Tautology or not tautology?

A science-based agenda for health-protective chemical assessments and decisions: overview and consensus statement

The manufacture and production of industrial chemicals continues to increase, with hundreds of thousands of chemicals and chemical mixtures used worldwide, leading to widespread population exposures and resultant health impacts. Low-wealth communities and communities of color often bear disproportionate burdens of exposure and impact; all compounded by regulatory delays to the detriment of public health. Multiple authoritative bodies and scientific consensus groups have called for actions to prevent harmful exposures via improved policy approaches. We worked across multiple disciplines to develop consensus recommendations for health-protective, scientific approaches to reduce harmful chemical exposures, which can be applied to current US policies governing industrial chemicals and environmental pollutants. This consensus identifies five principles and scientific recommendations for improving how agencies like the US Environmental Protection Agency (EPA) approach and conduct hazard and risk assessment and risk management analyses: (1) the financial burden of data generation for any given chemical on (or to be introduced to) the market should be on the chemical producers that benefit from their production and use; (2) lack of data does not equate to lack of hazard, exposure, or risk; (3) populations at greater risk, including those that are more susceptible or more highly exposed, must be better identified and protected to account for their real-world risks; (4) hazard and risk assessments should not assume existence of a "safe" or "no-risk" level of chemical exposure in the diverse general population; and (5) hazard and risk assessments must evaluate and account for financial conflicts of interest in the body of evidence. While many of these recommendations focus specifically on the EPA, they are general principles for environmental health that could be adopted by any agency or entity engaged in exposure, hazard, and risk assessment. We also detail recommendations for four priority areas in companion papers (exposure assessment methods, human variability assessment, methods for quantifying non-cancer health outcomes, and a framework for defining chemical classes). These recommendations constitute key steps for improved evidence-based environmental health decision-making and public health protection.

Toxic Substances Control Act (TSCA) Implementation: How the Amended Law Has Failed to Protect Vulnerable Populations from Toxic Chemicals in the United States

Exposures to industrial chemicals are widespread and can increase the risk of adverse health effects such as cancer, developmental disorders, respiratory effects, diabetes, and reproductive problems. The amended Toxic Substances Control Act (amended TSCA) requires the U.S. Environmental Protection Agency (EPA) to evaluate risks of chemicals in commerce, account for risk to potentially exposed and susceptible populations, and mitigate risks for chemicals determined to pose an unreasonable risk to human health and the environment. This analysis compares EPA's first 10 chemical risk evaluations under amended TSCA to best scientific practices for conducting risk assessments. We find EPA's risk evaluations underestimated human health risks of chemical exposures by excluding conditions of use and exposure pathways; not considering aggregate exposure and cumulative risk; not identifying all potentially exposed or susceptible subpopulations, and not quantifying differences in risk for susceptible groups; not addressing data gaps; and using flawed systematic review approaches to identify and evaluate the relevant evidence. We present specific recommendations for improving the implementation of amended TSCA using the best available science to ensure equitable, socially just safeguards to public health. Failing to remedy these shortcomings will result in continued systematic underestimation of risk for all chemicals evaluated under amended TSCA.

Lack of predictivity of the rat lethality (LD50) test for ecological and human health effects

The relationship between acute toxicity in rats (LD50 values) and indicators of potential health hazards in humans was investigated, based on a chemical population-based paradigm (i.e. the "chemical diversity approach"). These structure-activity relationship-based analyses indicate that high toxicity in rats (i.e. a low LD50 value) is not a good predictor of health effects in humans. In fact, it was found that high acute toxicity to minnows, as well as toxicity to cultured cells, showed significantly greater associations with the potential for health effects than rat LD50 values.

