5
Cancer

This chapter provides the committee’s assessment of the hazard identification and dose-response analysis for cancer endpoints in the 2022 Draft Assessment (EPA, 2022a). The 2022 Draft Assessment presents an evaluation of the evidence for cancer hazard and separately for dose-response in two parts: for the respiratory system and for nonrespiratory sites. This separation is appropriate and concordant with previous National Research Council (NRC) reports, which separately consider the portal-of-entry and systemic effects of formaldehyde (NRC 2011, 2014). These past reports define portal-of-entry effects as those arising from direct interaction of inhaled or ingested formaldehyde with the affected cells or tissues, and systemic effects as those that occur beyond tissues or cells at the portal of entry.

HAZARD IDENTIFICATION

The 2022 Draft Assessment includes detailed evaluation of the evidence from human, animal, and mechanistic studies that pertain to several specific cancer sites (Appendix A, Section A.5.9). Following systematic identification and evaluation of the relevant literature, EPA determined that for the portal-of-entry effects, cancers of the upper respiratory tract (i.e., nasopharyngeal cancer, sinonasal cancer, cancers of the oropharynx and hypopharynx, and laryngeal cancer) would be evaluated in detail. For systemic effects, EPA was determined that cancers of the lymphohematopoietic system (i.e., Hodgkin lymphoma, multiple myeloma, myeloid leukemia, and lymphatic leukemia) would also be evaluated in detail. EPA determined further that the evidence regarding the potential for formaldehyde exposure to cause cancers at other sites (i.e., lung, brain, bladder, colon, pancreas, prostate, skin) and non-Hodgkin lymphoma was highly limited, and therefore did not systematically evaluate these cancers.

In the past 20 years, several authoritative agencies and organizations have classified formaldehyde according to whether it poses a cancer hazard (Table 5-1). Unequivocal independent conclusions that formaldehyde is carcinogenic in humans have been reached by the International Agency for Research on Cancer (IARC) (2006, 2012), the National Toxicology Program (NTP) (2011), and the NRC (2014). The European Union (EU) Committee on Occupational Exposure Limits¹ concluded that formaldehyde poses a human cancer hazard, but with a threshold-based dose-response relationship. Two previous assessments have classified formaldehyde as a presumed² or potential (NIOSH, 1988) human carcinogen. The 2022 Draft Assessment concludes that the evidence demonstrates that formaldehyde inhalation causes cancer in humans and identifies nasopharyngeal and sinonasal cancers, as well as myeloid leukemia, as types of cancers with this level of evidence.

___________________

¹ Commission Directive (EU) 2017/164. 2017. Establishing a fourth list of indicative occupational exposure limit values pursuant to Council Directive 98/24/EC, and amending Commission Directives 91/322/EEC, 2000/39/EC and 2009/161/EU. Official Journal of the European Union L27:115–120.

² Commission Regulation (EU) No 605/2014. 2014. Official Journal of the European Union L167:36–49.

TABLE 5-1 Cancer hazard classifications of formaldehyde

LITERATURE IDENTIFICATION AND EVALUATION OF STUDY METHODS

While the 2022 Draft Assessment has been in development over several decades, most critical studies of cancer in humans and animals were published eight or more years ago. Studies on the potential mechanisms of formaldehyde carcinogenicity represent a dynamic body of literature. Methods and procedures for literature identification and evaluation of each study are presented in several parts of the 2022 Draft Assessment. Chapter 4.1 of the Assessment Overview presents a summary of the process, Sections 1.2.5 and 1.3.3 of the “Toxicological Review of Formaldehyde: Inhalation” provide more granular detail with references to the Appendix for additional information about respiratory tract and nonrespiratory tract cancers. The Appendix provides specific details about the population, exposure, comparator, and outcome (PECO) questions raised and associated inclusion and exclusion criteria for the human and animal studies of health effects, the bibliographic databases, search terms, and specific strategies used to search them (Appendix A, Sections A.4.7, A.5.5, and A.5.9). In addition, Appendix A, Section A.5.9 provides literature flow diagrams that summarize the results of the sorting process using the defined criteria and indicating the number of studies that were selected for consideration in the assessment through 2016. Because the completed 2017 draft assessment was suspended by EPA until 2021, EPA used a systematic evidence map (SEM) to conduct additional searches for any new (January 2016–May 2021) publications to be considered in updating the 2017 draft. The methods and results of this SEM process are provided in Appendix F. With respect to cancer endpoints and mechanistic evidence, several additional studies were identified, and were categorized as “possibly impactful” or “not impactful.”

STUDY EVALUATION

EPA’s evaluation of the studies identified through its literature searches was separated into several domains. Individual observational epidemiological studies were evaluated for several aspects of bias and sensitivity. An overall confidence classification (high, medium, or low confidence or not informative) was then developed by integrating the judgments for each category of bias and sensitivity for each study or for a specific analysis within a study, as detailed in Appendix A, Table A-28. Experimental animal studies were evaluated and assigned confidence ratings based on expert judgment regarding each study’s experimental details related to predefined criteria within five study feature categories: exposure quality, test subjects, study design, endpoint evaluation, and data considerations and statistical analysis. Explanation of these criteria is provided in Appendix A, Table A-29. Given the volume and diversity of endpoints of individual mechanistic studies, these studies were not evaluated systematically; instead, the focus was on those studies pertaining to specific well-established key events for formaldehyde-induced cancer: genotoxicity and cell proliferation. Despite the heterogeneity of mechanistic studies conducted over the years, EPA carefully considered, where appropriate, exposure assessment, study design, outcome ascertainment, and comparison groups for potential sources of bias and their potential impact. Evaluation of exposure in individual studies was considered especially relevant, and detailed description of the approaches to and criteria for exposure assessment are provided for observational epidemiological and animal studies.

SYNTHESES OF RESULTS AND SYNTHESIS JUDGMENTS

Synthesis of various lines of evidence (human, animal, and mechanistic) on the cancer-related effects of formaldehyde has previously been conducted by various authoritative agencies and organizations. Because numerous other agencies evaluated the same domains and synthesized results and judgments, a summary of the synthesis judgments in the 2022 Draft Assessment and

other published cancer hazard evaluations is provided in Tables 5.2–5.4. Overall, the synthesis judgments in the 2022 Draft Assessment are consistent with those of other authoritative agencies and organizations, while considering more endpoints and additional studies that have become available over time.

Cancers of the Upper Respiratory Tract

For upper respiratory tract cancers, the 2022 Draft Assessment first presents a synthesis for human, animal, and mechanistic evidence separately, and then overall for evidence integration.

