Uncertainty in cancer risk estimates

Uncertainty of Estimates of Cancer Risks Derived by Extrapolation from High to Low Doses and from Animals to Humans

The uncertainties associated with extrapolating model-based cancer risks from high to low doses and animal-based cancer risks to humans are examined. It is argued that low-dose linear extrapolation based on statistical confidence limits calculated from animal data is designed to account for data uncertainty, model-selection uncertainty, and model-fitting instability. The intent is to err on the side of safety, that is, overstating rather than understating the true risk. The tendency toward conservatism in predicting human cancer risks from animal data based on linear extrapolation is confirmed by a real-data analysis of the various sources of uncertainty involved in extrapolating from animals to humans. Along with the tendency toward conservatism, a high degree of overall uncertainty in the interspecies extrapolation process is demonstrated. It is concluded that human cancer risk estimates based on animal data may underestimate the true risk by a factor of 10 or may overestimate that risk by a factor of 1,000.

Ecological Risk Assessment at the Department of Energy: An Evolving Process

The United States Department of Energy (DOE) has facilities in 34 states, and many of these have chemical or radiological contami-nation that provides a potential risk to human or ecological health. Over the next few decades many of these sites will be cleaned up, and ecological risk assessment will be one tool used to make decisions about remediation and future land use. The DOE has developed an overall strategy for making remediation decisions that involves using risk assessment, with stakeholder input, although the final decisions are the Departments. The key elements of its ecological risk assessments involve valuing the severity and likelihood of occurrence of adverse ecological effects. It is currently using a process that incorporates descriptions of the environmental risk, and valuations of the severity and likelihood of an adverse outcome before, during, and after any remedial activity. The primary difficulty with the current DOE approach to risk has been a failure to use existing information to identify either species of concern or unique habitats at risk, and a lack of uniformity across the DOE complex. Nonetheless, the inclusion of ecological risk assessment in the decision-making process will help achieve one of the new missions of DOE: the protection and maintenance of biodiversity and healthy ecosystems at sites under its control.

Automated and reproducible read-across like models for predicting carcinogenic potency

Several qualitative (hazard-based) models for chronic toxicity prediction are available through commercial and freely available software, but in the context of risk assessment a quantitative value is mandatory in order to be able to apply a Margin of Exposure (predicted toxicity/exposure estimate) approach to interpret the data. Recently quantitative models for the prediction of the carcinogenic potency have been developed, opening some hopes in this area, but this promising approach is currently limited by the fact that the proposed programs are neither publically nor commercially available. In this article we describe how two models (one for mouse and one for rat) for the carcinogenic potency (TD50) prediction have been developed, using lazar (Lazy Structure Activity Relationships), a procedure similar to read-across, but automated and reproducible. The models obtained have been compared with the recently published ones, resulting in a similar performance. Our aim is also to make the models freely available in the near future thought a user friendly internet web site.

An evaluation of the accuracy of small-area demographic estimates of population at risk and its effect on prevalence statistics

Demographic estimates of population at risk often underpin epidemiologic research and public health surveillance efforts. In spite of their central importance to epidemiology and public-health practice, little previous attention has been paid to evaluating the magnitude of errors associated with such estimates or the sensitivity of epidemiologic statistics to these effects. In spite of the well-known observation that accuracy in demographic estimates declines as the size of the population to be estimated decreases, demographers continue to face pressure to produce estimates for increasingly fine-grained population characteristics at ever-smaller geographic scales. Unfortunately, little guidance on the magnitude of errors that can be expected in such estimates is currently available in the literature and available for consideration in small-area epidemiology. This paper attempts to fill this current gap by producing a Vintage 2010 set of single-year-of-age estimates for census tracts, then evaluating their accuracy and precision in light of the results of the 2010 Census. These estimates are produced and evaluated for 499 census tracts in New Mexico for single-years of age from 0 to 21 and for each sex individually. The error distributions associated with these estimates are characterized statistically using non-parametric statistics including the median and 2.5th and 97.5th percentiles. The impact of these errors are considered through simulations in which observed and estimated 2010 population counts are used as alternative denominators and simulated event counts are used to compute a realistic range fo prevalence values. The implications of the results of this study for small-area epidemiologic research in cancer and environmental health are considered.

