EPA REGION VII IRC
                              101612



      United States       Science Advisory      EPA-SAB-CASAC-01-003
      Environmental       Board (1400A)         December 2000
      Protection Agency     Washington DC         www.epa.gov/sab




6EPA REVIEW OF EPA's HEALTH



      ASSESSMENT DOCUMENT


      FOR DIESEL EXHAUST


      (EPA 600/8-90/057E)
      REVIEW BY THE CLEAN AIR

      SCIENTIFIC ADVISORY

      COMMITTEE (CASAC)

-------
                                    December 19, 2000

EPA-S AB-CAS AC-01 -003

Honorable Carol M. Browner
Administrator
U.S. Environmental Protection Agency
1200 Pennsylvania Avenue, NW
Washington, DC 20460

              Subject:      Clean Air Scientific Advisory Committee (CASAC) Review of EPA's
                           Health Assessment Document for Diesel Exhaust (EPA 600/8-
                           90/057E)

Dear Ms. Browner:

       The Clean Air Scientific Advisory Committee (CASAC, also referred to as the Committee) of
the EPA Science Advisory Board, supplemented by expert consultants (together referred to as the
Panel), met on October 12-13, 2000 to review the July 2000 draft document, Health Assessment
Document for Diesel Exhaust (EPA 600/8-90/057E), in a public meeting in Alexandria VA. This
draft document was prepared by EPA's National Center for Environmental Assessment (NCEA),
Washington, DC

       An SAB Subcommittee conducted an initial review of the diesel topic in 1990.  Subsequently,
CASAC reviewed drafts of the diesel health assessment document in 1995 and  1998, finding in both
cases that the document was not yet scientifically adequate for making regulatory decisions. A
consultation between the Panel and NCEA Staff (hereafter referred to as Staff)  was held on June 10,
1999 regarding the development of the next draft. On December 1,  1999, CASAC reviewed the draft
document and found it improved, but not sufficient to warrant closure.

       During the October 2000 meeting, numerous suggestions were offered for additional revisions
to improve the document's accurate and complete portrayal of current knowledge.  Several key issues
were discussed, and agreement between the Committee and Staff was reached on approaches to be
taken to making changes addressing all key issues.  Two issues engendered extended discussion.

       It was agreed that two approaches would be taken to characterizing the level of long-term
environmental exposure considered acceptably free from significant non-cancer health risk. A
reference concentration (RfC) would be derived as before, but would include an interspecies
uncertainty factor resulting in a value of approximately 5 ug/m3. It was agreed that linkages between
risks from diesel particulate matter (DPM) and ambient particulate matter (PM) would also be

-------
discussed, concluding that an annual national ambient air quality standard (NAAQS) for PM2.5 would
be considered adequately protective for long-term exposures to ambient DPM.

       The inclusion of a range of cancer risk values provides a perspective on the possible range of
lung cancer risk from environmental exposures was strongly debated.  There were concerns for the
conflict between inclusion of a range and the decision not to adopt a unit risk value for cancer, and for
the likely misuse of the values despite Agency  disclaimers.  It was agreed that the range would be
included, but accompanied by clear caveats and disclaimers concerning the uncertainty of the risk
values, the use of the values, and the fact that the possible lower end of the risk range includes zero.

       With mixed recommendations from its consultants, the Committee reached unanimous closure
on the document on October 13th, based on assurances by Agency staff that key revisions would be
made as agreed and attention would also be given to the numerous more minor issues raised by the
Panel.

       CASAC compliments EPA Staff for their strong effort to respond to the Panel's
recommendations in developing the revised draft.  With further, final revisions, the document will
constitute an acceptably accurate and complete summary of current knowledge concerning the health
effects of diesel emissions.

       We look forward to your response to the  advice we have given in this report and seeing the
final version of the Diesel Health Assessment so that it can provide scientific input to the future decisions
of the Agency.

                                    Sincerely,
                                    /s/
                                    Dr. Phillip K. Hopke, Chair
                                    Clean Air Scientific Advisory Committee

-------
                                         NOTICE
       This report has been written as part of the activities of the Science Advisory Board, a public
advisory group providing extramural scientific information and advice to the Administrator and other
officials of the Environmental Protection Agency. The Board is structured to provide balanced, expert
assessment of scientific matters related to problems facing the Agency.  This report has not been
reviewed for approval by the Agency and, hence, the contents of this report do not necessarily
represent the views and policies of the Environmental Protection Agency, nor of other agencies in the
Executive Branch of the Federal government, nor does mention of trade names or commercial products
constitute a recommendation for use.
Distribution and Availability: This Science Advisory Board report is provided to the EPA
Administrator, senior Agency management, appropriate program staff, interested members of the

-------
public, and is posted on the SAB website (www.epa.gov/sab). Information on its availability is also
provided in the SAB's monthly newsletter (Happenings at the Science Advisory Board). Additional
copies and further information are available from the SAB Staff.
                       U.S. Environmental Protection Agency
                               Science Advisory Board
                Clean Air Scientific Advisory Committee (CASAC)
                                 Diesel Review Panel
Panel Chair
Dr. Joe Mauderly1, Vice President, Senior Scientist, and Director of National Environmental
       Respiratory Center, Lovelace Respiratory Research Institute, Albuquerque, NM

CASAC Members2
Mr. John Elston, Administrator, Office of Air Quality Management, State of New Jersey, Department
       of Environmental Protection and Energy, Trenton, NJ

Dr. Philip K. Hopke3, R.A. Plane Professor of Chemistry, Clarkson University, Department of
       Chemistry, Potsdam, NY (CASAC Chair)

Dr. Eva J. Pell4, Steimer Professor of Agriculture Sciences, The Pennsylvania State University,
       University Park, PA

Dr. Arthur C. Upton, M.D., Director, Independent Peer Review, UMDNJ-CRESP, Environmental
       and Occupational Health Sciences Institute, New Brunswick, NJ

Dr. Sverre Vedal, M.D., University of British Columbia, Vancouver Hospital, Vancouver, BC,
       Canada

Dr. Warren White5, Senior Research Associate, Washington University, Chemistry Department, St.
       Louis, MO

CASAC Consultants6
Dr. David Diaz-Sanche^ Department of Medicine, UCLA, Los Angeles, CA
                                                                  *>
Dr. Eric Garshick, M.D., Staff Physician, Pulmonary and Critical Care Sectf&n, West Roxbury VA
       Medical Center, West Roxbury, MA

Dr. Roger O. McClellan, Advisor, Toxicology and Human Health Risk Analysis, and President
       Emeritus, Chemical Industry Institute of Toxicology (CIIT), Albuquerque, NM

-------
Dr. Gunter Oberdorstei; University of Rochester Medical Center, Department of Environmental
       Medicine, Rochester, NY

Dr. Leslie Stayner7, National Institute for Occupational Safety and Health (N1OSH), Risk Evaluation
       Branch, Taft Laboratories, Cincinnati, OH

Dr. Ron Wyzga, Electric Power Research Institute (EPRI), Palo Alto, CA

Science Advisory Board Staff
Mr. Robert Flaak, Designated Federal Official (DFO) and Team Leader, Committee Operations
       Staff, EPA Science Advisory Board (1400A), 1200 Pennsylvania Avenue, NW, US
       Environmental Protection Agency, Washington, DC 20460

Ms. Diana Pozun, Management Assistant, Committee Operations Staff, and Program Specialist,
       Office of the Staff Director, EPA Science Advisory Board (1400A), 1200 Pennsylvania
       Avenue, NW, US Environmental Protection Agency, Washington, DC 20460
1  Appointment as Chair of CASAC ended on October 30, 2000. Appointed ex officio Past Chair until September
        30,2001.
  CASAC Members are experts appointed by the Administrator to two-year terms to serve on the Clean Air
        Scientific Advisory Committee.
3  Appointed as Chair of CASAC on October 30, 2000.
4  Resigned from CASAC on September 28,2000.
5  Appointment as Member of CASAC ended on October 30,2000.
  CASAC Consultants are experts appointed by the Science Advisory Board Staff Director to a one-year term to
        serve on ad hoc Panels formed to address a particular issue; in this case, the CASAC Review of EPA's
        Health Assessment Document for Diesel Exhaust.
  Federal Expert.
                                               Ill

-------
                             TABLE OF CONTENTS


 1. EXECUTIVE SUMMARY                                                         1

 2. INTRODUCTION                                                               3
       2.1    Introduction                                                           3
       2.2    Charge                                                               4

 3. COMMENTS BY CASAC ON EACH CHAPTER                                     5
       3.1    Chapter 1: Executive Summary                                            5
       3.2    Chapter 2: Diesel Emissions, Characterization, Atmospheric Transformation,
                         and Exposures                                            5
       3.3    ChapterS: Dosimetry of Diesel Exhaust Particles in the Respiratory Tract          6
       3.4    Chapter 4: Mutagenicity of Diesel Exhaust                                   6
       3.5    ChapterS: Noncancer Health Effects of Diesel Exhaust                         7
       3.6    Chapter 6: Quantitative Approaches to Estimating Human Noncancer Health
                          Risks of Diesel Exhaust                                      7
       3.7    Chapter 7: Carcinogenicity of Diesel Exhaust                                 9
       3.8    Chapter 8: Dose-Response Assessment: Carcinogenic Effects                    9
       3.9    Chapter 9: Characterization of Potential Human Health Effects of Diesel
                          Exhaust: Hazard and Dose-Response Assessments              10

4. CONCLUSIONS                                                                12

REFERENCES CITED                                                             R-l

APPENDIX A - INDIVIDUAL PANELIST WRITTEN COMMENTS                      A-1
                                        IV

-------
                             1.  EXECUTIVE SUMMARY
       The Clean Air Scientific Advisory Committee (CAS AC), of the EPA Science Advisory Board,
supplemented by expert consultants met on October 12-13, 2000 to review the July 2000 draft
document, Health Assessment Document for Diesel Exhaust (EPA 600/8-90/057E), in a pubic
meeting in Alexandria, VA.  This review followed reviews of previous drafts in 1995, 1998, and 1999.
The Committee found the July 2000 draft, pending key revisions agreed upon at the meeting
(summarized below) and numerous minor editorial changes, to be an adequate summary of current
knowledge concerning the health effects of diesel engine emissions.

       The draft was improved over the last draft, and the Committee complimented Staff for its strong
effort to take the Panel's previous comments and recommendations into consideration. The Panel
approved of the general framework of (he document.  Two of the key issues raised at the last review
were satisfactorily addressed in accordance with the Committee's guidance. First, the revision
eliminated the use of a different health effect, immunological responses, as a basis for adjusting the
reference concentration (RfC) for non-cancer effects based on lung pathology.  Second, the revision
changed the descriptive characterization of cancer hazard from long term environmental exposures from
"highly likely" to "likely".

       Numerous suggestions were offered for  additional revisions throughout the document to
improve its accurate and  complete portrayal of current knowledge. Agreement between the Committee
and Staff was reached on approaches to making changes addressing all key issues.

       It was agreed that two approaches would be taken to characterizing the level of long-term
environmental exposure considered acceptably free from significant non-cancer health risk. An RiC
would be derived as before, but including a factor for uncertainty in interspecies extrapolation and
resulting in a value of approximately 5 ng/m3.  It  was agreed that linkages between risks from diesel
particulate matter (DPM) and ambient PM would also be discussed, concluding that an annual
NAAQS for PM2.5 would be considered adequately protective for ambient DPM.

       The inclusion of a range of cancer risk values to provide a perspective on the possible range of
lung cancer risk from environmental exposures was debated.  There were concerns that inclusion of the
range could be perceived as inconsistent with the decision not to adopt a unit risk value for cancer, and
for the likely misuse of the values despite Agency disclaimers.  It was agreed that the range would be
included, but accompanied by clear caveats and  disclaimers concerning the uncertainty of risk, the use
of the risk perspective values, and the fact that the possible lower end of the risk range includes zero.

-------
         With mixed recommendations from its consultants1, the Committee reached unanimous closure
on the document, pending assurances that the above key revisions would be made and attention would
be given to incorporating numerous more minor changes suggested in the individual Panelist's
comments.
  The closure options presented to the Panel by the Panel Chair were to: a) close on the draft document without needing to see it
again; or b) not close on the draft document if they desired to review additional rewrites (this latter option would require an additional
meeting). Three of the six consultants indicated that they wished to see how EPA characterized the results of the Panel discussions in
certain sections of the draft document, as a result, these three consultants were not able to fully close on the draft document. All six
of the CASAC Members present (the seventh Member had resigned prior to the review) choose to close on the draft document without
formally seeing it again, assuming that EPA would make a good faith effort to incorporate appropriate updates as discussed during the
meeting on October 12-13, 2000.

-------
                                2.  INTRODUCTION
2.1    Introduction

       The Clean Air Scientific Advisory Committee (CASAC) Diesel Review Panel (Members plus
expert Consultants) reviewed the Agency's revised draft Health Assessment Document for Diesel
Engine Exhaust (EPA, 2000) on October 12-13,2000 in Alexandria, VA. The draft review document
was prepared by the Agency's National Center for Environmental Assessment (NCEA) - Washington,
DC Office.

       The diesel topic was first reviewed by a Subcommittee of the Science Advisory Board in 1990.
This initial effort was followed in 1995 when the first CASAC Diesel Review Panel (Members plus
expert Consultants) conducted a peer review of the December 1994 version of the draft diesel
assessment. As a result of that review, the CASAC recommendations focused on: a) the use of
specific uncertainty factors in deriving the RfC (reference concentration) value for protecting from
adverse noncancer respiratory effects; b) the minimal scientific support for using rat bioassay data for
estimating human cancer risks; and c) the outdated nature of information in several chapters. The
Committee also made numerous suggestions and recommendations for improving the draft document,
asking to review the revised document when it was ready.  These recommendations are covered in
detail in the CASAC report of that review (CASAC, 1995).

       The resultant revised draft diesel assessment was then reviewed by a new CASAC Diesel
Review Panel at a meeting on May 5-6, 1998. At that meeting, NCEA provided CASAC with a  listing
that identified the  disposition of the significant recommendations that had been made by CASAC in
1995 (CASAC, 1995).  The CASAC Diesel Review Panel that was created for this 1998 review
included a number of Members and Consultants who served on the 1995 Panel as well as new
panelists to ensure that the composition of the review panel would be fresh and objective. This is  the
standard practice of the SAB and is consistent with the provisions of the Agency's 1994 Peer Review
Policy and the 1998 Peer Review  Handbook (EPA, 1998). Panelists were asked to provided written
comments on the  questions in the charge as well as specific chapters that they had been assigned for
review.  These recommendations are covered in detail in the  CASAC report of that review (CASAC,
1998).  The Committee could not  close on the draft and asked that the Agency make the requested
revisions and schedule another peer review meeting.

       On June 10, 1999, the Committee held a Consultation with NCEA Staff regarding the
development of the next revision of the draft diesel document (CASAC, 1999a). This was done as a
precursor to the next full peer review planned for later that year.  On December 1, 1999, CASAC
Diesel Review Panel met in Research Triangle Park, NC to conduct a peer review of the Agency's
revised draft Health Assessment Document for Diesel Engine Emissions (EPA, 1999). As a result of
that review, it was clearly apparent that Staff had made a strong effort to respond to the Panel's earlier

-------
recommendations in developing the revised draft. However, the number of major and minor remaining
criticisms and recommendations raised by the Panel once again precluded closure on the draft
document.  In particular, there was substantial concern for the approach taken to deriving the
uncertainty factors used in calculating the reference concentration value (RfC) for noncancer health
effects.  Secondly, the majority of the Panel disagreed with the Agency's use of the description "highly
likely" to portray cancer hazard from long-term environmental exposures. These recommendations,
and others, are covered in detail in the CASAC report of that review (CASAC, 1999b).  The CASAC
asked to see a further revision of the draft document, which is the subject of this current CASAC
report.

2.2     Charge

        The chief purpose of this review was to determine whether or not changes made in response to
previous CASAC guidance resulted in a revised document satisfactory  for closure. No other specific
charge questions or issues were raised by Staff.

-------
                3.  COMMENTS BY CASAC ON EACH CHAPTER
       In developing the final version of the document, we advise the Agency to consider the full range
of suggestions contained in the detailed written comments of the individual Panel members (see
Appendix A), the summary minutes of the meeting, and comments made during the meeting. Because a
full transcript of the meeting was not created, Staff are encouraged to interact with the Committee Chair
as may be necessary to ensure that key revisions are consistent with the intent of the agreements
reached at the meeting. Only  summary comments concerning key issues are contained in this report.

3.1     Chapter 1: Executive Summary

       This chapter will need to be examined carefully and revised as necessary to ensure that it
reflects changes made in subsequent chapters.

       It would be useful to include a statement on the motivation for developing this Hazard
Assessment Document and its relationship to regulatory decision-making.

       It would be useful  to briefly summarize emissions trends and the contribution of diesel emissions
to the nation's emissions inventories and pollutant burden.

3.2     Chapter 2: Diesel Emissions, Characterization, Atmospheric Transformation, and
       Exposures

       This chapter is much improved. The Panel offered several additional references for inclusion in
the final version.

       Tables summarizing emissions trends and the contribution of diesel particulate matter (DPM)
and nitrogen oxides to the total emissions inventory should be added to the chapter, in order to better
place the contribution of diesels into context.

       The chapter should make clearer that DPM is a subset of ambient PM, rather than something
separate from ambient PM, as the present wording appears to suggest in several places. It can be
clearly stated that although the composition and potential toxicity of DPM may differ from those of the
composite ambient PM, DPM is a ubiquitous component of ambient PM.
       Different techniques have been used for measuring elemental carbon and organic carbon, and
they yield significantly different results. Without specification of the measurement method, results from
different reports may not be directly comparable. Indications of the measurement method should be
added to citations of results from different studies and comparisons among them.
       The chapter appropriately mentions differences in fuels used in on-road, off-road, and railroad
diesel engines.  However this  and subsequent chapters fail to  tie these differences to their potential

-------
implications for health hazards. For example, would differences between on-road and railroad diesel
fuels have any implications for the possible interpretation or comparisons of epidemiological results from
truck drivers and railroad workers?  Stationary sources should also be discussed. It also does not
indicate that the changes made in on-road diesels to reduce their emissions may have also changed the
toxicological characteristics of those emissions (for example, reduced amounts of organic carbon
associated with the particles).

3.3    Chapter 3:  Dosimetry of Diesel Exhaust Particles in the Respiratory Tract

       The fact that DPM exhibit little hygroscopic growth has bearing on the estimation of deposition
in the respiratory tract, and should be mentioned.

       The portrayal of deposition should be expanded by considering the entire respiratory tract,
rather than just focusing on deposition in the lung.  Substantial deposition also occurs elsewhere, and
this knowledge is important to placing non-lung health effects into context (e.g., nasal deposition and
nasal immunological responses). Figures illustrating the relationship between regional deposition
fraction and particle size are readily available, and an example should be included.

       The discussion of different deposition models should be accompanied by a comparative
presentation of example results using the different models.  For example, it would help to place the
deposition assumptions and results of the model used to derive the RfC into context regarding other
broadly-accepted models like those of the NCRP or ICRP.

       The discussion of interspecies differences in the interstitialization of particles should note that,
although  differences between the proportion of particles in the interstitium of rats  and non-human
primates  have been observed, the observations have been at single times after chronic exposure, and it
is possible that comparable amounts of material enter the interstitium of rats, but move more rapidly to
lymph nodes and other locations than in primates. It should also be stated that the information cited in
regard to interstitialization is derived from high-dose exposures, and that no such information exists for
exposures at environmental levels. Points were raised in individual comments that would enhance the
discussion of how interstitialization is treated in the Yu model.

       This chapter still does not adequately portray the plausible doses of organic compounds to
airway or pulmonary cells that might result from either environmental or occupational exposures. Some
straightforward "order of magnitude" calculations could be added, and would help place the subsequent
information on high-dose in vitro mutagenicity test results into a more useful perspective.  This issue
was raised repeatedly in reviews of previous drafts, and has still not been adequately addressed.

3.4    Chapter 4:  Mutagenicity of Diesel Exhaust

-------
       It would be useful to bring the discussion of our understanding of the linkages between
mutagenicity and carcinogenicity up to the front of the chapter. This should be followed by the
information briefly describing the history of diesel-related mutagenicity research, which could be
usefully, but still succinctly, expanded as suggested by individual reviewer comments.

       As noted above, the issue of the relationship between doses used in mutagenicity assays and
plausible doses received from environmental exposures should be discussed. The chapter does a
disservice by not placing dose into better perspective.

3.5    Chapter 5: Noncancer Health Effects of Diesel Exhaust

       This chapter fails to mention the exposure or dose used in many of the studies cited. Without
some indication of dose, information on health responses has little value.  The relationship of
experimental doses to those from resulting from plausible human exposures is key to understanding
whether responses are likely to be of public health concern.

       The chapter contains a good breadth of information on non-cancer health issues; however,
there is insufficient depth of interpretation of much of the information.  The superficial treatment  of the
information leads to both under- and over-interpretations. A few examples follow. Many of the
cellular and chemical changes listed are biological markers that are consistent with asthma, but the
studies do not demonstrate that asthma was produced. Many of the differences between results of
animal studies  are more likely due to differences in study design, rather than to true interspecies
differences, as presently implied. The chapter appears ignorant of the implications of the important
differences between studies. The discussion of the increase in immune responses caused by pyrene
implies that pyrene is a concern. The discussion does not clarify that  pyrene was used as a single model
compound and was not compared to other compounds. It is not mentioned that other results indicate
that adjuvant activity is characteristic of multiple organic species,  and among multiple combustion
emissions.

       The chapter states wrongly that there have been no well-controlled chamber studies. Studies
are cited in the Panelists comments.

       The treatment of roadside asthma studies is too superficial. Present information should be
described more thoroughly, including a better distinction between asthma prevalence and exacerbation
of asthmatic responses.

       A number of key references to be considered for inclusion in the final document are offered in
individual Panelist's comments (see Appendix A).

3.6    Chapter 6: Quantitative Approaches to Estimating Human Noncancer Health  Risks of
       Diesel Exhaust

-------
        The presentation of information linking effects of DPM to those of ambient PM is improved, but
the comparison remains difficult. The contribution to the difficulty in making the linkage of the different
types of experimental approaches taken to DPM and ambient PM should be explained.  The
suggestions in individual Panelist's comments for additional citations for this material, and for discussing
their implications, should be incorporated.

