Review of the Formaldehyde Assessment in the National Toxicology Program 12th Report on Carcinogens (2014)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

3 Independent Assessment of Formaldehyde The second part of the committee’s task was to conduct an independent assessment of formaldehyde. The committee started with its peer review in Chapter 2 and the background document that supports the formaldehyde profile in the 12th RoC. It searched for additional peer-reviewed literature that had been published by November 8, 2013,1 and incorporated relevant human, experi- mental animal, and mechanistic studies into the independent assessment. The committee focused its attention on literature that contained primary data, but it also examined published review articles and reviews by other authoritative bod- ies to ensure that relevant literature was not missed and to ensure that all plausi- ble interpretations of primary data were considered. The committee considered comments and arguments presented to it during its first meeting, comments and documents received from other sources during the study process, and independ- ent literature searches carried out by National Research Council staff (see Ap- pendix D). The goals of the literature searches were to identify relevant litera- ture published around the time of the publication of the background document and later that may have missed inclusion in the 12th RoC and to identify any relevant literature that was published after the release of the 12th RoC. Each search covered the period from January 1, 2009 (the year in which the draft background document for formaldehyde was initially released; Bucher 2013), to November 8, 2013. Databases searched were PubMed, MEDLINE (Ovid), Em- base (Ovid), Scopus, and Web of Science. The search strategy for each database is described in Appendix D. After identifying the relevant body of literature up to November 8, 2013, the committee reviewed the primary data and applied the RoC listing criteria to human, experimental animal, and mechanistic studies. This chapter begins with a section on cancer studies in humans, which is followed by a section on cancer studies in experimental animals. The chapter then reviews toxicokinetic and metabolism literature and studies of mechanisms 1 The cutoff date for the literature search was chosen to allow the committee time to review the literature within the constraints of the project schedule. 66 

Independent Assessment of Formaldehyde 67 of carcinogenesis. It ends with a section that summarizes human, experimental animal, and mechanistic data and provides a conclusion and a listing recommen- dation for formaldehyde that is based on the listing criteria in the 12th RoC. The committee’s assessment of formaldehyde was guided by the RoC list- ing criteria, which were first introduced in the present report in Box 1-2. A sub- stance can be classified in the RoC as “reasonably anticipated to be a human carcinogen” if at least one of the following criteria is fulfilled (NTP 2010, p. iv):  “There is limited evidence of carcinogenicity from studies in humans, which indicates that causal interpretation is credible, but that alternative expla- nations, such as chance, bias, or confounding factors, could not adequately be excluded.”  “There is sufficient evidence of carcinogenicity from studies in experi- mental animals, which indicates there is an increased incidence of malignant and/or a combination of malignant and benign tumors (1) in multiple species or at multiple tissue sites, or (2) by multiple routes of exposure, or (3) to an unusu- al degree with regard to incidence, site, or type of tumor, or age at onset.”  “There is less than sufficient evidence of carcinogenicity in humans or laboratory animals; however, the agent, substance, or mixture belongs to a well- defined, structurally related class of substances whose members are listed in a previous Report on Carcinogens as either known to be a human carcinogen or reasonably anticipated to be a human carcinogen, or there is convincing relevant information that the agent acts through mechanisms indicating it would likely cause cancer in humans.” A substance can be listed as “known to be a human carcinogen” if “there is sufficient evidence of carcinogenicity from studies in humans, which indicates a causal relationship between exposure to the agent, substance, or mixture, and human cancer.” The RoC listing criteria are clear about the information needed to fulfill the criteria of sufficient evidence in experimental animals (see the sec- tion “Cancer Studies in Experimental Animals”). The type of information need- ed to meet the RoC listing criteria for limited or sufficient evidence in humans required more interpretation and expert judgment by the committee. To make the committee’s methods clear and transparent, the section “Cancer Studies in Humans” begins by describing the committee’s methodology for identifying and evaluating epidemiologic evidence and the committee’s interpretation and appli- cation of the listing criteria. CANCER STUDIES IN HUMANS Identification of Informative Epidemiologic Studies In its independent analysis of formaldehyde exposure and cancers, the committee first considered each of the epidemiologic studies cited in the back- ground document for formaldehyde. As discussed in Chapter 2, the National 

68 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Toxicology Program (NTP) did a thorough job of searching the literature for relevant human studies, so the committee used the background document as a starting point for its independent review. Second, the committee examined the results of the independent literature search described in Appendix D (see Box D- 1 and Figure D-1). One additional study (Coggon et al. 2014)—an update of Coggon et al. (2003)—was identified after the literature-search cutoff date and was included as part of the committee’s independent assessment. Third, the committee examined review articles, meta-analyses, and materials presented during its first meeting and during the study process. As part of its exclusion criteria (Box D-1), the committee based its assess- ment on the primary literature. It recognized that quantitative meta-analyses can be informative, but the heterogeneity of exposures in the primary literature on for- maldehyde makes it challenging to base any conclusions of causality on resulting summary estimates. The committee agrees with a previous National Research Council report that “meta-analysis can be a valuable method for summarizing evi- dence but can also be subject to variable interpretations depending on how litera- ture is selected and reviewed and data analyzed” (NRC 2011, p. 112). Evaluation of Epidemiologic Studies Several factors were considered in the evaluation of the strength of the ep- idemiologic literature. The principles of causal association, elaborated by Brad- ford Hill (Hill 1965), were used as a starting point for the evaluation of informa- tive epidemiologic studies. Of Bradford Hill’s original nine criteria, the committee focused on six: strength, consistency, specificity, temporality, biolog- ic gradient, and coherence. On the basis of the RoC listing criteria, plausibility was more relevant to supporting evidence from experimental animal studies and mechanistic data than to the evaluation of the epidemiologic evidence, and anal- ogy was not deemed to be a useful criterion for this topic. Coherence emerged as a particularly important criterion for similarity of findings among multiple study designs and populations (and is also related to consistency). The committee rec- ognizes that the Bradford Hill criteria can be useful guidelines for assessing causal association but agrees with NRC (2014, p. 91) that they “are by no means rigid guides to reaching ‘the truth’.” The committee also developed criteria for rating the quality and utility of epidemiologic studies and their exposure assessments, shown in Table 3-1. The development of the exposure-assessment evaluation is presented in detail in Ap- pendix C and summarized in Tables C-1 and C-2. In general, the committee judged a cohort or case–control study to be informative if it was large, had high and varied exposures that were systematically estimated, had reliably assessed cancer end points, and included credible comparison groups. Table 3-2 provides information about all the epidemiologic studies that the committee considered, including a description of the studies, a description of the exposure assessments used in each study, comments on strengths and limitations of the studies, and the committee’s determination of study quality (strong, moderately strong, or weak). 

Independent Assessment of Formaldehyde 69 TABLE 3-1 Criteria Used to Assess Epidemiologic Studies for Hazard Assessmenta Study Quality and Study Population, Design, Utility Classification Quality of Data, and Analysis Exposure Assessmentb Weak study: Modest or small population with Low discrimination between limited utility for few cases. Design limitations, exposed and control categories, hazard assessment; including broad case definition, qualitative or semiquantitative inconclusive; no duration of exposure, short evaluation, limited evidence of uninformative followup, limited data analysis substantial formaldehyde exposure Moderately strong Modest-sized population with few Moderate discrimination study: cases or a broad case definition; between high and low exposure somewhat useful for sufficient followup for latency; categories; substantial fraction hazard assessment standard data analysis of population probably highly exposed; qualitative, semiquantitative, or quantitative evaluation; use of duration of work as a proxy for exposure Strong study: Large population with many cases, High discrimination between highly useful for precise case definition, including high and low exposure hazard assessment subcategories; large number of categories, substantial fraction subjects with long-duration of population probably highly exposures; sufficient followup for exposed, detailed quantitative or latency; limited switching among highly selective semiquantitative exposure categories; sophisticated evaluation data analysis accounting for important potential confounders a The epidemiologic elements in the second column are not required to match with the exposure elements in the third column to define the study quality. b Exposure-assessment levels are based on the data presented in Appendix C and Table C-2. Source: Committee generated. The committee’s judgment of the strength of a study depended on both the epidemiologic design elements (the second column in Table 3-1) and the expo- sure-assessment dimensions (the third column in Table 3-1), which are some- what independent. A strong study might not have a highly developed exposure assessment. For example, several strong case–control studies of licensed em- balmers had no exposure assessments, but because the case definition required work as a licensed embalmer and that occupation has well-defined rules for practice (which define the exposure situation), the resulting studies were consid- ered to be strong or moderately strong. A well-designed study with a high- discrimination exposure assessment could be judged to be weak because few of the subjects were exposed to formaldehyde, as was the case, for example, in the textile studies. The overall strength of each study was assessed by considering all of the variables described in Table 3-1. 

70 TABLE 3-2 Description of Epidemiologic Studies Reviewed by the Committee Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd Andjelkovich et Cohort = 8,147 men; outcome: mortality; High-discrimination quantitative Followup since first exposure Moderately al. 1995 nasopharyngeal cancer = 1 case, exposure assessment; detailed work was short (≤19 years), total strong sinonasal cancer = 0 cases, history available for each study subject; duration of exposure was short Iron foundry workers lymphohematopoietic cancer = 15, extensive data from industrial-hygienist (≤17 years) from Michigan, USA leukemia = 5 cases; 3,929 workers with sampling, technical data from plant, potential exposure to formaldehyde for walk-through surveys, and job and task Although the study had a ≥6 months during 1960–1987; 83,064 descriptions; information assessed by an high-discrimination quantitative person-years for exposed and 40,719 industrial hygienist and assigned to high exposure assessment and the person-years for controls; a smoking- (median 1.5 ppm), medium (median 0.55 cohort was of a moderate size, history survey was administered via ppm), low (median 0.05 ppm), or no it was probably not large enough mail. formaldehyde-exposure categories; to detect risk of rare tumors, formaldehyde used in core-making such as nasopharyngeal cancer, operations in 1960–1987; all workers sinonasal cancer exposed to silica Armstrong et Population case–control; outcome: Low-discrimination qualitative exposure Formaldehyde exposure was Weak al. 2000 prevalent and incident cases; 282 cases assessment; exposure information limited (formaldehyde exposure with histologically confirmed gathered by structured interview to in only 9.0% of the sample, only General population nasopharyngeal cancers, ≥5 years of obtain complete dietary, residential, eight had accumulated ≥10 years of Maylasia residence in study area, and diagnosis in occupational history; exposures of exposure outside a 10-year 1987–1992; 282 cases and matched classified by broad Malaysian latency period); short latency controls identified from health-center occupational codes, industrial-hygienist period records in Kuala Lumpur and Selangor professional judgment among Malaysian Chinese Beane Freeman Cohort = 25,619; outcome: mortality from High-discrimination exposure Large, well-designed study Strong et al. 2009 lymphohematopoietic malignancy; all assessment; quantitative estimation and lymphohematopoietic types = 319 cases, job–exposure matrix used, but no No evidence of confounding by NCI study of US leukemia = 123 cases, myeloid leukemia = measurements after 1980; median other exposures chemical industry 88 cases; followup period: 1966–2004 exposure intensity was 0.3 ppm (range and plastics workers 0.01–4.3 ppm); median peak exposure Study was able to assess peak in 10 plants was about 2 ppm; about 25% were exposures exposed at >4ppm 

Beane Freeman Cohort = 25,619; outcome: mortality; High-discrimination exposure Large, well-designed study Strong et al. 2013 nasopharyngeal cancer = 10 deaths; assessment; extensive background data sinonasal cancer = 5 deaths; followup and samples; quantitative estimation and No evidence of confounding by NCI study of US period: 1966–2004; update of Hauptman job–exposure matrix used on the basis of other exposures chemical industry et al. (2004) extensive data, but no measurements and plastics workers after 1980; Beane Freeman et al. (2009) Study was able to assess peak in 10 plants reported the median exposure intensity exposures of 0.3 ppm (range 0.01–4.3 ppm); median peak exposure was about 2 ppm; about 25% were exposed at ˃4 ppm Bertazzi et al. 1989 Cohort = 1,332 men; outcome: mortality; Moderate-discrimination qualitative Evidence of increasing mortality Moderately hematologic neoplasms = 7 deaths, lung exposure assessment; cohort members from hematologic neoplasms strong Italian resin workers cancer = 24 deaths, larynx tumors = 6 worked in a department that used with longer latency; highest deaths; followup period: 1959–1986 formaldehyde; exposure intensity in increase in mortality was in many locations peaked at >3.0 ppm those who were employed during 1965–1969, an early period of high exposure Blair et al. 2001 Population-based leukemia case–control; Low-discrimination semiquantitative There were 513 incident cases, Weak outcome: incidence; 513 incident cases; exposure assessment for formaldehyde; but the study was judged to be General population ascertainment period: Iowa 1981–1983, broad job categories and industries; weak for assessing in Iowa and Minnesota 1980–1982 potential formaldehyde exposure was formaldehyde because the Minnesota categorized on a 4-point scale; likely number of cases with high high misclassification exposure (n = 3) was small, misclassification likely Checkoway et Case–cohort nested within cohort of Low-discrimination qualitative exposure Few workers exposed to Weak al. 2011 267,400 women textile workers; outcome: assessment (yes/no) for formaldehyde; formaldehyde (2 lung-cancer lung cancer incidence; 628 cases detailed job histories and job–exposure cases were exposed to Female textile diagnosed in 1989–1998 matrix used to assign detailed textile- formaldehyde) workers in Shanghai, dust and related exposures for all China workers for all years; exposure to formaldehyde was uncommon in these workers (Continued) 71 

72 TABLE 3-2 Continued Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd Coggon et al. 2014 Cohort = 14,008; outcome: mortality; Moderate-discrimination semiquantitative Cohort was small and Moderately nasopharyngeal cancer = 1 death, nose exposure assessment; work histories satisfactory for cancers that were strong Chemical workers in and nasal sinus cancer = 2 deaths abstracted from company employment more common, but probably too 6 British factories leukemia = 54 deaths, myeloid leukemia records; jobs were classified into five small to detect nasopharyngeal where formaldehyde = 36 deaths; followup period: 1941– exposure categories (background, low, and sinonasal cancers and only was produced or used 2012; update of Acheson et al. (1984) moderate, high, or unknown) by had moderate power to detect and Coggon et al. (2003) industrial-hygiene professional judgment; myeloid leukemia effects limited quantitative measurements available after 1970 covering many jobs, Authors reported a concern quantitative exposure assumed to be the about the quality of data when same before 1970 (although anecdotal, the they made exposure assignments reported exposures were much higher earlier in followup period); "high" exposure category was estimated to be over 2 ppm; no peak exposures identified; authors noted that there was some exposure to paraformaldehyde Dell and Teta 1995 Cohort = 5,932; outcome: mortality; Low-discrimination qualitative exposure Small study size had little power Weak nasopharyngeal cancer = 0 deaths, assessment; company job histories to detect risk of rare tumors Workers employed in sinonasal cancer = 0 deaths, collected; duration of employment used a Union Caribide lymphohematopoietic cancer = 28 deaths, as a surrogate for cumulative exposure; Few workers exposed to plastics leukemia and aleukemia = 12 deaths; some analysis of work department made formaldehyde manufacturing plant workers employed in 1946–1967; but limited by missing work data Limited exposure information in New Jersey followup through 1988; 5,932 males in the cohort (111 exposed to formaldehyde) Multiple concomitant exposures (raw materials used in the manufacturing process included asbestos [usually chrysotile], carbon black, epichlorohydrin, polyvinyl chloride, acrylonitrile, styrene, chemical additives [such as plasticizers, emulsifiers, and antioxidants]) 

Edling et al. 1987 Cohort = 521 men; outcome: mortality; Low-discrimination semiquantitative Small study size had little power Weak nasopharyngeal cancer = 0 deaths; exposure assessment; very limited to detect risk of rare tumors Workers in abrasive sinonasal cancer = 0 deaths, leukemia = formaldehyde exposure data from 1970s; manufacturing in 1 death; men with ≥5 years of two work areas had exposures; blue- Few workers exposed to Sweden employment in 1955–1983; followup collar workers assigned exposures; no formaldehyde period: 1958–1983 data on how many were exposed Limited exposure information Hall et al. 1991 Cohort = 4,512 men; outcome: Low-discrimination qualitative exposure Small study size had little power Weak mortality; nasopharyngeal cancer = 0 assessment on the basis of job title to detect risk of rare tumor UK pathologists cases, sinonasal cancer = 0 cases, (formaldehyde exposure was assumed leukemia = 4 cases; men identified in from cadavers); no discussion of High likelihood of 1973 Royal College of Pathologists exposure conditions was presented misclassification on exposure to membership list; followup period: 1974– formaldehyde; pathologists have 1987 less likelihood of exposures than embalmers Hansen and Olsen Cohort = 91,182 men with cancer, 2,041 Moderate-discrimination Study limited by lack of data on Weak 1995, 1996 men with longest work experience of semiquantitative exposure assessment; intensity of exposures and ≥10 years before the date of diagnosis of potentially exposed cases were identified internal plant operations Danish data-linkage cancer, 265 companies where as those with ≥10 years of blue-collar study identifying formaldehyde was used; outcome: work experience in formaldehyde-using Cohort had no or few cases of incident cancers in incidence; nasopharyngeal cancer = 4 companies; formaldehyde exposures some types of cancers, and this companies in which cases, cancer of the nasal cavity = 13 were ranked as low (white-collar jobs) limited its utility formaldehyde was cases, leukemia = 39 cases; cancer and high (blue-collar jobs) with no used diagnosed in 1970–1984; cases obtained wood-dust or high wood-dust exposure; from national cancer registry, linked to no workplace assessment of exposure national employment data and industry conditions or plant size were made, so reporting on chemical use high potential for misclassification by exposure intensity (for example, a large plant may only have a few workers out of a large workforce who are exposed) (Continued) 73 

74 TABLE 3-2 Continued Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd Hauptmann et Nested case–control; outcome: mortality; High-discrimination exposure assessment; No confounding by smoking Strong al. 2009 nasopharyngeal cancer = 4 cases, methods included quantitative lymphohematopoitic cancers = 168 reconstruction with statistical modeling, Strong trend with years in US funeral directors, cases, myeloid leukemia = 34 cases, sensitivity analyses; average exposure embalming; trends with average embalmers brain cancer = 48 cases; those who died intensity while embalming was 1.5–1.8 and peak exposure in 1960–1986; update of Hayes et al. ppm and average peak exposures was 8.1– (1990) 10.5 ppm depending on case group Hayes et al. 1986 Case–control; outcome: incidence; Moderate-discrimination qualitative Study limited by disagreement Moderately histological types of sinonasal cancer = exposure assessment; work history between exposure assignments strong General population 116 cases; cancer diagnosed in 1978– collected by interview included all jobs of 2 independent raters, but the in the Netherlands 1981; cases drawn from all six major held for 6 months or more; all jobs were association of formaldehyde hospitals for treatment of head and neck classified by industrial hygienists exposure and nasal cancer was tumors according to level and probability of similar for each rater formaldehyde exposure on 10-point scale; agreement between two raters was poor For sinonasal cancer, the study for adjacent scores, and this resulted in suggests an association between high potential for misclassification in formaldehyde and squamous- adjacent categories, which was rare for cell carcinoma, not high to low or low to high adenocarcinoma Hildesheim et Population case–control; outcome: Moderate-discrimination semiquantitative Considerable overlap in wood Moderately al. 2001 incidence; nasopharyngeal cancer = 375 exposure assessment; occupational history dust, formaldehyde exposures; strong cases; newly diagnosed, histologically data obtained by interview; exposures authors were concerned about General population confirmed nasopharyngeal cancer in were assigned to broad occupation codes greater misclassification for in Taiwan people younger than 75 years old who on basis of professional judgment of study formaldehyde than wood-dust were residents of Taipei City or County industrial hygienist; exposures were assignments for ≥6 months; cases identified at 2 classified from 0 (not exposed) to 9 tertiary-care hospitals; population-based (strong) according to probability, ˃95% of cases were positive for controls drawn from national housing intensity, and duration of formaldehyde Epstein Barr virus registry exposure; 74 cases exposed to formaldehyde; dietary factors and coexposure to cigarette smoking, wood dust, and solvents were assessed 

Levine et al. 1984 Cohort = 1,4777; outcome: mortality; Embalmers have well-defined, high Cohort was small and the study Moderately nasopharyngeal cancer = 0 deaths, exposures to formaldehyde; embalmer probably had little power to strong Licensed embalmers sinonasal cancer = 0 deaths, larynx = 1 exposure can be sharply discriminated detect risk of rare in Ontario, Canada death, lymphohematopoietic cancer = 8 from that of other job groups; job and nasopharyngeal and sinonasal deaths, leukemia = 4 deaths; 34,774 formaldehyde sources defined by cancers person–years of observation during regulations and training 1950–1977, 17,589 of which occurred ≥20 years since first licensure Li et al. 2006 Cohort = 267,400; outcome: incidence; Low-discrimination qualitative exposure Limited use of formaldehyde in Weak nasopharyngeal cancer = 67 cases, assessment for formaldehyde, which was textile operations; very few Chinese female sinonasal cancer = 10 cases; cases secondary to a primary evaluation of workers exposed (only 10 cases textile workers in identified in 1989–1998; 267,400 female textile production exposures; complete exposed to formaldehyde and 526 factories in textile workers drawn in 1925–1958 occupational history in textile industry none of the NPC cases were Shanghai was collected; factory profile form was classified as exposed) used by industrial hygienists in Shanghai to record for each factory production processes, types of workshops, and historical measurements of hazardous exposures since establishment of factory Luce et al. 1993 Case–control; outcome: incidence; Moderate-discrimination semiquantitative High correlation between wood Moderately sinonasal cancer = 207 cases; cases with exposure assessment; work history dust, formaldehyde exposure strong General population primary malignancies of the nasal cavity collected by interview; industrial hygienist limited ability to estimate in France and paranasal sinuses diagnosed in classified all jobs for probability of formaldehyde effect separately 1986–1988; cases obtained from 27 exposure (unexposed, possible, probable, hospitals, hospital and community definite); 107 cases with exposure to controls; analyses performed separately formaldehyde; formaldehyde for squamous-cell carcinoma and concentrations in exposed jobs estimated adenocarcinoma, the two major as low (<0.1 ppm), medium (0.1–1.0 histologic types ppm), high (>1.0 ppm); authors evaluated coexposures to wood dust (Continued) 75 

76 TABLE 3-2 Continued Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd Luce et al. 2002 Case–control; outcome: incidence; type High-discrimination exposure Statistical modeling used to Strong of nasopharyngeal cancer: assessment; uniform methods used in all evaluate effects of concurrent General populations adenocarcinoma = 195 cases, squamous- studies to gather detailed job wood-dust and formaldehyde in 7 countries cell carcinoma = 432 cases; cancer cases information; job titles and industries exposure. diagnosed in 1968–1990; pooled data coded uniformly; quantitative exposure from 12 case–control studies in seven data used to construct job–exposure countries matrix; hygienists assigned probabilities and intensities of formaldehyde exposure; cumulative exposure was principal summary measure of exposure; 192 cases with medium or high exposure to formaldehyde; authors evaluated effects of coexposures to wood dust Luo et al. 2011 Ecologic study; outcome: SEER lung- Low-discrimination semiquantitative Caution needed in interpreting Weak cancer incidence rates by county; data on exposure assessment; county-level ecologic associations as causal; General population age-adjusted lung-cancer incidence rates quantitative data on industrial release of high potential for in 13 US regions in 1992–2007; county-level correlation formaldehyde as proxy for general misclassification in large covered by SEER of Toxics Release Inventory data on population exposure in the county counties registries formaldehyde release with lung-cancer incidence rate from the SEER database Mahboubi et Population-based case–control study; Moderate-discrimination Large, well-conducted study; Moderately al. 2013 outcome: lung-cancer incidence; 1,595 semiquantitative exposure assessment; broad job titles limit strong male cases and 465 female cases; detailed job information gathered by discrimination General population interviews conducted in two periods: questionnaire; job titles and industries in Montreal, Canada 1979–1986 and 1996–2002. coded uniformly; hygienists assigned Little or no evidence of an confidence, relative concentration, and association with lung-cancer frequency of formaldehyde exposure; 99 incidence cases with “substantial” exposure to formaldehyde; authors evaluated effects of confounding by smoking and other exposures 

Meyers et al. 2013 Cohort = 11,043; outcome: mortality; High-discrimination quantitative Historical data on free Strong nasopharyngeal cancer = 0 deaths, exposure assessment; personal exposure formaldehyde in textile fabrics US garment-industry sinonasal cancer = 0 deaths, samples for formaldehyde from 549 strongly suggest that exposures workers lymphohematopoietic cancer = 107 randomly selected employees in five before 1970 were at least an deaths, leukemia = 36 deaths, myeloid different departments from the 1980s; order of magnitude higher than leukemia = 21 deaths; workers employed Pinkerton et al. (2004) reported exposures in the 1980s and later for ≥3 months after introduction of geometric mean 8-hr TWA of 0.09 ppm– (Elliot et al. 1987) formaldehyde-treated fabric into 0.20 ppm, overall geometric mean production process (1959 in facilities 1 concentration of 0.15 ppm; area Although the study design was and 2, 1955 in facility 3); followup monitoring showed that formaldehyde judged to be strong, the cohort through1998; update of Stayner et al. concentrations were essentially constant was probably not large enough (1985) and Pinkerton et al. (2004) without substantial peaks or intermittent to detect an effect for rare exposures cancers, such as nasopharyngeal cancer, sinonasal cancer Olsen and Case–control; outcome: incidence; Moderate-discrimination qualitative Only small numbers of cases Moderately Asnaes 1986 nasopharyngeal cancer = 293 cases, exposure assessment; employment ever exposed to formaldehyde strong sinonasal cancer = 466 cases; histories obtained from national pension, (13 cases of squamous-cell General population histologically confirmed cancer cases in population registries and exposure carcinoma; 17 cases of in Denmark 1970–1982; male cases and controls classified by job description, industry; adenocarcinoma ever exposed selected from Danish Cancer Registry each job rated by industrial hygienist as to formaldehyde); few with unexposed to formaldehyde, probably or formaldehyde exposure and certainly exposed, or unknown; wood- no wood-dust exposure products industry is widespread in Denmark No evidence of confounding by wood dust or smoking Ott et al. 1989 Nested case–control; outcome: mortality; Low-discrimination qualitative exposure Exposures not localized in Weak lymphohematopoietic cancer = 129 assessment for formaldehyde; broad job production areas, probably Two Union Carbide cases, leukemia = 59 cases; cases and plant departments with many resulting in likely broad facilities identified from review of causes of death exposures and few cases of formaldehyde misclassification among males from the Rinsky et al. exposure; formaldehyde exposure was (1987) cohort who died during 1940– assigned on the basis of work in a Multiple concomitant exposures 1978; Union Carbide facilities also department that used formaldehyde (raw materials used in the evaluated by Dell and Teta (1995) manufacturing process, including asbestos [usually chrysotile], carbon black, (Continued) 77 

78 TABLE 3-2 Continued Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd epichlorohydrin, polyvinyl chloride, acrylonitrile, styrene, chemical additives [such as plasticizers, emulsifiers, and antioxidants]) Partanen et al. 1993 Nested case–control; outcome: Moderate-discrimination qualitative Medium formaldehyde Moderately incidence; Hodgkin disease = 4, non- exposure assessment; methodology exposures likely, but study strong Finnish wood- Hodgkin lymphoma = 8, leukemia = 12; assigned exposure based on personal limited by small number of cases industry workers cancer cases diagnosed in 1957–1982 work histories and a job–exposure matrix that identified formaldehyde exposure; no average exposure intensity was provided Pesch et al. 2008 Industry-based case–control; outcome: Low-discrimination qualitative exposure Strong study of wood-dust Weak incidence; histologically confirmed assessment of formaldehyde; association with sinonasal German wood sinonasal cancer = 86 cases; recognized questionnaire collection of occupational cancer, but weak assessment of industry occupational disease diagnosed in 1994– history with additional data on wood- formaldehyde exposure 2003; cases identified from workers related exposures and chemical insured by Holz-BG insurance company treatments, including formaldehyde; Substantial exposure personal sampling for wood-dust exposure misclassification was likely in 1992–2002; expert industrial hygienists estimated wood-dust exposure to identify missing information and trends; crude assessment of formaldehyde exposures (yes/no) with no measurements; 47 cases exposed to formaldehyde (54.6%), an equal fraction of controls Richardson et al. Population-based case–control study of Low-discrimination semiquantitative Broad job categories; likely high Weak 2008 non-Hodgkin lymphoma and chronic exposure assessment; yes/no estimates of misclassification of exposure lymphocytic leukemia; outcome: formaldehyde exposure derived from job- General population incidence; non-Hodgkin lymphoma = history data and a job–exposure matrix in Germany 858 cases; newly diagnosed cases that that used broad job and industry groups occurred in 1986–1998 

Roush et al. 1987 Case–control; outcome: incidence; Low-discrimination semiquantitative Broad job categories; likely Weak nasopharyngeal cancer = 173 cases, exposure assessment; occupational misclassification General population sinonasal cancer = 198 cases; histories obtained from death certificates, in Connecticut histologically confirmed cases were from city directories; exposures were assigned Risk estimates adjusted for Connecticut Tumor Registry among to broad occupation codes on basis of smoking, race, and other risk males who died from any cause in 1935– industrial-hygienist professional factors 1975, controls from death certificates judgment; high exposure ≥1 ppm Siew et al. 2012 Cohort = 1.2 million working Finnish Moderate-discrimination quantitative Few cases with formaldehyde Moderately men; outcome: incidence; nose = 292 exposure assessment; occupation in 1970 exposure for three of the four strong Finnish general cases, nasal squamous-cell carcinoma = linked to job–exposure matrix to types of cancer investigated (17 population 167 cases, nasopharyngeal cancer = 149 estimate wood-dust exposure, cases of cancer of the nose, 9 cases; followup period: 1971–1995; data formaldehyde exposure, coexposures to cases of nasal squamous-cell linkage for all men born in 1906–1945 asbestos and silica; exposure assessment carcinoma, 5 cases of who were employed in 1970 completed by professional industrial nasopharyngeal cancer, and hygienists 1,831 cases of lung cancer with any exposure to formaldehyde) Significant lung cancer– formaldehyde association may have resulted from residual confounding by smoking, wood dust, asbestos, or crystalline silica Stellman et Cohort = 362,823 men enrolled in the Low-discrimination qualitative exposure High potential for Weak al. 1998 Cancer Prevention Study-II, 45,399 assessment; questionnaire given to self- misclassification in self- men employed in a wood-related identified wood workers and others with reporting exposure to American Cancer occupation, reported exposure to wood wood-dust exposure or people who formaldehyde Society Prevention dust, or both; outcome: cancer mortality; reported exposure to formaldehyde Study II sinonasal cancer = 1 death, (yes/no), asbestos nasopharyngeal cancer = 2 deaths, lymphohematopoietic cancer = 122 deaths, non-Hodgkin lymphoma = 51 (Continued) 79 

80 TABLE 3-2 Continued Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd deaths, Hodgkin lymphoma = 5 deaths, multiple myeloma = 20 deaths, leukemia = 46 deaths; followup period 1982–1988 Stern 2003 Cohort = 9,352 men; outcome: all Low-discrimination exposure Few cases with formaldehyde Weak mortality; nasal = 1 death, leukemia and assessment; personnel records were exposure; standardized mortality US tannery workers aleukemia = 16 deaths; included all reviewed, subjects were grouped into ratio for workers in finishing production workers employed for any five departments; semiquantitative department potentially exposed length of time at tannery A in 1940– potential exposure depended on to formaldehyde 1979 or at tannery B during 1940–1980; departments; Stern et al. (1987) reported followup through 1993; study is an that ambient formaldehyde was extension of Stern et al. (1987) measured in finishing department at time of study and was 0.5–7.0 ppm (mean 2.45 ppm) Stroup et al. 1986 Cohort = 2,317 men; outcome: all Moderate-discrimination exposure Exposure was defined aspect of Moderately mortality; buccal cavity and pharyngeal assessment; job structure strongly related job and varied according to type strong Anatomists living in cancer = 1 death, nasal cavity and to exposure; details available for of anatomist the United States sinuses = 0 deaths, lymphohematopoietic duration of association membership and cancer = 18 deaths, leukemia = 10 time period in which anatomists joined deaths, myeloid leukemia = 5 total the association, which were divided into deaths; men who joined American thirds to provide a crude surrogate of Association of Anatomists and lived in cumulative exposure to formaldehyde; United States during 1888–1969 information on research and teaching interests, department affiliations, and membership in other professional associations used to categorize each anatomist as specialist in gross anatomy, microanatomy, both, or neither; on basis of a review of reference materials and on discussions with anatomists who were 

familiar with laboratory techniques used in past, gross anatomists may have been exposed to formaldehyde more frequently than microanatomists Vaughan et Population–based case–control; Moderate discrimination Occupational-exposure Moderately al. 1986a outcome: incidence reported to cancer semiquantitative exposure assessment; prevalence was much lower than Strong registry; all incident cases of pharyngeal jobs obtained from interview histories in West et al. (1993) General population cancer (27 cases diagnosed during 1980– were assigned to broad occupation in western 1983) and sinonasal cancer (53 cases codes; likelihood and intensity of Only 3.5% of jobs had any Washington state diagnosed during 1979–1983) in exposure were assigned on basis of formaldehyde exposure (11 persons between 20–74 years old who industrial-hygienist professional cases of nasopharyngeal cancer resided in the study area judgment in a 4-category variable; and 12 cases of sinonasal cancer formaldehyde exposure associated with exposed to formaldehyde above making wood products background levels) Vaughan et al. Population–based case–control; outcome: Moderate-discrimination semiquantitative Although questionnaire data have Moderately 1986b incidence reported to cancer registry; all exposure assessment; subjects’ residential limited discrimination of past strong incident cases of nasopharyngeal cancer histories, including types of dwelling, exposures, living in a mobile General population (27 cases diagnosed in 1980–1983) and were determined from structured home has been associated with in western sinonasal cancer (53 cases diagnosed in telephone interview, which also collected high formaldehyde exposure in Washington state 1979–1983) in persons between the ages smoking, alcohol, and demographic period of about 1950 to of 20–74 who resided in the study area information; residential history since 1950 middle1980s included type of dwelling, use of urea- formaldehyde foam insulation, and occurrence of home renovation or new construction with particle board or plywood; information collected on lifetime occupational history to adjust for potential confounding Vaughan et al. 2000 Population-based case–control; outcome: High-discrimination quantitative Large, well-conducted study with Strong incidence; 196 newly diagnosed exposure assessment; detailed job, high-discrimination exposure General population nasopharyngeal cancer cases in 1987– industry data from structured interviews; assessment; no assessment of in catchment of 5 US 1993; cases were identified prospectively each job assessed on basis of industrial- peak exposures performed cancer registries (Continued) 81 

