Risk Assessment: What It Is, What It Can (and Can't) Do

This document discusses risk assessment and sets the discussion in the context of RISK*ASSISTANT™ for Windows, software that allows users to estimate health risks from chemicals in the environment in particular settings. (This overview appears in the RISK*ASSISTANT manual.) The principles and issues described, however, pertain to how risk assessment may be conducted -- and its results applied -- regardless of the tools used.

How Risk Contributes to Environmental Decision-Making

What Risks to Measure?

What Is an Adverse Effect?

Risk Assessment and the Real World

Uncertainties in Dose-Effect Relationships

Uncertainties in Exposure Relationships

Risk-Benefit Analyses

RISK*ASSISTANT and the Issues of Risk Assessment

Information on the capabilities and availability of RISK*ASSISTANT -- developed by Hampshire Research Institute -- is available from Thistle Publishing at http://www.ThistlePublish ing.com.

HOME

How Risk Contributes to Environmental Decision-Making

A health risk is the likelihood that one or more adverse effects will occur to a specific individual or group of individuals in a particular chemical-exposure situation. An environmental risk (not covered by the current version of RISK*ASSISTANT) is the likelihood that one or more adverse effects will occur to a species, to an ecosystem, or to part of the physical environment that is critical for the health of these systems (e.g., ozone in the stratosphere). View brief illustration.

Risk is determined by:

1. the inherent ability of the chemical to cause different adverse effects, qualitatively and quantitatively, at different amounts of exposure to the chemical (Chemical Hazards) and
2. the amount of exposure (Exposures) that occurs.

Usually, but not always, a higher exposure leads to more serious effects or to a greater likelihood that they will occur, so that the risks are higher. Conversely, at some low exposure level, the risks become insignificant or nonexistent.

Knowing the human health and environmental risks from toxic chemicals in the environment is important for many reasons. For example, knowledge of risks informs us about:

Estimates of risk are therefore essential for some environmental decisions, especially those involving protective and clean-up standards for contaminants already in the environment. Even then, risk estimates are but one factor for sound decision-making. Only in extreme cases will risks alone drive environmental decision-making. A zero risk compels no action. A great risk may compel immediate action.

In most cases, however, risk is only one of many factors that enter into environmental decisions. Other factors include:

Moreover, as this wide range of critical factors makes evident, risk is a poor substitute for prevention for decisions about toxic chemicals in waste or in products.

RISK*ASSISTANT is risk assessment, not environmental decision-making, software. Because risk and risk assessment are not precise scientific concepts and never will be, it is important to keep the use of RISK*ASSISTANT in perspective. Descriptions of some important issues with risk and risk assessment follow.

Return to top of Overview

What Risks to Measure?

Risks can only be estimated, not measured. But even if they could be measured, there is no consensus on what to measure.

RISK*ASSISTANT generates two types of risk estimate:

RISK*ASSISTANT uses these risk measures because they are the ones most often used in U.S. Environmental Protection Agency (EPA) risk assessments. They are by no means the only possible measures of health risk, however. Other ways to assess the same adverse health effect include:

The perspectives gained from different risk measures, for the same adverse effect, can vary greatly. For example, several measures that weigh risk for society as a whole may imply a small concern, but if they result from very high individual risk to a small number of people, that may justify a much greater concern.

Return to top of Overview

What Is an Adverse Effect?

A strong social consensus recognizes that certain health effects are adverse. Many of these are explicitly identified in state and federal environmental legislation. The federal Toxic Substances Control Act of 1976, for example, names these effects as priorities: "serious or widespread harm to human beings from cancer, gene mutations, or birth defects..." (Section 4(f)).

Other laws expand this list somewhat. Clearly, imminent death is inarguably an adverse effect. But no laws indicate whether any of the following (or hundreds of other) effects are "adverse":

Moreover, the social consensus changes over time as the scientific community develops more information and as environmental regulatory agencies enter into public dialogues on proposed actions based on identified effects.