For studies of human health effects, evidence is synthesized for cancer types based on the anatomical location of diagnoses commonly reported on death certificates because the histological type of each cancer is reported infrequently on death certificates.

• For nasopharyngeal cancer in humans, the 2022 Draft Assessment evaluates 20 primary epidemiological studies—12 case-control and eight cohort study designs. The text explains confidence level determinations for each individual study (Appendix F, Table 1-32). Two additional possibly impactful studies for this cancer type were identified through systematic evidence mapping (Table F-8), and both provided additional analyses of the studies that had already been evaluated. Overall, the consistency of the observed association was judged on the basis of 17 studies that were deemed informative (with various levels of confidence). The Draft Assessment states that 14 of the 17 studies reported increased risks of nasopharyngeal cancer with at least one metric of formaldehyde exposure, often with both clear statistical significance and exposure-response relationships. Also, EPA notes that results showing increased risks were consistently reported in populations from both high-risk areas (e.g., with endemic Epstein-Barr infection) and low/medium-risk areas, as well as across study populations with different proportions of histological cancer subtypes. The studies that did not report increased risks of nasopharyngeal cancer were deemed to have low confidence, and EPA discusses the basis for the confidence determinations for those studies. The overall synthesis judgment for consistency of the findings was that the majority of studies from different populations, in different locations and exposure settings, and using different study designs reported increased risks of nasopharyngeal cancer associated with formaldehyde exposure. In addition, EPA implies that the strength of the observed association, and the temporal and exposure-response relationships was also evident despite the variable magnitude of the relative risk estimates. Discussion of why potential impacts of selection bias, information bias, confounding bias, and chance could be excluded in the overall evidence synthesis is also provided. In the overall synthesis of epidemiological evidence, EPA concludes that the available human studies provide robust evidence of an association consistent with causation between formaldehyde exposure and increased risk of nasopharyngeal cancer.

TABLE 5-2 Formaldehyde Cancer Conclusions by Site: Evaluations of Epidemiological Data by EPA, the National Research Council (NRC), the International Agency for Research on Cancer (IARC), and the National Toxicology Program (NTP)

NOTE: ^a The NRC (2011) conclusions are based on integration over the three data streams—human, animal, and mechanistic evidence.

TABLE 5-3 Formaldehyde Cancer Conclusions: Evaluations of Animal Data by EPA, the National Research Council (NRC), the International Agency for Research on Cancer (IARC), and the National Toxicology Program (NTP)

NOTE: ^a The 2022 Draft Assessment (EPA, 2022a) uses the categories for the synthesis judgments for animal evidence in the 2022 IRIS Handbook (EPA, 2022b), using the terms robust (strong signal of effect with very little uncertainty), moderate (signal of effect with some uncertainty), slight (signal of effect with large amount of uncertainty), and indeterminate (signal cannot be determined for or against an effect).

TABLE 5-4 Formaldehyde Cancer Conclusions by Site: Evaluations of Mechanistic Data by EPA, the National Research Council (NRC), the International Agency for Research on Cancer (IARC), and the National Toxicology Program (NTP)

• For sinonasal cancer in humans, the 2022 Draft Assessment evaluates 20 primary epidemiological studies—7 case-control, 12 cohort study designs, and 1 pooled analysis. The draft explains confidence level determinations for each individual study (Table 1-32). No additional possibly impactful studies for this cancer type were identified through systematic evidence mapping (Appendix F, Table F-8). Overall, the consistency of the observed association was judged on the basis of 17 studies that were deemed informative (with various levels of confidence). These studies examined different populations, in different locations, under different exposure settings and used different study designs. Because of the extremely rare occurrence of this type of cancer in humans, a large number of informative studies were classified as having low confidence because of the small population size. The overall synthesis judgment for consistency of the findings was that the majority of the studies of different populations, in different locations and exposure settings, and using different study designs reported increased risk of sinonasal cancer associated with formaldehyde exposure that was unlikely to have been confounded by coexposure to wood dust. In addition, EPA implies that the strength of the observed association and temporal and exposure-response relationships was also evident. EPA concluded that the observations of multiple instances of very strong associations in different settings reduce the likelihood that chance, confounding, or other biases can explain the observed associations. In its overall synthesis of the epidemiological evidence, EPA concluded that the available human studies provide robust evidence of an association consistent with causation between formaldehyde exposure and increased risk of sinonasal cancer.
• For other respiratory tract cancers (oropharyngeal/hypopharyngeal and laryngeal) in humans, the 2022 Draft Assessment notes that there were fewer epidemiological studies to evaluate, and that it was difficult to discern the exact anatomical locations of the cancers that were evaluated in some studies. For oropharyngeal/hypopharyngeal cancer, EPA evaluated data from nine reports on six distinct study populations―four reports on three cohort studies and five reports on three case-control studies. For laryngeal cancer, the evaluation included 16 informative studies—12 cohort studies and 4 case-control studies. The 2022 Draft Assessment concludes that the available epidemiological studies provide only slight (for oropharyngeal/hypopharyngeal cancer) or indeterminate (for laryngeal cancer) evidence with which to assess the potential for an association between formaldehyde exposure and an increased risk of these cancers because of the challenges with consistency, strength, and temporal and/or exposure-response observations, as well as the potential impact of selection bias, information bias, confounding bias, and chance.

For studies of respiratory tract cancers in animals, evidence was synthesized for cancer types based on the histopathological classification. The 2022 Draft Assessment presents evidence tables for the experimental animal studies organized by study duration, focusing specifically on chronic exposure (≥1 year) and subchronic exposure (≥3 months) with long-term follow-up (typically assessed after ≥1 year), because of the latency of the expected tumors and their rarity in untreated animals. The challenges faced with evidence integration across studies in animals included species differences in observed effects that were attributed to interspecies differences in airway anatomy; in oral/nasal breathing patterns, including reflex bradypnea; and in other factors (e.g., mucus flow and production, as well as differences in the expression or distribution of enzymes involved in formaldehyde detoxification). Where data were available, the 2022 Draft Assessment also presents tumor incidence plots across the range of concentrations studied.

