I. The Source:

Anna Fan, Robert Howd, and Brian Davis, "Risk Assessment of Environmental Chemicals," ANNUAL REVIEW OF PHARMACOLOGY AND TOXICOLOGY Vol. 35 (1995), pgs. 341-368. The Annual Review of Pharmacology and Toxicology is published by Annual Reviews, Inc., 4139 El Camino Way, Box 10139, Palo Alto, CA 94303-0139. Editor: Arthur K. Cho; telephone: (415) 493-4400. Annual subscription rate: $47 ($52 foreign).

[NOTE: Risk assessment is widely used by government and industry to set "safe" levels for chemical exposures to workers and to the general public. Many dangerous technologies -- pesticides, medical waste incinerators, and soil-burning incinerators, for example -- are promoted as 'safe' based on risk assessments that evaluate the chemicals emitted and deem their risks 'acceptable.' It is therefore important to understand what risk assessment is, how it works, and what its shortcomings may be. It is not a simple topic, but an exceedingly important one. The "source" article this month is long and complex, but it describes the critical components of a risk assessment, thus preparing readers to ask good questions the next time a risk assessment comes their way.--P.M.]

II. Salient Facts from the Source:

[In the summary below, I have added explanations inside square brackets, attempting to clarify the text.--P.M.]

Risk assessment is an evolving process based mainly upon toxicology, but it is also based upon a broad background of knowledge in other fields: chemistry, physiology, molecular biology, environmental transport processes [how chemicals move through the environment], and applied statistics.

Risk assessment has been defined by the National Academy of Sciences as a four-step process:

Step 1: Hazard identification.

Hazard identification involves estimating chemical risks for acute (single dose), subchronic (a few doses), or chronic exposures for each possible toxic endpoint, such as liver damage [or birth defects, or immune system damage, or central nervous system damage, or reproductive system damage, or cancer, and any other such "endpoint"].

Step 2: Dose-response assessment.

Dose-response assessment usually includes extrapolation from high doses to low doses. [Dose-response assessment means determining what damage, and to which bodily systems, will occur as the dose of a chemical increases. Most people are familiar with the concept of dose-response; think of the effects from drinking one, two, or three glasses of wine. In general, greater dose leads to greater effect. To extrapolate means to project or extend known data into an area not known so as to arrive at knowledge, which is usually conjectural. Ethical considerations, embodied in the Nuremberg principle, established in international law after W.W. II, prevent testing of toxic chemicals on humans. Therefore, laboratory animals are tested instead. Because the number of animals tested is small, high doses are usually administered in the hope that, if there is a subtle effect from a chemical, it will be discerned even though the test group is small. As a result, laboratory animal findings must be extrapolated from the high doses given in the lab to the much lower doses to which humans would be exposed through contact with chemicals in the environment.]

Starting with information about the effects of high doses on various species of laboratory animals, many different techniques and models are used to estimate what the effects of low doses on humans might be.

Step 3: Exposure assessment.

Exposure assessment attempts [or should attempt] to determine how much of a chemical (or of all similar chemicals) is absorbed from all sources. Example: if the chemical is a pesticide, exposures might occur through food, groundwater, air, and through home and occupational uses.

Step 4: Risk characterization.

This final step in a risk assessment uses data from the three previous steps to make decisions regarding individual and population risks, blending the previous steps with information about the characteristics of the exposed population to describe the potential adverse outcomes and to describe the strength [validity] of both the evidence and the extrapolations. [In other words, the authors are saying that risk characterization takes information from hazard assessment, dose-response assessment, and exposure assessment, then adds information about the characteristics of the affected population -- How old are they? Do they eat a lot of fish? Are they generally malnourished or overweight? -- and combines it all together to determine an estimate of risk, usually expressed as a probability of a particular kind of harm over a specified period of time. For example, a typical estimate of risk might be expressed this way: a particular group of people is expected to endure one additional cancer for every 10,000 people, over and above the normal risk of cancer, as a result of chronic exposure to some toxic chemical in their drinking water during their lifetime of 70 years.]

The four components of risk assessment are carried out roughly in the order given, leading toward a quantitative estimate of the relative risk of a chemical exposure under specified conditions.

This review [the source article] addresses several areas of risk assessment that are receiving heightened attention, including neurotoxicity, immunotoxicity, reproductive and developmental toxicity, genotoxicity, carcinogenicity, and toxicokinetics [details of how toxicity occurs within the cells of bodily systems] and modeling [making mathematical approximations of physical, chemical and biological processes].

"Risk assessment is an evolving art based on sound scientific principles and judgments...."

