[[pp. 5806-5855]] Diesel Particulate Matter Exposure of Underground Metal and

Note: EPA no longer updates this information, but it may be useful as a reference or resource.
 [Federal Register: January 19, 2001 (Volume 66, Number 13)]
[Rules and Regulations]
[Page 5806-5855]
From the Federal Register Online via GPO Access [wais.access.gpo.gov]
[DOCID:fr19ja01-17]

[[pp. 5806-5855]] Diesel Particulate Matter Exposure of Underground Metal and
Nonmetal Miners

[[Continued from page 5805]]

[[Page 5805]]

were identified among more than 10,000 deaths in 1982 and 1983. For
comparison to these cases, a total of 1,085 controls was selected
(presumably at random) from the remaining deaths, restricted to those
who died from causes other than lung cancer, bladder cancer, or motor
vehicle accident. Information on work history, duration and intensity
of cigarette smoking, diet, and asbestos exposure was obtained from
next of kin. Detailed work histories were also obtained from pension
applications on file with the Teamsters Union.
    Both data sources were used to classify cases and controls
according to a job category in which they had worked the longest. Based
on the data obtained from next of kin, the job categories were diesel
truck drivers, gasoline truck drivers, drivers of both truck types,
truck mechanics, and dock workers. Based on the pension applications,
the principal job categories were long-haul drivers, short-haul or city
drivers, truck mechanics, and dock workers. Of the workers identified
by next of kin as primarily diesel truck drivers, 90 percent were
classified as long-haul drivers according to the Teamster data. The
corresponding proportions were 82 percent for mechanics and 81 percent
for dock workers. According to the investigators, most Teamsters had
worked in only one exposed job category. However, because of the
differences in job category definitions, and also because the next of
kin data covered lifetimes whereas the pension applications covered
only time in the Teamsters Union, the investigators found it
problematic to fully evaluate the concordance between the two data
sources.
    In the 1990 report, separate analyses were conducted for each
source of data used to compile work histories. The investigators noted
that ``many trucking companies (where most study subjects worked) had
completed most of the dieselization of their fleets by 1960, while
independent drivers and nontrucking firms may have obtained diesel
trucks later * * *'' Therefore, they specifically checked for
associations between increased risk of lung cancer and occupational
exposure after 1959 and, separately, after 1964. In the 1992 report,
the investigators presented, for the Union's occupational categories
used in the study, dpm exposure estimates based on subsequent
measurements of submicrometer elemental carbon (EC) as reported by
Zaebst et al. (1991). In the 1998 report, cumulative dpm exposure
estimates for individual workers were compiled by combining the
individual work histories obtained from the Union's records with the
subsequently measured occupational exposure levels, along with an
evaluation of historical changes in diesel engine emissions and
patterns of diesel usage. Three alternative sets of cumulative exposure
estimates were considered, based on alternative assumptions about the
extent of improvement in diesel engine emissions between 1970 and 1990.
A variety of statistical models and techniques were then employed to
investigate the relationship between estimated cumulative dpm exposure
(expressed as EC) and the risk of lung cancer. The authors pointed out
that the results of these statistical analyses depended heavily on
``very broad assumptions'' used to generate the estimates of cumulative
dpm exposure. While acknowledging this limitation, however, they also
evaluated the sensitivity of their results to various changes in their
assumptions and found these changes to have little impact on the
results.
    The investigators also identified and addressed several other
limitations of this study as follows:

    (1) possible misclassification smoking habits by next of kin,
(2) misclassification of exposure by next of kin, (3) a relatively
small non-exposed group (n = 120) which by chance may have had a low
lung cancer risk, and (4) lack of sufficient latency (time since
first exposure) to observe a lung cancer excess. On the other hand,
next-of-kin data on smoking have been shown to be reasonably
accurate, non-differential misclassification of exposure * * * would
only bias our findings toward * * * no association, and the trends
of increased risk with increased duration of employment in certain
jobs would persist even if the non-exposed group had a higher lung
cancer risk. Finally, the lack of potential latency would only make
any positive results more striking. (Steenland et al., 1990)

    The main results from the three reports covering this study are
summarized in the following table. All of the analyses were controlled
for age, race, smoking (five categories), diet, and asbestos exposure
as reported by next of kin. Odds ratios for the occupations listed were
calculated relative to the odds of lung cancer for occupations other
than truck driver (all types), mechanic, dock worker, or other
potentially diesel exposed jobs (Steenland et al., 1990, Appendix A).
The exposure-response analyses were carried out using logistic
regression. Although the investigators performed analyses under three
different assumptions for the rate of engine emissions (gm/mile) in
1970, they considered the intermediate value of 4.5 gm/mile to be their
best estimate, and this is the value on which the results shown here
are based. Under this assumption, cumulative occupational EC exposure
for all workers in the study was estimated to range from 0.45 to 2,440
g-yr/m3, with a median value of 373 g-yr/
m3. The estimates of relative risk (expressed as odds
ratios) presented for EC exposures of 373 g-yr/m3,
1000 g-yr/m3, and 2450 g-yr/m3
were calculated by MSHA based on the regression coefficients reported
by the authors for five-year lagged exposures (Steenland et al. 1998,
Table II).

BILLING CODE 4510-43-P

[[Page 5806]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.066

BILLING CODE 4510-43-C
    Under the assumption of a 4.5 gm/mile emissions rate in 1970, the
cumulative EC exposure of 2450 g-yr/m\3\ ( 6.1 mg-
yr/m\3\ dpm) shown in the table closely corresponds to the upper limit
of the range of data on which the regression analyses were based
(Steenland et al., 1998, p. 224). However, the relative risks (i.e.,
odds ratios) calculated for this level of occupational exposure are
presented primarily for purposes of comparison with the findings of
Johnston et al. (1997) and Saverin et al. (1999). At a cumulative dpm
exposure of approximately 6.1 mg-yr/m\3\, it is evident that the
Johnston models predict a far greater elevation in lung cancer risk
than either the Saverin or Steenland models. A possible explanation for
this is that the Johnston data included exposures of up to 30 years in
duration, and the statistical models showing an exposure-response
relationship allowed for a 15-year lag in exposure effects. The other
two studies were based on generally shorter diesel exposures and
allowed less time for latent effects. In Subsection 3.b.ii(3) of this
risk assessment, the quantitative results of these three studies will
be further compared with respect to

[[Page 5807]]

exposure levels found in underground mines.
    Several commenters noted that the HEI Expert Panel (HEI, 1999) had
identified uncertainties in the diesel exposure assessment as an
important limitation of the exposure-response analyses by Steenland et
al. (1998) and had recommended further investigation before the
quantitative results of this study were accepted as conclusive. In
addition, Navistar International Transportation (NITC) raised a number
of objections to the methods by which diesel exposures were estimated
for the period between 1949 and 1990 (NITC, 1999). In general, the
thrust of these objections was that exposures to diesel engine
emissions had been overestimated, while potentially relevant exposures
to gasoline engine emissions had been underestimated and/or unduly
discounted.\55\
---------------------------------------------------------------------------

    \55\ Many of the issues NITC raised in its critique of this
study depend on a peculiar identification of dpm exclusively with
elemental carbon. For example, NITC argued that ``more than 65
percent of the total carbon to which road drivers (and mechanics)
were exposed consisted of organic (i.e., non-diesel) carbon, further
suggesting that some other etiology caused or contributed to excess
lung cancer mortality in these workers.'' (NITC, 1999, p. 16) Such
lines of argument, which depend on identifying organic carbon as
``non-diesel,'' ignore the fact that dpm contains a large measure of
organic carbon compounds (and also some sulfates), as well as
elemental carbon. Any adverse health effects due to the organic
carbon or sulfate constituents of dpm would nonetheless be due to
dpm exposures.
---------------------------------------------------------------------------

    As mentioned above, the investigators recognized that these
analyses rely on ``broad assumptions rather than actual [concurrent]
measurements,'' and they proposed that the ``results should be regarded
with appropriate caution.'' While agreeing with both the investigators
and the HEI Expert Panel that these results should be interpreted with
appropriate caution, MSHA also agrees with the Panel ``* * * that
regulatory decisions need to be made in spite of the limitations and
uncertainties of the few studies with quantitative data currently
available.'' (HEI, 1999, p. 39) In this context, MSHA considers it
appropriate to regard the 1998 exposure-response analyses as
contributing to the weight of evidence that dpm exposure increases the
risk of lung cancer, even if the results are not conclusive when viewed
in isolation.
    Some commenters also noted that the HEI Expert Panel raised the
possibility that the method for selecting controls in this study could
potentially have biased the results in an unpredictable direction. Such
bias could have occurred because deaths among some of the controls were
likely due to diseases (such as cardiovascular disease) that shared
some of the same risk factors (such as tobacco smoking) with lung
cancer. The Panel presented hypothetical examples of how this might
bias results in either direction. Although the possibility of such bias
further demonstrates why the results of this study should be regarded
with ``appropriate caution,'' it is important to distinguish between
the mere possibility of a control-selection bias, evidence that such a
bias actually exists in this particular study, and the further evidence
required to show that such bias not only exists but is of sufficient
magnitude to have produced seriously misleading results. Unlike the
commenters who cited the HEI Expert Panel on this issue, the Panel
itself clearly drew this distinction, stating that ``no direct evidence
of such bias is apparent'' and emphasizing that ``even though these
examples [presented in HEI (1999), Appendix D] could produce misleading
results, it is important to note that they are only hypothetical
examples. Whether or not such bias is present will require further
examination.'' (HEI, 1999, pp. 37-38) As the HEI showed in its
examples, such bias (if it exists) could lead to underestimating the
association between lung cancer and dpm exposure, as well as to
overestimating it. Therefore, in the absence of evidence that control-
selection bias actually distorted the results of this study one way or
the other, MSHA considers it prudent to accept the study's finding of
an association at face value.
    One commenter (MARG) noted that information on cigarette smoking,
asbestos exposure, and diet in the trucking industry study was obtained
from next of kin and stated that such information was ``likely to be
unreliable.'' By increasing random variability in the data, such errors
could widen the confidence intervals around an estimated odds ratio or
reduce the confidence level at which a positive exposure-response
relationship might be established. However, unless such errors were
correlated with diesel exposure or lung cancer in such a way as to bias
the results, they would not, on average, inflate the estimated degree
of association between diesel exposure and an increased risk of ling
cancer. The commenter provided no reason to suspect that errors with
respect to these factors were in any way correlated with diesel
exposure or with the development of lung cancer.
    Some commenters pointed out that EC concentrations measured in 1990
for truck mechanics were higher, on average, than for truck drivers,
but the mechanics, unlike the drivers, showed no evidence of increasing
lung cancer risk with increasing duration of employment. NITC referred
to this as a ``discrepancy'' in the data, assuming that ``cumulative
exposure increases with duration of employment such that mechanics who
have been employed for 18 or more years would have greater cumulative
exposure than workers who have been employed for 1-11 years.'' (NITC,
1999)
    Mechanics were included in the logistic regression analyses
(Steenland et al., 1998) showing an increase in lung cancer risk with
increasing cumulative exposure. These analyses pooled the data for all
occupations by estimating exposure for each worker based on the
worker's occupation and the particular years in which the worker was
employed. There are at least three reasons why, for mechanics viewed as
a separate group, an increase in lung cancer risk with increasing dpm
exposure may not have been reflected by increasing duration of
employment.
    First, relatively few truck mechanics were available for analyzing
the relationship between length of employment and the risk of lung
cancer. Based on the union records, 50 cases and 37 controls were so
classified; based on the next-of-kin data, 43 cases and 41 controls
were more specifically classified as diesel truck mechanics (Steenland
et al., 1990). In contrast, 609 cases and 604 controls were classified
as long-haul drivers (union records). This was both the largest
occupational category and the only one showing statistically
significant evidence of increasing risk with increasing employment
duration. The number of mechanics included in the study population may
simply not have been sufficient to detect a pattern of increasing risk
with increasing length of employment, even if such a pattern existed.
    The second part of the explanation as to why mechanics did not
exhibit a pattern similar to truck drivers could be that the data on
mechanics were more subject to confounding. After noting that ``the
risk for mechanics did not appear to increase consistently with
duration of employment,'' Steenland et al. (1990) further noted that
the mechanics may have been exposed to asbestos when working on brakes.
The data used to adjust for asbestos exposure may have been inadequate
to control for variability in asbestos exposure among the mechanics.
    Third, as noted by NITC, the lung cancer risk for mechanics
(adjusted for age, race, tobacco smoking, asbestos exposure, and diet)
would be expected to increase with increasing duration of

[[Page 5808]]

employment only if the mechanics' cumulative dpm exposure corresponded
to the length of their employment. None of the commenters raising this
issue, however, provided any support for this assumption, which fails
to consider the particular calendar years in which mechanics included
in the study were employed. In compiling cumulative exposure for an
individual worker, the investigators took into account historical
changes in both diesel emissions and the proportion of trucks with
diesel engines--so the exposure level assigned to each occupational
category was not the same in each year. In general, workers included in
the study neither began nor ended their employment in the same year.
Consequently, workers with the same duration of employment in the same
occupational category could be assigned different cumulative exposures,
depending on when they were employed. Similarly, workers in the same
occupational category who were assigned the same cumulative exposure
may not have worked the same length of time in that occupation.
Therefore, it should not be assumed that duration of employment
corresponds very well to the cumulative exposure estimated for workers
within any of the occupational categories. Furthermore, in the case of
mechanics, there is an additional historical variable that is
especially relevant to actual cumulative exposure but was not
considered in formulating exposure estimates: the degree of ventilation
or other means of protection within repair shops. Historical changes in
shop design and work practices, as well as differences between shops,
may have caused more exposure misclassification among mechanics than
among long-haul or diesel truck drivers. Such misclassification would
tend to further obscure any relationship between mechanics' risk of
lung cancer and either duration of employment or cumulative exposure.
    (iv) Counter-Evidence. Several commenters stated that, in the
proposal, MSHA had dismissed or not adequately addressed epidemiology
studies showing no association between lung cancer and exposures to
diesel exhaust. For example, the EMA wrote:

    MSHA's discussion of the negative studies generally consists of
arguments to explain why those studies should be dismissed. For
example, MSHA states that, ``All of the studies showing negative or
statistically insignificant positive associations . . . lacked good
information about dpm exposure . . .'' or showed similar
shortcomings. 63 Fed. Reg. at 17533. The statement about exposure
information is only partially true, for, in fact, very few of any of
the cited studies (the ``positive'' studies as well) included any
exposure measurements, and none included concurrent exposures.

    It should, first of all, be noted that the statement in question on
dpm exposure referred to the issue of any diesel exposure--not to
quantitative exposure measurements, which MSHA acknowledges are lacking
in most of the available studies. In the absence of quantitative
measurements, however, studies comparing workers known to have been
occupationally exposed to unexposed workers are preferable to studies
not containing such comparisons. Furthermore, two of the studies now
available (and discussed above) utilize essentially concurrent exposure
measurements, and both show a positive association (Johnston et al.,
1997; Saverin et al., 1999).
    MSHA did not entirely ``dismiss'' the negative studies. They were
included in both MSHA's tabulation (see Tables III-4 and III-5) and (if
they met the inclusion criteria) in the two meta-analyses cited both
here and in the proposal (Lipsett and Campleman, 1999, and Bhatia et
al., 1998). As noted by the commenter, MSHA presented reasons (such as
an inadequate latency allowance) for why negative studies may have
failed to detect an association. Similarly MSHA gave reasons for giving
less weight to some of the positive studies, such as Benhamou et al.
(1988), Morabia et al. (1992), and Siemiatycki et al., 1988. Additional
reasons for giving less weight to the six entirely negative studies
have been tabulated above, under the heading of ``Best Available
Epidemiologic Evidence.'' The most recent of these negative studies
(Christie et al., 1994, 1995) is discussed in detail under the heading
of ``Studies Involving Miners.''
    One commenter (IMC Global) listed the following studies (all of
which MSHA had considered in the proposed risk assessment) as
``examples of studies that reported negative associations between [dpm]
exposure and lung cancer risk'':
     Waller (1981). This is one of the six negative studies
discussed earlier. Results were likely to have been biased by excluding
lung cancers occurring after retirement or resignation from employment
with the London Transit Authority. Comparison was to a general
population, and there was no adjustment for a healthy worker effect.
Comparison groups were disparate, and there was no adjustment for
possible differences in smoking frequency or intensity.
     Howe et al. (1983). Contrary to the commenter's
characterization of this study, the investigators reported
statistically significant elevations of lung cancer risk for workers
classified as ``possibly exposed'' or ``probably exposed'' to diesel
exhaust. MSHA recognizes that these results may have been confounded by
asbestos and coal dust exposures.
     Wong et al. (1985). The investigators reported a
statistically insignificant deficit for lung cancer in the entire
cohort and a statistically significant deficit for lung cancer in the
less than 5-year duration group. However, since comparisons were to a
general population, these deficits may be the result of a healthy
worker effect, for which there was no adjustment. Because of the
latency required for development of lung cancer, the result for ``less
than 5-year duration'' is far less informative than the results for
longer durations of employment and greater latency allowances. Contrary
to the commenter's characterization of this study, the investigators
reported statistically significant elevations of lung cancer risks for
``normal'' retirees (SMR = 1.30) and for ``high exposure'' dozer
operators with 15-19 years of union membership and a latency allowance
of at least 20 years (SMR = 3.43).
     Edling et al. (1987). This is one of the six negative
studies discussed earlier. The cohort consisted of only 694 bus workers
and, therefore, lacked statistical power. Furthermore, comparison was
to a general, external population with no adjustment for a healthy
worker effect.
     Garshick (1988). The reason the commenter (IMC Global)
gave for characterizing this study as negative was: ``That the sign of
the association in this data set changes based on the models used
suggests that the effect is not robust. It apparently reflects modeling
assumptions more than data.'' Contrary to the commenter's
characterization, however, the finding of increased lung cancer risk
for workers classified as diesel-exposed did not change when different
methods were used to analyze the data. What changed, depending on
modeling assumptions, was the shape and direction of the exposure-
response relationship among exposed workers (Cal-EPA, 1998; Stayner et
al., 1998; Crump, 1999; HEI, 1999). MSHA agrees that the various
exposure-response relationships that have been derived from this study
are highly sensitive to data modeling assumptions. This includes
assumptions about historical patterns of exposure, as well as
assumptions related to technical aspects of the statistical analysis.
However, as noted by the HEI Expert Panel, the study provides evidence
of a

[[Page 5809]]

positive association between exposure and lung cancer despite the
conflicting exposure-response analyses. Even though different
assumptions and methods of analysis have led to different conclusions
about the utility of this study for quantifying an exposure-response
relationship, ``the overall risk of lung cancer was elevated among
diesel-exposed workers'' (HEI, 1999, p. 25).
    Another commenter (MARG) cited a number of studies (all of which
had already been placed in the public record by MSHA) that, according
to the commenter, ``reflect either negative health effects trends among
miners or else failed to demonstrate a statistically significant
positive trend correlated with dpm exposure.'' It should be noted that,
as explained earlier, failure of an individual study to achieve
statistical significance (i.e., a high confidence level for its
results) does not necessarily prevent a study from contributing
important information to a larger body of evidence. An epidemiologic
study may fail to achieve statistical significance simply because it
did not involve a sufficient number of subjects or because it did not
allow for an adequate latency period. In addition to this general
point, the following responses apply to the specific studies cited by
the commenter.
     Ahlman et al. (1991). This study is discussed above, under
the heading of ``Studies Involving Miners.'' MSHA agrees with the
commenter that this study did not ``establish'' a relationship between
diesel exposure and the excess risk of lung cancer reported among the
miners involved. Contrary to the commenter's characterization, however,
the evidence presented by this study does incrementally point in the
direction of such a relationship. As mentioned earlier, none of the
underground miners who developed lung cancer had been occupationally
exposed to asbestos, metal work, paper pulp, or organic dusts. Based on
measurements of the alpha energy concentration at the mines, and a
comparison of smoking habits between underground and surface miners,
the authors concluded that not all of the excess lung cancer for the
underground miners was attributable to radon daughter exposures and/or
smoking. A stronger conclusion may have been possible if the cohort had
been larger.
     Ames et al. (1984). MSHA has taken account of this study,
which made no attempt to evaluate cancer effects, under the heading of
``Chronic Effects other than Cancer.'' The commenter repeated MSHA's
statement (in the proposed risk assessment) that the investigators had
not detected any association of chronic respiratory effects with diesel
exposure, but ignored MSHA's observation that the analysis had failed
to consider baseline differences in lung function or symptom
prevalence. Furthermore, as acknowledged by the investigators, diesel
exposure levels in the study population were low.
     Ames et al. (1983). As discussed later in this risk
assessment, under the heading of ``Mechanisms of Toxicity,'' this study
was among nine (out of 17) that did not find evidence of a relationship
between exposure to respirable coal mine dust and an increased risk of
lung cancer. Unlike the Australian mines studied by Christie et al.
(1995), the coal mines included in this study were not extensively
dieselized, and the investigators did not relate their findings to
diesel exposures.
     Ames et al. (1982). As noted earlier under the heading of
``Acute Health Effects,'' this study, which did not attempt to evaluate
cancer or other chronic health effects, detected no statistically
significant relationship between diesel exposure and pulmonary
function. However, the authors noted that this might have been due to
the low concentrations of diesel emissions involved.
     Armstrong et al. (1979). As discussed later in this risk
assessment, this study was among nine (out of 17) that did not find
evidence of a relationship between exposure to respirable coal mine
dust and an increased risk of lung cancer. As pointed out by the
commenter, comparisons were to a general population. Therefore, they
were subject to a healthy worker effect for which no adjustment was
made. The commenter further stated that ``diesel emissions were not
found to be related to increased health risks.'' However, diesel
emissions were not mentioned in the report, and the investigators did
not attempt to compare lung cancer rates in exposed and unexposed
miners.
     Attfield et al (1982). MSHA has taken the results of this
study into account, under the heading of ``Chronic Effects other than
Cancer.''
     Attfield (1979). MSHA has taken account of this study,
which did not attempt to evaluate cancer effects, under the heading of
``Chronic Effects other than Cancer.'' Although the results were not
conclusive at a high confidence level, miners occupationally exposed to
diesel exhaust for five or more years exhibited an increase in various
respiratory symptoms, as compared to miners exposed for less than five
years.
     Boffetta et al. (1988). This study is discussed in two
places above, under the headings ``Studies Involving Miners'' and
``Best Available Epidemiologic Evidence.'' The commenter stated that
``the study obviously does not demonstrate risks from dpm exposure.''
If the word ``demonstrate'' is taken to mean ``conclusively prove,''
then MSHA would agree that the study, viewed in isolation, does not do
this. As explained in the earlier discussion, however, MSHA considers
this study to contribute to the weight of evidence that dpm exposure
increases the risk of lung cancer.
     Costello et al. (1974). As discussed later in this risk
assessment, this study was among nine (out of 17) that did not find
evidence of a relationship between exposure to respirable coal mine
dust and an increased risk of lung cancer. Since comparisons were to a
general population, they were subject to a healthy worker effect for
which no adjustment was made. Diesel emissions were not mentioned in
the report.
     Gamble and Jones (1983). MSHA has taken account of this
study, which did not attempt to evaluate cancer effects, under the
heading of ``Chronic Effects other than Cancer.'' The commenter did not
address MSHA's observation that the method of statistical analysis used
by the investigators may have masked an association of respiratory
symptoms with diesel exposure.
     Glenn et al. (1983). As summarized by the commenter, this
report reviewed NIOSH medical surveillance on miners exposed to dpm and
found that ``* * * neither consistent nor obvious trends implicating
diesel exhaust in the mining atmosphere were revealed.'' The authors
noted that ``results were rather mixed,'' but also noted that ``levels
of diesel exhaust contaminants were generally low,'' and that ``overall
tenure in these diesel equipped mines was fairly short.'' MSHA
acknowledges the commenter's emphasis on the report's 1983 conclusion:
``further research on this subject is needed.'' However, the authors
also pointed out that ``all four of the chronic effects analyses
revealed an excess of cough and phlegm among the diesel exposed group.
In the potash, salt and trona groups, these excesses were
substantial.'' The miners included in the studies summarized by this
report would not have been exposed to dpm for sufficient time to
exhibit a possible increase in the risk of lung cancer.
     Johnston et al. (1997). This study is discussed in two
places above, under the headings ``Studies Involving Miners'' and
``Best Available Epidemiologic Evidence.'' MSHA disagrees with the
commenter's

[[Page 5810]]

assertion that ``the study does not support a health risk from dpm.''
This was not the conclusion drawn by the authors of the study. As
explained in the earlier discussion, this study, one of the few
containing quantitative estimates of cumulative dpm exposures, provides
evidence of increasing lung cancer risk with increasing exposure.
     Jorgenson and Svensson (1970). MSHA discussed this study,
which did not attempt to evaluate cancer effects, under the heading of
``Chronic Effects other than Cancer.'' Contrary to the commenter's
characterization, the investigators reported higher rates of chronic
productive bronchitis, for both smokers and nonsmokers, among the
underground iron ore miners exposed to diesel exhaust as compared to
surface workers at the same mine.
     Kuempel (1995); Lidell (1973); Miller and Jacobsen (1985).
As discussed later in this risk assessment, under the heading of
``Mechanisms of Toxicity,'' these three studies were among the nine
(out of 17) that did not find evidence of a relationship between
exposure to respirable coal mine dust and an increased risk of lung
cancer. The extent, if any, to which workers involved in these studies
were occupationally exposed to diesel emissions was not documented, and
diesel emissions were not mentioned in any of these reports.
     Morfeld et al. (1997). The commenter's summary of this
study distorted the investigators' conclusions. Contrary to the
commenter's characterization, this is one of eight studies that showed
an increased risk of lung cancer for coal miners, as discussed later in
this risk assessment under the heading of ``Mechanisms of Toxicity.''
For lung cancer, the relative SMR, which adjusts for the healthy worker
effect, was 1.11. (The value of 0.70 cited by the commenter was the
unadjusted SMR.) The authors acknowledged that the relative SMR
obtained by the ``standard analysis'' (i.e., 1.11) was not
statistically significant. However, the main object of the report was
to demonstrate that the ``standard analysis'' is insufficient. The
investigators presented evidence that the 1.11 value was biased
downward by a ``healthy-worker-survivor-effect,'' thereby masking the
actual exposure effects in these workers. They found that ``all the
evidence points to the conclusion that a standard analysis suffers from
a severe underestimate of the exposure effect on overall mortality,
cancer mortality and lung cancer mortality.'' (Morfeld et al., 1997, p.
350)
     Reger (1982). MSHA has taken account of this study, which
made no attempt to evaluate cancer effects, under the heading of
``Chronic Effects other than Cancer.'' As summarized by the commenter,
``diesel-exposed miners were found to have more cough and phlegm, and
lower pulmonary function,'' but the author found that ``the evidence
would not allow for the rejection of the hypothesis of health equality
between exposed and non-exposed miners.'' The commenter failed to note,
however, that miners in the dieselized mines, had worked underground
for less than 5 years on average.
     Rockette (1977). This is one of eight studies, discussed
under ``Mechanisms of Toxicity,'' showing an increased risk of lung
cancer for coal miners. As described by the commenter, the author
reported SMRs of 1.12 for respiratory cancers and 1.40 for stomach
cancer. MSHA agrees with the commenter that ``the study does not
establish a dpm-related health risk,'' but notes that dpm effects were
not under investigation. Diesel emissions were not mentioned in the
report, and, given the study period, the miners involved may not have
been occupationally exposed to diesel exhaust.
     Waxweiler (1972). MSHA's discussion of this study appears
earlier in this risk assessment, under ``Studies Involving Miners.'' As
noted by the commenter, the slight excess in lung cancer, relative to
the general population of New Mexico, was not statistically
significant. The commenter failed to note, however, that no adjustment
was made for a healthy worker effect and that a substantial percentage
of the underground miners were not occupationally exposed to diesel
emissions.
    Summation. Limitations identified in both positive and negative
studies include: lack of sufficient power, inappropriate comparison
groups, exposure misclassification, statistically insignificant
results, and potential confounders. As explained earlier, under
``Evaluation Criteria,'' weaknesses of the first three of these types
can reasonably be expected, for the most part, to artificially decrease
the apparent strength of any observed association between diesel
exposure and increased risk of lung cancer. Statistical insignificance
and potential confounders may, in the absence of evidence to the
contrary, be regarded as neutral on average. The weaknesses that have
been identified in these studies are not unique to epidemiologic
studies involving lung cancer and diesel exhaust. They are sources of
uncertainty in virtually all epidemiologic research.
    Even when there is a strong possibility that the results of a study
have been affected by confounding variables, it does not follow that
the effect has been to inflate rather than deflate the results or that
the study cannot contribute to the weight of evidence supporting a
putative association. As cogently stated by Stober and Abel (op cit.,
p. 4), ``* * * associations found in epidemiologic studies can always
be, at least in part, attributed to confounding.'' Therefore, an
objection grounded on potential confounding can always be raised
against any epidemiologic study. It is well known that this same
objection was, in the past, raised against epidemiologic studies
linking lung cancer and radon exposure, lung cancer and asbestos dust
exposure, and even lung cancer and tobacco smoking.
    Some commenters, have now proposed that virtually every existing
epidemiologic study relating lung cancer to dpm exposure be summarily
discredited because of susceptibility to confounding or other perceived
weaknesses. Given the practical difficulties of designing and executing
an epidemiologic study, this is not so much an objection to any
specific study as it is an attack on applied epidemiology in general.
Indeed, in their review of these studies, Stober and Abel (1996)
conclude that

    In this field * * * epidemiology faces its limits (Taubes,
1995). * * * Many of these studies were doomed to failure from the
very beginning.