        The information contained in the appendix on benchmark concentration analysis should be
discussed in a paragraph, rather than the present single sentence.

        The description of the derivation of the human equivalent concentration (HEC) needs further
clarification.  First, as agreed at the meeting, the section can be enhanced substantially by the addition
of a table listing the key input parameters and values used for rats and humans in the Yu model.
Second, one can determine from the appendix that the different patterns of rat exposures were
normalized to continuous exposures to derive the HEC, but this fact and the uncertainties thus
introduced into the extrapolation need to be stated clearly in the text. The statement that time averaging
was not part of the assessment appears in conflict with the use of time averaging in deriving the HEC.
Third, it should also be explained in the text why the relationship between rat exposure concentration
and HEC are not proportional over the concentration range listed in Table 6-2.

        The approach to be taken to characterizing a safe level for noncancer effects of long-term
exposure was discussed at length.  Staff had responded to the previous criticism of using an uncertainty
factor based on irnmunological responses to modify the no-effects HEC based on rat lung pathology by
removing the factor; thus increasing the reference concentration (RfC) from 5 to 14  ug/rn3. There was
general  agreement that irnmunological effects may possibly supersede lung pathology as the critical
health concern, but that:  1) irnmunological responses were more  likely a concern for acute, rather than
chronic, exposures;  and 2) present knowledge was insufficient for deriving a reference concentration
(RfC) for DPM based on immunological responses.  There was also general  agreement that it was
reasonable to retain  some form of uncertainty factor for interspecies extrapolation.

        Staff had also responded to the request to expand on the linkage between the safe level of
DPM and the proposed annual NAAQS for PM2 5.  It was generally agreed that some form of such a
comparison should be retained, but that current knowledge was insufficient to describe the relative
potencies of DPM and other components of PM25 with confidence.

       After discussing several options, agreement was reached between the Committee and Staff to
deal with these issues by presenting two perspectives, along with caveats regarding the uncertainties
involved.

       RfC for DPM: An RfC derived in the same manner as that in the draft and based on rat lung
       pathology will be included. An uncertainty factor of 3 for interspecies extrapolation will be

-------
used, together with a factor of 10 for differences in individual sensitivity, resulting in an RfC of 5
ug/m3.

DPM vs. PM2_5:  A discussion will be included which draws the general conclusion that, as
long as DPM continues to constitute it's approximate present proportion of PM2 5, an annual
PM2.5 standard would be considered adequately protective for DPM.

-------
3.7     Chapter?: Carcinogenicity of Diesel Exhaust

        The value of the Hill criteria for supporting the Agency's case for cancer causation was
questioned, especially considering their reliance on the Rothman interpretation of the criteria for
explaining why individual criteria did not have to be met. Perhaps only the consistency and biological
plausibility criteria have been clearly met; there was general agreement that the dose-response criterion
has not been clearly met.

        Latency of effect remains a key issue in interpreting the present database to indicate that diesel
emissions are carcinogenic in an exposure-related manner. Latency should be discussed more explicitly
in regard to the studies cited, indicating which studies provide useful information on latency, what range
of latency is present in each, and which studies provide no information on latency. This information
might best be provided in a table.

        The brief discussion of the California EPA and Crump analyses of the epidemiological data
should be expanded to describe  the differences, and the more recent analysis organized by the Health
Effects  Institute should also be included and described.

        There continued to be concern for the use of the rat inhalation studies as part of the weight of
evidence argument, as had been expressed during previous reviews. It was agreed that other
laboratory results provided sufficient biological plausibility for a cancer hazard.  Staff agreed to exclude
the rat inhalation results as part of the weight of evidence for carcinogenicity, although the summary
characterization of the evidence  would not be changed as a result.

        The conclusion in the weight of evidence section that diesel exhaust is likely to be carcinogenic
to humans is consistent with the  Committee's previous guidance.  There remained, however, mixed
views concerning: 1) the use of the term exhaust instead of citing a component, or components, of
exhaust; and 2) the further characterization, at any exposure condition. The former could be
misinterpreted to suggest that no clean-up strategy could reduce concern as long as something was
emitted, even if only water and carbon dioxide. The latter implies a clearer understanding of the
exposure-response relationship than presently exists, and should be clarified as a default assumption in
the absence of complete information.

3.8     Chapter 8: Dose-Response Assessment: Carcinogenic Effects

        The appendix on epidemiological studies warrants a summary paragraph, rather than the
present  single sentence.

        There should be a clearer, more methodical, more explicit summary of the uncertainties in
interpreting the epidemiological  database with respect to dose-response.  This summary will enhance
the reader's understanding of the caveats given later regarding the range of risk values.

                                               10

-------
       The Committee continued to agree with Staffs decision not to adopt a unit risk value.  It was
agreed that no single existing data set, or combination of existing data sets, allows for the calculation of
an estimate of unit cancer risk with acceptable confidence. It was felt unlikely that continued evaluation
of data sets from past epidemiological studies will resolve the uncertainties to a satisfactory degree, due
primarily to the lack of exposure information.  There were mixed views regarding the likelihood that
future studies will provide an acceptable unit risk value applicable to environmental exposures.

       The inclusion of a range of cancer risk values to provide a perspective of the possible range of
lung cancer risk from environmental exposures to DE was a major topic of discussion and a pivotal
issue in the Committee's decision to close on the document. The range first appeared in Chapter 8,
and appeared again in Chapter 9. Staff made clear its intent that its listing of the values  was not to be
interpreted as the Agency's endorsement of their use as unit risk values.  There were mixed views
among both the Committee and the Consultants regarding the appropriateness of including the range;
however, there was general agreement that, despite Agency disclaimers, the publication of the range
would likely be cited as endorsement of the values as unit risks for estimating cancer deaths.
Consensus was already established that no unit risk value could be calculated with  sufficient confidence
to be presented in the document  as an Agency position.  The difficult issue, therefore, was the conflict
between the use of the values to portray a possible range of risk and the agreement that no satisfactory
unit risk value could be selected.

       Staff emphasized that the range should be included in order to illustrate that the most likely
magnitude of risk is  sufficient to  be of public health concern; i.e., that the risk is  not negligible and
warrants continued action to control exposures to diesel emissions.  Staff did not find acceptable the
recommendations offered by  some Panelists for relying solely on a text characterization of likely risk.

       The Panel generally, but not unanimously agreed that the inclusion of the range of values would
not prevent a  recommendation for closure on the document, pending the accompanying inclusion of
satisfactory caveats and disclaimers.  The nature of the caveats was discussed and specific language
was offered by some Panelists. Although agreement was not reached on  specific language, it was
agreed that the disclaimers would include clear statements that: 1) that the values were  attended by
considerable unresolved uncertainty; 2) that their inclusion in the document did not constitute Agency
endorsement of their validity as unit risk values; 3) that the values are not proposed as useful for
estimating numbers of cancer deaths; and 4) that the range of potential risk from environmental
exposures was very broad and included at its lower bound the possibility of zero risk.

3.9    Chapter 9: Characterization of Potential Human Health Effects of Diesel Exhaust:
       Hazard  and Dose-Response Assessments

       This chapter largely recaps the preceding material.  Several comments  and recommendations
are contained in the individual comments of the Panelists, and should be considered in developing the
final document. Most of these comments repeat and reinforce issues raised in  other chapters.

                                              11

-------
       The basis for the statement that carcinogenicity is related to particle size is unclear. The
meaning of the statement, which only occurs in this chapter, is not explained and no data are given to
support the claim.  The statement should be removed.
                                              12

-------
                                   4. CONCLUSIONS
       The consultants offered mixed recommendations regarding closure on the document, with some
recommending re-review of some or all of the changes made in response to this review.

       The Committee was unanimous in recommending closure pending Staffs careful attention to all
comments offered at the meeting and in writing, and pending the revisions to which the Committee and
Staff agreed at the meeting. It was the Committee's view that, despite the considerable remaining
uncertainties regarding the health impacts of diesel emissions, the document as appropriately revised will
constitute a reasonable portrayal of current knowledge in the field.
                                             13

-------
                              REFERENCES CITED
CASAC. 1995.  Review of the Diesel Health Assessment, EPA-SAB-CASAC-LTR-95-003,
       Clean Air Scientific Advisory Committee, US EPA Science Advisory Board, Washington, DC.

CASAC. 1998.  CASAC Review of the Draft Diesel Health Assessment, EPA-SAB-CASAC-99-
       001. Clean Air Scientific Advisory Committee, US EPA Science Advisory Board,
       Washington,DC, October?, 1998.

CASAC. 1999a. Notification of a Consultation on the Diesel Health Assessment, EPA-SAB-
       CASAC-CON-99-005, Clean Air Scientific Advisory Committee, US EPA Science Advisory
       Board, Washington, DC, July 9, 1999.

CASAC. 1999b. CASAC Review of the Draft Diesel Health Assessment Document, EPA-SAB-
       CASAC-00-004. Clean Air Scientific Advisory Committee, US EPA Science Advisory
       Board, Washington, DC,  February 4, 2000.

EPA.  1998. Peer Review Handbook. EPA 100-B-98-001, Science Policy Council, Office of
       Research  and Development, US Environmental Protection Agency, Washington, DC, January
       1998.

EPA.  1999. Health Assessment Document for Diesel Emissions (EPA 600/8-90/057D), US EPA,
       Office of Research and Development, Washington, DC, November 1999.

EPA.  2000. Health Assessment Document for Diesel Exhaust (EPA 600/8-90/057E),US EPA,
       Office of Research and Development, Washington, DC, July 2000.
Please note:  The above referenced CASAC reports are available on the SAB website at
             www.epa.gov.sab under the REPORTS heading. The EPA reports are available from
             the Office of Research and Development (ORD) under www.epa.gov/ncea or
             www.epa.gov/ORD/spc/2polDrog.htm
                                       R-l

-------
      APPENDIX A - INDIVIDUAL PANELIST WRITTEN COMMENTS
      Note: These are the final written comments provided by individual Panelists following the
October 12-13,2000 meeting. They are included here to present the full range of opinion and to
document all edits suggested by panelists.
                                       A-l

-------
 APPENDIX A - TABLE OF CONTENTS

Comments by:	Page A-
Dr. Mauderly                                3
Dr. Hopke                                 12
Dr. White                                  13
Dr. Upton                                 19
Dr.Vedal                                  21
Dr. Diaz-Sanchez                            28
Dr. Garshick                                30
Dr. McClellan                              36
Dr. Oberdorster                             49
Dr. Wyzga                                 56
Dr. Stayner                                62
                   A-2

-------
                                  Joe L. Mauderly. DVM

GENERAL COMMENTS
       This revision is a substantial improvement over the last draft, and reflects a serious attempt to
incorporate the comments and guidance of the CAS AC panel.

       Overall, I would agree to close on the document in its present general form, pending
satisfactory edits by NCEA to address residual key issues and numerous lesser editorial items. The
majority of my comments address editorial faults and lack of clear statements or descriptions of studies,
rather than substantive issues.  However, some of these "minor" problems cloud the reader's
understanding of the facts. There is no reason the document can't be written well, as well as
adequately portraying the current state of knowledge and coming to the correct key conclusions.
Although one might avoid detailed editing in drafts, the final document will require careful examination
throughout for minor,  as well as major, changes.

       In the final version, the figures and tables need to be inserted into their proper places in the text.

SPECIFIC COMMENTS
Chapter 1:  Executive Summary

1-3, 7: This is a strange statement. Can you think of any portion of the U.S. population that is not
exposed to  "PM2 5 of which DE is a part"? Why not just say that everyone is exposed?

Chapter 2:  Diesel Emissions Characterization. Atmospheric Transformation, and Exposures

2-2, 35: Compression ignition is not just an example of a type of internal combustion engine in which
diesel fuel is burned - it is the only type. Just eliminate the "e.g.".

2-3, 4-10: This description is confusing. In line 7, you state that DPM "are considered fresh after
being emitted and undergoing aging".  I don't think that's what you mean. The point  is well taken that
there is fresh DPM and aged DPM, but the wording of the paragraph needs some attention.

2-8, 18: It does little good to refer to the "Zeldovich mechanism" unless you explain briefly what it is,
or at least give  a reference. Very few readers would have heard that term before.

2-9, 8 and 29 (and elsewhere): The terminology in sections describing DPM and ambient PM need to
be tightened up. DPM is a ubiquitous part of ambient  PM. Of course, if you analyze pure DPM it will
have a composition overlapping with,  but different on average than,  ambient PM.  The toxicity of
ambient PM samples and pure DPM may well be different. Regardless, DPM is part of ambient PM.
With a little attention,  the wording can be changed to more accurately explain your point. Taken to
ridiculous extreme, if one listed each subset of ambient PM as different from ambient PM, there

                                            A-3

-------
wouldn't be any ambient PM left!  The sloppy terminology tends to mislead the naive reader in thinking
that DPM and ambient PM are two completely different things.  The opportunity that is lost in the
present wording is to provide a clear and accurate perspective that the many materials comprising
ambient PM differ in nature; thus, the average composition necessarily differs from the composition of
any individual component. That's not difficult to state.

2-13, 31:  Shouldn't it read "and the maximum" instead of "or the maximum"?

2-22, 36:  Here  is the Zeldovich mechanism again. If it's worth citing twice, it's worth explaining, or at
least referencing, once.

2-25, 23:  If the "hydraulic-electronic unit injection" the system commonly called "common rail"? If so,
since "common rail" is the more commonly-used term, it would be good to put it in parentheses in this
sentence.

2-27, 6: First, has "DDEC" been defined? If not, the abbreviation is worthless.  Second, write out
Detroit Diesel Corporation. You don't use the abbreviation DDC consistently (eg, next page, line 30),
and you write out names of other engine manufacturers.

2-28, 33-34:  This sentence needs fixing. You use the term "two stroke" twice.

2-32,4: You make  the point that the sample from the tunnel could  be  taken as representative of heavy
engine emissions, and you may be correct.  However, you support that premise with two facts that are
not convincing. First, 25.7% doesn't seem like a "relatively large" percentage, as stated.  Relative to
what?  Second,  the number of heavy-duty vehicles passing through the runnel doesn't mean that the
sample was predominated by those vehicles. You would have to make the case on the basis of the
relative volume  of emissions from heavy and light-duty vehicles, but you don't make that case directly.

2-34, 15 and 31: Here is "DDC" again.

2-41,17:  Here, you use the term "intercooled". Elsewhere you use the term "aftercooled".  Neither
term is defined.  Are they the same?

2-42, 30:  The terminology here is not clear. Do you mean a decrease in the amount of emissions, or a
decrease in the range (number, type?) of "factors"?

2-44, 1: A given set of dilution tunnel conditions can model "what occurs" in the atmosphere, but only
under that one condition.  The point is not that dilution tunnel results can not be representative (ie,
accurate), the point is that no dilution tunnel condition can accurately represent the frail set of conditions,
or range of conditions that occur in the environment.  Without that clarification, a naive reader would
                                             A-4

-------
assume that dilution tunnel results have no value, rather than understanding correctly that their value is
limited to a specific set of conditions.

2-46, 34: Sentence needs to be fixed.

2-52, 6-12: There are conflicting statements here. It is stated that DPM have limited hygroscopic
growth when fresh, but have more when aged. It is then stated that DPM "do not appear to undergo
hygroscopic growth once emitted to the atmosphere". You can't have it both ways.

2-54, 35: Presumably, the first "PM" in the sentence should be "DPM".

2-57, 31-32:  It sure doesn't. Mel Zeldin (SCAQMD) has stated in public meetings that they believe
that today, only about 33% of EC in the basin is from DPM. This is a big difference from 67%. If you
want to be  contemporary, you might want to call Mel and include a quote here.

2-63, 21: I've heard of biomarkers for benzene exposure, but what biomarker is specific for particle-
associated benzene exposure? If there isn't one, then this statement is misleading.

2-67, 12-13:  This information needs clarification. If the freeway contributed 0.7 to 4.0 "excess"
DPM, what does "a maximum of 7.5" mean? Is this including background?  What measurement was
taken as "background"?

Chapter 3:  Dosimetry of Diesel Particulate Matter

3-1,  28:  Bad sentence - "here" is Chapter 3, not Chapter 2. Just say they are described in Chapter 2.

3-1,  29-30: It should read "—are the focal points—".

3-2,  15:  I believe that a larger range is portrayed in Chapter 2.

3-3,  25:  You already defined these abbreviations above.

3-3,  33:  Some agglomerated DPM are larger than 1 micron. Most are not, but the statement isn't true
as it  stands.

3-5, 24-25: The presentation of the issue of interspecies differences (or similarities) in deposition is
confusing.  The statement here, along with figure 3-1, indicates that there is no important species
difference in lung deposition. First the reader has no idea what is meant by similar "relative deposition"
Second, this premise conflicts with the three-fold differences in deposited dose given in Table 3-1. Are
you saying that all of the interspecies difference in deposited dose is due to differences in ventilation
rate? If not, then we need more explanation. If so, is that consistent with current best thinking.

                                             A-5

-------
3-6, 26: I think you mean "different" particles, not "specific" particles.

3-7, 34: Are you making the case that there would be no difference in clearance from either region for
DPM vs. 4 micron particles? Do we know that?

3-8, 18: It was exposure of rats to whole exhaust containing DPM, not just to "whole DPM".

3-8, 24: Again, the exposure was to whole exhaust containing these concentrations of DPM. The
same problem appears again in lines 36, 31, and 33.

3-9, 5: It would be more informative to state that a larger fraction of DPM translocated to the
interstitium in heavily-exposed primates than in heavily-exposed rats. We really don't have information
on whether the same difference might exist in animals exposed to environmental levels of DPM.

3-9, 16-18: It would be more informative to state that the reason it isn't relevant is that there were no
airway tumors in the rats. The rumors were all parenchymal, so nothing that happened in the airways
could have directly affected them.

3-10, 24:  Did Chan expose rats to whole exhaust containing this concentration of DPM, or just to
resuspended DPM, as the present wording implies?

3-11, 20:  I think there should be an opening parenthesis before "CxT".

3-12, lines 19, 29, and 32:  I believe that the exposures were to whole exhaust in all these cases.  Same
on P3-13, 11.

3-20, lines 29, 34:  Again, the exposures weren't just to DPM, they were to whole exhaust.

3-22, 18:  It should be "rodents" rather than "animals". The predominant site was not alveolar
macrophages in primates. 1 guess it might have been if you postulate that the interstitial burden was
contained  in "alveolar" macrophages that migrated there.  Are you making that case?

3-29,4: The sentence need fixed. "Some deposition conducting airways" doesn't make sense.

3-31, lines 6,7: So what precludes you from using a multiple-path model for deposition and some other
model for  clearance?

3-36, 29: Which "much  larger particle" are you referring to - 0.4 or 2.0 microns?

3-37,16: "Extimated" is  a misspelling.
                                            A-6

-------
3-44, Table 3-2:  Shouldn't the title read "—exposed to DPM in whole exhausf'? As in the text,
implying that the exposures were to only DPM is misleading.

3-48, figure legend:  The "3" at the end should be a superscript.

3-49, figure legend: It should be "mg DPM". not "mg PPM".
                                           A-7

-------
Chapter 4: Mutagenicity

4-3, 18:  First, what is "tar"? This is the only place in the document that uses that term. Regardless of
what the authors might have called it, you need to put the exposure material in context within the
terminology used in this document. In view of the preceding sentence, you must mean "extract".
Second, why cite a reference that dosed animal intraperitoneally with 2-4 g extract per kg? That would
be equivalent to 140-280 g in a 70 kg human!  A human wouldn't absorb that much extract in a
lifetime, much less a female during a single pregnancy!

4-3, 24:  Presumably the exposure was to whole exhaust, not just DPM. This is especially important
when you are talking about mutagenicity, given the amount of SVOCs that go through filters.

4-4, 17:  In what cells were the adducts measured?

4-5, 24:  Was the exposure to whole exhaust?

Chapter 5: Noncancer Health Effects of Diesel Exhaust

Note: This chapter does a particularly good job of providing summary statements at the end of each
section. The other chapters would do well to emulate this one in that regard.  On the other hand, this
chapter stands out as failing to mention the exposure or dose used in many of the studies cited.

5-3, 17-18:  How could an increase in nasal ascorbate prevent further oxidant stress in the "respiratory
tract"? Presumably, you mean only in the nose - or do you mean that you are assuming an increase in
ascorbate throughout the tract?

5-6, 23:  It would be better to have a separate heading for the "traffic" studies. They are really a
different type of study in which "exposure" is quantitated in terms of traffic rather than any airborne
material.

5-8,4: I know I'm from the provinces, but what is an "express tunnel", and would someone be sitting
in one - as the sentence implies?

5-9, 5-10, and 5-11:  The doses are not stated in several of these paragraphs. Specifically, the
paragraphs beginning on 5-9,11, 5-9, 34,  5-10, 26, 5-11, 9, and 5-11, 17. Without some indication
of dose, the information is of limited value.

5-21, 6-8: This statement doesn't derive from the studies presented in this section. If you want to
make a case for the organics (which is reasonable), then cite the studies demonstrating the relationship.
If you can't cite studies, don't make the statement.
                                             A-8

-------
5-40, 25: What was the exposure concentration?

5-41,16: What was the range of doses?

5-41,21: What was the dose? Why cite studies using intraperitoneal injection when there are several
studies dosing via the respiratory tract?

5-41,30: What was the dose?

5-42, 9: What was the dose? Why cite a footpad study when we have respiratory studies?

5-42,20 and 30; 5-44,  18 and 30; 5-45, 15 and 22:  What were the doses in these studies?

5-53, 27; and 5-54, 1:  What were the exposure concentrations in these studies?

