May 1993
                      WORKSHOP REPORT ON


                      EXPOSURE TO PCBs
                        September 14-15,1992
                 Research Triangle Park, North Carolina
                           Prepared by:

                     Eastern Research Group, Inc.
                        110 Hartwell Avenue
                        Lexington, MA 02173
                       Risk Assessment Forum
                  U.S. Environmental Protection Agency
                                                Printed on Recycled Paper

       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 recommendation for use.


Acknowledgments	 vi

Preface	 . . . .  . vii

       Carole A. Kimmel

2.     ISSUE PAPERS  	  2-1

       General Toxicity of PCBs	  2-1
       Carole Braverman

       Pharmacokinetics of PCBs	 .	2-7
       Linda S. Bimbaum

       PCB Structure-Activity Relationships and Developmental Toxicity	  2-11
       James McKinney

       PCB Developmental Neurotoxicity in Experimental Animals	  2-14
       Diane B. Miller

       Developmental Neurotoxicity of PCBs in Humans	  2-20
       Mary M. Prince


       3tl    Introduction	  3-1
       3.2    General Toxicity/Exposure	  3-3
             Michael Bolger
       3.3    Pharmacokinetics	  3-6
             Linda S. Bimbaum
       3.4    Structure-Activity Relationships	  3-12
             James McKinney
       3.5    Risk Assessment for Developmental Toxicity	  3-16
             Gary L. Kimmel

      Elaine Z. Francis, Chairperson

       4.1    Work Group Summaries  	  4-1
      4.2    General Discussion Summary	  4-1
      4.3    Panel Discussion Summary	  4-5

      Hugh A. Tilson, Chairperson

      5.1    Work Group Summaries	 5-1
      5.2    Panel Discussion Summary	 5-1
      5.3    General Discussion Summary	5-4


      6.1    Sufficiency of the Available Data for Risk Assessment .	 6-1

             Summary of Human Work Group Deliberations 	 6-1
                Mary M. Prince, Co-chaitperson
             Summary of Animal Work Group Deliberations		  6-14
                Carole A. Kimmel, Co-chairperson
             Comparison of Human and Animal Data	,	  6-21
                Carole A. Kimmel, Co-chairperson

      6.2    Research  Needs to Support Risk Assessment	  6-23

             Summary of Human Work Group Deliberations	  6-23
                Jane Adams, Co-chairperson
             Summary of Animal Work Group Deliberations 	  6-26
                Diane B. Miller, Co-chaitperson


             Jane  Adams	 A-l
             Donald G. Barnes  	A-4
             David Bellinger 	A-8
             Robert E. Bowman	A-12
             John  F. Brown, Jr	A-19
             Theo Colborn	A-36
             Eric Dewailly '.	A-41
             Kim N. Dietrich	A-44
             Beth  Gladen	A-52
             L. Earl Gray, Jr	A-68
             G. Jean Harry 	A-72
             Joseph L. Jacobson and Sandra W. Jacobson	A-75
             James McKinney  	A-88
             Diane B. Miller 	A-89
             John  A. Moore	A-93
             Richard E. Peterson	A-120
             Mary M. Prince 	A-126
             Walter Rogan	A-128

             Susan L. Schantz	A-131
             Richard Seegal	A-135
             Anne Sweeney 	A-140




             Comments	D-l
             Richard Seegal

             Neonatal PCB Exposure and Neurodevelopmental Deficit:
             An Outline of the Planned/Ongoing Dutch-German Cohort Study
             on the Validation of Early Indicators
             of Adverse Developmental Effects	V.. .". '.".'. '.	D-2
             Gerhard Winneke, E. Rudy Boersma, and Pieter J. Sauer

       The U.S. Environmental Protection Agency's (EPA's) Risk Assessment Forum
appreciates the time and contributions of all workshop participants.  Carole Kimmel of EPA's
Reproductive and Developmental Toxicology Branch (RDTB) coordinated EPA's workshop and
document development activities with the assistance of EPA's Risk Assessment Forum staff.

       Linda Birnbaum of EPA's Health Effects Research Laboratory (HERL) and Carole
Kimmel served as overall chairs for the workshop.  Linda Birnbaum, Carole Kimmel, Michael
Bolger (U.S. Food and Drug Administration), Carole Braverman (EPA Region 5), Elaine Francis
(EPA/Office of Research and Development), Gary Kimmel (EPA/RDTB), and Hugh Tilson
(EPA/HERL) served on the workshop technical panel.                          ,

       Linda Birnbaum presented the introduction  to the workshop; Carole Kimmel prepared
the workshop chair's summary; Jane Adams  (University of Massachusetts), Carole Kimmel,
Diane Miller (EPA/HERL), and Mary Prince (National Institute of Occupational Safety and
Health) prepared individual work group chairs' summaries. Eastern Research Group, Inc.
(ERG), an EPA contractor, provided organizational support and prepared this report.

       On September 14 and 15,1992, the U.S. Environmental Protection Agency's (EPA's)
Risk Assessment Forum sponsored a workshop on the developmental neurotoxic effects of
polychlorinated biphenyls (PCBs) (57 Federal Register 39200; 28 August 1992).  The meeting was
held in Research Triangle Park, NC, and was chaired by Linda Birnbaum and Carole Kimmel of
the Environmental Protection Agency. Participants from academia, industry, and state and
federal government brought expertise from a wide range of disciplines to the discussion.
Members of the public and EPA scientific staff attended the workshop as observers.

       This report collects workshop papers and discussion on principles and methods for
evaluating data from animal and human studies; The report also summarizes data and other
information discussed at the workshop for characterizing risk to human development, growth,
survival, and function following exposure to PCBs prenatally or to infants and children. EPA
compiled several issue papers (see Section 2) on various aspects of PCB toxicity and especially
on developmental neurotoxicity, as a framework for workshop discussion." -These issue papers
were distributed to all invited participants,  who then submitted pre-meeting comments (see
Appendix  A).           ,                  ,   '..                                •.'..'

       As outlined in the issue papers, the purpose of this workshop was to arrive at a general
"sense of the meeting" regarding the current state of the science on neurotoxic effects associated
with prenatal and perinatal exposure to PCBs.  EPA did not expect participants to reach a
common position on  all of the issues before the group. Because PCBs are present in air, water,
and food,  information developed at the workshop will assist the Agency in evaluating the effects
of PCBs in these media and will serve as a basis for protecting public health from PCBs
occurring in these media.

       Recent studies defined the issues selected for workshop analysis. Data from rodents and
monkeys have demonstrated that prenatal and perinatal PCB exposure results in neurotoxicity in
the offspring. Related effects have been reported in human studies. For example, human
poisonings (Yusho and Yucheng) have led  to developmental delays and impairment in
neurobehavioral indices in offspring of exposed women. Also, relatively low levels of exposure to
PCBs in cohorts in Michigan and North Carolina have suggested neurobehavioral deficits  in
infants and young children.  Thus, the public health consequences of exposure to developmental
neurotoxicants such as PCBs are potentially significant.

       These observations pose several questions regarding the use of PCB data for assessing
risk of neurotoxic effects because of prenatal or perinatal exposure:

       •     Are all PCBs alike in these effects or, if not, are any useful structure/activity
             relationships discernable?

       •     What  are the dose/response relationships?

       •     Are there populations at special risk due to elevated exposure or to inherent

       •     What  are the endpoints of greatest concern and of greatest sensitivity?

These questions are important because of the persistence of PCBs in environmental media such
as water and air, and the nature of the data available on PCBs and developmental neurotoxicity;
i.e., many studies are available on various mixtures but little or no information is available
regarding specific congener effects on the developing organism,  or the mechanism of action
of PCBs.

       This workshop report presents the chair's overall workshop summary; the issue papers;
summaries of the opening presentations; reports of the two work groups, including their
conclusions on human and animal developmental neurotoxicity data; and summaries of the two
panel discussions, one on information sufficient for risk assessment and the other on research
needs. The workshop agenda, list of panel members, participants  and observers, and pre-meeting
comments are provided in appendices.

       Some sections of the workshop chairs' summaries were edited for clarity; contributors
were not asked to follow a single format.  Relevant sections were reviewed by each workshop
chair and speaker.

       The views presented are those of each contributor, not the U.S. Environmental
Protection Agency.

                                    SECTION ONE

                      CHAIR'S OVERVIEW AND SUMMARY

                                     Carole A. Kimmel
                            V.S. Environmental Protection Agency
       This section summarizes the outcome of the PCB Developmental Neurotoxicity
Workshop in relation to the stated goals of the workshop. The first goal was to address the
question of whether currently available  data are sufficient for risk assessment.  The sense of the
meeting seemed to be that, at least in qualitative terms, the available data are sufficient.  In
other words, based on an evaluation of the strengths and weaknesses in the data and on the
consistency of effects seen in all species tested, including humans, there is sufficient information
to indicate that PCBs cause developmental neurotoxicity.  Interestingly, the data suggest that
prenatal exposure to PCBs may be more detrimental than postnatal exposure, even though the
level of exposure via breast milk is much greater than that occurring via placental  transfer.

       Quantitatively, the sufficiency of the available data was less certain,  although most
participants seemed to feel that the data derived using commercial mixtures could be applied to
risk assessment for commercial mixtures.  There was less agreement on the applicability of these
data to environmental exposures.  Since the effects of exposure to various commercial mixtures
appear to be very consistent in laboratory animals, and since these effects are also consistent with
those observed in humans exposed to environmental mixtures, most participants felt that data
derived using commercial mixtures should also be used to set limits on environmental exposure
to PCBs.

       The second goal of the workshop was to identify any additional information that might be
used to support the developmental neurotoxicity data in risk assessment.  Based on discussions at
the workshop, at least three types of data that could be helpful in this regard appear to
be available:

              (1)  information on the mix of congeners present in breast milk and/or other
       human tissues;

              (2)  data from fish feeding studies, wildlife studies, and other environmental
       exposure situations; and

              (3)  data on other types  of toxicity that might be relevant to  interpretation of the
       developmental data, particularly the neuroendocrine  effects of these compounds'.

       The third goal of the workshop  was to recommend research needed to address risk
assessment issues.  The overwhelming sense of the meeting was that there is a need for more
research on the developmental effects of PCBs. In particular, currently available data need to be
replicated in additional human studies,  monkey studies, and rodent studies. There is also a need
for comparative studies to determine whether the effects of various mixtures of congeners are
similar or different. These, in turn, should be linked with studies assessing the effects of

 environmentally relevant mixtures.  Mechanistic studies of PCBs are needed on a congener-
 specific basis, and there is also a need for studies designed to use comparable testing methods
 across multiple species.

        Various types of research approaches that might be pursued were also discussed at length
 in the workshop.  The sense of the meeting was that characterization of mixtures and of specific
 congeners are both important, and several approaches to this problem were discussed.  A very
 significant need is to characterize developmental exposures (i.e., what reaches the developing
 embryo/fetus/neonate) in relation to the mix of congeners present in the original commercial or
 environmental source. In addition,  rather than trying to tackle all 209 possible congeners, there
 was general agreement that it might make sense to first evaluate the approximately 30-35
 congeners that commonly occur in the environment, and to compare these results with the
 currently available data.  In addition, an SAR approach such as that described by Dr. McKinney
 could be taken to look at specific types of activity—e.g., in a prealbumin binding assay.

       A great deal of information is  needed on the mechanisms of action of the PCBs.
 Another approach could be to categorize various congeners on the basis of their chemical
 reactivities—e.g., dioxin-like, thyroid activity, dopaminergic, estrogenic, andrpgenic, and retinoid
 interaction—since all PCB effects do not appear to be linked to any single type of chemical
 interaction.  These reactivities could be evaluated initially in in vitro systems.  Although studies to
 investigate the role of these reactivities were considered important, there was also a sense that
 these investigations should be clearly linked to observed functional effects. Similarly, SAR
 approaches should also be used to the extent that they can be linked to functional effects.

       Further characterization of PCB effects" is also needed.  In this regard, more sensitive
 tests are needed to clarify the specific  types of effects occurring in various functional domains.
 Tin's information also can be extremely useful in studying mechanisms of toxicity. It was also
 suggested that greater consistency in the administration of tests—both within species and among
 species—is essential to comparing and interpreting data from different laboratories.  Finally, it is
 especially important to identify subpopulations that may be especially susceptible to PCB effects.
The current data suggest that the developing organism is extremely sensitive, especially during
 the prenatal period. The  sensitivity of developmental effects to various conditions of exposure
needs to be further characterized, however, in part by using tests- that are capable of detecting
changes in age-specific developmental processes.                    '.              '

                                   SECTION TWO

                                   ISSUE PAPERS

                              PRE-MEETING ISSUE PAPER

                                   WORKSHOP TOPIC

                             GENERAL TOXICITY OF PCBs

                          Prepared by Dr.  Carole Braverman, EPA


1.     To review general principles of polychlorinated biphenyl (PCB) toxicity.

2.     To define, where possible, the dose levels associated with nondevelopmental endpoints.


Although the purpose of this workshop is to discuss the role of PCBs as developmental toxicants,
PCB exposure is associated with a wide spectrum of other effects including hepatotoxicity,
neurotoxicity,  immunotoxicity and carcinogenicity.  A discussion of these other endpoints is
pertinent in that it allows for a comparison of the dose levels producing developmental endpoints
with the dose levels associated with other sensitive effects.

The relative scarcity of congener specific data complicates attempts to succinctly define PCB
toxicity.  PCBs were produced commercially  as complex mixtures of congeners, distinguished by
percent chlorine content and sold under trade names including Aroclor, Clophen, and Kanechlor.
Commercial mixtures vary in congener content from batch to batch and contain varying amounts
of contaminants including polychlorinated dibenzofurans. The poorly defined nature of the
commercial mixtures, weathering of PCB mixtures in the environment, differential potential of
the congeners to bioaccumulate and the evidence that individual congeners vary greatly in their
toxicity all point to the need for congener specific data in addition to mixture specific data.

Congener toxicity has been shown to depend  on both the degree of chlorination and the spatial
orientation of the chlorine atoms on the biphenyl ring.  It is generally agreed that for certain
endpoints, coplanar congeners are of the greatest toxicological significance, and that this toxicity
is mediated through a 2,3,7,8 tetrachlorodibenzo-p-dioxin-like interaction  with the aryl
hydrocarbon receptor. A toxic equivalency factor approach to the TCDD-like endpoints has been
proposed for the coplanar PCBs (Safe, 1990). For other endpoints the relationship between
toxicity and coplanar conformation is less clear.

Animal studies provide the basis for quantitative PCB risk assessment and species sensitivity
varies widely.  Currently available epidemiological studies assessing the impact of two poisoning
incidents in Japan and Taiwan (identified as the Yusho and Yu-cheng incidents, respectively), and
occupational exposure studies are inadequate for quantitative risk assessment but can provide
qualitative information.


Acneform dermatitis (chloracne) and hepatic dysfunction are associated with PCB exposure in
occupationally exposed populations (Kimbrough et al., 1987).  Increased mortality, skin lesions,
hepatotoxicity and weight loss are seen in animal studies of  PCB exposure. Skin lesions and
weight loss were seen in female rhesus monkeys ingesting diets containing 2.5 or 5.0 ppm
Aroclor 1248 in chronic studies (Barsotti et al., 1976).  Hepatic fluorescence, an indication of
porphyria, was detected in rats exposed to 30 ppm Aroclor 1254 in a 16-week study. (Zinkl,
1977).  Many of the symptoms of systemic toxicity are similar to symptoms produced by other
TCDD-like halogenated aromatic  hydrocarbons and are associated with the interaction of coplanar
congeners with the Ah receptor (Safe, 1990). A diet containing 0.3-3.0 ppm of a coplanar
congener (3,4,3',4'-TCB) produced chloracne, weight loss, and increased mortality in rhesus
monkeys while the same doses of a noncoplanar congener produced no effect (McNulty et  al.,


PCB exposure is associated with involution of the thymus and with impaired humoral and cellular
immunity.  The immunosuppressive effects of PCBs are believed to be mediated by the coplanar
congeners (Safe, 1990). Evidence of the influence of the Ah receptor on immune function was
shown in a study where pre-and post immunization doses of a coplanar congener (3,4,3',4'-TCB)
suppressed IgM response, caused  thymic atrophy and liver pathology and induced cytochrome P-
450 in Ah-f- but not in Ah- mice (Silkworth and Grabstein, 1982).  In the same study, a
noncoplanar congener (2,5,2',5'-TCB) did not produce any effect in either strain of mice at the
doses tested (10 and 100 mg/kg ip).

Tryphonas et al. (1991) examined the effect of chronic Aroclor 1254 exposure on the immune
function of rhesus monkeys at four dose levels ranging from 5 ug/kg-day to 80 ug/kg-day.  A
dose dependent decrease in anamnestic response to a T-cell dependent antigen (sheep red blood
cells) was demonstrated by  decreases in IgM titers at all doses tested once a week for 4 weeks
following immunization, and a statistically significant trend for decreased IgG with increased dose
levels for 3 out of 4 weeks.  Anamnestic antibody response to pneumococcal vaccine (which is
relatively independent of T-cell help and macrophage accessory functions) was not statistically
different from the control response.  Lymphocyte proliferation was reduced in response to
concanavalin-A and phytohemagglutinin (T-cell mitogens), while response to poke-weed mitogen
(a T and B cell mitogen) was unchanged.  Monocyte chemiluminescence in response to the
activator phorbol myristrate acetate (but not to the activator zymogen) showed a significant delay
in time to peak reading. These results suggest immunosuppressive effects in adult animals  at
doses of Arochor 1254 as low as 5 ug/kg-day and may reflect changes in T-cell or macrophage
function. No threshold has been established for immunotoxic effects of PCBs.


Women exposed to PCBs/PCDFs in the Yusho incident experienced menstrual cycle irregularities
suggesting altered ovarian function.  Animal studies support this association.  A single dose of 4
mg/kg-day Clophen 30 administered to four rhesus monkeys at ovulation was followed by
anovulatory cycles in two monkeys. Low estrogen and luteal phase progesterone concentrations
were also observed (Muller et al., 1978). Chronic administration of Aroclor 1248 to rhesus
monkeys at 2.5 ppm (approx 100 ug/kg-day) and 5.0 ppm (approx. 200 ug/kg-day) in the diet
resulted in increased menses duration, decreased estrogen and progesterone peaks, and increased
abortions (Allen et al., 1979).  In a 2-year study by Truelove et al. (1990), Aroclor 1254 did not
produce significant dose-dependent alteration in menses duration, menstrual cycle length or
progesterone or estrogen serum concentrations in rhesus monkeys, at doses up to 80 ug/kg-day.
This study is a sister study to the Tryphonas study and used the same dose levels and dose
delivery method.


PCB exposure has been linked to neurological effects in persons exposed in the Yusho and Yu-
cheng incidents and in occupationally exposed populations. Symptoms of PCB exposure  include
headache, numbness, altered peripheral nerve conduction velocity (Chen et al., 1985) and
decreased neurobehavioral function as measured by visual memory, problem solving and mean
choice reaction time (Kilburn et al., 1989). Nondevelopmental neurotoxic effects in rodents and
honhuman primates include decreases in dopamine function. There is evidence to suggest that
PCB neurotoxicity is associated not with the dioxin-like coplanar congeners, but with ortho and
ortho-para substituted congeners (Shain et al., 1991).  Seegal et al. (1991b) treated adult
nonhuman primates with either Aroclor 1016 or Aroclor 1260 at doses of 0.8,  1.6 or 3.2 mg/kg-
day for 20 weeks. Both Aroclors reduced concentrations of dopamine in the caudate, putamen,
and hypothalamus, but not in the globus pallidus or the hippocampus. Aroclor 1016, but not
Aroclor 1260, reduced dopamine concentrations in the substantia nigra.  Three ortho-substituted
nonplanar congeners (2,4,4', 2,4,2',4' and 2,5,2',5') accounted for more than 95% of the total
PCB residue found in the brain.  Lightly chlorinated congeners were more effective at reducing
dopamine concentrations than the more highly chlorinated congeners  (Seegal et al., 1990).


The carcinogenicity of PCBs has recently been reviewed in detail by Silberhorn et al. (1990).
Current epidemiological evidence does not provide  sufficient evidence of PCB carcinogenicity in
humans to establish a causal relationship, although several studies suggest a possible association
between PCB exposure and certain types of cancer  (Brown, 1987; NIOSH, 1991).
Carcinogenicity of the lower chlorinated mixtures in animals is a subject of debate, but there is
some evidence that Arochor 1254 and Kanechlor 500 induce liver tumors. Chronic
administration of Aroclor 1260 induces neoplastic nodules and hepatocellular carcinoma in rats.
EPA's current cancer potency factor estimate (7.7 per mg/kg-day)  is based on a study by
Norback and Weltman (1985)  in which Aroclor 1260 induced hepatocellular carcinoma in
Sprague-Dawley rats.  Some coplanar and some noncoplanar congeners have demonstrated
promoting ability, although with marked differences in potency (Silberhorn et al., 1990). More

 recent studies (Buchmann et al., 1991; Laib et al.,  1991) have identified several specific
 congeners with promoting activity and one congener with weak initiating activity. The initiating
 activity was demonstrated by a noncoplanar congener (2,2',4,5').

 Attempts to demonstrate genotoxicity of PCB congeners have been largely negative although
 there have been several recent positive findings.  Sargent et al.  (1992) examined the individual
 and combined effects of a coplanar and a noncoplanar PCB on the development of neoplasia in
 rat liver with or without the initiator diethyl nitrosamine (DEN).  Administration of 3,4,3',4'-
 TCB, together with 2,5,2',5'-TCB, produced a synergistic increase in chromosomal damage with
 or without DEN pre-treatment.  Induction of the P-450 enzymes by the planar congener was
 hypothesized to increase metabolism of the noncoplanar congener to a reactive arene oxide
 intermediate. Similar genotoxic results were previously demonstrated in lymphocytes (Sargent et
 al.,  1989) and increased chromosomal damage was  recently reported in an occupationally exposed
 population (Kalina et al.,  1991).


 A consideration of all lexicological endpoints associated with PCB exposure is crucial in
 determining which endpoints are of greatest sensitivity and of greatest concern.

 The deliberations will be guided by focusing on the following set of questions, for which a sense-
 of-the-meeting position will be sought.
How do the dose levels associated with developmental neurotoxicity compare to the dose
levels associated with 10"4 -10~6 cancer risk?                                  ,

Immunosuppression is associated with relatively low doses of PCBs.  Is there a threshold
for immunotoxicity, and how do the dose levels associated with this endpoint compare
with dose levels associated with developmental neurotoxicity?

Is there a correlation between congeners implicated in adult neurotoxicity and congeners
responsible for developmental neurotoxicity?

Brown, D.P.  1987. Mortality of workers exposed to polychlorinated biphenyls - an update.
Arch. Environ. Health. 42:333-339.

Buchmann A., et al. 1991.  Effects of PCBs in rat liver: Correlation between primary
subcellular effects and promoting activity.  Toxicol. Appl. Pharm. 111:454-468.

Chen R., et al.  1985.  Polychlorinated biphenyl poisoning: Correlation of sensory and motor
nerve conduction, neurologic symptoms, and blood levels of polychlorinated biphenyls,
quarterphenyls and dibenzofurans. Environ. Res. 37:340-348.

Kalina I., et al.  1991. Cytogenetic analysis of peripheral blood lymphocytes in workers
occupational!/ exposed to polychlorinated biphenyls.  Teratog. Carcinog. Mutagen. 11:77-82.

Kilburn K., et al.  1989.  Neurobehavioral dysfunction in firemen exposed to polychlorinated
biphenyls (PCBs):  Possible improvement after detoxification. Arch. Environ. Health. 44:345-

Kimbrough R., et al.  1987. Human health effects of polyehlorinated biphenyls (PCBs) and
polybrominated biphenyls (PBBs).  Ann. Rev. Pharmacol. Toxicol. 27:87-111.

Laib R.J., et al. 1991.  Hepatocarcinogenicity of polychlorinated biphenyl congeners. Toxicol.
Environ. Chem. 34:19-22.

McNulty, et al.  1980. Chronic toxicity of 3,3',4,4' and 2,2',5,5'-tetrachlorobipheriyls (PCBs)
in rhesus macaques.  Toxicol. Appl. Pharmacol. 56:182-190.

NIOSH Human Hazard Evaluation Report.  1991. Westinghouse Electric Corporation,
Bloomington, Indiana. HETA 89-116-2094.                                               •

Norback, D., and R.H. Weltman.  1985. Polychlorinated biphenyl induction of hepatocellular
carcinoma in the Sprague-Dawley rat. Environ. Health Perspect. 60:97-105.

Safe, S.  1990.  Polychlorinated biphenyls (PCBs), dibenzo-p-dioxins (PCDDs), dibenzofurans
(PCDFs), and related compounds: Environmental and mechanistic considerations which support
the development of toxic equivalency factors (TEFs). Crit. Rev. Toxicol. 21:51-88.

Sargent, et al.  1989. In vitro chromosome damage due to PCB interactions. Mutat. Research
224:79-88.                                                      '

Sargent, L.,  et al. 1992.  Ploidy and specific karyotypic changes during promotion with
phenobarbitol, 2,5,2',5'-tetrachlorobiphenyl, and/or 3,4,3',4'-tetrachlorobiphenyl hi rat liver.
Cancer Research 52:955-962.

Seegal, R., et al.  1990.  Lightly chlorinated ortho-substituted PCB congeners decrease dopamine
in non-human primate brain and in tissue culture.   Toxicol Appl. Pharmacol. 106:136-144.

Seegal, R., et al.  1991a.  Subchronic exposure of the rat to Aroclor  1254 selectively alters
central dopaminergic function.  Neurotoxicology 12:55-66.

Seegal, R., et al.  1991b. Comparison of the "effects of Aroclor 1016 and 1260 on non-human
primate catecholamine function. Toxicology 66:145-63.

Shain, et al.  1991.  Neurotoxicity of polychlorinated biphenyls: Structure-activity relationship of
individual congeners.  Toxicol. Appl. Pharmacol. 111:33-42.

Silberhorn, E. et al.  1990. Carcinogenicity of polychlorinated biphenyls: PCBs and PBBs. Crit.
Rev. Toxicol. 20:439-496.

Silkworth, et al. 1982.  Polychlorinated biphenyl immunotoxicity: Dependence on isomer
planarity and the Ah gene complex.  Toxicol. Appl. Pharmacol. 65:109-115.

Truelove, et al.  1990.  Effect of polychlorinated biphenyls on several endocrine reproductive
parameters in the female rhesus monkey. Arch Environ. Contamin. Toxicol. 19:939-943.

Tryphonas, et al.  1991. Effect of chronic exposure of PCB (Arochlor 1254) on specific and
nonspecific immune parameters in the rhesus (Macaco, mulatta) monkey.  Fund. Appl. Toxicol

                             PRE-MEETING ISSUE PAPER

                                  WORKSHOP TOPIC

                            PHARMACOKINETICS OF PCBs

                          Prepared by Dr. Linda Birnbaum, EPA
1.     To examine the pharmacokinetics of PCBs, both the commercial and environmental
       complex mixtures as well as the individual congeners.

2.     To discuss the issues of placental and lactational transfer and a comparison of
       embryo/fetal vs  maternal levels, as well as explore age-related differences in
       pharmacokinetic behavior of PCBs.

3.     To recommend  research needs to better characterize pharmacokinetic issues in assessing
       the special risk  of the developing organism to PCB exposure.


In order to extrapolate the risk of developmental neurotoxicity from exposure to PCBs, not only
must the hazard be identified and the mechanism elucidated, but the dose/response relationship
clearly defined.  PCBs represent a complex mixture of 209 possible congeners, of which
approximately 100 are present in the  commercial mixtures. The commercial  mixtures,  which have
been banned in the U.S. since 1977, involved different levels of chlorination.  The lower
chlorinated mixtures, which contained predominately the congeners with fewer chlorines, are
more rapidly degraded, both by environmental and metabolic processes.  In contrast, the higher
chlorinated mixtures tend to be extremely resistant to degradation, leading to their persistence in
the environment and bioaccumulation up theifood chain. The environmental mixtures present
today are not well represented by the commercial mixtures.  Since most of the exposure to the
general population occurs via consumption of food, market basket surveys provide the best
indication of ongoing exposure. Fish, dairy  products,  and meat provide the major dietary

PCBs are highly lipophilic compounds which are fairly well absorbed from the gastrointestinal
tract.  Depending on their lipophilicity, they may actually be absorbed via chylomicrons into the
lymphatics before entering into the systemic vasculature. The most highly halogenated congeners
(>6) are absorbed less well than the  lower chlorinated congeners, probably due to their poor
solubility. The PCBs are absorbed through the skin, but this represents a less important route
than the oral one. Few studies have  examined pulmonary absorption, but if PCBs enter the
respiratory tract they would tend to be well absorbed.

PCBs, once absorbed, are distributed via the blood bound to plasma proteins, especially the
lipoproteins. They initially distribute to highly perfused organs, such as the liver and muscle, and
then redistribute based on their lipophilicity to the more poorly perfused organs, such as the skin

  and adipose tissue.  The adipose tissue remains the major storage depot for the PCBs. Changes
  in this depot result in alterations- in distribution of the PCBs. Because of the sensitivity of the
  partitioning of PCBs between the serum and lipids in the various tissues, the PCBs levels are
  generally reported on a lipid-adjusted basis.

  PCBs can be eliminated by metabolism and/or excretion. In general, metabolism is a prerequisite
  for excretion.  For the higher chlorinated PCBs, excretion tends to be via the bile into the feces.
  Metabolites of lower chlorinated PCBs may be eliminated in the urine.  Because of the
  lipophilicity of the parent PCBs, lipid sinks such as milk, sebum, and oily secretions may result in
  passive elimination.   Some unmetabolized PCBs can also be passively eliminated across the gut
  into the intestinal lumen. Fecal excretion can be enhanced by dietary consumption of rice bran
  and cholestyramine.

 Metabolism of PCBs occurs via several mechanisms.  In general, lower chlorinated compounds
 are more rapidly metabolized than those with more chlorines. Chlorination on only one ring also
 leads to metabolism more readily than when both rings are  chlorinated. The most common
 metabolic pathway involves formation of an arene oxide leading to hydroxylation, preferentially
 at the para positions.  Two adjacent unsubstituted carbon atoms facilitate metabolism. In
 addition, chlorine migration and loss can occur during metabolism.  Direct insertion of a
 hydroxyl group has also been noted, especially at the meta position.  Hydroxylated PCB
 metabolites are usually conjugated, with glucuronidation, methylation and reaction with
 glutathione being major pathways leading to elimination.  While glucuronide and methyl
 conjugates are  readily excreted, glutathione conjugates are further metbolized via gut bacteria,
 leading to methylsulfonyl and methyl sulfide PCB metabolites, which concentrate in the lungs.
 Mammals appear to have greater ability to metabolize PCBs than do fish and birds.  Human
 metabolism of PCBs apears to resemble that of rodents and monkeys more closely than that of
 dogs, which have unusual ability to extensively metabolize the most persistent PCBs.  Since
 metabolism is the major determinant of elimination of PCBs, the half-life of highly chlorinated
 PCB mixtures in occupationally exposed workers appears to be from 6-12 years. Lower
 chlorinated PCBs have shorter half-lives.

 PCBs are readily transferred across  the placenta to the embryo/fetus, and can be detected in the
 amniotic fluid.  However, the major transfer of PCBs from mother to offspring occurs postnatally
 via lactation. .The lipid mobilization that  accompanies late pregnancy and lactation results in a
 dramatic decrease in the adipose tissue concentration of PCBs.  A nursing mother may reduce
 the concentration of PCBs in her tissues by 30-50% after 5-6 months of breast feeding.

 Pharmacokinetic studies have been conducted on a number of individual PCB isomers in rats,
 mice, guinea pigs, monkeys, and dogs.  In  vitro studies have examined metabolism of specific '
 PCB congeners  in human liver samples  as well as laboratory animals. In addition, the
 pharmacokinetic behavior of PCB isomers has been examined in senescent rodents and observed
 to reflect physiological changes in body composition and blood flow.  Little work has been
 conducted on pharmacokinetic differences between developing organisms and the adult. PCBs
induce the metabolism of related compounds, certain hormones,  and endogenous substrates.
This inductive response may change  as a function of age.


The most important pharmacokinetic issues concern understanding the tissue dosimetry, which is
associated with the biological response.  Of course, it is critical to understand the window of
sensitivity when examining developmental neurotoxicity issues.  If the effect is mediated
transplacentally, then very minute doses are bringing about the biological response.  For
extrapolation purposes, it is critical to understand which PCBs are involved in the response and
then study the dosimetric issues of those specific congeners.

The deliberations will be guided by focusing on the following set of questions, for which a sense-
of-the-meeting position will be sought.

1.     Which PCB congeners are responsible for the outcomes of interest?

2.     What are the pharmacokinetic properties of these PCBs - absorption?  distribution?
       metabolism? excretion?

3.     How persistent and bioaccumulative are the PCB congeners of interest?

4.     Can we develop a physiologically based pharmacokinetic model to predict the target
       tissue doses of the desired PCB,, in the embryo/fetus? following lactation in the young


Birnbaum, L.S. (1985) The role of structure in the disposition of halogenated aromatic
xenobiotics. Environmental Health  Perspectives, 61:11-20.

Jensen, A.A. (1987) PCBs, PCDDs, and PCDFs in human milk, blood, and adipose tissue.
Science in the Total Environment, 64: 59-293.

Kutz, F.W., P.H. Wood, D.P. Baltimore (1991) Organochlorine pesticides and polychlorinated
biphenyls in human adipose tissue.  Reviews in Environmental Contamination and Toxicology,
120: 1-85.

Matthews, H.B. and R.L. Dedrick (1984) Pharmacokinetics  of PCBs. Annual Review of
Pharmacology and Toxicology, 24: 85-103.

Muhlebach, S., P.A. Wyss and M.H. Bickel (1991) The use of 2,4,5,2',4',5'-hexachlorobiphenyl as
an unmetabolizable lipophilic model compound. Pharmacology and Toxicology, 69:410-415.

Schnellmann, R.G., A.E.M. Vickers, and I.G. Sipes (1985). Metabolism and  disposition of
polychlorinated biphenyls. In: Review of Biochemical Toxicology, Vol. 7 (E. Hodgson, J.R. Bend,
and R.M. Philpot, eds.), Elseveir, New York, pp. 247-282.

WHO/IPCS (1992) Environmental  Health Criteria for Polychlorinated Biphenyls and
Polychlorinated Terphenyls. In press.

Yakushiji, T. (1988) Contamination, clearance, and transfer of PCBs from human milk. Reviews
in Environmental Contamination and Toxicology, 101: 139-164.

                              PRE-MEETING ISSUE PAPER

                                   WORKSHOP TOPIC


                           Prepared by Dr. James McKinney, EPA



To examine the SAR studies related to developmental toxicity in animals and humans for
PCBs and related dioxins.

To focus on SARs for developmental neurotoxicity for PCBs in general as well as for
specific congeners.

To make recommendations regarding specific PCB structural features that determine the
developmental toxic potential in animals and humans.
The available literature across almost all of the toxic effects of PCBs suggest that the congeners
fall into two major categories or classes--(i) dioxin-like congeners that bind the Ah receptor and
induce associated biological responses and (ii) certain ortho-substituted congeners that bind very
poorly, if at all, to the Ah receptor.  The dioxin-like congeners include the non-ortho substituted
or "coplanar" type PCBs as well as certain monoortho-substituted PCBs.  This class  of PCBs may
reflect a stacking type of interaction in the receptor binding domain.  The other class generally
includes diortho  and above substituted congeners, but the underlying nature of their molecular
recognition in biological systems is not clear. It has been suggested that congeners of this type
because of steric constraints may prefer to undergo "cleft type" binding interactions such as has
been shown for their specific binding to the human thyroxine carrier protein prealbumin (and
possibly other prealbumin-like proteins).

There is an increasing body of evidence that PCBs and related compounds exert many of their
toxic effects through alteration in hormonal function (either changes in circulating concentrations
of hormones or in receptor number or affinity).  For example, some developmental neurotoxicity
might be associated with the anti-estrogenic properties of the dioxin-like PCB congeners which, in
turn, can affect neurochemistry such as dopaminergic function. In this case, the
structure-dependent potencies of dioxin and related compounds were similar to their Ah receptor
binding affinities. This also appears to be the case for Ah receptor antagonist ligands (such as
6-methyl-l,3,8-trichloro-dibenzofuran) that  show" a range of antiestrogenic activities but are
relatively nontoxic compared to dioxin. Another class of hydroxylated PCBs (presumed
metabolites) have been shown to bind specifically to the mouse uterine estrogen receptor and as
such represent potential agonists/antagonists for certain estrogen functions.  The
structure-dependent binding activities in this case are believed to be associated with  the close

  match of the phenolic ring with the A-ring in estradiol and the comformational restriction and
  possibly increased hydrophobic bulk of the molecule brought about by the presence of the
  ortho-chlorines and other substituents.

  The neurotoxic potential of PCBs, especially to the developing nervous system, has been indicated
  in epidemiological studies as well as through investigations using laboratory rodents and
  non-human primates. Studies to determine the exact nature of the PCBs responsible for the
  neurotoxic action indicated some common structural features suggesting the involvement of
  specific binding sites associated with dopamine synthesis (tyrosine hydroxylase activity).
  Although the structure-dependent neurotoxic activities appear to be associated with some degree
  of ortho chlorine substitution, it is not clear if this is a direct stereoelectronic effect on specific
 binding or an indirect effect, of reducing the binding to Ah receptor related proteins and that other
 structural requirements are more important in site recognition.  There are peculiar aspects of this
 SAR that are reminiscent of the SAR of PCBs  on binding to prealbumin.  It is interesting to note
 in this regard that there is some evidence to suggest that tyrosine hydroxylase can mediate the
 deiodination of thyroxine to triiodothyronine and such deiodinases appear to  be in the prealbumin
 family of proteins. la addition, the de novo synthesis of prealbumin is a major function of the
 choroid plexus of the brain and can thus directly influence the availability of thyroid hormones  to
 the central nervous system. PCBs can not only alter thyroid status by reducing circulating
 concentrations of thyroid hormones but may also interfere with the metabolism of tyrosine as well
 as with the metabolism of thyroxine (as structural analogs) at the tissue/cellular level.

 Biochemical thyroid status at the cellular level may be  a factor in the neurotoxicity  of PCBs.
 lodothyronines are known to be essential for brain development, and many investigators have
 provided evidence demonstrating effects of the hormones on immature brain ceU maturation,
 myelinogenesis, protein metabolism, nucleic acid metabolism, and electrical activity of the
 growing brain.  In addition, there are a number of well-documented examples of the regulation of
 single and multiple clusters  of genes through mutiple hormonal interactions.  Thyroidal, adrenal,
 and gonadal hormones are all necessary for the normal  differentiation of the nervous system and
 are involved in the ultimate expression of the target genes in several of these multi-hormonal
 regulating systems.

 Certain PCBs and PBBs are known to alter serum and/or tissue levels of Vitamin A. Limited
 studies show that the SARs  for altering hepatic retinoids differ from those for serum retinol
 implying the involvement of multiple mechanisms.  "Coplanar" type PCBs have pronounced
 effects on hepatic, renal, and serum retinoids whereas other biphenyls only decreased serum
 retinol level.

 PCBs can apparently alter hormonal functions by acting directly as hormonal agonists and
 antagonists with specific binding proteins or indirectly by altering the receptor number or affinity
 of other hormones involved in multihormonal regulating systems.  Therefore, some  PCB mixtures
may have complex interactive effects in biological systems.  Understanding these properties of
PCBs and how they are determined by structure would be an important focus of our SAR


1.     In view of the developing understanding of SARs for the developmental toxicity of PCBs
       and dioxins, what can we say with regard to the determining structural features that might
       underlie both their qualitative as well as quantitative behavior in developing biological

2.     An important question to consider is whether or not most PCB developmental toxicity is
       associated directly or indirectly with the thyromimetic properties of the PCBs.

3.     A direct search for specific PCB binding proteins in tissues is needed, particularly in
       relationship to PCB developmental toxicity.  This would include a better understanding of
       the structure-binding relationships of ortho-substituted PCBs thought to be involved in
      ' developmental neurotoxicity.
            i •                                     -     '                 •

Buff, K., and A. Brundl (1992). Specific binding of polychlorinated biphenyls to rat liver cytosol
protein.  Biochem. Pharmacol. 43(5): 965-970.

Chen, L.C.,  et al. (1992). Polychlorinated and polybrominated biphenyl congeners and retinoid
levels in rat tissue:  Structure-activity relationships. Toxicol.  Appl. Pharmacol. 1214:47-55.

Eayrs, J.T. (1971).  In "Hormones in Development" (M. Hamburgh and EJ.W. Bamngton, eds.),
Appleton, New York; also see J. T. Eayrs (1964) Arch. Biol.  75: 529.

Ford, D.H.,  and E.B. Cramer (1977). Developing nervous system in relation to thyroid hormones.
In "Thyroid Hormones and Brain Development" (G.D. Grave, ed.), Raven Press, New York, pp

Korach, K., et al. (1988).  Estrogen receptor-binding activity of polychlorinated hydroxybiphenyls:
Conformationally restricted structural probes. Molecular Pharmacology, 33: 120-126.

Safe, S., et al. (1991).  2,3,7,8-Tetrachlorodibenzo-p-dioxin and related compounds as
antiestrogens: Characterization and mechanism of action. Pharmacology and Toxicology, 69:

Shain, W., et al. (1991).  Neurotoxicity of polychlorinated biphenyls: Structure-activity
relationship of individual congeners.  Toxicology and Applied Pharmacology,  111: 33-42.

