x>EPA
           United States
           Environmental Protection
           Agency
            Office of Research and
            Development
            Washington, DC 20460
EPA/600/8-90/064
November 1988
Technical Support
Document on Risk
Assessment of Chemical
Mixtures

-------

-------
                                       EPA/600/8-90/064
                                         November 1988
Technical Support  Document on Risk
  Assessment of Chemical Mixtures
          Environmental Criteria and Assessment Office
         Office of Health and Environmental Assessment
            Office of Research and Development
            U.S. Environmental Protection Agency
                 Cincinnati, Ohio 45268

-------
                                  DISCLAIMER

    This  document  has  been  reviewed  In  accordance with  U.S.  Environmental
Protection  Agency  Policy  and approved  for  publication.   Mention of  trade
names or  commercial  products does not constitute  endorsement  or  recommenda-
tion for use.
                                      11

-------
                                    PREFACE

    The  preparation  of  the mixtures  Technical  Support  Document  (TSD)  was
recommended 1n  1985  by  the U,S, EPA Science  Advisory  Board  (SAB) panel that
reviewed  the   Agency's  mixtures  guidelines.   Following  completion  of  the
external review draft 1n December,  1987,  the  TSD was reviewed by both Agency
and  external  experts  In   the  field of  chemical  mixtures risk  assessment.
Among the  external reviewers  were  Ron  Wyzga  (EPRI), who  was  a  member of the
original SAB  review  panel  for  the mixtures guidelines,  and  Richard Cothern,
who Is currently a member  of the SAB.

    Unique  sections  of  the TSD Include: an  overview  of  available toxiclty
data  on complex  mixtures   and  binary  exposures  (ch.  2)  and  mechanisms  of
interaction (ch.  3), an estimate of the maximum  synergistic  effect observed
for environmental  chemicals (ch. 2),  an evaluation of  quantitative  methods
(statistics and  models)  that have  been  used  In  characterizing interactions
(ch.  4),  a summary  of  the U.S.  EPA's  Interaction  data base  (appendix  A),
recommendations  for  revisions  to  the  existing  mixtures guidelines  (ch.  5)
and  recommendations  for  future research  relevant  to  risk   assessment  (ch.
6).   The  two  most significant  conclusions  in this
available  literature is  extremely  inadequate  for
extent  of  synergism expected  from  environmental
document are 1)  that  the
 use  in  quantifying  the
 exposures,  and  2)  that
validation  of  in vitro  and short-term in  vivo studies  seems  to  offer  the
most promise for Improving risk assessments  of complex mixtures.

    The  first   draft  of  this  document  was  prepared  by  Syracuse  Research
Corporation under  contract no. 68-C8-0004  with chapters contributed  by  the
Department  of   Environmental  Health of  the  University  of Cincinnati  under
cooperative  agreement  no.  CR-813569-01-0,  and  by  staff  of  the  Agency's
Environmental  Criteria  and  Assessment  Office In Cincinnati.  The literature
search perfomed is current as  of August,  1988.
                                      ill

-------

-------
                              TABLE  OF  CONTENTS
                                                                       Page
1.  INTRODUCTION AND BACKGROUND 	  1-1

    1.1.   THE CHEMICAL MIXTURE GUIDELINES	  1-1
    1.2.   EXAMPLES OF THE U.S. EPA CHEMICAL MIXTURES RISK
           ASSESSMENT ACTIVITIES	  1-3
    1.3.   DEFINITIONS USED IN THIS DOCUMENT	1-8
    1.4.   OVERVIEW OF THIS DOCUMENT	1-10

2.  TYPES OF INFORMATION AVAILABLE	2-1

    2.1.   OVERVIEW	2-1
    2.2.   COMPLEX MIXTURES	  2-4

           2.2.1.   Overview	2-4
           2.2.2.   Ep1dem1olog1c Studies	  2-4
           2.2.3.   Whole Animal Bloassays	  2-5
           2.2.4.   In vitro Studies and Other Screening Tests. ...  2-8

    2.3.   MIXTURES OF CHEMICAL CLASSES	  2-13
    2.4.   SIMPLE MIXTURES, COMPONENTS AND TOXIC INTERACTIONS ....  2-14

           2.4.1.   Overview	  2-14
           2.4.2.   Measurements of Toxicant Interactions ......  2-16
           2.4.3.   The U.S. EPA Data Base on Toxic Interactions.  .  .  2-18

    2.5;   INTERACTIONS OF CARCINOGENS WITH OTHER COMPOUNDS . . .  .  .  2-18

           2.5.1.   Promoters and Cocarclnogens	  2-18
           2.5.2.   Inhibitors and Masking	  2-21

    2.6    QUANTIFICATION OF INTERACTIONS	  2-23

3.  AVAILABLE INFORMATION ON INTERACTION MECHANISMS	  3-1

    3.1.   OVERVIEW	  3-1
    3.2.   CHEMICAL INTERACTIONS	  3-4
    3.3.   PHARMACOKINETIC-BASED INTERACTIONS 	  3-5

           3.3.1.   Effects on Absorption	  3-5
           3.3.2.   Effects on Distribution 	  3-7
           3.3.3.   Effects on Excretion	3-8
           3.3.4.   Effects on Metabolism 	  3-8
           3.3.5.   Interactions at Receptor Sites or Critical
                    Cellular Targets	3-9
           3.3.6.   Promotion and Co-carc1nogen1c1ty	3-11
           3.3.7.   Interactions and Developmental Toxlclty .....  3-15

-------
                           TABLE OF CONTENTS (cont.)
                                                                        Page
 4.  MATHEMATICAL MODELS AND STATISTICAL TECHNIQUES	4-1

     4.1.   INTRODUCTION	4-1
     4.2.   DOSE ADDITION	4-2
     4.3.   RESPONSE ADDITION	.  4-5
     4.4.   GENERALIZED LINEAR MODELS. .	4-9
     4.5.   RESPONSE SURFACE MODELS	  4-13
     4.6.   SUMMARY OF INTERACTION DATA BASE	4-14

            4.6.1.   Description of the Mixtures Data Base Sample. .  .  4-9

     4.7.   CRITICAL ASSESSMENT EXAMPLE	4-20

            4.7.1.   Experimental Conditions 	  4-21
            4.7.2.   Discussion of Design	4-22
            4.7.3.   Discussion of Results 	  4-23

            SUMMARY	4-25

 5.  DISCUSSION AND REASSESSMENT OF THE GUIDELINES 	  5-1

     5.1.   OVERVIEW	5-1
     5.2.   COMPLEX MIXTURES                                            5-4
     5.3.   MIXTURES OF CHEMICAL CLASSES	5-9
     5.4.   SIMPLE MIXTURES, COMPONENTS AND TOXIC INTERACTIONS ....  5-12
     5.5.   MIXTURES OF CARCINOGENS WITH OTHER COMPOUNDS .......  5-15

 6.  RESEARCH NEEDS	  .  6-1

 7.  REFERENCES	7-1

APPENDIX A: Agency Data Base on Mixture Toxlclty	A-l
APPENDIX B: Diesel Exhaust Emissions and "Sufficient Similarity" ...  B-l
APPENDIX C: Analysis of the Sample Studies from
            the Interaction Data Base	C-l
APPENDIX D: References	  .  D-l

-------
No.
2-1
3-1
3-2
4-1
4-2
                       LIST OF TABLES
                          Title
Page
Summary of Interaction Data Base	2-19
Chemical and Biological Bases of Toxicant Interactions. . .  .  3-2
Mechanisms of Promotion and Co-carcinogenicity	.  .  3-13
Survey of interaction Studies Methodologies .... 	  4-15
Combined Results for CaDTPA and DMSA Inhibition of
Cd Toxicity	  4-24
                                     VII

-------

-------
                        1.   INTRODUCTION AND  BACKGROUND

    This  technical  support document  Is  a  supplement  to  the U.S.  Environ-
mental  Protection  Agency's  Guidelines  for   the  .Health  Risk Assessment  of
Chemical Mixtures published on  September  24, 1986  (U.S.  EPA,  1986a,  1987a).
This  document  was  developed 1n  response  to  a recommendation  of  the  Science
Advisory  Board  (SAB).    It discusses  available  toxlclty  and  Interaction
Information  useful   in   assessing  human  health  risks  from  mixtures.   In
addition,  applicable mathematical  models   and   statistical  techniques  are
reviewed  and  research   needs  are  identified.   The  results  of  the  above
information are discussed along with implications for the current guidelines.
1.1.   THE CHEMICAL MIXTURE GUIDELINES
    The mixtures guidelines are  Intended to guide  Agency  analysis  of infor-
mation relating to health  effects  data on  chemical  mixtures  in line with the
policies and procedures  established  in the  statutes administered by the U.S.
EPA.  They were developed  as  part  of  an interoffice  guidelines  development
program  under  the  auspices   of  the  Office  of  Health  and  Environmental
Assessment  (OHEA)  in  the Agency's Office of  Research  and Development.   They
reflect  Agency consideration  of  public  and  SAB  comments  on  the  Proposed
Guidelines  for the  Health Risk Assessment  of   Chemical  Mixtures  published
January 9, 1985 (50 FR 1170).
    These  guidelines  set forth  the  principles and  procedures to  guide U.S.
EPA  scientists in  the   conduct  of  Agency  risk  assessments,  and  to inform
Agency   decision   makers   and   the  public  about   these  procedures.    In
particular,  the guidelines  emphasize that  risk assessments will  be conducted
on   a  case-by-case  basis,  giving  full   consideration   to  all  relevant
scientific   information.    This   case-by-case  approach  means  that  Agency
                                    1-1

-------
 experts  review the  scientific  information on each  chemical  mixture and use
 the  most  scientifically  appropriate  interpretation  to  assess  risk.   The
 guidelines  also  stress  that  this  information  will  be  fully  presented in
 Agency  risk assessment  documents,  and that  Agency  scientists  will identify
 the  strengths  and weaknesses of each assessment by describing uncertainties,
 assumptions  and limitations, as  well  as   the  scientific  basis  and  rationale
 for  each assessment.
     Finally,  the guidelines  are  formulated  in  part  to bridge  gaps in  risk
 assessment  methodology  and data.   By  identifying  these gaps  and the Import-
 ance of  the missing  information to the risk assessment process, the U.S. EPA
 wishes  to  encourage  research and  analysis  that  will   lead   to   new   risk
 assessment methods and data.
    Work  on the  guidelines began  in  January  1984.   Draft  guidelines  were
 developed  by  an  Agency working  group  composed  of  expert  scientists   from
 throughout  the U.S.  EPA.   The  draft was   peer-reviewed by  expert  scientists
 in   the   fields  of   toxicology,  pharmacokinetics,  and   statistics   from
 universities,  environmental  groups,  industry, labor,  and  other  governmental
 agencies.  They were then  proposed  for public comment.   On November 9,  1984,
 the  Administrator  directed U.S.  EPA offices  to use the  proposed  guidelines
 In performing  risk assessments until final guidelines become available.
    After  the  close  of the  public comment  period,  Agency  staff  prepared
 summaries  of  the comments,  analyses  of  the major  issues  presented by  the
 commentors,  and  preliminary  Agency  responses  to  those  comments.   These
analyses were  presented to  review panels of  the  SAB.  The  guidelines  were
revised, where appropriate, consistent  with the SAB recommendations.
    The SAB  made  several comments and recommendations.  Among  the  recommen-
dations was  that the U.S.  EPA should  develop a  separate  technical  support
                                    1-2

-------
document  for  the  mixtures   guidelines.    The  SAB  pointed  out  that  the
scientific and  technical background  from which  these Guidelines must  draw
their  validity  Is  so  broad  and  varied  that  It  cannot  reasonably  be
synthesized within the framework of a  brief  set  of guidelines.   The SAB also
Identified the need  for  a technical support document  because  of the limited
knowledge  on  Interactions of  chemicals  1n biological systems  and  commented
that  progress  In  Improving  risk  assessment  will be  particularly  dependent
upon  progress  In  the  science  of  Interactions.   The  Identification  of
research needs was an  additional SAB  concern to  be addressed in this support
document.
1.2.   EXAMPLES OF THE U.S. EPA CHEMICAL MIXTURES RISK ASSESSMENT ACTIVITIES
    U.S.  EPA  personnel   were  directed   by   the  Administrator   to  use  the
guidelines   when   assessing   the   human   health  risks   from   mixtures  of
chemicals.   They  are to  be used In developing regulations under the various
statutes  for pollutants  that  are mixtures, such  as dlesel exhaust, coke oven
emissions, gasoline  and  gasoline vapors.   Another major  use Is In assessing
the  health risks at  hazardous waste  sites  where large  numbers of chemicals
are  frequently encountered.
     Many  of  the  statutes  that  govern  U.S.  EPA  activities  suggest  a  single
chemical  approach to  the regulation  of  toxic chemicals.   For  example, the
Clean Air  Act,  Clean  Water  Act and  Safe  Drinking  Water   Act  generally
instruct  the U.S. EPA to protect  public  health and the  environment through
regulation of specific  sources of  pollution or  establishment of  standards
and  allowable  levels for specific  contaminants.    In general, when developing
regulations  to implement these Acts,  the  U.S. EPA considers the human  health
hazards  of single chemicals.
     Some  statutes mention  chemical  mixtures, but  generally  in  combination
with the  term chemical   substance, as 1n "chemical  substance  or mixture."

                                     1-3

-------
These  statutory  discussions  do not provide one with a clear definition.  For
example,  the Toxic Substances Control Act  (TSCA)  defines the term  "mixture"
as  follows:
     "The  term 'mixture1 means any combination of  two  or more chemical
     substances  If  the combination does not  occur  1n nature and 1s  not,
     In  whole or  in part,  the result of a chemical  reaction; except  that
     such  term does Include any combination which occurs, 1n whole or In
     part,  as a  result  of a  chemical  reaction 1f  none  of  the chemical
     substances  comprising the combination  1s a  new chemical substance
     and  1f the combination could  have  been manufactured for commercial
     purposes  without  a  chemical   reaction  at  the  time  the  chemical
     substances comprising  the  combination were combined." (TSCA, sec. 3)

     Other  mixture-related  terms   are   also  not  clearly  defined  1n  the
statutes.    The  term  'hazardous   waste1   1s  defined  under  the  Resource
Conservation  and  Recovery  Act (RCRA)  as  a  solid  waste,  or  combination of
wastes  that   pose  a  substantial  hazard  to  human  health or  the  environment
when  Improperly   managed.   Hazardous  wastes  encountered  at  Inactive  or
abandoned  facilities  or  from emergency spill  situations  are  covered  under
provisions   of   the  Comprehensive  Emergency   Response,  Compensation  and
Liability  Act (CERCLA)  and the Superfund  Amendments and Reauthorlzatlon Act
of   1986   (SARA).   CERCLA1s   definition   of  hazardous  substance  Includes
substances and  mixtures as  defined under  a variety of  other  environmental
Acts.   For the  purposes of this technical  support  document,  definitions for
different  types  of mixtures  and  mixture interactions are presented  later in
this chapter.
    Perhaps  the  greatest  use  of  the  mixtures guidelines In the U.S.  EPA is
in  assessing human health  risk  at Superfund  sites.  These  sites  generally
contain  dozens  of chemicals  In   varying   concentrations.   The  Office  of
Emergency  and   Remedial   Response  (OERR)   utilizes   the   risk   assessment
guidelines,   and   particularly the  mixtures   and   exposure  guidelines  In
analyzing  public  health  impacts  of  remedial  alternatives  at   Superfund
                                    1-4

-------
hazardous waste sites.   OERR's  approach is outlined in  the  Stiperfund  Public
Health   Evaluation  Manual   (U.S.   EPA,  1986c).   The  manual   covers   two
elements:  baseline evaluations and  analysis  of  remedial alternatives.   OERR
is currently  revising  this manual to  ensure  that it is consistent  with the
final risk assessment guidelines.
    The  OERR  approach  for  mixtures  is  perhaps  the  most  structured of the
Agency  mixture approaches,  involving  five   specific  steps  for  determining
human health risk:
    1.   Selection of Indicator Chemicals
    2.   Estimation of Exposure Point Concentrations of Individual Chemicals
    3.   Estimation of Chemical Intakes
    4.   Toxicity Assessments
    5.   Risk Characterization for the Site

An  assumption in  this  process  is  that  there  are  no  data  on  the  specific
mixture  of concern, or a similar mixture.
    The  first step is  to select a  workable number  of  indicator chemicals.
When  the number  of chemicals found  at  a site is  determined to  be too  large
to  work  with (>10-15),  a  scoring  system   is  used to develop  a   list  of
indicator  chemicals on  which to  base  the  assessment.   The  scoring system
considers  toxicity information,  site  concentration data and environmental
mobility.   Use  of professional  judgment  is encouraged  to  add or  delete
chemicals  to  the  list.   Indicator scores  are used only for relative ranking
among the chemicals  present  and have  no meaning outside of  the context of
the   individual  chemical  selection  process.   From  the indicator  scores  a
smaller, more manageable  list of chemicals is selected.    .
     In   the second  step  of  this  process, baseline  environmental concentra-
tions  of  individual  chemicals  are  estimated  using   monitoring  data  and
                                     1-5

-------
modeling  to  estimate when  and how  human exposures  will  take  place.   The
Superfund  Exposure  Assessment Manual  describes  various  chemical  fate  and
transport models  that may be used for this step.
    The estimation  of the amount of  human  exposure to the selected contami-
nants  Is  the  next  step.  Concentrations  estimated in step two  are used to
calculate  separate  intakes  for  each  chemical  in  each environmental medium:
air, groundwater, surface water,  fish and soil.  These are summed, resulting
in  total  oral  exposure  and   total   inhalation  exposure.   Subchronlc  and
chronic durations are calculated separately.   In  some cases  intake calcula-
tions may be  based  on personal air  monitors and body burden data for exposed
Individuals.    Site-specific   considerations,   such  as  nonstandard  intake
values, are considered as appropriate.
    In  step  four, the toxicity  information  is  identified  that  will  be  used
with  results  of  the  exposure  assessment   in the   risk  characterization.
Toxicity values  for chronic and subchronic exposures  to  noncarcinogens,  and
carcinogenic  potency  factors for  potential  carcinogens are  located in avail-
able  Agency  sources.   Toxicity  data  may  be  developed  when  necessary.
Teratogenic chemicals are listed separately.
    The  final  step  involves  a  comparison between estimated exposures  and
toxicity values or  potency  factors.    For  the  noncarcinogenic  chemicals,  a
hazard  index  is  calculated (see  Section  5.4) for all  chemicals  for  each
medium of exposure.   Separate  hazard  indices,  by critical  effect, are recom-
mended when the overall  hazard index exceeds  unity.   The  mixture guidelines
suggest consideration  of all  types  of effects from a particular  chemical,
not just  the  "critical  effect,"  i.e.,  the effect  seen at the  lowest  dose.
Critical effect  information  is readily available  in U.S.  EPA  documentation,
while data on other  effects  may sometimes be  more  difficult  to  obtain.   For
                                    1-6

-------
potential carcinogens,  response addition for  Independently-acting  chemicals
at low  doses  1s the approach  recommended.   The manual  further  assumes  that
cancer risks are additive across all exposure routes.
    Following  these  five  steps,  It  Is  recommended  that  the risk  assessor
determine  the  validity  of  the  Initial  list  of  Indicator  chemicals.   In
addition, a written  summary of all  the significant  uncertainties  1s  recom-
mended as part  of  the  risk characterization  step.   Assumptions  were to have
been noted along the way  for  each step.   These public health evaluations are
used to  develop performance goals and analyses of  risks for  remedial  action
alternatives.
    Two  other  approaches  for  chemical  mixtures,  relative potency  and  toxic
equivalency  factors,  have  been  considered  and  utilized  by  U.S.   EPA  risk
assessors  and  are  discussed  In   Chapter 5   of  this  document.   Briefly,  a
relative potency method  for carcinogenic mixtures  is based on the assumption
that  the ratio  of  the  two potencies  Is  constant,  whether  It  is  based  on
comparisons between  human studies, jm  vivo  assays  or jm  vitro  assays.   The
results  of  human studies  are correlated with  those of  in \nyo  assays,  and
results  of  j_n  vivo  bioassays are  correlated with  the  results  of  jm  vitro
bioassays.   The  human   potency   of  a  poorly-studied mixture   can  then  be
estimated from its  \n vivo  (or  in vitro) potency  multiplied by the potency
ratios  of a  well-studied, similar  mixture.   The  toxic  equivalency  factor
approach has  been adopted by  U.S. EPA as an interim procedure for estimating
risks  associated  with   exposure   to  chlorinated  dioxins  and  dibenzofurans
(U.S.  EPA,  1987c).   This  method relies on in  vitro and in  vivo data  to
estimate  "toxic  equivalency  factors"  for   the   various  congeners  in  the
mixture.  These factors  then  express  the inferred toxicity or cancer risk of
poorly   studied congeners  in  terms  of  the  toxicity  of  a  well-studied
                                    1-7

-------
 congener,  and can  be used  in  an additive  model  to estimate  toxlclty of a
 mixture  of  these  congeners.
     Many of U.S.  EPA's regional  offices  are routinely  using the guidelines,
 with Superfund activities  being  the  primary application.   In  addition,  at
 least  one  region  is applying the guidelines in the NPDES permitting program,
 by  using additivity  when  the  pollutants have the  same mechanism of action.
 There  are  currently programs underway  in  the  U.S.  EPA  to Implement the risk
 assessment  guidelines in  all  appropriate  Agency  activities.   It  will take
 some time before  they are being fully applied  in all U.S. EPA operations.
 1.3.   DEFINITIONS  USED IN THIS DOCUMENT
     Consistent and  clear  terminology is critical in  the  discussion of chem-
 ical mixtures  risk  assessment.   Many different definitions  have been offered
 for  the  terms  used  with  toxicity of  chemical mixtures,  and most of these are
 discussed in the  body of  this  document.  Except  for these historical discus-
 sions, the  definitions below are  used  in this document.   These definitions
 are  oriented  toward  their  use  in   risk  assessment.   For  example,  the
 definition  of  a   mixture  actually   describes "mixed  exposures."   From  a
 toxlcologlc  standpoint,   however,  the  joint  exposures  are  similar to  the
 single exposure   (perhaps  time-varying) that  would  result  if  the  chemicals
were  physically  combined  into   a  true   chemical  mixture.   The  following
definitions are generally consistent  with those found in the literature:
Mixture:        Any  set  of  two  or more chemical  substances,  regardless  of
               their  sources, that may  jointly contribute  to  toxicity  in the
               target population.
Simple
Mixture:
A mixture containing two or  more identifiable components,  but
few  enough  that  the  mixture  toxicity  can  be  adequately
characterized by a combination of the component toxicities.
                                    1-8

-------
Complex
Mixture:
Similar
Mixtures:
Chemical
Classes:
 Interaction:
 Synergism:


 Antagonism:


 Potentiation;



 Inhibition:
A mixture containing so many components that any estimation of
its toxicity  based  on its  component  toxicities contains  too
much uncertainty and  error  to  be useful.   The chemical compo-
sition  may vary  unpredictably  over   time  or  with  different
conditions  under  which  the  mixture  is  produced.   Complex
mixture   components   may   be   generated   simultaneously   as
by-products  from a  single  source  or  process, intentionally
produced  as  a commercial product, or  may  co-exist because of
disposal  practices.   Risk  assessments of complex mixtures are
preferably  based   on  toxicity  and   exposure  data  on  the
complete mixture.  Gasoline  is an example.

Mixtures having the same components but  in slightly different
ratios,  or having most  components  in nearly  the  same ratios
with  only a  f<*w  different (more  or  fewer)  components,  and
displaying  similar  types   and degrees  of  toxicity.   Diesel
exhausts  from  different engines  are  an  example  of  similar
mixtures  (Appendix B).

Groups  of  compounds that are similar  in chemical structure and
biological  activity,  and which frequently occur  together in
the  environment, usually  because they  are  generated  by the
same  commercial process.   The composition  of these  mixtures
is  often well controlled,  so  that  the mixture can be treated
as  a  single  chemical.  Polychlorinated  biphenyls (PCBs) are
an  example.

The  circumstance in  which  exposure  to two  or more chemicals
results in a qualitatively or quantitatively  altered biolog-
 ical  response relative  to  that predicted from the actions of
the  components  administered  separately.   The multiple  chem-
 ical  exposures  may be  simultaneous  or sequential  in  time and
 the  altered  response may  be  greater  or  smaller  in magnitude
 (adapted from NRC, 1980).   For quantitative  evaluations, the
 "no-interaction"   prediction  is  based  on  dose  or  response
addition, as  appropriate.

 A response  to  a mixture  of  toxic chemicals  that is greater
 than that suggested by the  component  toxicities.

 A response to a mixture of toxic chemicals  that  is  less  than
 that suggested  by the component toxicities.

 A special case  of synergism  in  which one substance  does not
 have a toxic effect  on  a  certain  organ or  system,  but  when
 added  to another chemical   it makes  the latter much more  toxic.
                                                I
 A special case  of  antagonism in which one  substance  does not
 have a toxic effect  on  a  certain  organ or  system,  but  when
 added  to a toxic chemical  it makes  the latter less toxic.
                                     1-9

-------
 Masking:       The  situation In which  the toxic  effect  of one  chemical  1s
                not  displayed because of  functionally  competing effects from
                the  other chemical.   The most  striking  example  1s  when the
                carcinogenic  activity of  the  mixture Is  not observed  at the
                experimental  doses,  because  of  more  obvious  toxic  signs,
                particularly mortality, Induced by other toxic components.
 1.4.   OVERVIEW OF THIS DOCUMENT
     The  main body  of  this  report  discusses the  Information  available  on
 chemical  mixtures,   the mechanisms  by  which  chemicals   Interact,  and  the
 mathematical models  used to describe  toxicant  Interactions.  After a  brief
 Initial  description  of  the  terminology used  to  describe toxicant  Inter-
 actions,  Chapter  2  discusses  the  nature of  the  available Information  on
 three general categories  of mixtures:   complex mixtures,  mixtures  composed
 of  a  single class  of  chemicals  and   simple  mixtures.   This  section  1s
 Intended to Illustrate the differences between the  types of Information  that
 are available on  the  various categories  of mixtures but  Is not  Intended  to
 be a compendium of  all  available  Information on  all mixtures.    Emphasis  Is
 placed  on  the description  of the  tests  used  to assess the toxlclty of the
 mixture  as well as  the  available  methods and  feasibility  of  these methods
 for  quantitatively measuring Interactions of  the  components 1n  the mixture.
 This  chapter concludes with  discussions  of  additional  topics:   Interactions
 of  carcinogens  with  other  compounds,  some results  from  the Agency's   data
 base  on mixtures and quantitative measures of  Interactions.
    A  discussion  of  mechanisms  of  toxicant  Interactions  Is   presented 1n
 Chapter 3.  This section  discusses  the ways  In which compounds  may Interact:
 direct  chemical-chemical  reactions  that  result   1n  the  formation  of  a
 different  chemical  species  as  well  as   the  biological   bases  of  toxicant
 Interactions  such  as  effects  on  absorption,  distribution,  metabolism,  excre-
tion and  receptor  site  affinity.  This Is followed  In  Chapter 4 by a  review
of the mathematical  models  and  statistical  procedures  used  to  assess  toxic
                                    1-10

-------
Interactions, including dose  addition,  response  addition,  generalized linear
models, and response surface  models.  This  section  concludes  With a critical
review of  statistical  methods used in research  articles  that are covered 1n
the Agency's mixtures data base.
    Chapter 5 reassesses  the guidelines  in terms of  the  information summa-
rized  in   the previous chapters.   Following  the organization  of Chapter 2,
which  is  in  turn  dictated by the different types of information available on
the  various  chemical   classes,   this  chapter  separately  discusses  complex
mixtures,  similar  mixtures  and  simple  mixtures.    For-  complex  mixtures,
emphasis  remains  on  in  yjivo bioassays,  the  applicability of which can be
extended   by the  concept  of  sufficient  similarity,  as  illustrated  in
Appendix  B.   Recognizing  the highly variable nature of some  complex  mixtures
as  well  as  the difficulty and  expense  of  obtaining good in viyp.  bioassays,
the  relative  potency  method,  the  "toxic equivalency  factor"  method  and
analogous methods based  on  In  vitro  assays,  are more  strongly endorsed  than
 in   the   original  guidelines.   A  limitation  of  dose   addition   Is   also
discussed,  primarily  related  to limitations  of risk  assessment  of  single
 compounds.                              -
     This   document  concludes  with a  brief  outline  of  research  needed  to
 improve   or  validate  the risk  assessment  procedures  for  mixtures.   Because
 the reassessment of  the guidelines  relies heavily on the  use . of  In yUro
 tests, emphasis  is placed  on the validation of  such tests  using whole animal
 assays.                                                  -•   -   i
                                     1-11

-------

-------
                      2.   TYPES OF  INFORMATION AVAILABLE
2.1.   OVERVIEW
    This chapter  summarizes the  kinds of  Information  available on  various
categories   of  mixtures;   namely,  complex  mixtures,  chemical  classes  and
simple  mixtures.   Also  covered 1s  the  nature  and  utility of  Information
available on  the  Interactions  of carcinogens with  other  compounds  Including
discussions  of promotion,  cocardnogenldty, Inhibition  and masking.   The
focus  of  this chapter  1s  on  the  usefulness  as well  as the  limitations  of
available  data on mixtures;  for risk  assessment.   This  1s not  Intended  to
provide a comprehensive summary  of all  available  Information on  these topics
but  1s based  on  the  Information Included   1n the computerized data  base,
which  Is described 1n Section 2.4.3., Chapter 4 and Appendix A.
    Given  the  quality  and quantity of  the available  data on chemical Inter-
actions, few generalizations  can  be made concerning the likelihood, nature
or  magnitude of  Interactions.   Most Interactions   that  have been quantified
are  within  a  factor of  10 of  the expected activity  based  on the assumption
of  dose addition.  The limited available Information suggest;;  that  at least
some  interactions may  have thresholds  and that addltlvlty may be a plausible
assumption  at  low levels  of   environmental  exposure.   This  supposition  1s
reenforced  by  mechanistic  considerations  discussed in Chapter 3.  It must be
emphasized,  however,   that  these generalizations  are based  on  very limited
data.
     The information  available  on complex mixtures  is fundamentally different
in   design  and  focus   from that  on  simple  mixtures.   Studies on  complex
mixtures  generally  are  designed  to  characterize  the   toxic  properties  or
potency of  the  mixture  as  an  entity.   In this  respect, the  design  and
conduct of   such studies  do   not   differ  greatly from  studies on  single
                                     2-1

-------
 compounds.  As a consequence,  the  great  majority  of  the bloassays on complex
 mixtures  are  not useful  for  assessing potential   Interactions  of components
 In the mixtures.  In  some  cases, however,  sample  collection  or  concentration
 of complex mixtures prior  to  a  bloassay  may  cause changes  1n the composition
 of the  mixture, which  could  limit  the  applicability of  the  study  In  risk
 assessment.  This  factor, however,  1s not  greatly  different  from  problems
 that  can  be  encountered   In  the  preparation  and purification  of a  single
 compound prior to bloassay.
    Studies available on  simple mixtures are generally restricted to binary
 combinations  and are usually designed to measure  the  magnitude  of the Inter-
 action  among  the components  In the mixture.  The   study  design generally
 Includes  a control  group, one  or  more  groups  of subjects  exposed   to  each
 component  of  the mixture  at one or more  dose levels, and one or  more groups
 exposed  to one or more  doses  of all components at equal ratios.   The Inter-
 action  Is  generally reported   as  the  ratio  of the  observed response  to a
 response   predicted  by  the  assumption  of  dose   add1t1v1ty  (discussed 1n
 Chapter 4).
    Studies  on chemical classes are generally  similar  to those  on   complex
 mixtures.   For Instance,  most of  the  available  information on  mixtures of
 polychlorinated  blphenyls  (PCBs) comes  from bloassays on commercial mixtures
 of these  substances, and no quantitative  measures  have been attempted of the
 individual  components  as  to their  concentration  or  biological  activity.  A
 significant amount  of information  is available on individual  components of
many  complex  mixtures  and chemical  classes,  but   such  studies  are  not
directly useful In quantifying interaction.
    The restriction  to binary  mixtures  of  bloassays that attempt to quantify
mixture  Interaction,  and   the  virtual  absence  of   bloassays  on  complex
                                    2-2

-------
mixtures  or  mixture classes  that attempt  to define  such  Interactions,  Is
attributable to the  nature  of the experimental design  that  1<>  necessary for
quantifying Interactions.  Note the example given by Clayson (1984):
    "... If It was  wished  to examine the  Interactions  of  just  10 chem-
    icals  In  pairs  It  would  Involve  conducting 45  separate  bloassays
    plus  a further  10  for  the  single  chemicals.   If  1t was  deemed
    necessary  to  study  these pairs of  chemicals  In  just  5  different
    ratios  1t  would be  necessary  to undertake 255  separate bloassays.
    As there are estimated  to be  1n  excess of 25,000 chemicals  produced
    commercially  in  significant quantities,  examination even  in pairs
    becomes quite  impracticable with  about  313  million tests  1f  only
    one  ratio  Is  used  or  1.57 American  billion  tests  [sic] If  5  dif-
    ferent ratios were employed."

The difficulties  in  obtaining quantitative measures on  toxicant interactions
are exacerbated by  the  fact  that  many  of the studies on binary mixtures that
purport  to quantify  toxicant interactions  are  Improperly  designed  and the
reported  results  are  either  unlnterpretable or  are  difficult to  compare
among different studies.
    Studies on the  interactions  of carcinogens  with  other  compounds  share
many  of  the same difficulties  and  limitations as  those  discussed  above.   A
substantial body  of  data,  however,  has  accumulated  which  suggests  that some
compounds  may  markedly  modify  the  carcinogenic  potency  of  other compounds.
Although   the   early  investigations  focused  on  dermal  applications  and
enhancement of skin tumor  response, more recent studies  indicate  that such
interactions may  be relatively common  and affect cancer  induction at  other
sites.   Conversely,  some agents are known  to Inhibit  the carclnogenicity of
other  compounds.   The  inhibitory  activity of some materials can  vary as a
function  of time  of  application in  relation to the carcinogen as well as the
tumor site.
                                    2-3

-------
 2.2.    COMPLEX  MIXTURES
 2.2.1.    Overview.   Some  classes  of  chemical  mixtures,  such  as automotive
 emissions  and  coke  oven emissions, are  composed of  hundreds  of components
 produced  by  a  single  process  or  set  of  related  processes.    Some  of the
 components  may  be grouped Into similar classes while others may  not have any
 apparent  structural  or  toxlcologlc  similarity   to  other  elements  of the
 mixture.   While  toxlcologic  data may  be available  on some of  the mixture
 components  or classes  of components, the characterization of the toxidty of
 other agents  1n the mixture may  be  incomplete or nonexistent.    In addition,
 the  chemical  composition  of   such  mixtures  may  vary  over  time  or  as  a
 function  of  changes  in  conditions  (e.g.,   temperature  or  pressure)   under
 which  the  mixtures  are  generated.   For example,  it  has  been   demonstrated
 that malfunctioning  fuel injection  systems  in diesel  engine cars  can  cause
 increased mutagenicity and benzo[a]pyrene emissions  (Zweidinger, 1982).  As
 is the  case for data on Individual  toxic agents,  the quality and quantity of
 data  on  complex  mixtures  varies  markedly   among  different mixtures.   Few
 generalizations can  be made concerning  the  nature of  the available data or
 the applicability of these data for use in risk assessment.
 2.2.2.   Epldemiologic  Studies.  In a  few  Instances, human  exposures  to
 complex  mixtures  have  been sufficiently  high that  direct  human  data are
 available for quantifying risks from  exposure to the  mixtures  or  processes
 generating  the  mixtures.   This has most  often  been  the case  for  mixtures
 that  Induced cancer.   For  instance,  a  substantial   body of  epidemiologic
 literature  is available on the  carcinogenic  potency  of cigarette  smoke and
of coke  oven  emissions.  Such  epidemiologic  investigations,  while sometimes
allowing  for  quantifying of  risk   from exposure  to  the complex  mixture,
seldom  provide   information on  the  nature,   magnitude  or  significance  of
                                    2-4

-------
Interactions  among   the   components   1n   the  mixture.    Some  Interactions
Involving  exposure  to  complex mixtures  that  have  been  quantified  Include
those  between cigarette  smoke  and  asbestos   (Hammond  and Sellkoff,  1973;
Hammond et al., 1979; and  Sellkoff  et al.,  1968),  cigarette smoke and radia-
tion exposure (Lundln et  al.,  1969), as well  as cigarette smoke and  vitamin
A  (Dayal,  1980).   Even  these  examples,  however,  which are  the  best  studied
examples providing human data  on  Interactions  Involving  complex  mixtures,  do
not quantify  Interactions among components  In  the  complex  mixture but rather
measure Interactions between the complex mixture and another agent.
    In 1981,  a WHO  committee on health effects of  combined exposures 1n the
work,environment  concluded  the following:   "The dearth of  sound  epldemlolog-
1cal studies  to  date and  the  potential Importance of at  least  some  of the
possible Interactions between  occupational  and nonoccupatlonal environmental
factors attest  to the need  for  more work  1n  this  field"   (WHO,  1981).   The
more recent  literature  (e.g.,  Kopfler  and Craun,  1986;  WHO,  1983)  has not
substantially  Improved  the  prospect  of  developing  human  data  on  complex
mixtures that will  be useful  In  quantifying  component  Interactions.   Given
the difficulties  1n  assessing  and designing studies  to  measure  Interactions
1n  simple  binary mixtures  [as discussed  In  general  1n  Section  2.4.1.  and
discussed  specifically  In  terms  of  epldemlologlc studies  by  Andelman  and
Barnett (1986)],  human  data on  complex mixtures  are likely  to  remain  most
useful  for risk assessments on  the  complex  mixture  Itself  but  will seldom If
ever  be adequate for  the  quantitative  assessment  of  Interactions  among
components within the mixture.                                 i
2.2.3.    Whole  Animal  Bloassays.    For   most  groups   of  highly   complex
mixtures,  data on whole  animal bloassays  that are directly useful for  risk
assessment are not available.   Lewtas (1985), for  example,  has  reviewed the
                                    2-5

-------
 available  data  on  combustion  emissions  from  diesel  engines,   gasoline
 engines,  and  energy  combustion  sources   (wood  stoves,  oil  furnaces,  and
 utility   power  plants).   For  the   gasoline  and  diesel  engines,   the  most
 comprehensive  jm  vivo  data  are  from mouse  skin, tumor  initiation  studies,
 which  are usually not directly  used in  risk  assessment to estimate  carcino-
 genic  potency  in  humans.   While several  iji  vivo studies  have examined the
 carcinogenic   and  systemic  effects  of  diesel   exhaust,  the  data  base,
 Including epidemiologic  data, in general,  is  extremely limited (NAS, 1981).
 For  the  energy combustion  sources, no  in vivo  studies  are  available.   A
 large  body  of data, however,  is  available  on  these  and other mixtures using
 a  variety  of  in  vitro  test  systems.   This  information  is  discussed  in
 Section  2.2.4.  below, and  the potential  use  of these  data  in quantitative
 risk assessment of mixtures Is discussed in Chapter 5.
    Although  data  from  animal  studies  are  available .for  the few complex
 mixtures  that  have been identified  as  human carcinogens,  long-term in  vivo
 animal bloassays  on complex mixtures have  tended  to  follow rather  than lead
 ep1dem1olog1c  investigations  and have  focussed  on complex mixtures  such  as
 polycycllc  aromatic  hydrocarbons  (PAH),   coke   oven  emission,  and  diesel
 exhaust  (as discussed  in  Appendix  B)  for which data  on  human effects  or
 human  exposures  suggested  a  potential  hazard.   The   paucity  of  whole animal
 bloassay  data  on complex  mixtures  is  illustrated  by the   compilation  of
 cancer  risk  assessments   currently on   the   U.S.   EPA's  Integrated  Risk
 Information System  (IRIS).  Of the  95 risk  assessments currently on IRIS one
 is  for a technical  grade  mixture  of  hexachlorocyclohexane isomers,  one  is
 for  a  binary  mixture of  hexachloro-p-dioxins and  one is  for mixtures  of
xylene  Isomers.   Only two  assessments, nickel  refinery dust  and  creosote,
are for  complex mixtures.   The  assessment  of  one of  these mixtures, nickel
                                    2-6

