On the Relationship Between Carcinogenicity and Acute Toxicity

Parodi et al. ⁽¹⁾ and Zeise et al. ⁽²⁾ found a surprising statistical correlation (or association) between acute toxicity and carcinogenic potency. In order to shed light on the questions of whether or not it is a causal correlation, and whether or not it is a statistical or tautological artifact, we have compared the correlations for the NCI/NTP data set with those for chemicals not in this set. Carcinogenic potencies were taken from the Gold et al. database. We find a weak correlation with an average value of TD₅₀/LD₅₀= 0.04 for the non-NCI data set, compared with TD₅₀/LD₅₀= 0.15 for the NCI data set. We conclude that it is not easy to distinguish types of carcinogens on the basis of whether or not they are acutely toxic.     

Relative Potency of Chemical Carcinogens in Rodents

Extensive carcinogenesis data compiled by Gold et al. for 770 compounds tested in 2944 chronic bioassays in animals provided an opportunity to compare cancer rates across animal species for a wide variety of compounds administered by various routes of exposure. The comparisons in this paper are restricted to the most frequently tested species: rats, mice, and hamsters. When sufficient experimental data exist, Gold et al. provide estimates of the TD50 (the chronic dose rate expressed in mg/kg body weight/day which halves the actuarially adjusted percentage of tumor-free animals at the end of a standard lifetime experiment). Since the current practice generally is to base risk assessments upon the data set producing the highest cancer risk, the ratio of the minimum TD50's provides a measure of the relative potency between two species for each compound administered to animals by the same route. The geometric means of the ratios of minimum TD50's for rats:mice are 1/2.2 and 1/1.3 for diet and gavage, respectively. A mean ratio for rats: mice of 1/1.48 is obtained for compounds administered in the diet when the tumor site is the liver for both species. In general the minimum TD50 is lowest for the rat and highest for the hamster. Although limited data are available for inhalation studies, this route of administration resulted in the poorest agreement between rats and mice. In general, comparisons of minimum TD50's across the three rodent species are generally within a factor of 100 for a wide variety of compounds.     

Leukemia Risk Associated with Benzene Exposure in the Pliofilm Cohort. II. Risk Estimates

The detailed work histories of the individual workers composing the Pliofilm cohort represent a unique resource for estimating the dose-respoonse for leukemia that may follow occupational exposure to benzene. In this paper, we report the results of analyzing the updated Pliofilm cohort using the proportional hazards model, a more sophisticated technique that uses more of the available exposure data than the conditional logistic model used by Rinsky et al. The more rigorously defined exposure estimates derived by Paustenbach et al. are consistent with those of Crump and Allen in giving estimates of the slope of the leukemogenic dose-response that are not as steep as the slope resulting from the exposure estimates of Rinsky et al. We consider estimates of 0.3-0.5 additional leukemia deaths per thousand workers with 45 ppm-years of cumulative benzene exposure to be the best estimates currently available of leukemia risk from occupational exposure to benzene. These risks were estimated in the proportional hazards model when the exposure estimates of Crump and Allen or of Paustenbach et al. were used to derive a cumulative concentration-by-time metric.     

