Safety Assessment for Non-genotoxic Rodent Carcinogens: Curves, Low Dose Extrapolations, and Mechanisms in Carcinogenesis: Commentary

Carl L. Alden, DVM,

Monsanto T1A, 800 N. Lindberg Avenue St. Louis, MO 63167

Tel: (314) 694-7392

Fax: (314)694-4426



Probably the most profitable activities in the toxicology industry revolve around exploring the human risk of agents at low levels of exposure that are rodent carcinogens at highly exaggerated exposures or risk of agents for which the rodent data reveal weak or equivocal response. Unfortunately much of this effort occurs in the arena in which the power of our current assays are insufficient for hard data based decision making; thus debates rage on in a void relying on unsubstantiated untestable hypotheses and on shapes of dose response curves with extrapolations and modeling remote to actual data. Let us take a look at the problems and the potential solutions.

"If Pigs Had Wings Could They Fly?"

Academic institutions play an integral role in society creating new generations of toxicologists that assist industry in moving new products into the market place to improve the quality and quantity of human life. The research conducted by academia of course is driven by sources of funding, sometimes with little relevance to the real world. The academic research in turn drives concepts and structure for the toxicology training programs. Unfortunately, these research programs may drive concepts that may raise barriers to improved understanding of the cancer process rather than lower the barriers. As a contemporary example we recently interviewed a candidate whose government funded research involved studies of the mechanism of dioxin mammary carcinogenesis. The best evidence suggesting that dioxin protects against mammary carcinogenesis did not serve as a consequential deterrent in this research. This is a "world class" example of research on the subject of- "If pigs had wings, could they fly?" As a second example, the years of research on mechanisms of phenobarbital carcinogenesis, as described by my good friend Dr. Jim Klaunig, has undoubtedly consumed hundreds of millions of dollars yet has universal acceptance as irrelevant for humans.

Probing The Irrelevant

The phenobarbital type cancer response does not create any serious barrier to registration of a pharmaceutical since the phenomenon is widely recognized. The phenobarbital rodent cancer response occurs at exposures without exaggeration of human exposure thus obviating the value of low dose extrapolations. Note the similar recent multispecies/multitissue carcinogen "block buster" drug approvals that involve liver and endocrine tumorigenesis such as loratidine and atorvastatin. This unique rodent liver phenomenon can be predicted from two week animal tests in which P450 enzyme induction occurs in the absence of hepatocellular toxicity and the liver weight increases over 20% (growth perturbatuion). The liver cancer response is consistent in the mouse but weak in the rat. The bioassay may be negative or only weakly positive in the rat because of the power of the test to detect weak cancer influences. Hence, phenobarbital research cannot be justified in search for short-term predictive tests since these already exist.

Review of all drugs listed in the 1994 PDR reveals that the majority of drugs tested for carcinogenicity test positive and that the most frequent response involves liver associated with robust enzyme induction without toxicity. Through this review we are able to recognize that the rodent cancer bioassay has over a 70% false positive rate (Alden, 1998). This fact, coupled with the fact that the standard rodent test system does not even detect the only important (based on impact on population cancer rates) commercial product causing human cancer, specifically the cigarette (Coggins, 1998), really suggests the softness of the cancer testing exercise despite its profitability in the toxicology industry. As you move from the pharmaceutical industry to industries involving significantly larger margins between human and animal exposure rates you move substantially further into the realm of the irrelevant in cancer hazard identification and risk assessment activities. Unfortunately, in my opinion activities relevant to low dose extrapolations with mathematical models take studies of the irrelevant to the extreme. I am sorry to say that I do not need curves and models to estimate human risk, if I have good exposure data. The industry and regulatory expenditures on activities in this arena must be moved to more productive research strategies if our toxicology industry is to move effectively into the 21st century.

