Literature Identification

In the Main Assessment and Appendices, EPA documents the steps it followed to formulate a population, exposure, comparator, and outcome (PECO) statement; develop inclusion and exclusion criteria; and conduct its literature search. EPA updated its literature searches annually through September 2016, after which systematic evidence mapping was used to search studies through 2021. For its literature search, EPA used four databases, and curated reference lists in published reviews and other national or international health assessments of formaldehyde (see Appendix A, Section A-231), although the flow diagram shows that only two databases were used (Figures A-22 and A-263). Search terms included formaldehyde, paraformaldehyde, formalin, irritation, irritant, and irritants (Table A-31).

Recommendation 4.1 (Tier 2): To maximize transparency and facilitate replication, EPA should clarify the Medical Subject Headings (MeSH) terms used, list and justify any MeSH terms that were excluded (e.g., eye, ear, nose, or skin), provide the list of national and international reviews and assessments used to identify additional references, and provide more specific links to the Health & Environmental Research Online (HERO) database where the screening decisions are documented (see Appendix A, Section A-232, line 4).

Finding: EPA excluded outdoor exposure studies (Table A-32) without providing adequate justification. This exclusion was broadly applied to multiple health outcomes (e.g., sensory irritation [see Table A-32], pulmonary function [see Table A-42], and nervous system effects [see Table A-83]). In the Main Assessment (p. xxv) EPA states:

Publications were typically excluded if they contained no information about formaldehyde exposure or were descriptions of analytic methods using formaldehyde. Ambient levels of formaldehyde in outdoor air are significantly lower than those measured in the indoor air of workplaces or residences, and the exposure range was narrow in many epidemiological studies of ambient exposure (<0.005 mg/m³) limiting their sensitivity to find any associations with health outcomes even if they existed.

Excluding outdoor studies from the 2022 Draft Assessment may skew the evidence pool in the direction of higher exposure studies relative to the levels commonly experienced by the general population. The committee also found unconvincing EPA’s argument that studies with lower exposure levels may have a limited ability to detect associations between formaldehyde exposure and health effects.

The committee notes further that the assertion that outdoor levels of formaldehyde are universally below 0.005 mg/m³ may not correctly represent true population exposures for all geographies and time periods. For example, the Hanrahan et al. (1984) study, which is a pivotal sensory irritation study for this assessment, reported mean outdoor levels of 0.04 ppm (0.05 mg/m³) with high variation (standard deviation [SD] = 0.03 ppm). The committee noted that additional studies reported ambient levels at or above 0.005 mg/m³, as summarized in Table 4-2.

TABLE 4-2

Formaldehyde Levels in Outdoor Air Reported in Selected Studies.

Recommendation 4.2 (Tier 2): EPA should include the body of evidence from outdoor exposure studies at the preliminary stage to derive a more holistic and objective assessment of the scientific literature.

Finding: In the Main Assessment (Figure 2-3), the box representing typical formaldehyde levels depicts a sharp delineation between lower outdoor and higher indoor levels. In real-world exposure scenarios, that is unlikely to be the case; rather, a continuum of concentrations with overlapping outdoor and indoor levels is a more likely scenario (Table 4-2). Figure 2-3 could show representative ranges of outdoor and indoor formaldehyde concentrations levels more accurately (see Appendix E).

Study Evaluation

Outcome-specific criteria used to evaluate human studies of formaldehyde-induced sensory irritation included participant selection; information bias; potential for confounding; statistical analysis; and other considerations, such as exposure levels, contrast, duration of follow-up, and sensitivity of outcome assessment (Appendix A, Section A-233). Figure A-22 states that a total of 38 observational studies and 20 randomized controlled trials on humans were included for sensory irritation. Selection bias receives some attention in the evaluation, given the cross-sectional nature of some of the key observational studies.

Finding: The outcome-specific criteria EPA used to evaluate the human studies were generally appropriate but the application of the specified criteria across studies appears inconsistent, and it is not clear if the same set of quality criteria was applied uniformly across studies. The potential for selection bias is addressed for each of the studies. Determining the potential for selection bias in a study is challenging as the specific selection fractions that lead to potential bias are generally unknown. Cross-sectional studies are generally at risk for selection bias. The potential for selection bias may also be signaled by a low response rate and nonrepresentativeness of a study population on exposure and outcome frequency. The committee could not find a consistent approach in how EPA evaluated the potential for selection bias, for example, across the range of observed response rates.

Finding: The 2022 Draft Assessment described specific aspects of exposure assessment that EPA considered when evaluating individual studies, such as having a measurement protocol, duration of exposure measurement period, number of samples obtained, consideration of temperature and humidity, and percentage of measurements below the limit of detection(Appendix A, Section A-236). The text does not clarify whether systematic or random contributors to inaccuracy were assessed in evaluating information bias in individual studies, nor does it describe how the assessment took account of multiple sources of information bias. Finding: The 2022 Draft Assessment lacks clarity as to how the availability and characteristics of outcome assessment questionnaires were used to rate the confidence level of a study. For example, Mueller et al. (2013) provide a reference for their questionnaire, and EPA assessed this study as having a high confidence level. However, Green et al. (1989) do not provide a reference for their outcome assessment questionnaire, yet EPA categorized it as well as having a high confidence level (Main Assessment: Table 1-1, pp. 1–18). EPA categorized several studies (Hanrahan et al., 1984; Liu et al., 1991; Sexton et al., 1986) as having medium confidence because of concerns regarding the questionnaires used to assess symptoms without adequate explanation of their deficiencies.

Recommendation 4.3 (Tier 2): EPA should explicitly state what constitutes an adequate assessment of outcomes when a questionnaire is not cited, and explicitly provide the criteria used to determine the adequacy of a questionnaire. Information on these aspects of outcome assessment would facilitate replication of the EPA approach. It would be preferable for EPA to use age categories generally instead of ambiguous descriptors.

Finding: Some studies are potentially affected by multiple limitations. The approach used to classify the level of confidence for such studies is unclear. If there are three sources of potential bias with no apparent directionality (i.e., either over- or underestimation) for each bias, how will the quality of that study be determined? The committee considered as an example the evaluation of the study by Main and Hogan (1983). In Table 1-2, two problems with this study are identified, with one box fully colored. The classification scheme in Figure II would suggest this study is of medium confidence, but it is classified as having low confidence, presumably because the fully colored box for “other” signifies a greater “degree of limitation.” Additional explanations related to Figure II would make it easier to understand individual study confidence determinations, such as when there is only one potential source of bias, and that bias is likely to overestimate the exposure effect.

Recommendation 4.4 (Tier 2): To increase transparency, EPA should document how it assessed the potential for different types of biases, the directionality of resulting biases, and the number of biases, and state how each combination should be interpreted in terms of high, medium, low, or not informative study confidence.

Finding: The considerations listed for classification of study confidence and evaluation of each study by at least two independent experts is consistent with EPA’s state-of-practice methods.

Recommendation 4.5 (Tier 3): In the population selection criteria, the potential for selection bias could be assessed by considering the proportion of the eligible population invited to participate in the study and the proportion of the eligible population that was ultimately included in the analysis. EPA should state the criteria used to assess selection bias in the text, tables, and figures.

