracts, bixin or norbixin are expressed as total bixin plus norbixin.

The Committee at its fifty-third meeting concluded that intake of annatto extracts would exceed the ADI for bixin if all food products contained this additive at the levels proposed in the draft General Standard for Food Additives (GSFA) and recommended that the relevant data should be re-evaluated in 2001. The current exposure assessment is based on data provided by the Natural Colour Association (NATCOL) and the International Association of Colour Manufacturers (IACM).

In contrast to the previous assessment, the Committee at its current meeting also considered the use of annatto extracts in beverages.

The report combines figures for production (in kg) and the distribution of market share in different food categories (in %) for annatto extracts. The information submitted about the distribution of the additive in various categories is more precise than the data submitted in 1999.

Table 5. Estimated theoretical maximum level of annatto extract in food and beverages, by the budget method

^a Highest level in food and beverages according to the GSFA

United States

In the United States (Table 6), it should be noted that 50% (20 tons) of the total amount of bixin/norbixin used is added to cheese. The amount of annatto extract used in each food category is divided by the total population of the United States. The resulting contribution is about 0.38 mg/person per day, corresponding to 10% of the current ADI (5% for cheese). This result is consistent with those submitted in 1999 for the United States (7.7% and 12.8% using poundage data from 1987 and 1996 respectively).

Table 6. Estimated per capita intakes of bixin and norbixin in the United States, based on production

^a Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat (2002)

France, Germany and the United Kingdom

In France, Germany and the United Kingdom (Table 7), the total production of bixin/norbixin is about 7% of that in the United States, whereas the population of these three countries combined represents 70% of that of the United States. The resulting estimate is therefore 10 times lower, i.e. 0.037 mg/person per day, corresponding to about 1% of the current ADI (70% for cheese).

Table 7. Estimated per capita intakes of bixin/norbixin in France, Germany and the United Kingdom, based on production

^a Population Division of the Department of Economic and Social Affairs of the United Nations Secretariat (2002)

The author provides a comparison between several use levels in Europe and combines the "typical" use levels for bixin/norbixin with the data on household consumption in France. It should be noted that the "typical" use levels provided by industry are close to those described by Scotter et al. and are below the standards for both the European Union and Codex Alimentarius Commission (see Table 8).

Table 8. Maximum permitted and actual use levels of bixin/norbixin in food in the European Union (mg/kg)

ND, no data provided

^b All levels "as consumed" (i.e. after making up with milk/water)

The result of this estimation is a total average intake of 0.54 mg/person per day corresponding to 0.009 mg/kg bw per day. The 97.5th percentile was also determined for each food category and the overall exposure was calculated assuming that a consumer only consumes one food item at a high level of intake and the others at the mean level. This exercise provides a total maximum intake of 0.02 g/kg bw (34% of the ADI) for consumers with a high intake of breakfast cereals. Considering that this additive is distributed in a relatively large number of food categories, it is possible to make a similar calculation based on the consumption of two food categories at the 97.5th percentile (Rees & Tennant, 1994). Such an exercise would result in a total maximum intake of 0.03 mg/kg bw (46% of the ADI) for consumers with a high intake of both desserts and breakfast cereals. In the previous assessment, a dietary exposure of 0.034 mg/kg bw was found using the European Union authorizations, which are lower than the typical use levels considered in this monograph.

Table 9. Typical average intakes of bixon/norbixin in the USA

The resulting estimate is about 0.63 mg/person per day, corresponding to 16% of the current ADI. This value is lower than that resulting from the previous assessment but is consistent considering that in 1999 the maximum use levels were taken into account

This estimate was based on the Northern American model diet developed by the Market Research Corporation of America (MRCA), and on data from the United States Department of Agriculture (USDA) on portion size. These figures for food consumption were combined with both usual and maximum use levels provided by the United States colour industry (International Association of Color Manufacturers, IACM). It should be noted that even if food categories are not the same in the United States and in the European Union, no use at a level >50 mg/kg was described.

