PRINCIPLES OF RISK ASSESSM!
A Nontechnical Review
3-EPA
WORKSHOP ON RISK ASSESSMENT
United States Environmental Protection A^sncy
U.S. Environmental Protection Agency
Library, Room 2404 PM-211-A
401 M Street, S.W.
Washington, DC 20460
L.l5™;16"'1?1 \Wlto an understanding of
concepts and methods used to perform risk
assessments of environmental chemicals/"t £s md
as background Information at the Regional R14 Assessment
SKIS" 1n Atlanta' ** 29'30- 1«5- Hoti tfit"he
mateHal was prepared by a contractor (Envlroh
Corporation) and It doe, not necessSri y r^t EPA
-------
PRINCIPLES OF RISK ASSESSMENT
A Nontechnical Review
S-EPA
WORKSHOP ON RISK ASSESSMENT
United States Environmental Protection Agency
U.S. Environraer-tal Protection Agency
Library, Room 2404 PM-211-A
401 M Street, S.W.
Washington, DC 20460
-------
The materials presented here have been reviewed by personnel from the
United States Environmental Protection Agency. They do not, however, necessarily
reflect United States Environmental Protection Agency policy. The materials
were prepared primarily by:
ENVIRON CORPORATION
Washington. D.C.
-------
CONTENTS
Page
I. INTRODUCTION I-i
II. RISK AND RISK ASSESSMENT 11-1
Basic Concepts and Definitions 11-3
The Components of Risk Assessment 11-3
Dose 11-4
III. HAZARD IDENTIFICATION 11 1-1
Introduction 1 11—1
Toxicity Information from Animal Studies 11 1-1
The Use of Animal Toxicity Data 11 1-1
General Nature of Animal Toxicity Studies 111-2
Manifestations of Toxicity 111-4
Design and Conduct of Toxicity Tests 111-6
Design of Tests for Carcinogenicity 111-8
Conduct and Interpretation of Toxicity Tests 111-10
Categorization of Toxic Effects 1 1 1-11
Uncertainties in Evaluation of Animal
Carciriogenicity Test Results 111-12
Short—Term Tests for Carcinogens 111-13
Data from Human Studies 111-13
Hazard Identification: A Summary 111-16
IV. DOSE-RESPONSE EVALUATION IV-1
Introduction IV-1
Threshold Effects 1V1
Effects that May Not Exhibit Thresholds IV-3
The Carcinogenic Process 1V3
Potency and High-to-Low Dose Extrapolation IV-4
Interspecies Extrapolation IV—7
Dose—Response Evaluation: Summary IV-7
-------
2
CONTENTS (con’t)
Page
V. HUMAN EXPOSURE EVALUATION V-i
VI. RISK CHARACTERIZATION VI-i
APPENDIX: Toxic Effects on Organs or Other Target Systems
Introduction A—i
Liver A-i
Kidney A-2
Reproductive System A-3
Lungs A-5
Skin
Central Nervous System A-6
Blood A-8
Immune System A-9
Genetic Toxicology A-9
-------
I. INTRODUCTION
This report provides general background information for
understanding the types of scientific data and methods currently
used to assess the human health risks of environmental chemicals.
Human health risk is the likelihood (or probability) that a given
chemical exposure or series of exposures may damage the health of
exposed individuals. Chemical risk assessment involves the anal-
ysis of exposures that have taken place in the past, the adverse
health effects of which may or may not have already occurred. It
also involves prediction of the likely consequences of exposures
that have not yet occurred. This document is by no means a com-
plete survey of the complex subject of risk assessment, but it is
sufficiently comprehensive to assist conference participants in
dealing with the specific sets or data relevant to the case
study.
The report begins with a discussion of the four major compon-
ents of risk assessment and their interrelationships. This sec-
tion is followed by extensive discussion of these four major com-
ponents. Generally, each section focuses on the methods and
tests used to gather data, the principles used for data interpre-
tatiori, and the uncertainties in both the data and inferences
drawn from them. Throughout these discussions, key concepts
(e.g., exposure, dose, thresholds, and extrapolation) are defined
and extended descriptions provided.
Many of the principles discussed in this report are widely
accepted in the scientific community. Others (e.g., thresholds
for carcinogens, the utility of negative epidemiology data) are
controversial. In such cases we have attempted to describe the
various points of view and the reasons for them and have also
identified the viewpoint that seems to have been broadly adopted
by public health and regulatory officials.
Finally, the concepts and principles we describe here, al-
though broadly applicable, may not apply in specific cases. In
some instances, the data available on a specific chemical may
reveal aspects of its behavior in biological systems that suggest
a general principle (e.g., that data obtained in rodent studies
are generally applicable to humans) may not hold. In such in-
stances, the usual approach is to modify the risk assessment
process to conform to the scientific finding.
-------
II. RISK AND RISK ASSESSMENT
BASIC CONCEPTS AND DEFINITIONS
Risk is the probability of injury, disease, or death under
specific circumstances. It may be expressed in quantitative
terms, taking values from zero (certainty that harm will not
occur) to one (certainty that it will). In many cases risk can
only be described qualitatively, as “high, ulow,N Rtrivial.
All human activities carry some degree of risk. Many risks
are known with a relatively high degree of accuracy, because we
have collected data on their historical occurrence. Table 1
lists the risks of some common activities.
Table 1
ANNUAL RISK OF DEATH FROM SELECTED COMMON HUMAN ACTIVITIES 1
Nt er of Deaths
in Representative Lifetime
Year Individual Risk/Year Risk 2
Coal Mining
Accident 180 1.30 x iO or 1/770 1/17
Black lung disease 1,135 8 x iO or 1/125 1/3
Motor Vehicle 46,000 2.2 x 10 or 1/4,500 1/65
Truck Driving 400 1O or 1/10,000 1/222
Falls 16,339 7.7 x 10 or 1/13,000 1/186
Home Accidents 25,000 1.2 x iO or 1/83,000 1/130
1 Selected from Hutt (1978) Food, Drug, Cosmetic Law J . 33:558—589.
2 (stimated based upon 70—year lifetime and 45—year rk exposure.
The risks associated with many other activities, including
the exposure to various chemical substances, can not be readily
assessed and quantified. Although there are considerable histor-
ical data on the risks of some types of chemical exposures (e.g.,
the annual risk of death from intentional overdoses or accidental
exposures to drugs, pesticides, and industrial chemicals), such
data are generally restricted to those situations in which a
single, very high exposure resulted in an immediately observable
form of injury, thus leaving little doubt about causation.
Assessment of the risks of levels of chemical exposure that do
-------
11—2
not cause immediately observable forms of injury or disease (or
only minor forms such as transient eye or skin irritation) is far
more complex, irrespective of whether the exposure may have been
brief, extended but intermittent, or extended and continuous. It
is the latter type of risk assessment activity that is reviewed
in this report (although some review of acute poisoning is also
included).
As recently defined by the National Academy of Sciences, risk
assessment is the scientific activity of evaluating the toxic
properties of a chemical and the conditions of human exposure to
it in order both to ascertain the likelihood that exposed humans
will be adversely affected, and to characterize the nature of the
effects they may experience)
The Academy distinguishes risk assessment from risk manage-
ment; the latter activity concerns decisions about whether an
assessed risk is sufficiently high to present a public health
concern and about the appropriate means for control of a risk
judged to be significant.
The term “safe,” in its common usage, means “without risk.”
In technical terms, however, this common usage is misleading
because science can not ascertain the conditions under which a
given chemical exposure is likely to be absolutely without a risk
of any type. The latter condition——zero risk——is simply immea-
surable. Science can, however, describe the conditions under
which risks are so low that they would generally be considered to
be of no practical consequence to persons in a population. As a
technical matter, the safety of chemical substances——whether in
food, drinking water, air, or the workplace——has always been
defined as a condition of exposure under which there is a “prac-
tical certainty” that no harm will result in exposed individuals.
(As described later, these conditions usually incorporate large
safety factors, so that even more intense exposures than those
defined as safe may also carry extremely low risks). We note
that most “safe” exposure levels established in the way we have
described are probably risk-free, but science simply has no tools
to prove the existence of what is essentially a negative condi-
tion.
Another preliminary concept concerns classification of chemi-
cal substances as either “safe” or unsafe” (or as “toxic” and
“nontoxic”). This type of classification, while common (even
among scientists who should know better), is highly problematic
1 Risk Assessment in the Federal Government: Managing the Process
(Washington, D.C.: National Academy Press, 1983).
-------
11—3
and misleading. All substances, even those which we consume in
high amounts every day, can be made to produce a toxic response
under some conditions of exposure. In this sense, all substances
are toxic. The important question is not simply that of toxici-
ty, but rather that of risk——i.e., what is the probability that
the toxic properties of a chemical will be realized under actual
or anticipated conditions of human exposure? To answer the lat-
ter question requires far more extensive data and evaluation than
the characterization of toxicity. 2
THE COMPONENTS OF RISK ASSESSMENT
There are four components to every (complete) risk assess-
ment:
A. Hazard Identification——Involves gathering and evaluating
data on the types of health injury or disease that may
be produced by a chemical and on the conditions of expo-
sure under which injury or disease is produced. It may
also involve characterization of the behavior of a chem-
ical within the body and the interactions it undergoes
with organs, cells, or even parts of cells. Data of the
latter types may be of value in answering the ultimate
question of whether the forms of toxicity known to be
produced by a substance in one population group or in
experimental settings are also likely to be produced in
humans. Hazard identification is not risk assessment;
we are simply determining whether it is scientifically
correct to infer that toxic effects observed in one
setting will occur in other settings (e.g., are sub-
stances found to be carcinogenic or teratogenic in ex-
perimental animals likely to have the same result in
humans?).
B. Dose—Response Evaluation--Involves describing the quan-
titative relationship between the amount of exposure to
a substance and the extent of toxic injury or disease.
Data derive from animal studies or, less frequently,
from studies in exposed human populations. There may be
many different dose—response relationships for a sub-
stance if it produces different toxic effects under
2 Some scientists will claim that carcinogens display their toxic
properties under all conditions of exposure, and that there is
no “safe” level of exposure to such agents. This special prob-
lem receives extensive treatment in later sections.
-------
11—4
different conditions of exposure. The risks of a sub-
stance can not be ascertained with any degree of conf i—
dence unless dose—response relations are quantified,
even if the substance is known to be toxic.”
C. Human Exposure Evaluation——Involves describing the
nature and size of the population exposed to a substance
and the magnitude and duration of their exposure. The
evaluation could concern past or current exposures, or
exposures anticipated in the future.
D. Risk Characterization——Generally involves the integra-
tion of the data and analysis of the first three compo-
nents to determine the likelihood that humans will
experience any of the various forms of toxicity associ-
ated with a substance. (In cases where exposure data
are not available, hypothetical risk can be character-
ized by the integration of hazard identification and
does—response evaluation data alone.)
The next four sections elaborate on each of these components
of risk assessment. However, the concept of “dose, which under-
lies all the discussions to follow of both experimental animals
and human populations, is reviewed first.