5-57, 5: The wording states that pyrene adsorbed to DPM augments the adjuvant effect. That is not
what the study demonstrated. What you don't say is that the only single organic compound they tried
was pyrene. They used that as a model compound, and speculated that it may also act that way on
DPM.  While it is true that the studies showed that pyrene could have this effect, it is very misleading to
imply that pyrene adsorbed to DPM causes the effect, much less is chiefly responsible for the effect. It
is likely that several organic compounds could have this effect in pure form, and also  a reasonable
hypothesis that they do the  same when attached to DPM. You do a disservice, however, by citing the
study in a way that leads the reader to a misunderstanding of the facts. This whole issue is very
important, and it is appropriately cited in this chapter. There is no reason to cloud the issue by not
being clear  about what we do and don't know at this time. It  would also provide perspective to note
the dose of pyrene used to achieve this effect, and how that relates to doses expected from
environmental exposures.

5-59, 9: What was the dose in this study?

5-59,29: This statement is exactly backwards. There was less aggregation in the mice, not in the rats.

5-64, 9: It should say "—lung tumors in rats".

5-69, 25: It would probably be more precise to use "potency" rather than "efficacy".

Chapter 6:  Quantitative Approaches for Estimating Human Noncancer Risks of Diesel Exhaust

6-2,21: Is "adjuvancy" really a word?

6-4, 24: Do you really mean PM15, or do you mean PM10?

                                            A-9

-------
 6-11,19-21:  Is this really true? What about the Fred Miller et al. Model - isn't that one
 parameterized for humans and rats and currently available?
 6-12, 7: Didn't you use the retained lung burden as the "dose", rather than the air concentration?  An
 air concentration is an exposure metric, not a dose metric. I thought that the model(s) used for HEC
 compared species on the basis of retained PM.

 6-12: It would seem that  somewhere on this page, it should be made explicit that the extrapolation
 assumes the equivalence of CxT = CxT. That is, "dose" was adjusted across species assuming that a
 weekly non-continuous exposure could be extrapolated to a continuous exposure by dividing by the
 hourly ratio. This is a reasonable assumption under the circumstances, but is an important feature of the
 extrapolation. For example, this is an important assumption underlying the information given in 6-13,
 17-21. The reader can catch this point from footnote b in Table 6-2, but it should be made clear in the
 text.

 6-16,4: You could say that some studies report, or that studies occasionally report, but you probably
 shouldn't say that some studies occasionally report.

 6-17, 32-34: This sentence  is confusing. You note that environmental exposures can be above the
 RfC for short times, which is true. You then say that time-averaging is not part of the assessment. In
 calculating the RfC, you did time average the animal exposures to estimate and equivalent continuous
 human exposure.  All of the animal exposures were episodic, in the sense that none were continuous.
 Thus, an acceptance  of time  averaging is implicit in your method. Surely you don't think that anyone
 would be exposed to 14 ug/m3 continuously for a lifetime!  Overall, to say that time averaging was not
 part of the assessment is misleading.

 6-18, Ishinishi et al. Data: You should note that the two data sets are for LD and FID. Also, why is
 there a space between 1.84  and 3.72 lines in the HD  data set?

 6-20, Table 6-2:  One cannot tell from the table or the explanation why you lumped together 2.44 and
 6.3 exposure levels for the Nikula 1995 study.

 Chapter 7:  Carcinogenicity of Diesel Exhaust

 7-1, 23-25: You should note that these are estimates, or modeled data. Simple inserting "it was
 estimated" would suffice.

 7-2, 8-9:  I think you are confusing the reader here about the difference between "health risk" and
 "burden".  The health burden for DPM can't be larger than the health burden for PM.  However, the
health "risk" in terms of unit risks, could be much higher for DPM than for PM. There is enough
confusion on the street about this issue without contributing to it!
                                            A-10

-------
7-4, 15: The term "potentially exposed" is misused here. Of course the workers were exposed -
everyone is. The point is that these workers had potentially much higher exposures than average.

7-7, 7: Don't you mean "positive" association instead of negative? A cancer effect would be a positive
association.

7-64, 4: Imbalances in smoking prevalence are only one way smoking could be a confounder.  If
smoking acts synergistically with diesel exposure, then smoking could be an important confounder even
if smoking was similar among exposed and control populations.

7-65,  12-13: Why not just say that Hill provided a set of criteria? Others have used the criteria, but
they didn't "provide" them. The criteria that I've heard used by all during the past several years all
stemmed from the Hill list

7-67, 28: The references you list are all "rat" references.  Why not say that diesel exhaust has been
shown to cause cancers in rats, rather than in "animals".  If you really mean to be more inclusive, add
some other references.

7-132: The high concentration in the Mauderly mouse study is listed as 7.0 mg/m3. In the text and on
page 7-127, it is cited as 7.1 mg/m3. The reported concentration was 7.08 mg/m3, so 7.1 would be the
more  correct. At least be consistent.

7-139, Table 7-9: There is no "DHHS, 2000" in the reference list.

Chapter 8:  Dose-Response Assessment: Carcinogenic Effects

8-1,4: There is an "and" missing between "data" and "Discusses".

8-13,  17: Wouldn't the equivalent exposure be 21  ug/m3 instead of 20?

8-13,  28: The point here is not clear.  How are you separating on-road from non-road sources?
Surely you don't mean the this entire document has only discussed risk from on-road sources. The
hazards and risks described throughout the  document presumably include those from diesel exhaust
from all sources. How then do you propose that exposure to off-road sources confers some special
risk?  This doesn't make sense.

Chapter 9:  Characterization of Potential Human Health Effects of Diesel Exhaust: Hazard and Dose-
Response Assessments

9-3,2: It should read "—most other ambient PM -". The present wording suggests that DPM is not
part of ambient PM.

                                            A-ll

-------
9-9,25: It should read "—exposure to high concentrations of DE -".  As for cancer, non-cancer
effects have only been demonstrated in animals at high concentrations of exposure. Regardless of how
you extrapolate to humans, the fact remains that animals have only demonstrated effects at high
exposure levels.

9-9,30: Presumably, you mean "—mainly from rats -". I've looked inside lots of rats and I've never
seen any data in there!

9-13, 17-18: The meaning here is not  clear. What do you mean by stating that the carcinogenicity
appears related to particle size? Do you have data on carcinogenicity of diesel soot of different sizes
that supports this premise.  Regardless of what you might have meant here, the statement should be
clarified. It could be taken, for example, to suggest that ultrafine DPM are more carcinogenic than
"regular" DPM.  We have no data suggesting that.

9-13, 22: The meaning of this statement is not clear. I presume that you must mean that the organics
may have a greater "relative" importance at low exposure levels. If they have importance at all, then
they would have importance at all levels. You must mean that since elemental carbon is thought to be
important in the rat response at high levels, but there is no rat response  at low levels, the organics could
still be important in humans at low levels. That is a reasonable hypothesis, but that doesn't mean that
the organics are any less important  at high levels of exposure.

9-17, 25: You don't mean 0.14 ug/m3 here, you mean 0.14 mg/m3.  It would be better to use  14 ug/m3
to be consistent with the RfC chapter.

9-18, 21: It should read "-0- limit for DPM would be -".  If you are talking about "DE", the mass
concentration would be higher. I think you are just talking about DPM.

9-18, 25: Where else would DPM come from other than DE sources?

9-18, 26-29: This statement doesn't make sense. It could only apply if you some standard portion of
PM was DPM. If you want to speculate about relative toxicity, this isn't the way to do it. The whole
example doesn't make sense. If DPM and other ambient PM have the same toxicity, you could have
DPM at any concentration up to 15 ug/m3 and not exceed the toxic hazard of the 15 ug/m3  annual PM
standard. DPM constitutes a variable portion of PM.  It may be more  or less toxic than other  PM.  If
all PM were equally toxic, you could have any mix up to a total concentration of 15 ug/m3 and  meet the
intent of the annual standard.  The general point that not all PM are likely to be equally toxic and that
DPM may be more toxic than most is a reasonable one to make, but the present wording only  confuses
the issue.

9-19, 3-7:  I don't agree that the apparent congruence of the RiC and the  annual PM standard suggests
the validity of either.  I could argue equally well that the congruence suggests that one or both must be

                                            A-12

-------
wrong. Happening to come out with close to the same number may be reasonable and acceptable, but
that certainly doesn't support the validity of either number.  Remember, in the last draft you were
equally convinced that the RFC should be 3 times lower. Overall, you'd better let well enough alone
and not try to construct a circular argument for why the congruence imparts confidence.

9-24, 25: Just which area in the U.S. does not have ozone present? There is ambient ozone
everywhere.  The statement doesn't make sense. Now if you want  to argue from the basis of data in
hand that there is some synergism and that DPM may have greater effects as accompanying ozone
exposures increase, you may have a defendable point. However, that's not what the sentence says.
                                           A-13

-------
                                      Philip Hopke. PhD

Chapter 2

        They have made a very comprehensive review of the information that is available on diesel
emission rates, characteristics of those emissions and the amounts of diesel particles that can be
estimated to be present in the ambient air. One of the problems with this chapter is that it does not
provide any clear links to the health risks of diesel exhaust. It would be useful to provide some pointers
to the discussions that will come later in the document. For example, given the information they provide
with respect to the significant differences in railroad diesel fuel relative to highway and off-road diesel,
there should be some comment about how this might affect the epidemiology of railroad workers
relative to truckers.

        One aspect that is still missing in either chapter 2 or possible chapter 3  is the response of diesel
particles to the high humidity conditions it would encounter in the respiratory tract. One of the
important characteristics of diesel particles is that they do not demonstrate significant hygroscopic
growth. Hygroscopic growth has been examined by several investigators (Weingartner et al., 1993;
1997; Dua et al., 1999) and find that these particles do not exhibit growth.  In  fact, they appear to have
some decrease in size with higher humidity possibly through the collapse of the fractal aggregate
structure.  This point has been raised in prior reviews, but has yet to be added.  This raises questions
about their level of understanding of particle dosimetry.

References

Dua, S.K., P.K. Hopke and T. Raunemaa, (1999) Water, Air,  Soil Pollution 112: 247-257.

Weingartner, E., Burtscher, H., and Baltensperger, U. (1993). J. Aerosol Sci.  24:8371-8372.

Weingartner, E., Burtscher, H., and Baltensperger, U. (1997). Atmospheric Environ. 31:2311-2327.
                                            A-14

-------
                                  Warren H. White. PhD

Chapter 2

       In the subset of tables and figures that my interests led me to examine closely, I came across
several substantive errors.

Table 2-18. I can't find PAH emission rates broken out by different fuel types anywhere in Norbeck et
al. (1998c), the cited reference.  Moreover, the values in this table are 2 or 3 orders of magnitude
higher than those given by Norbeck or those in the HAD's unsourced Table 2-18.

Table 2-25. The values given for Rochester and Washington presumably belong in the [ig/m3 column
rather than the % column. The footnote symbols for MATES II and 'not available' are reversed.

Table 2-26. Anaheim has a quarter-million people, and is identified as 'urban'  in Table 2-25.  It is
unclear that the distinction urban/nonurban has any meaning within the LA basin.

Figure 2-33. The Schauer et al. (1999) study employed TOT rather than TOR for the EC
measurement. Rogge et al. (1993) employed a method different from the DRI TOR used for Zielinska
et al. (1998) and Norbeck et al. (1998).

Figure 2-34. (A) The plotted values (y-axis) give EC content as percent of total fine particulate
matter, not as percent of total carbon. For example, Schauer et al. (1999) report EC as 30.8% of fine
particle mass and OC as 19.7% of fine particle mass, for an EC/TC ratio of 61%. (B) I can't find in
Zielinska et al. (1998) the 33% EC content plotted for engine model year 1990. (C) All four data
references have coauthors, not just Zielinska.

TOR vs TOT

       Two different thermal-optical procedures are widely used to determine the fraction of a
sample's total carbon present in the reduced, or "elemental," form. These are variously identified as the
TOT/Sunset Labs/NIOSH and TOR/DRI/IMPROVE methods, and they are known to yield
substantially different EC/OC splits (Countess, 1990; Chow et al., 2000; Norris et al., 2000). Both
methods evolved from a common ancestor (Johnson et al., 1981)  over the course of a decade or more,
complicating the interpretation of trends and the integration of data from different studies, particularly
those from different years. This situation will likely continue, as the Agency has invested in two large
ambient monitoring networks (IMPROVE and PM2.5 speciation), the former employing the IMPROVE
method and the latter the NIOSH method.

       The distinction between SOF and thermal-optical determinations of OC is noted and discussed
in section 2.2.8.1.1, but the distinction between TOR and TOT determination of EC is overlooked in

                                           A-15

-------
section 2.2.8.1.3. Six lines in section 2.4.2.2 seem to be the only mention anywhere of this pervasive
source of ambiguity. As I note in my Chapter 9 comments, failure to distinguish between TOR and
TOT determinations of EC is an important source of uncertainty in estimates of ambient DPM
concentrations, which are central to this assessment.  Paul Solomon and other air monitoring people at
EPA are actively studying the disagreement between TOR and TOT, and should be consulted on this
subject.

EC as surrogate for DPM

       The discussion on page 2-57, lines 23-31, uses far too many significant figures. Neither the
64% cited for EC as a fraction  of DPM nor the 67% cited for the diesel share of EC differs in any
meaningful way from the fraction 2/3.  To claim that DPM = 1.04*EC rather than DPM j EC is absurd.
(Note that 'HC' is erroneously  substituted for 'EC' in line 31, and again in all three displayed equations
on page 2-58.) Over-interpretation of the approximation DPM j EC continues on page 2-58, and
concludes with the circular logic in lines 22-27, which overlook the fact that the surrogate calculations
are themselves based on the CMB apportionments supposed to validate.

References

J.C. Chow, J.G. Watson, D. Crow, D.H. Lowenthal, and T. Merrifield (2000) Comparison of
IMPROVE and NIOSH carbon measurements. Paper presented at PM2000, Charleston.

R.J. Countess (1990) Interlaboratory analyses of carbonaceous aerosol samples. Aerosol Science &
Technology 12,  114-121.

R.L. Johnson, J.J. Shah, R.A. Gary, and J.J. Huntzicker (1981) An  automated thermal-optical method
for the analysis of carbonaceous aerosol. In Atmospheric Aerosol: Source/Air Quality
Relationships, E.S. Macias and P.K. Hopke editors, American Chemical Society, Washington.

G.A. Norris, M.E. Birch, M.P. Tolocka, C.W. Lewis, P.A. Solomon, and J.B. Homolya (2000)
Comparison of particulate organic and elemental carbon measurements made with the IMPROVE and
NIOSH Method 5040 protocols. In existing HAD draft references.

Chapter 9

       This chapter has clearly benefitted from the added round of editing and revision.  I particularly
like its explanations of risk-assessment concepts and Agency guidelines. At the end of the day,
however, I find that I still don't understand the larger question of why the Agency needs this assessment
of diesel exhaust as  an air pollutant that is distinct from its chief constituents, fine particulate matter and
nitrogen oxides.  Is  the need a statutory one?  A consequence of overlapping regulatory paradigms? A
response to a singular practical issue?

                                           A-16

-------
       Given the PM2 5 NAAQS, why should we worry specifically about diesel PM? This chapter
notes as one reason that somewhat different constellations of health effects have been suggested for
PM2 5 and DPM. Another reason is that some components of PM2 5, such as silt particles and sulfuric
acid droplets, are chemically quite dissimilar to diesel exhaust particles.  But one can grant that it makes
sense to distinguish silt particles from engine exhaust and still question the need to distinguish between
the compression- and spark-ignition contributions to engine exhaust. Are gasoline and diesel exhaust
qualitatively so distinct in terms of particle composition?

       Diesel PM is distinctively rich in EC; the chapter gives a range of 50%-75% for the EC fraction
of DPM mass.  However recent measurements also find substantial EC fractions in gasoline exhaust.
Norbeck et al. (1998) found EC to contribute over 40% of PM emissions averaging 7 mg/mi from nine
1991-97 light duty gasoline vehicles  in southern California. Similarly Watson et al. (1998) found EC to
contribute over 40% of PM emissions from light duty gasoline vehicles during cold start operation in
Denver. Diesel exhaust also contains benzene, PAHs, and other carcinogens, but again so does
gasoline exhaust. Indeed, gasoline exhaust contains substantially higher proportions of the heavier
PAHs such as BaP (e.g. Watson et al.,  1998).

       One could argue that diesel and gasoline exhaust are different and variable mixtures containing
EC and a suite of organic toxicants in common.  Given the document's characterization of risk in terms
of orders of magnitude, are several-fold differences in EC content or larger differences in the relative
proportions of different individual PAHs actually significant?

Ambient exposures to DPM

       The document often interprets EC as a marker for diesel exhaust, but large ambiguities in its
measurement continue to confound the Agency and the scientific community.  There are currently two
different analytic methods for distinguishing EC from OC, variously identified as the TOT/Sunset
Labs/NlOSH and TOR/DRI/IMPROVE methods, and they are known to yield substantially different
EC/OC splits (Chow et al., 2000). (Note in Figure 2-34, for example, that the sole post-1990 DPM
sample showing less than 50% EC (Schauer et al., 1999) is also the  only one analyzed by TOT rather
than TOR.* ) Both methods evolved from a common ancestor (Johnson et al., 1981) over the course
of two decades, complicating the interpretation of trends and the integration of data from different
studies.  This situation will likely  continue, as the Agency has invested in two large ambient monitoring
networks (IMPROVE and PM2  5 speciation), one employing each method. It presents an overarching
problem for the Agency, and this chapter of this document presents a good opportunity to highlight its
generic importance.

       Additional  ambiguity in reports of EC as a fraction of total carbon is introduced by the OC
measurement, which is sensitive  to sampling conditions.  Schauer et al. (1999), for example, found 35%
less OC in their DPM when they sampled behind an XAD denuder to remove organics in the gas
phase.

                                           A-17

-------
*  [The caption of Figure 2-33, which claims that all studies employed TOR, is incorrect (Schauer et
al., 1999, page 1579).  The vertical axis of Figure 2-34 is also incorrectly labeled: the plotted EC
values represent % of total particle mass, not % of total carbon.]
       This document (and individual investigators) sometimes combine EC values from different data
sources without accounting for their origin in different measurement methods. Given the similarity of EC
contents in some gasoline PM to those in DPM, such combining of inconsistent source and ambient EC
measurements has the potential to yield sizeable errors in estimates of DPM exposures. I would
accordingly direct more emphasis on page 9-6 to the fact that all DPM concentrations in the middle
paragraph represent uncertain estimates rather than actual measurements.

Human evidence for carcinogenic effects

       This is a charged subject, in which a single line of 'spin' can squander the credibility purchased
with a paragraph of balanced discussion. There are a couple of points where the Agency seems to be
reaching.

9-11, 8+:  "Although some studies did not have information on smoking, confounding by
smoking is unlikely because the comparison populations were from the same socioeconomic
class."
       Is socioeconomic class such an accurate predictor of smoking? How fine a class structure are
we talking about here?  The previous draft (11/5/99) was more cautious on this point, concluding that
(7-81, 11+)".. a possibility remains that the statistical adjustment for smoking is not completely
effective, and residual confounding by smoking may persist to bias the measure of the diesel exhaust-
lung cancer association."

9-15, 8+:  "..examination of the available PM data has not resulted in the identification of a
cancer hazard for ambient PM, although there is some evidence indicating a possible
association between ambient PM and lung cancer. "
       'Some evidence' consists of three familiar studies (pages 7-2,3,4).  The first is Six Cities, which
reported a non-significant RR = 1.37 for lung cancer.  The HEI reanalysis (Krewski et al., 2000) found
this association, unlike those for all-cause and cardiopulmonary mortality, to disappear when he
accounted for occupational exposure. The second is the ACS study, which reported a significant RR =
1.36 for sulfates, but no association with fine particles. It takes considerable effort to construe these
results as linking cancer with the carbonaceous DPM that is  elsewhere the focus of this document.  The
final exhibit is the AH SMOG study, which I previously understood to have found no significant
associations between pollution and mortality. I now learn that recent analyses show a cigarette-like RR
= 23 for lung cancer in nonsmokers from PM10; unfortunately, the citations given for this stunning finding
are missing from the list of references. Please!  One can as reasonably assert that there is 'some
evidence'  indicating a possible association between air freshener and lung cancer.
                                            A-18

-------
Relationship between risks from PPM and ambient PM2;

       I appreciate the Agency's attempt to accommodate the Committee's previous recommendation
that it acknowledge the connection between DPM and PM2.5. After reading sections 9.5.1.2-4 I also
appreciate its previous lack of enthusiasm for the task, as the resulting discussion has a distinctly Alice-
in-Wonderland flavor. Risk analysis is the Agency's mission, however, and the Agency should be able
to articulate its logic more persuasively than it has in the following passages.

9-18, 18+:  "If one assumes that the adverse health effects of ambient fine particles are due
entirely to DPM, ... the upper-limit for DE would be 15 lg/m3. "

       But if DPM is "typically in the range of 10%" of PM2.5 mass (line 26), PM2.5 epidemiology then
implies that 15 lg/m3 DPM is associated with the (rather severe) health risks observed at 150 lg/m3
PM2 5. That is, if DPM is "exceptionally toxic" then we need DPM < 1.5 lg/m3 to obtain the protection
associated with the PM2.5 NAAQS.

9-18,26+:  "If one assumes that DPM is as toxic as other constituents of ambient PM2,s, then
ambient concentration to [sic] DPM needs to be below the range of 1.5 to 5 lg/m3 ... to achieve
the same protection for the annual average standard for ambient fine particles of 15 lg/m3. "
       By this same logic, if 1.5 lg/m3 is an acceptable lifetime exposure level for DPM, then it is not
an acceptable level for either on-road DPM or off-road DPM.  And if, say, 1 lg/m3 is an acceptable
lifetime exposure level for the on-road portion, then it is not an acceptable level for diesel busses. And
if.... etc. Alternatively, if back-yard barbeque smoke (the Chamber of Commerce's favorite pollutant)
is as potent as DPM and one-tenth as abundant in urban air, then back-yard barbeque smoke needs to
be below 0.15 lg/m3 to achieve the same protection.  Does the Agency really want to place itself in the
position of having to choose between barbeques and public health?

9-19, 5+: "This congruence of independent methods should also increase overall confidence in
these estimates .."
       Two of the three methods are hardly independent, resting directly on the PM2 5 NAAQS.
Their agreement (to within an order of magnitude) is a simple consequence of the agreement between
ambient DPM and PM2 5 concentrations (to within an order of magnitude).

