Goodbye to the bioassay

It is time to say goodbye to the standard two-year rodent bioassay.

Abstract

It is time to say goodbye to the standard two-year rodent bioassay. While a few, primarily genotoxic, compounds which are clearly associated with human cancer test positive in the bioassay, there is no science-based, sound foundation for presuming it provides either a valid broad (across different chemicals) capability for discerning potential human carcinogens or a valid starting point for making human risk assessment decisions. The two basic assumptions underlying the bioassay are: (1) rodent carcinogens are human carcinogens; and (2) results obtained at high doses are indicative of results that will occur at lower, environmentally relevant, doses. Both of these assumptions are not correct. Furthermore, a reevaluation of National Toxicology Program bioassay data has revealed that if the dose group size were increased from 50 to 200 rodents per group the number of bioassays deemed to be positive would increase from approximately 50% to very close to 100%. Thus, under the extreme conditions of the bioassay (e.g., high doses, lifetime exposure and, at times, a non-physiological route of administration) virtually all chemicals tested could be made into rodent carcinogens. In recent years there have been a number of proposals to move away from the standard bioassay. In particular, a recently formulated decision tree (Cohen, 2017), which places an emphasis on dose–response relationships and invites the use of MOA information, provides a sound basis for moving on from the bioassay and towards a rational approach to both identify chemicals which appear to have the potential to cause cancer in humans and take dose–response relationships into consideration in order to place the extent, if any, of the risk they might pose into proper perspective.

Introduction

The first indications that chemicals could cause cancer were based on observations in humans. John Hill, in 1761, found that people who used tobacco snuff exhibited an unusual high rate of nasal cancer,1 and in 1775 Percivall Pott described scrotal cancer as an occupational cancer in chimney sweeps, due to contamination of the skin by chimney soot.2 However, 140 years passed until, in 1915, experimental carcinogenesis in animals began with the demonstration that application of coal tar to the ears of rabbits caused malignant skin tumors.3 This observation lead to the use of animals as proxies for evaluating the potential of chemicals to cause human cancer. Justification for the use of animal testing to identify human carcinogens was based, in part, on observations that a number of compounds (typically exhibiting a potential to cause genotoxicity) which sound epidemiology studies indicate are clearly associated with human cancer have been shown to cause cancer in rodents.4 However, beyond this rather small group of chemicals, there is no convincing body of evidence to justify a statement that the opposite is true. The rodent bioassay has not been validated.

Long-term, approaching life-time, studies in rodents were viewed as a way to identify chemicals that had the potential to be carcinogenic to humans. A historical review of the rodent bioassay indicates that while several long-term bioassays were conducted in the 1930s, guidelines did not appear until the 1950s and it was not until the late 1960s that cancer bioassays as we know them today really began.5 At that time the National Cancer Institute (NCI) was charged with evaluating the potential of a rather large number of chemicals for carcinogenic activity.5 The NCI developed guidelines which were published in 1976.6 Chemicals were considered to be either carcinogens or noncarcinogens.7 Chronic rodent bioassay testing became more formalized with the development of the National Toxicology Program in 1984, the Food and Drug Administration's guidelines (referred to as the “Red Book”) in 1982 and revised in 1993, and the Environmental Protection Agency in 1987 and the Organization for Economic Cooperation and Development.5

Virtually all of the chemicals tested early-on in the bioassay were DNA-reactive electrophiles, or capable of being metabolized to a DNA-reactive (electrophilic) metabolite, able to form covalent adducts with DNA and referred to as genotoxic chemicals.8 A DNA adduct is not a mutation; however, it can be a substrate for mutagenesis depending upon the degree to which it is capable of causing miss-pairing of DNA-bases, the capacity for DNA repair and rate of cell proliferation. The belief that mutagenesis is the basis for carcinogenesis was enhanced when Bruce Ames and colleagues developed a bacteria-based assay for mutagens, showed that a number of carcinogens tested positive in this assay and proclaimed that “carcinogens are mutagens”.9 However, by the mid 1990s it became quite clear that the notion that all carcinogens are mutagens is overly simplistic since numerous non-mutagens were testing positive in the bioassay, e.g., saccharin, 1,1′-(2,2,2-trichloroethane-1,1-diyl)bis(4-chlorobenzene) (DDT), 2,3,7,8-tetrachlorodibenzodioxin (TCDD) and d-limonene.10 The first, highly attractive, proposed mechanism underlying carcinogenesis which explains how non-genotoxic chemicals, e.g., non-mutagens, can cause cancer was put forth by Cohen and Ellwein in 1991.11 These investigators hypothesized that chemicals which increased cell proliferation (either by acting as mitogens or causing necrosis leading to compensatory hyperplasia) could cause cancer in a secondary fashion due to errors in DNA replication (DNA polymerase is not 100% accurate) leading to accumulation of critical mutations. The practical significance here is that if the dose of the chemical in question is kept below that which leads to increased cell proliferation cancer is not expected to arise, e.g., the chemical exhibits a threshold.11