DOSE
Human exposures to substances in the environment may occur
because of their presence in air, water, or food. Other circum-
stances may provide the opportunity for exposure, such as direct
contact with a sample of the substance or contact with contami-
nated soil. Experiments for studying the toxicity of a substance
usually involve intentional administration to subjects through
the diet, air to be inhaled, or direct application to skin.
Experimental studies may include other routes of administration:
injection under the skin (subcutaneous), into the blood (usually
intravenoue), or into body cavities (intraperitoneal).
In both human and animal exposures, two types of measurement
must be distinguished:
1. Measurement of the amount of the substance in the
medium (air, diet, etc.) in which it is present or
administered.
2. Measurement of the amount received by the subject,
whether human or animal.
-------
11—5
It is critically important to distinguish these two types of
measures. The second measure, which is usually expressed as a
dose, is the critical factor in assessing risk. The first mea-
sure, along with other information, usually is essential if the
dose is to be established. It may be substituted or supple-
mented, however, in cases where environmental modeling or biomon—
itoring data are available.
The difference between these two measures is best described
by example. Suppose a substance is present in drinking water to
be consumed by an individual. To determine the individual’s dose
of this substance, it is first necessary to know the amount
present in a given volume of water. For many environmental sub-
stances, the amounts present fall in the milligram (mg, one—
thousandth of a gram = 1/28571 ounce) or microgram (ag, one-
millionth of a gram = 1/28,571,429 ounce) range. The analyst
will usually report the number of mg or ug of the substance
present in one liter of water, i.e., mg/i or g/l. These two
units are sometimes expressed as parts per million (ppm) or parts
per billion (ppb), respectively. 3
Given the concentration of a substance in water (say in ppm),
it is possible to estimate the amount an individual will consume
by knowing the amount of water he drinks. Time is another im-
portant factor in determining risk, so the amount of water con-
sumed per unit time is of interest. In most public health evalu-
ations, it is assumed that art individual consumes 2 liters of
water each day through all uses. Thus, if a substance is present
at 10 ppm in water, the average daily individual intake of the
substance is obtained as follows:
10 mg/liter x 2 liter/day = 20 mg/day
For toxicity comparisons among different species, it is necessary
to take into account size differences, usually by dividing daily
intake by the weight of the individual. Thus, for a man of aver-
age weight (usually assumed to be 70 kilograms (kg) or 154
pounds), the daily dose of our hypothetical substance is:
20 mg/day , 70 kg = 0.29 mg/kg/day
3 A liter of water weighs 1,000 g. One mg is thus one—millionth
the weight of a liter of water; and one g is one—billionth the
weight of a liter.
-------
11—6
For a person of lower weight (e.g., a female or child), the daily
dose at the same intake rate would be larger. For example, a 50
kg woman ingesting the hypothetical substance would receive a
dose of:
20 mg/day r 50 kg = 0.40 mg/kg/day
A child of 10 kg could receive a dose of 2.0 mg/kg/day, although
it must be remembered that such a child would drink less water
each day (say, 1 liter), so that the child’s dose would be:
10 mg/liter x 1 liter/day ? 10 kg = 1.0 mg/kg/day
Also, laboratory animals, usually rats or mice, receive a much
higher dose than humans at the same daily intake rate because of
their much smaller body weights (of course, rats and mice do not
drink 2 liters of water each day!).
These sample calculations point out the difference between
measurement of environmental concentrations and dose, at least
for drinking water. The relationships between measured environ-
mental concentrations and dose are nore complex for air and other
media. Table 2 lists the data necessary to obtain dose from data
on the concentration of a substance in water. Each medium of
exposure must be treated separately and some calculations are
more complex than in the dose per liter of water example.
-------
11—7
Table 2
TA AND ASSUMPTIONS NECESSARY TO ESTIMATE
HIJIAN WSE OF A WATER CONTAMINANT FROM IG4OWLEOGE OF ITS CONCENTRATION
Total Dae is Equal to the Sum of D aes fro. Five Routee
1. Direct Ingestion Through Drinking
mount of water consumed each day (generally aesused to be 2 liters for
adalta and 1 liter for 10 kg child).
Fraction of contaminant absorbed through wall of gastrointestinal tract.
Average human body weight.
2. Inhalation of Contaminants
Air concentrations resulting from showering, bathing, and other uses of
water.
Variation in air concentration over time.
ANount of contaminated air breathed during those activities that may lead
to volatilization.
Fraction of inhaled contaminant absorbed through lungs.
Average human body weight.
3. Skin Absorption from ter
Period of time spent washing and bathing.
Fraction of contaminant absorbed through the akin during washing and
bathing.
Average human body weight.
4. Ingestion of Contaminated Food
Concentrationa of contaminant in edible portions of various plants and
animals exposed to contaminated groun ater.
Abount of contaminated food ingested each day.
Fraction of contaminant absorbed through wall of gastrointestinal tract.
Average human body weight.
5. 9cm Absorption for Contaminated Soil
Concentrations of contaminant in soil exposed to contaminated
groundaster.
- ount of daily skin contact with soil.
Puount of aoil ingested per day (by children).
Absorption rates.
Average human body weight.
-------
11—8
It is important always to consider that a human may be
simultaneously exposed to the same substance through several
media. That is, a dose may be received through more than one
route of exposure (inhalation, ingestion, derrnal contact). The
“total dose” received by an individual is the sum of doses re-
ceived by each individual route (see the example in Table 2).
In some cases, it may not be appropriate to add doses in
this fashion. In these cases, the toxic effects of a substance
may depend on the route of exposure. For example, inhaled chrom-
iwn is carcinogenic to the lung, but it appears that ingested
chromium is not. In most cases, however, as long as a substance
acts at an internal body site (i.e., acts systemically rather
than only at the point of initial contact), it is usually con-
sidered appropriate to add doses received from several routes.
Two additional factors concerning dose require special atten-
tion. The first is the concept of ab8orption (or absorbed dose) .
The second concerns inferences to be drawn from toxicities ob-
served under one route of exposure for purposes of predicting the
likelihood of toxicity under other routes.
Absorption
When a substance is ingested in the diet or in drinking
water, it enters the gastrointestinal tract. When it is present
in air (as a gas, aerosol, particle, dust, fume, etc.) it enters
the upper airways and lungs. A substance may also come into
contact with the skin and other body surfaces as a liquid or
solid. Some substances may cause toxic injury at the point of
initial contact (skin, gastrointestinal tract, upper airways,
lungs, eyes). Indeed, at high concentrations, n st substances
will cause at least irritation at these points of contact. But
for many substances, toxicity occurs after they pass through
certain barriers (e.g., the wall of the gastrointestinal tract or
the skin itself), enter blood or lymph, and gain access to the
various organs or systems of the body. Figure 1 is a diagram of
some of the important routes of absorption. This figure also
shows that chemicals may be distributed in the body in various
ways and then excreted . (However, some chemical types--usually
substances with high solubility in fat, such as DOT——are stored
for long periods of time, usually in fat.)
-------
11—9
Figure 1
KEY ROUTES OF CHEMICAL ABSORPTION, DISTRIBUTION, AND EXCRETION
Some chemicals under chemical change (metabolism) within the cells of the body before excretion.
Toxicity may be produced by the chemical as introduced, or by one or more metabolites.
z
0
I-
I
I-
U)
I I
I I
I I
I... I
EXCRETION
-------
11—10
Substances vary widely in extent of absorption. The frac-
tion of a dose that passes through the wall of the gastrointes-
tinal tract may be very small (e.g., 1 to 10% for some metals) to
substantial (close to 100% for certain types of organic mole-
cules). Absorption rates also depend upon the medium in which a
chemical is present (e.g., a substance present in water might be
absorbed differently from the same substance present in a fatty
diet). These rates also vary among animal species and among
individuals within a species.
Ideally, estimating systemic dose should include considera-
tion of absorption rates. Unfortunately, data on absorption are
limited for most substances, especially in humans. As a result,
absorption is not always included in dose estimation (i.e., by
default, it is frequently considered to be complete). Sometimes
crude adjustments are made based on some general principles con-
cerning expected rates based on the molecular characteristics of
a substance.
Iriterspecies Differences in Exposure Route
As described later, a critical feature of risk characteriza—
tiori is a comparison of doses that are toxic in experimental
animals and the doses received by exposed humans. If humans are
exposed by the same route as the experimental animals, it is
frequently assumed (in the absence of data) that the extent of
absorption in animals and humans is approximately the same; under
such an assumption, it is unnecessary to estimate the absorbed
dose by taking absorption rate into account. Rowever, humans are
often exposed by a different route than that used to obtain tox-
icity data in experimental animals. If the observed toxic effect
is a systemic one, it may be appropriate to infer the possibility
of human toxicity even under the different exposure route. Be-
fore doing so, however, it is critical to consider the relative
degrees of absorption by different exposure routes. For example,
if a substance is administered orally to a test animal but human
exposure is usually by inhalation, knowledge of tne percentage
absorbed orally by the animal and by inhalation in humans is
necessary to properly compare human and animal doses. These
calculations and underlying assumptions are too complex for dis-
cussion here, but they should be kept in mind when risks are
being described.
In the following discussion of the components of risk assess-
ment, we shall use the term dose only as described. Many risk
assessors use the terms exposure and dose synonomously. In this
document, however, the term exposure describes contact with a
-------
h—h
substance (e.g., we say that animals are exposed to air contain-
ing 10 mg/rn 3 , of a compound), as well as the size of the dose,
the duration of exposure, and tne nature and size of the exposed
population. In our usage, exposure is a broader term than dose.
Although our usages of those terms are technically correct, it
should be recognized that some assessors use the term exposure to
mean dose (although the reverse is not true).
-------
III. EAZARD IDENTIFICATION
INTRODUCTION
Information on the toxic properties of chemical substances is
obtained from animal studies, controlled epidemiological investi-
gations of exposed human populations, and clinical studies or
case reports of exposed humans. Other information bearing on
toxicity derives from experimental studies in systems other than
whole animals (e.g., in isolated organs, cells, subcellular com-
ponents) and from analysis of the molecular structures of the
substances of interest. These last two sources of information
are generally considered less certain indicators of toxic poten-
tial, and accordingly, they receive limited treatment here.
Similarly, clinical studies or case reports, while sometimes
very important (e.g., the earliest signs that benzene was a human
leukeinogen came from a series of case reports), seldom provide
the central body of information for risk assessment. For this
reason, and because they usually present no unusual problems of
interpretation, they are not further reviewed here. Rather, our
attention is devoted to the two principal sources of toxicity
data: animal tests and epidemiology studies. These two types of
investigation are not only principal sources of data, but also
present interpretative difficulties, some rather subtle, some
highly controversial.
TOXICITY INFORMATION FROM ANIMAL STUDIES
The Use of Animal Toxicity Data
Animal toxicity studies are conducted based primarily on the
longstanding assumption that effects in humans can be inferred
from effects in animals. In fact, this assumption has been shown
to be generally correct. Thus, all the chemicals that have been
demonstrated to be carcinogenic in humans, with the possible
exception of arsenic, are carcinogenic in some although not all,
experimental animal species. In addition, the acutely toxic
doses of many chemicals are similar in humans and a variety of
experimental animals. This principle of extrapolation of animal
data to humans has been widely accepted in the scientific and
regulatory communities. The foundation of our ability to infer
effects in humans from effects in animals has been attributed to
the evolutionary relationships and the phylogenetic continuity of
animal species including man. Thus, at least among mammals, the
basic anatomical, physiological, and biochemical parameters are
similar across species.
