IPCS INCHEM Home



    INTERNATIONAL PROGRAMME ON CHEMICAL SAFETY


    ENVIRONMENTAL HEALTH CRITERIA 70






    PRINCIPLES FOR THE SAFETY ASSESSMENT OF
    FOOD ADDITIVES AND CONTAMINANTS IN FOOD








    This report contains the collective views of an international group of
    experts and does not necessarily represent the decisions or the stated
    policy of the United Nations Environment Programme, the International
    Labour Organisation, the World Health Organization, or the Food
    Agriculture Organization of the United Nations

    Published under the joint sponsorship of
    the United Nations Environment Programme,
    the International Labour Organisation,
    and the World Health Organization in
    collaboration with the Food and Agriculture
    Organization of the United Nations

    World Health Orgnization
    Geneva, 1987


         The International Programme on Chemical Safety (IPCS) is a
    joint venture of the United Nations Environment Programme, the
    International Labour Organisation, and the World Health
    Organization. The main objective of the IPCS is to carry out and
    disseminate evaluations of the effects of chemicals on human health
    and the quality of the environment. Supporting activities include
    the development of epidemiological, experimental laboratory, and
    risk-assessment methods that could produce internationally
    comparable results, and the development of manpower in the field of
    toxicology. Other activities carried out by the IPCS include the
    development of know-how for coping with chemical accidents,
    coordination of laboratory testing and epidemiological studies, and
    promotion of research on the mechanisms of the biological action of
    chemicals.

        ISBN 92 4 154270 5 

         The World Health Organization welcomes requests for permission
    to reproduce or translate its publications, in part or in full.
    Applications and enquiries should be addressed to the Office of
    Publications, World Health Organization, Geneva, Switzerland, which
    will be glad to provide the latest information on any changes made
    to the text, plans for new editions, and reprints and translations
    already available.

    (c) World Health Organization 1987

         Publications of the World Health Organization enjoy copyright
    protection in accordance with the provisions of Protocol 2 of the
    Universal Copyright Convention. All rights reserved.

         The designations employed and the presentation of the material
    in this publication do not imply the expression of any opinion
    whatsoever on the part of the Secretariat of the World Health
    Organization concerning the legal status of any country, territory,
    city or area or of its authorities, or concerning the delimitation
    of its frontiers or boundaries.

         The mention of specific companies or of certain manufacturers'
    products does not imply that they are endorsed or recommended by the
    World Health Organization in preference to others of a similar
    nature that are not mentioned. Errors and omissions excepted, the
    names of proprietary products are distinguished by initial capital
    letters.



CONTENTS

PRINCIPLES FOR THE SAFETY ASSESSMENT OF FOOD ADDITIVES AND CONTAMINANTS 
IN FOOD

FOREWORD

PREFACE

 1. INTRODUCTION

 2. HISTORICAL BACKGROUND

    2.1. Introduction
    2.2. Periodic review
        2.2.1. Concept of periodic review
        2.2.2. Mechanism of periodic review

 3. CRITERIA FOR TESTING AND EVALUATION

    3.1. Criteria for testing requirements
        3.1.1. Estimating exposure
        3.1.2. Predicting toxicity from chemical structure
        3.1.3. Other factors to consider when developing criteria
    3.2. Priorities for testing and evaluation
    3.3. Quality of data

 4. CHEMICAL COMPOSITION AND THE DEVELOPMENT OF SPECIFICATIONS

    4.1. Identity and purity
    4.2. Reactions and fate of food additives and contaminants in food
    4.3. Specifications

 5. TEST PROCEDURES AND EVALUATION

    5.1. End-points in experimental toxicity studies
        5.1.1. Effects with functional manifestations0
        5.1.2. Non-neoplastic lesions with morphological manifestations
        5.1.3. Neoplasms
        5.1.4. Reproduction/developmental toxicity
        5.1.5.  In vitro  studies
    5.2. The use of metabolic and pharmacokinetic studies in safety 
        assessment
        5.2.1. Identifying relevant animal species.
        5.2.2. Determining the mechanisms of toxicity
        5.2.3. Metabolism into normal body constituents
        5.2.4. Influence of the gut microflora in safety assessment
                5.2.4.1 Effects of the gut microflora on the chemical
                5.2.4.2 Effects of the chemical on the gut microflora
    5.3. Influence of age, nutritional status, and health status on the
        design and interpretation of studies
        5.3.1. Age
                5.3.1.1 History
                5.3.1.2 Usefulness of studies involving  in utero exposure
                5.3.1.3 Complications of aging
        5.3.2. Nutritional status

        5.3.3. Health status
        5.3.4. Study design
    5.4. Use of human studies in safety evaluation
        5.4.1. Epidemiological studies
        5.4.2. Food intolerance
    5.5. Setting the ADI
        5.5.1. Determination of the no-observed-effect level
        5.5.2. Use of the safety factor
        5.5.3. Toxicological versus physiological responses
        5.5.4. Group ADIs
        5.5.5. Special situations
        5.5.6. Comparing the ADI with potential exposure

6. PRINCIPLES RELATED TO SPECIFIC GROUPS OF SUBSTANCES

    6.1. Substances consumed in small amounts
        6.1.1. Food contaminants
        6.1.2. Food flavouring agents
    6.2. Substances consumed in large amounts
        6.2.1. Chemical composition, specifications, and impurities
        6.2.2. Nutritional studies
        6.2.3. Toxicity studies
        6.2.4. Foods from novel sources

REFERENCES

ANNEX I.    GLOSSARY

ANNEX II.   STATISTICAL ASPECTS OF TOXICITY STUDIES

REFERENCES TO ANNEX II

ANNEX III.  GUIDELINES FOR THE EVALUATION OF VARIOUS GROUPS OF FOOD 
            ADDITIVES AND CONTAMINANTS

REFERENCES TO ANNEX III

ANNEX IV.   EXAMPLES OF THE USE OF METABOLIC STUDIES IN THE SAFETY 
            ASSESSMENT OF FOOD ADDITIVES

REFERENCES TO ANNEX IV

ANNEX V.    APPROXIMATE RELATION OF PARTS PER MILLION IN DIET TO MG/KG PER
            DAY

ANNEX VI.   REPORTS AND OTHER DOCUMENTS RESULTING FROM MEETINGS OF
            THE JOINT FAO/WHO EXPERT COMMITTEE ON FOOD ADDITIVES


WHO TASK GROUP ON UPDATING THE PRINCIPLES FOR THE SAFETY ASSESSMENT OF
FOOD ADDITIVES AND CONTAMINANTS IN FOOD

1985 and/or 1986 JECFA Members

a,b,c,d,e Dr H. Blumenthal, Division of Toxicology, Center for
             Food Safety and Applied Nutrition, US Food and
             Drug Administration, Washington DC, USA
c,d       Dr I. Chakravarty, Department of Biochemistry and
             Nutrition, All India Institute of Hygiene and
             Public Health, Calcutta, India
c         Dr W.H.B. Denner, Food Composition and Information
             Unit, Food Sciences Division, Ministry of Agri-
             culture, Fisheries and Food, London, United
             Kingdom
c         Dr A.H. El-Sebae, Pesticides Division, Faculty of
             Agriculture, Alexandria University, Alexandria,
             Egypt
c         Professor P.E. Fournier, Hpital Fernand-Widal,
             Paris, France
a,c,d,f   Dr S. Gunner, Food Directorate, Health Protection
             Branch, Health and Welfare Canada, Ottawa,
             Ontario, Canada
d         Mr J. Howlett, Food Science Division, Ministry of
             Agriculture, Fisheries and Food, London, United
             Kingdom
c,d,f     Professor K. Kojima, College of Environmental Health,
             Azabu University, Sagamihara-Shi, Kanagawa-Ken,
             Japan
c         Dr W. Kroenert, Food Chemistry Division, Max von
             Pettenkofer Institute, Federal Office of Public
             Health, Berlin (West)
a,b,c,d,f Dr B. MacGibbon, Division of Toxicology, Environ-
             mental Pollution and Prevention, Department of
             Health   and   Social  Security,   London,  United
             Kingdom
c,d       Dr R. Mathews, Food Chemicals Codex, National
             Academy of Sciences, Washington DC, USA
d         Mrs I. Meyland, National Food Institute, Ministry of
             the Environment, Soborg, Denmark
c,d,e     Dr J. Modderman, Food Additives Chemistry Evaluation
             Branch, Center for Food Safety and Applied
             Nutrition, US Food and Drug Administration,
             Washington DC, USA
d         Dr G. Nazario, Ministry of Health, National Health
             Council, Rio de Janeiro, Brazil
c,d       Professor K.A. Odusote, College of Medicine, Univ-
             ersity of Lagos, Lagos, Nigeria
c,d       Professor F. Pellerin, Facult de Pharmacie de
             l'Universit Paris-Sud, Hpital Corentin Celton,
             Issy-les-Moulineaux, France
c,f       Dr P. Pothisiri, Food Control Division, Food and
             Drug Administration, Ministry of Public Health,
             Bangkok, Thailand

b,c,d     Professor M.J. Rand, Department of Pharmacology,
             University of Melbourne, Parkville, Victoria,
             Australia
a,b,d,e,g Dr P. Shubik, Green College, Oxford, United Kingdom
d         Dr A. Slorach, Food Research Department, The National
             Food Administration, Uppsala, Sweden
d         Dr V.A. Tutelyn, Institute of Nutrition, Academy of
             Medical Sciences of the USSR, Moscow, USSR

Secretariat

d         Dr Y.K. Al-Mutawa, Division of Public Health Labor-
             atory, Ministry of Public Health, Safat, Kuwait
a         Dr E.A. Bababunmi, University of Ibadan, Ibadan,
             Nigeria
e         Dr A. Br, Department of Vitamin and Nutrition
             Research, F. Hoffmann-LaRoche and Company, Ltd.,
             Basel, Switzerland
a,b,d,f   Dr J. Cabral, Unit of Mechanisms of Carcinogenesis,
             International Agency for Research on Cancer,
             Lyons, France
e         Dr J. Caldwell, St. Mary's Hospital Medical School,
             London, United Kingdom
a,e       Dr D.M. Conning, British Nutrition Council, London,
             United Kingdom
f         Dr J.L. Emerson, External Technical Affairs, The
             Coca-Cola Company, Atlanta, Georgia, USA
d         Mr A. Feberwee, Committee on Food Additives, Nutri-
             tion and Quality Affairs, Ministry of Agriculture
             and Fisheries, The Hague, The Netherlands
d         Professor C.L. Galli, Toxicology Laboratory,
             Institute of Pharmacology and Pharmacognosy,
             University of Milan, Milan, Italy
e         Dr M.J. Goldblatt, Consumer Nutrition Affairs,
             General Foods Corporation, White Plains,New
             York, USA
f         Dr W. Grunow, Divison of Food Toxicology, Max von
             Pettenkofer Institute, Federal Office of Public
             Health, Berlin (West)
d         Mr R. Haigh, Commission of the European Communities,
             Brussels, Belgium
a,d,f     Dr Y. Hayashi, Division of Pathology, Biological
             Safety Research Center, National Institute of
             Hygienic Sciences, Tokyo, Japan
b,d,e,g   Dr J. Herrman, Division of Food and Color Additives,
             Center for Food Safety and Applied Nutrition,
             US Food and Drug Administration, Washington DC,
             USA
f         Dr D. Krewski, Environmental Health Directorate,
             Health Protection Branch, Health and Welfare
             Canada, Ottawa, Ontario, Canada
e         Mr P.N. Lee, Consultant in Statistics and Adviser in
             Epidemiology and Toxicology, Surrey, United
             Kingdom

b         Dr M. Mercier, International Programme on Chemical
             Safety, World Health Organization, Geneva,
             Switzerland
d         Dr R.W. Moch, Center for Food Safety and Applied
             Nutrition, US Food and Drug Administration,
             Washington DC, USA
a         Dr V.H. Morgenroth, Chemicals Division, Environment
             Directorate, Organization for Economic Cooper-
             ation and Development, Paris, France
a         Dr I. Nir, Department of Pharmacology and Experi-
             mental Therapeutics, The Hebrew University
             Hadassah Medical School, Jerusalem, Israel
d         Dr E. Poulsen, National Food Institute, Institute of
             Toxicology, Soborg, Denmark
b,d,f     Dr A.W. Randell, Food Policy and Nutrition Division,
             Food and Agricultural Organization of the United
             Nations, Rome, Italy
d         Dr N. Rao Maturu, Joint FAO/WHO Food Standards
             Programme, Food and Agricultural Organization of
             the United Nations, Rome, Italy
e,f       Dr A.G. Renwick, Clinical Pharmacology Group, Univ-
             ersity of Southampton, Southampton, United
                Kingdom
e,f       Dr F.J.C. Roe, Consultant in Toxicology and Adviser
             in Experimental Pathology and Cancer Research,
             London, United Kingdom
d,e       Dr S.I. Shibko, Division of Toxicology, Center for
             Food Safety and Applied Nutrition, US Food and
             Drug Administration, Washington DC, USA
a,b,e,f   Dr V. Silano, Department of Comparative Toxicology,
             High Institute of Health, Rome, Italy
d         Professor A. Somogyi, Department of Drugs, Animal
             Nutrition and Residue Research, Institute for
             Vetinary Medicine, Berlin (West)
b,d       Professor R. Truhaut, Facult des Sciences Pharma-
             ceutiques et Biologiques de Paris Luxembourg,
             Laboratoire de Toxicologie et Hygiene Indus-
             rielle, Universit Rene Descartes, Paris, France
f         Dr G.J. Van Esch, National Institute for Public
             Health and Environmental Hygiene, Bilthoven,
             The Netherlands
a,b,d     Dr G. Vettorazzi, International Programme on Chemical
             Safety, World Health Organization, Geneva,
             Switzerland

e         Dr M.J. Wade, Division of Toxicology, Center for
             Food Safety and Applied Nutrition, Washington DC,
             USA
a,b,d,e,g Dr R. Walker, Department of Biochemistry, University
             of Surrey, Guildford, Surrey, United Kingdom



-----------------------------------------------------------------
a    Present  at strategy meeting, Oxford,  United Kingdom, 19-23
     September, 1983.
b    Present   at   pre-consultation  of   contributors,  Geneva,
     Switzerland, 29-31 May, 1985.
c    Member of JECFA-85, Geneva, Switzerland, 3-12 June, 1985.
d    Participant in JECFA-86, Rome, Italy, 2-11 June, 1986.
e    Consultant who contributed written material.
f    Submitter of written comments.
g    Member of Editorial Committee.

NOTE TO READERS OF THE CRITERIA DOCUMENTS

    Every  effort has been  made to present  information in  the
criteria  documents  as  accurately as  possible  without unduly
delaying their publication.  In the interest of all users of the
environmental  health  criteria  documents, readers  are  kindly
requested  to communicate any errors  that may have occurred  to
the Manager of the International Programme on  Chemical  Safety,
World  Health Organization, Geneva,  Switzerland, in order  that
they  may  be  included in  corrigenda,  which  will  appear  in
subsequent volumes.



                          *    *    *

FOREWORD


    The WHO activities concerned with the safety  assessment  of
food   chemicals   were  incorporated   into  the  International
Programme on Chemical Safety (IPCS) in 1980.  Since this time, a
keen  interest has developed  in all aspects  pertaining to  the
toxicological  evaluation  of  food additives  and contaminants,
including the methodological aspects.  These activities are part
of  the  responsibilities  of  the  Programme  insofar  that its
objectives  include the formulation  of "guiding principles  for
exposure  limits,  such as  acceptable  daily intakes  for  food
additives  and  pesticide  residues, and  tolerances  for  toxic
substances   in  food,  air,   water,  soil,  and   the  working
environment".

    The  present  publication  on  "Principles  for  the  Safety
Assessment  of Food Additives and Contaminants in Food" has been
developed in response to repeated recommendations by  the  Joint
FAO/WHO  Expert  Committee  on  Food  Additives  (JECFA).    Its
inclusion in the methodology section of the Environmental Health
Criteria  series will make it  readily available to both  Member
States and the food industry.

    The  IPCS gratefully acknowledges  the financial support  of
the  United  Kingdom Department  of  Health and  Social Security
(DHSS), and the US Food and Drug Administration (FDA), which was
indispensable for the completion of the project.

                         Dr M. Mercier
                         Manager
                         International Programme on
                           Chemical Safety

PREFACE

    For  the  last  thirty years,  the internationally sponsored
committee  known as the Joint  FAO/WHO Expert Committee on  Food
Additives  (JECFA) has played a major role in providing a unique
international   mechanism  for  the  identification  and  safety
assessment  of  food  chemicals, including  food additives, food
contaminants,  and  residues  of  veterinary  drugs.   With   no
regulatory  aspirations, this Committee has probably contributed
more  to the elaboration  of sound national  food regulation  in
this area than any other international body aimed at harmonizing
or  normalizing  often  divergent  national  approaches  to  the
problem  of  food safety,  food  technology, and  food  control.
JECFA  achieved  this  by  providing  recommendations  based  on
scientific  evidence  and by  establishing  a rational  model of
safety assessment that is widely reputed and accepted.

    Hundreds  of  highly skilled  international specialists have
given, and continue to give, freely of their time and talents to
foster  advances  in toxicological  methodologies and analytical
procedures, to consolidate accessible presentations of data, and
to  keep  abreast  with  scientific  developments,  which  often
requires readjustment of previous conclusions.  Through reports,
toxicological    monographs,    and    profiles   of    chemical
specifications  by  JECFA, national  food regulatory authorities
and  the Codex Alimentarius Commission are provided with all the
necessary elements for making the best decisions on the rational
use of chemicals in food.

    The  present  undertaking  has  several  precedents  in  the
history  of JECFA.  For example,  in 1957, the second  report of
the  Committee  elaborated  on  procedures  for  the  testing of
intentional  food additives to  establish their safety  for  use
and,  in 1960, the fifth report contained a series of guidelines
for  the  evaluation  of  the  carcinogenic  hazards   of   food
additives.   It should  also be  mentioned that,  in  1966,  the
Committee  commissioned  a  special scientific  group to develop
procedures  for investigating intentional and unintentional food
additives.  Finally, in 1981, after realizing that a significant
interval  of  time  had elapsed  since  previous  methodological
updatings, the Committee called for a state-of-the-art review of
methodology.   A favourable answer  was received from  the newly
established International Programme on Chemical Safety (IPCS), a
cooperative  programme  sponsored  by the  International  Labour
Organisation  (ILO),  the  United Nations  Environment Programme
(UNEP), and the World Health Organization (WHO).  It  should  be
noted  that the implementation of the recommendation by the IPCS
was significantly facilitated by the fact that the toxicological
component of the JECFA came within the scope of the programme.

    The contents of this publication are the result of sustained
efforts  of  an  IPCS Task  Group  during  a number  of meetings
including:  a  strategy  meeting  in  1983,  a  consultation  of
contributors  in 1985,  and the  JECFA Working  Groups at  their
annual meetings in 1985 and 1986.  The members of the Task Group
contributed  either written material  or comments, or  both.  An

Editorial  Board was responsible  for preparing the  final draft
for publication.  The Task Group benefited widely from the large
number of recommendations and observations on the methodology of
testing and assessing chemicals in food, found in  the  previous
reports of JECFA and related scientific groups.

    Thus,    this    publication    reflects   faithfully    the
recommendations of the Committee regarding the safety assessment
of food additives and contaminants by reaffirming  the  validity
of  recommendations that are  still appropriate, while  pointing
out  the problems associated  with those that,  in the light  of
modern  advances  in  methodology,  are  no  longer  valid.  New
recommendations  are also made, as  might be expected, with  the
advancing  state  of  the toxicological  sciences.  Particularly
enlightening are the section dealing with the principles related
to  the  safety  assessment  of  substances  consumed  in  large
amounts,  and Annexes  II and  IV dealing  with the  statistical
aspects of toxicity studies and examples of the use of metabolic
studies   in   the   safety  assessment   of   food   additives,
respectively.  An Index has also been included at the end of the
book.

    It  is  the  most earnest  wish  of  all concerned  with the
production  of this publication that it should make an important
contribution to the field of food toxicology and that it will be
found useful by the members of JECFA, national  food  regulatory
authorities, and industry, who are involved in  the  development
of  safety data and in making the consumer aware of the problems
of the safe use of food additives.

                                           Dr G. Vettorazzi


                                           Dr A. Randell

1.  INTRODUCTION

    This publication is concerned with reviewing the  basis  for
decision-making  by the Joint Expert Committee on Food Additives
(JECFA) of the Food and Agriculture Organization of  the  United
Nations  (FAO) and the World Health Organization (WHO).  Because
the toxicological and chemical characteristics of food additives
are the primary concern of the Committee, both aspects are dealt
with  in this publication, which  has been prepared by  WHO- and
FAO-appointed   consultants,  assisted  by   the  WHO  and   FAO
secretariats.   The twenty-eighth JECFA  report (1)  includes  a
summary of the areas that the Committee considered to be most in
need of evaluation.

    The major concerns of this monograph are with the testing of
chemicals in food and with the evaluation of the  test  results.
In  keeping with the approach developed during the past 30 years
by  JECFA, the recommendations  for test procedures  and  safety
assessment  are discussed in broad terms taking into account the
latest  scientific advances in  the relevant fields.   No effort
has  been  made to  provide  instructions for  test  procedures.
Differences  in national approaches to toxicological evaluations
exist,  and  some  variations in  the  data  submitted  must  be
considered  by  JECFA  in making  their  evaluations.   However,
certain basic data requirements are necessary in order to enable
an expert committee to make sound judgements.  The  elements  of
this base line are included in various sections of this report.

    In essence, the problems under consideration fall into three
general categories: first, the determination of the chemical and
toxicological  test  requirements for  individual chemicals that
are  added to or occur  in food; second, the  assessment methods
that  are  to be  applied; and third,  the updating of  the test
procedures and methods of assessment as the science progresses.

    Methods for testing and assessment have changed considerably
during  the life of  JECFA.  However, the  Committee has, by  no
means,  been  a  static  organization.   Not  only  have various
approaches  been continuously updated at the individual meetings
of  the Committee, but intervening meetings of scientific groups
have  been  held  to  consider  the  impact  of  new  scientific
developments  on the procedures used (2, 3).  Thus, every effort
has been made in this publication to record the views  of  JECFA
and to detail the changes that have come about with  the  course
of time.

    It  is recognized  that, with  advances in  science,  it  is
possible  to  obtain  more complete  toxicological  profiles for
individual  chemicals.  For example, it is becoming increasingly
easy to learn more about the disposition and metabolic  fate  of
xenobiotics.   Until  relatively  recently, the  toxicologist in
this  field had  to rely  almost entirely  on a  set of  routine
tests,  the results  of which  were then  assessed and  used  to
establish arbitrarily-determined safe levels.

    JECFA  has long recognized that  a number of factors  can be
used to determine test requirements; these include the structure
of  the  chemical, its  natural  occurrence in  foodstuffs,  its
metabolic  characteristics, and knowledge of its effects in man.
However,  systematic guidelines incorporating these factors have
not  been developed  by JECFA.   In recent  years, some  of  the
problems  posed  to  JECFA  have  concerned  the   testing   and
evaluation of additives and food ingredients that  are  consumed
in  large amounts.  Other key  problems in the determination  of
the  appropriate level of testing  involve the largest group  of
food  additives, the flavouring agents that are used, generally,
at  very  low levels  and that are  often "nature identical"  or
derived   from   natural   sources.    Therefore,   both   "high
consumption"   substances and flavouring agents are discussed in
detail in sections 6.1 and 6.2

    There  have been considerable changes  in laboratory studies
used in other areas of toxicology, which are yet to have a major
impact  on  the evaluation  of  food additives.   Of  particular
interest  are  the series  of mutagenicity/clastogenicity tests,
often, but not invariably, using sub-mammalian test organisms  in
 vitro.   The number,  diversity, and  uses of  these tests  have
increased  rapidly in the past  decade.  In general, such  tests
are  effective for measuring an intended genetic end-point.  How
effectively  these tests identify  chemical carcinogens is  much
less  clear.   In  the  absence  of  a  clear  correlation  with
carcinogenicity, it is difficult to know how such  tests  should
be  interpreted  and used  in  safety evaluations.   Even though
these  in vitro  tests may not be required for the evaluation  of
the  safety of  food additives,  it is  becoming more  and  more
frequent  for  chemicals  to be  tested  in  this way  for other
reasons, e.g., to detect potential environmental or occupational
hazards.   Thus, JECFA may  have to decide  on the relevance  of
such information (section 5.1.5).

    During the past decade, there has been a major  increase  in
the number of chemicals tested routinely for chronic toxicity in
standardized  in vivo tests.   Although  these tests  may not be
designed  to evaluate  food additives,  the results  have to  be
carefully  considered  at  JECFA  meetings.   Among  the   major
problems that occur are those arising from results obtained when
chemicals are administered to animals by routes other  than  the
diet  or  drinking-water.  For  example,  some years  ago, JECFA
faced the problem of assessing the significance of the induction
of  subcutaneous sarcomas at  the site of  injection of  certain
food chemicals into rodents.  It was found that many substances,
including inert plastics, could give rise to similar sarcomas on
implantation.   As a result,  the Committee concluded  that such
findings  could not be used in a definitive manner for assessing
the  safety of food  additives (4, pp.  16-17).  However,  these
findings  cannot be totally ignored  and may be an  indicator of
the  need  for further  carcinogenicity  studies using  the oral
route.   Another problem with interpretation arises with many of
the  more  recent  "routine"  in vivo  studies  that  record  the
enhancement  of a variety  of common "spontaneous"   tumours  in
rodents,  including lymphomas, hepatomas,  and pheochromocytomas
(section 5.1.3).

    The  scope of long-term  toxicity tests has  been  discussed
extensively  by JECFA.  For example, several food chemicals have
been  tested in 2-generation  studies rather than  the commonly-
used  single-generation  study.   While  the  use  of  this more
extensive  test is advisable under certain conditions, it should
not necessarily be a routine procedure (section 5.3).

    Many  of the chemicals of  concern to those responsible  for
food  safety evaluation are present  in food at very  low levels
and may be present as environmental contaminants or  may  result
from the migration of substances from food packaging or residues
from  the  use of  solvents,  pesticides, or  veterinary  drugs.
These situations often require very different approaches to test
requirements  than  those  used for  intentional  food additives
(Annex III).  One case, the use of anabolic agents in livestock,
has  posed various problems  for JECFA that  cannot be  answered
within  the scope of current procedures (Annex III).  JECFA will
soon be developing methodology for the testing and evaluation of
veterinary  drug  residues  in support  of  a  new committee  on
residues  of veterinary drugs in foods that has been established
by the Codex Alimentarius Commission.

    In assessing the significance of data, a major issue  to  be
resolved  concerns the distinctions  that should be  made  among
different   toxicological   manifestations.   The   carcinogenic
potential  of chemicals  has been  emphasized in  the  past  few
decades  to the exclusion of most other toxic end-points.  There
was  a general consensus that chemicals found to be carcinogenic
were  not appropriate as food additives at any level whatsoever.
More recently, however, it has become widely accepted  that  the
term  "carcinogen"  has  become  harder  and  harder  to  define
(section 5.1).  It is apparent that cancer can be induced  by  a
variety  of chemicals acting  by very different  mechanisms  and
that  the mechanism should  be an important  consideration  when
determining  whether  a safe  level  can be  established.  Other
questions  concern whether high-dose animal data are relevant to
human exposure to low levels, and how teratogenicity data in the
absence  of  maternal toxicity  are  to be  interpreted (section
5.1.4).

2.  HISTORICAL BACKGROUND

2.1.  Introduction

    The Joint FAO/WHO Expert Committee on Food Additives (JECFA)
was established following recommendations made to the Directors-
General  of FAO and WHO by the Joint FAO/WHO Expert Committee on
Nutrition  at it fourth  session (5), and  the subsequent  first
Joint FAO/WHO Conference on Food Additives was held  in  Septem-
ber, 1955 (6).  The terms of reference of the  earlier  meetings
of JECFA related to the formulation of general  principles  gov-
erning the use of food additives and consideration  of  suitable
uniform  methods for evaluating  the safety of  food  additives.
For  these purposes, food  additives were defined  by the  Joint
Conference  as "non-nutritive substances added  intentionally to
food,  generally in small quantities, to improve its appearance,
flavour,  texture,  or storage  properties."a   Following recom-
mendations of the third Joint FAO/WHO Conference on  Food  Addi-
tives  (8), these terms of  reference were broadened to  include
substances  unintentionally introduced into human food and JECFA
has   subsequently  considered  and  evaluated  such  materials,
including  growth promoters, components of  packaging materials,
solvents  used in food processing,  aerosol propellants, enzymes
used in food processing, and metals in foods.  Novel  foods  and
ingredients that may be incorporated into foods at levels higher
than  those previously envisaged  for food additives  have  also
been  referred to  JECFA and  pose special  problems  in  safety
evaluation,  which  will  be  discussed  later  in  this  report
(section 6.2).

    The first (9), second (10), and fifth Reports (4)  of  JECFA
established  principles for the use  of food additives and  made
recommendations on methods for establishing the safety-in-use of
food  additives  and  for  the  evaluation  of  the carcinogenic
hazards  of food additives.   From the outset,  JECFA recognized
that:

    "no single pattern of tests could cover adequately, but
    not wastefully, the testing of substances so diverse in
    structure  and  function  as food  additives"  and that
    "the  establishment  of  a uniform  set of experimental
    procedures that would be standardized and obligatory is
    therefore undesirable" (10).

----------------------------------

a   From  a practical standpoint, the "food additive" definition
    has  been  expanded  since the  time  it  was drafted,  as a
    variety   of   compounds,  including   nutritive  substances
    consumed  in  high  amounts,  have  been  brought  under the
    umbrella  of  food  additives.   Indeed,  the  second  Joint
    FAO/WHO  Conference on Food  Additives (7) recommended  that
    the  scope of  the JECFA  programme be  expanded beyond  the
    substances included in the original definition.

Accordingly, this Committee concluded "that it was only possible
to  formulate  general  recommendations with  regard  to testing
procedures."  The Committee also recognized that advances in the
basic  sciences  might  suggest new  approaches to toxicological
investigations and that these might be used immediately  by  the
scientist  but would take longer to become incorporated into any
officially  recommended testing procedures.  Subsequent meetings
of  JECFA  have  consistently  adopted  this  approach  and have
avoided  the adoption  of fixed  protocols for  the testing  and
evaluation  of all classes of intentional and unintentional food
additives.  This has had the advantage of allowing the Committee
to  respond to new  problems as they  have arisen, with  minimal
inertia, and to encompass non-routine and  ad hoc  studies in the
safety evaluation process.  Within this framework, the Committee
has found it possible to formulate guidelines for the evaluation
of   several  groups  of  intentional   and  unintentional  food
additives  that posed their  own peculiar problems;  several  of
these  guidelines, which serve  as specific examples  to support
general principles, are contained in Annex III.

    The  requirement to keep abreast  of scientific developments
in  toxicology  and  related scientific  disciplines implies the
need  for a periodic  review of testing  methodology.  Following
recommendations to this effect made by the eighth (11) and ninth
(12) meetings of JECFA, a WHO Scientific Group on Procedures for
Investigating  Intentional and Unintentional Food  Additives was
convened in 1966

    "to review, in the light of new  scientific  knowledge,
    the  criteria  used  in establishing  acceptable  daily
    intakes.  . . ."   and "to suggest  further studies  on
    toxicological  procedures  used  for the  evaluation of
    intentional  and unintentional food additives  in order
    to establish their safety to the consumer"  (2).

Subsequent meetings of JECFA have taken cognisance of the report
of  this Scientific Group and of the report of a more recent WHO
Scientific  Group on the Assessment of Carcinogenicity and Muta-
genicity of Chemicals (3).  Some aspects of the reports of other
WHO  Scientific Groups  on the  Principles for  the Testing  and
Evaluation of Drugs for Carcinogenicity (13), Mutagenicity (14),
and Teratogenicity (15)  are also pertinent to  the  methodology
of testing food additives.  However, significant developments in
the  science  of  toxicology  and  related  disciplines  led the
seventeenth  meeting of JECFA to recommend that "the methods and
procedures  for the toxicity testing of food additives should be
comprehensively  reviewed and brought into line with advances in
toxicology  and cognate disciplines"  (16).  This recommendation
was  reiterated  in  the reports  of  the  eighteenth  (17)  and
nineteenth  (18) meetings, the  latter of which  called for  the
convening  of an  appropriate meeting  for the  purpose  of  the
review, and reaffirmed at the twentieth meeting (19).

    Safety  evaluation of food  additives is a  2-stage process.
The  first  stage  involves  the  collection  of  relevant  data
including  the results of  studies on experimental  animals and,

where  possible, observations in man.  The second stage involves
the  assessment of data  to determine the  acceptability of  the
substance  as  a  food  additive.   While  the   recommendations
referred to in the preceding paragraph emphasize the  impact  of
scientific  advances on the first  testing stage, the impact  of
such  advances extends also to  the assessment stage.  This  was
made explicit in the twenty-first report of JECFA (20,  p.  31),
which stated that:

    "in  view of the rapid progress of the science of toxi-
    cology and the increasing refinement of evaluation pro-
    cedures,  the Committee felt  strongly that the  tradi-
    tional  concepts  of  setting ADIs,  the application of
    safety  factors, and the  relationship of these  safety
    factors to the observed toxicological manifestations in
    animal experiments should be reconsidered".

This recommendation was endorsed by the twenty-fourth meeting of
JECFA (21).

    Many  features of toxicity testing and evaluation of advent-
itious  food  additives and  contaminants  that fall  within the
terms of reference of JECFA, are common to pesticides  that  are
within  the scope  of the  Joint FAO/WHO  Meeting  on  Pesticide
Residues.  In recognition of this, the twenty-fifth  meeting  of
JECFA  (22)  recommended  that a  group  of  experts  should  be
convened,  as soon  as possible,  to study  the  application  of
advances  in methodology to the toxicological evaluation of food
additives and contaminants, and also of pesticide residues.  The
urgency  of  the  need  to  implement  this  recommendation  was
stressed  by the twenty-sixth (23) and twenty-seventh (24) meet-
ings of JECFA.

    In  response to the Committee's  repeated recommendations, a
meeting  of  a  group of  experts  to  study the  application of
advances  in methodology to the toxicological evaluation of food
additives  and contaminants was convened in September 1983.  The
objectives  of the meeting were to formulate specific recommend-
ations in order to bring up to date:

    (a)  the  principles  set out  in  earlier reports  of JECFA
         concerning  safety  evaluation in  relation to specific
         toxicological problems or specific chemical entities or
         groups;

    (b)  the  test methods used in  the toxicological evaluation
         of chemicals in food; and

    (c)  the   assessment   procedures   adopted  by   JECFA  in
         determining quantitative end-points.

    The  report  of the  Working  Group (Updating  Principles of
Methodology  for Testing and Assessing Chemicals in Food: Report
of a Strategy Meeting) (unpublished WHO document ICS(Food)/83.3)
and  working papers on  specific issues were  considered by  the

twenty-eighth  meeting  of  JECFA (1).   Several  questions were
identified  as  remaining  to be  considered,  including special
problems associated with:

    (a)  bulking agents and novel foods;

    (b)  food contaminants;

    (c)  animal feed additives and veterinary drug residues;

    (d)  test  methods  and  principles  (including  alternative
         methods of testing);

    (e)  testing for allergenicity;

    (f)  lesions  observed  in  bioassays that  are difficult to
         interpret  (a  number  of  examples  are  cited  in the
         report); and

    (g)  assessment  procedures; extrapolation and  quantitative
         assessment.

    The Committee recommended that a unified document  on  these
issues should be prepared for consideration by JECFA at a future
meeting.   The present publication has been prepared in response
to that recommendation.

    In  carrying out the review  of methodology for the  testing
and  evaluation of intentional and unintentional food additives,
the  working group has  taken notice of  recommendations, guide-
lines, and procedures adopted by national regulatory authorities
and  international/supra-national  organizations  including  the
Organization  for  Economic Cooperation  and Development (OECD),
the  International Agency for Research on Cancer (IARC), and the
European  Economic  Community  (EEC)  Scientific  Committee  for
Food.  It is recognized that, where possible, a unified approach
should  be adopted.  However, the purposes for which these other
bodies have formulated guidelines differ in detail from those of
JECFA,  and it is inappropriate  to adopt these without  modifi-
cation to meet the needs of JECFA.

2.2.  Periodic Review

2.2.1.  Concept of periodic review

    JECFA has indicated that, in discharging its duty  to  eval-
uate  the  safety-in-use  of intentional  and unintentional food
additives,  it may  be necessary  to carry  out a  periodic  re-
evaluation of substances previously assessed by the Committee.

    The  first JECFA meeting,  in looking ahead,  envisaged,  in
addition  to the continuing  evaluation of food  additives, that
there  would  be a  re-evaluation  process associated  with  the
programme  on food additive  safety assessment (9).   It stated:

    "Permitted  additives should be subjected to continuing
    observation  for  possible  deleterious  effects  under

    changing conditions of use.  They should be reappraised
    whenever  indicated by advances in  knowledge.  Special
    recognition  in  such  reappraisals should  be given to
    improvements in toxicological methodology."

    This principle was endorsed in the third (25), seventh (26),
eighth (11), and ninth reports (12) of JECFA.

    The  "need  for review  of  past recommendations"  was high-
lighted in the thirteenth JECFA report as follows (27, p. 22):

   "There  is a widespread but  fallacious belief that
    clearance of an additive for use in food constitutes an
    irrevocable decision.  Such a view renders a grave dis-
    service  to  the cause  of  consumer protection  for it
    fails to recognize the need for regular review  of  all
    safety evaluations."

    Periodic   review  of  past  decisions  on  safety  is  made
necessary by one or more of the following developments (27):

    (a)  A new manufacturing process for the food additive.

    (b)  A new specification.

    (c)  New data on the biological properties of the compound.

    (d)  New  data  concerning  the nature,  or  the  biological
         properties,  or both, of  the impurities present  in  a
         food additive.

    (e)  Advances  in scientific knowledge germane to the nature
         or mode of action of food additives.

    (f)  Changes  in consumption patterns or  level of use of  a
         food additive.

    (g)  Improved  standards of safety evaluation.  This is made
         possible  by new scientific  knowledge and the  quality
         and quantity of safety data considered necessary in the
         case  of new additives.   Since JECFA began  the  eval-
         uation  of  food  additives  in  1956,  the  paucity of
         information  available on many food  additives has been
         such  that  assessments  have often  been  difficult to
         make.   Tests of too  short duration, conducted  with a
         very  small  number  of animals  at  inappropriate dose
         levels,  and without adequate clinical, haematological,
         chemical,   or   histopathological  examinations   have
         frequently  been  encountered among  the data submitted
         for  evaluation.  Tests of this sort cannot be regarded
         as having permanent validity; with the passing of time,
         they  need to be supplemented by studies carried out in
         full accordance with the recommendations set out in the
         report  of the WHO  Scientific Group on  Procedures for
         Investigating   Intentional   and  Unintentional   Food
         Additives (2).

    It  should  also  be noted  that  the  second Joint  FAO/WHO
Conference  on Food Additives  (7) recommended that  it  (JECFA)
"should revise, if needed, the toxicological evaluation  of  all
additives   considered  in  previous  meetings   of  the  Expert
Committee."

    The seventeenth report of JECFA (16) reads, in part:

    "The  objective in assessing the  toxicological data on
    food additives is to ensure their safety for  the  con-
    sumer on the basis of all the evidence available to the
    Committee  at  the  time.  Future  results with present
    methods  or with techniques  yet to be  developed  will
    necessitate  reassessments that may lead  to changes in
    earlier decisions."

    Other  meetings  reaffirmed the  need  to take  advantage of
recent developments in toxicological techniques for research and
safety assessment.

    These   Committee  recommendations  and   observations,  the
rationale  set forth  in the  thirteenth report  (27) and  other
reports for the need to review past decisions, and  the  ensuing
years of progress in the science of toxicology and refinement of
research,   evaluation  procedures,  and   changing  consumption
patterns all point towards the advisability of  periodic  review
of this large class of substances.

2.2.2.  Mechanism of periodic review

    That a considerable amount of re-evaluation of substances is
already  carried out within the system is evident when the year-
to-year  agenda of JECFA is examined.  Food additives are reass-
essed when new biological and chemical data are  made  available
to  FAO and WHO.  In  fact, new data are  mandated on timetables
established by JECFA when temporary ADIs are  established  (sec-
tion 5.5.5). In addition, re-evaluations are made at the request
of Member States and by the Codex Alimentarius Commission.

    However,  for many additives,  the assessment has  not  been
conducted  using  the  more  recently  adopted  procedures   for
investigating  intentional and unintentional food  additives.  A
review of past decisions also reveals that some  additives  have
only had a cursory examination.  The evaluation of  these  addi-
tives may have been based on limited data.

    A   periodic  review  programme  on   substances  previously
reviewed  by JECFA should be constituted to reflect the changing
state-of-the-art and to provide the best possible  assurance  to
consumers of food additive safety.  However, a mechanism has not
yet  been developed for  the continuous systematic  updating  of
safety  information on food additives.   Of course, even in  the
absence  of a periodic review  programme, if new data  on a food
additive  raises suspicion of significant hazard, then immediate
re-evaluation is conducted.

    The  use of an international forum to devise and implement a
system for the periodic review of chemicals used in or  on  food
and  contaminants of food  could also be  of great economic  and
practical  value to Member  States.  It would  ensure a  uniform
approach  to  a  complex toxicological  problem,  duplication of
effort  would be  minimized, and  emphasis on  such a  programme
would give added reassurance to consumers throughout  the  world
that  the food  supply continues  to be  safe.  Perhaps  such  a
programme  could  be developed  in  cooperation with  the  Codex
Alimentarius Commission.

3.  CRITERIA FOR TESTING AND EVALUATION

    JECFA  has  always operated  on  the principle  that testing
requirements  for all  food additives  should not  be the  same.
Such  factors  as  expected toxicity,  exposure  levels, natural
occurrence  in  food (section  6.1),  occurrence as  normal body
constituents  (section  5.2),  use  in  traditional  foods,  and
knowledge of effects in man (section 5.4) should be  taken  into
account.  In relation to carcinogenic hazards, the Committee has
stated that "the scope of the test required should depend  on  a
number  of  factors, such  as the nature  of the substance,  the
extent  to which it might be present in food, and the population
consuming  it" (4).  More generally, the Committee has requested
data  on,  inter alia,  method(s) of manufacture, impurities, fate
in  food, levels of use  of additives in food,  and estimates of
actual  daily intake, and  concluded that such  information "was
important and relevant both for the toxicological evaluation and
for the preparation of specifications" (22).  However, difficul-
ties arise when an attempt is made to determine testing require-
ments  because  of  problems in  predicting toxicity, estimating
levels   of  food  additive  use  and  natural  occurrence,  and
obtaining  human data.  As discussed below, criteria for testing
requirements  can also be  used to allocate  priorities for  the
testing and evaluation of food chemicals.