    For important ethical reasons, however, tightly controlled lung
cancer experiments cannot be performed on humans. Therefore, despite
their inherent limitations, MSHA must rely on the weight of evidence
from epidemiologic studies, placing greatest weight on the most
carefully designed and executed studies available.
(b) Bladder Cancer
    With respect to cancers other than lung cancer, MSHA's review of
the literature identified only bladder cancer as a possible candidate
for a causal link to dpm. Cohen and Higgins (1995) identified and
reviewed 14 epidemiologic case-control studies containing information
related to dpm exposure and bladder cancer. All but one of these
studies found elevated risks of bladder cancer among workers in jobs
frequently associated with dpm exposure. Findings were statistically
significant in at least four of the studies (statistical significance
was not evaluated in three).

[[Page 5811]]

    These studies point quite consistently toward an excess risk of
bladder cancer among truck or bus drivers, railroad workers, and
vehicle mechanics. However, the four available cohort studies do not
support a conclusion that exposure to dpm is responsible for the excess
risk of bladder cancer associated with these occupations. Furthermore,
most of the case-control studies did not distinguish between exposure
to diesel-powered equipment and exposure to gasoline-powered equipment
for workers having the same occupation. When such a distinction was
drawn, there was no evidence that the prevalence of bladder cancer was
higher for workers exposed to the diesel-powered equipment.
    This, along with the lack of corroboration from existing cohort
studies, suggests that the excessive rates of bladder cancer observed
may be a consequence of factors other than dpm exposure that are also
associated with these occupations. For example, truck and bus drivers
are subjected to vibrations while driving and may tend to have
different dietary and sleeping habits than the general population. For
these reasons, MSHA does not find that convincing evidence currently
exists for a causal relationship between dpm exposure and bladder
cancer. MSHA received no public comments objecting to this conclusion.
    ii. Studies Based on Exposures to PM2.5 in Ambient Air.
Prior to 1990, the relationship between mortality and long-term
exposure to particulate matter was generally investigated by means of
cross-sectional studies, but unaddressed spatial confounders and other
methodological problems inherent in such studies limited their
usefulness (EPA, 1996).\56\ Two more recent prospective cohort studies
provide better evidence of a link between excess mortality rates and
exposure to fine particulate, although some of the uncertainties here
are greater than with the short-term studies conducted in single
communities. The two studies are the ``Six Cities'' study (Dockery et
al., 1993), and the American Cancer Society (ACS) study (Pope et al.,
1995).\57\ The first study followed about 8,000 adults in six U.S.
cities over 14 years; the second looked at survival data for half a
million adults in 151 U.S. cities for 7 years. After adjusting for
potential confounders, including smoking habits, the studies considered
differences in mortality rates between the most polluted and least
polluted cities.
---------------------------------------------------------------------------

    \56\ Unlike longitudinal studies, which examine responses at
given locations to changes in conditions over time, cross-sectional
studies compare results from locations with different conditions at
a given point in time.
    \57\ A third such study, the California Seventh Day Adventist
study (Abbey et al., 1991), investigated only TSP, rather than fine
particulate. It did not find significant excess mortality associated
with chronic TSP exposures.
---------------------------------------------------------------------------

    Both the Six Cities study and the ACS study found a significant
association between chronically higher concentrations of
PM2.5 (which includes dpm) and age-adjusted total
mortality.\58\ The authors of the Six Cities Study concluded that the
results suggest that exposures to fine particulate air pollution
``contributes to excess mortality in certain U.S. cities.'' The ACS
study, which not only controlled for smoking habits and various
occupational exposures, but also, to some extent, for passive exposure
to tobacco smoke, found results qualitatively consistent with those of
the Six Cities Study.\59\ In the ACS study, however, the estimated
increase in mortality associated with a given increase in fine
particulate exposure was lower, though still statistically significant.
In both studies, the largest increase observed was for cardiopulmonary
mortality.
---------------------------------------------------------------------------

    \58\ The Six Cities study also found such relationships at
elevated levels of PM10 and sulfates. The ACS study was
designed to follow up on the fine particle results of the Six Cities
Study, and also investigated sulfates separately. As explained
earlier in this preamble, sulfates may be a significant constituent
of dpm, depending on the type of diesel fuel used.
    \59\ The Six Cities study did not find a statistically
significant increase in risk among non-smokers, suggesting that non-
smokers might be less sensitive than smokers to adverse health
effects from fine particulate exposures; however, the ACS study,
with more statistical power, did find significantly increased risk
even for non-smokers.
---------------------------------------------------------------------------

    Both studies also showed an increased risk of lung cancer
associated with increased exposure to fine particulate. Although the
lung cancer results were not statistically significant, they are
consistent with reports of an increased risk of lung cancer among
workers occupationally exposed to diesel emissions (discussed above).
    The few studies on associations between chronic PM2.5
exposure and morbidity in adults show effects that are difficult to
separate from measures of PM10 and measures of acid
aerosols. The available studies, however, show positive associations
between particulate air pollution and adverse health effects for those
with pre-existing respiratory or cardiovascular disease. This is
significant for miners occupationally exposed to fine particulates such
as dpm because, as mentioned earlier, there is a large body of evidence
showing that respiratory diseases classified as COPD are significantly
more prevalent among miners than in the general population. It also
appears that PM exposure may exacerbate existing respiratory infections
and asthma, increasing the risk of severe outcomes in individuals who
have such conditions (EPA, 1996).
d. Mechanisms of Toxicity
    Four topics will be addressed in this section of the risk
assessment: (i) the agent of toxicity, (ii) clearance and deposition of
dpm, (iii) effects other than cancer, and (iv) lung cancer. The section
on lung cancer will include discussions of the evidence from (1)
genotoxicity studies (including bioavailability of genotoxins) and (2)
animal studies.
    i. Agent of Toxicity. As described in Part II of this preamble, the
particulate fraction of diesel exhaust is made up of aggregated soot
particles, vapor phase hydrocarbons, and sulfates. Each soot particle
consists of an insoluble, elemental carbon core and an adsorbed,
surface coating of relatively soluble organic compounds, such as
polycyclic aromatic hydrocarbons (PAHs). Many of these organic carbon
compounds are suspected or known mutagens and/or carcinogens. For
example, nitrated PAHs, which are present in dpm, are potent mutagens
in microbial and human cell systems, and some are known to be
carcinogenic to animals (IPCS, 1996, pp. 100-105).
    When released into an atmosphere, the soot particles formed during
combustion tend to aggregate into larger particles. The total organic
and elemental carbon in these soot particles accounts for approximately
80 percent of the dpm mass. The remaining 20 percent consists mainly of
sulfates, such as H2SO4 (sulfuric acid).
    Several laboratory animal studies have been performed to ascertain
whether the effects of diesel exhaust are attributable specifically to
the particulate fraction. (Heinrich et al., 1986, 1995; Iwai et al.,
1986; Brightwell et al., 1986). These studies compare the effects of
chronic exposure to whole diesel exhaust with the effects of filtered
exhaust containing no particles. The studies demonstrate that when the
exhaust is sufficiently diluted to nullify the effects of gaseous
irritants (NO2 and SO2), irritant vapors
(aldehydes), CO, and other systemic toxicants, diesel particles are the
prime etiologic agents of noncancer health effects. Exposure to dpm
produced changes in the lung that were much more prominent than those
evoked by the gaseous fraction alone. Marked differences in the effects
of whole and filtered diesel exhaust were also evident from general
toxicological

[[Page 5812]]

indices, such as body weight, lung weight, and pulmonary
histopathology.
    These studies show that, when the exhaust is sufficiently diluted,
it is the particles that are primarily responsible for the toxicity
observed. However, the available studies do not completely settle the
question of whether the particles might act additively or
synergistically with the gases in diesel exhaust. Possible additivity
or interaction effects with the gaseous portion of diesel exhaust
cannot be completely ruled out.
    One commenter (MARG) raised an issue with regard to the agent of
toxicity in diesel exhaust as follows:

    MSHA has not attempted to regulate exposure to suspected
carcinogens contained in dpm, but has opted instead, in metal/non-
metal mines, to regulate total carbon (``TC'') as a surrogate for
diesel exhaust, without any evidence of adverse health effects from
TC exposure. * * * Nor does the mere presence of suspected
carcinogens, in minute quantities, in diesel exhaust require a 95
percent reduction of total diesel exhaust [sic] in coal mines. If
there are small amounts of carcinogenic substances of concern in
diesel exhaust, those substances, not TC, should be regulated
directly on the basis of the risks (if any) posed by those
substances in the quantities actually present in underground mines.
[MARG]

    First, it should be noted that the ``suspected carcinogens'' in
diesel exhaust to which the commenter referred are part of the organic
fraction of the total carbon. Therefore, limiting the concentration of
airborne total carbon attributable to dpm, or removing the soot
particles from the diesel exhaust by filtration, are both ways of
effectively limiting exposures to these suspected carcinogens. Second,
the commenter seems to have assumed that cancer is the only adverse
health effect of concern and that the only agents in dpm that could
cause cancer are the ``suspected carcinogens'' in the organic fraction.
This not only ignores non-cancer health effects associated with
exposures to dpm and other fine particles, but also the possibility
(discussed below) that, with sufficient deposition and retention, soot
particles themselves could promote or otherwise increase the risk of
lung cancer--either directly or by stimulating the body's natural
defenses against foreign substances.
    The same commenter [MARG] also stated that ``* * * airborne carbon
has not been shown to be harmful at levels currently established in
MSHA's dust rules. If the problem is dpm, as MSHA asserts, then it is
not rationally addressed by regulating airborne carbon.'' MSHA's intent
is to limit dpm exposures in M/NM mines by regulating the submicrometer
carbon from diesel emissions--not any and all airborne carbon. MSHA
considers its approach a rational means of limiting dpm exposures
because most of the dpm consists of carbon (approximately 80 percent by
weight), and because using low sulfur diesel fuel will effectively
reduce the sulfates comprising most of the remaining portion. The
commenter offered no practical suggestion of a more direct, effective,
and rational way of limiting airborne dpm concentrations in M/NM mines.
Furthermore, direct evidence exists that the risk of lung cancer
increases with increasing cumulative occupational exposure to dpm as
measured by total carbon (Saverin et al., 1999, discussed earlier in
this risk assessment).
    ii. Deposition, Clearance, and Retention. As suggested by Figure
II-1 of this preamble, most of the aggregated particles making up dpm
are no larger than one micrometer in diameter. Particles this small are
able to penetrate into the deepest regions of the lungs, called
alveoli. In the alveoli, the particles can mix with and be dispersed by
a substance called surfactant, which is secreted by cells lining the
alveolar surfaces.
    The literature on deposition of fine particles in the respiratory
tract was reviewed in Green and Watson (1995) and U.S. EPA (1996). The
mechanisms responsible for the broad range of potential particle-
related health effects varies depending on the site of deposition. Once
deposited, the particles may be cleared from the lung, translocated
into the interstitium, sequestered in the lymph nodes, metabolized, or
be otherwise chemically or physically changed by various mechanisms.
Clearance of dpm from the alveoli is important in the long-term effects
of the particles on cells, since it may be more than two orders of
magnitude slower than mucociliary clearance (IPCS, 1996).
    IARC (1989) and IPCS (1996) reviewed factors affecting the
deposition and clearance of dpm in the respiratory tracts of
experimental animals. Inhaled PAHs adhering to the carbon core of dpm
are cleared from the lung at a significantly slower rate than
unattached PAHs. Furthermore, there is evidence that inhalation of
whole dpm may increase the retention of subsequently inhaled PAHs. IARC
(op cit.) suggested that this can happen when newly introduced PAHs
bind to dpm particles that have been retained in the lung.
    The evidence points to significant differences in deposition and
clearance for different animal species (IPCS, 1996). Under equivalent
exposure regimens, hamsters exhibited lower levels of retained dpm in
their lungs than rats or mice and consequently less pulmonary function
impairment and pulmonary pathology. These differences may result from a
lower intake rate of dpm, lower deposition rate and/or more rapid
clearance rate, or lung tissue that is less susceptible to the
cytotoxicity of dpm. Observations of a decreased respiration in
hamsters when exposed by inhalation favor lower intake and deposition
rates.
    Retardation of lung clearance, called ``overload'' is not specific
to dpm and may be caused by inhaling, at a sufficiently high rate, dpm
in combination with other respirable particles, such as mineral dusts
typical of mining environments. The effect is characterized by (1) an
overwhelming of normal clearance processes, (2) disproportionately high
retention and loading of the lung with particles, compared to what
occurs at lower particle inhalation rates, (3) various pathological
responses; generally including chronic inflammation, epithelial
hyperplasia and metaplasia, and pulmonary fibrosis; and sometimes
including lung tumors.
    In the proposed risk assessment, MSHA requested additional
information, not already covered in the sources cited above, on fine
particle deposition in the respiratory tract, especially as it might
pertain to lung loading in miners exposed to a combination of diesel
particulate and other dusts. In response to this request, NIOSH
submitted a study that investigated rat lung responses to chronic
inhalation of a combination of coal dust and diesel exhaust, compared
to coal dust or dpm alone (Castranova et al., 1985). Although this
report did not directly address deposition or clearance, the
investigators reported that another phase of the study had shown that
``particulate clearance, as determined by particulate accumulation in
the lung, is inhibited after two years of exposure to diesel exhaust
but is not inhibited by exposure to coal dust.''
    iii. Effects other than Cancer. A number of controlled animal
studies have been undertaken to ascertain the toxic effects of exposure
to diesel exhaust and its components. Watson and Green (1995) reviewed
approximately 50 reports describing noncancerous effects in animals
resulting from the inhalation of diesel exhaust. While most of the
studies were conducted with rats or hamsters, some information was also
available from studies conducted using cats, guinea pigs, and monkeys.
The authors also

[[Page 5813]]

correlated reported effects with different descriptors of dose,
including both gravimetric and non-gravimetric (e.g., particle surface
area or volume) measures. From their review of these studies, Watson
and Green concluded that:
    (a) Animals exposed to diesel exhaust exhibit a number of
noncancerous pulmonary effects, including chronic inflammation,
epithelial cell hyperplasia, metaplasia, alterations in connective
tissue, pulmonary fibrosis, and compromised pulmonary function.
    (b) Cumulative weekly exposure to diesel exhaust of 70 to 80
mg hr/m3 or greater are associated with the presence
of chronic inflammation, epithelial cell proliferation, and depressed
alveolar clearance in chronically exposed rats.
    (c) The extrapolation of responses in animals to noncancer
endpoints in humans is uncertain. Rats were the most sensitive animal
species studied.
    Subsequent to the review by Watson and Green, there have been a
number of animal studies on allergic immune responses to dpm. Takano et
al. (1997) investigated the effects of dpm injected into mice through
an intratracheal tube and found manifestations of allergic asthma,
including enhanced antigen-induced airway inflammation, increased local
expression of cytokine proteins, and increased production of antigen-
specific immunoglobulins. The authors concluded that the study
demonstrated dpm's enhancing effects on allergic asthma and that the
results suggest that dpm is ``implicated in the increasing prevalence
of allergic asthma in recent years.'' Similarly, Ichinose et al.
(1997a) found that five different strains of mice injected
intratracheally with dpm exhibited manifestations of allergic asthma,
as expressed by enhanced airway inflammation, which were correlated
with an increased production of antigen-specific immunoglobulin due to
the dpm. The authors concluded that dpm enhances manifestations of
allergic airway inflammation and that ``* * * the cause of individual
differences in humans at the onset of allergic asthma may be related to
differences in antigen-induced immune responses * * *.''
    The mechanisms that may lead to adverse health effects in humans
from inhaling fine particulates are not fully understood, but potential
mechanisms that have been hypothesized for non-cancerous outcomes are
summarized in Table III-6. A comprehensive review of the toxicity
literature is provided in U.S. EPA (1996).

BILLING CODE 4510-43-P

[[Page 5814]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.067

[[Page 5815]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.068

BILLING CODE 4510-43-C

[[Page 5816]]

    Deposition of particulates in the human respiratory tract may
initiate events leading to increased airflow obstruction, impaired
clearance, impaired host defenses, or increased epithelial
permeability. Airflow obstruction can result from laryngeal
constriction or bronchoconstriction secondary to stimulation of
receptors in extrathoracic or intrathoracic airways. In addition to
reflex airway narrowing, reflex or local stimulation of mucus secretion
can lead to mucus hypersecretion and, eventually, to mucus plugging in
small airways.
    Pulmonary changes that contribute to cardiovascular responses
include a variety of mechanisms that can lead to hypoxemia, including
bronchoconstriction, apnea, impaired diffusion, and production of
inflammatory mediators. Hypoxia can lead to cardiac arrhythmias and
other cardiac electrophysiologic responses that, in turn, may lead to
ventricular fibrillation and ultimately cardiac arrest. Furthermore,
many respiratory receptors have direct cardiovascular effects. For
example, stimulation of C-fibers leads to bradycardia and hypertension,
and stimulation of laryngeal receptors can result in hypertension,
cardiac arrhythmia, bradycardia, apnea, and even cardiac arrest. Nasal
receptor or pulmonary J-receptor stimulation can lead to vagally-
mediated bradycardia and hypertension (Widdicombe, 1988).
    Some commenters mistakenly attributed the sensory irritant effects
of diesel exhaust entirely to its gaseous components. The mechanism by
which constituents of dpm can cause sensory irritations in humans is
much better understood than the mechanisms for other adverse health
effects due to fine particulates. In essence, sensory irritants are
``scrubbed'' from air entering the upper respiratory tract, thereby
preventing a portion from penetrating more deeply into the lower
respiratory tract. However, the sensory irritants stimulate trigeminal
nerve endings, which are located very close to the oro-nasal mucosa and
also to the watery surfaces of the eye (cornea). This produces a
burning, painful sensation. The intensity of the sensory irritant
response is related to the irritant concentration and duration of
exposure. Differences in relative potency are observed with different
sensory irritants. Acrolein and formaldehyde are examples of highly
potent sensory irritants which, along with others having low molecular
weights (acids, aldehydes), are often found in the organic fraction of
dpm (Nauss et al., 1995). They may be adsorbed onto the carbon-based
core or released in a vapor phase. Thus, mixtures of sensory irritants
in dpm may impinge upon the eyes and respiratory tract of miners and
produce adverse health effects.
    It is also important to note that mixtures of sensory irritants in
dpm may produce responses that are not predicted solely on the basis of
the individual chemical constituents. Instead, these irritants may
interact at receptor sites to produce additive, synergistic, or
antagonistic effects. For example, because of synergism, dpm containing
a mixture of sensory irritants at relatively low concentrations may
produce intense sensory responses (i.e., responses far above those
expected for the individual irritants). Therefore, the irritant effects
of whole dpm cannot properly be evaluated by simply adding together the
known effects of its individual components.
    As part of its public comments on the proposed preamble, NIOSH
submitted a study (Hahon et al., 1985) on the effects of diesel
emissions on mice infected with influenza virus. The object of this
study was to determine if exposure to diesel emissions (either alone or
in combination with coal dust) could affect resistance to pulmonary
infections. The investigators exposed groups of mice to either coal
dust, diesel emissions, a combination of both, or filtered air (control
group) for various durations, after which they were infected with
influenza. Although not reflected by excess mortality, the severity of
influenza infection was found to be more pronounced in mice previously
exposed to diesel emissions than in control animals. The effect was not
intensified by inhalation of coal dust in combination with those
emissions.
    In addition to possible acute toxicity of particles in the
respiratory tract, chronic exposure to particles that deposit in the
lung may induce inflammation. Inflammatory responses can lead to
increased permeability and possibly diffusion abnormality. Furthermore,
mediators released during an inflammatory response could cause release
of factors in the clotting cascade that may lead to an increased risk
of thrombus formation in the vascular system (Seaton, 1995). Persistent
inflammation, or repeated cycles of acute lung injury and healing, can
induce chronic lung injury. Retention of the particles may be
associated with the initiation and/or progression of COPD.
    Takenaka et al. (1995) investigated mechanisms by which dpm may act
to cause allergenic effects in human cell cultures. The investigators
reported that application of organic dpm extracts over a period of 10
to 14 days increased IgE production from the cells by a factor of up to
360 percent. They concluded that enhanced IgE production in the human
airway resulting from the organic fraction of dpm may be an important
factor in the increasing incidence of allergic airway disease.
Similarly, Tsien et al. (1997) investigated the effects of the organic
fraction of dpm on IgE production in human cell cultures and found that
application of the organic extract doubled IgE production after three
days in cells already producing IgE.
    Sagai et al. (1996) investigated the potential role of dpm-induced
oxygen radicals in causing pulmonary injuries. Repeated intratracheal
instillation of dpm in mice caused marked infiltration of inflammatory
cells, proliferation of goblet cells, increased mucus secretion,
respiratory resistance, and airway constriction. The results indicated
that oxygen radicals, induced by intratracheally instilled dpm, can
cause responses characteristic of bronchial asthma.
    Lovik et al. (1997) investigated inflammatory and systemic IgE
responses to dpm, alone and in combination with the model allergen
ovalbumin (OA), in mice. To determine whether it was the elemental
carbon core or substances in the organic fraction of dpm that were
responsible for observed allergenic effects, they compared the effects
of whole dpm with those of carbon black (CB) particles of comparable
size and specific surface area. Although the effects were slightly
greater for dpm, both dpm and CB were found to cause significant,
synergistic increases in allergenic responses to the OA, as expressed
by inflammatory responses of the local lymph node and OA-specific IgE
production. The investigators concluded that both dpm and CB
synergistically enhance and prolong inflammatory responses in the lymph
nodes that drain the site of allergen deposition. They further
concluded that the elemental carbon core contributes substantially to
the adjuvant activity of dpm.
    Diaz-Sanchez et al. (1994, 1996, 1997) conducted a series of
experiments on human subjects to investigate the effects of dpm on
allergic inflammation as measured by IgE production. The studies by
Takenaka et al. (op cit.) and Tsien et al. (op cit.) were also part of
this series but were based on human cell cultures rather than live
human volunteers. A principal objective of these experiments was to
investigate the pathways and mechanisms by which dpm induces allergic
inflammation. The investigators found that the organic fraction of dpm
can enhance IgE production, but that the major

[[Page 5817]]

polyaromatic hydrocarbon in this fraction (phenanthrene) can enhance
IgE without causing inflammation. On the other hand, when human
volunteers were sprayed intranasally with carbon particles lacking the
organic compounds, the investigators found a large influx of cells in
the nasal mucosa but no increase in IgE. These results suggest that
while the organic portion of dpm is not necessary for causing
irritation and local inflammation, it is the organic compounds that act
on the immune system to promote an allergic response.
    Salvi et al. (1999) investigated the impact of diesel exhaust on
human airways and peripheral blood by exposing healthy volunteers to
diesel exhaust at a concentration of 300 g/m\3\ for one hour
with intermittent exercise. Following exposure, they found significant
evidence of acute inflammatory responses in airway lavage and also in
the peripheral blood. Some commenters expressed a belief that the
gaseous, rather than particulate, components of diesel exhaust caused
these effects. The investigators noted that the inflammatory responses
observed could not be attributed to NO2 in the diesel
exhaust because previous studies they had conducted, using a similar
experimental protocol, had revealed no such responses in the airway
tissues of volunteers exposed to a higher concentration of
NO2, for a longer duration, in the absence of dpm. They
concluded that ``[i]t therefore seems more likely that the particulate
component of DE is responsible.''
    iv. Lung Cancer. (1) Genotoxicity Studies. Many studies have shown
that diesel soot, or its organic component, can increase the likelihood
of genetic mutations during the biological process of cell division and
replication. A survey of the applicable scientific literature is
provided in Shirname-More (1995). What makes this body of research
relevant to the risk of lung cancer is that mutations in critical genes
can sometimes initiate, promote, or advance a process of
carcinogenesis.
    The determination of genotoxicity has frequently been made by
treating diesel soot with organic solvents such as dichloromethane and
dimethyl sulfoxide. The solvent removes the organic compounds from the
carbon core. After the solvent evaporates, the mutagenic potential of
the extracted organic material is tested by applying it to bacterial,
mammalian, or human cells propagated in a laboratory culture. In
general, the results of these studies have shown that various
components of the organic material can induce mutations and chromosomal
aberrations.
    One commenter (MARG) pointed out that ``even assuming diesel
exhaust contains particular genotoxic substances, the bioavailability
of these genotoxins has been questioned.'' As acknowledged in the
proposed risk assessment, a critical issue is whether whole diesel
particulate is mutagenic when dispersed by substances present in the
lung. Since the laboratory procedure for extracting organic material
with solvents bears little resemblance to the physiological environment
of the lung, it is important to establish whether dpm as a whole is
genotoxic, without solvent extraction. Early research indicated that
this was not the case and, therefore, that the active genotoxic
materials adhering to the carbon core of diesel particles might not be
biologically damaging or even available to cells in the lung (Brooks et
al., 1980; King et al., 1981; Siak et al., 1981). A number of more
recent research papers, however, have shown that dpm, without solvent
extraction, can cause DNA damage when the soot is dispersed in the
pulmonary surfactant that coats the surface of the alveoli (Wallace et
al., 1987; Keane et al., 1991; Gu et al., 1991; Gu et al., 1992). From
these studies, NIOSH concluded in 1992 that:

    * * * the solvent extract of diesel soot and the surfactant
dispersion of diesel soot particles were found to be active in
procaryotic cell and eukaryotic cell in vitro genotoxicity assays.
The cited data indicate that respired diesel soot particles on the
surface of the lung alveoli and respiratory bronchioles can be
dispersed in the surfactant-rich aqueous phase lining the surfaces,
and that genotoxic material associated with such dispersed soot
particles is biologically available and genotoxically active.
Therefore, this research demonstrates the biological availability of
active genotoxic materials without organic solvent interaction.
[Cover letter to NIOSH response to ANPRM, 1992].