82 TABLE 3-2 Continued Reference and Study Population Study Informationa Exposure Assessmentb Critique and Conclusionsc Study Qualityd in five population-based cancer registries hygienist professional judgment for in United States; controls identified by probability of exposure and, if exposed, random-digit dialing; expanded exposure the 8-hr TWA; estimated 8-hr TWA (low evaluation relative to Vaughan et al. <0.10 pm; moderate ≥0.10,<0.50 ppm; (1986a) and high ≥ 0.50 ppm); 13.2% of jobs had ≥10% probability of exposure; coexposure to wood dust was also assessed for each job Walrath and Cohort = 1,132 men; outcome: mortality; Embalmers make up group that has well- Although the cohort was small, Moderately Fraumeni 1983 nasopharyngeal cancer = 0 deaths, defined high exposures to formaldehyde; exposures likely to have been strong sinonasal cancer = 0 deaths, tasks and formaldehyde sources are substantial with good New York state lymphohematopoitic cancer = 25 deaths, defined by regulations, training; double discrimination and qualitative embalmers and leukemia = 12 deaths, myeloid leukemia licensure—embalmer and funeral distinctions between exposed funeral directors = 6 deaths, nonwhites had 3 deaths from director—has fewer exposure and not exposed lymphohematopoitic cancer; persons opportunities who died in 1925–1980; 1,132 white, Cohort probably not large male embalmers and funeral directors enough to detect risk of rare licensed in 1902–1980; no duration of cancers, such as sinonasal employment or length of licensure cancer, nasopharyngeal cancer available; persons who held only funeral director’s license were not included Walrath and Cohort = 1,007 men; cohort: mortality; Embalmers make up group that has well- Although the cohort was small, Moderately Fraumeni 1984 sinonasal cancer = 0 deaths, defined, high exposures to exposures likely to have been strong lymphohematopoietic cancer = 19 formaldehyde; tasks and formaldehyde substantial with good California state deaths, leukemia = 12 deaths, myeloid sources are defined by regulations, discrimination and qualitative licensed embalmers leukemia = 6 deaths; men who died in training; length of licensure used as distinctions between exposed 1925–1980; white male embalmers surrogate of length of employment and not exposed licensed in 1916–1976; 1,109 deaths; duration of licensure was available but Cohort probably not large not employment enough to detect risk of rare cancers, such as nasal cancer, nasopharyngeal cancers 

West et al. 1993 Population case–control; outcome: Moderate-discrimination Association with formaldehyde Moderately incidence; nasopharyngeal cancer = 104 semiquantitative exposure assessment; was stronger for participants strong General population cases; followup period: unknown; cases exposure (yes/no) assigned to specific who were positive for Epstein in the Philippines identified at Philippines General job groups on basis of industrial- Barr virus Hospital; two types of controls selected: hygienist professional judgment; for hospital (n = 104) and community those exposed, several duration variables No evidence of confounding or controls (n = 101) were calculated effect modification by wood dust or other exposures; estimates adjusted for age, sex, education, ethnicity a The study information includes the study type, size of cohort, outcome type, followup period or source of cases and ascertainment period, and prior studies of the same population. The study information also includes the total number of cases by cancer type, which may differ from the number of cases in other tables in Chapter 3 (Tables 3-3–3-7 give the number of cases exposed to formaldehyde). b The exposure-assessment information includes the overall discrimination strength of the study, key data (such as work histories, exposure data and data on jobs, tasks, operations, and key history dates), professional industrial-hygienist data analysis, classification of exposures and met- rics used, and data on coexposures. See Table 3-1 and discussion of exposure assessment in Appendix C for descriptions and definitions of terms used in this column. c The committee’s critique and conclusions include information on critical study strengths and limitations. d The committee’s judgment of the study quality according to the criteria that it developed and presented in Table 3-1. Abbreviations: ICD, International Classification of Diseases; NCI, National Cancer Institute; ppm, parts per million; SEER, Surveillance, Epi- demiology, and End Results program of the National Cancer Institute; TWA, time-weighted average. Source: Committee generated. 83 

84 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Tables 3-3–3-7 present the number of exposed cases for strong and mod- erately strong studies as a particularly useful indicator of study power. When both disease and exposure are rare, the number of exposed cases will be an im- portant determinant of power (Thomas 2009). The number of exposed cases also has merit because it allows a comparison of size (in the common sense that big- ger studies are more powerful) of both case–control and cohort studies. The def- inition and ascertainment of exposed differs among studies and within some studies, so it was sometimes necessary for the committee to make a judgment about which definition to use when choosing the data to present in Tables 3-3– 3-7. The reader is referred to the primary literature to view all data and summary measures of exposure reported by specific studies. As discussed in Chapter 2, particular attention was paid to the choice of summary measures of exposure. Ideally, an epidemiologist chooses the appropri- ate measure to summarize exposure data on the basis of an understanding or hy- pothesis about the pharmacokinetics and pharmacodynamics of the exposure-to- dose and dose-to-response processes (Checkoway et al. 2004; Smith and Kriebel 2010). The investigators studying the association between formaldehyde and can- cer have little information on which to base that choice. In practice, therefore, it is common and appropriate to test the associations by using several different sum- mary measures, including cumulative exposure, average exposure, duration of exposure, and peak exposure. It is expected that, on average, choosing the wrong metric will result in an underestimation of an association if one exists (Checkoway et al. 2004)—that is, it is not expected that choosing the wrong summary measure of exposure will create evidence of an association where one does not exist except by chance. Another factor that complicates the assessment of risks by alternative met- rics is the imprecision and other limitations of the exposure-intensity data on which the summary measures are based. As discussed above, those data are of- ten only approximations and are likely to have substantial uncertainty. That makes it even more difficult to assert with confidence that one summary meas- ure is more likely than another to be “correct”. For those reasons, the committee looked at the measures of association between cancer risk and all the available summary measures presented in each study rather than choosing or preferring one a priori. Furthermore, patterns in disease associations and associated confi- dence intervals from smaller studies that did not reach traditional significance— that is, a p value less than 0.05 and the exclusion of 1.0 from the 95% confi- dence interval (CI)—were not discarded in the committee’s evaluation of the literature; they were weighed as weaker but still relevant evidence of consisten- cy in the results. The committee reviewed the available literature on the topic of which ex- posure metrics are more appropriate for environmental and occupational cancer studies. There is a long history of using cumulative exposure (the product of average intensity and exposure duration) as the summary measure of exposure 

Independent Assessment of Formaldehyde 85 (Checkoway et al. 2004). Cumulative exposure tends to be proportional to dis- ease risk and loss of function due to nonmalignant respiratory diseases caused by dusts, such as coal dust, silica, and asbestos. Possibly because of that con- sistency, cumulative exposure has often been used as the summary measure of exposure for other exposures and other diseases, including cancer. But in the few cases in which data are adequate for examining the relative performance of different exposure metrics, it has been found that cumulative exposure is gener- ally not proportional to cancer risk and should not necessarily be assumed to be the correct summary measure of exposure for cancer risk. Evidence for this find- ing first came from the studies of Doll and Peto (1978) on smoking and lung cancer, which found that lung cancer risk was not directly proportional to cumu- lative tobacco exposure (packs/day smoked multiplied by the years of smoking). Cumulative exposure also does not appear to be an appropriate measure for evaluating asbestos exposure and risk of mesothelioma (Peto et al. 1982) and for both asbestos and silica and risk of lung cancer (Zeka et al. 2011). More recent- ly, Richardson (2009) showed that leukemia risk was not proportional to cumu- lative benzene exposure. In the absence of knowledge about which outcome measure is applicable, the committee concluded that there was no compelling reason to prefer findings for one of the standard exposure metrics mentioned above over another. And, as noted above, the pattern of findings on all available metrics should be evaluated, data permitting. Consistent with the RoC listing criteria, the committee used its expert sci- entific judgment to interpret and apply the listing criteria. Limited evidence was defined by the committee as evidence from two or more strong or moderately strong studies with varied study designs and populations that suggested an asso- ciation between exposure to formaldehyde and a specific cancer type, but whose limitations led the committee to conclude that alternative explanations—such as chance, bias, and confounding factors—could not be adequately excluded and that therefore a causal interpretation could not be accepted with confidence. Suf- ficient evidence was defined by the committee as consistent evidence from two or more strong or moderately strong studies with varied study designs and popu- lations that suggested an association between exposure to formaldehyde and a specific cancer type and for which chance, bias, and confounding factors could be ruled out with reasonable confidence because of the study methodologies and the strength of the findings. Consistent with those definitions, the presence of negative findings in other studies, especially weak studies, did not necessarily negate positive findings. Nasopharyngeal Cancer The committee reviewed the literature on epidemiologic studies of for- maldehyde and nasopharyngeal cancer (see Table 3-3). Vaughan et al. (2000) was a large multicenter case–control study that was conducted in a general 

86 TABLE 3-3 Studies of Nasopharyngeal Cancer and Formaldehyde Exposure Reference and Study Population No. NPC Cancer Cases in Exposed Findings (95% CI) Beane Freeman et al. 2013 NPC defined by ICD-8 147; number of OR for highest average intensity of exposure (≥1 ppm) = 11.54 cases identified from Tables 2–4 in the (1.38–96.81) NCI study of US chemical industry publication and plastics workers in 10 plants OR for highest peak exposure category (≥4 ppm) = 7.66 (0.94– n=8 62.34) and test for trend with increasing peak categories p < 0.005 OR for highest cumulative exposure category (≥5.5 ppm–years) = 2.94 (0.65–13.28) Hildesheim et al. 2001 Histologically confirmed NPC; number OR for >10 years of exposure = 1.60 (0.91–2.90) cases identified from Table 2 in the General population in Taiwan publication OR among formaldehyde-exposed subjects who were positive for Epstein Barr virus = 2.6 (0.87–7.70) Ever exposed to formaldehyde: n = 74 >20 years since first exposure: n = 55 Siew et al. 2012 Histologically confirmed NPC; number of RR (adjusted for wood-dust exposure) for any formaldehyde cases identified from Table 3 in the exposure compared with no formaldehyde exposure = 0.87 (0.34– Finnish general population publication 2.20) Any exposure to formaldehyde: n = 5 Vaughan et al. 1986a,b NPC defined by ICD code 146-149: OR (adjusted for smoking and race) for highest exposure score = number of cases identified from Tables 3 2.1 (0.6–7.8) General population of western and 5 in Vaughan et al. (1986a) and Table Washington state 2 in Vaughan et al. (1986b) OR for ≥10 years occupational exposure = 1.6 (0.4–5.8) n = 11 OR for ≥10 years of residence in mobile home = 5.5 (1.6–19.4) 

Vaughan et al. 2000 ICD-O codes used to classify according to OR for highest cumulative exposure category (1.10 ppm–years) three histologic groups of NPC; number of = 3.0 (1.3–6.6) General population in catchment of cases identified from Table 2 of the 5 US cancer registries publication Positive trend in disease frequency over categories of cumulative exposure (p = 0.033) Ever exposed: n = 79 Wood-dust exposure and smoking had little effect on the Duration >5 years: n = 55 relationship with formaldehyde West et al. 1993 Histologically confirmed NPC; number of OR for ≥25 years since first exposure = 4.0 (1.3–12.3) cases identified from Table 2 of the General population in the publication OR derived from the final model that was adjusted for concurrent Philippines effects of education, diesel and dust, smoking, processed meats, n = 26 fresh fish, mosquito coils, and herbal medicines (In some calculations in Table 2 of the publication, n = 27) Abbreviations: ICD, International Classification of Diseases; NCI, National Cancer Institute; NPC, nasopharyngeal cancer; OR, odds ratio; ppm, parts per million. Source: Committee generated. 87 

88 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens population. Incidence data were collected from the National Cancer Institute (NCI) Surveillance, Epidemiology, and End Results Program registries. The study was identified as a strong study (Table 3-1). There were 24 nasopharynge- al-cancer cases in the highest category of cumulative exposure, so this study was one of the largest that the committee reviewed for nasopharyngeal cancer. Its methods included a quantitative exposure assessment with moderate discrimina- tion of who was exposed and the intensity of exposure, and the study was con- ducted with a well-described expert assessment of formaldehyde exposures clas- sified by self-reported jobs of cases and controls. The estimation of the probability of exposure level or intensity of exposure in each job enabled the investigators to estimate lifetime cumulative exposure of each participant. There was evidence of increasing disease frequency with increasing exposure. The odds ratio (OR) was 3.0 (95% CI 1.3–6.6) for the highest cumulative exposure category (˃1.10 ppm-year) compared with nonexposed, and there was a signifi- cant trend (p < 0.001) in the association between nasopharyngeal cancer and an increasing probability of exposure and duration. Controlling for wood-dust ex- posure and smoking had little effect on the association. The association appeared to be restricted to squamous-cell carcinoma rather than undifferentiated and nonkeratinizing carcinoma, although this finding is limited by small numbers. The evidence from the Vaughan et al. (2000) study is supported by several other studies. The National Cancer Institute (NCI) industrial cohort study of mortality is one of the important additional sources of evidence. The committee judged the study to be strong. Since the completion of NTP’s assessment of formaldehyde in 2011, the NCI cohort has been updated with 10 additional years of followup: NTP’s substance profile for formaldehyde cited Hauptmann et al. (2004), and the update of that study is Beane Freeman et al. (2013). The evi- dence from the cohort continues to suggest that formaldehyde exposure is asso- ciated with an increase in the frequency of nasopharyngeal cancer, although even with the additional followup the numbers of exposed cases are small. There were 10 total deaths from nasopharyngeal cancer (and five total deaths from sinonasal cancer, as discussed below). Although small numbers of cases for rare cancers can be a limitation, even for strong studies, because of the high quality of the quantitative, high-discrimination exposure assessment and the design and conduct of the study, the overall results were considered strong, informative, and continue to be persuasive. In the Beane Freeman et al. (2013) study, there was evidence of increasing mortality with increasing exposure for all three exposure metrics evaluated: average, cumulative, and peak exposure (see Appendix C for discussion of exposure metrics). Compared with low exposure, those in the highest categories of each of those metrics had rate ratios of 11.54 (95% CI 1.38–96.81), 2.94 (95% CI 0.65–13.28), and 7.66 (95% CI 0.94–62.34), respec- tively. A strength of this study is that there was very little wood-dust exposure (only one case was thought to have had such exposure), so there is little concern that the results were confounded by wood dust (a well-known risk factor for nasopharyngeal cancer). 

Independent Assessment of Formaldehyde 89 Several studies were judged to be moderately strong and provided support for the finding of increased nasopharyngeal-cancer risk (Vaughan et al. 1986a,b; West et al. 1993; Hildesheim et al. 2001; Siew et al. 2012). Vaughan et al. (1986a,b) conducted a small population-based case–control study of nasopha- ryngeal cancer incident cases (n = 27 total cases) that were drawn from 13 coun- ties in western Washington state. Interviews with cases (or next of kin if cases were deceased) and controls provided information on occupation (Vaughan et al. 1986a) and residence (Vaughan et al. 1986b) from which estimates of formalde- hyde exposure were developed. There was a weak association between working in a job with formaldehyde exposure and incidence of nasopharyngeal cancer (OR for 10 years or more of exposure compared with none was 1.6, 95% CI 0.4– 5.8). There was somewhat stronger evidence of an association between living in a mobile home (a well-documented source of formaldehyde exposure) and inci- dence of nasopharyngeal cancer (OR for 10 years or more of residence com- pared with none was 5.5, 95% CI 1.6–19.4) (Vaughan et al. 1986b). West et al. (1993) conducted a moderately large population-based case– control study of incident cases of nasopharyngeal cancer in the Philippines. The exposure assessment appeared to be a well conducted, semiquantitative assess- ment with moderate discriminations of exposure and was based on blind expert evaluation of the reported job histories. Several metrics of formaldehyde expo- sure, particularly in the distant past, were positively associated with nasopha- ryngeal-cancer incidence. The authors gathered data on several potential con- founders, including wood dust, smoking, and dietary factors. In a final model that controlled for confounders, the authors reported that subjects first exposed to formaldehyde 25 years or more prior to diagnosis had an OR of 4.0 compared with never exposed (95% CI 1.3–12.3). Control for smoking and “dust” expo- sure did not weaken the association. A somewhat larger population-based case–control study of incident cases with a semiquantitative exposure that had moderate discrimination was conduct- ed in Taiwan by Hildesheim et al. (2001). The exposure assessment was similar to that of West et al. (1993) in that an industrial hygienist reconstructed each subject’s occupational history. There was an increased incidence of nasopharyn- geal cancer in the longest duration-of-exposure category (OR = 1.60, 95% CI 0.91–2.90), and there was some evidence that the association was stronger in subjects who were seropositive for Epstein Barr virus (OR = 2.6, 95% CI 0.87– 7.7). Siew et al. (2012) used several Finnish national databases to evaluate as- sociations between incidence of sinonasal, nasopharyngeal, and lung cancers and exposures to wood dust and formaldehyde. Cases of those cancers were di- agnosed among Finnish men during 1971–1995, and were linked to census data on occupations. A job-exposure matrix was used to estimate wood-dust and formaldehyde exposures for subjects based on their occupations. There were only five nasopharyngeal cancer cases with any formaldehyde exposure and the relative risk (RR) for any formaldehyde exposure compared to no formaldehyde exposure was 0.87. There was a wide confidence interval (95% CI 0.34–2.20). 

90 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens An industrial cohort study of mortality by Meyers et al. (2013) was judged to be a strong study because it was well-designed with a high-discrimination, quantitative exposure assessment and it included Poisson regression modeling to control for confounding; however, it contributed little information to the evalua- tion of formaldehyde exposure and nasopharyngeal cancer in that it was not suf- ficiently large to detect an effect for rare cancers such as nasopharyngeal cancer. There was only a little more than one death expected from nasopharyngeal can- cer (n = 1.33), and none were observed. Several studies that were judged to be moderately strong also contributed little information to the evaluation of nasopharyngeal cancer in that they had a small number of subjects who had nasopharyngeal cancer and were exposed to formaldehyde: Walrath and Fraumeni (1983, 1984), Levine et al. (1984), Stroup et al. (1986), Andjelkovich et al. (1995), and Coggon et al. (2014). Walrath and Fraumeni (1983) reported on proportionate mortality in 1,132 deaths of em- balmers in New York. The authors reported that there were no deaths from can- cer of the nasopharynx. The authors conducted a similar study of licensed em- balmers in California (Walrath and Fraumeni 1984) and again observed no deaths from nasal or nasopharyngeal cancer. The study by Levine et al. (1984) of 1,477 Ontario undertakers with 319 deaths from all causes found one death from cancer of the buccal cavity and pharynx (2.1 expected, standardized mor- tality ratio [SMR] and CIs not given). The authors did not report whether that death was from nasopharyngeal cancer or a different neoplasm. Stroup et al. (1986) reported a retrospective cohort study of mortality in 2,317 male Ameri- can anatomists. All or nearly all worked with embalming fluid, which contains formaldehyde and other volatile chemicals. One death from buccal cavity and pharyngeal cancer was observed (6.8 deaths expected, SMR = 0.2, 95% CI 0.0– 0.8). The authors did not report whether that death was from nasopharyngeal cancer or a different neoplasm. Andjelkovich et al. (1995) evaluated mortality in a subset of automotive iron-foundry workers in Michigan. The original cohort was 8,147 men, and the subcohort exposed to formaldehyde, 3,929 men. There was one death from nasopharyngeal cancer in the exposed group (no SMR or 95% CI reported). Coggon et al. (2014), an update of the industrial cohort study of mortality by Coggon et al. (2003), reported only one death from nasopharyn- geal cancer. Several studies did not contribute to the committee’s assessment of for- maldehyde exposure and nasopharyngeal cancer, because the committee judged the studies to be weak and inconclusive (see Tables 3-1 and 3-2). Roush et al. (1987) conducted a population-based case–control study of incidence in 173 men drawn from the Connecticut Cancer Registry who had a history of nasopha- ryngeal cancer and had died. Occupation was determined from death certificates and city directories. The probable level of formaldehyde exposure was deter- mined from job title, industry, specific employment, and year of employment. For the seven deaths in the highest exposure category—probably exposed to some level of formaldehyde for most of their working life and probably exposed at a high level for 20 years or more prior to death—the OR was 2.3 (95% CI 

Independent Assessment of Formaldehyde 91 0.9–6.0; two-sided, p = 0.100), adjusted for age at death, year of death, and availability of occupational information. ORs were given for 14 specific indus- try categories; none was statistically significant, although numbers were small. Coexposures and residential exposures to formaldehyde were not addressed. Dell and Teta (1995) reported a long-term mortality study of an industrial cohort of workers in a single plastics manufacturing and research and development (R&D) plant in the United States. Of 5,932 male employees, 111 had job as- signments that involved formaldehyde. The number of deaths in this small group was not stated, but none was from nasopharyngeal cancer. Hansen and Olsen (1995) investigated cancer incidence in an industrial cohort of men who were employed at 265 companies in Denmark in which formaldehyde exposure was identified. The authors reported standardized proportionate incidence ratios (SPIRs) adjusted for age and calendar period; the comparison group was the Danish population as reported to the Danish Cancer Register. Four cancers of the nasopharynx were reported (3.2 expected, SPIR = 1.3, 05% CI 0.3–3.2). Other coexposures were not reported or adjusted for. Stellman et al. (1998), in an update of the industrial cohort mortality study of the American Cancer Socie- ty (ACS) Cancer Prevention Study–II, found one cancer of the nasopharynx in study participants who had an occupational history of exposure to wood dust (OR = 0.44, 95% CI 0.06–3.29) and one in men who had worked in a wood- related occupation (OR 1.44, 95% CI 0.19–10.9). Coexposures were not report- ed. Armstrong et al. (2000) conducted a large population-based case–control study of nasopharyngeal-cancer incidence (282 cases, all cases were squamous- cell carcinomas) in predominantly Chinese Malaysians. The exposure assess- ment was qualitative, and the study found no evidence of an association with formaldehyde exposure. Limitations in exposure assessment may contribute to an explanation of the low reported prevalence of formaldehyde exposure (for example, only eight cases reported more than 10 years of exposure and more than 10 years of latency), or formaldehyde exposure may simply have been rare and at low in concentration in the population. In either case, the uninformative finding of this limited study does not weaken the apparent association between formaldehyde exposure and nasopharyngeal cancer. Li et al. (2006) conducted a large industrial cohort study of nasopharyngeal cancer incidence in female tex- tile workers in China that included a low-discrimination, qualitative exposure assessment for formaldehyde (years for ever exposed vs never exposed). The authors noted that there was a potential for formaldehyde exposure to be mis- classified. The study had some potential to be informative, but the investigators found few workers who had formaldehyde exposures—10 noncases and no cas- es were identified as having formaldehyde exposure. In summary, the committee found that epidemiologic studies provided ev- idence of a causal association between formaldehyde exposure and nasopharyn- geal cancer in humans. Evidence of an association was derived from a strong population-based case–control study (Vaughan et al. 2000), a strong industrial cohort study (Beane Freeman et al. 2013), and several moderately strong popu- lation-based case–control studies (Vaughan et al. 1986a,b; West et al. 1993; 

92 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Hildesheim et al. 2001; Siew et al. 2012). See Table 3-3 for important key measures of association. The conclusion was based on the strength, consistency, temporality, dose–response relationship, and coherence of the evidence and on the considerations presented in Table 3-1.The most informative epidemiologic studies were ones that were large, that estimated exposure systematically, that had credible comparison groups, and that assessed cancer end points reliably. Not all studies that were judged as strong or moderately strong were informative in the evaluation of the evidence on nasopharyngeal cancer, because of the rarity of tumors at this site and because the studies reported only a few or no deaths from nasopharyngeal cancer. Other studies had sufficient cases but had weak exposure evaluations. The weakest and least informative studies had limited exposure assessments and few or no cases of nasopharyngeal cancer. Sinonasal Cancer The committee reviewed the literature on epidemiologic studies of for- maldehyde and sinonasal cancer (see Table 3-4). The strongest study was the pooled population-based case–control study by Luce et al. (2002) that assessed incidence data. It provided evidence of an association between formaldehyde exposure and sinonasal cancer. As mentioned in Chapter 2, a pooled study dif- fers from a meta-analysis in that the data from the studies are combined into a single dataset by using the same or similar case definitions and exposure as- sessments; this is analogous to what is done in a multisite cohort study. The Luce et al. study was particularly valuable because a new exposure assessment was conducted to inform each of the 12 studies that were assembled for the pooled analysis. The exposure assessment was quantitative and had high dis- criminatory ability; it estimated the level of exposure (average air concentration) and probability of exposure. The exposure data permitted the investigators to analyze risks among categories of cumulative exposure. There was strong evi- dence of an association between adenocarcinoma and formaldehyde exposure. For example, the OR for sinonasal-cancer incidence was 3.0 (95% CI 1.5–5.7) in men who were in the highest tertile of cumulative formaldehyde exposure com- pared with no exposure. The comparable OR in women was 6.2 (95% CI 2.0– 19.7). The association between formaldehyde and squamous-cell carcinoma was weaker and showed little evidence of a trend. The association between formal- dehyde and adenocarcinoma was investigated for possible confounding or effect modification by wood-dust exposure. The researchers used multiple logistic regressions, including analysis of the level of wood-dust exposure as a covariate and stratification on wood-dust exposure, to examine the association between formaldehyde exposure and adenocarcinoma in those who had no wood-dust (or leather-dust) exposure. The results showed only a modest weakening of the for- maldehyde risk. In women, the OR for high cumulative exposure fell from 6.2 to 

TABLE 3-4 Studies of Sinonasal Cancer and Formaldehyde Exposure Reference and Study Population No. SNC Cases in Exposed Findings (95% CI) Hayes et al. 1986 Histologically confirmed ICD-9 160, 160.2–160.5; OR for squamous-cell carcinoma cases comparing any vs no two raters (A and B) for exposure; number of cases formaldehyde exposure = 3.0 (90% CI 1.3–6.4) for rater A, General population in the Netherlands identified from Tables 3 and 4 of the publication 1.9 (90% CI 1.0–3.6) for rater B Any formaldehyde exposure, low wood-dust OR for squamous-cell carcinoma cases comparing high vs no exposure: rater A, n = 15; rater B, n = 24 formaldehyde exposure (with low wood-dust exposure) = 3.1 (90% CI 0.9–10.0) for rater A, 2.4 (90% CI 1.1–5.1) for rater B Squamous-cell carcinoma with any formaldehyde exposure, low wood-dust exposure: rater A, n = 12; Rater B assigned proportionally more controls to rater B, n = 19 formaldehyde exposure compared with rater A; rating from both raters showed an increase in OR with increasing formaldehyde assignments Luce et al. 1993 Cancer of nasal cavity and paranasal sinuses ICD-9 OR for adenocarcinoma from possible, probable, or definite 160.0, 160.2–160.9; number of cases identified from formaldehyde exposure and no or low wood-dust exposure = General population in France Table 2 of the publication 8.1 (0.9–72.9) Adenocarcinoma with probable or definite exposure (male and female): n = 70 Squamous-cell carcinoma with probable or definite exposure (male and female): n = 18 Luce et al. 2002 Number of cases identified from Table 3 of the OR for adenocarcinoma (adjusted for age and wood- and publication leather-dust exposure) from high formaldehyde exposure, General populations of 7 countries male = 3.0 (1.5–5.7); female = 6.2 (2.0–19.7) Adenocarcinoma cases with medium or high exposure: n = 122 male; 5 female OR for adenocarcinoma from high formaldehyde exposure and no wood- or leather-dust exposure, male = 1.9 (0.5–6.7); Squamous-cell carcinoma cases with medium or high female = 11.1 (3.2–38.0) exposure: n = 70 male; 13 female OR for squamous carcinoma from high formaldehyde exposure, male = 1.2 (0.8–1.8) (Continued) 93 

94 TABLE 3-4 Continued Reference and Study Population No. SNC Cases in Exposed Findings (95% CI) Olsen and Asnaes 1986 Histologically confirmed ICD-7 160.0, 160.2–160.9; Ever vs never exposed to formaldehyde, pooled estimate for number of cases identified from Table 4 of the formaldehyde exposure adjusted for wood-dust exposure: General population in Denmark publication; most formaldehyde exposures occurred - Squamous-cell carcinoma of the nasal cavity and in Danish wood-working industry and few sinuses: OR = 2.3 (95% CI 0.9–5.8) formaldehyde cases not exposed to wood dust - Adenocarcinoma of nasal cavity and sinuses: OR = 2.2 (95% CI 0.7–7.2) Ever vs never exposed to formaldehyde: - Squamous-cell carcinoma: n = 13 ≥10 years since first exposure, pooled estimate for - Adenocarcinoma: n = 17 formaldehyde exposure adjusted for wood-dust exposure: - Squamous-cell carcinoma of the nasal cavity and sinuses: OR = 2.4 (0.8–7.4) - Adenocarcinoma of nasal cavity and sinuses: OR = 1.8 (0.5–6.0) Siew et al. 2012 Nasal cancer; number of cases identified from Table RR (adjusted for wood dust) for any formaldehyde exposure 3 in the publication compared with no formaldehyde exposure = 1.11 (0.66–1.87) Finnish general population Any exposure to formaldehyde: n = 17 Vaughan et al. 1986a,b SNC defined by ICD 160: number of cases identified OR (adjusted for age, sex, smoking, and alcohol) for number from Tables 3 and 5 in Vaughan (1986a) and Table 2 of years exposed: 1–9 years = 0.7 (0.3–1.4); ≥10 years = 0.4 General population in western in Vaughan (1986b) (0.1–1.9) Washington state Exposed to formaldehyde above background, n = 12 OR (adjusted for age, sex, smoking, and alcohol) for cumulative exposure score (all years): 5-19 = 0.5 (0.1–1.6); ≥20 years = 0.3 (0.0–2.3) OR (adjusted for age, sex, smoking, and alcohol) for cumulative exposure score (15-year lag period): 5-19 = 1.0 (0.3–2.9) Abbreviations: ICD, International Classification of Diseases; OR, odds ratio; ppm, parts per million; RR, relative risk; SNC, sinonasal cancer. Source: Committee generated.   

Independent Assessment of Formaldehyde 95 5.8 (95% CI 1.7–19.4), and males showed a similar reduction. A number of oth- er studies that were judged to be moderately strong contributed to the conclusion that this study was not anomalous. The two key strengths of the Luce et al. (2002) study are the great size and the high-quality exposure assessment; the other studies were smaller and had less adequate exposure assessments. All of the studies have their own limitations, but taken as a whole they provide corrob- orating evidence. The moderately strong studies identified by the committee that supported an association between exposure to formaldehyde and sinonasal cancer were Hayes et al. (1986), Olsen and Asnaes (1986), Vaughan et al. (1986 a,b), Luce et al. (1993), and Siew et al. (2012). Hayes et al. (1986) conducted a population- based case–control study of the incidence of histologically confirmed cases of sinonasal cancer in the Netherlands from 1978 to 1981. A low-discrimination, qualitative exposure assessment was conducted independently by two trained hygienists (rater A and rater B) who classified all jobs as to the level (intensity) and probability of formaldehyde (and wood-dust) exposure. The study was large enough to permit separate assessment of risks specifically for cases of squa- mous-cell carcinoma (there were at least 12 cases with formaldehyde exposure). For all sinonasal cancer combined, the OR was approximately doubled when the exposed were compared with the nonexposed; the CIs excluded 1.0. The authors stratified their analysis by wood-dust exposure (none and low vs high) and found that there were trends of increasing incidence with increasing level of formaldehyde exposure in the no or low wood-dust stratum. That pattern was more evident for squamous-cell carcinomas (there were not enough adenocarci- nomas in the group with low wood-dust exposure to permit this analysis). The OR was 3.1 (90% CI 0.9–10.0) for high formaldehyde exposure and low or no wood-dust exposure vs no formaldehyde exposure for rater A and 2.4 (90% CI 1.1–5.1) in the same category for rater B. Rater B assigned proportionally more controls to formaldehyde exposure compared with rater A. The rating from both raters showed an increase in OR with increasing formaldehyde exposure. Olsen and Asnaes (1986) was an update of Olsen et al. (1984). In the 1986 study, the authors conducted a population-based case–control study of incidence nested in the Danish cancer registry, and they included cancer controls. Den- mark has a large wood-working industry, which also includes some formalde- hyde exposures. As a result, few cases have formaldehyde exposure without wood-dust exposure. The study had a limited exposure assessment that was based on expert evaluation of job information. The exposure assessment was qualitative and was of moderate discrimination in its assessment in determining whether each subject had certainly or probably been exposed to formaldehyde. The authors in- vestigated separately the association between formaldehyde exposure and inci- dence of the two main histologic types of nasal and paranasal sinus cancer— squamous-cell carcinoma and adenocarcinoma. When the ever exposed to formal- dehyde were compared with the never exposed to formaldehyde, the ORs were very similar for the two subtypes; 2.3 (95% CI 0.9–5.8) for squamous-cell carci- noma and 2.2 (95% CI 0.7–7.2) for adenocarcinoma. Although limited by small 

96 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens numbers, there was evidence of increased incidence of adenocarcinoma from for- maldehyde exposure in subjects who were not exposed to wood dust (OR = 7.0, 95% CI 1.1–43.9). When the data were examined for 10 or more years since first exposure, the OR for squamous-cell carcinoma was 2.4 (95% CI 0.8–7.4) and the OR for adenocarcinoma was 1.8 (95% CI 0.5–6.0). Vaughan et al (1986a) undertook a population-based case–control study in Washington state of 53 incident cases of sinonasal cancer, including 12 in peo- ple thought to have had occupational exposure to formaldehyde. The authors found no evidence of increased risk with maximum exposure, number of years exposed, a cumulative exposure score, or the cumulative exposure score with a 15-year lag period. Vaughan et al. (1986b) used the same study group as Vaughan et al. (1986a) to examine the role of residential exposures and sinona- sal cancer. Evaluations were reported for people exposed in mobile homes (5 cases, OR = 0.6, 95% CI 0.2–1.7), people living for not more than 10 years in new or renovated housing with particle board or plywood (13 cases, OR = 1.8, 95% CI 0.9–3.8), and people living for 10 years or more in new or renovated housing with particle board or plywood (12 cases, OR = 1.5, 95% CI 0.7–3.2). The authors did not investigate coexposures except for lifetime smoking history and recent consumption of alcoholic beverages. Luce et al. (1993) conducted a large population-based case–control study (207 cases and 409 controls) of the incidence of sinonasal cancer in France. His- tologic data allowed separate investigations of adenocarcinoma and squamous- cell carcinoma. The exposure assessment was semiquantiative with moderate discrimination in that it was based on expert judgment without measurement data for assessment of jobs (which were classified by probability of exposure) and expert assessment of exposure frequency and intensity. The investigators started with a large case series: there were 38 adenocarcinoma cases that had more than 30 years of exposure to formaldehyde. The squamous-cell carcinoma series was somewhat smaller—five in the longest duration category. The study was limited in its ability to discriminate risks associated with potentially con- founded wood-dust and formaldehyde exposure, and nearly all cases that had formaldehyde exposure also had probable or definite wood-dust exposure; only four adenocarcinoma cases that had possible, probable, or definite formaldehyde exposure were believed to have had no or low wood-dust exposure (OR = 8.1, 95% CI 0.9–72.9). The authors also reported that the combination of wood dust plus formaldehyde exposure was associated with a higher risk of adenocarcino- ma than wood dust alone, although confidence intervals were wide because of the small number of cases. Siew et al. (2012), the cohort of Finnish men from a national database, was summarized above in the section on nasopharyngeal cancers. There were 17 cases of cancer of the nose and paranasal sinuses in Finnish men identified as having any occupational exposure to formaldehyde. There was a weak associa- tion of cancer in those who had any exposure to formaldehyde compared to no exposure to formaldehyde (RR = 1.11, 95% CI 0.66–1.87). 