-------
refinery dust,  Is  based on  epidemiologic data  rather  than animal  bioassay
                                                               !
data  (U.S.  EPA, .19875).   Similarly, although  the International Agency  for
Research on  Cancer  has  identified several industrial processes  that Involve
exposure to  complex  mixtures; and which  are  causally 'associated  with  cancer
in  humans  based  on epidemiologic  studies,   no  complex  mixtures have  been
designated as  carcinogens  based  solely  on the  results  of  animal  bioassays
(IARC, 1982).                                                  i
    As is  the  case  for  epidemlologic investigations, long-term  whole  animal
                                                               i
bioassays  on complex mixtures  can  be  useful  for risk  assessments on  the
complex mixture  itself but are  not,  and  from  a  practical  perspective cannot,
be  designed  for  the  quantitative measurement  of  interactions  among  compo-
nents  within  the  mixtures.  The  practical   difficulties  In  making  such
measurements for  complex mixtures are  an extension  of  those discussed  for
binary mixtures  In  Section  2.4.1.   In  addition, because  of  the variability
of complex mixtures over time or  with  different  conditions in the generation
of the mixture,  the few  bioassays that  are available  on  complex  mixtures are
                                                               i
not necessarily  applicable to all exposures to  the complex mixture.   This is
illustrated  in Appendix B for diesel  exhaust.
    Several  short-term jn.  vivo  assays  for carcinogenic activity  such  as the
mouse  skin  initiation/promotion assay  (Pereira, 1982a; Slaga  et al.,  1982),
rat  liver  focus bioassay (Herren-Freund and  Pereira, 1986;  Pereira, 1982b),
and  strain  A mouse  lung tumor  bioassay  (Haronpot  et al.,  1986;  Stoner  and
Shimkin,  1982)  have been  developed  for assessing  the  effects  of  mixtures.
Such  studies are normally  not used as  the sole  basis  for a quantitative risk
assessment  because  of  the  relatively  short periods  of  exposure  and  the
endpoints  that  are  measured.   Nonetheless,  because these  studies can  be
conducted more  rapidly and  less expensively  than standard chronic bioassays,
                                    2-7

-------
 they   can  be   applied  In  qualitative   or   quantitative  assessments  of
 Interactions.   Such  short-term \jn  vivo  tests more  closely approximate the
 chronic  in vivo assays  that are normally  used  In  risk assessments and thus
 may   have  more  Intuitive  appeal   than   _1n.  vitro  assays.   Nonetheless,
 comparative  analyses between the  results  of  such  short-term In vivo assays
 with  other short-term assays (Perelra and  Stoner, 1985) or  long-term  in vivo
 bloassays  (Herren-Freund  and Perelra, 1986) do not clearly  Indicate the such
 assays  are superior  to  some of  the in vitro assays  discussed below.   The
 short-term in vivo  assays   that  have been  developed  to date  focus  only  on
 sceenlng   tests  for  carcinogenic  activity.   Research articles  describing
 comparable tests  for  measuring  Interactions  1n  the  Induction  of   chronic
 toxic effects have not  been  located.
 2.2.4.   In. vitro Studies and Other  Screening Tests.   Certain aspects of the
 toxlclty  of  complex  environmental  mixtures have  been  evaluated extensively
 using In  vitro  assays and other  screening  tests.   Four types of assays have
 been  most often  used:   mlcroblal  mutagenldty,  cell  culture, embryo  bio-
 assays and plant cytogenetlcs.   The  endpolnts assessed In  these  assays are
 one  or  more  of  genotoxidty,  cytotoxlclty,  embryotoxlclty   and  Impaired
 development.  Although  the utility of many  of these assays  1n quantitatively
 or  qualitatively  assessing  the  in  vivo  biological  activity  of  single
 compounds  or  complex  mixtures  has  not   been  extensively  validated  (as
 discussed  1n  Section  5.1.), these  j.n. vitro  assays  are currently  the  only
 practical  approach  to  obtaining detailed  Information  on the  biological
 activity   of   complex   mixtures,   particularly   In   site-specific   and
 process-specific assessments.
    The Salmonella  hlstldlne auxotroph reversion  assay (Ames  et  al., 1975)
 has been   the most  widely used procedure   for  detection of  mutagenldty  of
complex  mixtures.    Numerous  environmental  mixtures,  as  entitles  or  after

                                    2-8

-------
 fractlonatlon,   have  been  tested  In  this  assay:   coal-llqulflcatlon and
 gasification products (Epler  et al.,  1978; Rao et  al.,  1980;  Schoeny  et al.,
 1981;  Houk  and  Claxton,  1986),  automotive  and dlesel  exhaust  (Hulslngh et
 al.,  1978;  Claxton and  Kohan,  1980), crude  shale  oil  (Epler et al., 1978),
 drinking  water  (Chrlswell  et  al.,  1978; Loper  and Lang,  1978),  cigarette
 smoke  (Kourl et al., 1978),  Industrial  effluents  (Commoner !et  al.,  1978;
 Douglas  et al.,  1983; HcGeorge et  al.,  1984),  urban ambient air partlculate
 and  extracts (Commoner  et al., 1978;  Butler  et al., 1984), sludge (Houk and
 Claxton,  1986)   and  waste-amended  soil  (Donnelly  et al.,  1983).   Mutagenlc
 activity   of mixtures  has  also  been  assessed  In  a  forward  mutation  In
 Salmonella  typhlmurlum   using  8-azaguan1ne   resistance   for   selection.
                                                               I
 Automotive exhaust (Claxton  and  Kohan,   1980),  oil shale and  water   samples
 (Whong  et  al.,   1983) and  coal  liquefaction  products  (Schoeny  et  al.,  1986)
 have produced positive results  in this assay.
    Fractlonatlon,  or separation of  the mixture  Into  chemically-related  or
 distinct constituents, has been utilized  to  define constituents  In  a  mixture
 more clearly and to determine which  compounds  are  responsible  for  mutagenlc
 activity.    Fractlonatlon  procedures   have  also  been   used  to  concentrate
 materials  and  to  reduce  toxlclty of whole  mixtures,  thus making  them  more
 amenable  to  assay.   Extraction  methods   (e.g.,  acid/base,  polar/nonpolar),
 however, may lead  to chemical  reactions  that could alter  the  components  of
 the mixture, thereby affecting the toxlclty.
    The application of reconstruction assays  can be useful In  assessing the
 effects of  fractlonatlon  procedures or In uncovering Interactions  among the
 fractions.    Thllly et  al.  (1983),  for  example,   Identified  the relative
abundance of constituents  In  partlculates  from  kerosene  combustion  (kerosene
 soot).   The  mutagenlc contribution  of the  14  most  abundant compounds was
                                    2-9

-------
 determined  1n the  Salmonella  forward mutation assay.  When  these chemicals
 were  combined  1n  appropriate  proportions  to approximate the "pure soot," the
 mutagenldty  of  the  reconstituted  kerosene  soot  was  equivalent  to  the
 original   soot   extract,  demonstrating  the   concentration   dependence  of
 mutagenldty  for  the mixture, that  Is,  addltlvlty.   By contrast,  a  similar
 study  of  fractionated coal hydrogenation materials  in which  the  sum  of the
 mutagenic activities  of  organic  extracts  was  compared  with  the activity from
 the whole  sample  and with  a  reconstituted  whole sample, Indicated a  depar-
 ture from addltivity  for some mixtures (Schoeny et al., 1986).
    DNA  repair-deficient  strains   of  Bacillus  subtil Is  (Donnelly  et  al.,
 1983)  and  Escherichia coll (Rossman  et  al.,  1984)  have been  used  to  detect
 alterations  in  DNA  induced   by wood-preserving  waste  and  by  urban  air
 partlculates,  respectively.   Assays  for  reverse  mutation  in  Saccharomyces
 cerevislae  (yeast)  (Douglas  et al.,  1983)  and assays  detecting  dominant  or
 recessive  lethals  in Paramecium tetraurelia  (SmUh-Sonneborn et  al.,  1983)
 have been less frequently used.
    Embryo  culture  assays  have  been  developed to examine  potential embryo-
 lethality,  malformation  and   growth/developmental   alterations   induced  by
 Individual  substances  and  complex  environmental mixtures.   Dumont  et  al.
 (1983)  developed  the Frog  Embryo  Teratogeniclty Assay:   Xenopus or  FETAX.
Coal  and  shale-derived synfuels   (Dumont  et  al.,   1983)   and  mine   water
discharge (Dawson et  al.,  1985)  have caused one or more of embryolethality,
gross  malformation   or   embryotoxicHy  in  frog  embryos  exposed  in   vitro.
Other embryo assays using the rat (Klein  et al.,  1983)  and  sea urchin  (Hose,
1985)  are being developed and  validated for  application to mixtures.
    Environmental  mixtures  have  been  evaluated in cell culture assays  as  to
their potential mutagenldty at specific  loci, as well  as for  their capacity
                                    2-10

-------
                                                               I
 to  Induce  sister  chromatid exchange  or chromosomal  aberrations.   Cytotox-
 iclty  has  also been evaluated as measured  by effects on cellular growth and
 division,  and  on  morphological,  cytochemical  and  biochemical  alterations.
 For  example,  Chinese hamster ovary (CHO) cells  have been used in determina-
 tions  of  the ability  of  complex environmental mixtures  to  produce cytotox-
 icity  and  mutagenicity at  the hypoxanthine  guanine  phosphoribosyl  trans-
 ferase locus  (Hsie  et  al.,  1978).   Subfractions  of crude synthetic oil (Hsie
 et  al.,  1978), coal gasification  condensate tar  (Cunningham  et  al.,  1984),
 oil  and coal  fly ash (Chescheir  et  al., 1980; Li et al., 1983), textile mill
 effluents  (Waters  et  al.,  1978),  diesel engine  exhaust (Chescheir et  al.,
 1980;  Li  et al.,  1983), retort  process  water  from crude shale oil (Strnlste
 et  al.,  1983), as  well as  coke oven  emissions,  roofing tar  and  cigarette
 smoke  condensate  (Li  et  al.,  1983),  have  produced positive  cytotoxic  and
 mutagenic responses  in this assay.
                                                               I
    Other  endpoirits  measured in CHO  cells  in  response to  complex  environ-
 mental   mixtures   have   included   mutagenicity   at   the   Naf~Kf-dependent
 ATPase  locus  using  ouabain  resistance  for  selection,  sister  chromatid
 exchange  and  chromosomal  aberrations.   Diesel  exhaust  particle  extract  (Li
 et al., 1983)  and  pulp and paper mill  effluents  (Douglas et al.,  1983)  were
 genotoxic In these assays.
    Cell   types such as  alveolar  macrophages  or  epithelial  tissue,  which
would  be  directly   exposed  to  environmental  agents,  have  been  used  to
evaluate  toxicity of mixtures.   In the  pulmonary  alveolar macrophage assays,
viability,  phagocytic ability, specific  enzyme activities and  ATP  levels  are
the  endpolnts  most  often evaluated.   The toxicity  of  jn vitro exposure  to
various fly ash particles (Waters et al., 1978;  Aranyi  et al., 1980; Humford
and  Lewtas,  1983),  liquid textile  mill  effluents  (Waters et  al.,  1978)  and
                                    2-11

-------
smelter dust (Aranyi et al.,  1980)  was  assessed using rabbit alveolar macro-
phages.   Fisher  et  al.  (1983)   used  bovine  alveolar  macrophages  for  the
analysis of coal and  oil  fly ash.  Unscheduled  DNA  synthesis,  an indication
of DNA  damage, was  induced in organ cultures  of hamster trachea! epithelium
exposed   to  coal-fired  fly  ash,  diesel  fuel   exhaust  and cigarette  smoke
condensate (Schiff et al., 1983).
    Other  less  frequently  used   cell  culture  assays  have  been applied  to
environmental  mixtures.   The BALB/c-313  cell  transformation  assay  showed
enhanced toxicity of drinking water organic  concentrate  fractions (Loper and
Lang,  1978).   The   rainbow  trout  gonadal   tissue   (RTG-2)  assay,  wherein
anaphase aberrations resulting from jji  vitro exposure are determined, showed
genotoxiclty of marine sediment samples (Kocan and Powell, 1984).
    Ijn  vitro  plant  assays have  been  used to  evaluate  various  environmental
mixtures.  Plants,  like animals, are  eukaryotlc organisms and may  have the
ability  to  convert  chemical  compounds  to biologically  active  species.   The
most widely  used  higher plant for testing genetic  toxicity has been Trades-
cantia.   Tradescantia  plant  systems   are  especially  useful   for  in  situ
environmental air exposure and the  testing of gaseous agents.   The induction
of somatic  mutation at  a particular  locus  is  measured  in  the Tradescantia
stamen  hair  system  as  a phenotypic change in pigmentation in  mature flowers
following exposure  of  the developing  floral tissue  (Schairer  et  al.,  1978,
1983).  Tradescantia exposed  jn  situ for  10  days to  ambient air pollution 1n
several cities in the  United  States  have shown positive results for mutagen-
Iclty  in this  assay  (Schairer  et al.,  1978,  1983).   In  the Tradescantia
mlcronucleus  test,  early prophase  I meiotic pollen mother  cells  of Trades-
cantia  plant  cuttings  are exposed and  the frequency of  mlcronuclei  (chromo-
somal  fragments)  determined  in  the  tetrads following  meiosis (Ma  et  al.,
1980, 1983;  Plewa,  1984).  Sewage sludge  from  several  cities   (Plewa, 1984),

                                    2-12

-------
shallow well  water  samples  and deep  well water  containing 226Ra, as  well
as  combustion  products   of   diesel  and   dlesel/soybean  oil  fueled  engine
exhaust fumes (Ma et al., 1980, 1983, 1984} were genotoxic In this assay.
    The barley  root  tip  cytogenetic system involves  scoring barley (Hordeum
vulqare)  root tip  cells in  germinating  seeds  at  anaphase for  detectable
aberrations following treatment  of  the seed (Constantin et  al.,  1980).   Fly
ash-aqueous  extracts  and  arsenic-contaminated  groundwater  have  produced
positive  results  1n  this  assay.    The  Arabidopsis  thaliana assay  (Redei,
1980) and the Soybean Spot Test  using  Glyclne  max {Vig, 1980),  while not yet
applied to complex environmental mixtures, detect  phenotyplc alterations in
the embryo  or mature  plant  indicative of  mutational  events
result of exposure of the seed.
2.3.   MIXTURES OF CHEMICAL CLASSES
    A mixture  of  a class  of  chemicals refers  to  a group of
are structurally  and biologically  similar  and which usually
occurring as a
compounds  that
occur  together
in the environment  because  they are produced by  the  same process.   Mixtures
of chemical classes, like the complex mixtures, may  contain tens  or hundreds
of  components.   Also,  as  with  the  complex  mixtures,   the  composition  of
                                                              i
similar mixtures may vary over  time because  of  environmental  partitioning or
different  conditions  of generation, use  and release.  Examples  of mixtures
of  chemical  classes  include  the chlorinated  dioxins, chlorinated dibenzo-
furans, chlorinated naphthalenes and chlorinated biphenyls.
    As with  the complex mixtures,  the  amount  of data available  on mixtures
of  chemical  classes  varies markedly,  but  the  types of  data are similar:
human data (generally data  from accidental exposures), animal  bloassay data,
and data from in vitro  assays.   The relative amounts  of  the various types of
data  are  dependent on  the  levels and   nature  of  human exposure  to  the
                                    2-13

-------
 mixtures,  the perceived  levels  of hazards associated  with exposure to  each
 mixture,  and certain  practical  considerations  that are associated with  some
 of  the  more  common  simple mixtures.
     For Instance,  PCBs have been  commercially  produced as several groups of
 similar mixtures  varying  In  the average  degree  (percent  by  weight)  of
 chlorlnatlon (U.S.   EPA,  1984).   For  the more  commercially  significant PCB
 mixtures,  such  as  Aroclor 1254 (54%  Cl)  and  1262  (62%  Cl),  whole animal
 bloassays  for carcinogenic  effects are  available on  the mixture  and  have
 been  used directly to  estimate  cancer  potency.  The  chlorinated dioxins,
 however,  have never been  used  as a commercial  product  but have occurred as
 contaminants  in  commercial products or as  combustion  by-products (U.S.  EPA,
 1985).  Consequently,  there  is  no "typical"  dioxin mixture, and whole animal
 bloassays  have been-conducted  only on certain individual dioxins  (such as
 2,3,7,8-tetrachlorod1benzo-p_-dioxin)  or  on  simple  mixtures  of  hexachlori-
 nated dioxins,  which  are difficult to separate  chemically.  Given technical
 problems  associated  with the synthesis and purification  of large quantities
 of  chlorinated  dioxins as well  as the undesirability  of  synthesizing  large
 quantities of them,  it  is  not  likely that many  more  whole animal bloassays
will  or should  be   conducted  on  this  class  of chemicals.   Much  research,
 however, has  been and  continues to be  conducted  using Vn  vitro bioassays to
 facilitate a better  understanding of  structure-activity  relationships  and
mechanisms of action of  chlorinated dioxins as  well  as many other classes of
simple  mixtures.  These  data  have been recently  reviewed  (Kociba  and Cabey,
1985) and  their application  to risk assessment  is an  active topic In  the
literature (U.S. EPA,  1987c; Eadon  et al.,  1986)  and  is discussed further in
Chapter  5.
                                    2-14

-------
 2.4.    SIMPLE  MIXTURES,  COMPONENTS  AND  TOXIC  INTERACTIONS
 2.4.1.    Overview.   The great  majority  of  studies  In  which  attempts  have
 been  made  to  assess  toxic  Interactions  quantitatively  have  used simple
 binary  mixtures.   Only a few studies  (Gullino  et a!., 1956; Hermens et a!.,
 1985a,b)  have  used  mixtures   of   over  10 compounds.   In  such  studies  of
 relatively  simple mixtures,,  approaches  to  the analysis  of  toxicant inter-
 actions  used  by  most  toxicologists  have been based on the assumption of dose
 addition  using simple experimental  designs involving a control group, groups
 exposed  separately  to each compound  at multiple doses  so  that the LOC s or
                                                                       50
 ED50i!  can  be  estiimated»  and   groups   exposed  at  multiple   doses  to  one
 mixture  of  all compounds  in  fixed  proportions.   The degree and nature of the
 toxic interaction  is  then expressed as  the ratio (or some transformation of
 the  ratio)  of  the observed  LD5Q  or  ED5Q  of  the mixture  to the  LQ  Q  or
 ED50  exPected  from tne  assumption  of  dose  addition.  This can  be referred
 to as the ratio of interaction  (R.I.) and expressed as
                          R.I. = ED5Q(Obs)/ED50(Exp)       Equation 2-1
A  ratio  of  Interaction  greater than  one is  associated with  antagonism  in
 that  the observed ED5Q  is greater than  expected  (i.e.,  less  toxic)  based
on  the  assumption of  dose  additlvity and the  measured  ED5Qs ' of  the compo-
nents in  the mixture.   Conversely, a ratio  of less than one  is  associated
with synerglsm.
                                                               ,!
    As discussed  by  Berenbaum  (1981, 1985a)  and detailed in  Chapter 4,  the
difficulty  in  demonstrating  significant  interaction  based on  studies  using
                                                               i
single ratios of interaction is primarily  one of experimental  design.   Since
the ratio of  interaction is  dependent  on the  proportions of  the  components
                                                               i
in  the  mixture,  a  test  has  the  best   chance  of  demonstrating  significant
interaction  if the mixture  giving   maximum interaction  is  selected.  If  the
                                    2-15

-------
combination  of toxicants  being  tested 1s  assumed  to evidence  a  pattern of
symmetric  Interaction,  a  mixture  of equltoxlc doses would be the best selec-
tion.   Even  with  this  simplifying  and  not  necessarily  valid assumption,
however, tests  based  on single ratios of  Interaction will not yield signifi-
cant  results  unless the magnitude  of  the  interaction  is  substantial and the
experimental variability  is minimal.
2.4.2.   Measurements  of  Toxicant  Interactions.   Keplinger  and  Deichmann
(1967)  used  the ratio of  interaction  to measure  the joint action of various
pesticides in mice.   In this  study, only  one mixture of each combination was
used, and  significant  interaction  was arbitrarily  defined as ratios of <0.57
for  synergism and  >1.75  for  antagonism.    Smyth et al.   (1969,  1970}  used  a
slightly modified  expression  of  the  ratio  of  interaction,  which resulted in
estimates  that  resembled  the shape  of  a   normal  distribution.   Significant
interaction  was  then  defined as  those  ratios   of  observed   to  predicted
LD5Qs  in  rats   that  were  beyond  1.96  standard  deviations  from   the  mean
ratio.  In studies  on the joint action of  pesticides  in  houseflies, Sun and
Johnson  (1960)  defined  the  cotoxicity coefficient as the  ratio  of  inter-
action  multiplied  by  100.   Significant  interaction was  estimated  in  this
study by taking repeated  measurements and  determining if  the  95% confidence
Interval of  the cotoxicity coefficients  included 100.  They reported  a  high
degree  of  synergism   for  a  mixture  of  8-(dimethoxyphosphinyloxy)  N,N-di-
methylcrotonamide and  sesamex  while methylparathion and  sesamex  were antago-
nistic.  Wolfenbarger   (1973)   used cotoxicity coefficients  to  estimate  the
joint  action  of  toxaphene-DDT   mixtures   in  insects.   Although  different
combinations   of  each  mixture  were  used   and  cotoxicity coefficients  were
derived for  each  combination, no  attempt  was made to derive  coefficents of
interaction,  as  defined and  discussed in  Section  4.3.,  which could  be  used
                                    2-16

-------
 to  characterize  the direction  and  magnitude of  the  Interaction  for  all
 combinations of the mixtures.
    All  of  the above approaches are  severely  limited  by their reliance on a
 single  Interactive  ratio.   As  discussed  by  Hewlett   (1969),  the  ratio  of
 interaction  is  characteristic  only  of a particular  combination of compounds.
 In  other words,  the  estimated value  of  the ratio of  interaction  will vary
 depending on the proportions of the toxicants present In the mixture.
    Another  limitation  in  the  use of  ratios  of Interaction is encountered in
 attempts  to demonstrate  statistical   significance.  The method used  by  Sun
 and  Johnson (I960), based  on  repeated measurements  of the  ratio  of inter-
 action,  may be the  least  objectionable;  however, because of  the  dependence
 of  the  ratio  of  interaction  on the  proportion  of the  components  in  the
 mixture,  the  estimate  of  interaction  is  valid  only for  the  particular
mixture  tested and  has  no  merit  in  assessing the  overall interaction charac-
 teristic  of the combination being  tested.  This  limitation  may be  particu-
 larly  misleading  for those compounds that  evidence asymmetric  interaction,
as discussed in  Chapter 4.   The approach  adopted by Kepiinger  and  Deichmann
 (1967)  1s  totally arbitrary and  makes no  attempt  to   establish a  criterion
 for  statistical  significance.   The method of Smyth et al.   (1969,  1970)  is
based  on  arbitrary selection  of test chemicals  that influence  the  criteria
 for interaction.  The other methods that use 95%  confidence  intervals of  the
LD5Qs  of  the mixture and  individual   components  (Marking  and Dawson,  1975;
Wolfenbarger,  1973)  are overly  sensitive to  both  endogenous and  exogenous
variance.   Marking  and   Dawson  (1975)   recognized   the   difficulty with
exogenous variance in stating that  "well planned  toxicity  tests  which result
 in  narrow confidence  Intervals  are  most  useful  1n  the  assignment  of  the
effects  of  chemical  mixtures."  If  endogenous variation  is  high  (that  is,
                                    2-17

-------
the  slope  of  the  log  dose-probH  response  line  is  low),  however,  even
well-designed   toxldty   tests   may  yield  95%  confidence  Intervals  that
preclude the detection of Interaction.
2.4.3.   The U.S.  EPA  Data  Base  on Toxic Interactions.  The Interaction data
base  was  constructed  to  determine the  nature  and extent  of  Information  on
component Interactions.  Host of  the  entries  are for  studies on two chemical
Interactions,  but a  few consider  combinations of two  mixtures.  The  data
base  currently covers  roughly 600  chemicals.   Host of  the studies  evaluate
the  Interactions  based on  mortality  following  acute  exposure.  Host  of  the
studies investigate  the  influence  of  one chemical  on  the toxlcity of another
(I.e.,  potentiation  or  inhibition),  where  the first is  administered at  a
nontoxic dose  (Table  2-1).   The statistical methods  used  in  these  studies
are  discussed  In  Chapter  4.    Details  of  the   data  base  are  given  in
Appendix A.
2.5.   INTERACTIONS OF CARCINOGENS WITH OTHER COHPOUNDS
2.5.1.   Promoters  and  Cocardnogens.   Only  13  years  after  Bauer   (1928)
proposed the  somatic  cell  mutation theory of  cancer, Rous and  Kidd  (1941)
and Berenblum  (1941a)  proposed that some forms  of  chemically induced  cancers
Involved a  two step  process.  Berenblum1s (1941b)  report  on the  enhancement
of benzo(a)pyrene Induced  carcinogenicity by  extracts of  Croton  tiqlium.  a
complex  mixture,   was   the   first   example  of   one  chemical  enhancing  the
carcinogenic activity  of  another.   With  improvements  in  chemical  techniques
for fractlonation  and  isolation,  the  active components of  Croton  resin  have
been identified (Hecker,  1968;  Van Duuren, 1969).  Since  1941, over  30  such
agents, Including all  extracts  or derivatives  from   Croton  oil,  have  been
Identified  (Van Duuren,  1976;  Pitot and Sirica, 1980; Fujiki  et  al.,  1979).
The best known of  these is TPA (12-o-tetradecanoyl-phorbol-13-acetate).
                                    2-18

-------
                                   TABLE 2-1



                       Summary of Interaction Data Base
Category
Duration



Interaction








Type
acute
subchronlc
chronic
lifetime
synergism
potentiation
antagonism
inhibition
additivity
no apparent interaction
masking
chemical synergism
unable to assess
Percentage
of Studies*
73
11
8.4
0.29
I
2.8
29
1.7
31
3.7
I 25
; 0.59
0.13
5.6
*Representing a total of 587 chemicals
                                    2-19

-------
     The extensive  and  complex literature on promotion  and  cocardnogeniclty
 has been  recently reviewed  by Bohrman  (1983),  Clayson  (1984),  Driver  and
 McLean (1986).  For  purposes  of  this document, promoters will  be defined  as
 agents which, when applied after but not before an  initiator,  act to  enhance
 the carcinogenicity of the  initiating  agent.   Cocarcinogens are taken  to  be
 agents that  may  enhance  carcinogenicity when  applied  before or at the  same
 time as the  initiator.  The  definitions  of cocarcinogens and  promoters  are,
 thus,  operational  and  depend  largely  on  the  design of  the experiments  In
 which  they are found to have an effect.  It is likely  that  cocarcinogens and
 promoters  may  have some  mechanisms  of  action in  common,  as  well  as  some
 unique modes  of enhancing  a  carcinogenic  response.
     As discussed  by Van Duuren (1976),  all  promoters can probably display  at
 least  some tumorigenlc activity  in  the absence of  a known  initiator.  This
 Is  to  be  expected "... if  one assumes  that  in  any group  of animals  there
 will be some that  have latent  tumor  cells,  either by earlier exposure to  an
 external  initiating agent  or by spontaneous conversion  of  normal  cells into
 latent tumor  cells...  If  this explanation  is  accepted, the question about
 'pure'  promoting  agents should be obsolete."   While  this  may be true within
 the  context  of  interpreting the  results  of an  initiation-promotion  assay,
 the  distinction between promoters  and  initiators  could have  a significant
 Impact on  risk  assessment.   Because  it  is generally accepted that Initiation
 1s  a  nonthreshold   (genotoxic)  phenomenon  and  promotion  is  probably  a
 threshold  (epigenetic)  phenomenon,  the distinction  between  "pure"  promoters
and  those  promoters that may also be weak  initiators may be crucial  to the
selection of  high-  to low-dose  extrapolation models,  as  discussed  further by
Clayton (1984) and  in Section 3.3.6.  and 5.5.
    While most  of  the  studies  on  chemical  promotion involve dermal or  oral
exposure,  Nettesheim et  al.   (1981)  documented   several  factors enhancing

                                    2-20

-------
carcinogenesis in the  respiratory  tract;  at least some of  these  were attri-
butable  to  initiation-promotion   processes   (e.g.,  promotion  of  urethane-
induced  pulmonary  tumors  in  mice by  phorbol esters  or  butylated  hydrqxy-
toluene).   In  addition  to   the   skin  and  respiratory  tract,  Initiation-
promotion has also been  observed  In  the  liver (2-AAF- or  DMN-phenobarbital),
bladder  (N-methyl-N-nitrosourea-sodium saccharin  or  cyclamate), gastrointes-
tinal  tract  (DMBA-TPA, dimethylhydrazlne-phenobarbital),  and  mammary glands
(DMBA-estrogens  or  phorbol esters), as  detailed  in  an extensive  review  by
Bohrman  (1983).   Some  epidemiologic  data  are  suggestive  of a  two-stage
Initiation-promotion  process   in   humans,  although  the  evidence   is  scanty
(Hakaraa,  1971;   Berenblum,  1979).   Thus,  promotion  may  be  a very  common
phenomenon that occurs among many chemicals and affects most species.
2.5.2,.   Inhibitors   and  Masking.   Some   compounds,   such   as   butylated
hydroxytoluene (BHT) and  other  antloxidants,  have been  shown to decrease the
development  of  tumors when administered  before  the  administration of known
carcinogens   (Ito  et  al., 1985;  King  and  McCay,   1983;   NRC,  1980).   In
addition,  a   compound  that causes  an  increase  in the mortality  rate could
result  in  a  decreased cumulative Incidence of  late  appearing tumors because
of competing  risks.
     In  the  case of compounds  that  apparently decrease  carcinogenic response
through  a "protective"  mechanism,  the  nature  of the  protective mechanisms
and  the dose-response  relationship  of the  protective  effect  have  not been
clearly  defined.   In   addition,  some  of  the   compounds  that  display  a
"protective"  effect  under one set  of circumstances may, in fact, enhance the
carcinogenic  response  under  different  conditions of exposure.  For  Instance,
BHT  reduces   carcinogenic responses when  administered  before  some carcino-
gens  but  enhances  carcinogenic  responses  when  administered after  other
                                     2-21

-------
 carcinogens.  The protective effect Is attributed  to  the  antloxldant  proper-
 ties of  BHT and the enhancement  to  production of a  metabolite of BHT  with
 promoting activity.  Any  attempt to  predict  the  Interaction  of  BHT with a
 specific  carcinogen  Is  complicated   because  BHT  Is   known  to  Inhibit  the
 mutagenlc activity of  benzo[a]pyrene  but  to  enhance  the mutagenlc activity
 of  aflatoxin Bl 1n the  Salmonella reverse mutation assay (Malklnson, 1983).
 The sequence  of exposure  Is  an  Important variable   for  other compounds as
 well.   Both phenobarbltal  and  cloflbrate, for example, enhance carcinogenic
 response when  administered subsequent to an  Initiator.   When administered
 concurrently with  an  Initiator, however,   phenobarbltal Inhibits tumor forma-
 tion whereas cloflbrate  enhances  tumor formation  (Moch1zuk1  et a!., 1983).
 In  addition to  variations  1n  the effects of  dose schedule  on carcinogenic
 Interaction,  the nature  of the Interaction may  also  vary with  the  site of
 action.   For example, Anderson  et al. (1983)  have noted  that PCBs (Aroclor
 1254)  Inhibit  the  development of  lung  tumors  but  enhance the development of
 liver tumors In  mice  Initiated with N-nltrosodlmethylamlne.
     As  with the quantifying of  cocarcinogenlclty  and promotion,  a  consis-
 tent  and  predlcable  pattern  of  Interaction  has  not  yet  emerged   1n  the
 assessment  of  compounds  that inhibit  carclnogenldty  (Schulte-Herman, 1985;
 Williams,  1984).  Until  such a pattern does  emerge,  It   Is  not  likely  that
 studies  such as those  described  above will  be used  to  modify quantitative
 risk assessments for chemical mixtures.
     Conversely,  both  inhibition  and masking may be  significant in  the inter-
 pretation  of  cancer  bioassay data on  mixtures.  For  instance,  Raabe  (1987)
 has  recently presented  an  analysis  of the dose-time-response relationships
 of  plutonium-239 in  causing deaths   from pneumonitis and   lung  cancer  In
beagle  dogs.   Deaths  from  pneumonitis  tended  to  occur at higher  doses  and
earlier  in  life than  deaths   from  cancer,  thus   masking the  carcinogenic

                                    2-22

-------
activity  seen  at   lower  doses.   Similar patterns  have  been  seen  In  the
results  of  many  cancer  bloassays  on  single  compounds  In  which  early
mortality from causes other  than  cancer  confounded  the Interpretation of the
results.  For bloassays  on  mixtures of compounds, the  results;  of masking of
carclnogenlcity because  of  early  mortality could  be  particularly significant
1f the  mixture  contains known  carcinogens.   For  example, human  exposure to
the mixture  at  concentrations  below the toxic  threshold could  result  In  a
significant  Increase  In the  risk of cancer  that  would not be  reflected In
the  animal   bloassay.    No  data,  however,   were   located that  specifically
address this Issue In the published literature.
2.6.    QUANTIFICATION OF INTERACTIONS                         '
    The  practical   or   quantitative  significance  of  toxic  Interactions  at
environmental levels  of exposure  Is difficult to  assess.  As  discussed In
previous  sections  of this  report  and  detailed further  In  Chapter  4,  most
published studies on Interactions are  not designed  to  quantify the magnitude
of the  interaction but  focus  primarily on  qualitatively  characterizing the
nature of the interaction.   In  addition,  quantitative  measurements of inter-
actions  can  only  be made  In  reference  to  a non-interactive  mathematical
model,  several  of  which are  discussed  in  Chapter  4 and  by  NAS  (1988a).
Thus,  the interpretation  of the data  in  determining whether  Interactions
                                  •
occur can be highly model  dependent.  The available models  also assume that
the  Interaction  among  the  compounds in  the  mixture is  consistent  over the
entire  range of  relevant   dose  levels.  An  important  consequence  of  this
assumption  is  that  the interaction  1s  assumed to  have  no threshold.   Few
data are available for assessing the validity of this assumption.
    The  majority  of  studies that  allow for  any  quantitative  estimate  of
                                                              I
interaction  involve acute  exposures  in  which death  or  some  other  severe
endpoint  Is  measured.   In  such  studies (Smyth et  a!.,  1970; Hermens  et al.,

                                    2-23                      !

-------
 1985a,b),  Interactions are  expressed  as  the  ratio of  the  observed  LD5Q to
 the   expected  LD5Q  based  on   the   assumption   of  dose  addltlvlty.   As
 discussed  In  Section  2.2.,  this  Is  often  referred  to  as  the ratio  of
 Interaction.   Most  reported ratios of  interaction  do  not  exceed a factor of
 2; the  largest reported variation 1s  a factor of 5  for an equivolume mixture
 of morpholine and  toluene  in  the study  by  Smyth et  al.  (1970).   Given the
 variations  Inherent in the  conduct  of bioassays,  the  significance  of these
 variations from  additivlty  is  unclear.   Few data are available regarding the
 variation  of  Interactions  among  bioassays  conducted  by  the  same  investi-
 gators  (Sun   and  Johnson,  1960),  and   no  interlaboratory  studies have been
 conducted.   Another source  of  uncertainty In assessing the  Implications  of
 these  ratios  of interaction  is  that  the  nature  and  magnitude of  inter-
 actions  for   severe toxic  effects  may  not  be  the same  as  those for  less
 severe  effects.   Furthermore,  interactions that occur at  high  doses  may not
 occur  in  the low-dose  region.   For  example,  the work of  Plaa  et  al.  (1982)
 on the  well-documented potentiation of  carbon  tetrachloride-induced  hepato-
 toxidty by  acetone suggests  that threshold  concentrations exist below which
 an enhancement  of toxicity may  not  occur.  As discussed  1n  Chapter  3, many
 of the  biologic  mechanisms  by  which  interactions  occur are also likely to be
 threshold phenomena.
    As  with  acute bioassays of  simple  binary mixtures, very  few  studies  on
 promotion or  interaction  were located  that  allow for the  quantification  of
 the interaction.   One exception  is the study summarized by  Pfeiffer  (1977)
 on interactions  of  carcinogenic  and noncarclnogenic PAHs.   This study, which
 Involved  3000   mice,   demonstrated  both  enhancement  and   Inhibition  of
 carcinogenic  activity.  Measured in  terms   of  the  observed  proportion  of
 responders  versus  the expected  proportion  of  responders,  variations  from
addltlvlty ranged up  to  a  factor  of  approximately 3.   Most  other  studies

                                    2-24

-------
using experimental animals  Involve far less elaborate  experimental  designs:
ethanol  and vinyl chloride  (Radlke et  a!.,  1981);  cyclopenteno[cd]pyrene  and
benzo[a]pyrene   (Cavallerl   et  al.,   1983);   and  diethylnitrosamine   and
phenobarbHone  or  alcohol   (Driver   and  McLean,   1986).    These  generally
discuss  or  provide  data  that  suggest  variations  from  additivity,  based  on
comparing the  observed vs.   the  expected proportion of responders, by  less
than  a  factor  of 10.   Because observed  response  rates  In  most  of  these
bloassays are over 10% and  must  be less than 100%, this observation may have
more  to  do  with  the design  and  limitations of most bloassays  than  with  the
quantitative  significance  of  Interactions.   No  quantitative  reviews   of
cocarclnogenlc activity or  promotion  efficiency  have been  encountered  1n  the
literature  that  attempt   a   systematic  and   consistent   analysis   of  the
available but diverse  animal  data  in order  to  estimate the significance  and
       '
magnitude of these phenomena for risk assessment.
    Epldemiologic studies on mixtures,  as discussed  In  Section 2.2.2.,  focus
on  measuring  relative risk  associated with exposure  to a  complex  mixture.
Occasionally,  Interactions   can  be  quantified  between  exposures  to  two
complex mixtures  or  one complex mixture  and another compound  or agent.   As
with measurements of interactions  from other types of  studies,; any quantita-
tive  estimate of  interaction  must  be made with  reference to  a non-inter-
active model.   For  example,, one of  the most studied examples  is the inter-
action  between   occupational  exposure  to  asbestos  fibers  and  cigarette
smoking  (Hammond  and  Selikoff, 1973; Hammond et al.,  1979;  Selikoff et al.,
1968, 1980).  In  the  study  by  Hammond et al. (1979), relative risks of about
                                                               I
5,  11,   and  53  were  noted for nonsmokers with  occupational  exposures  to
asbestos,  smokers with  no   occupational  exposure  to  asbestos,  and smokers
with  occupational exposure   to asbestos,  respectively.   As  discussed  in  the
                                                               i
mixture  guidelines,  this can  be interpreted  as  evidence for  a substantial

                                    2-25

-------
Interaction  (synerglstlc)  between  cigarette smoking and asbestos exposure 1f
an additive  risk  model  is  assumed  or as an Indication of no Interaction If a
multiplicative risk model  1s assumed.
    More  recently,  Steenland and  Thun (1986) have  reviewed  the measurement
of Interactions In  epldemlologlc studies  Including  a reappraisal of the data
on cigarette smoking  and exposures  to  asbestos,  radon  daughters,  arsenic or
chloromethyl  ethers.  As discussed by Steenland  and Thun (1986), synerglstlc
departures  from  an  additive risk  model have Important  public  health  conse-
quences  In  that  eliminating  exposure to one agent can result,  In  a greater
reduction In risk than  1f  no  synerglstlc  Interaction  occurred.   The  multi-
plicative  risk model,  on  the other hand,  1s   used  1n characterizing  the
etiology  of  a disease by determining  1f  one  risk, factor modifies  the  effect
of another  risk  factor.   Of the epldemlologlc studies  reviewed  by Steenland
and Thun  (1986),  the  Hammond et al.  (1979)  study showed the  greatest  devia-
tion, by  a  factor of  about 3.5,  from risk addltlvlty.  Other  deviations from
risk  addltlvlty  ranged from  a  factor  of  about   2  for  smoking  and  radon or
arsenic to  0.2  for  smoking and chloromethyl  ethers.   In no Instance did  the
observed  relative risk  for smoking  and the other  agent  exceed  the relative
risk  predicted  by the multiplicative  risk  model.  The  recent  reanalysls of
the combined  effects  of  cigarette  smoking  and exposure  to radon daughters In
the BIER  IV  report  (NAS, 1988b) also  noted  evidence for a  multiplicative or
a "submultlpHcatlve" model  (I.e., the risk was  greater  than  that predicted
by the  additive  risk model but   less  than  predicted by the  multiplicative
risk model)  for uranium miners, although  some support was found for a  supra-
multiplicative model.
                                    2-26

-------
              3.   AVAILABLE INFORMATION ON INTERACTION MECHANISMS
3.1.   OVERVIEW
    This  chapter  summarizes  Information  on  the  chemical  and  biological
                                                              I
mechanisms  by which  compounds  Interact.   Such  mechanisms  Include chemical-
chemical  interactions,  pharmacokinetic effects  and  interactions at receptor
site:;  and other  critical  cellular targets*   For  the most part,  effects of
different   types   (lethality,   narcosis,   enzyme   induction,   reproductive
effects)  or  effects  at different  sites  involve a common  set of mechanisms.
The  phenomena  of  promotion  and   cocarcinogenicity  have  been  extensively
studied  in  a  distinct body of  literature  and  may  involve a  complex  and
as-yet-not-fully  understood  series  of  mechanisms,  which  are discussed at the
end of this section.
                                                              I
    As stated  in  the  mixtures  guidelines,  toxicant interactions may be based
on any  of the processes  that  are  significant  to  the toxicologic expression
of a  single compound:  absorption, distribution, metabolism,  excretion  and
activity  at   cellular  site(s).    In  addition,   compounds   may  interact
chemically,  causing  a change  in  the  biological effect  or  they may Interact
by  causing   different  effects  at  different  receptor  sites.   Of  greatest
practical importance  is that most  of  these mechanisms  are saturable  as  are
most  kinetic   processes  for  single compounds.   Consequently,   many   of  the
interactions observed  at  high doses may  not be significant  in the low-dose
region.
    Table 3-1, which  summarizes   these  general modes  of  interaction  along
with  some  examples,  was  prepared  using a  modification  of the  basic  scheme
proposed  by  Veldstra  (1956).   As  detailed  in  an  extensive  review   by  WHO
(1981), "... the  available evidence from In  vitro  and animal  experiments  and
from  human  observations  has  shown  that a  limited number of  mechanisms  seem
                                    3-1

-------
                                   TABLE  3-1

            Chemical and Biological Bases of Toxicant Interactions
         (See  text for  discussion,  additional  examples  and  references)
                                                Examples
Bases of Interaction   Synerglsm or Potent1at1on
                                   Antagonism
Chemical
Biological
  Absorption
  Distribution
formation of nltrosamines
from nitrites and amines
neurotoxldty of EPN
(o-ethyl o-r-n1trophenyl
phenylphosphorothloate)
enhanced by aliphatic
hexacarbons due In part
to Increased skin absorp-
tion (Abou-Donla et a].,
(1985)
Increased lead levels
brain after treatment
with dlthiocarbamate/
thluram derivatives
(Oskarsson and L1nd,
1985)
In
         dimethyl hydrazine
         reacts jm vivo with
         pyridoxal phosphate
         (vitamin B6) to form
         a hydrazone, thus
         rapidly depleting
         tissue stores of this
         enzymatic cofactor
         (Cornish, 1969)
         dietary zinc inhibits
         lead toxldty 1n part
         by decreasing the
         percent dietary lead
         absorbed (Cerklewski
         and Forbes,  1976)
the mechanisms by which
selenium protects
against cadmium toxlc-
Ity Include decreasing
the concentration of
cadmium in Hver and
kidney and the redis-
tribution of cadmium In
the testes from the
low-to-high molecular
weight Cd-binding
proteins (Chen et al.,
1975)
                                    3-2

-------
                              TABLE 3-1  (cont.)
                                                Examples
Bases of Interaction   Synerglsm or  Potentiatlon
                                   Antagonism
  Excretion
  Metabolism
decreased renal excretion
of penicillin when co-
administered with pro-
benecld
organophosphorous com-
pounds (profenfos, sul-
profos, DEF) potentiate
the toxldty of fenval-
erate and malathlon by
Inhibiting esterase which
detoxifies many pyreth-
rold Insecticides (Gaughan
et al., 1980)
  Interaction at
  Receptor Sites
  (Receptor Antagonism)
no Information available
  Interaction Among    no Information available
  Receptor Sites
  (Functional Antagonism)
  Interaction at DNA   no information available
arsenic antagonizes the
effects of selenium in
part by enhancing the
biliary excretion of
selenium (Levander and
Argrett, 1969)

selenium inhibits 2-
acetylaminofluorene-
induced hepatic damage
and liver tumor inci-
dence in part by
shifting metabolism
toward detoxification
(ring hydroxylation)
relative to metabolic
activation (N-hydroxy--
lation) (Marshall
et al., 1979)

blocking of acetyl-
choline receptor sites
by atropine after
poisoning with organo-
phosphates

interaction of hlsta-
mine and norepinephrine
on vasodllation and
blood pressure
         i
Induction of DNA repair
by exposure to
alkylating agents
                                    3-3

-------
 to account for the majority of the  Important  known  biological  Interactions."
 In other words, the basic mechanisms by which  toxicants  interact  as  detailed
 by Veldstra  (1956)  are based on  classic  pharmacologic  principles that  have
 not changed  substantially  over  the  past  30 years.  While  most  of  the  best
 studied examples of  the  mechanisms  of compound  interactions  are still  from
 the pharmacologic literature on  therapeutic drugs  or abused substances  such
 as ethanol  (Seitz,  1985;  Puurunen  et al.,  1983), an  increasing number of
 examples  are available  showing  similar mechanisms  for  compounds  of occupa-
 tional  and  environmental  concern.
 3.2.    CHEMICAL  INTERACTIONS
    Many  cases of  direct chemical-chemical Interactions  lead  to  a  decrease
 1n toxlcologlc activity, and this is one  of  the  common  principles of anti-
 dotal  treatment.   Examples  Include  the use  of chelating agents  to complex
 with  metal  ions, the  inactlvation of  heparin by binding to  protamine,  and
 the  use of ammonia  orally as an  antidote to  the  ingestion of formaldehyde
 through  the  formation  of hexamethylenetetramine  (Goldstein et a!., 1974).
 This  class of  reactions  has been  referred  to  as  chemical   antagonism by
 Klaassen and  Doull (1980).
    Chemical  reactions  that  lead  to greater than additive effects have been
 less  frequently  documented.   One example  that   has  received  considerable
attention  is  the formation,  in  the  stomach,  of  nitrosamines  from nitrites
and  amines,  which  result  in an  increase  in both  toxic  and  carcinogenic
effects  (Welsburger  and  Williams,  1980;   U.S.  EPA,  1986a).   Other  examples
 Include the formation of  arsine and  stibine from  ores  containing  arsenic  and
antimony,  respectively,  which come  into   contact  with strong  acids in  the
stomach.  Thus, while antagonism may  be the most  widely  recognized result of
                                    3-4

-------
this mechanism for toxicant  Interaction  In  the  classic  pharmacologlc  litera-


ture, synerglsm or potentlatlon  also occur and may  be  as  significant  In the


environment.


