As commonly practiced today, risk assessment and risk
management consider exposures and risks in isolation from one
another, typically chemical-by-chemical. For example, risks
associated with air pollution are not put into the context of
concurrent risks associated with contaminated drinking water or
foodborne pesticide contamination. That fragmented approach to
risk characterization is mostly a result of the fragmentation of
responsibilities of different regulatory agencies and programs,
but it can also be attributed to the limitations in our knowledge
of the interdependence of different risks. Failure to account for
multiple and cumulative exposures is one of the primary flaws of
current risk assessment and risk management, according to
testimony received by the Commission from Michael McCloskey,
chairman of the Sierra Club, and others. Many people are
surprised to learn that scientists usually do not test mixtures
and that risk assessors and managers do not even try to account
for the full array of exposures and health (or ecologic) risks.
If the Commission's risk-management framework is implemented and
experience with testing and evaluating multiple chemical risks
increases, it should be feasible to move beyond fragmentation.
FINDING 3.4: Humans are exposed to many
chemicals and other potentially toxic agents in the environment,
but toxicity testing and regulations generally focus on one
chemical at a time, often just in air, water, or food. Most risk
assessments evaluate the toxicity of or risk associated with
individual chemicals and then combine them by simple addition to
estimate risk related to chemical mixtures. However, adding risks
ignores potential synergistic or antagonistic interactions that
could lead to underestimation or overestimation of total risk,
respectively. Knowledge of mechanisms of action can guide
judgments of whether risks related to combinations of particular
chemicals will be additive or independent.
RECOMMENDATION: Toxicity testing of complex
environmental mixtures of regulatory importance should be
performed for hazard identification and to generate comparative
potency estimates of human risk. For risk assessments involving
multiple chemical exposures at low concentrations, without
information on mechanisms, risks should be added. If the
chemicals act through separate mechanisms, their attendant risks
should not be added, they should be considered separately.
RATIONALE
Many complex mixtures--such as automobile exhaust, cigarette
smoke, and other combustion products--have hundreds or thousands
of chemical components. Attempting to identify and characterize
each component and then adding their risks is clearly
impractical. In those cases, the toxicity of the mixtures
themselves can be tested and their risks characterized on that
basis. For example, toxicity studies of diesel exhaust and other
emissions have been conducted by the Health Effects Institute,
jointly supported by EPA and motor vehicle manufacturers. The
valuable results of those studies and others, such as tests of
smoggy air from the Los Angeles basin, encourage the Commission
to recommend the testing of other important chemical mixtures.
Predicting a complex mixture's toxicity or risk can be
assisted by testing it in bioassay systems and comparing the
results with those of tests of similar mixtures of known toxicity
or risk. Bioassays that might be useful for testing mixtures
could range from mutation tests in microorganisms to evaluation
of effects on organs in culture or short-term tests of rodent
respiratory function. A database of methods, bioassays, and
biologic markers of effect and knowledge of the behavior of known
mixtures in those bioassays will be needed to facilitate risk
predictions for environmental mixtures. Such whole-mixture
testing could be considerably less expensive to perform than
routine monitoring for over 100 drinking-water contaminants, for
example, and might provide results that can be more easily
extrapolated to human toxicity and discussed with stakeholders.
The index of biotic integrity (see section 3.5) is another
example of the use of a bioassay to integrate effects of numerous
chemical exposures.
The experimental and epidemiologic database available for
generating estimates of comparative potency of mixtures is not
large. Most work has been applied to predicting lung-cancer
risks. For example, epidemiologic data are available on the
carcinogenic potencies of coke-oven emissions, coal roofing tar,
coal smoke, aluminum smelters, and cigarette smoke. The human
cancer risks of those emissions have been characterized and
compared with their potencies in experimental systems to estimate
the risks associated with mixtures that lack epidemiologic data,
including automotive emissions (diesel and gasoline), woodstove
emissions, residential oil-furnace emissions, and ambient air
particles; it is assumed that the relative carcinogenic potencies
observed in experiments would be similar for humans (Harris 1983,
Lewtas 1993).
