Conclusions and Recommendations
CLASSIFICATION OF TCDD AS CARCINOGENIC TO HUMANS
In its charge, the committee was requested to comment specifically on the U.S Environmental Protection Agency (EPA) conclusion that 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD, also referred to as dioxin) is best characterized as “carcinogenic to humans.” Both EPA and the International Agency for Research on Cancer (IARC), an arm of the World Health Organization (WHO), have established criteria for qualitatively classifying chemicals into various categories based on the weight of scientific evidence from animal, human epidemiological, and mechanism or mode-of-action studies. In 1997, an expert panel convened by IARC concluded that the weight of scientific evidence for TCDD carcinogenicity in humans supported its classification as a Class 1 carcinogen—“carcinogenic to humans.” In 1985, EPA classified TCDD as a “probable human carcinogen” based on the data available at the time, but in the latest Reassessment (2003),1 EPA concluded that TCDD was “best characterized as ‘carcinogenic to humans.” The National Toxicology Program (NTP 2000) also classified TCDD as “known to be a human carcinogen.”
After reviewing EPA’s 2003 Reassessment and other scientific information and in light of EPA’s recently revised 2005 Guidelines for Carcinogen
Risk Assessment (cancer guidelines), the committee concludes that the classification of TCDD as “carcinogenic to humans”—a designation suggesting the greatest degree of certainty about carcinogenicity—versus “likely to be carcinogenic to humans”—the next highest designation—is somewhat subjective and depends largely on the definition and interpretation of the criteria used for classification. The true weight of evidence lies on a continuum, with no obvious point or “bright line” that readily distinguishes between those two categories.
Referring to the specific definitions in EPA’s 2005 cancer guidelines for qualitative classification of chemical carcinogens, the NRC committee was split on whether the evidence met all the criteria necessary for classification of TCDD as “carcinogenic to humans,” although the committee unanimously agreed on a classification of at least “likely to be carcinogenic to humans.” The committee concludes that the weight of epidemiological evidence that TCDD is a human carcinogen is not strong, but the human data available from occupational cohorts are consistent with a modest positive association between relatively high body burdens of TCDD and increased mortality from all cancers. Positive animal studies and mechanistic data provide additional support for classification of TCDD as a human carcinogen. The committee recommends that EPA summarize its rationale for concluding that TCDD satisfies the criteria set out in its cancer guidelines for designation as either “carcinogenic to humans” or “likely to be a human carcinogen.”
If EPA continues to designate TCDD as “carcinogenic to humans” under the new guidelines, it should explain whether this conclusion reflects a finding that there is a strong association between TCDD exposure and human cancer or between TCDD exposure and key precursor events of TCDD’s mode of action (presumably aromatic hydrocarbon receptor [AHR] binding). If its finding reflects the latter association, EPA should explain why that end point (e.g., AHR binding) represents a “key precursor event.”
As noted above, the committee concludes that the distinction between these two categories is based more on semantics than on science and recommends that EPA focus its energies and resources on more carefully delineating the assumptions used in quantitative risk estimates for TCDD, other dioxins, and dioxin-like compounds (DLCs) derived from human and animal studies.
The committee agrees that other dioxins and DLCs are most appropriately classified as “likely to be carcinogenic to humans.” If EPA continues to classify TCDD as “carcinogenic to humans,” more justification will be required to explain why a mixture containing TCDD would not also meet the classification of “carcinogenic to humans.”
USE OF LOW-DOSE LINEAR VERSUS THRESHOLD (NONLINEAR) EXTRAPOLATION MODELS FOR QUANTITATIVE CANCER RISK ESTIMATIONS
The committee unanimously agrees that the current weight of evidence for TCDD, other dioxins, and DLCs carcinogenicity favors the use of nonlinear methods for extrapolation below the point of departure (POD) of mathematically modeled human or animal data. However, the committee recognizes that it is not scientifically possible to exclude totally a linear response at doses below the POD, so it recommends that EPA provide risk estimates using both approaches and describe their scientific strengths and weaknesses to inform risk managers of the importance of choosing a linear vs. nonlinear method of extrapolation. To the extent that EPA favors using default assumptions for regulating dioxin as though it were a linear carcinogen, such a conclusion should be made as part of risk management. EPA should strictly adhere to the distinction between risk assessment, which is a scientific activity, and risk management, which takes into account other factors.