• For squamous cell carcinomas (SSCs) in animals, the 2022 Draft Assessment evaluates six two-year exposure studies in rats; of these, five showed increases in SCCs that were restricted to the nasal cavity. One two-year study in mice also reported an increased SCC incidence. Nonlinear dose-response relationships were observed in these studies as was strain-to-strain variability among rats. EPA also concluded, based on the synthesis of evidence, that the locations of the induced SCCs were consistent with both the distribution of inhaled formaldehyde and the locations of other formaldehyde-induced nasal pathologies, with SCCs arising from the epithelium lining the airway and not from the underlying glandular epithelium.
• For other malignant and nonmalignant neoplasms and dysplasia, the database is less robust than that for SCCs. Other malignant neoplasms were considered rare but not incidental because they developed only after exposure to the highest formaldehyde concentrations. Other benign tumors of the respiratory tract have been reported following formaldehyde exposure in rats, but not in other species; they have been restricted to the nasal cavity and generally have taken more than 12 months of exposure to develop. Several studies reported increased incidence and severity of dysplasia in long-term formaldehyde inhalation studies in rats and mice (chronic or subchronic exposure, with observation periods of >12 months).
• Overall, the 2022 Draft Assessment concludes that in mice and several strains of rats, but not in hamsters, tumors of the respiratory tract (predominantly SCCs but including other epithelial and nonepithelial tumors) were consistently observed in chronic studies of formaldehyde at concentrations above approximately 6−7 mg/m³. Precancerous lesions and epithelial dysplasia were also observed, both at the anterior regions of the nasal cavity and sometimes at lower concentrations than those associated with increased tumor formation. The Draft Agreement also concludes that the development of these lesions, particularly SCCs, depended on the duration of observation, and, based on an increasing incidence and severity of lesions in animals exposed for longer periods of time, the duration of formaldehyde exposure.

For consideration of mode of action (MOA) for upper respiratory tract cancers, EPA synthesized the evidence to propose an integrated MOA. The strengths of this portion of the 2022 Draft

Assessment (across the various documents) are the comprehensive evaluation of the available evidence across multiple mechanisms, evaluation of the concordance of temporal and dose-response relationships, and an attempt to integrate the evidence using both MOA and adverse outcome network frameworks. The Draft Assessment acknowledges (Appendix A, p. A-771) that no formal, systematic approach was used to identify and evaluate the literature databases of studies examining mechanistic data relevant to interpreting the potential for formaldehyde to cause upper respiratory tract tumors. The reason given is that because key events are well established, a formal, systematic review approach to the mechanistic evidence may be redundant. The 2022 Draft Assessment evaluates evidence for genotoxicity, cellular proliferation, cytotoxicity and pathology in the nasal airways, and several other mechanisms separately and then provides an integration and summary of the MOA evidence. Key conclusions are that (1) strong, consistent evidence from rodent and nonhuman primate models supports the role of both direct (i.e., potentially DNA–protein crosslink or hypermethylated DNA adduct-associated) mutagenicity, as well as indirect genotoxicity, mutagenicity, and regenerative proliferation resulting from respiratory tissue pathology, in rodent upper respiratory tract carcinogenesis; (2) mutagenicity is presumed to be a relevant component of upper respiratory tract carcinogenesis in humans, supported by consistent observations of direct genotoxicity and mutagenicity from human epidemiological studies; and (3) increased nasal epithelial cell proliferation (in rats and nonhuman primates) coincides anatomically with progressive, proliferative lesions in the nasal/buccal epithelium and nasopharynx of chronically exposed humans. Finally, the Draft Assessment notes that mechanistic data provide strong and consistent evidence supporting the contribution of both direct genotoxicity and mutagenicity and cytotoxicity-induced regenerative proliferation as primary mechanistic events. EPA concluded that these mechanisms were highly relevant for informing quantification of nasal cancers in experimental animals following chronic formaldehyde exposure.

The 2022 Draft Assessment presents an integrated summary of evidence for upper respiratory tract cancers in both narrative and tabular formats (Table 1-43). Evidence judgments and hazard determinations are presented separately for nasopharyngeal cancer, sinonasal cancer, oropharyngeal/hypopharyngeal cancer, and laryngeal cancer. For each of these cancer types, human evidence, animal evidence, and other inferences were considered separately and then integrated into the overall hazard determination. Mechanistic evidence was integrated throughout each of these lines of evidence to support arguments about biological plausibility. A specific hazard determination (evidence demonstrates, evidence suggests, or inadequate evidence) was made for each of the upper respiratory tract cancers, and potential susceptible subpopulations were also addressed for these cancer types with the classification at the level of evidence demonstrates.

Lymphohematopoietic Cancers

For lymphohematopoietic cancers, the 2022 Draft Assessment focuses on clinical diagnoses of Hodgkin lymphoma, multiple myeloma, myeloid leukemia, and lymphatic leukemia in exposed humans, as well as relevant tumor findings in animals and mechanistic evidence. The evidence was integrated for the overall causality determinations.

For human health effect studies, evidence was synthesized for cancer types based on specific clinical diagnoses that were available in epidemiological studies, as recommended by the NRC (2011). EPA’s 2010 Draft Assessment made determinations of causality for lymphohematopoietic cancers in several groupings: “all LHP cancers,” “all leukemias,” and “myeloid leukemias.” The 2011 NRC committee criticized this approach because “it combines many diverse cancers that are not closely related in etiology and cells of origin.” The NRC committee recommended that EPA focus on “the most specific diagnoses available in the epidemiologic data, such as acute myeloblastic leukemia, chronic lymphocytic leukemia, and specific lymphomas.” Accordingly, EPA acknowledged in the section “Overview of Lymphohematopoietic Cancer Biology” that lymphohematopoietic cancers are a heterogeneous group of cancers that encompass a wide variety of leukemias and lymphomas, cancers that are often derived from cells of different origin, can demonstrate unique genetic abnormalities, may arise in different tissues, and may have distinct etiologic underpinnings. While acknowledging the challenges of deducing specific cancer diagnoses from epidemiological studies and differences in terminology among different versions of the International Classification of Diseases (ICD), EPA concluded that four specific types of lymphohematopoietic cancer (myeloid leukemia, lymphatic leukemia, multiple myeloma, and Hodgkin lymphoma) warranted the most attention.

