Review of Risk Characterization
The Reassessment1 (Part III, Chapter 6) considers risk characterization under a series of headings, many of which represent summaries of the inputs to risk characterization instead of the output of risk characterization and its formulation of advice to risk managers. For convenience, this chapter uses the same headings for the committee’s review of the Reassessment’s risk characterizations before presenting its conclusions. Because Chapter 6 in the Reassessment summarizes data from previous chapters, many of the committee’s comments here were raised in previous chapters of this report.
TCDD, Other Dioxins, and DLCs Can Produce a Wide Variety of Effects in Animals and May Produce Many of the Same Effects in Humans
Reassessment (Part III, pp. 6-1 to 6-4)
This introductory text sets the scene by stating:
Effects will likely range from detection of biochemical changes at or near background levels of exposure to detection of adverse effects with increasing severity as the body burdens increase above background levels. (Part III, p. 6-1, lines 28-30)
Clearly adverse effects, including, perhaps, cancer, may not be detectable until exposures contribute to body burdens that exceed current backgrounds by one or two orders of magnitude (10 to 100 times). (Part III, p. 6-2, lines 11-13)
The rationale for those statements is not clearly defined, although the Reassessment states later that few clinically significant effects were detected in the small number of human cohorts studied, nearly all members of which had body burdens significantly above background levels.
The text considers species differences in sensitivity to 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD, also referred to as dioxin), other dioxins, and dioxin-like compounds (DLCs) and concludes that “humans most likely fall in the middle rather than at either extreme of the range of sensitivity for individual effects among animals” (p. 6-2, lines 25-26). A general comparison across several species is not relevant to the focused issue of risk characterization. The comparisons of importance are those between humans and the species and strains used in the specific studies that revealed adverse effects at the lowest levels of exposure, the so-called critical effects (see below). This general statement on interspecies differences detracts from a critical and quantitative assessment of differences in sensitivity between humans and the species used in the key toxicological studies for risk characterization.
Overall the committee considered this introductory section to be reasonable but unfocused.
TCDD, Other Dioxins, and DLCs are Structurally Related and Elicit Their Effects Through a Common Mode of Action
Reassessment (Part III, pp. 6-4 to 6-5)
The text is uncontroversial and concludes that binding to the aromatic hydrocarbon receptor (AHR) appears to be necessary but is not sufficient to elicit the various TCDD-induced effects. The committee agrees with this conclusion.
EPA and the International Scientific Community Have Adopted Toxic Equivalency of TCDD, Other Dioxins, and DLCs as a Prudent Scientific Policy
Reassessment (Part III, pp. 6-5 to 6-6)
The text summarizes the current situation and is uncontroversial. Obviously, given the date of the Reassessment, the text has not considered planned updates to toxic equivalency factor (TEF) values and whether these
will be used in future risk assessments of TCDD, other dioxins and DLCs. The committee’s recommendations related to TEFs appear in Chapter 3.
Complex Mixtures of TCDD, Other Dioxins, and DLCs Are Highly Potent, Likely To Be Carcinogens
Reassessment (Part III, pp. 6-6 to 6-12)
The Reassessment states that “because TCDD, other dioxins, and DLCs always occur in the environment and in humans as complex mixtures of individual congeners, it is appropriate that the characterization [likely carcinogen] apply to the mixture” (p. 6-6, lines 19-21). Therefore, despite the attention given by EPA and hence by this committee to consideration of whether TCDD is “carcinogenic to humans” or “likely to be carcinogenic to humans,” the reality is that TCDD is always present as part of a mixture of TCDD, dioxins, and DLCs, and, therefore, the practical hazard characterization of human exposure to TCDD is in effect considered by EPA “likely to be carcinogenic.” In consequence, the focus on qualitative classification of the nature of the cancer hazard by EPA has been a somewhat futile exercise. The text then further discusses this issue and concludes that the “likely to be carcinogenic” classification could differ in strength, depending on the constituents in the mixture. Subsequent text reconfirms that TCDD is classified by EPA as “carcinogenic to humans” and outlines the evidence used to reach this conclusion, including the presence of “strong and consistent” evidence from occupational epidemiological studies (a point on which the committee does not agree; see Chapter 5). The committee concluded that such detailed consideration of hazard classification was of little value in a section on risk characterization, especially as the difference between “carcinogenic” and “likely to be carcinogenic” would not have a significant impact on the formulation of advice under risk characterization.
