Regulations (Preambles to Final Rules) - Table of Contents Regulations (Preambles to Final Rules) - Table of Contents
• Record Type: Occupational Exposure to 1,3-Butadiene
• Section: 6
• Title: Section 6 - VI. Quantitative Risk Assessment

VI. Quantitative Risk Assessment

A. Introduction

In 1980, the United States Supreme Court ruled on the necessity of a risk assessment in the case of Industrial Union Department, AFL-CIO v. American Petroleum Institute, 448 U.S. (607), the "Benzene Decision." The United States Supreme Court concluded that the Occupational Safety and Health (OSH) Act requires, prior to issuance of a standard, that the new standard be based on substantial evidence in the record considered as a whole, that there is a significant risk of health impairment at existing permissible exposure limits (PELs) and that issuance of the standard will significantly reduce or eliminate that risk. The Court stated that, before the Secretary of Labor can promulgate any permanent health or safety standard, he is required to make a threshold finding that a place of employment is unsafe in the sense that significant risks are present and can be eliminated or lessened by a change in practices. (448 U.S. 642) In 1981, the Court's ruling on the OSHA's Cotton Dust Standard (American Textile Manufacturers Institute v. Donovan, 452 U.S. 490 (1981)) reaffirmed its previous position in the Benzene Decision, that a risk assessment is not only appropriate, but that OSHA is required to identify significant health risk to workers and to determine if a proposed standard will achieve a reduction in that risk, and OSHA as a matter of policy agrees that assessments should be put into quantitative terms to the extent possible.

For this rulemaking, OSHA has conducted a quantitative risk assessment to estimate the excess risk for cancer and consequently for premature deaths associated with exposure to an 8-hour time-weighted-average (TWA), 5 days/week, 50 weeks/year, 45-year exposure to BD at concentrations ranging from 0.1 to 5 ppm, the range of permissible exposure limits (PELs) considered by OSHA in this rulemaking. The data used in the quantitative risk assessment were from a National Toxicology Program (NTP) chronic inhalation study in which B(6)C(3)F(1) mice of both sexes were exposed to either ambient air or BD exposure concentrations ranging from 6.25 to 200 ppm, known as NTP II. (Ex. 90) For seven gender-tumor site combinations, multistage Weibull time-to-tumor models were fit to these NTP II data. The best fitting models were chosen via a log-likelihood ratio test.

OSHA's maximum likelihood estimate (MLE) of the excess risk of developing cancer and subsequent premature death as a result of an 8-hour TWA occupational lifetime exposure to 2 ppm BD, the PEL proposed by OSHA in 1990, was 16.2 per 1,000 workers, based on the most sensitive gender-tumor site combination, female mouse lung tumors. If the occupational lifetime 8-hour time-weighted-average (TWA) exposure level is lowered to 1 ppm BD, based on female mouse lung tumors, the estimate of excess cancer and premature death drops to 8.1 per 1,000 workers. In other words, an 8-hour TWA lifetime occupational exposure reduction from 2 ppm to 1 ppm BD would be expected to prevent, on average, 8 additional cases of cancer and probable premature deaths per 1,000 exposed workers. Based on the individual tumor site dose-response data, which were best characterized by a 1-stage Weibull time-to-tumor model, (male-lymphoma, male-lung, female-lymphoma and ovarian), on average, one would expect there to be between 1 and 6 fewer excess cases of cancer per 1,000 workers based on a 8-hour TWA occupational lifetime exposure to BD at 1 ppm versus BD at 2 ppm. Estimates of leukemia deaths at the former 8-hour TWA PEL of 1,000 ppm of BD, for an occupational lifetime, are not presented because contemporary BD exposures are generally far lower than this level.

B. Assessment of Carcinogenic Risk

1. Choice of Data Base for Quantitative Risk Assessment

The choice of data provides the platform for a quantitative risk assessment (QRA). Either animal studies which evaluate the dose-response relationship between BD exposure and tumorigenesis or epidemiological dose-response data may be suitable sources of data.

Estimates of the quantitative risks to humans can be based on the experience of animals from a chronic lifetime exposure study. Chronic lifetime inhalation bioassays with rats and mice generally last 2 years or two-thirds of the lifespan of the animal. (Ex. 114) These types of studies provide insight into the nature of the relationship between exposure concentration, duration and resulting carcinogenic response under a controlled environment. Furthermore, some researchers have estimated a variety of measures of dose of BD, including inhaled and absorbed dose as well as BD metabolites, to estimate human risks based on the observed dose-response relationship of animals in a bioassay; the form of the dose used in a dose-response analyses is called the dose-metric.

The carcinogenicity of lifetime inhalation of BD was studied in Sprague-Dawley rats by the International Institute of Synthetic Rubber Producers (IISRP) and in B(6)C(3)F(1) mice by the National Toxicology Program. The IISRP sponsored a two-year inhalation bioassay of Sprague-Dawley rats performed at Hazelton Laboratories Europe (HLE). (Ex. 2-31) Groups of 110 male and female Sprague-Dawley rats were exposed for 6-hours per day, 5 days per week to 0, 1,000, or 8,000 parts per million (ppm) of BD. The males were exposed for 111 weeks and the females for 105 weeks. Statistically significant increased rates of tumors were found in both male and female rats. Among exposed male rats, there were increased occurrences of pancreatic and testicular tumors and among the exposed female rats there were higher incidence rates of uterine, zymbal gland, mammary and thyroid tumors than in the control groups.

The National Toxicology Program (NTP) has performed two chronic inhalation bioassays using B(6)C(3)F(1) mice. (Ex. 23-1; 90; 96) The first study, NTP I, was intended to be a two-year bioassay, exposing groups of 50 male and female mice to 0, 625, or 1,250 ppm of BD for a 6-hour day, 5 days/week. The study was prematurely curtailed at 60 weeks for the males 61 weeks for the females caused by an unusually high cancer mortality rate due to malignant neoplasms in multiple organs. Despite some weaknesses in the way the study was conducted, the results of this study show that BD is clearly carcinogenic in these mice, with statistically significant increases in malignant lymphomas, heart hemangiosarcomas, lung tumors, and forestomach tumors in comparison to the controls for exposed male and female mice. (Ex. 90) The second NTP BD chronic inhalation bioassay, NTP II, had groups of 70 (except for the group exposed to the highest concentration, which contained 90) male and female mice exposed to concentrations of 0, 6.25, 20, 62.5, 200 and 625 ppm for 6 hours/day, 5 days/week for up to 104 weeks. The NTP II bioassay provided lower exposures, closer to prevailing occupational exposure levels, than the NTP I and HLE chronic inhalation studies. The NTP II supported the pattern of carcinogenic response found in NTP I. Both male and female mice exposed to BD developed tumors at multiple sites including: lymphomas, heart hemangiosarcomas, and tumors of the lung, liver, forestomach, and Harderian gland (an accessory lacrimal gland at the inner corner of the eye in animals; they are rudimentary in man). Reproductive tissues were also adversely affected. Among the exposed males there were significant increases in tumors of the preputial gland; among females there were significant increases in the incidence of ovarian and mammary tumors.

In 1996, a retrospective cohort study by Delzell and co-workers of about 18,000 men who worked in North American synthetic rubber plants was submitted to OSHA. (Ex. 117-1) In this study researchers derived estimates of occupational exposure to BD using a variety of resources, such as work histories, engineering data, production notes, and employees' institutional memories. In their October 2, 1995 report Dr. Delzell et al., characterized their effort as follows:

Retrospective quantitative exposure estimation was done to increase the power of the study to detect associations and to assist with the assessment of the impact of specific exposure levels on mortality from leukemia and other lymphopoietic cancers. (Ex. 117-1)

In April 1996, Dr. Delzell expressed concern with possible discrepancies between estimated cumulative exposures and actual measurements. (Ex. 118-2) OSHA believes that in a well-conducted study, retrospective exposure estimates can be reasonable surrogates for true exposures; misclassifications or uncertainty can decrease the precision of the risk estimates derived from such a study, but the problem must be severe and widespread to invalidate the basic findings.

At the time of publication of the proposed standard on occupational exposure to BD (August 1990), only the NTP I mouse and HLE rat bioassays were available for quantitative risk assessments (QRA). Presented in Table V-9 is an overview of authorship and data sets used in the various QRAs submitted to the OSHA docket. With one exception, the rest of the QRA's in the BD Docket have relied on animal chronic exposure lifetime bioassays. Each of the five risk assessments discussed in the proposal based its quantitative risk assessment on one or both of the higher-exposure chronic bioassays (exposure groups exposed to BD concentrations ranging between 625-8,000 ppm). (Exs. 17-5; 17-21; 23-19; 28-14; 29-3; 32-27) The three QRAs conducted using bioassay data subsequent to the publication of the NTP II study used NTP II data with exposures of 6.25-625 ppm BD, closer to actual occupational exposures, for calculating their best estimates of risk. (Exs. 90; 118-1b; 32-16) A summary of each of the ten QRA's follows:


TABLE V-9.--SUMMARY TABLE OF QUANTITATIVE RISK ASSESSMENTS (QRAs) IN
                 ORDER OF THEIR REVIEW IN THE OSHA BD STANDARD
_____________________________________________________________________
  Exhibit  |                        Author                           |
___________|_________________________________________________________|
90.........|National Institute for Occupational Safety and Health    |
           |  (NIOSH)(Preliminary)...................................|
118-1b.....|NIOSH....................................................|
118-1......|NIOSH....................................................|
17-21......|United States EPA Carcinogen Assessment Group (CAG)......|
32-27......|California Occupational Health Program (COHP) of the     |
           |  California Department of Health services (CDHS)........|
32-16......|Shell Oil Corporation....................................|
17-5.......|United States EPA Office of Toxic Substances (OTS).......|
23-19......|ICF/Clement Inc..........................................|
29-3.......|Center for Technology, Policy, and Industrial Development|
           |  at the Massachusetts Institute of Technology...........|
28-14......|Environ Inc..............................................|
___________|_________________________________________________________|



