Drinking Water and Health,: Volume 1 (1977)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

if Chemical Contaminants: Safety And Risk Assessment The hazards of ingesting pollutants in drinking water can be assessed in two general ways: with studies of toxicity in the laboratory and with epidemiological studies. Studies in the laboratory may employ a variety of experimental systems, ranging from chemical ejects of pollutants on DNA, through exposure of bacterial or mammalian cells in culture, to lifetime feeding studies on experimental animals. They are prospective studies, in which relatively small numbers of cells or animals are exposed to characterized pollutants at known concentrations. Epidemiological studies deal with human populations. Their design is constrained by external circumstances, and they involve large numbers of people whose exposure to the pollutant in question is commonly uncertain and confounded by exposure to other pollutants. The aim of studies of both types is to allow the risk to man to be estimated. The first can give precise information on relatively high risks related to individual pollutants in this or that animal species to which human exposure and risk may be compared. The second can provide less precise information on the human risk related to one pollutant (isolated, it is hoped, from other pollutants). Toxicity data obtained from laboratory animals will generally have to be relied on for estimating human risk, if we are to control human exposure to carcinogens. Epidemiological studies have discovered causes of disease and can buttress, supplement, or contradict laboratory data. Imaginative comparisons between laboratory and epidemiological data 19

20 DRINKING WATER AND H"LTH are of the utmost importance, particularly in the area of metabolic pathways and fate of chemicals found to be carcinogens in animals. Efforts to develop rapid assays for mutagenesis and carcinogenesis have recently been greatly expanded. Methods that show promise include tests for mutagenicity that make use of bacterial, cell transformation, and organ culture systems. There appears to be high positive correlation between mutagenicity, as determined by some of these methods, and carcinogenicity in agents already studied (McCann et al., 1975; McCann and Ames, 1976; Ames, 1976~. The utility of these rapid methods will depend on experimental demonstration that their results are well correlated with those obtained from conventional long-term studies of carcinogenicity with well-designed animal systems. High priority should be given this research because it offers a reasonable probability of success in a relatively short time and at lower cost than long-term testing, and there is an urgent need for a primary screen for selecting compounds for long-term assay (DHEW, 19771. The committee is fully conscious of these modern methods for determining genetic and physiological phenomena. Their use, when appropriate, and their further development is strongly encouraged. Pollutants in water have many different effects. At- one extreme, they can impart a disagreeable taste or odor. This is quickly perceived by the community, and the process of characterization and identification of the offending pollutant is generally prompt and fairly straightforward. At the other extreme, the ejects on human health of a carcinogen present in drinking water will probably go undetected, particularly if it produces only a modest in-crease in the incidence of a common cancer. The major toxicological and epidemiological efforts should therefore be directed to characterizing and identifying pollutants whose biological effects include chronic, irreversible, and progressive diseases, such as cancer. It is necessary to develop risk estimates for large human populations of varied susceptibilities that are exposed to small concentra- tions of such toxic pollutants, including carcinogens. The development of safety factors for pollutants whose toxic ejects are reversible and nonprogressive involves empirical calculations based on past history of use and concentrations that appeared safe for the public. These safety factors are usually applied to the highest dose or concentra- tion at which no adverse eject was observed. The chosen dose or concentration is divided by a "safety factor" that varies over a wide range, depending on the adequacy of the data. Whether or not the "safety factor" approach can be used with pollutants that cause chronic, irreversible, and progressive disease in laboratory animals is controversial. Those who argue for safety factors,

Chemical Contaminants: Safety And Risk Assessment 21 and thereby thresholds, find it inconceivable that very small concentra- tions can cause a cancer to develop, claiming that body defenses can surely protect at doses smaller than the threshold value. Those who argue that safety factors are inadequate and that almost no thresholds can be determined, or theoretically developed, suggest that even one or a very few molecular events have a finite probability of initiating a successful malignant or neoplastic transformation in a cell, and that this can lead to a lethal cancer. Although one malignant cell can lead to death by cancer, many liver or kidney cells can be killed or damaged (but not malignantly transformed), without causing any detectable disease. Furthermore, man is never exposed to one carcinogen at a time, but is exposed to low concentrations of many at the same time. Accordingly, we have adopted a "nonthreshold" approach for estimat- ing risks from pollutants that have been shown to be carcinogenic in laboratory animals. Demonstration that a pollutant is carcinogenic, and application of nonthreshold risk estimates to it, do not imply that its use must be prohibited. Such a proscription might itself give rise to even greater risks to health or other disadvantages. In some cases, a net risk must be estimated, and society must attempt to use the pollutant in such a way as to minimize risk and maximize benefit. Nowhere is this better illustrated than in the use of chlorination to disinfect water. Chlorination controls pathogenic organisms, but introduces chloroform, which is carcinogenic in animal-test systems. Methods must be devised to minimize concentra- tions of chloroform and chlorinated hydrocarbons, from whatever source, in drinking water. But before alternative methods for control of pathogenic organisms are instituted, toxicological studies must show that they are as effective as, and pose no greater risk than, chlorination. We perceive that society is willing to accept some risks to health if the attendant benefits are demonstrably greater. Drinking water contains low concentrations of many chemicals, some of which, if ingested for a long time, could have delayed toxic effects. The insidious eject of chronic exposure to low doses of toxic agents is difficult to recognize, because often there are few early warning signs and, when signs are ultimately observed, the effects may have become irreversible. Subchronic toxicology studies may not offer reliable means for assess- ment of long-term toxic effects in animals, let alone extrapolation to chronic effects in man; hence, different considerations have to be applied in assessing risk. The methods and principles of acute toxicology do not offer any easy, straightforward methods for extrapolation of such experimental data to calculate risks for large human populations. Two important questions must be answered: What assay procedures .. .

22 DRINKING WATER AND H"LTH are required for a valid assessment of chronic toxicity of chemicals in experimental animals? How can the data from such procedures be extrapolated to estimate risks in humans? In dealing with these questions, we use the specific risk of cancer as our major example, although other toxicities are considered. This report seeks to summarize the state of the art in extrapolating to man the results of experiments on animals, chiefly in relation to carcinogenesis. EFFECTS ON HEALTH The purpose of drinking-water standards is to ensure protection from acute poisoning and from long-term, or "chronic," effects. In recent years, numerous short-duration, presumptive tests in vitro have been developed that may help to predict carcinogenesis. Nonetheless, studies of chronic toxicity continue to be required for safety evaluations. These are necessary because there is no general way to predict carcinogenic elects on the basis of the observed short-term elects of chemical- biological interactions. [The significance to health of the finding that a pollutant is mutagenic in the new test systems is unknown (See Drake, 1975; DHEW, 1977~. But, because evidence of the correlation between mutagenicity and carcinogenicity continues to accumulate, we suggest that a conservative safety factor be provisionally applied to the mutagenicity data and that, if new information (such as the results of a reliable carcinogenicity study) is lacking 4 yr from the time a mutagenici- ty study is completed, nonthreshold methods be used to establish risk.] Chronic exposures and chronic effects are different (Casarett, 1975~. The former means frequent ingestion over a long period of time. The latter implies injury that persists, either because the injury is irreversible or progressive, or because the exposure is prolonged and the rate of new injury exceeds the rate of repair. Chronic exposure in animals is generally considered to be at least half the life span. In man, it can be much less. Injury from chronic exposure may occur in at least three ways: by accumulation of the chemical to a critical concentration at sites of action sufficient to induce detectable injury; by accumulation of injury until physiological reserves can no longer compensate (i.e., repair is never complete); or after a long, latent period beginning with an exposure that has an unrecognized biological elect, and precipitates the eventual appearance of injury. In the first case, knowledge of the kinetics of chemical absorption, metabolism, and excretion obtained in short-term studies may allow computation of the amount of the toxic chemical that will accumulate in long-term use. Such investigations will improve the

Chemical Contaminants: Safety And Risk Assessment 23 usefulness of long-term, low-dose, chronic-exposure studies. To predict chronic accumulation, or latent development, of injury from the results of short-term tests requires knowledge of the kinetics of injury and repair. There are few, if any, substances for which such understanding is at hand. Reversibility of Chemical Injury Reversible elects disappear after exposure ends. The time required for return to normal should be a small fraction of the remaining lifetime of an organism. During the period of return to normal, the organism must be at no greater risk (than one that was never exposed) of further or other damage from other sources. For some elects, reversibility may be qualified by the normal lifetime of a specific cell or macromolecule that serves as the end point of the elect. A nonreversible effect is one in which the damage does not regress completely, or may progress after exposure ceases. Some erects of toxic chemicals are unmistakably irreversible. They include terata, malignant tumors, mutations in offspring of exposed animals, and some neurological changes. These are gross manifestations of specific chemical-cell interactions, and it is possible, or probable, that there are early reversible effects, either in the cellular process first affected or at intervening stages. Prediction of adverse effects from short-term studies is possible if the critical dose and the rate-limiting factors that determine reversibility are known. Without this knowledge, evaluation of toxicity will generally deal more with the possibility of irreversible effects than with speculatively reversible elects. Net reversibility varies from one tissue, species, strain, or individual to another. It is generally impossible to measure the specific processes involved in injury and repair in the standard toxicity-evaluation study. However, measurements of reversibility in short-term studies should provide useful information that may allow extrapolation to the longer term. The predictive value of such tests does not necessarily depend on the persistence of the chemical in the test organism. If the chemical produces a reversible erect and then is rapidly detoxified or excreted, it may be possible to compute the doses or schedules of exposure that would not produce cumulative and ultimately nonreversible or irreversible effects (see definitions below). But other factors might be overriding. For example, rapid reversibility after a single dose might not indicate the rate of reversal after repeated doses, if the first dose, in addition to the measured erect, altered the repair process or processes responsible for detoxification (Murphy, 1967~. To evaluate repeated-dose effects, sub

24 DRINKING WATER AND H"LTH chronic- and chronic-exposure studies should include groups of animals that are removed from exposure at selected intervals during the experiment. The rate of reversal of elects in these animals can be measured at intervals, or at some critical time after exposure ceases. If the chemical persists in the organism, quantification of rates of reversal is more complicated. Data are needed on absorption and disposition to correspond with data on rates and reversal of elects. Perspectives and Perceptions of Effects Whether an elect is reversible, nonreversible, or irreversible might be shown by experiments (or epidemiological studies) that include observa- tions during exposure-free "recovery"periods similar to those made at the end of the exposure period. Nonreversible and irreversible injuries are of greater concern than reversible injuries in evaluating human health elects. However, frequently recurring reversible injury may lead to greater morbidity than a nonreversible or irreversible injury that appears only late in life. Characterization of an effect as reversible implies that there is a dose below which health will not be compromised. This assumes, of course, that any subliminal cellular injury that is responsible for the manifest elect is also reversible. Full understanding of thresholds, margins of safety or safety factors, and extrapolations of estimated risk requires understanding the underlying cellular mechanisms of injury. An alternative to the safety-factor approach for reversible toxicity may be considered. The nonthreshold approach is attractive partly because of the idea that one transformed cell could lead to fatal neoplastic disease. What number of damaged (but not transformed) or killed kidney, liver, or lung cells is compatible with a healthy life? If these numbers, or fractions of total organs, could be estimated for a number of species, including man, and if experimental dose-response curves for fractional damage to all vital organs could be obtained, the numbers of damaged cells that are compatible with health could be estimated. This might constitute an initial approach to the development of rational risk estimates for toxic effects other than cancer. Clearly, a major research effort is needed. Where it was once common to refer to "no-effect doses" of chemicals and "safe" doses, it is now more appropriate to speak of"no-observed- adverse-e~ect" doses and "acceptable risk" when describing permissible use or exposure to chemicals. This change has been accompanied by an increasing concern for the health of the most susceptible individuals in