USE OF THE 1% RESPONSE LEVEL AS A POINT OF DEPARTURE FOR LOW-DOSE RISK ESTIMATION
The Reassessment adopts the benchmark dose (BMD) method to replace the traditional, less quantitative approach of using no-observed-adverse-effect level (NOAEL) and lowest-observed-adverse-effect level (LOAEL) to characterize noncancer effects. A BMD (or an effective dose) can be calculated mathematically from a fitted dose-response model and is not limited to the experimental doses. The BMD method is a significant advance in dose-response modeling, and EPA’s use of BMD is highly commendable. However, the determination of an ED at the 1% response level (ED01) for continuous noncancer effects is not without significant limitation. Specifically, the ED01 was the dose associated with a change in mean response away from the background level by 1% of the maximum possible total response range. For some noncancer end points, the significance of such a change may be difficult to identify both clinically and statistically and can be well within the variation of the control data. The biological significance of this magnitude of change represented by the ED01 values for different continuous noncancer end points should be evaluated.
The adoption of such a novel approach gave extremely low margin-of-exposure (MOE) values compared with background exposures and was used by EPA as justification for not analyzing and interpreting the MOE values for each end point and also for not using the massive dioxin database to set a reference dose (RfD).
In its evaluation of the ED01 used for cancer risk assessment, the committee concluded that EPA had not adequately justified use of the 1% response level as the POD for the analysis of either the epidemiological or the animal bioassay data. Even though it is necessary to demonstrate that the POD is within the range of the observed data, that by itself is not sufficient to justify use of the ED01. Other conditions, such as demonstrating that the POD is relatively insensitive to functional form (as noted in the cancer guidelines [EPA 2005a]), must also be satisfied. EPA should acknowledge the larger extrapolation from justifiable POD values down to environmentally relevant doses that would be necessitated by use of a higher response-level POD.
With regard to EPA’s review of the animal cancer bioassay data, the committee recommends that EPA establish clear criteria for the inclusion of different data sets. The reliance on data for one site from one gender of one species, as reported by a single study, does not adequately represent the full range of data available. The committee recommends that EPA consider the full range of data, including the new NTP animal bioassay studies on TCDD, for quantitative dose-response assessment.
For the various noncancer end points, EPA should describe more clearly how and why the ED01 values were determined in animals and translated to human equivalents. At the least, the risk assessment should provide more apparent and parallel calculations using a 5% response level as the POD to demonstrate the impact that this assumption might have on both the point estimates of risk at low doses and the range of uncertainty surrounding that point estimate. This recommendation applies to extrapolation for cancer risk estimates, for which an ED01 was also used, as well as for noncancer risk estimates.
Although the committee commends EPA’s extensive efforts on dose-response modeling of a large number of data sets, particularly those of noncancer end points, it is concerned that selection of the final model for computing POD values was not based on a statistical assessment of model goodness of fit, particularly at low doses. An inadequately fitted model could substantially alter extrapolation to low doses and therefore is a source of error that can result in significant uncertainty. The committee recommends using statistically rigorous methods for assessing model fit to control and reduce this source of uncertainty related to selection of a POD. Although the committee encourages EPA to use thorough statistical analyses of data, it also cautions that “statistical significance” does not always equate with “biological significance,” and thus sound scientific judgment, in addition to statistical analysis, is a critical element of data interpretation.
CHARACTERIZATION OF UNCERTAINTY FOR RISK ESTIMATES
Overall, the committee found that the Reassessment qualitatively addressed many sources of uncertainty and variability but that it failed quantitatively to sufficiently address uncertainty and variability that resulted from the numerous decisions EPA made in deriving point estimates of risk in the comprehensive risk assessment. In contrast, EPA used concerns about uncertainties and uncertainty factors as part of the justification for not setting an RfD for noncancer effects (see Chapter 7 for further discussion).
The committee recommends that EPA provide statistical estimates of the upper-, lower-, and central-bound risk estimates for all quantitative risk estimates. In light of the magnitude of this uncertainty, the committee considers identification of a point estimate value for the dioxin cancer slope factor (CSF), even a point estimate designated as an upper bound, to confer a false sense of precision. EPA should identify the sources of uncertainty and quantitatively characterize their impact on the probability distribution that describes the set of plausible CSF values. If necessary, EPA should acknowledge that the information available is not sufficient to support designation of a meaningful point estimate.