Additionally, altered epigenetic parameters leading to aberrant gene expression that can facilitate the transformation of a normal cell into a frank malignancy is another mechanism underlying carcinogenesis which is not, necessarily, based on mutagenesis.12–15

Assumptions underlying the use of the bioassay are not valid

The use of the rodent bioassay as a predictor of the ability of the chemical in question to cause cancer in humans is based on two assumptions, which should be questioned: (1) the animal model result is relevant to humans (interspecies extrapolation); and (2) the tumor response at doses used in animal models are relevant to human exposure levels (dose extrapolation).16 The validity of these assumptions for the two-year bioassay had been based on the tumor response for potent DNA reactive carcinogens. However, for non-genotoxic chemicals, one or both of these assumptions have been demonstrated to be incorrect.17 The use of a linear no threshold (LNT) assumption for carcinogens appears to have its basis in radiation genetics and recent, in-depth reviews provide convincing evidence that this conjecture is not valid.18,19

The National Toxicology Program (NTP) Board of Scientific Counselors reviewed the NTP in 1992.20 There were three working groups and the report of the Carcinogenesis Working Group formed the basis of a subsequent publication.21 An aspect of the use of bioassay results which drew particular attention was the implicit assumptions that underlie a linear extrapolation from doses in the vicinity of the highest dose used (the maximum tolerated dose (MTD)): (1) pharmacokinetics are not dose-dependent; (2) the dose–response relationship is linear; (3) DNA repair is not dependent on dose; (4) response is not dependent on age; and (5) the test doses need not bear a relationship to human exposure (e.g., there is no need to employ doses which are in the vicinity of anticipated human exposure). These assumptions were evaluated in a critical manner. The Board concluded that “the implicit assumptions underlying extrapolation from the MTD … do not appear to be valid”.20,21 Thus, it is wrong to place mathematics in front of biology, the biology should drive the mathematics.

Too many rodent carcinogens

It is instructive to reflect on the observation that a high percentage of chemicals tested in the bioassay are positive only at the MTD. Approximately 50% of chemicals that were bioassayed were deemed to be rodent carcinogens.22,23 Additionally, all too frequently the NTP carcinogens would not have been considered to be carcinogens if the MTD was not included in the study. For example: (1) an evaluation of the carcinogenic effects detected when the diet was the route of administration, the route used in a majority of NTP bioassays,23 revealed that two-thirds of the positive bioassays were positive only when the MTD was employed;24 and (2) an evaluation of 127 positive bioassays (using various routes of administration) indicated that in 43% of these tests statistically significant increases in tumor incidences occurred only at the MTD.23 Furthermore, in those bioassays where a positive effect was observed at a dose below the MTD, this “lower” dose (typically one-half to one-fourth of the MTD) is so close to the MTD relative to the multiple orders of magnitude associated with extrapolation down to a dose that is purported to represent, for example, a one in a million added cancer risk that, for practical purposes, the extrapolation can be considered as based on one data point. There is no valid biological basis for doing this. In this context, it is of interest to consider that: (1) in freshman chemistry laboratory we are taught to not extrapolate beyond a standard curve; and (2) drawing a straight line from a single data point would not be deemed acceptable in a basic geometry class.