Evidence Synthesis and Judgment

The 2022 Draft Assessment concludes that there is robust evidence for sensory irritation from controlled human exposure studies, as well as epidemiological studies. It also concludes that robust evidence exists from animal studies, and there are established mechanisms. The 2022 Draft Assessment indicates further that there is robust evidence for a specific mode of action (MOA) underlying the association between inhalation of formaldehyde and sensory irritation. Formaldehyde exposure directly or indirectly stimulates trigeminal nerve endings in the respiratory epithelium, which has been highlighted as the dominant pathway for causing this outcome. The supporting evidence is based largely on animal studies, but EPA’s interpretation is that the suggested MOA identified in animals is also relevant to humans.

Finding: EPA provides a reasonable basis, one that is consistent with its state-of-practice methods, for labeling as robust the evidence relating formaldehyde exposure and sensory irritation.

Overall Conclusions About the Hazard Descriptor

EPA’s judgment was that, taken together, the evidence demonstrates that inhalation of formaldehyde causes sensory irritation. This judgment was based on four high- and medium-confidence studies of symptom prevalence in humans in residential settings, numerous high- and medium-confidence acute controlled human exposure studies, and numerous high- and medium-confidence occupational studies.

Finding: EPA’s overall hazard conclusions are supported by the scientific evidence and are consistent with EPA’s state-of-practice methods.

Dose-Response Evaluation

In section 2.1.1 of the Main Assessment (p. 2-2), EPA states that only high- and medium-confidence studies were chosen for POD analysis. This section also reports that emphasis was placed on the characteristics of the study population, the accuracy of formaldehyde exposure, the severity of the observed effects, and the exposure levels. Human epidemiological studies that evaluated groups most representative of the general population were preferred, as were studies that reported complete results and are unlikely to have alternative explanations.

Recommendation 4.6 (Tier 2): EPA should clarify and clearly state the criteria used to select the studies for dose-response analysis of noncancer endpoints.

EPA excluded three of the six studies selected for POD derivation: one (Liu et al., 1991) that reported partial results, another (Mueller et al., 2013) that did not observe an exposure–response relationship, and a third (Lang et al., 2008) for which the adverse response level was difficult to define. For the remaining three studies, EPA derived a POD. EPA stated that of these three studies, it had less confidence in those of Kulle et al. (1987) and Andersen and Molhave (1983), which were controlled exposure studies for which the PODs were an order of magnitude higher than that for the study by Hanrahan et al. (1984), which was the final selection.

Finding: EPA’s favoring of well-conducted and -reported epidemiologic studies over controlled human exposure studies was justifiable for sensory irritation.

Literature Identification

Table A-41 in Appendix A summarizes the search terms used for PubMed and Web of Science. Table A-42 (p. 314) summarizes the PECO inclusion and exclusion criteria. Only human studies with indoor inhalation exposure and formaldehyde measurements were included. Outcomes were restricted to FVC, FEV₁, FEF, and PEFR.

A total of 53 studies were identified and evaluated for consideration (Appendix A, p. 313, line 15). This total apparently represents studies rather than individual publications since some studies are described in multiple publications (e.g., Broder et al., 1988a,b,c). Among these 53 studies were 42 observational epidemiology and 11 controlled human exposure studies (Appendix A, Figure A-23). Publication dates for these studies ranged between 1975 and 2015.

Finding: The inclusion and exclusion criteria listed in Appendix A, Table A-42 are consistent with EPA’s state-of-practice methods.

Study Evaluation

Methodological issues considered in the evaluation of studies are provided in the Main Assessment (p. 95), as well as in the Assessment Overview (Section 3.2.2). They include pulmonary function measures (with a table of definitions [Table 1-5]); a preference for studies that follow American Thoracic Society (ATS) guidelines or provide detailed protocol and reference equation information; and a preference for pulmonary function measures normalized by race or ethnic origin, gender, age, and height. EPA mentions smoking as a potential confounder and the need to consider a referent group when evaluating change across a work shift or laboratory session. Appendix A presents a table (Table A-43) of criteria used to categorize study confidence for epidemiological studies as high, medium, low, and not informative. This table considers two aspects: (1) exposure, and (2) study design and analysis. Neither the Main Assessment nor the Assessment Overview references Table A-43. Table A-44 provides EPA’s evaluation of all epidemiological studies of pulmonary function, organized by study type and then alphabetically by author. Some studies have multiple confidence ratings— for example, for preshift versus cross-shift outcomes, longitudinal versus cross-lab change, or comparison with community referents versus change during embalming. Table A-45 covers the controlled human exposure studies. This table is organized by confidence level and then year of publication, with the confidence categories being medium (randomized, with results fully reported) (five publications), low (incomplete reporting of results or blinding not described, with multiple exposure levels) (four publications), and low (no randomization, with blinding not discussed) (two publications).

Evidence Synthesis and Judgments

Section 1.2 of the Main Assessment, “Synthesis of Evidence for the Effects on the Respiratory System,” addresses synthesis between and within evidence streams for all noncancer outcomes. Section 1.2.2, “Pulmonary Function,” begins with an overview providing brief comments on aspects of the literature identification and inclusion, followed by brief summaries of the evidence synthesis, mechanistic support, and hazard evaluation. Detailed synthesis information begins with the unnumbered subsection on PDF p. 96 “Pulmonary Function Studies in Humans.” The synthesis discussion is blended with the study evaluation tables (Tables 1-6 to 1-9) also presented in this section, and the subsection is organized by exposure duration (acute, intermediate, and long-term) and study design features: for acute, controlled human exposure and observational within the work shift or anatomy lab section; for intermediate, anatomy labs; and for long-term, occupational and then residential. Appendix A provides several figures (A-24 to A-26) that show results from studies describing change in pulmonary function measures during a work shift or anatomy lab session, with one figure for each outcome measure (FEF, FEV₁, FVC, etc.). This topic ends with Table A-46, summarizing each of those studies. The following narrative provides specific comments relevant to the synthesis discussion for each study grouping presented in the Main Assessment.

Acute controlled human exposure studies. This subsection notes that while formaldehyde exposure has not been shown to induce pulmonary function deficits in nonexercising healthy volunteers, small but statistically significant deficits have been observed in studies with two or more 15-minute exercise regimens, although not in studies with shorter exercise segments. The overview at the beginning of Section 1.2.2 (Main Assessment, p. 93) indicates that the controlled human exposure studies “consistently did not observe changes.”

Acute epidemiological studies: Changes in pulmonary function across a work shift or anatomy course lab session. This subsection considers studies of work shift changes among multiple occupations (plywood workers, chemical industries, funeral workers), as well as students in anatomy lab sessions. The workers were assumed to have had prior formaldehyde exposure, while the students were not. The text states that many of the studies “observed pulmonary function declines over the course of the workday or lab.” Most studies did not consider change in an unexposed referent group; studies that did include a referent group showed a change in pulmonary function on average in that group, although studies varied with respect to the direction observed, and this additional potentially insightful detail is not provided.