                               PRE-MEETING ISSUE PAPER

                                    WORKSHOP TOPIC


                             Prepared by Dr. Diane Miller, EPA
To evaluate the available data base on the developmental neurotoxicity of PCBs, to
identify data gaps that hinder a decision regarding the developmental neurotoxicity of the
PCBs and to recommend research to eliminate these deficiencies.

To evaluate the role of structure-activity in the developmental neurotoxicity of the PCBs.
That is, are non-coplanar PCBs, especially the ortho-substituted and less highly chlorinated
members of the series, more or less likely to be neurotoxic to the developing nervous
system than TCDD-like congeners?

To determine if maternal factors (e.g. undernutrition, liver damage, porphyria, altered
endocrine function, vitamin deficiency, etc.) play a significant role in the developmental
neurotoxicity  of the PCBs.

 Alterations in growth and spontaneous activity, delays in sensorimotor maturation, neuromuscular
 deficits, abnormal motor coordination, and compromised learning can result as a consequence of
 developmental exposure to PCBs in experimental animals.  Experimental data concerning the
 developmental neurotoxicity of the PCBs are available from a number of animal species including
 several species of rat and mouse, nonhuman primates and fowl (reviewed in Tilson et al., 1990,
 and Golub et al., 1991). Offspring of humans exposed to complex mixtures containing PCBs as
 well as other halogenated aromatic hydrocarbons also display neurological and developmental
 abnomalMes suggestive of compromised neural development. Thus, a primary issue concerning
 the PCBs is their developmental neurotoxicity. Of particular concern is the mechanism (or
 mechanisms) by which a particular congener or a complex mixture of PCBs alters development of
 the nervous  system.                                                                    ;

 While these animal and human studies were able to clearly identify the hazards to the nervous
 system associated with developmental exposure to the PCBs, most of the studies are descriptive in
nature. Due to the primary emphasis on hazard identification, exposures were of long duration
and maintained throughout the developmental period. In most instances, functional or behavioral
indices were used to indicate compromised development of the nervous system and few studies
have attempted to identify the particular changes in neural substrates (e.g., neuropathological,
structural or neurochemical alterations) responsible for the observed behavioral change.

For example, many studies used motor activity measures as a primary means for determining the
impact of PCBs on neural development.  Motor activity is defined as the frequency of movements
over a period of time and is quantifiable.  It can be recorded automatically (e.g., the number of
times a photocell is interrupted, or some other means of detecting movement; see Agrawal et al.,
1981) or manually, such as in the open field test (e.g., number of squares crossed; see Lilienthal et
al., 1990).  Motor activity is an integrated behavior pattern, that is, it involves the coordinated
participation of sensory, motor, and integrative processes.  Consequently, it is considered to be an
appropriate measure for the detection of CNS alterations and is one  of the test procedures
included in standard neurotoxicity testing (see EPA, 1991).  Increases and decreases in activity or
a failure to decrease activity within a test session or with repeated testing (i.e., habituation) are all
considered indicative of possible compromised CNS development. All of the preceeding changes
in activity have been reported as a function of developmental exposure to PCBs (see Table 2 in
Tilson et al., 1990 review).  Changes in motor activity (especially those that are  long-lasting or
that appear to be irreversible) are considered to be important in the detection of CNS insult
because they are known to accompany damage to or alterations in the development of specific
brain structures or areas (striatum, hippocampus, etc.).  Further, motor activity changes can occur
as a direct effect of a chemical and can accompany changes in neurotransmitter  levels engendered
by blocked uptake or release of that neurotransmitter.

As poor cognitive performance in humans resulted from gestational  exposure to  PCBs, many
investigators have employed endpoints designed to evaluate learning and/or memory as a means
for detecting brain alterations  after developmental exposure to PCBs (see Table  3 in the Tilson et
al., 1990 review).  Learning and/or memory are measured indirectly  with learning considered as
any lasting change in behavior due  to experience.  Memory is considered to be the persistence of
a learned behavior over time.  Further, as both must be inferred, changes in behavior resulting
from  motivational, sensory or motor impairment must be eliminated or controlled for to insure
compromised performance is really the result of alterations in learning and memory.  In the study
of PCBs, avoidance procedures have been used predominately with rodents and  fowl species.  The
more limited work with primates has utilized appetatively motivated tasks involving discrimination
of shape or position. A few investigators have  used maze procedures or operant schedule-
controlled performance to evaluate learning and memory  (see Table 3 in Tilson  et al., 1990).
Avoidance tasks require the subject to  emit a specific response (e.g., press a lever, move to an
adjacent chamber, etc.) in response to a signal (e.g., tone) that indicates an aversive event  (e.g.,
shock) will occur within a specific period of time (see  Lilienthal et al., 1991, and Pantaleoni et al.,
1988). Many studies investigating the developmental consequences of PCBs have reported
alterations in the acquisition of avoidance behavior with both superior as well as inferior
performance relative to controls  being  observed. A change in either direction can be indicative of
insult to the CNS as lesioning of many different brain areas will facilitate acquisition of avoidance
because  pre-existing response tendencies (e.g., freezing in response to shock) are eliminated.

In studies utilizing discrimination procedures, the subject is required to emit a response (e.g.,
lifting an object) indicating differentiation between stimuli (e.g., color, shape, etc., of an object)
for a reward (food, etc.).  In reversal-discrimination procedures the meaning of the stimuli are
switched following acquisition of the original discrimination.  As with the  avoidance studies in
rodents,  the primate studies utilizing discrimination procedures reported both facilitation and
impairment of learning as a function of developmental exposure to PCBs (e.g., see Schantz et al.,

 1989). Several of these same authors (Levin et al., 1992) have recently proposed that the pattern
 of facilitation and deficits in reversal-discrimination procedures following toxicant exposure may
 provide clues as to the brain areas affected.  For example, the facilitation of color and shape
 reversal learning observed following developmental exposure to PCBs is similar to that found
 following cortical lesions in the monkey. However, no neuropathological data from monkeys
 treated with PCBs developmentally exists to confirm or challenge, this hypothesis.

 There has been little or no attempt made to determine exactly how complex mixtures of PCBs or
 a specific congener can interfere with brain function or development.  There is some confusion,
 for example, as to whether dibenzofurans contribute significantly to the neurotoxicity observed
 with PCB mixtures. Nor has there been a concentrated emphasis  on the role a particular
 developmental stage (e.g., day of gestation, etc.) may play in the neurotoxicity displayed.
 However, several previous studies reviewed by Tilson (e.g., Pantaleoni et al., 1988) indicate the
 prenatal period may be the most sensitive period for exposure.  The recent work of Peterson's
 laboratory (Mabry et al., 1992) indicating alterations of male sexual behavior and reproductive
 development after a single exposure to TCDD on gestational day 15 are therefore of particular
 interest.  Finally, only a few recent studies concerning the neurotoxicity of the PCBs have
 considered the endocrine and systemic effects of certain of these compounds as possible
 determinants of the observed neurodevelopmental effects or behavioral changes (see Lilienthal et
 al., 1990, and Seegal & Shain, 1992).


 The deliberations concerning the neurotoxicity of the PCBs will be guided by the following
 questions.  Sense-of-the-meeting positions among participants will be sought.
Do PCBs cause structural or neurochemical (e.g., neurotransmitters, neural growth factors,
etc.) alterations of the nervous system that can account for the types of developmental
delays and behavioral/functional alterations that are observed following gestational and/or
perinatal exposure to these compounds? Does exposure during different periods of
development result in specific patterns of neurotoxicity?

The PCBs as well as other halogenated aromatic hydrocarbons share structural properties
with various hormones, possess hormonal activity and can impact on gonadal, adrenal and
thyroid systems.  In some instances their impact on hormonal systems involves modulation
at the level  of gene transcription rather than alterations at the hormone receptor level.
Interference with development of these systems or exposure  of the developing organism to
agents with  gonadal, steroidal or thyroidal properties can compromise  development of the
nervous system that is manifested by structural as well as behavioral alterations.  Several
related questions are prompted by the endocrine aspects of the PCBs.  Can the endocrine
properties of various PCBs alter CNS development either by acting on the maternal
compartment (e.g., cause hypothyroidisrn in dam) or by a direct effect on the fetus (e.g.,
alter development of thyroid axis or prevent normal interactions of thyroid hormone and
brain development)?  Will PCBs with TCDD-like properties  compromise gonadal status
and thus alter or impair sexual differentiation of the CNS? Does developmental exposure
to PCBs alter the morphology or development of sexually dimorphic areas of the nervous

       system?  Are there any human studies suggesting alterations in sexually dimorphic
       behaviors as a function of exposure to PCB mixtures?

3.     Is there a homology across species regarding the neurobehaviofal or neurodevelopmental
       changes observed in humans following exposure to complex mixtures of PGBs and the
       changes observed in animal subjects exposed to known doses of PCBs? For example, are
       delays in reflex development or the appearance of specific developmental landmarks (e.g.,
       eye opening) in experimental animals exposed to PCBs during development predictive of
       specific effects in humans?

4.     A tremendous problem in determining the neurotoxicity of the PCBs is the large number
       of congeners that must be tested.  Strategies for dealing with the problem of predicting the
       general toxicity of the halogenated aromatic hydrocarbons, including the PCBs, have
       involved the use of toxicity equivalents with TCDD serving as the standard or most toxic
       congener.  Toxic potency is related to aryl hydrocarbon hydroxylase (AHH) induction
       capabilities and binding to the putative Ah receptor. Will such a strategy be useful in
       determining the neurotoxicity of PCBs?  Should other toxic equivalency factors be
       included such as the ability to  affect the thyroid axis (e.g., thyroxine equivalents)? Are
       there in vivo bioassays (e.g., hormone changes, thymus  weight changes, etc.) or in vitro
       procedures (PC12 cultures, etc.) available that will provide guidance as to the potential
       developmental neurotoxicity of specific congeners or complex mixtures of the PCBs?

5.     The pharmacokinetics of the PBC congeners are markedly affected by the position of the
       chlorines on the parent biphenyl structure.  Commercial mixtures are identified by their
       percent chlorine content expressed on a weight basis. Is there a relationship between the
       distribution and deposition of specific congeners to brain and the ability of the congener to
       alter CNS development?  That is, are target tissue levels of a particular congener of use in
       predicting the neurotoxicity of that congener?  Are the distributional and metabolic
       properties of the various congeners altered as a function of being members of a complex
       mixture?  That is, does the neurotoxicity of a given congener change if exposure occurs in
       a complex mixture format?

6.     Some PCB congeners are rapidly metabolized and excreted (e.g., 3,4,3'4'-
       tetrachlorobiphenyl) but produce permanent aberrations in CNS  function and structure.
      • The arene oxide intermediates are  hypothesized to be involved in the metabolism of the
       PCBs in mammals and the production of these metabolites is increased with increased
       metabolism of PCBs.  Arene oxides are known to cause cellular necrosis.  Is there
       evidence to suggest that the congeners capable of producing more arene oxides are more
       likely to cause neurotoxicity to the developing nervous  system?  Are other metabolites of
       the PCBs likely to be developmentally neiirotoxic?


Agrawal, A. K., Tilson, H. A. and Bondy, S.C. 3,4,3'4'-tetrachlorobip'hehyl given to mice
prenatally produces long-term decreases in striatal dopamine and receptor binding sites in the
caudate nucleus. Toxicol. Letters 8:417-424,1981.

 Chen, L.-C., Berberian, L, Koch, B., Mercier, M., Azais-Braesco, V., Glauert, H. P., Chow, C. K.
 and Robertson, L. W. Polychlorinated and polybrominated biphenyl congeners and retinoid levels
 in rat tissues:  Structure-activity relationships.  Toxicol Appl. Pharmacol.  114:47-55,1992.

 Golub, M. S., Donald, J. M. arid Reyes, J. A.  Reproductive toxicity of commercial PCB mixture^:
 LOAELs and NOAELs from animal studies. Environ. Health Perspect. 94:245-253,1991.

 Landers, J. P. and Bunce, N. J.  The Ah receptor and the mechanism of dipxin toxicity:  Biochem.
 1.276:273-587,1991.                                      :   -        v     -     ^.

 Levin, E. D., Schantz, S. L. and Bowman, R. E. Use of the lesion model for examining toxicant
 effects on cognitive, behavior. Neurotoxicol. TeratOl. 14:131-141,1992.             :

 Lilienthal, H. and Winneke, G.  Sensitive periods for behavioral toxicity of polychloritiated
 biphenyls: Determination by cross-fostering in rats. Fundam. Appl.  Toxicol.  17:368-375,1991.

 Lilienthal, H., Neuf, M., Munoz, C. and Winneke, G. Behavioral effects of pre- and postnatal
 exposure to a mixture of low chlorinated PCBs in rats. Fundam. Appl. Toxicol.  15:457-467,1990.

 Mably, T. A., Moore, R. W., Goy, R. W. and Peterson, R. E. In utero  and lactational exposure of
 male rats to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol. Appl. Pharmacol. 114:108-117,1992.

 Pantaleoni, G. C., Fanini, D., Sponta, A. M., Palumbo, G., Giorgi, R., Adams, P. M. Effects of
 maternal exposure to  polychlorobiphenyls (PCBs) on Fl generation behavior in the rat.  Fundam.
 Appl. Toxicol.  11:440-449,1988.

 Rickenbacher, U., McKinney, J.  D., Oatley, S. and Blake, C. C. F. Structurally specific binding
 of halogenated  biphenyls to thyroxine transport protein.  J. Medical Chem.  29:641-648,1986.

 Safe, S.  Polychlorinated biphenyls (PCBs), dibenzo-p-dioxins (PCDDs), dibenzofurans (PCDFs),
 and related compounds: Environmental and mechanistic considerations which support the
 development of toxic  equivalency factors.  CRC Grit. Reviews Toxicol.  21:51-88,1990.

 Safe, S.  Polychorinated biphenyls (PCBs)  and polybrominated biphenyls (PBBs):  Biochemistry,
 toxicology, and mechanism of action.  CRC Crit. Reviews Toxicol. 13:319-395,1984.

 Schantz, S. L., Levin, E. D., Bowman, R. E., Heironimuss, M. P. and Laughlin, N. K. Effects of
 perinatal PCB exposure on discrimination - reversal learning in  monkeys. Neurotoxicol.  Teratol.

 Seegal, R. F., and Shain, W. Neurotoxicity of polychorinated biphenyls:  The role of ortho-
 substituted congeners  in altering neurochemical function. In The Vulnerable Brain and
Environmental Risks,  Eds. R. Isaacson and K. Jensen. IN PRESS, Plenum Press.

Syracuse Research Corporation. Toxicological Profile for Polychlorinated Biphenyls. U. S.
Department of Health and Human Services, Agency for Toxic Substances and Disease Registry,
U. S. Government Printing Office:  1991-537-123.

Tilson, H. A., Jacobson, J. L. and Rogan, W. J. Polychlorinated biphenyls and the developing
nervous system:  Cross-species comparisons. NeurotoxicoL TeratoL  12:239-248,1990.

Tilson, H. A., Davis, G. J., McLachlan, J. A. and Lucier, G. W. The effects of polychlorinated
biphenyls given prenatally on the neurobehavioral development of mice.  Environ. Research

U. S. Environmental Protection Agency.  Pesticide assessment guidelines, subdivision F.  Hazard
evaluation: human and domestic animals. Addendum 10:  Neurotoxicity, series 81, 82, and 83.
Office of Pesticides and Toxic  Substances, Washington, DC.  EPA 540/09-91-123. Available
from: NTIS, Springfield, VA.  PB91-154617,1991.

                              PRE-MEETING ISSUE PAPER

                                   WORKSHOP TOPIC


                           Prepared by Dr. Mary M. Prince, NIOSH
 1.     To examine the epidemiologic evidence relating polychlorinated biphenyl (PCB) exposures
       and increased risk for developmental neurotoxicity in human infants and children.

 2.     To evaluate the consistency and concordance of the epidemiologic studies to findings from
       experimental animal studies.

 3.     To evaluate the impact of analytic methods of exposure assessment for PCBs and related
       contaminants on interpreting results from epidemiologic studies.

 4.     To recommend future approaches to estimating the risk of PCB-related developmental
       toxicity in human infants and  children using appropriate measures of motor and cognitive
       Given the strengths and weaknesses of the epidemiologic studies on the effects of
       PCBs on infant development, how does one evaluate these results in relation to the
       animal data on reproductive and developmental toxicity to gain a consensus on
       whether there is an excess risk of these adverse outcomes due to low level PCB

Developmental neurotoxicity of PCBs has been extensively studied in rodents, monkeys, and
humans using measures of motor and cognitive function (Tilson et al., 1990).  Concerns regarding
developmental and reproductive toxicity were stimulated by studies of accidental poisonings in
Japan and Taiwan from cooking oils contaminated with PCBs and related compounds. These
studies found that pregnant women who ingested these contaminated oils reported a higher
incidence of intrauterine growth retardation among their offspring (Wong et al., 1981; Rogan et
al., 1988; Yen et al., 1989). The relevance of the Asian poisoning situation to the U.S. studies of
the general population is questionable for the following reasons:  (1) there were relatively high
exposures of PCBs, which were contaminated with highly toxic polychlorinated dibenzofurans
(PCDFs); and (2) the exposed pregnant women in these studies had overt symptoms of PCB
toxicity.  These findings are also consistent with effects of lower birth weight and gestational age
from occupational exposures to PCBs (Taylor et al., 1989).  However, data based on these
populations cannot be directly extrapolated to  the lower exposures from environmental sources.

Most of the evidence for developmental toxicity due to chronic low level environmental PCB
exposures in humans is derived from epidemiologic studies in North Carolina (Gladen et al., 1988,
1991; Rogan et al., 1986), and Michigan (Fein et al., 1984; Jacobson et al., 1985, 1990, 1990a).
Both studies have been extensively reviewed to assess whether the observed results can be
explained by chronic low level PCB exposure or to differences in conduct, design, and analyses of
these studies.  Both infant populations were found to have PCB-related developmental deficits
from prenatal exposure.  However, the largest of the two studies (e.g., North Carolina, NC) failed
to show that this effect persisted past 2 years of age.  On the other hand, the Michigan (MI) study
reported persistent motor and cognitive deficits in children at age four from prenatal exposures (as
measured from cord blood levels).


The disparity in the results of these studies has raised questions regarding the etiologic link
between low level PCB  exposure and neuro-developmental deficits. These include concerns about
selection bias, as well as issues related to lack of, or inconsistent control of, very important
confounding variables in the MI studies (Fein et al., 1984; Jacobson et al., 1985, 1990, 1990a).
However, other important factors may explain why these different populations produce seemingly
conflicting results.  A direct comparison of results from these 2 populations (Gladen et al.» 1991;
Jacobson  et al., 1990) is not possible given major differences in  (a) PCB exposure definitions  and
analytic quantification, (b) endpoints measured, and (c) control for confounding variables.  A
major challenge in comparing these studies is that PCBs represent a mixture of chemicals
consisting of up to 209  congeners, having varying degrees of chlorination, persistence, and
toxicity.   PCBs with lower degrees of chlorination tend to be more rapidly excreted while higher
chlorinated isomers are  retained in body lipid stores (Tilson et al., 1990). These issues must be
systematically evaluated and addressed while weighing the evidence for or against the hypothesis
that PCB  exposure at low levels produces neurobehavioral deficits in offspring.

The deliberations will be guided by focusing on the following set of questions, for which a sense-
of-the-meeting position  will be sought:

1.     Could incomplete control of potential confounding (e.g.,  maternal smoking, alcohol
       consumption, and weight) explain the positive findings from the MI study?         ,

2.     How different are the two populations in relation to the distribution of highly chlorinated
       PCBs (e.g., congeners most likely to accumulate in the body over time)?

3.     Could the different analytic methods for measuring total  PCBs in these studies mask
       differences in the magnitude of exposures and/or congener-specific composition of PCBs
       between the populations?

4.     What would be  the impact of having different congener distributions in each of the
       populations on the magnitude of the  association between PCB levels and developmental or
       reproductive toxicity?

 5.     Prenatal (in utero) exposures in these studies were different with respect to the
        physiological compartment sampled (e.g., NC used a measure of average milk PCB at
        birth up to endpoint measurement, while ME used serum cord blood). Could this explain
        the discrepancies between the results of these studies with respect to the association
        between prenatal PCB exposure and specific endpoints of interest?

 6.     The functional domains affected by PCBs appeared to be different in the NC
        (psychomotor performance) and MI (short term memory) studies.  Does this inconsistency
        weaken the evidence for an association between PCB exposure and neurobehavioral
        deficits, or can this disparity be explained by congener-specific differences in the
        population and/or differences in the type and comparability of test instruments across

 7.     About 80% of the cord serum levels in the NC  study were below the level of detection as
        compared  to 66% of the ME sample. What is the effect of having a large proportion of
        PCB measures in the non-detectable range on the estimation of risk and internal validity of
        each of these studies?

 8.     In view of these issues, how does one evaluate the consistency of the epidemiologic data
        to effects observed in well-controlled animal studies, for which there is good evidence for
        PCB-related effects on cognition and motor function?  Major questions should focus on
        the comparability of the experimental  animal data to the human studies with respect to the
        neurobehavioral and reproductive outcomes measured, the different PCB measures used for
        assessing pre- and post-natal exposures, and  characteristics of the PCB exposures in these


 Fein GG, Jacobson JL, Jacobson SW (1984): Prenatal exposure to polychlorinated biphenyls on
 birth size and gestational age: J.  Pediatrics, 105:315-320.

 Jacobson JL, Jacobson SW, Humphrey HE  (1990):  Effects of PCBs  and related compounds on
 growth and activity in children. Neurotoxicol. Teratol., 12:319-326.

 Jacobson JL, Jacobson SW, Humphrey HE (1990a):  Effects of in utero exposure to
 polychlorinated biphenyls  and related contaminants on cognitive functioning in young children. J.
 Pediatrics, 116:38-45.

Jacobson SW, Fein GG, Jacobson JL, et al. (1985). The effect of intrauterine PCB exposure on
visual recognition  memory.  Child Dev. 56:853-860.

Gladen BC, Rogan WJ, Hardy P, et al. (1988): Development after exposure to polychlorinated
biphenyls and dichlorodiphenyl dichloroethene (DDE) transplacentally and through human milk
J.  Pediatrics,  113:991-995.

Gladen BC and Regan WJ (1991): Effects of perinatal polychlorinated biphenyls and
dichlorodiphenyl dichloroethene on later development.  J. Pediatrics, 119:58-63.

Rogan WJ, Gladen BC, McKinney JD, et al. (1986): Neonatal effects of transplacental exposure
to PCBs and DDE. J. Pediatrics, 109:335-341.

Rogan WJ, Gladen BC, Hung KY et al. (1988): Congenital poisoning by polychlorinated
biphenyls and their contaminants in Taiwan. Science, 241:334-336.

Taylor PR, Stelma JM, Lawrence CE (1989): The relation of polychlorinated biphenyls to birth
weight and gestational age in the offspring of occupationally exposed mothers. Am. J. Epid.,

Tilson HA, Jacobson JL, Rogan WJ (1990): Polychlorinated Biphenyls and the developing
nervous system: Cross-species comparisons. Neurotoxicol. Teratol., 12:239-248.

Wong KG, Hwang MY (1981). Children born to PCB poisoned mothers.  Clin. Med. (Taipei),

Yen YY, Lan SJ, Ko YC, and Chen CJ (1989).  Follow-up study of reproductive hazards  of
multiparous women consuming PCB-contaminated rice oil in Taiwan.  Bull. Environ. Contam.
Toxicol. 43:647-655.


                                  SECTION THREE

       Interest in the developmental neurotoxicity of polychlorinated biphenyl (PCB) compounds
continues to increase, based in part on a growing body of research that suggests a linkage
between prenatal or perinatal exposure to PCBs and neurotoxicity in the offspring. To explore
issues related to risk assessment in this area, a technical panel of EPA and FDA personnel was
established under the aegis of the Risk Assessment Forum. This panel, chaired by Dr. Carole
Kimmel, organized a two-day workshop on "Developmental Neurotoxic Effects Associated with
Exposure to PCBs," which  was held in Research Triangle Park, North Carolina, on September 14
and 15, 1992. At this workshop, experts in the areas of developmental toxicity, neurotoxicity,
PCBs, pharmacokinetics, structure-activity relationships,  and epidemiology were invited to discuss
issues related to the developmental neurotoxicity of PCBs, to make recommendations on the
utility of existing data for risk assessment purposes, and  to identify areas of additional research
needed to improve risk assessment in this area:  Twenty-five individuals representing academic
institutions, industry, and various federal agencies participated in the workshop, along with seven
members of the technical panel responsible for organizing the meeting (see Appendix C).

       An overview of the background and goals of the workshop was provided by Dr. Linda
Birnbaum, who chaired the opening session. After welcoming attendees and thanking them for
agreeing to participate in the workshop, Dr. Birnbaum described a number of factors that had
influenced the decision to hold this meeting. She noted  that activities undertaken by several
other scientific bodies in the past year had generated considerable interest in PCBs and their
possible role as developmental neurotoxicants. For example, an October 1991 review conducted
by the Canadian Department of Health and Welfare found evidence of extreme developmental
toxicity in monkeys exposed to Aroclor 1254, a commercial PCB mixture. Although the number
of offspring had been inadequate to support an evaluation of the developmental neurotoxicologic
effects of PCBs per se, the  nature and severity of the effects that were observed convinced many
people of the need for more research on the developmental neurotoxicity of these compounds.
Similarly, a workshop held by the World Wildlife Fund in March of 1992 had examined the role
of endocrine disrupters, including PCBs, in the environment. Data presented at this workshop
on the structural and behavioral effects  of endocrine disrupters in various species of wildlife were
of concern to many researchers, since wildlife often serve as sentinels  for humans in their
responses to environmental toxicants. Finally, a third impetus for the current workshop was
provided by EPA's ongoing reassessment of the environmental risks posed by dioxin and
dioxin-like compounds. Given that certain  PCBs are dioxin-like  in their effects, the Agency
considers it critical that the latest and best  science regarding potentially toxic effects of PCBs be
incorporated into the dioxin reassessment wherever possible.

       Against this background, Dr. Birnbaum stated that one purpose of the workshop would
be to evaluate the body of  scientific data that is currently available regarding the developmental
effects  of PCBs, with particular emphasis on data related to the effects of these compounds on

neurobehavioral development.  She noted that, in general, three types of information comprise
the existing data base:

       •      recent studies of the offspring of rodents and monkeys exposed to PCBs, in which
              prenatal and perinatal exposure to these compounds have been demonstrated to
              produce neurotoxicity in the offspring;

       •      studies of human poisoning episodes in Japan and Taiwan, in which
              developmental delays and neurobehavioral deficits have been observed in the
              offspring of women exposed to PCB-contaminated rice oils; and

       »      cohort studies in Michigan and  North Carolina in which exposure to relatively low
              levels of PCBs has been associated with neurobehavioral deficits in infants and
              young children.

In general, data from the poisoning episodes have been considered of limited  utility for risk
assessment purposes, both because of the high doses of PCBs involved and because highly toxic
polychlorinated dibenzofurans are known also  to have been present in the contaminated oils.  In
terms of the available data, therefore, the workshop was expected to  focus most heavily on
studies of PCB toxicity in animals and on the epidemiologic studies of low-level PCB exposure
conducted in Michigan and North Carolina.

       To determine the adequacy of these  data for risk assessment purposes, Dr. Birnbaum
suggested that a number of questions would need to be addressed by the workshop, including
the following:

       •      Question #1:  Are all PCBs alike in their effects on neurobehavioral development? If
              not, are any structure/activity relationships discernible?

It is clear, for example, that some PCBs are dioxin-like in their behavior, while others are not.
One issue that might be addressed by the workshop, therefore, is whether there are specific
structural  features that might be used to group congeners in ways that are relevant to their
effects on various toxic or behavioral endpoints.

       •      Question #2:  What are the dose-response relationships?

Dr. Birnbaum noted that this question is often complicated by the fact that most environmental
exposures to PCBs involve mixtures of congeners with varying toxicities.

       •      Question #3:  Are there populations at special risk due to elevated exposure or to
              inherent sensitivity?

In some ways, Dr. Birnbaum noted, the designation of effects as developmental assumes  an
enhanced  susceptibility on the part of the developing embryo, fetus, or neonate.  Enhanced
susceptibility  and elevated exposure are not mutually exclusive, however. The suckling infant, for
example, might be at increased  risk both  as a result of increased exposure (due to the high
concentration of PCBs in breast milk) and as a result of enhanced susceptibility (to the extent
that toxic effects are developmental stage-specific).

       •      Question #4:  What are the endpoints of greatest concern? Of greatest sensitivity?

It might be important to focus on both of these questions, Dr. Birnbaum suggested, since the
endpoints of greatest concern might or might not be the same as the most sensitive endpoints.

       While noting that discussions at the workshop were expected to cover a broad range of
topics, Dr. Birnbaum indicated that the organizers would prefer discussion to focus on two main
issues. The first is whether currently available information on the developmental neurotoxicity of
PCBs is sufficient to be used for risk assessment purposes. Second, participants were asked to
identify areas in which major uncertainty remains regarding the developmental neurotoxicity of
PCBs and, wherever possible, to suggest research that might resolve these uncertainties and
thereby improve the scientific basis of the risk assessment.

       Dr. Birnbaum said that these two areas of discussion could be thought of more generally
as what the science is already telling us,  and  what we still need to know.

       Dr. Birnbaum stressed that  the organizers did not view the workshop as a "consensus
conference" in the sense that no one expected participants to reach a common position on all of
the issues before the group.  The organizers  did, however, think that it would be useful to
attempt to arrive at a general "sense of the meeting" regarding the current state of the science.
An attempt to prioritize identified research needs would also be of particular value to the
sponsors of the meeting.  Finally, Dr. Birnbaum reminded participants that the workshop was
intended to address scientific rather than policy issues, and that in this regard the focus of the
meeting should remain on scientific issues  related to risk assessment rather than on policy issues
having more to do with risk management.

       Following several  administrative  announcements, Dr. Birnbaum concluded her
introductory presentation by reviewing the proposed agenda for the two-day meeting (Appendix
B). She noted that the first morning of the workshop would be devoted to a series of formal
presentations on general background issues related to PCBs.  During the afternoon session,  the
workshop was scheduled to break into two smaller work groups that would evaluate the animal
and human toxicity databases, respectively. Then, on the second day of the meeting, the whole
group would reconvene to discuss the, conclusions and recommendations reached by both work
groups, particularly as those conclusions  relate to the assessment of developmental neurotoxic
risks associated with exposure to these compounds.  Finally, at the end of the second day, a
"sense  of the meeting" summary would be presented by Dr. Carole Kimmel for consideration by
the workshop as a whole.
       Michael Bolger, Center for Food Safety and Nutrition, FDA

       Dr. Bolger's presentation focused on the general toxicology of PCBs in terms of the
endpoints that have been used in hazard and risk assessments previously conducted by the FDA,
EPA, and other public health agencies.  Specifically, Dr. Bolger noted that the purpose of his
presentation was to demonstrate how knowledge about the toxic effects of PCBs has been used
to derive tolerable daily intakes and other values of regulatory significance.

        To begin his presentation, Dr. Bolger showed a diagram of the biphenyl structure that is
 common to all PCBs (see Figure 1).  Chlorine substitution at the ortho, meta, and/or para
 positions of either ring in the biphenyl structure can potentially produce any of 209 possible PCB
 congeners.  He noted that in both industrial and environmental exposure situations, it is usually a
 mixture of multiple PCB congeners that is present.

Figure 1
               meta     ortho
meta    ortho
                                                       4' y para
ortho      meta
       To provide an historical perspective on the dietary intake of PCBs, Dr. Bolger presented
 data from the FDA's Market Basket Survey, which has monitored dietary levels of Aroclor 1254,
 a commercial PCB mixture, over the past 20 years. Between 1971 and 1990, a 10- to  20-fold
 reduction in the dietary burden of PCBs was observed, as the average adult intake of Aroclor
 1254 declined from an average of 6.9 /*g/day to about 0.05 /ig/day. PCB  data from the Market
 Basket Survey are not broken down by congener, however.

       Dr. Bolger then discussed a recent draft monograph prepared by the Agency for Toxic
 Substances and Disease Registry (ATSDR), in which the dose-response information from
 numerous toxicity studies of laboratory animals were graphically analyzed to determine which
 endpoints are of greatest concern following various types of exposure to  PCBs. In all of these
 tables, NOAELs and LOAELs obtained in various species were plotted for a variety of
 endpoints.  Endpoints addressed in this analysis included death, various forms of systemic toxicity
 (i.e., respiratory, cardiovascular, gastrointestinal, hematologic, hepatic, renal, dermal, ocular, and
 other), and impairments in neurologic, developmental, and reproductive  function. Endpoints
 were considered more sensitive to PCBs as the reported NOAEL or LOAEL for the end-
 point decreased.

       In analyzing studies of acute exposure, which was defined  to include exposures of 14 days
 or less, the ATSDR study found that the endpoints most sensitive to  PCB exposure were hepatic
 and developmental toxicity, both of which had NOAEL/LOAELs of approximately 1.0 mg/kg/day
 or less.  In studies involving intermediate exposure, which was defined to include periods of
 exposure between 15  and 364 days, the most sensitive endpoints included gastrointestinal,
 immunologic, hepatic, dermal/ocular, neurologic, developmental, and reproductive effects, all of
which were observed  in the dose range of 0.1 mg/kg/day. In chronic exposures of 365  days or

more, immunologic and developmental effects were the most sensitive endpoints, with NOAELs
and LOAELs in the 0.01 mg/kg/day range. Dr. Bolger pointed out that the application of
standard uncertainty factors to these data produces a reference dose of approximately 10"5
mg/kg/day. This dose, in turn, falls well within the range of doses that comprise  the estimated
upper bound human cancer risk estimates associated with chronic exposure to these compounds.
For additional information and analysis of the available toxicity data, Dr. Bolger referred
workshop participants to a review recently published by Tilson, Jacobson, and Rogan.

       Dr. Bolger also presented data from a recent paper by Golub and coworkers in which the
doses of PCBs associated with reproductive  toxicity were compared to doses associated with
other types of toxic effects. At high doses of these compounds, toxic effects involving numerous
organ systems are seen; the most sensitive of these was altered liver histology, which occurred at
doses of approximately 5.0 mg/kg/day.  At lower doses, in the 0.5 to 5.0 mg/kg/day range, virtually
all evidence of toxicity involved reproductive function; toxic effects noted at these dose levels
included decreased conceptions, postnatal, thyroid dysfunction, postnatal developmental delay,
lower birth weights, and other related  effects.  In their analysis of the data, Golub and colleagues
determined LOAELs for the reproductive effects of PCBs as 0.25 mg/kg/day for  rodents and
0.008 mg/kg/day for non-human primates, levels significantly below the LOAELs associated with
toxicity to other organ systems. ..-....•

       Noting that Dr. McKinney would be presenting detailed information on structure-activity
work related to PCBs later in the morning, Dr. Bolger said that he would limit his
structure-activity focus to the question of whether it is appropriate to use toxic equivalency
factors (TEFs) in assessing the toxic potential of specific congeners.  He showed data from the
work of Safe and colleagues, who concluded that a TEF approach may be appropriate for
coplanar, dioxin-like PCBs, but probably not for non-dioxin-like compounds (e.g., the
ortho-substituted, non-coplanar congeners).  Functional differences between dioxin-like and
non-dioxin-like  PCBs have also been demonstrated in the work of Seegal and colleagues, who
have shown  that dioxin-like PCBs have relatively little effect on dopamine levels  in the brain,
while ortho-substituted congeners are highly active in depleting dopamine stores  in certain brain
regions.  Dr. Bolger noted that, because bur understanding of the mechanisms underlying the
functional differences among PCBs is so limited, we may simply not know enough to incorporate
this information in hazard and risk assessments  of these compounds at this time.

       As an example of previous efforts to quantify the risks associated with exposure to  PCBs,
Dr. Bolger presented a compilation of the data used by the FDA in deriving tolerance levels for
PCBs in fish and other foods.  In that  analysis, the FDA relied on the three rodent cancer
studies that were available at the time and on a study of primate reproductive performance
conducted by Barsotti and Allen. These data were combined with information from.the Japanese
Yusho poisoning, which was at the time thought to be a PCB problem, in a  risk assessment
paradigm that examined the lifetime risk associated with tolerances ranging from 1 to  10 ppm
PCBs in fish. Using a classic uncertainty factor approach on data available from the Yusho
poisoning, 99% upper confidence limits on the lifetime risk per 100,000 population* were
calculated for 50th and 90th percentile fish-eaters at each assumed tolerance level. These
estimates of risk were then integrated  with other risk management factor's to set  a final tolerance
level of 2 ppm for PCBs in fish.

       Dr. Bolger then presented data from the review by Tilson, Jacobson, and Rogan, in which
the reference doses calculated for various species using different endpoints are compared. He
noted that in rodent studies, reference doses generally fall in the range of 10"1 to 10~2 mg/kg/day;
in monkey studies, the values are more typically in the 10"4 to 10"s mg/kg/day range; and in
human studies they drop even lower, to the range of 10"6 mg/kg/day.  For purposes of
comparison, Dr. Bolger noted that the one-in-a-million carcinogenicity risk estimate for PCBs
currently stands  at the 10~7 mg/kg/day dose level, and the current EPA reference  dose, which is
based on reproductive and developmental effects in monkeys, is 10~4 mg/kg/day.

       To conclude his overview of PCB hazard and risk assessments, Dr. Bolger presented data
from an assessment recently performed for the Uniform Great  Lakes Basin Fish Consumption
Advisory Protocol by researchers in the Wisconsin Department of Health.  This assessment was
based on work in the laboratories of Jacobson and Rogan estimating the threshold PCB
concentration in maternal breast milk at 1.0 and 3.4 j^g/g fat, respectively. In response to a
question from a  workshop participant, Dr. Rogan indicated that in both laboratories threshold
PCB concentrations were estimated graphically, based on the shape of the dose-response curves.
Assumptions incorporated into the Great Lakes analysis were that exposure to PCBs was chronic,
that PCBs have a biological half-life of one year, that maternal  weight averaged 60 kg with 25%
body fat, and that the concentration of PCBs in body fat stores  is equivalent to their
concentration in milk fat.  Based on these assumptions, tolerable body burdens of PCBs were
calculated to be  7.5 or 25 mg, depending on the threshold concentration used. These values, in
turn, correspond to reference doses of 3 x 10"5 and 1 x  10~4 mg/kg/day, which fall into  roughly the
same range as the reference doses  that have been calculated using other types of data.
       Linda S. Bimbaum, Health Effects Research Laboratory, EPA

       Dr. Birnbaum's presentation focused on the pharmacokinetic properties of PCBs, which
she predicted will be an increasingly important area of research as people learn more about risk
assessment in this area. Noting that much of the work in this area has in the past focused on
complex industrial mixtures of PCBs, Dr. Birnbaum suggested that there will continue to be
significant gaps in our understanding of this whole class of compounds until we are able to
identify those mechanisms of toxicity that are specific to a single congener or group of congeners.

       To begin, Dr. Bimbaum discussed the structural similarities and dissimilarities that exist
among the various PCB congeners. What these molecules all have in common, she noted, are
two benzene rings linked by a carbon-carbon bridge. Because of this double-ring structure,
halogenation at any or all of 10 possible sites has the potential to produce 209 possible  PCB
congeners.  Depending on the pattern of halogenation, PCBs may be symmetrical or
asymmetrical  molecules.  In the literature,  halogenation at the 2 or 6 position of either  ring is
referred to as ortho substitution; halogenation at the 3 or 5 position as meta substitution; and
halogenation at the  4 or 4' position is referred to as para substitution.  In most cases, chlorine is
the halogen of concern, although there have been environmental episodes  involving
polybrominated biphenyls (PBBs) as well.  In certain situations, iodinated compounds exhibiting
properties similar to PCBs and PBBs have also been observed.

       Different patterns of halogen substitution are known to affect the three-dimensional
structure of PCBs in ways that also affect their biological activity. Coplanar PCBs are those in
which halogenation leaves the molecule relatively flat; these are the compounds that are
generally dioxin-like in their activity. Other patterns of halogenation, particularly those involving
substitutions at the ortho positions, tend to twist and bend the molecule, diminishing or
eradicating entirely its dioxin-like activity.

       Most PCBs in the environment originated as components of industrial mixtures that were
produced and used in many different parts of the world prior to 1977, when these substances
were banned  in the U.S. Dr. Birnbaum noted that, of the 209 possible PCB congeners, there are
about 100 that are routinely present in industrial mixtures and about 60 that are commonly
found in the environment.  It is important to keep in mind, however, that the mixtures of. PCBs
found in the environment and in humans are not equivalent to any of the industrial mixtures that
initially entered the environment.  This is mainly due to the extensive alteration of these
compounds that takes place as they pass through the environment and eventually into
human bodies.

       The names and numbering systems that are used to refer to various industrial mixtures of
PCBs reflect both the site of production and the degree of chlorination of biphenyl compounds
in these mixtures.  Commercial mixtures produced in the U.S. are known as Aroclors,  for
example, while  those produced in France, Germany, and Japan are referred to as Phenoclors,
Clophens, and Kanechlors, respectively.  Within each group of compounds, the numbering  of
individual mixtures reflects the percentage of the molecular weight that is  due to chlorine.  In
Aroclor 1254, for example, 54% of the molecular weight is  due to chlorine, while in Aroclor
1248, chlorine accounts for 48% of the molecular weight. In general, therefore, the higher the
number of the mixture in any given series, the greater the percentage of highly chlorinated
congeners present in the mixture.  This is an important consideration, since the degree of
chlorination determines many of the physical properties of these chemicals. The less chlorinated,
lower molecular weight PCBs tend to be clear,  mobile oils,  for example, while increasing
chlorination produces compounds  that are gradually more discolored; sticky, and immobile. At
the highest levels  of chlorination, these compounds form viscous resins and may even solidify.