Tiered chemical testing: a value of information approach

In December 2000 the EPA initiated the Voluntary Children's Chemical Evaluation Program (VCCEP) by asking manufacturers to voluntarily sponsor toxicological testing in a tiered process for 23 chemicals selected for the pilot phase. The tiered nature of the VCCEP pilot program creates the need for clearly defined criteria for determining when information is sufficient to assess the potential risks to children. This raises questions about how to determine the "adequacy" of the existing information and assess the need to undertake efforts to reduce uncertainty (through further testing). This article applies a value of information analysis approach to determine adequacy by modeling how toxicological and exposure data collected through the VCCEP may be used to inform risk management decisions. The analysis demonstrates the importance of information about the exposure level and control costs in making decisions regarding further toxicological testing. This article accounts for the cost of delaying control action and identifies the optimal testing strategy for a constrained decisionmaker who, absent applicable human data, cannot regulate without bioassay data on a specific chemical. It also quantifies the differences in optimal testing strategy for three decision criteria: maximizing societal net benefits, ensuring maximum exposure control while net benefits are positive (i.e., benefits outweigh costs), and controlling to the maximum extent technologically feasible while the lifetime risk of cancer exceeds a specific level of risk. Finally, this article shows the large differences that exist in net benefits between the three criteria for the range of exposure levels where the optimal actions differ.

Synergy between systemic toxicity and genotoxicity: relevance to human cancer risk

The health risk manager and policy analyst must frequently make recommendations based upon incomplete toxicity data. This is a situation which is encountered in the evaluation of human carcinogenic risks as animal cancer bioassay results are often not available. In this study, in order to assess the relevance of other possible indicators of carcinogenic risks, we used the "chemical diversity approach" to estimate the magnitude of the human carcinogenic risk based upon Salmonella mutagenicity and systemic toxicity data of the "universe of chemicals" to which humans have the potential to be exposed. Analyses of the properties of 10,000 agents representative of the "universe of chemicals" suggest that chemicals that have genotoxic potentials as well as exhibiting greater systemic toxicity are more likely to be carcinogens than non-genotoxicants or agents that exhibit lesser toxicity. Since "genotoxic" carcinogenicity is a hallmark of recognized human carcinogens, these findings are relevant to human cancer risk assessment.

On interspecies correlations of carcinogenic potencies

It has been established in the literature that constraints on the designs of experiments used to estimate carcinogenic potencies cause overestimation of true biological interspecies correlations of such potencies. This article explores the potential for appreciable underestimation of interspecies correlations, due to the experimental error that occurs in the estimation of carcinogenic potencies.

Threshold of estimated toxicity for regulation of indirect food additives

In response to the objectives of this ATSDR workshop, 2 new procedures are described for assessing the safe use of indirect food additives. First, this workshop provided a timely forum in which to describe a newly proposed Threshold of Regulation (T/R) Policy under which the Food and Drug Administration (FDA) would exempt certain indirect food additives from the formal premarket petition process. Second, this workshop offered an opportunity to discuss 2 circumstances in which Quantitative Structure Activity Relationship (QSAR) methodologies might be utilized in the future to provide decision-support information for substances used in food-contact articles.

The performance of short-term tests in identifying potential germ cell mutagens: a qualitative and quantitative analysis

A retrospective analysis was undertaken to assess the performance of selected short-term tests in the discrimination of mammalian germ cell mutagens and nonmutagens using data derived from the U.S. Environmental Protection Agency/International Agency for Research on Cancer Genetic Activity Profile (EPA/IARC GAP) and EPA GENE-TOX databases. The short-term tests selected were gene mutation in Salmonella (S. typhimurium), cultured mammalian cell gene mutation and chromosomal aberrations, and mammalian bone marrow cytogenetics (micronucleus and chromosomal aberrations). These are the first level tests used in the EPA mutagenicity testing guidelines. The results of this analysis showed good sensitivity of short-term in vitro tests for mammalian cell gene mutation (96%) or chromosomal aberrations (92%) in identifying germ cell mutagens, while the sensitivity of tests for gene mutation in S. typhimurium was lower (79%). Bone marrow micronucleus or chromosomal aberration assays in vivo each displayed a sensitivity of 96%. Thus, both the in vitro and in vivo tests may be used effectively to screen chemicals for potential germ cell mutagenicity. In contrast, the in vitro tests mentioned above performed poorly in discriminating putative germ cell nonmutagens, giving results for specificity at or below what is expected due to chance alone (50-11%). The bone marrow assays were more efficient in this regard, the micronucleus test yielding a specificity of 63% and the chromosomal aberrations assay 64%. The mouse bone marrow micronucleus test also performed well on a quantitative basis, responding at or below the lowest effective doses tested in the mouse dominant lethal assay. Regression analysis of the mean lowest effective doses of chemicals evaluated in vivo showed approximately 1:1 linear correlations for mouse germ cell assays (heritable translocation vs dominant lethal or specific locus tests) as well as for mouse bone marrow assays (micronucleus vs chromosomal aberration). The results suggest the value of the bone marrow micronucleus test as an assay for potential germ cell mutagenicity and the dominant lethal test as a relatively inexpensive choice for confirmation of germ cell damage. The sensitivity of the in vitro assays investigated and the discriminatory capability of the in vivo bone marrow assay affirmed the utility of these tests within the framework of the EPA mutagenicity testing guidelines.