_____________________________________________________________________
  Exhibit  |                        Data-set                         |
___________|_________________________________________________________|
90.........|NTP II(a) bioassay (preliminary).                        |
118-1b.....|NTP II bioassay.                                         |
118-1......|Delzell et al. epidermiological study.                   |
17-21......|NTP I(b) and HLE(c) bioassays; Epidemiological based on  |
           |  Fajen Exposure Data.                                   |
32-27......|NTP I; HLE bioassays Epidemiological based on Fajen      |
           |  Exposure Data.                                         |
32-16......|NTP I, NTP II and HLE bioassays.                         |
17-5.......|NTP I bioassay.                                          |
23-19......|NTP I bioassay.                                          |
29-3.......|NTP I and HLE bioassays.                                 |
28-14......|HLE bioassay.                                            |
___________|_________________________________________________________|
  Footnote(a) NTP II, The National Toxicology Program, Technical
Report 434, 2-year bioassay of B(6)C(3)F(1) mice to 5 exposure groups
receiving between 6.25 and 625 parts per million (ppm) of BD
  Footnote(b) NTP I, The National Toxicology Program, prematurely
terminated longtime bioassay of B(6)C(3)F(1) mice to 2 exposure groups
receiving either 625 or 1,200 ppm of BD
  Footnote(c) HLE, Hazelton Laboratories Europe's, lifetime bioassay
of Sprague Dawley rats, exposed groups received 1,000 ppm of BD or
8,000 ppm of BD

NIOSH-Quantitative Risk Assessments based on NTP II

In the early 1990's, two QRAs were conducted sequentially by the National Institutes for Occupational Safety and Health (NIOSH). One was a preliminary and the other a final, with the latter using final pathology data for histiocytic sarcomas and one particular type of lymphoma from NTP II. In 1991, NIOSH submitted a preliminary QRA using the then preliminary NTP II tumor pathology data for various individual organ sites (8 from the female mice and 6 from the male mice) to estimate excess cancer risk at different BD exposures over an occupational lifetime. (Ex. 90) For all gender-tumor site analyses, NIOSH excluded the 625 ppm exposure group in its best estimate of risk since the plethora of competing tumors(6) in this high exposure group provide less information for a dose-response analysis of individual tumor sites than do data from some of the lower exposure groups. Another reason for the exclusion was that the dose-time-response relationship in mice is saturated for exposures above 500 ppm and the data would thus provide very little additional information for low dose extrapolation. NIOSH's QRA relied on an allometric conversion of body weight to the three-quarters power, (mg/kg)(3/4), and equated a 900-day-old mouse to a 74-year old human. To avoid duplication of risks, NIOSH presented only maximum likelihood estimates based on the aggregate of all types of lymphomas even though dose-response data were also available for the lymphocytic lymphoma subset.

__________

Footnote(6) Competing tumors refers to the lack of opportunity of a later developing tumor to express itself due to the occurrence of early developing lethal tumor; Among the 625 ppm exposure group lymphocytic lymphomas were mortal early developing tumors which prevented later developing disease such as heart hemangiosarcomas from possibly developing.

Of the fourteen gender-tumor site data sets NIOSH modeled to extrapolate animal data to humans, 12 (86%) yielded excess risks greater than 2 cancer deaths per 1,000 workers, given an 8-hour TWA lifetime occupational exposure of 1 ppm BD. Estimates of excess risks to workers based on the best fitting models for each of the six dose-time-response relationships for male tumor sites were between 0.4 and 15.0 per 1,000 workers assuming an 8-hour TWA, 45 year occupational exposure to 1 ppm BD. Among estimates based on male mice's dose-response data, the lowest and highest excess risk estimates were from the heart hemangiosarcoma and Harderian gland dose-response relationships, respectively. For estimates of excess risk based on either gender's set of individual tumor dose-response relationships, only the heart hemangiosarcoma data predicted a risk of less than 1 per 1,000 workers with an occupational lifetime exposure of 1 ppm: these data predicted 0.4 and 3 x 10(-3) excess cancer cases per 1,000 workers based on the best fitting models for male and female mice, respectively.

Based on tissue sites in females, the excess risk estimates for 8-hour TWA occupational lifetime exposure to 1 ppm BD range between 4 and 31 per 1,000 workers.

NIOSH presented its findings for lifetime exposure to 2 ppm as follows:

Based on tumors at the most sensitive site, the female mouse lung [assuming (mg/kg)(3/4) conversion], our maximum likelihood estimates of the projected human increased risk of cancer due to a lifetime occupational exposure to BD at a TWA PEL of 2 ppm is approximately 60 in 1,000 (workers). (Ex. 90)

For the linear models, if scaling were on a (mg/kg) basis rather than the (mg/kg)(3/4) used by NIOSH for allometric conversion, the revised estimate of excess cancer risk for an 8-hour TWA occupational lifetime exposure to 2 ppm BD would decrease approximately 6 fold to 9.2 per 1,000 workers based on the same female mouse lung tumor data.

In 1993, NIOSH finalized its estimates of excess risk caused by occupational exposure based on the tumorigenesis experience of mice in the NTP II study. (Ex. 118-1B) The rounded maximum likelihood estimates (MLE) from the final QRA are presented in Table V-10. NIOSH expanded the gender-tumor sites to include histiocytic sarcoma for both male and female mice. NIOSH chose to present only its risk estimate based on lymphocytic lymphoma, rather than an assessment based on the aggregate of lymphomas. In the preliminary and final NIOSH QRAs, 1-stage time-to-tumor models" rounded estimates of risk associated with lifetime exposure to 1 ppm BD ranged from 1 to 30 excess cancer cases per 1,000 workers, with estimates based on the male-lymphocytic lymphoma and the female-lung dose-response data providing the lower and upper ends of the range of risk, respectively.

As part of its sensitivity analyses, NIOSH derived the estimates of risk based on (1) equating a human lifespan to a mouse equivalent age of 784 days, a figure OSHA has used, and (2) equating a human lifespan to a mouse lifespan of 900 days (a figure more often used by NIOSH.) The best estimates of risk equating human lifespan to a mouse lifespan of 784 days were lower, by about one-third, than those assuming a human lifespan equivalency to 900 days for the mouse, all else held constant.


TABLE V-10.--NIOSH'S(a) FINAL QUANTITATIVE RISK ASSESSMENT'S (QRA)
    MAXIMUM LIKELIHOOD ESTIMATES (M.L.E.S)(b) PER 1,000 WORKERS OF
    LIFETIME EXCESS RISK DUE TO AN OCCUPATIONAL(c) EXPOSURE TO 1 PPM
    OF BD USING BEST FITTING MODELS, AS DESIGNATED BY NUMBER OF STAGES
    OF THE WEIBULL TIME-TO-TUMOR MODEL
_____________________________________________________________________
                 Gender-tumor site                  | MLE, Final QRA |
                                                    |     (Stages)   |
____________________________________________________|________________|
Male mouse:                                         | .............. |
   Forestomach......................................|        0.03(2) |
   Harderian gland..................................|          10(1) |
   Heart hemangiosarcoma............................|         0.5(2) |
   Histiocytic sarcoma..............................|           8(1) |
   Liver............................................|           4(1) |
   All Lymhoma......................................|            NA  |
   Lympocytic lymphoma..............................|         0.9(1) |
   Lung.............................................|          10(1) |
Female mouse:                                       | .............. |
   Forestomach......................................|           5(1) |
   Harderian Gland..................................|           7(1) |
   Heart hemangiosarcoma............................|    3x10(-3)(3) |
   Histiocytic sarcoma..............................|          10(1) |
   Liver............................................|           7(1) |
   All lymhoma......................................|            NA  |
   Lympocytic lymphoma..............................|           9(1) |
   Lung.............................................|          30(1) |
   Mammary..........................................|           4(1) |
   Ovarian..........................................|           9(1) |
____________________________________________________|________________|
  Footnote(a) Based on NTP II, excluding the 625 ppm exposure
category, equating a 900-day-old mouse to a 74-year old human and
assuming an allometric conversion of (mg/kg)(3/4).
  Footnote(b) Rounded to one significant figure.
  Footnote(c) Occupational lifetime is an 8-hour
time-weighted-average, 40-hours per week, 50-weeks per year,
time-weighted-average (TWA) for 45-years.


The Carcinogen Assessment Group QRA

The Carcinogen Assessment Group (CAG) and the Reproductive Effects Assessment Group of the Office of Health and Environmental Assessment at the United States Environmental Protection Agency (EPA) also conducted an assessment of the mutagenicity and carcinogenicity of BD. (Ex. 17-21) In its quantitative risk assessment, CAG used both male and female response data from the two chronic bioassays available at the time, NTP I with B(6)C(3)F(1) mice and the HLE Sprague Dawley rat study. The CAG analysis is based on EPA's established procedures for quantitative risk analyses, which fit the total number of animals with significantly increased or highly unusual tumors with the linearized multistage model and use the upper 95% confidence interval. Mice dying before week 20 and rats dying during the first year of the study (before the observation of the first tumor) were eliminated from the analysis to adjust for non-tumor differential mortality.

The dose-metric was based on a preliminary report by the Lovelace Inhalation Toxicology Research Institute of its six-hour exposure study in B(6)C(3)F(1) mice and Sprague Dawley rats at different concentrations of BD, roughly corresponding to the concentrations used in NTP I and HLE, with total internal BD equivalent dose expressed as a function of inhalation exposure concentration. Then CAG estimated the amount and percent of BD retained for various exposure concentrations in these bioassays. These internal dose-estimates were then extrapolated to humans based on animal-to-human ppm air concentration equivalence.

CAG adjusted risk estimates from the mouse study by a factor of (study duration/lifetime)(3) to account for less-than-lifetime observations, since the NTP I study was prematurely terminated at 60 weeks for males and 61 weeks for females due to predominating cancer mortality. CAG extrapolated the short lifespan mouse data to an expected mouse lifetime, 104 weeks, in order to estimate lifetime risk to humans.

CAG estimated all risks based on continuous exposure to BD, 24 hours per day, 365 days per year, for a 70-year lifetime. The incremental unit risk estimates for the female mouse were about eight times as high as those for the female rat; for the males, the incremental unit risk estimate for mice was about 200 times as high as for rats. The CAG final incremental unit risk estimate of 0.64 (ppm)(-1) is based on the geometric mean of the upper-limit slope estimates for male and female mice and would predict an upper limit of 640 excess cancers per 1,000 people exposed to 1 ppm continuously throughout their lifetime, 70 years. Extrapolating this same estimate to an equivalent 45-year working lifetime of 240 work days per year at an 8-hour TWA exposure to 1 ppm BD would yield an upper-limit risk estimate of 90 excess cancers per 1,000 workers. If the working day is assumed to require one-half (10m(3)) the daily tidal volume, the total amount of air inhale, the excess would be 135 cancers per 1,000 workers.

