Sulfur mustard as a carcinogen: Application of relative potency analysis to the chemical warfare agents H, HD, and HT

REGULATORY

TOXICOLOGY

ANDPHARMACOLOGY

l&1-25

(1989)

Sulfur Mustard as a Carcinogen: Application of Relative Potency Analysis to the Chemical Warfare Agents H, HD, and HT

ANNETTA

P. WATSON,

TROYCE

D. JONES, AND GUY D. GRIFFIN

Health and Safety Research Division, Oak Ridge National Laboratory, P.O. Box 2008, Oak Ridge, Tennessee 378314101

Received July 28, 1988

A relative potency method for assessing potential human health effects from exposures to relatively untested chemicals is presented and documented. The need for such a method in evaluating the carcinogenic potential of the chemical warfare agent sulfur mustard (agent HD) from a limited data base is specifically addressed. The best-estimate potency factor for sulfur mustard relative to benzo[a]pyrene is I .3, with an interquattile range of0.6 to 2.9. The method is applied to (1) the estimated fence-boundary air concentrations of mustard during operation of a proposed agent incinerator at Aberdeen Proving Ground (APG), Maryland, and (2) the current approved general population exposure level of 1 X 1Om4mg HD/m3 and the occupational exposure level of 3 X 10m3mg HD/m3. Maximum estimates of excess lifetime cancer risk for individuals at sites along the APG boundary range between 3 X 10e8 and 1 X lo-‘. Lifetime cancer risk estimates < 10e6 are not now regulated by the U.S. Environmental Protection Agency or the Food and Drug Administration. Maximum estimates of excesslifetime cancer risk assuming daily exposure to the approved standards during the proposed 5 years of incinerator operation are on the order of 10m5for the general public and 10m4for the worker population. These values are considered upper limit estimates. 0 1989 Academic PK~S$ 1~.

INTRODUCTION The Department of Defense Authorization Act of 1986 (PL 99- 145) directed and authorized the Secretary of Defense to destroy the United States stockpile of lethal unitary (active agent packaged in the munition; as opposed to binary weapon design) chemical munitions and agents by September 30,1994; the Act was amended in 1988 to permit operations testing of commercial-scale incinerator design and to allow for unitary munition disposal completion by September 30, 1997. The inventory of material in this category includes the organophosphate agents GA, GB, and VX as well as the vesicant (blister) agents H, HD, HT (various formulations of sulfur mustard) and Lewisite (an organic arsenical). These agents are presently stored at eight separate locations in the continental United States as bombs, cartridges, mines, projectiles, 0273-2300189 $3.00 Copyright 0 1989 by Academic Press, Inc. All rights ofreproduction in any form reserved.

2

WATSON,

JONES, AND GRIFFIN

FIG. 1. Unitary chemical weapon stockpile distribution throughout the continental United States.

rockets, spray tanks, and ton containers (see Fig. 1). The current method of choice for agent destruction is high-temperature ( 1130- 1400°C) incineration. While each of these agents is acutely lethal at sufficiently high doses, much interest in evaluating potential delayed or latent effects of long-term, low-dose rates has been expressed by the public, the scientific community, and the Department of the Army. It is thought that such low-level doses may conceivably be received by members of the general public or occupational populations during normal operation of incineration disposal. At present, the consequences of such hypothetical exposures are unquantified. A good candidate for comparative analysis of latent effects is sulfur mustard (bis[2chloroethyl] sulfide, CAS No. 505-60-2, C4H&12S), a recognized carcinogen (IARC, 1975). Sulfur mustard is the principal vesicant component of agents H (Levinstein mustard); HD (distilled mustard), and HT (a plant-run mixture of about 60% HD, ~40% stabilizer “T” [bis(Zchloroethyl thioethyl)ether (CsH&120S2)] and a variety of sulfur impurities). Because of its status as a warfare agent, sulfur mustard has not been widely evaluated in standard laboratory biological test systems. Although limited in scope, sufficient data exist to make use of a recently developed analytical technique whereby the toxicity of a compound without epidemiologically derived risk estimates (an “interviewing compound”) can be compared with the toxicity of wellcharacterized compounds (“reference compounds”) to derive a “relative potency factor” (RPF) (Jones et al., 1985; Jones et al., 1988). The RPF can then be used in subsequent estimates of risk and assessment of exposure standards. An analysis of sulfur mustard by this technique follows. EXPOSURE

HISTORY

The record of human cancer induction after exposure to sulfur mustard (agents H or HD) is based on very few data sets describing the response of soldiers and weapons

CARCINOGENIC

POTENCY

OF

SULFUR

MUSTARD

3

plant workers exposed to toxic concentrations under wartime conditions. In some cases, victims were exposed to atmospheric concentrations of mustard (some estimates are as great as 50 to 70 mg/m3) sufficiently high to induce signs of acute toxic response shortly after chemical exposure (e.g., skin blistering, chemical burns of the eyes and mucous membranes, and respiratory distress). Retrospective studies of these victims are summarized below. In soldiers exposed to mustard in World War I, there is evidence of a positive association between mustard exposure and increased risk of developing lung cancer 10 to 40 years postexposure. However, the increased risk is not large (IARC, 1975; Case and Lea, 1955; Beebe, 1960; Norman, 1975). British pensioners exposed to a mustard agent were compared to ( 1) war pensioners who had chronic bronchitis, and (2) a group of amputees. Neither of the latter two groups had been exposed to mustard. Mortality due to lung cancer was elevated in both the mustard gas group and the chronic bronchitis group when compared to the amputee group. On the basis of this study, the increase in lung cancer could not be uniquely attributed to mustard agent exposures alone (Case and Lea, 1955). One interpretation of these data is that lung tissue damaged by either chronic bronchitis (or the etiologic agent that generated chronic bronchitis) or battlefield mustard agent exposures was more likely to become malignant than nondamaged tissue. Mortality from nonmalignant causes of death in both the group exposed to mustard and the group with bronchitis were much higher than that observed in either the general population or amputee group, particularly during the period before 50 years of age. This hints at the possibility that the study groups were different from the reference set in other ways. It is possible that the factors associated with nonmalignant causes of death may also have affected the observed mortality from lung cancer. American soldiers who experienced mustard agent exposures were also compared to two other groups: (1) soldiers who had pneumonia during the influenza outbreak of 19 18, but who were not exposed to mustard, and (2) wounded soldiers, who had neither pneumonia nor mustard exposures. The number of respiratory cancer cases in the mustard-exposed group was somewhat elevated, compared to the other two groups (ratio of observed to expected cases: mustard gas, 1.47; pneumonia, 0.81; wounded, 1.15 (Beebe, 1960)). A further study of this same group of American soldiers, involving an additional 10 years of follow-up (Norman, 1975) did not change the conclusions of the previous study. Norman (1975) found that the relative risk of death from lung cancer among those poisoned by mustard was 1.3 compared to controls, with 95% confidence limits of 0.9- 1.9. Norman (1975) concludes that this increased risk does not make a strong case for a carcinogenic effect of mustard, probably because an insufficient dose was received. The effect of tobacco use was examined among a limited subgroup of veterans for which smoking histories were available (Norman, 1975). The relative risk of lung cancer mortality among cigarette smokers who were exposed to mustard agents was approximately equal to that of the population of gassed veterans who stated they did not smoke (4.3 vs 4.4). Norman concludes that there was no evidence that mustard exposure and cigarette smoking “acting together produced either a smaller or larger relative risk of death from lung cancer than the sum of their separate effects.” Thus, for the limited data set available, there appears to be no substantiation for a synergistic effect between cigarette smoking and mustard exposure.

4

WATSON,

JONES,

AND

GRIFFIN

Indirect, but more definitive evidence of the carcinogenic action of mustard in humans comes from occupational exposures experienced in poison gas factories active prior to and during World War II. From 1929 through 1945, the Japanese Army operated a poison gas factory on Okuno-jima, an island in the Inland Sea and south of Hiroshima Prefecture (Wada et al., 1962a, 1962b; Nishimoto et al., 1983, 1986). This factory produced Lewisite, mustard gas, hydrocyanic acid, diphenylcyanarsine (sneezing gas), chloroacetophenone (tear gas), and phosgene for use in the invasion and occupation of mainland China. At peak capacity, this facility produced 450 tons/ month of mustard gas, 50 tons/month each of Lewisite, diphenyl cyanarsine, and hydrocyanic acid, and 25 tons/month of chloroacetophenone. Approximately 1000 individuals were employed in the manufacture of these warfare chemicals during the period of maximum production (1937-1942) (Wada et al., 1962a). Thus, Japanese workers experienced exposures to multiple poisonous agents other than mustard, although mustard was produced in much larger quantities than the other agents. The atmospheric concentrations of mustard in the mustard production areas of the factory (estimated to be 50-70 mg/m3) were great enough to produce most of the acute symptoms of mustard poisoning in workers. Even vegetation (grasses and trees) around the mustard factory died. In 1952, 5 years after the facility was dismantled and stocks destroyed, “the mustard-like odor could still be detected in some areas” (Wada et al., 1962a). During factory operation, some protective clothing was worn, but it was neither kept in good repair nor stocked in sufficient quantity. Other industrial hygiene measures were haphazard at best and many workers suffered severe exposures (Wada et al., 1962a; Inada et al., 1978). Retrospective studies ofthese workers have shown definite increases in respiratory cancer among workers who produced mustard agent when compared to office workers at the same factory (Wada et al., 1962b; Wada et al., 1968; Nishimoto et al., 1983; IARC, 1975; and Tokuoka et al., 1986). Some highlights from these studies include: (1) increased mortality due to respiratory cancer in mustard gas production workers (i.e., 33 deaths in a group of 495 workers compared to 0.9 deaths expected in a group of this size [Wada et al., 1962b]); (2) evidence of a dose-response relationship between mustard exposure and subsequent development of respiratory cancer (i.e., when poison gas production workers were compared to office staff or other factory workers not involved with poison gas production or shipping) (Nishimoto et al., 1983; Yamakido et al., 1985); (3) a lung cancer standardized mortality ratio (SMR) of approximately 3 for workers (all males) directly involved with poison gas production. A SMR ofthis magnitude indicates that the number of deaths in this occupational population was approximately three times that of unexposed males living in the surrounding area (Nishimoto et al., 1983); and (4) histological evidence that most of the respiratory cancers arose in the proximal part of the respiratory tract (e.g., pharynx, larynx, trachea), as would be expected from inhalation of a carcinogenic gas (Tokuoka et al., 1986; Nishimoto et al., 1983). In groups of mustard-exposed Japanese war factory workers who developed lung cancer, the average length of employment was 5.7 years (Tokuoka et al., 1986). Besides the respiratory tract cancers observed, a recent report indicates an increase in digestive tract neoplasms in these Japanese workers, although no further details