       Dr. Birnbaum pointed out  that the chemical and physical properties of PCBs also affect
their phannacokinetic profiles by influencing the ways in which these chemicals are absorbed,
distributed, metabolized, and excreted from the body. Absorption of PCBs is determined mainly
by the relative lipophilicity (i.e., fat solubility) of various congeners.  As a class, PCBs are highly
lipophilic compounds; this feature enhances their absorption into the body, since absorption
depends in large measure on the ability of a chemical to cross fatty membranes.  Among PCBs,
the less chlorinated compounds are the most readily absorbed. With increasing chlorination,
these compounds  may become  so lipophilic that they actually become "stuck" in the fatty
membranes of the body. At the same time, the tendency toward increased viscosity and even
solidification means that highly chlorinated PCBs do not readily enter solution, which is another
prerequisite for absorption.

       Because of their relatively large size and high lipophilicity, PCBs appear to be more
readily absorbed by oral than by dermal routes of exposure.  Noting that little work examining
the absorption of PCBs via the inhalation route has been undertaken to date, Dr. Birnbaum
speculated that PCBs would probably be absorbed relatively well across the lung parenchyma if,

 for example, one were to inhale paniculate matter upon which PCBs were present.  She also
 noted that relative rates of absorption can be somewhat deceptive, since with chronic exposure
 the absorption of even small amounts of these chemicals will eventually produce a significant
 body burden.

        Following their absorption into the bloodstream, PCBs are distributed throughout the
 body by carrier molecules in the blood. Unlike many other blood-borne substances, which rely
 on albumin as a  carrier,  PCBs are carried through the blood mainly by lipoprotein molecules.
 Because of their high affinity for a fatty environment, however, PCBs rapidly partition out of the
 blood and into body tissues. As a result of this rapid partitioning, the concentration of highly
 chlorinated PCBs can be hundreds of times higher in fatty tissues than in the blood, where they
 are usually present only  in very small quantities.  .             '

     .   With -chronic  exposure, PCBs tend to become most concentrated in the fatty tissues of
 the body; the major storage depots for these compounds, in decreasing order of importance, are
 the adipose tissue, liver,  and skin. The distribution of PCBs is modulated somewhat by body
 composition, in the sense that an organism with a higher percent body fat will have a greater
 ability to store PCBs  than an organism With less body fat.  At the same time, changes in the  fat
 composition of the body can produce a redistribution of these compounds. Studies have  shown,
 for example, that animals fed weight-reducing diets tend to mobilize their fat stores, at the same
 time mobilizing PCBs that are stored in their fatty tissues.  As a result of this  mobilization,' PCB
 molecules  re-enter the bloodstream and are carried to the liver; where they may undergo
 metabolism and be eliminated from the body. Similarly, the mobilization  of lipids that occurs
 during lactation can result in high concentrations of PCBs being excreted  from the body in
 breast milk.

       Elimination of PCBs from the body occurs by a combination of both'metabolism and
 excretion.  Although some passive elimination of PCBs occurs, the main routes of excretion are
 dependent upon metabolic changes that increase the  water solubility of these compounds.
 Passive elimination of PCBs does occur, however, both as a result of diffusion across the  gut wall
 and as a result of the partitioning of these compounds into sebum, sweat,  and  other bodily
 secretions. Passive elimination mechanisms also  account for the sex-linked routes of PCB
 excretion, including their elimination in mammalian breast milk and in the fat-rich portion of the
 eggs of fish, birds, reptiles,  and amphibians.  For active elimination processes,  molecular weight is
 the main determinant of the route of elimination; smaller molecules tend  to be eliminated in the
 urine, while larger molecules are eliminated mainly via the bile into the feces.  Urinary sampling
 is generally not a good method for assessing exposure to PCBs, since more highly substituted
 molecules and those that undergo conjugation reactions in the liver are far more likely to be
 eliminated  via the fecal route.

       Like many of their physical and chemical properties, the metabolism of PCBs is governed
by both the number and position of halogens on the biphenyl ring. The enzymes responsible  for
metabolizing PCBs are most effective  in degrading molecules in which there are two  adjacent
unsubstituted carbon atoms. In general, therefore, less chlorinated compounds are metabolized
much more rapidly than compounds with a greater number of chlorine  substituents.  For similar
reasons, compounds in which the chlorine substituents are localized to a single ring are generally
more readily metabolized than those with chlorine substituents on both rings.  Because
metabolism is the rate-limiting step in the elimination of PCBs, variations  in the susceptibility of

 PCBs to metabolic enzymes produces a wide range of half-lives for these compounds. In
 rodents, for example, some hexachlorobiphenyls (i.e., those with adjacent unsubstituted carbon
 atoms)  have half-lives as short as. 12 hours, while others (i.e., those without.adjacent
 unsubstituted carbon atoms) have half-lives that are longer than the lifetime of the rodent.
 Hexachlorobiphenyls that are very difficult to metabolize include 2,4,5,2',4',5'-hexachlorobiphenyl,
 which accounts for about 60% of the body burden of PCBs in humans.

        In terms of the specific metabolic reactions involved in the degradation and elimination
 of PCBs, Dr. Birnbaum pointed out that the most important step is the initial  oxidative
 metabolism of these compounds by cytochrome P450 enzymes.  This oxidative  reaction usually
 leads either to the incorporation of a hydroxyl group into the PCB structure or to the formation
 of an epoxide intermediate. The phenolic and epoxide intermediates, in turn,  can then undergo
 various other types of metabolic transformations  that promote their removal from the body.
 Conjugation of the phenolic chemical with glucuronic acid or sulfates, for example, produces
 metabolites that are readily excreted from the body, while conjugation of the epoxide with
 glutathione produces a whole series of sulfur-containing metabolites that may  go on to become
 concentrated in the lungs.  Other metabolic transformations that PCBs can undergo in the body
 include dechlorination reactions and direct insertion of hydroxyl groups.  In recent years,
 microbial species capable of degrading PCBs have also been identified.

        Metabolic disposition of PCBs is further complicated by the presence of many different
 types of cytochrome P450 enzymes, many of which have different patterns of substrate specificity.
 Dr. Birnbaum distinguished among two main types of P450 enzymes involved in the oxidation of
 PCBs as follows:

        •     the CYP1A family, which are induced by dioxin and related compounds and which
              tend to preferentially  oxidize the 2 or 2' ortho position; and

        •     the CYP2B  family, which are induced by phenobarbital  and  related compounds
              and which tend to preferentially oxidize the 4 or 4' para position.

 Noting  that the 3 or 3' meta position is the site at which direct insertion of a hydroxyl group is
 most likely to occur, Dr. Birnbaum said that it is not yet clear which P450 enzyme or group of
 enzymes is responsible for this type of oxidative reaction.

        In addition,  there appear to be distinct differences in the  ability of various species to
 metabolize PCBs.  In general, mammals are more proficient in metabolizing these compounds
 than are birds, which are more proficient than fish.  Among mammals, metabolic efficiency in
 eliminating PCBs is  roughly equivalent among humans, rodents,  and rats.  Dogs are by far the
 most proficient mammals studied in terms of their ability to eliminate PCBs, even the highly
 chlorinated congeners such as 2,4,5,2',4',5'-hexachlorobiphenyl that tend to  persist for long
 periods of time in other mammalian species.  The importance of these comparisons, Dr.
 Birnbaum noted, is  that the half-lives of these  chemicals in the body is inversely related to the
 ease with which they are metabolized, since chemicals that are not readily metabolized will
 persist in the body.   In response to a question  from one of the participants, Dr. Birnbaum
 indicated that the half-lives of various congeners  present after exposure to  Aroclor 1260 have
 been estimated to range between one and twelve years, although she said that  she believes these
. estimates are based  on relatively weak data. The important point, she suggested, is that the total

 body burden of PCBs generally increases with age, due mainly to the fact that these chemicals
 persist for such long periods of time in our bodies.

        As our understanding of the pharmacokinetics  of these compounds has increased, it has
 been possible to begin to put together physiologically-based pharmacokinetic models for PCBs.
 The advantage of such models, Dr. Birnbaum noted, is that they have the potential to offer some
 predictive power in extrapolating from one kind of exposure scenario or dose situation to
 another, or from one species to another. In attempting to develop these models, she said,  people
 have begun to look at different variables in order to determine which are chemical-dependent,
 which are species-dependent, and which are neither chemical- nor species-dependent.
 Physiologic parameters such as blood flow rate and organ volume are generally not
 chemical-dependent, for  example, although they are certainly species-dependent.  Partition
 coefficients, on the other hand, are clearly chemical-dependent, since they depend so heavily on
 the lipid solubility of the chemical of interest. Metabolic parameters are likely to be chemical-,
 species-, and even tissue-dependent, since they will vary depending on the number and type of
 P450 enzymes present. Dr.  Birnbaum noted that, in order for reliable physiologically-based
 models of PCB behavior  to be developed, these and many other parameters have to be measured
 or extrapolated for inclusion in the model.

        Another factor that needs to be taken into account in the development of
 physiologically-based models of PCB pharmacokinetics, Dr. Birnbaum noted, is the fact that what
 we are attempting to model is  a highly dynamic system in which exposure to PCBs has the
 potential to change many of the relevant parameters.  It is known, for example, that PCBs act to
 induce their own metabolism, which means that the pharmacokinetics of chronic exposure may
 differ in very important ways from the pharmacokinetics of acute exposure to these compounds.
 In addition, it is also known that exposure to PCBs can alter the body's metabolic handling  of
 various endogenous compounds, including thyroid hormones, vitamin A, steroids and fatty acids.
 As a result of these changes, organisms that have been exposed to PCBs differ from naive
 organisms in many pharmacokinetically important ways.

       Dr. Birnbaum concluded her presentation with a brief review of the state of knowledge
 about the pharmacokinetics of PCBs as they relate to the potential developmental neurotoxicity
 of these compounds.  To  fully understand the pharmacokinetic properties of PCBs in the
 developing organism and  neonate, it is also important to understand the mechanisms by which
 PCBs are  transferred from the mother to the offspring. It is known, for example, that the
 mobilization of fat resources that occurs  during the process of lactation can produce a 30-50%
 decrease in the mother's total body burden of PCBs.  In comparison, movement of PCBs  across
 the placenta is thought to be very limited, although even a small transplacental gradient can
 represent significant exposure due to the very small size of the embryo or fetus.  In addition, it is
 possible that there may one or  more "windows" of enhanced sensitivity to these compounds  in
 association with particular developmental periods. In terms of the ability of the fetus to
 metabolize PCBs, little is  known,  except that the capacity to metabolize these compounds
appears relatively  limited  until  fairly late in development.  In response to a question from a
participant, Dr. Birnbaum noted that we know almost nothing about whether important PCB
metabolites are able to traverse the placental membranes.

       Noting that there are significant gaps in our understanding of the pharmacokinetic
properties of PCBs that relate to their potential as developmental neurotoxicants, Dr. Birnbaum
identified four critical questions in this area as follows:

       •      Which PCBs are developmentally neurotoxic?

       •      Is there a critical window of sensitivity during which the fetus or neonate is at
              increased risk, either to the effects of PCBs in general or to the effects of specific
              congeners or groups of congeners?

       •      What are the pharmacokinetic profiles of the critical PCBs, alone and in
              combination with other congeners and metabolites?

       •      How can we integrate the information that is currently available regarding the
              toxicity of these compounds to develop physiologically-based models that offer
              predictive value in the developmental system?

       Following Dr. Birnbaum's formal presentation, there was some discussion among
participants of issues related to the developmental pharmacokinetics of PCBs.  One participant
pointed out that the work of Barsotti  and colleagues has suggested that, at least in primates, the
mix of congeners seen in the infant at birth can be very  dissimilar to the mix of congeners seen
in the mother's blood.  In general, this participant noted, less chlorinated PCBs tended to be
present in neonatal blood samples in proportions higher than one would predict based on the
mix of compounds present in the mother's blood. Dr. Birnbaum agreed that these data  pose
interesting questions about the persistence of the lower  chlorinated compounds in this setting,
given the relative ease with which these compounds should  theoretically be metabolized. One
possibility, she suggested, is that the PCBs in question may be bound to proteins that interfere
with their metabolism, although she noted that neither this nor other mechanisms that might
explain this persistence phenomenon have yet been explored in any detail.

       Another participant pointed out that much  of what is known about the pharmacokinetics
of PCBs is based either on exposure to commercial mixtures or on exposure to a single synthetic
congener. This person suggested that neither of these situations  is truly representative of the
majority of human exposures, which generally involve exposure to combinations of congeners that
have been filtered through the  environment  and through the food chain.  Because of this, most
of the compounds with short half-lives may already have been metabolized by the time human
exposure occurs.

       The same participant questioned the extent to which lipid mobilization is a factor in the
activation of PCBs stored in the body, noting that there  is no evidence that even dramatic weight
reduction can reduce the total body burden of PCBs in humans.  A more useful construct, this
participant suggested, is to think about PCB distribution in terms of an equilibration by solubility
parameters, in which the amount of PCBs in the blood is simply proportional to the amount of
acetone-soluble material present.  Dr. Birnbaum noted that there has been some recent  success
in using a cholestyramine and rice bran regimen to reduce the body burden of PCBs in humans,
but she agreed that the human  data in this area are much less compelling than the studies that
have been done in rodents.

        Dr. Birnbaum disagreed that solubility parameters alone are adequate to describe the
 partitioning of PCBs between the blood and tissues, however, noting that the ability of these
 compounds to undergo partitioning is complicated by the existence of binding proteins. This is
 particularly true in the case of the dioxin-like PCBs, since those chemicals are known to induce
 production of their own binding proteins.

        James McKinney, Health Effects Research Laboratory, EPA

        Dr. McKinney began his talk by noting that structure-activity researchers are interested
 less in the structural features of ?CBsperse than in the ways in which these structural features
 might account for the reactivities that underlie the toxicity of these chemicals. Noting that there
 are only a limited number of reactivities that these chemicals can have, he suggested that  looking
 at three of the most important types of reactivity might provide important insights into the
 mechanisms of toxicity of the various PCB congeners.  These three types of reactivity are:

        »      stacking interactions, which have to do with the coplanarity associated with certain
               types of dioxin-like activity;

        »      polarizability interactions, which have to do with the importance of lateral  and
               non-lateral patterns of halogenation; and

        •      oxidative metabolic reactions, which have to do with the  production of metabolites
               that may  or may not have toxic properties of their own.

       The stacking interaction appears  to be the form of reactivity that is most relevant to
 binding at the aromatic hydrocarbon  (Ah) receptor.  Dr. McKinney explained that the stacking
 interaction model is based on separation distances determined using energy minimization
 calculations. In diagrams and in computer-generated graphics depicting the van  der Waals
 surfaces of coplanar and non-coplanar PCBs, the dioxin molecule appears to stack more tightly
 with the Ah receptor than does a coplanar PCB molecule, which in turn stacks more tightly than
 an ortho-substituted, non-coplanar PCB molecule. The tightness of the  stacking  interaction (a
 function of both steric and electronic factors), therefore, appears to be related to the likelihood
 of a molecule's producing dioxin-like effects via interactions with the Ah receptor. Dr.
 McKinney also showed data indicating that the Ah receptor binding constants (expressed as a
 ratio) predicted for a series of PCBs using the computational chemistry  approach correlates very
 closely with experimentally-determined EC50 values for these same compounds. This correlation
 suggests that the model does a pretty good job of predicting the Ah receptor-binding properties
 of PCBs on the basis of their relative reactivities in terms  of stacking interactions.

       Since most PCBs in  the environment are ortho-substituted and since ortho substitution
 has been implicated in neurotoxicity, an important question concerns the implications of this
 substitution pattern for PCB reactivity. Dr. McKinney noted that the tendency for ortho
 substitution to interfere with the stacking interaction is interesting in a number of different ways.
 Structurally, ortho substitution restricts rotation around the pivot bond of the molecule, making
 it more difficult for the molecule to assume a coplanar shape and introducing a more rigid,
steroid-like configuration. In its most extreme form (i.e., in 2,4,6,2',4',6'-hexachlorobiphenyl)

ortho substitution can cause the phenyl rings to become locked essentially at right angles to one
another.  Functionally, because these conformational changes reduce the ability of the molecule
to bind to the Ah receptor, ortho substitution diminishes the dioxin-like activity of the
non-coplanar PCB congeners.  Since ortho substitution appears to be positively correlated with
the neurotoxic properties of PCBs, it is suggested that the non-coplanarity of ortho-substituted
congeners may play a role in the neurotoxicity of these compounds.

       Having described the relationship between stacking interactions and relative affinity for
the Ah receptor, Dr. McKinney turned his attention to the second major type of PCB reactivity,
the polarization interaction. He noted that the longest polarization vector in the biphenyl
molecule is the one that passes through the 4 and 4' positions, since halogenation at either of
these two positions would be expected to produce the most dramatic, effect on molecular
polarizability.  To a lesser extent, chlorine substitution at the lateral (meta) positions would also
be expected to produce significant polarization of the molecule.  Chlorine substitution at the
ortho positions would be expected to have the least significant effect on polarization of the
biphenyl molecule. .Polarization appears to have important functional consequences, since it is
related not only to the binding properties of these compounds (ring stacking as well as chlorine
interactions), but also to physicochemical characteristics such as their relative lipid solubilities.

       To explore the structure-activity implications of PCB chlorine polarization,
Dr. McKinney and colleagues  developed a model system to predict  patterns of PCB binding to
prealbumin, a protein that is known to be involved in the transport  of thyroid hormones to the
central nervous system. Because prealbumin has been studied extensively using x-ray
crystallographic techniques, a great deal is known about its two equivalent thyroid hormone
binding sites. Dr. McKinney noted that there is a very snug, "hand-in-glove" match between the
shape of the thyroid hormone  molecule and the shape of the channel that serves as the
prealbumin  binding site. The  hormone is literally engulfed by the prealbumin  protein, with the
phenolic ring and lateral substituents of the hormone fitting tightly into "pockets"  in the binding
domain of the prealbumin  molecule.

       Dr. McKinney then showed a series of slides in which the same computer  techniques
were used to simulate the binding of various PCB congeners to the  prealbumin molecule. In
general, the more polarized PCB molecules (i.e., those with chlorine substituents in the meta
and/or para positions)  tend to fit very snugly into the binding site, while those with ortho
substitutions bind to prealbumin much more loosely, if indeed they  bind,at all.  More specifically,
Dr. McKinney listed the following as results of SAR studies examining patterns of PCB binding
to prealbumin:                                                                  \

       •      The best-binding PCBs are those that have only lateral chlorine substituents (e.g.,
              the 3,5- and 3,4,5- patterns of substitution); these  molecules bind to prealbumin
              with an affinity that is four to eight times higher than T4, the  endogenous  ligand.

       •      Ortho substitution does not interfere with binding much, as long as lateral
              chlorine substituents are present (e.g., the 2,3,4,5- pattern of substitution); these
              molecules bind to prealbumin with an affinity that is two to three times higher
              than the endogenous ligand.                         ,               >

         »      The unsubstituted biphenyl molecule and the fully ortho-substituted molecule (i.e.,
                the 2,2',6,6'- pattern of substitution) show no binding activity at all; the addition
                of a single lateral chlorine substituent (e.g., the 2,4,6- pattern of substitution),
                however, restores binding to  a level of about half that seen with the endogenous

        Noting that lateral substitution is generally thought to be associated with the persistence
 of PCBs in the tissues, Dr. McKinney said that it seems reasonable to predict that the most
 persistent PCBs are also likely to be those that bind most avidly to the prealbumin molecule.

        Dr. McKinney also showed data comparing the prealbumin binding of various PCB
 congeners with their binding to a "prealbumin-like" nuclear binding protein in rat liver tissue.
 With few exceptions, patterns  of binding to the nuclear protein were very similar to patterns of
 prealbumin binding. In fact, the affinity of the laterally substituted PCBs that could also be
 coplanar was even greater for the nuclear binding protein  than the affinity of these same
 compounds for the prealbumin molecule.  This similarity led Dr. McKinney and colleagues to
 predict that stacking and polarization interactions might explain a good deal of the behavior of
 thyroid hormone in binding to nuclear receptors.  Their suspicion appears  to have been
 confirmed by experiments in which thyroid hormone binding constants predicted by a simple
 stacking model were found to  correlate  closely with published receptor binding data for these
 same hormones.

       Based on the similarities  in their binding affinities for both prealbumin  and for nuclear
 receptor proteins, Dr. McKinney and colleagues have also proposed that certain PCBs and
 thyroid hormones may have similar kinds of  molecular  recognition properties.  In addition to the
 stackable rings and sites of potential lateral substitution that these compounds have in common,
 it is clear that they share some of the same binding proteins. In fact, activity differences among
 PCBs appear  to correlate in some ways with  differences in the activity  of the T3 and T4 thyroid
 hormones. PCBs that exhibit significant stacking reactivities seem more like T3 in their binding
 affinities, while highly polarized,  laterally substituted PCBs behave more like T4.  In addition, as
 one might expect, the dioxin receptor is  of the T3 variety.  Although he acknowledged that little is
 currently known about structure-activity  relationships between PCBs and retinoids, Dr. McKinney
 suggested that it will  be interesting to explore whether similar binding considerations underlie
 functional relationships among the PCBs, thyroid hormones, and vitamin A metabolism.
 Another interesting area of exploration might be the extent to which different types-of PCBs
 function  as thyroid hormone agonists and/or  antagonists, resulting in qualitatively
 different toxicities.

       The third and final  type of PCB reactivity that Dr. McKinney addressed in his talk  was
 the potential for oxidative metabolism to produce PCB metabolites with significant toxic
 properties of their own.  He noted, for example, that a series of hydroxylated compounds, all of
which represent potential metabolites of the less chlorinated PCBs, have been demonstrated to
be capable of binding to the estrogen receptor and therefore capable of exerting estrogenic kinds
of activity.  The PCBs with the  greatest estrogen-binding activity are those with ortho
substitutions, which Dr. McKinney had previously described as conferring an almost steroid-like
rigidity on the PCB molecule.  Thus, in addition to having the potential to act as thyroid

hormone agonists or antagonists, some PCBs may also have the potential to act as agonists or
antagonists in the estradiol system.

       Dr. McKinney concluded his presentation by discussing some of the potential implications
of this structure-activity information for understanding the developmental neurotoxicity of
various types of PCBs. He noted, for example, that in the work of Seegal and colleagues, the
patterns of chlorine substitution in compounds exhibiting neurotoxic effects are almost identical
to the patterns  of chlorination associated with increasing affinity for the prealbumin binding site
(the main exception being light chlorination). Dr. McKinney said that this would suggest to him
that neurotoxic effects of PCBs might depend mainly on the degree of lateral halogenation of the
various congeners, particularly those that also have ortho chlorine substituents. Because of their
ortho substituents, these compounds might be steered away from stacking interactions with the
Ah receptor, leaving them available for binding to prealbumin or structurally similar molecules.
Heavily chlorinated PCBs with some ortho substitution might still compete well for the Ah
receptor or related proteins.

       A  final  observation that Dr: McKinney offered was a working hypothesis regarding
additional implications of lateral chlorination that might be.worthy of further study. If it is
correct that laterally substituted PCBs function as T4 analogs, these compounds could also be
expected to antagonize the effects of T4 through competitive  binding interactions.  To Dr.
McKinney, this suggests that it is at least theoretically possible for the presence of PCBs to
produce what he referred to as chemically-induced hypothyroidism at the cellular level.  At the
same time, competitive interactions between PCB molecules  and T4 could disrupt a whole range
of processes that involve T4 binding, including deiodinase reactions, tyrosine hydroxylase
reactions, and dopamine receptor regulation. Further study will be needed to establish direct
linkages between PCB "laterality" and various types of toxicity.

       Following Dr. McKinney's presentation, there was some discussion among workshop
participants regarding various aspects of the data he presented. One participant asked whether
binding to prealbumin is affected by the structure of the molecule at the end opposite  the site of
lateral substitution. Dr, McKinney responded that the trailing end of the molecule appears to be
relatively-unimportant for binding, noting that even a halogenated phenol will bind reasonably
well to prealbumin.

       Another participant asked Dr. McKinney to respond to an issue that one of the
pre-meeting commenters had raised in pointing out that the molecules responsible  for thyroid
hormone  transport vary from species to species. - Dr. McKinney said that it is his impression that
such differences are mainly in the structure  and function of thyroxine-binding globulin,  which
serves as  a reservoir from which thyroid hormone can be mobilized in emergency situations.  He
said that there  appear to be fewer species-related differences in prealbumin, which he believes is
the real "workhorse" protein in terms of routine transport of thyroid hormone to the cells,
although he noted that his understanding of this process may not be correct. The individual  who
had made the original pre-meeting comment was not able to add any further information to
this discussion.

         Gary L. Kimmel, Reproductive and Developmental Toxicology Branch, EPA

         Dr. Kimmel began by noting that, unlike the previous speakers, who had addressed the
 question of PCB toxicity from a data collection standpoint, his presentation would focus on how
 the Agency uses scientific data in order to assess the risks associated with exposure to a
 particular compound or class of compounds.  In addition to describing the risk assessment
 process itself, Dr. Kimmel indicated that he would also spend a fair amount of time describing
 the ways in which the Agency goes about determining whether a given data base is or is not
 sufficient to support the development of a reference dose or some other kind of quantitative
 analysis of risk.

        Dr. Kimmel said that the Agency's general approach to risk assessment is based on a
 paradigm originally set forth by the  National Research Council (NRC) in 1983.  During the
 process of developing risk assessment guidelines for developmental toxicity, it became
 increasingly clear that some  modification to the NRC approach would be required in order to
 adequately address the risks  associated  with compounds that act on various non-cancer
 endpoints. The modifications initiated  as a part of the developmental toxicity guidelines have, in
 fact, turned out to be applicable to most other non-cancer endpoints as well.

        The first step in the risk assessment process is to pull together all available information
 related to the identification of hazards associated with the chemical of interest, as well as any
 dose-response data that may be available for the chemical. As  part of this step,  the database
 itself is characterized in terms of its ability to  serve as a basis for a scientifically  meaningful
 assessment of risk.  The strength and completeness of available data are major points of concern
 at this stage of the assessment. If it is determined that the available data are sufficient, an  effort
 is then  made to quantify the  risks associated with the chemical of concern.  Traditionally, this has
 involved the calculation of a  reference dose, although Dr. Kimmel noted that the Agency has
 recently begun to move toward other types of quantitative expressions of risk, such as benchmark
 doses and biologically-based  dose-response models.

        Once the evaluation of available data has been completed, the risk estimate is combined
 with the results of an exposure  assessment, which attempts to describe,  again quantitatively, levels
 of actual or presumed human exposure to the  chemical under study.  The process culminates in
 risk characterization, which attempts to  integrate information about  the potential hazards of a
 chemical with information about levels of exposure that are likely to occur in the environment or
 in specific settings of concern. This risk characterization, in turn, becomes one of the factors
 that a risk manager considers in deciding whether and how to deal with a potentially toxic
chemical. Other factors that  may influence a risk manager's decision at this point in the process
include  the social and legal implications  of various  risk management alternatives.

       In terms of developmental toxicity per se, there are four basic types of adverse effects that
the Agency considers manifestations of developmental toxicity. These include:

       •      prenatal or early postnatal death;

       •      structural abnormalities;

       •      growth alterations (whether manifested in the animal as a whole or in the growth
              of particular organs); and                                                  v

       »      functional deficits (which can include behavioral alterations and/or abnormal
              organ system function).

An important factor in distinguishing developmental effects from other forms of toxicity is the
period during which the exposure occurs, regardless of the time at which the effects of exposure
appear. For an adverse effect to be classified as developmental, exposure may have occurred any
time prior to conception, during prenatal development, or postnatally  until the time of sexual
maturation.  Exposures that occur prior to conception may involve either parent.  Adverse effects
of the exposure may become apparent at any time during the lifespan of the organism; an  effect
is classified as developmental based on the time of exposure rather than the time at which the
effects of exposure become apparent.

       Following his discussion of the Agency's definition of developmental toxicity, Dr. Kimmel
noted that there is also a very distinct set of assumptions that the Agency uses in evaluating a
developmental toxicity database.  These assumptions include the following:

       •      An adverse effect observed  in an animal is assumed to indicate some potential
              risk for humans.

       •      All manifestations of developmental toxicity are of concern (i.e., no assumption is
              made  that one type of developmental effect is of greater  concern than another).

       •      The types of effects seen in animal studies are not assumed to be identical with
              effects of the chemical in humans.  That is, although adverse effects in animals
              are accepted as evidence of a potential risk for humans, it is not assumed that the
              specific effect in humans will be the same as the specific  effect observed in  the

       •      The most appropriate or sensitive animal  species is used to estimate  risk in
              humans.  Where human data are available, they should be used. Where human
              data are not available, selection of a most appropriate  animal species should be
              made  on the basis of metabolic similarities,  similarities in reproductive function,
              and so on. In  cases in which it is not clear which animal species is most
              appropriate, the species that is most sensitive to the effects of the chemical should
              be used.

       •      It is generally assumed that dose-response curves for non-cancer endpoints
              include a threshold below which the adverse effect can be expected not to occur
              as a result of exposure to the chemical.

       Dr. Kimmel then described some of the factors that go into the Agency's consideration of
the strength of the database during the hazard identification and dose-response phases of the risk
assessment process. Many of these factors have to do with study design; such factors include the
resolving power of the available studies, the relevance of routes  and timing of exposure
considered, the appropriateness of the doses selected, the number and types of endpoints

  studied, the timing of the examination, and the number of species tested.  Other considerations
  that might influence a conclusion about the strength of the available database include whether
  and to what extent observed effects have been replicated and/or have been supported by the
  results of pharmacokinetic, mechanistic, or structure-activity studies.

        For the benefit of participants less familiar with human toxicdlogic studies, Dr. Kimmel
  described some of the strengths and weaknesses associated with various types of observational
  epidemiologic designs. In cohort studies, for example, a group that is defined on the basis of
  exposure is followed over time.  Cohort studies allow for the possibility of examining multiple
  health effects, but these types of studies are generally very expensive to conduct. In case-control
  studies, on the other hand, the  study group is defined by a selected health effect and the
  exposure histories of individuals exhibiting that effect are  reconstructed. Because the focus of
  the study is limited to a single effect, case-control studies are often used'to investigate relatively
  rare outcomes and tend to be less expensive than cohort studies. A third type of epidemiologic
 study known as an ecologic study takes a more general approach.  Ecologic studies define a
 particular segment of the population on the basis of a general  characteristic such as  zip code or
 distance  from a specified site; then, researchers look for any obvious trends in the incidence of
 adverse effects among segments of the selected population. Although ecologic studies do not
 have the power of cohort or case-control studies, they provide  a relatively inexpensive and quick
 way of getting rough information about a population of interest. In this sense, ecologic studies
 arc useful mainly for purposes of generating hypotheses that can then be tested  in other, more
 powerful types of studies.

        In evaluating epidemiologic studies, Dr. Kimmel noted  that there are a number of
 potential sources of bias that must be taken into account.  It is important, for example, to avoid
 selection bias in assigning specific individuals to the exposed or control population. Similarly, it
 is important to control for information bias, in which an individual who is providing information
 to the researcher may be biasing his or her responses for one reason or another. Statistically, it
 is important to consider whether the study is capable of detecting a true effect and to what
 extent the outcome of the study depends on factors such as population size or the level of risk'
 involved. Finally, the results of these studies must be adjusted  to account for factors called
 confounders and effect modifiers, which can affect the results of the study simply because they
 are themselves interrelated.

       Having discussed in some detail the various^ factors used to evaluate the strength of
 human toxicity studies, Dr. Kimmel then described some of the standard protocols used to
 conduct developmental toxicity studies in animals. A standard Segment II study  design, for
 example,  begins with a timed mating. Exposure begins at the beginning of organogenesis, which
 occurs in  the rat at about the same time as implantation. Exposure continues throughout the
 period of organogenesis. Just prior to birth, pregnancy outcome and maternal and fetal
 endpoints are evaluated. Dr. Kimmel noted that this type of study design has been used for
 many years and has provided a great deal of information, including a wealth of control data and  -
 a general  understanding of developmentally-sensitive endpoints. He also noted, however, that it
 is important to keep in mind the limitations of this and similar  study designs. As an example,
 Dr. Kimmel pointed out that researchers frequently and erroneously compare the second
 trimester of development in the  rat with the second trimester of development in  the human fetus.
In actuality, everything that occurs over the first two trimesters of development in the rat occurs
well within the first trimester of  human gestation.  Such differences can be extremely  important

when one is attempting to look at the effects of a chemical on the development of specific organ
systems, including the nervous system.  In this regard, even defining those effects that constitute
endpoints of developmental toxicity is likely to provide a significant challenge for work groups
looking at both the human and the animal data in this area.

       In terms of developmental neurotoxicityper.se, Dr. Kimmel related a series of conclusions
reached nearly three years ago by a workshop on risk assessment in the area of developmental
neurotoxicity.  Among the conclusions reached by this workshop were the following:

      ' »      Cross-species comparability for qualitative effects is strongly supported. That is,
              despite the fact that the definition of developmental toxicity does not require a
              one-to-one correspondence of effects, it is usually the case that developmental
              neurotoxic effects are similar from one species to another.

       •      Since no single category of function was found to be routinely the most sensitive,
              a battery of functions should be included in a developmental neurotoxicity testing

       •      Any reliable effect in animals should be considered  an adverse effect for
              extrapolating to humans.                                             ,

       •      Permanent as well as transient effects should be considered adverse for
              extrapolation to humans.

       »      Developmental effects in the presence or absence of maternal  toxicity should be
              considered adverse.

Based in part  on these conclusions,  the workshop on developmental neurotoxicity agreed  that,
despite its limitations,  the RfD approach was an appropriate one for determining a level below
which no increase in developmental neurotoxicity is expected. The workshop  also concluded,
however, that  measures of target organ dosage would probably be useful in reducing the
uncertainties associated with cross-species extrapolation.
  . ;•            •
       To conclude his presentation, Dr. Kimmel briefly reviewed  the criteria for evaluating a
health-related database as set out in the December 1991 EPA risk assessment guidelines  for
developmental toxicity. He noted that the document specifies that the characterization should be
carried out on the database as a whole, rather than focusing on a single study or group of studies
that address a particular effect or level of exposure.  The document also provides guidance for
judging a potential developmental hazard within the context of the expected human exposure or
dose of the compound. From the standpoint of these criteria, there are three categories into
which the health effects database can potentially be judged to fall:

       •  "    sufficient human evidence In these cases, data from epidemiologic studies provide
              evidence that is convincing enough for the scientific community to judge that a
              causal relationship between  exposure and toxicity is or is not supported. Although
              single case studies or case reports  may provide supporting evidence for this ,
              conclusion,  they are not sufficient by themselves to establish an exposure as a
              particular human risk.      ,                         ,

        •     sufficient animal evidence In these cases, data from animal studies and/or limited
               human data provide convincing evidence upon which the scientific community can
               judge the potential  for developmental toxicity. Evidence of toxicity obtained from
               a single well-conducted animal study showing developmental toxicity is sufficient
               to carry out a risk assessment, while studies in at least two species and evaluating
               a variety of endpoints are necessary in order to judge that a compound does not
               pose a significant risk to humans.

        •     insufficient evidence  Examples of cases in which the evidence could be judged
               insufficient include situations in which there are simply no data available, data are
               obtained from studies of questionable design, data frorp a single species suggest
               no adverse  developmental effects, data are available only from short-term toxicity
               testing, and the only data available are from pharmacokinetic, metabolic, or
               structure-activity tests.

        Following Dr.  Kimmel's presentation, participants engaged in a discussion of some of the
 issues he had raised.  In response to a question from one participant regarding who decides what
 constitutes a well-conducted study, Dr. Kimmel noted that the developmental toxicity risk
 assessment guidelines offer a fair amount of guidance regarding the  number of animals that need
 to be used, the number of dose groups that need to be tested, and so on.  In response to a
 follow-up question from the same participant, however, Dr. Kimmel  noted that studies often do
 not  follow protocols from the testing guidelines per se, and that risk assessors often find
 themselves in a position of having to work with whatever data are available. Similarly, in vitro
 and structure-activity studies can represent an important complement to the available toxicity
 data, even though a risk assessment would not generally be based on the results of these types of
 studies alone.

        Another participant asked where infertility fits 'into the spectrum of developmental
 toxicity, given that the inability to produce a conceptus has clear developmental implications.
 Noting that infertility is treated as an example of reproductive rather than developmental toxicity,
 Dr. Kimmel acknowledged that the Agency has made a  somewhat arbitrary distinction in this
 area in order to deal with the data coming in. In the same sense, defining the end of
 development as puberty is also an arbitrary decision. Another participant  asked who would
 address a situation in which, for example,  the organism developed normally through puberty but
 then exhibited an effect such as premature aging.  Noting that developmental effects are defined
 on the basis of the period during  which exposure occurs, Dr. Kimmel said that this effect would
 be considered developmental if it were clearly linked to  an exposure that occurred before puberty,
 even though the effects of the exposure were not apparent  until after  the end of development.

       Following this discussion, observers were given an opportunity to present questions or
 comments on any of the material  discussed during the morning session  of the workshop.  One
 observer noted that the Agency appears to demand more from negative than from positive
 toxicity studies; this individual asked Dr. Kimmel whether any relative weighting is done based on
 an assessment of the power of the available studies. Dr. Kimmel responded that, ideally, every
study should be assessed for power.  It is a common misconception that risk management
decisions are usually based on studies submitted to the Agency specifically  for risk assessment

purposes, when in fact risk characterization more often has to rely on studies in the literature
that were designed for very different purposes.

       In a follow-up question, the same observer asked how conflicting evidence of toxicity is
dealt with in the risk assessment process  (e.g., if one study finds an effect to be significant and
another does not). Dr. Kimmel responded that this is one of the most difficult questions that the
Agency has to deal with, and that in most cases it is necessary to rely on the best available
scientific judgment. Wherever possible, it is usually also very instructive to  try to sort out the
basis of the differences in the results obtained, since strain or species differences in
responsiveness to a chemical often provide valuable information about mechanisms of the
chemical's  toxicity.

       Another observer asked Dr. Kimmel how the Agency deals with situations in which
exposure involves a number of chemicals with similar properties and similar effects, as appears to
be the case in Great Lakes area women who have PCBs, dioxins, and related compounds in their
breast milk. Dr. Kimmel noted that  in some cases (e.g., in evaluating the effects of dioxins  and
dibenzofurans) it has been possible to develop toxic equivalency factors that are useful for this
purpose. In general, the Agency currently treats exposure to complex mixtures on an additive
basis, unless there is some clear reason why additivity  should not be assumed. Another workshop
participant noted that this is one of the questions that is currently being addressed in the dioxin
reassessment, since there is  at least some evidence that interactions other than additivity may
also be important.

       Following this discussion, Dr. Birnbaum thanked workshop participants and observers  for
an interesting discussion of background issues related  to PCBs.  She reiterated that the overall
focus of the workshop would be to look at  the available data and to assess the adequacy of  the
database for risk assessment purposes. At the same time, it would be important for the group to
identify areas of needed research that could, strengthen the scientific base in this area.  Following
several logistical announcements, Dr. Birnbaum announced the time and place for each of the
afternoon work group meetings, and the  opening session of the workshop was adjourned.


                                   SECTION FOUR

                              DISCUSSION SESSION:
                             FOR RISK ASSESSMENT

                                  Chair: Elaine Z. Francis
                            U.S. Environmental Protection Agency
       Dr. Elaine Francis opened the second day of the workshop by reviewing the previous
 day's agenda, which she described as having provided the background for discussions scheduled
 to take place over the remainder of the meeting. During the morning of the second day, she
 noted, participants would be asked to review the available data on the developmental
 neurotoxicity of PCBs, with a view toward determining the adequacy of the data for risk
 assessment purposes.  Then, during the final session of the meeting, the group would be asked to
 identify gaps in the available data, wherever possible prioritizing the research needed to improve
 EPA's ability to conduct risk assessments in this area.

       Noting that the previous day had ended with the adjournment of the workshop into work
 groups charged with reviewing the strengths and weaknesses of the available human and animal
 experimental data, Dr. Francis asked two of the co-chairs to present brief summaries of the
 issues discussed in their respective groups.  Dr. Mary Prince presented a summary of the
 discussion that had taken place in the work group charged with reviewing the  available human
 data, and Dr. Carole Kimmel presented a summary of the discussion that had taken place in the
 animal data group, as well as a comparison of animal and human data. (See Section 6.1 of this
 report for the work group co-chairs' written summaries of these discussions.)
       Following these presentations, Dr. Francis opened the discussion to other workshop
 participants, noting that the purpose of the session was to assess the utility of the available data
 for risk assessment purposes.   Toward this end, she briefly reviewed criteria related to the
 sufficiency of available data that had been previously described by Dr. Gary Kimmel during his
. presentation on the risk assessment process.  She  noted that the adequacy of the data for hazard
 identification is an important consideration in this regard, as is the availability of data that can be
 used to establish the existence  of meaningful  dose-response relationships between PCBs and their
 putative effects.

