Michael J. DiBartolomeis, PhD
INTRODUCTION
Most people generally are aware that voluntary or involuntary exposure to chemicals and other hazardous substances can cause harm to their health or to the health of their children and the unborn fetus. Taken at the minimum necessary dosages, however, some chemicals, such as medicines, are also beneficial to human health. Manufacturing with chemicals has resulted in some new products and technologies that have, arguably, benefited society as a whole by creating new jobs, developing less costly and more durable consumer products and building materials, and improving communication and transportation. However, the true cost of the production, use, and disposal of these synthesized chemicals to the environment and human health is unknown and difficult to quantify. Furthermore, we know that hazards in the workplace associated with chemical exposure often are greater than the hazards from exposure to environmental pollutants. Many other factors also play a role, including poverty and employment status, which affect nutrition and access to health care, violence, smoking, and drug use. Scientists and policymakers still do not know the exact degree to which human health problems can be attributed to environmental pollution and how much should be attributed to other environmental factors or lifestyle choices.
In the early 1970s, the level of concern for the safety of the food supply, air, drinking water, and working environment intensified, and new laws were passed and regulations promulgated to help control and restrict the level of pollutants released into the environment. Many of these regulations were based on observed or predicted human health effects of exposure to hazardous materials either in the environment, in the food or water supplies, or in the workplace. Despite these efforts, some contend that not enough is being done to clean up and maintain a healthy environment, whereas others believe that these concerns are exaggerated or unwarranted.
Given the scientific uncertainties involved in evaluating the impact of environmental stressors on human health, it is prudent public health practice to reduce or eliminate preventable exposures to hazardous substances when an activity raises the risk of harm to human health or the environment, even if cause-and-effect relationships have not been fully established. This is the guiding principle behind the precautionary approach to risk management, a familiar component of international and European environmental law. Furthermore, environmental protection programs should effect empowerment within individuals and communities and raise the consciousness about their health, their environment, and multicultural issues. In the United States, these are particularly important given the rapidly changing demographic face of the nation, the ongoing problems associated with environmental pollution, and the increased production and use of chemicals.
RISK AS A DECISION-MAKING FACTOR
Environmental decision making is a multidimensional process. Policies and laws that are written to address concerns about environmental pollution, occupational hazards, and the protection of human health usually rely on information taken from a myriad of sources, some of which are process-based and others of which are value-based or based on a systematic analysis. Table 50–1 provides examples of some factors that might be considered in formulating a decision on an environmental problem.
Table 50–1. Examples of decision-making factors that might be considered in formulating environmental policy.
Although it is only one tool that might be used in the overall decision-making process, government agencies more often than not consider risk first when making decisions on mitigation, control, enforcement, or regulation of chemicals released into the environment. By definition, risk is the probability or chance that a desired or unwanted action, circumstance, or event will result in loss or harm. It can apply to almost any activity or event, such as the likelihood for injury when playing a sport or driving a car, the chance for developing a disease from exposure to pathogens or chemicals, or the possibility for property damage from a natural catastrophe. This chapter focuses on human health risks and how to evaluate risk. Risk assessment methodology also has been developed and applied to evaluate the impact of pollution on the environment and ecosystems and, to a lesser degree, on quality-of-life issues. In the context of human health, risk is the probability that adverse health effects, ranging from death to subtle biochemical changes, may occur because of exposure to a hazardous substance. Risk also might be thought of as voluntary or involuntary. Smoking, for example, is both a voluntary and an involuntary risk. It is voluntary because the smoker might choose to begin smoking. It is involuntary because second-hand smoke can cause harm to nonsmokers and also because nicotine is addictive, and it is difficult to stop smoking even if the user wants to.
Risk assessment is a means or methodology to quantify risk, but it is important to recognize that it is a process and not a science. The process of risk assessment uses scientific data, statistical and mathematical methodology, and expert judgment to characterize the probability for an adverse outcome. In its most basic form, risk assessment is the process through which toxicology data collected from animal studies and human exposure studies are combined with information about the degree of exposure to predict the likelihood that a particular adverse response will be seen in an individual or a population.
Historically, the results of risk assessments have been used to regulate chemical production, use, and release into the environment or food supply. For example, risk assessment methodologies have been used to set standards for pesticide residues in food, chemical contaminants in drinking water, indoor and ambient air standards, and exposure limits for contaminants found in consumer products and other media. However, risks might be assessed differently among agencies, and there are actually only a few “environmental agencies” that assess environmental or occupational health risks. These agencies attempt to make decisions based on data supported with scientific judgment. Some agencies also are mandated to consider future or multiple risks. With the exception of the application of pesticides in agriculture, risk assessment has not been used widely as a basis for setting workplace exposure standards.
GENERAL RISK ASSESSMENT PROCESS
Elements of the Model
The risk-based model for environmental priority setting generally follows a two-tiered approach. The first tier is to evaluate the size and scope of the potentially hazardous situation and quantify the level of risk posed by the hazard (risk assessment). The National Research Council defines risk assessment as a four-step process developed to aid in the evaluation of the safety of synthetic chemical use or the exposure to humans from chemicals in the environment. The four steps of risk assessment are hazard identification, dose-response assessment, exposure assessment, and risk characterization. In conducting health risk assessments, a number of representative questions about each environmental problem are asked (Table 50–2).
Table 50–2. Standard steps to conducting a health risk assessment.
The results of a risk assessment then are used to help determine which risks need to be addressed or managed. This second tier is called risk management, and it uses a value-based approach to determine what level of risk to human health will be considered significant and to formulate options for identifying, selecting, and implementing actions to prevent, reduce, or maintain risks below that level. Risk management considers risk along with other technical (such as technical or methodological feasibility), economic, legal, and social factors.