CARCINOGENIC

POTENCY

OF

SULFUR

MUSTARD

5

are indicated (Yamakido et al., 1985). The mustard production workers also experienced skin damage (pigmented and depigmented spots) and skin cancers (basal cell carcinoma and Bowen’s disease (Inada et al., 1978)). These lesions were not observed among employees in the same factory who were not engaged in mustard production. In addition, a correlation between length of employment in mustard gas production and severity of skin lesions was noted. These results provide strong evidence for an association between mustard agent exposure and certain skin cancers; again, it should be noted that doses were large, and the duration of exposure was long term (mean of 9 years employment for those with skin lesions studied by Inada et al. (1978)), and workers were exposed to other intensely irritating warfare agents as well as toxic organic chemicals. Epidemiological studies of German poison gas factory workers have also indicated increased incidence of malignant respiratory cancers (I-ohs, 1975; IARC, 1975). Klehr (1984) reported multiple skin tumors (i.e., basal cell carcinoma, Bowen’s disease) in German workers who were exposed to sulfur mustard and nitrogen mustard without adequate protection during the dismantling of a poison gas factory. Individuals also experienced necrotic and gangrenous skin ulcerations, which had a tendency to spread and be resistive to therapy. A study of British workers employed in a mustard gas factory during World War II revealed a significant increase in diagnosed cases of fatal and nonfatal laryngeal carcinoma among factory workers when compared to a control population (2 deaths attributable to cancer of the larynx as compared to 0.4 expected out of a worker population of 502) (Manning et al., 1981). Individuals with laryngeal cancer had worked in the mustard gas manufacturing process for 4 to 5 years. Although lung cancer deaths were also elevated in the occupational group when compared to the appropriate control, the increase was not statistically significant. The authors suggest that differences in the cancer risk of the British and Japanese groups could have been due to better industrial hygiene practices in wartime England. Based upon available but limited human data, IARC concludes that: “There is evidence for an increased incidence of cancers of the respiratory tract in men exposed to mustard gas” (IARC, 1975). Thus, mustard is included in a group of chemicals which are carcinogenic for humans, since data are sufficient to support a causal relationship between exposure to the chemical and subsequent cancer induction (Saracci, 198 1; Nelson, 198 1). The foregoing studies, although significant and important, do not provide useful dose-response information for estimating the carcinogenic activity of mustard agents. This is particularly true for the inhalation studies. In fact, in the case of the human epidemiological data, no well-documented estimates of dose can be made. Given this predicament, investigators who wish to quantify cancer risk from mustard exposure are forced to rely on laboratory animal data (also limited) and extrapolate to an estimated human response in some systematic manner. An approach that has proved useful in evaluating other poorly characterized compounds is summarized below. Example calculations are provided. DEVELOPMENT

OF RELATIVE

POTENCY

CONCEPT

Many chemicals are of concern to human health, but only a few possess epidemiologically derived risk estimates. Because it is necessary for regulators to estimate per-

6

WATSON,

JONES, AND GRIFFIN

missible concentrations of pollutants in water and air without the benefit of complete data bases, numerous decision protocols have been developed to obtain an approximate idea of the relative hazard of an untested chemical or a well-studied chemical under untested conditions. To give an idea of the gap between what is characterized and what is not, it is important to note that permissible concentrations for only approximately 65 priority pollutants have been estimated in U.S. Environmental Protection Agency (U.S. EPA) Water Quality Criteria Documents (Sittig, 1985). The Carcinogen Assessment Group (CAG) of US. EPA has estimated relative carcinogenic potencies for approximately 59 chemicals in the health assessment document series (e.g., see U.S. EPA, 1987a). Both CAG and the water criteria work are ongoing activities, and additional chemicals are expected to be analyzed at a rate of a few per year. In contrast, common industrial effluents contain thousands of compounds, only a fraction of which have been sufficiently evaluated to support assignment of a permissible concentration estimate. There are many different and useful systems available to analyze the hazards of unknown materials or test conditions; each method has inherent strengths and weaknesses. Most are designed for specific purposes and are unsuitable for general application across broad classes of chemicals. We have reviewed major features of approximately 40 of these systems (Jones et al., 1985) and find the following undesirable characteristics: (1) dependence on expert judgment for data selection and analysis, which leads to much controversy between experts when comparing different chemicals, different exposures, or establishing remedial priorities; (2) arbitrary safety factors that change from chemical to chemical; (3) very limited selection of test data-even limited selection of data from one dose-response experiment (i.e., not all data from a given study are used [Anderson and U.S. EPA-CAG, 19831); (4) subject to false-positive and false-negative conclusions (note that the probability of false conclusions are greatly reduced by increasing the amount of test data used in the analysis); (5) arbitrary combination of categorical numerical subscores; (6) inaccuracy when compared with the benchmarks (because of the extensive analytical effort of numerous experts in developing them, benchmark values are assumed to be reasonable predictors in this analysis); and (7) lack of a reasonable estimate of possible uncertainty between methodologies and in evaluations performed by a given method. Of the many decision methods available, most were devised for ranking, relative scoring, or prioritizing and are not compatible with quantitative risk assessment. Exceptions include methods based on CAG models, the use of an index that relates the potency of each carcinogen in rodents to human exposures (Ames et al., 1987), and relative potency-based methods (Jones et al., 1985; Jones et al., 1988). Formerly, U.S. EPA used the Hazard Ranking System (HRS) developed by the Mitre Corporation to rank chemical hazards (U.S. EPA, 1982a). For toxicological considerations, the HRS relied on the Sax Index of Toxicity (Sax, 1975, 1979), which was derived from categorical considerations of the intensity and duration of the toxic

CARCINOGENIC

POTENCY

OF SULFUR MUSTARD

7

response. The Sax Index does not consider the dose required to cause an effect. Thus, the HRS was not suited to quantitative risk assessments and has since been found to be deficient in other toxicological respects (ICF Inc., 1987). U.S. EPA is currently promulgating two basic approaches to the assessment of hazard. The first is the “reportable quantity” (RQ) concept (U.S. EPA, 1987b) devised to aid in the reporting and management of chemical spills. Decision criteria for RQs are based on six considerations: aquatic toxicity, ignitability, reactivity, acute toxicity, chronic toxicity, and cancer induction. The assignment of categories by this method is not compatible with quantitative risk analysis. The second basic approach promulgated by U.S. EPA is a semiquantitative one in that CAG models are used for individual carcinogens to derive “risk specific doses” (i.e., RSDs) (U.S. EPA, 1986a). For chemicals not currently known to be carcinogenic in and oftheir own properties, U.S. EPA postulates a no-effect threshold and assumes that small doses of these individual chemicals cause no harm. These threshold values are known as RlDs (i.e., reference doses). However, this regulatory approach does not take into account the known biological activity of promoters or potentiators of carcinogenesis. Humans are not usually exposed to individual compounds in isolation, but to complex mixtures that may very well include initiators (e.g., subclinical doses of carcinogens) and/or promoters, as well as carcinogens and cocarcinogens. In addition, it is clear that some materials currently listed as known or suspected human carcinogens (e.g., arsenic, benzene, and DDT) likely require the prior action of an initiator before the primary cancer lesion can develop. In our opinion, it is a matter for serious concern that U.S. EPA’s current position fails to properly consider human exposures to complex mixtures and the probable potentiation of cancerous lesions by generally toxic chemicals (which may act via either genetic or epigenetic mechanisms). The current regulatory approach may greatly underestimate potent human health effects resulting from exposures to such mixtures. Compensatory cell proliferation, generated by hyperplastic or toxic insults, is essential for complete initiation of carcinogenesis and to permanently fix any molecular lesions associated with the initiation. “Fix” is used in the classical biological sense, and is used to describe a permanent tissue change. Experiments have shown that promotion performed at any subsequent time will amplify fixed initiation events and result in neoplasm development (Jones et al., 1983, 1988). Some chemicals contribute only to the toxic response represented by compensatory cell proliferation; a complete carcinogen would demonstrate the capacity to both initiate and promote. Benzo[a]pyrene, DMNA, and bischloromethyl ether (BCME) are all well-known examples of complete carcinogens; although less fully characterized, mustard agent is suspected to be a complete carcinogen as well. The relative potency approach that we propose postulates that all chemicals inducing compensatory cell proliferation through hyperplastic or hypoplastic effects may act to enhance cancer rates as measured in a cohort subjected only to normal homeostatic mechanisms (Jones et al., 1983; Jones, 1984). We consider the relative potency method to be a direct index of a compound’s activity as a promoter. Our “RASH” (Rapid Screening of Hazard) approach converts all documented toxic chemical effects to some effective dose of a reference chemical (usually the polynuclear aromatic hydrocarbon benzo[a]pyrene (B[a]P), C20Hr2) (Jones et al., 1985; Jones et al.,

