Caballero B. (ed.) Encyclopaedia of Food Science, Food Technology and Nutrition. Ten-Volume Set

Подождите немного. Документ загружается.

liking rating may remain poor in the presence of the

concept, but given the information about the product

benefits, the purchase intent may increase dramatic-

ally.

Frequency of Use

0020 Another type of acceptance measurement is the esti-

mated frequency of use. Market researchers usually

do studies to estimate volume, or amount to be con-

sumed, not just overall liking. Thus, a measure of

frequency reveals whether the product is highly

liked but might be consumed infrequently, highly

liked and consumed frequently, etc. Frequency is

measured in at least two ways. One instructs the

consumer to assign a number (e.g., times per month

or times per year) to show the estimated number of

times that the person would eat the product. Another

gives the consumer a number of eating occasions (e.g.,

10), and instructs the consumer to assign the number

of times out of the next set of consumption occasions

when the individual would eat the product. This ap-

proach is often used with the consumer given an array

of products, and instructed to assign to each product

the number of eating occasions out of the next 10 (or

whatever number). This is called the constant sum

method, and is used to determine the impact of

introducing a product on the frequency of consump-

tion of other products with which it competes for

consumption.

0021 Of all the acceptance measures, purchase intent is

probably the most popular, and the one on which many

marketing decisions are based. The typical purchase

intent scale comprises a limited number of points (the

most common being the five-point scale; Table 1).

Rather than computing an average purchase score,

the market researcher most typically computes the

distribution of purchase intent ratings across the

population of consumers who rated the product. This

generates a set of incidence statistics, treated as per-

centages. The term ‘top-two box’ refers to the propor-

tion of consumers who, having evaluated a specific

product, rate it as definitely or probably would pur-

chase.

0022 Purchase intent is important for a number of

reasons. First, it asks the consumer to describe what

action he or she would take – purchase or not pur-

chase. Second, many researchers create a normative

database from many such studies, so that when

they determine the purchase intent rating for a new

product they can also estimate the likelihood of

success. Third, a number of proprietary research pro-

cedures, aimed at estimating volume sales, use pur-

chase intent as part of the input information to the

model.

Purchase Ratings Coupled with Price Information

0023Purchase intent scales are often used along with

pricing information. Although in theory price infor-

mation should not affect product liking, it should

affect purchase intent. Thus, a product may be very

well liked, but the purchase intent may increase if

the product is offered at a low price, and decrease

if the same product is offered at a high price. This

implicit demand curve can be captured by a simple

equation: Purchase intent ¼k

(liking

)  (price

Typically the exponent for price, n, is much higher

in absolute value (and negative) than the exponent for

liking, (m, which is positive). This means that small

changes in price have more effect in lowering pur-

chase intent than the same small proportional change

in liking has in raising purchase intent.

Acceptance as Replacement of a Current Product

0024Another aspect of acceptance deals with the issue of

whether the product would replace or simply be

added to the consumer’s choices of foods to be pur-

chased. Although not strictly a question of degree of

liking, replacement questions deal with the issue of

how the food will be used. Marketers always search

for products that will be added to a person’s set of

foods, rather than one that has to fight to oust an-

other food. It is much easier to add to a person’s list

than to replace a food that already has a history of

being purchased.

Appropriateness – Day-Part and Situation

0025Food acceptance possesses a significant cognitive

component, viz., that foods have defined parts of

the day when they will be eaten, and other day-parts

when they are not customarily eaten. Much of this is

cultural, and labile, explaining why now cola drinks

are being consumed for breakfast. Consumers have

no trouble identifying the appropriate day-parts

for specific products, and often use implicit rules

to assign a product to breakfast, lunch, dinner,

and snack. Appropriateness and eating context is

emerging as an important aspect of food acceptance.

The same food can be highly acceptable for dinner,

but not particularly acceptable for breakfast (e.g.,

soup for North American consumers).

Relevance of the Product as Part of a Line of

Related Products

0026From the manufacturer’s point of view a single prod-

uct (e.g., a single flavor) has several aspects of

acceptance. One is the acceptance of the single prod-

uct by itself. Another is the degree to which the

product augments a line of already existing or

2578 FOOD ACCEPTABILITY/Market Research Methods

planned products. For instance, a fruit-flavored cereal

may be highly acceptable itself without reference to

other products, but may be redundant. That is, con-

sumers satisfied with one fruit-flavored cereal may

not really be interested in purchasing another, equally

liked fruit-flavored cereal featured in the same line.