A third tier of the risk assessment model, risk communication, was added later with the intent of linking risk assessors with the public by presenting information in the most effective way. In communicating risk to the public, some questions that might be asked include: Is the information clearly relevant to and understandable by the affected public? Does the information respond to the public’s concerns? What are the limitations of the risk assessment? Despite the best efforts of the risk assessors to communicate the results of a risk assessment to the public, it is clear that risk communication is an afterthought in the process. More recently, as the emphasis for addressing environmental pollution issues has been placed on the affected communities (ie, disproportionate risk and environmental justice), the importance of involving the public earlier in the process has been realized.
Scope of Risk Assessment
Health risk assessments can be conducted for any hazard for which there are adequate toxicologic (from animal or human exposures) or epidemiologic data and either measured or estimated exposure in an individual or population. The spectrum of health effects described in toxicologic and epidemiologic studies is quite broad and might include acute, subchronic, and/or chronic effects following exposure to a chemical or chemical mixture. Acute adverse health effects usually are observed a few hours after a single high-level exposure (or dose) or after several high-level exposures over a short period of time. Although some health effects, such as delayed neuropathy or developmental toxicity, might be observed days or even months after a single high-level exposure to a chemical, chronic health effects usually are observed following repeated low-level exposures over many years (up to a lifetime in animals), and subchronic health effects usually are observed from repeated doses over 30-90 days in animals and for up to about 1 year in humans.
Table 50–3 presents some typical toxicologic endpoints used for risk assessment. It should be mentioned that for some toxic effects, the length and level of exposure might not be limited to any one category, and in fact, there is some overlap. As a general rule, a risk assessment does not exclude any toxicologic effect that is clearly caused by the chemical exposure. In cases when there is ambiguity in the data or the data are incomplete, it is generally a responsible approach to assume that the health effect is related to the chemical exposure until more data become available that clearly show an alternative cause of the adverse health effect.
Table 50–3. Common toxicologic endpoints reported in animal and human exposure studies that are used for quantitative health risk assessment.
This approach to toxicity testing is resource-intensive, time-consuming, and cannot effectively account for the toxicity of complex chemical mixtures. Furthermore, the results of whole animal toxicity testing provide little information on the variability in human susceptibility and the mechanism by which a chemical exerts its toxic effects. Because of these and other reasons, the demand for complete toxicity testing of tens of thousands of chemicals in commerce is not being met. Proposals to address the inadequacies of the current testing system include focusing on upstream biochemical events and cellular changes that might lead to the downstream observable effects in whole animal studies. Using predictive, high throughput in vitro assays, individual chemicals and chemical mixtures could be evaluated for relevant perturbations of key early biochemical and cellular changes that are thought to initiate “toxicity pathways” leading to gross pathological changes and disease. If this vision is implemented, current toxicity testing models would be phased out while new rapid, high throughput methods are developed resulting in the more efficient testing of all chemicals in a timely, cost-effective fashion.
RISK ASSESSMENT STEPS
Hazard Identification
To begin a risk assessment, hazard identification is the step in which it is determined whether exposure to an agent could (at any dose) cause an increase in the incidence of adverse health effects (eg, cancer, birth defects, or neurotoxicity) in humans. Many factors are considered in this determination, and depending on the toxicologic endpoint of concern, there might be specific additional factors to consider. A compound’s chemical and physical properties need to be known to be able to evaluate its fate in the environment and biological systems (eg, stability, half-life for elimination), the potential for bioaccumulation, possible routes of metabolism, and the likely toxicity of the compound. Also, factoring in the potential for human exposure and the likely routes of exposure is important to prioritize chemicals for hazard assessment.
If human exposures and toxicity are well documented, identification of a hazard is relatively easy; it can be more complicated when only experimental data in animals are available. In general, the criteria used in a risk assessment to identify a threat to human health from animal data include the number of animal species affected, the dose at which the animals are affected, the existence of a dose-response relationship, the severity of the effect, and for some agents, whether the toxicity observed in the animal is relevant to humans.
For individual chemicals and chemical mixtures, multiple health effects frequently are observed following dosing in animals or exposure to humans. For example, as required under the Federal Insecticide, Fungicide and Rodenticide Act (FIFRA), registrants must submit data from a standard battery of experimental toxicity tests that include acute, subchronic, and chronic studies for all pesticide active ingredients. Each pesticide usually exhibits some consistent toxicologic effects in different species that are related or unrelated to the pesticidal action of the chemical. In addition, there also might be either nonspecific toxicity or species-specific effects that occur at comparable doses or at higher or lower doses than the consistent toxicologic effects.
The spectrum of toxicity exhibited by a chemical in a battery of tests can be considered a “hazard profile” that might or might not be consistent with other structurally related chemicals or chemicals that exhibit comparable mechanisms of action.
For some toxicologic endpoints, additional consideration needs to be given to fully characterize or profile the hazard. For carcinogens, it is also important to consider the number and types of tumors occurring in the animals, the target organs affected, the background incidence (usually regarded as historical controls), the time-to-tumor response, the formation of preneoplastic lesions, and the genotoxicity (including mutagenicity) of the chemical. For carcinogens, there might not be consistency among species for tumor type, and there might be positive data in one species and negative data in another. Depending on the final use of a risk assessment, it is often prudent to accept the results from positive studies even if there are negative studies in order to take a precautionary approach to protecting public health.
To address the concern of equivocal data, a “weight-of-evidence” approach might be taken. A weight-of-evidence approach considers the complete data set (including all negative and positive results) as a whole in order to gain an appreciation of the scientific certainty of the identification process. This process includes all available data, regardless of the source, and evaluates the results of the studies in a qualitative manner to develop a sense of consistency or inconsistency in the data set. A meta-analysis approach, on the other hand, involves compiling data from comparable experiments (ie, similar experimental design, statistical power, reporting details, and overall quality) and evaluating the data set in a quantitative, statistical context. Epidemiologic data from several comparable studies sometimes are examined using meta-analysis, as are data from multiple carcinogen bioassays in animals.