8

WATSON,

JONES, AND GRIFFIN

1988). Thus, the total biological insult of a complex mixture may be analyzed as an effective dose of one chemical-B[a]P. In our opinion, the biological response data characterizing B[a]P toxicity are of sufficient quality and number to provide a sound basis for dose-response estimation. Compounds for which comparable data are unavailable have not been, and should not be, used as reference compounds. Based on what is known at this time, this method is prudent and reasonable with respect to safety. The RASH calculations do not support the argument that a human population will host the calculated effect. The appropriate interpretation is that if the effect does occur, the RASH index is a realistic predictor. Because RASH incorporates the equivalent dose concept relative to the reference chemical and since the CAG carcinogenic potency factor p has been derived as “an estimated upper-95% confidence limit of the carcinogenic potency of [a] chemical” (U.S. EPA, 1986b), any quantitative estimate based on RASH would also be likely to reflect the upper 95% limit of the CAG maximum likelihood estimate. This approach is well supported by U.S. EPA guidelines for carcinogen toxicity assessment (U.S. EPA, 1986b), which recommend use of a Q* based on animal study data to determine an upper-bound estimate of human cancer risk. We consider the Q+ derived for B[a]P to be reasonably stable. The method we have derived makes use of the availability of single-source documents for extensive toxicity information. At present, the most useful document for this purpose is the Registry of Toxic Effects of Chemical Substances (RTECS), published annually by the U.S. Department of Health and Human Services (e.g., Tatken and Lewis, 1985). Other sources exist and are described in Jones et al. (1985). The toxicological potency of a chemical to be assayed (i.e., the interviewing chemical) is compared not only to the toxicity of the primary standard [B[a]P], but also to one or more of several secondary standards (benzene [C,Hh], cadmium [Cd], and N-nitroso dimethylamine [DMNA] [C2H6N20]). We have found this process to compare favorably with the results of laborious consideration by expert committees in deriving permissible exposures (Jones et al., 1988). This relative potency or hazard assessment approach was designed for ease of evaluation, and to be comparable within an approximate order of magnitude to the risk estimates developed for about 80% of the chemicals evaluated by different expert committees, including the EPA-CAG. The degree of comparability is more appropriately considered to be a measure of precision, because a goal of the RASH method was to simulate the decision-making process of expert committees; their standard but uncertain values are used by RASH for comparison. It should be noted that the expert value is usually an estimate based on the upper 95% confidence limit for a dose-response derived from an animal bioassay. The upper 95% confidence limit is commonly extended to derive estimates of potential human risk, however, we consider such treatment to be an inappropriate and erroneous extrapolation of available data. In most applications, the uncertainty in the extrapolation to man is dominated by interspecific differences and variations in exposure schedules or experimental design. Variance from the 95% upper confidence limit relative to a maximum likelihood estimate on an animal-based dose-response would likely be a small fraction of the total uncertainty, although it is widely cited as if it were the composite uncertainty (all components of uncertainty are generally not known). The RTECS data used to derive RASH estimates includes responses observed in both in viva and in vitro test systems as well as a wide variety of treatment

CARCINOGENIC

POTENCY

OF

SULFUR

MUSTARD

9

regimens. As a result, we consider the RASH method (which makes use of the entire array of available bioassay data) to encompass more of the explicit elements of experimental uncertainty than the selected few bioassays used by the various expert committees. The RASH range (interquartile range if the number of ratios is sufficiently large) of uncertainty associated with the best estimate of maximum plausible risk serves to indicate how the response of an untested heterogeneous population (i.e., man) may deviate from the central tendency of bioassay results. GUIDELINES

FOR RTECS

COMPARISONS

An example of a relative comparison follows: According to the RTECS listings, one may find that x (mg/kg) of a chemical has produced a particular effect such as LDL, (i.e., low acute mortality) in a particular species and y (mg/kg) of B[a]P or some other reference chemical tested by some other investigator was required to induce LD, in the same species. The potency of the first chemical relative to the standard or reference chemical would be y/x. Thus, if the reference chemical was considered by the National Institute for Occupational Safety and Health (NIOSH) Criteria Documents, CAG, or some other regulatory agency to be safe at a concentration in water of 1 mg/liter, then the unregulated chemical could be limited to (1 mg/liter)/( y/x). Likewise, if NIOSH or U.S. EPA considers that the reference chemical is safe at a concentration of 1 mg/m3 in air, then the unregulated chemical could be limited to ( 1 mg/m3M Y/d. This illustrates the first and most common way of computing relative potency values: l

comparing

different doses required to induce the same level of effect.

The other basic method is: l comparing different levels of effects resulting from equal doses, and/or different exposure times.

We have developed several general guidelines, which are summarized in Table 1. The RASH user of data from any large data base, including RTECS, should be aware that any limitations in the original study will be carried into the RASH estimate of potency. If sufficient time and resources are available, it is advisable to evaluate the original sources for data quality (confidence limits, sufficient numbers of test animals, appropriate exposure conditions, etc.). Even so, the user should understand that comparison across even closely related species (i.e., rats to mice) may not, in reality, be 1: 1. In actual fact, there is no clear guideline for interspecies tolerance across broad categories of test compounds. Comparison of very dissimilar mammalian species (i.e., rats to chimpanzees) should be avoided unless there is no other choice. We chose the median potency as characteristic of the interviewing chemical relative to the standard (i.e., B[a]P in this paper) because the nature of the underlying statistical distribution(s) of toxicological data is unknown. Selection of a mean value would assume normality of all interrelated sources of data in the data base used; it has been previously determined that not all toxicological data are normally distributed (Glass, 1987). Similarly, we chose to use the interquartile range describing dispersion about the median as a practical estimate of uncertainty. Other spans could be used, such as

10

WATSON,

JONES, AND GRIFFIN TABLE 1

SELECTEDPROCEDURESUSEDTOSTANDARDIZETHERASHMETHOD OFCOMPUTINGRELATIVEPOTENCYVALUES l

l

l

l

l

l

Generally attempt to match each test on the interviewing chemical with only one test on the standard chemical. If a test on the interviewing chemical does not match a test on the primary standard, then use a secondary standard and correct the estimate to the primary standard. If a listing in RTECS for the standard chemical seems inconsistent with other entries for the standard, select a replacement value from the secondary standards. If a listing for the interviewing chemical seems inconsistent, it may be rejected if appropriate. Because the median relative potency is used and the interquartile range is a measure of uncertainty, a small percentage of erroneous values should not significantly change either the rank of relative toxicities or the median potency of a chemical. Gross errors are possible when only a few test values are available and a large fraction are inaccurate or not representative of disease processes in man. When few tests are available for the interviewing chemical, it may be desirable to match with the primary standard and with the secondary standards in order to reduce the total variance deriving from the choice of a standard-generally, this step is unnecessary. With fewer than six matches, use the extreme spread instead of the interquartile range as a measure of uncertainty.

l

Use milligram per kilogram units. Potencies on a molar or parts per million basis have different scales,

l

Match tumors across species only as a last resort.

l

l

l

l

l

l

Lethality can generally be matched across all mammals unless novel pharmacological processes are involved. When many tests are available for the interviewing chemical and for the standard, it may be desirable (for some applications) to use equal numbers of matches from mutation tests, reproductive tests, tumor assays,toxicity comparisons, and irritation tests. This may be especially important if a chemical appears to be very potent in some categories and weak in others. More generally, each subset of relative potency values seems to have the same range as the set from which it is drawn. Do not compare mammals with species from other phylogenetic classes(such as amphibians, fish, birds, etc.). Similar routes of administration may be matched if other experimental conditions remain fairly constant, e.g., subcutaneous with intramuscular or stomach tube with oral. When such matches are attempted, it is necessary to match treatment schedules closely. When considering tumor studies, treatment schedules lasting 4 to 12 weeks can be considered equal if the tumor count is taken many months after treatment ceases. When treatment intervals are long (e.g., more than 26 weeks), the time treatment schedules should match as closely as possible. For example, one test spanning 80 weeks may be matched with a test spanning 65 weeks but should probably not be matched with a test spanning 120 weeks. Generally, when test intervals are long we have attempted matches only when 0.8 < T, /T2 < 1.2. Tumor studies are most difficult to compare. Even when test data are available, the test protocols may be too different to attempt to compute a relative potency. Or, test conditions may be so different that relative potency comparisons are of unknown validity. Fortunately (and generally), relative potency values obtained from comparing tumor studies closely resemble relative potency values from other toxicity endpoints. When evaluating effects of treatments over time, make equivalent dosage comparisons (e.g., 2 &liter for 2 hr is postulated to be approximately equivalent to 1 pg/liter for 4 hr). This linearization of time should not be applied to tumor studies.

CARCINOGENIC

POTENCY

OF

SULFUR

MUSTARD

11

90%. For many chemicals the effect would not be statistically different; but for some chemicals, the 90% span could be very large. In such cases, the spread could reflect inflated observational ranges due to random error combined with results obtained from extremely sensitive or extremely resistant test models. We are fully aware of the variable quality of data available for analysis. In practical terms then, with all other variables comparable,

where RPF is the relative potency factor, DBtalP is the dose of B[a]P producing a given endpoint in a given test system, and & is the dose of “test” or “interviewing” chemical that produces the same endpoint in the same system. If a good match of experimental conditions is not obtained with the primary standard data set, perhaps a better comparison can be made with the toxicity data available for one of the secondary standards (benzene, cadmium, or N-nitrosodimethylamine). If so, then

RPF=(F)(g), where D, is the dose of the secondary standard producing the desired endpoint given test system. The ratio

has been determined

in a

for each secondary standard. The values are as follows: *

= 0.005 benzene

2

=0.079

$f=

= 0.23.