• of three publications that reported exposure-response trends describing the effect estimates of all association between formaldehyde exposure and risk of myeloid leukemia (Table 1-59). Studies were divided into those that provided population-level exposure assessment (five studies, three of which were classified as having low and two as having medium confidence), and individual-level exposure assessment (eight studies, four classified as having low, one as having medium, and three as having high confidence). While EPA makes a causal determination for myeloid leukemia overall (acute and chronic), it notes that in studies with separate estimates by subtype, risks were elevated for both acute and chronic myeloid leukemias, with the association for the chronic subtype appearing to be as strong as or stronger than that for acute myeloid leukemia. With respect to the strength of the associations, EPA concludes that overall, studies with higher-quality exposure data, based on individual-level exposure assessment, generally reported higher relative effect estimates. With respect to the temporal relationship of the observed associations, EPA acknowledges that while myeloid leukemia cancer diagnoses were made after the individuals had exposure to formaldehyde, evidence is mixed with respect to the latency period and “time since first exposure.” Evaluation of exposure-response relationships is challenging because of the differences in categorizing exposures across studies and cohorts; however, EPA concludes that three high-confidence studies demonstrated significant exposure-response trends, even though several additional studies showed no significant relationship. Discussion of why the potential impact of selection bias, information bias, confounding bias, and chance could be excluded in the overall evidence synthesis is also provided. In the overall synthesis of epidemiological evidence, EPA concludes that the available epidemiological studies provide robust evidence of an association consistent with causation between formaldehyde exposure and increased risk of myeloid leukemia.
• For lymphatic leukemia in humans, the 2022 Draft Assessment evaluates nine primary epidemiological studies—two case-control and seven cohort study designs. EPA notes that the diagnosis of lymphatic leukemia in the published studies was largely presented in a way that made it difficult to separate the results into acute and chronic lymphocytic leukemia. The 2022 Draft Assessment explains the confidence level determinations for each individual study in Table 1-61. The overall synthesis judgment for consistency of the findings is that informative studies consistently identified null associations—that is, evidence indicative of no association between formaldehyde exposure and the risk of developing or dying from lymphatic leukemia. In addition, EPA concludes that the strength of the null association, lack of temporal concordance, and exposure-response relationships were also evident. EPA concludes that, despite consistent observations of genotoxicity in peripheral blood lymphocytes, exfoliated buccal cells, or nasal mucosal cells in several occupational studies, these data were not sufficient to overturn the judgment on the lack of human evidence for lymphatic leukemia. EPA concludes that the available epidemiological studies provide indeterminate evidence with which to assess the carcinogenic potential of an association between formaldehyde exposure and an increased risk of lymphatic leukemia.
• For multiple myeloma in humans, the 2022 Draft Assessment evaluates 14 primary epidemiological studies—five case-control and nine cohort study designs. The 2022 Draft Assessment explains confidence level determinations for each individual study in Table 1-62. The overall synthesis judgment for the consistency of the findings is that there were generally mixed results, with some studies showing nonsignificant

• increases in risk and others showing nonsignificant decreases in risk. A number of challenges with exposure assessment and other methodological issues are described in detail for all the studies. In addition, EPA concludes that the strength of the associations was inconsistent, and there was limited evidence with which to evaluate temporal relationships or exposure-response relationships, with one study reporting inverse relationships with duration of exposure and time since first exposure. The 2022 Draft Assessment concludes that there were consistent observations of genotoxicity. Overall, EPA concludes that the available epidemiological studies provide slight evidence of an association consistent with causation between formaldehyde exposure and an increased risk of multiple myeloma, primarily with respect to peak exposure.
• For Hodgkin lymphoma in humans, the 2022 Draft Assessment evaluates 15 primary epidemiological studies—one case-control and 14 cohort study designs; only 12 of these studies were deemed informative. The 2022 Draft Assessment explains confidence level determinations for each individual study in Table 1-63. The overall synthesis judgment for consistency of the findings is that the results of the 12 informative studies were not consistent. One possible explanation presented by EPA for their lack of consistency is that the long-term survival rate for Hodgkin lymphoma is far higher than that for other lymphohematopoietic cancers, thus raising a question about the use of cancer mortality as a surrogate for cancer incidence. In addition, EPA concludes that strength of the associations was inconsistent, with effect estimates being highly variable and predominantly less than 1 (unity). Only one study showed a temporal relationship, and EPA cites lack of corroboration of this finding in other studies. Two studies that had the data to evaluate exposure-response relationships showed opposite results. The 2022 Draft Assessment concludes that observations of genotoxicity were consistent. Overall, EPA concludes that the available epidemiological studies provide slight evidence of an association consistent with causation between formaldehyde exposure and an increased risk of Hodgkin lymphoma.

Finding: With respect to cancer hazard identification for lymphohematopoietic cancers in humans exposed to formaldehyde, EPA used appropriate methods to synthesize the current state of the science and presents conclusions regarding the hazard identification analysis that accord with its framework and criteria, and are based on expert judgment with the support of the available scientific evidence.

For studies of lymphohematopoietic cancers in animals, EPA concluded that the database for drawing conclusions about causality is limited because most animal studies of formaldehyde did not evaluate tissues beyond the respiratory tract. One chronic-duration inhalation study that reported detailed information on formaldehyde-induced leukemia or lymphoma in rodents (Battelle Columbus Laboratories, 1982) was deemed indeterminate because its results remain unpublished. Specifically, in this high-confidence chronic study, the incidence of lymphoma in female mice increased from 16 percent in the control group (0 ppm) to 22 percent in the high-dose group (17.6 ppm) (p-value, 0.06); in contrast, the incidence of leukemia in female rats decreased from 9 percent in the control to 6 percent in the high-dose group (p-value, 0.006). This study did not examine

lymphohematopoietic tissues at intermediate doses (2.5 and 6.9 mg/m³). Several additional rat and mouse studies that did examine lymphohematopoietic tissues and did not find statistically significant increases in leukemia or lymphoma were considered to have medium and low confidence because of methodological and/or reporting deficiencies. EPA deemed the overall database to be indeterminate for drawing conclusions regarding the potential for formaldehyde exposure to induce lymphohematopoietic cancers in rodent bioassays.

In considering MOA, EPA concluded that no MOA(s) has been established for how formaldehyde inhalation may result in lymphohematopoietic cancers. Instead, evidence was evaluated and synthesized to present plausible mechanisms for lymphohematopoietic cancers through inhalation exposure. This approach is consistent with the conclusions of previous National Academies committees that evaluated the cancer hazard of formaldehyde (Table 5-4). According to the NRC (2011, p. 5), for example, “data are insufficient to conclude definitively that formaldehyde is causing cytogenetic effects at distant sites … a mechanism that would explain the occurrence of cytogenetic effects in circulating blood cells has not been established. That gap in mechanistic understanding is particularly problematic because the data strongly suggest that formaldehyde is not available systemically in any reactive form. Thus, the committee can only hypothesize that the observed effects result from an unproven mechanism in portal-of-entry tissues.” The 2014 NRC committee stated that experimental “findings provide plausible mechanistic pathways supporting a relationship between formaldehyde exposure and cancer, even though the potential mechanisms of how formaldehyde may cause such systemic effects are not fully understood” (NRC, 2014, p. 17.)