Uncertainty in Cancer Risk Estimates

Several existing databases compiled by Gold et al.^(1–3) for carcinogenesis bioassays are examined to obtain estimates of the reproducibility of cancer rates across experiments, strains, and rodent species. A measure of carcinogenic potency is given by the TD₅₀ (daily dose that causes a tumor type in 50% of the exposed animals that otherwise would not develop the tumor in a standard lifetime). The lognormal distribution can be used to model the uncertainty of the estimates of potency (TD₅₀) and the ratio of TD₅₀'s between two species. For near-replicate bioassays, approximately 95% of the TD₅₀'s are estimated to be within a factor of 4 of the mean. Between strains, about 95% of the TD₅₀'s are estimated to be within a factor of 11 of their mean, and the pure genetic component of variability is accounted for by a factor of 6.8. Between rats and mice, about 95% of the TD₅₀'s are estimated to be within a factor of 32 of the mean, while between humans and experimental animals the factor is 110 for 20 chemicals reported by Allen et al.⁽⁴⁾ The common practice of basing cancer risk estimates on the most sensitive rodent species-strain-sex and using interspecies dose scaling based on body surface area appears to overestimate cancer rates for these 20 human carcinogens by about one order of magnitude on the average. Hence, for chemicals where the dose-response is nearly linear below experimental doses, cancer risk estimates based on animal data are not necessarily conservative and may range from a factor of 10 too low for human carcinogens up to a factor of 1000 too high for approximately 95% of the chemicals tested to date. These limits may need to be modified for specific chemicals where additional mechanistic or pharmacokinetic information may suggest alterations or where particularly sensitive subpopu-lations may be exposed. Supralinearity could lead to anticonservative estimates of cancer risk. Underestimating cancer risk by a specific factor has a much larger impact on the actual number of cancer cases than overestimates of smaller risks by the same factor. This paper does not address the uncertainties in high to low dose extrapolation. If the dose-response is sufficiently nonlinear at low doses to produce cancer risks near zero, then low-dose risk estimates based on linear extrapolation are likely to overestimate risk and the limits of uncertainty cannot be established.     

Route-to-Route Extrapolation of the Toxic Potency of MTBE

MTBE is a volatile organic compound used as an oxygenating agent in gasoline. Inhalation from fumes while refueling automobiles is the principle route of exposure for humans, and toxicity by this route has been well studied. Oral exposures to MTBE exist as well, primarily due to ground-water contamination from leaking stationary sources, such as underground storage tanks. Assessing the potential public health impacts of oral exposures to MTBE is problematic because drinking water studies do not exist for MTBE, and the few oil-gavage studies from which a risk assessment could be derived are limited. This paper evaluates the suitability of the MTBE database for conducting an inhalation route-to-oral route extrapolation of toxicity. This includes evaluating the similarity of critical effect between these two routes, quantifiable differences in absorption, distribution, metabolism, and excretion, and sufficiency of toxicity data by the inhalation route. We conclude that such an extrapolation is appropriate and have validated the extrapolation by finding comparable toxicity between a subchronic gavage oral bioassay and oral doses we extrapolate from a subchronic inhalation bioassay. Our results are extended to the 2-year inhalation toxicity study by Chun et al. (1992) in which rats were exposed to 0, 400, 3000, or 8000 ppm MTBE for 6 hr/d, 5 d/wk. We have estimated the equivalent oral doses to be 0, 130, 940, or 2700 mg/kg/d. These equivalent doses may be useful in conducting noncancer and cancer risk assessments.     

Issues Concerning the Estimation of the TD50

The TD50 (or tumorigenic dose rate 50) is a generally accepted measure of the carcinogenic potency of a chemical in a particular strain of animal. This paper discusses error in the estimation of the TD50 caused by intercurrent mortality and error resulting from consideration of only significant TD50's. Using computer simulations, we found that treatment-related toxicity had only a small effect on estimating the TD50, with errors seldom exceeding 5%. The TD50 is sensitive to changes in tumor lethality with errors ranging to as high as 50%. Many of these errors were significantly different from zero and the results suggest that potency estimation could be improved by basing the estimates upon the tumor incidence rate rather than upon the tumor death rate when an estimate of tumor lethality is obtainable.     