-------
111—2
However, although the general principle of inferring efeects in
humans from effects in experimental animals is well founded,
there have been a number of exceptions. Many of these exceptions
relate to differences in the way various species handle a chemi-
cal to which they are exposed and to differences in metabolism,
distribution and pharmacokinetics of the chemical. Because of
these potential differences, it is essential to evaluate all
interspecies differences carefully in inferring human toxicity
from animal toxicologic study results.
In the particular case of evaluation of long-term animal
studies conducted primarily to assess the carcinogenic potential
of a compound, certain general observations increase the overall
strength of the evidence that the compound is carcinogenic. With
an increase in the number of tissue sites affected by the agent,
there is an increase in the strength of the evidence. Similarly,
an increase in the number of animal species, strains, and sexes
showing a carcinogenic response will increase the strength of the
evidence of carcinogenicity. Other aspects of importance are the
occurrence of clear-cut dose—response relationships in the data
evaluated; the achievement of a high level of statistical signif-
icance of the increase of tumor incidence in treated versus con-
trol animals; dose—related shortening of the time-to-tumor occur-
rence or time—to-death with tumor; and a dose—related increase in
the proportion of tumors that are malignant. The following sec-
tions describe the general nature of animal toxicity studies,
including major areas of importance in their design, conduct, and
interpretation. Particular consideration will be given to the
uncertainties involved in the evaluation of their results.
General Nature of Animal Toxicity Studies
Toxicity studies are conducted to identify the nature of
health damage produced by a substance 4 and the range of doses
over which damage is produced. The usual starting point for such
investigations is a study of the acute (single—dose) toxicity of
a chemical in experimental animals. Acute toxicity studies are
necessary to calculate doses that will not be lethal to animals
used in toxicity studies of longer durations. Moreover, such
4 We use the term substance to refer to a pure chemical, to a
chemical containing impurities, or to a mixture of chemicals.
It is clearly important to know the identity and composition of
a tested substance before drawing inferences about the toxicity
of other samples of the same substance that might have a some-
what different composition.
-------
111—3
studies will give one estimate of the compound’s comparative
toxicity and may indicate the target organ system for chronic
toxicity (e.g., kidney, lung, or heart). Toxicologists examine
the lethal properties of a substance and estimate its LD 50
(lethal dose, on average, for 50% of an exposed population). In
a group of chemicals, those exhibiting lower LD 50 s are more
acutely toxic than those with higher values. group of well-
known substances and their LD 50 values are listed in Table 3.
Table 3
APPROXIMATE ORAL L0 50 a IN RATS F R A
OROUP OF WELL—KNOWN CHEMICALS
Chemical LDç (mq/kg )
Sucrose (table sugar) 29,700
Ethyl alcohol 14,000
Sodium chloride (common salt) 3,000
Vitamin A 2,000
Vanillin 1,580
Aspirin 1,000
Chloroform 800
Copper sulfate 300
Caffeine 192
Phenobarbital, sodium salt 162
DDT 113
Sodium nitrite 85
Nicotine 53
.Aflatoxin 81 7
Sodium cyanide 6.4
Strychnine 2.5
1 Selected from NIOSH, Registry of Toxic Effects of Diemical
Substances , 1979. Results reported elsewhere may differ.
2 Cospounda are listed in order of increasing toxicity——i.e.,
sucrose is the least toxic and strychnine is the most toxic.
-------
111-4
LD 50 studies reveal one of the basic principles of toxi-
cology: not all individuals exposed to the same dose of a sub-
stance will respond in the same way. Thus, at a dose of a sub-
stance that leads to the death of some experimental animals,
other animals dosed in the same way will get sick but will re-
cover, and still others will not appear to be affected at all.
We shall return to this point after a fuller discussion of other
forms of toxicity.
Each of the many different types of toxicology studies has a
different purpose. Animals may be exposed repeatedly or contin-
uously for several weeks or months (subchronic toxicity studies)
or for close to their full lifetimes (chronic toxicity studies)
to learn how the period of exposure affects toxic response. In
general, the reasons to conduct toxicity studies can be summar-
ized as follows:
• Identify the specific organs or systems of the body
that may be damaged by a substance.
• Identify specific abnormalities or diseases that a
substance may produce, such as cancer, birth defects,
nervous disorders, or behavioral problems.
• Establish the conditions of exposure and dose that give
rise to specific forms of damage or disease.
• Identify the specific nature and course of the injury
or disease produced by a substance.
• Identify the biological processes that underlie the
production of observable damage or disease.
The laboratory methods needed to accomplish many of these
goals have been in use for many years, although some methods are
still being developed. Before describing some of the tests, it
is useful to say more about the various manifestations of toxi-
city.
Manifestations of Toxicity
Toxic responses, regardless of the organ or system in which
they occur, can be of several types. For some, the severity of
the injury increases as the dose increases. Thus, for example,
some chemicals affect the liver. At high doses they may kill
liver cells, perhaps so many as to destroy the liver and thus
cause the deaths of some or all experimental subjects. As the
dose is lowered, fewer cells may be killed, but they may exhibit
other forms of damage, causing imperfections in their function-
ing. At lower doses still, no cell deaths may occur and there
-------
111—5
may be only slight alterations in cell function or structure.
inally, a dose may be achieved at which no effect is observed,
r at which there are only biochemical alterations that have no
nown adverse effects on the health of the animal (although some
oxicologists consider any such alteration, even if its long—term
onsequences are unknown, to be “adverse,” there is no clear
onsensus on this issue.) One of the goals of toxicity studies
;s to determine the “no observed effect level” (NOEL), which is
he dose at which no effect is seen; the role of the NOEL in risk
issessment is discussed later.
In other cases, the severity of an effect may not increase
iith dose, but the incidence of the effect will increase with
increasing dose. In such cases, the number of animals experienc-
ing an adverse effect at a given dose is less than the total
number, and, as the dose increases, the fraction experiencing
adverse effects (i.e., the incidence of disease or injury) in-
creases; at sufficiently high dose, all experimental subjects
will experience the effect. The latter responses are properly
characterized as probabilistic. Increasing the dose increases
the probability (i.e., risk) that the abnormality will develop in
an exposed population. Often with toxic effects, including can-
cer, both the severity and the incidence increase as the level of
exposure is raised. The increase in severity is a result of
increased damage at higher doses, while the increase in incidence
is a result of differences in individual sensitivity. In addi-
tion, the site at which a substance acts (e.g., liver, kidney)
may change as the dose changes.
Generally, as the duration of exposure increases, both the
NOEL and the doses at which effects appear decrease; in some
cases, new effects not apparent upon exposures of short duration
become manifest.
Toxic responses also vary in degree of reversibility. In
some cases, an effect will disappear almost immediately following
cessation of exposure. At the other extreme, some exposures will
result in a permanent injury——for example, a severe birth defect
resulting from a substance that irreversibly damages a fetus at a
critical r ment of its development. Most toxic responses fall
somewhere between these extremes. In many experiments, however,
the degree of reversibility cannot be ascertained by the investi-
gator.
Seriousness is another characteristic of a toxic response.
Certain types of toxic damage are clearly adverse and are a def-
inite threat to health. However, other types of effects observed
during toxicity studies are not clearly of health significance.
For example, at a given dose a chemical may produce a slight
-------
111—6
increase in red blood cell count. If no other effects are ob-
served at this dose, it will not be at all clear that a true
adverse response has occurred. Determining whether such slight
changes are significant to health is one of the critical issues
in assessing safety that has not been fully clarified.
Design and Conduct of Toxicity Tests
Toxicity experiments vary widely in design and conduct.
Although there are relatively well standardized tests for various
types of toxicity (e.g., National Cancer Institute carcinogen-
icity bioassays) developed by regulatory and public agencies in
connection with the premarket testing requirements imposed on
certain classes of chemicals, large numbers of other tests and
research—oriented investigations are conducted using specialized
study designs (e.g., carcinogenicity assays in fish). In this
section, we present a few of the critical considerations that
enter into the design of toxicity experiments. However, there
are numerous variations on the general themes we describe.
Selection of Animal Species
Rodents, usually rats or mice, are the most commonly used
laboratory animals for toxicity testing. Other rodents (e.g ,
hamsters and guinea pigs) are sometimes used, and many experi-
ments are conducted using rabbits, dogs, and such nonhuman pri-
mates as monkeys or baboons. For example, although nonhuman
primates may be chosen for some reproductive studies because
their reproductive systems are similar to that of humans, rabbits
are often used for testing dermal toxicity because their shaved
skin is more sensitive.
Rats and mice are the most common choice because they are
inexpensive and can be handled relatively easily. Furthert re,
such factors as genetic background and disease susceptibility are
well established for these species. The full lifespans of these
smaller rodents are complete in two to three years, so that the
effects of lifetime exposure to a substance can be measured rela-
tively quickly (as compared to the much longer—lived dog or
monkey).
Dose and Duration
An LDSQ using high doses of the substance is frequently the
first toxicity experiment performed. After completing these
experiments, investigators study the effects of lower doses
-------
111—7
administered over longer periods. The purpose is to find the
range of doses over which adverse effects occur and to identify
the NOEL for these effects (although the latter is not always
sought or achieved). A toxicity experiment is of limited value
unless a dose of sufficient magnitude to cause some type of
adverse effect within the duration of the experiment is achieved.
If no effects are seen at all doses administered, the toxic
properties of the substances will not have been characterized,
and the investigator will usually repeat the experiment at higher
doses or will extend its duration. 5
Studies are frequently characterized according to the dura-
tion of exposure. Acute toxicity studies involve a single dose,
or exposures of very short duration (e.g., 8 hours of inhala-
tion). Chronic studies involve exposures for near the full hf e—
times of the experimental animals. Experiments of varying dura-
tion between these extremes are referred to as subchronic stud-
ies.
Number of Dose Levels
Although it is desirable that many different dose levels be
used to develop a well characterized dose—response relationship,
practical considerations usually limit the number to two or
three, especially in chronic studies. Experiments involving a
single dose are frequently reported and leave great uncertainty
about the full range of doses over which effects are expected.
Controls
No toxicity experiment is interpretable if control animals
are omitted. Control animals must be of the same species,
strain, sex, age, and state of health as the treated animals, and
must be held under identical conditions throughout the experi-
ment. (Indeed, allocation of animals to control and treatment
groups should be performed on a completely random basis.) Of
course, the control animals are not exposed to the substance
under teste
5 Some substances with extremely low toxicity must be administered
at extremely high levels to produce effects; in many cases, such
high levels will cause dietary maladjustments leading to an
adverse nutritional effect that confounds interpretation. As a
practical matter, the highest level of a compound fed to an
animal in toxicity studies is 5% of the diet, even if rio toxic
effect is seen at this level.
