H.S. 469: Risk Analysis
MODULES
|
A. Introduction |
|
B. |
|
C. |
|
D. |
|
E. |
Modules
Objective: to provide a model for analyzing the various risks associated with environmental health.
Risk analysis is a broad term that represents a collection of approaches and disciplines devoted to all aspects of risk issues. At a minimum, risk analysis includes 1) risk assessment, 2) risk communication, and 3) risk management (all defined below).
Risk assessment is the characterization of adverse effects from exposure to hazards. Probably the simplest example of this characterization is to say "the risk of cancer from a lifetime of exposure to "chemical X" is greater than one out of a million." More formally, risk assessment includes four steps defined below: hazard identification, dose response assessent, exposure assessment, and risk characterization.
Hazard identification is to determine whether a particular agent is causally linked to particular health effects. For example, does this chemical cause cancer?
Dose-Response Assessment is to determine the relationship between the magnitude of exposure and the probability of occurrence of health effects in question. For example, one ounce of "chemical X" will kill 50% of laboratory mice.
Exposure Assessment is to determine the extent of human exposure (this is especially useful both before or after the application of regulatory controls). For example, after the Clean Air Act revisions have been put into place, the exposure to the average citizen to "chemical X" is 50% of the allowable standard.
Risk characterization is to describe the nature and often the magnitude of human risk, including attendant uncertainty. For example, "chemical X" may cause cancer deaths in anywhere from 3 to100 people in Los Angeles over the next 20 years.
Risk communication is an interactive exchange of information and opinions among individuals, groups, and institutions regarding risk.
Risk management is the evaluation, selection, and implementation of alternative risk control actions.
Assignment: consider a current health issue using this model.
Send your assignment (and read others) to: hypernews
Test your knowledge with a: quiz
For more information, try: Introduction
Chapter 1: Defining Risk Analysis
1.1 Introduction: what is risk?
A logical introduction to an introductory text should include a basic definition of the field. However, in the case of risk analysis, this task is more difficult than you might expect. The literature abounds with various and often conflicting definitions. If we consider some of these conflicts, we may gain a better insight on the nature and dynamics of this field.
Accordingly, our first task is to define the term "risk." Webster's dictionary gives us a fair picture when it defines risk as "the possibility of suffering harm or loss." In the first part of this definition, "possibility" suggests the work of statisticians. In the last part of this definition, "harm or loss" suggests the work of health professionals (if we took a broader view, we could also include engineers, public policy analysts, ecologists, and various other scientists). Perhaps most interesting is Webster's use of the word "suffering." Note that we could drop this term from the definition and still have a workable guide (i.e., "the possibility of harm or loss.") However, its inclusion suggests the work of psychologists, sociologists, or medical practitioners: how else would we measure "suffering"? Furthermore, perhaps a few lawyers would have a few things to argue about how to measure suffering!
In any event, the underlying message here should be clear: the study of "risk" should always be multidisciplinary. Even the simplest definition shows us the formidable multiple challenges we face.
In 1985, the National Academy of Sciences assembled a group to provide a more substantial definition of the field, and my paraphrase of their definition is: the probability and magnitude of a hazard. A closer look at this definition shows it to be consistent with Webster's dictionary, but it uses a language more in keeping with the working professionals in the field.
An interesting counter-definition is provided by Professor Douglas Brown-Crawford of the University of North Carolina: "a rational belief that an event might possibly occur in light of existing evidence." What makes this interesting is that the definition begins with a more psychological bent to the field. As we shall see, this is more in keeping with the risk communication literature. Equally interesting is that "existing evidence" sounds as much like a legal as a scientific perspective. Brown-Crawford challenges us with the old notion that witch doctors are quite rational if you grant them their assumptions! Risk analysis, then rests on assumptions which may later prove to be false. Alas, such is true with all of science!
Rather than using this definition to dismiss all of risk analysis, it provides us with a critical theme that will be repeated throughout this text: we must be clear and explicit on the assumptions in a risk analysis. One final exercise can help reveal additional insights into how use the term "risk." By using a thesaurus, we can arrive at interesting synonyms for this term. For example, synonyms that derive from the term risk are: possibility --> possible --> capable --> ability --> power --> skill or talent !