Independent Assessment of Formaldehyde 97 The recently updated NCI industrial cohort study of mortality was judged to be strong, but the number of sinonasal-cancer cases was small (Beane Free- man et al. 2013). There were five deaths from sinonasal cancer in this large co- hort (three deaths in the exposed population compared to 3.3 expected deaths). There was no evidence of increased mortality from this cancer, but because of the small numbers of expected deaths from sinonasal cancer, little weight was given to these findings. Meyers et al. (2013), an update of Pinkerton et al. (2004), was also judged to be a strong industrial cohort study of mortality, but it contributed little infor- mation because of its size; there were only 0.95 cases of sinonasal cancer ex- pected and none were observed. The authors investigated mortality in 11,043 workers in three garment plants (Meyers et al. 2013). There were no deaths from sinonasal cancer among in 3,915 deaths reported. Additional details were not provided. Several studies were judged to be moderately strong, but they contributed little information to the evaluation of sinonasal cancer because few subjects who had sinonasal cancer had been exposed to formaldehyde: Walrath and Fraumeni (1983, 1984), Levine et al. (1984), Stroup et al. (1986), and Coggon et al. (2014). The studies by Walrath and Fraumeni (1983, 1984) were described in the nasopharyngeal-cancer section above; the results of the two studies were not informative for evaluating sinonasal cancer, because no cases were reported. The study by Levine et al. (1984) of a cohort of 1,477 Ontario undertakers with 319 deaths from all causes found no deaths from cancer of the nose, middle ear, or sinuses (0.2 deaths expected, SMR and CIs not given). Stroup et al. (1986) reported a retrospective cohort study of mortality in 2,317 male American anat- omists. All or nearly all worked with embalming fluid, which contains formal- dehyde and other volatile chemicals. None of the 738 deaths was from cancer of the nasal cavity or sinuses (0.5 deaths expected, SMR = 0, 95% CI 0.0–7.2). Coggon et al. (2014) completed a long-term study of mortality in a cohort of 14,014 men in six British plants where formaldehyde was produced or used. In the group of workers whose jobs that were classified as having potential formal- dehyde exposure, there were two deaths from cancer of the nose and nasal si- nuses (2.8 deaths expected from US national rates, SMR = 0.71, 95% CI 0.09– 2.55). Coexposures were not discussed. Several studies did not contribute to the committee’s assessment of for- maldehyde exposure and sinonasal cancer, because the committee judged the studies to be weak and inconclusive (see Tables 3-1 and 3-2). Roush et al. (1987) conducted a population-based case–control study of incident cases in 198 men in the Connecticut Cancer Registry who had a history of sinonasal cancer and died. Occupation was determined from death certificates and city directo- ries. Probable level of formaldehyde exposure was determined from job title, industry, specific employment, and year of employment. The OR for the seven deaths in the highest exposure category was 1.5 (95% CI 0.6–3.9) (adjusted for age at death, year of death, and availability of occupational information). ORs were given for 14 specific industry categories, and none was statistically signifi- 

98 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens cant, but the numbers were small. Coexposures and residential exposures to formaldehyde were not addressed. Dell and Teta (1995) reported a long-term study of mortality in a cohort of industrial workers in a single plastics manufacturing and R&D plant in the Unit- ed States. Of 5,932 male employees, 111 had job assignments that involved formaldehyde. The number of deaths in this small group was not stated, but none was from sinonasal cancer. Hansen and Olsen (1995, 1996) conducted a study in a large national can- cer cohort of industrial workers and reported SPIRs. The authors obtained gov- ernment employment data on blue-collar workers employed in Danish industries who were identified as having used formaldehyde and linked those data with cancer-registry data. A national product register was used to identify workers in broad industries in which formaldehyde was used and formaldehyde exposure was likely. The records were used to determine a moderate-discrimination, sem- iquantitative metric of formaldehyde exposure: duration of work with potential formaldehyde exposure. A similar approach was used to determine wood-dust exposure at the industry level by identifying industrial classification codes that corresponded with jobs that used wood products. Only 13 cases of cancer of the nasal cavity were reported to the national cancer registry (compared with 5.2 deaths expected on the basis of the proportionate distribution of all cancers combined) in men whose longest job was in a company that used formaldehyde. The investigators calculated an SPIR as an estimate of the rate ratio; for nasal cancer, the SPIR was 2.3 (95% CI 1.3–4.0). When the data were limited to blue- collar workers in formaldehyde-using industries in which wood products were not used, the SPIR increased to 3.0 (95% CI 1.4–5.7). Stellman et al. (1998), in an update of the industrial cohort mortality study of the ACS Cancer Prevention Study-II, found one death from sinonasal cancer in men who had wood-dust exposure and found no evidence of an association with formaldehyde. Stern (2003) completed a study of mortality in an industrial cohort of 9,352 tannery workers in jobs that often included formaldehyde expo- sure; one death from cancer of the nasal cavity was reported (SMR not given). Pesch et al. (2008) conducted an industry-based case–control study of incident cases of adenocarcinoma of the nasal cavity and paranasal sinuses in the German wood industry (86 male cases, 204 controls). In the group of workers who were exposed to formaldehyde and wood products, eight cases were exposed to for- maldehyde before 1985 (OR = 0.46, 95% CI 0.14–1.54), and 39 cases were ex- posed to formaldehyde in 1985 or later (OR = 0.94, 95% CI 0.47–1.90). Because both cases and controls were exposed to wood dust, a recognized cause of si- nonasal cancer, extension to the general population is uncertain. The committee found that epidemiologic studies provided evidence of a causal association between formaldehyde and sinonasal cancer in humans. Evi- dence of an association was derived from the strong pooled case–control studies of sinonasal cancer (Luce et al. 2002) and several moderately strong population- based case–control studies (Hayes et al. 1986; Olsen and Asnaes 1986; Vaughan et al. 1986a.b; Luce et al. 1993; Siew et al. 2012). See Table 3-4 for important 

Independent Assessment of Formaldehyde 99 key measures of association. The conclusion was based on the strength, con- sistency, temporality, dose–response relationship, and coherence of the evidence and on the considerations presented in Table 3-1.The most informative epidemi- ologic studies were the ones that were large, that estimated exposure systemati- cally, that had credible comparison groups, and that assessed cancer end points reliably. The studies that did not find associations were usually too small to de- tect an effect for these rare cancers or used methods of exposure assessment that had little ability to discriminate exposures, and they did not provide convincing evidence that there were sufficient numbers of highly exposed subjects. Lymphohematopoietic Cancers The committee reviewed the literature on a potential association between formaldehyde exposure and lymphohematopoietic cancers. This section begins with a discussion of methodologic considerations in exposure assessment in studies of lymphohematopoietic cancers and then discusses in greater detail studies in industrial cohorts and studies in embalmers and others in the funeral trade, anatomists, and pathologists. Data from studies that the committee judged to be strong and moderately strong and informative are presented in Tables 3-5 (industrial workers), 3-6 (funeral workers, embalmers, pathologists, and anato- mists), and 3-7 (general population). Methodologic Considerations in Exposure Assessment in Studies of Lymphohematopoietic Cancers In the substance profile for formaldehyde, NTP considered the most in- formative primary studies for the evaluation of lymphohematopoietic cancers to be the study of mortality in the large NCI cohort of formaldehyde-industry workers (Beane Freeman et al. 2009) and the NCI nested case–control mortality study of embalmers and funeral directors, which was based on a cohort of funer- al-industry workers (Hauptmann et al. 2009). Those were judged to be the strongest studies because of the high quality of the quantitative exposure as- sessments, which included assignments of participants into exposure categories with high discrimination. When large occupational cohorts are used to study relatively rare cancer, subpopulations are drawn from several worksites of varying size to obtain suffi- cient cases. Although the worksites have exposure to formaldehyde as a com- mon feature, they can have large differences in exposure conditions even if the job titles and types of operations are the same (see Appendix C for a more de- tailed discussion). Beane Freeman et al. (2009) conducted a comprehensive ex- posure assessment, which increases confidence that valid exposure–response trends can be derived from the diverse industries and exposure conditions. Both the formaldehyde-industry (Beane Freeman et al. 2009) and funeral- industry (Hauptmann et al. 2009) cohorts included extensive separate evalua- 

100 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens tions of occupational exposures, their determinants, and modeling approaches to reconstructing unmeasured historical exposures.2 The exposure studies of the formaldehyde-industry cohort were reported by Blair et al. (1986, 1990) and Hauptmann et al. (2004). The exposure studies of the funeral-industry cohort were reported by Stewart et al. (1992). The committee recognized that those additional exposure studies were keys to the strength of the epidemiologic stud- ies. Because Beane Freeman et al. (2009) and Hauptmann et al. (2009) were critical for the formaldehyde assessment of lymphohematopoietic cancers, this section elaborates on their approaches. The exposure assessments for the formaldehyde-industry and funeral- industry cohorts were designed to determine exposures associated with job titles and worksites listed in the work histories of the study subjects so that exposures and subjects could be linked. Historical changes in job activities and in the for- maldehyde industry produced substantial differences in temporal profiles of ex- posure. Industrial exposures have declined considerably since the early 1970s as a result of process changes and engineering controls of process emissions. The exposures in the Beane Freeman et al. (2009) study changed (more in some jobs than in others), and the data suggest that exposures in the 1960s were much higher than those after 1970 (Blair et al. 1986, 1990). Embalming-fluid emis- sions of formaldehyde have probably changed little, but local exhaust ventilation was added in some funeral homes and was estimated to have reduced exposure by 50–90% (Stewart et al. 1992). Exposures in the industrial and embalming settings were described by time-weighted averages (TWAs) and short-term measurements. The short-term measurements were used to capture brief (15 minutes) intense exposures called peaks. Although peaks are part of the distribution of short-duration concentra- tions that contribute to the longer TWA measurements, they might not correlate well with the overall average (Blair and Stewart 1990), as was seen in the Beane Freeman et al. study (2009). Blair and Stewart (1990) also noted that exposure metrics can differ among manufacturing plants because in some plants everyone is exposed but in others only half the workforce is in areas with exposure or be- cause similar work areas had lower exposures. As explained in Appendix C, the summary measures of exposure (which are also called exposure metrics or dose metrics) used in epidemiologic studies are weighting schemes applied to summarize the complex temporal profiles of personal exposure histories. In that application, they are analogous to the con- cept of dose applied in toxicologic studies, but there is no universal dose metric that applies to all toxic responses, including carcinogenesis. Some dose metrics are not appropriate for the underlying biology, and when an inappropriate metric is used, a weaker or no dose–response relationship will usually be observed 2 Appendix C provides a general summary of exposure assessments, the rationale for estimating exposures on the basis of physical principles, and a description of methods for measuring airborne formaldehyde exposures. 

Independent Assessment of Formaldehyde 101 (Blair and Stewart 1990; Smith and Kriebel 2010). Although cumulative expo- sure is the most common dose metric for chronic, minimally reversible disease processes, it is probably not the optimal dose measure for studying cancer (Smith and Kriebel 2010), as noted above. A fundamental feature of cumulative exposure is that it gives equal weight to long, low-intensity exposures and short, high-intensity exposures, which may not be biologically appropriate for cancer biology. A lag time until effects are observed may also be included in the expo- sure metric to account for an induction period between the first exposure to for- maldehyde and the diagnosis of cancer. That period includes any delay from first exposure to the exposure that initiated the cancer, the time from initiation through the biologic events that led to malignant change, and the time required for that change to produce signs or symptoms that result in diagnosis. Those steps are commonly thought to require at least 10 years for solid cancers in adults, perhaps less for leukemia and lymphomas. Epidemiologic models that use exposure metrics for peak exposures hy- pothesize an underlying nonlinear damage process in which exposures at low concentrations have little or no effect and exposures at high concentrations pro- duce disproportionate effects. That might indicate a threshold process, or some protective process might be overwhelmed or a damaging secondary process might occur. When the mode of action is unknown, it is common for epidemiol- ogists to try several exposure metrics, such as cumulative exposure and peak exposure that have different biologic implications (Blair and Stewart 1990). The mechanistic process associated with the cumulative exposure and peak exposure metrics appear to be different, and conceptually the metrics should be useful for obtaining insight about the possible mechanism of the ef- fects. Unfortunately, the precision of estimated metric values is often limited by sparse historic data and the cost of making measurements, variation of exposure between subjects, process and material variation in the industrial operations, and business and economic variations in the demand for a product. If the precision is too limited, it may not be possible to determine which metric is the strongest. Data quality and extrapolation approaches may favor one dose metric over an- other. Thus, as discussed above, it is common for epidemiologists to calculate several different exposure metrics, such as cumulative exposure, average expo- sure, and the occurrence or frequency of peaks. When data and resources are limited, epidemiologists often use simpler metrics, such as years of work in a job, categories of ever exposed vs never exposed on the basis of job title or work location, or sometimes even ‘ever having worked in an exposed industry’. In addition to the NCI formaldehyde-industry study (Beane Freeman et al. 2009) and the NCI nested case–control study (Hauptmann et al. 2009), Meyers et al. (2013), an update of Pinkerton et al. (2004), was considered to have strong methods (Table 3-2). The study investigated mortality in an industrial cohort of garment workers. The authors relied on earlier studies of the same sites by Stayner et al. (1985, 1988), Acheson et al. (1984), and Gardner et al. (1993). 

102 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Semiquantitative exposure estimates were developed on the basis of small num- bers of measurements, job activities, and reports of sensory irritation in jobs or work locations. There were also several moderately strong studies of limited utility in in- dustrial workers (Bertazzi et al. 1989; Partanen et al. 1993; Andjelkovich et al. 1995; Coggon et al. 2014) and embalmers, anatomists, or pathologists (Walrath and Fraumeni 1983, 1984; Levine et al. 1984; Stroup et al. 1986). Those had smaller populations and less discriminating exposure assessments and as a result contributed less to the evidence of an association between formaldehyde and lymphohematopoietic cancers than did the strong studies. Most of the smaller studies used job information alone to define those who were “exposed”—an approach that has little ability to discriminate among people with varied levels of exposures. Duration of exposure obtained from occupational histories was used as a semiquantitative exposure metric, but again, duration alone does not discriminate among exposures that have different intensities. Population-based case–control studies have the most serious problem of exposure misclassification because they draw from the broad mixture of person- al and industrial activities throughout the population in a wide area. For exam- ple, the broad job categories of “mortician” and “undertaker” include embalmers (the most highly exposed) but also include a number of less exposed occupa- tions. People in some of those other occupations may occasionally do embalm- ing, but less frequently, and embalming is not one of their main job activities. The categories also include funeral directors, who usually do not embalm. And differences are related to the size of funeral homes’ businesses. Use of narrow, well defined, specific job titles, such as a focus on embalmers, can greatly re- duce misclassification even without specific measurements. Studies of Industrial Cohorts Exposed to Formaldehyde Table 3-5 provides the studies of industrial cohorts exposed to formaldehyde that the committee judged to be strong or moderately strong. As already stated, the NCI industrial-worker cohort mortality study is large, well conducted, and in- formed by a quantitative, high-discrimination exposure assessment (Beane Free- man et al. 2009). The investigators collected mortality data on workers employed in US chemical factories that used formaldehyde during 1966–2004. The study was the largest in terms of numbers of exposed cancer cases—there were 286 he- matologic-malignancy cases, including 116 leukemia cases, and 44 of the leuke- mia cases were classified as myeloid leukemia. Exposure levels varied widely over time and among plants; the estimated overall median daily exposure was 0.3 ppm. The manufacturing plants produced a various of products, including formaldehyde (plants 2, 7, and 10), formaldehyde resins and molding compounds (plants 1, 2, and 7–10), molded plastic products (plants 8 and 9), photographic film (plants 4 and 5), decorative laminates (plant 6), and plywood (plant 3) (Blair et al. 1990). 

TABLE 3-5 Lymphohematopoietic Cancers: Industrial Workers No. Cancer Cases in Exposed Findings (95% CI) All All Lymphohematopoi Myeloid Lymphohematopoietic Reference and Study Population etic Cancer Leukemia Leukemia Cancer Leukemia Myeloid Leukemias Andjelkovich et al. 1995 7 2 — SMR = 0.59 (0.23–1.21) SMR = 0.43 — (0.05–1.57) US iron-foundry workers (Number of cases from Table 3 of the publication) Beane Freeman et al. 2009 286 116 44 peak >4 ppm: RR = 1.37 peak >4 ppm: peak >4 ppm: RR = (1.03–1.81), trend with RR = 1.42 (0.92– 1.78 (0.87–3.64) NCI study in US chemical increasing peak 2.18), trend with workers exposure increasing peak highest peak category exposure before 1994: RR = (Number of cases from Table 1 2.79 (1.08–7.21), p of the publication) trend = 0.02 Bertazzi et al. 1989 7 — — SMR = 7/3.9 = 1.8 — — (0.72–3.70) Italian resin workers (Number of cases from Table 3 of the publication) Coggon et al. 2014 — 18 9 — high exposure high exposure: ≥1 year: OR = 0.59 OR = 1.26 (0.39– UK chemical workers (0.23–1.50) 4.08) (Number of cases from Table 6 of the publication) (Continued) 103 

104 TABLE 3-5 Continued No. Cancer Cases in Exposed Findings (95% CI) All All Lymphohematopoi Myeloid Lymphohematopoietic Reference and Study Population etic Cancer Leukemia Leukemia Cancer Leukemia Myeloid Leukemias Meyers et al. 2013 107 36 21 SMR = 1.11 (0.91–1.34) ≥10 years of ≥10 years of exposure and ≥20 exposure and ≥20 Update of Pinkerton et al. (2004) years since first years since first exposure: SMR = exposure: SMR = US garment workers 1.74 (1.10–2.60) 1.90 (0.91–3.50) (Number of cases from Table 2 16–19 years exposure of the publication) vs none: SRR = 6.42 (1.40–32.30); test for trend with increasing duration: p = 0.01 Partanen et al. 1993 7 2 — OR = 2.49 (0.81–7.59) OR = 1.40 (0.25– — 7.91) Finnish wood-industry workers (Number of cases from Tables 1 and 3 of the publication) Abbreviations: CI, confidence interval; OR, odds ratio; RR, relative risk; SMR, standardized mortality ratio; SRR, standardized rate ratio. Source: Committee generated. 

Independent Assessment of Formaldehyde 105 That complexity might have introduced problems of noncomparability among the plants, but a thorough reconstruction of historical formaldehyde average and peak exposures was conducted consistently for all sites until 1980. Good-quality histor- ical data on potential confounders were also assembled from plant records and interviews of long-term employees. Because it pooled data from many plants, the study was powerful enough to detect effects that would not be measurable in plant-by-plant analyses. The formaldehyde exposure assessment was conducted only for jobs held until 1980. Thus, there is likely to have been more error in the exposure assignments in the later time period; in the primary analyses, exposure after 1980 was assumed to be zero. Two sensitivity analyses were conducted to evaluate the effect of that assumption on the results. About one-fourth of the NCI industrial-worker cohort was estimated to have experienced peak exposures of at least 4.0 ppm (Beane Freeman et al. 2009). A 1999 Agency for Toxic Substances and Disease Registry literature review found that the threshold for mild to moderate human eye, nose, and throat irritation by formaldehyde ranged from 0.4 to 3 ppm in 17 laboratory studies (ATSDR 1999). Thus, the highest peak exposure category (greater than 4 ppm) was above the irritation threshold, and at this level about 50–100% of sub- jects would have experienced an irritation response. There was evidence of increased risk of myeloid leukemia with increasing formaldehyde exposure (Beane Freeman et al. 2009). The evidence was strong- est when the peak-exposure metric was used, weaker when average exposure was used, and very weak when the effect of cumulative exposure was assessed. In the primary analysis (which assumed zero exposure for all jobs after 1980), the RR of myeloid leukemia increased with increasing exposure. Compared with those who had peak exposures less than 2.0 ppm, the RR in those who had peak exposures from 2.0–4.0 ppm was 1.30 (95% CI 0.58–2.92) and in those who had peak exposures of at least 4.0 ppm, 1.78 (95% CI 0.87–3.64). The data also show the expected pattern wherein the RRs for the highest peak category com- pared with the lowest peak category increased as the tumor category was nar- rowed—the RR of all lymphohematopoietic cancers was less than that of all leukemias grouped, and the RR of all leukemias grouped was less than that of myeloid leukemias grouped. The associations were weaker when average expo- sure was used as the summary measure of exposure than when peak exposure was used, but the trends were similar. A modest increase in RRs was observed among categories of increasing average exposure. The RR increased from the group of all lymphohematopoietic cancers to the grouping of all leukemias, and the RR increased further from the grouping of all leukemias to the grouping of myeloid leukemia. Beane Freeman et al. (2009) investigated the sensitivity of their results to the assumption of zero exposure after 1980 by censoring all persons who were still exposed in 1979 (this resulted in a loss of about 5% of the person–time of followup). The resulting effect estimates were stronger for both peak and aver- age exposure metrics. For example, the RR for the highest peak exposure cate- gory increased from 1.79 (cited above) to 2.64 (95% CI 1.12–6.20), and the 

106 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens trend among categories was also stronger (p = 0.03). The authors reported that there were stronger associations with exposures in the distant past, which may be explained either by higher air concentrations or by a relatively short latency for formaldehyde-induced leukemia. There was evidence to support the former explanation; exposures in the plants were much higher before 1970 than in later years when exposure controls were instituted (Stewart et al. 1986). The possibil- ity of a relatively short latency (compared with that of solid tumors) is supported by two studies of the association between benzene and leukemia (Silver et al. 2002; Glass et al. 2004). In both cohorts, the RR of leukemia after benzene ex- posure decreased with increasing follow up, and the authors proposed that this is likely due to a relatively short latency for the effects caused by benzene. Beane Freeman et al. (2009) reported that for the period up to 1994, the RR for the highest peak-exposure category compared with the lowest was 2.79 (95% CI 1.08–7.21), and there was evidence of an increasing trend among cate- gories (p = 0.02). It is not clear why Beane Freeman et al. (2009) found an asso- ciation with peak exposure and not with cumulative exposure. The committee noted that there were only 10 cases of myeloid leukemia in the highest cumula- tive exposure category, which was defined as at least 5.5 ppm-years. That is not very many cases and not a very high level of exposure. As a result, this finding is not strong evidence against an association between formaldehyde and myeloid leukemia. As noted earlier in this chapter, the alternative exposure metrics of peak, average, and cumulative exposure are expected to be proportional to the inci- dence of a disease as related to different biologic mechanisms or pathways. A complicating factor that must also be considered is the effect of exposure as- sessment errors on the resulting summary measures. However, it cannot be pre- dicted with any confidence which exposure metric would be expected to be closer to the “truth” in the investigation of formaldehyde and cancer. Therefore, the committee assessed peak, average, and cumulative exposure with equal weight on its overall evaluation. More precise studies in the future may be able to resolve this issue. Hodgkin lymphoma was strongly associated with peak exposure (RR = 3.96, 95% CI 1.31–12.02) when the subgroups with the highest and lowest peak exposure were compared. A positive association with multiple myeloma was also observed when the highest and lowest peak-exposure subgroups were com- pared (RR = 2.04, 95% 1.01–4.12). For both outcomes, there was evidence of a trend of increasing mortality with increasing peak exposure. The findings on Hodgkin lymphoma and multiple myeloma are potentially important for further investigation, but the committee did not find additional evidence of these associ- ations in other studies. An important strength of the NCI industrial-cohort study was its ability to investigate possible confounding by other chemical exposures (antioxidants, asbestos, benzene, carbon black, dyes and pigments, hexamethylenetetramine, melamine, phenol, plasticizers, urea, and wood dust); none was found. Beane Freeman et al. (2009) specifically investigated a potential confounding effect of 

Independent Assessment of Formaldehyde 107 benzene by excluding all workers who were known to have been exposed to benzene, and the results were not changed. Plant heterogeneity was investigated and found not to be an important factor in the results. There were some limita- tions. Despite the size of the study, the numbers of deaths in some categories of rare neoplasms were still small, and this limited the power to detect associations in the smallest subgroups. The magnitude of the exposure–response associations changed over time, and it is not possible without strong a priori assumptions to distinguish alternative explanations, such as disease latency, changes in expo- sures associated with changes in industrial operations and engineering controls, or time-dependent measurement uncertainties. The committee concluded that although those limitations exist, the study was of high quality. The careful and clearly documented design and analysis reduced the likelihood that the results could be explained by bias. As noted, the authors investigated important sources of confounding and found no important evidence of confounding that might seriously undermine their results. Chance is an unlikely explanation given the consistent patterns of increased RR among exposure categories and tumor categories noted above. Thus, the committee determined that the findings are relevant to evaluating an association between formaldehyde exposure and myeloid leukemia. Additional evidence of an association between formaldehyde exposure and lymphohematopoietic cancers in workers who were exposed during industrial operations was found in the National Institute for Occupational Safety and Health (NIOSH) study of garment workers. Meyers et al. (2013) updated earlier reports by Stayner et al. (1988) and Pinkerton et al. (2004) on mortality in a co- hort of 11,043 industrial workers who were exposed to formaldehyde in three garment-manufacturing plants. The cohort was considerably smaller than the NCI formaldehyde-industry cohort (21 myeloid-leukemia deaths compared with 44 in the NCI cohort). The study methods included a high-discrimination, quan- titative exposure assessment for current exposures that was performed during the early 1980s, which was an important strength of the study, but it did not cover the full period of exposures. The investigators did not attempt to estimate earlier exposures. The only known source of formaldehyde exposure was off-gassing from treated fabrics (which were produced elsewhere), so the amount of free formaldehyde in the fabric was a primary determinant of the workroom expo- sure (Elliot et al. 1987). Before 1970, the free-formaldehyde content of the fab- ric was estimated to be over 4,000 ppm; by 1980, the fabric concentrations had been reduced to 100–200 ppm. The air concentration measured in the work- rooms in 1984 (geometric mean exposure, 0.15 ppm) was a result of off-gassing of the 100–200 ppm in the fabric. The ratio of fabric content to air content was about 1,000:1. Assuming that the ratio is fairly constant, fabric that contained 4,000 ppm probably produced an air concentration of about 4 ppm before 1970. However, the investigators did not make use of that simple estimate of earlier exposure; they merely noted that air exposure was likely to have been higher before 1970. Goldstein (1973) reported that industry efforts to reduce formalde- hyde levels in work rooms by reducing the amount of resin in the fabric resulted 

108 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens in a decreased from 10 ppm in 1968 to 2 ppm in 1973. Formaldehyde air con- centrations were found to be similar between plants and across departments within the same plant. TWA concentrations were reported in a fairly narrow range (0.09–0.20 ppm), and there was little evidence that short-term peaks ex- ceeded the mean. Given the relatively homogenous exposure scenario, it was reasonable to use all employed workers as the exposed group and to compare their mortality with that in the general population. They used years of work from the workers’ company job histories to approximate cumulative exposure and implicitly assumed that each year had roughly the same intensity of exposure, so the cumulative exposures of the workers who entered the cohort before 1970 were substantially underestimated. The committee considered Meyers et al. (2013) to be a strong study for the evaluation of formaldehyde and myeloid leukemia. The study found evidence of an association with myeloid leukemia. The committee reviewed the evidence from both Meyers et al. (2013) and Pinkerton et al. (2004) together because the only important difference between them was that the former had 10 more years of followup (through 2008 instead of 1998). As noted earlier, some evidence in the literature on benzene and leukemia suggests risks decrease with increasing followup (Silver et al. 2002; Glass et al. 2004), and this pattern was observed in the two analyses of the NIOSH garment workers cohort. With followup through 1998, the SMR for all leukemia in those who had an exposure duration of 10 years or more and whose time since first exposure was 20 years or more was 1.92 (95% CI 1.08–3.17); with 10 additional years of followup, the SMR de- creased to 1.74 (95% CI 1.10–2.60). For myeloid leukemia, the SMR for the same exposure definition as above with followup through 1998 was 2.55 (95% 1.10–5.03); with followup through 2008, it was 1.90 (95% CI 0.91–3.50). There was little evidence of increased mortality from lymphocytic leukemia in either reports of the NIOSH garment-workers cohort (Pinkerton et al. 2004; Meyers et al. 2013). The Meyers et al. (2013) report included additional Poisson regression modeling of the data on all leukemia and myeloid leukemia. Those analyses enabled better control of confounding and a more thorough investigation of al- ternative exposure metrics than were available in Pinkerton et al. (2004). There was a strong positive trend in mortality with increasing duration of formalde- hyde exposure (p = 0.01). The standardized rate ratio for 16–19 years of expo- sure was 6.42 (95% CI 1.40–32.20), although the rate ratio dropped in the long- est duration category, at least 19 years. Again, that decrease may reflect the pattern of decreasing risk with extended followup. The garment workers’ coexposures were generally different (lint particles and cleaning-solvent vapors) from those of the NCI formaldehyde-industry co- hort, and this reduced the likelihood that an unmeasured confounder would ex- plain both associations. No other potentially carcinogenic exposures were identi- fied in the plants. As noted above, the exposure assessment had some important limitations. However, the committee agreed with the authors that it is reasonable to assume relatively constant exposure intensity throughout the period of em- 

Independent Assessment of Formaldehyde 109 ployment. On balance, the committee concluded that the finding of an associa- tion between formaldehyde exposure and an association with myeloid leukemia was unlikely to have been explained by an unknown bias or confounder, and chance was an unlikely explanation given the pattern of statistically significant findings. Coggon et al. (2014), an industrial cohort study of mortality in UK chemi- cal workers, was judged to be moderately strong. The publication was an update of Coggon et al. (2003) and included 12 additional years of followup and more than 2,000 additional deaths. The earlier study included very few leukemia deaths and did not provide data specifically on myeloid leukemia. In some re- spects, Coggon et al. (2014) is similar to the NCI formaldehyde-industry study, but it is smaller and provides less information on its exposure assessment. The 2014 update included substantially fewer exposed myeloid-leukemia deaths; for example, there were nine deaths with “high” exposure in Coggon et al. (2014) and 19 deaths in Beane Freeman et al. (2009) with peaks greater than or equal to 4.0 ppm. Coggon et al. (2014) benefited from a semiquantitative exposure as- sessment that provided moderate discrimination among jobs with varied expo- sure intensities. Work histories were abstracted from employment records. Each job was classified into one of five exposure categories—background, low, mod- erate, high, or unknown—by an industrial hygienist who used professional judgment. Quantitative environmental measurements were available after 1970 that covered many jobs, but the authors judged the data insufficient to estimate cumulative exposure or other formal metrics. Exposures were assumed to be the same before 1970 (although anecdotally reported exposures were much higher earlier in the followup period). Peak exposures were not evaluated, nor were temporal trends evaluated or estimated. The authors reported that “each job title [within a factory] was assigned to the same exposure category across all time periods” (Coggon et al. 2014). More than 95% of subjects were exposed before the middle 1980s, and less than 5% of the cohort was still working after the middle1980s. The authors extended the followup of a previously reported cohort of 14,014 men (Acheson et al. 1984; Gardner et al. 1993) who had worked in six plants where formaldehyde was made or used. Mortality was compared with national rates in England and Wales and, in some cases, local rates. Coggon et al. (2014) mention several coexposures, but they do not provide details or report adjusted rates. In the most detailed exposure–response analysis, a nested case– control study, ORs for myeloid leukemia were estimated for four categories of exposure intensity and for a duration 5 years before disease onset. No analysis by duration, cumulative exposure, or other standard continuous exposure metric was presented. CIs for the effect estimates were wide and included the null val- ue. An effect of the size observed in the NCI cohort would probably not have been detectable, so although the results were not inconsistent with those of Beane Freeman et al. (2009), Hauptmann et al. (2009), and Pinkerton et al. (2004), the committee determined that, on balance, the study was generally in- conclusive. 

110 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens The committee judged three additional studies of small industrial cohorts that evaluated formaldehyde and lymphohematopoietic cancers to be moderately strong (Bertazzi et al. 1989; Partanen et al. 1993; Andjelkovich et al. 1995). Each was based on only a handful of cases. Two of the three yielded some evi- dence of an association with lymphohematopoietic cancers (Bertazzi et al. 1989 and Partanen et al. 1993). Bertazzi et al. (1989) reported on cancer mortality in an industrial cohort of 1,330 male workers who produced formaldehyde resins, including 219 for whom specific work histories could not be determined. Among the 179 deaths, there were seven from lymphohematopoietic cancer; 3.9 deaths were expected from national rates and 4.9 deaths expected from local rates, but regardless of which standard was used, the observed excess could have been due to chance. For the entire category of lymphohematopoietic cancers, the authors reported an SMR of 5.35 (95% CI 1.56–14.63) in plastic-resin workers who had formaldehyde exposures during 1965–1969, a period that had no expo- sure controls and therefore likely high exposure. Formaldehyde exposures be- fore 1975 were often greater than 2.4 ppm (3.0 mg/m3). Duration of work in the plant was often short. There was no discussion of possible coexposures. The seven cases of lymphohematopoietic cancer were not further categorized, so no analyses for leukemia was possible. Partanen et al. (1993) conducted a small industrial nested case–control study of the incidence of lymphoma and leukemia in Finnish wood-industry workers who were exposed to formaldehyde. There were only two exposed leukemia cases (type unspecified) with an adjusted OR for formaldehyde exposure of 1.40 (95% CI 0.25-7.91). The Andjelkovich et al. (1995) industrial cohort study of foundry workers examined mortality in 3,929 men who had potential exposure to formaldehyde for at least 6 months during their work in a single automotive iron foundry. Comparisons were with the US population and with workers in the plant who were not exposed to formalde- hyde. There were two deaths from leukemia (type not specified) in exposed workers and three deaths from leukemia in unexposed workers. The study was too small to be informative. Studies of Embalmers and Others in the Funeral Trade, Anatomists, and Pathologists Table 3-6 summarizes the studies that the committee judged to be strong or moderately strong that investigated embalmers and others in the funeral trade, anatomists, and pathologists. NCI assembled and followed a cohort of inactive or deceased embalmers and funeral directors (Hauptmann et al. 2009). The study is particularly useful for evaluating the association between formaldehyde expo- sure and cancer because of the likelihood of high exposures and a high-quality exposure assessment that was conducted by Stewart et al. (1992) and extended by Hauptmann et al. (2009). The authors conducted a nested case–control analy- sis of data on the cohort, using mortality as the outcome measure. The case 

TABLE 3-6 Lymphohematopoietic Cancers: Funeral Workers, Embalmers, Pathologists, and Anatomists No. Cancer Cases in Exposed Key Measures of Association (95% CI) All All Reference and Study Lymphohematopoietic Myeloid Lymphohematopoietic Myeloid Population Cancer Leukemia Leukemia Cancer Leukemia Leukemia Hauptmann et al. 2009 168 44 33 Ever embalm: OR = 1.4 Ever embalm: Ever embalm: (lymphohematopoietic (0.8–2.6) OR = 3.0 (1.0–9.5) OR = 11.2 US funeral directors, malignancy of (1.3–95.6) embalmers nonlymphoid origin) Highest level of (Number of cases identified all exposure from Tables 1 and 2 of the metrics had publication) p<0.05 Levine et al. 1984 8 4 — O/E = 1.2 (0.53–2.43) O/E = 1.6 (0.44–4.10) — ON provincial licensed embalmers (Number of cases identified from Table 1 of the publication) Stroup et al. 1986 18 10 3 SMR = 1.2 (0.7–2.0) SMR = 1.5 (0.7–2.7) SMR = 8.8 (1.8–25.5) US anatomists (Number of cases identified from Table 3 of the publication) Walrath and Fraumeni 25 12 6 PMR = 1.2 (0.79–1.79) PMR = 1.4 (0.73–2.47) PMR = 1.5 1983 (0.54–3.19) NY state-licensed embalmers 111 (Continued) 

112 TABLE 3-6 Continued No. Cancer Cases in Exposed Key Measures of Association (95% CI) All All Reference and Study Lymphohematopoietic Myeloid Lymphohematopoietic Myeloid Population Cancer Leukemia Leukemia Cancer Leukemia Leukemia (Number of lymphohemtopoietic and leukemia cases identified from Table 3 of the publication; number of cases of myeloid leukemia noted on page 408 of the publication) Walrath and Fraumeni 19 12 6 PMR = 1.2 (0.73–1.90) PMR = 1.8 (0.90–3.04) PMR = 1.5 1984 (0.55–3.26) PMR for ≥20 years of CA state-licensed licensure = 2.2 embalmers (Number of lymphohemtopoietic and leukemia cases identified from Table 3 of the publication; number of cases of myeloid leukemia noted on page 4640 of the publication) Abbreviations: CI, confidence interval; O/E, observed/expected; OR, odds ratio; PMR, proportionate mortality ratio; SMR, standardized mor- tality ratio. Source: Committee generated. 

