NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

National Research Council (US) Committee on Comparative Toxicity of Naturally Occurring Carcinogens. Carcinogens and Anticarcinogens in the Human Diet: A Comparison of Naturally Occurring and Synthetic Substances. Washington (DC): National Academies Press (US); 1996.

Cover of Carcinogens and Anticarcinogens in the Human Diet

Carcinogens and Anticarcinogens in the Human Diet: A Comparison of Naturally Occurring and Synthetic Substances.

Show details

5Risk Comparisons

Previous chapters of this report have discussed the ways in which dietary carcinogens are identified and documented the presence in the human diet of both naturally occurring and synthetic substances that may possess carcinogenic potential. In this chapter, we discuss the relative risks posed by natural and synthetic dietary carcinogens.

Throughout this report, the term diet is used to refer to foods and beverages consumed intentionally and customarily in the U.S., not as a result of accident or deprivation. As in previous chapters, it is convenient to differentiate among constitutive, derived, acquired, pass-through, and added naturally occurring food chemicals. The definition of a carcinogen adopted in this report is that used by the International Agency for Research on Cancer, namely any agent capable of increasing the incidence of malignant neoplasia. Operationally, the committee treats as carcinogens those agents classified in certain IARC categories (i.e., 1, 2A, and 2B) and in the National Toxicology Program's (1994) Annual Report as known to be or reasonably anticipated to be carcinogenic.

The level of risk associated with a carcinogenic agent depends on both the potency of the agent and on the level of exposure to that agent: Carcinogenic potency can be estimated using clinical and epidemiologic data on humans or toxicologic data derived from animal bioassays. Exposure to carcinogenic agents present in the diet depends on both food consumption patterns and the concentration of those agents in foods consumed. Food consumption data can be collected through the use of food diaries, or by using questionnaires designed to gauge the frequency with which specific foods are consumed or to identify by recall those foods recently consumed. Concentrations of carcinogenic agents in the food supply can be determined by analytic techniques, such as chemical analyses for pesticide residues present on foods.

Inferences about dietary cancer risks are complicated by several factors. Diet is a complex mixture containing a large number of micro-and macroingredients. Components of the diet may interact with one another in a synergistic or antagonistic way. Some dietary components, such as aflatoxin, might increase cancer risks, whereas others, such as fruits and vegetables rich in antioxidants, might reduce cancer risk.

Food-consumption patterns can be highly variable even among individuals in the same population subgroup. Food consumption varies depending on availability, ethnic customs, age, economics, and other factors. Chemical contaminants, extraneous matter, and pesticide residues can be present in food at variable concentrations. Food composition and products derived from preparation and processing of food are also variable. In addition to variability in dietary intakes, individuals may also vary with respect to their susceptibility to food components with carcinogenic potential. Each of these sources of dietary variability can effect individual exposures to food chemicals, as well any associated risks.

Estimates of potential dietary cancer risks are subject to considerable uncertainty; for example, epidemiological studies have failed to provide unambiguous evidence of the effects of dietary fat on cancer risk. Estimates of potential cancer risks associated with low levels of individual food chemicals derived on the basis of laboratory results are highly uncertain. The application of animal cancer test data to humans requires extrapolation from the high doses used in laboratory studies to much lower doses corresponding to concentrations in the human diet, and extrapolation from animals to humans. The joint effects of ingestion of multiple agents in the form of complex dietary mixtures are also difficult to define. Consequently, in evaluating dietary cancer risks, it is important that both uncertainty and variability be recognized and, if possible, characterized.

Recognizing that estimates of cancer risk are uncertain, this chapter focuses on the following questions:

  • Does diet contribute to an appreciable proportion of human cancer?
  • What are the relative contributions of naturally occurring and synthetic agents to dietary cancer risk?
  • Are there significant interactions between either synthetic or naturally occurring carcinogens and anticarcinogens in the diet?

To determine whether synthetic or natural chemicals classified as carcinogens pose the greater risk, it is necessary to know 1) the identity of the carcinogens present in the diet; 2) levels of ingestion of specific dietary carcinogens, both natural and synthetic; and, 3) the carcinogenic potency of these chemicals. Although this information might be used to evaluate the potential risks associated with individual food chemicals, it is more difficult to evaluate the overall risk posed by carcinogens present in the diet as a whole. The human diet is a complex mixture of food chemicals that interact in ways that are not generally well understood. Consequently, much of the discussion in this chapter of the comparative risks of naturally occurring and synthetic carcinogens present in the diet will focus on individual substances rather than mixtures.

The levels of exposure to dietary carcinogens vary widely, depending on food-consumption patterns and dietary concentrations of carcinogenic substances. Because consumption patterns vary among individuals, it is important to consider the range of exposures within the population of interest, particularly those persons with high dietary intakes of naturally occurring or synthetic carcinogens.

For purposes of risk comparison, a quantitative measure of the potency of naturally occurring and synthetic carcinogens is required. A widely used measure of carcinogenic potency is the TD50' defined as the level of exposure resulting in an excess lifetime cancer risk of 50% (Peto et al. 1984, Sawyer et al. 1984). The TD50 can be derived from either epidemiologic or toxicologic investigations. Because the TD50 is often derived from high-dose experimental data, it does not necessarily provide an appropriate basis for making inferences about cancer risks at low levels of exposure. To obtain a measure of carcinogenic potency that is closer to human exposure levels, the committee also used the TD01 as an index of carcinogenic potency. Because risk is a function of exposure and potency, the ratio of exposure to potency has been proposed as a means of comparing the relative risk of exposure to different carcinogens (Ames and Gold 1987).

This chapter reviews existing data on the comparative potency of naturally occurring and synthetic carcinogens. Specifically, the committee compiled a database on the carcinogenic potencies of 37 natural and 70 synthetic carcinogens known to occur in the diet. These substances were identified as being carcinogenic in animals or humans by either the U.S. National Toxicology Program or the International Agency for Research on Cancer. All of these chemicals were classified by the NTP as known or reasonably anticipated to be carcinogens or by IARC as known (Group 1), probable (Group 2A), or possible (Group 2B) human carcinogens. Although the potency of naturally occurring dietary carcinogens as a group was on average greater than that of the synthetic carcinogens, the potencies of both types vary widely with considerable overlap. As discussed above, they are also subject to considerable uncertainty. Based on this limited number of chemicals, which might not represent the universe of naturally occurring and synthetic dietary carcinogens, it appears difficult to distinguish between the potencies of the two classes.

The overall contribution of diet to the human cancer burden is also considered. Although tobacco and diet are thought to account for the large majority of human cancer, the contribution of the diet is less well understood than that of tobacco and is subject to far greater uncertainty as to the attributable risk fraction. Recent reviews of the causes of human cancer have suggested that synthetic carcinogens present in the diet might be responsible for a very small fraction of the human cancer burden (Ames et al. 1995, Higginson 1988), due in part to regulations that have limited the use of pesticides, including those with carcinogenic potential, and preclude the use of carcinogenic substances as direct food additives. In terms of risk from food substances, calories and fat may represent the most important naturally occurring dietary constituents. Food chemicals produced naturally by plants for self-defense have not been investigated to the same extent, and therefore the degree to which they contribute to human cancer is less clear.

Monitoring Food Consumption

Sources of Information

Pesticides in the Diets of Infants and Children (NRC 1993a) addresses issues of food and water consumption in the U.S. population. Directed primarily at the pediatric population, the report discusses approaches to quantifying food and water consumption in the population at large, and the limitations of methods for food consumption monitoring.

National food surveys are conducted by the U.S. Department of Agriculture (USDA) and the Department of Health and Human Services. The USDA's Human Nutrition Information Service (HNIS) conducts a comprehensive Nationwide Food Consumption (NFC) Survey about every ten years. In the interim, the service conducts Continuing Surveys of Food Intakes of Individuals (CSFII). Both the 1977-1978 and the 1987-1988 NFC Surveys were reviewed in Pesticides in the Diets of Infants and Children. However, the 1987-1988 survey was considered less reliable for estimating dietary exposures because of the low response rate (34%). Other limitations of the 1987-1988 USDA survey are discussed by the Government Accounting Office (1991).

Although subject to serious limitations, the USDA surveys provide the only comprehensive data currently and publicly available on food consumption by people of all ages. Some surveys focus on segments of the U.S. population: for example, the 1985-1986 CSFII emphasized women 19-50 years old and their children ages 1-5, a sample of low-income women and their children, and in 1985 only, men ages 19-50 years. In the CSFII 1989, 1990, and 1991 surveys, data were collected on individuals of both sexes in all age classes, with response rates higher than those of the 1987-1988 NFC Survey (over 50%). The results of the 1989 and 1990 surveys are available commercially (Technical Assessment Systems 1995a,b) and on computer media from the National Technical Information Service.

There is substantial uncertainty in such food consumption data due to a variety of factors, including recall bias, measurement error, and recording errors. The fact that the numbers of people surveyed are relatively small and response rates poor makes generalization to the U.S. population at large difficult. The optimal system for collecting and validating data on food consumption has yet to be developed. Ideally, complete and accurate records of the types and quantities of food consumed by the survey respondents could be used as a reference against which different surveys providing estimates of consumption could be compared. Because of the problems in measuring actual food intake, however, validation studies have in the past focused on the comparisons of results obtained from different surveys using different data collection methods.

Another series of food consumption surveys that provides nationwide data is conducted by DHHS's National Center for Health Statistics (NCHS). Since 1960, the center has conducted seven health examination surveys of the U.S. population. The National Health and Nutrition Examination Surveys (NHANES), including the recently completed NHANES III, were designed to obtain representative information on the health and nutritional status for the U.S. population through health and medical histories, dietary interviews, direct physical examinations, and laboratory measurements. NFC and NHANES surveys deal primarily with nutritional considerations and are less useful for evaluating ingestion of naturally occurring chemicals and food additives.

NHANES I (1971-1974) and NHANES II (1976-1980) sought data on medical conditions, especially nutrition-related disorders (obesity, growth retardation, anemia, diabetes, atherosclerotic cardiovascular diseases, hypertension, and deficiencies of vitamins or minerals). Both were directed at the civilian, noninstitutional population. (Excluded were the homeless, residents of hotels, rooming houses, dormitories, Native American reservations, military posts, prisons, hospitals, and residential treatment centers for drug addiction, alcoholism, and obesity.) Both surveys covered the 48 contiguous states, although Alaska and Hawaii were included in NHANES II. These surveys are discussed in detail in the NRC report Diet and Health (NRC 1989a). NHANES II data have been used to evaluate the proportion of the population at risk for deficiencies of vitamin A, vitamin C, folate, iron, zinc, and protein.

The target population for NHANES III was the U.S. civilian, noninstitutional population aged 2 months or older. The survey design called for a stratified sample of counties, blocks, and persons randomly selected from households. National samples were drawn during 1988-1991 and 1991-1994. Eighty-one counties were selected from 26 states; from these, approximately 40,000 persons of all races were selected, and about 30,000 agreed to participate in the medical examination. Precise estimates of health characteristics were needed for relatively small population subgroups (children, older persons, black, and Mexican Americans), which were subject to oversampling.

Some of the 30 topics investigated in NHANES III were high blood pressure, high blood cholesterol, obesity, passive smoking, lung disease, osteoporosis, HIV, hepatitis, helicobacter pylori, immunization status, diabetes, allergies, growth and development, blood lead, anemia, food sufficiency, and dietary intake, including fats, antioxidants, and nutritional blood measures. Results from NHANES III are being analyzed by NCHS and are not yet available.

Although NHANES data are extensive and derive from a broad range of measurements, the data are of limited use in the study of chronic diseases, in part because of the sample size, response rates, and recall bias. NHANES provides only cross-sectional data on a periodic basis. Diet and Health provides an in-depth discussion of the limitations of NHANES data.

A number of factors need to be considered when using food composition and consumption data in estimating dietary cancer risks. For example, food composition databases do not contain data on the concentration of many of the potential carcinogenic constituents found in foods (USDA 1992). Information on macronutrients and on certain micronutrients with carcinogenic or anticarcinogenic potential is available for a large variety of foods. However, data on many other microconstituents, naturally occurring and synthetic, are lacking. For example, little information is available on plant biocides, non-nutritive plant products with anticarcinogenic potential, or compounds generated by cooking. Furthermore, databases frequently do not take into consideration variability among food samples, population groups, and individuals, or consumption patterns that vary over time.

Sources of Variation in Food Composition and Consumption

Individuals vary in their dietary habits, people of different ages have different dietary requirements, and the concentration of food constituents can differ substantially. This section addresses some of the sources of variation in exposure to carcinogens in the diet.

Dietary risk assessments should take into consideration that food samples can vary greatly in composition (see Chapter 2). For example, plants may produce natural chemicals for purposes of self-defense when they come in contact with certain pests (Harborne 1993). The concentrations of these substances in a particular food can vary considerably among samples, depending on the extent of the stress to the plant prior to harvesting. Factors such as storage, cooking, and pesticide application rates also have effects on food composition. This variability could result in substantial seasonal, geographic, and individual variation among food samples and, consequently, among human exposure levels. Similarly, cultivars of fruits and vegetables may differ in the content of naturally occurring constituents. For example, a cultivar of Idaho potatoes had to be taken off the market when it was found to contain toxic levels of the neurotoxin solanine (IFBC 1990).

Dietary assessments should also allow for consumption patterns that may vary among population subgroups (defined in terms of sex, ethnicity, income, and other characteristics) (Kolonel et al. 1983, USDA 1987). Using data from the 1977-1978 USDA survey, Pesticides in the Diets of Infants and Children concluded that infants and children consume more calories relative to body weight than adults, eat far less-varied diets, and consume far greater amounts of milk in some form. Because of the lack of diversity of infant diets, infants can consume much greater quantities of certain foods than adults: the average 1-year-old consumes approximately 40-fold more apple juice relative to body weight than the average adult (Murdoch et al. 1992). Other population subgroups whose dietary habits differ from those of the general population include vegetarians and religious groups with special dietary restrictions. However, such groups may also exhibit nondietary differences from the general population, with respect to other factors such as socio-economic status and smoking habits (Lyon et al. 1980).

Water is a major component of food (see Chapter 2) containing trace levels of a number of chemicals and should be considered in any analysis of dietary risk. This fact is particularly important when estimating the risks to children from dietary exposures. Pesticides in the Diets of Infants and Children considered three types of water in its analysis of dietary risks: water intrinsic to food, tap water added to food during preparation, and the direct consumption of tap water. The report indicates that dietary sources of water represented by fruits, liquids (especially fruit juices and milk), and vegetables are greater for infants than for older children or adults, and should be considered in estimating the risk from dietary exposure to potential human carcinogens. Such age-dependent differences in dietary patterns need to considered when evaluating lifetime cancer risks (Goddard et al. 1995).

Dietary patterns can change markedly over time with changes in food preferences and the introduction of new foods. For example, artificial sweeteners were unknown until the discovery of saccharin in 1879 (cf. Arnold et al. 1983); however, since its approval for widespread use as an artificial sweetener, aspartame has become a common constituent of the diets of many Americans. In addition, fabricated and genetically engineered foods introduced in recent years have also afforded consumers with new dietary choices.

Finally, food consumption data and survey methods need to be standardized to make them more useful in estimating exposures and determining risk. Currently, there is no simple, uniform method for conversion of a food, as consumed, to its components in terms of raw agricultural constituents. In addition, surveys should be coordinated among concerned organizations and carried out in a timely manner in order to identify trends in food and water consumption.

Factors Affecting Susceptibility

Individuals within a subgroup may vary with respect to their susceptibility to the toxic effects of those agents. Most carcinogens, whether naturally occurring or synthetic, will be metabolized in the body to a greater or lesser degree by different individuals. Some of these substances may be activated to their carcinogenic derivatives, while others are detoxified. Some of the enzymes involved in these reactions, such as certain of the cytochrome P450 mixed function oxidases, are not only inducible but also encoded by polymorphic genes (Idle et al. 1992). As a result, individuals vary in their susceptibility to carcinogens (Omenn et al. 1990).

Prescription and over-the-counter drugs may affect specific constituents of food, such as cholesterol and fat. Several million people take cholesterol-lowering agents, and new drugs are being developed to block lipid absorption. Information on the use of pharmaceutical products that may affect dietary cancer risks is therefore of interest.

Dietary Exposure to Potential Carcinogens and Anticarcinogens

Despite a substantial degree of measurement error in assessing food intakes, as well as limitations in the current food composition and consumption databases, useful estimates of human exposure to some naturally occurring constituents of foods can often be derived. Estimating exposures to synthetic agents is more problematic. Residues of pesticides, chemicals added in processing, and carcinogens produced during cooking can be extremely variable in foods and are usually present in microquantities. In addition, exposure estimates based on dietary intakes (as opposed to serum or tissue measurements) do not account for the bioavailability of food constituents, which depends on many factors, such as other foods consumed at the same time, and the manner in which constituents are structurally bound in food.

In the sections that follow, dietary exposure levels to naturally occurring and synthetic carcinogens and anticarcinogens are discussed. Major food sources and concentrations found in those foods are discussed in detail in Chapter 2.

Naturally Occurring Carcinogens

Table 5-1 classifies chemicals identified as naturally occurring carcinogens that may be present in the diet into five categories: constitutive, derived, acquired, pass-through, and added. The chemicals listed in each category have been classified by IARC or NTP as carcinogens. The remainder of this section discusses potential exposure to these substances and focuses on a few agents in each class for which relatively high intakes are expected.

Table 5-1. Naturally Occurring Animal and Human Carcinogens That Might Be Present in U.S. Diets.

Table 5-1

Naturally Occurring Animal and Human Carcinogens That Might Be Present in U.S. Diets.

Constitutive Exposures

Included in this group are the sex hormones (e.g., estradiol 17Image img00010.jpg, estrone, acetate, progesterone, testosterone), metabolic intermediates in plants (e.g., acetaldehyde), caffeic acid, and natural furocoumarins (5-methoxy and 8-methoxypsoralen). The sex hormones are present in fairly low quantities in consumed meats, although exposure to some can be increased above natural levels through their legal and illegal use as growth promoters in meat production.

Acetaldehyde is a metabolic intermediate in the formation of ethanol during anaerobic respiration, a process which plant tissues can only tolerate for brief periods. Acetaldehyde has been identified as a volatile component of essential oils from a variety of fruit and spice plants, and as a natural constituent of numerous edible berries and other fruits. Acetaldehyde is also formed in animals, during the intracellular oxidation of ethanol, and is present at relatively low concentrations in meat. However, the extent that low levels of acetaldehyde in foods poses a carcinogenic risk is unclear.

The risk for oral exposure appears to be substantially less than for inhalation exposures (ILSI, 1993). For this reason, estimates of the amount of exposure to acetaldehyde via the diet are not made.

Caffeic acid, a metabolic precursor of lignin, a structural polymer found in all land plants, is ubiquitous in the food supply. It accumulates primarily in conjugated forms, which can be hydrolyzed to free caffeic acid in the digestive trace (see Chapter 2). In the conjugated form, it is widely distributed in fruits and vegetables. Concentrations of these conjugates and free caffeic acid have been measured in a variety of food plants by Herrmann (1989) and colleagues. The most recently reported values for caffeic acid in these food sources are assumed to be the most reliable, because of the use of improved analytical techniques involving gas and high performance liquid chromatography. Using the concentrations reported by Herrmann and colleagues and USDA food consumption data, rough estimates of human intake are obtained and presented in Table 5-2.

Table 5-2. Concentrations of Caffeic Acid, Selected Foods, and Predicted Intake of Caffeic Acid from These Foods in the General Population.

Table 5-2

Concentrations of Caffeic Acid, Selected Foods, and Predicted Intake of Caffeic Acid from These Foods in the General Population.

Table 5-2 reports ranges rather than single values for caffeic acid intake, to reflect both the uncertainty and variability in caffeic acid exposure. Caffeic acid exists primarily as conjugates (esters and glucosides) in unprocessed food plants (see Chapter 2); while these conjugates may be converted to free caffeic acid during food processing and after ingestion, the extent of the conversion is unknown. Human ingestion of chlorogenic acids gave rise to urinary metabolites of caffeic acid in one study, but the amount of the conversion was not measured (Booth et al. 1957). Although other work has indicated that cholorogenic acids are hydrolyzed in the digestive tract of the rat before caffeic acid appears in the bloodstream, the amount of hydrolysis was not reported (Czok et al. 1974). A more recent study of the metabolism of caffeic acid by humans concluded that the phenolic acid is extensively and rapidly metabolized (i.e., within 4 hours), but accounted for only 11% of the acid administered (Jacobson et al. 1983). In calculating average intakes of caffeic acid for the general population, 100% hydrolysis of caffeic acid conjugates was assumed in deriving the upper bound and 10% hydrolysis in deriving the lower bound. Estimates of average daily intake of caffeic acid from food for the general population were 0.02-0.2 mg/kg. Daily intakes for high consumers of produce can be considerably greater (e.g., the 95th percentile estimates for children aged 1-6 are 0.2-3 mg/kg). Estimates of caffeic acid intake from coffee for moderately high consumers are also considerably higher: 0.9-9 mg/kg for the general population.

Most exposure to natural furocoumarins is derived from limes and other citrus and umbelliferous plants, with per capita exposure estimated to be 1.3 mg per day (Wegstaff 1991).


Included in this group are compounds generated by cooking (e.g., polycyclic aromatic hydrocarbons [PAHs], heterocyclic amines, benzene, and glycidaldehyde), and those compounds that appear in preserved or cured foods (e.g., nitrosamines) and in technologically altered foods.

PAHs are present in a variety of prepared foods. They can be endogenously produced, or enter food through environmental contamination from both natural (e.g., forest fires) and anthropogenic sources (IARC 1983). The potential for high exposures to benzo(a)-pyrene, one of the most potent carcinogenic PAHs, is greatest with consumption of charred meats, smoked fish, vegetable oils, tea, roasted coffee, and some fruits and vegetables.

The formation of heterocyclic amines during the cooking of proteinaceous foods was discussed earlier. Data are limited concerning the intake of these substances. Layton et al. (1995) analyzed the consumption of foods containing five of the principal heterocyclic amines by 3,563 persons who provided 3-day dietary records in a USDA-sponsored survey conducted in 1989. They calculated average intakes (ng/kg per day) of the five principal heterocyclic amines as follows: PhIP, 16.64; AC, 5.17; MeIQx, 2.61; DiMeIQx, 0.81; and IQ, 0.28. One of the study authors indicates that these calculated intake levels are probably (within a factor of 5) those consumed by the average person, but that the intake for high consumers of meats cooked ''well-done" (95th percentile) could be considerably greater than the average value (J.S. Felton, Lawrence Livermore National Laboratory, personal communication).