In the hazard identification phase of a health risk assessment, there is often a need to separate statistical significance from biologic significance. Statistical significance might exclude effects of biologic significance, and in the case where several studies demonstrate comparable biologic effects with varying statistical significance, the effect still might be considered for risk assessment. In the dose-response assessment step, other criteria would be applicable to help discern the mechanism of toxic action and the use of the data for quantitative purposes. Furthermore, there are toxicologic endpoints for which biologic relevance is not known or difficult to define (eg, increased immunologic activity without obvious clinical signs of toxicity). Therefore, the risk assessor might attempt to define the term adverse effector at least segregate an effect that is clearly adverse from one for which the data are equivocal. The validity of this exercise is open to scientific debate, and there are many examples where the difference between adverse and nonadverse is not at all clear for a toxicity endpoint.
Dose-Response Assessment
Dose-response evaluations define the relationship between the dose of an agent and the observance or expected occurrence of a specific toxicologic effect. A dose-response evaluation usually requires extrapolation from doses administered to experimental animals to the exposures expected from human contact with the agent in the environment or in the workplace. When evaluating toxicologic effects in animals, it is generally assumed that at a given dose the animal response to a chemical will be nearly identical to the human response. This approach is reasonably accurate for chemicals that exhibit a threshold dose-response curve and which are eliminated from the body fairly rapidly (ie, short biologic half-life). If available, human exposure/dosing data from occupational or environmental exposures might be useful to better characterize the dose-response relationship of a chemical and its toxic effect. Data from human volunteer studies for exposure to hazardous substances are less desirable because of the generally poor study design, inherent bias of the subjects or the investigators, lower statistical power, and questionable ethical context.
Chemicals are thought to exhibit two types of dose-response relationships, those exhibiting a threshold for toxicity and those that do not. For chemicals that exhibit a threshold, the basic principle is that a specific dose level can be identified below which no toxic effect would be observed. The conventional approach to selecting dose levels for risk assessment of chemicals that exhibit a threshold for toxicity is to first identify the most sensitive endpoint from all studies and then to identify the highest no-observed-adverse-effect level (NOAEL) for that endpoint from the data collected from comparable studies. If no NOAEL can be identified (because of a dose selection that did not find a dose level at which no effect was observed), then the lowest observed adverse effect level (LOAEL) is substituted. In the case where a LOAEL and not a NOAEL is used for risk assessment, additional uncertainty is inherent in the calculation of risk that should be accounted for in the risk characterization step (see “Risk Characterization” below).
Alternatively, for chemicals that exhibit a toxicity threshold, a benchmark dose (BMD) methodology might be better suited with certain data sets in which a NOAEL cannot be clearly established. In this method, a toxicologic effect is first identified, such as a percentage of animals exhibiting a response or a percentage of decrease or increase in an enzymatic activity. Second, a benchmark response level is selected (eg, a response rate of 5% or 10%), and a mathematical model is applied to the data. The fitted curve then is used to designate the corresponding BMD. A lower limit on the BMD confidence level often is chosen as the NOAEL equivalent. This BMD confidence level then is used for risk assessment calculations by applying the appropriate safety/ uncertainty factors.
The methods for dose-response extrapolation employed for carcinogens are different. It is widely assumed that for chemicals that induce tumors, no threshold for toxicity exists. However, we do not fully understand the mechanism(s) of action for all chemical carcinogens. Chemical initiators and promoters have been identified in experimental studies, and for these, a postulated genotoxic mechanism of action appears to be reasonable. For other chemicals that induce tumorigenesis in laboratory animals, the evidence supporting a genotoxic mechanism of action is equivocal or negative, and other mechanisms, such as cytotoxicity or disruptions in physiologic processes that affect hormone levels or immunologic response, have been postulated.
To describe the dose-response curve for carcinogens at the low doses expected for human occupational or environmental exposures, it is often necessary to extrapolate from the relatively high doses used in cancer bioassays (typically in rodents).
Most low-dose extrapolation models are derived from assumptions of the statistical distribution of the data (eg, log-probit, Mantel-Bryan, logit, and Weibull), the postulated mechanism of carcinogenicity (eg, linear one-hit, gamma multihit, and Armitage-Doll multistage), or some other parameter (eg, time to tumor, pharmacokinetic, and biologically based). The carcinogenic process typically is described mathematically by a set of elementary biologic events, most often as part of a multistage process, and the effect of carcinogens on these processes is assumed to be the simplest possible (eg, described by a chemical reaction rate). Therefore, the dose-response relationship described by these mathematical models usually will be as arbitrary as the assumptions made for the biologic processes.
There are several mathematical models that usually will fit the animal cancer bioassay data. Because these models use different formulas and assumptions for predicting the chemical’s carcinogenic potency, they might yield different results at the doses to which humans are exposed depending on the characteristics of the dose-response curve and the assumed mechanism of carcinogenicity (Figure 50–1). For most carcinogens, the one-hit and linearized multistage models are applied to the animal cancer bioassay data in order to estimate cancer potency in humans. These models were developed based on our understanding that ionizing radiation and genotoxic chemicals exhibit a linear, or nearly linear, response in the low-dose region. When presenting the results of the dose-response assessment for carcinogens, the upper-bound risk from the cancer models are provided as well as the upper and lower bounds of the risk. The objective of the bounding techniques is to attempt to account for the statistical uncertainty in the results of the animal tests.
Figure 50–1. The fit of most dose-response models to data in the observable range is generally similar (left plot). However, because of the differences in assumptions on which the equations are based, the risk estimates at low doses can vary dramatically between the different models (right plot).
There are chemicals for which there are positive cancer bioassay data but negative or equivocal genotoxicity data. There is an ongoing debate in the scientific community as to the mechanism of tumorigenesis for these agents. For example, the chloro-s-triazine herbicides (eg, atrazine, simazine, and cyanazine) induce mammary tumorigenesis, but the data for genetic toxicity are equivocal. There is some evidence that these chemicals disrupt endocrine function at the level of the hypothalamus-pituitary-ovarian axis, although they do not bind estrogen receptors. Therefore, a threshold dose-response for the triazine herbicides has been proposed, but no clear mechanism of action has been demonstrated. Other examples of chemical carcinogens for which there is ongoing debate as to the mechanism of action include chlorinated solvents such as chloroform and chlorinated polycyclic aromatic compounds such as 2,3,7,8-tetrachlorodibenzo-p-dioxin.