3.1.  Criteria for Testing Requirements

    The   establishment   of  principles   for  determining  the
appropriate amount of data that will be required  to  adequately
evaluate  the safety of additives at their estimated consumption
levels  is urgently needed  to ensure consistency  in  decision-
making by the Committee and to provide guidance to  sponsors  of
food  chemicals.   Both  exposure data  and  potential  toxicity
should be important considerations in the establishment of these
principles.   Consideration of only one of these elements to the
exclusion  of the other  leaves serious deficiencies.   If  only
exposure  data are used, then  no consideration is given  to the
wide  range  of  toxicities  observed  among  chemicals  and  no
advantage is taken of the vast amount of bioassay  data  already
in existence.  If, on the other hand, only toxicity information,
predicted  or known, is  used, then chemicals  with known  toxic
properties or those related to chemicals of known high toxicity,
particularly  carcinogenicity,  would automatically  require the
most  data,  with  no  consideration  given  to  relatively  low
exposure levels.

3.1.1.  Estimating exposure

    For  the purpose of this publication, exposure is defined as
the total intake of a chemical substance by human  beings.   For
the  majority of substances evaluated by JECFA, the primary mode
of  exposure is through ingestion  of the substance in  the food
supply.

    Estimates of exposure used by Committees in  previous  years
are of three general types:  per capita  estimates, estimates from
dietary  food intake surveys, and analytical values from market-
basket/total-diet  surveys.   For  a detailed  discussion of the
advantages and the use of these different types of estimates see
reference no. 28.

    The  per capita  approach is an  estimated value that  repre-
sents  the exposure level if a food additive or contaminant were
equally  distributed  across  a population.   For example, a  per
 capita  intake for a nation  may be calculated  by dividing  the
total  yearly  production  volume,  corrected  for  imports  and
exports,  of a chemical  used in food,  within a nation,  by the
national  population.  Another form of  per capita  intake may be
computed  from a nation's  per capita  disappearance of a certain
food commodity multiplied by the usual level of an  additive  or
contaminant in the food commodity. These  per capita  intakes can
be converted to daily intake per kilogram body weight.

    In  some countries, dietary  surveys are performed  on food-
stuffs consumed by a representative group of individuals, within
a national population, over a short period of time, e.g., 1 - 14
days.   The intake of an additive or contaminant, per food type,
can be calculated by multiplying the usual additive  or  contam-
inant  level  in  each type of food by the dietary intake of the
food.   The intakes per food type can then be summed to derive a
total  additive  or contaminant  intake.   An advantage  of  the
dietary  survey approach is that additive or contaminant intakes
for  selected subpopulations, such  as different age  groups  or
high-frequency consumers of certain foodstuffs, may be computed,
depending on the specificity of the dietary survey.

    When  considering  intakes  computed by  the  dietary survey
approach,  the tenth meeting of JECFA (29, pp. 23-24) reaffirmed
the validity of calculating the average daily intake of  a  food
additive  based on: (a)  levels arising from  good technological
practice;  (b)  average  consumption  of  foods  containing  the
additive;  and  (c) average  body  weight.  This  Committee also
noted at the time that, while data on average  food  consumption
were available from many countries, high consumption  data  were
available from only two countries.  The Committee  recognized  a
special  need  for determining  how much of  a food additive  is
likely  to  be  consumed by  groups  that  have a  high level of
consumption and strongly recommended that every effort should be
made to obtain such information on food consumption.

    The  fourteenth meeting of  JECFA (30) considered  methodol-
ogies  for computing additive intakes from dietary food surveys,
and  recognized the importance  of experimental design  so  that
collective  data can be used for calculating reliable intakes on
an individual basis.  The Committee noted difficulties in common
descriptors  for  food items,  when  information is  gathered in
surveys  performed by different organizations,  and in obtaining
confidential information about food additive use from industry.

    Market-basket  surveys  (also  called  total-diet   surveys)
involve  analyses of representative diets for the usual level of
additive or contaminant in the diet.  The analyses may  be  per-
formed  on  food  mixtures  or  on  individual  foodstuffs.  The
selection  of foods represents a normal diet for a certain popu-
lation,  such as the typical  daily diet for a  certain nation's
average  consumer.   Market-basket  surveys  can  be  used   for
estimating  the actual  level of  additive or  contaminant in  a
selected total diet, which is of value for substances present in
food at levels that are less than the amounts  added.   However,
the difficulties of analyses usually restrict this  approach  to
estimations of average intakes of contaminants in samples repre-
sentative  of the average dietary  habits of a nation's  general
population  rather  than  estimations of  intakes  for  selected
subpopulations.  In this regard, data on certain contaminants in
food  are  available  from the  Global  Environmental Monitoring
System (GEMS).

    These  procedures are useful for estimating exposure to food
additives in individual countries.  However, accurate estimation
is  much  more  difficult when  attempted  on  a  global  scale.
Clearly,  consumption of a food additive will not be the same in
two countries in which it is regulated with  differing  restric-
tions  or with very different food consumption patterns.  To use
exposure  estimates on such a  scale as a criterion  for testing
requirements or for setting priorities for the testing  of  food
additives  is an extremely ambitious exercise that would require
extensive resources.

3.1.2.  Predicting toxicity from chemical structure

    Chemical structure determines to a great extent the attitude
of the toxicologist towards a compound.  As a result, there have
been many efforts to systematize the use of  chemical  structure
as a predictor of toxicity.  The use of such  relationships  has
been  suggested  by JECFA  with  certain classes  of  flavouring
agents (section 6.1.2), and chemical structure is  an  important
consideration  in the selection of compounds for carcinogenicity
testing.   Structure/activity relationships also form  the basis
for establishing group ADIs (section 5.5.4).

    Structure/activity   relationships   appear  to   provide  a
reasonably   good  basis  for   predicting  toxicity  for   some
categories  of compounds, primarily carcinogens, which are char-
acterized  by  specific  functional groups  (e.g., nitrosamines,
carbamates,  epoxides,  and  aromatic amines)  or  by structural
features  and  specific  atomic arrangements  (e.g.,  polycyclic
aromatic hydrocarbons and aflatoxins).  However, all these chem-
ical  groups have some members  that do not seem  to be carcino-
genic or are only weakly so.  Since structure/activity relation-
ships  are  better established  for  carcinogens than  for other
toxic  end-points,  dependence  on such  predictions  emphasizes
suspect carcinogens at the expense of other forms  of  potential
toxicity.   However, as more  chemicals are tested  for toxicity
and other end-points are identified in the future, the data base
will become larger, which should permit more  valid  comparisons
between structure and toxicity among more classes of compounds.

    In terms of carcinogenic substances, another system that has
sometimes  been used  for predictive  purposes is  a battery  of
tests  for genotoxicity (possible applications of such tests are
discussed in section 5.1.5).

3.1.3.  Other factors to consider when developing criteria

    The  value of structure/activity relationships  and exposure
data  in  determining  the extent  of  testing  required may  be
considerably  enhanced  by collateral  information on metabolism
and pharmacokinetics.  It has been previously accepted that:

    "if a series of chemical analogues can be shown to give
    rise  to the same metabolic product. . . it may be suf-
    ficient to carry out toxicological studies on  a  suit-
    able  representative of the series" and "where adequate
    biochemical  and toxicological data on  closely related
    chemicals  are  available,  the objective  (of toxicity
    tests)  becomes the detection of any deviation from the
    established pattern.  This can usually be determined by
    intensive studies of a few months duration  when  these
    are adequately designed and evaluated"  (2).

More recently (31), JECFA has concluded that:

    "if  the chemical structure of a compound under consid-
    eration  did  not closely  resemble  that of  any known
    toxic  or carcinogenic compound, and,  if the toxicolo-
    gical data on it, its metabolites, and  its  homologues
    did not give any cause for concern, these  less  exten-
    sive toxicological data might be used for  the  evalua-
    tion  of the compound.  . . .   In the evaluation  of a
    series   of  structurally-related  compounds,  complete
    toxicological data should be available for at least one
    member  of the series.   Other compounds in  the series
    should  be evaluated on the  basis of these data,  plus
    data on their natural occurrence and metabolism, and on
    the toxicology of their homologous compounds."

    These principles can form the basis for determining the lim-
ited amount of testing that may be required for  compounds  that
are  closely  related  structurally.  If  the toxicological data
base is adequate for the homologous compounds and suggests a low
intrinsic toxicity, metabolic and pharmacokinetic data alone may
be sufficient to make an evaluation of a related compound.

    The  results  of  studies on  absorption,  distribution, and
metabolism  may either increase  or decrease the  health concern
from  the use of the  additive.  For example, a  relatively non-
toxic additive may be transformed by liver enzymes into  a  sub-
stance  with  a much  greater  toxic potential,  or  vice versa.
Correlations  between  structure  and activity  will often auto-
matically  include these considerations, because substances of a
particular  class will often be absorbed, distributed, and meta-
bolized in similar ways.  However, this will not always  be  the
case,  and  these  parameters should  be specifically considered
when making such correlations.

    Other  factors  influence the  extent  and type  of  testing
required  for  safety assessment.   For  example, the  need  for
extensive  testing may be  mitigated when the  substance  occurs
naturally in food and has a history of human use or when  it  is
metabolized into normal body constituents (section 5.2.3).  More
extensive  testing in animals may be necessary when the additive
will  be used in special  populations at risk, such  as pregnant
women and very young infants (section 5.3).  Human  testing  may
be needed if problems of intolerance arise (section 5.4.2).  The
types  of  end-points,  as discussed  in  section  5.1, must  be
considered in any criteria system that is established.

    The  development of criteria  for determining the  extent of
required testing is worthy of extensive future study.  Its value
would be considerably enhanced by including it in the context of
a priority-setting scheme, as discussed below, because additives
should not be considered in isolation from one another.

3.2.  Priorities for Testing and Evaluation

    The primary basis for establishing the list of substances to
be  considered by  JECFA is  the recommendations  of  the  Codex
Committee  on  Food  Additives (CCFA)   and  Member Governments.
However,  Committees  have recognized  the  need for  the estab-
lishment  of a "priority list  as a means of  selecting the most
relevant compounds for future evaluation.  In order to establish
priorities  for  the  toxicological testing  and  evaluation  of
intentional and unintentional food additives", JECFA recommended
at the twenty-second (32) and twenty-third (31)  meetings  that:

    "FAO  and  WHO  convene an  inter-disciplinary group of
    experts  to  establish  an inventory  of compounds that
    have  not yet been fully evaluated and to classify them
    in  terms of their  potential hazard to  health on  the
    basis of toxicological knowledge and extent of use."

The  Committee has  recognized that  the most  obvious need  for
allocating priorities is for the testing and  safety  evaluation
of food flavouring agents (19).  Committees continue  to  stress
the need to establish priorities for testing and evaluating food
additives (24,33).

    One  basis  for  establishing priorities  for  testing  food
additives  is  by using  an index based  on exposure levels  and
predicted  toxicity.   For  examples of  approaches  using these
parameters, see references 34-38.

    As  discussed  in section  3.1.1,  valid exposure  data  are
extremely  difficult to develop.  However, comparative levels of
consumption  may be sufficient for the purpose of setting prior-
ities for the testing of food additives.  Therefore, even though
accurate  consumption  estimates of  wide geographical relevance
will  probably never be achieved, the lesser requirement of com-
parative estimates may be achievable to the extent necessary for
JECFA's  use, if the Committee  decides to develop the  informa-
tion.

    Because  of  the  semiquantitative  nature  of  much  of the
biological  data  available  for predicting  toxicity,  rigorous
analytical or statistical interpretation is not always possible.
Therefore, expert interpretation and evaluation of the  data,  a
time-consuming  and expensive procedure, must be integrated into
any  automated decision-making  mechanism that is developed.  To
ensure  maximum  usefulness, the  priority-setting system should
take  account  of all  available  toxicity and  other biological
information,  including metabolic and  human data.  A  properly-
devised system will be capable of considering new data  and  can
be modified using modern data processing methods and equipment.

3.3.  Quality of Data

    In  recent years, various  national regulatory agencies  and
international  bodies have instituted  codes of Good  Laboratory
Practice  (GLP), the  aim of  which is  to help  underwrite  the
validity  of studies by ensuring  that they can be  verified and
reproduced.   GLP  codes  require  the  maintenance  of  certain
records  regarding  the  performance of  studies, including data
from  chemical and toxicological  tests, which help  ensure full
documentation of the conduct and results of  studies.   However,
GLP  codes  are  not a  substitute  for  scientific quality;  an
inappropriate study may be conducted according to GLP standards.
On the other hand, a study that does not meet GLP  criteria  may
still be scientifically sound.

    JECFA  has always judged studies  on their merits, the  main
criteria  being  that  the  study  was:  (a)  carried  out  with
scientific  rigor,  and (b)  reported  in sufficient  detail  to
enable comprehensive evaluation of the validity of the results.

    Usually,  studies  that  are  published  in  the  scientific
literature  are subjected to  peer review prior  to publication,
and  after publication, the  results are open  to refutation  or
confirmation.  Unpublished reports, on the other hand,  are  not
necessarily  subjected to this scrutiny.  For this reason, JECFA
has  repeatedly  recommended  that  data  brought  before  it be
published (10, p. 6; 12, p. 7; 39, p. 7; 40).  However, in point
of fact, JECFA does review many high-quality studies that remain
unpublished  for  proprietary  and  other  reasons.   Also,  the
Committee  often  requests  unpublished raw  data when published
reports  do  not  include  sufficiently  detailed  data  for  an
adequate  safety review.  Studies  performed in compliance  with
GLP   codes  provide  added   assurance  that  the   quality  of
unpublished  data  is  acceptable.   For  these  reasons,  it is
appropriate that JECFA experts continue to consider all the data
brought  before  them,  published,  and  unpublished,  and  make
decisions about the validity of these data on  an  ad hoc  basis.
This  means that the studies reviewed by the Committee should be
fully documented.

4.  CHEMICAL COMPOSITION AND THE DEVELOPMENT OF SPECIFICATIONS

    The  proper  safety evaluation  and  use of  food  additives
requires  that they be chemically  characterized. Therefore, the
Committees  review data relating  to the identity  of additives,
impurities that may be present, and possible  reaction  products
that   may   arise   during  storage   or   processing.   "JECFA
specifications"   are  then  elaborated, taking  these and other
factors into account.

4.1.  Identity and Purity

    To establish the chemical identities of food  additives,  it
is necessary to know the nature of the raw materials, methods of
manufacture, and impurities (22).  This information is  used  to
assess the completeness of analytical data on the composition of
additives  and to  assess the  similarity of  materials used  in
biological  testing  with  those  commercially  produced.   From
information  on  raw  materials  and  methods  of   manufacture,
potential  impurities  in  commercially  manufactured   chemical
materials due to carry-over of contaminants in raw materials and
by-products of the manufacturing process can be predicted.

    To  evaluate biological testing data  from multiple studies,
JECFA must have information on the chemical composition  of  the
tested  materials, which necessitates manufacturing information.
Analytical data on the chemical composition of materials used in
biological  testing  should be  more  detailed than  a  standard
presentation of chemical specifications.  Furthermore, materials
used   in  biological  testing   should  be  representative   of
substances  manufactured by actual commercial  processes so that
the   materials   administered  to   experimental  animals  will
represent those ingested by consumers.

    A  food  additive  may be  a  single  chemical substance,  a
manufactured  complex  chemical  mixture, or  a natural product.
The  need  for  complete information  on  chemical  composition,
including  description,  raw materials,  methods of manufacture,
and analyses for impurities, is equally valid for each  type  of
additive.    However,  implementation  of  the  requirement  for
chemical  composition data  may vary  depending on  the type  of
substance.   For additives that are  single chemical substances,
it  is virtually impossible  to remove all  impurities in  their
commercial   production;   therefore,  analyses   are  generally
performed on the major components and predicted impurities, with
the highest significance placed on potentially toxic impurities.
For  commercially manufactured complex  mixtures, such as  mono-
and   diglycerides,  information  is  needed  on  the  range  of
substances  commercially produced, with emphasis on descriptions
of  manufacturing processes, supported by analytical data on the
components  of the different commercial  products.  Natural pro-
ducts  present particularly difficult problems  because of their
biological variability and because the chemical constituents are
too  numerous for regular  analytical determinations; thus,  the
analyst  is starting with  an "unknown".  For  additives derived
from  natural products, it is vital that the sources and methods

of manufacture are precisely defined.  Chemical composition data
should  include  analyses for  general chemical characteristics,
such  as proximate analyses  for protein, fat,  moisture, carbo-
hydrate,  and mineral content,  and analyses for  specific toxic
impurities  carried over from raw materials or chemicals used in
the manufacture of the substance.  Further information necessary
for  the evaluation  of "novel  foods", which  are usually  sub-
stances derived from natural products, is provided  in  sections
6.2.1 and 6.2.4.

4.2.  Reactions and Fate of Food Additives and Contaminants in Food

     Biological testing of food chemicals must relate  to  their
presence  in food as consumed.   This is an important  consider-
ation  when added substances  undergo chemical change  in  food.
Therefore, data are necessary on the reactions and fate of addi-
tives  or contaminants in  food and their  effects on  nutrients
(22).

    Certain  food additives perform  their functional effect  by
reaction  with undesirable food constituents (e.g., antioxidants
react  with oxygen in food and EDTA reacts with trace metals) or
by  reactions  that  modify food  constituents  (e.g., potassium
bromate  reacts  with  dough constituents).   Food additives may
also degrade under certain conditions of food processing, though
such degradation is detrimental to their functional effect.  For
example,  the sweetener aspartame  is transformed to  a  diketo-
piperazine  derivative at rates varying with the acidity and the
temperature  of the food.  In previous evaluations of "reactive"
additives,  Committees  have  evaluated  analyses  for  additive
reaction products in food, as consumed, and  biological  testing
data  on either specific  reaction products or  samples of  food
containing  the  reaction  products as  consumers  would  ingest
them.

    For  all intentional food additives proposed for evaluation,
Committees request submission of four types of data  related  to
reactivity:

    (a)  the general chemical reactivity of the additive;

    (b)  its  stability  during  storage and  reactions in model
         systems;

    (c)  reactions of the additive in actual food systems; and

    (d)  the  additive's fate in living systems.  These data are
         important  for  relating  biological test  data  to the
         actual use of the additive in food.

    Processing aids are substances that come into  contact  with
food  during processing and  may unintentionally become  part of
food because of their incomplete removal.  Committees have eval-
uated a number of processing aids, such as  extraction  solvents
and  enzyme preparations, for their safety in use.  When evalua-
ting a processing aid, information should be provided on its use

and either analytical data on the amount of the  processing  aid
carried over into food or a computed estimate of the  amount  to
be  expected in food.  In some cases, a component of the proces-
sing aid may have the greatest potential for biological effects,
such as ethylenimine leaching from polyethylenimine, an immobil-
izing  agent  used  in  the  preparation  of  immobilized enzyme
preparations.

    Contaminants   in  food  evaluated  by  previous  Committees
include environmental contaminants and substances migrating from
food packaging.  Of environmental contaminants, metals have been
considered  the most often.   Committees request information  on
the  chemical forms of  metals in the  food supply (e.g.,  ionic
form and/or covalently bonded chemical form) and  their  concen-
tration  distribution  in  the  food  supply,  as  determined by
analyses  of  food or  experimental  models for  carry-over from
environmental  sources.   For  contaminants  derived  from  food
packaging,  data are required on the identification of chemicals
migrating from the packaging material and concentrations in food
(analysed or estimated from migration modelling studies).

4.3.  Specifications

    Specifications  are a necessary  product of Committee  eval-
uations, the purposes of which are 3-fold:

    (a)  to  identify the substance  that has been  biologically
         tested;

    (b)  to ensure that the substance is of the quality required
         for safe use in food; and

    (c)  to reflect and encourage good manufacturing practice.

    The  first Joint FAO/WHO  Conference on Food  Additives  (6)
established  a programme for the collection and dissemination of
information  on the chemical, physical, pharmacological, toxico-
logical, and other properties of individual food additives.  The
first two meetings of the Joint Expert Committee,  in  preparing
reports  on  "General  Principles  Governing  the  Use  of  Food
Additives"  (9) and "Procedures for the Testing  of  Intentional
Food  Additives  to  Establish  Their  Safety  for  Use"   (10),
recommended  that specifications should be  prepared, citing the
need for:

    (a)  limiting impurities in food;

    (b)  identifying materials used in toxicity testing; and

    (c)  ensuring  that the additive tested is the additive used
         in practice.

The third meeting of JECFA was devoted in its entirety  to  dev-
eloping  principles governing the elaboration  of specifications
and developing provisional specifications for the first group of
additives evaluated by the Committee (25).

    JECFA specifications are minimum requirements for the compo-
sition  and quality of food-grade additives, allowing for accep-
table  variation in their production (18).  These specifications
are meant to be used internationally and to the extent that data
are  available, specifications are elaborated  to cover suitable
products  manufactured in various parts of the world.  The third
meeting  considered the value  of specifications with  regard to
protection for the consumer, advice to regulatory organizations,
standards for the food industry, and establishment of safety for
use  (relative to identification of materials used in biological
testing  in  comparison  with materials  produced for commercial
use).  The format for specifications established by this meeting
continues to be used in current JECFA specifications,  that  is,
the  additive  is  identified by  synonym,  definition (chemical
name,  formula, relative molecular mass, etc.), and description,
its functional uses are listed, tests of identity and impurities
are  provided, and an assay  for the major component(s)  is pro-
vided.  The third meeting of JECFA, recognizing  that  practical
specifications  could  not  specify every  impurity, limited the
scope of impurity tests to constituents of commercially produced
substances  that:   (a)  were related  to  the  safe use  of the
additive; (b)  might affect the usefulness of the  additive;  or
(c)  would serve as an indicator of good manufacturing practice.
Finally,  the meeting concluded  that, for specifications  to be
acceptable on a global basis, they must be subject to continuing
review and evaluation to take into account the  presentation  of
new   information,   particularly  with   respect  to  different
manufacturing processes and improved analytical methods.

    In  detailing the purposes  of JECFA specifications,  Commi-
ttees  have, through the years, refined the scope of their spec-
ifications  and provided  advice on  how they  should  be  used.
Specifications  developed by each  Committee should be  read  in
conjunction with the report of that Committee.   JECFA  specifi-
cations  apply  to  the  material(s)  that  was  toxicologically
reviewed  and take into  account the uses  of the additive  (17,
41).   Periodic review of specifications is required, because of
changes  in patterns of additive  use, in raw materials,  and in
methods  of  manufacture.   Comments on  JECFA specifications by
national and international organizations are valuable sources of
information for periodic review (18, 27, 29).

    JECFA  specifications in their entirety  describe substances
of food-grade quality, and as such, they are directly related to
toxicological   evaluations and to good  manufacturing practice.
However,  though  specifications  may include  criteria that are
important for commercial users of additives, they do not include
requirements  that  are of  interest  only to  commercial  users
(42).

    Differences  may  exist  between specifications  prepared by
national and international organizations; however, the Committee
is  not aware of any  information indicating that these  differ-
ences incur health risks for consumers (23).   JECFA  specifica-
tions are meant to be minimum requirements for the safe  use  of
additives,  and not every component of commercially manufactured

chemical substances is subject to an impurity test  (11).   Test
requirements  in JECFA specifications  are sufficient to  ensure
the  safe use of commercially manufactured food additives.  Sub-
stances  of  higher  chemical  purity  (e.g.,  analytical  grade
reagents)  are not excluded from  use in food, even  though such
substances may deviate somewhat from the identification tests in
the  specifications, provided that they meet the stated require-
ments  for specified purity tests and are otherwise suitable for
use as food additives (18).

    Since 1956, the meetings of JECFA have  designated  specifi-
cations  as either full or tentative.  Specifications were given
the  tentative  designation  from  the  third  to  twenty-second
meetings  because either the  chemistry data needed  to  prepare
specifications were not adequate or a temporary ADI was assigned
to  the additive.  At,  and since, the  twenty-third meeting  of
JECFA, the tentative designation has been assigned only when the
data necessary for preparing specifications were insufficient.

    JECFA  policy has been  to prepare specifications  for  sub-
stances  added to food,  whenever constituents of  the substance
had  the potential to be present in food.  Initially, specifica-
tions  were prepared only  for intentional food  additives  that
were  added directly, to accomplish a functional effect in food.
The  fourteenth  JECFA  (30) evaluated  extraction solvents, the
first  group of   "processing aids" that had  been  reviewed  by
JECFA.   This Committee concluded that, although extraction sol-
vents  are substantially removed  from food, evaluation  of  the
conditions of safe use of these solvents depends on the identity
and  purity  of  the solvents.   Therefore, JECFA specifications
were  prepared.  Since then,  specifications have been  reported
for other processing aids such as anti-foaming/defoaming agents,
clarifying  agents,  decolourizing agents,  enzyme preparations,
filtering  aids, packing gases,  propellants, lubricants/release
agents,  odour/taste-removing  agents,  and yeast  "food" (yeast
nutrients).  The twenty-seventh JECFA (24) decided that chemical
reagents used in the preparation of food additives or processing
aids (such as glutaraldehyde in the preparation  of  immobilized
enzyme  preparations or acetic  anhydride in the  manufacture of
modified  starches) do not usually  need specifications.  Carry-
over  of these reagents or  their contaminants into food  may be
controlled in the specifications for purity of the  additive  or
processing aid.

    Food  additives may be marketed  as formulated preparations,
such  as a mixture of  a main ingredient with  a solvent vehicle
and emulsifier.  Specifications refer only to each ingredient in
the  formulated preparation as  individual commercially-manufac-
tured food additive substances.  Mixtures should not  be  formu-
lated  in such a  way that the  absorption or metabolism  of any
ingredient  is altered so that  the biological data are  invali-
dated  (12, 42).  Added  substances such as  anticaking  agents,
antioxidants,  and stabilizers may also influence the results of
tests  given  in  specifications.  Therefore,  in its nineteenth
report,  JECFA recommended that manufacturers  of food additives
should indicate the presence of such added substances (18).

    In considering whether specifications apply to food additive
quality  as manufactured or  as received, JECFA  has decided  to
prepare  specifications to cover  the normal shelf-life  of  the
product.   Limits are set  for decomposition products  that  may
form  during normal storage.   Manufacturers and users  of  food
additives  should ensure good  packaging and storage  conditions
and  use good handling practices to minimize deleterious changes
in  quality and  purity (18).   Information on  changes  in  the
composition of food additives during storage should be submitted
for evaluation by the Committee.

    In addition to periodic reviews to examine  the  consistency
of  specifications  within  classes of  similar additives, JECFA
periodically  reviews specification test  methods to update  the
analytical  methodology  of  specifications.  Two  summaries  of
specification test methods have been published (43,  44),  which
provide  guidelines  for  the application  and interpretation of
specification  requirements  and  test methods.   JECFA has made
considerable  progress in adopting modern analytical methodology
for  specification tests, whenever equipment  and other supplies
needed  to perform  the tests  are accessible  on  a  world-wide
basis.  However, because JECFA specifications are elaborated for
world-wide  use, certain analytical methods  involving recently-
developed  techniques or equipment cannot be included until such
techniques are available on an international scale.  Alternative
methods  of analysis can be used to test products for conformity
with  specifications, provided that  the methods and  procedures
used produce results of equivalent accuracy and specificity.

    In order to foster international agreement on specifications
for  food-grade  substances,  JECFA seeks  comments  from member
countries  and international organizations.  The  Codex Aliment-
arius  Commission  systematically  provides  comments  on  JECFA
specifications  through  the  Codex Committee  on Food Additives
(CCFA)  and  endorses  certain JECFA  specifications  as  "Codex
Advisory   Specifications".   The  systematic  review  of  JECFA
specifications by the CCFA has provided JECFA with valuable data
on novel manufacturing processes, previously unknown impurities,
updated  methodology, and advice on  the format of JECFA  speci-
fications.

    Although  JECFA  specifications  and  those  of  the   Codex
Alimentarius  Commission  are elaborated  for  many of  the same
purposes,  the interpretation of  these purposes may  result  in
differences  in specific requirements  or test methods  for  the
same  food-grade substance.  In replying to suggested changes in
JECFA  specifications from the CCFA or other interested parties,
it  may be decided  to amend existing  specifications, providing
that  the  requested changes  do  not significantly  lessen  the
assurance  of  food-grade  quality  embodied  within  the  JECFA
specifications and that the requested change conforms  with  the
principles  for  elaboration  of specifications  established  at
previous meetings.  A requested change in an existing full JECFA
specification must be supported by scientific data.

5.  TEST PROCEDURES AND EVALUATION

5.1.  End-Points in Experimental Toxicity Studies

    There  are virtually no findings  in experimental toxicology
that  can be simply extrapolated to man without careful thought.
During  the past two decades, there has been an increasing emph-
asis  on carcinogenesis as a manifestation of chemical toxicity.
Most of the other manifestations of chemical  intoxication,  for
example,  immunosuppression, have, by comparison, received rela-
tively  little attention.  This  has resulted in  an  unbalanced
approach by the toxicologist in which the emphasis on end-points
bears a less and less obvious relationship with disease patterns
in  man.  For example, a  survey of the recommendations  for the
evaluation  of food  additives has  revealed that  little or  no
attention  is paid to  the detection of  cardiovascular lesions,
even  though these lesions are  the most common cause  of fatal-
ities  in  the  human  population  in  developed  countries.  In
addition, certain lesions commonly found in rodents that are the
primary  targets of the toxicologist do not have any counterpart
in  man.  It seems reasonable  that an effort should  be made to
relate  toxicological findings more carefully to the human situ-
ation,  recognizing that this will  be a long-term project.   In
the meantime, when conducting experimental animal tests, special
attention should be paid to alterations that indicate  a  poten-
tial  for the  test compound  to adversely  affect  the  cardio-
vascular,   immunological,  reproductive,  or   central  nervous
systems.   If  such alterations  are  detected, they  should  be
investigated  further using special studies  aimed at clarifying
their significance.

    The end-points discussed in this section have  been  grouped
for convenience into effects with:

    (a)  functional manifestations only;

    (b)  non-neoplastic morphological characteristics;

    (c)  neoplastic manifestations; and

    (d)  reproduction/developmental manifestations.

In  view  of  the large  number  of  effects encountered,  it is
possible  to summarize only  some of the  specific observations.
However,  situations  that  have become  controversial are dealt
with in more detail.

    Finally, this section concludes with a short  discussion  on
the role of short-term  in vitro  tests in the  safety assessment
of food additives.

5.1.1.  Effects with functional manifestations

    Generalized weight loss, although having causes that are not
solely  physiological  (section  5.5.3),  does  not  necessarily
involve  any  particular  pathological lesions  (section 5.5.3).

Reduced weight gain has played a major role as an  end-point  in
toxicological  determinations in various  ways.  In a  sense, it
has often been used for determining various empirical indices in
the absence of other manifestations.  The procedures followed by
JECFA  for determining an  ADI demand that  a no-observed-effect
level  should be established.  For this level to be established,
it  is necessary to establish an effect level and, when all else
has failed, a generalized decrement in weight gain has been used
for  this  purpose,  provided reduced  food  intake  is not  the
obvious cause.  The other major use of a decrease in weight gain
has been in establishing a maximum tolerated dose (MTD)  (for  a
definition of the MTD, see Annex I).

    Among  the  commonest effects  observed  in studies  on food
additives  is a laxative  effect; the physiological  reasons for
this  are usually quite apparent  and can be taken  into account
when  considering  the appropriate  levels  of use  of additives
causing  this effect.  In  most instances, additives  have  been
permitted  that cause laxation at  high levels in man  when they
have  been otherwise non-toxic  and can be  used effectively  at
levels at which laxation does not occur.

    Although  a  great  deal has  been  said  about the  need to
evaluate  food additives and  contaminants for the  induction of
possible  behavioural changes, JECFA has hitherto devoted little
time  to evaluating such changes.   Since it has been  suggested
that  certain food constituents can  produce behavioural changes
in  man, JECFA will, in the future, undoubtedly have to consider
such effects.  Unfortunately, good animal models have  not  been
developed, and objective human data are difficult to obtain.  It
is  not possible to recommend  a simple series of  tests at this
time, primarily because there is no clear consensus on the kinds
of studies that should be performed nor on the interpretation of
the results.

    Intolerance  to food additives should always be considered a
possibility, even though tests for reactions to  food  additives
are  not part of  the normal data  package that JECFA  considers
when  assessing  new  food  additives.   Even  when  evidence of
widespread  intolerance to a food additive appears, it may still
be  very difficult to determine a cause-and-effect relationship.
Problems  include  the often-anecdotal  nature  of much  of  the
evidence, psychological factors, and problems with developing an
adequate  central  data  collection system.   These  points  are
discussed in more detail in section 5.4.

5.1.2.  Non-neoplastic lesions with morphological manifestations

    A number of lesions are relatively frequently observed in  in
 vivo   studies, particularly in rodents, that often give rise to
controversy.

    Non-specific  liver  enlargement  or  hypertrophy  has  been
discussed  by  a WHO  scientific group (2,  pp. 18-19).  In  the
past,  this occurrence has been considered to be a manifestation
of  toxicity.  More recently, it has been realized that this can

often  be a physiological  response involving the  induction  of
microsomal  enzymes  in  the  detoxification  process  that   is
reversible on removal of the test compound.

    The  formation  of  calculi in  the  urinary  bladder  is  a
frequent  finding in rodent  studies.  Often, the  formation  of
calculi may be followed by the formation of bladder tumours.  It
is not uncommon to find calculus formation in one rodent species
and  not in another.  Under  these circumstances, the nature  of
the  calculi can sometimes be associated with specific metabolic
changes  that have led to  their formation.  This, in  turn, may
allow for a scientifically-based extrapolation to man, providing
that human clinical studies are possible.

    Caecal enlargement, a common finding in rodent studies, is a
normal  finding  in  rodents maintained  on  standard laboratory
diets  under germ-free conditions.  It is also a common response
of  rodents, especially rats, to diets that include non-nutrient
substances  (e.g., certain permitted food colours and saccharin)
or certain nutrients in excessive concentrations (e.g., modified
starches, plant gums, lactose, and various polyols).

    Most of the enlargement is attributable to increased luminal
contents;  in addition,  the weight  of the  caecal  wall  after
washing out the luminal contents is usually marginally more than
normal.  In haematoxylin- and eosin-stained sections, the caecal
wall shows no remarkable features, and there is no evidence that
caecal  enlargement predisposes to any  form of neoplasm of  the
caecum.  Caecal enlargement may be due to osmotic  effects,  but
its  mechanism is not  well understood.  In  some cases,  caecal
enlargement  is an incidental  finding, with the  primary effect
being  nephrocalcinosis  (23,  pp.  11-12).   Various  forms  of
mineral deposition occur in the kidneys of  laboratory  rodents,
more  commonly  in  rats  than  in  mice  or  hamsters.   Unless
appropriate  diagnostic staining or chemical analysis is carried
out, it is not strictly justifiable to refer to these changes as
renal "calcinosis", though most of the mineral deposits  do,  in
fact,  contain  calcium  in  one  form  or   another.    Mineral
deposition  can take place in almost any part of the nephron and
deposition may predominate in one or more areas of  the  kidney.
The  main forms of  renal mineralization are  basement  membrane
mineralization,  corticomedullary nephrocalcinosis, pelvic neph-
rocalcinosis, and nephrocalcinosis associated with acute tubular
nephrosis.   All these forms  of nephrocalcinosis may  co-exist,
and one and the same agent may cause more than one form of neph-
rocalcinosis.

    Magnesium  deficiency in standard laboratory  diets undoubt-
edly contributes to the high incidence of corticomedullary neph-
rocalcinosis  in rats.  Excessive dietary phosphate and possibly
excessive  dietary calcium may  predispose to pelvic  nephrocal-
cinosis.   Such observations lead  to the conclusion  that  more
attention needs to be paid in the future to the  formulation  of
diets  for  rodents  with  respect  to  physiologically-relevant
levels of calcium, magnesium, and phosphorus.

    Testicular  atrophy, which is sometimes observed in rodents,
may occur as a result of reduced caloric intake, frequently as a
result  of  the  addition of  an  unpalatable  chemical  to  the
animal's  food or drinking-water.  This  should be distinguished
from  testicular atrophy resulting from the direct action of the
chemical  on  the testicular  cells.   This distinction  can  be
achieved by undertaking paired feeding studies.  It is important
that  the function of the testes be investigated in reproduction
studies when atrophy is detected.

    Manifestations  of vitamin deficiencies, notably of the fat-
soluble  vitamins, are sometimes  observed in studies  on agents
that  may be fat  solvents and are  only partly absorbed  in the
gastrointestinal  tract;  an  example  of  such  a  substance is
mineral  oil.  Another effect that is sometimes observed is dis-
colouration of mesenteric lymph nodes after feeding  a  coloured
substance.  This is a normal physiological response  and  should
not  be considered a toxic end-point, as long as it is not asso-
ciated with proliferative reactions.

    Hormonally-associated  effects occur with  certain additives
and  require  special  endocrinological evaluations.   Recently,
JECFA  has  been faced  with the task  of evaluating the  use of
certain  anabolic agents used  in the raising  of meat-producing
animals (22, 23).  These agents result in the presence  of  low-
level  residues  of  hormonally-active compounds  in  meat.  The
evaluation  of  the potential  toxic  effects of  such compounds
requires  knowledge  of  the levels  of naturally-occurring com-
pounds with similar effects (Annex III).

5.1.3.  Neoplasms

    The  most difficult decisions facing  the toxicologist arise
from  the varied end-points found in long-term  in vivo carcino-
genesis studies.  These problems have existed for a  long  time,
but  they have been greatly  exacerbated by the large  number of
effects  observed  in  the many  rote  tests  now  performed  on
chemicals including many that may be present in the food supply.
These  chemicals  include  certain direct  food  additives, some
processing   and  carrier  solvents,  components   of  packaging
materials, and a variety of contaminants.

    Oncologists  now  generally  recognize that  different mech-
anisms  of carcinogenesis exist with  different chemicals acting
on different tissues in the body (45, 46).  Many believe that it
may  be possible to determine tolerance levels for some carcino-
gens,  though this is still not possible with any degree of cer-
tainty with the majority of them.  The view that a tolerance may
be set for carcinogens giving rise to tumours through  either  a
hormonal  mechanism or by the  formation of bladder calculi  has
been expressed by a WHO scientific working group (3, p. 11).

    The perception that a chemical for which there  is  evidence
of "carcinogenicity" in any system should not be  permitted  for
use as a food additive at any dose whatsoever has  become  wide-
spread.   Although this philosophy has never been promulgated or

officially adopted by JECFA, it has, in practice, influenced the
Committee's  approach to decision making.  Probably more experi-
mental work has been undertaken in cancer research over the past
two  decades, since this view was first established, than in all
the  preceding  years,  and clearly  there  is  a great  need to
clarify the issue in terms of practical interpretation.

    The  assessment of the  evidence for the  carcinogenicity of
chemicals  is a major issue  to be resolved by  JECFA.  Not only
have  many  chemicals  been tested  by  a  variety of  routes of
administration  that may not  be relevant to  food additive  use
(such  as the repeated injection of food colours and other addi-
tives  in rats and mice  with consequent development of  subcut-
aneous  sarcomas at the site  of injection (27)), but,  in addi-
tion, new end-points are continually revealed and interpretation
becomes more confusing as studies become more and more detailed.
Positive results may be obtained due, for example, to a carcino-
genic  impurity.  The extrapolation of such data has become very
complex.  One possibility, to make the term  "carcinogen"   more
generalized,  clearly would not solve the problem of how best to
interpret these data.

    Much of this issue centres around the meaning of the various
types  of enhancement of the  tumours that occur in  the rodents
used for  in vivo  bioassays, since the rodents in common use, the
mouse  and  the  rat, develop  extremely  high  incidences of  a
variety of tumours in the untreated state.  Many  reports  indi-
cate  that one or more  of these tumours has  an increased inci-
dence  or has appeared  earlier (or both)  in treated,  compared
with  untreated animals.  One problem in interpreting the signi-
ficance of these tumours is the difficulty in  deciding  whether
these  naturally-occurring tumours are spontaneously  induced or
whether the agent is able to induce them.  The problem  is  fur-
ther confounded by the fact that the incidence of tumours in the
untreated control animals varies considerably with time.   As  a
result, there is now a debate as to the importance of historical
as well as concurrent control animals.  It appears without doubt
that both such controls are of importance (especially  when  the
historical control data come from the same laboratory, using the
same standardized diet, and do not go back in history  beyond  5
years  of the study under consideration)  and that, if the chem-
ical  in question has only enhanced the incidence of a commonly-
seen  tumour to a  level seen in  historical controls, then  the
level of concern will be much less than would otherwise  be  the
case.

    The  evaluation of studies in which these commonly-occurring
tumours  are  a  complicating  factor  need  careful  individual
assessment.   The  tumours  that have  given  rise  to the  most
controversy in recent years are hepatomas (particularly  in  the
mouse),  pheochromocytomas in the rat (see below), lymphomas and
lung  adenomas  in the  mouse,  pancreatic adenomas  and gastric
papillomas in the rat, and certain endocrine-associated tumours,
including  pituitary, mammary, and thyroid tumours, in both rats
and  mice.  Some of  these tumours, such  as hepatomas and  lung
adenomas, may occur in the majority of untreated animals.

    With the exception of lymphomas, some of which  are  virally
associated,  the  endocrine-associated  tumours,  and   possibly
hepatomas in high-incidence strains of mice, which  may  involve
oncogenes  (47), there is  no clue as  to the origin  of tumours
that  occur  commonly  in experimentally-used  rodents.  Indeed,
there is not even any cogent speculation as to the mechanisms by
which these tumour incidences are increased.

    Adrenal  medullary lesions in rats provide a good example of
the  problems  encountered  in interpreting  the significance of
high tumour incidences.  An overview of the literature indicates
that untreated rats of various strains may exhibit  widely  dif-
fering  incidences  of lesions  described as "pheochromocytomas"
(24,  48, 49).  There are  no clear criteria for  distinguishing
between  prominent foci of hyperplasia and benign neoplasms, and
pathologists differ in the criteria that they use  for  disting-
uishing between benign and malignant adrenal medullary tumours.