Dose influences mechanism and over a wide range of doses one can expect that mechanism will change with changing doses. Thus, a carcinogenic effect observed at a high dose is not necessarily expected to occur at lower doses,25,26 especially when dealing with nongenotoxic compounds.27 It is likely that critical, limiting steps in a mechanistic pathway might become overwhelmed with increasing dose leading to the emergence of new modalities of toxic tissue injury at these higher doses.28,29 Indeed, a variety of examples of dose-dependent transitions in mechanisms of toxicity have been reported.28,29 This includes a demonstration of the influence of DNA repair on non-linear dose-responses for mutation.30 Thus, one should not presume there is no threshold for mutagenicity.18,19,30

A key, very legitimate concern regarding the validity of the bioassay as a predictor of human carcinogenicity is that positive results are basically a high dose artifact, e.g., secondary to toxicity which would not occur at lower doses, particularly doses in the vicinity of what humans might be exposed to.31,32 Nongenotoxic chemicals that cause cancer in rodents at high doses typically increase cell proliferation in their target organ(s). A biologically-based description of carcinogenesis was used to demonstrate that increased cell proliferation can account for the carcinogenicity of nongenotoxic compounds and that an increased rate of cell proliferation also plays a role in the carcinogenic dose–response for genotoxic chemicals.11,33 Thus, doses below that which are capable of causing increased cell proliferation are expected to not be carcinogenic, e.g., there is a threshold. The important role that increased cell proliferation can play in carcinogenesis is not incompatible with the view that a chemical's ability to affect epigenetic parameters is crucial, too. An analysis of over 150 NTP bioassays was performed with the aim of estimating the probability that a statistically significant (p < 0.01) dose–response trend would be obtained at one or more sites in either sex of rats or mice if 200 rather than 50 rodents were used per dose group.34 In this series of chemicals 97/156 (62%) were considered to be rodent carcinogens by the NTP. With an increase of the number of animals per dose group from 50 to 200 it was estimated that 92% would be deemed to be rodent carcinogens. Many of the chemicals are not genotoxic. The analysis suggests that almost all of the chemicals evaluated would be viewed as carcinogens when tested at the maximum tolerated dose (MTD) with a larger sample size.34 Thus, the rodent bioassay could be just a screen for cytotoxins at the MTD rather than a screen specifically for carcinogenicity.34

What about evidence of chemoprevention?

An aspect of the rodent bioassay which typically is not afforded the attention it deserves is the fact that it often results in tumor decreases as well as increases. An analysis of 31 NTP bioassays indicated a similar frequency of tumor increases and decreases.35 However, an evaluation of 218 NTP bioassays revealed that the vast majority (<90%) of the chemicals tested showed at least one statistically significant (p < 0.05) decrease in site-specific tumor incidence.36 The anticarcinogenic responses in rodent bioassays were shown to not be explained by random effects and the authors of this study point out that: (1) the bioassay is actually rather insensitive for detecting anticarcinogenicity because so few tumor types have background rates high enough to be able to detect a statistically significant decrease; and (2) the high doses used might so perturb the animal's physiology that the responses, both positive and negative, have little meaning for lower doses.37 There have been some suggestions that anticarcinogenic effects observed are due to decreased body weight and/or decreased survival of treated rodents.35,36 However, there are a number of chemicals that are anticarcinogenic with no indication of weight or survival depression and many chemicals cause weight or survival decreases with no apparent anticarcinogenic effects.38 Clearly, there has been insufficient discussion regarding how to factor anticarcinogenic effects into the overall evaluation of bioassay results.

Rodents are not small humans

Concordance between outcomes in rodents and humans is a fundamental tenet of the rodent bioassay. However, there are numerous examples which demonstrate that this assumption is not correct. Defining the mode(s) of action by which chemicals induce tumors in laboratory animals has become a key to judgments about the relevance of such tumor data for human risk assessment. Mode of action analytical frameworks depend on both qualitative and quantitative evaluations of relevant data and information: (1) presenting key events in the animal mode of action, (2) developing a “concordance” table for side-by-side comparison of key events as defined in animal studies with comparable information from human systems, and (3) using data and information from mode of action analyses, as well as information on relative sensitivity and exposure, to make weight-of-evidence judgments about the relevance of animal tumors for human cancer assessments.39 The International Program on Chemical Safety (IPSC) has developed a framework for analyzing the relevance of a cancer mode of action (MOA) for humans.40 A variety of nongenotoxic compounds which cause mouse liver tumors (e.g., phenobarbital), a classical activator of the constitutive androstane receptor (CAR);41,42 sulfoxaflor;43 and compounds which activate the nuclear receptor peroxisome proliferator-activated receptor-alpha (PPAR-alpha)44 have been shown to act through a MOA which is not relevant to humans and, therefore, these compounds are not considered to be potential human carcinogens.