Physiologically based pharmacokinetic (PBPK) models are used by some risk assessors to predict the human response from rodent data. These models attempt to quantitatively account for the various differences between the test species and humans by considering body weight, metabolic capacity and products, respiration rate, blood flow, fat content, and a number of other parameters (Figure 50–2). Confidence in the results of physiologically based pharmacokinetic models often relies on some untestable assumptions, such as the delivered dose of an unstable metabolite to a target organ. While PBPK models have been developed for a variety of industrial chemicals (eg, chlorinated solvents) and pesticides (eg, malathion), application of the results of these analyses for risk assessment is still not clearly defined. Biologically based approaches to estimating cancer risk are also being developed that allow for the incorporation of biologic factors such as the number of mutations required for malignancy and the role of target-cell birth and death processes in the accumulation of these mutations. A key element is a quantitative description of how the carcinogen affects the cellular birth, death, and mutation rates. At this time, however, most of the information needed to perform these analyses is not yet available.
Figure 50–2. Simplified diagram of a general compartmental physiologically based pharmacokinetic model. The (a) absorption, (b) distribution, (c) metabolism, (d) storage, and (e) elimination of an internalized xenobiotic are described by a series of mathematical interrelationships. Physiologically based pharmacokinetic models yield information such as the predicted change in the amount of a chemical in a given organ over time depending on the data input (eg, rate constants for transport, distribution, respiration, metabolism, and excretion, as well as the chemical and physical properties of the chemical). The compartments are intended to represent, as best as possible, actual anatomic structures, defined with respect to their volumes, blood flows (perfusion rate), chemical binding (partitioning) characteristics, and ability to metabolize and excrete the chemical of interest. For risk assessment purposes, these models are used primarily to predict and compare target tissue doses for different exposure situations in different animal species.
The results of human exposure (eg, epidemiology) studies also might provide useful data to supplement the animal cancer bioassay data or offer an independent assessment of the dose-response of a chemical and its effect in humans. The design of human exposure studies, however, often limits use of the results of such studies for risk assessment purposes because the degree of uncertainty in estimating exposures is greater and the statistical power of the studies is usually lower than for experimental animal studies.
Exposure Assessment
For there to be a health risk, there must be both inherent toxicity and exposure to a chemical. In other words, the prevention or elimination of the exposure to a toxic substance would result in zero risk. Because the total elimination of chemical exposure often is not feasible or practical, the exposure assessment step in a risk assessment is used to estimate the magnitude and probability of uptake from the environment by any combination of oral, inhalation, and dermal routes of exposure. The results of the exposure assessment are quantitative doses presented in the amount of the chemical per unit of body weight per unit of time (eg, mg/kg per day).
Early in the exposure assessment, the population at risk needs to be identified by determining who would be exposed to the chemicals of concern. The size of the exposed population depends on the proximity of the population to the source. For example, there is a high potential for exposing large numbers of people if the chemical is in drinking water or air. On the other hand, if the contamination is confined to an enclosed area (eg, indoor workplace), the population affected is likely to be smaller. In characterizing an exposed population, it is important to consider age, gender, health status, and race and cultural diversity within that population because individuals differ in sensitivity and susceptibility to a chemical hazard.
The primary routes of exposure to chemicals in the environment are inhalation of particulates, dusts, and vapors; dermal contact with contaminated surfaces (eg, soils or contaminated vegetation); use of consumer products (eg, paints and plastic containers); and ingestion of contaminated food, water, and contaminated surfaces (ie, hand-to-mouth transfer). Workplace exposures also result from inhaling, ingesting, and making contact not only with contaminated media but also with concentrated solutions or mixtures of industrial chemicals. Despite recent advances in protective clothing and gear, labeling instructions, and properly engineered ventilation systems, the potential for workplace exposures is still significantly higher than most environmental exposures.
Estimates of human exposure might be based on analytic measurements of samples taken from environmental or workplace monitoring, direct measurements of human exposure, or mathematical (predictive) models. Although direct measurements of human exposure are the most precise methods for detecting exposure in an individual or population, these methods are costly, require specialized instruments, and are time consuming. More frequently, exposure estimates are based on mathematical models. Numerous methodologies for estimating the human uptake of contaminants have been proposed and refined in recent years. Models have been developed and used to predict the movement of chemicals in the environment (eg, in air, groundwater, or surface water), transfer from contaminated surfaces (eg, carpet or clothing, hand-to-mouth), and deposition onto edible fruits and vegetables. Physiologically based pharmacokinetic models also are used to predict the rate of absorption, metabolism, and distribution of a chemical in the body. Some retrospective studies of human exposure rely on surveys and the recall of the exposed persons. This latter method, while often necessary, is the least reliable and one reason that data from some epidemiologic studies cannot always be used for quantitative risk assessment.
In quantifying exposure doses, the number of exposed persons at each of the anticipated dose levels is described, as well as the upper and mean estimates of exposure. The best approach is to develop exposure scenarios that examine a range of potential or actual exposures for individuals, populations, and subpopulations. Depending on the use of the risk assessment, it might be adequate to estimate only doses from a single chemical exposures from a single source of the chemical. More often, multiple chemical exposures from multiple sources should be evaluated and aggregated, despite the relative complexity of doing this.