Independent Assessment of Formaldehyde 113 subjects were 6,808 embalmers and funeral directors who died during January 1, 1960–January 1, 1986, and deaths were included if they had an underlying or contributory cause identified as lymphohematopoietic cancers of lymphoid origin (99 cases) or nonlymphoid origin (48 cases). Myeloid leukemia (34 cases) was analyzed as a separate subgroup. The control subjects were identified ran- domly from people in the funeral industry who died of other causes, excluding cancers of the buccal cavity and pharynx, of the respiratory system, and of the eye, brain, or other parts of the nervous system. A quantitative exposure assess- ment was conducted by using information on workplaces and job tasks drawn from interviews with former co-workers and next of kin (Hauptmann et al. 2009) and a NIOSH air-monitoring study (Stewart et al. 1992). All subjects had interview job histories that indicated funeral home or not, embalming or not, and funeral-home ventilation characteristics, which were the predominant factors that affected exposures. The authors found that the average exposure intensity during embalming was 1.7 ppm. The study group was relatively large: there were 34 myeloid-leukemia deaths in the latest followup (33 had “ever embalmed”) (Hauptmann et al. 2009), nearly as many as the 44 in the NCI formaldehyde-industry cohort (Beane Free- man et al. 2009). The findings of Hauptmann et al. (2009) point strongly toward an association between formaldehyde exposure and myeloid leukemia, although measures of associations were stronger in the broad category of all lymphohema- topoietic cancers and all leukemias. The simplest exposure metric—distinguishing ever vs never embalming—was moderately associated with increased mortality from all lymphohematopoietic cancers (OR = 1.4, 95% CI 0.8–2.6), more strongly associated with mortality from all leukemias (OR = 3.0, 95% CI 1.0–9.5), and strongly associated with increased myeloid leukemia mortality (OR = 11.2, 95% CI 1.3–95.6). There was a trend of increasing mortality with increasing duration of embalming (p = 0. 02), rising to OR = 13.6 (95% CI 1.6–119.7) when the group that had more than 34 years of embalming was compared with the group that had never embalmed. There was also a clear trend (p = 0.04) with increasing peak ex- posure, which is a metric similar to the one that Beane Freeman et al. (2009) found to be associated with myeloid leukemia in the different setting of the NCI industri- al-cohort workers. In the highest peak-exposure category (greater than 9.3 ppm), the OR was 13.0 (95% CI 1.4–116.9) compared with no exposure. Another simi- larity to the findings of Beane Freeman et al. (2009) was that there was not a clear trend of increasing mortality with increasing cumulative exposure (p = 0.19). Hauptmann et al. (2009) found no evidence of an association between formaldehyde exposure and leukemia of lymphoid origin. The specificity within the broader grouping increased the committee’s confidence that the results were not likely to be due to an unknown bias. A striking finding of the study was that of the 34 myeloid-leukemia cases, only one did not ever embalm. The ratio of 33:1 contrasts with the ever: never embalming ratio of roughly 4:1 in controls (the exact numbers were 210:55). The 4:1 ratio is a simple way to see the asso- ciations noted above by using different exposure metrics, but it created a meth- odologic limitation for the authors in that the unexposed reference group only 

114 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens had one case. That limitation reduced the precision of the OR reported above. To investigate the effect, the authors repeated the analyses with an enlarged “unexposed” group, which included those who reported fewer than 500 embalm- ing procedures in their career. As expected, the measures of association in the redefined reference group were lower than those reported above, but the patterns were very similar. For example, the OR for those who reported more than 34 years of embalming was 3.9 (95% CI 1.2–12.5) compared with the OR of 13.6 reported above. Strengths of Hauptmann et al. (2009) were that high exposures were readi- ly identified and there were good supporting data on the range for exposure as- signments (Stewart et al. 1992). The model used by the authors explained a high percentage of variability of exposure measurements (74%) (Hauptmann et al. 2009). Errors in quantification would probably not affect the relative ranking of individual exposure histories, especially in the high-exposure category. There was no evidence of confounding by smoking, and few additional chemicals that might confound the association with formaldehyde were involved. In addition, the authors did not adjust for possible changes in work or employer; this could lead to overestimates or underestimates of exposure. The total duration of em- balming work was estimated for all subjects, but some exposure information was missing. Exposures from large spills were important for peaks but infrequent and generally not recorded. The authors also noted that “there was a considera- ble amount of missing data that required imputation for analyses” (Hauptmann et al. 2009, p. 1697). However, sensitivity analyses suggested that the key find- ings were unaffected by the absence of some data points. On balance, the committee concluded that Hauptmann et al. (2009) was a strong study. The committee did not identify any important biases that might have explained the key finding of an association between formaldehyde and myeloid leukemia. The authors persuasively demonstrated that confounding was an unlike- ly explanation. In addition, the clear pattern of associations with multiple increas- ing exposure metrics and after several sensitivity analyses makes it unlikely that chance could have explained the findings. Several small studies of embalmers (Walrath and Fraumeni 1983, 1984; Levine et al. 1984) and anatomists (Stroup et al. 1986) in the 1980s provided sup- porting evidence and were judged to be moderately strong. Each study had only a handful of leukemia deaths and inadequate exposure assessment that was based on the high likelihood of job exposure to formaldehyde and documentation of years of work. Three of the four studies found a pattern of increasing mortality from leukemia in general and from myeloid leukemia specifically, although few were statistically significant; Walrath and Fraumeni (1983, 1984) and Stroup et al. (1986) provided data on myeloid leukemia as the cause of death. Walrath and Fraumeni reported proportionate mortality ratios (PMRs) and proportionate cancer mortality ratios (PMCRs) in a cohort of embalmers in New York State (1983) and California (1984). The PMRs for all leukemias combined were 1.2 (based on 12 deaths) and 1.8 (based on 12 deaths) in New York and Cali- fornia, respectively. Confidence intervals were not given in the publication, but 

Independent Assessment of Formaldehyde 115 they were calculated by the committee (see Table 3-6). There was a small excess in PMRs among workers who had less than 20 years of experience and a statisti- cally significant excess in those who had more than 20 years. The authors noted that embalming fluid contains potentially carcinogenic substances other than for- maldehyde. Levine et al. (1984) studied mortality in a cohort of 1,477 licensed under- takers in Ontario and found four deaths from leukemia, not further specified (2.5 deaths expected, SMR not given).The authors also presented a brief analysis of mortality in formaldehyde-exposed men in eight plants and cohorts of pathologists and anatomists; when the results were combined with their own study of undertakers, 53 leukemia deaths were observed and 44 deaths expected. The publication does not provide additional details. Stroup et al. (1986) reported a retrospective cohort mortality study of 2,317 anatomists, who are exposed to a wide array of solvents, stains, and pre- servatives, including formaldehyde. The authors found 10 deaths from leukemia (6.8 deaths expected, SMR = 1.5, 95% CI 0.7–2.7). Information on potential confounders and biases was not presented, but the authors suggested that low SMRs for smoking-related cancers and cirrhosis of the liver suggested that co- hort members used cigarettes and alcohol less than the general population. Other Studies Potentially Relevant to Formaldehyde and Lymphohematopoietic Hematologic Cancers The committee reviewed all other studies in the background document for formaldehyde for evidence bearing on the question of the carcinogenicity of formaldehyde. Studies that were reviewed were judged to be weak and contrib- uted no informative evidence to this review of lymphohematopoietic cancers were those by Edling et al. (1987), Ott et al. (1989), Hall et al. (1991), Dell and Teta (1995), and Stern (2003). Each was small with a low-discrimination expo- sure assessment that did not permit reliable estimation of an association between formaldehyde exposure and any of the types of cancers of interest. The study by Edling et al. (1987) was a cohort study of mortality that focused on abrasives and leather tanneries, respectively, and formaldehyde constituted a secondary exposure. Hall et al. (1991) updated a study of mortality in a cohort of 4,512 British pathologists (Harrington and Oakes 1984) and found four deaths from leukemia (2.63 deaths expected, SMR = 1.52, 95% CI 0.41–3.89). Followup was nearly complete. Coexposures were not discussed. Dell and Teta (1995) and Ott et al. (1989) studied the same large chemical plants that manufacture a variety of chemicals; few people were exposed to formaldehyde, and the broad job titles limited the specificity of exposure assignments. Dell and Teta (1995) reported on mortality in a cohort of 5,932 male employees in a plastics manufacturing and R&D facility in New Jersey. SMRs for leukemia and aleukemia were 0.98 in hourly employees (12 deaths observed, 12.31 deaths expected, 95% CI 0.50– 1.70) and 1.98 in salaried employees (11 deaths observed, 5.56 expected, 95% 

116 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens CI 0.99–3.54) in salary employees. Numerous possible coexposures were men- tioned by the authors. The text reports eight leukemia deaths (three expected) in the R&D workers, but does not include details. Dell and Teta (1995) provided no data on lymphohematopoietic cancers and formaldehyde. Ott et al. (1989), building on a cohort mortality study by Rinsky et al. (1987), conducted a nested case–control study of mortality in male workers in two chemical-manufacturing facilities and an R&D center in New Jersey. The four causes of death that they studied included nonlymphocytic leukemia. Controls were group-matched on decade of first employment and survival. Exposure was assessed on the basis of departmental usage; coexposures were numerous. There were two cases of non- lymphocytic leukemia (2.6 expected, SMR not given). The Stern (2003) study followed mortality in a cohort of workers in two leather tanneries. It had no formal assessment of formaldehyde exposure, and workers were exposed to many toxic agents, including possible carcinogens. Comparisons were with both US and state rates. There were 16 deaths from leukemia and aleukemia (22 deaths expected according to US rates, SMR = 0.72, 95% CI 0.41–1.18). Results in the two tanneries were similar, as were SMRs based on state rates. There was little evidence of a trend with years of employment. The study did not break down leukemia mortality to permit assessment of the myeloid subgroup. The committee also identified several studies based on general-population registries or surveys that it judged to be weak and that contributed little or no evi- dence to this review of lymphohematopoietic cancers. Blair et al. (2001) was a population-based case–control study of 513 incident cases and 1,087 matched con- trols. It focused on agricultural risk factors in leukemia cases drawn from cancer registries in Iowa and Minnesota. The authors investigated workers who had job- related chemical exposures. In those whose work histories suggested low or high formaldehyde exposure, the ORs for chronic myeloid leukemia were 1.3 in the low-exposure category (7 cases, 95% CI 0.6–3.1) and 2.9 in the high-exposure category (1 case, 95% CI 0.3–24.5). Coexposures were numerous. Richardson et al. (2008) conducted a population-based case–control study of non-Hodgkin lym- phoma and chronic lymphocytic leukemia incidence in Germany. Semiquantita- tive estimates of formaldehyde exposure derived from job-history data, and a job– exposure matrix were weakly positively associated with non-Hodgkin lymphoma and chronic lymphocytic leukemia, but confidence intervals were wide and in- cluded the null. The study did not address myeloid leukemia. Hansen and Olsen (1995), which was a Danish cancer incidence study, was described earlier because it found an increased incidence of sinonasal can- cer in formaldehyde-exposed workers. The authors reported an SPIR for leuke- mia in men who worked in 265 factories that imported or manufactured formal- dehyde. They found 39 leukemia deaths (47.0 deaths expected, SPIR = 0.8, 95% CI 0.6–1.6). Coexposures were not investigated. The exposure definition used in the study (being a blue-collar worker in a company that was registered with the government as a user of formaldehyde) probably led to substantial misclassifica- tion with the likely consequence of underestimation of true risks. Another limi- tation of the study was that it did not report results separately for leukemia 

Independent Assessment of Formaldehyde 117 types. For all leukemia types combined, the study did not find evidence of an increased incidence in formaldehyde-exposed workers, although the confidence interval was wide (SPIR = 1.0, 95% CI = 0.6–1.4). Stellman et al. (1998) analyzed cancer mortality in members of the ACS Cancer Prevention Study II, a very large prospective industrial cohort study. Mortality was examined after 6 years in 45,399 men who had reported being employed in wood industries or occupationally exposed to wood dust and 362,823 who did not report such exposures. Thirty-two leukemia cases were observed in those who reported wood-dust exposure (SMR = 0.90, 95% CI 0.63–1.30), and 14 were observed in the partially overlapping group in wood- related occupations (SMR 1.08, 95% CI 0.6–1.85). The exposure assessment for formaldehyde was by self-report alone, which is likely to be of poorer quality than an expert review and job–exposure matrix. Furthermore, the authors did not report results for subtypes of leukemia. As a result, this study was judged to be of little utility for the committee’s assessment. Summary of Evidence on Lymphohematopoietic Cancers In summary, the committee concluded that the epidemiologic studies pro- vided evidence of a causal association between formaldehyde and myeloid leu- kemia in humans. Evidence of an association was derived from two strong in- dustrial cohorts (Beane Freeman et al. 2009; Myers et al. 2013), one strong cohort of embalmers (Hauptmann et al. 2009), and several moderately strong cohorts from the chemical industry (Coggon et al. 2014) and the funeral trade (Walrath and Fraumeni 1983, 1984; Stroup et al. 1986). See Tables 3-5 and 3-6 and Figures 3-1 and 3-2 for key measures of association supporting this conclu- sion. The conclusion was based on the strength, consistency, temporality, dose– response relationships, and coherence of the evidence according to the quality criteria presented in Table 3-1. To present data from the studies, it was necessary to choose a particular exposure definition; however, it is important to note that, in its evaluation of the body of evidence, the committee did not choose a single exposure metric a priori for analysis. Instead, it looked at the full set of exposure metrics and their asso- ciations with disease. Figure 3-1 emphasizes a pattern noted earlier—that is, in the studies that were large enough and detailed enough to present associations between formal- dehyde and the “nested” case definitions of all types of lymphohematopoietic cancers, all leukemias, and myeloid leukemia, the measures of association tend- ed to increase as the definition was narrowed (the data points for the nested sets of case definitions are linked by a solid line in Figure 3-1).The figure also illus- trates that the stronger and larger studies generally reported stronger associations with formaldehyde and were more likely to present confidence bounds for their 

118 FIGURE 3-1 Summary of strrong and moderately strong studies of fo ormaldehyde and lym mphohematopoietic ccancers. Note: Data ppoints con- nected by a line indicate results from the same stud dy according to the saame exposure metricss but for different tum mor sites. 

Independent Assessment of Formaldehyde 119 effect estimates that excluded the null. Measures of association between formal- dehyde exposure and myeloid leukemia are represented in Figure 3-2 for all studies that reported this association. There is a pattern of positive findings from studies that were judged to be large and strong studies. Low-precision studies, such as those with a small cohort, only a few cases, or limited exposure assessments, may provide some useful data on risk estimates if several studies were performed. When several small populations are studied using a good design, the measures of association would not be expected to be the same. They would have a distribution that would cluster around the overall risk value for the population; some estimates would be above that value and some would be below that value. If the risk estimates for formaldehyde exposure and myeloid leukemia showed a distribution that was shifted above 1.0 so that few studies showed RRs below 1.0, that pattern of results suggests that there may be a causal relationship between exposure and disease risk. The closer the risk values cluster around 1.0 (some above and some below), the less likely it is that a relationship exists. In Figures 3-1 and 3-2, nearly all RRs are above 1.0, which suggests that a relationship exists. That argument does not imply that all studies are equal. Strong studies make more precise estimates of the RR and are more useful in assessing factors that may affect the RR compared with weaker studies. Strong studies should not produce large RRs when the relationship is weak or absent unless there is a bias in the data. Reference = strong study quality Beane Freeman et al. 2009 = moderately strong study quality = weak study quality Hauptmann et al. 2009 Meyers et al. 2013 Coggon et al. 2014 Walrath and Fraumeni 1983 Walrath and Fraumeni 1984 Stroup et al. 1986 Blair et al. 2001 1 10 1 Risk Estimate FIGURE 3-2 Summary of key findings from all studies that reported associations be- tween formaldehyde and myeloid leukemia. 

120 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens As noted above, the informative epidemiologic studies were the ones that were large, that estimated exposure systematically, that had credible comparison groups, and that assessed cancer end points reliably. Studies that did not find associations between exposure and myeloid leukemia were usually too small to detect an effect, did not break out results for myeloid leukemia, or used methods of exposure assessment that resulted in exposure misclassification. A single, large, high-quality study (Beane Freeman et al. 2009) found evidence of in- creased risk of Hodgkin lymphoma and multiple myeloma in those who had a history of high peak exposures. Those findings do not appear to be supported by other epidemiologic evidence and, in the committee’s view, constitute insuffi- cient evidence of effects. Cancer at Other Sites The committee conducted a literature search (see Appendix D) to identify studies that examined associations between formaldehyde and cancers at other sites (Table 3-7). Four studies were identified that reported measures of associa- tion between formaldehyde and lung cancer. Two of the studies were judged to be moderately strong (Siew et al. 2012; Mahboubi et al. 2013) and two studies were judged to be weak (Checkoway et al. 2011; Luo et al. 2011). TABLE 3-7 Other Cancer Sites Reference and Study No. Lung Cancer Cases in Findings (95% CI) Population Exposed Checkoway et al. 2011 Number of cases identified from Hazard ratio for ≥10 years Table 3 of the publication formaldehyde exposure = 2.1 Female textile workers in (0.4–11.0) Shanghai, China Cases with ≥10 years of formaldehyde exposure: n = 2 Luo et al. 2011 Not relevant; unit of analysis RR for counties with any was county formaldehyde release vs none = General population in 13 US 1.14 (1.05–1.24) regions covered by SEER registries Mahboubi et al. 2013 Number of cases identified from OR for pooled population Table 3 of the publication comparing substantial with no General population in Montreal, exposure = 0.88 (0.63–1.24) Canada Cases with “substantial” exposure: n = 99 No evidence of trend with duration, time since first exposure Siew et al. 2012 Number of cases identified from RR for any formaldehyde Table 3 of the publication exposure = 1.18 (1.12–1.25) Finnish general population Cases with any formaldehyde exposure: n = 1,831 Abbreviations: CI, confidence interval; OR, odds ratio; RR, relative risk; SEER, Surveil- lance, Epidemiology, and End Results program of the National Cancer Institute. Source: Committee generated. 

Independent Assessment of Formaldehyde 121 Mahboubi et al. (2013) published a large case–control study of lung cancer and formaldehyde exposure. The authors used a long-running study of lung can- cer in Montreal that was based on incident cases gathered during two time peri- ods: 1979–1986 and 1996–2002. The well-described exposure assessment meth- ods were based on a detailed questionnaire on jobs and duties performed. Trained occupational hygienists evaluated each questionnaire, blinded to case and control status, on three dimensions of formaldehyde exposure: confidence (possible, probably, definite); relative concentration (low, medium, high); and frequency of use in a normal week (low, medium, high). The study was relative- ly large; there were 99 cases with exposure to formaldehyde that were judged by the occupational hygienists to be “substantial” exposures. The study found little to no evidence of incidence of lung cancer associated with any of the formalde- hyde exposure measures. The study investigated potential confounding by smok- ing, and none was found. The study was able to evaluate effects separately in men and women, and no effect was observed in either gender. It was also able to stratify on the three primary histologic types of lung tumors (squamous cell, small cell, and adenocarcinoma) and, again, there was no evidence of an asso- ciation with formaldehyde exposure for any type. Siew et al. (2012) established a population-based cohort of all Finnish men who were born during 1906–1945 and followed the cohort for cancer incidence by linking to data in the Finnish Cancer Registry. They used the men’s occupa- tions reported to the 1970 national census to estimate occupational exposures to a wide array of chemicals, including formaldehyde, and found that men who developed lung cancer were 18% more likely to have jobs that involved expo- sure to formaldehyde than men who did not develop lung cancer (RR = 1.18, 95% CI 1.12–1.25). That finding was positive, and the size of the study (more than 30,000 lung-cancer cases) resulted in tight confidence limits, but the au- thors were doubtful of the finding because of the likelihood that they were una- ble to control fully for confounding by smoking and by concurrent exposures to other strong lung carcinogens, particularly asbestos. The committee concurred with those concerns. Checkoway et al. (2011) had a strong study design, but the committee judged it to be weak for the purposes of this assessment because few cases were exposed to formaldehyde. The study was a large industrial case-cohort study (628 incidence lung-cancer cases) of Chinese female textile workers and it had detailed exposure assessment. However, the prevalence of formaldehyde expo- sure was low, and only two cases had 10 years or more of formaldehyde expo- sure. The resulting measure of association was imprecise: the hazard ratio for 10 or more years of formaldehyde exposure was 2.1 (95% CI 0.4–11). Luo et al. (2011) conducted a population-based ecologic study of incident cases in US counties. They linked lung-cancer incidence from the Surveillance, Epidemiology, and End Results Program cancer registries to US Environmental Protection Agency Toxics Release Inventory data on formaldehyde emissions from industries. They found that a county’s lung-cancer rate was positively as- sociated with releases of formaldehyde (and chromium and nickel). For exam- 

122 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens ple, the RR was 1.18 (95% CI 1.05–1.33) when nonmetropolitan counties that had any formaldehyde release were compared with counties that had no formal- dehyde release. The results are intriguing, but, as the authors note, evidence from individual-level studies is needed to support the finding. The committee concluded that the newly identified studies do not provide enough evidence to indicate a causal association between formaldehyde and lung cancer. There remains a good possibility that confounding factors explain the increase in lung cancer reported in some formaldehyde studies. In addition, the studies yielded no epidemiologic evidence that indicated an association be- tween formaldehyde exposure and cancer at other sites. CANCER STUDIES IN EXPERIMENTAL ANIMALS This section reviews the evidence of carcinogenicity in experimental ani- mal studies and applies the NTP criteria to produce the committee’s independent evaluation. In reviewing the evidence, the committee looked at primary litera- ture and considered analyses in other reviews, including those by the Interna- tional Agency for Research on Cancer (IARC 1982, 1995, 2006a) and NTP (2010, 2011). To capture studies that may have been published concurrently with the completion of the background document for formaldehyde up to 2013, the committee undertook an independent literature search. See Appendix D (Box D-2 and Figure D-2) for more information. Studies of Low Power for Detecting Malignancies Some bioassays discussed in the section “Studies of Cancer in Experi- mental Animals” of NTP’s background document for formaldehyde are of lim- ited adequacy to evaluate the carcinogenicity of formaldehyde (Table 3-8). Some of the studies were designed to follow up on studies that found carcino- genicity, for example, to explore hypotheses related to etiology or to look for differences in activity in different species. Those studies have findings of inter- est in considering progression to carcinogenesis, but they had low power to de- tect malignancy, mostly because they were not of sufficient duration. In addi- tion, some studies have small groups, particularly the studies that used monkeys (Rusch et al. 1983; Monticello et al. 1989). All the studies that were of low power to detect malignancies were inhala- tion studies except that by Tobe et al. (1989), which exposed animals to formal- dehyde via drinking water. Tobe et al. had a relatively small group (20 male and 20 female) at the start of the study; all the animals in the high-dose group receiv- ing 5,000 ppm of formaldehyde in drinking water and a substantial fraction in the low-dose groups receiving 200 ppm of formaldehyde in drinking water (46.9% of males and 33.7% of females) died before the end of the study, alt- hough survival in the group receiving 1,000 ppm of formaldehyde in drinking water was relatively good. Mortality began within the first month of the study. With the small initial group and substantial noncancer mortality in the high- and 

Independent Assessment of Formaldehyde 123 low-dose groups, the study has little overall power for evaluating the oral car- cinogenicity of formaldehyde. Additional studies published decades ago that were identified from bioassay tabulations (for example, the US Public Health Service 149 series Survey of Compounds Which Have Been Tested for Carcino- genicity) were also of short duration and had other deficiencies (Garschin and Schabad 1936; Watanabe et al. 1954; Muller et al. 1978), as discussed in more detail in Chapter 2. TABLE 3-8 Studiesa of Low Power for Detecting Malignancies Findings of Interest in Species Limitations Formaldehyde-Treated Animals Reference C3H mice  Examined only lung; no Basal-cell hyperplasia, Horton et al. 1963 examination of nose epithelial stratification,  Study terminated for most squamous-cell metaplasia, groups at 35 weeks and atypical metaplasia in  Small group in single animal trachea and major bronchi group allowed to live longer Wistar rats  Short duration (13 weeks) Proliferative lesions in nasal Woutersen et  Small group (10 male and and olfactory epithelium al. 1987 10 female) Wistar rats  Short duration (13 weeks) Disarrangement, hyperplasia, Wilmer et al. 1989  Histopathology only of nasal squamous metaplasia with cavity keratinization of epithelium Wistar rats  Short duration (1 year) Increased basal-cell hyperplasia Appelman et  Small group (10 male) and squamous-cell metaplasia al. 1988  Only nasal cavity examined Wistar rats  Relatively small initial group Forestomach hyperkeratosis, basal Tobe et al. 1989 (20 male and 20 female) and high and squamous-cell hyperplasia; mortality glandular stomach hyperplasia Wistar rats  Short duration (32 weeks) 8 of 10 treated rats with Takahashi et  Small group (10 male) forestomach papilloma, none al. 1986 in controls Fischer rats  Short duration (26 weeks) Increased squamous-cell Rusch et al. 1983  Relatively small group (20 metaplasia and hyperplasia, male and 20 female) basal-cell hyperplasia at high doses Syrian golden  Short duration (26 weeks) No significant findings Rusch et al. 1983 hamsters  Small group (10 male and 10 female) Cynomolgus  Short duration (26 weeks) Squamous-cell metaplasia and Rusch et al. 1983 monkeys  Small group (6 male) hyperplasia of nasal turbinates  Age unknown Rhesus  Short duration (1–6 weeks) Mild degeneration and Monticello et monkeys  Small group (9 male) squamous-cell metaplasia of al. 1989 nasal epithelium; increased cell proliferation rate a All studies conducted by inhalation except studies by Tobe et al. (1989) and Takahashi et al. (1986), which were via drinking water. Source: Committee generated. 

124 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens The study by Takahashi et al. (1986), which exposed male Wistar rats to formaldehyde in water at 5,000 ppm for 32 weeks is notable. Although it was of short duration, eight of 10 exposed rats and no control animals developed forestomach papilloma. The formaldehyde group was serving as a reference group in a study of the effect of formaldehyde on N-methyl-N′-nitro-N- nitrosoguanidine carcinogenicity. Because of the very short study duration, the finding of tumors is particularly notable. The two studies conducted in nonhuman primates are also noteworthy. They were of short duration and used small numbers of animals, but both studies demonstrated clear cellular and proliferative lesions of the nasal turbinates. Ru- sch et al. (1983) reported squamous-cell metaplasia and hyperplasia in the high- dose exposure group of six cynomolgus monkeys exposed to formaldehyde at 2.95 ppm 22 hours/day, 7 days/week for 26 weeks. Monticello et al. (1989) ex- posed rhesus monkeys to formaldehyde at 6 ppm 6 hours/day, 5 days/week for 1 week (n=3) or 6 weeks (n=3). The authors reported increased rates of nasal epi- thelial cell-proliferation with squamous-cell metaplasia of the transitional and respiratory epithelia of the nasal passages and squamous-cell metaplasia of the respiratory epithelia of the trachea and large airways of the bronchial tree. Even though those findings do not reflect overt carcinogenesis, they are highly remi- niscent of the preneoplastic epithelial lesions of the nasal cavity that were ob- served to precede nasal malignancies in chronic rat studies. Evidence from Informative Studies Chapter 2 discusses whether the committee found NTP’s evaluation of the evidence and application of its criteria scientifically sound. The committee’s independent application of the NTP criteria emphasizes studies that are designed with greater sensitivity to detect an effect. Table 3-9 shows the highest-quality inhalation studies in boldface. They all had relatively large groups (90 animals or more), handled test material adequately, and included well-defined compari- son groups (Kerns et al. 1983; Sellakumar et al. 1985; Monticello et al. 1996). The studies were all conducted in rats. In each, formaldehyde caused high inci- dences of rare malignant nasal tumors (squamous-cell carcinomas) at air- chamber concentrations of 10–15 ppm; these tumors are rarely seen in carcino- genesis bioassays and can be characterized as occurring “to an unusual degree” with respect to incidence. It is noteworthy that none of the animals in control groups in any of the long-term exposure studies had a tumor of this type. The Kerns et al. (1983) study was among the group of highest-quality studies. That experiment had a robust finding of squamous-cell carcinoma in both male and female rats, and the incidences were also increased to an unusual degree. The initial report of this study (Battelle 1981) stated there was a significant increase in bone marrow hyperplasia in rats following exposure to formaldehyde. The short-term exposure study by Feron et al. (1988) did not achieve statistical sig- nificance (p = 0.1 by Fisher exact comparison between the top dose group and controls). 

TABLE 3-9 Nasal Squamous-Cell Carcinoma in Long-Term Inhalation Studies of Formaldehyde1 Concentrations in Air (Incidences) Species and Study Duration Strain (week)2 Sex No SCC Effect SCC Other Findings Reference Mouse 104 M 0 ppm (0/109) 14.3 ppm (2/104) Epithelial dysplasia and squamous Kerns et al. 1983; B6C3F1 2 ppm (0/100) metaplasia in high- and middle-dose groups; Battelle 1981 5.6 ppm (0/106) epithelial hyperplasia at high doses F 0 ppm (0/114) — Dysplasia in high- and middle-dose groups; 2 ppm (0/114) squamous metaplasia in the high-dose group 5.6 ppm (0/112) 14.3 ppm (0/119) Rat Wistar 130 (13 weeks of M 0 ppm (0/45) 10 ppm (1/44) One carcinoma in situ and two polypoid Feron et al. 1988 exposure)3 20 ppm (3/44) adenomas at 20 ppm 120 M 0 ppm (0/26) 0.1 ppm (1/26) — Woutersen et al. 1 ppm (1/28) 1989 10 ppm (1/26) Rat F344 104 M 0 ppm (0/118) 5.6 ppm (1/119) Four high-dose animals with other nasal Kerns et al. 1983 2 ppm (0/118) 14.3 ppm (51/117*) malignancies F 0 ppm (0/114) 5.6 ppm (1/116) One high-dose female with other nasal 2 ppm (0/118) 14.3 ppm (52/115*) malignancy 104 M 0 ppm (0/90) 6 ppm (1/90) Nasal malignancies in one animal at 10 Monticello et al. 0.7 ppm (0/90) 10 ppm (20/90*) ppm and one animal at 15 ppm; polypoid 1996 2 ppm (0/96) 15 ppm (69/147*) adenomas in 14 animals at 15 ppm 120 M 0 ppm (0/32) 15 ppm (13/32*) An additional 3 rats at 15 ppm with Kamata et al. 1997 0.3ppm (0/32) squamous-cell papilloma 2 ppm (0/32) (Continued) 125 

126 TABLE 3-9 Continued Concentrations in Air (Incidences) Species and Study Duration Strain (week)2 Sex No SCC Effect SCC Other Findings Reference Rats Sprague Life M 0 ppm (0/99) 15 ppm (38/100*) Two treated rats with other nasal Sellakumar et al. Dawley malignancies; 10 with squamous-cell 1985 papillomas 104 F 0 ppm (0/15) 12.4 ppm (1/16) Squamous-cell metaplasia or dysplasia in 10 Holmström et al. exposed rats 1989 Hamster Life M 0 ppm (0/132) — Minimal hyperplastic and metaplastic Dalbey 1982 Syrian 10 ppm (0/88) response Golden 30 ppm (0/50) *Statistically significant, p < 0.0001 by pairwise Fisher exact comparison. 1 Well-conducted studies with relatively large groups are in boldface. 2 All exposures were for 6 hours/day, 5 days/week except the Dalbey (1982) study in hamsters, which had one group at 5 hours/day, 5 days/week and one group at 5 hours/day, 1 day/week. 3 13 weeks of exposure followed by a long period of no exposure. Results of experiments with shorter exposure times not tabulated. Abbreviation: ppm, parts per million; SCC, squamous-cell carcinoma. Source: committee generated. 

Independent Assessment of Formaldehyde 127 In addition to the findings of the robust rat studies, Kerns et al. (1983) car- ried out a study in mice. Nearly all 17 high-dose mice that survived 24 months had nasal lesions (dysplasia and metaplasia), and two also had squamous-cell carcinoma. As noted by the authors and in the background document for formal- dehyde, that finding is sufficient to demonstrate the potential for these tumors in the mice exposed by inhalation when put into the context of evidence for this site in the rat and when the rarity of the tumor is considered. The findings of squamous-cell carcinoma in long-term studies that exposed mice and rats via inhalation are supported by the preneoplastic lesions (for example, squamous metaplasia with keratinization of epithelium) and other nasal lesions found in the shorter-term studies. The study using hamsters found no effect (Dalbey 1982). The Kerns et al. (1983), Kamata et al. (1997), and Sellakumar et al. (1985) inhalation studies included histopathologic examinations of non–respiratory tract tissues; the other inhalation studies did not. Kerns et al. (1983) was report- ed in full in the Battelle (1981) report to the Chemical Industry Institute of Toxi- cology. The Battelle report discusses findings of leukemia and lymphoma that were not found to be exposure-related. However, diffuse multifocal bone mar- row hyperplasia in formaldehyde-exposed animals was increased in both treated males (six of 114 controls vs 26 of 111 treated, p = 0.0001) and females (seven of 113 controls vs 28 of 115 treated, p = 0.0001). Kamata et al. (1997) and Sellakumar et al. (1985) reported no statistically significant nonrespiratory tu- mor findings but provided no detail regarding other non–respiratory tract histo- pathology. The database for evaluating oral exposure to formaldehyde is less robust than for inhalation exposure. Three studies exposed rats to formaldehyde via drinking water over long periods (Til et al. 1989; Soffritti et al. 1989, 2002). The studies are described at length by IARC (2006a) and NTP (2010). The study by Til et al. (1989) exposed Wistar rats to formaldehyde that was generated with 95% pure paraformaldehyde and 5% water. The adminis- tered drinking-water concentrations were 0, 20, 260, and 1,900 mg/L; the initial groups were 70 animals per sex at each dose; and the interim sacrifices occurred at 53 and 79 weeks. The intestines were not examined histologically in the mid- dle- and low-dose groups but were in the high-dose group. The authors found no increases in cancer incidence in the gastrointestinal tract. A male in the low-dose group and a female in the control group had gastric papilloma. Nearly all male (seven out of 10) and female (five out of nine) animals in the highest-dose group had epithelial hyperplasia of the forestomach, and substantial fractions had focal hyperkeratosis of the forestomach and hyperplasia of the glandular stomach. In contrast, in the 32-week study by Takahashi et al. (1986), noted above in the discussion of the low-power studies, eight of 10 male Wistar rats exposed via drinking water to formaldehyde at 5,000 mg/L had stomach papilloma. The ex- posure level in the Takahashi et al. (1986) study was higher than in the Til et al. (1989) study. 

