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.
Keeping Pace with Science and Engineering. 1993. Pp. 189-220. Washington, DC: National Academy Press. Formaldehyde Science: From the Laboratory to the Regulatory Arena Susan W. Putnam and John D. Graham In 1979 the Chemical Industry Institute of Toxicology (CIITJ, a re- search organization funded primarily by a consortium of chemical corpora- tions, released the interim results of rodent bioassays indicating that expo- sure to formaldehyde (HCHO) causes nasal cancer in rats. This implication of the chemical as an animal carcinogen raised immediate questions about its potential as a human carcinogen. Over the next decade, several regula- tory agencies struggled with the problem of what to make of these data, and, ultimately, what to do about formaldehyde. While not immediately accepted in all circles, CIIT's initial bioassay data on ECHO ultimately became the foundation of risk assessments used by all of the relevant regulatory agencies. The data left many questions unanswered, however, especially the mechanisms of carcinogenicity and their implications for assessing human risk. In an effort to answer these questions, CIIT expanded its formaldehyde research program into pharma- cokinetics and mechanism of action. These new studies explored biologic issues not only in the rat but also in primates, a species more similar to humans in physiology and metabolism. Simultaneously, a large epidemiologic study provided new insights into HCHO exposure and human cancer. The pharmacokinetic and mechanistic data have not been readily ac- cepted in the regulatory arena. There also has been controversy over the interpretation of the latest epidemiologic information. The question-of how to interpret all these data in risk assessment has been a hot regulatory issue. Should the new methodologies and results be considered established sci- ence, acceptable and appropriate for regulatory decisions? 189
190 Science and Engineering - Initial CIIT bioassay results 1 975 19~80 ~ - Clll's initial delivered dose (DPX) studies ~ + on rats 1 -1985 - SAB recommends special panel on DPX ~ - EPA special panel concludes DPX data: are premature - Blair epidemiologic study ~ _1 990 - CllT's refined DPX studies, rats and monkeys - _ _ - _ _ ~ - SAB recommends against using DPX data in I ~ _ risk assessment - ClFr's cell proliferation studies SUSAN W. PUTNAM AND JOHN D. GRAHAM Policy and Regulation - CPSC bans UFFI - EPA (Gorsuch) will not designate HCHO a priority chemical under TSCA - Court overturns CPSC ban - EPA (Ruckelshaus) prioritizes HCHO - EPA draft risk assessment to SAB - Court ruling forces OSHA action - OSHA lowers standard from 3 ppm to 1 ppm - EPA finalizes risk assessment - EPA updates risk assessment with DPX; submits to SAB · - California HCHO risk assessment - OSHA standard to 0.75 ppm FIGURE 1 Timeline of significant events in formaldehyde regulation. The way in which one interprets the new science can make a big differ- ence in human risk estimates. If the new animal results are incorporated into risk assessments for formaldehyde using one of the most prevalent interpretations of the data, the human risk estimates for the chemical are 10 to 100 times lower than if they are not included. Advocates of the animal data champion the scientific advances that they embody; opponents fear the data underrate the risk that formaldehyde poses for humans. This case study examines the efforts (and reluctance) to incorporate the biologic data into the regulatory arena. Tracing the scientific and regulatory histories of fo~aldehyde, the study explores the key issues and current debate concerning the appropriateness of the new science for use in regula- tory action. Figure 1 shows the timeline of significant events in the formal- dehyde story. BACKGROUND A UBIQUITOUS AND IRRITATING CHEMICAL Formaldehyde is an omnipresent and versatile chemical that is pro- duced naturally by the human body (Cascieri and Clary, 19921. It is manu- factured and used by industry to create a variety of products, such as plas i,..
FORMALDEHYDE SCIENCE 191 tics, adhesive resins for plywood and particleboard, permanent-press fab- rics, and numerous household products. It is also widely used as an em- balming fluid. In addition, HCHO is generated by multiple sources, such as incomplete combustion, and the chemical is a major component of cigarette smoke. As a consequence, formaldehyde is ubiquitous in the ambient air of both indoor and outdoor environments, often in significant concentrations. The irritant properties of formaldehyde have long been recognized. Consumers exposed to products containing the chemical, as well as workers employed in industries using the substance, may be susceptible to irritation of the eyes, skin, or respiratory system as a result of exposure to the chemi- cal. Because of these irritant effects, much research had been focused on the acute toxicity of formaldehyde (National Research Council tNRC], 19801. The acute effects of the chemical have been fully acknowledged by the regulatory agencies. Based on irritant effects, the Occupational Safety and Health Administration (OSHA) adopted, in 1972, a worker exposure stan- dard for the chemical of 3 parts per million (ppm) (29 C.F.R. 1910.1000tb] Table Z-2~. The Consumer Product Safety Commission (CPSC) was simi- larly troubled by the irritant effects that consumers were experiencing in their homes. Of particular concern was consumer exposure to urea-formal- dehyde foam insulation that was being installed in many homes by the late 1970s to conserve energy. These were often mobile or factory-built homes for elderly or low income residents. THE INDICTMENT- IS FORMALDEHYDE A HUMAN CARCINOGEN? Although the irritant effects of formaldehyde were well established, few studies had explored the question of the chemical's carcinogenicity. The CIIT rodent bioassay was the first persuasive indication of formaldehyde's carci- nogenic potential. The release of the interim bioassay results in 1979 un- leashed a new and volatile chapter in the scientific history of formaldehyde. The results spawned a wealth of both scientific and regulatory debate. The CIIT study was a two-year inhalation bioassay exposing both rats and mice to several exposure levels of HCHO (Kerns et al., 1983~. The bioassay results are reported in Table 1. By the end of the study, half of the 206 rats exposed to the highest concentration of the chemical, 14.3 ppm, had developed squamous cell carcinomas of the nasal cavity. Two rats developed similar tumors at the next lower exposure level, 5.6 ppm. At the lowest exposure levels, no malignant tumors were apparent. Neither were there many tumors in the mice, even at the 5.6 and 14.3 ppm concentrations. There was an elevated incidence of polypoid adenomas (benign tumors) in the experimental rats at all the exposure levels, although no dose-response relationship was apparent.
92 SUSAN W. PUTNAM AND JOHN D. GRAHAM TABLE 1 Cancer Incidence in Rodents Following Inhalation of Formaldehyde Formaldehyde Concentration (ppm)a Number of Tumors/Animals at Riskb Rats Mice 0 0/208 (0%) 0/72 (0%) 2 0/210 (0%) 0/64 (0%) 6 2/210 (1%) 0/73 (0%) 15 103/206 (50%) 2/60 (3.3%) NOTE: Inhalation was for 6 hours/day, 5 days/week. The study was initiated with 960 Fischer-344 rats and 960 B6C3F1 mice, evenly divided by sex into treatment groups. Duration of the study was 24 months. aTarget concentrations. Actual average measured concentrations were 0, 2.0, 5.6, and 14.3 ppm. bActual number of animals exposed to formaldehyde up to, and including, the interval when the first squamous cell carcinomas were found (11-12 months for rats; 23-24 months for mice). SOURCE: Gibson (1983, p. 297). The results of the long-term bioassay experiments were interpreted to demonstrate that, at least for Fischer-344 rats (a specific strain) exposed to high concentrations of the chemical, formaldehyde was an animal carcino- gen. This conclusion was later corroborated by other studies and expanded to include the induction of similar squamous cell tumors in different strains of rats (Albert et al., 1982; Tobe et al., 1985~. The CIIT data raised many questions. Of particular concern was the nonlinear tumor response as the HCHO concentrations were increased. The highest exposure level, 14.3 ppm, was two and a half times higher than the next level (5.6 ppm), yet there was a fiftyfold difference in tumor response. To complicate matters further, there were no malignant tumors at the lowest exposure levels. What did this imply for the shape of the dose-response curve, particularly at low concentrations? Did this suggest that a threshold exists for the chemical's carcinogenic activity? Was formaldehyde "safe" at low exposure levels? Another concern was raised about the paucity of tumors in the mice. Were the squamous cell carcinomas specific to the physiology and metabo- lism of the rat (e.g., the rat is an obligatory nose breather), or might other species have similar tumor responses? With the rat and mouse tumor rates differing so greatly, how should the results be extended to other species? Most important, given such issues, how should these data be interpreted for humans, who were most generally exposed to chronic low concentrations of