Return to top of Overview

Risk Assessment and the Real World

Risks from environmental chemicals cannot be measured directly, but only estimated. Real risks can only be estimated after damage has occurred. The risk assessment process seeks to predict risks by evaluating:

1. the inherent ability of the chemical to cause various adverse effects, qualitatively and quantitatively, at different amounts of exposure to the chemical (estimated Chemical Hazards) and
2. the amount of exposure (estimated Exposures) that occurs.

View illustration.

In addition to the many issues associated with risk, risk assessments inevitably incorporate uncertainties, often extremely large, in their estimates of exposure and risk. Most of these uncertainties arise because scientific measurements needed to determine risk precisely cannot be made in the actual individuals, exposure setting, and time frame of interest. Alternatives are used, from which answers must be extrapolated to the setting of interest. The magnitude of the uncertainties from such extrapolations is typically unknown. Thus, when any risk assessment is complete, no one knows how close its estimates are to the exposures and risks present in the real world. RISK*ASSISTANT cannot quantify such uncertainties, although it estimates the uncertainty associated with some parts of the process. It is critical that the extent of uncertainties in risk assessment not be forgotten when the results are used.

Return to top of Overview

Uncertainties in Dose-Effect Relationships

A measure of chemical hazard (the ability of a chemical to cause adverse effects) is the dose-effect relationship: the relationship between the amount of a chemical received in an exposure (dose) and the nature and/or severity of the resulting toxic response (effect). {In contrast, dose-response relationship is the relationship between the amount of a chemical received in an exposure (dose) and the percentage of the exposed group that shows a pre-defined effect from that dose (response).}

Dose-effect data on toxic chemicals usually come from laboratory experiments or from analysis of the effects in humans who have been exposed in a non-experimental setting (e.g., accidents or workplace exposures). Most laboratory data come from studying the effects of exposing nonhuman species to different amounts of the chemical over time. In some cases, inferences are made based on the effects observed in bacteria and other microscopic species or in human cells grown in the laboratory. Human data generally come from doctors' reports of treating victims of unintended exposures or from scientists (epidemiologists) studying patterns of effects in exposed populations.

These dose-effect data raise two major categories of uncertainties:

1. How does the reported dose-effect relationship compare to that for the actual exposed population in any risk assessment?

A rat, mouse, or cell is clearly not a human being. Moreover, only limited data are available upon which to base well-informed extrapolations from different animal species to humans (although this situation is slowly improving). Further, most local populations are diverse in their response to chemicals because of differing genetic make-up, age, habits, occupations, health status, diet, etc. In contrast, many laboratory species are homogeneous, in order to elicit reproducible results and to limit variability in response. Even a worker population is typically healthier and more uniform than the average local community exposed to environmental chemicals.

Extrapolations from animals to humans may often be more reliable than those from previously exposed humans. Many confounding variables are associated with "uncontrolled" humans and their exposures, in contrast to the well-characterized and controlled conditions that can be created in laboratory studies. (Human experimentation is not done with suspected toxic materials for ethical reasons.)

2. Typically, relatively high exposures are used experimentally, to cause statistically significant effects. Accidents and workplace conditions, in human-based reports, also often involve high exposures. How do the effects reported in those contexts compare with effects at the much lower doses that typify human exposures in environmental settings of interest in risk assessment?

When no effect is observed in laboratory animals or in a limited number of exposed people, is there no risk of human effects at such a level? Or is there a risk lower than can be seen in a small group but potentially significant for a large exposed human population? For example, a 1% incidence of any disease would be impossible to detect in a study of 25 animals, but would represent more than 2 million cases if the entire U.S. population were exposed. What assumptions should be made about dose-effect relationships at exposure levels where effects cannot be excluded but are unmeasurable? The answer is we don't know. In the absence of direct data, scientists attempt to predict the mechanisms that cause the observed adverse effects and to estimate from those mechanisms what happens at these lower doses.