The estimate provided by the sponsor is based on data from the United Kingdom (National Diet and Nutrition Survey for adults and children). A typical concentration of bixin/norbixin (see Table 8) is assigned to each relevant food and then multiplied by the amount of food consumed on each eating occasion to generate an intake figure. This results in an average estimate of 0.008 and 0.04 mg/kg bw, corresponding to 12% and 62% of the ADI for adults and children respectively. At the 97.5th percentile, the estimated intake was 0.02 and 0.1 mg/kg bw, corresponding to 34 and 160% of the ADI for adults and children, respectively.

Additional information was provided by Food Standards Australia New Zealand (FSANZ). Two data sets describe the exposure to bixin/norbixin in Australia and New Zealand by combining data on individual food consumption (consumers only) with both typical and maximum use levels for the food additive. Percentages of market share for "coloured" foodstuffs were used when available. It should be noted that the Australian maximum permitted levels are higher than those of the United States or the European Union (10–100 versus 1–50 mg/kg of food). Moreover, the industry use levels are higher than the "typical" use levels described by the sponsor both in the United States and in the European Union (1–100 versus 1–50 mg/kg of food). The resulting estimate is, for consumers only, a dietary exposure of 10 and 28 mg/day at the mean and the 95th percentile, respectively, using Australian permitted levels. Using concentrations provided by industry, the estimated exposure would be between 8 and 29 mg/day at the mean and the 95th percentile, respectively. The last calculations performed by FSANZ used the permitted levels for Codex Alimentarius and the resulting estimate of the exposure is therefore between 57 and 144 mg/day at the mean and the 95th percentile, respectively. Considering the fact that the range for Codex standards is between 10 and 600 mg/kg of food, such a result is not surprising.

Considering both the range of average intake of bixin (0.03–0.4 mg/day) and the 97.5th percentile (1.5 mg/day), a comparison was done with the ADIs for various extracts established by the Committee. Table 10 summarizes this exercise, assuming each time that all the bixin is provided by the same category of annatto extract.

Table 10. Comparison between the current levels of exposure to bixin and ADIs for various annatto extracts

The new data confirmed earlier findings that there appears to be at least partial absorption of bixin and norbixin, and that the pigments in watersoluble preparations of annatto are more readily absorbed than those in oilsoluble preparations.

Bixin was not detected in plasma after oral administration of norbixin to rats, suggesting that norbixin is not converted to bixin in the body. cis-Norbixin appears to be readily converted to trans-norbixin. The more polar acid norbixin is absorbed to a greater extent than the less polar bixin. The presence of norbixin in plasma after administration of annatto B and E suggests that bixin may be converted to norbixin in the body, but these preparations also contain norbixin that could have accounted for the norbixin levels in plasma. No bixin was detected in the urine following administration of any of the annatto extracts, but extremely small amounts of norbixin (<3% of the dose) were found in urine after administration of annatto F, and traces (<0.01%) after administration of annatto E. Bixin and norbixin are mostly cleared from the plasma within 24 h. The total percentage of the dose of bixin and norbixin absorbed cannot be determined from the available data. Although studies have shown that about half of the administered dose appears in the faeces, it is not possible to determine whether this is because the pigments are not absorbed and pass through the gastrointestinal tract, or whether part of the dose is absorbed and is excreted in the bile.

In humans given a commercial preparation containing 16 mg bixin and 0.5 mg norbixin, the concentrations of norbixin in blood were higher than those of bixin and persisted for longer, since bixin could not be detected in the plasma 8 h after a single oral dose whereas norbixin reached a peak after 4 h and was still detected after 48 h. It is unclear whether conversion of bixin to norbixin occurs in the body.