If this conclusion is correct, it follows that dpm itself, and not only
its organic extract, can cause genetic mutations when dispersed by a
substance present in the lung.
    One commenter (IMC Global) noted that Wallace et al. (1987) used
aged dpm samples from scrapings inside an exhaust pipe and contended
that this was not a realistic representation of dpm. The commenter
further argued that the two studies cited by Gu et al. involved
``direct application of an unusually high concentration gradient'' that
does not replicate normal conditions of dpm exposure.
    MSHA agrees with this commenter's general point that conditions set
up in such experiments do not duplicate actual exposure conditions.
However, as a follow-up to the Wallace study, Keane et al. (op. cit.)
demonstrated similar results with both exhaust pipe soot and particles
obtained directly from an exhaust stream. With regard to the two Gu
studies, MSHA recognizes that any well-controlled experiment serves
only a limited purpose. Despite their limitations, however, these
experiments provided valuable information. They avoided solvent
extraction. By showing that solvent extraction is not a necessary
condition of dpm mutagenicity, these studies provided incremental
support to the hypothesis of bioavailability under more realistic
conditions. This possibility was subsequently tested by a variety of
other experiments, including experiments on live animals and humans.
    For example, Sagai et al. (1993) showed that whole dpm produced
active oxygen radicals in the trachea of live mice, but that dpm
stripped of organic compounds did not. Whole dpm caused significant
damage to the lungs and also high mortality at low doses. According to
the investigators, most of the toxicity observed appeared to be due to
the oxygen radicals, which can also have genotoxic effects.
Subsequently, Ichinose et al. (1997b) examined the relationship between
tumor response and the formation of oxygen radicals in the lungs of
mice injected with dpm. The mice were treated with sufficiently high
doses of dpm to produce tumors after 12 months. As in the earlier
study, the investigators found that the dpm generated oxygen radicals,
even in the absence of biologically activating systems (such as
macrophages), and that these oxygen radicals were implicated in the
lung toxicity of the dpm. The authors concluded that ``oxidative DNA
damage induced by the repeated DEP [i.e., dpm] treatment could be an
important factor in enhancing the mutation rate leading to lung
cancer.''
    The formation of DNA adducts is an important indicator of
genotoxicity and potential carcinogenicity. Adduct formation occurs
when molecules, such as those in dpm, attach to the cellular DNA. These
adducts can negatively affect DNA transcription and/or cellular
duplication. If DNA adducts are not repaired, then a mutation or
chromosomal aberration can occur during normal mitosis (i.e., cell
replication) eventually leading to cancer cell formation. IPCS (1996)
contains a survey of animal experiments showing DNA adduct induction in
the lungs of experimental animals exposed to diesel

[[Page 5818]]

exhaust.\60\ MSHA recognizes that such studies provide limited
information regarding the bioavailability of organics, since positive
results may well have been related to factors associated with lung
particle overload. However, the bioavailability of genotoxic dpm
components is also supported by human studies showing genotoxic effects
of exposure to whole dpm. DNA adduct formation and/or mutations in
blood cells following exposure to dpm, especially at levels
insufficient to induce lung overload, can be presumed to result from
organics diffusing into the blood.
---------------------------------------------------------------------------

    \60\ Some of these studies will be discussed in the next
subsection of this risk assessment.
---------------------------------------------------------------------------

    Hemminki et al. (1994) found that DNA adducts were significantly
elevated in lymphocytes of nonsmoking bus maintenance and truck
terminal workers, as compared to a control group of hospital mechanics,
with the highest adduct levels found among garage and forklift workers.
Hou et al. (1995) reported significantly elevated levels of DNA adducts
in lymphocytes of non-smoking diesel bus maintenance workers compared
to a control group of unexposed workers. Similarly, Nielsen et al.
(1996) found that DNA adducts were significantly increased in the blood
and urine of bus garage workers and mechanics exposed to dpm as
compared to a control group.
    One commenter (IMC Global) acknowledged that ``the studies
conducted by Hemminiki [Hemminiki et al., 1994] showed elevations in
lymphocyte DNA adducts in garage workers, bus maintenance workers and
diesel forklift drivers'' but argued that ``these elevations were at
the borderline of statistical significance.'' Although results at a
higher level of confidence would have been more persuasive, this does
not negate the value of the evidence as it stands. Furthermore,
statistical significance in an individual study becomes less of an
issue when, as in this case, the results are corroborated by other
studies.
    IMC Global also acknowledged that the Nielsen study found
significant differences in DNA adduct formation between diesel-exposed
workers and controls but argued that ``the real source of genotoxins
was unclear, and other sources of exposure, such as skin contact with
lubricating oils could not be excluded.'' As is generally the case with
studies involving human subjects, this study did not completely control
for potential confounders. For this reason, MSHA considers it important
that several human studies--not all subject to confounding by the same
variables--found elevated adduct levels in diesel-exposed workers.
    IMC Global cited another human study (Qu et al., 1997) as casting
doubt on the genotoxic effects of diesel exposure, even though this
study (conducted on Australian coal miners) reported significant
increases in DNA adducts immediately after a period of intense diesel
exposure during a longwall move. As noted by the commenter, adduct
levels of exposed miners and drivers were, prior to the longwall move,
approximately 50% higher than for the unexposed control group; but
differences by exposure category were not statistically significant. A
more informative part of the study, however, consisted of comparing
adducts in the same workers before and after a longwall move, which
involved ``intensive use of heavy equipment, diesel powered in these
mines, over a 2-3 week period.'' MSHA emphasizes that the comparison
was made on the same workers, because doing so largely controlled for
potentially confounding variables, such as smoking habits, that may be
a factor when making comparisons between different persons. After the
period of ``intensive'' exposure, statistically significant increases
were observed in both total and individual adducts. Contrary to the
commenter's characterization of this study, the investigators stated
that their analysis ``provides results in which the authors have a high
level of confidence.'' They concluded that ``given the * * * apparent
increase in adducts during a period of intense DEE [i.e., diesel
exhaust emissions] exposures it would be prudent to pay particular
attention to keeping exposures as low as possible, especially during
LWCO [i.e., `longwall change out'] operations.'' Although the commenter
submitted this study as counter-evidence, it actually provides
significant, positive evidence that high dpm exposures in a mining
environment can produce genotoxic effects.
    The West Virginia Coal Association submitted an analysis by Dr.
Peter Valberg, purporting to show that ``* * * the quantity of
particle-bound mutagens that could potentially contact lung cells under
human exposure scenarios is very small.'' According to Dr. Valberg's
calculations, the dose of organic mutagens deposited in the lungs of a
worker occupationally exposed (40 hours per week) to 500 g/
m\3\ of dpm would be equivalent in potency to smoking about one
cigarette per month.\61\ Dr. Valberg indicated that a person smoking at
this level would generally be classified a nonsmoker, but he made no
attempt to quantify the carcinogenic effects. Nor did he compare this
exposure level with levels of exposures to environmental tobacco smoke
that have been linked to lung cancer.
---------------------------------------------------------------------------

    \61\ The only details provided for this calculation pertained to
adjusting 8-hour occupational exposures. Dr. Valberg adjusted the
500 g/m\3\ concentration for an 8-hour occupational
exposure to a supposedly equivalent 24-hour continuous concentration
of 92 g/m\3\. This adjustment ignored differences in
breathing rates between periods of sleep, leisure activities, and
heavy work. Even under the unrealistic assumption of homogeneous
breathing rates, the calculation appears to be erroneous, since (500
g/m\3\)  x  (40 hours/week) is nearly 30 percent greater
than (92 g/m\3\)  x  (168 hours/week). Also, Dr. Valberg
stated that the calculation assumed a deposition fraction of 20
percent for dpm but did not state what deposition fraction was being
assumed for the particles in cigarette smoke.
---------------------------------------------------------------------------

    Since the commenter did not provide details of Dr. Valberg's
calculation, MSHA was unable to verify its accuracy or evaluate the
plausibility of key assumptions. However, even if the equivalence is
approximately correct, using it to discount the possibility that dpm
increases the risk of lung cancer relies on several questionable
assumptions. Although their precise role in the analysis is unclear
because it was not presented in detail, these assumptions apparently
include:
    (1) That there is a good correlation between genotoxicity dose-
response and carcinogenicity dose-response. Although genotoxicity data
can be very useful for identifying a carcinogenic hazard,
carcinogenesis is a highly complex process that may involve the
interaction of many mutagenic, physiological, and biochemical
responses. Therefore, the shape and slope of a carcinogenic dose-
response relationship cannot be readily predicted from a genotoxic
dose-response relationship.
    (2) That only the organic fraction of dpm contributes to
carcinogenesis. This contradicts the findings reported by Ichinose et
al. (1997b) and does not take into account the contribution that
inflammation and active oxygen radicals induced by the inorganic carbon
core of dpm may have in promoting lung cancers. Multiple routes of
carcinogenesis may operate in human lungs--some requiring only the
various organic mutagens in dpm and others involving induction of free
radicals by the elemental carbon core, either alone or in combination
with the organics.
    (3) That the only mutagens in dpm are those that have been
identified as mutagenic to bacteria and that the

[[Page 5819]]

mutagenic constituents of dpm have all been identified. One of the most
potent of all known mutagens (3-nitrobenzanthrone) was only recently
isolated and identified in dpm (Enya et al., 1997).
    (4) That the mutagenic components of dpm have the same combined
potency as those in cigarette smoke. This ignores the relative potency
and amounts of the various mutagenic constituents. If the calculation
did not take into account the relative amounts and potencies of all the
individual mutagens in dpm and cigarette smoke, then it oversimplified
the task of making such a comparison.
    In sum, unlike the experimental findings of dpm genotoxicity
discussed above, the analysis by Dr. Valberg is not based on empirical
evidence from dpm experiments, and it appears to rely heavily on
questionable assumptions. Moreover, the contention that active
components of dpm are not available in sufficient quantities to cause
significant mutagenic damage in humans appears to be directly
contradicted by the empirical evidence of elevated DNA adduct levels in
exposed workers (Hemminki et al., 1994; Hou et al., 1995; Nielsen et
al., 1996; Qu et al., 1997).
    (2) Animal Inhalation Studies. When dpm is inhaled, a number of
adverse effects that may contribute to carcinogenesis are discernable
by microscopic and biochemical analysis. For a comprehensive review of
these effects, see Watson and Green (1995). In brief, these effects
begin with phagocytosis, which is essentially an attack on the diesel
particles by cells called alveolar macrophages. The macrophages engulf
and ingest the diesel particles, subjecting them to detoxifying
enzymes. Although this is a normal physiological response to the
inhalation of foreign substances, the process can produce various
chemical byproducts injurious to normal cells. In attacking the diesel
particles, the activated macrophages release chemical agents that
attract neutrophils (a type of white blood cell that destroys
microorganisms) and additional alveolar macrophages. As the lung burden
of diesel particles increases, aggregations of particle-laden
macrophages form in alveoli adjacent to terminal bronchioles, the
number of Type II cells lining particle-laden alveoli increases, and
particles lodge within alveolar and peribronchial tissues and
associated lymph nodes. The neutrophils and macrophages release
mediators of inflammation and oxygen radicals, which have been
implicated in causing various forms of chromosomal damage, genetic
mutations, and malignant transformation of cells (Weitzman and Gordon,
1990). Eventually, the particle-laden macrophages are functionally
altered, resulting in decreased viability and impaired phagocytosis and
clearance of particles. This series of events may result in pulmonary
inflammatory, fibrotic, or emphysematous lesions that can ultimately
develop into cancerous tumors.
    IARC (1989), Mauderly (1992), Busby and Newberne (1995), IPCS
(1996), Cal-EPA (1998), and US EPA (1999) reviewed the scientific
literature relating to excess lung cancers observed among laboratory
animals chronically exposed to filtered and unfiltered diesel exhaust.
The experimental data demonstrate that chronic exposure to whole diesel
exhaust increases the risk of lung cancer in rats and that dpm is the
causative agent. This carcinogenic effect has been confirmed in two
strains of rats and in at least five laboratories. Experimental results
for animal species other than the rat, however, are either inconclusive
or, in the case of Syrian hamsters, suggestive of no carcinogenic
effect. In two of three mouse studies reviewed by IARC (1989), lung
tumor formation (including adenocarcinomas) was increased in the
exposed animals as compared to concurrent controls; in the third study,
the total incidence of lung tumors was not elevated compared to
historical controls. Two more recent mouse studies (Heinrich et al.,
1995; Mauderly et al., 1996) have both reported no statistically
significant increase in lung cancer rates among exposed mice, as
compared to contemporaneous controls. Monkeys exposed to diesel exhaust
for two years did not develop lung tumors, but the short duration of
exposure was judged inadequate for evaluating carcinogenicity in
primates.
    Bond et al. (1990a) investigated differences in peripheral lung DNA
adduct formation among rats, hamsters, mice, and monkeys exposed to dpm
at a concentration of 8100 g/m\3\ for 12 weeks. Mice and
hamsters showed no increase of DNA adducts in their peripheral lung
tissue, whereas rats and monkeys showed a 60 to 80-percent increase.
The increased prevalence of lung DNA adducts in monkeys suggests that,
with respect to DNA adduct formation, the human lungs' response to dpm
inhalation may more closely resemble that of rats than that of hamsters
or mice.
    The conflicting carcinogenic effects of chronic dpm inhalation
reported in studies of rats, mice, and hamsters may be due to non-
equivalent delivered doses or to differences in response among species.
Indeed, monkey lungs have been reported to respond quite differently
than rat lungs to both diesel exhaust and coal dust (Nikula, 1997).
Therefore, the results from rat experiments do not, by themselves,
establish that there is any excess risk due to dpm exposure for humans.
However, the human epidemiologic and genotoxicity (DNA adduct) data
indicate that humans comprise a species that, like rats, do suffer a
carcinogenic response to dpm exposure. This would be consistent with
the observation, mentioned above, that lung DNA adduct formation is
increased among exposed rats but not among exposed hamsters or mice.
Therefore, although MSHA recognizes that there are important
differences between rats and humans (as there are also between rats and
hamsters or mice), MSHA considers the rat studies relevant to an
evaluation of human health risks.
    Reactions similar to those observed in rats inhaling dpm have also
been observed in rats inhaling fine particles with no organic component
(Mauderly et al., 1994; Heinrich et al., 1994, 1995; Nikula et al.,
1995). Rats exposed to titanium dioxide (TiO2) or pure
carbon (``carbon black'') particles, which are not considered to be
genotoxic, exhibited similar pathological responses and developed lung
cancers at about the same rate as rats exposed to whole diesel exhaust.
Carbon black particles were used in these experiments because they are
physically similar to the inorganic carbon core of dpm but have
negligible amounts of organic compounds adsorbed to their surface.
Therefore, at least in some species, it appears that the lung cancer
toxicity of dpm may result largely from a biochemical response to the
core particle itself rather than from specific, genotoxic effects of
the adsorbed organic compounds.\62\
---------------------------------------------------------------------------

    \62\ NIOSH commented as follows: ``Data cited by MSHA in support
of this statement are not comparable. Rats were exposed to dpm at 4
mg/m\3\ for 2 years (Mauderly et al. 1987; Brightwell et al. 1989),
in contrast to rats exposed to TiO2 at 250 mg/m\3\ for
two years [reference to article (Lee et al. 1985) not cited by
MSHA]. It is not apparent that the overload mechanism that is
proposed to be responsible for tumors in the TiO2 exposed
rats could also have been responsible for the tumors seen in the dpm
exposed rats at 62-fold lower exposure concentrations.'' In the
reports cited by MSHA, levels of TiO2 and/or carbon black
were commensurate with dpm levels.
---------------------------------------------------------------------------

    One commenter stated that, in the proposed risk assessment, MSHA
had neglected three additional studies suggesting that lung cancer
risks in animals inhaling diesel exhaust are unrelated to genotoxic
mechanisms. One of these studies (Mauderly et al., 1996) did not
pertain to questions of

[[Page 5820]]

genotoxicity but has been cited in the discussion of mouse studies
above. The other two studies (Randerath et al., 1995 and Belinsky et
al., 1995) were conducted as part of the cancer bioassay described in
the 1994 article by Mauderly et al. (cited in the preceding paragraph).
In the Randerath study, the investigators found that no DNA adducts
specific to either diesel exhaust or carbon black were induced in the
lungs of rats exposed to the corresponding substance. However, after
three months of exposure, the total level of DNA adducts and the levels
of some individual adducts were significantly higher in the diesel-
exposed rats than in the controls. In contrast, multiple DNA adducts
thought to be specific to diesel exhaust formed in the skin and lungs
of mice treated topically with organic dpm extract. These results are
consistent with the findings of Mauderly et al. (1994, op cit.). They
imply that although the organic compounds of diesel exhaust are capable
of damaging cellular DNA, they did not inflict such damage under the
conditions of the inhalation experiment performed. The report noted
that these results do not rule out the possibility of DNA damage by
inhaled organics in ``other species or * * * [in] exposure situations
in which the concentrations of diesel exhaust particles are much
lower.'' In the Belinsky study, the investigators measured mutations in
selected genes in the tumors of those rats that had developed lung
cancer. This study did not succeed in elucidating the mechanisms by
which dpm and carbon black cause lung tumors in rats. The authors
concluded that ``until some of the genes involved in the
carcinogenicity of diesel exhaust and carbon black are identified, a
role for the organic compounds in tumor development cannot be
excluded.''
    The carbon-black and TiO2 studies discussed above
indicate that lung cancers in rats exposed to dpm may be induced by a
mechanism that does not require the bioavailability of genotoxic
organic compounds adsorbed on the elemental carbon particles. Some
researchers have interpreted these studies as also suggesting that (1)
the carcinogenic mechanism in rats depends on massive overloading of
the lung and (2) that this may provide a mechanism of carcinogenesis
involving a threshold effect specific to rats, which has not been
observed in other rodents or in humans (Oberdorster, 1994; Watson and
Valberg, 1996). Some commenters on the ANPRM cited the lack of a link
between lung cancer and coal dust or carbon black exposure as evidence
that carbon particles, by themselves, are not carcinogenic in humans.
Coal mine dust, however, consists almost entirely of particles larger
than those forming the carbon core of dpm or used in the carbon black
and TiO2 rat studies. Furthermore, although there have been
nine studies reporting no excess risk of lung cancer among coal miners
(Liddell, 1973; Costello et al., 1974; Armstrong et al., 1979; Rooke et
al., 1979; Ames et al., 1983; Atuhaire et al., 1985; Miller and
Jacobsen, 1985; Kuempel et al., 1995; Christie et al., 1995), eight
studies have reported an elevated risk of lung cancer for those exposed
to coal dust (Enterline, 1972; Rockette, 1977; Howe et al., 1983;
Correa et al., 1984; Levin et al., 1988; Morabia et al., 1992; Swanson
et al., 1993; Morfeld et al., 1997). The positive results in five of
these studies (Enterline, 1972; Rockette, 1977; Howe et al., 1983;
Morabia et al., 1992; Swanson et al., 1993) were statistically
significant. Morabia et al. (op cit.) reported increased risk
associated with duration of exposure, after adjusting for cigarette
smoking, asbestos exposure, and geographic area. Furthermore, excess
lung cancers have been reported among carbon black production workers
(Hodgson and Jones, 1985; Siemiatycki, 1991; Parent et al., 1996).
After a comprehensive evaluation of the available scientific evidence,
the World Health Organization's International Agency for Research on
Cancer concluded: ``Carbon black is possibly carcinogenic to humans
(Group 2B).'' (IARC, 1996).
    The carbon black and TiO2 animal studies cited above do
not prove there is a threshold below which dpm exposure poses no risk
of causing lung cancer in humans. They also do not prove that dpm
exposure has no incremental, genotoxic effects. Even if the genotoxic
organic compounds in dpm were biologically unavailable and played no
role in human carcinogenesis, this would not rule out the possibility
of a genotoxic route to lung cancer (even for rats) due to the presence
of the particles themselves. For example, as a byproduct of the
biochemical response to the presence of particles in the alveoli, free
oxidant radicals may be released as macrophages attempt to digest the
particles. There is evidence that dpm can both induce production of
reactive oxygen agents and also depress the activity of naturally
occurring antioxidant enzymes (Mori, 1996; Ichinose et al., 1997; Sagai
et al., 1993). Oxidants can induce carcinogenesis either by reacting
directly with DNA, or by stimulating cell replication, or both
(Weitzman and Gordon, 1990). Salvi et al. (1999) reported acute
inflammatory responses in the airways of human exposed to dpm for one
hour at a concentration of 300 g/m\3\. Such inflammation is
associated with the production of free radicals and could provide
routes to lung cancer with even when normal lung clearance is
occurring. It could also give rise to a ``quasi-threshold,'' or surge
in response, corresponding to the exposure level at which the normal
clearance rate becomes overwhelmed (lung overload).
    Oxidant activity is not the only mechanism by which dpm could exert
carcinogenic effects in the absence of mutagenic activity by its
organic fraction. In its commentary on the Randerath study discussed
above, the HEI's Health Review Committee suggested that dpm could both
cause genetic damage by inducing free oxygen radicals and also enhance
cell division by inducing cytokines or growth hormones:

    It is possible that diesel exhaust exerts its carcinogenic
effects through a mechanism that does not involve direct
genotoxicity (that is, formation of DNA adducts) but involves
proliferative responses such as chronic inflammation and hyperplasia
arising from high concentrations of particles deposited in the lungs
of the exposed rats. * * * Phagocytes (macrophages and neutrophils)
released during inflammatory reactions ``produce reactive oxygen
species that can damage DNA. * * * Particles (with or without
adsorbed PAHs) may thus induce oxidative DNA damage via oxygen free
radicals. * * * Alternatively, activated phagocytes may release
cytokines or growth factors that are known to increase cell
division. Increased cell division has been implicated in cancer
causation. * * * Thus, in addition to oxidative DNA damage,
increased cell proliferation may be an important mechanism by which
diesel exhaust and other insoluble particles induce pulmonary
carcinogenesis in the rat. [Randerath et al., 1995, p. 55]

    Even if lung overload were the primary or sole route by which dpm
induced lung cancer, this would not mean that the high dpm
concentrations observed in some mines are without hazard. It is
noteworthy, moreover, that dpm exposure levels recorded in some mines
have been almost as high as laboratory exposures administered to rats
showing a clearly positive response. Intermittent, occupational
exposure levels greater than about 500 g/m\3\ dpm may
overwhelm the human lung clearance mechanism (Nauss et al., 1995).
Therefore, concentrations at the even higher levels currently observed
in some mines could be expected to cause overload in some humans,
possibly inducing lung cancer by a mechanism

[[Page 5821]]

similar to what occurs in rats. In addition, a proportion of exposed
individuals can always be expected to be more susceptible than normal
to clearance impairments and lung overload. Inhalation at even moderate
levels may significantly impair clearance, especially in susceptible
individuals. Exposures to cigarette smoke and respirable mineral dusts
may further depress clearance mechanisms and reduce the threshold for
overload. Consequently, even at dpm concentrations far lower than 500
g/m\3\ dpm, impaired clearance due to dpm inhalation may
provide an important route to lung cancer in humans, especially if they
are also inhaling cigarette smoke and other fine dusts simultaneously.
(Hattis and Silver, 1992, Figures 9, 10, 11).
    Furthermore, as suggested above, lung overload is not necessarily
the only route to carcinogenesis in humans. Therefore, dpm
concentrations too low to cause overload still may present a hazard. In
humans exposed over a working lifetime to doses insufficient to cause
overload, carcinogenic mechanisms unrelated to overload may operate, as
indicated by the human epidemiologic studies and the data on human DNA
adducts cited in the preceding subsection of this risk assessment. It
is possible that overload provides the dominant route to lung cancer at
high concentrations of fine particulate, while other mechanisms emerge
as more relevant for humans under lower-level exposure conditions.
    The NMA noted that, in 1998, the US EPA's Clean Air Scientific
Advisory Committee (CASAC) concluded that there is ``no evidence that
the organic fraction of soot played a role in rat tumorigenesis at any
exposure level, and considerable evidence that it did not.'' According
to the NMA, this showed ``* * * it is the rat data--not the hamster
data--that lacks relevance for human health assessment.''
    It must first be noted that, in MSHA's view, all of the
experimental animal data on health effects has relevance for human
health risk assessment--whether the evidence is positive or negative
and even if the positive results cannot be used to quantify human risk.
The finding that different mammalian species exhibit important
differences in response is itself relevant for human risk assessment.
Second, the passage quoted from CASAC pertains to the route for
tumorigenesis in rats and does not discuss whether this does or does
not have relevance to humans exposed at high levels. The context for
the CASAC deliberations was ambient exposure conditions in the general
environment, rather than the higher occupational exposures that might
impair clearance rates in susceptible individuals. Third, the comment
assumes that only a finding of tumorigenesis attributable to the
organic portion of dpm would elucidate mechanisms of potential health
effects in humans. This ignores the possibility that a mechanism
promoting tumors, but not involving the organics, could operate in both
rats and humans. Induction of free oxygen radicals is an example.
Fourth, although there may be little or no evidence that organics
contributed to rat tumorigenesis in the studies performed, there is
evidence that the organics contributed to increases in DNA adduct
formation. This kind of activity could have tumorigenic consequences in
humans who may be exposed for periods far longer than a rat's 3-year
lifetime and who, as a consequence, have more time to accumulate
genetic damage from a variety of sources.
    Bond et al. (1990b) and Wolff et al. (1990) investigated adduct
formation in rats exposed to various concentrations of either dpm or
carbon black for 12 weeks. At the highest concentration (10 mg/m\3\),
DNA adduct levels in the lung were increased by exposure to either dpm
or carbon black; but levels in the rats exposed to dpm were
approximately 30 percent higher. Gallagher et al. (1994) exposed
different groups of rats to diesel exhaust, carbon black, or
TiO2 and detected no significant difference in DNA adduct
levels in the lung. However, the level of one type of adduct, thought
to be derived from a PAH, was elevated in the dpm-exposed rats but not
found in the control group or in rats exposed to carbon black or
TiO2.
    These studies indicate that the inorganic carbon core of dpm is not
the only possible agent of genetic damage in rats inhaling dpm. After a
review of these and other studies involving DNA adducts, IPCS (1996)
concluded that ``Taken together, the studies of DNA adducts suggest
that some organic chemicals in diesel exhaust can form DNA adducts in
lung tissue and may play a role in the carcinogenic effects. * *
*however, DNA adducts alone cannot explain the carcinogenicity of
diesel exhaust, and other factors, such as chronic inflammation and
cell proliferation, are also important.''
    Nauss et al. (1995, pp. 35-38) judged that the results observed in
the carbon black and TiO2 inhalation studies on rats do not
preclude the possibility that the organic component of dpm has
important genotoxic effects in humans. More generally, they also do not
prove that lung overload is necessary for dpm-induced lung cancer.
Because of the relatively high doses administered in some of the rat
studies, it is conceivable that an overload phenomenon masked or even
inhibited other potential cancer mechanisms. At dpm concentrations
insufficient to impair clearance, carcinogenesis may have followed
other routes, some possibly involving the organic compounds. At these
lower concentrations, or among rats for which overload did not occur,
tumor rates for dpm, carbon black, and TiO2 may all have
been too low to make statistically meaningful comparisons.
    The NMA argued that ``MSHA's contention that lung overload might
``mask'' tumor production by lower doses of dpm has been convincingly
rebutted by recognized experts in the field,'' but provided no
convincing explanation of why such masking could not occur. The NMA
went on to say:

    The [CASAC] Panel viewed the premises that: a) a small tumor
response at low exposure was overlooked due to statistical power;
and b) soot-associated organic mutagens had a greater effect at low
than at high exposure levels to be without foundation. In the
absence of supporting evidence, the Panel did not view derivation of
a quantitative estimate of human lung cancer risk from the low-level
rat data as appropriate.

MSHA is not attempting to ``derive a quantitative estimate of human
lung cancer risk from the low-level rat data.''
    Dr. Peter Valberg, writing for the West Virginia Coal Association,
provided the following argument for discounting the possibility of
other carcinogenic mechanisms being masked by overload in the rat
studies:

    Some regulatory agencies express concern about the mutagens
bound to dpm. They hypothesize that, at high exposure levels,
genotoxic mechanisms are overwhelmed (masked) by particle-overload
conditions. However, they argue that at low-exposure concentrations,
these organic compounds could represent a lung cancer risk. Tumor
induction by mutagenic compounds would be characterized by a linear
dose-response and should be detectable, given enough exposed rats.
By using a ``meta-analysis'' type of approach and combining data
from eight long-term rat inhalation studies, the lung tumor response
can be analyzed. When all dpm-exposed rats from lifetime-exposure
studies are combined, a threshold of response (noted above) occurs
at approximately 600 g/m\3\ continuous lifetime exposure
(approximately 2,500 g/m\3\ of occupational exposure).
Additional statistical analysis of only those rats exposed to low
concentrations of dpm confirms the absence of a tumorigenic effect
below that threshold. Thus, even data in rats (the most sensitive
laboratory species) do not support the hypothesis that particle-
bound organics cause tumors.

[[Page 5822]]

    MSHA finds that this analysis relies on several questionable and
unsupported assumptions and that, for the following reasons, the
possibility remains that organic compounds in inhaled dpm may, under
the right exposure conditions, contribute to its carcinogenic effects:

    (1) The absence of evidence for an organic carbon effect is not
equivalent to evidence of the absence of such an effect. Dr. Valberg
did not demonstrate that enough rats were exposed, at levels
insufficient to cause overload, to ensure detection of a 30- to 40-
percent increase in the risk of lung cancer. Also, the normal lifespan
of a rat whose lung is not overloaded with particles may, because of
the lower concentrations involved, provide insufficient time for the
organic compounds to express carcinogenic effects. Furthermore, low
bioavailability of the organics could further reduce the likelihood
that a carcinogenic sequence of mutations would occur within a rat's
relatively short lifespan (i.e., at particle concentrations too low to
cause overload).
    (2) If the primary mechanism for carcinogenesis requires a reduced
clearance rate (due to overload), then acute exposures are important,
and it may not be appropriate to represent equivalent hazards by
spreading an 8-hour occupational exposures over a 24-hour period. For
example, eight hours at 600 g/m\3\ would have different
implications for lung clearance than 24 hours at 200 g/m\3\.
    (3) Granting that the rat data cannot be used to extrapolate risk
for humans, these data should also not be used to rule out mechanisms
of carcinogenesis that may operate in humans but not in rats.
Clearance, for example, may operate differently in humans than in rats,
and there may be a gradual rather than abrupt change in human overload
conditions with increasing exposure. Also, at least some of the organic
compounds in dpm may be more biologically available to the human lung
than to that of the rat.
    (4) For experimental purposes, laboratory rats are deliberately
bred to be homogeneous. This is done, in part, to deliberately minimize
differences in response between individuals. Therefore, individual
differences in the threshold for lung overload would tend to be masked
in experiments on laboratory rats. It is likely that human populations
would exhibit, to a far greater extent than laboratory rats, a range of
susceptibilities to lung overload. Also some humans, unlike the
laboratory rats in these experiments, place additional burdens on their
lung clearance by smoking.
    One commenter (MARG) concluded that ``[t]here is * * * no basis for
extrapolating the rat results to human beings; the animal studies,
taken together, do not justify MSHA's proposals.''
    MSHA is neither extrapolating the rat results to make quantitative
risk estimates for humans nor using them, in isolation, as a
justification for these regulations. MSHA does regard it as
significant, however, that the evidence for an increased risk of lung
cancer due to chronic dpm inhalation comes from both human and animal
studies. MSHA agrees that the quantitative results observed for rats in
existing studies should not be extrapolated to humans. Nevertheless,
the fact that high dpm exposures for two or three years can induce lung
cancer in rats enhances the epidemiologic evidence that much longer
exposures to miners, at concentrations of the same order of magnitude,
could also induce lung cancers.

3. Characterization of Risk

    After reviewing the evidence of adverse health effects associated
with exposure to dpm, MSHA evaluated that evidence to ascertain whether
exposure levels currently existing in mines warrant regulatory action
pursuant to the Mine Act. The criteria for this evaluation are
established by the Mine Act and related court decisions. Section
101(a)(6)(A) provides that:

    The Secretary, in promulgating mandatory standards dealing with
toxic materials or harmful physical agents under this subsection,
shall set standards which most adequately assure on the basis of the
best available evidence that no miner will suffer material
impairment of health or functional capacity even if such miner has
regular exposure to the hazards dealt with by such standard for the
period of his working life.

    Based on court interpretations of similar language under the
Occupational Safety and Health Act, there are three questions that need
to be addressed: (a) Whether health effects associated with dpm
exposure constitute a ``material impairment'' to miner health or
functional capacity; (b) whether exposed miners are at significant
excess risk of incurring any of these material impairments; and (c)
whether the rule will substantially reduce such risks.
    Some commenters argued that the link between dpm exposure and
material health impairments is questionable, and that MSHA should wait
until additional scientific evidence becomes available before
concluding that there are health risks due to such exposure warranting
regulatory action. For example, MARG asserted that ``[c]ontrary to the
suggestions in the [proposed] preamble, a link between dpm exposure and
serious illness has never been established by reliable scientific
evidence.'' \63\ MARG
---------------------------------------------------------------------------

    \63\ MARG supported this assertion by claiming that ``[t]he EPA
reports which MSHA references in its preamble were found `not
scientifically adequate for making regulatory decisions concerning
the use of diesel-powered engines' by EPA's Clean Air Scientific
Advisory Committee. [reference to CASAC (1998)]'' Contrary to MARG's
claim, CASAC (1998) did not review any of the 20 EPA documents MSHA
cited in the proposed preamble. Instead, the document reviewed by
CASAC (1998) was an unpublished draft of a health risk assessment on
diesel exhaust (EPA, 1998), to which MSHA made no reference. Since
MSHA has not relied in any way on this 1998 draft document, its
``scientific adequacy'' is entirely irrelevant to this rulemaking.
    In response to the 1998 CASAC review, EPA modified its draft
risk assessment (EPA, 1999), and CASAC subsequently reviewed the
1999 draft (CASAC, 2000). CASAC found the revised draft much
improved over the previous version and agreed that even
environmental exposure to diesel emissions is likely to increase the
risk of lung cancer (CASAC, 2000). CASAC endorsed this conclusion
for dpm concentrations in ambient air, which are lower, by a factor
of more than 100, than the levels observed in some mines (see Fig.
III-4).