An overview of the report: correlation between carcinogenic potency and the maximum tolerated dose: implications for risk assessment

Current practice in carcinogen bioassay calls for exposure of experimental animals at doses up to and including the maximum tolerated dose (MTD). Such studies have been used to compute measures of carcinogenic potency such as the TD50 as well as unit risk factors such as q1 * for predicting low-dose risks. Recent studies have indicated that these measures of carcinogenic potency are highly correlated with the MTD. Carcinogenic potency has also been shown to be correlated with indicators of mutagenicity and toxicity. Correlation of the MTDs for rats and mice implies a corresponding correlation in TD50 values for these two species. The implications of these results for cancer risk assessment are examined in light of the large variation in potency among chemicals known to induce tumors in rodents.

The value of animal test information in environmental control decisions

Value of information (VOI) analytic techniques are used to evaluate the benefit of performing animal bioassays to provide information about the cancer potency of specific chemical compounds. These tools allow the identification of the conditions in which the cost of reducing uncertainty about potency, by performing a subchronic or chronic bioassay, is justified by the benefit of having improved information for making control decisions. The decision analytic results are readily scaled to apply to a range of human contact rates (exposures) and a variety of control strategies. The sensitivity of results to uncertainty about animal to human extrapolation and the design of the bioassay is explored. An evaluation of the possible gains in general understanding about the mechanisms of carcinogenicity resulting from chronic bioassays is beyond the scope of this approach.

How tautological are interspecies correlations of carcinogenic potencies?

Crouch and Wilson demonstrated a strong correlation between carcinogenic potencies in rats and mice, supporting the extrapolation from mouse to man. Bernstein et al., however, show that the observed correlation is mainly a statistical artifact of bioassay design. Crouch et al. have a comeback. This paper will review the arguments and present some new data. The correlation is largely (but not totally) tautological, confirming results in Bernstein et al.

Chemicals classified by IARC: an investigation of some of their toxicological characteristics

Chemicals classified by the IARC to its Groups 1, 2A, 2B and 3 were examined in an attempt to identify characteristics of their behaviour in experimental studies of carcinogenicity, genotoxicity and acute, mammalian toxicity that correlate with those categories. Only those agents for which carcinogenic potency information was available were studied. For both mice and rats, greater proportions of chemicals were potent carcinogens if they had been categorized in Group 1 (human carcinogens) than if they had been put into one of the other categories. Not surprisingly, there was a weak association between carcinogenic potency and acute toxicity. Mice were especially sensitive to tumour induction by halides, while the lower sensitivity of rats to any carcinogenic effect of halides could be due in part to the higher systemic toxicity of halides in this species: a reduced differential of toxic and carcinogenic doses decreases the dose window in which carcinogenic effects may be demonstrated. It was notable that the human carcinogens were active in those genotoxicity tests with higher specificity for identifying rodent carcinogens. Predictive assays for carcinogenicity considered to have high specificity were in vivo cytogenetic, hepatocyte unscheduled DNA synthesis and Salmonella (5 commonly used strains) and mammalian cell hprt locus mutation assays. None of the relationships was strong enough to form the basis of a simple categorization process, but they could serve to alert investigators to chemicals of special toxicological interest and importance.