As used by the insurance industry, risk is the amount an insurance company stands to lose (especially in dollars). Ultimately, it would appear that risk is tied to money, power, and prestige. Other synonyms include: luck, fortune, play, and even dare! Again, the warning should be clear, the true meaning of risk depends on who is using the term.
1.2 Analysis
After considering the definition of risk, we can now make a better effort at defining risk analysis: a collection of approaches and disciplines devoted to all aspects of risk issues.
This definition is deliberately broad and open, because the emergence of new scientific disciplines are sure to bring new insights to the nature of risk. As this term is used by most professionals and within the Society for Risk Analysis, the field includes risk assessment, risk communication, and risk management. Thus, our task must now be extended to include these subsets of risk analysis.
Risk assessment is defined as: the characterization of adverse effects from exposure to hazards. Characterization is intended to include, at a minimum: probability, uncertainties, analytic techniques, and models.
Risk communication is defined as: an interactive exchange of information and opinions among individuals, groups, and institutions regarding risk.
Risk management is defined as: the evaluation, selection, and implementation of alternative risk control actions.
1.3 Risk Assessment
The four major segments of risk assessment are defined below.
1. Hazard identification: to determine whether a particular agent is causally linked to particular health effects.
2. Dose-Response Assessment: to determine the relation between the magnitude of exposure and the probability of occurrence of health effects in question.
3. Exposure assessment: to determine the extent of human exposure, often before and after applying regulatory controls or other changes.
4. Risk characterization: to describe the nature and often the magnitude of human risk, including attendant uncertainty.
Given the broad range of disciplines involved in examining risks, it should be clear that this text cannot replace more substantial studies in such areas as statistics, toxicology, and communication. Nevertheless, it can provide a map for interpreting the developments in these respective fields. For example, epidemiology is "the study of the distribution and determinants of disease in humans." As such, epidemiology is an input to risk analysis in two important respects. First, its results can be compared to toxicological tests (often in non-human test animals) to corroborate final conclusions about risk. Second, epidemiology as a field may be less inclined to make policy recommendation, while one of the major purposes of risk analysis is to guide public or corporate policy.
Thus, risk analysis does more than simply predict risks. It should also compare risks; it should compare measures of risks; it should compare methods for measuring risks; and its ultimate purpose is to inform policy.
1.4 More Definitions
Throughout this text, the following terms will follow their traditional definitions. It should be noted that these definitions are not universal, and clarification may be required in how these terms are used within a give risk assessment.
A risk group is the group for which a risk assessment is being conducted.
An experimental group is a study group (toxicological or epidemiological) used to ascertain risk to the risk group.
Nonzero threshold toxicants have a NOEL (no observed effect level).
Zero threshold toxicants do not have a NOEL (carcinogens?).
An experimental dose range includes the dose in the experimental group. If this dose is higher than risk group,
we must extrapolate to the risk group.
1.5 Risk Communication and Risk Management
There will always be limitations of risk assessment to policy decisions. First, risk assessment is "only one aspect" of policy decisions. Second, risk assessments must always acknowledge the "judgments" that are part of its conclusions. Risk assessment typically provides the following:
1. characterize the expected health effects
2. estimate the probability (risk) of health effects
3. estimate the number of cases
4. suggest an acceptable concentration from the standpoint of risk
However, the outputs of a risk assessment are for regulatory decisions. The major impetus for risk analysis is federal legislation and executive orders. Some measures call for controlling "unreasonable risk," others specify "de minimus non curat lex" ("the law is not concerned with trivial matters").
A major influence on de minimus risk cam in the Supreme Court decisions on the regulation of benzene. The court ruled that de minimus risk was somewhere "between 1 in billion and 1 in a thousand." Some agencies have interpreted this to mean 1 in a thousand, while others have essentially split the difference (1 in a million).