California Occupational Health Program (COHP) QRA

In 1990, five years after the CAG conducted its quantitative risk assessment, the California Occupational Health Program (COHP) produced its estimates of risk with a similar assessment of the carcinogenicity of BD, using the same available bioassays, with more recent information on BD risk in humans, pharmacokinetic (PK) modeling, and animal low exposure absorption efficiency. (Ex. 32-16) Using three separate dose-metrics for each bioassay and multistage models to characterize the basic dose-response relationship, CAG presented several quantitative estimates of incremental lifetime unit risks. Quantal lifetime response multistage models were fit to the data. COHP, like NIOSH, used the individual data with a multistage Weibull time-to-tumor model to characterize the dose response relationship. COHP stated that it also fit Mantel-Bryan and log-normal models to the data, and that the multistage models gave a better fit; the results obtained with these other models were not reported.

COHP performed calculations on each primary tumor site separately, and also did calculations on the pool of primary tumors that showed significantly increased tumor incidences. For their main dose-metric, COHP refined the CAG approach, using a revised estimate of low-exposure absorption via inhalation. COHP also included an estimate of the PK model derived BD monoepoxide metabolites, but de-emphasized their use by stating that these were "presented for comparative purposes only." The third dose-metric was straight ppm for animal-to-human species conversion (adjusting for duration of exposure). COHP stated:

(COHP) followed standard EPA practice and assumed that a certain exposure concentration in ppm or mg/m(3) in experimental animals was equivalent to the same exposure concentration in humans. (Ex. 32-16)

Like CAG, COHP also adjusted for less than lifetime survival in the NTP I mouse study, by using a cubic power of time, (study duration/lifetime)(3). COHP's potency estimate adjustment for the male mouse study with 60-week survival was 5.21; for the 61-week female mouse survival the adjustment was 4.96.

With all the combinations of sites, species, sexes, models, and dose-metrics, COHP presented over 60 potency estimates for the rat and over 100 for the mouse. As with the CAG and other analyses, the estimates based on NTP I were typically one to two orders of magnitude greater than those based on the rat for similar dose-metrics, models and total tumors. COHP chose the estimates based on the male mouse as final indicators of human risk based on the "superior quality of the mouse study." From these estimates, using the quantal form of the multistage model, COHP chose "the upper bound for plausible excess cancer risk to humans." COHP's final cancer potency estimate of 0.32 (ppm)(-1) presented in units of continuous lifetime exposure, is based on all significant tumors in the male mouse and uses the internal BD equivalent dose conversion factor of 0.54 mg/kg-d/ppm for the mouse and animal-to-human ppm equivalency. COHP's final potency estimate was one-half the value of 0.64 (ppm)(-1) calculated by the CAG; the difference is due mainly to a low exposure absorption modification by COHP. The continuous lifetime exposure potency factor converts to a working lifetime risk of 45 to 67 excess cancers per 1,000 workers, exposed to 1 ppm of BD at an 8-hour TWA over a 45 year working lifetime.

COHP, like CAG, attempted to determine whether its animal-based risk extrapolation could predict the leukemia mortality observed in epidemiology studies. Following the approach employed by CAG in its analyses of the Meinhardt (1982) study, the COHP compared its estimates of risk from bioassays to the then most recent epidemiological studies of Downs et al. (1987) and Matanoski and Schwartz (1987). Both COHP and CAG used MLEs based on mouse lymphoma for comparing the animal-derived potency estimates with the occupational response. In addition, neither COHP nor CAG used the upward adjustment factor of approximately 5 to correct for the less-than-lifetime duration of NTP I. Because neither of these epidemiology studies (Downs et al. (1987) or Matanoski and Schwartz (1987)) had recorded exposure estimates, the COHP relied on 8-hr TWA estimates of 1 and 10 ppm taken at different but similar plants reported by Fajen et al. (1986). For lifetime unit risk estimates, COHP used the initial MLE of 0.0168 (ppm)(-1) derived from the male mouse lymphoma analysis, unadjusted for less-than-lifetime survival. This part of the analysis also assumed that a lymphocytic outcome in the animals would equate to leukemia death in humans. These assumptions yielded a range of 6 to 21 predicted lymphocytic cancer deaths (for 1 and 10 ppm exposures) versus the 8 observed by Downs et al.

Office of Toxic Substances (OTS) QRA

The Office of Toxic Substances (OTS), U.S. Environmental Protection Agency (EPA) conducted a quantitative risk assessment using only the NTP I data. (Ex. 17-5) The reasons cited for this choice include: (1) The mouse is a more sensitive test species for BD than the rat; (2) a quality control review had been done for the mouse bioassay at the time OTS wrote its risk assessment whereas none was available for the rat bioassay; (3) greater amount of histopathological data was available for the NTP I study than for the HLE rat study; and (4) the type of BD feedstock used by NTP I had a much lower dimer concentration than the BD used by HLE (increased dimer concentration results in the lowering of availability of BD for metabolism to the mono- and di-epoxides, which are thought to be the carcinogenic agents). To compensate for early termination of the NTP I study, OTS adjusted dose by a factor of (study duration/lifetime).(3) Butadiene ppm exposure concentration was used as the measure of dose and mouse-to-human species extrapolation was also on a ppm equivalence basis. OTS estimated cancer risks based on heart hemangiosarcoma and pooled tumors (grouping of sites showing statistically significant elevated incidence rates) tumors using a 1-stage quantal model. Workplace exposures to BD were converted to estimated lifetime average daily doses. Since the NTP I study was curtailed at 61 weeks, tumor incidence rates were adjusted for survival by life-table methods. Cancer risks were based on administered dose of BD and not delivered dose to various target organs. (Ex. 17-5) Estimated 95% upper confidence-limits for the excess risk of cancer from an occupational lifetime exposure to an 8-hour TWA of 1 ppm BD, for 240 days/year for 40 years, ranged between 10 and 30 per 1,000 workers, based on pooled tumor incidence for female and male animals, respectively.

ICF/Clement Estimates

In 1986, ICF/Clement (ICF) estimated the risk of cancer associated with occupational exposure to BD. (Ex. 23-19) ICF determined that only the NTP I data were suitable for a risk assessment based on animal data, (NTP II data were not available at that time) based on ICF/Clement's concern over the discrepancies between HLE's summary statistics and individual counts. ICF chose to use individual tumor type data for some of its analyses. ICF fitted a linearized multistage quantal model to the NTP I data. Based on a preliminary study by Bond (a senior toxicologist at the Chemical Industry Institute of Toxicology), ICF adjusted the NTP I exposure concentrations for percent retention which varied inversely from 100% at 1 ppm to 5% at 1,000 ppm.

ICF assumed ppm as the proper dose-metric and ppm to ppm for the mouse-to-human species extrapolation factor. (Exs. 23-86; 23-19) The 95% upper confidence limit estimates of risk based on pooled female tumor data with a lifetime occupational exposure was 200 per 1,000 workers at 1 ppm BD, and 400 per 1,000 workers at 5 ppm BD; the non-proportionality reflects the assumption of lower percentage retentions at higher concentrations.

Massachusetts Institute of Technology (MIT) QRA

Hattis and Wasson at the Center for Technology, Policy, and Industrial Development at MIT conducted pharmacokinetic/mechanism-based analyses of the carcinogenic risk associated with BD. (Ex. 29-3) The analyses include both HLE and NTP I data. Key elements, such as partition coefficients for blood/air and tissue/blood, were not available to be measured and had to be estimated. The best estimate of excess risk of cancer given a lifetime occupational exposure of 1 ppm BD 8-hr TWA was 5 per 1,000 workers based on the NTP I female mouse data set, incorporating pharmacokinetic models which set the blood/air partition coefficient to 0.2552. Based on the HLE female rat data with a blood/air partition coefficient of 0.2552, an excess risk was estimated to be 0.4 additional cases of cancer for every 1,000 workers at an 8-hour TWA, occupational lifetime exposure to 1 ppm BD.

Environ QRA

Environ conducted a quantitative risk assessment based on the HLE rat bioassay data. (Ex. 28-14) Environ noted that the relatively high BD concentrations of the earlier bioassays (HLE with groups exposed to 8,000 and 1,000 ppm BD and NTP I with exposures of 1,250 and 625 ppm BD) made it difficult to extrapolate risks to the relevant, lower exposure levels of BD in occupational settings. Environ stated that among B(6)C(3)F(1) mice, metabolic saturation occurs with 8-hour TWA BD concentrations greater than 500 ppm; thus, the time-dose-response relationship is different at higher doses than at lower doses. Environ stated that the methodological problems and the high early mortality shown in the NTP I data contributed to the uncertainty of its relevance to human risks and therefore chose to use the HLE rat bioassay data instead. Environ believes that human metabolism of BD is more similar to that in the Sprague-Dawley rat than in the B(6)C(3)F(1) mouse. Extrapolated risks were based on estimates of absorbed dose, expressed in mg/kg, as defined in the Bond et al. (1986) absorption study. (Ex. 23-86) Environ used the HLE female rats to estimate the extra lifetime risk of developing cancer given an occupational lifetime 8-hr TWA exposure to 1 ppm BD. Using MLEs from multistage, Weibull, and Mantel-Bryan models, based on the total number of female rats with significantly increased tumors, Environ's predicted occupational lifetime risks were 0.575 (Multistage), 0.576 (Weibull), and 0.277 (Mantel-Bryan) per 1,000 workers.

Shell Oil Company QRA

Shell Oil Company estimated excess cancer risks by the multistage quantal and the Weibull time-to-tumor models based on female heart hemangiosarcomas and pooled malignant tumors from the NTP II study. Shell estimated human risks based on various assumptions, correcting for BD retention and/or relative human epoxide dose. Shell stated that the Weibull time-to-tumor model better characterized risks since it was able to fully utilize available dose-response data, including time until onset of tumors and latency (time from initiation until detection of tumor). (Ex. 32-27) Shell used

* * * crude time-to-tumor data consisting of early deaths to 40-weeks, 40-week interim sacrifices, deaths to 65-weeks, 65-week interim sacrifices, death to 104-weeks and terminal sacrifices * * * in-lieu of individual animal data [for NTP II data]. (Ex. 32-27)

OSHA believes that the true dose-response relationship is obscured by Shell's use of crude time-to-tumor data and its grouping of early deaths to 40 weeks, deaths to 65 weeks and deaths to 104 weeks; instead, dose-time-tumor response data for each individual mouse should have been used.