EPA reasons in the 2022 Draft Assessment that a network of mechanistic events or pathways is more suitable than a linear MOA to support the biological plausibility for many cancers, including lymphohematopoietic cancers. Specifically, EPA organized the evidence around the following mechanistic events that support the biological plausibility of formaldehyde exposure–induced lymphohematopoietic carcinogenesis: formaldehyde-induced DNA damage in peripheral blood leukocytes; impacts other than genotoxicity on immune cell populations in peripheral blood in humans and inflammation/immune dysfunction; systemic oxidative stress; and other health effects outside of the respiratory system, including developmental and reproductive toxicity (for which hazard classification was that the evidence indicates that effects in humans are likely). Each of these plausible mechanistic events is discussed in the general style of a narrative review, and the relevance of each mechanistic event to lymphohematopoietic carcinogenesis is considered in greater detail. Alternative hypotheses and gaps in the current knowledge base are specifically acknowledged. Summary conclusions for each hypothesized mechanistic event are then further summarized in Table 1-66, with explicit statements on human relevance and weight-of-evidence conclusions and considerations for biological plausibility. Overall, EPA concludes that the available evidence supports some events that could contribute to plausible mechanistic pathways relating formaldehyde exposure to lymphohematopoietic carcinogenesis; however, the database was insufficient to support the evaluation or development of any specific MOA. EPA further concludes that while the available mechanistic database has limitations, this does not detract from the strength of the association between formaldehyde exposure and myeloid leukemia in epidemiological studies. This conclusion is identical to that drawn by the NRC (2014).

The 2022 Draft Assessment presents an integrated summary of evidence for lymphohematopoietic cancers in both narrative and tabular formats (Table 1-67). Evidence judgments and hazard determinations are presented separately for myeloid leukemia, multiple myeloma, Hodgkin lymphoma, and lymphatic leukemia. For each of these cancer types, human evidence, animal evidence, and other inferences are considered separately and then integrated into the overall hazard determination. Mechanistic evidence is integrated throughout each of these lines of evidence to present arguments about biological plausibility. A specific hazard determination (evidence demonstrates, evidence suggests, or inadequate evidence) is made for each of the included lymphohematopoietic cancers, and potential susceptible subpopulations are also addressed for the cancer types with an evidence demonstrates determination.

DOSE-RESPONSE ANALYSIS OF CANCER EFFECTS OF FORMALDEHYDE

EPA performed dose-response analysis for nasopharyngeal cancer and myeloid leukemia outcomes. For these cancer outcomes, EPA made a causal determination that the evidence demonstrates that inhalation of formaldehyde causally increases risk for these malignancies in humans. The 2022 Draft Assessment does not include a third cancer outcome in this evidence category—sinonasal cancer—because adequate quantitative data for dose-response analysis were lacking.

EPA concluded that the large and diverse body of mechanistic, pharmacokinetic, whole-animal, and human evidence was sufficient to conduct the technically complex dose-response analysis for the nasopharyngeal cancer and myeloid leukemia outcomes. For nasopharyngeal cancer, the 2022 Draft Assessment first analyzes the epidemiological data and estimates a cancer “inhalation unit risk,” which is the extra risk (above background) caused by lifetime exposure to an increase of 1 µg/m³ unit of formaldehyde.³ Multiple alternative estimates are also derived from animal bioassay data, and reflect various assumptions about the established and potential mechanisms by which formaldehyde causes nasal cancer, while also addressing pharmacokinetic considerations. Because of a lack of suitable animal data and uncertainties in mechanistic understanding, the unit risk for myeloid leukemia is based only on epidemiological data.

In its review of the 2010 Draft Formaldehyde Assessment, the 2011 NRC committee made a number of recommendations and comments related to methods, assumptions, and other decision points about the cancer dose-response analysis, including the following (NRC, 2011):

• “The draft IRIS assessment does not provide adequate narratives regarding selection of studies and endpoints for derivation of unit risks. The committee strongly recommends that EPA develop, state, and systematically apply a set of selection criteria for studies and cancer endpoints.” (p. 145)

___________________

³ Can also be expressed as extra risk per ppm.

• “The committee recommends that EPA … Update the dose-response analysis in the IRIS assessment when findings from the update of the NCI cohort on solid cancers become available.” (p. 88)
• “the committee recommends that the CFD [computational fluid dynamic]-based approach also be used to extrapolate to low concentrations, that the results be included in the overall evaluation, and that EPA explain clearly its use of CFD modeling approaches (p. 44)
• “The committee recommends that EPA provide alternative calculations that factor in nonlinearities associated with the cytotoxicity-compensatory cell proliferation mode of action and assess the strengths and weaknesses of each approach.” (p. 59)
• “Overall, the committee finds EPA’s approach to calculating the unit risks [for systemic cancers] reasonable. However, EPA should validate the Poisson dose-response models for NPC, leukemia, and Hodgkin lymphoma mortality with respect to adequacy of model fit, including goodness of fit in the low-dose range, (log) linearity, and absence of interactions of covariates with formaldehyde exposure. EPA is strongly encouraged to conduct alternative dose-response modeling by using Cox regression or alternative nonlinear function forms.” (p. 145)

The present committee considered the recommendations of the 2011 NRC committee, as well as the guidelines presented in EPA’s 2022 IRIS Handbook (EPA, 2022b) and 2005 Guidelines for Carcinogen Risk Assessment (EPA, 2005a), in evaluating whether appropriate methods were used in developing the 2022 Draft Assessment to synthesize the current state of the science and determine whether the dose-response analysis for the effects of formaldehyde by inhalation with respect to cancer outcomes was supported by the scientific evidence and appropriately conducted.

INHALATION UNIT RISK FOR NASOPHARYNGEAL CANCER FROM EPIDEMIOLOGICAL DATA

The 2022 Draft Assessment first presents the unit risk estimates derived from epidemiological data, and then presents the analysis based on animal bioassay data and mechanistic, anatomical, and physiological considerations to support the dosimetry using computational fluid dynamic modeling of dosimetry and biologically based dose-response modeling.