Intermediate-duration exposures (<1 year) among anatomy/pathology students. The discussion of the three panel studies published by two sets of authors highlights that there were different results for spirometry measures (FVC, FEV₁, FEV₁/FVC, and FEF_25-75) in one study versus change in PEFR measures in two other studies. Interpretation of those studies was challenged by intermittent exposures, student absences, and decreasing formaldehyde concentrations over the quarter. These studies are not mentioned further in any of the summaries (i.e., the beginning of Section 1.2.2 of the Assessment Overview or the integrated summary of evidence in the Main Assessment).

Long-term formaldehyde exposure in occupational settings. This subsection addresses two types of study design: cross-sectional (or prevalence) studies and longitudinal studies. Note that the reference to cross-sectional studies (Main Assessment, p. 102) points to studies under the “prevalence studies” heading in Table 1-7; it is left to the reader to make this connection. EPA concludes that overall, these studies show evidence of decrements in pulmonary function associated with formaldehyde exposure, particularly given that many studies could be biased toward no association.

For the cross-sectional studies, challenges are highlighted in the text: selection bias (healthy worker effect and survivor [lead time] bias), a community-based reference group in one study, higher prevalence of other exposures affecting pulmonary function in the referent group). Nonetheless, EPA notes that most studies observed associations of formaldehyde with deficits in pulmonary function before the work shift at the beginning of the work week.

The discussion of the longitudinal results from three studies with lead authors Nunn, Alexandersson, and Lofstedt (p. 1–47, PDF p. 106), highlights that the four- to six-year duration of these studies, their small sample sizes, and the potential for exposure-related loss to follow-up leading to selection bias are challenges, but notes that nonetheless, some pulmonary function declines were reported. Three studies are mentioned briefly, and then an additional two reference studies are covered (Lee and Fry, 2010; Redlich et al., 2014). They report a formaldehyde-associated age-related decline in FEV₁ among nonsmokers that was 50 percent greater compared with the expected rate of age-related decline. The remainder of the discussion concerns duration of work in an exposed job and its association with pulmonary function. EPA concluded that results “seem to support a conclusion that occupational exposure may result in declines of FEV₁ and FEF over time” (PDF p. 53 of the Assessment Overview), but did not deem them to be consistent across studies.

Residential exposures among adults. This section addresses four studies (covered in seven publications), noting the challenge of comparing them given their different approaches to pulmonary function assessment and reporting. Overall, EPA concluded that “adults in general did not experience declines in pulmonary function at average formaldehyde levels less than 0.05 mg/m³” (PDF p. 117). Much of the discussion in this subsection focuses on Krzyzanowski et al. (1990) because it was rated as having high confidence.

Residential and school exposures among children. This discussion also focuses on Krzyzanowski et al. (1990), whose results for children aged <15 years show a decrease in PEFR associated with increased formaldehyde exposure. Results from this paper are reproduced in Figure 1-6, which shows the modeled decrease in PEFR per unit of formaldehyde exposure; the figure would benefit from a more in-depth discussion given the later use of this study for dose-response analysis. The general discussion highlights the low exposures in these studies and the small exposure contrast in one, although it was difficult to determine the exposure concentration reported by Wallner et al. (2012).

Finding: Within study groupings, EPA equates a lack of statistical significance with a lack of decline in pulmonary function (e.g., reference to Lofstedt PDF p. 98, line 19), and makes statements that are poorly supported by the evidence (e.g., reference to the size of differences and lack of precision on PDF p. 101, line 7).

Finding: Evidence is presented inconsistently across the six sections describing results from different exposure durations and study populations. Finding: Overall, while EPA’s distillation and synthesis of the pulmonary function evidence is challenging to follow, the judgments presented in the Main Assessment are supported by the scientific evidence and consistent with EPA’s state-of-practice methods.

Finding: Given the wording in Section 3.2.3, it is difficult to understand the basis for some of the evidence synthesis judgments as presented in the Assessment Overview. For example, EPA does not provide the specific evidence and decision frameworks used to support the following statement: “overall the longitudinal analyses appear to be inconsistent, but while hindered by a lack of sensitivity, seem to support a conclusion that occupational exposure may result in declines of FEV1 and FEF over time” (p. 53, line 17). Further, some aspects of the study synthesis discussion are not clearly defined. For example, what does it mean that studies observed “inconsistent responses” (Assessment Overview p. 52, line 10)? Does this mean that reported results are in different directions, or that some results are statistically significant, while others are not?

Recommendation 4.7 (Tier 2): EPA should clarify the basis for its synthesis judgments and provide additional information about the studies on which they are based, such as the formaldehyde levels observed, as well as the exposure ranges or other measure of variability. The study summary tables (Tables 1-6 to 1-9) should be updated to provide an organized distillation of the points made in the evidence synthesis text.

Recommendation 4.8 (Tier 2): If the Assessment Overview is retained (see Recommendations 2.1 and 2.2 in Chapter 2), EPA should harmonize its presentation of evidence synthesis with the presentation in the Main Assessment. In particular, the evidence synthesis section of the Assessment Overview could be updated to build upon the first three paragraphs of the “Integrated Summary of Evidence for Pulmonary Function” section in the Main Assessment (PDF p. 134).

EPA’s synthesis also considers the MOA evidence for decrements in pulmonary function. As summarized in Figure 1-7, the most plausible mechanisms are indirect activation of sensory nerve endings in the lower respiratory tract and/or increases in airway eosinophils. There are also possible changes in the upper respiratory tract that may contribute to this outcome. Table 1-10 summarizes the studies that provide the most informative mechanistic evidence regarding decrements in pulmonary function following formaldehyde exposure.

Finding: The MOA considerations and mechanistic evidence are clearly presented and appropriately documented.

Overall Conclusions About the Hazard Descriptor

EPA concluded that, based on moderate human evidence, long-term inhalation of formaldehyde is likely causal for decreases in pulmonary function (i.e., EPA applied the evidence indicates rating). EPA deemed Krzyzanowski et al. (1990) to have the strongest design and methods, providing evidence supported by a more limited study in schools conducted by Wallner et al. (2012). The narrative and table also mention “several studies of workers with long-term exposure to >0.2 mg/m³” without giving references. EPA also concluded that the evidence is inadequate to determine the causal effect of formaldehyde exposure on acute and intermediate-term time scales. This judgment does not give much weight to the evidence that pulmonary function decrements occurred in controlled human exposure settings among participants who exercised at least 30 minutes (Green et al., 1987, 1989), or that some individuals exhibited clinically significant deficits (Green et al., 1987).

Finding: EPA’s evidence integration judgments are supported by the scientific evidence and are consistent with EPA’s state-of-practice methods.

Dose-Response Evaluation

EPA’s confidence in the human study used to derive the POD (Krzyzanowski et al., 1990) was high. This cross-sectional study of residential exposure found a linear relationship between higher formaldehyde exposure and decreased PEFR among children exposed to average concentrations of 0.032 mg/m³ (26 ppb). EPA applied benchmark dose modeling to calculate the concentration at which a 10 percent decrement in pulmonary function would be expected; EPA considered a 10 percent decrement to be the benchmark response (BMR). A benchmark concentration (BMC)₁₀ (0.033 mg/m³) and benchmark concentration lower bound (BMCL)₁₀ (0.021 mg/m³) were subsequently determined from the regression coefficient from a random effects model of PEFR among children reported by the study authors. A single uncertainty factor to account for variability among humans (UF_H) of 3 was applied to the BMCL₁₀ to derive a cRfC of 0.007 mg/m³.