Shell did not explain why it chose one model over the other. For example, without explanation, Shell dropped the highest exposure group, 625 ppm, when estimating lifetime occupational risk for all of its Weibull time-to-tumor models and dropped additional dose groups when using some multistage quantal models. Moreover, estimates of excess risk were presented only for 5-stage Weibull time-to-tumor models, although there is no discussion of correct model specifications. For example, no reasons are given for choosing the 5-stage model rather than another. Also, Shell does not support its estimation that the latency between the induction of a tumor and its observation is for the pooled female mice malignant tumors and 40-weeks for the female mice heart hemangiosarcomas.

Based on the Shell analyses, extrapolating from pooled malignant female mice tumors, assuming 10% human BD retention efficiency at 2 ppm, and on a 5-stage Weibull time-to-tumor model, one would expect 18 excess cancers per 1,000 workers given an 8-hour TWA occupational lifetime exposure of 2 ppm BD. Based on the same data set, but assuming a mouse-to-human species conversion factor based on an epoxide ratio of 590 (mouse-to-monkey) in addition to a 10% BD retention efficiency factor, the estimate of excess risk of cancer drops to 0.3 cases per 1,000 workers with an 8-hour TWA occupational lifetime exposure of 2 ppm. Using the same pooled malignant female mice tumors, but assuming the blood epoxide estimates of the Dahl et al. study and an 8-hour TWA lifetime occupational BD exposure of 2 ppm, the estimate of excess risk of cancer is slightly lower, 0.24 per 1,000 workers. The excess risk estimates based on female hemangiosarcomas and a 5-stage Weibull time-to-tumor model and occupational lifetime exposure to 2 ppm of BD were: (a) 6.4 x 10(-8) (assuming a 10% BD retention factor); (B) 6.2 x 10(-15) (assuming a 10% BD retention factor and an epoxide ration of 590); and (c) 1.3 x 10(-11) (assuming the blood epoxide estimates of the Dahl et al. study).

Shell also presented the Environ Inc. QRA based on the HLE Sprague-Dawley rat bioassay and made similar adjustments for BD retention and blood epoxide to those it made for the NTP II B(6)C(3)F(1) mice data. As had Environ, Shell stated that the dose-response of the rat is more relevant than that of the mice in predicting risk in humans. Shell concluded that the risk estimates derived from HLE Sprague Dawley rat data should be given greater weight than those based on the B(6)C(3)F(1) mouse data.

NIOSH's QRA Based on the Delzell et al. Study

NIOSH estimated the excess risk of workers developing leukemia based on the Delzell et al. preliminary estimates of occupational exposure categories of a retrospective cohort study. (Exs. 117-1; 118-1) NIOSH derived excess risks from the best fitting relative risk (RR) model, the square root model, as fit by Delzell et al. who adjusted for age, years since hire, and calendar period. The preferred final model specified by Delzell et al. was:

Relative Risk=1+0.17 x (BD ppm-years)(0.5)

Under this model the age-cause specific leukemia death rates (ACSDR) are a function of cumulative occupational exposure up to that age. The occupational ACSDRs are a multiplicative function of background ACSDR times the BD-caused relative increase (0.17 * BD ppm-years) in leukemia. These total ACSDRs were then applied to an actuarial program which adjusted for competing risks to estimate lifetime excess risk of leukemia associated with 45-year 8-hour TWA occupational exposures for a number of PELs for BD. Estimates of background rates of leukemia and all causes of death were taken from the mortality rates for all males, 20 to 65 years of age, from the 1989 Vital Statistics of the United States. This model estimates the excess risk of leukemia death, given an occupational lifetime exposure of 2 ppm of BD, as 11 per 1,000 workers. Lowering the 8-hour TWA occupational lifetime BD PEL to 1 ppm, on average, one would expect there to be 8 excess leukemia deaths per 1,000 workers over a working lifetime.

In most animal bioassays, exposure to chemical carcinogens is usually associated with an elevated tumor incidence at only one or two target tissues. BD is of great concern because significantly increased incidences of tumors at multiple sites and doses were observed in both rats and mice.

OSHA's final risk assessment is based upon the NTP II bioassay. (Exs. 90; 96) In NTP II, the following tumor sites' incidence rates were elevated: Heart, lymph nodes, lung, forestomach, Harderian gland, preputial gland, liver, ovaries and mammary gland. The NTP II bioassay was preferred over the NTP I mouse and the HLE rat bioassay for several reasons. First, most of the exposure levels for NTP II (6.25, 20, 62.5 and 200 ppm) were closer to current occupational exposure levels than were those in the other bioassays (625; 1,000 and 8,000 ppm); studies with higher than typical occupational exposure concentrations may lead to difficulties in extrapolating the effects to the lower concentrations of BD which typically occur in current occupational settings. Furthermore, for doses (625 to 8,000 ppm) above the metabolic saturation level of 500 ppm, the biologically effective doses are not proportional to ppm exposure concentrations. Second, the NTP II mice were successfully randomized to exposure groups and their individual pathology reports were consistently coded. The randomization of the bioassay mouse population lends to the internal validity of the study through the similar composition of experimental and control groups. Third, Good Laboratory Practices were followed, as verified by audits. Fourth, there was a clear dose-response relationship for several cancer sites. Fifth, since the carcinogenic mechanism is still unknown, OSHA conservatively estimates excess risk to humans based on the experience of the more sensitive animal species unless there is specific evidence indicating that the choice of that species is inappropriate. Sixth, risk assessment results based on the preliminary findings from the most recent epidemiologic study suggest that the B(6)C(3)F(1) mouse is a reasonable species to use for quantitative risk assessment. (Ex. 118-1) For its risk assessment, OSHA has focused exclusively on those tumor sites that are scientifically pertinent. From the NTP II study, the range of excess cancer risk associated with a lifetime occupational exposure to BD is estimated based on the dose-response relationships of four target tissues, three common to both genders: Heart (hemangiosarcoma), lung, and lymphoma, and one, ovarian tumors, observed in one gender only. OSHA's focus on these four individual target tissues is based not on an objection to the use of other tissue tumors and sites but rather on the judgment that the chosen animal sites are appropriate because they include both rare (e.g., heart hemangiosarcoma) and common tumors (e.g., lung) and those sites with the lowest (heart hemangiosarcoma) and highest incidence rates (lymphatic).

Three of the target organs chosen for the QRA demonstrated a significantly elevated tumor incidence in both male and female animals; ovarian tumor incidence was also significantly elevated in female animals. For both male and female mice, heart hemangiosarcomas were selected for modeling because there is virtually no background incidence of heart hemangiosarcoma among untreated mice in the NTP control population; only 0.04% of unexposed B(6)C(3)F(1) mice develop heart hemangiosarcoma, and thus any observed increase in the incidence of heart hemangiosarcoma could be attributed to BD exposure. (Ex. 114, p. 121) The earlier developing lymphocytic lymphoma caused a significant number of mice to die. Therefore, leaving mice are left at risk for the later developing tumor, heart hemangiosarcoma. (Ex. 114, p. 123) This situation is known as competing risk (the lack of opportunity for later developing tumors to express themselves because an earlier developing tumor has already caused the death of the animal. The occurrence of heart hemangiosarcomas in the NTP study is even more notable because of these competing risks.

In the absence of definitive, pharmacokinetic information, OSHA has estimated excess risks to humans based on the most sensitive species-sex-tumor site. Lung tumors are the most sensitive sites for both male and female B(6)C(3)F(1) mice and, as such, were included in OSHA's final risk assessment.

Ovarian tumors are an example of the group of reproductive tumors which also had significantly increased incidence rates among the animals in the NTP II bioassay. Other significantly increased incidence rates were seen in testicular, preputial and mammary tumors.

The increased risk of developing leukemia that has been observed in the epidemiological studies suggests that lymphomas might be the most relevant tumor site in animals for estimating the quantitative cancer risk to workers. Some have suggested that the high rate of lymphoma among B(6)C(3)F(1) mice might have been due to the presence of the murine retro virus (MuLV) and have asserted that the presence of this virus in B(6)C(3)F(1) mice may be partially responsible for the incidence of thymic lymphoma. For example, in 1990, Dr. Richard Irons reported,

A major difference between NIH Swiss and B(6)C(3)F(1) mice is their respective exotropic retro viral background (MuLV) * * * Chronic exposure to BD (at 1250 ppm) for up to a year resulted in a fourfold difference in the incidence of thymic lymphoma between B(6)C(3)F(1) mice and NIH Swiss mice * * * The role of endogenous retro virus (MuLV) in the etiology of chemically induced murine leukemogenesis is presently not understood. (Ex. 23-104)

Dr. Melnick of the National Toxicology Program testified during his public hearing statement,

In terms of the difference in response between the B(6)C(3)F(1) mouse or the NIH Swiss Mouse, you must be aware that the study is not a complete cancer study. It's a one-year exposure. We do not know the full response in the NIH Swiss mouse if it were conducted as a cancer study (about 2-years). (Tr. 1/16/91, p. 382)

Furthermore, NIOSH stated: "It is not known whether the retro virus activation mechanism is operative at the lower exposure concentrations of 1,3-butadiene [below 1250 ppm]." (Ex. 90) There is no information in the record to show that retrovirus insertion into the B(6)C(3)F(1) mice of the NTP II study led to the induction of lymphoma. Nor is there information indicating that the murine retro virus may have led to an enhancement of butadiene-induced lymphomas in B(6)C(3)F(1) mice. The development of thymic lymphoma in BD-exposed NIH Swiss mice that do not have this endogenous virus argues against the virus alone inducing the lymphomas observed in the BD-exposed B(6)C(3)F(1) mice. (Ex. 23-104) Tables V-11 and V-12 show the breakdown of microscopically examined tissues included in OSHA's QRA, by exposure concentration and death disposition of female and male mice. As illustrated in the tables, microscopic examination varied by tissue type, exposure group, means of death, and gender. Microscopic examinations of all tissues were made for all natural deaths, and moribund and terminal sacrifices, irrespective of exposure group.

