Chemical Carcinogenesis
BEING MORE REALISTIC ABOUT CHEMICAL CARCINOGENESIS
First of a five-part series on cancer risk assessment
Chris F. Wilkinson
Department of Entomology
Cornell University
Ithaca, NY, 14853
Cancer is a special disease. Cancer of one type or another will eventually claim
the life of one in every four or five Americans; indeed, few will escape the
suffering of losing a friend or family member to the disease. Second only to
heart disease as a leading cause of death in the United States, cancer is
responsible for close to 500,000 deaths per year. Quite apart from its
importance as a factor in human mortality, the very thought of cancer arouses a
special dread in their minds of most individuals.
Public awareness and fear of cancer in the United States intensified during the
late 1960s as a result of widely publicized associations between human cancer
and a number of environmental factors associated with modern technology. In
particular, attention was focused on the possible carcinogenic risks associated
with the many products and byproducts of the chemical industry, and there was
enormous public pressure for prompt legislative action to regulate human
exposure to potential carcinogens.
As a result, legislation directed toward the protection of human health and the
environment has increased dramatically during the past two decades.
Approximately 30 such laws have been enacted, and although they differ in their
objectives and regulatory authority, most are designed to control the
carcinogenic and other health risks of chemicals introduced into commerce,
released into the environment, or encountered in the workplace.
Despite substantial progress in the past few years, basic understanding of
carcinogenic mechanisms and the major factors causing human cancer still leaves
much to be desired. There remains considerable uncertainty in current procedures
for identifying and regulating potential human carcinogens. Unfortunately, the
development of an acceptable policy for the regulation of chemical carcinogens
constitutes a particularly troublesome problem, because it seeks to match
argument based on uncertain science against the deep-seated human fear of
cancer. It is not surprising that the question of how to regulate cancer risks
in the United States has become a highly divisive issue in which the limited
amount of good science that is available either is not used to maximum advantage
or rapidly becomes lost in a tangle of emotions and subjective value judgments.
Human exposure
It is frequently alleged, and widely and uncritically accepted by a large
segment of the public, that the United States is currently in the midst of a
veritable explosion of human cancer; futhermore, we are told, the situation
continues to deteriorate. Although different statistical sources use different
methods of analysis and data presentation , and although selective analyses can
be used to support particular viewpoints, it is difficult to understand how the
data can be interpreted as being indicative of a cancer epidemic.
No one will argue that more people are dying from cancer each year; some 433,795
Americans died of cancer in 1982, a 56% increase over the 278,562 deaths that
occurred in 1962 (1). This is a sobering figure. However, when the trends are
adjusted for the increased size of the population and for changes in age
distribution the total cancer mortality rate from 1962 to 1982 has increased
8.7% (0.4%/year) and cancer incidence rate form 1973 to 1981 has increased 8.5%
(1.1%/year) (1).
When site-specific analyses are conducted, the most obvious trends that become
apparent during the past 35 years are the enormous increase in the incidence of
lung cancer (greater than 200%) and the marked decrease in the incidence of
stomach cancer; the incidence of cancers at most other sites has remained more
or less constant. Lung cancer is of such dominance that if the mortality with
which it is associated is excluded from the overall cancer mortality data over
the last 30 years or so, the 8% increase in mortality becomes a 13% decrease
(1). Indeed, when the effects of lung and skin cancer were excluded, a steady
decline appeared in overall cancer mortality over the last 50 years in people
under 65 years of age, according to a study by Doll and Peto (2). There is some
evidence that, as a result of underreporting in the past, age-adjusted
mortalities from many types of cancer (except lung cancer) have been declining
significantly for decades (3).
Consequently, the only real evidence of a cancer epidemic in the American
population (similar effects are also observed in other countries) is in relation
to lung cancer, which is now generally considered to result primarily from
smoking cigarettes (4). Sadly, this largely preventable disease is responsible
for almost 30% of all human cancer deaths in the United States; it has been the
major killer in males for some 20 years and now surpasses breast cancer as the
major cause of cancer mortality in women.