Chemical Contaminants: Safety And Risk Assessment 25 the population, besides that of the average individual. Many scientists now distinguish between injuries produced by chemicals for which there is likely to be a threshold dose and elects (e.g., carcinogenesis and mutagenesis) for which there is likely to be either no threshold dose or no way presently known to estimate one for large, heterogeneous popula- tions. [A report of another committee of the National Academy of Sciences expresses some doubt about the validity of the threshold concept for any type of biological erect (NAS, 1975~; see also Drake et al., 1975, and Hoel et al., 1975.] From another point of view, Well (1972), considering statistics and judgment in safety evaluation, wrote: "No matter what the biological erect, at some concentrations under some sets of conditions, a dose level must exist below which no biological damage will occur during the life-span of the great majority of men. No matter how small the dose, however, one, or a few, of millions of subjects may exhibit the critical response." It is more prudent to treat some kinds of toxic elects 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 these are elects that result from an initial, chemically induced alteration in cellular genetics that is transmitted by cell propagation. Carcinogenesis and mutagenesis are examples in which a single cell transformation has the theoretical potential for irreversibility, which might involve self-propagation, even in the absence of further exposure. Other injuries may become self-propagating-e.g., advanced stages of cirrhosis but they are usually preceded by detectable injury that is reversible. The initial erects should have a dose-response threshold, inasmuch as the nature of cellular injury that precedes them can be detected while the injury is still reversible. Some forms of injury may be strictly cumulative, because the cells in which they occur are not repaired or replaced. (For example, destruction of enough neurons leads to a decrease in central nervous system function.) Congenital malformations appear to be irreversible. In this case, injury occurs from exposure during only a brief period. In addition, it is probable that a threshold dose could be estimated from adequate experimental or epidemiological data. Current knowledge of the proper principles for extrapolating toxicolog- ical data from high dose to low dose, and from one species to another, is inadequate. Nonetheless, standards for drinking water must be devel- oped. Whenever possible, a maximal no-observed-adverse-effect dosage should be identified. Three major categories of erects should be

26 DRINKING WATER AND H"LTH considered, and different ways of arriving at standards can be proposed for each. Irreversible (SeIf-Propagating) Effects (These are likely to become life-threatening even after exposure has ceased) 1. Genetically self-propagating ejects, e.g., somatic or germ-cell mutation that culminates in a malignant neoplasm or is transmitted to later generations: Assume no threshold, assume a linear dose-response at low doses, and estimate risk. Set standard at something other than zero only if exposure cannot be eliminated by reasonable means, or if material has no safer substitute, and if it has great utility or social value. An acceptable degree of risk arrived at by a case-by-case consideration involving numerous scientific, technological, economic, and societal issues and values should determine the permitted dose. Nonreversible Effects 1. Ejects that are sequels to probably detectable, reversible injury, but that may become self-propagating (such as cirrhosis): If a threshold can be demonstrated, use it as an upper limit, with application of an appropriate safety factor. If not, proceed as in "Irreversible Ejects." 2. Death of irreplacable cells, cumulative with continued exposure, e.g., central nervous system disease, as in exposure to methyl mercury. If a threshold can be demonstrated, use it as an upper limit, with application of an appropriate safety factor. If not, proceed as in "Irreversible Effects." Reversible Effects 1. Life-threatening or major morbidity, e.g., inhibition of a vital enzyme system. If a threshold can be demonstrated, use it as an upper limit. If not, proceed as in "Irreversible Ejects." 2. Minor morbidity, e.g., sensory irritation without histological change. Determine the range of sensitivities and choose an upper limit low enough to minimize occurrence in the population. 3. No detectable functional or sensory decrement, but possibly predictive precursor of more serious effect, e.g., plasma cholinesterase inhibition, or small increase in liver enzymes in plasma. Proceed as in 2.

Chemical Contaminants: Safety And Risk Assessment 27 IRREVERSIBLE TOXICITY Many factors make the assessment of long-term health risks to human populations difficult for example: 1. The sensitivity of the test systems used to detect carcinogenic ejects depends on the number of animals used in each test and on the duration of their survival. 2. Any series of experiments will yield false-positive and false-negative results. 3. Detection of neoplastic changes in treated animals requires exten- sive gross and microscopic examination of many tissues by trained people. 4. Time, resources, and money required to conduct an adequate test are all substantial. Controversies have arisen because of the above problems and because of inadequate testing for long-term effects. False-positive results can cause unnecessary public concern and the removal of useful materials from the market, and false-negative results can endanger the health of large groups of people. Long-term ejects are particularly difficult to detect and treat because they are discovered only after many years, by which time they are often irreversible. The main question to be answered is: "Within the limitations of present-day capabilities, what are the minimal requirements for an adequate test (on experimental animals) of the long-term effects of potentially toxic agents, and how can these results be used to estimate possible risk to the human population?" Summary of Principles for Extrapolating Animal Toxicity To Humans Despite wide gaps in our knowledge of the metabolism and ultimate fate of chemicals in man, properly conducted experiments will yield results that can improve our estimates of the risk to human populations from long-term exposures. Many mechanisms for chemical carcinogenesis have been postulated. If the mechanism involves somatic mutation or alteration, there is no threshold dose for long-term exposure; if the mechanism is unknown, it is prudent to assume that DNA damage is involved. The idea that there is a "safe" dose of such chemicals may be conceptually valid, but "safety" cannot be established by any experimental method now available. Every dose

28 DRINKING WATER AND H"LTH should be regarded- as carrying some risk. A "most probable risk" can be estimated by appropriate statistical treatment of the results of experi- ments on animals, and once the benefits of use of a chemical have been defined and estimated, it is possible to weigh the health risks against the health benefits. The balance between them should then be the overriding consideration in regulating the amounts of such substances in the environment. The method used in classical toxicology for determining safe doses for short-term exposure of humans to drugs is to estimate a maximum exposure that is tolerated without adverse ejects in a group of animals, and to apply a safety factor. This procedure is valid only for estimating the risk of reversible toxic effects. "No-observed-adverse-effect dose" is a better term, because it makes clear that the exposure can often be a function of the size of the experiment the larger the experiment, the lower this dose can be. Studies in laboratory animals must be used to predict the safety of environmental chemicals. Human epidemiological studies cannot be used to predict nor assure safety, for several reasons: 1. Epidemiology cannot tell what effects a material will have until after humans have been exposed. One must not conduct what might be hazardous experiments on man. 2. If exposure has been ubiquitous, it may be impossible to assess the ejects of a material, because there is no unexposed control group. Statistics of morbidity obtained before use of a new material can sometimes be useful, but when latent periods are variable and times of introduction and removal of materials overlap, historical data on chronic ejects are usually unsatisfactory. 3. It is usually difficult to determine doses in human exposures. 4. Usually, it is hard to identify small changes in common ejects, which may nonetheless be important if the population is large. 5. Interactions in a "nature-designed" experiment usually cannot be controlled. With the possible exception of arsenic and benzene, the known human carcinogens are carcinogenic in some laboratory species. Therefore, animal studies of carcinogenesis in laboratory animals are useful for predicting effects in man. Thus, for ethical and practical reasons, data derived by using animals for toxicity testing are essential for protecting the public from harmful effects of new chemicals in the environment and probably also necessary for evaluating the potential harm of "old" chemicals. By the same token,

Chemical Contaminants: Safety And Risk Assessment 29 epidemiological surveillance studies are necessary for detecting the errors that will surely arise from use of the animal studies alone. Thus, epidemiological studies are both a last line of defense and a means for verifying and adjusting the conclusions from animal studies. The General Problem Of Extrapolation The knowledge and insight that provide a basis for more successful extrapolation are rapidly increasing. The value of tests on laboratory animals is most easily estimated when the chemical agents tested are ultimately administered to, or confront man in a manner similar to the animal exposure, as in the drug-development process. The sequence of animal tests of a new chemical agent, after toxicological studies, continues with studies conducted in order to determine: the mechanism through which the laboratory animals respond to the agent, the nature of its metabolism in tissues and organs, and the rates and routes of elimination of the agent and its derivatives. Thus, damage observed in an organ of an animal provides clues that lead to an understanding of the metabolism and organ involvement of the substance in humans. Similarities and differences between humans and animals can be noted, and the validity of the laboratory-animal test systems can be better estimated. This approach is most useful for observing early elects that occur soon after the substance is administered. The use of such data for assessing long-term effects in humans has many difficulties. When a mouse or man is exposed to a chemical, a number of events can occur that can greatly influence the final reaction, which may appear as the observed toxic elect. These events include: absorption; distribu- tion and storage; metabolism, excretion, and reabsorption; arrival at the site of action; reaction with the biological receptor; and interaction with other constituents of the environment. They can be compared among various animal species and among strains and individuals. Anatomical, biochemical, physiological, pharmacological, and pathological differ- ences and similarities can and have been identified, and there have been efforts to characterize systematically the differences and similarities between species for some compounds and classes of compounds. These are appropriate subjects for research. Chemicals to which man is exposed can be divided into two classes: those deliberately administered for therapeutic, diagnostic, or nutritive purposes, which contribute to health, and those with uses that do not directly benefit health, but reach man through a variety of routes. To some extent, the acute toxicity of the first class can be observed directly in man. If the chemicals are already in use, the laboratory-animal

30 DRINKING WATER AND H"LTH toxicity tests provide a type of pre-screening or early warning system. However, chemicals of the second class are generally not tested in man, and their potential hazard can be estimated only through laboratory animal testing, epidemiological studies, or observation. Extrapolation of data from laboratory animals to man is difficult to treat in a systematic and comprehensive fashion. Although vast amounts of data have been accumulated on the toxicity of compounds in animals and the ejects of drugs and other chemicals on human health, experiments that generate the data tend to be so diverse that comparison is usually impossible. Good quantitative data on the toxicity of compounds in man are rare. Most commonly, there is neither knowledge of the amount of a chemical to which man was exposed nor of the incidence and severity of toxic effects. Toxicologists rely heavily on "experience" for successful extrapolation of toxicity data to man. Unfortunately, experience is interpreted differently by different people and thus is a variable guide. For some of the most serious toxic ejects, it can also be very misleading. For example, some cancers may take decades to develop in man after exposure to a carcinogen, so it is difficult for investigators to develop experience that will permit them to link exposure to ultimate eject. Tests for mutagenic effects in the human population are inadequate today, and it is possible that such ejects from chemicals are already appearing. However, who has had enough "experience" to infer that man has changed because of some specific chemical in, say, the last 50 yr? Statistical observation of small changes is usually difficult. For example, if 20% of all pregnant women used a chemical that caused stillbirths in 5% of the women who used it, the resulting increase in stillbirths would be 1% (0.20 x 0.05 = 0.01), and it is likely that it would not even be noticed. If 5% of all pregnant women used a chemical that caused a delayed eject in 20% of their offspring, this also would probably escape notice. Even if the chemical caused a 10% increase in a most common form of cancer in the offspring of the 5% of women (e.g., cancer of the colon, which would mean an increase of 200 deaths per year), the eject would very likely go undetected. Specific Problems In Extrapolation Experiments on laboratory animals are generally performed under highly standardized conditions, with controlled diet, temperature, humidity, and light-dark cycles, and usually with genetically homogeneous animals, such as inbred mice, rats, or beagles. Thus, one obtains reasonably precise and reproducible information on the toxicity of a substance under