The court's decision leaves unanswered a variety of questions. For example, if you add all our involuntary exposures, 100 chemicals would yield a risk of 1 in 10,000. Furthermore, if we compare these risks to voluntary activities, there is a major discrepancy: parachuting has a risk of 4 in 100, professional stunt men a risk of 2 in 10, and climbing Mount Everest a risk of over 1 in 10. At what point an activity becomes voluntary is unclear.
1.6 Types of statistical risk
Given in diversity of issues in interpreting risk assessments, it is instructive to consider the various fundamental types of error in statistical measures. The purpose of this consideration is not simply to dismiss the validity of any risk assessment, but to better interpret the results from a policy perspective.
We begin with the most classic consideration of statistical error, the type 1 (or alpha) error. This is the false rejection of a null hypothesis. Classic statistical theory holds that a null hypothesis states there is no difference between anexperimental group and a control group. The burden of proof is on the scientific investigator to show that there is a difference between these two groups, thus establishing a significant effect. In the language of statistics, we must reject the null hypothesisin order to establish that an effect is statistically significant. Thus, the rejection of a null hypothesis results in acceptance of scientific theory. Since scientists do not want to accept false theories (which would threaten the very integrity of science), our first concern is to minimize the chances of rejecting a null hypothesis and later discovering that the null hypothesis was true. In other words, accepted theories should pass the strictest standards, thereby preserving the integrity of science. In the language of risk, type 1 error tells us that we must prove something is unsafe. This is contrasted with type 2 error, defined as the false acceptance of the null hypothesis. Consider the consequences of type 2 error: if we accept the null hypothesis and state there is no statistically significant effect, what if we later discover that our conclusions were incorrect? We could miss opportunities for protecting public health! While scientists would be protected from making inappropriate claims of health effects when none exist, it is equally true that many more people could die before science finally has the statistical confidence to declare a health hazard. Of course, we can decrease both type 1 and 2 error if we have more powerful studies with larger sample sizes and more sensitive tests, but many scientific issues are midstream, and all too many scientists are in intellectual paralysis while waiting for stronger proof of an effect. In the language of risk, type 2 error reminds us that we must also consider proving that something is safe. This is at the heart of so many important policy issues. Activists are more concerned about type 2 errors, while classical scientists and corporations that often must devote huge resources to diverse public issues are more concerned about type 1 errors. When a scientific issue is fraught with uncertainty, this simply highlights the distinction between these two errors and creates a fundamental impasse in public policy.
This impasse has been the subject of great scrutiny by scholars and professionals over the years, and more attention has been devoted in the recent era to type 3 error, defined as "asking the wrong question. For example, we may have looked at cancer when the real problem is reproductive hazard. If we had asked a well focused question, we may have minimized errors of the previous categories. Agencies are prone to type 3 error because their focus is usually predicated by law. Similarly, type 4 error emphasizes use of the wrong method, particularly compelling in light of the previous discussion about different inputs to risk analysis. For example, we may be using chemistry to solve a social problem, or vice versa. Consider this: who is in a position to diagnose this problem? Generalists!!! As specialists use the tools of their discipline on an array of risk issues, their expertise belies the ignorance they may have for other specialties. It has often been said that the greatest development of recent scientific research is to have multi-disciplinary research. This speaks to the heart of type 4 error, although a multi-disciplinary group is not the same as an integrated group of generalists (recall the Tower of Babel).
Finally, type 5 error is reaching the wrong conclusion, sometimes described as the right diagnosis followed by wrong medicine, or as good science followed by bad policy. Scientists who are reluctant to be involved in public policy may be valuable in the lab but sit in frustration while poor policy unfolds. Science groups with social agenda often come under sharp criticism, but this approach speaks to the heart of type 5 error.
We finish this chapter with a consideration of some of the fundamental uncertainties in any risk assessment.
1. dose and exposures
2. short observation periods in experimental studies
3. inappropriate route of exposure
4. interactions (synergism, antagonism, potentiation)
5. misdiagnosis
6. misclassification of exposures (people are mobile)
7. inter - species extrapolation
8. extrapolation to low dose
9. improper control groups (healthy worker effect)
10. differences between experimental groups and risk groups
11. age, sex, confounders (smoking), etc.