Reanalysis of Dose-Response Data from the Iraqi Methylmercury Poisoning Episode

Applying a hockey stick parametric dose-response model to data on late or retarded development in Iraqi children exposed in utero to methylmercury, with mercury (Hg) exposure characterized by the peak Hg concentration in mothers'hair during pregnancy, Cox et al. calculated the "best statistical estimate" of the threshold for health effects as 10 ppm Hg in hair with a 95% range of uncertainty of between 0 and 13.6 ppm.⁽¹⁾A new application of the hockey stick model to the Iraqi data shows, however, that the statistical upper limit of the threshold based on the hockey stick model could be as high as 255 ppm. Furthermore, the maximum likelihood estimate of the threshold using a different parametric model is virtually zero. These and other analyses demonstrate that threshold estimates based on parametric models exhibit high statistical variability and model dependency, and are highly sensitive to the precise definition of an abnormal response. Consequently, they are not a reliable basis for setting a reference dose (RfD) for methylmercury. Benchmark analyses and statistical analyses useful for deriving NOAELs are also presented. We believe these latter analyses—particularly the benchmark analyses—generally form a sounder basis for determining RfDs than the type of hockey stick analysis presented by Cox et al. However, the acute nature of the exposures, as well as other limitations in the Iraqi data suggest that other data may be more appropriate for determining acceptable human exposures to methylmercury.     

Dose‐Response Modeling for Inhalational Anthrax in Rabbits Following Single or Multiple Exposures

There is a need to advance our ability to characterize the risk of inhalational anthrax following a low‐dose exposure. The exposure scenario most often considered is a single exposure that occurs during an attack. However, long‐term daily low‐dose exposures also represent a realistic exposure scenario, such as what may be encountered by people occupying areas for longer periods. Given this, the objective of the current work was to model two rabbit inhalational anthrax dose‐response data sets. One data set was from single exposures to aerosolized Bacillus anthracis Ames spores. The second data set exposed rabbits repeatedly to aerosols of B. anthracis Ames spores. For the multiple exposure data the cumulative dose (i.e., the sum of the individual daily doses) was used for the model. Lethality was the response for both. Modeling was performed using Benchmark Dose Software evaluating six models: logprobit, loglogistic, Weibull, exponential, gamma, and dichotomous‐Hill. All models produced acceptable fits to either data set. The exponential model was identified as the best fitting model for both data sets. Statistical tests suggested there was no significant difference between the single exposure exponential model results and the multiple exposure exponential model results, which suggests the risk of disease is similar between the two data sets. The dose expected to cause 10% lethality was 15,600 inhaled spores and 18,200 inhaled spores for the single exposure and multiple exposure exponential dose‐response model, respectively, and the 95% lower confidence intervals were 9,800 inhaled spores and 9,200 inhaled spores, respectively.     

Physiologically-Based Pharmacokinetic Modeling of Benzene in Humans: A Bayesian Approach

Benzene is myelotoxic and leukemogenic in humans exposed at high doses (>1 ppm, more definitely above 10 ppm) for extended periods. However, leukemia risks at lower exposures are uncertain. Benzene occurs widely in the work environment and also indoor air, but mostly below 1 ppm, so assessing the leukemia risks at these low concentrations is important. Here, we describe a human physiologically-based pharmacokinetic (PBPK) model that quantifies tissue doses of benzene and its key metabolites, benzene oxide, phenol, and hydroquinone after inhalation and oral exposures. The model was integrated into a statistical framework that acknowledges sources of variation due to inherent intra- and interindividual variation, measurement error, and other data collection issues. A primary contribution of this work is the estimation of population distributions of key PBPK model parameters. We hypothesized that observed interindividual variability in the dosimetry of benzene and its metabolites resulted primarily from known or estimated variability in key metabolic parameters and that a statistical PBPK model that explicitly included variability in only those metabolic parameters would sufficiently describe the observed variability. We then identified parameter distributions for the PBPK model to characterize observed variability through the use of Markov chain Monte Carlo analysis applied to two data sets. The identified parameter distributions described most of the observed variability, but variability in physiological parameters such as organ weights may also be helpful to faithfully predict the observed human-population variability in benzene dosimetry.     