This section continues by presenting the upper bound of the cancer risk estimate of 1 × 10−3 per pg of toxic equivalent quotient (TEQ)/kg of body weight/day for both background intakes and incremental intakes above background. The value is based on the range of cancer slope factors (CSFs) developed from linear modeling of the occupational cohort data. The Reassessment states, “Evaluations of shape parameters…for biochemical effects that can be hypothesized as key events in a generalized dioxin mode-of-action model do not argue for significant departures from linearity below a calculated ED01, extending down to at least one to two orders of magnitude lower exposure” (p. 6-8, lines 31-33, to 6-9, lines 1-2). This sentence appears to be critical to the risk characterization approach for cancer adopted by EPA, but there is no scientific assessment of the strength of the available evidence to support that statement, of the ability of the model-fitting meth-
ods used by EPA to detect a departure from linearity, were one to exist; or of the indications of nonlinearity in the dose-response relationships for many noncancer effects also considered to be mediated via an interaction with the AHR. The Reassessment attempts to explain the decision to use a linear approach (Part III, p. 5-15, lines 27-29) by stating that “The linear default is selected on the basis of the agent’s mode of action when the linear model cannot be rejected and there is insufficient evidence to support an assumption of nonlinearity.” Quantitative evidence of nonlinearity below the point of departure (POD), the ED01 (effective dose), will never be available because the POD is chosen to be at the bottom end of the available dose-response data. As discussed in Chapter 5 of this report, EPA should give greater weight to knowledge about the mode of action and its impact on the shape of the dose-response relationship. The committee considers that the absence of evidence that argues against linearity is not sufficient justification for adopting linear extrapolation, even over a dose range of one to two orders of magnitude or to the assumption of linearity through zero, which would not normally be applied to receptor-mediated effects. This view is supported by the results of the recent cancer bioassay (NTP 2004), which was not available to EPA at the time of the Reassessment but which could have a major impact on the risk assessment approach adopted by EPA.
The text compares the current estimate with previous EPA estimates and uses previous evaluations with the same approach to support the outcome of the Reassessment. The difference between EPA’s evaluation and that of the Food and Agricultural Organization of the United Nations (FAO)/World Health Organization (WHO) Joint Expert Committee on Food Additives (JEFCA) (which considered TCDD to be a nongenotoxic carcinogen and an uncertainty factor approach to be adequate to account for both cancer risk and noncancer effects) was thought to “reflect differences in science policy” (p. 6-11, line 2). The Reassessment did not attempt to explain why EPA has chosen to use an uncertainty factor approach for the risk characterization of other nongenotoxic, receptor-mediated carcinogens with a known mode of action (such as for thyroid carcinogens) but not for TCDD. The Reassessment suggests that a margin-of-exposure (MOE) approach should be adopted for both cancer and noncancer effects but does not explore the implications of the estimated MOE for cancer or the ability of the MOE approach to refine the advice for population groups.
The use of different methods for the risk characterization of end points that result from the same basic underlying mode of action is scientifically illogical, a conclusion that seems to be supported by EPA in an earlier part of the Reassessment (Part III, p. 5-3, lines 18 to 28).
Although the Reassessment defines a slope factor and cancer risk estimate, it does not spell out clearly the health implications of emphasizing
this approach for the U.S. population. The Reassessment states that the slope factor has a “public health-conservative nature” (p. 6-9, line 14), but such risk management considerations should not be used to support an approach to risk characterization or detract from selection of the most appropriate, scientifically justifiable approach.
The text discusses the consistency of the present slope factor with previous EPA evaluations and further discusses the issue of hazard classification, as if the decision to define a CSF affected the hazard classification (or vice versa), which illustrates the lack of clarity and focus in this part of the Reassessment.