       Much of the discussion during this session centered on the extent to which the results of
 experiments in animals treated with commercial mixtures of PCBs can be used to estimate the
 risk of developmental neurotoxicity to humans,  whose exposure is not to commercial mixtures,
 but  rather to combinations of congeners found in  the environment and/or in breast milk. One

 participant observed that, if a decision is made that commercial mixtures of PCBs are not
 particularly relevant to human exposures, it is not clear what mix of congeners might be more
 relevant. Another participant disagreed, noting that it is now possible to synthesize a "cocktail"
 of PCBs that is very similar to breast milk on a congener-by-congener basis.  The same
 participant noted that this mix of congeners looks like something between Aroclor 1254 and
 Aroclor 1260, except that most of the less chlorinated compounds are omitted from the mixture.

        Another participant noted that some preliminary work has also been done to characterize
 the mix of congeners present in the diet as part of the FDA's ongoing market basket studies.
 The same participant who had suggested the use of a  synthetic mixture questioned the
 assumption that diet is the source of greatest exposure to PCBs, noting, for example, that the
 body burden of these compounds is not quite doubled even in avid consumers of sport fish.  This
 same participant suggested that the dermal route may account for significantly more exposure to
 PCBs than is commonly thought, given the large volume of these compounds that still exists in
 sources that may contribute to dermal exposure, such  as industrial solvents and old electrical
 parts. Noting that PCB levels typically do not correlate with the  levels of classic bioconcentrators
 like DDE and other cyclodiene pesticides, this participant suggested that it is not at all certain
 that PCBs in humans are derived exclusively or even mainly from dietary sources.  A third
 participant argued that the analogy between PCBs and individual pesticides such as DDE is not a
 useful one, however, since the type and amount of PCBs in the environment has changed
 dramatically over the past 20 years.

        One of the co-chairs of the animal data work group expressed some surprise about
 concerns regarding the use of data from human exposure studies, since most of the discussion in
 the animal data work group had focused on the need for information more clearly relevant to
 environmental exposures.  She noted that, because animal studies have been conducted almost
 exclusively using the commercial mixtures, many participants in the animal data group had
 expressed some reservation about the use of this data  for risk assessment purposes.

        One participant thought it important to distinguish between the use of animal data for
 hazard  identification and its use in more quantitative assessments of risk. This participant
 suggested that the exposure of animals to commercial  mixtures of PCBs clearly offers much that
 is relevant to humans in terms of the hazard identification.  This person thought it less clear,
 however, whether or to what extent these data should be used to establish dose-response
 relationships that are held to be relevant to human exposures to PCBs in the  diet, in breast milk,
 or in the environment.

        Another major topic of discussion during this session was whether risk assessments
 should be conducted on a congener-specific basis,  or whether total PCBs can be considered an
 appropriate measure of dose for risk assessment purposes. One participant predicted that, if
 examined, there will turn out to be very little heterogeneity in the distribution of PCB congeners
 in humans. Another concurred, noting that breast milk samples from three or four different
 parts of the world look fairly similar in terms  of their congener make-up. A third participant
 also  agreed, suggesting that exposure to PCBs via the aquatic route, in particular, has become
significantly more homogeneous in the years since the production and distribution of these
compounds was curtailed.  One of these individuals suggested that, to the extent that the
distribution of congeners is unlikely to vary significantly, total PCBs might represent an
appropriate surrogate for dose. This same participant  expressed the view that total PCB levels in

humans would almost certainly represent a more relevant measure of dose than levels of an
industrial mixture in animals.

       One of the co-chairs of the human data work group held that the use of total PCB levels
as a measure of dose overlooks a number of potentially important considerations, including
differences in the pharmacokinetic and toxicologic profiles of the various congeners and the
potential for interactions that may occur among congeners in an environmental mixture.  She
suggested that, from an epidemiologic point of view, it is almost impossible to establish a
dose-response relationship without first having some idea of what the active agent is.  This same
participant noted that a number  of important assumptions are imbedded in the use of total PCBs
as a measure of dose, and suggested that it will be important to state these assumptions explicitly
if this is the approach that the Agency decides to take in addressing  the problem of exposure to
these compounds.

       Another participant agreed that it is not possible to predict the effects of a given
exposure without knowing the identity of the active agent, its mechanism of toxicity, and
something about the relationship between administered dose and the concentration attained at
the site of action.  This person noted, however, that there are very few compounds or classes of
compounds that have been studied as thoroughly as PCBs. He suggested that if the Agency is
unable to perform  a risk assessment on the basis of the available data for PCBs, it is unlikely to
be able to conduct risk assessments for any other chemicals,  either.  This participant also
expressed the opinion that it will be difficult for the Agency  to  stimulate significant amounts of
new research on these  compounds, given the large body of data that already exist.  Noting that
.there are also limitations on the number of congeners and potential  mechanisms that can
practically be studied, this participant suggested that at some point it will simply be up to the
Agency to decide that enough positive data exists to stake out a defensible regulatory position.

       Another participant suggested that cost can be a significant factor in decisions about
whether to perform congener-specific analyses, noting that these analyses generally cost up to  10
times more than measurements limited to total PCBs. Another person disagreed that all
methods of congener-specific analysis are this expensive, and suggested that in his  experience the
hesitation to undertake congener-specific analysis often has more to  do with uncertainty about
how to interpret the results. This participant said that it is not unusual for researchers to
undertake a congener-specific analysis, only to be so intimidated by the complexity of the results
that they end up reporting their  results in terms of total PCBs after all.  Noting that the pattern
of congeners present can be highly specific to a particular site, including specific bodies of water,
this participant also expressed the view that congener-specific analysis is also necessary to
"fingerprint" sources of PCB exposure.

       One person suggested that it is also important to keep in mind that there may be multiple
mechanisms through which PCBs exert their  toxic effects. As an example, this participant
suggested that it seems highly  unlikely that an effect observed in animals treated with Aroclor
1016 would be caused by the same congener  as the same effect observed in animals treated with
Aroclor 1254. To the extent that one accepts that multiple mechanisms of toxicity may be
operating, this person thought that congener-specific analysis also offers the best hope for sorting
out some of the important mechanistic relationships that may exist among congeners.

        As a compromise between total PCB and congener-specific risk assessment, one partici-
 pant wondered whether there might be smaller groupings of PCBs that could be used for risk
 assessment purposes. Another participant thought not, suggesting that the current interest in
 different classes of PCBs may be as misguided as the historical tendency to view all PCBs as
 acting through the Ah receptor. A third participant disagreed with this position, noting that
 most  people would recognize at least two distinct classes of PCB congeners, and suggested that
 in terms of their biological effects there may be even more classes or subclasses of compounds.
 Tin's same participant speculated that it may turn out to be specific combinations of congeners,
 rather than the concentration of any single compound or of total PCBs, that correlates best with
 observed changes in  behavior. This participant went on to suggest that, if this is the case, undue
 attention to total PCB levels may actually obscure the underlying relationship between exposure
 and observed effects.

        Another participant noted that structure-activity studies suggest that there may be
 important distinctions between the coplanar PCBs,  which appear to bind mainly via stacking
 interactions, and ortho-substituted compounds, which appear to bind mainly through lateral
 interactions.  Given the strong association that appears to exist between ortho substitution, the
 tendency for lateral interactions, and the potential to cause neurotoxic effects, this person
 suggested that a classification of PCBs based on these types of chemical reactivities might be
 meaningful for risk assessment purposes. Citing the example of toxic equivalency factors that
 have been used to compare the potency of various dioxin-like molecules, he further suggested
 that development of  a "thyroxine equivalency factor" reflecting the affinity of the various
 congeners for the prealbumin binding site might represent a good first step in this direction.

       In response to a question regarding the likelihood that there are other reactivities that
 might be useful for purposes of classifying PCBs, this participant expressed the view that other
 reactivities may exist, but he predicted that the laterality interaction will turn out to account for
 many important functional properties of these compounds, including their effects on multiple
 hormone systems. In response to a second question, this participant said that to his knowledge
 the prealbumin binding assay has not been performed with mixtures of PCB congeners, but he
 noted  that there is no obvious reason why this assay could not be used to test mixtures.

       Another participant expressed some concern about the proposed prealbumin binding
 assay,  noting that displacement of thyroxine from its binding site is only one of several
 mechanisms by which PCBs could act to decrease circulating levels of thyroid hormone.  Given
 the assay's inability to account for alternative mechanisms such as the induction of hepatic
 enzymes and direct effects of PCBs on the thyroid, this participant wondered about the predictive
 value of an assay that addresses only one such mechanism.  The individual proposing the
 thyroxine binding assay agreed that the thyroid hormone system is an exceedingly complex one.
 Noting that PCBs appear to act as thyroid hormone antagonists in some settings but as agonists
 in others, he agreed that the assay may not tell the whole story, but suggested that a thyroxine
 equivalent value might provide a useful first cut in looking at the effects of PCBs on the thyroid
 hormone system. The person who had raised the question noted that the prealbumin binding
 assay may be especially relevant  in terms of the neurotoxic effects of PCBs,  since conversion of
T, to T3 clearly occurs in the brain. The other participant agreed, noting that prealbumin also
 appears to be synthesized  in the brain, suggesting to him that this protein may play an important
 role in controlling the access of thyroid hormones (and by extension PCBs) to the central
 nervous system.

       Regarding the need for additional systems of PCB classification, one paiielist pointed .out
that neither of the methods currently used to classify PCBs (i.e., by .Aroclor standards and by
degree of coplanarity) has proven very useful in assessing the neurotoxic potential of these
compounds. This person suggested that our ability to move forward in this area depends on
finding ways of segregating out the properties of PCBs that are relevant to their toxicity profiles;
otherwise the debate will continue to be focused more on the mix of congeners involved than on
their individual or collective toxicity profiles.

       Another participant suggested that our understanding of PCB toxicity might benefit from
an approach that attempts to characterize these chemicals  from what he described as a combined
"top-down" and "bottom-up" approach. In the "top-down" component of this approach, efforts
would be focused on characterizing the effects of a congener mix designed to approximate
human exposure, while the "bottom-up" component would focus more on the characterization  of
individual congeners or groups of congeners that share important chemical properties. Even-
tually, the two approaches would presumably meet, revealing in the process something about the
mechanism or mechanisms of PCB toxicity.

       At the conclusion of this discussion, Dr. Francis opened the floor to comments from
observers, but none  were forthcoming. Dr. Francis complimented workshop participants on the
breadth and depth of the morning's discussion and suggested that it would serve as a useful
framework for the panel discussion scheduled to follow the morning break.
       Following a short break, Dr. Francis reconvened the meeting by reminding participants
that one of the stated purposes of the workshop was to obtain a sense of the group regarding the
adequacy of the available data for risk assessment.  Toward this end, she reviewed the categories
of sufficiency established by EPA in its developmental toxicity risk assessment  guidelines.  She
noted that, unlike the previous discussions, in which the animal and human data were considered
separately, the discussion of sufficiency should attempt to focus on the PCB toxicity database as
a whole.

       To stimulate discussion of these issues, Dr. Francis proposed a series of questions for
consideration by a six-member panel of workshop participants (see Appendix C). These
questions were:

       •      Are the data sufficient or insufficient for risk assessment purposes? What are the
              strengths  and weaknesses of the experimental animal data?  Of the human data?

       •      Is it appropriate to use data on commercial mixtures of PCBs as surrogates for
              other PCB mixtures? As surrogates for environmental exposures? Is  the use of
              such surrogates appropriate for purposes of risk assessment?

       •      How should the developmental neurotoxic effects of PCBs be considered in
              relation to other forms of toxicity?

        »      Are there data to suggest that there are populations at special risk due to elevated
               levels of exposure to PCBs or to inherent sensitivity?

        The panel discussion began with consideration of whether the available human data are
 sufficient to support the more quantitative aspects of risk assessment. Noting that there have
 historically been some concerns about the extent to which potential confounders were addressed
 in both the Michigan and North Carolina studies, one of the co-chairs of the human data work
 group commended the researchers involved in these studies for the additional analyses both
 groups had done in preparation for the workshop. Strengths of the available  human data, this
 person suggested, include the fact that similar effects were  observed in two very different
 populations as well as the fact that both studies involved the use of analytic techniques and
 cognitive tests that were state-of-the-art at the time the studies were performed.  The main
 shortcoming of the human data was the large number of non-detects in both the Michigan and
 North Carolina cohorts.  While acknowledging that more precise analytic methods have been
 developed since the time these studies were conducted, she suggested that the large number of '
 non-detects leaves many unanswered questions about dose-response relationships and/or the
 existence of a threshold at or below the limits of detection  of the analytic methods used  in
 these studies.                •   ,               -'  •   .  ,  • .        '•     -    •  ,         ,  ,.y

        Another panelist expressed some concern  about the potentially confounding effect of
 methyl mercury on the results of the Michigan study, given  that methyl mercury has long been
 known to pose a neurotoxic risk to individuals in fish-eating populations.  While acknowledging
 that this and other studies  provide strong qualitative evidence of PCB toxicity, this panelist was
 concerned that the role of methyl mercury  as a potential confounder might render the Michigan
 study less useful for rnore quantitative" estimatesI of risk.                                  !

        One person expressed the opinion that it is not deficiencies of the epidemiologic  studies
 themselves that is at issue, but rather the purposes for which risk assessors and risk managers are
 attempting to use these studies.  He suggested that the real issue is not whether PCBs can impact
 neurological function, which they obviously can, but whether the available evidence of toxicity  in
 humans meets the definition of sufficiency established by the-Agency in its risk assessment
 guidelines. In this panelist's opinion, neither the North Carolina nor the  Michigan  data meets
 the "cpnvincing evidence" criterion embodied in'the Agency's definition of sufficiency. This does
 not mean that the data should be ignored, he emphasized, but only that they are not sufficient to
 serve as the basis for a quantitative assessment of risk.

       Another panelist noted that EPA regional  staff often find themselves in the position of
 having to make'decisions based on limited amounts of data; this individual suggested that, as a ;
 result, there is a strong bias among risk managers that all available information should be taken"
into account in any characterization of risk.  From a risk management perspective, this person '
thought it better to rely on imperfect data than to ignore an effect that it  might be difficult or
impossible to go back and address later on, when better data do become available.  Given that a '
reference dose  for one of the Aroclors has  already been added to the Agency's toxicity database,
this participant suggested that risk managers are probably already using this data in their
decision-making process.                   '            •  '           -

       One panelist expressed the view that the database clearly supports a conclusion that PCBs
pose a risk of developmental neurotoxicity,  but noted that it is difficult to see  how the available

 data could be used in a quantitative risk assessment paradigm.  On the one hand, she thought
 that concerns raised about the limitations of both the human and animal data were legitimate
 ones, while on the other hand she agreed that it is important for risk managers to make
 regulatory decisions on the basis of all available data.  On balance, this panelist felt it
 appropriate to use the information now available to set limits on exposure to PCBs, although
 perhaps only with a number of caveats that would clearly reflect the uncertainties in the
 existing database.

        One panelist suggested that, given the availability of new data on the congener mix in
 breast milk samples, the Michigan study might be a more appropriate choice than the North
 Carolina study for quantitative risk assessment purposes. In response to this suggestion, one of
 the researchers involved in the Michigan study noted that congener levels had not been directly
 measured in the breast milk of women whose offspring took part in the study.  Rather, congener
 levels had been extrapolated on the basis  of consistencies between the mix of congeners seen in
 the serum of four-year old study participants and the mix of congeners present in breast milk
 samples from other populations of post-partum women. Given the almost total lack of more
 directly obtained congener-specific information, this individual suggested that a more appropriate
 question might be the extent to which total PCB levels can legitimately be used for risk
 assessment purposes.

        Following up on the previous comment, one panelist expressed his belief that the
 aggregate of all available data clearly shows that PCBs are not all alike.  Not only are there
 qualitative differences in the  effects produced by different congeners, but congeners producing
 qualitatively similar effects may  do so only at significantly different dose levels.  Because of this,
 he maintained, any broad statement about the toxicity of PCBs as  a class is likely to be
 very misleading.

       The same panelist noted that the vast majority of congener-specific work that has been
 done on PCBs—including work addressing the developmental effects of these compounds—has
 focused almost exclusively on the biological and/or toxicologic properties of the dioxin-like
 congeners. Noting that the neurotoxic effects of PCBs do not seem to be mediated via
 dioxin-like interactions with the Ah receptor, this panelist suggested that it will be very important
 to establish ways of discriminating  among  PCBs that encompass this diversity of effects. In the
 meantime, he suggested,  there is little in the way of baseline data against which to compare the
 results obtained following exposure to the non-dioxin-like congeners.

       Responding to the previous comment, another panelist found it a source of little concern
 that some toxicities may be due to  dioxin-like congeners and others may  not, noting that it may
 well be  the case that more  than one mechanism of toxicity is operating during development.
 Another participant agreed, suggesting also that the tendency  of a  congener to produce
 deleterious effects may depend not only on the structure of the congener but also on the timing
 of the exposure. He cited the example of the 3,4,3',4'- congener, which has been demonstrated
 to exert adverse effects on the developing  nervous system but  which has no effect on the
 behavior or neurochemistry of adult animals even at very high doses.  In order to detect these
 types of time-sensitive effects, this  person  suggested that it will probably be necessary to conduct
comparative studies designed to look at differences in perinatal  and adult responsiveness to  the
various congeners.

       Regarding the need for congener-specific analysis, one participant asked whether it is
reasonable to expect to find less chlorinated congeners in any exposed population, given the
rapid rate at which these compounds are eliminated.  Another participant responded that
measurement of PCBs in the urine can provide a good measure of concurrent exposure, which
might help address the issue of persistence of the less chlorinated congeners.  A third participant
noted that there is at least one less chlorinated compound—the 2,4,4'-congener—that does show
up in environmental  samples, in breast milk, and in perinatally exposed children;  because of this,
it is probably not correct to assume, as some have,  that the congeners present in Aroclor 1016
are irrelevant to human environmental exposures.  The individual who had initially raised the
issue agreed with this statement, noting that there is some recent evidence to suggest that at least
some of the less chlorinated congeners may remain stable in human tissue for 20 years or more.

       There was also some discussion of the extent to which the animal data might offer a
clearer picture of PCB toxicity for risk assessment purposes.  One panelist suggested that a
strength of the animal data is that all  of the commercial mixtures  exhibit some sort of
developmental neurotoxic effect, noting that this consistency  of effect is an important piece of
information for risk assessment purposes. A second panelist echoed this sentiment and noted
that despite differences in the  results of individual  studies, the human and animal database as a
whole strongly suggests a correlation between PCB exposure  and the potential for neurotoxic or
adverse neurodevelopmental outcomes.  A third panelist strongly disagreed with this contention,
arguing that the ability of very different commercial mixtures to produce the same effects actually
complicates the risk assessment picture, since there is little or no overlap of congeners in the
mixtures demonstrated to produce these effects.  This panelist also suggested that there are a
number of other important questions that can be raised about the utility of the animal data for
risk assessment purposes, including the paucity of clean dose-response data, the general lack of
replication in this area, the small numbers of animals employed in most studies, and the
possibility of cross-contamination, particularly in association with studies looking  at relatively low
doses of these compounds.  In a follow-up to the preceding comment, one participant pointed
out that the mixtures in the Aroclor 1000 series are fractional cuts of the more highly chlorinated
1200 series mixtures, so it is reasonable to expect that there would be some overlap among
congeners present in mixtures  from the two different  series.

       Panelists were divided in their responses to the question of whether commercial mixtures
could be used as surrogates for environmental  exposure to PCBs.  One panelist said that the use
of commercial mixtures is appropriate, but for purely non-scientific reasons. In her view, the
need to have some kind of measure or limit on exposure to these  compounds outweighs the
uncertainty associated with the available data.  A second panelist agreed, noting that addressing
uncertainty is part of what a risk manager does with the output of any risk assessment.  A third
panelist held that commercial mixtures are an appropriate surrogate for environmental exposures
because it is mixtures of congeners that are invariably encountered in the environment.  Another
participant differed sharply with this assessment, noting that  commercial mixtures usually also
contain several  dozen congeners that people are unlikely to encounter in environmental
exposures. In this person's view, attempting to define and characterize a standard mixture that
more or less duplicates environmental exposure would be a far more rational approach than
continuing to rely on the results of studies using commercial  mixtures.

       One panelist suggested that it is difficult to  know what mix of congeners might represent
an appropriate surrogate for environmental exposures, since  we know so little about the mix of

congeners that typically make their way into humans.  Another participant disagreed, noting that
thousands of chromatograms have been analyzed to answer precisely this question. Based on
these studies, he suggested that a relatively small number of congeners—probably 30 or 35
compounds—probably account for as much as 99% of the PCBs present in human serum  and
tissue samples, and that there is little geographic variation in the mix of congeners present.
Another person agreed with this assessment, noting that in his extensive experience with
high-resolution chromatography, there appears to be little variability in the mix of PCBs present
in human samples, except perhaps in the immediate vicinity of specific point sources. For the
most part, the distribution of various PCB congeners corresponds closely to the relative
proportion of each congener contained in all of the industrial mixtures produced prior to  1977.

       One  participant disagreed with the statement that there is little geographic difference in
the mix of congeners appearing in humans, noting that the pattern of congeners detected  in
samples from the Eskimo population, for example, is significantly different from the pattern of
congeners seen in some of the other populations  that have been studied. Noting that fish
consumption appears to influence the total body burden of PCBs when intake exceeds about 200
grams/day, this person suggested that the North Carolina cohort may be more representative of
the U.S. population than the Michigan cohort, since PCB levels in the Michigan cohort may have
been elevated as a result of that population's higher levels of sport fish consumption.

       Other panelists agreed that the use of commercial mixtures as surrogates for
environmental exposures is a questionable proposition.  One said that there is no way to answer
this question in generic terms, since the adequacy of specific studies and the availability of
dose-response information should determine the  applicability of the data on a case-by-case basis.
Another panelist expressed concern about using data from exposures that occurred 20 years ago
to predict the results of current environmental  exposures, particularly without a better
understanding of how the mix of congeners present in the environment may have changed
over time.

       There was also some discussion of the extent to which exposure to the commercial
mixtures themselves remains a problem that needs to be addressed.  One panelist noted that it is
not unusual  for the mix of congeners present at a Superfund or RCRA site to resemble one of
the commercial mixtures. Another speculated that there are probably hundreds of millions of
pounds of PCBs still in use in transformers, which means that the potential for occupational
exposure to  the commercial mixtures continues to exist. In both of these situations, he suggested
that the results obtained with commercial mixtures can be very useful in setting cleanup levels or
determining tolerable levels of exposure to these compounds.  One panelist observed that in the
case of exposure to commercial mixtures, the animal data is clearly the best information
available, since virtually all of the animal studies  conducted to date have involved exposure to
these mixtures.

       Regarding the issue of how neurotoxicity should be viewed relative to the other toxic
effects of PCBs,  most panelists  felt that other effects are important, but  that a  case for regulatory
action  could be made on the basis of the evidence for developmental neurotoxicity alone. One
panelist said that she has what she described as a nagging worry about the immunotoxic effects
of these compounds, while another suggested that effects on .thyroid function clearly merit
further consideration.  Another suggested that limits on our knowledge of the mechanisms
involved largely preclude any clear statement about the relationship between developmental  and

other types of toxicity, except to note that developmental neurotoxic effects appear to occur at
levels far below those associated with more generalized toxicity.

       One panelist expressed the opinion that no clear evidence of a causal relationship
between levels  of PCBs in the blood and adverse neurological effects exists, noting that the
observed correlation could be accounted for equally well by postulating a third biochemical
parameter  that accounts for both of these phenomena.  He presented data suggesting that serum
iron levels  might represent this third parameter, and noted that a number of other developmental
toxicants have been associated with a concomitant inhibition of heme protein synthesis and a
reduction in whole-body levels of cytochrome P450. This individual maintained that it is at least
possible  that disturbances in  iron metabolism could account for the observed neurologic effects,
and that concurrent changes  in cytochrome P450 levels may account for a responsive rather than
causative increase  in serum PCB levels.

       Another panelist reiterated his concern that developmental neurotoxicity appears to differ
from other types of PCB toxicity in that it does not appear to be mediated via interactions with
the Ah receptor, but he did not speculate about  how this difference should be  addressed.  Still
another panelist pointed out  that this pattern of toxicity is not unlike that seen with other
developmentally active agents, including lead, which produces developmental neurotoxicity at
levels much lower  than those associated with more generalized toxic effects.

       Other workshop participants raised additional issues in this area.  One noted that, at least
among transplacentally exposed children, there does not appear to be any strong correlation
between  the appearance  of physical stigmata and delays in neurological development,  suggesting
that these effects probably occur more or less independently of one another. Another participant
suggested that the  reproductive system may prove a useful one for further characterizing
differences in the toxicity of the various congeners, since in this system the less chlorinated
compounds appear to exert estrogen-like effects  on sexual differentiation, while the more highly
chlorinated congeners tend to produce more dioxin-like effects. A third participant suggested
that it may also be important to pay attention to the relationship between observed effects and
the duration of exposure to PCBs, since there may be significant differences in the effects
occurring after 'a brief but intense exposure and those that occur following a longer-term,
lower-level exposure.  In this sense, he suggested that the most important effect may be
determined to a large extent  by the type of exposure that a given risk assessment is attempting
to model.

       There was insufficient time for the panelists to address the issue of whether there are
populations at special risk due to elevated levels of exposure or inherently increased sensitivity
to PCBs.

                                   SECTION FIVE

                              DISCUSSION SESSION:

                                  Chair: Hugh A. Tilson
                            U.S. Environmental Protection Agency

       Dr. Hugh Tilson reconvened the meeting by noting that the second of the stated purposes
of the day's deliberations was to obtain a sense of the group regarding the program of research
that is needed to improve the .scientific basis for evaluating the risk of developmental
neurotoxicity associated with PCBs. Toward this end, Dr. Tilson asked two of the co-chairs to
present brief summaries of the research needs identified in their respective work groups. Dr.
Jane Adams presented a summary of the discussion of research needs that had taken place in the
group charged with reviewing the available human data, and Dr. Diane Miller presented a
summary of the discussion that had taken place in the animal data work group.  (See Section 6.2
of this  report  for the work group co-chairs' written summaries of these discussions.)

       Following these presentations, Dr. Tilson listed a number of issues for consideration by a
six-member panel of workshop participants (see Appendix.C).  Noting that EPA is assigned the
task of dealing with the available data in a risk- assessment context, Dr. Tilson asked the panel to
try to focus their comments on how any recommended research could be expected to reduce the
uncertainty in assessments  of risk that are based-on the available data.  Dr. Tilson indicated that
existing reference doses for PCBs are based on endpoints other than  developmental
neurotoxicity, and suggested that it might also be useful for the panel to recommend/any
additional research that might be necessary to justify basing a reference, dose on the  types of
effects discussed at this meeting.             ,

       Noting that it was his sense that the question of how to deal with mixtures is one of the
most pressing in this area, Dr. Tilson suggested that it might be useful for panelists to discuss the
types of research needed to deal with this issue. He said that it was not clear to him, for
example, whether any consensus had been reached regarding the need for direct comparisons
between commercial and environmental mixtures of PCBs, or between the effects associated with
exposure to coplanar versus non-coplanar compounds. In addition, he wondered whether there
was a clear sense of the group regarding the  relative merits of the "top-down" and "bottom-up"
approaches that had been discussed earlier in the meeting. Finally, he reiterated that it would be
very important for panelists to consider ways in which any recommended research might or might
not be expected to  improve the scientific basis of risk assessment in this area.

       One panelist suggested that there did seem to be a consensus regarding the desirability of
conducting animal studies with PCB mixtures that more closely approximate the mix of

 congeners thought to exist in the environment. This same panelist suggested that it might also
 be useful to further characterize the various combinations of congeners involved in different
 types of human exposure to PCBs.  To the extent that differences in these exposures can be
 correlated with differences in the effects observed, she suggested that it might also be possible to
 get at the mechanisms underlying differences in the toxicity profiles of specific congeners  or
 groups of congeners.

       Noting that there is a relatively large database upon which to base the conclusion  that
 developmental exposure to PCBs poses a hazard, another panelist agreed that what is really
 needed is more information about the nature of environmental exposures. Another panelist
 concurred, suggesting that risk assessment should proceed using whatever data are currently
 available, but that future studies should be conducted using a mix of congeners that more closely
 approximates our best guess about the congeners involved in real-world exposure to
 these compounds.

       Another person suggested that the effects of commercial mixtures of PCBs remains an
 important area of study, since exposure to these mixtures continues to occur in settings such as
 Superfund sites and accidental spills.  In addition, he noted that the original  source of most
 environmental exposures was the commercial mixtures.  This panelist conceded that the appeal of
 congener-specific experiments is mitigated somewhat by the fact that congener-specific exposures
 do not occur.  In addition, given the interactions that might be expected to occur, he agreed that
 it may turn out that exposure to a single congener in isolation  does not produce the same result
 as exposure to the same congener as part of an environmental mixture.

       Another panelist argued that congener-specific information would be nice to have  from a
 basic science perspective, but that it is largely irrelevant to the risk assessment process, since
 exposure  to a single congener almost never occurs. Given that exposure is always to mixtures,
 this person expressed the view that  even a full characterization of each individual congener would
 not tell us much about how these substances produce the toxic effects that have been observed.
 She speculated that to some extent the call for congener-specific data may be little more than a
 ploy to preclude regulatory action based on the large body of data that is already available.

       Noting that it would be both impractical and meaningless to  attempt to  look  at all
 possible combinations of the 209 PCB congeners, one panelist suggested that a more rational
 approach might be to pick a half-dozen or so from the 30 or 35 congeners to which people are
 commonly exposed and to study these specific compounds relatively  exhaustively in an animal
 system. This individual thought that by using characteristics such as persistence in the body,
 laterality, and coplanarity to select certain compounds for study, it should be possible to learn a
 great deal about the nature of differences in the toxicity of different congeners or mixtures of
 congeners. Another panelist agreed, noting that it would probably be very instructive to
 determine how the dose-response profiles for some of the individual congeners or mixtures of
congeners compare with the profiles of environmental mixtures. In this way, he suggested that it
 might be possible to determine whether a particular subset of PCBs is responsible for some or all
of the toxic effects observed.  This person also  suggested that the toxicity of individual congeners
should be studied in the reproductive, endocrine, and immunologic systems as well as the
nervous system, and that these studies should be undertaken in multiple species.

       Another person expressed the view that regulatory action should not await the results of
congener-specific studies, but that congener-specific research may offer the best hope for
improved understanding of the mechanisms of toxicity that underlie the broad range of effects
produced by these compounds. A second panelist agreed that pursuing the mechanisms by which
these substances exert their toxic effects should be high on the research agenda.  He noted that,
in order to predict the effect of a given dose of a substance, it is necessary to  have both some
understanding of the mechanisms of toxicity at the target organ and some idea Of the relationship
between administered  dose and the concentration attained at the target site.  At this point, he
suggested, we don't even know what the active agents are, much less how they exert their toxic
effects. A third panelist agreed that better characterization of the mechanisms of developmental
toxicity is needed, and suggested that the distinctions people have tried to draw among the
chemical reactivities of different subclasses of congeners might be particularly useful in
this regard.

       One panelist suggested that  for regulatory purposes it is not clear whether a reference
dose derived from the developmental neuroto'xicity data will  really be very useful, since the
reference dose  derived from cancer data seems likely to continue to drive the PCB risk
assessment process. While acknowledging that the application of uncertainty  factors might
produce a reference dose that approaches the number derived from the cancer data, this
individual thought that adoption of  a reference dose based on a non-cancer endpoint is highly
unlikely unless  the methods used to calculate carcinogenic risk undergo significant changes.

       Dr. Tilson asked each of the panel members whether they thought that the research  they
were recommending had to be completed before risk assessment could legitimately be
undertaken on  the developmental neurotoxicity of these compounds. One panelist expressed the
view that the available data are adequate and that either the Michigan or North Carolina study
could reasonably be used for this purpose.  Another panelist concurred with this view, noting
that there is enough consistency between the results of the Michigan and North Carolina studies
to set at least a provisional reference dose at this time.

       Another panelist suggested that the available data should be used for risk assessment
purposes, but only with the qualification that more information is needed, including more
replication of the studies upon which the assessment is based. A second person objected strongly
to the repeated calls for more replication, noting that slide after slide of studies replicating one
another's results had been presented throughout the meeting. This person shared the view of a
previous commentor that there is far more replication in the PCB literature than in most other
areas of toxicology. Another panelist thought that the need for  replication also extends to the
human studies,  particularly since currently available  analytic methods could easily address some
of the major sources of uncertainty associated with the Michigan and North Carolina  data.
There was some discussion of additional human studies that are currently getting underway, but
this panelist expressed some concern regarding the power of the study designs that have thus far
been proposed.

       Another panelist  expressed some reservation about the large degree of uncertainty that
would be associated with any risk assessment based on the data  that are currently available.
Noting that there is a  considerable amount of heterogeneity both in the chemicals themselves
and in  the endpoints they appear to affect, this individual suggested that further research should

 be directed more specifically toward reducing the uncertainty associated with whatever number
 or range of numbers the preliminary risk assessment produces.

        At the conclusion of this discussion, Dr. Tilson summarized the panel's comments by
 noting that, while there appears to be some hesitancy about going ahead with a risk assessment,
 most panelists believe that this is a reasonable thing to do, since the human studies are probably
 adequate for this purpose.  He noted that most panelists seemed to feel that additional research
 is needed, both to identify the mix of congeners that best approximates environmental  exposures
 and to refine our understanding of dose-response relationships and the mechanisms by which
 these compounds exert their toxic effects.  Finally, he suggested that the panel seemed to be in
 general agreement that any reference dose set on the basis of currently available data should
 be revisited as more information about the toxic effects of environmental mixtures
 becomes available.

       Following this summary of the panel's deliberations, the floor was opened to comments
 from other workshop participants.  One participant suggested that there is a need to devise some
 basis for applying a general determination of PCB toxicity to the large variety of mixtures that
 may be encountered in the environment. Noting that in the case of cancer the equivalency
 position is that all PCBs are equally potent, this person suggested that such an assumption is
 clearly not warranted in the case of developmental neurotoxicity. He suggested that a program
 of research is needed to establish some sort of toxic equivalency system for PCBs, perhaps based
 on the types of structure-activity relationships proposed at this meeting.  The panel  chair
 suggested that it will be much more difficult to develop an equivalency system for the
 developmental neurotoxicity of PCBs, since no one has a clue as to what the mechanism of
 toxicity actually is. Another panel member noted that there had been extensive discussion
 throughout the workshop of structure-activity categories  that could be used as a first
 approximation of distinctions  that may prove relevant to the varying biological effects of these
 compounds.  A second participant agreed, noting that the alternative is to assume that all PCBs
 are equally potent, which he thought would be scientifically indefensible given what  we already
 know about differences in the toxicity of the various congeners.

       Another panel member disagreed strongly with the implication  that further research is
 needed before risk assessment can proceed in this area, noting that the objections raised  at this
 meeting are the same types of stalling tactics that were used to delay the regulation  of lead,
 methyl mercury and other developmental n euro toxicants acting by mechanisms that  remain
 largely unknown to this day. She argued that the issue is not whether anyone believes that these
 compounds are all equivalent in their effects, but rather whether it is reasonable to  assume that
 they are all equivalent in order to perform a first-pass  risk assessment. She .maintained that such
an approach is reasonable, given that the scientific research to establish meaningful  distinctions
among these compounds would certainly be free to continue. One of the participants who had
raised the question of equivalency objected to this characterization of the preceding argument,
noting that no one had suggested doing nothing, but rather that some people thought it would be
difficult to defend applying a single risk value to all PCBs.

       Another panelist suggested that an alternative way of thinking "about this issue might be
to conclude that a measure of total PCBs represents a reasonable marker for whatever is toxic in
the mixture. This individual felt that at least for now this seems to be a reasonable assumption,
particularly given the consistency of results that people have obtained using total PCBs as a
surrogate for dose.

       There was also considerable discussion regarding the identification of specific congeners
or metabolites that may play a role in the  developmental  effects of PCB mixtures. One
participant expressed a concern that the compounds of greatest interest may not even have been
present at the time serum sampling was performed in the existing epidemiologic studies.  She
speculated  that the greater effect of prenatal than postnatal exposure in the North Carolina study
might mean that a shorter-lived compound than those present in breast milk was responsible for
at least some of the developmental effects observed.  One participant suggested that analysis of
fetal cord samples should have been able to  address this question, but another thought that this
would only be the case if the substance producing damage during gestation remained in the
maternal blood at term.  Two participants  stated that the mix of congeners in maternal blood
varies little over the course of gestation, but one of the panelists observed that even if levels  do
remain  stable in the maternal system,  access of these compounds to the active  fetal tissue is likely
to be highly variable over time.   This person suggested that multiple time point sampling would
be the best way to answer this question, since this approach would address most directly  the
possibility that specific windows of vulnerability occur during development.

       Regarding the need for replication of the human studies, one participant suggested that it
might be useful to establish a standard or  at least model protocol for these types of
investigations: Citing the lead literature as an example of an area in which some standardization
has occurred, this participant  suggested that  the adoption of common time points for
measurement, the use of similar measures of exposure, and the use of a common battery of
developmental tests all have the potential  to mitigate the kinds of problems that can arise in
trying to compare the data from different epidemiologic studies. One of the panelists thought
that this was an excellent idea, while another expressed some concerns about excessive
standardization.  While agreeing that some standardization, particularly of the  functional tests, is
desirable, this person thought that it is also important for investigators who have a hunch or  a
particular type of expertise to have the freedom to explore new areas of inquiry.

       Following up on the preceding comment, another  participant suggested that it might also
be useful to consider the extent to which variables related to the administration of functional
tests may influence the results of human studies. Noting  that many investigators fail to take into
account the difficulties that can arise in administering  these tests, this person suggested that
there needs to be some way of sensitizing investigators  to the potentially confounding effects of
variables such as  ambient light and the speed with which the test is conducted that on the surface
may seem relatively unimportant.  Another panelist found it interesting that investigators doing
human studies 'may not be sufficiently attuned  to these  types of variables, noting that much effort
over many years has been devoted to the standardization  of behavioral testing in animals.

       Another area that a participant suggested might benefit from greater standardization  is
the collection and preservation of serum and tissue samples. This person noted that much of the
uncertainty associated with the Michigan and North Carolina data could be resolved if samples
from the study populations were available  for analysis  today.

        Following this discussion, Dr. Tilson asked the panelists to comment on the extent to
 which it is or is not necessary to establish a mechanism of PCB neurotoxicity before proceeding
 with a risk assessment. The panelists were generally in agreement that risk assessment should
 proceed using the data that are currently available, but that investigation of the mechanisms of
 toxicity should continue to be a high research priority.  One panelist also suggested that
 proceeding with the risk assessment may itself stimulate research on the mechanisms of PCB
 toxicity, since the imposition of a provisional reference dose will provide a strong incentive for
 those who disagree with the assessment to fill some of the gaps in the data.

        Dr. Tilson also noted that an issue that came up in both the human and animal data
 work groups was the need for further characterization  of the effects observed following exposure
 to PCBs. Using the example of the relatively simple, screening-level tests  used to measure motor
 function in some of the human studies, Dr. Tilson noted that there are more sensitive tests that
 could be used to further characterize this effect.  Noting that more sensitive tests might offer a
 better estimate  of the NOAEL or LOAEL for these compounds, Dr. Tilson asked the panelists
 to comment both on the types of work that need  to be done in this area and the extent to which
 risk assessment  should or should not be postponed until the results of more sensitive tests
 are available.

       Panelists generally  agreed that this type of work should also be a high research priority,
 but that there is no need to delay risk assessment pending its outcome. Within this context,
 however, a number of panelists  suggested specific areas of research that they thought especially
 deserving of further attention.  One panelist brought up the  effects of PCBs on dopamine levels
 in the brain, noting that this effect is certainly interesting,  but that it has not yet been correlated
 with any functional endpoint. Another participant agreed, suggesting that  the functional
 consequences of these effects will not be understood until changes in dopamine levels  can be
 correlated with an effect in the intact animal that can in turn be related to disturbances in a
 specific dopaminergic pathway.  This person also  suggested that it will be important to look
 carefully at dose-response relationships in this system,  since  at least some  of the  available
 evidence suggests that dopaminergic effects may occur  only at doses associated with more
 generalized forms of toxicity. One of the panelists noted that, although it  is correct that no
 direct correlation between  changes in dopamine levels  and functional deficits has thus  far been
 established, it is interesting to note  that the many of the endpoints affected by PCBs are precisely
 those that one would want  to examine in attempting to confirm a postulated effect on
 dopaminergic function. Another participant suggested  that additional information on the
 functional significance of changes in brain dopamine should be available shortly, since  studies
 designed to examine functional endpoints and dopamine levels in the same animals have  recently
 begun in her laboratory.

       The same participant who raised the issue of the need for dose-response evaluations of
 PCB effects on dopamine also suggested that similar types of studies are needed  to clarify the
 functional consequences of PCB effects on circulating levels of thyroid hormones.  In the  absence
 of any evidence suggesting functional impairments that are clearly thyroid hormone-mediated,
 this participant expressed concern about the advisability of developing thyroxine equivalents or
similar equivalency factors for these  compounds.  As in the case of dopaminergic effects, he held
 that the functional relevance  of these changes should be established before the changes
 themselves can be used to define subclasses of PCBs in any meaningful  way.

       One person suggested that the reproductive effects of these compounds, which are also
mediated via the central nervous system, need to be characterized further, as do centrally
mediated effects ,on immunologic and thyroid function. Another pointed out that further
characterization of the affected endpoints may provide important clues regarding the mechanisms
of PCB toxicity, particularly if endpoints are selected to involve neurophysiologic or
neurochemical  systems that are themselves well characterized.