Formulas for estimating exposures from environmental and workplace chemicals can be applied in order to quantify dose levels for risk assessment. These formulas require entering values for physiologic and activity parameters such as breathing rate (resting and/or under exertion), daily water ingestion, food intake, body weight or size, and other factors that depend on the age, gender, physical well-being, and habits of the individual. Factors such as drug interactions, physical debilitation, stage in development (eg, fetus, perinatal, or infancy), and smoking status, for example, might increase susceptibility and sensitivity to a chemical exposure and should be documented and considered in the exposure assessment if possible. Values for body weight, breathing rate, and body size are obtained from tables that normalize the data and present mean and statistical bounds on the data. Often, values for parameters such as water ingestion and food intake are obtained from regional or even national surveys and therefore are not specific to a particular community, ethnicity, or lifestyle. For a more precise or defining exposure assessment for a specific population or individual, it is necessary to gather more specific data for entering into the exposure formulas.
Application of statistical analyses to the exposure data set might be necessary to determine the distribution of data because environmental and occupational data might be lognormally distributed rather than conform to a Gaussian distribution.
Depending on the exposed population and the problem, exposure estimates might need to be made for different subpopulations (eg, children and infants, pregnant women, and the infirm) because these individuals are differentially susceptible, exhibit different activities patterns, or are particularly sensitive for a number of reasons. For a purely statistical description of a population, stochastic or “likelihood of risks” approaches were developed to characterize exposures using models that replicate randomness in exposure. The probabilistic techniques can characterize a range of potential exposures and their likelihood of occurrence.
Some chemicals persist for many years in the environment, whereas others degrade rapidly. The environmental fate of chemicals depends on several factors, for example, the chemical and physical properties of the substance, the potential for movement through various environmental media (eg, groundwater and porous soils) or storage (eg, binding of chemicals to sediments), the rate of degradation in the environment (eg, by sunlight, soil and water microbes, and evaporation), and the potential for bioaccumulation and biomagnification. Some chemicals such as the polychlorinated aromatic hydrocarbons (eg, polychlorinated biphenyls and dichlorodiphenyltrichloroethane [DDT]) can persist in the environment for 50 or more years, whereas other chemicals (eg, some organophosphorus pesticides) degrade relatively rapidly and will persist for weeks or a few months. Lipophilic chemicals (eg, methyl mercury) in the environment are stored in the tissues of animals, most notably fish and, through a process called biomagnification, increase (sometimes to concentrations hundreds of times greater than the original environmental levels) as the stored chemicals move up the food chain. Therefore, although direct human exposures to chemical contaminants might be reduced when chemicals are degraded rapidly, there is certainly significant exposure potential for even those chemicals that exist for only a few days in the environment (eg, agricultural workers) or that start out at low concentrations but bioaccumulate in the food chain (eg, contaminated fish).
New technologies and advances in analytic instrumentation and methodologies now allow for the detection of very small quantities of exogenous (xenobiotic) chemicals in blood, urine, hair, feces, exhaled breath, and fat and other tissues (ie, biomonitoring). Measurement of chemical residues at parts per trillion (ppt) levels and even lower is now possible in biologic tissues (as well as in environmental media). For many chemicals, biomonitoring results represent a direct indicator of either acute or chronic exposure to a chemical. These direct measurements offer a better alternative to assessing exposure than using any mathematical models. Furthermore, environmental monitoring also has benefited from these advances in technology, although the presence of mixtures of chemicals and the matrices in which these chemicals reside tend to complicate and interfere with environmental measurements at low levels. As field measurement techniques are further refined, less reliance will need to be placed on mathematical models for predicting the distribution of chemicals in the environment.
Risk Characterization
In the risk characterization, the risk assessor summarizes and interprets the information collected from the previous three steps, presents a quantitative estimate of the human health risk(s), and identifies (and quantifies when possible) the uncertainties in these risk estimates. This process allows the risk assessor to identify the greatest individual and population health risks and promulgate health-based action levels to protect individuals and populations from further exposure or to prevent immediate- or long-term injury. Estimated risks depend on the measured or estimated exposure duration and can be calculated either retrospectively (ie, the release of the chemical or the exposure has already occurred) or prospectively (ie, as a means to prevent a release or the exposure from happening). It is appropriate and often necessary in a risk characterization to estimate both noncancer and cancer risks for a chemical exposure and to evaluate multiple exposure scenarios to aid in the determination of the necessary mitigation steps.
For chemical toxicity endpoints that clearly exhibit a threshold dose-response curve, reference exposure levels (RELs), defined as threshold exposure levels below which no adverse health effects are anticipated, can be calculated. These reference levels are comparable with the EPA’s reference doses (RfDs) or reference concentrations (RfCs).
RELs are derived by identifying and dividing the NOAEL (or BMD) by uncertainty factors to account for inadequacies in the database, incomplete scientific knowledge, and protection of more sensitive individuals (Table 50–4). The application of uncertainty factors offers a margin of safety to consider when developing mitigation options or regulatory standards. Some uncertainty factors can be considered default values when adequate physiologic or toxicologic information does not exist to provide a more precise estimate of uncertainty.
Table 50–4. Uncertainty factors that may be applied in calculating risk-based exposure levels.
For carcinogens, unless a threshold for toxicity is clearly demonstrated, it is assumed that the dose-response is linear with no “no risk” level. For these chemical agents, a cancer potency is calculated, and the probability for excess individual cancer risk is estimated based on exposure estimates. The determination as to what is an “acceptable” (or de minimis) cancer risk is a value-based decision, and often a range of risk is presented for comparative purposes.
Documented differences in physiology and toxicology between species may be used to modify RELs and, to a lesser degree, cancer risk estimates to better reflect the human exposure and predicted response to the chemical. The concept of ensuring a margin of safety between exposure and toxicity still should apply, however, even when a more precise estimate of uncertainty can be made. In particular, some subpopulations (eg, the developing fetus, infants, and children) may be more sensitive or differentially susceptible to a chemical exposure. It is difficult to predict with accuracy the effects of a chemical exposure to such an individual compared with the average, healthy adult in the population. Frequently, gender, race, or other genetic traits also may affect an individual’s sensitivity. The risk characterization step should take into account the differences in individuals and subpopulations and uncertainties in the data and methodology.