DMNA

In order to compare the potency of our primary standard with other compounds, it is important to consider its U.S. EPA grouping relative to other carcinogens. Benzo[a]pyrene is classified as a B2 carcinogen by U.S. EPA weight-of-evidence criteria (U.S. EPA, 1986b). A B2 carcinogen is considered a “probable human carcinogen” based on “sufficient evidence of carcinogenicity in animals” but “inadequate evidence of carcinogenicity in humans” (U.S. EPA, 1986b, Exhibit B-2, p. 163). Some other B2 compounds are aidrin, auramine, carbon tetrachloride, chlordane, chloroform, DDT, ethylene dibromide, and formaldehyde (U.S. EPA, 1986b, Exhibit A-4, pp. 137-140). The next two categories above B2 are: (1) B 1: “Probable human carcinogen” based on “limited evidence of carcinogenicity in humans from epidemiologic studies.” Examples (inhalation as exposure route) include acrylonitrile, beryllium, and cadmium.

12

WATSON,

JONES, AND GRIFFIN

(2) A: “Human carcinogen” based on “sufficient evidence from epidemiologic studies to support a causal association between exposure and cancer.” Examples (inhalation as exposure route) are arsenic, asbestos, benzene, benzidine, chromium VI, diethylstilbestrol (DES), nickel, and vinyl chloride. “Mustard gas” is grouped by IARC with arsenic, asbestos, diethylstilbestrol, and vinyl chloride in a category that approximates the U.S. EPA Category A (IARC “Group 1,” Saracci, 198 1). This IARC class also includes industrial processes (such as nickel refining) that are considered carcinogenic to humans. APPLICATION

OF RASH

METHOD

TO MUSTARD

DATA

Since the principal concerns expressed by the public regarding possible long-term, low-dose mustard exposure focus on this compound’s carcinogenicity, we have limited our assessment to an analysis of tumorigenicity data only. In so doing, we are knowingly restricting the present analysis to a subset of an already small data base for the sake of rigor. Exposure to toxic concentrations of sulfur mustard also results in other acute and well-known mammalian responses inherent to most vesicant agents, such as chemical bums of the skin and respiratory tissues and eye irritation. These effects are not assessed in the current evaluation. Of the published literature on sulfur mustard, we have found only two groups of investigators that present documentation of mustard’s positive tumorigenic potential. These comprise both injection and inhalation exposures of laboratory rodents: intravenous exposures of strain A mice (Heston, 1950), subcutaneous exposure of strains A, C3H, and C3Hf mice (Heston, 1953) and the inhalation exposure of strain A mice (Heston and Levillain, 1953) and laboratory rats (McNamara et al., 1975). There are two other investigations (Fell and Allsopp, 1948; Berenblum, 1929 as cited in Fell and Allsopp, 1948) where liquid solutions of sulfur mustard were repeatedly applied directly to the skin of laboratory mice for as long as 278 days. The maximum cumulative dose documented was 2.0 mg/individual (Fell and Allsopp, 1948); no tumors were observed in the 12 months following cessation of exposure. While the test population in the latter work was too small for a definitive finding, these data do indicate that skin cancer induction does not appear to be a concern at the tested doses. Careful reading of the former reports caused us to eliminate Heston and Levillain (1953) from further consideration. The treatment protocol (a strip of filter paper saturated with 0.01 cc mustard and suspended in an 8-liter desiccator in which treatment groups were exposed for 15-min intervals) imposed a massive (in the mg/m3 range) acute dose on the test population. Immediate toxic response of chemical bums to the skin around the ears, nose, and eyes was first noted, followed by several direct-exposure fatalities (13 deaths out of 80 individuals, or 16.3%) and reduced weight gain among the survivors for a period of 10 days postexposure. These data are not comparable to those of modem standard tests for tumorigenicity, where care is taken to expose test populations to sublethal concentrations. The atmospheric concentration to which these mice were exposed is also questionable, and the investigator freely admitted that the degree of mustard evaporation from the filter-paper strip was unknown. For these reasons, data from the Heston and Levillain (1953) report was not included in the present analysis. Heston’s other two reports employ injection as the route of exposure (Heston, 1950; Heston, 1953). While this route is not directly applicable to expected human

CARCINOGENIC

POTENCY

OF SULFUR

MUSTARD

13

exposures during handling of mustard containers and operation of a mustard incinerator, there is little question about actual dose received by test animals under this treatment protocol. Strain A mice were selected for their genetic susceptibility to pulmonary tumor induction (“incidence of spontaneous pulmonary tumors . . . is approximately 90% in animals 18 months of age and 50% in animals 12 months of age” (Heston, 1942)). At the time of Heston’s 1950s work, strain A mice were considered a sensitive indicator for carcinogenicity testing, particularly by intravenous exposure. Strain C3H had been developed for its susceptibility to spontaneous mammary tumors as well as production of subcutaneous tumors following exposure to polycyclic hydrocarbon carcinogens (Heston, 1953). Strain C3Hf is similar to C3H, but does not carry the susceptibility to high mammary tumor induction (Heston, 1953). In both of these reports, Heston had difficulty developing a dose regimen that the animals could tolerate and his early experimental groups exhibited high mortality (approximately 50% within a few days of treatment). This experience was used to scale down doses so that the later test groups could survive long enough to develop tumors. Again, Heston’s early high mortality results were not incorporated into the present analysis. For calculational purposes, the dose(s) at which tumorigenicity was noted is considered a toxic dose (TD). Values of TD derived from Heston’s work are presented in Table 2. Noted tumorigenicity included significant differences in pulmonary tumor incidence and average number of nodules between the test group and controls in the iv exposed strain A mice, and significantly greater incidence of neoplasms at the injection site in mice exposed subcutaneously. Subcutaneous exposure of sulfur mustard did not increase the occurrence of pulmonary tumors in strain A mice. Three of the sarcoma types observed in the subcutaneous test group were successfully transplanted through 13 to 44 transplant generations and are considered malignant (Heston, 1953). The single remaining reference that examines laboratory animal exposure to mustard in air is that of McNamara and his colleagues, who performed toxicological evaluations of several warfare agents in the Biomedical Laboratory at Aberdeen Proving Ground, Maryland (APG; McNamara et al., 1975). Two atmospheric exposures were used: (1) 0.001 mg/m3, 24 hr/day, 5 days/week, for up to 1 year, and (2) 0.1 mg/m3 for 6.5 hi-/day + 0.0025 mg/m3 for 17.5 hi-/day, 5 days/week for up to 1 year. Test animals included dogs, rabbits, guinea pigs, Sprague-Dawley-Wistar rats, and strains ICR Swiss-albino and A/J mice (A/J mice are an inbred strain with a high frequency of spontaneous pulmonary tumors) (McNamara et al., 1975). Both at the low (0.00 1 mg/m3) and high (intermittent exposure to 0.1 mg/m3) exposures, no tumors (0%) were noted in guinea pigs, rabbits, or dogs. No significant numbers of tumors were noted for mice at either exposure level (3.2 and 2.2%, respectively, as compared to the unexposed mouse tumor incidence of 5.6%). Rats in the low-exposure group also exhibited no significant numbers of tumors (10.4%) when compared to unexposed controls (7.4%). However, rats exposed to the elevated concentration developed significant numbers of skin cancers at 12 weeks of exposure. At the end of the 52-week study period, 25 of the 57 rats (43.9%) in the elevated exposure group exhibited agent-attributable cancers (the majority of which were squamous cell carcinomas or primary epidermal tumors). When corrected for spontaneous tumor pro-

14

WATSON,

JONES,

AND

TABLE

GRIFFIN

2

LITERATURE-DERIVED VALUES OF TOXIC DOSE (MG/KG) FOR SULFUR MUSTARD TUMORIGENICITY Exposure regimen Four iv injections on alternate days of 1:10 dilution of 0.06 to 0.07% solution in distilled water Six sc injections at weekly intervals of0.05 cc of 0.05% solution in olive oil Six sc injections at weekly intervals of 0.05 cc of 0.05% solution in olive oil Five SCinjections at weekly intervals of 0.05 cc of 0.05% solution in olive oil Six SCinjections at weekly intervals of 0.05 cc of 0.05% solution in olive oil 0.1 mg/m3 approx 6.5 hr/day plus 0.0025 mg/m3 approx 17.5 hr/day, 5 days/week. up to 1 year

Toxic dose

Reference

Strain A mice

3.5 mg/kg”

Heston, 1950

Strain C3H mice

6.8 mg/kg’

Heston, 1953

Strain C3Hf mice

7.5 mg/kg’

Heston, 1953

Strain A mice

6.3 mg/kg’

Heston, 1953

Strain C3H mice

6.0 mg/kgd

Heston, 1953

Test animal

Sprague-Dawley-Wistar

Rats

12.2 mg/kg

McNamara et al., 1975

” From Fig. 1 (Heston. 1950). initial weight of treated mice was 2 1 g (female) and 24 g (male). Dose estimate assumed 0.02 kg. ’ From Fig. 1 (Heston. 1950). assume 22 g. ‘ From Fig. 1 (Heston, 1950), assume 20 g. Experiment replicated twice at same dose on strain A. ’ From Fig. 1 (Heston, 1950). an all-male test group, assume 25 g. Experiment replicated twice at same dose on strain C3H.

duction in controls, the 52-week treatment group demonstrated a significant tumor incidence of 36%. For calculational purposes, the staggered atmospheric exposure received by the elevated group was converted to an estimated dose (mg/kg) with standard assumptions of adult rat body weight (Lewis and Sweet, 1985), respiration frequency, and tidal volume (Jones and Walsh, 1985). Exposure to this concentration was considered intermittent (i.e., 5 days/week) (see Table 2). Compared to most compounds we evaluate with the RASH approach, available data for mustard tumorigenicity are sparse. We prefer a wider distribution of scaled doses over exposure periods approximating the lifetime of the test animal. In addition, only one published report presented data for the exposure route of interest (i.e., atmospheric) (McNamara et al., 1975). Nevertheless, it is always rare to find completely adequate data sets; the original investigators usually had different objectives than ours. The only option is to acknowledge limitations and use what is available in a reasonable manner. We consider the analyses presented below to fall into this category. Examples of relative potency factor calculations follow. The RTECS 1983- 1984 Supplement (Lewis and Sweet, 1985) was used as the principal data source document for primary and secondary standards. Tumorigenic data for B[a]P were also gleaned from independent literature searches. Guidelines provided in Table 1 apply.