-------
111—8
Route of Exposure
Animals are usually exposed by a route that is as close as
possible to that through which humans will be exposed, because
the purpose of most such tests is to produce the data upon which
human safety decisions will be based. In some cases, however,
the investigator may have to use other routes or conditions of
dosing to achieve the desired experimental dose. For example,
some chemicals are administered by stomach tube (gavage) because
they are too volatile or unpalatable to be placed in the animals’
diets at the high levels needed for toxicity studies.
Specialized Designs
Generally, the toxicologist exposes test animals and simply
records whatever effects happen to occur under the conditions of
the experiment. If, however, it is decided that certain highly
specific hypotheses about a substance are to be tested (e.g.,
does the substance cause birth defects or does it affect the
immune system?), certain specialized designs must be used. Thus,
for example, the hypothesis that a chemical is teratogenic
(causes birth defects) can be tested only if pregnant females are
exposed at certain critical times during pregnancy.
One of the most complex of the specialized tests is the
carcinogenesis bioassay . These experiments are used to test the
hypothesis of carcinogenicity——that is, the capacity of a sub-
stance to produce tumors. Because of the importance of the car-
cinogenesis bioassay, a full section is devoted to it. We shall
then discuss, in turn, controversial issues in the design of
animal tests and interpretation of test results.
Design of Tests for Carcinogenicity
In a National Cancer Institute (NCI) carcinogenicity bioas-
say, the test substance is administered over most of the adult
life of the animal, and the animal is observed for formation of
tumors. The general principles of test design previously dis-
cussed apply to carcinogenicity testing, but one critical design
issue that has been highly controversial requires extensive dis-
cussion. The issue is the concept of maximum tolerated dose
(tTPD), which is defined as the maximum dose that an animal spe-
cies can tolerate for a major portion of its lifetime without
significant impairment of growth or observable toxic effect other
than carcinogenicity. Cancer can take most of a lifetime to
develop, and it is thus widely agreed that studies should be
designed so that the animals survive in relatively good health
for a normal lifetime. It is not so widely agreed, however, that
-------
111—9
the MTD, as currently used, is the best way to achieve this
objective. The MTD and half the MTD are the usual doses used in
the NCI carcinogenicity bioassay.
The main reason cited for using the MTD as the highest dose
in the bioassay is that experimental studies are conducted on a
small scale, making them “statistically insensitive,” and that.
very high doses overcome this problem. For practical reasons,
experimental studies are carried out with relatively small groups
of animals. Typically, 50 or 60 animals of each species and sex
will be used at each dose level, including the control group. t
the end of such an experiment, the incidence of cancer as a func-
tion of dose (including control animal incidence) is tabulated by
the examining pathologists. Statisticians then analyze the data
to determine whether any observed differences in tumor incidence
(fraction of animals having a tumor of a certain type) are due to
random variations in tumor incidence or to treatment with the
substance.
In an experiment of about this size, assuming none of the
control animals develop tumors, the lowest incidence of cancer
that is detectable with statistical reliability is in the range
of 5%, or 3 animals with tumors in a test group of 60 animals.
If control animals develop tumors (as they frequently do), the
lowest range of cancer incidence detectability is even higher.
cancer incidence of 5% is very high, yet ordinary experimental
studies are not capable of detecting lower rates and most are
even less sensitive.
MTD advocates argue that inclusion of high doses will com-
pensate for the weak detection power of these experiments. By
using the MTD, the toxicologist hopes to elicit any important
toxic effects of a substance and ensure that even weak carcin-
ogenic effects of the chemical will be detected by the study.
MTD critics do not reject the notion that animal experiments may
be statistically insensitive, but rather are concerned about the
biological, implications of such high doses.
Concerns about use of MTDs can be summarized: (1) the
underlying biological mechanisms that lead to the production of
cancer maychange as the dose of the carcinogen changes; (2) cur-
rent methods for estimating an MTD for use in an experiment do
not usually take these mechanisms into account; (3) the biologi-
cal mechanisms at work under conditions of actual human exposure
may be quite different from those at work at or near the MTO; and
(4) therefore, observations at or near an MTD (as determined by
current methods) may not be qualitatively relevant to conditLons
of actual human exposure.
-------
111—10
Many agree that greater attention should be paid to develop-
ing data on the underlying mechanisms of carcinogenicity and
their relation to dose. Also, a range of doses should be includ-
ed in carcinogenicity testing to assess whether physiological
mechanisms that would normally detoxify the chemical are over-
whelmed at an MTD. These biological considerations have consid-
erable merit, but they are frequently disregarded in designing
studies and interpreting data. Although there are occasional
attempts to develop a more biologically relevant definition of
MTD, most current tests (e.g., those carried out by the National
Toxicology Program) use a definition of MTD that does not take
biological mechanisms into account.
This state of affairs is not likely to change. Those who
promote the use of MTD, as currently defined, frequently argue
that if the highest dose used was not the MTD, failure to observe
a carcinogenic effect in a given experiment does not permit the
conclusion that the tested substance is not carcinogenic. A
similar argument is made if the survival of the test animals did
not approximate their full lifetimes.
Conduct and Interpretation of Toxicity Tests
Many factors must be considered in the conduct of toxicity
tests to ensure their success and the utility of their results.
In evaluating the results of such tests, certain questions must
be asked about the design and conduct of a test to ensure criti-
cal appraisal. The major questions include the following:
1. Was the experimental design adequate to test the hypo-
thesis under examination?
2. Was the general conduct of the test in compliance with
standards of good laboratory practice?
3. Was the dose of test compound correctly determined by
chemical analysis?
4. Was the test compound adequately characterized with
regard to the nature and extent of impurities?
5. Did the animals actually receive the test compound?
6. Were animals that died during the test adequately exam-
ined?
7. How carefully were test animals observed during the
conduct of the test?
-------
‘Il—il
8. What tests were performed on the animals (e.g., blood
tests, clinical chemistry tests) and were they ade-
quately performed?
9. If the animals were examined histopathologically (i.e.,
detailed pathological examination based on sections
taken from individual tissues), was the examination
performed by a qualified pathologist?
10. Was the extent of animal and animal tissue examination
adequate?
11. Were the various sets of clinical and pathology data
properly tabulated?
12. Were the statistical tests used appropriate and were
they adequately performed?
13. Was the report of the test sufficiently detailed so
that these questions can be answered?
A proper evaluation would ensure that these and other types
of questions were examined and would include a list of qualifica-
tions on test results in areas where answers were missing or
unsatisfactory.
Categorization of Toxic Effects
Toxicity tests may reveal that a substance produces a wide
variety of adverse effects on different organs or systems of the
body or that the range of effects is narrow. Some effects may
occur only at the higher doses used, and only the most sensitive
indicators of a substance’s toxicity may be manifest at the lower
doses.
The toxic characteristics of a substance are usually catego-
rized according to the organs or systems they affect (e.g., liv-
er, kidney, nervous system) or the diseases they cause (e.g.,
cancer, birth defects). The most commonly used categorizations
of toxicity are briefly described in Appendix I.
Although there are uncertainties associated with most evalu-
ations of animal toxicity data, there are some special problems
associated with interpretation of carcinogenicity data. Because
these problems are the source of much controversy, we afford them
special attention in the next section.
-------
111-12
Uncertainties in Evaluation
of Animal Carcinogenicity
Test Results
One area of uncertainty and controversy concerns the occur-
rence of certain types of tumors in control animals. In most
animal, experiments, control animals will also develop tumors, and
interpretation of such experiments depends on comparing the inci-
dence of tumors in control animals with that observed in treated
animals. In some instances, this is not as straightforward as it
may seem. For example, the lifetime incidence of lung tumors in
a certain strain of male mice, untreated with any substance, may
vary from a low of about 2% to a high of about 40%; the average
rate is about 14%. Suppose that, in a particular experiment,
male mice treated with a substance exhibited a 35% incidence of
lung tumors, and control animals exhibited an incidence of 8%.
Statistical analysis of such data would show that the treated
animals experienced a statistically significant increase in tumor
incidence, and the substance producing this effect might be la-
beled a lung carcinogen.
Further analysis of the incidence data suggests that such a
statistical analysis may be misleading. The 35% incidence ob-
served in treated animals is within the range of tumor incidence
that is normally experienced by male mice, although the particu-
lar group of male mice used as controls in this experiment exhib-
ited an incidence in the low end of the normal range. Under such
circumstances, use of the simple statistical test of significance
might be misleading and result in the erroneous labeling of a
substance as a carcinogen.
Another major area of uncertainty arises in the interpreta-
tion of experimental observations of benign tumors. Some types
of tumors are clearly malignant; that is, they are groups of
cells that grow in uncontrolled ways, invade other tissues, and
are frequently fatal. There is usually no significant contro-
versy about such tumors, and pathologists generally agree that
their presence is a clear sign that a carcinogenic process has
occurred. - Other tumors are benign at the time they are observed
by pathologists, and it is not always clear they should be con—
idered indicators of a carcinogenic process. Some tumors will
remain benign for the lifetime of the animal, but in some cases
they have been observed to progress to malignancy. Generally,
the numbers of animals with benign tumors that are thought to be
part of the carcinogenic process are combined with those having
malignancies to establish the total tumor incidence. Many path-
ologists disagree with such combining, and there appears to be no
end to the controversy in this area. The issue has been espe-
cially controversial in connection with tumors found in rodent
livers.
-------
111—13
Short—Term Tests for Carcinogens
The lifetime animal study is the primary method used for
detecting the carcinogenic properties of a substance. In recent
years, other experimental techniques have become available and,
although none is yet considered definitive, they may provide
important information.
Short—term tests for carcinogenicity measure effects that
empirically or theoretically appear to be correlated with carcin-
ogenic activity. These tests include assays for gene mutations
in bacteria, yeast, fungi, insects, and mammalian cells; mamma-
lian cell transformation assays; assays for DNA damage and re-
pair; and in vitro (outside the animal-—e.g., bacterial cells as
in the Ames mutagenicity assay) and in vivo (within the animal)
assays for chromosomal mutations in animals’ cells. In addition
to these rapid (test—tube) tests, several tests of intermediate
duration involving whole animals have been used. These include
the induction of skin and lung tumors in mice, breast cancer in
female certain species of rats, and anatomical changes in the
livers of rodents.
Other tests are used to determine whether a substance will
interact with the genetic apparatus of the cell, as some well-
known carcinogens apparently do. However, not all substances
that interact with DNA have been found to be carcinogenic in
animal systems. Furthermore, not all animal carcinogens interact
directly with genetic material.
These short-term tests are playing increasingly important
roles in helping to identify suspected carcinogens. They provide
useful information in a relatively short period, and may become
critical screening tools, particularly for selecting chemicals
for long—term animal tests. They may also assist in understand-
ing the biological processes underlying the production of tumors.
They have not been definitively correlated with results in animal
models, however, and regulatory agencies and other public health
institutions do not consider positive or negative results in
these systems as definitive indicators of carcinogenicity or the
lack thereof, but only as ancillary evidence.