3.3.   PHARMACOKINETIC-BASEI) INTERACTIONS


    Many  examples of  toxicant  Interactions  are  based  on  alterations  In


patterns of absorption, distribution, excretion or  metabolism  of  one  or more


compounds  1n  the  mixture.  Several  reviews  of  these factors  in  the  assess-


ment  of multiple  chemical  exposures  are  available (Anderson and  Clewell,


1984; Plaa  and  Hewitt, 1981;  WHO,  1981; Wlthey,  1981).   All  of  these kinds


of  Interactions  essentially alter  the bioavailability  of  thejtoxic  agent(s)


to   the  cellular  slte(s)  without  qualitatively  affecting  the  toxicant-


receptor site Interaction.                                    ;


3.3.1.   Effects  on  Absorption.  Most  kinds of Interactions based  on alter-


ations  In  absorption  involve  vehicle  effects,  the  chemical formation  of


poorly  absorbed  conjugates  or complexes,  or  decreases  in gastrointestinal


motility.   Examples  of  such  effects  have  been  noted  for oral and  dermal
                                                              i

exposures.                                                    j
                                                              i

     For  example,  the  dermal  toxicity  of TCOD  adsorbed on  charcoal  is


considerably  less than  that  of TCDD  solubllized   in  a  lipophilic  medium.


This is presumably due to  the reduced  availability of  the charcoal-adsorbed


TCDD for  absorption  by  biological  systems (Polger  and  Schlutter,  1980).


Conversely,  dimethyl  sulfoxide, a  commonly used vehicle  1n  dermal  toxlcity


studies,  is  known  to facilitate  the absorption  of many organic  compounds
                                                              i

across  the skin,  thus causing apparent  potentiatlon when compared  with less


lipophilic  vehicles  (Goldstein  et  a!., 1974).   A  similar mechanism appears


to  be involved  1n the enhancement  of the neurotoxlclty of  o-ethyl-o-4-n1tro-


phenyl  phenylphosphonothioate by various aliphatic  hydrocarbons when applied
                                                              j

derraally  to hens  (Abou-Donia et  a!., 1985).                   i

                                                              !
                                                              !

                                     3-5

-------
     The acute oral  toxicity  of many  compounds  Is  substantially affected  by
 the vehicle used, and many of  these effects are probably due  to differences
 1n rate of absorption.   For  example,  clloqulnol  administered orally Is able
 to complex with  many metals,  facilitating  their  absorption,  and  has been
 Implicated  In an  outbreak of heavy metal-Induced subacute myelo-optlc  neuro-
 pathy In Japan (Tjalve,  1984).   By  contrast there are examples of compounds
 that  form poorly  absorbed complexes  after oral administration such as  tetra-
 cycllne and  calcium carbonate,  as  well  as  cholestyramlne  and cholesterol
 (Goldstein  et al.,  1974).   Some  compounds  given  orally,  such  as  codeine,
 morphine,  atroplne and  chloroqulne,  decrease the  rate  of  gastric  emptying,
 thus  decreasing  the rate  of  absorption  of  orally  administered  compounds.
 For  the most  part,  such Interactions  usually lead  to decreases In effects,
 because of the slower  rate  of  absorption rather  than  Increases  In effects
 because of more complete  absorption  (Levlne,  1973).
    As  discussed  by Wlthey (1981) and confirmed In  the  literature  reviewed
 for   this  report,  there  are  no examples   of   toxlcologlcally  significant
 changes  In absorption  associated with  the  Inhalation of  mixtures.   Murphy
 (1964)  reported  Increased carboxyhemoglobln  levels  In mice and rats exposed
 to  an ozone-CO mixture compared with  CO  alone.   The exact  mechanism of this
 response, however,  has yet to  be determined.  Anderson and  Clewell  (1984),
 In  their  review  of  pharmacoklnetic  Interactions  and Inhalation  modeling,
 cite  several   examples of Interactions  based on effects  on   metabolism  but
 none  based on absorption.   It  has   been  hypothesized,  however,  that  one
mechanism by  which particulates  such  as  ferric  oxide serve  as  respiratory
cocarclnogens  for  benzo[a]pyrene (B[a]P) Is  by Increasing  residence  time  In
the lung and,  thus, allowing for  more complete  absorption of  the  compound.
Alternatively, If  absorbed on  particles, the  B[a]P  1s  taken up by  macro-
phages  that  have  been   shown  to be  capable  of  metabolizing B[a]P  to  a

                                    3-6

-------
proximate  carcinogen.   In  addition  it  has been  shown that  cocarcinogenic
particles  facilitate  uptake  of   the  adsorbed  chemical  carcinogen  across
phospholipid bilayer membranes  (Lakowitz and Hylden, 1978; Leikowitz  et  a!.,
1980).
3.3.2.   Effects on  Distribution.  Distribution  can  play a role  in  compound
Interactions if a more active agent  is  displaced from  an  inactive  site  to a
primary  receptor  site  by a  less  active  or  inactive  agent.  One  of  the  best
documented  examples  of this  kind of activity  is  the  displacement  of  anti-
coagulants  from plasma  proteins   by  compounds  such  as barbiturates,  anal-
gesics,  antibiotics  or  diuretics  (Goldstein   et  al.,  1974).   Similarly,
tri-o-tolyl  phosphate  decreases  the binding  of  paraoxon to  nonvital  tissue
in rat liver and  plasma,  consequently  increasing the  toxicity of paraoxon in
rats  (Lauwerys  and  Murphy,  1969).   Since  body  fat  represents  a  major
nonvital storage site  for many  lipophilic xenobiotics,  it  may be anticipated
that  compounds  that cause  fat  mobilization could  result  in  similar  poten-
tiating effects (WHhey, 1981).
    Recently,  Oskarsson and  Lind (1985) demonstrated  that  dithiocarbamates
and  tetramethylthiuram  disulflde may  complex  with   lead  and  selectively
increase  the  accumulation   of  lead   in  the  brain.   While  the  toxicologic
significance  of this  interaction has not  yet  been  demonstrated it  can be
reasonably  presumed  that  this effect on distribution is likely to  lead  to a
synergistic  effect   on the  CNS  effects  of lead.  A  related  mechanism was
proposed by  Larsson  et al.  (1976) for  the teratogenic  effect of maneb, which
is antagonized  by zinc acetate,  suggesting that the  teratogenic activity of
maneb  is  due  to  the   binding  of zinc,  causing embryonic  zinc deficiency.
Host,  examples  cited  above,  however,  result in greater  than  additive effects
— synergism or potentiation.
                                    3-7

-------
 3.3.3.    Effects on  Excretion.   Most examples  of  excretion as  a basis  for
 toxicant  Interaction  Involve  compounds  that  are  eliminated  through  the
 kidneys.  For Instance,  probenedd  or carinamide both competitively  Inhibit
 the elimination  of penicillin, thus prolonging or potentiating its desirable
 therapeutic  effect.  Similarly,  phenylbutazone  inhibits  the renal excretion
 of  hydroxyhexamide, which  can  cause undesirably prolonged hypoglycemia.  If
 a   toxicant   Is  eliminated  through  the  kidneys,   a  stimulation of   renal
 elimination   can  cause  an  antagonistic   effect,   as   is  seen  with   the
 coadmlnistration  of  phenobarbital  and   sodium  bicarbonate   in  which   the
 increased  urine  alkalinity  induced by  the  bicarbonate   ion  increases  the
 excretion  of  phenobarbital  (Goldstein et al.,  1974).
    A   less   direct   effect  on   renal  elimination   has  been  suggested  by
 Herschberg and  Sierles  (1983)  for  the   substantial  potentiation  of   the
 toxlclty  of  lithium,   which   is   eliminated   through   the   kidneys,  by
 indomethacin.  These  investigators  suggest  that the potentiation  is  due to
 the  inhibition  of  prostaglandin  synthesis  by  Indomethacin,  which  in  turn
 causes  vasoconstriction and a decrease in  the  renal excretion of lithium.
    As  summarized  by  WHO (1981), several  drugs  and  other  chemicals are also
 able  to compete  for  biliary  excretion.   Yamada et  al.  (1986)  have  demon-
 strated  that  quinidine  has a  marked  inhibitory effect  on  the  presystemic
 elimination  of  ajmaline  by  the  liver when  both compounds  are administered
 concurrently to rats;  similar observations  have been noted  in humans.
 3.3.4.   Effects  on  Metabolism.   Altered  patterns   of  compound  metabolism
 have been  shown  to be the bases  of  many toxicant  interactions  (Anderson and
 Clewell, 1984;  WHO,  1981).  A  major enzyme  system  involved in  such  inter-
actions  is   liver  microsomal  mixed-function  oxidases   (MFO),   which   are
 Involved 1n the  activation  or detoxification of  a wide  variety  of compounds.
                                    3-8

-------
Both  the types  (e.g.,  different  forms of  cytochrome P-450)  and  levels of
metabolic enzymes  can be  induced by agents such as phenobarbHal, and enzyme
activity can  be  inhibited  by agents such as piperonyl butoxide  (Goldstein et
al.,  1974).   Thus, depending on whether or  not the toxicant is activated or
detoxified,  inducers  or inhibitors  of  this  enzyme system  may cause syner-
gistic/potentiating  effects;  or  antagonistic  effects  (Freeman and  Hayes,
1985; -Leonard  et al., 1985).   Toxicant interactions involving the MFO can be
                                                               i
complex  and  depend  on  both  dose  and  duration of exposure,  with  some
compounds  causing  an  initial  inhibition  of enzyme  activity  followed  by  a
marked induction of activity  (WHO,  1981).   Although liver microsomal MFO are
the most commonly  studied  enzymes involved  in  toxicant  Interactions,  MFO in
other  tissues  may also  play an  important role in toxicant Interactions as
may  other   enzyme   systems,  such  as  alcohol  and aldehyde dehydrogenases,
monarnine and  diamine  oxidases,  dehydrochlorinases, azo and nit.ro reductases,
hydrolases  and  enzyme  systems  involved   in  conjugation  reactions.   For
instance, ethanol  serves  as an  antagonist of  the  toxic  effects of methanol
by  acting  as   a   competitive   inhibitor   of  alcohol  dehydrogenase,  thus
suppressing  the  formation  of  formaldehyde  and   formic  acid  from  methanol
(Goldstein et al.,  1974).                                      !
3.3.!).   Interactions at Receptor  Sites or  Critical  Cellular ;Targets.   All
of the biological  modes  of toxicant  interactions  discussed  above  -•- absorp-
tion,,  distribution,  excretion  and  metabolism  --  are essentially  disposi-
tional,  affecting  the amourit(s) of  toxicant(s) reaching  the  primary  recep-
tor(
-------
1nterat1ons.  The antagonistic nature of  Interactions  that  occur  at  the same
receptor site was discussed by Veldstra (1956):
    ...we may say that  the  effect  of  a  combined action of two compounds
    at the  same  site of primary action will  not  result  in  a synergism,
    but will,  generally,  even be unfavorable.  The  competition  for  the
    receptor  will  usually  decrease  the  frequency  of  the   best  inter-
    actions,  and with  decreasing  intrinsic  activity  of   one  of  the
    components the combined  action  will more and more take the  form of
    a competitive antagonism.

Examples of  such interaction Include the  antagonistic effects of oxygen  on
carbon  monoxide,  atroplne  on   cholinesterase inhibitors   and  naloxone  on
morphine  (Goldstein  et a!.,  1974).  The  antagonistic consequences   of  this
kind  of  toxicant  Interaction  are so  consistent  that  it   has  been  termed
"receptor  antagonism"  by  Klaassen  and  Doull  (1980)  and  "pharmacological
antagonism" by Levine  (1973).   While  it seems  conceivable  that  one  compound
could increase the  intrinsic activity  of  another compound   by modifying  the
receptor site -- analogous to the  effect  of  modulators on regulatory enzymes
-- such Interactions  have not been demonstrated.
    Interactions  of  agents  among receptor sites  are also  thought to result
primarily  in  antagonistic  effects  and  has been  referred to as  "functional
antagonism" by both  Klaassen  and Doull  (1980)  and  Levine (1973).  This kind
of Interaction 1s  most commonly defined  as  two or  more  compounds acting  on
different  receptor sites  and causing opposite effects  on  the  same  physio-
logical function.   Examples  Include  the opposite  effects  of hlstidlne  and
noreplnephrlne on  vasodilation  and  blood  pressure  and  the  antlconvulslve
effects of barbiturates on many  compounds  that cause convulsions.  Neverthe-
less,  that  Interactions among receptor  sites uniformly result in  an  antago-
nistic response  is not  certain,  particularly when the receptor  sites act  on
                                    3-10

-------
different  physiological  systems.    The  rationale   for   this  statement  was

presented by Veldstra (1956):
                                                              i

    The  sites  of  action  for  two  compounds  having the  same type  of
    activity may be different.  This  is  the  case when the effect  can be
    caused either by  a  direct stimulation or  by  the annihilation of an
    inhibition.   In  both  cases,   the   combination   of  two  compounds,
    linked in  parallel  or in series, as it  were, may well  result  in a
    synergistlc effect.   When the  components  of a  combination  possess
    different  sites  of  action  and  different  types of  activity,  no
    plausible  prediction  about  the  possibility  of  synergism  can  be
    made, unless their mode of action is well known.

                                                              i
    While examples  of  such Interactions have not been well  characterized in

the literature, the potentlation  of carbon  tetrachlorlde  by chlordecone may
                                                              ,i
be  at  least   partly  mediated  by  an  inhibition of  hepatocellular  repair

(Lockard et a!., 1983).                                       j

    Another possible  illustration of  Veldstra's  argument  is  presented in the

work  of  Alstott et  al.  (1973),  who examined  the  acute  lethal  effects  of

combinations of  1-methylxanthine  and ethanol  on mice,  and  noted  two  basic

kinds  of effects:   kidney  dysfunction  and  increased  respiratory  rate  and

depth.   In  animals  exposed  to mixtures  in  which  the  ratio of  1-methyl-

xanthine to ethanol was  relatively  high,  antagonism  of acute lethal toxicity

was observed;  however,   in mixtures  in   which  the same ratio !was  relatively

low, a synergism of acute  lethal toxicity was  observed.   This indicates that
                                                              i -
in cases where toxicants  interact at  more than one  cellular  site, the nature

of  the  Interaction  can  be either  antagonistic  or  synergistic.   The  compli-

cating factor of the  "asymmetric" pattern  of interaction  observed by Alstott

et al. (1973) is discussed in greater detail  in Chapter 4.     |

3.3.6.   Promotion  and  Cocarcinogenicity.   Mechanistic  studies   on  promo-

tion and cocarcinogenicity have been active  areas of  research over the past

decade.  The extensive  literature has been the  subject of  several  comprehen-

sive general reviews  (Slaga, 1984;   Williams,  1984;  Yamasaki,, 1984)  as  well


                                    3-11                      j        .

-------
 as  reviews  that  have  focussed  on  specific  topics  such  as  hepatocarcino-
 genesls  (Pltot  et  al.,  1982;  Schulte-Hermann,   1985),  the  Inhibition  of
 cellular communication by promoters  (Trosko  et  al., 1983), the  induction  of
 superoxlde anions  by promoters (Troll and Wiesner,  1985), and  the  binding  of
 promoters to protein kinase  C  in  cell membranes  (Hecker,  1985).
     Various   investigators   have  used  different  but  generally overlapping
 mechanistic  schemes  to categorize  the  types  of  information on promotion and
 cocarclnogenicity.    Table  3-2  summarizes  the  mechanisms  based  on  the
 approach taken by  Williams (1984),  who also provides many specific examples.
 As  mentioned earlier in  this  chapter, one possible mechanism of  cocarcino-
 genesls   is   to  increase  cellular   exposure  to   an   initiating   substance.
 Particulate  ferric oxide can  serve as an  effective vehicle for delivery  of
 an  adsorbed  carcinogen, such  as  benzo[a]pyrene  to  the  target  organ, namely
 lung.   The  particles  are  subject  to  phagocytosis  by pulmonary alveolar
 macrophages,  which  can  elute the  benzo[a]pyrene  (Autrup  et  al.,  1979),
 transport  the  compound  to  a  distant  site  or   metabolize  it.   Likewise
 solvents  may  also  serve  as  cocarcinogens  by   increasing  efficiency  of
 carcinogen delivery.
    Agents  may  serve  as  cocarcinogens  by affecting  the  metabolism  of  a
 procarcinogen  such that  a  more active metabolite  or that a greater quantity
 of reactive  metabolites  is maintained in  the  cell.  This can be accomplished
 by  Induction  of  metabolic  enzyme   systems  as  described  previously  or  by
 depletion  of or competition with detoxification  systems.   An  example of  a
 compound  with  this  latter  activity  is  diethyl  maleate, which is known  to
 deplete  liver  of glutathione,  a  cellular nucleophile.   Depletion  of  gluta-
 thione Increases hepatotoxicity (and  presumably  the potential  for hepatocar-
cinogenicity) of aflatoxin Bl (MgBodile et al.,  1975).
                                    3-12

-------
                                   TABLE  3-2                    j
                Mechanisms  of  Promotion and Co-carcinogenicity*
Co-carc1nogenes1s:
                                                               i
         1.  Increased uptake of carcinogen                    \

         2.  Increased proportion of carcinogen activation
                                                               [
         3.  Depletion of competing nucleophiles

         4.  Inhibition of the rate or fidelity of DNA repair  |

         5.  Enhancement of the conversion of DNA lesions to permanent
             alterations

Promotion:                                                     }

         1.  Enhancement of expression of neoplastic phenotype
                Inhibition of differentiation

         2.  Stimulation of cell proliferation
                                     - -                         1-
                Cytotoxlclty
                Hormone Effects

         3.  Cell membrane effects                             [
                                                               I
                Induction of proteases                         [
                Inhibition of Intercellular communication
                                                               [
         4.  Immunosuppression                                 j

*Source:  Williams,  1984                                       i
                                    3-13

-------
    Another  possible mechanism of cocarclnogenesls  takes  place at the level
 of  DNA damage.   It Is known that certain compounds can act as  co-mutagens In
 In  vitro  systems: norharman  for aniline  and benzo[e]pyrene  for benzo[a]-
 pyrene.   These  Interactions could take place  at  any of a number of steps In
 the mutagenlc  process,  Including enhancement of mutagenlc metabolite produc-
 tion.   It 1s known, however,  that DNA  that 1s being actively  transcribed Is
 more  susceptible to damage  than  Is  "resting"  DNA.   It seems plausible then
 that  some agents  could  enhance Initiation by making the DNA more susceptible
 to damage, for  example,  by holding It 1n a single stranded configuration, or
 by Increasing transcription.
    Interference  with  error-free DNA  repair  Is  another  way  In  which  a
 cocarclnogen  could  work.    Induction  of  an  error-prone  repair system  by
 DNA-damag1ng agents  is  a well-documented  phenomenon  In EscheMchla coll.   In
mammalian  cells,   certain  systems,  such  as that  responsible  for repair  of
alkylation  damage, also  appear  to  be  inducible (Swenberg  et  a!.,  1982).
There  1s,  however, no evidence  as  yet  of  an  error-prone  repair  system that
could  be  turned  on by either a DNA-damaging or a co-mutagenic agent.   It  has
been  reported  that some agents  reduce  the rate  of  DNA synthesis, including
repair  synthesis.   Such  a  reduction  in rate of repair could  have the effect
of  increasing   the  number  of  permanent  DNA   alterations  or  mutations
 {Williams, 1984).  It  has  been  reported  that a compound,  3-aminobenzene,
which  inhibits  the activity  of  the  poly(ADP-ribose)  polymerase  specific  to
DNA  repair  enhances  the   formation  of  liver   foci   initiated  by  another
compound  (Takahaski et a!., 1982).
    The  classic   two-stage  initiation-promotion  sequence  proposed by  early
Investigators  (Berenblum,  1941a,b) is  more likely  to reflect  experimental
design  constraints  than  two  simple  discrete  mechanistic  stages.    Slaga
(1984) has described two separate stages  in promotion  in  which  the initiated

                                    3-14

-------
cell develops  to  a benign tumor, as  well  as  two stages of progression.  The
first  stage  of progression  Is  that  In which a  benign  tumor  develops Into a
malignant tumor, while  In  the  second  stage the malignant tumor metastaslzes;
each of these stages may Involve different mechanisms of Interactions.
    As mentioned In Chapter  2,  the  practical  significance of the distinction
between tumor  Initiation  and tumor  promotion  Is  that  the former Is commonly
regarded as  having no threshold while the  later 1s  often thought to display
a  threshold  below which  no  tumor promotion  will occur  (Driver  and McLean,
1986).   This view,  however,  has  been  challenged  by  Yamasaki  (1984),  who
claimed that the data are  not  adequate to determine If promotion evidences a
true dose-threshold.   Rather,   1t was suggested  that  because at  least  some
stages  of  promotion  are  reversible,  promoters  display  a  dose-schedule
threshold (I.e., the dosing  schedule  Is  of greater  Importance than the total
administered dose) that  Is  different  from  that of  InHators,--or complete
carcinogens.                                                    ;
                                                                i
    The  Implications  of  mechanisms  of   promotion  for  risk assessment  are
further  complicated  by the  fact  that  some  compounds  can  interact  with
promoters  to  Increase  or  diminish  promoting  efficiency  (Schulte-Hermann,
1985;  Slaga,  1984;  Williams,  1984).   For  example,  Sleight  (1985)  has
reported   that   3,3',4,4',5,5'-hexabromob1phenyl   enhances   the   promoting
efficiency of 2,2',4,4',5,5'-hexabromoblphenyl  and  that this  may explain why
commercial  mixtures  of  polybromlnated  blphenyls have  a greater  promoting
ability than any of the Individual congeners.                   ]
3.3.7,.    Interactions and Developmental Toxiclty.  Developmental  toxldty  1s
Indicated by many  different types  of endpolnts  Including death,  structural
abnormality, altered  growth  and  functional  deficits  (U.S.  EPA,  1986b).
                                    3-15

-------
These  various  endpolnts are  likely to  arise  as a  consequence of any  of a
number of  cellular processes Including  mutations,  membrane changes,  changes
1n  gene  expression,  or   other  events  leading  to  cell   death.   There  Is
potential  for  Interaction  to occur  at  any of these  processes  that would be
manifested as  Increases or  reduction  1n developmental measurements.   It Is
generally  assumed  that there are  dose  thresholds  for  developmental  effects
based  on  the  rationale that  the  embryo  has  some  capacity  for repair "of
damage or  replacement  of  dead cells.  Interactions could  have  the effect of
raising or lowering this threshold as for other systemic effects.
                                   3-16

-------
              4.  MATHEMATICAL MODELS AND STATISTICAL TECHNIQUES
4.1.   INTRODUCTION
    Thrs  chapter  presents  a  review  and  evaluation  of  some  representative
statistical  methods  for   the  assessment  of  toxic  responses  to  mixtures.
There  are four different  classes  of  methods,  described as  Follows:   dose
addition,  response  addition,  generalized linear models  and  response surface
models.   The  theoretical  framework for  each  class is discussed,  and varia-
tions  within  each  class are described.   Some recently  proposed  methods  for
use  in  analysis of mixtures data  are  also presented, along with an evalua-
tion of applications of statistical methods in the mixtures literature.
    Interaction is defined statistically as the  effect  of two  or  more treat-
ments applied jointly that cannot  be predicted from the  average responses of
the  separate  factors.   This  concept  of  dependence of the  effect of  one
factor  on the  level  of  another  factor  is  a fundamental  scientific  idea.
When  interaction  is  present,   the result  of  two or  more  factors  applied
jointly  may   result   in  either  positive  or   negative   deviations  from  the
expected  result for each  factor  taken  one at  a  time.  As noted in  Chapter  1
and  Appendix  A,  when  a   large  positive  deviation  is  present,  the  common
                                                               i
biological terminology  used  is  synergism.    When  a  negative  deviation  Is
present,  antagonism  is  said  to be  present.   In  the special  case where  a
deviation  occurs when the two factors are applied together, but  one factor
by itself  has no  effect,  the positive deviation is  called potentiatlon,  and
the negative  deviation is  called inhibition.
    The  above  definitions  are  contingent  on  how  the expected  (or   "no-
interaction")  effects  are defined  (Berenbaum,  1985a).  There are  two general
classes  of models  for  joint   action  that   assume  no  interaction.   These
classes describe either  dose addition  or  response addition.
                                    4-1

-------
4.2.    DOSE ADDITION
    Dose addition,  or  simple similar action (Finney, 1971), assumes that the
compounds  in a  mixture act  as  if  they  are dilutions  or  concentrations of
each  other.   For  example,  in a  binary  mixture, a  dose  containing z, units
of  compound  1   and z~ units  of  compound  2  would,  under  dose addition,
behave  exactly   as  a  dose  of (z1+pz2)  units, of  compound  1  alone,  where  p
1s  the  potency  of compound  2  relative  to compound 1.  In particular, assume
the two compounds have parallel   regression  lines  of probits on log doses as
follows:
                               YI = a.j +  3  logZ                        (4-1)
                               Y2 = a2 +  13  logZ                        (4-2)
where  Z  is  the  dose.  Simple  similar  action  is  said to  occur when  the
response  to  a  mixture containing  amounts  Z, and Z2  units  of  compounds  1
and 2,  respectively, has a response probit of the form
                            Y = a] + 13 log(Z14-pZ2)                     (4-3)
Alternatively,  if the  mixture is  a total  dose Z  of  the two  compounds in
proportions f1 and f2 then the mixture has a response probit of the form:
                        Y  =  a-j  +  13 Iog(f^pf2)  +  13  logZ                 (4-4)
Note  that  the assumption of parallelism  is  implicit in the  formulation of
this model (4-1  and 4-2).
    One method  for  testing  for dose additivity  Is  to assess  the adequacy of
fit of  the model  (4-4)  rewritten as
                               Y  =  a3  +  6 logZ                         (4-5)
Alternatively,  when  several mixtures  of different  proportionate  concentra-
tions  are  to  be  tested, several  different  estimates of  p  can  be obtained.
                                    4-2

-------
The sum of squares between  observed  and  predicted response from equation 4-4
                                                               i
can then  be  minimized with  respect  to a, B  and p, and  an  overall estimate
of p  is  found.  Testing  for  dose  additivity is  then  done by comparing this
sum of  squares against one  where  values of p were  estimated separately for
each dose series.                                              j
    Finney  (1971)  has  also  proposed the -following  model   to  be used  for
assessment of Interaction:
                 Y = a + 13 1og(f1+f2p + K(f.,f2p)0-5) + B logZ          (4_6)
where a,  B,   p are  as  deflined  before,  Z  is  the  sum of Z,  |and  Z~,  and  K
is  the  coefficient  of  interaction.   A  positive  value  of  K  Indicates
synergism; a zero value, simple additivity; and a negative value, antagonism.
    This model assumes a  constant  Interaction  throughout  the entire range of
proportions  of  individual  components.    In  order  to  allow  for  a  less
restrictive assumption, Durkin (1981) made the following modification:
             Y = <* * B log(f1+f2p+(K1f1+K2f2)(f1f2p)°-5) + B logZ      (4-7)
The properties of  this  model, however, have yet  to be critically evaluated.
Durkin  (1981) also  proposes  several statistical  methods  for  testing  for
                                                               j
departures from  simple additivity.   For  example, the model  for symmetrical
interactive action (Finney1;; model  from equation 4-6):         i
                                                -p)0-5]               (4-8)
where  !„  is  the  observed  LC5Q  for  the mixture  and  Z,  is  the LC,-0  for
compound 1, can be fit using weighted linear regression analysis.  Similarly,
the model  for  asymmetrical interactive action  (Durkin1 s  model  from equation
                                                              j
4-7),
                 (1/Z3)  = (l/Z1)[f14-f2p+(K1f1^K2f2)(f1f2p)0-5] j         (4-9)
                                    4-3

-------
 can also be fit by similar means.  The hypothesis that
                                 Kfp0'5)/!., = 0
 (4-10)
                                       or
                                         0.5,
                            (K1f1+K2f2)(pu'3)/Z1 = 0                    (4-11)
 can then  be tested.   If  the relevant  hypothesis  1s  not rejected,  then  the
 data are  consistent  with simple  addltlvity.   Let  SSE] denote  the sum  of
 squares error  for  the simple addltlvity  model,  and SSE2 the  sum  of squares
 error  for  the relevant Interactive model.   Define then
     where:   $3 = (SSE-j  - SSE2)/(M - g)
             S2 = SSE2/(n - (M + 1))
 where n 1s the  number  of measurements, g is the number of  parameters  1n  the
 model for  simple  similar action  and  H  is  the number  of  parameters  in  the
 Interactive model.  Thus,  this F  statistic  has M-g  and  n-(M+l) degrees  of
 freedom.   Again,  the  properties  of this  method  have not  been rigorously
 evaluated.
     Another  method proposed  for  testing for simple additivlty  is to  divide
 the   observed   LC5Q  of   the   mixture   by  the   LC5Q   predicted   from   simple
 additlvity  (Durkin,  1981).   This  method is  to be used  when LC,.  estimates
                                                                 3\J
 of components  and  mixtures vary substantially,  especially those from experi-
 ments  conducted at different times, thus obscuring  trends  to nonlinearity.
 Under  the  null  hypothesis  of simple  similar  action  the  observed  LC™   of
                                                                        DU
 the mixture will equal the LC5Q predicted from equation 4-5.  Explicitly
                                    = cj>1 -t-
= 1
(4-12)
where  1   and   2  can  be  considered   the   proportions  of  the  mixture
toxicity  attributable   to   compounds  1  and   2,   respectively.    Under  a
                                    4-4

-------
hypothesis of Interaction such as given by equations 4-8 or 4-9, then
                                                                       (4-13)
or
               Z3pred
                                                                       (4-14)
This heuristic method does not  have  set  rules  for determination of statlstl-
                       j


cal significance and  the  method has not been  rigorously  evaluated.   Several



variations on this approach have been discussed 1n Section 2.4.J

                                                               i

    The  dose addition  model  can  be  extended beyond  two substances.   The



mathematics  of  such  models,  however, are  even  more  complicated, and  data

                                                               -I

requirements  for  fitting  these  models  Increase  substantially  as  well.



Therefore, using  current  Information, these models  can be of  practical  use



only with mixtures of relatively few component chemicals.



    Plackett  and  Hewlett  (1952)  criticize  the  dose  addition i model   on  two



points.   First,  the  parameter  K  Is  Inadmissible for  certain values of  Z,



and . !„.    Second,  this  model  assumes  that  p,   the potency  of  compound  2

                                                               i

relative to  compound  1,  Is fixed  and  constant  for all  organisms under study,



a condition  that they feel 1s unnecessarily restrictive.