       Responding to a different part of Dr. Tilson's question, one participant expressed the
view that increasing the precision of NOAEL and LOAEL estimates  should not necessarily be a
goal of research in this area.  This person suggested that, it would probably be more useful to
think about PCBs in term's of the levels of risk associated with different levels of exposure than'
to attempt  to define a specific threshold or no-effect level.

       At the conclusion of this discussion, Dr. Tilson  opened the floor to comments from
observers.  One observer referred back to the discussion  of the ability of different commercial
mixtures to cause similar effects, which at least one participant had suggested might pose a
problem for risk assessment.  The observer suggested that the uniformity of effects should
actually increase the level of confidence in  an assessment that relies on data from commercial
mixtures to predict the risks associated with environmental exposures.  This observer also
expressed support  for the panel's conclusion that more information is needed both for individual
congeners and  for  mixtures of congeners to refine  any preliminary assessment of risk that  can be
made on the basis  of the available data.  He also suggested that additional  research may be
needed to identify  specific populations with increased susceptibility to the effects of PCBs, noting
that the apparent multiplicity of toxic mechanisms  may mean that there are also multiple
subpopulations  in whom specific types of effects are more likely to occur.


                                    SECTION SIX
       This section summarizes work group discussions on the adequacy of the available human
and animal data bases on developmental neurotoxicity for risk assessment purposes, with special
attention to future research needs. The work group co-chairs prepared these summaries, and
supplemented them in some cases, with references to the literature.

                                  Mary M. Prince, Ph.D.
                    National Institute for Occupational Safety and Health
                                     Cincinnati, OH
       Most of the evidence for developmental toxicity due to chronic low-level environmental
PCB exposure in humans is derived from epidemiologic studies conducted in North Carolina
[NC] (Gladen et  ail., 1988, 1991; Rogan et al., 1986b) and Michigan [MI] (Fein et al., 1984;
Jacobson et al., 1985, 1990, 1990a).  The NC and MI studies were extensively reviewed for this
workshop to assess whether differences in the observed results are actually due to chronic
low-level PCB exposure or whether these results could be explained by differences in the design,
conduct, or analysis of these studies. Both infant populations were found to have developmental
deficits related to prenatal PCB exposure, but these effects persisted only in the MI study.
Furthermore, the developmental deficits in the MI study were cognitive (Jacobson et al., 1990a,
1985), while the NC study reported psychomotor  deficits in infants and children  up to 2 years of
age (Rogan et al., 1986b; Rogan, 1991).  Neither  study found there to be significant relationships
between postnatal exposure to PCBs and neurobehavioral deficits at birth or later in life.

       The deliberations of the work group focused on questions raised in and comments
received on the pre-meeting issue paper by Dr. Mary Prince (see Section 2 of this report).  Each
study was evaluated in terms of the following points:

       •     study design;

       •     analysis of study results;

       •     exposure assessment; and

       •     issues related to the interpretation of results.

 For ease of discussion, study methodology was divided into design and analysis components.
 Issues related to the interpretation of results were considered in terms of the various
 developmental endpoints measured in these studies and the relationship of these endpoints to
 prenatal and postnatal PCB exposure.  The epidemiologic evidence as a whole was evaluated
 with respect to its usefulness for both qualitative and quantitative risk assessment.

    ,: ',  To provide an  appropriate perspective for the work group's deliberations, relevant
 background information  on the two study populations was summarized in a series of tables used
 to-organize the discussion. Table 1 summarizes differences between the two  cohorts with respect
 to sample size, source  of PCB exposure, and method of maternal recruitment. The MI
 researchers interviewed women about their fish consumption after delivery, while the NC
 researchers recruited mothers before delivery. Actual recruitment of subjects into the MI cohort
 was limited to women  who reported that they consumed at least 11.8 kg over a 6-year period. As
 noted in the 1990 Jacobson et al. report,  contaminated fish consumption was  defined as the sum
 of annual lake fish consumed, weighted by the PCB content of the species consumed, in,the
 present or past, whichever was greater. A random sample of women who did not consume fish
 was also invited to participate in the study. This sample represented 4.6% of all women
 reporting no Lake Michigan fish consumption (i.e., the "non-fish consumers").

       The two study populations were similar with respect to socioeconomic status (SES),
 maternal  education, race, age, and mean  length of breast feeding.  In the MI  study, 52% of the
 infants were male, 43% were firstborn, and 6 infants were the larger member of twin pairs.
 Although 90% of the exposed infants were designated  "at medical risk,"  risk factors were not
 systematically under-represented among infants  in the  unexposed group (Fein et al.', 1984). The
 MI researchers examined a total of 73 potential confounders, including demographic background,
 history of pregnancy and delivery, obstetrical medication, and fetal exposure to drugs, alcohol,
 caffeine, and nicotine.  They also examined variables such as parity of child, marital status,
 maternal weight,  education, HOME scores, maternal vocabulary, maternal employment,  nursery
 school attendance of child, and familial stress. Exposure to other environmental contaminants
 such as lead, DDT, and PBB was also considered in 4-year-old children  (Jacobson et al., 1990).

       The NC cohort of women was predominantly white (92%) and well educated (53%
 college educated).  The NC researchers also examined several potential  confounders in their
 analysis: maternal  factors such as age, race, education, occupation,  smoking, and drinking, as
 well as child's sex, gestational age, birth weight, head circumference at birth, jaundice, duration
 of breast feeding, and parity (Gladen et al., 1988).  PCB  levels in breast milk  tended to be higher
 in women who were older, regularly drank alcohol, arid were primiparous (Rogan et al.,  1986b).

       The deliberations of the work group were guided by several  questions, each of which is
 discussed in the paragraphs that  follow.

       •     Question:  Could incomplete control for potential confounders (such as alcohol,
             maternal smoking, and weight) explain the positive findings in these studies?

       In the pre-meeting comments, the primary investigators involved  in the NC and MI
studies provided additional information on their analysis  of confounding factors and issues
related to possible selection bias  (see Appendix A of this report). The investigators pointed out
that there were very few variables that were related to both the outcome and PCB exposure, and

                              TABLE 1


                      STUDY DESIGN FEATURES
North Carolina Study
Michigan Study
General population
(Few fish eaters,- 21%)
                   ~ 900 Mothers recruited
                     from 4 hospitals before
                     delivery (1978-82)
                   856 infants recruited
Women living in the vicinity
of Lake Ml (Grand Rapids &
Muskegon). Over-sampled for
women who had consumed
lake fish.

8484 women screened from 3
hospitals after delivery
(1980-81):  77% reported
moderate to high fish

313 infants recruited:
   242 exposed infants
   71 nonexposed infants

 none of these significantly changed the main conclusions of their studies.  During the work
 group's deliberations, the following covariates were brought up for discussion:

         »      socioeconomic status (SES) of the study populations;

         •      iron status of the mothers and infants;

         •      selection bias due to attrition; and

         •      methyl mercury contamination.

        Based on the additional SES data submitted by the investigators, the two cohorts did not
 appear to differ by social class.  In the MI study, the families ranged from upper middle class to
 working class, but fewer lower class families participated in the study.  The women who
 participated in the NC study were also highly educated.

        Iron status was initially raised as a potential confounder because of similarities between
 the behavioral and cognitive effects thought to result from  iron deficiency and those observed
 following PCB exposure.  Related to this issue were data suggesting that serum iron may be
 strongly correlated with the rate of PCB clearance.  It was  hypothesized that, because
 iron-deficient subjects have lower levels of P450-2B (an iron-heme protein), these individuals
 might exhibit lower PCB clearance rates and higher steady state accumulations of PCBs. The
 concern was raised that if iron deficiency produces the same deficits in cognitive performance as
 those associated with PCB accumulation, then iron deficiency could account for at least some of
 the correlations seen in the epidemiologic data.

        There was, however, a general consensus among work group members that confounding
 due to iron deficiency was not likely to be relevant in these study  populations.  First,  the women
 in these study populations received very good prenatal care and would have been treated for iron
 deficiency.  The degree of maternal iron deficiency necessary to cause postnatal cognitive deficits
 in the offspring was not clearly delineated in these discussions.  The issue of iron deficiency was
 proposed to be more relevant to postnatal iron status.  Since postnatal PCB exposure was not
 related  to any neurobehavioral deficits at any age in either  of these studies, iron deficiency would
 not be expected to alter the interpretation of study results.  In addition,  there was some
 skepticism about the relationship between iron deficiency and the types of psychomotor effects
 observed in these studies. There were some suggestions that pre- or postnatal iron deficiency
 would be more likely to affect cognition than psychomotor performance.

       Work group members agreed that selection bias due to attrition is an issue in any
 prospective epidemiologic study.  Attrition was apparently not a major problem in either the MI
 or NC study, however, since attrition did not appear to be related  to exposure in either study. In
 addition, trends in the MI study  suggest that attrition may have produced a population that was
 actually higher in SES than the population initially recruited into the study. It was suggested
 that drop-outs in this study were mainly families of lower SES, presumably because more
 educated parents had greater interest in obtaining information on  their child's cognitive
 development and may have placed greater value on research in general.  Because of this, it is
likely that any bias due to attrition would have tended to be in the direction of decreased risk of
psychoneurological deficits.  It has been  suggested in previous epidemiologic studies that the

ability to detect relatively small health effects (if such effects truly exist) is influenced by the
overall "risk-status" of the population (Rothman and Poole, 1988).

       The role of methyl mercury exposure as a potential confounder cannot be ruled out in
the MI study since fish consumption was a source of PCB exposure. The issue of whether
methyl mercury levels were high in Lake Michigan fish was a point of contention that was not
fully resolved in the work group.  However,  during the discussion, it was mentioned (without
reference) that the effects observed in the MI study were not as severe as the neurobehavioral
effects seen in studies of methyl mercury exposure.  In one set of pre-meeting comments, a
WHO 1990 Environmental Health Criteria document regarding methyl mercury exposure and
fetal risk is cited.  This document states that a peak maternal hair mercury level above 70
micrograms/g  is associated with a high risk  (more than 30%) of neurological disorders in
offspring. Since mercury levels were not measured in the MI cohort, one can only speculate
whether the effects observed are  due to mercury or PCB exposure.

       In summary, the group as a whole felt that, for the variables collected and examined in
both studies, inadequate control of confounding was not a likely  explanation for the positive
results.  However, the possibility  of confounding due to unknown or unmeasured risk factors  or
confounders could not be definitely ruled out in either study.

       •      Question: Do the  two populations  differ in their congener distributions of PCBs
              and how might this influence the interpretation of the study results?

       It is clear from information previously reviewed that the sources of PCB exposure
differed in the MI and NC populations.  Although there are no published data regarding the
congener composition of PCBs in the NC cohort, Jacobson et al. (1989) have published some
data on the pattern of congeners present in  children's sera at 4 years of age. Overall PCB
elution patterns in these samples closely matched patterns reported in serum samples from
regular fish eaters and other adults (Humphrey, 1987). Although many different PCB mixtures
have been deposited in the environment (Jensen and Sundstrom, 1974), these data confirm that
certain congeners preferentially accumulate  in both adults and children. In the MI cohort
(Jacobson et al., 1989), the most  prevalent congeners were the same as in human milk and serum
samples evaluated in previous studies (Parkinson et al., 1980; Safe et al., 1985; Yakushiji et
al., 1979).

       Most of the work group participants  indicated that congener-specific data on both
populations would have to be available to adequately address this question. There was some
speculation that background PCB congener  distributions should be relatively similar  across
geographic areas,  but it was not clear whether data exist to support this hypothesis.

       •     Question:  Could different analytic methods for total PCB measurement account
              for differences in  exposure magnitude or mask differences in congener-specific
              distributions in the study populations?

       Based on the exposure profiles of the two populations, it would appear at first glance that
maternal serum PCB levels were lower in the MI study than in the NC study (see Table 2).  It
was difficult to compare exposure levels between the studies because of differences in the
analytic methods used. The MI researchers used an adaptation of the Webb-McCall method

                               TABLE 2

                       PCB EXPOSURE PROFILES1
 measure for Ml)
 measure for NC)
           North Carolina2
           1.8 ppm (median)
           (1800 ppb)

           MAX= 15.8 ppb;
           95th Percentile:
            3.6 ppm
Michigan Study3
0.836 ± 0.38 ppm (mean)
(836 ppb)

(range: 0.18 - 2.6 ppm)
           ~ < 4.3 ppb
           (max = 410 ppb)
                    80% below
                    detection limit
2.5 ± 1.9 ppb

(range: 0-12.3 ppb)


26.3% of "fish exposed" kids4:  PCBs
< 3 ng/mL (limit of detection)
           < 12.0 ppb

            97% below level
            of detection.
           9.06 PPB
 5.5 ± 3.7 PPB PCB
PCB (post-natal)
          NOT MEASURED
 2.1 ± 3.3 ppb (0-23: range).
DDE concentrations differed irt these populations (Jacobson et al..
1989 Rogan et al., 1986a);
  ppm = parts per million;
  ppb = parts per billion;
  milk sample PCB levels are all lipid adjusted

Represent median values (Rogan et al., 1986a)
Represent mean values (Fern et al., 1984; Jacobson et al., 1989)

Children whose mothers had exposure to PCBs through fish
consumption (Jacobson et al., 1989)

(calibration standards 1016 and 1260; Webb and McCall, 1973), while the NC researchers used a
method developed by McKinney (calibration standard Aroclor 1254; McKinney et al., 1984). The
NC study reported median maternal blood levels between 7 and 9.1 ppb (Rogan et al., 1986), as
compared with 5.5 ppb in the MI study (Fein et al., 1984).  Quantification of total PCBs in the
MI study was estimated on the basis of a dozen peaks (Schwartz et al., 1983), while estimates in
the NC study were based,.on two chromatographic peaks (125 and 146). It was suggested that if
PCB values had been estimated by the same methods in both studies, exposure levels in the MI
population would probably have turned out to be marginally higher than in the NC population.
During work group discussions, there was also speculation  that background PCB congener
distributions would be generally similar across geographic areas.  However, it is not clear whether
published data currently exist to support  this hypothesis.

       •      Question: Different physiological compartments were sampled for PCBs in the
              two populations.  Does this apparent inconsistency weaken the  evidence for an
              association between PCB  exposure and neurobehavioral deficits?

       A direct comparison of the results of the MI  and NC studies is difficult in view of their
differing definitions of prenatal exposure. Table 3 summarizes the study definitions for prenatal
PCB exposure and lists the compartments sampled in the NC and MI studies.  Both studies
measured cord blood, maternal serum, and breast milk PCB levels. Both studies—but
particularly the NC  study—had difficulties in detecting PCBs in cord blood. The MI group
additionally measured serum .levels in 4-year-old children.  The NC group attempted to measure
placental levels but  found very few samples  in which PCBs were detectable. The MI study
examined cord blood PCB levels directly; the NC study did not,  due to analytic limitations in
detecting cord blood PCBs, which were below the level of  detection in 88% of samples.
Therefore,  the NC researchers constructed a measure of transplacental PCB exposure using
several postpartum measures which were found to be highly correlated with one another (Rogan
et al.,  1986a). Researchers involved in the NC  study indicated that, for women having  detectable
cord and placental PCB levels (about 12% and 3%, respectively), the correlations between these
measures and breast milk PCB levels were found to  be fairly good. Based on this, prenatal
exposure was assumed to correlate well with cord serum PCB levels.

        There was general concern among work group members that there may have been some
overlap of pre- and postnatal compartments in  the NC study. However, it was not clear whether
any such overlap might explain the differences  in the results of these two studies.  To resolve this
question, a comparison of the results using breast milk and cord blood levels as measures of
prenatal PCB exposure would be necessary.  Since archived samples are not available, however,
this is not feasible.

        Another issue addressed was whether the large proportion of cord samples (about 66%)
in which PCBs were not detectable in the MI study could have accounted for differences in the
results of the two studies. The NC researchers could not use their cord blood values due to the
large proportion  of non-detects among the samples.  It was pointed out during discussions that
the analytic methods used in both studies were state-of-the-art 'at the time.  Nondetectable levels
are no longer a problem given improvements in current analytic procedures for measuring PCBs.

        Many participants suggested that this problem would not affect the validity of the study,
but would rather decrease the study's power. The high proportion of non-detects would also

                             TABLE 3

               North Carolina Study
                         Michigan Study
Sampled for


 Common scale used to
 account for different com-
 partments sampled and

POST-PARTUM measures
are surrogates of
prenatal exposure


* during & prior to pregnancy

* Weighted Sum of annual
  Lake Ml fish consumption;
  weights assigned according
  to average contaminated
  levels in lake fish.


affect the ability of the study to discern a dose-response relationship, particularly at lower dose
levels.  Although the relative proportions of non-detects would not necessarily account for
differences in the results directly, the large number of non-detects did impact on the NC
investigators' ability to examine developmental function in relation to cord blood PCB levels.
PCB levels in cord blood are a more direct measure of prenatal exposure than breast milk, and
the use of cord blood levels would also obviate the risk of pre- and postnatal compartmental
mixing. These differences in definitions of pre- and postnatal exposure make it difficult to
directly compare the results of the two studies.

       •      Question:  Different functional'domains were affected in these studies with respect
              to prenatal PCB exposure.  How does this affect the interpretation of study

       A major issue of concern in terms  of the concordance of results across studies was the
observation that different functional domains appear to have been affected by prenatal PCB
exposure in the MI and NC studies.  The results of the two studies are summarized in Table 4.
No association between postnatal exposure to PCBs and developmental deficits was apparent in
either study. However, the MI study found significant reductions in birth weight, gestational age,
and head circumference, all of which were significantly related to prenatal PCB exposure as
measured  by cord blood levels and total fish consumption.

       The NC study did not find any  significant relationship between these endpoints and PCB
.exposure.  However, the NC study found more hypotonicity and hyperreflexia on the Brazelton
Neonatal Behavioral Assessment  Scale (BNBAS) with increasing prenatal PCB exposure.  This
psychomotor impairment was not observed in the MI study, which reported that maternal fish
consumption (measure of prenatal PCB exposure) correlated with deficits in autonomic maturity,
increases in the number of abnormal reflexes, and poorer lability of states. No effect was
observed to correlate with cord PCB measures.

       At 3 and 6 months, the NC study failed to find any association between PCB levels and
growth or  morbidity. Lower psychomotor scores were observed on the Bayley scales at 5-7
months of age in the NC study. Although the MI study found a tendency towards poorer
performance on the Bayley scales at 5  months, the change was not statistically significant.  At 7
months, the MI researchers found that fish consumption and cord blood  PCB indices (measures
of prenatal exposure) were associated with decreased visual recognition memory and that this
effect was not mediated by neonatal factors (e.g., low birth weight or smaller head
circumference). The NC researchers, who performed more regular follow-up examinations,
continued to find psychomotor deficits  on the Bayley scales at 12, 18, and 24 months. However,
these effects did not persist beyond 2 years of age. The MI group observed memory and verbal
deficits on the McCarthy scales at 4 years  of age.

       Thus, the major difference in the effects observed in the two studies is in the functional
domains affected by prenatal PCB exposure. The MI researchers found clear associations
between prenatal  PCB exposure and cognitive deficits, whereas the NC researchers consistently
observed effects on  psychomotor development.

       The MI group measured neurobehavioral function  at  narrower intervals of age, which
may have improved  detection sensitivity.  Furthermore, the MI researchers chose a domain that

                  TABLE 4
At Birth

3-6 months

5-6 months

7 months

12, 18, 24


Gestational age
birth weight

BNBAS scores

Growth and

Bayley scales

Bayley scales


School grades
(grades 3-5)




YES -(6 mos;



YES -shorter
YES -lower
YES -smaller

& autonomic
# abnormal reflexes,
range of state
tendency toward
poorer performance
@ 5 mos., NOT


YES (prenatal
PCB & memory &
verbal function


is recognized as the most sensitive measure of infant cognition—visual recognition.  No
comparable tests were conducted in the NC study. Another reason for the apparent
inconsistency between the results of the studies may be that tests for various neurobehavioral
domains are not the same for different ages. Therefore, testing instruments may not be
comparable over time in the population. Alternative explanations  could include the possibility of
different congener distributions in the population or other unknown factors related  to inherent
differences in the design or conduct of the studies.

       •      Question: Is there sufficient epidemiologic evidence for risk assessment purposes?

       While methodologies and results differed in the two studies, there was general agreement
that the epidemiologic data suggest that  low-level PCB exposure affects both psychomotor and
cognitive endpoints.  However, differences in analytic methods, study design, and endpoints
measured over time make it difficult to directly compare these studies. Although
congener-specific PCB data would be preferable, quantitative risk assessment using  total PCB
measures  (with some assumptions) may be possible at this time. Based on the fact  that
congeners differ with respect to their toxicity, more precise evaluations of dose-response
relationships  for particular congeners or known mixtures of congeners would provide stronger
evidence for  a causal association between exposure to  PCBs and developmental toxicity.

       In terms of qualitative risk assessment, it was concluded that the studies were of sound
quality and that both were well designed and well executed. The fact that two independent,
well-designed studies showed that low-level exposure to PCBs was  associated with psychomotor
and cognitive deficits in humans suggests that prenatal PCB exposure is of public health concern.
However,  the epidemiologic evidence cannot stand alone, and must be appraised in terms of its
consistency with the results of animal studies and with the results of future epidemiologic studies.
Fein G.G., J.L. Jacobson, and S.W. Jacobson. 1984. Prenatal exposure to polychlorinated
biphenyls: effects on birth size and gestational age. /. Pediatrics 105(2): 315-320.

Gladen, B.C., W. J. Rogan, P. Hardy, et al. 1988. Development after exposure to
polychlorinated biphenyls and dichlorodiphenyl dichloroethene (DDE) transplacentally and
through human milk. /. Pediatrics 113:  991-995.

Gladen, B.C. and W. J. Rogan.  1991.  Effects of perinatal polychlorinated biphenyls and
dichlorodiphenyl  dichloroethene on later development.
/. Pediatrics 119: 58-63.

H.E.B. Humphrey.  1987.  The human population—an ultimate receptor for aquatic
contaminants. Hydrobiologia 149: 75-80.                            ,

Jacobson J.L., S.W. Jacobson, and H.E. Humphrey.  1990. Effects of PCBs and related
compounds on growth and activity in children.  Neurotox. Terat. 12: 319-326.

 Jacobson, J.L., S.W. Jacobson, and H.E. Humphrey. 1990a. Effects of in utero exposure to
 polychlorinated biphenyls and related contaminants on cognitive functioning in young children.
 /. Pediatrics 116: 38-45.

 Jacobson J.L., S.W. Jacobson, H.E. Humphrey, S.L. Schantz, M.D. Mullin, and R. Welch. 1989.
 Determinants of polychlorinated biphenyls (PCBs), polybrominated biphenyls (PBBs), and
 dichlorodiphenyl trichloroethane (DDT) levels in the sera of young children.  Am. J. Public
 Health 79: 1401-1404.

 Jacobson S.W., G.G. Fein, J.L. Jacobson, et al.  1985.  The effect of intrauterine PCB exposure
 on visual recognition memory.  Child Dev. 56: 853-860.

 Jensen, A.A. and G. Sundstrom. 1974.  Structure and levels of most chloro-biphenyls in two
 technical PCB products and in human tissue.  Ambio. 3: 70-76.

 McKinney, J.D., L. Moore, A. Prokopetz, and D.B. Walters.  1984.  Validated extraction and
 cleanup  procedures for polychlorinated biphenyls and DDE in human body fluids and infant
 formula.  /. Assoc. Off. Anal.  Chem.  67: 122-129.

 Parkinson, A., L.W. Robertson, and S. Safe.  1980. Reconstituted human breast milk PCBs as
 potent inducers  of aryl hydrocarbon hydroxylase. Biochem. Biophys.. Res. Commun. 96: 882-889.

 Rogan, W.J., B.C. Gladen, K.Y. Hung, S. Koong, Y. Shin, J.S. Taylor, Y.C. Wu, D. Yang, N.B.
 Ragan, and C.C. Hsu.  1988.  Congenital poisoning by polychlorinated biphenyls and their
 contaminants in Taiwan.  Science 241: 334-336.

 Rogan, W.J., B.C. Gladen, J.D. McKinney, N. Carreras, P. Hardy, J. Thullen, J. Tingelstad, and
 M. Tully. 1986a. Polychlorinated biphenyls (PCBs) and dichlorodiphenyl dichloroethane (DDE)
 in human milk:  effects of maternal  factors and previous lactation. Am. J. Public Health 76:

 Rogan, W.J., B.C. Gladen, J.D. McKinney, et al.  1986b. Neonatal effects of transplacental
 exposure to PCBs and DDE.  /. Pediatrics 109: 335-341.

 Rothman, K.J. and C. Poole.  1988.  A strengthening programme for weak associations. Int. J.
 EpidemioL 17: 955-959.

 Safe, S., L. Safe, and M. Mullin.  1985.  Polychlorinated biphenyls:  congener-specific analysis of
 a commercial mixture and human milk extract. /. Agric. Food  Chem. 33: 24-29.

 Schwartz, P.M., S.W. Jacobson, G. Fein, J.L. Jacobson,  and H.A. Price.  1983.  Lake Michigan
 fish consumption as a source of polychlorinated  biphenyls in human cord serum, maternal serum
 and milk. Am. J. Public Health 73: 293-296.

 Webb, R.G. and A.C. McCall. 1973.  PCB standards for electron capture gas chromatography.
J. Chromatogr. Set.  11: 366-373.      .

Yakushiji, T., I. Watanabe, K. Kuwabara, S. Yoshida, I. Koyama, and N. Kunita. 1979. Levels
of polychlorinated biphenyls (PCBs) and organochlorine pesticides in human milk and blood
collected in Osaka Prefecture from 1972 to 1977. Int. Arch. Occup. Environ. Health  43: 1-15.


                                 Carole A. Kimmel, Ph.D.
                           U.S. Environmental Protection Agency
                                     Washington, DC
       The animal studies work group discussed a number of issues related to the animal data
available.  Participants were reminded that the goals of the workshop were:

              (1)     to ascertain whether sufficient data already exist in available animal
       studies to determine the risk associated with exposure to PCBs for the developing CNS;

              (2)     if the available data are only partially adequate, to determine what
       additional information would be needed to conduct a risk assessment.

       In evaluating the data, a number of design issues were considered, as outlined in EPA's
Developmental Toxicity Risk Assessment Guidelines (1991) and as listed below:

       •      standard design considerations;

       •      several doses and vehicle controls;

       »      adequate numbers of animals;

       »      randomization of animals to dose groups;

       «      litter considered statistical  unit;

       »      replicate design;

       •      pharmacological challenge;

       •      tests with moderate background variability (coefficients of variation);

       •      battery of tests, testing of both sexes;

       •      critical periods include both prenatal and postnatal periods up to the time of
              sexual maturity;

       •      effect may vary depending on time and degree of exposure;

       •      effect may be transient/reversible  or permanent;

       •      between- and/or within-species homology for the  effect; and

       •     maternal factors as mediator of effect.

The discussions of this work group were based on questions raised in and comments received on
the pre-meeting issue paper by Dr. Diane Miller. The tables  from Tilson et al. (1990) were used
to review the data and to add any new information available.  Three questions were asked in
each case:

              (1) Are the effects due to PCBs?

              (2) Are they adverse effects?

              (3) Do any of the new data change the LOAELs and/or NOAELs?

       Table 1 summarizes the data available on neuromotor function. In addition to the data
reviewed by Tilson et al. (1990), data from Lilienthal et al. (1990) with Clophen A-30 in Wistar
rats (0.4 and 2.4 mg/kg/day) was considered.  These data indicated a NOAEL of 0.4 mg/kg/day
for PCB effects on activity in an open field. However, the lowest LOAEL was still .014
mg/kg/day,  from a study of Aroclor 1248 in Rhesus monkeys,  with no NOAEL.

       Although not unanimous, the work group consensus was that the data showed consistent
evidence of altered neuromotor function in all species  tested. Activity was the measure most
frequently used.  Although the direction of changes in  activity was not consistent, this may be
accounted for by differences in time of exposure and age at testing. The measures used in these
studies were considered to be screening measures, i.e., global measures of neuromotor function.
In addition, it was recognized that activity may be modulated by a number of factors other than
alterations  in the  nervous  system.

       Table 2 (adapted from Table 3 in the Tilson et al. 1990 paper) summarizes the data on
learning and performance  effects of PCBs. Three studies had been published since the summary
by Tilson et al. (1990). A study by Lilienthal et al. (1990) indicated changes in active avoidance
and visual discrimination  at 2.4 mg/kg/day, with  a NOAEL of 0.4 mg/kg/day. A study by
Lilienthal and Winneke (1991) indicated that these  effects were primarily due to prenatal
exposure. A study by Schantz et al. (1989) on Aroclor 1016 in Rhesus monkeys reported effects
on spatial and shape discrimination at 0.03 mg/kg/day,  with a NOAEL of 0.008 mg/kg/day.
Participants pointed out that there were differences in the types of tests used to measure
cognition in the different  species.  In  Rhesus monkeys, spatial discrimination was affected, and
the effects were different from those  caused by TCDD. In rodents, active avoidance was the
measure  most frequently used.

       It was suggested that some alterations  in learning and performance may be due to
differences in activity level. Also, in many cases, only male rodents were tested.  There were also
differences in the motivation used to  examine cognition between species; in primates, positive
reinforcement was used, whereas in rodents, negative motivation (shock) was most often used.
The work group consensus, again not unanimous, was that the data showed consistent alterations
in cognition/learning in all species tested. There were  some reservations expressed about the
monkey data because concurrent controls were from a different source than the treated animals.
Other issues were raised about cross-contamination from other  chemicals being studied at the
same time in the  facility.  In addition, one participant indicated that the amount of PCBs
consumed by the  monkey infants is likely to have been underestimated since these  animals begin
eating the mother's food at 2 months and increase their intake with age. Because of this, infant




w Z3

^ o

•M t?



0 0



o *-••
u ca

e_ -
o ex



• t_
rH -HY-
cd cuco

0} •#
c c: oT
cd cd 3

§S5 i-H
0) 0 C
COT- ca o

o o u

01 CO tO

C C £=
5 5 S
3 O

-rz o
cc m




3= co
O O}



rH 0
cd ex

0) CX
O.-— •

CO If)
o -
o — *




r* r
rH rH
cd cd

4-> 4-*
0) (V
cd cd

ICQ §co
OOl 00)
COT- CflT-

>, >>
•H -H
O 0
cd cd
<- *-

(O 0)
,<-; ,cz
0 0

rH rH
cd cd








- O
. 0)

CO 4->

CC— '





cd O)
t— tN-

> a>
E 0
-*-H w


CM in




4-> 0
CC —





at co

edo '








M ;.
>"S •
0 erf
cd a.


cd cd

' CX
to >
a. o
X M t-
o to«t-
i nj J2
if) "D • — '



•H cd
U_ E_




O O>
a> cz
H- E
4-> CO
•H 0)
cd cd
o e

otf CO

T- tOT-
ed cd O

c -

cd .•— .
O -H
^~ cd
i 01

/; 1
i :


rH (s.

>-» '
>o c
sa 5
e ex
-»-H CO

-CM >>
^- (D T- cd
T"" T3
cd cdO CM
•a-o-r- A

- -3
« t;'
= «
•H S .
w y*
o cd
CM +-> CO
cd a>
X 4-* 0
to o ^^
CM >>cd
i CdrH CM
T- "O ' — • CO


rH 5
£ £
o a
t. c
a> j
*< H

     O)      -HT-
     h~      rHO)
_    t»      •HO)
T-OT-      _lT-
   n4J o
   n o i-







C -M
H 0





cu --
o m


*-» *-*
£_ •
O Q.



rH -
Ctf •
•H Ctf




•H O

i— ^o
•H tO
e E

CN ^
i 0)

rH C
Cd O «
-K e CD
Cd E
•H co
t- 0 t_
CU 4-1 3


co LO
O •
• CM



CO >*
=3 CU
.E 0

O • • Tf
! 0 I I 10 CM T- • O

T- CO *"*
0 O
T- O O ^t CM CM CM
O O t- CM T- CO
*4* CO CO CO O)
S- O C) O) O)
O O T- i- f- •»-
CO T- O)
O) .,_ ^ - ^ fs, «
<0 (O T- •» CO • ' • • O) •
O) O) •» E Y— ctfctfctf ftf
•H CD -H 4-J -H • -H
-, « Ctf IE • tQ CU Q> CD fH CD
•H h- ca
rH rH-MotfctfO) -H -H-H rH
CtfCtfCD t-EEE4-»etf
t- -M O O O 0> J=
•4-* -f-J N >EE
CD « ~a Q. OS £_*r--H-Hctfctf
4-* CC CDET3 •« ' -H 4- -M 4-» "D "O
oCtf Q E-H-H 0)Cd CU 00-H -HC
OTJ t- O o O^CUQ. cOcd O CO
fH CC f"* rH -H CD ^ ^> "**• N £5 CU CU CD CD ^ ^* ^ """I
CO E CO ^-> O Cd O O) Cd *Hl > O "O O "O O Ctf O Ctf 4-J
£->H -M -O Cti CD CDCDCD
i CU W ' <$ E"O >4-» O t. O OT3 -H-O •H'O > ^ > +-1 •
CO DJ-H CO W oO4JrH 4-<4J O Q-OCl.O4-'»H4-r*-
C0.EO CO -> 3Ctf Ctf<- CtfSW Ectf -HCtf -HCtf Ctf«- Cd-*
.E ^r jr
4-» 4-» 4-J CO W M tO W M CO O
C C C >i >» >» >> >>>»>»CO

to cM^^tco-r- o o o m m
rHOiHrH^TrH rH rH
4^ X V V 4^ 4-» Ctf Cd Ctf4-'Ctf4-'
Ctfcuctfcdcdc04-' 4-J 4-'Cd4^ctf
•HCO-H-H^-H E'E E4-* E-H
£_ O i- t- CO 1- CD <3> CDCOCDC-
Q.-— ' O. O.Q.O.O.
*—. C CO CO
SO. 4-< CO— Cd — » ^-^ ^-»
— - M •— •* * °- E O O O Q-
E EOr- Q. in O^HO-H -H^-Q.
O.vO. E Q_CUT-4-'CM-l-'4-'<
Q. ^SQ.«-CL CM O XrHC_ Ctf Cd CtfCMO
^t ^f CO W Q. CO O O X 4-> X 4-J 4-« CO
coin JL in o CM T- o cot- to o co o
O' o'°'° "" ' T->.CU^}-CDCMcd CD*J-« rH
•CM .O «OO v O ictfjDiO)irHCMD)-LO ^
O'^' o%~'°*~'CNJ ^ ^ LO-O*»^CM— -v-'»-'COs-'Os— * Ctf
— v S* cT
Cd M tO O)
fH rH 3 O)
IH — ' 4-" t-
N— ' etf *— *
cd t.
CM , CO ctf L. •
O.O rH
0.-H Ctf
>> itf 4-<
CU 4->^ 01
rH CO CD 4->rH
S "* « CDCO C ctf =
cd -^ 4-1 f— eu co O
O co eyed -nEt.w
CD . i 0 I- rH m CD "H
W >» -H =J CD EC- Ll^n " t! cuH
3OrHSo) jz ed euosocc
tO^:-H edto OCO t- -H Et.4-»H-iCD
J=0=J O Q-Ctf -Hetf Q -H OE>-Ho4-»
CCEO rH C0t_ U. t. O 5 Ol-eU-H
E O 3CC4-" O
•H . O ed-M
ErHf- CUi-l'O
tO .4. O O 3 OrH e CD
*- ^ ± CO O"O 3-HT3
5 S g . 2 < ^ S g^ o-1«
^ i T oiH ^0-o.ccn
i- , Si. !c i? c OTCO-HM
% 0 = « oS ^ " |, "«"=^>
§ o S "3 £•§.§• Sro" " "1
° ° co c +•"•••< in tSooceccS
.< 4 SS £ 6-17 "" -° W.MJ»P.«

 monkeys probably take in a variable amount of PCB from food as well as in maternal milk, and
 these two sources may involve different mixtures of congeners.  These issues were addressed by
 the principal investigators, who indicated that cross-contamination with other chemicals under
 study was not a problem, and that tissue levels were measured in the infants to get an estimate of
 their exposures.  Following this discussion, most participants felt that despite shortcomings
 mainly related to issues concerning the source of concurrent controls, the monkey data were
 consistent with those in other species and should' be used for risk assessment.

       Data summarized more recently by Golub et al. (1991) also indicated the potential for
 effects of prenatal PCB exposure on thyroid and reproductive function, as well as other
 endocrine effects. Work group participants discussed these data, but felt that additional work is
 needed to more clearly elucidate the effects of PCBs on these systems. One participant raised
 the issue of using data related to PCB effects on wildlife, and it was agreed that these data
 should be considered in the risk assessment process.  The same participant indicated that studies
 are underway in which fish with defined concentrations of individual congeners and total PCBs
 are being fed to other animals.  These studies should be very useful in defining environmental
 exposures as well as the reproductive and neuroendocrine or neurotoxic effects of PCBs.

       The work group briefly discussed issues of pharmacokinetics, mechanisms of action and
 structure-activity  relationships in the context of the data on developmental neurotoxicity of PCBs,
 but came to no specific conclusions.   -

       The work group discussion about the quality of the available animal studies concluded
 that there is no reason to disqualify any of the data, although there was some dissent concerning
 the primate data  as indicated above. The work group decided that sufficient data were available
 for risk assessment.  This agreement was fairly strong for the specific commercial mixtures tested,
 but there was less agreement that the data could be used to draw conclusions about
 environmental mixtures.  Participants indicated that more information is needed on the congener
 make-up of environmental  PCB exposures, and suggested that data may be available from human
 studies. Comparisons of environmental  and commercial mixtures—and analyses of what crosses
 the placenta following exposure to both  types of mixtures—could be very helpful in determining
 the importance of exposure to various congeners.  One participant indicated  that there is a need
 for risk assessments on the commercial mixtures because of the potential for soil contamination
 from surface impoundments of discarded capacitors and other sources of PCBs.  The question
was also raised concerning  other contaminants—in  this case, for example, the polychlorinated
 dibenzofurans—and their contribution to any toxicity from such  sources.

       In terms of the  dose-response relationships, several uncertainty factors were discussed,
as follows:

       •      animal to human extrapolation;

       •      intraspecies  extrapolation;

       •      less-than-lifetime to lifetime exposure (the work group questioned whether this is
              a significant concern when prenatal exposure is judged the most sensitive period);

       •      the use of environmental enrichment (bioconcentration) factor(s) in the
              application of commercial mixture data to environmental exposures (the work
              group noted that it is not entirely clear which congeners bioconcentrate).

       There was also some discussion of whether  an additional uncertainty factor should be
applied to assessments using the rodent data since  the behavioral methods used to test rodents
are considered less sensitive than those used in primates.  The work group did not make any
judgments about the size of these uncertainty factors, or exactly when they should be applied.
Bowman, R.E., and M.P. Heironimus. 1981. Hypoactivity in adolescent monkeys perinatally
exposed to PCBs and hyperactive as juveniles.  Neurobehav. Toxicol. Teratol. 3: 15-18.

Bowman, R.E., M.P. Heironimus, and J.R. Allen.  1978.  Correlation of PCB body burden with
behavioral  toxicology in monkeys. Pharmacol Biochem. Behav., 9: 49-56.

Bowman, R.E., M.P. Heironimus, and D.A. Barsotti.  1978. Locomdtor hyperactivity in PCB
exposed rhesus monkeys. Neurotoxicology 2: 251-268.

Chou, S.M., T. Miike, W.M. Payne, and GJ. Davis.  1979. Neuropathology of "spinning
syndrome"  induced by prenatal intoxication with a PCB in mice. Ann. NYAcad. Scl,

Golub, M.S., J.M. Donald, and J.A. Reyes.  1991.  Reproductive toxicity of commercial PCB
mixtures: LOAELs and NOAELs from animal studies. Environ. Health Perspect.  94: 245-253.

Koja, T., T. Fujisaki, T. Shimizu, C. Kishita, and T. Fukuda. 1978. Changes of gross behavior
with polychlorinated biphenyls (PCB) in immature rats. Kagoshima Daigaka Igaka Zasshi, 30:

Koja, T., C. Kishita, T. Shimizu, T. Fujisaki, M. Kitazoro, and T. Fukuda.  1979. Effects of
polychlorinated biphenyls (PCB) on the gross behavior of immature rats and influence of drugs
upon them. Kagoshima Daigaka Igaka Zasshi,  31: 315-319.

Kreitzer, J.F.,  and G.H. Heinz.  1974. The effect of  sublethal dosages of five  pesticides and a
polychlorinated biphenyl  on the avoidance response  on Coturnix quail chicks.  Environ. Pollut. 6:

Lilienthal, H. and G. Winneke.  1991. Sensitive periods for behavioral toxicity of polychlorinated
biphenyls:  Determination by cross-fostering in rats.  Fundam. Appl. Toxicol.  17: 368-375.

Lilienthal, H., M. Neuf, C. Munoz, and G. Winneke.  1990. Behavioral effects of pre- and
.postnatal exposure to a mixture of low chlorinated PCBs in rats, Fundam. Appl. Toxicol.  15:

 Mele, P.C., R.E. Bowman, and E.D. Levin. 1986. Behavioral evaluation of perinatal PCB
 exposure in rhesus monkeys: fixed-interval performance and reinforcement-omission.
 Neurobehav. Toxicol. Teratology 8: 131-138.