Examination of cytochrome P450 enzymes in liver samples at the end of 90-day studies in rats fed with one of several annatto preparations (annatto B, E and F) revealed that annatto B and E are inducers of CYP1A2. There was no evidence that any of the annatto extracts was a phenobarbital-type inducer, or was an inducer of CYP2E1. There were only slight increases in CYP3A1 and CYP3A2. All three annatto extracts induced CYP4A, particularly in male rats. The pattern of induction of CYP1A2 and CYP4A by different preparations in different sexes indicates that these effects are independent. Annatto F caused the greatest induction of CYP4A and there was an increase in the number of mitochondria observed by electron microscope with annatto C, observations which are consistent with the action of peroxisome proliferators. Liver weight increases were not related to the increase in cytochrome P450 enzymes. The absence of any further hypertrophy between days 28 and 90 of treatment, and the absence of pathological changes in the liver, may be compatible with metabolic adaptation.

Studies in mongrel dogs fed a chloroform-extracted preparation of annatto in glucose and similar studies in rats and mice fed an ethanol-extracted preparation of annatto, were considered not to be relevant to the extracts being evaluated.

Studies of genotoxicity in vitro revealed equivocal and inconsistent positive results only at concentrations that exceeded solubility or at concentrations that were cytotoxic. Since the results of tests on analytical grade bixin and norbixin were negative, some weak positive results obtained with the concentrated annatto extracts in bacterial tests in the absence of an endogenous metabolic activation system were considered to be caused by other components in the annatto preparations. Results of tests for mutagenicity in mammalian cells and for chromosomal aberration were inconsistent. Weak positive results at toxic concentrations were noted for some preparations in tests for mutagenicity in mammalian cells in the absence of an endogenous metabolic activation system, whereas weak positive results were noticed only in the presence of an endogenous metabolic activation system in tets for chromosomal aberration. Studies in mice receiving annatto B, E and F preparations did not demonstrate any potential to cause genetic damage in a test for micronucleus formation in bone marrow in vivo.

In its previous evaluations, the Committee concluded that annatto extracts are not carcinogenic. This conclusion was based on the results of tests with annatto preparations containing low concentrations of bixin. No new studies of carcinogenicity have become available, but in a study of the initiation/promotion of liver carcinogenesis, annatto C did not increase the incidence of preneoplastic lesions. Together with the results of the tests for genotoxicity and the absence of proliferative lesions in the short-term tests for toxicity, this is supportive of earlier conclusions.

A study of developmental toxicity in rats fed an annatto extract with a bixin content (28%) comparable to that of annatto E at doses of up to 500 mg/kg bw per day (equal to 140 mg of bixin/kg bw per day) confirmed the absence of developmental toxicity at this dose.

Studies of subchronic toxicity demonstrated that the annatto extracts tested, annatto B, C, E and F, have low toxicity, as only non-specific toxicity was reported at the higher doses tested. As is the case with genotoxicity, it is not clear whether the non-specific toxicity was attributable to bixin or norbixin or to other components present in the extracts. A common feature was the increase in absolute and relative liver weights at high and intermediate doses, which was accompanied in some cases by centrilobular hepatocellular hypertrophy.

The NOEL for annatto B was 16 000 ppm (equal to 1311 mg and 1446 mg of extract/kg bw per day for males and females respectively, corresponding to 1170 mg or 1290 mg/kg bw per day expressed as bixin, and 21 mg or 23 mg/kg bw per day expressed as norbixin) on the basis of urinary effects (elevated concentrations of protein in urine and crystals in urine sediment).

The NOEL for annatto C was 1000 ppm (69 mg and 76 mg/kg bw per day for males and females respectively, corresponding to 63 mg or 70 mg/kg bw per day expressed as norbixin) on the basis of increases in liver weight accompanied by hepatocellular hypertrophy and necrosis.

The NOEL for annatto E was identified as 10 000 ppm in the diet (734 mg and 801 mg/kg bw per day for males and females respectively, corresponding to 172 mg or 180 mg/kg bw per day expressed as bixin, and 8 mg or 8.8 mg/kg bw per day expressed as norbixin) on the basis of increases in thyroid and kidney weights and decreased spleen weights.