    Rats  fed  ad libitum  on  highly nutritious  diets  tend  to
develop  a  wide  variety  of  neoplasms,  particularly  of  the
endocrine  glands,  in  much  higher  incidences  than   animals
provided  with enough food to  meet their nutritional needs  but
not  enough to render them  obese.  The adrenal medulla  is just
one of the sites affected by overfeeding.   Controlled  feeding,
especially early in life, reduces the life-time  expectation  of
developing  either  hyperplasia  or  neoplasia  of  the  adrenal
medulla in rats.

    A  complicating factor in assessing  carcinogenicity studies
is  the question of how  to consider benign tumours.   If benign
and malignant tumours are observed in an animal tissue and there
is  evidence of  progression from  the benign  to the  malignant
state, then it is appropriate to combine the tumour types before
performing  statistical  analysis.   Assessment of  the relative
numbers  of benign  and malignant  tumours at  the various  dose
levels  in the study can help determine the dose response of the
animal to the compound under test.  On the other hand,  if  only
benign tumours are observed and there is no indication that they
progress  to malignancy, then, in  most cases, it is  not appro-
priate  to consider the compound to be a frank carcinogen, under
the  conditions of the  test (this finding  may suggest  further
study).   Often, how benign tumours should be considered is much
less clear.  Some clarification can be achieved  by  classifying
and analysing tumours on the basis of their  histogenic  origin.
This  is helpful, not only  for determining the significance  of
benign  tumours,  but  also for  preventing  different malignant
tumours  occurring in the same organ from being grouped together
for  statistical  analysis.   For further  discussion  of  these
points, see (50), pp. 226-230.

    The  results of statistical analyses are often misunderstood
and misused.  An effect may be statistically significant but not
be  of any biological  significance, because the  animal's well-
being  is not affected by its occurrence.  On the other hand, an
event that is of biological significance, such as the occurrence
of  one  or  two tumours of a very rare type in treated animals,

may  not be  significant by  the usual  battery  of  statistical
tests.   This  difference  between  biological  and  statistical
significance  underscores  the  need for  critical  analysis  of
statistical  results  rather than  the  blind acceptance  of the
numbers  obtained.  The statistical  aspects of the  design  and
interpretation of toxicity studies are discussed in Annex II.

    Generally, it is becoming increasingly difficult to identify
a  substance as a  carcinogen with confidence.   In  particular,
when  animals with a  high background incidence  of tumours  are
involved,  it is extremely  difficult to know  when to draw  the
line  between a result  that indicates that  a compound that  is
potentially  hazardous for man has been discovered compared with
a  compound  that  is  merely  an  experimental  curiosity.  The
International  Agency  for  Research on  Cancer  (IARC)  reviews
evidence  for the carcinogenicity  of chemicals on  a continuous
basis  and  drafts  monographs  on  many  groups  of substances.
However,  faced  with  the difficulty  of  separating  different
levels   of  "carcinogenicity",  IARC  working   groups  on  the
Evaluation  of  the Carcinogenic  Risk  of Chemicals  to  Humans
designate  compounds as being  possessed of either  "limited" or
"sufficient"  evidence  for carcinogenicity.   Given the natural
desire  of the food additive  toxicologist to be as  cautious as
possible,  this  terminology is  not  very practical.   After  a
compound  has  been  designated as  possessing "limited carcino-
genicity",  it is very difficult from a regulatory point of view
to approve it as a food additive, even if extensive further work
on the compound shows it to be safe at expected levels  of  con-
sumption with no other evidence of carcinogenicity.

    The  decisions  that  are being  made  on  the basis  of the
present  state  of  knowledge  of  carcinogenesis,  may,  in the
future,  prove to have been  excessively conservative.  However,
it  is now possible to make certain reasoned decisions, provided
that  each  instance where  carcinogenicity  is the  problem  is
examined  individually and all  relevant factors are  taken into
account.

5.1.4.  Reproduction/developmental toxicity

    Most food additives are consumed by men and women during the
reproductive stages of their lives and by pregnant and lactating
women.  Some food additives are also consumed by infants.  Thus,
a  thorough safety evaluation requires  that the effects of  the
substance  on  reproductive  performance  and  development  from
fertilization  through  weaning  be studied.   JECFA recognizes,
however, that it is unrealistic to expect that such  studies  be
performed in all cases (sections 5.3.1 and 5.3.4).

    Adverse  effects  on  reproduction may  be expressed through
reduced  fertility or sterility  in either the  parents or  off-
spring  due to morphological, biochemical, or physiological dis-
turbances.   Adverse  effects  on development  may  be expressed
through  structural  or  functional abnormalities  due to either
mutations  or  to  biochemical  or  physiological  disturbances.
Mutations  may occur in either somatic or germ cells.  Mutations

in  male or  female germ  cells represent  potentially the  most
long-lasting and severe effects on the human population  that  a
chemical could cause.

    Adverse  effects on reproduction  or development induced  by
chemicals may be expressed immediately or they may  be  delayed,
sometimes  for  many  years, as  exemplified  by  transplacental
carcinogens (section 5.3.1.2).

    Structural  or functional abnormalities  are most likely  to
develop  during embryogenesis, the period  of development during
which  cells  differentiate  into  the  various  organ  systems.
Typical  teratogenicity studies investigate the effects of expo-
sure  to test  substances during  this period.   Effects due  to
exposure  during fetogenesis, the developmental period after the
organ  systems have formed, generally involve growth retardation
and  functional disorders, though the external genitalia and the
central  nervous system are  also susceptible to  injury  during
this  period (51, 52).   Such structural or  functional  abnorm-
alities  often do not become obvious until some time after birth
and, in some cases, not until adulthood.

    Neonatal development may be influenced by the consumption of
milk  containing  chemicals  (or their  metabolites)  that  were
consumed  by  the  mother.   Agents  may  also  affect  neonatal
development by influencing maternal behaviour, hormonal balance,
or  nutrition.  Direct neonatal exposure to xenobiotic compounds
also occurs, but is less common, since JECFA considers  it  pru-
dent that food intended for infants younger than 12 weeks of age
should not contain any additives (42).

    Guidelines  for  reproductive  toxicity investigations  have
been  developed by various legislative and international organi-
zations, including the US Food and Drug Administration (US FDA),
US  Environmental Protection Agency (US EPA), the United Kingdom
Committee  on Safety of  Medicines (CSM), Committee  on Toxicity
(COT),  and  Pesticides  Safety Precautions  Scheme  (PSPS), the
Japan  Ministry  of  Agriculture, Forestry  and Fisheries (MAFF)
and  Ministry of  Health and  Welfare (MHW),  the  World  Health
Organization  (WHO),  the Organization  for Economic Cooperation
and  Development (OECD)  (all listed  in (51))  and  the  Inter-
national  Programme on Chemical Safety (IPCS) (53).  A review of
the methodology for assessing the effects of chemicals on repro-
ductive  function has been published  under the auspices of  the
IPCS and the Scientific Committee on Problems of the Environment
(SCOPE)  of the International Council of Scientific Unions (54).
The procedures described in these publications are  designed  to
assess  the reproductive and developmental toxicity potential of
test compounds using lower mammals as model systems.  These pro-
cedures generally involve combining various stages of  the  life
cycle in one test, as it is usually not practicable  to  examine
the effects of a chemical in each separate stage of  the  repro-
ductive  cycle.  An exception is  the so-called "teratogenicity"
study,  where exposure is limited to the period of organogenesis
(see below).

    The  goal of reproduction/developmental toxicity  studies is
to  assess whether the organism  is more sensitive to  the agent
under test during its reproductive and developmental stages than
during  its adult phase.   Therefore, the highest  dose of  food
chemical that is administered is generally the amount that would
be  expected to cause slight  maternal toxicity, and the  lowest
dose  is  the  amount that is not expected to cause an effect in
either  the  mother or the  conceptus.  If profound toxicity  is
observed in the offspring at the high dose (the dose that causes
only  slight maternal toxicity),  then the conclusion  would  be
that  the substance is more toxic for offspring than for adults.
This conclusion would be reinforced by the appearance of adverse
effects in the conceptus at the mid- and/or low-dose levels.  On
the  other hand, if the  test substance injures reproduction  or
development at levels comparable with levels that cause toxicity
in  adults, then no  special concern should  be attached to  the
results of the reproduction/developmental toxicity studies.

    Single-generation  and multi-generation reproduction studies
are  useful  for  assessing potentially  deleterious  effects on
reproduction  and development through parturition and lactation.
However,  because of the  long-term exposure inherent  in  these
studies,  detoxifying enzymes may  be induced in  mothers before
embryogenesis  takes  place.   Under  these  circumstances,  the
observed  toxicity  would  be  understated.   Studies  in  which
mothers  are exposed to the  test substance only during  organo-
genesis, as in teratogenicity studies, reduce the possibility of
the mother adapting to the test compound.

    The  range  of effects  arising  from maternal  exposure  to
chemicals during organogenesis includes:

    (a)  death and resorption of the embryo;

    (b)  teratogenic  defects  (malformations  of  a  structural
         nature);

    (c)  growth  retardation  or specific  developmental delays;
         and

    (d)  decreased postnatal functional capabilities (55).

Which  of these effects will  be expressed depends on  the level
and gestational timing of the dosage of the food  chemical,  and
the duration of the period of treatment (56).  Thus, a substance
given at one dose level may result in growth retardation, while,
at  a higher level, it may result in death and resorption of the
embryo.  Sometimes, the slope of the dose-response curve of, and
between,  these effects is very steep, making the interpretation
of  the studies very difficult.   Because all of these  outcomes
are  unacceptable, the most important  consideration when evalu-
ating  these studies should not be which effect is observed, but
rather,  at what dose level  the adverse effect became  evident.
This dosage information can then be used to set exposure limits.
Because  teratogenic effects  are only  one part  of  the  total

spectrum  of the embryotoxic effects that should be investigated
in  such studies, a better term for "teratogenic studies"  might
be  "embryotoxicity studies".  In those rare situations when the
studies  are performed during  the period of  fetal development,
the term "fetotoxicity studies" should be used.

    The  appropriate  role  of  studies  involving  in utero  and
neonatal  exposure  in  the  evaluation  of  food  additives  is
discussed in section 5.3.

5.1.5.   In vitro  studies

    In  recent years, a great  deal of effort has  gone into the
development  of  in vitro  test systems. Generally, these systems
are  segregated according  to two  kinds of  functions:  (a)  to
reveal  whether a particular kind of toxicity is produced by the
agent  under study; or  (b) to help  elucidate the mechanism  of
toxicity  displayed by a chemical.   The former tests are  being
developed  to  serve both  as  predictors of  toxicity  (section
3.1.2)  and  as  substitutes for  complex, lengthy  in vivo  pro-
cedures.   The latter are more directly focused than the former,
and  their  value has  been clearly demonstrated  as a means  of
establishing the metabolic mechanisms at the organ,  tissue,  or
cellular level (section 5.2).

    Much effort has been devoted to the development of  in vitro
test systems based on isolated cells, tissues, and organs.  Some
of  these systems are reported  to be related to  specific toxic
end-points  such as mutagenicity and  carcinogenicity (e.g., DNA
damage and repair in mammalian cells, covalent binding  to  DNA,
cell  transformation, mitotic recombination, and gene conversion
in  yeast  (57))  and  to  embryotoxicity  (e.g.,  whole  embryo
cultures,  cultures of embryonic tissues, teratocarcinoma cells,
and embyronated eggs (58)).

    The number, diversity, and use of these tests have increased
rapidly  in  the  past decade  and  are  likely to  continue  to
increase  in the future.  However, correlations among results of
various  in vitro  tests  and  reported correlations  between the
results of batteries of short-term  in vitro  tests  and  in vivo
carcinogenesis  bioassays (which have been the primary thrust of
these assays)  are not high. Such short-term  in vitro  tests are
generally  effective  at  measuring their  intended genetic end-
point,  i.e., mutagenicity in the particular system under study.
However,  the relevance of  mutagenic effects to  food  additive
toxicity  has not  been established,  and the  results  of  many
current  in vitro  test  procedures  do  not  relate  to  genetic
effects  in mammalian reproductive tissues.  Neither is it clear
how well these tests identify chemical carcinogens or  how  they
should  be used in the absence of corroborative data on carcino-
genicity.

    In a similar fashion, culture techniques designed to measure
prenatal  toxicity are extremely  useful for research  purposes,
but,  at their present stage  of development, they are  not very
suitable  for  screening  (58).   By  excluding  the   maternal-

placental-fetal  relationship, such dissected systems permit the
compound  to reach the  target directly (membrane  systems  that
provide  biological barriers are missing) without permitting the
potential  moderating or activating  influences of the  maternal
tissues.

    Attention  should be paid  to scientific developments  in  in
 vitro  test systems.  However, because of  the many experimental
uncertainties  and controversial issues surrounding the efficacy
of  these tests as predictors  of specific toxic end-points,  it
would  be  inappropriate  for JECFA  to  request  that all  food
additives brought before it should be subjected to such tests on
a  systematic basis.  On the  other hand, data obtained  with  in
 vitro  systems sometimes help to clarify the mechanism of action
of  chemicals  observed in  in vivo  systems.   Therefore,  JECFA
should  continue  to  determine the  relevance  of  available  in
 vitro  data on an  ad hoc  basis  when  assessing the  safety  of
specific compounds.

5.2.  The Use of Metabolic and Pharmacokinetic Studies in Safety Assessment

    Chemical   toxicity  results  from  reactions   between  the
ingested  toxic chemical, or its  metabolites, with constituents
of  the body.  Therefore,  the complete safety  evaluation of  a
substance such as a food additive must consider  its  metabolism
and pharmacokinetics.  Unfortunately, a great deal more has been
said  and  recommended  in  this  area  than  has been  done  in
practice.   The importance of metabolic and pharmacokinetic data
in  the proper planning  and interpretation of  in vivo  toxicity
testing  of chemicals is obvious, but the fact is that such data
are either inadequate or unavailable to aid in  the  interpreta-
tion  of the majority of long-term studies undertaken with chem-
icals, including food additives.

    Detailed  metabolic studies have gained  added importance in
determining  the  extent  of appropriate  toxicological  testing
since  the  advent of  novel and modified  foods.  This is  con-
sidered in detail in section 6.2.  However, it is  necessary  to
repeat some general aspects of the subject here.

    Biochemical  studies play two  separate roles in  the safety
evaluation of chemicals.  These are:

    (a)  to design animal studies by identifying the appropriate
         species  for, and helping to  determine the appropriate
         level of, testing; and

    (b)  to  extrapolate  experimental  animal toxicity  data to
         human  beings, by elucidating the mechanism of toxicity
         of the chemical, thus facilitating the establishment of
         a no-observed-effect level; a comparison of biochemical
         data   between  experimental  animals  and   man  helps
         determine  the  relevance  of any  toxicity observed in
         animals.

    The  ingested chemical itself may exert a toxic effect, or a
metabolite(s)   may  be  the  toxic  agent.   Many  polar,  non-
lipophilic  chemicals  are rapidly  metabolized and/or excreted,
while  lipophilic  compounds may  be  stored, excreted  into the
bile,  or metabolized into more  polar, water-soluble compounds,
which are eliminated from the body, in the urine,  more  rapidly
than the ingested additive.

    Absorbed  substances, except those that  enter the lymphatic
system,  are transported directly  to the liver  via the  portal
vein.   Many substances that  are metabolized in  the liver  are
transported via the hepatic vein to the kidneys to  be  excreted
in  the  urine.   Through enterohepatic  circulation,  some sub-
stances that are conjugated in the liver are excreted  with  the
bile,  reabsorbed, and then  excreted once again  in either  the
bile or the urine.

    Metabolism, primarily involving enzymatic reactions, may:

    (a)  convert  the  additive into  a  body constituent  or  a
         source of energy;

    (b)  lead to the detoxification of the ingested chemical and
         the excretion of its metabolites; or

    (c)  result  in  activation  of the  chemical  into reactive
         intermediates  that  then  react most  importantly with
         glutathione, tissue proteins, RNA, or DNA.

Biotransformation  reactions are catalysed by  intra- and extra-
cellular enzymes and by enzymes of the microflora of the gastro-
intestinal tract.  Knowledge of the rates of formation, reaction
with  tissue components, and excretion of various metabolites is
essential  for full understanding of the disposition and elimin-
ation  of the chemical  from the body  and of the  mechanism and
extent of its toxicity.

    This  section contains a general  discussion of the role  of
metabolism  and pharmacokinetic data in the safety assessment of
food additives.  Simple guidelines have not been  generated,  as
it is doubtful that such guidelines are feasible  or  desirable.
Several food additives that have been extensively  studied  bio-
chemically  are  discussed  in  Annex  IV.   These  examples are
designed  to provide an appreciation  of the value and  problems
involved  with investigating the  metabolic bases for  the  bio-
logical responses to food additives.

5.2.1.  Identifying relevant animal species

    The  occurrence of interspecies  differences in response  to
foreign   compounds  complicates  the  extrapolation  of  animal
toxicity data to human beings.  The resolution of  this  problem
depends on an understanding of such interspecies  variations  in
the  disposition  of  ingested  compounds.   In  this   context,
disposition  is  meant  to encompass  metabolism  and  pharmaco-
kinetics.

    The  rates of absorption,  rates and sites  of distribution,
and rates and routes of excretion determine  the  concentration-
time  profiles of  the parent  molecule and  metabolites in  the
various  tissues and organs of the body.  The overall biological
response  is  thus the  product of the  fluxes of the  unchanged
molecule  and  its metabolites  occurring  in the  animal  under
examination.   Definition of the pharmacokinetic properties of a
food additive may require various routes of administration.  The
influence  of any vehicle to be used in long-term studies should
be determined, because the vehicle may influence the absorption,
metabolism, or toxicity of the test compound.

    In  order to extrapolate reliably  from animals to man,  the
ideal situation would be one in which the tissues of the animals
and  of man would be exposed to identical fluxes of the compound
and  its  metabolites.   This  requires  that  the  qualitative,
quantitative,  and  kinetic aspects  of  the disposition  of the
compound be the same in animals and man.  This  ideal  situation
is probably never achieved because of species variations.

    The goal should be to select a species for testing  that  is
the  most closely related to  man in terms of  the metabolism of
the  compound  under  study,  using  a  route  of administration
similar to the anticipated human exposure.  However, the list of
species  that may realistically  be used in  a toxicity test  is
very  limited  because  of  problems  of  availability,  lack of
background   pathological   and  physiological   knowledge,  and
experimental  convenience.  Thus, it is unlikely that a suitable
metabolic  model species will  fulfill other important  criteria
used  to  select test  species.   Given these  facts,  metabolic
studies  used  for  species selection  should  be  prospectively
performed  only on species  suitable for toxicity  testing.  The
species is then selected that is closest to the human  being  in
terms  of the  metabolism of  the compound.   This,  of  course,
requires  knowledge about its metabolism in the human being.  In
many cases,  in vivo  human studies are not possible.

    In  general, the required metabolic  and kinetic information
can best be obtained from  in vivo  studies. Pharmacokinetics can
obviously  only  be  examined  in vivo, since  these studies deal
with  whole animal phenomena.   In vitro  studies, such  as organ
perfusions  and tissue cell incubations,  provide useful inform-
ation  in some cases, but they do not provide information on the
absorption,  distribution, and excretion of  chemicals.  The use
of  isolated cells of human tissues may prove acceptable in some
cases,  because,  even  considering their  inherent limitations,
these systems may be the only ones available for  examining  the
metabolism  of compounds that  cannot be administered  to  human
subjects.

    In  a long-term toxicity  test, the attainment  of a  steady
state  depends on the relationship between the kinetic variables
of  the compound  and the  dose interval.   When a  compound  is
administered  continuously in the  diet, an approximate  "steady
state" will sometimes be established.  Therefore, marked differ-
ences seen in animals in single-dose pharmacokinetic studies may

change,  or even disappear with  long-term administration.  How-
ever,  a steady state  will never truly  be achieved because  of
diurnal patterns of dietary intake by common laboratory animals.
On  the other hand, compounds that are rapidly absorbed and have
a short half-life will show wide temporal variations  in  plasma
concentrations,  as  will  compounds that  are  administered  by
gavage or capsule.

    In  many  cases,  the metabolic  profile  of  a compound  is
determined by the amount administered, as well as the species in
question.   The use of very  high doses in toxicity  testing may
give  metabolic  patterns,  and therefore  biological responses,
that  are  patently  unrepresentative  of  the  situation  to be
expected  with actual  levels of  exposure.  Thus,  data on  the
influence  of dose level on metabolism in the test animal should
be  generated  to  determine whether  absorptive,  metabolic, or
excretory  processes  may have  a  threshold.  An  animal model,
apparently  suitable  at  one level  of  exposure,  may be  less
suitable at a different level.

    Consideration  should be given  to the possibility  that the
metabolism of the compound will differ between  long-term  tests
and  short-term  metabolic studies.   This  could be  because of
adaptation  by  gut microflora,  which  is discussed  in section
5.2.4,  or  because  of the  induction  of  enzyme systems  that
metabolize the substance.

    An aim of toxicity testing should be to examine the possible
activities  in animals  of all  of the  human metabolites  of  a
compound  that may induce toxicity.  In many cases, this is best
achieved  by  combining data  from  several animal  species,  to
include all the metabolites of interest.  In  interpreting  such
data, the closest attention should be paid to  the  similarities
of the mechanisms of toxicity in the various animal species, and
also to the possibility that toxicity may  involve  interactions
between the parent compound and its metabolite(s), which may not
be the same in all species and may be irrelevant for man.

5.2.2.  Determining the mechanisms of toxicity

    A great deal of research has been performed in an attempt to
explain  the mechanism(s) by  which certain test  chemicals have
given  rise to particular  lesions.  In the  majority of  cases,
these  studies  have  concerned  the  development  of   tumours.
Clearly,  it is likely to  be most difficult to  find any simple
mechanistic  answer to carcinogenesis.   However, it is  usually
possible  to  determine  factors  of  importance  for  a  safety
evaluation,  which are not necessarily complete solutions to the
mechanism of action.

    If  an  additive  has  been  determined  to  be  an   animal
carcinogen,  it is extremely difficult  to show that it  is safe
for  human beings.  It is not easy to establish its safety, with
retrospective  metabolic  and  pharmacokinetic studies,  that an
exclusively   secondary  mechanism  is  operative   and  that  a
threshold  exists below which the  use of the additive  is safe.

If  a  carcinogenic  impurity is  present  in  an additive,  the
impurity should be either removed or limited to such  an  extent
that  consumption of the additive  does not pose a  carcinogenic
risk  for consumers.   The level  of the  impurity in  the  food
additive should be controlled within specifications.

    In  some cases, toxicity may  occur as a result  of the test
compound  or a metabolite displacing  endogenous substrates from
carrier proteins or receptor sites.  Where such  mechanisms  are
indicated,  ad hoc  studies of  relative  binding affinities  can
form  a useful adjunct  to routine pharmacokinetic  studies  and
assist  in  establishing  safe   levels  of  exposure.   In vitro
studies  are useful for  determining the mechanism  of toxicity,
especially  when covalent binding to  cellular macromolecules is
involved.   In  such  cases, comparative  binding  studies using
preparations of metabolizing enzymes from various animal species
are desirable.

5.2.3.  Metabolism into normal body constituents

    The  metabolism  of an  ingested  compound into  normal body
constituents  does not provide  assurance that the  substance is
safe.  Not only are many metabolites toxic (e.g., most excretory
products), but there are limits to the body's ability to process
even  relatively non-toxic metabolites.  These  limits should be
known,  and,  if  a toxic  threshold  for  a substance  has been
identified,  its acceptance as  a food additive  will depend  on
controlling  intake so  that toxic  levels are  not ingested  by
human beings.

    Knowledge  that a substance  is a natural  metabolite or  is
metabolized  into normal body constituents  is of great help  in
evaluating its safety.  However, without concomitant information
on  the  kinetics  of  the  production  and   disappearance   of
metabolites, the extent to which such information can be used is
limited.

    When  the additive provides only a small increment in levels
of metabolites compared with the ordinary consumption  of  food,
then  the questions about  safety are greatly  simplified.   The
report  of the WHO  Scientific Group on  Procedures for  Invest-
igating  Intentional and Unintentional Food  Additives (2, p. 7)
considered this situation and concluded that:

    "if  the biochemical evidence  shows that the  additive
    makes  only a small contribution  to existing metabolic
    pools from food components or in the tissues, there may
    be  no need for  detailed toxicological studies  on it,
    provided that it conforms to adequate specifications."

JECFA  has also considered  this situation in  some detail  (29,
pp.12-13) and has suggested that:

    "any  food additive that  is completely broken  down in
    the food or in the gastrointestinal tract to substances

    that are common dietary or body constituents  might  be
    satisfactorily  evaluated. . .  on the basis  of appro-
    priate biochemical and metabolic studies alone. . .".

This  report summarizes the evidence  required in such cases  as
follows:

    (a)  "evidence  that the substance is readily broken down in
         the  food or in  the gastrointestinal tract  to  common
         food constituents under the conditions of use;

    (b)  evidence to indicate the main factors concerned in this
         breakdown, e.g, pH and enzymes;

    (c)  evidence,  preferably  including studies  on human sub-
         jects,  that  the  material,  when  given  in  moderate
         amounts and under conditions similar to those that will
         prevail  if used as a food additive, is absorbed to the
         same  extent as the  food materials to  which it  gives
         rise,  and does not  interfere with the  absorption  of
         other nutrients;

    (d)  evidence  that  unhydrolysed or  partly hydrolysed mat-
         erial  does  not occur  in  significant amounts  in the
         stools,  and that it does not cumulate in body tissues;
         and

    (e)  evidence that the most important food components in the
         additive  are  metabolized and  utilized as effectively
         when  administered  in  composite form  as  when  given
         separately, and that overloading does not occur."

    As  long  as  adequate evidence  along  the  above lines  is
presented, the Committee concluded that:

    "the  food additive is  handled in the  body  in a  way
    that  is not significantly different from that required
    for  the component food  materials.  If so,  no toxico-
    logical studies need be demanded, since the problem now
    becomes  one  involving  the toxicology  of foods them-
    selves rather than the toxicology of a food additive."

The Committee allocated ADIs to such substances,  calculated  on
the  basis that the  food additive would  not increase the  food
component  into which it is  converted by more than  about 5% of
the quantity in an average diet (29, p. 13).

    These  general principles have been accepted by the eleventh
and  seventeenth meetings of JECFA (41, pp. 8-9; 16, p. 31), the
second  of which confirmed  that, if biochemical  evidence shows
that  the  sole  effect of  the  additive  is to  make  a  small
contribution  to existing metabolic loads  from food components,
there is no need for detailed toxicological  studies.   Examples
cited  in the various  reports include sucrose  esters of  fatty
acids,  lactic and fatty acid ester of glycerol, and some esters
used as food flavours.

    These  principles are still  valid.  However, in  some JECFA
reports, the combined evidence of breakdown in food and  in  the
gastrointestinal  tract was considered.  In contrast, studies on
the  stability/breakdown pathways of the additive in food, under
the  proposed conditions of  use, may be  needed to ensure  that
significant quantities of toxic products are not  formed  during
food processing or storage, either through transformation of the
additive  or through its  reaction with food  constituents.  All
procedures  designed to measure metabolites must be accurate and
they  must have a  high level of  sensitivity for the  compounds
under  consideration,  to  draw  the  conclusion  that   further
toxicity studies are not necessary.

5.2.4.  Influence of the gut microflora in safety assessment

    The  gut microflora may  influence the outcome  of  toxicity
tests  in  a  number of  ways,  reflecting  their importance  in
relation  to the nutritional status  of the host animal,  to the
metabolism  of  xenobiotics  prior  to  absorption,  and  to the
hydrolysis  of biliary conjugation  products.  JECFA has  recog-
nized this, and has drawn attention to the usefulness of studies
on  metabolism, involving the intestinal  microflora, in toxico-
logical evaluation (30, p. 7; 18, p. 10).

    Interactions that may occur between food additives  and  the
bacterial  flora of the  gastrointestinal tract should  be  con-
sidered  both in terms of  the effects of the  gut microflora on
the chemical and the effects of the chemical on the  gut  micro-
flora.   Because the gut microflora  are important in the  meta-
bolic  fate and toxicological  activity of some  food additives,
the  safety  assessment of  food  additives should  include  the
possibility  that gut microflora modify the host response to the
food  additive and/or that  the food additive  is affecting  the
host microflora.

5.2.4.1     Effects of the gut microflora on the chemical

    The  spectrum of metabolic activity  shown by the gut  flora
contrasts markedly with that of the host tissues.  While hepatic
metabolism  of foreign compounds  is predominantly by  oxidation
and  conjugation  reactions,  the gut  bacteria  perform largely
reductive and hydrolytic reactions, some of which appear  to  be
unique to the gut flora.  Typical reactions include:

    (a)  the  hydrolysis  of  glycosides (including  glucuronide
         conjugates), amides, sulfates, and sulfamates;

    (b)  the  reduction of double  bonds and functional  groups;
         and

    (c)  the  removal of functional  groups such as  phenol  and
         carboxylic acid moieties.

Thus, from a structural point of view, many food  additives  are
potential substrates for microbial metabolism.

    The  gut bacteria are  situated principally in  the terminal
parts  of the intestinal  tract, and thus  highly  lipid-soluble
compounds  that are  absorbed in  the upper  intestine will  not
undergo  bacterial  metabolism.  However,  tissue metabolism may
give rise to conjugates that are excreted into the bile and thus
available  for bacterial hydrolysis.  Clearly,  then, the design
of  appropriate investigations with  the gut microflora  must be
linked closely to  in vivo  studies on absorption and metabolism.
 In vitro  incubation of the food additive and/or its metabolites
with the bacteria of the caecum or faeces is a useful but diffi-
cult  technique, with considerable potential  for the generation
of spurious data.  Some of the pitfalls of prolonged incubations
are that:

    (a)  the  use of a nutrient medium may allow the growth of a
         non-representative bacterial population; while

    (b)  the  use of a non-nutrient medium may act as a powerful
         selective  force for organisms able to use the additive
         as a source of carbon and energy.

    There  are three  primary  in vivo  methods  for studying the
role of the gut microflora in the metabolism of a compound:

    (a)  parenteral administration of the compound, which should
         result  in  decreased  microbial metabolism  of  poorly
         absorbed polar compounds, compared with oral dosing;

    (b)  studies  on animals in  which the bacterial  flora  are
         reduced  by  the  use  of  antibiotics  or  by surgical
         removal of the caecum; and

    (c)  studies  on germ-free animals  and on (formerly)  germ-
         free   animals  contaminated  with  known   strains  of
         bacteria.

    A number of factors may influence the  metabolic  activation
of  foreign chemicals by the  host microflora (see reference  59
for an expansion of these points):

    (a)  Host species

    Species differences exist in the number and type of bacteria
found  in the gut and  in their distribution along  the gut.  In
this respect, the rat is a poor model for man, since significant
numbers of bacteria occur in the upper intestinal tract  of  the
rat, whereas this region is almost sterile in man.

    (b)  Individual variations

    There  is  a great  deal  of variability  among  individuals
within a species in the extent to which some  compounds  undergo
metabolism  by the gut flora.  Many of these variations probably
arise  from differences  in the  enzymatic capacity  of the  gut
flora rather than in the delivery of the chemical to  the  lower
intestine.   Thus, if,  in animal  studies, a  food additive  is

shown to be metabolized by the gut flora to an entity of toxico-
logical significance, it is essential that its metabolic fate be
characterized in the human being.

    (c)  Diet

    The  composition of the gut flora depends on the diet, which
may  influence the  extent of  microbial metabolism  of  a  food
additive.

    (d)  Medication

    The  widespread oral administration of  medications, such as
antibiotics and antacids, in the human population, is a cause of
variations in metabolism by the gut microflora.

    (e)  Metabolic adaptation

    The metabolic capacity of the gut flora is far more flexible
than that of the host.  Thus, long-term adminstration of foreign
chemicals  can lead to changes in both the pattern and extent of
microbial metabolism of the chemical.  Because prior exposure to
the  compound under test  may significantly alter  the metabolic
potential  of the gut  microflora, metabolic studies  should  be
performed not only on previously unexposed animals but  also  on
animals  that have been  exposed to the  test compound for  some
time.  For the same reason, any  in vitro  studies should be per-
formed  with caecal contents that have been collected both prior
to and during long-term animal feeding studies.

5.2.4.2     Effects of the chemical on the gut microflora

    During  high-dose animal feeding studies, the gut microflora
may be affected in two ways:

    (a)  Development of antibacterial activity

    A  weak antibacterial activity may  become significant after
long-term intake of near-toxic doses of a food  additive.   This
may  manifest itself either  as an alteration  in the number  of
bacteria  present, which  can be  measured directly,  or  as  an
abnormal microbial metabolic pattern.  The latter can be studied
by  measurement of certain endogenous  metabolites produced only
by  the gut flora,  such as phenol  and p-cresol, which  provide
indirect  evidence  of  alterations  in  the  gut  flora.   Such
information  may also be of value in the interpretation of other
variables such as nitrogen balance.

    (b)  Increased substrate for gut microflora

    The food additive may act directly as a substrate  for  bac-
terial growth.  This can be readily illustrated  by  appropriate
high-dose pharmacokinetic studies, coupled  with  in vitro  meta-
bolic studies on the gut flora.  Alternatively, the  food  addi-
tive may inhibit digestion or absorption of other dietary compo-
nents  so  that these  become available to  the bacteria in  the
lower intestine in increased amounts.

    Increased  amounts of substrates in the lower intestine pro-
vide  an increased osmotic  effect in the  caecum, which may  be
detectable  as caecal enlargement  (section 5.1.2).  The  reason
for caecal enlargement must be studied before  the  significance
of the lesion can be assessed since it may indicative of:

    (i)     abnormal  osmotic balance with consequent  changes in
            permeability  to minerals in the  caecum, which could
            lead to nephrocalcinosis;

    (ii)    microbial metabolism of nutrients, which could result
            in the formation of potentially toxic metabolites and
            abnormalities in the nitrogen balance; or

    (iii)   microbial  metabolism  of  the food  additive,  which
            could lead to the formation of toxic products.

5.3.  Influence of Age, Nutritional Status, and Health Status in the Design 
and Interpretation of Studies

    Animal toxicity studies are generally performed with healthy
animal  populations that are in  a state of over-nutrition  in a
protected  environment.   This  basic procedure  is altered only
when  there is a specific reason for doing so, for example, when
nutritional factors are being studied.

    In  order to establish the safety of food additives, experi-
mental  protocols  have  tended towards  more  universal designs
encompassing  populations during all  stages of the  life cycle,
e.g.,  reproduction  studies  are often  included  in  long-term
studies.  Such protocols are intended to mimic the type of expo-
sure in the bulk of the human population.  Margins of safety and
medical  advice are used  to protect subpopulations  at  special
risk for one reason or another.

5.3.1.  Age

    It is well known that the age of a test animal can influence
the toxic response to a substance being tested.  For example, an
enzyme  activity that is  involved in the  metabolism of a  sub-
stance in an adult may be virtually absent in an immature animal
or  vice versa.  Thus, a compound that is metabolized to  a  less
toxic  metabolite  in an  adult animal would  be more toxic  for
young  animals  lacking  the appropriate  enzyme activity; obvi-
ously,  the reverse would be true for a substance metabolized to
a  more  toxic  metabolite.  Differences  in sensitivity between
mature  and young animals  may also result  from differences  in
intestinal  flora as  observed, for  example, in  the growth  of
distinctive flora in the upper intestine in human  infants  that
render  them sensitive to nitrate.  Greater sensitivity may also
arise in young animals, because of the incomplete  formation  of
intestinal,  blood-brain, or other tissue  barriers, which leads
to  the passage of  potentially harmful substances  through  the
barriers.

5.3.1.1     History

    The  WHO  Scientific  Group on  Procedures for Investigating
Intentional  and  Unintentional  Food  Additives  discussed  the
effects  of age on toxicity  (2, pp.  10-12) and  found that "in
general, the young animal is more sensitive to the toxic effects
of  exposure to  chemicals". Among  the reasons  cited  for  the
increased sensitivity in neonates were differences in  the  dis-
tinctive  flora of the upper bowel and differences in the levels
of  the "drug-metabolizing enzymes", which are frequently low in
the neonate.  Attention was drawn to interspecies differences in
the  neonatal levels and in age-related changes in the levels of
these  enzymes.   The  Scientific Group  stated  that "pertinent
information derived from reproduction (multi-generation) studies
provides some assurance on the safety of compounds that might be
present  in the diet of babies" but felt that "since babies con-
stitute  a special population, close observation of epidemiology
in this group is an important practical aspect of the evaluation
of  the effects of exposure."  The Scientific Group also saw the
need  for "further information on the development of enzyme sys-
tems  in the  human young,  with particular  emphasis  on  those
enzymes  responsible  for  dealing with  foreign compounds." The
report  concluded  (2, p. 23)  that  "useful information  may be
obtained from studies in newborn or young animals,  from  repro-
duction  studies, and from  biochemical studies" and  called for
further  research on "the development  of enzyme systems in  the
human young, with particular emphasis on those  enzymes  respon-
sible  for dealing with  foreign chemicals"  (2,  p. 25).   With
respect to the latter research, the Scientific  Group  concluded
that "this information is essential in assessing the  safety  of
additives in baby food."

    Subsequently, the tenth Report of JECFA (29, p.  24)  recom-
mended  that a special  subcommittee of JECFA  should be  estab-
lished  to study the special  problems arising from exposure  of
infants  and young children to  food additives.  In response  to
this  recommendation, an FAO/WHO  meeting on Additives  in  Baby
Foods was convened in 1971, and the report of this  meeting  was
included as Annex 3 to the fifteenth JECFA Report (42,  pp.  29-
37).   A distinction was made, on developmental grounds, between
children  up to 12 weeks of age and children over 12 weeks.  The
Subcommittee  considered  it  prudent  that  food  intended  for
infants under 12 weeks of age should not contain  any  additives
at all.  However, if it were deemed necessary to  use  additives
in food intended for young infants, the  Subcommittee  concluded
that  "particularly  for  infants under  12 weeks, toxicological
investigations  should be more extensive and include evidence of
safety to young animals."

    With  respect  to  contaminants, the  Subcommittee concluded
that "the establishment of acceptable residue levels  of  pesti-
cides  or other contaminants  likely to be  present in milk  and
cereals  for infant foods should be based on toxicological eval-
uation in very young animals" (42, p. 31).  The report also made
observations  on  particular  classes  of  food  additives  (42,

pp. 32-33).   The vulnerability of very young infants was recog-
nized,  and guidelines on toxicological  testing were formulated
(42, p. 34).  These include the following:

    (a)  Before a food additive is regarded as safe for  use  in
         food  intended for infants up  to 12 weeks of  age, the
         toxicological  studies  should  be extended  to include
         animals in the corresponding period of life.

    (b)  It  is  difficult  to recommend  precise  toxicological
         testing  procedures until more basic  research has been
         undertaken.   There are also difficulties  in selecting
         appropriate  species.   In these  circumstances, short-
         term studies should be conducted in several species and
         should  include the oral administration of the additive
         under  test,  at suitable  dose  levels, to  newly born
         animals  up to  and including  the end  of the  weaning
         period.

    (d)  When life-span studies and multi-generation studies are
         carried  out, they should  be extended to  include oral
         adminstration  of  the  food additive  at suitable dose
         levels to a proportion of animals from the day of birth
         throughout the pre-weaning period.

    The  practical difficulties and  cost of implementing  these
recommendations  on a routine basis would be immense, involving,
as it would, artificial feeding of litters of newborn laboratory
animals.   However, in situations in  which young infants are  a
target  population  for an  additive,  it seems  reasonable that
studies such as these should be performed.

    When  considering glutamate, the fourteenth  report of JECFA
(30, p. 8) noted that:

    "any  attempt to interpret these data in terms of human
    neonates  and infants involves  the problem of  how far
    developmental stages in animal species and in  man  can
    be  considered equivalent in relation  to vulnerability
    to possible effects of food additives.  Relevant infor-
    mation would be of considerable value."

    The  sixteenth (60) and  twentieth meetings (19,  p. 22)  of
JECFA  recommended that a review  should be made of  the special
problems  arising from the exposure  of infants and children  to
contaminants  in food.  This review was conducted at the twenty-
first  meeting  (20,  pp.  9-12),  which  also  considered  food
additives.  The Committee stated that:

    "scientific  evidence  indicates that  newborn and very
    young  children are particularly sensitive to the harm-
    ful  effects of foreign chemicals" due to,  inter alia,
    "immaturity of enzymatic detoxifying mechanisms, incom-
    plete  function  of  excretory organs,  low  levels  of
    plasma proteins capable of binding toxic chemicals, and

    incomplete  development of physiological  barriers such
    as the blood-brain barrier.  Moreover, there appears to
    be  a general vulnerability of rapidly growing tissues,
    which  is  particularly  important with  regard  to the
    developing nervous system."

    The  Committee  reiterated  that food  intended  for infants
under  12 weeks  of age  should (with  certain  exceptions)  not
contain  any  additives but  that  "in assessing  food  additive
safety,  the  question  of  potential  special  hazards  for the
newborn and infants should be kept in mind"  and  "toxicological
and  metabolic studies of  food additives should  always include
investigations  that  permit the  evaluation  of safety  for the
newborn  and the infant."  The Committee stated further that "in
order  to gain more information  about the long-term effects  of
exposure  in utero  and in  the  post-natal  period,  appropriate
methodology  must  be  developed," and  the Committee emphasized
that  "short-  and  long-term effects  of  exposure  in utero and
during lactation should be taken into account for food additives
and contaminants evaluated by the Committee" and  "this  evalua-
tion might include a request for appropriate animal studies."

    Implicit in these statements is a call for metabolic studies
on  neonates  and for  toxicological studies involving  in utero
exposure  followed  by  long-term studies.   Indeed, the twenty-
second  report of  JECFA (32,  p. 30)  reaffirmed the  need  for
testing  the effects of exposure  to food additives and  contam-
inants  in utero  and on  neonates  during  suckling.    However,

    "in  view of the complexity of the testing procedures,"
    the  Committee recommended that  "WHO should convene  a
    meeting of experts to assess: (a) the degree of any in-
    crease  in  the  sensitivity of  toxicological  testing
    afforded  by  exposure  in utero through  lactation; and
    (b) the need to include such expo-sure in toxicological
    tests  as a means  of increasing public  health protec-
    tion."

Criteria  for  determining  whether  in utero exposure  is  to be
included  in such studies  should include such  information that
the chemical crosses the placental barrier and/or is secreted in
breast milk.  The Committee recommended further that:

    "the  experts should also propose  the most appropriate
    guidelines  for  experimentation, taking  into account:
    (a)   the  dosages used  and  the relative  exposure of
    mother and fetus to the agent under study; (b) the pos-
    sibility of combining this modified long-term test with
    reproduction studies; (c) the length of the studies re-
    quired; and (d) the most appropriate species to use."