Miscellaneous details and typos

9-2,27+: Hasn't 'microns' become politically incorrect? Aren't we now supposed to say
'micrometers'?

9-3, 7: It's  a fact, not just an expectation, that 'Some geographic areas have a higher percentage of
DPMinPM2.5.' Period.
                                           A-19

-------
9-3, 9: The 1982 estimates cited here are all for different sites in the same urban California region,
greater Los Angeles.

9-3, 18: How can 'some DPM organic constituents' be higher than the upper limit of the range for the
'organic carbon portion of DPM'?

9-13, 22: I think the author means to say that the relative importance of adsorbed organics may
increase at lower exposure levels.

9-17,25: The NOAEL is 0.14 mffigrams per cubic meter, not micrograms.

9-19, 18: According to the destinction drawn at the top of page 9-13,1 think the author means
'mode(s)' rather than 'mechanism(s)' here.

9-20, 12: 'DE' should modify 'exposure' rather than 'workers'.

Additional references

R.L. Johnson, J.J. Shah, R.A. Gary, and J.J. Huntzicker (1981) An automated thermal-optical method
for the analysis of carbonaceous aerosol.  In Atmospheric Aerosol: Source/Air Quality Relationships.
E.S. Macias and P.K. Hopke editors, American Chemical Society, Washington.

D. Krewski, R.T. Burnett, M.S. Goldberg, K. Hoover, J. Siemiatycki, M. Jerrett, M. Abrahamowicz,
and W.H. White (2000) Reanalysis of the Harvard Six Cities Study and the American Cancer Society
Study of particulate air pollution and Mortality.  Health Effects Institute,  Cambridge.

J.C. Chow, J.G. Watson, D. Crow, D.H. Lowenthal, and T. Merrifield  (2000) Comparison of
IMPROVE and N1OSH carbon measurements. Paper presented at PM2000, Charleston.
                                           A-20

-------

-------
                                      Arthur Upton. IMP

Chapter 4

        The revisions that have been made in Chapter 4 respond effectively to the following criticisms
which were leveled against the preceding draft: 1) this version of the report contains an appropriately
expanded discussion of current information on the mutagenicity of particles having little or no organic
content, including evidence for the involvement of reactive oxygen species as demonstrated by the work
of Driscoll and others; 2) this version also includes a discussion of the mutagenicity of oxygen radicals,
which are thought to contribute to the tumoigenic effects of chronic high-level exposure to diesel
exhaust particles on the rat lung; and 3) this version cites the work of Wallace, Keane, et al. at NIOSH
on the mutagenic activity of whole DPM, as was recommended.

        This version of the report does not adequately respond, however, to the criticism that it should
place the high does used in mutagenicity assays in context relative to the does likely to occur from
inhalation.

        As concerns editorial issues, the statement on page 4-4, line 24 that "both adduct and
mutagenicity were highest among the 16 most exposed" is confusing in that it appears to contradict the
preceding sentence, which states that hrpt mutant frequencies in workers did not differ from those in
controls.

        Pages 6-2 to 6-6: in addition to the references cited, the document should cite the newly
published article by Daniels, Dominici, Samet, and Zeger (Am. J. Epidemiol. 152: 397-406,2000),
which reports  that the data for the 20 largest U.S. cities are best fitted by a linear-nonthreshold
relationship between PM10 and daily mortality from all causes and from cardiorespiratory causes; and
which concludes that "linear models without a threshold are appropriate for assessing the effects of
particulate air pollution even at current levels".

        Pages 6-6 to 6-7 and page 6-21: if daily mortality may vary as a linear-nonthreshold function of
particulate air pollution, ought it not  to be included among the non-cancer health risks considered in this
chapter?

        Page 7-114, line 25: the statement here  and elsewhere (e.g., page 8-3, line 13, and page 9-15,
line 1) in the document that "the mode of action for DE-induced lung tumors in rats is  sufficiently
understood" is not scientifically justified. I would suggest another wording; e.g.,  "Although the mode of
action by which DE may pose a risk of lung cancer for humans is not fully known, the tumor-inducing
action of DE on the rat lung appears to depend on particle overloading of the lung and is therefore
judged to be irrelevant for purposes of assessing the risks of ambient levels of DE, as discussed in
Section 7.4".
                                             A-21

-------
       Page 8-14, line 7: since DE is not known with certainty to be a carcinogen for humans, the
statement should be reworded to read: "...pose a lifetime cancer risk ranging from a lower limit of zero
to an upper limit of 10-5 to 10-3".

       Page 9-9, Section 9.4.2: shouldn't daily mortality be included here, or at least mentioned?

       Page 9-14, line 13: "its" should be changed to "the relevant".

       Page 9-14, line 15: "appears to" should be changed to "may".
                                             A-22

-------
                                      Sverre Vedal. MD

Chapter 7. Carcinogenicity.

Hill criteria
       The application of the Hill criteria (pp. 65-68) for assessing the likelihood of causation is
improved in this version of the document. It should be appreciated that modem views on the usefulness
of the Hill criteria, such as those expressed by Rothman as referenced in the document, have
significantly limited the usefulness of the criteria for this purpose. In brief, the general  point of these
views is that none of the Hill criteria, except for temporality, need to be satisfied by an association that
is, in fact, a causal association. That is, none are necessary.  Specifically, neither a strong association,
specificity of effect, dose-response, plausibility nor consistency are required. One wonders about the
utility of applying the Hill criteria when so little is gained by their use.  Regarding the strength of an
association, in the document it is noted that the  strength of the association is "modest  to weak".
Although some studies observe relative risks of 2.0 or greater, the overall estimate across studies is
approximately 1.3 to 1.4, which I would argue is a weak association.  The Hill criteria that are probably
met are the consistency criterion and the biological plausibility criterion.  It is not clear that the dose-
response criterion is met. Specificity is partially met, although studies have either not addressed other
effects or have lacked power to do so. The  temporality criterion is assumed and not  tested. The
conclusion based on the discussion of the Hill criteria, and as repeated in the Weight of Evidence
section (p.  110), seems forced. I would recommend basing arguments for causality, when using the
epidemiological data, not on the Hill criteria, but rather on those characteristics of studies that determine
validity.

Latency
       Latency was discussed under the Hill heading  of "temporality", which is only partly stretching
the intent of this criterion. Nevertheless, latency is seldom directly addressed for any  of the studies
except in Table 7-1 where it is noted that "no latency  analyses" were performed for some of the
studies. Given the relatively recent use of diesel with  respect to end of the follow-up period in most of
the epidemiological studies, there is a sense that latency is short.  Explicit assessment  of latency for each
study should have been attempted in the document, noting those studies where it was not possible to
address the issue because of lack of information.  At issue here, obviously, is whether the observed
effects with the present latency periods are implausible, and  therefore whether the effects observed are
due to uncontrolled confounding.

       In order for investigators to address  latency in a more satisfactory manner, it would be
interesting to perform analyses in which latency is explicitly examined. For example, one could restrict
the study to include only subjects with  latency of less than 10 years, in which case if similar effects were
observed, one would argue that diesel  exposure was not likely responsible.  This would also weaken
the Hill "temporality" criterion.  Other strata of latency could also be  examined, realizing that latency,
                                             A-23

-------
age, and duration of exposure would likely be correlated.  Assessing duration of exposure, as a
measure of "dose", is not the same as assessing latency
                                             A-24

-------
PM studies
       It is relevant to include descriptions of the observational studies of PM in this chapter, as was
done in this version of the document.  The PM "cohort" studies that included lung cancer mortality as an
outcome (the 6-Cities Study, the ACS Study, and the AHSMOG Study) resulted in some interesting,
but ultimately puzzling, findings.  In the 6-Cities Study, deaths due to lung cancer were increased in the
most polluted relative to the least polluted city, although the association was not statistically significant.
The number of lung cancer deaths was not reported.  In the ACS Study, lung cancer deaths increased
with increasing sulfate concentrations in men, but were not increased with increasing  PM2.5 in the subset
of cities with data on PM2 5.  Although the number of lung cancer deaths was not reported, they were
likely 300 times larger in the ACS Study than in the 6-Cities Study, based on the relative number of
deaths in the two studies.  In the AHSMOG  Study, with 20 female and 16 male lung cancer deaths in
this nonsmoking population, ozone, SO2 and PM)0 were associated with lung cancer deaths in men,
while SO2 and PM10 were associated in women.  At this time, the validity of the findings of these
studies with respect to lung cancer is questionable, as is therefore their relevance to diesel
carcinogenicity. A more critical review of these findings would be  welcomed. Nevertheless,  the final
statement (p.4 , line ) places these studies in  reasonable perspective.  However, it is not clear how this
reasonable statement follows from the description of the "cohort" studies. Also on p.4 (line 5), why is
the ACS finding with respect to PM2 5, which is a negative finding,  consistent with the findings of the 6-
Cities Study?

Minor comments and editorial issues:

page
2      In paragraph 2, why do the PM estimates represent an "upper limit" for estimates of DPM if
       DPM is only part of the PM mix?  That is, say if the rest of PM has no carcinogenic effects, it
       would be counterintuitive to conclude that any observed effects of PM would be an upper limit
       on effects  of DPM. I may be misinterpreting what  is meant, and if so, some clarification is
       needed (see also comments for chapter 6).

       Also in this paragraph, mention is made of evidence from the cohort studies that these provide
       evidence on chronic effects. I am not sure they provide any evidence on chronic vs. acute
       effects.

50     The second paragraph is reproduced verbatim on  p. 51. Although possible,  please review.

52     "tskas" line 20

54     The last sentence of this page needs rewording to  clarify (double negative, etc.)

60     The meaning of sentence beginning line 4 is not  clear.
                                            A-25

-------
64     It is not clear how insufficient latency would result in increased RRs (Iine33).

67     Do the Garshick and Steenland studies really have adequate latency (line 6)?  In chapter 8 it is
       noted that adequacy of the latency period is a problem for both.

99     In the last paragraph, it is argued that human lungs (and mouse lungs?) could be more sensitive
       to the carcinogenic effects of PAHs than rat lungs, since only in human and mouse lungs do
       many lung cancers exhibit mutations of p53 and K-ras genes.  However, the high dose DPE
       inhalation studies show that in fact rats are relatively sensitive and mice relatively insensitive.
       This paragraph needs clarification.

10     1 Section 7.4.2, as noted, refers only to particle overload conditions. It should be emphasized
       that the inflammatory mechanisms are largely relevant only in this setting.

106   Line 7 "increasing" is confusing, since the effects occur with decreasing particle size.

110   My reading of the application of the Hill criteria section beginning on p. 65 is that they add little
       to a conviction for the causality of diesel exhaust.

112   Is "limited" on line 23  too strong an adjective?

133   Table  7-4 needs a source attribution (Dasenbrock, 1996).

       In the Weight of Evidence section, reference is again made to the chronic exposure rat data,
which should not be used to argue for causality.

Chapter 8. Dose-response assessment: carcinogenic effects.

       This chapter is generally well done.  The estimates of lifetime risk based on the human
observational data, of course, assume that the estimated effects are valid, that is, are unconfounded.  At
the very least,  a qualifier to this discussion should be added, since otherwise we would be concluding
that DE is a definite carcinogen, and we are merely trying to quantify the dose-response relationship. It
seems unlikely that cigarette smoking is the unifying confounder. However, the relatively short latency
remains worrisome, suggesting either that the effects would be stronger with an adequate latency
period, or that the effects are confounded and have no relationship to latency. This latter issue raises
the concern that relative risks are in fact 1.0, which would clearly invalidate the range of estimated
effects that was proposed. This is related to the issue that exposure in these studies is not "diesel
exhaust"  but rather employment category.

       Based on these reservations regarding the validity of the  effect estimates in the occupational
cohorts, and even legitimate concerns that the estimates may in fact be 1.0,1 would not be in favor of a

                                             A-26

-------
 quantification of risk that ranges from 1 x 10 "5 to 1 x 10 "3.  If such a quantification of risk is deemed
 important, I would recommend using the range zero to 1 x 10 "3. Future work should include addressing
 effects in cohorts with clearly inadequate latency to determine if estimated effects remain elevated, and
 hence not attributable to diesel exposure.  If effects are not elevated for subjects with clearly inadequate
 latency, and adequate latency has been shown for other studies, then I would be much more
 comfortable with the validity of the occupational cohort effect estimates.

        It was my understanding that another significant activity that is underway (first paragraph, p.l 1)
 is an extension of the period of follow-up for the Garshick cohort study, an activity which has the
 potential of addressing the latency issue.

        With respect to cigarette smoking, on page 4 (line 16) it is noted that adequate smoking
 adjustment could have significant impacts on the estimates of effect, which is true. Nevertheless, when
 this has in fact been done, the impacts are small. Also, it is stated (line 21) that traditional statistical
 analyses are unable to adjust for the possibility that smokers may be more susceptible to DE effects (a
 notion regarding susceptibility, by the way, which if maintained should be justified somewhat better).
 However, this can be handled easily in a logistic regression, and in many other types of analysis, through
 introduction of interaction terms (between smoking and DE exposure) in the models, assuming that
 smoking status is known. This point therefore needs clarification.

       Also regarding smoking, the sentence beginning on line 22 notes that control for smoking is a
 greater problem for case-control than for cohort studies because most lung cancer cases are also
 smokers.  However, the same is true for lung cancers detected in a  cohort study. That is  therefore not
 the reason that smoking might be more difficult to control in a case-control study. A prospective cohort
 study is generally preferable to a case-control study since good smoking data, and data on other
 potential confounders, are easier to obtain, and the adequacy of the control group is not an issue. The
 point regarding data on confounders cannot be maintained for a retrospective cohort study such as
 those that address the DE issue.

 Minor comments and editorial issues:

page
 2      Should be "causal" (line 27)

       The use of the high concentration animal data to motivate causation should be dropped.
7Add "and risk of lung cancer." to linelS to follow "exposure...".

Chapter 5. Non-cancer health effects of diesel exhaust.

page
                                             A-27

-------
2      The claim (line 2) that there have been "no well-controlled chamber" studies is out of date.  The
       two Salvi papers referenced on p.7, and now the Nightingale study reported in Am J Respir
       Crit Care Med in 2000, are examples of such studies.

6      The description of the "roadway" studies (beginning Iine23) is one-sided in the sense that no
       negative studies are included. Some studies show no association with asthma prevalence,
       although there may be associations with asthma exacerbation.

64     The claim beginning on line 6 concluding that the "principal noncancerous health hazard to
       humans posed by exposure to DE is a structural or functional injury to the lung" is debatable
       given the, in my opinion, more compelling data in both humans and animals on the effects of DE
       on allergic responses.

70     Regarding the Conclusions, in keeping with my observation above on allergic effects, these
       effects seem more significant than either of those described in the other two conclusions:  that
       noting increases in symptoms (an inconsistent finding) and that on chronic effects in humans and
       animals.  Also, given the weight placed on both inflammation and fibrosis resulting from chronic
       exposures in the RfC in Chapter 6, the third conclusion (p.71) should include  fibrosis as an
       important effect in the animal studies.

       Reference to the Ishinishi (1988) study should be made in a consistent fashion in this chapter
and chapter 6. This reference is critical to the NOAEL in chapter 6, but it is referred  to as the
"Research Committee for HERP  Studies" (1988) in chapter 5 (Table  5-6, for example).

Minor comments and editorial issues:

page.
65     line 15, "incidents" should be "incidence"

66     The two sentences beginning on lineSO, taken together, are confusing. The first notes that
       short-term DPM exposures have no apparent health effects, whereas the second details effects
       on the lung at "lower levels of DE". Which is it? Maybe the meaning is that these latter
       responses are not really health effects. Nevertheless, clarification is needed.

Chapter 6. Quantitative approaches to estimating human non-cancer health risks of diesel
exhaust.

       As noted in my comments on chapter 5, the animal  data on  effects of chronic exposure are not
as compelling as the human and animal data on allergic responses. The use of the animal data on effects
of chronic exposures for calculating an RfC does not therefore make use of the best data for estimating
a reference concentration for DE.  The motivation for not using these allergy data is that they "are

                                            A-28

-------
considered inadequate for dose-response evaluation" (p.9 line 16).  But, as noted below, absence of
dose-response information does not preclude a study from providing useful information for determining
a reference concentration.

       With respect to the chronic animal studies, ignoring studies that do not provide information on
dose-response (for example, p. 10, line 31) ignores valuable information in deriving an RfC. Even a
study that made use of only one DPM concentration, since effects at that concentration are either
consistent or inconsistent with the observations from studies that included a dose-response evaluation,
would seem to be relevant to the RfC derivation. For example, should a study that observed no effects
at the one concentration  evaluated, if that concentration, say, were above the NOAEL observed in
studies assessing dose-response, not be considered relevant?

       It should be emphasized that we have a relatively high degree of confidence in the RfC, given
that what we are highly confident of with an RfC is that the effects of concern do not occur at
concentrations lower than the RfC.  The EPA has gone to great pains in the incorporation of the
uncertainty factors to ensure this type of confidence. This is not to suggest that effects are likely to
occur at concentrations even several fold higher than the RfC, since there is no upper bound. That is,
the RfC could be a gross underestimate, but higher RfCs would not give us the same degree of
confidence that effects do not occur at concentrations lower than these RfCs.

Minor comments and editorial issues:
cage

1      The reference to  "upper limit" (line 30) here is confusing. The point being made, I believe, is
       that observed PM effects are the upper limit of effects attributable to DPM.  As noted, DPM
       comprises only a fraction of PM. If DPM is more toxic than other components of PM
       (unlikely, but possible), the observed effects of PM would not represent an upper limit of DPM
       effects.  Any point intended here needs to be clarified.

4      Line 22. The effect estimates in the 6-Cities study are  weak, not large. Estimated public health
       impact should not be confused with strength of association.

16     The point regarding "congruence of estimates" is ingenuous, but is now moot based on the
       discussions at the CASAC meeting.

       The reference to  the Ishinishi study (1988) should be made in a consistent manner in this
chapter and in chapter 5.

Chapter 9. Characterization of potential human health effects of diesel exhaust: hazard and
dose-response assessments.
                                            A-29

-------
page
11     Again, the use of the Hill criteria argument (line 12) is not a very compelling one regarding a
       causal association in the case of DE.

12     The use of the rat data here (line 13) is puzzling.  The data are not useful for dose-response
       analysis because of the overload argument. But because this mechanism is felt, with
       justification, not to play a role in possible cancer pathogenesis in humans, it seems also that the
       data should not be used to support the presence of a cancer hazard in humans.

19     Line 3.  The point about "congruence of estimates" here  (and p. 22, Iine28) is ingenuous. It is
       difficult to be reassured about the reasonableness of the RfC based on the PM2.5 annual
       standard when the RfC is heavily influenced by an uncertainty factor of 10 (one order of
       magnitude). Further, the DPM component of the PM2.5 standard is approximately 10%, or 1.5
       ug/m3, rather than 15 ug/m3.

       Line 16. The rat data are used again.

Minor comments and editorial issues:

page

5      DPM should be PM, I believe.

14     Line 33. I would add "...and other potential confounders..." after "...the effects of smoking...".
       Similarly for chapter 1, p.4, line 22.

16     Line 28 (and p. 18, line 20):  see discussion on "upper bound" in chapter 6 comments
       (reference top. 1).

23     Line 30. It is unclear what "this assessment" refers to. It seems to refer to the RfC, in which
case an uncertainty factor of 10 to account for susceptible subgroups has already been incorporated.
This therefore does not  assume that it applies to "average health adults". This point also applies to
chapter 1 (p.6, line 4).

Chapter 1.  Executive summary.

       See comments for chapter 9 referring to p. 6, line 4, and p. 4, line 22..
                                             A-30

-------
                                  David Diaz-Sanchez, PhD

General comments

       As with each new draft, I believe this is a further improvement from the previous one and is a better
representation of the health risks  of diesel. I believe that the more cautious approach taken in this draft
towards  making overstatements on both carcinogenic and non-cancer effects is warranted until more
quantitative studies are published.

Chapter 5

       Despite the limitations cited below I think this chapter is satisfactory with minor editing.

       The agency  should be commended in amplifying the number and range of articles on non-cancer
effects of diesel exhaust. Indeed I believe that they have been overzealous in achieving this aim. Thus, while
the study by Brunekreef et al. (2000) cited on Page 5-7, 5 is interesting and potentially very important, it
should be noted that it has  yet to be peer-reviewed. Similarly, Madden et al., is listed under the references
as "submitted".

       Despite the  breadth of the articles cited there seems again to be little depth in understanding the
significance of these results. For example: what is the  significance of a change in IgE or goblet cell
hyperplasia or mast cell influx or cytokine changes in animal models? Without a statement that these are
key changes and markers of asthma, the reader is left with the impression that diesel induces a variety of
immunological changes but has no idea what relevance this has to health effects.

       Unfortunately, this  lack ofunderstanding leads to incorrect conclusions such as that on 5-45, 30
that "diesel exhaust has minimal effects on the immune status of rats and guinea pigs" while it does have an
effect in mice. This implies that there is inter-species variation, however, as stated in the comments to the
last draft  report,  the lack of response seen in the studies done on rats and guinea pigs  (Dziedzic  1981,
Mentnech 1984, Bice 1985) are to be expected since they were only performed in the absence of an
allergen unlike those done on mice where an allergen (ovalbumin, house dust mite, pollen etc.) was used.

       The organization of this chapter appears arbitrary. For example why is Yang et al., (1997) placed
under acute exposure when this is an in vitro assay? Similarly, why are the  studies by Terada et al (1997
and 1999) performed on human cells in the Laboratory Animal section (5.1.2.3.6) and  not the "Human
cell culture studies" section(5.1.1.1.4)? Why is Takano et al (1997) under acute exposure, when all other
instillation experiments are  under  5.1.2.3.6? Why are studies  by Rudell et al., in a separate section  than
those by Salvi et al., when these exposures were performed by the same group under the same conditions
and measured similar or related outcomes.
                                            A-31

-------
       5-57,5  "notably pyrene" implies that pyrene is more active than other constituents in the organic
matter, this is probably an over-interpretation. Pyrene is used in the studies cited (Suzuki et al. 1993,
1996.) as a model chemical and is not compared to others. It should be noted in vitro studies suggest that
phenanthrene, benzo (a) pyrene and TCDD can affect immunogenicity and allergenicity. It may also be of
relevance to note that other combustion material such as fly ash and second-hand smoke has been shown
to have similar adjuvant effects as DPM in animal models.