The committee recommends that EPA more completely characterize uncertainty associated with cancer risk estimates inferred from the epidemiological data (1) by taking into account the full range of ED values statistically consistent with the data (not only the central and lower estimates); (2) by considering alternative PODs; (3) by considering biologically plausible alternative dose-response functional forms consistent with the data; and (4) by considering uncertainty associated with the half-life estimates of TCDD in humans for the purpose of back-extrapolating exposures in occupational cohort studies.
The Reassessment did not provide details about the magnitudes of the various uncertainties surrounding the decisions that EPA makes about dose metrics (e.g., the impact of species differences in percentage body fat on the steady-state concentrations present in nonadipose tissues). The committee recommends that the Reassessment use simple physiologically based pharmacokinetic models to define and characterize the uncertainty of any differences between humans and rodents in the relationship between total body burden at steady state (as calculated from the intake, half-life, and bioavailability) and tissue concentrations; EPA should modify the estimated human equivalent intakes when necessary. While PBPK modeling may itself introduce uncertainty, the process of building the PBPK model should help to reduce the far greater uncertainty and likelihood of error that arises when PBPK considerations are not included explicitly. Many opportunities exist to further characterize sources of uncertainty and variability related to the dose metric choices, and the committee recommends that EPA improve
the Reassessment by providing a clear evaluation of the impacts of possible choices on the risk estimates.
The committee recommends that EPA make greater use of mechanistic information to assess the biological plausibility of different mathematical models, use more rigorous criteria (e.g., goodness-of-fit test) for selecting a model for deriving a POD, and clearly identify the benchmark response level of toxicological significance for noncancer end points.
USE OF TOXIC EQUIVALENCY FACTORS FOR RISK ESTIMATION OF DLCS AND MIXTURES OF DLCS
Overall, even given the inherent uncertainties, the toxic equivalency factor (TEF) method provides a reasonable, scientifically justifiable, and widely accepted method to estimate the relative toxic potency of dioxins, other than TCDD, and DLCs, relative to TCDD, on human and animal health. However, the Reassessment should acknowledge the need for better uncertainty analysis of the TEF values. The committee also supports a previous recommendation from the EPA Science Advisory Board (SAB) “that, as a follow up to the Reassessment, EPA should establish a task force to build consensus probability density functions for the … chemicals for which TEFs have been established, or to examine related approaches such as those based on fuzzy logic.”
USE OF BODY BURDEN AS THE PRIMARY DOSE METRIC FOR CROSS-SPECIES EXTRAPOLATION
Although the committee agrees that use of body burden as the dose metric is the most reasonable and pragmatic approach at the present time, a number of uncertainties in using body burden to develop risk estimates should be addressed. The magnitudes of the various uncertainties are not clearly defined. The most significant impact is the species differences in percentage body fat on the relationship between body burden and the concentrations present in nonadipose tissues. An analysis of the impact of possible uncertainties in the dose metric on the final risk estimates would be informative.
It remains to be determined whether the current WHO TEFs, which were developed to assess the relative toxic potency of a mixture to which an organism is directly exposed by dietary intake, are appropriate for body burden toxic equivalent quotient (TEQ) determinations, which are derived from the concentrations of different congeners measured in body fat. If body burdens are to be used as the dose metric, a separate set of body burden TEFs should be developed and applied for this evaluation. Without these corrected values, the overall TEQs estimated by use of intake TEFs might be substantially in error.
EPA’S EXPOSURE ASSESSMENT OF TCDD, OTHER DIOXINS, AND DLCS IN THE UNITED STATES
To assess total emissions of TCDD, other dioxins, and DLCs, EPA used a “bottom-up” approach, which attempted to identify all source categories and then estimated the magnitude of emissions for each category. However, a “top-down” approach that attempts to account for the levels measured in receptors (e.g., people, animals, and plants) could give rise to substantially different information. Such alternative approaches are likely to give rise to significantly different estimates of the historical levels of dioxin emissions. Both approaches come with uncertainties, and EPA could benefit significantly from using them simultaneously to set plausible bounds on the historical and current trends in emissions.