Study Selection

The 2022 Draft Assessment uses the Beane Freeman et al. (2013) study in its derivation of unit risk for nasopharyngeal cancer from epidemiological data, stating that it is the only study with sufficient individual exposure data. This study was a follow-up of a large National Cancer Institute (NCI) retrospective cohort made up of workers from 10 U.S. plants in formaldehyde-using industries. Another advantage of this study noted in the assessment is that it included internal comparisons, which minimized the bias associated with a potential healthy-worker effect. In the analysis of the NCI cohort (Beane Freeman et al., 2013), several different exposure metrics were used—peak exposure, average intensity, cumulative exposure, and duration of exposure. This study was published after the 2010 Draft Assessment was released. Other studies that could have been considered for dose-response analysis are generally described as smaller in population size, not having an appropriate internal comparison group, or not reporting adequate exposure metrics for modeling.

Endpoint Selection

The endpoint selected for dose-response analysis from the Beane Freeman et al. (2013) study was mortality from nasopharyngeal cancer. Although the 2022 Draft Assessment also contains a judgment of a causal link (evidence demonstrates) between formaldehyde exposure by inhalation and sinonasal cancer, the latter outcome was not used to conduct a dose-response analysis because the study identifies a total of only five deaths from nose and nasal sinus cancer.

Dose Metric Selection

Individual exposure assessments were available for each worker in the Beane Freeman et al. (2013) study, enabling the development of various exposure estimates of interest, including peak

short-term exposures, frequency of peak exposures, duration spent in jobs with formaldehyde exposure, and cumulative exposure. EPA selected cumulative exposure as the metric for its characterization of dose-response for deriving unit risk from the Beane Freeman et al. (2013) study. Cumulative exposure was estimated from time in a specific job multiplied by the time-weighted average concentration for that job category.

The 2022 Draft Assessment characterizes frequency of peak exposure as highly uncertain because it was estimated based on assumptions made by the exposure assessors for the study. Although EPA stated that the peak exposure metric produced the strongest exposure-response relationships, it did not use this metric because it is not based on exposure concentration measurements. In addition, EPA expressed uncertainty about how to use peak exposure-based estimates in predicting risks for lifetime exposure to continuous low exposures. Because the average exposure metric did not account for duration and duration does not account for level of exposure, these metrics also were not used.

Model Selection and Fit

EPA relied on dose-response modeling performed by Beane Freeman et al. (2013). Poisson regression models for incidence with a log-linear function were applied. Multiple lag periods were modeled to account for the latency period for solid tumors, with a 15-year lag period ultimately being adopted. Stratification was done by calendar year, age, sex, race, and pay category (salary vs. hourly wage). The authors note that the Poisson regression and Cox proportional hazards models yield essentially similar results when age is adjusted for in the cohort. The low-exposure group served as the reference population to address unmeasured confounding associated with nonexposed workers and potential differences in socioeconomic status. The use of low rather than zero exposure as the reference group made little difference in the observed trend for the cumulative exposure dose metric, with the level of statistical significance being quite similar regardless of whether the zero-exposure group was included (p = 0.07) or excluded (p = 0.06). The effect was more pronounced in the dose-response trend associated with the peak exposure metrics (p = 0.005 under exposed person-years and p = 0.10 when the zero-exposure group was used as the reference group).

The lead author of the study provided EPA with the regression coefficients reflecting the relative risk per cumulative exposure unit (i.e., per ppm × year) (Beane Freeman et al., 2013, pp.

2–49). The coefficients were nearly identical for the regression with the zero-exposure group included vs. excluded (0.431 vs. 0.439 per ppm × year).

Point of Departure and Inhalation Unit Risk for Nasopharyngeal Cancer Mortality

A point of departure (POD) is an exposure level that is drawn from the dose-response curve fit to observed data. In cancer risk assessment, it is used as a basis for extrapolation to lower concentrations than those in the study used in its derivation. In this case, the POD is used to derive the slope of the concentration versus cancer response curve. That slope is called the “inhalation unit risk,” and is based on the assumption that there is not a threshold concentration below which the cancer risk is zero. The unit risk estimate enables the calculation of “extra risk” from low incremental increases in concentration.

EPA applied the regression coefficients from Beane Freeman et al. (2013) in a life-table program that accounts for competing causes of death to predict extra risk of mortality from nasopharyngeal cancer at different concentrations of formaldehyde. These calculations also used a 15-year lag period. The upper 95 percent confidence bounds on the extra-risk estimates at different concentrations were derived from the reported standard errors on the regression coefficients, and below 0.01 ppm linearly decreased with decreasing dose (Table 2-17, p. 2-50).

EPA used the same approach to calculate the POD, in this case the lower-bound estimate of concentration associated with an extra risk of 0.05 percent. Because background mortality rates for nasopharyngeal tumors are very low, the 2022 Draft Assessment explains that using a higher extra risk would be inappropriate. For example, using an extra risk of 1 percent, typically used for epidemiological data, would result in a relative risk of 53, an upward extrapolation from the observed data, and using 0.1 percent would yield a relative risk of 6.2, on the high end of the observed range in the Beane Freeman et al. (2013) study. At 0.05 percent risk, the relative risk estimate is 3.6, which is in the observed range of the study. The POD associated with the extra-risk level of 0.05 percent is 0.11 ppm.

EPA calculated the unit risk value by dividing the benchmark extra-risk level of 0.05 percent by the POD—that is, the lower-bound estimate of the concentration associated with that extra risk. The unit risk EPA calculated for nasopharyngeal cancer mortality was 3.7 × 10^-3 per mg/m³ (4.5 × 10^-3 per ppm).

Inhalation Unit Risk for Nasopharyngeal Incidence

Because nasopharyngeal cancer has a favorable survival rate (the 2022 Draft Assessment cites 51 percent at five years in the 1990s in the United States), EPA also calculated an incidence-based unit risk. In making this calculation, EPA used the same approach used to derive the mortality extra-risk estimates. However, instead of using the mortality statistics for this cancer, EPA used incidence values from NCI’s Surveillance, Epidemiology, and End Results (SEER) program for the period 2000–2010. The same approach was used to calculate the POD of 0.055 ppm at an extra risk of 0.05 percent, which in turn was used to derive the unit risk estimate of 7.4×10^-3 per mg/m³ (9.1 × 10^-3 per ppm) for formaldehyde-induced nasopharyngeal cancer incidence.

The 2022 Draft Assessment (pp. 2–53) also presents a risk calculation to consider the plausibility of the unit risk for nasopharyngeal incidence. Estimates of formaldehyde-related nasopharyngeal cancer cases in the United States were derived and compared with actual case numbers of this rare cancer. Under the assumption that the U.S. population is exposed to 5 ppb formaldehyde for 75 years, the annual number of incident cases of nasopharyngeal cancer in the United States was estimated to be 180. Using a higher formaldehyde concentration estimate of 20 ppb resulted in an estimate of 730 cases annually. These numbers were compared with the 2,300 actual incident cases per year in the United States—a much greater number.