Goodbye bioassay

In light of the discussion presented above it is not surprising to see that, in recent years, a variety of publications have made compelling statements to the effect that the time has come to cease performing/requiring the standard 2-year rodent bioassay because it does not provide reliable data for assessing the potential of chemicals to cause human cancer. The mouse arm of the bioassay is viewed as being not scientifically justifiable for the safety assessment of pesticides,45 while some go further and take the position that inclusion of the mouse bioassay is no longer necessary when dealing with nongenotoxic compounds.46 A strong case has been made regarding the bioassay as being no longer necessary for the evaluation of chemicals to potentially cause human liver tumors.47 Additionally, a commentary concerning cancer risk assessment at the US Food and Drug Administration's Center for Drug Evaluation and Research raises the question as to whether the era of the 2-year bioassay is drawing to a close.48

The serious concerns regarding the ability to identify the significance, if any, that the standard 2-year bioassay has for human safety assessment have led to some carefully thought out proposals for alternatives. An evaluation of pharmaceuticals indicated that the value added by conducting a rat two-year carcinogenicity study is dubious, at best, for compounds that lack: (1) histopathologic risk factors for rat neoplasia in chronic (e.g., 6 month) toxicology studies, (2) evidence of hormonal perturbation, and (3) positive genetic toxicology results. Based on these criteria, a decision paradigm was proposed which has the potential to eliminate over 40% of rat two-year testing on new pharmaceuticals without compromise to patient safety.49 This approach is supported by a thorough study which evaluated carcinogenicity data of medical products for human use that have been authorized by the European centralized procedure (CP) between 1995 and 2009.50 The majority of tumor findings in rodent carcinogenicity studies were considered not to be relevant for humans. Appropriately, the authors questioned the value of the rodent bioassay and called for a revision of the carcinogenicity testing paradigm.50 Additionally, a recent study focused on the rat arm of the bioassay, regarding pharmaceuticals, concluded that juxtaposing a consideration of the pharmacological properties of the compound in question with histopathology findings following treatment for 6-months could enhance the ability to define conditions under which two-year rat carcinogenicity studies would not add value.51

Consideration of alternatives to the standard rodent bioassay extends beyond a focus on pharmaceuticals. A retrospective study involving an evaluation of sub-chronic toxicity studies of 163 nongenotoxic chemicals in comparison with tumor responses in 2-year rat carcinogenicity studies supported that notion that chemicals showing no histological risk factors for neoplasia in a sub-chronic study do not require further testing in a bioassay.52 A tiered strategy for cancer hazard identification has been proposed.53 The authors indicate they are well aware of the fact that, at this time, we do not have a good handle on short-term approaches for detecting nongenotoxic compounds that might be carcinogenic. In addition to genotoxic potential, they rely on histopathology, evidence of hormone perturbation and, as yet to be developed, toxicogenomics approaches to discern nongenotoxic carcinogens.53 My emphasis on the “as yet to be determined toxicogenomics approaches” is not intended to connote criticism. While surely being a worthy pursuit, the task of developing a human-relevant toxicogenomics approach to discern nongenotoxic chemicals which are potential human carcinogens will be, to put it mildly, very difficult. Additionally, the authors53 focus on hazard identification and we know that the necessary transition to safety assessment requires an understanding dose–response relationships. The dose makes the poison. The tiered strategy is well thought out and the authors deserve a lot of credit for putting it forward.

In this author's view, the decision tree for evaluating the potential of chemicals to cause a cancer response in humans proposed by Cohen (2017) is the current best, comprehensive proposal for moving away from the standard two-year rodent bioassay.17 It builds upon and advances earlier proposals.49–53 The decision tree starts with an evaluation of a chemical's potential to be DNA-reactive (i.e., a genotoxic compound) If the answer is “yes” it might be a human carcinogen, and there is a need to take human exposure and dose–response relationships into consideration. If the answer is “no”, then one asks if the compound causes significant immune suppression and/or significant estrogenic activity (I would phrase it in a broader sense, hormonal perturbation) and/or significant increased cell proliferation. If the answer to any of these is “yes” then the compound is presumed to be a possible human carcinogen, depending on human exposure and dose–response relatonships. If the answers to all these are “no”, the compound is presumed to be a noncarcinogen with regard to humans. This decision tree,17 which places an emphasis on dose–response relationships and invites the use of MOA information, provides a sound basis for moving on from the bioassay and towards a rational approach to both identify chemicals which have the potential to cause cancer in humans and take dose–response relationships into consideration in order to place the extent, if any, of the risk they might pose into proper perspective. Thus, we have a sound basis for saying “goodbye to the bioassay”.