For each gender-exposure-group, 10 animals were sacrificed at 40 and 65 weeks. Microscopic evaluations were not made for all tissue types among interim sacrifices (40 and 65 weeks). Among early sacrifices (40 weeks) for the 6.25 and 20 ppm exposure groups, there were no microscopic examinations of the relevant tissues. For the 65-week female sacrifices at the 6.25 and 20 ppm dose levels only lung and ovarian tissues were examined microscopically. No microscopic evaluations were made for male 65-week sacrifices at the 6.25 ppm exposure level, but at the 20 ppm exposure level, animals were microscopically examined for heart hemangiosarcoma and lung cancer. Male and female interim sacrifices exposed to 62.5 ppm of BD were not microscopically examined for heart hemangiosarcoma.

Only observations confirmed by microscopic examination were included in the analyses. Among natural deaths for some gender-tissue combinations, there were a few animals for which tissues were not available. Tissue unavailability was due to autolysis (cell destruction post death) and missing tissues due to the delay between accident and discovery.


     TABLE V-11.--TYPES OF TISSUES MICROSCOPICALLY EXAMINED BY
                  CONCENTRATION DOSE AND DISPOSITION BROUPS AMONG
                  FEMALE MICE FROM NTP(a)
____________________________________________________________________
Concentration  |  Natural death and moribund  |                     |
    ppm        |          sacrifice           |   Week 40 sacrifice |
_______________|______________________________|_____________________|
0..............|lymphoma, heart(b), lung,     |lymphoma, heart, lung|
               |   ovaries                    |  overies.           |
6.25...........|lymphoma, heart, lung,        |none(c)..............|
               |   ovaries                    |                     |
20.............|lymphoma, heart, lung,        |none.................|
               |   ovaries                    |                     |
62.5...........|lymphoma, heart, lung,        |lymphoma, lung,      |
               |   ovaries                    |  overies............|
200............|lymphoma, heart, lung,        |lymphoma, heart, lung|
               |   ovaries                    |  overies.           |
_______________|______________________________|_____________________|


____________________________________________________________________
Concentration  |     Week 65 sacrifice      |   Terminal sacrifice  |
    ppm        |                            |                       |
_______________|____________________________|_______________________|
0..............|lymphoma, heart, lung,      |lymphoma, heart, lung, |
               |   ovaries                  |  overies.             |
6.25...........|lung, ovaries...............|lymphoma, heart, lung, |
               |                            |  ovaries.             |
20.............|lung, ovaries...............|lymphoma, heart, lung, |
               |                            |  ovaries.             |
62.5...........|lymphoma, heart, lung,      |lymphoma, heart, lung, |
               |   ovaries                  |  overies.             |
200............|lymphoma, heart, lung,      |lymphoma, heart, lung, |
               |   ovaries                  |  overies.             |
_______________|____________________________|_______________________|
  Footnote(a) These organs and tissue types are those contained in the
OSHA risk assessment and do not reflect all ot the types of tissues
which were microscopically examined.
  Footnote(b) Heart, specifically Heart hemangiosarcoma.
  Footnote(c) None of the four tissue types used in the OSHA
quantitative risk assessment were microscopically examined.


  TABLE V-12.--TYPES OF TISSUES MICROSCOPICALLY EXAMINED BY
               CONCENTRATION DONE AND DISPOSITION GROUPS AMONG
               MALE MICE FROM NTP(a)
____________________________________________________________________
Concentration  |  Natural death and moribund  |                     |
    ppm        |          sacrifice           |   Week 40 sacrifice |
_______________|______________________________|_____________________|
0..............|lymphoma, heart(b), lung,     |lymphoma, heart, lung|
6.25...........|lymphoma, heart, lung,        |none(c)..............|
20.............|lymphoma, heart, lung,        |none.................|
62.5...........|lymphoma, heart, lung,        |lymphoma, lung,      |
200............|lymphoma, heart, lung,        |lymphoma, heart, lung|
_______________|______________________________|_____________________|


____________________________________________________________________
Concentration  |     Week 65 sacrifice      |   Terminal sacfrifice |
    ppm        |                            |                       |
_______________|____________________________|_______________________|
0..............|lymphoma, heart, lung,      |lymphoma, heart, lung, |
6.25...........|none........................|lymphoma, heart, lung, |
20.............|heart, lung.................|lymphoma, heart, lung, |
62.5...........|lymphoma, heart, lung.......|lymphoma, heart, lung, |
200............|lymphoma, heart, lung,......|lymphoma, heart, lung, |
_______________|____________________________|_______________________|
  Footnote(a) These organs and tissue types are those contained in the
OSHA risk assessment and do not reflect all of the types of tissues
which were microscopically examined.
  Footnote(b) Heart, specifically heart, hemangiosarcoma
  Footnote(c) None of the four tissue types used in the OSHA
quantitative risk assessment were microscopically examined.

2. Measure of Dose

The mechanism of cancer induction by BD is unknown for both rodents and humans. One or more of the metabolites of BD, epoxybutene, diolepoxybutane and diepoxybutane, are suspected as being responsible for the carcinogenic response in at least some of the cancers. However, which of the metabolites may be responsible for how much of the carcinogenic response has yet to be determined. Bond suggests that epoxybutene and diepoxybutane may be responsible for carcinogenic responses. (Ex. 32-28) Dr. Bond wrote:

If carcinogenic response is elicited by a metabolite, as has been suggested, mice because of their higher rate of metabolism, might be expected to yield a greater (carcinogenic) response than rats. (Ex. 17-21)

Because there are different theories about which metabolites of BD are responsible for the various carcinogenic responses, some risk assessments have characterized carcinogenic risk as a result of type of dose: External, absorbed, or retained. In the BD proposal (55 FR 32736), OSHA calculated the (14)C-BD equivalents that were retained in mice at the conclusion of a 6-hour exposure period and incorrectly labeled the level as "absorbed dose." This does not necessarily represent all the BD absorbed through inhalation exposure. (Ex. 34-1) The metabolic and pharmacokinetic properties of BD have not been fully characterized for either humans or animals. Despite the absence of a generally accepted pharmacokinetic model, some metabolic information can still be applied to OSHA's QRA. The overall rate of BD metabolism in B(6)C(3)F(1) mice is approximately linear at external concentrations up to 200 ppm; BD metabolism increases sublinearly as concentrations increase until it is saturated at 625 ppm. (Ex. 90) Bond reported that epoxybutene is one of the putative carcinogenic metabolites for which metabolism in the B(6)C(3)F(1) mouse becomes saturated at 500 ppm; thus, the B(6)C(3)F(1) mouse is unable to eliminate epoxybutene as quickly above 500 ppm. Bond suggests that above 500 ppm direct quantitative extrapolation of risk from mouse studies may not be justified. (Ex. 23-86) Therefore, the 625 ppm exposure group was excluded from OSHA's risk assessment. Similarly, NIOSH and Shell did not include the 625 ppm exposure group in their best estimates of risks using NTP II data. However, NIOSH did include the 625 ppm dose group in its sensitivity analyses to see how the inclusion of the data would affect the specification (the form and number of dose explanatory variables e.g., d, d(2), d(3), etc.) of the model and the estimates of risk. (Ex. 90)

3. Animal-to-Human Extrapolation

A QRA based on a mouse bioassay requires setting values for some mouse and human variables, including those used in animal-to-human extrapolations. The values of these variables were chosen before conducting the analyses. In OSHA's quantitative risk assessment, a mouse's life span was assumed to be 113 weeks. Mice were 8 weeks old at the beginning of the study and were exposed for up to 105 weeks. OSHA assumes workers will have an average lifespan of 74 years and an occupational lifetime, working 5 days/week, 50 weeks/year, of 45 years. In the NTP II study, the average male mouse weighed 40.8 grams and female mouse weighed 38.8 grams. (Ex. 90) Mice were assigned breathing rates of 0.0245 l/min. Breathing rates of workers (for an 8-hour workday) were set at 10 m(3)/8-hr.

OSHA has chosen to use a straight mg/kg, body weight to the first power, (BW)(1), intake as the animal-to-human species extrapolation factor for dose equivalence. Other BD QRAs employed various extrapolation factors such as ppm equivalence, (mg/kg)(3/4) equivalence, BD mono-epoxide blood levels between mice and monkey equivalence, and BD total body equivalence in (mg/kg)(2/3). OSHA believes that the evidence for the use of any of the alternative extrapolation factors is persuasive, although the Agency believes that body weight extrapolation is appropriate in this case because of the systemic nature of the tumors observed in both animal bioassays. This conversion of body weight, (BW)(1) , produces estimates of risk which are lower than those derived using (BW)(3/4), everything else held constant. For example, with a linear, 1-stage model, if OSHA used the (BW)(3/4) conversion, holding all other elements constant, one would expect the estimates of excess risk to humans to be about 6.5 times higher than if the (BW) extrapolation factor had been used because of the weight of the experimental species (between 38.8 and 40.8 grams), and their breathing rate. For the quadratic (2-stage) and cubic (3-stage) models, the effect of relying on the (BW)(3/4) conversion rather than the (BW)(1), holding all else constant, would be to increase the predicted excess human risk more than 6.5 fold. (Ex. 90)

4. Estimation of Occupational Dose

It is necessary to estimate the development of cancer at a variety of occupational doses. This requires occupational doses to be converted into units comparable to those used to measure the animal experimental dose. As discussed earlier, OSHA first converted animal experimental exposures measured in ppm into occupational intake dose measured in (mg/kg).

An exposure of 1 ppm BD is converted into an equivalent exposure measured in mg/m(3) using the equation:


             Molecular Weight BD
1 ppm BD = -------------------------- X density of air
            Molecular Weight of Air

             54.1 mg/mole
1 ppm BD = ------------------ = 2.21 BD mg/m(3)
            24.45 mole/m(3)


  Given a worker weighing 70 kg, breathing 10 m(3) of air per 8-hour
day, and exposed to air containing Y ppm BD, the inhaled dose of BD in
mg/kg is given by:

                                       mg/m(3)   10 m(3)
Y(mg.kg)BD inhaled = Y (ppm)BD x 2.21 -------- x --------
                                         ppm       70 kg

Using the above formula, one can calculate the estimated equivalent inhaled BD exposure among workers based on the exposure concentrations for animals (See Table V-13).