CARCINOGENIC

For comparable

intravenous

POTENCY

OF SULFUR MUSTARD

15

exposure in the mouse, D BLaIP:TD = 10 mg/kg DHD: TD = 3.5 mg/kg

RPF

=

+

= 8

=

test

2.86

f

(see Table 3).

When data could be matched for a secondary standard such as N-nitrosodimethylamine, relative potency was estimated in the following manner: RPF=

(F)(2).

As previously stated, the median potency of DMNA +=

relative to B[a]P is

= 0.23.

DMNA

For secondary standard subcutaneous exposure in the mouse, DDMNA: TD = 24.0 mg/kg DHD: TD = 7.5 mg/kg RPF = (0.23)

(see Table 3).

The above process was repeated for each of the TD values presented in Table 2 with comparable tumorigenic data. Benzo[a]pyrene and DMNA were the primary and secondary standards, respectively (see Table 3 for individual values and documentation of sources). Note that there was no good match of McNamara’s mustard inhalation data under strict adherence to the rules for tumor study comparison as outlined in Table 1. Even though tumor development is the common endpoint used throughout the present analysis, the incidence of tumor indication in the paired test populations varies. The authors are well aware of the concerns some investigators have regarding our treatment of these data as like pairs. However, the reader should understand that small increments in dose often generate strong response effects once a critical threshold of time and/or concentration is attained. An excellent example of this obstudy, a chemical servation is the data gathered during the “ED 01” (“Megamouse”) carcinogenesis dose-response experiment designed to determine the dose of certain carcinogens necessary to induce tumors in 1% of the mouse population (N > 24,000) tested (Society of Toxicology Task Force, 1981). Incidence of bladder cancer in BALB/C female mice exhibited sharp increases at dose increments of 2-acetylaminofluorene between 15 and 25 ppm. Thus, we consider the comparison of tumor incidence results presented in Table 3 to be appropriate. If all RASH computation guidelines (Table 1) are strictly applied with regard to treatment time, only the first two values presented in Table 3 (i.e., 2.9 and 11.1) would be valid. The median of this array would be 7.0 as an estimate of mustard potency relative to B[a]P. If treatment time is disregarded in order to minimize uncertainty due to data selection, the number and range of RPF estimates is considerably expanded. Thirty-four RPF values result and are ranked as follows: 0.3,0.3,0.3,0.3,

16

WATSON,

JONES, AND GRIFFIN TABLE 3

RELATIVEPOTENCYESTIMATESDERIVEDFROMMUSTARDTUMORDATA Chemicals HD

BblP HD

BblP HD

W#’ HD

BblP HD WIP HD

BMP HD

BblP HD

BblP HD

BblP HD

WIP HD

WIP HD

BblP HD

BW HD

BblP HD

B[alP HD

BblP HD

BblP HD

WP HD DMNA HD DMNA HD DMNA HD DMNA

Biological test

Type of estimate”

ivn-mus ivn-mus ivn-mus ivn-rat scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus

TDss T&5 T&S TDm TD, Tb-65 TD, TDe-65 TD, TDes-65 TD, TDms TD, The9 TD, T@-69 TD, -W-m TD, T&L69 TD7 TD, TD, TD-, TD, TD, TD, TD, TD, TD, Tb TD, TD, TD, TJ% TD, TD, TD.u TD7 TL TD, TD,, TD, TJ-L

Dose 3.5 mg/kg/7D-I 10 mg/kgb 3.5 mg/kg/‘JD-I 39 mg/kg/6D-1’ 6.8 mg/kg/6W-I 9 mg/kgd 1.5 mg/kg/6W-I 9 Wkd 6.25 mg/kg/SW-I 9 wlkgd 6.0 mg/kg/6W-I 9 wlkd 6.8 mg/kg/6W-I 15 mg/kg’ 7.5 mg/kg/6W-I 15 mg/kg’ 6.25 mg/kg/6W-I 15 mg/kg’ 6.0 mg/kg/6W-I 15 mg/kg’ 6.8 mg/kg/6W-I 2 m&s” 7.5 mg/kg/6W-I 2 mglk$ 6.25 mg/kg/6W-I 2 m&s 6.0 mg/kg/6W-I 2 w/k’ 6.8 mg/kg/6W-I 2 mg/kg/l2M-IP 7.5 mg/kg/6W-I 2 mg/kg/ 12M-Ig 6.25 mg/kg/6W-I 2 mg/kg/ 12M-P 6.0 mg/kg/6W-I 2 mg/kg/ 12M-1” 6.8 mg/kg/6W-I 24.0 mg/kgh 7.5 mg/kg/6W-I 24.0 mg/kg” 6.25 mg/kg/5W-I 24.0 mg/kgh 6.0 mg/kg/6W-I 24.0 mg/kgh

Potency relative to BlalP 2.86 11.14 1.32 1.20 1.44 1.50 2.21 2.00 2.40 2.50 0.30 0.27 0.32 0.33 0.30 0.27 0.32 0.33 0.81 0.74 0.88 0.92

CARCINOGENIC

POTENCY

17

OF SULFUR MUSTARD

TABLE 3-Continued Chemicals HD

BblP HD

JWIP HD

WP HD

WG' HD

WP HD

BblP HD

BblP HD

BblP HD

BblP HD

BblP HD

BblP HD

WalP

Biological test

Type of estimate’

scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-mus scu-ham scu-mus scu-ham

TD, T&4 TQ T&4 Tb T&4 TD, T&., TD7 -WV TD, ‘-WV Tb TD93 TD7 T&3 TJA TD,z-64 TD7 T&-64 TD7 =k-m TD, TDn-64

scu-mus scu-ham scu-mus scu-ham scu-mus scu-ham scu-mus scu-ham scu-mus scu-ham scu-mus scu-ham

Dose 6.8 mg/kgJ6W-I 160 mg/kg’ 7.5 mg/kg/6W-I 160 mg/kg’ 6.25 mg/kg/6W-I 160 mg/kg’ 6.0 mg/kgJ6W-I 160 mg/kg’ 6.8 mg/kg/6W-I 20 mg/kg’ 7.5 mg/kg/6W-I 20 mg/kg’ 6.25 mg/kg/SW-I 20 mg/kg’ 6.0 mg/kg/6W-I 20 mg/kg’ 6.8 mg/kg/6W-I 4.5 mg/kgk 7.5 mg/kg/6W-I 4.5 mglkgk 6.25 mg/kg/6W-I 4.5 mg/kgk 6.0 mg/kg/6W-I 4.5 mg/kgk

Potency relative to B[a]P 23.5 21.3 25.6 26.7 2.94 2.67 3.20 3.33 0.66 0.53 0.72 0.75

a Subscript indicates tumor incidence. ’ Andervont ( 1940); pulmonary tumors in strain A mice approx 16 weeks after single injection. ’ Pataki and Huggins (1969); mammary tumors in female Sprague-Dawley rats receiving three iv injections over 6 days. Sacrificed approx 48 days after first injection. d Buening et al. (1983); injection site tumors in male C3H/tCum mice receiving single injection of B[a]P in either DMSO or trioctanoin carrier. Tumor incidence at 15 months postinjection; variation due to carrier. ’ Nagata et al. (197 1); injection site tumors in female ddN mice given single injection of B[a]P in either tricaprylin or benzene carrier. Termination at either 250 or 430 days post-treatment. f Bryan and Shimkin (1943); injection site tumors in male C3H mice given single injection of B[a]P in tricaprylin carrier; 3.7 to 9.0 month latency. g Hueper et al. (1962); injection site tumors in C57 black mice. Repeated monthly injections for 12 months. Tumor incidence at 12 months after last injection. h Toth and Shubik (1967); pulmonary tumors in AKR mice treated at birth with single injection. Latent period of 36-47 weeks. ’ Toth and Shubik (1967); pulmonary tumors in AKR mice treated at birth with single injection. Latent period of 36-47 weeks. JHalberstaedter (1939); polymorph cell sarcomata in golden hamsters after 3-month latency period. k Homburger et al. (1972); injection site sarcomas among various inbred hamster strains after 20-60 week latency; single injection.