DATA FROM HUMAN STUDIES
Information on adverse health effects in human populations
is obtained from four sources: (1) summaries of self—reported
symptoms in exposed persons; (2) case reports prepared by medical
personnel; (3) correlational studies (in which differences in
-------
111-14
disease rates in human populations are associated with differ-
ences in environmental conditions); and (4) epidemiological stud-
ies. The first three types of study can be characterized as
descriptive epidemiology and are often useful in drawing atten-
tion to previously unsuspected problems. Although they cannot
identify a cause—and—effect relationship, they have value in
generating hypotheses that can be further tested. Epidemiologic
studies involve comparing the health status of a group of persons
who have been exposed to a suspected agent with that of a compar-
able nonexposed group.
Most epidemiology studies are either case—control studies or
cohort studies . In case—control studies, a group of individuals
with a specific disease is identified and an attempt is made to
ascertain commonalities in exposures they may have experienced in
the past. The carcinogenic properties of DES were discovered
through such studies. In cohort studies, the health status of
individuals known to have had a common exposure is examined to
determine whether any specific condition or cause of death is
revealed to be excessive, compared to an appropriately matched
control population. Benzene leuicemogenesis was established with
studies of these types. Generally, epidemiologists have turned
to occupational settings or to patients treated with certain
drugs to conduct their studies.
When epidemiological investigations yield convincing re-
sults, they are enormously beneficial because they provide infor-
mation about humans under actual conditions of exposure to a
specific agent. Therefore, results from well—designed, properly
controlled studies are usually given more weight than results
from animal studies in the evaluation of the total database.
Although no study can provide complete assurance that no risk
exists, negative data from epideiniological studies of sufficient
size can be used to establish the level of risk that exposure to
an agent almost assuredly will not exceed.
Although epidemiology studies are powerful when clearcut
differences exist, several points must be considered when their
results are interpreted:
• Appropriately matched control groups are difficult to
identify, because the factors that lead to the exposure
of the study group (e.g., occupation or residence) are
often associated with other factors that affect health
status (e.g., lifestyle and socioeconomic status).
• It is difficult to control for related risk factors
(e.g., cigarette smoking) that have strong effects
on health.
-------
111—15
• Few types of health effects (other than death) are
recorded systematically in human populations (and even
the information on cause of death is of limited relia-
bility). For example, infertility, miscarriages, and
mental illnesses are not as a rule systematically re-
corded by public health agencies.
• Accurate data on the degree of exposure to potentially
hazardous substances are rarely available, especially
when exposures have taken place in the past. Estab-
lishing dose—response relations is thus frequently
impossible.
• For investigation of diseases that take many years
to develop, such as cancer, it is necessary to wait
many years to ascertain the absence of an effect.
Of course, exposure to suspect agents could continue
during these extended periods of time and thereby
further increase risk.
• The statistical detection power of epidemiological
studies is limited, unless very large populations are
studied.
For these reasons, epidemiological studies are subject to
sometimes extreme uncertainties. It is usually necessary to have
independent confirmatory evidence, such as a concordant result in
a second epidemiological study, or supporting data from experi-
mental studies in animals. Because of the limitations of epi-
demiology, negative findings must also be interpreted with cau-
tion 6
6 i is important to recognize the limitations of negative epide—
miological. findings. A simple example reveals why this is so.
Suppose a drug that causes cancer in one out of every 100 people
exposed to 10 units is released for use (no one is aware of the
risks). Moreover, the average time required for cancer to
develop from 10 units’ exposure is 30 years (not uncommon for a
carcinogen). After the drug has been in use for 15 years, an
epidemiologist decides to study its effects. Me locates the
death certificates of 20 people who took the drug, but finds
little information on their dosage. Some took the drug when it
was first released, others not for several years after its
release. The health records, which are incomplete, reveal no
excess cancer in the 20 people when compared to an appropriate
control group. Is it correct to conclude that the drug is not
carcinogenic?
-------
111—16
HAZARD IDENTIFICATION: A SUMMARY
For some substances the available database may include sub-
stantial information on effects in humans and experimental
animals, and may also include information on the biological mech-
anisms underlying the production of one or more forms of toxi-
city. In other cases, the database may be highly limited and may
include only a few studies in experimental animals.
In some cases, all the available data may point clearly in a
single direction, leaving little ambiguity about the nature of
toxicity associated with a given compound; in others, the data
may include apparently conflicting sets of experimental or epide-
miological. findings. It is not unusual for a well-studied com-
pound to have conflicting results from toxicity tests. If the
tests are performed properly, positive tests results usually
outweigh negative test results. Confusion may be compounded by
the observation that the type, severity, or site of toxicity may
vary with the species of animal exposed. Although it is gen-
erally accepted that results in animals are and have been useful
in predicting effects in humans, such notable exceptions as
thalidomide have occurred. This complex issue, briefly mentioned
here, must be considered for each compound examined.
The foregoing discussion of hazard evaluation was derived
for exposures to a single toxic agent. Humans are rarely exposed
to only one substance: commercial chemicals contain impurities,
chemicals are used in combinations, and lifestyle choices (e.g.,
smoking, drinking) may increase exposure to mixtures of chemi-
cals. When humans are exposed to two or more chemicals, several
results may occur. The compounds may act independently; that is,
exposure to the additional chemical(s) has no observable effect
on the toxic properties of the substance. Toxic effects of chem-
icals may be additive; that is, if chemical A produces 1 unit of
disease and chemical B produces 2 units of disease, then exposure
to chemicals A and B produces 3 units of disease. Exposure to
combinations of chemicals may produce a greater than additive
(synergistic) effect; that is, exposure to chemicals A and B
produces more than 3 units of disease. Finally, chemicals may
reduce thedegree of toxicity of each other (antagonism); that
is, exposure to chemicals A and B produces less than 3 units of
disease. Hazard evaluation of mixtures of chemicals is complex
and not standardized.
A proper hazard evaluation should include a critical review
of each pertinent data set and of the total database bearing on
toxicity. It should also include an evaluation of the inferences
-------
111—17
about toxicity in human populations who tnight be exposed. At
this stage of risk assessment, however, there is no attempt to
project human risk. For the latter, at least two additional sets
of analyses must be conducted.
-------
IV. DOSE-RESPONSE EVALUATION
INTRODUCTION
The next step in risk assessment is to estimate the dose—
response relationships for the various forms of toxicity exhib-
ited by the substance under review. Even where good epidemiolo-
gical studies have been conducted, there are rarely reliable
quantitative data on exposure. Hence, in most cases dose-
response relationships must be estimated from studies in animals
which immediately raises three serious problems: (1) animals are
usually exposed at high doses, and effects at low doses must be
predicted, using theories about the form of the dose-response
relationship; (2) animals and humans often differ in suspectibil—
ity, if only because of differences in size and metabolism; and
(3) the human population is very heterogeneous, so that some
individuals are likely to be more susceptible than average.
Toxicologists conventionally make two general assumptions
about the form of dose—response relationships at low doses. For
effects that involve alteration of genetic material (including
the initiation of cancer), there are theoretical reasons to be-
lieve that effects may take place at very low dose levels; sever-
al specific mathematical models of dose—reponse relationships
have been proposed. For most other biological effects, it is
usually assumed that Nthresholdu levels exist. However, it is
very difficult to use such measures to predict safe” levels in
humans. Even if it is assumed that humans and animals are, on
the average, similar in intrinsic susceptibility, humans are
expected to have more variable responses to toxic agents. We
discuss these and other issues at length in the following subsec-
tions.
THRESHOLD EFFECTS
It is widely accepted on theoretical grounds, if not defini-
tively proved empirically, that most biological effects of chemi-
cal substances occur only after a threshold dose is achieved. In
the experimental systems described here, the threshold dose is
approximated by the no—observable—effect level or NOEL.
It has also been widely accepted, at least in the process of
setting public health standards, that the human population is
likely to have much more variable responses to toxic agents than
are the small groups of well—controlled, genetically homogeneous
-------
IV-2
animals ordinarily used in experiments. Moreover, the NOEL is
itself subject to some uncertainty (e.g., how can it be known
that the most serious effects of a substance have been identi-
fied?). For these reasons, standard—setting and public health
agencies protect populations from substances displaying threshold
effects by dividing experimental NOELs by large Nsafety factors.”
The magnitude of safety factors varies according to the nature
and quality of the data from which the NOEL is derived; the seri-
ousness of the toxic effects; the type of protection being sought
(e.g., are we protecting against acute, subchronic, or chronic
exposures?); and the nature of the population to be protected
(e.g., the general population, or populations——such as workers-—
expected to exhibit a narrower range of susceptibilities). Safe-
ty factors of 10; 100; 1,000; and 10,000 have been used in vari-
ous circumstances.
NOBLe are used to calculate the Acceptable Daily Intake
(ADI) for humans (which goes by other names in some circum-
stances) for chemical exposures. The ADI is derived by dividing
the experimental NOEL, in mg/kg/day, for the toxic effect appear-
ing at lowest dose, by one of the safety factors listed above.
The ADI (or its equivalent) is thus expressed in mg/kg/day. For
example, a substance with a NOEL from a chronic toxicity study of
100 mg/kg/day may be assigned an ADI of 1 ing/kg/day, for chronic
human exposure. The concentration of the substance-—be it pesti-
cide, food additive, or drinking water contaminant--permitted in
various media must be determined by taking into account the vari-
ous uses to which the material has been or will be put, the pos-
sible routes of exposure, and the degree of human contact. The
permitted concentrations, sometimes called tolerances or crite-
ria, are assigned to ensure the ADI is not exceeded.
This approach has been used for several decades by such
federal regulatory agencies as FDA and EPA, as well as by such
international bodies as the World Health Organization and by
various committees of the National Academy of Sciences.
Although there may be some biological justification for
assuming tk)e need for safety factors to protect the more sensi-
tive members of the human population, there is very little scien-
tific support for the specific safety factors used. They are
arbitrarily chosen to compensate for uncertainty and, in fact,
could be seen as policy rather than scientific choices.
There is no way to determine that exposures at ADIs esti-
mated in this fashion are without risk. The ADI represents an
acceptable, low level of risk but not a guarantee of safety.
Conversely, there may be a range of exposures well above the ADI,
perhaps including the experimental NOEL itself, that bears no
-------
IV-3
risk to humans. The “NOEL—safety factor” approach includes no
attempt to ascertain how risk changes below the range of experi-
mentally-observed dose—response relat4ons.
The assessment of low dose Nrisksu from threshold agents are
discussed in Section VI on Risk Characterization.
EFFECTS THAT MAY NOT EXHIBIT THRESHOLDS
At present, only agents displaying carcinogenic properties
are treated as if they do not display thresholds (although a few
scientists suggest that some teratogens and mutagens may behave
similarly). In somewhat more technical terms, the dose—response
curve for carcinogens in the human population achieves zero risk
only at zero dose; as the dose increases above zero, the risk
immediately becomes finite and thereafter increases as a function
of dose. Risk is the probability of cancer, and at very low
doses the risk can be extremely small (this will vary according
to the potency of the carcinogen). In this respect, carcinogens
are not much different from agents for which hDIs are established
(i.e., the most that can be said about an ADI is that it repre-
sents a very low risk, not that it represents the condition of
absolute safety).