Causes of human cancer
Not very many years ago it was generally believed that cancer was caused by a
limited number of discrete chemical, physical, or biological (e.g., viruses)
agents. Today, cancer is recognized as a highly complex, multifactorial disease
caused, in part, by endogenous metabolic or other imbalances associated with age
or genetic makeup and, in part, by a wide variety of exogenous factors including
diet, lifestyle, and exposure to ionizing radiation and chemicals of natural or
man-made origin.
As a result of early epidemiological studies suggesting the predominance of
exogenous over hereditary factors in many human cancers, the search for causes
of cancer was focused on the physical environment (4). At a time of intense
public awareness of the potentially adverse impacts of technology on the
environment it was perhaps inevitable that human cancer would be attributed to
the many drugs, pesticides, plastics, food additives, and other materials
generated by the chemical industry. Unfortunately, despite the absence of
supporting data, a large segment of the public continues to believe that most
human cancers are directly associated with exposure to synthetic chemicals.
There is now general consensus that the personal and cultural habits of
individuals are the predominant determinants of human cancer (2,4,5). Thus
cigarettes smoking alone accounts for about 30% of all male cancer in the United
States, and other "bad" habits such as consumption of alcohol and sexual
promiscuity may cause an additional 10%. Diet is a highly variable factor, and
its importance can be expected to change markedly with geographic and ethnic
background. It is estimated that factors associated with diet are responsible
for about 35% of human cancer (2,5). Even in the most highly industrialized
countries, it appears that very few cancers can be attributed to exposure to
synthetic chemicals. Occupational exposure to a variety of chemicals or
industrial processes probably accounts for no more than 5% of human cancer, and
the total contribution of environmental pollution is estimated to be only 1-2%
(2).
Naturally occurring carcinogens
The fact that human cancer incidence has not changed significantly during the
last 50 years provides convincing evidence for the existence of a variety of
long-established cancer risk factors not associated with modern technology.
Consonant with the epidemiological association of cancer with diet, there is
increasing evidence that the human diet contains substantial amounts of a wide
variety of natural mutagens and carcinogens (6,7). Many of these -- such as the
hydrazine derivatives in mushroom species, pierine, and safrole in black pepper
and other plants, theobromine in cocoa and tea, chlorogenic acid in coffee, the
pyrrolizidine alkaloids in many species, and the potent mycotoxins such as
aflatoxin -- are established mutagens and animal carcinogens and often occur in
plants at concentrations of 2-10% by weight (6,7). Clearly, a multitude of other
naturally occurring materials of unknown toxicity remain to be isolated and
characterized.
Other potent carcinogens are produced as a result of cooking various foods.
These include the materials present in charred or browned protein and the
mutagenic pyrolysis product methyglyoxal, which is present in coffee (6-8).
It is important to consider our current obsession with identifying, regulating,
and generally worrying about what often appear to be trivial carcinogenic risks
associated with many synthetic chemicals against the truly overwhelming
background of natural carcinogens. The common belief that everything "natural"
is good is simply not true, and we must consider naturally occurring chemicals
as potentially important causative factors in human cancer. Indeed, in view of
the fact that our total daily intake of natural carcinogens could exceed our
intake of synthetic materials by as much as 10,000-fold (6,7), it is highly
unlikely that, for the general population, the combined carcinogenic effects of
all synthetic chemicals can ever be distinguished from the natural background.
What is a carcinogen?
By analogy with previous experience with infectious diseases, the initial
concept of a carcinogen was of some discrete physical, chemical, or biological
entity. This view was strengthened by early studies showing that, under certain
conditions, cancer could indeed be induced by exposing animals to single test
chemicals. Today, although most of the public continues to subscribe to the
concept of discrete causal agents of cancer, the scientist's view of what
constitutes a carcinogen has become far more complex.
Cancer is now considered to be the end result of a multistage process in which a
large number of endogenous and exogenous factors interact, simultaneously or in
sequence, to disrupt normal cell growth and division (9,10). Cancer, therefore,
is a complex disease that may involve a number of different mechanisms.
Consequently chemical carcinogenicity should not be considered as an inherent
property of a chemical but rather as an outcome of the interaction of a chemical
with a complex biological system influenced by many factors.