Finding: EPA’s derivation of BMCL was based on nonasthmatic children and nonmorning exposures. This practice was inconsistent with EPA’s state of the practice of using more vulnerable subpopulations for risk estimation.

Recommendation 4.9 (Tier 2): EPA should provide additional justification for why the most vulnerable subpopulations were not used for risk estimation, and should consider using the data from children with asthma that are provided in Krzyzanowski et al. (1990).

Literature Identification

PubMed search terms related to respiratory tract pathology in humans included hyperplasia, metaplasia, nasal mucosa, occupational diseases, respiratory tract diseases, rhinitis, and mucociliary. Some broad terms (e.g., the MeSH term pathology) were not used in the search for human literature.

A total of 1009 citations were screened (title and abstract) for assessment of respiratory tract pathology in humans, and 12 studies were ultimately included in the review (Appendix A, p. A-390). A total of 1678 citations were screened (title and abstract) for the assessment of respiratory tract pathology in animals, and 41 toxicology studies were ultimately included in the review (Appendix A, p. A-393).

Finding: For human studies, the comparator of the PECO is defined as evaluated outcome associations with formaldehyde exposure; it is unclear what specific comparisons were made. The comparator in the PECO statement used for screening the animal studies is undefined.

EPA’s review focused on histopathological endpoints and signs of pathology in nasal tissues.

Finding: It is unclear why EPA did not consider more distal effects in the respiratory tract since studies performed in nonhuman primates have reported changes in the respiratory epithelium of the trachea and major bronchi (Monticello et al., 1989). The exclusion criteria for human respiratory tract pathology included exclusion of studies reporting rhinitis. This exclusion is surprising since rhinitis was a search term used to find human studies. Moreover, it is unclear whether this exclusion criterion resulted in exclusion of studies with histologic evidence of inflammation.

Recommendation 4.10 (Tier 2): EPA should provide an explicit description of the comparator used in screening human and animal studies, and resolve discrepancies between search terms and inclusion and exclusion criteria.

Study Evaluation

Outcome-specific criteria used to evaluate human studies of formaldehyde-induced respiratory pathology included assessment of the exposure, participant selection and comparability, possibility of confounding, analysis and completeness of results, and study size (Appendix A, Table A-57). EPA evaluated whether studies describing histologic results provided an explanation of how tissues were evaluated and scored. EPA downgraded cross-sectional studies among occupational cohorts since workers may have been less sensitive to the irritant properties of formaldehyde then the general population. EPA considered gender and smoking were considered by EPA to be potential confounders for pathological endpoints.

Outcome-specific criteria used to assess the animal studies included sample size, inadequate reporting of lesion incidence and/or severity, combining of multiple lesions, inadequate sampling of the respiratory tract, and short (<1 year) exposure duration or follow-up. EPA also evaluated the source of the formaldehyde (e.g., commercial grade, paraformaldehyde, or formalin). Coexposure to methanol (found in some formalin products) was not considered to be a major confounding factor for identifying effects of inhaled formaldehyde on respiratory pathology. According to EPA, “a sample size of less than 10 was considered a significant limitation”; however, this criterion was applied to the Holmstrom et al. (1989) study even though this study had 16 animals/group.

Finding: EPA’s outcome-specific criteria are generally appropriate and consistent with its state-of-practice methods. It is unclear whether the downgrade for cross-sectional studies among occupational cohorts is relevant for histopathologic outcomes. Evidence supporting consideration of gender as a confounder is not provided. For animal studies, sample size limitations were applied inconsistently. In general, EPA was also inconsistent in providing a rationale for each study quality domain (Appendix A, Table A-59), reducing transparency.

Recommendation 4.11 (Tier 2): EPA should provide a consistent rationale for each study quality domain used in the assessment.

Evidence Synthesis and Judgments

EPA identified no high-confidence and four medium-confidence human occupational studies. Histological changes in the respiratory tract seen in the latter four studies were associated with formaldehyde exposures ranging from 0.1 to 2.5 mg/m³ (Table 1-25). EPA downgraded occupational studies for this outcome. EPA’s rationale for this downgrade was that survivor bias impacted studies, specifically, “current workers who likely were less sensitive ‘survivors’ of the long-term respiratory irritant effects of formaldehyde, which would cause survival bias and an attenuation of comparisons between exposed and comparison groups” (pp. 1–153). Although this downgrade was applied, EPA did not critically evaluate whether a healthy worker survivor effect was present in these studies.

EPA’s evaluation of animal studies focused on the incidence of hyperplasia and metaplasia after formaldehyde inhalation. Only studies judged by EPA to be of high and medium confidence are included in the synthesis and evidence tables in the Main Assessment. Long-term studies were also deemed more relevant for the assessment. EPA concluded that “a clear relationship between formaldehyde exposure duration and the development of squamous metaplasia and, to a lesser extent, hyperplasia” (pp. 1–161) could be drawn from the experimental animal studies. EPA characterizes formaldehyde-induced squamous metaplasia in the rat nasal cavity as minimally adverse, with no clear discussion of this determination.

EPA identified a single study conducted in rhesus monkeys as having medium confidence. This study reported hyperplasia and metaplasia in the larynx, trachea, and carina after a 6-week exposure to 7.4 mg/m³ formaldehyde (Monticello et al., 1989). EPA concluded that these data may suggest that the “monkey nose is less efficient than the rodent nose at scrubbing formaldehyde from inhaled air” (pp. 1–160). Although this conclusion is plausible, a more likely explanation relates to differences in breathing patterns between rodents and nonhuman primates (obligatory nasal versus oronasal).

EPA reviewed mechanistic data extensively and found robust data for several endpoints, including the binding of formaldehyde to macromolecules, alterations in mucociliary function, epithelial damage or dysfunction in the upper respiratory tract, increased cell proliferation in the upper respiratory tract, and trigeminal nerve stimulation. These data were used to support EPA’s conclusion that epithelial cell injury results in squamous metaplasia in the upper respiratory tract.

Finding: The judgments presented in the Main Assessment are supported by the scientific evidence and consistent with EPA’s state-of-practice methods.

Overall Conclusions About the Hazard Descriptor

EPA concluded that there was robust evidence that inhaled formaldehyde exposure can induce histopathologic lesions in the nasal cavity and other portions of the upper respiratory tract. Lesions were dependent on both the formaldehyde concentration and, to a lesser extent, duration of exposure. This conclusion was based on numerous high- and medium-confidence studies of chronic and subchronic exposure duration of multiple experimental animal species. EPA also concluded that the data provide moderate evidence that inhaled formaldehyde induces histopathological lesions in the human upper respiratory tract. This determination was based on four medium-confidence human epidemiological studies (Ballarin et al., 1992; Boysen et al., 1990; Edling et al.,1988; Holmstrom et al., 1989). These studies showed that participants exposed to average formaldehyde levels between 0.05 and 0.6 mg/m³ had higher average histopathology scores than those of their respective comparison group. Overall, the strength of the evidence for hyperplasia and squamous metaplasia includes robust evidence from animal studies and moderate human evidence from observational epidemiological studies, and strong support for a plausible MOA based primarily on mechanistic evidence from experimental animals. EPA’s overall conclusion is that the evidence demonstrates that inhalation of formaldehyde causes respiratory tract pathology in humans given the appropriate exposure circumstances.