Uncertainties in the CIIT Model for Formaldehyde-Induced Carcinogenicity in the Rat: A Limited Sensitivity Analysis–I

Scientists at the CIIT Centers for Health Research (Conolly et al., 2000, 2003; Kimbell et al., 2001a, 2001b) developed a two-stage clonal expansion model of formaldehyde-induced nasal cancers in the F344 rat that made extensive use of mechanistic information. An inference of their modeling approach was that formaldehyde-induced tumorigenicity could be optimally explained without the role of formaldehyde's mutagenic action. In this article, we examine the strength of this result and modify select features to examine the sensitivity of the predicted dose response to select assumptions. We implement solutions to the two-stage cancer model that are valid for nonhomogeneous models (i.e., models with time-dependent parameters), thus accounting for time dependence in variables. In this reimplementation, we examine the sensitivity of model predictions to pooling historical and concurrent control data, and to lumping sacrificed animals in which tumors were discovered incidentally with those in which death was caused by the tumors. We found the CIIT model results were not significantly altered with the nonhomogeneous solutions. Dose-response predictions below the range of exposures where tumors occurred in the bioassays were highly sensitive to the choice of control data. In the range of exposures where tumors were observed, the model attributed up to 74% of the added tumor probability to formaldehyde's mutagenic action when our reanalysis restricted the use of the National Toxicology Program (NTP) historical control data to only those obtained from inhalation exposures. Model results were insensitive to hourly or daily temporal variations in DNA protein cross-link (DPX) concentration, a surrogate for the dose-metric linked to formaldehyde-induced mutations, prompting us to utilize weekly averages for this quantity. Various other biological and mathematical uncertainties in the model have been retained unmodified in this analysis. These include model specification of initiated cell division and death rates, and uncertainty and variability in the dose response for cell replication rates, issues that will be considered in a future paper.     

A Simple Upper Limit for the Sum of the Risks of the Components in a Mixture

Natural or manufactured products may contain mixtures of carcinogens and the human environment certainly contains mixtures of carcinogens. Various authors have shown that the total risk of a mixture can be approximated by the sum of the risks of the individual components under a variety of conditions at low doses. Under these conditions, summing the individual estimated upper bound risks, as currently often done, is too conservative because it is unlikely that all risks for a mixture are at their maximum levels simultaneously. In the absence of synergism, a simple procedure is proposed for estimating a more appropriate upper bound of the additive risks for a mixture of carcinogens. These simple limits also apply to noncancer endpoints when the risks of the components are approximately additive.     

Extrapolation of Carcinogenicity Between Species: Qualitative and Quantitative Factors

Prediction of human cancer risk from the results of rodent bioassays requires two types of extrapolation: a qualitative extrapolation from short-lived rodent species to long-lived humans, and a quantitative extrapolation from near-toxic doses in the bioassay to low-level human exposures. Experimental evidence on the accuracy of prediction between closely related species tested under similar experimental conditions (rats, mice, and hamsters) indicates that: (1) if a chemical is positive in one species, it will be positive in the second species about 75% of the time; however, since about 50% of test chemicals are positive in each species, by chance alone one would expect a predictive value between species of about 50%. (2) If a chemical induces tumors in a particular target organ in one species, it will induce tumors in the same organ in the second species about 50% of the time. Similar predictive values are obtained in an analysis of prediction from humans to rats or from humans to mice for known human carcinogens. Limitations of bioassay data for use in quantitative extrapolation are discussed, including constraints on both estimates of carcinogenic potency and of the dose-response in experiments with only two doses and a control. Quantitative extrapolation should be based on an understanding of mechanisms of carcinogenesis, particularly mitogenic effects that are present at high and not low doses.     

The Bullwhip Effect—Impact of Stochastic Lead Time,Information Quality,and Information Sharing: A Simulation Study

We use a simulation model called ‘SISCO’ to examine the effects in supply chains of stochastic lead times and of information sharing and quality of that information in a periodic order‐up‐to level inventory system. We test the accuracy of the simulation by verifying the results in Chen et al. (2000a) and Dejonckheere et al. (2004). We find that lead‐time variability exacerbates variance amplification in a supply chain, and that information sharing and information quality are highly significant. For example, using the assumptions in Chen et al. (2000a) and Dejonckheere et al. (2004), we find in a numerical experiment of a customer‐retailer‐wholesaler‐distributor‐factory supply chain that variance amplification is attenuated by nearly 50 percent at the factory due to information sharing. Other assumptions we make are based on interviews or conversations with managers at large supply chains.     