Chemical Contaminants: Safety And Risk Assessment 31 specified conditions, in animals from specific genetic pools. However, large and diverse segments of the human population may be exposed to the compound. Humans are not genetically homogeneous: They live under various environmental conditions, they eat a great variety of foods, and they are exposed to large and increasing numbers of substances. We would like to protect the whole population its most sensitive members as well as the average. Therefore, environmental and genetic variability must be considered in the process of extrapolation. Man is generally exposed to toxic pollutants in the water supply through his gastrointestinal tract, and laboratory animals should be exposed in the same way. However, some problems inherent in the use of animals must be kept in mind when one is using animal studies to predict the carcinogenic potential of a substance for man. SIZE OF ANIMALS Size tends to determine the rate of distribution of substances in the body; in large animals, they are distributed more slowly and tend to persist longer. In general, large animals metabolize compounds more slowly than do small animals. Large mammals have many more susceptible cells; thus, if the carcinogenic event is rare, it would be more likely to occur in a large mammal than in a small one (a mouse experiment using 1,60~2,000 animals represents about the same number of susceptible cells as are contained in one man). Obviously, however, a mouse is not the equivalent of a very small man; nor is a man the equivalent of a very large mouse. In the mouse, the ratio of cardiac output per minute to blood volume is 1:1. In man, the ratio is 1:20. Thus, blood circulates approximately 20 times as rapidly in a mouse as in a man. (This difference is not a peculiar species difference, but typifies the differences between very small and relatively large animals.) The implication is that the time a substance spends in the plasma (excluding metabolism) is longer in a larger mammal than in a smaller one. Thus, for the same milligram-per-kilogram dose, human tissues are exposed to a substance for a much longer time than mouse tissues. This is consistent with data obtained in studies of anticancer drugs, which showed that on a milligram-per-kilogram basis-a mouse required 12 times as much drug to respond as did man, a rat 6 times as much, and a dog and a monkey 2- 3 times as much. When the data were expressed on a milligram-per- square-meter basis, the differences between species were sharply reduced. In addition, the cell-division rate is greater in small animals: e.g., the cycle time of mouse gut or marrow cells is about half that of man. The life span of a man is 35 times that of a mouse. Thus, there are many more cell

32 DRINKING WATER AND H"LTH divisions in man, and therefore a greater opportunity for neoplastic change. T hese observations suggest that small mammals that are routinely used for toxicity testing are often more resistant than man to toxic compounds. This implies that small animal systems are likely to produce many false- negative results, and has important implications for establishing safety factors or using "conservative" techniques for extrapolation. NUMBER OF ANIMALS The number of exposed animals is important. Typically, about 102-103 experimental animals are tested, whereas the population of humans to be protected may be 108-109. The human population is genetically heterogeneous, whereas test populations of animals, as a rule, are relatively inbred. Genetic traits can affect many aspects of susceptibility to pollutants. Selectivity influences the test-animal population in that only healthy, vigorous animals are started on test, whereas the exposed human population may contain subpopulations that are ill and weak. ENVIRONMENTAL DIFFERENCES Nutritional differences and the physical environment (heat, light, ionizing radiation, etc.) can affect response to pollutants. The chemical environ- ment from drugs to air pollutants-with its overwhelming number and types of substances, provides the possibility for synergistic toxicity. Synergistic effects have occurred with therapeutic drugs, are well known in experimental carcinogenesis, and are an ever-present danger. ABSORPTION Rates of absorption of chemicals through the gastrointestinal tract are generally comparable in laboratory animals and man. The barriers within the gastrointestinal tract and the cell membranes, which prevent the free passage of organic compounds, are quite similar among mammalian species. These barriers are either the cells themselves or the intercellular junctions, which are "closed" or "tight" and which, in general, force substances to move through cells, rather than between them. It should be recognized, however, that there are differences between one animal and another and between man and animal. The ratio of surface area to volume in the gastrointestinal tract, and the transit time through it, often differ among and within species. These factors affect absorption. The pH

Chemical Contaminants: Safety And Risk Assessment 33 diners in various parts of the gastrointestinal tract in different species; this can affect the absorption of some partially ionized compounds. The possibility that intact proteins can be absorbed in neonates of some species suggests that absorption in the neonate can be nonspecific. The presence of bacteria in the gastrointestinal tract can affect absorption indirectly. Bacteria may convert the original substance into one that is more or less absorbable, and thus alter the apparent toxicity of a chemical, or they may convert a nontoxic substance into a toxic one. For example, reduction of nitrate to nitrite permits the formation of highly carcinogenic nitrosamines by reaction with secondary amines from the diet. The bacterial populations in the gastrointestinal tract vary between species and within species, and in the same individual from time to time. often chancing with changes in diet. , ~ c} DISTRIBUTION AND STORAGE Once a compound is absorbed, it is distributed throughout the body, largely by the circulatory system. This gives the opportunity for binding to plasma protein or tissue, and storage in such depots as fat and bone. There can also be d.ifferences between species with respect to plasma protein and tissue binding. These differences do not seem to be of major importance, but small mammals, such as mice, tend to bind substances less extensively than man. Research on such subjects as storage depots is needed. METABOrIC DIFFERENCES The primary organs that metabolize substances are the liver, the lungs, and the kidneys. With respect to hepatic metabolism often considered to be the most important small mammals generally metabolize substances more rapidly than large mammals, and herbivores more rapidly than carnivores. There are other interspecific differences; this is a field in which research is active. EXCRETION AND REABSORPTION There are few comparable data contrasting renal or hepatic excretory rates of substances in mice, rats, dogs, monkeys, and man. What data there are suggest that small mammals excrete compounds via the kidneys considerably more rapidly than large mammals. There are too few data to show systematic differences in hepatic excretion and the extent of reabsorption after hepatic excretion. Some substances are excreted

34 DRINKING WATER AND H"LTH through the hepatic system more rapidly by rodents than by dogs and other large animals. After excretion via the bile, reabsorption from the gastrointestinal tract may occur. DIFFERENCES IN RECEPTOR SITES The action of cellular and intracellular membranes is also of concern. An unaltered substance or its metabolites, after distribution through the body, must pass through a variety of cellular and intracellular mem- branes to arrive at the site of action. Many of these membranes are of major importance in determining the ultimate toxicity of the substance. The blood-brain and blood-testicular barriers, for instance, electively modulate the chemical composition of the fluid surrounding the central nervous system cells and the developing sperm cells. These barriers can also avert or delay the arrival of potentially toxic compounds at the sites of action. Similarly, the placenta acts as a barrier between the maternal circulation and the developing fetus. In addition to the passive barriers, active-transport systems influence the movement of foreign organic compounds. This is a subject in which research is moving rapidly, and there is a great need for unified systematic theories on interspecific relationships. In general, if a substance reaches a site of- action, the ultimate toxic reaction seems to be consistent among species. Develop- ment of a lesion will then follow this reaction. With some of the more serious manifestations, such as malignant change and mutagenic change, development of a lesion may depend on the extent of damage (e.g., severely damaged cells will die; generation time of the cells and rates of repair processes may be affected). For instance, the generation time of the rapidly proliferating cell population in the mouse is about half of that in man; therefore, cancer is likely to appear earlier in the mouse. Design Of Laboratory Experiments On Animals In designing assay procedures, the route and type of human exposure should be duplicated as closely as possible. Because it is often impossible, a priori, to determine specific susceptibility-and because man may often be the most susceptible species more than one species of laboratory animal should be used. Short-lived species, such as rats and mice, should be chosen so that tests can be completed in a short time. Doses should be selected for long-term studies, so that exposure to the test substance causes little or no weight loss or shortening of life of animals under test. Sufficiently great exposures in animals need to be used to compensate for the possibly higher sensitivity of the human population and the

Chemical Contaminants: Safety And Risk Assessment 35 uncertainties involved in extrapolation. However, the dose should not be so high as to produce acute toxic ejects. Some logical and statistical considerations in the design and conduct of the animal experiments affect the extrapolation. Two major considera- tions are the structure of the experiments how many doses, how spaced, inbred vs. outbred animals and the size of the experiment how many animals at each dose. If several doses are used, it may be possible to derive a dose-response relationship. The faith that can be placed in an experimental result is also related to the size of the experiment. An experiment using 1,000 animals that shows no untoward effects at a high dose elicits much more confidence that there truly is no effect than does an experiment using 10 animals even if it is conducted at a similarly high dose. A low response rate, 1-2%, is almost impossible to establish with an experiment that uses fewer than 500 animals. If the animals were spread out among several doses (some of which might give a 0.1% or 0.01% response), the number of animals needed to yield reliable figures would be in the hundreds of thousands. Man is genetically heterogeneous and lives in a heterogeneous environment, so some experimenters feel more confident in their extrapolations if there are non-inbred animals in their experiments. Using such animals generally requires larger experiments to give reproducible results. One approach to this problem is to conduct many experiments, in several species, using inbred animals of each species. Current Capability to Extrapolate More and more patterns that are useful for extrapolation to man are being recognized and can be identified in the course of studying pharmacological disposition of a substance. Most of the differences that have been observed suggest that man is more sensitive than the usual experimental animal, and this must be kept in mind in establishing permissible exposures for humans. There are great difficulties in comparing the median animal to the not- so-average man. Man is not genetically homogeneous and is usually exposed to a much wider range of environmental conditions than the usual experimental animal. Differences in environmental factors are known to affect the toxicity of a substance. Differences in the genetic makeup of the individual can affect toxicity. These must be considered in the extrapolation of laboratory-animal toxicity data to man. We must predict for, and protect, the highly sensitive individual as well as the average or median person. Because of the multitude of man-made chemicals, the different habits and life-styles of populations, and the

36 DRINKING WATER AND H"LTH different eating habits of populations, there is considerable variation in the intake of, or exposure to, environmental pollutants. These must also be considered in establishing permissible exposures to environmental agents. In the last decade, large genetic differences in metabolism of toxic agents have been found in man, and study of these ejects has led to a new branch of pharmacology pharmacokinetics. The range of variabil- ~y Is ~us~ra~ea oy fine observation teat, In a group of about 20 patients, the steady-state plasma concentrations of two tricyclic antidepressants varied by about 30-fold, and the half-time of disappearance from the plasma varied about 10-fold (Hammer and Sjoqvist, 1967~. Also, the plasma concentration of isonicotinic acid hydrazide has been found to vary by a factor of almost 100 in different People. and this annears to be genetically controlled (Weber, 19761. ~ 1 1 . ~ ~ ~ . ~ ~ . ~ . ~ . - r - -r--~ ~ -rr ~~~ ~ Many examples of unexpected toxic reactions to therapeutic agents have been caused by interactions of compounds that were normally safe when given alone (Conney and Burns, 1972~. The likelihood that such synergistic interactions will occur with environmental pollutants is high. An early and striking example of unexpected synergistic interactions (hypertensive crisis) occurred in patients who, while taking monoamine oxidase inhibitors, ate cheese that contained bioactive amines. Although theoretically predictable, this effect had not been suggested in the literature before it was observed in patients. Many sources of variation within the population of any species, such as age, sex, pregnancy, disease, can affect the response to foreign substances. An example is the difference between sexes in metabolism and toxicity of acetylsalicylic acid (Menguy et al., 1972~. In addition, such environmental variables as temperature, lighting, barometric pressure, humidity, and diet are known to affect the toxicity of environmental pollutants. For example, the oral absorption of pentobarbital is greatly altered by the prior ingestion of a small cheese sandwich. In this case, peak pentobarbital concentration in plasma is diminished by about 50%, and the duration of the elective plasma concentration is doubled (Bush et al., 1966~. Another difficulty in extrapolating from laboratory animals to man is the dose-duration problem. The lifetime of a mouse is about 2 yr. A man's life expectancy is close to 70 yr, and a woman's 5 yr longer. Because man usually excretes compounds more slowly than mice, this means that man, with a very small daily intake, can develop a greater accumulation of a compound over many years. It is currently impossible to estimate the ejects of all the genetic and environmental variability in the human population. ~,,c, .