To the marketer this means that the product is liked

but redundant. A practical outcome would be that the

item is rejected as part of the line. A substantial

amount of consumer research funds is devoted to

line optimization, or creating the proper combination

of items in such a way that the line maximizes the

number of consumers who would be satisfied by at

least one of the products offered in the line. Market

researchers call the solution of this problem the TURF

issue, or ‘total unduplicated reach and frequency.’

Product acceptance here is viewed as an issue that

transcends the more conventional rating of liking, be-

cause it calls into account the presence of, and reaction

to, other potential entries in the same product line.

Acceptance in Light of Sensory Satiety and

Boredom/Habituation

0027 Very few products enjoy continued consumer accept-

ance at constant level (given a constant amount of

advertising and promotion support). Sooner or later

consumers tire of most products, and consume the

product less frequently. This reduction in consump-

tion comes about two ways – through sensory-specific

satiety, and through boredom and habituation. A food

loses its acceptance very quickly after eating, so that

during a meal we tire of a food and stop eating it. This

is sensory-specific satiety – we may have room for

dessert even though we don’t want to continue eating

the main course. Habituation and boredom occur

when the consumer simply tires of a product, and

stops selecting it and/or eating it, even if the food

has not been consumed recently. To some degree the

more salient the sensory characteristics of the prod-

ucts, the more rapidly the product becomes boring. A

good example of this is the difference between steak

and hamburger. Steak wins against hamburger time

after time in preference tests. Yet people tire more

rapidly of steak after multiple consumptions, perhaps

because of the salient sensory factors. An even more

dramatic example of this occurs in the case of bread

versus steak. Bread almost never wins in acceptance

tests against steak, but we can eat bread day after day,

meal after meal. We cannot consume steak to the

same degree.

Branded versus Blind Product Acceptance

0028 Most, but not all product testing is done in a blinded

environment, wherein the researcher does not identify

the particular brand. Branding (the identification of

the brand), however, may make a great deal of differ-

ence in the acceptance rating when the product is

already on the market. For many common products

the brand may add substantially to the acceptance

rating. The difference between the blind and the

branded product shows the impact of the brand itself.

In developmental research it is easy to test products

branded because the researcher has control over the

product, and can instruct R&D to create products in

a form that cannot be identified. It is not so easy with

in-market products. When the researcher wants to

test competitor products that exist in the market,

attempts to ‘blind’ or disguise the product may

require unusual precautions. These have included

shaving off an imprint on the product, testing the

product in a red glass in red light (to prevent visual

identification), or somehow cutting the product in

order to eliminate a tell-tale shape that would imme-

diately signal the brand.

Implementing a Study of Food Acceptance

– Design and Field Considerations

0029The importance of acceptance as a precursor for

product success has led to a large number of proced-

ural recommendations and rules of thumb. Some of

these are based in science, others are simply dictated

by fiat through company policy.

Choice of Panelists who Participate

0030The composition of the sample generally is critical for

a valid acceptance study. At the early stage of evalu-

ation the particular choice of panelists is not too

critical, other than the panelists accept the product.

At the later stages, usually conducted by the market

researcher, sampling the proper individuals becomes

more important. The financial implications of a

wrong decision (so-called downside risk) grow rap-

idly as the product comes closer to being launched in

the marketplace. That is, if the study uses the wrong

consumers, then the results are immediately suspect

and the business decision may be faulty, even if the

study is executed perfectly. The consumers may be

chosen to be current users of the product, users of

the competitor products, and lapsed users (individ-

uals who used to use the product but no longer do so).

In the case of children who consume the product but

do not buy it, the panelists may comprise a group of

parents and children (not necessarily related), in

order to determine how the ‘gatekeeper’ (parent) re-

sponds, and how the ultimate consumer (child)

responds. The conventional market research sample

comprises some or all of these individuals, in suffi-

ciently large base sizes of individuals so that the

ratings from each subgroup can be analyzed.

FOOD ACCEPTABILITY/Market Research Methods 2579

What Type of Questionnaire is Most Appropriate for

the Acceptance Test?

0031 Most market researchers embed the acceptance scale

(whether liking or purchase) in a larger set of ques-

tions dealing with other aspects of the food, such as

the sensory and image aspects. For market researchers

product acceptance is only one part of the picture

(albeit an important part). Other important aspects

include image of the product, when the product will

be used, etc. The key exception here is a claims test,

whose goal is to demonstrate that one product is

preferred to another. The rule of thumb for a claims

test is that the researcher should limit the question to

liking or purchase intent alone, and not deal with any

other attributes that could cloud the findings, and

therefore the resulting claim.

Where Should the Acceptance Test be Conducted?