Establishing the level of safety concern for chemicals in food without the need for toxicity testing

There is demand for methodologies to establish levels of safety concern associated with dietary exposures to chemicals for which no toxicological data are available. In such situations, the application of in silico methods appears promising. To make safety statement requires quantitative predictions of toxicological reference points such as no observed adverse effect level and carcinogenic potency for DNA-reacting chemicals. A decision tree (DT) has been developed to aid integrating exposure information and predicted toxicological reference points obtained with quantitative structure activity relationship ((Q)SAR) software and read across techniques. The predicted toxicological values are compared with exposure to obtain margins of exposure (MoE). The size of the MoE defines the level of safety concern and should account for a number of uncertainties such as the classical interspecies and inter-individual variability as well as others determined on a case by case basis. An analysis of the uncertainties of in silico approaches together with results from case studies suggest that establishing safety concern based on application of the DT is unlikely to be significantly more uncertain than based on experimental data. The DT makes a full use of all data available, ensuring an adequate degree of conservatism. It can be used when fast decision making is required.

Ratcheting up cancer potency estimates

The current paradigm for cancer risk assessment in the United States (U.S.) typically requires selection of representative rodent bioassay dose-response data for extrapolation to a single cancer potency estimate for humans. In the absence of extensive further information, the chosen bioassay result generally is taken to be that which gives the highest extrapolated result from the "most sensitive" species or strain. The estimated human cancer potency is thus derived from an upper-bound value on animal cancer potency that is technically similar to an extreme value statistic. Thus additional information from further bioassays can only lead to equal or larger cancer potency estimates. We here calculate the size of this effect using the collected results of a large number of bioassays. Since many standards are predicated on the value of the cancer potency, this effect is undesirable in producing a strong counter-incentive to performing further bioassays.

An evaluation of benchmark dose methodology for non-cancer continuous-data health effects in animals due to exposures to dioxin (TCDD)

The U.S. Environmental Protection Agency (EPA) has conducted extensive reviews and analyses of health effects associated with exposures to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and related compounds. Because the carcinogenicity of TCDD has received considerable attention from EPA and others, this paper focuses on animal data for non-cancer health effects that sometimes appear to be almost as sensitive as cancer to TCDD exposures. Benchmark dose (BMD) methodology can be used to identify point-of-departure (POD) estimates for use in derivation of reference doses or evaluation of margins of exposure. However, selection of an appropriate BMD methodology for assessment of non-cancer data, which are usually continuous (non-quantal), needs to be considered. One option available for a benchmark dose is to use a small percentage change in the mean response relative to the estimated maximum effect of TCDD at large doses. The benchmark based on a change estimated to equal 1% of the estimated maximum change from background to the asymptotic response at large doses (denoted as the relative ED01) was used by EPA in a reassessment of TCDD health risks. A lower confidence limit (LED01) could serve as a point of departure for setting a reference dose (RfD). This is a somewhat arbitrary effect level, generally within the background range of variation among unexposed animals, with an unknown risk. An alternative approach is recommended in which the risk of abnormal levels can be estimated. For continuous-data effects, a low and/or high percentile (e.g., 1st and/or 99th) in unexposed control animals can be used to define abnormal (not necessarily adverse) levels. From a dose-response curve and the standard deviation, it is possible to estimate the excess risk (proportion) of animals with abnormal levels as a function of dose for normally distributed levels. With this approach, the risk-based benchmark dose (BMD01) represents the dose with an estimated excess risk of 1% of the animals in the abnormal range rather than an arbitrary change in the value of the measured endpoint. Values for the relative and risk-based benchmark doses are computed from published data for a variety of non-cancer health effects associated with exposure to TCDD. For the 30 cases investigated, the BMD01 tended to vary around the lowest experimental dose tested, whereas the relative ED01 tended to be about a factor of three below the lowest dose, and the BMD01 was more precisely estimated than the ED01 as reflected by narrower confidence intervals. The BMDL01 values were on average more than fivefold higher than the corresponding LED01 values. However, these values still provide a conservative assessment for POD assessment, because the BMDL01 tends to be about an order of magnitude lower (more conservative) than the no-observed-adverse-effect level. This analysis demonstrates the potential impact of alternative choices in benchmark dose methodology. In combination with selection of appropriate adverse health effect endpoint(s) and studies, use of the risk-based BMD results in identification of more valid and meaningful POD estimates for non-cancer effects compared to the use of the relative ED approach.

