PRINCIPLES OF TOXICOLOGY

462 RISK ASSESSMENT
characterization of potential risks posed by the chemicals but also in developing effective strategies
for managing the risks.
A particularly important aspect of the risk characterization is a discussion of the uncertainties
associated with the risk estimates. Each individual step in the risk assessment process is a potential
source of uncertainty. Many of these are discussed throughout the chapter, and include uncertainty
associated with estimating exposure (e.g., measurement errors, uncertainty in selecting the best
exposure models, uncertainty regarding exposure conditions that will exist in the future) as well as
determining safe levels of exposure (e.g., uncertainty regarding the shape of the dose–response
relationship in the low-dose region and extrapolating results from animals to humans, uncertainty that
the most sensitive health effect has been identified, uncertainty regarding ways that multiple chemicals
might interact). These need to be articulated in the risk characterization so that an appreciation of the
level of confidence and conservatism in the risk estimate can be gained. Without a discussion of
uncertainty, risk assessment results are often perceived as being more precise than they really are,
which could lead to misuse. Minimally, uncertainties should be discussed qualitatively, identifying the
source or nature of each uncertainty and how, in a general way, it could affect the risk estimation (i.e.,
whether the approach taken, in view of the uncertainty, is likely to contribute to an over- or
underestimation of risk). A semiquantitative perspective is helpful, in which the implications of model
and assumption choices on the risk estimate are described in rough, order-of-magnitude terms. As
discussed in Section 18.6, probabilistic techniques can be used to provide more precise quantitative
expression of the uncertainties associated with risk estimates, as well as a description of variability in
risks encountered in exposed populations. This approach is attractive in that it offers a much richer
characterization of the risks and uncertainties than do the more traditional risk estimation techniques.
It is technically demanding, however, requiring much greater time, resources, data, and technical
expertise.
18.6 PROBABILISTIC VERSUS DETERMINISTIC RISK ASSESSMENTS
Risk assessments are rarely performed for a single individual. Instead, risk assessments are designed
primarily to characterize risks to populations of individuals. Many—perhaps most—of the factors that
affect risk can vary from one person to another. Differences in body weight, inhalation rate, frequency
and duration of contact with contaminated media, and even sensitivity to toxicity, are examples of
factors that can lead to different risks among individuals, even if the concentration of chemical to which
they are all exposed is the same. Theoretically, there is no single risk for a particular exposure
circumstance, but rather as many different risk values as there are individuals in the exposed population.
In confronting the issue of variability in risk assessment, the traditional approach used for regulatory
purposes has been to simply characterize the risk to individuals within the population with the greatest
exposure. A deterministic approach to risk calculation is used, where a single value is selected for each
exposure variable and a single risk estimate is produced. The exposure assumptions are chosen to
represent the plausible upper bound of exposure, and the risk estimate is said to be associated with
reasonable maximal exposure (RME) or high-end exposure. The development of a single, high-end
risk estimate for regulatory use is consistent with the goal of regulatory agencies to develop risk
management strategies protective of the entire population. For perspective, a deterministic approach
may also be used to develop an estimate of risk for the average individual in the population, that is, a
central tendency estimate of the risk. The problem with this approach is that it provides little
information on the extent to which risk varies within the population. For example, while exposure
values may be selected to develop a high-end estimate of risk, it is difficult to know whether this value
is merely conservative or extreme. Does it represent an exposure that might be exceeded by 1 or 2 out
of every 10 individuals, or does it represent an exposure circumstance so extreme that it is unlikely
ever to take place? This vagueness regarding the degree of conservatism in deterministic risk estimates
undermines their value and creates controversy regarding their use in regulatory decision making.