4.3.   RESPONSE ADDITION                                       ; -  •



    Response  addition models  were  first  proposed by  Bliss (1939).   In  the



original representation for "Independent joint action"  of the  two chemicals,



Bliss  (1939) assumed that the  two  chemicals acted  on  different physiologic



systems.  This assumption  can be generalized  to functional Independence  of



the two  separate effects,  even  1f  they are on the same organ system.   Define



the following:
                                    4-5

-------
         D, = dose of chemical 1
         PI m proportion of animals responding to D_

and  similarly  for  chemical  2.   Bliss  (1939) noted  that the  proportion  of
animals  that respond  to  the mixture  depends  not  only  on  P,  and P?  but
also  on   the  correlation   between   the   two  distributions   of  Individual
tolerances to the  chemicals.   If there 1s parallelism  In the susceptibility
to the  two chemicals  so that the correlation  1s  1,  I.e., If  the ordering of
the  animal sensitivities  to chemical  1   1s   the  same  as  the  ordering  for
chemical  2,  then  the  most  toxic  chemical will  elicit  the  response  first.
The mixture response 1s then:
                                P = max(P1,P2)
(4-15)
As  noted  1n   the  U.S.  EPA  (1986a)  mixture  guidelines,  1f  the  tolerance
correlation is -1, I.e.,  if  the  animal  least sensitive to chemical 1 is most
sensitive to chemical 2, and so on throughout the range of sensitivity, then
                               P  =  min(P1i-P2,  1)
(4-16)
For P<1, equation 4-16 is the simplest response addition formula:
                                  P  '  Pl  *  P2
(4-17)
Other tolerance  correlations  give P  values  between these extremes.   If  the
correlation  is  0,  then  one  obtains  the  familiar  model  for  statistical
independence:
                                    4-6

-------
                             P = Pl * P2 ' (P1P2)
    Hewlett and Plackett (1964) discuss a  different  class  of models based on
combining  responses  of  two  chemicals.   Instead  of  counting
        (4-18)
the number  of
animals  responding,   they  model  the  number  of  tissue  receptors  that  are

affected by  the  chemicals.   Their fundamental assumption  Is  that  the tissue
                                                                        .
damage can  be  described by  chemical  complexlng of  the  tissue  receptor  with
                                                              i
the administered  chemical.   The manner of competition  of chemical  molecules

for  these   tissue  receptors;  is  assumed  to  be  described by  laws  of  mass
                                                              I
action,  so  that  key  model  parameters  are  the  chemicals'   dissociation

constants for  complexes.   Their model also  assumes that  a  quanta! response

occurs  only when  an  underlying  graded  response,  E,  exceeds  some critical

threshold,   E .   The  model  assumes  that  the joint  action of  two  compounds
             \f
Is the  result  of  competition  for  the same set of  receptors. !  Using  Hewlett

and  Plackett 's  (1964)  notation,  let  m,  and m-  be  the reciprocals   of  the

dissociation  of  the  receptor-compound  complexes  for  compounds   1   and  2,

respectively,    and    let   u,    amd   w^    be     the    respective    molar

concentrations of  these  compounds at their respective  sites  of action.   The

graded response to compounds 1  and 2 Is

                                                          o-)           (4-19)
                               ,,
                                                 -,-1
where
                                              1,2,
A.  is  the  intrinsic activity  of  compound  i,  and  [r]  is  the  total  molar

concentration  of  receptors.   Let  u. .
                                     [-1

-------
     When  compound 1 1s an  agonist and compound 2  Is an  antagonist  (Case 1),
 quanta! response  occurs when
                               to-, /(I
                                                i
(4-20)
 If   w^   Is  log-normally   distributed   In  the  population  of   Individuals
 considered  then  the model  for  the  normal  equivalent  deviate  of  response Is
                         Y = e + e'log (o.j/0
 (4-21)
If  the  relation  between  the acting  concentration  of  compound  1,  u.,  and
the administered amount, Z.,  1s assumed  to  be
                                  i -
then equation 4-21 becomes
                                                                        (4-22)
Y =
                            a-, + B1 log(Z.,/(l
(4-23)
If both compounds are agonists both elicit the  same maximum response  (Case  2)
then quanta! response occurs when
                        ,u,
(4-24)
where  K  1s   the  critical  graded  response  for  compound  1  alone  and  for
compound 2 alone.  Thus quanta! response occurs when
                                   i        i
                               (0,7(0,  + (o-Ao,, > 1                        (4-25)
                                    4-8

-------
Let
              .  Then the nonresponse proportion is
                               q  =
                                            < 1}
(4-26)
which   can   be   evaluated   If   w-,   and   o>2  have   a   blvariate  normal
distribution.  If  the maximum  response  attainable by compound  1  1s greater
than that attainable by compound 2 (Case 3), then quanta! response occurs If
               n^  -i- n2o>2)/(l +• m-jo^ + m2&>2) > n^o-j/O + m^
                                                                       (4-27)
As   n2-»0,   Case   1   results.    As   (n2/m2)-^(n1/m-1 ),   Case   2   results.
Otherwise,  subjecting this  model  to  the  same  derivations  as  for  Case  1
results  1n  "a model  which has  doubtful  practical  value  on  account  of the
number of parameters Involved" (Hewlett and Plackett, 1964).
    The parameter q Is evaluated  by  integrating the blvariate normal density
function  over  the appropriate  region.   Subsequent analysis of dose-mortality
data then uses the  log-dose-probit response  line, which is curvilinear under
independent  action  of the two  compounds  and  1s skewed  upward  as response
increases and  the  correlation coefficient between  the  action tolerances for
the  two  compounds  decreases.   Bliss  (1939)  notes  that curvilinearity  of  a
                                                               i
dose-response  curve  is  difficult  to  test  in experimental  data,  and  Ourkin
(1981) attributes the paucity  of studies with  examples  of response addition
to this difficulty.
    Similar to the dose addition  models,  the  response addition models  can be
easily  generalized  to  more  than two  chemicals.   The  complexity of  such
models  and  the  accompanying  extreme data  requirements,  however,  make  such
models of little practical use.
                                    4-9

-------
4.4.   GENERALIZED LINEAR MODELS
    For  the ordinary  linear  model,  "Interaction" Is  taken  by statisticians
to mean  a  departure from response additivity assuming  a  normal  distribution
of  the response  variable.   For  generalized  linear  models  (e.g.,  logistic,
log-linear,  log-probit  and  multistage models),  interaction  is taken  to mean
a departure from additivity for  a  transformation of  the  response  variable.
For Instance, In the log-linear model
                           = m
                                      b, t d.,, 1=0,1, j=0,l
(4-28)
where   a0-b0-d00-d10=d01-0.
                               and   where   p..   denotes   the   proportion
responding in  the  group  receiving dose level 1 of compound  1  and  dose  level
j  of  compound  2,  the   d,,   term  describes  the  presence  and  extent  of
Interaction between compounds 1 and  2.  Similarly,  in the logistic  model
                               m  m +  a1  +  b^  +  d^,  1=0,1,  j=0,l
                                                                       (4-29)
                                                               ll
                                                                    descr1bes
where   p^   is   as    before,    a0=b(fdoo=d10=d01=0'   and
the Interaction.
    Although  the  examples  of  generalized  linear  models  given  above  are
applicable only to experiments with  simple binary  mixtures, these  models  can
be extended to experimentation with  three or more  compounds.   The  difficulty
in doing  so  is  not in the mathematics, but rather  time and expense  Incurred
in the conduct of appropriately designed factorial  experiments.
    Use of  fractional  factorial  designs can be  used more economically,  but
still  can  be  lengthy  if  whole  animal  lifetime  studies  are conducted.
                                    4-10

-------
Moreover,  fractional   designs  also  assume  that  one  or  more  higher  order
interactions  are  zero,  when  information  on  all  interactions  may be  the
object of the exercise.
                                                                i
    Nonlinear terms  can  also be incorporated  into  generalized linear models
•
and  the  Box-Tidwell  fitting technique  can be applied  to  obtain parameter
estimates  (McCullagh  and  Nelder,  1983).   In  particular,  if c|(X;e)  is  the
covarlate of  interest where e  is  unknown, the expansion  of g(X;0) about an
initial value eQ is obtained to derive the  linear approximation,

                     g(X;e) - g(X;e0) *  (e-e0)[ag/ae]0=eo              (4_3Q)

Therefore, if the model contains a nonlinear term of the form   :
                                    Bg(X;e)
then replace it by two linear terms of the  form
                                    IJu •«• yv
where  u = g(X;eQ)
       v = [ag/ae]e=e                                           i
       Y = B(e-e0)                                              |
The estimation procedure for e is then iterative as follows:
    1.  Fit the generalized  linear model with  covariates u and v
    2.  Obtain e, = e0+Y/B as; the improved  estimate             [
    3.  Iterate to convergence
McCuTlagh  and  Nelder  (1983) noted  that this  technique  is highly useful and
probably  under-used,  but cautioned  that this  method  is  not appropriate for
the   inclusion   of   many   nonlinear   terms  since  the  estimates  of  these
parameters  will  have large  sample  variances and  will  usually  be highly
correlated with the linear parameters and  possibly with each  other.

                                    4-11                        ;

-------
    Elashoff  et  al.   (1987)  describe  a  modification  of  the  proportional
hazards model  to  allow for the Incorporation of  competing  risk  for  death  to
evaluate interactions  between two  chemicals  in  a  2x2 carcinogeniclty experi-
ment.  For  the analysis of tumor  incidence  data, they  test  for  Interaction
using the additivity index (Wahrendorf et al.,  1981)  as follows:
                                                                       (4-31)
                                                                           10
where  qQQ  Is  the  background  probability of  not  developing  a  tumor,  q
and  qQ1  are  the  probabilities  of developing a tumor when  compounds  1  and 2
are  administered alone,  respectively,  and  q,,  is  the  probability of  not
developing  a  tumor  when  compounds 1  and 2 are administered concurrently.  If
I>0, synergy  is  said  to  have occurred, and if I<0  then antagonism is said to
have occurred.
    The time-to-death data  is  important to consider in  addition to the tumor
Incidence  data when  lethal  nontumorigenic  toxicity in  the doubly  exposed
group  relative  to the singly exposed  groups  is  excessive since It can  cause
a negative bias in I.  Therefore, they used the proportional hazard model

           Pr(survival without tumor at T years for treatment ij) =
                        exp[-jj(h'.j(t) + hQO(t)) dt],                 (4-32)

where h1 represents the  incremental  force  of  mortality  due  to treatment.   To
test for interaction they use the null model
                       h'10(t) + h'Q1(t) - h'n(t) = 0
                                                                       (4-33)
                                   4-12

-------
A  test  statistic developed  by Korn  and Liu  (1983), which  uses a  Mantel-



Haenszel approach,  1s  then used  to  test for no  Interaction with respect  to



time to death.



    Generalized  linear  models  have also been proposed for  multi-effect  data



on  the  complete  mixture.   The  responses  are  graded  (nonquantal)  and  the



overall  toxlclty  of   the  mixture   1s   assigned  to a  severity  category



(Hertzberg, 1987).   A  link function  transforms the  response  frequencies  for



each dose  In  each severity category, and the  transformed  response for these



ordered categories  Is  then regressed on a  linear  function  of|dose (duration
                                                              i


could  also  be  Included  as  a  covarlate).   For  example,  1f   effects  are
                                                              l


categorized  as  "none,"  "mild,"   "moderate" and  "severe,"  and 1f  "mild"



effects  were   considered  tolerable,  then  one  could determine  the risk  of



"moderate  or  severe" effects  for  a given mixture  dose.   The model 1s similar



to  those discussed  previously.  For  example, for  the logistic  link function,



the counterpart  to equation 4-29  1s
                                      - b*[ log(D) - log(D)
(4-34)
where  D now  denotes  the dose  of  the complete mixture,  the  overbar denotes




the  mean  of  log(D),  and  j  denotes  the  severity  category.,   The  response
                                                              I
                                                              I


variable G 1s a function of the mixture dose D and represents the cumulative




response  frequency at  dose D,  I.e., the  organisms  responding  at  severity




level  j or  less.   If  P.  Is the  proportion of animals  responding  to dose D
                        K


at severity  level  k, then the transformed response Is
                                                                       (4-35)
                                     4-13

-------
 The  probabilistic risk  estimate  from such a  model  is obtained by  inverting
 the  link  function, to  give  the  risk  of an  effect worse than category j,
                   p(J>j) =  1  -  expCF^D}] /  (1  t
(4-36)
where  F  represents  the right-hand  side of equation 4-34.
    A  mixture of  chemicals  is  likely  to  Induce several  different  kinds of
effects  1n  different organs.   Applying the  previously  discussed  response
models  to  each kind  of  effect,  even if data were available  on the complete
mixture, would  generate  several  dose-response  curves, and would require some
statistical  combination  algorithm  to address  the  multiplicity  of  effects.
The recasting of the risk  problem using severity categories  is mathematic-
ally simpler, and  also avoids  the  difficult  issue of correlation of specific
toxic  effects across  species.   The  risk  assessor  then  evaluates  only  the
risk of  general systemic toxicity,  e.g.,  the  risk  of  unacceptable  effects.
This procedure  also allows  the  toxicologist to  assign  multiple effects to a
higher  severity category.   For  example, "mild"  effects  in  several  diverse
organs   and   tissues  would  be  deemed   "moderate"   and   unacceptable  when
considered as a composite toxic  response.
4.5.   RESPONSE SURFACE MODELS
    When a 2x2  factorial design  is  used  to  study the interaction of  two com-
pounds,  no information is gained about how  the  response changes with changes
in the magnitude  of exposure to both compounds  1 and 2.   If  the dose ranges
for  these  compounds  are  more  completely  studied, however,  the  economic
requirements   increase as well.   For instance,  with  only  three  nonzero doses
of each  compound in a binary mixture,  16 treatment groups must  be studied if
all possible  combinations of  the   two  compounds are used.   An  alternative
                                    4-14

-------
approach Is motivated  by  conceptualizing the response to  the  joint exposure
as  forming  a  surface  over  the  experimental plane  with peaks  and valleys.
Designs that maximize  or minimize  this  surface  by sequential  exploration are
called response surface models.  They are  most  frequently  used 1n Industrial
experimentation where  the  response can  be  measured quickly and where a small
number of factors  are  to  be combined.   Thus, their  utility for the study of
mixtures  of  even  moderate  complexity  or  for  use  1n  long-term  toxldty
studies Is questionable.
4.6.   SUMMARY OF INTERACTION DATA BASE
                                                               i
    A  survey  of the  statistical  methods  utilized In studies  pertaining to
mixtures was  conducted using  those  papers  Included In the U.S.  EPA Inter-
action data  base  (U.S. EPA, 1988) as well  as  papers retrieved: subsequent to
the construction of  the data base.  A  total  of 462 relevant references were
                                                               i
Included  In this  survey,  which  also  examined  the  type of  mixture studied
(binary,  simple,  or complex),  whether   the  study was descriptive  or mecha-
nistic  In  Its  approach,  and whether  the mixture  Included carcinogenic
                                                               i
compounds.   A relevant reference  was  considered  one  In which both methods
and data  were presented,   I.e., abstracts  and reviews  were not Included.  Of
the  331  references  contained In  the   Interaction  data  base,  307  were con-
sidered  relevant.   An additional  155   studies  were  also  Included  In  this
survey (Table  4-1).
    A  summary of  the  types of studies  examined and  the  statistics used In
each  Is  presented In Table  4-1.  Individual  columns are  used  for   those
papers  found  In the  data base and  those  not  Included  so  that an  exclusive
analysis  of the, data  base can  be  made  separately.  The first  group of  cate-
gories  pertains to  the  general  characteristics of  each  Individual study.
                                     4-15

-------
                                    TABLE 4-1
                  Survey of Interaction Studies Methodologies*
                                  Data Base
                       Other
Total
Percent
 Number of studies
           307          155
Nature of Individual Studies
                                                             462
Binary mixture
Simple mixture
Complex mixture
Descriptive
Mechanistic
Noncardnogen
Carcinogen
Statistical
Student's t-Test
No Statistics
Statistics Not Specified
Analysis of Variance
Ch1 -Square Test
Neumann-Keuls Test
Mann-Whitney U Test
Wilcoxon Test
Duncan's Multiple Range Test
Fisher Exact Test
Tukey's Test
Dunnet's Test
Kruskall-WalUs Test
Least Significant Difference
Finney Additivity Formula
F Test
Scheffe's Test
2-Sample Rank Test
Fisher-Yates Test
Mantel-Haenszel Procedure
294
24
17
276
61
261
46
Breakdown of
85
85
71
34
17
13
7
3
9
4
3
3
3
3
1
2
1
1
0
0
150
16
7
136
50
118
37
Individual
52
35
35
16
12
3
5
7
1
4
3
2
2
1
3
1
2
0
1
1
444
40
24
412
111
379
83
Studies
158
119
106
52
29
16
12
10
10
8
6
5
5
4
4
3
3
1
1
1
96.1
8.7
5.2
89.2
24.0
82.0
18.0

34.2
25.8
22.9
11.3
6.3
3.5
2.6
2.2
2.2
1.7
T . 3
1 1
1*1
1 .1
0.9
0.9
0.6
0.6
0.2
0.2
0.2
*Refer to text for explanation of individual categories.
                                    4-16

-------
Interactions  result  from a  binary  mixture  (two  constituents),  a  simple
mixture (more than  two  but less than dozens  of  Identifiable components),  or
a complex mixture  (dozens or  more constituents, many  of  which  are unidenti-
fied or present  1n low  concentration).   Several  studies used  more  than one
type of mixture  Involving,  In most cases, the effect  of  one compound on the
Interaction  of  two  other compounds.   In other  Instances,  the  Interaction
between two  single  components,  e.g., carbon  tetrachloride  and  phenobarbltal
(binary mixture), as well as  the  Interaction  between a single compound and a
mixture of compounds, e.g., carbon  tetrachlorlde  and PCBs (complex mixture),
would be Investigated 1n  the  same  study.   The total  number  of evaluations  1s
then much larger  than  the number of references,  although It Is obvious that
an overwhelming number of evaluations pertain only to binary mixtures.
    These studies  are  also segregated as  to whether  they  analyze  an Inter-
action mechanistically,  descriptively  or  both.   A descriptive study  1s one
that  only  looks  at  one or  more  toxic endpolnt(s)  to  characterize  the
magnitude of  the Interaction without  examining  the underlying  cause(s) for
the  Interaction.   Such   endpoints  commonly  Include  LD5Q  values,  serum
enzyme levels, and  sleeping  times.   Mechanistic studies, on  the  other hand,
attempt  to   quantify  changes  in  the  absorption,  distribution,  metabolism,
                                                              i
excretion,  receptor  binding,   or  physical  characteristics  of  a  compound.
Examples  of  mechanistic endpoints   Include  urinary  metabolite  profile,
Intestinal  absorption,  hepatic enzyme  activities, and  tissue  distribution.
Several  studies  incorporate  both approaches  by  attempting  to correlate  a
change in toxicity  with  the biological or  chemical bases  of the Interaction.
For  example,  several studies  have examined  the  effects of certain  enzyme
inducers such as  phenobarbital, 3-methylcholanthrene, or PCBs  with  a change
                                    4-17

-------
 In hepatotoxldty Induced by carbon tetrachlorlde.  Table  4-1  Indicates  that
 61 studies  (412 descriptive  + 111 mechanistic  - 462  total)  utilized  both
 strategies.
     Finally,   the  number  of  studies   involving  carcinogenic  endpoints  was
 determined.   A carcinogen  study  is defined as one  in which a determination
 of tumor frequency,  latency or  incidence  is  made.   Studies  in  which known
 carcinogens   were  used   but were   not   of  sufficient  duration   for  tumor
 formation  were  included in  the noncarcinogen  category.    Unlike  the other
 categorizations,  a  study was classified  as  either carcinogen or noncarcino-
 gen but  not both.
    The  use  of  statistical  methods as  specifically stated  in  either  the
 methods  section,  in  tables  or figures,  or in the text was  tabulated for each
 study.   As  reflected  1n Table  4-1,  the most  widely  used  procedure  is  the
 Student  t-test,  which was  utilized in  over one-third of  the studies.  This
 test was  frequently  used 1n  conjunction  with  other  methods such  as analysis
 of  variance   (ANOVA).  Most  often,  however, the  t-test  was  the  only  method
 employed.  A  noteworthy  finding was that one-quarter of  the  studies  1n  the
 survey  contained no  reference  to  any  statistical procedures.   In  addition,
 nearly  23%  of  the  studies  did not specify  the  type  of  statistical  tests
 used.  In these  cases,  either the  p values were  given in  the  text  or  in  the
 footnotes to  tables  or  figures  without explanation or the use  of statistics
was  referenced to another  source.   In  one study,  the  authors  stated  that
 "statistical  comparisons  were made  by  standard  procedures"  (Cerklewski  and
Forbes,  1976).  Table  4-1  indicates   that  83%  of   those  studies  examined
either used  no statistics,  did not specify the  statistical methodology  or
used Student's t-test.
                                    4-18

-------
    The other statistical tests employed  in  these  studies  are also listed in

Table  4-1.   Because many  studies  used  more than  one procedure,  the  total

number of  individual  tests  is  greater than  the  total number of  studies In
                                                               !
the  survey.   Nearly  37%  of the  studies used  a  method  other  than or  in

addition to  the Student t-test.   No  attempt will  be made here  to define or

characterize each  method nor critically  assess the  appropriateness of  these

tests  for  interaction studies except  for  the use  of Finney's (1971) equation

(equation  4-8   with  p=0)  for  joint  toxic  action.   Four  studies  used  this

additivity  model   to  calculate the predicted  LD5Q values  for  a  number of
                                                               i
binary  mixtures.    Ratios  of  predicted  to   observed  LD5Qs   were   calculated

and  a determination was made as  to  the  significance of the  deviance   from

additivity.   Keplinger   and  Deichmann (1967)  determined the  acute toxicity

Induced  by combinations of  two and  three pesticides and reported  that while

most  of  the combinations  induced  essentially  additive effects  in mice and

rats,  there were  cases  of less than  or  more than  additivity.   Pairs  of 27

industrial  chemicals  tested for   joint  toxic  Interaction  demonstrated   that

the  additive model  reasonably  predicted the  toxiclties of a  majority of

these  binary mixtures  (Smyth et al., 1969).   Departures from additivity  were

reported by Withey and  Hall (1975) who  investigated  the joint toxic action

of  perchloroethylene with  benzene or toluene and by Freeman  and Hayes (1985)
                                                               i
who  observed the  potentiation of  acute acetonitrile  toxicity  by  acetone.

     A handful   of  other  studies  has also attempted to quantify toxic inter-

actions  in  terms  of   deviation   from  an additive  response, j  An  undefined
                                                               i
additive model  was employed by Woolverton  and Balster  (1981) to  Investigate

 the   effects   of   combined   ethanol  and   1,1,1-trichloroethane   exposure.
                                     4-19

-------
 Wysocka-Paruszewska et  al.  (1980) used  the  coefficient of combined  action,
 defined as  "the ratio  of the  calculated  LDcn  on  the  basis  of  LD,.  of a
                                              bu                      50
 single  compound  to  the  experimental  LD50"   to  evaluate  the  toxicity  of
 thiuram in  combination  with  several  other  pesticides.  Derr  et al. (1970)
 used  a  response  addition approach  in which  the mean  heart  or body  weights
 for   individual   treatment groups   (minus  control  values)  were  added   to
 calculate  the expected  combined  response to  cobalt  (cobaltous chloride) and
 ethanol  exposures.   The  observed and calculated weights  were then  compared
 using  a  Student t-test.   The  effects  of  prophylactic  protection   against
 cyanide  intoxication  were evaluated  using  potency  ratios  defined  as the
 LDgo   of  KCN  with  antagonist(s)  divided   by  the  LD5Q  of  KCN   without
 antagonlst(s)  (Way  and  Burrows,  1976).   The  results  of the  above   studies
 were  varied  in  that  additive,  potentiated  and  antagonistic  effects  were
 observed depending  on the  mixture components and concentrations.
 4.6.1.   Description  of  the Mixtures  Data  Base Sample.  The  use  of   statis-
 tics  in  the U.S. EPA mixtures  data base has been described  in the previous
 section.  A  10%  random  sample  of papers from the U.S.  EPA mixtures data base
 was taken  to  review  the  quality  of  experimental design, use  of statistics
 and ensuing  conclusions.   The  sample was  stratified by classification of
 type of  statistics  used;  there  were 32 papers  assessed.  A detailed critique
 of these papers  is  contained in Appendix C.   It is  important to note that if
 an investigator  used  a  poor experimental  design or  inappropriate statistical
 analyses,  the conclusions regarding  the interaction  are  suspect.   Unfortu-
 nately,  it is  impossible  to determine  if  the  conclusions are correct  without
access to the raw data for re-analysis.
    In summary,  there was  no  use of statistics  in 8 studies,  the statistics
used were  not specified in 7,  no statistics  were given  in 2  abstracts,  and
                                    4-20

-------
no quantitative  data were  given  1n 1 paper.   Of  the remaining  papers,  the

ones  that  described  their  statistical  methods,  the  methods  used  were

Inappropriate  In  9  and there  was  no baseline  control  1n 4 papers.   In  one

paper, the design and  use  of  statistics  were appropriate with  the conclusion

justified.                                                    !

4.7.   CRITICAL ASSESSMENT EXAMPLE                            ;

    As a  further assessment  of the  quality of statistical analysis  in  the

mixtures  literature,  one  paper was  selected  for Intensive scrutiny.   The
                                                              j
study by  Eybl  et al.  (1984) was chosen  because of Its detailed descriptions
                                                              I
of the toxlcologlc and statistical methods employed.

    Eybl  et  al.  (1984)   Investigated   the  Influence  of  several  chelating

agents  on  the acute  toxlclty of  cadmium  (Cd).   As will  be  shown  in  the

following  discussion,  the  experimental  design  and  the  statistical  methods
                                                              i
                                                              i
used  were  Inappropriate  for  characterizing  the   Interaction  for  risk

assessment  purposes, and  were in  fact  inadequate for  some of the authors'

goals as  well.  Eybl  et  all.  (1984) examined effects  on mice  and rats; only

the  mouse  experiments are  discussed here.   Characteristics |  common  to  the
                                                              I
mouse test  series were as  follows:


       species:    male mice  (SPF,  Velaz Prague),  20.22  g  body  weight
       route:      i.v. (single injection)
       chemicals:  Cd  with any of  six chelatlng agents or  combinations
       endpolnt:   survival rate at 10 days


4.7,1.    Experimental  Conditions.   The  first  series studied;  the  effect of

single  chelatlng  agents  on  survival   of   mice   injected  with  CdCl2.   The
 conditions  were as  follows:
     Groups:
20 mice per exposure group
                                     4-21

-------
     Exposure:
     Statistical method:
toxicant-    CdCl2»2.5H20  single  subcutaneous
Injection  20 mg/kg;  Inhibiting agent-  single
1ntraper1toneal  Injection  at a molar  ratio  of
25:1 (chelator:CdCl2)
unspecified,   probably   Fisher's  exact   test
(Fisher's   exact,   Chl-square,   t-test   all
mentioned 1n Methods section)
    The  conditions for the  second,  third and  fourth  series  were similar to
 those  of  the  first  series..  The second  series  Included  three  dose levels
 (molar ratio  of  1:1,  2:1,  5:1).  The third series used one dose level (molar
 ratio  5:1) but two treatment  sequences (simultaneous vs.  2  hours after the
 Cd  Injection).   The  fourth  series used  one  dose level  for  single chelator
 effects  and a different  dose  level  for  effects of  two  chelators together.
 For example,  ZnDTPA  and  DMSA were tested  Individually  at a  molar  ratio of
 2:1, while  the combination ZnDTPA+DMSA was tested at a ratio of 1:1:1.
 4.7.2.   Discussion  of  Design.   This  first  series  seems  to  have  been
 Intended only  to  screen for the most  effective Inhibitors (antidotes) of Cd
 toxldty.   Cadmium Is  always administered at the same dose,  and  each of the
 chelators  1s  administered at  only one dose  level.   Consequently,  no dose-
 related  Interaction  can be  determined.   The  authors apparently  assume  that
 the data are  similar  to data on treatment regimens  for a disease, where here
 the disease is  Cd toxlcity  and the  treatment  is one of  the  chelators.   The
 "disease-treatment" interpretation, however,  requires  the assumption that Cd
 lethality  occurs  only at   20  mg/kg  or   more,  and  that  the  administered
chelator levels  are the  standard  antidote dosages.   None of these  assump-
tions   has  been demonstrated in this  paper.    Consequently, any  conclusions
are then specific to the doses used.
                                    4-22

-------
    No models  were presented  by the authors  as a  means  of estimating  the
"expected" response from  the Cd-chelator combination.  Models  of  the  inter-
action between a  chelator  aind Cd cannot be applied  to  the  data as presented
since results  are not given  for a control group  (no CD,  no  chelator),  nor
for exposure to a  chelator  alone (no  Cd).   A  key unstated  assumption is that
all the chelators are administered at nontoxic doses.
    The  statistical  test  used  is  not stated, but can  be  assumed to  be  the
Chi-square  or  Fisher's  exact  test.   These  tests  are  consistent  with  the
interpretation of  the experiment as  if  it were the  treatment; of  a  disease.
As  further  confirmation,  use of  Fisher's  .exact  test  in  recalculating  the
significance  levels  showed  agreement with  Eybl's  published  values  (Eybl et
a!.,  1984,  Table  1)  except  for the group 4  to  group 6  comparison,  which
should  show p=0.02,  I.e.,  it should be  footnoted  by an asterisk to  denote
p<0.05, not p<0.01.                                            ;
                                                               i
    The   preceding  comments;  also  apply   to  the  other  test  series.   In
                                                     ,
addition, the  doses  (molar ratios) used  in the second and third  series were
the  same for all  chelators,  regardless  of each one's  inhibitory effective-
ness.  The  doses  used in  the fourth series were selected to provide the  same
number  of moles  of  mixed  chelators  as  used  in  each individual  test.   No
model  has been located  that uses such  a dose  selection  in a mixture  study,
and  the  authors do not provide  any  justification for  these doses.
4.7.3.    Discussion  of Results.   The reported  results  for all  four   series
include   the  survival fraction  (n/20)   and significance level  (percent) of
various  differences  in survival  rate.   Several comparisons  are  made  in the
first,   second  and   fourth  test  series  with  no  adjustment   for  multiple
comparisons.   The  Importance of  the multiple comparison  problem is   easily
demonstrated  with.the first  series.  Note  that two comparisons are reported
                                     4-23

-------
between  chelators,  suggesting  that  all  chelator  survival  rates may have been
compared  with  one another  but  only a  few  comparisons  were reported.  There
are  6!/(2!  x  4!)  = 15  such  palrwlse  comparisons  In addition  to  the  six
comparisons  between  each  chelator and  Cd alone.   At  a  decision significance
of  0.05,  one  of  the  21   comparisons  can  turn  out  to   appear  significant
through  random chance alone.  So  one of the  six  significant  findings could
be  circumstantial  and not  due  to actual differences in  inhibition.   If  the
decision   rule  is  to  require  significance  of  at  least  0.05,  then  the
chelators  showing  survival  increase  at  a significance of  0.01  or  lower would
probably  be  significant   after   the multiple  comparison  adjustment.   The
finding  that was  significant at  0.05 but not 0.01  is suspect.   In the fourth
series,  the multiple comparison  issue  is  not as  strong,  since only  six
possible  comparisons  could  be  made;   the  reported  significance  levels,
however, are still inaccurate.
    In  addition  to  compensating  for   multiple  comparisons,  the  analysis
should  have  used  survival  time  (when  the  animals died)  instead of  the  end
survival  fraction.   In  addition  to  using  more  data   1n  the  statistical
analysis,  comparing  survival  curves  would  have  also provided  more  informa-
tion for  studies  on  the mechanism and  pharmacokinetics of  inhibition by  the
chelators.
    Use of a single  dose  level  in the  first  series  Is  justified  for  screen-
ing purposes.   The analysis of  the  second  series should have combined  the
dose levels, instead  of merely reporting pairwise  comparisons.   For  example,
if  the  different  series  are  assumed to be comparable and the groups  com-
bined,  then  the  dose-response  data  appear as  in Table  4-2.  The  dose selec-
tion for  the  fourth series  could  then have been  made  according   to  some
interaction model  so that  the  response to  the  combined  chelators could  be
                                    4-24

-------
                                  TABLE 4-2                   ;

        Combined  Results  for  CaOTPA and DMSA  Inhibition  of  Cd Toxicity3
Chelator
Doseb
Survival
Surviving/Total
CaDTPA





DMSA


1
1
2
5
5
25
1
2
5
60.0
60.6
80.0
81.8
86.7
85.0
3.0
35.0
100.0
1 12/20
i 20/33
16/20
27/33
1 13/15
17/20
1/33
7/20
33/33
aSource: Eybl et a!., 1984

bDose is molar ratio of chelator:Cd with Cd administered as  CdCl2»2.5H20  at
 20 mg/kg                                                     ;
                                    4-25

-------
 predicted.   For  example,  for a response addition model where  Independence of
 action  1s  assumed, the doses  used  In Individual testing would be duplicated
 In  the mixed  test (1f the  molar ratio  of  chelator:Cd of  2:1  was  used for
 each  single  chelator  test,  then   the  mixed  exposure  ratio  of  chelator:
 chelator:Cd  should be 2:2:1).   For  a dose addition model where similarity of
 action  1s  assumed,  the  mixed exposure  would use doses  scaled  according to
 potency,  where  the  summed  scaled   dose  of  the  mixture  would  have  been
 previously tested  for one of  the single  chelators.   Instead,  since the dose
 selection  was  not  justified by the  authors,  and since  no  predictive model
was  presented,  the authors'  conclusion that  "the  additive effect  of these
 two  chelatlng  agents was  demonstrated" 1s  false.   In general,  the  conclu-
 sions  throughout  this paper  are much weaker  than  they could have  been had
adequate design and analysis been Implemented.
4.8.   SUMMARY
    In summary, statistical  methods  that  have been  used for  assessing Inter-
actions  among  components  1n  chemical  mixtures have  been  examined.   This
review  Indicates  that proper  experimental  design Is  Infrequently  utilized,
and that statistical  techniques  are  rarely chosen appropriate  to the  experi-
mental  data.   In  particular,  current  techniques   for  Investigating  the
presence and extent  of  Interactions  in  complex mixtures  are  Inadequate,
Impractical or Impossible  to apply.   At best, practical  design  and  analysis
techniques  can  be applied  to  characterize  Interactions  in  the  experimental
dose ranges only among constituents  of simple mixtures.
                                    4-26

-------
               5.   DISCUSSION  AND  REASSESSMENT OF  THE  GUIDELINES
5.1.   OVERVIEW                                               j
                                                              i
    This  chapter  reviews  and reevaluates  the  current Agency  guidelines  on
mixtures  based  on the  Agency's  experience  In  applying these  guidelines  as
well  as  considerations  of new information that has  been obtained  and  new
approaches  that  have  been  proposed  since the  guidelines were  developed.
Revisions  suggested  in  this  chapter along  with  other comments  received  by
                                      •
the Agency will be considered for  future incorporation into the guidelines.
    Based  on  the mechanistic  considerations summarized in Chapter  3,  toxic
interactions may  modify significantly the  toxic  and  carcinogenic  potency  of
environmental contaminants.   The  types of  information  available  for quanti-
tatively assessing the  magnitude  of  such  Interactions  as  reviewed 1n Chapter
2.,  however,  are  not  extensive.  While  appropriate mathematical  models  and
statistical  techniques  are available  to  quantify some simple  binary inter-
actions,  these  methods  cannot  be extended  to  complex mixtures  because  the
data  requirements  of  such extensions  lead  to experimental designs  that  are
impractical.  In  addition, mathematical models  for  quantifying promotion  and
cocardnogenlc  efficiency  that could  be used  to systematically  assess  and
compare  the  quantitative  significance  of  these  phenomena   have  not  been
developed.   Those quantitative estimates of compound  interactions  that  can
be  made  suggest that most  interactions  are within a  factor  of 10  of  those
that would be predicted  based on  the assumption of  no Interaction.  The data
on which this generalization  is based, however, are limited.
    The  preferred approach presented  in  the guidelines for  conducting risk
assessments  on mixtures  Is   to  use  Vn  vivo  toxicity data  on  the mixture
itself based  on the  route of exposure and  duration period of concern.   This
                                    5-1

-------
 remains  the preferred approach,  as long as  certain  factors such as masking
 of  toxic or carcinogenic effects are considered.  Nonetheless, this approach
 will  not  be  practical  In  most  cases  because  adequate toxidty  data are
 available  on only a few complex  mixtures.   While the concept of  "sufficient
 similarity"  may  be able  to extend this approach  somewhat, this approach will
 stm  be restricted  to  a  few  well-studied groups of  complex mixtures (see
 Appendix B).
    The  use  of an assumption of dose or response addltlvlty as the basis for
 risk  assessments on mixtures  remains a  useful,  and  1n  many  cases  the only
 practical,   approach.    Some   mechanistic   considerations  suggest   that
 addltlvlty  may  be a  plausible assumption  1n  the  low-dose  region  because
 thresholds   for  many  types  of  Interactions  are  expected  to   exist.   In
 addition, many acute bloassays  on  binary or simple mixtures suggest that the
 dose  addltlvlty   often adequately  accounts  for mixture toxlcity based  on
 gross toxic  endpolnts.   Nonetheless,  the credibility  of this approach dimin-
 ishes as the number of components  1n  the  mixture Increases  because for many
mixtures  the  toxlcity and perhaps  the  Identity of  all components  are not
 known.
    Alternatives   to any  of the  above  approaches  are  being developed  and
 explored by  the  Agency and other groups  to  more fully  utilize the extensive
In.  vitro and  short-term In vivo data on  many mixtures.  Two  such  alterna-
tives,   the   "comparative   potency   approach"   and   the  "toxic   equivalency
factor," were not discussed 1n the guidelines.
    The  "comparative  potency approach"  attempts to'calibrate the Ijn  vitro
potency  of  groups of complex  mixtures  to  the  limited   jm  vivo  potency
estimates of these mixtures.   Once a relationship between  in vitro and  In
vivo potency has  been  demonstrated, the results  of jin  vitro assays  on  other
                                    5-2

-------
related  complex  mixtures can  be estimated.   As  discussed below  1n Section
5.3.,, this approach  has  been applied to  the  carcinogenic  potency of combus-
                                               1
tlon emissions and  can  be regarded as  a  more formal  and quantitative exten-
sion of  "sufficient  similarity."  As with  the direct  application  of suffi-
cient similarity, care must  be taken to  ensure  that  the approach Is applied
only to mixtures that are likely to exert effects by the same mode of action.
    The  "toxic equivalency factor" method Involves  estimating the potency of
less well  studied components In a mixture  relative  to the potency of better
studied components,   using data  from  comparable types  of in vitro and 1_n vivo
assays.  So  far,  this  method has been  used only  to  estimate  the toxlclty of
                                                               j
mixtures  of  chlorinated  dloxlns  and  dlbenzofurans   (a  group  of  similar
compounds) by using  the  considerable data on  the  Ui  vitro activity of these
compounds.   The  toxiclty of the mixture 1s   then  estimated  by  summing  the
products of  the  equivalency  factors  and  concentrations  of  the  components 1n
the mixture.  An  estimate of the In vivo potency of  the mixture can be made
by multiplying this  sum  of the  products by  the ^n vivo, potency of the refer-
ence compound,  I.e., the compound  that  served  as  the  basis  for  estimating
the  toxic equivalency   factors  (2,3,7,8-TCOD  In  the  case  of  mixtures  of
                                                               i
chlorinated dloxins).  This  approach can  thus  be  regarded  as  an extension of
the  assumption  of  dose  addltlvlty   and  like   dose  add1t1v1ty  must   be
restricted to compounds  that act by the same mechanism.
    Both of  the  above  approaches are likely  to  prove  useful  as alternatives
or bases  for comparison  with  risk assessments using  the  hazard  index based
on dose or response  additiv'ity  as given in  the guidelines.  As with any type
                                                               i
of analysis  based on jji vitro data, confidence  in  these methods  will  vary
with  the  degree  to  which   the 1_n  vitro  analyses  have  been validated  as
predictors of in vivo responses.                               >
                                    5-3

-------
    None  of  the above considerations fundamentally alter  the basic approach
recommended  In  the  original  guidelines.    All  of  these considerations  do
reinforce  the  underlying  principle  of  the  guidelines:  "No  single approach
can  be  recommended  to  risk  assessments  for  multiple  chemical  exposures.
Given  the complexity  of  this  issue and  the  relative paucity  of  empirical
data from which sound generalizations  can be constructed, emphasis  must  be
placed on  flexibility, judgment, and a  clear articulation  of  the assumptions
and limitations in any risk assessment that  is developed."
5.2.   COMPLEX MIXTURES
    For complex mixtures,  It  is  not  likely that  toxic or  carcinogenic inter-
actions will or can  be quantified  using the  mathematical  constructs given in
Chapter  4.   As discussed  in  Chapter 4 and  illustrated in Section  2.4.,  the
types  of  experimental  designs  that are  required  for meaningfully quantifying
interactions for single pairs  of chemicals are prohibitively  complex for  the
routine  assessment  of  chronic  effects.  For mixtures  containing  tens  or
hundreds  of  chemicals, the proportions  of which can  vary  over  time or among
sources  of  generation,   elaborate  bioassays  for  quantifying  interactions
among components are impractical.
    The  guidelines  currently  recommend   using  data  on   the mixture  or  a
"sufficiently  similar" mixture for  the risk assessment.   In  general terms,
the  determination  of  sufficient  similarity  should  consider   the  chemical
composition  of the mixture,  any  variation  in the chemical  composition,  as
well as  the  toxlcologic  properties of  the mixture  components  and  fractions.
The criteria for  determining  "sufficient similarity"  are  intentionally vague
and are  likely to  vary depending on  the nature  and quality of  the  available
data,  the toxlcologic  endpoint, and the  extent  of human exposure.   A  case
study  applying the concept of  "sufficient similarity" is given  in Appendix
B.  Using this approach,  a  risk assessment  can be conducted  if the mixture

                                    5-4

-------
on  which  adequate  toxlcologlc  data  are  available  is judged  sufficiently
similar  to  the  mixture for which a  risk assessment  is  desired.   For certain
classes  of  complex mixtures on which human  or  animal  data  are available on a
relevant  route  of exposure and  are  adequate  for conducting  a  quantitative
risk  assessment  (e.g., coke  oven  emissions),  the assessment  of  "sufficient
similarity" should be a useful approach.
    For  many  other  classes of complex mixtures,  however,  such In  vivo data
are not  available or  if available are not  by  a route  of  exposure  likely to
occur  in  the  environment.   As currently written, the guidelines  suggest, in
the  absence of  "sufficient   similarity,"  that an  additivlty assumption be
used  for similar-acting components  after  assessing  whether data are  suffi-
cient  for quantifying  any component  interactions.   In practice,  this  will
normally  lead  to an  additivlty assumption.   If the  mixture contains  many
                                                               i
chemicals,  it is  also  likely  that adequate toxicity data  will not  be  avail-
                                                               i
able  on  some of  the  components.   Furthermore,  for  some  highly complex or
highly variable mixtures,  not all of the  chemical  components may  be  known.
The Agency  recognizes  that as the number of components increases and  as the
number of  components lacking  adequate toxicity  data  increases, confidence 1n
the risk assessment diminishes.                                !
    The  use of a  comparative  potency method may sometimes  be preferable to a
simple  additivity assumption  in  cases where  the  criteria  for  sufficient
similarity  are   not  met.   This  method,  as  applied  to  carcinogens^  was
presented by Albert et  al. (1983)  and  was  further refined by  Lewtas (1985).
The underlying  assumption  is  that  relative potencies  among j_n  vivo  and In
vitro bioassays  are constant:
                                  RP-, = kRP2                            (5-1)
                                                               I
where  RP,  and RP~ are  the relative  potencies  of a  compound  or mixture In
                                                                             .
bioassays  1  and 2,  respectively, and k is a  constant.  It is also assumed