TABLE V-13.--ESTIMATE OF TOTAL HUMAN INHALED DOSE OVER A WORKDAY FOR
                        VARIOUS EXPOSURE LEVELS OF BD
_____________________________________________________________________
                                              Estimate of total human
      Exposure concentrations (ppm)              inhaled BD over a
                                              workday (mg/kg/8-hours)
_____________________________________________________________________
200............................................                63.2
62.5...........................................                19.8
20.............................................                 6.3
5..............................................                 1.6
2..............................................                 0.6
1..............................................                 0.3
____________________________________________________________________

5. Selection of Model for Quantitative Risk Assessment

In the proposal (55 FR 32736), OSHA estimated excess risk using a quantal form of the multistage model (in a reparameterized form as calculated by GLOBAL83), which based estimates of risk to humans on the experience of the group rather than the individual. Three of the later risk assessments, Shell, NIOSH, and COHP, used a Weibull time-to-tumor form of the multistage model to fit the mouse bioassays. (Exs. 32-27; 90; 32-16) Time-to-tumor models use more of the available information than quantal multistage models to characterize time until the development of each observable tumor, and extrapolate risks, based on an occupational dosing pattern. Since significant increases in tumor incidence occurred at multiple sites in the NTP II bioassay and a time-to-tumor model takes these competing risks into account, a time-to-tumor method is preferred over a quantal model. (Ex. 118-1B) Therefore OSHA used a Weibull time-to-tumor form of the multistage model to characterize the risks of development of observable tumors, using the software package, TOX__RISK Version 3.5 by ICF Kaiser. The model predicts the probability, P(t,d), of tumor onset with dose pattern d by time t. It adjusts for competing causes of death prior to time t.

The Weibull time-to-tumor model is a multistage model based on the theory of carcinogenesis developed by Armitage and Doll. This theory of carcinogenesis is based on the assumption that a single line of stem cells must pass through a certain number of stages sequentially for the development of a single tumor cell. In the reparameterized form of the model used here, a k stage model is described by a polynomial of degree k, with all dose parameters greater than or equal to zero. The number of stages necessary for a model to be correctly specified varies by type of tumor, animal, and exposure agent, or any combination of the three.

Both the MLE and the 95% upper limit of the risk of developing cancer in various tissues per 1,000 workers by time t are calculated. The 95% upper bound is the largest value of excess risk that is consistent with the observed data with two-sided 95% confidence intervals. The 95% upper bound is computed based on the Weibull time-to-tumor model for which the parameters satisfy:

-2 (Log likelihood-Log likelihood(max))< /=2.70554

Where: Log likelihood(max) is the maximum value of the log-likelihood

A 1-stage model is linear in dose; a 2-stage model is quadratic in dose; a 3-stage Weibull model is cubic in dose. Below is a mathematical representation of a 3-stage Weibull time-to-tumor model:

P(t,d)=1-exp [-(q(0) + q(1d) + q(2)d(2) + q(3)d(3))(t-t(0)(z))]

where: t(0) designates the time of onset of the tumor, t is the variable for time the tumor was observed and is assumed to follow a Weibull distribution; d is the dose-metric and is multistage; z is a parameter to be estimated, constrained between 1 and 10; the background parameter q(o) and the dose parameters, q(1), q(2), q(3), are constrained to be non-negative. Constraining the dose parameters to zero or greater is biologically based, since the dose parameters are proportional to the mutation rates of the successive stages in the development of a tumor cell. The Weibull time-to-tumor model provides reasonable fits for about 75% of the tissues in the NTP historical control data base, but the precision of the fit to the dose-response data depends on the specific agent. (Ex. 90) Four forms of the model, one less than the number of exposure groups, for each gender-outcome were fit to the data. The correct specification of the model, the number of stages, is determined by the fit of the model to the data. The likelihood ratio test identifies which model is a better fit by determining if the log-likelihood of a model is significantly greater than another model's value. The 1-, 2-, 3- and 4-stage Weibull time-to-tumor models for each gender-outcome combination were ordered according to the value of their log-likelihood. If the log-likelihood of the higher stage model is significantly greater than that of the next lower stage model's log-likelihood, one would reject the null hypothesis (the additional stage does not create a model that better characterizes the data) and conclude that the higher stage model is a significantly better predictor of the estimates of risk in the observed range than is the lower stage model.

The steps of the likelihood ratio test are as follows:

For example, assuming an alpha of 0.05, and 1 degree of freedom (the difference in the number of parameters from 1-stage and 2-stage models), the critical value would be 3.84.


  Fail to Reject H(0) if:

2 (log likelihood(1-stage)-log likelihood(2-stage))< 3.84

  Reject H(0) if:

2 (log likelihood(1-stage)-log likelihood(2-stage))>/=3.84

If two times the difference of the log likelihood values of the nth stage model and the nth + 1 stage model was less than 3.84, then the additional stage would be deemed unnecessary for goodness of fit; on the grounds of parsimony, the lower stage model would be used for the risk assessment. Otherwise, the higher stage model would be judged a better fit than the lower stage one and the process would continue.

While the likelihood ratio test is suitable for testing the significance of the next higher degree dose parameter, the biologically reasonable constraint on the background incidence parameter q(0) and dose parameters that they be non-negative q(1), q(2), q(3)>=0,--may impair the log-likelihood ratio test's power to determine statistical significance.

The incidences of lymphoma, heart hemangiosarcoma, lung and ovarian tumors are shown in Tables V-14 and V-15 for males and females, respectively. The TOXRISK Weibull time-to-tumor model requires that the tumor context be described for each observation. Outcomes can be put into three context categories: (1) Censored, no tumor; (2) rapidly fatal tumor; and (3) observed, tumor incidental to the animal's survival. Since OSHA was predicting the time until onset of tumor, assuming no lag time between onset and detection of tumor, t(0) was set to zero. Therefore, estimates of risk to humans based on the contribution to the likelihood of either a rapidly fatal or incidental tumor are mathematically the same.

Tables V-16 and V-17 show the Weibull time-to-tumor model estimates of log-likelihoods, the shape parameters, intercept and dose coefficients for relevant target tissues for male and female mice, respectively. The relative performance of various staged models for a specific target tissue-gender are enumerated in the log-likelihood values. It should be noted that some of the tissue-gender combination's log-likelihood values do not vary even though there is a change in the number of the stages in the model. For example, the log-likelihood values for models of all lymphoma for males and lung tumors for males and females are -6.986 E+1, -1.763 E+2, -1.626 E+2, respectively, regardless of the specification, number of stages, in the model. OSHA concluded that the 1-stage models were preferred.

As identified in Tables V-16 and V-17, only heart hemangiosarcoma models are non-linear. This is consistent with NIOSH's results when fitting Weibull time-to-tumor models to these gender-tumor combinations. The quadratic (2-stage) model for males and the cubic (3-stage) model for females better characterized the dose-response relationship in modeling time to detection of heart hemangiosarcoma than did the linear models. The higher stage model necessary to fit the heart hemangiosarcoma data is driven by the absence of cases in the two lower exposure groups, shown in Tables V-14 and V-15. Unlike the other tissues studied, there were no cases of heart hemangiosarcoma in the control and lowest exposure groups for both male and female mice. Both male and female mice had similar heart hemangiosarcoma tumor rates, almost 30%, among the 200 ppm exposure groups. The intercepts, q(0), were zero for models of both male and female mice based on the dose-response of heart hemangiosarcomas. This is consistent with what one would expect, given the absence of background incidence rates of heart hemangiosarcomas.


TABLE V-14.--UNIVARIATE ANALYSIS OF HEART, LUNG, AND ALL LYMPHOMA
             NEOPLASMS BY EXPOSURE LEVEL OF 1,3-BUTADIENE AMONG
             NTP II MALE MICE ANALYZED IN THE TIME-TO-TUMOR MODELS
_______________________________________________________________________
                                   |
                                   |                Outcome
                                   |___________________________________
          Neoplasm                 |          |             |
                                   |Tumor n(a)| Censored(c) |
                                   | (%N(b))  |  n (%N)     | Total N
___________________________________|__________|_____________|__________
All lymphoma, 0 ppm................|  4 (5.7) |  66 (94.3)  |   70
All lymphoma, 6.25 ppm.............|  3 (6.0) |  47 (94.0)  |   50
All lymphoma, 20 ppm...............|  8 (16.0)|  42 (84.0)  |   50
All lymphoma, 62.5 ppm.............| 11 (15.9)|  58 (84.1)  |   69
All lymphoma, 200 ppm..............|  9 (12.9)|  61 (87.1)  |   70
Heart hemangiosarcoma, 0 ppm.......|  0 (0)   |  70 (100)   |   70
Heart hemangiosarcoma, 6.25 ppm....|  0 (0)   |  49 (100)   |   49
Heart hemangiosarcoma, 20 ppm......|  1 (1.7) |  59 (98.3)  |   60
Heart hemangiosarcoma, 62.5 ppm....|  5 (8.6) |  53 (91.4)  |   58
Heart hemangiosarcoma, 200 ppm.....| 20 (29.4)|  48 (70.6)  |   68
Lung tumor, 0 ppm..................| 22 (31.4)|  48 (68.6)  |   70
Lung tumor, 6.25 ppm...............| 23 (46.9)|  26 (53.1)  |   49
Lung tumor, 20 ppm.................| 20 (33.3)|  40 (66.7)  |   60
Lung tumor, 62.5 ppm...............| 33 (47.8)|  36 (52.2)  |   69
Lung tumor, 200 ppm................| 42 (60.0)|  28 (40.0)  |   70
___________________________________|__________|_____________|__________
  Footnote(a) n is number of microscopically determined outcomes per
tumor-context, gender, exposure-group outcome site combination.
  Footnote(b) N is the total number of gender, exposure-group, outcome
site combination which were microscopically examined.
  Footnote(c) Tumor's context is C (censored); animals were
microscopically examined and no tumor was found at this site.