Finding: The evidence integration judgments made by EPA are supported by the scientific evidence and consistent with EPA’s state-of-practice methods.

Dose-Response Evaluation

EPA’s confidence in the two studies used to derive PODs was high. Four human studies EPA judged as having medium confidence provided additional support. The PODs derived by EPA were based on lesions seen at Level 1 in the rat nasal cavity. EPA used lesion incidence data (Woutersen et al., 1989; Kerns et al., 1983) to model the dose-response relationship. EPA found that the 24-month data for Level 1 (Table 1-26) could not be modeled because of the steep dose-response relationship seen in the Kerns et al. (1983) study, so it modeled the 18-month incidence data and obtained a BMCL₁₀ of 0.448 mg/m³ (Table 2-5).

EPA used a computational fluid dynamic (CFD) model (Kimbell et al., 2001) to estimate formaldehyde flux at the Level 1 cross section of the F344 rat nose. EPA found that the average flux in the Level 1 region corresponding to the BMCL₁₀ of 0.448 mg/m³ determined for the Kerns et al. (1983) study was estimated to be 685 pmol/mm²-hr. A human CFD model (Kimbell et al., 2001) was then used to estimate the formaldehyde exposure concentration (0.484 mg/m³) that would result in a similar formaldehyde flux in the human nose (Table 2-5; Appendix B, Section B.1.3). This value was subsequently adjusted for continuous exposure (6 hours/24 hours) × (5 days/7 days) to provide a human POD_ADJ of 0.086 mg/m³.

A human POD_ADJ of 0.094 mg/m³ was obtained for squamous metaplasia at nasal Level 1 using lesion incidence data from Woutersen et al. (1989). Since a CFD model for Wistar rats was unavailable, EPA based the POD for the Woutersen et al. (1989) study on parts per million equivalence (pp. 2–19).

EPA’s confidence in the POD calculation based on Woutersen et al. (1989) was medium, while confidence based on Kerns et al. (1983) was low. EPA stated that it assigned lower confidence to the POD derived from Kerns et al. (1983) because the calculation involved an extrapolation well below the tested formaldehyde concentrations, and the BMC was based on the 18-month exposure rather than the 24-month exposure, where responses were greater in magnitude. EPA did not explain why its confidence in the POD based on the Woutersen et al. (1989) study was medium.

An uncertainty factor (UF)A interspecies uncertainty of 3 was used to account for animal-to-human variation; a UF_H of 10 was used to account for human variation; and a UF_S subchronic uncertainty factor of 3 was applied to account for extrapolation to chronic exposure in the Kerns et al. (1983) study. Incidence data for squamous metaplasia at Level 1 from the Kerns et al. (1983) study are reported in Table 1-26. EPA’s rationale for including a UF_S of 3 rests on its inability to model the 24-month data from the Kerns et al. (1983) study, prompting use of the 18-month data. EPA acknowledges that “a lower POD would have been expected if the 24-month data could have been modeled” (p. 74), and EPA states further that “while exposure duration is important to the development of this lesion, such effects appear to be more dependent on exposure concentration” (p. 75). In addition, an 18-month study in rats is a chronic-duration study that would not typically necessitate the use of a subchronic UF.

Two cRFCs were calculated for respiratory tract pathology. The first was estimated using data from Woutersen et al. (1989), had a composite UF of 30 (UFA = 3, UF_H = 10), and was 0.003 mg/m³. The second cRFC was estimated using data from Kerns et al. (1983), had a composite UF of 100 (as discussed above), and was 0.0009 mg/m³. EPA chose the organ-specific RfC for respiratory tract pathology of 0.003 mg/m³ based on the Woutersen et al. (1989) study. EPA considered the completeness of the database for respiratory tract pathology to be high.

Finding: EPA’s dose-response evaluation was well documented and transparent. Steps used are consistent with EPA’s state-of-the-practice.

Literature Identification

EPA’s literature search strategy is described in detail in the 2022 Draft Assessment, and was informed by consultation with the outside experts. Table A-48 in Appendix A describes the PECO criteria. The experts advised EPA on study inclusion/exclusion criteria, diagnosis instruments used for immune outcomes, confidence assessments for specific studies, ages of participants, and disease mechanisms. The use of validated questionnaire instruments (e.g., International Study of Arthritis and Allergies in Children [ISAAC]; Asher et al., 1995) and diagnostic tests (e.g., skin prick) to specify diagnostic quality as criteria to enhance confidence determinations compared with studies without such validated instruments was appropriate. Studies on asthma involving very young (<5 yrs) and elderly (>75 yrs) subjects were excluded or downgraded since outside experts stated that respiratory conditions other than asthma may be the basis of symptoms in these age groups. Thirty-six studies (27 observational and 9 randomized controlled trials) were chosen for toxicological review pertaining to immune conditions.

Finding: The considerations and methods adopted by EPA were appropriate given the often poor specificity of diagnostic classifications in the published literature on these conditions. Study choice decision metrics are adequately described.

Finding: The terms “very young” and “very old” (as exclusion criteria) are not defined, and several studies of school-age children were used to inform the strength-of-evidence determination. While age cutoffs of <5 and >75 years are cited in the Assessment Overview and Appendices as exclusion criteria, the Main Assessment does not describe age cutoffs, and it is unclear how these cutoffs were made actionable in study choice as some studies of infants and children <5 feature in asthma evaluations in particular. “Infants” (again, a specified age needs to be given each time this word is used) were deemed to have nonspecific symptoms (wheeze and infection) compared with adults and were not considered in the formulation of PODs or synthesis judgments. Likewise, the “elderly” (assumed to mean >75, but again, age needs to be specified each time the word is used) were not considered because of the potential interference of alternative-mechanism diseases such as chronic obstructive pulmonary disease (COPD). In general, EPA divided evaluations into “child” and “adult” studies, but age cutoffs for these categories need to be stated and applied consistently.

Finding: EPA’s decision to exclude food allergy in response to airborne formaldehyde while including eczema is not explained. The rationale for excluding formaldehyde-specific immunoglobulin E (IgE) as an endpoint (because of its rarity in evaluation) is explained appropriately.

Finding: The inclusion/exclusion criteria for animal studies are described in confusing terms. While the search terms emphasize biological mechanisms as relevant (biomarkers such as immunoglobulin G (IgG), IgE, cytokines, hypersensitive reactions), EPA also states that studies “describing the development of immunological or allergy animal models” were not included (Appendix A, p. A-337). The 2022 Draft Assessment documents that animal models are generally helpful only in immediate pathological response to formaldehyde rather than the development of disease states that recapitulate human diseases, but a clear reason for excluding animal models of “the development of immunological or allergy” outcomes is not provided.

Recommendation 4.13 (Tier 2): EPA should explicitly state its rationale for age-based exclusions and define the terms “very young” and “very old,” better justify the rationale for excluding allergic contact dermatitis and food allergy as outcomes of interest, and provide the rationale for excluding animal models of “the development of immunological or allergy” outcomes (unless such studies do not exist or are inadequate).