128 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens In a series of experiments in Sprague Dawley rats, Soffritti et al. (1989, 2002) administered formaldehyde via drinking water. The studies included full histologic examination of all tissues. In the first study (Soffritti et al. 1989), formaldehyde of unspecified purity was administered to 25-week-old breeders (20 controls and 18 treated) at 2,500 mg/L in water. The offspring were exposed in utero via the dam and then postnatally via water for 104 weeks. In the breed- ers, no stomach or intestinal tumors were observed in the controls, whereas stomach tumors were observed in one treated female (benign) and one treated male (malignant). In the offspring, similarly, there were no stomach or intestinal tumors in the control animals (59 males and 49 females). However, in treated offspring (36 males and 37 females), a variety of benign and malignant gastroin- testinal tumors were observed at a low incidence, including malignant leiomyo- sarcoma, which is exceedingly rare in these animals. Leiomyosarcoma was ob- served in stomach tissues in one treated female and one treated male and in intestinal tissue of five treated females (statistically significant at p = 0.01) (IARC 2006a, NTP 2010). In addition, nonleiomyosarcoma gastrointestinal tu- mors were observed in two males (one benign and one malignant) and one fe- male (malignant). Soffritti et al. (2002) later followed up with a long-term drinking-water study with multiple exposure groups and groups with lower exposures than in the earlier (Soffritti et al. 1989) study: 0, 10, 50, 100, 500, 1,000, and 1,500 mg/L; 50 animals of each sex per group, except for the controls, which had a group size of 100. Four treated males developed leiomyosarcoma at 10 mg/L (forestomach, one animal), 1,000 mg/L (glandular stomach, one animal), and 1,500 mg/L (intestine, two animals), and seven treated females developed leio- myoma at 10 mg/L (two animals), 50 mg/L (one animal), and 1,500 mg/L (three animals) or leiomyosarcoma at 50 mg/L (one animal). None of the 200 untreated control animals (100 male and 100 female) had these tumors. Soffritti et al. (2002) also reported an increased incidence of hemolympho- reticular tumors in some groups. The finding is of interest, but there is uncertain- ty about it because of the changing counts of the tumors in earlier study reports (as noted by IARC 2006a), the pooling of tumors of different cellular origins, and recent questions raised about the evaluation of this class of tumors by this laboratory (Malarkey and Bucher 2011; Gift et al. 2013). Total mammary tu- mors also increased with increasing dose in the females; this, too, involved pool- ing of tumors of different origins (for example, adenocarcinoma and liposar- coma). Although noteworthy, the findings of hemolymphoreticular and mammary tumors are not used in the committee’s independent evaluation. Committee Evaluation in the Context of the Report on Carcinogens Listing Criteria Applying the NTP criteria to the bioassay data for formaldehyde, the committee draws the following conclusions about exposure to formaldehyde in experimental animals: 

Independent Assessment of Formaldehyde 129 1. Multiple species and multiple tissue types affected by the exposure:  Multiple species: Increase in malignant tumors in rats (F344 rats [Kerns et al. 1983; Monticello et al. 1996; Kamata et al. 1997], Sprague Dawley rats [Sellakumar et al. 1985; Soffritti et al. 1989], and Wistar rats [Feron et al. 1988; Woutersen et al. 1989]) and mice (B6C3F1 mice [Kerns et al. 1983]).  Multiple tissue types: Malignancies of nasal epithelium (mostly squamous-cell carcinoma) (Kerns et al. 1983; Sellakumar et al. 1985; Feron et al. 1988; Woutersen et al. 1989; Monticello et al. 1996; Kamata et al. 1997) and gastrointestinal tract (leiomyosar- coma) (Soffritti et al. 1989 [offspring]; Soffritti et al. 2002 [adults]). 2. Carcinogenicity by multiple routes of exposure: Inhalation (Kerns et al. 1983; Sellakumar et al. 1985; Feron et al. 1988; Woutersen et al. 1989; Monti- cello et al. 1996; Kamata et al. 1997) and oral (Soffritti et al. 1989 [offspring]; Soffritti et al. 2002 [adults]). 3. Carcinogenicity to an unusual degree with respect to incidence, site, type of tumor, or age at onset: Nasal tumors are rare in untreated rats and in multiple studies occurred in treated rats at relatively high incidence (Kerns et al. 1983; Monticello et al. 1996). The committee concludes that there is sufficient evidence that formalde- hyde is carcinogenic in experimental animals. TOXICOKINETICS This section outlines multiple aspects of the toxicokinetics of gas-phase formaldehyde. The most likely route of exposure in humans is inhalation, and the committee has focused on this route. Information on the reactivity and me- tabolism of formaldehyde is followed by specific information on endogenous vs exogenous formaldehyde levels and on the inhalation dosimetry of this gas, par- ticularly as related to the potential for absorption into the bloodstream and sys- temic distribution. The current report focuses on formaldehyde gas; however, it is worth noting that paraformaldehyde powder is used in some embalming and chemical applications. These uses may produce exposures to airborne particles of paraformaldehyde in addition to gas-phase formaldehyde. There is currently a dearth of information on human health effects associated with exposure to para- formaldehyde particles. Reactivity and Metabolism Formaldehyde is a volatile, organic, one-carbon aldehyde that exists as a gas at room temperature. It is water-soluble and reacts reversibly with water to form methanediol, which is the principle aqueous form in tissues after exposure to formaldehyde (Fox 1985). It can self-polymerize to form paraformaldehyde, 

130 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens which is a solid at room temperature that has the ability to break down when heated to release the monomer. It also reacts reversibly with amine and sulfhy- dryl groups, and this may ultimately result in cross-links between macromole- cules. The inherent chemical reactivity of gas-phase formaldehyde is important to note because it plays a key role in its interaction with many macromolecules and cellular processes. The innate chemical reactivity of formaldehyde allows it to act as a cross-linking agent to fix tissue for pathological analysis and as a re- actant in the synthesis of numerous industrial products. Those same chemical properties can, in part, explain its numerous toxic properties. Formaldehyde is reactive because its carbonyl atom acts as an electrophile, which reacts reversi- bly with nucleophilic sites on cell membranes, amino groups on proteins and DNA, and thiol groups on such biochemicals as glutathione (Bolt 1987). The native reactivity of formaldehyde contributes to the well-established irritant properties of formaldehyde. Studies have found formaldehyde to cause dermatitis on dermal exposure and both eye and nasal irritation on inhalation exposure (Paustenbach et al. 1997). The nasal sensitization does not appear to be related to concentrations of glutathione–formaldehyde dehydrogenase; this indi- cates that formaldehyde itself, not metabolic products, is the irritant (Zeller et al. 2011b). Formaldehyde also reacts with macromolecules—a feature that has been used extensively to detect exogenous exposure to formaldehyde through meas- urement of formaldehyde–DNA adducts (ATSDR 1999; Lu et al. 2011) and proteins (Edrissi et al. 2013a). The reaction of formaldehyde with cellular com- ponents contributes to the sensitization of people to formaldehyde, which is manifested as allergic reactions and alterations in a person’s immune system (Costa et al. 2013; Hosgood et al. 2013; Lino-dos-Santos-Franco et al. 2013). Although the mechanism is unclear, several reports associate formaldehyde with induction of an occupational asthmatic response in exposed people (Tang et al. 2009; McGwin et al. 2011) and in animal models (Wu et al. 2013). Formaldehyde is rapidly absorbed and biotransformed extensively at the point of contact after ingestion or inhalation. It is primarily oxidatively biotrans- formed by glutathione-dependent formaldehyde dehydrogenase (FDH), official- ly named alcohol dehydrogenase 5 (ADH5), and S-formyl-glutathione dehydro- genase to formic acid (IARC 2006a). Formic acid can be ionized to formate and excreted via the kidney, further biotransformed to CO2 and exhaled, or con- densed with tetrahydrofolate and enter the one-carbon pool (IARC 2006a). In one study, 70% of a 14C-labeled formaldehyde dose was found to be excreted as [14C]CO2 within 12 hours, and the remainder entered the one-carbon pool, where it was incorporated into biomolecules in the body (Buss et al. 1964). Formalde- hyde dehydrogenases are ubiquitous in all tissues, including the respiratory tract, with no distinct “regional” differences in the biotransformation of formaldehyde (Casanova-Schmitz et al. 1984; Thompson et al. 2008). The biotransformation of formaldehyde is similar in all species tested. The rapid biotransformation of formaldehyde at the point of contact limits the access of formaldehyde systemi- cally. 

Independent Assessment of Formaldehyde 131 Endogenous vs Exogenous Sources Formaldehyde exposure has both exogenous and endogenous sources. It is produced intracellularly as a component of the one-carbon pool intermediary metabolism pathways. It is also the product of metabolism of drugs and other exogenous compounds (NTP 2010; NRC 2011). Because formaldehyde is nor- mally present in tissues, the toxicokinetics of exogenous formaldehyde exposure must be evaluated in the context of the relatively large amounts of formaldehyde (near 0.1 mM) that are endogenously present. Measurement of tissue formalde- hyde is somewhat difficult because of its volatility and reactivity. Many tech- niques rely on extraction followed by mass spectrometry (for example, Heck et al. 1982). Those methods provide a measure of free and reversibly bound for- maldehyde but do not differentiate between the two. Formaldehyde, through the one-carbon pool, is metabolically incorporated into tissue macromolecules. Therefore, simple use of 14C-labeled formaldehyde does not provide a direct measure of the distribution of parent exogenously administered formaldehyde (NTP 2010; NRC 2011). As noted above, because of its reactivity, formaldehyde may form DNA–protein cross-links, DNA–DNA cross-links, and protein or DNA adducts (Lu et al. 2010a; NTP 2010; NRC 2011; Edrissi et al. 2013b). Those moieties have the advantage of being more stable and longer-lasting than formaldehyde itself and have been used as biomarkers of cellular exposure to formaldehyde. It is important to recognize that use of the moieties (for example, DNA–protein cross-links) as biomarkers of cellular formaldehyde delivery does not require a direct link to tumorigenesis. The endogenous formaldehyde concentration in whole blood of rodents and nonhuman primates is about 0.1 mM. The concentration in tissues is proba- bly somewhat higher (NTP 2010; NRC 2011). That value represents free plus reversibly bound formaldehyde. Information on the fraction of blood formalde- hyde that is free vs bound is not available. Whether from endogenous or exoge- nous sources, formaldehyde is extensively metabolized to formate via formalde- hyde dehydrogenase as described above. Inhalation Dosimetry Because inhalation is the most likely route of exposure to formaldehyde, an understanding of the fate of inhaled formaldehyde is critical for evaluation of its toxicity. As would be expected for a water-soluble highly reactive gas (Kim- bell 2006), inhaled formaldehyde is effectively removed from the airstream. Thus, it is expected that formaldehyde will be efficiently removed from the air- stream in the first airways with which it comes into contact, either the nose dur- ing nose breathing or the tracheobronchial airways during mouth breathing. Wa- ter-soluble reactive gases may be absorbed efficiently in the mouth and pharynx during mouth breathing (Frank et al. 1969); although this is likely to occur with formaldehyde, it has not been confirmed experimentally. Experimental studies in the dog (Egle 1972) indicate greater than 95% deposition of inhaled formal- 

132 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens dehyde in the nose, lower respiratory tract, and total respiratory tract. A pub- lished abstract (Patterson et al. 1986) provides similar data on nasal deposition in the rat. Numerous state-of-the-art inhalation dosimetry mathematical models have been directed toward dosimetry of inhaled formaldehyde. They have recently been extensively and appropriately reviewed (NRC 2011). The models suggest that inhaled formaldehyde is not deposited uniformly throughout the nose, but local areas, “hot spots”, receive a higher delivery of the dose than other areas. Those areas correlate closely, in the rat, with areas in which DNA–protein cross- link studies indicate high cellular delivery and with areas in which tumors are most likely to arise. Models suggest that rates of localized delivery to small re- gions in the human nose may be similar to those observed in rats exposed at the same concentration (Kimbell et al. 2001). The modeling prediction adds weight to the idea that formaldehyde may pose a carcinogenic hazard to the human nose. Models suggest that, despite the existence of localized hot spots within the nose, nasal deposition efficiency averaged over the entire nose is lower in hu- mans or nonhuman primates than in rats, leading to greater penetration of in- haled formaldehyde to the lower respiratory tract. That is supported by DNA– protein cross-links studies that suggest higher cellular delivery of inhaled for- maldehyde to the trachea and mainstream bronchi in nonhuman primates than in rats (Heck et al. 1989; Casanova et al. 1991). Unlike the obligate nose-breathing rodent, humans are capable of mouth breathing; this would greatly increase the delivery of inhaled formaldehyde to the lower airways. The airway epithelium is metabolically active. Of relevance to formalde- hyde disposition within nasal tissues is the presence of ADH5/FDH. The meta- bolic pathways offer an effective clearance mechanism for formaldehyde. Only formaldehyde that escapes metabolism is available for binding to tissue macro- molecules or potentially available for absorption into the blood. Like all meta- bolic pathways, formaldehyde metabolism demonstrates saturation kinetics. As saturation occurs, the likelihood of reaction of formaldehyde with tissue macro- molecules or of penetration of formaldehyde to deeper tissues increases. On the basis of modeling efforts and DNA–protein cross-link assessments, saturation kinetics may occur at concentrations above 2 ppm in the rodent nose. Specifical- ly, a nonlinear relationship between inspired concentration and DNA–protein cross-links in the nose is observed at exposure concentrations of 6 ppm or high- er, greatly exceeding what would be expected for a linear increase from the DNA–protein cross-links observed at concentrations of 2 ppm or lower (NTP 2010; NRC 2011). Absorption into Blood The disposition of formaldehyde in airway tissues and distribution throughout the body are important for understanding the potential for tissue inju- ry in airways or distant tissues. As previously noted, formaldehyde reacts readily and reversibly with sulfhydryl and amine moieties. Formaldehyde reacts revers- 

Independent Assessment of Formaldehyde 133 ibly with water to form methanediol, with the equilibrium strongly favoring me- thanediol. As outlined by Georgieva et al. (2003), it is not likely that the dissoci- ation of methanediol to form formaldehyde is rate-limiting (in contrast with the reaction with macromolecules), so this process is not critical for determining formaldehyde disposition in nasal tissues (NRC 2011). Because formaldehyde reactions are reversible, it is possible that an individual formaldehyde molecule, if it is not metabolically degraded, may shuttle from one binding site to another. Therefore, an individual endogenous formaldehyde molecule could be distribut- ed away from its site of formation, and an individual exogenous formaldehyde molecule could be distributed to tissues away from its site of first contact. That would occur only if the formaldehyde molecule escaped metabolic transfor- mation. Because ADH5/FDH is ubiquitously expressed, including expression in red blood cells, the likelihood of metabolic transformation is high, and this low- ers the likelihood of penetration to distant tissues through the bloodstream. Anatomic features of the airways are highly relevant to the potential for absorption into the blood and systemic distribution of formaldehyde (NRC 2011). The air–blood barrier of the nose and large tracheobronchial airways con- sists of a mucous lining layer overlying a pseudostratified columnar mucociliary epithelium. Residing below the basement membrane, the submucosal space of the nasal airways is highly vascularized. In the nose, a superficial capillary layer is present just below the basement membrane (Figure 3-3). This relationship is important for evaluation of formaldehyde disposition in the nose. Presumably, the target cells for tumorigenesis in the nasal airways are the basal cells that reside on the basement membrane. Immediately below the basement membrane are the vessels of the superficial capillary layer of the nose. The total epithelial thickness in the nose depends on the site but is generally less than 0.05 mm in rodents and humans (Schroeter et al. 2008). A similar structure exists with re- spect to the nasal associated lymphoid tissue (NALT), which resides just below the basement membrane (Figure 3-3). On the basis of mathematical modeling and estimation of the rates of reac- tion and metabolism, it has been estimated that formaldehyde would penetrate to some depth in nasal tissues (see Figure 3-4) (Georgieva et al. 2003). Specifical- ly, the modeling efforts suggest that the formaldehyde concentration at the depth of 0.05 mm (below the basement membrane) is greater than 50% of the concen- tration at the mucus–tissue interface. Thus, the concentration–tissue depth pro- file appears to have a shallow slope. Formaldehyde is clearly cytotoxic to the nasal epithelium, and the nasal epithelial basal cells are probably the target for nasal tumorigenesis; this indicates that reactive formaldehyde penetrates to this depth in the nose. Given the shallow slope of the concentration–tissue depth profile, it is likely that toxicologically significant concentrations of formalde- hyde penetrate somewhat deeper to the superficial capillary layer of the nose, inasmuch as these capillaries are adjacent to the basement membrane and basal 

134 Review R of the Fo ormaldehyde Asssessment in the N NTP 12th Repoort on Carcinogeens FIGUR RE 3-3 Schematiic representation n of the structuree of the nasal muucosa of the resppir- atory eppithelium and follicle-associate fo d epithelium. F For both epitheliia, a concentratiion gradientt for exogenous formaldehyde during d inhalationn exposure will exist with conceen- trations at the superficial layer (closestt to the airstream m) being higher than concentratiion in deepeer layers. As outlined in the texxt, this gradient iis due to the reaaction of formaldde- hyde wiith tissue substrates or metaboliism via ADH5/F FDH. It is worthh noting that baasal cells, a target for formaaldehyde-induced carcinogenesiss, lie immediateely above the baase- ment membrane m and caapillaries and nasal associated lyymphoid tissue ((NALT) lie imm me- diately below b the basem ment membrane. Source: NRC 20011, p. 32. cells (ssee above). Thu us, at sufficien nt airborne conccentrations, bioologically signnif- icant concentrations c of formaldehy yde may be prresent in the nnasal submucoosa and cap pillary bed. It should be reccognized, how wever, that the presence of fo for- maldeh hyde in the nassal submucosa and capillary bbed does not ittself indicate thhat biologiically significaant concentratio ons of formalddehyde penetraate via the bloood- stream to distant tissu ues. A toxicok kinetic approacch could be foormulated to essti- mate th he exposure co oncentrations th hat would be reequired to raisse systemic bloood formald dehyde substaantially above endogenous concentrationns. To the com m- mittee’s knowledge, thatt has not beeen performed. 

Indepen ndent Assessmeent of Formaldeehyde 1135 FIGUR RE 3-4 Model-baased estimates of o exogenous forrmaldehyde conncentration in naasal tissues during d inhalation n exposure to 6 ppm p formaldehyyde. Tissue conccentrations increaase quickly from 0.1 to 0.5 minutes after th he onset of expoosure as a quasii-steady state is es- tablisheed. Readily appaarent is the pred diction that the formaldehyde concentration aat a depth off 50 μm, measurred from the muccus:tissue interfaface, is fairly sim milar to the conceen- tration at a the interface itself. Source: Georgieva G et al. 2003. Reprintedd with permissioon; copyright 2003, Inhalattion Toxicology. Distribution D off Inhaled Form maldehyde The T nose receiv ves about 1% of cardiac outtput, and mathhematical models suggestt that about on ne-third of nasaal circulation ((0.33% of totaal cardiac outpuut) may peerfuse the supeerficial capillary layer (Gloeede et al. 20111). Venous bloood from thhe nose is ultimmately mixed withw the system mic venous bloood. On the baasis of relattive perfusion rates, blood from fr the entiree nose is diluteed by a factor of 100 (beecause the nosse receives 1% % of the cardiacc output) with systemic venoous blood; blood from th he superficial capillary c layer is diluted by a factor of aboout 300 wiith systemic veenous blood beefore distributiion to the bodyy. From that per- spectiv ve, it can be ap ppreciated that although the cconcentration oof an inhaled xxe- nobiotiic in the nasall capillary bloo od may be higgh, its concenttration is greaatly reduced d (by a factor of 100–300) as a blood from the nose mixees with system mic venouss blood. The un nderlying struccture of the larrge tracheobronnchial airwayss is similarr to that of thee nose; thus, th he relationshipps described abbove are qualiita- tively similar s for thee lower airway ys. The entire tracheobronchhial tree receivves about 1% 1 of cardiac output o (Gloedee et al. 2011). A As for the nosee, any xenobiootic absorbeed into the traccheobronchial circulation of the large airw ways is diluted by a factorr of about 100 as the venouss output from thhe airways mixxes with the syys- temic venous v blood. Although A it is theoretically t possible that ann individual exxogenous form mal- dehydee molecule cou uld be distributed away froom the portal of entry, mass- 

136 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens balance and kinetic arguments and experimental data strongly suggest that this does not occur to a great extent. Specifically, multiple studies that used different conceptual approaches, from simple mass-balance estimates (Heck and Casano- va 2004; Nielsen et al. 2013) to more detailed pharmacokinetic analysis (Franks 2005), universally support the conclusion that the amount of formaldehyde that is inhaled (at reasonable exposure concentrations) and absorbed into circulation is much lower than the endogenous amounts in circulation. Analytic studies did not observe a large increase in the total content of formaldehyde in blood or tissue above the endogenous concentrations during inhalation exposure (NTP 2010; NRC 2011). Published literature, relying on gas chromatographic and mass spectrometry techniques, indicates that blood formaldehyde (measured as free plus reversibly bound) is not increased in the rat, monkey, or human by inhalation exposure to formaldehyde (Heck et al. 1985; Casanova et al. 1988). Studies that use bound formaldehyde as a biomarker and that rely on dual- labeled formaldehyde also did not observe an increase in tissue formaldehyde during inhalation exposure in any tissue except the nose (Lu et al. 2011; Moeller et al. 2011; Edrissi et al. 2013b). Contrary to these findings are findings of for- maldehyde adducts in the blood of exposed individuals. One study reported in- creases in blood albumin–formaldehyde adducts in workers exposed to formal- dehyde (Pala et al. 2008); another reported increases in formaldehyde– hemoglobin adducts (Bono et al. 2006). Mass-balance arguments call the validi- ty of those findings into question (Nielsen et al. 2013), specifically that the amount of formaldehyde that would be required to raise albumin adducts or he- moglobin adducts to the levels reported is much greater than the amount that was inhaled. Recent well-designed studies have relied on dual labeled formaldehyde to measure formaldehyde–DNA adducts as a biomarker of delivered dose of exog- enous formaldehyde for comparison with endogenous concentrations (Lu et al. 2010a,b; Moeller et al. 2011). They indicate that endogenous formaldehyde– DNA adducts are ubiquitous throughout the body. Increased exogenous formal- dehyde–DNA adducts are observed in nasal tissues of rodents and nonhuman primates after inhalation exposure to formaldehyde, and this validates the sensi- tivity of the technique. High concentrations of exogenous formaldehyde–DNA adducts are not observed in distal tissues, including bone marrow, after formal- dehyde inhalation. Those experiments provide strong evidence that formalde- hyde exposure at the concentrations used (up to 15 ppm) does not result in sub- stantial delivery of exogenous formaldehyde to nonrespiratory tissues. The results have recently been confirmed by using formaldehyde–lysine adducts as biomarkers instead of formaldehyde–DNA adducts (Edrissi et al. 2013b). MECHANISMS OF CARCINOGENESIS The mechanisms of carcinogenesis of formaldehyde have been the subject of intense research for decades, and a large evidence base is available from 

Independent Assessment of Formaldehyde 137 which to draw inferences and conclusions. Despite the wealth of information available on a variety of test systems, from naked DNA (that is, DNA without any associated proteins) to experimental animals and exposed humans, it is still being debated what mechanistic events take place in tissues that have been sug- gested as targets for formaldehyde-associated carcinogenesis. Such debate is informed, in large part, by the considerations of formaldehyde toxicokinetics, inasmuch as formaldehyde is both a highly reactive molecule and an endoge- nously formed compound produced in the course of normal cellular metabolism. There is evidence that exogenously administered formaldehyde is responsible for noncancer and cancer effects at the portal of entry, such as nasal mucosa or other parts of the upper aerodigestive tract, depending on the mode of admin- istration and breathing patterns. It has been more controversial whether formal- dehyde itself or products of its biotransformation may reach tissues that do not come into direct contact with inhaled or ingested formaldehyde in experimental animals or humans, and a detailed discussion of the available evidence is pro- vided under the section “Toxicokinetics” above. There is general agreement that systemic delivery of formaldehyde is unlikely (NRC 2011), but it is also true that various toxicity phenotypes (for example, genotoxicity and mutagenicity in circulating blood cells, changes in the number of circulating cells and bone mar- row cells, and gene expression changes in blood) have been found in cells and tissues that are not in direct contact with exogenously administered formalde- hyde. That apparent inconsistency notwithstanding, the committee concurs with the conclusions drawn by the National Research Council Committee to Review EPA’s Draft IRIS Assessment of Formaldehyde (NRC 2011) that it is important to differentiate between systemic delivery of formaldehyde and systemic effects. It is possible that the “systemic delivery of formaldehyde is not a prerequisite for some of the reported systemic effects seen after formaldehyde exposure. Those effects may result from indirect modes of action associated with local effects, especially irritation, inflammation, and stress” (NRC 2011, p. 36). The present committee found that the most sensible characterization of the adverse health effects of formaldehyde and associated mechanisms is that pro- posed by NRC (2011). Specifically, a wide array of the adverse outcomes that have been associated with formaldehyde exposure are best classified into portal- of-entry and systemic categories, which are defined as follows: portal-of-entry effects are effects that arise from direct interaction of inhaled or ingested for- maldehyde with the affected cells or tissues; systemic effects are effects that occur beyond tissues or cells at the portal of entry. The committee notes, how- ever, that it is plausible that some of the systemic effects, most notably genotox- icity in circulating blood cells, may have resulted from the exposure of these cells at the portal-of-entry tissues (for example, lymphoid tissue in the nasal mucosa). As discussed in previous sections, the committee relied on the background document for formaldehyde, published reviews, and assessments performed by other authoritative bodies to ensure that relevant literature was captured up to the publication of the 12th RoC. It also considered literature, comments, and 

138 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens arguments provided during its open session and submitted by other sources dur- ing the duration of the study. The committee carried out its own literature search (see Appendix D) for publications that are pertinent to the major postulated modes of carcinogenic action of formaldehyde (genotoxicity, cell proliferation and apoptosis, and effects on the immune system). The committee’s exclusion criteria and detailed search strategies for studies related to genotoxicity and mu- tagenicity are presented in Box D-3 and for studies related to immune effects are presented in Box D-4. Literature trees were used to document identification and selection of the literature evidence (Figures D-3 and D-4). The general question that the committee addressed was, What is the evidence that the following mechanistic events—genotoxicity and mutagenicity or effects on the hematolog- ic system—are part of the overall mode of action of formaldehyde-associated carcinogenicity? The outcomes of the searches and the evidence available in the background document for formaldehyde (NTP 2010) were evaluated together and are detailed below. The committee notes that because of the limitations of time and resources several of the mechanisms that have been proposed by NTP (2011) to explain the carcinogenicity of formaldehyde (such as cytotoxicity followed by compen- satory proliferation and oxidative stress) have not been evaluated by conducting new literature searches. In the course of the review of the substance profile for formaldehyde in the NTP 12th RoC (see Chapter 2), the committee found that the mechanism of cytotoxicity followed by compensatory cell proliferation is a well-established portal-of-entry mechanism that is not controversial. On the con- trary, oxidative stress is a mechanistic event that has not been addressed in detail and on which the evidence base is too small to draw firm conclusions. The committee focused its attention on the mechanistic evidence that is related to genotoxicity and mutagenicity, hematologic effects, and data from toxicoge- nomic studies, which reflects broad biologic responses and is thus informative as both the overall effect and specific pathways that may be perturbed by exposure to formaldehyde. The RoC does not present quantitative assessments of risks of cancer as- sociated with the substances listed. Therefore, the committee did not explicitly take into consideration the issue of the dose or concentration of formaldehyde that was applied or evaluated in each study. The background document for for- maldehyde contains extensive information on the doses and concentrations used in various studies, and, where it is available, the committee notes dose- dependent and time-dependent trends in the new studies that have been pub- lished since June 10, 2011. Finally, the committee notes that although the mode of action of a chemi- cal substance is an important component of decision-making to protect human health, the guidelines established by various national and international agencies that conduct such assessments differ in how such information is gathered, pre- sented, and evaluated (Box 3-1). The guidance documents of IARC, the US En- vironmental Protection Agency (EPA), and the International Programme on Chemical Safety (IPCS) are informative, but the committee’s charge (see Ap- 

Independent Assessment of Formaldehyde 139 pendix B) was to integrate the level-of-evidence conclusions and to consider all relevant information in accordance with RoC listing criteria. In that respect, for each listed substance, the RoC includes studies of genotoxicity and of biologic mechanisms. The listing criteria are used to guide the evaluation of the human, animal, and mechanistic evidence. The listing criteria specifically state that “da- ta derived from the study of tissues or cells from humans exposed to the sub- stance in question, which can be useful for evaluating whether a relevant cancer mechanism is operating in humans” (NTP 2010, p. iv), constitute one of the lines of evidence used to support whether there is sufficient or limited evidence of carcinogenicity from studies in humans. BOX 3-1 Guidance from Various Agencies on the Use of Mechanistic and Other Relevant Data The IARC Monographs Program operates under the general guidance of a preamble, which specifies that a working group is to consider mechanistic and other relevant data because they “may provide evidence of carcinogenici- ty and also help in assessing the relevance and importance of findings of cancer in animals and in humans” (IARC 2006b, p. 15). The preamble out- lines “scientific principles, rather than a specification of working procedures” (p. 1), for the experts who participate in the development of each monograph. It notes that “the procedures through which a Working Group implements these principles are not specified in detail” (p. 1). The EPA Guidelines for Carcinogen Risk Assessment (EPA 2005) state that the agency’s assessments should discuss the available information on the modes of action and associated key events of chemicals under evalua- tion. Specifically, the assessments aim to address several questions pertain- ing to the extent and quality of the evidence on the hypothesized mode of action. The questions include sufficiency of supporting information from test animals, relevance to humans, and any information that may suggest that particular populations or life stages can be especially susceptible to the hy- pothesized mode of action. It is noted, however, that “in the absence of suffi- ciently, scientifically justifiable mode of action information, EPA generally takes public health-protective, default positions regarding the interpretation of toxicologic and epidemiologic data” (EPA 2005, p. 1-10). IPCS developed a mode-of-action relevance framework for the analysis of mechanistic evidence on chemical carcinogens in experimental animals and its relevance to humans (Boobis et al. 2008). The framework calls for determining whether the weight of evidence based on experimental observa- tions is sufficient to establish a hypothesized mode of action. A series of key events causally related to the toxic effect are then identified using an ap- proach based on the Bradford Hill criteria and compared qualitatively and quantitatively between experimental animals and humans. 

140 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Genotoxicity and Mutagenicity The data available to examine the potential role of genotoxicity and muta- genicity of formaldehyde are extensive. Those effects are likely to be relevant for all cancer sites that have been associated with formaldehyde exposure. Near- ly all aspects of genotoxicity and mutagenicity have been studied with formal- dehyde, so assertive conclusions can be drawn from the available evidence. The committee collated the evidence on all the mechanistic events that make up the genotoxic mode of action into separate tables (see Appendix E). In each table, the committee separated studies by type of the model system, includ- ing a clear division between the portal-of-entry and systemic effects in in vivo studies. Publications that have evaluated a particular mechanistic event and found evidence supporting or refuting each were included. In addition, a sum- mary table (Table 3-10) was constructed to present the totality of the evidence available on each mechanistic event in each experimental model system. Overall, the evidence on genotoxicity and mutagenicity of formaldehyde resulted from studies that evaluated DNA adducts (Table E-1), DNA–DNA cross-links (Table E-2) and DNA–protein cross-links (Table E-3), DNA strand breaks (Table E-4), mutations (Table E-5), sister-chromatid exchanges (Table E- 6), micronuclei (Table E-7), and chromosomal aberrations (Table E-8). Several published studies have also examined the DNA-repair responses to formalde- hyde-induced DNA damage. Owing to the paucity of data, the model systems used in these studies, and the scope of the present committee’s charge, that in- formation was not included in the evaluation. Similarly, the committee found that although some reports examined the possible role of genetic polymorphisms in the genotoxic potential of formaldehyde or ensuing adverse outcomes, the overall database was not robust and did not provide strong evidence that human variability factors (genetic polymorphisms) may be critical for drawing conclu- sions. All studies included in Appendix E were examined in full text (including translations, where applicable) by at least two committee members, who inde- pendently determined whether a given study observed an important effect or lack thereof with respect to the phenotype named in each table. Studies were categorized as positive if a statistically significant effect was observed. Studies were categorized as negative if the results reported an absence of a particular effect (that is, no statistically significant difference from the appropriate control group). Although the committee members exercised their scientific judgment in categorizing studies and determining their relevance to each phenotype, the committee did not perform a formal quality assessment of each individual study, whether it was categorized as positive or negative. The committee members also did not make judgments about the study design or methodology, recognizing that all the studies had been subjected to some form of peer review before publi- cation. 