                                    5-5

-------
that  a single  number  1s  sufficient  to  characterize  the  response  In  each
assay  and  that  species  show parallel response within  an  assay.   Using these
assumptions,  the  results  of 1_n  vivo mixture bloassays from  which quantita-
tive  risk  assessments  can  be  made are  correlated  with  the  quantitative
results of  in_ vitro bloassays.  This correlation  can  be  used as a "calibra-
tion curve" to  estimate the  jji vivo rate  of response of similar compounds or
mixtures when only  quantitative In vitro results  are  available.   Using this
approach,  Albert  et  al.   (1983)   reported  that  estimates  of  comparative
potency  for coke oven  emissions,  roofing  tar  and cigarette smoke  based on
several  in 'vitro  bloassays   (Salmonella  mutagenlclty assay,  L5178Y  mouse
lymphoma cell mutagenlclty assay and a  sister chromatld exchange assay) were
within  a   factor  of  <2  of  estimates  of  comparative  potency  based  on
epldemlologlc data  for  lung cancer.   Using additional data  from  mouse skin
tumor  Initiation  studies, Albert  et  al.   (1983)  proposed  unit  lung cancer
risks  for  dlesel  and  gasoline engine exhaust  partlculates  based  on  the
relative potencies  of  these  partlculates  In In  vitro  assays.   Lewtas (1985)
extended  this  analysis  to  Include emissions from various  energy combustion
sources.
    As  discussed  by both   Albert  et  al.  (1983)  and  Lewtas  (1985),  the
relative potency  approach makes  several assumptions  concerning  mechanisms of
action  and  dose-response  relationships among the various  types of  Iji  vivo
and  _1n_ vitro  bloassays that  are  used.   These assumptions  and  the corre-
sponding uncertainties must  be weighed against  the  assumption  of  and uncer-
tainties  1n dose or response  addition.   The  relative potency  approach  Is
attractive  because  data  on   the  mixture  of   concern   can  be  generated
relatively  quickly   and  Inexpensively.   In addition,  given  the  Increasing
amount  of  data  available on  the effects  of  mixtures  in  in  vitro  tests,  as
                                    5-6

-------
discussed  In  Section  2.2.,  and the dearth of information on  the magnitude of
toxic  interactions  In vivo, the relative  potency  method offers one approach
to  the problem of complex  mixtures  that  is  amenable to experimental  testing
and validation.
    The  use  of the relative  potency method or  other  approaches  based on In
vitro  or  short-term  in vivo  bioassays seems  to  be  potentially  useful for
assessing  the  biologic  activity  of complex  mixtures.   Only  limited data,
however,  are  available  for  supporting the  quantitative  correlation  of i_n
vitro  and  j[n  vivo relative  potencies and the data that are available  suggest
that  the correlation  between  biological  activity  in  the jjn  vitro assay and
the  In  vivo  assay will not  be  uniform  for  all  types  of  mixtures.   For
Instance,  Salmonella  are known to  be particularly  sensitive to the mutagenlc
effects  of nitropyrene  by virtue of  the organism's  endogenous  nitroarene
reductase  (Mermelstein  et  a!., 1981).   A  comparative potencyjjudgment  of  a
nitropyrene-containing  mixture  based  solely  on  Salmonella mutation  data
would  likely overestimate eukaryotic mutagenic or  tumorlgenic activity.
    An empirical  approach  to  selecting  the  most  appropriate  i_n  vitro assay
for applying  the relative  potency approach  could be  based  on the  use  of  a
battery  of screening  tests, including in  vitro  assays  and  short-term in yivo
assays (NAS,  1988a).   The  quality of the  correlation  in biological  activity
between  the  screening tests;  and  the known in  vivo  relative  potencies  of  a
related  group of  complex mixtures  could  then  serve as  a guide in determining
the most  appropriate  assays   for  applying  the  relative  potency  method  to
other   related complex  mixtures.   The  scientific  validity  of applying  the
                                                               j
relative potency  method  based solely on empirical correlations  is question-
able,   however,  particularly  when multiple  pair-wise  comparisons are  made
among  several in. vivo  and in  vitro  assays.   An  alternative to  multiple
                                    5-7

-------
pair-wise comparisons  has  been  proposed  by DuMouchel and Harris (1983) using
Bayeslan  statistical  methods  to combine the  results  of  multiple jm vivo and
!n  vitro assays.  Nonetheless,  confidence 1n  the  use  of  any In  vitro  or
short-term  i_n  vivo assay  for estimating  environmental  risk will  depend  on
the  extent  to  which  the  assay  reflects  the  mechanism  of  action  and
pharmacoklnetlcs  of  the mixture.   For many  in  vitro assays,  which provide
only an exogenous activating system, this confidence may be limited.
    Furthermore,  1n  many  Instances, the dose-response curves  within in vivo
or in.  vitro assays  for even pure  chemicals are  not  linear over a wide range
of  concentrations or  doses.   Consequently,  a  single  meaningful   "potency"
term will  not  be appropriate  for  comparing  arrays  of nonlinear curves.   If
the  "potency"  Is expressed as  an  estimate of single  slope parameters taken
from the  mid-range or linear portion  of the  dose-concentration curve of the
in  vitro bioassay and such  values are correlated  with  linearized potency
terms  from  in.  vivo  bioassays  with   relatively  few dose  groups  and  small
numbers  of  animals  per  group, the   errors  associated   with   the  estimated
potencies  are  likely  to   be  high  and the significance of any correlation
questionable.
    Notwithstanding these  limitations  and  concerns, the  use of the compara-
tive potency method  or some  analogous approach  based on  in vitro  or short-
term in. vivo  tests  may be  the only  practical   method  for assessing  risks
posed  by  complex mixtures  on  which adequate long-term  in  vivo studies  are
not  available.  The  extent  to  which  the use  of such   an approach  can  be
considered  scientifically  valid  or simply  the application  of  a risk manage-
ment decision  scheme  Is   likely to  vary  depending  on   the  quality  of  the
correlations in  biological activity and  the degree to which a  clear associa-
tion can be  made between mechanisms of action in the screening assays and  in
                                    5-8

-------
the  In  vivo  effect  Induced  by  the  mixture.   Depending On  the number  of
compounds In the mixture of concern and  the  adequacy of the toxlcologic data
                                                             i
on  these  compounds,  1t may  be most  reasonable  to use both the  comparative
                                                             \
potency method  as  well as the  assumption of dose or  response  additivHy  to
gauge the variability between  the two methods and better  express the uncer-
tainty 1n the risk assessment.                                i
    This approach  has generally  been  applied only  to  carcinogenic  effects.
An  application  to  noncancer  health effects  could be reasonably  made  1f the
mechanisms of  action  were similar between  the  effect of concern and  the  jn.
vitro or short-term In vivo bloassays  proposed and If data were adequate for
assessing the  constancy and  the correlation  In potencies  between the short-
term and long-term assays.
5.3.   MIXTURES OF CHEMICAL CLASSES
    As discussed  In Section  2.2., mixtures  of  chemical classes  differ  from
complex mixtures  In  that  the   compounds  in  the  former category are  struc-
turally and  toxlcologically related.   Some  types  of  mixtures  of  chemical
classes are produced  and used as  a mixture  following a reasonably consistent
and well-defined  procedure.   Examples of  such  mixtures Include  the  various
commercial  polychlorinated  and  polybrominated   blphenyls,   toxaphene  and
chlorinated  naphthalene.   Other  types  of such  mixtures  are chemically and
toxicologically  related  compounds  that  are  usually  found  together   In  the
environment but can  vary  substantially in the proportions of  the components
depending on  the  source  of the mixture.   Examples  of  these latter  mixtures
include polychlorinated and  polybrominated  dioxlns  and  dibenzofurans.   This
distinction  between  these  two  mixture   types  Is  Intended  to   reflect  the
different types  of  data  that are  available  or  might  reasonably  be  obtained
on mixtures of chemical classes.
                                    5-9

-------
     Because  some mixtures  are  reasonably  consistent- and   limited  in  the
 diversity  of  their   composition,   data  are  available  on  their  different
 commercial formulations  (e.g.,  Aroclor 1264).   When  data are not  available
 on a specific  formulation,  the  formulation lacking data may  often  be  suffi-
 ciently similar to a formulation for which data  are available so  that  a  risk
 assessment can  be  conducted by  analogy.   For such mixtures, It thus seems
 reasonable to  continue  to conduct  risk  assessments  using  toxicity data on
 the mixture  as  the   preferred  approach.   Nonetheless, data  may  sometimes
 suggest  that   differential  rates  of  environmental  decay  or environmental
 partitioning  of  the  mixture  components  may  lead  to human  exposures  to a
 mixture that  is not representative  of  the  mixture on which the risk assess-
 ment  was  originally  based.   In  such cases,  quantitative structure activity
 relationships  or approaches  based  on  the  relative  potency method  discussed
 above  may have  merit.   Such  modifications  to  the current approach have  not
 been conducted  as  yet  by the Agency and examples  of such approaches have  not
 been  encountered  in  the  literature.  If  such  approaches  are  used,  their
 validity  will   be  dependent,  as with the relative potency  approach,  on   the
 degree  to  which the approach can be  validated with in  vivo data.
    Other  mixtures,  such  as  the   chlorinated  dioxins and  dibenzofurans,
 require a  different approach  since  "typical"  formulations or compositions do
 not  exist  and  thus  the  multiple chronic bioassays  may be  not  be feasible.
The Agency has  proposed  an interim  procedure  for  estimating risks associated
with exposure to chlorinated  dioxins and  dibenzofurans (U.S.  EPA,  1987c).  A
 similar approach has  been  used by  the New York   State  Department  of  Health
 (Eadon  et  a!.,  1986).   As with the  relative  potency  approach, these methods
rely on in vitro  or ;acute  in vivo  data.   Rather than  using such data  to
assess  the toxicity of  the  mixture of  concern,  however,  these  approaches
                                    5-10

-------
estimate  "toxic  equivalency  factors"   for   the   various  congeners  In  the

mixture based  on  acute or  In  vitro data and validate  the  relationship with

the available data on  chronic  or  subchronlc  toxicity.   The  toxic equivalency

factors  can  then  be  used  to  assess  the hazard  posed by  exposure to  any

combination of  the congeners in any ratio.   To  do this, the concentration of

each component  in  the  mixture  is multiplied  by  the  toxic  equivalency factor

of  that  component.   This product expresses  the concentration of  the compo-

nent as an  equivalent  concentration of  the  reference  compound.   The equiva-

lent concentrations  for all components  are  then  added.  This  total repre-

sents  an  estimate  of   exposure  to  the  mixture  1n  terms  of the  reference

compound.   This  transformed  exposure  estimate  is  then  multiplied by  the

potency  of  the  reference  compound   (2,3,7,8-TCDD   in  the  case  of  the

chlorinated  dioxins)  to obtain  an  overall  estimate  of risk.   Depending on
                                                          - i
the quality  of  the monitoring  data  and  exposure assessment,  U.S. EPA (1987c)
 ' . -       •>       "                       -                    '           J*
also   provides  recommendations   for   modifying  the   risk   assessment.   As
        f        -                   •                       ;  ,        i
reviewed by  U.S.  EPA (1987c),  several  other  countries  and  organizations have

adopted similar approaches for the  chlorinated dioxins.

    The  relative  potency  approach  and  the  toxic equivalency  approach  are
                                                          • !  /
similar  in  that both  use  types  of  data to  assess and  quantify  the  toxicity

of  mixtures  that  are  not  often used to quantify the  risk  from exposure to

single  chemicals  (I.e., acute  data,  data from atypical routes  of  environ-

mental exposure and jjjn vitro data).  They differ, however,  in that the toxic
                                                t
equivalency  approach rests  explicitly on the assumption of  dose or  response

additivity;  this  method should  be applied  only  to compounds that have the

same  mode of   action  or  act  Independently, and  does  not  account  for  any

potential   interactions.    If   significant   interactions  do  occur  in  the
                                    5-11

-------
 mixture,  as  appears  to be  the case with  the  promotion efficiency of  poly-
 bromlnated  blphenyls  (Sleight,  1985),  the  toxic  equivalency approach  could
 result 1n risk assessments  that are misleading.
     The  relative  potency  approach, while  not  explicitly based  on  simple
 similar action, assumes a  linear nonthreshold response  as 1t 1s applied  to
 carcinogens  by Albert  et  al.  (1983).  In  that  the relative potency method,
 however,  Is  conducted  on  mixtures and  validated  using  In.  vivo  data  on
 mixtures, the  possibility  to  account  for  Interactions  Is  not  excluded.  A
 combination  of the  relative potency and  toxic  equivalency approaches  could
 Improve confidence in  risk  assessments of similar mixtures  and  mixtures  of
 chemical  classes.
     In applying either the relative  potency  or  the toxic  equivalency  factor
 methods,  care must be  taken to  ensure  that the compounds are not only  chem-
 ically but  also biologically  similar.   Taking  an example  from  Hehlman and
 Witz   (1986),  a mixture of  ketones  containing  methyl-n-butyl   ketone and
 methyl  isobutyl ketone  would  be similar  only  superficially because methyl-
 n-butyl   ketone,  unlike  methyl  isobutyl   ketone,  is   a  potent  peripheral
 neuropathic  agent.  The  failure to  account  for  the neurotoxic  potency of
 methyl-n-butyl  ketone,  which  is  toxicologically more  similar to n-hexane and
 2,5-hexadione  than to  other  ketones,  could lead to an erroneous risk assess-
 ment.   While  this  type  of  potential error can occur in  dealing  with single
 chemicals with an incompTe.te  data  base   (e.g.,  lack   of   a  teratogenicity
                       \
 study),  the  potential  for  this  type of  error  is higher  when dealing with
mixtures  and  using  data that  are  normally  considered   inadequate  for  con-
 ducting risk assessments on single compounds.
 5.4.   SIMPLE MIXTURES, COMPONENTS AND TOXIC INTERACTIONS
    In  the guidelines  for mixtures,  the Agency  has proposed using addltivity
assumptions  when  data  are  not  available  on the  mixture   of  concern  or  a

                                    5-12

-------
reasonably  similar  mixture,  and  when  the components  are  mechanistically
similar or  Independent.   For toxic  agents with  thresholds,  a Hazard  Index
(HI) Is recommended  based on the  assumption  of dose additivity, and  can  be
expressed  as follows:                                             •
                      HI  = E1/AL1 + E2/AL2  + ...  +  En/AI_n                (5-2)
where E Is  the  level  of  exposure and AL Is the acceptable level of  exposure.
The reference dose  (RfD), an estimate  (with uncertainty  spanning perhaps  an
order  of  magnitude)  of a  dally  exposure to the  human  population  (Including
sensitive  subgroups)  that  1s  likely  to be without  an appreciable  risk  of
deleterious effects during a lifetime,  Is  recommended for use as the "accep-
table  level"  (AL)  1n  order to standardize  Agency  risk  assessments.  Since HI
Is  dlmenslonless,   use  of  the  RfD  means  that  exposure  (E)  must   then  be
presented  In  similar  units  as  dally  intake  (mg/kg/day).   For carcinogens,
the  recommended equation  Is  based  on  a simple  addition of  risks.   At  low
risk levels, this equation simplifies to
P =
                                                                        (5-3)
where  P  is  the expected response,  D is the dose  (level of exposure) and B is
an  estimate of response rate (usually a plausible upper bound called a slope
factor).   In the low-dose region where responses are linear, equation 5-3 is
considered   to  be  a  reasonable  approximation.  At  higher  levels  of  risk,
nonlinearity and competing risks would need  to be considered.   In  addition,
the guidelines also suggest some simple interactive models by which nonaddi-
tive joint action could  be  considered, while recognizing that adequate data
for using such models will  usually  not be  available.
     Since the publication of the guidelines,  the literature on joint action
has  not   suggested  any  fundamental  revisions  to  the  above   approach.
Berenbaum  (1985a)  has  suggested   a   general  approach estimate  of expected
                                     5-13

-------
 responses under the assumption  of  addHlvlty.  Seller  and  Scott  (1987)  Illus-
 trated a method for partitioning attributable  risks  under  either  the assump-
 tion of addltlvity or  using  data adequate  for  quantifying  Interactions.   The
 available data base on  the magnitude  of  toxic  Interactions  for  environmental
 contaminants  has  not,  however, changed  substantially.   In  most cases,  an
 estimate of  risk for  exposure to  a  chemical  mixture will  be  based on  an
 addHlvlty assumption,  except  In  those cases  where, chronic mixture data  or
 an appropriate surrogate approach  (e.g.,  relative potency)  are available.
     The addltlvity assumptions  presented In  equations  5-2  and 5-3 do, none-
 theless,  have  serious  shortcomings.  As applied  to toxicants,  equation  5-2
 Implies that as the acceptable  level  1s approached or  exceeded, the  level  of
 concern Increases  linearly (e.g.,  an HI of 50  1s  of  twice as  much concern  as
 an HI  of  25)  and  In  the same manner  for  all  mixtures.   As  the  mixtures
 guidelines   note,   these  Implications  are  Incorrect.   RfDs  (the  values
 recommended  for  use  as  acceptable levels)  do  not  have  equal  accuracy  or
 precision,   and are  not  based on   the   same   severity  of  toxic  effect.
 Moreover,  slopes  of  dose-response  curves In  excess  of  the  RfO 1n theory are
 expected  to  differ  widely.   The  determinations  of accuracy,  precision or
 slope  are exceedingly  difficult because   of  the general  lack  of  toxlclty
 data.   Severity of  endpolnt, however,  1s  often  known.   For example,  with
 fluoride and  selenium 1t  Is  known  that relatively narrow excursions  above or
 below   the   RfD can  cause   severe  adverse   effects   through  toxlclty   or
 deficiency,  respectively.   Among other compounds,  the margins of safety or
 error  are  thought  to   vary  because of  differences  1n the  quality of  the
available data or the  relationships  of  dose  and  time of  exposure to  the
 Incidence,  severity  or  intensity  of  effects.   Some  of   these  sources   of
variability  and uncertainty  have  been discussed  in  the literature  (Crump,
1984;  Dourson  and Stara,  1983; Lu,  1985;  Rulis, 1987),  but approaches  to

                                    5-14

-------
quantifying  these differences  among  chemicals  have  not  been adopted  for
single  compounds, and  this  inhibits modification  and  improvement  of  the
current approach for  the assessment of mixtures of systemic toxicants.
    For carcinogens,  equation 5-3 may  be overly conservative  because  upper
bounds rather than estimates  of expected  risk  are  added.   This  limitation is
                                                               I
recognized but a  practical  alternative has not been proposed.   As  discussed
in  the  Agency's  guidelines  for  carcinogens,   upper  bounds on  risk are  used
because of  the  substantial  uncertainties  involved  in  high- to  low-dose  and
species-to-species  extrapolation.   Conversely,  as  discussed  by  Berenbaum
(1985b),  synergistic  interactions between  carcinogens may  result  in  dose-
response  curves   that  are  steeper  in the  low-response  region  than  in  the
experimentally  observable  region.    In   such  cases,   the  assumptions   of
linearity  and  additivity could  underestimate risk.   This can  also  be  the
                                                               I
case  in  heterogeneous  responding  populations  (Margosches et  a!.,   1981).
Mechanisms for low-dose synergism have not  been  proposed; in  fact,  Thorslund
and  Charnley  (1987)   show that under  the multistage  theory,  experimentally
determined  synergism  will  not  significantly   differ from the  low-dose  risk
           '
estimate based on additivity.                                  [
5.5.   MIXTURES OF CARCINOGENS WITH OTHER COMPOUNDS
                                                              '[
    The  enhancement   (by  promotion   or   cocarcinogenicity),   inhibition   or
masking of the carcinogenic activity  of  known  or  unidentified  carcinogens in
complex mixtures  is   only briefly discussed  in  the  Agency's   guidelines  on
mixtures.   While  potentially  of great  practical  importance (Reif, 1984),  few
specific proposals have been made to  assess and quantify such  interactions.
    Even with all  the work  that has   been done on  tumor  promoters and cocar-
cinogens,  much of  which  is  summarized by Lucier and Hook (1983), systematic
                                    5-15

-------
 and  predictive relationships  for  expressing and  measuring enhancement have
 not  yet emerged.   Given  the complexities of promotion/cocarcinogenicity, It
 Is  not surprising  that no  clear  approach for  incorporating  these concepts
 into   a risk   assessment  methodology   has  been  recommended.   While  some
 approaches  to  low-dose extrapolation have  been  recommended  which consider
 the  effect of  promoters  on  the initiator  dose-response relationship (Burns
 et  a!., 1983), no  dose-response  models  that  consider  variations  in  both
 doses  of  the   initiator  and doses  of  the  promoter  have  been  proposed  or
 applied to complex mixtures.   As  discussed by  Stara  et al.  (1983), several
 questions must  be answered before such applications are  likely to be made:
    How specific  and  consistent  are  Initiator-promoter   interactions?
    Does  the promoting efficiency  of  a compound vary  with  initiating
    agents  and,  If  so,  does  this   variation   follow  a consistent  or
    predictable pattern?
    How does exposure to  multiple  promoting agents affect the promoting
    efficiency  of  the individual promoters?   If addltlvHy  is a reason-
    able assumption,  which  type of  additivlty  might  be expected based
    on  what we  know about the mechanism  of promotion?
    How does promoting  efficiency vary  with  the duration of exposure to
    the initiator and the promoter?
    Is  there any  validity   in  using  promotion  data  from  one  route  of
    administration  to predict  promoting activity  from another  route of
    exposure?

These  questions  remain  largely  unanswered.   Until  answers  or  reasonable
assumptions are proposed,  progress  in directly applying promotion/cocarcino-
genlcity data  to quantitatively modifying  risk assessments for  mixtures  is
likely  to be minimal.
    A similar situation exists  with compounds  that cause an apparent inhibi-
tion of or  protection  from  chemically-induced  carcinogenicity.   As  reviewed
by NRC  (1980),  very few examples of  this type  of interaction  have been  noted
and the nature  of the interaction can vary  with the  time course of exposure.
                                    5-16

-------
More recently, in  reviewing  the literature on tumor promotion  of  the liver,
Hermann (1985) cites a  few additional studies showing a  decrease  of  preneo-
                                                                i
plastic  liver  foci  after  prolonged   treatment  with  some anti-oxidants  or
hypolipidemic  compounds  and  suggests  that  such  "anti-promoters"  may  have
potential in  the  control of cancer.   While  such a  prospect  is encouraging,
the data currently available are not  sufficient  for  quantifying the dose and
time relationships  for  tumor  inhibition.  Until such  data become  available,
the presence  of  tumor inhibitors  in  mixtures are not  likely  to be  used  in
quantitatively modifying the risk  assessment  unless  they  are  incorporated in
a comparative potency assessment.                               !
    The problem  of masking of  the carcinogenic  activity  of  some  components
in a mixture  that  is due to simple dilution  or  to  the toxic  but noncarcino-
genic  activity of other  components   in  the  mixture  is  less   difficult  to
address  than  either  enhancement   or  inhibition  of  carcinogenicity.   One
component of  this  problem simply  is  to account for competing  risks.   As  In
the  example cited  in  Section  2.5.  from Raabe  (1987),  this  problem  is not
unique  to  mixtures.   The  statistical  methods  for   accounting  for  competing
risks  in animal bioassays  are  available in the literature (Altschuler, 1970;
Hoel,  1972;  Peto  et  al.,  1972; Peto,  1974)  and are  Incorporated  into some
commercially  available  statistical  programs  for   the   analysis  of  cancer
bioassay  data  (e.g.,  MULTI-WEIB  by  Howe  and  Crump,  n.d.).   In  other
Instances,  an  adequate chronic  study  showing  no  carcinogenic  activity may be
available  on  a mixture  that contains  known  carcinogens.  While  the guide-
lines  state that  data on the mixture of  concern are preferred, to additivity
assumptions  based  on  the  known activity  of  the components  in the mixture,
the  analysis  of  such a  "negative"  bioassay must consider whether  a carcino-
genic  response  would  have  been  expected given  the  doses  and  numbers  of
                                    5-17

-------
experimental animals  used.   As  with masking due  to  toxlclty,  masking due to
dilution  1s  not unique  to  mixtures but  Is essentially  Identical  to evalu-
ating  the  significance  of  negative  and  positive   results  from  different
bloassays of a single compound.
                                   5-18

-------
                              6. RESEARCH NEEDS



    For complex mixtures, similar mixtures and mixtures  of  chemical  classes,

the kinds  of  research  needs  vary  depending on  the specific approach  to  be

taken  in  developing  the  risk assessment.   For   instance,  U.S. EPA  (1987c)

proposed  the   following  research   needs  for  better  validating  the  toxic

equivalency factor approach for  chlorinated  dioxins and dibenzofurans:


    1.  Validation and completion of in vitro test data.

    2.  Investigation of  the relationships  between short-term  in. vivo
        and in  vitro tests and  the chronic  toxic  endpoints  of concern
        (I.e.,  carcinogenlcity,  reproductive  toxicity,  immunotoxlcity
        and other significant human health effects).             i
                                                                       .
    3.  Additional data on pharmacodynamics  and metabolite toxicity.
                                          '
    4.  Development  of  additional  short-term assays which  can  test the
        mechanistic  hypotheses  underlying the toxic  equivalency factor
        approach.

                                                                 i
Since  this approach  may  also prove  useful  for   other classes  of  compounds,

such  as  the  bromlnated  dioxins  and dibenzofurans,  comparable  studies  on

these  classes of compounds might also be added to the above list.
                                                                 I
    Research  needs  for  the comparative potency  method  are  somewhat similar.

Currently,  the  relationship  is  validated by  comparing  the  in  vitro and in.

vivo  relative  potencies  of  relatively  few  mixture classes.   Confidence in

this  method could be improved if  the basis  for  the comparison was broadened

to  include not only  relative potency estimates  from  human studies but also

potency  estimates  from  animal  bioassays.    In   addition,  a more extensive

comparison including not only data on mixtures  but  also data   on Individual
                                                                 i
compounds  would help to strengthen  this approach.

     Both   the  toxic  equivalency  factor  and  comparative  potency methods are

generally  applied only  to carcinogens.   While  the in  vitro tests  on which
                                     6-1

-------
 these methods are currently based are  probably  only  appropriate  for  carcino-
 gens, other short-term assays have been developed  for  other  endpolnts  (e.g.,
 teratogenlclty and cytotoxlcity) that may be applicable  to the assessment  of
 the noncardnogenlc  toxlclty  of mixtures.  Given  the  diversity of  mixtures
 1n  the  environment,  the  validation  of  a  battery of  short-term  assays  to
 assess the systemic effects of mixtures could serve as  a valuable  adjunct  to
 the addltlvlty assumption.
     In Implementing this research,  the validation of screening tests must  be
 recognized as  a complex process.  As discussed with respect  to several kinds
 of assays  (Brown  et al.,  1979;  Purchase  et al.,  1976;  Rlnkus  and  Legator,
 1979,  1980; Rosenkranz and  Po1r1er,  1979;  Suglmura et al., 1976) validations
 require  not only  careful  criteria  for assessing  false  positive  and  false
 negatives   but  also  a  consideration  of  the  class  of  compounds  used   to
 validate   the   assay   and   the   limitations  that   this   may  Impose  on  the
 usefulness  of   the  assay for other  classes  of  chemicals.  In  addition,  the
 proposal  to use any screening  test  Is greatly supported  by the demonstration
 that  the  mechanisms  of  action are  similar  for  the toxic  effect  of  concern
 and  the  response   observed  1n  the   screening  test.    Depending  on  use  and
 consistency of  the  screening test,  greater  attention  may need to be given to
 the  statistical  analyses of the assay  results  (Gart  et al.,  1979;  Frome  and
 DuFrain,  1986) so  that  the errors  and uncertainty  In  any analysis can  be
more explicitly Identified.
    As noted In Chapter  5,  the use of  the  addltlvlty assumption  1s somewhat
restricted  by   the  approach currently  used for  risk  assessment   of  single
systemic  toxicants.   While  an  Improvement  of  this  situation appears  to  be
more a matter  of analysis  than  the  generation  of additional  data, 1t  1s  an
area  that  must  be  addressed If  an  Improvement  In the  application  of  the
assumption of addltlvlty 1s  to  be made.

                                    6-2

-------
    As  discussed  In  Section  2.6.  and  Chapter   4,  several  nonlnteractlve
models can be  applied  to the diverse kinds  of  quantitative Interaction data
that  are  available.   In addition,  appropriate  statistical methods  have  not
been  applied to  much  of the data  that  are  available,  and the limitations of
some  of the  available  information preclude  the application  of  any quantita-
tive  model. Consequently, no generalizations can  be  made on the quantitative
significance of  interactions at  normal environmental  levels.   This problem
could  be  at  least  partially  addressed  by a  detailed  reanalysis of  the
available  data  by  applying a  variety  of   noninteractive  models  to  derive
quantitative interactive coefficents.                     .'
    In  conducting risk  assessments  for  single compounds,  both carcinogens
and  systemic  toxicants, the Agency  uses conservative  but  plausible assump-
tions  concerning  extrapolations   from   high to  low  doses  and  species  to
species,  and  concerning modeling of time-to-effects  data.   Concern has been
expressed  both  within  the Agency  and   by  other  elements  of  the  scientific
community  that  the  use  of  dose or response  additivity combining  such conser-
vative  risk  assessments for  individual chemicals could  lead to  implausibly
conservative  risk estimates  for   complex  mixtures.   This  limitation  in the
use   of  an  additivity  assumption  is   one   of  the  reasons  that  the   Agency
prefers  using  data oh, the mixture of concern  or  a  sufficiently similar
mixture and  has  used  the relative potency method or toxic  equivalency  factor
approach  for complex  mixtures.  Nonetheless, additivity  assumptions will be
used  for  many  risk assessments on mixtures, and the need  to  develop alterna-
tive  risk assessment  procedures  or testing strategies  is recognized.   The
Agency  is  currently  reviewing   the  recent  recommendations of  NAS  (1988a)
                                    "\
along with other approaches.                              '
                                     6-3

-------

-------
                                7.  REFERENCES




Abou-Donia, M.B., H.M.  Makkawy  and G.M. Campbell.  1985.   Pattern  of  neuro-

toxicity of ni-hexane, methyl  n-butyl  ketone,  2,5-hexanediol,  and 2,5-hexarte-

dlorie alone  and  In  combination with  o-4-nitrophenyl  phenylphosphonothioate

in hens.  3.  Toxicol. Environ. Health.  16: 85-100.


                                                              I

Albert, R.E.,  0.  Lewtas, 8.  Nesnow,  T.W.  Thorslund and  E.  Anderson.   1983.

Comparative potency  method  for  cancer  risk assessment  application  to  diesel

participate emissions.  Risk Analysis.  3(2):  101-117.

                                                              i

Alstott, R.L., H.E.  Tarrant  and R.B.  Forney.   1973.  The acute toxicities of

1-methylxanthine, ethanol,  and  1-methylxanthine/ethanol  combinations  in the

mouse.  Toxicol. Appl.  Pharmacol.  24:  393-404.



Altshuler,  B.   1970.   Theory  for  the measurement  of  competing   risks  in

animal  experiments.  Math. Biosci.  6:  1-11.                  !



Ames,   B.N.,  J.  McCann  and  E.   Yamasaki.    1975.   Methods  for  detecting

carcinogens  and  mutagens with the  Salmonella/mammalian-microsome mutagen-

icity test.   Mutat.  Res.  31: 347-364.


                                                             j

Andelman,  3.B.  and  M.  Barnett.  1986.  Feasibility study  to relate arsenic

in  drinking  water  to   skin  cancer  in  the United  States.  In_:  Environmental
                                                              !
Epidemiology,  Kopfler,  F.C. and  G.F. Craun,  Ed.  Lewis  Publishers,   Inc.,

Chelsea, HI.  p.  89-107.
                                     7-1

-------
 Andersen, H.E.  and H.G.  Clewell,  III.   1984.   Pharmacoklnetlc  Interactions
 of mixtures.   Proceedings  of Conference on Environmental  Toxicology  (14th),
 Dayton, Ohio, November 15-17.  AD-A146 400.

 Anderson, L.M.,  K.  van Havere  and J.M. Budlnger.   1983.   Effects of  poly-
 chlorinated  blphenyls  on lung and  liver tumors Initiated in  suckling mice by
 H-nHrosodlmethylamlne.   J.  Natl.  Cancer  Inst.  71(1):  157-163

 Aranyl,  C.,  J. Bradof,  D.E.  Gardner  and  J.L.  Hulslngh.  1980.  In vitro and
 in vivo  evaluation  of  potential  toxldty  of  Inudstrlal  particles.   In.:
 Short-term  Bloassays  In  the Analysis of Complex Environmental  Mixtures II.
 Plenum  Press,  NY.   p.  431-444.

 Autrup,  H.,  C.C. Harris,  P.W.  Schafer,  B.F. Trump,  6.D. Stoner  and  I.  Hsu.
 1979.   Uptake of benzo(a)pyrene-ferr1c oxide partlculates by human pulmonary
 macrophages  and  release  of  benzo(a)pyrene  and  Us  metabolites  (40536).
 Proc. Soc. Exp. Blol.  Med.   161: 280-284.

 Bauer,  K.H.    1928.    MutatlonstheoMe  der   GeschurelstEntstehung.   Ubergang
 von Korperzellen  1n Geschurlestzulen   durch  Gln-anderung.   Tulluls Springer,
 Berlin.

 Berenbaum, M.   1981.   Criteria  for  analyzing  Interactions  between  biolog-
 ically active agents.  Adv. Cancer Res.  35:  269-335.

Berenbaum, H.C.  1985a.  The  expected  effect of a combination of  agents:  The
general  solution.  J. Theor.  Blol.   114:  413-431.
                                    7-2

-------
Berenbaum,  M.   1985b.    Consequences   of  synergy  between   environmental
carcinogens.  Environ.  Res.   38:  310-318.                      i

                        '
Berenblum,  I.   1941a.   The  mechanism  of  cardnogenesls:  A  study  of  the
significance of  cocarcinogenlc action  and related  phenomena.   Cancer  Res.
1: 807-814.
                                                               I
                                                               i
Berenblum,  I.   1941b.   The  cocarcinogenlc action  of  croton  resin.   Cancer
Res.  1: 44-47.

Berenblum,  I.   1979.   Theoretical  and  practical  aspects  of  the  two-stage
mechanism of cardnogenesls.   In: Carcinogens:  Identification  and  Mechanisms
of  Action,  A.C.  Griffin and  C.R.  Shaw, Ed.   Published  by  Raver  Press,  New
York.  p. 25-36.
                                                               I
                                                               I
BUss,  C.I.   1939.  The  toxlclty of  poisons  applied  jointly.   Ann.  Appl.
B1ol.  26: 585-615.
Bohrman,  J.S.   1983.   Identification and  assessment of  tumor-promoting  and
                                                               i
cocarcinogenlc  agents:  State-of-the-art in  vitro methods.   CRC  Cr1t.  Rev.
Toxlcol.  11(2): 121-167.
Brown,  M.M.,  G.S.  Wassom,  H.V.  Mailing, M.D.  Shelby and  E.S.  Von  Halle.
1979.   Literature  survey of bacterial, fungal, and  Drosophlla  assay  systems
used  1n the  evaluation  of  the  selected chemical  compounds  for  mutagenlc
activity.  J. Natl. Cancer Inst.  62(4): 841-871.              i
                                    7-3

-------
 Burns,  F.,  R.  A.E.  Altschuler  and  E.  Morris.   1983.   Approach  to  risk
 assessment  for  genotoxlc  carcinogens  based  on  data  from  the  mouse  skin
 Initiation-promotion model.  Environ. Health Perspect.  50: 309-320

 Butler,  G.P.,  T.J.  Knelp,  F.  Mukai  and  J.M.  Dalsey.   1984.    Inter-urban
 Variations  In  the  Hutagenlc   Activity  of  the  Ambient   Aerosol   and  Their
 Relations to  Fuel  Use  Patterns,  in.:  Short-Term Bioassays in the  Analysis of-
 Complex  Environmental  Mixtures  IV, Waters,  Sandhu,  Lewtas,  Claxton, Strauss
 and Nesnow, Ed.  Plenum Press,  New York.   p. 233-246.

 Caval1er1, E.,  A.  Munhal,  E.  Rogan,  S.  Salmasl and K. Patil.  1983.  Syncar-
 cinogenlc effect of the  environmental   pollutants  cyclopenteno[cd]pyrene  In
 mouse skin.   Carclnogenesis.  4(4): 393-397.

 Cerklewski,  F.L. and R.M.  Forbes.  1976.   Influence  of dietary  zinc on lead
 toxlclty In'the rat.  J. Nutr.  106: 689-696.

 Chen, R.W., P.O. Whanger  and  P.M. Weswlg.   1975.  Selenium-induced redistri-
 bution  of cadmium  binding to   tissue  proteins:   A  possible  mechanism  of
 protection against cadmium toxicity.  Bioinorg. Chem.   4:  125-133.

 Chescheir,  G.M.,  N.E.  Garrett,   J.D.   Shelburne,  3.L.  Huisingh  and  M.D.
Waters.   1980.   Mutagenic  effects   of  environmental  particulates  in  the
 CHO/HGPRT  system.    JJK  Short-Term  Bioassays  in  the  Analysis   of  Complex
 Environmental  Mixtures  II,  Waters, Sandhu,  Huisingh,  Claxton  and Nesnow,  Ed.
Plenum Press,  New York.  p. 337-350.
                                    7-4

-------
Chriswell,  C.D.,  B.A.  Glatz,  J.S.Fritz  and  H.O.  Svec.   1978.   Mutagenlc
analysis of drinking water.   Ijri:  Application of Short-Term Bloassays  1n  the
Fractlonatlon  and  Analysis   of   Complex   Environmental   Mixtures,   Waters,
Nesnow,  Hulsingh,  Sandhu   and  Claxton,   Ed.    Plenum   Press,   New  York.
p. 477-494.                                                       i

Claxton, L.  and  M.  Kohan.   1980.   Bacterial mutagenesls  and  the evaluation
of  mobile-source  emissions.   In.:   Short-Term  Bloassays  In  the  Analysis  of
                     •                            •                 f >      • •  .
Complex  Environmental   Mixtures  II,  Waters,  Sandhu,  Hulsingh,  Claxton  and
Nesnow, Ed.  Plenum Press, New York.  p. 299-318.
Clayson,   D.B.    1984.
55(Suppl. 2): 35-51.
Co-carc1nogenes1s.    Acta   Pharmacol.   Toxlcol.
Commoner,  B.,  A.J.  VHhayathll and P.  Dolara.   1978.   Mutagenlc analysis of
complex  samples  of aqueous  effluents,  air partlculates,  and  foods.   In:
Application  of  Short-Term Bloassays  In the  Fractlonatlon and  Analysis of
Complex   Environmental   Mixtures,  Waters,  Nesnow,   Hulsingh,  Sandhu  and
                                                                  i   ,'"
Claxton,  Ed.   Plenum Press, New York.   p. 529-570.                :
                                                                  i    .    '. • .

Constantin,  M.J.,  K.  Lowe, T.K.  Rao,  F.W.  Larimer  and  J.L.  Epler.   1980.
The  detection  of potential genetic hazards In complex environmental mixtures
using  plant cytogenetics  and  mlcroblal  mutagenesls  assays.    In.: Short-Term
Bloassays In  the  Analysis of  Complex  Environmental  Mixtures  II,  Waters,
Sandhu,   Hulsingh,   Claxton   and   Nesnow,   Ed.    Plenum  Press,  New  York.
p.  253-266.
                                     7-5

-------
 Cornish, H.H.   1969.   The  role  of vitamin  B,  1n  the  toxiclty  of  hydra-
 zlnes.   Ann.  N.Y.  Acad.  Ac1.   166:  136-145.

 Crump,  K.S.   1984.   A  new method  for determining  allowable dally  Intakes.
 Fund. Appl. Toxlcol.   4: 854-871.

 Cunningham,   H.L.,   D.A.  Haugen,   F.R.   Klrchner  and   C.A.  Rellly.   1984.
 Toxlcologlc   responses  to a  complex  coal  conversion  by-product:  Mammalian
 cell  mutagenlcity and  dermal  carcinogenldty.  In:  Short-Term Bloassays in
 the  Analysis   of  Complex Environmental Mixtures IV,  Waters,  Sandhu,  Lewtas,
 Claxton,  Strauss and  Nesnow, Ed.  Plenum  Press,  New York.  p. 113-124.

 Dawson,   D.A.,  C.A.  McCormlck  and   3.A.   Bantle.    1985.    Detection  of
 teratogenic  substances  1n acidic  mine water  samples using  the  frog embryo
 teratogenesls  assay -  Xenopus (FETAX).  J. Appl. Toxlcol.  5(41): 234-244.

 Dayal, H.H.   1980.   Additive  excess risk  model for epidemiologic Interaction
 in retrospective studies.  J.  Chron. Dis.   33: 653-660.

 Derr, R.F.,  H. Aaker,  C.S.  Alexander  and  H.T.  Nagasawa.  1970.   Synergism
 between cobalt and ethanol on rat growth rate.  J.  Nutr.  100: 521-552.

 Donnelly, K.C.,  K.W.  Brown  and B.R.  Scott.   1983.   The use  of  short-term
bioassays to  monitor  the environmental  impact  of land treatment of  hazardous
wastes.    In.:  Short-term  Bioassays  in  the Analysis of  Complex  Environmental
Mixtures   III, Waters,   Sandhu,  Lewtas,  Claxton,  Chernoff and Nesnow,  Ed.
Plenum Press,  New York.  p. 59-78.
                                    7-6

-------
Douglas, G.R.,  E.R.  Nestmann,  A.B. McKague,  et al.  1983.   Mutagenlclty  of

pulp and  paper  mill  effluents:  A comprehensive  study  of Complex  mixtures.