The NOEL for annatto F was 1000 ppm in the diet (79 mg and 86 mg/kg bw per day for males and females respectively, corresponding to 33 mg and 36 mg/kg bw per day expressed as norbixin) on the basis of increased kidney weights, haematological changes and alterations in serum proteins.

The differences in NOELs for annatto B (bixin) and annatto C (norbixin) indicate that extracts that mainly contain norbixin are more potent than those containing mainly bixin. The potencies of the different extracts cannot be explained on the basis of their bixin and/or norbixin contents. Other components in the extracts might contribute to or be responsible for the effects noted, and/or the differences in potency might have arisen from differences in bioavailability of the extracts.

A number of studies on possible allergic potential in humans were available, but the results of oral challenges were inconclusive because of inadequate study design or lack of statistical significance.

A revision of the previous intake estimate was performed on the basis of typical use levels of extracts expressed as bixin and norbixin provided by industry. These levels were combined with various data on national intake. The resulting average intake is observed to be between 0.03 mg and 0.40 mg per day. Estimated intake for consumers with a high intake reaches 1.50 mg per day, on the basis of data from the United Kingdom.

The Committee could not establish a generic ADI for the various annatto extracts on the basis of the data submitted and therefore established a temporary ADI for each of the individual preparations tested. With the application of a 200-fold safety factor to the NOEL for each of the annatto preparations, the following temporary ADIs were allocated:

• Annatto B: 0–7.0 mg/kg bw (based on NOELs of 1311 mg and 1446 mg/kg bw per day in male and female rats respectively).
• Annatto C: 0–0.4 mg/kg bw (based on NOELs of 69 mg and 76 mg/kg bw per day in male and female rats respectively).
• Annatto E: 0–4.0 mg/kg bw (based on NOELs of 734 mg and 801 mg/kg bw per day in male and female rats respectively).
• Annatto F: 0–0.4 mg/kg bw (based on NOELs of 79 mg and 86 mg/kg bw per day in male and female rats respectively).

No data on the potential toxicity of annatto D or annatto G were available, and no ADI could be established. An additional safety factor of 2 was applied to the NOELs, because of deficiencies in the database.

Comparison of the estimated intakes with the temporary ADI values was performed assuming that each annatto extract was a unique source of bixin/norbixin. These simulations show in each case that the estimated exposure for adults is <20% of the corresponding temporary ADI.

The Committee requested additional information to clarify the role that the non-pigment components of the extract play in the expression of the qualitative and quantitative differences in toxicity of the various extracts. In addition, the Committee requested data on the reproductive toxicity of an extract, such as annatto F, which contains norbixin.

Alves de Lima, R.O., Azevedo, L., Ribeiro, L.R. & Salvadori, D.M.F. (2003) Study on the mutagenicity and antimutagenicity of a natural food colour (annatto) in mouse bone marrow cells. Food Chem. Toxicol., 41, 189–192.

Aranez, A.T. & Bayot, E. (1997) Genotoxicity of pigments from seeds of Bixa orellana L. (atsuete). II. Determined by lethal test. Philipp. J. Sci., 126, 163–173.

Aranez, A.T. & Rubio, R.O. (1996) Genotoxicity of pigments from seeds of Bixa orellana L. (Atsuete) I. determined by allium test. Philipp. J. Sci., 125, 259–269.

Banzon, R.B., & Aranez, A. T. (1984) Studies on the mutagenicity of pigments of the seeds of Bixa orellana L. (atsuete). Nat. Appl. Sci. Bull. 36, 161–182.

BIBRA Working Group (1991) Annatto extracts: toxicity profile. BIBRA Toxicology International.