FORMALDEHYDE SCIENCE 193 HCHO, although levels in the workplace and home were often within a factor of ten of 5.6 ppm? NEW REVELATIONS In an effort to answer these and other questions, major scientific inves- tigations on formaldehyde continued in several critical areas. Some of this research, such as the studies exploring pharmacokinetics and mechanisms of cancer, offered groundbreaking methodologies and ideas that were at the frontiers of science. Other work, such as the several major epidemiologic studies, contributed much larger data sets and analytical power than were previously available. Taken together, the studies offered significant new insights on the potential human carcinogenicity of formaldehyde. Delivered Dose (DPX) One of the major areas of formaldehyde research at CIIT was pharma- cokinetics. Using biochemical methods, scientists can trace the uptake, metabolism, and distribution of HCHO in the body, including the interac- tion of the chemical with DNA in the target cells of various tissues. The first pharmacokinetic studies, performed by Casanova, Heck, and their colleagues at CIIT, explored the relationship between the amount of formaldehyde that the animals were exposed to the administered dose- and the dose of the chemical that actually reached the target tissue cells in the rat nose the delivered dose. Using radiolabeled HCHO, the research- ers measured the amount of covalent binding of the chemical to rat nasal mucosal DNA and formation of DNA-protein cross-links) (Casanova-Schmitz et al., 1984~. It was hypothesized that the covalent binding would not only be a surrogate measure of the delivered concentration of the chemical, but that it was also related in some way to the observed nasal tumors in the original rodent studies. The covalent binding was assessed for acute (6- hour) exposures in the Fischer-344 rat. The concentrations of administered formaldehyde were similar to those used in the initial rodent bioassay. The results of these studies indicated that the delivered dose was not linearly related to the administered concentration of the chemical. This nonlinearity was particularly manifest for the lower exposures, as Figure 2 indicates. For example, the concentration of formaldehyde covalently bound to DNA at 6 ppm was 10.5-fold higher than at the 2 ppm exposure level. In other words, the DNA binding at the 2 ppm level was significantly lower than the amount predicted by drawing a straight line from 6 ppm to the Orlgln. These data were less than ideal for use in risk assessment. It is difficult to extrapolate from short-term exposures in small groups of animals to the
94 SUSAN W. PUTNAM AND JOHN D. GRAHAM 0.8 0.7 _ ' 0.6( :,,,:t ~, ~ 0.5 / I ~/ cam 0.4 _ ~ ~ a a 0~3 ~ / 0 ~5 - 1/ =o 02 j;j~= 0.1 o 4 6 8 10 12 14 16 Administered Dose (ppm) FIGURE 2 Delivered dose results. SOURCE: Casanova-Schmitz et al. (1984, p. 38~. chronic exposures generally experienced by humans. However, the indica- tion of nonlinearity in the DNA-protein cross-link (DPX) data held signifi- cant implications for the cancer risk associated with formaldehyde. If these data were an accurate representation of the dose reaching the tissue after exposure to low concentrations of the chemical, then there may actually be less response at low doses than the linear curve predicted. This idea was in fact illustrated in a risk assessment performed by Starr and Buck of CIIT. The analysis, summarized in Table 2, showed that cancer risk estimates using the new DPX data were at least an order of magnitude lower than those using the administered doses (Starr and Buck, 1984~. Subsequent studies on glutathione depletion explored the effect of glu- tathione as a defense mechanism in the formation of DNA-protein cross- links (Casanova and Heck, 19871. Other work examined the "isotope ef- fect" in the use of formaldehyde labeling to quantify the DNA cross-linking process (Heck and Casanova, 1987), and augmented the measurement of this phenomenon by the use of improved experimental procedures and tech- nology (Casanova et al., 1989~. This new work built upon the initial con- cept explored in the earlier delivered-dose studies and developed superior measurement schemes to quantify the results. From CIIT's perspective, these additional studies served to further support the nonlinear relationship between the administered and delivered doses of the chemical. The CIIT scientists argued that this relationship may be even more nonlinear than was initially perceived.
FORMALDEHYDE SCIENCE 195 In addition to the pharmacokinetic experiments in rats, CIIT researchers conducted similar experiments on DNA-protein cross-links in six rhesus monkeys (Casanova et al., 1991~. Since there are marked differences be- tween humans and rats with respect to nasal anatomy and respiratory physi- ology, monkeys were selected as experimental subjects in an effort to use a species more closely related to humans. The rhesus monkey is biochemi- cally and physiologically similar to humans, particularly in the important target site areas for formaldehyde (e.g., the nose and upper respiratory tract). The results of the monkey studies were quite similar to what was pre- dicted using the rat studies and interspecies scaling factors (e.g., body weight). There were important differences in the results, however. The monkey data indicated a lower rate of DNA-protein cross-link formation than in the rat, particularly for the low concentrations of formaldehyde. Monkeys also exhibited a wider distribution of cross-links and lesions in the upper respi- ratory tract than did the rats (Casanova et al., 1991~. The monkey DPX data also held significant implications for HCHO cancer risks. The lower rate of cross-link formation and larger target area in the monkey were used to predict a lower cancer risk for primates, and, ultimately, an even lower one for humans than was estimated using the rat data. For example, in the 1987 EPA risk assessment for formaldehyde that did not incorporate the pharmacokinetic data, the upper-bound risk of hu- man cancer at 0.1 ppm exposure was 1.6 x 10-2. In the 1991 analysis, the upper-bound risk estimate at the same exposure level using the rat DPX data was 2.8 x 10-3. Using the monkey-based data, the risk estimate was 3.3 x 10-4. The DPX data, especially those for the monkey, result in a much smaller estimate of human risk than do the initial bioassay data: a sixfold TABLE 2 Formaldehyde Risk Assessmenta Maximum Exposure Likelihood Upper 95% Level Dose Estimate Confidence (ppm) Metric of Risk Limit 0.1 Administered 2.5 x 10-7 1.6 x 10-4 Delivered 4.7 x 10-9 6.2 x 10-5 1.0 Administered 2.5 x 10-4 1.8 x 10-3 Delivered 4.7 x 10-6 6.2 x 10-4 aThree-stage multistage model fitted to CIIT rat bioassay data using either administered or delivered dose as the dose metric. Risk is the lifetime prob- ability of a malignant tumor for chronic exposure at the stated level. SOURCE: Starr and Buck ( 1984).
96 SUSAN W. PUTNAM AND JOHN D. GRAHAM decrease with the rat data, a fiftyfold decrease with the monkey data (EPA, 1991~. Other models using the pharmacokinetic data on HCHO consistently produce similar or even lower estimates (Starr, 1990~. Mechanisms of Carcinogenesis While pharmacokinetic studies are useful, the key area of scientific investigation on formaldehyde involved the mechanisms of carcinogenesis. What are the biological mechanisms responsible for the growth of the nasal carcinomas in rats? At the heart of the research lay two viable hypotheses: mutagenesis and cell proliferation. Scientists ascribing to the mutagenicity theory argue that contact be- tween formaldehyde and the DNA in cells could induce mutations in the critical genes that could ultimately result in the growth of cancerous tu- mors. There have been several studies exhibiting the mutagenic potential of the chemical (Goldmacher and Thilly, 1983~. Proponents of this work sug- gested that the nasal carcinomas seen in the rat bioassays may be the result of the mutagenic capability of formaldehyde. How-ever, some scientists saw formaldehyde as only a weak mutagen (Consensus Workshop on Formalde- hyde, 1984~. Because the chemical is so highly reactive in the body, it is difficult to accurately explore its mutagenic potency. Mutagenicity is important in assessing human risk. If one believes in the mutagenic capability of formaldehyde to the extreme, then every mol- ecule of the chemical has the potential to induce mutations in the DNA with which it comes in contact. While cancer development is a multistage pro- cess, there may be other processes going on in the body to which the chemi- cal exposure adds the final step. Assuming repair mechanisms are imper- fect, even low exposures to formaldehyde could induce cancerous tumors. No concentration would then be safe for human exposure. Proponents of this view argue that there may be a linear dose-response curve; if so, there would be some potential risk at even the smallest exposure levels. Although there were no tumors seen at the low doses in the animal bioassays, the potential for tumors occurring in a larger sample of rats tested at these concentrations could not be ruled out. Another theory of formaldehyde carcinogenesis focuses on cell prolif- eration. Proponents argue that with the increased cell growth and division induced by the onslaught of a toxic substance, there is an increased fre- quency of spontaneous DNA mutations. Also, because of the high rate of cell reproduction, there is less time for DNA defense mechanisms to repair the critical mutations that could lead to a carcinogenic response. Elevated rates of cell proliferation may enhance both the likelihood of interaction of formaldehyde with DNA and the fixation of adducts before DNA repair could occur (Monticello et al., 1989~. Since cell proliferation is often con - -
FORMALDEHYDE SCIENCE 197 sidered a nonlinear process, the implications for risk assessment are pro- found. Several studies in the late 1980s and early l990s (e.g., Monticello and Morgan, 1990; Monticello et al., 1991; Swenberg, 1986) explored this phe- nomenon. Researchers at CIIT examined the relationship between the rate of formaldehyde-induced cell proliferation and the cancerous tumors seen in the initial bioassays. The studies demonstrated that acute and subchronic exposure to formaldehyde induced nasal epithelial lesions and increases in surface cell proliferation rates in rat nasal passages at HCHO concentrations of 6 ppm and higher. As Figure 3 indicates, significant elevations in cell proliferation rates were not detected at concentrations lower than 6 ppm. This result correlated well with the lack of tumors observed below the 5.6 ppm concentration in the earlier bioassay. Increased cell proliferation rates beginning at 6 ppm were also seen in monkeys (Monticello et al., 1989~. CIIT scientists interpreted the correlation between tumor responses in the original bioassay and sustained increases in cell proliferation as evi- dence that cell proliferation plays a causative role in the carcinogenic pro- cess. If this were the case, then the absence of malignant tumors below the 6 ppm exposure level may be related to the lack of detectable increases in cell proliferation rates. The implication of this hypothesis is that there may 70t 65 60 - 55 a, 50 a) 45 AL 40 a' 35 a) 30 c, 25 20 15 10 5 o Tumor Incidence, 24 mot (Kerns et al., 1983) Cell Proliferation, 12 mot (Monticello, 1990) E} Cell Proliferation, 18 mot (Monticello, 1990) 0 2 1 1 1 ,4 I 1 1 1 1 1 1 1 11 eK 10 ° 9 ~ 8 ~a) 7 ° 6 _ a) 5 () 4 3 2 1 lo 14 16 4 6 8 10 12 Formaldehyde Concentration (ppm) FIGURE 3 Tumor incidence and cell proliferation in rats exposed to formalde- hyde. *Cell proliferation is measured by the mean unit length labeling index at nasal level II (fold increase over control). SOURCE: Environmental Protection Agency (1991, p. 301.