18

WATSON,

JONES, AND GRIFFIN

0.3,0.3, 0.3,0.3, 0.6, 0.7,0.7, 0.7, 0.75, 0.8, 0.9,0.9, 1.2, 1.3, 1.4, 1.5, 2.0, 2.2, 2.4, 2.5, 2.7, 2.9, 2.9, 3.2, 3.3, 11.1, 21.3, 23.2, 25.6, and 26.7. The median value ofthis latter array is 1.3 and the interquartile range is 0.6 to 2.9. The narrow interquartile range suggests that the dispersion about the median estimate of mustard relative potency is less than’an order of magnitude. The entire range of estimated relative potency values covers two orders of magnitude and is small compared to the range determined for other compounds scored by this procedure (Jones et al., 1988). The RPFs from these two arrays, i.e., 7.0 and 1.3, are not significantly different from one another and illustrate the stability of the median statistic as an estimate of central tendency. Use of a large array, where each endpoint for the test compound is compared with compatible endpoints for the standard compound(s), can be further justified as a means of minimizing the uncertainty produced by the existence of a small data set for the test compound (sulfur mustard in this case). A weighting bias may result from this procedure. The authors are aware of this possibility and would prefer more and better data to characterize potency under a greater variety of test conditions. Until (and unless) such data are generated, the authors consider that analysis of an array of ratios is a reasonable compromise. The median represents a carcinogenic (via tumor production) potency value for sulfur mustard relative to B[a]P (“relative potency”); the interquartile range is a crude measure of uncertainty stemming from experimental design, random error, and biological variability. In other words, the limited animal data available indicate that sulfur mustard is likely to approximate B[a]P in its ability to produce malignant tumors in man (assuming that mouse, rat, and hamster data are indicative of human response). Heston had speculated along these lines when he pointed out that comparable lung tumor response was observed in strain A mice after iv injection exposure to similar doses of either sulfur mustard or the polynuclear aromatic hydrocarbon methylcholanthrene (C2,Hi6) (Heston, 1950). (Note that methylcholanthrene has a RPF = 1.5 (Jones et al., 1988).) The authors obtained consistent results by using bischloromethyl ether (BCME; CAS No. 542-88-1, C2H4C120) as a secondary standard in estimating an RPF for sulfur mustard. Both compounds exhibit comparable modes of action in biological systems and a reasonable argument could be made that similar mechanisms are responsible for cellular damage observed after exposure to either BCME or sulfur mustard. Both compounds are alkylating agents, powerful lung and eye irritants, both can generate necrotic skin lesions following direct contact (Gosselin et al., 1984; Windholz et al., 1983; Medema, 1986; Vogt et al., 1984; Papirmeister et al., 1984), and both are classified as Group I carcinogens by IARC (Saracci, 198 1). Careful comparison of BCME toxicity data provided in RTECS (Lewis and Sweet, 1985; Go&in et al., 1984) with that available for mustard (Watson et al., 1988) generates an array of 12 values for sulfur mustard relative potency with a median (i.e., RPF) equal to 0.6 and an interquartile range of 0.2 to 0.9. The median RPF value of 0.6 for sulfur mustard with BCME as the secondary standard is strikingly similar (i.e., within an order of magnitude) to values estimated immediately above ( 1.3 and 7.0) and derived from a comparison with B[a]P and DMNA data. Expert committee deliberations of toxicological potency (e.g., U.S. EPA-CAG, American Conference of Governmental Industrial Hygienists) rarely exhibit less than an order-of-magnitude uncertainty; variability in potency determinations is usually much greater (Jones et al., 1988).

CARCINOGENIC

APPLICATION

POTENCY

OF SULFUR MUSTARD

OF RELATIVE POTENCY TO RISK ESTIMATION

An accepted model for estimating old dose-response relationship is

19

FACTOR

lifetime cancer risk assuming a linear nonthreshRisk = (p)(O),

where Risk is the additional individual lifetime risk of developing cancer based on a lifetime of continuous exposure to dose D of a compound with the potency factor Q*. Units of p are dose reciprocal, i.e., [(mg/kg)/day]-‘, and units of dose are [(mg/ kg)/day]. In the current analysis, the risk estimate is a measure of potential cancer incidence (i.e., tumorigenicity and not cancer deaths). Common assumptions for inhalation exposure are a 70-kg person inhaling 20 m3/day (Anderson and U.S. EPACAG, 1983). Values of p are based on the upper 95% confidence limit of the linearized dose response for animal test results judged by expert selection to be most representative of man. Thus, any cancer risk estimate derived by use of p will represent an upper bound. The value of p used in subsequent calculations,

is the inhalation potency factor for B[a]P documented in the Superfund Public Health Evaluation Manual (U.S. EPA, 1986a, Exhibit C-4, p. C-24). The authors applied this risk model to recent estimates of potential mustard releases expected from normal operation of a hypothetical agent incinerator at APG (as prepared by the U.S. Army Environmental Hygiene Agency (Hunter and Oliverson, 1987)). Under reasonable assumptions of local meteorology, stack gas temperature and fugitive emission release, the expected annual concentrations of HD at selected receptor sites along the APG facility boundary range between 0.0 13 X lop6 mg/m3 and 0.037 X lop6 mg/m3. (Hunter and Oliverson, 1987, Table 3). Thus, the following risk estimate may be performed. For the estimated minimum concentration at the APG site boundary, Min dose = (0.013 X 10-6?#$$)(&--) = 0.0037 x 1o-6 e. Similarly,

the Maxdose=O.OllOX

10p6-mglkg day .

In the following estimates of lifetime cancer risk, we have adjusted the mustard RPF (derived from iv and subcutaneous exposure data) by the absorption coefficient of B[a]P inhalation (0.29) (Jones et al., 1987). The absorption coefficient of an injection route is assumed to equal 100%.

20

WATSON,

For Risk = (Q*)(O)(RPF),

JONES, AND GRIFFIN

and an RPF = 1.3,

and Lifetime

Risk at max dose = 2.6 X lo-*.

For an RPF of 7.0, Lifetime

Risk at min dose = 0.05 X 1Oe6 = 5.0 X lo-‘,

Lifetime

Risk at max dose = 0.14 X lop6 = 1.4 X lo-‘.

and

Thus, the estimated excess individual lifetime cancer risk due to inhalation of potential sulfur mustard emissions at receptor sites along the Edgewood Arsenal boundary ranges between approximately 1 X 10e8 and 1 X 10w7.These values assume lifetime inhalation of HD at concentrations ranging between 0.013 and 0.037 ng/m3 (e.g., 0.0 13 and 0.037 X 10e6 mg/m3). Scheduled operation of the agent incinerator for approximately 5 years rather than a 70-year lifetime will further reduce the calculated risk by a factor of 14 (i.e., a range of 6.6 X 10-l’ to 1.0 X 10e8). Lifetime cancer risk estimates on the order of 1O-6 are considered insignificant by most individuals and regulatory authorities. Releases that generate cancer risks equal to values less than 10V6 are not usually regulated by the U.S. EPA or the Food and Drug Administration (U.S. FDA). For comparison, the limit of regulatory acceptability for some pesticides in foodstuffs is a lifetime cancer risk of 10m4(Norman, 1987). Similar calculations can also be used to evaluate the excess lifetime cancer risk from exposure to the existing approved general population (GPL) and occupational exposure levels as established by Surgeon General’s Working Groups (Department of Health and Human Services) (GPL, 1 X lop4 mg HD/m3 for a 72-hr time-weighted average; occupational, 3 X 10e3 mg HD/m3 for an 8-h time-weighted average) (U.S. DHHS, 1987). The resulting risk estimates developed with assumptions of repeated exposures, body mass, and inspiration as above, are presented in Table 4. COMPARISON

WITH

OTHER

REGULATORY

STANDARDS

A recent examination of the last decade’s federal regulatory decisions in which some measure of risk was estimated prior to promulgation demonstrates that protective action for carcinogen exposure was always invoked if the maximum estimated individual lifetime risk exceeded 4 X low3 (Travis et al., 1987). There has been only one instance where protective action was taken to reduce estimated lifetime cancer risk at levels below 1 X lop6 (U.S. FDA control of DMNA in baby bottle manufacture, R = 4 X lo-* (Travis et al., 1987)). Reasonable maximum estimated lifetime cancer risks for 5 years of incinerator operation and daily exposure to mustard at the general population and occupational time-weighted average (Risk = 2.5 X 10e5 and 2.6 X 10p4, respectively) fall into the intermediate category described above (i.e., be-

CARCINOGENIC

POTENCY

OF SULFUR

MUSTARD

21

TABLE 4 ESTIMATED EXCESS LIFETIME CANCER RISKS FROM POPULATION EXFQSURE TO POTENTIAL SULFUR MUSTARD EMISSIONS DURING AGENT INCINERATION Period of exposure Estimated exposure APG facility boundaryd 0.013 X 10mhmg/m’ (min) 0.037 X 10m6mg/m3 (max) Approved levels’ General public: 1 X low4 mg/m3 (72-hr TWA) Occupational: 3 x IO-’ mg/m’ (8-hr TWA)

Human lifetime”,b

Period of incinerator operatio&

9.3 x 1om9to 5.0 x lo-” 2.6 x IO-* to 1.4 X lo-’

6.6 x lo-“to 3.6 x 1O-9 1.9 x 1o-9 to 1.0 x 1o-8

6.7 x 1O-5 to 3.5 X 10m4

4.8 x 10m6 to2.5 X 1O-5

6.7 x lo-“ to 3.6 X 10m3

4.7 x 10m5 to 2.6 x 10-4’

’ Assume 70 years of continuous exposure (24 hr/day) for general public and 50 years of daily exposure (8 hr/day) for occupational population at the given concentration, 70-kg body mass, 20 m3/day inspiration rate. ’ Range bounded by minimum and maximum estimates of the relative potency factor (RPF) for sulfur mustard (i.e.. RPF = 1.3.7.0). The best RPF estimate is 1.3. ’ Assume 5-year period of incinerator operation. d Minimum and maximum atmospheric concentrations calculated for Aberdeen Proving Ground (APG) facility boundary during incinerator operation (Hunter and Oliverson, 1987). ’ Established by Surgeon General’s Working Groups of the U.S. Department of Health and Human Services (U.S. DHHS, 1987). Time-weighted average (TWA). ‘Assume worker exposure to the occupational limit for entire shift every day, 5 days/week, 50 weeks/ year.

tween 4 X 10e3 and 1 X 10-(j). Agency guidelines used in regulating carcinogens with risk estimates between these two bounds are: ( 1) taken (2) (3) (4) above

size of affected population (in small populations, regulatory action has not been for individual risk values below I X 10e4), use of best available control technology (BAT), availability of substitutes, and cost effectiveness of regulation, which is primarily determined by (2) and (3) (Travis et al., 1987).