In general, a thorough characterization of risk also should discuss background concentrations of the chemical in the environment and in human tissue, pharmacokinetic differences between the animal test species and humans (the results of a PBPK or another biologically based model are useful here), the effect of selecting specific exposure parameters, the level of uncertainty in the methods (ie, calculations and statistical analyses), and other factors that can influence the magnitude of the estimated risks. Furthermore, areas for which additional research is needed also should be identified (eg, data gaps).
EXAMPLE OF THE APPLICATION OF RISK ASSESSMENT METHODOLOGY
The general approach to calculating risk for noncancer and cancer endpoints is illustrated below for the pesticide and environmental contaminant dibromochloropropane (DBCP). California promulgates maximum contaminant levels (MCLs) for drinking water contaminants that are based in part on public health goals. In deriving an MCL, which is a regulatory standard, costs, benefits, and technical feasibility (eg, of detection or mitigation) must be considered. A public health goal is developed based on a risk calculation, consideration of the uncertainty in the methods and the data, and taking into account the most sensitive or susceptible individuals (eg, infants and children). The public health goal is developed in order to protect public health, but it is not a regulatory standard like an MCL and therefore is not enforceable.
DBCP was used extensively as a soil fumigant and nematocide in the United States until 1977, when its registration as a pesticide was suspended. Although it is no longer manufactured commercially or used in this country, groundwater contamination still exists in the San Joaquin Valley and other agricultural regions in California. Exposure to DBCP occurs from the use of tap water as a source of drinking water, as well as in preparing foods and beverages. It is also used for bathing or showering and for washing, flushing toilets, and other household uses resulting in potential dermal and inhalation exposures.
Noncancer Health Effects
DBCP induces testicular damage and infertility, as evidenced by numerous studies of occupational exposures, described as reduced (oligospermia) or no sperm counts (azoospermia), altered sperm motility, damage to the seminiferous tubules, and hormonal disruption. Testicular toxicity is reported most frequently and appears to occur at lower exposures than that of other noncancer endpoints (ie, it is the most sensitive noncancer toxicity endpoint). In experimental animal studies, the highest NOAEL of 0.025 mg/kg per day is identified for adverse testicular effects in the male rabbit. Using this information, the calculation of an REL (or public health goal), in this case defined as C mg/L for a noncarcinogenic effect of DBCP, follows the equation
where NOAEL is no observed adverse effect level, BW is body weight (a default value of 70 kg [154.3 lb] for an adult male is used), RSC is relative source contribution (the sole anticipated source of exposure is groundwater, and therefore, 80% is used as input for DBCP), UF is the uncertainty factor (10 to account for interspecies extrapolation, 10 for use of subchronic NOAEL, and 10 for potentially sensitive human subpopulations), and W is daily water consumption rate (a daily water consumption rate of 6 liter equivalents [Leq] is used because direct ingestion accounts for approximately one-third of the total exposure from household use of DBCP contaminated water, and the remaining two-third is from dermal and inhalation exposure).
The risk of noncancer health effects from drinking DBCP-contaminated water can be determined by calculating the hazard index, which is the ratio of human exposure to the REL. If the hazard index is less than 1, an adequate margin of safety exists. If the hazard index is equal to or greater than 1, the estimated exposure is equal to or greater than the REL, and further examination of the public health implications is required. Applying this method for DBCP, a hazard index of greater than 1 would be achieved when drinking water levels exceed 0.2 ppb.
Carcinogenic Effects
DBCP also causes cancer in experimental animals, and there is some suggestive evidence from human exposure studies. For risk assessment purposes, the development of squamous cell carcinomas of the stomach in female mice is used to calculate a carcinogenic potency of 7 (mg/kg-d)-1. To calculate the cancer potency, the multistage model was fit to the animal carcinogenicity dose-response data, and the 95% upper confidence limit on the linear term (q1*) was used. This estimate in animals is adjusted to a lifetime potency, assuming that potency tends to increase with the third power of the observation time in a bioassay. The estimate of lifetime animal carcinogenic potency is converted to an estimate of potency in humans by the factor (70 kg/animal body weight)1/3. This conversion follows from the assumption that a dose rate calculated as daily intake of DBCP divided by (body weight)2/3 has the same potency in rodents and humans. Using this cancer potency, the calculation of an REL (C) for DBCP in drinking water using the cancer endpoint follows the equation:
where BW is adult body weight (the default of 70 kg [154.3 lb] for an adult man), R is de minimis level for lifetime excess individual cancer risk (a default of 10-6), CSF is cancer potency (q1*) of 7 (mg/kg-d)-1 for the development of squamous cell carcinomas of the stomach in female mice, and W is daily volume of water consumed in liter equivalents (Leq) per day.
Therefore, for DBCP, an individual excess cancer risk of 1 × 10–6 (1 in 1 million) would be exceeded when drinking water levels are above 1.7 ppt. It is clear from the results of this risk assessment that the drinking water level considered more health protective is the one based on the cancer endpoint.
DISCUSSION
Quantitative risk assessment has been the foundation for environmental decision making in the United States for almost 40 years. If risk assessment and risk management are to remain the key factors in environmental decision making, “value” choices in the risk evaluation process should be made explicit, and policymakers must recognize the limitations of quantitative risk assessment. Furthermore, the design and results of the risk assessment must be described clearly in the context of the environmental problem. In other words, the context within which the “science” of risk assessment is performed should shape how scientific information is used and interpreted.
Limitations of Using Risk Assessment for Environmental Decision Making
There is an ongoing debate concerning the limitations of using risk assessment results in environmental decision making. The primary complaints include
1. Risk assessment is not solely “science based” but incorporates judgments and values that are limited by a high degree of uncertainty.
2. Conventional risk assessment methods do not account for the disproportionate risk burdens borne by certain communities, nor do they account for the impacts of cumulative and multiple exposures in toxic hot spots or to groups of people (eg, farm workers and their families).