198 SUSAN W. PUTNAM AND JOHN D. GRAHAM be some level-a threshold level below which no stimulation or increased cell proliferation, and hence no cancer, occurs. Like the DPX data, these new data suggested a nonlinear dose-response curve for formaldehyde at low concentrations. In contrast to those believing in the mutagenic poten- tial of the chemical, proponents of cell proliferation argued that exposure to formaldehyde may be safer at low concentrations than is predicted by linear models. Cell proliferation data have as yet been difficult to incorporate into risk assessment procedures. Critics assert that while tumor formation as a result of HCHO exposure is preceded by increased cell proliferation, increased cell proliferation may not necessarily cause tumor formation. Also, the current risk assessment process uses mathematical models based on the principles of mutagenesis. The models incorporate a linear dose-response curve, which projects some level of risk even at very low concentrations of the chemical. Preliminary speculation, however, theorizes that if the cell proliferation data with their potential threshold level were incorporated into the analysis, the risk at low concentrations would be reduced, possibly to zero. The risk estimates for human cancer would be lower. Using both the cell proliferation and delivered-dose data, the overall implication for human risk assessment is that inhalation of small doses of formaldehyde may not be as harmful as risk assessors at regulatory agencies had originally pre dicted. Another possibility is that cell proliferation exacerbates the incidence of tumors at high doses even though mutagenicity plays some role at low doses. Mutagenesis and cell proliferation are not mutually exclusive ideas. The hypothesis that both these factors play a role is consistent with linearity at low doses and curvature at high doses. If both mechanisms are operating, the unknown slope of the dose-response curve at low doses is critical to risk assessment. Epidemiology While a variety of studies were being conducted on laboratory animals, there were several new studies investigating HCHO's effects on humans as well. These epidemiologic studies offered large data sets with which to explore the carcinogenic potential of the chemical, including any nasal tu- mors similar to those seen in rats. Many epidemiologic studies have been conducted on formaldehyde over the past several decades. Using both cohort and case-control designs, the studies focused primarily on groups of workers in a variety of occupational settings. The results of these studies were interpreted as suggesting that HCHO may be a human carcinogen for certain groups of professionals (Consensus Workshop on Formaldehyde, 19841. Based on these studies in conjunc
FORMALDEHYDE SCIENCE 199 lion with the animal data the International Agency for Research on Cancer (IARC) and the U.S. Environmental Protection Agency (EPA) concluded that limited evidence existed for an association between formaldehyde and human cancer, and classified the chemical as a "probable" human carcino- gen. As with many epidemiologic studies, however, the human studies of formaldehyde were criticized as suffering from certain limitations. The major drawbacks were perceived to include small sample sizes, insufficient follow-up of the exposed populations, a low statistical power to detect small relative risks for rare cancers, and an inability to separate the effects of formaldehyde from the myriad confounding substances to which the subject populations were simultaneously being exposed especially cigarette smoke and particulates (EPA, 1987~. Despite these problems, the epidemiologic data base for formaldehyde was much larger than many up to that date and suggested directions for further study. In an effort to provide more conclusive epidemiologic data on formal- dehyde, several large studies were conducted on the chemical in the mid- 1980s (Blair et al., 1986; Stayner et al., 1988; Vaughan et al., 1986~. These studies were designed specifically to detect moderate elevations in underly- ing cancer risk among populations with HCHO exposure in the workplace. According to EPA's interpretation, all three studies revealed statistically significant elevations in the risk of site-specific upper respiratory cancers with measures of HCHO exposure (EPA, 1991~. These studies were not without their critics, however, particularly for their failure to control for cigarette smoking. The most notable and widely cited of these studies was that conducted by Blair and colleagues at the National Cancer Institute, DuPont, and Monsanto (Blair et al., 1986~. The Blair cohort study followed more than 26,500 workers (approximately 600,000 person-years of data) in 10 plants that either produced or used formaldehyde. The HCHO exposure level of the workers varied, probably averaging about 0.5 ppm for the exposed members of the cohort as a whole. The study did not control for smoking. As shown in Table 3, the results of the study indicated slight excesses of cancer in the upper respiratory tract and lungs for persons occupationally exposed to formaldehyde. The results also suggested that the risk of cancer in the nasopharyngeal and sinonasal cavities may be enhanced with simulta- neous exposure to particulates or wood dust. The authors determined, how- ever, that the tumors did not show a consistently rising incidence with level of exposure. They interpreted the results of the study as providing little evidence that cancer mortality was associated with formaldehyde exposure at levels experienced by the workers in the plants under focus. The new study was not without controversy, however. Some members of the study's external review panel, as well as labor unions and other
200 SUSAN W. PUTNAM AND JOHN D. GRAHAM TABLE 3 Mortality from Selected Cancers for White Men Cancer Observed Expected SMRa Respiratory System Exposed 215 192 112 Nonexposed 53 56 95 Buccal Cavity and Pharynx Exposed 18 19 96 Nonexposed 3 6 54 Larynx Lung Exposed 128 142 Nonexposed 43 b Exposed 201182 111 Nonexposed 4953 93 NOTE: The number of persons involved was 16,962 exposed, 5,562 nonexposed. Number of person-years of exposure: 378,497 exposed, 100,237 nonexposed. aStandardized mortality ratio (SMR) is the number of observed cancers divided by the number of expected cancers. An SMR of 100 is interpreted as the norm; the observed number of cancers is the same as the expected number. Given the occupational cohort of the study, however, there may be some bias due to the "healthy worker effect," whereby the study population of workers may be healthier than the average population overall. If this phenomenon were present, then fewer cancers would be expected for the healthy study population. This would in turn raise the standard mortality ratios for the study results. bSMR is not presented when the observed and expected numbers are less than 5. SOURCE: Blair et al. (1986, from Table 2, pp. 1075-1077). groups, criticized the negative interpretation of the results and accused the authors of collaborating too closely with industry. The lack of a consistent dose-response relationship was not seen as detracting from the positive findings presented by the study. Critics of Blair's interpretation also argued that the statistically significant excess of lung cancer in workers (who had an expo- sure latency of at least 20 years) contradicted the authors' characterization of little or no evidence of carcinogenesis (Jasanoff, 1990~. Blair and colleagues subsequently further evaluated some of the data. They stated, for example, that even though the numbers were small, the pattern for nasopharyngeal cancer suggested that simultaneous exposure to formaldehyde and particulates may be a risk factor (Blair et al., 19871. This conclusion was met with opposition as well, however, this time from the owners of the plant where most of the tumors occurred (Collins et al., 1987, 1988~. Despite the controversies that ensued over its interpretation, the Blair _
FORMALDEHYDE SCIENCE 201 study is perceived to be an important piece of epidemiologic research. Con- sidering the study's size, cost, and methodological care, it is doubtful that a similar study of such proportions will soon be conducted on formaldehyde. The data set yielded by the study is enormous and is still being evaluated (e.g., Blair et al., 1990~. The new epidemiologic data hold important implications for risk as- sessment as well. Although the data are difficult to quantify for risk assess- ment estimates, some analyses have been done. The EPA, for example, developed some rough comparisons (using the Blair data) suggesting that observed human risks were concordant with animal-based risk estimates using administered dose levels. The cancer risk estimates using human data were, however, significantly higher than those obtained using the DPX models (EPA, 1990~. THE CONSEQUENCES THE AGENCIES RESPOND To relate the story of regulatory action on HCHO to the new scientific information, it is necessary to include a brief synopsis of agency reactions to the initial bioassay data. Several regulatory agencies have jurisdiction over the chemical, most notably the Consumer Product Safety Commission, the Occupational Safety and Health Administration, and the Environmental Protection Agency. Each of the agencies faced the same scientific data sets, yet their responses were curiously unique. While agencies were aware of the irritant effects of the chemical, it was the release of the CIIT bioassay data in 1979 that brought formaldehyde to the regulatory forefront. The demonstration that the chemical was an ani- mal carcinogen injected a new level of both urgency and controversy into the regulatory debate. Cancer policy guidelines that had been adopted in the late 1970s stated that, in the absence of adequate human data, regulators were to interpret positive and valid animal bioassay data as providing pre- sumptive evidence of human carcinogenic risk (U.S. Regulatory Council, 1979~. Given the high level of economic investment tied to such a widely used chemical, there was a natural hesitancy in the agencies to regulate. The new cancer guidelines, however, required that the agencies take a new look at the chemical and its potential risk to humans. Although many of the agency deliberations and decisions concerning formaldehyde were concurrent and highly intertwined, the events of each agency will be treated separately here in an effort to make the story less confusing. Also, while there were some interagency efforts to coordinate their approach to HCHO risk assessment, such as the Interagency Regula- tory Management Council Workgroup on Formaldehyde, many of the result- ant decisions appeared to be quite unique to specific agencies. Finally, HCHO has had a long and involved history at many of the agencies. For the . ~