0032 There are two test venues – central location and home

use. A central location venue requires that the con-

sumer come to a central site in order to evaluate the

product. The site may be a shopping mall or a special-

ized test facility. A central location site affords the

greatest control over the evaluation, because a trained

food professional and/or interviewer can ensure that

the product is properly prepared and served. The other

test venue is one’s home, wherein the product is sent

home (e.g., by mail) or given to the panelists at a

shopping mall to take home. The product is prepared,

eaten, and often rated at home. The panelist may then

be telephoned to get the ratings, or may mail back (or

e-mail back) the completed questionnaire.

Some Typical Best Practices for Choosing

Consumers who Participate

0033 Conventionally, market researchers try to use large

representative consumer panels. Often (but not

always) this consists of testing panelists in different

markets, and of different ages. Historically, most food

products have been evaluated by females, except for

products specifically designed for males (which are

relatively few). In more recent years it has come to be

realized that geographical dispersion does not guar-

antee representativeness in the population, because

there are more fundamental ways in which people

differ that transcend geography. One newly emerging

concept is the existence of sensory preference seg-

ments, where individuals in a segment show rather

similar sensory preferences to product features. These

members of the same segment may live in different

locations, have different incomes, education, and pur-

chase history. The jury is still out on how to improve

the consumer panelist selection process in light of

these overwhelming differences among sensory pref-

erence segments.

Anchoring the Scale

0034Some practitioners feel that every point on the accept-

ance rating scale should be anchored with a verbal

descriptor, because they feel that it is the descriptive

term, not the numerical scale, that is important.

Other practitioners feel that the scale itself is relevant,

and that the only anchors need be at the top and the

bottom of the scale.

The Position of the Acceptance Rating in the Order

of Attribute Questions

0035Market researchers have not adopted any hard and

fast rules about the nature of the questionnaire, and

the location of the acceptance measure within the

questionnaire. Some researchers feel strongly that

the acceptance rating should come first in order for

that rating not to be influenced by other attribute

ratings. Other researchers feel just as strongly that

the acceptance rating should come last, because

otherwise the consumers will attempt to justify their

acceptance rating by having all other attribute ratings

fall into line. The majority of researchers pay little

attention to this issue, but the issue arises from time to

time.

Appropriate Statistics for Ratings

0036Most market researchers come from and/or operate

within a sociology-oriented system, where the em-

phasis is on percentage of people who exhibit certain

behaviors (such as accepting a food). Thus the major-

ity of market researchers working with purchase

intent scales use percent measures (percent rating

the product as probably or definitely buy), with the

appropriate statistical analysis being that used for

percentages. In some cases (especially with liking

ratings), the researchers will use means and standard

deviations, so the appropriate statistics there are the

ones used in parametric analyses (e.g., analysis of

variance, regression).

See also: Dietary Surveys: Surveys of Food Intakes in

Groups and Individuals; Food Acceptability: Market

Research Methods; Nutritional Assessment: Importance

of Measuring Nutritional Status; Satiety and Appetite:

The Role of Satiety in Nutrition; Sensory Evaluation:

Sensory Characteristics of Human Foods; Food

Acceptability and Sensory Evaluation; Practical

Considerations; Sensory Difference Testing; Wines: Wine

Tasting

2580 FOOD ACCEPTABILITY/Market Research Methods

Further Reading

Amerine MA, Pangborn RM, and Roessler EB (1965)

Principles of Sensory Evaluation of Food. New York:

Academic Press.

Balintfy J, Duffy WJ and Sinha P (1974) Modeling food

preferences over time. Operations Research 22: 711–727.

Chakrapani C (ed.) (2000) Marketing Research: State-

of-the-Art Perspectives. Chicago, IL: American Market-

ing Association.

Graf E and Saguy IS (1991) Food Product Development.

New York: Van Nostrand Reinhold/AVI.

Meiselman HL (1978) Scales for measuring food prefer-

ence. In: Petersen MS and Johnson AH (eds) Encyclo-

pedia of Food Science, pp. 675–678. Westport: Avi.

Meiselman HL (ed.) (2000) Dimensions of the Meal: The

Science, Culture, Business and Art of Eating. Gaithers-

burg, MD: Aspen.

Moskowitz HR (1983) Product Testing and Sensory Evalu-

ation of Foods: Marketing and R&D Approaches. West-

port, CT: Food and Nutrition Press.

Moskowitz HR (1985) New Directions for Product Testing

and Sensory Analysis of Foods. Trumbull, CT: Food and

Nutrition Press.

Moskowitz HR (1994) Food Concepts and Products: Just-

In-Time Development. Trumbull, CT: Food and Nutri-

tion Press.

Smithies RA and Buchanan BS (1990) Substantiating a taste

claim. In: Proceedings of the NAD Workshop. New

York: Council of Better Business Bureaus, and the

Advertising Research Foundation.