Although beyond the scope of the review of the EPA Reassessment, the committee noted that it would be useful for EPA to set up an active congener-specific database of typical concentrations for the whole range of polychlorinated dibenzo-p-dioxins (PCDDs), polychlorinated dibenzofurans (PCDFs), and dioxin-like polychlorinated biphenyls (PCBs) (included in the WHO TEF list) present in food. This database should be based on a compendium of all available data that would be updated on a regular basis with new data as they are published in the peer-reviewed literature. Such a database should have clear requirements of data quality and traceability (e.g., chemical analysis, representative and targeted sampling, representative of consumer exposure, presentation of data, and handling and presentation of nondetects).
The committee suggests that in the future EPA define a strategy for collecting samples and reanalyzing archived samples to answer a number of remaining questions about exposure trends and to fill some important data gaps.
EPA’S EVALUATION OF IMMUNOTOXICITY OF TCDD, OTHER DIOXINS, AND DLCS
Present clinical findings are inconclusive about whether or in what way TCDD, other dioxins, and DLCs are immunotoxic in humans, and EPA acknowledges that human data are sparse. A series of studies from a Dutch children’s cohort showed an association between prenatal exposure to DLCs and changes in immune status. The effects were modest, and laboratory values did not fall significantly outside the full range of normal. Some clinically relevant adverse effects seen in this perinatal study are also seen at higher levels of exposure, although these do not seem to persist. A number of animal studies suggest that the developing immune system is especially sensitive. In light of the large database showing that TCDD, other dioxins,
and DLCs are immunotoxic in laboratory animal studies—together with limited human data—EPA is prudent in concluding that these compounds are likely to be human immunotoxicants in the absence of more definitive human data.
However, EPA’s conclusion that TCDD and related compounds are immunotoxic at “some dose level” by itself is inadequate. At a minimum, a section or paragraph should be added that discusses the immunotoxicology of TCDD, other dioxins, and DLCs in the context of current AHR biology.
Likewise, some discussion should also be included on the strengths and weaknesses of using genetically homogeneous inbred mice to characterize immunotoxicological risk in the genetically variable human population. Expanding the discussion to include the above crucial points would provide additional balance to Part III, Integrated Summary and Risk Characterization.
Additional comments and recommendations relating to the use of specific data sets for risk assessment of the immunotoxic effects of TCDD, other dioxins, and DLCs are provided in Chapter 6.
EPA’S EVALUATION OF REPRODUCTIVE AND DEVELOPMENTAL TOXICITY OF TCDD, OTHER DIOXINS, AND DLCS
As clearly described in the Reassessment, embryonic and fetal development and reproductive effects are sensitive end points of TCDD toxicity in rodents. It is clear that the fetal rodent is more sensitive than the adult rodent to adverse effects of TCDD. Comparable human data are generally lacking, and the sensitivity of humans to these end points is less apparent.
The committee recommends that EPA address more thoroughly how the effective doses used in the animal pregnancy models relate to human reproductive and developmental toxicity and risk information, including TEFs and TEQs. For available human clinical data on reproductive and developmental end points, EPA should establish formal principles of, and a formal mechanism for, evidence-based classification and systematic statistical review, including meta-analysis when possible.
Finally, EPA should discuss the dose-response effects of TCDD, other dioxins, and DLCs on the adult female reproductive system that result in endocrine disruption in animals. Based on the dose-response data provided in these studies, the impact on human risk assessment should be presented.
EPA’S EVALUATION OF OTHER TOXIC END POINTS
In general, the committee determined that the Reassessment adequately addressed the available data on whether exposures to TCDD, other dioxins, and DLCs are likely to be significant risk factors for other toxic end points, such as chloracne, thyroid function, liver function, diabetes, lipid
disorders, and cardiovascular diseases. In humans, the relationship between dioxin exposure and risk of individual, clinically significant, noncancer end points remains uncertain, except for chloracne.
The overall conclusions in the Reassessment about noncancer risks due to exposure to TCDD, other dioxins, and DLCs are, in general, cautiously stated, and the uncertainty of suspected relationships is acknowledged. Nonetheless, the limitations of individual human studies are not uniformly addressed, and the broad 95% confidence intervals accompanying some reported statistically significant effects are not discussed in the context of the uncertainty that these broad confidence limits imply. Conversely, statistically nonsignificant effects are sometimes highlighted, presenting an implied potential for unobserved detrimental effects without a firm evidence base. For available human clinical data for other noncancer end points, EPA should establish formal principles of, and a formal mechanism for, evidence-based classification and systematic statistical review, including meta-analysis when possible.