It is useful here to point out that systematic evaluation of standards governing the pesticide content of foodstuffs has shown that the U.S. EPA currently accepts some oncogenic pesticides at concentrations estimated to pose a cancer risk on the order of 10e4 (Norman, 1987; U.S. EPA, 1982b). One example is ethylene bisdithiocarbamate fungicides, widely used on crops and omamentals and as a slimicide in cooling systems and paper mills (Risk = 5 X 10m5to 5 X 10e4) (U.S. EPA, 1982b)). This estimate is illustrative of many enforcement standards established prior to widespread application of quantitative risk analysis methodologies. Additional perspective can be obtained by considering other industrial risks. In August of 1987, the National Council of Radiation Protection (NCRP) expanded the

22

WATSON,

JONES, AND GRIFFIN

concept of comparable risk as a basis for occupational radiation exposure limits in that “the level of protection provided for radiation workers should, as far as possible, be comparable with that in other single ‘safe’ industries” (NCRP, 1987, p. 6). Examination of historical data maintained by the National Safety Council (NSC) defined “safe” industries “as those having an associated annual fatality accident rate of one or less per 10,000 workers, i.e., an average annual [fatality] risk of 10-4” (NCRP, 1987, p. 8). For comparison, a decidedly unsafe industry such as mining or quarrying exhibits an annual accidental fatality rate of 6 X 10e4 in the United States (NSC, 1985, as cited in NCRP, 1987). There is obviously still room for improvement, but the annual risks of accidental death associated with these safe and unsafe industries illustrate rates that have previously been considered acceptable ( 1 X 1Om4per year) or not (6 X 10e4 per year). The Occupational Safety and Health Administration (OSHA) takes a similar view in making a determination on the safety of a given workplace, and does not generally regulate workplace cancer risks where the predicted lifetime risk is less than 1 X lop3 (this value roughly corresponds to a worker lifetime [45year] fatal accident risk in a “safe” industry) (Rodericks et al., 1987). The estimated maximum excess risk (i.e., 2.6 X 10m4) of developing a malignancy (not necessarily fatal) after 5 years of continuous occupational exposure to currently allowable quantities of sulfur mustard is comparable in value to that of the accidental fatality risk (i.e., 1.0 X 10-4) in a nominally “safe” or moderately safe industry for a l-year work period. However, the human health consequences of the two risk estimates (i.e., case of malignancy vs accidental death) are strikingly different, as are the bases for their conclusion. The “safe” industry fatality number is derived from actual data, while the mustard incinerator worker health risk estimate is derived from maximum extrapolations of animal data and a calculated upper limit model fit. The resulting incinerator worker cancer risk is a clear maximization. It has been 12 years since McNamara and his colleagues first proposed the current general public and occupational standards for mustard based on extrapolation from 1 year of atmospheric exposure of a rat colony at 1 X 1O-3 mg/m3 (McNamara et al., 1975). He and his colleagues proposed a 10X protective factor for the general public (72-hr time-weighted average) and a 1 X protective factor for workers (8-hr timeweighted average). McNamara’s work remains the only mammalian in vivo longterm, low-dose atmospheric exposure experiment of record with mustard agent and was performed with protocols appropriate at the time. In the intervening years, experimental guidelines have changed while individual investigators and select committees have developed methods to estimate maximum lifetime risks of low dose exposures. One method has been applied in the above analysis. The resulting maximum estimates of lifetime cancer risk are not large, but would not be considered negligible by everyone. ACKNOWLEDGMENTS The authors acknowledge useful comments and discussion provided during preparation of this analysis by Max D. Morris (Engineering Physics and Mathematics Division, Oak Ridge National Laboratory (ORNL); Bruce A. Owen, Clay E. Easterly, Alan R. Hawthorne, and Phillip J. Walsh (all of Health and Safety Research Division, ORNL); Barbara Bass (EA Engineering, Science, and Technology, Inc., Sparks, MD); Sanford Leffingwell (Center for Environmental Health, CDC, Atlanta, GA) and William Lijinsky (Laboratory of Chemical and Physical Carcinogenesis, National Cancer Institute, Frederick, MD). Oak

CARCINOGENIC

POTENCY

OF SULFUR

MUSTARD

23

Ridge National Laboratory operated by Martin Marietta Energy Systems, Inc., under contract DE-ACOS84OR21400 with the U.S. Department of Energy.

REFERENCES AMES, B. N., MACAW, R.. AND GOLD, L. S. (1987). Ranking possible carcinogenic hazards. Science 236, 271-280.

ANDERSON, E. L., AND EPA-CAG (1983). Quantitative approaches in use to assesscancer risk. Risk Anal. 3,277-295.

H. B. (1940). Biologic testing of carcinogens. II. Pulmonary-tumor induction technique. J. Nat/. Cancer Inst. 1,225-239. BEEBE, G. (1960). Lung cancer in World War I veterans: Possible relation to mustard-gas injury and 1918 influenza epidemic. J. Natl. Cancer Inst. 25, 123 l-1252. BERENBLUM, I. (1929). The modifying influence of dichloroethyl sulfide on the induction of tumors in mice by tar. J Pathol. Bacterial. 32,425-434 (as cited in Fell and Allsopp, 1948). BRYAN, W. R., AND SHIMKIN, M. B. (1943). Quantitative analysis of dose-response data obtained with three carcinogenic hydrocarbons in strain C3H male mice. J. Natl. Cancer Inst. 3,503-53 1. BUENING, M. K., LEVIN, W., WOOD, A. W., CHANG, R. L., AGRANAT, I., RABINOVITZ, M., BUHLER, D. R., MAH, H. D., HERNANDEZ, O., SIMPSON, R. B., JERINA, D. M., CONNEY, A. H., MILLER, E. C., AND MILLER, J. A. (1983). Fluorine substitution as a probe for the role of the 6-position of benzo(a)pyrene in carcinogenesis. J. Nat/. Cancer Inst. 71,309-3 15. CASE, R., AND LEA, A. (1955). Mustard gas poisoning, chronic bronchitis and lung cancer. An investigation into the possibility that poisoning by mustard gas in the 19 14- 19 18 war might be a factor in the production of neoplasia. Brit. J. Prev. Sot. Med. 9,62-72 (as cited in IARC 1975). FELL, H., AND ALLSOPP, M. (1948). The effect of repeated applications of minute quantities of mustard gas on the skin of mice. Cancer Res. 8, 177- 18 1. GLASS, L. R. (1987). Use of Short-Term Test Systems for the Prediction of the Hazard Represented by Potential Chemical Carcinogens. Ph.D. thesis, School of Public Health, University of Texas, Houston, Texas. GOSSELIN, R. E., SMITH, R. P., HODGE, H. C., AND BRADDOCK, J. E. (1984). Clinical Toxicology of Commercial Products, 5th ed. Williams & Wilkins, Baltimore. HALBERSTAEDTER, L. (1939). A benzpyrene tumour strain in hamsters with tendency to metastasis formation. Nature(London) 144,377-378. HESTON, W. E. (1942). Inheritance of susceptibility to spontaneous pulmonary tumors in mice. J. Natl. Cancer Inst. 3,79-82. HESTON, W. E. (1950). Carcinogenic action of the mustards. J. Natl. Cancer Inst. 11,4 15-423. HESTON, W. E. (1953). Occurrence of tumors in mice injected subcutaneously with sulfur mustard and nitrogen mustard. J. Natl. Cancer Inst. 14, 13 1- 140. HESTON, W. E., AND LEVILLAIN, W. D. (1953). Pulmonary tumors in Strain A mice exposed to mustard gas. Proc. Sot. Exp. Biol. Med. 82,457-460. HOMBURGER, F., HSUEH, S-S, KERR, C. S., AND RUSS~ELD, A. B. (1972). Inherited susceptibility of inbred strains of Syrian hamsters to induction of subcutaneous sarcomas and mammary and gastrointestinal carcinomas by subcutaneous and gastric administration of polynuclear hydrocarbons. Cancer Res. ANDERVONT,

32,360-366.

HUEPER, W. C., KOTIN, P., TABOR, E. C., PAYNE, W. W., FALK, H., AND SAWICKI. E. (1962). Carcinogenic bioassay on air pollutants. Arch. Pathol. 74,89-107. HUNTER, C. H., AND OLIVERSON, F. W. (1987). Air Pollution Consultation No. 43-21-1229-87, Evaluation of Multiple Incinerator Air Quality Impacts. EdgewoodArea. Aberdeen Proving Ground, MD, May 1987. Dept. ofthe Army, U.S. Army Environmental Hygiene Agency, Aberdeen Proving Ground, MD 210105422.

ICF Inc. ( 1987). Discussion of Options for Revising the Hazard Ranking System (HRS) Toxicity Factor. U.S. EPA Publication No. 105NCP-HRS-6-2. ICF Inc., 1850 K Street NW, Washington, DC 20006. INADA, S., HIRAGUN, K., SEO, K., ANDYAMURA, T. (1978). Multiple Bowen’s Disease observed in former workers of poison gas factory in Japan, with special reference to mustard gas exposure. J. Dermatol. 5, 49-60.

24

WATSON,

JONES, AND GRIFFIN

International Agency for Research on Cancer (IARC) (1975). Monograph on the Evaluation ofthe Carcinogenic Risk of Chemicals to Man from Aziridines, N-, S- and O-mustards and Selenium. Vol. 9, WHO, Geneva, Switzerland. JONES,T. D., GRIFFIN, G. D., AND WALSH, P. J. (1983). A unifying concept for carcinogenic risk assessments. J. Theor. Biol. 105,35-6 1. JONES, T. D. (1984). A unifying concept for carcinogenic risk assessments: Comparison with radiationinduced leukemia in mice and men. Health Phys. 47,533-558. JONES, T. D., WALSH, P. J., AND ZEIGHAMI, E. A. (1985). Permissible concentrations of chemicals in air and water derived RTECS entries: A “RASH” chemical scoring system. J. Toxicol. Ind. Health 1,2 13234.

