distributions. For risk assessment he advocated considering the variety of estimates that are generated from the use of several forms of two-stage models. T. Thorslund thought that the two-stage model is a desirable start on a new way of estimating risks in the regulatory process. He indicated that the data required by Moolgavkar are rarely available and that considerable training is required to use the model.
APPLICATION OF THE TWO-STAGE MODEL TO ANIMAL DATA
S.M. Cohen presented the third major address of the workshop based largely on his own research. Cohen initially indicated that he feels that the current bioassay procedure was a good way to screen for carcinogens (he doesn't know a better way), but that it does not provide sufficient information on the mechanism of action to be useful in biological-based cancer risk assessment. He and his colleagues undertook the problem from the engineering simulation approach.
Cohen agreed fundamentally with the two-stage model described by Moolgavkar. The dividing cell has the greatest susceptibility of a genetic mistake and if cancer arises from genetic mistakes the two factors that influence tumorigenesis are increasing the rate at which genetic events occur or increasing the number of times a given cell divides. The initiated and transformed populations of cells can be augmented in size or in proliferative rates by genetic or nongenetic factors.
Noteworthy aspects of the mouse ED01 study of 2-acetylaminofluorene (2-AAF) were outlined. In the liver, the development of cancer is linear, whereas it is not linear in the bladder, but there appears to be a linear dose-response relationship with respect to DNA adduct formation in both the liver and bladder. This occurs because of a difference in the pharmacokinetics related to the development of tumors at the two different sites. In the liver, 2-AAF is metabolized to the N-hydroxyl intermediate and then to the reactive sulfur-containing metabolite. Because the initiated cells or the ''cells in the foci" apparently do not metabolize 2-AAF, only the first genetic event occurs in the liver. There is no compound-induced increase in cell proliferation. In contrast, in the bladder the N-hydroxylation occurs as in the liver, but then an N-glucuronide is formed. This glucuronide is excreted into the urine where it is hydrolyzed and can lead to DNA adduct formation in the urothelium. Thus
in the bladder, the reactive metabolite causes both initiation and proliferation. Proliferation is only observed above 60 ppm, although DNA adduct formation occurs at dietary concentrations as low as 5ppm. Proliferation in the bladder appears to be essential for tumor formation, however. In conclusion, a two-stage model may be used for either case, but understanding the pharmacodynamics or oncodynamics of the particular tumor-target site is a prerequisite.
Saccharin, which causes bladder cancer in the rat but not in the mouse, hamster, or nonhuman primate, served as an example of a nongenotoxic agent. The male rat is affected to a greater degree than the female. Several types of information were offered to support the importance of proliferation in the development of bladder tumors with saccharin. First, bladder tumors develop principally if the material is given early in life, when cell division in the bladder is normally high. The labeling index (indicator of cell proliferation) is 10 percent at birth, 1.5 percent at seven days, and 0.1 percent at 21 days of age in the rat. Saccharin, co-administered with a cell-proliferating agent (promoter), increases tumor production. Thus, an increase in cell division in the bladder, by irritation as from saccharin or by a promoter, can result in bladder cancer.
The formation of crystals in the urine is the second factor in saccharin induced tumors. The solubility of the saccharin salt is an important aspect in tumor development and can be dependent on the acidity (pKa) of the salt. If the urine is made acid, crystals are not formed in the urine and tumors do not develop. If alkaline, saccharin leads to the formation of silicate crystals in the urine. These crystals irritate the bladder epithelium to cause an increase of cell turnover rate. This increased proliferative rate is considered responsible for increasing the rate of spontaneous mutation and thus for the induction of bladder cancer.
This sequence of events does not occur with significant frequency in the female rat, nor at all in the mouse. In the mouse, the pH of the urine is not changed. Presumably, in nonhuman primates and humans at doses that are not otherwise toxic, the increase in cellular division would not occur. Several other materials, melamine and uracil, that cause an increase in bladder epithelial cell division by forming crystals (although the crystals are of a different nature) were cited additionally to support the apparent relationship between bladder cancer and urinary crystals.
With other nongenotoxic carcinogens, such as dioxin, it is important
to determine whether they mediate cancer through a receptor mechanism. Such a mechanism can induce cancer by increasing cell proliferation as in the case of a hormonal-related receptor, or it may play a role in affecting the immune system via a receptor site on the human-leukocyte antigen (HLA).
In conclusion, just as an indication of mutagenicity does not necessarily indicate that a material is carcinogenic, neither does the ability of an agent to produce cell proliferation at a high dose indicate that it will be carcinogenic (or carcinogenic at lower doses). Therefore, the determination of the dose-relationship to cellular proliferation is important when considering the likelihood of risk.
A regulator's point of view was presented by W. Farland of the USEPA. He said that risk assessment is not merely the estimation of the risk of cancer. The models presented are useful in considering both the quantitative and qualitative assessment of risk and may lead to the establishment of situational and conditional carcinogenic risks. Conditional carcinogens are those that could cause cancer at some dose, whereas situational carcinogens are those that cause cancer only under certain circumstances. Thus, situational carcinogens are important only if the situations under which they cause cancer are likely to occur as a result of conditions. The issues of benign tumors, target specificity, or other mechanisms of action (e.g., decreased immune surveillance) and anticarcinogenesis may have a place in consideration of the two-stage model. Finally, he indicated that the EPA has considered biologically-based modeling in those few cases where sufficient information is available—particularly in the area of characterization of risk.
C. Barrett followed and commented from the viewpoint of a molecular oncologist on the multiple causes of cancer, suggesting that many mechanisms may be involved. A chemical might cause cancer by inducing a heritable mutation on one critical gene, by inducing heritable epigenetic changes in critical genes, or by clonal expansion of one heritable alteration. Additionally, he suggested that one compound might cause cancer through several pathways.
To emphasize the complexities, Barrett said that if the same processes are operating in humans and rodents but rodent tumors have shorter latent periods, then either there are fewer steps or the rates of transitions from step to step are faster in rodents than in humans. Individual tumors in patients may have anywhere from zero to 10 chromosomes
showing loss of heterozygosity (loss of suppressor genes); that is, tumors develop individually.
Three ways by which a substance can influence a multistep carcinogenic process are (1) it can induce heritable mutations in the critical genes (directly or indirectly), (2) it can cause heritable epigenetic changes in critical genes, and (3) it can cause one heritable alteration that increases clonal expansion. Substances acting late in the process may be producing secondary mutational events rather than clonal expansion. Adaptation and potentiation must also be taken into account.
Barrett also cautioned us of the difficulties of generating dose-response curves for mitogenesis and of defining mutagenesis. Cell-cycle control genes and genetic instability are as yet little understood but potentially important, as is also transcriptional control. He concluded that cancer is multicausal, multistep, multigenic, and probably multimechanistic.
In the subsequent general discussion, W. North remarked on the richness of possible modeling approaches and Moolgavkar agreed that we should continue to use the old models until we have more experience with the new ones. K. Crump pointed out that the models would be more helpful if we had comparable data in humans.