Thomson DMH (ed.) (1989) Food Acceptability. Barking,

Essex: Elsevier Applied Science.

Van Trijp JCM (1995) Variety-Seeking in Product Choice

Behavior – Theory with Applications in the Food

Domain, Wageningen, the Netherlands: Mansholt

Studies.

Wu L (ed.) (1989) Product Testing with Consumers for

Research Guidance. Philadelphia: American Society for

Testing and Materials.

FOOD ADDITIVES

Safety

S A Jordan, Food Directorate, Ottawa, Ontario, Canada

Background

0001 The relationship between humans and food additives

spans thousands of years. The initial use of materials

added to foods was likely for the practice of food

preservation. Food preservation in ancient times

took a variety of forms, from purely physical pro-

cesses such as drying to the addition of substances

such as smoke, salt, vinegar, or alcohol. Archeological

and written evidence of food preservation reveals the

use of these substances as far back as 5000 years.

While the preservation of food was a critical activity

for ancient populations, it was not the only reason for

adding substances to food. The ancient origins of

spices as flavoring agents is well known; however, it

is known that naturally derived materials such as

saffron and certain berry juices were used by the

Egyptians and Romans to color foods. Today, how-

ever, synthetic additives have largely supplanted their

natural counterparts in producing desirable technical

effects.

0002 Generally, food additives are defined as substances

added to food to produce a specified and desired tech-

nical effect. In some jurisdictions, the definition is

somewhat broader and may include ‘indirect food

additives’ such as packaging materials, which may

be incorporated into food unintentionally. Taking the

USA as an example, at the present time, there are about

3000 food additives approved for use, while Canada,

for example, regulates only about 400. While there

are obviously international differences as to what is

considered a food additive, this summary deals with

premarket safety evaluation procedures that may

apply to chemicals that can be incorporated into

food either by direct or indirect addition. Examples

of direct food additives are provided in Table 1.

Benefits of Direct Food Additives

0003The benefits of food additives are many. An obvious

benefit of using food additives is increased safety of

products. The use of antimicrobial agents to prevent

proliferation of bacteria or molds that may result in

food poisoning contributes significantly to providing

tbl0001Table 1 Examples of food additive categories

Anticaking agents Emulsifiers Preservatives

Antioxidants Enzymes Processing aids

Antimicrobial agents Firming agents Propellants

Colors Formulation aids Stabilizers

Curing agents Fumigants Surface finishing

agents

Drying agents pH control agents Nonnutritive

sweeteners

FOOD ADDITIVES/Safety 2581

a safe food supply. Other ultimate benefits of food

additives include a greater choice of foods in the mar-

ketplace, the availability of lower-priced foods and

enhanced organoleptic properties of food products.

The Acceptable Daily Intake

0004 While food additives provide numerous benefits for

consumers and manufacturers, there are potential

risks associated with the addition of chemicals which

may have widespread use. Foods generally are con-

sidered to be safe when consumed in an ad libitum

fashion for long periods of time, and the addition of

substances to food must not cause undue risk when

consumed in this fashion. In order to minimize any

potential adverse effects of a new food additive, a

thorough safety assessment is required. This type of

assessment is aimed at determining a safe level of

intake of material. This level is termed an acceptable

daily intake (ADI), a concept first proposed by the

Joint FAO/WHO Expert Committee on Food Addi-

tives (JECFA) in 1961. The ADI is the level of intake

by a human of a food additive, per day, for which no

adverse effects are expected when consumed over an

entire lifetime. The acceptable daily intake for food

additives usually is expressed as milligrams of the

substance ingested per kilogram of body weight.

0005 The concept of the ADI is based on the premise

that, for most compounds, there exists a discrete

threshold of exposure above which adverse effects

may be produced; however, the concept of thresholds

may not hold for substances that are able to produce

cancer. Substances that are known or suspected of

being carcinogenic to humans (especially genotoxic

carcinogens) are not allowed as food additives.

0006 Toxicity studies on specific substances, conducted

as part of an industry submission of data to govern-

ment agencies, or literature searches of earlier infor-

mation, are geared to identify levels of these

substances that produce no adverse effects (or, indeed,

no effects at all). Acronyms used to identify such

levels are NOEL (no observed effect level), NOAEL

(no observed adverse effect level), LOEL (lowest ob-

served effect level), and LOAEL (lowest observed

adverse effect level). Any measured level of effect

from experimental studies is associated with uncer-

tainty in terms of how it can be applied to possible

responses in ‘real-life’ situations. An ADI, applicable

to humans, is calculated by dividing the NO(A)EL or

LO(A)EL, usually derived from animal studies, by

numerical factors in order to reduce the uncertainties

associated with the available scientific data.