A meeting of experts has not yet been convened to consider these
issues.

    A  document of potential  benefit to JECFA  is one that  has
recently  been developed by the  IPCS and the Commission  of the

European  Communities  concerning the  principles for evaluating
health risks from chemicals during infancy and childhood.  Among
the objectives of this activity were to:

    (a) "investigate whether and when there is a  need  for
    specific  approaches  when  evaluating the  health risk
    associated  with exposure to  chemicals . .  .   during
    infancy  and  childhood";  and  (b)  "identify  further
    developments  in methodology that are necessary for the
    assessment  of health risks associated with exposure to
    chemicals during the early period of life" (61).

5.3.1.2     Usefulness of studies involving  in utero  exposure

    In  order to evaluate the usefulness of  in utero studies, it
is important to review the available toxicity data  relating  to
this  issue.   Most  studies  including  in utero  exposure  have
involved  the  use of  known  carcinogens.  The  IARC  monograph
"Transplacental  Carcinogenesis"  (62)  provides information  on
much of the earlier research on these substances.   In  general,
the  reports  indicate  that, although  transplacental  carcino-
genesis  could occur,  there have  been no  instances  in  which
compounds that were carcinogenic for the offspring were not also
carcinogenic for the adult and  vice versa.  However,:

    (a)  the  carcinogenic effects in the offspring can occur at
         sites  different  from  those observed  in  the  parent
         (e.g.,  in transplacental rat studies, ethylnitrosourea
         shows  striking neuro-oncoselectivity, not  observed in
         the  parent, and the induction of unique tumours of the
         vagina   in   young   women  has   been  observed  with
         diethylstilboestrol treatment of mothers given the drug
         for pregnancy maintenance); and

    (b)  the fetus may be more susceptible to tumour development
         than the adult rat (observed with ethylnitrosourea).

    On  the other hand, in some cases, adults are more sensitive
than  young offspring, as  with certain nitrosamines  (62,  63),
suggesting  that  enzyme  systems capable  of  converting  these
compounds into their ultimately carcinogenic forms are not fully
developed in fetal tissue.

    Although  transplacental  carcinogenesis has  been the major
interest of  in utero studies, evidence is accumulating that non-
carcinogenic  substances  may  be the  cause  of  a  variety  of
biochemical  and other toxic  effects in the  developing  fetus.
Some  of  this information  has come from  studies of the  toxic
effects  of environmental contaminants such as methylmercury and
PCBs.   Poisoning episodes of  methylmercury in Japan  and  Iraq
indicate  that  the developing  fetus  shows toxic  symptoms  at
levels  at which the mother is asymptomatic.  This appears to be
because of selective localization of methylmercury in the brains
of exposed fetuses rather than a higher sensitivity of the fetus
itself  (64).  In the case of PCBs, monkeys with body burdens of
PCBs  derived  from  previous exposure  produced  offspring that

showed  behavioural and learning deficiencies;  it was estimated
that  approximately  40%  of the  body  burden  of PCBs  in  the
offspring was derived from placental transfer (65). A finding of
equal or greater interest in this study was that 60% of the body
burden  of PCBs was transferred  postnatally in the milk.   This
result  is consistent with  the finding that  the only route  of
excretion of certain chemicals from the human body, particularly
the  halogenated hydrocarbons that accumulate in fat, may be via
breast milk.  Breast-fed infants, therefore, may be  exposed  to
very  high levels of these  compounds, far exceeding the  accep-
table  daily intake (ADI)  or the provisional  tolerable  weekly
intake  (PTWI) (66), pointing to the critical need for obtaining
data during the neonatal phase for these types of compounds.

    The  results of transplacental  rat studies have  also shown
the  possibility that the  course of enzyme  development in  the
fetus may be markedly altered by exposure to foreign substances.
This  so-called programming may alter the time of development of
specific  enzymes or change the  pattern of development of  sex-
dependent  enzymes, i.e., male offspring may develop enzyme pro-
files  more  characteristic  of female  offspring (67).  Studies
designed  so that the progeny of exposed parents are used in the
long-term phase serve as a fitness test to detect subtle effects
of this type.

    Possible  differences  in  the placenta  structure  in human
beings  and  experimental animals  should  be considered  in the
interpretation of  in utero  studies.  Structural differences may
result in significantly different rates of transfer of chemicals
across  the placental membrane in  experimental animals compared
with  human beings.  This  should be considered  when  selecting
appropriate  dose  levels  during the  in utero  phase  of animal
studies.

5.3.1.3     Complications of aging

    The   Scientific  Group  on  Procedures   for  Investigating
Intentional  and Unintentional Food Additives (2) concluded that
it  is  "better  to  carry  out  toxicity  studies  before   the
complications of senescence arise" but, nevertheless, called for
"more  basic information . . . on toxicity in aged as well as in
young animals."

    Older  animals  may  be  especially  sensitive  to   certain
substances,  because of reduced functioning of vital organs such
as  the kidney  or liver.   The ability  to  metabolize  certain
substances  in the liver may  decrease in aged animals  (68-71),
resulting  in an accumulation of toxic substances and consequent
effects  that would  not be  seen in  young animals.   Normally-
occurring lesions of old age such as tumours or  kidney  lesions
may mask subtle compound-related pathology and may  render  aged
animals   a  poor  model  system  for  assessing  some  lesions.
Conversely, some lesions may require long exposure to develop or
may only be manifested in older animals.  Such old  animals  are
used routinely, and more research and documentation  are  needed
in this area.

5.3.2.  Nutritional status

    JECFA has not directly addressed the issue of over-nutrition
in  laboratory animals.  Excessive  food intake and,  in  parti-
cular,  ad libitum  feeding of animals can  complicate the inter-
pretation  of  studies.   Considerable research  indicates  that
altered  caloric intake and qualitative changes in diet can have
a  profound effect on various disease processes, particularly on
the  occurrence of neoplasms  (72, 73).  Additional  research is
needed  leading  to  better nutritional  designs of experimental
models for safety evaluation.

    In  considering the effects  of nutritional status  on toxi-
city, the WHO Scientific Group on Procedures  for  Investigating
Intentional  and  Unintentional  Food Additives  (2,  pp. 12-13)
recognized that nutritional status can influence toxicity, posi-
tively  or negatively, depending on the substance, but felt that
"it is wise to maintain all the animals on a diet that is nutri-
tionally  adequate in every way,  unless there is some  specific
reason for doing otherwise."  While noting that "further work is
needed  on the  effects of  various states  of  malnutrition  or
undernutrition  on  the  toxicity manifested  by  chemical  com-
pounds,"  the  Scientific Group  concluded  that "an  effort  to
simulate  conditions  of  malnutrition  in  man  .  . .  is  not
considered  advisable  in  routine toxicological  investigations
intended  for the evaluation of  safety" and "the evaluation  of
safety is best carried out by using healthy animals on adequate,
balanced diets."

    In  the seventeenth JECFA  report (16, p. 31),  it is  noted
that  reactions  of food  additives  with food  constituents may
affect  the  nutritional  value  of  the  food  and  that   this
possibility  can  be studied  by  chemical or  biological  assay
methods (section 4.2).  However, it is further reported that "it
may be necessary to undertake a toxicological  investigation  of
treated  food  materials;  here,  a  margin  of  safety  may  be
introduced  by  conducting  the test  with  food  that has  been
deliberately  over-treated to a measured extent."  In addressing
this   problem  of  interaction   between  additives  and   food
constituents in the twenty-fourth report of JECFA  (21,  pp. 10-
11), it is pointed out that such reactions may occur during food
manufacture, storage, and cooking, and it is re-emphasized

    "that  a better perspective of the safety of food addi-
    tives  would be gained if information on their manufac-
    ture and technological use were more readily available.
    Such information should cover . .  . any available data
    on  the chemical fate of  each additive in those  foods
    and on the effects of additives on nutrients . .  .  ."
    It  "may even  be necessary  sometimes to  carry out  a
    study  on the technological versus  nutritional effects
    of certain additives and to present this information to
    the Committee."

Similar  conclusions were reached by the twenty-fifth meeting of
JECFA (22, pp. 11-12).

    In the twenty-fourth report of JECFA (21, p. 10), increasing
concern  is also expressed  with "the development  of  materials
designed as substitutes for normal components of food" and noted
that  "questions of nutritional adequacy arise in such instances
and must not be overlooked."  The Committee believed  that  "the
problems  associated with the designing of tests to assess toxi-
city  of  these substances  and  with their  interpretation  and
extrapolation to man require special consideration".

    In  considering  the  particular case  of  acceptable  daily
intakes  (ADIs) with regard to  nutrients such as ascorbic  acid
used  as food additives, it is noted in the eighteenth report of
JECFA  that the lower  limits corresponding to  the requirements
for  such nutrients are  determined by expert  committees  "con-
cerned  with adjusting the  ADI if a  food additive is  shown to
interfere  with nutritional requirements in one form or another"
(17, footnote p. 9).

    Interference  with  nutritional  requirements can  occur  by
antagonizing  the  normal  physiological function  of a vitamin,
trace  metal, or other micronutrient, either through destruction
of   the  micronutrient  before  ingestion   (such  as  thiamine
destruction   by  sulfur  dioxide)  or   through  antagonism  or
inactivation  in the body after ingestion.  Testing regimes with
both  nutritionally-supplemented and unsupplemented animals will
show  whether the antagonism is reversible and will separate the
toxic  potential of the  agent under study  from its effects  on
nouriture.  Such studies will aid in an ultimate  assessment  of
safety  in that  normal levels  of the  micronutrient  in  human
populations may be so in excess of absolute requirements  as  to
make  this effect of  the agent of  little consequence.  On  the
other hand, if the nutrient is often at marginal levels in human
diets, then clearly its evaluation will have to take this effect
into account.

    Other  specific nutritional problems have been considered in
relation  to phosphates (23, p. 13) and metals occurring in food
(23, p. 14).  The problems of the former include  alteration  of
dietary calcium:phosphorus (Ca:P) ratios with consequent compli-
cations  of,  for  example, nephrocalcinosis,  and the Committee
recommended that "further studies should be carried out  on  the
consequences of high dietary intakes of phosphate,  with  parti-
cular  reference to the  Ca:P ratio."  The  association  between
caecal enlargement and nephrocalcinosis also indicates that fur-
ther  complications may arise in  relation to calcium and  phos-
phorus absorption due to other food additives.

    With  respect  to  the presence  of  metals  in  foods,  the
Committee noted that:

    "toxicological  evaluation of metals in food calls  for
    carefully balanced consideration of. . . ( inter  alia) 
    . . . nutritional  requirements, including  nutritional
    interactions with  other constituents of food  in  res-
    pect of .  .  .  absorption,  storage in the  body  and
    elimination" and in the  case  of  essential  elements,

    tentative  tolerable daily  intakes "should not be con-
    strued as an indication  of  any change in  recommended
    daily  requirements,   but  as  reflecting  permissible
    human exposure. . . ." (23, pp. 14-15).

5.3.3.  Health status

    Health  status  of  test animals  is  of  key importance  in
assessing  the results of any toxicity study.  Health of animals
should be routinely monitored during testing.  Animals  in  poor
health  from a viral  or bacterial infection  may be  especially
sensitive to the test substance.  Early deaths  from  infectious
diseases may leave insufficient time for chronic toxicity of the
test compound such as carcinogenicity to manifest itself.  Path-
ological lesions from infectious disease may also mask compound-
related pathology.  For example, lung toxicity could  be  masked
by a respiratory infection.  The reverse can also  occur.   Cer-
tain respiratory infections in the rat predispose the  rat  lung
to  lymphoreticular neoplasms (74: acesulfame potassium, pp. 22-
23).   Antibiotics and  other drugs  should not  be used  unless
absolutely  necessary  to  control infections  because their use
complicates the interpretation of the study.

5.3.4.  Study design

    When  designing  toxicity  studies on  food  chemicals where
factors of age, nutritional status, and animal health are likely
to  affect the results, the investigator should design the study
appropriately  with foreknowledge of  these factors, bearing  in
mind the population likely to be exposed.  For example,  if  the
substance  is to be  used in infant  formulas or baby  foods, an
appropriate  animal model to  mimic the human  infant should  be
used.  The miniature pig may be a useful model in  this  regard,
because  it can be bottle-fed and many aspects of its metabolism
are similar to those of man.

    Most  food additives are also consumed by pregnant women, so
the  factors  discussed above  with respect to  in utero exposure
should  be  considered  when  assessing  their  safety.   Animal
studies designed to parallel human exposure should  include  the
important phases of exposure that occur during fetal development
and  suckling of the  infant.  Exposure of  the test animal  can
then involve both the parent compound and  maternal  metabolites
that  can  either  cross the  placental  barrier  or  enter  the
mother's milk; it will also permit the assessment of metabolites
formed  in the developing  embryo, which may  differ from  those
formed in the maternal system.

    The need for guidelines, as recommended by the twenty-second
JECFA  (32, p. 30), continues  (section 5.3.1.1).  The  exposure
level  and its relationship  to the no-observed-effect  level in
animal  studies, the type  of additive (e.g.,  nutritive  versus
non-nutritive),  information about whether the  additive crosses
the  placental barrier or  is secreted in  milk, and other  data
relating  to  the  reproductive  or  developmental  toxicity  of
similar compounds should be considered in determining  the  need
for  in utero studies.

    The age of test animals is an important factor  to  consider
when designing carcinogenicity studies (75, pp. 57-107).  If the
study  is  terminated too  soon,  the possibility  of  detecting
carcinogenicity  that manifests itself late in the animal's life
span  is lessened.  Conversely,  in studies with  long  exposure
times  (more than 104  weeks in rats  and mice), the  background
incidence   of  naturally-occurring  tumours  may   "swamp  out"
compound-related  tumours occurring at  some sites.  One  way to
resolve  this problem is to add additional groups of animals for
interim  sacrifice.   However,  this  increases  the  cost   and
complexity  of the  study.  Knowledge  of the  test strain  with
regard  to longevity and  tumour incidence is  necessary in  the
design  and interpretation of  carcinogenicity studies.  In  any
case,  final termination should take place while there are still
enough  survivors  among  the exposed  animals  and  concomitant
controls to make a statistical evaluation.

    It  generally is not feasible to test all food additives for
their  effects on all age  groups and disease states.   In cases
where certain populations, such as phenylketonuric patients, are
known to be sensitive to a food additive, warning labels or edu-
cation  through other means  may be necessary.   When substances
are included in special medical foods used in  treating  certain
diseases,  it would be prudent  to examine closely any  reported
physiological  or  toxicological  effects of  the substances, to
determine  whether they can be  safety ingested by the  intended
population.   Specialized testing on animal models may be neces-
sary.   If the  additive is  to be  used in  infant formulae  or
"junior  foods", this fact should be kept in mind at the time of
the safety assessment.  This point is often  overlooked.   Large
amounts  of such a  food additive may  be consumed, because  the
formula may constitute the entire diet of the infant and because
infants take in much more food than an adult on a kg body weight
basis.   If specific subpopulations  are identified as  being at
higher  risk than the  general population, these  groups can  be
protected  by adjusting the ADI to take their special needs into
account.

    In general, the safety of a food additive, as far as limited
special  populations not readily identified  are concerned, must
rely  on  the conservatisms  built  into the  safety  assessment
process, the analysis of the data, and the safety  factors  used
in setting an ADI.

5.4.  Use of Human Studies in Safety Evaluation

    Human studies are not normally included in the data packages
that  JECFA reviews  in its  evaluation of  new food  additives.
However, the Committee recognizes the value of human  data,  has
sometimes  requested such data,  and has always  used it in  its
evaluations when available.  Data from controlled human exposure
studies  are useful in confirming the safety indicated by animal
studies  after the establishment  of ADIs.  Such  data are  also
useful  in subsequent periodic  reviews, and might  facilitate a
re-evaluation   of  the  safety  factors  that  are  applied  in
calculating ADIs.

    Investigation  in human subjects  was addressed by  the  WHO
Scientific Group on Procedures for Investigating Intentional and
Unintentional Food Additives (2, pp. 9-10).  The Group felt that

    "prediction and prevention of possible toxic hazards to
    the community that might arise from the introduction of
    a  chemical  into the environment can be made more cer-
    tain if information from meaningful  studies  in  human
    subjects  is  available."  Three  particular aspects of
    toxicology  were  identified  in  this connection, "the
    choice  of the most appropriate animal species for. . .
    the prediction of human responses; secondly, the inves-
    tigation  of a  reversible  specific effect observed in
    the  most sensitive animal species to determine whether
    it represents a significant hazard to man; thirdly, the
    study of effects specific to man."

The Group pointed to:

    "the need, at a relatively early stage, to  obtain  in-
    formation  on  the   absorption,  distribution,  metab-
    olism,  and elimination of the chemical in  human  sub-
    jects, since this makes it  possible  to  compare  this
    information  with  that obtained in various animal spe-
    cies  and to choose the species that are most likely to
    have a high predictive value for human responses."

This  need has been reiterated  by subsequent meetings of  JECFA
(27,  p. 23; 16,  p. 31; 32,  p. 13) and  in  WHO  Environmental
Health  Criteria  6 (76).   However,  the WHO  Scientific  Group
acknowledged that "it is necessary to have  adequate  short-term
toxicological  information  in  several species  before even low
doses  of a new chemical are administered to human subjects" (2,
p. 9).

    In  relation  to  ascertaining  whether  the  safety  margin
predicted  from animal data is  valid, the WHO Scientific  Group
decided  that it might  be helpful to  administer a chemical  to
human volunteers, but emphasized the conditions that  should  be
fulfilled  with regard to such a study (2, p. 10).   Inter alia,
these conditions include:

    (a)  The effect or effects studied should be reversible.

    (b)  The  dose levels used should  be based on full  inform-
         ation  of the toxicological properties of the substance
         in animals.

    (c)  The  investigation should be terminated immediately the
         effect has been unequivocally demonstrated.

    With regard to effects specific to man, the  WHO  Scientific
Group  (2,  p. 10)  considered  it  unacceptable to  study  such
effects  by  means  of volunteers  (in  an  analogous manner  to
clinical  trials  with  drugs) but  thought  that  toxicological
studies could be made on those who are occupationally exposed to

the chemical or in patients suffering from accidental poisoning.
A  need was identified  for "more critical  epidemiological  and
toxicological  investigations in such situations."  Such studies
could  be of particular value in relation to hypersensitivity or
other idiosyncratic reactions since no suitable animal model has
yet  been  developed.   In  relation  to  hypersensitivity,  the
seventeenth  and eighteenth meetings  of JECFA (16;  17,  p. 10)
stated  that  "no  approval would be given for the use of a sub-
stance  causing  serious  or widespread  hypersensitivity  reac-
tions".   However,  such information  can  be derived  only from
studies on human beings.

    The  WHO  Scientific Group  has  raised an  apparent contra-
diction  in its different  recommendations with regard  to  con-
firming  animal  studies  and investigating  effects specific to
man.   As stated above,  the Group recommended  that  controlled
human studies be performed to confirm animal studies,  but  that
it  is inappropriate to study effects specific to man by the use
of human volunteers.  This is all the more  perplexing,  because
controlled  human  studies,  despite their  limitations, are the
only  means available, at present,  for studying effects in  man
that  are not observed in animals.  JECFA may wish to reconsider
the  question  of using  human  volunteers to  identify specific
responses, which would be done only after the usual  battery  of
toxicological  investigations had been completed.   The words of
Paget (77) are cogent in this regard:

    "The question  is  not  whether or not  human  subjects
    should  be  used  in  toxicity  experiments  but rather
    whether  such  chemicals,  deemed  from animal toxicity
    studies to be relatively safe, should be released first
    to controlled, carefully monitored groups of human sub-
    jects,  instead of  being  released indiscriminately to
    large populations with no monitoring and with little or
    no opportunity to observe adverse effects."

    The  ethical problems associated with  toxicological studies
on  human beings have been  reviewed succinctly in WHO  Environ-
mental Health Criteria No. 6 (76, pp. 41-42).

    Information  relating to human  exposure to a  food additive
during  its  pre-marketing stage  can  be obtained  through  the
health  monitoring  of employees  coming  into contact  with it,
either  in the laboratory  or the manufacturing  plant.  Because
the  route  of exposure  in such a  situation is through  either
contact  with the  skin or  vapour in  the lungs,  immunological
sensitivity  and anaphylactoid reactions  (mediator-release ana-
phylaxis-like reactions), often involving histamine release, are
the  adverse effects most likely  to occur.  Thus, any  observa-
tions  indicating  the potential  for  these effects  should  be
recorded at the time they are observed.

5.4.1.  Epidemiological studies

    Most  studies of  the effects  of food  additives  on  human
populations  are performed after the additive has been placed on

the  market.  In nearly all  cases, the impetus for  the perfor-
mance  of human studies on a food additive is that the safety of
its  use has been brought  into question for one  reason or ano-
ther.   For example, retrospective investigations  have revealed
effects  such as "beer drinkers  cardiomyopathy", resulting from
exposure to cobalt salts.  Adverse findings in these studies may
be used for bringing an additive back to JECFA for re-evaluation
of its safety.

    Epidemiological  studies  designed  to assess  the safety of
food  additives have been  performed in several  instances, but,
generally,  definitive results have not been obtained because of
the lack of sensitivity of such studies and problems  in  ident-
ifying  control  populations.  For  example, long-term low-level
nitrite exposure has been very difficult to  study  epidemiolog-
ically  because of its ubiquitous nature and the consequent dif-
ficulty  of finding subpopulations with little or no exposure to
nitrite.   With  saccharin,  an extensive  data  base  involving
retrospective  epidemiological studies and  case-control studies
has  been developed.  This  data base has  been generated  using
different  subpopulations  located  in  different   geographical
areas,  and the results  for human bladder  cancer have  usually
been negative (78).

    In many cases, the purpose of an epidemiological study is to
confirm  in human beings a positive finding observed in animals.
Thus,  when food additives are used extensively and when exposed
or  unexposed populations cannot be identified, negative results
are  usually  deemed  not to  be  of  much value  for regulatory
purposes.   This is because epidemiological studies, are, on the
whole, less sensitive than well-designed animal feeding studies.
However,  when the number  of individuals studied  becomes  very
large,  this lack of  sensitivity is somewhat  ameliorated,  and
safety  decisions can be made on the basis of the human studies.
For  example, in  the case  of saccharin,  negative  results  in
epidemiological  studies have been considered important by JECFA
for deciding that its continued use is acceptable (1).

    Of  course, much more  can be said  about a positive  result
than  a negative one,  especially with epidemiological  studies,
which  are  usually  relatively  insensitive.   An  undetectable
adverse effect in a study involving a few  thousand  individuals
could  affect a very large  number of people in  a population of
hundreds of millions.

5.4.2.  Food intolerance

    For  the purposes of  this discussion, food  intolerance  is
defined as a reproducible, unpleasant reaction to a food or food
ingredient,  including  reactions due  to immunological effects,
biochemical  factors  such  as enzyme  deficiencies, and anaphy-
lactoid  reactions, which often include histamine release.  Food
allergy, sometimes used synonymously with food sensitivity, is a
form  of food intolerance in which there is evidence of abnormal
immunological reaction to the food.  Immunological reactions may
be further characterized on the basis of the timing of the onset

of symptoms following ingestion of the offending food and on the
type  of response involved.  Reactions  occurring within minutes
to hours of food ingestion are characterized as immediate aller-
gic  reactions, which are  mediated by Immunoglobulin  E  (IgE),
while reactions beginning several hours to days after food expo-
sure  are characterized as  delayed allergic, or  cell-mediated,
reactions.

    Various dietary factors may be responsible for  food  intol-
erance.   These may be naturally-occurring  dietary constituents
or, in some cases, food additives.  Two notable examples of food
additives  that have been  implicated are tartrazine,  which may
induce  urticaria and bronchoconstriction in asthmatic patients,
and  sodium  metabisulfite,  which  has  been  associated   with
bronchospasm,  flushing,  hypotension,  and even  death  due  to
anaphalaxis  after ingestion by some  asthmatic patients.  Mono-
sodium  glutamate  (MSG)  gives  rise  to  "Chinese   Restaurant
Syndrome",  manifested  largely  by violent  headache.   Certain
subpopulations  appear to be sensitive to MSG, but the mechanism
is  unknown.  Despite these examples, there is little to suggest
that  food additives are likely  to cause more problems  of food
intolerance than are components naturally present in food.

    Satisfactory  animal models to  predict food intolerance  in
human  beings have not been  developed.  At the same  time, many
difficulties  are associated with human  studies, and interpret-
ation  is difficult  at least  partly because  of the  anecdotal
nature  of much  of the  evidence.  Any  interpretation of  food
intolerance  is complicated by psychological  factors, making it
extremely important that blind trials be performed to assess the
nature of the problem.

    The  most  unambiguous  method of  demonstrating food intol-
erance  is to use challenge feeding in a double-blind study; the
diagnosis  of food intolerance  can only be  established if  the
symptoms  disappear with an elimination diet and if a controlled
challenge then leads to either recurrence of symptoms or to some
other clearly identified change associated with the intolerance.
If  delayed allergic reactions  are being studied,  such effects
may  take several weeks  to disappear and  then redevelop  after
challenge  feeding.  Challenge feeding is most reliable when the
ingestion  of food is  associated with development  of  symptoms
within one to two hours.

    No  oral food challenge, even if blind, can be perfect for a
number of reasons.  Presentation of food in capsules  may  avoid
the  possibility of reactions in  the mouth, pharynx, and  oeso-
phagus  and may decrease early digestion of the food by salivary
enzymes.  Small amounts of food may be regurgitated or eructated
and  identified by taste  and smell.  Unknown  relationships may
exist  between suspected foods  and periods of  abstinence  from
that food before challenge.  The presence of other  foods  eaten
with a suspected food may have facilitated or  inhibited  diges-
tion and absorption (79).


    The simplest and most commonly used test  for  demonstrating
IgE antibodies is the direct skin test.  However, this  test  is
unreliable  as used, because  standard dosages of  food extracts
have  not  been  developed, and,  with sufficiently concentrated
food  extracts,  it  is possible  to  evoke  positive skin  test
results in any person tested.  Therefore, skin  testing  results
should  be  verified by  testing  the extract  on  non-sensitive
individuals (80).

    Other methods for testing IgE antibodies to food include the
 in  vitro  radioimmunoassay and the leukocyte  histamine release
assay  (81).  The former assay is limited by an inadequate stan-
dardized  reporting  procedure,  making a  comparison of results
between investigators very difficult (82).  The latter has found
only limited application, because it requires fresh  blood,  and
only a limited number of allergens can be tested from  a  single
blood aliquot.

    Published  studies concerning the usefulness  of either skin
testing or immunoassay to diagnose clinical adverse reactions to
food  have shown a  marked discrepancy in  results.  In most  of
these  studies, the investigators  have relied on  the  clinical
history  for  determining  the false-positive  or false-negative
rates  for skin tests or immunoassays.  Such reports are unreli-
able.  Negative results obtained in skin tests  or  immunoassays
should  be treated with  more confidence than  positive  results
(83, 84).

    If  evidence of widespread  intolerance to a  food  additive
appears  in a country that permits its use, procedures should be
established  for the centralized reporting  of such information,
if one is not already in place.  Medical professionals should be
alerted,  and  appropriate  medical tests  performed on affected
individuals  to  determine the  nature  of intolerance.   If the
problem  arises with an additive previously considered and given
an  ADI by JECFA,  ideally the results  will be  relayed  to the
Committee  so that  the safety  of the  additive can  be  recon-
sidered.   Remedies may range from no action to a recommendation
that  the additive be removed from the market.  Factors, such as
its  natural occurrence in food, should be taken into account in
such  deliberations.   Because  food intolerance  is  not spread
throughout  the general population,  but is restricted  to small
subpopulations or individuals, one of the usual remedies  is  to
label  the  food containing  the  additive prominently,  so that
sensitive individuals can avoid it.

5.5.  Setting the ADI

    Almost any substance at a high enough test level  will  pro-
duce  some  adverse effect  in  animals.  Evaluation  of  safety
requires  that this potential  adverse effect be  identified and
that adequate toxicological data be available to  determine  the
level at which human exposure to the substance can be considered
safe.


    At  the time of its first meeting, JECFA recognized that the
amount  of an additive used  in food should be  established with
due  attention to "an adequate  margin of safety to  reduce to a
minimum any hazard to health in all groups of consumers" (9, pp.
14-15).   The  second JECFA,  in  outlining procedures  for  the
testing  of intentional food additives to establish their safety
for  use, concluded that  the results of  animal studies can  be
extrapolated to man, and that

    "some  margin of  safety is desirable to allow for  any
    species  difference  in  susceptibility, the  numerical
    differences between the test animals and the human pop-
    ulation exposed to the hazard, the  greater variety  of
    complicating disease processes in the human population,
    the difficulty of estimating the human intake, and  the
    possibility of synergistic action among food additives"
    (10, p. 17).

This conclusion formed the basis for establishing  the  "accept-
able  daily  intake", or  ADI, which is  the end-point of  JECFA
evaluations for intentional food additives.  In the  context  in
which  JECFA uses  it, the  ADI is  defined as  an estimate  (by
JECFA)  of the amount  of a food  additive, expressed on  a body
weight basis, that can be ingested daily over a lifetime without
appreciable health risk.

    The ADI is expressed in a range, from 0 to an  upper  limit,
which  is considered  to be  the zone  of acceptability  of  the
substance.   JECFA expresses the  ADI in this  way to  emphasize
that the acceptable level it establishes is an upper  limit  and
to encourage the lowest levels of use that  are  technologically
feasible.

    Substances  that accumulate in the body are not suitable for
use  as food additives (39,  p. 8).  Therefore, ADIs are  estab-
lished only for those compounds that are  substantially  cleared
from the body within 24 h.  Data packages should  include  meta-
bolism  and excretion studies designed to provide information on
the cumulative properties of food additives.

    JECFA generally sets the ADI of a food additive on the basis
of the highest no-observed-effect level in animal  studies.   In
calculating  the ADI, a  "safety factor" is  applied to the  no-
observed-effect level to provide a conservative margin of safety
on account of the inherent uncertainties in extrapolating animal
toxicity  data to potential effects  in the human being  and for
variation  within the human species.   When results from two  or
more  animal studies are available, the ADI is based on the most
sensitive animal species, i.e., the species that  displayed  the
toxic effect at the lowest dose, unless metabolic  or  pharmaco-
kinetic  data are available  establishing that the  test in  the
other species is more appropriate for man (section 5.5.1).

    Generally,  the ADI is established  on the basis of  toxico-
logical information and provides a useful assessment  of  safety
without  the need for  data on intended  or actual use  and con-
sumption.   However, in setting ADIs, an attempt is made to take

account of special subpopulations that may be  exposed.   There-
fore,  general  information  about exposure  patterns  should be
known at the time of the safety assessment (section 5.5.6).  For
example,  if a food additive  is to be used  in infant formulae,
the  safety assessment is not complete without looking carefully
at safety studies involving exposure to very young animals.

    JECFA uses the risk assessment process when setting the ADI,
i.e.,  the level of "no  apparent risk" is set  on the basis  of
quantitative  extrapolation from  animal  data to human  beings.
Generally,  JECFA does not undertake risk management, in that it
leaves  it  to  national  governments  to  use  the quantitative
assessments  in a manner  appropriate to their  own  situations.
However,  this has not always  been the case, in  that sometimes
JECFA has taken into consideration, in peripheral ways, benefits
(e.g.,  hydrogen peroxide as an alternative to pasteurization in
developing countries (12; 21)) and economic need  (e.g,  polymer
packaging  materials that contain potentially hazardous migrants
should  be limited to  situations where no  satisfactory  alter-
natives exist (1)).  In this context, risk assessment  and  risk
management are more broadly used than, e.g., they are often used
in the context of carcinogenesis.

5.5.1.  Determination of the no-observed-effect level

    A  determination of a  no-observed-effect level for  a study
depends  primarily on the proper  selection of doses, such  that
the highest dose produces an adverse effect that is not observed
at the lowest dose.  Several dose levels are used  to  determine
the dose-effect relationship.  Knowing the nature of  the  toxic
response  to a  compound at  the high  level, a  more  confident
assessment  of  a no-observed-effect  level  at the  lower  test
levels  can  be  made by  focusing  more  clearly on  the target
tissues.   Great care must be  taken in dose selection,  because
the  no-observed-effect level must  be one of  the  experimental
doses; it is not an inherent property of the animal system.  For
a  discussion of  items to  consider when  selecting doses,  see
reference 75, pp. 9-49.

    The  following discussion concerns the  performance of long-
term  studies, because  these studies  are the  type most  often
performed  in support of  intentional food additives,  and  they
give  rise  to much  controversy.   However, 90-day  studies are
sometimes  sufficient for establishing safety,  as, for example,
with  substances that are closely  related to food additives  of
known   low  toxicity  (section  5.5.4).   Many  of  the  points
discussed  below  in  relation  to  long-term  studies  are also
appropriate for shorter studies when such studies serve  as  the
basis for safety determinations.

    When  long-term  studies  are indicated,  short-term  range-
finding studies should first be performed to ensure  the  proper
selection  of dosage regimen.   Care must be  taken in  applying
this  approach  to  dose  level  selection,  because  doses that
produce  signs of toxicity in  short-term testing may be  rever-
sible  on  more long-term  exposure.   In such  situations,  the

highest dose selected from range-finding studies may not produce
an adverse effect with long-term exposure, precluding  a  deter-
mination  of the no-observed-effect  level in the  longer  study
(the  significance of this transient effect should be taken into
account  when  evaluating the  data).   A situation  of  perhaps
greater frequency is one in which the dosages in  the  long-term
study  are too high,  so that, even  the lowest dose  results in
adverse  effects,  and  a  no-observed-effect  level  cannot  be
established.

    Ideally, in a long-term study, the high-dose level should be
sufficiently  high to elicit  signs of toxicity  without causing
excessive  mortality or some exaggerated pharmacological effect,
such  as sedation.  Although doses of non-nutritive additives as
high  as 5%  of the  total diet  do not  always produce  adverse
effects,  higher doses should  not be tested,  because they  may
produce  a significant nutritional imbalance.   Therefore, if no
adverse effects are observed at 5% of the diet, this dose should
be  considered the no-observed-effect level.  On the other hand,
nutritive  additives may be fed  at higher doses as  long as the
nutritional  balance is effectively  preserved in both  the test
animal and controls (section 6.2.3).

    Ordinarily,  the  middle  dose  should  be  selected  to  be
sufficiently high to elicit minimal toxic effects or  it  should
be  set midway  between the  high and  low doses.   However,  if
significant   differences   exist  in   the  pharmacokinetic  or
metabolic profile of the test substance between the high and low
doses, then an additional dose should be included in  the  study
to provide more assessment points.

    The  lowest  dose  should  not  interfere  with  morphology,
development,  normal  growth,  or longevity  or  produce adverse
functional alterations.

    The determination of an adverse effect in a particular study
depends on the doses tested, the types of  parameters  measured,
and  the ability to distinguish between real adverse effects and
false  positives.  If, for  example, only a  slight change in  a
particular  parameter is noted at  the highest dose that  is not
observed at the lower doses, then it is difficult to distinguish
between a real adverse effect and a spurious  positive  finding.
In  addition,  a  reduction  in  body-weight  gain  coupled with
decreased  food  consumption is  difficult  to interpret  as  an
adverse  effect,  because  palatability  of  the  chow  might be
affected  by the presence of  high levels of the  test compound.
However,  as noted in  section 5.1.1, generalized  decrement  in
weight  gain has sometimes been used for setting an effect level
in the absence of other toxic manifestations.

    When  two or more  studies are performed  on an additive  in
different  animal species, no-observed-effect levels  are calcu-
lated  from  each  study.  The  overall no-observed-effect level
used  for calculating the  ADI is the  no-observed-effect  level
from  the animal  study that  displayed a  toxic effect  at  the
lowest dose.  The species on which this study was  performed  is

then considered to be the most sensitive species.  This approach
is  reasonable when the animal studies are of similar length (in
relation  to the expected life span of the species) and quality,
and  no other data relating  to this issue are  available.  How-
ever, if the quality of one study is obviously superior  to  the
others and/or the studies differ with respect to  length  (long-
term  versus short-term), extra  weight should be  given to  the
longer  better-quality studies when determining  the overall no-
observed-effect  level.   If metabolic  and pharmacokinetic data
are available, the species most similar to man with  respect  to
the toxic effect should be used in calculating the  overall  no-
observed-effect level, rather than the most sensitive species.

5.5.2.  Use of the safety factor

    The  safety  factor  has  been  used  by  JECFA,  since  its
inception.   It is  intended to  provide an  adequate margin  of
safety for the consumer by assuming that the human being  is  10
times more sensitive than the test animal and that  the  differ-
ence  of sensitivity within the human population is in a 10-fold
range.  In determining an ADI, a safety factor is applied to the
no-observed-effect  level  determined  in an  appropriate animal
study.

    JECFA traditionally uses a safety factor of 100 (10 x 10) in
setting  ADIs based on long-term  animal studies, i.e., the  no-
observed-effect level is divided by 100 to calculate the ADI for
an  additive.  The no-observed-effect level is usually expressed
in  terms of mg compound per kg body weight per day, and the ADI
is expressed in the same units.  A food additive  is  considered
safe for its intended use if its human exposure is less than, or
is  approximately,  the  same as  the  ADI.   The ADI  generally
includes  both its natural occurrence and deliberate addition to
food  (17, pp. 8-10), except when the substance occurs naturally
in  a chemical form different from that employed as a food addi-
tive,  or when its  natural occurrence was  not considered  when
setting  the ADI and the substance naturally present in the diet
contributes  significantly  to its  total  intake (as  with nit-
rates).  Because in most cases, data are extrapolated from life-
time animal studies, the ADI relates to life-time use  and  pro-
vides a margin of safety large enough for toxicologists  not  to
be  particularly  concerned  about short-term  use  at  exposure
levels  exceeding  the ADI,  providing  the average  intake over
longer periods of time does not exceed it.

    National  governments  are  responsible for  regulating food
additives in such a way that consumption from natural occurrence
and deliberate addition to food does not exceed the ADI for each
additive  that is permitted.   As stated by  the WHO  Scientific
Group  on Procedures for Investigating  Intentional and Uninten-
tional  Food Additives (2, p. 6), "it is desirable that national
governments  should maintain a check on the total intake of each
food additive, based on national dietary surveys,  to  determine
whether  the total load  in the diet  approaches the  acceptable
daily  intake."  Individual governments  have the discretion  of

determining whether they will base their regulatory decisions on
the  "average"  consumer  or  the  "high"   consumer   of   food
additives.

    A safety factor of 100 should not be  considered  immutable.
When  setting the ADI, various test data and judgemental factors
should be considered.  These include:

    (a)  Inadequate data base

    In  this case,  a larger  safety factor  may be  appropriate
(section 5.5.5).

    (b)  Reversibility  of the observed effect in 
         embryotoxicity studies

    If  irreversible  developmental  effects, such  as  skeletal
abnormalities (as opposed to retarded skeletal growth), are seen
in the fetuses of animals administered the substance  in utero, a
study on a second species is indicated.  If similar irreversible
effects are not confirmed in the second animal  species,  pharm-
acokinetic  studies would be  useful to determine  relevance  to
human beings.  Judgement would then be needed to set  an  appro-
priate safety factor.  If frank teratogenic effects are observed
in  both studies, judgement  would be needed  to decide  whether
either  a larger safety factor should be considered or it should
be recognized that the use of the substance as a  food  additive
is  not appropriate.  If  only reversible developmental  effects
are  seen, such as retarded skeletal and soft tissue development
or decreased fetal weight, the usual safety factor of 100 may be
applied.

    (c)  Age-related effects in reproduction studies

    Such  studies may demonstrate  different toxic responses  in
young animals compared with older ones.  Metabolic  studies  may
demonstrate  that the differences in sensitivity are due to such
factors  as incomplete development  of enzyme systems  used  for
metabolizing  xenobiotic compounds or differences  in intestinal
flora.   Safety factors should be set on the basis of the target
population.  If young children are likely to consume  the  addi-
tive,  the ADI should be  based on the no-observed-effect  level
from the phase of the study in which young animals were exposed,
if  the no-observed-effect  level was  lower than  in the  adult
phase.   If, on the other  hand, it is shown  that children will
not be exposed to the additive, it may be appropriate to set the
ADI  on the basis of the no-observed-effect level established in
the adult phase of the study.

    (d)  Finding of carcinogenicity

    Carcinogens vary in the magnitude of risk they  present  for
man,  because they act via different mechanisms.  Even though no
basis   exists  for  the   exact  extrapolation  of   risk  from
experimental  animals to man, the degrees of risk from different
carcinogens  can often be inferred from the data.  However, with

the  present state  of knowledge,  it would  be  appropriate  to
consider the use of a carcinogenic substance as  an  intentional
food  additive  only  under very  restricted circumstances.  For
example,  if cancer is shown  to be a secondary  effect, such as
bladder  tumours occurring secondary to the induction of bladder
stones,  and there is  evidence of a  threshold below which  the
additive is safe, then it would be appropriate to use  a  safety
factor  for determining the safe  level of use of  the additive.
Under  extenuating circumstances, such as  an unambiguous demon-
stration that the health benefits exceed the risk, it  may  also
be possible to use a carcinogenic additive.

    (e)  If reasons exist for setting a lower safety factor

    If  toxicity and dose-response  effects in human  beings are
known,  such data should take precedence over extrapolation from
animal  studies; a 10-fold safety factor would be appropriate if
there  is no evidence that human sensitivity to the agent varies
more  than 10-fold among individuals.  A lower safety factor may
also  be appropriate when the additive is similar to traditional
foods,  is  metabolized  into normal  body  constituents, and/or
lacks  overt  toxicity.  Also,  a  100-fold safety  factor often
would not provide a high enough level of nutrients  required  to
satisfy nutritional needs and to maintain health  (toxicity  for
some essential nutrients such as Vitamin A, Vitamin  D,  certain
essential  amino acids, and iron  may be reached at  levels less
than 10 times higher than those recommended for  optimal  nutri-
tion).   A substance  that serves  as a  significant  source  of
energy  in  the human  diet obviously cannot  fit into the  con-
straints of a 100-fold safety factor.

    The use of standardized safety factors based on no-observed-
effect  levels for establishing the  acceptable level of use  of
food  additives  is  a crude  procedure,  given  the known  wide
variability  in toxic responses.  For example, the nature of the
dose response usually is not used.  In part, this is  a  reflec-
tion  of the fact that good dose-response data are not available
for many compounds.  Attempts to use the dose-response behaviour
of  compounds in establishing quantitative  end-points must con-
tend with this limitation.