Chemical Contaminants: Safety And Risk Assessment 37 Although gains are being made in predicting ejects of pollutants in man, safety testing must, for the present, continue to rely on the results of administration of high doses of substances in test animals to provide a basis for protecting the human population. THRESHOLDS Biological Considerations Whether or not a particular eject follows a dose-response relationship that has a threshold depends entirely on the mechanism of the eject. Many ejects have thresholds. For example, the gastrointestinal-radiation syndrome, acute drug toxicity, and radiation or drug control of some tumors all have dose-response curves that show thresholds. The curves are sigmoid, and below a particular dose there is zero probability of producing the effect, because it requires many independent occurrences and will not occur until the number of such events exceeds some critical value. The gastrointestinal-radiation (or drug) syndrome is a case in point. An animal will not die until the number of intestinal crypt cells that have been killed exceeds some value that is critical to the integrity of the organ. Any radiation or drug dose that kills fewer cells than this critical number can be considered to be "safe" (at least for this one syndrome). We are used to thinking in terms of thresholds and sigmoid dose- response curves. For example, if it costs $4,000 to buy an automobile, we do not imagine that we will have a 50% chance of buying the same vehicle for $2,000. If 100 aspirin tablets constitute a lethal dose, we do not calculate that we will have a 1% chance of dying if we swallow a single tablet. Because we know the mechanisms underlying these events, we expect thresholds to the dose-response curves, and indeed they are evident. However, other ejects may well not have threshold dose-e~ect relationships. If an eject can be caused by a single hit, a single molecule, or a single unit of exposure, then the eject in question cannot have a threshold in the dose-response relationship, no matter how unlikely it is that the single hit or event will produce the effect. Mutations in prokaryotic and euka~yotic cells can be caused by a single cluster of ion pairs produced by a beam of ionizing radiation. We would expect that mutations can be caused by a single molecule or perhaps group of molecules in proximity to the DNA. The necessary conclusion from this result is that the dose-response relationship for radiation and chemical

38 DRINKING WATER AND H"LTH mutagenesis cannot have a threshold and must be linear, at least at low doses. It is one step further to correlate mutagenesis with carcinogenesis. Nevertheless, the evidence is strong that there is a close relationship between the two (McCann et al., 1975; McCann and Ames, 1976; Ames, 1976; DHEW, 1977). We therefore conclude that, if there is evidence that a particular carcinogen acts by directly causing a mutation in the DNA, it is likely that the dose-response curve for carcinogenesis will not show a threshold and will be linear with dose at low doses. CONSIDERATION OF THE DOSE-RESPONSE RELATIONSHIP In considering the possibility of thresholds for carcinogenesis, it is important to understand that there is no agent, chemical or physical, that induces a form of cancer in man that does not occur in the absence of that agent. In other words, when there is exposure to a material, we are not starting at an origin of zero cancers. Nor are we starting at an origin of zero carcinogenic agents in our environment. Thus, it is likely that any carcinogenic agent added to the environment will act by a particular mechanism on a particular cell population that is already being acted on by the same mechanism to induce cancers. This reasoning implies that only if it acted by a mechanism entirely different from that already operating on the tissue could a newly added carcinogen show a threshold in its dose-response curve. EXAMINATION OF EXPERIMENTAL DOSE-RESPONSE CURVES The most extensive information on carcinogenesis both in experimental animals and in humans is with ionizing radiation (BEIR,1972~. Although there is evidence implicating thresholds in some animal tissues (Walburg, 1974), thresholds have in general not been established for most tissues. If such thresholds exist, they occur at sufficiently low doses that it would require massive, expensive, and impracticable experiments to establish them. In view of the common finding-for example of a linear dose- response relationship (unaffected by dose-rate) for cancer induction in animals by high LET radiation, it is unlikely that such thresholds exist. Linearity is not essential to the no-threshold argument since nonlinear, dose-response relationships do not necessarily imply the existence of thresholds. Recent reviews by Barendsen (1975) and Brown (1976) suggest that the dose-response curve for mutation induction, the production of chromo

Chemical Contaminants: Safety And Risk Assessment 39 some aberrations, and the induction of tumors in mammals is linear with low-LET radiation up to about 50-100 reds. Brown (1976) concluded from the available human data that this also applies to man. Such findings argue against a threshold for ionizing radiation. Because many carcinogenic agents act like radiation in producing mutations, chromo- some aberrations, and cell killing, we see this as an additional argument against the likelihood of thresholds in the dose-response curves of these agents. HETEROGENEITY OF THE POPULATION The human population in the United States- the population we are trying to protect is a large, diverse, and genetically heterogeneous group exposed to a variety of toxic agents. Genetic variability to carcinogenesis is well documented (Strong, 1976), and it is also known that individuals who are deficient in immunological competence (for genetic or environ- mental reasons) are particularly susceptible to some forms of cancer (Cottier et al., 1974~. It seems, therefore, that even if we were to postulate an average threshold for a particular cancer induced by a particular agent, we would in practice need a series of thresholds for different individuals. It would be extremely difficult to establish a single threshold. We conclude from these arguments that, despite all the complexities of chemical carcinogenesis, thresholds in the dose-response relationships do not appear to exist for direct-acting carcinogens. If they do exist, they are unlikely to be detected and, hence, impossible to use. This means that there can be no totally "safe" exposure to a particular carcinogen, nor can the term "margin of safety" have any meaning. Any dose of a carcinogen must be considered to be associated with a risk, and even if that risk is vanishingly small, it must be estimated. Statistical Considerations Many quantitative theories of carcinogenesis that have been proposed relate the frequency of detectable tumors to the intensity of exposure to the carcinogen. The purposes of these theories are twofold: to elucidate the mode of action of the carcinogen and the nature of the neoplastic change, and to estimate from animal experimentation the risk to human populations exposed to environmental concentrations of the carcinogen. One of the earliest quantitative theories was that of Iversen and Arley (Iversen and Arley, 1950; Arley and Iversen, 1952~. Their model was basically a one-step transition process, in which a single "normal" cell is

40 DRINKING WATER AND H"LTH regarded as having some probability of being transformed to a cancer cell. Iversen and Arley assumed that this transition probability was a linear function of the amount of the carcinogen, the intercept of this linear function representing the background or spontaneous transition probability, as would be obtained if none of the carcinogen were present. After transition to a cancer cell, Iversen and Arley assumed, growth of the clone could be represented by a pure birth process with a birth rate independent of the initial amount of the carcinogen. It was assumed that the clone would become a detectable tumor when it reached a given size. This model is commonly referred to as the one-hit model and implies that the expected number of tumors within a lifetime will depend only on the total dose and not on the pattern of exposure. The mathematical forms of several hit-theory models have been reviewed by Turner (1975), including generalizations that take account of variations in susceptibility and the number of critical targets. Both Nordling (1953) and Stocks (1953) carried the Iversen-Arley model a step further when they proposed theories in which a single cell can generate a tumor only after it has undergone more than one change or mutation; these could be termed multievent theories of carcinogenesis. They assumed that, within some time-frame, the probabilities of transition from one state to the next were the same and proportional to both time and the concentration of the carcinogen. Like Iversen and Arley, Nordling and Stocks assumed that the growth time of the tumor (after the last event had occurred) was either independent of the carcinogen or negligible. When the number of necessary changes is about 6 or 7, this model yields age-specific, cancer-incidence rates that are proportional to the fifth or sixth power of age and in this respect is consistent with human incidence data. However, the model also predicts that cancer incidence is proportional to the same high powers of the concentration of the carcinogen; this is not in agreement with the results of human and animal data. To avoid this discrepancy, Armitage and Doll (1954, 1961) modified the theory of Nordling and Stocks by assuming that the probabilities of transition between events were not all equal. They also assumed that only some of the transitional events depended on the carcinogen and that the remainder had a probability of spontaneous transition independent of the concentration of the carcinogen in question. With this modification, the model became consistent with both human and animal data that showed tumor incidence as related to either dose or the square of dose, but not higher powers. It should be noted that the theories of Nordling, Stocks, and Armitage and Doll are based on the concept that carcinogenesis has a single-cell origin, whereas a theory proposed by Fisher and Holloman (1951) is multicellular in concept.

Chemical Contaminants: Safety And Risk Assessment 41 They proposed that a tissue of N cells must contain at least K cells that have undergone some transformation if a tumor is to occur in the tissue. This theoretical approach leads to a model very similar in form to the multievent model. Other multievent theories have been proposed that incorporate the concepts of cell death or loss of ability to divide (Burch, 1960), and modifications that permit cells in intermediate stages to grow more rapidly than normal (Armitage and Doll? 1957~. Crump et al. (1976) discussed many of these models of carcinogenesis from the viewpoint of low-dose kinetics. They made two basic assump- tions: that the cancer process is single-cell in origin, possibly with multiple steps between initiation and complete alteration, and that the growth period of the completely altered cell is basically independent of the degree of exposure. For direct carcinogenic processes, in which the agent or its metabolite acts at the cellular level to produce an irreversible change, they concluded that most models of carcinogenesis will be linear for low doses. In addition, they showed that, if it can be assumed that the environment contains carcinogens that act in conjunction with the test agent, then all the models thus far proposed will be linear for low doses. In all these theories, the emphasis is mainly on the stochastic nature of the changes involved in the carcinogenesis process. The role played by . . . . . - the carc~nogen ~s cons~dered to a much lesser degree. It is commonly assumed that transitional events in the process that are attributable to the carcinogen occur with probabilities proportional to the degree of exposure. This is undoubtedly a gross oversimplification of the actual process. The actual exposure is no doubt modified by absorption, distribution, metabolism, and excretion of the chemical substance, and the effective exposure should probably be the actual concentration of the carcinogen at and within the target cells. Other factors that may affect delivery of the carcinogen to intracellular compartments are membrane permeability and enzyme binding Therefore, the "elective dose" may well be some complex function of the actual exposure and the biochemi- cal and physiological characteristics of the host. Most of these mathemat- ical models incorporate the dose as it is actually administered in animal experiments or human exposure. The function relating administered dose to "effective dose," if it is not a simple case of proportionality, can have a profound effect on the dose-response relationship. As a simple example, consider a linear model relating dose (effective dose) to response. If the effective dose is proportional to the administered dose, then a linear model obtains for administered dose versus response. If the dose relationship is concave, which would obtain if the incremental increase in elective dose decreases with higher doses, then the relationship between administered dose and response would also be concave. Thus, the various