"Not all considerations in risk assessment are specified in any one set of laws or procedural manuals; many are built upon the training and experience of the toxicologists who assess the risk. Thus there may be significant variations in the conduct and results of risk assessments."

General toxicity considerations

Protecting the public from toxic effects of environmental chemicals primarily involves considering the mechanisms of toxicity at low exposure levels, and likely biological effects from such exposures. Potential cumulative, irreversible effects such as carcinogenesis and neurotoxicity receive the most attention. Developmental effects are also intensively investigated, because a single exposure during a sensitive period in gestation [gestation is the time from conception to birth] can have lifelong detrimental effects. Preventing exposure to mutagens is important for the same reason. [Emphasis added; we return to this point below. Mutagens are toxins that can damage the genes, sometimes causing cancer or other illnesses in the affected person, but also sometimes causing damage that may be inherited by the person's offspring.]

In the study of problems other than cancer, the goal is to protect an exposed human population against any adverse effects by estimating a safe dose level, called a reference dose. The "safe" or "reference" [sometimes called "rfd"] dose is calculated as follows: The highest dose that doesn't cause toxicity (the no observed effect level, or NOEL) or, if that is unavailable, the lowest dose observed to have a toxic effect (the lowest observed effect level, or LOEL) in an animal study is divided by safety or uncertainty factors. [In other words, scientists take what they think are 'safe' levels for animals and they divide it, thus making it even smaller, and call it a "safe' dose for humans.]

In calculating the reference dose, safety factors of ten have commonly been applied to account for (a) uncertainty caused by extrapolation between species [for example, from rats to humans], for (b) variations in sensitivity among humans, for (c) extrapolation from a LOEL to a NOEL (when necessary), and for (d) estimating a safe chronic dose when only acute or subchronic toxicity levels are known. The total uncertainty factor used may thus vary from 10 to 10,000, depending on the available data; the most common value is 100. [In other words, in the typical case, a 'safe' dose is determined for, say, a rat, and that dose is divided by 100 to establish a 'safe' dose for a human.]

Uncertainty factors are used to account for both the inherent variability in humans (e.g. individual differences in size, age, exposure, metabolic rate, disease states, genetic susceptibility, etc.) and uncertainty resulting from undefined [unknown] toxic effects. "Some uncertainty may remain unquantifiable," the authors say.

Exposure to multiple chemicals

"Because of the near-infinite number of potential mixtures, it is impossible to experimentally test for all interactions, and relevant studies of an effect or interaction [between multiple chemicals] are not usually available."

Neurotoxicity assessment

Comprehensive evaluation of neurological competence [health of the central and peripheral nervous systems] is difficult and expensive. Dozens of procedures to analyze various nervous system functions have been developed over the last few decades to evaluate both acute and chronic effects of chemicals, but no comprehensive test battery has emerged as a standard. "Information on neurotoxicity for most environmentally significant chemicals is therefore fragmented and difficult to interpret, and there are no simple guidelines for use of the data in risk assessment." The US EPA [Environmental Protection Agency] has been in the process of formulating neurotoxicity [testing] guidelines for several years [but has not finalized those guidelines.]

Immunotoxicity testing

Risk assessment of immunotoxic chemicals is a new and challenging area. Reasons for its significance include:

(a) recent emphasis on the human health implications of the immune system; (b) public awareness that chemicals or biological agents can alter immune responses, as seen in acquired immune deficiency syndrome (AIDS); (c) the ability of immune function tests to detect changes at very low doses; (d) the fact that pesticides, a major class of chemicals, are known to modulate [modify] immune system response; (e) special vulnerability to immune system distress among young, old, pregnant, and malnourished individuals; (f) the potential for development of hypersensitivity, including Multiple Chemical Sensitivity (MCS); [MCS is a poorly-understood, and therefore controversial, disorder in which patients complain of a wide range of symptoms after exposure to low levels of chemicals that give off odors, such as perfumes, new carpets, etc.]; (g) and the current lack of regulatory data requirements and risk assessment guidance for immunotoxic effects.

Impaired immune response can result in increased susceptibility to, or decreased resistance to, bacterial, fungal, and viral infections, and to some forms of cancer.

The immune system is a complex network of lymphoid organs [lymph system] and cells in circulating blood and elsewhere in the body which interact to generate the immune responses. No single test is adequate to measure the immunologic effects of chemicals on those multiple system components.