Complex mixtures seemingly from the same source can vary
considerably. For example, neither automobile engines nor
gasolines are identical, so automobile exhaust is likely to vary
substantially among sources and over time. The composition of air
pollution varies with time of day and time of year, not to
mention geographic location and source, so the toxicity of such
mixtures is likely to vary considerably. Probabilistic approaches
to describing the variability of composition within a class of
mixtures and the relationship between that variability and
toxicity should be explored.
Most of the information that is available on interactions
among chemicals comes from human occupational studies and from
rodent bioassays. Those studies generally evaluate doses that are
much higher than the low, environmental doses commonly
encountered. Dose is important because interactive effects
(either synergistic or antagonistic) depend heavily on dose;
therefore, characterizing interactions that occur at one set of
doses (such as those used in a rodent bioassay) is likely to
provide very little information about interactions at very
different doses (such as those generally encountered in the
environment). "High" doses for combined effects are
defined as those at which statistically significant increases in,
for example, cancer incidence, are observed in either laboratory
or occupational studies. For the most part, exposure to chemical
mixtures in the environment occurs at "low"
doses--typically, one-thousandth (or less) of the doses at which
toxicity is observable in rodent bioassays or in epidemiologic
studies of highly exposed workers. The difference between
exposures observed to cause adverse effects and actual
environmental exposures is called the margin of exposure (EPA
1996) (see section 3.1.1).
The combined effects of exposure to chemicals in a mixture are
determined by how individual components of the mixture affect the
biological processes involved in toxicity. Components of a
mixture can affect biological processes in many ways. For
example, anything that affects the absorption, distribution,
metabolism, or elimination of a chemical will affect the amount
of that chemical that is available to react with DNA or other
cellular targets. Because all chemical-biological interactions
are the result of reactions of many molecules at many cellular
sites, a mathematical dose-response model of a response that
depends on such mechanisms would have to be nonlinear at low
doses. Such logic strongly suggests that any disease process that
depends on such interactions is only marginally important in
small exposures. Only at high doses of one or more mixture
components--such as cigarette smoke, alcohol, and some substances
in occupational exposures--is the combined effect likely to be
greater than the sum of the individual effects. For example,
occupational exposure to asbestos is associated with a mortality
ratio for lung cancer of 5 (that is, in comparison to persons not
occupationally exposed to asbestos) and smoking with a mortality
ratio for lung cancer of 10; but asbestos workers who smoke have
a mortality ratio for lung cancer of 50, not 15. The risk of
liver cancer risk associated with aflatoxin is increased markedly
by hepatitis B virus infection.
The National Academy of Sciences report Complex Mixtures
(NRC 1988) also concluded that effects of exposures to agents
with low response rates usually appear to be additive. The
experimental evidence that can be used to infer effects at low
doses appears to support the assumption that low-dose additivity
does not underestimate, and in most cases probably overestimates,
risk (see, for example, Ikeda 1988). When the individual
components of a chemical mixture exhibit different kinds of
toxicity or have different biological mechanisms of toxicity,
they do not interact--they act independently at low doses. In
that case, risk is not equal to the sum of the individual risks,
each risk should be considered independently. Experiments have
shown that when groups of unrelated chemicals with unrelated
targets of toxicity were administered to rodents simultaneously
at doses equal to their separate NOAELs
(no-observed-adverse-effect levels), no effects were observed;
each chemical acted independently, not additively or
synergistically (Jonker et al. 1990, Groten et al. 1994). The
same is true of groups of chemicals with the same target but
different mechanisms of action (Jonker et al. 1993). Studies in
which similar chemicals with similar mechanisms and targets were
administered simultaneously indicate that antagonism, not
additivity or synergism, is the usual outcome (Falk and Kotin
1964, Schmähl et al. 1977), thereby reducing the effect expected
from even a single one of those chemicals.