42 DRINKING WATER AND H"LTH dose-response curves that have been observed may not be indicative of different carcinogenic processes once the agent has reached the target cell, but rather may indicate different functions relating administered dose to elective dose. This problem will probably relate more to chemical carcinogenesis, as opposed to radiation-induced cancer. Even if a normal cell has been transformed to some intermediate stage in the carcinogenic process, this would not necessarily mean that cancer . ~ .. .. ~ ~ . must occur. cell repair or recovery or some other response trom the immune mechanisms may be sufficient to stop or reverse the process before the final stage is reached. In addition, the death of these transformed cells may occur before the process has a chance to continue toward the eventual cancer. This is one of the major arguments in favor of the existence of a threshold. However, if there is some probability that these recovery mechanisms will not complete their role before the occurrence of another event or transformation, this type of threshold will not exist. Thresholds may be considered from two viewpoints an "actual" threshold, which is an exposure below which any carcinogenic response attributable to the specific agent is impossible, and a "practical" threshold, which is an exposure below which an attributable carcinogenic response is highly unlikely. In discussing carcinogenic thresholds, Mantel (1963) has argued that it is immaterial whether or not "actual" thresholds exist, and that one should consider the "practical" thresholds when estimating human risk. Mantel and Bryan (1961) stated that a risk of cancer of 1 in I08 could be thought of as a "practical" threshold and that efforts should be made to estimate exposures that would produce no more than this risk. Using mathematical models that relate the latency period (time between initiation of exposure and appearance of cancer) to the exposure, Jones (1976) suggested that a "practical" threshold could be defined as an exposure for which the latency period is beyond the normal life span. Experimental or observational evidence of the existence of an "actual" threshold is usually presented in the form of a dose-response graph, in which the percentage of animals with tumors or the average number of tumors per animal is plotted against the dose of the carcinogen. Either the existence of doses that do not lead to an increase in tumor incidence over controls, or the extrapolation of such curves to low doses that apparently would result in no tumor increase, is cited as an indication of the existence of a threshold below which no response is possible. This type of reasoning is an exercise in self-deception. In the first case, failure to observe positive responses does not guarantee that the probability of response is actually zero. From a

Chemical Contaminants: Safety And Risk Assessment 43 statistical viewpoint, zero responders out of a population of size N is consistent at the 5% significance level with an actual response probability between zero and approximately 3/N (e.g., when N = 100 and zero responders are observed, the true probability of response may be as high as 3~01. In the second case, when an observed plot of dose against tumors is extrapolated downward to find a no-effect dose, it is assumed that the observed dose-response relationship, usually linear, will obtain through- out the entire spectrum of doses and that one threshold exists for the entire population at risk. The assumption of linearity throughout the entire dose spectrum can easily lead to erroneous conclusions. For example, consider the true dose-response relationship shown by the dashed curve in Figure II- 1. This curve is convex, which would be consistent with a multievent theory of carcinogenesis in which more than one event is affected by the carcinogenic agent. This type of convex dose-response relationship, when sampled in an animal experiment over only a part of the dose range, could be thought to imply the existence of a threshold if simple linear extrapolation is used. If the animal experiment is performed at doses between A and B. one could conclude that a threshold exists at dose do; if the experiment is performed at doses between B and C, the conclusion Could be that a higher threshold exists at dose d2. In fact, with convex dose-response curves, simple linear extrapolation will always imply the existence of a threshold, if the experiment is performed over any range of doses that appears to produce linearity between dose and response. In addition, the assumption of one threshold is unrealistic. It is much more likely that, if thresholds do exist, not all members of the population have the same one. The human population is a very diverse, genetically heterogeneous group that is exposed in different degrees to a large variety of toxic agents. Many different disorders may affect the frequency of mutational events in specific tissues. Disorders characterized by chromo- somal instability have been found to be predisposed to malignancy. Patients with xeroderma pigmentosum are highly susceptible to ultravio- let-induced skin cancer, and it has been found that they have deficient DNA repair mechanisms (Robbing et al., 1974~. In Bloom's syndrome, there is an immune deficiency and increased risk of leukemia and cancer of the colon (German, 1972~. These systems may provide a model in which the risk of mutation and subsequent malignancy after exposure to an environmental carcinogen may be genetically determined. If malig- nancy is the result of a series of mutational events, there must be subpopulations at various degrees of risks or with various thresholds for the carcinogenic agent. Therefore, the search for thresholds should not be

44 DRINKING WATER AND HEALTH - m m o cr is o An LU o _: 1 / 1 1 - // A do B C DOSE LEVEL FIGURE II-I Linear extrapolation from a convex dose-response curve. a search for one specific no-effect dose applicable to all members of the population at risk; rather, the problem is to find many thresholds for each of the subgroups within the population. This variability in thresholds or susceptibility to carcinogenic agents has been shown by Mantel et al. (1961) to induce an increased convexity in dose-response curves at low doses. They demonstrated that a linear  dose-response curve with a fixed threshold will become convex at low exposures, if the individual thresholds are allowed to vary. Therefore, the extrapolation of observed dose-response curves, when the individual thresholds actually vary in the population, will, at best, simply lead to the average threshold of the population at risk. This estimate of the average threshold will have little practical utility, because many members of the population will have individual thresholds below this value. In addition,

Chemical Contaminants: Safety And Risk Assessment 45 if we are willing to assume that threshold variability is the likely state in nature, then from a statistical viewpoint it is practically impossible to distinguish between mathematical models that hypothesize different thresholds and multievent models that hypothesize no thresholds, because the shapes of the two models can be very similar, and in some cases identical. For example, a one-hit, dose-response model with a threshold may be written as, P(d ~ A) = In d < -e-<X(d - A) d ~ ,< where P(d l A) represents the dose-induced response rate at a dose level d and A is the threshold below which no response can occur. If we assume that the population consists of individuals with different thresholds, and that these thresholds vary according to some probability distribution F(~), then the variable-threshold model is simply the convolution of P(d l A) with F(A), P(d ~ =r~ P(d ~ \)dF()~) If, for computational simplicity, we choose F(~) to be an exponential probability distribution, then this variable-threshold model takes the mathematical form, P`d'=1_ e _~/,Be-~d 1 - ~x/,B The interesting aspect of this particular mathematical model is that, as the ratio a/,B approaches unity, the model becomes P(d) = 1 - (1-ord ye which is the mathematical form of a two-hit model for dose response. Discrimination among these models on the basis of animal experiments is often impossible. These three models were fitted to data from an experiment by Terracini et al. (1967), in which dietary concentrations of dimethylnitrosamine (DMN) of () 20 ppm were fed to female rats. The experiment was continued for 120 weeks, and the appearance of liver tumors was the response variable. The data are shown in the following table.

46 DRINKING WATER AND H"LTH DMN in Diet (ppm) 0 2 Number responders/at risk 0/29 0/18 4/62 2/5 Response rate O.O~O O.O~O 6.5% 40~O 10 20 15/23 65% The fixed-threshold and-variable-threshold models, in addition to the two-hit model, all fit these data equally well. There is no statistical basis on which to prefer one over the others. Experimental results like these, although appearing to give evidence of a threshold, will provide no statistical evidence either in favor of, or opposed to, the existence of such thresholds. Therefore, statistical analysis of standard animal carcinoge- nicity experiments is not and probably will never be in a position to resolve the threshold question. There are too many "biologically reasonable" mathematical models, both implying and denying the existence of thresholds, that will fit the observed results. The quantitative theories of carcinogenesis that have been proposed are all stochastic. They consider the probabilistic aspects of the occurrence of some series of events. If one is willing to assume that these events have transition probabilities that can be affected by the carcinogen (no matter how small the concentration); that the systems of distribution and metabolism have some chance of allowing some amount of the carcinogenic agent to reach the target cells (no matter how small the chance or the amount); and that the repair and recoverer systems may not do a perfect job (no matter how small the chance that this will happen), then there will be no exposure that will have a zero probability of leading to a cancer. In addition, when considering the possibility of carcinogenic thres- holds, one should keep in mind that no agent has been found to induce a type of cancer that has not been previously described. It is possible, perhaps even likely, that many carcinogenic agents act by the same mechanism on the same target cells. This would imply that, inasmuch as there are many carcinogenic agents in our everyday environment, some additional carcinogen could act in a simple additive manner and thus that any exposure would simply be added to the background. This means that for a population already exposed to carcinogenic agents, any additional carcinogen would simply increase the expected tumor inci- dence in a continuous manner, no matter how low the exposure to the additional agent. Therefore, despite all the complexities of the mech- anisms of chemical carcinogenesis, because of genetic variation among members of the population at risk and because statistical analysis cannot resolve the question one way or the other, the search for an "actual" carcinogenic threshold is probably fruitless, and any human exposure to

Chemical Contaminants: Safety And Risk Assessment 47 a carcinogen should be considered to be associated with some risk, no matter how small that risk may be. The current mathematical models that relate exposure to attributable risk are, at best, extremely crude. Much work needs to be done to refine these theories. HIGH-DOSE TO LOW-DOSE EXTRAPOLATION Dose-Response Models Animal experiments must be performed at doses high enough to produce tumors, and environmental concentrations must be low enough to produce few, if any, tumors. Therefore, to estimate the probability of response at dose levels outside the experimental range, it is necessary to make an assumption concerning the form of the dose-response relation- ship at low doses. As noted above, many quantitative theories of carcinogenesis have been proposed that relate the occurrence of detectable tumors to both the quantity and the duration of exposure to the carcinogenic agent. Most of these theories start with the premise that the carcinogenic process consists of one or more stages that occur at the cellular level, but that not all of these stages are related to the carcinogen. These stages may be cell mutations or other biological or chemical events and may be monocellular or multicellular in origin. The probability of transition to an event related to the carcinogen is assumed to be proportional to the exposure. The exact nature and causes of these events are largely unknown. However, the most important aspect of these quantitative theories of carcinogenesis is that most of them lead to mathematical models for which the probability of tumor occurrence is generally related to a polynomial function of dose. For low exposure, the region of most importance, they are all well approximated by a simple linear function of dose (Iversen and Arley, 1950; Fisher and Holloman, l9S l; Muller, 1951; Nordling, 1953~. This class of dose-response models may be considered as models that are "linear at low dose." Other mathematical dose-response models have been proposed for this problem of extrapolating from high-doses to low doses, the most notable being the log-probit method of Mantel and Bryan (1961~. These types of models have little biological justification in what is known about the carcinogenic process. In addition, some require use of preselected parameters chosen without regard to the particular experimental situation or results. A dose- response model selected for extrapolation purposes should, at the very least, be consistent with current knowledge of the carcinogenic process. The mathematical models that relate dose to response rate fall into two