Implications of newly recognized relationships between mutagenicity, genotoxicity and carcinogenicity of molecules

The CASE structure-activity relational method was used to predict the mutagenicity, cytogenotoxicity, carcinogenicity, sensory irritation, male rat-specific alpha 2 mu-nephrotoxicity and maximum tolerated dose of a population of molecules (N greater than or equal to 1300). These chemicals were then sorted out by their predicted responses to specific tests and sub-populations of molecules with different prevalence with respect to described endpoints were constructed, i.e. 0-100% prevalences of mutagens, rodent carcinogens and SCE inducers. The predicted properties of these populations were analyzed and the overlap among tests was determined. The method also permits the determination of the dependence among assays and the level of false-positive and false-negative predictions.

Carcinogenic potency correlations: real or artifactual?

It has been suggested by various investigators that observed interspecies correlations of carcinogenic potency estimates based on bioassays are to a degree due to tautologous aspects of experimental designs. Some authors have even suggested that carcinogenic potency estimates derived from chronic bioassays are little more than experimental design artifacts. In this paper it is argued that constraints imposed by the use of the one-hit model to estimate carcinogenic potency are more restrictive than potential experimental design constraints.

Quantitative factors in chemical carcinogenesis: variation in carcinogenic potency

The quantitative assessment of toxicological data on the carcinogenic potential of chemicals requires consideration of a number of factors, including mathematical models of the mechanism of carcinogenic action and pharmacokinetic models for the metabolic activation of the parent compound to its reactive metabolite. In this article, the use of such models in estimating carcinogenic potency and in predicting risks at low levels of exposure is discussed, along with other factors involved in the evaluation of carcinogen bioassay data. The Carcinogenic Potency Database (CPDB) established by Gold et al. (1984, Environ. Health Perspect. 58, 9-322) is used to illustrate the application of quantitative approaches to carcinogenic risk assessment and to examine the variation in the potency of chemical carcinogens. Based on an analysis of 585 experiments selected from the CPDB, the risk-specific (10(-6) doses (RSDs) obtained by linear extrapolation from the TD50 were generally within a factor of 5-10 of those derived from the linearized multistage model. The RSDs obtained by linear extrapolation from the TD50 are roughly log-normally distributed with a median of about 20-90 ng/kg/day, depending on the subset of the CPDB considered. This distribution has been used by Rulis (1986, in Food Protection Technology (C. W. Felix, Ed.), pp. 29-37, Lewis, Chelsea, MI) to explore the concept of a threshold of regulation for chemical carcinogens present in the environment at low levels.

Important recent advances in the practice of health risk assessment: implications for the 1990s

Health risk assessments have been so widely adopted in the United States that their conclusions are a major factor in many environmental decisions. The procedure by which these assessments are conducted is one which has evolved over the past 10-15 years and a number of short-comings have been widely recognized. Unfortunately, improvements in the process have often occurred more slowly than advancements in technology or scientific knowledge. Recent significant advances for more accurately estimating the risks posed by environmental chemicals are likely to have a dramatic effect on the regulation of many substances. Each of the four portions of risk assessment (hazard identification, dose-response assessment, exposure assessment, and risk characterization) has undergone significant refinement since 1985. This paper reviews some of the specific changes and explains the likely benefits as well as the implications. Emphasis is placed on the improved techniques for (a) identifying those chemicals which may pose a human cancer or developmental hazard, (b) using statistical approaches which account for the distribution of interindividual biological differences, (c) using lognormal statistics when interpreting environmental data, (d) using physiologically based pharmacokinetic models for estimating delivered dose and for scaling up rodent data, (e) using biologically based cancer models to account for the seven or more apparently different mechanisms of chemical carcinogenesis, (f) describing the severity of the public health risks by considering those portions of the population exposed to various concentrations of a contaminant, and (g) reviewing how criteria for acceptable risk have been influenced by the number of exposed persons. The net benefit of these improvements should be a reduction in the uncertainty inherent in current estimates of the health risks posed by low level exposure to carcinogens and developmental toxicants.