 Overmann, S.R., J. Kostas, L.R. Wilson, W. Shain, and B. Bush.  1987.  Neurobehavioral and
 somatic effects of perinatal PCB exposure to rats. Environ. Res. 44: 56-70.

 Pantaleoni, G., D. Fanini, A.M. Sponta, G. Palumbo, R. Giorgi, and P.M. Adams.  1988. Effects
 of maternal exposure to polychlorobiphenyls (PCBs) on Fl generation behavior in the rat. Fund
Appl. Toxicol  11: 440-449.

 Schantz, S.L.,  E.D. Levin, R.E.  Bowman, M.P. Heironimus, and N.K. Laughlin. 1989. Effects of
 perinatal PCB exposure on discrimination reversal learning in monkeys.  Neurotoxicol. Teratol
 11: 243-250.

 Shiota, K.  1976. Postnatal behavioral effects of prenatal treatment with PCBs (polychlorinated
 biphenyls)  in rats.  Okajimas Fol. Anat. Jpn., 53: 105-114.

Storm, I.E., J.L. Hart, and  R.F. Smith.  1981.  Behavior of mice after'pre- and postnatal
exposure to Arochlor 1254. Neurobehav. Toxicol. Teratology, 3:  5-9.

Tilson, H.A., GJ. Davis, J.A. McLachlan, and G.W. Lucier. 1979. The effects of
polychlorinated biphenyls given prenatally on the neurobehavioral development of mice.
Environ. Res.,  18: 466-474.

Tilson, H.A., J.L. Jacobson, W.J. Rogan.  1990.  Polychlorinated biphenyls and the developing
nervous system:  cross-species comparisons. Neurotoxicol. Teratology 12: 239-248.

U.S. Environmental Protection Agency.  1991.  Guidelines for developmental toxicity risk
assessment; notice. Fed. Regist.  56: 63798-63826.


                                    Carole A. Kimmel
                           U.S. Environmental Protection Agency
                                  .   Washington, DC
       A comparison of developmental neurotoxic effects was made for humans and animals in
four categories of function. As shown in Table 1, there were generally very similar types of
effects in humans, monkeys, and rodents, despite differences in the exposures used (i.e., whether
they were commercial mixtures or environmental exposures). For example, delayed motor
development and altered motor control were seen in all species.                             .

       Cognitive function, was also impaired, although the specific types of test paradigms used
differed across species. In humans, impaired visual recognition memory and deficits in
short-term memory or attention were reported.  In primates, altered discrimination reversal and
operant performance were seen.  The tests used in rodents were less sensitive and were
considered more screening-level tests.  In these tests, alterations in operant performance, delayed
maze  learning and alterations in retention of avqidance acquisition were reported.  Thus,
impairment in memory and attention seemed to be a common finding in many of the studies.   ,

       In terms of motivation and arousal, alterations  in responsiveness or activity were also '
common findings in the species tested.  These effects may have had an impact on the changes
seen in cognitive function reported in several studies.                          ,      ,

       Interestingly,  even though the types of effects seen were similar, the effective doses and
NOAELs were quite dissimilar across species. Rodents were least sensitive, with an approximate
NOAEL of 2 x 10"1 mg/kg/day, based on studies using Aroclor 1254 exposure, whereas for
monkeys, the NOAELs were almost two orders of magnitude lower  (8 x 10'3 mg/kg/day), based
on studies using Aroclor 1016. Based on estimates reported by Tilson et.al. (1990), humans
appear to be most sensitive, with an approximate NOAEL for environmental exposures of 2.7 x
10"5 mg/kg/day—two  additional orders of magnitude below that for monkeys.  The differences in
species sensitivity may be related to variations in the sensitivity of the  testing paradigms used in
different species, and/or differences in the toxicity of the various commercial mixtures or
environmental exposures  used in various studies.


1 •
•4— •

cq "2
m s
Summary of
Laboratory A





ni i
o o
rH rl
Q) 0
•o o

Jj £j
•P -P
S g

v -d
W 0)
id a)
rH -P
(1) rH
O ft

0 O
rH rl
(D O
•d o
rl rl
-P -P
•d T3
(U (U
!>1 rt
id a
rH -P
01 rH
Q rt!


^4 M
0 0
-p -p
o o
e s
•d -d
0) 0
rH 4J
0) rH
Q <

M id
O D^
rl -rl O
id o
•P Q) -H
^ l—j | *
rl Q) 0) d
0) N -p O
ft Id Q) -H
0 6 r| 4J
T3 T! "d 01
(D 0) 0) -rl
rl >. rl 3
0) id d) o1
4J rH 1J 0
rH Q) rH (d
rt! Q rt!


rl "^
CD id
•p a
d -rl
oj 01
Itered op
Itered di
rt! rt{

1 recogniti
ro 4J -P
3 S-) -P
01 o id
•rl X)
> U) rl
•d -d
0) >i Q) >i
Vt Vl M V-l
•rl O -H 0
id S id S
ft Q) ft 0)

^ £5
a o
o S
CJ fe






Itered ac



""-•H '


"5i ,
^4 '
*• — .



I— 1

o ,
CJ 1~l
o £
^ 0
1(8 rt!
d |-n
o 5*
Tl -3
Q) M
01 3
r"i *M
^3 W

a 0)
§ 01

species; t
"' d
5 °
ay study in
6 exposure,


• using
j± 3
5 -P
^ 01

"Based c
based o:


                                   Jane Adams, Ph.D.
                                University of Massachusetts
                                       Boston, MA
       This summary of the deliberations of the work group concerning the identification of
research needs relevant to the human neurotoxicity of PCBs is organized around the original
questions posed to the work group by the Chair (see summary by Dr. Mary Prince, earlier in this
section). For ease of presentation, certain questions are jointly discussed.

       •      Question: Could positive and/or conflicting findings be due to incomplete control
              of possible confounding variables?

       As discussed by Dr. Prince in her summary of the work group's deliberations, traditional
sources of confoundment (maternal SES, smoking/ drinking status, weight,  selection bias due to
attrition) were deemed to have been adequately examined by previous studies.  The work group
recommended, however, that additional variables should be directly examined in future studies.
These include:

              (1)    measurements of pre- or perinatal levels of lead in cord blood and
       postnatal measurement of blood lead levels beginning at birth, again between 18-24
       months, and at regular intervals through the conclusion of PCB monitoring;

              (2)    similarly monitoring levels of methyl mercury;

              (3)    measuring iron levels during pregnancy, but especially postnatally; and

              (4)    monitoring thyroid functioning during childhood.

Measurements of lead and methyl mercury levels were recommended given the possible
contamination by these environmental agents that are known to be neurotoxic.  Iron status was
raised as a  possible confounding variable and, as pointed out by Dr. Prince, was generally not felt
to have confounded the findings of previous studies.  This  conclusion was based mainly on the
fact that all women in these studies received prenatal care, and iron 'deficiency would have been
detected and treated. In addition, there was incongruity between the nature of developmental
effects induced by iron deficiency and those associated with PCB exposure. Nevertheless,  the
low cost and possible benefit of measuring iron levels argued for its inclusion. Finally, the work
group supported  the monitoring of thyroid function because certain thyroid hormones may be
active in placental/fetal  transport of PCB metabolites.

       •      Question:  Might exposures to different congener distributions in Michigan and
              North Carolina account for differences in study results?

        The work group's discussion of this issue is presented in detail in Dr. Prince's summary.
 Generally, the work group felt it parsimonious to assume similarity in the congener profiles
 present in Michigan and North Carolina given the absence of data to suggest otherwise.  With
 respect to the need for the measurement of congener profiles in future studies of the
 neurotoxicity of PCBs, the work group was divided into two camps.  One opinion was that
 without knowledge of the specific congeners of greatest developmental relevance, the cost of
 congener-specific analysis would not be justified by the limited information to be gained. The
 other group  felt that the information to be gained when the relevant congeners are identified
 justifies the measurement of congener distributions now for use in the future. This group held
 that these samples  would allow later determination of congener profiles  and, therefore, better
 definition of differences in exposure characteristics between studies involving different human
 samples.  The research needed to determine which congeners are most active in the production
 of developmental toxicity was recognized to be extensive and expensive,  and it was deemed most
 appropriate that this research be conducted using animal models, despite their shortcomings.

        •     Question: What compartments might be measured to best relate prenatal
              exposure to postnatal endpoints?   .

        During both the Michigan and North Carolina studies, the sensitivity of analytical
 techniques precluded the detection  with adequate resolution of PCBs in  cord blood or placental
 samples.  This is now possible and the work group felt it important to  measure from these
 compartments as well as breast milk and maternal sera in future studies. The work group was
 split, however, with respect to how often the prenatal  samples should be taken. Many felt that
 cord blood, perinatal maternal serum, and postnatal breast milk assays were adequate. Their
 rationale was that PCBs accumulate over time and, during pregnancy, remain relatively constant
 in the maternal system. Thus, the exposure level of the conceptus would be relatively stable.
 The more teratologically-focused members  of the work group argued that sampling of maternal
 serum should be  conducted at multiple points during gestation to assist in defining periods of
 greatest vulnerability. It was acknowledged, however, that information about placental transfer
 dynamics  and placental/fetal pharmacokinetics at different stages of pregnancy as well as
 identification of the active agents of developmental neurotoxicity would be critical to this process.
 As a result, most work group members agreed that research in this area  will continue to depend
 on extrapolation (with difficulty) from animal models.

        •      Question: Do analytic techniques limit  our ability to characterize human

       As previously discussed, the  work group agreed that the sensitivity of available analytic
 techniques had been a problem in the past, but that newer techniques allow fully adequate
 characterization of PCB exposure.  It was said more than once that it is now possible to get
whatever degree of resolution  a research budget will allow. The work group suggested that the
 need for congener-specific analysis would depend on the purpose of any given study.

       •   •   Question: What functional domains should be examined in future studies?

       The work group  endorsed the continued use of a battery of tests,  with the selection of
tests guided by neuropsychological models that relate specific functional domains to specific
neural systems.  Given that existing  data suggest the psychomotor and attention/memory

processing systems to be vulnerable to PCB exposure, it was suggested that these systems, in
particular, should be further explored.  Such exploration should utilize more demanding tests in
order to challenge the functional integrity of these systems.  The work group also discussed the
importance of longitudinal assessment, not just in terms of assessing behavior at different ages,
but also in assessing behavior within similar domains at different ages so that meaningful
comparisons of changes in behavioral characteristics could be made over time.

       Also with respect to the nature of the tests selected for use in developmental
neurotoxicology studies, it was suggested that methods that measure-a "process" rather than just
the outcome of processing may provide greater sensitivity. For example, it may be better to
examine the rate of response acquisition on a measure of learning rather than simply noting
whether or not a criterion was finally met.

       The work group also suggested that investigators should include measures of
neurophysiologic function because  these indices are only minimally influenced by social/economic
variables.  The continued need for attention to maternal educational and occupational status was
endorsed, along with the need to include these data as covariates in all analyses.

       Finally, the work group discussed the fact that behavioral systems are not the only
systems subject to the effects of developmental toxicants, and that other systems warrant
inclusion in postnatal evaluations.  In the case of PCBs, the integrity of thyroid and immunologic
function may warrant investigation.


                                     Diane Miller, Ph.D.
                            U.S. Environmental Protection Agency
                                Research Triangle Park, NC
        One goal of the workshop was to delineate research needed to address risk assessment
 issues. The following is a summary of the discussion of research needs that took place in the
 animal data work group.

        The biggest data gaps identified were the need for developmental neurotoxicity
 data concerning:

               (1)    the mixture or mixtures of PCBs that occur in the environment and/or the
        mixture to which the developing organism  is exposed (i.e., the "exposure mixture"), and

               (2)    specific congeners.

 There are limited data concerning the developmental neurotoxicity of specific congeners or the
 developmental neurotoxicity of the mix of congeners likely to be involved in actual environmental
 exposure.  Congener-specific information is important because knowledge of how a specific
 congener acts is the only way to determine which of the congeners in a specific mixture is
 responsible for the observed developmental neurotoxicity.

        The importance of obtaining data that relates to both data gaps was recognized.  Further,
 the best plan of attack would be to obtain data in  both areas in parallel. That is, neither data
 gap was considered more important than the other.  Thus, inquiry should proceed from both a
 "top-down" and "bottom-up" approach.  The requirements  for the study of the "exposure mixture"
 were based on many of the concerns voiced by work group members regarding the design and
 conduct of studies that make up the existing database. Although many thought that additional
 information on commercial mixtures is not needed, it was noted that exposure to these mixtures
 may still occur because of waste-site cleanup, problems with disposal  of capacitors, and so on.
 Regional laboratories of the EPA still receive queries concerning the health  risks of exposure to
 commercial mixtures even though PCBs  are no longer manufactured.

       Many felt that some information  on the composition  of the "exposure mixture" should be
 obtained prior to initiating further research on the effects  of PCBs. Certain criteria were
 identified that should be used in determining the composition of the "exposure mixture" prior to
 actual evaluation of this mixture or mixtures for its developmental effects.  Some expressed the
view that the mix of congeners to which the developing organism is exposed may not be the same
 mixture that is found in the environment, although some participants predicted that the
composition of the two mixtures is likely to be similar. Others suggested that, whether exposure
occurs to a commercial mixture or an environmental mixture, what eventually accumulates in the
animal will be the same.  Others indicated that the environmental mixture certainly was not the
same  composition as any of the commercial  mixtures and that the possibility of bioconcentration
of certain congeners in the parts of the food chain  should be taken into account.  Most felt that
the composition of the "exposure mixture" should be  determined in part using bioavailability data

and not solely on analyses of soil samples or foodstuffs (fish, milk, etc.). However, some thought
that valuable information could be obtained by this type of analysis,  since wildlife and certain
human populations receive major exposure to  PCBs through fish.  Further, some information on
the developmental neurotoxicity of environmental mixtures has been obtained by feeding
PCB-laden fish from Lake Ontario to pregnant laboratory rats, followed by an evaluation of
offspring for behavioral alterations/developmental neurotoxicity.

       However, others suggested that even the mixture present in the environment is not
necessarily the one to which the developing organism is exposed.  Exposures having significance
for development can occur prior  to pregnancy, during pregnancy, or during lactation.
Environmental processes and maternal metabolic and pharmacokinetic factors shape the
composition of the "exposure  mixture." As a result, there  is no. substitute for an analysis of what
actually reaches  the developing organism and  when.  Consequently,  there is a great need to
determine the composition of this human prenatal and postnatal "exposure mixture" prior to the
characterization  of its neurotoxicity.

       Organs or tissues related  to the fetus or to delivery of nutrients to the fetus and/or the
infant should be characterized in terms of the levels  of various PCB congeners present. There
has been some characterization of the composition of the  PCBs in human breast milk.  Because
the composition  of this mixture is likely different from that found in the maternal  blood
compartment, the placenta, and the fetal blood supply, however, there should be an attempt to
determine the congener make-up of cord blood as well as placenta.  Cord blood levels would
need to be lipid-adjusted.  Some  thought cord blood would be a better measure of actual
prenatal exposure, since the presence of PCBs in breast milk does not necessarily  indicate
prenatal exposure. Breast milk content is determined by what has been stored in maternal fat
and mobilized along with fat stores at the time of parturition.  Some indicated that it would be
difficult to determine  actual exposures based on an analysis of placental tissue due to the low
lipid.content of this tissue.

       Further,  it is unknown what role metabolism occurring in the mother, the fetus, or the
infant plays in determining the actual mixture to which the developing nervous system is exposed.
Excretion of PCBs from the body occurs primarily by metabolism  in the liver and depends  on the
degree of chlorination of the  congener. Little information is available on the accumulation and
bioactivity of these metabolites during pregnancy and development.  To this  end there is a great
need for the development of  radiolabeled PCBs so that metabolic profiles can be determined.  It
should be determined whether a  particular effect is due to a specific congener or its metabolites.
Many felt that studies should be  conducted with radiolabeled parent compounds in order to
develop a profile of the parent compound versus metabolites in cord blood, placenta, the fetal
compartment, and breast milk. These studies could be conducted in a rodent species.   Others
cautioned that there are differences in PCB metabolism among the  commonly used experimental
species—the rat, the dog, and the monkey.  Others noted  that the bioactivity of some metabolites
could be determined using known structure-activity relationships, synthetic metabolites  (some are
available) and in vitro techniques  (e.g., tissue culture, study of binding to specific binding
proteins, etc.).

       Once the constituents of the actual "exposure mixture" were  determined, studies would be
conducted with this partially defined mixture.  Many felt that it would be best to make  up an
"exposure cocktail" of those congeners most commonly seen  in cord blood and focus on

  determining the developmental neurotoxicity of this mixture in a specific species, probably the
  rat. A high-quality study was defined to include the assessment of multiple endpoints related to
  the development of the nervous system  and other body systems suspected to be impacted by
  developmental exposure, strict attention to study conduct and design, and an emphasis on the
  determination of dose-response relationships.  Exposure should be chronic—that is, before
  pregnancy, throughout pregnancy, and postnatally.  Animals would be sampled from multiple
  litters at a large number of time points and multiple endpoints would be evaluated. All agreed a
  study of this caliber has not yet been conducted. Data generated in such a study could be used
  to identify critical periods and to select critical congeners for future research.  Such a study
  should not be delayed, despite the difficulty in obtaining a complete characterization of the
  "exposure mixture" and/or in  identifying the mechanism or mechanisms responsible for
  developmental alterations observed in the .available database. Work could begin based on the
  information that is currently available. For example, it is known that only 30-35 of the 209
  possible congeners occur in the environment,  thus eliminating 174 congeners from immediate

        Interpretation of existing data  regarding the comparability of developmental effects seen
  following exposure to different PCB mixtures  or congeners is compromised because such
  comparisons have not been made within a single study.  Thus, comparative data generated in the
  same study or at the least in the same laboratory utilizing the same endpoints was deemed
  essential in determining the comparability of effects. Further, replication of some of the existing
  human, rodent, and primate findings are all needed. In future studies, offspring of both sexes
 must be evaluated.  In many of the existing  rodent studies, only male offspring have been
 studied. All agreed that female cyclicity and possible interaction of sex hormones with chosen
 endpoints was not an adequate excuse for limiting evaluation to one sex.  The evaluation of
 female offspring is particularly important given recent findings concerning alterations in male
 sexual development  at both the CNS and reproductive levels following limited gestational
 exposure to dioxin.  Problems related to cyclicity could be overcome by appropriate attention to
 these issues in study design (e.g., noting the stage of cycle at the time of the study and
 determining whether there is an interaction  with the endpoint under study).

        Another area of intense discussion concerned the appropriateness  of various  endpoints
 for evaluation. Endpoints chosen for evaluation should include those that would  reveal the most
 sensitive effects as well as those that represent "screening" measures.  Although there was
 agreement that sensitive endpoints should be evaluated, many thought that it would be difficult
 to determine a priori which effects would be the most sensitive.  Some expressed the view that
 endpoints should be  selected on a mechanistic basis; others suggested that there is not enough
 information regarding mechanisms of PCB toxicity to allow mechanism to dictate selection of
 endpoints.  Most agreed that the available data suggest that there is more than one mechanism
 in operation, and that the available information could be used to dictate endpoint selection.
 Thus, endpoints related to the immune, thyroid, and reproductive systems would be included,
 since these systems have been  shown to be affected in either  adult or developmental  studies of
 the PCBs. There should be a link between endpoints chosen and the functional status of the
 animal; endpoints must not be studied in isolation.  For example, if serum levels  of T3 and T4
were used as markers of change in the  thyroid  system, there must also be an attempt to
determine whether alterations  in T3 and T4 lead to compromised thyroid function.  That is, it
would be important to determine whether these alterations result in a hypo- or hyperthvrdkK -
animal.                                                                     r    .

       Because the existing database concerning the developmental neurotoxicity of the PCBs
consists mainly of neurobehavioral data, there was much debate concerning which behavioral
endpoints should be used in future research. It was agreed that, as in other systems, there must
be an attempt to relate nervous system endpoints with the function of the system. There was
also agreement that many of the existing rodent studies .have relied on what can be considered
"screening" measures that can be affected by changes in systems other than the nervous system.
These studies have generally used global measures  of neuromotor function such as activity. For
example, many of the studies that have suggested PCBs to be developmentally neurotoxic.in the
rodent have,used motor activity as an endpoint, and changes in activity following developmental
exposure may also account for the changes in learning and memory endpoints observed in this
species.  However, primate work has also identified activity/reactivity and learning/memory as
areas affected by exposure to PCBs.   Many participants  felt that  it would be most useful to
examine endpoints that are more closely linked to specific aspects of the PCB toxicity, rather
than relying exclusively on more global screening measures.  There should also  be attempts to
design and execute  studies comparing the effects of PCBs (congeners or defined mixtures) on
multiple species using endpoints that assess the same functional domains in each species.  For
example, spatial memory tasks could be evaluated in humans, monkeys, and rodents.

       Another area of intense discussion was the  strategy to be  used in choosing congeners  for
study. All agreed that  more congener-specific  data are needed. The available data have been
derived mainly from work with commercially available mixtures.   The commercial PCB mixtures
are comprised of different congeners chosen from the 209 possible congeners, with the final
composition of each commercial mixture determined by  the physical and chemical properties
desired.  There are limited data concerning the general toxicity of most of the 209 possible
congeners, and even less data concerning the developmental neurotoxicity of these compounds.
However, most felt that enough information is available to group or classify congeners, with these
groupings then serving as a guide for future research.

       Congeners could be grouped in several ways, including similarities in their structure or
chemical reactivity. In the; structure-activity approach, congeners could be grouped according to
the number or type of substitutions  to the parent structure,  the coplanarity or non-coplanarity of
the substitute compound, and so-on.  Studies could then be  conducted to determine whether  any
link exists between a particular change in structure and a specific activity such as thyroid binding.
In some cases it would be possible to use in vitro systems to systematically select congeners or
specific mixtures of congeners for study. Congeners can also be grouped along activity or
reactivity lines. For example, there are a number of congeners considered to be dioxin-like, and
others appear to have  effects on specific hormonal or neurochemical systems.  Congeners have
also been grouped using structure-activity criteria.  Thus, studies  could be targeted to explore a
particular aspect of PCB toxicity, with a particular  congener.or several congeners serving as
"surrogates" for that particular  characteristic.  The available information concerning these systems
and brain development and function would be used to design more focused studies, with an
emphasis on endpoints related to the activity in question.

        Several different examples of the reactivity approach were discussed by work group
members.  For example, recent work has suggested that PCBs can have actions on dopaminergic
systems both in vivo and in vitro and that dopaminergic activity is linked to structure of the
congener.  Such information could be used to  design studies comparing the. ability of various
congeners or mixtures of congeners to affect the development of brain systems or

  behaviors/functions predominantly affected or controlled by dopaminergic pathways.  Other work
  has identified dioxin as having effects on male sexual development; no information is as yet
  available regarding female sexual development.  One or several of the "dioxin-like" PCB
  congeners could thus be chosen as surrogates and their effects on sexual development could be
  assessed using behavioral, biochemical, and structural endpoints linked to sexual development
  and function.

        Most participants thought that the bulk of new data concerning the developmental
  neurotoxicity of the "exposure mixture"  or specific congeners should be collected using a rodent
  model.  Once  adequate information has been obtained regarding sensitive endpoints and sensitive
  periods in the rodent model, follow-up work  may be required using a primate model. Issues that
  must be considered in future primate work include:

               (1)    choice of the primate (Rhesus, marmoset, etc.), which would be partly
        dictated by metabolic considerations (i.e., knowledge of differences among rodents,
        humans, and nonhuman primates in metabolizing the "exposure mixture" or a particular
        congener); and

               (2)    study design and conduct considerations to avoid problems that have
        prevented unequivocal acceptance of the  data from previous primate  studies in this  area.

 Although participants did agree that the available data from studies conducted with primates
 were sufficient at this time for risk assessment, they also felt these studies did have flaws.
 Problems to be avoided in future studies include cross-contamination of food between exposed
 and npnexposed populations, errors in the estimation of levels of compound to which treated
 offspring have  been exposed, and inadequate  or inappropriate selection of controls.

       Susceptible populations  were also discussed in the context of research needs. Research is
 needed to 'determine which stages of development are the most vulnerable to PCB insult. For
 some commercial mixtures and  particular congeners,  the prenatal period appears more sensitive,
 but it was agreed that there is limited information on this point. Also, there have been no direct
 comparisons of specific  congeners or mixtures during different periods  of development.
 Researchers must be cognizant of species differences  in the degree of development of the
 nervous system at particular time  points.  For example, the rodent and the primate brain are at
 different developmental stages during the period  of lactation, and the maturity of the brain  at
 birth differs in  rodents,  humans, and primates. Distribution of PCBs to the brain following
 developmental  exposure depends on the time  of insult.  Such differences pose problems in
 creating exposure scenarios that will mimic human exposure.  No solution to the  problem was
 reached, but it was agreed that these issues should be taken into account when designing studies
 of the developmental neurotoxicity of PCBs.  Susceptible periods of development might also
 include exposure of sperm or egg, and this possibility  should be considered when  evaluating
susceptible periods.




                    Pre-meeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research Triangle Park, NC
                      September 14-15,  1992

                           Jane Adams
                     Department of  Psychology
                   University of Massachusetts
                     100 Morrissey  Boulevard
                        Boston, MA 02125
Pharmacokinetics of PCBs

     I read this paper to obtain a general orientation and am'not
an expert in this area.  I have two comments on general
presentation and one on the stated objectives for the work group.
First, it would be helpful to provide the reader with a figure
showing the prototypical structure of PCBs in order to facilitate
visualization of various configurations.  Second, the paragraph
on placental transfer and body clearance is awkward perhaps due  .
to its brevity.  It implies significance to the quantity of
compound transferred postnatally vs. prenatally, rather than
addressincr differential substrate sensitivity.  A follow-up
statement on neonatal "load" obtained through the "milk deposit"
to the infant would also help mutualize the paragraph.

     The questions posed for the work group appear to represent
priority issues and to be comprehensive, but feasible.  Clearly,
the first streamlining step must be to agree upon the "outcomes
of interest".  In addressing the questions, it would promote
utilization of the information if the work group could provide
information on the metabolic and cellular activity of the
selected congeners, especially interactions with hormonal and
other neuroactive agents.  This will facilitate
predictions/understanding of the physiological and developmental
consequences of exposure at different ages,-as well as the
identification of individuals at particularly high risk.

PCB Developmental Neurotoxicity in Experimental Animals

     This paper captures issues of critical relevance and
priority and poses fundamental questions not just to an
understanding of the developmental toxicity of PCBs but to issues
in developmental toxicity in general.  The questions posed are of
great importance, but also of great scope.  It is likely that the
discussion will identify further research needs as opposed to
providing complete answers.  In the process, however, important
information will be synthesized from a broad base,and necessary
steps will be taken toward resolution of issues.   Comments on
specific questions are discussed below.

 Question 1. Interest is focused on the relationships between
 stage of exposure and expression of neurotoxicity and on
 anatomic,^neurochemical, and functional relationships.  Given the
 lipid mobilization that occurs at the end of pregnancy,  this
 period should not be overlooked as one of importance.  With
 respect to neural and functional relationships,  work group
 members will all be aware of the great complexities inherent in
 establishing these relationships.  The paragraphs on activity and
 learning measures suggest a recognition of important variables .
 but a desire for oversimplification.   As pointed out, activity
 represents a final common pathway for many contributing
 interactive system's: emotional reactivity,  inappropriate
 modulation of motor output,  sensory disturbances relevant to
 species-specific exploratory behavior,  and habituation processes,
 all^may influence activity level.  It is well-recognized that the
 environmental and temporal conditions of measurement as  well as,
 developmental age at measurement contribute to the measured
 activity level.   Therefore,  many variables  must  be considered to
 resolve apparent contradictions in reported activity levels.   It
 is therefore recommended that the panel evaluate data only from
 studies that meet a specified set of  design characteristics
 (sample size with respect to variance in endpoint,  etc)  and more
 standard approaches to activity measurement.   What may appear to
 be contradictory findings often are quite orderly when all
 appropriate variables are considered.   Likewise,  many
 relationships must be understood to make since out of
 neurochemical/functional relationships.   Certain neuroactive
 compounds (perhaps the'majority)  play a role  in  both inhibitory
 and excitatory systems,  so the system and receptor
 characteristics  must always  be kept in  focus.  Whole brain
 neurochemistry may not be insightful.

      Similar comments apply  to the interpretation of apparently
 contradictory findings on learning abilities.  An animal  whose
 fundamental deficit is increased reactivity may  be more reactive
 to adverse  stimuli and more  "motivated"  to  avoid them.  This  may
 facilitate  performance in some paradigms  and  reduce  it in others,
 often interacting with the nature of  the  response required as
 well  as  the stimulus  conditions.

 2.  Question number 5.   Part  of  question number 5  is  somewhat
 unclear:  is it the relationship between the affinity for  the
 adult brain and  the ability  of  the  congeners  to  disrupt CNS
 embryogenesis which is being addressed?   If so,  is the adult
 brain the target  tissue  to be examined or the developing  embryo?
 I  have assumed that  the  predictive  relationship being sought  is
 between  the affinity  for the adult  brain as a target tissue and
 CNS dysfunction as  an outcome measure following prenatal
 exposure.   Clarification of  this question will be helpful.

Developmental Neurotoxicity  of  PCBs in Humans

     Important issues have been raised regarding the
interpretation of human  findings and exposure characteristics of

the low exposure samples which have been studied.   While all of
the issues raised are clearly important, it may not be fruitful
to place priority focus on these issues and on the contradictory
data from studies in children, without first holding a discussion
of the appropriate role to be played by reliable findings of
disrupted infant behavior alone.  Since the infant and 4 year
behavioral evaluations are not analogous and since more demanding
tasks may be necessary to demonstrate effects in children, these
data should perhaps not be weighted more heavily than the infant
data.  One solid finding is that deficits in the competence of
infant performance are produced by prenatal exposure to PCBs.
Should not this effect be viewed as an important index of risk?
If compensation has occurred by the age of 4 with respect to
gross levels of behavioral performance, are we willing to say
that it has not been at some cost to other relevant processing
abilities not yet evaluated or perhaps not yet demanded by
childhood cognition?                  •

                      Pre-Meeting Comments for
            Workshop on Developmental Neurotoxic Effects
                  Associated with Exposure to PCBs
                     Research Triangle Park,  NC
                       September 14-15, 1992

                          Donald 6.  Barnes
                      Science Advisory Board
                U.S.  Environmental Protection Agency
                          401 M Street,  SW
                       Washington, D.C.  20460

 General Comments

 1.  Kudos

      The background materials for the workshop have  been well-'
 prepared and hold the promise of making this a focused,  effective
 effort.  Congratulations to the organizers and the issue paper
 writers.                                       ,

 2.  Fundamental considerations

     (I submit that research to date has implicated "PCBs"
 (variously defined)  as a neurotoxicant.  The challenge before us
 is  to^test this hypothesis with sufficient rigor that  we can
 identify  the agent(s)  involved and  assess to what, if  any,  extent
 environmental levels of "PCBs" pose neurotoxic risk--pre-,  peri-,
 or  post-natally—primarily.to human populations.

     _Scientists and regulators alike are  frustrated  by having
 sufficient information to be concerned, but  insufficient
 knowledge upon which to design definitive experiments  or take
 specific  regulatory action.   The sources  of  this frustration
 include a series of tough questions:

      a.    "Who is this guy called 'PCBs'?11

      1) Analytical methods

      Over the years  various  techniques  and technologies  have been
used  to determine the  amount  of  "PCBs"  in a  sample.  There has
been  considerable variation  in the  results reported by different


     One  output  of the workshop should be a  specific
recommendation that  the NIST or ASTM sponsor a round-robin
analytical testing of PCB  analytical procedure(s).   This could
either be a test  of  a single,  common procedure or a performance
test of the participants' preferred method.

     2)  Environmental samples,vs. commercial mixtures

     On the one hand, scientists generate data on commercial
mixtures of PCBs, using them to assess the potency and risks
posed by exposure to these mixtures.  However, scientists are the
first to say that the mixture of PCBs to which the vast majority
of humans are exposed can be significantly different_from the
commercial mixtures.  Therefore, treating (from a scientific or
regulatory point of view) environmental mixtures of PCBs which
have been "weathered", transported through the air or water,
and/or biotransformed in various organisms is an assertion that
should be challenged. •


     Therefore, the workshop should design criteria {and
experiments to test the criteria) whereby the similarity of
environmental samples to commercial mixtures of PCBs can been

     b.   "These fellows each has his own unique personality, but
          what happens when you get them together in a group?"

     Scientists have more recently tested for the presence and
effects of isolated PCB congeners, in analysis of environmental
samples and toxicology testing situations, respectively.
However, there is evidence that certain congeners can interact
additively, synergistically, or antagonistically, depending upon
the congeners and relative concentrations involved.


     The workshop should design an overall strategy of research
in PCBs that will integrate both the  "mixture testing"  (top-down)
approach and the "congener testing"  (bottom-up) approach into a
single coordinated research plan, with the goal of "meeting in
the middle" to help us understand the likely effects of exposure
to environmental samples.


     The workshop should try to reach consensus on the degree to
which inter-congener  interactions will cause mixtures of
environmental  "PCB"  samples to deviate from strict additivity.
In addition, the workshop should design a set of experiments
which would test the  consensus view.

     c.    "Who's responsible: the guy left at the scene of the
          crime, or  the guy who came and went"?

     The authors of  the paper often cite PCB residue data in
different organs of  the body.  Does the presence of these
congeners imply  that  they are cause, of the effect, or should the
cause more correctly  be ascribed to short-lived congeners that

 may have had their negative effect in the process of being
 eliminated?                                          <  •

      For example, ortho-substituted congeners have been found in
 the brains of affected animals (Seegal et al, 1990),  but  the
 rapidly metabolized 3,4,3',4'-TCB was found to affect dopamine
 and dopamine receptor binding sites (Agrawal, et al,  1981)
 Which is the real culprit?      •
 Specific comments

 General Toxicity of PCBs

      a.   DL vs.  NDL effects:  Is the distinction real?

      Do the workshop participants agree with the assertion that
 PCS endpoints can be divided between those  which are  "dioxin"-
 like (DL)  and those which are  non-"dioxin"-like  (NDL)?   If so,
 what are those respective endpoints?
 TEFs for PCBs do not  perform as well  as  TEFs  for
 chlorinated dibenzo-p-dioxins and dibenzofurans
     Note  that  the TEF approach is  not  fool-proof. ' For example,
the application of often-used TEFs  to the  congeners in Arochlor
1254 leads to predictions  of  carcinogenicity  that are not borne
out by  the tests (Safe,  1992).   This is may not be so surprising
since fewer of  the criteria of use  of TEFs  (Barnes, et'al, 1992)
are met in the  case of PCBs than they are  in  the case of

     c.    How does one handle the anticarcinogenic effects?
           Silberhorn acknowledges the "antitumor activity of
           PCBs".   How  does this,  and the antagonistic effects
           identified by Safe,  factor into  the characterization of
           risk?  Further,  low doses of 2,3,7,8-TCDD (Kociba,
           circa  1978 and NCI,  circa 1979)  seemed to indicate an
           anticarcinogenic effect at low doses for some tumors.
           Some researchers have  linked this to the effect of
           2,3,7,8-TCDD on  estrogenic activity and response.
           Similar  DL behavior  should be expected for certain
           PCBs.  What  are the  implications of this fact?

     d.     Is there more blood  in  those epidemiological turnips?
          Has a meta-analysis been conducted of the PCB epi

Pharmacofcinetics of  PCBs
We have to use some caution in pursuing the implied
reductionist approach of looking for congener-specific
causes for toxicity.  This may be true, but we should

          retain some healthy skepticism and design further
          experiments to test the, assumption.

     b.   Does this bundle of knowledge suggest therapeutic
          approaches to treating PCB toxicity in humans?

PCS Structure Activity Relationships and Developmental Toxicity

     a.   We should use this information to generate testable

     b.   We should strengthen the link between SAR and
          development neurtox.  What are some explicit
          experiments that would' test this hypothesis?

Developmental Neurotoxicity of PCBs in Humans

 -   a.   Can the workshop participants identify additional
          populations for study?

     b.   On a scale of 1-10 how strong is the epi evidence?  How
          does it compare to some other epi-driven effect; e.g.,
          tobacco and lung cancer?

PCB Developmental Neurotoxicity in Experimental Animals

     a.   Some of these effects occur only at high doses compared
          to environmental exposures?  How should this be
          characterized in the risk characterization section of
          the risk assessment?

     b.   How does fetal metabolism compare to adult metabolism
          for the congeners in question?

                     Pre-Meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                Associated with. Exposure to PCBs
                    Research Triangle Park,  NC
                      September 14-15, 1992

                         David Bellinger
                Neuroepidemiology Unit-Gardner 457
                       Children's Hospital
                       300 Longwood Avenue
                         Boston,  MA 02115
Developmental Neurotoxicity of PCBs in Humans

Question 1. As with all observational studies of free-ranging
populations, the possibility of incomplete adjustment for
confounding bias must be considered. Although such bias generally
is viewed as creating Type I errors (as would pertain to the
Michigan study), it should also be considered a potential cause
of Type II error (as would pertain to the North Carolina study).

     Several points for discussion regarding confounder control
in the two studies are as follows:

     a. As described in Fein et al. 1984 (J. Peds 105: 315-320),
     mothers of exposed and nonexposed infants differed in terms
     of several measured (and thus probably unmeasured)  factors
     such as alcohol and caffeine intake during pregnancy. In
     commenting on the sampling strategy employed,  Paneth (199?)
     (Health and Environment Digest, 4-6) argued that rather than
     randomly sampling the unexposed population, the
     investigators should have matched unexposed to exposed women
     on characteristics likely to produce confounding. He also
     argued that a much larger unexposed group would have
     increased statistical power,  providing more information
     about developmental experience among non-PCB exposed

     b. Although recommendations differ with respect to the best
     strategy for_identifying potential confounders, many
     biostatisticians feel that relying on p-values alone,
     especially one as restrictive as  p<.10,  increases the risk
     of missing true confounders (Dales & Ury,  1978; Kleinbaum et
     al.,  1982). In our lead work,  we  found that variables not
     related to either exposure or outcome at p<.10 can
     substantially affect the exposure coefficient  when included
     in a regression, model.  We have adopted a multi-pronged
     approach to identifying potential confounders.  We include,
     first,  factors that meet the  empirical criterion of
     association with both exposure and outcome at  a liberal p-
     value of <.25  and,  second,  factors that,  on the basis of
     subject matter considerations  seem likely
     antecedents/correlates  of the outcome of interest.  Inclusion

     of such factors may increase precision with which the
     exposure coefficient is estimated by reducing the regression
     error term. In Table 1 of Jacobson et al. (1990, J Peds) the
     R2 values  for models of McCarthy Scale scores are quite low,
     below 15%, when adjustment is made for only 3 control
     variables: maternal age, gravidity, and examiner. It seems
     likely that if adjustments were made for maternal IQ, family
     SES and other such variables, the R2  value would jump to 30-
     35%. The cord serum PCB estimate could be driven in either
     direction by adjustment for these factors.

     SES is reported to be a correlate of cord serum PCB in
analysis of some endpoints  (S. Jacobson et al., Child Dev 1985
study reporting on Fagan Test data) but not others (e.g., J.
Jacobson et al., J Peds 1990 reporting on the McCarthy Scales).

     In some early reports  (Fein et al., 1984), the MI
investigators report including as potential confounders only
those variables for which the incidence exceeded 15% within each
category. It is not clear if this restriction was carried over
into the analysis of later endpoints. Although low frequency can
pose estimation problems, especially in analysis of dichotomous
outcomes where one level of a variable perfectly predicts the
outcome, it does not necessarily mean that a variable cannot be a

     It may be informative to explore the sensitivity of the PCB-
development associations in the MI cohort to variations in
approach to confounder selection.

     c. It appears that some factors included as control factors
     in both studies may have been coded differently. This could
     result in differences in the extent to which PCB-develppment
     associations were adjusted for these factors in the two
     studies. For instance, alcohol use during pregnancy was
     apparently coded as absolute alcohol per day in the MI study
     and >1 drink per week/not in the NC study. If the raw data
     are available in a form that would permit receding so as to
     be consistent across studies, any impact of differences in
     coding strategy could be assessed.

     d. Neither study mentions assessing children's iron status.
     Given the similarities in the behavioral/cognitive
     correlates of iron deficiency and PCB excess, assessment of
     the confounding bias attributable to this factor seems
     warranted  (if these data are available).

     Another factor to consider is the extent to which the
cohorts differ in general level of developmental risk. Despite
efforts to adjust for other risks in regression analyses,
measurement error and other uncertainties make the task of
compensating for inequalities between cohorts in the nature and
severity of competing risks very difficult. Studies of other
toxicants seem to suggest that, effect modification by other

risks notwithstanding, detecting subtle toxicant impact is easier
if the level of background risk attributable to such competing
factors  is  low. Is  it possible that the MI cohort is generally at
lower risk  of developmental problems than the NC cohort?

Question 3. Only  if archived samples from the two studies are
available and can be analyzed by the same methods can this
possibility be evaluated.

Question 4. This  question can be answered only if we know the
degree to which the developmental/reproductive effects are
congener-specific.  If they are, and the congener distributions
differ in the NC  and MI cohorts, estimates of the nature and
magnitude of the  PCB-outcome associations observed will differ as
well. Unless serum  and milk samples from the 2 cohorts have been
archived, it would  seem that at this point only animal studies
can help to address the issue of the role of congener-specificity
in accounting for study differences. If no specificity is evident
among animals, this would suggest that we need look elsewhere for
an explanation.

