Advantages and Disadvantages of Bioassays that Use the MTD
The current animal bioassay was designed as a qualitative screen for carcinogenicity and noncarcinogenicity. It typically does not yield information about the carcinogenicity of materials at doses much below the EMTD (e.g., lower than MTD/2). Current bioassay designs usually include a control group, a group exposed at the MTD, and one to two additional doses, the lowest of which is MTD/10 to MTD/2. Thus, testing is generally conducted in a relatively narrow range of doses below the MTD. Effects observed at these doses might or might not be relevant to human exposure at environmental concentrations.
There are a number of advantages to including the MTD in long-term rodent bioassays. The MTD is the dose most likely to induce tumors; as a result, its use provides information about tumor type and about which organs are sensitive to the carcinogenic effects of a test substance. This information can provide a basis for designing followup studies for characterizing the biologic mechanisms through which cancer is produced. Knowledge of target organs can also assist epidemiologists in designing studies among human populations exposed to a chemical, although species can differ in the tissue that responds to a chemical.
When bioassays are conducted in more than one animal species, use of the MTD provides a consistent basis for interspecies comparisons. By evaluating the differences in response in different species, strains, and sexes of animals, we improve our ability to extrapolate the data
from the animal tests to humans. For the few chemicals on which data exist, potency estimates calculated from animal data have been reported to be highly correlated with similar estimates made from human data; this increases confidence that animal data can be used to predict results in humans (Allen et al., 1988).
Another advantage of the MTD is that tests benefiting from the greater sensitivity of the MTD are more likely to give positive or negative results that can be a starting point for structure-activity correlation analyses. In addition, use of the MTD provides information on a number of end points of toxicity in addition to cancer.
A disadvantage of animal bioassays as they are currently performed is that they generally are not designed specifically to provide information on biochemical and physiologic mechanisms operating during the production of tumors. However, the MTD bioassay might yield clues about mechanisms, which can aid in designing mechanistic studies. For example, preliminary, short-term testing is conducted before the bioassay, primarily to determine how much of a chemical can be administered to animals without decreasing their lifespan through causes other than cancer (i.e., to estimate the MTD). Consequently, mild, non-life-threatening toxicity is common in groups of animals receiving the EMTD. It can lead to changes in food consumption, recurrent cytotoxicity in specific organs, hormonal imbalance, or combinations of these and other effects. Those 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) and could be used to provide the first clues to an agent's biologic mechanisms. Those mechanisms are discussed below.
If a chemical alters physiologic processes the alterations can influence delivery of the chemical or its metabolites to a target site or its clearance from a target site. An example is the effect of large lung burdens of particles on clearance of particles inhaled later (Lee et al., 1985). When animals are exposed to high concentrations of insoluble particles, their lungs rapidly accumulate particle burdens large enough to overwhelm the lungs' normal clearance mechanisms; this increases the rate of accumulation of particles inhaled later. The phenomenon has been reported for diesel-soot particles (Mauderly et al., 1987) and titanium dioxide (Lee et al., 1985). In such a situation, lung tumors might be induced in laboratory animals even by particles that are nontumorigenic at lung burdens
acquired in more relevant human exposures. It is not clear whether this secondary process exhibits a predictable dose-response relationship.
The magnitude of exposure to xenobiotic compounds is known to affect the pathways by which they are metabolized. Metabolic enzymes can be characterized by their affinity for substrates and their capacity for metabolizing them. At low doses, high-affinity low-capacity enzymes can be expected to play the major role in metabolism of a chemical; at high doses, low-affinity high-capacity enzymes will be major contributors. If the metabolic pathway of a chemical is a low-capacity pathway and produces the carcinogenic metabolite, as is true for the metabolism of benzene (Sabourin et al., 1990) and vinyl chloride (Maltoni et al., 1981), exposures of animals to high doses of the chemical, particularly in bolus doses, can lead to an underestimation of the potential for tumor production at lower doses.