Moving Into The Future

Now having trashed the livelihood of toxicologists involved with carcinogenesis in the academic and industrial settings we can set the stage for effective movement into the future, without clinging to outdated paradigms. While I am an advocate of full employment for toxicologists (but perhaps not statisticians) I also am an advocate of the constant struggle to move to activities considered to be relevant based on today's and yesterday's learnings. Clearly the paradigm of cancer as a one to three step process is designed as pabulum for the masses or at least the students. This is a gross oversimplification and irrelevant in the human setting. The quest for the unitary trigger for cancer has created an over simplistic paradigm through research using potent genotoxic agents. In the cancers best characterized in humans the process can be recognized to represent a complex matrix of multiple phenotypic and genotypic changes involving continuing cell replication, failed growth control and escape from senescence creating the fertile field for carcinogenesis but not one discrete or possibly two or three hits as suggested in the initiation/promotion paradigm.

Continuation of research on phenobarbital carcinogenesis will not serve to further the well being of any human and certainly does not contribute to the quality of life of the rodent. Similarly, the focus on the shape of the curve is a practice in escapism at best, or at worst, simply the current battle ground for attack by pseudoscience activists whose goal is to prevent any new molecules from entering the market place as well as to push civilization back into the quality and quantity of life recorded a century ago.

Amazingly the actual activities in hazard identification and safety assessment are little changed in the past thirty years despite the explosive growth of understanding relevant to molecular mechanisms of carcinogenesis. Actually today's practices are only incrementally improved over the middle ages in which the royalty fed the closest midget or dog a bite of their food before they ate to preclude poisoning. Exceptions exist, for example in the acceptance and application in risk assessment of the unique alpha2uglobulin carcinogenesis in male rats. This one was easy since the driving force was d-limonene, a natural constituent of foods know to be healthful such as oranges. The d-limonene example represents one of the first, if not the first example in which the FDA did not move to regulate under the Delaney Clause against a food additive rodent tumorigen. (Alden, 1986) Can you imagine stamping oranges as cancerous in male rats as some gas stations did on the gas pumps. Talk about irrelevance. Because of the current political impact of extremist environmentalists we probably could not have used the same data base to gain approval of a synthetic chemical unless it were a drug that alleviated human suffering. Note that there is no "no effect level" for toxicity nor carcinogenicity associated with d-limonene indicating that the dose response curve and extrapolations were irrelevant in this example also, as they are with phenobarbital. Another good example of leveraging mechanistic understanding for refining the risk assessment would be with the drug omeprazole in which the mechanism of gastric cancer is understood and also demonstrated not to occur in humans.

Again this is probably an easy one, since this is an extremely effective anti-ulcer drug that gives immediate relief from suffering in humans. Actually the oldest example (that I am aware of) in which good mechanistic research was leveraged to ascertain human safety of a nongenotoxic synthetic chemical that induces rodent cancer at hundreds of thousands fold exposure increase over potential human exposure involved research with nitrilotriacetic acid, a detergent builder. This research supported the first example in which the EPA found a synthetic environmental chemical that is nongenotoxic to not represent a human health hazard back in the late 1970's. (Anderson, and Alden 1982) One has to wonder if this would be possible today because of the environmental extremists.

The success of industry in improving the quality and quantity of human life can be found in the statistics in which people are living longer than ever and cancer rates are declining (Joint Report, 2000). This does not serve as an endorsement for maintaining status quo, however. We must progress the efficiency of our practices in response to the dramatically improved understanding of the cancer process. Eliminating the confirmed irrelevant mouse bioassay with movement to genetically engineered models by the FDA and pharmaceutical industry is one example of progress beginning to occur. Similar initiatives with the rat bioassay hold promise to eliminate the gross nonpredictiveness of the rodent cancer response to models more relevant to humans. There are over 60,000 chemicals in commerce in the United States and probably at least three times that number in Europe. The statistics on rodent cancer bioassays clearly indicate that half of these chemicals using today's antiquated paradigms in hazard identification would be or are rodent carcinogens. However less than 100 of these chemicals are identified as human carcinogens (Alden, 1998). This clearly represents the strongest support for Dr. Ed Calabrese's mantra (1999) that things bad at high dose are good for you or at minimum are inconsequential at exposures relevant to human real life conditions. Improving existing models represents only part of the equation to improve practices in cancer risk assessment. The other major component will require continuing research on mechanisms of carcinogenesis. However, current understandings in mechanism are substantive and sufficient for standardizing testing practices in refinement of the cancer risk assessment as follows.