3. Risk assessment as a two-tiered approach separates risk assessment from management as a means to insulate the “objectivity” of risk assessment from value-laden management decisions. This approach is criticized by scientists and philosophers of science for being unrealistic in that no practice of science is purely objective. Some social scientists argue that risk assessors cannot be completely immune to the political factors of the institutions within which they operate.
4. Risk assessment leads to regulatory delays; that is, “paralysis by analysis.”
5. Focusing on the quantitative aspects of risk does not provide enough information on the qualitative aspects, such as anxiety about the future, involuntariness of exposure, and equity concerns.
6. Risk assessment is used primarily to justify certain amounts of pollution, whereas the goal should be pollution elimination, prevention, or environmental sustainability (ie, leaving sufficient resources and a clean environment for future generations).
7. The process is disempowering (undemocratic) and often neglects the public participation and social values needed to make good decisions about environmental priorities. Inclusion of “risk communication” in the latter stages of the risk assessment process not only is a poor use of an important information resource (ie, the affected community itself), but it also clouds the process, making it difficult to understand and reproduce.
8. Environmental decisions based on risk comparisons with regulatory benchmarks often are viewed with skepticism by those who are affected the most. This is particularly problematic when those who are at the greatest risk do not substantially benefit from the stressor.
Does the Dose Really Make the Poison?
Students of toxicology will no doubt read somewhere in a textbook that “the dose makes the poison.” While there are applications where this statement holds true, in general it oversimplifies what we know of the toxicity of chemicals in living organisms. This often leads to misunderstanding by laypersons or misuse by some in an attempt to downplay the impact of environmental pollutants and other chemicals on humans. Although the phrase the dose makes the poison has applicability for laboratory experiments where all variables are tightly controlled, there are some notable exceptions. The timing of exposure during pregnancy rather than the dose is more critical for chemicals that cause birth defects; therefore, it is the timing that makes the poison for these chemicals. As noted previously, chemical carcinogens that cause genetic damage or mutations in DNA are thought to have no safe dose; therefore any dose makes the poison for these chemicals. Other chemicals trigger receptors in cells at very low doses and can change the activities of the cell or the signals to other cells.
For humans, there are additional reasons why the statement “the dose makes the poison” does not adequately address the risk of health damage. For example, the statement does not account for the wide-ranging variations in the human population, including sensitive, susceptible, and vulnerable populations or individuals. For example, human defense and repair mechanisms will vary in individuals depending on factors such as age, physical state, gender, race, nutritional status, etc. Therefore, the effective toxic dose will not only vary from person to person, it could also vary within an individual. Furthermore, no individual is exposed to a single chemical from a single source from a single route of exposure at the same dose over a lifetime. People are exposed to multiple chemicals in a limitless number of combinations and doses daily such that over a lifetime (starting at least at conception) it is likely the doses required for an individual chemical to exert toxicity will be highly variable.
Finally, carcinogens and some chemicals that cause noncancer health effects even at the lowest doses (eg, lead) do not exhibit thresholds for toxicity. For these chemicals, determining a level that “won’t hurt anyone” requires a risk-based (probability-based) evaluation and by definition this is a subjective (not science-based) determination. It must account for the value system of the person being impacted. In other words, people will rightfully have different opinions regarding what level of risk is acceptable to them depending on their own values. Under these circumstances, the dose that “makes the poison” is subjective and dependent on an individual’s own personal tolerance and acceptance levels.
Individual versus Population Risks
Some risk assessments or decisions based on risk assessments rely on measures of population risks; that is, measures of the additional incidence of some adverse impact in the affected population. In this situation, assessing and comparing risks for a potentially hazardous situation using population risks alone might not identify it as an environmental priority. For example, if arsenic were to leach from an abandoned toxic waste site into a nearby waterway, it could present alarmingly high individual risks. The total population risk associated with this situation, however, might be very small if only a small number of people depended on that water supply. A circular construct emerges: Waste sites and industrial facilities that often are located in poor communities and communities of color are not subject to stringent intervention or remedial action because the population risks (as opposed to individual risks of those exposed) are seen as minimal. By using population risk as the benchmark, policymakers might justify not taking action on the basis of the lesser benefits of mitigation to the overall population. Using average population risk for ranking without also looking at maximum individual risk is an economic or policy choice, not a “scientific” decision.
The use of aggregate statistics and population risk measures does not routinely account for “hot spots,” that is, geographic areas where residents experience greater environmental risks or locations where multiple exposures to hazardous substances and associated risks occur over time. In addition, risk assessments do not routinely account for differences in individual susceptibilities to toxic substances and chemical-chemical interactions in mixtures. Some attempts have been made by the EPA to develop guidance to incorporate these and other considerations in the risk assessment process. Nevertheless, inclusion of these issues is not yet widely practiced.
Public Involvement
Collaboration among the business community and industrial sector, the general population, and government agencies is required for effective involvement of the public. Although public participation is now generally accepted in diverse policy fields, it is still not addressed adequately in science-based environmental decision making such as risk assessment and risk management.
Environmental agencies should develop and implement plans to involve the public in the decision-making process and recognize that public participation can be seen as a solution to some environmental problems in and of itself, but only when the public is involved as a full and equal partner, not as an adversary. This includes maximizing meaningful participation in the review of agencies’ activities and progress in accomplishing the objectives of promoting long-term planning for sustaining a healthy environment and workplace. To accomplish this, public participation needs to be initiated early in the hazard evaluation process and incorporated into the decision-making process. Furthermore, education is a key component to effective public involvement, and therefore, technical information should be easily accessible to the public and translated, if necessary, into the residents’ and workers’ primary language(s).
Research Needs
More research needs to be done to better understand the risks that environmental and workplace pollution poses, including
1. Completing the toxicity database for many substances released in large quantities into the air, water, land, and workplace or as contaminants in food and other consumer products.
2. Making available data describing actual human exposures to most pollutants.
3. Developing risk assessment methods further. For example, methods to assess cumulative risk from multiple chemical exposures and the effects of chemicals on the endocrine, nervous, and immune systems are necessary to understand better the full spectrum of hazards posed by environmental pollutants and occupational hazards.