TABLE 3-10 Summary of Published Studies on the Genotoxic and Mutagenic Effects of Formaldehyde in Test Systems and Organisms1 DNA Adducts DDX DPX Strand breaks Mutations SCE MN CA Cellfree systems + (7/0) + (3/0) + (3/0) Nonmammalian model organisms + (6/0) +** Mammalian in vitro Rodent + (1/0) + (14/1) + (6/2) +/- (3/2) + (9/0) + (4/0) + (5/0) Human + (2/0) + (23/0) + (8/0) + (6/0) + (6/0) + (4/0) + (6/2) Mammalian in vivo: Rodent + (2/0) + (8/0) - (0/1) +/- (1/1) -/+ (1/2) + (1/0) portal-of-entry effects Primate + (1/0) + (2/0) Human +/- (11/3) Mammalian in vivo: Rodent - (0/1) +/- (2/2) +/- (2/1) + (1/0) - (0/2) -/+ (4/5) - (2/5) systemic* effects Primate - (0/1) - (0/2) Human +# (1/0) + (3/0) + (9/2) -/+ (7/9) + (18/3) +/- (11/5) 1 Total numbers of studies demonstrating effect or lack thereof are indicated in parentheses. See Appendix E for data that support this summary table: DNA adducts (Table E-1), DNA–DNA cross-links (Table E-2), DNA–protein cross-links (Table E-3), DNA strand breaks (Table E-4), mutations (Table E-5), sister-chromatid exchanges (Table E-6), micronuclei (Table E-7), and chromosomal aberrations (Table E-8). +: all or most of the studies indicate the effect. +/-: most of the studies indicate the effect, although many show lack thereof. -/+: most of the studies indicate lack of the effect, although many positive studies have been published. -: all or most of the studies indicate lack of the effect. *The committee acknowledges that although most investigators consider the effects on circulating-blood mononucleated cells as systemic be- cause cells for the analyses were collected from the systemic circulation, it is plausible that the cells had been exposed to formaldehyde in the nose through lymphoid tissue in the mucosa. **The results are overwhelmingly positive for point mutations and overwhelmingly negative for frame-shift mutations. # M1G adduct has been postulated to be the result of secondary DNA damage caused by formaldehyde-associated oxidative stress. Abbreviations: DNA, deoxyribonucleic acid; DDX, DNA–DNA cross-links; DPX, DNA–protein cross-links; SCE, sister-chromatid exchanges; MN, muconuclei; CA, chromosomal aberrations. Source: Committee generated. 141 

142 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens The committee’s work was informed by the Bradford Hill criteria (Hill 1965) for determining causality between exposure to formaldehyde and findings of genotoxicity and mutagenicity. Although those criteria have been proposed for determinations of causality in epidemiologic studies, they do not all apply to the evaluation of the mechanistic evidence. As noted in EPA guidelines (EPA 2005, p. 2-13), “one . . . cannot simply count up the numbers of studies reporting statistically significant results or statistically non-significant results for carcino- genesis and related MOAs [modes of action] and reach credible conclusions about the relative strength of the evidence and the likelihood of causality.” Thus, the committee, upon systematizing the available mechanistic evidence pertain- ing to the genotoxicity and mutagenicity of formaldehyde into tables, appraised the evidence by using the general guidance of the “causal criteria” (EPA 2005) to determine its overall strength for drawing conclusions about causality for each of the mechanistic events identified in the tables. Because the body of evi- dence on genotoxicity and mutagenicity of formaldehyde is very large, the mechanistic synthesis does not contain many citations to the individual publica- tions; all the evidence is presented in multiple tables. Owing to the challenge of establishing whether and how formaldehyde can exert point-of-entry and systemic effects, the committee chose to evaluate cau- sality for each of the mechanistic events in three broad categories: 1) Effects on the naked DNA or on the DNA of nonmammalian organisms or mammalian cells in vitro. 2) Effects observed on the portal-of-entry tissues of animals or humans exposed to formaldehyde. 3) Systemic effects in animals or humans exposed to formaldehyde. The latter two are most relevant to the determination of the cancer-hazard classification according to the RoC listing criteria, which call for conclusions to be based on the information “derived from the study of tissues or cells from hu- mans exposed to the substance in question” (NTP 2011, p. 198). Again, the committee acknowledges that although most investigators consider the effects on circulating blood mononucleated cells to be systemic because cells for the analyses were collected from the systemic circulation, it is plausible that these cells have been exposed to formaldehyde in the nose through lymphoid tissue in the mucosa. Effects of Formaldehyde on Naked DNA or on DNA of Nonmammalian Organisms or Mammalian Cells in Vitro The totality of the evidence overwhelmingly shows that when formalde- hyde is added to naked DNA or nonmammalian organisms or mammalian cells are incubated in the presence of formaldehyde, DNA adducts (Table E-1), cross- links (Tables E-2, E-3), strand breaks (Table E-4), mutations (Table E-5), and 

Independent Assessment of Formaldehyde 143 clastogenic damage (Tables E-6, E-7, and E-8) are found. Studies were conduct- ed in different types of model systems and have produced consistent results. The evidence of genotoxicity and mutagenicity of formaldehyde comes from studies where different model systems were tested and various molecular techniques were used to evaluate the effects. Because all studies evaluated in this category used formaldehyde, specificity of the effects being caused by for- maldehyde has been firmly established. In addition, many studies used appropri- ate positive and negative controls, and this further strengthens the specificity of the association. The temporal relationship of the observed association is clear in that the studies evaluated genotoxic and mutagenic effects after DNA or cells came into contact with formaldehyde. Dose–response relationships between genotoxic and mutagenic effects and formaldehyde were observed in studies that had appropriate designs. For example, DNA–protein cross-links were formed in a concentration–response manner in human lymphoblastoid cell lines (Ren et al. 2013), epithelium-like human lung cells (Speit et al. 2010), and isolated human lymphocytes (Neuss et al. 2010a,b). Similar observations were made in whole- blood cultures for sister-chromatid exchanges, micronuclei, and chromosomal aberrations (Schmid and Speit 2007; Ren et al. 2013). The committee concludes that the genotoxic and mutagenic mode of ac- tion of formaldehyde in studies of naked DNA, studies of DNA from nonmam- malian organisms, and studies of mammalian cells in vitro is consistent, strong, and specific to the formaldehyde exposure. Both temporal and dose–response relationships have been established. This mechanistic event is relevant to human cells because all the genotoxic effects observed in studies of naked DNA, non- mammalian model organisms, or cells from rodents have been also observed in human cells, either established cell lines or primary cells. Effects on the Portal-of-Entry Tissues of Animals or Humans Exposed to Formaldehyde Because various studies reviewed by the committee may have used differ- ent routes of administration of formaldehyde and because of the differences in breathing patterns among rodents and humans, the committee considered the following anatomic regions as points of entry: nasal passages, oral cavity and upper aerodigestive tract, and forestomach (in gavage studies). The committee identified no studies that evaluated DNA–DNA cross-links or sister-chromatid exchanges in exposed rodents or humans at the portal of entry, so these mecha- nistic events were not considered in this section. Most of the evidence of genotoxic and mutagenic effects at the portal of entry, depending on the end point studied, is from studies of laboratory rodents and exposed humans. Several reports evaluated pertinent mechanistic events in nonhuman primates. Studies of DNA adducts (Table E-1), even though the data- base is not large, showed that formaldehyde-induced DNA damage is consistent- ly observed in both rodents (Lu et al. 2010a, 2011) and nonhuman primates 

144 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens (Moeller et al. 2011). Similarly, consistent evidence from a large number of studies of rodents and nonhuman primates demonstrates formation of DNA– protein cross-links (Table E-3). Positive and negative findings, albeit from a small number of studies of formaldehyde exposure of rodents, are equally divid- ed for strand breaks (Table E-4), mutations (Table E-5), micronuclei (Table E- 7), and chromosomal aberrations (Table E-8). In humans exposed to formalde- hyde, formation of micronuclei was examined in cells at the portal of entry, and 11 of 14 studies demonstrated a positive association (Table E-7). Overall, the findings are consistent with genotoxic and mutagenic effects of formaldehyde observed in naked DNA, in the DNA of nonmammalian organisms, and in mammalian cells in vitro. Evidence of genotoxicity and mutagenicity of formaldehyde in exposed humans is strong, even though several studies reported no induction of micronu- clei. The positive observations were made in studies of diverse groups of sub- jects that were exposed to formaldehyde. Various assays have been used to evaluate the mechanistic events, and statistical significance of the effects was established in the positive studies. In rodent and nonhuman primate studies, formaldehyde exposures were well documented (for example, purified reagent-grade formaldehyde was used). Furthermore, several studies of DNA damage have used 13C-labeled formalde- hyde (Lu et al. 2010a, 2011; Moeller et al. 2011), which shows that the genotox- ic effects of formaldehyde occur at the portal of entry. In human studies, many investigators established the association between formaldehyde and these mech- anistic events through exposure monitoring, albeit most of the studies were of occupational cohorts and the presence of other agents cannot be excluded. Some of the studies that found no evidence of micronuclei in portal-of-entry tissues from humans (Speit et al. 2007; Zeller et al. 2011a) is evidence that questions the association in controlled exposures of volunteers to formaldehyde. Studies of rodents and nonhuman primates provide strong evidence for a temporal relationship of the observed association because the genotoxic and mutagenic effects were observed after exposure to formaldehyde. In many hu- man studies, temporality was established by collecting samples before and after exposure in the workplace. Studies of rodents and nonhuman primates provide strong evidence of concentration–response relationships in the genotoxicity of formaldehyde at the portal of entry (Lu et al. 2010a, 2011; Moeller et al. 2011). The concentrations of formaldehyde used in the studies (around 1–10 ppm) are comparable with or an order of magnitude higher than those documented in human occupational exposures. The shape of the concentration–response curve of several biomarkers of genotoxicity in the portal-of-entry tissues in rodents is nearly identical with that for tumorigenesis in the noses of rodents (Swenberg et al. 2013). The committee concludes that the genotoxic and mutagenic mode of ac- tion of formaldehyde in the portal-of-entry tissues of animals or humans ex- posed to formaldehyde is supported by the experimental evidence. Several nega- tive studies notwithstanding, the evidence is consistent, strong, and specific with 

Independent Assessment of Formaldehyde 145 respect to an association following exposure to formaldehyde. Both temporal and exposure–response relationships have been established, most strongly in the studies of experimental animals (rodents and nonhuman primates). This mode of action is relevant to humans because statistically significant increases in the number or frequency of micronuclei, known biomarkers of clastogenesis, have been observed in most, but not all, of the studies of portal-of-entry tissues from humans exposed to formaldehyde. Systemic Effects in Animals or Humans Exposed to Formaldehyde Systemic effects are effects that occur outside cells or tissues that come in- to direct contact with exogenous formaldehyde. Most studies in the systemic- effects category examined genotoxic and mutagenic effects of formaldehyde in circulating blood mononucleated cells unless stated otherwise. The committee acknowledges, however, that although most investigators consider the effects on circulating blood mononucleated cells as systemic because cells for the analyses were collected from the systemic circulation, it is also plausible that these cells were exposed to formaldehyde in the nose through lymphoid tissue in the muco- sa. Most of the experimental evidence that is available for drawing conclu- sions about systemic genotoxic and mutagenic effects of formaldehyde comes from studies in humans exposed to formaldehyde, mostly in occupational set- tings. Fewer experimental-animal (for example, rodent) studies have been con- ducted, and only two studies of nonhuman primates examined some of the mechanistic events in question. Overall, the database pertaining to this question is most consistent in exposed humans in whom formaldehyde exposure- associated DNA–protein cross-links (Table E-3), strand breaks (Table E-4), mi- cronuclei (Table E-7), and chromosomal aberrations (Table E-8) were detected in most of the studies. Data on sister-chromatid exchange formation in response to exposure to formaldehyde in humans are almost equally divided for and against (Table E-6). In studies in rodents, there is little positive evidence of clas- togenic effects of formaldehyde on circulating blood cells but some evidence of strand breaks and mutations. Studies of nonhuman primates found no evidence of the increased formation of DNA adducts in bone marrow after exogenous administration of 13C-labeled formaldehyde (Moeller et al. 2011) or the presence of DNA–protein cross-links in the most distal regions (lung parenchyma) of the respiratory tract (Casanova et al. 1991). Evidence of genotoxicity and mutagenicity of formaldehyde in exposed humans is strong because various assays were used to evaluate these effects, data come from a number of independent laboratories around the world, and the positive studies were conducted on humans exposed in a variety of occupational settings (for example, pathologists, embalmers, and anatomy students). The neg- ative human studies also contribute important information in that the diversity of 

146 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens the study designs and occupational and laboratory-based exposures is apprecia- ble. The studies of rodents and nonhuman primates used controlled exposures to purified reagent-grade formaldehyde, and some studies even used controlled exposures to 13C-labeled formaldehyde, which increases the specificity of the negative observations. Human studies were largely in occupational exposure scenarios in which formaldehyde was the primary—not the only—agent and other chemical (for example, solvent) or physical (for example, wood-dust) ex- posures were possible. Formaldehyde-associated DNA–protein cross-links were found in three human studies (Table E-3); however, most of the end points that were evaluated in the positive studies, such as strand breaks (Table E-4) and clastogenic effects (Tables E-6, E-7, and E-8), are difficult to attribute specifi- cally to formaldehyde. Thus, the specificity of the observed positive associations is somewhat uncertain. In many—not all—positive human studies, a temporal relationship was es- tablished by collecting samples before and after exposure in the workplace (Lin et al. 2013) or by considering the extent of employment in an occupation in which formaldehyde exposure is very likely (Viegas et al. 2010; Ladeira et al. 2011; Souza and Devi 2014). Some studies of rodents and nonhuman primates provide strong evidence of lack of a dose–response relationship in the formation of exogenous formaldehyde-induced DNA adducts (Lu et al. 2010a, 2011; Moeller et al. 2011). Recent studies that evaluated DNA–protein cross-links, however, show dose-dependent increases in this biomarker of genotoxicity in tissues (bone marrow, liver, spleen, and testes) that are not in direct contact with inhaled formaldehyde (Ye et al. 2013). Some of the positive human studies found a relationship between the clastogenic effects of formaldehyde and expo- sure duration (Viegas et al. 2010; Ladeira et al. 2011; Souza and Devi 2014) or dose (Jiang et al. 2010). The committee concludes that the systemic genotoxic and mutagenic mode of action of formaldehyde is sufficiently supported by the evidence from studies of humans exposed to formaldehyde. The committee acknowledges that reporting bias against negative results could be a limitation of its approach to reviewing the mechanistic evidence (NRC 2014); however, that limitation does not detract from the conclusion that formaldehyde can induce systemic genotox- ic changes. The evidence is consistent and strong, albeit it is difficult to establish unequivocal specificity of the effects following exposure to formaldehyde in the human studies. Whereas the committee recognizes some inconsistencies among data in experimental animals and humans and among genotoxicity biomarkers, this variability does not undermine the committee’s conclusion. Both temporal and exposure–response relationships have been demonstrated in studies of hu- mans exposed to formaldehyde. This mode of action is relevant to humans be- cause most of the positive evidence comes from studies of humans exposed to formaldehyde. The data do not exclude the possibility of other modes of action but strongly suggest a causal relationship between exposure to formaldehyde and human cancer. 

Independent Assessment of Formaldehyde 147 Hematologic Effects The systemic effects of formaldehyde exposure and the association with hematopoietic malignancies have been a source of debate, and there has been much interest in the hematologic effects of formaldehyde exposure. Several re- cent studies have evaluated the effects of formaldehyde on circulating hemato- poietic cells, and a number of them were published after the release of the NTP 12th RoC. In this section, the committee focuses on changes in hematopoietic- cell number or function—that is, “hematologic effects”. It did not consider genotoxicity studies and studies of altered gene expression because they are covered in other sections of this chapter. In addition, given that few studies have been designed to address the clinical significance of hematologic effects, to ad- dress the mechanisms by which hematologic effects may arise after exposure, or to address mechanisms that contribute to adverse health effects (including can- cer), these topics were not considered by the committee. The focus of this sec- tion is on evaluation of recently available evidence related to the hematologic effects of formaldehyde in human and animal exposure studies and evidence that is available from in vitro studies. Hematologic Effects in Humans Exposed to Formaldehyde Hematologic effects of formaldehyde include effects on cells of the hema- topoietic system that are circulating in the peripheral blood, are present in hema- tologic tissues (such as bone marrow, lymph nodes, and spleen), or are present in other tissues, whether at the portal of entry or not. The available data primari- ly reflect the hematologic consequences of exposure to inhaled formaldehyde in humans without addressing the mechanism or health consequences of the find- ings. Many studies have addressed the hematologic effects of exposure to for- maldehyde in humans (Tables 3-11 and 3-12). Six studies that examined inhala- tion exposures of formaldehyde in humans reported decreases in overall white blood cells, and three reported decreases in red cells and platelets. Studies have also reported many other hematologic effects, such as increases in monocytes, eosinophils, and some T-cell subsets and decreases in neutrophils and T-cell function. It should be noted that several studies have reported contrasting find- ings in the same hematologic characteristic, such as increases vs decreases in total lymphocyte concentration and T-, B-, and NK-cell subsets. Given that for- maldehyde exposure concentrations, durations, and sources varied greatly among studies, it is difficult to reconcile those results. However, taken as a whole, the body of evidence demonstrates consistently that exposure of humans to inhaled formaldehyde is associated with an array of hematologic effects. 

TABLE 3-11 Recent Studies of Hematologic Effects of Formaldehydea 148 Main Hematologic Findings Model Subjects Exposure Sample (Excluding Genotoxicity)b Reference Inhalation Workers (43 Factory workers exposed Peripheral blood tested for Extension of Zhang et al. (2010) using the same Hosgood et al. exposure in formaldehyde- to formaldehyde– lymphocyte subsets subjects and reporting additional assays. Total NK- 2013 humans exposed, 51 age- melamine resins cell and T-cell counts were 24% and 16% lower, and sex-matched compared with respectively, in exposed workers. Decreased counts in controls) workers without exposed workers were observed for CD8+ T cells, formaldehyde exposure; CD8+ effector memory T cells, and regulatory T cells. mean formaldehyde B-cell numbers did not differ significantly. exposure 1.28 (0.63– 2.51) ppm vs <0.03 ppm Workers (43 Factory workers exposed Peripheral blood measures Reanalysis of Zhang et al. (2010) data. Differences Gentry et al. 2013 formaldehyde- to formaldehyde– (complete blood count and in blood measures when examined in context of exposed, 51 age- melamine resins; WBC differential) population averages for Chinese and general and sex-matched exposures same as populations and when controlled for potential controls) Hosgood et al. (2013) confounders (for example, suspected thalassemia trait) suggest that effects attributed to formaldehyde are not clinically significant. Concerns were raised regarding relevance of CFU-GM assays to AML stem-cell biology. Male workers Factory workers in Blood samples measured Percentage of lymphocytes was increased 13% in Aydin et al. 2013 (46 formaldehyde- two medium-density for lymphocyte subsets, formaldehyde-exposed workers. Absolute numbers exposed, 46 fiberboard-producing immunoglobulins, and percentages of T cells (17% and 6%, respectively) controls) plants; measured complement proteins, and NK cells (48% and 34%, respectively) were higher, formaldehyde levels; and TNFα concentrations IgG (23%) and IgM (27%) in exposed workers were 8-hour TWA = 0.20 ± statistically lower, TNFα was significantly higher 0.06 ppm (0.10–0.33 (308%). No significant differences in white blood cell, ppm) erythrocytes, hemoglobin, neutrophils, or monocytes were observed. Workers (35 Pathology anatomy Blood sample measured for Overall, 30% decrease in percentage of B cells Costa et al. 2013 formaldehyde- workers with >1 year lymphocyte subsets (T, B, (CD19+) found in formaldehyde-exposed workers exposed, 35 exposure in four and NK cells) and MN, compared with controls (p < 0.05). Decreased B-cell controls) hospitals in Portugal SCE, and TCR mutations percentage was significant in multivariate analysis 

and nonexposed (including sex, smoking, and age) (p = 0.014). T cells administrative workers in (CD3+) and helper T cells (CD3+/CD4+) increased same facilities; 8-hour when analyzed by formaldehyde exposure (p = 0.002 TWA mean exposure = and 0.006, respectively) and in multivariate analysis 0.36 ± 0.03 ppm (range (p = 0.024 and 0.037, respectively). NK cells 0.23–0.69 ppm) (CD16+/CD56+) decreased on basis of individual exposure levels (p < 0.001) and in multivariate analysis (p < 0.001). Female workers Workers, formaldehyde- Blood samples measured Apoptotic cells after PHA stimulation were mean of Jakab et al. 2010 (37 formaldehyde- exposed women in four for apoptosis, proliferation, 77% higher in formaldehyde-only exposed workers exposed, 37 controls) pathology departments HPRT function, UV-induced compared with controls. Lectin labeling index and in Hungary; 8-hour TWA DNA synthesis, CA, SCE, variant frequency, measures of HPRT function, were mean exposure = 0.9 and T-cell activation marker significantly increased and decreased, respectively, in mg/m3 measured in three CD71 after PHA stimulation formaldehyde-exposed workers. CD71 expression on of four sites; 16 subjects in vitro T cells and BrdU incorporation were not significantly identified as having changed. exposure to organic solvents in addition to formaldehyde were analyzed separately Inhalation Male Balb/c mice Inhaled formaldehyde at Blood measured for Formaldehyde exposure led to a significant decrease Zhang et al. 2013 exposure in 0, 0.5, 3 mg/m3, 8 complete blood count (cell (p<0.05) in white blood cells, red blood cells, and animals hours/day, 5 days/week types and hemoglobin), BM lymphocytes after exposure to 0.5 mg/m3 of (5 days on, 2 days off), for histology, ROS, GSH, formaldehyde (43%, 7%, and 39%, respectively), and 13 days cytochrome 1A1, GSTT1, 3.0 mg/m3 of formaldehyde (52%, 27%, and 43% NFkB, TNFα, and IL-1b respectively). Platelet counts were significantly increased (p<0.05) after exposure to formaldehyde at 0.5 mg/m3 (109%) and 3.0 mg/m (67%). Monocytes and granulocytes were not significantly changed. At a formaldehyde exposure of 0.5 mg/m3 and 3.0 mg/m3, ROS levels in BM increased by 31% and 102%, respectively; CYP1A1 increased by 8% and 37%, respectively; and GSTT1 decreased by 0% and13%, (Continued) 149 

150 TABLE 3-11 Continued Main Hematologic Findings Model Subjects Exposure Sample (Excluding Genotoxicity)b Reference Inhalation respectively. At 3.0 mg/m3 of formaldehyde, NFkB exposure in increased by 34%, and inflammatory cytokines were animals increased—TNFα by 42% and IL-1b by 98%. Female C57BL/6 Inhaled formaldehyde at BM, lymph node, spleen, Formaldehyde-exposed mice showed 30% increase in Kim et al. 2013 0, 5, 10 ppm, 6 liver, and lung measured for percentage of T cells (CD3+), 38% increase in CD8+ T hours/day, 5 days/week, cell types and NK function cells, and 28% decrease in B cells (B220+) in spleen at 14 days of exposure 10 ppm, but absolute numbers were not significantly different. No change in percentage of CD4+ or CD8+ T cells in BM, lymph nodes, liver, or lung. Percentage of NK cells (NK1.1+) in lung was decreased in concentration-dependent manner (decrease of 19% at 5 ppm and 58% at 10 ppm) and returned nearly to normal in 2 weeks after last formaldehyde exposure. Absolute numbers of NK cells were reduced in lung, but total leukoctye numbers were not changed at 10 ppm. Total number of cells present in BAL was increased >20-fold in formaldehyde-exposed mice, but absolute number of NK cells was decreased by over 65%, as were Ly49 receptor expression levels on NK cells. Similarly, percentage and total NK cells and Ly49 expression were decreased in spleen in a time-dependent manner, but no change in total splenocytes was observed. IFNg, perforin, and CD122 were decreased in NK cells from lung and spleen of formaldehyde-exposed mice, and LPS-mediated increase in these proteins was inhibited after formaldehyde exposure in lung. NK cytolytic activity (chromium release assay) of splenic NK cells was decreased at 2–3 weeks of formaldehyde exposure. Decrease in NK-cell numbers (approximately 30%) and function were seen in tumor-bearing mice exposed to formaldehyde. Decreases in NK viability and differentiation in vitro were also observed. 

Outbred female Inhaled formaldehyde Blood measured for blood Of blood-cell types and hemoglobin, formaldehyde- Katsnelson et al. white rats at 12.8 ± 0.69 mg/m3, cell types, hemoglobin, MN, exposed rats had statistically significant differences 2013 4 hours/day, 5 and multiple serum proteins (p < 0.05) in percentage of lymphocytes (11% increase) days/week, 10 and amino acids and percentage of segmented neutrophils (31% weeks decrease). Female Wistar rats Inhaled formaldehyde, Blood and bone marrow Sham-control rats were part of a larger study of female Lino-dos-Santos- nebulized at 0.32%, 90 samples measured for cell sex hormone effects on formaldehyde-induced airway Franco et al. minutes/day for 3 subsets; BAL fluid inflammation. Formaldehyde exposure in these control 2011 consecutive days leukocytes rats showed a 111% increase in WBC, including mononuclear and neutrophil subsets in BAL fluid. Sham-control rats had 197% increase in WBC, but there was >70% decrease in BM cell numbers in formaldehyde-exposed rats. >19-fold increase in degranulated mast cells was seen in lungs of formaldehyde-exposed control rats. In vitro Primary expanded 0–150 mcM Cell growth and cell cycle Formaldehyde exposure suppressed in vitro human Ji et al. 2013 studies human erythroid formaldehyde in tissue distribution erythroid progenitor cell expansion in dose-dependent progenitor cells from culture manner. PBMCs Primary expanded 0–3,200 µM Morphology, viability, NK-cell viability, cytolytic activity, and perforin Li et al. 2013 human NK cells from formaldehyde in tissue apoptosis, cytotoxicity secretion were decreased above 800 micromolar. PBMCs culture examined at 10, (killing tumor-cell activity), 30, 60, and 120 minutes cytokine and cytolytic proteins, and secretion of NK cells were evaluated Primary mouse BM 0–200 mcM Viability (MTT assay) BM MSCs demonstrated cytotoxicity >75 micromolar. She et al. 2013 MSCs formaldehyde in tissue culture Human 0–200 mcM Viability (AnnexinV binding FANCD2-deficient lymphoblastoid cell line was Ren et al. 2013 lymphoblastoid cell formaldehyde for 24 and PI staining) statistically more sensitive to formaldehyde-induced lines hours in tissue culture cell death than FANCD2-expressing control. (Continued) 151 

152 TABLE 3-11 Continued Main Hematologic Findings Model Subjects Exposure Sample (Excluding Genotoxicity)b Reference Primary human 0–1.152 mg/mL Viability (trypan blue and Statistically significant decreases in viability seen at Pongsavee 2011 lymphocytes from formaldehyde after PHA MTT assay) formaldehyde concentrations above 0.036 mg/mL. 30 volunteers stimulation for 72 hours a The studies in this table were identified through the committee’s literature search. See Appendix D for more details of the search. b All reported findings are significant with p <0.05. Abbreviations: AML, acute myeloid leukemia; B, bursa-derived cells; BAL, bronchoalveolar lavage; BM, bone marrow; BrdU, bromodeoxyur- idine; CA, chromosomal aberrations; CD, cluster of differentiation; CFU-GM, colony-forming unit-granulocyte-macrophage; CYP1A1, cyto- chrome P450, family 1, subfamily A, polypeptide 1; DNA, deoxyribonucleic acid; FANCD2, fanconi anemia group D2 protein; GSTT1, gluta- thione s-transferase theta 1; HPRT, hypoxanthine-guanine phosphoribosyltransferase; GSH, glutathione; IFNg, interferon gamma; IgG, immunoglobulin G; IgM, immunoglobulin M; IL-1b, interleukin-1 beta; LPS, lipopolysaccharide; Ly49 - killer cell lectin-like receptor subfam- ily A; mg/m3, milligram per cubic meter; mg/mL, milligrams per milliliter; MN, micronucleus test; MSC, mesenchymal stem cell; MTT, me- thylthiazol tetrazolium; NFkB, nuclear factor kappa-light-chain-enhancer of activated B cells; NK, natural killer cells; PBMC, peripheral blood mononucleated cell; PHA, phytohemagglutinin; ppm, parts per million; ROS, reactive oxygen species; SCE, sister-chromatid exchange; T, thymus cells; TCR, T-cell receptors; TNFa, tumor necrosis factor alpha; TWA, time-weighted average; UV, ultraviolet; WBC, white blood cell count. Source: Committee generated. 

TABLE 3-12 Studies Grouped by Hematologic Effects Model Cell Type Hematologic Effectsa Reference Inhalation WBC ↓ Total WBC Qian et al. 1988; Kuo et al.1997; Tang and Zhang 2003; Cheng et al. 2004; exposure Tong et al. 2007; Zhang et al. 2010 in humans ↑ Percentage of lymphocytes Aydin et al. 2013 ↓ Total lymphocytes Zhang et al. 2010 ↓ CFU formation T cells ↓ Total T cells and CD8+ T cells Ying et al. 1999; Ye et al. 2005; Hosgood et al. 2013 ↓ CD4+ T cells Ying et al. 1999 ↑ CD4/CD8 ratio Ying et al. 1999; Ye et al. 2005 ↑ CD26+ activated T cells Madison et al. 1991 ↑ T cells Aydin et al. 2013; Costa et al. 2013 Impaired mitogen-induced proliferation of lymphocytes Vargova et al. 1992 ↑ PHA-induced apoptosis Jakab et al. 2010 NK cells ↓ NK cells Costa et al. 2013; Hosgood et al. 2013 ↑ NK cells Aydin et al. 2013 B cells ↑ B cells percentage Ying et al. 1999; Ye et al. 2005 ↓ B cell percentage Costa et al. 2013 ↑ autoantibodies and anti-FA-albumin conjugates Madison et al. 1991 ↑ IgM/IgA Qian et al. 1988 ↓ IgG/IgM Aydin et al. 2013 Erythrocytes ↓ erythrocyte count and hematocrit level Lyapina et al. 2004 ↓ hemoglobin level Yang 2007 ↑ MCV Zhang et al. 2010 Neutrophils ↓ spontaneous respiratory burst activity Lyapina et al. 2004 ↑ susceptibility to infection 153 (Continued) 

154 TABLE 3-12 Continued Model Cell Type Hematologic Effectsa Reference Monocytes ↑ monocytes in indoor FA+nitrogen dioxide exposure Erdei et al. 2003 Eosinophils ↑ eosinophils Qian et al. 1988 Platelets ↓ platelets Tong et al. 2007; Yang 2007; Zhang et al. 2010 Inhalation WBC ↓ WBC Brondeau et al. 1990; Zhang et al. 2013 exposure ↑ WBC Lino-dos-Santos-Franco et al. 2011 in animals ↓ lymphocytes Zhang et al. 2013 ↓ lymphocyte viability Pongsavee 2011 ↑percentage lymphocytes Kim et al. 2013 ↓ bone marrow cell numbers Lino-dos-Santos-Franco et al. 2011 ↑ bone marrow cell numbers Battelle 1981 T cells ↑ percentage of T cells and CD8+ T cells Kim et al. 2013 NK cells ↓ total and percentage of NK cells Kim et al. 2013 ↓ IFNg, perforin, and CD122 in NK cells. ↓ cytolytic activity and NK differentiation ex vivo B cells ↓ B cells Kim et al. 2013 Neutrophils ↓ segmented neutrophils Katsnelson et al. 2013 Erythrocytes ↓ erythrocytes Zhang et al. 2013 Platelets ↑platelets Zhang et al. 2013 In vitro T cells ↓ IFNg and IL-10 in stimulated human T cells Sasaki et al. 2009 studies B cells ↓ viability of human lymphoblastoid cells Ren et al. 2013 NK cells ↓ NK cell viability, cytolytic activity, and perforin secretion Li et al. 2013 Erythrocytes ↓ expansion of human erythroid progenitor cells in vitro Ji et al. 2013 MSCs ↓ viability of bone marrow stromal cells She et al. 2013 a All significant effects reported with p <0.05. Source: Committee generated. 

Independent Assessment of Formaldehyde 155 Given the variability of blood measures in any person over time and the heterogeneity among people in a population, it is difficult to find statistically significant changes in blood measures in human studies. Thus, it is notable that despite the inherent limitations of studying hematologic measures, over 14 re- cently published studies reported statistically significant hematologic effects on multiple hematopoietic-cell types. Although there are valid concerns about some results in individual studies (for example, the authors of one study used the con- sequences of the thalassemia trait for mean corpuscular volume to explain the findings), it is unlikely that most of these studies have been confounded by such issues. In light of the numerous studies that have reported significant differences in multiple measures, there is a strong association between inhaled formalde- hyde exposure in humans and hematologic effects. Although confounding exposures may complicate the interpretation of some studies, most of the studies documented efforts to identify possible con- founding factors. Several studies were conducted in occupations in which for- maldehyde was probably the predominant exposure during the period of study. One study showed that hematologic changes occurred in individual subjects over a limited period of exposure (Ying et al. 1999). Thus, the hematologic effects observed in those studies establish a specific association with inhaled formalde- hyde in humans. Establishing the temporal relationship of exposure and effect is difficult in most human-exposure studies. Several studies report an association between duration of employment and exposure to formaldehyde, and an 8-week anatomy-laboratory exposure study (Ying et al. 1999) supports a temporal rela- tionship. There is evidence from one human study that supports a biologic gra- dient of formaldehyde exposure and hematologic effects. In this study, increases in T cells and decreases in NK cells were proportional to formaldehyde exposure level (Costa et al. 2013). Those findings are supported by findings in animal- exposure studies (see below). Hematologic Effects in Animals Exposed to Formaldehyde Experimental-animal studies are informative with regard to the specificity, temporal relationship, and exposure–response relationship between formalde- hyde and hematologic effects. It can be argued that rodents and humans differ in the mechanics of inhalation, the physiology of hematopoietic-cell turnover, and DNA-repair mechanisms. Therefore, results of animal studies were evaluated as supporting data, whereas the human data presented above are considered the primary source of evidence of potential associations of formaldehyde exposure and hematologic effects. Six studies addressed the hematologic effects of exposure to formaldehyde in animals in vivo, of which four were published after the publication of the 12th RoC (Tables 3-11 and 3-12). There is poor agreement between individual stud- ies as to the direction of hematologic effects induced by inhaled formaldehyde in animals. In particular, increased or decreased effects on total white-cell counts, 

156 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens total lymphocyte counts, and percentage limit the ability to interpret the results. In addition, other hematologic effects have been reported in only one study, so the consistency of the findings cannot be assessed. The committee finds limited evidence of consistent hematologic effects in the few available studies of for- maldehyde-exposed animal models alone. In the experimental-animal studies, the associations that were observed were often strong in magnitude or level of statistical significance, although the clinical and biologic significance is unknown (Katsnelson et al. 2013; Kim et al. 2013). Thus, the strength of those specific associations is quite high, even if the consistency of the findings is limited. As is expected in experimental-animal studies, the observed multiple hematologic effects can be closely linked to the tested agent, and this establishes a specific association with formaldehyde. By their nature, the animal-exposure studies establish the temporal relationship be- tween inhaled formaldehyde exposure and multiple hematologic effects. In par- ticular, specific hematologic effects were shown to depend on the duration of exposure (Kim et al. 2013). Two animal studies reported multiple hematologic measures, and effects on them were proportional to formaldehyde concentrations (Kim et al. 2013; Zhang et al. 2013). The results suggest an exposure–response relationship between formaldehyde exposure and hematologic effects. Hematologic Effects on Isolated Animal or Human Cells In vitro studies of hematologic effects are of limited utility because they evaluate a nonintact hematopoietic system, which ignores the complex interplay between various cell types and the vascular and lymphohematopoietic organs. Such studies do not account for the complex dynamics between the portal of entry and the systemic distribution of formaldehyde. The committee examined six studies that reported cytotoxic effects on or functional consequences for hematopoietic cells or bone marrow stromal cells, of which five were published after publication of the 12th RoC (Table 3-10 and 3-11). All six studies reported deleterious effects of formaldehyde exposure on T cells, B cells, NK cells, or bone marrow stromal cells; this suggests that formal- dehyde may have hematologic effects if it comes into direct contact with these cell types. However, given the unclear relevance of direct exposure in in vitro studies, particularly exposure to formaldehyde, the committee concludes that although the available literature demonstrates a deleterious effect of formalde- hyde exposure on hematologic cells in vitro, it is difficult to draw firm conclu- sions regarding the hematologic effects of formaldehyde on isolated animal and human cells. The direct effects reported on several hematopoietic cell types raise important questions, but additional studies are needed that account for the phys- iologic exposure of hematopoietic cells to formaldehyde and its metabolites and for poorly understood systemic consequences. 