Preformed N-nitroso compounds may be present in the diet, mainly in foods cured with nitrate or nitrite. Cured meats and beer are the most important sources of nitrosamines. In 1981, the National Research Council (NRC 1981) estimated the daily intake of nitrosamines from dietary sources to be 1.1 Image img00014.jpgg; the same estimate was obtained for 1979 for N-nitrosodimethylamine intake by males in Germany (Preussmann 1984). Estimates of N-nitrosodimethylamine intake from food have recently been published for European countries, and U.S. intake levels are expected to be similar. Daily intakes of N-nitrosodimethylamine from food were estimated to be roughly 0.1 Image img00014.jpgg (≈ 0.0014 Image img00014.jpgg/kg-bw) in eastern France (Biaudet et al. 1994), 0.2 Image img00014.jpgg (≈ 0.003 Image img00014.jpgg/kg-bw in West Germany; Tricker et al. 1991), and, for all volatile nitrosamines, less than 0.1 Image img00014.jpgg in the Netherlands (Ellen et al. 1990).

N-nitrosodimethylamine levels in beer of 1-5 ppb were common in the early 1980's (Scanlan 1983), and in 1981 a National Research Council (NRC 1981) committee estimated a daily consumption of 0.9 Image img00014.jpgg (≈ 0.027 Image img00014.jpgg/kg-bw) from a level of 2.8 Image img00014.jpgg/l in beer (Scanlan and Barbour 1991). A recent analysis of nearly 200 U.S. and Canadian beers by Scanlan and Barbour (1991) found a mean concentration level of 0.074 Image img00014.jpgg/kg; current daily consumption (apparently for beer consumers) of N-nitrosodimethylamine from beer was estimated to be around 0.026 Image img00014.jpgg (≈ 0.00037 Image img00014.jpgg/kg-bw), or about 3% of the value a decade ago. The decrease in exposure from the 1981 NRC value was due to measures taken to reduce the formation of N-nitrosodimethylamine in malt. Concentrations of roughly 0.5 Image img00014.jpgg/kg were observed for a few types of beer, and thus high consumers of those beers would be exposed to relatively high levels of N-nitrosodimethylamine. Concentrations roughly double those observed by Scanlan and Barbour (1991) were reported for analyses of 170 retail samples of beer by Massey et al. (1990) (range <0.1 to 1.2 Image img00014.jpgg/kg, with mean of 0.2 Image img00014.jpgg/kg).

In addition to preformed N-nitroso compounds, humans are exposed to a wide range of nitrogen-containing compounds and nitrosating agents that can react in vivo to form N-nitroso compounds (Bartsch 1991), including N-nitrosodimethylamine (Pignatelli et al. 1991). Residual nitrites in cured meats and fish are an important source of nitrosating agents in the stomach (NRC 1989a). Potentially endogenous formation can lead to exposures substantially higher than from direct ingestion of preformed compounds. The above calculations do not account for endogenous formation of N-nitroso compounds.

Urethane is formed naturally in fermented beverages and foods, such as alcoholic beverages, leavened bread, soy sauce, yogurt, and olives, and has also been measured in milk (Battaglia et al. 1990, Zimmerli and Schlatter 1990, Dunn et al. 1991). High levels in beverages have been associated with the use of urea as a yeast food and the use of the antimicrobial agent diethylpyrocarbonate. Both of these uses are now prohibited in the U.S. Tabulations of reported measured levels in most fermented foods are in the low parts per billion, with the mean values falling below 1 ppb for cheese, milk, and yogurt, and below 10 ppb for bread (Battaglia et al. 1990, Dunn et al. 1991). Somewhat higher levels have been observed for soy sauce, with mean values reported ranging from 4.4 to 18 ppb. Single samples of other fermented foods also indicate possible levels in the low ppb range: olives (1.1 ppb), sauerkraut (0.3 ppb), orange juice (1.5 ppb), apple vinegar (3.3 ppb). Assuming that a level of 1 ppb occurs in milk products, taking mean levels reported in the literature for the other food items, and using the results from the USDA Continuing Survey of Food Intakes by Individuals (as codified by TAS 1995a,b), estimates for daily intake of urethane in food were obtained: 1.4 × 10-5 mg/kg for the general population, and 4.4 × 10-5 mg/kg for children aged 1-6, with an upper 95th percentile for this group of 9.5 × 10-5. The greatest contribution was from milk products, for which data are scanty and an upper bound concentration estimate was used; thus, the intake results should be seen as upper bound estimates. The estimate for the general population is in the range of that for the Swiss population published by Zimmerli and Schlatter (1991; 10-20 ng/kg-bw, i.e., 10 to 20 × 10-5 mg/kg). Zimmerli and Schlatter report intakes moderately higher than those received from food are associated with consumption of wine and spirits; moderate consumption of wine (95th percentile for general population) was associated with approximately a 5-fold increase over the mean population level. However, moderate daily consumption of stone fruit distillates (30 ml/day) can increase exposure by roughly 60 fold (Zimmerli and Schlatter 1991) to roughly 0.01 mg/kg.


Included in this group are naturally occurring chemicals absorbed by food organisms from the environment (e.g., lead) and mycotoxins (e.g., aflatoxin, ochratoxin, sterigmatocystin, and toxins derived from Fusarium moniliforme). Of the agents in this group studied for carcinogenicity and found to be carcinogenic, mycotoxins are predominant and have the greatest human exposure potential. Exposures to the mycotoxins aflatoxin, fumonisin B1, and sterigmatocystin are described below:

  • Aflatoxin. Levels of aflatoxin in crops vary geographically and over time, with the southeastern United States frequently referred to as an area where high levels in corn can occur. Field corn, which is primarily used for animal feed and milling, has been the primary concern. Sweet corn is considered to be of no major consequence as a source of dietary exposure to aflatoxin. Human consumption of peanuts and peanut products are the other major source of aflatoxin exposure. FDA routinely samples aflatoxin in corn and peanuts for human consumption and the American Peanut Product Manufacturers, Inc. has established a surveillance system for peanuts destined for human consumption. Using the results of the FDA surveillance for 1984-1989, and of a USDA survey (year not specified), the FDA recently estimated for a 60 kilogram person an average intake for aflatoxin B1 of 17 ng/day from corn and peanut products. The upper 90th percentile estimate was 40 ng/day (Springer 1994). Subgroups who regularly consume large amounts of certain corn products (e.g., grits, corn tortillas), in areas where aflatoxin contamination is frequent, may be exposed to larger quantities.
  • Fumonisin B1. Data on fumonisin occurrence are being developed and at present are limited. In a recent tabulation by Pohland (1994), fumonisin B1 was present in a majority of samples of field corn products, with levels as follows: corn meal, average approximately 1 ppm (47/48 positive); corn flour, 0.12-0.25 (2/2); grits, average 0.2 (5/5); corn bran cereals, 0.06 - 0.33 (5/5); corn flakes 0.01 - 0.055 (4/19); fiber cereal 0.06-0.13 (2/2 positive); hominy, 0.06 (1/1); masa 0.017 (1/1); popcorn (0.01 - 0.06 (6/8); puffed corn, 0.79-6.1 (6/6); torilllas 0.06-0.12 (2/4); tortilla chips 0.03 -0.32 (4/6). With respect to whole corn, fumonisin B1 occurred in 9 of 27 frozen corn samples (0.08 - 0.35 ppm), 33 of 73 canned (0.03-0.34 ppm), and 2 of 16 sweet corn samples (0.07 and 0.79 ppm). From the USDA 1978 survey data and the above concentrations, crude estimates of average daily exposures to fumonisin B1 were calculated to fall between 2 × 10-6 to 4 × 10-5 mg/kg-day. Subgroups consuming corn in areas subject to high contamination may be exposed to substantially greater levels.
  • Sterigmatocystin and ochratoxin A. Data on sterigmatocystin occurrence and exposure are difficult to locate. IARC (1976) notes that sterigmatocystin has been found as a natural contaminant of green coffee beans (1.1 ppm) and wheat (0.3 ppm), and that it can be identified in salami inoculated with Aspergillus versicolor. It can also be isolated from cultures of A. versicolor found on country hams. Calculation of exposure estimates thus awaits better and more complete data. Ochratoxin A is a contaminant of stored grain. It appears to be relatively uncommon in the U.S., because the winter storage climates are fairly cold (Miller 1994). However, occasional outbreaks occur. IARC (1993) notes that pork products, contaminated via feed grains, can be a significant human dietary source of ochratoxin A.


Inorganic metals and organic contaminants, such as polycyclic aromatic hydrocarbons (PAHs), can enter the food supply following uptake by plants and animals from the environment. Levels of some PAHs in vegetables decrease, for example, with increased distance from industrial centers and highways (Shibamato and Bjeldanes 1993). In fact, environmental contamination is seen as a major source of relatively high PAH levels observed in vegetable oil (Shibamoto and Bjeldanes 1993), fresh vegetables, meats, seafoods, oils, grains, and fruits (IARC 1983). Vegetation, especially that with a high lipid content, can be a major vehicle for removing PAHs from the atmosphere (Simonich and Hites 1994). Examples of benzo(a)pyrene levels due to uptake include: cereal (0.2-4 ppb), grain (0.7-2.3 ppb), flour (dried, 4 ppb), lettuce (2.8-12.8 ppb), margarine (0.9-36 ppb), coconut fat (0.9-43 ppb), and sunflower oil (different reports: 0.2; 5; 29-62 ppb).

Intentional Food Additives and Constituents of Spices

Intentional food additives are plotted in Figure 5-1 in decreasing order of annual per capita disappearance, which reflects usage in food. Disappearance exceeds actual human intake because of wastage and losses due, for example, to volatilization and leaching during processing, storage, distribution, and final preparation. The data were compiled in 1980 from NAS surveys, supplemented and checked against independently acquired information. Dietary patterns change only slowly; e.g., sucrose, the dominant caloric sweetener in 1980, has been replaced in part by low-fructose corn syrup, but this has not altered the overall pattern of ingredient usage.

Figure 5-1. Rank ordering of food ingredients by per capita annual disappearance.

Figure 5-1

Rank ordering of food ingredients by per capita annual disappearance.

In this figure (5-1), section 1 of the curve contains the caloric sweeteners, major functional ingredients such as acidifiers, but no spices or flavors. Section 2 contains pH-adjusting agents, processing aids, and a few major spices and flavors, such as vanilla extract, the synthetic flavor vanillin, and black pepper. Section 3 consists largely of minor spices and herbs, essential oils, and the major synthetic flavors. Section 4 of the curve, below a few milligrams per year, contains most of the flavors added to foods and a few of the smaller-volume, intentionally added micronutrients. The median intake of the entire group is approximately 1 mg per person per year.

Some substances, such as nitrate and nitrite, occur as natural components and as intentional additives. Nitrate is reduced endogenously to nitrite. Although there is no evidence that either nitrate or nitrite alone is carcinogenic, nitrite consumed with nitrosatable amines results in the endogenous formation of carcinogenic nitrosamines.

Dietary nitrate intake is estimated to vary from about 75 to 270 mg (NRC 1981). The extent to which nitrate is reduced endogenously to nitrite depends on gastric acidity and the nature and number of bacteria present. Dietary nitrite intakes are much lower than those of nitrate.

Vegetables are the primary source of nitrate and nitrite in food, although cured meat and dairy products also contribute to overall exposure. Concentrations of nitrate in vegetables depend on agricultural practices, the temperature and light in which they are grown, the concentrations of nitrate in the soil, fertilizers, and water used to grow the vegetables, and on storage conditions (NRC 1981).

The concentrations of nitrate and nitrite in cured-meat products depend on the curing process and on the amounts added as preservatives. Concentrations of nitrite in bacon, for example, can be as high as 120 ppm, the maximum allowed by law (9CFR 318.7B). Nitrate and nitrite are used as preservatives because of their ability to inhibit the growth of Clostridium botulinum (NRC 1981). However, improved manufacturing processes have led to a steady decline in the concentrations of nitrate and nitrite in preserved meats; in fact, nitrate is now used only rarely.

Dairy products contain low concentrations of nitrate and nitrite in general, rarely exceeding 5 mg/kg in milk (NRC 1981).

Approximately 100 spices and herbs are used for dietary purposes in the United States, most in very small quantities. A majority of these are available commercially and are regulated under the Food, Drug and Cosmetic Act. Some are grown in home herb gardens, and others are gathered wild. Teas made from herbs gathered by amateurs are a well-recognized cause of human poisonings. The carcinogenic properties of most of these spices have not been evaluated. Most are generally recognized as safe by the U.S. Food and Drug Administration.

The animal carcinogens so far identified in spices and flavors are listed with their plant sources and levels of occurrence in Appendix A. The appendix notes only that there has been some degree of carcinogenicity testing with some positive results reported. Among the plant constituents, those associated with spices include benzaldehyde, capsaicin, estragole, eugenol (also reported to be anticarcinogenic), and safrole. Constitutive substances reported to be inhibitors of carcinogenesis are listed in Appendix B. The comparative carcinogenic potency of constitutive and nonconstitutive substances is addressed later in this chapter.

Traditional Foods

The International Agency for Research on Cancer has identified more than 69 agents capable of causing cancer in humans, including the following whole foods and beverages: bracken fern (or fiddleheads), Chinese salted fish (Cantonese style), hot maté, betel quid (with tobacco leaves), and alcoholic beverages. The carcinogenicity of these agents was established through epidemiologic studies. With the exception of alcoholic beverages, risks of these agents are confined to relatively small groups in the United States. (For example, Chinese salted fish is used in some communities to wean infants.) However, alcoholic beverages are consumed by a relatively large subpopulation at moderate to high levels.

Recent advances in genetic engineering are expected to lead to the development of new technologically altered foods in the future. Very limited quantities of the first such technologically altered food, the Flavr/Savr™ tomato, were released for public consumption in 1994. The introduction of such novel foods may affect overall food consumption patterns by augmenting or displacing consumption of traditional foods.

Synthetic Carcinogens

Table 5-3 classifies synthetic carcinogens found in the diet into six categories: pesticide residues, potential animal drug residues, packaging or storage container migrants, residues from food processing, colors, and direct food additives. The chemicals listed under each category have been classified by IARC or NTP as carcinogens. In the remainder of this section, potential exposures to these categories of chemicals will be discussed.

Table 5-3. Synthetic Animal and Human Carcinogens That Might Be Present in the Diet.

Table 5-3

Synthetic Animal and Human Carcinogens That Might Be Present in the Diet.

Pesticide Residues in Foods

Pesticide residue data can be obtained from a variety of sources, including the FDA, state regulatory agencies, the food-processing industry, retail distributors, the agricultural chemical industry, and food commodity associations. Although all these sources of information are useful, no one source is necessarily preferable to another for purposes of assessing exposure. Residue analyses are complex, difficult to perform, and expensive. All data should be judged in this context.

Residue levels depend on several factors, the most important of which are the percentage of crop acreage treated with the pesticide, the sampling design of the survey, and the analytical limit of detection for the pesticide in that food. Other factors include the stability of the chemical; the time between pesticide application, harvest, and sampling; and the degree of post-harvest processing.

Processing may increase, decrease, or have no effect on the concentration levels of pesticide residues in foods. Washing the raw foods tends to reduce residues, blanching reduces them further, and canning reduces them even further (Elkins 1989). For example, malathion levels in tomatoes were reduced 99% when subjected to all three processes. However, ethylene thiourea (ETU) levels in frozen turnips increased 94.5% as a result of maneb degradation during cooking in a saucepan. Similarly, when the plant-growth regulator Alar was used in apple production, the concentration of unsymmetric dimethyl hydrazine was several-fold greater in apple juice and apple sauce than in fresh apples, because of the breakdown of Alar.

In 1993, the NRC Committee on Pesticides in the Diets of Infants and Children determined that FDA's market basket sampling and analysis provided the most comprehensive residue data at this time. More than 100 different pesticides were reviewed by the NRC using the FDA surveillance data. Pesticides were detected most frequently in fresh fruits and vegetables, such as apples, peaches, pears, bananas, peas, green beans, and carrots. Detectable levels were found in less than 10% of the samples for most crop-pesticide combinations with 2-year samples larger than 25. The percentage of positive detections ranged from 0.3% for captan on carrots and peas to 50% for benomyl on peaches. Of those residues that were detected, most were well below the EPA tolerance levels. Only six crop-pesticide combinations had maximum residues exceeding EPA tolerance levels, with all mean concentrations of detected residues well below these levels.

Veterinary Drug Residues

The U.S. FDA (1985) regulates the concentrations of veterinary drugs that are known to be carcinogenic. Specifically, it requires analytical methods capable of detecting drug residues in edible meat at concentrations in the human diet that have estimated lifetime cancer risks of less than 10-6. For some of the hormone residues, FDA prohibits detectable levels in edible tissues. Scheuplein (1990) estimates the total daily dose of these animal drug residues to be less than 100 ppb.

Packaging Materials

Humans may be exposed to trace quantities of chemicals that migrate into food from packaging materials. Regulations governing packaging components and their constituents are based primarily on the characteristics and uses of these materials. These characteristics and uses are very different from those of direct additives, and also from those of the indirect additives (e.g., pesticidal residues). The primary purpose of packaging is to protect the food it contains from air, moisture, light, contamination, attack by pests, tampering, quality loss, and physical damage. Packaging also can carry information and advertising, and it may aid in dispensing the food. Serving these functions requires materials that typically are insoluble, inert, and specifically intended not to enter the food or to affect it in any way other than to protect it. Moreover, the packaging regulations provide a list of thousands of options. There is no way of knowing which particular sets of options, employed in packaging each of many tens of thousands of foods, are in use across the entire food supply at any particular time. As a result, human exposures to such agents are difficult to estimate.

The most common restrictions on packaging components and their constituents are the following: good manufacturing practices (GMP); a provision that they may be used ''in an amount not to exceed that required to accomplish the desired technical effect"; and provisions that confine usage to particular technical effects (e.g., preservatives or emulsifiers). In a few instances, such as some monomers used in fabricating plastic resins, the regulation will limit the amount used or the level remaining in the packaging component or constituent to below a particular quantity. This limitation is usually based on extraction tests using specified methods. It is intended to ensure that when humans consume food that has been in contact with a component that might contain a questionable substance, such exposure cannot exceed a value that will provide a reasonable assurance of safety. The U.S. Food and Drug Administration has proposed to set a threshold for regulation to correspond to a toxicologically insignificant exposure level for food-contact substances, as described later in this chapter.

Residues from Food Processing

Carcinogenic residues can sometimes form as a result of processing. For example, cooking can increase the level of ethylene thiourea, a degradation product of maneb and related pesticides, and was observed to dramatically increase the levels of unsymmetric dimethyl hydrazine in apple products due to the degradation of Alar. Methylene chloride residues introduced during the production of decaffeinated coffee were the subject of considerable concern.

Direct Food Additives

A variety of additives are allowed into the final food product to assure its safety in the package. It is important to note that this vast array of substances are strictly regulated by the FDA. Among these additives are antioxidants.

One of these direct food additives is the antioxidant butylated hydroxyanisole (BHA), which is generally recognized as safe by the FDA at levels less than 0.02% of the fat content of food. It is also codified as a prior-sanctioned ingredient for use in food-packaging materials. Products that may contain BHA include breakfast cereals, potato flakes, poultry and meat products, sausage, shortenings, oils, food packaging materials, dessert and beverage mixes, glazed fruits, chewing gum, and flavoring agents. Based on the NRC (1979) report, Use and Intake of Food and Color Additives, estimates of mean BHA intakes for various age groups ranged from 0.12 to 0.35 mg/kg day. At the 90th percentile, intakes ranged from 0.27 to 0.76 mg/kg-day It should be noted that an ad hoc expert panel of the Federation of American Societies for Experimental Biology excluded the most recent NRC estimates of BHA intake data from consideration because it was felt that the estimates (which were about 25% the previous NRC values) were tempered by lack of survey compliance data.



There is a paucity of data on the amounts and kinds of dietary fiber in foods. Although data for crude fiber are available in food-composition tables, this is an inadequate indicator of dietary fiber because the method of analysis for crude fiber involves treatment of foods with acids and alkalies that destroy many of the components of dietary fiber (NRC 1989). The first surveys to include an estimate of dietary fiber were the 1985 and 1986 CSFIIs (USDA 1987). In these surveys, dietary fiber included the insoluble fraction (neutral detergent fiber) and soluble fraction (such as gums in cereal grains and pectin in fruits and vegetables). Foods highest in dietary fiber include whole (unrefined) grains and breads made from them, legumes, vegetables, fruits, nuts, and seeds. According to the 1985 CSFII Survey, average intake of dietary fiber per day for women 19 to 50 years of age was 10.9g; for children 1 to 5 years old, 9.8g (both based on 4 days of intake); and for men 19 to 50 years old, 18g (based on a 1-day intake).

The 1986 CSFII (USDA 1987) indicated that women in the west and midwest sections of the U.S. had higher intakes of dietary fiber than those in the south or northeast.


Several nutrients, including vitamins A, C, D, E, folic acid, calcium, selenium, and iron, have been extensively studied in cancer chemoprevention. Intake of these agents, particularly vitamins A, C, calcium, and iron, has been estimated in food consumption surveys. Data on the intake of vitamins D, E (alpha-tocopherol or alpha-TE), selenium, and folic acids in foods are less complete.

A previous NRC report (1989a) reviewed the intake of vitamins A, C, D, E, folic acid, calcium, and iron from major food sources in relation to the recommended daily allowances (RDAs) for these substances. The RDA of vitamin A ranges from 400 to 1,000 Image img00014.jpgg depending on age, with additional supplements recommended for pregnant and lactating women. Although this RDA is not achieved by all individuals, the 1985 CSFII indicated that the majority of the population appeared to consume the recommended levels. Intakes tended to be lower among low-income groups and higher in the western United States. RDAs for vitamin C range from 35 to 60 mg, again with supplements for pregnant and lactating women. Both the 1985 CSFII and the 1977 NFCS indicated average intakes of this vitamin exceeding the RDA, with intakes positively correlated with economic status. RDAs for vitamin D range from 5 to 10 Image img00014.jpgg; however, its intake is only estimated in national surveys monitoring food consumption, because little information is available on vitamin D in foods. The RDAs for vitamin E range from 3 to 4 mg for children to 11 mg for lactating women. Vitamin E intakes were first reported in the 1985 CSFII, with average intake levels near or above the RDA.

The RDAs for folic acid range from 35 to 40 Image img00014.jpgg for infants to 800 Image img00014.jpgg for pregnant women. Data from the 1985 CSFII indicate average intakes of 305 Image img00014.jpg/day and 189 Image img00014.jpgg/day for men and women 19 to 50 years of age, respectively. Results for folic acid are limited by the inherent variability in laboratory methods of analysis of folacin in foods, and by the high percentage of folacin concentrations in foods that were imputed rather than measured.

Calcium RDAs range from 360 mg to 1,200 mg in children 11 to 18 years of age and in pregnant or lactating women. Mean intakes of calcium were less than the RDA in most population subgroups, although men 19 to 50 years of age consumed an average of 115% of the RDA in the 1985 CSFII.

RDAs for iron range from 10 to 18 mg. While the mean intake among children 1 to 5 years old was 78% of the RDA, only 4% of women met or exceeded the RDA.

Estimates of selenium intake for the U.S. population range from 0.071 to 0.152 mg selenium/day (Schrauzer and White 1978, Welsh et al. 1981, FDA 1982, Levander 1987, Schubert et al. 1987, Pennington et al. 1989).