The Reassessment recognizes that “the shape of the dose-response curve below the range of observation can be inferred only with uncertainty” (p. 6-12, lines 1 to 2), and therefore the Reassessment should have given equal weight and critical evaluation to the derivation of a CSF and to the calculation of the MOE with a discussion of the adequacy of the MOE of an exposure in relation to the remaining uncertainties.
In future revisions of the Reassessment, aspects that should be discussed for both approaches include the known mode of action, the adequacy of the occupational cohorts to represent the whole population, the integration of data from the animal cancer bioassays (including the most recent study) in relation to the spectrum of cancers detected, and the shape of the dose-response relationship.
Use of a MOE Approach to Evaluate Risk for Noncancer and Cancer End Points
Reassessment (Part III, pp. 6-12 to 6-18)
Despite the title of this section, there is no focused discussion in the Reassessment of the MOE in relation to cancer or to each of the end points, using exposure data relevant to that end point (Part III, Appendix A, Table A-1). In addition, there is no discussion of the areas of uncertainty that would need to be taken into account for each study and end point. For example, the MOE for cancer, and possibly for immune and neurodevelopmental effects, would be based on epidemiological data, whereas MOEs for noncancer effects would be based largely on data from animal studies. Other issues that should be discussed in the interpretation of the MOE for each end point are the relevance of the effect to the general population and to population groups and life stages and, most important, the clinical significance of the magnitude of the effect detected at the ED01 (if this is retained as a point of comparison on the dose-response relationship; see committee comments earlier in this report).
The Reassessment concludes that setting a reference dose (RfD) is not
appropriate because of the relatively high background levels of exposure compared with effect levels (Reassessment, Part III, p. 6-14) and that “any RfD that the Agency would recommend using a traditional approach for setting an RfD using uncertainty factors to account for limitations of knowledge is likely to be below—perhaps significantly below (by a factor of 10 or more)—current background intakes and body burdens.” EPA thus concluded that setting an RfD would be of little value for evaluating possible risk management options when the RfD has already been exceeded by average background exposure. This issue is not resolved by simply replacing the RfD by an MOE without analyzing whether the estimated MOE is adequate for that particular end point based on the data used to derive the point of comparison on the dose-response relationship.
The magnitude of this apparent problem arises from two aspects of the Reassessment: the use of ED01 as the point of comparison with exposure for continuous variables, which in many cases is two orders of magnitude below the lowest-observed-adverse-effect level (LOAEL) (Reassessment, Appendix A, Table A-1), and the use of the usual default uncertainty factors despite the wealth of data available on TCDD. The issue of the derivation and suitability of the ED01 for continuous variables was discussed in Chapter 2 of this report. EPA has not justified its use for risk assessment and its replacement of traditional measures, such as the no-observed-adverse-effect level (NOAEL), LOAEL, or BMD10 (benchmark dose) or BMD05, as a point for establishing an RfD or for comparison with human exposures by calculation of MOEs. Selection of appropriate uncertainty factors is discussed further below.
An additional problem identified by the Reassessment is that “the calculation of an RfD (with its traditional focus on a single “critical” effect) distracts from the large array of effects associated with similar body burdens of dioxin” (Part III, p. 6-14, lines 20-22). This statement appears to contradict EPA’s well-established approach of focusing on the critical effect as a basis for setting health protective values. The problem applies to some degree to many other chemicals that show multiple effects over a narrow dose range and does not invalidate the selection of a “critical effect” from the TCDD database, which would lead to a more focused discussion and risk characterization. A critical effect of postnatal reproductive changes after in utero exposure was identified from the available dose-response data by the European Scientific Committee on Food (SCF 2000, 2001) and by JECFA (2002). These bodies concluded that developing a health-based guidance value (equivalent to an RfD) from reproductive effects in rats after in utero exposure would also cover the risks of other effects (including cancer) detected at slightly higher body burdens. In principle, the case of TCDD is no different from that of other contaminants that produce multiple adverse effects. Because the exposures of a proportion of the U.S. population would
be above any RfD, it would have been useful for EPA to define the nature and magnitude of the risks at different levels of intake, the groups of the population most at risk, and the major sources of exposure for any at-risk groups. Alternatively, if MOEs were calculated for noncancer effects, then the risk characterization should describe the nature of the adverse effect and the uncertainties and variability inherent in both the BMD (ED) estimate and the relevant exposure estimate. It would have been useful if MOE values had been calculated and discussed for different exposure scenarios.