0007 Types of uncertainties that typically result from the

evaluation of experimental data are the extrapolation

of data derived from animals studies to humans, the

susceptibility of human subpopulations that may be

overly sensitive to the compound under study, the use

of data from which only a LOEL is available, the

extrapolation of less than chronic data to chronic

exposure, and any other inadequacies or deficiencies

of the data. The uncertainty factor for a lack of

chronic data would not normally apply to current

data requirements for assessing new food additives,

since chronic studies in animals would need to be

submitted to the regulatory authority. The number

of these ‘uncertainty factors’ (also sometimes termed

‘safety factors’) that are applied, and their numerical

values, depend on the situation and the available

data, and are assessed on a case-by-case basis. Table 2

outlines various uncertainty factors currently in use

and their usual numerical value.

0008The application of uncertainty factors is somewhat

subjective; however, validation exercises have pro-

vided evidence that the value of 10 is adequate for

less than chronic to chronic extrapolation and LOEL

to NOEL extrapolation, in most cases. Still, the use

of uncertainty factors is usually tailored to suit the

particular strengths and weaknesses of the database

being assessed. Once an ADI has been established, it

should not be considered an absolute threshold. Peri-

odic consumption of an additive at levels above the

ADI may be without appreciable risk, depending on

the circumstances. As the level of exposure increases

above the ADI, however, the risk of adverse effects

increases as well. Owing to the subjective nature of

the uncertainty factor approach in the derivation

of ADIs, other mathematical approaches have been

proposed.

0009The criticisms levelled at the uncertainty factor

approach include the fact that the NOEL in scientific

studies is simply one of the dose levels and may not

reflect the true NOEL for the species in question.

Also, the size of the study affects the size of the

tbl0002Table 2 Typical uncertainty factors and their values

Uncertainty Uncertainty

factor value

Extrapolation from the average animal to the

average human

Protection of sensitive populations of humans

(intraspecies variation in susceptibility)

No NOEL, only a LO(A)EL identified from scientific

studies

Extrapolation of data from less than chronic

exposure to chronic or lifetime intakes (however,

data on chronic exposure would be required for

current food additive submissions)

Inadequacies of the data (e.g., certain types of

studies missing, severity of effects associated

with a LOAEL)

2–10

2582 FOOD ADDITIVES/Safety

NOEL. Studies with large numbers of animals tend to

provide a lower NOEL, and vice versa. This, in fact,

penalizes the company for conducting a larger study,

since a lower NOEL will be associated with a lower

ADI. Another criticism is that the NOEL-uncertainty

factor approach ignores any dose–response curve gen-

erated by multiple-dose studies. Included in the alter-

native approaches to the NOEL-uncertainty factor

approach is the concept of benchmark dose, defined

as the lower confidence limit on a dose producing

some predetermined small increase (e.g., 1% or

10%) in response rate over the background level for

a specific effect. Other variations on this method have

been proposed; however, all rely to some degree on

the use of uncertainty factors (e.g., for interspecies

extrapolation). Advantages of alternative methods

are that they take into consideration the dose–re-

sponse and the variability of data, and in some cases

allow the assessment of risk at exposures below the

ADI. The disadvantages of these alternative methods

are that they do not entirely avoid the use of arbitrary

uncertainty factors and that most toxicological stud-

ies do not provide sufficient information to adequately

characterize the dose–response curve. Based on these

concerns, many organizations rely much more on the

traditional NOEL-uncertainty factor approach in cal-

culating appropriate ADIs for food additives.

0010 Along with the assessment of an ADI for a food

additive, another important step is the determination

of how much of the additive is, or is likely to be,

consumed in the general population. If the estimated

daily intake is sufficiently above the ADI, concerns

may be raised regarding the use of that additive in

general commerce. Estimated daily intakes are calcu-

lated knowing in what foods the additive will be used

and the amounts of those food products consumed by

different populations of consumers. The use of pro-

posed food additives can be limited to specific food

categories to limit the consumption, if this is seen as

necessary by the regulatory authority.

Toxicological Assessment of Food

Additives

0011 The toxicological assessment of proposed food addi-

tives encompasses a wide variety of tests ranging from

relatively simple in vitro studies to complex multi-

generation animal studies. The tests required for

each proposed additive will vary depending on the

nature of the additive itself. For proposed additives

that are the same as either naturally and commonly

consumed components of food or normal constitu-

ents of the human body (e.g., lecithin), the testing

requirements may be much less stringent compared

with additives which are novel synthetic compounds.