Study Evaluation

EPA chose to adhere to the recommendations of the outside experts in applying evidence status to those studies that (1) had more robust measures of disease (using appropriate biomarkers and validated questionnaires), (2) included higher exposure levels, or (3) included a prospective study design. The one prospective study (in which formaldehyde exposures preceded disease [Smedje and Norback, 2001]) that was based on low exposure values (many were below the analytical chemical limit of detection) was elevated to medium confidence given its longitudinal, prospective nature. Children were deemed to be “more sensitive” at lower exposures given observed effects, and were carried through the evaluation to yield distinct POD and cRfC values separate from those for adults.

Finding: The outcome-specific criteria used to evaluate human studies were generally appropriate and in line with expert recommendations. However, EPA’s approach appears to contradict the expert panel’s advice that children’s allergic symptoms are more liable to misdiagnosis than those of adults, particularly for asthma in infants and young children.

Recommendation 4.14 (Tier 2): EPA should include a specific statement on the age at which asthma diagnosis is considered valid to justify the age exclusions for young children, as well as the category of “the elderly.”

Evidence Synthesis and Judgments

EPA concluded based on a moderate level of human evidence (and slight in animals) that inhalation of formaldehyde causes an increased risk of prevalent allergic conditions and asthmatic symptoms and decreased control of asthmatic symptoms, with evidence from occupational studies in the range of exposure values >0.1 mg/m³ and from schools and homes at 0.03–<0.1 mg/m³. EPA made these judgments after carrying out the synthesis of substantial evidence on allergic conditions, asthma, and lower respiratory tract infections (in young children), with a large number of medium-confidence studies being used to justify them. These studies included 8 studies on allergies in children and/or adults (one high- and three medium-confidence) and 15 studies on asthma in children or adults (2 high- and 13 medium-confidence), which together provide a preponderance of evidence that formaldehyde influences these phenotypes in humans. Two studies on asthma control (in known asthmatic subjects; pp. 1–112 of the Main Assessment) provided high and medium confidence, respectively, with Venn et al. (2003) showing strong trends. The section on “Controlled Acute Exposure” describes studies of short-term acute exposures to humans.

Finding: While EPA accorded some studies of acute exposures high confidence, those studies were not deemed generalizable to other populations (pp. 1–114) because of brief exposures not relevant to the chronic-exposure scenario. A short and useful section (pp. 1–122) considers effect modification by tobacco smoke, family history of atopy, and known atopy status.

Regarding animal studies, the synthesis was indeterminate for allergy and asthma outcomes based on the lack of appropriate models that recapitulate the symptoms and physiology of human allergy and asthma conditions. Most of the animal data is based on the ovalbumin (OVA) allergen model, which does not recapitulate the entirety of human pathophysiology. The animal model data did aid in advancing understanding of MOA.

Finding: EPA’s use of animal data and the resulting synthesis judgments are appropriate, described thoroughly, and consistent with EPA’s state-of-practice methods. The relative risk size of 1.2 for exposures of around 0.04–0.06 mg/m³ is small, but these conditions are common, making this an important outcome. A specific study (Matsunaga et al., 2008) suggests higher risks, including a two-fold higher risk for eczema, and a three-fold higher risk for allergy-like symptoms from a childhood classroom study that is helpful for determining a POD.

In the section “Evidence on MOA for Immune-related Conditions,” EPA concludes that a definitive MOA could not be identified, but components of the MOA are clear. Formaldehyde is responsible for airway inflammatory changes and remodeling that can contribute to respiratory immune-related conditions. The mechanisms for allergic sensitization are less clear. Reliable human data on changes in production of antibodies to formaldehyde and its protein-adducting metabolites is lacking. Figure 1-12 describes the cascades of immune alterations leading to effector-level changes that result in the pathologic observed formaldehyde hazards. Many of the pathway components have uncertain interrelationships, and the evidence is at moderate to slight levels. These immune mechanisms are complex and largely beyond the scope of the 2022 Draft Assessment, and are described in light of several high- to medium-confidence studies in relation to formaldehyde MOA. These animal studies provide strong evidence for some aspects of formaldehyde’s relevant activities—for example, bronchoconstriction and eosinophil activation via inflammatory mediators such as tachykinins, antibodies, Th2-related cytokines, and white blood cell changes. Summaries of these changes are provided in Tables 1-22 and 1-23, with extensive discussion in Appendix A, Section A.5.6.

Finding: EPA provides an appropriately nuanced discussion of the complex MOA for immune-related conditions and identifies many gaps in knowledge, consistent with its state-of-practice methods.

Overall Conclusions About the Hazard Descriptor

Overall, based primarily on moderate human evidence as well as slight animal evidence from mechanistic studies supporting biological plausibility (including molecular and cellular inflammatory changes and evidence of hypersensitivity), EPA concluded that the evidence indicates that inhalation of formaldehyde likely causes increased risk of prevalent allergic conditions and prevalent asthma symptoms, as well as decreased control of asthma symptoms, given appropriate exposure circumstances.

Finding: EPA’s hazard determination judgments are appropriate given the scientific evidence described and are consistent with EPA’s state-of-practice methods.

Dose-Response Evaluation

EPA deemed six studies on allergic conditions, six studies on asthma, and two studies on control of asthma in asthmatic persons eligible for POD derivation. Two studies with the highest confidence from each of these three categories were ultimately selected for POD identification. No observed adverse effect level (NOAEL) and lowest observed adverse effect level (LOAEL) values were chosen from the highest-quality study in children (as a more sensitive population [Annesi-Maesano et al., 2012]) and the highest-quality study in adults (Matsunaga et al., 2008).

The study in children (Annesi-Maesano et al., 2012) used to derive an allergy POD also provided the NOAEL for asthma along with a second study in children, and the asthma control PODs both pertain to asthma in children. EPA’s final judgments on confidence in these PODs were high for allergy in children and medium for asthma control. No judgments or statements about PODs were derived from studies in adults, who, as noted, are generally less sensitive than children. EPA provides reasoning for not providing a POD for adults given the lack of quantification in the studies evaluated, wide confidence intervals, and dichotomous analyses with variable exposure levels. EPA provides no POD evaluation for contact exposures to formaldehyde.

Uncertainty factors were applied to the PODs to derive cRfCs. Because NOAELs were used for most of the PODs, a LOAEL-to-NOAEL UF (UF_L) of 1 was applied. A UF_H of 3 was generally applied, since most studies assessed exposures and outcomes in potentially sensitive populations such as children or pregnant women. All other uncertainty factors were applied at 1 (apart from a UF_S of 3 that was applied for the one study in which exposure was measured during pregnancy [Matsunaga et al., 2008]).

Finding: EPA’s conclusions and its choice of UFs are consistent with its state-of-practice methods, given the confidence levels and quality of the chosen studies. It is appropriate that asthma control PODs pertain to asthma in children, as no studies of asthma control in adults were of high enough quality.

Literature Identification

The steps EPA followed in conducting its literature search and PECO assessment for inclusion and exclusion of studies are described in the Main Assessment and the Appendices. The initial search was conducted in October 2012, with yearly updates through September 2016. A systematic evidence map identified literature from 2017 to 2021. Inclusion and exclusion criteria for human and animal studies are summarized in Appendix A, Tables A-89 and A-90, respectively. The text indicates that 20 human and 35 animal studies were identified for inclusion.