4. Considering subpopulations that bear disproportionate risks (that is, “hot spots”), which must be incorporated into any new and/or existing site-specific risk assessments.
5. Developing methods to assess the societal distribution of environmental and occupational health risks in the context of achieving environmental justice.
6. Devoting resources to measuring population exposures to toxicants, including from microenvironments, from accidental releases, and among highly exposed groups.
7. Increasing the capacity to identify and prevent future impacts on public health and the environment from emerging risks.
Other Models for Environmental Decision Making
Applying scientific knowledge and judgment to address environmental issues requires universal strategies as well as some fundamental changes in the status quo of environmental decision making. In other words, more consideration should be given to alternative science or value-based processes proposed or used to address environmental and occupational hazards.
One alternative model used to support environmental decision making, predominantly in European countries, is the precautionary principle. This approach does not exclude making estimates of risk, but the burden of proof is levied on the polluter rather than the affected public. In fact, it has been argued that the precautionary principle should be viewed as a complement to science to be invoked when a lack of scientific evidence means that the outcomes are uncertain. In applying the precautionary principle, ethical and value-based aspects should be weighed equally with the science. The key element to the precautionary principle is that action should be taken in the face of uncertainty rather than delaying action until more “evidence” is generated.
Other options include technology-based approaches that require retooling or reformulating industrial processes to use fewer or lesser amounts of hazardous materials or by substituting them with safer alternatives. The EPA is already mandated to incorporate pollution prevention into its implementation plans under the Toxic Substances Control Act and the Clean Air Act, whereas the reduction or elimination of hazardous pesticide use has lagged behind. These approaches apply the principles of hazard identification without necessarily relying on a risk-based assessment because the ultimate goal is to achieve elimination of hazardous materials and prevention of environmental and workplace exposures. In banning the chemicals DDT, polychlorinated biphenyl (PCB), and lead in gasoline, pollution prevention is achieved without allowing for some level of “negligible risk."
Public pressure, public right-to-know laws, and civil suits also have achieved a certain degree of success in influencing environmental decision making. For example, California’s Proposition 65, approved by a wide margin in 1986 as an initiative to address growing concerns about exposures to toxic chemicals, is an example of a public right-to-know law that also empowers citizens to “blow the whistle” on polluters. Currently, more than 700 chemicals are listed as reproductive or developmental toxicants or carcinogens. Proposition 65 is an effective mechanism for reducing certain exposures that may not have been controlled adequately under existing federal or state laws. It also provides a market-based incentive for manufacturers to remove listed chemicals from their products. Furthermore, because of Proposition 65, information regarding the dangers of exposure to certain chemicals in more susceptible subpopulations is widely disseminated. In 2005, California passes another right-to-know law, the California Safe Cosmetics Act, which is the first law in the country requiring manufacturers of cosmetic products to publicly disclose harmful ingredients used in their products. Almost 100 chemicals known or suspected to cause cancer, reproductive effects, and/or birth defects are used in cosmetic product formulations.
REFERENCES
Biomonitoring California: http://oehha.ca.gov/multimedia/biomon/index.html.
California Safe Cosmetics Program: http://www.cdph.ca.gov/programs/cosmetics/Pages/default.aspx.
Cote I: Advancing the next generation of health risk assessment. Enviorn Health Perspect 2014;120:1499 [PMID: 22875311].
Mumtaz, M: Application of physiologically based pharmacokinetic models in chemical risk assessment. J Toxicol 2012;2012:904603 [PMID: 22523493].
National Toxicology Program, High Throughput Screening Initiative: http://ntp.niehs.nih.gov/?objectid=05F80E15-F1F6-975E-77DDEDBDF3B941CD.
Office of Environmental Health hazard Assessment. Public health goals for drinking water. http://www.oehha.ca.gov/water/phg/allphgs.html.
Proposition 65: http://www.oehha.ca.gov/prop65.html.
Steenland K: Risk estimation with epidemiologic data when response attenuates at high-exposure levels. Environ Health Perspect 2011;119:831 [PMID: 21220221].
U.S. Environmental Protection Agency, Cancer Risk Assessment Guidelines: http://www.epa.gov/cancerguidelines/.
U.S. Environmental Protection Agency, Environmental Laws and Regulations: http://www.epa.gov/lawsregs/index.html.
SELF-ASSESSMENT QUESTIONS
Select the one correct answer to each question.
Question 1: Risk
a. is the anxiety that an event will result in loss or harm
b. may be thought of as voluntary but not involuntary
c. does not include the probability of adverse health effects
d. can apply to almost any activity or event
Question 2: Risk assesment
a. is solely “science-based”
b. incorporates only values with a high degree of certainty
c. avoids regulatory delays
d. is a process and not a science
Question 3: Exposure assessment
a. is used to estimate the magnitude and probability of uptake from the environment by any combination of oral, inhalation, and dermal routes of exposure
b. presents results in qualitative, not quantitative, terms
c. identifies the population at risk by determining who has elevated blood levels of toxic chemicals
d. does not need to consider proximity of the population to the source
Question 4: A precautionary approach to decision making
a. avoids long delays in taking action when there is uncertainty in the existing data
b. supercedes risk assessment as a decision-making factor for federal regulators
c. requires government to prove harm before it takes action
d. is supported by the chemical industry
Question 5: A risk management decision
a. involves evaluating the impact of risk assessment on medical research funding
b. is solely based on empirical data generated by impartial scientists and analysts
c. requires an impacted population or individual to assign an acceptable risk factor of their exposure
d. considers risk along with cost, technical feasibility, societal benefits, and political climate
Question 6: Reference exposure levels (RELs) are
a. defined as median exposure levels below which no adverse health effects are anticipated
b. derived by identifying and dividing the NOAEL (or BMD) by uncertainty factors
c. not modified by differences in physiology and toxicology between species
d. solely based on empirical data generated by impartial scientists and analysts