Boobis, A.R. (2002) Effects of dietary annatto blends on hepatic P450 apoprotein levels in the rat. Unpublished report from Imperial College, London, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Bowman Research Ltd. (2002a) Pharmacokinetic studies with annatto blends B, E and F in the rat. Unpublished report No. 0206 from Bowman Research Ltd, Newport, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Bowman Research Ltd. (2002b) Preparation of microsomes from rats livers from toxicology studies with annatto blends B, E and F. Unpublished report No. 0202 from Bowman Research Ltd, Newport, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Fernandes, A.C.S., Almeida, C.A., Albana, F., Laranja, G.A.T., Felzenszwalb, I., Lage, C.L.S., de Sa, C.C.N.F., Moura, A.S. & Kovary, K. (2002) Norbixin ingestion did not induce any detectable DNA breakage in liver and kidney but caused a considerable impairment in plasma glucose levels of rats and mice. J. Nutr. Biochem., 13, 411–420.

Fuglsang, G., Madsen, C., Saval, P. & Osterballe, O. (1993) Prevalence of intolerance to food additives among Danish school children. Pediatr. Allergy Immunol., 4, 123–129.

Fuglsang, G., Madsen, C., Halken, S., Jorgensen, M., Ostergaard, P.A. & Osterballe, O. (1994) Adverse reactions to food additives in children with atopic symptoms. Allergy, 49, 31–37.

Germano, M.P., DePasquale, R., Rapisarda, A., Monteleone, D., Keita, A. & Sanogo, R. (1997) Drugs used in Africa as dyes: 1. Skin absorption and tolerability of Bixa orellana L. Phytomedicine, 4, 129–131.

Hamama, A.A., Hewedi, F.M., Ibrahim, M.H. & Abdel-Rahim, E.A. (1987) Effect of synthetic and natural food colorants on subcellular fraction enzymes in albino rats. J. Drug Res. Egypt, 17, 231–238.

Hamama, A.A., Hewedi, F.M., Ibrahim, M.H. & Abdel-Rahim, E.A. (1991) Effect of synthetic and natural food pigments on some energetic metabolites in albino rats. J. Drug Res. Egypt, 20, 117–123.

Haveland-Smith, R.B. (1981) Evaluation of the genotoxicity of some natural food colours using bacterial assays, Mut. Res., 91, 285–290.

Hirose, S., Yaginuma, N. & Inada, Y. (1972) Energised state of mitochondria as revealed by the spectral change of bound bixin. Arch. Biochem. Biophys., 152, 36–43.

Huntingdon Life Sciences Ltd (2000a) Annatto B: Preliminary toxicity study by dietary administration to CD rats for 4 weeks. Unpublished report No. ATE010/990103 from Huntingdon Life Sciences Ltd, Huntingdon UK. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2000b) ATE011/990104, Annatto D: Preliminary toxicity study by dietary administration to CD rats for 4 weeks. Unpublished report No. ATE011/990104 from Huntingdon Life Sciences Ltd, Huntingdon UK. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2001a) Annatto E: Preliminary toxicity study by dietary administration to CD rats for 4 weeks. Unpublished report No. ATE012/990105 from Huntingdon Life Sciences Ltd, Huntingdon, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2001b) Annatto F: Preliminary toxicity study by dietary administration to CD rats for 4 weeks. Unpublished report No. ATE013/990106 from Huntingdon Life Sciences Ltd, Huntingdon, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2001c) Annatto G: Preliminary toxicity study by dietary administration to CD rats for 4 weeks Unpublished report No. ATE014/990107 from Huntingdon Life Sciences Ltd, Huntingdon, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2002a) Annatto B: Toxicity study by dietary administration to CD rats for 13 weeks. Unpublished report No. ATE015/003817 from Huntingdon Life Sciences Ltd, Huntingdon, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2002b) Annatto E: Toxicity study by dietary administration to CD rats for 13 weeks Unpublished report No. ATE016/002344 from Huntingdon Life Sciences Ltd, Huntingdon, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Huntingdon Life Sciences Ltd (2002c) Annatto F: Toxicity study by dietary administration to CD rats for 13 weeks. Unpublished report No. ATE018/004678 from Huntingdon Life Sciences Ltd, Huntingdon, England. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inada, Y., Hirose, S., Yaginuma, N. & Yamashita, K. (1971) Spectral changes of bixin upon interaction with respiring rat liver mitochondria. Arch. Biochem. Biophys, 146, 366–367.