Mechanisms of Carcinogenesis in Humans and Rodents

1. Tissue Growth/Cell Replication Perturbation
Dr. Klaunig does a superb job of summarizing the contemporary paradigms on experimental mechanisms of nongenotoxic carcinogenesis in the rodent. In the search for solutions to our soft processes in cancer risk assessment though we need to also quickly consider the human. All of the chemicals/agents known to cause human cancer have been reported to fall into one or more of four categories based on biological effects as follows: 1. radioactive, 2. cytotoxic, 3. hormonal, and 4. genotoxic. As discussed by Dr. Klaunig, each of these biologic events are also clearly linked to risk for rodent carcinogenesis. Dr. Klaunig effectively presents the ubiquitous signal of altered tissue growth/cell replication perturbation maintained over the life span of the animal as the singular unifying biologic effect among all nongenotoxic carcinogens. I have carefully followed the experimental cancer literature for over 25 years in search for an exception to his generalization. No creditable exception exists. Please recognize that the occurrence of these events in short term rodent studies does not prove that rodents will get cancer if exposure is continued over the lifetime at levels causing growth/replication perturbation. However, this does not compromise the validity of the association in the slightest. This simply reflects the limited power of the bioassay to detect weak carcinogenic influences. The limited power of the bioassay is not problematic since dose settings involve highly exaggerated conditions and the rodent is recognized as over predictive. Frequently the nongenotoxic mechanisms for rodent carcinogenesis are low incidence phenomenon. The traditional rodent bioassay does not have the power to detect carcinogenic incidences of less than 10%. Only the clueless will not recognize the opportunity presented for improving the effectiveness of cancer risk assessment by this association of tissue growth/cell replication perturbation and cancer. If you do not induce the tissue growth/cell replication perturbation then you do not induce the cancer. The growth perturbation is easily and quickly identifiable experimentally. The technology exists or is close at hand to enable similar determination in humans, where at least with pharmaceutics, exposure is acceptable and quantifiable. With environmental chemicals simply identifying an acceptable margin between human exposure and the rodent threshold exposure for tissue growth/cell replication perturbation should suffice to protect the public based on real data.

2. Oxidative Stress
Mechanisms of toxicity relevant to rodent carcinogenesis are ably reviewed. The specific mechanism inducing the growth perturbation undoubtedly influences the potency of the carcinogenic response. Clearly Dr. Klaunig's focus on oxidative stress relevant to cancer is well placed and likely represents the most important mechanism in nongenotoxic carcinogenesis. I will continue to critique Dr. Klaunig for clinging to the irrelevant phenobarbital type response in the discussion for the reasons stated cryptically above. Phenobarbital clearly causes lifetime sustained tissue growth/cell replication perturbation characterized by decreased apoptosis and mitosis with cell hypertrophy but clearly does not act through a mechanism related to increased cell replication as most phenobarbital antiques would maintain since exposure only through the stage of increased replication does not lead to cancer in the mouse (McClain, personal communication). The suggestion that phenobarbital might act through oxidative stress mechanisms has been hypothesized and tested decades ago- so long ago that I cannot even remember the researchers. The technology exists to definitely and quickly test this unlikely hypothesis. But who cares what the phenobarbital type mechanism consists of in causing rodent cancer? Unfortunately the phenobarbital mechanism is the most frequent cause of rodent carcinogenesis in the pharmaceutical industry where half of new development candidates are enzyme inducers based on a survey by DIA. However the tracks are easily identified and not rate limiting today, at least in pharmaceutical development as discussed above.

3. Other Mechanisms
The listing of inhibition of gap junction and the description of altered methylation as mechanisms in nongenotoxic rodent carcinogenesis are not particularly satisfying in that the evidence does not exist differentiating a causal role from an effect role in the inexorable cancer process. No doubt these events are occurring as prerequisite in the process but this does not imply causality in carcinogenesis.