A Bayesian Semiparametric Model for Radiation Dose‐Response Estimation

In evaluating the risk of exposure to health hazards, characterizing the dose‐response relationship and estimating acceptable exposure levels are the primary goals. In analyses of health risks associated with exposure to ionizing radiation, while there is a clear agreement that moderate to high radiation doses cause harmful effects in humans, little has been known about the possible biological effects at low doses, for example, below 0.1 Gy, which is the dose range relevant to most radiation exposures of concern today. A conventional approach to radiation dose‐response estimation based on simple parametric forms, such as the linear nonthreshold model, can be misleading in evaluating the risk and, in particular, its uncertainty at low doses. As an alternative approach, we consider a Bayesian semiparametric model that has a connected piece‐wise‐linear dose‐response function with prior distributions having an autoregressive structure among the random slope coefficients defined over closely spaced dose categories. With a simulation study and application to analysis of cancer incidence data among Japanese atomic bomb survivors, we show that this approach can produce smooth and flexible dose‐response estimation while reasonably handling the risk uncertainty at low doses and elsewhere. With relatively few assumptions and modeling options to be made by the analyst, the method can be particularly useful in assessing risks associated with low‐dose radiation exposures.     

Comparison of the Dependence of the TD50 on Maximum Tolerated Dose for Mutagens and Nonmutagens

The relationship between the minimum TD50 (i.e., the TD50 measured at the most sensitive site) and the maximum dose administered (maxD) in rodent carcinogenicity bioassays was investigated separately for mice and rats. The relationship between log(1/TD50) and log(1/maxD) was analyzed as a function of (1) mutagenicity and (2) the statistical significance cutoff for selecting the minimum TD50 values. For rat bioassays, the variance of log(1/TD50) is larger and the correlation of log(1/TD50) with log(1/maxD) is weaker for mutagens than for nonmutagens, suggesting that the relationship between minimum TD50 and MTD is, in general, stronger for nonmutagens than for mutagens. The difference in correlation does not depend on the TD50 statistical significance cutoff, but the difference in variance is not significant for the most stringently selected dataset. For mouse bioassays, no significant mutagen/nonmutagen differences in log(1/TD50) variance are found. A significantly weaker correlation of log(1/TD50) with log(1/maxD) for mutagens in comparison to nonmutagens occurs only for the dataset with minimum TD50 chosen at the least stringent level, suggesting that this difference may be due to chance variation. We also looked for changes in correlation and regression parameters as a function of mutagenic potency in Salmonella; the variance of log(1/TD50) and its correlation with log(1/maxD) are not found to vary in a consistent manner. Taken as a whole, our results indicate that (1) mutagenicity is a determinant of the TD50/maxD relationship in rats and (2) any effect that mutagenicity may exert on the TD50/maxD relationship in mice is unimportant.(ABSTRACT TRUNCATED AT 250 WORDS)     

One-Hit Models of Carcinogenesis: Conservative or Not?

One-hit formulas are widely believed to be "conservative" when used to analyze carcinogenesis bioassays, in the sense that they will rarely underestimate risks of cancer at low exposures. Such formulas are generally applied to the lifetime incidence of cancer at a specific site, with risks estimated from animal data at zero dose (control), and two or more additional doses that are appreciable fractions of a maximum tolerated dose. No empirical study has demonstrated that the one-hit formula is conservative in the sense described. The Carcinogenesis Bioassay Database System contains data on 1212 separate bioassays of 308 chemical substances tested at exactly three evaluable doses. These provided sufficient data to examine 8432 specific combinations of cancer site with sex, species, and chemical. For each of these we fitted a one-hit formula to the zero and maximum dose data points, then examined the relation of the fitted curve to the incidence rate observed at the mid-dose, with and without adjustment for intercurrent mortality. Both underestimates and overestimates of risk at mid-dose occurred substantially more often than expected by chance. We cannot tell whether such underestimates would occur at lower doses, but offer six biological reasons why underestimates might be expected. In a high percentage of animal bioassays, the one-hit formula is not conservative when applied in the usual way to animal data. It remains possible that the one-hit formula may indeed be conservative at sufficiently low doses (below the observational range), but the usual procedure, applied to the usual dose range, can be nonconservative in estimating the slope of the formula at such low doses. Risk assessments for regulation of carcinogens should incorporate some measure of additional uncertainty.     