Li: Short-term  Bioassays  in the  Analysis  of Complex Environmental  Mixtures

III,  Waters,  Sandhu,  Lewtas,  Claxton,  Chernoff  and  Nesnow,  Ed.   Plenum

Press, New York.  p. 431--460.



Dourson,  M.L.   and  J.  Sltara.   1983.   Regulatory  history] and  experimental

support  of  uncertainty  (safety)  factors.   Reg.  Toxlcol.  Pharmacol.   3:

224-238.



Driver,  H.E.  and   A.E.M.  McLean.   1986.    Dose-response relationships  for

Initiation  of  rat   liver  tumours  by  dlethylnltrosamlne and  promotion  by

phenobarbitone or alcohol.  Food Chem. Toxlcol.   24(3):  241-245.



Dumont, J.N., T.W.  Schults,  M.V.  Buchanan and  6.L.  Kao.   1983.  Frog embryo
                                                           j
teratogenesls assay:  Xenopus  (FETAX)  --  A  short-term assay  applicable  to
                                                           I
complex environmental mixtures.   In.:  Short-term Bloassays 1m  the Analysis  of

Complex   Environmental   Mixtures   III,  Waters,   Sandhu,   Lewtas,   Claxton,

Chernoff and Nesnow, Ed.  Plenum Press, New York.  p. 393-406.



DuMouchel,  W.H. and  J.E.  Harris.   1983.   Bayes  methods for  combining  the
                                                           j
results  of  cancer  studies  In humans  and  other  species.   J. Am.  Statls.

Assoc.  78(382): 293-315.
Durkin, P.R.   1981.   An  approach to the analysis of toxicant interactions in

the  aquatic environment.   Proc.  4th Ann. Symp.  Aquatic  Toxicology American

Society for Testing and Materials,   p. 388-401.
                                    7-7

-------
Eadon,  G.A.,  L.  Kamlnsky,  3.   Silkworth   et  al.   1986.   Calculation  of
2,3,7,8-TCDD  equivalent  concentrations of  complex  environmental  contaminant
mixtures.  Environ. Health. Perspect.  70: 221-227.

Elashoff,  R.H.,  T.R.  Fears  and  H.A.  Schnelderman.    1987.   Statistical
analysis  of  a   carcinogen  mixture  experiment.   I.  Liver carcinogens.   3.
Natl. Cancer Inst.  79: 509-526.

Epler, 3.L.,  B.R.  Clark,  C.-H. Ho, M.R.  Guerln and  T.K.  Rao.   1978.   Short-
term  bloassay of  complex  organic mixtures:  Part II,  Hutagenlclty  testing.
In: Application  of Short-Term  Bloassays  1n  the  Fractlonatlon  and  Analysis of
Complex   Environmental   Mixtures,   Waters,   Nesnow,  Hu1s1ngh,   Sandhu  and
Claxton, Ed.  Plenum Press, New York.  p. 269-290.

Eybl, V.,  3. Sykora,  3.  Koutensky  et  al.   1984.   Interaction of  chelatlng
agents  with  cadmium  1n  mice  and  rats.   Environ.  Health  Perspect.   54:
267-273.

Flnney,   D.3.   1971.   Problt Analysis,  3rd  ed.    Cambridge  University  Press,
Cambridge, Great Britain.  333 p.

Fisher,   G.L.,  K.L.  McNeil! and  C.3.  Democko.   1983.   Application of  bovine
macrophage bloassays  In  the analysis  of toxic agents  In  complex mixtures.
In.; Short-term  Bloassays In the  Analysis of Complex Environmental  Mixtures
III,  Waters,  Sandhu,  Lewtas,  Claxton,  Chernoff  and  Nesnow,  Ed.    Plenum
Press, New York.  p.  257-268.
                                    7-8

-------
Freeman, 3.J. and  E.P.  Hayes.  1985.   Acetone  potentiation of acute  aceto-


nitrile toxicity In rats.   J.  Toxlcol.  Environ.  Health.   15(5):  609-621.





Frome,  E.L.  and  R.J.  DuFrain.   1986.   Maximum  likelihood  estimation  for


cytogenetlc dose-response  curves.   Biometrics.   41(1):  73-84.

                                                             !|



Fujiki,  H.,  M.  Mori, M.  Nakayasu, M.  Terada  and  T.  Suglmura.   1979.   A


possible  naturally occurring  tumor promoter,  tile  ocldin  B   from  strepto-
                                                             i

myces.  Blochem. Blophys.  Res. Commun.   90: 976-983.





Gart,  J.J.,  J.A.  DIPaolo  and  P.J.  Donovan.   1979.  Mathematical  models  and


the  statistical  analyses  of  cell  transformation  experiments.   Cancer  Res.


39: 5069-5075.





Gaughan,  L.C.,   J. Engel  and  J.  Casida.  1980.   Pesticide  Interactions:


Effects  of  organophosphorous pesticides  on  the metabolism,  toxlclty  and


persistence  of  selected  pyrethrold Insecticides.  Pestle.  Blochem.  Physlol.


14: 81-85.                                                   ;





Goldstein, A., L.  Aronow  and  S.M.  Kalman.   1974.  Principles  of Drug Action:


The  Basis  of Pharmacology,  2nd  ed.   John Wiley  and  Sons, Inc., New York.


854 p.
Gullino,  P., M.  Winitz,  S.M.  Birnbaum,  J.  Cornfield,  M.C.  Otey  and  J.P.


Greenstein.   1956.   Studies  on the  metabolism of  amino acids  and related
                                                              i

compounds  In vivo.   I.  Toxicity  of essential amino  acids,  individually and


in  mixtures,  and  the  protective  effect of   L-arginine.   Arch.  Biochem.


Biophys.   64: 319-332.                                        ;




                                    7-9

-------
 Hakama,  H.  1971.  Epldemiologic evidence for multi-stage theory of carcino-
 genesls.   Int.  J.  Cancer.   7:  557-564.

 Hammond,  E.G.  and  I.J.  Sellkoff.   1973.   Relation of  cigarette  smoking to
 risk  of  death of asbestos-associated disease among Insulation workers In the
 United States.   Mount  Sinai  School of Medicine, New York.  p. 312-317.

 Hammond,  E.G.,  I.J.   Selikoff  and  H.   Seldman.   1979.   Asbestos  expsoure,
 cigarette  smoking  and  death  rates.   New  York  Acad.  Sciences,   New  York.
 p. 473-490.

 Hecker,  E.   1968.   Cocarcinogenic   principles  from the  seed  oil  of  Croton
 tlqluum and from other Euphorbiaceae.  Cancer Res.  28: 2338-2349.

 Hecker, E.  1985.   Cell  membrane associated protein kinase  C  as receptor of
 dlterpene  ester  co-carcinogens  of the tumor promoter  type and  the  phenotype
 expression of tumors.  Arzneimittelforschung.  35: 1890-1903.

 Hermann, R.S.  1985.  Tumor promotion  in  the liver.   Arch.  Toxlcol.   57:
 147-158.

Hermens, J.,  P.  Leeuwangh  and  A. Musch.  1985a.   Joint  toxiclty of mixtures
of groups  of  organic  aquatic pollutants to  the guppy  (Poecllia  reticulata).
Ecotoxlcol. Environ. Safety.  9:321-326.
                                    7-10

-------
Hermens,  J.,  E.  Broekhuyzen,  H. Canton and  R.  Wegman.   1985b.  Quantitative
structure  activity relationships  and  mixture  toxicity  studies  of  alcohols
and  chlorohydrocarbons:   Effects   on   growth  of  Daphnia  magna.    Aquatic
Toxicol.  6: 209-217.                                         j

Herren-Freund, S.L. and  M.A.  Pereira.   1986.  Carcinogenicity of by-products
of  disinfection   in  mouse and  rat  liver.    Environ.  Health  Perspect.   69:
59-65.
                                                              .i

Herschberg,  S.N.  and  F.S.   Sierles.   1983.   Indomethacin-induced  lithium
toxicity.  Am. Fam. Physician.  28(2):  155-157.

Hertzberg,  R.    1987.   Discussion  of   "Development  of  models  for  combined
toxicant  effects" by  E.  Christensen.   In.:   Current  Assessment  of  Combined
Toxicant  Effects.  ASA/EPA  Conferences on  Interpretation  of  Environmental
Data, Hay 5-6, 1986.  EPA 230-03-87-027.

Hewlett,  P.S.   1969.   Measurement  of  the  potencies   of  drug  mixtures.
Biometrics.  25:  477-487.

Hewlett,  P.S.  and R.L.  Plackett.   1964.    A unified  theory  for  quanta!
response to mixtures of drugs: Competitive action.  Biometrics.  20:  566-575.

Hoel, D.G.   1972.  A  representation  of mortality  data  by  competing  risks.
Biometrics.  28:  475-488.
                                    7-11

-------
 Hose,  J.E.   1985.   Potential  uses  of. sea  urchin  embryos  for  identifying
 toxic  chemicals: Description  of  a bioassay  incorporating  cytologic,  cyto-
 genetic  and  embryologic  endpoints.   J. Appl.  Toxicol.  5(4): 245-254.

 Houk,  V.S.  and L.D. Claxton.  1986.   Screening  complex  hazardous wastes for
 mutagenic  activity  using a  modified version  of the  TLC/Salmonella assay.
 Mutat. Res.   169: 81-92.

 Howe,  R.B. and K.S.  Crump,   n.d.  MULTI-WEIB: A computer program  to extrapo-
 late  time  to  tumor  animal  toxidty data   to  low doses.   K.S.  Crump  and
 Company, Clement Associates, Inc.,  Ruston, LA.

 Hsie,  A.M.,  J.P. O'Neill,  G.R.  San  Sebastian,  et al.   1978.   Quantitative
 mammalian  cell genetic toxicology: Study of  the cytotoxicity  and mutagenic-
 ity of seventy individual environmental agents related to energy technologies
 and  three  subfractions  of  a  crude  synthetic oil  in  the  CHO/HGPRT system.
 In.: Application  of  Short-Term  Bioassays  in  the Fractionation and Analysis of
 Complex  Environmental   Mixtures,   Waters,    Nesnow,   Huisingh,   Sandhu  and
 Claxton, Ed.   Plenum Press, New York.  p. 291-316.

 Huisingh, 0.,  R.  Bradow,  R. Jungers, et al.   1978.  Application of bioassay
 to  the characterization  of diesel  particle  emissions,   in: Application of
 Short-Term Bioassays  in  the  Fractionation  and Analysis of  Complex Environ-
mental Mixtures,  Waters, Nesnow,  Huisingh,  Sandhu and  Claxton,  Ed.   Plenum
 Press, New York.  p. 381-418.
                                    7-12

-------
 IARC  (International  Agency for Research  on  Cancer).   1982.  IARC Monographs
 on  the Evaluation of  the Carcinogenic  Risk of Chemicals  to Humans.  Chem-
 icals,   Industrial   Processes   and  Industries  Associated  with  Cancer  In
 Humans.  IARC Monographs, Lyon, France.   Volume 1 to 29.  4:  14-22.
                                                            •i
 Ito,  N., S.  Fukushima  and H.  Tsuda.   1985.  Carcinogenicity and modification
 of  the  carcinogenic  response  by BHA,  BHT, and other antloxldants.  CRC Crit.
 Rev.  Toxlcol.  15(2): 109-150.                              ;
Keplinger,  M.L.  and W.B.  Delchmann.   1967.  Acute  toxicity  of combinations
                                                            i
of pesticides.  Toxicol. Appl. Pharmacol.  10: 586-595.
King,  M.M.  and  P.B.   McCay.    1983.    Modulation  of  tumor
possible  mechanisms  of  inhibition  of  mammary  carcinogenesi
antioxidants.  Cancer Res,.  43(5): 2485s-2490s.
 incidence  and
s  by  dietrary
Klaassen,  C.O.  and  0.  Doull.   1980.   Evaluation  of  safety:  Toxicologic
evaluation.   In.:  Toxicology:  The  Basic  Science of  Poisons,  J.  Doull,  C.D.
Klaassen  and M.O.  Amdur,   Ed.   Macmillan  Publishing Co., Inc., New  York.
p. 11-27.
Klein,  N.W.,  C.L.  Chatot,  J.D.  Plenefisch  and  S.W.  Carey.   1983.   Human
serum  teratogenicity  studies  using in  vitro cultures  of  rat embryos.   ITK
Short-term Bloassays  in  the Analysis  of Complex  Environmental Mixtures  III,
Waters, Sandhu, Lewtas, Claxton, Chernoff and Nesnow,  Ed.   Plenum Press,  New
York.  p. 407-416.
                                    7-13

-------
 Kocan,  R.H.  and D.B. Powell.  1984.   Anaphase  aberrations:  An In vitro test
 for  assessing  the genotoxlcity of Individual chemicals and complex mixtures.
 In.:  Short-Term Bloassays  In  the Analysis of  Complex  Environmental  Mixtures
 IV,  Waters,  Sandhu,  Lewtas, Claxton,  Strauss and  Nesnow,  Ed.   Plenum Press,
 New  York.  p. 75-86.

 Kodba,  R.3.  and   0.  Cabey.   1985.   Comparative  toxlclty  and  biologic
 activity  of  chlorinated  dlbenzo-p-dloxlns  and  furans  relative  to  2,3,7,8-
 tetrachlorodlbenzo-p-dloxln (TCDD).  Chemosphere.  14(6/7): 649-660.

 Kopfler,  F.C.   and   G.F.  Craun.   1986.   Environmental Epidemiology.   Lewis
 Publishers, Inc., Chelsea, HI.  284 p.

 Korn, K.L. and  P.Y.  Liu.  1983.  Interactive effects  of  mixtures of stimuli
 In life table analysis.  Biometrlka.  70: 103-110.

 Kourl, R.E., K.R. Brandt,  R.G.  Sosnowskl, L.M.  Schechtman  and  W.F. Benedict.
 1978.  In.  vitro  activation of cigarette smoke  condensate  materials  to  their
mutagenlc  forms.  In.:  Application of  Short-term Bloassays  In  the Fractlona-
 tlon  and  Analysis  of   Complex  Environmental  Mixtures,  Waters,   Nesnow,
Hulslngh, Sandhu and Claxton, Ed.  Plenum Press, New York.   p.  495-512.

Lakowlcz,  J.R. and J.L.  Hylden.   1978.   Asbestos-mediated  membrane update  of
benzo(a)pyrene observed by fluorescence spectroscopy.  Nature.   275:  446-448.
Lakowlcz,  3.R.,  D.R.  Bevan  and   S.C.   Rleman.    1980.    Transport   of   a
carcinogen,  benzo(a)pyrene,  from partlculates  to  Upld bllayers.   Blochem.
Blophys. Acta.  629: 243-258.

                                    7-14

-------
Larsson,  K.S.,  A. Claes,  E.  Cekanova  and  M. Kjellberg.   1976.   Studies of
Teratogenlc  Effects  of  the  Dlthiocarbamates Maneb,  Mancozeb,  and Proplneb.
Laboratory of Teratology,  Karollnska Instltutet, S-104011 Stockholm, Sweden.
Teratology.  14: 171-184.
                                                                   1

Lauwerys,  R.R.  and  S.D.  Murphy.   1969.   Interaction  between paraoxon  and
trl-o-tolyl phosphate In rats.  Toxicol. Appl. Pharmacol.  14: 348-357.
                                                                   i
                                                                   i
Leonard,  T.B.,  D.G.  Morgan  and  J.6.  Dent.   1985.   Ran1t1d1ne-acetam1nophen
Interaction: Effects  on acetam1riophen-1nduced hepatotoxicity  In  Fischer  344
rats.  Hepatology.  5(3): 480-487.
                                                                   I
                                                                   i
                                                                   i
Levlne,  R.E.   1973.    Pharmacology:   Drug  Actions   and  Reactions.   Little,
Brown and Company, Boston, MA.  412 p.

Levander, 0. and  L.  Argrett.  1969.   Effects of arsenic,  mercury,  thallium
and  lead on selenium  metabolism  1n  rats.    Toxicol.  Appl.  Pharmacol.   14:
308-314.                                                           |

                                                                   i
Lewtas,  J.   1985.   Development  of a  comparative potency method  for  cancer
risk assessment  of complex mixtures  using   short-term  1_n vivo and jji  vitro
bloassays.  Toxicol.  Indust.  Health.  1(4):  193-203.
LI,  A.P.,   A.L.  Brooks,  C.R.  Clark,  R.W.  Shimlzu,  R.L.  Hanson  and  J.S.
Dutcher.  1983.  Mutagenlclty testing of  complex  environmental  mixtures  with
Chinese hamster  ovary cells.   In:  Short-term Bloassays  In  the Analysis  of
Complex  Environmental  Mixtures   III,   Waters,  Sandhu,  Lewtas, '  Claxton,
Chernoff and Nesnow, Ed.   Plenum Press,  New York.   p. 183-196.

                                    7-15

-------
Lockard,  V.G.,  M.M.  Harlhara  and  R.M.  O'Neal.   1983.   Chlordecone-lnduced
potentlatlon  of  carbon  tetrachlorlde  hepatotoxlclty:  A  morphometrlc  and
biochemical study.  Exper. Mol. Pathol.  39: 246-255.

Loper,  3.C.  and  D.R.   Lang.    1978.    Mutagenlc,  carcinogenic,  and  toxic
effects of  residual  organlcs  In  drinking  water.   In: Application  of  Short-
Term  Bloassays  1n  the  Fractlonatlon  and  Analysis of Complex  Environmental
Mixtures, Waters,  Nesnow, Hulslngh, Sandhu  and Claxton, Ed.   Plenum  Press,
New York.  p. 513-528.

Lu,  F.C.   1985.   Safety assessments  of  chemicals  with threshold  effects.
Reg. Toxlcol. Pharmacol.  5(4): 460-465.

Luder, G.W.  and 6.E.R.  Hook, Ed.  1983.   Proceedings  of the  Symposium on
Biological  Tests  1n  the  Evaluation of  Mutagenlcity and Carclnogenlclty  of
A1r Pollutants with  Special  Reference  to Motor Exhausts and  Coal  Combustion
Products.  Environ. Health Perspect.  p. 1-341.

Lundln, F.E.,  Jr.,  J.W. Lloyd and  E.M.  Smith.  1969.   Mortality  of  uranium
miners  In relation  to  radiation  exposure, hard-rock  mining and  cigarette
smoking.  Health Physics.  16:  571-578.

Ma, T.H., V.A. Anderson and  S.S. Sandhu.  1980.   A  preliminary  study  of  the
clastogenic  effects  of  diesel  exhaust  fumes  using  the Tradescantia  micro-
nucleus  bioassay.    In;  Short-Term  Bioassays  in  the   Analysis  of  Complex
Environmental Mixtures  II, Waters,  Sandhu,  Huisingh,  Claxton  and Nesnow,  Ed.
Plenum Press, New York.   p.  351-358.
                                    7-16

-------
Ma, T.H., W.R.  Lower,  P.O.  Harris, et al.  1983.   Evaluation  by  the  Trades-
cantla-micronucleus test  of the mutagenicity  of Internal  combustion  engine
exhaust  fumes  from dlesel  and  diesel-soybean oil  mixed  fuels.  In:  Short-
term  Bioassays  in  the   Analysis  of  Complex  Environmental  Mixtures  III,
Waters, Sandhu, Lewtas, Claxton, Chernoff and Nesnow,  Ed.   Plenum Press,  New
York.  p. 89-99.

Ma,  T.H.,  M.M.  Harris,   V.A.  Anderson,  et  al.   1984.   Tradescantja.-micro-
nucleus  (Trad-MCN)  tests on  140  health-related  agents.   Mutat.  Res.   138:
157-167.                                                       i
             .
                                                               t
                                                               i
Malkinson, A.M.   1983.   Review:  Putative mutagens  and  carcinogens  in  foods.
III. Butylated hydroxytoluene (BHT).   Environ. Mutagen.  5(3):  353-362.
Margosches, E.H.,  M.E.  Samuhel and  3.C.  Bailar,  III.  1981.   Some  implica-
tions  of  heterogenlcity  for  low dose  extrapolation.   Envirohmetrics.   p,
                                                               i
81.  SIAM/SIMS.                                                I
Marking, L.L. and  V.K.  Dawson.   1975.  Method for  Assessment  of  Toxicity or
Efficacy of  Mixtures  of  Chemicals.   USDI, Fish and  Wildlife  Service,  Bureau
of  Sport  Fisheries  and  Wildlife,  Washington,  DC,  Investigations  in  Fish
Control, No. 67, p. 1-8.
                                                               i
                                                            '
Maronpot,  R.R., M.B.  Shimkin,  H.P.  Witschi,  L.H.  Smith and  O.M.  Cline.
1986.   Strain A mouse pulmonary tumor test results  for  chemicals  previously
tested  1n  the  National  Cancer  Institute  carcinogenidty  tests.   J.  Nat.
Cancer  Inst.  76(6): 1101-1102.
                                    7-17

-------
Marshall,  M.,  M.  Arnott, H. Jacobs and  A.  Griffin.   1979.  Selenium effects
on  the  carcinogenidty  and  metabolism of  2-acetylamlnofluorene.   Cancer
Lett.  7:  331-338.

HcCullagh,  P.  and  J.A.  Nelder.   1983.  Generalized  Linear  Models.   Chapman
and Halle, New York.  p. 199-208.

HcGeorge,  L.O.,  J.B.  Louis,   T.B.   Atherholt  and   G.J.  McGarrlty.   1984.
HutagenlcHy  analyses of  Industrial   effluents:  Results  and  considerations
for  Integration  Into  water  pollution  control  programs.   In.:  Short-Term
Bloassays  1n  the  Analysis  of  Complex  Environmental  Mixtures  IV,  Waters,
Sandhu,  Lewtas,  Claxton, Strauss  and Nesnow, Ed.   Plenum Press, New  York.
p. 247-268.

Mehlman,  M.A.   and   G.  W1tz.   1986.    Health risk   assessment  of  chemical
mixtures.  Proc.-APCA Annual Meet.  1986.  79th (Vol. 1): 1-17.

Mermelsteln, R., O.K. Klriazldes, M.  Butler.  E.G. McCoy  and  H.S.  Rosenkranz.
1981.  The extraordinary mutageniclty of  nltropyrenes  In bacteria.   Mutat.
Res.  89: 187-196.

MgBodUe,  M.U.K., M.  Holscher  and R.A.  Neal.   1975.  A  possible  protective
role for  reduced  glutathione in  aflatoxln  B,  toxicity:  Effect of  pretreat-
ment  of  rats  with  phenobarbital  and   3-methylcholanthrene  on   aflatoxin
toxicity.  Toxicol.  Appl. Pharmacol.   34: 128-142.
                                    7-18

-------
 Mochlzukl,  Y.,  F.  Kazunorl  and  N.  Sawada.   1983.   Effect  of  simultaneous
 administration  of  clofibrate  with  dlethylnltrosamlne  on  hepatic  tumorl-
 genesls  In  the rat.   Cancer  Lett.   19: 99-105.                !
 Mumford,  3.L.  and  J.   Lewtas.   1983.   Sample  collection  and  preparation
 methods affecting  mutagenldty  and  cytotoxlclty of coal fly ash.   in: Short-
 term  Bloassays  1n  the  Analysis  of  Complex  Environmental! Mixtures  III,
 Waters, Sandhu,  Lewtas,  Claxton,  Chernoff and Nesnow, Ed.  Plenum Press, Mew
 York.  p. 39-58.
Murphy,  S.D.   1964.   A review  of effects  on  animals  of
exhaust  and  some  of  Its  components.   J  Air  Pollut.  Cont.
303-308.
exposure  to  auto
  Assoc.   14(8):
NAS  (National  Academy of Science).  1981.   Impacts  of  Diesel
Duty  Vehicles.   Health  Effects  of Exposure  to  Diesel  Exhc
Academy Press, Washington, DC.
  -Powered  Light-
  :ust.   National
NAS (National Academy of Science).  1988a.   Complex  Mixtures:  Methods  for  In
Vivo Toxlclty Testing.  National Academy Press, Washington,  DC.  227 p.
NAS  (National  Academy of Science).   1988b.   Biological Effects of  Ionizing
Radiation, IV.  National  Academy Press, Washington,  DC.        \

                                                              i
Netteshelm, P., D.C.  Topping  and  R.  Jamasbi.  1981.  Host and  environmental
factors  enhancing  carclnogenesls  In  the  respiratory  tract.    Ann.   Rev.
Pharmacol. Toxlcol.   21:  133-163.
                                    7-19

-------
NRC (National  Research  Council).   1980.  Principles of Toxlcologlcal  Inter-
actions  Associated  with  Multiple  Chemical   Exposures.    National   Academy
Press, Washington, DC.

Oskarsson,  A.  and  B. Llnd.   1985.   Increased  lead levels  In  brain  after
long-term treatment  with  lead and dlthlocarbamate or thluram  derivatives  In
rats.   Acta. Pharmacol.  Toxlcol.  56(4): 309-315.

Perelra, H.A.  1982a.  House  skin bloassay for chemical carcinogen.   0.  Am.
Coll.  Toxlcol.  1(1): 47-82.

Perelra,  H.A.   1982b.   Rat  liver  foci  bloassay.   J.  Am.  Coll.  Toxlcol.
1(1):  101-117.

Perelra, H.A.  and G.D.   Stoner.   1985.   Comparison  of  rat Hver  foci  assay
and strain  A  mouse lung  tumor assay to  detect  carcinogens: A  review.   Fund.
Appl.  Toxlcol.  5(4): 688-699.

Peto,   R.  1974.   Guidelines on the  analysis  of tumor rates and death rates
In experimental animals.   Br.  J. Cancer.  29:  101-105.

Peto,   R., P.  Lee  and W.  Paige.   1972.   Statistical  analysis of  the  bloassay
of continuous carcinogens.  Br. J. Cancer.   26: 258-261.

Pfelffer, E.H.  1977.  Oncogenlc  Interaction  of carcinogenic and noncardno-
genlc  polycycllc   aromatic  hydrocarbons  In  mice.   IARC   Scl.  Publ.  1977.
16(A1r Pollut. Cancer Han Proc. Hanover  Int. Carclnog.  Meet.,  2nd): 69-77.
                                    7-20

-------
Pilot,  H.C.  and A.E.  Sirica.   1980.  The  stages  of initiation  promoters  in
hepatocarcinogenesis.   Biocihim. Biophys.  Acta.   605:  191-215.  ;

Pitot,  H.C.,  T.  Goldsworthy, S. Moran,  A.E.  Sirica and W.J.  McArdie.   1982.
Properties of  incomplete carcinogens and promoters  in  hepatocarcinogeneisis.
    -: ' iA^j '.:... .'- ' '•..  . • " '         •      - .      - • -•    "'•     •   ,      ,
Carcinog. Compr. Surv.   7: 85-98.
                                                               '[
Plaa,  6.L.  and  W.R.  Hewitt.   1981.   Methodological  approaches  for  inter-
                                                               I
action  studies:  Potentiation  of  haloalkane-induced  hepatotoxicity.    In.:
Workshop  on  the  Combined  Effects  of  Xenobiotics.  Associate  Committee  on
Scientific  Criteria  for Environmental  Quality.   National  Research Council
,\ '."".'.     i.* ..-*-•         •-     1 ,                      • ' -1       -- • .     .     •
Canada, 1981.  p. 67-96.

Plaa,  G.L.,   W.R.   Hewitt,   P.  du  Souich,  6.   Caille  and  S.  Lock.    1982.
Isopropanol and  acetone potentiation of carbon  tetrachloride-lnduced hepato-
toxicity:  Single  versus  repetitive  pretreatments  in  rats.   J.  Toxlcol.
Environ. Health.  9: 235-250.
Plackett,  R.L.  and P.S.  Hewlett.   1952.   Quanta!  responses
poisons.  J. Roy. Stat. Ser.  B14{2): 141-163.
to mixtures  of
Plewa, M.3.   1984.   Plant  genetic  assays to  evaluate  complex environmental
mixtures.  In.: Short-Term Bioassays  in  the Analysis of Complex Environmental
Mixtures  IV,  Waters,   Sandhu,   Lewtas,   Claxton,   Strauss  and  Nesnow,  Ed.
Plenum Press, New Yprk.  p.  45-64.
                                    7-21

-------
Polger, H. and C. Schlutter.  1980.   Influence  of  solvents  and  adsorbents  on
dermal  and  Intestinal  absorption of  TCDD.   Food  Cosmet.  Toxlcol.   18(5):
477-481.

Purchase, I.F.H., E.  Longstaff,  J. Ashby, et  al.   1976.   Evaluation  of  Six
Short-Term Tests  for  Detecting  Organic Chemical Carcinogens  and  Recommenda-
tions for Their Use.   Imperial Chemical  Industries,  Ltd.,  Central  Toxicology
Laboratory,  Alderly  Park, Cheshire, UK.  p. 624-627.

Puurunen, J.,  P.  Huttunen  and   J.  Hlrvonen.    1983.   Interactions  between
ethanol and acetylsallcycllc acid  1n  damaging the  rat  gastric mucosa.   Acta.
Pharmacol. Toxlcol.   52(5): 321-327.

Raabe,  0.6.    1987.    Three-dimensional   dose-response  models  of  competing
risks and natural  life span.  Fund. Appl. Toxlcol.   8:  465-473.

Radlke,  H.J.,  K.L.   Stemmer and  E.   Blngham.   1981.  Effect  of  ethanol  on
vinyl chloride carclnogenesls.   Environ.  Health  Perspect.   41: 59-62.

Rao,  T.K.,   J.L.  Epler,   H.R.  Guerln,  B.R. Clark,  and   C.-H.  Ho;    1980.
Mutagenldty of  nitrogen  compounds  from  synthetic  crude  oils:  Collection,
separation,   and  biological  testing.    In.:  Short-Term   Bloassays   In  the
Analysis  of  Complex  Environmental Mixtures  II,  Waters,   Sandhu,  Hulslngh,
Claxton and  Nesnow,  Ed.  Plenum Press, New York.  p.  243-252.
                                    7-22

-------
Redei, G.P.  1980.  Arabldopsls assay  of  environmental  mutagens.   In:  Short-

Term Bloassays In the Analysis of  Complex Environmental  Mixtures  II,  Waters,

Sandhu,  Hulslngh,  Claxton   and   Nesnow,  Ed.   Plenum   Press,   New   York.

p. 211-232.                                                    ;



Relf, A.E.  1984.   Synerglsm  In  carcinogenesis.  J. Natl.  Cancer  Inst.   73:

25-39.



Rlnkus,  S.J.  and  M.S.   Legator.    1979.   Chemical  characterization  of  465

known or suspected  carcinogens and  their  correlation with mutagenlc activity

In the Salmonella typhimuriuin system.  Cancer Res.  39:  3289-3318.



Rlnkus,  S.3.  and M.S.  Legator.   1980.   The  need for  both In vitro and  jji
                                                               i
vivo systems In mutagenlclty  screening.   In:  Chemical Mutagens,,  Vol.  6,  F.J.

de Serres  and  A. Mollcender, Ed.   Plenum Publishing Corporation,  New  York.

p. 365-473.



Rosenkranz, H.S.  and  L.A.  Poirier.   1979.   Evaluation of  the  mutagenicity

and  DNA-modlfying  activity of  carcinogens and  noncarcinogens  in  mlcrobial
                                                               i
systems.  3. Natl.  Cancer Inst.   p. 873-891.                   '
                                                               [


Rossman, T.G., L.W.  Meyer, J.P. Butler  and 3.M. Daisey.   1984.   Use of  the

microscreen assay  for airborne particulate organic matter.   I_n:  Short-Term

Bioassays  in  the  Analysis  of  Complex   Environmental  Mixture:;  IV,  Waters,

Sandhu,  Lewtas,  Claxton, Strauss  and  Nesnow, Ed.   Plenum Press,  New  York.

p. 9-24.
                                    7-23

-------
Rous,  P.  and  J.G.   Kldd.   1941.   Conditional  neoplasms  and  subthreshold
neoplastlc states.  J. Exp. Med.  73: 365-390.

Rulls, A.M.   1987.  Safety  assurance margins  for food additives currently In
use.  Reg. Toxlcol. Pharmacol.  7: 160-168.

Schalrer,  L.A.,  J. Van't  Hof,  C.G. Hayes,  R.M. Burton and  F.J.  de Serres.
1978.  Measurement  of biological  activity of  ambient  air mixtures  using  a
mobile  laboratory  for  in  situ   exposures:   Preliminary  results   from  the
Tradescantla  plant  test  system.  In.: Application  of  Short-Term Bloassays In
the  Fractionatlon  and Analysis of  Complex  Environmental Mixtures,  Waters,
Nesnow,   Hu1s1ngh,  Sandhu   and  Claxton,  Ed.    Plenum   Press,   New  York.
p. 419-440.

Schalrer,  L.A.,  R.C.   Sautkulls and  N.R.  Tempel.  1983.   A search  for  the
Identity  of   genotoxic  agents  In  the  ambient  air  using the  Tradescantla
bloassay.  In,: Short-term Bloassays  1n  the Analysis of  Complex Environmental
Mixtures  III, Waters,  Sandhu,  Lewtas,  Claxton,   Chernoff  and Nesnow,  Ed.
Plenum Press, New York.  p. 211-228.

Schlff, L.J.,  S.F.  Elliott, S.J. Moore,  M.S.  Urcan and  J.A.  Graham.   1983.
Unscheduled DNA synthesis In hamster  trachea!  epithelium  exposed in  vitro to
chemical   carcinogens  and environmental  pollutants.   In;  Short-term Bloassays
1n  the  Analysis  of  Complex  Environmental   Mixtures   III,  Waters,  Sandhu,
Lewtas,   Claxton,  Chernoff  and   Nesnow,  Ed.   Plenum   Press,   New  York.
p. 277-284.
                                    7-24

-------
Schoeny,  R.,  D.  Warshawsky,, L.  Hollingsworth,  M. Hund and  G.  Moore.   1981.
Mutagen1c1ty  of   products  from  coal  gasification  and  liquefaction  1n  the
Salmonella mlcrosome assay.  Environ. Mutagen.  3: 181-195.

Schoeny,  R.,   P.  Warshawsky  and  6.  Moore.   1986.   Non-additive  mutagenic
responses by  components of  coal-derived  naturals.  ACS  Symposium on Chemical
Basis  for  lexicological  Response  in  Synthetic Fuels.   American  Chemical
Society Division of Fuels Chemistry, New York, NY.  Vol. 31:  147-155.

Schulte-Herman,  R.   1985.   Tumor  promotion  in  the  liver.   Arch.  Toxicol.
ISSN 0340-5761.  57(3): 147--158.

Seller, F.A.  and  B.R.  Scott.   1987.  Mixtures of toxic agents  and  attribut-
able risk calculations.  Risk Analysis.  7(1): 81-90.

Seitz,  H.K.   1985.   Alcohol  effects  on  drug-nutrient interactions.   Drug
Nutr. Interact.  4(1-2): 143-146.

Selikoff,  I.J.,   B.C.   Hammond  and J.  Churg.   1968.    Asbestos  exposure,
smoking, and neoplasia.  G.  Am. Med. Assoc.  204(2):  104-110.

Selikoff,  I.,  H.  Sedman   and  E.  Hammond.   1980.    Mortality  effects  of
cigarette smoking among amosite  asbestos  factory workers.   ;j.  Nat.  Cancer
Inst,,  65: 507-513.                                           :
Slaga, T.J.   1984.   Multistage skin carcinogenesis:  A  useful model for  the
study  of  the  chemopreventlon  of  cancer.   Acta  Pharmacol  Toxicol.   55(2):
107-124.
                                    7-25

-------
Slaga, T.J.,  S.M.  Fisher,  L.L.  Triplett  and S.  Nesnow.  1982.  Comparison of
complete  carcinogenesis  and  tumor  initiation  in  mouse skin:  Tumor  initia-
tion-promotion  a  reliable short  term assay.   J.  Am.  Coll.  Toxicol.   1(1):
83-89.

Sleight,  S.   1985.   Effects  of PCBs and  related  compounds  on hepatocardno-
genesls in rats and mice.  Environ. Health Perspect.  60: 35-39.

Smlth-Sonneborn, J.,  E.A.  McCann  and R.A. Palizzi.  1983.   Bioassays  of  oil
shale process waters  in  Paramecium  and  Salmonella.   Ir»:  Short-term Bioassays
In  the  Analysis   of  Complex  Environmental  Mixtures   III,  Waters,  Sandhu,
Lewtas,   Claxton,   Chernoff   and   Nesnow,  Ed.   Plenum  Press,   New  York.
p. 197-210.

Smyth, H.F.,  C.S. Weil,  O.S.  West and  C.P.  Carpenter.   1969.  An exploration
of  joint   toxic  action:   I.  Twenty-seven  industrial  chemicals  intubated  in
rats in all possible pairs.  Toxicol. Appl. Pharmacol.   14:  340-347.

Smyth, H.F.,  C.S. Weil,  G.S.  West and  C.P.  Carpenter.   1970.  An exploration
of  joint  toxic  action:  II.  Equitoxic  versus equivolume mixtures.   Toxicol.
Appl. Pharmacol.  17: 498-503.

Stara, J.F.,  D. Hukerjee, R.  McGaughy,   P.  Durkin  and M.L.  Dourson.   1983.
The  current  use of  studies  on promoters  and cocarcinogens  In  quantitative
risk assessment.  Environ. Health  Perspect.  50: 359-368.
                                    7-26

-------
Steenland, K.  and  M.  Thun.  1986.   Interaction  between tobacco  smoking  and
occupational   exposures  In  the causation  of  lung  cancer.   3.  Occup.  Med.
28(2): 110-118.                                               !
Stoner, G.D.  and  M.B.  Shlmkln.  1982.   Strain  A mouse  lung  tumor  bloassay.
J. Am. Coll. Toxlcol.  1(1):  143-169.

Strnlste,  G.F.,  J.M. Blngham,  W.D. Spall,  G.W. Nlckols,  R.t.  Oklnaka  and
D.J.-C. Chen.   1983.  Fractlonatlon of  an oil  shale  retort process  water:
Isolation  of  photoactive  genotoxlc components.   In.: Short-term  Bloassays  1n
the Analysis  of  Complex  Environmental  Mixtures   III, Waters,  Sandhu,  Lewtas,
Claxton, Chernoff and Nesnow, Ed.  Plenum Press,  New York.   p. 139-152.
Suglrnura,  T.,  S.  Sato,  M.  Nagao,  et al.   1976.   Fundamentals  1n  Cancer
Prevention.  Univ.  of  Tokyo  Press, Tokyo/Un1v.  Park Press, Baltimore,  MD.
p. 191-215.                                                   •

Sun, Y.P. and E.R. Johnson.  1960,   Analysis  of  joint action of Insecticides
against housefHes.  J. Ecori.  Entomol.   53:  887-892.
                                                              i
Swenberg, J.A.,  M.A.  Bedell,  K.C.  Billings,  D.R.  Umbenhauer and A.E.  Pegg.
1982.   Cell-specific   differences  1n   06-alkylguan1ne  DNA  repair   activity
during  continuous exposure  to  carcinogen.   Proc.   Natl.   Acad.  Sc1.   79:
5499-5502.
Takahashl S., T.  Ohnlshl,  A.  Denda and Y. Konlshl.   1982.   Enhancing  effect
of  3-amlnobenzamlde  on  Induction  of  y-glutamyl   transpeptldase  positive
foci In rat Hver.  Chem-Blol. Interact.   39:  363-368.

                                    7-27                       '•

-------
Thllly,  W.6.,  0.   Longweel  and  B.A.  Andon.    1983.   General  approach  to
biological analysis  of  9 complex mixtures.  Environ.  Health.  Perspect.   48:
129-136.

Thorslund,  T.  and   G.  Charnley.   1987.   Use  of  the multistage  model  to
predict  the  carcinogenic response  associated  with  time-dependent  exposures
to multiple  agents.  In:  Current Assessment  of Combined Toxicant  Effects.
ASA/EPA  Conferences  on  Interpretation of Environmental Data,  May  5-6,  1986.
EPA 230-03-87-027.

Tjalve,  H.  1984.   The  aetiology of SMON may  Involve  an  Interaction between
cHoqulnol and environmental metals.  Med. Hypotheses.   15(3):  293-299.

Troll, W. and  R.  Wlesner.   1985.  The  role of  oxygen  radicals as  a possible
mechanism of tumor promotion.  Ann.  Rev. Pharmacol.  Toxlcol.   25:  509-520.

Trosko,  O.E.,  C.C.  Chang and A. Medcalf.  1983.  Mechanisms  of tumor promo-
tion:  Potential  role of Intercellular  communication.   Cancer  Invest.   1(6):
511-526.

U.S.   EPA.   1984.  Health Effects  Assessment  for Polychlorinated  Biphenyls.
Prepared by the  Office  of Health and  Environmental  Assessment, Environmental
Criteria and  Assessment Office, Cincinnati, OH  for the Office of  Emergency
and   Remedial   Response,   Washington,   DC.     EPA   540/1-86-004.    NTIS
PB86-134152/AS.
                                    7-28

-------
U.S.  EPA.   1985.   Health Assessment  Document  for Polychlorlnated Dlbenzo-p-
Dloxlns.   Office  of  Health  and  Environmental  Assessment,  Environmental
Criteria  and Assessment  Office,  Cincinnati,  OH.   EPA/600/8--84/014F.   NTIS
PB86-122546.