Traditionally, the development of cancer has been divided into two major stages,
initiation and promotion. Initiation describes the process whereby a chemical or
other agent damages the DNA of the cell, and promotion refers to the subsequent
progression and proliferation of the "transformed" cell through a variety of
pathological states (e.g., hyperplasia, neoplasia) leading eventually to a
malignant tumor (9,10). It is now recognized that initiation and promotion each
consist of several stages and may involve distinct mechanisms; some of these
stages are reversible and some are not, but probably all are susceptible to a
variety of modulating factors through which they may be enhanced or inhibited.
Initiation can result directly from a mutagenic effect of the chemical (or its
metabolite) on DNA, or indirectly from chronic cytotoxicity (resulting in cell
turnover and natural errors in cell replication), the activation of cellular
oncogenes, or other mechanisms. Initiation can be modulated by factors that
change the efficiency of DNA repair or immune surveillance, and, in the case of
chemicals that require metabolic activation, initiation will be affected by
factors that modify metabolism.
Because these and a large number of other physiological (e.g., age, sex, hormone
balance, nutritional status) and exogenous (e.g., stress, dietary fiber, and
fat) factors are often key determinants in the development of cancer, should
they be defined as carcinogens? Certainly an excess level of a natural hormone
can be just as important a causal agent in human cancer as a pesticide residue.
How to define carcinogen is not simply of academic importance; it has profound
conceptual and practical implications in relation to the development of a
reasonable national policy for the regulation of chemical carcinogens. It is
also of importance to a confused public that constantly is being barraged by
news of an ever-lengthening list of "carcinogens."
The International Agency for Research in Cancer (IARC) has defined chemical
carcinogenesis as "the induction by chemicals of neoplasms that are not usually
observed, the earlier induction by chemicals or neoplasms that are usually
observed, and the induction by chemicals of more neoplasms than are usually
found" (11). Although this is a useful operational definition, it does not
attempt to address the fundamental distinction between direct-acting carcinogens
and those acting indirectly through complex interactions with the test organism
(12).
Clearly, a classification of carcinogens based on their mechanisms of action
would be preferable, because, in some cases, this might provide an opportunity
to adopt a more appropriate regulatory approach. A tilt in this direction is
indicated by increasing use of terms like "genotoxic" and "nongenotoxic" (or
"epigenetic") carcinogens to distinguish chemicals capable of damaging DNA from
those apparently acting by other mechanisms (12,13). Unfortunately, current
bioassay procedures do not allow us to classify all carcinogens according to
their modes of action.
Although there is mounting evidence that some chemicals are acting through
nongenotoxic mechanisms for which thresholds or no-effect dose levels might be
anticipated, current regulatory policy requires that all are treated as though
they are genotoxic "complete" carcinogens. In 1983, a somewhat simplistic and
premature attempt by the EPA to place "epigenetic" carcinogens in a lower
regulatory risk category than those considered to be "genotoxic" was greeted by
such an uproar or dissent that it was quickly dropped from further consideration
(14). A more fruitful approach, recommended by the Office of Science and
Technology and others, evaluates each chemical on a case-by-case basis using all
available data to understand its mechanism of action (9,15,16).
The search for chemical carcinogens
Efforts to identify chemicals likely to pose a potential cancer threat to humans
have intensified in recent years and have relied mainly on the results of
chronic bioassays with animals, short-term in vitro tests for genotoxicity, and
epidemiological studies in human populations. Chronic bioassay with laboratory
animals, mainly rats and mice, remains the major and most practical experimental
procedure for identifying potential carcinogens. It is also a procedure beset by
many practical and theoretical uncertainties relating to both the design and
conduct of the test itself and the subsequent interpretation of the data.
Major limitations of chronic animal bioassays are that they are inherently
insensitive and highly variable in nature. Furthermore, there is always a great
deal of uncertainty associated with the fact that all such tests ultimately
require the extrapolation of data obtained under one set of conditions (i.e.,
with rodents exposed to very high doses in the laboratory) to predict those
likely to occur under an entirely different set of conditions (i.e., with humans
exposed to very low doses in the real world). Such dose and species
extrapolations are particularly troublesome in the generation of quantitative
estimates of human cancer risk.