The Carcinogenic Process
If a particular type of damage occurs to the genetic mate-
rial (DNA) of even a single cell, that cell may undergo a series
of changes that eventually result in the production of a tumor;
however, the time required for all the necessary transitions that
culminate in cancer may be a substantial portion of an animal’s
or human’s lifetime. Carcinogens may also affect any number of
the transitions from one stage of cancer development to the next.
Some carcinogens appear capable only of initiating the process
(these are termed irtitiators”). Still others act only at later
stages, the natures of which are not well known (so—called promo—
tors may aát at one or more of these later stages). And some
carcinogens may act at several stages. Some scientists postulate
that an arbitrarily small amount of a carcinogen, even a single
molecule, could affect the transition of normal cells to cancer-
ous cells at one or more of the various stages, and that a great-
er amount of the carcinogen merely increases the probability that
a given transition would occur. Under these circumstances there
is little likelihood of an absolute threshold below which there
is no effect on the process (even though the effect may be ex-
ceedingly small).
-------
IV-4
This description of the carcinogenic process is still under
extensive scientific scrutiny and is by no means established.
However, it is by far the dominant model and it has substantial
support. This multistage model has influenced the development of
some of the models used for dose—response evaluation. Before
discussing these models further, it is useful to review the ex-
perimental dose—response information obtained from bioassays and
to discuss why models of the dose—response relation are needed.
Potency and High—to—Low Dose Extrapolation
The following example, drawn from Rodricks and Taylor, 7
illustrates the need for high—to—low dose extrapolation. Assume
that a substance has been tested in mice and rats of both sexes
and been found to produce liver cancer in male rats. A typical
summary of the data from such an experiment might be as follows:
Lifetime Incidence Lifetime
Lifetime Daily of Liver Cancer Probability of
Dose in Rats Liver Cancer
0 mg/kg/day 0/50 0.0
125 mg/kg/day 0/50 0.0
250 mg/kg/day 10/50 0.20
500 mg/kg/day 25/50 0.50
1000 mg/kg/day 40/50 0.80
The incidence of liver cancer is expressed as a fraction,
and is the number of animals found to have liver tumors divided
by the total number of animals at risk. The probability (P) of
cancer is simply the fraction expressed as a decimal (e.g., 25/50
= 0.50).
Although there is no—effect” at 125 mg/kg/day, the response
is nevertheless compatible with a risk of about 0.05 (5%) because
of the statistical uncertainties associated with the small nuxn—
bers of animals used.
This experiment reveals that if humans and rats are about
equally susceptible to the agent, an exposure of 250 mg/kg/day in
humans will increase their lifetime risk by 20%; if 1,000 people
were to be exposed to this substance at this dose for a lifetime,
then 200 of these people will be expected to contract cancer from
this substance. This is an extremely high risk and obviously one
7 ”Application of Risk Assessment to Food Safety Decision-Making,”
Regulatory Toxicology & pharmacology (1983), 3:275—307.
-------
‘v-s
that no one would sanction. However, it is near the low end of
the range of risks that can be detected in animal experiments.
To continue with the illustration, assume that it is possi-
ble to estimate the daily dose of the chemical in the human popu-
lation. For the present example, assume that the exposed human
population receives a dose of 1.0 mg/kg/day. It thus becomes of
interest to know the risk to male rats at 1.0 mg/kg/day.
There is a great difference between the doses used experi-
mentally and the dose of interest. The risks that would likely
exist at a dose of 1.0 mg/kg/day are quite small and to determine
whether they exist at all would require enormous numbers of ani-
mals (perhaps hundreds of thousands). tt is thus necessary under
these circumstances to rely on means other than experimentation
to estimate potential risk.
Scientists have developed several mathematical models to
estimate low dose risks from high dose risks. Such models de-
scribe the expected quantitative relationship between risk C?)
and dose Cd), and are used to estimate a value for P (the risk)
at the dose of interest (in our example, the dose of 1.0 mg/kg/
day). The accuracy of the projected P at the dose of interest,
d, is a function of how accurately the mathematical model de-
scribes the true, but immeasurable, relationship between dose and
risk at the low dose levels.
These mathematical models are too complex for detailed expo-
sition in this document. Various models may lead to very differ-
ent estimations of risk. None is chemical—specific; that is,
each is based on general theories of carcinogenesis rather than
on data for a specific chemical. None can be proved or disproved
by current scientific data, although future results of research
may increase our understanding of carcinogenesis and help in
refining these models. Regulatory agencies currently use one—
hit, multistage, and probit models, although regulatory decisions
are usually based on results of the one—hit or multistage models.
They also use multihit, Weibull, and logit models for risk
assessment,
If these models are applied to the data recorded earlier for
the hypothetical chemical, the following estimates of lifetime
risk for male rats 8 at the dose of 1.0 mg/kg/day are derived:
8 A11 risks are for a full lifetime of daily exposure. The life-
time is the unit of risk measurement because the experimental
data reflect the risk experienced by animals over their full
lifetimes. The values shown are upper confidence limits on risk
(data drawn from Rodricks and Taylor, 1983).
-------
IV-6
Model Applied Lifetime Risk at 1.0 mg/kg/day
One—hit 6.0 x i0 Cone in 17,000)
Multistage 6.0 x 10—6 (one in 167,000)
Multihit 4.4 x i0 (one in 230,000)
Weibull 1.7 x 10—8 (one in 59 million)
Probit 1.9 x 10 10 (one in 5.2 billion)
There may be no experimental basis for deciding which esti-
mate is closest to the truth. Nevertheless, it is possible to
show that the true risk, at least to animals, is very unlikely to
be higher than the highest risk predicted by the various models.
In cases where relevant data exist on biological mechanisms
of action, the selection of a model should be consistent with
the data. In many cases, however, such data are very limited,
resulting in great uncertainty in the selection of a model for
low dose extrapolation. At present, understanding of the mecha-
nism of the process of carcinogenesis is still quite limited.
Biological evidence, however, does indicate the linearity of
tumor initiation, and consequently linear models are frequently
used by regulatory agencies.
The one—hit model always yields the highest estimate of low
dose risk. This model is based on the biological theory that a
single uhit of some minimum critical amount of a carcinogen at a
cellular target——namely, DNA——can initiate an irreversible series
of events that eventually lead to a tumor.
The multistage model, which yields risk estimates either
equal to or less than the one—hit model, is based on the same
theory of cancer initiation. However, this model can be more
flexible, allowing consideration of the data in the observable
range to influence the extrapolated risk at low dose. It is also
based on the multistage theory of the carcinogenic process and
thus has a plausible scientific basis. EPA generally uses the
linearized multistage model for low dose extrapolation because
its scientific basis, although limited, is considered the strong-
est of the currently available extrapolation models. This model
yields estimates of risk that are conservative, representing a
plausible upper limit for the risk. In other words, it is un-
likely that the actual risk is higher than the risk predicted
under this model.
The probit model incorporates the assumption that each indi-
vidual in a population has a tolerance dose and that these
doses are distributed in the population in a specified certain
way, The other models have more complex bases; because none is
-------
IV-7
widely used we shall not discuss them. None of the models, as
currently used, incorporates a threshold dose for an exposed
population.
Interspecies Extrapolation
For the majority of agents, dose—response evaluation primar-
ily involves the analysis of tests that were performed on labor-
atory animals, because useful human data are generally not avail-
able. In extrapolating the results of these animal tests to
humans, the doses administered to animals must be adjusted to
account for differences in size and metabolic rates. Differences
in metabolism may influence the validity of extrapolating from
animals to man if, for example, the actual material producing the
carcinogenic effect is a metabolite of the tested chemical, and
the animal species tested and humans differ significantly in
their metabolism of the material.
Several methods have been developed to adjust the doses used
in animal tests to allow for differences in size and metabolism.
They assume that human and animal risks are equivalent when doses
are measured in:
• Milligrams per kilogram body weight per day
• Milligrams per square meter of body surface area per
day
• Parts per million in the air, water, or diet
• Milligrams per kilogram per lifetime.
Currently, a scientific basis for using one extrapolation method
over another has not been established.
DOSE-RESPONSE EVALUATION: A SUMMARY
For substances that do not display carcinogenic properties,
or for the noncarcinogenic effects of carcinogens, dose—response
evaluation consists of describing observed dose—response rela-
tions and identifying experimental NOEL5. NOELs can be used to
establish ADI5, or can be used for the type of risk character-
ization described in Section VI.
For carcinogens, various models are applied to project the
dose—response curve from the range of observed dose—responses to
-------
IV-8
the range of expected human doses. After the known or expected
human dose is estimated (Section V) carcinogenic risk can be
characterized (Section VI). Although the models in use yield a
range of dose—response relations, it is highly likely that the
projections of the more protective models will not underestimate
risk, at least to experimental animals, and they may strongly
overestimate it. None of the models includes a threshold. In a
few cases, dose-response data are available from human epidemi—
ology studies and may be used in lieu of animal data for low dose
extrapolation.
It appears that certain classes of carcinogens do not possess
the capacity to damage DNA (they are not genotoxic); in our ear-
her discussion of the carcinogenic process, such substances
would affect only late stages in the process. Some scientists
maintain that such (nongenotoxic) carcinogens must operate under
threshold mechanisms. Many of the reasons for such a hypothesis
are sound, but no general consensus has yet emerged on this mat-
ter. It is nevertheless possible that some classes of carcino-
gens could be treated in the same way noncarcinogena are treated
for purposes of establishing ADIs.
-------
V. HUMAN EXPOSURE EVALUATION
Assessment of human exposure involves estimation of the num-
ber of people exposed and the magnitude, duration, and timing of
their exposure. In some cases, it is fairly straightforward to
measure human exposure directly, either by measuring levels of
the hazardous agents in the ambient environment or by using per-
sonal monitors. In most cases, however, detailed knowledge is
required of the factors that control human exposure, including
those factors which determine the behavior of the agent after its
release into the environment. The following types of information
are required for this type of exposure assessment:
• Information on the factors controlling the production
of the hazardous agent and its release into the envi-
ronment.
• Information on the quantities of the agent that are
released, and the location and timing of release.
• Information on the factors controlling the fate of the
agent in the environment after release, including fac-
tors controlling its movement, persistence, and degrad-
ation. (The degradation products may be more or less
toxic than the original agent.)
• Information on factors controlling human contact with
the agent, including the size and distribution of vul-
nerable human populations, and activities that facili-
tate or prevent contact.
• Information on human intakes.
The amount of information of these types that is available
varies greatly from case to case and is difficult to discuss in
general terms. For some agents, there is fairly detailed infor-
mation on the sources of release into the environment and on the
factors controlling the quantities released. However, for many
agents there is very limited knowledge of the factors controlling
dispersion and fate after release. Measurements of transport and
degradation in the complex natural environment are often diff i—
cult to conduct, so it is more co on to rely on mathematical
models of the key physical and chemical processes, supplemented
with experimental studies conducted under simplified conditions.