Finding: EPA’s process for literature identification, PECO assessment, and study inclusion/exclusion was generally transparent and in line with EPA’s state-of-practice methods for the female and male reproductive and developmental outcomes. However, search terms related to birth defects and teratology, such as congenital anomalies, were not included in the search.

Study Evaluation

EPA’s evaluation of human studies of female reproductive or developmental toxicity resulted in two medium-confidence occupational studies of spontaneous abortion, two low-confidence studies of congenital malformations, two medium-confidence studies of decreased birthweight and head circumference, and five low-confidence studies of fecundability and spontaneous abortion. Low-confidence animal studies of female reproductive or developmental toxicity had mixed findings for several outcomes. For male reproductive toxicity, the evaluation included one medium-confidence human occupational study of sperm motility and other outcomes, and one low-confidence human study of sperm count and morphology. Multiple high-or medium-confidence animal studies using mice or rats contributed to an assessment of histopathological lesions of the testes or epididymes, sperm count, testosterone levels, and organ weight change.

Finding: EPA’s outcome-specific criteria for evaluating human studies are consistent with EPA’s state-of-practice methods. A summary of key assessment factors specific to human reproductive and developmental outcomes is not provided.

Finding: For human studies, EPA applied the appropriate general criteria for study quality (e.g., epidemiologic biases), consistent with its state-of-practice methods, as were the conclusions for individual studies (e.g., one medium-confidence study for effects on time to pregnancy and two medium-confidence studies for effects on spontaneous abortion).

Finding: For animal studies, EPA applied the appropriate criteria for study quality (e.g., test substance, dose(s), test animals, evaluation endpoints), consistent with its state-of-practice methods, as were the conclusions for individual studies (e.g., two medium-confidence studies and three high-confidence studies for male reproductive effects). Potential updates for two high-confidence studies (Ozen et al., 2002, 2005) and one medium-confidence study (Sapmaz et al., 2018) are detailed in Appendix E. However, the potential reassignment of these three studies would not affect synthesis judgments and the next assessment steps since all medium- and high-confidence studies were considered in the next steps.

Evidence Synthesis and Judgments

The discussion of reproductive and developmental toxicity in the 2022 Draft Assessment is based on an evaluation of all female reproductive and developmental outcomes combined into a single group. With this treatment as a single group, the mix of human evidence, including medium-confidence studies (with uncertainty due to random error or bias), would place the evidence within the moderate level based on the framework for strength-of-evidence judgments. The justification for combining outcomes was that it would be difficult to distinguish underlying pathogenic events that could yield a delayed recognized pregnancy or fetal loss. Although this rationale may apply generally for delayed time to pregnancy and spontaneous abortion, the broader category of female reproductive and developmental toxicity also encompasses human studies of other “later” developmental outcomes, such as congenital malformations and birthweight.

Finding: Within EPA’s framework for synthesizing results from human studies and making judgments, the conclusion of moderate evidence for female reproductive or developmental toxicity is justified. However, combining all reproductive and developmental outcomes in a single group is an oversimplification. EPA could consider separating its evidence synthesis section into early and late events.

Assessment of animal studies revealed indeterminate evidence for developmental toxicity and separately, indeterminate evidence for female reproductive toxicity. All evaluated studies had low confidence with methodological limitations, the majority of which were due to a lack of information about test substance or use of formalin, which can contain methanol, a known developmental and reproductive toxicant. For developmental effects, EPA found no direct evidence of biological plausibility; however, oxidative stress and/or hormone disruption are noted as possible indirect linkages. For female reproductive effects, EPA notes that the biological plausibility of neuroendocrine-mediated mechanisms involving the hypothalamic–pituitary–gonadal–axis is consistent with the alterations of reproductive hormones identified in the low-confidence rodent formaldehyde studies.

Finding: The conclusion of indeterminate evidence for developmental toxicity and female reproductive toxicity in animals is consistent with EPA’s state-of-practice methods.

For male reproductive toxicity, EPA judged the evidence from human studies to be slight. There was one medium-confidence study, but the sparseness of the available evidence for multiple reproductive and developmental outcomes and the associated uncertainty provided a reasonable basis for the strength-of-evidence judgment as slight.

Finding: The conclusion of slight evidence for male reproductive toxicity in humans is consistent with EPA’s state-of-practice methods.

For animal studies of male reproductive toxicity, EPA judged the evidence to be robust based on six medium- or high-confidence studies conducted by three research teams using five cohorts of rats or mice. The text on p. 82 of the Assessment Overview discusses six medium- or high-confidence studies, but Table A-36 in Appendix A lists only five studies because one row combines two studies. Nonetheless, it is reasonable to conclude that these six studies were well conducted, although all used high formaldehyde concentrations (>5 mg/m3). The studies found adverse testes and epididymis histopathology, decreased sperm count, altered sperm motility and morphology, and decreased serum testosterone. In addition, several low-confidence studies produced consistent results on male reproductive toxicity, although they were also conducted with very high formaldehyde levels (most above >12 mg/m³). Evidence on the MOA for formaldehyde and male reproductive toxicity is lacking, but indirect effects of oxidative stress and heat shock protein induction were noted in testes or epididymes of exposed rats in the medium- and high-confidence studies.

Finding: Information provided in the text and tables for animal studies of male reproductive toxicity is inconsistent (see Appendix E).

Overall Conclusions About the Hazard Descriptor

The judgment that the evidence indicates that inhalation of formaldehyde likely causes increased risk of developmental or female reproductive toxicity in humans (given the appropriate exposure circumstances) was based on moderate human evidence and indeterminate animal evidence for developmental or female reproductive toxicity.

Finding: The conclusion that the evidence indicates that inhalation of formaldehyde likely causes increased risk of developmental or female reproductive toxicity in humans is consistent with EPA’s state-of-practice methods.

The judgment that the evidence indicates that inhalation of formaldehyde likely causes increased risk of reproductive toxicity in men (given the appropriate exposure circumstances) was based on slight human evidence and robust animal evidence for male reproductive toxicity.

Finding: The conclusion that the evidence indicates that inhalation of formaldehyde likely causes increased risk of reproductive toxicity in men is consistent with EPA’s state-of-practice methods.

Dose-Response Evaluation

For female reproductive and developmental toxicity, dose-response estimation was based on a single medium-confidence study with a time-to-pregnancy endpoint (Taskinen et al., 1999). The 8-hour time-weighted average (TWA) for the intermediate (middle) exposure group was selected as a NOAEL. A UF_H of 10 was applied to the developmental toxicity POD.

Finding: The rationale for study selection, the POD determination, uncertainty factors, and cRfC derivation, including confidence levels, were consistent with EPA’s state-of-practice methods.

For male reproductive toxicity, dose-response estimation was based on two high-confidence studies in rats exposed to paraformaldehyde for 13 weeks that assessed relative testes weight and serum testosterone endpoints. For decreased testes weight (Ozen et al., 2002), a LOAEL of 12.3 mg/m³ was adjusted for continuous exposure based on the experimental paradigm to yield a POD_ADJ of 2.93 mg/m³. A final uncertainty factor of 3000 was applied to the male reproductive toxicity testes weight POD. For decreased serum testosterone and decreased mean seminiferous tubule diameter likely associated with decreased serum testosterone (Ozen et al., 2005), a BMCL_1SD of 0.208 mg/m³ was calculated, resulting in a POD_ADJ of 0.05 mg/m³. A final uncertainty factor of 300 was applied to the male reproductive toxicity decreased testosterone POD.