48 DRINKING WATER AND HEALTH categories: dichotomous-response models, for which the important experimental fact is whether a particular condition, such as a malignant tumor, is present by a particular time (in chronic exposure experiments, this time is usually the animal's normal lifetime); and time-to-response models, for which the actual time to occurrence of the particular condition is known for each animal in the experiment. (For the latter type of model, the relationship of time since initiation of exposure, to response can be ascertained together with the relationship between exposure and response). Each type of model is completely specified, except for some unknown parameters that are to be estimated from the experimental results. These estimates and the functional form of the assumed dose- response model will provide an estimate of the risk attributable to the carcinogen at any desired dose. In the dichotomous-response situation, the multievent theory of carcinogenesis proposed by Armitage and Doll (1961) leads to a mathematical model that relates the probability of response P(a, to the dose dby P(d) = 1 -e (~0 + Ald + \2~2 + . . . + Ak~k) where k represents the number of transitional events in the carcinogenic process that are related to the carcinogen under test and Ao, Al, A2,.~.,Ak are unknown nonnegative parameters. As with the other "linear-at-low- dose" models, for small enough values of the exposure d, this dose- induced response rate will be approximately equal to Ad assuming that A0 is the "background" rate). Therefore, in extrapolating from high dose levels to low doses, the risk attributable to the carcinogen, after correcting for background, will depend on the magnitude of the coefficient, A'. When estimating risk with this model, the function is to be fitted to the experimental data by some such procedure as maximal likelihood (Guess and Crump, 1976~. A point estimate of the attributable risk may be obtained from the model with the estimated parameters, but to incorporate the vagaries of random sampling, it would be prudent to include the upper-statistical confidence limit on this risk estimate. Appropriateness of Data for Low-Dose Extrapolation Before extrapolation is attempted, considerable attention must be given to the appropriateness of the experimental data. Bioassay procedures must be of high quality in order to avoid misleading risk estimates. Also, oral administration of the carcinogen is necessary, because we are concerned with estimating the risks associated with drinking-water

Chemical Contaminants: Safety And Risk Assessment 49 consumption. The experimental animals should receive a lifetime administration of the carcinogen, because this corresponds to the human situation that we are attempting to estimate. It is not correct to assume that the tumor yield in a partial lifetime study will be equivalent to a lifetime study with the same total dose. This follows from the multistage models for carcinogenesis which assume that the cancer-incidence rate is proportional to a power (often 3 to 5) of duration of exposure. It may be possible to extrapolate from partial to full-lifetime exposure, but it must be done on a sound biological basis. Serious problems also develop from the actual data generated by the assay. These problems have to do with the need for some observed dose- response information. Often, the experimental doses are very high, and this produces cancers in almost all the experimentally treated animals. This may be because the assay protocols are designed for the determina- tion of carcinogenicity, and not for low-dose extrapolation and dose- response estimation. If all of the treated animals produce high incidences of tumors, one has no useful idea of where the dose-response curve actually is (particularly the lower convex portion of the curve). The true dose-response curve could be badly underestimated even with the use of a straight-line extrapolation. Therefore, one must have some informa- tion concerning doses at which the tumor incidence is in the lower convex portion of the dose-response curve. Although low-dose extrapolations are out of the question for high- incidence data, one can calculate some relative dose information that can be used in expressing concerns over possible adverse health ejects. Essentially, the exposure that is of interest in man is converted to the equivalent dose in animal study. The ratio of the lowest experimental dose to the equivalent human dose is then calculated. Clearly, the closer this ratio is to unity, the greater the concern. One final point should be made. The extrapolations should generally be conducted on total tumor yields, as opposed to specific tumor types, because we wish to estimate total cancer risk in man. INTERACTIONS A major concern in carcinogenesis, and in assessing the safety of a single material, lies in the answer to the question "How does this material behave in the presence (or absence) of other materials?" Is the response to two materials the sum of the individual responses? Do the ejects come about as if doses were added? Is there more (or less) than additivity? Do some things cancel each other out?

50 DRINKING WATER AND H"LTH Some experience in man begins to give answers to these questions. The few known interactive ejects in man are related to exposures to cigarette smoking combined with some other exposure industrial or social. These are the combined ejects of: cigarette smoking (inhaled); asbestos exposure (Seliko~, 19681; inhalation of radon, as in the case of the uranium miners (Lundin, 1971~; and the use of alcohol (Rothman, 1972~. All of these show risks that are more than additive. For example, the heavy smoker has about two times the risk of the nonsmoker of dying of cancer of the oral cavity. The heavy drinker-heavy smoker has about 15 times the risk of the nonsmoker, nondrinker. Reports from Great Britain support the idea that the recent reduction in male lung cancer mortality, particularly in Greater London, with small changes in cigarette smoking and large changes in air pollution have come about through an interaction between urban air pollution and cigarette smoking (Lawther, 1976). Animal experimentation (Berenblum, 1947) showed an initiator-pro- moter eject, in which each material applied separately (to the skin of test animals) had little or no carcinogenic effect. The two materials, applied in the appropriate sequences, however, were highly carcinogenic. At one time it appeared that there were materials that were purely initiators (they started the process) and there were others that were purely promoters (they moved the process along to frank cancer). More recent views suggest that some materials are both initiators and promoters, sometimes acting as one, sometimes as the other, and sometimes as both, giving a "complete" carcinogen (Berenblum, 1974~. The exact nature of initiation and promotion is not clear, and there is evidence that there is initiation-inhibition as well (Falk, 1971~. Thus, it may be possible that the carcinogenic eject of one material may be inhibited by another material. Genetic susceptibility can be included in the initiator-promoter model. Increased genetic susceptibility may be providing an initiator step, while increased genetic resistance will look like initiation-inhibition. Some instances of this are evident in man. For example, ultraviolet radiation leads to skin cancer, mostly in people with fair skin, light hair and light eyes (Urbach, 1969~. We know that these latter characteristics are genetically determined and so we speak of genetic susceptibility. Dark-skinned, dark-eyed, dark-haired people have genetic resistance. It is not nearly so obvious, nor do we know the genetics involved, in the case of the higher susceptibility to breast cancer among women from "breast cancer families." Women who come from families where a mother, sisters, or the fathers' sisterts) have breast cancer are at higher (twofold to threefold) risk than women from non-breast

Chemical Contaminants: Safety And Risk Assessment 51 cancer families (Lilienfeld, 1963~. There is nothing so obvious nor so directly involved as skin pigmentation to give any leads here. A possibly unifying concept concerning interactions is the "fertile ground" idea. If the carcinogen falls on fertile ground, it will lead to cancer growth. If not, no cancer. This implies a whole set of questions about the nature of the "fertile ground." Some animal experimentation seems to imply that, with the induction of mammary tumors in the rat by hormonal stimulation, the ground is made fertile by earlier viral infection (Shellabarger, 1976~. Similar results seem to be shown in the induction of (mouse) cancers through the combination of malaria infection and exposure to the Moloney mouse-tumor virus (Wedderburn, 1970~. A similar mechanism has been suggested for the development of Burkitt's lymphoma in children in East Africa, in malaria infection associated with appropriate viral infection (Morrow, 1976~. Another way that fertile ground can arise is through a deficiency interaction. Here, the absence of some element, material, or vitamin may reduce the ability of repair mechanisms, so that a level of a carcinogen that produces little or no cancer in a population capable of repair produces much cancer in a population of repair-deficient individuals. An example is the high cancer rate in Xeroderma pigmentosum patients people with a low ability for DNA repair (Robbing, 1974; Cleaver, 1968~. A similar mechanism (on a grosser level) may be involved in the high incidence of cancer of the esophagus in central Asians, whose bread and tea diet may not contain the elements necessary to permit repair of tissue affected by low levels of a non~dentified carcinogen (Mahboubi, 1973~. Much of the classical toxicology of dose-response assumes a distribu- tion of sensitivities in a population. Individuals with thresholds lower than the dose given will respond by developing the condition. Individuals with higher thresholds will not. The existence of dose-response curves that are not simple step-functions (zero response at dose d, 100970 response at dose d + e, where e is some very small increment of dose) suggests these host-environment interactions. Individuals with high gluathione and sulfLydryl levels in their livers may be able to "handle" (to metabolize to some nonactive form) toxicants much better than individu- als with low levels (Gillette, 1974~. According to Brown (1975), no carcinogen has been found to induce a cancer of an entirely new, never-before-seen histological type. A possible explanation of this is that many carcinogencic agents or processes act through the same ultimate mechanism in the same target cells. He writes " . . . since there are many carcinogenic agents in our . . . en- vironment, [a new] carcinogen could act in a simple, additive manner and thus any level of exposure would . . . be added to the . . . existing

52 DRINKING WATER AND H"LTH background." This kind of interaction phenomenon has important implications for the models used in extrapolation. It implies that these models may not assume any threshold for a new material in a world already populated by many carcinogens. It also leads to the conclusion by Crump et al. (1976) that "all models of carcinogenesis thus far proposed will be linear at low doses." The discovery of interactions will require more sophisticated experi- mental techniques than are now being used. Testing combinations of materials multiplies the number of tests that must be done (100 materials, tested two at a time, will require 100X 99/2! tests; tested three at a time will require 100 x 99 x 98/3! tests). New techniques for multiple testing will have to be developed. Uncovering deficiency interactions will require entirely new and different approaches. SUMMARY Chemical Contaminants: Safety and Risk Assessment Large populations are repeatedly exposed to potentially toxic contami- nants in the drinking water in minute amounts over many months or years, or over whole lifetimes. Delayed, essentially irreversible, ejects can occur. Methods and criteria of classical, conventional toxicology do not offer reliable means for assessing long-term toxic ejects such as carcinogenesis in man by extrapolation from animal data; hence, novel considerations have to be applied in assessing risk. The insidious ejects of chronic exposure to low doses of toxic agents is difficult to recognize, because there are few, if any, early warning signs and, when signs are ultimately observed, they often imply irreversible ejects. For example, cancer induction in experimental animals, even with the most potent carcinogenic chemicals, requires at least several months and in many instances a whole lifetime. There are as yet no easy, straightforward methods for extrapolating even chronic-exposure experi- ~nental data to calculate risks to large human populations. Teratogenic ejects are easier to establish by animal experimentation, but there are similar uncertainties in extrapolating to human populations. Mutagenic ejects are difficult to assess experimentally in mammals, and such ejects are particularly insidious, in that they appear only in later generations. Various measures used in assessing acute toxicity such as LD~o, LD50, and maximal tolerated dose are generally found to be quantita- tively similar among most animals. On the basis of dose-per-unit of body surface, toxic ejects in man are in the same range as those in experimental animals, such as mouse, rat, hamster, dog, and monkey. On a body-weight basis, man is generally more vulnerable than the