Question 5. This  is a theoretical possibility and can be
approached  empirically. Measuring the various exposure indices
used by  either the  Ml or NC study on a fresh cohort will indicate
the degree  to which the indices are redundant. It seems quite
possible that they  reflect differ aspects of PCB kinetics,
accounting  for their differing associations with development.

Question 6. The basis for the apparent inconsistency between the
findings of the MI  and NC studies'cannot be determined until the
comparability of  congener distributions in the two cohorts is
ascertained and differences in terms of confounder
parameterization  and selection are resolved. It seems possible to
reject instrument differences as the in view of the fact that in
both studies the NBAS, Bayley Scales, and the McCarthy Scales
were used.  Since  only the MI study employed the more detailed
assessments of memory and information processing (Fagan Test,
Sternberg visual  search;and recognition memory test, and visual
discrimination task), it is possible that the findings in the NC
cohort would have been similar had these tasks been administered.

     In  fact, there are some rather striking points of similarity
in the performance  of the children in the two studies on the
McCarthy Scales.  In both, Verbal and Memory subscale scores were
the most' strongly associated with PCB exposure (although with
postnatal exposure  in the NC cohort and prenatal in the MI
cohort). The trend  was not monotonic in the NC study,  with
children in the highest exposure category (12+)  scoring higher
than children with  moderate exposures.  For both Verbal and Memory
scales,   however, it looks as though a threshold exists between
the first two categories (0-2 and 2-5).  It is not clear how many
observations are contained in the 12+ category.  Although they
appear to disrupt a clear dose-effect relationship,  perhaps there
are reasons not to  give:their data much weight.  The same trend

appears  to hold with respect to the DDE-McCarthy association. Is
this group unusual  in  some respect  (e.g., sociodemographic
characteristics,  feeding  style or duration)? In the MI cohort,
the children who  were  breast-fed longer had the highest 4 year
PCB body burden,  highest  quality rearing environment, and best
performance on the  McCarthy Scales. Under these circumstances,
incomplete adjustment  for rearing environment quality could
obscure  an inverse  association between PCB and performance. HOME
scores  (or similar  scores) were not obtained in the NC study, and
this could account  for this U-shaped relationship.

Question 7. The impact depends on the shape of the dose-effect
relationship between cord serum PCB and development  (i.e.,
whether  it's linear on the measured scale, log-linear, "hockey-
stick"), and whether the  detect level is less than, greater than,
or equal to any threshold level that may exist. If the detection
limit was equivalent in the two studies, the greater percentage
of detects in the MI sample would provide greater power to
discern  an association, should one exist, by increasing the
precision of the  risk  estimates associated with elevated PCB
levels.  If 66% of  the cord serum levels were below detection in
the MI sample, it seems unlikely that a log transformation would
normalize the distribution. Were regression diagnostics carried
out to examine the  adequacy of fit for this parameterization of
exposure? This issue is less relevant for the NC analyses, which
used exposure categories  (were these ordered or unordered
indicator variables?)

Question 8. Until issues  bearing on the comparability of the MI
and NC studies can  be  resolved, conclusions about the body of
epidemiologic data  cannot be drawn. It therefore seems premature
to attempt to relate these data to animal data.

PCB Developmental Neurotoxicity in Experimental Animals

Question 3. Several issues complicate drawing cross-species
parallels in the  nature of neurobehavioral changes, beyond those
pertaining to exposure characterization (congener, dose,  timing,
duration, etc.).  There is suggestive evidence that the form taken
by the putative effects of biologic insult change over time as
development ensues.  For  example,  low birthweight/premature
infants  may initially  manifest deficits in psychomotor skills as
infants  and toddlers,  but more cognitive,  specifically
linguistic, deficits as preschoolers. As another example,  in one
lead study, children with elevated blood lead levels at age 2
years manifested  deficits in visual-motor integration skills at
age 5 years,  but  deficits in more verbally-based skills at age 10
years. The extent to which this is attributable to differences in
the psychometric  characteristics of instruments available for
assessing children  at  different ages or to real evolution in the
manner in which the underlying pathologic process is expressed is
not clear. In any event,   this possibility complicates comparison
of different human  studies,  as well as any comparison of  human
and animal studies.

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research Triangle Park, NC
                       September 14-15, 1992

                         Robert E. Bowman
                     Harlow Primate  Laboratory
                     University of  Wisconsin
                       22 N.  Charter Street
                         Madison,  WI 53715
General Toxicity of PCBs

     Rhesus Monkey, Birth Weight, vs. Dose of Aroclor 1016

     Experiment.conducted by Dr. J. Allen arid Dr.. D. Barsotti
(then a graduate student) working at the U.W. Biotron.

     We sent data  to Dr. John Cicmanec, EPA', Cincinnati, who
arranged statistical analysis.

     Infants were  conceived and born to mothers fed 0, 25, 250' or,
1000 ppb A1016.  As usual, there was a gender difference in birth
weights  (male heavier), in which the only statistical dose effect
was a lower birthweight at 1000 ppb  (422 & 423 gm for the females
and males respectively).  At 0, 25, & 250 ppb of A1016, for which
doses the data gave no meaningful indication of ari effect of
A1016 on birthweight,  the 13 females averaged 479gm and the 12
males averaged 522 gm.  This was similar to the means of 478 gm-
and 501 gm we observed for 196 control females and 168 control
males born in Harlow Primate Lab from 1978-1988.   •         '  "'•'•'"'

     Rhesus Monkey, Endometriosis, vs. Dose of 2,3,7,S^TCDD

     Adult female  rhesus monkeys, 8 per group, were exposed to 0,
5 or 25 ppt TCDD for about 3.5 years (5 ppt) and about 4 years
(25 ppt).  This part of the experiment was conducted by Dr. J.
Allen, Dr. Merle Evenson and Dr. D. Barsotti  (then a graduate
student) working at the U.W. Biotron.  In early 1983, I then  '•""'•
received the monkeys and all the lab records on them, and have
conducted the experiment since the end of the TCDD exposure,
maintaining the monkeys at the Harlow Primate Lab.

     Starting in about 1989, at which time these monkeys were
about 20 years old, the first of these monkeys showed overt
indications of severe  endometriosis.  By early 1992, we had seen
4 cases of such endometriosis.  Then, with a collaboration
arranged by the Endometriosis Association, involving Dr. Dan
Martin and Dr. Paul Dmowski, all 17 monkeys surviving of this set
were subjected to  laparoscopy  (June 20, ''92),- and the incidence

and stage of endometriosis was assessed in each animal.  A
statistically significant regression was seen between severity of
endometriosis and cumulative dose of TCDD consumed by each
monkey.  See Fig. 1, attached.

     Note that this was as long delayed effect, emerging about 9
years post exposure to the TCDD, at roughly 8 half-lives post
exposure-.  The dose-effect is significant if tested between the 5
ppt and control groups, and is also significant if tested between
the 25 and 5 ppt groups.  Hence, it is not an artifact of one of
the groups showing an idiosyncratic, response.

     Via a collaboration between myself, Dr. Dan Martin, Dr.
Jeanne Becker, and Sherry Rier  (a graduate student with Dr.
Becker), we have a grant proposal pending to study immune and
hormonal function in these monkeys, seeking mechanistic features
related to dose differential induction of endometriosis in these

     Rhesus Monkey, Endometriosis, vs. Dosage with PCBs

     A literature search on endometriosis revealed to us that an
abstract had been .published reporting increased incidence of
severe endometriosis in rhesus monkeys treated with PCBs (cited
by Paul Dmowski:  Campbell, Wong, Tryphonas et al., Is simian
endometriosis an effect of immunotoxicity?, Proc. Ont. Assocj.

     I have been in contact with Dr. John Campbell and Dr. Ken
Ruehl, pathologists involved in this study, but have not seen
their data to know the correspondence between their observations
with PCBs, and ours with TCDD.  It appears the effect developed
•sooner in their PCB-treated monkeys, than in our TCDD ones.

Pharmacokinetics of PCBs

     No comments.

PCB Structure Activity Relationships and Developmental Toxicity

     No comments.

Developmental Neurotoxicity of PCBs in Humans

     No comments.

PCB Developmental Neurotoxicity in Experimental Animals

     My first comments address possible maternal factors.

     Rhesus Monkey, Effects of TCDD on Reproductive Success

     This experiment was begun by Dr. J. Allen and D. Barsotti.
Their data subsequently came to me, and I analyzed it,  along with

doing further work.  The females dosed at 25 ppt were
significantly impaired in the number of viable offspring they
produced, but those at 5 ppt were not.  See these data in
Chemosphere 18(1-6); 243-252, 1989.

     Rhesus Monkey, Infant Weight Gain during Nursing, vs
     Maternal Dose of TCDD

     In our TCDD-treated female monkeys, cited above, the second
and third cohorts of offspring bred from them gained
significantly less weight while nursing (about 10% less^ weight
gain), which was true at both 5 and 25 ppt.  From data collected
on cohort III, I suspect the TCDD treated mothers were impaired
in the amount of lipid in their milk  (three 25-ppt mothers were
studied, and their milk lipid content was 2.5%, whereas 3.5-4% is
generally reported).  However, the offspring rapidly gained
weight to control levels when weaned to normal diets.  The
literature on behavioral development under caloric restriction
suggest that this small an effect on body weight gain would not
likely produce significant behavioral effects.

     RhesusMonkeys, Cognitive Learning Tasks

     We studied much the same battery of learning tasks in TCDD-
exposed monkey offspring, as we studied in PCB-exposed monkey
offspring.  The pattern of behavioral effects is not the same.
This suggests significant differences in the mechanism of
toxicity between PCB mixtures and 2,3,7,8-TCDD.  Hence, it
suggests a significant amount of•PCB toxicity is not via the Ah
locus.  Of course, not all TCDD toxicity itself may be via the Ah
locus.  Regrettably, I do not have any notable insights to offer
at this time on this TCDD/PCB difference.

o    o  0



     Fig.  1.   The  relationship between the degree of
endometriosis  and  the  exposure to TCDD in our monkeys is best
shown as here,  by  linear  regressions of assessment  (via
laparoscopy) of the  severity  of  endometriosis by a  standard
clinical scale,  regressing versus the dose of TCDD.  The dose of
TCDD used  here is  essentially the cumulative TCDD dose  (dose x
duration), but virtually  the  same regression coefficient was
obtained by regressing versus the mean daily dose rather than
cumulative dose.   Moreover, the  linear regression was lower when
regressed  against  nominal dose rather than against  the assessed
dose of TCDD shown here,  a further indication of a  real effect of
the actual dose of TCDD.

     Reminder:   This endometriosis manifested itself at 7-1.0
years post-TCDD exposure.

     The left  hand panel  shows that the dose-effect function is
linear from zero TCDD  to  25 ppt  TCDD experimentally added to the
daily diet, doing  regression  on  all three groups.

    _The middle panel  shows there was a significant increase in
the incidence  of severe endometriosis between zero and 5 ppt TCDD
added to the daily diet,  at 0.05 one-tailed, comparing only the
control and 5  ppt  groups.  A  one-tailed test is justified here, .
because the controls are  so close to zero, there is no way in
this experiment to test for less endometriosis than that shown by
controls.                                              .

     The right  hand panel shows  that there was also a significant
increase in the incidence of  severe endometriosis between 5 ppt
of TCDD and 25 ppt TCDD in the daily diet, comparing only the 5
and 25 ppt groups.

     No one experiment can be definitive.  However,  when one sees
such a dose-effect function between successive pairs of dose
levels (groups), as well as overall,  it is highly likely that the
dose is the variable that accounts for the function.  .

     Neither the differential in number of offspring born to
these monkeys,  nor the differential in number of fat biopsies
done on them,  accounts for the increased severity of the
endometriosis in them,  for neither of these variables even came
close to correlating significantly with the severity of the

              Supplemental Premeeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research  Triangle  Park, NC
                      September 14-15,  1992

                          Robert Bowman
                       Emeritus, Psychology
                    Harlow Primate Laboratory
                University of Wisconsin - Madison
                     22 North Charter Street
                       Madison, WI  53715

                          Susan Schantz
               Institute for Environmental Studies
            University of  Illinois at  Urbana-Champaign
                     1101 West Peabody Drive
                        Urbana, IL  61801
     Dr. Moore's pre-meeting comments raise certain concerns
about the rigor of the Wisconsin monkey studies,that yielded the
offspring we tested in our neurobehavioral studies.  He concludes
that the published data suggest cross contamination between the
Aroclor 1016, Aroclor 1248 and PCB studies and thus invalidate
the studies for purposes of risk assessment.  Our comments
correct several inaccuracies in Dr. Moore's citing of the data.
He cites as his sources Barsotti (1980, PhD Dissertation),
Barsotti and Van Miller (1984) and Allen et al. (1976) .  The
comments below are responses to his points 1-6.

     1. Re "adequate controls."  Barsotti  (1980)  addresses this
issue  (pp 43-44), identified the groups of monkeys by date of
arrival in the lab and the nature of treatment (her Table 2-2),
and showed that the animals used in the control group did not
differ from those in the exposed-groups on menstrual cycle
length, hormone levels, number 'of conceptions or number of live
births  (her Table 2-4).  The conception rate and live birth rate
were at or near 100% in every group.  Since there was no evidence
of impaired reproductive capacity in any group, it is unclear
what Dr. Moore's concern is.

     2. Re assay levels of Aroclor 1016 in the monkeys prior to
dosing.  The mean of 0.69 ppm cited by Dr. Moore takes into
account only 5 of the 8 monkeys that were sampled  (those with
tissue levels above the detection limit).  When samples below the
detection limit are included, the mean is 0.44 ppm (undetecteds
set to the detection limit).

     3. Aroclor 1016 in infants at birth was reported as 1.54 ppm
in controls, nearly as high as the 1.65 ppm reported.in the
infants from mothers exposed to 0.25 ppm.  Again,  the stated mean
of 1.54 ppm is based on only 2 of the 6 control monkeys that were
sampled.  The other 4. controls were below the detection limit.

The mean for all 6 controls is actually 0/52 ppm (undetected^ set
to the detection limit).  The attached graph shows how these data
for the controls compare to those for the treated groups.

     4. "As discussed in the thesis there was a third treated
group in the original design that was never reported in the
published paper because of the cross contamination with PBBs.
The treatment group at issue was exposed to 0.025 ppm Aroclor
1016 in the maternal diet.  Their exposure to PBBs has been
acknowledged by the researchers and is documented in Barsotti
(1980).  The mixup occurred because PBB-containing feed was
mislabeled, not because the Aroclor 1016 feed was contaminated
with PBBs.  There is no evidence of cross contamination of the
diets (i.e. no evidence of PCBs on chromatograms from monkeys in
the other two Aroclor 1016 groups or on chromatograms from
samples of the Aroclor 1016 diets).  The unfortunate mixup with
one dose group does not negate the findings from the other two

     5. "GC tracings... reveal... congeners... from exposure to a
PCB other than 1016...probably Aroclor 1248...also studied in the
Wisconsin facility^"  Barsotti (1980) discussed this issue,
pointing out that the peaks in question (Webb McCall 125 and 146)
are also seen in control monkeys and probably result from
background exposure to higher chlorinated PCBs present in
standard monkey chow.  If there had been recent exposure of the
Aroclor 1016 and control monkeys to Aroclor 1248 several other
peaks representing lower chlorinated congeners present in Aroclor
1248 would also be visible on the GCs.

     6. The 1016 study cites 3.37 ppm PCB at birth in infants of
mothers getting 1 ppm in the diet, but the 1248 paper reports 2.8
ppm PCB in infants of mothers getting 2.5 ppm.  This compares
apples and oranges.  The 1016 data are for-PCBs in lipid,  while
the earlier 1248 data are on a whole tissue wet weight basis.
The numbers represent PCBs in skin samples taken from infants at
birth.  The lipid content of the samples was variable and
averaged well less than 30%.  Hence, the values are quite
consistent with the intended exposures.

                     Premeeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research Triangle Park, NC
                      September 14-15,  1992

                        John F. Brown, Jr.
                      Health Research Center
                     General Research Center
                         General Electric
                Corporate Research and Development
                     P.O. Box 8,  Building Kl
                      Schenectady,  NY  12301
General Toxicity of PCBs

     Since many of the statements made in this review have
already received comments from others, I shall confine my remarks
to just two points:

A.  Composition of the commercial mixtures

     It is stated, without reference, that the commercial PCB
products•(e.g., Aroclors) were ill-defined mixtures, varying
widely from batch to batch in congener content.  This is simply
not true.  The-gas chromatograms of different Aroclor specimens
produced over at least the last 20 years of production show
remarkably little variation in relative peak heights, which are
indicative of congener distribution.  The reported
quantifications of those peak heights may vary appreciably from
analyst to analyst, but not the peak heights themselves.  The
levels of PCDFs in Aroclor samples do show considerable batch-to-
batch variation, but -usually only within a range (i.e., 0-5 ppm)
that is, too low to have toxicological significance.

B.  Utility of occupational exposure studies

     I do not believe that any survey of the general toxicity of
PCBs should simply brush aside the substantial body of literature
that has accumulated regarding the effects of occupational
exposure.  The occupationally exposed study groups included many
individuals who had measured PCB levels 10-100 times those of the
general population, and hence can serve as far more sensitive
indicators of any effects of PCB exposure.  Thus, the general
absence of significant health effects [Fischbein et al., 1979;
Brown et al., 1991] in these occupational groups is an important
consideration.  Their tissue levels were also in the range of
those in many of the experimental animals, and hence provide a
basis for interspecies comparisons of effects.  Finally, as we
shall show in the next section, the studies of occupationally-
exposed populations can provide data on fundamental
pharmacodynamic relationships, including those needed to define

 the true significance of the statistical associations revealed by
 the epidemiological studies.

 Pharmacokinetics of PCBs

      Writing as one who, has been deeply involved in the subject,
 I found Birnbaum's statements of goals and issues to present an
 excellent review of the available literature and current
 concerns.  However, I do have some additional recent results that
 are highly pertinent to the questions raised,  and hence should be
 presented here.

 A.   Interspecies differences in PCB metabolism

      A PCB congener mixture basically constitutes a substrate
 test panel that can be used to categorize the metabolic
 selectivity pattern of an enzyme or organism,  the pattern being
 revealed simply by the loss of a particular set  of peaks from the
 gas chromatogram of the residual PCBs.   Such patterns have
 permitted the differentiation of numerous strains of aerobic PCB-
 degrading bacteria in soil and water specimens,  and of anaerobic,
 PCB-dechlorinating, microbial strains in aquatic sediments
 [Brown,  1987] .   Thus far,  we have also checked but the metabolic
 patterns of 55  species of aquatic fauna [Brown,  1992],  a variety
 of  birds and mammals (ongoing)  and over 200 PCB-exposed capacitor
 or  transformer  workers [Brown et al.,  1989,  1990].   In sharp
 contrast to the situation with bacteria,  all higher animals  show
 either or both  of just two metabolic patterns, designated P4501A-
 or  2B-like on the basis of their similarity to those effected by
 the individual  cytochrome P450  isozymes in vitro [Brown,  1992] .
 The P4502B-like metabolic pattern is shown by most crustaceans, a
 few fish species,  and almost all birds  and mammals,  including the
 mouse,  sheep, and normal  human.   Metabolism by P450lA-like
 isozymes,  which are regulated by the Ah-receptor,  is  seen in many
 fish species, some porpoises,  and,  along with P4502B-like
 metabolism,  in  a number of animal species,  including the rat,
 probably the monkey,  and also the chloracnegenic  human,  where the
 abnormal PCB metabolism was first recognized and  designated
 '^Pattern A"  by  Masuda in  1974.   Interestingly  enough,  this PCDF-
 induced  pattern never appears in normal (non-chloracnegenic)
 humans,  even in occupationally  exposed  individuals with serum PCB
 levels over  1,000  ppb.  Evidently,  in the human,  as  in  the mouse
 [Alvares  et  al.,  1982]  AhR-mediated enzyme induction by the
 commercial Aroclor mixtures is much more  difficult than in the
 rat.  This explains the differences between  the dioxin  toxic
 equivalency  factors (TEFs)  found for PCBs  in the  rat by Safe  and
 those found  in  the mouse by Birnbaum, and indicates that  a
 different  set of  PCB TEF-values  may be 'needed  for each  species.

     The  situation in  the  rhesus monkey regarding AhR-induction
by Aroclors  is  still  ambiguous.   The chromatograms in Barsotti's
 thesis and paper  show a P450lA-like metabolic pattern in  some of
the animals  dosed  with Aroclor 1016  (which does not induce P4501A

in the rat except after prolonged very high doses)  but not in
others.  These animals, as pointed out by Moore,  are now known to
have come from a group in the Wisconsin primate lab where
Aroclors 1016 and 1248, polybrominated biphenyl (PBB), and dioxin
(TCDD) were all under test simultaneously.  Subsequent
chromatograms showed some of the control monkeys carried.1016,
some of the 1016-fed monkeys carried PBB, and others of the
1016-fed and TCDD-fed groups showed the pigmentation
characteristic of TCDD-induced chloracne.  Thus,  it is uncertain
whether the P4501A-induction and the chloracne (both AhR-mediated
responses) seen in the monkeys resulted from cross-contamination
with Aroclor 1248 or TCDD.

     Since the offspring of these same monkeys were the ones used
by Levin and Schantz in their studies of cognitive behavior,
there is a corresponding ambiguity in determining whether the
impairment of learning ability should be attributed to the 1248
or the TCDD.  What can be said on the basis of the observations
on both the children of the yucheng poisoning victims and the
Wisconsin monkeys is that when an AhR-mediated response becomes
large enough to produce manifest chloracne symptoms,
neurodevelopmental effects may also occur.  Since such a response
in the human to PCB alone has not yet been seen in even the most
heavily exposed workers, the risk of such an outcome from low
level exposure must be negligible.

B.  Differences between Aroclors in accumulability

     The published data reviewed in Birnbaum's paper deals
entirely with the acute and subchronic pharmacodynamic behavior
of the PCBs, whereas the chronic behavior, or long term
accumulability, is of greater concern for risk assessment.  In
order to determine the chronic behavior of the various
commercially used PCB mixtures (e.g., Aroclors),  or of the
altered compositions found in environmental specimens, we have
drawn upon individual PCB congener clearance data from a variety
of sources, but especially from the worker populations under
surveillance at GE, and used them to calculate accumulation vs.
time curves for the various PCB compositions in both normal and
chloracnegenic humans  [Brown et al., 1992].  Figure 1 shows the
curves as log-log plots of relative Aroclor accumulation in
normal  (i.e., non-chloracnegenic) continuously.dosed humans
exhibiting the geometric mean metabolic clearance rates of our
capacitor worker population over the period 1977-1983.  It is
evident from the figure that some of the highest Aroclors are
mostly retained for life in the body, some of the lowest are
mostly cleared within days, and those in between show an
intermediate behavior, gradually approaching, but never quite
reaching, steady-state levels.

     Table 1 shows the calculated lifetime (70 year)
accumulations of the various Aroclors for both normal and
chloracnegenic humans.  It may be noted that the latter
accumulations are much smaller, owing to two additional processes


  O CO
  co 13
  E .2

  3 «
  o >
  < -t
Q. cd
0) Q.
  0) (0

• ^


5 § $2 g
O T- *- T-
r- O CO CM
O T- i- CM
CO CO t- i-
T- co ^ co
IO f**« T- O>
r^ CO 00 CO
10 O 0 0
SO CM 00
T- T- T- 1—
_ 0

in the chloracne patients, namely, induction of P4501A as well as
P4502B, and markedly accelerated non-metabolic clearance [Brown
et al., 1992].  However, if the clearance rates be normalized to
the values for Aroclor 1260, nearly identical sets of relative
human 70-year accumulation, or RHA(70),  values are obtained.
These show that the most widely used PCB product,  Aroclor 1242,
has only about 5% of the chronic accumulation potential of
Aroclor 1260, which is the current standard for risk assessment.
We now have a computer program for calculating accumulation
curves like those of Figure 1 from capillary GC analytical data
for any PCB composition, and suggest that such calculations be
used in risk assessment rather than presuming all Aroclors to
present equal risk.

     In order to perform similar calculations for chronic
accumulation in other species it is necessary to determine first,
whether the species is showing clearance behavior like that of a
normal or a chloracnegenic human, and second, how the absolute
clearance rates compare with those in the human groups.  Thus
far, we have found that Aroclor 1254-dosed mice show the same
(P4502B-like) pattern as the capacitor workers, but with a
15-fold greater clearance rate, whereas rats show the mixed
induction pattern of the chloracne patients and a clearance rate
roughly 15-fold faster.

C.  The use of composition vs. response relationships to
    distinguish among toxic mechanisms

     A different way of portraying the results of the
pharmacodynamic calculations is shown in Figure 2, where the
three curves terminating is the upper right-hand corner which
show the relative PCB accumulations after dosing for five days in
a mouse, two years in a rat, or 70 years in a normal human, are
plotted against the chlorine content of the Aroclor.  The curve
in the middle shows the relative dioxin toxic equivalency, as
determined by AHH induction, in the rat.  The curves on the left
are more hypothetical; the upper one shows the relative levels of
PCB metabolites that would be expected in a mouse after five
days' dosing if the metabolites themselves were not eliminated,
while the lower one sketches out possible relative levels of
metabolites that were themselves metabolizeable.

     At the August DIOXIN  '92 meeting, important new information
was presented on the structures of two classes of metabolites:
the methyl sulfones that accumulate in lung and liver tissue
[Letcher et al., 1992] and the hindered biphenols that accumulate
in serum owing to binding to the thyroxine binding protein
[Klasson-Wehler et al.,, 1992] .  In both cases, the strongly
accumulated metabolites were those that would be derived from
congeners occurring most abundantly in Aroclor 1254.  Thus, plots
of the relative accumulations of both these types of metabolites
should resemble that shown for AHH induction.



•*= "co
"5  3
 co  c
 CD  c
CM o

esuodsey eAi

      Having the sorts of model curves shown in Figure 2, it is
 possible to review the results of any-toxicological assessment
 where_several different Aroclors had been compared and
 immediately determine whether the activity vs. composition
 profile matches that expected for   (a) an effect mediated by a
 toxic metabolite, other than one of a selectively bound type  (b)
 one mediated by either the AhR receptor or a selectively bound
 metabolite, or (c) one linked to the simple accumulation of
 persistent PCB congeners, such as P4502B induction.

      When we did this for the Davis and Safe [1989] data for
 immunotoxicity in mice,  as indicated by the suppression of the
 PFC response,  we got a very nice match with the curve for
 five-day accumulation in the mouse,  indicating an effect mediated
 by some correlate of total accumulation.   This was quite
 unexpected, since Birnbaum has found this response in the mouse
 to be exquisitely sensitive to dioxin,  so that if there had been
 any tendency at all for the PCB mixtures  to act on the mouse.
 Ah-receptor we should have expected concordance with the curve
 for AHH induction.  Conversely,  for acute immunotoxicity in the   -
 rat,  as indicated by thymic involution, an abundance of data from
 Safe s  group does indicate correlation with the curve for AHH
 induction.   More  interestingly though,  all chronic studies of
 Aroclor effects in rats  show correlations for  persistence rather
 than TEQ.   The available data for hepatotoxicity,  tumor
 promotion,  tumorigenesis,  and carcinogenesis  [IEHR,  1991]  all
 seem to track  the curve  labelled "2  yr. rat."   A possible
 mechanistic basis  for the observed correlation between PCB
 accumulability and tumorigenicity may be  that  the tumorigenic
 ettects  in  heavily dosed rats  are linked  to induction  of P4502B
 rather  than 1A.   The  more persistent PCB  congeners  are all
 roughly equi-potent as agents  for P4502B  induction, which has
 been  correlated with  liver tumor  promotion in  rats  [Lubet  et
 S-L . ,  1989] .

      Finally  we have the  data of  the Seegal group  showing  that
 Aroclors 1016,  1248,  and  1260 all  seem to  produce quite  similar
 effects on  the  levels of dopamine  in the brains of  rats  and
monkeys.  If true, this would suggest a response vs. composition
 relationship like  that shown by the  dotted line on Figure 2
 implying^mediation by metabolizeable metabolites, i.e.,  not'those
of the_hindered biphenol type that are associated with the
thyroxine-binding protein of the serum, but perhaps other
biphenols instead.

D.  The pharmacokinetic basis for the statistical correlations
between PCB accumulation and neurodevelopmental indices observed
in epidemiological studies

     During the past decade there has been much discussion and
debate over the discoveries of statistical correlations between
the accumulated levels of PCBs in body fluids, such as serum,
cord blood, or milk, and various indices of children^ mental
performance.  Recognizing that any statistical association may
arise from covariance of the correlated parameter with some
confounding variable, much attention has been given, by both the
discoverers of the correlations and their critics, to the
possible contributions of confounders that might affect the
neurodevelopmental indicators.  The ongoing debate over this_
issue is presented in many of the premeeting and commentary issue
papers submitted by others.  Remarkably though, no attention
whatsoever has been directed at the confounding variables
affecting the other member of the correlated pair, namely, the
accumulation of PCB in the body fluid sampled.

      Table 1 shows that the average composition and accumulation
 characteristics of all the Aroclors produced were .approximately '
 equal to that of Aroclor 1248.  Figure 1 shows that a woman of
 child-bearing age, say 30 years, with a constant lifetime
 environmental exposure to a PCS mixture of that general
 composition,  would be fairly close to having reached a steady-
 state level of PCB in her body.  A steady-state level would mean
 that her rate of PCB clearance, mainly by metabolism, Vm,  had
 reached her rate of intake,  or dosage, Vd.  Now, at  any point  in
 time her rate of PCB metabolism will be directly proportional  to
 her body burden of PCB,   (PCB)b, and to her total  level of PCB-
 metabolizing cytochrome ,P450 enzyme activity,  (E)b,  as  expressed
 in terms of activity at unit PCB concentration in body fat, and
 inversely proportional to her total mass  of  body fat,  (F)b,  and
 to the infinite time accumulability of the particular PCB
 composition taken up,  Ac.  Thus, the near-steady^state
 relationship  present after -30 years' exposure can be described
 by the relationship:
  - Vm =
     For those disturbed by  the  slight  approximation  required
because of the not-quite-complete  attainment  of  steady-state
conditions, explicit  equations for the  curves shown in Figure 1
are given in Brown  et al.  [1992] and could be used to put^egn'. 1
into a precise form.  However, the resulting  equation would be so
complex as to obscure the basic  relationship  between  the major
variables shown in  eqn.  1.

     Proceeding on, we know  that PCBs are distributed among body
tissues and fluids  strictly  according to their content of neutral
lipids, L,  so that the  lipid-adjusted  PCB concentration in any
body compartment will be the same,  and, incidentally, may also
serve as a measure of the pharmacological availability of that
PCB for any PCB-mediated effect  [Brown  and Lawton, 1984].  Thus,
following the convention of  using  square brackets to  represent
concentrations,  we may write:
[PCB]SL = [PCB]s  -
    JSL    [L]S     (F)b

where  [PCB]SL represents the  PCB concentration in serum  lipid  and
[PCB]B and [L]s the measured  concentrations of PCB and neutral
lipids in the serum.  Combining eqn. 1 and 2, the terms for body
fat cancel out and we have:
FPTRl  ~
     This relationship indicates that the measured PCB level in
serum  (or in any other body fluid or tissue) must be directly
proportional to the mean intake rate and accumulability of the
PCB taken up, and also to the neutral lipid content of the
specimen sampled, and inversely proportional to the total level
of PCB-metabolizing enzyme activity in the body.  Thus,
covariance of a neurodevelopmental response with any one of these
four parameters could account for an observed correlation.

     Covariance of  [PCB]s with PCB composition,  as indicated by
Ac,  is certainly a possibility that should be checked,  and one
that should not be too difficult to assess if the original
analytical raw data is still available.  It is our experience,
however, that the PCBs in the general environment away from point
sources are fairly uniformly mixed, and we hence would hot expect
a great deal of variance in Ac.

     Covariance of  [PCB]S with environmental dose rate,  Vd, is
already known to be relatively unimportant from the Jacobson's
data on the Michigan populations, where, as pointed out by Paneth
[1991] the correlation coefficients  (r2)  for Vd, as estimated
from fish consumption, with maternal and infant serum PCB were
0.08 and 0.09, respectively. This means that variances in serum
lipid,  [L]s and/or P450 level,  (E)b,  along with those due to
random error, could account for as much as 90% of the observed
variance in mental performance.

     Covariance of  [PCB]S with [L]s   is another possibility that
should be checked, particularly since it has been previously
shown  to completely account for the statistical associations
between  [PCB]S  and serum triglycerides,  serum cholesterol,  SGPT,
and serum alkaline phosphatase  [Lawton et al., 1985], all of
which  were originally taken as evidence for PCB effects on liver
function.  The demonstration of  [L]s as a confounding variable
involved conducting the regressions using  [PCB]SL  rather than
[PCB]g as the  independent variable.  When this was done the
associations with serum triglycerides, cholesterol, and the
enzymes vanished.  What this showed was that the liver conditions
that led to the elevated serum enzymes and serum lipids caused
the serum PCB levels to rise in proportion to the serum lipids,

 rather than the PCB themselves having a causative effect on the
 underlying condition.

      In the case of the neurodevelopmental correlations, however,
 a much more likely source of covariance is in the parameter (E)b.
 The whole body P4502B level is obviously not an easy parameter to
 measure in a human population survey; however,  some measurable
 correlates of (E)b are available.   In our  initial  survey of  the
 associations between biochemical parameters and PCB levels in
 capacitor workers robust negative associations  between 1976 PCB
 levels and serum iron were observed.  These associations were
 reported without comment [Lawton et al.,  1985]  because at the
 time there seemed to be no way of explaining why PCB should be
 associated with iron deficiency.  Since then, we have been
 examining the possible associations between the various clinical
 measurements and the rate of PCB clearance over the 1976-1988
 period.   Again,  serum iron has emerged from the multiple
 regressions as a strongly correlated variable,  indicating a
 relationship between serum iron and PCB clearance rate,  which
 must be  determined by (E)b.  Evidently, iron-deficient subjects
 have lower levels of P4502B {an iron-heme  protein),  and hence
 lower clearance rates, and hence elevated  steady-state
 accumulations of PCB.

      What makes  this significant is that iron deficiency produces
 exactly  the same,  frequently transient,  deficits in cognitive
 performance as those that have been associated  with PCB
 accumulation [Scrimshaw,  1991].   Thus,  iron deficiency should
 account  for at least part of the correlations seen in the
 Michigan and North Carolina study groups.   Another possible  cause
 would be lead poisoning,  which produces  similar
 neurodevelopmental effects.   Lead is known to interfere  with
 porphyrin synthesis,  and hence should also lead to reduced  P450

    _  In  summary,  this  analysis  of  PCB pharmacokinetics  indicates
 a likely cause of  the  statistical  association between PCB
 accumulation  and neurodevelopmental  deficits to be  a  covariance
 between  PCB accumulation  and the  effects of either  iron
 deficiency, lead,  or other  agents  that can affect  cytochrome P450
 levels.   It also suggests that  the correlation  could  be made more
 specific by using  the  serum lipid  levels of lower  PCBs rather.
 than  gross serum levels of  total PCBs  in the regressions.

E.  Answers to questions posed

      *n view of the analysis presented in.the above four  sections
I believe that the four questions  posed may be  irrelevant to the
public health problem before us.   Instead,  we should be asking
now far we can go in using a suitably measured  PCB level  in
environmentally exposed populations as an  indicator of agents
affecting ferro-heme protein synthesis.

PCB Structure-Activity Relationships and Developmental Toxicity

A.  SARs for human neurotoxicity

     By way of cross-reference we simply restate the conclusion
of the pharmacokinetic analysis presented in section 2D above,
which is that the basic biochemical lesion in the affected
members of the populations studied would appear to have been a
disruption of ferro-heme protein synthesis, which produces
parallel effects on PCB accumulation and cognitive performance,
and hence leads to the statistical correlations originally
interpreted as a causative effect.  This conclusion renders
irrelevant the questions originally posed regarding the SARs for
that effect.

B.  SARs for other PCs-related effects

     Again by way of cross-reference, I call attention to section
2C above, including Figure 2, which shows how composition-
activity relationships may be used to distinguish among
alternative mechanistic hypotheses.

C.  SARs for reduction in dopamine production

     One important structure-activity correlation that was not
dealt with in the above section 2C arises  from the tissue culture
studies of Shain, Seegal et al. using various PCB congeners on.
PC12 pheochromocytoma cells to determine which congeners might be
responsible for the reductions in brain dopamine levels produced
in rats and monkeys by administration of Aroclors.  These are the
studies that led to the unexpected finding that the dioxin-like
non-ortho-substituted  ("coplanar") congener, 3,3',4,4'-tetra-
chlorobiphenyl was inactive in producing the effect, but instead
that the lower mono- and diortho-substituted congeners were

     The design of this study was marred by a fatal flaw which I
called to the attention of the investigators during a site visit
in October 1991.  This flaw arises from the fact that in all
cases the PCB concentrations used were above the reported
solubility limits  [Shiu and Mackay,  1986].  Under such conditions
PCBs form dispersions of microcrystals or  (usually) microdroplets
that gradually settle out on the bottom of the container, or  on
any layer of cells thereon.  Additional data  [Seegal et
al., 1992, Figure 2] shows that this is indeed what happened  in
the authors' experiments:  the cells harvested from the culture
dishes contained 50-300% PCB on a protein  basis, far more than
could have been internalized.

     It  is known that coating  cells  with inert fluids can depress
metabolic function via a purely  "physical" toxicity  [Bar, 1987].
It is also reasonable to assume that a separate oil phase in  the
tissue culture system would  remove any fat-soluble vitamins or
growth factors from  the medium, and  hence  also depress cellular

  functions.   Evidently,  one or both of  these  effects must have
  happened to the pheochromocytoma  (PC12)  cells,  except with  those
  congeners,  such as  the  3,4,3',4'-,  3,4,5,3',4'-, and 2,6,2',6'-
  chlorobiphenyls,  which  crystallize readily and  hence would  not
  oil out  on  the  cells.

      _Since  the  pharmacological availability  of  a substance  does
  not increase if present in quantities  above  its saturation  limit,
  the availabilities  of the  PCBs tested  must have been the same for
  all of the  non-zero test dosages used, and any pharmacological
  S£feS*s  produced_by internal  PCBs  rather than external coatings
  should be determinable  by  extrapolating the  dose-response curve
  back to  the (near zero) dose  corresponding to the solubility
  limit.   Inspection  of the  authors' Figure 2  [Seegal et- al.,  1990]
  shows that  any_such effect  of internalized PCBs on dopamine level
 must be  indistinguishable  from zero.   This is surprising, since
  the monkey  brains were  showing responses in  animals where the
 pharmacological availabilities (as indicated by the estimated
  levels in neutral lipid) ranged between about 80 and 300¥ lO'6
  their values in saturated solutions.  Thus,  the pheochromocytoma
 cells must be at  least  105-fold less responsive to internalized
 PCBs than monkey brain  cells in terms of the dopamine depression
 response, and hence worthless as experimental models.  The only '
 response to PCBs exhibited by this cell line is simply a physical
 response to external coatings of liquid PCBs  deposited from
 supersaturated media.

      The irrelevance of the tissue culture data does  not
 necessarily undermine Seegal and Shain's basic conclusion that
 their observed depressions in brain dopamine  levels were not '
 caused by congeners  with dioxin-like activity.  They have reported
 quantitatively similar effects for various Aroclors,  albeit  in
 aitferent animals. As  noted above,  in connection with our Figure
 ^, this is inconsistent  with an AhR-mediated  effect, which should
 be very much more pronounced with  Aroclors 1248  and 1254  than
 with 1016 and 1260.  At the same time, we must note that  Seegal
 observed  no  symptoms of  impaired neurological function in even
 the  most  heavily dosed monkeys (nor were they seen in  PCB-exposed
 capacitor workers; [Fischbein et al., 1979])  indicating  that the
 observedDepressions in  animal brain dopamine levels were within
 the  physiologically  tolerated range.  The authors' justification
 for  reporting these  observations as examples  of  true
 neurotoxicity was apparently the assumption that actual
 neurotoxic effects were  being observed  in  all the other studies,
 and hence that the mild  physiological responses  seen in their
 animals must have represented the initial  stages of such effects.

 Developmental Neurotoxicity of PCBs  in  Humans

     Solely  by way of cross-reference,  I point out the
     aJ°kin?t:L2  analYsis Presented  as subsection D of my response
bSt-iJSn1??*  ab°Vei'  1]?dicatinsr that the apparent correlation
between PCB  accumulation and neurodevelopmental deficits may well

arise from the fact that factors causing depression in. cytochrome
levels, such as iron deficiency, and possibly lead poisoning as
well, not only produce the observed cognitive defects in
children, but also elevations in serum PCB accumulation, owing to
retarded metabolic clearance. Thus, the measured PCB level should
be an indicator of an underlying biochemical lesion, not a
causative agent.

PCB Developmental Neurotoxicity in Experimental Animals

     Again, solely by way of cross-reference, I point out the
comments made in sections 2A and 3A above, indicating that true,
albeit quite small, cognitive defects associated with exposures
to TCDD, PCDFs, and/or PCBs have only been seen in monkeys or
humans showing manifest chloracne-symptoms and P4501A induction,
indicating that they were undergoing AhR-mediated responses.
Because of the wide variation among animal species in the
sensitivity of their AhR-mediated response symptoms to the
commonly encountered PCB-congener mixtures  (e.g., Aroclors),;it
is quite possible that some animal species may indeed response
via this mechanism to PCBs presented at modest levels. In the
human, however, there is no evidence of AhR-mediated responses  •
(e.g., chloracne or PCB clearance pattern alterations indicative
of P4501A induction) even in the most heavily exposed workers,
where serum PCB levels ranged above 1,000 ppb. Accordingly, there
is no basis for assuming that an AhR-mediated neurotoxic response
to PCBs presented at environmental levels would occur in humans.