It is reasonable to suppose that significant immunologic changes in exposed individuals will produce adverse outcomes. This concept is supported by human data from individuals with genetic or virus-induced (e.g., AIDS) immunodeficiencies and from those being therapeutically immunosuppressed (e.g., organ transplant recipients). [In other words, there is evidence from human experience that people tend to get sick when their immune systems are damaged.] In addition, a good correlation between immune test results and altered host resistance has been demonstrated in animal studies. [In other words, data from animal studies confirm that a damaged immune system can give rise to disease.]

To date, little immunotoxicity information is available for environmental chemicals. In the US EPA Integrated Risk Information System (IRIS), which provides summary toxicity data on over 200 chemicals, only one chemical has a reference dose based on immunotoxicity. The National Academy of Sciences in its recent study of pesticides in the diets of infants and children, recommended immunotoxicity assessment as part of a complete pesticide safety evaluation. Regulatory agencies such as the US EPA and the US FDA [Food and Drug Administration] are developing testing guidelines and data requirements for immunotoxicity testing. Formal guidelines for the use of immunotoxicity data in risk assessment do not exist.

Developmental toxicity assessment

Developmental toxicity is toxicity that adversely affects offspring through maternal exposure, with effects becoming apparent during the period from birth to sexual maturity. Adverse developmental effects include structural abnormalities, growth alteration, functional defects, and death. The most common test for developmental toxicity involves exposing females, typically rodents and rabbits, to a toxic chemical in the air, water or diet at various stages of pregnancy. By limiting exposure to various stages of gestation, it is possible to define the period of development that is most vulnerable.

Reproductive toxicity assessment

Reproductive toxicity is toxicity that adversely affects any aspect of male or female reproductive function. Effects may be observed as changes in reproductive cells or organs, in endocrine functions, or in behavior. [The endocrine system regulates bodily functions, including growth, development, and sexual characteristics, via hormones.] Because they affect reproductive cells, (germ line, meiotic cells, or gametes), genotoxic effects may be included in reproductive toxicity. [When they divide, meiotic cells give rise to gametes; germ line and gametes are sex cells, i.e., sperm cells or female egg cells, damage to which can be inherited by the offspring.] A mutational event in such a cell could theoretically be caused by a single molecule [of a toxic chemical]. This possibility is the basis for assertions that some development and reproductive toxins are non-threshold agents. [Non-threshold agents are harmful chemicals for which there is no threshold, no dose below which damage will not occur; for non-threshold agents, the only completely 'safe' dose is zero.]

Reproductive toxicity can be assayed [assessed, or tested] in a two-generation rodent study.

In pregnant women, adverse affects may occur in the fetus at dose levels too low to affect the mother, because the fetus is more sensitive to certain effects. Adverse fetal affects that occur at dose levels below those that cause maternal toxicity are clearly of concern. A recent study by the National Academy of Sciences [PESTICIDES IN THE DIETS OF INFANTS AND CHILDREN (Washington, D.C.: National Academy Press, 1993)] assumes that a safety factor of 1000 below the mother's NOEL may be necessary to protect the fetus.

Genetic toxicity (genotoxicity) assessment

Risk assessment for genetic effects is both important and difficult. The outcome of 5-10% of live human births is a significant birth defect. It has been estimated that among all congenital effects [birth defects] about 20% have environmental causes, 20% have predominantly genetic causes, and 60% result from a combination of the two. [In other words, environmental factors contribute to 20% + 60% = 80% of all birth defects. There were 4.1 million live births in the U.S. in 1993, according to the 1994 edition of the Statistical Abstract of the United States.]

In addition to problems that are severe enough to be recognized as birth defects, genetic changes can lead to more subtle results: a child may lead a less healthy life, be more susceptible to disease, or have a shortened productivity and life span.

A major problem in hazard identification is to determine the relevance to human health of the hundreds of available genotoxicity assays [tests].

Animal assays for germ line mutations [inheritable mutations] are costly and difficult, while human germ line data are never available.

Direct chemical interactions with genes represent non-threshold phenomena. [In other words, the only safe dose is zero.]

The US EPA has developed guidelines for mutagenicity risk assessment and in a series of papers has attempted to demonstrate appropriate risk assessment procedures, using ethylene oxide as an example. These results were offered in order to stimulate discussion and progress in this area, but a consensus on genetic risk assessment is not yet in view. [In other words, there is no agreed-upon way of assessing genetic risks, for the purposes of risk assessment, and agreement on such methods is unlikely to be reached soon.]