While there is an emphasis on genotoxic potential of compounds of interest, it needs to be understood that not all genotoxic compounds are expected to be carcinogens and not all carcinogens are expected to be genotoxic compounds capable of providing substrates for mutagenesis. Mutagenesis does not equal carcinogenesis, carcinogenesis involves more than mutagenesis, and altered epigenetic parameters play a key role in the transformation of normal cells into frank malignancies.12–15 However, a thorough evaluation (which includes giving due consideration to dose–response relationships and the existence of thresholds) of the genotoxic potential of compounds in question, with an understanding that high quality in vivo data trumps in vitro data, is warranted. Irrespective of the cancer issue, we do not want to expose humans to significant levels of genotoxic compounds.

Conclusions

It is time to say goodbye to the standard two-year rodent bioassay. The two basic assumptions underlying the bioassay are: (1) rodent carcinogens are human carcinogens; and (2) results obtained at high doses are indicative of results that will occur at lower, environmentally relevant, doses are not correct. In this context, it is ironic to note that the first author of a key paper which describes an algorithm for low-dose extrapolation of carcinogenicity data is an individual named “Guess”.54 While a few, primarily genotoxic, compounds which are clearly associated with human cancer test positive in the bioassay, there is no science-based, sound foundation for presuming it provides either a valid broad (across different chemicals) capability for discerning potential human carcinogens or a valid starting point for making human risk assessment decisions. However, compounds which test negative under the extreme conditions of the rodent bioassay are highly likely to not be capable of causing cancer in humans. Though I do not see a compelling reason to perform the rodent bioassay to focus on the negative data. In recent years there have been a number of proposals to move away from the standard bioassay. In particular, the decision tree devised by Cohen (2017), which places an emphasis on dose–response relationships and invites the use of MOA information, provides a sound basis for moving on from the bioassay and towards a rational approach to both identify chemicals which have the potential to cause cancer in humans and take dose–response relationships into consideration in order to place the extent, if any, of the risk they might pose into proper perspective.17

Conflicts of interest

The author has no conflict of interest to declare. The drafting of this paper was not funded by any agency in the government, private or not-for-profit sectors.

Biography

Jay I. Goodman

Dr Jay I. Goodman, a past President of the Society of Toxicology, is a professor of Pharmacology and Toxicology at Michigan State University and a member of the University's Center for Integrative Toxicology. He is a Diplomate of the American Board of Toxicology and a Fellow of the Academy of Toxicological Sciences. Dr Goodman's research interests are focused on discerning mechanisms underlying, and biomarkers for, non-genotoxic chemical-induced carcinogenesis, and testing the hypothesis that susceptibility to carcinogenesis is related inversely to the capacity to maintain the normal epigenetic status. Dr Goodman is an Associate Editor of Toxicology Sciences, and of Regulatory Toxicology and Pharmacology. Dr Goodman served as a member of the Advisory Committee to the Director of the Centers for Disease Control and Prevention, the Board of Scientific Counselors of the National Toxicology Program, the Board of Directors of the American Board of Toxicology, the Board of Directors of the Academy of Toxicological Sciences, Chair of the Board of Trustees of the International Life Sciences Institute's (ILSI) Health and Environmental Sciences Institute (HESI), and the Board of Scientific Counselors, NIH, National Institute of Environmental Health Sciences. Dr Goodman's honors and awards include: Distinguished Alumnus Award, Long Island University, College of Pharmacy, 1998; Distinguished Alumnus Award, Doctoral Program in Pharmacology, The University of Michigan, 2000; Recipient of the John Barnes Prize Lecture, awarded by the British Toxicology Society, 2005; Recipient of the George H. Scott Memorial Award, awarded by the Toxicology Forum, 2007; Recipient of the Society of Toxicology's Merit Award, 2014; Recipient of the International Society of Regulatory Toxicology and Pharmacology's International; Achievement Award, 2014. Dr Goodman holds a Ph.D. in Pharmacology from The University of Michigan, and was a postdoctoral fellow in the McArdle Laboratory for Cancer Research, University of Wisconsin.