JONES, T. D., AND WALSH, P. J. (1985). Animal studies and prediction of human tumors can be aided by graphical sorting of animal data. Amer. J. Ind. Med. 7,185-2 17. JONES, T. D., OWEN, B. A., AND WELLS, S. M. (1987). Evaluation of Human Health Risks from Mixtures of Hazardous Chemicals and Radionuclides, ORNL/RAP-38. Oak Ridge National Laboratory, Oak Ridge, TN 37831-6101. JONES, T. D., WALSH, P. J., WATSON, A. P., OWEN, B. A., BARNTHOUSE, L. W., AND SANDERS, D. A. (1988). Chemical scoring by a Rapid Screening Hazard (RASH) method. Risk Anal. 8,99-l 18. KLEHR, N. (1984). Cutaneous late manifestations in former mustard gas workers. Z. Hautkrankh. 59, 1161-1170. LEWIS, R. L., SR., AND SWEET, D. V. (1985). Registry of Toxic Effects of Chemical Substances (RTECS). 1983-1984 Supplement, Vol. I, A-G. DHHS (NIOSH) Publication No. 86-103, PHS CDC, NIOSH, Cincinnati, OH 45226. L0HS, K. (1975). Delayed Toxic Eficts of Chemical Warfare Agents. SIPRI, Almqvist & Wiksells, Stockholm and New York. MCNAMARA, B. P., OWENS, E. J., CHRISTENSEN,M. K., VOCCI, F. J., FORD, D. F., AND ROZIMAREK, H. (1975). Toxicological Basis for Controlling Levels of Mustard in the Environment. Edgewood Arsenal Special Publication EB-SP-74030, Aberdeen Proving Ground, MD 2 10 10. MANNING, K., SKEGG, D., STELL, P., AND DOLL, R. (198 1). Cancer of the larynx and other occupational hazards of mustard gas workers. Clin. Otolaryngol. 6, 165- 170. MEDEMA, J. (1986). Mustard gas: The science of H. Nucl. Biol. Chem. Defense Tech. Int. 1,66-7 1. NAGATA, C., TAGASHIRA, Y., INOMATA, M., AND KODAMA, M. (197 1). Effect of iodine on the carcinogenicity of 3,4benzopyrene. Gann 62,309-3 14. National Council on Radiation Protection and Measurements (NCRP) (1987). Recommendations on Zonizing Radiation Exposure Limits. NCRP Report No. 9 1, NCRP, 79 10 Woodmont Ave., Bethesda, MD. National Safety Council (NSC) (1985). Accident Facts 1984. National Safety Council, Chicago, IL (as cited in NCRP 1987). NELSON, W. (198 1). Cancer prevention: Environmental, industrial and occupational factors. Cancer 47, 1065-1070. NISHIMOTO, Y., YAMAIUDO, M., SHIGENOBU, T., ONAII, K., AND YUKUTAKE, M. (1983). Long-term observation of poison gas workers with special reference to respiratory cancers. J. Uoeh 5,89-94. NISHIMOTO, Y., YAMAKIW, M., SHIGENOBU, T., YUKETAKE, M., AND MATSUSAKA, S. (1986). Cancer of the respiratory tract observed in workers having returned from a poison gas factory. Gun to Kaguku Ryoho 13,1144-l 148. NORMAN, J. ( 1975). Lung cancer mortality in World War I veterans with mustard gas injury: 19 19- 1965. J. Natl. Cancer Inst. 54,3 1 l-3 17. NORMAN, C. (1987). Regulating pesticides: The Delaney paradox. Science 236, 1054-1055. PAPIRMEISTER, B., GROSS,C., PETRALI, J., AND HIXSON, C. (1984). Pathology produced by sulfur mustard in human skin gratts on athymic nude mice. 1. Gross and light microscope changes. J Toxicol. Cutaneous Ocul. Toxicol. 3,37 l-39 1. PATAKI, J., AND HUGGINS, C. (1969). Molecular site of substituents of benz(a)anthracene related to carcinogenicity. Cancer Res. 29,506-509. RODERICKS, J., BRETT, S., AND WRENN, G. (1987). Significant Risk Decisions in Federal Regulatory Agencies. ENVIRON Corporation, Washington, DC. SARACCI, R. (198 1). The IARC monograph program on the evaluation ofthe carcinogenic risk ofchemicals to humans as a contribution to the identification of occupational carcinogens. In Banbury Report, Vol. 9-Quantification of Occupational Cancer (R. Peto and M. Schneiderman, Eds.), pp. 165-176. Cold Spring Harbor Laboratory.

CARCINOGENIC

POTENCY

OF SULFUR MUSTARD

25

SAX, N. I. (1975). Dangerous Properties of Industrial Materials, 4th ed., Van Nostrand-Reinhold, New York, NY. SAX, N. I. (1979). Dangerous Properties ofIndustrial Materials, 5th ed., Van Nostrand-Reinhold, New York, NY. SITTIG, M. (1985). Handbook of Toxic and Hazardous Chemicals and Carcinogens, 2nd ed., Noyes Publications, Park Ridge, NJ. Society of Toxicology Task Force ( 198 1). Reexamination of the ED,,, Study-Risk assessment using time. Fundam. Appl. Toxicol. 1,88-123. TATKEN, R. L., AND LEWIS, R. J., SR. (1985). Registry of Toxic of Chemical Substances, 1982-1983 ed. U.S. Department of Health and Human Services, Centers for Disease Control, National Institute for Occupational Safety and Health, Cincinnati, OH 45226. TOKUOKA, S.. HAYASHI, Y., INAI, K., EQWA, H., AOKI, Y., AKAMIZU, H., ETO, R., NISHIDA, T., OHE, K., KOBUKE, T., NAMBU, S., TAKEMOTO, T., Kou, E., NISHINA, H., FUJIHARA, M., YONEHARA, S., TSUYA, T., SUEHIRO, S., AND HORIUCHI, K. (1986). Early cancer and related lesions in the bronchial epithelium in former workers of mustard gas factory. Actu. Pathol. Japan 36,533-542. TOTH, B., AND SHUBIK, P. (1967). Carcinogenesis in AKR Mice Injected at birth with benzo(a)pyrene and dimethylnitrosamine. Cancer Res. 27,43-S 1. TRAVIS, C. C., RICHTER, S. A., CROUCH, E. A. C., WILSON, R., AND KLEMA, E. D. (1987). Cancer risk management: A review of 132 federal regulatory decisions. Environ. Sci. Technol. 21,415-420. U.S. Environmental Protection Agency (U.S. EPA) (1982a). National oil and hazardous substances contingency plan. Final rule. Fed. Regist. 47,3 1,180-3 1,202. U.S. Environmental Protection Agency (U.S. EPA) (1982b). Ethylene bisdithiocarbamates (EBDC’s); determination concluding the presumption against registration; notice of availability of decision document; notice of determination; notice of availability. Fed. Regist. 47,47,669-47,674. U.S. Environmental Protection Agency (U.S. EPA) ( 1986a). Hazardous waste management system; identification and listing of hazardous waste; notification requirements; reportable quantity adjustments. Proposed rule. Fed. Regist. S&21,648-21,685. U.S. Environmental Protection Agency (U.S. EPA) (1986b). Superfund Public Health Manual. Office of Emergency and Remedial Response Report No. EPA/540/i-86/066 (October 1986). U.S. Environmental Protection Agency (U.S. EPA) (1987a). Health Assessment Document for Beryllium. EPA/600/8-84/026F. U.S. Environmental Protection Agency (U.S. EPA) (198713). Reportable quantity adjustments. Proposed rule. Fed. Regist. 52,8 140-8 17 1. U.S. Department of Health and Human Services (U.S. DHHS) (1987). Recommendations for protecting the health and safety against potential adverse effects of long-term exposure to low doses of agents: GA, GB, VX, Mustard (H, HD, HT), and Lewisite (L). Fed. Regist. 52,48,458-48,46 1. VOGT, R., DANNENBERG, A., SCHOFIELD, B., HYNES, N., AND PAPIRMEISTER, B. (1984). Pathogenesis of skin lesions caused by sulfur mustard. Fundam. Appl. Toxicol. 4, S7 l-S83. WADA, S., NISHIMOTO, Y., MIYANSHI, M., KATSUTA, S., NISHIKI, M., YAMAD, A., TOKUOKA, S., UMISA, H., AND NAGAI, M. (1962a). Review of Okuno-jima poison gas factory regarding occupational environment. Hiroshima J. Med. Sci. 11,75-80. WADA, S., NISHIMOTO, Y., MIYANSHI, M., KATSUTA, S., NISHIKI, M., YAMADA, A., TOKUOKA, S., UMISA, H., AND NAGAI, M. (1962b). Malignant respiratory tract neoplasms related to poison gas exposure. Hiroshima J. Med. Sci. 11,8 l-9 1. WADA, S., MIYANASHI, M., NISHIMOTO, Y., KAMBE, S., AND MILLER, R. (1968). Mustard gas as a cause ofrespiratory neoplasia in man. Lancet 1, 1161-l 163. WATSON, A. P., AMBROSE, K. A., GRIFL~N, G. D., MUNRO, N. B., AND WATERS, L. C. (1988). Toxicity of Warfare Agents and Their Breakdown Products. Appendix Bin ChemicalStockpile Disposal Program Final Programmatic Environmental Impact Statement, Vol. 3, pp. B- 1 to B-2 12. Program Manager for Chemical Demilitarization, Aberdeen Proving Ground, MD 2 1010-540 1. WINDHOLZ, M., BUDAVARI, S., BLUMETTI, R. F., AND OTTERBEIN, E. S. (1983). The Merck Index: An Encyclopedia of Chemicals, Drugs and Biologicals. 10th ed., Merck and Co., Inc., Rahway, NJ. YAMAKIDO, M., SHIGENOBU, T., KAMITSUNA, A., NISHIMOTO, Y., AND YUKUTAKE, M. (1985). The causes of death in the retired workers of Okuno-jima poison gas factory. Japan J. Med. 24,394.