Non-Nutritive Constituents

Several investigations suggest that non-nutrient components of plants consumed in the diet contribute significantly to cancer prevention. Of particular interest are members of the flavonoid class, such as quercetin and kaempferol glycosides, which are widely distributed in foods. Others include isoflavonoids and phytoestrogens. However, unlike the nutrients, intakes of non-nutritive constituents have not been extensively studied. No quantitative exposure information is available on the anticarcinogenic non-nutritive constituents. Their numerous conjugated forms in plants have not been investigated, and estimates of exposure are further complicated because these constituents are hydrolyzed in the gut to other products.

Comparisons of Exposure Predictions for Naturally Occurring and Synthetic Carcinogens

Predicted exposure levels to some of the naturally occurring and synthetic carcinogens in the diet are provided in Tables 5-4 and 5-5. These estimates are subject to considerable uncertainty, and should be considered ballpark figures, accurate to at most an order of magnitude.

Table 5-4. Predicted Daily Intake Levels of Selected Naturally Occurring Carcinogens in the Diet.

Table 5-4

Predicted Daily Intake Levels of Selected Naturally Occurring Carcinogens in the Diet.

Table 5-5. Predicted Daily Exposure Levels of Selected Synthetic Carcinogens in the Diet.

Table 5-5

Predicted Daily Exposure Levels of Selected Synthetic Carcinogens in the Diet.

Measures of Carcinogenic Potency

To estimate cancer risks, information is required on the carcinogenic potency of the agent of interest, in addition to the level of exposure to the agent. Recent developments in measuring carcinogenic potency are related to the TD50 index introduced by Peto et al. (1984a) and Sawyer et al. (1984). Formally, the TD50 is defined as the dose that reduces the proportion of tumor-free animals by 50% at a specified point in time. Letting P(d) denote the probability of a tumor occurring at dose d, the TD50 is that dose d that satisfies the equation

Image img00002.jpg

where R(d) is the extra risk over background at dose d. Note that the TD50 is inversely related to potency, in the sense that the more potent the carcinogen, the lower the TD50. Thus, once the dose-response relationship P(d) has been determined, the TD50 may be estimated from equation (1).

More generally, the TD100p is the dose corresponding to an excess risk of 0<p<1. In this chapter, the TD01 will also be considered as the basis for comparing the potency of naturally occurring and synthetic carcinogens. The TD01 is closer to human exposure levels, yet still not sufficiently low that it depends strongly on the dose-response model chosen to represent P(d).

Sawyer et al. (1984) used a hazard function of the form

Image img00003.jpg

where 0(t) denotes the baseline hazard at time t in the absence of exposure. Use of this function leads to an essentially linear dose-response relationship

Image img00004.jpg

at a fixed time T, in the absence of mortality from causes other than tumor occurrence. Other hazard functions based on either the empirical or biologically based dose-response models could be used to accommodate nonlinear dose-response relationships (Krewski et al. 1992).

Ideally, the time of tumor occurrence would be used as the basis for statistical inference about the hazard function for tumor induction. Unfortunately, because most tumors in laboratory animals are unobservable before sacrifice, the tumor onset time is generally unknown. Sawyer et al. (1984) avoided this problem by using the time of death of those animals with tumors as a proxy for the time of tumor onset. Although this assumption is appropriate for tumors that are rapidly lethal, Portier and Hoel (1987) found that estimates of the TD50 are quite sensitive to assumptions about tumor lethality.

Finkelstein and Ryan (1987) addressed this problem by using the methods of Peto et al. (1980) for combining tumor mortality and prevalence data to estimate the TD50. Bailar and Portier (1993) used survival-adjusted quantal response data as described by Bailar and Portier (1988) to estimate the TD50. Dewanji et al. (1993) used Weibull models for the time of tumor onset, the time to death due to tumor occurrence, and the time to death from competing risks to estimate the TD50. This latter approach can be applied to rapidly lethal or incidental tumors, or to those of unknown lethality.

Gold et al. (1984, 1986, 1987, 1990, 1993, 1995) have tabulated the TD50 values for a large number of chemical carcinogens in their Carcinogenic Potency Database (CPDB). The TD50 values in the CPDB were calculated using the statistical methods developed by Sawyer et al. (1984). All TD50 values are expressed in common units of mg/kg body weight/day, and corrected for intercurrent mortality whenever individual animal survival times are available. The CPDB currently includes data on over 4,000 experiments on more than 1,000 different chemicals. The carcinogenic potency of these chemicals expressed in terms of the TD50 varies by more than 10 million fold.

Although the TD50 represents a useful measure of carcinogenic potency, it is based on experimental observation at doses generally well above those to which humans may be exposed. Another measure of potency, which has been used by the U.S. Environmental Protection Agency (1986a), is the estimate of the linear term q1 in the multistage model

Image img00005.jpg

where P(d) denotes the probability of tumor occurrence at dose d. Because the extra risk P(d) - P(O) is approximated by R(d) = q1d at low doses, the value of q1 may be used to estimate the risk associated with environmental exposures to a dose d of a carcinogen.

In practice, an upper confidence limit q1* on the value of q1 is used, because of the instability of the maximum likelihood estimate of the linear term in the multistage model (Crump 1984). This application is commonly referred to as the linearized multistage (LMS) model and represents an upper bound on risk. This procedure was used to estimate the TD01, later in this report, for comparing the carcinogenic potencies of naturally occurring and synthetic carcinogens.

Another approach for estimating low-dose risks is the model-free extrapolation (MFX) method proposed by Krewski et al. (1991). This procedure assumes that the dose-response curve is linear only at low doses and is based on a series of secant approximations to the low-dose region. Upper confidence limits on the slope of the dose-response curve based on MFX are generally close to the values of q1* obtained from the LMS model.

Although the preceding measures of carcinogenic potency are useful, they are subject to certain limitations. The TD50 is based on data obtained at high doses and may not reflect potency at low doses. Krewski et al. (1990) noted that the TD50 will exceed the highest dose used in animal cancer tests in some cases and thus require extrapolation outside the range of the experimental data. Because the TD50 can vary appreciably with the period of exposure (Dewanji et al. 1993), it is important that TD50s used for comparative purposes be standardized. When data on the dependency of cancer risk on age and exposure time are not available, standardization of TD50 values in the Carcinogenic Potency Database is done under the assumption that risk increases in proportion to time to the second power. Inferences about carcinogenic potency at low doses require extrapolation of animal bioassay data to low doses outside of the experimentally observable range. In the absence of information to the contrary, extrapolation is generally performed under the assumption that the dose-response relationship is linear at low doses.

To avoid low-dose and high-dose extrapolation, an intermediate measure of potency such as the TD05 or TD10 is preferred. The TD01 represents perhaps the lowest indicator of carcinogenic potency that remains near the experimentally observable response range and is used in the comparison of potencies of naturally occurring and synthetic carcinogens presented in a later section of this chapter. Van Ryzin (1980), Farmer et al. (1982), and Gaylor et al. (1994) used the TD01 as the basis for linear extrabolation to lower doses. Krewski et al. (1993) showed that the TD50s of rodent carcinogens, based on either the one-stage or multistage models, are tightly clustered around their MTDs. For the one-stage model, TD01 = TD50/69.

Correlation Between Cancer Potency and Other Measures of Toxicity

Several investigators have noted a strong correlation between the TD50 and the MTD (Bernstein et al. 1985, Gaylor, 1989, Krewski et al. 1989, Reith and Starr 1989, Freedman et al. 1993). This correlation (r=0.924) is illustrated in Figure 5-2 using a sample of 191 carcinogens selected from the CPDB (Krewski et al. 1993). Although the TD50 values used in this analysis are based on an essentially linear dose-response model, the correlation is equally high when it is obtained with TD 50s estimated using the Armitage-Doll multistage model. Bernstein et al. (1985) attributed this high correlation to the limited range of possible TD50 values relative to the MTD, and to the large variation among MTDs for different chemicals. Subsequent analytical results by Kodell et al. (1990) and Krewski et al. (1993) support this argument.

Figure 5-2. Association between carcinogenic potency and MTD.

Figure 5-2

Association between carcinogenic potency and MTD. Source: NRC 1993b.

Krewski et al. (1989) noted that the values of q1* derived from the linearized multistage model fitted to 263 data sets were also highly correlated with the maximum doses. As with the TD50, this association between q1* and the MTD occurs as a result of the limited range of values that q1* can assume once the MTD is established. This correlation is illustrated in Figure 5-3 using the same data presented in Figure 5-2. As indicated in Figure 5-3, there is a strong negative correlation between q1 and the MTD. Thus, the MTD has a strong influence on measures of carcinogenic potency at both high and low doses.

Figure 5-3. Association between upper bounds on low-dose slope and MTD.

Figure 5-3

Association between upper bounds on low-dose slope and MTD. Source: NRC 1993b.

The absence of points in the upper lefthand triangular region in Figure 5-2 is due to the lower limit on the number of tumors required for statistical significance observed in the exposed groups. This limit implies that highly toxic chemicals of weak carcinogenic potency would likely go undetected in a standard bioassay, because such agents would not yield a measurable excess of tumors at the MTD (NRC 1993b). Crouch et al. (1987) attribute the absence of points in the lower right hand triangular region (Figure 5-2) to a lack of chemicals with extremely high potencies relative to their MTDs. An informal search for such ''supercarcinogens" conducted by the National Research Council (1993b) failed to reveal any points clearly embedded in this region.

The relationship between acute toxicity and carcinogenic potency has been the subject of several investigations. Parodi et al. (1982) found a significant correlation (r = 0.49) between carcinogenic potency and acute toxicity. Zeise et al. (1982, 1984, 1986) reported a high correlation between acute toxicity, as measured by the LD50, and carcinogenic potency. Metzger et al. (1989) reported somewhat lower correlations (r = 0.6) between the LD50 and TD50 for 264 carcinogens selected from the CPDB. McGregor (1992) calculated the correlation between the TD50 and LD50 for different classes of carcinogens considered by IARC. The highest correlations were observed in IARC Group 1 carcinogens (i.e., known human carcinogens) with r = 0.72 for mice and r = 0.91 for rats, based on samples of size 9 and 8, respectively. Goodman and Wilson (1992) calculated the correlation between the TD50 and LD50 for 217 chemicals that they classified as being either genotoxic or nongenotoxic. The correlation coefficient for genotoxic chemicals was approximately r = 0.4 regardless of whether rats or mice were used, whereas the correlation coefficient for nongenotoxic chemicals was approximately r = 0.7. Haseman and Seilkop (1992) showed that chemicals with low MTDs (i.e., high toxicity) were somewhat more likely to be rodent carcinogens that chemicals with high MTDs, but this association was limited primarily to gavage studies.

Travis et al. (1990a,b, 1991) investigated the correlation between the TD50 and composite indices based on mutation, toxicity, reproductive anomalies, and tumorigenicity data derived from the Registry of Toxic Effects of Chemical Substances (RTECS) (Sweet, 1987). This analysis confirmed the previously reported correlation of r = 0.4 between mutagenic potency in the Ames assay and the TD50 estimated from rodent carcinogenicity studies (McCann et al. 1988, Piegorsch and Hoel 1988); the correlation of r = 0.7 between LD50 and TD50 was also somewhat greater than that reported by Metzger et al. (1989). Using all of the RTECS assays, the correlation of the composite potency index with the TD50 was r = 0.80, 0.87, or 0.79, depending on whether data for rats, mice, or the most sensitive species were used.

Hoel et al. (1988) explored the relationships between toxicity and carcinogenicity for 99 chronic studies conducted in rodents. Only seven of the 53 carcinogenicity studies reporting positive results exhibited target organ toxicity that could have been the cause of observed carcinogenic effects. Those findings suggested that only a small number of chemical carcinogens that induce tumors in rodents solely by indirect mechanisms may be limited.

Tennant et al. (1991) examined 31 chemicals, of which 22 were classified as rodent carcinogens. Regardless of their mutagenic potential, chronic toxicity was not always sufficient to produce neoplasia. This suggests that effects other than mutagenicity or toxicity may be responsible for the carcinogenicity of some chemicals.

Gaylor and Gold (1995) used the correlation between the MTD and cancer potency to obtain preliminary estimates of cancer risk based on the 90-day MTD and not on 2-year bioassays in rodents. Using results from studies conducted by the National Toxicology Program, estimates of the regulatory virtually safe dose (VSD), corresponding to an estimated maximum lifetime cancer risk of less than one per million, were obtained by linear low-dose extrapolation for 139 rodent carcinogens. Based on these cases, the geometric mean ratio of the MTD to the VSD was 740,000. Of the 139 cases examined, only five ratios differed from the mean ratio by more than a factor of 10. Here, the VSD is roughly a small fraction of the rodent test doses. A similar result can be obtained using the acute toxicity measure the LD50, compared with a dose corresponding to an estimated risk of less than one per million (Zeise et al. 1984). Hence, it may not be worthwhile to conduct a chronic bioassay when levels of human exposure are so low as to suggest negligible risk (see the section below on toxicologically insignificant exposures).

Interpretation of Carcinogenic Potency

Little information is currently available on the relative potencies of various carcinogens and anticarcinogens in the human diet. Although reasonable estimates may be determined for a few substances, such as aflatoxin, most of the data are sparse. Although precise potency data for humans on each substance of interest would be ideal, an absolute measure is not essential for comparative purposes, and a relative measure could serve.

Nondietary exposures may influence the degree of toxicity of dietary carcinogens and anticarcinogens. For example, cigarette smoking induces aryl hydrocarbon hydroxylase, one of the cytochrome P450 isozymes, which in turn may affect the metabolism of substances in the diet.

As noted earlier in the section, most information on the potencies of carcinogens in humans is inferred from animal experiments. In fact, the U.S. EPA Guidelines for Carcinogen Risk Assessment (1986a) specify that if there are adequate data demonstrating an agent's carcinogenicity in laboratory animals, it should be regulated as a human carcinogen. However, the issue of extrapolating the results of animal bioassays to humans for regulatory purposes is controversial. The current animal bioassay was designed as a qualitative screen for carcinogenicity; it was not designed to provide information on an agent's biologic mechanisms of action or on its dose-response characteristics at low doses. Most cancer bioassays use the MTD, which is generally much higher than environmental exposure levels likely to be encountered by humans. The MTD is used in cancer studies to increase the likelihood of detecting any potential carcinogenic effects of a chemical in experiments involving small numbers of laboratory animals. The problem with using such high doses is that they may cause other effects that could influence the likelihood of tumor formation; for example, high doses can lead to changes in food consumption, recurrent cytotoxicity in specific organs, hormonal imbalance, or combinations of these and other effects (NRC 1993b). Such effects have been associated with both increases and decreases in tumor incidence in laboratory animals (Reitz et al. 1980, 1990; Turnbull et al. 1985; Roe 1988). Thus, dose-response data from high-dose studies can be of limited value for low-dose risk prediction (NRC 1993b). In addition, testing at high doses that result in lower body weights may reduce the incidence of some types of tumors (Turturro et al. 1993).

When quantitative estimates of human cancer risk must be made on the basis of animal data, an interspecies scaling factor may be employed. Gaylor and Chen (1986) have shown that estimates of cancer potency in rats and mice are correlated, although potencies can differ by a factor of 100-fold or more in some cases. Although quantitative interspecies extrapolation has been done on the basis of body weight or surface area in the past, Travis and White (1988) have suggested that an intermediate scaling based on body weight to the 3/4 power may be more appropriate. Watanabe et al. (1992) note that the available data do not permit one to distinguish between surface area or 3/4 scaling and that different chemicals may warrant different scaling procedures. Physiologic pharmacokinetic models offer a more sophisticated approach to interspecies extrapolation through the use of predicted tissue doses in the species of interest (NRC 1987). However, whenever human risks are inferred from animal test data, considerable uncertainty may remain about the magnitude of the risk.

Estimating Human Cancer Risks

Risk-Estimation Methods

Estimates of human cancer risks are based primarily on findings from human epidemiologic studies or results from laboratory tests involving experimental animals. Epidemiologic studies have the advantage of providing information on carcinogenic risks directly in humans. However, the results of such studies may be subject to biases due to confounding factors or errors in exposure and disease ascertainment. Nonetheless, these studies have been used to identify 69 agents capable of causing cancer in humans. Of these, only a few (including salted fish, alcoholic beverages, and aflatoxin B1) are present in the U.S. diet. Epidemiologic studies may not be sufficiently sensitive to detect the hazards to humans from exposures to low levels of carcinogens in the environment. In contrast, toxicologic studies can achieve greater sensitivity through the use of high doses and because they can be designed to control for potential sources of bias. These gains are offset to a certain extent, however, when extrapolating from high to low doses and from animals to humans. Nonetheless, about 600 chemicals have been shown to demonstrate carcinogenic potential in animal cancer tests (Gold et al. 1995).

Quantitative predictions of human cancer risk are generally obtained by fitting a suitable dose-response model to either toxicologic or epidemiologic data. Many epidemiologic studies provide little information on dose-response relationships, making it difficult to estimate risks at exposure levels different from those observed in the original study. When dose-response data are available, a linear model is often sufficient to describe the observed dose-response relationship.

Toxicologic experiments involving only two or three relatively high doses also provide limited information on dose-response. Important questions have been raised about the relevance to humans of results observed at the high doses used in laboratory studies. Ames and Gold (1990a) argue that cancer risks may be increased at high doses through the induction of mitogenic effects that would not be expected to occur at low doses. After reviewing the use of the MTD in animal cancer tests, a committee of the National Research Council (1993b) concluded that results obtained at the MTD need to be interpreted with care, although a majority of that committee saw no clear alternative at the time.

Attempts have been made to strengthen cancer risk-assessment methods through the use of biologically based risk assessment models (Goddard and Krewski 1995). Biologically based models are appealing because the model parameters have a direct biological interpretation. Such models can also lead to a greater understanding of cancer mechanisms when they fail to fit the data. In such cases, different mechanistic assumptions must be made to obtain a parsimonious description of the data.

As noted previously, the linearized multistage procedure does not enable one to account for tissue growth and cell kinetics. These features can be incorporated into multistage modeling at the expense of achieving a somewhat over-parametrized model. For this reason, simpler biologically based models, such as the two-stage clonal expansion model described by Moolgavkar and Luebeck (1990) are of value. This model is based on the assumption that two mutations are required to transform a normal cell into a cancer cell; provision is made both for natural growth of normal tissues, and for selective clonal expansion of initiated cells that have sustained the first mutation. This model has recently been used in a biologically based reanalysis of data on lung cancer in Colorado uranium miners exposed to radon and cigarette smoke (Moolgavkar et al. 1993). Other biologically based models incorporating data on metabolic activation (Krewski et al. 1994) and receptor binding (Kohn and Portier 1993) have been proposed in an attempt to build a comprehensive pharmacokinetic/pharmacodynamic risk assessment model (Andersen et al. 1993).

A clear understanding of the process of carcinogenesis and considerable data are required if the results of biologically based approaches are to be viewed with confidence. Lacking that data necessary to develop biologically based models, empirical approaches to cancer risk estimation are often invoked. If the dose-response curve is assumed to be linear in the low-dose region, estimates of human cancer risk at low doses can be obtained using the linearized multistage model described by Crump (1984), or the model-free approach to linear low-dose extrapolation discussed by Krewski et al. (1991). In practice, these two methods yield similar results. Linear extrapolation from the TD50 also yields similar predictions when estimating cancer risk at low doses (Krewski 1990).

Uncertainty Analysis

In the past, methods for quantitative cancer risk assessment focused on a single estimate of risk—either a best estimate or an upper-bound estimate derived from the most likely model of carcinogenesis and for the expected exposure conditions. However, such estimates are subject to variability and uncertainty. For example, both food consumption patterns and concentrations of carcinogenic substances present in the diet can vary widely, leading to appreciable variation in individual intake of dietary chemicals. Uncertainty about the mechanisms of carcinogenic action and the relative sensitivity of animals and humans leads to uncertainty about the risk associated with a given level of exposure.

Recently, attempts have been made to evaluate both variability and uncertainty in cancer risk assessment. Pesticides in the Diets of Infants and Children (NRC 1993a) proposed methods to describe the variation in food-consumption levels and in pesticide residues found in foods. This analysis was accomplished by combining consumption and residue distributions to arrive at a single intake distribution of pesticide residue in the diet, reflecting both sources of variation.

Using existing databases, Gaylor et al. (1993) examined the uncertainties in cancer potency estimates resulting from experiment, strain, and species variability. They concluded that estimates of potency for human carcinogens, based upon animal data, are likely to be within a factor of 110 for most cases. Cancer rates in humans are commonly estimated by basing cancer risk estimates on the most sensitive rodent species, strain, and sex, and using an interspecies dose scaling factor based on body surface area. However, this practice appears to overestimate cancer rates in humans by about an order of magnitude (Gaylor et al. 1993). Hence, for chemicals where the dose-response relationship is nearly linear at doses below the experimental range, cancer risk estimates based on animal data are not necessarily conservative. Sources of underestimation not fully addressed by these comparisons include saturable pharmacokinetics (as seen, for example, with vinyl chloride and tetrachloroethylene (Bois et al. 1994)); less-than-lifetime animal bioassay (e.g., diethylnitrosamine (Peto et al. 1984b); human occupational data (e.g., benzidine (IARC 1982)); and human heterogeneity (e.g., 4-aminobiphenyl (Bois et al. 1995)). For a more complete discussion of underestimation, see Portier and Kaplan (1989), Finkel (1995), and Zeise (1989).

Portier and Kaplan (1989) analyzed the uncertainty surrounding the values of model parameters and its impact on model-based predictions of metabolites reaching target tissues. This analysis was conducted on model parameters in physiologically based pharmacokinetic models for methylene chloride. They demonstrated that uncertainty surrounding the values of the model parameters implied considerable uncertainty in tissue doses of the proximate carcinogen. Similar analyses have been conducted by Farrar et al. (1989), Bois et al. (1990), and Hetrick et al. (1991). Methods for uncertainty analysis are described in a recent text by Morgan and Henrion (1990).

Krewski et al. (1995) proposed a general methodology for analyzing uncertainty, variability, and sensitivity in physiologically based pharmacokinetic models. This approach takes into account a number of natural constraints on the model parameters, such as the upper limit on the sum of tissue volumes imposed by total body volume. Using this approach, it is possible to evaluate the uncertainty in model-based predictions of tissue dose attributable to the uncertainty associated with individual model parameters, as well as the total uncertainty conferred by all model parameters. The impact of interindividual variability in model parameters is assessed, including physiological parameters that are related to body weight, which varies considerably among individuals. Methods are also proposed for identifying sensitive parameters that play a critical role in model-based predictions concerning metabolites. These analyses may guide efforts to improve the reliability of estimates derived for the most critical model parameters.

Finkel (1995) analyzed the uncertainties associated with the cancer risks of aflatoxin and alar (a synthetic organic growth regulator formerly used on apples). Although single estimates of risk suggested aflatoxin to be 18 times more potent than Alar, the risks of these two substances could not be distinguished when all of the relevant uncertainties were taken into account.