TABLE V-15.--UNIVARIATE ANALYSIS OF HEART, LUNG, ALL LYMPHOMA AND
             OVARIAN NEOPLASMS BY EXPOSURE LEVEL OF 1,3-BUTADIENE
             AMONG NTP II FEMALE MICE ANALYZED IN THE TIME-TO-TUMOR
             MODELS
_______________________________________________________________________
                                 |
                                 |                Outcome
                                 |_____________________________________
        Neoplasm                 |             |            |
                                 |Tumor n(a)   |Censored (c)|
                                 |   (%N(b))   |  n (%N)    | Total N
_________________________________|_____________|____________|__________
All lymphoma, 0 ppm..............| 10 (14.3)   | 60 (85.7)  | 70
All lymphoma, 6.25 ppm...........| 14 (28.0)   | 36 (72.0)  | 50
All lymphoma, 20 ppm.............| 18 (36.0)   | 32 (64.0)  | 50
All lymphoma, 62.5 ppm...........| 10 (14.3)   | 60 (85.7)  | 70
All lymphoma, 200 ppm............| 19 (27.1)   | 51 (72.9)  | 70
Heart hemangiosarcoma, 0 ppm.....|  0 (0)      | 70 (100)   | 70
Heart hemangiosarcoma, 6.25 ppm..|  0 (0)      | 50 (100)   | 50
Heart hemangiosarcoma, 20 ppm....|  0 (0)      | 50 (100)   | 50
Heart hemangiosarcoma, 62.5 ppm..|  1 (1.7)    | 58 (98.3)  | 59
Heart hemangiosarcoma, 200 ppm...| 20 (28.6)   | 50 (71.4)  | 70
Lung tumor, 0 ppm................|  4 (5.7)    | 66 (94.3)  | 70
Lung tumor, 6.25 ppm.............| 15 (25.0)   | 45 (75.0)  | 60
Lung tumor, 20 ppm...............| 19 (31.7)   | 41 (68.3)  | 60
Lung tumor, 62.5 ppm.............| 27 (38.6)   | 43 (61.4)  | 70
Lung tumor, 200 ppm..............| 32 (45.7)   | 38 (54.3)  | 70
Ovarian tumor, 0 ppm.............|  1 (1.4)    | 68 (98.6)  | 69
Ovarian tumor, 6.25 ppm..........|  0 (0)      | 59 (100)   | 59
Ovarian tumor, 20 ppm............|  0 (0)      | 59 (100)   | 59
Ovarian tumor, 62.5 ppm..........|  9 (12.9)   | 61 (87.1)  | 70
Ovarian tumor, 200 ppm...........| 11 (15.7)   | 59 (84.3)  | 70
_________________________________|_____________|____________|__________
  Footnote(a) n is number of microscopically determined outcomes per
tumor-context, gender, exposure-group outcome site combination.
  Footnote(b) N is the total number of gender, exposure-group,
outcome site combination which were microscopically examined.
  Footnote(c) Tumor's context is C (censored); animals were
microscopically examined and no tumor was found at this site.


TABLE V-16.--MAXIMUM LIKELIHOOD ESTIMATES OF MODEL COEFFICIENTS FROM
VARIOUS STAGES OF WEIBULL TIME-TO-TUMOR MODELS USING THREE TUMOR
RESPONSES OF MALE MICE IN THE NTP II STUDY, EXCLUDING 625 PPM EXPOSURE
GROUP; SELECTION OF SPECIFICATION OF MODEL IS BASED ON LIKELIHOOD
                           RATIO TEST


__________________________________________________________________
    Neoplasm                 |  Stage(a) | Log-likelihood |  Z(b) |
_____________________________|___________|________________|_______|
Heart hemangiosarcoma........|     W1    | -7.061         | 9.810 |
Heart hemangiosarcoma........|  (c)W2    | -2.783 E-1     |    10 |
Heart hemangiosarcoma........|     W3    | -2.712 E-1     |    10 |
Heart hemangiosarcoma........|     W4    | -2.659 E-1     |    10 |
All lymphoma.................|  (c)W1    | -6.986 E+1     | 4.743 |
All lymphoma.................|     W2    | -6.986 E+1     | 4.743 |
All lymphoma.................|     W3    | -6.986 E+1     | 4.743 |
All lymphoma.................|     W4    | -6.986 E+1     | 4.743 |
Lung tumor...................|  (c)W1    | -1.763 E+2     | 3.318 |
Lung tumor...................|     W2    | -1.760 E+2     | 3.413 |
Lung tumor...................|     W3    | -1.760 E+2     | 3.143 |
Lung tumor...................|     W4    | -1.760 E+2     | 3.413 |
_____________________________|___________|________________|_______|


__________________________________________________________________
    Neoplasm                 |    q(0)   |   q(1)    |    q(2)    |
_____________________________|___________|___________|____________|
Heart hemangiosarcoma........|0.00       |8.306 E-23 |            |
Heart hemangiosarcoma........|0.00       |0.00       |3.071 E-25  |
Heart hemangiosarcoma........|0.00       |1.058 E-24 |2.636 E-25  |
Heart hemangiosarcoma........|0.00       |1.119 E-24 |2.664 E-25  |
All lymphoma.................|2.709 E-11 |6.136 E-13 |            |
All lymphoma.................|2.709 E-11 |6.136 E-13 |0.00        |
All lymphoma.................|2.709 E-11 |6.136 E-13 |0.00        |
All lymphoma.................|2.709 E-11 |6.136 E-13 |0.00        |
Lung tumor...................|1.132 E-7  |2.636 E-9  |            |
Lung tumor...................|7.674 E-8  |1.253 E-9  |3.134 E-12  |
Lung tumor...................|7.674 E-8  |1.253 E-9  |3.134 E-12  |
Lung tumor...................|7.674 E-8  |1.253 E-9  |3.134 E-12  |
_____________________________|___________|___________|____________|


_____________________________________________________
    Neoplasm                 |    q(3)   |   q(4)    |
_____________________________|___________|___________|
Heart hemangiosarcoma........|           |           |
Heart hemangiosarcoma........|           |           |
Heart hemangiosarcoma........|2.057 E-28 |           |
Heart hemangiosarcoma........|0.00       |9.626 E-31 |
All lymphoma.................|           |           |
All lymphoma.................|           |           |
All lymphoma.................|6.540 E-33 |           |
All lymphoma.................|0.00       |0.00       |
Lung tumor...................|           |           |
Lung tumor...................|           |           |
Lung tumor...................|0.00       |           |
Lung tumor...................|0.00       |0.00       |
_____________________________|___________|___________|
  Footnote(a) Stage of time-to-tumor model; W1, Weibull 1-stage
time-to-tumor model; W2, Weibull 2-stage time-to-tumor model; W3,
Weibull 3-stage time-to-tumor model; W4, Weibull 4-stage
time-to-tumor model.
  Footnote(b) Z is the shape parameter; it is bounded, (1< =z< =10).
  Footnote(c) Selected Model.


TABLE V-17.-- MAXIMUM LIKELIHOOD ESTIMATES OF MODEL COEFFICIENTS FROM
   VARIOUS STAGES OF WEIBULL TIME-TO-TUMOR MODELS USING FOUR TUMOR
   RESPONSES OF FEMALE MICE IN THE NTP II STUDY, EXCLUDING 625 PPM
   EXPOSURE GROUP; SELECTION OF SPECIFICATION OF MODEL IS BASED ON
                      LIKELIHOOD RATIO TEST

__________________________________________________________________
    Neoplasm                 |Stage(a)| Log-likelihood |   Z(b)  |
_____________________________|________|________________|_________|
Heart hemangiosarcoma........|  W1    |   -2.097 E+1   |  4.957  |
Heart hemangiosarcoma........|  W2    |   -8.745       |  6.126  |
Heart hemangiosarcoma........|  W3(c) |   -4.866       |  6.770  |
Heart hemangiosarcoma........|  W4    |   -4.267       |  7.011  |
Ovarian tumor................|  W1(c) |   -6.140 E+1   |  2.857  |
Ovarian tumor................|  W2    |   -6.069 E+1   |  4.079  |
Ovarian tumor................|  W3    |   -6.069 E+1   |  4.079  |
Ovarian tumor................|  W4    |   -6.069 E+1   |  4.079  |
All lymphoma.................|  W1(c) |   -5.724 E+1   |  6.857  |
All lymphoma.................|  W2    |   -5.501 E+1   |  7.143  |
All lymphoma.................|  W3    |   -5.426 E+1   |  7.230  |
All lymphoma.................|  W4    |   -5.401 E+1   |  7.258  |
Lung tumor...................|  W1(c) |   -1.626 E+2   |  3.416  |
Lung tumor...................|  W2    |   -1.626 E+2   |  3.416  |
Lung tumor...................|  W3    |   -1.626 E+2   |  3.416  |
Lung tumor...................|  W4    |   -1.626 E+2   |  3.416  |
_____________________________|________|________________|_________|


______________________________________________________________
    Neoplasm                 |   q(0)   |   q(1)   |   q(2)   |
_____________________________|__________|__________|__________|
Heart hemangiosarcoma........|   0.00   |4.356 E-13|          |
Heart hemangiosarcoma........|   0.00   |   0.00   |2.222 E-17|
Heart hemangiosarcoma........|   0.00   |   0.00   |   0.00   |
Heart hemangiosarcoma........|   0.00   |   0.00   |   0.00   |
Ovarian tumor................|1.407 E-8 | .031 E-9 |          |
Ovarian tumor................|5.397 E-11|7.075 E-12|1.399 E-13|
Ovarian tumor................|5.397 E-11|7.075 E-12|1.399 E-13|
Ovarian tumor................|5.397 E-11|7.075 E-12|1.399 E-13|
All lymphoma.................|3.453 E-15|1.338 E-16|          |
All lymphoma.................| 1.18 E-15|2.577 E-18|2.453 E-19|
All lymphoma.................|7.758 E-16|6.847 E-18|   0.00   |
All lymphoma.................|7.360 E-18|7.359 E-18|   0.00   |
Lung tumor...................|2.096 E-8 |2.096 E-9 |          |
Lung tumor...................|2.090 E-8 |2.090 E-9 |   0.00   |
Lung tumor...................|2.090 E-8 |2.096 E-9 |   0.00   |
Lung tumor...................|2.090 E-8 |2.096 E-9 |   0.00   |
_____________________________|__________|__________|__________|


___________________________________________________
    Neoplasm                 |   q(3)   |   q(4)   |
_____________________________|__________|__________|
Heart hemangiosarcoma........|          |          |
Heart hemangiosarcoma........|          |          |
Heart hemangiosarcoma........|8.088 E-21|          |
Heart hemangiosarcoma........|2.637 E-22|1.368 E-3 |
Ovarian tumor................|          |          |
Ovarian tumor................|          |          |
Ovarian tumor................|0.00      |          |
Ovarian tumor................|0.00      |0.00      |
All lymphoma.................|          |          |
All lymphoma.................|          |          |
All lymphoma.................|7.809 E-22|          |
All lymphoma.................|0.00      |3.387 E-24|
Lung tumor...................|          |          |
Lung tumor...................|          |          |
Lung tumor...................|0.00      |          |
Lung tumor...................|0.00      |0.00      |
_____________________________|__________|__________|
  Footnote(a) Stage of time-to-tumor model; W1, Weibull 1-stage
time-to-tumor model; W2, Weibull 2-stage time-to-tumor model; W3,
Weibull 3-stage time-to-tumor model; W4, Weibull 4-stage time-to-tumor
model.
  Footnote(b) Z is the shape parameter; it is bounded, (1< =z< =10).
  Footnote(c) Selected Model.