Also, some regulatory authorities have variable

minimal data requirements, depending on the

expected toxicity of the additive and its anticipated

consumer exposure. In this ‘Level of Concern’ system,

increasing anticipated exposure triggers additional

data requirements. As the potential toxicity of the

product increases (based on structural characteris-

tics), the exposure-based cut-off points for increased

data requirements are lower. The potential toxicity is

determined by comparing the structure of the pro-

posed additive with those of known biological activ-

ity. Substances of low concern would include normal

cellular constituents such as carbohydrates and fats,

aliphatic hydrocarbons which are saturated and

contain few functional groups, and inorganic chem-

icals, which are normal cellular constituents or inert

gases. Those substances of high concern include aro-

matic hydrocarbons and halogenated and heterocyc-

lic chemicals.

0012In all cases, the conduct of toxicity studies, and the

interpretation of data derived from these studies, may

reveal the need for further testing, which may not

have been apparent at the outset. In some cases, the

regulatory authority may see the need for further

testing in order to address specific concerns. Since

food additives generally are meant to be consumed

in a long-term, ad libitum manner, the studies that

play the most important role in assessing the safety of

proposed food additives are those aimed at determin-

ing potential chronic and/or carcinogenic, along with

reproductive effects. Other toxicological tests con-

tribute to the overall understanding of the possible

adverse effects of a substance on biological systems.

Adequate dose levels in studies allow the definition of

a dose–response and the assignment of a NOEL to

the compound under study. The high dose employed

in these studies should elicit some adverse effects.

With any type of toxicological testing, there must

be adequate methodologies employed to insure the

resulting data are meaningful and to reduce any po-

tential criticisms, which may put the results in dis-

pute. Even so, animal methodologies do not always

allow extrapolation to human exposure scenarios.

For instance, the potential food additive exposure of

human newborns and infants through artificial milk

substitutes is not mimicked in rodent models where

newborns are nourished only via maternal milk. Sev-

eral regulatory authorities have issued guidelines for

the appropriate conduct of toxicity tests. Both the

Organisation for Economic Cooperation and Devel-

opment (OECD) and the US Food and Drug Adminis-

tration (FDA) have published documents in this area

with the FDA’s detailed Redbook II currently under-

going revision. The OECD’s Guidelines for the

Testing of Chemicals are comprehensive in their

FOOD ADDITIVES/Safety 2583

coverage of a wide variety of testing methodologies

and are updated as needed. The OECD guidelines

have been endorsed by member countries, allowing

the acceptance of data between all countries belonging

to the OECD. In addition to the aforementioned guide-

lines, the concept of Good Laboratory Practice is im-

portant in assuring that testing laboratories are able to

generate valid data. Some regulatory bodies publish

principles of Good Laboratory Practice.

Types of Toxicity Tests

0013 Short-term tests for mutagenicity/genotoxicity The

potential of a test chemical to interact with or damage

DNA (genotoxicity) and have carcinogenic activity

can be assessed using a variety of tests, following

exposure in vitro or in vivo, with cells obtained from

microbial, mammalian, and human sources. Usually, a

battery of short-term tests for food additives will in-

clude a test for microbial gene mutations, chromo-

somal aberrations in mammalian cells in vitro, and

chromosomal aberrations in mammalian cells in vivo.

0014 In vitro tests include the bacterial Salmonella

typhimurium reverse mutation assay (Ames test) to

detect base-pair substitutions and frameshift muta-

tions. Mammalian cell lines include those from

mouse lymphoma cells, various tissues from Chinese

hamsters, and from human cells to measure chromo-

somal aberrations and gene mutations. Other in vitro

tests sometimes conducted include DNA repair (un-

scheduled DNA synthesis) or sister chromatid ex-

changes and cell transformation. In vivo tests

include those to detect chromosomal aberrations (in-

cluding micronucleus formation) in either circulating

blood cells or bone marrow cells, and DNA

repair assays.

0015 In vitro tests require an exogenous metabolic acti-

vation system that may not reflect metabolism in live

animals. These tests tend to be overly sensitive and

therefore may overpredict activity. However, they are

generally based on well-established, easy-to-perform

assays that are reproducible and can incorporate high

doses of test substance. In vivo genotoxicity assays

have the advantage of having an intact metabolic

system and the ability to assess concurrently target

organ distribution of the test substance and its metab-

olites. Also, in concert with other animal toxicity

tests, the results of these types of assays are more

readily applicable for extrapolation to humans. How-

ever, in vivo tests may be insensitive, requiring the

expense of large numbers of animals and, frequently,

large amounts of test material. Agents that cause

cancer by nongenotoxic mechanisms will not be

detected by genotoxicity tests, and these assays do

not replace bioassays that are aimed at detecting

cancer.

