Click for next page ( 2


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 1
1 Executive Summary Risk assessment of chemicals constitutes the linchpin for many regu- latory decisions. In this sixth volume of Drinking Water and Health, the Safe Drinking Water Committee reviews the state of the art in the de- veloping field of risk assessment and provides some risk assessments for drinking water exposure to 14 compounds of interest to the U.S. Envi- ronmental Protection Agency (EPA3 as the agency develops its regulatory program. Reflecting the tendency of past toxicological studies to focus on cancer as the main health effect of concern, most of the risk assessments in this report were estimated for carcinogenicity. Wherever data were adequate, however, the committee also developed risk assessments for noncancer end points. In 1977, the first volume of this series contained the advice that both carcinogenicity and other health effects should be evaluated when studying exposures to contaminants in drinking water and elsewhere. Until quite recently, however, the assessment of these noncancer diseases has re- mained fairly primitive, characterized by relatively little work on mod- eling, methods, and materials. For some common and widespread effects of exposure, such as reproductive and developmental impairments and neurotoxicity, risk assessments have been conducted only infrequently. A major portion of this report is devoted to an overview of these noncancer diseases and general advice on methods for estimating risks associated with chemical pollutants. There is also extensive discussion of the many factors that must be considered in the risk estimation process. The committee has thus laid the groundwork for innovative approaches to assessing cancer and noncancer risks, including the assessment of risks 1

OCR for page 1
2 DRINKING WATER AND HEALTH to development, reproduction, and necrologic functioning. The committee also assessed methods for using epidemiological data in conducting risk assessments and methods for extrapolating results of laboratory animal inhalation studies to estimate human risk from ingestion. In addition to the use of safety factors, the committee has described mathematical mod- eling approaches currently available or under development for the as- sessment of noncancer risks. DEVELOPMENTAL TOXICITY Developmental effects encompass embryo and fetal death, growth re- tardation, and malformations, all of which can be highly sensitive to chemical exposures. Up to 25% of all pregnancies end in spontaneous abortion, but relatively little is known about the causes of these terminated pregnancies. When defects that become apparent only later in life are included, the frequency of major and minor malformations in the U.S. population increases to about 16%, but here again, few causes have been firmly established. Applying the results of developmental toxicity studies to risk assessment remains problematic, because the underlying events leading to impaired development are very complex. Some of these events may be reversible, whereas others clearly are not. Moreover, the most appropriate animal models for extrapolating developmental toxicity data to humans have not yet been identified. Major factors that influence developmental toxicity include timing of exposure and different patterns of dose response for growth retardation, embryo death, and teratogenicity in laboratory ani- mals. These must all be considered when examining data from mammalian, nonmammalian, and in vitro studies to estimate human risks from devel- opmental toxicants. Risk assessors must also determine whether to select a no-observed-effect level (NOEL) or the lowest-observed-effect level (LOEL) and must choose a model for estimating the risks of developmental toxicity. These decisions depend on the extent of the data, whether data on humans are available, and the potential associated hazard. The com- mittee recognized that several different models (e.g., the one-hit model) could be used for estimating developmental toxicity but that much addi- tional research needs to be done to establish their relative biological re- liability. REPRODUCTIVE TOXICITY Reproductive toxicity may affect anything within a continuum of events, ranging from germ cell formation and sexual functioning in the parents through sexual maturation of the offspring. The relationship between ex- posure and reproductive dysfunction is highly complex, because exposure