U.S.  EPA.   1986a.   Guidelines  for  the  Health  Risk Assessment  of  Chemical
Mixtures.  Federal Register.  51(185): 34014-34025.
U.S. EPA.   1986b.   Guidelines  for the Health Assessment  of
mental Toxicants.  Federal Register.  51(185): 34028-34040.
Suspect Develop-
U.S.  EPA.   1986c.   Superfund Public  Health  Evaluation  Manual.   Office  of
Emergency and Remedial Response,  U.S. EPA, Washington, DC.  EPA/540/1-86/060.

U.S. EPA.  1987a.  The Risk Assessment;Guidelines of 1986.  EPA/600/8-87/045,
                                      i
p. 3-1 to 3-16.

U.S.  EPA.   1987b.   Integrated Risk  Information System  (IRIS).   Carclnogen-
Iclty  Assessment for  Lifetime Exposure  to Nickel  Refinery Dust.   Online.
(Verification   date   04/01/87.)    Office   of  Health   and   Environmental
Assessment, Environmental Criteria and Assessment Office,  Cincinnati, OH.
U.S. EPA.   1987c.   Interim procedures  for  estimating risks  associated  with
exposures  to mixtures  of  chlorinated d1benzo-p_-d1ox1ns  and  dlbenzofurans
(CDDs and DCFs).  Risk Assessment Forum, Washington,  DC.   EPA 625/3-87/012.
                                    7-29

-------
U.S.  EPA.   1988.   MIXTOX.  Studies  on  toxlclty of mixtures  and Interacting
chemicals.   User's  guide.   Environmental  Criteria  and  Assessment  Office,
Cincinnati, OH.

Van Duuren, B.L.   1969.   Tumor-promoting agents In two-stage carclnogenesls.
Prog. Exp. Tumor Res.  11: 31-68..

Van Duuren,  B.L.   1976.   Tumor-promoting and cocarclnogenlc  agents  1n chem-
ical  carclnogenesls.  In.: Chemical  Carcinogen,  ACS Monograph 173,  Chapter 2,
American Chemical Society, Washington, DC.  p. 24-51.

Veldstra,  H.   1956.  Synerglsm and  potentlatlon  with special  reference to
the combination of structural analogues.  Pharmacol. Rev.   8: 339-387.

V1g,  B.  1980.  Soybean  system  for  testing  the  genetic effects of Industrial
emissions and liquid effluents.   In:  Short-term Bloassays  1n the Analysis of
Complex Environmental Mixtures  II,  Waters,  Sandhu, Lewtas  Hulslngh,  Claxton,
Nesnow, Ed.  Plenum Press, New York.  p. 233-242.

Wahrendorf, G., R.  Zentgraf  and C.C. Brown.  1981.   Optimal  designs  for the
analysis  of  Interactive effects  of  two  carcinogens or  other  toxicants.
Biometrics.  37: 45-54.
Waters, M.D.,  J.L.  Hulslngh  and N.E. Garrett.   1978.   The  cellular  toxlclty
of complex  environmental  mixtures.  In: Application  of Short-Term Bloassays
1n the Fract1onat1on and  Analysis  of Complex Environmental  Mixtures,  Waters,
Nesnow,  Hulslngh,   Sandhu  and  Claxton,   Ed.   Plenum  Press,   New  York.
p. 125-170.

                                    7-30

-------
Way,  J.L.  and  G.  Burrow:;.   1976.   Cyanide  Intoxication:  Protection  with

chlorpromazlne.  Toxicol. Appl.  Pharmacol.  36: 93-97.

                                                             I


Weisburger,  H.O.   and  G.M.  Williams.    1980.   Chemical  carcinogens.   In.:
                                                             i
Toxicology: The Basic Science  of  Poisons,  J.  Doull, C.D. Kliaassen  and  M.O.

Amdur, Ed.   Hacmlllan Publishing Co., Inc., New York.  p. 84-138.

                                                             i

WHO   (World   Health  Organization).    1981.   Health   Effects   of  Combined

Exposures  in  the  Work  Environment.   World  Health Organization  Technical

Report Series 662.  p. 1-76.

                                                             j

WHO  (World  Health  Organization).  1983.   Environmental Health  Criteria 27.

Guidelines  on  Studies   In  Environmental   Epidemiology.   IPCS  International

Programs on Chemical Safety,  WHO, Geneva.   351  p.            .

                                                             j

Whong, W.Z.,  3.0. Stewart, B.C.  Adamo  and T.  Ong.   1983.   Mutagenic detec-

tion  of  complex environmental mixtures  using the Salmonella/arablnose-resis-

tant  assay sytem.   Mutat. Res.   120:  13-19.                  |
                                                             i
                                                             !

Williams, G.M.  1984.   Modulation of chemical  carcinogenesis by xenobiotics.

Fund. Appl. Toxicol.  ISSN 0272-0590.  4(3 pt 1): 325-344.


                                                             i

Withey,  J.R.   1981.   Toxlcodynamics  and  biotransformation.   in:  Inter-

national Workshop of the  Assessment of  Multichemical  Contamination, Milan,

Italy.   Draft copy  courtesy of J. Withey.
Withey,  J.R.  and  3.W.  Hall.   1975.   The   joint toxic  action  of perchloro-

ethylene with benzene or toluene in rats.  Toxicology.  4: 5-15.


                                    7-31

-------
Wolfenbarger,  D.A.   1973.   Synerglsm  of  toxaphene-DDT  mixtures  applied
topically  to  the  bollworm  and  the  tobacco  budworm.   3.  Econ.  Entomol.
66(2): 523-524.

Woolverton, W.L.  and  R.L.  Balster.   1981.  Behavioral and  lethal  effects  of
combinations  of  oral  ethanol  and  inhaled  1,1,1-trichloroethane  in  mice.
Toxlcol. Appl. Pharmacol.  59: 1-7.

Wysocka-Paruszewska,  B,  A.  Osicka,  J. Brzezinski  and  I.  Gradowska.   1980.
An  evaluation  of the  toxlcity  of thiuram  1n combination with  other  pesti-
cides.  Arch.  Toxicol.  Suppl. 4(Further  Studies  in the Assessment of  Toxic
Action): 449-451.

Yamada, K.,  A. Yatsuzuka, M.  Yasuhara,  et  al.  1986.   Mechanisms of  phar-
macoklnetlc Interaction  between  ajmaline  and quinidlne in  rats.   J.  Pharma-
coblo-Dyn.  9: 347-351.

Yamasakl,  H.   1984.  Tumor promotion  —  Mechanisms and implication  to  risk
estimation.  Acta. Pharmacol. Toxicol.  55(2): 89-106.

Zweldlnger, B.B.   1982.   Emission factors  from diesel  and  gasoline  powered
vehicles:   Correlation with  the  Ames  test.   In:  Toxicological  Effects  of
Emissions   from Diesel Engines,  J.  Lewtas,  Ed.  Elsevier Science  Publishing
Co., The Netherlands,  p. 83-96.
                                    7-32

-------
             APPENDIX A



AGENCY DATA BASE ON MIXTURE TOXICITY
                A-l

-------
OVERVIEW
    The  toxic  Interaction  data   base  contains  Information  obtained  from
literature  searches  of all  published  studies  on Interactions  between  toxic
chemicals  (U.S.  EPA,  1988).  The  goal  Is  to be complete,  not  merely  repre-
sentative,  so  that analysis  of  the data,  e.g.,  for trends  across  chemical
classes,  can  be  performed  If   desired.   This  version  does  not  contain
extensive  quantitative  data.  This constraint Is consistent  with  the  Agency
Guidelines  for  the Health  Risk  Assessment of  Chemical  Mixtures  (U.S.  EPA,
1986),  which  do  not   recommend  any   quantitative method   for   Including
Interaction  data  Into a risk  assessment.   As  a result,  the  current version
of  the  data  base  1s   most useful  for  a qualitative  evaluation  of  the
potential types of toxic Interaction between two environmental chemicals.
    The  data  base  package  contains   a   User's  Guide,  diskettes  (IBM  PC
compatible), and  a table  for Interpreting the  CASSI codes  (CAS,  1980) for
the  reference  citations.   The  data base  Is In  dBASE III  Plus  format.   The
access programs are compiled dBASE programs, and  can be  run without the need
for  dBASE  III Plus.   The  data  base 1s available  from  the  Risk  Assessment
Contacts  In each of the U.S. EPA's Regional Offices.
                                    A-2

-------
DATA BASE STRUCTURE
                                                              i
    The data base Includes  13  data fields, which are described  In  detail  In

the next section.  The structure 1s as  follows:
    Field Name
Type
Width
CAS-ONE
CHPD-ONE
CAS-TWO
CMPD-TWO
RTE-EXP
SPECIES
SEQUENCE
DUR-EXP
SITE
EFFECTS
INTERACT
AUTHOR
REFERENCE
Character
Character
Character
Character
Character
Character
Character
Character
Character
Character
Character
Character
Character
12
30
12
30
7
7
8
7
7
10
7
30
26
Description
                                           CAS No.  of first  chemical
                                           First chemical  name
                                           CAS No.  of second chemical
                                           Second chemical name
                                           Exposure route      \
                                           Animal species      !
                                           Treatment sequence
                                           Exposure duration
                                           Site of  adverse effects
                                           Type of  adverse effects
                                           Type of  Interaction
                                           First two authors (or  et al.)
                                           Reference code, volume:page
    An example  of  Input  format and corresponding on-screen  computer  display

Is Illustrated In Figure  1.

DESCRIPTION OF DESIRED FIELDS                                 |

Compounds

    Each compound Is listed as either  Compound I or  Compound II.

CAS Numbers

    CAS numbers corresponding to the above mentioned compounds are Included.

Routes of Exposure (codes provided - Table 1)
                                                              i
    The  routes  of  administration  are  specified  for each  compound and  are

listed  1n  order,   I.e.,   Compound  I  first  and  Compound   II  second.   For

example,  In  the  study by Short et al.  (1977),  vinylldene  chloride was given

via Inhalation while  d1sulf1ram was  administered  orally.  In  Figure  1,  this

Is Illustrated as follows:

                                   IHL; ORL
                                    A-3

-------
                                               CAS No.: 75-35-4
                                               CAS No.: 97-77-8
a)  Input Data Sheet

Compound I:  Vlnylldene chloride
Compound II:  D1sulf1ram
Route of Exposure:  IHL; ORL
Species:  HUS
Treatment Regimen:  II; I and SIM
Duration:  ACU
Site:  WBY
Effects:  HOR(I)
Qualitative Assessment:  INH
Reference:   Short, R.D., Winston,  J.M.,  Minor, J.L., Hong, C.,  Selfter,  0.,
and Lee, C.  1977.  Toxlclty  of  vlnylldene  chloride  In  mice and rats  and  Us
alteration  by various treatments.  0. Toxlcol. Environ.  Health.   3:  913-921.
b) Corresponding On-Screen Display
75-35-4       VINYLIDENE CHLORIDE
97-77-8       DISULFIRAH
Route:  IHL;ORL
Site:   WBY
                    Species:
                    Effects:
Author:  SHORT,RD ETAL
HUS
HOR(I)
Sequence:  II;I&SIM   Duration:  ACU
Interaction:  INH
Ref:  JTEHD6 (1977) 3:913-21
                                   FIGURE 1
            Example of  Interaction  Data Showing Original Coded Data
                         and On-Screen Representation
                                    A-4

-------
     TABLE 1



Route of Exposure
GAV
IAT
IAL
IBR
ICE
ICV
IDR
IDU
IHL
IMP
IMS
IPC
IPL
IPR
- gavage
- Intraarterial
- Intraaural
- 1ntrabronch1al
- Intracerebral
- 1ntracerv1cal
- Intradermal
- Intraduodenal
- Inhalation
- Implant
- Intramuscular
- Intraplacental
- Intrapleural
- Intraperltoneal
IRN
ISC
ISP
ITR
IVG
IVN
OCU
ORL
PAR
REC
SCU
SKN
UNR

- Intrarenal
- Intrascapular
- Intrasplnal
- Intratracheal
- Intravaglnal
- Intravenous
- ocular
- oral (dietary)
- parenteral
- rectal
- subcutaneous
- skin ;
- unreported
i
1
     A-5

-------
 When  the same exposure route 1s used for both compounds, the route Is listed
 only  once.   For example,  1f lead and zinc were both administered orally, the
 Input would  read
                                      ORL
 Species  (codes  provided - Table  2)
    The  species utilized In  the  study of  Interest
 Treatment Sequence
    This  field  specifies whether the compounds of Interest were administered
 simultaneously  or  sequentially.   If administration was sequential,  the order
 of  administration  Is  specified by  the  number  of  the compound.   In  Figure 1,
 "treatment  regimen"  Indicates  that dlsulflram was  administered  before (II;
 I)  and simultaneously  with   (SIM) vlnylldene chloride.   If  the  two  compounds
 had been administered concurrently, the format would read
                                      SIM
 Duration of Study
    The  duration  of  the  study  of  Interest 1s  classified as  either  acute,
 subchronlc, chronic or lifetime where
         acute = <14 days (ACU)
         subchronlc = >14 days but  not <90 days (SCH)
         chronic = >90 days   (CHR)
         lifetime = lifetime (LIF)
Sites  (codes provided In Table 3)
    The  site  or sites  affected  by the  compound  of Interest are entered  In
this  field.   In  Figure 1,  the  observed endpolnt  was  decreased  survival,
which  Is  considered  a whole  body effect.  Thus,  the  site  of  the effect  Is
coded  as  WBY.  Duration 1s  defined as  the  period between the beginning  of
                                    A-6

-------
                    TABLE  2
                    Species
CAT  -  cat
CTL  -  cattle
CHD  -  clilld
DOG  -  adult dog
DOM  -  domestic animals (goat, sheep, horse)
GRB  -  gerbll                              ;.
GPG  -  guinea pig
HAH  -  hamster
HMN  -  human
INF  -  Infant
MKY  -  monkey                              i
MUS  -  mouse
PIG  -  pig                                 ;
                                            i
RBT  -  rabbit                              j
RAT  -  rat                                 ;
SQL  -  squirrel
                    A-7

-------
      TABLE 3
Site/Organ Affected
ADR
BID
BHR
BRN
BRS
CAR
CER
CNS
COL
CVS
EAR
EHB
END
EYE
FAT
FET
GEN
GIT
HED
HRT
KDN
LIH
LNG
- adrenals
- blood
- bone marrow
- brain
- breast
- carcass
- cervix
- central nervous system
- colon
- cardiovascular system
- ears
- embryo
- endocrine
- eyes/ocular
- fatty tissue
- fetus
- genitals (external)
- gastrointestinal tract
- head
- heart
- kidney
- limbs
- lung
LVR
LYM
MMB
HSK
HTH
NSL
OVR
PAN
PLC
PNS
PUL
RBC
SEN
SKN
SOI
SPL
TES
THM
THR
UNS
UTS
WBY

- liver
- lymphocyte
- mucous membrane
- musculoskeletal
- mouth
- nasal passageways
- ovary
- pancreas
- placenta
- peripheral nervous system
- pulmonary system
- red blood cells
- gen. sensory
- skin
- site of Injection
- spleen
- testes
- thymus
- thyroid
- unspecified
- uterus
- whole body

      A-8

-------
treatment and the  time  when  the endpolnt assay  Is  conducted.   In  teratology
studies, exposure  during  gestation  1s  considered chronic to the life  of  the
fetus.
Effects (codes provided In Table 4)                           !
                                                             i
    The  effects  observed  at  the above-mentioned  site  or  sites.   In  cases
where  only  one  compound produces  an  effect  (potentlatlon,  no  apparent
Interaction,  Inhibition),  the compound number  Is placed  In  parentheses after
the code for effect.                                         I
    For  example,  In  Figure  1,  the  effect  of  Interest  Is  a  vlnylldene
chloride-Induced Increase In mortality (MOR).  Thus the "effects" field reads
                                    MOR(I)
    In  cases  where both  compounds  cause an effect at a  given  site (antago-
nism, addl-tlvHy,  synerglsm)  or opposite effects at a given  site  (masking),
the Interacting compounds are not listed 1n parentheses after  the effect.
Type of Interaction
    In  an  attempt  to   characterize  toxicant  Interactions,   a  scheme   of
classification  (see  Figure 2  for  an  outline) has been  devised  to distinguish
between  the  various  types   of  Interactions  encountered  1n  the  existing
literature.  The scheme Is as follows:
Both Compounds I and II Produce a Given Effect at a Given Site.
                                                             i
    1)   Additive  -  The  magnitude  of  the  effect  observed  In  the
         presence  of both  compounds 1s not quantitatively greater  or
         less than the  sum of  effects produced  by  each  compound alone.
         For  example,  both aldrln and aramlte cause Increased mortality
         when  administered  Individually  to  mice.   When  administered
         together,   the   observed  mortality  Is  equal   to  the  sum  of
         mortalities    observed   for   each    compound   administered
         Individually (simple response addition).            :j

                                     A-9

-------
                                    TABLE 4
                                Nature of Effect
 ABS  -  absorption  altered
 ALR  -  allergic  responses
         (I.e.,  hypersensltlvlty)
 COR  -  corrosive effects
         (burns, desquamatlon)
 DDP  -  drug  dependence
 DE6  -  degenerative changes
 DEP  -  depression  of function
 DIS  -  distribution altered
 ELI   -  elimination altered
 ENZ   -  enzyme activity altered
 EXC   -   excretion altered
 FUN   -   functional Impairment
 HEM   -   hematologlc changes
 HMR   -   hemorrhage
 IRR  -   Irritation
MET  -  metabolism altered
 MOR  -  mortality
 MUT  -  mutagenlc
 NBH  -  neurobehavloral  effects
 NEO  -  neoplastlc
 NPY  -  neuropathy
 OCC  -  ocular effects
 PIG  -  pigmentation changes
 PRO  -  prollferatlve changes
 REP  -  reproductive effects
 RET  -  retention altered
 STI   -  stimulation of function
 SUR   -  survival/viability altered
 TEM   -   temperature changes
 TER   -   teratogenlc
 UNS   -   unspecified effects
WGT   -  weight altered
                                    A-10

-------
A.  Both compounds produce a given effect at a given site or sites

    1.  Additive (ADD)
    2.  Antagonism (ANT)
    3.  Synerglsm (SYN)                                       ;
B.  Only one compound produces a given effect at a given site or sites

    1.  Inhibition (INH)                                      [
    2.  No Apparent Influence (NAI)
    3.  Potentlatlon (POT)
C.  Neither compound alone produces a given effect  but  when  placed together,
    an effect Is seen - Chemical Synerglsm (CSV)
                                                              !


D.  Compounds I and  II  produce  opposite effects at the same  site  or  sites  -
    Masking (MSK)
E.  Unable to assess (UTA)
                                   FIGURE 2

                       Types of  Interaction with Codes
                                    A-ll

-------
    2)   Antagonism  - The magnitude  of the  effect  In  the  presence of
         both  compounds  I  and II Is  less  than  would be expected 1n the
         case  of  addltlvlty.   For  example,  both 2,4-D butyl and 2,4,5-T
         butyl  produce  teratogenlc  effects  and  fetal mortality  when
         administered  alone;   however,   the  effects  seen  when  both
         compounds are administered together  are less severe than those
         seen  when  either  2,4-D butyl or  2,4,5-T  butyl Is administered
         alone,  and  hence,   less   than  expected  under the  addltlvlty
         assumption.
    3)   Synerglsm -  The  effect seen In the  presence  of both compounds
         1s  quantitatively  greater  than would  be expected  In  the  case
         of  addltlvlty.   For example  both  PCB  and  vlnylldlne  fluoride
         cause  an  alteration  In   enzyme  activity  In  the   Hver  when
         administered  Individually.   When  administered  together,  the
         effect Is quantitatively greater  than  would be expected In the
         case of addltlvlty.

Only One Compound Produces  a  Given Effect at a Given  Site.
    The  following  three  classifications  are  special  cases of  the  three
discussed previously:  addition, antagonism and  synerglsm.
    1)   No Apparent  Influence  - A noneffectlve compound, II, does  not
         modify  ostensibly   the  effect  produced  by  compound  I.   For
         example,  acrylamlde-promoted  neuropathy  Is  unaffected  by  the
         co-administration  of  cortlsol.   Cortlsol alone has no  effect
         upon  the peripheral  nervous system.   Thus,  cortlsol  has  no
         apparent  Influence  on the acrylamlde-promoted neuropathy.
                                    A-12

-------
     2)    Inhibition  -  The  noneffectlve  compound,  II,   quantitatively
          Inhibits  the  effect  produced  by  Compound  I.   An  tsxample of
          inhibition  Is  presented  in  Figure  1.   Vinylidene  chloride
          (compound   I)  caused . an  Increase  In  mortality,  which  was
          inhibited  by  co-administration  of disulflram  (compound  II).
          When  administered  alone,  dlsulfiram  had  no  effect  on  the
          survival  rate;  thus,  the Interaction was classified as  Inhibi-
          tion  rather than  "antagonism" or  "masking."
     3)    Potentiatlon  -  The noneffectlve  compound,  II,  enhances  the
          magnitude of  the  effect  produced  by compound I.   An example of
          potentlatlon   is    vinylidene   chloride-promoted  degenerative
          changes  in  the  liver, which  are enhanced quantitatively by the
          co-administration  of acetone.    Under  the  conditions  of  the
          experiment, acetone alone has no  effect upon the  liver.
                                                              i-
Masking
     The assessment of  "masking"  is reserved for  the  instance when compounds
I  arid II  produce opposite  effects  at  the  same site or sHes  and  either
diminish  or override the effects  of each other.   For example, zinc alone has
been  shown  to  cause  an  Increase  in  S-aminolevul1n1c  acid  dehydratase
activity  (ALA-D)  In  red blood cells,  while ethanol  alone causes  a  decrease
In ALA-D  In  red blood cells.  Co-administration  results   in  a  rise  in  ALA-D
quantitatively  similar  to  that   observed  when  zinc  was  administered  by
itself.   Thus,  on  the  Input sheet,  "Effects"  would  read  "ENZ"  and  qualita-
tive assessment would read "MSK."
Unable to Assess
    This   Is  used  for  studies  that  are  poorly  designed or  insufficiently
detailed  to discern the nature  of  the  Interaction.
                                    A-13

-------
 Reference
    The  Input data sheets  contain  the complete reference  (see  Figure 1  for
 an  example).   The data  base Includes  only the  first  two  authors   (second
 author  1s  "et  al."  1f more  than  two  authors),   year,  reference code  and
 volume:page  numbers.
 GENERAL COMMENTS
    In  most  cases,   Identical  data  generated  by  the  same laboratory  but
 reported  In  more  than one reference were  not repeated  In  the data base.   In
 addition,  results  reported  In  the text without  accompanying data were  not
 used  because an adequate  evaluation of the Interaction could  not be made.
This  data  base Is  only concerned  with  effects resulting  from  excess expo-
 sures, e.g., studies examining  the consequences  of  feeding  diets deficient
 1n an essential nutrient were not Included.
    In  general,  because  of  a  widespread  lack  of  adequate  statistical
methodology  In  the studies  reviewed,  assessing  the qualitative  relationships
between compounds  was often  difficult.   In many  cases,  It was left  to  the
judgment of  the reviewer whether  an  Interaction  existed at all and,  1f  so,
how to  classify It  according to the  scheme presented  above.   It  should  be
emphasized that this  data base should be  used  only as  a tool to  direct  the
user to the literature currently available  regarding toxicant Interactions.
                                    A-14

-------
                     APPENDIX B
DIESEL EXHAUST EMISSIONS AND "SUFFICIENT SIMILARITY"
                        B-l

-------
             DIESEL EXHAUST EMISSIONS AND "SUFFICIENT SIMILARITY"

    An Important concept  1n  the  Guidelines  for  the Health Risk Assessment of
Chemical  Mixtures  (U.S.  EPA,  1986)  Is the  use  of data  on  similar  mixtures
for  a  risk  assessment  on  the  mixture   of  concern.   This  procedure  Is
predicated  on  the  determination  of   "sufficient similarity"  between  the
mixtures.   In  brief,  If  health effects  data  on  a   similar  mixture  are
available,  1t  must be  determined  1f  the  mixture on which there  are data Is
or  1s  not  sufficiently  similar  to the  mixture of concern  to allow  a risk
assessment.    This  determination  should   Include  consideration   of  the
component  proportions  of  the  mixtures  as  well  as  any  toxlcologlc  or
pharmacoklnetlc data  on the components or  the mixtures  that  would assist In
assessing  the  significance of any chemical differences  between  the similar
mixture  and  the  mixture  of concern.   The  determination  of  "sufficient
similarity"  should  be  made  on  a  case-by-case  basis  In  light  of  the
uncertainties  associated  with using  data  on a  dissimilar  mixture  and with
using  other   approaches   such   as addltlvlty.    (For   further   Information
concerning  the applicability  of  the sufficient  similarity approach,  refer to
the guidelines.)
    Diesel  engine  exhaust represents  a family  of  complex mixtures  that  are
generated   with  varying   compositions  depending  on   different  temporal,
emission  source,   or  operating  condition  characteristics.   Because  dlesel
engine emissions were expected to make  a  significant contribution  to  urban
pollution,  the  U.S.   EPA  Instituted  a  major  research  program  aimed  at
quantifying the potential  health  and  environmental Impacts  of dlesel-powered
light-duty  vehicles  (U.S. EPA,  1979).   The purpose  of  this exercise  Is  to
use  these  data   to  determine   whether  dlesel   emissions   from  different
                                     B-2

-------
 sources  are sufficiently  similar  to warrant  their  use  for  the purposes of
 predicting  the  health  effects  of unknown dlesel emissions as outlined In the
 guidelines.                                                    I
    The  U.S.  EPA  research program  was  designed  to determine  the relative
 mutagenlc  and  carcinogenic  potency  of  extractable  organlcs;   from  dlesel
 partlculate  emissions  compared  with  particle-bound  organlcs  from  other
 environmental  emissions  (gasoline  engines,  cigarette  smoke  condensate,  and
                                                               i
 coke  oven  and  roofing  tar  emissions)  (Lewtas et  a!.,  1981).   The  mobile
 source  samples  selected  for  this  study  Included a  heavy-duty Caterpillar
 3304  dlesel engine,  three light-duty dlesel  passenger  car  engines  (Datsun
 Nissan  220C,  Oldsmoblle  350,  and  Volkswagen turbocharged  Rabbit),  and  a
 gasoline  catalyst  Mustang  II.   All dlesel engines were operated on the same
 lot  of  No. 2  dlesel  fuel.   In  addition,   all  vehicles  (except for  the
 Caterpillar)  were   operated   on  a  chassis   dynamometer   under  Identical
 conditions  using   the  highway  fuel  economy   test  cycle  (HWFET).   Particle
 samples from all  engines  were collected with  a dilution  tunnel  In which  the
 hot exhaust  was diluted,  cooled,  and filtered through  Pallflex  T filters.
All  samples were  extracted  by  a  Soxhlet  apparatus  with  dlchloromethane,
which was removed  by evaporation under dry nitrogen.
    The test matrix consisted  of  the following bloassays:  reverse mutation
 In  Salmonella   typhlmurlum;   sister  chromatld  exchange  (SCE)   In  Chinese
hamster ovary (CHO)  cells; gene mutation In L5178Y  mouse lymphoma cells  and
BALB/c  3T3  (3T3);  viral  enhancement of  transformation  In  Syrian  hamster
embryo  (SHE) cells;  oncogenic  transformation  In   3T3  cells;  and  skin  tumor
Initiation  In SENCAR mice.  Where  possible,  these bloassays were  conducted
such that a positive dose-response relationship was  observed over at  least
                                                               i
three doses above spontaneous  levels.   Comparative  potency rankings of  the
                                    .8-3

-------
 samples  were determined  based on  the  Initial linear  slope  of the response
 curve.   Where  dose-response  data  were  not  obtained,  the  lowest effective
 dose  (LOEL)  tested was  determined.
    A wide  range of  activity was observed 1n S^. typhlmurlum  strains TA98 and
 TA100 (Table B-l).   The majority of  the  activity  associated with the dlesel
 samples  was  direct  acting  while  the addition  of  a  mammalian  activation
 system Increased  the activity  of the  gasoline engine sample (Claxton, 1981).
    All  of   the  emission  samples gave  positive mutagenlc responses  both In
 the presence and absence  of metabolic activation using  the  criteria  of the
 L5178Y  mouse lymphoma  thymldlne klnase  (TK)  locus forward  mutation  assay.
 The dlesel  engine emissions  were more  cytotoxlc  In  the absence of metabolic
 activation  while cytotoxldty  Increased  In  the presence  of  activation with
 the Mustang  emissions.  Among  the  dlesel  engines,  the  Nissan emission sample
 was the  most cytotoxlc while  the Caterpillar  sample was the least cytotoxlc
 with  a potency below that of the gasoline engine (Mitchell et a!., 1981).
    Curren  et al.  (1981) assayed  the Caterpillar,  Nissan,  and  Oldsmoblle
 dlesel  samples,   and  the  Mustang  gasoline  sample  1n  the BALB/c  3T3  muta-
 genesls  assay.    Although  several   Individual  doses of the  dlesel  extracts
 Induced  a  significant  increase in  ouabain-reslstant  mutants,  none of  the
 samples yielded a dose-dependent Increase in mutation  frequency.   Based on a
 determination of mutation frequency  for   the  dose ranges  tested, both  the
 Nissan  and  Mustang   samples  were  significantly  mutagenic  both  with  and
without  metabolic activation  while  the  Caterpillar  and  Qldsmobile  samples
were  not significantly different from controls.
    Definitive  conclusions  concerning  the   DNA-damaging  capabilities   of
dlesel emissions  as  measured  by the  SCE  test  are  difficult to  reach  given
                                     B-4

-------
                                 •  TABLE B-l

              Specific Activities at TOO yg of Organic Material*
               In Salmonella typhlmurium Strains TA98 and TAiOO


Sample

Caterpillar
Nissan
Oldsmobile
VW

Mustang


+S9

59.3
1367.1
318.7
297.5

341.9

TA98
-S9
Diesel
65.9
1225.2
614.8
399.2
Gasoline
137.8

TAIOO
tS9

115.2
881.7
169.9
426.0

228.0


-S9

167.8
1270.1
247.5
641.6
-
196.5
*Source:  Claxton, 1981
                                     B-5

-------
 that the  results  are based on  single  experiments.   However, It  Is  signifi-
 cant that  an observed  Increase In  SCE  frequencies  In  CHO cells  following
 exposure to  all  except the  Oldsmoblle sample  In  the absence of activation
 Indicates   that  these   samples  contain  one  or  more  components  that  are
 direct-acting chromosome-damaging agents.  Although  the significance of the
 differences  among  these  dlesel  and  gasoline  emission  samples  cannot  be
 Inferred from the  data, this test gave the following  qualitative  comparative
 potency   ranking:   Nissan  >  Rabbit,  Mustang  »  Caterpillar,  Oldsmoblle
 (Mitchell  et  al.,  1981).
     Two  assays,  one  measuring  morphologic  transformation  In 3T3 cells  and
 the  other  measuring  viral  enhancement  of transformation In  SHE  cells,  were
 used to  observe the  effects of  gasoline and dlesel  emissions  on oncogenlc
 transformation.  As with  the  3T3 mutation assay,  dose-related, responses  1n
 transformation  frequency  1n 3T3  cells  were not  observed   for  any  of  the
 samples  (Caterpillar, Nissan,  OldsmobUe,  and Mustang).   Both the Nissan and
 Mustang  samples  Induced  a  significant number  of transformed  foci  In  the
 absence  of  metabolic  activation  while only emissions  from the  Mustang had  a
 transformation frequency  significantly  greater than that of  controls  In  the
 presence of metabolic activation (Curren et al.,  1981).
     In  the viral   enhancement  assay,  the  Nissan  appeared   to  be  the  most
 potent followed  by the  Rabbit  and Mustang,  which were equlpotent,  and  the
 OldsmobUe  and  Caterpillar according to  the  lowest  effective  concentration
 tested that Induced significant  enhancement  (Casto et al.,  1981).  However,
 1f the data from three  separate  experiments  were combined  to determine  the
 slope of the  pooled  dose-response curve  for each  sample,   the  comparative
potency ranking would be:  Nissan, Mustang > Rabbit  >  Oldsmoblle'(Caterpillar
1s considered negative).   Because the variation  1n response  between  the
                                     B-6

-------
 three   experiments   was  significant  (r    value   as   low  as  0.18),  each
 experiment  was  analyzed  separately  and   the  experiment  resulting  1n  the
          2
 highest  r  was  used to  determine  the  following potency  ranking:  Nissan >
 Rabbit  > Oldsmoblle, Mustang  (Lewtas, 1983).   Despite these variations, the
 ranking  for  light-duty  dlesel  engine samples remains fairly constant: Nissan
 > Rabbit > Oldsmoblle.
    In  the  skin tumor  Initiation  assay   In  SENCAR  mice,  the  four  dlesel
 samples  varied  significantly  1n  the tumorlgenlc  responses  they produced,
 ranging  In  activity  from  0 to  5.7  paplllomas/mouse  (Nesnow  et  al.,  1982).
                                                                i
 Paplllomas were  produced'1n .all  samples except  for  the Caterpillar.  Excess
 tumor  multiplicity  activities  1n  paplllomas  per  mouse  at  1   mg  of extract
 were calculated  as  follows: Nissan - 0.59,  Oldsmoblle - 0.31,   Rabbit - 0.24,
                                                                i
 and Mustang -  0.17  (Albert  et  al., 1983).   These data Indicate that only the
                                                                i
 Nissan  extract can  be  considered  a strong  tumor  Initiator,   with  activity
 similar to that of roofing tar.                                 !
    A comparison  of these  test systems reveals  that,  1n  general,  there  Is a
 consistency  1n the  comparative  potency of these  extracts  wHth  the  Nissan
 sample  the  most active and the  Caterpillar  sample  the least   potent  In  all
                                                                j
 bloassays.   The  main  Issue,  however,  Is  whether  these  data  demonstrate
 sufficient  biological  similarity  among  the  different  samples   to  warrant
 their  use  In  predicting  the  effects of  other dlesel  mixtures.    Based  on
 these data,  results from   heavy-duty  dlesel engine  (Caterpillar) emissions
would severely  underestimate  the effects of  a  light-duty dlesel  engine  and
                              i
 should not be  used  for  that purpose.  Within the light duty class of  dlesel
engines,  there  appears  to  be   reasonably  close   agreement   between   the
Oldsmoblle and Rabbit engines  while  the  Nissan Is  considerably more  potent.
                                                                  •
Because  of, the  Nissan  data,  It  would  not be  prudent to  assume that  all
                                     B-7

-------
 light-duty dlesel  engine  emissions  are  sufficiently  similar  as  to their
 biological effects.
     The  available  Information  on  components  of  the  four  dlesel  and  one
 gasoline   emission  samples  Indicates  a  wide  range  of  organic  extractable
 material   (Table  B-2).   Benzo(a)pyrene  (BaP)  content  per  mg extract  also
 varied considerably,  from 0.0002 to 0.11%  (Lewtas  et  a!., 1981).  Nesnow et
 al.  (1982)  state that  the  tumor  data  from  the SENCAR mouse  skin tumor
 Initiation assay  cannot  be explained  solely by BaP content since there Is no
 significant  relationship  between  tumor  Incidence  and  BaP  content  In  each
 complex mixture  (Including dlesel and  gasoline engine,  roofing tar, and coke
 oven  emissions).   They  estimate that  BaP  accounts for  only 20-30%  of  the
 activity  seen  and that  other constituents must  be contributing  toward  the
 tumorlgenlc activity.   Whether  this  contribution Is  through  Interaction  or
 direct component activity cannot be determined from the available  data.
    It  1s  evident  that   the  available component  data  does not  meet  the
 sufficient  similarity  criteria at least In  the case of BaP  content  for  the
Nissan and  Mustang samples.   If BaP can account  for.no more  than  30% of  the
tumorlgenlc  activity  of   the   mixture,  It   Is   apparent  that   additional
component   Information   Is   necessary   before   the   Issue   of   sufficient
constituent similarity can be accurately evaluated.
                                     B-8

-------
                     TABLE  B-2                    ''
Results of Extraction and Benzo(a)pyrene Analysis*
Benzo(a)Dvrene
Extractable
Matter nq BaP riq BaP
Sample Source percent mg extract mg partlculate
Diesel CAT 26-27 2
NISSAN 4-8 1173
OLDS 12-17 2
VW RAB 18 26
Gasoline MUSTANG 39-43 103
*Source: Lewtas et al., 1981
0.5
96.2
0.4
4.6
44.1

                      B-9

-------

-------
                          APPENDIX  C                   i



ANALYSIS OF THE SAMPLE STUDIES FROM THE INTERACTION DATA BASE
                            C-l

-------
     Thirty-two studies were selected from the U.S. EPA  Interaction data base
 (see Appendix  A)  for  detailed evaluation  of  the statistical  methods that
 were employed In determining  the  type  of toxic Interaction.  The evaluation
 Included the appropriateness of the statistical method used and the correct-
 ness of  Interpretation  of  the  statistical  results.   The  10% random  sample
 was  stratified by  the type of  statistics  used.   The following  text  Is the
 evaluation  of each  study.
     Carlson (1973)  pretreated  rats with either phenobarbltal  (PB), 3-methyl-
 cholanthrene (3-MC),  saline or  corn oil vehicle, then  exposed  them to air,
 1,1,1-trlchloroethane   or   1,1,2-trlchloroethane.   Endpolnts  assessed  were
 Hver  and  body  weights,   serum glutamlc  oxaloacetlc  transamlnase  (S60T),
 serum  glutamlc pyruvlc transamlnase  (SGPT)  and liver glucose-6-phosphatase.
 Analysis  was  by  2-way analysis  of variance  to  assess  differences  between
 pretreatments,  Inhalation  treatments and  the Interactions  between  the two.
 The  analysis was appropriate.   No  differences were  found  In liver  or body
 weights,  and  1t  was  concluded  that  3-MC  did not  potentiate  the  hepato-
 toxldty  of  the  trlchloroethanes,   but  PB  did and  enhancement  was  greater
 with 1,1,2-trlchloroethane than with 1,1,1-trlchloroethane.
     Short et  al.  (1977) exposed  mice and rats to  continuously Inhaled air  or
 1,1-dlchloroethylene  (VDC),  then  to one of  dlsulflram,  dlethyldHhlocarba-
 mlde   (DDC),   thlram,   aptelne,   methlonlne,   N-acetylcystelne,   SKF   525-A,
 cobaltous chloride, phenoxybenzamlne, propanolol,  Vitamin  C  or to  Vitamin  £.
 Endpolnts assessed were death,  organ damage  as  assessed  by serum enzymes and
 hlstopathology, changes  In  Hver and  kidney, and radioactivity In  protein.
 Statistical  methods  used  were  calculation   of  the   LCcn  and LTcn  for VDC
                                                       bU        bu
 for  assessment  of survival, and  the 2  sample  rank  test and  Fisher's  exact
 test for  the  other  endpolnts.   No methods were used  to  control  for  multiple
comparisons.  Survival  analysis would have been more  appropriate to  use.   No
                           %
                                    C-2

-------
negative control  group  was  present.  Repeated measures analysis  should  have
been used  to assess  the  changes across  time In  SGOT  and SGPT.   They  con-
cluded that  dlsulflram  reduced the  severity  of  the lethal,  hepatotoxlc  and
renal  effects  of VDC  In mice,  that DDC  and thlram  protect; mice  from  the
lethal effects  of VDC,  and that  the dlthlocarbamates protected  against  the
toxlclty of VDC.                                            .  ;
    Castro et al.  (1974)  exposed rats to SKF 525-A, Sch  5705,  Sch. 570.6,  Sch
5712, CFT  1201,  Lilly  18947,  DPEA,  promethlzlne or vehicle control, then to
                                    Endpolnts   assessed  were!  ethylmorphlne
N-demethylase activity  (EM-ase),  cytochrome P-450 activity,  peroxldatlon of
CC1,,  or   olive  on   vehicle.
   4
liver  mlcrosomal  llplds,  CC1.  concentration  In  liver,  protein  concentra-
tion, and  NADP-llnked  Isocltrlc  dehydrogenase (ICD) activity Mn  plasma.   In
examination of  the  time  course  of CCl^  concentration,  they  used  Student's
t-test and  the Mann-Whltney-U test  for comparison  at  each time point.   In
this  situation the  groups  were CC1.  alone  vs.  CC1. and other  compound,
                                     «t                HI
sometimes  varying  the  level  of  CC14.   To  examine the  effects  on  EM-ase,
ICD and  P-450,  they made  comparisons  via  2~way ANOVA.  To  examine  the  time
course of  llpld  peroxldatlon and body  temperature,  they used a  2-way  ANOVA
at  each  time  point,  whereas  a  repeated measures  analysis would have  been
correct.   They concluded  that  "although these compounds tested are  known  to
Inhibit  cytochrome  P-450   dependent   drug-metabolizing   enzymes   In  liver
mlcrosomes, they apparently  do not evoke  their  protective  effects  by slowing
the  elimination  of  CC1.."   These conclusions  are  appropriate  In   light  of
the methods used.
    Andrews et  al.  (1977)  examined  the  effects of  toluene
disposition and  hemopoletle toxlclty  of  [H3]benzene, particularly  red  cell
59Fe  Incorporation  as  a   measure  of  erythropolesls.   Two
                                    C-3
                                                              on  metabolism,
                                                               2x2  factorial