General principles for risk assessment of living modified organisms: Lessons from chemical risk assessment

Modern biotechnology has led to the development and use of Living Modified Organisms (LMOs) for agriculture and other purposes. Regulators at the national level are increasingly depending on risk assessment as a tool for assessing potential adverse effects of LMOs on the environment and human health. In addition, the Cartagena Protocol on Biosafety, an international agreement expected to enter into force in the near future, requires risk assessment as the basis for decision-making regarding import of some LMOs. While LMO risk assessment is relatively new, there are other risk assessment disciplines which have developed over longer time periods. The field of assessment of the environmental and human health risks of chemicals is particularly well developed, and is similar in application to LMO risk assessment. This paper aims to draw lessons for LMO risk assessment from the vast experience with chemical risk assessment. Seven general principles are outlined which should serve as a useful checklist to guide assessments of risks posed by LMOs.

Comparison of cancer risk estimates based on a variety of risk assessment methodologies

The EPA guidelines recommend a benchmark dose as a point of departure (PoD) for low-dose cancer risk assessment. Generally the PoD is the lower 95% confidence limit on the dose estimated to produce an extra lifetime cancer risk of 10% (LTD(10)). Due to the relatively narrow range of doses in two-year bioassays and the limited range of statistically significant tumor incidence rates, the estimate of the LTD(10) is constrained to a relatively narrow range of values. Because of this constraint, simple, quick estimates of the LTD(10) can be readily obtained for hundreds of rodent carcinogens from the Carcinogenic Potency Database (CPDB) of Gold et al. Three estimation procedures for LTD(10) are described, using increasing information from the CPDB: (A) based on only the maximum tolerated dose (the highest dose tested); (B) based on the TD(50); and (C) based on the TD(50) and its lower 99% confidence limit. As expected, results indicate overall similarity of the LTD(10) estimates and the value of using additional information. For Method (C) the estimator based on the [[(TD(50))(0.36) x (LoConf)(0.64)]/6.6] is generally similar to the estimator based on the one-hit model or multistage model LTD(10). This simple estimate of the LTD(10) is applicable for both linear and curved dose responses with high or low background tumor rates, and whether the confidence limits on the TD(50) are wide or tight. The EPA guidelines provide for a margin of exposure approach if data are sufficient to support a nonlinear dose-response. The reference dose for cancer for a nonlinear dose-response curve based on a 10,000-fold uncertainty (safety) factor from the LTD(10), i.e., the LTD(10)/10,000, is mathematically equivalent to the value for a linear extrapolation from the LTD(10) to the dose corresponding to a cancer risk of <10(-5) (LTD(10)/10,000). The cancer risk at <10(-5) obtained by using the q(1)(*) from the multistage model, is similar to LTD(10)/10,000. For a nonlinear case, an uncertainty factor of less than 10,000 is likely to be used, which would result in a higher (less stringent) acceptable exposure level.

A procedure for developing risk-based reference doses

Reference doses (RfDs) for toxic substances based on a no observed adverse effect level (NOAEL) or lowest observed adverse effect level (LOAEL) are established to restrict human exposures to only nontoxic or minimally toxic levels. In order to calculate a risk-based RfD, for the point of departure it is necessary to replace a NOAEL or LOAEL by a benchmark dose (BMD) estimated to be associated with a specified level of estimable risk in or near the low end of an experimental dose range. Then the RfD is calculated by dividing the BMD by a series of uncertainty factors. Among these uncertainty factors is one for interindividual sensitivity, typically assigned a value of 10. If information is available on interindividual sensitivity, this default factor can be replaced with a factor expected to provide protection for a specified proportion of a population. Examination of published databases suggests interindividual effects often to be approximately log-normal. For example, in order to illustrate the procedure for establishing a risk-based RfD, a standard deviation of the logarithm (base e) of individual sensitivity of 1.7 was selected, i.e., a factor of 5.5. This value is near the upper range of values reported in the literature (D. Hattis et al., 1999, in "Characterizing Human Variability in the Risk Assessment Process," ILSI Press, Washington, DC). Using this information in combination with an RfD based on a benchmark dose associated with a specified level of risk, the risk at the RfD can be estimated. For example, a benchmark dose associated with a risk of 10% divided by 60, to account for interindividual variation, is expected to limit risk at the RfD to about 1 in 10,000. If the standard deviation of the logarithm (base e) of individual sensitivity is 1.2, a more typical value, the divisor is approximately 20. These would replace an RfD having an unknown risk based on the LOAEL divided by 100. Or, an RfD with a specified level of risk could be estimated. The estimate of risk can be improved for a specific case by replacing an overall estimate of the standard deviation for interindividual variability by an estimate of the standard deviation for a particular class of chemicals and/or biological endpoint, if available. Risks can be substantially lower for smaller values of interindividual variability. Determination of an RfD still would require an additional uncertainty factor for animal to human extrapolation.