Inveresk (2001a) Annatto type B: Testing for mutagenic activity with Salmonella typhimurium TA1535, TA1537, TA98, TA100, TA102, TA104 and Escherichia coli WP2uvrA (pMK101). Unpublished report no 20072 from Inveresk Research, Scotland, submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001b) Annatto Type E: Testing for mutagenic activity with Salmonella typhimurium TA1535, TA1537, TA98, TA100, TA102, TA104 and Escherichia coli WP2uvrA (pMK101). Unpublished report No. 20073 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001c) Annatto type F: Testing for mutagenic activity with Salmonella typhimurium TA1535, TA1537, TA98, TA100, TA102, TA104 and Escherichia coli WP2uvrA (pMK101). Unpublished report No. 20074 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001d) Bixin standard: Testing for mutagenic activity with Salmonella typhimurium TA1535, TA1537, TA98, TA100, TA102, TA104 and Escherichia coli WP2uvrA (pMK101). Unpublished report No. 20075 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001e) Norbixin standard: Testing for mutagenic activity with Salmonella typhimurium TA1535, TA1537, TA98, TA100, TA102, TA104 and Escherichia coli WP2uvrA (pMK101). Unpublished report No. 20076 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001f) Annatto type B: Chromosomal aberration assay with Chinese hamster ovary cells in vitro. Unpublished report No. 19870 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001g) Annatto type E: Chromosomal aberration assay with Chinese hamster ovary cells in vitro. Unpublished report No. 19920 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001h) Annatto type F: Chromosomal aberration assay with Chinese hamster ovary cells in vitro. Unpublished report No. 19988 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001i) Annatto type B: Mouse lymphoma cell mutation assay. Unpublished report No. 19767 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001j) Annatto type E: Mouse lymphoma cell mutation assay. Unpublished report No. 19928 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2001k) Annatto type F: Mouse lymphoma cell mutation assay. Unpublished report No. 19929 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2002a) Annatto type B: Micronucleus test in bone marrow of CD-1 mice, 0 h and 24 h oral dosing and 48 h sampling. Unpublished report No. 21816 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2002b) Annatto type E: Micronucleus test in bone marrow of CD-1 mice, 0 h and 24 h oral dosing and 48 h sampling. Unpublished report No. 21846 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Inveresk (2002c) Annatto type F: Micronucleus test in bone marrow of CD-1 mice, 0 h and 24 h oral dosing and 48 h sampling. Unpublished report No. 21979 from Inveresk Research, Scotland. Submitted to WHO by Annatto Interest Group (AIG), Cork, Ireland.

Ishidate, M., Sofuni, T. & Yoshikawa, K. (1981) Chromosomal aberration tests in vitro as a primary screening tool for environmental mutagens and or carcinogens. In: Inui, N., Kuroki, T., Yamada, M.A. & Heidelberger, C., eds, Mutation, promotion and transformation in vitro. Gann Monograph. Vol. 27, Tokyo: Japanese Scientific Society Press, p. 95.

Ishidate, M., Sofuni, T., Yoshikawa, K., Hayashi, M., Nohmi, T., Sawada, M. & Matsuoka, A. (1984) Primary mutagenicity screening of food additives currently used in Japan. Food Chem. Toxicol., 22, 623–636.

Jewell, C. & O’Brien, N.M. (1999) Effect of dietary supplementation with carotenoids on xenobiotic metabolizing enzymes in the liver, lung, kidney and small intestine of the rat. Br. J. Nutr., 81, 235–242.