OSHA's Estimates of Risk

The estimates from OSHA's quantitative risk assessment based an 8-hour TWA, occupational lifetime, working 5 days/week, 50 weeks/year, for 45 years, at various BD PELS are shown in Table V-18. The MLEs of excess risk of material impairment of health per 1,000 workers for cancer, based on tumors of various tissue sites and the 95% upper bounds, are presented. Various 8-hour TWA PELS, ranging from 0.1 to 5 ppm, are presented to provide a context in which to evaluate the OSHA final rule PEL of 1 ppm and to explore the feasibility of other PELS, including the proposed PEL of 2 ppm. Risks at the former BD 8-hour TWA PEL, 1,000 ppm, are not presented in Table V-18. Although risks could be estimated for an occupational lifetime exposure to an 8-hour TWA of 1,000 ppm of BD from the linear models, there is little relevancy to estimating the true risk at an 8-hour PEL for BD at 1,000 ppm for an occupational lifetime, since 8-hour TWA BD exposures have been generally far lower than 1,000 ppm.

Although the estimates of carcinogenic outcomes differ, excess risks derived from tumor sites common to both male and female B(6)C(3)F(1) mice had the same relative ranking from lowest to highest risk estimates by target tissues (heart hemangiosarcomas < lymphomas < lungs) within each gender group. After a lifetime occupational exposure to BD at the proposed 8-hour TWA PEL of 2 ppm based on the above model fits to these three individual tumor sites, one would expect between 2.7 x 10(-4) to 16.2 excess cancer cases per 1,000 workers, depending on which gender-tumor site dose-response relationship is used as the basis for the extrapolation to human occupational excess risks. Decreasing the BD 8-hour TWA PEL from 2 to 1 ppm, results in a reduction of the range of estimates of excess risk of cancer to between 3.4 x 10(-5) to 8.1 cases per 1,000 workers.

The estimate of excess cancer risk based on male mouse lymphoma is 1.3 per 1,000 workers at an 8-hour TWA for an occupational lifetime exposure to 1 ppm BD. Extrapolating from female mouse lymphoma data results in an estimate of 6.0 extra cancer deaths per 1,000 workers at a BD 8-hour TWA PEL of 1 ppm for an occupational lifetime of exposure.

Extrapolating from the most sensitive site, the female mouse lung, based on the 1-stage Weibull time-to-tumor model, with an 8-hour TWA PEL of 2 ppm of BD for an occupational lifetime, one would expect 16 excess cancer cases per 1,000 workers. Lowering the PEL to 1 ppm would cut the expected number of excess cancers in half to 8 cases, based on the same gender-tumor site. Based on male lung tumors, the estimate of excess cancer deaths for an 8-hour TWA exposure to 2 ppm BD over an occupational lifetime was 12.8 per 1,000 workers; lowering the 8-hour TWA occupational lifetime exposure level to 1 ppm BD decreases the estimate of excess cancer risk to 6.4 per 1,000 workers, a reduction of 6 cancer cases per 1,000 workers.

OSHA's estimates of premature occupational leukemia deaths based on the 1-stage Weibull time-to-tumor models for the following outcome sites: All lymphoma, lung tumors, and ovarian tumors, ranged between 1.3 and 8.1 per 1,000 workers. Similarly, NIOSH's 14 estimates of the excess risk of death due to leukemia, based on 1-stage Weibull time-to-tumor models, as a consequence of exposure to an 8-hour TWA of 1 ppm BD over an occupational lifetime, ranged between 0.9 and 30 cases per 1,000 workers. The preliminary estimate of 8 per 1,000 from the Delzell et al. study is concordant with this range of animal-based estimates. OSHA acknowledges that there is uncertainty in the Delzell et al. estimate, perhaps due to the natural sampling variability present in any epidemiologic study plus the possibility of extra-binomial uncertainty stemming from exposure misclassification. While this uncertainty makes it difficult to say whether quantitative risk estimates would be adjusted up or down relative to animal-based estimates, this suggestion is far less important than the basic conclusion that the Delzell et al. study reinforces earlier estimates. Even if refinement of exposures caused the Delzell et al. estimate to move up or down by even as much as a factor of 5 or more, it would not change this qualitative, and roughly quantitative, agreement.


TABLE V-18.--MAXIMUM LIKELIHOOD ESTIMATES (MLE) AND NINETY-FIVE
             PERCENT UPPER BOUNDS OF LIFETIME EXTRA TISK TO DEVELOP AN
             OBSERVABLE TUMOR PER 1,000 WORKERS DUE TO AN 8-HOUR TWA
             FOR AN OCCUPATIONAL LIFETIME(a) OF EXPOSURE TO
             1,3-BUTADIENE, USING NTP II BIOASSAY(b) AND THE BEST
             FITTING WEIBULL TIME-TO-TUMOR MODELS
_________________________________________________________________
                         |      |  8-hour time-weighted average  |
                         |      |        concentration(c)        |
                         |      |________________________________|
                         |Stages|     0.1 ppm   |     0.2 ppm    |
                         |      |_______________|________________|
                         |      |  MLE  |  95%  |  MLE  |   95%  |
                         |      |       |U.B.(d)|       |   U.B. |
_________________________|______|_______|_______|_______|________|
Male mice:               |      |       |       |       |        |
  Heart Hemangiosarcoma..|   2  |(e)< 0.1|  0.2  |(e)< 0.1|  0.4   |
  All lymphoma...........|   1  |  0.1  |  0.2  |  0.3  |  0.5   |
  Lung tumor.............|   1  |  0.7  |  0.1  |  1.3  |  2.0   |
Female mice:             |      |       |       |       |        |
  Heart Hemangiosarcoma..|   3  |(f)< 0.1|(f)< 0.1|(f)< 0.1| < 0.1   |
  Ovarian tumor..........|   1  |  0.1  |  0.3  |  0.3  |  0.5   |
  All lymphoma...........|   1  |  0.6  |  0.9  |  1.2  |  1.8   |
  Lung tumor.............|   1  |  0.8  |  1.2  |  1.6  |  2.4   |
_________________________|______|_______|_______|_______|________|


_________________________________________________________________
                         |      |  8-hour time-weighted average  |
                         |      |        concentration(c)        |
                         |      |________________________________|
                         |Stages|     0.5 ppm   |      1 ppm     |
                         |      |_______________|________________|
                         |      |  MLE  |  95%  |  MLE  |   95%  |
                         |      |       |  U.B. |       |   U.B. |
_________________________|______|_______|_______|_______|________|
Male mice:               |      |       |       |       |        |
  Heart Hemangiosarcoma..|   2  |(e)< 0.1|  0.9  |(e)< 0.1|  1.8   |
  All lymphoma...........|   1  |  0.6  |  1.1  |  1.3  |  2.3   |
  Lung tumor.............|   1  |  3.2  |  4.9  |  6.4  |  9.8   |
Female mice:             |      |       |       |       |        |
  Heart Hemangiosarcoma..|   3  |(f)< 0.1|  0.2  |(f)< 0.1|  0.5   |
  Ovarian tumor..........|   1  |  0.7  |  1.3  |  1.4  |  2.6   |
  All lymphoma...........|   1  |  3.0  |  4.6  |  6.0  |  9.2   |
  Lung tumor.............|   1  |  4.1  |  6.1  |  8.1  | 12.2   |
_________________________|______|_______|_______|_______|________|


_________________________________________________________________
                         |      |  8-hour time-weighted average  |
                         |      |        concentration(c)        |
                         |      |________________________________|
                         |Stages|      2 ppm    |      5 ppm     |
                         |      |_______________|________________|
                         |      |  MLE  |  95%  |  MLE  |   95%  |
                         |      |       |  MLE  |       |   U.B. |
_________________________|______|_______|_______|_______|________|
Male mice:               |      |       |       |       |        |
  Heart Hemangiosarcoma..|   2  |(e)< 0.1|  3.6  |  0.4  |  9.1   |
  All lymphoma...........|   1  |  2.5  |  4.5  |  6.3  | 11.2   |
  Lung tumor.............|   1  | 12.8  | 19.4  | 31.7  | 47.9   |
Female mice:             |      |       |       |       |        |
  Heart Hemangiosarcoma..|   3  |(f)< 0.1|  1.0  |(f)< 0.1|  2.4   |
  Ovarian tumor..........|   1  |  2.8  |  5.2  |  6.9  | 13.0   |
  All lymphoma...........|   1  | 12.0  | 18.3  | 29.7  | 45.0   |
  Lung tumor.............|   1  | 16.2  | 24.1  | 40.00 | 59.4   |
_________________________|______|_______|_______|_______|________|
  Footnote(a) Occupational lifetime, working 5 days/week, 50
weeks/year, for 45 years.
  Footnote(b) Using data from NTP II for the following exposure
groups: 0,6.25, 20, 62.5 and 200 ppm; the 625 ppm exposure group was
excluded.
  Footnote(c) Estimated lifetime excess risk for cancer assuming:
mouse life-span of 113 weeks, male mouse body weight of 40.8g; female
mouse body weight of 38.8g; worker's breathing rate is 1.25 m(3)/hr;
mouse to human risk extrapolated in mg/kg-day equivalent units.
  Footnote(d) 95% U.B., 955 Upper Bounds is the largest value of
excess risk that is compatible with the animal response data at a
confidence level of 95%.
  Footnote(e) MLEs ranged from 1.5u10-(4) to 6.0u10-(2)
  Footnote(f) MLEs ranged from 3.4u10-(8) to 4.3u10-(3)


Regulations (Preambles to Final Rules) - Table of Contents Regulations (Preambles to Final Rules) - Table of Contents