0016Toxicokinetics/pharmacokinetics Results from

these studies are useful in the interpretation of data

derived from other investigations and provide infor-

mation on the absorption, distribution, metabolism,

and excretion of the test substance. This information

also is useful in the development of further studies.

For example, if it is known that a substance passes

through the gut unabsorbed, the design of other

studies can be tailored accordingly.

0017Acute toxicity tests Generally, acute oral toxicity

tests are used where no information on a particular

substance is available. The data derived from these

single, high-dose, studies are limited, but provide in-

formation on maximum tolerated doses and may give

information on the mechanism of action of the com-

pound and the organ systems that may be affected.

0018Short-term, repeated-dose, toxicity tests Usually

conducted for 28 (or sometimes 14) days, these

short-term, oral, multiple-dose studies are designed

to gather preliminary information on the effects of the

daily administration of test substances on such par-

ameters as hematology, clinical chemistry, and path-

ology. Also, the data derived can be used to narrow

the focus of future studies to any identified target

organs and to determine appropriate doses for fur-

ther, longer-term, studies. Short-term tests generally

are conducted on a rodent (mouse or rat) and a non-

rodent species (usually dogs).

0019Subchronic toxicity tests Subchronic studies are

usually conducted for 90 days and generally involve

parameters and testing methodologies similar to

shorter-term studies. Again, both rodent and non-

rodent species are employed in subchronic studies.

These investigations will reveal significantly more

information on physiological and pathological effects

of the test chemical than will shorter-term studies.

Focused studies on other parameters such as immu-

notoxicology and neurotoxicology are included in

subchronic testing strategies.

0020Chronic and carcinogenicity studies These two

types of studies may be conducted separately; how-

ever, if a sufficient number of animals are used, the

chronic and carcinogenicity aspects may be combined

in a single study. The length of the studies is to include

the majority of the lifespan of the animal; however,

animals in the chronic exposure arm of the study may

be removed earlier. Generally, two rodent species are

employed (usually rats and mice), and the study is

conducted for 2 years with daily dosing of multiple

dose levels. In some cases, animals are also exposed

during gestation. These long-term studies are the

2584 FOOD ADDITIVES/Safety

most comprehensive, with a considerable variety of

parameters and toxic endpoints being examined and

the requirement of high standards of analysis, includ-

ing pathology and statistics. NOELs derived from

chronic studies are important, since the exposure

scenario can be extrapolated to the real-life, long-

term, human situation with food additives. Important

goals of all long-term studies are to develop dose–

response information and to identify clear LOELs

and NOELs that can be used for ADI calculations.

(See Carcinogens: Carcinogenicity Tests.)

0021 Reproduction, teratology, and developmental tox-

icity studies The aim of these investigations is to

determine the potential effects of the test substance

on the ability of animals of both sexes to produce

viable and healthy offspring and on the subsequent

development of these progeny. Most studies in this

regard involve continuous dosing through two gener-

ations of test animals (rats or mice). Males are dosed

in advance of mating to elucidate any effects of the

test compound on one complete spermatogenic cycle.

Females should be dosed for at least two estrous

periods before attempted mating. In this way, the

first-generation offspring are exposed from the

moment of conception to the weaning of their own

(second-generation) offspring. Reproductive param-

eters measured include gonadal function, mating be-

havior, time to conception, fetal viability, parturition,

and lactation. Developmental parameters include

morbidity and mortality, growth, and behavior. As

part of the reproductive studies, or investigated in

its own right, the potential of the test material to

produce physical or functional birth defects must be

determined. These teratology studies are conducted

by exposing the fetuses during in utero development

for the time that corresponds to the development of

various organ systems. Just before expected partur-

ition, the fetuses are examined for skeletal and

visceral abnormalities.

0022 Other toxicity tests The need for further studies may

become apparent during other investigations. These

include focused studies on the potential toxicity of the

substance on the immune or nervous systems.

0023 Human studies After significant evaluation in

animals, human testing is sometimes conducted.

Studies in humans usually are proposed in the case

of food additives that are expected to have a wide use

by a large proportion of consumers (e.g., sugar or fat

substitutes). While toxicological studies obviously are

not conducted in people, these human studies provide

information on possible unwanted effects (such as the

gastrointestinal tolerability of the additive) resulting

from more realistic doses and use patterns than can be

achieved with animal studies.

Interpretation of Toxicological Data

0024The data derived from the aforementioned studies

should provide sufficient information to enable the

calculation of an ADI for the particular proposed

food additive. It is important, however, to keep in

mind that it is not unusual for different professional

opinions to exist on the significance of toxicity of

a substance based on the same data. Reasons for

divided opinion are varied and may include differing

views on the potential severity of effects noted and on

the extrapolation of animal data to the human situ-

ation. In some cases, debate has occurred as to the

meaning of results from animal studies where effects

noted (including tumor development) are present in

biological structures that have no counterpart in

humans or where the metabolism of the compound

differs between animals and humans. Well-known

examples of food additive substances that have pro-

duced significant debate are saccharin and butylated

hydroxyanisole (BHA).