202 SUSAN W. PUTNAM AND JOHN D. GRAHAM sake of brevity, this study does not attempt to cover the complete story at each agency. Instead, there is an effort to cover the regulatory history most related to the new scientific information at issue. The EPA receives the greatest focus here because that was where much of the controversy over the new studies ensued. Consumer Product Safety Commission After the release of the rodent bioassay data, the Consumer Product Safety Commission took the most immediate and ambitious regulatory ac- tion. Buoyed by the conclusion that the chemical should be assumed to pose a carcinogenic risk to humans (Federal Panel on Formaldehyde, 1982), the commission banned the use of urea-formaldehyde foam insulation (UFFI) in all homes and schools in 1982 (CPSC, 1982~. The decision was sup- ported by a quantitative assessment indicating a risk of up to 1.8 additional cases of human cancer for every 10,000 homes using urea-formaldehyde foam insulation. The ban was subsequently overturned the following year by the Fifth Circuit Court of Appeals, on the court's determination that the CPSC had failed to support its action with substantial evidence. The court decision stated that the CPSC had erred in relying solely on the CIIT study to establish the risk of human carcinogenicity (Gulf South Insulation v. CPSC, 1983~. Nevertheless, the early efforts of the CPSC represent an ambitious and prompt response to the demonstration of the chemical's car . . . . cinogen~c~ty In Enema s. The CPSC's response to the initial CIIT pharmacokinetic model and data on formaldehyde, on the other hand, was less than enthusiastic. The commission's chief scientist responsible for formaldehyde, Murray Cohn, was one of the most vocal opponents to the use of the initial Casanova studies in the risk assessment process. As the lead author of a published critique of the CIIT studies (Cohn et al., 1985), Cohn was concerned about both the relevance of the short-term experiments to the chronic exposures experienced by humans and the relationship of the DNA-protein cross-links to carcinogenesis. Also, he perceived that there was too much "background noise" in the studies to sort out what was really going on. His conclusion was that the methodologies used were weak and that the data were too preliminary to include in the risk assessment process (personal communica- tion with Cohn, 1992~. With the release of the more recent pharmacokinetic studies, the out- look of the commission-or at least of its key scientist has changed. To- day Cohn argues that the new studies incorporate improved procedures that have helped to clear up some of the methodological issues that CPSC was concerned about in the original Casanova research. The commission has not performed an official risk assessment using a pharmacokinetic model
FORMALDEHYDE SCIENCE 203 and the new data, but it has done unofficial, back-of-the-envelope analyses that do incorporate them. Were a new HCHO regulatory issue to arise, the new analyses would be ready to go (personal communication with Cohn, 1992). It remains to be seen whether formaldehyde will resurface as a key issue for the CPSC. Urea-formaldehyde foam insulation has been largely phased out of the market, as well as removed from many of the homes in which it was used. The commission is currently working with industry on a voluntary emissions standard for urea-formaldehyde particleboard, but this has not required that a formal new risk assessment be done. Occupational Safety and Health Administration In contrast to the CPSC's prompt action following the release of the bioassay results, the Occupational Safety and Health Administration was slower to respond to the CIIT data. Leaders at the agency argued that the cancerous tumors in the bioassay results appeared at a higher exposure level than that faced by humans in the workplace (Graham et al., 19881. It was not until a federal district court ordered OSHA to formally consider initiat- ing a proceeding for a new permanent rule on formaldehyde-in response to a suit against the agency brought by the United Auto Workers (UAW) and other groups seeking an emergency temporary standard for HCHO- that the agency took any formal action on the chemical. The proposed rulemaking procedure for HCHO that ensued in 1985 was based entirely on the chemical's irritant effects, however, not on its cancer risk (OSHA, 19853. It was not until 1987 that the agency used the CIIT bioassay data in regulatory proceedings, when it lowered its standard for formaldehyde from 3 ppm to 1 ppm as an 8-hour, time-weighted average (OSHA, 1987b). The agency's quantitative risk assessment incorporated the CIIT data and indicated that by lowering the standard from 3 to 1 ppm, the lifetime risk of cancer per 100,000 workers would decrease in the range of 70 to 570 cases. The reduction in cancer risk was not inexpensive. OSHA estimated that the annual cost of the new standard would be approximately $62 million. This cost was judged to be economically feasible, however, based on an industry-specific analysis of revenue and profit ratios. The new standard was also perceived to be technologically feasible. The agency foresaw that many of the industries could pass on some of the costs of compliance to purchasers of their products, and the new standard would not adversely affect the viability of most of the firms in the relevant industries. Also, the new standard was not perceived to have a significant adverse effect on a substantial number of small entities nor on the economy as a whole (OSHA, 1987c). -
204 SUSAN W. PUTNAM AND JOHN D. GRAHAM While the 1987 rulemaking process for formaldehyde used the bioassay data, it did not incorporate a pharmacokinetic model and the DPX data into OSHA's risk assessments. OSHA asserted that the DPX studies provided qualitative evidence that HCHO could be a carcinogen, but stated that such information could be used with great caution in quantitatively assessing risk. The agency cited the lack of chronic exposure data, a tumor incidence in rats inconsistent with DNA binding results, a paucity of data identifying DNA cross-links, and several other issues as reasons for not incorporating the new studies into its analysis (OSHA, 1987a). As with the CPSC, OSHA had trouble, not with the concept of using the delivered-dose data, but with the methodologies used in the initial CIIT DPX studies. The most recent OSHA rule on formaldehyde, issued in May 1992 in response to a court remand of the 1987 standard, lowered the standard for occupational exposure to 0.75 ppm (OSHA, 1992~. Using the same risk assessment developed in 1987, the agency estimated that from 0.2 to 72 cancers (depending on whether the maximum likelihood estimate or the upper confidence limit was used) would be avoided over the next 45 years with the standard lowered to 0.75 ppm. Again, while there was a substantial cost to industry, the standard was judged to be both economically and tech- nologically feasible. As in 1987, OSHA again did not include a pharmacokinetic model or the DPX data in its analysis. The agency stated that its reasons behind this omission remain the same as they were in 1987. Moreover, since advocates of the new studies did not claim that they would justify a different standard, the agency did not believe that it was necessary or appropriate to reopen the rulemaking record to include them (OSHA, 1992~. In essence, OSHA has decided that, at least for now, the 1987 risk assessment without the pharmacokinetic model is sufficient. Redoing the risk estimates with the new data would require reopening the rulemaking process. Such an action is financially draining on both the agency and the other parties involved. The agency worked with industry and the relevant unions to resolve the problems with the 1987 standard and to develop a new standard that was economically feasible and satisfactory to all parties. If the agency were pressed into new rulemaking procedures in the future, industry would undoubtedly push for including the pharmacokinetic model and DPX data. While industry has so far been able to accommodate the increasingly restrictive OSHA regulations for HCHO, it would use the new data to defend its position that no further reductions in emissions standards are necessary (personal communication with Formaldehyde Institute, 1992~. As it stands, however, industry is anxious to achieve regulatory certainty and seems willing to live with the current risk assessment (personal commu- nication with OSHA, 1992~. -
FORMALDEHYDE SCIENCE 205 Environmental Protection Agency From the perspective of the new scientific studies and how to interpret them, the regulatory story on formaldehyde is perhaps the most interesting at the Environmental Protection Agency. Under the Toxic Substances Con- trol Act of 1976 (TSCA), the EPA is authorized to regulate chemicals pos- ing an unreasonable risk to health. The agency was slow to act on the initial bioassay data for HClIO, however. Under Administrator Anne Gorsuch, the agency elected not to designate formaldehyde as a priority chemical under Section 4(f) of TSCA and did not initiate any regulatory activity. In 1983 the agency sponsored a work- shop of experts on the chemical in an effort to evaluate the available data and reach some consensus on HCHO. The group failed to resolve key technical disagreements, however, and could not agree on the carcinogenic properties and potential of formaldehyde (Consensus Workshop, 1984~. It was not until the following year, after William Ruckelshaus had taken over as administrator and the agency was facing a lawsuit from the Natural Resources Defense Council and the United Auto Workers, that the EPA reconsidered its decision and assigned high priority to HCHO. The designation of formaldehyde as a priority chemical spurred the agency to take action on several fronts. One of these was the initiation of a continuing regulatory investigation into exposure to HCHO in homes. Over the next several years, risk managers at EPA explored various control op- tions for the chemical and the attending costs for these options. In conjunction with the risk management strategy, the agency simulta- neously developed a quantitative risk assessment of the health effects of formaldehyde, using both the CIIT bioassay data and the available epidemio- logic studies. Written by the Office of Toxic Substances (OTS), the 1985 analysis concluded that the chemical should be seen as a carcinogen in rats and, based on epidemiologic data perceived as "limited," may be a human carcinogen (EPA, 1985~. With the high levels of scientific uncertainty surrounding the chemical and its colorful history, the EPA sent the risk assessment document to its extramural review panel, the Science Advisory Board (SAB). One of the key issues of concern to OTS was how to interpret and incorporate the model using the pharmacokinetic data on formaldehyde into the risk assess- ment process. The agency had chosen not to include this model, but was seeking advice on the issue. The Environmental Health Committee of the SAB gave the document a generally favorable review. Its response to the issue of incorporating a pharmacokinetic model and data, however, is of particular interest. Several members of the review panel were familiar with the CIIT studies and strongly believed that a model using the data should be incorporated into the analy . ~