---------------------------------------------------------------------------
continued as follows:

    Precisely because the scientific evidence * * * is inconclusive
at best, NIOSH and NCI are now conducting a * * * [study] to
determine whether diesel exhaust is linked to illness, and if so, at
what level of exposure. * * * MARG is also funding an independent
parallel study.
    * * * Until data from the NIOSH/NCI study, and the parallel MARG
study, are available, the answers to these important questions will
not be known. Without credible answers to these and other questions,
MSHA's regulatory proposals * * * are premature * * *.''

    For reasons explained below, MSHA does not agree that the
collective weight of scientific evidence is ``inconclusive at best.''
Furthermore, the criteria for evaluating the health effects evidence do
not require scientific certainty. As noted by Justice Stevens in an
important case on risk involving the Occupational Safety and Health
Administration, the need to evaluate risk does not mean an agency is
placed into a ``mathematical straitjacket.'' [Industrial Union
Department, AFL-CIO v. American Petroleum Institute, 448 U.S. 607, 100
S.Ct. 2844 (1980), hereinafter designated the ``Benzene'' case]. The
Court recognized that regulation may be necessary even when scientific
knowledge is not complete; and--

so long as they are supported by a body of reputable scientific
thought, the Agency is free to use conservative assumptions in
interpreting the data * * * risking error on the side of
overprotection rather than underprotection. [Id. at 656].

[[Page 5823]]

    Moreover, the statutory criteria for evaluating health effects do
not require MSHA to wait for incontrovertible evidence. In fact, MSHA
is required to set standards based on the ``best available evidence''
(emphasis added).
a. Material Impairments to Miners' Health or Functional Capacity
    MSHA recognizes that there is considerable disagreement, among
knowledgeable parties, in the interpretation of the overall body of
scientific research and medical evidence related to human health
effects of dpm exposures. One commenter for example, interpreted the
collective evidence as follows:

    * * * the best available scientific evidence shows that diesel
particulate exposure is associated with serious material impairment
of health. * * * there is clear evidence that diesel particulate
exposure can cause lung cancer (as well as other serious non-
malignant diseases) among workers in a variety of occupational
settings. While no body of scientific evidence is ever completely
definitive, the evidence regarding diesel particulate is
particularly strong * * *. [Michael Silverstein, MD, State of
Washington Dept. of Labor and Industries]

    Other commenters, including several national and regional
organizations representing the mining industry, sharply disagreed with
this interpretation. For example, one commenter stated that ``[i]n our
opinion, the best available evidence does not provide substantial or
credible support for the proposal.'' Several commenters argued that
evidence from within the mining industry itself was especially
weak.\64\ A representative of one mining company that had been using
diesel equipment for many years commented: ``[t]o date, the medical
history of our employees does not indicate a single case of lung
cancer, chronic illness, or material impairment of health due to
exposure to diesel exhaust. This appears to be the established norm
throughout the U.S. coal mining industry.'' This commenter, however,
submitted no evidence comparing the rate of lung cancer or other
material impairment among exposed miners to the rate for unexposed
miners (or comparable workers) of similar age, smoking habits, and
geographic location.
---------------------------------------------------------------------------

    \64\ At the public hearing on May 11, 1999, a commenter
representing MARG suggested there is evidence that miners exposed to
dpm experience adverse health effects at lower-than-normal rates.
According to this commenter, ``[s]ignificantly, the human studies
conducted in the mining industry reveal a negative propensity for
diesel particulate matter-related health effects.'' These studies
drew comparisons against an external reference population and failed
to adjust for the ``healthy worker effect.'' (See MSHA's discussion
of this effect, especially as manifested in the study by Christie et
al., 1995, in Subsection 2.c.i(2)(a) of this risk assessment.)
---------------------------------------------------------------------------

    With due consideration to all oral and written testimony, comments,
and evidence submitted during the rulemaking proceedings, MSHA
conducted a review of the scientific literature cited in Part III.2.
Based on the combined weight of the best available evidence, MSHA has
concluded that underground miners exposed to current levels of dpm are
at excess risk of incurring the following three kinds of material
impairment: (i) Sensory irritations and respiratory symptoms (including
allergenic responses); (ii) premature death from cardiovascular,
cardiopulmonary, or respiratory causes; and (iii) lung cancer. The next
three subsections will respectively explain MSHA's basis for linking
these effects with dpm exposure.
    i. Sensory Irritations and Respiratory Symptoms (including
allergenic responses). Kahn et al. (1988), Battigelli (1965), Gamble et
al. (1987a), and Rudell et al. (1996) identified a number of
debilitating acute responses to diesel exhaust exposure. These
responses included irritation of the eyes, nose and throat; headaches,
nausea, and vomiting; chest tightness and wheeze. These symptoms were
also reported by miners at the 1995 workshops and the public hearings
held on these proceedings in 1998. In addition, Ulfvarson et al. (1987,
1990) reported evidence of reduced lung function in workers exposed to
dpm for a single shift. The latter study supports attributing a portion
of the reduction to the dpm in diesel exhaust. After reviewing this
body of literature, Morgan et al. (1997) concluded ``it is apparent
that exposure to diesel fumes in sufficient concentrations may lead to
[transient] eye and nasal irritation'' and ``a transient decline of
ventilatory capacity has been noted following such exposures.''
    One commenter (Nevada Mining Association) acknowledged there was
evidence that miners exposed to diesel exhaust experienced, as a
possible consequence of their exposure, ``acute, short-term or
`transitory' irritation, such as watering eyes, in susceptible
individuals * * *''; but asserted that ``[a]ddressing any such
transient irritant effects does not require the Agency's sweeping,
stringent PEL approach [in M/NM mines].''
    Although there is evidence that such symptoms subside within one to
three days of no occupational exposure, a miner who must be exposed to
dpm day after day in order to earn a living may not have time to
recover from such effects. Hence, the opportunity for a so-called
``reversible'' health effect to reverse itself may not be present for
many miners. Furthermore, effects such as stinging, itching and burning
of the eyes, tearing, wheezing, and other types of sensory irritation
can cause severe discomfort and can, in some cases, be seriously
disabling. Also, workers experiencing sufficiently severe sensory
irritations can be incapacitated or distracted as a result of their
symptoms, thereby endangering themselves and other workers and
increasing the risk of accidents. For these reasons, MSHA considers
such irritations to constitute ``material impairments'' of health or
functional capacity within the meaning of the Act, regardless of
whether or not they are reversible. Further discussion of why MSHA
believes reversible effects can constitute material impairments can be
found above, in Subsection 2.a.2 of this risk assessment.
    The best available evidence also points to more severe respiratory
consequences of exposure to dpm. Significant statistical associations
have been detected between acute environmental exposures to fine
particulates and debilitating respiratory impairments in adults, as
measured by lost work days, hospital admissions, and emergency room
visits (see Table III-3). Short-term exposures to fine particulates, or
to particulate air pollution in general, have been associated with
significant increases in the risk of hospitalization for both pneumonia
and COPD (EPA, 1996).
    The risk of severe respiratory effects is exemplified by specific
cases of persistent asthma linked to diesel exposure (Wade and Newman,
1993). Glenn et al. (1983) summarized results of NIOSH health
evaluations among coal, salt, trona, and potash miners and reported
that ``all four of the chronic effects analyses revealed an excess of
cough and phlegm among the diesel exposed group.'' There is persuasive
evidence for a causal connection between dpm exposure and increased
manifestations of allergic asthma and other allergic respiratory
diseases, coming from recent experiments on animals and human cells
(Takenaka et al., 1995; Lovik et al., 1997; Takano et al., 1997;
Ichinose et al., 1997a). Based on controlled experiments on healthy
human volunteers, Diaz-Sanchez et al. (1994, 1996, 1997), Peterson and
Saxon (1996), and Salvi et al. (1999) reported significant increases in
various markers of allergic response resulting from exposure to dpm.
    Peterson and Saxon (1996) reviewed the scientific literature on the
relationship between PAHs and other products of fossil fuel combustion
found

[[Page 5824]]

in dpm and trends in allergic respiratory disease. They found that the
prevalences of allergic rhinitis (``hay fever'') and allergic asthma
have significantly increased with the historical increase in fossil
fuel combustion and that laboratory data support the hypothesis that
certain organic compounds found in dpm ``* * * are an important factor
in the long-term increases in the prevalence in allergic airway
disease.'' Similarly, much of the research on allergenic responses to
dpm was reviewed by Diaz-Sanchez (1997), who concluded that dpm
pollution in the ambient environment ``may play an important role in
the increased incidence of allergic airway disease.'' Morgan et al.
(1997) noted that dpm ``* * * may be partly responsible for some of the
exacerbations of asthma'' and that ``* * * it would be wise to err on
the side of caution.'' Such health outcomes are clearly ``material
impairments'' of health or functional capacity within the meaning of
the Act.
    ii. Premature Death from Cardiovascular, Cardiopulmonary, or
Respiratory Causes. The evidence from air pollution studies identifies
death, largely from cardiovascular, cardiopulmonary, or respiratory
causes, as an endpoint significantly associated with acute exposures to
fine particulates (PM2.5--see Table III-3). The weight of
epidemiologic evidence indicates that short-term ambient exposure to
particulate air pollution contributes to an increased risk of daily
mortality (EPA, 1996). Time-series analyses strongly suggest a positive
effect on daily mortality across the entire range of ambient
particulate pollution levels. Relative risk estimates for daily
mortality in relation to daily ambient particulate concentration are
consistently positive and statistically significant across a variety of
statistical modeling approaches and methods of adjustment for effects
of relevant covariates such as season, weather, and co-pollutants. The
mortality effects of acute exposures appear to be primarily
attributable to combustion-related particles in PM2.5 (such
as dpm) and are especially pronounced for death due to pneumonia, COPD,
and IHD (Schwartz et al., 1996). After thoroughly reviewing this body
of evidence, the U.S. Environmental Protection Agency (EPA) concluded:

    It is extremely unlikely that study designs not yet employed,
covariates not yet identified, or statistical techniques not yet
developed could wholly negate the large and consistent body of
epidemiologic evidence * * *. [EPA, 1996]

    There is also substantial evidence of a relationship between
chronic exposure to fine particulates (PM2.5) and an excess
(age-adjusted) risk of mortality, especially from cardiopulmonary
diseases. The Six Cities and ACS studies of ambient air particulates
both found a significant association between chronic exposure to fine
particles and excess mortality. In some of the areas studied,
PM2.5 is composed primarily of dpm; and significant
mortality and morbidity effects were also noted in those areas. In both
studies, after adjusting for smoking habits, a statistically
significant excess risk of cardiopulmonary mortality was found in the
city with the highest average concentration of PM2.5 as
compared to the city with the lowest. Both studies also found excess
deaths due to lung cancer in the cities with the higher average level
of PM2.5, but these results were not statistically
significant (EPA, 1996). The EPA concluded that--

    * * * the chronic exposure studies, taken together, suggest
there may be increases in mortality in disease categories that are
consistent with long-term exposure to airborne particles and that at
least some fraction of these deaths reflect cumulative PM impacts
above and beyond those exerted by acute exposure events * * * There
tends to be an increasing correlation of long-term mortality with PM
indicators as they become more reflective of fine particle levels.
[EPA, 1996]

    Whether associated with acute or chronic exposures, the excess risk
of death that has been linked to pollution of the air with fine
particles like dpm is clearly a ``material impairment'' of health or
functional capacity within the meaning of the Act.
    In a review, submitted by MARG, of MSHA's proposed risk assessment,
Dr. Jonathan Borak asserted that ``MSHA appears to regard all
particulates smaller than 2.5 g/m3 as equivalent.''
He argued that ``dpm and other ultra-fine particulates represents only
a small proportion of ambient particulate samples,'' that ``chronic
cough, chronic phlegm, and chronic wheezing reflect mainly
tracheobronchial effects,'' and that tracheobronchial deposition is
highly dependent on particle size distribution.
    No part of Dr. Borak's argument is directly relevant to MSHA's
identification of the risk of death from cardiovascular,
cardiopulmonary, or respiratory causes faced by miners exposed to high
concentrations of dpm. First, MSHA does not regard all fine
particulates as equivalent. However, dpm is a major constituent of
PM2.5 in many of the locations where increased mortality has
been linked to PM2.5 levels. MSHA regards dpm as presenting
a risk by virtue of its comprising a type of PM2.5. Second,
the studies MSHA used to support the existence of this risk
specifically implicate fine particles (i.e., PM2.5), so the
percentage of dpm in ``total suspended particulate emissions'' (which
includes particles even larger than PM10) is not relevant.
Third, the chronic respiratory symptoms listed by Dr. Borak are not
among the material impairments that MSHA has identified from the
PM2.5 studies. Much of the evidence pertaining to excess
mortality is based on acute--not chronic--ambient exposures of
relatively high intensity. In the preceding subsection of this risk
assessment, MSHA identified various respiratory symptoms, including
allergenic responses, but the evidence for these comes largely from
studies on diesel emissions.
    As discussed in Section 2.a.iii of this risk assessment, many
miners smoke tobacco, and miners experience COPD at a significantly
higher rate than the general population. This places many miners in two
of the groups that EPA (1996) identified as being at greatest risk of
premature mortality due to particulate exposures.
    iii. Lung Cancer.  It is clear that lung cancer constitutes a
``material impairment'' of health or functional capacity within the
meaning of the Act. Therefore, the issue to be addressed in this
section is whether there is sufficient evidence (i.e., enough to
warrant regulatory action) that occupational exposure to dpm causes the
risk of lung cancer to increase.
    In the proposed risk assessment, MSHA noted that various national
and international institutions and governmental agencies had already
classified diesel exhaust or particulate as a probable human
carcinogen. Considerable weight was also placed on two comprehensive
meta-analyses of the epidemiologic literature, which had both found
that the combined evidence supported a causal link. MSHA also
acknowledged, however, that some reviewers of the evidence disagreed
with MSHA's conclusion that, collectively, it strongly supports a
causal connection. As examples of the opposing viewpoint, MSHA cited
Stober and Abel (1996), Watson and Valberg (1996), Cox (1997), Morgan
et al. (1997), and Silverman (1998). As stated in the proposed risk
assessment, MSHA considered the opinions of these reviewers and agreed
that no individual study was perfect: even the strongest of the studies
had limitations when viewed in isolation. MSHA nevertheless concluded
(in the proposal) that the best available epidemiologic studies,
supported by experimental data

[[Page 5825]]

showing toxicity, collectively provide strong evidence that chronic dpm
exposure (at occupational levels) actually does increase the risk of
lung cancer in humans.
    Although miners and labor representatives generally agreed with
MSHA's interpretation of the collective evidence, many commenters
representing the mining industry strongly objected to MSHA's
conclusion. Some of these commenters also expressed dissatisfaction
with MSHA's treatment, in the proposed risk assessment, of opposing
interpretations of the collective evidence--saying that MSHA had
dismissed these opposing views without sufficient explanation. Some
commenters also submitted new critiques of the existing evidence and of
the meta-analyses on which MSHA had relied. These commenters also
emphasized the importance of two reports (CASAC, 1998 and HEI, 1999)
that both became available after MSHA completed its proposed risk
assessment.
    MSHA has re-evaluated the scientific evidence relating lung cancer
to diesel emissions in light of the comments, suggestions, and detailed
critiques submitted during these proceedings. Although MSHA has not
changed its conclusion that occupational dpm exposure increases the
risk of lung cancer, MSHA believes that the public comments were
extremely helpful in identifying areas of MSHA's discussion of lung
cancer needing clarification, amplification, and/or additional
supportive evidence.
    Accordingly MSHA has re-organized this section of the risk
assessment into five subsections. The first of these provides MSHA's
summary of the collective epidemiologic evidence. Second is a
description of results and conclusions from the only two existing peer-
reviewed and published statistical meta-analyses of the epidemiologic
studies: Bhatia et al. (1998) and Lipsett and Campleman (1999). The
third subsection contains a discussion of potential systematic biases
that might tend to shift all study results in the same direction. The
fourth evaluates the overall weight of evidence for causality,
considering not only the collective epidemiologic evidence but also the
results of toxicity experiments. Within each of these first four
subsections, MSHA will respond to the relevant issues and criticisms
raised by commenters in these proceedings, as well as by other outside
reviewers. The final subsection will describe general conclusions
reached by other reviewers of this evidence, and present some responses
by MSHA about opposing interpretations of the collective evidence.
    (1) Summary of Collective Epidemiologic Evidence. As mentioned in
Section III.2.c.i(2)(a) and listed in Tables III-4 and III-5, MSHA
reviewed a total of 47 epidemiologic studies involving lung cancer and
diesel exposure. Some degree of association between occupational dpm
exposure and an excess rate of lung cancer was reported in 41 of these
studies: 22 of the 27 cohort studies and 19 of the 20 case-control
studies. Section III.2.c.1(2)(a) explains MSHA's criteria for
evaluating these studies, summarizes those on which MSHA places
greatest weight, and explains why MSHA places little weight on the six
studies reporting no increased risk of lung cancer for exposed workers.
It also contains summaries of the studies involving miners, addresses
criticisms of individual studies by commenters and reviewers, and
discusses studies that, according to some commenters, suggest that dpm
exposure does not increase the risk of lung cancer.
    Here, as in the earlier, proposed version of the risk assessment,
MSHA was careful to note and consider limitations of the individual
studies. Several commenters interpreted this as demonstrating a
corresponding weakness in the overall body of epidemiologic evidence.
For example, one commenter [Energy West] observed that ``* * * by its
own admission in the preamble * * * most of the evidence in [the
epidemiologic] studies is relatively weak'' and argued that MSHA's
conclusion was, therefore, unjustified.
    It should first be noted that the three most recent epidemiologic
studies became available too late for inclusion in the risk assessment
as originally written. These three (Johnston et al., 1997; Saverin et
al., 1999; Bruske-Hohlfeld, 1999) rank among the strongest eight
studies available (see Section III.2.c.1(2)(a)) and do not have the
same limitations identified in many of the other studies. Even so, MSHA
recognizes that no single one of the existing epidemiologic studies,
viewed in isolation, provides conclusive evidence of a causal
connection between dpm exposure and an elevated risk of lung cancer in
humans. Consistency and coherency of results, however, do provide such
evidence. An appropriate analogy for the collective epidemiologic
evidence is a braided steel cable, which is far stronger than any of
the individual strands of wire making it up. Even the thinnest strands
can contribute to the strength of the cable.
(a) Consistency of Epidemiologic Results
    Although no epidemiologic study is flawless, studies of both cohort
and case-control design have quite consistently shown that chronic
exposure to diesel exhaust, in a variety of occupational circumstances,
is associated with an increased risk of lung cancer. Furthermore, as
explained earlier in this risk assessment, limitations such as small
sample size, short latency, and (usually) exposure misclassification
reduce the power of a study. These limitations make it more difficult
to detect a relationship even when one exists. Therefore, the sheer
number of studies showing a positive association readily distinguishes
those studies criticized by Taubes (1995), where weak evidence is
available from only a single study. With only rare exceptions,
involving too few workers and/or observation periods too short to have
a good chance of detecting excess cancer risk, the human studies have
shown a greater risk of lung cancer among exposed workers than among
comparable unexposed workers.
    Moreover, the fact that 41 out of 47 studies showed an excess risk
of lung cancer for exposed workers may itself be a significant result,
even if the evidence in most of those 41 studies is relatively weak.
Getting ``heads'' on a single flip of a coin, or two ``heads'' out of
three flips, does not provide strong evidence that there is anything
special about the coin. However, getting 41 ``heads'' in 47 flips would
normally lead one to suspect that the coin was weighted in favor of
heads. Similarly, results reported in the epidemiologic literature lead
one to suspect that the underlying relationship between diesel exposure
and an increased risk of lung cancer is indeed positive.
    More formally, as MSHA pointed out in the earlier version of this
risk assessment, the high proportion of positive studies is
statistically significant according to the 2-tailed sign test. Under
the ``null hypothesis'' that there is no systematic bias in one
direction or the other, and assuming that the studies are independent,
the probability of 41 or more out of 47 studies being either positive
or negative is less than one per ten million. Therefore, the sign test
rejects, at a very high confidence level, the null hypothesis that each
study is equally likely to be positive or negative. This means that the
collective results, showing increased risk for exposed workers, are
statistically significant at a very high confidence level--regardless

[[Page 5826]]

of the statistical significance of any individual study.
    MSHA received no comments directly disputing its attribution of
statistical significance to the collective epidemiologic evidence based
the sign test. However, several commenters objected to the concept that
a number of inconclusive studies can, when viewed collectively, provide
stronger evidence than the studies considered in isolation. For
example, the Engine Manufacturers Association (EMA) asserted that--

[j]ust because a number of studies reach the same conclusion does
not make the collective sum of those studies stronger or more
conclusive, particularly where the associations are admittedly weak
and scientific difficulties exist in each. [EMA]

    Similarly, IMC Global stated that

* * * IMC Global does not consider cancer studies with a relative
risk of less than 2.0 as showing evidence of a casual relationship
between dpm exposure and lung cancer. * * * Thus while MSHA states
[in the proposed risk assessment; now updated to 41 out of 47] that
38 of 43 epidemiologic studies show some degree of association
between occupational dpm exposures and lung cancer and considers
that fact significant, IMC Global does not. [IMC Global]

    Although MSHA agrees that even statistically significant
consistency of epidemiologic results is not sufficient to establish
causality, MSHA believes that consistency is an important part of
establishing that a suspected association is causal.\65\ Many of the
commenters objecting to MSHA's emphasis on the collective evidence
failed to distinguish the strength of evidence in each individual study
from the strength of evidence in total.
---------------------------------------------------------------------------

    \65\ With respect to the IMC Global's blanket rejection of
studies showing a relative risk less than 2.0, please see also the
related discussions in Subsection 2.c.i(2)(a) above, under the
heading of ``Potential Confounders,'' and in Subsection 3.a.iii(3)
below, entitled ``Potential Systemic Biases.''
---------------------------------------------------------------------------

    Furthermore, weak evidence (from just one study) should not be
confused with a weak effect. As Dr. James Weeks pointed out at the
public hearing on Nov. 19, 1998, a 40-percent increase in lung cancer
is a strong effect, even if it may be difficult to detect in an
epidemiologic study.
    Explicable differences, or heterogeneity, in the magnitudes of
relative risk reported from different studies should not be confused
with inconsistency of evidence. For example, as described by Silverman
(1998), one of the available meta-analyses (Bhatia et al., 1998)
``examined the primary sources of heterogeneity among studies and found
that a main source of heterogeneity is the variation in diesel exhaust
exposure across different occupational groups.'' Figures III-5 and III-
6, taken from Cohen and Higgins (1995), respectively show relative
risks reported for the two occupations on which the most studies are
available: railroad workers and truck drivers.
    Each of these two charts compares results from studies that
adjusted for smoking to results from studies that did not make such an
adjustment. For each study, the point plotted is the estimated relative
risk or odds ratio, and the horizontal line surrounding it represents a
95-percent confidence interval. If the left endpoint of a confidence
interval exceeds 1.0, then the corresponding result is statistically
significant at a 95-percent confidence level.
    The two charts show that the risk of lung cancer has consistently
been elevated for exposed workers and that the results are not
significantly different within each occupational category. Differences
in the magnitude and statistical significance of results within
occupation are not surprising, since the groups studied differed in
size, average exposure intensity and duration, and the time allotted
for latent effects.

BILLING CODE 4510-43-P

[[Page 5827]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.069

[[Page 5828]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.070

BILLING CODE 4510-43-C

[[Page 5829]]

    As documented in Subsection 2.c.i(2)(a) of this risk assessment,
all of the studies showing negative associations were either based on
relatively short observation or follow-up periods, lacked good
information about dpm exposure, involved low duration or intensity of
dpm exposure, or, because of inadequate sample size or latency
allowance, lacked the power to detect effects of the magnitude found in
the ``positive'' studies. Boffetta et al. (1988, p. 404) noted that, in
addition, studies failing to show a statistically significant
association--

    * * * often had low power to detect any association, had
insufficient latency periods, or compared incidence or mortality
rates among workers to national rates only, resulting in possible
biases caused by the ``healthy worker effect.''

    Some commenters noted that limitations such as insufficient
duration of exposure, inadequate latency allowance, small worker
populations, exposure misclassification, and comparison to external
populations with no adjustment for a healthy worker effect may explain
why not all of the studies showed a statistically significant
association between dpm exposure and an increased prevalence of lung
cancer. According to these commenters, if an epidemiologic study shows
a statistically significant result, this often occurs in spite of
methodological weaknesses rather than because of them. MSHA agrees that
limitations such as those listed make it more difficult to obtain a
statistically significant result when a real relationship exists.
(b) Best Available Epidemiologic Evidence
    As explained above, it is statistically significant that 41 of the
47 available epidemiologic studies reported an elevated risk of lung
cancer for workers exposed to dpm. MSHA finds it even more informative,
however, to examine the collective results of the eight studies
identified in Section III.2.c.i(2)(a) as providing the best currently
available epidemiologic evidence. These studies, selected using the
criteria described earlier, are: Boffetta et al. (1988), Boffetta et
al. (1990), Bruske-Hohlfeld et al. (1999), Garshick et al. (1987),
Garshick et al. (1988, 1991), Johnston et al. (1997), Steenland et al.
(90, 92, 98), and Saverin et al., (1999). All eight of these studies
reported an increased risk of lung cancer for workers with the longest
diesel exposures and for those most likely to have been exposed,
compared to unexposed workers. Tables showing the results from each of
these studies are provided in Section III.2.c.1(2)(a).
    The sign test of statistical significance can also be applied to
the collective results of these eight studies. If there were no
underlying association between exposure to diesel exhaust and an
increased risk of lung cancer, or anything else systematically favoring
a positive result, then there should be equal probabilities (equal to
one-half) that any one of these eight studies would turn out positive
or negative. Therefore, under the null hypothesis that positive and
negative results are equally likely, the probability that all eight
studies would show either a positive or a negative association is
(0.5)\8\ = 0.0039, or 0.39 percent. This shows that the collective
results of the eight studies comprising the best available
epidemiologic evidence are statistically significant at a confidence
level exceeding 99 percent (i.e., 100-2 x 0.39).
    When the risk of disease or death increases in response to higher
cumulative exposures, this is described by a ``positive'' exposure-
response relationship. Like consistency of results, the existence of a
positive exposure-response relationship is important in establishing
that the exposures in question actually cause an increase in risk.
Among the eight studies MSHA has identified as comprising the best
available epidemiologic evidence, there are five that provide evidence
of increasing lung cancer risk with increasing cumulative exposure:
Boffetta, et al. (1990), Bruske-Hohlfeld et al. (1999), Johnston et al.
(1997), Saverin et al. (1999), and Steenland et al. (1990, 1992, 1998).
The results supporting such a relationship are provided in the table
accompanying discussion of each of these studies in Section
III.2.c.i(2)(a).
    Although some have interpreted the results from the two studies by
Garshick et al. as also providing evidence of a positive exposure-
response relationship (e.g., Cal-EPA, 1998), this interpretation is
highly sensitive to the statistical models and techniques used to
analyze the data (HEI, 1999; Crump 1999). Therefore, for purposes of
this risk assessment, MSHA is not relying on Garshick et al. (1987) or
Garshick et. al (1988, 1991) to demonstrate the existence of a positive
exposure-response relationship. MSHA used the study for purposes of
hazard identification only. The Garshick studies contributed to the
weight of evidence favoring a causal interpretation, since they show
statistically significant excesses in lung cancer risk for the exposed
workers.
    The relative importance of the five studies identified in
demonstrating the existence of a positive exposure-response
relationship varies with the quality of exposure assessment. Boffetta
et al. (1990) and Bruske-Hohlfeld et al. (1999) were able to show such
a relationship based on the estimated duration of occupational exposure
for exposed workers, but quantitative measures of exposure intensity
(i.e., dpm concentration) were unavailable. Although duration of
exposure is frequently used as a surrogate of cumulative exposure, it
is clearly preferable, as many commenters pointed out, to base
estimates of cumulative exposure and exposure-response analyses on
quantitative measurements of exposure levels combined with detailed
work histories. Positive exposure-response relationships based on such
data were reported in all three studies: Johnston et al. (1997),
Steenland et al. (1998), and Saverin et al. (1999).
(c) Studies With Quantitative or Semiquantitative Exposure Assessments
    Several commenters stressed the fact that most of the available
epidemiologic studies contained little or no quantitative information
on diesel exposures and that those studies containing such information
(such as Steenland et al., 1998) generated it using questionable
assumptions. Some commenters also faulted MSHA for insufficiently
addressing this issue. For example, one commenter stated:

    * * * the Agency fails to highlight the lack of acceptable (or
any) exposure measurements concurrent with the 43 epidemiology
studies cited in the Proposed Rule. * * * the lack of concurrent
exposure data is a significant deficiency of the epidemiology
studies at issue and is a major factor that prevents application of
those epidemiology results to risk assessment. [EMA]

    MSHA agrees that the nature and quality of exposure information
should be an important consideration in evaluating the strength of
epidemiologic evidence. That is why MSHA included exposure assessment
as one of the criteria used to evaluate and rank studies in Section
2.c.1(2)(a) of this risk assessment. Two of the most recent studies,
both conducted specifically on miners, utilize concurrent, quantitative
exposure data and are included among the eight in MSHA's selection of
best available epidemiologic evidence (Johnston et al., 1997 and
Saverin et al., 1999). As a practical matter, however, epidemiologic
studies rarely have concurrent exposure measurements; and, therefore,
the commenter's line of