    In  the broadest sense,  the procedures used  by JECFA  take
into  account the nature of  the biological effects observed  in
animal bioassays only to the extent that a distinction  is  made
between  carcinogens and non-carcinogens, i.e.,  ADIs are estab-
lished  for  non-carcinogens,  while most  carcinogens  are con-
sidered to unacceptable for use as intentional  food  additives.
Otherwise,  the nature of the observed effect is not an explicit
component  of  the  quantitative assessment  of  food additives.
However,  the nature of  the effect and  a determination of  its
significance are often implicitly considered by scientists, when
reviewing the data.

    JECFA  should take these and other factors into account when
determining   acceptable   daily  intakes   of  food  additives.
However,  in situations  in which  little is  known  beyond  the

empirical finding of toxicity in animal studies, the traditional
approach  for  calculating ADIs  would  seem to  be appropriate.
This may be an issue for future consideration by JECFA.

5.5.3.  Toxicological versus physiological responses

    When  analysing  a toxicological  study  and setting  a  no-
observed-effect  level,  a  distinction must  be  drawn  between
reversible changes that are due entirely to normal physiological
processes  or  homeostasis-maintaining mechanisms,  and to toxic
responses  themselves  (section  5.1).  Examples  of  the former
include:   laxative  effects  from osmotic  or  faecal overload,
liver  hypertrophy  and  microsomal enzyme  induction  from high
doses  of substances metabolized  by the liver,  decreased  body
weight  gain or  caecal enlargement  from high  levels  of  non-
nutritive  substances,  alteration  in  renal  weight  that   is
directly related to the amount of water being processed  by  the
kidney, and decreased growth rate and food  consumption  related
to  the  dietary  administration of  an  unpalatable  substance.
However, care must be taken in interpreting these  changes,  and
they  should not automatically be dismissed as being unimportant
from  a toxicological point  of view.  For  example,  microsomal
enzyme induction in the liver may result in alterations  in  the
metabolism of compounds unrelated to the administered substance,
which could result in a toxic effect.  A decrease in the rate of
body-weight  gain coupled with a corresponding reduction of food
intake  could  be due  to toxic anorexia,  rather than a  palat-
ability defect.

    The  dose at which the effect occurs should be compared with
the  amount of the substance consumed by human beings.  Thus, it
would  ordinarily be acceptable to permit the use of a substance
that causes diarrhoea only at very high levels of consumption in
rats,  but  the  use of  such  a  substance should  be  severely
restricted  or not permitted  if it causes  diarrhoea at  normal
levels of consumption in human beings.  Sometimes, physiological
adaptation may progress through overload to frank toxicity.

    Further studies are indicated in situations in which  it  is
difficult  to  draw a  clear distinction between  a toxic and  a
physiological response.  Special studies such as paired feeding,
caloric  balance comparisons between food  consumption and body-
weight  gain, or,  in the  case of  reproduction studies,  cross
fostering,  can be performed  to decide issues  such as  reduced
food  intake and reduced body-weight gain related to unpalatable
test  substances.  Metabolic and pharmacokinetic  studies may be
of  use in providing information on the distribution of the test
compound  and its metabolites or  the dose at which  a change in
metabolism occurs.

5.5.4.  Group ADIs

    If  several compounds that display similar toxic effects are
to  be considered for  use as food  additives, it may  be appro-
priate in establishing an ADI to consider the group of compounds
in order to limit their cumulative intake.  For  this  procedure

to be feasible, the additives must be in the same range of toxic
potency.   Flexibility should be  used in determining  which no-
observed-effect  level is to be used in calculating the ADI.  In
some  cases, the average  no-observed-effect level for  all  the
compounds  in the group  may be used  for calculating the  group
ADI.  A more conservative approach is to base the group  ADI  on
the  compound  with  the lowest  no-observed-effect  level.  The
relative  quality and length of studies on the various compounds
should be considered when setting the group ADI.  When  the  no-
observed-effect  level for one of  the compounds is out  of line
with the others in the group, it should be treated separately.

    When  considering the use of a substance that is a member of
a series of compounds that are very closely  related  chemically
(e.g.,  fatty acids), but for which toxicological information is
limited,  it may be possible to base its evaluation on the group
ADI established for the series of compounds.  This procedure can
only be followed if a great deal of toxicological information is
available  on at least one member of the series and if the known
toxic  properties of the  various compounds fall  along a  well-
defined continuum.  Interpolation, but not extrapolation, can be
performed by this procedure.  The use of this procedure by JECFA
represents  one of the few situations in which the Committee has
used  structure/activity  relationships  in its  safety  assess-
ments.

    In some instances, group ADIs can be  established  primarily
on  the basis of metabolic information.  For example, the safety
of esters used as food flavours could be assessed on  the  basis
of  toxicological  information  on their  constituent  acids and
alcohols, provided that it is shown that they are quantitatively
hydrolysed in the gut.

    The  calculation of  a group  ADI is  also  appropriate  for
compounds  that cause additive  physiological or toxic  effects,
even if they are not closely related chemically.   For  example,
it  may be appropriate  to establish a  group ADI for  additives
such  as bulk  sweeteners that  are poorly  absorbed  and  cause
laxation.

5.5.5.  Special situations

    There are occasions when JECFA considers the use of  an  ADI
in numerical terms not to be appropriate.  This situation arises
when the estimated consumption of the additive is expected to be
well below any numerical value that would ordinarily be assigned
to it.  Under such circumstances, JECFA uses the term  "ADI  not
specified".   The Committee defines this  term to mean that,  on
the  basis  of  available data  (chemical,  biochemical, toxico-
logical,  and other), the total  daily intake of the  substance,
arising  from  its use  at the levels  necessary to achieve  the
desired  effect and from its acceptable background in food, does
not,  in  the opinion  of the Committee,  represent a hazard  to
health.   For  that reason,  and for the  reasons stated in  the
individual evaluations, the establishment of an ADI in numerical
form is not deemed necessary (e.g., 1, Annex II).   An  additive

meeting  this criterion must be  used within the bounds  of good
manufacturing  practice,  i.e.,  it  should  be  technologically
efficacious  and should be used at the lowest level necessary to
achieve this effect, it should not conceal inferior food quality
or adulteration, and it should not create a  nutritional  imbal-
ance  (16, pp. 10-11).  That  the background occurrence  of  the
chemical  must be taken  into account in  the evaluation of  its
safety was articulated by the WHO Scientific Group on Procedures
for  Investigating Intentional and Unintentional  Food Additives
(2, p. 7).

    JECFA has encountered several situations in which either the
body  of data before it  on a new additive  was limited, or  the
safety  of a food  additive for which  the Committee  previously
assigned an ADI was brought into question by the  generation  of
new  data.  When the Committee  feels confident that the  use of
the substance is safe over the relatively short period  of  time
required  to generate and evaluate  further safety data, but  is
not  confident that its  use is safe  over a lifetime,  it often
establishes  a "temporary" ADI, pending the submission of appro-
priate  data to resolve the  safety issue on a  timetable estab-
lished  by  JECFA.   When  establishing  a  temporary  ADI,  the
Committee  often uses a higher-than-usual safety factor, usually
increasing  it by a factor of 2.  The additional biochemical and
toxicological  data required for the establishment of an ADI are
clearly  stated,  and a  review of these  new data is  conducted
before the expiration of the provisional period.

    This  approach  seems  to  have  worked  reasonably  well in
practice  in that it  has encouraged necessary  research without
creating  any known safety  problems.  In many  cases, long-term
studies are requested, but timetables are not met,  which  means
that  JECFA has had to extend temporary ADIs for long periods of
time.   JECFA has withdrawn ADIs,  in instances where data  were
not forthcoming, as a safety precaution.

5.5.6.  Comparing the ADI with potential exposure

    When  establishing ADIs, collateral exposure  information is
often  useful for determining  the relationship between  the two
values.  The agreement between exposure and acceptable daily in-
take helps to determine whether an "ADI not specified" should be
established.  Exposure information is also indispensable when:

    (a)  performing  risk assessments for food  contaminants and
         processing aids; and

    (b)  assessing  the safety of  added substances that  may be
         naturally  present in food to  determine their relative
         contributions to the diet (17, pp. 8-10).

    In  order  to  accurately compare  exposure  and  acceptable
intake,  similar  assumptions should  be  used for  making  each
estimate or, at least, the differences and similarities  in  the
estimates  should be  understood.  For  example, if  an  ADI  is
computed from lifetime dosage, then the estimated human exposure

should  represent lifetime exposure to the additive.  Sometimes,
acceptable intakes are computed for specific age groups  or  for
certain  dosage  conditions  when short-term  exposure should be
limited, such as with certain food additives that cause laxative
effects  at  high dose  levels.   Under such  circumstances, the
estimated  human exposure should represent the same age group or
dosage  conditions.  In practice, however, estimates of exposure
do not represent exposure for individual consumers in  the  same
way  that  toxicological  data represent  dosages for individual
animals.  Data bases on food and food additive  intakes  provide
composite data for subpopulations, such as the  average  dietary
habits of a particular nation's population.

    For  effective comparison of exposure estimates with accept-
able intakes, the assumptions used to compute exposure estimates
should  always be  stated.  Data  on the  functional  use(s)  of
intentional food additives and information on approaches used to
compute  intake estimates, such  as analytical studies  on  food
constituents  or migration (carry-over) models  for certain con-
taminant situations, should be provided if possible.

    Each estimate of exposure represents a facet of actual human
exposure  and, thus, each estimate  represents useful scientific
data.   However, it is  not possible to  describe specific  pro-
cedures for estimating exposure for all food additive  and  con-
taminant situations.  However, JECFA is able to provide guidance
by  describing the types of estimation procedures that have been
accepted  by previous Committees, which are discussed in section
3.1.1

6.  PRINCIPLES RELATED TO SPECIFIC GROUPS OF SUBSTANCES

6.1.  Substances Consumed in Small Amounts

    Many  of the substances that come before JECFA for its eval-
uation  are present  in food  in only  trace  amounts.   Testing
requirements  generally  take  these low  exposures into account
(section  3.1).   However,  as discussed  below,  special safety
concerns  are  raised by  the use of  many of these  substances,
despite the low exposures to both the parent compounds and their
residues.

    In  some cases, these substances have no technological func-
tion in the food itself.  Some are used in food processing.  For
example,  residues  from extraction solvents used  inter alia  in
extracting fats and oils, in defatting fish and other meals, and
in  decaffeinating coffee and  tea may be  present in the  final
food  product because of incomplete  removal.  The same is  true
for enzymes and immobilizing agents (and their  residues)   used
in  immobilized enzyme preparations.  Residues  arising from the
use  of xenobiotic anabolic agents and from the use of packaging
materials may also occur in food.

    Residues belonging to all these classes of  substances  have
been  evaluated by JECFA, and the Committee has developed guide-
lines concerning their safety evaluation.  The guidelines, which
are reproduced in Annex III, are intended to serve  as  examples
of guidance by JECFA for evaluating these specific categories of
substances.   Further discussion of  such substances and  others
follows in section 6.1.1.

    Flavouring  agents constitute a category  of substances that
have  a functional effect  in food, but  are generally added  in
small  amounts.  The safety evaluation of flavours has presented
special problems for JECFA, and these are discussed in detail in
section 6.1.2.

 6.1.1.  Food contaminants

    JECFA  has considered the  presence of food  contaminants on
many  occasions since 1972, when mercury, lead, and cadmium were
first  assessed (60, pp.  11-24).  These food  contaminants have
included,  in  addition  to heavy  metals, environmental contam-
inants such as mycotoxins, impurities arising in food additives,
solvents  used in food processing,  packaging material migrants,
and  residues arising  from the  use of  animal  feed  additives
and/or  veterinary drugs.  Each of these classes of food contam-
inants  possesses its own unique  characteristics and evaluation
requirements.  Thus, JECFA has recognized through the years that
evaluation  principles should pertain  to classes or  groups  of
contaminants  rather than to  food contaminants  in toto.   JECFA
has published guidelines, which are reproduced in Annex III, for
the  evaluation  of  various  classes  of  contaminants;   these
guidelines are still valid.

    At  the  time that  JECFA  considered mercury,  cadmium, and
lead,  in  1972,  it  established  the  concept  of "provisional
tolerable weekly intake" (PTWI), which is a departure  from  the
traditional  ADI concept (60, pp. 9-11).  JECFA has continued to
use this concept, with some modifications, ever since.

    ADIs  are intended to be  used in allocating the  acceptable
amounts  of  an  additive for  necessary technological purposes.
Obviously,  trace contaminants have no intended function, so the
term  "tolerable"  was  seen as  a  more  appropriate term  than
"acceptable",  which signifies permissibility rather than accep-
tability  for the intake of  contaminants unavoidably associated
with  the  consumption  of otherwise  wholesome  and  nutritious
foods.

    In  this convention, tolerable  intakes are expressed  on  a
weekly  basis, because the  contaminants given this  designation
may  accumulate within the body  over a period of  time.  On any
particular  day,  consumption  of food  containing above-average
levels  of the contaminant may exceed the proportionate share of
its  weekly  tolerable  intake.  JECFA's  assessment  takes into
account  such daily variations, its real concern being prolonged
exposure to the contaminant, because of its ability  to  accumu-
late within the body over a period of time.

    The  use of the  term "provisional" expresses  the tentative
nature  of the evaluation,  in view of  the paucity of  reliable
data on the consequences of human exposure at levels approaching
those with which JECFA is concerned.

    A tolerable intake, as defined above, represents the maximum
acceptable level of a contaminant in the diet; the  goal  should
be  to limit exposure to the maximum feasible extent, consistent
with  the PTWI.  However,  potent carcinogens, such  as  certain
mycotoxins,  cannot be made to fit within the confines of a PTWI
because, using the traditional approach, safe levels  cannot  be
set.   JECFA addressed  this issue  in 1978  and introduced  the
concept  of an "irreducible  level", which it  defined as  "that
concentration of a substance which cannot be eliminated  from  a
food without involving the discarding of that  food  altogether,
severely  compromising the ultimate  availability of major  food
supplies" (32, pp.  14-15).

    Another  JECFA end-point, the "provisional maximum tolerable
daily  intake"  (PMTDI) has  been established for  food  contam-
inants that are not known to accumulate in the body, such as tin
(23),  arsenic (24), and styrene (1).  The value assigned to the
PMTDI represents permissible human exposure as a result  of  the
natural  occurrence of the  substance in food  and in  drinking-
water.

    In   1982,  JECFA  decided  to  change  the  methodology  of
assessment  for trace elements that are both essential nutrients
and unavoidable constituents of food, such as copper  and  zinc.
The Committee concluded that, in such situations, the use of one

number  for expression of  the tolerable intake  was not  suffi-
ciently  informative, so expression  in a range  was  instituted
(23,  Annex II).  The lower value represents the level of essen-
tiality  and the upper value  the PMTDI.  Thus, the  upper value
should  not  be  construed  as  representing  its  normal  daily
requirement.

    With  the  use  of increasingly  sensitive  analytical tech-
niques,  it is becoming clear that many food additives also con-
tain trace levels of carcinogenic contaminants.  JECFA addressed
this issue in 1984 when it considered the low-level migration of
carcinogenic  contaminants  from  food packaging  materials (1).
The Committee did not consider it appropriate to  allocate  ADIs
on  the basis of the  information available.  For further  eval-
uation,  the twenty-eighth JECFA stated  that it would need  the
following information:

    (a)  the lowest levels of potential migrants from within the
         polymeric system(s) that are technologically attainable
         with  improved manufacturing processes for food-contact
         materials;

    (b)  the resulting levels of the migrants in foods;

    (c)  the intake of the foods; and

    (d)  the  most  appropriate  statistical  design  that  will
         enable  the implications for  health to be  interpreted
         from adequate and relevant toxicological data.

In  the meantime, the Committee recommended that "human exposure
to  migrants from food-contact  materials be restricted  to  the
lowest levels technologically attainable" (1, p. 23).

6.1.2.  Food flavouring agents

    The  special problems associated with  the safety evaluation
of  food  flavouring agents  have been apparent  to JECFA for  a
considerable  time  and  the  tenth  meeting  of  the  Committee
recommended that "further meetings of JECFA should  be  convened
to  draw up specifications for flavouring substances, . . . used
as  food additives, and  to evaluate the  toxicological  hazards
involved  in their  use" (29).   The Committee  referred to  the
extent  and  magnitude of  these problems in  the report of  the
eleventh  meeting (41) but, despite  the time that has  elapsed,
the  problems remain largely unresolved.  Much of the difficulty
arises  from the fact that a very large number of substances are
used as food flavouring agents, many of which occur  in  natural
products,  but  their level  of use is  generally low and  self-
limiting.    The   compounds  and   other  materials  (extracts,
oleoresins,  essential oils) used to impart flavour to foods can
be classified into four groups (19), viz:

    (a)  artificial  substances  unlikely to  occur naturally in
         food;

    (b)  natural  materials not normally consumed as food, their
         derived  products, and the  equivalent nature-identical
         flavourings;

    (c)  herbs  and  spices,  their derived  products,  and  the
         equivalent nature-identical flavourings; and

    (d)  natural  flavouring substances obtained  from vegetable
         and  animal  products  and normally  consumed  as  food
         whether  processed  or  not, and  their synthetic equi-
         valents.

    Most  of  these  food  flavouring  materials  have  not been
subjected  to detailed and comprehensive  toxicity tests, though
with  the flavouring agents in  classes (c) and (d)  above there
may  be a long history of use and limited evidence of safety-in-
use.   In some cases, there may also be evidence of adverse con-
sequences of human exposure such as hypersensitivity  and  idio-
syncratic  intolerance, which have been observed with capsaicin,
zingiberin,  and menthol.  However, the natural origin of a food
flavouring agent is no guarantee of safety, nor does traditional
use of a material constitute unequivocal evidence of safety; the
flavour  saffrole is of natural  origin (oil of sassafras)   and
had a long history of use before it was demonstrated to be hepa-
toxic  and carcinogenic (22).  Consequently,  natural flavouring
compounds  cannot be exempted from  the toxicological evaluation
applicable  to food flavouring  agents in general.   Conversely,
there is no basis for assuming that compounds that interact with
gustatory  or olfactory receptors  are more likely  to  interact
with  other physiological receptor sites than non-flavoured com-
pounds.   In principle, therefore, the safety evaluation of food
flavouring  compounds is similar  to that for  other food  addi-
tives, and this fact was recognized in the twentieth  report  of
JECFA  (19).  However, this Committee  concluded that evaluation
should  be  flexible  and "may  require  extensive toxicological
testing  or be made simply from available data".  In the evalua-
tion process, the work of other bodies, such as national govern-
ments and the Council of Europe, should be considered.

    In  view of the very large number of substances used as food
flavouring agents and the fact that they are  generally  applied
in low and self-timing concentrations in foods, it is considered
impractical  and unreasonable to require that each food flavour-
ing  material be  subjected to  the same  and extensive  toxico-
logical  evaluation within a  reasonable period (32).   Further-
more,  human data may  be available that  have a bearing  on the
extent of toxicity testing required and the urgency  with  which
data  from such  tests are  needed.  Thus,  there is  a need  to
develop  a  rational basis  for  allocating priorities  for  the
testing  and safety evaluation of food flavouring agents and for
determining  the  extent  of  the  testing  required.   Previous
Committees  have drawn attention to  this need on several  occa-
sions (19, 32, 31, 24).  Several of the factors  that  influence
the  allocation of priorities are also relevant to the extent of
testing  required and  some of  these are  discussed in  general

terms  elsewhere in this report (sections 3, 5.2, and 5.4).  The
following   discussion  is  specifically  concerned   with  food
flavouring compounds.

    Factors  that  should be  considered  in the  allocation  of
priorities  and  the  determination  of  the  extent  of testing
required include:

    (a)  nature and source of the material;

    (b)  data  on  usage  and on  the  extent  and frequency  of
         exposure of the average consumer and of subpopulations,
         who  may  be  highly exposed  (including  exposure from
         natural  sources such as herbs and spices but excluding
         bizarre eating behaviour);

    (c)  structure/activity  relationships and the similarity of
         compounds  of known toxicological and biochemical prop-
         erties;

    (d)  similarity to compounds of known biological activity in
         metabolism and pharmacokinetics;
    (e)  information  from short-term tests for mutagenicity and
         clastogenicity;

    (f)  prior experience of human use; and

    (h)  toxicological  status/regulatory status previously det-
         ermined  by  national  regulatory  agencies  or  supra-
         national  organizations  such as  the European Economic
         Community  (EEC) Scientific Committee for Foods and the
         International Agency for Research on Cancer (IARC).

    Information  on  the nature  and  source of  the  flavouring
material  is  clearly  necessary  in  order  to  assess  overall
exposure to its components, and such information may  assist  in
the  allocation  of priorities  and  level of  testing required.
Data obtained on synthetic compounds may assist in  the  evalua-
tion  of limited data on extracts or essential oils that contain
the compounds naturally.  For example, knowledge of  the  compo-
sition  of an  extract or  essential oil  may indicate  that  no
further  testing  is necessary;  conversely,  the presence  of a
known  toxic compound  as a  major component  in an  extract  or
essential  oil (e.g., saffrole in  oil of sassafras, thujone  in
oil  of wormwood, tansy, etc.)  may alter the priority  given to
that  extract  or  essential oil  and  determine  the nature  of
specific  toxicity tests required.  It must be stressed that any
assessment  of  priorities will  depend  on the  availability of
adequate specifications which, in the case of  complex  mixtures
such  as spice extracts,  may include maximum  limits for  known
toxic constituents.

    In  considering the extent of exposure, the eleventh meeting
of  JECFA (41) suggested that flavouring compounds with an esti-
mated  per capita  consumption exceeding 3.65 mg per  annum (sus-
tained  intake exceeding 10 g per day), and/or use in food at a

level higher than 10 mg/kg food, should be given priority.  How-
ever, exposure must be considered in the light of  other  avail-
able  information and it is inappropriate to classify flavouring
compounds for priority evaluation solely on the basis  of  expo-
sure levels.

    In the absence of other data or considerations,  a  combina-
tion of exposure levels and structure/activity relationships can
be  used to establish both  priorities and requirements for  the
level of testing.  A number of schemes that take  these  factors
into account have been proposed to accomplish this (34 - 38).

    The  general  principles regarding  structure/activity rela-
tionships  are discussed in section 3.1.2.  With particular ref-
erence  to food flavouring compounds, these principles have been
applied already by JECFA in a limited number of cases,  both  in
relation  to  high  levels  of  concern  (flavouring   compounds
structurally-related  to  saffrole)  and acceptance  of  limited
toxicological  data (simple structural analogues, homologues, or
derivatives  such  as esters).   In  the latter  case,  detailed
toxicological data on one member of a group of related compounds
together  with metabolic and pharmacokinetic data can be used to
allocate an ADI or group ADI to structurally  related  compounds
without  the need for  further testing.  On  this basis, a  sub-
stantial number of esters used as food flavouring  agents  would
warrant  the  allocation  of a  low  priority  for  testing  and
acceptance for an ADI on the basis of metabolic  studies  alone.
This  would be true in  cases in which flavouring  compounds are
shown  to be rapidly  and quantitatively hydrolysed  to  toxico-
logically  known alcohols and  acids, the safety  of which  have
been established.

    The  use of short-term tests for mutagenicity/clastogenicity
cannot at present be considered a substitute for carcinogenicity
bioassays  and hence negative results  in such tests need  to be
considered  in the light of the total toxicological data base in
allocating  priorities.  In general, such negative results would
not  justify  waiving  the requirement  for  long-term  carcino-
genicity studies on food flavouring compounds, but they  may  be
useful  supplementary information to that discussed above.  Con-
versely,  positive results in  a suitable battery  of short-term
mutagenicity/clastogenicity   tests  would  indicate   a  higher
priority for more detailed testing.

    Prior  experience of human use will influence the allocation
of  priorities,  as  indicated  earlier,  and  even  the limited
evidence  of safety-in-use may  support judgements based  on the
other  criteria discussed above, to  allocate a low priority  to
testing  or to accept  limited toxicological data.   Conversely,
observations of human idiosyncratic intolerance and/or allergies
would indicate a need for adequate information relating  to  the
extent and severity of the problem.

    Systematic  treatment of the above  data (structure/activity
relationships,  exposure, and human data) and the application of
reasonable  criteria for the  adequacy of existing  data,  could

provide  a useful approach in focusing effort on substances that
should  receive priority from a scientific point of view.  Other
criteria  that influence the  priority that JECFA  allocates  to
food  flavours include requests from, or previous evaluation by,
national governments, the European Economic Community Scientific
Committee  for Foods, the  Codex Alimentarius Committee  on Food
Additives,  and the need  to re-evaluate previously  established
temporary  ADIs.  Earlier JECFAs have recommended the setting up
of  a working  group of  experts specifically  to  consider  the
allocation  of priorities for the testing and evaluation of food
flavours,  and this recommendation  still appears to  be  appro-
priate.   There is an  implicit need for  adequate resources  to
perform this task and cooperation with other  interested  groups
to avoid wasteful duplication of effort.

6.2.  Substances Consumed in Large Amounts

    The  safety assessment of  substances that are  consumed  in
relatively  large amounts presents a number of special problems.
Such  materials  include  defined chemical  substances  such  as
sorbitol and xylitol (23, 24), modified food ingredients such as
modified starches (23), and foods from novel sources.

    The  safety assessment of such substances should differ from
that of other food additives, such as colouring  and  flavouring
agents, and antioxidants, for the following reasons:

    (a)  Many  will have a  high daily intake  and, thus,  minor
         constituents  and processing impurities assume greater-
         than-usual significance.

    (b)  Even though they are often structurally similar or even
         identical to natural products used as food and thus may
         appear   to  be  of  low  toxicity,  many  may  require
         extensive toxicity testing, because of their high daily
         intake.

    (c)  Some may be metabolized into normal body constituents.

    (d)  Some substances, particularly foods from novel sources,
         may replace traditional foods of nutritional importance
         in the diet.

    (e)  Many  are complex mixtures rather than defined chemical
         substances.

    (f)  The  difference between the quantity that can be fed to
         animals  in feeding tests  and the amount  consumed  by
         human beings is often relatively small.

6.2.1.  Chemical composition, specifications, and impurities

    Thorough  chemical  analysis  should be  performed  on high-
consumption  substances to measure  potential impurities and  to
provide  information  on  nutritional adequacy,  especially when
such substances replace traditional food.

    It  is not  possible to  provide a  checklist  of  necessary
chemical   studies  to  cover  all  high-consumption  compounds.
However, the substance should be subjected to a  full  proximate
analysis and particular attention should be paid to  the  points
discussed in the following paragraphs.

    Because  the  intake  of undesirable  impurities concomitant
with  the  intake  of high-consumption  materials is potentially
high,  special effort should be made to identify the impurities.
Information  on the production process,  including the materials
and procedures involved, will point to the types of contaminants
for which limits may need to be specified.   The  specifications
should be accompanied by details of product variability  and  of
the  analytical  methods used  to  check the  specifications and
details  of the  sampling protocols.   If the  substance  is  so
complex  that  comprehensive product  specifications on chemical
composition  are impracticable (as it  might be for a  microbial
protein), the description of the substance in the specifications
may include relevant aspects of its manufacturing  process.   If
manufacturing data are based on production on a pilot scale, the
manufacturer  should demonstrate that, when produced in a large-
scale plant, the substance will meet the  specifications  estab-
lished on the basis of pilot data.

    The  permissible  limits for  impurities  may in  some cases
correspond  to the levels accepted  for natural foods that  have
similar  structure  or  function, or  that  are  intended to  be
replaced by the new material.  If the substance is prepared by a
biological  process,  special attention  should  be paid  to the
possible occurrence of natural toxins (e.g., mycotoxins).

    The substance should be analysed for the presence  of  toxic
metals.  Depending on the intended use, analysis for  metals  of
nutritional significance may also be appropriate.

    If  the  nature of  the  substance or  manufacturing process
indicates the possible presence of naturally-occurring or adven-
titious  antinutritional  factors (phytate,  trypsin inhibitors,
etc.),  or toxins (haemagglutinins, mycotoxins, nicotine, etc.),
the product should be analysed for them  specifically.   Biolog-
ical  tests, either as part of the nutritional evaluation in the
case  of  enzyme inhibitors  or more specifically  as part of  a
mycotoxin  screening programme, will provide useful back-up evi-
dence concerning the presence or absence of these contaminants.

    Finally,  if under the intended  conditions of use the  sub-
stance  may be unstable or is likely to interact chemically with
other food components (e.g., degradation or rearrangement of the
substance  during heat processing),  data should be  provided on
its  stability  and reactivity.   The  various tests  should  be
conducted  under conditions relevant to the use of the substance
(e.g.,  at the acidity and temperature of the environment and in
the presence of other compounds that may react).

6.2.2.  Nutritional studies

    With  some substances, particularly with novel foods, nutri-
tional studies may be necessary to forecast the likely impact of
their introduction on the nutritional status of  consumers.   In
addition  to affecting the nutritional content of the diet, such
substances   may   influence  the   biological  availability  of
nutrients  in  the diet.   The  nutritional consequences  of the
introduction  of such a substance in the diet can only be judged
in  the light of  information about its  intended use.  As  much
information  as  possible  should therefore  be  obtained  about
potential markets and uses, and the likely  maximum  consumption
by particular subpopulations should be estimated.

6.2.3.  Toxicity studies

    When  testing  high-consumption  additives,  animals  should
generally  be fed the  highest levels possible  consistent  with
palatability  and nutritional status.  Therefore,  before begin-
ning  such studies, it  is desirable to  investigate the  palat-
ability of the test diet in the test animals.  If a palatability
problem  is encountered,  it may  be necessary  to increase  the
amount  of the test substance  to the required level  gradually.
Paired-feeding  techniques should be used, if the problem cannot
be overcome.  It should always be borne in mind that  there  are
practical  limits to the  amounts of certain  foods that can  be
added to animal diets without adversely affecting  the  animal's
nutrition and health.

    To  ensure that the nutritional status of the test animal is
not distorted, the test and control diets should have  the  same
nutritive  value in terms of both macronutrients (e.g., protein,
fat, carbohydrate, and total calories) and micronutrients (e.g.,
vitamins and minerals).  When feeding substances at high levels,
it  is  usually advisable  to  formulate diets  from  individual
ingredients  (rather than adding the test material to a standard
laboratory  diet)  to provide  the same nutrient  levels in  the
control  and test diets.  Comprehensive nutrient analyses of the
test and control diets should be performed to ensure  that  they
are comparable nutritionally.  Sometimes nutritional studies are
advisable  before toxicological studies are  performed to ensure
that test diets are correctly balanced.  Without due  regard  to
nutritional  balance,  excessive  exposure may  investigate  the
toxicological  end-points of long-term dietary  imbalance rather
than the toxic effects of the substance.

    The establishment of a precise no-observed-effect level will
not  usually be feasible on account of the relative non-toxicity
of   high-consumption  additives  and  the  impracticability  of
achieving  an  adequate  safety margin  between the no-observed-
effect  level in animals  and the expected  consumption of  such
substances  by  human  beings.  Therefore,  it  is  particularly
important  that the variables  for assessing the  safety of  the
substance,  such  as body  weight,  food and  water consumption,
haematological parameters, ophthalmology, blood chemistry, urine

analysis, faecal analysis, mineral and vitamin excretory levels,
etc.,  are chosen  carefully to  include monitoring  of all  its
possible toxic effects.

    Metabolic studies are useful and necessary for assessing the
safety  of  high-consumption additives.   With complex mixtures,
studies  on the  metabolic fate  of every  constituent would  be
impracticable.  However, if contaminants or minor components are
suspected as the cause of toxicity, their metabolism  should  be
investigated.  Consideration should also be given to the effects
that the new constituents may have on the ability of the host to
withstand  other  toxic  agents,  for  example,  effects  on the
metabolism of xenobiotic compounds.  If the material, or a major
component of it, consists of a new chemical compound  that  does
not  normally occur in  the diet (e.g.,  a novel  carbohydrate),
studies  of the  metabolic fate  of the  new compound  would  be
appropriate.

    If  biochemical  and metabolic  studies  show that  the test
material is completely broken down in the food or in the gastro-
intestinal tract to substances that are common dietary  or  body
constituents,  then other toxicity studies may not be necessary.
The results of metabolic studies can stand on their own if it is
shown:  that  breakdown  into these  common  constituents occurs
under the conditions of normal consumption of the material, that
the material contributes only a small proportion of these common
constituents in the daily diet, and that side  reactions  giving
rise to toxic products do not occur.

    Analysis of urine and faeces may provide  important  inform-
ation relating to changes in normal excretory  functions  caused
by  the  test  substance.  For  example,  the  gut flora  may be
altered, or preferential loss of a mineral or vitamin may occur,
resulting  in detrimental  effects on  the health  of  the  test
animals.   If the substance is  incompletely or not degraded  by
the  digestive enzymes of  the stomach or  the small  intestine,
appreciable  concentrations may be found in the faeces or in the
distal gut compartments.  As a result, changes in the absorption
of  dietary constituents or changes in the composition and meta-
bolic activity of the intestinal flora may be observed.  Because
of  species-dependent  anatomical  differences in  the digestive
tract and because of considerable differences in the composition
of the basal diet, such effects may occur only in man but not in
rodents,   or  vice versa.   Therefore,  short-term   biochemical
studies  should be performed in animals and man (if possible) in
which variables likely to be affected by the test  compound  are
examined in detail.  It is especially important  to  investigate
questions  relating to whether the eventual effects are progres-
sive  or transient and whether they occur in subjects exposed to
the  compound for the first time and/or in subjects adapted to a
daily  intake of the substance.  Clearly, no standard design for
such studies can be devised.  Only a thorough knowledge  of  the
nutritional  and biochemical literature  can serve as  a  guide-
line.

    When  establishing an ADI, the traditional concept of a 100-
fold  safety factor cannot  operate when the  human  consumption
level is high and feeding studies do not produce adverse effects
(except  for effects arising from the physical properties of the
additive,  such as its bulk  and hydrophilicity), even when  the
substance is added to the diet in the maximum  possible  propor-
tion,  consistent with reasonable nutrition.  In such cases, new
approaches are indicated, including setting the ADI on the basis
of a smaller safety factor, which may be permissible  when  fac-
tors  such as similarity  to traditional foods,  metabolism into
normal  body  constituents, lack  of  overt toxicity,  etc., are
considered.  For a compound, such as a bulking agent,  that  may
influence the nutritional balance or the digestive physiology by
its mere bulk and which may be absorbed from the gut only incom-
pletely or not at all, it may be more appropriate to express the
dosage level in terms of the percentage inclusion in  the  diet.
In  using this approach, a direct comparison between the propor-
tion in the human diet, with a small safety factor, can be made.
If several similar types of compounds are likely to be consumed,
a  group ADI (limiting the  cumulative intake)  should be  allo-
cated.

    The results of human studies, which are discussed  in  rela-
tion  to  novel  foods in section 6.2.4, may permit the use of a
lower safety factor than that obtained from animal studies.

    Separate  toxicological tests should be performed on toxico-
logically  suspect impurities or minor components present in the
test  material.  If any observed  toxicity can be attributed  to
one of the impurities or minor components, its level  should  be
controlled by use of specifications and manufacturing controls.

6.2.4.  Foods from novel sources

    Over  the past 25 years,  a series of developments  has made
possible  the  production  of foods  from unconventional sources
(e.g.,  fungal mycelia and  yeast cells) and  foods produced  by
genetic  techniques.  These foods are  intended for consumption,
either directly, or after simple physical modification  to  pro-
vide a more acceptable product.  They may be consumed  in  large
amounts,  even by infants and children, particularly if they are
permitted for use as protein supplements in  otherwise  protein-
deficient diets.

    Complete  chemical identification of such  materials may not
be  feasible, but specifications  are necessary to  ensure  that
levels of potentially hazardous contaminants, such as mycotoxins
and  heavy metals,  and other  substances of  concern,  such  as
nucleic acids, are kept to a minimum.  Toxicological evaluations
must be closely related to well-defined materials,  and  evalua-
tions may not be valid for all preparations from the same source
material, if different processing methods are used.

    When  a novel food is intended to replace a significant por-
tion of traditional food in the diet, its likely impact  on  the
nutritional  status of consumers requires special consideration.

The  influence of the introduction  of the new substance  on the
nutrient  composition of the diet  as a whole should  be identi-
fied,  particularly with respect to groups such as children, the
elderly,  and "captive populations", e.g., hospital patients and
school  children.  In order not  to adversely affect the  nutri-
tional quality of the diet, it may be necessary to  fortify  the
substance with vitamins, minerals, or other nutrients.

    The nutritional value of the novel food should  be  assessed
initially  from its chemical  composition with respect  to  both
macronutrients  and  micronutrients,  taking  into  account  the
effects  of any further  processing and storage.   The  possible
influence  of components in the  novel food, such as  antinutri-
tional factors (e.g., inhibitors of enzyme activity  or  mineral
metabolism),  on the nutritional value or keeping quality of the
remainder of the diet should also be established.

    Depending on the nature and intended uses of the novel food,
studies  in animals  may be  needed to  supplement the  chemical
studies.   If the novel  food is intended  to be an  alternative
significant supply of protein, tests on its protein quality will
be  necessary.  In vivo  studies will  also be needed when  it is
appropriate  to determine (a)  the availability of  vitamins and
minerals  in the novel food in comparison with the food it would
replace;  and (b) any interaction the novel food might have with
other  items of  the diet  that would  reduce the  whole  diet's
nutritional  value.  If the  novel food is  expected to play  an
important role in the diet, it may be necessary to  verify  that
the  results of  animal studies  can be  extrapolated  to  human
beings  by  measuring the  availability  of nutrients  to  human
subjects.

    In  most cases, novel foods constitute a large percentage of
the daily diet in animal studies because they are of a non-toxic
nature.   Therefore,  the  considerations discussed  in  section
6.2.3 apply to the toxicological testing and evaluation of foods
from novel sources.  The sixteenth and seventeenth  meetings  of
the FAO/WHO/UNICEF Protein Advisory Group (PAG) and  the  United
Kingdom  Department of Health  and Social Security  (DHSS)  have
developed  guidelines on the  development and testing  of  foods
from  novel sources, which  should be consulted  for a  detailed
discussion (85 - 87).

    With certain classes of substances, such as foods  that  are
modified  by recombinant DNA or hybridization techniques to pro-
duce  what effectively are new cultivars or modified traditional
foods,  the use of  the ADI is  not appropriate.  However,  with
many  foods produced from novel  sources, the use of  the ADI is
appropriate,  because  these  foods bear  little relationship to
foods  that have been consumed traditionally.  The allocation of
an ADI is useful to permit the establishment  of  specifications
to ensure microbiological purity and to control chemical contam-
inants.

    After the appropriate animal tests have been performed and a
tentative ADI has been established, human volunteer  studies  to

test for specific human effects should be undertaken.  The first
human  study should involve  the feeding of  a single meal  con-
taining the novel food at a known dose level to one volunteer at
a  time.  If no harmful effects are observed with several volun-
teers,  studies involving the  feeding of the  novel food for  a
short  period (initially about four weeks with follow-up studies
of longer duration) should be performed.  Different diets incor-
porating different levels of the novel food should be related to
the anticipated levels of human exposure.  The closest attention
should be paid to matching groups with respect to  age,  height,
weight, sex, alcohol intake, and smoking habits.  In addition to
having  normal control  groups, it  may be  useful  to  organize
studies in which the test groups are fed diets incorporating and
not  incorporating the novel food in sequential periods, so that
each volunteer acts as his own control; blind  crossover  trials
are the most satisfactory.  Once it has been determined that the
novel  food is  tolerated well  by volunteers  at fixed  dietary
levels,  it may  be useful  to feed  it  ad libitum, for a  short
period of time, in order to assess its acceptability.

    If the novel food is intended for use by a certain community
or  section of the  community (e.g., among  a particular  ethnic
group  or by diabetic  patients), at least  one study should  be
conducted in the group of people for whom the food is intended.

    It  may be necessary to conduct allergenicity studies on the
novel  food because of  its composition (e.g.,  if it is  highly
proteinaceous) or because the results of animal or human feeding
studies  suggest that the food might produce hypersensitivity in
some  people.  Important information can be gained by monitoring
the health of workers coming into contact with the  novel  food,
such  as  laboratory staff  and  employees in  the manufacturing
plant.  In order to detect possible allergenicity of  the  novel
food in the general population, it will generally  be  essential
to monitor a large number of people.

    Large-scale  acceptability  and  marketing trials  should be
undertaken  only after the novel  food's safety has been  demon-
strated by the studies indicated above.  It may be  most  useful
to restrict the trial to a defined geographical area.  The local
medical services responsible for the area in which the substance
is  tested  should  be alerted  so  that  they may  take it into
account  when evaluating any  unusual disease patterns  that may
appear during or after the test period.  Because  large  numbers
of  people will be involved in the trials, it may be possible to
obtain  information about rare food  intolerance (e.g., allergic
reactions)  that may  not have  been observed  in earlier  human
studies.  The extent to which health monitoring should  be  per-
formed  will  depend  on the  nature  of  the substance  and the
results of previous toxicological investigations.



REFERENCES

 1. FAO/WHO   (1984)  Evaluation of  certain food additives  and
    contaminants.  Twenty-eighth  report  of the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 710).

 2. WHO   (1967)   Procedures for  investigating intentional and
    unintentional  food  additives.  Report of  a WHO Scientific
    Group,  Geneva,  World  Health Organization  (WHO  Technical
    Report Series No. 348).

 3. WHO    (1974)    Assessment   of  the   carcinogenicity  and
    mutagenicity of chemicals. Report of a WHO Scientific Group,
    Geneva,  World  Health  Organization (WHO  Technical  Report
    Series No. 546).

 4. FAO/WHO   (1961)  Evaluation of the  carcinogenic hazards of
    food  additives. Fifth report  of the Joint  FAO/WHO  Expert
    Committee  on Food Additives (FAO  Nutrition Meetings Report
    Series No. 29; WHO Technical Report Series No. 220)  (out of
    print).

 5. FAO/WHO    (1955)    Joint   FAO/WHO  Expert   Committee  on
    Nutrition.  Fourth  report  (FAO Nutrition  Meetings  Report
    Series No. 9; WHO Technical Report Series No. 97).

 6. FAO/WHO   (1956)  Joint FAO/WHO Conference on Food Additives
    (FAO  Nutrition Meetings Report Series No. 11; WHO Technical
    Report Series No. 107).

 7. FAO/WHO   (1963)   Second  Joint FAO/WHO  Conference on Food
    Additives  (FAO Nutrition Meetings Report Series No. 34; WHO
    Technical Report Series No. 264).