With respect to human noncancer end points, the committee determined that the Reassessment text should be revised to include the relevant, more recent data and, when appropriate, the quality and data uncertainty of the studies referenced. When the mechanism is established, currently available and newly available human clinical studies should be subject to such systematic review and formal evidence-based assessment. The quality of the available evidence should be reported, and the strength or weakness of a presumptive association should be classified according to currently accepted criteria for levels of evidence.
New studies on effects of TCDD on the developing vascular system suggest that this system could be a highly sensitive target and suggest that this area be identified as an important data gap in the understanding of the potential adverse effects of TCDD, other dioxins, and DLCs.
EPA’S OVERALL APPROACH TO RISK CHARACTERIZATION
As discussed above, EPA used linear extrapolation from the POD (the ED01) derived from the cancer epidemiological studies and animal bioassays to calculate a CSF. The selection of the default linear extrapolation approach was one of the most critical decisions in the Reassessment, but the decision to use this approach was not supported by a scientifically rigorous argument, nor was there a balanced presentation of arguments that would support the calculation and interpretation of a MOE with the same data. The committee determined that a balanced presentation of available data could support the use of a nonlinear model consistent with a receptor-mediated mode of action with subsequent calculations and interpretation of
MOE values. (For cancer risk assessment, the threshold approach should be used in addition to the linear approach.)
Because EPA decided not to define an RfD, the Reassessment lacked detailed risk characterization information—for example, the proportion of the population with intakes above the RfD; detailed assessment of population groups, such as those with occupational exposures; and the contributions of major food sources and other environmental sources for those people with high intakes. The lack of such a focus in the Reassessment results in a diffuse risk characterization that is difficult to follow and that does not provide clear advice to risk managers.
The committee points out particular areas that could be improved in Part III of the Reassessment. In particular, the risk characterization chapter of the Reassessment should describe concisely and clearly the following aspects.
The effects seen at the lowest body burdens that are the primary focus for any risk assessment—the “critical effects.”
The modeling strategy used for each noncancer effect modeled, paying particular attention to the critical effects, and the selection of a point of comparison based on the biological significance of the effect; if the ED01 is retained, then the biological significance of the response should be defined and the precision of the estimate given.
The precision and uncertainties associated with the body burden estimates for the critical effects at the point of comparison, including the use of total body burden rather than modeling steady-state concentrations for the relevant tissue.
The committee encourages EPA to calculate RfDs as part of its effort to develop appropriate margins of exposure for different end points and risk scenarios, including the proportions of the general population and of any identified groups that might be at increased risk (See Table A-1 in the Reassessment, Part III Appendix, for the different effects; appropriate exposure information would need to be generated.) Interpretation of the calculated values should take into consideration the uncertainties in the POD values and intake estimates.
Consideration of individuals in susceptible life stages or groups (e.g., children, women of childbearing age, and nursing infants) who might require estimation of a separate MOE using specific exposure data.
Distributions that provide clear insights about the uncertainty in the risk assessments, along with discussion of the key contributors to the uncertainty.
The committee recognizes that it will require a substantial amount of effort by EPA to incorporate all the changes recommended in this review;
however, it does not advocate a substantial expansion in the length of the Reassessment. Rather, the committee encourages EPA to address the major concerns raised in this review and to finalize the current Reassessment as quickly, efficiently, and concisely as possible. The committee agreed that it is important for EPA to recognize that new advances in the understanding of the toxicity of TCDD, other dioxins, and DLCs could require reevaluation of key assumptions in the risk assessment document. The committee recommends that EPA routinely monitor new scientific information with the understanding that future revisions may be required to maintain a risk assessment that is based on current state-of-the-science. However, the committee also recognizes that stability in regulatory policy is important to the regulated community and thus expects that science-based changes in regulatory policy on TCDD, other dioxins, and DLCs will be invoked only in the face of compelling new information that would warrant revision of its final risk assessment. Such substantial gains in knowledge are not likely to occur frequently.