Abstract
The major provocative conjecture of Dr. Cohen’s article is contained in the subtitle: “The Two-Year Bioassay Is No Longer Necessary.” I truly wish that I could board this train; I cannot. Neither, however, do I wish to lay on the tracks to derail an attempt to shift from a tired toxicological paradigm: the two-year rodent carcinogen bioassay. The difficulty appears to be the need to accept Cohen’s specific mechanistic views of carcinogenesis for the alternative he suggests to be viable.
Everyone would appreciate having at their disposal bioassays that could predict carcinogenesis, mechanisms included, using less time, labor, and treasure than expended by the currently employed two-year rodent models and their kin. Cohen’s proposal here for hepatocellular carcinogens is an approach that might go some distance in reaching that goal after some rethinking. Yet one disturbing exigency comes to mind that might present an impasse: how would quantitative risk assessment or potency ranking be gleaned from short-term studies? The biological unit of time for carcinogenesis is not celestial, but physiologic, best correlated with species life span. The rationale of the two-year rodent studies is that they employ the majority of life span without significant compromise of study viability, which can be thwarted for many reasons. Early nonneoplastic lesions are often considered to be indicators of the eclipse of the viable chronic dose. Moreover, carcinogenesis, per se, will not be adequately realized, either qualitatively or quantitatively, by studies employing only a small fraction of rodent lifetime. I am skeptical that employing early nonneoplastic lesions will give a valid quantitative measure of potency, which should be the primary regulatory parameter.
There may also be other technical concerns with Cohen’s approach. They are likely to be with the experimental details and, most importantly, over reaching a consensus on utility. Balanced regulation, guided by risk assessment, is best based on the use of scientific models that have a reasonable level of consensus. To do other than this results in a degree of bickering not compatible with public confidence in science-based regulation. Moreover, one has only to view the history of the Interagency Coordinating Committee on the Validation of Alternative Methods and its international counterparts, European Centre for the Validation of Alternative Methods and the Japanese Center for the Validation of Alternative Methods, to know that it is not just an ephemeral premonition to worry that consensus about alternative models, such as the Cohen proposal, will be elusory.
I have more general concerns with Cohen’s approach that I would like to air.
From experience, I have learned that convincing risk managers to accept a radically different model (e.g., a “threshold” model) to provide for “safety” from carcinogens in the environment is formidable, despite the evidence for the alternative model. This is particularly so when the risk-management default is a venerable no-threshold model. As long as there is this dichotomy of “no threshold” for carcinogens and “threshold” for all other toxicity, including the ostensible nongenotoxic carcinogens, I believe it will be difficult to achieve a twenty-first-century solution to this dilemma.
I believe it is too simplistic to divide up the universe of carcinogens into “genotoxic” and “nongenotoxic” categories and to ascribe “no biological threshold” and “biological threshold,” respectively, to their collective dose responses. The eternal debate over the existence of thresholds is not resolvable by science in any general sense. We ought not to focus our attention on this enigma. The linear, no-threshold dose-response paradigm for carcinogens is deduced virtually entirely from the one-hit radiation model for cancer, which makes this specific prediction. The one-hit model has been viewed for decades as materially overly simplistic, if not seriously amiss, particularly for orally administered chemical compounds. Yet the prediction of linear, no-threshold dose-response for carcinogens remains a sacred cow. After five-plus decades, we still adhere to this dubious characterization. Instead of one-hit, we now say “genotoxic.”
The experimental irreversibility and stability of early events in carcinogenesis may have made it easy to blame somatic mutation(s) as responsible. But a mutation is not the only stable irreversibility in cellular biology. All differentiation is essentially irreversible without any somatic mutation. Moreover, altered DNA can be repaired by various mechanisms. Gene expression, not DNA sequence, is responsible for the phenotypic behavior of a cell. Phenotypic change from one cell behavior (normal) to another (neoplastic) depends on both genes and the selective expression of genes. Epigenetic research is experiencing a renaissance today based on the fact that DNA sequence fails to explain much about the behavior of organisms and cells; we probably should not conclude that we know the mechanistic details of carcinogenesis before anyone really knows the mechanisms of differentiation and dedifferentiation.
For example, diethylstilbestrol (DES), taken by women during pregnancy, induces vaginal cancer in their female offspring, but only after a latency period lasting until after puberty. Presumably, the hormonally driven cell proliferation starting at that period abets the carcinogenesis as Cohen would predict. Researchers have been attempting to demonstrate genotoxicity for DES for decades but have been largely unsuccessful. Even so, the irreversible imprint delivered in utero on estrogen-sensitive tissue in the vagina suggests something other than mutation is responsible for the memory of this encounter until puberty. I am not entirely sure what mode a risk assessment process for DES should take. Scientifically, I would not recommend a linear, no-threshold model in any event, even though this has been the default for carcinogens that are inducers of “irreversible” insults. The time to reconsider the appropriateness of this venerable default has long past. But it is particularly apropos now that the scientific community is on the verge of reinvigorating the linear, no-threshold default using a therapy of “genotoxicity.”
I suspect that Cohen’s model to predict carcinogenesis outcome from nonneoplastic lesions in subchronic animal studies has merit without invoking mechanistic rationales. However, this approach confers a sense of déjà vu from all the attempts to use a potpourri of short-term tests to predict carcinogenesis. While these tests have enjoyed considerable success for predicting qualitative outcome, they all are deficient in one critical aspect: the ability to predict the dose-response of the carcinogens in question. This potency quantification, not the qualitative determination (yes, no), is the parameter needed for estimating the risk and having a rational regulatory approach to carcinogens in the environment. The Cohen model presented here might be the exception to the general case that these short-term studies fail to predict potency. This hypothesis is experimentally verifiable and should be challenged with this outcome in mind.