Independent Assessment of Formaldehyde 157 Conclusions and Considerations for Hematologic Effects The committee concludes that the association of inhalation formaldehyde exposure and diverse hematologic effects is supported by evidence from human studies. Studies in experimental animals provide some additional support. The consistency of individual hematologic effects varied among multiple human and animal studies, and many reported decreases in hematologic measures. The strength of the association in multiple reports of hematologic effects in multiple populations is convincing. The specificity of findings in exposed humans is challenging, but select human studies and experimental-animal studies support the specificity of the association. The temporal relationship is adequately ad- dressed in most studies, and the biologic gradient is addressed in some studies, particularly in animal studies. Taken as a whole, the body of evidence from studies of exposed humans and animals indicates broad and strong associations between exposure to inhaled formaldehyde and hematologic effects. Toxicogenomics Toxicogenomics is the study of gene-expression changes elicited by a tox- icant. The committee reviewed recent toxicogenomic publications to gain a bet- ter understanding of changes in gene expression after formaldehyde exposure. The committee looked specifically at toxicogenomic studies and identified eight publications that had microarray data. Those publications provided information on the genomewide expression of mRNA transcripts in humans, experimental animals, or cultured cells after exposure to formaldehyde. Five of the publica- tions were identified through the committee’s independent literature search for genotoxicity and mutagenicity studies (Andersen et al. 2010; Zeller et al. 2011a; Cheah et al. 2013; Neuss et al. 2010b; Kuehner et al. 2013) (see Figure D-4), and two additional publications were identified from the reference lists of those relevant publications (Hester et al. 2003; Andersen et al. 2008). One publication was identified during the committee’s secondary ad hoc effort to identify rele- vant literature (Rager et al. 2013). Five of the eight publications described expo- sures in humans or experimental animals (Hester et al. 2003; Andersen et al. 2008, 2010; Zeller et al. 2011a; Rager et al. 2013), and the remaining three used cell culture (Hester et al. 2003; Neuss et al. 2010b; Cheah et al. 2013). The eight studies are described in more detail in this section and in Table 3-13. Zeller et al. (2011a) used volunteer human subjects to examine tran- scriptomal changes in nasal inferior turbinate biopsies and peripheral blood samples after inhalation of formaldehyde vapor at up to 0.8 ppm 4 hours/day for 5 days. This is the only study that the committee identified that attempted to examine both portal-of-entry and systemic transcriptomal effects of formalde- hyde. The authors reported that 27 mRNA transcripts were differentially ex- pressed between exposed and nonexposed conditions in the nasal specimens. In 

158 TABLE 3-13 Transcriptomal Profiling Studies Model Subjects Exposure Sample Criteriaa Main conclusions Reference Animals or Human volunteers: Formaldehyde Before and after 2-fold or 1.5-fold; p Formaldehyde exposure affected mRNA Zeller et al. 2011a humans male nonsmokers vapor exposure (paired) < 0.05 (paired t); no expression in nasal biopsy or blood samples or ex-smokers nasal biopsy (inferior FDR correction only marginally. There were 2–17 and 25– Up to 0.8 ppm turbinate); venous 67 differentially expressed genes identified 4 hours/day for whole blood in biopsies with 2.0- and 1.5-fold difference 5 days criteria, respectively. Results identified 0–9 and 6–39 differentially expressed genes in 3 groups the blood with 2.0- and 1.5-fold difference (5–8/group) criteria, respectively. Differentially expressed genes identified in the three exposure groups showed little overlap. No significant specific pathways involving differentially expressed genes were apparent. When FDR cutoff (less than 10%) was applied in addition to 1.5-fold change cutoff, no differentially expressed genes were detected. Nonhuman Formaldehyde Nasal epithelial 1.5-fold; p < 0.05 Low (2 ppm) and high (6 ppm) doses of Rager et al. 2013 primates: male vapor tissue from (ANOVA); FDR formaldehyde changed 3 and 13 micro-RNA Cynomolgus maxilloturbinate corrected q < 0.1 expressions, respectively. Suppression of macaques 0 (n = 2), 2 region collected transcriptional targets of most significantly (n = 3), and 6 ppm by necropsy increased miRNA (miR-125b) was (n = 3) 6 hours/day confirmed by real-time PCR. Induction of for 2 days transcriptional targets of most robustly decreased miRNA (miR-142-3p) was also confirmed by real-time PCR. Four miR-125b targets encoding proapoptotic regulators BAK1, CASP2, MAP2K7, and MCL1 b were downregulated. Thus, formaldehyde 

exposure disrupts miRNA expression in nasal epithelium and probably affects apoptosis. Rats: Formaldehyde Nasal surface 2-fold; Benjamini- Exposure to formaldehyde at 2 ppm caused Andersen et male F344/CrlBR vapor epithelial cells (lateral Hochberg; FDR induction of genes involved in cellular stress al. 2010 meatus < 0.05 responses—thiol transport/reduction, 0, 0.7, 2, 6, 10, and and nasoturbinate inflammation, and cell proliferation—at all 15 ppm 6 hours/day encompassing area exposure durations. Exposure to for 1, 4, 13 weeks between levels II and formaldehyde at 6 ppm or greater resulted in III) selectively changes in expression of genes involved in (15 per dose per isolated by incubating cell-cycle regulation, DNA repair, and time) necropsy tissues in apoptosis. protease mixture Rats: Formaldehyde Nasal surface 1.5-fold; No differentially expressed genes were Andersen et male F344/CrlBR vapor or instillation epithelial cells Benjamini- detected after exposure to formaldehyde al. 2008 (lateral meatus and Hochberg; FDR vapor at 0.7 ppm. Exposure at 2 and 6 ppm Vapor: 0, 0.7, 2, nasoturbinate < 0.05 resulted in up to 15 and 54 differentially and 6 ppm 6 encompassing area expressed genes, respectively, at different hours/day for between levels II and timings over the course of the 3-week 5 days/week for up III) selectively exposure. Exposure at 15 ppm caused 745 to 3 weeks (5 per isolated by differentially expressed genes within 24- dose per time) incubating necropsy hour period, and exposure by instillation tissues in protease (400 mM x 40 µL per nostril) caused 2,553 Vapor: 15 ppm mixture differentially expressed genes within 24- for 6 hours hour period. About 75% of differentially (10 exposed, expressed genes caused by exposure at 15 5 controls) ppm were also affected by exposure via instillation, and these genes were enriched in Instillation: gene ontology categories of wound response, 400 mM x 40 μL apoptotic regulation, inflammation, and per nostril, 6 hours receptor tyrosine kinase signaling. (10 exposed, 5 controls) (Continued) 159 

160 TABLE 3-13 Continued Model Subjects Exposure Sample Criteriaa Main conclusions Reference Rats: Formaldehyde Nasal epithelial cell Benjamini- Exposure to formaldehyde caused Hester et al. 2003 male F344 instillation lysis by direct Hochberg; FDR differential gene expression. These genes instillation of Trizol < 0.05 or 0.1 were enriched in pathways relevant to 400 mM reagent xenobiotic metabolism, cell cycle, apoptosis, formaldehyde and DNA repair. (n = 3) or water (n = 4) x 40 μL per nostril, 24 hours Cell Primary culture 20 or 100 μM for 2 Total cell lysate 2-fold; p < 0.05 Exposure to 100 and 200 μM formaldehyde Neuss et al. 2010b culture human nasal hours; 50, 100, 200 (t test); no FDR for 4 hours changed expression of 153 and epithelial cells μM for 4 hours; correction 887 genes, respectively. Exposure to 50 μM (commercial 100 or 200 μM for formaldehyde for 24 hours with 4 repeats product, derived 24 hours; 20 or 50 changed expression of 143 genes. Less from three μM for 24 hours than 10 differentially expressed genes were Caucasian women) with 4 consecutive observed with all other conditions. Genes repeats; no upregulated by exposure to 200 μM exposure control formaldehyde for 4 hours were enriched for apoptosis regulation and stress response. Human A549 lung- 0 or 83.2 μM for Total cell lysate 1.5-fold; Exposure to 83.2 μM formaldehyde for 2 Cheah et al. 2013 cancer cell line 2 hours Benjamini- hours caused 66 differential gene (adenocarcinoma, Hochberg;FDR expressions, which were enriched for alveolar basal < 0.05 apoptosis regulation, transcription, and epithelial) DNA damage (upregulated genes) or transcription (downregulated genes). Human TK6 B 0, 50, 100, or 200 Total cell lysate 1.5-fold and Exposure to 50 μM formaldehyde did not Kuehner et al. 2013 lymphoblastoid uM for 4 or 24 2-fold; p < 0.05 cause significant transcriptomal changes. cells hours (t test); FDR < 0.1 Exposure to 200 μM formaldehyde caused (multi-variable 2,147 and 2,502 differentially expressed permutation test) genes after 4 or 24 hours of exposure, respectively. Exposure to 100 μM formaldehyde for 4 hours caused 1,367 differentially expressed genes, whereas 

exposure to the same concentration of formaldehyde for 24 hours caused only 2 differentially expressed genes. Genes upregulated after exposure to 200 μM formaldehyde for 24 hours were enriched for transcription, transport, protein phosphorylation, signal transduction, and apoptosis. Abbreviation: FDR, false discovery rate. a Criteria for defining differentially expressed genes. b MCL1 isoform 1 is antiapoptotic, whereas isoform 2 is proapoptotic. 161 

162 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens the blood specimens, statistically significant differential expression of 11 mRNA transcripts was observed. However, the authors concluded that these were “mi- nor” effects that reflected assay variability and that inhalation of formaldehyde did not cause alterations in the expression of genes in either the nasal or blood samples. In the absence of appropriate negative exposure control groups, appro- priate positive controls, or detailed power-analysis discussion, the committee was unable to determine whether the results of this study supported the absence of transcriptomal effects after exposure to formaldehyde or whether the study design provided sufficient discovery power in light of the small number of study subjects (six to eight per group). Rager et al. (2013) examined maxilloturbinate necropsy specimens of na- sal epithelial tissues from macaques and observed significant changes in expres- sion of micro-RNAs after exposure to formaldehyde at 6 ppm 6 hours/day for 2 days. Using real-time quantitative polymerase chain reaction methods, the au- thors confirmed significant induction of miR-125b expression and concomitant suppression of its target mRNA transcripts, including proapoptotic genes BAK1, CASP2, MAP2K7, and MCL1. Two other studies examined transcriptomal effects in nasal epithelial cells of F344 rats that were exposed to formaldehyde via vapor or instillation into the nostrils (Hester et al. 2003; Andersen et al. 2010). These studies collectively demonstrated that exposure to formaldehyde, either by inhalation (2 ppm or higher for 6 hours or longer) or by intranasal instillation (40 μL of a 400 mM solution for 6 hours or longer), resulted in significant changes in expression of the mRNA transcripts that encode proteins involved in cell-cycle regulation, DNA repair, wound response, inflammation, and regulation of apoptosis. In comparison, data obtained after exposure to lower doses of formaldehyde were mostly insignificant. Three cell-culture experiments—one that used primary cultures of human nasal epithelial cells (Neuss et al. 2010a), one that used human A549 lung alveo- lar basal epithelial cancer cells (Cheah et al. 2013), and one that used human TK6 lymphoblastoid cells (Kuehner et al. 2013)—demonstrated significant for- maldehyde-related changes in expression of mRNA transcripts that encode pro- teins involved in apoptosis regulation, stress response, transcription, DNA dam- age, transport, and signal transduction. Relatively high concentrations of formaldehyde—greater than 83.2 μM for 2 hours (Cheah et al. 2013) or greater than 100 μM for 4 hours (Neuss et al. 2010a; Kuehner et al. 2013)—resulted in transcriptomal changes, whereas exposure to lower concentrations of formalde- hyde did not have detectable effects even after prolonged exposure. The committee found multiple studies that reported transcriptional re- sponses in nasal cavity epithelial cells from experimental animals exposed to formaldehyde vapor at doses of 2 ppm or greater. The transcriptomal responses were indicative of cell apoptosis, DNA damage, and proliferation, which are relevant to carcinogenesis. The committee notes that the doses are relevant to occupational human exposure to formaldehyde. The committee did not identify studies that considered the transcriptomal effects of chronic, low-dose exposure 

Independent Assessment of Formaldehyde 163 to formaldehyde in the nasal epithelial cells, peripheral blood, or any other tis- sues of human or animal models. SUMMARY OF EVIDENCE The statement of task specifically asked the committee to “integrate the level-of-evidence conclusions, and considering all relevant information in ac- cordance with the RoC listing criteria, make an independent listing recommen- dation for formaldehyde and provide scientific justification for its recommenda- tion” (Appendix B). The committee notes that the term integrate does not have a standard definition in the context of hazard assessment. The committee under- stood the term in its conventional sense of bringing together parts into a whole. To be listed as “reasonably anticipated as a human carcinogen” or “known to be a human carcinogen”, the RoC listing criteria only requires information to be integrated across human studies or across animal studies, and supporting infor- mation can be derived from mechanistic studies. Mechanistic information “can be useful for evaluating whether a relevant cancer mechanism is operating in people” (NTP 2010, p. iv), but a known mechanism is not required for a sub- stance to be listed in the RoC. In the subsections below, the committee summa- rizes human, experimental animal, and mechanistic information on nasopharyn- geal and sinonasal cancer and myeloid leukemia. Summaries were not presented for other kinds of cancer because of a lack of strong evidence that formaldehyde exposure causes other kinds of cancer in humans. Nasopharyngeal and Sinonasal Cancers The committee found clear and convincing epidemiologic evidence of an association between formaldehyde exposure and nasopharyngeal cancer and sinonasal cancer in humans. On the basis of evidence of an association between nasopharyngeal cancer and exposure to formaldehyde in two strong studies—a large case–control study (Vaughan et al. 2000) and a large cohort study (Beane Freeman et al. 2013)—and other supporting studies that were judged to be mod- erately strong (Vaughan et al. 1986a,b; West et al. 1993; Hildesheim et al. 2001; Siew et al. 2012), the committee concludes that the relationship is causal and chance, bias, and confounding factors can be ruled out with reasonable confi- dence. For sinonasal cancer, there is evidence of an association based on a strong, well-conducted pooled case–control study (Luce et al. 2002) and other, corroborating studies that were judged to be moderately strong (Hayes et al. 1986; Olsen and Asnaes 1986; Vaughan et al. 1986a,b; Luce et al. 1993; Siew et al. 2012). The committee concludes that the relationship between formaldehyde and sinonasal cancer is causal and chance, bias, and confounding factors can be ruled out with reasonable confidence. Several well-conducted studies in experimental animal models demon- strate an increase in nasal squamous-cell carcinoma after inhalation exposure to formaldehyde (Kerns et al.1983; Sellakumar et al. 1985; Monticello et al. 1996). 

164 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Two of the studies used F344 rats (Kerns et al. 1983; Monticello et al. 1996), and one used Sprague Dawley rats (Sellakumar et al. 1985). The evidence is corroborated by other rat studies (Feron et al. 1988; Soffritti et al. 1989; Woutersen et al. 1989; Kamata et al. 1997) and by a mouse study (Kerns et al. 1983). Although there are limitations in extrapolating findings on nasal tumors in rodents to nasopharyngeal and sinonasal cancer in humans, the experimental- animal evidence indicates that exposure to inhaled formaldehyde is associated with carcinogenic effects on tissues at the portal of entry. Inhalation of formaldehyde at sufficient concentrations substantially in- creases formaldehyde to above the total endogenous concentration in tissues at the portal of entry in both animal and human studies. There is experimental evi- dence that, due to its chemical reactivity, formaldehyde exerts genotoxic and mutagenic effects and cytotoxicity followed by compensatory cell proliferation at the portal of entry3 in animals and humans exposed to formaldehyde; this pro- vides biologic plausibility of a relationship between formaldehyde exposure and cancer. The evidence on formaldehyde-associated DNA adducts, DNA–protein cross-links, DNA strand breaks, mutations, micronuclei, and chromosomal aber- rations is consistent, strong, and specific. In addition, both temporal and expo- sure–response relationships have been established, most strongly in studies of rodents and nonhuman primates. Myeloid Leukemia The committee found clear and convincing epidemiologic evidence of an association between formaldehyde exposure and myeloid leukemia. There may also be an increase of other lymphohematopoietic cancers, although the evi- dence is less robust. On the basis of three strong studies with widely different coexposures (the NCI formaldehyde-industry cohort [Beane Freeman et al. 2009], the NIOSH garment-worker cohort [Meyers et al. 2013], and the NCI funeral-industry cohort [Hauptmann et al. 2009]) and several moderately strong studies (Walrath and Fraumeni 1983, 1984; Stroup et al. 1986; Coggon et al. 2014), the committee concludes that there is a causal association between for- maldehyde exposure and myeloid leukemia. Chance, bias, and confounding fac- tors can be ruled out with reasonable confidence given the consistent pattern of association in the larger studies that had good exposure assessment. Although multiple lines of reasoning and experimental evidence indicate that it is unlikely that inhalation exposure to formaldehyde will increase formal- dehyde to substantially above endogenous concentrations in tissues distant from the site of entry, there is a robust database of experimental studies of systemic4 3 Defined as effects that arise from direct interaction of inhaled or ingested formalde- hyde with cells or tissues. 4 Defined as effects that occur beyond cells or tissues that have direct interaction with inhaled or ingested formaldehyde. 

Independent Assessment of Formaldehyde 165 mechanistic events that have been observed after exposure to formaldehyde. The committee notes that it is plausible that some of the systemic effects, notably findings of genotoxicity and transcriptional changes in circulating blood cells, may have resulted from the exposure of the cells at the portal of entry (for ex- ample, lymphoid tissue in the nasal mucosa). The mechanistic events that were considered by the committee as relevant to the plausibility of formaldehyde- associated tumors beyond the portal of entry included genotoxicity and muta- genicity, hematologic effects, and effects on gene expression. Overall, in mech- anistic studies of experimental animals and exposed humans, the evidence is largely consistent and strong. As shown in Table 3-10, a majority of the mam- malian in vivo studies resulted in positive findings compared to negative find- ings (60 and 38 studies, respectively), particularly in humans (49 and 19 studies, respectively). Both temporal and exposure–response relationships have been demonstrated in studies of humans and animals exposed to formaldehyde. The committee concludes that these 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. It would be desirable to have a more complete understanding about how formaldehyde exposure may cause systemic effects, but the lack of known mechanisms should not detract from the findings of an association between formaldehyde exposure and myeloid leukemia in epidemi- ology studies. The animal cancer bioassay literature provided some information relevant to myeloid leukemia. One drinking water study (Soffritti et al. 2002) reported a significant increase in lymphohematopoietic cancers following long-term expo- sure to formaldehyde in drinking water, but there is uncertainty regarding the finding. Of the three inhalation studies that included histopathologic examina- tions of non–respiratory tract tissues, two did not report leukemia (Sellakumar et al. 1985; Kamata et al. 1997). The full laboratory report (Battelle 1981) of a third study (Kerns et al. 1983) discussed findings of leukemia and lymphoma that were not found to be compound related. However, diffuse multifocal bone marrow hyperplasia in rats exposed to 15 ppm of formaldehyde for 18 months was increased in both treated males (p = 0.0001) and females (p = 0.0001). Alt- hough the Battelle finding was not a finding of malignancy, it does indicate that long-term inhaled formaldehyde may cause effects in bone marrow. CONCLUSIONS AND LISTING RECOMMENDATION The committee identified and evaluated relevant, publicly available, peer- reviewed literature on formaldehyde, including attention to literature published between June 10, 2011 (the release date of the substance profile for formalde- hyde in the 12th RoC), and November 8, 2013. The committee applied NTP’s established RoC listing criteria to the scientific evidence on formaldehyde from 

166 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens studies of humans, studies of experimental animals, and other studies relevant to mechanisms of carcinogenesis. The type of information needed to meet the criteria for sufficient evidence in experimental animals is clear and transparent, as outlined in the section “Can- cer Studies in Experimental Animals”. In contrast, the RoC listing criteria do not provide detailed guidance about how evidence should be assembled to meet the requirement of limited evidence or sufficient evidence of carcinogenicity from studies in humans, except to note that limited evidence cannot exclude alterna- tive explanations, such as chance, bias, or confounding factors, and to note that conclusions should be based on “scientific judgment, with consideration given to all relevant information” (NTP 2010, p. iv). In the section “Cancer Studies in Humans”, the committee used scientific judgment to develop an approach to assessing the epidemiology evidence. The approach included careful review of individual studies, selection of studies that were most informative, and evalua- tion of informative studies on the basis of the strength, consistency, temporality, dose-response, and coherence of the evidence and on the considerations present- ed in Table 3-1. The committee notes that evidence in experimental animals and a known mechanism of action is not required by the RoC listing criteria in making a list- ing recommendation that a substance is known to be a human carcinogen if the evidence from studies in humans is sufficient and indicates an association be- tween exposure and human cancer. Also, and importantly, the RoC listing crite- ria require an association in only one type of cancer to make the determination. On the basis of the information summarized directly above for nasopharyngeal cancer, sinonasal cancer, and for myeloid leukemia, the committee makes its independent determinations as follows:  There is sufficient evidence of carcinogenicity from studies of humans based on consistent epidemiologic findings on nasopharyngeal cancer, sinonasal cancer, and myeloid leukemia for which chance, bias, and confounding factors could be ruled out with reasonable confidence.  There is sufficient evidence of carcinogenicity in animals based on ma- lignant and benign tumors in multiple species, at multiple sites, by multiple routes of exposure, and to an unusual degree with regard to type of tumor.  There is convincing relevant information that formaldehyde induces mechanistic events associated with the development of cancer in humans, spe- cifically genotoxicity and mutagenicity, hematologic effects, and effects on gene expression. Because there is sufficient evidence of carcinogenicity from studies in humans that indicates a causal relationship between exposure to formaldehyde and at least one type of human cancer, the committee concludes that formaldehyde should be listed in the RoC as “known to be a human carcinogen”. 

Independent Assessment of Formaldehyde 167 REFERENCES Acheson, E.D., H.R. Barnes, M.J. Gardner, C. Osmond, B. Pannett, and C.P. Taylor. 1984. Formaldehyde in the British chemical industry. An occupational cohort study. Lancet 1(8377):611-616. Andersen, M.E., H.J. Clewell, E. Bermudez, G.A. Willson, and R.S. Thomas. 2008. Ge- nomic signatures and dose-dependent transitions in nasal epithelial responses to inhaled formaldehyde in the rat. Toxicol. Sci. 105(2):368-383. Andersen, M.E., H.J. Clewell, E. Bermudez, D.E. Dodd, G.A. Willson, J.L. Campbell, and R.S. Thomas. 2010. Formaldehyde: Integrating dosimetry, cytotoxicity, and genomics to understand dose-dependent transitions for an endogenous compound. Toxicol. Sci. 118(2):716-731. Andjelkovich, D.A., D.B. Janszen, M.H. Brown, R.B. Richardson, and F.J. Miller. 1995. Mortality of iron foundry workers. IV. Analysis of a subcohort exposed to formal- dehyde. J. Occup. Environ. Med. 37(7):826-837. Appelman, L.M., R.A. Woutersen, A. Zwart, H.E. Falke, and V.J. Feron. 1988. One-year inhalation toxicity study of formaldehyde in male rats with a damaged or undam- aged nasal mucosa. J. Appl. Toxicol. 8(2):85-90. Armstrong, R.W., P.B. Imrey, M.S. Lye, M.J. Armstrong, M.C. Yu, and S. Sani. 2000. Nasopharyngeal carcinoma in Malaysian Chinese: Occupational exposures to par- ticles, formaldehyde and heat. Int. J. Epidemiol. 29(6):991-998. ATSDR (Agency for Toxic Substances and Disease Registry). 1999. Toxicological Profile for Formaldehyde. U.S. Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry, Atlanta, GA [online] . Available: http://www.atsdr.cdc.gov/toxprofiles/tp111.pdf [accessed Sept. 23, 2013]. Aydin, S., H. Canpinar, U. Undeger, D. Güc, M. Çolakoğlu, A. Kars, and N. Başaran. 2013. Assessment of immunotoxicity and genotoxicity in workers exposed to low concen- trations of formaldehyde. Arch. Toxicol. 87(1):145-153. Battelle. 1981. Final Report on a Chronic Inhalation Toxicology Study in Rats and Mice Exposed to Formaldehyde. Prepared by Battelle Columbus Laboratories, Colum- bus, OH, for the Chemical Industry Institute of Toxicology (CIIT), Research Tri- angle Park, NC. CIIT Docket No. 10922. Beane Freeman, L.E., A. Blair, J.H. Lubin, P.A. Stewart, R.B. Hayes, R.N. Hoover, and M. Hauptmann. 2009. Mortality from lymphohematopoietic malignancies among workers in formaldehyde industries: The National Cancer Institute Cohort. J. Natl. Cancer Inst. 101(10):751-761. Beane Freeman, L.E., A. Blair, J.H. Lubin, P.A. Stewart, R.B. Hayes, R.N. Hoover, and M. Hauptmann. 2013. Mortality from solid tumors among workers in formalde- hyde industries: An update of the NCI cohort. Am. J. Ind. Med. 56(9):1015-1026. Bertazzi, P.A., A. Pesatori, S. Guercilena, D. Consonni, and C. Zocchetti. 1989. Carcino- genic risk for resin producers exposed to formaldehyde: Extension of follow-up [in Italian]. Med. Lav. 80(2):111-122. Blair, A., and P.A. Stewart. 1990. Correlation between different measures of occupational exposure to formaldehyde. Am. J. Epidemiol. 131(3):510-516. Blair, A., P. Stewart, M. O'Berg, W. Gaffey, J. Walrath, J. Ward, R. Bales, S. Kaplan, and D. Cubit. 1986. Mortality among industrial workers exposed to formaldehyde. J. Natl. Cancer Inst. 76(6):1071-1084. 

168 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Blair, A., P.A. Stewart, and R.N. Hoover. 1990. Mortality from lung cancer among work- ers employed in formaldehyde industries. Am. J. Ind. Med. 17(6):683-699. Blair, A., T. Zheng, A. Linos, P.A. Stewart, Y.W. Zhang, and K.P. Cantor. 2001. Occu- pation and leukemia: A population-based case-control study in Iowa and Minneso- ta. Am. J. Ind. Med. 40(1):3-14. Bolt, H.M. 1987. Experimental toxicology of formaldehyde. J. Cancer Res. Clin. Oncol. 113(4):305-309. Bono, R., M. Vincenti, T. Schiliro, E. Scursatone, C. Pignata, and G. Gilli. 2006. N- Methylenvaline in a group of subjects occupationally exposed to formaldehyde. Toxicol. Lett. 161(1):10-17. Boobis, A.R., J.E. Doe, B. Heinrich-Hirsch, M.E. Meek, S. Munn, M. Ruchirawat, J. Schlatter, J. Seed, and C. Vickers. 2008. IPCS framework for analyzing the rele- vance of a noncancer mode of action for humans. Crit. Rev. Toxicol. 38(2):87-96. Brondeau, M.T., P. Bonnet, J.P Guenier, P. Simon, and J. de Ceaurriz. 1990. Adrenal- dependent leucopenia after short-term exposure to various airborne irritants in rats. J. Appl. Toxicol. 10(2): 83-86. Bucher, J.R. 2013. Follow-up Questions. Material submitted by the NAS Committee on Review of the Formaldehyde Assessment in the NTP 12th RoC and the NAS Committee on Review of the Styrene Assessment in the NTP 12th RoC, April 2, 2013. Buss, J., K. Kuschinsky, H. Kewitz, and W. Koransky. 1964. Enteric resorption of for- maldehyde [in German]. N-S Arch. Exp. Pathol. Pharmakol. 247:380-381. Casanova, M., H.d'A. Heck, J.I. Everitt, W.W. Harrington, Jr., and J.A. Popp. 1988. For- maldehyde concentrations in the blood of rhesus monkeys after inhalation expo- sure. Food Chem. Toxicol. 26(8):715-716. Casanova, M., K.T. Morgan, W.H. Steinhagen, J.I. Everitt, J.A. Popp, and H.D. Heck. 1991. Covalent binding of inhaled formaldehyde to DNA in the respiratory tract of rhesus monkey: Pharmacokinetics, rat to monkey interspecies scaling, and extrapo- lation to man. Fundam. Appl. Toxicol. 17(2):409-428. Casanova-Schmitz, M., T.B. Starr, and H.D. Heck. 1984. Differentiation between meta- bolic incorporation and covalent binding in the labeling of macromolecules in the rat nasal mucosa and bone marrow by inhaled [14C]- and [3H]formaldehyde. Tox- icol. Appl. Pharmacol. 76(1):26-44. Cheah, N.P., J.L. Pennings, J.P. Vermeulen, F.J. van Schooten, and A. Opperhuizen. 2013. In vitro effects of aldehydes present in tobacco smoke on gene expression in human lung alveolar epithelial cells. Toxicol. In Vitro 27(3):1072-1081. Checkoway, H., N. Pearce, and D. Kribel. 2004. Research Methods in Occupational Epi- demiology, 2nd Ed. Oxford: Oxford University Press. Checkoway, H., R.M. Ray, J.I. Lundin, G. Astrakianakis, N.S. Seixas, J.E. Camp, K.J. Wernli, E.d. Fitzgibbons, W. Li, Z. Feng, D.L. Gao, and D.B. Thomas. 2011. Lung cancer and occupational exposures other than cotton dust and endotoxin among women textile workers in Shanghai, China. Occup. Environ. Med. 68(6):425-429. Cheng. Z., Y. Li, B. Liang, and C. Wang. 2004. Investigation of formaldehyde level and health of personnel in clinical pathology. J. Bengbu. Med. Coll. 29(3): 266-267. Coggon, D., E.C. Harris, J. Poole, and K.T. Palmer. 2003. Extended follow-up of a cohort of British chemical workers exposed to formaldehyde. J. Natl. Cancer Inst. 95(21):1608-1615. Coggon, D., G. Ntani, E.C. Harris, and K.T. Palmer. 2014. Upper airway cancer, myeloid leukemia, and other cancers in a cohort of British chemical workers exposed to for- maldehyde. Am. J. Epidemiol. 179(11):1301-1311. 

Independent Assessment of Formaldehyde 169 Costa, S., J. Garcia-Leston, M. Coelho, P. Coelho, C. Costa, S. Silva, B. Porto, B. Laffon, and J.P. Teixeira. 2013. Cytogenetic and immunological effects associated with oc- cupational formaldehyde exposure. J. Toxicol. Environ. Health A 76(4-5):217-229. Dalbey, W.E. 1982. Formaldehyde and tumors in hamster respiratory tract. Toxicology 24(1):9-14. Dell, L., and M.J. Teta. 1995. Mortality among workers at a plastics manufacturing and research and development facility: 1946-1988. Am. J. Ind. Med. 28(3):373-384. Doll, R., and R. Peto. 1978. Cigarette smoking and brochial carcinoma: Dose and time relationships among regular smokes and lifelong non-smokers. J. Epidemiol. Community Health 32(4):303-313. Edling, C., B. Jarvholm, L. Andersson, and O. Axelson. 1987. Mortality and cancer inci- dence among workers in an abrasive manufacturing industry. Br. J. Ind. Med. 44(1):57-59. Edrissi, B., K. Taghizadeh, and P.C. Dedon. 2013a. Quantitative analysis of histone mod- ifications: Formaldehyde is a source of pathological N6-formyllysine that is refrac- tory to histone deacetylases. PLoS Genet. 9(2):e1003328. Edrissi, B., K. Taghizadeh, B.C. Moeller, D. Kracko, M. Doyle-Eisele, J.A. Swenberg, and P.C. Dedon. 2013b. Dosimetry of N6-formyllysine adducts following [13C2H2]- formaldehyde exposures in rats. Chem. Res. Toxicol. 26(10):1421-1423. Egle, J.L., Jr. 1972. Retention of inhaled formaldehyde, propionaldehyde, and acrolein in the dog. Arch. Environ. Health 25(2):119-124. Elliot, L.J., L.T. Stayner, L.M. Blade, W. Helperin, and R. Keenlyside. 1987. Formalde- hyde Exposure Characterization in Garment Manufacturing Plants: A Composite Summary of Three in-depth Industrial Hygiene Surveys. Division of Surveillance, Hazard Evaluations and Field Studies, National Institute for Occupational Safety and Health, Cincinnati, OH. EPA (U.S. Environmental Protection Agency). 2005. Guidelines for Carcinogen Risk As- sessment. EPA/630/P-03/001F. Risk Assessment Forum, U.S. Environmental Protec- tion Agency, Washington, DC [online]. Available: http://www.epa.gov/raf/publicati ons/pdfs/CANCER_GUIDELINES_FINAL_3-25-05.PDF [accessed Jan. 23, 2014]. Erdei, E., J. Bobvos, M. Brozik, A. Paldy, I. Farkas, E. Vaskovi, and P. Rudnai. 2003. Indoor air pollutants and immune biomarkers among Hungarian asthmatic chil- dren. Arch. Environ. Health 58(6):337-347. Feron, V.J., J.P. Bruyntjes, R.A. Woutersen, H.R. Immel, and L.M. Appelman. 1988. Nasal tymours in rats after short-term exposure to a cytotoxic concentration of formaldehyde. Cancer Lett. 39(1):101-111. Fox, E.M. 1985. Urea formaldehyde foam insulation: Defusing a time-bomb. Am. J. Law Med. 11(1):81-104. Frank, N.R., R.E. Yoder, J.D. Brain, and E. Yokoyama. 1969. SO2 (35S labeled) absorp- tion by the nose and mouth under conditions of varying concentration and flow. Arch. Environ. Health 18(3):315-322. Franks, S.J. 2005. A mathematical model for the absorption and metabolism of formalde- hyde vapour by humans. Toxicol. Appl. Pharmacol. 206(3):309-320. Gardner, M.J., B. Pannett, P.D. Winter, and A.M. Cruddas. 1993. A cohort study of workers exposed to formaldehyde in the British chemical industry: An update. Br. J. Ind. Med. 50(9):827-834. Garschin, W.G., and L.M. Schabad. 1936. About atypical proliferation of the bronchial epithelium with the introduction of formalin into the lung tissue [in German]. Z. Krebsforsch. 43(1):137-145. 

170 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Gentry, P.R., J.V. Rodricks, D. Bachand, C. Van Landingham, A.M. Shipp, R.J. Alber- tini, and R. Irons. 2013. Formaldehyde exposure and leukemia: Critical review and reevaluation of the results from a study that is the focus for evidence of biological plausibility. Crit. Rev. Toxicol. 43(8):661-670. Georgieva, A.V., J.S. Kimbell, and P.M. Schlosser. 2003. A distributed-parameter model for formaldehyde update and disposition in the rat nasal lining. Inhal. Toxicol. 15(14):1435-1463. Gift, J.S., J.C. Caldwell, J. Jinot, M.V. Evans, I. Cote, and J.J. Vandenberg. 2013. Scien- tific considerations for evaluating cancer bioassays conducted by the Ramazzini Institute. Environ. Health Perspect, 121(11-12):1253-1263. Glass, D.C., M.R. Sim, L. Fritschi, C.N. Gray, D.J. Jolley, and C. Gibbon. 2004. Leuke- mia risk and relevant benzene exposure period – Re: Followup time on risk esti- mates, Am. J. Ind. Med. 42:481-489, 2002 [letter]. Am. J. Ind. Med. 45(2):222- 223. Gloede, E., J.A. Cichocki, J.B. Baldino, and J.B. Morris. 2011. A validated hybrid com- putational fluid dynamic-physiologically based pharmacokinetic model for respira- tory tract vapor absorption in hum and rat and its application to inhalation dosime- try of diacetyl. Toxicol. Sci. 123(1):231-246. Goldstein, H.B. 1973. Textiles and the chemical industry: A marriage. J. Am. Assoc. Text. Chem. Color. 5(10):209–214. Hall, A., J.M. Harrington, and T.C. Aw. 1991. Mortality study of British pathologists. Am. J. Ind. Med. 20(1):83-89. Hansen, J., and J.H. Olsen. 1995. Formaldehyde and cancer morbidity among male em- ployees in Denmark. Cancer Causes Control 6(4):354-360. Hansen, J., and J.H. Olsen. 1996. Occupational exposure to formaldehyde and risk of cancer [in Danish]. Ugeskr. Laeger. 158(29):4191-4194. Harrington, J.M., and D. Oakes. 1984. Mortality study of British pathologists 1974-80. Br. J. Ind. Med. 41(2):188-191. Hauptmann, M., J.H. Lubin, P.A. Stewart, R.B. Hayes, and A. Blair. 2004. Mortality from solid cancers among workers in formaldehyde industries. Am. J. Epidemiol. 159(12):1117-1130. Hauptmann, M., P.A. Stewart, J.H. Lubin, L.E. Beane Freeman, R.W. Hornung, R.F. Herrick, R.N Hoover, J.F. Fraumeni Jr., A. Blair, and R.B. Hayes. 2009. Mortality from lymphohematopoietic malignancies and brain cancer among embalmers ex- posed to formaldehyde. J. Natl. Cancer Inst. 101(24):1696-1708. Hayes, R.B., J.W. Raatgever, A. de Bruyn, and M. Gerin. 1986. Cancer of the nasal cavity and paranasal sinuses, and formaldehyde exposure. Int. J. Cancer 37(4):487-492. Hayes, R.B., A. Blair, P.A. Stewart, R.F. Herrick, and H. Mahar. 1990. Mortality of U.S. embalmers and funeral directors. Am. J. Ind. Med. 18(6):641-652. Heck, H., and M. Casanova. 2004. The implausibility of leukemia induction by formal- dehyde: A critical review of the biological evidence on distant-site toxicity. Regul. Toxicol. Pharmacol. 40(2):92-106. Heck, H.A., E.L. White, and M. Casanova-Schmitz. 1982. Determination of formalde- hyde in biological tissues by gas chromatography/mass spectrometry. Biomed. Mass. Spectrom. 9(8):347-353. Heck, H.A., M. Casanova-Schmitz, P.B. Dodd, E.N. Schachter, T.J. Witek, and T. Tosun. 1985. Formaldehyde (CH2O) concentrations in the blood of humans and Fischer- 344 rats exposed to CH2O under controlled conditions. Am. Ind. Hyg. Assoc. J. 46(1):1-3. 