206 SUSAN W. PUTNAM AND JOHN D. GRAHAM sis (Transcript of Environmental Health Committee meeting, 1985~. The SAB recommended that a special review panel, independent of the agency, be formed to evaluate the pharmacokinetic data (Environmental Health Committee, 1985~. In collaboration with a scientific consulting firm, Life Systems, Inc., the EPA convened a panel of seven scientists from a variety of disciplines to review the CIIT pharmacokinetic studies. The group was to review the new data to determine whether they were appropriate both for use in assessing the health risks of formaldehyde and for incorporation into the risk assess- ment document (Life Systems, Inc., 1986~. In January 1986 the panel re- leased its report, which was very critical of the CIIT research. Of particular concern were the assumptions and procedures used in the pharmacokinetic studies, particularly the analytic methods used to measure the tissue dose of the chemical. While recognizing that the work represented an important first step toward the introduction of intracellular dosimetry into the risk assessment process, the report concluded that the research did not provide a basis for quantifying risk (Life Systems, Inc., 1986~. CIIT scientists con- tested the panel's conclusion, to no avail (Casanova et al., 1986~. In 1987 the OTS issued a final risk assessment document (EPA, 1987~. Again, a pharmacokinetic model and the DPX data were not included. A1- though several members of the Science Advisory Board, as well as CIIT scientists, did not agree with the conclusions of the special panel, the EPA heeded the panel's recommendation that the data should not be used for risk quantification. The agency maintained its original position that the use of the data for risk assessment was "premature" (DiCarlo, undated). The agency accepted the 1987 analysis as a working document and shifted formalde- hyde over to OTS risk managers to develop policy options for reducing potential residential cancer risks (Putnam, 1991~. Before sending their options analysis to senior EPA decision makers in 1989, however, OTS staff explored what further developments had emerged in formaldehyde science. They wanted to ensure that the office was famil- iar with the latest data on the chemical. Believing that the recent pharma- cokinetic studies were important, the OTS decided that the risk assessment should be updated to include a pharmacokinetic model and the new CIIT data. In the updated risk assessment document released in 1991, the OTS included both the rat and monkey DPX data. The authors of the analysis determined that the more recent CIIT research had clarified most of the concerns expressed by the previous special panel. They concluded that, although some uncertainties remain about the pharmacokinetic data, their use in risk estimates is appropriate. This decision to incorporate the DPX data was reinforced by remaining uncertainties in the epidemiologic data. While the Blair and other large studies provided more human data than had
FORMALDEHYDE SCIENCE 207 previously been available, the OTS determined that there were still not enough epidemiologic exposure data to quantify a dose-response relation- ship for the chemical and, hence, it could not calculate low-dose human cancer risk (EPA, 1991~. To make the risk assessment as biologically plau- sible as possible, the OTS strongly favored the use of the DPX data (per- sonal communication with OTS, 1992~. The OTS also recognized the im- portance of the new cell proliferation data, although the authors determined that they could not yet incorporate these data into their current risk assess- ment models. Again the OTS sent its risk assessment document to be reviewed by the Environmental Health Committee of the agency's Science Advisory Board. In an initial consultation (a less formal review), many of the members of the panel reported favorably on the document and its use of a pharmacokinetic model in the analysis. The document and its review were mired in contro- versy, however. The Science Advisory Board was criticized by the United Auto Workers and other environmental groups as being unbalanced in scien- tific opinion and institutional representation, and biased in its meeting pro- cedures. Of particular concern to these groups was what was perceived to be a heavy weighting of panel members who favored the use of a pharmaco- kinetic model and the DPX data, and a lack of members with an epidemiologic background. The Environmental Health Committee met for formal review of the formaldehyde risk assessment update in July 1991. This time the panel included three new members, each of whom was known to have a conserva- tive view on the use of a pharmacokinetic model in the risk assessment process. The incorporation of such a model and the DPX data was clearly one of the most contentious issues at the meeting, and the panel had a difficult time trying to reach consensus (personal communication with SAB, September 1992~. The SAB report presented a mixed picture. The panel recognized the advances brought by the pharmacokinetic data, but stated that the use of delivered dose measures in quantitative risk assessment remained equivocal, except as a measure of exposure. It also pointed out that because of weak- nesses in the epidemiologic data, such as possible confounding with simul- taneous exposure to particulates, only point estimates with extensive uncer- tainties were likely to be available. The report recommended that the risk estimates based on the DPX data be compared with those derived from the most appropriate epidemiologic studies, such as the Blair study. The panel advocated that the risk assessment document should present estimates and models based on rats, monkeys, and humans, and then compare the advan- tages and disadvantages of each data source and model (Science Advisory Board, 1992~. On the EPA regulatory front, risk managers at OTS realized that the -
208 SUSAN W. PUTNAM AND JOHN D. GRAHAM incorporation of the pharmacokinetic model and DPX data into the risk assessment would lower the estimates of human cancer risk for exposure to formaldehyde. This would in turn lower the perceived benefits of strong regulatory action, at least for those exposed to the low levels typically found in homes. The OTS risk managers began to focus on identifying potential risk reduction actions for HCHO oriented to avoiding noncancer human effects in residential settings. Most recently, while the OTS is working on the latest version of the risk assessment document, risk managers have also been pursuing a regula- tory strategy to address the issue of indoor air quality posed by HCHO emissions from urea-formaldehyde pressed wood building materials, cabi- nets, and furniture. This includes supporting the revision of the emissions standard set by the American National Standards Institute for urea-formal- dehyde particleboard flooring products from 0.3 to 0.2 ppm (Greenwood, 1992). THE STATES ENTER THE RING The risk assessors at OTS were not the only regulatory proponents of the biologic data for formaldehyde. While the EPA and its Science Advi- sory Board were struggling over the appropriateness of the data for the agency's risk assessment, the state of California released its own assessment of the health effects of the chemical. This document similarly explored the use of the new biologic data to estimate HCHO's cancer risk to humans. Like the most recent EPA document, the California analysis used the DNA cross-links (although California used only the rat DPX data) as an indicator of tissue dose. The risk assessment also explored the use of cell proliferation data in developing a range of risk estimates. The authors concluded that the use of these data yielded a more accurate risk assessment (California Air Resources Board, 1992~. The California analysis incorporated a variety of different assumptions than the analysis done by the EPA. One of the key assumptions involved the use of tumors in the analysis: the EPA focused specifically on nasal cancer from the rat data; California tried to account for other cancers in the respiratory tract as well as lung cancer using the rat data and animal-to- human scaling factors (personal communication with California Air Re- sources Board, 19921. The resultant unit risk for a lifetime of exposure estimated by the analysis 7 x 10-3 ppm~i was several times higher than those proposed by the federal agency. Moreover, the California analysis used only the rat data, not the monkey data, for the interspecies extrapola- tion to human risk. California has been criticized for its interpretation of the new biologic data in its risk assessment. Of particular note is the belief by several CIIT -
FORMALDEHYDE SCIENCE 209 scientists that California misused their data in the risk models. For ex- ample, critics have argued that the California model has a strong conserva- tive bias and overpredicts the human risk from formaldehyde. This bias results from such factors as the model's overprediction of DPX levels ob- served ire rats exposed to low formaldehyde concentrations, the interspecies extrapolation procedures used in the model, and the failure to adequately account for formaldehyde's effects on cell replication in target tissues (Starr, 1991~. Furthermore, the critics chided the California model for disregard- ing the monkey DPX data. While EPA's risk assessment was parallel with the models developed by Starr and others at CIIT, California chose to incor- porate different assumptions and modeling techniques. The California Air Resources Board has declared HCHO a toxic air contaminant. The risk assessment has gone to the state's Office of Adminis- trative Law for review. What regulatory action, if any, the state will ulti- mately pursue remains to be seen. THE DEBATE From almost the moment the first pharmacokinetic data were released, they generated controversy in the scientific and regulatory arenas. The key issue was how the results of the studies were to be interpreted for use in quantitative risk assessment. Sides were quickly chosen over whether the data should be used in these analyses to evaluate the risk that formaldehyde poses for humans. Critics of the early pharmacokinetic data were quick to take issue with many of the methodological and design components of the research. Oppo- nents challenged the relevance of the short-term studies for the risk esti- mates associated with the chronic exposure experienced by humans. They were also concerned not only about the measurement of the DNA-protein cross-links but also about the relationship of the cross-links to carcinogenesis. The critics argued that this relationship had not been defined for formalde- hyde, nor, indeed, for other chemical carcinogens. Despite the further studies that have been released furnishing additional pharmacokinetic and mechanistic data for the scientific debate, critics re- main. Opponents2 contend that while the science of pharmacokinetics is a reasonable and promising field in principle, the data on formaldehyde are not yet of sufficient quality to permit the incorporation of a pharmacoki- netic model into formal risk analyses. They argue that while the scientific methodologies are improving, the DNA adduct and other measurements are not yet strong enough to be included in human risk estimates. Also, given the uncertainties about the data that still remain, the numbers obtained in new analyses using these data may not be any closer to the "truth" than the previous work. Critics contend that if we base risk estimates on the DPX · ..