 8. FAO/WHO    (1974)   Report  of   the  Third  Joint   FAO/WHO
    Conference   on   Food   Additives  and   Contaminants  (FAO
    Miscellaneous  Meeting  Reports  Series  No.  ESN:MMS  74/6;
    WHO/Food Add./74.43).

 9. FAO/WHO   (1957)   General  principles governing  the use of
    food  additives. First report  of the Joint  FAO/WHO  Expert
    Committee  on Food Additives (FAO  Nutrition Meetings Report
    Series No. 15; WHO Technical Report Series No. 129).

10. FAO/WHO   (1958)  Procedures for the  testing of intentional
    food  additives to establish  their safety for  use.  Second
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 17; WHO
    Technical Report Series No. 144) (out of print).

11. FAO/WHO   (1965)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: food
    colours  and  some antimicrobials  and antioxidants.  Eighth
    Report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 38; WHO
    Technical Report Series No. 309) (out of print).

12. FAO/WHO   (1966)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    antimicrobials,   antioxidants,   emulsifiers,  stabilizers,
    flour-treatment  agents, acids, and  bases. Ninth report  of
    the  Joint FAO/WHO Expert  Committee on Food  Additives (FAO
    Nutrition  Meetings  Report  Series No.  40;  WHO  Technical
    Report Series No. 339) (out of print).

13. WHO   (1969)  Principles for  the testing and  evaluation of
    drugs for carcinogenicity. Report of a WHO Scientific Group,
    Geneva,  World  Health  Organization (WHO  Technical  Report
    Series No. 426).

14. WHO    (1971)    Evaluation   and  testing   of   drugs  for
    mutagencity:  principles  and  problems.  Report  of  a  WHO
    Scientific  Group,  Geneva,  World Health  Organization (WHO
    Technical Report Series No. 482).

15. WHO   (1967)   Principles  for  the  testing  of  drugs  for
    teratogencity.  Report  of  a WHO  Scientific Group, Geneva,
    World  Health Organization (WHO Technical  Report Series No.
    364).

16. FAO/WHO   (1974)   Toxicological evaluation  of certain food
    additives  with  a  review  of  general  principles  and  of
    specifications.  Seventeenth  report  of the  Joint  FAO/WHO
    Expert  Committee on Food Additives  (FAO Nutrition Meetings
    Report Series No. 53; WHO Technical Report Series  No.   539
    and corrigendum) (out of print).

17. FAO/WHO   (1974)   Evaluation  of  certain  food  additives.
    Eighteenth report of the Joint FAO/WHO Expert  Committee  on
    Food Additives (FAO Nutrition Meetings Report Series No. 54;
    WHO Technical Report Series No. 557 and corrigendum).

18. FAO/WHO   (1975)  Evaluation of certain food additives: some
    food  colours,  thickening  agents, smoke  condensates,  and
    certain  other  substances.  Nineteenth report  of the Joint
    FAO/WHO  Expert Committee on  Food Additives (FAO  Nutrition
    Meetings Report Series No. 55; WHO Technical  Report  Series
    No. 576).

19. FAO/WHO   (1976)   Evaluation  of  certain  food  additives.
    Twentieth  report of the  Joint FAO/WHO Expert  Committee on
    Food  Additives (FAO Nutrition Meetings Report Series No. 1;
    WHO Technical Report Series No. 599).

20. FAO/WHO   (1978)   Evaluation  of  certain  food  additives.
    Twenty-first report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 617).

21. FAO/WHO   (1980)   Evaluation  of  certain  food  additives.
    Twenty-fourth  report of the Joint  FAO/WHO Expert Committee
    on Food Additives (WHO Technical Report Series No. 653).

22. FAO/WHO   (1981)   Evaluation  of  certain  food  additives.
    Twenty-fifth report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 669).

23. FAO/WHO   (1982)   Evaluation of certain  food additives and
    contaminants.  Twenty-sixth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 683).

24. FAO/WHO   (1983)  Evaluation of  certain food additives  and
    contaminants.  Twenty-seventh  report  of the  Joint FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 696 and corrigenda).

25. FAO/WHO   (1962)  Specifications for identity  and purity of
    food     additives    (antimicrobial    preservatives    and
    antioxidants).  Third  report  of the  Joint  FAO/WHO Expert
    Committee  on  Food  Additives.  These  specifications  were
    subsequently  revised  and  published as  Specifications for
    identity  and  purity  of food  additives.  I. Antimicrobial
    preservatives  and antioxidants, Rome, Food  and Agriculture
    Organization of the United Nations (out of print).

26. FAO/WHO   (1964)  Specifications for the identity and purity
    of   food  additives  and  their  toxicological  evaluation:
    emulsifiers,  stabilizers,  bleaching  and maturing  agents.
    Seventh report of the Joint FAO/WHO Expert Committee on Food
    Additives  (FAO Nutrition Meetings Report Series No. 35; WHO
    Technical Report Series No. 281) (out of print).

27. FAO/WHO   (1970)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    food  colours, emulsifiers, stabilizers,  anticaking agents,
    and certain other substances. Thirteenth report of the Joint
    FAO/WHO  Expert Committee on  Food Additives (FAO  Nutrition
    Meetings Report Series No. 46; WHO Technical  Report  Series
    No. 445).

28. GEMS  (1985)  Guidelines for the study of dietary intakes of
    chemical  contaminants,  Geneva,  World Health  Organization
    (GEMS:  Global Environmental Monitoring System)  (WHO Offset
    Publication No. 87).

29. FAO/WHO   (1967)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    emulsifiers  and  stabilizers and  certain other substances.
    Tenth report of the Joint FAO/WHO Expert Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 43; WHO
    Technical Report Series No. 373).

30. FAO/WHO     (1971)     Evaluation    of   food    additives:
    specifications for the identity and purity of food additives
    and their toxicological evaluation: some extraction solvents
    and   certain  other  substances;   and  a  review   of  the
    technological   efficacy   of  some   antimicrobial  agents.
    Fourteenth report of the Joint FAO/WHO Expert  Committee  on
    Food Additives (FAO Nutrition Meetings Report Series No. 48;
    WHO Technical Report Series No. 462).

31. FAO/WHO   (1980)   Evaluation  of  certain  food  additives.
    Twenty-third report of the Joint FAO/WHO Expert Committee on
    Food  Additives  (WHO Technical  Report  Series No.  648 and
    corrigenda).

32. FAO/WHO   (1978)  Evaluation of  certain food additives  and
    contaminants.  Twenty-second  report  of the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 631).

33. FAO/WHO   (1985)  Evaluation of  certain food additives  and
    contaminants.  Twenty-ninth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 733).

34. NCI   (1974)  An automatic procedure  for assessing possible
    carcinogenic  activity of chemicals  prior to testing,  Palo
    Alto,  California, Stanford Research Institute (prepared for
    the  National  Cancer Institute,  Contracts N01-CP-33285 and
    NIH-NCI-71-2045).

35. CRAMER, G.M., FORD, R.A., & HALL, R.L.   (1978)   Estimation
    of  toxic  hazard: a  decision  tree approach.  Food Cosmet.
    Toxicol., 16: 255-276.

36. FOOD SAFETY COUNCIL  (1980)  Proposed system for food safety
    assessment.  Final report of the Scientific Committee of the
    Food Safety Council, Washington DC, Food Safety Council.

37. FDA    (1982)   Toxicological  principles  for   the  safety
    assessment of direct food additives and color additives used
    in  food, Washington DC, Bureau  of Foods, US Food  and Drug
    Adminstration.

38. DHSS   (1982)  Guidelines for  the testing of  chemicals for
    toxicity,  London, Department of Health and Social Security,
    Committee  on  Toxicity  of  Chemicals  in  Food,   Consumer
    Products  and the Environment  (Report on Health  and Social
    Subjects No. 27).

39. FAO/WHO  (1962)  Evaluation of the toxicity of a  number  of
    antimicrobials  and antioxidants. Sixth report  of the Joint
    FAO/WHO  Expert Committee on  Food Additives (FAO  Nutrition
    Meetings Report Series No. 31; WHO Technical  Report  Series
    No. 228) (out of print).

40. VETTORAZZI,  G., ed.  (1980)  Handbook of international food
    regulatory  toxicology. I. Evaluations, New York, SP Medical
    and Scientific Books.

41. FAO/WHO   (1968)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    flavouring  substances and non-nutritive  sweetening agents.
    Eleventh  report of the  Joint FAO/WHO Expert  Committee  on
    Food Additives (FAO Nutrition Meetings Report Series No. 44;
    WHO Technical Report Series No. 383).

42. FAO/WHO  (1972)  Evaluation of food additives: some enzymes,
    modified    starches,   and   certain    other   substances:
    toxicological evaluations and specifications and a review of
    the  technological efficacy of some  antioxidants. Fifteenth
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 50; WHO
    Technical Report Series No. 488).

43. FAO   (1983)   Guide  to  specifications:  general  notices,
    general  methods, identification tests, test  solutions, and
    other   reference  materials,  Rome,  Food  and  Agriculture
    Organization of the United Nations (FAO Food  and  Nutrition
    Paper No. 5, Rev. 1).

44. FAO   (1978)   Guide  to  specifications:  general  notices,
    general  methods, identification tests, test  solutions, and
    other   reference  materials,  Rome,  Food  and  Agriculture
    Organization of the United Nations (FAO Food  and  Nutrition
    Paper No. 5).

45. IARC  (1983)  Approaches to classifying chemical carcinogens
    according   to  mechanism  of  action.  Joint  IARC/IPCS/CEC
    Working   Group  report,  Lyons,  International  Agency  for
    Research  on  Cancer  (IARC Internal  Technical  Report  No.
    83/001).

46. FEDERAL  REGISTER  (1985)  Chemical carcinogens: a review of
    the  science and its  associated principles. Fed.  Reg., 50:
    10371-10442 (March 14).

47. FOX,  T.R. & WATANABE, P.G.  (1985)  Detection of a cellular
    oncogene  in  spontaneous  liver  tumors  of  B6C3F1   mice.
    Science, 228: 596-597.

48. RIBELIN,   W.E.,  ROLOFF,  M.V.,  &   HOUSER,  R.M.   (1984)
    Minimally      functional     rat     adrenal      medullary
    pheochromocytomas. Vet. Pathol., 21: 281-285.

49. ROE,  F.J.C.  &  BAR,  A.   (1985)   Enzootic  and epizootic
    adrenal  medullary proliferative disease of  rats: influence
    of  dietary factors which  affect calcium absorption.  Human
    Toxicol., 4: 27-52.

50. NTP   (1984)   Report of  the  NTP Ad Hoc Panel  on Chemical
    Carcinogenesis  Testing and Evaluation, Washington DC, Board
    of  Scientific  Counselors, National  Toxicology Program, US
    Department of Health and Human Services.

51. BAEDER,  C., WICKRAMARATNE, G.A.S., HUMMLER, H., MERKLE, J.,
    SCHON,  H., & TUCHMANN-DUPLESSIS, H.  (1985)  Identification
    and assessment of the effects of chemicals  on  reproduction
    and   development  (reproductive  toxicology).   Food  Chem.
    Toxicol., 23: 377-388.

52. KOCHENOUR,  N.K. (1984) Adverse pregnancy outcome: sensitive
    periods,  types of adverse outcomes,  and relationships with
    critical   exposure  periods.  In:   The  new  frontier   in
    occupational  and  environmental health  research, New York,
    Alan R. Liss, pp. 229-235.

53. WHO   (1984)  EHC 30: Principles for evaluating health risks
    to  progeny  associated  with exposure  to  chemicals during
    pregnancy, Geneva, World Health Organization, 123 pp.

54. VOUK,  V.B.  &  SHEEHAN,  P.J.,  ed.   (1983)   Methods  for
    assessing   the   effects   of  chemicals   on  reproductive
    functions, New York, John Wiley and Sons.

55. WILSON,  J.G.   (1973)   Environment and  birth defects, New
    York, London, Academic Press.

56. JOHNSON,  E.M.  &  CHRISTIAN,  M.S.   (1984)   When   is   a
    teratology study not an evaluation of teratogenicity? J. Am.
    Coll. Toxicol., 3: 431-434.

57. WHO   (1985)  EHC 47:  Summary report on  the evaluation  of
    short-term  tests for carcinogens (Collaborative study on  in
    vitro  tests), Geneva, World Health Organization, 74 pp (In:
    Ashby,  J., de Serres, F.J.,  Draper, M., Ishidate, M.,  Jr,
    Margolin, B.H., Matter, B.E., & Shelby, M.D.,  ed.  Progress
    in mutation research, Amsterdam, Elsevier, Vol. 5).

58. NEUBERT,  D.   (1982)   The  use  of  culture  techniques in
    studies  on  prenatal  toxicity. Pharmacol.  Ther., 18: 397-
    434.

59. ROSLAND,  I.R.  &  WALKER, R.   (1983)  The gastrointestinal
    tract  in food toxicology.  In: Conning, D.M.,  &  Lansdown,
    A.B.G., ed. Toxic hazards in food, London, Croom  Helm,  pp.
    183-274.

60. FAO/WHO   (1972)  Evaluation of  certain food additives  and
    the  contaminants  mercury,  lead,  and  cadmium.  Sixteenth
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 51; WHO
    Technical Report Series No. 505 and corrigendum).

61. WHO   (1986)  EHC 59: Principles for evaluating health risks
    from  chemicals during infancy and early childhood: the need
    for  a special approach, Geneva,  World Health Organization,
    73 pp.

62. TOMATIS,   L.,   MOHR,  U.,   &   DAVIS,  W.,   ed.   (1973)
    Transplacental  carcinogenesis, Lyons, International  Agency
    for Research on Cancer (IARC Scientific Publication  No.  4)
    (Proceedings   of  a  Meeting   held  at  the   Medizinische
    Hochschule,   Hanover,  Federal  Republic  of  Germany,  6-7
    October, 1971).

63. ALTHOFF,  J.  & GRANDJEAN,  C.   (1979)   In vivo  studies in
    Syrian  golden  hamsters:  a transplacental  bioassay of ten
    nitrosamines. Natl Cancer Inst. Monogr. Ser., 51: 251-255.

64. HARADA, Y.  (1968)  In: Kutsana, M., ed.  Minamata  disease,
    Kumamoto, Japan, Kumamoto University Press, pp. 93-117.

65. BOWMAN,  R.E.,  HEIRONIMUS,  M.P., &  BARSOTTI, D.A.  (1981)
    Locomotor  hyperactivity  in  PCB-exposed  rhesus   monkeys.
    Neurotoxicity, 2: 251-268.

66. ROGAN,   W.J.,  BAGNIEWSKA,  A.,   &  DAMSTRA,  T.    (1980)
    Pollutants  in breast  milk. N.  Engl. J.  Med., 302:  1450-
    1453.

67. LAMARTINIERE,  C.A., LUTHER, M.A.,  LUCIER, G.W., &  ILLSEY,
    N.P.   (1982)   Altered  imprinting of  rat  liver monoamine
    oxidase by  o,p'- DDT and methoxychlor. Biochem. Pharmacol.,
    31: 647-651.

68. KATO,  R. & TAKANAKA, A.  (1968)  Effect of phenobarbital on
    electron  transport system, oxidation and reduction of drugs
    in  liver microsomes of rats of different ages. J. Biochem.,
    63: 406-408.

69. KATO,  R. & TAKANAKA, A.  (1968)  Metabolism of drugs in old
    rats.  I. Activities of NADPH-linked  electron transport and
    drug-metabolizing  enzyme systems in liver microsomes in old
    rats. Jpn. J. Pharmacol., 18: 381-388.

70. KATO,  R. & TAKANAKA, A.  (1968)  Metabolism of drugs in old
    rats.  II. Metabolism  in vivo  and  effects of drugs in  old
    rats. Jpn. J. Pharmacol., 18: 389-396.

71. SCHMUCKER, D.L. & WANG, R.K.  (1980)  Age-related changes in
    liver  drug-metabolizing  enzymes. Exp.  Gerontol., 15: 321-
    329.

72. TANNENBAUM,  A.  &  SILVERSTONE, H.   (1953)   Nutrition  in
    relation to cancer. Advances Cancer Res., 1: 451-501.

73. ROSS,  M.H., BRAS, G., & RAGBEER, M.S.  (1970)  Influence of
    protein and caloric intake upon spontaneous tumour incidence
    of the anterior pituitary gland of the rat. J.  Nutr.,  100:
    177-189.

74. FAO/WHO   (1981)   Toxicological evaluation  of certain food
    additives (WHO Food Additives Series No. 16).

75. GRICE,  H.C., ed.  (1984)  The selection of doses in chronic
    toxicity/carcinogenicity    studies    and    age-associated
    (geriatric)  pathology:  its  impact on  long-term  toxicity
    studies,   New  York,  Springer-Verlag  (Current  Issues  in
    Toxicology.  Sponsored  by  the International  Life Sciences
    Institute).

76. WHO   (1978)  EHC 6:  Principles and methods  for evaluating
    the  toxicity  of chemicals.  Part  I, Geneva,  World Health
    Organization, 264 pp.

77. PAGET,  G.E.   (1970)   The  design  and  interpretation  of
    toxicity tests. Methods Toxicol., 1970: 1-10.

78. MORGAN, R.W. & WONG, O.  (1985)  A review of epidemiological
    studies  on artificial sweeteners  and bladder cancer.  Food
    Chem. Toxicol., 23: 529-533.

79. ATKINS,  F.M. & METCALFE,  D.D.  (1984)  The  diagnosis  and
    treatment of food allergy. Ann. Rev. Nutr., 20: 233-255.

80. BOCK,  S.A. & MAY, C.D.   (1983)  Adverse reactions to  food
    caused  by sensitivity. In: Middleton, E., Jr, Reed, C.E., &
    Ellis, E.F., ed. Allergy principles and practice,  2nd  ed.,
    St. Louis, Missouri, C.V. Mosby Company, p. 1415.

81. SIRAGANIAN, R.P. & HOOK, W.A.  (1980)  Histamine release and
    assay methods for the study of human allergy. In: Rose, N.R.
    &  Friedman, H., ed. Manual of clinical immunology, 2nd ed.,
    Washington DC, American Society of Microbiology, p. 808.

82. SANTRACH, P.J., PARKER, J.L., JONES, R.T., & YUNGINGER, J.W.
    (1981)    Diagnostic  and  therapeutic  applications   of  a
    modified  RAST and comparison with the conventional RAST. J.
    Allergy clin. Immunol., 67: 97.

83. BOCK,  S.A., LEE, W.Y.,  REMIGO, L.K., &  MAY, C.D.   (1978)
    Studies  of hypersensitivity reactions  to foods in  infants
    and children. J. Allergy clin. Immunol., 62: 327.

84. SAMPSON, H.A. & ALBERGO, R.  (1984)  Comparison  of  results
    of  skin  tests, RAST,  and double-blind, placebo-controlled
    food challenges in children with atopic dermatitis. J. clin.
    Immunol., 74: 26.

85. Protein  advisory group report on the FAO/WHO/UNICEF protein
    advisory  group 16th meeting,  Geneva, 8-11 September,  1969
    (Unpublished PAG Meeting Report Document 3.14/5).

86. Protein  advisory group report on the FAO/WHO/UNICEF protein
    advisory  group  17th meeting,  New  York, 25-28  May,  1970
    (Unpublished PAG Meeting Report Document 3.14/8).

87. Memorandum  on  the  testing of  novel  foods, incorporating
    guidelines   for  testing  by  the   advisory  committee  on
    irradiated  and novel foods, Department of Health and Social
    Scottish  Home  and  Health Department,  Welsh  Office,  and
    Department  of Health and Social Services, Northern Ireland,
    December, 1984.


ANNEX I.  GLOSSARY

I.1  Abbreviations Used in this Document

ADI:      Acceptable Daily Intake (see definition)

CCFA:     Codex  Committee on Food  Additives (see definition  of
          Codex Alimentarius Commission)

COT:      Committee on Toxicity (United Kingdom)

CSM:      Committee on Safety of Medicines (United Kingdom)

EEC:      European Economic Community

EPA:      Environmental Protection Agency (USA)

FAO:      Food and Agriculture Organization of the United Nations

FDA:      Food and Drug Administration (USA)

FSC:      Food Safety Council, Washington, DC, USA

GEMS:     Global Environmental Monitoring System

GLP:      Good Laboratory Practice

IARC:     International Agency for Research on Cancer

IPCS      International Programme on Chemical Safety

JECFA:    Joint  FAO/WHO Expert Committee on  Food Additives (see
            definition)

LD50      Lethal Dose, median

MAFF:     Ministry of Agriculture, Forestry and Fisheries (Japan)

MHW:      Ministry of Health and Welfare (Japan)

MTD:      Maximum Tolerated Dose (see definition)

OECD:     Organisation for Economic Cooperation and Development

PMTDI:    Provisional Maximum Tolerable Daily Intake (see definition)

PSPS:     Pesticides Safety Precautions Scheme (United Kingdom)

PTWI:     Provisional Tolerable Weekly Intake (see definition)

SCOPE:    Scientific  Committee on Problems of the Environment of
            the International Council of Scientific Unions

UNICEF:   United Nations Childrens' Fund

WHO:      World Health Organization


I.2  Definitions of Terms Used in this Document

Acceptable  daily intake:  An estimate by JECFA of the amount of
a food additive, expressed on a body weight basis, that  can  be
ingested daily over a lifetime without appreciable  health  risk
(standard man = 60 kg).

Acceptable daily intake not allocated:  See no ADI allocated.

Acceptable daily intake not specified:  A term applicable  to  a
food substance of very low toxicity which, on the basis  of  the
available   data  (chemical,  biochemical,   toxicological,  and
other), the total dietary intake of the substance  arising  from
its  use at the levels  necessary to achieve the  desired effect
and  from  its acceptable  background in food  does not, in  the
opinion  of  JECFA,  represent a  hazard  to  health.  For  that
reason,  and for reasons  stated in individual  evaluations, the
establishment  of  an  acceptable  daily  intake  expressed   in
numerical  form is not  deemed necessary.  An  additive  meeting
this   criterion   must  be  used  within  the  bounds  of  good
manufacturing  practice,  i.e.,  it  should  be  technologically
efficacious  and should be used at the lowest level necessary to
achieve this effect, it should not conceal inferior food quality
or   adulteration,  and  it  should  not  create  a  nutritional
imbalance.

Codex  Alimentarius  Commission:   The Commission  was formed in
1962  to implement the  Joint FAO/WHO Food  Standards Programme.
The Commission is an intergovernmental body made up of more than
120  Member Nations, the delegates  of whom represent their  own
countries.   The Commission's work of harmonizing food standards
is carried out through various committees, one of which  is  the
Codex Committee on Food Additives (CCFA).  JECFA serves  as  the
advisory  body  to  the  Codex  Alimentarius  Commission  on all
scientific matters concerning food additives.

Conceptus:    All  products  of  conception   derived  from  and
including  the  fertilized ovum  at  any time  during pregnancy,
including the embryo or fetus and embryonic membranes.

Developmental  toxicity:  Any adverse  effects induced prior  to
attainment   of  adult  life,   including  effects  induced   or
manifested in the embryonic or fetal period and those induced or
manifested postnatally (before sexual maturity).

Effect:  A biological change in an organism, organ, or tissue.

Elimination  (in metabolism):  The  expelling of a  substance or
other  material  from  the body  (or  a  defined part  thereof),
usually  by a process of  extrusion or exclusion, but  sometimes
through metabolic transformation.

Embryo/fetotoxicity:   Any toxic effect on  the conceptus resul-
ting  from prenatal exposure, including structural or functional
abnormalities or postnatal manifestation of such effects.

Embryonic  period:  The  period from  conception to  the end  of
major  organogenesis.  Generally, the organ  systems are identi-
fiable at the end of this period.

Enterohepatic  circulation:  Intestinal reabsorption of material
that  has been excreted  through the bile  followed by  transfer
back  to the liver,  making it available  for biliary  excretion
again.

Fetal period:  The period from the end of embryogenesis  to  the
completion of pregnancy.

Food  allergy:   A form  of food intolerance  in which there  is
evidence  of  an abnormal  immunological  reaction to  the food.
"Immediate  allergic  reactions"  are those  which  occur within
minutes  to hours after ingestion  of the offending food,  while
reactions  beginning several hours  to days after  food exposure
are characterized as "delayed allergic reactions".

Food intolerance:  A reproducible, unpleasant reaction to a food
or  food  ingredient,  including reactions  due to immunological
effects,  biochemical factors such  as enzyme deficiencies,  and
anaphylactoid reactions that often include histamine release.

Group  acceptable  daily  intake:  An  acceptable  daily  intake
established  for a group of compounds that display similar toxic
effects, thus limiting their cumulative intake.

Irreducible  level (of a food  contaminant):  That concentration
of  a substance which cannot  be eliminated from a  food without
involving  the  discarding  of that  food  altogether,  severely
compromising the ultimate availability of major food supplies.

JECFA:   JECFA is a technical committee of specialists acting in
their  individual  capacities.   Each  JECFA  is  a  separately-
constituted  committee, and when either the term "JECFA" or "the
Committee"  is used, it is  meant to imply the  common policy or
combined output of the separate meetings over the years.

Long-term toxicity study:  A study in which animals are observed
during  the whole life span (or the major part of the life span)
and  in which exposure to the test material takes place over the
whole  observation time or a substantial part thereof.  The term
chronic toxicity study is used sometimes as a synonym for "long-
term toxicity study".

Maximum tolerated dose:  A term in common use in carcinogenicity
testing meaning a dose that does not shorten life expectancy nor
produce  signs  of  toxicity other  than  those  due  to  cancer
(operationally, the MTD has been set as the maximum  dose  level
at  which a substance induces  a decrement in weight  gain of no
greater than 10% in a subchronic toxicity test).

No ADI allocated:  Terminology used by JECFA in situations where
an  ADI is not established  for a substance under  consideration
because (a) insufficient safety information is available; (b) no
information  is available on its food use; or (c) specifications
for identity and purity have not been developed.

No-observed-effect  level:  The greatest concentration or amount
of  an agent, found by study or observation, that causes no det-
ectable,  usually adverse, alteration of  morphology, functional
capacity, growth, development, or lifespan of the target.

Novel  food:   A  food or  food  ingredient  produced  from  raw
materials not normally used for human consumption or  food  that
is  severely modified by the  introduction of new processes  not
previously used in the production of food.

Processing  aid:  A substance  added to food  during processing,
but subsequently removed.  Traces of a processing aid may remain
with the food.

Provisional  maximum tolerable daily intake:  The end-point used
by  JECFA for contaminants  with no cumulative  properties.  Its
value represents permissible human exposure as a result  of  the
natural  occurrence of  the substance  in food  and in  drinking
water.   In the case of  trace elements that are  both essential
nutrients  and  unavoidable constituents  of  food, a  range  is
expressed,   the   lower   value  representing   the   level  of
essentiality and the upper value the PMTDI.

Provisional  tolerable  weekly  intake:  The  end-point  used by
JECFA for food contaminants such as heavy metals with cumulative
properties.   Its  value  represents  permissible  human  weekly
exposure  to those contaminants unavoidably  associated with the
consumption of otherwise wholesome and nutritious foods.

Reproductive  effects:  To test for  the effects of exposure  to
low  levels of chemicals exceeding  the life span of  one gener-
ation,  tests have been developed  covering several reproductive
cycles.  In the three-generation test, the animals  are  exposed
through three complete reproductive cycles (starting with the F0
generation  at weaning).  These tests, which include exposure  in
 utero and  through the milk,  have been used  in particular  for
assessing toxic effects related to reproduction.

Safety  factor:  A factor applied  by JECFA to the  no-observed-
effect  level to  derive an  acceptable daily  intake  (the  no-
observed-effect  level  is  divided  by  the  safety  factor  to
calculate the ADI).  The value of the safety factor  depends  on
the  nature of the toxic effect, the size and type of population
to   be  protected,  and   the  quality  of   the  toxicological
information available.

Short-term  toxicity study:  An animal study (sometimes called a
subacute or subchronic study) in which the effects  produced  by
the  test  material, when  administered  in repeated  doses  (or
continuously in food or drinking-water) over a period  of  about
90 days, are studied.

Temporary  acceptable daily intake:  Used by JECFA when data are
sufficient  to conclude that use  of the substance is  safe over
the  relatively short period  of time required  to generate  and
evaluate further safety data, but are insufficient  to  conclude
that  use of the substance  is safe over a  lifetime.  A higher-
than-normal  safety factor is used when establishing a temporary
ADI  and  an  expiration  date  is  established  by  which  time
appropriate data to resolve the safety issue should be submitted
to JECFA.

Teratogen:    An  agent  which,  when  administered  prenatally,
induces permanent abnormalities in structure.

Teratogenicity:   The property (or potential)  to produce struc-
tural malformations or defects in an embryo or fetus.

Transplacental  carcinogenesis:  The appearance of  neoplasia in
the  progeny  of  females  exposed  to  chemical  agents  during
pregnancy.



ANNEX II.  STATISTICAL ASPECTS OF TOXICITY STUDIES

II.1  Summary

    Statistical  design and analysis  should aim at  eliminating
sources  of potential bias  and minimizing the  role of  chance.
The application of these principles in the  experimental  design
and  conduct  of toxicological  studies  is discussed  under  10
headings:  choice of species,  dose levels, number  of  animals,
duration  of the study, accuracy  of determinations, stratifica-
tion,  randomization, adequacy of control  groups, animal place-
ment, and data recording.  A number of general considerations to
be borne in mind when conducting statistical analyses  are  also
discussed:  experimental and observational units,  types of res-
ponse  variable, types of between-group comparisons, stratifica-
tion,  age adjustment, multiple  observations per animal,  hypo-
thesis testing and probability values, and multiple comparisons.
Finally,  some recommended methods  of statistical analysis  are
summarized.

II.2  Introduction

    These  guidelines  are  intended primarily  to  provide  the
experimental  scientist without statistical  qualifications with
some  insight into statistical aspects of toxicological studies.
Considerations relating to the design and conduct of studies and
to  the analysis and  interpretation of results  are  discussed,
emphasizing the principles involved rather than the mathematical
details.   While  the  experimentalist  should  have  sufficient
information  to deal with many standard situations, the need for
the  advice of an expert statistician, when dealing with toxico-
logical  data,  cannot  be overemphasized.   Scientific journals
frequently contain papers describing studies in which  the  con-
clusions of the author(s) cannot be supported because  of  defi-
ciencies  in  statistical  methodology, which  could  have  been
avoided had the advice of a qualified statistician  been  avail-
able to the researcher.

II.3  Sources of Difference Between Treated and Control Groups

    An  objective  of  many  toxicity  studies  is  to determine
whether a treatment elicits a response.  However,  the  observa-
tion of a difference in response between a treated and a control
group  does not necessarily mean that the difference is a result
of  the treatment.  There are two other potential causes of dif-
ference, bias and chance.

    Bias  implies  systematic  differences other  than treatment
between the groups, in other words, failure to compare like with
like.   Properly  conducted  studies analysed  appropriately can
eliminate bias.

    Chance  factors cannot be  wholly excluded, because  identi-
cally-treated  animals will not all respond identically, however
carefully  the study is conducted.  While it is impossible to be

absolutely  certain that even  very extreme differences  in res-
ponse  are not due  to chance, appropriate  statistical analysis
will  allow the experimentalist to  assess the probability of  a
"false  positive",  that is,  the  probability of  the  observed
difference having occurred had there been no effect of treatment
at all.  The smaller the probability, the greater the confidence
of  having found a  real effect.  To  improve the likelihood  of
detecting  a true effect with confidence, it is necessary to try
to  minimize the role  of chance by  seeking to ensure  that the
"signal" can be recognized above the "noise".

II.4  Experimental Design and Conduct

    Ten  aspects of the experimental design and conduct of toxi-
city  studies are considered below, the first six being involved
primarily with minimizing the role of chance and the  last  four
being  particularly relevant to the avoidance of bias.  For con-
venience,  the principles are  illustrated with reference  to  a
long-term carcinogenicity study.

II.4.1  Choice of species

    While  maximizing the "signal"  means avoiding a  species in
which the response of interest is very rare, the use of an over-
responsive species also has problems.  Thus, to achieve the same
level of statistical significance in comparing a  treated  group
with  a  5%  response and  a  control  group with  a 0% response
requires  only one tenth as  many animals as when  the responses
are  55 and 50%, respectively.   Furthermore, it is not  certain
that an increased incidence of a lesion that is a  common  spon-
taneous  finding in the animal  species used (such as  pituitary
tumours  in  Wistar rats)  provides  biological evidence  of  an
effect  that can be extrapolated  to other species.  Other  con-
siderations  related to the choice  of species, whether they  be
practical  (short life span, small size, availability, existence
of  detailed knowledge of the species) or more theoretical (bio-
chemical,  physiological, or anatomical  similarity to man),  do
not really pose statistical problems.

II.4.2  Dose levels

    Dose  selection is an important and controversial element in
the  development of a protocol for a toxicity bioassay.  On bio-
logical grounds, it would be ideal to test only at  dose  levels
comparable  with those to  which human beings  are exposed.   On
statistical and economic grounds, this is not  usually  practic-
able because the effect will be too small to detect without very
large numbers of animals.  To avoid the possibility  of  missing
an  effect that would occur in a small proportion of millions of
exposed human beings in a study on hundreds or even thousands of
animals,  it is  normally appropriate  to test  animals at  dose
levels  many times higher than the maximum human exposure level.
Then,  assuming any effect that exists is dose-related, the dem-
onstration of a non-significant increase in response at  a  high
dose level, though not providing evidence of absolute safety (an

impossible goal), can give reasonable grounds for believing that
any  effects that might  occur at a  very much lower  dose level
would be, at most, very slight.

    A  particular problem with this  procedure is to decide  how
high  the dose level  should be.  In  long-term  carcinogenicity
studies,  the dose should  clearly be one  that is not  so great
that  the animals  die from  toxic effects  before they  have  a
chance  to get cancer.   On the basis  of these principles,  the
International  Agency for Research on Cancer (1) has recommended
that  the high dose should  be one expected on  the basis of  an
adequate short-term study to produce some toxicity  when  admin-
istered  for the duration of  the study, but should  not induce:
(a)  overt toxicity, i.e., appreciable  death of cells or  organ
disfunction,  as determined by  appropriate methods; (b)   toxic
manifestations  that are predicted materially to reduce the life
span  of the animals, except  as a result of  the development of
neoplasms; or (c) 10% or greater retardation of body-weight gain
compared with that in control animals.

    If the substance seems completely non-toxic, the  high  dose
may  represent about 5% of the diet, or even more for substances
such as some nutritive food ingredients.

    It  is important  to have  more than  one dose  level for  a
number  of reasons.  One  is to compensate  for the  possibility
that  a misjudgment has occurred  and that the highest  dose may
prove  to be toxic.  A second is that the metabolic pathways may
differ at the various dose levels.  A third reason is  that  the
whole  point of the study may be to obtain dose-response inform-
ation.   Finally, it may be  necessary to ensure that  an effect
does not occur at dose levels in the range to be used by man.

II.4.3  Number of animals

    The  number of animals  to be used  is clearly an  important
determinant of the precision of the findings.   The  calculation
of the appropriate number depends on:

    (a)  The critical difference, i.e., the size of  the  effect
         to be detected;

    (b)  the  false positive rate,  i.e., the probability  of an
         effect  being detected when  none exists (known  as the
         "type I error" or the " alpha level"); and

    (c)  the  false negative rate,  i.e., the probability  of no
         effect  being detected when one of exactly the critical
         size exists (known as the "type II error" or  the  " 
         level").

A  reduction in any  of these factors  means an increase  in the
number of animals required.

    The  method of calculation of the number depends on the exp-
erimental design and the type of statistical analysis envisaged.
Tables  are available for a  number of standard situations.   To
give an idea how the numbers depend on the  critical  difference
and on the alpha and  levels, Tables 1 and 2 give examples of two
common situations, both of which are related to a study in which
there is a control and a treated group.  The first is related to
a continuous variable assumed to be normally  distributed,  with
the  critical difference  expressed in  terms of  the number  of
standard  deviations () by which the treated group differs from
the  control group.  Thus, given  a control response known  from
past experience to have a mean value of 50 units with a standard
deviation  of 20 units, two  groups of 36 animals  each would be
needed to have a 90% chance ( = 0.10)  of detecting  a  differ-
ence  in  response of 10  units (delta = 10/20 = 0.5) at the 95% 
confidence level (alpha = 0.05).

    In the second situation, two proportions are compared.  Here
the numbers of animals depend not only on the ratio  of  propor-
tions, but also on the assumed proportion in the controls. Thus,
when  the control response is expected to be 10%, the numbers of
animals required in each group to detect an  increased  response
by a factor of 1.5 (r) is 920, assuming again an alpha level  of
0.05  and a  level of 0.1,  whereas if the control  response is
expected to be 50%, the numbers required would be 79 per group.

    For  more complex situations,  the advice of  a professional
statistician  should be sought, though a general rule is that to
increase  precision  (i.e., decrease  the  size of  the critical
difference) by a factor n, the number of animals  required  will
have to be increased by a factor of approximately n squared.
Table 1.  Number of animals required in each of a control and treated
group in order to have a probability (1 - ) of picking up a difference
of delta standard deviations as significant at the 100 (1 - alpha) percent
confidence level for a normally distributed variable
-----------------------------------------------------------------------------
Single-side test a    alpha = 0.005       alpha = 0.025       alpha = 0.05
Double-sided test a   alpha = 0.01        alpha = 0.05        alpha = 0.1
 =                0.01   0.1   0.5     0.01   0.1  0.5     0.01   0.1   0.5
-----------------------------------------------------------------------------
delta =   0.5      100    63    30       76    44   18       65    36    13
          0.75      47    30    16       35    21    9       30    17     7
          1.0       28    19    10       21    13    6       18    11     5
          1.5      15     11     7       11     7             9     6
          2.0      10      8     5        7     5             6
-----------------------------------------------------------------------------
a   See  section  II.5.7 for  definitions  of single-sided  and double-sided
    tests.

Table 2.  Number of animals required in each of a control and treated
group in order to have a probability (1 - ) of picking up a
proportional increase by a factor r as significant at the 100 (1 - alpha)
percent confidence level for a binomially distributed variable
-----------------------------------------------------------------------------
Single-side test a   alpha = 0.005       alpha = 0.025       alpha = 0.05
Double-sided test a  alpha = 0.01        alpha = 0.05        alpha = 0.1
 =                0.01   0.1  0.5      0.01   0.1   0.5    0.01   0.1   0.5
-----------------------------------------------------------------------------
Control level = 10%

r =       1.25     7679  4754  2120     5871  3358  1228    5039  2737   865
          1.5      2103  1302   581     1608   920   337    1380   750   237
          2.0       613   380   170      469   268    98     403   219    69

Control level = 20%

r =       1.25     3353  2076   926     2563  1466   536    2200  1195   378
          1.5       902   558   249      689   395   145     592   322   102
          2.0       253   157    70      193   111    41     166    90    29

Control level = 50%

r =       1.25      757   469   209      579   331   122     497   270    86
          1.5       181   112    50      138    79    29     119    65    21
          2.0        37    23    10       28    16     6      24    13     5
-----------------------------------------------------------------------------
a   See  section  II.5.7 for  definitions  of single-sided  and double-sided
    tests.
    When  a number of  treatments are to  be tested in  a study,
each to be compared with a single untreated control group, it is
advisable  that more animals  be included in  the control  group
than in each of the treated groups, because the precision of the
control  group results is  relatively more important.   When all
groups  are of equal interest, it is appropriate to have approx-
imately the square root of k times as many animals in  the  con-
trol group as in each of the k treated groups.

    One  point  frequently  misunderstood  by  the  experimental
scientist  is  related  to the  number  of  animals required  in
studies  in which more than  one treatment is investigated  in a
crossed  design.  If, for example,  compounds A and B  are being
compared,  and each is dissolved in two different solvents, in a
2  x 2 design with 4 groups, calculations of sample size to gain
an overall verdict on the difference between the  two  compounds
should  generally be  based on  the overall  numbers of  animals
treated  with each compound  for both solvents  combined, unless
there  is reason to  expect compound-solvent interaction,  i.e.,
that  the  compound A/compound  B  difference depends  on  which
solvent  is  used.  Conversely,  if it has  been decided that  2
groups  of 100 animals each are sufficient for attaining a given
level  of precision concerning the  differences in effects of  a
treatment,   additional  information  and  another   factor  (or
factors)  of  interest can  be  obtained without  requiring  any
additional animals.

II.4.4  Duration of the study

    The  duration of  the study  can also  markedly  affect  the
sensitivity  of  tests.  This  is  particularly so  in long-term
carcinogenicity studies in which the great majority  of  cancers
are  seen  in the  latter half of  an animal's lifetime.   Thus,
while  studies should not  be terminated too  early, it is  also
important  that they do not go on too long.  This is because the
last few weeks or months may produce relatively little data at a
disproportionate cost, and diseases of extreme old age may be of
little interest in themselves but may render it  more  difficult
to  detect tumours and  other conditions that  are of  interest.
Where  the  study is  of the prevalence  of an age-related  non-
lethal condition observable only at death that ultimately occurs
in  all  or  nearly all  of  the  animals, early  termination is
required.   In  this  situation  the  greatest  sensitivity   is
obtained when the average prevalence is about 50%.

II.4.5  Accuracy of determinations

    Accuracy  of observations is clearly important in minimizing
error.   The  advent of  good  laboratory practice  and  quality
control  units has done much to improve the quality of recording
observations,  but the  quality of  the study  still depends  on
interested and diligent personnel.

II.4.6  Stratification

    To detect a treatment difference with accuracy,  the  groups
being compared should be as homogeneous as possible with respect
to  other known causes of  the response of interest.   Consider,
for  example,  a set  of animals thought  to be homogenous  (but
which, in fact, consist of two genetically different substrains)
in which the following measurements of body weight were obtained
in groups of 10 treated and 10 control animals,  the  underlined
readings relating to the first of the two substrains:

    Control:  181 192 217 290 321 292 307 347 276 256

    Treated:  222 249 232 284 270 215 265 378 328 391.

If  the  substrain  is ignored,  the  variability  of  the  data
increases  so  that  more controls  are  required  to  detect  a
treatment effect and, if unequal numbers of each  substrain  are
present  in each group, may bias the comparison.  In the example
given the means are as follows:

                  Substrain 1      Substrain 2         Total

     Control         196.7            298.4            267.9
     Treated         248.1            365.7            283.4
     Difference       51.5             67.2             15.5

Although  in each substrain the treatment results in an increase
in  body weight  of over  50 units,  the greater  number of  the
lower-weight  strain  in  the  treated  groups  means  that  the
difference observed is much less.