[[Page 5830]]

reasoning would exclude nearly all of the available studies from this
risk assessment--including all six of the negative studies. Since
Section 101(a)(6) of the Mine Act requires MSHA to consider the ``best
available evidence'' (emphasis added), MSHA has not excluded studies
with less-than-ideal exposure assessments, but, instead, has taken the
quality of exposure assessment into account when evaluating them. This
approach is also consistent with the recognition by the HEI Expert
Panel on Diesel Emissions and Lung Cancer that ``regulatory decisions
need to be made in spite of the limitations and uncertainties of the
few studies with quantitative data currently available'' (HEI, 1999;
p.39).
    The degree of quantification, however, is not the only relevant
consideration in evaluating studies with respect to exposure
assessment. MSHA also considered the likely effects of potential
exposure misclassification. As expressed by another commenter:

    * * * [S]tudies that * * * have poor measures of exposure to
diesel exhaust have problems in classification and will have weaker
results. In the absence of information that misclassification is
systematic or differential, in which case study results would be
biased towards either positive or no-effect level, it is reasonable
to assume that misclassification is random or nondifferentiated. If
so, * * * study results are biased towards a risk ratio of 1.0, a
ratio showing no association between diesel exhaust exposure and the
occurrence of lung cancer. [Dr. James Weeks, representing UMWA]

In her review of Bhatia et al. (1998), Silverman (1998) proposed that
``[o]ne approach to assess the impact of misclassification would be to
exclude studies without quantitative or semiquantitative exposure
data.'' According to Dr. Silverman, this would leave only four studies
among those considered by Dr. Bhatia: Garshick et al. (1988),
Gustavsson et al. (1990), Steenland et al. (1992), and Emmelin et al.
(1993).\66\ All four of these studies showed higher rates of lung
cancer for the workers estimated to have received the greatest
cumulative exposure, as compared to workers who had accumulated little
or no diesel exposure. Statistically significant results were reported
in three of these four studies. Furthermore, the two more recent
studies utilizing fully quantitative exposure assessments (Johnston et
al., 1997; Saverin et al., 1999) were not evaluated or otherwise
considered in the articles by Drs. Bhatia and Silverman. Like the other
four studies, these too reported elevated rates of lung cancer for
workers with the highest cumulative exposures. Specific results from
all six of these studies are presented in Tables III-4 and III-5.
---------------------------------------------------------------------------

    \66\ Emmelin et al. (1993) was considered but excluded from the
meta-analysis by Bhatia et al. (1998) for reasons explained by the
authors.
---------------------------------------------------------------------------

    Once again, the sign test of statistical significance can be
applied to the collective results of the four studies identified by Dr.
Silverman plus the two more recent studies with quantitative exposure
assessments. As before, under the null hypothesis of no underlying
effect, the probability would equal one-half that any one of these six
studies would turn out positive or negative. The probability that all
six studies would show either a positive or a negative association
would, under the null hypothesis, be (0.5) 6 = 0.0156, or
1.56 percent. This shows that the collective results of these six
studies, showing an elevated risk of lung cancer for workers estimated
to have the greatest cumulative exposure, are statistically significant
at a confidence level exceeding 96 percent (i.e., 100-2 x 1.56).
    As explained in the previous subsection, three studies showing
evidence of increased risk with increasing exposure based on
quantitative or semi-quantitative exposure assessments are included in
MSHA's selection of best available epidemiologic evidence: Johnston et
al. (1997), Steenland et al. (1998), and Saverin et al. (1999). Not
only do these studies provide consistent evidence of elevated lung
cancer risk for exposed workers, they also each provide evidence of a
positive exposure-response relationship--thereby significantly
strengthening the case for causality.
(d) Studies Involving Miners
    Eleven studies involving miners are summarized and discussed in
Section 2.c.i(2)(a) of this risk assessment. Commenters' observations
and criticisms pertaining to the individual studies in this group are
also addressed in that section. Three of these studies are among the
eight in MSHA's selection of best available epidemiologic evidence:
(Boffetta et al., 1988; Johnston et al., 1997; Saverin et al., 1999).
All three of these studies provide evidence of an increased risk of
lung cancer for exposed miners. Although MSHA places less weight on the
remaining eight studies, seven of them show some evidence of an excess
lung cancer risk among the miners involved. The remaining study
(Christie et al., 1995) reported a greater all-cause SMR for the coal
miners involved than for a comparable population of petroleum workers
but did not compare the miners to a comparable group of workers with
respect to lung cancer.
    The NMA submitted a review of six of these studies by Dr. Peter
Valberg, who concluded that ``[t]hese articles do not implicate diesel
exhaust, per se, as strongly associated with lung cancer in miners * *
* The reviewed studies do not form a consistent and cohesive picture
implicating diesel exhaust as a major risk factor for miners.''
Similarly, Dr. Jonathan Borak reviewed six of the studies on behalf of
MARG and concluded:

    [T]he strongest conclusion that can be drawn from these six
studies is that the miners in those studies had an increased risk of
lung cancer. These studies cannot relate such increased [risk] to
any particular industrial exposure, lifestyle or combination of such
factors.

Apparently, neither Dr. Valberg nor Dr. Borak disputed MSHA's
observation that the miners involved in the studies they reviewed
exhibited, overall, an excess risk of lung cancer. It is possible that
any excess risk found in epidemiologic studies may be due to extraneous
unknown or uncontrolled risk factors (i.e., confounding variables).
However, neither Drs. Valberg or Borak, nor the NMA or MARG, offered
evidence, beyond a catalog of speculative possibilities, that the
excess lung cancer risk for these miners was due to anything other than
dpm exposure.
    Nevertheless, MSHA agrees that the studies reviewed by Drs. Valberg
and Borak do not, by themselves, conclusively implicate dpm exposure as
the causal agent. Miners are frequently exposed to other occupational
hazards associated with lung cancer, such as radon progeny, and it is
not always possible to distinguish effects due to dpm exposure from
effects due to these other occupational hazards. This is part of the
reason why MSHA did not restrict its consideration of evidence to
epidemiologic studies involving miners. What implicates exposure to
diesel exhaust is the fact that diesel-exposed workers in a variety of
different occupations, under a variety of different working conditions
(including different types of mines), and in a variety of different
geographical areas consistently exhibit an increased risk of lung
cancer.
    Drs. Valberg and Borak did not review the two studies that utilize
quantitative dpm exposure assessments: Johnston et al. (1997) and
Saverin et al. (1999). In recently received comments Dr. Valberg,
writing for the NMA brought up four issues on the Saverin et al. 1999.
These issues were potential exposure misclassification, potential flaws
in the sampling method, potential smoker

[[Page 5831]]

misclassification, and insufficient latency. Two of these issues have
already been extensively discussed in section 2.c.i.2.a.ii and
therefore will not be repeated here. Dr. Valberg suggested that the
potential flaw in the sampling method would tend to over-estimate
exposure and that there was insufficient latency. If, in fact, both of
these issues are relevant, they would act to UNDERESTIMATE the lung
cancer risk in this cohort instead of OVERESTIMATE it. MSHA regards
these, along with Boffetta et al. (1988), Burns and Swanson (1991),\67\
and Lerchen et al. (1987) to be the most informative of the available
studies involving miners. Results on miners from these five studies are
briefly summarized in the following table, with additional details
provided in Section 2.c.1(2)(a) and Tables III-4 and III-5 of this risk
assessment. The cumulative exposures at which relative risks from the
Johnston and Saverin studies are presented are equivalent, assuming
that TC constitutes 80 percent of total dpm. The cumulative dpm
exposure of 6.1 mg-yr/m 3 is the multiplicative product of
exposure duration and dpm concentration for the most highly exposed
workers in each of these two studies.
---------------------------------------------------------------------------

    \67\ Listed in Table III-5 under Swanson et al., 1993.

BILLING CODE 4510-43-P

[[Page 5832]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.071

BILLING CODE 4510-43-C

[[Page 5833]]

    Although MSHA places less weight on the studies by Burns and
Swanson and by Lerchen than on the other three, it is significant that
the five best available studies involving miners all support an
increased risk of lung cancer attributable to dpm exposure.
(2) Meta-Analyses
    MSHA recognizes that simply tabulating epidemiologic studies as
positive or negative can sometimes be misleading. There are generally a
variety of outcomes that could render a study positive or negative,
some studies contain different analyses of related data sets, some
studies involve multiple comparisons of various subgroups, and the
studies differ widely in the reliability of their results. Therefore,
MSHA is not limiting its assessment of the epidemiologic evidence to
such a tabulation or relying only on the sign test described above.
MSHA has also considered the results of two statistical meta-analyses
covering most of the available studies (Lipsett and Campleman, 1999;
Bhatia et al., 1998). These meta-analyses weighted and pooled
independent results from those studies meeting certain inclusion
requirements to form overall estimates of relative risk for exposed
workers based on the combined body of data. In addition to forming
pooled estimates of the effect of diesel exposure, both meta-analyses
analyzed sources of heterogeneity in the individual results and
investigated but rejected publication bias as an explanation for the
generally positive results reported. Both meta-analyses derived a
statistically significant increase of 30 to 40 percent in the risk of
lung cancer, attributable to occupational dpm exposure.
    Lipsett and Campleman (1999) systematically analyzed and combined
results from most of the studies summarized in Tables III-4 and III-5.
Forty-seven studies published between 1957 and 1995 were identified for
initial consideration. Some studies were excluded from the pooled
analysis because they did not allow for a period of at least 10 years
for the development of clinically detectable lung cancer. Others were
excluded because of bias resulting from incomplete ascertainment of
lung cancer cases in cohort studies or because they examined the same
cohort population as another study. One study was excluded because
standard errors could not be calculated from the data presented. The
remaining 30 studies, contributing a total of 39 separate estimates of
exposure effect (for distinct occupational groups within studies), were
analyzed using a random-effects analysis of variance (ANOVA) model.
    Potential effects of publication bias (i.e., the likelihood that
papers with positive results may be more likely to be published than
those with negative results) were investigated by plotting the
logarithm of relative risk estimated from each study against its
estimated precision, as expressed by the inverse of its standard error.
According to the authors, the resulting ``funnel plot'' was generally
consistent with the absence of significant publication bias, although
there were relatively few small-scale, statistically insignificant
studies. The investigators performed a further check of potential
publication bias by comparing results of the included studies with the
only relevant unpublished report that became available to them during
the course of their analysis. Smoking-adjusted relative risks for
several diesel-exposed occupations in the unpublished study were,
according to the investigators, consistent with those found in the
studies included in the meta-analysis.
    Each of the 39 separate estimates of exposure effect was weighted
by a factor proportional to its estimated precision. Sources of
heterogeneity in results were investigated by subset analysis--using
categorical variables to characterize each study's design, target
population (general or industry-specific), occupational group, source
of control or reference population, latency, duration of exposure,
method of ascertaining occupation, location (North America or Europe),
covariate adjustments (age, smoking, and/or asbestos exposure), and
absence or presence of a clear healthy worker effect (as manifested by
lower than expected all-cause mortality in the occupational population
under study).
    Sensitivity analyses were conducted to evaluate the sensitivity of
results to inclusion criteria and to various assumptions used in the
analysis. This included (1) substitution of excluded ``redundant''
studies of the same cohort population for the included studies and (2)
exclusion of studies involving questionable exposure to dpm. An
influence analysis was also conducted to examine the effect of dropping
one study at a time, to determine if any individual study had a
disproportionate effect on results of the ANOVA.
    The pooled relative risk from all 39 exposure effects (estimated
from 30 studies) was RR = 1.33, with a 95-percent confidence interval
(CI) extending from 1.21 to 1.46. For the subgroup of 13 smoking-
adjusted exposure effects (nine studies) from populations ``most likely
to have had substantial exposure'' to dpm, the pooled effect was RR =
1.47, with a CI from 1.29 to 1.67. Based on the all of the various
analyses they conducted, the authors concluded:

    Although substantial heterogeneity existed in the initial pooled
analysis, stratification on several factors substantially reduced
heterogeneity, producing subsets of studies with increased relative
risk estimates that persisted through various influence and
sensitivity analyses. * * *
    In studies that adjusted for confounding by cigarette smoking,
not only did the positive association between diesel exhaust
exposure and lung cancer persist but the pooled risk estimate showed
a modest increase, with little evidence of heterogeneity.
    * * * [T]his meta-analysis provides quantitative evidence
consistent with several prior reviews, which have concluded that the
epidemiologic evidence supports a causal relationship between
occupational exposure to diesel exhaust and lung cancer. [Lipsett
and Campleman, 1999]

    The other meta-analysis was conducted by Bhatia et al. (1998) on
epidemiologic studies published in peer-reviewed journals between 1957
and 1993. In this analysis, studies were excluded if actual work with
diesel equipment ``could not be confirmed or reliably inferred'' or if
an inadequate latency period was allowed for cancer to develop, as
indicated by less than 10 years from time of first exposure to end of
follow-up. Studies of miners were also excluded, because of potential
exposure to radon and silica. Likewise, studies were excluded if they
exhibited selection bias or examined the same cohort population as a
study published later. A total of 29 independent results on exposure
effects from 23 published studies were identified as meeting the
inclusion criteria.
    To address potential publication bias, the investigators identified
several unpublished studies on truck drivers and noted that elevated
risks for exposed workers observed in these studies were similar to
those in the published studies utilized. Based on this and a ``funnel
plot'' for the included studies, the authors concluded that there was
no indication of publication bias.
    After assigning each of the 29 separate estimates of exposure
effect a weight proportional to its estimated precision, Bhatia et al.
(1998) used a fixed-effects ANOVA model to calculate pooled relative
risks based on the following groupings: all 29 results; all case-
control studies; all cohort studies; cohort studies using internal
reference populations; cohort studies making external comparisons;
studies adjusted for smoking; studies not adjusted for smoking; and
studies grouped by occupation (railroad workers,

[[Page 5834]]

equipment operators, truck drivers, and bus workers). Elevated risks of
lung cancer were shown for exposed workers overall and within every
individual group of studies analyzed. A positive duration-response
relationship was observed in those studies presenting results according
to employment duration. The weighted, pooled estimates of relative risk
were identical for case-control and cohort studies and nearly identical
for studies with or without smoking adjustments.
    The pooled relative risk from all 29 exposure effects (estimated
from 23 studies) was RR = 1.33, with a 95-percent confidence interval
(CI), adjusted for heterogeneity, extending from 1.24 to 1.44. For just
the smoking-adjusted studies, it was 1.35 (CI: 1.20 to 1.52); and for
cohort studies making internal comparisons, it was 1.43 (CI: 1.29 to
1.58). Based on their evaluation of the all the analyses on various
subgroups, Bhatia et al. (1998) concluded that the elevated risk of
lung cancer observed among exposed workers was unlikely to be due to
chance, that confounding from smoking was unlikely to explain all of
the excess risk, and that ``this meta-analysis supports a causal
association between increased risks for lung cancer and exposure to
diesel exhaust.''
    The pooled relative risks estimated in both meta-analyses equal
1.33 and exceed 1.4 for studies making internal comparisons, or
comparisons to similar groups of workers. Both meta-analyses found
these results to be statistically significant, meaning that they cannot
be explained merely by random or unexplained variability in the risk of
lung cancer that occurs among both exposed and unexposed workers.
Although both meta-analyses relied, by necessity, on an overlapping
selection of studies, the inclusion criteria were different and some
studies included in one meta-analysis were excluded from the other.
They used different statistical models for deriving a pooled estimate
of relative risk, as well as different means of analyzing heterogeneity
of effects. Nevertheless, they derived the same estimate of the overall
exposure effect and found similar sources of heterogeneity in the
results from individual studies.\68\ One commenter observed that--
---------------------------------------------------------------------------

    \68\ Several commenters suggested that because the two meta-
analyses both received direct or indirect funding from the same
governmental agency, they were not independently conducted. These
commenters speculated that Dr. Allan Smith, a co-author of Cal-EPA
(1998) and Bhatia et al. (1998), contributed to both meta-analyses.
Although an earlier version of Lipsett and Campleman (1999) appeared
as an appendix to Cal-EPA (1998), commenters provided no evidence
that Dr. Smith contributed anything to that appendix. Dr. Smith is
not listed as a co-author of Lipsett and Campleman (1999).

    Lung cancer relative risks for occupational ``control groups''
vary over a range from 0.4 to 2.7 * * *. Therefore, the level of
relative risks being reported in the dpm epidemiology fall within
---------------------------------------------------------------------------
this level of natural variation. [IMC Global]

This argument is refuted by the statistical significance of the
elevation in risk detected in both meta-analyses in combination with
the analyses accounting for heterogeneity of exposure effects.
    The EMA objected that MSHA's focus on these two meta-analyses
``presents an incomplete picture because the counter-arguments of
Silverman (1998) were not discussed in the same detail.'' IMC global
also faulted MSHA for dismissing Dr. Silverman's views without adequate
explanation.
    In her review,\69\ Dr. Silverman characterized Bhatia et al. (1998)
as a ``careful meta-analysis'' and acknowledged that it ``add[s] to the
credibility that diesel exhaust is carcinogenic * * *.'' She also
explicitly endorsed several of its most important conclusions. For
example, Dr. Silverman stated that ``[t]he authors convincingly show
that potential confounding by cigarette smoking is likely to have
little impact on the estimated RRs for diesel exhaust and lung
cancer.'' She suggested, however, that Bhatia et al. (1998)
``ultimately do not resolve the question of causality.'' (Silverman,
1998)
---------------------------------------------------------------------------

    \69\ Silverman (1998) reviewed Bhatia et al. (1998) but not
Lipsett and Campleman (1999) or the earlier version of that meta-
analysis (Lipsett and Alexeeff, 1998) cited in MSHA's proposed
preamble.
---------------------------------------------------------------------------

    Dr. Silverman imposed an extremely high standard for what is needed
to ultimately resolve the question of causality. The precise question
she posed, along with her answer, was as follows:

    Has science proven causality beyond any reasonable doubt?
Probably not. [Silverman, 1998, emphasis added.]

Neither the Mine Act nor applicable case law requires MSHA to prove
causality ``beyond any reasonable doubt.'' The burden of proof that Dr.
Silverman would require to close the case and terminate research is not
the same burden of proof that the Mine Act requires to warrant
protection of miners subjected to far higher levels of a probable
carcinogen than any other occupational group. In this risk assessment,
MSHA is evaluating the collective weight of the best available
evidence--not seeking proof ``beyond any reasonable doubt.'' \70\
---------------------------------------------------------------------------

    \70\ It is noteworthy that, in describing research underway that
might resolve the issue of causality, Dr. Silverman stressed the
need for studies with quantitative exposure measurements and stated
that ``underground miners may, in fact, be the most attractive group
for study because their exposure to diesel exhaust is at least five
times greater than that of previously studied occupational groups.''
(Silverman, 1998) She then mentioned a study on underground miners
in Germany that had recently been initiated. The study of German
underground potash miners (Saverin et al., 1999), published after
Dr. Silverman's article, utilizes quantitative exposure measurements
and is included in MSHA's selection of best available epidemiologic
evidence (see Section 3.a.iii(1)(a) of this risk assessment). MSHA
also includes in that selection another underground miner study
utilizing quantitative exposure measurements (Johnston et al.,
1997). The 1997 study was available prior to Dr. Silverman's article
but is not listed among her references.
---------------------------------------------------------------------------

    The EMA objected to MSHA's reliance on the two meta-analyses
because of ``* * * serious deficiencies in each'' but did not, in
MSHA's opinion, identify any such deficiencies. The EMA pointed out
that ``most of the original studies in each were the same, and the few
that were not common to each were not of significance to the outcome of
either meta-analysis.'' MSHA does not regard this as a deficiency.
Since the object of both meta-analyses was to analyze the available
epidemiologic evidence linking dpm exposure with lung cancer, using
defensible inclusion criteria, it is quite understandable that they
would rely on overlapping information. The principal differences were
in the types and methods of statistical analysis used, rather than in
the data subjected to analysis; and MSHA considers it informative that
different approaches yielded very similar results and conclusions. It
is noteworthy, moreover, that both of the meta-analyses explicitly
addressed the EMA's concern by performing analyses on various different
sub-groupings of the available studies. The sensitivity of results to
the inclusion criteria was also explicitly investigated and considered.
MSHA believes that the conclusions of these meta-analyses did not
depend on unreasonable inclusion or exclusion criteria.
    The EMA also argued that--

    [a] meta-analysis cannot compensate for basic deficiencies in
the studies used to create the meta-analysis, and this fact is not
clearly stated by MSHA. Instead, MSHA follows the tack of the meta-
analysis authors, who claim that the meta-analysis somehow overcomes
deficiencies of the individual studies selected and presents a
stronger case. This is simply not true. [EMA]

MSHA agrees that a meta-analysis cannot correct for all deficiencies
that may be present in individual studies. It

[[Page 5835]]

can, however, correct for certain types of deficiencies. For example,
individual studies may lack statistical power because of small study
populations. By pooling results from several such studies, a meta-
analysis may achieve a level of statistical significance not attainable
by the individual studies. Furthermore, both of the meta-analyses used
well-defined inclusion criteria to screen out those studies with the
most severe deficiencies. In addition, they both found that it was the
more rigorous and technically more valid studies that reported the
strongest associations between excess lung cancer and dpm exposure.
They also performed separate analyses that ruled out inflationary
effects of such ``deficiencies'' as lack of a smoking adjustment. For
example, Lipsett and Campleman (1999) reported a pooled RR = 1.43 for
20 smoking-adjusted results, as compared to a pooled RR = 1.25 for 19
results with no smoking adjustment.
    IMC Global and MARG submitted five specific criticisms of the meta-
analyses, to which MSHA will respond in turn.

(1) Publication Bias

    * * * both studies * * * rely only on published studies. * * *
the authors rely on statistical analysis in an attempt to uncover
possible publication bias. * * * the only safeguard to protect
against possible publication bias is to seek out unpublished results
* * *. [IMC Global]

    Both meta-analyses compared the results of published and
unpublished studies and found them to be similar. Bhatia et al. (1998)
found several unpublished studies of lung cancer among truck drivers
that ``* * * were not included in our analysis; however the risk ratios
of these studies are similar to the [sic] those in published studies
among truck drivers.'' (Bhatia et al., p. 90) Lipsett and Campleman
(1999) checked ``[s]moking-adjusted relative risks for several diesel-
exposed occupations'' in an unpublished report on U.S. veterans and
found them ``* * * consistent with those reported here.'' They remarked
that ``although publication bias cannot be completely ruled out, it is
an unlikely explanation for our findings.'' (Lipsett and Campleman, p.
1015) In addition to comparing results directly against unpublished
studies, both meta-analyses used the statistical method of ``funnel
plots'' as an indirect means of checking for the existence of
significant publication bias. It should also be noted that MSHA did not
exclude unpublished studies from this risk assessment.

(2) Selection Bias

    * * * [the] meta-analyses have to provide a much more convincing
rationale as to why all miners were excluded even when the
confounders that are mentioned are not likely or important, for
example in studies conducted in potash and salt mines. * * * IMC
Global sees no reason why the older studies of potash workers
[Waxweiler et al., 1973] and more recent studies on New South Wales
coal miners [Christie et al., 1995] should not be included * * *.
[IMC Global]
    Studies were selectively included or excluded, without good or
sufficient explanation. [MARG]

    Contrary to the commenters' characterization, both meta-analyses
listed each study excluded from the analysis of pooled relative risk
and gave a good reason for its exclusion. For example, both meta-
analyses excluded studies that failed to allow for a minimum 10-year
latency period for lung cancer to develop after first exposure. With
respect to the exclusion of all studies on miners, Bhatia et al. (1998)
pointed out that ``[s]ince studies of miners often indicate higher
relative risks for lung cancer than those considered in this meta-
analysis, this was a conservative exclusion.'' Even if studies on
miners had been considered, Waxweiler et al. (1973) and Christie et al.
(1995) would have been excluded from both meta-analyses because of
their failure to meet the 10-year minimum latency requirement.

(3) Lack of Actual Exposure Data

    * * * [N]ondifferential exposure or disease misclassification
can sometimes produce bias away from the null * * * Thus, tests for
heterogeneity performed in both these meta-analyses won't detect or
correct this problem. [IMC Global]

    Lipsett and Campleman acknowledged that ``[e]xposure
misclassification is a problem common to all studies of cancer and
diesel emissions. In no case were there direct measurements of
historical diesel exhaust exposures of the subjects.'' However, as Dr.
Silverman pointed out in her review, ``* * * this bias is most likely
to be nondifferential, and the effect would probably have been to bias
point estimates toward the null value. Thus the summary RR of 1.33 may
be an underestimate of the true lung cancer effect associated with
diesel exposure.'' (Silverman, 1998)

(4) Smoking as a Confounder

    * * * The use of data manipulation and modeling adjustments in
both these meta-analyses cannot rectify the flaws in the initial
studies. [IMC Global]
    * * * misclassification of this exposure [cigarette smoking]
could result in residual confounding of individual studies and,
consequently, meta-analyses, of those studies. [MARG]

    Contrary to the commenter's suggestion, neither of the meta-
analyses made any attempt to manipulate or adjust the data in order to
rectify what the commenter regards as ``flaws'' in the way smoking or
other potential confounders were treated in the initial studies. Both
meta-analyses, however, compared the pooled RR for studies with a
smoking adjustment to the pooled RR for studies without any such
adjustment. Both meta-analysis calculated a pooled RR for the smoking-
adjusted studies greater than or equal to that for the unadjusted
studies. In addition, Bhatia et al. (1998) analyzed the impact of the
smoking adjustment for the subgroup of studies reporting results both
with and without such an adjustment and found that the ``small
reduction in the pooled RR estimates would not be consistent with a
major effect from residual confounding.'' Dr. Silverman concluded that
``[t]he authors convincingly show that potential confounding by
cigarette smoking is likely to have little impact on the estimated RRs
for diesel exhaust and lung cancer.'' (Silverman, 1998)

(5) Inadequate Control in the Underlying Studies for Diet

    As noted by Lipsett and Campleman, ``Diet may also confound the
diesel-lung cancer association.'' The researchers also caution that
this risk factor was not controlled for in the nearly 50 diesel
studies they examined. [MARG]

    Since inhalation is the primary route of dpm exposure, and the lung
is the primary target organ, MSHA considers potential dietary
confounding to be of minor importance in the diesel-lung cancer
association. Lipsett and Campleman acknowledged that diet might be a
relevant consideration for long-haul truck drivers, but stated that
``diet would probably not be an important confounder in studies of
other occupations, particularly those using internal or other
occupationally active reference populations.'' Studies making internal
comparisons, or comparisons to similar groups of workers, are unlikely
to be seriously confounded by dietary differences, because the groups
of workers being compared are likely to have very similar dietary
habits, on average. The pooled relative risk for cohort studies making
comparisons internally or to other active workers was 1.48 (95% CI =
1.28 to 1.70). (Lipsett and Campleman, 1999, Table 3) This was
considerably higher than the pooled RRs for studies making comparisons
against regional or national populations, where dietary differences

[[Page 5836]]

(and also differences with respect to other potential confounders)
would be more important.
(3) Potential Systematic Biases
    Citing failure to account for dietary differences as an example,
some commenters argued that the meta-analyses may simply propagate
weaknesses shared by the individual studies. These commenters contended
that many of the studies MSHA considered in this risk assessment share
methodological similarities and that, therefore, a ``deficiency''
causing bias in one study would probably also bias many other studies
in the same direction. According to these commenters, no matter how
great a majority of studies report a 30- to 40-percent increase in the
risk of lung cancer for exposed workers, the possibility of systematic
bias prevents the collective evidence from being strong or sufficient.
    Although this point has some theoretical foundation, it has no
basis in fact for the particular body of epidemiologic evidence
relating lung cancer to diesel exposure. The studies considered were
carried out by many different researchers, in different countries,
using different methods, and involving a variety of different
occupations. Elevated risk was found in cohort as well as case-control
studies, and in studies explicitly adjusting for potential confounders
as well as studies relying on internal comparisons within homogeneous
populations. The possibility that systematic bias explains these
results is also rendered less plausible by results from studies of a
radically different type: the elevated risk of lung cancer associated
with chronic environmental exposures to PM2.5 (Dockery et
al. 1993; Pope et al., 1995).
    Furthermore, the commenters advancing this argument presented no
evidence that the studies shared any deficiencies of a type that would
systematically shift results in the direction of showing a spurious
association. As explained in Subsection 2.c.i(2)(a), exposure
misclassification, healthy worker effect, and low power due to
insufficient latency generally have the opposite effect--systematically
diluting and masking results. Although many studies may share a similar
susceptibility to bias by dietary differences or residual smoking
effects,\71\ there is no reason to expect that such effects will
consistently bias results in the same direction, across all occupations
and geographic regions.
---------------------------------------------------------------------------

    \71\ The term ``residual smoking effects'' refers to the
potentially confounding effects of smoking that may remain after a
smoking adjustment has been made.
---------------------------------------------------------------------------

    Associations between dpm exposure and excess lung cancer are
evident in a wide variety of occupational and geographical contexts,
and it is unlikely that all (or most) would be biased in the same
direction by lifestyle effects. There is no reason to suppose that, in
nearly all of these studies, exposed subjects were more likely than
unexposed subjects to have lifestyles (apart from their occupations)
that increased their risk of lung cancer. On the other hand, exposures
to other occupational carcinogens, such as asbestos dust, radon
progeny, and silica, could systematically cause studies in which they
are not taken into account to exhibit spurious associations between
lung cancer and occupational diesel exhaust exposures. Silica dust and
radon progeny are frequently present in mining environments (though not
usually in potash mines), and this was the reason that studies on
miners were excluded from the two meta-analyses.
    IMC Global argued that because of the possibility of being misled
by systematic biases, epidemiologic evidence can be used to identify
only those hazards that, at a minimum, double the risk of disease
(i.e., RR  2.0). IMC Global explained this viewpoint by
quoting an epidemiologist as follows:

    * * * [E]pidemiologic methods can only yield valid documentation
of large relative risks. Relative risks of low magnitude (say, less
than 2) are virtually beyond the resolving power of the
epidemiologic microscope. We can seldom demonstrably eliminate all
sources of bias, and we can never exclude the possibility of
unidentified and uncontrolled confounding. If many studies--
preferably based on different methods--are nevertheless congruent in
producing markedly elevated relative risks, we can set our
misgivings aside. If however, many studies produce only modest
increases, those increases may well be due to the same biases in all
the studies. [Dr. Samuel Shapiro, quoted by IMC Global]

    It is important to note that, unlike IMC Global, Dr. Shapiro did
not suggest that results of RR  2.0 be counted as ``negative.'' He
contended only that low RRs do not completely rule out the possibility
of a spurious association due to unidentified or uncontrolled
confounding. More importantly, however, this restriction would allow
workers to be exposed to significant risks and is, therefore,
unacceptable for regulatory purposes. For purposes of protecting miners
from lung cancer, certainty is not required; and an increase in the
relative risk of less than 100 percent can increase the absolute risk
of lung cancer by a clearly unacceptable amount. For example, if the
baseline risk of lung cancer is six per thousand, then increasing it by
33 percent amounts to an increase of two per thousand for exposed
workers.
    IMC Global went on to argue that--

    * * * only a few of these studies have relative risks that
exceed 2.0, and some of the studies that do exceed 2.0 exhibit
biases that make them unsuitable for rulemaking purposes in our
opinion. * * * Thus, in IMC Global's opinion, the epidemiologic
evidence demonstrates an artificial association that can be
explained through common biases probably due to smoking habits and
lifestyle factors. [IMC Global]

    This line of reasoning leaps from the possibility that systematic
biases might account for observed results to a conclusion that they
actually do so. Furthermore, after proposing to allow for possible
biases by requiring that only relative risks in excess of 2.0 be
counted as positive evidence, IMC global has ignored its own criterion
and discounted results greater than 2.0 for the same reason. Contrary
to IMC Global's claim that ``only a few of the studies have relative
risks that exceed 2.0,'' Tables III-4 and III-5 show 23 separate
results greater than 2.0, applying to independent categories of workers
in 18 different studies.
    According to Stober and Abel (1996), the potential confounding
effects of smoking are so strong that ``residual smoking effects''
could explain even statistically significant results observed in
studies where smoking was explicitly taken into account. MSHA agrees
that variable exposures to non-diesel lung carcinogens, including
relatively small errors in smoking classification, could bias
individual studies. However, the potential confounding effect of
tobacco smoke and other carcinogens can cut in either direction.
Spurious positive associations of dpm exposure with lung cancer would
arise only if the group exposed to dpm had a greater exposure to these
confounders than the unexposed control group used for comparison. If,
on the contrary, the control group happened to be more exposed to
confounders, then this would tend to make the association between dpm
exposure and lung cancer appear negative. Therefore, although smoking
effects could potentially distort the results of any single study, this
effect could reasonably be expected to make only about half the studies
that were explicitly adjusted for smoking come out positive. Smoking is
unlikely to have been responsible for finding an excess prevalence of
lung cancer in 17 out of 18 studies in which a smoking adjustment was
applied. Based on a 2-tailed sign test, this possibility can be

[[Page 5837]]

rejected at a confidence level greater than 99.9 percent.
    Even in the 29 studies for which no smoking adjustment was made,
tobacco smoke and other carcinogens were important confounders only to
the extent that the populations exposed and unexposed to diesel exhaust
differed systematically with respect to these other exposures. Twenty-
four of these studies, however, reported some degree of excess lung
cancer risk for the diesel-exposed workers. This result could be
attributed to other occupational carcinogens only in the unlikely event
that, in nearly all of these studies, diesel-exposed workers happened
to be more highly exposed to these other carcinogens than the control
groups of workers unexposed to diesel.
    Like IMC Global, Stober and Abel (1996) do not, in MSHA's opinion,
adequately distinguish between a possible bias and an actual one.
Potential biases due to extraneous risk factors are unlikely to account
for a significant part of the excess risk in all studies showing an
association. Excess rates of lung cancer were associated with dpm
exposure in all epidemiologic studies of sufficient size and scope to
detect such an excess. Although it is possible, in any individual
study, that the potentially confounding effects of differential
exposure to tobacco smoke or other carcinogens could account for the
observed elevation in risk otherwise attributable to diesel exposure,
it is unlikely that such effects would give rise to positive
associations in 41 out of 47 studies. As stated by Cohen and Higgins
(1995):

    * * * elevations [of lung cancer] do not appear to be fully
explicable by confounding due to cigarette smoking or other sources
of bias. Therefore, at present, exposure to diesel exhaust provides
the most reasonable explanation for these elevations. The
association is most apparent in studies of occupational cohorts, in
which assessment of exposure is better and more detailed analyses
have been performed. The largest relative risks are often seen in
the categories of most probable, most intense, or longest duration
of exposure. In general population studies, in which exposure
prevalence is low and misclassification of exposure poses a
particularly serious potential bias in the direction of observing no
effect of exposure, most studies indicate increased risk, albeit
with considerable imprecision. [Cohen and Higgins (1995), p. 269].

    Several commenters identified publication bias as another possible
explanation for the heavy preponderance of studies showing an elevated
risk of lung cancer for exposed workers. As described earlier, both of
the available meta-analyses investigated and rejected the hypothesis of
significant publication bias affecting the overall results. This was
based on both a statistical technique using ``funnel plots'' and a
direct comparison between results of published and unpublished studies.
Commenters presented no evidence that publication bias actually exists
in this case. After the 1988 NIOSH and 1989 IARC determinations that
diesel exhaust was a ``potential'' or ``probable'' human carcinogen,
negative results would have been of considerable interest, and, in the
absence of any evidence specifically applying to dpm studies, there is
no reason to assume they would not have been published.

(4) Causality

    MSHA must draw its conclusions based on the weight of evidence. In
the absence of any statistical evidence for differential confounding or
significant publication bias, the weight of epidemiologic evidence
strongly favors a causal connection. On the one side, it is evident
that virtually all of the studies that adjusted for smoking and other
known confounders, or controlled for them by comparing against similar
groups of workers, showed positive associations (i.e., relative risk or
odds ratio > 1.0). Also on this side of the balance are all eight of
the studies MSHA identified as comprising the best available human
evidence. These include three studies reporting positive exposure-
response relationships based on quantitative dpm exposure assessments:
two recent studies specifically on underground miners (one coal and one
potash) and one on trucking industry workers.\72\ On the other side of
the balance is the possibility that publication bias or other
systematic biases may have been responsible for some unknown portion of
the overall 30- to 40-percent elevation in lung cancer risk observed--a
possibility that, while conceivable, is based on speculation. After
considering other viewpoints (addressed here and in the next
subsection), MSHA has accepted what in its view is the far more likely
alternative: that the vast majority of epidemiologic studies showed an
elevated risk in association with occupational exposures to diesel
exhaust because such exposures cause the risk of lung cancer to
increase. The toxicity experiments discussed in Subsection 2.d.iv of
this risk assessment support the causal interpretation that MSHA has
placed on the associations observed in epidemiologic studies.
---------------------------------------------------------------------------

    \72\ These studies (respectively: Johnston et al., 1997; Saverin
et al., 1999; Steenland et al., 1998) are discussed in detail in
Subsection 2.c.i(2)(a) of this risk assessment.
---------------------------------------------------------------------------

    In this risk assessment, MSHA is basing its conclusions primarily
on epidemiologic studies. However, the results obtained from animal
studies confirm that diesel exhaust can increase the risk of lung
cancer in some species and help show that dpm (rather than the gaseous
fraction of diesel exhaust) is the causal agent. The fact that dpm has
been proven to cause lung cancer in laboratory rats only under
conditions of lung overload does not make the rat studies irrelevant to
miners. The very high dpm concentrations currently observed in some
mines could impair or even overwhelm lung clearance for miners already
burdened by respirable mineral dusts, thereby inducing lung cancer by a
mechanism similar to what occurs in rats (Nauss et al., 1995). It must
also be noted, however, that most of the human studies show an
increased risk of lung cancer at dpm levels lower than what might be
expected to cause overload. Therefore, the human studies suggest that
overload is not a necessary condition for dpm to induce or promote lung
cancer among humans. Salvi et al. (1999) reported marked inflammatory
responses in the airways of healthy human volunteers after just one
hour of exposure to dpm at a concentration of 300 g/m\3\.
Animal studies provide evidence that inhalation of dpm has related
effects, such as induction of free oxygen radicals, that could promote
the development of human lung cancers by mechanisms not requiring lung
overload. (See Sec. III.2.d.iv(2).)
    Similarly, the weight of genotoxicity evidence helps support a
causal interpretation of the associations observed in the epidemiologic
studies. This evidence shows that dpm dispersed by alveolar surfactant
can have mutagenic effects, thereby providing a genotoxic route to
carcinogenesis that is independent of overloading the lung with
particles. After a comprehensive review of the evidence, IPCS (1996)
concluded that both the particle core and the associated organic
materials have biological activity. The biological availability of
carcinogens present in the organic portion of dpm may, however, differ
significantly in different species. Chemical byproducts of
phagocytosis, which occurs even when the lung is not overloaded, may
provide another genotoxic route. Inhalation of diesel emissions has
been shown to cause DNA adduct formation in peripheral lung cells of
rats and monkeys, and increased levels of human DNA adducts have been
found in association with occupational exposures. (See Sec.
III.2.d.iv(1)) None of this evidence

[[Page 5838]]

suggests that a lung cancer threshold exists for humans exposed to dpm,
despite its importance in the rat model. Nor does this evidence suggest
that lung overload is necessary for dpm to induce lung cancer in
humans. Indeed, lung overload may be only one of many mechanisms
through which lung cancer is produced in humans.
    Results from the epidemiologic studies, the animal studies, and the
genotoxicity studies are coherent and mutually supportive. After
considering all these results, MSHA has concluded that the
epidemiologic studies, supported by the experimental data establishing
the plausibility of a causal connection, provide strong evidence that
chronic occupational dpm exposure increases the risk of lung cancer in
humans.
    In a review, submitted by MARG, of MSHA's proposed risk assessment,
Dr. Jonathan Borak asserted that MSHA's determination that results from
the epidemiologic and toxicity studies were ``coherent and mutually
reinforcing'' involved circular reasoning. He supported this assertion
by incorrectly attributing to MSHA the view that ``most of the
individual [epidemiologic] studies are not very good'' and that their
suggestion of an association between dpm and lung cancer is ``made
credible in light of the animal data.'' To complete his argument that
MSHA relied on circular reasoning, Dr. Borak then suggested that the
epidemiologic data provided MSHA's sole basis for considering the
animal data relevant to humans. In a similar vein, Kennecott Minerals
claimed there was an ``absence of toxicological support for
epidemiologic findings that are themselves inconclusive.''
    Contrary to Dr. Borak's assertion, MSHA has not characterized most
of the epidemiologic studies as ``not very good.'' Nor has MSHA
suggested that the epidemiologic evidence would not be credible or
plausible in the absence of supporting animal data. As Dr. Borak
correctly noted, MSHA acknowledged that ``none of the existing human
studies is perfect'' and that ``no single one of the existing
epidemiological studies, viewed in isolation, provides conclusive
evidence of a causal connection * * *.'' That a study is not
``perfect,'' however, does not imply that it is ``not very good.''
MSHA's position has consistently been that, as demonstrated by the two
available meta-analyses, the collective epidemiologic evidence is not
merely credible but statistically significant and indicative of a
causal association. Although MSHA views the toxicity data as supporting
and reinforcing the epidemiologic evidence, MSHA believes that the
collective epidemiologic evidence is highly credible in its own right.
    Furthermore, MSHA does not consider the animal data relevant to
humans simply because of the positive epidemiologic evidence. The
animal evidence is also credible in its own right. As MSHA has
repeatedly pointed out, dust concentrations in some mines have been
measured at levels of the same order of magnitude as those found to
have caused lung cancer in rats. Such high exposures, especially when
combined with occupational exposures to respirable mineral dusts and
exposures to particles in tobacco smoke, could overload the human lung
and promote lung cancer by a mechanism similar to that hypothesized for
rats. (Hattis and Silver, 1992, Figures 9, 10, 11). Also, many of the
animal experiments have elucidated genotoxic effects that, while
apparently not responsible for the excess lung cancers observed for
rats, may be responsible for some or all of the excess risk reported
for humans.
    MSHA has not relied on circular reasoning. If either the animal
data or the toxicity data had failed to show any link between dpm and
effects implicated in the induction or promotion of lung cancer, then
MSHA's conclusion would have been weakened. The existence of
experimental evidence confirming that there is such a link is not
imaginary and is logically independent of the epidemiologic evidence.
Therefore, contrary to Dr. Borak's characterization, the ``coherency
and reinforcement'' arising from the epidemiologic, animal, and
genotoxicity data are not the product of circular reasoning. A more apt
description is that the three sources of evidence, like three legs of a
tripod, support the same conclusion.
    Many commenters argued that a causal connection between dpm
exposure and an increased human risk of lung cancer should not be
inferred unless there is epidemiologic evidence showing a positive
exposure-response relationship based on quantitative measures of
cumulative dpm exposure. MSHA does not agree that a quantitative
exposure-response relationship is essential in establishing causality.
Such a relationship is only one of several factors, such as consistency
and biological plausibility, that epidemiologists examine to provide
evidence of causality. As mentioned earlier, however, there are three
studies providing quantitative exposure-response relationships. One of
these studies (Steenland et al., 1998) controlled for age, race,
smoking, diet, and asbestos exposure, but relied on ``broad
assumptions'' to estimate historical exposure levels from later
measurements. Two of the studies, however, (Johnston et al., 1997, and
Saverin et al., 1999) utilized measurements that were either
contemporaneous with the exposures (Johnston) or that were made under
conditions very similar to those under which the exposures took place
(Saverin). Both of these studies were conducted on underground miners.
The Saverin study used exposure measurements of total carbon (TC). All
three of the studies combined exposure measurements for each job with
detailed occupational histories to form estimates of cumulative dpm
exposure; and all three reported evidence of increasing lung cancer
risk with increasing cumulative exposure.
    Several commenters, expressing and endorsing the views of Dr. Peter
Valberg, incorrectly asserted that the epidemiologic results obtained
across different occupational categories were inconsistent with a
biologically plausible exposure-response relationship. For example,
MARG argued that--

    It is biologically implausible that, if dpm were (causally)
increasing lung cancer risk by 50% for a low exposure (say, truck
drivers), then the lung cancer risk produced by dpm exposure in more
heavily exposed worker populations (railroad shop workers) would
fall in this same range of added risk. The added lung-cancer risk
for bus garage workers is half that of either railroad workers or
truck drivers, but dpm concentrations are considerably higher.
[MARG]

Earlier, MARG had argued to the contrary that, due to their lack of
concurrent exposure measurements, these studies could not reliably be
used for hazard identification. MARG then attempted to use them to
perform the rather more difficult task of making quantitative
comparisons of relative risk. If cumulative exposures are unknown, as
MARG argued elsewhere, then there is little basis for comparing
responses at different cumulative exposures.
    In an analysis submitted by the West Virginia Coal Association, Dr.
Valberg extended this argument to miners as follows:

    * * * If dpm concentrations for truck drivers is in the range of
5-50 g/m3, then we can assign the 0.49 excess
risk (Bhatia's meta-analysis result) to the 5-50 g/
m3 exposure. Hence, dpm concentrations for miners in the
range of 100-2,000 g/m3 should have yielded
excess risks forty times larger, meaning that the RR for exposed
miners would be expected to be about 21 (i.e., 1 + 19.6), whereas
reported risk estimates are less than 3 (range from 0.74

[[Page 5839]]

2.67). Such an utter lack of concordance argues against a causal
role for dpm in the reported epidemiologic associations.

    Based on a similar line of reasoning, IMC Global asserted that ``*
* * the assumptions that MSHA used to develop [Figure III-4] * * * do
not do make sense in the context of a dose-response relationship
between lung cancer and dpm exposure.'' This was one of the reasons IMC
Global gave for objecting to MSHA's comparison (in Section III.1.d) of
exposure levels measured for miners to those reported for different
occupations. IMC Global proposed that, as a consequence of this
argument, MSHA should delete this comparison from its risk assessment.
    MSHA sees three major flaws in Dr. Valberg's argument and rejects
it for the following reasons:
    (1) The argument glosses over the important distinction between
exposure concentrations (intensity) and cumulative exposure (dose).
Total cumulative exposure is the product of intensity and duration of
exposure. Depending on duration, high intensity exposure may result in
similar (or even lower) cumulative exposure than low intensity
exposure. Furthermore, different industries, in different nations,
introduced diesel equipment at different times. The studies being
considered were carried out in a variety of different countries and
covered a variety of different historical periods. Therefore, the same
number of years in different studies can correspond to very different
durations of occupational exposure.
    Many of the miners in the studies Dr. Valberg considered may have
been occupationally exposed to dpm for relatively short periods of time
or even not at all. Various forms of exposure misclassification would
tend to obscure any exposure-response relationship across industries.
Such obscuring would result from both exposure misclassification within
individual studies and also variability in the degree of exposure
misclassification in different industries.
    Furthermore, the exposure levels or intensities assigned to the
various occupations would not necessarily be proportional to cumulative
exposures, even if the average number of years of exposure were the
same. Different job conditions, such as longer-than-average work hours,
could have major, variable impacts on cumulative exposures. For
example, lower dpm concentrations have been measured for truck drivers
than for other occupationally-exposed workers. But as a group, the
truck drivers who were studied, due to their work conditions, may have
been in their trucks for longer than the standard 40-hour work week and
therefore have larger cumulative dpm exposures. These truck drivers
commonly congregated in parking areas and slept in their trucks with
the engines idling, thereby disproportionately increasing their
cumulative dpm exposures compared to miners and other types of workers.
    (2) The commenters advancing this argument assumed that an
exposure-response relationship spanning occupations at different levels
of exposure intensity would take the form of a straight line. This
assumption is unwarranted, since carcinogens do not necessarily follow
such a simple pattern across a broad range of exposure levels. There is
little basis for assuming that the relationship between cumulative dpm
exposures and the relative risk of lung cancer would appear as a
straight line when plotted against exposure levels that may differ by a
factor of 100. Steenland et al. (1998) reported a better statistical
``fit'' to the data using a model based on the logarithm of cumulative
exposure as compared to simple cumulative exposure. Even across the
relatively limited range of exposures within the trucking industry, the
logarithmic exposure model exhibits pronounced curvature towards the
horizontal at the higher cumulative exposures (Steenland et al., 1998,
Fig. 5). If this model is extrapolated out to the much higher exposures
currently found in underground mining, then (as shown in Subsection
3.b.ii(3)(b) of this risk assessment) it diverges even more from a
straight-line model. Toxicological evidence of curvature in the dose-
response relationship has also been reported (Ichinose et al., 1997b,
p. 190).
    Furthermore, the exposure-response pattern may depend on other
aspects of exposure, besides how much is accumulated. For example, the
National Research Council (NRC) has adopted a risk model for radon-
induced lung cancer in which the relative risk (RR) at any age depends
on both accumulated exposure and the rate (reflecting the intensity of
exposure) at which total exposure was accumulated. In this model, which
was derived empirically from the epidemiologic data, exposures
accumulated over long time periods at relatively low rates result in a
greater risk of lung cancer than the same total exposures accumulated
over shorter time periods at relatively higher rates (NRC, 1999). A
similar effect for dpm could cause apparent anomalies in the pattern of
relative risks observed for occupations ranked simply with respect to
the intensity of their average exposures.
    (3) Mean exposures and relative risks reported for miners involved
in the available studies were mischaracterized. Although dpm levels as
high as 2000 g/m3 have been measured in some mines,
the levels at most mines surveyed by MSHA were substantially lower (see
Figures III-1 and III-2). The average levels MSHA measured at
underground mines were 808 g/m3 and 644 g/
m3 for M/NM and coal mines using diesel equipment for face
haulage, respectively (Table III-1). However, these were not
necessarily the levels experienced by miners involved in the available
studies. The mean TC exposure concentration reported by Saverin et al.
(1999), for work locations having the highest mean concentration, was
390 g/m3--corresponding to a mean dpm concentration
of about 490 g/m3. In the only other study
involving miners for which exposure measurements were available,
Johnston et al. (1997) reported dpm concentrations for the most highly
exposed category of workers (locomotive drivers), ranging from 44
g/m3 to 370 g/m3. Therefore,
the mean dpm concentration experienced by the most highly exposed
miners involved in these two studies was not ``forty times larger''
than the level imputed to truck drivers, but closer to seven times
larger.\73\ Applying Dr. Valberg's procedure, this yields an
``expected'' relative risk of about 4.4 for the underground miners who
happened to work at mines included in these particular studies (1 +
7 x (0.49)). Miners exposed at higher levels would, of course, face a
greater risk.
---------------------------------------------------------------------------

    \73\ The estimate of seven times larger dpm exposure in miners
is the result of averaging data from Saverin et al. (1999) with data
from Johnston et al. (1997) and comparing the combined average miner
dpm exposure to the average truck driver dpm exposure.
---------------------------------------------------------------------------

    Dr. Valberg asserted that the highest relative risk reported for
miners was 2.67 (from Boffetta et al., 1988). Dr. Valberg failed to
note, however, that the upper 95-percent confidence limit for miners'
relative risk in this study was 4.37, so that this result hardly
qualifies as an ``utter lack of concordance'' with the 4.4 ``expected''
value for miners. Furthermore, even higher relative risks for miners
have been reported in other studies. Burns and Swanson (1991) reported
5.0 for operators of mining machinery, with an upper 95-percent
confidence limit of 16.9. The relative risk estimated for the most
highly exposed miners in the study by Johnston et al. (1997) was either
5.5 or 11.0, depending on the statistical model used. These results
appear to be quite consistent with the data for truck drivers.

[[Page 5840]]

(5) Other Interpretations of the Evidence

    After reviewing the same body of scientific evidence as MSHA, Dr.
Peter Valberg came to a very different conclusion with respect to the
likelihood of causality:

    Flawed methodology (lack of adequate control for smoking);
values for relative risks (``RR'') that are low and often not
statistically elevated above 1.0; inadequate treatment of sources of
variability; reliance on multiple comparisons; and inadequate
control over how authors choose to define dpm exposure surrogates
(that is, job category within a profession, cumulative years of
work, age at time of exposure, etc.), all undermine the assignment
of causality to dpm exposure.

    On the other hand, many scientific organizations and governmental
agencies have reviewed the available epidemiologic and toxicological
evidence for carcinogenicity and, in accordance with MSHA's conclusion,
identified dpm as a probable human carcinogen--at levels far lower than
those measured in some mines--or placed it in a comparable category.
These include:

YEAR

    2000  National Toxicology Program (NTP);
    1999  (tentative) U.S. Environmental Protection Agency (EPA)
    1998  (tentative) (American Conference of Governmental
Industrial Hygienists (ACGIH); Currently on Y2K NIC list. Probable
vote in 10/2000.
    1998  California Environmental Protection Agency (Cal-EPA);
    1998  Federal Republic of Germany;
    1996  International Programme on Chemical Safety (IPCS), a joint
venture of the World Health Organization, the International Labour
Organization, and the United Nations Environment Programme;
    1989  International Agency for Research on Cancer (IARC);
    1988  National Institute for Occupational Safety and Health
(NIOSH).

    Nevertheless, several commenters strongly objected to MSHA's
conclusion, claiming that the evidence was obviously inadequate and
citing scientific authorities who, they claimed, rejected MSHA's
inference of a causal connection. In some cases, views were
inaccurately attributed to these authorities, and misleading quotations
were presented out of context. For example, the Nevada Mining
Association stated that its own review of the scientific literature led
to--

    * * * the only reasonable conclusion possible: there is no
scientific consensus that there is a causal link between dpm
exposure and lung cancer. The HEI [1999 Expert Panel] report
concludes that the causal link between diesel exhaust and lung
cancer remains unproven, and that further study and analysis are
clearly required. [Nevada Mining Assoc.]

Although HEI (1999) recommended further study and analysis for purposes
of quantitative risk assessment, the report contains no findings or
conclusions about the ``causal link.'' To the contrary, the report
explicitly states that the panel ``* * * was not charged to evaluate
either the broad toxicologic or epidemiologic literature concerning
exposure to diesel exhaust and lung cancer for hazard identification
purposes, which has been done by others.'' (HEI, 1999, p. 1)
Furthermore, the HEI panel ``* * * recognize[d] that regulatory
decisions need to be made in spite of the limitations and uncertainties
of the few studies with quantitative data currently available.'' (HEI,
1999, p. 20)
    MARG, along with the Nevada Mining Association and several other
commenters, mischaracterized the Expert Panel's findings as extending
beyond the subject matter of the report. This report was limited to
evaluating the suitability of the data compiled by Garshick et al.
(1987, 1988) and Steenland et al. (1990, 1992, 1998) for quantitative
risk assessment. Contrary to the characterization by these commenters,
HEI's Expert Panel explicitly stated:

    [The Panel] was not charged to evaluate the broad toxicologic or
epidemiologic literature for hazard identification purposes, which
has been done by others. State, national, and international agencies
have all reviewed the broader animal and human evidence for
carcinogenicity and, in either their draft or final reports, have
all identified diesel exhaust as [a] probable human carcinogen or
placed it in a comparable category.'' [HEI, 1999, p. 1]

The Panel then identified most of the organizations and governmental
institutions listed above (HEI, 1999, p. 8).
    One commenter (MARG) also grossly misrepresented HEI (1999) as
having stated that ``the available epidemiologic work has `study design
flaws, including uncontrolled, confounding and lack of exposure
measures, leading to a lack of convincing evidence.' '' (MARG post-
hearing comments) The opinion falsely attributed to HEI was taken from
a sentence in which HEI's Diesel Epidemiology Expert Panel was
describing opinions expressed in ``[s]ome reviews critical of these
data.'' (HEI, 1999, p. 10) The Panel did not suggest that these
opinions were shared by HEI or by any members of the Panel. In fact,
the cited passage came at the end of a paragraph in which the Panel
cited a larger number of other review articles that had ``discusse[d]
this literature in depth'' and had expressed no such opinions. In the
same paragraph, the Panel confirmed that ``[t]he epidemiologic studies
generally show higher risks of lung cancer among persons occupationally
exposed to diesel exhaust than among persons who have not been exposed,
or who have been exposed to lower levels or for shorter periods of
time.'' (HEI, 1999, p. 10)
    Several commenters noted that the U.S. EPA's Clean Air Scientific
Advisory Committee (CASAC) issued a report (CASAC, 1998) critical of
the EPA's 1998 draft Health Assessment Document for Diesel Emissions
(EPA, 1998) and rejecting some of its conclusions. After the HEI (1999)
Expert Panel report was published, the EPA distributed a revised draft
of its Health Assessment Document (EPA, 1999). In the 1999 draft, the
EPA characterized human exposures to diesel exhaust as ``highly
likely'' to be carcinogenic to humans at ambient (i.e., environmental)
exposure levels. After reviewing this draft, CASAC endorsed a
conclusion that, at ambient levels, diesel exhaust is likely to be
carcinogenic to humans. Although CASAC voted to recommend that the
designation in the EPA document be changed from ``highly likely'' to
``likely,'' this change was recommended specifically for ambient rather
than occupational exposures. The CASAC report states that ``[a]lthough
there was mixed opinion regarding the characterization of diesel
emissions as `highly likely' to be a human carcinogen, the majority of
the Panel did not agree that there was sufficient confidence (i.e.,
evidence) to use the descriptor `highly' in regard to environmental
exposures.'' (CASAC, 2000, emphasis added)
    MSHA recognizes that not everyone who has reviewed the literature
on lung cancer and diesel exposure agrees about the collective weight
of the evidence it presents or about its implications for regulatory
decisions. IMC Global, for example, stated:

    After independently reviewing most [of the] * * * epidemiologic
studies, the literature reviews and the two meta-analyzes, IMC
Global believes * * * MSHA has misrepresented the epidemiologic
evidence in the Proposed Rule. The best conclusion that we can reach
based on our review of this information is that different reputable
studies reach conflicting conclusions * * *. [IMC Global]

IMC Global continued by expressing concern that MSHA had ``dismissed''
opposing arguments critical of the positive studies, especially
``regarding lack of statistical significance; small magnitudes of
relative risk * * *; and the impact of confounding factors, especially
smoking * * *. [IMC Global]''

[[Page 5841]]