Human toxicity potentials for life-cycle assessment and toxics release inventory risk screening

The human toxicity potential (HTP), a calculated index that reflects the potential harm of a unit of chemical released into the environment, is based on both the inherent toxicity of a compound and its potential dose. It is used to weight emissions inventoried as part of a life-cycle assessment (LCA) or in the toxics release inventory (TRI) and to aggregate emissions in terms of a reference compound. Total emissions can be evaluated in terms of benzene equivalence (carcinogens) and toluene equivalents (noncarcinogens). The potential dose is calculated using a generic fate and exposure model, CalTOX, which determines the distribution of a chemical in a model environment and accounts for a number of exposure routes, including inhalation, ingestion of produce, fish, and meat, and dermal contact with water and soil. Toxicity is represented by the cancer potency q1* for carcinogens and the safe dose (RfD, RfC) for noncarcinogens. This article presents cancer and noncancer HTP values for air and surface-water emissions of 330 compounds. This list covers 258 chemicals listed in U.S. Environmental Protection Agency TRI, or 79 weight-% of the TRI releases to air reported in 1997.

New issues in cancer risk assessment

When a nonlinear dose-response at low doses can be justified, an acceptable daily intake for a carcinogen can be obtained by dividing a benchmark dose, associated with a low incidence of tumors in animals, by uncertainty factors to account for animal-to-human extrapolation, human variability, and risk reduction from a low observed adverse-effect level. This approach can utilize mechanistic information to justify smaller uncertainty factors than typical default values of 10. If a nonlinear dose-response cannot be justified, traditional linear extrapolation from the benchmark dose to zero sometimes gives similar results. This suggests a unified risk-assessment procedure based on uncertainty factors. The issue of cross-species extrapolation based on the risk relative to background risks, rather than excess risk, is examined. The relative risk approach reduces the estimates of cancer risk in humans based on common rodent tumors, such as the liver in some strains of mice.

Regulatory cancer risk assessment based on a quick estimate of a benchmark dose derived from the maximum tolerated dose

The proposed U.S. Environmental Protection Agency carcinogen risk assessment guidelines employ a benchmark dose as a point of departure (POD) for low-dose risk assessment. If information on the carcinogenic mode of action for a chemical supports a nonlinear dose-response curve below the POD, a margin-of-exposure ratio between the POD and anticipated human exposure would be considered. The POD would be divided by uncertainty (safety) factors to arrive at a reference dose that is likely to produce no, or at most negligible, cancer risk for humans. If nonlinearity below the POD is not supported by sufficient evidence, then linear extrapolation from the incidence at the POD to zero would be used for low-dose cancer risk estimation. The carcinogen guidelines suggest that the lower 95% confidence limit on the dose estimated to produce an excess of tumors in 10% of the animals (LTD10) be used for the POD. Due to the relatively narrow range of doses in 2-year rodent bioassays and the limited range of statistically significant tumor incidence rates, the estimate of the LTD10 obtained from 2-year bioassays is constrained to a relatively narrow range of values. Because of this constraint, a simple, quick, and relatively precise determination of the LTD10 can be obtained by the maximum tolerated dose (MTD) divided by 7. All that is needed is a 90-day study to establish the MTD. It is shown that the LTD10 determined by this relatively easy procedure is generally within a factor of 10 of the LTD10 that would be estimated using tumor incidence rates from 2-year bioassays. Estimates of cancer potency from replicated 2-year bioassays, and hence estimates of cancer risk, have been show to vary by a factor of 4 around a median value. Thus, there may be little gain in precision of cancer risk estimates derived from a 2-year bioassay, compared to the estimate based on the MTD from a 90-day study. If the anticipated human exposure were estimated to be small relative to the MTD/7 = LTD10, there may be little value in conducting a chronic 2-year study in rodents because the estimate of cancer risk would be low regardless of the results of a 2-year bioassay. Linear extrapolation to a risk of less than 1 in 100,000 and use of an uncertainty factor, e.g., of 10,000, would give the same regulatory "safe dose." Linear extrapolation to a virtually safe dose associated with a cancer risk estimate of less than one in a million would be 10 times lower than the reference dose based on the LTD10/10,000.

Health risk assessment practices in the U.S. Food and Drug Administration

The U.S. Food and Drug Administration (FDA) regulates a wide variety of consumer products. Safety issues involve chemical and microbial contaminants in food, biologies, and medical devices; side effects from prescription and nonprescription drugs; residues of animal drugs in food; and radiation from electronic devices. Because of this wide diversity, the legal standards, rules, and policies governing the regulation of these products differ considerably. Hence, risk assessment and risk management practices within the FDA are of necessity quite diverse. This paper presents a summary of risk assessment practices at each of the product centers of the FDA (Center for Food Safety and Applied Nutrition, Center for Drug Evaluation and Research, Center for Biologics Evaluation and Research, Center for Devices and Radiological Health, and Center for Veterinary Medicine) and of the development of risk assessment procedures at the National Center for Toxicological Research.