The final uncertainty factors were derived for the two male reproductive endpoints using the following assumptions. First, a UF_A of 3 was applied to both endpoints to account for residual toxicodynamic uncertainties in interspecies extrapolation. Second, a UF_S of 10 was applied to both endpoints to approximate the potential effect of chronic exposure because the studies were conducted over a subchronic duration. Third, a UF_L of 10 was applied to the relative testes weight endpoint, which was based on a LOAEL. Fourth, a UF_H of 10 was applied to both endpoints to account for the limited variability in susceptibility factors encompassed by these typical studies of inbred laboratory animal populations.

Finding: The rationale for exclusion of studies from the dose-response assessment is consistent with EPA’s state-of-practice methods and includes analysis of pooled tissues, short exposure duration, single exposure level, and nonpreferred endpoint assessment.

Literature Identification

The steps EPA followed to conduct its literature search and PECO assessment for inclusion and exclusion of studies are documented in the Main Assessment and the Appendices. A total of 4338 articles were screened for inclusion in the assessment of these outcomes, with 147 being considered for hazard identification (40 human, 60 animal, and 47 in vitro and noninhalation studies) (Appendix A, Figure A-25, p. A-591).

Finding: The process for literature identification and inclusion/exclusion criteria for the review of nervous system effects was transparent.

Finding: EPA elected to exclude headache as a human health outcome for evaluation because of the subjectivity of outcome reporting (Appendix A, Table A-83). Considering that other endpoints included in this assessment, such as sensory irritation, depend on subjective self-reporting, excluding headache based on this criterion is inconsistent.

Recommendation 4.15 (Tier 3): EPA should include studies with headache as an outcome to maintain consistency with other health effect categories. Alternatively, a stronger rationale should be provided for exclusion of headache other than its perceived subjectivity. Headache could be combined with other self-reported neurotoxicity outcomes.

Study Evaluation

Human studies of nervous system outcome-specific criteria were evaluated for strengths and limitations based on principles of epidemiologic study quality, including methods for exposure assessment; windows of exposure; sample size; and potential for selection bias, information bias, and confounding (Tables 1-44 and 1-45; Appendix A, Tables A-84 and A-85). EPA rated some human studies examining neurobehavioral outcomes such as memory impairment and deficits in concentration (Bach et al., 1990; Kilburn et al., 1985, 1987) as having overall low confidence. EPA also examined whether there were associations between formaldehyde and amyotrophic lateral sclerosis (ALS) or mortality from neurological disease. These studies estimated formaldehyde exposure based on estimated job exposures. For these included studies, occupational formaldehyde exposure was based on self-report or job exposure matrix (JEM) estimations of level and probability of exposure according to occupational history from tax records. Furthermore, three of the included studies dichotomized formaldehyde exposure into ever versus never exposed, although each attempted to account for duration and timing of exposure.

Finding: Overall, criteria for evaluating human studies were consistent with EPA’s state-of-practice methods. However, it is not clear whether the study quality criteria were applied uniformly across studies.

In evaluating animal studies, EPA considered several factors, including possible confounding due to coexposure to methanol, especially when high exposures (>10 mg/m³) were involved; the potential influence of irritation or changes in olfaction on behavioral measures, with preference given to behavioral studies with a period of latency between exposure and endpoint testing of at least 2 hours; blinding of the outcome assessors for nonautomated assessments; and a preference for outcomes that were deemed sensitive and specific for nervous system effects (Appendix A, Table A-86). Duration of exposure was also considered; however, studies of short-term or even acute duration were not considered to be less informative. Three animal studies (Aslan et al., 2006; Sarsilmaz et al., 2007; Sorg et al., 1998) were deemed to have medium confidence.

Finding: Where explanations are provided for confidence assessments for more recently published literature, EPA gives no details about the basis for confidence decisions for controlled exposure studies in humans (Appendix A, Table A-85).

Finding: Regarding human studies, study quality criteria (i.e., selection bias, sample size, exposure assessment, and confounding) were consistent with EPA’s state-of-practice methods. However, the reasons for final conclusions on study confidence are not always clear, and there are inconsistencies in the explanations for the confidence ratings. Potential updates to specific studies (e.g., Bellavia et al., 2021; Kilburn, 2000; Pinkerton et al., 2013; Schenker et al., 1982; see also Seals et al., 2017 and Peters et al., 2017) are detailed in Appendix E.

Finding: Regarding animal studies, study quality criteria were consistent with EPA’s state-of-practice methods.

Evidence Synthesis and Judgments

EPA summarizes its evidence synthesis for nervous system effects of formaldehyde in Table 1-50. For ALS, EPA determined that the human evidence was slight, that the animal evidence was indeterminate, and that an effect of formaldehyde on ALS would be surprising in terms of biological plausibility because of a lack of systemic distribution (and a lack of relevant mechanistic studies in humans). For developmental neurotoxicity, the human evidence was indeterminate, and the animal evidence was considered slight, although some evidence of relevant molecular and neurochemical changes in animals provided biological plausibility. For several types of neurobehavioral effects, the human evidence was considered indeterminate or slight, and the animal evidence was considered slight, with some animal evidence that provided potential biological plausibility.

Finding: With human studies having medium and high confidence for ALS and low or not informative confidence for neurobehavioral outcomes (e.g., memory, mood changes), along with medium confidence for animal studies on developmental neurotoxicity, the basis for the overall evidence judgment is not well articulated.

The 2022 Draft Assessment concludes that the underlying MOA for neurotoxicity for inhaled formaldehyde exposures is unknown. Although the text acknowledges that it is not likely that inhaled formaldehyde would be transported directly to the central nervous system, there are several potential indirect mechanisms for formaldehyde to impact the nervous system. Most notable is repeated sensory responses, which is recognized as a neurogenic pathway for immune and respiratory outcomes. However, the inflammatory responses could also elicit nervous system outcomes.

Finding: The 2011 NRC committee found the discussion of the neurogenic MOA in the 2010 Draft Assessment to be vague and speculative. In the 2022 Draft Assessment, EPA provides a much more thorough overview of potential MOAs without speculative conclusions.

Overall Conclusions About the Hazard Descriptor

The 2022 Draft Assessment notes that evidence on noncancerous nervous system effects is weak and lacking, and thus concludes that the evidence suggests, but is not sufficient to infer that inhaled formaldehyde may lead to adverse neurological outcomes in humans. This assessment is based on limited animal studies having medium confidence and several human studies having medium and high confidence for ALS and low or not informative confidence for neurobehavioral outcomes (e.g., memory, mood changes).

Finding: The conclusion that the evidence suggests but is insufficient to infer significant long-term neurotoxicity from inhaled formaldehyde in humans is consistent with EPA’s state-of-practice methods, given the number of studies with final assessments of confidence in the not informative to low range, and very few being assessed as having medium or high confidence.

Dose-Response Evaluation

EPA did not derive a POD or cRfC for nervous system effects following formaldehyde inhalation.