    There  are two ways to take account of the substrain differ-
ence and to achieve a more precise answer.  One is to  use  sub-
strain  as a "stratifying variable" at the analysis stage.  This
involves  carrying out separate  analyses at each  level of  the
variable  considered and combining  the results for  an  overall
conclusion  about the treatment  effect.  However, it  does  not
preclude  the possibility that the proportions of each substrain
in  each group are so  different that the data  provide substan-
tially  less  comparative  information than  might  otherwise be
achieved.  In the extreme case, if all control animals  were  of
substrain  1 and all  treated animals were  of substrain 2,  the
study would be worthless to determine whether  differences  were
due  to treatment or  substrain.  To obviate  this  possibility,
substrain  can be used as a "blocking factor" in the design.  In
this  case, animals in each  substrain are allocated equally  to
control and treated groups.  Although this removes bias,  it  is
still  necessary to treat  strain as a  stratifying variable  in
statistical analysis to increase precision.

    When  more than one known  factor affects the response,  all
can  be taken into account  simultaneously.  Both at the  design
stage,  or  retrospectively in  the  analysis, the  results  are
treated  at each combination of levels of the factors.  Thus, to
block  for substrain, sex, and  room where 3 experimental  rooms
were  needed to house the animals, 12 mini-studies, one for each
of  the  2 (substrains)   x 2 (sexes)  x 3 (rooms)  combinations
would be set up.

II.4.7  Randomization

    Random allocation, or randomization, of animals in treatment
groups  is an  essential of  good experimental  design.  If  not
carried  out, it  is not  strictly possible  to tell  whether  a
difference  between groups  is a  result of  differences in  the
treatment  applied or is due  to some other relevant  factor.  A
fundamental  on which statistical  methodology is based  is that
the  probability of a particular response occurring is equal for
each  animal,  regardless of  group.   The ability  to randomize
easily  is  a  major advantage  that  animal  studies have  over
epidemiological studies.

    The  process  of  randomization  eliminates  bias,  so  that
statistical  analysis  is  concerned  only  with  assessing  the
probability  of an observed difference happening by chance.  The
smaller the probability, the more it suggests a  true  treatment
effect.   The procedure used for  randomization should genuinely
ensure  that all possible  assignments of animals  to  treatment
groups  are equally probable.  Such equal probabilities are best
achieved  with  pseudorandom  numbers,  as  found  in  tables or
produced   by  computer,  it  being  difficult  to  ensure  that
apparently random devices such as dice or playing  cards  really
are  random.  Randomization should never be based on a system of
testing animals haphazardly, as they come, and assigning them to
successive treatment groups.  Not only do human beings  find  it
virtually  impossible to generate random  sequences unaided, but

it  is  well known  that the first  animals selected may  differ
markedly  from  the last,  who are more  active and avoid  being
caught.

    In many experimental situations it is adequate  to  randomly
allocate  all the animals to treatment groups, but, in some, the
technique of stratified random sampling is preferred.   In  this
technique,   the  animals  are  first   divided  into  subgroups
("strata"),  according  to  factors  known  or  believed  to  be
strongly  related  to the  response,  with random  allocation to
treatment  groups then being  carried out within  each  stratum.
Sex is normally treated as a stratifying variable.  In  a  large
study  in which animals  are delivered in  batches, batch  could
also be treated in this way, each batch forming a smaller study,
the results from which can be combined in the analysis.

    The above discussion on randomization and stratification has
been  concerned  primarily with  the  allocation of  animals  to
treatment  groups.  The same  principles apply to  anything that
can affect the recorded response.  Thus, in a  two-group  study,
measurements  of some biochemical  parameter should not  be made
for the first group in the morning and for the second  group  in
the  afternoon.  While the major part of such potential bias can
be averted fairly easily by various simple procedures,  such  as
doing  alternate  measurements  on treated  and control animals,
randomization is preferable.  Although many different procedures
throughout  a  study  (feeding, weighing,  observation, clinical
chemistry,  and pathological examinations) require consideration
in  this way, the same  random number can usually  be applied to
all the procedures.  Thus, if the cage position of  the  animals
is  randomly allocated  and does  not depend  on treatment,  the
animals can always be handled in the same cage sequence.

II.4.8  Adequacy of control groups

    The  principle of comparing like with like implies that con-
trol groups should be randomly allocated from the  same  control
source as the treatment groups.  While historical  control  data
can  be of  value in  the interpretation  of rarer  findings  in
treated animals, there is so much evidence of quite large syste-
matic  differences in response between  apparently identical un-
treated  control groups  tested at  different times  that it  is
often  impossible to be sure whether a difference seen between a
treated group and a historic group is really due to treatment at
all.

    It  is also  essential to  be sure  that the  treated  group
differs  from  the  control  group  only  with  respect  to  the
treatment  of interest.  Thus,  if a treatment  is applied in  a
solvent,  an  untreated  control  is  not  a  proper  basis  for
comparison, as one cannot be sure whether  observed  differences
are a result of the treatment or the solvent.  In this case, the
appropriate  control group  would be  one in  which animals  are
given only the solvent.

II.4.9  Animal placement

    The  general underlying requirement to avoid systematic dif-
ferences  between groups other than their treatment also demands
that attention be given to the question of animal placement.  If
all  treated animals are placed  on the highest racks  or are at
one end of the room, differences in heating, lighting, or venti-
lation  may produce effects  that are erroneously  attributed to
treatment.   Such systematic differences should be avoided, and,
in  many cases, randomization  of cage positions  is  desirable.
This  may not be  possible in some  circumstances, such as  with
studies  involving volatile agents where cross-contamination can
occur.

II.4.10  Data recording

    The application of the principle of comparing like with like
means the avoidance of systematic bias in data  recording  prac-
tices.  Two distinctly different types of bias can  occur.   The
first  is a systematic shift in the standard of measurement with
time,  coupled with a tendency  for the time of  measurements to
vary  from treatment to treatment.  The second is that awareness
of the treatment may affect the values recorded by the measurer,
consciously  or subconsciously.  The second bias is circumvented
by the animals' treatment not being known to the measurer, i.e.,
the  readings being carried  out "blind".  Although  not  always
practical  (that an animal  is treated may  be obvious from  its
appearance),  laboratories should organize their  data recording
practices  so that, at  least for subjective  measurements,  the
observations are made "blind".

    The  problem of avoidance of bias due to differences in time
of  observation is a  particularly important one  in histopatho-
logical assessment, especially for the recording of lesions of a
graded  severity, and in large  studies in which the  slides may
take the pathologist more than a year to read.  When  more  than
one  pathologist reads the  slides, there should  be  discussion
between them as to standardization of terminology and data to be
recorded, and each should read a random or a stratified  set  of
the slides to avoid bias.

II.5  Statistical Analysis - General Considerations

    In  the simplest situation, animals are randomly assigned to
a  treated, or a control,  group and one observation  is made on
each  animal, the objective of the statistical analysis being to
determine whether the distribution of responses in  the  treated
group  differs from that in  the control group.  Before  summar-
izing  some of the appropriate techniques for analysis, a number
of  more general  points underlying  the choice  of the  correct
method and interpretation of the results will be discussed.

II.5.1  Experimental and observational units

    In  the simple example cited  above, the animal is  both the
"experimental unit" and the "observational unit".  This  is  not

always  so.  In the  case of feeding  studies, the cage,  rather
than  the animal, is usually the experimental unit in that it is
the cage, rather than the animal, that is assigned to the treat-
ment.   In the case  of histopathology, observations  are  often
made from multiple sections per animal in which case the section
rather  than the animal is the observational unit.  For the pur-
pose of determining treatment effects by the  methods  described
below, it is important that each experimental unit provides only
one item of data for analysis, as the methods are all  based  on
the  assumption  that  individual data  items  are statistically
independent.  If multiple observations per experimental unit are
made, these observations should be combined in some suitable way
into  an overall observation  for that experimental  unit before
analysis.  Thus, in a study in which 20 animals were assigned to
two  treatment groups of 10  animals each and in  which measure-
ments of the weight of both kidneys were made  individually,  it
would  be wrong to carry out an analysis in which the 20 weights
in group 1 were compared with the 20 weights in group 2, because
the  individual kidney weights are not independent observations.
A valid method would be to carry out an analysis  comparing  the
10  average kidney weights in group 1 with the 10 average kidney
weights in group 2.

II.5.2  Types of response variable

    Responses  measured in toxicological studies can normally be
classified as being one of three types:

    (a)  Presence/absence:  A response either occurs  or it does
         not.

    (b)  Ranked: A response may be present in  various  degrees.
         Thus,  severity  may  be classed  as  minimal,  slight,
         moderate, severe, or very severe.

    (c)  Continuous:  A response may  take any value,  at  least
         within a given range.

    Each  type of response demands  a different sort of  statis-
tical technique.  While analysis of presence/absence data, often
referred to as "contingency table analysis", can be  applied  to
ranked  or continuous data by defining values above a given cut-
off  point  as "present",  this  is not  generally  recommended,
because it wastes information.

    Continuous   data  are  usually  analysed   by  "parametric"
methods,  which assume that the  statistical distribution under-
lying the response variable (or some transformation of it, e.g.,
its logarithm) has a specific form, traditionally the well-known
bell-shaped Normal or Gaussian distribution.  While such methods
are best when the distribution assumed is correct, they can give
misleading conclusions when the assumption is grossly incorrect.
For  this reason, when there is doubt about the underlying dist-
ribution, it is often preferable to analyse continuous  data  by
methods  appropriate  for  ranked data,  since  these "non-para-
metric"  methods make no  such underlying assumption  and  their
conclusions are generally more valid.



II.5.3  Types of between-group comparisons

    While  in the two-group  study only one  comparison is  pos-
sible,  this is not the case when more than two groups are being
compared.   Two particularly important types of test made in the
k (> 2) group situation are the test for heterogeneity  and  the
test for dose-related trend.

    The  test for heterogeneity  determines whether, taken  as a
whole,  there  is significant  evidence  of departure  from  the
(null) hypothesis that the groups do not differ in their effect.
It  is generally applicable but is not very informative, because
it does not specifically take into account the likely pattern of
response.

    The  test  for  dose-related  trend  is  applicable  only in
studies  in which the groups receive different doses of the same
substance  (or have some other natural ordering).  It determines
whether  there is a tendency for response to rise in relation to
the dose of the test substance.  Graphically, the test for trend
can  be  seen  as determining  whether  a  sloped straight  line
through  the  dose (x-axis)/response  (y-axis) relationship fits
the data significantly better than a horizontal  straight  line.
That  the  trend statistic  is  significantly positive  does not
necessarily imply that the treatment increases response  at  all
dose levels, though it is a particularly good test if  the  true
situation is a linear non-threshold model.  A trend  test  often
detects a significant true effect when individual comparisons of
treated  groups with  the control  group fail  to give  signifi-
cance.

    Sometimes,  there is significant departure from trend, e.g.,
when  there  is significant  heterogeneity  but no  evidence  of
trend.  This may arise because a response increases with dose at
lower  dose levels and then reduces at high dose levels, perhaps
due  to competing risks.   In this situation,  it may be  appro-
priate  to test for trend  using only the data  from the control
and lower-dose groups.

II.5.4  Stratification

    In  the simplest  situation all  the animals  in  the  study
differ  systematically  only  with  respect  to  the   treatment
applied.   However, often there are a number of sets of animals,
each of which differs systematically only in respect  to  treat-
ment,  but  where  the  characteristics  of  sets  differ.   The
commonest  situation  relates to  male  and female  animals, but
there are many other possibilities, such as different conditions
under  which the  response variable  is measured.   While it  is
often useful to look within each set of animals or "stratum", as
it  is often called,  to determine the  effects of treatment  in
each situation, it is also useful to determine whether,  on  the
basis  of the data from  all the strata, an  effect of treatment
can  be seen overall.   In some situations,  a relatively  small
number  of animals in each stratum can make it difficult to pick

up  an  effect of  treatment  as significant  within  individual
strata,  and a clear  result can be  seen only when  results are
combined over the strata.  The essence of stratification lies in
making comparisons within strata and then accumulating treatment
differences  over  strata.  Pooling  data  over strata  and then
making  a single comparison  can lead to  erroneous conclusions.
To illustrate this, consider a hypothetical study in  which,  in
one  batch of  animals, 5/10  controls and  12/30  test  animals
responded,  while in a second batch, 6/30 controls and 1/10 test
animals  responded.  If batch  were ignored, it  would be  noted
that  11/40 controls compared with 13/40 test animals responded,
leading  to the erroneous  conclusion that treatment  tended  to
increase response.  An appropriate analysis would consider batch
as a stratifying factor and note that within both batch  1  (50%
versus  40%) and batch 2 (20% versus 10%)  the response  in  the
control  group was higher than  that in the test  group so that,
combining these two differences, a conclusion would  be  reached
that treatment tended to decrease response.

II.5.5  Age adjustment

    For  many  conditions,  such as  tumour incidence, frequency
increases  markedly with age  (and concomitantly with  length of
exposure to the agent), and the overall frequency in a treatment
group  can depend as much on the proportion of animals surviving
a long time as on the actual ability of the treatment  to  cause
the  condition.   To adjust  for  differential survival,  age is
usually  treated as a stratifying variable so that between-group
comparisons  are made of animals at similar ages, results of the
comparisons being combined over the different age  strata.   Age
adjustment  is  normally applied  to presence/absence conditions
and,  as  discussed at  length by Peto  et al. (2),  the correct
method depends on the context of observation of  the  condition.
There are three different situations:

    (a)  Conditions  visible in-life:  Here comparisons are made
         of  the number of  animals developing the  condition in
         the  time period as a  proportion of those without  the
         condition at the beginning of the period.

    (b)  Conditions visible only at death and assumed  to  cause
         death (fatal):  Here comparisons are made of the number
         of  animals dying from the condition in the time period
         as  a proportion of those alive at the beginning of the
         period.

    (c)  Conditions  visible only at  death and assumed  not  to
         cause death (incidental):  Here comparisons are made of
         the number of animals dying with the condition  in  the
         time period as a proportion of all those dying  in  the
         time period.

II.5.6  Multiple observations per animal

    When  multiple observations are made on one animal there are
a  number of additional types of statistical analysis, depending

on  the  experimental  situation  and  the  objectives.   It  is
impossible to cover all the possibilities in this  summary,  but
the following situations are of reasonable frequency:

    (a)  Association   between  variables:   Here  two  or  more
         different variables are recorded on an animal  and  the
         objective  is  to  determine  whether  the  values  are
         independent or are correlated.

    (b)  Variation  in  association between  variables by treat-
         ment:  For each treatment group an indicator  of  asso-
         ciation between variables is calculated and the statis-
         tical  problem is to test whether this indicator varies
         significantly by treatment group.

    (c)  Multiple  observations of the same  variable:  For body
         weight and clinical chemistry data it is common to take
         measurements  on the same animals  at regular intervals
         throughout  a  long-term  study.   While  between-group
         comparisons  can be carried out on the basis of data at
         any specific point in time, this limits the  amount  of
         information  in  any  one analysis.   Statistical tech-
         niques are also available to compare groups  with  res-
         pect to change in response between two points  in  time
         or, more generally, with respect to the general pattern
         of response over a period of time.

    (d)  Within-animal comparisons:  In most toxicological stud-
         ies, different animals receive the different treatments
         and comparisons are made on a between-animal basis.  In
         some  studies, the same  animal receives more  than one
         treatment.   In such studies,  it is important  to  use
         appropriate  statistical methods based on within-animal
         comparisons.

II.5.7  Hypothesis testing and probability values

    Reports of toxicity studies often include statements such as
"the relationship between treatment and blood glucose levels was
statistically  significant (P = 0.02)".  What does this actually
mean?  Three points must be made.

    First, there is a difference in meaning  between  biological
and  statistical significance.  It is  quite possible to have  a
relationship  that is unlikely  to have happened  by chance  and
therefore   statistically  significant  but  of   no  biological
consequence  at all, the  animals' well-being being  unaffected.
On  the other hand an  observation may be biologically,  but not
statistically, significant, such as when one or two  tumours  of
an  extremely rare type  are seen in  treated animals.   Overall
judgement of the evidence must take into account both biological
and statistical significance.

    Second, "P  = 0.02" does not mean that the  probability  that
there  is no treatment effect is 0.02.  The true meaning is that
given  the treatment actually  had no effect  whatsoever (or  to

phrase  it  more technically,  under  the null  hypothesis)  the
probability of observing a difference as great or  greater  than
that actually seen is 0.02.

    Third,  there  are  two types  of  probability (P)-value.  A
"one-sided"  (or  one-tailed)  P-value  is  the  probability  of
getting,  by chance  alone, a  treatment effect  in a  specified
direction as great or greater than that observed.  A "two-sided"
(two-tailed) P-value is the probability of obtaining  by  chance
alone, a treatment difference in either direction,  positive  or
negative, as great or greater than that observed.  Whenever a P-
value  is quoted, it should  be made clear which  is being used.
Normally,  two-tailed  P-values are  appropriate.  However, when
there  is  prior  reason to  expect  a  treatment effect  in one
direction  only, a one-tailed  P-value is normally  used.  If  a
one-tailed   P-value  is  used,  differences   in  the  opposite
direction to that assumed should be ignored.

    While  a P-value of 0.001  or less can, on  its own, provide
very  convincing  evidence  of  a  true  treatment  effect, less
extreme  P-values such as P = 0.05 should be viewed as providing
indicative  evidence  of  a  possible  treatment  effect,  to be
reinforced or supported by other evidence.  If the difference is
similar  to one found in  a previous study, or  if the response,
based on biochemical considerations, is expected, a less extreme
P-value would suffice than if the response was unexpected or not
found at other dose levels.

    Some  laboratories,  when presenting  results of statistical
analyses,   assign  an  almost  magical  relevance  to  the  95%
confidence  level (P > 0.05), simply marking results significant
or not significant at this level.  This is very  poor  practice,
because  it gives insufficient  information and does  not enable
the  distinction  between  an  undeniable  effect  and  one that
requires other confirmatory evidence.  While it is not necessary
to give P-values exactly, it is essential to give some  idea  of
the  degree of confidence.  A useful method is to use plus signs
to  indicate positive differences  (and minus signs  to indicate
negative  differences), with ++++ meaning P < 0.001, +++ meaning
0.001 < P < 0.01, ++ meaning 0.01 < P < 0.05, and + meaning 0.05
<  P < 0.1.   This makes it  easier to assimilate  findings when
results for many variables are presented.

II.5.8  Multiple comparisons

    Toxicological  studies frequently involve  making treatment/
control  comparisons for large numbers of variables.  If no true
treatment  effect exists, it is possible that, purely by chance,
one or more variables will show differences significant  at  the
95%  confidence  level.   For  example,  with  100   independent
variables, at least one variable would show  significance  99.4%
of  the time.  Because of  this, it has been  suggested that the
critical value required to achieve significance should  be  made
more  stringent with increasing numbers of variables studied, so
that,  in testing at the  95% confidence level, 19  times out of
20, all the variables in the test show  non-significance.   This

approach is not recommended, because frequently in toxicological
studies a compound has only one or two real effects and  has  no
effect  on a  large number  of other  variables  studied.   Such
multiple comparison tests would make it much more  difficult  to
demonstrate  statistical significance for the  real effects.  In
any case, there is something unsatisfactory about a situation in
which  the  relationship between  a  treatment and  a particular
response  arbitrarily depends on which other response happens to
be investigated at the same time.  For this reason, no reference
is made below to any such procedures.

II.6  Statistical Analysis - Some Recommended Methods

    A  number of recommended methods of statistical analysis are
listed below.  For mathematical details the reader  is  referred
to  Peto et al.  (2) and to  Breslow & Day  (3) for analysis  of
presence/absence data, to Siegel (4) and to Conover (5) for non-
parametric  analysis, and to Johnson  & Leone (6) and  Bennett &
Franklin  (7) for analysis of continuous data.  When details are
not found in these volumes, specific references are given.

II.6.1  Presence/absence data

II.6.1.1    Between-animal comparisons

Individual group     Fisher exact test (unstratified data) 2 x 2
comparisons          corrected chi-squared test (stratified or
                     unstratified data)

Heterogeneity        2 x k chi-squared test (stratified or
                     unstratified data)

Dose-related         Armitage test (stratified or unstratified
trend                data).

See reference 3 for details of stratified tests and for tests of
constancy  of association over strata.  See reference 2 for age-
adjusted tests.

II.6.1.2    Within-animal comparisons

Individual group     McNemar test or sign test
comparisons

Heterogeneity        Cochran test

Association          Fisher exact test
between variables    2 x 2 corrected chi-squared test

II.6.2  Ranked data

II.6.2.1    Between-animal comparisons

Individual group     Mann Whitney U test
comparisons

Heterogeneity        Kruskal-Wallis one-way analysis of variance

Dose-related         See reference 8
trend

II.6.2.2    Within-animal comparisons

Individual group     Wilcoxon matched pairs signed-rank test
comparisons

Heterogeneity        Friedman two-way analysis of variance

Dose-related         Page test
trend

Association          Spearman's rank correlation coefficients
between variables

II.6.3  Continuous data

    Methods assume normality and homogeneity of variance between
groups.

    Before using the methods:

Test for outliers    See reference 9

Test for             Bartlett test
homogeneity
of variance

    Consider  transformation of data by logarithms and/or square
roots if untransformed data show heterogeneity of variance.

    If variance is still heterogeneous after transformation, use
methods of ranked data.

II.6.3.1    Between-animal comparisons

Individual group     Students t-test
comparisons

Heterogeneity        One-way analysis of variance

Dose-related         Linear regression analysis
trend

II.6.3.2    Within-animal comparisons

Individual group     Paired t-test
comparisons

Heterogeneity        Two-way analysis of variance

Dose-related         Linear regression analysis
trend

II.6.3.3    Association between variables

Variation in         Pearson correlation coefficient
association          Analysis of convariance
between variables
over groups

Change in            Analysis of variance to assess difference at
variable             second time-point after adjusting for first.
over time

REFERENCES TO ANNEX II

 1. FERON,  V.C., GRICE, H.C., GRIESEMER, R., & PETO, R.  (1980)
    Basic requirements for long-term assays for carcinogenicity.
    In:   Long-term   and   short-term  screening   assays   for
    carcinogens:  a  critical  appraisal,  Lyons,  International
    Agency for Research on Cancer, pp. 21-83 (IARC Monographs on
    the  Evaluation  of the  Carcinogenic  Risk of  Chemicals to
    Humans, Suppl. 2).

 2. PETO,  R., PIKE,  M.C., DAY,  N.E., GRAY,  R.C., LEE,  P.N.,
    PARISH, S., PETO, J., RICHARDS, S., & WAHRENDORF, J.  (1980)
    Guidelines  for  simple,  sensitive significance  tests  for
    carcinogenic  effects in long-term animal  experiments.  In:
    Long-term and short-term screening assays for carcinogens: a
    critical appraisal, Lyons, International Agency for Research
    on Cancer, pp. 311-426 (IARC Monographs on the Evaluation of
    the Carcinogenic Risk of Chemicals to Humans, Suppl. 2).

 3. BRESLOW,  N.E. & DAY,  N.E.  (1980)  Statistical  methods in
    cancer  research. I. The  analysis of case-control  studies,
    Lyons,  International  Agency  for Research  on Cancer (IARC
    Scientific Publication No. 32).

 4. SIEGEL,   S.   (1956)   Non-parametric  statistics  for  the
    behavioural sciences, New York, McGraw-Hill Book Company.

 5. CONOVER,  W.J.  (1980)  Practical non-parametric statistics,
    2nd ed., New York, John Wiley and Sons.

 6. JOHNSON,   N.L.  &  LEONE,  F.C.    (1964)   Statistics  and
    experimental   design   in  engineering   and  the  physical
    sciences, New York, John Wiley and Sons, Vol. 1.

 7. BENNETT, C.A. & FRANKLIN, N.L.  (1954)  Statistical analysis
    in chemistry and the chemical industry, New York, John Wiley
    and Sons.

 8. MARASCUILO,  L.A.  &  MCSWEENEY, M.   (1967)  Non-parametric
     post  hoc comparisons for trend.  Psychol. Bull., 67:  401-
    412.

 9. BARNETT,  V. & LEWIS,  T.  (1978)  Outliers  in  statistical
    data, New York, John Wiley and Sons.

ANNEX III. GUIDELINES FOR THE EVALUATION OF VARIOUS GROUPS OF 
FOOD ADDITIVES AND CONTAMINANTS

    These  guidelines  have been  established  by JECFA  and are
reproduced here for easy reference.  They are valid  within  the
context  in which they were generated, and are intended to serve
as  examples  of  guidance  by  JECFA  for  evaluating  specific
categories of substances.

1.  Enzyme  Preparations Used in Food  Processing (adapted from:
    1, p. 49; 2)

    (a)    Toxicological evaluation

    For   the   purpose  of   toxicological  evaluation,  enzyme
preparations used in food processing can be grouped into 5 major
classes:

    (i)    Enzymes  obtained from edible tissues  of animals com-
           monly used as foods.  These are regarded as foods and,
           consequently,  considered  acceptable,  provided  that
           satisfactory  chemical and microbiological  specifica-
           tions can be established.

    (ii)   Enzymes  obtained  from  edible  portions  of  plants.
           These  are also regarded  as foods and,  consequently,
           considered   acceptable,  provided  that  satisfactory
           chemical  and  microbiological  specifications can  be
           established.

    (iii)  Enzymes  derived  from microorganisms  that are tradi-
           tionally accepted as constituents of foods or are nor-
           mally used in the preparation of foods. These products
           are  regarded  as foods  and, consequently, considered
           acceptable,  provided  that satisfactory  chemical and
           microbiological specifications can be established.

    (iv)   Enzymes  derived  from  non-pathogenic  microorganisms
           commonly   found  as  contaminants  of  foods.   These
           materials are not considered as foods.  It  is  neces-
           sary  to establish chemical and microbiological speci-
           fications  and to conduct short-term  toxicity studies
           to  ensure the absence of  toxicity.  Each preparation
           must  be  evaluated individually  and  an ADI  must be
           established.

    (v)    Enzymes derived from microorganisms that are less well
           known.   These  materials  also require  chemical  and
           microbiological   specifications  and  more  extensive
           toxicological  studies, including a long-term study in
           a rodent species.

    Safety  assessments for enzymes belonging to classes i - iii
will  be the  same regardless  of whether  the enzyme  is  added
directly  to food or is  used in an immobilized  form.  Separate

situations  should  be considered  with  respect to  the enzymes
described in classes iv and v:

    (i)    Enzyme  preparations  added  directly to  food but not
           removed.

    (ii)   Enzyme preparations added to food but removed from the
           final  product  according to  good manufacturing prac-
           tice.

    (iii)  Immobilized  enzyme  preparations that  are in contact
           with food only during processing.

    For (i) above, an ADI should be established to  ensure  that
levels  of  the enzyme  product present in  food are safe.   The
studies  indicated  in  these  guidelines  are  appropriate  for
establishing  ADIs (the guidelines  were originally drafted  for
this  situation).   For  (ii), an  ADI  "not  specified" may  be
established,  provided  that a  large  margin of  safety  exists
between  possible  residues  and their  acceptable  intake.  For
(iii), it may not be necessary to set an ADI for  residues  that
could occur in food as a result of using the immobilized form of
the  enzyme.  It is acceptable  to perform the toxicity  studies
relating to the safety of the enzyme on the  immobilized  enzyme
preparation,  provided that information  is given on  the enzyme
content in the preparation.

    (b)    Specifications for identity and purity

    Prior to revising existing specifications and developing new
specifications  for enzyme preparations for food processing, the
following data are necessary:

    (i)    A comprehensive description of the main enzymatic act-
           ivity (or activities), including the Enzyme Commission
           number(s) if any.

    (ii)   A list of the subsidiary enzymatic activities, whether
           they perform a useful function or not.

    (iii)  A clear description of the source.

    (iv)   A  list of non-enzymatic  substances derived from  the
           source material(s), with limits where appropriate.

    (v)    A list of added co-factors, with limits  where  appro-
           priate.

    (vi)   A  list of carriers  and diluents, with  limits  where
           appropriate.

    (vii)  A  list of preservatives  present from manufacture  or
           deliberately added, with limits where appropriate.

    (c)    Immobilizing agents

    A  number  of  procedures involving  different chemical sub-
stances  are  used  for immobilizing  enzymes.   These processes
include  microencapsulation (e.g., entrapment in gelatin to form
an  immobilized complex), immobilization  by direct addition  of
glutaraldehyde,  immobilization by entrapment in  porous ceramic
carrier,  and complexation with agents such as DEAE-cellulose or
polyethylenimine.   Several agents may  be used in  the immobil-
izing  process.  Substances derived  from the immobilizing  mat-
erial  may be in  the final product  due to either  the physical
breakdown  of  the immobilized  enzyme  system or  to impurities
contained  in  the  system.  The  amount  of  data necessary  to
establish  the safety of the  immobilizing agent depends on  its
chemical  nature.  The levels of  residues in the final  product
are expected to be extremely low.

    Some of the substances used in the preparation  of  immobil-
izing  systems are extremely  toxic.  The levels  of these  sub-
stances  or their contaminants  permitted in the  final  product
should  be at the lowest  levels that are technologically  feas-
ible,  provided that these levels are below those of any toxico-
logical concern.  An ADI will not be established.

2.  Natural  and Synthetic Food Colours (Adapted from: Reference
    1, Annex 6, p. 50)

    For  toxicological  evaluation,  natural colours  should  be
considered as falling within three main groups:

    (a)  A  colour isolated in a chemically unmodified form from
         a recognized foodstuff and used in the  foodstuff  from
         which  it is extracted at levels normally found in that
         food.   This  product could  be  accepted in  the  same
         manner  as  the food  itself  with no  requirement  for
         toxicological data.

    (b)  A  colour isolated in a chemically unmodified form from
         a  recognized foodstuff but used at levels in excess of
         those  normally found  in that  food or  used in  foods
         other  than  that from  which  it is  extracted.   This
         product  might  require the  toxicological data usually
         demanded   for  assessing  the  toxicity  of  synthetic
         colours.

    (c)  A  colour isolated from  a food source  and  chemically
         modified  during  its  production or  a  natural colour
         isolated  from a non-food source.  These products would
         also require a toxicological evaluation similar to that
         carried out for a synthetic colour.

    It is recognized that natural colours may be  reproduced  by
chemical  synthesis  but  it is  noted  that  "nature-identical"
colours  produced by chemical  synthesis may contain  impurities
warranting toxicological evaluation similar to that required for
a synthetically produced food colour.

    The toxicological evaluation of synthetic food colours would
require the following minimum data:

    (a)  Metabolic studies in several species, preferably inclu-
         ding  man.  These should include studies on absorption,
         distribution,  biotransformation, and elimination,  and
         an  attempt should be  made to identify  the  metabolic
         products in each of these steps.

    (b)  Short-term  feeding  studies in  a non-rodent mammalian
         species.

    (c)  Multi-generation reproduction/teratogenicity studies.

    (d)  Long-term   carcinogenicity/toxicity  studies  in   two
         species.

3.  Solvents Used in Food Processing (Adapted from: Reference 1,
    Annex 6, pp.  50-51)

    Extraction solvents are used  inter alia  in the extraction of
fats  and oils, defatting  fish and other  meals, and in  decaf-
feinating  coffee and  tea.  They  are chosen  mainly for  their
ability  to dissolve the  desired food constituents  selectively
and  for their volatility, which enables them to separate easily
from  the extracted material  with minimum damage.   The  points
raised by their use relate to:

    (a)  toxicity of their residues;

    (b)  toxicity of any impurities in them;

    (c)  toxicity  of substances such as solvent stabilizers and
         impurities that may be left behind after the solvent is
         removed; and

    (d)  toxicity  of any substances produced  as a result of  a
         reaction between the solvent and food ingredients.

    Before  any extraction solvent can be evaluated, information
is required on:

    (a)  identity  and  amount  of  impurities  in  the  solvent
         (including  those that are formed, acquired, or concen-
         trated owing to continuous reuse of the solvent);

    (b)  identity and amount of stabilizers and other additives;
         and

    (c)  toxicity   of  residues  of  solvents,  additives,  and
         impurities.

    Impurities are particularly important because there are wide
differences in the purities of food grade and  industrial  grade
solvents.   The food use  of extraction solvents  is  frequently
much  less than the industrial  use, and hence their  food-grade

requirements  may  receive  insufficient consideration,  both in
food  use and in toxicological testing.  Furthermore, the impur-
ities  or stabilizers may  not have the  same volatility as  the
solvent itself, and as a result, these may be left behind in the
food  after the solvent is removed.  Finally, the possibility of
any  solvent,  impurity,  stabilizer, or  additive reacting with
food ingredients should be checked.

    When  biological and toxicological data raise doubts about a
substance's safety, two approaches are possible: (a) to  set  an
ADI for the substance or (b) to discourage its  use  altogether.
Even when data indicate a wide margin of safety for a substance,
or  when there is  a paucity of  toxicological data on  the sub-
stance, but no problems concerning the impurities, residues, and
any  chemical reaction with food ingredients, it would be appro-
priate to limit the use of the substance to the minimum possible
level.

    When the data on a substance indicate the presence  of  cer-
tain  impurities in the  tested material, considerable  problems
arise in its evaluation.  This is especially true if industrial-
grade  rather than  food-grade material  has been  used  in  the
toxicological  study.  For example, when evaluating the solvents
1,1,1-trichloroethane,   trichloroethylene,   and   tetrachloro-
ethylene, it was noted that the toxicological data indicated the
presence  of  certain  known toxic  and carcinogenic substances.
The  interpretation  of  these data  became  extremely difficult
because  industrial-grade material had been used in the studies.
Only food-grade material should be used in toxicological studies
and the impurities in the material should be fully identified.

    Carrier  solvents raise somewhat different issues.  They are
used  for dissolving and  dispersing nutrients, flavours,  anti-
oxidants,  emulsifiers, and a wide variety of other food ingred-
ients and additives.  With the exception of carrier solvents for
flavours,  they tend  to occur  at higher  levels in  food  than
extraction  solvents,  mainly  because frequently  no attempt is
made  to remove them,  and because some  of them are  relatively
nonvolatile.   Since carrier solvents are  intentional additives
and  are often not removed from the processed food, it is impor-
tant  to evaluate their safety  together with the safety  of any
additives or stabilizers in them.

4.  Residues Arising from Use of Xenobiotic Anabolic  Agents  in
    Animal  Feed (Adapted from: Reference 1, p. 13 and reference
    3, p. 15)

    Many studies have established the importance and efficacy of
anabolic  agents for meat  production.  Two categories  of  com-
pounds are used - namely:

    (a)  hormones   that   are  identical   to  those  occurring
         naturally  in food-producing animals and  human beings,
         including the esters of these hormones; and

    (b)  xenobiotic  compounds, such as derivatives of hormones,
         synthetic  compounds  with hormonal  activity, natural-
         product hormonally active agents that are not identical
         with human endogenous hormones, and derivatives of such
         compounds.

    The  toxicological evaluation of residues of anabolic agents
that  are present in  human food obtained  from animals  treated
with  these agents must take into account whether the residue is
identical to a human endocrine hormone.  In the latter case, the
possible  endocrinological effects and carcinogenic potential of
the residue must be closely examined.

    Chemically  modified hormones, hormonally active agents from
plants,  and  synthetic  anabolic agents  present  the following
specific problems:

    (a)  extreme  potency  and  consequently the  need to ensure
         minimal residues;

    (b)  potential tumorigenic activity; and

    (c)  the  presence of their  metabolites in animal  products
         that  might  be  of endocrinological  or  toxicological
         consequence.

    The  evaluation  for acceptance  of  the use  of  xenobiotic
anabolic  agents  in animal  food  production resembles  in many
respects the evaluation of pesticides, since the  two  essential
elements required are:

    (a)  adequate, relevant toxicological data, and

    (b)  comprehensive  data  about  the  kinds  and  levels  of
         residues  when  the  substances are  used in accordance
         with  good  animal  husbandry practice,  which requires
         evidence  as to the efficacy of the anabolic agent, the
         amounts  used to produce the effect, the residue levels
         based on field trials, and information about methods of
         analysis  of  residue levels  that  could be  used  for
         control or monitoring purposes.

5.  Metals in Food (Adapted from: Reference 1, pp. 14-15)

    Toxicological  evaluation  of  metals  in  foods  calls  for
carefully balanced consideration of the following factors:

    (a)  nutritional  requirements, including nutritional inter-
         actions  with  other  constituents of  food  (including
         other metals when the interactions are nutritionally or
         toxicologically  relevant) in respect of, for instance,
         absorption, storage in the body, and elimination;

    (b)  the results of epidemiological surveys and formal toxi-
         cological  studies,  including interactions  with other
         constituents  of food (including other  metals when the

         interactions are nutritionally or toxicologically rele-
         vant), information about pharmaceutical and other medi-
         cinal  uses,  and  clinical observations  on  acute and
         chronic  toxicity  in  human experience  and veterinary
         practice;

    (c)  total intake on an appropriate time basis (e.g., daily,
         weekly,  yearly  or  lifetime) from  all sources (food,
         water,  air) of metals  as normal constituents  of  the
         environment, as environmental contaminants, and as food
         additives of an adventitious or deliberate nature.

    The  tentative tolerable daily intakes  proposed for certain
metals   by  the  Committee  provide  a  guideline  for  maximum
tolerable  exposure.  In the  case of essential  elements, these
levels exceed the normal daily requirements, but this should not
be construed as an indication of any change in  the  recommended
daily  requirements.  In  the case  of both  essential and  non-
essential   metals,  the  tentative  tolerable  intake  reflects
permissible human exposures to these substances as a  result  of
natural   occurrence  in  foods   or  various  food   processing
practices, as well as exposure from drinking-water.

    It  is important that the proposed tolerable intakes are not
used  as  guidelines for  fortifying  processed food,  since the
currently   accepted  values  for  required   daily  intake  are
sufficient to meet the known nutritional requirements.

6.  Packaging Materials (Adapted from: Reference 4, pp. 22-23)

    Many  substances may become food contaminants as a result of
the  use  of  food-contact materials.   The Committee considered
that  the general principles governing the use of food additives
established  by  the  WHO  Scientific  Group  on  Procedures for
Investigating  Intentional and Unintentional Food Additives, and
the  WHO Scientific  Group on  the Assessment  of  the  Carcino-
genicity and Mutagenicity of Chemicals, should be applied in the
overall  evaluation of substances migrating into food from food-
contact materials.

    Many  such materials  are made  of polymer  systems and  the
polymers  themselves are usually  inert, non-toxic, and  do  not
migrate  into  food.   However, monomers,  which  are inevitably
present  in the polymeric materials,  residual reactants, inter-
mediates,  manufacturing aids, solvents, and  plastic additives,
as  well as the products  of side reactions and  chemical degra-
dation,  may be present.  These substances may migrate into food
and  may be toxic.   Migration from food-contact  materials  may
arise  during  processing,  storage, and  preparation operations
such as heating, microwave cooking, or treatment  with  ionizing
radiation.    For   evaluation  purposes,   information  on  the
following is required:

    (a)  the  chemical identity and toxicological  status of the
         substances that enter food;

    (b)  the  possible exposure, details of which can be derived
         from   migration  studies  using   suitable  extraction
         procedures, and/or the analysis of food samples; and

    (c)  the  nature and  amount of  food in  contact  with  the
         packaging materials, and the intake of such food.

It  is necessary,  in many  instances, to  recommend that  human
exposure  to migrants from food-contact  materials be restricted
to  the lowest levels  technologically attainable.  One  way  to
achieve  this is to draw  up strict specifications limiting  the
quantities  in the materials.  It is also necessary to determine
whether food processing has an effect in generating  the  poten-
tially toxic substances in food-contact materials.

    Generally, evaluation will require the following:

    (a)  the lowest levels of potential migrants from within the
         polymeric system(s) that are technologically attainable
         with  improved manufacturing processes for food-contact
         materials;

    (b)  the resulting levels of the migrants in food;

    (c)  the intake of the foods; and

    (d)  the  most  appropriate  statistical  design  that  will
         enable  the implications for  health to be  interpreted
         from adequate and relevant toxicological data.

A  monitoring programme should  be established, with  a view  to
supplementing  the existing data on human exposure and providing
a  means  of  demonstrating a  reduction  in  such  exposure  as
techniques improve.  Priority in the programme should  be  given
to  the  substances with  the  greatest potential  for adversely
affecting human health.

REFERENCES TO ANNEX III

1.  FAO/WHO   (1982)  Evaluation of  certain food additives  and
    contaminants.  Twenty-sixth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 683).

2.  FAO/WHO   (1986)  Evaluation of  certain food additives  and
    contaminants.  Twenty-ninth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 733).

3.  FAO/WHO   (1981)   Evaluation  of  certain  food  additives.
    Twenty-fifth report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 669).

4.  FAO/WHO   (1984)  Evaluation of  certain food additives  and
    contaminants.  Twenty-eighth  report  of the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 710).

ANNEX IV.  EXAMPLES OF THE USE OF METABOLIC STUDIES IN THE 
SAFETY ASSESSMENT OF FOOD ADDITIVES

    As indicated in section 5.2, it is not feasible or desirable
to  develop simple guidelines for  pharmacokinetic and metabolic
studies.  In view of this, the examples given below  of  several
food  additives  and  contaminants on  which  a  great  deal  of
biochemical work has been done will serve to highlight the value
and  many  of  the problems  associated  with  the use  of these
studies  for determining mechanisms.  Clearly,  further research
will be needed to solve these problems.

1.  Sodium Cyclamate

    This compound represents a unique situation in toxicology in
that  it has been generally  agreed that levels of  a metabolite
rather  than the parent  compound should be  used for the  usual
safety  determinations.  The twenty-sixth Committee allocated an
ADI of 0 - 11 mg/kg body weight to cyclamate, calcium and sodium
salts, expressed as cyclamic acid (1).

    It   has  been  shown   that  a  metabolite   of  cyclamate,
cyclohexylamine  (an active pressor amine), appears in the urine
in  variable  amounts after  variable  time intervals  from  the
administration of cyclamate in rats (2, 3).  This metabolite has
been found to be produced by bacterial action in  the  intestine
(3, 4), but only after intestinal flora have undergone undefined
adaptive changes (2).  Therefore, it normally appears only after
a  latent  period.  However,  in  certain human  subjects,  some
immediate converters have been found to be present (2,  5).   In
both animals and man, the ability of intestinal flora to convert
cyclamate to cyclohexylamine varies widely with time in the same
individual.  The number of individuals able to convert cyclamate
to cyclohexylamine and the level at which this conversion occurs
have been factored into ADI determinations using  averages  from
some  studies (6).   However, it  is difficult to obtain  really
consistent figures, and those in use represent compromises.  The
ADI  is  based on  subsequent  studies, which  demonstrated that
cyclohexylamine, administered orally, induced testicular atrophy
in rats (7, 8).