    MSHA has addressed these three issues, as they relate to the
evaluation of individual studies, in Section 2.c.i(2)(a) of this
preamble. The argument that confounding factors such as smoking may
have been systematically responsible for the positive results was
discussed above, under the heading of ``Potential Systematic Biases.''
Statistical significance of the collective evidence is not the same
thing as statistical significance of individual studies. Application of
the sign test, as described Subsection 3.a.iii(1) above, is one way
that MSHA has addressed statistical significance of the collective
evidence. Another approach was also described above, under the heading
of ``Meta-Analyses.''
    IMC Global quoted Morgan et al. (1997) as concluding that
``[a]lthough there have been a number of papers suggesting that diesel
fumes may act as a carcinogen, the weight of the evidence is against
this hypothesis.'' This conclusion was based largely on the authors'
contention, shared by IMC Global, that the epidemiologic results were
inconsistent and of insufficient strength (i.e., RR  2.0) to rule out
spurious associations due to potential confounders. MSHA, on the other
hand, interprets the epidemiologic studies as remarkably consistent,
given their various limitations, and has argued that the strength of
evidence from individual studies is less important than the strength of
evidence from all studies combined. Dr. Debra Silverman has referred to
the ``striking consistency'' of this evidence. (Silverman, 1998)
    Ironically, Morgan et al. point out many of the very limitations in
individual studies that may actually explain why the studies do not
yield entirely equivalent results. The 1997 Morgan article was written
before the meta-analyses became available and resolved many, if not
all, of the apparent inconsistencies in the epidemiologic results.
Since none of the existing human studies is perfect and many contain
important limitations, it is not surprising that reported results
differ in magnitude and statistical significance. The meta-analyses
described earlier showed that the more powerful and carefully designed
studies tended to show greater degrees of association. MSHA has
addressed the joint issues of consistency and strength of association
above, under the heading of ``Consistency of Epidemiologic Evidence.''
    The Engine Manufacturers Association (EMA) quoted Cox (1997) as
concluding: ``* * * there is no demonstrated biological basis for
expecting increased risk at low to moderate levels of [diesel]
exposure.'' (Cox, 1997, as quoted by EMA] The EMA, however, prematurely
terminated this quotation. The quoted sentence continues: ``* * * low
to moderate levels of exposure (those that do not lead to lasting soot
deposits, chronic irritation, and perhaps GSH enzyme depletion in the
lung).'' MSHA does not regard concentrations of dpm exceeding 200
g/m\3\ as ``low to moderate,'' and the EMA presented no
evidence that the effects Dr. Cox listed do not occur at the high
exposure levels observed at some mines. Salvi et al. (1999) reported
marked inflammatory responses in the airways of healthy human
volunteers after just one hour of exposure to dpm at a concentration of
300 g/m\3\. The deleted caveat ending the quotation is
especially important in a mining context, since mine atmospheres
generally contain respirable mineral dusts that may diminish clearance
rates and contribute to meeting thresholds for chronic irritation and
inflammation leading to oxidative damage. Based on miners' testimony at
the public hearings and workshops, there is, in fact, reason to believe
that exposed miners experience lasting soot deposits and chronic
irritation as a result of their exposures.
    With respect to the epidemiologic evidence, the EMA quoted Dr. Cox
as concluding: ``* * * among studies that demonstrate an increased
relative risk, it appears plausible that uncontrolled biases in study
design and data analysis methods can explain the statistical increases
in relative risk without there being a true causal increase.'' (Cox,
1997, quoted by EMA) Dr. Cox refers to non-causal explanations for
positive epidemiologic results as ``threats to causal inference.'' In
considering Dr. Cox's discussion of the evidence, it is important to
bear in mind that his purpose was ``* * * not to establish that any (or
all) of these threats do explain away the apparent positive
associations between [dpm] and lung cancer risk * * * but only to point
out that they plausibly could * * *.'' (Cox, 1997, p. 813) Dr. Cox's
stated intent was to identify non-causal characteristics of positive
studies that could potentially ``explain away'' the positive results.
This is a relatively simple exercise that could misleadingly be applied
to even the strongest of epidemiologic studies. As stated earlier, no
epidemiologic study is perfect, and it is always possible that unknown
or uncontrolled risk factors may have given rise to a spurious
association. Neither the EMA nor Dr. Cox pointed out however, that
there are characteristics common to the negative studies that plausibly
explain why they came out negative: insufficient latency allowance,
nondifferential exposure misclassification, inappropriate comparison
groups (including healthy worker effect, negative confounding by
smoking or other variables. A similar approach could also be used to
explain why many of the positive studies did not exhibit stronger
associations. As observed by Dr. Silverman, ``an unidentified negative
confounder may have produced bias across studies, systematically
diluting RRs.''

b. Significance of the Risk of Material Impairment to Miners

    The fact that there is substantial and persuasive evidence that dpm
exposure can materially impair miner health in several ways does not
imply that miners will necessarily suffer such impairments at a
significant rate. This section will consider the significance of the
risk faced by miners exposed to dpm.
i. Meaning of Significant Risk
(1) Legal Requirements
    The benzene case, cited earlier in this risk assessment, provides
the starting point for MSHA's analysis of this issue. Soon after its
enactment in 1970, OSHA adopted a ``consensus'' standard for exposure
to benzene, as authorized by the OSH Act. The standard set an average
exposure limit of 10 parts per million over an 8-hour workday. The
consensus standard had been established over time to deal with concerns
about poisoning from this substance (448 U.S. 607, 617). Several years
later, NIOSH recommended that OSHA alter the standard to take into
account evidence suggesting that benzene was also a carcinogen. (Id. at
619 et seq.). Although the ``evidence in the administrative record of
adverse effects of benzene exposure at 10 ppm is sketchy at best,''
OSHA was operating under a policy that there was no safe exposure level
to a carcinogen. (Id., at 631). Once the evidence was adequate to reach
a conclusion that a substance was a carcinogen, the policy required the
agency to set the limit at the lowest level feasible for the industry.
(Id. at 613). Accordingly, the Agency proposed lowering the permissible
exposure limit to 1 ppm.
    The Supreme Court rejected this approach. Noting that the OSH Act
requires ``safe or healthful employment,'' the court stated that--

    * * * `safe' is not the equivalent of `risk-free' * * * a
workplace can hardly be considered ``unsafe'' unless it threatens
the

[[Page 5842]]

workers with a significant risk of harm. Therefore, before he can
promulgate any permanent health or safety standard, the Secretary 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. [Id., at 642,
italics in original].

The court went on to explain that it is the Agency that determines how
to make such a threshold finding:

    First, the requirement that a `significant' risk be identified
is not a mathematical straitjacket. It is the Agency's
responsibility to determine, in the first instance, what it
considered to be a `significant' risk. Some risks are plainly
acceptable and others are plainly unacceptable. If, for example, the
odds are one in a billion that a person will die from cancer by
taking a drink of chlorinated water, the risk clearly could not be
considered significant. On the other hand, if the odds are one in a
thousand that regular inhalation of gasoline vapors that are 2%
benzene will be fatal, a reasonable person might well consider the
risk significant and take appropriate steps to decrease or eliminate
it. Although the Agency has no duty to calculate the exact
probability of harm, it does have an obligation to find that a
significant risk is present before it can characterize a place of
employment as ``unsafe.'' [Id., at 655].

The court noted that the Agency's ``* * * determination that a
particular level of risk is `significant' will be based largely on
policy considerations.'' (Id., note 62).
    Some commenters contended that the concept of significant risk, as
enunciated by the Supreme Court in the Benzene case, requires support
by a quantitative dose-response relationship. For example, one
commenter argued as follows:

    * * * OSHA had contended in * * * [the benzene] case that
``because of the lack of data concerning the linkage between low-
level exposures and blood abnormalities, it was impossible to
construct a dose-response curve at this time''. 448 U.S. at 632-633.
The court rejected the Agency's attempt to support a standard based
upon speculation that ``the benefits to be derived from lowering''
the permissible exposure level from 10 to 1 ppm were `likely' to be
`appreciable'.'' 448 U.S. at 654.
    One year after the Benzene case, the Court in American Textile
Mfr's Inst. v. Donovan, 452 U.S. 490 (1981), upheld OSHA's ``cotton
dust'' standard for which a dose-response curve had been established
by the Agency. The Court relied upon the existence of such data to
find that OSHA had complied with the Benzene mandate, stating: ``In
making its assessment of significant risk, OSHA relied on dose-
response curve data * * * It is difficult to imagine what else the
agency could do to comply with this Court's decision in the Benzene
case.'' Id. at 505, n. 25. See also Public Citizen Research Group v.
Tyson, 796 F. 2d 1479, 1496, 1499 (D.C. Cir. 1986) (where a dose
response curve was constructed for the ethylene oxide standard and
the agency [had] gone to great lengths to calculate, within the
bounds of available scientific data, the significance of the risk);
United Steelworkers of America v. Marshall, 647 F. 2d 1189, 1248
(D.C. Cir. 1980), cert. denied, 453 U.S. 913 (1981) (where in
promulgating a new lead standard ``OSHA amassed voluminous evidence
of the specific harmful effects of lead at particular blood levels
and correlated these blood lead levels with air lead levels'').
[NMA]

    A dose-response relationship has been established between exposure
to PM2.5 (of which dpm is a major constituent) and the risk
of death from cardiovascular, cardiopulmonary, or respiratory causes
(Schwartz et al.,1996; EPA, 1996). Furthermore, three different
epidemiologic studies, including two carried out specifically on mine
workers, have reported evidence of a quantitative relationship between
dpm exposure and the risk of lung cancer (Johnston et al., 1997,
Steenland et al., 1998, Saverin et al., 1999). However, the Secretary
has carefully reviewed the legal references provided by the commenters
and finds there is no requirement in the law that the determination of
significant risk be based on such a relationship. The cited court
rulings appear to describe sufficient means of establishing a
significant risk, rather than necessary ones. Indeed, as stated earlier
in this section, the Benzene court explained that:

    * * * the requirement that a ``significant'' risk be identified
is not a mathematical straitjacket. It is the Agency's
responsibility to determine, in the first instance, what it
considered to be a ``significant'' risk. * * * the Agency has no
duty to calculate the exact probability of harm * * *.

    The Agency has set forth the evidence and rationale behind its
decision to propose a rule restricting miner exposure to dpm, obtained
an independent peer review of its assessment of that evidence,
published the evidence and tentative conclusions for public comment,
held hearings, kept the record open for further comments for months
after the hearings, and re-opened the record so that stakeholders could
comment on the most recent evidence available. Throughout these
proceedings, the Agency has carefully considered all public comments
concerning the evidence of adverse health effects resulting from
occupational dpm exposures. Based on that extensive record, and the
considerations noted in this section, the Agency is authorized under
the statute and relevant precedents to act on this matter--despite the
fact that a more conclusive or definitively established exposure-
response relationship might help address remaining doubts among some
members of the mining community.
    As the Supreme Court pointed out in the benzene case, the
appropriate definition of significance also depends on policy
considerations of the Agency involved. In the case of MSHA, those
policy considerations include special attention to the history of
extraordinary occupational risks leading to the Mine Act. That history
is intertwined with the toll to the mining community of silicosis and
coal workers' pneumoconiosis (CWP or ``black lung''), along with
billions of dollars in Federal expenditures.
(2) Standards and Guidelines for Risk Assessment
    Several commenters suggested that this risk assessment, as
originally proposed, deviated from established risk assessment
guidelines, because it did not provide a sufficiently quantitative
basis for evaluating the significance of miners's risks due to their
dpm exposures. One of these commenters (Dr. Jonathan Borak) maintained
that a determination of significant risk based on a ``qualitative''
assessment ``has no statistical meaning.''
    MSHA recognizes that a risk assessment should strive to provide as
high a degree of quantification and certainty as is possible, given the
best available scientific evidence. However, in order to best protect
miners' health, it is not prudent to insist on a ``perfect'' risk
assessment. Nor is it prudent to delay assessing potentially grave
risks simply because the available data may be insufficient for an
ideal risk assessment. The need for regulatory agencies to act in the
face of uncertainty was recognized by the HEI's Diesel Epidemiology
Expert Panel as follows: ``The Panel recognizes that regulatory
decisions need to be made in spite of the limitations and uncertainties
of the few studies with quantitative data currently available.'' (HEI,
1999) When there is good, qualitative evidence--such as the sight and
smell of heavy smoke--that one's house is on fire, an inference of
significant risk may be statistically meaningful even without
quantitative measurements of the smoke's density and composition.
    Moreover, as will be demonstrated below, the question of whether a
quantitative assessment is or is not essential is, in this case, moot:
this risk assessment does, in fact, provide a quantitative evaluation
of how significant the risk is for miners occupationally exposed to
dpm.

[[Page 5843]]

ii. Significance of Risk for Underground Miners Exposed to dpm
    An important measure of the significance of a risk is the
likelihood that an adverse effect actually will occur. A key factor in
the significance of risks that dpm presents to miners is the very high
dpm concentrations to which a number of those miners are currently
exposed--compared to ambient atmospheric levels in even the most
polluted urban environments, and to workers in diesel-related
occupations for which positive epidemiologic results have been
reported. Figure III-4 compared the range of median dpm exposure levels
measured for mine workers at various mines to the range of medians
estimated for other occupations, as well as to ambient environmental
levels. Figure III-7 presents a similar comparison, based on the
highest mean dpm level observed at any individual mine, the highest
mean level reported for any occupational group other than mining, and
the highest monthly mean concentration of dpm estimated for ambient air
at any site in the Los Angeles basin.\74\ As shown in Figure III-7,
underground miners are currently exposed at mean levels up to 10 times
higher than the highest mean exposure reported for other occupations,
and up to 100 times higher than the highest mean environmental level
even after adjusting the environmental level upwards to reflect an
equivalent occupational exposure.
---------------------------------------------------------------------------

    \74\ For comparability with occupational lifetime exposure
levels, the environmental ambient air concentration has been
multiplied by a factor of approximately 4.7. This factor reflects a
45-year occupational lifetime with 240 working days per year, as
opposed to a 70-year environmental lifetime with 365-days per year,
and assumes that air inhaled during a work shift comprises half the
total air inhaled during a 24-hour day.
---------------------------------------------------------------------------

    Given the significant increases in mortality and other acute health
effects associated with increments of 25 g/m3 in
fine particulate concentration (see Table III-3), the relative risk of
acute effects for some miners (especially those already suffering
respiratory problems) appears to be extremely high. Acute responses to
dpm exposures have been detected in studies of stevedores, whose
exposures were likely to have been less than one tenth the exposure of
some miners on the job. Likewise, the risk of lung cancer due to dpm
exposure would appear to be far greater for those underground miners
who are exposed at such high levels than for other workers or general
urban populations.

BILLING CODE 4510-43-P

[[Page 5844]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.072

BILLING CODE 4510-43-C

[[Page 5845]]

    Several commenters asserted that current dpm exposures in
underground mines are lower than they were when MSHA conducted its
field surveys and that MSHA had not taken this into account when
assessing the significance of dpm risk to miners. A related comment was
that MSHA had not designed its sampling studies to provide a
statistically representative cross section of the entire industry but
had nevertheless used the results in concluding that the risk to
underground miners was significant.
    In accordance with Sec. 101.(a)(6) of the Mine Act, MSHA is basing
this risk assessment on the best available evidence. None of the
commenters provided evidence that dpm levels in underground metal/
nonmetal mines had declined significantly since MSHA's field studies,
or provided quantitative estimates of any purported decline in average
dpm concentrations, or submitted data that would better represent the
range of dpm concentrations to which underground miners are typically
exposed at the present time. Although MSHA's field studies were not
designed to be statistically representative in a way that can be
readily quantified, they were performed at locations selected,
according to MSHA's best engineering judgement, to be typical of the
type of diesel equipment used. Furthermore, as will be shown below,
MSHA's evaluation of the significance of risks presented to underground
metal/nonmetal miners by their dpm exposures does not rely on the
highest levels, or even the average levels, that MSHA has measured. As
documented in Section 1.d of this risk assessment, some of the highest
of MSHA's measurements were made as recently as 1996-1997. It is
important to note, as is shown below, the cancer risks of dpm exposure
are clearly significant even at a concentration of 300 g/
m3--less than half of the average level that MSHA observed
in its field studies. Therefore, MSHA believes that a reduction in
exposure of more than 50 percent in the last couple of years is highly
implausible.

BILLING CODE 4510-43-P

[[Page 5846]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.073

BILLING CODE 4510-43-C

[[Page 5847]]

    Earlier in this risk assessment, MSHA identified three types of
material impairment that can result from occupational exposures to dpm.
The next three subsections present the Agency's evaluation of how much
of a risk there is that miners occupationally exposed to dpm will
actually incur such consequences. Each part addresses the risk of
incurring one of the three types of material impairment identified
earlier.
(1) Sensory Irritations and Respiratory Symptoms (Including Allergenic
Responses)
    It is evident from the direct testimony of numerous miners working
near diesel equipment that their exposures pose a significant risk of
severe sensory irritations and respiratory symptoms. This was
underscored during the workshops and public hearings by several miners
who noted that such effects occurred immediately and consistently after
episodes of intense exposure (Section 2.b.i). There is also persuasive
experimental evidence that exposure at levels found in underground
mines frequently cause eye and nose irritation (Rudell et al., 1996)
and pulmonary inflammation (Salvi et al., 1999). Section 2.a.ii and
3.a.i of this risk assessment explain why these effects constitute
``material impairments'' under the Mine Act and why they threaten
miners' safety as well as health. Therefore, it is clear that even
short-term exposures to excessive concentrations of dpm pose
significant risks.
    MSHA's quantitative evaluation of how significant the risks of
sensory irritations and respiratory symptoms are for miners is limited,
by the quantitative evidence available, to acute respiratory symptoms
linked to fine particulate exposures (PM2.5) in ambient air
pollution studies. MSHA recognizes that, for miners exposed to dpm,
this type of risk cannot be quantified with great confidence or
precision based on the available evidence. This is because
PM2.5 is not solely comprised of dpm and also because
miners, as a group, have different demographic and health
characteristics from the general populations involved in the relevant
studies. However, MSHA believes that the quantitative evidence suffices
to establish a lower bound on the significance of this type of risk to
miners exposed to dpm. Even at this lower bound, which is likely to
substantially underestimate the degree of risk, the probability that a
miner's occupational exposure to dpm will cause adverse respiratory
effects is clearly significant.
    As shown in Table III-3, the risk of acute lower respiratory tract
symptoms has been reported to increase, at a 95-percent confidence
level, by 15 to 82 percent (RR = 1.15 to 1.82) for each incremental
increase of 20 g/m\3\ in the concentration of PM2.5
in the ambient air. This means that the relative risk estimated for a
given PM2.5 concentration ranges between (1.15)\k\ and
(1.82)\k\, where k = the concentration of PM2.5 divided by
20 g/m\3\. For example, for a PM2.5 concentration
of 40 g/m\3\, the RR is estimated to be between (1.15)\2\ and
(1.82)\2\, or 1.32 to 3.31. MSHA believes that part of the reason why
the range is so wide is that the composition of PM2.5 varied
in the data from which the estimates were derived.
    MSHA acknowledges that there are substantial uncertainties involved
in converting 24-hour environmental exposures to 8-hour occupational
exposures. However, since mining often involves vigorous physical
activity (thereby increasing breathing depth and frequency) and sleep
is characterized by reduced respiration, it is highly likely that
miners would inhale at least one-third of their total 24-hour intake of
air during a standard 8-hour work shift. If it is assumed that the
acute respiratory effects of inhaling dpm at a concentration of 60
g/m\3\ over an 8-hour workshift are at least as great as those
at a concentration of 20 g/m\3\ over a 24-hour period, then it
is possible to estimate a lower bound on the relative risk of such
effects.
    Based solely on the fact that dpm consists almost entirely of
particles much smaller than 2.5 micrometers in diameter, the dpm would
be expected to penetrate the lower respiratory tract at least as
effectively as PM2.5. Also, given the complex chemical
composition of dpm, and its generation within a confined space, there
is no reason to suspect that dpm in an underground mining environment
is less potent than ambient PM2.5 in inducing respiratory
symptoms. Under these assumptions, a short-term environmental exposure
to PM2.5 at a concentration of 20 g/m\3\ would
correspond to a short-term occupational exposure to dpm at a
concentration of 60 g/m\3\. Consequently, the RR at an
occupational exposure level of Y g/m\3\ would equal the RR
calculated for an ambient exposure level of 20 x (Y/60) g/
m\3\. For example, the relative risk (RR) of acute lower respiratory
symptoms at an occupational exposure level of 300 g/m\3\ dpm
would, at a minimum, correspond to the RR at an ambient exposure level
equal to 5 x 20 g/m\3\ PM2.5. (See Table III-3) A
dpm concentration of 300 g/m\3\ happens to be the level at
which Salvi et al. (1999) found a marked pulmonary inflammatory
response in healthy human volunteers after just one hour of exposure.
    Under these assumptions, the risk of lower respiratory tract
symptoms for a miner exposed to dpm for a full shift at a concentration
of 300 g/m\3\ or more, would be at least twice the risk of
ambient exposure (i.e., RR = (1.15)\5\ = 2.01). This would imply that
for miners exposed to dpm at or above this level, the risk of acute
lower respiratory symptoms would double, at a minimum. The Secretary
considers such an increase in risk to be clearly significant.
(2) Premature Death From Cardiovascular, Cardiopulmonary, or
Respiratory Causes
    As in the case of respiratory symptoms, the nature of the best
available evidence limits MSHA's quantitative evaluation of how large
an excess risk of premature death, due to causes other than lung
cancer, there is for miners exposed to dpm. As before, this evidence
consists of acute effects linked to fine particulate exposures
(PM2.5) in ambient air pollution studies. Therefore, the
analysis is subject to similar uncertainties. However, also as before,
MSHA believes that the quantitative evidence suffices to place a lower
bound on the increase in risk of premature mortality for miners
occupationally exposed to dpm. As will be shown below, even this lower
bound, which is likely to substantially underestimate the degree of
increase, indicates that a miner's occupational exposure to dpm has a
clearly significant impact on the likelihood of premature death.
    Schwartz et al. (1996) found an average increase of 1.5 percent in
daily mortality associated with each increment of 10 g/m\3\ in
the daily concentration of fine particulates. Higher increases were
estimated specifically for ischemic heart disease (IHD: 2.1 percent),
chronic obstructive pulmonary disease (COPD: 3.3 percent), and
pneumonia (4.0 percent). The corresponding 95-percent confidence
intervals for the three specific estimates were, respectively, 1.4% to
2.8%, 1.0% to 5.7%, and 1.8% to 6.2%, per increment of 10 g/
m\3\ in daily PM2.5 exposure. Within the range of dust
concentrations studied, the response appeared to be linear, with no
threshold. The investigators checked for but did not find any
consistent or statistically stable relationship between increased
mortality and the atmospheric concentration of ``course'' respirable

[[Page 5848]]

particles--i.e., those with aerodynamic diameter greater than 2.5
micrometers but less than 10 micrometers.
    As explained earlier, it is highly likely that miners would inhale
at least one-third of their total 24-hour intake of air during a
standard 8-hour work shift. Therefore, under the same assumptions made
in the previous subsection, the 24-hour average concentrations of
PM2.5 measured by Schwartz et al. are no more potent, in
their impact on mortality risk, than eight-hour average concentrations
that are three times as high. As discussed in Section 2.a.iii of this
risk assessment, underground miners may be less, equally, or more
susceptible than the general population to the acute mortality effects
of fine particulates such as dpm. However, miners who smoke tobacco
and/or suffer various respiratory ailments fall into groups identified
as likely to be especially sensitive (EPA, 1996). Consequently, for
such miners occupationally exposed to dpm, the relative risk of each
type of premature mortality would be at least equal to the
corresponding lower 95-percent confidence limit specified above.
    Therefore, MSHA estimates that, on average, each increment of 30
g/m\3\ in the dpm concentration to which miners are exposed
increases the risk of premature death due to IHD, COPD, and pneumonia
by a factor of at least 1.4 percent, 1.0 percent, and 1.8 percent,
respectively. As noted earlier, these estimates are based on the
evidence of acute effects linked to fine particulate exposures
(PM2.5) in ambient air pollution studies. A lower bound on
the increased risk expected at an occupational dpm concentration
greater than 30 g/m\3\, is obtained by raising the relative
risks equivalent to these factors (i.e., 1.014, 1.01, and 1.018) to a
power, k, equal to the ratio of the concentration to 30 g/
m\3\. For a concentration of 300 g/m\3\, k = 10; so MSHA
estimates the lower bounds on relative risk to be: (1.014)\10\ = 1.149
for IHD; (1.01)\10\ = 1.105 for COPD; and (1.018)\10\ = 1.195 for
pneumonia. This means that for miners exposed to dpm at or above this
level, MSHA expects the risks to increase by at least 14.9 percent for
IHD, 10.5 percent for COPD, and 19.5 percent for pneumonia. The
Secretary considers increases of this magnitude to be clearly
significant, since the causes of death to which they apply are not rare
among miners.
(3) Lung Cancer
    In contrast to the two types of risk discussed above, the available
epidemiologic data can be used to relate the risk of lung cancer
directly to dpm exposures. Therefore, the significance of the lung
cancer risk can be evaluated without having to make assumptions about
the relative potency of dpm compared to the remaining constituents of
PM2.5. This removes an important source of uncertainty
present in the other two evaluations.
    There are two different ways in which the significance of the lung
cancer risk may be evaluated. The first way is based on the relative
risk of lung cancer observed in the best available epidemiologic
studies involving miners (identified as such in Subsections 3.a.iii(1)
(b) and (d) of this risk assessment). As will be explained below, this
approach leads to an estimated tripling of lung cancer risk for miners
exposed to dpm, compared to a baseline risk for unexposed miners. The
second way is to calculate the lung cancer risk expected at exposure
levels MSHA has observed in underground mines, assuming a specified
occupational lifetime and using the exposure-response relationships
estimated for underground miners by Johnston et al. (1997) and Saverin
et al. (1999). As will be explained further below, this second approach
yields a wide range of estimates, depending on which exposure-response
relationship and statistical model is used. All of the estimates,
however, show at least a doubling of baseline lung cancer risk,
assuming dpm exposure for a 45-year occupational lifetime at the
average concentration MSHA has observed. Most of the estimates are much
higher than this. If the exposure-response relationship estimated for
workers in the trucking industry by Steenland et al. (1998) is
extrapolated to the much higher exposure levels for miners, the
resulting estimates fall within the range established by the two mine-
specific studies, thereby providing a degree of corroboration. Since
lung cancer is not a rare disease, the Secretary considers even the
very lowest estimate--a doubling of baseline risk--to represent a
clearly significant risk.
    Both of these methods provide quantitative estimates of the degree
by which miners' risk of lung cancer is increased by their occupational
dpm exposures. The estimate based on exposure-response relationships is
more refined, in that it ties the increased risk of lung cancer to
specific levels of cumulative dpm exposure. However, this added
refinement comes at the price of an additional source of uncertainty:
the accuracy of the exposure-response relationship used to calculate
the estimate. This additional uncertainty is reflected, in MSHA's
evaluation, by a broad range of relative risk estimates, corresponding
to the range of exposure-response relationships derived using different
statistical models and epidemiologic data. The next two subsections
present the details of MSHA's two approaches to analyzing lung cancer
risk for miners exposed to dpm, along with MSHA's responses to the
relevant public comments.
(a) Risk Assessment Based on Studies Involving Miners
    As one commenter pointed out, the epidemiologic evidence showing an
elevated risk of lung cancer for exposed workers is mostly based on
occupations estimated to experience far lower exposure levels, on
average, than those observed in many underground mines:

* * *[U]nderground coal, metal and non-metal miners face a
significant risk of lung cancer from occupational exposure to diesel
particulate. Numerous epidemiologic studies of workers exposed to
levels far below those experienced by coal, metal and non-metal
miners have found the risk for exposed workers to be 30-50% greater
than for unexposed workers. [Washington State Dept. of Labor and
Industries]

    Indeed, although MSHA recognizes that results from animal studies
should be extrapolated to humans with caution, it is noteworthy that
dpm exposure levels recorded in some underground mines (see Figures
III-1 and III-2) have been well within the exposure range that produced
tumors in rats (Nauss et al., 1995).
    Both existing meta-analyses of the human studies relating dpm
exposure and lung cancer excluded studies on miners but presented
evidence showing that, averaged across all other occupations, dpm
exposure is responsible for an increase of about 40 percent in lung
cancer risk (See Section 3.a.iii(2) of this risk assessment). Even a
40-percent increase in the risk of lung cancer would clearly be
significant, since this would amount to more than two cases of lung
cancer per year per thousand miners at risk, and to an even greater
risk for smoking miners. The best available evidence, however,
indicates (1) that exposure levels in underground mines generally
exceed exposures for occupations included in the meta-analyses and (2)
that lung cancer risks for exposed miners are elevated to a greater
extent than for other occupations.
    As Dr. Valberg and other commenters pointed out, the epidemiologic
studies used in the meta-analyses involved much lower exposure levels
than those depicted for mines in Figures III-1 and III-2. The studies
supporting a 40-percent excess risk of lung cancer were

[[Page 5849]]

conducted on populations whose average exposure is estimated to be less
than 200 g/m\3\--less than one tenth the average concentration
MSHA observed in some underground mines. More specifically, average
exposure levels in the two most extensively studied industries--
trucking (including loading dock workers) and railroads--have been
reported to be far below the levels observed in underground mining
environments. For workers at docks employing diesel forklifts--the
occupational group estimated to be most highly exposed within the
trucking industry--the highest average dpm concentration reported was
about 55 g/m\3\ EC at an individual dock (NIOSH, 1990). As
explained in Subsection 1.d of this risk assessment, this corresponds
to less than 150 g/m\3\ of dpm, on average. Published dpm
measurements for railworkers have generally also been less than 150
g/m\3\ (measured as respirable particulate matter other than
cigarette smoke). The reported mean of 224 g/m\3\ for hostlers
displayed in Figure III-7 represents only the worst-case occupational
subgroup (Woskie et al., 1988). In contrast, in the study on
underground potash miners by Saverin et al. (1999), the mean TC
concentration measured for production areas was 390 g/m\3\--
corresponding to a mean dpm concentration of about 490 g/m\3\.
As shown in Table III-1, the mean dpm exposure level MSHA observed in
underground production areas and haulageways was 644 g/m\3\
for coal mines and 808 g/m\3\ for M/NM.
    In accordance with the higher exposure levels for underground
miners, the five studies identified in Section III.3.a.iii(1)(d) as
comprising the best available epidemiologic evidence on miners all show
that the risk of lung cancer increased for occupationally exposed
miners by substantially more than 40 percent. The following table
presents the relative risk (RR) of lung cancer for miners in these
studies, along with the geometric mean based on all five studies:

------------------------------------------------------------------------
                                                                Relative
                                                                risk of
                            Study                                 lung
                                                                 cancer
------------------------------------------------------------------------
Boffetta et al., 1988........................................       2.67
Burns & Swanson, 1991........................................       5.03
Johnston et al., 1997 (mine-adjusted model applied at highest       5.50
 cumulative exposure)........................................
Lerchen et al., 1987.........................................       2.1
Saverin et al., 1999 (highest vs least exposed)..............       2.17
geometric mean...............................................       3.2
------------------------------------------------------------------------

    As shown in this table, the estimated RR based on these five
studies is 3.2 for miners exposed to dpm. In other words, the risk of
lung cancer for the highly exposed miners is estimated to be 3.2 times
that of a comparable group of occupationally unexposed workers. The
geometric mean RR remains 3.2 if the two studies on which MSHA places
less weight (by Burns & Swanson and by Lerchen) are excluded from the
calculation. This represents a 220-percent increase in the risk of lung
cancer for exposed miners, in contrast to the 40-percent increase
estimated, on average, for other occupationally exposed workers. The
Secretary believes that a 40-percent increase in the risk of lung
cancer already exceeds, by a wide margin, the threshold for a clearly
significant risk. However, a 220-percent increase to more than three
times the baseline rate is obviously of even greater concern.
    Some commenters questioned whether increased lung cancer risks of
this magnitude were plausible, since they were not aware of any
unusually high lung cancer rates among workers at mines with which they
were familiar and which used diesel equipment. There are several
reasons why an elevated risk of lung cancer might not currently be
conspicuous among U.S. miners exposed to dpm. Lung cancer not only may
require a latency period of 30 or more years to develop, but it may
also not develop until beyond the normal retirement age of 65 years.
Cases of lung cancer developing after retirement may not all be known
to members of the mining community. Also, in a population that includes
many tobacco smokers, it may be difficult to discern cases of lung
cancer specifically attributable to dpm exposure when they first begin
to become prevalent. Two commenters expressed some of the relevant
considerations as follows. Although they were referring to coal miners,
the same points apply to M/NM miners.