Chemical Contaminants: Safety And Risk Assessment 53 experimental animal, probably by a factor of ~12. Comparative studies have shown generally that absorption, metabolism, and excretion of various drugs are slower, dose-for-dose, in man; that there is greater retention of such drugs; and that higher concentrations occur in body fluids and tissues in man than in small mammals. With an awareness of these quantitative differences, appropriate safety factors can be applied to calculate relatively safe therapeutic dosages for man. These methods and principles of classical toxicology are useful for assessing toxic effects that are reversible and nonprogressive. They are much less useful in dealing with the problems of chronic irreversible toxicity or the effects of long- term exposure. This subject has not been considered widely in the past. From the review of available information, two major questions emerge: "What types of experimental-assay procedures are required for a valid assessment of chronic toxicity of chemicals in experimental animals?" "How can such data be extrapolated to estimate risks in humans?" In dealing with these questions, our recommendations are restricted to a specific risk namely, cancer with the understanding that the same considerations will apply at least partially to the problems of mutagenesis and teratogenesis. Furthermore, we consider only carcinogens whose mechanisms involve somatic mutations. Some principles that underlie efforts to assess the irreversible effects of long-continued exposure to carcinogenic substances at low dose rates are outlined below. Principle 1 EFFECTS IN WAS, PROPERLY QUALIFIED, ~ ITALIC - LE TO ~ This premise underlies all of experimental biology and medicine, but because it is continually questioned with regard to human cancer, it is desirable to point out that cancer in men and animals is strikingly similar. Virtually every form of human cancer has an experimental counterpart, and every form of multicellular organism is subject to cancer, including insects, fish, and plants. Although there are differences in susceptibility between different animal species, between different strains of the same species, and between individuals of the same strain, carcinogenic chemicals will affect most test species; and there are large bodies of experimental data that indicate that exposures that are carcinogenic to animals are likely to be carcinogenic to man, and vice versa. Evidence that circumstances leading to cancer induction in humans are also applicable to experimental animals stems from the very first observation of chemical carcinogenesis the appearance of cancer of the

54 DRINKING WATER AND H"LTH scrotum in chimney sweeps, observed by the British surgeon, Percival Pott, in 1775. It was not until modern times that a substance implicated in human cancer was found to be carcinogenic in animals, when the Japanese scientists, K. Yamagiwa and K. Ichikawa, found in l91S that extracts from coal tar cause cancer when applied to the skin of experimental animals. Many pure carcinogenic chemicals have since been isolated from a wide variety of"tars" derived from incomplete combustion of organic matter, such as coal, wood, and tobacco. There is little doubt that these and other chemicals, alone or in combination, are responsible for the greatly increased incidence of lung cancer among smokers. With the possible exception of arsenic and benzene, all known carcinogens in man are also carcinogenic in some species, although not in all that have been tested. Principle 2 METHODS DO NOT NOW EXIST TO ESTABLISH A THRESHOLD FOR LONG TERM EFFECTS OF TOXIC AGENTS With respect to carcinogenesis, it seems plausible at first thought, and it has often been argued, that a threshold must exist below which even the most toxic substance would be harmless. Unfortunately, a threshold cannot be established experimentally that is applicable to a total population. A time-honored practice of classical toxicology is the establishment of maximal tolerated (no-e~ect) doses in humans based on finding a no-observed-adverse-e~ect dose in chronic experiments in animals, and to divide this dose by a "safety factor" of, say, 100, to designate a "safe" dose in humans. There is no scientific basis for such estimations of safe doses in connection with carcinogenesis. For example, even if no tumors are obtained in an assay of 100 animals, this means only that at a 95% confidence level, the true incidence of cancer in this group of animals is less than 3%. Even if we were to carry out the formidable task of using 1,000 animals for assay and no tumors appeared, we could only be 95% sure that the true incidence were less than 0~3%. Obviously, 0.3% is a very high risk for a large human population. In fact, there are no valid reasons to assume that false-negative results of carcinogenicity tests are much less frequent than false-positive ones. To dismiss all compounds that did not induce tumors in one or two mouse and rat experiments as noncarcinogenic is wrong. Labeling as "carcinogens" all substances that gave rise to increased incidence of tumors is justified only if there is conclusive evidence of a causal relationship. The "relative risk" of compounds that are not found to

Chemical Contaminants: Safety And Risk Assessment 55 induce tumors in animal experiments must also be considered. But this requires evaluation of data other than those collected in chronic toxicity studies on rodents. Experimental procedures of bioassay in which even relatively large numbers of animals are used are likely to detect only strong carcinogens. Even when negative results are obtained in such bioassays, it is not certain that the agent tested is unequivocally safe for man. Therefore, we must accept and use possibly fallible measures of estimating hazard to man. This reasoning leads us to the statement of Principles 3 and 4. Principle 3 THE EXPOSURE OF EXPERIMENTAL ANIMALS TO TOXIC AGENTS IN HIGH DOSES IS A NECESSARY AND VALID METHOD OF DISCOVERING POSSIBLE CARCINOGENIC HAZARDS IN MAN The most commonly expressed objection to regulatory decisions based on carcinogenesis observed in animal experiments is that the high dosages to which animals are exposed have no relevance in assessment of human risks. It is therefore important to clarify this crucial issue. Practical considerations in the design of experimental model systems require that the number of animals used in experiments on long-term exposure to toxic materials will always be small, compared with the size of the human populations similarly at risk. To obtain statistically valid results from such small groups of animals requires the use of relatively large doses so that effects will occur frequently enough to be detected. For example, an incidence as low as 0.01% would represent 20,000 people in a total population of 200 million and would be considered unacceptably high, even if benefits were sizable. To detect such a low incidence in experimental animals directly would require hundreds of thousands of animals. For this reason, we have no choice but to give large doses to relatively small experimental groups and then to use biologically reasonable models in extrapolating the results to estimate risk at low doses. Several methods of making such calculations have been considered and used, but we think that the best method available to us today is to assume that there is no threshold, and that the incidence of tumors is directly proportional to dose. However, it is important to recognize that such calculations may give either too small or too large an estimate risk. The actual risk to humans might be even greater over a human lifetime, because it is about 35 times that of a mouse; and there is evidence that the risk of cancer increases rapidly with the length of exposure. Moreover, experimental assays are conducted under controlled dietary and environ

56 DRINKING WATER AND H"LTH mental conditions with genetically homogeneous animals, whereas humans live under diverse conditions, are genetically heterogeneous, and are likely to include subpopulations of unusual susceptibility. It should be emphasized that these general considerations give only a minimal estimate of human risk; moreover, they do not take into consideration differences in susceptibility between species. For example, beta-naphthylam~ne is well established as a human carcinogen on the basis of epidemiologic studies of occupationally exposed workers, whereas experiments have not shown it to be carcinogenic in the hamster, which is relatively resistant. Not all substances that cause a given incidence of cancer in the rat are equally carcinogenic for man. This means that chronic-toxicity studies, which are imperfect assay systems for carcinogenicity testing, should not be used as the sole criterion in the assessment of risk. Principle 4 MATERIAL SHOULD BE ASSESSED IN TERMS OF HUMAN RISK, RATHER THAN AS SAFE OR UNSAFE The limitations of the current experimental techniques do not allow us to establish safe doses, but with the help of statistical methods we may be able to estimate an upper limit of the risk to human populations. To calculate such a risk, we need data to estimate population exposure; a valid, accurate, precise, and reproducible assay procedure in animals; and appropriate statistical methods. Several general guidelines may be presented. First, no rigid, generally applicable procedure can be recommended for testing all toxic agents. Substances diner too much in their overall effects, and the design of appropriate assays will ultimately have to be left to the well-informed judgment of expert investigators. If substances that affect large populations are found to be carcinogenic, experiments of much wider scope may have to be conducted to obtain more detailed information on their possible ejects in humans. As a pragmatic guideline, it would be desirable to test a compound for carcinogenicity in at least two species, such as the mouse and the rat, and the strains of animals used should have a rather low incidence of spontaneous tumors under the conditions of the test. It is important to include "positive" controls, with known carcinogens, under the same conditions used for the test animals. This has been a point of considerable controversy. Experiments should be conducted over as much of the lifetime of the experimental animal as possible. The highest dose should be the

Chemical Contaminants: Safety And Risk Assessment 57 maximum that is tolerated without shortening the lifespan through causes Other than cancer. Every animal, whether it dies during the exposure period or is sacrificed at the end of the experiment, should be examined grossly and microscopically, and all toxic effects (not only cancer) should be noted. Risk constitutes but half the essential comparison that should be made in the assessments of human hazard; the other half is benefit to the exposed population of the agent for which hazard has been identified. Decisions cannot involve merely the risk. Statements of benefits should include the nature, extent, and recipient of the benefit. Technology has always been associated with some risk. But the acceptability of risk should depend on the specific benefits derived, the nature of the population exposed, and the availability of practical alternatives. It is not possible to guarantee a risk-free society; nor is a risk-free society necessarily the best society. It is often necessary to accept the risks of chemicals such as drugs and pesticides when the benefits warrant their use. Risks imposed on persons who gain no benefits are generally not acceptable. Personal choice and personal values enter into the risk- benefit comparison. For major benefits- for example, in the treatment of otherwise incurable or incapacitating diseases much higher risks are allowable than otherwise. An important principle in risk-benefit assess- ment is that each person must be allowed the widest possible choice supported by full information on risks, as well as benefits so that intelligent choices may be made. The benefit portion of the equation should be well defined by knowledgeable experts, and based on data at least as good as the risk data. It is important, therefore, that the benefit-risk comparisons be established with the active cooperation of people who are qualified to assess the usefulness of a substance and the consequences to those in need of it, as well as to the population at large. Finally, mankind is already exposed to many carcinogens whose presence in the environment cannot be easily controlled. In view of the nature of cancer, the long latent period of its development, and the irreversibility of chemical carcinogenesis, it would be highly improper to expose the general population to an increased risk if the benefits were small, questionable, or restricted to limited segments of the population. Principles To Be Used for Noncarcinogens and Nonmutagens The nature and reversibility of the toxic eject must be considered. For carcinogens that are not shown to be mutagens, some sort of extrapolation must be postulated.

58 DRINKING WATER AND H"LTH For noncarcinogens for which it seems likely that there are thresholds for toxic ejects, the acceptable dose should be below the threshold. If a threshold cannot be shown, the acceptable dose must be related to the data from animal experimentation and consideration of the seriousness of the toxic effects, as well as the likelihood and ease of reversibility, the variability of the sensitivity of the exposed population, and the economic and health-related importance of the material. RESEARCH RECOMMENDATIONS Research must be supported to develop an understanding of the mechanisms by which water pollutants produce toxic effects. This includes pharmacokinetics, tox~cation-detoxication mechanisms, and biochemical and pathological mechanisms of action. Estimates of margin of safety can be made more precisely and rationally as more is known about what happens to a chemical in the body and what the chemical or its metabolites do to the body. The results of such research also are necessary to develop rapid, inexpensive, accurate screening tests for various critical forms of toxicity. It is recognized that much of this research is going on, but the Committee is convinced that more must be done. In protecting the population of the United States from environmental pollution there is no more important or potentially- productive effort than the support of this kind of research. Since these studies are long-term in nature and must be closely coupled to basic biomedical research, they should be supported primarily by research rather than regulatory agencies. There are many research needs in the field of chronic disease epidemiology. Manpower is in critically short supply. There are critical problems of data resources. Research on statistical methods and mathematical models for estimat- ing low dose ejects should be encouraged. Statistical work is pratically nonexistent for effects other than carcinogenesis. Although a consider- able effort has been expended on dose-response estimation for carcino- genesis, very little has been done on species variability and susceptible subgroups. This area could at least be studied from an empirical standpoint so that there would be a better understanding of the precision of low-dose risk estimates. These recommendations are summarized below: 1. Studies of the physiological and biochemical mechanisms by which the toxic substances in water produce their ejects.