       5-40, 23 should read IgGl not Igl

Chapter 6

       Despite the fact that almost 2/3 of articles published in the last 10 years on the non-cancer effects
of diesel have dealt with immunological changes, and that acute exposures maybe of more relevance than
lifetime exposures, I agree with the authors that  the lack of dose-response information makes taking a
quantitative approach using these criterion premature. I applaud the authors for including a guidance value
for DPM by treating it as a subset oftotalPM2.5. This is a necessity given the health outcome studies and
previous statements in the documents such as 5-63, 5 "diesel exhaust toxicity results from a mechanism that
is analogous to that of other relatively inert particles".  The inclusion of Appendices B and C are most
useful.

       6-6, 22 states that DPM is typically in the range of 10%, while the executive summary (1-2, 7)
gives the figure as 6%. There should be consistency throughout the document.

       6-14,12 suggests that children have a greater susceptibility to DPM but 5-59,21 states that there
is no  evidence that the youth of an individual enhances the risk. A similar argument and contradiction is
made for pre-existing conditions such as emphysema.

       6-6,23 states that large numbers of ultrafine particles may make DPM disproportionately toxic. If
the authors are going to make such a sweeping statement they should provide references or refer back to
other parts of the document.
                                             A-32

-------
                                  Eric Garshick, MD, MOH

Chapter 1: Executive Summary

       General comments: This chapter needs to be rewritten to reflect the changes that will be made
in the rest of the document in response to the comments of CASAC. In order to prevent confusion
regarding what is known about diesel exhaust inhalation as a potential hazard (a great deal of
information) with the true risk of an adverse health effect at environmental levels (very little information),
these terms need to be carefully defined, and the definitions should be repeated in he executive
summary and in key chapters. Differentiation between ambient cancer hazard and defining ambient
cancer risk is made on the top of page 1-5, but I would consider noting this earlier in this section.

       Page 1-4, line 15-17: One might extend the title of this section to "Carcinogenic Effects -
Hazard Identification".

       Consider ending the sentence after "inhalation" with a period instead of saying "at any exposure
condition" since the phrase "at any exposure condition" is qualified in the next paragraphs. I am
concerned that the uncertainties regarding the phrase  "at any exposure condition" is not accurately
conveyed. It is worth noting that accurate exposure-response information, particularly at low levels of
exposure is not available.

       Consider these sentences: " The studies on which this is based comes from epidemiologic
studies of workers with occupational exposure to diesel exhaust. Although in some cases there is
overlap between occupational and environmental exposures, it is not possible to establish a link
between a specific exposure level and lung cancer risk with confidence and therefore determine the
magnitude of the risk that occurs at specific environmental levels'. The point of this paragraph is to state
that there is uncertainty about the level of risk that occurs at a given level of exposure. Although it is
agency policy to consider that there is a risk at environmental levels, it is not possible to state the
magnitude of risk that occurs.

       Page 1-4, line 33-34: Consider this change: "hazard extends to ambient environmental levels"
to "some hazard extends to environmental levels". I do not think it is necessary to include the word
ambient and environmental in the same sentence.

       Page 1-4, line 36: change "it is prudent public health policy to presume a cancer hazard for DE
at any exposure" to "it is prudent public health policy to presume a cancer hazard for DE at low levels
of exposure, although it is not possible to determine the magnitude of the risk." I am concerned the
sentences as written will be taken out of context.

       Page 1-5, lines  12-19: consider adding to this section "The range of the actual risk may
approach and include zero"

                                            A-33

-------

-------
       Page 1-5, Sources of Uncertainties: This is a good section and can be referred to at the
beginning of the Executive Summary so that persons do not forget the limitations inherent in the
interpretation of these data.

Chapter 3: Dosimetry Of Diesel Participate Matter

       Page 3-20, line  18: Comments on the section entitled "Relevance to Humans": Consider
expanding this paragraph to describe the relevance of the rat model to the occurrence of non-cancer
health effects in humans. This seems important because the rat model was abandoned for both hazard
identification and formal risk assessment for lung cancer, but is used to estimate an RiC value for non-
cancer health effects.

ChapterS: Noncancer Health Effects

       This chapter is much improved, but there are still some points that need to be clarified. It is
important to be precise when relating the relevance of experimental data to potential mechanisms of
human disease. At times results are generalized and their relevance is overstated. Some the health
effects attributable to PM alone Add: PM document to support some of non-cancer health effects in a
qualitative sense.

       Page 5-4, lines 23-24: The results summarized by the sentence "Miners with a history of
smoking had an increased number of decrements over the shift than non-smokers did" needs to be
restated since it is not clear what an "increased number of decrements" refers to. Presumably it refers to
a greater decrement in pulmonary function over a work shift observed in smokers compared to non-
smokers.

       Page 5-6, line 35: Presumably  truck traffic counts were associated with a decrement in lung
function in these children. A stronger point can be made, rather than saying just "lung function was
associated with truck traffic  density". Relate roadway studies to potential diesel exposure.

       Page 5-7, line 1: A sentence could be added to describe what is meant by "black smoke" in
this study and how it relates  to diesel exhaust.

       Page 5-12, lines 4-6: These lines state that DPM has the potential to cause inflammatory and
immunological responses typical of asthma. Taken at face value, this statement implies that DPM can
cause clinical asthma in humans, which is a syndrome characterized by airway inflammation, bronchial
hyperreactivity, and recurrent episodes of wheezing, breathlessness, chest tightness, and  coughing, as
well as airway  inflammation. Evidence has not been produced in humans that diesel exhaust exposure
results in asthma, and I  am concerned that these sentences may be taken out of context.  It seems more
appropriate to be more circumspect regarding the evidence for a link between the inhalation of DPM
and asthma in humans.

                                           A-35

-------
       The controlled inhalation experiments in humans summarized in the chapter document increased
numbers of polymorphonuclear lymphocytes in BAL samples, the expression of the genes than result in
the production of IL-8 and IL-5 (inflammatory mediators), and bronchial biopsies with an increase in
neutrophils, mast cells, and lymphocytes. Nasal lavage studies reveal that DPM can act as an adjuvant
to stimulate specific IgE responses and can result in the production of various inflammatory mediators in
the nasal lavage cells. Therefore, a more precise way to summarize these data obtained from studies in
humans and from using human cells may be that DPM can cause an increase in markers of inflammation
in the nose and airway, and that some of these markers of inflammation are also observed in asthma.

       The study that described the 3 cases of asthma following high dose exposure to diesel exhaust
has been taken out of the section on "Immunological Effects" but would be worth noting in the section
called "Short-term Exposures" since the mechanism was due to the short term inhalation of high
concentrations of exhaust rather than an immunologic mechanism. The study is still listed in Table 5-1
but is not described in the text.

       Page 5-12, line 8: The words "increase their effectiveness" appears  at the end of the sentence.
It is not clear what "effectiveness" means here. Presumably it means the effects of DPM in enhancing
the IgE response, but in this left to the reader's imagination.

       Page 5-61, line 10-15: This sentence regarding susceptibility to inhaled particles is not
supported by data that is presented in this document.  In fact, two experimental studies did not support
the increased susceptibility of developing lungs or emphysema studied in rat  lungs. The results of these
experiments are dismissed as irrelevant to human exposure.

       Page 5-66, lines 2-3: I'm not sure what the phase "doubling of a minor restrictive airway
disease" means. This needs to be explained since airway disease does not typically result in a restrictive
ventilatory defect.

       Page 5-66, lines 20-22: This should be restated to say that DPM appears to have the potential
to induce airway inflammation in humans without disease, and in one exposure study, peripheral blood
changes were noted. The sentence as written does not seem to accurately capture the experimental
evidence noted from human studies. How is pulmonary inflammation different from airway inflammation
here?

Chapter 6: Quantitative Approaches to Estimating Human Noncancer Health Risks of Diesel
Exhaust

       This chapter is more clearly written than in the past and the discussion about the PM standards
and health effects enhances the rationale for the development of the RfC. The message that I take from
the chapter as it was originally written is that a range of possible RfC values are suggested, ranging from
15 ug/m3 based  on the PM standard, a value of 14 ug/m3 based on animal data extrapolated to humans,

                                           A-36

-------
and a possible value of 1.5 ug/m3 to 5 ug/m3. However, it is not possible to endorse the lower values
over the higher values because of the uncertainty of the data used to obtain these values. It also would
be reasonable to talk about the relevance of the rat data for the derivation of the RfC to contrast this
decision with the same rat data that has been rejected to for the assessment of carcinogenicity. I agree
that data based on allergy may be the basis of future standards.

        Page 6-14, lines 8-11: This sentence regarding persons who may be more susceptible to diesel
exhaust seems speculative, and should be justified.

        Line 25, page 6-16: It is not clear what is meant by "The lexicological database for DE is
relatively complete".  There seems to be much uncertainty regarding many questions.

        It is reasonable to add the uncertainty factor of 3 to the calculation of the RfC as discussed at
the meeting.

Chapter 7:  Carcinogenicity of Diesel Exhaust

        Overall, this chapter is more clearly written than previous versions and an improvement.
Enhanced by discussion of studies of PM and lung cancer. Agree with bottom line

        Page 7-3, lines 28-29: Relative risks quoted seem large here. The overall whereas the odds
ratios for the long haul drivers was 1.31 (95% CI=0.81-2.11), and for the short haul drivers was 1.27
(0.83-1.93). The long haul drivers drove mainly diesel trucks, whereas the short haul drivers drove
gas-powered trucks. The similarity in odds ratios and exposure levels between the short haul and long
haul drivers suggests that much of the driver's exposures come from the roadway. Other issue is one of
latency, and knowing for sure the types of truck driven and for how long, given that the deaths were
collected 1982-83. This doesn't negate observations that diesel particles have the potential to cause
lung cancer, but raises the issue that workers who work as professional drivers are exposed to particles
of a variety of combustion sources.

        Page 7-39, lines 9 and 10: The sentence "As far as qualitative risk assessment is concerned,
this study is still considered to be positive and strong" is confusing since the quantitative risk assessment
performed is not used in assessing risk in the document (due to limitations of the exposure data), and
the study has already been described in the previous section. This paper doesn't add any independent
information to the discussion of the previous paper.

        Page 7-63, line 7: A statement is made that occupational data (presumably job title) is a poor
surrogate for the  true underlying exposure distribution. A more precise statement would be that
variability in actual lifetime exposure to diesel exhaust in an occupational cohort may not be reflected in
differences  in job title, and there might be considerable variability in actual exposure despite similar job
titles.

                                             A-37

-------
       Page 7-63, line 8: A statement is made that death certificate information is inaccurate regarding
the diagnosis of lung cancer, whereas on the next page it is stated (appropriately) that lung cancer
diagnosis on the death certificate is generally accurate. This is contradictory.

       Page 7-67: Biologic gradient: The dose-response relationship between diesel exhaust and lung
cancer is uncertain since the biologic gradient has not been defined well in the literature.

       Page 7-110, first paragraph: A well  as the factors noted in this paragraph, an overall limitation
of the diesel-lung cancer literature is that there has not been a study conducted with workers with
documented long-term exposure to diesel exhaust and long-term follow-up.

       Page 7-113: The designation likely carcinogen is consistent with previous assessments, as is the
group Bl designation (Page 7-112). Further justification of how the "upper end" of the spectrum
designation differs from a "lower end" of the spectrum designation would add clarity for me in
understanding why the terminology for upper end is used.

       Page 7-113, line 4: The comment "likely to be carcinogenic at any exposure condition" is based
on EPA's science policy,  noted in the sections below this statement. However, taken out of this
context, it implies that we know for certain that there is a measurable lung cancer risk at all
environmental conditions. A better characterization might be "likely to be carcinogenic to humans by
inhalation, but with greater uncertainty about the magnitude of cancer risk at environmental and low-
level occupational conditions". To summarize, statements regarding risk at environmental levels should
be emphasize the uncertainty about what is known about actual risk.

       Page 7-114, line 9: The statement regarding environmental exposure resulting in a  cancer
hazard is repeated, but the qualifier "may" has been added, in contrast to the statement made on the
previous page. A reason for uncertainty can include the lack of a study conducted with workers with
documented long-term exposure to diesel exhaust and long-term follow-up.

       It was recommended at the meeting that the rat data regarding carcinogenicity be removed from
hazard identification because the mechanism of lung cancer in the rat (particle over load) is not
observed in humans. I pointed out to the committee that the rat data had been part of hazard
identification (but not risk assessment) since the document was first written approximately 10 years ago.
I agree with the consensus to mention this on page 7-113, but not to make it part of the formal hazard
assessment criteria.

Chapter 8: Dose-Response Assessment: Carcinogenic Effects

       Page 8-2,  lines 25-26: consider adding "at some levels of environmental exposure, although the
specific levels cannot be specified"  and "may pose a cancer hazard". The word hazard has an
administrative meaning, if unclear, it can be interpreted to mean risk.

                                             A-38

-------
        Page 8-5, line 13: Should read "between 10 to 20 years of work" in 1959.

        Page 8-10, lines 14-18: The methods used to obtain the cumulative exposure values are
controversial (exposure weighted using emission factors and assumptions about vehicle miles traveled in
the absence of a more comprehensive exposure assessment), and the listing of these cumulative
exposure levels is misleading if additional research proves them to be incorrect.

        Page 8-11: Section on cancer risk: I prefer a narrative discussion of risk, and then showing that
environmental levels have the potential to overlap at the lower range of occupational exposure, and
saying that this merits control. I do not think that these risk levels should be estimated given the
uncertainty about dose needed to result in cancer. In the absence of study of persons with years of
exposure and follow-up, with some idea of exposure, I can't justify these calculations. While these are
interesting, I think it remains speculative to do this.

        I believe that the risk calculations presented here grossly over simplifies the overall approach
needed to understand the relationship between dose of diesel and lung cancer. We do not know the
nature of the complex exposures experienced by workers in diesel exhaust exposed jobs. These
workers were not only exposed to diesel particles, but to other particles as well. We need to
understand the nature of what is captured in the job codes used in the epidemiologic studies that has
come to mean cancer risk. If the risk numbers were used, it would be complete to note that the range
of risk may approach zero as well.

Chapter 9: Hazard and Dose-Response

        This chapter needs to be rewritten to reflect changes made  in the rest of the document.

       Page 9-13, lines 17-18: Carcinogenic activity is related to the small sizes of the particles. Is this
known for sure in humans?

       Page 9-14: Add that a limitation of the diesel literature is the need to follow-up of a cohort with
well-characterized exposure over many years.

       Page 9-20 to 9-21:1 believe that these risk calculations as too uncertain, and favor substitution
of prose describing the overlap in occupational and environmental exposures, and the uncertainty
regarding the lifetime dose of diesel necessary for the development of cancer.

       Page 9-22, line 24: Would say "It is presumed  that low-level exposure may cause chronic
inflammation	:" Previously statements like this were phrased with greater uncertainty.
                                            A-39

-------
                                 Roger O. McClellan, DVM
                                                          *
A.     General Comments

       The document with several significant revisions will be acceptable for use in regulatory decision-
making. In general, the document is improved over earlier versions. However, with a preparation time that
now extends over more than a decade, preparation having started in the late  1980s, it is inevitable that
some portions of this document have become  out-dated as other portions have been updated and
improved. The key elements of the document concern the two major types of health effects attributed to
diesel exhaust exposure; a) cancer and b) non-cancer health effects principally related to effects on the
respiratory and cardiac systems.

       The document adequately reviews potential non-cancer hazards. This version of the document is
improved in that it provides a better linkage than in past revisions, to the substantial body of information
available on the health hazards of Particulate  Matter (PM)  exposure. This information has been reviewed,
and is currently being updated, as part of the process for setting the National Ambient Air Quality Standard
(NAAQS) for PM.  The derivation of a RFC of 14  jag/m3 essentially equivalent to the annual  PM2.5
NAAQS of 15 (ig/m3 is appropriate.  Indeed, I find no compelling evidence to not use the NAAQS PM2.5
value for the RFC for diesel exhaust particulate matter. In my opinion, this value is conservative, i.e. more
likely is overly stringent because for chronic exposure it is dependent upon key epidemiological studies for
which exposure extended over earlier decades when air quality measures, and especially particulate matter
concentrations, were substantially higher than the time periods immediately prior to the assessment of the
health impacts. In my opinion, the use of the historical ambient air concentrations suggests the reported
exposure-response functions may have been biased to the high side by a factor of 2 or more.

       The 15 (ig/m3 value for an RFC for diesel based on human data is clearly preferable to using the
value of 14 jig/m3 or the suggested revised value of 5 Jig/m3 (arrived at by extrapolation from laboratory
animal data).  That extrapolation  process is uncertain and the results  are strongly dependent on the
exposure levels selected for use in the original animal toxicity studies. If more exposure levels had been
studied above the lowest levels  studied by Manderly et al and  Ishinishi et al it is quite likely that a higher
NOAEL would have been found.

       At the meeting,  EPA staff suggested they would revise the RFC to 5  (ig/m3.  Although the
extrapolation process is uncertain, the use of two uncertainty factors (10 for intraspecies  - human  -
variability and 3 for rat to human toxicodynamic extrapolation) very adequately accounts for uncertainty.
Thus, the use of a factor of 30 to cover uncertainty results in an RFC with an associated high degree of
confidence.

       The second endpoint of concern, cancer, is more uniquely related to diesel exhaust particulate
matter than PM in general.  This relates principally to  concern  for the complex mixture  of organic
compounds associated with the readily respirable diesel exhaust particles. Without question, when the

                                             A-40

-------
organic fraction is removed from diesel exhaust particles with vigorous extraction methods and used in high
concentrations in biological systems multiple effects are observed.  These include cell damage, DNA
damage, and mutations. What is frequently overlooked are the small quantities of the organics that are
actually deposited in the human respiratory tract even with chronic exposure to diesel exhaust. As noted
on page  3-29 of the document, it is calculated that continuous exposure to 1 Hg/m3  of diesel exhaust
particulate matter results in annual deposition of 3 (ig (the report actually uses a precisely calculated value
of 2.94 jag) of polycyclic aromatic hydrocarbons (PAH) in the total lung volume. This is an important
reference statement that should be repeated in several places in the report.

       Consideration of this small quantity of PAH deposited helps one appreciate why with occupational
exposures orders  of magnitude higher than 1 Hg/m3 even in the best epidemiological studies the result is
a weak signal for excess lung cancer (a relative risk on the order of 1.4).  The ability to detect a weak
signal of excess lung cancer risk is complicated by multiple factors.  The most significant of these is the fact
that many individuals  in the epidemiological studies were cigarette smokers and smoking has associated
relative risk on the order of 10.0. Thus, the epidemiologist is challenged to try to tease out an effect for
diesel that is only about l/20th of the much more significant risk factor—smoking. Nonetheless, the
presence of a weak cancer-causing signal for diesel exhaust does raise public health concerns because of
the large population exposed at low concentrations.

       A critical issue is the potential for translating the weak signal into an estimate of potency; i.e., a unit
risk estimates. The calculation of potency requires knowledge of both health outcome and exposure of the
populations studied in the epidemiological studies.  This includes exposures that occurred over decades
prior to the observation  of lung cancer. Unfortunately, the actual level of exposures are not known and it
is impossible to reconstruct them with a high degree of certainty.  I do not have  confidence in the
reconstructed values and, thus, urge that the reconstructed exposure values not be used to either develop
unit risk estimates. In my opinion, the reconstructed  exposure values are sufficiently uncertain that I do
not favor their use even in "back of the envelope" approaches to create surrogate unit risk factors as done
in the document on page 8-13.

       In my opinion, the evidence is sufficient to characterize diesel exhaust occupational exposure at high
levels as a likely human carcinogen. The risks at ambient levels of exposure are unknown and cannot be
quantified. I do not think it is necessary to stretch the bounds of science to create quantitation that conveys
a level of certainty that is not consistent with the available science.

       On page xxi, there is a brief accounting of past versions and reviews of the document. Missing
from the account is the first draft and a consultative CASAC review that took place in 1990 or early 1991.
This CASAC review should be cited to make the record complete.
                                             A-41

-------
Chapter 1.  Executive Summary

       In general, the executive summary is adequate. However, it could be improved with some
modest changes.

1.      Page 1-2: It may be useful to cite quantitative data from EPA's Trends report in diesel
emissions of particulate matter.  Indeed, it may be appropriate to use a table that shows diesel
emissions compared to other sources as in Table 3 of Lipfert (1998) which is based on EPA data.

2.      Page 1-4, line 23-24. The references to "intensive evidence for the induction of lung cancer in
the rat from chronic inhalation exposure to high concentrations of DE" must be qualified with reference
made to the overload phenomena and the lack of relevance for estimation of human hazard.

3.      It is critical that the Executive Summary very accurately reflect the contents of the report,
including the changes proposed during the course of the October 12-13 meeting.

Chapter 2.  Diesel Emissions

1.      Page 2-11, line 5. Beyond referencing EPA's Trends report, I suggest a table be included that
shows diesel particulate emissions compared to other sources as I suggested for the Executive
Summary (see Lipfert, 1998). I feel there is great value in showing data for all major sources of
particulate emissions and oxides  of nitrogen. This type of comparison helps place the diesel issue in
perspective.

2.      Page 2-15, lines 15-19. If available, information should be provided on the sulfur content of
railroad-grade  diesel fuel compared to on-road fuel.

3.      Page 2-15, lines 20-23.  The discussion should be expanded to include consideration of
lubricating oil contributions to the solvent extractable organics and not just PAHs.

4.      Page 2-43, Figure 2-37.  Information should be provided on where relative to the origin of the
particles in the combustion cylinder and  exhaust system the number/mass particle size measurements
were made.  Kittleson may have more recent data on how number/mass size distribution changes with
distance and time relative to point of formation. (See also pg 2-45, lines 25-28.)