Such models have been developed in considerable detail for radio—
isotopes, but have not yet been developed in comparable detail
for other physical and chemical agents.
-------
V- 2
In comparison with toxicology and epidemiology, the science
of exposure assessment is still at a very early stage of develop-
ment. Except in fortunate circumstances, in which the behavior
of an agent in the environment is unusually simple, uncertainties
arising in expos ire assessments are often at least as large as
those arising in assessments of inherent toxicity.
Once these various factors are known human data can be esti-
mated, as described earlier. The dose, its duration and timing,
and the nature and size of the population receiving it are the
critical measures of exposure for risk characterization.
-------
VI. RISK CHARACTERIZATION
The final step in risk assessment involves bringing together
the information and analysis of the first three steps. Risk is
generally characterized as follows:
1. For noncarcinogens, and for the noncarcinogenic effects
of carcinogens, the margin-of-safety (MOS) is estimated
by dividing the experimental NOEL by the estimated
daily human dose.
2. For carcinogens, risk is estimated at the human dose by
multiplying the actual human dose by the risk per unit
of dose projected from the dose—response modelling. A
range of risks might be produced, using different mod-
els and assumptions about dose-response curves and the
relative susceptibilities of humans and animals.
Although this step can be far more complex than is indicated
here, especially if problems of timing and duration of exposure
are introduced (as they no doubt need to be in the present case),
the MOS and the carcinogenic risk are the ultimate measures of
the likelihood of human injury or disease from a given exposure
or range of exposures.
The ADIS described earlier are not measures of risk; they
are derived by imposing a specified safety factor (or, in the
above language, a specified MOS). Our purpose here is not to
specify an ADI, but to ascertain risk. There is no means availa-
ble to accomplish this for noncarcinogens. The MOS is used as a
surrogate for risk: as the MOS becomes larger, the risk becomes
smaller. At some point, most scientists agree that the MOS is so
large that human health is almost certainly not jeopardized. The
magnitude of the MOS needed to achieve this condition will vary
among different substances, but its selection would be based on
factors similar to those used to select safety factors to estab-
lish ADIs. -
-------
Appendix
TOXIC EFFECTS ON ORGANS AND OTHER TARGET SYSTEMS
-------
Appendix
INTRODUCTION
To understand the potential toxic effects of chemicals, it is
useful to understand the toxic effects (i.e., measurable effects)
on endpoints that are commonly observed in animals, including
humans. While the following discussion is presented by organ or
system, chemicals frequently affect more than one organ and can
produce a variety of endpoints. Concentration of the chemical,
duration of exposure, and route of exposure are three of the
factors that can influence the potential toxic effect.
LIVER
A major function of the liver is metabolism——i.e., the bio-
chemical conversion of one substance into another for purposes of
nutrition, storage, detoxification, or excretion. The liver has
multiple mechanisms for each of these processes, and interference
with any of the processes can lead to a toxic effect. Chemicals
that damage the liver are termed hepatotoxic.” Toxic endpoints
of the liver can include lipid (e.g., fat) accumulation, jaun-
dice, cell death (necrosis), cirrhosis, and cancer. In addition,
chemicals that increase the level of metabolic enzymes, i.e.,
enzyme inducers, can dramatically affect the toxicity of other
compounds.
The accumulation of lipids, primarily triglycerides, is re-
lated to the liver’s conversion of sugars and carbohydrates into
fat for storage (or vice versa for energy production during star-
vation). Chemicals that increase the rate of triglyceride syn-
thesis, decrease the rate of triglyceride excretion, or both can
lead to an accumulation of lipids in the liver and a concomitant
decrease of triglycerides in the blood. While the effects of
lipid accumulation in the liver are not known, a fatty liver is
generally regarded as an indication of an injury to the organ.
Jaundiàe is a frequent endpoint when the excretory functions
of the liver are impaired; the yellow cast of the skin is caused
by the retention in the blood of the yellow bile pigments that
would normally be excreted. Since blood that has absorbed com-
pounds from the gastrointestinal tract passes through the liver
before the rest of the body, the liver is a major site for the
removal of nutrients and toxicants. Elimination of the absorbed
toxicants can occur in the feces via the bile. In addition to
-------
A-2
bile acting as a mechanism of excretion, bile salts aid in the
absorption of nutrients that are not water soluble. Thus, im-
pairing liver function can affect absorption of compounds. Fi-
nally, the liver is also a site of the destruction of aged red
blood cells. Jaundice is an indicator of liver malfunction.
Necrosis, or cell death, can occur from multiple causes.
There are many mechanisms by which toxicants can directly or
indirectly inhibit required cell functions. The liver has a
limited ability to regenerate destroyed cells. Chronic destruc-
tion of cells, however, may lead to cirrhosis of the liver in
which the normal liver cells (hepatocytes) are replaced by al-
tered cells and connective tissue such as collagen.
A wide variety of chemicals have been shown to cause liver
cancers in laboratory animals. Exposure to vinyl chloride has
been associated with liver cancers in humans. The theories and
uncertainties of carcinogenesis are discussed in the main text.
As a major site of metabolism and and detoxification, the
liver contains enzyme systems that biochemically alter compounds.
Many of these processes facilitate excretion by making the com-
pound n re polar, i.e., highly charged (e.g., cytochrome P—450
systems) or attaching polar groups to the compound (e.g., gluta—
thione, glycuronyl, or sulfo—transferases). The speed at which
this occurs depends on the amount of enzyme present; the amount
of enzyme can be increased by exposure to certain chemicals
called inducers. If a nonmetabo].ized compound is toxic, exposure
to an inducer may decrease the toxic effect by increasing the
rate at which the compound is metabolized. If the compound needs
to be metabolized to be toxic, however, exposure to an inducer
may increase the toxic effect by increasing the rate of its meta-
bolism.
KIDNEY
As an organ whose major function is the elimination of toxi—
cants and çther waste products, the kidney can be considered
a complex, elaborate filter. The kidney concentrates wastes for
elimination and retains nutrients and water that are useful to
the body. The kidney can metabolize and detoxify some of the
same compounds as the liver, although the rate of metabolism is
usually slower. Compounds that injure the kidney are called
renal toxicants. Some renal toxicants may cause cell death
(necrosis) or cancer. In addition, the kidney produces chemicals
necessary for homeostasis (maintenance of the body’s balance of
functions) and responds to the sympathetic nervous system. To
efficiently remove the body’s waste, the kidneys must process
-------
A- 3
large volumes of blood. Thus, the first level of susceptibility
of the kidney is that which changes the flow of fluids. This
change can be mechanical-—e.g., kidney stones or puncturing
vesicles——or chemicals that dilate or constrict the passages.
The complexity of the kidney’s filtering function makes it
susceptible to a number of toxicants. Although some of the fil-
tering requires no energy or special enzymes since the flow is
from high to low concentrations, much of the selection is to a
higher concentration than in the blood and is performed by en-
zymes that may be affected by chemicals. Excessive elimination
of water, salts, or other nutrients can be as harmful as failure
to eliminate wastes. Furthermore, because the kidneys concen-
trate some toxicants, the effective dose of toxicants to the
kidneys may be higher than that for the rest of the body. Toxi-
cants that cause necrosis can also impair renal function. Fail-
ure of the kidneys to filter properly is frequently detected by
an increase in wastes in the blood or an increase in nutrients in
the urine.
The ability of the kidney to metabolize compounds has not
been studied as extensively as has metabolism in the liver. The
presence of inducible metabolic enzyme systems is known. Other
specific metabolic functions occur in the kidney. Finally, be-
cause the kidney produces compounds that are necessary for other
body functions, damage to the kidney may affect other organ sys-
tems.
REPRODUCTIVE SYSTEM
Reproductive toxicology involves at least three organisms
(both male and female parents and their offspring) and consists
of many steps and stages. Toxic effects to the reproductive
system can be classified into three general endpoints: impaired
ability to conceive, failure of the conceptus to survive, and
production of abnormal offspring.
Problems with conception usually result from impaired produc-
tion of the sperm or egg. The formation of sperm (spermatogene—
sis) is continuous in the male and requires a series of steps.
Chemicals that interfere with these steps may prevent sperm pro-
duction and cause sterility, reduce sperm production, or result
in abnormal sperm that have reduced capacity to fertilize. Al-
though in mammals all eggs are formed before birth, their final
maturation occurs in cycles after puberty. Chemicals, e.g.,
contraceptives, can impede this process. Mature sperm and egg,
as well as proper biochemical and physiological conditions within
the body, are required for fertilization.
-------
A-4
Viability of the conceptus depends on a series of steps, in-
cluding implantation and development of the amniotic sac and
placenta. Death of the conceptus, whether at the early embryonic
stage or later fetal stage, can be caused by a variety of factors
including chemicals. Such chemicals are labeled Nembryotoxicu
and Nfetotoxic,u respectively.
Chemicals that cause defects in development and result in
abnormal offspring are called teratogens. Defects range from
abnormal skeletal or muscle structure and mental retardation, to
metabolic malfunctions, to subtle malfunctions that may not be
noticed during a normal life.
Functionally, for the developing mammal to be exposed, the
chemical must pass through two barriers: the mother and the
placenta. If a given dose of a compound is sufficiently toxic to
kill the mother, resultant toxic effects on the offspring will
not be observed. Although this statement may seem trivial, its
converse is an important principle in teratogenesis. The more
dangerous teratogens are those which affect the developing organ-
ism at concentrations that are significantly lower than those
that affect the adult mother.
Although the placenta was once thought to be a rather strong
barrier, many chemicals have been found to cross to the con—
ceptus. Depending on the compound, the final concentration may
be higher in the mother, higher in the conceptus, or equal in
mother and conceptus. Moreover, the placenta is not inert but is
capable of metabolizing some chemicals into either more or less
toxic substances. Metabolism may also affect the flow of com-
pound across the placenta.
Timing has two critical aspects in teratogenesis: timing of
the dose during development and parallel timing of developing
systems. Time of exposure to the potential teratogen may not
only determine which developing system is affected but also
whether the compound will have any effect at all. For each de-
veloping system there is a critical period, usually between three
and twelve weeks in the human, during which the system is parti-
cularly sensitive to chemically induced abnormal development.
Although terata may form after this period, the abnormalities are
usually less severe.
The second aspect of timing involves the relative rate of
development of each of the organ systems. To produce a well-
formed offspring, development must be well orchestrated. As with
a symphony, the pace must be parallel in all sections. Nerves
cannot attach to muscles that are not present; cleft palate in
laboratory animals is frequently caused by events occurring out
-------
A-5
of sequence. If all the developing systems were equally re-
tarded, the result might be an immature, but not malformed fetus.
LUNGS
The major function of the lungs is to exchange oxygen and
carbon dioxide between blood and air. This same mechanism can
facilitate entry and exit of other compounds from the body. In
addition, the lungs have the ability to alter some chemicals
metabolically. Damage to the lung can range from irritation and
constriction, to cell death (necrosis), edema, or fibrosis, to
cancer.