0025Saccharin, a sugar substitute that was generally

regarded as safe in the USA until 1972, has been

linked to bladder cancer in rats when exposed

in utero or from birth, with males affected more than

females. A NOAEL of 3% saccharin in the diet was

established in a large two-generation study in rats,

and no tumors were noted with 1% saccharin in the

diet. The factors contributing to the development of

bladder neoplasms by saccharin are species-specific.

While saccharin is nongenotoxic, it has been sug-

gested that its ability to cause bladder cancer results

from the complexing of saccharin ions with urinary

proteins with precipitation and the formation of crys-

tals. The crystals may act to damage the walls of the

bladder with increased cell proliferation. The protein

suspected of being involved in the crystal formation

(a-2-microglobulin) is present in higher concentra-

tions in the urine of male rats compared with female

rats, mice, and humans. This and other species differ-

ences in urine composition suggest that humans may

not be susceptible to the carcinogenic action of sac-

charin. Epidemiologic studies in human populations

using saccharin have not revealed any increase in

bladder cancer.

0025BHA is a phenolic antioxidant used mainly in fats

and oils. It has been shown to have a low acute

toxicity. In chronic exposure investigations, no con-

sistent adverse effects have been noted with up to

several per cent BHA in the diet, with the exception

of forestomach hyperplasia and tumor formation.

Also, it has been shown to have a low potential of

being genotoxic. Malignant tumors of the forestomach

FOOD ADDITIVES/Safety 2585

have been noted in rats fed 2% BHA in the diet for 2

years, which equals almost 1 g of BHA per kilogram

of body weight. No such tumors were noted when the

rats were fed 0.5% BHA in their diet. In other rodent

species with forestomachs, BHA has produced hyper-

plasia, and benign tumors have been noted in the

hamster forestomach. No species other than the ham-

ster and rat have been shown to develop carcinomas

in this organ. The carcinogenic effect in certain

rodent species appears to be due to indirect, rather

than direct, effects of BHA on DNA. Since the forest-

omach is an organ that is not present in humans, the

relevance of the findings in animals is questionable. In

addition, the effects noted in the positive animal

models were observed at very high doses, and a

threshold for the formation of carcinomas in rodents

was identified.

0026 Because of new developments in the scientific lit-

erature or in the interpretation of toxicological data,

it is essential that regulators keep up to date with the

toxicological literature. The generation of substantial

new evidence may allow a refinement or reassessment

of the safety of specific food additives.

Emerging Issues

0027 The continuing evolution of the food industry means

that new types of products come into use on a con-

tinuing basis. Novel types of food additives and pack-

aging methods and materials arise that present unique

challenges to the toxicologist and regulatory author-

ities in terms of appropriate testing strategies.

0028 In recent years, certain plants and microorganisms

have been modified through biotechnology, including

genetic modification. Food additives, such as enzymes,

derived from genetically modified organisms (or

the addition of the modified organisms themselves)

present challenges in safety assessment. Issues to con-

sider in the assessment of these types of food additives

include the exposure to foreign DNA and proteins

(and thus considerations of allergenicity), increased

exposure to these types of products as the market

increases, and the necessity of microbiological purity

if products are derived from modified microorgan-

isms. (See Biotechnology in Food Production; Genet-

ically Modified Foods.)

0029 The advent of ‘functional foods’ (foods containing

ingredients demonstrated to have specific health

benefits, in addition to providing basic nutrients and

nutritional benefits) has led regulatory authorities not

only to develop ways of regulating health claims as-

sociated with these products but also to consider the

inherent safety of these products. Globalization of

markets can result in new types of food and food

ingredients being introduced. The addition of

medicinal-type herbs, such as Echinacea species or

St. John’s Wort (Hypericum perforatum) to conven-

tional foods, for perceived health benefits, necessi-

tates an assessment of the safety of these new food

products, since herbal ingredients may be consumed

in ways that are very different from their use in

traditional systems of medicine.

0030That young children are different from adults in

many aspects of their physiology is well known, and

this must be taken into account in the assessment of

food additives that may be consumed by this popula-

tion. In recent years, the applicability of the ADI-

uncertainty factor approach to insuring the safety of

food additives consumed by children has come under

scrutiny, and this subject is one of ongoing discussion

in many countries.

See also: Biotechnology in Food Production;

Carcinogens: Carcinogenicity Tests; Genetically

Modified Foods