If the high-affinity low-capacity metabolic enzymes play a protective role in an organism, and the low-affinity high-capacity enzymes produce reactive, potentially toxic intermediates, then administration of high doses might cause a shift in metabolism to the more toxic pathway. For example, administration of high doses of methylene chloride to mice causes a disproportionate increase in metabolism by the glutathione transferase pathway, and Andersen et al. (1987) suggested that production of reactive metabolites from this pathway was responsible for induction of liver tumors in mice. Similarly, Reitz et al. (1984a,b) noted that, when conjugation of the male rat bladder tumorigen o-phenylphenol (Hiraga and Fujii, 1981, 1984) was saturated by administration of high doses, a shift in metabolic pathways was associated with production of more reactive metabolites. In these cases, overestimation of the potential for tumor production at lower doses would be expected.
Several chemicals are known to induce hormonal imbalances, which might in turn induce tumors by secondary mechanisms. Many of those compounds alter Phase I or Phase II metabolizing enzymes (Sipes and Gandolfi, 1991) or both, and some—such as diethylstilbestrol, methyltestosterone, zearalenones, and retinoids—mimic endogenous hormones (Nebert, 1991). In addition, a small number of nonmutagenic, carcinogenic chlorinated polycyclic hydrocarbons, such as 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and polychlorinated biphenyls (PCBs), exhibit hormone-like activity through receptor-driven mechanisms. One of the primary cellular actions of TCDD is through a cytosolic nuclear protein
designated as the Ah (arylhydrocarbon) receptor, which is analogous in some ways to steroid hormone receptors in both structure and function (Poland and Knutson, 1982; Umbreit and Gallo, 1988). The binding of the ligand to the receptor activates the receptor, which translocates to the nucleus, binds to a response element on the DNA, and serves as a transacting growth regulatory factor (Whitlock, 1986; Landers and Bunce, 1991). Receptor theory and recent evidence supporting hormone and TCDD receptor theory (Stephenson, 1956; Jordan and Murphy, 1990; DeVito et al., 1991) support the argument that the process might require complex occupancy of multiple receptors and nuclear binding sites and that the process is compound-specific and saturable (Safe, 1986). As with hormone action, there is some reason to believe that responses to low doses of TCDD-like molecules might not be sufficient to trigger physiologic responses, but high dose exposures to receptor ligands could lead to aberrant growth, cytotoxicity, and increased risk of cancer. Work now in progress, however, suggests that receptor binding might not by itself explain all the dose-response effects observed at low doses (G. Lucier, NIEHS, pers. comm., 1992).
Some chemicals (thioureas and sulfonamides) are thought to produce thyroid neoplasia secondary to hormonal imbalance in rodents through a variety of mechanisms, including altered thyroid hormone synthesis or hormonal metabolism (Paynter et al., 1988; Hill et al., 1989). Other chemicals that produce thyroid tumors are hepatic microsomal enzyme inducers in rodents at high doses and alter thyroid function by increasing the hepatic disposition of thyroid hormone (Oppenheimer et al., 1968; Cavalieri and Pitt-Rivers, 1981; Hill et al., 1989). Decreased serum thyroid hormone concentrations could result in a compensatory increase in pituitary TSH, which in turn might exert a tumor-promoting effect (Hiasa et al., 1982) or an increase in thyroid neoplasia (McClain, 1989). Because small amounts of thyroxine will block the tumor-promoting effect of a microsomal enzyme inducer, such as phenobarbital, this effect is presumed to be secondary to hormonal imbalance, as opposed to a direct tumor-promoting or direct carcinogenic effect in the thyroid (McClain et al., 1988, 1989; McClain, 1989).
Reserpine has been reported to produce adrenal medullary tumors in male rats, mammary tumors in female mice, and seminal vesicle tumors in male mice (DHEW, 1979). All three tumor types might be secondary to the neurogenic effects of reserpine. Reserpine increases cell prolifer-
ation in the adrenal medulla (Tischler et al., 1988, 1989); that this can be prevented by unilateral denervation of the adrenal gland indicates that the cell proliferation is probably a neurogenically mediated reflexive response to catecholamine depletion (Tischler et al., 1991). Reserpine and other neuroleptic agents increase serum prolactin via inhibition of dopaminergic neurotransmission in the hypothalamus. Thus, the mammary tumors might be secondary to increased serum prolactin.