The air not only contains a variety of gases but also small
suspended particulates and liquid aerosols. The fate and, there-
fore, potential to cause damage, for each physical state depends
on the size and composition of the inhaled substance. An analogy
is often drawn between the airways of the respiratory passages
and the structure of a tree. In both, the starting point has a
large diameter and branches into more numerous but increasingly
smaller appendages. Given the size of the passage and the fact
that large particles fall out of suspension faster, larger in-
haled particulates and droplets will generally deposit in the
upper respiratory tract. Deposition is also affected by the
breathing pattern——for example, how fast and how deep.
The lung contains other mechanisms for handling inhaled sub-
stances including secretions, the mucociliary escalator, and
macrophages. Secretions, including mucus, can facilitate trans-
port of compounds across the lungs, between the air and blood.
The mucociliary escalator consists of mucus and hairlike projec-
tions in the upper respiratory passages. The latter move so that
particles that have been deposited are transported up the passage
until they can be swallowed. Substances that either affect the
mucus or inhibit the cilia movement can impair this process.
Macrophagea are a type of mobile cell that can engulf particles.
Lungs-facilitate exchange in both directions between air and
blood; thus, they can be equally efficient in absorption or ex-
cretion from the body. Whether a given substance is concentrated
in the blood or in the lung air or is at equal concentrations on
both sides depends on several factors, including its solubility
in water and ability to be bound to proteins in the blood. Fur—
thermore, lungs are able to metabolize some chemicals. These
changes may alter the chemical properties and, therefore, the
transport of the chemical.
-------
A-6
Chemicals that irritate the lung can lead to discomfort.
Although the effects of exposure to irritants are usually revers-
ible, chronic exposure may lead to permanent cell damage. The
normal, necessary exchange of gases across the lung can be im-
paired by compounds that constrict the respiratory passages,
affect secretions or other normal functions, or physically remain
in the lung. Substances that cause necrosis, edema (excessive
fluid retention), or fibrosis (a change in cell type and composi-
tion) will impair lung function. Exposure to some substances,
such as cigarette sixke, asbestos, and arsenic, can lead to im-
paired lung function and cancer.
SKIN
Skin is a barrier between the internal organism and the ex-
ternal environment. It prevents loss of body fluids, regulates
body temperature, and prevents entry of many substances. How—
ever, the skin is a route of entry for some toxicants. Dermal
toxicants can cause irritation, sensitization, pigmentation
changes, chloracne, ulcerations, and cancer.
The skin can also be a major route of entry for other sub-
stances-—for example, some pesticides and solvents. Moreover,
abrasions or cuts on the skin can compromise the barrier. Com-
pounds that are absorbed through the skin may affect other
systems——for example, organophosphate pesticides that affect the
nervous system. Similarly, compounds that enter by other routes
may affect the skin——for example, the oral ingestion of arsenic
causes dermal changes.
Irritation, rashes, and itching are common toxic reactions to
dermal exposures. Chemical sensitizers may cause an allergic
reaction that becomes more severe with continued exposure to
light. Folliculitis (damage to the hair follicles) and acne are
other common skin disorders. Chloracne is a particular form of
acne that is often caused by exposure to chlorinated hydrocar-
bons. Compounds can change skin pigmentation. Skin keratoses
(hardening-or scaling) or ulcers are additional toxic responses.
Skin cancer may be caused by dermal contact with some agents or
systemic administration of others.
CENTRAL NERVOUS SYSTEM
The major function of the central nervous system (C 1S) is
communication. Control of reflexes, movement, sensory informa-
tion, autonomic functions (e.g., breathing), and intelligence are
-------
A-7
controlled by the CNS. These functions can be impaired by toxi-
cants. Damage to the nervous system can occur in the brain or
other nerve cell bodies, to nerve processes that extend through
the body, to the rnyelin sheaths that cover these processes, and
at the nerve—nerve or nerve—muscle junctions. Damage to nerve
cell functions are often called “neuropathies.
As in other cells, damage to the cell body of a neuron (nerve
cell) can result in impaired function or death. The brain is
partially protected by the blood—brain barrier. Like other phy-
siological barriers, this one has proven more permeable than
originally thought, although it does block or reduce the passage
of some substances to the brain. In contrast, certain substan-
ces, such as organic mercury, have been shown to concentrate in
the CNS.
Axons are long processes that conduct impulses from the nerve
cell body; they can span much of the length of an animal. Sever-
ing the axon can destroy transmission of signals along the nerve.
Because electrical signals are transmitted by charged elements
(ions), chemicals that change the permeability of the cell mem-
brane to ions can also impair transmission of the signal.
Myelin is the insulating cover of axons. Special cells,
called Schwann cells, form myelin by wrapping themselves in many
layers around the axons. Chemicals can either destroy the myelin
or decrease its amount, both of which decrease the insulation and
impair signal transmission. Furthermore, demyelination of nerves
can cause a degeneration of the axon. These effects take time to
occur, even if damage is caused by a single exposure. Thus, the
effect may be delayed and not immediately associated with the
exposure.
Transmission of signals between nerves or from a nerve to a
muscle occurs across a space or junction. Chemical compounds
that are stored in vesicles at the nerve endings carry the signal
across the junctions. Exposure to chemicals may accelerate or
inhibit release of these vesicles, mimic the compounds that are
released from the vesic].es, or block the receptors that react to
release of the compounds. Any of these responses will distort
the signal.
Subjective or behavior neurological toxicology may be the
most difficult toxicological effects to assess. While generally
accepted that exposure to some chemicals can cause headaches,
fatigue, or irritability, it is difficult to determine whether
such symptoms are caused by chemical exposure, lack of sleep,
depression, or other factors. Although these symptoms may be
mild and difficult to assess, they are frequently an early warn-
ing of exposure to a toxicant.
-------
A-8
Behavioral changes are often caused by damage to the nervous
system. In laboratory animals, such damage may be as precise and
fatal as failure of pups to nurse. Mental retardation and learn-
ing disabilities are other measurable behavioral changes. Chemi-
cal alteration of behavior is the basis for psychological drug
therapy. Thus, although they are difficult to assess, behavioral
changes should not be ignored.
BLOOD
Transport of oxygen, carbon dioxide, and other materials is
the major function of blood. The hematopoietic system, which
includes organs and tissues that produce, transport, and filter
blood, interacts with the cells of all other systems. Toxicity
can occur to developing blood cells, existing cells, or the hema-
topoietic organs.
In the human being and other mammals, blood cells are formed
in bone marrow; the three major types of blood cells are formed
by branches from a common precursor cell. Red blood cells con-
tain hemoglobin and transport oxygen and carbon dioxide. White
blood cells function as part of the immune system. Platelets are
necessary for blood clotting. Chemicals toxic to bone marrow can
affect blood formation. Depending on the stage and cell affect-
ed, any or all of the major blood cells may be decreased in nuin—
ber. Abnormal increases in production of certain blood cells are
also possible, as in leukemia (excess white cells).
Blood plasma contains a number of proteins, ions, and other
compounds. Changes in the chemical composition of blood may
indicate a toxic response. Furthermore, some chemicals bind to
plasma proteins. Changes in plasma protein composition could
affect the effective concentration of a toxicant.
The normal function of the hemoglobin in circulating red
blood cells is critical to the transport of oxygen to and carbon
dioxide from all cells in the body. Reduced oxygen supply can be
very detrimental; the effects resulting from oxygen deprivation
vary with the site of action. Chemicals can affect hemoglobin by
chemically oxidizing the heme group (causing methemoglobin) or by
denaturing the hemoglobin (which may lead to the formation of
Heinz bodies).
Two other heinatopoietic organs that may be affected are the
spleen and heart. The former removes old or damaged red blood
cells from circulation. The rate and efficiency of the heart’s
pumping action can be altered by many causes. Chemicals that
-------
A-9
constrict or dilate the blood vesicles can also affect circu-
latory function.
IMMUNE SYSTEM
Recognition and protection against foreign substances in the
body is handled by the immune system. Rapid advances are being
made in immunology research; therefore, current knowledge may
soon be obsolete. Three types of cells (inacrophages, B lympho-
cytes, and T lymphocytes) are part of the body’s immune response.
These cells interact at the peripheral lymphoid organs (lymph
nodes, spleen, and tonsils). Exposure to chemicals may activate
or supress the immune system.
The cells involved in the immune system are formed in bone
marrow; hence, chemicals that affect bone marrow may impair im-
mune function. One type of cell engulfs foreign matter, especi-
ally bacterial and viruses, by phagocytosis. Another type pro-
duces the five classes of antibodies. A third type produces
polypeptides, such as interferon, that are important for some
immune responses; this type of cell is also involved in cell—
mediated immunity, such as contact dermatitis, and may partially
regulate the function of antibody—producing cells.
Chemicals may stimulate immune responses by several mecha-
nisms including acting as allergens or by stimulating production
of interferon. Chemicals may also suppress immune response; im—
munosuppressants result in an increased susceptibility to infec-
tion and may result in an increased susceptibility to some forms
of cancer.
GENETIC TOXICOLOGY
The integrity of genetic material (DNA) in all cells is crit-
ical to cell function and may be affected by some toxic agents.
Damage may take several forms: alteration in the chemical compo-
sition of DNA, change in the physical structure of DNA, or addi-
tion or deletion of chromosomes. Effects of genetic toxicity can
range from no observable effect to cancer. Genetic toxicity has
become a popular endpoint for toxicity testing because test re—
suits can be obtained relatively rapidly and inexpensively.
Genetic damage can occur by many mechanisms; the results are
generally classified in three groups: mutations, clastogenic
events, and aneuploidy. Mutagens are substances that change the
-------
A-b
chemical structure of DNA. Since DNA is “read TM to provide inf or-
rnation necessary for cell function and proliferation, mutations
may cause a misreading, leading to cell damage. Clastogens cause
a break in one or more strands of DNA and a physical rearrange-
ment of its parts. Depending on where the break occurs, clasto—
gens may affect cell proliferation or the production of cell
proteins. Aneuploidy is an addition or deletion of the number of
chromosomes; a commonly known aneupboidy is Down’s syndrome
(Mongolism) in which there is an extra chromosome. Aneuploidy is
often caused by chemicals that affect cell division.
Genetic toxicology is often considered with carcinogenicity
since many carcinogens are mutagens and testing for cnutagenicity
is easier than testing for carcinogenicity. Genetic toxicants,
however, can have many effects. Much of the DNA in cells is
quiescent. Since skin cells do not produce hemoglobin, there
will be little damage if instructions for producing hemoglobin
are damaged in a skin cell. Such events are called silent muta-
tions. Genetic damage can alter cell proteins and, therefore,
normal functioning of cells. Improper cell function may lead to
cell death or cancer. Finally, if the damage is in the reproduc-
tive system, genetic toxicants can cause reproductive failure or
abnormal offspring.
A variety of genetic toxicology tests have been developed in
recent years. Many are performed in vitro (outside the whole
animal——e.g., the Ames mutagenicity assay) and use cells grown in
liquids; some are performed in vivo (within the animal). These
tests are often referred to as short—term testing and require
less time, and therefore, less money. Typically, short—term
tests take days to months as contrasted with several years re-
quired for carcinogenicity testing.
------- |