Carcinogenicity assessment

Risk assessments evaluate carcinogenicity differently from other toxic effects. The primary reason for this is the treatment of carcinogenicity as having no threshold and all other toxicity as having a threshold. The basis for this hypothesis is that mutations in a few critical genes can lead to the loss of regulation of cell division. Thus a few molecules of a carcinogen can produce an unregulated cell that divides to form a clone [an identical twin cell, which then divides to form two more identical cells, and so on] and ultimately a tumor composed of unregulated cells. [In other words, if a cell loses the ability to regulate itself, i.e., to stop growing, it can eventually form a cancer.] The consequence of this assumption is that any dose of a carcinogen, no matter how small, should produce a calculable risk. [In sum, carcinogens are assumed to be non-threshold toxins.]

Genotoxic chemicals should also lack a threshold, but because of the difficulties in risk assessment described in the previous section, they are not evaluated in the same manner nor with the same concern as carcinogens.

Carcinogens are regulated at risk levels in the range of 1 X 10-4 (one in 10,000) to 1 X 10-6 (one in a million). Such levels are based on rodent studies with about 50 males and 50 females in each dose group. Because male and female outcomes are treated separately, the limits of resolution for these assays are on the order of a few percent for a response. [In other words, because the size of the test group is small, such a test can only discern gross effects -- effects that occur in several percent of the tested animals. If an effect were to occur in only one in 10,000 animals or one-in-a-million, such tests would not be able to discern that effect.]

III. Our Interpretation:

This article was written by qualified scientists who are enthusiastic about risk assessment and who use it regularly in their work for the California Environmental Protection Agency (Cal EPA).

Nevertheless, the article highlights major shortcomings of risk assessment. For example:

Risk assessment is not really a science; it is an art that weaves together various strands of information, some of which have been gathered by scientific methods. The authors say risk assessment is "an evolving art based on sound scientific principles and judgments...." But "judgments" are not scientific because they are not reproducible from one laboratory to another; different risk assessors will make different judgments when faced with the same data (or the same lack of data).

Yet risk assessment is routinely characterized in the media and elsewhere as a scientific enterprise -- a characterization that is misleading at best, and intentionally deceptive at worst.

The authors acknowledge that different risk assessors will reach different conclusions because risk assessments are completed using differing techniques: they say, "Not all considerations in risk assessment are specified in any one set of laws or procedural manuals; many are built upon the training and experience of the toxicologists who assess the risk. Thus there may be significant variations in the conduct and results of risk assessments." Here we see again that risk assessment is not a science, but an art.

Risk assessment has traditionally emphasized the carcinogenicity of chemicals. But as the authors point out, there are many kinds of harm that chemicals can cause besides cancer. A risk assessment needs to consider neurological damage; damage to the immune system and the endocrine system; developmental damage; reproductive damage; and genetic damage.

They say, for example, "Developmental effects are also intensively investigated, because a single exposure during a sensitive period in gestation can have lifelong detrimental effects. Preventing exposure to mutagens [chemicals that cause inheritable genetic changes] is important for the same reason." In other words, a single exposure to a toxic chemical at the "wrong time" during development can cause permanent damage to a fetus or a young animal or child.

There will always be aspects of the toxicity of a chemical that are not known, so "uncertainty factors" will be applied to try to protect public health from unknown hazards. The authors say, "The total uncertainty factor used may thus vary from 10 to 10,000, depending on the available data; the most common value is 100." But, the authors say, "Some uncertainty may remain unquantifiable." If uncertainties cannot be quantified, then one cannot know whether a safety factor of 10 or 100 or 1000 is "adequate" to protect public health and safety. One can guess and hope, but guesswork and hope are not science -- they belong in the realm of art or personal philosophy.

The effects of multiple exposures cannot be known: The authors say, "Because of the near-infinite number of potential mixtures, it is impossible to experimentally test for all interactions, and relevant studies of an effect or interaction are not usually available." So risk assessors study the effects of one chemical at a time, as if people and animals were never exposed to multiple chemicals simultaneously. This is a major flaw in decision-making that is based upon risk assessment. In the real world, people are exposed to multiple chemicals all the time -- prescription drugs, cigarette smoke, automobile exhaust, multiple pesticides, food additives, indoor air pollution from building materials, carpets, and so on.

The likelihood of nervous system damage cannot be reliably evaluated because there is no standard way to test for it: The authors say, "Dozens of procedures to analyze various nervous system functions have been developed over the last few decades to evaluate both acute and chronic effects of chemicals, but no comprehensive test battery has emerged as a standard. Information on neurotoxicity for most environmentally significant chemicals is therefore fragmented and difficult to interpret, and there are no simple guidelines for use of the data in risk assessment."

The likelihood of damage to the immune system cannot be reliably evaluated because, as the authors say, "No single test is adequate to measure the immunologic effects of chemicals on those multiple system components." And, they say, where data do exist, the meaning of the data is not clear: "Human data concerning chemically induced modulations [modifications] of immune-response parameters are scarce. Where modulations have been observed, the biological significance has been unclear."