Two prominent mechanisms need to be addressed that are not mentioned and that have been recognized as potentially relevant in causality of cancer. The first of these is the concept of lysosomal overload with escape of lysosomal constituents that have been demonstrated to have genotoxic potential specifically RNAases and DNAases. Two prominent examples of toxicity with lysosomes as the target illustrate this potential including the alpha2uglobulin nephropathy syndrome in which the lysosome is overloaded with the cytotoxic low molecular weight protein, alpha2uglobulin. The toxicity mechanism in this syndrome may well be the intracellular escape of lysosomal enzymes. (Alden, 1989) The other example is the toxicity associated with orally administered nitrilotriacetic acid (NTA) in which again, there is lysosomal swelling possibly with lysosomal enzyme release as the mechanism of cell toxicity. In this syndrome the toxicity has been identified as prerequisite in the regenerative and proliferative response ultimately resulting in neoplasia. The toxicity has been identified as inexorably linked with zinc overload in the proximal tubular epithelium but the exact mechanism of toxicity has not been further identified to date (Anderson, and Alden 1982). More recent publications on the iron salt of NTA given via intravenous infusion indicate a completely different spectrum of injury than through NTA given orally, no matter the salt form. The intravenous iron salt of NTA may well act via an oxidative stress mechanism but does not have a lysosome target.

The second hypothetical mechanism involves the ubiquitous messenger in biologic systems, nitric oxide. Nitric oxide, when combined with superoxide can form the free radical peroxynitrite. The peroxynitrite radical has been demonstrated to have genotoxic potential and may well play a role in carcinogenesis especially those cancers associated with chronic inflammatory processes such as cirrhosis (Felley-Bosco, 1998). This mechanism can be considered as an extension of the oxidative stress mechanism paradigm, thus not a distinctly different process.


The most frequent biologic events associated with multispecies multisite rodent nongenotoxic carcinogenesis include robust liver P450 enzyme induction without liver toxicity. This cancer process is not relevant for humans as exemplified by phenobarbital. A host of other pharmaceuticals including loratidine and atorvastatin with current sales in the billions of dollars attest to the lack of relevance of our rodent models acting via this specific process. Hopefully the mechanism of phenobarbital will no longer draw research monies in further probing of the irrelevant in human carcinogenesis. The known human carcinogens not linked with genetic injury all perturb tissue growth/replication in humans as well as in rodent short-term tests at doses ultimately linked with cancer. Clearly identifying exposures perturbing growth/replication versus exposures not perturbing growth provides an effective tool for demonstrating the threshold for cancer risk in refining the risk assessment as capably described by Dr. Klaunig. Hopefully we can begin to increasingly use data based decisions in risk assessment versus hypothetical low dose extrapolations and mathematical modeling, building on the wealth of current understanding in human and rodent cancer processes, in arenas beyond pharmaceutics.


Alden, C.L. Lifetime mouse cancer tests- the end of an era? Toxicol. Pathol., 26:4, 584, July, 1998.

Alden, C. A review of unique male rat hydrocarbon nephropathy. Toxicol. Pathol. 14:1, 1986.

Alden, C. Male rat specific alpha2uglobulin nephropathy and renal tumorigenesis. Nephrotox., 1989. Editors: P.Bach and T. Lock. Plenum Publishing.

Anderson, R., and Alden, C. The effects of nitrilotriacetate on cation disposition and ruinary tract toxicity. Fd. Chem. Toxic., 20, 105, 1982.

Calabrese, E. Chemical hormesis: its historical foundations as a biological hypothesis. Toxicol. Pathol., 27:2, 195-216, March, 1999.

Coggins, C. A review of chronic inhalation studies with mainstream cigarette smoke in rats and mice. Toxicol. Pathol., 26:3, 307-314, May, 1998.

Felley-Bosco, E. Role of nitric oxide in genotoxicity: Implication for carcinogenesis. Cancer Metastasis Rev., 17:25-37, March, 1998. Joint Report. NCI, NAACCR, and CDC, 2000.