210 SUSAN W. PUTNAM AND JOHN D. GRAHAM data, and they turn out to be incorrect, we might be seriously underestimat- ing human risk and endangering the health of the population. Champions of using a pharmacokinetic model and the new data, on the other hand, perceive the additional information as bringing us closer to the truth. They argue that the DPX data significantly improve our under- standing of what is happening to the tissues in the body. Because of the naturally occurring defense mechanisms, such as mucociliary systems, there is little covalent binding at the lower exposures. However, as the concen- trations increase, the defense mechanisms saturate and the amount of form- aldehyde reaching the DNA increases disproportionately. These data yield a more accurate measure of dose, and in turn provide for a more accurate risk assessment. Furthermore, the glutathione and isotope research im- proved many of the earlier methodological issues and unknowns with the data, the monkey data support the effects seen in the rat and have the potential to serve as a more appropriate species surrogate for humans; and, finally, chronic DPX studies are currently being conducted with results soon to be released. Proponents of the cell proliferation data similarly profess the impor- tance of the data. They argue that carcinogenesis is a multistage process. Sustained cell proliferation, which fixes unstable adducts into mutations, is necessary to get cancerous growths. Just as the DPX data support the view of a nonlinear relationship between administered and delivered dose for formaldehyde, the cell proliferation data posit a nonlinearity in the dose response curve. Proponents argue that it is important to be as realistic as possible and incorporate this nonlinearity into the risk assessment process, not just rely on the customary defaults of the current risk assessment mod- els-such as the no-threshold, low-dose linear response curve assumed with mutagenesis. While the cell proliferation data are currently difficult to incorporate into the risk assessment process, the champions of the data assert that it is important to develop new models and paradigms that would facilitate their use. In the regulatory arena, risk assessors may be reluctant to use the new data in their risk analyses. Regulatory agencies have traditionally taken a conservative stance, choosing to err on the side of being overly protective of human health. They may be hesitant to use the new data, knowing that the resultant risk estimates will be smaller than if the data are not incorporated. This lowering of the risk estimate is particularly pertinent to the low-dose exposure levels typically faced by chronic human exposure and traditionally the focus of regulatory policy. Furthermore, risk assessors receive little guidance on what data to use in their analyses. Legislative mandates gov- erning the agencies do not provide many criteria as to when to depart from the traditional models and data sets and incorporate new information into the risk assessment estimates (Rosenthal et al., 19921.
FORMALDEHYDE SCIENCE OBSERVATIONS 211 The formaldehyde case offers a number of interesting observations on the use of new science in the regulatory arena. The first of these observa- tions illustrates an apparent mismatch between the generation of new sci- ence and regulatory decisions. As the HCHO case demonstrates, regulatory decisions may be only weakly influenced by new data. The CPSC acted promptly on the initial CIIT bioassay data, but it took a federal court deci- sion for OSHA to finally use the bioassay data in its risk assessment and rulemaking process. Although it is now 14 years since the publication of CIIT's bioassay results, the EPA still lacks a settled regulatory policy on HCHO. None of the three agencies has yet officially incorporated the new findings on HCHO into their risk estimates or regulatory decisions. The second point of interest centers on the observation that much of the debate about the new information concerns not the validity of the new sci- ence, but how to interpret it for risk assessment and regulatory use. For example, some scientists interpret the lower rate of DNA-protein cross-link formation and larger target areas seen in the monkey DPX data (as opposed to the rat) as predicting a lower cancer risk for primates, and, ultimately, a lower one for humans than was estimated using the rat data. Other scien- tists, however, have interpreted these same data with their larger target area as supporting the observation of upper airway and lung cancer in people seen in the Blair study, in contrast to the nasal cancer seen in the rodent bioassays. This interpretation of the monkey data would lead to a higher cancer risk for humans than that estimated from the rat data. This matter is further complicated by the problem that several pieces of the new data from different scientific disciplines appear to be in opposi- tion to each other. A prevalent interpretation and use of the DPX data lowers the risk estimates for human cancer; a prevalent interpretation of the epidemiologic data raises the risk estimates for human cancer. In essence, use of different pieces of the new information results in contrasting risk estimates. Those on one side of the scientific-and regulatory-debate can cite some of the new scientific information in defense of their views, while those on the other side cite other pieces of the new data. Unfortunately, the debate over the interpretation and regulatory use of the new science on formaldehyde is also tainted by how the numbers play out in the risk assessment process. It can be argued that because the incor- poration of a pharmacokinetic model and the DPX data may lower the risk estimates, red flags were immediately raised in several camps. Those charged with protecting human health in the regulatory arena worried about the new data because they may lower the estimated risk of exposure to formalde- hyde. While supporting the use of new science in principle, their job ulti- mately compels them to oppose lowering the risk estimates if there remains .
212 SUSAN W. PUTNAM AND JOHN D. GRAHAM any uncertainty in the interpretation on which they can base their argument. Similarly, industry representatives welcomed the new data as supporting their position that exposure to formaldehyde at current levels is "safe," perhaps even overly protective. Had the numbers played out in the opposite direction, it is possible that the players would have taken different sides in the regulatory debate. Although the resulting risk estimates should not ultimately affect scien- tists' perception of the data, it is difficult to imagine that scientists can always divorce themselves from the results of the analyses. In a perfect world, scientists are solely interested in the generation and interpretation of "good" science, regardless of other issues such as their sponsorship or how the numbers play out which may be the case with many scientists today. Unfortunately, the possibility exists that, while scientists may not be inti mately involved in the politics of the regulatory arena, they may have an inherent level of conservatism that dictates their perspective on new data. It is unfair to speculate on potential differences in the scientific debate were the risk assessment numbers to be more conservative using the new data, but it is impossible not to wonder about it. A third observation involves the source of the new data. Much of the DPX and cell proliferation data on formaldehyde emanated from the Chemi- cal Industry Institute of Toxicology, a research institution supported largely by industry dollars. Not only did the data come from a single institute, but the funds to develop them flowed from industries with a distinct interest in lowering the risk estimates for the chemical. On the one hand, one could argue for more studies from diverse institutions to either refute or corrobo- rate the CIIT data for a more balanced generation of the science. But, on the other hand, the size and cost of these types of studies make such a wish somewhat unrealistic. Because industry has such a large stake in formalde- hyde, companies are willing to support major research operations on the chemical. It is unlikely that government agencies would have either the resources or the personnel necessary to undertake such projects. CIIT scientists are also ardent advocates of using the new science in risk assessment. In many of the papers describing their studies, CIIT re- searchers fervently champion the incorporation of the study results into the risk assessment and regulatory process. They are proud of the work that they have done on HCHO and would like it to be used to further the regula- tory process as well as the scientific environment. Several of the scientists have developed risk assessments themselves using the latest available data, and CIIT scientists frequently attend public hearings (such as SAB meet- ings) on the data. A fourth key observation is the level of "politics" involved in incorpo- rating new science into the regulatory arena. This issue is well illustrated by the SAB panel review of the 1991 EPA risk assessment on HCHO. - -
FORMALDEHYDE SCIENCE 213 Initially, in its preliminary review of the document, the panel supported the agency's use of the DPX data. The SAD, however, received much criticism from the UAW and others who opposed the DPX data and who had an important stake in not lowering the human risk estimate. Several additional members were subsequently added to the SAB panel, and, at the new review of the. rick n~.cm~nt the croon was snlit on whether to incorporate the DPX data. Also, not only is the clout of powerful interested players of note here, but why the UAW would get so involved with an agency that has only secondary jurisdiction in areas concerning occupational risk. One might speculate that if the debate over the new data were to disappear and the EPA were to use them, then maybe OSHA would similarly feel compelled to incorporate them in its next risk assessment as well. If this were the case, perhaps it would be in the best interest of the labor unions to protect their flanks and to discourage the use of the new data at a variety of points in the regulatory process, not just at OSHA. CONCLUSION The path that new scientific information travels from the laboratory to the regulatory arena is neither straightforward nor predictable. The time- table involved is often a long one. Numerous factors, such as scientific uncertainty, technical feasibility, and economic and political viability, con- found and entangle the incorporation of new science into the decision-mak- ing process. Furthermore, there are a myriad of players who interact with and complicate the process. In the simplest or "naive" depiction of this process, new science gets incorporated into agency risk assessments of the relevant risk. These risk assessments then become the basis for regulatory decisions (see Figure 4~. In an ideal world, it would be nice if the regulatory process were as simple and efficient. Regulatory reality, however, tells a different story. There are numerous perturbations that can disrupt and convolute this simple model. The first complication to the model arises from the fact that science can bypass the risk assessment process and directly influence regulatory deci- sions (Figure 5~. The scientific data, such as the HCHO epidemiologic data, may be difficult to quantify and incorporate into a formal risk assess- ment. While risk assessors may disagree over the quality of the science and its acceptability for their analyses, the mere existence of the new science may perturb risk managers to worry about the "truth" of the issue at hand. Moreover, the development of new science may serve to slow down the regulatory decision-making process, while risk assessors and risk managers debate its regulatory appropriateness. The regulatory model can be further complicated by the role that addi ~.