Inhalation Unit Risk for Nasal Cancer from Animal Bioassay Data

Study Selection

EPA also derived inhalation unit risk estimates from two long-term animal studies of formaldehyde in F344 rats (Kerns et al., 1983; Monticello et al., 1996). EPA selected these two studies in part because both reported exposure-dependent incidence of nasal squamous cell carcinoma (SCC). EPA further decided to combine the two studies for dose-response analysis because the combined data had a wider exposure range and larger numbers of animals, providing more robust dose-response information and greater statistical power.

Cancer Endpoint Selection

EPA selected nasal SCC as the cancer endpoint and used time to tumor as the outcome measure for the dose-response assessment.

Dose-Response Modeling

EPA used a computational fluid dynamic (CFD) model for formaldehyde airflow in the nasal passage of rats to characterize regional exposure to formaldehyde (Kimbell et al. 2001a, 2001b; Schroeter et al., 2014). This dosimetry of local exposure of formaldehyde inhalation facilitates biologically based dose-response (BBDR) modeling of SCC incidence. It was also to derive dose surrogates for use in statistical modeling of the dose-response. Because formaldehyde was deemed to be a direct-acting mutagen, EPA also used a physiologically based pharmacokinetic (PBPK) model (Subramaniam et al., 2007) to model DPC as a function of formaldehyde flux. In addition, EPA conducted dose-response modeling of cell proliferation as a precursor outcome of SCC using data from Monticello et al. (1991, 1996).

EPA presents results from several dose-response models for SCC incidence.

• EPA fit the multistage Weibull time-to-tumor model (MSW) to individual animal data from the CIIT animal cancer bioassays. The dose metric was formaldehyde flux derived using the CFD model.
• EPA fit a Weibull model to grouped incidences after adjusting for censoring using Kaplan-Meier survival (Schlosser et al., 2003). EPA presents results for two different dose metrics used by Schlosser et al.: formaldehyde flux and DPC derived from the PBPK modeling. EPA presents the fits using these two dose metrics for the best-fitting Weibull model, which had a nonzero intercept on the dose axis, the so-called “Weibull with threshold” result.
• EPA calibrated a BBDR model with time-to-tumor data. The BBDR model was based on the two-stage clonal expansion (TSCE) model (Moolgavkar et al., 1988), but incorporates DPC as the molecular dose that depended on formaldehyde flux predicted by the

• CFD model. DPC tissue concentration was in turn calculated using a PBPK model developed by Conolly et al. (2000). EPA presents results from two different modeling assumptions regarding cell kinetics as a function of exposure.

Benchmark Dose Modeling

Using the dose-response models, EPA produced estimates of benchmark concentration (BMC) and benchmark concentration lower bound (BMCL); a summary is presented in Table 2-22. EPA used two versions of the BBDR model—one based on a conservative prediction of formaldehyde flux of inhalation exposure in conjunction with historical control cancer rates of inhalation exposure studies, and the other based on the prediction of a monotonic increase of formaldehyde flux along with the historical control rates used in the first version.

Uncertainty and Variability

EPA discusses extensively uncertainties associated with dose-response analysis and BMC estimation. The discussion includes the use of precursor outcomes (cell proliferation and hyperplasia) for BMC estimation, low-dose extrapolation using BBDR models, model selection, and statistical uncertainties in a BMC estimate within a model. The focus of the discussion is on the use of BBDR for extrapolation, with a list of seven major sources of uncertainty (Table 2-24). EPA concludes that the sources that substantially impacted uncertainty were rat cell labeling data, cell division rates, the assumption that SCC was a fatal tumor, the use of historical controls, and the division and death rates of initiated cells. EPA further concludes that for human extrapolation, the BBDR models suffered two major limitations (Conolly et al., 2004), and did not provide robust measures of human nasal SCC risk at any exposure concentration. Therefore, EPA did not use the BMCL derived from the BBDR models for human risk extrapolation.

The focus of EPA’s discussion is on the use of the BBDR model for extrapolation. EPA’s conclusion on the sources of uncertainty with major impact is based largely on the sensitivities of the output of the BBDR models to model inputs related to the associated parameters (e.g., Table 2-25). Despite such a discussion of the uncertainties and variabilities from multiple sources, the unit risk estimates from the BMCs and BMCLs are remarkably similar (Table 2-26) across the models for the animal data.

Selection of a Unit Risk Estimate for Nasal Cancer

EPA derived inhalation unit risk estimates using the human nasopharyngeal cancer data and animal SCC data. The inhalation unit risk was 7.4 × 10^-3 per mg/m³ based on the nasopharyngeal cancer data and 8.9 × 10^-3 to 1.8 × 10^-2 per mg/m³ based on the animal SCC data, and these results

are comparable. Based on EPA’s Guidelines for Carcinogen Risk Assessment (EPA, 2005a), which prefers human data for derivation of risk estimates, EPA designated the inhalation unit risk derived from human nasopharyngeal cancer data as the preferred inhalation unit risk.

Uncertainties and Confidence in the Preferred Unit Risk Estimate for Nasal Cancers

Inhalation Unit Risk for Myeloid Leukemia from Epidemiological Data

The methods and rationale for derivation of the unit risk estimate for myeloid leukemia are similar to or the same as those for nasopharyngeal tumors discussed earlier. Differences that are particular to myeloid leukemia are emphasized in this section.

Study Selection

As was done for nasopharyngeal cancers, the most recently published primary study of leukemia on the large NCI cohort (Beane Freeman et al., 2009) was chosen as the basis for developing the unit risk for myeloid leukemia. This was the only study with appropriate data for dose-response modeling. The exposure assessment in Hauptmann et al. (2009) was deemed deficient because worker histories were obtained from the next of kin, resulting in lower confidence in such data, and 30 percent of subjects were missing detailed work histories.

Inhalation Unit Risk Value for Myeloid Leukemia

As with the nasopharyngeal cancers, the 2022 Draft Assessment notes that the Poisson regression and Cox proportional hazards models yield essentially the same results when age is well characterized and adjusted for. Also, the regression coefficient estimates from the Poisson regression were provided to EPA by the lead author on the study publication (Table 2-30, footnote “Source”).