Mechanistic Considerations

Methods currently used to calculate cancer potencies generally do not incorporate mechanistic considerations. The methods are based on the assumption that if a chemical increases tumor rates at high doses in animal bioassays, it may increase cancer risk among human populations exposed to lower levels found in the environment or the diet. This is an important consideration because chemicals that increase tumor rates at high doses by increasing cell proliferation as a result of tissue damage may not be carcinogenic at low doses. Some chemicals in the diet that test positive in cancer bioassays using high doses may not necessarily be carcinogenic at doses likely to be encountered by humans in the environment. Therefore, studies of the carcinogenicity of some chemicals at high doses are of limited value in assessing the dose-response relationships occurring at low exposure levels that are likely to be encountered in the human diet (Cohen and Ellwein 1990, Cohen et al. 1991).

In addition, some chemicals induce cancer in rodents by mechanisms that are likely not to be operative in humans (Cohen and Ellwein 1990, Boorman et al. 1994). An example is the induction of urinary bladder cancer in male rats as a consequence of microcrystal formation after ingestion of high doses of sodium salts. Although these chemicals demonstrate carcinogenic activity in rats, they may pose little or no carcinogenic risk to humans.

The International Agency for Research on Cancer (IARC 1992) has provided guidelines on the use of mechanistic information in carcinogenic risk assessment. These guidelines allow for the possibility that certain animal carcinogens are unlikely to be human carcinogens, and are hence considered unclassifiable (Group 3) with respect to human carcinogenicity. The guidelines also allow for the classification of an agent as a human carcinogen (Group I) in the absence of positive epidemiologic evidence. Such classification could occur in the presence of strong animal evidence for carcinogenicity and molecular evidence that the agent is likely to be effective in humans.

Toxicologically Insignificant Exposure Levels

The concept of toxicologically insignificant exposures (TIE) in the regulation of certain chemical substances found in food is currently receiving considerable attention (Rulis 1986, 1989, 1992; FDA 1993), particularly indirect food additives that migrate into food from packaging materials. Under the assumption that a new chemical in a particular class is no more toxic than the most toxic chemicals previously in that class, an upper bound on the potency of the new substance can be inferred. In turn, this upper bound on potency can be used to establish a toxicologically insignificant exposure for that chemical, below which only trivial risks are assumed to arise. This concept is discussed in more detail by Munro (1990).

The concept of toxicological insignificance has advantages and disadvantages in safety assessment. Because inferences on the potential toxicity of the substance in question are made in relation to the toxicity of other chemicals, the need to perform extensive toxicity studies is eliminated. However, the assumption that a new chemical is as potent as the more potent substances in the selected reference class of chemicals may impart a high degree of conservatism in evaluating its safety, especially when this assumption is applied to all new chemicals. In addition, the possibility that the new chemical of interest may be more potent than any in the reference class cannot be ruled out.

After analyzing approximately 220 chronic toxicity studies collected from the literature, Frawley (1967) proposed a toxicologically insignificant level of exposure of 0.1 ppm for food additives and other ingredients of the human diet. Since then, several hundred chemicals have been tested for tumorigenicity in rodents. With the increasing emphasis on tumorigenicity as an endpoint in chemical regulation, and the recognition that tumorigens are more stringently regulated than other chemicals, additional studies have examined the possibility of establishing a TIE level for potentially tumorigenic indirect additives. One such attempt is that of Rulis (1992), who proposed a threshold of regulation for indirect additives (packaging materials) of 500 ppt, based on his analysis of the Carcinogenic Potency Database described earlier. However, contamination by some agents with moderately high carcinogenic activity could be a concern at this level.

Risks of Joint Exposures and Mixtures

Risk assessment methods for joint exposures and mixtures have been examined in detail by the National Research Council (1988) and the International Agency for Research on Cancer (Vainio et al. 1990). The effects of a specific dietary constitutent may be influenced by the concentrations of other carcinogens or anticarcinogens in the diet. This fact complicates the task of estimating cancer risk for a complex mixture such as food. That is, substances may not act independently, and dietary constituents may interact. Examples of such interaction are provided by the recent studies by the National Toxicology Program (NTP 1995) designed to evaluate the effect of diet restriction on chemical-induced carcinogenicity. Chronic exposure studies conducted by the NTP usually involve administration of the test chemical to rodents maintained on the basal diet ad libitum for a period of about two years. Animals maintained under these conditions typically develop obesity; in contrast, diet restriction results in leaner animals that live longer.

In the series of studies conducted by the NTP, groups of 50-60 rats and mice per dosage group were fed diet either ad libitum or in amounts that restricted mean body weights to approximately 85% of the ad libitum bodyweight. After adjusting for intercurrent mortality, the effect of butyl benzyl phthalate on the incidence of pancreatic tumors was substantially higher in animals allowed to feed ad libitum and killed after 2 years, as compared with those on restricted feed. Specifically, pancreatic tumors were not observed in animals exposed to the compound but on restricted feed and killed after 2 years, and were only observed in exposed animals on restricted feed and killed at 30 months. Salicylazosulfapyridine caused an increased incidence of liver neoplasms in male mice fed ad libitum relative to the ad libitum-fed and weight-matched controls. This increase did not occur in the restricted feed protocols; in fact, decreases in combined hepatocellular carcinomas and adenomas were observed in salicylazosulfapyridine-exposed animals on restricted feed and sacrificed after 2 years, and decreases in hepatocellular carcinomas in those sacrificed after 3 years.

A further example is the interaction of aflatoxin and hepatitis B in the induction of liver cancer, as observed in epidemiologic investigations of humans exposed to both agents. Liver cancer incidence rises more rapidly for those individuals exposed to aflatoxin and testing positive for the hepatitis B virus than for those exposed to aflatoxin but testing negative for hepatitis. Cancer potency estimates differ by approximately a factor of 10 for the two groups (Wu-Williams et al. 1992). These findings are consistent with results from molecular studies (Bressae et al. 1991, Hsu et al. 1991).

Vainio et al. (1990) note that interactions among chemicals may occur during absorption, distribution, metabolism, and excretion. Compounds may interact chemically to create new toxic components or to change the bioavailability of mixture components. The matrix in which mixture components exist, such as soil, also may alter the carcinogenicity of a mixture. Evidence from animal and human studies indicates that interactive effects may cause deviations from simple additivity of risks, although these deviations are usually much less than one order of magnitude. Vainio et al. (1990) conclude that the most relevant quantitative data for making risk estimates for complex mixtures comes from epidemiologic studies of populations exposed to the complete mixture, not from studies of individual components.

Similarly, the EPA Guidelines for the Health Risk Assessment of Chemical Mixtures (1986b) state that the carcinogenic effects of a mixture can best be examined by testing the mixture. In the absence of such information, test results from a similar mixture may be adequate. Estimates of carcinogenic potencies for mixtures are not usually available, whereas estimates of potencies for the primary components of a mixture may be obtained in a number of cases. Thus, the EPA guidelines further state that if estimates of the effects of interactions among the components are available, the cancer risk for the mixture can be predicted. If it can be assumed that such interactions are negligible, the cancer risk for the mixture is generally estimated from the sum of the risks of the individual components. This practice does not imply that response additivity should be considered a universal phenomenon that occurs without exception, but that it should be a default position in the absence of evidence to the contrary (Krewiski and Thomas 1992). Where a series of related chemicals causes cancer by the same mechanism, and the relative potency of such chemicals is known, estimates of cancer risk associated with a mixture of such chemicals may be based on the sum of the effective doses of each component.

The presence of a chemical in a mixture may increase or decrease the carcinogenic potency of other components through the induction of detoxification enzymes or DNA repair, competition for receptors, or saturation of metabolic pathways. Experimental studies involving simultaneous exposure to high doses of two carcinogens have demonstrated both synergistic and antagonistic effects with respect to cancer risk. Interaction among components at low doses, on the other hand, may be minimal, due in part to the lack of competition for reactive agents. However, it is possible that two innocuous chemicals can combine to form a toxic compound, even at low doses.

Increased cell proliferation provides an increased opportunity for genetic events that lead to cancer. The concentration of an individual component of a mixture may not be adequate to cause mitogenesis by itself. However, the cumulative concentration of several components that increase mitogenesis by the same process may result in an increase in cell division. Hence, the cancer risk for a mixture could exceed the sum of the cancer risks of its components.

The occurrence of spontaneous tumors indicates that there are sufficient background levels of exogenous carcinogens, endogenous factors, and spontaneous tumorigenic events to produce tumors. The addition of a chemical that augments the background carcinogenic process will increase cancer risk to an extent that is proportional (linearly related) to the chemical dose at low doses. In this case, the use of a response-additive model is inappropriate, and a dose-additive model should be used (NRC 1989b). Krewski et al. (1995) reviewed this issue of additivity to background doses and low-dose linearity of risk. As the number of carcinogenic components increases in a mixture, the chance of augmenting a carcinogenic process increases; this process, in turn, increases the probability that the relationship between cancer risk and dose of the mixture is linear at low doses. Where the dose-response relationship is nearly linear at low doses, a response-additive and a dose-additive approach will give essentially the same result (NRC 1989b).

Apart from any chemical or biological interactions, the cancer risk from two chemicals acting on the same stage of a multistage process is the sum of the risks of the individual chemicals (Gibb and Chen 1986). Continuous, lifetime exposure of two chemicals acting on different stages of a multistage process results in multiplicative relative risks (Gibb and Chen 1986). For less than lifetime exposure, Brown and Chu (1989) show that the effect of two chemicals acting on the first and penultimate stages of a multistage process depends upon the age of the exposed animal and duration of the exposure. Nearly additive relative risks result when the two exposure periods occur simultaneously and are of short duration, while multiplicative relative risks result from lifetime exposure.

Kodell et al. (1991) describe the nature of interactions possible under the commonly used (approximate) form of the Moolgavkar-Venzon-Knudson two-stage clonal expansion model of carcinogenesis. In this analysis, interaction among initiators (agents that increase the first-stage mutation rate), promoters (agents that increase the rate of proliferation of initiated cells that have sustained the first mutation), and completers (agents that increase the second-stage mutation rate) was evaluated in terms of age-specific relative risk. It was found that joint exposure to two initiators or to two completers resulted in additivity of the age-specific relative risks, whereas joint exposure to an initiator and a completer or to an initiator and a promoter led to a multiplicative relative risk relationship. Supramultiplicative relative risk can occur with simultaneous exposure to two promoters. These results assume that there are no chemical or biological interactions between the two agents involved.

Kodell and Gaylor (1989) discuss the difficulties of distinguishing between additive and multiplicative relative risks for carcinogens. For the small, relative risks that would be expected at low doses, estimates of cancer risk for a mixture are approximately additive (NRC 1988). This is readily seen by a simple example. Suppose the relative risks for three components of a mixture are 1.01, 1.02, and 1.03. Using a multiplicative relative risk model, the relative risk for the mixture is 1.01 × 1.02 × 1.03 = 1.061. The additional risk of 0.061 is approximately the sum of the additional risk from each component 0.01 + 0.02 + 0.03 = 0.060. For most low-dose situations, there is no need to distinguish between additive and multiplicative risks. Thus, at low doses, the risk for a mixture generally can be approximated by the sum of the risks of the components.

For chemicals that can be grouped according to structural properties, such as polycyclic aromatic hydrocarbons, dibenzo-p-dioxins, and dibenzofurans, the combined carcinogenicity of a mixture can be estimated on the basis of a potency-equivalence approach based on a representative chemical of the class (NRC 1989b).

The combined effect of eating fruits and vegetables results in a reduction in some types of cancer. Both carcinogens and anticarcinogens may occur in a single food (e.g., citrus fruits contain vitamin C, which has anticarcinogenic properties, but also quercetin, a mutagen and possible carcinogen), and certainly both types of substances can be consumed in a single meal (NRC 1989b). Accordingly, individuals are exposed not only to mixtures of carcinogens, but also to mixtures of carcinogens and anticarcinogens. The net effect of such complex exposures is difficult to predict and is best determined empirically. Thus, although fruits and vegetables contain substances that may be carcinogenic in humans (Stolz et al. 1984), they also contain protective substances, and epidemiologic data strongly suggest that the risk of cancer is significantly reduced among individuals who consume higher amounts of these foods (Steinmetz and Potter 1991). It is not known whether the net beneficial effect of these foods results from interactions between the carcinogens and the anticarcinogens, or whether the combined beneficial effect simply exceeds the carcinogenic effect. However, from a public health perspective, people are encouraged to consume these foods because of the anticarcinogenic effects, which outweigh any carcinogenic risks posed by naturally occurring constituents.

Dietary Cancer Risks

Overall Impact of Diet on Cancer

The NRC Committee on Diet and Health (NRC 1989a) systematically examined the evidence relating dietary components to the occurrence of cancer. That committee found suggestive evidence linking specific cancers to the diet (e.g., alcoholic beverage intake and increases in esophageal and colorectal cancers), but the committee did not estimate the overall effect of diet on cancer mortality and incidence rates. Doll and Peto (1981) gave as a best estimate that 35% of all cancer mortality reported in the U.S. is related to diet, with an uncertainty range of 10-70%. Doll (1992) recently asserted that ''the estimate that the risk of fatal cancer might be reducible by dietary modification by 35 percent remains a reasonable estimate." In discussing these and other estimates, the Committee on Diet and Health noted that "because few relationships between specific dietary components and cancer risk are well-established, it is not possible to quantify the contribution of diet to individual cancers (and thus to total cancer rates) more precisely." Similarly, the earlier NRC Committee on Diet, Nutrition, and Cancer (1982) found that cancers of most major sites are influenced by dietary patterns, but concluded that "the data are not sufficient to quantitate the contribution of diet to the overall cancer risk or to determine the percent reduction in risk that might be achieved by dietary modification."

After reviewing all the evidence available to date, Ames et al. (1995) conclude that 1) reduction of smoking, 2) avoidance of intense sun exposure and high levels of alcohol consumption, 3) control of infections, and 4) increased consumption of fruits and vegetables, as well as increases in physical activity are likely to reduce the risks of specific cancers. The authors also suggest that reduced consumption of red meat may decrease the incidence of colon and prostate cancer and that modification of sex hormone levels might have some effect on breast cancer incidence.

Impact of Dietary Constituents on Human Cancer

Although only a few individual carcinogens in the diet have been identified through epidemiologic studies (e.g., aflatoxin and arsenic), these studies have identified dietary components as either contributing to or reducing cancer risk. Carcinogenic and anticarcinogenic dietary components strongly supported by epidemiologic studies at the time of the NRC Committee's report on Diet and Health in 1989, are shown in Table 5-6.

Table 5-6. NRC Committee on Diet and Health: Findings on Diet and Cancer.

Table 5-6

NRC Committee on Diet and Health: Findings on Diet and Cancer.

Many commonly consumed foods also contain chemical substances that have both carcinogenic and anticarcinogenic potential. Cooked beef steak contains a number of derived pyrolysis products such as benz(a)anthracene that have been shown to be carcinogenic in animals (cf. IARC 3:45, 32:135). Heterocyclic amines formed during cooking of beef steak, such as 2-amino-3-methylimidazo[4,5-f]quinoline (IQ), are also in low concentrations ranging up to 20 ng/g (IARC 40:261, 56:165). Beef steak may also contain traces of naturally occurring carcinogens acquired from the environment, including arsenic and ochratoxin A, a mycotoxin resulting from feed contamination (IARC 10:191, 31:191, 42:262, 56:489). Synthetic carcinogens such as 2,3,7,8-tetrachlorodibenzo-p-dioxin and other chlorinated polycylic hydrocarbons may also be present in beef steak as a consequence of contamination of packaging materials (IARC 15:41). In addition to trace levels of these carcinogenic substances, beef steak contains anticarcinogenic substances such as conjugated linoleic acid and selenium.

Many other foods contain carcinogenic and anticarcinogenic constituents. For example, broccoli may contain trace levels of arsenic, a widely distributed natural contaminant. Broccoli also contains chlorogenic and neochlorogenic acids; both are metabolized to caffeic acid, which is found in broccoli at concentrations up to 10 mg/kg. Anticarcinogenic substances found in broccoli include ascorbic acid, indole and other isothiocyanates, and sulforaphane.

Assessing dietary cancer risks requires consideration of both these types of agents, carcinogenic and anticarcinogenic. Ames et al. (1993b) have suggested that antioxidants such as ascorbate, tocopherol, and carotenoids found in fruits and vegetables may reduce cancer risk. Calorie restriction may also be effective in reducing cancer risk (Pariza and Boutwell 1987, Youngman et al. 1992).

Role of Calories and Fat

Excess Calories

Although animal models have shown effects of caloric intake on tumor incidence, especially a protective effect of reduced energy intake (Pariza and Boutwell 1987), in human beings the relationship of caloric intake to cancer risk has not yet been clarified. Because caloric intake and the intake of macronutrients (especially fat) are highly correlated in individuals, an effect of calories needs to be distinguished from one of fat or other macronutrients. Unfortunately, many early epidemiologic studies of diet and cancer did not assess total caloric intake, and others did not adjust for calories when examining the effects of fat or other macronutrients (NRC 1989a). Moreover, recent studies that examined total energy intake have not been consistent in their findings. For example, studies have shown both positive associations (West et al. 1989) and no association (Goldbohm et al. 1994) for total energy intake and colon cancer. Others have found independent effects of fat (or other macronutrients) on cancer risk after adjusting for caloric intake (Van't Veer et al. 1990, Willett et al. 1990, Giovannucci et al. 1993).

Epidemiologic evidence suggests that energy expenditure is also related to cancer risk, particularly inverse associations between physical activity and cancers of the colon, breast, and prostate (Peters et al. 1989, Whittemore et al. 1990, Lee et al. 1991, Lee et al. 1992, Friedenreich and Rohan 1995). The association of both caloric intake and physical activity with cancer of the colon suggests that net energy balance may be an important determinant of risk for some cancers. Moreover, positive associations have been found between obesity and the risk of cancer of the endometrium, colon, and breast (Henderson et al. 1983, Hunter and Willett 1993, Potter et al. 1993). Although the determinants of obesity are not fully understood (Berry et al. 1987), this association offers additional support to the hypothesis that net energy balance is related to cancer risk. In general, epidemiologic studies suggest that high caloric intake, low levels of physical activity, and obesity cause a moderate (i.e., 2- or 3-fold) increase in cancer risk.

Estimates of per capita caloric intake in the U.S. show relatively little change over the period 1909-1985, because increases in fat intake were largely offset by decreases in carbohydrate levels, while protein intake remained relatively constant (NRC 1989a). Data from food consumption surveys in recent years show variation in energy intake between sexes and among age groups. For example, in the 1977-1978 Nationwide Food Consumption Survey (see earlier section of this chapter on Sources of Information), men 75 years and older averaged 1,866 kcal per day, whereas women in the same age group consumed 1,417 kcal per day. Men in the age group 15-18 years of age, in contrast, consumed an average of 2,568 kcal per day.


Many epidemiologic studies have found a positive association between dietary fat intake and cancer at certain sites, such as the colon, prostate, and breast (NRC 1989a). Ecological studies, for example, have shown a positive relationship between per capita fat intake level by country and corresponding incidence or mortality rates for several cancers. In some instances, the associations were stronger for animal fat than for vegetable fat consumption (Rose et al. 1986, Prentice and Shephard 1990).

The findings from these ecological analyses have not been confirmed in analytical epidemiologic studies based on data collected directly from individuals. Thus, whereas a combined analysis of 12 case-control studies found a weak positive association between fat intake and breast cancer risk (Howe et al. 1990), 6 large cohort studies found no significant elevation in the risk of breast cancer among women with the highest levels of fat intake (Hunter and Willett 1993). Epidemiologic data on diet and prostate cancer are less abundant but suggest overall a positive association between dietary fat, especially animal or saturated fat, and risk of this cancer (Nomura and Kolonel 1991). Colon cancer has also been associated with saturated or animal fat, in both case-control and cohort studies (Whittemore et al. 1990, Willett et al. 1990). Cancers at other sites, such as the ovary and lung (Goodman et al. 1988, Risch et al. 1994), have also been associated with dietary fat, but the evidence at present is very limited.

In some populations, fat intake is highly correlated with the consumption of meat, a major source of saturated fats in the diet. Furthermore, some recent studies found strong positive associations between the intake of meat, especially red meat, and the risk of cancers of the prostate and colon (Giovannucci et al. 1993, Le Marchand et al. 1994). Meat may contain carcinogens other than fat, because cooking such foods at high temperatures (as in broiling or frying) can generate potentially carcinogenic polycyclic hydrocarbons and heterocyclic amines (Sugimura 1985). Thus, a causal relationship between dietary fat and cancer in humans has not yet been conclusively established.

Animal studies also show a relationship of dietary fat to cancer, although the only fatty acid that has been shown unequivocally to enhance carcinogenesis in animal models is linoleic acid (see Chapter 2). Although dietary fat and calories appear to interact in modifying cancer risk in animals (Welsch et al. 1990, see Chapter 2), an independent effect of fat has also been demonstrated (Birt et al. 1993).

Per capita intake of fat in the U.S. appears to have increased throughout the period 1909-1913 to 1985. During this time, estimated saturated fat intake rose from about 50 g/day to 65 g/day.

As a percent of total calories, fat has remained relatively constant (approximately 40%) during the past twenty years (NRC 1989a). However, the data for these apparent trends are difficult to interpret, because survey methods have varied over time.

Risk Estimates Derived from Epidemiologic Studies

In only a few cases can dietary risks be estimated from human data. The example of aflatoxin shows the difficulties in risk estimation, even for a well-studied case with human data. Several investigators have estimated the cancer risk associated with aflatoxin B1 exposure of individuals living in southern China. The range of estimates varies considerably, depending upon whether additive or multiplicative models were chosen and whether the exposed individuals also tested positive for the hepatitis B virus. Using a relative risk model, Hoseyni (1992) calculated a lifetime risk for liver cancer of 2.5 per million per ng/kg body weight/day for aflatoxin B1 in hepatitis-negative individuals and 62 per million per ng/kg body weight/day in hepatitis-positive individuals. Wu-Williams et al. (1992) calculated a lifetime risk of primary hepatocellular carcinoma for males of 5.6 per million per ng/kg body weight/day using a multiplicative relative risk model, and 46 per million per ng/kg body weight/day using an interactive risk model. Risk levels were at least an order of magnitude higher for hepatitis B-positive individuals. Bowers et al. (1993) estimated a lifetime cancer risk of 9 per million per ng/kg body weight/day and 230 per million per ng/kg body weight/day for hepatitis B-negative and -positive individuals, respectively. Aflatoxin B1 exposure levels are generally on the order of a few ng/kg body weight/day. Thus, primary hepatocellular carcinoma risks from aflatoxin appear to be on the order of one per 10,000, which would account for a fraction of the liver cancer in the U.S. However, those infected with hepatitis B (which, for example, is endemic to the Asian immigrant community) could face liver cancer risks on the order of 10-3 to 10-2.

Data from rats, the species most sensitive to aflatoxin, shows carcinoma potency estimates from aflatoxin is considerably greater than that derived from epidemiologic data (i.e., cancer potency estimates of 5,000 to 20,000 (mg/kg-day)-1 for the rat versus 5 to 50 (mg/kg-day) -1 for the hepatitis B negative humans).