Peak Exposures in Epidemiologic Studies and Cancer Risks: Considerations for Regulatory Risk Assessment

We review approaches for characterizing “peak” exposures in epidemiologic studies and methods for incorporating peak exposure metrics in dose–response assessments that contribute to risk assessment. The focus was on potential etiologic relations between environmental chemical exposures and cancer risks. We searched the epidemiologic literature on environmental chemicals classified as carcinogens in which cancer risks were described in relation to “peak” exposures. These articles were evaluated to identify some of the challenges associated with defining and describing cancer risks in relation to peak exposures. We found that definitions of peak exposure varied considerably across studies. Of nine chemical agents included in our review of peak exposure, six had epidemiologic data used by the U.S. Environmental Protection Agency (US EPA) in dose–response assessments to derive inhalation unit risk values. These were benzene, formaldehyde, styrene, trichloroethylene, acrylonitrile, and ethylene oxide. All derived unit risks relied on cumulative exposure for dose–response estimation and none, to our knowledge, considered peak exposure metrics. This is not surprising, given the historical linear no‐threshold default model (generally based on cumulative exposure) used in regulatory risk assessments. With newly proposed US EPA rule language, fuller consideration of alternative exposure and dose–response metrics will be supported. “Peak” exposure has not been consistently defined and rarely has been evaluated in epidemiologic studies of cancer risks. We recommend developing uniform definitions of “peak” exposure to facilitate fuller evaluation of dose response for environmental chemicals and cancer risks, especially where mechanistic understanding indicates that the dose response is unlikely linear and that short‐term high‐intensity exposures increase risk.     

Risk Assessment of Nongenotoxic Carcinogens Based upon Cell Proliferation/Death Rates in Rodents

Increased cell proliferation increases the opportunity for transformations of normal cells to malignant cells via intermediate cells. Nongenotoxic cytotoxic carcinogens that increase cell proliferation rates to replace necrotic cells are likely to have a threshold dose for cytotoxicity below which necrosis and hence, carcinogenesis do not occur. Thus, low dose cancer risk estimates based upon nonthreshold, linear extrapolation are inappropriate for this situation. However, a threshold dose is questionable if a nongenotoxic carcinogen acts via a cell receptor. Also, a nongenotoxic carcinogen that increases the cell proliferation rate, via the cell division rate and/or cell removal rate by apoptosis, by augmenting an existing endogenous mechanism is not likely to have a threshold dose. Whether or not a threshold dose exists for nongenotoxic carcinogens, it is of interest to study the relationship between lifetime tumor incidence and the cell proliferation rate. The Moolgavkar–Venzon–Knudson biologically based stochastic two-stage clonal expansion model is used to describe a carcinogenic process. Because the variability in cell proliferation rates among animals often makes it impossible to detect changes of less than 20% in the rate, it is shown that small changes in the cell proliferation rate, that may be obscured by the background noise in rates, can produce large changes in the lifetime tumor incidence as calculated from the Moolgavkar–Venzon–Knudson model. That is, dose response curves for cell proliferation and tumor incidence do not necessarily mimic each other. This makes the use of no observed effect levels (NOELs) for cell proliferation rates often inadmissible for establishing acceptable daily intakes (ADIs) of nongenotoxic carcinogens. In those cases where low dose linearity is not likely, a potential alternative to a NOEL is a benchmark dose corresponding to a small increase in the cell proliferation rate, e. g., 1%, to which appropriate safety (uncertainty) factors can be applied to arrive at an ADI.