Furthermore, immune system toxicity data are truly scarce: The authors say, "To date, little immunotoxicity information is available for environmental chemicals. In the US EPA Integrated Risk Information System (IRIS), which provides summary toxicity data on over 200 chemicals, only one chemical has a reference dose based on immunotoxicity." There are over 60,000 chemicals currently in commercial use, so having 'safe' doses established for only one chemical based on immune system toxicity gives an indication of how far we have to go.

Toxicity to human genes is important, but difficult to assess. As the authors say, "Developmental effects are also intensively investigated, because a single exposure during a sensitive period in gestation can have lifelong detrimental effects. Preventing exposure to mutagens is important for the same reason."

The authors say, it has been estimated that among all congenital effects [birth defects] about 20% have environmental causes. In other words, environmental factors alone cause 20% of all birth defects. There are about 4 million live births each year in the U.S., so there are between 200,000 and 400,000 serious birth defects each year. Of these, 40,000 to 80,000 are caused by environmental influences and not by inherited genetic problems, according to the authors.

In addition to problems that are severe enough to be recognized as birth defects, genetic changes can lead to more subtle results: a child may lead a less healthy life, be more susceptible to disease, or have a shortened productivity and life span.

There is no way to quantify such subtle effects, and no way to avoid them except by eliminating exposures to genotoxic chemicals. Risk assessment thus gives a false sense of security.

But genotoxic chemicals are very difficult to identify: The authors say, "A major problem in hazard identification is to determine the relevance to human health of the hundreds of available genotoxicity assays [tests].

"Animal assays for germ line mutations [inheritable mutations] are costly and difficult, while human germ line data are never available," the authors say.

"Direct chemical interactions with genes represent non-threshold phenomena," the authors say. [In other words, the only safe dose is zero; if a large population is exposed to a non-threshold chemical, some damage will occur, even at low doses.]

"...a consensus on genetic risk assessment is not yet in view." [In other words, there is no agreed-upon way of assessing genetic risks, for the purposes of risk assessment, and agreement on such methods is unlikely to be reached soon.]

Therefore, it seems proper to conclude that testing of chemicals for genotoxicity is very important, but cannot be done at present, and is unlikely to be done in the near future.

Furthermore, it is known that some chemicals can harm the nervous system, but risk assessors have no standard way to evaluate chemicals for such damage.

Furthermore, it is known that some chemicals can harm the immune system, but risk assessors have no standard way to evaluate chemicals for such damage.

Furthermore, it is known that humans are exposed to multiple chemicals simultaneously, but risk assessors have no way to evaluate the effects of several chemicals simultaneously.

Furthermore, the authors have presented evidence that developmental toxicants, reproductive toxicants, and genetic toxicants are all non-threshold chemicals. Carcinogens are also considered to be non-threshold chemicals. For these classes of hazardous chemicals, the only safe dose is zero, which means we must prevent exposures not rationalize and justify them via risk assessment.

And lastly, in our 25 years of experience advising citizen groups, we have never seen a risk assessment that conforms to the guidelines implied in this article. For example, we have never seen a risk assessment that actually took into account the particular characteristics of an exposed population; "averages" are used, not particulars. Therefore, though this article defines a risk assessment as a rather thorough investigation, in our experience, the "rules" for carrying out risk assessments are more honored in the breach than in the observance.

In sum, risk assessment is not a technique that can protect the public from toxic chemicals. Instead, it is a technique that gives the public a false sense of security, while allowing health damage to occur. People, including scientists, who claim to be able to determine 'safe' doses of chemicals based on risk assessment techniques are deceiving themselves and -- worse -- misleading the public.

IV. Contacts:

Anna Fan, senior author, is chief of the Pesticides and Environmental Toxicology Section, Office of Environmental Health Hazard Assessment, California Environmental Protection Agency: (510) 540-3063.

Robert Howd is a staff toxicologist with the Pesticides and Environmental Toxicology Section, Office of Environmental Health Hazard Assessment, California Environmental Protection Agency: (510) 540-3063.

Brian Davis is with the Office of Scientific Affairs, Office of Toxic Substances Control, California Environmental Protection Agency: (916) 327-2493.

Mary O'Brien is staff scientist with Environmental Research Foundation; she lives in Eugene, Oregon, where she can be reached at (503) 485-6886. O'Brien is writing a book criticizing society's excessive reliance on risk assessment for decision-making, and proposing that we evaluate all available alternatives before making decisions--a simple change with far-reaching ramifications.