    Because the latency period for lung cancer is so long, and
diesel-powered equipment has only been used extensively in U.S. coal
mines for about 25 years, the epidemic may well be progressing
unnoticed. [UMWA]
    If dpm exposure will cause cancer, there is a huge population of
miners here in the West that have already been exposed. Considering
the latency periods indicated by MSHA, these miners should be
beginning to develop cancers. [Canyon Fuels]
(b) Risk Assessment Based on Miners' Cumulative Exposure
    Although it is evident that underground miners currently face a
significant risk of lung cancer due to their occupational exposure to
dpm, there are certain advantages in utilizing an exposure-response
relationship to quantify the degree of risk at specific levels of
cumulative exposure. As some commenters pointed out, for example, dpm
exposure levels may change over time due to changes in diesel fuel and
engine design. The extent and patterns of diesel equipment usage within
mines also has changed significantly during the past 25 years, and this
has affected dpm exposure levels as well. Furthermore, exposure levels
at the mines involved in epidemiologic studies were not necessarily
typical or representative of exposure levels at mines in general. A
quantitative exposure-response relationship provides an estimate of the
risk at any specified level of cumulative exposure. Therefore, using
such a relationship to assess risk under current or anticipated
conditions factors in whatever differences in exposure levels may be
relevant, including those due to historical changes.
(i) Exposure-Response Relationships from Studies Outside Mining
    Stayner et al. (1998) summarized quantitative risk assessments
based on exposure-response relationships for dpm published through
1998. These assessments were broadly divided into those based on human
studies and those based on animal studies. Depending on the particular
studies, assumptions, statistical models, and methods of assessment
used, estimates of the exact degree of risk varied widely even within
each broad category. However, as presented in Tables III and IV of
Stayner et al. (1998), all of the very different approaches and methods
published through 1998 produced results indicating that levels of dpm
exposure measured at some underground mines present an unacceptably
high risk of lung cancer for miners--a risk significantly greater than
the risk they would experience without the dpm exposure.\75\
---------------------------------------------------------------------------

    \75\ In comments submitted by MARG, Dr. Jonathan Borak asserted
that MSHA had ``misrepresented the findings of a critical study'' by
stating that all methods showed an ``unacceptably high risk'' at
exposure levels found at some mines. Dr. Borak claimed that Stayner
et al. (1998) had described an analysis by Crump et al. ``that
reached an opposite conclusion.'' Dr. Borak failed to distinguish
between a finding of high risk and a finding of changes in that risk
corresponding to changes in estimated exposures. The findings to
which Dr. Borak referred pertained only to the exposure-response
relationship within the group of exposed workers. Garshick (1981),
Crump (1999), and HEI (1999) all noted that the risk of lung cancer
was nevertheless elevated among the exposed workers, compared to
unexposed workers in the same cohort, and they all identified
reasons why the data used in this study might fail to detect a
positive exposure-response relationship among the exposed workers.

---------------------------------------------------------------------------

[[Page 5850]]

    Quantitative risk estimates based on the human studies were
generally higher than those based on analyses of the rat inhalation
studies. As indicated by Tables 3 and 4 of Stayner et al. (1998), a
working lifetime of exposure to dpm at 500 g/m\3\ yielded
estimates of excess lung cancer risk ranging from about 1 to 200 excess
cases of lung cancer per thousand workers based on the rat inhalation
studies and from about 50 to 800 per thousand based on the
epidemiologic assessments. Stayner et al. (1998) concluded their report
---------------------------------------------------------------------------
by stating:

    The risk estimates derived from these different models vary by
approximately three orders of magnitude, and there are substantial
uncertainties surrounding each of these approaches. Nonetheless, the
results from applying these methods are consistent in predicting
relatively large risks of lung cancer for miners who have long-term
exposures to high concentrations of DEP [i.e., dpm]. This is not
surprising given the fact that miners may be exposed to DEP [dpm]
concentrations that are similar to those that induced lung cancer in
rats and mice, and substantially higher than the exposure
concentrations in the positive epidemiologic studies of other worker
populations.

    Restricting attention to the exposure-response relationships
derived from human data, Table IV of Stayner et al. (1998) presented
estimates of excess lung cancer risk based on exposure-response
relationships derived from four different studies: Waller (1981) as
analyzed by Harris (1983); Garshick et al. (1987) as analyzed by Smith
and Stayner (1991); Garshick et al. (1988) as analyzed by California
EPA (1998); and Steenland et al. (1998). Harris (1983) represented
upper bounds on risk; and all of the other estimates represented the
most likely value for risk, given the particular data and statistical
modeling assumptions on which the estimate was based. Three different
ranges of estimates were presented from the California EPA analysis,
corresponding to various statistical models and assumptions about
historical changes in dpm exposure among the railroad workers involved.
As mentioned above and in the proposed version of this risk assessment,
the low end of the range of estimates was 50 lung cancers per 1000
workers occupationally exposed at 500 g/m\3\ for a 45-year
working lifetime. This estimate was one of those based on railroad
worker data from Garshick et al. (1988).
    Several commenters objected to MSHA's reliance on any of the
exposure-response relationships derived from the data compiled by
Garshick et al. (1987) or Garshick et al. (1988). These objections were
based on re-analyses of these data by Crump (1999) and HEI (1999),
using different statistical methods and assumptions from those used by
Cal-EPA (1998). For example, the NMA quoted HEI (1999) as concluding:

    At present, the railroad worker cohort study * * * has very
limited utility for QRA [quantitative risk assessment] of lifetime
lung cancer risk from exposure to ambient levels of diesel exhaust *
* * [NMA, quoting HEI (1999)]

    From this, the NMA argued as follows:

    What then is the relevance of this data to the proceedings at
issue? Simply put, there is no relevance. The leading epidemiologist
[sic], including Dr. Garshick himself, now agree that the data are
inappropriate for conducting risk assessment. [NMA]

    MSHA notes that the HEI (1999) conclusion cited by the NMA referred
to quantitative risk assessments at ambient, not occupational, exposure
levels. Also, HEI (1999) did not apply its approach (i.e.,
investigating the correlation between exposure and relative risk within
separate job categories) to the Armitage-Doll model employed by Cal-EPA
in some of its analyses. (Results using this model were among those
summarized in Table IV of Stayner et al., 1998). Therefore, the
statistical findings on which HEI (1999) based its conclusion do not
apply to exposure-response relationships estimated using the Armitage-
Doll model. Furthermore, although HEI concluded that the railroad
worker data have ``very limited utility for QRA * * * at ambient
levels'' [emphasis added], this does not mean, even if true, that these
data have ``no relevance'' to this risk assessment, as the NMA
asserted. Even if they do not reliably establish an exposure-response
relationship suitable for use in a quantitative risk assessment, these
data still show that the risk of lung cancer was significantly elevated
among exposed workers. This is the only way in which MSHA is now using
these data in this risk assessment.
    In the proposed risk assessment, MSHA did not rely directly on the
railroad worker data but did refer to the lowest published quantitative
estimate of risk, which happened, as of 1998, to be based on those
data. MSHA's reasoning was that, even based on the lowest published
estimate, the excess risk of lung cancer attributable to dpm exposure
was clearly sufficient to warrant regulation. If risk assessments
derived from the railroad worker data are eliminated from
consideration, the lowest estimate remaining in Table IV of Stayner et
al. (1998) is obviously even higher than the one that MSHA used to make
this determination in the proposed risk assessment. This estimate
(based on one of the analyses performed by Steenland et al., 1998) is
89 excess cases of lung cancer per year per thousand workers exposed at
500 g/m\3\ for a 45-year working lifetime.
    HEI (1999) also evaluated the use of the Steenland data for
quantitative risk assessment, but did not perform any independent
statistical analysis of the data compiled in that study. Some
commenters pointed out HEI's reiteration of the cautionary remark by
Steenland et al. (1998) that their exposure assessment depended on
``broad assumptions.'' The HEI report did not rule out the use of these
data for quantitative risk assessment but suggested that additional
statistical analyses and evaluations were desirable, along with further
development of exposure estimates using alternative assumptions. MSHA
has addressed comments on various aspects of the analysis by Steenland
et al., including the exposure assumptions, in Section 2.c.i(2)(a) of
this risk assessment.
    One commenter noted that Steenland et al. (1998) had recognized the
limitations of their analysis and had, therefore, advised that the
results ``should be viewed as exploratory.'' The commenter then
asserted that MSHA had nevertheless used these results as ``the basis
for a major regulatory standard'' and that ``[t]his alone is sufficient
to demonstrate that MSHA's proposal lacks the necessary scientific
support.'' [Kennecott Minerals]
    The Secretary does not accept the premise that MSHA should exclude
``exploratory'' results from its risk assessment, even if it is granted
that those results depend on broad assumptions possibly requiring
further research and validation before they are widely accepted by the
scientific community. Steenland et al. (1998) estimated risks
associated with specific cumulative exposures, based on estimates of
historical exposure patterns combined with data originally described by
Steenland et al., 1990 and 1992. Regardless of whether the cumulative
exposure estimates used by Steenland et al. (1998) are sufficiently
reliable to permit pinpointing the risk of lung cancer at any given
exposure level, the quantitative analysis indicates that as cumulative
exposure increases, so does the risk. Therefore, the 1998 analysis adds
significantly to the weight of evidence supporting a causal

[[Page 5851]]

relationship. However, MSHA did not use or propose to use exposure-
response estimates derived by Steenland et al. (1998) as the sole basis
for any regulatory standard.
    The exposure-response relationships presented by Steenland et al.
were derived from exposures estimated to be far below those found in
underground mines. As Stayner et al. (1998) point out, questions are
introduced by extrapolating an exposure-response relationship beyond
the exposures used to determine the relationship. The uncertainties
implicit in such extrapolation are demonstrated by comparing results
from two statistical models based on five-year lagged exposures--one
using simple cumulative exposure and the other using the natural
logarithm of cumulative exposure (Steenland et al., 1998, Table II).
    Assuming that, on average, EC comprises 40 percent of total
dpm,\76\ the formula for calculating a relative risk (RR) using
Steenland's simple cumulative exposure model is RR =
exp(0.4 x 0.389 x CumExp), where CumExp is occupationally accumulated
dpm exposure (expressed in mg-yr/m\3\), ignoring the most recent five
years. Again assuming EC=0.4 x dpm, the corresponding formula using
Steenland's Log(CumExp) model is: RR =
exp(0.1803 x (Log(0.4 x 1000 x CumExp + BG)-Log(BG))), still ignoring
occupational dpm exposure in the most recent five years.\77\
---------------------------------------------------------------------------

    \76\ The assumption is that, on average, EC = TC/2 and TC =
0.8 x dpm.
    \77\ BG, expressed in g-yr/m\3\, accounts for an
assumed background (i.e., non-occupational) EC exposure level of 1.0
g/m\3\. At age 70, after a 45-year worklife and an
additional 5-year lag after retirement, BG is assumed to equal 70
g-yr/m\3\. ``Log'' refers to the natural logarithm, and
``exp'' refers to the antilogarithm of the subsequent quantity.
---------------------------------------------------------------------------

    The risk estimates from these two models are similar at the
cumulative exposure levels estimated for workers involved in the study,
but the projected risks diverge markedly at the higher exposures
projected for underground miners exposed to dpm for a 45-year
occupational lifetime. For example, a cumulative dpm exposure of 2.5
mg-yr/m\3\ (i.e., 45 years of occupational exposure at an average dpm
concentration of about 55.6 g/m\3\) is within the range of
cumulative exposures from which these exposure-response relationships
were estimated. At this level of cumulative exposure, the models (both
lagged five years) yield relative risk estimates of 1.48 (based on
simple cumulative exposure) and 1.64 (based on the logarithm of
cumulative exposure, with BG=70 g-yr/m\3\). On the other hand,
45 years of occupational exposure at an average dpm concentration of
808 g/m\3\ amounts to a cumulative dpm exposure of 36,360
g-yr/m\3\, or about 36.4 mg-yr/m\3\. At this level, which lies
well beyond the range of data used by Steenland et al. (1998), the
simple and logarithmic exposure models produce relative risk estimates
of about 300 and 2.6, respectively.
    Despite the divergence of these two models at high levels of
cumulative exposure, they can provide a useful check of excess lung
cancer risks estimated using exposure-response relationships developed
from other studies. For highly exposed miners, the Steenland models
both produce estimates of lung cancer risk within the range established
by the two miner studies discussed below. This corroborates the upper
and lower limits on such risk as estimated by the various statistical
models used in those two studies.
(ii) Exposure-Response Relationships from Studies on Miners
    As described in Section 2.c.i(2)(a) of this risk assessment, two
epidemiologic studies, both conducted on underground miners, provide
exposure-response relationships based on fully quantitative dpm
exposure assessments. Johnston et al. (1997) conducted their study on a
cohort of 18,166 underground coal miners, and Saverin et al. (1999)
conducted theirs on a cohort of 5,536 underground potash miners. Each
of these studies developed a number of possible exposure-response
relationships, depending on the statistical model used for analysis
and, in the case of Saverin et al. (1999), inclusion criteria for the
cohort analyzed. For purposes of this risk assessment, MSHA has
converted the units of cumulative exposure in all of these exposure-
response relationships to mg-yr/m\3\.
    Two exposure-response relationships derived by Johnston et al.
(1997) are used in this risk assessment, based on a ``mine-adjusted''
and a ``mine-unadjusted'' statistical model. In both of these models,
cumulative dpm exposure is lagged by 15 years.\78\ This reflects the
long latency period required for development of lung cancer and means
that the most recent 15 years of exposure are ignored when the relative
risk of lung cancer is estimated. The exposure-response relationships,
as reported by the investigators, were expressed in terms of g-hr/m\3\
of cumulative dpm exposure. MSHA has converted the exposure units to
mg-yr/m\3\ by assuming 1920 work hours per year.
---------------------------------------------------------------------------

    \78\ The 15-year lagged mine-unadjusted and mine-adjusted models
are respectively denoted by M/03 and M/06 in Table 11.2 of Johnston
et al. (1997). As explained earlier, the individual mines considered
in this study differed significantly with respect to both dpm
exposures and lung cancer experience. The investigators could not
determine exactly how much, if any, of the increased lung cancer
risk associated with dpm exposure depends on other, unknown factors
differentiating the individual mines. The mine-adjusted model
allocates a significant number of the lung cancers otherwise
attributable to dpm exposure to the ``norm'' for specific mines.
Therefore, if the differences in lung cancer prevalence between
mines is actually due to corresponding differences in mean dpm
exposure, then this model will mask a significant portion of the
risk due to dpm exposure. After adjusting for miners' age and
smoking habits, the mine-unadjusted model attributes differences in
the prevalence of lung cancer between mines to corresponding
differences in mean dpm exposure. However, the mine-adjusted model
has the advantage of taking into account differences between mines
with respect to potentially confounding factors, such as radon
progeny and silica levels.
---------------------------------------------------------------------------

    Two different methods of statistical analysis were applied by
Sauverin et al. (1999) to both the full cohort and to a subcohort of
3,258 miners who had worked underground, in relatively stable jobs, for
at least ten years. Thus, the investigators developed a total of four
possible exposure-response relationships from this study. Since they
were based on measurements of total carbon (TC), these exposure-
response relationships were expressed in terms mg-yr/m\3\ of cumulative
TC exposure. MSHA has converted the exposure units to mg-yr/m\3\ of
cumulative dpm exposure by assuming that, on average, TC comprises 80
percent of total dpm.
    The following table summarizes the exposure-response relationships
obtained from these two studies. Each of the quantitative relationships
is specified by the unit relative risk (RR) per mg-yr/m\3\ of
cumulative dpm exposure. To calculate the relative risk estimated for a
given cumulative dpm exposure (CE), it is necessary to raise the unit
RR to a power equal to CE. For example, if the unit RR is 1.11 and CE =
20, then the estimated relative risk is (1.11)\20\ = 8.1. Therefore,
the estimated relative risk of lung cancer increases as CE increases.
For the two Johnston models, CE does not include exposure accumulated
during the 15 years immediately prior to the time in a miner's life at
which the relative risk is calculated.

[[Page 5852]]

Exposure-Response Relationships Obtained From Two Studies on Underground
                                 Miners.
------------------------------------------------------------------------
                                                             Unit RR per
                Study and statistical model                   mg-yr/m\3\
                                                                 dpm
------------------------------------------------------------------------
Saverin et al. (1999)\1\:
  Poisson, full cohort.....................................        1.024
  Cox, full cohort.........................................        1.089
  Poisson, subcohort.......................................        1.110
  Cox, subcohort...........................................        1.176
Johnston et al. (1997)\2\ :
  15-year lag, mine-adjusted...............................        1.321
  15-year lag, mine-unadjusted.............................       1.479
------------------------------------------------------------------------
\1\ Unit RR calculated from Tables III and IV, assuming TC = 0.8 x dpm.
\2\ Unit RR calculated from Table 11.2, assuming 1920 work hours per
  year.

    For example, suppose a miner is occupationally exposed to dpm at an
average level of 500 g/m3. Then each year of
occupational exposure would contribute 0.5 mg-yr/m3 to the
miner's cumulative dpm exposure. Suppose also that this miner's
occupational exposure begins at age 45 and continues for 20 years until
retirement at age 65. Consequently, at or above age 65, this
hypothetical miner would have accumulated a total of 10 mg-yr/
m3 of occupational dpm exposure. According to the Saverin-
Cox-subcohort model, the relative risk estimated for this miner after
retirement is RR = (1.176)10 = 5.1. This means that, at or
above age 65, the retired miner's risk of lung cancer is estimated (by
this model) to be about five times that of another retired miner having
the same age and smoking history but no occupational dpm exposure.
    Since the two Johnston models exclude exposure within the last 15
years, it is instructive to calculate the relative risk using these
models for the same hypothetical retiree at age 75. Since this miner
retired at age 65, immediately after 20 years of occupational exposure,
the cumulative exposure used in applying the Johnston models must be
reduced by the 2.5 mg-yr/m3 accumulated from age 60 to age
65. Therefore, according to the Johnston mine-adjusted model, the
relative risk estimated for this retired miner at age 75 is RR =
(1.321)7.5 = 8.1. At age 80 or above, however, this model
predicts that the relative risk would increase to RR =
(1.321)10 = 16.2.
    The six exposure-response relationships obtained from these two
studies establish a range of quantitative risk estimates corresponding
to a given level of cumulative dpm exposure. This range provides lower
and upper limits on the risk of lung cancer for workers exposed at the
given level, relative to similar workers who were not occupationally
exposed. The lower limit of this range is established by Saverin's full
cohort Poisson model. Therefore, the lowest estimate of relative risk
after 45 years of occupational dpm exposure is RR =
(1.024)45 x 0.644 = 2.0 at a mean concentration of 644
g/m3 or RR = (1.024)45 x 0.808 = 2.4 at
mean concentration of 808 g/m3. These exposure
levels correspond to the averages presented in Table III-1 for
underground coal and underground M/NM mines, respectively.
    A relative risk of 2.0 amounts to a doubling of the baseline lung
cancer risk, and all of the models project relative risks of at least
2.0 after 45 years of exposure at these levels. Therefore, MSHA expects
that underground miners exposed to dpm at these levels for a full 45-
year occupational lifetime would, at a minimum, experience lung cancer
at a rate twice that of unexposed but otherwise similar miners. Five of
the six statistical models, however, predict a relative risk much
greater than 2.0 after 45 years at a mean dpm concentration of 644
g/m3. The second-lowest estimate of relative risk,
for example, is RR = (1.089)45 x 0.644 = 11.8, predicted by
Saverin's full cohort Cox model.\79\
---------------------------------------------------------------------------

    \79\ Some commenters contended that MSHA cannot establish a
reliable exposure-response relationship because of potential
interferences in MSHA's dpm concentration measurements. More
specifically, some of these commenters claimed that MSHA's dpm
measurements in underground coal mines were significantly inflated
by submicrometer coal dust.
    As explained in Subsection 1.a of this risk assessment, the
sampling device MSHA used to measure dpm in underground coal mines
was designed specifically to allow for the submicrometer fraction of
coal dust. Both the size-selective and RCD methods are reasonably
accurate when dpm concentrations exceed 300 g/
m3. Moreover, neither of these methods was used to
establish the exposure-response relationships presented by Saverin
et al. (1999) or Johnston et al. (1997).
---------------------------------------------------------------------------

    In the next subsection of this risk assessment, relative risks will
be combined with baseline lung cancer and mortality data to estimate
the lifetime probability of dying from lung cancer due to occupational
dpm exposure.
    (iii) Excess Risk at Specific dpm Exposure Levels. The ``excess
risk'' discussed in this subsection refers to the lifetime probability
of dying from lung cancer resulting from occupational exposure to dpm
for 45 years. This probability is expressed as the expected excess
number of lung cancer deaths per thousand miners occupationally exposed
to dpm at a specified level. The excess is calculated relative to
baseline, age-specific lung cancer mortality rates taken from standard
mortality tables. In order to properly estimate this excess, it is
necessary to calculate, at each year of life after occupational
exposure begins, the expected number of persons surviving to that age
with and without dpm exposure at the specified level. At each age,
standard actuarial adjustments must be made in the number of survivors
to account for the risk of dying from causes other than lung cancer.
    Table III-7 shows the excess risk of death from lung cancer
estimated across the range of exposure-response relationships obtained
from Saverin et al. (1999) and Johnston et al. (1997). Estimates based
on the 5-year lagged models from Steenland et al. (1998) fall within
this range and are included for comparison. Based on each of the eight
statistical models, the excess risk was estimated at four levels of dpm
exposure: 200 g/m3, 500 g/m3,
644 g/m3 (the mean dpm concentration observed by
MSHA at underground coal mines, as shown in Table III-1), and 808
g/m3 (the mean dpm concentration observed by MSHA
at underground M/NM mines, as shown in Table III-1).

BILLING CODE 4510-43-P

[[Page 5853]]

[GRAPHIC] [TIFF OMITTED] TR19JA01.074

BILLING CODE 4510-43-C

[[Page 5854]]

    All of the estimates in Table III-7 assume that occupational
exposure begins at age 20 and continues until retirement at age 65.
Excess risks were calculated through age 85 as in Table IV of Stayner
et al. (1998). Table III-7 differs from Table IV of Stayner et al. in
that results from Johnston et al. and Saverin et al. are substituted
for results based on the two studies by Garshick et al. Nevertheless,
at 500 g/m3, the range of excess risks shown in
Table III-7 is nearly identical to the range (50 to 810 g/
m3) presented in Table IV of Stayner et al. (1998).
    MSHA considers the exposure levels shown in Table III-1 to be
typical of current conditions in underground coal mines using diesel
face equipment. At the mean dpm concentration observed by MSHA at
underground M/NM mines (808 g/m3), the eight
estimates range from 83 to 830 excess lung cancer deaths per 1000
affected miners. At the mean dpm concentration observed by MSHA at
underground coal mines (644 g/m3), the estimates
range from 61 to 811 excess lung cancer deaths per 1000 affected
miners. MSHA recognizes that these risk estimates involved
extrapolation beyond the exposure experience of the miner cohorts in
Saverin et al. (1999) and Johnston et al. (1997). However, the degree
of extrapolation was less for those two studies than the extrapolation
that was necessary for the diesel-exposed truck drivers in Steenland et
al. The lowest excess lung cancer risk in dpm exposed miners found in
Table III-7 is 61/1000 per 45-year working lifetime. Based on the
quantitative rule of thumb established in the benzene case, this
estimate indicates a clearly significant risk of lung cancer
attributable to dpm exposure at current levels. [Industrial Union vs.
American Petroleum; 448 U.S. 607, 100 S.Ct. 2844 (1980)].
c. The Rule's Expected Impact on Risk
    MSHA strongly disagrees with the views of some commenters who
asserted that the proposed rules would provide no known or quantifiable
health benefit to mine workers. On the contrary, MSHA's assessment of
the best available evidence indicates that reducing the very high
exposures currently existing in underground mines will significantly
reduce the risk of three different kinds of material impairment to
miners: (1) Acute sensory irritations and respiratory symptoms
(including allergenic responses); (2) premature death from
cardiovascular, cardiopulmonary, or respiratory causes; and (3) lung
cancer. Furthermore, as will be shown below, the reduction in lung
cancer risk expected as a result of the rule can readily be quantified
based on the estimates of excess risk at exposure levels given in Table
III-7.
    Using exposure-response relationships and assumptions described in
Subsections 3.b.ii(1) and 3.b.ii(2) of this risk assessment, MSHA
estimated lower bounds on the significance of risks faced by miners
occupationally exposed to dpm with respect to (1) acute sensory
irritations and respiratory symptoms or (2) premature death from
cardiovascular, cardiopulmonary, or respiratory causes. MSHA expects
the rules to significantly and substantially reduce all three kinds of
risk. However, MSHA is unable, based on currently available data, to
quantify with confidence the reductions expected for the first two
kinds. A 24-hour exposure at 20 g/m\3\ may not have the same
short-term effects as an 8-hour exposure at 60 g/m\3\.
Furthermore, this concentration is only 30 percent of the maximum dpm
concentration that MSHA expects once the rules are fully implemented
and represents an even smaller fraction of average dpm concentrations
many underground miners currently experience. It is unclear whether the
same incremental effects on acute respiratory symptoms and premature
mortality would apply at the much higher exposure levels found in
underground mines. Additionally, as MSHA suggested in the proposed
preamble and several commenters repeated, the toxicity of dpm and
PM2.5 may differ because of differences in composition.
Finally, underground miners as a group may differ significantly from
the populations for which the PM2.5 exposure-response
relationships were derived.
    Therefore, MSHA's quantitative assessment of the rule's impact on
risk is restricted to its expected impact on the third kind of risk--
the risk of lung cancer. The rule will limit dpm concentrations to
which miners in underground M/NM mines are exposed. The rule will limit
these dpm concentrations to approximately 200 g/m\3\ by
limiting the measured concentration of total carbon to 160 g/
m\3\. Assuming that, in the absence of this rule, underground M/NM
miners would be occupationally exposed to dpm for 45 years at a mean
level of 808 g/m\3\, the following table contains the
estimated reductions in lifetime risk expected to result from full
implementation of the rule, based on the various exposure-response
relationships obtained from Saverin et al. (1999) and Johnston et al.
(1997). These estimates were obtained by calculating the difference
between the corresponding estimates of excess lung cancer mortality, at
808 g/m\3\ and 200 g/m\3\, shown in Table III-7. The
Regulatory Impact Analysis (RIA), presented later in this preamble,
contains further quantitative discussion of the benefits anticipated
from this rule.

 Reduction in Lifetime Risk of Lung Cancer Mortality Expected as Result
 of Reducing Exposure Level From 808 g/m\3\ to 200 g/
                                  m\3\.
------------------------------------------------------------------------
                                                             Expected
                                                           reduction in
                                                            lung cancer
               Study and statistical model                  deaths per
                                                           1000 affected
                                                             miners\1\
------------------------------------------------------------------------
Saverin et al. (1999):
Poisson, full cohort....................................              68
Cox, full cohort........................................             507
Poisson, subcohort......................................             600
Cox, subcohort..........................................             620
Johnston et al. (1997):
15-year lag, mine-adjusted..............................             487
15-year lag, mine-unadjusted............................            317
------------------------------------------------------------------------
\1\ Calculated from Table III-7.

    Although the Agency expects that health risks will be substantially
reduced by this rule, the best available evidence indicates that a
significant risk of adverse health effects due to dpm exposures will
remain even after the rule is fully implemented. As explained in Part V
of this preamble, however, MSHA has concluded that, due to monetary
costs and technological limitations, the underground M/NM mining sector
as a whole cannot feasibly reduce dpm concentrations further at this
time.

4. Conclusions

    MSHA has carefully considered all of the evidence and public
comment submitted during these proceedings to determine whether dpm
exposures, at levels observed in some mines, present miners with
significant health risks. This information was evaluated in light of
the legal requirements governing regulatory action under the Mine Act.
Particular attention was paid to issues and questions raised by the
mining community in response to the Agency's ANPRM and NPRM and during
workshops on dpm held in 1995. Based on its review of the record as a
whole, the agency has determined that the best available evidence
warrants the following conclusions:
    1. Exposure to dpm can materially impair miner health or functional
capacity. These material impairments include acute sensory irritations
and respiratory symptoms (including allergenic responses); premature
death

[[Continued on page 5856]]
Notices For
2009
2008
2007
2006
2005
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
[[pp. 5806-5855]] Diesel Particulate Matter Exposure of Underground Metal and

Local Navigation