The examples noted above indicate that failure to account for biologic mechanisms of action of many chemicals that elicited tumors when they were tested in bioassays at their MTDs could lead to errors in qualitative and quantitative predictions about human carcinogenesis.
The specificity and sensitivity of animal bioassays are also important considerations for evaluating the predictability of bioassay results for humans. Sensitivity refers to the ability of bioassays to detect true human carcinogens, and specificity refers to their ability to avoid mistaking substances that do not cause cancer in humans for carcinogens. The sensitivity of bioassays, as well as can be determined at present, is very high. All the known human carcinogens adequately tested thus far (about 39) have been carcinogenic in one or more animal species (Huff and Rall, 1992), although target organs are often inconsistent among species. The specificity of animal bioassays, however, cannot be evaluated now because information on human noncarcinogens is insufficient to make comparisons. The default assumption has been that evidence supporting or refuting carcinogenic activity in animal studies should be considered applicable to humans until better information is obtained.
The most likely explanation for the small numbers of carcinogens and noncarcinogens identified in humans is the relative insensitivity of clinical and epidemiologic methods and their great difficulty in demonstrating causality (Karstadt et al., 1981). It is common for epidemiologists to propose associations between exposures to environmental substances and later cancer formation, and for experimentalists to demonstrate biologic plausibility and causality of the associations through animal studies.
An alternative explanation for the finding of many more rodent carcinogens than human carcinogens is that animal bioassays conducted at the MTD are too sensitive or of higher sensitivity than human epidemiologic studies of exposure at lower doses. One way of examining the question of sensitivity is to consider the responses of animals to fractions of the MTD. Among the difficulties in doing so are that the EMTD
might be an overestimate or underestimate of the true MTD, that the fractions of the highest dose that are tested are still high doses, and that, as the dose is reduced, the statistical power to detect an effect is also reduced. Hoel et al. (1988) reported that, among a group of 52 animal carcinogens tested at multiple doses, 34 (65%) showed statistically significant effects at lower doses. All but three of the remaining 18 substances also increased tumor incidences in a lower-dose group, compared with incidences in controls, but the increases were not statistically significant.
Another possible explanation is that the high proportion of carcinogens found in animal studies reflects bias in the selection of substances for testing. Substances can be selected for a variety of reasons, such as widespread human exposure, commercial use, or prior suspicion of carcinogenicity. In one study in which prior suspicion of carcinogenicity was evaluated as an important selection criterion, Griesemer (NIEHS, pers. comm., 1991) found that, of 255 substances tested because they were suspected of carcinogenicity, 169 (66%) were carcinogenic in animals; of 127 substances tested for other reasons, 26 (20%) were carcinogenic in animals.
One consequence of such selection bias is relative confidence in our ability to identify noncarcinogens. If increased cancer incidences are not detected in animals exposed to a substance at the MTD, one might conclude that the substance is noncarcinogenic for the species-sex-strain combination being tested (within the limits of sensitivity of the test), or that the carcinogenic potential of the substance is too low for carcinogenicity to be detected (under the conditions of the test). Failure to observe statistically significant increases in tumors in a standard set (two species and both sexes) of bioassays performed at the MTD has become the operational definition of noncarcinogen . It is not possible to apply that operational definition in bioassays in which a dose substantially below the MTD was used as the highest dose tested. In such cases, use of a higher dose or the MTD might have revealed a carcinogenic effect. Of course, chemicals that satisfy the operational definition of noncarcinogen might be shown to have carcinogenic potential if tested in larger numbers of animals or in additional strains or species. But the definition has proved useful in categorizing chemicals for regulatory purposes.
The idea that substances that are carcinogenic at very high doses might not be carcinogenic at lower doses requires the assumption that