OCR for page 1
Executive Summary 3 of the mother, father, or both may influence reproductive outcome. In addition, critical exposures may include maternal exposures long before or immediately prior to conception or exposure of the mother and fetus during gestation. Sexual development and maturation of the male and female reproductive systems are highly susceptible to general environmental exposures, but they involve vastly different durations and stages of vulnerability. The committee noted that a variety of xenobiotic substances, including poly- cyclic aromatic hydrocarbons and certain antineoplastic agents (e.g., adri- amycin, 5-fluorouracil, bleomycin, vincristine), have been associated with reproductive toxicity in animals, but the applicability of these findings to humans is not known, since only limited relevant studies have been con- ducted in humans. Risk assessment of reproductive toxicity necessarily relies on the results of in vivo and in vitro models of animal studies during the distinct stages of reproduction when important developments occur. Mutagenic activity in the germ cells of test animals must be regarded as evidence of potential germ cell toxicity in people. Demonstrated ability to cause somatic cell mutation, combined with evidence that a substance can reach and interact with the germ cells, is evidence of potential mu- tagenicity in humans. Thus, the committee recommends that a battery of short-term tests, including the dominant lethal assay and the heritable translocation test, be conducted to confirm activity in germ cells. Among the factors that should be studied are DNA damage and repair in sper- matocytes, sperm morphology, and distribution and binding sites. Un- fortunately, there have been few attempts to develop assays for measuring chemically induced mutations in the germ cells of females. Accurate information concerning the potential mutagenic hazards of chemicals in humans has been difficult to obtain for several reasons, but primarily because genetic alterations such as recessive mutation are not easily observed in humans. Chromosome damage in the embryo may be the undocumented and unrecognized cause of an unknown proportion of spontaneous abortions. Two factors complicate the assessment of mutagenic risk for humans: there has been little systematic thinking about the factors that influence the determination of safety factors to be applied to experimental data, and mathematical models of the mutagenic process often cannot be applied because of insufficient experimental data. Consequently, the committee did not estimate risks of germ cell mutagenesis for the compounds re- viewed in this report. NEU ROTOXICITY The nervous system is especially vulnerable to chemical insult. This recent finding has led to an increased interest in the toxicology of that

OCR for page 1
4 DRINKING WATER AND HEALTH system. More than 850 chemicals have been identified as known neuro- toxicants in humans or animals; many of them are pesticides. Of the 350,000 cases of pesticide intoxication that occur worldwide annually, most have involved some form of neurotoxicity. Neurotoxicity data have served as the bases for threshold limit values (TEVs) recommended by the American Conference of Governmental Industrial Hygienists for 30% of all frequently encountered industrial chemicals. I;eurotoxicity in humans encompasses a vast variety of effects, ranging from cognitive, sensory, and motor impairments to immune system def- icits. Because of this complexity, classification of chemical neurotoxic action is constantly evolving. Although tentative, broad, and subject to change as knowledge increases, a classification by cellular target and period of peak vulnerability has been proposed by the committee. As with developmental and reproductive toxicity, major considerations affecting risk assessment for neurotoxicants include the permanence of effects, the amount and timing of exposure, the mechanisms underlying the toxicity, and heightened sensitivity of various subpopulations. Also, as for other chronic noncancer diseases, neurological impairment has rarely been studied in humans (with the important exception of lead-poisoning studies and the adverse effects of a few clinical agents). Nevertheless, there are valid in viva and in vitro models, and others are being developed. Human neurological disorders can be modeled in many animals given doses of chemical substances for a test period simulating human exposure. In the absence of data on humans, therefore, convincing demonstration that a substance is neurotoxic in a test animal can be regarded as evidence that the agent may be neurotoxic in humans. The application of data from studies in animals to estimate human risks of necrologic disease is complicated by several factors: major differences in the degree to which neurotoxic responses occur in animals and in humans as a result of exposures to comparable levels of a given chemical; the existence of thresholds for some- end points in some species that may not be clearly expressed in other species; incompleteness of information on dose-response relationships for most environmental chemicals (unlike that for therapeutic agents); and the influence of timing and exposure patterns on outcomes. For these reasons, the committee advised caution in assessing risk for neurotoxicants and suggested that such assessments might best be applied to distinct necrologic processes and segments of the nervous system. CARCINOGENESIS Current theories and experimental research in animals indicate that there are at least three distinct stages in the development of carcinogenesis: initiation, promotion, and progression. Each of these stages apppears to