Risk Estimates Derived from Toxicological Studies

Cancer risk estimates can also be derived from toxicological studies, although such estimates are uncertain due to the need to extrapolate from high to low doses, and from animals to humans. An example of risk estimation relying on toxicological information for the cancer potency estimate is given here for illustrative purposes. Gaylor and Kadlubar (1991) investigated the cancer risk to humans exposed to (polycyclic) heterocyclic amines formed during the cooking of food. Cancer potency estimates were derived from rodent data. With one exception, the animal bioassays consisted of only one dose level and controls. Hence, the potency estimates were based on the one-hit model. Cancer potency estimates for several of the heterocyclic amines were as high as one in a 100,000 per ppb in the total diet. Data on exposures in the human diet were sketchy, with ranges from 0.02 to 83 ppb in the total diet. Based on these data, it was estimated that the upper limit on cancer risk in the total average diet due to heterocyclic amines formed during cooking is on the order of one per 10,000.

Recently, Layton et al. (1995) examined the cancer risk of heterocyclic amines in cooked foods, establishing concentrations of heterocyclic amines from published literature. Average consumption of foods containing heterocyclic amines were obtained from 3-day dietary records of a random survey of 3563 individuals conducted by the USDA in 1989. Cancer potencies were taken from Bogen (1994), which accounted for the induction of tumors at multiple tissue sites. An upper bound estimate at the incremental cancer risk of heterocyclic amines produced by cooking is 1.1 × 10-4, using cancer potencies based on body surface area. Nearly half of the risk was due to the ingestion of 2-amino-1-methyl-6-phenylimidazo[4,5-6] pyridine (PhIP). Consumption of beef and fish products accounted for about 80% of the total risk. The authors calculated that the ingestion of these dietary heterocyclic amines might account for at most a small fraction (0.25%) of colorectal cancers in the United States.

Apportionment of Dietary Cancer Risk

Lutz and Schlatter (1992) recently estimated cancer risks to humans from known animal and human carcinogens present in the diet. Human exposure estimates were based on average daily intake estimates for dietary carcinogens in Switzerland. The total cancer risk was compared with the number of cancer cases attributed by epidemiologists to dietary factors, about 80,000 cases per 1 million people. The investigators state that, "Except for alcohol, the known dietary carcinogens could not account for more than a few hundred cancer cases." This estimate is only a small fraction of the risk estimate provided by Doll (less than 0.01%). In deriving these estimates, Lutz and Schlatter did not consider potential interactions among the known human carcinogens, the unspecified (and unknown) carcinogens, and dietary macronutrients, nor did they consider the multifactorial nature of cancer causation in humans. Consideration of these factors could result in substantially greater risk estimates than those provided.

Because their estimate proved to be negligible, Lutz and Schlatter (1992) searched for other possible contributors to risk and concluded that the vast majority of dietary cancers may be due to overnutrition.

As already discussed, epidemiologic studies have demonstrated an association between excess calories and the occurrence of breast, colon, and prostate cancer. Furthermore, studies in animals have shown that dietary restriction results in significant reductions in spontaneous and induced cancer incidence, as well as increases in tumor incidence in animals fed ad libitum when compared to dietary-restricted control animals. These studies suggest that the increased tumor incidence may be attributed to the excess food intake.

To test this assumption, Lutz and Schlatter determined that the average calorie intake in Switzerland between 1985 and 1987 for the age group 20-39 was 2315 kcal/person/day (excluding alcohol). Basal requirements for this age group are 1963 kcal/person/day. Thus, the investigators concluded that the average Swiss adult is overnourished by about 5.5 kcal/kg/day, or 1.9 g food/kg body weight/day. Using a TD50 of 16g/kg/d for excess feed in rats, the authors estimate that 60,000 cancer cases in a population of one million could be attributed to excess food intake in Switzerland, noting that "this value is provocatively close to the number of cancer cases not explained by the human dietary chemical carcinogens."

Scheuplein (1990) examined the contribution of various food categories to overall human cancer risk. His estimates of daily intakes of selected food categories are shown in Table 5-7. The intake of a food category that is calculated to be carcinogenic is also shown. This intake was calculated by multiplying the food intake by the proportion estimated to be carcinogenic, yielding "unverified estimates that seem reasonable." Assuming equal potencies (i.e., equal risk per mg of carcinogen) across food categories, the naturally occurring carcinogens account for 99.8% of human dietary cancer. Assuming that the potencies of traditional food, spices and flavors, charred protein, and mycotoxins are 0.01, 0.1, 100, and 1000 times the potencies of synthetic chemicals, respectively, Scheuplein estimated that naturally occurring carcinogens account for approximately 91% of the cancers attributable to the human diet.

Table 5-7. Daily Intakes of Dietary Carcinogens in Selected Food Categories.

Table 5-7

Daily Intakes of Dietary Carcinogens in Selected Food Categories.

Scheuplein's attempt to quantitate the contribution of dietary risk factors to human cancers demonstrates the gaps in our present knowledge. There is a need to systematically investigate the role of dietary modulation of malignant neoplasia. Using available data, plausible assumptions, and mathematics, he effectively challenges the scientific community to prove or disprove his hypotheses. Although the Surgeon General's Report on Nutrition and Health (1988) concludes that many food factors are involved, it cites in particular the ''disproportionate consumption of food high in fats, often at the expense of food high in complex carbohydrates and fiber—such as vegetables, fruits, and whole grain products that may be more conducive to health."

Risks of Naturally Occurring Versus Synthetic Carcinogens in the Diet

Potency of Naturally Occurring and Synthetic Carcinogens

Distribution of Potency

Only a small fraction of chemicals, natural or synthetic, has been adequately tested for carcinogenicity. Over 13 million chemicals have been assigned CAS numbers by the Chemical Abstracts Service. Testing has been undertaken for less than 0.01% of those, a number of these because of suspicions regarding their potential carcinogenicity.

Both naturally occurring and synthetic agents currently found in the U.S. diet have been identified as carcinogens by the International Agency for Research on Cancer (IARC) or the National Toxicology Program (NTP). Appendix B lists all agents identified by IARC as group 1, 2A, or 2B carcinogens or by the NTP as known or reasonably anticipated to be carcinogens. This list of carcinogens is divided into four tables: agents that may be encountered in U.S. diets (Table B-1), agents formerly encountered in U.S. diets but no longer (Table B-2) agents rarely or accidentally encountered in U.S. diets (Table B-3), and agents unlikely to be present in U.S. diets (Table B-4). When available, TD01 values are provided for these agents. The committee also attempted to determine whether an agent listed in the appendix should be considered as synthetic or naturally occurring as defined in Chapter 1. Naturally occurring agents are further subclassified as constitutive, derived, acquired, or added. The subclassification of naturally occurring agents was sometimes arbitrary, with some agents belonging in more than one category.

Table 5-8 compares the geometric means of the TD01 values for various synthetic versus naturally occurring carcinogens found in U.S. diets. As depicted in Figure 5-4, TD01 values for the constitutive agents are higher than the TD01 values for the nonconstitutive agents, indicating that the constitutive agents are less carcinogenic. This may be an artifact of selection bias, however, since few constitutive agents have been identified and TD01 values are available on even fewer of these. In addition, because of the uncertainty in the TD 01 estimates, the differences are not considered to be of practical significance. As shown in Figure 5-4, a comparison of TD01 values for synthetic versus naturally occurring agents indicates that the latter appear to be somewhat more potent, with lower TD01 values. Because the selection of the agents for testing is not random, and the number of naturally occurring and synthetic dietary agents tested is relatively small, the extent to which these results can be generalized is not known.

Table 5-8. Comparison of TD01 Values for Selected Synthetic and Naturally Occurring Carcinogens in the Diet.

Table 5-8

Comparison of TD01 Values for Selected Synthetic and Naturally Occurring Carcinogens in the Diet.

Figure 5-4. TD01 values for selected synthetic and naturally occurring carcinogens in the diet.

Figure 5-4

TD01 values for selected synthetic and naturally occurring carcinogens in the diet.

Interpretation of Results

Figure 5-4 supports the conclusion that, quantitatively, the carcinogenic activities of naturally occurring and synthetic dietary agents do not appear to differ substantially. Although the considerations involved in making this inference have been mentioned, a more extensive discussion of issues arising in the comparative carcinogenicity of dietary naturally occurring and synthetic carcinogens is now given.

Route of Exposure. Several naturally occurring carcinogens inherent in commonly eaten foods were identified as carcinogenic on the basis of findings reported in occupational studies, in which workers were exposed to high concentrations, or from inhalation bioassays conducted in animals. Cancer potencies may differ, depending on the route of exposure. This appears to be the case for several of these agents. For example, the cancer activity associated with oral exposure appears to be considerably less than that for inhalation exposures for asbestos, crystalline silica, hexavalent chromium, cadmium, nickel, formaldehyde, and acetaldehyde; furthermore, the degree to which oral exposures to these agents poses a cancer risk is currently the subject of debate. Cancer potencies for these naturally occurring agents as a result of inhalation are not representative of the cancer potencies associated with dietary intake and thus were not used in the comparisons. The fact that the FDA has permitted the addition of acetaldehyde to various foods and formaldehyde to milk (as a vitamin D carrier) underscores the fact that carcinogenic activity associated with the two routes is understood to be substantially different. Any of the above-mentioned agents, if active by the oral route, would have oral TD01 values considerably higher than the inhalation values.

Definition of Naturally Occurring. The distinction made here is between those agents not known to occur in nature (as noted by IARC) and the rest of the chemical world. Thus, benzo(a)pyrene and a number of other pyrolytic products are classified as naturally occurring dietary carcinogens, even though relatively large dietary contributions can result from plant uptake of fossil fuel combustion byproducts.

Selection Bias. The agents for which cancer potency estimates are available do not represent a random sample of the universe of synthetic and naturally occurring agents. For the most part, cancer testing has been guided by suspicions of carcinogenicity from noted similarities to other compounds recognized as carcinogens, structure-activity considerations, mechanistic data, and indications from epidemiologic studies (Rosenkranz 1992). Finding one had actor typically leads to the identification of other bad actors. For example, the finding that mutagenic products are formed during cooking has stimulated considerable research, including the conduct of cancer bioassays. As a result, this class of agents is relatively well identified (see Table 5-1). Similarly, observations of an increased incidence of bladder cancer among workers exposed to synthetic dyes and dye intermediates led to the discovery of the carcinogenicity of benzidine and related dyes; consequently, cancer potency estimates are available for a number of these agents. In contrast, only a few constitutive agents, normally present in U.S. diets, have been well studied by the oral route and identified as carcinogenic. The extent to which these are representative of all carcinogenic constituents is unknown. There may be literally millions of constitutive agents. Caffeic acid is one of the prevalent constitutive carcinogens, but as Lutz and Schlatter (1992) point out, the compound may be carcinogenic only at high doses (see, for example, Ito et al. 1991); thus it is improbable that the risks derived from animal bioassay data using standard techniques are representative of actual risks for human beings. The estimated carcinogenic potency of caffeic acid is therefore unlikely to be indicative of its activity under conditions of most human exposures. Further, whether caffeic acid actually poses a human cancer risk is unclear.

Uncertainty in Potency Estimates. The risks predicted using cancer potency values such as the reciprocal of the TD01 are often described as upper bound estimates. It is widely recognized that these risk predictions are not precise and that the uncertainty inherent in risk calculations is usually large. Ratios of cancer potencies for dissimilar agents can be subject to larger uncertainty. Statistical techniques can be applied to estimate the error in deriving potency parameters from bioassay data (e.g., the error in deriving the Image img00010.jpg parameter). Areas that are uncertain and not as amenable to quantification include interspecies extrapolation, estimation of lifetime risks from experiments conducted for less than a lifetime, and the application of potency parameters to exposure levels and patterns considerably different from those in the bioassay that has served as the basis for calculating the potency. In addition, comparisons of potencies for multiple agents in a mixture carry even greater uncertainty, as was discussed previously.

HERP Approach

A number of recent papers by Ames and colleagues have discussed the relative importance of naturally occurring and synthetic carcinogens with respect to dietary cancer risk (Ames et al. 1987, 1990a,b; Ames and Gold 1990a,b,c, 1995; Gold et al. 1992, 1993b, 1994). In these papers, the investigators compare the ratios of human exposure to animal cancer potency for different agents, noting that the result serves as a guide to priority setting (Ames et al. 1987). One of these ratios, the Human Exposure/Rodent Potency index (HERP), is constructed by dividing an estimate of human exposure by the TD50. The larger the value of the HERP index, the closer the level of human exposure is to the dose estimated to cause a 50% excess cancer risk in animals (Gold et al. 1990, 1992). Based on a tabulation of HERP indices for various carcinogens using data from the Carcinogenic Potency Database, these investigators assert that the risks associated with exposure to natural carcinogens is greater than many synthetic agents. For example, the HERP index for 8-methoxypsoralen in a quarter of a parsnip (0.06) is larger than the HERP index for chloroform in one liter of tap water (0.001). However, Ames and Gold (1990b) offer assurances that the exposure to individual natural and synthetic carcinogens is low, and there is no cause for alarm for those who eat a balanced diet.

Ames and his associates investigated agents they identified as naturally occurring and synthetic pesticides to determine their role in the induction of cancer in humans (Ames et al. 1990a,b; Gold et al. 1993a). They calculated "that 99.99% (by weight) of the pesticides in the American diet are chemicals that plants produce to defend themselves" (Ames et al. 1990b). In another study Ames et al. (1993) compared the HERP indices for agents they identified as natural plant pesticides to the indices of synthetic organic pesticides used to enhance agricultural productivity. On the basis of these results, they concluded that dietary exposure to naturally occurring pesticides, weighted by carcinogenic potency, is greater than dietary exposure to synthetic pesticides. They further noted that a high proportion of "natural pesticides" were positive in animal cancer tests and concluded that naturally occurring and synthetic chemicals present in the diet are equally likely to test positive in animal cancer tests, and that the potential risks of dietary residues of synthetic organic compounds are insignificant.

Comparisons of HERP indices among different types of naturally occurring carcinogens have also been made (Gold et al. 1993, 1994). For example, on the basis of HERP comparisons Gold et al. (1994) conclude that "possible hazards from HA (heterocyclic amines) rank below those of most 'natural pesticides' and products of cooking and food preparation; synthetic pesticide residues also rank low."

In evaluating comparisons of HERP indices, certain scientific limitations need to be considered. One is that most comparisons are made on the basis of carcinogens in a single serving, rather than on the basis of average daily exposure. After comparing on the basis of average exposures, Perera and Boffetta (1988) note that "the HERP scores of many manmade pollutants are comparable to those of naturally occurring carcinogens in the diet." A second limitation is that some agents identified by Ames et al. as natural pesticides or food carcinogens and for which HERP indices were calculated lack sufficient evidence of carcinogenicity. For example, Ames et al. (1990b) list 27 agents as carcinogenic plant chemicals; in an earlier report (1987) they identify 11 compounds as 'natural pesticides' and dietary toxins. However, only 7 of the first and 5 of the second were considered by IARC to be Group 1 or 2 carcinogens, or by NTP as known to be or reasonably anticipated to be carcinogenic. A number of the compounds were classified by IARC as not classifiable as to its carcinogenicity to humans (i.e., IARC Group 3) and a few were not reviewed by IARC. Consequently, there is an uneven comparison of synthetic agents that have been formally recognized as carcinogens to naturally occurring agents with insufficient evidence. Third, the analysis is not comprehensive in that only a few selected synthetic and naturally occurring agents are compared, including some obscure exposures to natural agents (e.g., the high HERP index for a serving of comfrey root and symphytine in comfrey-pepsin tablets is compared to the low HERP index for daily intake of EDB in grains averaged over the U.S. population). Fourth, as is common to a number of assessments, the increased risks to subpopulations are not addressed, subpopulations with considerably higher exposure than the calculated average exposure (or single serving in some cases) or those with identifiable increased susceptibility. Fifth, as Wartenberg and Gallo (1990a,b) and Hoel (1990) point out, all carcinogens are implicitly assumed to act by similar mechanisms, and the approach fails to account for the nonlinear regions of the dose-response curve, even in known and understood cases such as vinyl chloride and AF-2 (Littlefield et al. 1980). Sixth, the fraction of untested natural and synthetic agents that would be identified as carcinogenic after adequate testing is unknown. Ames et al. (1990b) note that half of the agents they identify as natural carcinogens test positive in rodent bioassays. Of chemicals selected for carcinogenicity testing by the NTP, 68% of those chosen because they were suspected to be carcinogenic were found to be positive in at least one sex/species, in contrast to 21% of those selected primarily on the basis of exposure considerations (Fung et al. 1993). A smaller fraction was found to meet the criteria for sufficient evidence in animals by IARC. It is unclear what the finding would be for a random selection of agents from the synthetic and natural world. Finally, because the diet is a complex mixture of agents containing both carcinogenic and anticarcinogenic components, evaluation of individual ones in isolation can be misleading, particularly since some agents possess dual activity. With respect to the latter, the HERP index for caffeic acid suggests high hazard, but experimental evidence indicates the possibility for protective effects at dietary levels.

In a more recent publication by Gold and colleagues (Gold et al. 1992) the second criticism mentioned above is partially addressed in that some chemicals in foods are compared on the basis of average servings for a particular food. Summing HERP indices for the same chemical (e.g., adding d-limonene in black pepper and orange juice), and then ranking HERP indices from highest to lowest for agents they identify as natural and synthetic pesticides results in the following: caffeic acid, d-limonene, safrole, mix of hydrazines, catechol, DDT (before ban in 1972), UDMH (in 6 oz of apple juice), allyl isothiocyanate, DDE, 8-methoxypsoralen, glutamyl-p-hydrazinobenzoate, EDB (before 1984 ban), carbaryl, toxaphene, p-hydrazinobenzoate, dicofol, -methylbenzyl alcohol, lindane, chlorobenzilate, chlorothalonil, folpet, captan. Removing those agents classified by IARC as "not classifiable as to their carcinogenicity in humans" results in the following list, with naturally occurring agents in italic: caffeic acid, safrole, mix of hydrazines, DDT (before ban in 1972), UDMH, DDE, 8-methoxypsoralen, glutamyl-p-hydrazinobenzoate, EDB (before 1984 ban), toxaphene, p-hydrazinobenzoate, -methylbenzyl alcohol, lindane, chlorothalonil, folpet.

Additional Comparisons

The committee made an additional attempt to compare the potential risks of naturally occurring and synthetic carcinogens in food, focusing on a selected number of substances that may be found in relatively high concentrations in the diet. These comparisons were based on an exposure potency index (EPI) defined as the ratio of dietary exposure to carcinogenic potency. Carcinogenic potency is expressed in terms of the TD01, then estimated to induce an excess lifetime cancer risk of 1%. This is analogous to the HERP index used by Ames and Gold (1987), with the TD50 replaced by the TD01. The TD01 is used here because it is closer to human exposure levels, yet not so low as to be highly model dependent. Because potency rankings based on the TD50 and TD01 are generally similar (Krewski et al. 1990), risk comparisons based on the EPI are expected to be similar to those based on the HERP. The TD01 is readily available for the series of compounds assembled by the committee in Appendix B.

Many criticisms of the comparison done by Ames and colleagues apply to the comparisons in Table 5-9 as well. Only a few of the dietary carcinogens listed in Tables 5-1 and 5-3 are considered. The committee focused on some agents for which exposures are believed to be high relative to other carcinogenic agents. Average dietary exposures are difficult to estimate, and upper and exposures even more so due to the paucity of consumption and concentration data available on which to base such estimates. Nonetheless, these crude estimates suggest that exposure to individual synthetic and natural chemicals might fall within a comparable range.

Table 5-9. Exposure/Potency Index for Selected Naturally Occurring and Synthetic Carcinogens in the Diet.

Table 5-9

Exposure/Potency Index for Selected Naturally Occurring and Synthetic Carcinogens in the Diet.

Some of the putative carcinogens in foods (excluding excess fat) are ingested in greater quantities than the others (milligrams rather than micrograms per day). These include caffeic acid, a naturally occurring compound shown to be both carcinogenic and anticarcinogenic in animal models, and BHA, a synthetic chemical; both of these are consumed at comparable levels in the human diet. While potentially carcinogenic to humans if consumed at levels producing cancer in animal bioassays, these simple phenolic compounds are potentially anticarcinogenic at lower levels, operating through an antioxidant mechanism (see Chapter 2). The degree, therefore, that caffeic acid poses a cancer risk to humans is unclear. Although the consumption of fruits and vegetables containing caffeic acid is considered to be protective against cancer, extremely high levels of coffee consumption may pose a risk. The available scientific information does not enable us to resolve this issue.

Table 5-9 illustrates two other important points. First, some consumer subgroups might be exposed to relatively high levels of naturally occurring carcinogens, for example, high consumers of stone fruit distillates and well-done meats. When averaged out over the population, these levels may appear unimpressive, but for the individual high consumer the risk might be significant. Second, reductions in exposures to individual dietary carcinogens have been significantly reduced after the identification of these agents as carcinogens, for both synthetic agents (e.g., DDT and UDMH) and natural ones (e.g., nitrosodimethylamine and urethane). Not shown in the table are several agents for which past levels of exposure are likely to have been very high (e.g., arsenic, benzyl violet 4B, and dihydrosafrole). This suggests caution when interpreting the results of exposure calculations for dietary carcinogens.

Summary And Conclusions

Humans are exposed to a wide variety of chemical substances, naturally occurring and synthetic, in their diets, although the carcinogenic potential of most of these substances has not been evaluated. Diets high in calories and fat are associated with increased cancer risks, although the mechanisms by which these substances increase risks are not well understood. Other dietary substances known to increase cancer risk in animals and possibly in humans include the mycotoxins (e.g., aflatoxins) and the heterocyclic amines, which are formed when meat is cooked at high temperatures. Recent investigations have focused on the possibility that some naturally occurring compounds produced by plants may possess carcinogenic properties. In total, such natural toxins appear to be present in the diet in greater quantities than residues of synthetic organic pesticides. Unlike most naturally occurring dietary constituents, synthetic ones such as direct and indirect food additives and pesticide residues are highly regulated, with stringent limits placed on their allowable levels of synthetic chemicals in foods.

The risks associated with dietary carcinogens depend on the carcinogenic potency of the substance and its level of ingestion. At present, dietary epidemiologic studies have identified a small number of specific components capable of causing cancer in humans. Only a few of these agents, such as aflatoxin and arsenic compounds, occur in food and water. Exposure to most of the other known human carcinogens occurs primarily through nondietary pathways. Laboratory studies have identified a much larger number of synthetic and naturally occurring agents capable of causing cancer in experimental animals. However, prediction of human cancer risks based on laboratory results is uncertain because extrapolations must be made from high to low doses and from animals to humans. Thus, important questions remain about the relevance of findings from animal studies for predicting human cancers. In particular, rodent carcinogens, some of which may act by increasing cell proliferation only at high doses, may pose little or no risk at low doses. Although the use of the maximum tolerated dose (MTD) in rodent carcinogenicity studies has been questioned, a recent review of its use by the National Research Council (1993b) failed to identify a suitable alternative. In the future, short-term tests for DNA adduct formation, mutation, and cellular proliferation, all critical factors in carcinogenesis, may provide a stronger basis for evaluating carcinogenic risks in the absence of adequate epidemiologic data.