Chemical Contaminants: Safety And Risk Assessment 59 2. Development of rapid, inexpensive, and precise tests to identify substances that may produce important toxic ejects at low doses and dose rates. 3. Epidemiological studies of chronic disease. 4. Research on statistical methods and analytical models for describ ing and estimating the ejects of long exposure to low doses of toxic substances. Studies should not be limited to carcinogenesis and should consider, also, differences between species, and particularly sensitive subgroups in the population. REFERENCES Ames, B.N. 1976. Carcinogenicity tests. (Letters) Science 191 :241-245. Arley, N., and S. Iversen. 1952. On the mechanism ofexperimentalcarcinogenesis.VI. Hit theory interpretation of some experiments of Berenblum and Shubiko Acta Path. Microbial. Scand.31: 164 171. Armitage, P., and R. Doll. 1954. The age distribution of cancer and a multi-stage theory of carcinogenesis. Br. J. Cancer 8: 1-12. Armitage, P., and R. Doll. 1957. A two-stage theory of carcinogenesis in relation to the age distribution of human cancer. Br. J. Cancer 11: 161-169. Armitage, P., and R. Doll. 1961. Stochastic models for carcinogenesis. In Proceedings of the Fourth Berkeley Symposium on Mathematical Statistics and Probability, vol. 4, pp. 19- 38. University of California Press, Berkeley and Los Angeles. Barendsen, G.W. 1975. The effectiveness of small doses of ionizing radiations for the induction of cell reproductive death, chromosomal changes and malignant transforma- tion. Presented at 5th Symposium on Microdosimetry, Verbania Pallanza, Italy, Sept. 22. Berenblum, I., and P. Shubik. 1947. A new quantitative approach to the study of the stages of chemical carcinogenesis in the mouse's skin. Br. J. Cancer 1:383-391. Berenblum, I. 1974. Carcinogenesis as a Biological Problem. North Holland Publ. Co., Amsterdam/Oxford. 376 pp. Brown, C.C. 1975. Personal Communications. Brown, J.M. 1976. Linearity versus nonlinearity of dose-response for radiation carcinogen- esis. Health Physics 31:231-245. Burch, P.R.J. 1960. Radiation carcinogenesis: A new hypothesis. Nature 185:135-142. Bush, M.T., G. Berry, and A. Hume. 1966. Ultra-short-acting barbiturates as oral hypnotic agents in man. Clin. Pharmacol. Ther. 7:373-378. Casarett, L.J., and J. Doull, eds. 1975. Toxicology: The Basic Science of Poisons, pp. 11-25. Macmillan Publ. Co., New York. Cleaver, J.E. 1968. Defective repair replication of DNA in xeroderma pigmentosum. Nature 218:652~56. Conney, A.H., and J.J. Burns. 1972. Metabolic interactions among environmental chemicals and drugs. Science 178:576-586. Cattier, H., M.W. Hess, H.U. Keller, P. Luscieti, and B. Sordat. 1974. Immunological deficiency states and malignancy. In Interaction of Radiation and Host Immune Defense  Mechanisms in Malignancy. Proceedings of a Conference at Greenbrier, West Virginia, March, pp. 3(~4.

60 DRINKING WATER AND H"LTH Crump, K.S., D.G. Hoel, C.H. Langley, and R. Peto. 1976. Fundamental carcinogenic processes and their implications for low dose risk assessment. Cancer Res. 36: 2973-2979. DHEW. 1977. Approaches to determining the mutagenic properties of chemicals: Risk to future generations. Report of Committee to Coordinate Toxicology and Related Programs, Department of Health, Education and Welfare. Obtainable from: Box 12233, Research Triangle Park, N.C. Drake, J., Chairman. 1975. Environmental mutagenic hazards, Committee 17 Report. Science 187:503-514. Falk, H.L. 1971. Anti-carcinogenesis-An alternative. In B.L. van Duuren and B.A. Rubin, eds. Inhibition of Carcinogenesis. Progress in Experimental Tumor Research 14: 10 137. S. Karger, New York. Fisher, J.C., and J.H. Holloman. 1951. A new hypothesis for the origin of cancer foci. Cancer 4:916-918. German, J. 1972. Genes which increase chromosomal instability in somatic cells and predispose to cancer. Prog. Med. Genet. 8:61-101. Gillette, J.R., J.F. Mitchell, and B.B. Brodie. 1974. Biochemical mechanism of drug toxicity. Ann. Rev. Pharmacol. 14:271-288. Guess, H.A., and K.S. Crump. 1976. Low dose rate extrapolation of data from animal carcinogenicity experiments-Analysis of a new statistical technique. Math. Biosci. 32: 15-36. Hammer, W., and F. Sjoqvist. 1967. Plasmalevelsofmonomethylatedtricyclicantidepres- sants during treatment with ipramine-like compounds. Life Sci. 6: 1895-1903. Hoel, D.G., D.W. Gaylor, R.L. Kirschstein, U. Saffiotti, and M.A. Schneiderman. 1975. Estimation of risks of irreversible, delayed toxicity. J. Toxicol. Environ. Health 1: 133- 151. Iversen, S., and N. Arley. 1950. On the mechanism of experimental carcinogenesis. Acta Pathol. Microbiol. Scand. 27:773~03. Jones, H. 1976. Dose-effect relationships in carcinogenesis and the matter of threshold of carcinogenesis. Presented at the NIEHS Conference on the Problems of Extrapolating the Results of Laboratory Animal Data to Man and of Extrapolating the Results from High Dose Level Experiments to Low Dose Level Exposures, Pinehurst, N. C., March 10-12. Lawther, P.J., and R.E. Waller. 1976. Air pollutants and lung cancer-Recent trends in England and Wales. Presented at the Cold Spring Harbor Symposium. Lilienfeld, A.M. 1963. The epidemiology of breast cancer. Cancer Res. 23: 1503-1513. Lundin, F.E., J.K. Wagoner, and V.E. Archer. 1971. Radon Daughter Exposure and Respiratory Cancer. Quantitative and Temporal Aspects. NIOSH/NIEHS Joint Mono- graph 1. U.S. Department of Health, Education, and Welfare. Mabboubi, E., J. Kmet, P.J. Cook, N.E. Day, P. Ghadirian, and S. Salmasizadeh. 1973. Oesophageal cancer studies in the Caspian Littoral of Iran: The Caspian cancer registry. Br. J. Cancer 28: 197-214. Mantel, N., and W.R. Bryan. 1961. Safety testing of carcinogenic agents. J. Nat. Cancer Inst.27:455-470. Mantel, N., W.E. Heston, and J.M. Gurian. 1961. Thresholds in linear dose-response models for carcinogenesis. J. Nat. Cancer Inst. 27:203-215.  Mantel, N. 1963. Symposium on chemical carcinogenesis. Part IV. The concept of threshold in carcinogenesis. Clin. Pharmacol. Therap. 4: 104 109. McCann, J., E. Choi, E. Yamasaki, and B. N. Ames. 1975. Detection of carcinogens as mutagens in the Salmonella/microsome test: Assay of 300 chemicals. Proc. Nat. Acad. Sci. 72:5135-5139.

Chemical Contaminants: Safety And Risk Assessment 61 McCann, J., and B. N. Ames. 1976. Detection of carcinogens as mutagens in the Salmor~ella/microsome test: Assay of 300 chemicals: Discussion. Proc. Nat. Acad. Sci. 73: 95~954. Menguy, R., L. Desbaillets, W.F. Masters, and S. Okabe. 1972. Evidence for a sex-linked difference in aspirin metabolism. Nature 239: 102-103. Morrow, R.H., A. Kisuule, M.C. Pike, and P.G. Smith. 1976. Burkitt's lymphoma in Mengo Districts of Uganda: Epidemiologic features and their relationship to malaria. J. Nat. Cancer Inst. 56:479~83. Muller, H.J. 1951. Radiation damage to the genetic material. Sci. Prog. 7:93-94. Murphy, S.D. 1967. Malathion inhibition of esterases as a determinant of malathion toxicity. J. Pharmacol. Exp. Ther. 156:352-365. National Academy of Sciences-National Research Council. Division of Medical Sciences. 1972. The Effects on Populations of Exposure to Low Levels of Ionizing Radiation. Washington, D.C.217 pp. National Academy of Sciences. 1975. Principles for Evaluating Chemicals in the Environ- ment. Washington, D.C. 454 pp. Nordling, C.O. 1953. A new theory on the cancer inducting mechanism. Br. J. Cancer 7:68- 72. Robbins, J.H., K.H. Kraemer, M.A. Lutzner, B.W. Festoff, and H.G. Coon. 1974. Xeroderma pigmentosum. An inherited disease with sun sensitivity, multiple cutaneous neoplasms, and abnormal DNA repair. Ann. Int. Med. 80:221-248. Rothman, K.J., and A.Z. Keller. 1972. The effect of joint exposure to alcohol and tobacco on risk of cancer of the mouth and pharynx. J. Chron. Dis. 25:711-716. Selikoff, I.J., E.C. Hammond, and J. Churg. 1968. Asbestos exposure, smoking and neoplasia. J. Am. Med. Assoc. 204: 106-112. Shellabarger, C.J. 1976. Modifying factors in rat mammary gland carcinogenesis. In J.M. Yuhas, R.W. Tennant, and J. D. Regan, eds. Biology of Radiation Carcinogenesis, pp. 31~3. Raven Press, New York. Stocks, P. 1953. A study of the age curve for cancer of the stomach in connection with a theory of the cancer producing mechanism. Br. J. Cancer 7:407-417. Strong, L.C. 1976. Susceptible subgroups. Presented at the NIEHS Conference on the Problems of Extrapolating the Results of Laboratory Animal Data to Man and of - Extrapolating the Results from High Dose Level Experiments to Low Oose Level Exposures. Pinehurst, N.C., March 10-12. Terracini, B., P.N. Magee, and J. Barnes. 1967. Hepatic pathology in rats on low dietary levels of dimethyloitrosamine. Br. J. Cancer 21 :559-565. Turner, M.D. 1975. Some classes of hit-theory models. Math. Biosci. 23:219-235. Urbach, F. 1969. Biological Effects of Ultraviolet Radiation. Pergamon Press, Oxford, England. Walburg, H.E., Jr. 1974. Experimental radiation carcinogenesis. In J. T. Lett, H. Adler, and M. Zelle, eds. Advances in Radiation Biology, vol. 4, p. 209-254. Academic Press, New York. Weber, W.W. 1976. Genetic variability and extrapolation from animals to man: Some perspectives on susceptibility to chemical carcinogenesis fom aromatic amines. Presented  at the NIEHS Conference on the Problems of Extrapolating the Results of Laboratory Animal Data to Man and of Extrapolating the Results from High Dose Level Experiments to Low Dose Level Exposures, Pinehurst, N.C., March 1~12. Wedderburn, N. 1970. Effect of concurrent malarial infection on development of virus- induced lymphoma in Balb/c mice. Lancet 2: 111~1116.

62 DRINKING HER MID H"~H ail, C.S. 1972. S~1ishcs vs. s^V racers and sdent~c ju_nt ~ He eviction of =~ far man. l~col. Pal. Sol. 21:45~3. Witty, A., and J.B. KeDer. inactive ~eodes of Crests. in press. Yummy, K., a~ K. Iffy. 1915. ~h~n~De Stumble Oar die Pa~o~nese tier Epi~elial~s~ste. Mitt. red. Fat. go 15: 295~.