The Reassessment discusses the approaches adopted by the SCF and WHO (IPCS1998b) and by the Agency for Toxic Substances and Disease Registry (ATSDR 1998) (Reassessment, Part III, pp. 6-16 to 6-18). The more recent JECFA (2002) evaluation was considered with reference to the date of the meeting in 2001. The Reassessment highlights three sources of differences:
An initial focus on cancer or noncancer effects.
The use of intake or body burden.
The “safety” or “uncertainty” factors used.
The recent evaluations by SCF and JECFA used the approach proposed by the WHO International Programme on Chemical Safety (IPCS) in which the usual 10-fold default uncertainty factors are subdivided into toxicokinetic and toxicodynamic subfactors, which can be replaced when suitable chemical-specific data are available (IPCS 1994, 1999, 2004; WHO 2005). In this approach, the 10-fold interspecies factor is divided into 4.0 for toxicokinetics and 2.5 for toxicodynamics, the product (10) being used in the absence of chemical-specific data to replace either of the default values; the 10-fold human variability factor is subdivided equally into 3.2 for toxicokinetics and 3.2 for toxicodynamics. Subdividing the 10-fold default uncertainty factors was done by EPA in its recent evaluation of boron (EPA 2004b), although a slightly different split of the 10-fold interspecies factor was used. (This difference between EPA and WHO would not have altered significantly the health-based guidance value for TCDD, other dioxins, and DLCs that was derived by SCF and JECFA.)
The Reassessment considers briefly the rationale for the uncertainty factors used in the recent SCF and JECFA evaluations. These evaluations concluded that interspecies differences in toxicokinetics had been taken into account by the use of body burden as the dose metric instead of the external dose, and therefore this subfactor would become 1 instead of 4.0. The Reassessment does not discuss the explanation of why SCF and JECFA concluded that “no uncertainty factor needed to be applied for differences in toxicodynamics between experimental animals and humans and for interindividual variation [in toxicodynamics] among humans” (Reassess-
ment, Part III, p. 6-17, lines 21-23). The rationale given in the JECFA monograph (JECFA 2002, p. 590) was that, in general, rats are more sensitive than humans to the adverse effects of TCDD, and therefore the interspecies factor might be less than 1, but that “it cannot be excluded that the most sensitive humans might be as sensitive to the adverse effects of TCDD as rats were in the pivotal studies. Therefore, it was concluded [by the JEFCA] that no safety factor in either direction need to be applied for differences in toxicodynamics among humans.” In other words, any possible variability in toxicodynamics among humans would be compensated for by the higher inherent sensitivity of the rat strains used in the pivotal studies compared with average humans, and each of these subfactors would become 1. Of the four aspects for which the usual 100-fold uncertainty factor is applied, the only one for which data was considered to be inadequate related to human variability in toxicokinetics. SCF and JECFA applied the default value of 3.2 for human variability in toxicokinetics.
While the approach adopted by the SCF and JECFA is open to criticism because of its simplicity, the attempt to incorporate the wealth of data on TCDD into the risk assessment process contrasts with EPA’s assumption that default values would be used, and hence the RfD would be below the current levels of exposure. The Reassessment (Part III, p. 6-18, lines 6-9) states, “In particular, the focus on accounting for residual toxicodynamic differences in cross-species scaling and interindividual variability in the general population to account for sensitive individuals, including children” would suggest larger uncertainty factors than have been proposed by these groups if EPA were to set an RfD. However, EPA does not discuss how the usual uncertainty factors might be modified using the TCDD database and does not give an analysis of the uncertainty factors that it would use and justification for their use. The Reassessment does not discuss whether or not the EPA considered how the uncertainty factors or other aspects of risk characterization could be revised based on probabilistic approaches.