Activity profiles of carcinogenicity data: application in hazard identification and risk assessment

Animal cancer data play a primary role in human risk assessment due to the limited epidemiological data. The current database of test results from the NCI/NTP rodent bioassays provide valuable information concerning the carcinogenic potential of hundreds of environmental agents. An approach is presented to reduce and graphically display these data as activity profiles to allow visualization and assessment of tumor response trends across multiple parameters, e.g. species, sex, target site, and route of exposure. Spreadsheet graphics are used to construct the profiles organized on the multiple parameters of carcinogenicity in a format that enables comparative analysis among chemicals. Several example applications are described to illustrate the value of activity profiles in hazard identification and risk assessment. The NCI/NTP data used in developing this concept are from the Carcinogen Potency Database (CPDB) complied by Gold et al. (Environ. Health Perspect. 103 (Suppl. 8) (1995) 3-122). Computer links to the underlying details in the CPDB are maintained such that specific histopathologies at individual tumor sites, duration of the study, dose-response data, and notes related to diet, survival, treatments, and the authors evaluation are available to aid in the assessment process. The profiles display carcinogen potency based on the tumorigenic dose rate 50 (TD50), i.e. the chronic dose rate that would induce tumors in half of the test animals at the end of their standard lifespan adjusting for spontaneous tumors. The TD50 values provide an index for establishing a relative potency ranking of the chemicals for any specific parameter, such as species or target site. An example ranking of hepatocarcinogens is presented to illustrate relative potencies for chemical analogs. The rank order indicates that the degree and type of halogenation of alkanes has a direct bearing on the carcinogenic potency of these compounds.

A pharmacokinetic analysis of interspecies extrapolation in dioxin risk assessment

This study entails a pharmacokinetic analysis of the relationship between the external dose of 2,3,7,8-tetrachlorodibenzo-p-dioxin (dioxin, TCDD) and resulting concentrations of TCDD in internal tissues and organs of humans and rodent species. The methodology is based on the development and testing of physiologically based pharmacokinetic models for several rodent species and humans. The results indicate that the relationship between the external dose of TCDD and resulting TCDD concentrations in liver and adipose tissue of humans and various species of rats and mice can vary by as much as 725 fold, illustrating that humans and experimental animals differ considerably in their ability to convert external dosages of dioxin to tissue concentrations. Interspecies scaling factors are reported to express the differences in tissue concentrations of dioxin between mice, rats and humans in response to an equivalent external dose. The significance of these findings for conducting human cancer and ecological risk assessments is discussed. It is recommended that pharmacokinetic differences be considered explicity in risk estimation, while separately recognizing interspecies differences in pharmacodynamics (sensitivity).

Risk estimation--an overview: disease risk estimation based upon animal bioassays

Much more could be written about the issues addressed here, as well as about issues that are not even mentioned. The goal was to present a brief overview of some of the techniques and issues in quantitative health risk assessment based upon animal data. Hopefully, this overview will provoke some attention to specific in risk assessment that require more research. Perhaps the bibliographic references given will lead to other papers.

Reproducibility of the dose-response curve of steroid-induced cleft palate in mice

Pregnant CD-1 mice were exposed to cortisone acetate at doses ranging from 20 to 100 mg/kg/day on days 10-13 by oral and intramuscular routes. Multiple replicate assays were conducted under identical conditions to assess the reproducibility of the dose-response curve for cleft palate. The data were fitted to the probit, logistic, multistage or Armitage-Doll, and Weibull dose-response model separately for each route of exposure. The curves were then tested for parallel slopes (probit and logistic models) or coincidence of model parameters (multistage and Weibull models). The 19 replicate experiments had a wide range of slope estimates, wider for the oral than for the intramuscular experiments. For all models and both routes of exposure the null hypothesis of equality of slopes was rejected at a significant level of p < 0.001. For the intramuscular group of replicates, rejection of slope equality could in part be explained by not maintaining a standard dosing regime. The rejection of equivalence of dose-response curves from replicate studies showed that it is difficult to reproduce dose-response data of a single study within the limits defined by the dose-response model. This has important consequences for quantitative risk assessment, public health measures, or development of mechanistic theories which are typically based on a single animal bioassay.