5.      Page 2-61. Exposure: to provide some historical perspective, this section might include one or
more figures from Lipfert (1998). (See Figures 5 and 6.) Consideration of these historical levels of
PM is needed to appropriately interpret the findings of epidemiological studies of both PM and diesel
exhaust particulate matter.
                                            A-42

-------
 Chapter 3.  Dosimetry of Diesel Particulate Matter

        This chapter in its present form is marginally adequate for use in evaluating the health hazards of
 exposure to diesel exhaust.  The chapter could be substantially improved with modest additions and
 changes.  Specific suggestions are given below:

 1.      The present chapter does not represent a balanced coverage of dosimetry information on
 humans and laboratory animals. Obviously, our primary interest is in humans and, thus, human data
 should be emphasized when it is available with laboratory animal data used only when required to fill
 gaps in our knowledge of humans.  Page 3-2, lines 3-7 needs to be revised to recognize that it is only
 necessary to become concerned with "human equivalent concentrations" derived from laboratory
 animals when sufficient human data are not available.

 2.      The chapter should give broader coverage of the total respiratory tract because inhaled diesel
 exhaust particles deposit on the total tract from the nares to the alveal: not exclusively the lung as stated
 on page 3-1, lines 30-31.

 3.      Section 3.3.1 on Deposition Mechanisms is adequate but could be enhanced with use of the
 well-known (to experts in the field, but not all readers) figures on deposition mechanisms. The section
 appears to understate the deposition of diesel exhaust particles in the nares by diffusion.

 4.      Section 3.3.1.1 would be enhanced by inclusion of a figure showing regional deposition of
 particles in humans as influenced by particle size.  This general view should be presented before the
 present Figure 3-1. It might also be useful to show similar curves for the most commonly studied
 laboratory animal species such as the rat. These are available from earlier reviews by Schlesinger and
 others.

 5.      Section 3.3.2 on Particle Clearance and Translocation could be enhanced by including a figure
 showing typical clearance patterns (as a function of time after single exposures) for humans and perhaps
 the rat.  The date of Bailey et al (1982) might be used. It is cited on pg. 3-9, lines 33-35.

 6.     A version of Figure 3-5 (referenced on page 3-19) should be used that includes the predicted
 lung burdens for the 3.5 and 7.1 Hg/m3 levels for comparison with the observed lung burdens.  This
 helps illustrate the impact of the impaired clearance and overload.  I can provide a figure if the staff
 cannot locate one.

 7.     Section 3.3 should include a discussion of the  NCRP and ICRP respiratory tract models. This
should include one or more  illustrations of model results for chronic exposure. The modeling results
 should then be compared with the predictions made by Xu and Yu, 1987 and presented in Table 3.1.
This kind of comparison was requested in the last CASAC review.
                                            A-43

-------
8.     The information presented in Section 3.5.4 on Bioavailability/Deposirion of Organics needs to
be verified and the results compared with those developed using the NCRP and 1CRP models. All
relevant parameters including assumptions should be explicitly stated.  The results presented in this
section are very important to later discussion involving effects measured in vivo as well as in in vitro
assays.  The fact that exposure to 1 (Jg/m3 of diesel exhaust continuously only results in deposition of
about 3 (ig of PAHs per year in the total human lung helps reconcile the results of in vitro assays with
very large doses and the weak cancer signal in heavily exposed occupational populations.

9.     The summary section (3.6) on modeling is  inadequate. The information requested above (items
8 and 9) could be presented here. However, a quantitative treatment and comparison of the models is
required.

10.    The footnote to Table 3-1 has typos related to expressions of surface area  in cm3 rather than
cm2. Reference is made to human data on total lung volume, total airway surface area and the surface
area of the unciliated airways.  Similar data should be presented for rats and hamsters. In addition,
other critical respiratory parameters should be presented for all three species (humans, rats and
hamsters).

11.    The source for Figure 3.3 is unclear.

Chapter 4. Mutagenicity

       This chapter in its present form is adequate for use in evaluating the health hazards of exposure
to diesel exhaust. The chapter would be substantially improved by including
(a) a contextual setting as to why mutagenicity of diesel exhaust is of interest, (b) a historical context,
and (c) a context for considering the relatively large doses of diesel  exhaust particulate extracts used in
the various in vitro and in vivo assays. Recall the EPA calculation that continuous exposure to  1 Jig/m3
of DPM results in annual deposition of 3  )Jg of PAHs in the total human lung. This result from Chapter
3 should be summarized in Chapter 4.

1.     Page 4-1. A brief paragraph should be added concerning the linkage between mutagenicity
and carcinogenicity. Hence, the basis for the high degree of interest in the mutagenicity of diesel
exhaust. This is done later but it would be helpful as part of a "road map" up front.

2.     Page 4-1. A brief paragraph should be added providing a historical context for the research on
mutagenicity of diesel exhaust particle extracts.  This would include a reference to the paper of Kotin et
a] (1955) that describes the presence of aromatic hydrocarbons in diesel  exhaust and the
carcinogenicity of the extracts when painted on mouse skin. (Kotin, P.; Falk, H. L.; and Thomas, M.,
1995. Aromatic Hydrocarbons: 111. Presence in the Particulate Phase  of diesel engine exhausts and
the carcinogenicity of exhaust extracts, hid. Health H:  113-120.)  The results reported in this paper
coupled with the availability of the Ames test in the 1970s triggered EPA's early efforts to evaluate the

                                             A-44

-------
mutagenicity of diesel exhaust. Note should be made of two aspects of EPA's early program: (a) its
orientation to attempting to use a comparative potency approach (short-term mutagenicity and cell
transformation assays, skin painting/carcinogenicity assays, short-term animal studies, and
epidemiological evidence on roofing tar, coke oven emissions, and cigarette smoking) to estimate diesel
exhaust risks and (b) biological activity-directed chemical fractination/characterization. This would
include references to several papers by Roy Albert who championed the comparative potency
approach and Joellen Lewtas who guided the biological activity directed chemical fraction work.  This
work is covered starting at the bottom of page 4-1, but should be introduced in opening paragraph of
the chapter.  A reference on the comparative potency approach that might be included is Cuddihy, R.
G.; W. C. Griffith, and R. O. McClellan, 1984, Health Risks from Light Duty Diesel Vehicles,"
Environmental Science Technical 18: 14A - 21 A. A related reference is McClellan, R. O., R. G.
Cuddihy, W. C. Griffith, and J. L. Manderly (1989); Integrating Diesel Data Sets to Assess the Risks
of Air Pollutants in Assessment of Inhalation Hazards. (U. Mohr, D. R. Bates, D. L. Dungworth, P. N.
Lee, R. O. McClellan, and FJC Roe, page 3-22 ILSI Monograph International Life Science Institute,
Springer-Verlag, Berlin, Germany.  Although the comparative potency approach has been replaced, the
use of direct, but highly uncertain epidemiological evidence, the approach still deserves brief coverage.

3.     A brief paragraph should be included to place the dosages used in the mutagenicity assays in to
perspective relative to plausible human exposures.  This would include  use of Figure 3 from McClellan
(1986)  (McClellan, R. O., "Opening Remarks: Toxicological Effects of Emissions from Diesel Engines
(1986). In: Carcinogenicity and Mutagenicity of Diesel Engine Exhaust.  N. Ishinishi, A. Koizum, R.
O. McClellan and W. Stober, Eds. Pg. 3-8, Elsevier Science Publishers, Amsterdam.)  Reference
should also be made here to the material presented on page 3-29 of Chapter 3 on Dosimerry. That
section notes that an individual inhaling 1 ^g diesel exhaust particulate matter/m3 for  1 year would
deposit 420  ng of DEP and 2.94 jig of PAHs in their lung.

4.     Page 4-1:  Close to the reference of proceedings of symposiums on the health effects of diesel
emissions, reference should also be made to the Health Effects Institute document that reviewed the
health effects of diesel exhaust. Several chapters in that excellent review cover the issue of genotoxicity
and its  linkages to carcinogenocity.

5.     Page 4-4:  A recent paper from Japan on gene mutations evaluated in rodents with a marker
gene and exposure to diesel exhaust should be cited. I do not have this reference at hand.  Although it
notes an increase in gene mutations in the lung it fails to cite the classical studies by Driscoll and
Oberdorster which show that the mechanisms of mutagenic response to soot particles is likely via an
inflammation/oxidative stress pathway.

Chapter 6.    Quantitative Approaches to Estimating Human Non-Cancer Health Risks of
Diesel  Exhaust
                                            A-45

-------
       This chapter is acceptable for use in evaluating the human health hazards of exposure to diesel
exhaust.  It could benefit from rigorous editing to improve the clarity and make it more concise. The
following points require specific attention.

1.      Page 6-4, line 32:  Reference should be made to the recent re-analysis of the Harvard Six-
Cities Study and the American Cancer Society Study performed under Health Effects Institute
sponsorship by a team led by Professor Dan Krewski. The results of the re-analysis should be briefly
summarized.  In addition, the recent multi-city study performed by Professor Jon Samet and associates
under Health Effects Institute sponsorship should be cited and the results briefly summarized.

       In discussing the results of the Harvard Six Cities and the American Cancer Society Study,
reference should be made to Lipfert (1998), who has emphasized the importance of considering
historical changes in air concentrations of particulate when estimating exposure-response relationships.
Lipfert (1995) indicates that when historical TSP values were substituted for more current values in a
re-analysis of the Harvard Six Cities Study the regression coefficient for the exposure-response
relationship was reduced by a factor of 2.6. (Lipfert, F. W. (1995) "Estimating Air Pollution -
Mortality Risks from Cross-Sectional Studies: Prospective vs. Ecological Study Designs." In:
Particulate Matter: Health and Regulatory Issues, pp 78-102. AWMA Pub. No. VIP 49.
Proceedings of the International Specialty Conference, Pittsburgh, PA.) More recently, Lipfert (1998)
has reviewed data on historical changes in particle concentrations and discussed how these changes
should be considered in the analysis and interpretation of American Cancer Society Study reported by
Pope et al. Lipfert (1998) indicates that the failure to use this exposure data from the earlier time
period biased  Pope et al's regression coefficients for exposure-response relationships upwards by a
factor of 2 to 5.5. He indicates regression coefficients based on air quality data for the earlier time
period could have yielded coefficients in the range of 0.0013 to 0.0035 per Jig/m3. (Lipfert, F. W.
(1998), "Trends in Airborne Particulate Matter in the United States." Appl. Assup. Environ.  Hyg. 3.
(6):  370-384). The time trends in ambient air concentrations of particulate and other pollutants need to
be considered in evaluating health effects studies associating effects in morbidity and mortality with
long-term exposure to pollution. It may also be informative to consider the changing patterns of
emission sources.

2.      Page 6-6, lines 27-34:  The approach outlined in this section has very little support and should
be eliminated.  If the paragraph is left in it should be placed after more plausible approaches.  On line
27 the word "view" should be replaced by "assume" and on line 28 the word "treat" should be replaced
by "assume."  If the approach taken here were to be used it would also  have to assume that particles
other than DPM are without any health effects.

3.      Page 6-7, lines 9-12: The approach proposing apportionment of the 15 Jig/m3 NAAQS for
PM2.5 to different sources, in this case to diesel, is inappropriate. This is not an appropriate part of the
setting of an inhalation reference concentration and this sentence should be  removed. Instead, I might
suggest "The approach of assuming equal potency for PM2.5 and diesel exhaust suggests that an

                                             A-46

-------
appropriate inhalation reference concentration for DPM is 15 ug/m3, a value equal to the annual
NAAQSforPM25."

4.     Page 6-9, lines 4-6: The reader is provided a one sentence pointer to a 10 page appendix (B)
on "Benchmark Concentration Analysis of Diesel Data." If this Appendix (B) is worthy of inclusion,
then it should be presented more clearly in the text. I suggest the appendix remain and that a paragraph
be used to introduce the approach on page 6-9 and then a paragraph used on page 6-13 or 6-14 to
indicate clearly why the Benchmark approach is not being used.

5.     Page 6-10, lines 22-24: This sentence should be removed from the text and added as a
footnote to Table 6-1.

6.     Table 6-1:  The footnote notes for the table need to be reviewed and revised. A key footnote
was place in the text (see above) and others are apparently note used.

7.     Page 6-11, HEC Derivation:  This section suffers from inadequacies noted in the Dosimetry
chapter and is excessively long and complicated. Part of the  difficulty relates to excessive reliance on
the model of Yu (1991) and the failure to make quantitative use of NCRP and 1CRP models for
particle deposition  and clearance.  The section could be improved if reference were made to a key
table as I have suggested for inclusion in the Dosimetry chapter that would detail all of the relevant input
parameters (including assumptions) for rats and humans used to calculate a "Human Equivalent
Concentration" for diesel exhaust particles. The table would  show input data and output for the Yu,
NCRP and 1CRP models and be accompanied by text explaining similarities and differences in results.
The present Dosimetry chapter does not contain such a table.

       The existing section 6.5.2 HEC Derivation could be substantially shortened if a table such as
described above were included in the report.

8.     Table 6-2:  This table could be improved by including a column that normalizes all the rat
exposure data to mg/m3 or Jig/rn3 based on continuous exposure.  This is the first step in evaluating the
HEC. Burying this normalization in the HEC values given mystifies the HEC evaluation process.
Indeed, I am concerned that excessive reliance on the Yu model has unnecessarily complicated the
HEC calculation process and may have introduced distortions in the basic data.  For example, the ratio
of the exposure concentration (not normalized to continuous  exposure) to the HEC for the lowest
exposure level (0.35 mg/m3) for the Manderly et al (1987a) study is about 10 to 1 while for the 3.5 and
7.0 mg/m3 levels it is nearly 2 to 1. For the Ishinishi et al (1988) study the  change in this data from the
lowest exposure concentration to the highest shifts from 3 to 1  to about 0.8 to 1. The difference
between the Manderly et al and Ishinishi et al studies at a given exposure concentration is primarily due
to the difference in exposure time (35 hours per week versus 96 hours per week).  Presumably, the
remainder of the differences including the  shift from low to high exposure concentrations results from
Yu's modeling of overload. I submit that Yu and the EPA staff have gotten carried away with

                                            A-47

-------
modeling, perhaps best characterized as "mathematical gymnastics," and would do better by focusing
on the basic data.  I suspect it may not change the basic RFC calculation since reliance is based on the
low exposure concentration data. The key question is whether one feel comfortable saying that rats
exposed to 0.46 mg/m3, 16 hours/day, 6 days/wk (normalized 0.38 mg/m3) and humans exposed to
0.144 mg/m3 are equivalent exposures in terms of dose to lung. This difference appears about right and
is largely related to differences in fractional deposition for rats and humans. This element of simplicity is
lost in the mathematical details of Appendix A.

9.      Pages 6-13, line 16:  A brief paragraph should be added here explaining why the Human
Equivalent Concentration to rat exposure concentration relationship is not proportional over the range
of exposures described.  This should include a brief statement as to why the shift occurs, is it all
attributable to overload? If so, what is the evidence for "overload" occurring in humans? This situation
might be clarified when a comparison is made to results from use of the 1CRP and NCRI models.

10.    Page 6-13, lines 18-19:  Change to read "with no observed adverse effect levels as high as
0.144 mg/m3.

11.    Page 6-13, lines 20-21:  Change to read "in the continuum from 0.33 to 1.95 ug/m3."

12.    Page 6-13: Much of the discussion on this page  and page 6-14 fails to recognize that the
relationships observed are in part a  function of study design. For example lines 35-36 note that the
highest no-effect HECs of 0.128 mg/m3 and 0.144 mg/m3 are nearly five-fold above other no-effect
levels of 0.032 and 0.038 mg/m3. So what!  If Manderly et al had elected to  study a concentration
higher than 0.35 mg/m3, 7 hr/day, 5 day/wk, than one could have calculated a NOAEL HEC closer to
0.144 mg/m3. Likewise, it is quite possible that if studies had been conducted at levels above that
identified as the highest NOAEL but below the LOAEL, an ever-higher NOAEL might have been
identified.  This should be clearly stated in the text.

13.    Page 6-13, line 36 and page 6-14, line 3: The basis for electing to not use the BMCL]0 needs
to be expanded. One could argue that a calculated BMCL,0 of 0.37 ug/m3 combined with a HEC-
NOAEL of 0.144  mg/m3 provides a basis for selecting a value of 0.25 to 0.30 mg/m3 as a starting point
for calculations of an RFC.

14.    Page 6-15, line 19: Expand to read "0.46 mg/m3, 16 hours/day/6days/week"

15.    Page 6-17, line 13: Remove the phrase "the apportionment estimates of 1.5 - 5 ug/m3". As
discussed earlier, the issue of apportionment is not a part of the science underlying selection of an RFC.

16.    Page 6-17, line 28: Revise to read "this DE RFC value of 14 ug/m3 or a revised DPM-RFC of
5 ug/m3 is reasonably congruent with  the annual PM2.5 NAAQs of 15 ug/m3 established to protect
against adverse effects of ambient air fine particles typical of the current U.S. environment.  If a revised

                                            A-48

-------
DPM-RFC is calculated using an uncertainty factor of 10 for intraspecies variability and 3 for
interspecies (toxicodynamic considerations), then I think the resulting RFC should be viewed as having
a high degree of confidence.

17.    Page 6-21: Expand to read "0.46 mg/m3,  16 hours/day, 6 days/week, a NOAEL."

18.    Appendix A:  This is an extraordinarily detailed presentation that would benefit greatly from
inclusion of an easy-to-read summary. Such a summary would include a table such as I have
advocated elsewhere, summarizing all of the critical input and output parameters for modeling the
disposition of particles in rats and humans exposed to diesel exhaust particles. It may be necessary to
modify Appendix A to include details related to use of the ICRP and NCRP models.

Chapter 7. Carcinogenicity of Diesel Exhaust

1.     Page 7-2, line 32:  As noted elsewhere, in considering the Harvard Six Cities and American
Cancer Society Studies, it is important to make note of the historical changes in air quality (Lipfert,
1998).

2.     Page 7-5, lines 8-10:  The figures presented here on sales of diesel-powered trucks need to be
qualified by making reference to the portion of the on-road truck fleet that was dieselized. Perhaps this
portion of Chapter 7 can be better linked to Chapter 2 on diesel emissions.

3.     Page 7-113, Weight-of-Evidence: This section needs to be rewritten. In my opinion, it is not
appropriate to use the rat lung tumor data inhalation study as part of the weight of evidence for human
carcinogenicity. It is totally inappropriate to include data from rats given intratracheal inhalations as part
of the weight of evidence in the face of the compelling mechanistic data from well-conducted inhalation
studies showing a lack of relevance of the rat data for assessing human hazard for diesel exhaust
particulate matter exposures.  Indeed, the most relevant rat data are  from the many rats  studied at levels
not producing an overload and these were negative for cancer induction (Valberg).

4.     Page 7-113, line 27-31: In my opinion the following statement is not adequately supported and
should be eliminated "Nonetheless, available data indicates that DE-induced lung carcinogenicity seems
to be mediated by mutagenic and non-mutagenic events by both the particles and associated organic
compounds, although a role for the organics in the gaseous phase cannot be ruled out.  Given that there
is some evidence for a mutagenic mode of action, a cancer hazard is presumed at any exposure level."
In my opinion, the argument  for a role for mutagenic events has not been clearly articulated and, thus,
the case for a linear extrapolation to levels of a few lig/m3 has not been made. Consequently, this
assumption is clearly a default and should be stated as such. The summary statement provided on page
114, lines 29-30 is appropriate. (Perhaps it was written by a different author.)

Chapter 8.  Dose-Response Assessment: Carcinogenic Effects

                                            A-49

-------
1.      Page 8-2, lines 29-31: While it has been shown unequivocally in several studies that DE can
cause benign, and malignant lung tumors in rats in a dose-related manner following chronic inhalation
exposure to sufficiently high concentrations, it has also been shown that low, but still substantial
exposures to diesel exhaust do not cause an excess of lung cancer in rats (Valberg). Further, it has
been clearly demonstrated that the rat lung tumors occur via a mode of action (particle overload,
inflammation, reactive oxygen species, mutations and lung cancer) that is unlikely to be operative in
humans at ambient levels of exposure.  Thus, the quoted sentence should be removed.  The sentence is
not consistent with lines 13-21 on page 8-3. The EPA staff should  accept the evidence for human
cancer risk from diesel exhaust for what it is—weak—and avoid overstating the evidence.

2.      Page 8-3, line 27:  I suggest adding:  "In addition, it is difficult to ascertain the role of ambient
particulate matter that was at substantially higher concentrations for much of the life of the
occupationally exposed individuals than was the case when the epidemiological studies were conducted
(Lipfert, 1998).

3.      Page 8-4, line 10:  I suggest you insert "For example, relative risks on the order of 10 are
frequently observed for lung cancer in  smokers compared to relative risks for diesel exposed workers
for up to about 1.5"

4.      Page 8-5, line 4:  I suggest it be revised to read studies have 'reconstructed quantitative
historical exposure data" to emphasize the exposure data was reconstructed. I suggest that any time
reconstructed historical exposure data is cited it should be specifically cited as reconstructed historical
exposure data.

5.      Page 8-6, lines 4-20: It would be appropriate to give the confidence intervals  for the odds ratio
of 1.64 and  1.41.  In addition, the odds ratios and the associated confidence intervals for smoking as
reported by the author should be given. This helps provide the reader with perspective as to the weak
evidence for cancer risks from diesel exhaust exposure vs. cigarette smoking.

6.      Page 8-9, line 8:  add "lack of knowledge of whether the men actually drove diesel trucks."

7.      Page 8-9, lines 9-17: I think the authors  should state more  strongly the high degree of
uncertainty associated with retrospectively reconstructing exposures for 1948 to 1983 from 1990
exposure assessments.

8.      Page 8-10, line 9:  I think the value is "about 15-fold" rather than five-fold.

9.      Page 8-10, lines 12-24:  I suggest that any time an exposure value is presented, especially
when presented to three  significant figures, it should be prefaced by "estimated." And, as noted
elsewhere, it should be stated that ambient particulate exposures, including EC, were undoubtedly much
higher for a major portion of the life of the individual.

                                             A-50

-------
 10.    Page 8-11, lines 4-7: With reference to activities underway, it would be appropriate to note
 both the Australian miner study and the N1OSH miner study.