The Reassessment does not evaluate critically the extent of species differences in target organ sensitivity, especially in relation to the pivotal studies and critical effects. Overall, there is inadequate discussion of the relative affinities of the AHRs in rats and humans and of the possible impact of polymorphisms in AHR and other sources of sensitivity differences within humans. The Reassessment (Part III, p. 6-18, line 4-5) states, “Traditional approaches that might be applied by EPA or that have been applied by ATSDR would likely require additional information to support the choice or removal of uncertainty factors as performed by WHO, SCF and JECFA.” However, there is no critical discussion of the limitations of the available data that might be used to move away from the traditional
uncertainty factors. The Reassessment does not give the rationale for EPA’s decision not to replace the default uncertainty factors by chemical-specific data, despite the enormity of the TCDD database.
The Reassessment (Part III, p. 6-18, lines 10-14) concludes that “any composite uncertainty factor greater than 10 applied to effect levels based on body burden … would result in TDI or MRLs below the current background intakes. The use of uncertainty factors in the range of 30 to 100 or more, as traditionally used by EPA, would result in values even further below some current background body burdens.”
The Reassessment concludes the risk characterization section (Part III, p. 6-34) by stating that the MOEs based on body burden are less than 1 for enzyme induction in rats and mice and less than 4 for developmental effects in rats and endometriosis in nonhuman primates. The reader is left to compare those values with uncertainty factors in the range of 30 to 100, which EPA would traditionally use, with no clear and concise guidance on the interpretation of this information. However, these judgments are based on the nontraditional use of ED01 in place of a BMD5, BMD10, NOAEL, or LOAEL.
Children’s Risk from Exposure to TCDD, Other Dioxins, and DLCs May Be Increased, but More Data Are Needed to Address This Issue
Reassessment (Part III, pp. 6-18 to 6-21)
The Reassessment highlights the greater susceptibility of in utero, perinatal, and neonatal life stages on the basis of animal and human epidemiological data. The Reassessment does not clarify the additional data that would be required before an RfD could be established or before definitive advice could be given about the adequacy or inadequacy of the MOE for adverse effects detected in animal studies after in utero exposure. Following these general doubts about the possible heightened susceptibility of neonates and children, the Reassessment comments on the greater exposure of nursing infants and children but concludes that, because the risk characterization is based on body burden, the short-term intake levels will have little impact on risk compared with overall lifetime exposure. The committee noted that EPA did not define the MOE for these life stages and that, overall, this section raises concerns about hypothetical, additional, undefined susceptibility while allaying concerns about the considerably greater exposures of infants and children compared with adults.
Background Exposures to TCDD, Other Dioxins, and DLCs Need To Be Considered When Evaluating Hazard and Risk
Reassessment (Part III, pp. 6-21 to 6-23)
This section of the Reassessment provides a summary of the extent of background exposures but does not adequately integrate the information into an MOE or an estimate of population cancer risk using the slope factor.
Evaluating the Exposure of “Special” Populations and Developmental Stages Is Critical to Risk Characterization
Reassessment (Part III, pp. 6-23 to 6-25)
The Reassessment describes sources of variability in exposure and intake—for example, contaminated poultry feed, increased consumption of fish, and occupational exposures. This section concludes that a high intake would be about three times the population mean, but again there is no quantification of the MOE or any attempt to link high-exposure groups to specific end points (except for breast-feeding, which is covered in the following section).
Breast-Feeding Infants Have Higher Intakes of TCDD, Other Dioxins, and DLCs for a Short but Developmentally Important Part of Their Lives but the Widely Recognized Benefits of Breast-Feeding Outweigh the Risks
Reassessment (Part III, pp. 6-26 to 6-27)
The Reassessment reiterates the information on breast-feeding and points out that the average daily intake by the infant over the first year of suckling would be 87 times the adult daily intake. It correctly points out that this would not result in an 87-fold higher body burden because of the rapid increase in body weight and more rapid elimination. The Reassessment reiterates the advantages of breast-feeding in general and concludes that reevaluation of TCDD, other dioxins, and DLCs does not alter the previous advice, especially because the risk assessment is based on body burden. While not disagreeing with the conclusion, the committee considers the Reassessment to be superficial on this point. It does not support its position with well-founded evidence, it does not consider the impact of body composition (e.g., percent body fat) on distribution of the body burden in infants, and, most important it makes no attempt to compare the intakes by infants with the doses producing adverse effects in the relevant
animal studies (that is, on those involving in utero exposure and subsequent assessment of developmental parameters in early life).