Juhlin, L. (1981) Recurrent urticaria: clinical investigation of 330 patients. Br. J. Dermatol., 104, 369–381.

Kawachi, T., Komatsu, T., Kada, T., Ishidate, M., Sasaki, M., Sugiyama, T. & Tazima, Y. (1980) Results of recent studies on the relevance of various short-term screening tests in Japan. In: Williams, G.M., Kroes, R., Waaijers, H.J. & van de Poll, K.W., eds, The Predictive Value of Short-Term Screening Tests in Carcinogenicity Evaluation. Amsterdam: Elsevier/North Holland Biomedical Press, pp. 253–267.

Kawachi, T., Yahagi, T., Kada, T., Tazima, Y., Ishidate, M., Sasaki, M. & Sugiyama, T. (1981) Cooperative program on short-term assays for carcinogenicity in Japan. IARC Scientific Publication, 27, 323–331.

Kovary, K., Louvain, T.S., Costa e Silva, M.C., Albano, F., Pires, B.B.M., Laranja, G.A.T., Lage, C.L.S. & Felzenswalb, I. (2001) Biochemical behaviour of norbixin during in vitro DNA damage induced by reactive oxygen species. Br. J. Nutr., 85, 431–440.

Kuramoto, Y., Yamada, K., Tsuruta, O., & Sugano, M. (1996) Effect of natural food colourings on immunoglobulin production in vitro by rat spleen lymphocytes. Biosci. Biotechnol. Biochem., 60, 1712–1713.

Levy, L.W., Regelado, E., Navarette, S. & Watkins, R.H. (1997) Bixin and norbixin in human plasma: determination and study of the absorption of a single dose of annatto food color. Analyst, 122, 977–980.

Lück, H. & Rickerl, E. (1960) Food Additives and mutagenic effects—6th Report. Examination of food dyes allowed and first suggested in West Germany for mutagenic effects on Escherichia coli. Zeitschrift fur Lebenmitteluntersuchung und Forschung, 112, 157.

Mikkelsen, H., Larsen, J.C. & Tarding, F. (1978) Hypersensitivity reactions to food colours with special reference to the natural colour of annatto extract (butter colour). Arch. Toxicol., 1, 141–143.

Morrison, E.Y.St A. & West, M.E. (1982) A preliminary study of the effects of some West Indian medicinal plants on blood sugar levels in the dog. West Indian Med. J., 31, 194–197.

Morrison, E.Y.St A. & West, M.E. (1985) The effect of Bixa orellana (annatto) on blood sugar levels in the anaesthetised dog. West Indian Med. J., 34, 38–42.

Morrison, E.Y.St A., Smith-Richardson, S., West, M., Brooks, S.E.H., Pascoe, K. & Fletcher, C. (1987) Toxicity of the hyperglycaemic-inducing extract of the annatto (Bixa orellana) in the dog. West Indian Med. J., 36, 99–103.

Morrison, E.Y.St A., Thompson, H., Pascoe, K., West, M. & Fletcher, C. (1991) Extraction of an hyperglycaemic principle from the annatto (Bixa orellana): a medicinal plant in the West Indies. Trop. Geogr. Med., 43, 184–188.

Murthy, M.S.S. (1979) Induction of gene conversion in diploid yeast by chemicals; correlation with mutagenic action and its relevance in genotoxicity screening. Mutat. Res., 64, 1–17.

Nish, W.A., Whisman, B.A., Goetz, D.W. & Ramirez, D.A. (1991) Anaphylaxis to annatto dye: a case report. Ann. Allergy, 66, 129–131.

Paumgartten, F.J.R., De-Carvalho, R.R., Araujo, I.B., Pinto, F.M., Borges, O.O., Souza, C.A.M. & Kuriyama, S.M. (2002) Evaluation of the developmental toxicity of annatto inthe rat. Food Chem. Toxicol., 40, 1595–1601.