Chronic bioassays: relevance to quantitative risk assessment of carcinogens

Conventional carcinogenic potency estimates for chemicals have been compared across rodent species. The high correlations previously demonstrated between maximum likelihood estimates (MLE) for chemicals identified as positive carcinogens in both rats and mice (r = 0.83, n = 83) are shown to occur also for chemicals that were negative in both species (r = 0.85, n = 51), for chemicals causing tumors in rats but not mice (r = 0.55, n = 15), and for those causing tumors in mice but not rats (r = 0.68, n = 25). Corresponding upper-bound estimates of carcinogenic potency are also highly correlated across rodent species. The correlations arise from (i) the strong interspecies correlation between maximum doses tested in chronic bioassays, (ii) the small group sizes utilized, and (iii) the narrow range of doses typically tested. These factors constrain conventional ML and upper-bound potency estimates to lie very close to the inverse maximum doses tested, irrespective of the bioassay's qualitative outcomes. The potency estimates are thus artifacts of experimental design and would seem to provide little information on the actual human cancer risks from chemical exposure. Risk assessment models can reveal true potency differences across species, but they must incorporate relevant mechanistic information before quantitative risk extrapolation from rodents to humans is scientifically defensible.

Experimental design constraints on carcinogenic potency estimates

The multistage model is used by U.S. regulatory agencies to calculate estimates of the carcinogenic potency (beta) of chemicals; the data for these estimates are generally obtained from chronic rodent bioassays. Three quantities characterize each group tested in the chronic bioassay: the dose level, the sample size, and the number responding to the dose. The dose levels tested are fixed by conventional protocols; the typical National Toxicology Program (NTP) experimental design calls for use of the maximum tolerated dose (MTD), one-half and one-fourth MTD, plus a control group. Only rarely are doses even one order of magnitude less than the MTD utilized in chronic bioassays. This experimental design constraint on dose selection limits the possible values of beta that can arise from multistage model analyses of chronic bioassay data. Sample size is also constrained by the experimental design of the chronic bioassay; the typical sample size in NTP studies is 50 animals. Occasionally, fewer animals are used, but only rarely are more. Thus, the multistage model which theoretically has three variable quantities with which to estimate carcinogenic potency, has in practice only one: the incidence of treatment-related response. Even this can vary within only a narrow range determined by sample size, the control incidence, and the level of statistical significance desired. The net result of these design constraints is that carcinogenic potency estimates derived from multistage-model analyses of chronic bioassay data may vary within only a narrow range surrounding the inverse maximum dose tested. We have illustrated this by calculating the largest possible finite potency estimates that could have arisen from the experimental designs used to test 82 mouse carcinogens in chronic bioassays. On average these maximum potency estimates were within one order of magnitude of the inverse maximum dose tested. We thus conclude that the chronic rodent bioassay, in and of itself, is altogether inadequate as a data source for estimating the risk to humans from exposure to carcinogenic chemicals.

Species comparisons in evaluating carcinogenicity in humans

Some species and strains of experimental animals have such unique mechanisms of developing cancer that the extrapolation of such bioassay results to the human situation would be fraudulent. This fraudulent extrapolation could occur both qualitatively and quantitatively. Although it will be expensive, species other than the rat, mouse, and hamster should be tested, and tested at wider dose ranges than presently used, before risk assessors will have sufficient data to make legitimate risk estimates. Both species- and strain-unique mechanisms and pharmacokinetic information must be made available to the risk assessors before their estimates can be any better than "guesstimates." As more and more data become available, it will become essential that newer techniques of visualization of the data be used in order to evaluate the weight of evidence that an animal carcinogen is or is not a human carcinogen.

Statistical analysis of Salmonella test data and comparison to results of animal cancer tests

A quantitative framework for the analysis of results of the Salmonella (Ames) test is presented, and the relationship between mutagenesis and carcinogenesis is examined. Color graphics are used for the Salmonella data to describe variability, and trends across multiple chemicals and test conditions. Positivity in the Salmonella test, using statistical criteria to classify results, is compared to positivity in carcinogenesis bioassays for 48 chemicals tested in NCI/NTP-sponsored programs. Sensitivity of the Salmonella test across 5 tester strains was 91% (21/23), while specificity was only 36% (9/25). Results were most concordant for TA100 Aroclor-induced rat S9: sensitivity was 87%, specificity 64%. The correlation of mutagenic potency and carcinogenic potency was 0.41 (p less than 0.001) for 80 chemicals, using results from both the general published literature and the NCI/NTP-sponsored programs. After removal of 3 extreme values, the correlation was 0.24 (p = 0.04).