214 SUSAN W. PUTNAM AND JOHN D. GRAHAM Science Risk Assessment FIGURE 4 Naive model of regulatory process. Regulatory Decisions tional groups play in the process. One key player to interact in the model is the courts (Figure 6~. The courts can interact with the regulatory decision- making process by influencing either the risk assessment stage or regulatory decisions. Many regulatory actions-or inactions-frequently land in the courts, where judges ultimately decide their legality and appropriateness. The formaldehyde case demonstrates the power of the courts on several fronts. For example, the federal district court forced future regulatory ac- tion by ordering OSHA to consider formally initiating a new rulemaking process for formaldehyde, in response to a suit brought against the agency by several groups seeking an emergency temporary standard for the chemi- cal. On the other hand, the court of appeals overturned CPSC's 1982 ban on urea-formaldehyde foam insulation, ruling the CPSC had failed to sup- port its action with substantial evidence. As illustrated in the HCHO case, the courts can be either a compelling or an inhibiting force in the regulatory process. Another key influence in the process is the prevailing presidential phi- losophy. This philosophy can impact the process again at either the risk assessment or decision-making stage (Figure 7~. It can also impact the courts through presidential appointment of judges. The influence of presi- dential philosophy is clearly depicted in the formaldehyde case through the "regulatory reduction" ideology of the Reagan administration. At both OSHA and EPA, agency leaders following this ideology were slow to act on HCHO in the early 1980s. It took powerful influences the courts in the case of OSHA, agency turmoil resulting in a new administrator in the case of EPA- to spur regulatory action at the agencies. The Republican antiregulatory L, Regulatory I Decisions Science I ~ Risk Assessment . ~ 1~ ~ FIGURE 5 Model of regulatory process: Complication #1.
FORMALDEHYDE SCIENCE Science ~ Fit sk Assessment ~ Hi_ ~ Courts Regulatory Decisions r FIGURE 6 Model of regulatory process: Complication #2. 215 philosophy further permeated the process, however, by requiring that the Office of Management and Budget (and later the Council on Competitive- ness) approve potential agency regulations. A fourth complication to the naive regulatory model is presented by the powerful influence of various interest groups. These groups, such as indus- try associations or public interest advocates, can impact the model at nu Presiclential Philosophy Science T Risk Assessment 1 ' '' Courts l Regulatory Decisions L FIGURE 7 Model of regulatory process: Complication #3. at,
216 Presidential Philosophy Science · Assessment rail :- -_ cOUdS = Interested Groups FIGURE 8 Model of regulatory process: Complication #4. SUSAN W. PUTNAM AND JOHN D. GRAHAM Regulatory Decisions _ . merous points (Figure 8~. The formaldehyde case illustrates the influence of these groups at many junctures. CIIT, an industry-sponsored research institute, generated both the initial animal bioassay and subsequent DPX data on formaldehyde. Furthermore, CIIT scientists not only adamantly advocated the incorporation of these data into agency risk assessments but also performed and published their own risk assessments using the data. Other groups, such as the United Auto Workers and several environmental groups, took OSHA to court over its initial inaction on HCHO. These groups, as well as the Formaldehyde Institute, again sued the agency over its 1987 regulatory decision (the 1.0 ppm standard) on the chemical. At EPA, both industry and labor groups lobbied the agency on the relevant data to incorpo- rate into the risk assessment process and gave testimony to the agency's Sci- ence Advisory Board review of formaldehyde. At every step in the formalde- hyde story, powerful interest groups have exerted their influence.
FORMALDEHYDE SCIENCE 217 Perhaps the ultimate issue illustrated by these models and their increas- ing complications is, Who decides when new science should be incorpo- rated into the regulatory process? Who converts science on the "frontier" into "settled" science, acceptable for risk assessment and regulatory deci- sions? Is it the agencies the risk assessors, the risk managers? Is it an external panel of scientists who reach some consensus on the data's accept- ability? Should Congress legislate what scientific methodologies or data sets must be used? Should it be left up to the courts to decide? There will always be debate over what science is acceptable or what methodologies are valid. Furthermore, as the new OSHA standards for formaldehyde illus- trate, regulatory actions do not come cheaply; there are significant eco- nomic as well as health issues involved in using or not using new scientific information. The issue ultimately becomes twofold. First, who is qualified to make these judgments? Second, who will be held accountable for the consequences of these judgments? NOTES 1. When cells are exposed to formaldehyde, there may occur a cross-linking between DNA and proteins in the cells. These cross-links can then be counted as a measure of the HCHO dose that has actually reached the cells the delivered dose (as opposed to the administered dose). 2. The views expressed here on the most recent formaldehyde biologic data by both oppo- nents and proponents of the data were gathered from informal personal communications with a number of scientists knowledgeable on the issue, many of whom wished their comments to remain anonymous. This includes comments on the new data from some of the panel of scientists originally interviewed for their opinion on the initial pharmacokinetic data (Casanova- Schmitz et al., 1984) in Hawkins and Graham (1988). REFERENCES Albert, R. E., A. R. Sellakmur, S. Laskin, M. Kuschner, N. Nelson, and C. A. Snyder. 1982. Gaseous formaldehyde and hydrogen chloride induction of nasal cancer in the rat. Jour- nal of the National Cancer Institute 68:597-603. 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. Journal of the National Cancer Institute 76:1071-1084. Blair, A., P. A. Stewart, R. N. Hoover, J. F. Fraumeni, Jr., J. Walrath, M. O'Berg, and W. Gaffey. 1987. Cancers of the nasopharynx and oropharynx and formaldehyde exposure (letter to the editor). Journal of the National Cancer Institute 78:191. Blair, A., P. A. Stewart, and R. N. Hoover. 1990. Mortality from lung cancer among workers employed in formaldehyde industries. American Journal of Industrial Medicine 17:683- 699. California Air Resources Board, Office of Environmental Health Hazard Assessment. 1992. Cancer Risk Assessment for Airborne Formaldehyde. Proposed Identification of Formal- dehyde as a Toxic Air Contaminant: Part B. Health Assessment. Sacramento, Calif.:California Air Resources Board. act