Independent Assessment of Formaldehyde 171 Heck, H.A., M. Casanova, W.H. Steinhagen, J.I. Everitt, K.T. Morgan, and J.A. Popp. 1989. Formaldehyde toxicity: DNA-protein cross-linking studies in rats and non- human primates. Pp. 159-164 in Nasal Carcinogenesis in Rodents: Relevance to Human Risk, V.J. Feron, and M.C. Bosland, eds. Wageningen: Pudoc. Hester, S.D., G.B. Benavides, L. Yoon, K.T. Morgan, F. Zou, W. Barry, and D.C. Wolf. 2003. Formaldehyde-induced gene expression in F344 rat nasal respiratory epithe- lium. Toxicology 187(1):13-24. Hildesheim, A., M. Dosemeci, C.C. Chan, C.J. Chen, Y.J. Cheng, M.M. Hsu, I.H. Chen, B.F. Mittl, B. Sun, P.H. Levine, J.Y. Chen, L.A. Brinton, and C.S. Yang. 2001. Occupational exposure to wood, formaldehyde, and solvents and risk of nasopha- ryngeal carcinoma. Cancer Epidemiol. Biomarkers Prev. 10(11):1145-1153. Hill, A.B. 1965. The environment and disease: Association or causation? Proc. R. Soc. Med. 58(5):295-300. Holmström, M., B. Wilhelmosson, and H. Hellquist. 1989. Histological changes in the nasal mucosa in rats after long-term exposure to formaldehyde and wood dust. Ac- ta Otolaryngol. 108(3-4):274-283. Horton, A.W., R. Tye, and K.L. Stemmer. 1963. Experimental carcinogenesis of the lung. Inhalation of gaseous formaldehyde or an aerosol coal tar by C3H mice. J. Natl. Cancer Inst. 30:31-43. Hosgood, H.D., III, L. Zhang, X. Tang, R. Vermeulen, Z. Hao, M. Shen, C. Qiu, Y. Ge, M. Hua, Z. Ji, S. Li, J. Xiong, B. Reiss, S. Liu, K.X. Xin, M. Azuma, Y. Xie, L. Beane Freeman, X. Ruan, W. Guo, N. Galvan, A. Blair, L. Li, H. Huang, M.T. Smith, N. Rothman, and Q. Lan. 2013. Occupational exposure to formaldehyde and alterations in lymphocyte subsets. Am. J. Ind. Med. 56(2):252-257. IARC (International Agency for Research on Cancer). 1982. Chemicals, Industrial Pro- cesses and Industries Associated with Cancer in Humans: An Updating of IARC Monographs Volumes 1 to 29. IARC Monographs on the Evaluation of the Car- cinogenic Risks to Humans Supplement 4. Lyon, France: IARC [online]. Availa- ble: http://monographs.iarc.fr/ENG/Monographs/suppl4/Suppl4.pdf [accessed June 10, 2013]. IARC (International Agency for Research on Cancer). 1995. Wood Dust and Formalde- hyde. IARC Monographs on the Evaluation of the Carcinogenic Risks to Humans Vol. 62. Lyon, France: IARC [online]. Available: http://monographs.iarc.fr/ENG/ Monographs/vol62/mono62.pdf [accessed June 10, 2013]. IARC (International Agency for Research on Cancer). 2006a. Formaldehyde. Pp. 39-325 in Formaldehyde, 2-Butoxyethanol and 1-tert-Butoxypropan-2-ol. IARC Mono- graphs on the Evaluation of the Carcinogenic Risks to Humans Vol. 88. Lyon, France: IARC [online]. Available: http://monographs.iarc.fr/ENG/Monographs/vol 88/mono88.pdf [June 10, 2013]. IARC (International Agency for Research on Cancer). 2006b. IARC Monographs on the Evaluation of the Carcinogenic Risks to Humans: Preamble. Lyon, France: IARC [online]. Available: http://monographs.iarc.fr/ENG/Preamble/CurrentPreamble.pdf [accessed June 10, 2013]. Jakab, M.G., T. Klupp, K. Besenyei, A. Biro, J. Major, and A. Tompa. 2010. Formalde- hyde-induced chromosomal aberrations and apoptosis in peripheral blood lympho- cytes of personnel working in pathology departments. Mutat. Res. 698(1-2):11-17. Ji, Z., X. Li, M. Fromowitz, E. Mutter-Rottmayer, J. Tung, M. Smith, and L. Zhang. 2013. Formaldehyde induces micronuclei in mouse erythropoietic cells and suppresses the expansion of human erythroid progenitor cells. Toxicol. Lett. 224(2):233-239. 

172 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Jiang, S., L. Yu, J. Cheng, S. Leng, Y. Dai, Y. Zhang, Y. Niu, H. Yan, W. Qu, C. Zhang, K. Zhang, R. Yang, L. Zhou, and Y. Zheng. 2010. Genomic damages in peripheral blood lymphocytes and association with polymorphisms of three glutathione S- transferases in workers exposed to formaldehyde. Mutat. Res. 695(1-2):9-15. Kamata, E., M. Nakadate, O. Uchida, Y. Ogawa, S. Suzuki, T. Kaneko, M. Saito, and Y. Jurokawa. 1997. Results of a 28-month chronic inhalation toxicity study of formal- dehyde in male Fisher-344 rats. J. Toxicol. Sci. 22(3):239-254. Katsnelson, B.A., T.D. Degtyareva, L.I. Privalova, I.A. Minigaliyeva, T.V. Slyshkina, V.V. Ryzhow, and O.Y. Beresneva. 2013. Attenuation of subchronic formaldehyde inhalation toxicity with oral administration of flutamate, glycine and methione. Toxicol. Lett. 220(2):181-186. Kerns, W.D., K.L. Pavkov, D.J. Donofrio, E.J. Gralla, and J.A. Swenberg. 1983. Car- cinogenicity of formaldehyde in rats and mice after long-term inhalation exposure. Cancer Res. 43(9):4382-4392. Kim, E.M., H.Y. Lee, E.H. Lee, K.M. Lee, M. Park, K.Y. Ji, J.H. Jang, Y.H. Jeong, K.H. Lee, I.J. Yoon, S.M. Kim, M.J. Jeong, K.D. Kim, and H.S. Kang. 2013. Formalde- hyde exposure impairs the function and differentiation of NK cells. Toxicol. Lett. 223(2):154-161. Kimbell, J.S. 2006. Nasal dosimetry of inhaled gases and particles: Where do inhaled agents go in the nose? Toxicol. Pathol. 34(3):270–273. Kimbell, J.S., R.P. Subramaniam, E.A. Gross, P.M. Schlosser, and K.T. Morgan. 2001. Dosimetry modeling of inhaled formaldehyde: Comparisons of local flux predic- tions in the rat, monkey, and human nasal passages. Toxicol. Sci. 64(1):100-110. Kuehner, S., K. Holzmann, and G. Speit. 2013. Characterization of formaldehyde's geno- toxic mode of action by gene expression analysis in TK6 cells. Arch. Toxicol. 87(11):1999-2012. Kuo, H., G. Jian, C. Chen, C. Liu, and J. Lai. 1997. White blood cell count as an indicator of formaldehyde exposure. Bull. Environ. Contam. Toxicol. 59(2):261–267. Ladeira, C., S. Viegas, E. Carolino, J. Prista, M.C. Gomes, and M. Brito. 2011. Genotox- icity biomarkers in occupational exposure to formaldehyde--the case of histo- pathology laboratories. Mutat. Res. 721(1):15-20. Levine, R.J., D.A. Andjelkovich, and L.K. Shaw. 1984. The mortality of Ontario under- takers and a review of formaldehyde-related mortality studies. J. Occup. Med. 26(10):740-746. Li, Q., Q. Mei, T. Huyan, L. Xie, S. Che, H. Yang, M. Zhang., and Q. Huang. 2013. Ef- fects of formaldehyde exposure on human NK cells in vitro. Environ. Toxicol. Pharmacol. 36(3): 948-955. Li, W., R.M. Ray, D.L. Gao, E.D. Fitzgibbons, N.S. Seixas, J.E. Camp, K.J. Wernli, G. Astrakianakis, Z. Feng, D.B. Thomas, and H. Checkoway. 2006. Occupational risk factors for nasopharyngeal cancer among female textile workers in Shanghai, Chi- na. Occup. Environ. Med. 63(1):39-44. Lin, D., Y. Guo, J. Yi, D. Kuang, X. Li, H. Deng, K. Huang, L. Guan, Y. He, X. Zhang, D. Hu, Z. Zhang, H. Zheng, X. Zhang, C.M. McHale, L. Zhang, and T. Wu. 2013. Occupational exposure to formaldehyde and genetic damage in the peripheral blood lymphocytes of plywood workers. J. Occup. Health 55(4):284-291. Lino-dos-Santos-Franco, A., M. Correa-Costa, A.C. Durao, A.P. de Oliveira, A.C. Breithaupt-Faloppa, J. de Almeida Bertoni, R.M. Oliveira-Filho, N.O. Camara, T. Marcourakis, and W. Tavares-de-Lima. 2011. Formaldehyde induces lung inflam- mation by an oxidant and antioxidant enzymes mediated mechanism in the lung tissue. Toxicol. Lett. 207(3):278–285. 

Independent Assessment of Formaldehyde 173 Lino-dos-Santos-Franco, A., J.A. Gimenes-Júnior, A.P. Ligeiro-de-Oliveira, A.C. Breithaupt-Faloppa, B.G. Acceturi, L.B. Vitoretti, I.D. Machado, R.M. Oliveira- Filho, S.H. Farsky, H.T. Moriya, and T. Tavares-de-Lima. 2013. Formaldehyde inhalation reduces respiratory mechanics in a rat model with allergic lung inflam- mation by altering the nitric oxide/cyclooxygenase-derived products relationship. Food Chem. Toxicol. 59:731-738. Lu, K., L.B. Collins, H. Ru, E. Bermudez, and J.A. Swenberg. 2010a. Distribution of DNA adducts caused by inhaled formaldehyde is consistent with induction of nasal carcinoma but not leukemia. Toxicol. Sci. 116(2):441-451. Lu, K., W.J. Ye, L. Zhou, L.B. Collins, X. Chen, A. Gold, L.M. Ball, and J.A. Swenberg. 2010b. Structural characterization of formaldehyde-induced cross-links between amino acids and deoxynucleosides and their oligomers. J. Am. Chem. Soc. 132(10):3388-3399. Lu, K., B. Moeller, M. Doyle-Eisele, J. McDonald, J.A. Swenberg. 2011. Molecular do- simetry of N2-hydroxymethyl-dG DNA adducts in rats exposed to formaldehyde. Chem. Res. Toxicol. 24(2):159-161. Luce, D., M. Gerin, A. Leclerc, J.F. Morcet, J. Brugere, and M. Goldberg. 1993. Sinona- sal cancer and occupational exposure to formaldehyde and other substances. Int. J. Cancer. 53(2):224-231. Luce, D., A. Leclerc, D. Begin, P.A. Demers, M. Gerin, E. Orlowski, M. Kogevinas, S. Belli, I. Bugel, U. Bolm-Audorff, L.A., Brinton, P. Comba, L. Hardell, R.B. Hayes, C. Magnani, E. Merler, S. Preston-Martin, T.L. Vaughan, W. Zheng, and P. Boffetta. 2002. Sinonasal cancer and occupational exposures: A pooled analysis of 12 case-control studies. Cancer Causes Control 13(2):147-157. Luo, J., M. Hendryx, and A. Ducatman. 2011. Association between six environmental chemicals and lung cancer incidence in the United States. J. Environ. Public Health (2011): Art. 463701. Lyapina, M., G. Zhelezova, E. Petrova, and M. Boev. 2004. Flow cytometric determina- tion of neutrophil respiratory burst activity in workers exposed to formaldehyde. Int. Arch. Occup. Environ. Health 77(5):335-340. Madison, R.E., A. Broughton, and J.D. Thrasher. 1991. Immunologic biomarkers associ- ated with an acute exposure to exothermic byproducts of a ureaformaldehyde spill. Environ. Health Perspect. 94: 219-223. Mahboubi, A., A. Koushik, J. Siemiatycki, J. Lavoué, and M.C. Rousseau. 2013. As- sessment of the effect of occupational exposure to formaldehyde on the risk of lung cancer in two Canadian population-based case-control studies. Scand. J. Work Environ. Health 39(4):401-410. Malarkey, D.E., and J.R. Bucher. 2011. Summary Report of the National Toxicology Pro- gram and Environmental Protection Agency‐Sponsored Review of Pathology Mate- rials from Selected Ramazzini Institute Rodent Cancer Bioassays, November 29, 2011. National Toxicology Program [online]. Available: http://ntp.niehs.nih.gov/N TP/About_NTP/Partnerships/International/SummaryPWG_Report_RI_Bioassays.pdf [accessed Mar. 12, 2014]. McGwin, G., J. Lienert, and J.L. Kennedy, Jr. 2011. Formaldehyde exposure and asthma in children: A systematic review. Cien. Saude Colet. 16(9):3845-3851. Meyers, A.R., L.E. Pinkerton, and M.J. Hein. 2013. Cohort mortality study of garment industry workers exposed to formaldehyde: Update and internal comparisons. Am. J. Ind. Med. 56(9):1027-1039. Moeller, B.C., K. Lu, M. Doyle-Eisele, J. McDonald, A. Gigliotti, and J.A. Swenberg. 2011. Determination of N2-hydroxymethyl-dG adducts in the nasal epithelium and 

174 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens bone marrow of nonhuman primates following 13CD2-formaldehyde inhalation exposure. Chem. Res. Toxicol. 24(2):162-164. Monticello, T.M., K.T. Morgan, J.I. Everitt, and J.A. Popp. 1989. Effects of formalde- hyde gas on the respiratory tract of rhesus monkeys. Pathology and cell prolifera- tion. Am. J. Pathol. 134(3):515-527. Monticello, T.M., J.A. Swenberg, E.A. Gross, J.R. Leininger, J.S. Kimbell, S. Seilkop, T.B. Starr, J.E. Gibson, and K.T. Morgan. 1996. Correlation of regional and non- linear formaldehyde-induced nasal cancer with proliferating populations of cells. Cancer Res. 56(5):1012-1022. Muller, P., G. Raabe, and D. Schumann. 1978. Leukoplakia induced by repeated deposi- tion of folmalin in rabbit oral mucosa. Long –term experiments with a new ‘oral tank”. Exp. Pathol. 16(1-6):36-42. Neuss, S., K. Holzmann, and G. Speit. 2010a. Gene expression changes in primary hu- man nasal epithelial cells exposed to formaldehyde in vitro. Toxicol. Lett. 198(2):289-295. Neuss, S., B. Moepps, and G. Speit. 2010b. Exposure of human nasal epithelial cells to formaldehyde does not lead to DNA damage in lymphocytes after co-cultivation. Mutagenesis 25(4):359-364. Nielsen, G.D., S.T. Larsen, and P. Wolkoff. 2013. Recent trend in risk assessment of formaldehyde exposures from indoor air. Arch. Toxicol. 87(1):73-98. NRC (National Research Council). 2011. Review of the Environmental Protection Agen- cy's Draft IRIS Assessment of Formaldehyde. Washington, DC: National Acade- mies of Science. NRC (National Research Council). 2014. Review of EPA’s Integrated Risk Information System (IRIS) Process. Washington, DC: The National Academies Press. NTP (National Toxicology Program). 2010. Report on Carcinogens Background Docu- ment for Formaldehyde, January 22, 2010. U.S. Department of Health and Human Services, Public Health Service, National Toxicology Program, Research Triangle Park, NC [online]. Available: http://ntp.niehs.nih.gov/ntp/roc/twelfth/2009/Novem ber/Formaldehyde_BD_Final.pdf [accessed July 17, 2013]. NTP (National Toxicology Program). 2011. Formaldehyde. Pp. 195-205 in Report on Car- cinogens, 12th Ed. U.S. Department of Health and Human Services, Public Health Service, National Toxicology Program, Research Triangle Park, NC [online]. Avail- able: http://ntp.niehs.nih.gov/ntp/roc/twelfth/profiles/formaldehyde.pdf [accessed Ju- ly 17, 2013]. Olsen, J.H., and S. Asnaes. 1986. Formaldehyde and the risk of squamous cell carcinoma of the sinonasal cavities. Br. J. Ind. Med. 43(11):769-774. Olsen, J.H., S.P. Jensen, M. Hink, K. Faurbo, N.O. Breum, and O.M. Jensen. 1984. Oc- cupational formaldehyde exposure and increased nasal cancer risk in man. Int. J. Cancer 34(5):639-644. Ott, M.G., M.J. Teta, and H.L. Greenberg. 1989. Lymphatic and hematopoietic tissue cancer in a chemical manufacturing environment. Am. J. Ind. Med. 16(6):631-643. Pala, M., D. Ugolini, M. Ceppi, F. Rizzo, L. Maiorana, C. Bolognesi, T. Schiliro, G. Gil- li, P. Bigatti, R. Bono, and D. Vecchio. 2008. Occupational exposure to formalde- hyde and biological monitoring of Research Institute workers. Cancer Detect. Prev. 32(2):121-126. Partanen, T., T. Kauppinen, R. Luukkonen, T. Hakulinen, and E. Pukkala. 1993. Malig- nant lymphomas and leukemias, and exposures in the wood industry: An industry- based case-referent study. Int. Arch. Occup. Environ. Health 64(8):593-596. 

Independent Assessment of Formaldehyde 175 Patterson, D.L., E.A. Gross, M.S. Bogdanffy, and K.T. Morgan. 1986. Retention of for- maldehyde gas by the nasal passages of F344 rats. Toxicologist 6:55. Paustenbach, D., Y. Alarie, T. Kulle, N. Schachter, R. Smith, J. Swenberg, H. Witschi, and S.B. Horowitz. 1997. A recommended occupational exposure limit for formal- dehyde based on irritation. J. Toxicol. Environ. Health 50(3):217-263. Pesch, B., C.B. Pierl, M. Gebel, I. Gross, D. Becker, G. Johnen, H.P. Rihs, K. Donhuijsen, V. Lepentsiotis, M. Meier, J. Schulze, and T. Bruning. 2008. Occupa- tional risks for adenocarcinoma of the nasal cavity and paranasal sinuses in the German wood industry. Occup. Environ. Med. 65(3):191-196. Peto, J., H. Seidman, and I.J. Selikoff. 1982. Mesothelioma mortality in asbestos work- ers: Implications for models of carcinogenesis and risk assessment. Br. J. Cancer 45(1):124-135. Pinkerton, L.E., M.J. Hein, and L.T. Stayner. 2004. Mortality among a cohort of garment workers exposed to formaldehyde: An update. Occup. Environ. Med. 61(3):193-200. Pongsavee, M. 2011. In vitro study of lymphocyte antiproliferation and cytogenetic effect by occupational formaldehyde exposure. Toxicol. Ind. Health 27(8):719-723. Qian, R.J., P.H. Zhang, T.L. Duang, and N.L. Yao. 1988. Investigation on occupational hazards of formaldehyde exposure [abstract]. Ind. Hyg. Occup. Dis. 14(2):101. Rager, J.E., B.C. Moeller, M. Doyle-Eisele, D. Kracko, J.A. Swenberg, and R.C. Fry. 2013. Formaldehyde and epigenetic alterations: MicroRNA changes in the nasal epithelium of nonhuman primates. Environ. Health Perspect. 121(3):339-344. Ren, X., Z. Ji, C.M. McHale, J. Yuh, J. Bersonda, M. Tang, M.T. Smith, and L. Zhang. 2013. The impact of FANCD2 deficiency on formaldehyde-induced toxicity in human lymphoblastoid cell lines. Arch. Toxicol. 87(1):189-196. Richardson, D.B. 2009. Multistate modeling of leukemia in benzene workers: A simple approach to fitting the 2-stage clonal expansion model. Am. J. Epidemol. 169(1):78-85. Richardson, D.B., C. Terschuren, and W. Hoffmann. 2008. Occupational risk factors for non-Hodgkin's lymphoma: A population-based case-control study in Northern Germany. Am. J. Ind. Med. 51(4):258-268. Rinsky, R.A., A.B. Smith, R. Hornung, T.G. Filloon, R.J. Young, A.H. Okun, and P.J. Landrigan. 1987. Benzene and leukemia: An epidemiologic risk assessment. N Engl. J. Med. 316(17):1044-1050. Roush, G.C., J. Walrath, L.T. Stayner, S.A. Kaplan, J.T. Flannery, and A. Blair. 1987. Nasopharyngeal cancer, sinonasal cancer, and occupations related to formalde- hyde: A case-control study. J. Natl. Cancer Inst. 79(6):1221-1224. Rusch, G.M., J.J. Clary, W.E. Rinehart, and H.F. Bolte. 1983. A 26-week inhalation tox- icity study with formaldehyde in the monkey, rat, and hamster. Toxicol. Appl. Pharmacol. 68(3):329-343. Sasaki, Y., T. Ohtani, Y. Ito, M. Mizuashi, S. Nakagawa, T. Furukawa, A. Horii, and S. Aiba. 2009. Molecular events in human T cells treated with diesel exhaust particles or formaldehyde that underline their diminished interferon-gamma and interleukin- 10 production. Int. Arch. Allergy Immunol. 148(3):239-250. Schmid, O., and G. Speit. 2007. Genotoxic effects induced by formaldehyde in human blood and implications for the interpretation of biomonitoring studies. Mutagenesis 22(1):69-74. Schroeter, J.D., J.S. Kimbell, E.A. Gross, G.A. Willson, D.C. Dorman, Y.M. Tan, and H.J. Clewell III. 2008. Application of physiological computational fluid dynamics models to predict interspecies nasal dosimetry of inhaled acrolein. Inhal. Toxicol. 20(3):227-243. 

176 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens Sellakumar, A.R., C.A. Snyder, J.J. Solomon, and R.E. Albert. 1985. Carcinogenicity of formaldehyde and hydrogen chloride in rats. Toxicol. Appl. Pharmacol. 81(3 Pt 1):401-406. She, Y., Y. Li, Y. Liu, G. Asai, S. Sun, J. He, Z. Pan, and Y. Cui. 2013. Formaldehyde induces toxic effects and regulates the expression of damage response genes in BM-MSCs. Acta Biochim. Biophys. Sin. 45(12):1011-1020. Siew, S.S., T. Kauppinen, P. Kyyronen, P. Heikkila, and E. Pukkala. 2012. Occupational exposure to wood dust and formaldehyde and risk of nasal, nasopharyngeal, and lung cancer among Finnish men. Cancer. Manag. Res. 4:223-232. Silver, S.R., R.A. Rinsky, S.P. Copper, R.W. Homung, and D. Lai. 2002. Effects of follow- up time on risk estimates: A longitudinal examination of the relative risks of leuke- mia and multiple myeloma in a rubber hydrochloride cohort. Am. J. Ind. Med. 42(6):481-489.Smith, T.J., and D. Kriebel. 2010. A Biologic Approach to Environ- mental Assessment and Epidemiology. New York: Oxford University Press. Smith, T.J., and D. Kriebel. 2010. A Biologic Approach to Environmental Assessment and Epidemiology. New York: Oxford University Press. Soffritti, M., C. Maltoni, F. Maffei, and R. Biagi. 1989. Formaldehyde: An experimental multipotential carcinogen. Toxicol. Ind. Health 5(5):699-730. Soffritti, M. F. Belpoggi, L. Lambertin, M. Lauriola, M. Padovani, and C. Maltoni. 2002. Results of long-term exposreimental studies on the carcinogeneicity of formalde- hyde and acetaldehyde in rats. Ann. N.Y. Acad. Sci. 982:87-105. Souza, A., and R. Devi. 2014. Cytokinesis blocked micronucleus assay of peripheral lymphocytes revealing the genotoxic effect of formaldehyde exposure. Clin. Anat. 27(3):308-312. Speit, G., P. Schutz, J. Hogel, and O. Schmid. 2007. Characterization of the genotoxic potential of formaldehyde in V79 cells. Mutagenesis 22(6):387-394. Speit, G., S. Neuss, and O. Schmid. 2010. The human lung cell line A549 does not devel- op adaptive protection against the DNA-damaging action of formaldehyde. Envi- ron. Mol. Mutagen. 51(2):130-137. Stayner, L., A.B. Smith, G. Reeve, L. Blade, L. Elliott, R. Keenlyside, and W. Halperin. 1985. Proportionate mortality study of workers in the garment industry exposed to formaldehyde. Am. J. Ind. Med. 7(3):229-240. Stayner, L.T., L. Elliott, L. Blade, L., R. Keenlyside, and W. Halperin. 1988. A retrospec- tive cohort mortality study of workers exposed to formaldehyde in the garmet in- dustry. Am. J. Ind. Med. 13(6):667-681. Stellman, S.D., P.A. Demers, D. Colin, and P. Boffetta. 1998. Cancer mortality and wood dust exposure among participants in the American Cancer Society Cancer Preven- tion Study II (CPS-II). Am. J. Ind. Med. 34(3):229-237. Stern, F., J. Beaumont, W. Halperin, L. Murthy, B. Hills, and J. Fajen. 1987. Mortality of chrome leather tannery workers and chemical exposure in tanneries. Scand. J. Work Environ. Health 13(2):108-117. Stern, F.B. 2003. Mortality among chrome leather tannery workers: An update. Am. J. Ind. Med. 44(2):197-206. Stewart, P., A. Blair, D. Cubit, R. Bales, S. Kaplan, J. Ward, W. Gaffey, M. O'Berg, and J. Walrath. 1986. Estimating historical exposures to formaldehyde in a retrospec- tive mortality study. Appl. Ind. Hyg. 1(1):34-41. Stewart, P.A., R.F. Herrick, C.E. Feigley, D.F. Utterback, R. Hornung, H. Mahar, R. Hayes, D.E. Douthit, and A. Blair. 1992. Study design for assessing exposures of embalmers for a case-control study. Part I. Monitoring results. Appl. Occup. Envi- ron. Hyg. 7(8):532-540. 

Independent Assessment of Formaldehyde 177 Stroup, N.E., A. Blair, and G.E. Erikson. 1986. Brain cancer and other causes of death in anatomists. J. Natl. Cancer Inst. 77(6):1217-1224. Swenberg, J.A., B.C. Moeller, K. Lu, J.E. Rager, R.C. Fry, and T.B. Starr. 2013. Formal- dehyde carcinogenicity research: 30 years and counting for mode of action, epide- miology, and cancer risk assessment. Toxicol. Pathol. 41(2):181-189. Takahashi, M., R. Hasegawa, F. Furukawa, K. Toyoda, H. Sato, and Y. Hayashi. 1986. Effects of ethanol, potassium metabisulfite, formaldehyde and hydrogen peroxide on gastric carcinogenesis in rats after initiation with N-methyl-N'-nitro-N- nitrosoguanidine. Jpn J. Cancer Res. 77(2):118-124. Tang, L.X., and Y.S. Zhang. 2003. Health investigation on workers exposed to formalde- hyde. Occup. Health 19(7):34-35. Tang, X., Y. Bai, A. Duong, M.T. Smith, L. Li, and L. Zhang. 2009. Formaldehyde in China: Production, consumption, exposure levels, and health effects. Environ. Int. 35(8):1210-1224. Thomas, D.C. 2009. Some special-purpuse design. Pp. 92-109 in Statistical Methods in Environmental Epidemiology, 1st Ed. Oxford: Oxford University Press. Thompson, C.M., R.P. Subramanian, and R.C. Grafstrom. 2008. Mechanistic and dose considerations for supporting adverse pulmonary physiology in response to for- maldehyde. Toxicol. Appl. Pharmacol. 233(3):355-359. Til, H.P., R.A. Woutersen, V.J. Feron, V.H. Hollanders, H.E. Falke, and J.J. Clary. 1989. Two-year drinking-water study of formaldehyde in rats. Food Chem. Toxicol. 27(2):77-87. Tobe, M., K. Naito, and Y. Kurokawa. 1989. Chronic toxicity study on formaldehyde administered orally to rats. Toxicology 56(1):79-86. Tong, Z.M., S.X. Zhu, and J. Shi. 2007. Effect of formaldehyde on blood component and blood biochemistry of exposed workers. Chin. J. Ind. Med. 20(6):409-410. Vargova, M., S. Janota, J. Karelova, M. Barancokova, and M. Sulcova. 1992. Analysis of the health risk of occupational exposure to formaldehyde using biological markers. Analusis 20(8):451-454. Vaughan, T.L., C. Strader, S. Davis, and J.R. Daling. 1986a. Formaldehyde and cancers of the pharynx, sinus and nasal cavity: I. Occupational exposures. Int. J. Cancer 38(5):677-683. Vaughan, T.L., C. Strader, S. Davis, and J.R. Daling. 1986b. Formaldehyde and cancers of the pharynx, sinus and nasal cavity: II. Residential exposures. Int. J. Cancer 38(5):685-688. Vaughan, T.L., P.A. Stewart, K. Teschke, C.F. Lynch, G.M. Swanson, J.L. Lyon, and M. Berwick. 2000. Occupational exposure to formaldehyde and wood dust and naso- pharyngeal carcinoma. Occup. Environ. Med. 57(6):376-384. Viegas, S., C. Ladeira, C. Nunes, J. Malta-Vacas, M. Gomes, M. Brito, P. Mendonca, and J. Prista. 2010. Genotoxic effects in occupational exposure to formaldehyde: A study in anatomy and pathology laboratories and formaldehyde-resins production. J. Occup. Med. Toxicol. 5(1):25. Walrath, J., and J.F. Fraumeni, Jr. 1983. Mortality patterns among embalmers. Int. J. Cancer 31(4):407-411. Walrath, J., and J.F. Fraumeni, Jr. 1984. Cancer and other causes of death among em- balmers. Cancer Res. 44(10):4638-4641. Watanabe, F., T. Matsunaga, T. Soejima, and Y. Iwata. 1954. Study of the carcinogenici- ty of aldehyde. 1. Experimentally produced rat sarcomas by repeated injections of aqueous solution of formaldehyde [in Japanese]. Gan. 45(2-3):451-452. 

178 Review of the Formaldehyde Assessment in the NTP 12th Report on Carcinogens West, S., A. Hildesheim, and M. Dosemerci. 1993. Non-viral risk factors for nasopharyn- geal carcinoma in the Philippines: Results from a case-control study. Int. J. Cancer 55(5):722-727. Wilmer, J.W., R.A. Woutersen, L.M. Appelman, W.K. Leeman, and V.J. Feron. 1989. Sub- chronic (13-week) inhalation toxicity study of formaldehyde in male rats: 8-hour in- termittent versus 8-hour continuous exposures. Toxicol. Lett. 47(3):287-293. Woutersen, R.A., L.M. Appelman, J.W. Wilmer, H.E. Falke, and V.J. Feron. 1987. Sub- chronic (13-week) inhalation toxicity study of formaldehyde in rats. J. Appl. Toxi- col. 7(1):43-49. Woutersen, R.A., A. van Garderen-Hoetmer, J.P. Bruijntjes, A. Zwart, and V.J. Feron. 1989. Nasal tumors in rats after severe injury to the nasal mucosa and prolonged exposure to 10ppm formaldehyde. J. Appl. Toxicol. 9(1):39-46. Wu, Y., H. You, P. Ma, L. Li, Y. Yuan, J. Li, X.Ye, X. Liu, H. Yao, R. Chen, K. Lai, and X. Yang. 2013. Role of transient receptor potential ion channels and evoked levels of neuropeptides in a formaldehyde-induced model of asthma in BALB/c mice. PLoS One 8(5):e62827. Yang, W.H. 2007. Hemogram of workers exposed to low concentration of formaldehyde. Pract. Prev. Med. 14(3):792-799. Ye, X., W. Yan, H. Xie, M. Zhao, and C. Ying. 2005. Cytogenetic analysis of nasal mu- cosa cells and lymphocytes from high-level long-term formaldehyde exposed workers and low-level short-term exposed waiters. Mutat. Res. 588(1):22-27. Ye, X., Z. Ji, C. Wei, C. McHale, S. Ding, R. Thomas, X. Yang, and L. Zhang. 2013. Inhaled formaldehyde induces DNA-protein crosslinks and oxidative stress in bone marrow and other distant organs of exposed mice. Environ. Mol. Mutagen. 54(9):705-718. Ying, C.J., X.L. Ye, H. Xie, W.S. Yan, M.Y. Zhao, T. Xia, and S.Y. Yin. 1999. Lympho- cyte subsets and sister-chromatid exchanges in the students exposed to formalde- hyde vapor. Biomed. Environ. Sci. 12(2):88-94. Zeka, A., R. Gore, and D. Kriebel. 2011. The two-stage clonal expansion model in occu- pational cancer epidemiology: Results from three cohort studies. Occup. Environ. Med. 68(8):618-624. Zeller, J., S. Neuss, J.U. Mueller, S. Kühner, K. Holzmann, J. Högel, C. Klingmann, T. Bruckner, G. Triebig, and G. Speit. 2011a. Assessment of genotoxic effects and changes in gene expression in humans exposed to formaldehyde by inhalation un- der controlled conditions. Mutagenesis 26(4):555-561. Zeller, J., A. Ulrich, J.U. Mueller, C. Riegert, S. Neuss, T. Bruckner, G. Triebig, and G. Speit. 2011b. Is individual nasal sensitivity related to cellular metabolism of for- maldehyde and susceptibility towards formaldehyde-induced genotoxicity? Mutat. Res. 723(1):11-17. Zhang, L.P., X.J. Tang, N. Rothman, R. Vermeulen, Z. Ji, M. Shen, C. Qiu, W. Guo, S. Liu, B. Reiss, L.B. Freeman, Y. Ge, A.E. Hubbard, M. Hua, A. Blair, N. Galvan, X. Ruan, B.P. Alter, K.X. Xin, S. Li, L.E. Moore, S. Kim, Y. Xie, R.B. Hayes, M. Azuma, M. Hauptmann, J. Xiong, P. Stewart, L. Li, S.M. Rappaport, H. Huang, J.F. Fraumeni, Jr., M.T. Smith, and Q. Lan. 2010. Occupational exposure to for- maldehyde, hematotoxicity, and leukemia-specific chromosome changes in cul- tured myeloid progenitor cells. Cancer Epidemiol. Biomarkers Prev. 19(1):80-88. Zhang, Y,, X. Liu, C. McHale, R. Li, L. Zhang, Y. Wu, X. Ye, X. Yang, and S. Ding. 2013. Bone marrow injury induced via oxidative stress in mice by inhalation expo- sure to formaldehyde. PLoS One 8(9):e74974.