Estimating dietary exposures requires knowledge of food consumption patterns, which vary substantially among individuals and over time. Despite periodic nationwide surveys, food consumption data are somewhat limited, particularly for the young and the elderly. The concentrations of chemical substances with carcinogenic potential in specific foods, including both naturally occurring and synthetic pesticides, are not well characterized and can vary widely.

In addition to agents that may increase cancer risk, the human diet also includes substances that may reduce the risk of cancer. These include broad classes of foods such as fruits and vegetables, as well as specific substances such as beta carotene and vitamin A. Because the human diet is a complex mixture of many constituents, dietary cancer risk assessment needs to take into account the potential for synergistic and antagonistic interactions among dietary components.

Methods for estimating cancer risk rely largely on epidemiologic and toxicological data. Epidemiologic studies provide direct information on cancer risks in humans but are subject to certain limitations, including inadequate exposure data and confounding due to exposure to multiple agents. Toxicological studies can be conducted under controlled conditions but provide only indirect information on human cancer risks. Several hundred chemicals have been found by the IARC and NTP to have sufficient evidence of carcinogenicity in animals.

The potency of carcinogenic agents may be quantified by the reciprocal of the TD50 or TD01, the doses estimated to produce a 50% or 1% excess lifetime cancer risk, respectively. The Carcinogenic Potency Database, which contains TD50 values for about 600 rodent carcinogens, serves as a useful source of information for comparing carcinogenic potency values.

The fact that the diet is a complex mixture raises important considerations when conducting cancer risk assessments. Evidence from both epidemiologic and toxicological studies indicates that interactive effects may occur. Hence, the effect of altering the concentration of one dietary substance may influence the carcinogenic (or anticarcinogenic) activity of another. The extent of such interactions among dietary constituents is not yet well characterized.

To evaluate the relative potency of naturally occurring and synthetic carcinogens in the diet, the committee compiled data on the carcinogenic potency of about 233 chemical carcinogens, including 65 naturally occurring substances and 168 synthetics. Of those with potency values, 37 naturally occurring and 60 synthetic agents were identified in U.S. diets. The potency values were derived from bioassay data tabulated in the CPDB using an expedited estimation procedure (as described by Hoover et al. 1995, Cal/EPA 1992) or, if available, drawn from regulatory agencies that had performed a more in-depth analysis. The data set included agents identified by IARC as having sufficient evidence of carcinogenicity in humans or animals or by the NTP as known or reasonably anticipated to be human carcinogens. The committee recognized that this represented a somewhat select group of substances, but considered it instructive to analyze the relative potency on this sample.

In this analysis, carcinogenic potency was expressed in terms of the reciprocal of the TD01, defined as the dose inducing an excess lifetime cancer risk of 1%. Results of the analysis indicate that the average potency of the naturally occurring carcinogens was somewhat higher than that of the synthetic carcinogens. The average of the few potencies for constitutive agents was roughly the same as for synthetic agents. Potencies for the derived, acquired, and pass-through agents were higher than the synthetic and constitutive agents, suggesting greater carcinogenic activity. However, there was wide variation in potency within the groups studied, with almost complete overlap of the potency ranges. In light of this variation, the uncertainty associated with estimates of cancer potency, and the potential problem of selective sampling, the committee concluded that these data failed to establish a clear difference between the potency of naturally occurring and synthetic carcinogens present in the human diet.

In addition to looking at potency data, characterization of risk requires a consideration of exposure. Although considerable information is available on food consumption patterns and on the concentrations of certain carcinogenic substances in foods, this information was insufficient to compare the risks posed by the naturally occurring and synthetic carcinogens considered previously. Furthermore, the committee decided that there was insufficient information on the carcinogenic potential of ''natural pesticides" evolved by plants to conclude that such agents pose a greater dietary cancer risk than synthetic pesticides.

The HERP index, proposed by Ames and Gold, compares the human exposure to a rodent carcinogen with the potency in a bioassay (TD50), expressed as a percentage: HERP = (human exposure/TD50) × 100. Although not a direct measure of risk, larger HERP values may be viewed with greater concern than lower values. Ames and his colleagues have argued that HERP values for naturally occurring carcinogens often exceed those for synthetic carcinogens, largely because of the greater consumption of a few naturally occurring substances such as caffeic acid, which occurs in high concentrations in commonly consumed foods such as lettuce, apples, and coffee. Although Ames suggests that naturally occurring carcinogens present in the diet may pose a greater risk than synthetic dietary carcinogens, he notes that dietary exposure to both kinds is low, and should be of little concern when ingesting a balanced diet.

The observation that the HERP indices for naturally occurring carcinogens tend to exceed those for synthetic carcinogens is broadly consistent with the committee's calculation that the average potency of naturally occurring carcinogens exceeds that of synthetic carcinogens. However, the committee also notes that the analyses of HERP indices performed by Ames and colleagues are based on a select number of agents and include a number of compounds for which there is not sufficient evidence of carcinogenicity. Based on average daily consumption data, HERP indices for natural and synthetic carcinogens appear to be somewhat comparable.

Several investigators have attempted to assess the proportion of the human cancer burden attributable to different sources. In 1981, Doll and Peto estimated with some confidence that about 30% of all human cancer could be attributed to tobacco smoking. They provided as a best guess that 35% was due to dietary factors, although the precise components of diet contributing to this risk are not well understood. There also exists a high degree of uncertainty about the impact of diet on the human cancer burden, with Doll and Peto citing a plausible range of uncertainty of 10-70%. Subsequent reviews have largely supported this initial analysis.

Based on the information summarized in this chapter, the committee reached the following conclusions.

  • Diet contributes to an appreciable portion of human cancer.
    Authoritative reviews have speculated that diet plays a role in about a third of all human cancer. Although the actual figure could be somewhat higher or lower, these reviews have identified diet as a major contributor to the human cancer burden.
    In most of these reviews, individual dietary constituents responsible for increased cancer risk were not identified. Nonetheless, a substantial body of evidence suggests that diets high in calories and fat appear to increase cancer risk. Human studies indicate that aflatoxin and alcoholic beverages increase the cancer risk. Animal studies have shown that a number of other chemical constituents present in food, both naturally occurring and synthetic, increase the risk of cancer.
    Although the precise components of diet responsible for increased cancer risks are generally not well understood, it is felt that individual chemical constituents of the human diet, both naturally occurring and synthetic, represent generally low cancer risks. The degree to which these individual agents in aggregate pose a risk is unclear. However, excess calories and fat appear to have a substantial impact on human cancer.
  • The human diet contains both naturally occurring and synthetic agents that may affect cancer risk.
    Constituents of food present in the human diet that have been shown to increase cancer risk in animal and human studies may be naturally occurring or synthetic. The risks associated with both depend on the level of exposure and the potency of these chemicals, as well as the susceptibility of the host.
  • Excess calories and fat appear to contribute substantially to the human cancer burden.
    In animal studies, diets low in calories and fat are associated with a reduction in cancer risk. Although less conclusive, epidemiologic evidence in humans indicates that excess calories and fat play a role in the occurrence of colon, prostate, and possibly breast cancer. In terms of cancer causation, current evidence suggests that the contribution of calories and fat outweighs that of all other individual food chemicals, both naturally occurring and synthetic.
  • The potencies of known naturally occurring and synthetic carcinogens present in the human diet do not differ appreciably.
    The committee compared the potencies of 233 chemical carcinogens, 65 of which are naturally occurring and 168 are synthetic. Potency was expressed in terms of the reciprocal of the TD01 derived from rodent carcinogenicity studies. Although the average potency of these naturally occurring substances was greater than that of the synthetic agents, the range in potency observed within both groups was broad, with the distributions of potency spanning comparable ranges and encompassing some six orders of magnitude. Since these results are based on a select group of chemicals, it is difficult to generalize to a larger set of substances.
  • Taking calories and fat into consideration, the human cancer risk from naturally occurring substances in the diet exceeds that of synthetic dietary carcinogens. However, if calories and fat are excluded from consideration, the data now available do not permit a firm conclusion about the relative risks of naturally occurring and synthetic food chemicals. Nonetheless, the committee felt that it is plausible that naturally occurring chemicals present in food pose a greater cancer risk than synthetic chemicals.
    Cancer risk is determined by the potency of the carcinogen and the level of exposure. The mechanism by which cancer is induced can also influence risk, although these mechanisms appear similar for naturally occurring and synthetic substances. Calories and fat represent major components of the human diet and play an important role in the dietary contribution towards human cancer. Naturally occurring chemicals are present in the food supply in much larger quantities than synthetic chemicals. Although only a limited number of naturally occurring food chemicals have been tested in rodent bioassays for carcinogenic activity, the proportion of them demonstrating carcinogenic properties in such tests is comparable to the proportion of synthetic chemicals that are carcinogenic. The committee also felt that further testing of carefully selected naturally occurring food chemicals would result in the identification of additional carcinogens.
  • The great majority of individual naturally occurring and synthetic food chemicals are present in human diet at levels so low that they are unlikely to pose an appreciable cancer risk.
    Although the human diet contains both naturally occurring and synthetic carcinogens, these chemicals are generally present at very low concentrations, and are thus unlikely to pose an appreciable cancer risk. Application of pesticides to food crops is subject to stringent control. Thus, most foods contain no detectable levels of such residues and the residues that are detected are generally present at levels well below established tolerances. Similarly, with the exception of a few agents such as caffeic acid, naturally occurring carcinogens, including "natural pesticides," are also present at very low levels.
  • The human diet contains anticarcinogens that reduce cancer risk.
    The human diet contains a number of nutrients, including fibre and micronutrients such as the antioxidant vitamins A, C and E, that appear to reduce cancer risk. In addition, several non-nutritive agents, including the isoflavonoids, phenolic acids, and isothio-cyanates, have been shown to demonstrate anticarcinogenic effects. Diets low in calories and fat are also associated with reduced cancer risk.
    Epidemiologic studies have associated diets rich in fruits and vegetables with reduced rates of human cancer. These foods are sources of a number of constitutive chemicals with anticarcinogenic properties.
  • Carcinogens and anticarcinogens present in the diet may interact in a variety of ways, which are not fully understood.
    Fruits and vegetables contain both anticarcinogenic and carcinogenic constituents. Because such foods are associated with reduced cancer risk in humans, the anticarcinogenic properties would appear to be greater than their carcinogenic properties.

A global assessment of dietary cancer risks is more difficult. Because interactions between dietary carcinogens and anticarcinogens can occur, it is difficult to predict overall dietary risks based on an assessment of the risks for individual components. Although risk assessment methods for mixtures are available, the information on interactions among dietary constituents is limited. The large number of dietary constituents present in foods further complicates the evaluation of potential interactions.

Overall Conclusions

Although diet clearly contributes importantly to human cancer, the committee found it difficult to identify the specific components of diet that serve to increase or decrease cancer risk. Although excess calories and fat appear to represent the single most important component of the human diet that increases cancer risk, the epidemiological evidence in this regard is not entirely consistent. Most other dietary constituents suspected of impacting upon human cancer, both naturally occurring and synthetic, account individually for a small proportion of the diet, and may pose correspondingly low risks. Much of the available evidence on these compounds derives from studies in animals, and may or may not be indicative of a human cancer risk.

The presence of anticarcinogenic substances in many foods needs to be considered in any evaluation of dietary cancer risks. Fruits and vegetables contain antioxidant vitamins that have demonstrated anticarcinogenic properties in laboratory studies; diets rich in fruits and vegetables are in fact associated with reduced cancer risks in humans. Many commonly consumed foods such as beef steak and broccoli contain low concentrations of a number of both carcinogenic and anticarcinogenic substances.

On the basis of the information available at this time, the committee found it particularly difficult to judge whether naturally occurring or synthetic dietary components (exclusive of excess calories and fat) present the greater cancer risk to humans. Much of the information on the carcinogenic potential of these substances derives from animal tests conducted at high doses, results of which are difficult to translate directly to humans. Nonetheless, the committee found that a series of selected naturally occurring and synthetic dietary components appeared to span the same range of carcinogenic potency, and that naturally occurring and synthetic substances can cause cancer by similar mechanisms. The committee also noted that the fractions of naturally occurring and synthetic substances that are positive in animal cancer tests are comparable, and that the dietary concentrations of many naturally occurring animal carcinogens such as caffeic acid exceed those of most synthetic carcinogens present in or on food. These observations supported the committee's conclusion that natural components of the diet may prove to be of greater concern than synthetic components with respect to cancer risk, although additional evidence is required before this conclusion can be drawn with certainty.