 11.    Page 8-12, lines 27-29: It is inappropriate to state a 5% background lifetime lung cancer risk
 for the U.S. population without giving data showing how non-uniformly this risk is distributed across the
 total population.  For example, the lifetime lung cancer risk for non-smokers is about 1% compared to
 10% for smokers with about 85% of the cases in this total population attributable to cigarette smoking.
 Since a relative risk approach is being used it follows that 85% of the estimated cases attributed to
 diesel exposure will occur in smokers.  Thus, for non-smoking workers the excess risk attributable to
 diesel exhaust exposure would be about 0.4% and for smokers about 4%.  In similar fashion the
 subsequent risk numbers need to be lowered by a factor of 5 for non-smokers and elevated by a factor
 of 2 for smokers. And one could go on and on playing a game of "mathematical gymnastics. The more
 one plays the game of "mathematical gymnastics," the more obscure the underlying data becomes and
 the greater the likelihood of losing contact with reality. I suggest that we call time-out on "mathematical
 gymnastics" and agree the existing epidemiological data are inadequate for quantitation.  The facts are
 that we have a weak signal for potency and a potentially large population exposed to low levels of
 diesel exhaust on the order of a fig/m3.

 12.    Page 8-13, lines 28-29: If the Agency is  concerned about non-road sources of diesel exhaust,
 then why didn't they include more information on non-road sources. The statement seems to imply that
 only a portion of the U.S. population is exposed to diesel exhaust from non-road sources. Intuitively, I
 suspect the non-road sources contribute generally to exposure across the U.S.  But the real question is
 where is the data? The statement "children who may be more sensitive to early life exposure" appears
 to be almost a throw-away.  Again, what is the evidence?

 13.    Page 8-13, line 30 and earlier:  I am concerned about the excessive quantitation based on very
 limited data. I doubt that this paragraph sufficiently qualifies the "mathematical gymnastics" exercises. If
these numbers are left in, I suggest words be added as follows - "There is a low degree of confidence in
the risk numbers cited and, therefore, they should  not be construed in any way be equivalent to the unit
risk values the agency has sometimes calculated."

 14.    Page 8-14, lines 6-7: I suggest it would be appropriate to reword as follows - "Nevertheless,
these analyses indicate that environmental exposures on the order of a ^tg/rn3 may pose a lung cancer
hazard even  though a level of zero risk cannot be  excluded."

 15.    Page 8-15, line 4: Remove the reference to environmental cancer risks from DE ranging from
 10-5tolO'3.

 16.    At several places in this chapter it is very important to emphasize that the basic epidemiological
data shows an association between employment in the trucking and railroad industry and a weak cancer
signal. It is important to emphasize that the epidemiological studies were not based on exposure to

                                           A-51

-------
DPM but rather job classification, duration of employment, etc., and an evaluation of association with
lung cancer. In a second step, estimates of exposure to DPM are introduced as a means of
extrapolation to ambient air concentrations and the general population. There is a high degree of
uncertainty in extrapolating from these very complex occupational situations to the general population
and from high levels of exposure to DPM and many other agents to substantially lower ambient levels of
DPM.
                                            A-52

-------
APPENDIX C

       As noted earlier in my comments, the estimation of exposure becomes a critical part of
calculating the potency (i.e., slope of the dose-response function) for past exposures increasing
mortality.  This point should be made in Appendix C when the Harvard Six Cities and American
Cancer Society Studies are cited. It will be appropriate to reference the Lipfert (1995 and 1998)
papers and note the impact of using estimates of exposure from historical air quality data. It may be
appropriate to include one or more of the figures from the Lipfert (1998) paper.
                                            A-53

-------
                                  Giinter Oberdorster, PhD

Chapter 3: Dosimetry of Diesel Particulate Matter

       This revised chapter is significantly improved, and the comments made at the previous review have
generally been addressed. References to the PM document have been made throughout the Chapter, and
some figures have been included.  It would still be useful to include also a figure showing the deposition
efficiency in humans over the whole particle range from ultrafines to 100 |lm, for example, as it has been
derived by 1CRP. The comments made below relate to a few additional issues which should be answered
and incorporated in this draft.

       Page 3-1. line 2: I suggest to delete "clearly" and start the sentence with "Animals and humans ...."

       Page 3-3. line 6: I suggest to delete "many".

       Page 3-6. lines 25/26: The insignificant rate of clearance by dissolution pertains only to the core
of DPM, not the whole DPM. Furthermore, there is also a significant difference between rats and humans
withrespect to the importance of clearance by dissolution. Due to the long alveolar macrophage-mediated
clearance in humans — in  contrast to rats — dissolution of even "poorly soluble particles" can be a
significant contributor for humans as has been pointed out by Kreyling.

       Page 3-7. line 3: Clearance of PSP deposited in the oral passages is not by coughing.

       Page 3-8. lines 5/6: One study is indicated here to show a difference in tracheal transport with
respect to  age.  The  reference is missing, and can that one study be used to generalize age-related
differences?

       Page 3-9. lines 4-10: It is speculated here as to why PAH on the diesel particles may have an
effect in humans but not in rats. One possibility suggested here is the greater interstitial localization of
inhaled diesel particles in the primate lung compared to rat lungs, based on the paper by Nikula et al. One
has to  be cautious with this interpretation of the paper since translocation rates from the alveolar to the
interstitium maybe quite different between rats and primates, similar to the alveolar macrophage-mediated
clearance rates being much faster in rats  than in humans. Thus, an evaluation of interstitial  vs. alveolar
particles made at one point in time only (at the end of a two-year study in rats and monkeys) probably does
not give an accurate picture of the kinetics of such translocation and the importance of the interstitial
compartment of one vs. the other species.

       Page 3-16. lines 11/12: It should be emphasized here that the studies by Adamson and Bowden
used high doses  by intratracheal instillation and that this will lead to significant direct particle-type I cell
interactions in the lung. The mechanism of interstitial access of particles at very high doses given as a bolus
is likely quite different from that at much lower inhaled doses.

                                            A-54

-------
        Page 3-16. lines 24-29: As mentioned before, the study by Nikula et al. needs to consider that
only one point in time after exposure to different particles in rats and monkeys was used to determine the
relative ratios of dust accumulation in the alveolar space vs. the pulmonary interstitium. As is well known
from other studies, there is significant accumulation of dusts inhaled at higher concentration by rats in their
tracheobronchial lymphnodes, /'. e., these particles have to be transported there via interstitial access; if the
clearance rate from the interstitium into lymphatic tissues is faster in rats than in monkeys this could account
for the observation that relatively less interstitial particles were found in the rat as compared to monkeys.
This may give rise to an incorrect conclusion with respect to species differences of the persistence of
interstitial vs. alveolar particles.

        Page 3-18. line 13: Change "pathophysical" to "pathophysiological".

        Page 3-19. lines 16-18: Reference is made here to Figure 3-5. There is a typo in the figure legend
and in the labelling of the ordinate, change PPM to DPM.

        Page 3-19.  lines 30/31: The revised version of the Pock model for rats by Stober is described
here, including an interstitial space compartment for retained particles that increases greatly.  This model
appears to be in contrast to the conclusions  by Nikula et al. if one reads the document in its present form.
However, as suggested above, the  data by Nikula et al. represent only one point  in time and may have to
be viewed in the context of different interstitial clearance rates between rats and monkeys.

        Page 3-20.  line 8: The overloading effect has also been noted in mice and hamsters, not only in
rats. See studies by Muhle et al.

        Page 3-20. lines 23/24: It  is not only a suggestion that macrophage-mediated clearance is slower
in humans than in rats, but it is, indeed, a fact that has been reaffirmed repeatedly.

        Page 3-20.  line 25:   It should be clarified what significant differences in  macrophage loading
between species are alluded to here.

        Page 3-21.  lines 28/29: It is noted here that no lung cancer was reported among miners with
apparent particle overload. However, it should also be noted that the same is true for rats where particle
lung overload can occur without the induction of lung tumors. See for example  the study by Lee et al.
where 50 mg/m3 of TiO2 exposure over two  years resulted in significant particle overload but no increased
lung tumors.

        Page 3-22.  lines  32-35:   In addition  to the suggestion that surface characteristics of alveolar
macrophages  are altered so they adhere  to each other,  it has also been suggested that  alveolar
macrophages activated by phagocytized particles release chemotactic factors that in turn attract other
macrophages leading to cluster formation. This was suggested by Bellmann et al.  (J. A erosol Science 21:
                                             A-55

-------
377-380, 1990) where a clearance and retention model with a specific alveolar macrophage compartment
is described.

       Page 3-23. line 32: Change "of to "at".

       Page 3-29. line 10: I suggest to change "calculate" to "estimate".

       Page 3-36. lines 21-36: This paragraph describes clearance rates for poorly soluble particles in
humans for larger particles and an adjustment to smaller ones. However, the adjustment being used for the
smaller particles are clearance rates determined for the rat by Snipes (1979). Moreover, the clearance
rates for these smaller particles are taken from an annual report of ITRI, i.e., these data were not peer-
reviewed. It is, thus, not understandable why the rat clearance rates should be used for humans; and it is
also not clear which larger particle size alluded to in line 29 is meant, is it 0.4 (Im or 2 |im? Likewise, it
is not clear why this choice would underestimate rather than overestimate (line 30) the correct clearance
rate for DPM.

       Page 3-37. lines 24/25: It is pointed out here that the Yu model has a significant clearance into the
lymphatic system, i.e., via interstitial compartment. This needs to be kept in mind, because later in the
document (see below) it is stated that the Yu model does not have an interstitial compartment.  (See also
publication by Hsieh and Yu, 1998, Inhalation Tox.).

       Page 3-38. lines 22-33:  Line 22 mentions the lack of an interstitial compartment in the Yu model,
which is not quite correct since an interstitial pathway in the Yu model for lymphatic clearance is certainly
provided. In this paragraph the work of Kuempel and Nikula is also mentioned as providing compelling
evidence on the significance of an interstitialization process in primates. Again, as stated before, rats also
have a  significant amount of particles cleared into the interstitium and the regional lymph nodes, and this
clearance is considered in the Yu model as well since in this model the alveolar clearance rate consists of
the macrophage-mediated and the interstitial clearance.  Specifically in the particle overload situation, this
pathway appears to be rather rapid in rats. Thus, the difference betweenrats and humans maybe the rate
of clearance into and out  of the interstitium, similar to the differences in alveolar macrophage-mediated
clearance rates between rats and humans.

       Line 30 states that the  findings in retired coal miners are  consistent with the existence of an
interstitial compartment.  However, these findings  are also  quite consistent  with overload induced
retardation of alveolar macrophage-mediated clearance, as has been pointed out before.

       Page 3-39. lines 34/35: The lack of the interstitial compartment in the Yu model is again addressed
here which may have to be revised.

       Page 3-40. lines 17/18:  The statement here that dissolution is insignificant for poorly soluble
particles compared to clearance as an intact particle is not necessarily true for humans because of the much

                                            A-56

-------
longer alveolar macrophage-mediated clearance in humans vs. rats. This was pointed out by Kreyling
repeatedly.

       Page 3-40. lines 22: The prominence of interstitialization poorly soluble particles in primates vs.
rodents should be considered in light of possibly much faster clearance rates for this pathway in rodents,
as mentioned several times before.

       Page 3-40. line 25:  Replace "may" with "is". It is, indeed, a fact that prolonged exposure to high
concentrations of particles will lead to particle overload.

       Also, replace "appears" with "is" in line 30. It has been well established that, indeed, macrophage-
mediated clearance is slower in humans than in rats.

       Page 3-41. line 31:  The MMAD of 0.2 |J,m should more appropriately be called a mass median
thermodynamic diameter.

       In general, this chapter should also emphasize more the variability in outcome between different
models with respect to deposition and clearance as well as the biological individual variability in particle
deposition and clearance which can lead to significant differences in estimates of retained particle burdens.
Thus, a cautionary note that the use of the C. P.Yu model should not be viewed as an endorsement that
it is the most appropriate one to use. The predictions derived from this model as well as from other models
still have potentially large uncertainties.

Chapter 4: Mutagenicity.

       This chapter does not include revisions that were requested repeatedly in all previous reviews of
the document, i.e., addressing the issue of very high doses which were used in mutagenicity assays relative
to doses that can reasonably be  expected to occur in vivo.   The EPA should take this suggestion a bit
more seriously and include a statement in this section regarding the high dose levels being used.

Chapter 9: Characterization of Potential Human Health Effects of Diesel Exhaust;
Hazard and Dose-Response  Assessments

       Based on the extensive discussions that the Panel had  at the CAS AC meeting this chapter will be
revised accordingly by EPA. Most of the following comments are repetitive since they were submitted
prior to these discussions and have been addressed at the meeting.

       Obviously, EPA  staff is struggling in  this chapter to find  the right wording for the degree of
carcinogenicity for diesel exhaust which sometimes leads to awkward sentences. For example, statements
like "the human evidence for potential carcinogenicity for DE is judged to be strong but less than sufficient"
(page 9-11, lines 15/16),  or "DE is likely to be carcinogenic to humans by inhalation at any exposure

                                            A-57

-------
condition" (page 9-13, lines 29/30).  Another sentence states (page 9-12, lines 13-15) that "the animal
evidence provides additional support for identifying a potential cancer hazard to humans, but is considered
not suitable for subsequent dose-response analysis and estimation of human risk with DE".

       This needs to be resolved: If the evidence is not sufficient, it cannot be that strong; carcinogenicity
of DE at any  condition is hard to accept; and if the animal evidence cannot be used for human risk
extrapolation (because of an irrelevant mechanism) then they cannot provide additional support for
identifying a potential cancer hazard to humans.

       The derivation of an RfC — summarized in this chapter — is not quite clear, specifically the
justification that the value of 14 [ig/m3 for the RiC agrees with the NAAQS of 15 [ig/m3, and at the same
time it being close to the 1.5 - 5.0 (ig/m3 derived from the apportionment of the PM2.5 standard. The RiC
apparently was derived by using just one uncertainty factor of 10 for more  susceptible parts of the
population, whereas no factor for rat to human extrapolation was used. This is justified by the use of a
dosimetric extrapolation model and by a statement in Chapter 6 that rats are more sensitive than humans.
However, the  use of the dosimetric deposition, retention and clearance model addresses only to some
extent the uncertainty for interspecies extrapolation, since deposition and clearance models in general have
some degree of uncertainty.  Thus, the use of such model does not eliminate completely the need for an
uncertainty factor which is normally 10 (3 for toxicokinetics, 3 for pharmacodynamics). In view of data
by Rudell et al. (1999; see below) showing that humans exposed for 1 hr. to 0.3 mg/m3 show significant
inflammatory cell responses in lung lavage, one could argue that humans are at least as sensitive or more
sensitive thanrats, and that an additional factor of 3 for rat to human extrapolation is justified. That would
indeed bring the present RfC  of 14 (ig/m3 down to around 5 [ig/rn3 for DE. This is a very low value, and
I consider the  confidence in  this  RiC to be moderate to low,  given that it is based on a number of
assumptions.

       Attached are the abstracts of two publications by the Swedish group, with Rudell as the first author
(Rudell et al.,  Efficiency of automotive cabin air filters to reduce acute health effects of diesel exhaust in
human subjects.  Occup &  Environ Med. 56(4): 222-231,  1999; Rudell  et al., Bronchoalveolar
inflammation after exposure to diesel exhaust:  comparison between unfiltered and particle trap filtered
exhaust. Occup & Environ Med. 56(8): 527-534, 1999).  The studies show that human subjects exposed
to diluted diesel exhaust at 300 (J-g/m3 for 1 hr. experience significant inflammatory responses with respect
to lung lavage cells.  If one contrasts this with results from a two-year rat inhalation study with diesel
exhaust (Henderson er a/., \988,FAATii: 546-567)at the lowest  exposure concentration of 350 Jlg/m3, it turns
out that in the rat study exposure at this concentration for 3 months (intermediate sacrifice timepoint) did
not show any significant inflammatory responses in the lung lavage.  Thus, a greater sensitivity of rats with
respect to non-cancer effects  cannot be assumed, as is stated in the present draft of the document.

       The studies by Rudell et al. (attached  abstracts) should also be included in Chapter 5 of the
document, Non-Cancer Health Effects  of Diesel Exposure.  These authors show,  in addition to
inflammatoryresponses in humans after short-termexposures,that by including a particle trap in the  exhaust

                                            A-58

-------
the inflammatory response in humans was still the same and that only the inclusion of an activated carbon
filter eliminated the response.  This result demonstrates that not the particles but the gaseous exhaust
components are responsible for the inflammatory cell response.

       The decision not to use the rat lung rumor response data for human extrapolation is in line with the
recommendation of CASAC since the rat tumors are induced by mechanisms which certainly will not be
operating at low environmental exposures in rats or in humans due to the existence of a threshold in particle
overload studies. There is some speculation in this chapter and in other parts of the document (Chapter
8) as to why there may be differences between induction of lung tumors in rats at high concentrations and
a suggested carcinogenicity for DE for humans at low environmental concentrations.  This difference is
hypothesized to be due to mutagenic and genotoxic effects of organic compounds in the gas and particle
phase and other effects related to induction of reactive oxygen species by the organics.
       However, one other aspect that is not discussed in the document relates to potentially significant
differences in the exposure  atmospheres of the chronic high dose rat studies vs. the low environmental
levels of diesel exhaust. At low environmental concentrations, as is described in Chapter 2 of the document
(Figure 2-37), there are  two distinct particle modes of diesel exhaust, the ultrafine particles and the
accumulation mode  particles.  In contrast, the high concentrations in the mg/m3 range used  as primary
dilution in the animal exposure studies are likely to have resulted in rapid coagulation of the ultrafine
particles onto the accumulation mode. Therefore, it is possible that the particles inhaled by the experimental
animals were only  the larger (about 0.2 |lm) accumulation mode particles.  Rapid coagulation of the
ultrafine particles at several mg/m3 of diesel exhaust canbe assumed from homogeneous and heterogeneous
coagulation processes (Hinds. Aerosol Technology. John Wiley .New York, 1982), and heterogeneous  coagulation can
be one to several orders of magnitude greater than homogeneous coagulation (NRC, 1979). Furthermore,
depending on the time and site of cooling ofthe hot gases in the dilution tunnel to room temperature could
lead to significant quenching of ultrafine particles.

       The chemical composition ofthe accumulation mode and  the ultrafine mode particles of diesel
exhaust are probably quite different (Kittelson, D.B. and Watts, W. Nanoparticle emissions from engines.
In:  Nanoparticles: Applications in Materials Science and  Environmental Science and Engineering. Natl.
Science Foundation, p. 17-21,2000; ISSN 1436-509X), which could  be  another important difference
between the experimental animals vs. humans.  The chemical composition of DPM  described in this
document (Chapter 2) is derived from filter samples with collection of both particle modes. They consist
of elemental and organic carbon compounds. In fact it is  stated in Chapter 2 that elemental carbon is the
major component of diesel exhaust, contributing approximately 50-85% ofthe diesel particulate mass. In
contrast, newer studies using particle size selective sampling show that ambient ultrafine particles consist
mainly of organic carbon compounds. (European Aerosol Conference 2000; Hughes et al., Environ. Sci
& Technol. 32, No. 9, 1998).  If ultrafine diesel particles as part of the ambient nuclei mode also consist
of organics, this would point to an important difference between low environmental exposure  of humans
to diesel (ultrafine +  accumulation mode particles) and high experimental exposures of rats to diesel (only
accumulation mode particles). Although speculative, this suggested difference could be  included into this
document together with other hypotheses about differences in response to diesel betweenrats and humans

                                            A-59

-------
(e.g., Section 7.4.5 - Integrative Hypothesis for Diesel-Induced Lung Cancer). Ultrafine particles (-20
nm) have a very high predicted deposition efficiency in the alveolar region of the respiratory tract, in fact
per unit surface area they have an approximately fifty times higher deposition in the conducting airways
compared to the alveolar region (ICRP  1994 model), and deposition is significantly greater than for
accumulation mode particles.  Significant differences of exposures oftarget cells between humans and rats
may, therefore, occur at low and high exposure levels.

       Some specific comments on Chapter 9 include: Avoid the use of microns, it should be micrometer
(page 9-2, lines 27-28).

       Page 9-12. lines 27/28: I suggest to add that the mutagenicity assays were performed at very high
doses.

       Page 9-17. line 25: Change microgram to milligram.

       Withrespectto reaching closure of this Health Assessment Document for Diesel Exhaust, I am in
favor of it, with the strong recommendation that EPA staff includes the changes that were discussed and
agreed upon at the CASAC meeting on October 12/13,2000.
                                            A-60

-------
                                  Ronald E. Wyzga, Sc. D.

Overall Comments:

        The document is much improved over earlier documents. The document is well organized, well
written, and works systematically to a logical conclusion. During the course of the meeting, several changes
were recommended for the document. Unfortunately, there was no record of the changes to be made.
If the document incorporates the changes and caveats that my colleagues and I suggested, I believe that
the document is ready for public release.

Chapter 1:

        Executive Summary.  Comments reflect those given below.  I will bring specific comments to
meeting.

Chapter 2:

        This chapter is missing several recent references; they should be reviewed and incorporated into
the document:

Brown, J. E.; Clayton, M. J.; Harris, D. B.; King, F. G., Jr. (2000) Comparison of the particle size
distribution of heavy-duty diesel exhaust using a dilution tailpipe sampler and an in-plume sampler during
on-road operation. J. Air Waste Manage. Assoc. 50: in press.

Christoforou, C. S.; Salmon, L. G.; Hannigan, M. P.; Solomon, P. A.; Cass, G. R. (2000) Trends in fine
particle concentration and chemical composition in southern California. J. Air Waste Manage. Assoc. 50:
43-53.

Cobum, T. C. (2000) Statistical analysis of on-road particulate matter emissions from diesel vehicles.
Inhalation Toxicol.  12(suppl.): 23-33.

Janssen, N. A. H.; de Hartog, J. J.; Hoek, G.; Brunekreef, B.; Lanki, T.; Timonen, K. L.; Pekkanen, J.
(2000) Personal exposure to fine particulate matter in elderly subjects: relation between personal, indoor,
and outdoor concentrations. J. Air Waste Manage. Assoc. 50: 1133-1143.

Kinney, P. L.; Aggarwal,  M.; Northridge, M. E.; Janssen, N.  A. H.; Shepard, P. (2000) Airborne
concentrations ofPM2.5 and diesel exhaustparticlesonHarlemsidewalks:acommunity-based pilot study.
Environ. Health Perspect.  108: 213-218.