218 SUSAN W. PUTNAM AND JOHN D. GRAHAM Casanova-Schmitz, M., T. B. Starr, and H. d'A. 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 [l4C]- and [3H] formaldehyde. Toxicology and Applied Pharmacology 76:26~4. Casanova, M., T. B. Starr, and H. d'A. Heck. 1986. Comments on the final report of the panel reviewing the CIIT pharmacokinetic data on formaldehyde. Research Triangle Park, N.C. Casanova, M., and H. d'A. Heck. 1987. Further studies of the metabolic incorporation and covalent binding of inhaled [3H]- and [14C] formaldehyde in Fischer-344 rats: Effects of glutathione depletion. Toxicology and Applied Pharmacology 89: 105-121. Casanova, M., D. F. Deyo, and H. d'A. Heck. 1989. Covalent binding of inhaled formalde- hyde to DNA in the nasal mucosa of Fischer 344 rats: Analysis of formaldehyde and DNA by high-performance liquid chromatography and provisional pharmacokinetic inter- pretation. Fundamental and Applied Toxicology 12:297-417. Casanova, M., K. T. Morgan, W. H. Steinhagen, J. I. Everitt, J. A. Popp, and H. d'A. Heck. 1991. Covalent binding of inhaled formaldehyde to DNA in the respiratory tract of Rhesus monkeys: Pharmacokinetics, rat-to-monkey interspecies scaling, and extrapola- tion to man. Fundamental and Applied Toxicology 17:409-428. Cascieri, T. C., and J. J. Clary. 1992. Formaldehyde-oral toxicity assessment. Comments in Toxicology 4:295-304. Cohn, M. S., F. J. DiCarlo, A. Turturro, and A. G. Ulsamer. 1985. Letter to the editor. Toxicology and Applied Pharmacology 77:363-364. Collins, J. J., J. C. Caporossi, and H. M. D. Utidjian. 1987. Letter to the editor. Journal of the National Cancer Institute 78: 192-193. Collins, J. J., J. C. Caporossi, H. M. D. Utidjian. 1988. Formaldehyde exposure and nasopha- ryngeal cancer: Re-examination of the National Cancer Institute Study and an update of one plant (letter to the editor). Journal of the National Cancer Institute 80:376-377. Consensus Workshop on Formaldehyde. 1984. Report on the Consensus Conference on Formaldehyde. Environmental Health Perspectives 58:323-381. Consumer Product Safety Commission. 1982. Ban of urea-formaldehyde foam insulation. Federal Register 47(April 2):14366-14419. DiCarlo, F. undated. Memorandum on the Expert Review of CIIT Pharmacokinetic Data on Formaldehyde. Washington, D.C.: U.S. Environmental Protection Agency. Environmental Health Committee. 1985. Review Report of the Draft Document, Preliminary Assessment of Health Risks to Garment Workers and Certain Home Residents from Ex- posures to Formaldehyde. Washington, D.C.: U.S. EPA Science Advisory Board. Federal Panel on Formaldehyde. 1982. Report of the Federal Panel on Formaldehyde. Envi- ronmental Health Perspectives 43: 139-168. Gibson, J. E., ed. 1983. Formaldehyde Toxicity. Washington, D.C.:Hemisphere Publishing Corp. Goldmacher, V. S., and W. G. Thilly. 1983. Formaldehyde is mutagenic for cultured human cells. Mutation Research 116:417-422. Graham, J. D., L. C. Green, and M. J. Roberts. 1988. In Search of Safety. Cambridge, Mass.: Harvard University Press. Greenwood, M. A., Office of Prevention, Pesticides, and Toxic Substances, U.S. EPA. 1992. Letter to William McCredie at the National Particleboard Association, July. Gulf South Insulation v. Consumer Product Safety Commission. 701 Fed.2d. 5th Circuit. 1137 (1983). Hawkins, N. C., and J. D. Graham. 1988. Expert scientific judgment and cancer risk assess- ment: A pilot study of pharmacokinetic data. Risk Analysis 4:615-625. Heck, H. d'A., and M. Casanova. 1987. Isotope effects and their implications for the covalent binding of inhaled [3H]- and [14C] formaldehyde in the rat nasal mucosa. Toxicology and Applied Pharmacology 89:122-134. ~,
FORMALDEHYDE SCIENCE 219 Jasanoff, S. 1990. The Fifth Branch. Cambridge, Mass.: Harvard University Press. Kerns, W. D., K. L. Pavkov, D. J. Donofrio, E. J. Gralla, and J. A. Swenberg. 1983. Carcino- genicity of formaldehyde in rats and mice after long-term inhalation exposure. Cancer Research 43 :4382-4392. Life Systems, Inc. January 1986. Expert Review of Pharmacokinetic Data: Formaldehyde. Washington, D.C. Monticello, T. N. 1990. Formaldehyde-induced pathology and cell proliferation. Ph.D. Dissertation submitted to the Department of Pathology, Duke University, North Carolina. Monticello, T. N., and K. T. Morgan. 1990. Correlation of cell proliferation and inflammation with nasal tumors in F-344 rats following chronic formaldehyde exposure. American Association of Cancer Research 31:139. Monticello, T. M., K. T. Morgan, J. I. Everitt, and J. A. Popp. 1989. Effects of formaldehyde gas on the respiratory tract of rhesus monkeys. American Journal of Pathology 134:515- 527. Monticello, T. N., F. J. Miller, and K. T. Morgan. 1991. Regional increases in rat nasal epithelial cell proliferation following acute and subchronic inhalation of formaldehyde. Toxicology and Applied Pharmacology 111 :409-421. National Research Council. 1980. Formaldehyde-An Assessment of Its Health Effects. Washington, D.C.: National Academy Press. Occupational Safety and Health Administration. 1985. Occupational exposure to formalde- hyde. Federal Register, vol. 50, no. 74, pp. 15179-15184, April 17, 1985. (To be codified in 29 CFR Part 1910.) Occupational Safety and Health Administration. 1987a. Conclusions regarding the pharmaco- kinetic model. Federal Register, vol. 52, pp. 46226-46227. Occupational Safety and Health Administration. 1987b. Occupational exposure to formalde- hyde. Federal Register, vol. 52, no. 233, pp. 46168-46171, December 4, 1987. (To be codified in 29 CFR Parts 1910 and 1926.) Occupational Safety and Health Administration. 1987c. Summary of the regulatory impact and regulatory flexibility analysis. Federal Register, vol. 52, pp. 46237~6242. Occupational Safety and Health Administration. 1992. Summary and explanation of the final amendments. Federal Register, vol 57, no. 102, pp. 22292-22328, May 27, 1992. (To be codified in 29 CFR Part 1910.) Putnam, S. W. 1991. Formaldehyde. Pp. 127-158 in Harnessing Science for Environmental Regulation, J. D. Graham, ed. New York: Praeger Publishers. Rosenthal, A., G. M. Gray, and J. D. Graham. 1992. Legislating acceptable cancer risk from exposure to toxic chemicals. Ecology Law Quarterly 19:269-362. Science Advisory Board, U.S. Environmental Protection Agency. 1992. Review of the Office of Toxic Substances' Draft Formaldehyde Risk Assessment Update by the Environmental Health Committee. September. Washington, D.C.: U.S. Environmental Protection Agency. Starr, T. B. 1990. Quantitative cancer risk estimation for formaldehyde. Risk Analysis 10:85-91. Starr, T. B. November 1991. Comments on the California Air Resources Board proposed identification of formaldehyde as a toxic air contaminant, September 1991, Draft SRP Version. Starr, T. B., and R. D. Buck. 1984. The importance of delivered dose in estimating low-dose cancer risk from inhalation exposure to formaldehyde. Fundamental and Applied Toxi- cology 4:740-753. Stayner, L. T., L. Elliott, L. Blade, R. Keenlyside, and W. Halperin. 1988. A retrospective cohort mortality study of workers exposed to formaldehyde in the garment industry. American Journal of Industrial Medicine 13:667-681. Swenberg, J. A., E. A. Gross, and H. W. Randall. 1986. Localization and quantitation of cell
220 SUSAN W. PUTNAM AND JOHN D. GRAHAM proliferation following exposure to nasal irritants. Pp. 291-298 in Toxicology of the Nasal Passages, C. S. Barrow, ed. New York: Hemisphere Publishing Corporation. Tobe, M., T. Kaneko, Y. Uchida, E. Kamata, Y. Ogawa, Y. Ikeda, and M. Saito. 1985. Studies of the Inhalation Toxicity of Formaldehyde. Report No. TR-85-0236. Tokyo: National Sanitary and Medical Laboratory Service. Transcript of the Environmental Health Committee Review of the Draft Risk Assessment Document on Formaldehyde. June 26, 1985. U.S. EPA Science Advisory Board. U.S. Environmental Protection Agency. 1985. Preliminary Assessment of Health Risks to Garment Workers and Certain Home Residents from Exposure to Formaldehyde. Wash- ington, D.C.: U.S. Environmental Protection Agency. Draft. U.S. Environmental Protection Agency. 1987. Assessment of Health Risks to Garment Work- ers and Certain Home Residents from Exposure to Formaldehyde. Washington, D.C.: U.S. Environmental Protection Agency. U.S. Environmental Protection Agency. 1990. Formaldehyde Risk Assessment Update. Washington, D.C.: U.S. Environmental Protection Agency. U.S. Environmental Protection Agency. 1991. Formaldehyde Risk Assessment Update. Washington, D.C.: U.S. Environmental Protection Agency. U.S. Regulatory Council. 1979. Statement on regulation of chemical carcinogens. Federal Register, vol. 44, p. 60038. Vaughan, T. L., C. Strader, S. Davis, and J. R. Dating. 1986. Formaldehyde and cancers of the pharynx, sinus, and nasal cavity: I. Occupational exposures and II. Residential expo- sures. International Journal of Cancer 38:677-688. _ -