  • Ames, B.N., and L.S. Gold. 1990. a. Chemical carcinogenesis: too many rodent carcinogens. Proc. Natl. Acad. Sci. U.S.A. 87:7772-7776. [PMC free article: PMC54830] [PubMed: 2217209]
  • Ames, B.N., and L.S. Gold. 1990. b. Dietary carcinogens, environmental pollution, and cancer: some misconceptions. Med. Oncol. Tumor Pharmacother. 7:69-85. [PubMed: 2232941]
  • Ames, B.N., and L.S. Gold. 1990. c. Misconceptions on pollution and the causes of cancer. Angew. Chem. Int. Ed. Engl. 29:1197-1208.
  • Ames, B.N., R. Magaw, and L.S. Gold. 1987. Ranking possible carcinogenic hazards. Science 236:271-280. [PubMed: 3563506]
  • Ames, B.N., M. Profet, and L.S. Gold. 1990. a. Nature's chemicals and synthetic chemicals: comparative toxicology. Proc. Natl. Acad. Sci. U.S.A. 87:7782-7786. [PMC free article: PMC54832] [PubMed: 2217211]
  • Ames, B.N., M. Profet, and L.S. Gold. 1990. b. Dietary pesticides (99.99% all natural). Proc. Natl. Acad. Sci. U.S.A. 87:7777-7781. [PMC free article: PMC54831] [PubMed: 2217210]
  • Ames, B.N., M.K. Shigenaga, and T.M. Hagen. 1993. a. Oxidants, antioxidants, and the degenerative diseases of aging. Proc. Natl. Acad. Sci. U.S.A. 90:7915-7922. [PMC free article: PMC47258] [PubMed: 8367443]
  • Ames, B.N., L.S. Gold, and W.C. Willett. 1995. The causes and prevention of cancer. JAMA, Special Issue on Cancer. Pp. 1-9.
  • Andersen, M.E., J.J. Mills, M.L. Gargas, L. Kedderis, L.S. Birnbaum, D. Neubert, and W.F. Greenlee. 1993. Modeling receptor-mediated processes with dioxin: Implications for pharmacokinetics and risk assessment. Risk Anal. 13:25-36. [PubMed: 8383868]
  • Arnold, D.L., D. Krewski, and I.C. Munro. 1983. Saccharin: A toxicological and historical perspective. Toxicology 27:179-256. [PubMed: 6353664]
  • Bailar, A.J., and C.J. Portier. 1988. Effects of treatment-induced mortality and tumor induced mortality on tests for carcinogenicity. Biometrics 44:417-431. [PubMed: 3390507]
  • Bailar, A.J., and C.J. Portier. 1993. An index of tumorigenic potency. Biometrics (in press).
  • Bartsch, H. 1991. N-Nitroso compounds and human cancer: Where do we stand? Pp. 1-10 in Relevance to Human Cancer of N-Nitroso Compounds, Tobacco Smoke and Mycotoxins, I.K, editor; . O'Neill, J. Chen, editor; , and H. Bartsch, editor. , eds. International Agency for Research on Cancer (IARC) Scientific Publications No. 105. Lyon: IARC;
  • BATF (Bureau of Alcohol, Tobacco, and Firearms). 1987. Final cumulative information 1986-1987. Ethyl carbamate testing by all ATF laboratories.
  • BATF (Bureau of Alcohol, Tobacco, and Firearms). 1988. Cumulative information 1988. Ethyl carbamate testing by all ATF laboratories.
  • Battaglia, R., H.B. Conacher, B.D. Page. 1990. Ethyl carbamate (urethane) in alcoholic beverages and foods: A review. Food Addit. Contam. 7(4):477-496. [PubMed: 2203651]
  • Benz, R.D. and H. Irausquin. 1991. Priority-based assessment of food additives database of the U.S. Food and Drug Administration center for food safety and applied nutrition. Environ. Health Perspect. 96:85-89. [PMC free article: PMC1568244] [PubMed: 1820284]
  • Bernstein, L., L.S. Gold, B.N. Ames, M.C. Pike, and D.G. Hoel. 1985. Some tautologous aspects of the comparison of carcinogenic potency in rats and mice. Fund. Appl. Toxicol. 5:79-86. [PubMed: 3987994]
  • Berry, E.M., editor; , S.H. Blondheim, editor; , H.E. Eliahou, editor; , and E. Shafrir, editor. , editors. 1987. Recent Advances in Obesity Research. London: V. John Libby. 397 pp.
  • Biaudet, H., T. Mavelle, and G. Debry. 1994. Mean daily intake of N-nitrosodimethylamine from foods and beverages in France in 1987-1992. Food Chem. Toxicol. 32:417-421. [PubMed: 8206440]
  • Birt, D.F., H.J. Pinch, T. Barnett, A. Phan, and K. Dimitroff. 1993. Inhibition of skin tumor promotion by restriction of fat and carbohydrate calories in SENCAR mice. Cancer Res. 53:27-31. [PubMed: 8416746]
  • Bogen, K.T. 1994. Cancer potencies of heterocyclic amines found in cooked foods. Food Chem. Toxicol. 32:505-515. [PubMed: 8045455]
  • Bois, F.Y., L. Zeise, and T.N. Tozer. 1990. Precision and sensitivity of pharmacokinetic models for cancer risk assessment: Tetrachloroethylene in mice, rats, and humans. Toxicol. Appl. Pharm. 102:300-315. [PubMed: 2300971]
  • Bois, F.Y., A. Gelman, J. Jiang, and D.R. Maszle. 1994. A toxicokinetic analysis of tetrachloroethylene metabolism in humans. Technical Report. California Environmental Protection Agency, Office of Environmental Health Hazard Assessment, Berkeley, CA.
  • Bois, F.Y., G. Krowech, and L. Zeise. 1995. Modeling human interindividual variability in metabolism and risk: The example of 4-aminobiphenyl. Risk Analysis 15(2):205-213. [PubMed: 7597257]
  • Boorman, G.A., R.R. Maronpot, and S.L. Eustis. 1994. Rodent carcinogenicity bioassay: Past, present, and future. Toxicologic Pathology 22(2):105-111. [PubMed: 7973358]
  • Booth, A.N., O.H. Emerson, F.T. Jones, and F. DeEds. 1957. Urinary metabolites of caffeic and chlorogenic acids. J. Biol. Chem 229:51-59. [PubMed: 13491557]
  • Bowers, J., B. Brown, J. Springer, L. Tollefson, R. Lorentzen, and S. Henry. 1993. Risk assessment for aflatoxin: An evaluation based on the multistage model. Risk Anal. 13:637-642. [PubMed: 8310162]
  • Bressac, B., M. Kew, J. Wands, and M. Ozturk. 1991. Selective G to T mutations of p53 gene in hepatocellular carcinoma from Southern Africa. Nature 350:429-431. [PubMed: 1672732]
  • Brown, C.C., and K.C. Chu. 1989. Additive and multiplicative models and multistage carcinogenesis theory. Risk Anal. 9:99-105. [PubMed: 2710973]
  • Clinton, W.P. 1985. The chemistry of coffee. Pp. 87-92 in Proceedings of the 11th International Scientific Colloquium on Coffee, Loma, Togo. Paris: Association Scientifique International du Café.
  • Cohen, S.M., and L.B. Ellwein. 1990. Cell proliferation in carcinogenesis. Science 249:1007-1011. [PubMed: 2204108]
  • Cohen, S.M., D.T. Purtilo, and L.B. Ellwein. 1991. Pivotal role of increased cell proliferation in human carcinogenesis. Mod. Pathol. 4:371-382. [PubMed: 2068065]
  • Crouch, E., R. Wilson, and L. Zeise. 1987. Tautology or not tautology? J. Toxicol. Environ. Health 20:1-10. [PubMed: 3806695]
  • Crump, K.S. 1984. An improved procedure for low-dose carcinogenic risk assessment from animal data. J. Environ. Pathol. Toxicol. Oncol. 5:339-348. [PubMed: 6520736]
  • Czok, G., W. Walter, K. Knoche, and H. Degener. 1974. Uber die Resorbierbarkeit von Chlorogensaure durch die Ratte. Zeit. fur Ernahrungswissenschaft 13:108-112. [PubMed: 4458297]
  • Dennis, M.J., A. Burrell, K. Mathieson, P. Willetts, and R.C. Massey. 1994. The determination of the flour improver potassium bromate in bread by gas chromatographic and ICP-MS methods. Food Addit. Contam. 11(6):633-639. [PubMed: 7895868]
  • Dewanji, A., D. Krewski, and M.J. Goddard. 1993. A Weibull model for estimation of tumorigenic potency. Biometrics 49:367-377. [PubMed: 8369373]
  • DHHS (U.S. Department of Health and Human Services). 1982. Third Annual Report on Carcinogens. DHHS, Public Health Service, NTIS PB83-135855. December.
  • DHHS (U.S. Department of Health and Human Services). 1991. Sixth Annual Report on Carcinogens. DHHS, Public Health Service.
  • Doll, R. 1992. The lessons of life. Keynote address to the Nutrition and Cancer Conference. Cancer Res. (Suppl.) 52:2024s-2029s. [PubMed: 1311987]
  • Doll, R., and R. Peto. 1981. The causes of cancer: Quantitative estimates of avoidable risks of cancer in the United States today. J. Natl. Cancer Inst. 66:1191-1308. [PubMed: 7017215]
  • Dunn, A., A.G. Salmon, and L. Zeise. 1991. Urethane: Background. Pp. 4-9 in Risks of Carcinogenesis from Urethane Exposure, A.G. Salmon, editor; and L. Zeise, editor. , eds. CRC Press.
  • Elkins, E.R. 1989. Effect of commercial processing on pesticide residues in selected fruits and vegetables. J. Assoc. Off. Anal. Chem. 72(3):533-535. [PubMed: 2745385]
  • Ellen, G., E. Egmond, J.W. VanLoon, E.T. Sahertian, and K. Tolsma. 1990. Dietary intakes of some essential and non-essential trace elements, nitrate, nitrite and N-nitrosoamines, by Dutch adults: Estimated via a 24-hour duplicate portion study. Food Addit. Contam. 7(2):207-221. [PubMed: 2354740]
  • EPA (U.S. Environmental Protection Agency). 1986. a. Guidelines for carcinogen risk assessment. Fed. Reg. 51:33992-34003.
  • EPA (U.S. Environmental Protection Agency). 1986. b. Guidelines for the Health Risk Assessment for Chemical Mixtures. Federal Register 51(185):34014-34025.
  • EPA (U.S. Environmental Protection Agency). 1989. a. Ethylene bisdithiocarbamates; Notice of Preliminary Determination to Cancel Certain Registrations, Notice of Availability of Technical Support Document and Draft Notice of Intent to Cancel. Federal Register 54(243):52158-52185.
  • EPA (U.S. Environmental Protection Agency). 1989. b. Preliminary Determination to Cancel Certain Daminozide Product Registrations; Availability of Technical Support Document and Draft Notice of Intent to Cancel. Federal Register 54(99):22558-22573.
  • EPA (U.S. Environmental Protection Agency). 1992. Ethylene bisdithiocarbamates (EBDCs); Notice of Intent to Cancel; Conclusion of Special Review. Federal Register 57:7484.
  • Farmer, J.H., R.L. Kodell, and D.W. Gaylor. 1982. Estimation and extrapolation of tumor probabilities from a mouse bioassay with survival/sacrifice components. Risk Anal. 2:27-34.
  • Farrar, D., B. Allen, K. Crump, and A. Shipp. 1989. Evaluation of uncertainty in input parameters to pharmacokinetic models and the resulting uncertainty in output. Toxicol. Lett. 49:371-385. [PubMed: 2690411]
  • FDA (U.S. Food and Drug Administration). 1982. a. FDA compliance program report of findings. FY 79 total diet studies—adult (7305.002). Washington, DC: U.S. Department of health and Human Services, Food and Drug Administration. PB83-112722.
  • FDA (U.S. Food and Drug Administration). 1982. b. FDA compliance program report of findings. FY 79 total diet studies—infants and toddlers (7305.002). Washington, DC: U.S. Department of health and Human Services, Food and Drug Administration. PB82-260213.
  • FDA (U.S. Food and Drug Administration). 1985. Sensitivity of Method Carcinogen Policy. Congressional Federal Register 40(No. 211) §45530-45553, October 31, 1985.
  • FDA (U.S. Food and Drug Administration). 1993. Food Additives: Threshold of Regulation for Substances Used in Food-Contact Articles. Federal Register 58:52719-52729.
  • Finkel, A. M. 1995. Toward less misleading comparisons of uncertain risks: The example of aflatoxin and alar. Environmental Health Perspectives 103:376-385. [PMC free article: PMC1519109] [PubMed: 7607139]
  • Finkelstein, D.M., and L.M. Ryan. 1987. Estimating carcinogenic potency from a rodent tumorigenicity experiment. Appl. Stat. 36:121-133.
  • Frawley, J.P. 1967. Scientific evidence and common sense as a basis for food-packaging regulations. Food Cosmet. Toxicol. 5:293-308. [PubMed: 4861454]
  • Friedenreich, C.M., and T.E. Rohan. 1995. A review of physical activity and breast cancer. Epidemiology 6:311-317. [PubMed: 7619942]
  • Fung, V.A., J. Huff, E.K. Weisburger, and D.G. Hoel. 1993. Predictive strategies for selecting 379 NCI/NTP chemicals evaluated for carcinogenic potential: Scientific and public health impact. Fund. Appl. Toxicol. 20:413-436. [PubMed: 8314458]
  • Fürst, P., C. Fürst, and K. Wilmers. 1991. Body burden with PCDD and PCDF from food. Pp. 133-142 in Banbury Report 35: Biological Basis for Risk Assessment of Dioxins and Related Compounds, M.A. Gallo, editor; , R.J. Scheuplein, editor; , and C.A. van der Heijden, editor. , eds. Cold Spring Harbour Laboratory Press.
  • Gallo, M.A., R. Scheuplein, and C.A. van der Heijden. 1991. Banbury Report 35: Biological Basis for Risk Assessment of Dioxins and Related Compounds. Cold Spring Harbour Laboratory Press. Pp.179-212.
  • Gaylor, D.W. 1989. Preliminary estimates of the virtually safe dose for tumors obtained from the maximum tolerated dose. Reg. Toxicol. Pharmacol. 9:101-108. [PubMed: 2717786]
  • Gaylor, D.W., and J.J. Chen. 1986. Relative potency of chemical carcinogens in rodents. Risk Anal. 6(3):283-290. [PubMed: 3602501]
  • Gaylor, D.W., and F.F. Kadlubar. 1991. Quantitative cancer risk assessment of heterocyclic amines in cooked foods. Pp. 229-236 in Mutagens in Food: Detection and Prevention, H. Hayatsu, editor. , ed. Boca Raton, FL: CRC Press, Inc.
  • Gaylor, D.W., J.J. Chen, and D.M. Sheehan. 1993. Uncertainty in cancer risk estimates. Risk Anal. 13:149-154. [PubMed: 8502788]
  • Gaylor, D.W., and L.S. Gold. 1995. Quick estimates of the regulatory virtually safe dose based on the maximum tolerated dose for rodent bioassays. Regul. Toxicol. Pharmacol. 22:57-63. [PubMed: 7494904]
  • Gaylor, D.W., R.L. Kodell, J.J. Chen, J.A. Springer, R.J. Lorentzen, and R.J. Scheuplein. 1994. Point estimates of cancer risk at low doses. Risk Analysis 14:843-850. [PubMed: 7800868]
  • Gibb, H.J., and C.W. Chen. 1986. Multistage model interpretation of additive and multiplicative carcinogenic effects. Risk Anal. 6:167-170. [PubMed: 3615986]
  • Giovannucci, E., E.B. Rimm, G.A. Colditz, M.J. Stampfer, A. Ascherio, C.C. Chute, and W.C. Willett. 1993. A prospective study of dietary fat and risk of prostate cancer. J. Natl. Cancer Inst. 85:1571-1579. [PubMed: 8105097]
  • Goddard, M.J., and D. Krewski. 1995. The future of mechanistic research in risk assessment: Where are we going and can we get there from here? Toxicology. 102:53-70. [PubMed: 7482562]
  • Goddard, M.J., D.J. Murdoch, and D. Krewski. 1995. Temporal aspects of risk characterization. Inhalation Toxicology 7:1005-1018.
  • Gold, L.S., L. Bernstein and B.N. Ames. 1990. The importance of ranking possible carcinogenic hazards using HERP. Risk Anal. 10:625-33. [PubMed: 2287789]
  • Gold, L.S., C.B. Sawyer, R. Magaw, G.M. Backman, M. de Veciana, R. Levinson, N.K. Hooper, W. Havender, L. Bernstein, R. Peto, M.C. Pike, and B.N. Ames. 1984. A carcinogenic potency database of the standardized results of animal bioassays. Environ. Health Perspect. 58:9-319. [PMC free article: PMC1569423] [PubMed: 6525996]
  • Gold, L.S., T.H. Slone, G.M. Backman, S. Eisenberg, M. Da Costa, M. Wong, N.B. Manley, L. Rohrback, and B.N. Ames. 1990. Third chronological supplement to the Carcinogenic Potency Database: Standardized results of animal bioassays published through December 1986 and by the National Toxicology Program through June 1987. Environ. Health Perspect. 84:215-285. [PMC free article: PMC1567635] [PubMed: 2351123]
  • Gold, L.S., T.H. Slone, B.R. Stern, N.B. Manley, and B.N. Ames. 1992. Rodent carcinogens: Setting priorities. Science 258:261-265. [PubMed: 1411524]
  • Gold, L.S., N.B. Manley, T.H. Slone, G.B. Garfinkel, L. Rohrbach, and B.N. Ames. 1993. a. The fifth plot of the Carcinogenic Potency Database: Results of animal bioassays published in the general literature through 1988 and by the National Toxicology Program through 1989. Environ. Health Perspectives 100:65-135. [PMC free article: PMC1519579] [PubMed: 8354183]
  • Gold, L.S., T.H. Slone, B.R. Stern, N.B. Manley, and B.N. Ames. 1993. b. Possible carcinogenic hazards from natural and synthetic chemicals: Setting priorities in Comparative Environmental Risk Assessment, C.R. Cothern, editor. , ed. Boca Raton: Lewis Publishers.
  • Gold, L.S., T.H. Slone, N.B. Manley, and B.N. Ames. 1994. Heterocyclic amines formed by cooking food: Comparison of bioassay results with other chemicals in the Carcinogenic Potency Database. Cancer Letters 83:21-29. [PubMed: 8062218]
  • Goldbohm, R.A., P.A. van den Brandt, P. van't veer, H.A.M. Brants, E. Dorant, F. Sturmans, and R.J.J. Hermus. 1994. A prospective cohort study on the relation between meat consumption and the risk of colon cancer. Caner Res. 54:718-723. [PubMed: 8306333]
  • Goodman, M.T., L.N. Kolonel, C.N. Yoshizawa, and J.H. Hankin. 1988. The effect of dietary cholesterol and fat on the risk of lung cancer in Hawaii. Am. J. Epidemiol. 128:1241-1255. [PubMed: 3195565]
  • Goodman, G., and R. Wilson. 1992. Comparison of the dependence of the TD50 on maximum tolerated dose for mutagens and nonmutagens. Risk Anal. 12:525-534. [PubMed: 1480798]
  • Government Accounting Office (GAO). 1991. Pesticides: Food Consumption Data of Little Value to Estimate Some Exposures. GAO/RCED-91-125. U.S. General Accounting Office, Resources, Community and Economic Development Division, Washington, DC.
  • Harborne, J.B. 1993. Introduction to Ecological Biochemistry, 4th edition. London: Academic Press. 318pp.
  • Haseman, J.K., and S.K. Seilkop. 1992. An examination of the association between maximum-tolerated dose and carcinogenicity in 326 long-term studies in rats and mice. Fund. Appl. Toxicology 19:207-213. [PubMed: 1516777]
  • Henderson, B.E., J.T. Casagrande, M.C. Pike, T. Mack, and T. Rosario. 1983. The epidemiology of endometrial cancer in young women. Br. J. Cancer 47:749-756. [PMC free article: PMC2011352] [PubMed: 6860544]
  • Herrmann, K. 1978. Ubersicht uber nichtessentielle inhaltsstoffe der germusearten III. Mohren, Sellerie, Pastinaken, Rote Ruben, Spinat, Salat, Endivien, Trieibzichorie, Rhabarber und Artischock. Z Lebensm Unter.-Forsch 167:262-273. [PubMed: 716640]
  • Herrmann, K. 1989. Occurrence and content of hydroxycinnamic and hydroxybenzoic acid compounds in foods. Crit. Rev. Fd. Sci. Nutri. 28(4):315-347. [PubMed: 2690858]
  • Hetrick, D.M., A.M. Jarabek, and C.C. Travis. 1991. Sensitivity analysis for physiologically based pharmacokinetic models. J. Pharmacokinet. Biopharm. 19:1-20. [PubMed: 2023106]
  • Higginson, J. 1988. Changing concepts in cancer prevention: Limitations and implications for future research in environmental carcinogenesis. Cancer Research 48:1381-1389. [PubMed: 3278801]
  • Hoel, D. 1990. Assumptions of the HERP index. Risk Anal. 10(4):623-624. [PubMed: 2287788]
  • Hoel, D.G., J.K. Haseman, M.D. Hogan, J. Huff, and E.E. McConnell. 1988. The impact of toxicity on carcinogenicity studies: Implications for risk assessment. Carcinogenesis 9:2045-2052. [PubMed: 3052903]
  • Hoseyni, M.S. 1992. Risk assessment for aflatoxin: III. Modeling the relative risk of hepatocellular carcinoma. Risk Anal. 12:123-128. [PubMed: 1315449]
  • Howe, G.R., T. Hirohata, T.G. Hislop, J.M. Iscovich, J.-M. Yuan, K. Katsouyanni, F. Lulian, E. Marulini, B. Moda, T. Rohan, P. Tonialo, and Y. Shunghang. 1990. Dietary factors and risk of breast cancer: Combined analysis of 12 case-control studies. J. Natl. Cancer Inst. 82:561-569. [PubMed: 2156081]
  • Hsu, I.C., R.A. Metcalf, T. Sun, J.A. Welsh, N.J. Wang, and C.C. Harris. 1991. Mutational hotspot in the p53 gene in human hepatocellular carcinomas. Nature 350:427-428. [PubMed: 1849234]
  • Hunter, D.J., and W.C. Willett. 1993. Diet, body size, and breast cancer. Epidemiol. Rev. 15(1):110-132. [PubMed: 8405195]
  • Idle, J.R., M. Armstrong, A.V. Boddy, C. Boustead, S. Cholerton, J. Cooper, A.K. Daly, J. Ellis, W. Gregory, H. Hadidi, C. Hofer, J. Holt, J. Leathart, N. McCracken, S.C. Monkman, J.E. Painter, H. Taber, D. Walker, and M. Yule. 1992. The pharmacogenetics of chemical carcinogenesis. Pharmacogen. 2:246-258.
  • ILSI (International Life Sciences Institute). 1993. Approaches for Establishing A Safe Intake Level for Ingested Acetaldehyde. Washington, D.C.: ILSI North America.
  • Inoue, S., K. Ito, K. Yamamoto, and S. Kawanishi. 1992. Caffeic acid causes metal-dependent damage to cellular and isolated DNA through H202 formation. Carcinogenesis 13:1497-502. [PubMed: 1394830]
  • IARC (International Agency for Research on Cancer). 1976. IARC Monographs on the Evaluation of Carcinogenic Risk of Chemicals to Man. Pp. 51-72 in Some Naturally Occurring Substances, vol. 10. Lyon, France: IARC.
  • IARC (International Agency for Research on Cancer). 1982. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Volume 29: Some Industrial Chemicals and Dyestuffs. WHO, IARC. 416 pp.
  • IARC (International Agency for Research on Cancer). 1983. Laboratory decontamination and destruction of carcinogens in laboratory wastes: some polycyclic aromatic hydrocarbons. IARC Scientific Publications No. 49. [PubMed: 6305830]
  • IARC (International Agency for Research on Cancer). 1992. Mechanisms of Carcinogenesis in Risk Identification, H. Vainio, editor; , P.N. Magee, editor; , D.B. McGregor, editor; , and A.J. McMichael, editor. editors. WHO, IARC. pp. 579-599 in IARC Scientific Publications No. 116.
  • IARC (International Agency for Research on Cancer). 1993. IARC Monographs on the Evaluation of Carcinogenic Risks to Humans. Lists of IARC Evaluations. WHO, IARC, Lyon. Vol. 56.
  • IFBC (International Food Biotechnology Council). 1990. Toxicol. Pharmacol. 12:526-528.
  • Ito, N., M. Hirose, and S. Takahashi. 1991. Cellular proliferation and stomach carcinogenesis induced by antioxidants. Pp. 43-52 in Chemically Induced Cell Proliferation: Implications for Risk Assessment., eds. Wiley Liss, editor. .
  • Jackel, S.S. 1994. Update on potassium bromate. Cereal Food World 39(10):772.
  • Kodell, R.L., and D.W. Gaylor. 1989. On the additive and multiplicative models of relative risk. Biom. J. 31:359-370.
  • Kodell, R.L., D.W. Gaylor, and J.J. Chen. 1990. Carcinogenic potency correlations: Real or artifactual? J. Toxicol. Environ. Health 32:1-9. [PubMed: 1987359]
  • Kodell, R.L., D. Krewski, and J.M. Zielinski. 1991. Additive and multiplicative relative risk in the two-stage clonal expansion model of carcinogenesis. Risk Anal. 11:483-490. [PubMed: 1947353]
  • Kohn, M.C., and C.J. Portier. 1993. Effects of the mechanism of receptor-mediated gene expression on the shape of the dose-response curve. Risk Anal. 13(5):565-572. [PubMed: 8259447]
  • Kolonel, L.N., A.M.Y. Nomura, M.W. Hinds, T. Hirohata, J.H. Hankin, and J. Lee. 1983. Role of diet in cancer incidence in Hawaii. Cancer Res. 43:2397s-2402s. [PubMed: 6831463]
  • Krewski, D. 1990. Measuring carcinogenic potency. Risk Analysis 10:615-617. [PubMed: 2287787]
  • Krewski, D., and R.D. Thomas. 1992. Carcinogenic mixtures. Risk Anal. 12:105-113. [PubMed: 1574610]
  • Krewski, D., D. Murdoch, and J.R. Withey. 1989. Recent developments in carcinogenic risk assessment (with discussion). Health Phys. 57 (Supplement 1):313-325. [PubMed: 2691452]
  • Krewski, D., M.J. Goddard, and J.R. Withey. 1990. Carcinogenic potency and interspecies extrapolation. Pp. 323-334 in Cancer and the Environment, Pard D: Carcinogenesis. M.L. Mendelson, editor; and R.J. Albertini, editor. , eds. New York: Wiley-Liss.
  • Krewski, D., M. Szyszkowicz, and H. Rosenkranz. 1990. Quantitative factors in chemical carcinogenesis: variation in carcinogenic potency. Regul. Toxicol. Pharmacol. 12:13-29. [PubMed: 2217915]
  • Krewski, D., D. Gaylor, and M. Szyszkowicz. 1991. A model-free approach to low-dose extrapolation. Environ. Health Perspect. 90:279-285. [PMC free article: PMC1519485] [PubMed: 2050073]
  • Krewski, D., M.J. Goddard, and J.M. Zielinski. 1992. Dose-response relationships in carcinogenesis. Pp. 579-599 in Mechanisms of Carcinogenesis in Risk Identification, H. Vainio, editor; , P.N. Magee, editor; , D.B. McGregor, editor; , and A.J. McMichael, editor. , eds. A.J. Lyon: IARC.
  • Krewski, D., D.W. Gaylor, A.P. Soms, and M. Syzszkowicz. 1993. Correlation between carcinogenic potency and the maximum tolerated dose: Implications for risk assessment. Pp. 111-171 in Issues in Risk Assessment. Washington, DC: National Academy Press.
  • Krewski, D., J.R. Withey, L.-F. Ku, and M.E. Andersen. 1994. Applications of physiologic pharmacokinetic modeling in carcinogenic risk assessment. Environmental Health Perspectives 102(Suppl. 11):37-50. [PMC free article: PMC1566756] [PubMed: 7737040]
  • Krewski, D., Y. Wang, S. Bartlett, and K. Krishnan. 1995. Uncertainty, variability, and sensitivity analysis in physiologic pharmacokinetic models. Journal of Biopharmaceutical Statistics in press. [PubMed: 8580927]
  • Layton, D.W., K.T. Bogen, M.G. Knize, F.T. Hatch, V.M. Johnson, and J.S. Felton. 1995. Cancer risk of heterocyclic amines in cooked foods: An analysis and implications for research. Carcinogenesis 16(1):39-52. [PubMed: 7834804]
  • Lee, I.M., R.S. Paffenbarger, Jr., and C. Hsieh. 1991. Physical activity and risk of developing colorectal cancer among college alumni. J. Natl. Cancer Inst. 83:1324-1329. [PubMed: 1886158]