As a result of the increasing reliance of U.S. regulators on precise numerical
estimates of theoretical, upper-bound, human cancer risk (e.g., 1 in a million
or, worse, 1.3 or 1.33 in a million) and the matter-of-fact way in which these
estimates are reported as real risks by the media, there are many who believe we
have exquisitely sensitive testing capabilities. Nothing could be further from
the truth. In a typical two-year rodent oncogenicity study utilizing a total of
about 600 animals, a cancer occurring at a frequency of 5 in every 1000 would
almost certainly go unnoticed. The practical implications of this are
considerable because a cancer frequency of 5 in 1000 translates into more than 1
million cases of cancer in the current U.S. population.
In attempts to increase the sensitivity of the animal bioassay, high exposure
levels at or approaching the maximum tolerated dose (MTD) are employed and,
indeed, are required by most regulatory guidelines (17,18). The importance of
the MTD in ensuring a successful outcome to the search for carcinogens is
illustrated by the fact that of a group of 52 chemicals judged positive in NTP
(National Toxicity Program) chronic bioassays, two-thirds would not have been so
classified had the high dose selected been one-half of the MTD actually used
(19). Does this increased "power of detection" really reflect true carcinogenic
potential, or is it a false positive resulting from cytotoxicity or
dose-dependent differences in metabolism and pharmacokinetics? Does it have any
relevance to assessing low-dose effects in any species?
Regulatory policy continues to cling to the concept that there is no finite
threshold below which carcinogens will not exert an effect. Consequently,
although the true shape of the dose-response curve at doses lower than those
actually used in the test is not known, low-dose effects can only be estimated
by extrapolation to zero of effects observed at high doses (9,18,20). It should
be noted that although chronic bioassays typically involve two or three doses
spanning perhaps one order of magnitude, extrapolations are often four, five, or
more orders of magnitude below the experimental range. Such extrapolations would
not even be attempted in most areas of science.
The question of how to extrapolate has led to the development of a number of
statistical models for the estimation of low-dose effects, and noisy debate over
which is most appropriate has overemphasized this aspect to the problem. It has
also led to the generation of precise mathematical risk estimates that are
simply not justified by the quality of the toxicology data from which they are
derived (21). The selection of the model to be used often dominates the results;
although most models are in general agreement in the range of experimentally
observed responses, they may provide estimates of low-dose responses that vary
to several orders of magnitude (16,22). New and more biologically realistic
models currently being developed will be described in future articles in this
series (22,23).
It should be noted that risk estimates prepared by federal agencies are
described as upper-bound estimates, not actual estimates, of risk. Real risks
are judged to be below the upper-bound values and could be as low as zero. This
is seldom made explicit in regulatory agency communications and is certainly not
understood by either the media or the general public.
There is no question that the "no threshold" concept for carcinogens -- and the
"zero tolerance," Delaney-type philosophy with which it is associated -- will
continue to cause endless grief as long as it is a part of regulatory policy.
Regardless of theoretical arguments for adopting a "no threshold" approach, it
seems clear from a practical standpoint that thresholds must exist. If this were
not the case the human race would have been long extinct as a result of exposure
to natural carcinogens. Surely our regulatory policy must be based on realism
rather than theory.
These are but some of the many sources of scientific uncertainty that severely
limit current capabilities of interpreting the results of chronic animal
bioassays for cancer. Others relate to properly identifying, quantifying, and
assessing the relevance to humans of a variety of animal tumor types (pathology)
and to evaluating the significance of tumors that occur spontaneously in several
strains of test animals. Some of these uncertainties will undoubtedly be
obviated as our understanding of the molecular biology of cancer and the
mechanisms of carcinogenesis continue to improve. Some will not, however, and
the need to extrapolate with respect to both dose and species will continue to
present serious problems in evaluating the relevance of animal tests data to
humans.
During the past two decades, a variety of short-term tests for genotoxicity have
been developed to augment chronic animal bioassays for carcinogenesis (24).
These include the well-known Ames Salmonella test for mutagenicity and several
other in vitro and in vivo assays based on a number of genotoxic end points such
as sister chromatid exchange, unscheduled DNA synethesis, and chromosome
aberrations.
Many of these short-term tests have been widely used, but it has become
increasingly apparent that there is often little correlation between the results
of these tests and those of the chronic bioassays (25). Furthermore, there is
considerable inconsistency among the results of different short-term tests
themselves. The reasons for the former undoubtedly relate to the fact that
carcinogenicity can occur through a variety of both genotoxic and nongenotoxic
mechanisms and that many of the short-term tests simply fail to replicate the
overall metabolic and pharmacokinetic conditions that exist in the intact test
species. Differences between the results of the short-term tests probably
reflect the different biological systems involved.