Many Dioxin Sources Have Been Identified and Emissions to the Environment Have Been Reduced
Reassessment (Part III pp. 6-27 to 6-29)
This summary of previously presented information on sources and emissions is adequate. (The committee noted, however, that it is largely irrelevant to this part of the Reassessment because it does not consider or contribute to risk characterization.) See Chapter 4 for the committee’s recommendations on sources and emissions.
TCDD, Other Dioxins, and DLCs Dioxins Are Widely Distributed in the Environment at Low Concentrations Primarily as a Result of Air Transport and Deposition
Reassessment (Part III, pp. 6-29 to 6-30)
This summary of previously presented information on sources and emissions is adequate. (The committee noted, however, that it is largely irrelevant to this part of the Reassessment because it does not consider or contribute to risk characterization.)
Environmental Levels, Emissions, and Human Exposures Have Declined During Recent Decades
Reassessment (Part III, p. 6-30)
This summary of previously presented information on sources and emissions is adequate. (The committee noted, however, that it is largely irrelevant to this part of the Reassessment because the data are not interpreted in the context of risk characterization.)
Risk Characterization Summary Statement
Reassessment (Part III, pp. 6-30 to 6-34)
This section provides a reasonable summary of the preceding parts of Chapter 6 of Part III.
CONCLUSIONS AND RECOMMENDATIONS
The committee considered Chapter 6 of Part III of the Reassessment to be the most important section, but in many ways it was the weakest and least scientifically rigorous in its support of the decisions made.
EPA used linear extrapolation from the POD, the ED01, derived from the cancer epidemiological studies to calculate a CSF. The resulting cancer risk estimate of 1 × 10−3 per pg TEQ/kg of body weight per day for both background intakes and incremental intakes above background was considered by EPA to be the most appropriate approach. Using a linear extrapolation approach in the Reassessment was one of the most critical decisions by EPA. Use of this approach was not supported by a scientifically rigorous argument, nor was there a balanced presentation of arguments using the same data to support the calculation and interpretation of an MOE. EPA did not adequately discuss the risk management implications of the cancer risk estimate, which might be interpreted to indicate the need to reduce the current exposure of the general population between 10-fold and 1,000-fold to limit the calculated cancer risk between 1 in 10,000 and 1 in 1,000,000. Such a use and interpretation of the slope factor would require EPA to consider the validity of the linear model over many orders of magnitude.
The Reassessment stated that it used an MOE approach for noncancer effects, but the discussion did not focus on the MOE values for different adverse effects. An important improvement over past EPA practice was the reliance in the Reassessment on an estimated ED (BMD) for noncancer effects rather than on the traditional NOAEL and LOAEL as the POD. An ED can be calculated mathematically from a fitted dose-response model and is not limited to the experimental doses, thus representing a significant advance in dose-response assessment. However, the computation of the ED01 for continuous noncancer effects was critical, where the ED01 was not the dose associated with a 1% incremental incidence of an adverse effect but was the dose associated with a change in the mean response from the background level that was 1% of the maximum possible total response range. EPA made no attempt to present the biological significance of such changes for each of the different continuous end points of studies subject to dose-response modeling.
The adoption of such a novel approach gave extremely low general MOE values 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 TCDD database to identify an RfD.
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, and contributions of the major food sources for those individu-
als 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 guidance to risk managers.
The Reassessment should describe 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, 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 an estimation of a separate MOE using specific exposure data.
Distributions that provide clear insights about the uncertainty in the risk assessments, along with discussion about the key contributors to the uncertainty.