OCR for page 1
Executive Summary 5 be influenced by such factors as age, heredity, diet, metabolic activity, and previous and current exposures to xenobiotic compounds, which may interact and potentiate, antagonize, or synergistically influence the de- velopment of carcinogenesis. The committee examined the mouse skin as a model for chemical carcinogenesis and the model's additional utility for studying the mechanisms of action of initiators and promoters. Rat and mouse liver models have also led to useful discoveries about stages of tumor development, and promotion has been found in other tissues, such as the lung and breast. Some promoters, such as asbestos, estrogens, and some constituents of cigarette smoke, appear to be tissue-specific, whereas others, such as 2-AAF (2-acetylaminofluorene), can act as promoters in the mouse bladder but as initiators in the liver of the same animal. The effect of a promoter can vary, depending on the sequence in which it is administered with the initiator. Therefore, it may be desirable to test potential promoters and inhibitors of promotion in vivo in several species and strains, to examine several organs for the response, and to test different sequences of administration. Recent investigations into the role of oncogenes seem to support the multistage theory of carcinogenesis. Of the approximately 20 oncogenes thus far discovered, most have been either of two types, immortalizing genes or transforming genes, which may be regarded as the biological counterparts of initiation and promotion in tumor induction. A third type of oncogene recently identified may cause a cancer cell to metastasize. Ideally, then, the risk modeler should consider at least three stages of tumor induction to be consistent with human and experimental evidence. Risk models must also take into account the diversity of promoting agents and their multiplicity of action at the cellular level. The immune system is particularly sensitive to the toxic effects of xenobiotic agents and can be suppressed or overstimulated by both non- carcinogens and carcinogens. Either effect may lead to increased cancer incidence. Suppression of the immune response typically produces de- creased resistance to viruses, parasites, bacteria, and tumor cell grafts, whereas enhancement results in the development of autoimmune diseases and hypersensitivity reactions. The most widely used tests for carcinogen evaluation and regulation are long-term animal bioassays, which are considered to be qualitative predictors of response in humans. However, animal bioassays are not only time consuming but also are difficult to conduct, are extremely expensive, do not necessarily predict the target of action in humans, and cannot duplicate the wide range of susceptibilities, simultaneous exposures, and genetic backgrounds of the human experience. Short-term cell culture in vitro assays, requiring only a fraction of the time, can test for toxic end points manifested during the initiation phase of carcinogenesis in the widely accepted multistage theories of carcino-

OCR for page 1
6 DRINKING WATER AND HEALTH genesis. Other tests, such as cell transformation assays, examine toxic end points for later events in carcinogenesis. There is a need to develop short-term tests for promotion. The committee concluded that a battery of tests is necessary for screening chemicals for genetic toxicity. It cautions that false-negative and false-positive results may occur because the tests do not reflect the complexity of interactions in the whole animal and that although knowledge of the chemical's prin- cipal mechanism of action is crucial to the selection of short-term tests with meaningful predictive value, this knowledge is rarely available. De- spite the fact that in vitro tests are technically difficult to design and interpret, the committee believes that they have considerable potential and should be developed further. DOSE-ROUTE EXTRAPOLATION It is difficult to estimate the risks of exposure to volatile compounds in drinking water because of the nature of their chemical properties. The principal route of exposure is ingestion, although the dermal and inhalation routes may account for fairly large exposures during bathing, showering, cooking, and other activities involving the use of water. Volatile com- pounds cannot easily be added to water for testing purposes because their solubility in water is limited. In addition, the volume of water consumed by a rodent each day is insufficient to provide an adequate simulation of the water-conveyed exposure of humans for many volatile compounds. Consequently, a substantial amount of data on many volatile substances derives from subchronic or chronic toxicological studies of inhalation exposures. When results of inhalation toxicity studies in animals are used to predict the risks associated with human consumption of contaminated drinking water, both a dose-route and interspecies extrapolation of the toxicity data must be performed. The committee found that inhalation studies, in which chemicals are absorbed at fairly uniform rates over a specified exposure period, are not very different from ideally designed drinking water studies, in which chemicals are absorbed at a variable but moderate rate over the course of a day. In fact, well-designed inhalation toxicity studies may serve as excellent experimental models for deriving drinking water stan- dards for a variety of volatile chemicals. The committee developed an innovative model for extrapolating from inhalation to ingestion exposures, taking into account the metabolic and pharmacokinetic differences between these different primary routes of exposure. The model also incorporates time-dependent physiological and metabolic changes, scaling data from the studies on the rat up to a standard, healthy, young adult human male. The committee did not address problems