To avoid the problem of deciding which, if any, of the short-term tests is the
most useful, it has become common practice to evaluate genotoxicity on the basis
of the total weight of evidence from a battery of such tests.
Ashby has criticized this nonscientific approach (26) and has suggested a
common-sense strategy that would reduce the number of short-term tests required
to demonstrate genotoxic potential to just one or two in vitro tests (e.g.,
Ames) and one or two in vivo tests (e.g., mouse bone marrow micronucleus assay).
Doubt regarding the utility of many of the short-term tests for genotoxicity has
also been raised by a recent NTP study that compared the results of chronic
carcinogenesis bioassays and four short-term in vitro (not in vivo) tests for
each of a group of 73 chemicals evaluated in the NTP/NCI program (27). Major
conclusions from this rather limited study were
- That a positive result in the Ames test carries a high probability (70%) that a
chemical will be associated with carcinogenic activity,
- That use of a battery of short-term tests does not improve predictability above
that provided by Ames test,
- That a tier system does not add to the utility of the tests, and
- That the short-term tests correlated better with each other than with the
chronic bioassay.
Human carcinogens
The IARC has developed a qualitative classification scheme that divides
chemicals into groups with varying degrees of evidence for carcinogenicity in
humans (28). These groupings are based on the strength of positive evidence
available from animal studies, short-term tests, and human epidemiology, the
evidence in each category being related as sufficient, limited, or inadequate.
It is of interest that, of some 600 chemicals, chemical mixtures, or processes
evaluated by IARC expert working groups, only 23 chemicals and 7 processes are
considered causally associated with cancer in humans (Group 1). An additional 14
chemicals (Group 2A) are considered probably carcinogenic to humans. No new
human carcinogens have been identified by IARC during the past 10 years; with
improving tests procedures and decreasing levels of human exposure, it seems
unlikely that many more will be identified.
The IARC has not been able to establish strong links with human cancer for more
than a handful of chemicals, but it is equally true that very few chemicals
tested are given an unequivocally clean bill of health. Of the first 192
bioassays conducted by the NCI, 98 were judged positive, 91 were judged
suggestive or inconclusive, and only 3 were considered negative (29). The NCI
guidelines require tests to be conducted on both sexes for each of two rodent
species. A chemical is judged "positive" if it yields a positive response in one
well-conducted test (i.e., in one sex of one species). A "negative"
classification requires a uniformly negative response in each of the four tests.
If the test results are equivocal or if the tests themselves are considered to
have been inadequately designed or conducted, the conclusion with respect to
carcinogenic potential is usually suggestive or inconclusive. Unfortunately, no
matter how tenuous the data, these latter terms are frequently interpreted as
positive by nonscientists who are unable to understand that a negative can never
be demonstrated experimentally.
Regulatory goals and directions
The development of a sound regulatory policy that not only protects the public
against the potentially adverse health effects of chemicals but also creates the
incentive for industry to develop new materials of real benefit to society is,
indeed, a difficult task. It is particularly difficult when, as is almost always
the case today, the major focus of concern is cancer.
Despite the fact that there is no epidemic of human cancer in the United States
and despite the fact that, of the cancer that does occur, only a very small
percentage can be attributed to synthetic chemicals, we continue to pour
billions of dollars worth of time, effort, and resources into attempts to
identify carcinogens. Hindered by the scarcity and uncertainty of the science
and complicated by the inevitable involvement of policy and value judgments, the
results of such efforts are seldom clear-cut or cost-effective.
The current obsession for regulating carcinogenic risk in the United States
seems to be based more on the public's perception of risk and fear of cancer
than on risks that actually can be demonstrated. We are caught up in a vicious
circle in which, in attempting to respond to public pressure, regulators are
focusing on and identifying increasingly smaller risks that in turn further
alarm the public and create yet more pressure to regulate. We seem unable, in a
regulatory sense, to distinguish toxicological trivia from more clear-cut
problems, and as a society we spend our time worrying about cancer risks that
are orders of magnitude smaller than those risks most of us face driving to work
each day.