-------
 experiments  were  conducted  on  mice,  with the  administration  of  0  or  880
 mg/kg benzene and 0  or  1720 mg/kg toluene as one experiment, and  the admin-
 istration  of  0  or  440  mg/kg  benzene and  0  or  1720  mg/kg  toluene  as  the
 other.   Endpolnts assessed  were benzene metabolites  In  urine,  expressed as
 percent  administered  dose  and  as  benzene equivalents,  percent 59Fe utili-
 zation,  exhaled  [3H]benzene, and  levels  of  [3H]benzene  1n  liver, spleen,
 epldldymal  fat pads, blood,  and  bone  marrow.   Although the experiments were
 conducted  as  2x2 factorials,  making  the  use  of 2-way  ANOVA  appropriate,
 Student's  t-tests  were  In  fact  used.    Furthermore,  the  time   course  of
 accumulation  of  [3H]benzene  In tissues was  analyzed  by t-tests  at  each time
 point,  whereas  repeated measures  analyses were appropriate.  The authors
 concluded  that toluene  reduced  the level of  urinary  metabolites  of benzene
 and   also   reduced   the  benzene-Induced  Inhibition   of  erythrocyte  S9Fe
 uptake.  These conclusions  are  consistent with  the  results  of  the statis-
 tical methods  used,  but  may  not  be valid due  to the Increased likelihood of
 false positives with  these methods.
    Friedman  and  Eaton  (1978) studied  the effects of  an  Inhibitor of  mixed
 function oxldase  (MFO)  activity,  plperonyl butoxide  (PB),  on methylmercury
 (MM)  toxlclty.   Rats were  fed  diets containing  either  0,  20  or 40  ppm
methylmercury, and either 0,  0.5  or 1.0% PB.   Endpolnts assessed were weight
gain and mortality.   No  statistical methods and  no dose-response models  were
used.  The authors  conclude  that  "PB  synergises  MM  poisoning in a  dose-
dependent fashion."
    Blanclflorl  et  al.   (1967)   examined  the  effects  of  estrogen  on  the
pathway  'through  which   chemical   carcinogens  exert   their   action.    Both
ovarlectomlzed and  intact mice  were given  estrone  at 0,  500 or   1000  tig/9.
drinking  water,  and  the mice were administered either  nothing,  or one  of
                                    C-4

-------
9,10-dimethyl-l,2-benzanthracene   (DMBA),   1,2:5,6-dibenzanthracene   (DBA),
20-methylcholanthrene  (MC)  or  3,4-benzopyrene  (BP)  at  0.5% In almond  oil,
twice  weekly  until  8  weeks  of  age.    Endpoints   assessed  were  mammary
carcinoma,  survival,  gastric  tumor,   ovarian  tumor,   leukemia,   and  lung
tumor.   Although  no  statistics  were  used,   the   authors  conclude  that
"administration of oestrone Increased the  Incidence  of  mammary carcinomas  In
both  Intact  and  ovarlectomlsed  mice when  DBA or MC were  the  carcinogens;
only  a  minimal effect  was obtained  with  BP  and the  result with DMBA was
equivocal,"  "squamous  carcinomas  of   the  forestomach  occurred   when  the
carcinogen was BP or DMBA,," and "DBA with oestrone Induced ovarian  tumours."
    Cone and  Netteshelm (1973)  Investigated  the effects  of high   levels  of
vitamin A  on  the toxlclty of  3-methylcholanthrene  (MCA)  In  the respiratory
tract  epithelium  of  the  rat.   All  animals  received  vitamin  A,   either  In
doses of  17,  87  or  1740  yg/week, and  either  0  or  5 mg MCA.  The endpolnt
assessed was  respiratory  tract  tumor.   Although no statistics were  given,
the  authors  concluded  that   vitamin  A  has  an Inhibitory  effect  on  the
                                                            i
development of respiratory tract tumors.
    Daoud  and  Griffin   (1980)  Investigated  the effect  of  retlnoic acid,
butylated  hydroxytoluene  (BHT),  selenium  (Se)  and   sorbic  add  on  azo-dye
                                                            I
hepatocarcinogenesls  1n  the  rat.  All  animals  received  a;  diet  containing
0.05%  3'-methyl-4-dimethylam1noazobenzene  (3'-MeDAB),  and  either  nothing,
0.05% BHT, 1 or 2%  sorbic  acid,  0.02%  retlnoic acid  or  2 or  4 ppm  Se,  but  no
                                                            i
combinations or the  latter four compounds.  The  endpoint  assessed  was liver
carcinoma.   Although no  statistics  were  given, the  authors conclude  that
                                                            I
protection was  "afforded the  animals given  the azo  compound  by the  dietary
supplementation with either retlnoic add or  BHT."
                                    C-5

-------
     Schlede et al.  (1970)  examined  the stimulatory effect of  benzo(a)pyrene
 (BaP)  and phenobarbltal pretreatment on the biliary excretion  of BaP metabo-
 lites  In the rat.  Animals were pretreated with either BaP,  phenobarbltal or
 vehicle,  then   received  either   10  or   300   vg   of   14C-labeled  BaP
 (BaP-14C).  The endpolnt  assessed  was  the  rate of  excretion of metabolites
 of   BaP-14C  Into  bile.   Although   no  statistics  were  given,  the  authors
 conclude that  "pretreatment of  rats with  BaP  or  phenobarbltal prior to the
 l.v.  Injection of  10 or  300  VS of  BaP-14C  enhances   the  rate of excretion
 of metabolites of BaP-14C  Into the bile."
     Ito  et  al.   (1973)  examined  the  effect  of  polychlorlnated  blphenyls
 (PCBs)   on  tumorlgenesls  by  benzene  hexachlorlde  (BHC)   1n  mouse  liver.
 Animals  received  diets containing 0,  100,  250 or 500  ppm  PCBs,  alone  or 1n
 conjunction  with  0,  100  or 250 ppm of a,  13  or  Y-BHC.   Endpolnts assessed
 were  hlstopathology  of   the  liver,   liver  weight  and  body  weight,   No
 statistics  were  given.   The  authors conclude  that  "PCBs  themselves  Induced
 hepatic  neoplasms  In  mice  and also  promoted  the Induction  of  tumors  by
a-BHC and B-BHC."
    Hagos et al.  (1974)  describe the effect of  cadmium  pretreatment  on the
 nephrotoxlc  action  and  kidney  uptake  of  mercury  1n  male  and  female  rats.
Animals  were  pretreated with  either 0 or  2x2.46 mg/kg CdCK,  then  treated
with  0,   0.5,  1.0  or  1.5 mg/kg  HgCl2.    Endpolnts  assessed  were vg  Kg2*
 In  kidneys/100 g  bw,  and  severity  of tubule  damage.   No  statistics  were
given.   The authors  state  that there was  a  "significant  sex  difference
observed  1n effect  of  Cd2*  pretreatment  on   the  uptake  of  Hg2*  by  the
kidneys" and a "protective effect  of Cd pretreatment against tubular  damage
caused by mercury."
                                    C-6

-------
    Moxon and DuBois  (1939)  Investigated the Influence of  arsenic  and  other
                                                             i
elements on  the  toxlclty of  selenlferous  grains  In the  rat,;   Twelve groups
of animals  were  given  diets  containing selenlferous  wheat and 11  of  these
                                                             I
groups were  given  5  ppm of one of  the  following  elements In drinking water:
W, F, Mo,  As,  Cr,  V, Cd, Zn, Co, U, N1.   A  thirteenth group received a diet
containing  selenium-free wheat.   The   elements  In drinking water  were  not
given In  conjunction with a diet containing  selenium-free  wheat.   Endpolnts
assessed were  survival  and Se  content  In  liver.   No  statistics were given.
A further  experiment was then  conducted  since  1t appeared  from the Initial
experiment  that   "tungsten  and  arsenic,   especially  the  latter,  reduced
selenium toxlclty in  some manner."  Animals  were  fed  diets  containing either
                                                             i
selenium-free or  selenlferous wheat,  then drinking water  containing either
nothing,  2.5 ppm W  or  2.5  ppm As.   Endpolnts  assessed were survival  and
liver damage,  and again ino  statistics were given.   The  authors  concluded
that  F,  Mo, Cr,  Cd,  Zn, Co, N1 and  U given with  Se  caused an Increase In
mortality,  that  W  seemed to  reduce  the mortality rate of  rats with Se,  and
As prevented Se  poisoning symptoms, although  1t did  not prevent liver damage
appreciably.
    Thlnd and Blery  (1974)  Investigated the  antagonism of renal anglographlc
effects  of  cadmium  by  zinc  In the  dog.   In  one  group,   animals  received
Intrarenal  doses  of  cadmium acetate (2-20 mg)  and zinc  chloride  (2-20 mg).
The  animals 1n  a  second group received  a  series of arterlograms  for  the
following   exposure   sequence:   control,   vasoactlve   drug   (anglotensln,
eplnephrlne, noreplnephrlne),  cadmium  acetate +  vasoactlve drug,  vasoactlve
drug, zinc  chloride  + cadmium  acetate  + vasoactlve drug;  this sequence  was
then  repeated   In   each  dog   with  two  different   vasoactlve drugs.   No
quantitative data  were -presented;  Instead, the authors presented  the radio-
graphs from the  anglograms;.   The authors conclude that "pretreatment of  the
                                                             I
                                    C-7

-------
           for  Injected  p-xylene  and  the  LC5Q  for  Inhaled  p-xylene  were
renal  vasculature with  zinc  Ions 1n  the present  study  effectively blocked
the acute Inhibitory effects of cadmium Ions In the kidney of normal dogs."
    Drew  and Fouts  (1974)  studied the effects  of pretreatment  with  either
phenobarbltal  (PB),  3-methylcholanthrene  (3-MC),  or chlorpromazlne (CPZ), on
p-xylene  toxiclty  In  rats.   Animals  were subsequently  either  exposed  to
p-xylene  vapors  or were  Injected  with p-xylene.   The  authors  allude  to the
presence  of  control groups,  but  the  nature of  these groups  1s  not stated.
The  LDc
       3
calculated  for  the control  groups and the pretreatment groups.   This  work
was presented  In  abstract.   No data were  given,  and  the statistical methods
used were not  specified.  The authors conclude that  the pretreatments  raise
the  LCgo of  Inhaled  p-xylene,  whereas   only  3-MC  Increases  the LD5Q  of
Injected p-xylene.
    Dletz  (1980)   Investigated   the   roles of   2~butanol,   2-butanone  and
2,3-butaned1ol  In potentiating  CC14  hepatotoxlclty.   Rats were  pretreated
with  various dosages  (unspecified)  of one of these  three compounds,  'then
                                           In  addition,  some  animals   were
pretreated  with  pyrazole.   Endpolnts  assessed   were  SGPT,  glucose-6-phos-
phatase activity  and trlglycerlde  concentration.   This  work  was presented 1n
abstract  form.   No data  were  given,  and  the  statistical methods  used  were
not specified.   The authors concluded  that the  capability  of  2-butanol  and
2-butanone   to  potentiate  CC1.  hepatotoxlclty  Is  due  to  their  further
metabolism to 3-hydroxy-2-butanone and 2,3-butanedlol.
    Bhargava and  Way (1974)  examined  the  effect  of l-phenyl-3-(2-th1azolyl)-
2-th1ourea (PTT)  on morphine analgesia, tolerance and  physical  dependence In
the mouse.   Animals previously  rendered tolerant  to morphine were pretreated
with PTT  or  vehicle,  and brain  uptake of  noreplnephMne,  dopamlne,  copper,
were  administered  one  dose  of  CC1..
                                    C-8

-------
serotonin, acetylchollne, choline and morphine were  assessed,  The  statisti-
cal methods  used  were not specified.   The  authors state that the  analgesic
effect of  morphine  was  potentiated  by  PIT, but  this  effect was not  corre-
lated  with  changes  1n  brain  levels  of noreplnephrine,  dopamine,  copper,
serotonin, acetylchollne and  choline.
    Gupta and Gupta  {1977}  Investigated the effect of  the  insecticide endo-
sulfan on  pentobarbltone sleep  time and concentration of  pentobarbltone  In
blood  and  brain  In  rats.  Animals received 0,  1, 2.5 or 5  irng/kg  endosulfan
for 7  or  15  days,  then all animals received  50  mg/kg  pentobarbltone.   The
statistical methods used were not specified.  The higher  doses  of  endosulfan
(2.5  and  5  mg/kg)  were  associated  with  Increased  Induction   time  and
decreased  sleep  time.  There was no  change In  pentobarbltone concentrations
1n blood and brain.
                                                             i
    Gunn  et  al.  (1968)  studied compounds  that  potentially  were  protective
against   cadmium  toxlcity.   Mice   received  CdCl2  (doses | ranged  between
0.0055 and 0.0664 mM/kg) In  conjunction with control,  an  arolno  add (either
alanlne,   arginine,   asparaglne,  cystelne,  glyclne,   Isoleuclne,  lyslne,
methlonlne,  proline,  serlne, threonlne,  vallne,  leuclne  or phenylalanlne),
2,3-d1mercaptopropanol  (BAL),  selenium  dioxide  or  zinc  acetate.   Endpoints
were  death  at  7 days,  and  percent Cd  uptake  1n  various  organs (testes,
kidneys,  heart,   lungs,  pancreas,  spleen and  gastrointestinal  tract).   The
statistical  methods  used were not  specified.  The authors found that cadmium
destroyed  the  testis, and, while this destruction was prevented by cystelne,
that  lethality  was  increased by cystelne.   Moreover,  BAL,  selenium and zinc
also  protected  the  testis from  cadmium, but did not affect levels of kidney
cadmium  nor  toxlcity  of cadmium.  Furthermore,  none of the other amino acids
was  protective against  cadmium damage  of   the  testis or  increased  cadmium
toxicity.
                                     C-9

-------
     Jernlgan  and Harbison  (1982)  Investigated  the  role  of  2,5-hexanedlone
  (2,5-HD)  In  halocarbon hepatotoxlclty.  Mice were pretreated  with  corn oil.
  2,5-HD  or  phenobarbHal  sodium  (PB),  then  subsequently received one  of  the
following  halocarbons:    CDC13,
                                          ,   CC14.   trlchloroethylene   (TCE),
 1,1,2-trlchloroethane   (TRI),   or   perchloroethylene   (PERC).    Endpolnts
 assessed  were  hepatic  cytochrome  P-450,  NADPH  cytochrome  c  reductase,
 aniline   hydroxylatlon,   p-nltroanlsole   0-demethylatlon   and   amlnopyrlne
 N-demethylatlon, and  serum  alanlne amlno  transferase  activity.   Statistical
 comparisons were made using analysis of  variance.   Because  of  the absence  of
 a control  group for the halocarbons,  and  the authors' continual  testing  by
 comparison  back  to  the   corn   oil   pretreatment  group,  this  paper   Is
 Insufficient  to assess  the  potentlatlon  of  any  of  these  halocarbons   by
 2,5-HD.    The  authors  conclude  that   ketone  potentlatlon  of  CHCl^-lnduced
                                                                    O
 hepatotoxlclty was  demonstrated  1n mice and that  pretreatment with 2,5-HD
 can  potentiate the  hepatotoxlclty of other  halocarbons.
     Snyder et  al. (1981) studied  the effect  of  ethanol 1ngest1on on hemato-
 toxldty  of Inhaled  benzene 1n mice.  The Inhalatlon-lngestlon groups  were
 as  follows:   air  + water,  air  +  5%  ethanol,  air +•  15%  ethanol,  300 ppm
 benzene +  water,  300 ppm benzene  + 5% ethanol, 300 ppm benzene + 15% ethanol.
 Endpolnts  assessed  were  body weight  and  blood  counts.   The   statistical
 methods used were not  stated.   The authors  find  that the "results indicate a
 true potenttatlon of the toxic effects  of benzene by ethanol."
    Csallany and Ayaz (1978b) assessed the effects of N02  and  vitamin  E  in
mice.  Animals  were exposed first  to either  filtered  air or 0.5  ppm  NO  or
1.0 ppm N02,  then to 0, 30  or  300 ppm  vitamin E or 30  ppm N.N'-dlphenyl-p-
phenylenedlamlne  (DPPD).    Endpolnts   studied  were  body   weight,   tissue
weights,   llpofusdn pigment  (LFP) concentration  In  tissues,  and  survival
                                    C-10

-------
rates.   The  statistical  methods  used  were  not  specified.   The  authors
                                                              i.
conclude  that  continuous  low  level  NO   exposures do  not  result  In  higher
concentrations of  tissue  organic solvent soluble  LFP,  but N02 does  have  an
overall  detrimental  effect  on  animals,  as measured  by lowered  whole-body
                                                              |        :
weights and survival  rates.
    El-hawarl (1978) examined  the potentlatlon  of  dlbromoethane  (EOB) toxlc-
Hy by  dlsulflram (DS),  thlram, dlethyldlthlocarbamate  and  carbon dlsulflde
1n  mice.   Animals  were  pretreated  with either  DS,  thlram, 'dlethyldlthlo-
carbamate  or CS  ,  then  treated with  EOB.   Only  one   dose  of   the  various
                                                              -i
pretreatment compounds was administered.  It was  unclear what control groups
were  present.  Endpolnts assessed were  SGPT, SGOT,  blood urea nitrogen (BUN)
levels  and  survival.   The  report  Is  In  abstract  form.   No  statistical
                                                              j
methods  are stated.   The authors  conclude that  pretreatment  with  any  of
these compounds enhances EDO toxldty.
    Agarwal  et  al. (1983) studied  the  Interactions of  CBr4  and  chlordecone
(CD)  1n  rats.  Animals were  fed diets  containing  either 0 or 10 ppm CD, then
                                                              !
were  Injected  with  either   0, 25,  50,  100  or  125   vs.  CBr4«   Endpolnts
assessed   were   urine   parameters,   Including  volume,   osmolallty,   blood,
protein,  glucose;  p-amlnohlppurate  (PAH) and  tetraethylammonlum (TEA) levels
1n  the renal  cortex;  SGOT  and SGPT  levels.   Since  the authors  felt  that
CBr.  was acting  like  a  nephrotoxln,  a second experiment was  undertaken  1n
which animals  were fed  diets  containing either  0 or  10 ppm CD,  then were
                                                              i
administered  either  vehicle  or  54  yS. CC14  or   75,   125  or  175 mg  CBr4.
Endpolnts  measured were  PAH  and TEA  levels  In  renal  cortical  slices.   In
both   experiments,  statistical  methods  were   not   specified,  although
comparisons  were  made  to  control  groups.   The  authors  concluded  that
chlordecone  did  not  modify  rerial  slice  response,  and that  CD does  not
potentiate  CBr4 hepatotoxlclty.                               i

                                    C-ll                      I

-------
     Berlin  and Lewander  (1965)  Investigated  the effect  of 2,3-dlmercapto-
 propanol  (BAL)  on brain  uptake  of mercury 1n  mice  given mercuric chloride.
 In  the  acute experiment, animals  were  given  0.5 mg/kg Hg,  then  either  0 or
 0.3 mg/kg BAL.   Endpolnts  measured  were  tissue  Hg concentration.   In  the
 chronic  experiment,  animals  were   given  1 mg/kg 203HgCl   and  either 0  or
 2  mg/kg  bw  BAL  for   16  days.    Again,  the  endpolnts  assessed  were  Hg
 concentrations In tissue.  There were no negative control  groups  (no BAL, no
 mercuric  chloride).  The statistical methods  used were not  specified.   The
 authors concluded that  BAL does not affect  Hg  elimination.
     deFerreyra et al.  (1983)  assessed  the potentlatlon  of CC1.  necrosis  by
 cystelne and tryptophan,  both  alone and together, In the  rat.   Experimental
 groups  were  as  follows:   control,  CC14,  cystelne,  tryptophan,  cystelne  +•
 CCl^,   tryptophan  +  CC14,  and  cystelne   +  tryptophan  +  CC1..    Endpolnts
 measured were ICO levels and degree of  liver necrosis.  Although  the  authors
 state  that a 2-way analysis of variance was  used, H 1s unclear  what  groups
 were compared.   All  other  comparisons  were   to   the  control  group, using
 unspecified   statistical  methods.   This  experiment  Is  missing  one  experi-
 mental  group  (cystelne  + tryptophan);  otherwise a 3-way analysis  of variance
 would  have been the  correct  procedure.   The authors  conclude that adminis-
 tration  of cystelne but not  tryptophan decreased ICD, and when both cystelne
 and  tryptophan were given  together,  a  "marked protective effect Is observed."
     Agarwal  and Mehendale  (1984)  studied  the  potentlatlon  of  CC1   hepato-
 toxlclty  by   chlordecone  (CD)   In ovarlectomlzed rats.  Animals  were either
 sham operated or  ovarlectomlzed, then fed diets  containing either 0. or  10
 ppm  CD,   then  received   either  25  ^  CC14 or  vehicle.    Endpolnts  measured
were  S6PT,  SGOT,   Isodtrlc   dehydrogenase (ICD)  and   ornlthlne  carbamyl
transferase  (OCT)  activity.   The authors  used  student's t-test  and one-way
                                    C-12

-------
analysis  of  variance,  making   comparisons   to   controls,  : although  3-way
analysis of  variance  would have  been  the correct method  to Investigate  the
                                                             I
Interplay  of these  compounds.   They  also  measured  biliary  excretion  of
phenolphthaleln  glucuronlde  (PG)  over  time, which would  have  been correctly
analyzed by  a  repeated  measures analysts.   The  authors  conclude  that  "CD
Induced  potentlatlon  of  CC14  hepatotoxlclty  In ovarlectomlzed rats  was  not
significantly   enhanced  as   compared   to  earlier  observations   In  Intact
females."
    Klnnamon and Bunce (1965)  examined  the  effects of copper,  molybdenum and
zinc  on 6SZn   tissue  distribution  and  excretion  In  the  rat.   There  were
eight experimental  groups  consisting of  the  combinations of 0 or 100 mg/kg
                                                                           .
Cu, 0  or 800 mg/kg Ho  and 0  or  5000  mg/kg Zn  In feed.   Eridpolnts assessed
                                                             I
were  body  weight,  weight  gain,  feed  consumption,  percent  Zn retention  In
tissues  and  percent   Zn  excretion In   urine.    Comparisons   were  made  to
                                                             i
controls,  using t-tests,  although  a 3-way analysis  of  variance  would  have
been  the correct  procedure.   The authors concluded  that  "Zn,  not Mo or  Cu,
                                                             I
significantly Influences tissue distribution and excretion of tracer Zn."
    Jaeger  and  Murphy   (1973)  studied the  effects  of 1,1-dlchloroethylene
(1,1-DCE),  cortlcosterone  or  acroleln   on  barbltuate  action  1n  the  rat.
Animals  were  pretreated  with either 0 or  400  mg/kg 1,1-DCE,! then were given
either  pentobarbltal   (PB) or  hexobarbltal  (HB).   Endpolnts  assessed  were
sleep  time,  barbltuate  concentration In  serum  and brain,  and  serum cortlco-
                                                             i
sterone  concentration.    A  second  and  third  experiment  were conducted  1n
which  pretreatment  was   either  0  or   25  mg/kg   cortlcosterone  or  3 mg/kg
acroleln,  assessing  the  same endpolnts.   Statistical  techniques  employed
Included t-test,  analysis  of variance and  regression,  although It 1s Impos-
sible  to tell  which  technique (t-test or  analysis  of  variance)  was used to
make  certain comparisons.   There was no  negative  (untreated)  control 1n any
                            .
                                     C-13                     !

-------
 experiment.   The  authors  concluded  that  both 1,1-DCE  and  cortlcosterone
 alter  PB-1nduced,  but  not  HB-lnduced,   sleep   time,   and   that   acroleln
 Increases both PB and HB sleep time.
     Hasumura  et  al.   (1974)  Investigated  the  effect  of  chronic  ethanol
 consumption on  CC14 hepatotoxlclty  In  the rat.   Experimental animals  were
 fed diets consisting of ethanol (36% of total  calories),  and  control  animals
 were pair-fed  diets In  which ethanol  had  been  Isocalorlcally  replaced  by
 carbohydrate.    Animals   then  received  either  0.5  mil/kg  CC1.   or  mineral
 oil.   Endpolnts  measured  were serum  ornlthlne carbamyl  transferase  (SOCT),
 S6PT,  blUrubln,  total   llpld,  cytochrome  P-450,  amlnopyrine N-demethylase
 activity,  and  glucose-6-phosphatase  activity.  Paired  Student's  t-test was
 used  to compare  ethanol  to pair-fed control,  whereas  a  randomized complete
 block  design would have been  the  correct  procedure to make this comparison,
 allowing  for  multiple  comparisons.   The  authors  concluded  that  "chronic
 ethanol  administration  to  rats  potentiates  CC1.  hepatotoxlclty,"  although
 they did  not use  methods  that  would allow  for  the assessment of Interactions.
    Harbison and  Becker  (1971) examined the effect  of  treatment  with pheno-
 barbltal  (PB)  or  SKF 525A on  dlphenylhydantoln (DPH) disposition  on pregnant
 mice.   All  animals  received  100 mg/kg  DPH, after  pretreatment.. with  either
 control,  PB  or SKF 525A.   Endpolnts measured  were  DPH  metabolism In plasma,
 placenta,  fetus,  amnlotlc  fluid,  liver,  brain, fat,  and muscle  over time.
 They used Student's  t-test  to  compare  each pretreatment group with DPH alone
 at  each  time  point,  although  a repeated  measures analysis would  have  been
 the  correct procedure.   They  concluded that  pretreatment with  PB  enhanced
 the  metabolism of  DPH,  with  decreased  plasma  DPH  and  DPH-1nduced  terato-
 genldty  and in  utero  deaths, while  pretreatment  with  SKF  525A  decreased
metabolism of  DPH,  with Increased plasma  DPH  and  DPH-lnduced  teratogenlclty
and 1n utero deaths.
                                    C-14

-------
    Csallany and  Ayaz (1978a)  Investigated  the effects of  Intermittent  N02
exposure and vitamin  E In  rats.   Animals  were fed either vitamin E deficient
(0  ppm),  normal  (30  ppm),  or  high  (300  ppm) diets,  then were  exposed  to
either  air  or  15  ppm  NO   for  either 5  or  18  weeks.   Endpoints  assessed
Included   methemoglobin    levels,   histopathology,    Upofuscin   pigment
concentration in  tissue and  fatty acid component  in  lung  tissue.   Student's
t-test was used to make  comparisons  between the  treatment  groups,  although
It  was  not  always  clear which  groups  were being  compared.   Analysis  of
variance  techniques would  have  been  correct.   The  authors concluded  that
"Intermittent NO- exposure,  under  the described  conditions,  did not  cause
ultimate changes of the biochemical parameters measured."    ;
    Derr et al.  (1970)  examined the synergism between  cobalt and  ethanol  on
rat. growth rate.   Water,  allowed  ad  libitum, was replaced with, either 0  or
10% ethanol,  and either  0  or 1  mg Co/10 ma. H20.   Endpoints measured  were
body  weight,  hematocrit,  heart  weight, heart  Zn, and  heart-to-body  weight
ratio.  Student's  t-test  was used  to compare  the  various  groups, and  an
additive model was  used to calculate  an expected  body weight for the ethanol
                                                             i
•H Co  group,  which was then  compared  with the observed  body weight  for  that
group by  Student's  t-test.   A  two-way  analysis  of variance would have  been
the correct  procedure.   The  additive  model   that  was used added  the  weight
deviations from  the control  in  order  to predict  the  weight  deviation  of  the
two chemicals combined.  The  authors did  not provide  any biological  justifi-
caton for  such a  model.   Even a  simple  method using relative potencies  would
be  better  justified.   The authors'  conclusion was  that ethanol  and  cobalt
have a synerglstic effect.
                                    C-15

-------

-------
                                  APPENDIX D

                                  REFERENCES

                                                             i

Agarwal, A.K.  and  H.M. Mehendale.   1984.   Chlordecone potentlation of  CC14

hepatotoxlclty In ovarlectomlzed rats.   Toxicology.   29:  315-323.


Agarwal, A.K.,  W.O.  Berndt  and H.M. Mehendale.  1983.   Possible  nephrotoxlc

effect  of   carbon   tetrabromlde  and  Us   Interaction  with   chlordecone.

Toxlcol. Lett.  17:  57-62.

                                                             i
Albert, R.E.,  J.  Lewtas,  S. Nesnow,  T.W.  Thorslund and E.  Anderson.   1983.

Comparative  potency method  for  cancer  risk  assessment; application  to  dlesel

participate emission.  Risk Anal.  3: 101-117.

                                                             I
Andrews,  L.S., E.W.  Lee,  C.M.  Wltmer,  J.3.  Kocsls  and  R. Snyder.   1977.

Effects of  toluene  on  the  metabolism, disposition and  hemapoletlc  toxldty

of [3H]benzene.  Blochem.  Pharmacol.  26:  293-300.
                                                             i
                                                             I
Berlin, M. and T. Lewander.  1965.   Increased brain uptake  of mercury  caused

by  2,3-dimercaptopropanol  (BAL)  In  mice   given   mercuric  chloride.    Acta

Pharmacol. Toxicol.   22:  1-7.


Bhargava,  H.N. and   E.L.  Way.   1974.   Effect of l-phenyl-3-(2-thiazolyl)-2-

thiourea,  a  dopamine (3-hydroxylase  inhibitor,  on  morphine  analgesia,  toler-

ance  and physical dependence.   3. Pharmacol.  Exp. Ther.  190;: 165-175.
                                    D-l

-------
B1anc1f1or1,  C., F.  Caschera, F.E.  Giornelll-Santmi  and  E.  BucciarelH.
1967.   The action  of  oestrone and  four  chemical carcinogens  In  Intact and
ovarlectomlzed BALB/C/Cb/Se mice.  Br. 3. Cancer.  21: 452-459.

Carlson,  6.P.   1973.   Effects  of  phenobarbltal  and  3-methylcholanthrene
pretreatment  on  the  hepatotoxldty  of 1,1,l-tr1chloroethane  and  I,l,2-tr1~
chloroethane.  Life Sci.  13:  67-73.

CAS.   1980.   Chemical  Abstracts  Service  Source Index  1907-1979.   Chemical
Abstracts Service, Columbus, OH.

Casto,  B.C.,  G.6.  Hatch, S.L.  Huang,  3.  Lewtas, S.  Nesnow  and  M.D.  Waters.
1981.   Mutagenlc and  carcinogenic potency of extracts of  dlesel and  related
environmental emissions;  in y.l.trq mutagenesls  and  oncogenlc transformation.
Environ. Int.  5: 403-409.

Castro, 3.A., E.C.  de  Ferreyra,  C.R.  de  Castro, O.M.  de  Fenos, H.  Sasame and
3.R.  Gillette.   1974.   Prevention  of carbon  tetrachlorlde-lnduced  necrosis
by  Inhibitors  of drug  metabolism —  Further  studies ;on  their  mechanism of
action.  Blochem. Pharmacol.  23: 295-302.

Claxton,  L.D.    1981.   Mutagenlc  and carcinogenic  potency  of  extracts  of
dlesel  and  related environmental  emissions;  Salmonella bloassay.   Environ.
Int.  5: 389-391.
Cone, H.V. and  P.  Netteshelm.   1973.  Effects of  vitamin  A  on 3-methylchol-
anthrene-lnduced squamous  metaplasias and  early  tumors  1n  the  respiratory
tracts of rats.  3. Natl. Cancer Inst.  50:  1599-1606.
                                    D-2

-------
CsalTany, C.S. and  K.L.  Ayaz.   1978a.  The effects of  Intermittent  nitrogen


dioxide  exposure on  vitamin  E-def1c1ent  and  -sufficient  rats.   Toxlcol.


Lett.  2: 97-107.





Csallany, C.S.  and  K.L.  Ayaz.  1978b.   Long-term N0?  exposure  of mice  In


the  presence and  absence  of  vitamin  E.   I.  Effect  on  body  weights  and


llpofusdn In pigments.   Arch.  Environ.  Health.  N/D:  285-291.
                                                            i




Curren,  R.D.,  R.E.  Dourl,  C.M.  K1m  and  L.M.  Schechtman.  1981.   Mutagenlc
                                                            i

and  carcinogenic potency  of  extracts  of dlesel  and related  environmental


emissions;  simultaneous  morphological   transformation  and mutagenesls   1n
                                                            i

BALB/c 3T3 cells.  Environ. Int.   5: 411-415.





Daoud, A.H.  and A.C.  Griffin.  1980.   Effect  of  retlnolc add,  butylated


hydroxytoluene,  selenium and  sorblc  acid  on  azo-dye  hepatocarc1nogenes1s.


Cancer Lett.  9: 299-304.                                   :



                                                            i

de Ferreyra, E.C.,  O.M.  de Fenos and J.A. Castro.  1983.   Tryptophan  poten-


tlation  of  the  late  cystelne  preventive effects  In carbon  tetrachlorlde-


Induced necrosis.  Res.  Comm. Chem.  Pathol.  Pharmacol.  40:  515-518.





Derr,  R.F.»  H.  Aaker,  C.S.  Alexander and  H.T.  Nagasawa.  1970.   Synerglsm


between cobalt and ethanoll on rat growth rate.   J.  Nutr.   100:  521-552.



                                                            i

D1etz, F.K.   1980.  The .role  of  2-butanol  and  2-butanone metabolism  1n  the


potentlatlon of  carbon  tetrachlorlde  Induced hepatotoxlclty.   D1s.  Abstracts


Int. Part B (Biology).  41: 150.                             j
                                    D-3

-------
 Drew,  R.T.  and J.R.  Fouts.  1974.  The effects of Inducers on acute p-xylene
 toxldty.   Toxlcol. Appl.  Pharmacol.   29: 111-112.

 El-hawar1,  A.M.   1978.   Potentlatioh of  dlbromoethane  (EDB)   toxlclty  by
 dlsulflram,  thlram,  dlethyldlthlocarbamate and  carbon  dlsulfide.   Pharma-
 cology.  20: 213.

 Friedman,   H.A.   and   L.R.  Eaton.    1978.    Potentlatlon  of  methylmercury
 toxldty by plperonyl butoxlde.  Bull. Environ. Contam. Toxlcol.   20: 9-16.

 Gunn,  S.A., T.C.  Gould and  W.A.D.  Anderson.   1968.   Selectivity  of  organ
 response  to cadmium  Injury  and various   protective  measures.   J.  Pathol.
 BacteMol.  96: 89-96.

 Gupta, P.K.  and  R.C.  Gupta.   1977.   Influence of endosulfan  on  pentobarbl-
 tone  sleeping  time  and blood  and   brain  concentrations  In  male rats.   J.
 Pharm. Pharmacol.  29: 245-246.

 Harbison, R.D.  and B.A. Becker.   1971.  Effects  of phenobarbltal  or  SKF  525A
 pre'treatment on  dlphenylhydantoln  disposition  In  pregnant mice.   Toxlcol.
Appl. Pharmacol.   20:  573-581.

 Hasumura, Y;,  R.  Teschke  and  C.S.  Lleber.   1974.   Increased carbon  tetra-
 chlorlde hepatotoxlclty and Its mechanism after  chronic  ethanol consumption.
Gastroenterology.  66: 415-422.   ,
                                    D-4

-------
Ito, N.,  H.  Nagasaki,  M. Aral, S.  Makulra,  S.  Suglhara  and K.  Hlrao.  1973.
Hlstopathologlc  studies  on liver  tumor 1 genes Is Induced  In  mice  by technical
polychlorlnated  blphenyls  and Its  promoting effect on  liver  tumors  Induced
by benzene hexachlorlde.  J.:Natl. Cancer Inst.  51: 1637-1646.

Jaeger,  R.J.  and  S.D.  Murphy.   1973.   Alterations  of   barbltuate  action
following  1,1-dlchloroethylene,  cortlcosterone  or  acroleln.    Arch.   Int.
                                                              i
Pharm. Ther.  205: 281-292.
                                                              I

Jernlgan, J.D.  and  R.D. Harbison.   1982.   Role of blotransformatlon In  the
potentlatlon  of halocarbon hepatotoxlclty  by 2,5-hexaned1one.   J.  Toxlcol.
Environ. Health.  9: 761-781.
                                                              i
Klnnamon, K.E.  and  G.E.  Bunce.   1965.   Effects  of  copper, molybdenum  and
zinc on zlnc-65  tissue distribution  and  excretion  In  the rat.   J. Nutr.   86:
225-230.

Lewtas,  0.   1983.   Evaluation of  the mutagenlcHy  and carclnogenlcHy  In
motor  vehicle  emissions  1n short-term bloassays.   Environ.  Health  Perspect.
47: 141-152.         ,                                         j  •

Lewtas, J.,  R.L. Bradow, R.H.  Jungers, et al.   1981.   Mutagenic  and carcino-
genic  potency  of  extracts  of  dlesel  and  related  environmental  emissions;
study  design,   sample,   generation,  collection  and  preparation.   Environ.
Int.  5: 383-387.
                                    D-5

-------
Hagos, L., M.  Webb  and W.H. Butler.  1974.  The  effect  of  cadmium pretreat-
ment  on  the  nephrotoxlc  action and  kidney uptake  of mercury  In male  and
female rats.  Br. J. Exp.  Pathol.  55: 589-594.

Mitchell, A.D., E.L. Evans, H.M. Jotz,  E.S.  Rlcdo,  K.E.  Mortelmans  and  V.F.
Simmon.  1981.  Mutagenlc and carcinogenic potency of  extracts  of  dlesel  and
related  environmental   emissions;   jn_  vitro  mutagenesls  and   DNA  damage.
Environ. Int.  5: 393-401.

Hoxon, A.L.  and K.P.  DuBols.   1939.   The  Influence of arsenic and  certain
other  elements  on   the  toxlclty  of   selenlferous   grains.   J.  Nutr.   18:
447-457.

Nesnow,  S.,  L.L.   Trlplett  and  T.J.   Slaga.   1982.   Comparative  tumor-
Initiating  activity  of  complex  mixtures  from  environmental  partlculate
emissions on SENCAR mouse skin.   J. Nat. Cancer Inst.  68: 829-834.

Schlede,  E.,  R.  Kuntzman  and  A.M.  Conney.   1970.    Stimulatory  effect  of
benzo(a)pyrene  and  phenobarbltal  pretreatment  on the  biliary  excretion  of
benzo(a)pyrene metabolites In the rat.  Cancer  Res.   30:  2898-2904.

Short,  R.D.,  G.M.  Winston, G.L.  Minor,  C.  Hong,   J.  Selfter  and C.  Lee.
1977.  Toxlclty  of  vlnylldlne  chloride  1n  mice and  rats and Us  alteration
by various treatments.   J. Toxlcol. Environ.  Health.   3:  913-921.
Snyder, C.A.,  K.A.  Baarson, B.D. Goldstein  and  R.E. Albert.   1981.   Inges-
tlon  of  ethanol  Increases  the  hematotoxlclty  of  Inhaled  benzene In  C57B1
mice.  Bull. Environ. Contam. Toxlcol.  27: 175-180.

                                    D-6

-------
Thlnd,  G.S.  and D.N. B1ery.   1974.   Antagonism of  renal  angliographlc effect
of cadmium by zinc.  Invest.  Radio!.   9:  386-395.

U.S. EPA.   1979.   The  Diesel Emissions Research  Programs.  EPA-625/9-74-004.
Center  for  Environmental  Research  Information,  Cincinnati,  OH.   (Cited  In
Lewtas et a!.,  1981)

U.S.  EPA.   1986.   Guidelines  for  the  Health  Risk Assessment  of  Chemical
                                                              I
Mixtures.  Federal Register.   51(185): 34014-34025.

U.S.. EPA.   1988.   MIXTOX.   Studies  on toxlclty  of  mixtures and  Interacting
chemicals.    User's  Guide.   Office of  Health  and  Environmental  Assessment,
Environmental Criteria and Assessment Office,  Cincinnati,  OH.
                     **U.S. GOVIiRNMJENT PRINTING OFFICE: 19 9 0- 7 * 8 - 1 5 yi 0 * >t 5
                                    D-7

-------

-------