OCR for page 1
Executive Summary 7 of extrapolating from healthy rodents to very young or old humans or to other susceptible human subpopulations. THE USE OF EPIDEMIOLOGICAL STUDIES IN RISK ASSESSM ENTS The strength of epidemiology lies in its use of direct human observa- tions. Its major limitation is that Epidemiological observations of humans are rarely made under adequately controlled conditions. Accordingly, the design, analysis, and interpretation of Epidemiological studies require spe- cial care to ensure that observations are valid and that populations under study are comparable. Epidemiological studies are best suited to confirm- ing past risks, but they can also be used for estimating future risks when the substance of concern resembles another substance for which adequate data exist. Epidemiological studies are also valuable for generating hy- potheses about possible etiologic agents, but they can rarely pinpoint precise causes. Negative Epidemiological studies require careful interpre- tation, especially where low-level exposures or risks are involved. The committee noted that for many chemicals to which people are exposed, few or no Epidemiological data are likely to become available because of these limitations. Epidemiological studies are of value principally for conducting quali- tative risk assessments of specific chemicals or industrial processes. Quan- titative risk assessments are possible only when the data have been generated in well-conducted studies of well-defined cohorts whose exposures to a specific substance have been carefully measured or estimated and when sufficient time has elapsed from first exposure to the expression of disease. Such conditions are rarely encountered, however, since much human ex- posure involves complex mixtures. There are even problems in developing risk assessments for the general population from data on well-studied industrial hazards. For example, one cannot directly extrapolate from studies of healthy workers to the general population, which includes hy- persusceptible people, the very young and the elderly, and others of less than robust health. Moreover, industrial exposures occur over an 8- to 10-hour workday, ceasing on weekends, whereas environmental exposures can be lower but continuous. The committee also noted that epidemio- logical studies of small populations are seldom able to detect risks from low exposures, which may nevertheless have significant long-term public health effects. A FRAMEWORK FOR RISK ASSESSMENT The theoretical basis for risk assessments discussed in the first Drinking Water and Health report has been extended considerably by the present

OCR for page 1
~ DRINKING WATER AND HEALTH committee's detailed consideration of cancer and noncancer diseases that should be included in risk assessments. The first volume distinguished between injuries produced by chemicals likely to have a threshold and those for which no threshold can safely be presumed to exist: "It is more prudent to treat some kinds of toxic effects that may be self-propagating or strictly cumulative, or both, as if there were no threshold and to estimate the upper limits of risk for any given exposure." Included among self- propagating effects were alteration in cellular genetics transmitted by cell propagation and injuries such as the destruction of neurons. For effects other than carcinogenicity and mutagenicity, earlier com- mittees and the present committee have relied on widely accepted safety factors that reflect the level of uncertainty associated with a particular set of toxicological data: an uncertainty factor of 10 is applied when there are valid results obtained from humans, a factor of 100 is applied when there are data from valid animal studies but human data are not available or are scanty, and a factor of 1,000 is applied when human data are not available and where the animal data are very limited. As in the past, the committee endorsed basing the assessments on a 70-kg human and an average daily consumption of 2 liters of drinking water. The committee agreed that noningestion exposure (such as exposures that occur during bathing) may be important, but it did not directly estimate such exposures. Although risks were not assessed for the 10-kg child who drinks propor- tionately more water (1 liter/day) than the adult, the committee noted that risks posed by carcinogens in drinking water would be higher for children than in adults. The committee relied on conversion factors based on body surface area rather than on body weight for extrapolating animal data to humans, since effects have been more directly correlated to that factor in cancer che- motherapy studies. For the noncancer end points considered by the com- mittee, much additional work must be done, both in the identification of suitable animal models for estimating human risk and in the development of theoretical models for using animal data, once obtained. The committee is encouraged by recent activity in these areas. The committee concluded that data from animal studies are useful in qualitative risk assessment and for setting priorities. When such data are used for quantitative estimates of risk, there is less consensus about the most appropriate models, data sets, and conversion factors. Despite these uncertainties, however, there are compelling arguments favoring the use of animal data for quantitative risk assessments. In assessing risk of exposure to carcinogens, the committee noted the importance of interspecies variations, interindividual variability in the human population, and a variety of other factors identified by the Office of Science and Technology Policy in 1984 and discussed by the State of