Ramadan, Z.; Song, X.-H.;Hopke, P.  K. (2000) Identification of sources of Phoenix aerosol by positive
matrix  factorization. J. Air Waste Manage. Assoc. 50: in press.

                                            A-61

-------
Singh, R. B.; Huber, A. H.; Braddock, J. M. (2000) Development of a microscale emissions factor model
for particulate matter (MicroFacPM) for predicting real-time motor vehicle emission.  Presented at:
PM2000: particulate matter and health--the scientific basis for regulatory decision-making, specialty
conference & exhibition; January; Charleston, SC. Pittsburgh

   The article by Christoforou et al. documents the nature of the changes in exposure that are occurring
over time. These data would be important to include in the document, especially in section 2.2.8.1.3.

     The paper by Kinney et  al. shows that under some circumstances short-term exposures to diesel
particles can be high.

Other comments:

       The nature and significance of exposure patterns should be highlighted in the report. I realize that
data are limited here; it would be useful for the document to emphasize the need to develop data to give
us a good understanding of the magnitude and nature of personal exposures to diesel exhaust.

Chapter 5:

       First of all this chapter is missing several key references:

Abe, S.; Takizawa, H.; Sugawara,!.; Kudoh, S. (2000) Diesel exhaust (DE)-induced cytokine expression
in human bonchial epithelial cells. Am. J. Respir. Cell Mol. Biol. 22: 296-303.

Baeza-Squiban, A.; BonvaUot, V.; Boland, S.; Marano, F. (1999) Airborne  particles  evoke an
inflammatory response in human airway epithelium. Activation of transcription factors. Cell  Biol. Toxicol.
15: 375-380.

Donaldson, K. (2000) Nonneoplastic lung responses induced in experimental animals by exposure to
poorly soluble nonfibrous particles. Inhalation Toxicol. 12: 121-139.

Green,  F. H. Y. (2000) Pulmonary responses to inhaled poorly soluble particulate in the human. In:
Gardner, D. E., ed. ILSI Risk Science Institute Workshop: The Relevance of the Rat Lung Response to
Particle Overload  for Human Risk Assessment; March, 1998. Inhalation Toxicol. 12: 59-95.

Kilbum, K. H. (2000) Effects of diesel exhaust on neurobehavioral and pulmonary functions.
Arch. Environ. Health 55: 11-17.

Mauderly, J. L.; Bice, D. E.; Cheng, Y. S.; Gillett,N. A.; Henderson, R. F.; Pickrell, J. A.; Wolff, R. K.
(1989)  Influence of experimental pulmonary emphysema on toxicological effects from inhaled nitrogen
                                            A-62

-------
dioxide and diesel exhaust. Cambridge, MA: Health Effects Institute; report no. HEI-RR-89/30. Available
from: NTIS, Springfield, VA; PB90-247347.

Nightingale, J. A.; Maggs, R.; Cullinan, P.; Donnelly, L. E.; Rogers, D. F.; Kinnersley, R.; Chung, K. F.;
Barnes, P. J.; Ashmore, M.; Newman-Taylor, A. (2000) Airway inflammation after controlled exposure
to diesel exhaust particulates. Am. J. Respir. Crit. Care Med. 162: 161-166.

Nikula, K. J. (2000) Rat lung tumors induced by exposure to selected poorly soluble nonfibrous particles.
Inhalation Toxicol. 12: 97-119.

Nikula, K. J.; Vallyathan, V.; Green, F. H. Y.; Hahn, F. F. (2000) Influence of dose on the distribution of
retained particulate material in rat  and human lungs. Presented at: PM2000: particulate matter and
health~the scientific basis for regulatory decision-making,  specialty conference & exhibition; January;
Charleston, SC. Pittsburgh, PA: Air & Waste Management Association.

Nordenhall, C.; Pourazar, J.; Blomberg, A.; Levin, J.-O.;  Sandstrom, T.;  Adelroth, E. (2000) Airway
inflammation following exposure to diesel exhaust: a study  of time kinetics using induced sputum. Eur.
Respir. J.  15: 1046-1051.

Ostro, B.;  Chestnut, L.; Vichit-Vadakan, N.; Laixuthai, A. (1999) The impact of particulate matter on daily
mortality in Bangkok, Thailand. J. Air Waste Manage. Assoc. 49(Sp. Iss.  SI): PM-100-107.

Valavanidis, A.; Salika, A.; Theodoropoulou, A. (2000) Generation of hydroxyl radicals by urban
suspended particulate air matter. The role of iron ions. Atmos. Environ. 34: 2379-2386.

Ye, S.-H.; Zhou, W.; Song, J.; Peng, B.-C; Yuan, D.; Lu,  Y.-M.; Qi, P.-P. (2000) Toxicity and health
effects of vehicle emissions in Shanghai. Atmos. Environ. 34: 419-429.

       All of the above should be incorporated into the document.  I don't believe that any of them would
dramatically change the conclusions or even tone of the report.  They could, however, lead to a greater
discussion of the relationship between diesel particle  and PM health effects.  Although this is discussed in
the report, the discussion is rather superficial.  The report should delve into issues, such as, is there reason
to believe that diesel particles are as, more, or less toxic than PM genetically. What health effects  are seen
for PM? and diesel particles? Have studies looked at the same endpoints for both groups? Were the results
similar? What experiments need to be done to examine the role of the diesel  particle fraction in PM health
effects? I believe there is a general disconnect between these two mixtures: diesel emissions have generally
been related to  carcinogenic and  chronic endpoints; PM, in general, has generally been linked to acute
endpoints, especially mortality and cardiovascular endpoints.  I believe that  there is a need to examine the
role of diesel emissions in some of the endpoints identified as being associated with PM exposures.
                                            A-63

-------
       In the tables, I would like to see exposure levels broken down into the components of C x T. It
would be interesting to examine other metrics besides the product of Haber's law to see if other patterns
might emerge.
                                           A-64

-------
 Chapter 6:

        The discussion of the development of the PM-2.5 NAAQS misses the key issues as far as diesel
 particles are concerned. See my comments on Chapter 5.  In many ways I believe that this section would
 be more appropriately placed in Chapter 5 than in Chapter 6, which should focus on the quantitative dose-
 response estimation. Given the lack of parallel studies for diesel exhaust and PM, I have problems with
 statements, such as those on p. 6-17, stating that "the congruence of estimates attests to the reasonableness
 of data used and the judgments made in the RiC process...." I believe the similarity in numbers is due to
 coincidence.

        Otherwise the material in this chapter is presented in a straightforward and logical approach that
 is consistent with EPA methods for calculating the RfC.
 I have no major problem with the chapter as written.  Perhaps the rationale for choice of the uncertainty
 safety factors could be clarified.

        I would argue, however, that the toxicological database for DE may not be "relatively complete".
 P. 6-16, 1. 25. The PM literature suggests that peak exposures are associated with a variety of health
 responses, some of which have not been thoroughly investigated for DPM. (See above comments on
 Chapter 5.)  Studies, which examine the relationship between some of these responses  (e.g., heart rate
 variability, other cardiovascular parameters) and peak DPM exposure should be encouraged.  The RfC
 is concerned largely with chronic exposures; there could be a need to say something about the  relative
 safety of acute exposures; it is my understanding that such a measure is under development; at some point,
 this measure could be applied to diesel emissions.
 Chapter 7:

        This chapter presents the key studies in a clear and consistent manner.  There are areas where
 elaboration could be helpful. For example, there is scant discussion on p 7-19 on the Crump et al. and
 CalEPA analyses.  This should be expanded to highlight  the differences.  I also believe that the Health
 Effects Institute held a workshop to help  resolve the  differences in these studies.  The discussion of the
 difference in these two studies is more appropriate here than in Chapter 8, which should focus on the
 development of dose-response relationships. Since these studies received greater attention in other reviews,
 e.g., that of the State of California, it is important that they be fully discussed here so that the reader realizes
 that they were fully considered here.

        On page 7-110 it could also be noted that the epidemiology literature considers exposures to diesel
 emissions from older technologies.  It should be noted that there could be a difference in the carcinogeniciry
 associated with the older technologies may or may be the same for the newer technologies. The emissions
 characteristics are different, and until studies are done on these technologies, it is difficult  to make any
judgment.
                                             A-65

-------
Chapter 8:

       I agree with the Agency's decision not to derive a unit risk estimate for diesel exhaust. In addition
to the arguments given here, I would add that we are dealing with a complex mixture, the composition of
which is changing over time.
Hence it is unclear whether a unit risk estimate derived from specific conditions of exposure (pattern of
exposure and composition of DE) could be extrapolated to different conditions of exposure.

       I have a great many concerns about Section 8.4. I believe the perspectives can articulate societal
concerns about DE and cancer without a fixed number, which can be easily misinterpreted and  misused.
 Since numbers will be retained, I believe that all of the caveats associated with them should be highlighted
and that there should be an explicit statement in the report which states: The risk numbers presented here
are done so to help place the exposures to diesel emissions in perspective compared to other exposures.
The risk numbers are highly uncertain for the many reasons given below. Accordingly these number should
not be used to estimate the numbers of cancers associated with diesel emissions. Suchnumbers would be
misleading.  The estimates are based upon the following assumptions: 1.) there is an association between
exposure to diesel emissions and lung cancer; 2.) the underlying dose-response relationship  between
exposure and cancer incidence is linear; 3.) the risk estimates involve using data from occupational studies,
where the influence of smoking is unclear and where exposures have not been measured quantitatively; 4.)
study exposures are tied to older technologies; current  exposures probably reflect  a mix with  some
emissions from newer technologies, which have not been evaluated; 5.) risk numbers are derived using
methodologies designed to be protective; hence they may be an upper bound; for any one  individual, the
risk may be zero.  On the other hand, the upper range of some  public exposures to diesel emissions
exceeds the lower range of occupational exposures which have been related to excess lung cancers.

       Specific comments:  Can the exposure levels in Table 8-1 be augmented to provide some idea of
the year during which these exposures occurred? I raise this for two reasons: 1.) it may give some idea of
exposure trend; 2.) it gives us some idea of how we may have to compensate for differences in DE
composition over time as we learn more about these differences

Chapter 9:

       This chapter appears to have two objectives: summarize all that was presented earlier and then
interpret the conclusions made by the document. The document fairly accomplishes the first objective; I
would like to see further clarification and caveats associated with the second  objective, particularly  in the
document's hazard characterization of carcinogenicity. The document mentions the changing character of
DE over time and notes that the evidence for carcinogenicity is largely tied to older technologies (some of
which still exist in the real world, and many of which exist in the less developed world.) The key issue is
the extent to which we can extrapolate across technologies. We just don't know, and it is important to
articulate that fact.  Hence whenever it is stated that DE  is a probable human carcinogen, the statement
should be caveated that the statement is based upon studies in which exposures to DE were tied to older

                                            A-66

-------
technologies. The extent to which this statement is true for technologies which have been introduced
recently is unknown. 1 would add such a statement to section 9.4.2.2.5 and 9.5.2.

       I also have some problems with section 9.5.1.. Given the disparity of the information used to derive
the PM-2.5 standard and the RfC, I find it difficult to make the comparisons made here. (Incidentally the
units on p. 9-17,1.25 should be mg not ug.) I also do not agree with the confidence of the RiC assessment
is "medium". First of all, I don't know how to define "medium"; secondly until we have more data from
more experiments, including more studies with humans, with different exposure patterns and at dose levels
closer to ambient exposures, we have considerable uncertainty.  I would drop line 31 on p.  9-17.

I'm also uncomfortable with the discussion beginning line 32,, page 9-20 through line 13, page 9-22.  The
document fairly summarizes what is know, wisely rejects the temptation to quantify in the face of great
uncertainties, and makes a coherent and logical argument on how to characterize the carcinogenicityofDE.
Why deviate from this practice here.  We have large uncertainties about the exposure levels in the studies
of concern;  moreover the nature of exposures (patterns and composition of DE) are different in the studies
and in the public domain.  See my comments on Chapter 8.

Appendix C:

I am disappointed with this appendix. I had hoped to see some discussion of the linkages betweenPM and
DE and what is know about their health effects.
                                           A-67

-------
                                     Leslie Stayner, PhD

Chapter 6: Quantitative Approaches to Estimating Human N7onCancer Health Risks of Diesel
Exhaust

       My only major concern with this chapter was the justification for dropping the use of an uncertainty
factor for the interspecies extrapolation. In my comments on the previous draft of this document, I
expressed a concern that EPA had only used a safety factor of 3 for interspecies extrapolation rather than
the conventional factor of 10.  The rationale given for this was that a pharmacokinetic (PBPK) model was
used, and thus this removed the uncertainty related to species differences in kinetics that was assumed to
be a factor of 3.  However, this logic presumes that there is no uncertainty in the PBPK model, which is
clearly not the case for the Yu model for reasons that are now discussed in the document.

       I was therefore quite surprised when I found that this current draft not only dropped the factor of
3 for pharmacokinetics, but also dropped the other factor of 3 for pharmacodynamics. The rationale for
dropping the factor for pharcodynamic differences was that rats were more sensitive to the non-cancer
health effects of diesel exhaust exposure than humans. This statement was unreferenced, but appears to
be in part based on a misinterpretation of an ILSI report. This report did draw this conclusion for lung
cancer, but not for non-malignant respiratory diseases which is the health outcome in this analysis.  It also
appears to be based on a discussion in Chapter 3 of the histology findings of the NIOSH 2-year coal
dust/DEP study in rats and primates (at exposures equivalent to the PELs; Lewis et al. 1989).  Nikula et
al. (1997) used tissue sections from the Lewis et al. study to compare particle retention patterns in rat and
primate lungs (and found retention was primarily in alveoli in rats, and in interstitium in primates).  On page
3-16 it is noted that the alveolar response was more severe in the rats, while the interstitial response was
greater in the monkey; yet it is concluded that the rat is more sensitive.  MOST IMPORTANTLY, this
document and Nikula et al. ignore the fact that  the rats were exposed for nearly a full  lifetime, but  the
primates were exposed for only a fraction of their lives.

       I was quite satisfied when the decision  was made at our meeting  to restore the factor of 3  for
pharcodynamic  differences, but  still believe  that there is  some uncertainty related to pharmacokinetic
differences. Of course, it would be difficult to argue on scientific grounds for a factor of 10 or 3 and this
is  probably a policy call. I do believe the resulting RFC value of
5 (ig/rn3 should be reasonably protective of public health.

Other Comments

       I agree with the decision to change the level of confidence in the RFC to medium, in the sense that
I am reasonably confident that exposure to this level will not present a substantial risk of non-malignant
respiratory diseases in the general public.
                                            A-68

-------
        In discussing the use of the PM2.5 standard as an alternative means ofjustifying an RFC, it should
be  stressed that this level of exposure may still be associated with a substantial risk.  If possible, it would
be useful to present an estimate of risk for PM2.5 at the 15 [ig/m3 level to illustrate this point. I definitely
agree with the decision to drop the analysis based on source apportionment, which was confusing, and
based on an unsupportable assumption that diesel  exhaust is responsible for all of the health  effects
associated with PM2.5 exposure.

        We continue to have some concerns about the adequacy of the Yu model for predicting human
doses, based on the work of Kuempel et al. (2000) in coal miners. If our experience with studies of coal
miners applies to diesel exhaust exposed workers, then the Yu model would  grossly underestimate the
doses  in humans at environmental exposure levels. The suggestion that was made by Dr. McClellan to
consider using the ICRP and NCRP models as an alternative for humans seems very appropriate.

        Page 614, 3rd para,  2nd sentence - Why would this assessment be for individuals of average health
in their adult years.  Doesn't EPA generally consider sensitive subpopulations in their assessments (e.g. the
young and elderly)? The factor of 10 that was used for intra-human variability would seem to at least in
part be used to address this  concern.

        Page 614, last sentence - This statement should be dropped for the reasons discussed above.

Chapter 8: Dose-response Assessment: Carcinogenic Effects

        The revisions of this chapter are entirely consistent with concerns raised in the review of the
previous draft of this document. In particular:

1. The revised document moved the review of previous risk analysis from the main body of the report to
the  Appendix, and deleted  the table summarizing these  risk estimates.  This was in response to  the
comments from some of the reviews and public comments of the previous draft.  The chapter simply states
that "Appendix D provides a summary review of dose-response assessments conducted to date by other
organizations and investigators." I think this chapter should at least briefly summarize these previous efforts
and refer the reader to the Appendix for further information.

2. The document suggests that it is not possible to perform a dose-response  analysis and therefore to
develop a unit risk at this time based on the epidemiologic studies of railroad workers or truck drivers.  I
agree with this position, although hopefully the situation will be improved in the near future. We have been
working with staff of the Office of EPA Mobile Sources to improve the estimates ofhistoric diesel exhaust
particulate exposures in the  N1OSH study of teamster workers.  We expect that this effort will result in
significant improvements in our confidence in dose-response analyses using this study, although substantial
uncertainties in exposure will certainly remain. We are also working collaboratively with Dr. Eric Garshick
on the update of his study of railroad workers, and this may also help to make this study more suitable for
dose-response analyses.

                                            A-69

-------
3. A added a section to this draft that provides a "Perspective" on the significance of the cancer risk using
a rather  crude method  for determining  the  bounds of the risk that  would be consistent with  the
epidemiologic  studies of workers exposed to DEP.  This  change was responsive to at least one of the
CASAC reviewer's comments (myself).  The analysis is based on the following assumptions:  1) a linear
dose-response relationship,

2) that the estimate of relative risk (1.4) from the meta-analyses of the occupational studies is unbiased, and
3) that the range of exposures in the epidemiologic studies were between 4 and 1740 micrograms per cubic
meter. Each one of these assumptions needs to be clearly stated in the document (which it is not currently
the  case for all of these assumptions), and I believe some  discussion of the reasonableness of these
assumptions is warranted.  Personally, I believe that they are all fairly easy to justify, and I would offer the
following arguments for this position.

Linearity

       The past practice of EPA and other  U.S. agencies has been to assume that the dose-response
relationship was linear when there was an absence of evidence to the contrary. I believe clearly that this
is the position we are currently in, which is that we simply don't know if the dose-response relationship is
linear or not. The fact that we are uncomfortable with using the existing epidemiologic studies for a dose-
response assessment clearly suggests that we can't say anything at this time about the shape of the dose-
response curve based on these studies.  Although it is true  that we do not find an excess of lung cancer
among coalminers who are clearly exposed to overload concentrations, I don't believe one should interpret
this as indicating that there is a threshold for particulate exposures and lung cancer in humans. First of all,
many of these studies are SMR studies which rely on comparisons with the general population. Smoking
maynegatively bias these studies, since coal miners are not permitted to smoke while working and thus may
have lower background lung cancer risks  than the general population. Furthermore, its not clear whether
the mechanism of action of diesel exhaust in humans is related to overload as it appears to be in rats and
its quite possible that it is instead related to the genotoxic activity of the poly aromatic hydrocarbons that
are adsorbed onto the diesel particles.

       The toxicologic studies are also clearly uninformative with respect to the question as to whether the
dose-response relationship in humans for diesel exposure and lung cancer risk is linear or not. If we accept
the argument that the rat studies are not useful for estimating lung cancer risk in humans, then obviously one
can not use these studies to argue that there is a threshold or non-linear dose-response in humans.

The Validity of the Meta-Analysis Relative Risk Estimate

       It is generally impossible in epidemiologic studies to fully exclude the possibility of confounding.
It is, of course, possible  that the true meta-relative risk in these studies  should be larger, smaller or even
one (i.e., no effect).  The fact that EPA  and others have concluded that the weight of the evidence is
consistent with diesel exhaust particulate exposure being a human lung carcinogen would lead one to at least

                                             A-70

-------
accept that the true relative risk must be greater than one. Even if it were less than 1.4 this would not have
a great effect on the range of risk estimates that EPA has presented in the draft assessment. For example,
if the true relative risk were 1.2 this would only reduce the risk estimates by a factor of2, which is hardly
of any significance in this context.  Thus overall, I would argue that a relative of risk of 1.4 is unlikely to
grossly overestimate the risk, could be an underestimate and that although the true risk might be 0 that this
unlikely.
                                             A-71

-------
Range of Exposures in the Epidemiologic Studies is Between 4 and 1740

       Ideally the estimates of exposure used in this analysis should reflect that probable range of average
exposures for the cohort studies that were included in the meta-analysis.  The range used by EPA is clearly
too broad for the studies included in the meta-analysis, and I am particularly concerned about the validity
of the upper end of the range (1740).  EPA cites the HEI1995 report for the range, and this appears to
come from a statement in a paper by Watts on page 121 of the report.  This review includes exposure
assessments of miners, which is the only occupational group with exposures as high  as 1740
llg/m3.  However, both of the meta-analyses cited for the  relative risk estimate of 1.4 excluded
studies of miners  from their meta-analyses. Thus the use of 1740 Jig/m3 as an upper bound
estimate of exposures was clearly inappropriate. The highest reported average exposure from the
epidemiologic studies appears to be 191  [ig/m3, which is the  level for hostlers in the railroad study by
Garshick et al.

Other Minor Comments

Page 8-7, 2nd para, 1st sentence -1 don't think its true to say that cohort studies are usually used for dose-
response assessments.  Nested case-control studies and even  sometime population based case-control
studies may be equally or more valuable.

Page 8-8, 2nd para, 3rd sentence - It is stated that "an ideal dose-response analysis would account for ages
when exposure to DE began and terminated . .  . using exposure intensity over age rather than cumulative
exposure".  I don't understand the logic here.  Accounting for what age exposure began and terminated
may or may not be important depending on whether these variables modify the dose-response relationship.
Using exposure intensity may or may not result  in a better model than using cumulative exposure. I don't
understand what is meant by "exposure intensity over age".

Page 8-10,  2nd para, last sentence -1 simply do not understand what the authors are trying to say here.

Page 8-13,  2nd para, last sentence - I don't believe that these sources of uncertainty impact this analysis
in "opposite directions".  The true dose-response could be either supra-linear, sub-linear or linear. The
actual exposure levels for the exposures is clearly a gross overestimate as discussed above, and would lead
to an underestimation of risk.
                                           A-72

-------