In the absence of evidence to the contrary, risk assessors (including those in EPA) will most often assume:

1. for cancer and other genetic effects: a. there is no dose below which one can be sure that no adverse effects occur, and b. at low doses, the relationship between dose and probability of effect is directly proportional (linear),
2. for most other health effects, there is some dose below which the likelihood of adverse effects is essentially zero.

Thus, for each chemical, RISK*ASSISTANT makes use of available hazard data on (1) the predicted proportional relationship for cancer effects at low doses and/or (2) a reference dose below which non-cancer adverse effects are not expected.

In addition, users can provide their own hazard data.

Return to top of Overview

Uncertainties in Exposure Assessments

Many similar uncertainties complicate the comparison of real world exposures to those that can be predicted from limited knowledge of local conditions and are used to make exposure estimates. The magnitude of many of these uncertainties is equally unknown.

Further, local monitoring data often do not exist. Even when they do, they may not represent well the quantity and quality of local exposures. In some cases, data are available only on releases of chemicals to the environment, not on chemical concentrations in the environment. Moreover, existing data measure past or current exposures that may or may not accurately represent the future. In these circumstances, models are often used as substitutes for direct monitoring to predict exposure.

Developing assessments that predict local exposures is at the heart of what RISK*ASSISTANT does. RISK*ASSISTANT strengthens the risk assessor's ability to address these uncertainties. Its broad flexibility for incorporating data on chemical concentrations in the environment enables the user to consider a range of possible exposure cases. It also includes powerful sensitivity analysis capabilities, to test immediately the impact of different assumptions on exposures and risks.

RISK*ASSISTANT incorporates a model for predicting long-term environmental concentrations from data on releases to air (Win-ISC2). With local data on air emissions, this model allows users to project contaminant concentrations at specific points in the vicinity of local sources. Similarly, RISK*ASSISTANT incorporates a model for calculating concentrations at points downstream of a discharge of toxic chemicals on any U.S. river or stream. RISK*ASSISTANT also accepts the results of other models that users may choose.

Again, it is critical to keep in mind the extent of uncertainties in exposure assessment. Also, some models are better than others, and the assumptions used in any model about local exposure conditions may be well informed or far off-base, greatly affecting the results. Responsible users will take these considerations into account. View illustration.

Return to top of Overview

Risk-Benefit Analysis

A major use of health risk assessments has been as a component of risk-benefit analyses. RISK*ASSISTANT will help supply a risk component for such an analysis, but it is not a risk-benefit system.

Risk-benefit analyses require not only an assessment of existing risks, but predictions of future risks under several different environmental management assumptions. Some of these differing assumptions will reflect various environmental quality standards or levels of contamination. For these, estimating the change in risk may be relatively straightforward. If, on the other hand, the environmental management option is a pollution control standard or product restriction, estimating the change in risk will be much harder, because the results depend on predicting how the world will respond to the environmental management action. Typically, government and regulated industries have had very limited success in predicting such responses.

Risk-benefit analyses also require an assessment of the economic and other consequences of alternative risk management decisions. (RISK*ASSISTANT does not address economic and non-health risk consequences.) Economic impact and other regulatory impact assessments often have as great or greater uncertainties than those in risk assessment, and they rely on the same ability to predict the world's response to such decisions.

Return to top of Overview

RISK*ASSISTANT and the Issues of Risk Assessment

RISK*ASSISTANT lets you see the impact of different assumptions in your exposure and risk assessments, but it cannot eliminate the large uncertainties in such assessments. RISK*ASSISTANT cannot establish that risk is indeed the issue, nor does it address non-risk elements such as the economic costs of adverse health effects. No software can assess all factors relevant to a local situation, but RISK*ASSISTANT can and does use standard approaches to generate estimates of exposure and risk using detailed, locally relevant information and test them against many alternative assumptions. View illustration.

Return to top of Overview

HOME

If you have questions about the Hampshire Research Institute or RISK*ASSISTANT™, send e-mail to info@hampshire.org.