OCR for page 1
Executive Summary 9 California and the EPA in their development of guidelines on cancer risk assessment at the end of the same year. The committee also examined models for extrapolating from high animal doses to estimate the responses of humans at more frequently encountered low exposures to carcinogenic materials. These models include the tolerance distribution models (logistic, log-normal, probity, the so-called hitness models, time-to-tumor occur- rence (or to death from cancer) models, and models derived from the multistage hypothesis of cancer causation (e.g. the Armitage-Doll model). A recently developed mathematical model (i.e., the Moolgavkar-Knudson model) extends the multistage concepts by adding the concept of a dif- ferential birth-death process for cells in a preneoplastic stage. The com- mittee believes that this model has the potential for improving quantitative risk assessments for cancer-causing materials. Most experts agree that current techniques for assessing cancer risk cannot generate a single precise estimate of human risk and that risks may best be expressed in terms of ranges or confidence intervals. One way of ensuring that such estimates are not overinterpreted is to state the as- sumptions underlying specific assessments and to discuss the uncertainties surrounding the numerical estimates, including both point estimates and upper confidence limits for all extrapolation models. The committee noted that these models diverge chiefly on the matter of low-dose extrapolation. For carcinogens assessed in this study, the committee generally relied on the multistage model, believing that this model most nearly reflects biological mechanisms of cancer, despite the existence of substantial individual differences in sensitivity to carcinogens and potential for de- toxification. TOXICOLOGICAL ASSESSMENT OF 14 COMPOUNDS Quantitative risk assessment includes four distinct components: hazard identification, exposure assessment, dose-response assessment, and char- acterization of human risk at projected levels and patterns of exposure. Such assessments have been done for approximately 205 drinking water contaminants in the previous five volumes of this series. In this volume, compounds that demonstrate important toxicological issues within the research community were selected from those on the regulatory agenda of the EPA. In Chapters 2 through 8, the committee has suggested several methods for risk assessment. Examples of the use of most of these methods are provided in the evaluations of individual contaminants. Pharmacokinetic principles have not been applied to risk assessment, however, due to a general lack of data to support this exercise for the compounds reviewed in this volume. The committee hopes that potential gains from using

OCR for page 1
TO DRINKING WATER AND HEALTH pharmacokinetic information to extrapolate response from high to low doses can be explored in the future to improve risk estimation techniques. The 14 compounds reviewed by the committee were, in order of dis- cussion, acrylamide, aldicarb, diallate, sulfallate, dibromochloropropane, 1,2-dichloropropane, 1,2,3-trichloropropane, 1,3-dichloropropene, di(2- ethylhexyl) phthalate, mono(2-ethylhexyl) phthalate, ethylene dibromide, nitrofen, pentachlorophenol, and trichlorfon. For each compound, the committee reviewed all available data on metabolism, health effects in humans and laboratory animals, mutagenicity, carcinogenicity, and ter- atogenicity. Cancer risk was estimated for several substances, and when data were adequate, the committee also developed risk assessments for noncancer health effects. When adequate evidence suggested that a chemical was not likely to be a carcinogen or a mutagen, the committee identified levels of exposure that would not be expected to cause adverse health effects in humans, i.e., suggested no-adverse-response levels, or SNARLs, using the uncertainty factors described above. The committee realized the ex- perimental difficulty of determining NOELs. In addition, a different NOEL or LOEL may be determined for each end point examined; the more sensitive end points will have lower values. As discussed in Chapters 3 and 8, for example, data on reproductive toxicity produced lower estimates of LOELs than did data from subchronic toxicity studies for 35% of compounds tested, the same estimates for another 35%, and higher esti- mates for 30%. For some compounds reviewed in Chapter 9, the data base was sufficient for estimating the magnitude of inter- or intraspecies variability and, thus, for selecting a safety factor on the basis of that estimation. For carcinogens, the committee used the multistage model to estimate both the maximal likelihood estimate and the upper 95% confidence limit of risk associated with a daily exposure to 2 liters of drinking water containing 1-~g/liter concentrations of the carcinogen. The committee recognizes that the assessment and the management of risk are two distinct responsibilities; the former is chiefly scientific, whereas the latter requires policy judgments beyond the purview of any scientific committee. Thus, in the development of standards, the risk estimates provided in this report are offered as general guidance to policymakers for evaluating the risks to the public posed by exposures to contaminants in drinking water.