Surely the time has come to pause and take serious stock of our regulatory goals
and directions. We have limited resources, and we must concentrate these on
resolving real problems that require immediate attention. Despite the inherent
difficulties it entails, we must address the issue of what constitutes a
significant health risk; in developing policy, we must balance this against what
we as a nation can afford in terms of remedial action to reduce risks. We cannot
afford to go blindly along, throwing large amounts of money into attempts to
resolve imaginary problems. Instead, we must carefully identify and rank the
areas of real health concern and develop appropriate strategies by which the
associated risks can be avoided or minimized.
References
(1) Bailar, K.C. III; Smith, E.M. N. Engl. J. Med. 1986, 314, 1226-32.
(2) Doll, R.; Peto, R. The Causes of Cancer; Oxford University Press: New York,
1981.
(3) Gori, G.B.; Lynch , C.J. Reg. Toxicol. Pharmacol. 1986, 6(3), 261-73.
(4) Higginson, J.; Muir, C.S. J. Natl. Cancer Inst. 1979, 63 1291-98.
(5) Higginson, J. Environ. Mutagen. 1983, 5, 929-40.
(6) Ames, B.N. Science 1983, 221, 1256-64.
(7) Ames, B.N.; Magaw, R.; Gold, L.S. Science 1987, 236, 271-80.
(8) Sugimura, T.; Sato, S. Cancer Res. 1983, 43, 2415s.
(9) Office of Science and Technology Policy. "Chemical Carcinogens: A Review of
the Science and its Associated Principles"; Fed. Regist. 1985, 50, 10371-442.
(10) Farber, E. Cancer Res. 1984, 44, 5463-74.
(11) International Agency for Research on Cancer (IARC). "IARC Monographs on the
Evaluation of the Carcinogenic Risk of Chemicals to Humans"; IARC: Lyons,
France, 1982; Vol. 29, p. 16.
(12) Weisburger, J; Williams, G. In Toxicology: The Basic Science of Poisons,
2nd ed.; Doull, J.; Klaassen, C.D.; Amdur, M. O., Eds.; MacMillan: New York,
1980; pp. 84-138.
(13) Weisburger, J. Jpn. J. Cancer Res. 1985, 76, 1244-46.
(14) Marshall, E. Science 1983, 220, 36-37.
(15) Squire, R.A.. Science 1981, 214-877-80.
(16) Panel report; Science 1984, 225, 682-87.
(17) International Life Science Institute (ILSI). In Current Issues in
Toxicology; Grice, H.C., Ed.; Springer-Verlag: New York, 1984; pp. 9-49.
(18) Environmental Protection Agency (EPA). "Guidelines for Carcinogen Risk
Assessment"; Fed. Regist. 1986, 51, 33993-34014.
(19) Haseman, J. K. Fund. Appl. Toxicol. 1985, 5, 66-78.
(20) National Toxicology Program (NTP). "Report of the NTP Ad Hoc Panel on
Chemical Carcinogenesis Testing and Evaluation"; U.S. Dept. of Health and Human
Services: Washington, D.C. 1984.
(21) Krewski, D.; Van Ryzin, J. In Statistics and Related Topics; Csorgo, M. et
al., Eds.; Elsevier/North Holland: Amsterdam, 1981; pp. 201-31.
(22) Menzel, D. ES&T, in press.
(23) Sielken, R. ES&T, in press.
(24) Weisburger, J.H.; Williams, G. In Chemical Carcinogens; 2nd ed.; Searle,
C.E., Ed.; American Chemical Society: Washington, D.C., 1984; Vol. 2, pp.
1323-73.
(25) Shelby, M.D.; Stasiewicz, S. Environ. Mutgen. 1984, 6, 871-76.
(26) Ashby, J. Mutagenesis 1986, 1, 3-16.
(27) Tennant, R..W. et al. Science 1987, 236, 933-41.
(28) International Agency for Research on Cancer (IARC). "IARC Monographs on the
Evaluation of the Carcinogenic Risk of Chemicals to Humans"; IARC: Lyons,
France, 1982; Supplement 4, pp. 7-23.
(29) Hottendorf, G.H.; Pachter, J. Toxicol. Pathol. 1982, 10, 22-26.