    The  primary reason for  the decision by  JECFA and  various
national  regulatory bodies to agree to the use of the levels of
this  metabolite  rather than  the  parent compound  for toxico-
logical evaluation purposes appears to have been the  nature  of
the metabolite, in this instance, a compound that  is  pharmaco-
logically-active  relative to the parent compound and is capable
of  inducing testicular  atrophy in  rats (9).   However, it  is
questionable  that the readiness of  all these bodies to  accept
this unique approach in this situation is, in fact, appropriate.
The  presence  of the  metabolite  certainly cannot  be ignored;
however, it would seem more logical to demand that  the  effects
in  question  (testicular  atrophy)  be  demonstrated  following
feeding  the  parent  compound (cyclamate).   This  approach  is
complicated  by the inconsistent nature of the appearance of the
metabolite.

    The  individuality of the  response has necessitated  a most
conservative  attitude  that  has raised  the  important general
question  of how the  toxicologist can best  protect  vulnerable
individuals.   In  addition, this  example  has pointed  to  the
importance  of studying metabolism by gut flora in toxicological
evaluations.  Unquestionably, variations in gut flora are one of
the  more important determinants of species differences, and the
example  of cyclamate has  pointed the way  to studies in  which
this factor has been reduced to a minimum.

2.  Sodium Saccharin

    JECFA has reviewed the safety of saccharin many  times.   In
1984,  the twenty-eighth Committee  allocated a temporary  group
ADI  of 0 -  2.5 mg/kg body  weight to saccharin,  including its
calcium,  potassium, and sodium  salts (10).  Subsequent  to the
demonstration  in  some  long-term toxicity  studies that sodium
saccharin  could induce tumours of  the urinary bladder in  male
rats  at high dose levels, much work was undertaken in an effort
to  explain this  phenomenon.  Only  two of  the  many  reported
findings  will be discussed in this section, as illustrations of
certain general principles.

    On the basis of the first series of studies, it  would  seem
unquestionable that sodium saccharin is not metabolized  in  the
rat; this seems to be generally applicable to human  beings  and
other  species (11).  There  is no postulated  theoretical mech-
anism  of chemical carcinogenesis  that could fit  this picture.
The  second series of studies  have established the most  inter-
esting finding that, although saccharin is not  metabolized,  it
can  modify the metabolic pathway of certain normal constituents
in  the  diet.  A  dose-related  increase in  certain tryptophan
metabolites - notably indoxyl sulfate - was found in  the  urine
of  saccharin-treated  rats  (12).  In  contrast,  these effects
could  not be demonstrated  in human beings  consuming saccharin
(13).   In  view  of previous  interest  in  the association  of
tryptophan and its metabolites with bladder tumour induction and
its occurrence in man, this observation was of  great  interest.
Although the further steps in this series of investigations have
not  succeeded in establishing a convincing relationship between
the  carcinogenicity finding and  the metabolism of  tryptophan,
nevertheless,  a most important general  principle in toxicology
has  been demonstrated that remains  to be exploited.  The  fact
that  a  compound  that is  not  metabolized  could  change  the
metabolism  of other compounds provides  a basis for studies  of
mechanism  of action  not considered  in the  past.   It  seemed
likely,  initially, that the  possible bacteriostatic action  of
saccharin  might be affecting the gut flora; although this seems
to  be a possible practical explanation, further study is needed
to explain the whole picture.

3.    o -Phenylphenol (OPP)

    This compound is a fungicide widely used on oranges, the use
of  which results in low  residue levels as a  food contaminant.
The Joint Meeting on Pesticide Residues (JMPR) has  allocated  a

temporary  ADI to OPP (and  its sodium salt) of  0 - 0.02  mg/kg
body weight (14).  It has been chosen as an example of  a  situ-
ation  in which extensive metabolic studies have been correlated
with toxicological findings.

    OPP  has been found to  give rise to tumours  of the urinary
bladder when fed to rats at relatively high levels in  the  diet
(15).  Two metabolites, the glucuronide conjugate of OPP and the
sulfate  ester conjugate  of OPP,  have been  identified in  the
urine of rats after administration of different levels  of  OPP.
A  third metabolite, tentatively  postulated to be  a conjugated
dihydroxybiphenyl compound, has been reported after the adminis-
tration  of a  high level  of OPP,  the same  level required  to
induce  bladder  tumours, but  not  after the  administration of
lower levels of OPP (16).  From this observation,  some  invest-
igators  have concluded that, at lower levels of administration,
carcinogenicity  does not occur, because  the "proximate carcin-
ogen"  (i.e., the  high-level  metabolite) has  not been formed.
This fascinating study is, unfortunately, incomplete.  There are
disputes  as to  the absence  of the  "active"  metabolite  (the
conjugated   dihydroxybiphenyl  compound)   at  the  lower  dose
levels; the detection limits of the various  metabolites,  which
are  not clearly delineated in  the available literature, are  a
matter  of  considerable  concern.  In  addition,  the  "active"
metabolite  has no special  chemical characteristics to  suggest
that it would conform to any of the current theories  of  action
of chemical carcinogens.  These results show the difficulties of
proving the mechanism of a tumourigen so that a safe dose can be
established, even though the available evidence points  in  that
direction.   It is clear  that much more  information is  needed
before  the postulated change  in metabolism can  be related  to
carcinogenesis.

4.  Methylene Chloride

    This compound, which is used as a food extraction solvent in
some  countries,  has been  the  subject of  intensive  carcino-
genicity,  mutagenicity,  metabolism, pharmacokinetic,  and epi-
demiological studies, but questions still exist about its safety
of use.  Because of the inadequacies of the studies available at
that  time, the twenty-sixth Committee  withdrew the previously-
allocated ADI for methylene chloride (17).

    The safety of methylene chloride was brought  into  question
by a long-term inhalation study that produced very rare salivary
gland sarcomas in male rats in an apparent  dose-response  rela-
tionship at 1500 and 3500 mg/kg air; an increased  incidence  of
tumours  was  not observed  in female rats  or in hamsters  in a
parallel study (unpublished studies by the Dow Chemical Company,
Midland, Michigan, USA, 1980).  Also, preliminary results from a
mouse  inhalation study indicate an increased incidence of liver
and lung neoplasms at 1000 and 4000 mg methylene chloride/kg air
(18).  However, arguments have been made on the basis  of  meta-
bolic studies that the maximum tolerated dose (MTD) was exceeded
in  these long-term inhalation  studies in rats  and mice  (19).
Bioassays  in which methylene chloride was administered at lower

levels by inhalation or in drinking-water have not resulted in a
significant  increase in malignant tumours  (unpublished studies
by the Dow Chemical Company, Midland, Michigan, USA, 1982 and by
the National Coffee Association, USA, 1982).

    Methylene  chloride  is  metabolized via  two pathways.  The
principle  site  of  metabolism is  the  liver  in  all  species
studied,  including  man.   One  pathway  involves  glutathione,
giving  rise to formaldehyde, which  is oxidized to formic  acid
and then carbon dioxide.  The other pathway is mediated by cyto-
chrome P-450 and involves dehydrochlorination to carbon monoxide
and  hydrogen chloride.  One of  the intermediates in the  first
pathway,  a glutathione conjugate,  has been implicated  as  the
DNA-reactive  species responsible for the  apparent mutagenicity
of  methylene chloride in some tests.  However, there is no evi-
dence of alkylation in animals (20, 21, 22).

    There  is  a  linear discontinuity  in  metabolite formation
(carbon  dioxide and carbon  monoxide) as exposure  to methylene
chloride  increases.  For example,  it has been  shown that,  on
inhalation  of 174  mg methylene  chloride/m3 (50  ppm), 95%  is
metabolized,  while at  1750 mg/m3 air  (500 ppm),  only 69%  is
metabolized and at 5200 mg methylene chloride/m3 air (1500 ppm),
only 45% is metabolized to carbon dioxide and  carbon  monoxide.
Both oral and inhalation studies show that saturation  of  meta-
bolism  occurs in all species examined (rat, mouse, hamster, and
man) (19).

    Greater  amounts of methylene chloride  are metabolized when
the compound is presented in drinking water than when  the  same
daily dose is gavaged in a single dose, either in corn oil or in
water.   Administration of  a large  number of  doses  in  small
amounts, such as when methylene  chloride is administered in the
drinking-water,  yields greater amounts of metabolites than when
the total amount is given at one time, such as by  gavage.   The
vehicle used in gavage studies also plays a role in  the  clear-
ance  of methylene chloride  from various tissues;  for example,
the  compound is dissipated  from both blood  and liver in  less
than 2 h after administration by gavage in water compared with a
residence time of about 8 h in venous blood and over 25 h in the
liver after administration by gavage in corn oil (19).

    Despite  tremendous efforts to study  this compound biochem-
ically,  a clear  picture of  the mechanism  of  its  biological
effects  has not emerged.  This shows the difficulties of devel-
oping sufficient biochemical data to set a safe dose level for a
substance  that causes cancer at  high dose levels.  The  satur-
ation effect and the occurrence of tumours at high  dose  levels
may  be related.  An  encouraging point is  that primates  meta-
bolize  chlorinated solvents  to a  lesser extent  than rats  or
mice; thus, less of the glutathione reactive intermediate, which
has  been  postulated as  being  responsible for  the  genotoxic
effects  of methylene chloride, should be present in man than in
the  animals exhibiting the  deleterious effects.  Finally,  the

differences  in  rates  of  metabolism  of  methylene  chloride,
depending  on the route of administration, point to the need for
very  careful assessment of the appropriate route of administra-
tion to mimic exposure in man.

5.  Trichloroethylene

    This  chemical  is  an  extraction  solvent.   It  has  been
reviewed  by JECFA,  but an  ADI has  not been  allocated  (17).
Trichloroethylene has been found to cause an increased incidence
of hepatocellular carcinomas in mice, but not in  rats  (23,24).
The  earlier  bioassays  were  performed  with  industrial-grade
trichloroethylene, which contained epoxide stabilizers, at least
one of which is a potent mutagen (section 3 in Annex III).  How-
ever,  the results of later studies using non-epoxide stabilized
material  confirmed the results  of the earlier  studies,  indi-
cating that the stabilizers were not responsible for the hepato-
carcinogenicity observed in mice.

    Trichloroethylene  has  been subjected  to  a great  deal of
metabolic  and pharmacokinetic research on mice and rats, and an
interesting story is emerging, which could explain  the  differ-
ence  in response between these species and the relevance to man
of the rodent bioassays.

    Electron micrographs of liver tissue from mice that had been
dosed  with a high level of trichloroethylene for 10 days showed
a  proliferation  of peroxisomes;  significant proliferation was
not observed in rats after the administration of the same amount
of  trichloroethylene (25).  Biochemical studies have shown that
cyanide-insensitive  acyl CoA oxidase, an  enzyme present within
the  peroxisome that ultimately  produces hydrogen peroxide,  is
enhanced   5 -  16-fold  in  the  peroxisome-proliferated  state
compared  within the control.  However, only a small increase in
catalase  activity, which catalyses  the conversion of  hydrogen
peroxide  to water, has been  observed in the proliferated  cell
(25, 26, 27).  It has been postulated that the large increase in
acyl  CoA oxidase activity coupled with the marginal increase in
catalase activity leads  to an increased steady-state concentra-
tion of hydrogen peroxide within the liver cell,  which  results
in  cytotoxicity and DNA  damage and eventually  cancer in  mice
(28).

    Trichloroethylene   appears   to  be   metabolized  via  the
cytochrome  P-450 system, in  both the rat  and mouse (29,  30).
The  major metabolite, which  is ultimately converted  to carbon
dioxide,  is trichloroacetic acid.   Apparently, the enzymes  in
this  pathway  are  not induced to the same extent in rats as in
mice,  as shown by  gavage studies with  trichloroethylene; mice
show linear kinetics with respect to metabolite formation over a
wide dose range, while rats show saturation kinetics; saturation
is  observed  at  low levels  relative  to  amounts required  to
observed  peroxisome proliferation and the  induction of hepato-
cellular  carcinomas in mice.  In contrast, when trichloroacetic
acid  is administered to rats and mice, a similar dose-dependent
large increase in peroxisomes is observed in both species.  This

suggests  that trichloroacetic acid is  the proximate peroxisome
proliferator,  and the reason that proliferation is not observed
after  the administration of  trichloroethylene to rats  is that
not  enough  of the  acid is produced  to cause the  effect.  If
peroxisome  proliferation  (and  a consequent  increased steady-
state  concentration of hydrogen peroxide  in the cell) is  res-
ponsible for hepatocellular carcinoma induction, then trichloro-
acetic  acid  should  be  a  hepatocellular  carcinogen  in both
species.  This hypothesis is at presently being tested (28).

    Where  does man fit into the picture?  Preliminary work with
hepatocytes  isolated from  mice, rats,  and man  show that  the
human  hepatocyte is much more like that of the rat than that of
the mouse in terms of its ability to  convert  trichloroethylene
to trichloroacetic acid; in fact, the human hepatocyte  is  even
less  active in converting trichloroethylene  to trichloroacetic
acid  than that of the  rat (28).  In addition,  when trichloro-
acetic acid is administered to cultured human hepatocytes, there
is  no  evidence of  peroxisome  proliferation, as  measured  by
cyanide-insensitive  acyl  CoA  oxidase activity.   In contrast,
studies  on  mouse and  rat  hepatocytes have  registered large,
dose-related  increases  in  this enzyme  activity.   These data
suggest  that the  mouse bioassay  data showing  an increase  in
hepatocellular carcinomas after the administration of trichloro-
ethylene are not appropriate for human beings because:

    (a)  trichloroethylene is not converted into trichloroacetic
         acid at a high enough rate in man to  cause  peroxisome
         proliferation; and

    (b)  trichloroacetic  acid does not  appear to cause  perox-
         isome  proliferation in man at  the levels at which  it
         causes the effect in rats and mice.

Further data are needed to obtain a firm conclusion  that  could
withstand regulatory scrutiny.

6.  Estragole

    This  chemical  is a  naturally-occurring anisole derivative
that is used as a flavouring agent.  Estragole has been reviewed
by JECFA, but an ADI has not been allocated (31).  It  has  been
found  to be a mouse carcinogen at a dose level of approximately
500  mg/kg  body  weight per  day  (32,  33).  In  contrast, the
estimated  human  daily  intake of  estragole  in  the  diet  is
approximately 1 g/kg body weight (349).

    One  of the routes of  metabolism of estragole is  through a
hydroxylated intermediate, 1'-hydroxyestragole (35).  This "act-
ivated"  intermediate is likely to  undergo esterification reac-
tions  with cellular constituents  to form electrophilic  conju-
gates.  It has been postulated that the carcinogenic effect seen
with estragole is due to the formation of the "proximate carcin-
ogen",  1'-hydroxyestragole, which reacts to  form the "ultimate
carcinogen", the electrophilic conjugate (32, 33).

    Studies  have  been  performed in  which  the  level of  1'-
hydroxyestragole  has been measured in the urine of mice exposed
to various levels of estragole.  At the low dose, 0.5 mg/kg body
weight,  1 - 2% of the dose was excreted as 1'-hydroxyestragole,
while, at the high dose, 1000 mg/kg body weight, 10 - 15% of the
ingested  dose  of estragole  was  excreted as  the hydroxylated
compound.  In studies on human volunteers, fed  1 g  estragole,
0.3% of the dose was excreted as 1'-hydroxyestragole (34).

    These  and  other data  suggest that only  at very high  and
overwhelming  levels of estragole are significant amounts of the
activated  intermediate formed, and that there appear to be over
6  orders  of  magnitude  difference  between  levels   of   the
intermediate  in the high-dose mouse study and the level present
in  man at normal levels of consumption.  This hypothesis, if it
holds,  leads to the conclusion that the human carcinogenic risk
from the ingestion of normal levels of estragole is negligible.

REFERENCES TO ANNEX IV

 1. FAO/WHO   (1982)  Evaluation of  certain food additives  and
    contaminants.  Twenty-sixth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 683).

 2. RENWICK,   A.G.  &  WILLIAMS,  R.T.   (1972)   The  fate  of
    cyclamate in man and other species. Biochem. J.,  129:  869-
    879.

 3. BICKEL,  M.H., BURKARD, B.,  MEIER-STRASSER, E., &  VAN  DEN
    BROEK-BOOT,   M.   (1974)   Entero-bacterial   formation  of
    cyclohexylamine in rats ingesting cyclamate. Xenobiotica, 4:
    425-439.

 4. DRASSAR,  B.S., RENWICK, A.G., & WILLIAMS, R.T.  (1972)  The
    role  of  the  gut flora  in  the  metabolism of  cyclamate.
    Biochem. J., 129: 881-890.

 5. ASAHINA,  M.,  YAMAHA,  T.,  WATANABE,  K.,  &  SARRAZIN, G.
    (1971)    Excretion  of  cyclohexylamine,  a  metabolite  of
    cyclamate, in human urine. Chem. Pharm. Bull.  (Tokyo),  19:
    628-632.

 6. RENWICK, A.G.  (1983)  The fate of cyclamate in man and rat.
    In:  Transcripts of the European  Toxicological Forum, 18-22
    October 1983, Geneva, pp. 301-312.

 7. GAUNT, I.F., SHARRATT, M., GRASSO, P., LANSDOWN,  A.B.G.,  &
    GANGOLLI,  S.D.  (1974)  Short-term toxicity  of cyclohexyl-
    amine  hydrochloride in the rat. Food Cosmet.  Toxicol., 12:
    609.

 8. MASON,  P.L. & THOMPSON, G.R.  (1977)  Testicular effects of
    cyclohexylamine  hydrochloride  in  the rat.  Toxicology, 8:
    143.

 9. FAO/WHO   (1978)   Evaluation  of  certain  food  additives.
    Twenty-first report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 617).

10. FAO/WHO   (1984)  Evaluation of  certain food additives  and
    contaminants.  Twenty-eighth  report  of the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 710).

11. RENWICK,  A.G.   (1985)   The disposition  of  saccharin  in
    animals  and man: a  review. Food Chem.  Toxicol., 23:  429-
    435.

12. SIMS, J. & RENWICK, A.G.  (1985)  The  microbial  metabolism
    of tryptophan in rats fed a diet containing  7.5%  saccharin
    in  a two-generation protocol. Food Chem. Toxicol., 23: 437-
    444.

13. ROBERTS, A. & RENWICK, A.G.  (1985)  The effect of saccharin
    on the microbial metabolism of tryptophan in man. Food Chem.
    Toxicol., 23: 451-455.

14. FAO/WHO  (1985)  Pesticide residues in food. Report  of  the
    Joint  Meeting of  the FAO  Panel of  Experts  on  Pesticide
    Residues  in Food and the Environment and a WHO Expert Group
    on  Pesticide Residues (FAO Plant  Production and Protection
    Paper No. 68).

15. HIRAGA,  K. & FUJII, T.  (1984)  Induction of tumours of the
    urinary bladder in F-344 rats by dietary  administration  of
    o-phenylphenol. Food Cosmet. Toxicol., 22: 865-870.

16. REITZ,  R.H.,  FOX,  T.R.,  QUAST,  J.F.,  HERMANN,  E.A., &
    WATANABE, P.G.  (1983)  Molecular mechanisms involved in the
    toxicity  of orthophenylphenol and  its sodium salt.  Chem.-
    Biol. Interact., 43: 99-119.

17. FAO/WHO   (1983)  Evaluation of  certain food additives  and
    contaminants.  Twenty-seventh  report  of the  Joint FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 696 and corrigenda).

18. FOOD   CHEMICAL   NEWS    (1985)   Methylene   chloride   is
    carcinogenic  in inhalation tests  conducted for NTP,  March
    25, p. 38.

19. KIRSCHMAN,  J.   (1984)  Methylene  chloride: safety testing
    overview.  In: Transcripts of the European Toxicology Forum,
    18-21 September 1984, Geneva, pp. 197-210.

20. AHMED, A. & ANDERS, M.  (1976)  Metabolism of dihalomethanes
    to  formaldehyde and inorganic halide. I.  In vitro  studies.
    Drug Metab. Dispos., 4: 357-361.

21. AHMED,  A., ET AL.  (1980)  Halogenated methanes: metabolism
    and toxicity. Fed. Proc., 39: 3150-3155.

22. GREEN,  T.  (1984)  Genotoxicity of  methylene chloride. In:
    Transcripts   of   the  European   Toxicology  Forum,  18-21
    September 1984, Geneva, pp. 211-215.

23. NCI   (1976)   Carcinogenesis bioassay  of trichloroethylene
    (CAS  Registry No. 79-01-6), National Cancer Institute (DHEW
    Publication No. (NIH) 76-802).

24. NTP   (1983)   Draft  report  abstracts  on  nine   chemical
    carcinogenesis  bioassays, National Toxicology  Program, pp.
    767-768 (Chemical Regulation Report No. 6).

25. ELCOMBE,   C.R.,   ROSE,   M.S.,  &   PRATT,   I.S.   (1985)
    Biochemical,  histological,  and ultrastructural  changes in
    rat   and  mouse  liver  following   the  administration  of
    trichloroethylene: possible relevance to species differences
    in  hepatocarcinogenicity.  Toxicol.  appl. Pharmacol.,  79:
    365-376.

26. LALWANI, N.D., REDDY, M.K., MANGKORNKANOL-MARK, M., & REDDY,
    J.K.    (1981)    Induction,  immunochemical   identity  and
    immunofluorescence   localization  of  an  80 000-molecular-
    weight    peroxisome-proliferation-associated    polypeptide
    (polypeptide  PPA-80) and peroxisomal enoyl-CoA hydratase of
    mouse liver and renal cortex. Biochem. J., 198: 177-186.

27. REDDY,  J.K., LALWANI, N.D.,  QURESHI, S.A., REDDY,  M.K., &
    MOEHLE,   C.M.   (1984)   Induction  of  hepatic  peroxisome
    proliferation in non-rodent species, including primates. Am.
    J. Pathol., 114: 171-183.

28. ELCOMBE,  C.R.   (1984)   In: Transcripts  of  the  European
    Toxicology  Forum,  18-21  September 1984,  Geneva, pp. 134-
    144.

29. GREEN,  T.  & PROUT,  M.S.   (1985)  Species  differences in
    response to trichloroethylene. II. Biotransformation in rats
    and mice. Toxicol. appl. Pharmacol., 401-411.

30. PROUT,  M.S.,  PROVAN, W.M.,  &  GREEN, T.   (1985)  Species
    differences    in   response   to    trichloroethylene.   I.
    Pharmacokinetics   in   rats   and  mice.   Toxicol.   appl.
    Pharmacol., 79: 389-400.

31. FAO/WHO   (1981)   Evaluation  of  certain  food  additives.
    Twenty-fifth report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 669).

32. DRINKWATER,  N.R., MILLER, E.C., MILLER, J.A., & PITOT, H.C.
    (1976)   The hepatocarcinogenicity of  estragole (1-allyl-r-
    methoxybenzene) and 1'-hydroxyestragole in the mouse and the
    mutagenicity  of  1'-acetoxyestragole  in bacteria.  J. Natl
    Cancer Inst., 57: 1323-1331.

33. MILLER, E.C., SWANSON, A.B., PHILLIPS, D.H., FLETCHER, T.L.,
    LIEM, A., & MILLER, J.A.  (1983)  Structure-activity studies
    of  the  carcinogenicities  in the  mouse  and  rat of  some
    naturally occurring and synthetic alkenylbenzene derivatives
    related  to safrole and  estragole. Cancer Res.,  43:  1124-
    1134.

34. BERNARD,  B. & CALDWELL,  J.  (1984)  Significance  of dose-
    dependent  metabolism of flavour chemicals  for their safety
    evaluation.  In:  Transcripts  of  the  European  Toxicology
    Forum, 18-21 September 1984, Geneva, pp. 124-131.

35. SWANSON,  A.B., MILLER, E.C.,  & MILLER, J.A.   (1981)   The
    side-chain    epoxidation    and   hydroxylation    of   the
    hepatocarcinogens  safrole  and  estragole and  some related
    compounds  by  rat  and  mouse  liver  microsomes.  Biochim.
    Biophys. Acta, 673: 505-516.


ANNEX V.  APPROXIMATE RELATION OF PARTS PER MILLION IN THE DIET 
TO MG/KG  BODY WEIGHT PER DAY a
--------------------------------------------------------------------------------
Animal         Weight    Food con-   Type of      1 ppm in     1 mg/kg body
                (kg)     sumed per     diet         food =       weight per
                          day (g)               (mg/kg body     day = (ppm
                          (liquids               body weight      of diet)
                          omitted)                per day)
--------------------------------------------------------------------------------
Mouse           0.02          3                      0.150            7

Chick           0.40         50                      0.125            8

Rat (young)     0.10         10        Dry           0.100           10
                                       labor-
Rat (old)       0.40         20        atory         0.050           20
                                       chow
Guinea-pig      0.75         30        diets         0.040           25

Rabbit          2.0          60                      0.030           33

Dog            10.0         250                      0.025           40
--------------------------------------------------------------------------------
Cat               2         100                      0.050           20
                                       Moist,
Monkey            5         250        semi-         0.050           20
                                       solid
Dog              10         750        diets         0.075           13

Man              60        1500                      0.025           40
--------------------------------------------------------------------------------
Pig or sheep     60        2400                      0.040           25
                                       Rela-
Cow             500        7500        tively        0.015           65
(maintenance)                          dry
                                       grain
Cow             500      15 000        forage        0.030           33
(fattening)                            mix-
                                       tures
Horse           500      10 000                      0.020           50
--------------------------------------------------------------------------------
a   Lehman, A.J.  (1954)  Association of Food and Drug Officials
    Quarterly  Bulletin, 18: 66.  The  values in this table  are
    average figures, derived from numerous sources.

Example:  What is the value in ppm and mg/kg body weight per day  of 0.5% 
          substance  X mixed in  the diet of  a  rat?

Solution: I.  0.5% corresponds to 5000 ppm.

          II. From  the table, 1 ppm in the diet of a rat is equivalent to 
              0.050 mg/kg body weight per day.  Consequently,  5000  ppm is  
              equivalent to 250 mg/kg body weight per day (5000 x 0.050).

ANNEX VI.  REPORTS AND OTHER DOCUMENTS RESULTING FROM MEETINGS 
OF THE JOINT FAO/WHO EXPERT COMMITTEE ON FOOD ADDITIVES

 1. FAO/WHO   (1957)   General  principles governing  the use of
    food  additives.  First report  of the Joint  FAO/WHO Expert
    Committee  on Food Additives (FAO  Nutrition Meetings Report
    Series  No. 15; WHO Technical Report Series No. 129) (out of
    print).

 2. FAO/WHO   (1958)  Procedures for the  testing of intentional
    food  additives to establish  their safety for  use.  Second
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 17; WHO
    Technical Report Series No. 144) (out of print).

 3. FAO/WHO   (1962)  Specifications for identity  and purity of
    food  additives (microbial preservatives  and antioxidants).
    Third report of the Joint FAO/WHO Expert Committee  on  Food
    Additives.  These  specifications were  subsequently revised
    and published as Specifications for identity and  purity  of
    food   additives.   I.   Antimicrobial   preservatives   and
    antioxidants,  Rome,  Food and  Agricultural Organization of
    the United Nations (out of print).

 4. FAO/WHO   (1963)  Specifications for identity  and purity of
    food  additives (food colours).  Fourth report of  the Joint
    FAO/WHO   Expert   Committee   on  Food   Additives.   These
    specifications  were  subsequently revised  and published as
    Specifications  for identity and  purity of food  additives.
    II.  Food colours, Rome, Food  and Agricultural Organization
    of the United Nations (out of print).

 5. FAO/WHO   (1961)  Evaluation of the  carcinogenic hazards of
    food  additives. Fifth report  of the Joint  FAO/WHO  Expert
    Committee  on Food Additives (FAO  Nutrition Meetings Report
    Series  No. 29; WHO Technical Report Series No. 220) (out of
    print).

 6. FAO/WHO  (1962)  Evaluation of the toxicity of a  number  of
    antimicrobials  and antioxidants. Sixth report  of the Joint
    FAO/WHO  Expert Committee on  Food Additives (FAO  Nutrition
    Meetings Report Series No. 31; WHO Technical  Report  Series
    No. 228) (out of print).

 7. FAO/WHO   (1964)  Specifications for the identity and purity
    of   food  additives  and  their  toxicological  evaluation:
    emulsifiers,  stabilizers,  bleaching  and maturing  agents.
    Seventh report of the Joint FAO/WHO Expert Committee on Food
    Additives  (FAO Nutrition Meetings Report Series No. 35; WHO
    Technical Report Series No. 281) (out of print).

 8. FAO/WHO   (1965)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: food
    colours  and  some  antimicrobials and  antioxidants. Eighth
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 38; WHO
    Technical Report Series No. 309) (out of print).

 9. FAO/WHO   (1965)  Specifications for identity and purity and
    toxicological   evaluation   of   some  antimicrobials   and
    antioxidants  (FAO Nutrition Meetings Report Series No. 38A;
    WHO/Food Add/24.65 (out of print).

10. FAO/WHO   (1966)  Specifications for identity and purity and
    toxicological  evaluation  of  food colours  (FAO  Nutrition
    Meetings Report Series No. 38B; WHO/Food Add/66.25).

11. FAO/WHO   (1966)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    antimicrobials,   antioxidants,   emulsifiers,  stabilizers,
    flour-treatment  agents, acids, and  bases. Ninth report  of
    the  Joint FAO/WHO Expert  Committee on Food  Additives (FAO
    Nutrition  Meetings  Report  Series No.  40;  WHO  Technical
    Report Series No. 339) (out of print).

12. FAO/WHO     (1967)    Toxicological   evaluation   of   some
    antimicrobials,   antioxidants,   emulsifiers,  stabilizers,
    flour-treatment  agents,  acids,  and bases  (FAO  Nutrition
    Meetings Report Series No. 40A,B,C; WHO/Food Add/67.29).

13. FAO/WHO   (1967)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    emulsifiers  and  stabilizers and  certain other substances.
    Tenth report of the Joint FAO/WHO Expert Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 43; WHO
    Technical Report Series No. 373).

14. FAO/WHO   (1968)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    flavouring  substances and non-nutritive  sweetening agents.
    Eleventh  report of the  Joint FAO/WHO Expert  Committee  on
    Food Additives (FAO Nutrition Meetings Report Series No. 44;
    WHO Technical Report Series No. 383).

15. FAO/WHO  (1968)  Toxicological evaluation of some flavouring
    substances   and   non-nutritive   sweetening  agents   (FAO
    Nutrition   Meetings   Report   Series  No.   44A;  WHO/Food
    Add/68.33).

16. FAO/WHO   (1969)   Specifications and  criteria for identity
    and  purity of some flavouring  substances and non-nutritive
    sweetening  agents (FAO Nutrition Meetings Report Series No.
    44B; WHO/Food Add/69.31).

17. FAO/WHO   (1969)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    antibiotics.  Twelfth  report  of the  Joint  FAO/WHO Expert
    Committee  on Food Additives (FAO  Nutrition Meetings Report
    Series No. 45; WHO Technical Report Series No. 430).

18. FAO/WHO   (1969)  Specifications for the identity and purity
    of  some antibiotics (FAO  Nutrition Meetings Report  Series
    No. 45A; WHO/Food Add/69.34).

19. FAO/WHO   (1970)  Specifications for the identity and purity
    of  food additives and their  toxicological evaluation: some
    food  colours, emulsifiers, stabilizers,  anticaking agents,
    and certain other substances. Thirteenth report of the Joint
    FAO/WHO  Expert Committee on  Food Additives (FAO  Nutrition
    Meetings Report Series No. 46; WHO Technical  Report  Series
    No. 445).

20. FAO/WHO   (1970)   Toxicological  evaluation  of  some  food
    colours,  emulsifiers,  stabilizers, anticaking  agents, and
    certain  other  substances  (FAO Nutrition  Meetings  Report
    Series No. 46A; WHO/Food Add/70.36).

21. FAO/WHO   (1970)  Specifications for the identity and purity
    of  some food colours, emulsifiers,  stabilizers, anticaking
    agents,  and  certain  other food  additives  (FAO Nutrition
    Meetings Report Series No. 46B; WHO/Food Add/70.37).

22. FAO/WHO   (1971)  Evaluation of food  additives:  specifica-
    tions  for the  identity and  purity of  food additives  and
    their toxicological evaluation: some extraction solvents and
    certain  other substances; and a review of the technological
    efficacy of some antimicrobial agents.  Fourteenth report of
    the  Joint FAO/WHO Expert  Committee on Food  Additives (FAO
    Nutrition  Meetings  Report  Series No.  48;  WHO  Technical
    Report Series No. 462).

23. FAO/WHO  (1971)  Toxicological evaluation of some extraction
    solvents   and  certain  other  substances   (FAO  Nutrition
    Meetings Report Series No. 48A; WHO/Food Add/70.39).

24. FAO/WHO   (1971)  Specifications for the identity and purity
    of  some  extraction  solvents and  certain other substances
    (FAO  Nutrition  Meetings  Report Series  No.  48B; WHO/Food
    Add/70.40).

25. FAO/WHO  (1971)  A review of the technological  efficacy  of
    some  microbial agents (FAO Nutrition Meetings Report Series
    No. 48C; WHO/Food Add/70.41).

26. FAO/WHO  (1972)  Evaluation of food additives: some enzymes,
    modified    starches,   and   certain    other   substances:
    toxicological evaluations and specifications and a review of
    the  technological efficacy of some  antioxidants. Fifteenth
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 50; WHO
    Technical Report Series No. 488).

27. FAO/WHO   (1972)  Toxicological evaluation of  some enzymes,
    modified   starches,  and  certain  other   substances  (FAO
    Nutrition  Meetings Report Series No. 50A; WHO Food Additive
    Series No.  1).

28. FAO/WHO   (1972)  Specifications for the identity and purity
    of  some enzymes and certain other substances (FAO Nutrition
    Meetings Report Series No. 50B; WHO Food Additive Series No.
    2).

29. FAO/WHO  (1972)  A review of the technological  efficacy  of
    some  antioxidants  and  synergists (FAO  Nutrition Meetings
    Report Series No. 50C; WHO Food Additive Series No. 3).

30. FAO/WHO   (1972)  Evaluation of  certain food additives  and
    the  contaminants  mercury,  lead,  and  cadmium.  Sixteenth
    report  of  the  Joint  FAO/WHO  Expert  Committee  on  Food
    Additives  (FAO Nutrition Meetings Report Series No. 51; WHO
    Technical Report Series No. 505 and corrigendum).

31. FAO/WHO   (1972)  Evaluation of mercury,  lead, cadmium, and
    the food additives amaranth, diethylpyrocarbamate, and octyl
    gallate (FAO Nutrition Meetings Report Series No.  51A;  WHO
    Food Additives Series No. 4).

32. FAO/WHO   (1974)   Toxicological evaluation  of certain food
    additives  with  a  review  of  general  principles  and  of
    specifications.  Seventeenth  report  of the  Joint  FAO/WHO
    Expert  Committee on Food Additives  (FAO Nutrition Meetings
    Report  Series No. 53; WHO  Technical Report Series No.  539
    and corrigendum) (out of print).

33. FAO/WHO   (1974)   Toxicological evaluation  of certain food
    additives   including  anticaking  agents,   antimicrobials,
    antioxidants,   emulsifiers,  and  thickening   agents  (FAO
    Nutrition Meetings Report Series No. 53A; WHO Food Additives
    Series No. 5).

34. FAO/WHO   (1978)  Specifications for the identity and purity
    of  thickening  agents,  anticaking agents,  antimicrobials,
    antioxidants,  and emulsifiers (FAO Food and Nutrition Paper
    No. 4).

35. FAO/WHO   (1974)   Evaluation  of  certain  food  additives.
    Eighteenth report of the Joint FAO/WHO Expert  Committee  on
    Food Additives (FAO Nutrition Meetings Report Series No. 54;
    WHO Technical Report Series No. 557 and corrigendum).

36. FAO/WHO   (1975)   Toxicological  evaluation  of  some  food
    colours,  enzymes, flavour enhancers, thickening agents, and
    certain  other food additives (FAO Nutrition Meetings Report
    Series No. 54A; WHO Food Additive Series No. 6).

37. FAO/WHO   (1975)  Specifications for the identity and purity
    of  some food colours, flavour enhancers, thickening agents,
    and  certain food additives  (FAO Nutrition Meetings  Report
    Series No. 54B; WHO Food Additives Series No. 7).

38. FAO/WHO   (1975)  Evaluation of certain food additives: some
    food  colours,  thickening  agents, smoke  condensates,  and
    certain  other  substances.  Nineteenth report  of the Joint
    FAO/WHO  Expert Committee on  Food Additives (FAO  Nutrition
    Meetings Report Series No. 55; WHO Technical  Report  Series
    No. 576).

39. FAO/WHO   (1975)   Toxicological  evaluation  of  some  food
    colours,  thickening  agents,  and certain  other substances
    (FAO  Nutrition  Meetings Report  Series  No. 55A;  WHO Food
    Additive Series No. 8).

40. FAO/WHO   (1976)  Specifications for the identity and purity
    of  certain  food  additives (FAO  Nutrition Meetings Report
    Series No. 55B; WHO Food Additive Series No. 9).

41. FAO/WHO   (1976)   Evaluation  of  certain  food  additives.
    Twentieth  report of the  Joint FAO/WHO Expert  Committee on
    Food  Additives (FAO Food  and Nutrition Series  No. 1;  WHO
    Technical Report Series No. 599).

42. FAO/WHO   (1976)   Toxicological evaluation  of certain food
    additives (WHO Food Additives Series No. 10).

43. FAO/WHO   (1977)  Specifications for the identity and purity
    of  some food additives (FAO  Food and Nutrition Series  No.
    1B; WHO Food Additive Series No. 11).

44. FAO/WHO   (1978)   Evaluation  of  certain  food  additives.
    Twenty-first report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 617).

45. FAO/WHO   (1977)  Summary of  toxicological data of  certain
    food additives (WHO Food Additives Series No. 12).

46. FAO/WHO   (1977)  Specifications for the identity and purity
    of   some  food  additives,  including   antioxidants,  food
    colours,  thickeners,  and  others (FAO  Nutrition  Meetings
    Report Series No. 57).

47. FAO/WHO   (1978)  Evaluation of  certain food additives  and
    contaminants.  Twenty-second  report  of the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 631).

48. FAO/WHO   (1978)  Summary of  toxicological data of  certain
    food  additives and contaminants  (WHO Food Additive  Series
    No. 13).

49. FAO/WHO   (1978)  Specifications for the identity and purity
    of  certain food additives (FAO Food and Nutrition Paper No.
    7).

50. FAO/WHO   (1980)   Evaluation  of  certain  food  additives.
    Twenty-third report of the Joint FAO/WHO Expert Committee on
    Food  Additives  (WHO Technical  Report  Series No.  648 and
    corrigenda).

51. FAO/WHO   (1980)   Toxicological evaluation  of certain food
    additives (WHO Food Additive Series No. 14).

52. FAO/WHO   (1979)  Specifications for the identity and purity
    of food colours, flavouring agents, and other food additives
    (FAO Food and Nutrition Paper No. 12).

53. FAO/WHO   (1980)   Evaluation  of  certain  food  additives.
    Twenty-fourth  report of the Joint  FAO/WHO Expert Committee
    on Food Additives (WHO Technical Report Series No. 653).

54. FAO/WHO   (1980)   Toxicological evaluation  of certain food
    additives (WHO Food Additives Series No. 15).

55. FAO/WHO   (1980)  Specifications for the identity and purity
    of  food  additives (sweetening  agents, emulsifying agents,
    and  other food additives) (FAO Food and Nutrition Paper No.
    17).

56. FAO/WHO   (1981)   Evaluation  of  certain  food  additives.
    Twenty-fifth report of the Joint FAO/WHO Expert Committee on
    Food Additives (WHO Technical Report Series No. 669).

57. FAO/WHO   (1981)   Toxicological evaluation  of certain food
    additives (WHO Food Additives Series No. 16).

58. FAO/WHO   (1981)  Specifications for the identity and purity
    of   food  additives  (carrier  solvents,   emulsifiers  and
    stabilizers,  enzyme  preparations, flavouring  agents, food
    colours,  sweetening agents, and  other food additives  (FAO
    Food and Nutrition Paper No. 19).

59. FAO/WHO   (1982)  Evaluation of  certain food additives  and
    contaminants.  Twenty-sixth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 683).

60. FAO/WHO   (1982)   Toxicological evaluation  of certain food
    additives (WHO Food Additives Series No. 17).

61. FAO/WHO   (1982)  Specifications for the identity and purity
    of  certain food additives (FAO Food and Nutrition Paper No.
    25).

62. FAO/WHO   (1983)  Evaluation of  certain food additives  and
    contaminants.  Twenty-seventh  report  of the  Joint FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 696 and corrigenda).

63. FAO/WHO   (1983)   Toxicological evaluation  of certain food
    additives  and contaminants (WHO  Food Additives Series  No.
    18).

64. FAO/WHO   (1983)  Specifications for the identity and purity
    of  certain food additives (FAO Food and Nutrition Paper No.
    28).

65. FAO/WHO   (1983)  Guide to specifications - General notices,
    general  methods, identification tests, test  solutions, and
    other  reference materials (FAO Food and Nutrition Paper No.
    5, Rev. 1).

66. FAO/WHO   (1984)  Evaluation of  certain food additives  and
    contaminants.  Twenty-eighth  report  of the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 710).

67. FAO/WHO   (1984)   Toxicological evaluation  of certain food
    additives  and contaminants (WHO  Food Additives Series  No.
    19).

68. FAO/WHO   (1984)  Specifications for the identity and purity
    of food colours (FAO Food and Nutrition Paper No. 31/1).

69. FAO/WHO   (1984)  Specifications for the identity and purity
    of food additives (FAO Food and Nutrition Paper No. 31/2).

70. FAO/WHO   (1986)  Evaluation of  certain food additives  and
    contaminants.  Twenty-ninth  report  of  the  Joint  FAO/WHO
    Expert  Committee  on  Food Additives  (WHO Technical Report
    Series No. 733).

71. FAO/WHO   (1986)   Toxicological evaluation  of certain food
    additives  and contaminants (WHO  Food Additives Series  No.
    20).

72. FAO/WHO   (1986)  Specifications for the identity and purity
    of  certain food additives (FAO Food and Nutrition Paper No.
    34).


    See Also:
       Toxicological Abbreviations