  • Lee, I.M., R.S. Paffenbarger, and E.C. Hseieh. 1992. Physical activity and risk of prostate cancer among college alumni. Am. J. Epidemiol. 135:169-179. [PubMed: 1536133]
  • Le Marchand, L., L.N. Kolonel, L.R. Wilkens, B.C. Myers, and T. Hirohata. 1994. Animal fat consumption and prostate cancer: A prospective study in Hawaii. Epidemiology 5:276-282. [PubMed: 8038241]
  • Levander, O.A. 1987. A global view of human selenium nutrition. Annu. Rev. Nutr. 7:227-250. [PubMed: 3300734]
  • Littlefield, N.A., J.H. Farmer, D.W. Gaylor, and W.G. Sheldon. 1980. Effects of dose and time in a long-term, low-dose carcinogenic study. J. Environ. Pathol. Toxicol. 3:17-34. [PubMed: 7365379]
  • Lutz, W.K., and J. Schlatter. 1992. Chemical carcinogens and overnutrition in diet-related cancer. Carcinogenesis 13:2211-2216. [PubMed: 1473226]
  • Lyon, J.L., J.W. Gardner, and D.W. West. 1980. Cancer risk and lifestyle: Cancer among Mormons (1967-1975). Pp. 273-290 in Genetic and Environmental Factors in Experimental and Human Cancer, H.V. Gelboin, editor; , B. MacMahon, editor; , T. Matsuhara, editor; , T. Sugimura, editor; , S. Takayama, editor; , and H. Takebe, editor. , eds. Tokyo: Japanese Scientific Society Press.
  • Massey, R., M.J. Dennis, M. Pointer, and P.E. Key. 1990. An investigation of the levels of N-nitrosodimethylamine, apparent total N-nitroso compounds, and nitrate in beer. Food Addit. Contam. 7(5):605-615. [PubMed: 2253805]
  • McCann, J., L.S. Gold, L. Horn, R. McGill, T.E. Graedel, and J. Kaldor. 1988. Statistical analysis of Salmonella test data and comparison to results of animal cancer tests. Mutat. Res. 205:183-195. [PubMed: 3285186]
  • McGregor, D.B. 1992. Chemicals classified by IARC: their potency in tests for carcinogenicity in rodents and their genotoxicity and acute toxicity. Pp. 323-352 in Mechanisms of Carcinogenesis in Risk Identification. H. Vainio, editor; , P.N. Magee, editor; , D.B. McGregor, editor; , and A.J. McMichael, editor. , eds. International Agency for Research on Cancer. Lyon, France. [PubMed: 1428089]
  • Metzger, B., E. Crouch, and R. Wilson. 1989. On the relationship between carcinogenicity and acute toxicity. Risk Anal. 9:169-177. [PubMed: 2762603]
  • Miller, J.D. 1994. Mycotoxin Overview. Presented at the Toxicology Forum Special Meeting on Mycotoxin Health Risk, Control and Regulation. Capital Hilton, Washington DC, February 23-24.
  • Moolgavkar, S.H., and G. Luebeck. 1990. Two-event model for carcinogenesis: Biological, mathematical, and statistical considerations. Risk Anal. 10(2):323-341. [PubMed: 2195604]
  • Moolgavkar, S.H., E.G. Luebeck, D. Krewski, and J.M. Zielinski. 1993. Radon cigarette smoke, and lung cancer: A re-analysis of the Colorado Plateau uranium miners'data. Epidemiology 4(3):204-217. [PubMed: 8512985]
  • Morgan, K.J., V.J. Stults, and M.E. Zabik. 1982. Amount and dietary sources of caffeine and saccharin intake by individuals ages 5 to 18 years. Regul. Toxicol. Pharmacol. 2:296-307. [PubMed: 7185096]
  • Morgan, M.G., and M. Henrion. 1990. Uncertainty: A Guide to Dealing with Uncertainty in Quantitative Risk and Policy Analysis. Cambridge: Cambridge University Press.
  • Munro, I.C. 1990. Safety assessment procedures for indirect food additives: an overview. Regul. Toxicol. Pharmacol. 12:2-12. [PubMed: 2217916]
  • Murdoch, D.J., D. Krewski, and J. Wargo. 1992. Cancer risk assessment with intermittent exposure. Risk Analysis 12:569-577. [PubMed: 1480801]
  • National Institute on Alcohol Abuse and Alcoholism. 1988. Alcohol Epidemiologic Data System. Surveillance Report #10. Apparent per capita alcohol consumption: National, state, and regional trends, 1977-1986. U.S. Department of Health and Human Services.
  • NRC (National Research Council). 1978. Saccharin: Technical Assessment of Risks and Benefits. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1979. a. Use and Intake of Food and Color Additives. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1979. b. The 1977 Survey of Industry on the Use of Food Additives. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1981. The Health Effects of Nitrate, Nitrite, and N-Nitroso Compounds. National Academy of Sciences. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1982. Diet, Nutrition and Cancer. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1987. Pharmacokinetics in Risk Assessment. Pp. 441-468 in Drinking Water and Health, Volume 8, Washington, DC: National Academy Press.
  • NRC (National Research Council). 1988. Complex Mixtures: Methods for In Vivo Toxicity Testing. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1989. b. Drinking Water and Health. Vol. 9. Selected Issues in Risk Assessment. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1989. a. Diet and Health: Implications for Reducing Chronic Disease Risk. Food and Nutrition Board, Committee on Diet and Health. Washington, DC.: National Academy Press.
  • NRC (National Research Council). 1993. b. Issues in Risk Assessement: Use of the Maximum Tolerated Dose in Animal Bioassays for Carcinogenicity. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1993. a. Pesticides in the Diets of Infants and Children. Washington, DC: National Academy Press.
  • NRC (National Research Council). 1994. Science and Judgment in Risk Assessment. Washington, DC: National Academy Press.
  • National Toxicology Program (NTP). 1995. DRAFT NTP Technical Report on the Effect of Dietary Restriction on Toxicology and Carcinogenesis Studies in F344/N Rats and B6C3F1 Mice. NTIP Technical Report Series No. 460. NIH Publication No. 95-3376. U.S. Department of Health and Human Services, Public Health Service, National Institutes of Health, Research Triangle Park, NC.
  • Nomura, A., and L.N. Kolonel. 1991. Prostate cancer: A current perspective. Epidemiol. Rev. 13:200-227. [PubMed: 1765112]
  • Pariza, M.W., and R.K. Boutwell. 1987. Historical perspective: Calories and energy expenditure in carcinogenesis. Amer. J. Clin. Nutrit. 45(Suppl):151-156. [PubMed: 3541566]
  • Parodi, S., M. Taningher, P. Boero, and L. Santi. 1982. Quantitative correlations amongst alkaline DNA fragmentation, DNA covalent binding, mutagenicity in the Ames test and carcinogenicity, for 21 compounds. Mutat. Res. 93:1-24. [PubMed: 7038465]
  • Pennington, J.A., B.E. Young, and D.B. Wilson. 1989. Nutritional elements in U.S. diets: Results from the Total Diet Study, 1982 to 1986. J. Am. Diet Assoc. 89:639-644. [PubMed: 2723289]
  • Perera, F., and P. Boffetta. 1988. Perspectives on comparing risks of environmental carcinogens. JNCI 80:1282-1293. [PubMed: 3050138]
  • Peters, R.K., D.H. Garabrant, M.C. Yu, and T.M. Mack. 1989. A case-control study of occupational and dietary factors in colorectal cancer in young men by subsite. Cancer Res. 49:5459-5468. [PubMed: 2766308]
  • Peto, R., M.C. Pike, N.E. Day, R.G. Gray, P.N. Lee, S. Parish, J. Peto, S. Richards, and J. Wharendorf. 1980. Guidelines for simple, sensitive significance tests for carcinogenic effects in long-term animal experiments. Pp. 311-426 in IARC Monographs on the Evaluation of the Carcinogenic Risk of Chemicals to Humans: Supplement of Long-Term and Short-Term Screening Assays for Carcinogens: A Critical Appraisal. International Agency for Research on Cancer. Lyon, France. [PubMed: 6935185]
  • Peto, R., M. Pike, L. Bernstein, L.S. Gold, and B. Ames. 1984. a. The TD50: A proposed general convention for the numerical description of the carcinogenic potency of chemicals in chronic exposure animal experiments. Environ. Health Perspect. 58:1-9. [PMC free article: PMC1569426] [PubMed: 6525991]
  • Peto, R., R. Gray, P. Brantom., and P. Grasso. 1984. b. Nitrosamine carcinogenesis in 5120 rodents: Chronic administration of sixteen different concentrations of NDEA, NDMA, NPYR, and NPIP in the water of 4400 inbred rats with parallel studies on NDEA alone of the effect of age of starting (8, 6, or 20 weeks) and of species (rats, mice hamsters). Pp. 617-665 in N-Nitroso Compounds: Occurrence, Biological Effects and Relevance to Human Cancer, I.K. O'Neill, editor; , R.G. Borstell, editor; , C.T. Miller, editor; , J. Long, editor; , and H. Bartsch, editor. , eds. IARC Scientific Publications No. 57. Lyon France: IARC. [PubMed: 6533057]
  • Piegorsch, W.W., and D.G. Hoel. 1988. Exploring relationships between mutagenic and carcinogenic potencies. Mutat. Res. 196:161-175. [PubMed: 3419441]
  • Pignatelli, B., Malaveille, C. Chen, A. Hautefeuille, P. Thuillier, N. Muñoz, B. Moulinier, F. Berger, H. DeMontclos, H. Oshima, R. Lambert, and H. Bartsch. 1991. N-Nitroso compounds, genotoxins and their precursors in gastric juice from humans with and without precancerous lesions of the stomach. In Relevance to Human Cancer of N-Nitroso Compounds, Tobacco Smoke and Mycotoxins, I.K. O'Neill, editor; , J. Chen, editor; , H. Bartsch, editor. , eds. International Agency for Research on Cancer (IARC) Scientific Publication No. 105. Lyon: IARC. [PubMed: 1855844]
  • Pohland, A.E. 1994. Worldwide Occurrence of Fumonisins. Proceedings of Toxicology Forum Special Meeting on Mycotoxin Health Risk, Control, and Regulation, Feb. 23-24, 1994. Washington, DC: Toxicology Forum.
  • Portier, C.J., and D.G. Hoel. 1987. Issues concerning the estimation of TD50. Risk Anal. 7:437-449. [PubMed: 3444931]
  • Portier, C.J., and N.L. Kaplan. 1989. Variability of safe dose estimates when using complicated models of the carcinogenic process. A case study: Methylene chloride. Fund. Appl. Toxicol. 13:533-544. [PubMed: 2612786]
  • Potter, J.D., M.L. Slattery, R.M. Bostick, and S.M. Gapstur. 1993. Colon cancer: A review of the epidemiology. Epidemiol. Rev. 15(2):499-545. [PubMed: 8174669]
  • Prentice, R.L., and L. Sheppard. 1990. Dietary fat and cancer: Consistency of the epidemiologic data, and disease prevention that may follow from a practical reduction in fat consumption. Cancer Causes and Control 1:81. [PubMed: 2102280]
  • Preussmann, R. 1984. Occurrence nad exposure to N-nitroso compounds and precursors. Pp. N-Nitroso Compounds: Occurrence, Biological Effects and Relevance to Human Cancer, I.K. O'Neill, editor; , R.C. von Borstel, editor; , C.T. Miller, editor; , J. Long, editor; , and H. Bartsch, editor. , eds. International Agency for Research on Cancer (IARC) Scientific Publication No. 57. Lyon: IARC.
  • Ranum, P. 1992. Potassium bromate in bread baking. Creal Food World (37(3):253-258.
  • Reith, J.P., and T.B. Starr. 1989. Experimental desing constraints on carcinogenic potency estimates. J. Toxicol. Environ. Health 27:287-296. [PubMed: 2754755]
  • Reitz, R.H., P.G. Watanabe, M.J. McKenna, J.F. Quast, and P.J. Gehring. 1980. Effects of vinylidene chloride on DNA synthesis and DNA repair in the rat and mouse: A comparative study with dimethylnitrosamine. Toxicol. Appl. Pharmacol. 52:357-370. [PubMed: 7368211]
  • Reitz, R.H., A.L. Mendrala, R.A. Corley, J.F. Quast, M.L. Gargas, M.E. Andersen, D.A. Staats, and R.B. Connolly. 1990. Estimating the risk of liver cancer associated with human exposures to chloroform using physiologically-based pharmacokinetic modeling. Toxicol. Appl. Pharmacol. 105:443-459. [PubMed: 2237918]
  • Risch, B., and K. Herrmann. 1988. Hydroxyzimtsäure-Verbindungen in citrus-früchten. Z Lebensm Unters.-Forsch 187:530-534.
  • Risch, H.A., M. Jain, L. Marrett, and G.R. Howe. 1994. Dietary fat intake and risk of epithelial ovarian cancer. J. Natl. Cancer Inst. 86:1409-1415. [PubMed: 8072035]
  • Roe, F.J.C. 1988. How do hormones cause cancer? Pp. in Theories of Carcinogenesis. O. H. Iversen, editor. , ed. New York: Hemisphere Publishing Corporation.
  • Rose, D.P., A.P. Boyar, and E.L. Wynder. 1986. International comparisons of mortality rates for cancer of the breast, ovary, prostate, and colon, and per capita food consumption. Cancer. 58:2363. [PubMed: 3768832]
  • Rosenkranz, H.S. 1992. Structure-activity relationships for carcinogens with different modes of action. Pp. 271-277 in Mechanisms of Carcinogenesis in Risk Identification, H. Vainio, editor; , P.N. Magee, editor; , D.B. McGregor, editor; , and A.J. McMichael, editor. , eds. International Agency for Research on Cancer Scientific Publications No. 116, Lyon.
  • Rulis, A.M. 1986. De minimis risk and the threshold of regulation. Pp. 29-37 in Food Protection Technology, C.W. Felix, editor. , ed. Chelsea, MI: Lewis Publishers.
  • Rulis, A.M. 1989. Establishing a threshold of regulation. Pp. 271-278 in Risk Assessment in Setting National Priorities, J.J. Bonin, editor; , and D.E. Stevenson, editor. , eds. New York: Plenum.
  • Rulis, A.M. 1992. Threshold of regulation: Options for handling minimal risk situations. Pp. 132-139 in Food Safety Assessment. ACS Sympos. Series 484.
  • Sawyer, C., R. Peto, L. Bernstein, and M. Pike. 1984. Calculation of carcinogenic potency from long-term animal experiments. Biometrics 40:27-40. [PubMed: 6733232]
  • Scanlan, R.A. 1983. Formation and occurrence of nitrosamines in food. Cancer Research 43:2435-2440. [PubMed: 6831466]
  • Scanlan, R.A., and J.F. Barbour. 1991. N-Nitrosodimethylamine content of US and Canadian beer. In Relevance to Human Cancer of N-Nitroso Compounds, Tobacco Smoke and Mycotoxins, I.K. O'Neill, editor; , J. Chen, editor; , and H. Bartsch, editor. , eds. International Agency for Research on Cancer (IARC) Scientific Publication No. 105. Lyon: IARC.
  • Scheuplein, R.J. 1990. Perspectives on toxicological risk—an example: Food-borne carcinogenic risk. Pp. 351-371 in Progress in Predictive Toxicology. D.B. Clayson, editor; , I.C. Munro, editor; , P. Shubik, editor; , and J.A. Swenberg, editor. , eds. Elsevier Science Publ., B.V.
  • Schrauzer, G.N., and D.A. White. 1978. Selenium in human nutrition: Dietary intakes and effects of supplementation. Bioinorg. Chem. 8:303-318. [PubMed: 647060]
  • Schubert, A., M. M. Holden, and W.R. Wolf. 1987. Selenium content of a core group of foods based on a critical evaluation of published analytical data. J. Am. Diet Assoc. 87:285-299. [PubMed: 3819246]
  • Shibamoto, T., and L.F. Bjeldanes. 1993. Introduction to Food Toxicology. New York: Academic Press.
  • Simonich, S.L., and R.A. Hites. 1994. Importance of vegetation in removing polycyclic aromatic hydrocarbons from the atmosphere. Nature 370:49-51.
  • Springer, J. 1994. Risk Assessment: FDA's most recent risk assessment. Presented at the Toxicology Forum Special Meeting on Mycotoxin Health Risk, Control and Regulation. Capital Hilton, Washington DC, February 23-24.
  • Steinmetz, K.A., and J.D. Potter. 1991. Vegetables, fruit, and cancer. I. Epidemiology (Review). Cancer Causes Control 5:325-357. [PubMed: 1834240]
  • Stich, H.F. 1991. The beneficial and hazardous effects of simple phenolic compounds. Mutation Research 259:307-324. [PubMed: 2017215]
  • Stich, H.F. 1992. Teas and tea components as inhibitors of carcinogen formation in model systems and man. Preventive Medicine 21:377-84. [PubMed: 1614999]
  • Stolz, D.R., B. Stavric, R. Stapley, R. Klaassen, R. Bendall, and D. Krewski. 1984. Mutagenicity screening of foods. II. Results with fruits and vegetables. Environ. Mut. 6:343-354. [PubMed: 6376087]
  • Sugimura, T. 1985. Carcinogenicity of mutagenic heterocyclic amines formed during the cooking process. Mut. Res. 150:33-41. [PubMed: 3889618]
  • Sugimura, T. 1986. Studies on environmental chemical carcinogenesis in Japan. Science 233:312-318. [PubMed: 3088728]
  • Sweet, D.V., editor. , ed. 1987. Registry of Toxic Effects of Chemical Substances (RTECS). 1986 Edition. National Center for Occupational Safety and Health. Washington, D.C.
  • Takayama, S., Y. Nakatsure, M. Masuda, H. Ohgaki, S. Sato, and T. Sugimura. 1984. Demonstration of carcinogenicity in F344 rats of 2-amino-3-methylimidazo[4,5-f]quinoline from broiled sardine, fried beef and beef extract. Gann 75:467-570. [PubMed: 6468834]
  • Tanaka, T., W.S. Barnes, G.M. Williams, and J.H. Weisburger. 1985. Multipotential carcinogenicity of the fried food mutagen 2-1mino-3-methylimidazo[4,5-f]quinoline in rats. Jpn. J. Cancer Res. 76:570-576. [PubMed: 3928552]
  • Tanaka, T., T. Kojima, T. Kawamor, A. Wang, M. Suzui, K. Okamoto, and H. Mori. 1993. Inhibition of 4-nitroquinoline-1-oxide-induced rat tongue carcinogenesis by the naturally occurring plant phenolics caffeic, ellagic, chlorogenic and ferulic acids. Carcinogenesis 14:1321-5. [PubMed: 8330344]
  • Technical Assessment Systems. 1995. a. Exposure®. Detailed Distributional Dietary Exposure Analysis. Washington, D.C.: Technical Assessment Systems, Inc.
  • Technical Assessment Systems. 1995. b. TASDIET®. International Diet Research System. Washington, D.C.: Technical Assessment Systems, Inc.
  • Tennant, R.W., M.R. Elwell, J.W. Spalding, and R.A. Griesemer. 1991. Evidence that toxic injury is not always associated with induction of chemical carcinogenesis. Molecular Carcinogenesis 4:420-440. [PubMed: 1793481]
  • Toda, S., M. Kumura, and M. Ohnishi. 1991. Effects of phenolcarboxylic acids on superoxide anion and lipid peroxidation induced by superoxide anion. Planta Medica 57:8-10. [PubMed: 1648246]
  • Travis, C.C., and R.K. White. 1988. Interspecies scaling of carcinogenic potency. Risk Analysis 8:119-125. [PubMed: 3375501]
  • Travis, C.C., S.A. Richter Pack, A.W. Saulsbury, and M.W. Yambert. 1990. a. Prediction of carcinogenic potency from toxicology data. Mutat. Res. 241:21-36. [PubMed: 2333083]
  • Travis, C.C., A.W. Saulsbury, and S.A. Richter Pack. 1990. b. Prediction of cancer potency using a battery of mutation and toxicity data. Mutagenesis 5:213-219. [PubMed: 2385175]
  • Travis, C.C., L.A. Wang, and M.J. Waehner. 1991. Quantitative correlation of carcinogenic potency with four different classes of short-term test data. Mutagenesis 6:353-360. [PubMed: 1795638]
  • Tricker, A.R., B. Pfundstein, E. Theobald, R. Preussmann, and B. Spiegelhalder. 1991. Mean daily intake of volatile N-nitrosoamines from foods and beverages in West Germany in 1989-1990. Food Chem. Toxicol. 29:729-732. [PubMed: 1761251]
  • Turnbull, G.J., P.N. Lee, and F.J.C. Roe. 1985. Relationship of body weight to longevity and to risk of development of nephropathy and neoplasia in Sprague-Dawley rats. Food Chem. Toxicol. 23:355-362. [PubMed: 2989133]
  • Turturro, A., P.H. Duffy, and R.W. Hart. 1993. Modulation of toxicity by diet and dietary macronutrient restriction. Mutation Research 295:151-164. [PubMed: 7507554]
  • USDA (U.S. Department of Agriculture). 1987). Nationwide Food Consumption Survey. Continuing Survey of Food Intake of Individuals. Women 19-50 years and their children 1-5 years, 4 days, 1985. Report No. 85-4. Nutrition Monitoring Division, Human Nutrition Information Service. Hyattsville, MD:
  • USDA (U.S. Department of Agriculture). 1992. Nutrient Database for Standard Reference. Release 10 Tape. Human Nutrition Information Service. Hyattsville, MD:
  • Vainio, H., M. Sorsa, and A.J. McMichael. 1990. Complex Mixtures and Cancer Risk. IARC Sci. Publ. No. 104. International Agency for Research on Cancer. Lyon, France:
  • Van Ryzin, J. 1980. Quantitative risk assessment. J. Occup. Med. 22(5):321-326. [PubMed: 7381612]
  • Van't Veer, P., F.J. Kok, H.A.M. Brants, T. Ockhuizen, F. Sturmans, and R.J.J. Hermus. 1990. Dietary fat and the risk of breast cancer. Int. J. Epidemiol. 19:12-18. [PubMed: 2351506]
  • Watanabe, K., F.Y. Bois, and L. Zeise. 1992. Interspecies extrapolation: A reexamination of acute toxicity data. Risk Analysis 12(2):301-310. [PubMed: 1502377]
  • Wartenberg, D., and M.A. Gallo. 1990. a. The fallacy of ranking possible carcinogenic hazards using the TD50. Risk Analysis 10(4):609-613. [PubMed: 2287786]
  • Wartenberg, D., and M.A. Gallo. 1990. b. A response to comments. Risk Analysis 10(4):629-633.
  • Wattenberg, L.W., J.B. Coccia, and L.K.T. Lam. 1980. Inhibitory effects of phenolic compounds on benzo(a)pyrene-induced neoplasia. Cancer Res. 40:2820-2823. [PubMed: 7388831]
  • Wegstaff, J. 1991. Dietary exposure to furocoumarins. Regulatory Toxicology and Pharmacology 14:261-272. [PubMed: 1771268]
  • Weinberg, R.A. 1989. Oncogenes, antioncogenes and the molecular bases of multistep carcinogenesis. Cancer Res. 49:3713-3721. [PubMed: 2660980]
  • Welsch, C.W. 1992. Cancer Research. 52(Suppl. 7):20405.
  • Welsch, C.W., J.L. House, B.L. Herr, S.J. Eliasberg, and M.A. Welsch. 1990. Enhancement of mammary carcinogenesis by high levels of dietary fat: A phenomenon dependent on ad libitum feeding. JNCI 82:1615-1620. [PubMed: 2136369]
  • Welsh, E.A., J.M. Holden, W.R. Wolf, et al. 1981. Selenium in self-selected diets of Maryland residents. J. Am. Diet Assoc. 79:277-285. [PubMed: 7264113]
  • West, D.W., M.L. Slattery, L.M. Robison, K.L. Schuman, M.H. Ford, A.W. Mahoney, J.L. Lyon, and A.W. Sorensen. 1989. Dietary intake and colon cancer: Sex- and anatomic site-specific associations. Amer. J. Epidemiol. 130:883-894. [PubMed: 2554725]
  • Whittemore, A.S., A.H. Wu-Williams, M. Lee, et al. 1990. Diet, physical activity, and colorectal cancer among Chinese in North America and China. J. Natl. Cancer Inst. 82:915-926. [PubMed: 2342126]
  • WHO (World Health Organization. 1979. DDT and Its Derivatives. Environmental Health Criteria 9. Geneva: WHO.
  • WHO (World Health Organization. 1988. PCBs, PCDDs and PCDFs in Breast Milk: Assessment of Health Risks. Environmental Health Series 29. Europe, Copenhagen: WHO, Regional Office.
  • WHO (World Health Organization. 1991. Lindane. Environmental Health Criteria 124. Geneva: WHO International Programme on Chemical Safety.
  • WHO (World Health Organization. 1993. Polychlorinated Biphenyls and Terphenyls (Second Edition). Environmental Health Criteria 140. Geneva: WHO, International Programme on Chemical Safety.
  • Willett, W.C., M.J. Stampfer, G.A. Colditz, B.A. Rosner, and F.E. Speizer. 1990. Relation of meat, fat, and fiber intake to the risk of colon cancer in a prospective study among women. N. Engl. J. Med. 323:1664-1672. [PubMed: 2172820]
  • Winter, M., and K. Herrmann. 1986. Esters and glucosides of hydroxycinnamic acids in vegetables. J. Agri. Food Chem. 34:616-620.
  • Wu-Williams, A.H., L. Zeise, and D. Thomas. 1992. Risk assessment for aflatoxin B1: A modeling approach. Risk Anal. 12:559-567. [PubMed: 1336206]
  • Yeh, F.-S. M.-C. Yu, C.-C. Mo, S. Luo, M.J. ong, and B.E. Henderson. 1989. Hepatitis B virus, aflatoxins, and hepatocellular carcinoma in southern Guangxi, China. Cancer Res. 49:2506-2509. [PubMed: 2539905]
  • Youngman, L.D., J.Y. Park, and B.N. Ames. 1992. Protein oxidation associated with aging is reduced by dietary restriction of protein or calories. Proc. Natl. Acad. Sci. U.S.A. 89:9112-9116. [PMC free article: PMC50075] [PubMed: 1409611]
  • Zeise, L., R. Wilson, E.A.C. Crouch, and M. Fiering. 1982. Use of toxicity to estimate carcinogenic potency. In Health and Environmental Effects Document: Non-Regulatory and Cost Effective Control of Carcinogenic Hazard. Report to the Department of Energy, Health and Assessment Division, Office of Energy Research. Energy and Environmental Policy Center, Harvard University, Cambridge, September.
  • Zeise, L., R. Wilson, and E.A.C. Crouch. 1984. Use of acute toxicity to estimate carcinogenic risk. Risk Anal. 4(3):187-199.
  • Zeise, L., E.A.C. Crouch, and R. Wilson. 1986. A possible relationship between toxicity and carcinogenicity. J. Am. Coll. Toxicol. 5:137-151.
  • Zeise, L. 1989. Issues in State risk assessment. Pp. 135-144 in Proceedings—Pesticides and Other Toxics: Assessing Their Risks, J.E. White, editor. , ed. University of California; at Riverside.
  • Zhou, Y.C., and R.L. Zheng. 1991. Phenolic compounds and as analog as superoxide anion scavengers and antioxidants. Biochem. Pharmacol. 42:1177-1179. [PubMed: 1653565]
  • Zimmerli, B., and J. Schlatter. 1991. Ethyl carbamate: Analytical methodology, occurrence, formation, biological activity and risk assessment. Mutation Research 259:325-350. [PubMed: 2017216]
Copyright 1996 by the National Academy of Sciences. All rights reserved.
Bookshelf ID: NBK232623


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (2.5M)

Related information

  • PMC
    PubMed Central citations
  • PubMed
    Links to PubMed

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...