Philp, J.M. (1981) Studies on three commercial extracts of annatto (Ref D76/001) Unpublished report by Unilever Research Division, (cited in Annex 1, reference 59).

San-Ei Gen F.F.I. Inc. (2002a) 90 day oral toxicity study of annatto C (norbixin) in rats. Unpublished report No. 0138 from San-Ei Gen F.F.I. Inc, Osaka, Japan, submitted to WHO.

San-Ei Gen F.F.I. Inc. (2002b) Reversed mutation test of annatto C and F in bacteria (Ames test) Unpublished reports No. 0214 and 0215 from San-Ei Gen F.F.I. Inc, Osaka, Japan, submitted to WHO.

San-Ei Gen F.F.I. Inc. (2002c) Electron microscopy of the liver in rats fed annatto C for two weeks. Unpublished report No. 0174 from San-Ei Gen F.F.I. Inc, Osaka, Japan, submitted to WHO.

San-Ei Gen F.F.I. Inc. (2002d) Rat medium term liver carcinogenesis bioassay of annatto C (norbixin). Unpublished report No. 0211 from San-Ei Gen F.F.I. Inc, Osaka, Japan, submitted to WHO.

Sasaki, M., Sugimura, K., Yoshida, M.A. & Abe, S. (1980) Cytogenetic effects of 60 chemicals on cultured human and Chinese hamster cells. Senshokutai, 20, 574–584.

Scotter, M.J., Castle, L., Honeybone, C.A. & Nelson, C. (2002) Method development and analysis of retail foods for annatto food colouring material. Food Addit. Contam., 19, 205–222.

Thresiamma, K.C., George, J. & Kuttan, R. (1996) Protective effect of curcumin, ellagic acid and bixin on radiation induced toxicity. Indian J. Exp. Biol., 34, 845–847.

Thresiamma, K.C. & Kuttan, R. (1997) Effects of antioxidants on lung fibrosis induced in rats. J. Clin. Biochem. Nutr., 22, 125–129.

Thresiamma, K.C., George, J. & Kuttan, R. (1998) Protective effect of curcumin, ellagic acid and bixin on radiation induced genotoxicity. J. Exper. Clin. Canc. Res., 17, 431–434.

van Esch, G.J., van Genderen, H. & Vink, H.H. (1959) Uber die chronische Verträglichkeit von Annattofarbstoff. Zeitschrift-Lebensmittel-untersuchung undForschung, 111, 93–108.

Vien, N.K., Hattel, T., Justesen, O. & Norhom, A. (1987) Oral challenge with food additives. Contact Dermatitis, 17, 100–103.

Villar, R., Calleja, J.M., Morales, C. & Caceres, A. (1997) Screening of 17 Guatemalan medicinal plants for platelet antiaggregant activity. Phytother. Res., 11, 441–445.

Young, E., Patel, S., Stoneham, M., Rona, R. & Wilkinson, J.D. (1987) The prevalence of reaction to food additives in a survey population. J. R. Coll. Physicians Lond., 21, 5–11.

Zhao, W., Han, Y., Zhao, B, Hirota, S., Hou, J. & Xin, W. (1998) The effect of carotenoids on the respiratory burst of rat peritoneal macrophages. Biochim. Biophys. Acta, 1381, 77–88.

ENDNOTES:

1  To ensure clarity, the Committee adopted the designations B, C, D, E, F, G, as employed in the submitted information, to refer to the different annatto extracts under evaluation.

    See Also:
       Toxicological Abbreviations
       Annatto extracts (FAO Nutrition Meetings Report Series 46a)
       Annatto extracts (WHO Food Additives Series 6)
       Annatto extracts (WHO Food Additives Series 17)
       Annatto extracts (WHO Food Additives Series 44)
       ANNATTO EXTRACTS (JECFA Evaluation)