Utilizing cumulative exposure as the exposure metric in modeling relative risk for myeloid leukemia resulted in a nonsignificant trend (p = 0.44) compared with the use of peak exposure (p = 0.07). The 2022 Draft Assessment lays out the likelihood of significant underreporting of myeloid leukemia and cites studies supporting the estimate that one-third to one-half of nonspecified leukemias on death certificates could be myeloid leukemias by hospital diagnosis (Percy et al., 1981, 1990). Other/unspecified leukemia amounted to about 30 percent of all leukemia cases in the NCI cohort. To address the issue of likely underreporting of myeloid leukemia, EPA grouped the subcategories of myeloid leukemia and “other/unspecified leukemia.” This grouping resulted in a more significant association between these leukemias and cumulative exposure to formaldehyde (p = 0.1) compared with myeloid leukemia alone (p = 0.44). EPA also considered all leukemias in deriving inhalation unit risk estimates and discusses the limitations of this approach in the Draft Assessment.

EPA used the same approach it used for nasopharyngeal cancer (lifetable, application of regression coefficient, selection of POD, and extra-risk benchmark) to derive inhalation unit risk estimates for myeloid leukemia mortality. Because of uncertainty regarding MOA, EPA used a default linear extrapolation following the 2005 Guidelines for Carcinogen Risk Assessment (EPA, 2005a). As with the nasopharyngeal cancer data, incidence-based inhalation unit risk were also derived using the SEER registry data, this time from 2006–2010.

The 2022 Draft Assessment includes three sets of unit risk values for leukemias—myeloid leukemia, all leukemia, and myeloid plus other/unspecified, with the final set identified as the preferred estimates. The uncertainties in the analysis are described, including the nonsignificant increase in risk of myeloid leukemia with increased cumulative exposure and the stronger finding for the other groupings of leukemias. EPA concluded that there is low confidence in the inhalation unit risk estimate for myeloid leukemia and did not include it in the overall inhalation unit risk estimate for formaldehyde.

Uncertainty in the Myeloid Leukemia Inhalation Unit Risk

The 2022 Draft Assessment provides a detailed discussion of the uncertainty associated with the myeloid leukemia inhalation unit risks. The discussion focuses on the statistical significance of the association between myeloid leukemia mortality risk and various exposure metrics due to underreporting of myeloid leukemia. Other important sources of uncertainty include a lack of mechanistic support for myeloid leukemia, uncertainty about the true but unknown shape of the dose-response relationship and its data manifestation, and exposure misclassification. For example, the trend p-values for peak exposure are 0.07 and 0.50 for myeloid leukemia and other/unspecified leukemia, respectively. In contrast, the trend p-values for cumulative exposure are 0.44 and 0.14 for myeloid leukemia and other/unspecified leukemia, respectively.

Human-Based Unit Risk Estimates for Potential Increased Early-Life Susceptibility

Decision to Apply Adjustment for Increased Early-Life Susceptibility

Age-dependent adjustment factors (ADAFs) were used to adjust for potential increased susceptibility resulting from early-life exposure, in accordance with the Supplemental Guidance for

Assessing Susceptibility from Early-life Exposure to Carcinogens (EPA, 2005b). The rationale for adjusting for increased early-life susceptibility was that the mutagenic mode of action was established for nasopharyngeal cancers, and the relative risk from formaldehyde exposure is independent of age within the adult age window. For earlier years, relative risk was assumed to be a factor of 10 larger from birth to <2 years and a factor of 3 larger for ages >2 to 16 years. Beyond the 16th birthday, the adult relative risk was assumed.

While there is uncertainty in the degree to which nonmutagenic processes may also contribute to the carcinogenic activity of formaldehyde inhalation at the point-of-entry tissues, there is sufficient evidence to support the assumption that a mutagenic MOA is involved in the carcinogenesis of formaldehyde in the upper aerodigestive tract in humans. Furthermore, there are no formaldehyde-specific data to inform adjustments for early-life exposure, which further supports use of the default procedure as discussed in the Supplemental Guidance for Assessing Susceptibility from Early-life Exposure to Carcinogens (EPA, 2005b). Finally, this approach is concordant with the Guidelines for Carcinogen Risk Assessment (EPA, 2005a) which states (p. 3-22):

Approach to Adjustment

An estimate of “adult-only unit risk” was derived by applying the lifetable approach discussed earlier for nasopharyngeal cancers to ages greater than 16 years. This result was scaled by multiplying by 70/54 years to create “adult-based unit risk estimates for nasopharyngeal cancer for use in ADAF calculations and risk estimate calculations involving less-than-lifetime exposure scenarios” (Table 2-38). Calculations were then made using this “adult-based” inhalation unit risk and ADAFs to derive a total lifetime inhalation unit risk estimate assuming exposure from birth to 70 years of age. The resulting estimate is termed a “lifetime unit risk estimate” (p. 2-99). The “adult based” unit risk estimate is identified as the “preferred inhalation unit risk estimate,” and the “lifetime unit risk estimate” is termed the “ADAF-adjusted unit risk estimate” (p. 2-102).

Confidence in the Unit Risk Estimate

EPA assigned the “preferred” unit risk estimate for nasopharyngeal cancer incidence an overall confidence level of medium. The 2022 Draft Assessment acknowledges the substantial uncertainty in the dose extrapolation, especially in light of endogenous formation of the chemical and the possible effect on uptake. On the other hand, EPA acknowledges the strength of the large NCI study and the quality of its exposure assessment. A major uncertainty is the inability to include myeloid leukemia in the unit risk estimate because of the quality of the data available for dose-response analysis. The Draft Assessment also describes various sources of uncertainty having to do with the exposure assessment in the NCI study, the selection of the model, and the exposure metric used for the dose-response modeling.

Appropriate Exposure Circumstance

This chapter follows the approach described in the “General Assessment Organization” section of the 2022 Draft Assessment of adding the modifier “under appropriate exposure circumstances” to each of the credible hazards identified. It is noted in this orienting section of the document that “the ‘appropriate exposure circumstances’ alluded to during hazard identification in

___________________

⁴ Integrated Risk Information System (IRIS) Glossary. Available at: https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&cad=rja&uact=8&ved=2ahUKEwiH9ZC73p7-AhVKjIkEHdUJC4wQFnoECBEQAQ&url=https%3A%2F%2Fwww.epa.gov%2Firis%2Firis-glossary&usg=AOvVaw29QeD8x7j-7RHqbA05CH2B.

Section 1 are more fully evaluated and defined through dose-response analysis in Section 2 (including, depending on the evidence available, the derivation of toxicity values).”

REFERENCES