Alvares, A.P., Ueng, T-H, and Eiseman, J.L.  (1982)
Polychlorinated biphenyls  (PCBs) inducible monooxygenases in  •
rabbits and mice: Species and organ specificities. Life Sci. 30,

Bar, R.  (1987) In Biocatalysis  in Organic Media  (C. Laane, J.
Tramper, M.D. Lilly, Eds.), Elsevier, Amsterdam, The Netherlands",
pp.  147-156.

Brown, J.F. Jr., and Lawton, R.W.  (1984) Polychlorinated biphenyl
(PCB) partitioning between.adipose tissue and serum. Bull.
Environ. Contam. Toxicol. 33, 227-280.

Brown, J.F. Jr., Bedard, D.L.,  Brennan, M.J., Carnahan, J.C.,
Feng, H., and Wagner, R.E.  (1987)  Polychlorinated biphenyl
dechlorination in aquatic sediments. Science 236, 709-712.

Brown, J.F. Jr., Lawton, R.W.,  Ross, M.R.,  Feingold, J., Wagner,
R.E., and Hamilton S.B.   (1989) Persistence  of PCB  congeners in
capacitor workers and yusho patients. Chemosphere 19, 829-834.

 Brown, J.F. Jr., Lawton, R.W., Ross, M.R., and Wagner, R.E.
 (1990) Serum PCS as a permanent record of PCB exposure and
 response.  In Proceedings:  1989 EPRI PCB Seminar (G. Addis,
 Ed.), Electric Power Institute, Palo Alto, "California, •pp. 9-49-1
 through 9-49-5.

 Brown, J.F. Jr., Lawton, R.W., Ross, M.R., and Feingold, J.
 (1991)  Assessing the human health effects of PCBs.  Chemosphere
 23,  1811-1815.                     ••'.-..'.

 Brown, J.F. Jr.  (1992)  Metabolic alterations of PCB residues in
 aquatic cytochrome P450IA- and P450lIB-like activities.  Marine
 Environ.  Res., in press.

 Brown, J.F. Jr., Lawton, R.W., Ross, M.R., O'Donnell, M.K., and
 Hamilton,  S.B.  (1992)   PCB accumulability versus toxic
 equivalency as quantitative indicators of relative risk.
 Organohalogen Compounds 10, 301-304 (extended abstracts of DIOXIN
 92  Symposium) .

 Davis, D.,  and Safe,  S. (1989) Dose-response immunotoxicities of
 commercial  polychlorinated biphenyls (PCBs)  and their interaction
 with 2,3,7,8-tetrachloro-dibenzo-p-dioxin. Toxicol.  Lett.  48,

 Fischbein,  A.,  Wolff, M.S., Lilis,  R.,  Thornton,  J.,  and
 Selikoff, I.J.  (1979) Clinical findings among PCB-exposed
 capacitor manufacturing workers.  Ann.  N.Y, Acad.  Sci . 320

 IEHR.  (1991) Reassessment  of liver  findings  in five  PCB studies
 in rats.  Institute  for Evaluating  Health Risks,  Washington,  DC,
 July 1.

 Klasson-Wehler,  E., Kuroki,  H., Athanasiadou,  M.,  and Bergman,  A.
 (1992)  Selective retention of hydroxylated  PCBs  in blood.
 Organohalogen  Compounds 10,  121-123  (extended abstracts  of  DIOXIN
 '92  Symposium) .         ,

 Lawton, R.W., Ross, M.R.,  Feingold,  J.,  and  Brown, J.F.  Jr.
 (1985) Effects  of PCB exposure on biochemical  and hematological
 findings in capacitor workers.  Environ.  Health Perspec
 60,165-184.                                               ,

 Letcher, R., Norstrom,  R.,  Bergman,  A.,  and Muir,  D.  (1992)
Methyl sulfone-PCB and  -DDE metabolites  in polar  bears — •
Comparison  to parent compounds  in the diet. Organohalogen
Compounds 8, 357-360  (extended  abstracts of DIOXIN 92 Symposium) .
                   R'W" Ward' J'M" Rice' J-M" and Diwan, J.M.
(1989) Induction of cytochrome P450b and its relationship to
liver tumor promotion. J. Am. Cell Toxicol. 8, 259-268.

Schrimshaw, N.S.  (1991) Iron deficiency. Sci.  Am. 265(4), 46-52.

Seegal, R.F., Bush,  B., and Shain, W. substituted PCB congeners
decrease dopamine in non-human primate brain and in tissue
culture.  Toxicol. Appl. Pharmacol. 106, 136-144.

Seegal, R.F., Bush,  B., and Shain, W.   (1990) Lightly chlorinated
ortho-substituted PCB congeners decrease dopamine in non-human
primate brain and in tissue culture.  Toxicol.  Appl. Pharmacol.

Seegal, R.F., Bush,  B., and Shain, W.   (1992) Neurotoxicology, of
ortho-substituted polychlorinated biphenyls. Chemosphere

Shiu, W.Y. and Mackay, D.  (1986) A critical review of aqueous
solubilities, vapor pressures, Henry's Law constants, and
octanol-water partition coefficients of the
polychlorinated-biphenyls. J. Phys. Chem. Ref.  Data 15, 911-929.

                      Pre-meeting Comments  for
            Workshop on Developmental  Neurbtoxic Effects
                 Associated with Exposure to PCBs
                     Research Triangle Park, NC
                       September 14-15, 1992

                            Theo  Colborn
                       • World Wildlife Fund
                   1250 Twenty-Fourth Street,  NW
                       Washington, DC 20037

      The major topics for the pre-meeting issue papers were well
 chosen and presented.  As a result,  the selection of background
 material provided a solid foundation on which to address the
 primary discussion questions:  risk assessment  concerning PCB
 developmental neurotoxicity; and the scientific uncertainties and
 research recommendations relevant to prenatal and perinatal
 exposure to PCBs.   One topic, however, that has only recently
 begun to receive recognition as  an issue of concern,  is  the
 endocrine disruptive effects of  a large number  of chemicals,
 including PCBs.   With this in mind,  the following comments are
 presented for further consideration:

      (1)  The endocrine disruptive effects  of  PCBs need more
      attention,  and in light of  the  focus  of  this Workshop,  more
      discussion  and research is  specifically  needed on the
      possible neuroendocrine effects  of  PCBs.

      (2) _A body  of literature from the field  of wildlife
      toxicology  and epidemiology supports  the hypothesis that
      PCBs are functional teratogens.

      (3)  In the  real world,  exposure  to  PCBs  commences
      prezygotically and continues  throughout  one's lifetime.  .
      Wildlife provide a  real world animal  model of exposure  from
      the  pregestational  experience through gestation  and
      maturation  and support the  hypothesis that PCBs  are
      neuroendocrine disrupters.

Neuroendocrine effects of PCBs

      The  Jacobsons and Rogan studies  on  PCB exposed offspring
furthered_the use  of functional  endpoints  for measuring  changes
in cognitive  and motor development.   Future research  in  human
behavior  will undoubtedly lead to  the creation and acceptance of
more  tests  to measure  other difficult-to-recognize functional
deficits  as well.   In  light  of the in vivo and in  vitro  evidence
concerning  the endocrine  disuptive effects of PCBs  (Ford and
Cramer, 1977,  and  Shain  et  al, 1991,  among reprints in the issue
paper on"PCB  Activity  Relationships and Developmental Toxicity";
Dieringer et  al.,  1979; Kubiak et al., 1989; Gray  et al., in
press), a series of  endpoints or landmarks of endocrine

development are needed to determine if, indeed, PCBs have an
effect on maturation and fertility in humans.

     Landmarks (biologic markers) of sexual development should
become standard measures of developmental neurotoxicity.   These
might include ontological events that are expressed over
discreet, narrow time frames during .maturation.  Biologic markers
of neuroendocrine toxicity are needed that can also serve as
early predictors of long-term delayed effects that may not be
fully manifested until adulthood or senescence.  These markers
might be events or changes that take place at the molecular,
cellular, or physiological level, not necessarily at the
oganismal or behavioral level.

Wildlife as a model for PCB neuroendocrine toxicity

     It is.pointed out in the issue paper on "Pharmacokinetics of
PCBs" that metabolism of PCBs operates via the same mechanism in
fish, birds, and mammals.  It is also recognized that the avian
and mammalian endocrine systems develop similarly.  Wildlife,
therefore, should provide an excellent model for real-world human
exposure to PCBs including the antagonistic, additive,
synergistic, and complementary effects of PCB congener mixes and
concommitant contamination by other compounds that affect the
same systems      .     .      •

     The obvious effects 'of egg-shell thinning and overt
mortality among wildlife.have abated in recent years as the
result of restrictions placed on the use and production of a
number chemicals, including PCBs.  However, other; less obvious,
but equally devastating health effects are still stressing
freshwater and marine wildife populations dependent upon
contaminated fish.  These effects are most often manifested in
the offspring of the exposed animals and many of the effects are
endocrine related.  A number of independent wildlife studies
using biologic markers in aquatic birds and marine mammals have
demonstrated associations between PCBs and concentrations of
plasma retinol, and thyroid hormones; concentrations of highly
carboxylated hepatic porphyrins  (HCHPs); AHH and EROD activity
reported as dioxin toxicity equivalents; and T4/T3 ratios.
Thyroid activity and HCHP levels in herring gulls have been
plotted over a geographic pollution gradient.

     Following a decade of unregulated dumping of industrial and
agricultural chemicals into the Great Lakes, major wildlife
population crashes commenced in the mid 1950s within the system.
In each case, the affected populations held significantly high
concentrations of PCBs and other chemicals.  Marquenie and
Reijnders estimated in 1989 that less than 1% of the global
production of PCBs had reached the oceans.  Major marine mammal
die-offs in the Northern hemisphere commenced in 1987, and like
the Great Lakes scenario the affected populations held
significantly high concentrations of PCBs and other
organochlorine chemicals.  Mortality in most cases was attributed

 to viral infections; a new canine-distemper-like virus specific
 for seals and another for dolphins.   Only toothed whales that    ;
 consume fish have exhibited similar die-offs.   Crude;evidence
 suggests that the immune competency of the animals in the
 affected populations is compromised.  There is much disagreement
 over whether the mammal immune systems are compromised as a
 result of viruses or contamination.   Preliminary studies with
 dolphins suggest that immunological  changes are associated with
 the presence of organochlorine chemicals.   Reijnders (1986)  found
 an association between PCB in fish that common seals consumed and
 reproductive failure around the stage of implantation in the
 seals.  He induced the same effects  in American mink using
 equivalent diets of fish and PCBs (25ug PCB/day).

 Real-world exposure to PCBs and other chemicals

      Hindsight has taught us that we have  not  known.what to look
 for when dealing with developmental  toxicants.   Gross  birth
 defects,  acute mortality,  and cancer have  held our attention
 while other_less-visible effects have been overlooked  in both
 human and wildlife epidemiology.   Experience has  revealed that
 the least sensitive (higher-dose)  endpoints can oftentimes mask
 more sensitive (lower-dose)  endpoints.   For example,  in 1983  a  ',
 cross-disciplinary team of researchers  discovered that wasting
 among Forster's terns chicks on an island  in Lake Michigan
 commenced during embryonic development  and led to 35%  mortality
 by day 17 (Kubiak et al,  1989).   Hatchability,  birth weight,  and
 weight gain were below normal when compared with  an inland colony
 not'dependent_upon Great Lakes fish.   In 1988  under lower
 exposure conditions,  wasting appeared not  to commence  until  day
 17 when young chicks began to die, and  emaciation led  to the  same
 mortality by day 31 (Harris,  1990).   In this case,  the onset  of
 the most sensitive endpoint,  wasting, was  delayed for  more than
 two weeks.   If hatchability had been the only  endpoint  measured
 and the researchers had returned home early in  the study,  the
 same incidence of chick mortality and overall  effect on the
 population that had been observed in 1983  would have been missed.
 Egg AHH enzyme induction activity was 2175 dioxin  TEQs  in 1983
 (22.2  ppm PCB med.)  and 913  dioxin TEQs  in the  1998  (7.3  ppm  PCB
 med.)  —   95% of the enzyme  activity was attributed  to  PCBs.   In
 1983,  an egg-switching component  of  the  study revealed  that lack
 of parental  concern contributed to poor  survival of  the
 offspring, as well  as  wasting.

 Other  PCB  endocrine-related  comments

     A number of  the reprints provided with  the issue papers made
 reference  to  "sex skewed results"  in  the presence of PCBs, such
 as  a greater  incidence of  cancer among female rats, sex-dependent
promotion  leading to more  altered  foci in  females', and
differences in  enzymatic activation and deactivation in a number
of  species that are  sex-linked, to mention a few.  Of relevance
is  the  fact that  nearly all of the double-crested cormorants with

severely deformed crossed-bills collected around the Great Lakes
in 1990 and 1991 were phenotypic females (n=100+).   These birds
came from geographic areas with demonstrated high PCBs
concentrations.  Chemical analyses and genetic sex determinations
are not completed on the birds.

     The pharmacokinetic model provided in the pre-meeting
literature did not graphically display the compartmentalization
of PCBs in reproductive tissues or organs (see p.88 in Matthews
and Dedrick, 1984 in the issue paper "Pharmacokinetics of PCBs".)
Attention is given in the other issue papers to post-conception
PCB concentrations in reproductive tissues such as placentae,
cord blood, and breast milk and the developmental neurotoxicity
associated with this exposure.  Along these lines,  Bush and
coworkers  (1986) reported the effect of three PCB congeners on
human sperm function.  Further research should investigate if
there are adverse effects from the presence of PCBs in folllcular
fluid, semen, sperm, etc. that can be separated from effects that
come later during gestational exposure.

     Much of what has been learned concerning transgenerational
exposure in wildlife and humans over the past decade was the
product of multidisciplinary research.  As the scope of future
studies broadens a better understanding of the
neuroendocrinological-immunological connections of PCBs will
undoubtedly surface.


Bush,  B.,  A.  Bennett,  J.  Snow.   1986.   Polychlorobiphenyl
congeners,  p,p'-DDE, and sperm function in humans.  Arch.
Environ. Contam.  Toxicol.  13:517-527.

Dieringer,  C.S.,  C.A.  Lamartiniere,  C.M.  Schiller, and G.W.
Lucier.  1979.  Altered ontogeny of  hepatic  steroid-metabolizing
enzymes by pure polychlorinated  biphenyl  congeners.  Biochem.
Pharmacol.  28:2511-2514.

Gray Jr.,  L.E., J.  Ostby,  R. Marshall,  and J. Andrews.   In press.
Reproductive  and  thyroid  effects of  low level polychlorinated
biphenyl  (Arochlor  1254)  exposure:   Repressed sex accessory
glands in  hypothyroid  rats with  normal  levels of serum and
testicular testosterone.

Harris, H.J.  1990.  Marshes, Forster's terns, and
microcontaminants in Green Bay.   Paper  presented at  "Preserving
Great Lakes Wetlands:   An  Environmental Agenda."  Conference
sponsored  by  the Great  Lakes Wetlands Policy Consortium, Buffalo,
New York.  May  15.

Kubiak, T.J., H.J.  Harris, L.M.  Smith,  T.R. Schwartz, D.L.
Stalling,  J.A.  Trick, L. Sileo,  D.E. Docherty, and T.C. Erdman.
1989.  Microcontaminants and reproductive impairment of the
Forster's  tern  on Green Bay, Lake Michigan-1983.  Arch. Environ.
Contam. Toxicol. 18:706-727.

Marquenie,  J.M. and P.J.H. Reijnders. 198"9.  PCBs, an increasing
concern for the marine  environment.  Report for International
Council for the Exploration of the Sea. 5p.

Reijnders,  P.J.H. 1986.  Reproductive failure in common seals
feeding on fish from polluted coastal waters.  Nature.  324:456-

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposure to PCBs
                    Research Triangle Park,  NC
                      September 14-15, 1992

                          Eric Dewailly
                   Community Health Department
                    Laval University Hospital
                      2050 blvd St-Cyrille
                         CANADA  G1V 2K8

General Toxicity of PCBs

     Ortho-substituted nonplanar congeners seem to be more
neurotoxic than coplanar PCBs in monkeys.  Are these effects due
to their specific accumulation in the brain or to their specific
toxicity as reported in vitro (or both)?

     What do we know about the relationship between the amount of'
Ah receptor in the cells of different fluids/organs and the site-
specific induction of Ah-mediated effects i.e. carcinogenicity,
immunotoxicity?  Is there a consensus regarding the absence of Ah
receptor in the CMS?

     More research is needed on the relationship between gonadal
and thyroid function during the development of CMS, with special
emphasis on the effects of low exposures  (low disturbances).

     For various species including man, it should be useful to
compare NOAEL or LOAEL for endocrine effects, immunotoxicity and
Carcinogenicity.                                                 ,

Developmental neurotoxicity of PCBs in humans

Question 1

     Any epidemiologic study designed to assess the developmental
neurotoxicity of PCBs needs appropriate control of the numerous
confounding factors.  For the Michigan cohort, the fact that the
average mother weight before pregnancy was less in the fish eater
group than in the control group is important.  Mother weight is
link both with exposure  (higher concentration in a smaller
adipose tissue compartment) and birth weight.  The question of
alcohol is also really important.  Although the relationship
between alcohol consumption and birth weight is well known, some
authors have reported associations between alcohol consumption
and PCB blood levels.  Rogan and Gladen in 1985 (Environ. Health
Perspect. .£0:215-221) reported that PCB levels in breast milk was
positively associated with alcohol consumption.  Stark in 1986
reported that the PCB blood levels of workers were significantly
higher among heavy drinkers  (> 2 drinks/day: 11.4 |lg/L)  than

 among non drinker (6.4 fig/L)  (Env.  Res.  41:  174-183).   Baker
 (1980)  and Kreiss (1981)  also discuss  these  associations.

      We also found this relation in 1986 (unpublished  data)  when
 comparing the sum of 5 PCB congeners in  the  blood of 565  french
 adults.   PCB blood levels in  the non drinker group was 4.2  [Xg/L
 compared with 6.0 jig/L for drinkers (P < 0.001).   This associa-
 tion  was highly significant in a regression  model (adjustment  for.
 age and sex).   In a recent Quebec survey of  536 breast milk
 samples,  PCB levels where also associated with alcohol consump-
 tion  (P < 0.001).

      Finally for the NC and MI cohorts,  a potential confounding
 toxin which correlate strongly with both PCB and  neurotoxic
 effects  (both motor and neurobehavioral)  is  methylmercury.   Most
 fish-eating population (fresh water fish) exposed to PCBs are
 also  exposed to methylmercury.

      WHO-1990  - Environmental Health Criteria 101:  Methymercury
 P 103 "The fetus is especially at risk.   Recent evidence shows
 that  a peak maternal hair mercury level  above 70  |J.g/g  a high risk
 (more than 30  %)  of neurological disorder in the  offspring.
 There is a need for epidemiological studies  on children exposed
 in utero to levels of methylmercury corresponding to a peak
 maternal hair  mercury level below 20 M-g/g, in order to screen  for
 those effects  only detectable by available psychological and
 behavioral tests."

 Question 2

      Because of the food  chain effect, it is possible  that  fish,
 eaters are exposed to highly  chlorinated PCB congeners compared
 with  the general population.

      In  Arctic Quebec (eskimo population) the 2,4,4'CB was  only
 detected in 2  of the 109  breast  milk samples from the  general
 population but was detected in all  96  milk  samples of the
 general  population.   PCB  153  (2,2',4,4',5,5') represents 40  % of
 all congeners  in eskimo population  and less  than  20 %  in the
 general  population.   The  same trend was  observed  in our fishermen
 study of  the St.  Lawrence river.

 Question 3

      No  comments

 Question 4

      Since  fish-eating populations  (sport fishermen,  natives,
 commercial  fishermen) appear  mostly exposed  to highly  chlorinated
 PCBs  and  nonplanar neurotoxic PCB congeners  have  short half-
 lives, then it  is  possible  that  neurodevelopmental effects are
not the preeminent risk for these population.  In order to study

 PCB-induced neurodevelopmental  effects, occupationally  exposed
 population could be more  appropriate  (if they  still exist).

      From a public  health point of view, the need to assess  PCB
 exposure  on a congener specific basis  is not obvious.   In  the
 last  few  years,  large progress  have been made  in our ability to
 conduct chemical analyses and toxicological assessment  of  PCBs.
 We  are submerge  by  new hypothesis on mechanisms of action, new
 TEFs  (dioxins, furans,  coplanar PCBs)  and by "new" metabolites
 (methylsulfone PCBs,  hydroxy  PCBs).  I personally 'doubt that the
 new information  on  the specific toxicity of each congener  and
 their metabolites will bring  us closer to answering the questions
 of  public health concern.

      In the real world, humans  are exposed to  numerous
 organochlorines  and heavy metals through their food  (mainly  fatty
 food  and  fish products).   There is a need for  evaluating the
 toxicity  of complex mixtures  representative of those found in the
-various regional ecosystems  (Great lakes, St.  Lawrence,  Arctic,
 etc.) .

 Question  5

      On a fat basis,  no major differences are  expected.  However,
 cord  blood measurement is more  accurate to assess prenatal
 exposure  (see CRC book -  Jensen 1991). Question 3 is more
 relevant  for this issue.

 Question  7

      The  power of a cohort design is low when  only 20%  or  34% of
 the information  on  exposure is  available.  This reenforce  the
 need  for  new cohort studies conducted  on highly exposed popula-
 tions, using the new  laboratory capacities now available.

 General comment

      The  questions  asked  by Dr. Prince are highly relevant.  I
 would like to mention how the question of pre  vs post natal
 exposure  is crucial for the public health authorities and  the
 population.  Whereas  there is no direct and effective possibility
 to  reduce exposure  of the fetus, infant exposure can be lowered
 by  discouraging  mothers from  breast-feeding. However more  work is
 needed to balance risks and benefits of breast-feeding.

                    Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposure to PCBs
                    Research Triangle Park,  NC
                      September 14-15, 1992

                         Kim N. Dietrich
                     University of Cincinnati
                       College of Medicine
                Department of Environmental  Health
            Division of Biostatistics and Epidemiology
                   Cincinnati, Ohio 45267-0056
Developmental Neurotoxicity of PCB's in Humans and Animal Models

     As a researcher whose experience has been in metals as
opposed to organochlorines, many, of my comments reflect, "some of
the problems encountered in pediatric neurotoxicology in general
and should not be construed as specific to the difficulties
encountered in estimating the risks of perinatal PCB' exposures

     It is quite evident from the matters raised in the "issues  .
papers" as well as from the series of published research reports
from the North Carolina and Michigan studies that the
epidemiologies of PCB and metals neurotoxicity share many similar
challenges.1  These demands include:   (1)  accurate establishment
of dose (biomarkers of environmental exposure), and the
differentiation of prenatal from postnatal exposures;   (2) ,
accurate assessment of an adequate number and range of
sociodemographic and biomedical background factors which may
potentially confound the observed dose-effect associations;  (3)
avoidance of bias in the ascertainment of study subjects; 4.)
Application of a rational data analytic strategy which balances  •,
the risks of both Type I and Type II error (i.e., over- or under-
control); (5) understanding the multiple ways in which health
effects may be expressed when dealing with a toxicant that.  ,
affects a number of organs and physiological systems. 6.)
Ultimately,  the greatest/challenge is to detect what are likely1
to be very subtle neurbpsychological deficits that are difficult .
to identify in an individual child, but which may have
significant public health implications for large populations-of
children.2  Working within the limits  of  their original  designs
and taking into account other practical considerations,  it is my
opinion that both the Michigan3 and North Carolina4 studies have
endeavored to respond to these challenges.

     There is one very important difference between the current
data-bases available on the health effects of low level:Pb.and
PCB exposures. Establishing the veracity of low-level PCB     '.'-,;.'
exposure effects is handicapped by the small number.of high
quality studies which have been conducted as well as by the fact
that PCB's represent a mixture of exposures which may vary

between investigations. Clearly, the Yusho and Yu-Cheng episodes
do not help us here owing to the high levels of exposure and the
fact that the observed health-effects cannot be specifically
attributed to the Kanechlors because of the probability of
simultaneous perinatal exposure to the highly toxic PCDF's.

     It is not uncommon to find inconsistency in the results of
pediatric environmental health studies. However, with a
sufficient number of investigations it is possible to take
advantage of state-of-the-art critical reviews and statistical
techniques such as meta-analyses to help to resolve uncertainties
in assessing population risks.5'6  At this point in time,  the
small number of low-level PCB exposure studies makes such an
approach difficult if not impossible. Nevertheless, a discussion
of the differences between the two major prospective studies
conducted to date may reveal reasons for their somewhat
discrepant findings.  More importantly, such deliberations should
guide future research efforts in this area.

     I will address some of the issues brought up by Dr.  Prince,
NIOSH and Dr. Miller, EPA, where applicable.

Incomplete control of potential confounding (e.g./ maternal
smoking, alcohol consumption, and weight)

     The question of control for substance abuse and other
maternal organismic factors has been a source of contention in
the area of Pb neurotoxicity as well.7   Consumption of  alcohol
prior to and during pregnancy is strongly associated with
exposure to PCB's in the Michigan study  (see Table I in Fein, et
al.8).   The methods  reportedly used to  quantify this are
described in a later publication.3  Developed  by Kuzma  and
Kissinger, this method allows a dose calculation (AA/Day, Week,
Month).  Although this methodology is widely accepted,  some
degree of under-reporting is to be expected. It is not clear from
the Michigan publications whether alcohol was treated as a
dichotomous (yes/no) variable early in the study (e.g., as
apparently in Fein,  et al.8) .  or whether the method of  Kuzma and
Kissinger was used in all published analyses.  I was not able to
discern how maternal alcohol consumption was quantified in the
North Carolina study in any of the papers I recently
reviewed.4'9"12   It  appears  as  though it  may have been as a
dichotomous (0,1)  variable in multivariable regression analyses.
Consumption of large amounts of alcohol during pregnancy clearly
presents a hazard to the fetus.  In my own studies of Pb, I have
not found the presence of this covariate to affect the magnitude .
or statistical significance of the toxic effect. Indeed,  in my'
sample of non-alcoholic women in Cincinnati, cigarette smoking
during pregnancy is the only chemical substance other than Pb
which is associated in a dose-dependent manner with various
neurobehavioral endpoints, including both school-age cognitive
and motor development (Lead references available upon request).

     Maternal  stature  and weight are commonly significant
predictors of  infant size at birth. I cannot determine if this
was one of the 73 potential confounding variables in the Fein, et
al.8 analyses,  but the covariable appears in a later paper
reporting on analyses  of postnatal growth in stature and weight.
It would be customary  to include this variable in any study
investigating  the environmental determinants of fetal growth and
maturation. If not, an important question would be whether it is
likely to be a bona fide confounder. Intuitively, one would
expect heavier consumption of  fish  (i.e., increased exposure to
dietary PCBs)  to be associated with lower maternal weight, but
not necessarily reduced stature. Interestingly, recent studies of
heavy consumers of marine mammals and pisciverous fish suggest
that the adverse reproductive  and developmental effects
associated with prenatal methylmercury exposure may be diminished
by the added nutritional benefits of a seafood diet
(Methylmercury references available upon request).

How different  are the  two populations in relation to the
distribution of highly chlorinated PCB's?

     Congener-specific analyses would have to be conducted.

What would be  the impact of having different congener
distributions  in each  of the populations on the magnitude of the
association between PCB levels and developmental or reproductive

     As far as I can tell, most of the studies in laboratory
animals have been performed with commercial mixtures of PCB's.
However, the studies that have been performed with individual
congeners seem to suggest substantial differences in their
toxicity and biological action.  Commercial PCB mixtures
clearly-differ from the residues in human tissue and human milk,
with those remaining in the body being believed to be the most
toxic and biologically active.  Undoubtedly, the extent to which
the more toxic congeners are present is going to affect the
dose-response  curve.   The degree of contamination with PCDF's is
also a consideration as concentrations of PCDF's vary among the
Aroclors produced in the United States. The lack of  .
congener-specific information in human studies makes risk
assessment very difficult.

Prenatal exposures were characterized differently in these

     The results of laboratory animal studies and human studies
suggest that in utero  exposure to PCB's presents the greatest   ,•
risk to neurobehavioral development, despite the fact that,
quantitatively, the infant can absorb substantially more PCB's
from mother's milk postnatally. However,  as others have pointed
out,12 exposure is much greater during embryo-fetal development

on the basis of dose per body mass.  The point here is that
prenatal and postnatal exposures are not equivalent, and the
mixing of' different physiological compartments spanning the
prenatal and postnatal periods of development could well lead to
different results.

     My reading of the Michigan and North Carolina studies
indicates that both groups collected and analyzed cord and
maternal serum.  In addition, the North Carolina study sampled
placentae. The fact that many of the cord serum PCB levels were
below detectable limits may have discouraged NIEHS investigators
from using this as a single estimate of dose.  However, a
reanalysis of the North Carolina data which abandons the omnibus
estimate of dose  (and its complicated physiological assumptions)
and focuses on cord and maternal serum values may be of value in
resolving this question.

The functional domains affected by PCB's appeared to be different
in the NC (psychomotor performance) and MI (short-term memory)

     In its early examinations of infant neurobehavioral status,
the Michigan study13 utilized an innovative methodology to assess
visual recognition memory.  The authors of this test claim that
it has substantially better predictive validity than standard
infant tests (e.g., the Bayley Scales of Infant Development as
used in the North Carolina study) ,14  Therefore, the early
results of the Michigan study were somewhat startling in that
they suggested that prenatal low-level exposures to PCB's may
have a lasting impact on cognitive function.  The North Carolina
investigation used the Bayley Scales of Infant Development which
is a well standardized assessment of infant neuromotor as well as
cognitive development.

     The differences in the early findings of these two studies
appear to be an effect on motor functions in one '(North Carolina)
and cognitive functions (e.g., early visual recognition memory)
in the other (Michigan). However,  an examination of Table III in
Gladen, et al.12 shows negative trends, though statistically
non-significant, for transplacental and breast milk PCB exposures
and the Mental Development Index.   Perhaps an item analysis of
the MDI in the North Carolina study might reveal significant
associations between PCB exposures and performance on Bayley
items reflecting memory or accommodation to novel stimuli.  Also,
I believe that the Bayley Scales were administered in the
Michigan study as part of the original protocol, but were not
reported.  A complete analysis and report of these data from
Michigan would be helpful for comparison with the North Carolina
results.  There is a brief reference to unpublished analyses of
the Michigan Bayley data in a review paper comparing the human
and laboratory animal data (Tilson, et al., A/eurotoxicol Teratol
12:239-248)  which suggests a perinatal PCB exposure effect on
fine-motor functions.

      There may be real discrepancies between the early results of
 these two studies but further analyses are definitely,needed to
 reach such a conclusion.

      There is a clearer discrepancy between the reported findings
 of the Michigan3  and North Carolina4 studies during the'late
 preschool/early school-age periods.   This may be due,  in part,'to
 differences in the characterization of dose as' discussed earlier.
 I admit that I had some difficulty discerning the N and  therefore
 the statistical power the North Carolina investigators had to   '
 detect a small effect of  low level PCB'exposure in their "latest
 report.4  They  state that they obtained scores  (report' cards,
 McCarthy Scales)  on 645 children at  3 years,  628 at 4 years,  and
 636 at 5 years of age,  etc.   However,  this is not consistent  with'
 Table I which suggests that  there was a substantial amount of
 missing developmental data in the earlier years (e.g., 3-5),  and ?
 Figures 1 and 2 do not provide Ns by ages.  I  would appreciate a

      The differences  in the  sociodemographic  backgrounds of these
 cohorts also needs further exploration, clarification and
 discussion.   Previous  epidemiologic  studies suggest that the
 ability to  detect small health effects is influenced by  the
 overall "risk-status"  of  the sample.15  The North Carolina cohort
 consisted of a highly educated cohort of women in which  88% chose
 to at least  attempt breastfeeding.9  As the authors acknowledge
 in several  of their papers,  this  is  hardly a  representative
 sample.   The Michigan  sample is  described as  predominantly urban
 and middle  class4.  I would like to see  some more specific details
 on these background factors  such as  HOME (Home Observation for
 Measurement  of the Environment)  scores if available, Social Class
 Index scores,  etc.  The possible  role of dietary differences  also
 need to be  considered  in  evaluating  the positive findings  in  the
 Michigan study with regard to perinatal exposures  to PCB's  and
 growth when_compared  to the  negative results  reported out  of
 North Carolina.   Diet would  seem to  be a particularly critical
 factor when  the reported  index of growth is weight  as in the
 Michigan study.16  For  example,  families which  supplement  their
 total  dietary protein  intake with sportfish are  likely to have
 leaner diets.

     Neither  study appears to investigate the  possibility of
 statistical  interactions  in  their data.  For example, are
 differences  in  sensitivity related to  diet, gender  or social
 class?   Such  interactions have been  reported  in  both the lead  and
methylmercury literature  in  both  humans  and animals.  It is an
 important question  in risk assessment  as  standards  are usually
based  on  the  most  sensitive  segments of  the population.

Laboratory animal models

     With regard to the animal studies and their relevance to the
human situation of perinatal PCB exposure, it is noteworthy that
for many environmental toxicants including MeHg, Pb and PCB's,
monkeys have been shown to respond to the lowest doses (i.e.,
those doses closer to general population exposures).  It is also
worth noting that higher order neurpbehavioral processes have
shown much greater consistency across both Pb and PCB studies
than other types of outcomes.17'18 Future work in this area should
continue and include long-term follow-up in primate colonies.
Given the complicated assumptions necessary to estimate a
reference dose from the available human epidemiological data,18
controlled studies of very low level perinatal PCB exposures in
primates take on added importance.  A concentration on sensitive
species (e.g., rhesus monkeys) has been recommended for further
work in this area.19                       .   -


1.   Dietrich K.N., and Bellinger b.C.   (in press)  Assessment of
     neurobehavioral  development in studies of the effects of
     fetal exposures  to toxic agents.  In H.L. Needleman and B.C.
     Bellinger  (Eds.)  Prenatal Exposure to Environmental Agents:
     Developmental  Consequences.  Baltimore:  Johns Hopkins
     University  Press.

2.   Weiss, B.   Behavioral  influences of environmental toxicants.
     In J.F. Kavanagh (Ed.)  Understanding Mental Retardation:
     Research Accomplishments and New Frontiers.   Baltimore:
     Paul H. Brookes  Publishing Company.

3.   Jacobson, J.L. et al.   (1990)  Effects of in utero exposure
     to polychlorinated biphenyls and related contaminants on
     cognitive functioning  in young children.  Journal of
     Pediatrics  116:38-45.

4.   Gladen, B.C. et  al.   (1991).  Effects of perinatal ,'
     polychlorinated  biphenyls and dichlorodiphenyl dichlorethene
     on later development.  Journal of Pediatrics 119:58-63.

5.   Hammond, P.B., and Dietrich, K.N.   (1990)  Lead exposure in
     early life:  Health Consequences.  Reviews of Environmental
     Contamination  and Toxicology 115: 91-124.

6.   Needleman H.L.,  and Gatsonis C.  (1990)  Low-level lead
     exposure and the IQ children:  A meta-analysis of modern
     studies.  Journal of the American Medical Association
     263:673-678.                             »
                   *>                          ...        ,

7.   Ernhart, C.B.  (1992)  A critical review of low level  ,
     prenatal lead  exposure in the human: Effects on the
     developing  child.  Reproductive Toxicology 6:21-40.

8.   Fein et al.  (1984)  Prenatal exposure to polychlorinated
     biphenyls:  Effects on birth size and gestational age.
     Journal of  Pediatrics 105:315-320.

9.   Rogan et al.   (1986)   Polychlorinated biphenyls (PCBs) and
     dichlorodiphenyl dichlorethene (DDE) in human milk:   Effects
     on maternal factors and previous lactation.  American
     Journal Of  Public Health 76:172-177.

10.  Rogan et al.   (1986)   Neonatal effects of transplacental
     exposure to PCBs and DDE.  Journal of Pediatrics 109:335-341.

11.  Rogan et al.   (1987)   Polychlorinated biphenyls (PCBs) and
     dichlorodiphenyl dichloroethene (DDE)  in human milk:
     Effects on growth, morbidity,  and duration of lactation.
     American Journal of Public Health 77:1294-1297.

12.  Gladen, B.C. et al.  '(1988)  Development after exposure to
     polychlorinated biphenyls and dichlorodiphenyl
     dichloroethene transplacentally and through human milk.
     Journal of Pediatrics 113:991-995.

13.  Jacobson, S.W. et al.  (1985)  The effect of iritrauterine
     PCB exposure on visual recognition memory.  Child
     Development 56:853-860.

14.-  Pagan et al.   (1986)  Selective screening device for the
     early detection of normal or delayed cognitive development
     in infants at risk for later mental retardation.  Pediatrics

15.  Rothman, K.J., and Poole, C.   (1988)  A strengthening
     programme for weak associations. International Journal of
     Epidemiology 17:955-959.

16.  Jacobson, J.L. et al.  (1990)  Effects of exposure to PCB's
     and related compounds on growth and activity in children.
     Neurotoxicology and  Teratology 12:319-326.

17.  Davis, J.M. et al.   The comparative developmental
     neurotoxicity of lead in humans and animals.
     Neurotoxicology and  Teratology 12:215-229.

18.  Tilson et al.  (1990)  Polychlorinated biphenyls and the
     developing nervous system: Cross-species comparisons.
     Neurotoxicology and  Teratology 12:239-248.

19.  Agency for Toxic Substances  and Disease Registry   (1992)
     Toxicological  Profile for  Selected PCB's  (Update).  U.S.
     Department of Health and Human Services, Public Health  *

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposure to PCBs
                    Research Triangle Park, NC
                       September 14-15,  1992

                            Beth Gladen
               Statistics and Biomathematics Branch
                          Mail  Drop B3-02
       National Institute of Environmental Health Sciences
                          P.O. Box 12233
                 Research Triangle Park, NC 27709
General  Toxicity of PCBs

      I don't  think we have a. very  good  idea of  the  dose  levels
 (or  congeners)  associated with developmental neurotoxicity  in
humans.                                       .

Pharmacokinetics of PCBs

      I have no  useful comments other  than the observation that  in
our  study, PCB  (and DDE)  concentrations  in breast milk declined,
on average, 20% over 6 months of lactation.

PCB  Structure Activity Relationships  and Developmental Toxicity

      I have a question.   It appears from what I read that several
links in plausible chains have been shown (this group of PCBs
tends to bind here,  things that bind  here should disrupt that,
etc.), but that there is  no test yet  of  the final predictions
 (that administration of  a specific PCB to test animals should
produce  a_specific effect).   Is that  indeed the case?  The  kind
of mechanistic  understanding based on structure discussed here  is
important, but  needs validation.

Developmental Neurotoxicity of PCBs in Humans

Deliberations and Outputs:

1)   In  the NC  study,  there were few  strong predictors of PCB
level and thus  few good candidates for confounders;  Maternal
smoking  and weight  were not  significantly related to PCBs in our
data; those who drank alcohol  had ,PCB levels about 13% higher
than those of non-drinkers.   Confounding cannot occur unless the
factor is related to PCBs.   We did include both smoking and
alcohol  as confounders anyway;•we did not include weight.

     The MI investigators.report checking smoking and alcohol as "
potential confounders.  They were apparently-not related to PCBs
in their data,  so  they did not  include them in further analyses.

     Based on this, uncontrolled confounding (at least by these
factors) does not appear to be a problem.

2)   There are no data on specific congeners in the NC study.  We
have only a summary measure of total PCBs.  The MI results are
reported based only on total PCBs; based on the papers, it does
not appear that they have congener data available (except for
sera from the children at 4 years).   It is quite tempting to
think that the congener distribution in NC would differ from that
in MI since the MI sample included heavy fish eaters, but there
are no data to support or refute this.         •

3)   The NC study used a method devised by McKinney using two
peaks to quantitate total PCBs.

     The MI study used the Webb-McCall method with Aroclors 1016
and 1260 as standards.                                         '

     We asked a chemist (Mary Wolff of Mt.-Sinai) to look at a
few of the NC chromatograms and guess what the results for total
PCBs would have been had the Webb-McCall 1260 method been used
(she actually used 1254, but said the differences would be
negligible).  Some guessing was involved since one of the peaks
needed for the MI method was hard to see on our chromatograms.
The upshot was that comparability between the two studies could
be achieved by dividing the NC numbers by two.  It should
definitely be kept in mind that this was an informal look, at a
small sample.  However, it provides a rough guideline.

     MI had a median neonatal milk PCB  (fat basis, 1260 standard)
of 0.7 ppm.  NC had a median of 1.8 ppm  (McKinney), which would
become 0.9 ppm.  Thus, roughly, the groups have comparable total
PCB levels.

     Again, the congener composition is not available.

4)   With no data, we can only speculate.

5)   In the NC study, we estimated the extent of prenatal
exposure to a child by averaging all the 'samples we had from its
mother.  Maximally, this would be one placenta sample, one cord
serum sample, two maternal serum samples, and about eight milk
samples.   (Note that all of these are surrogates, since they are
all post-partum measures and the interest is in pre-partum
exposure, probably to specific congeners at specific times.)
Results from all available samples on the woman were put on a
common scale and averaged.  Scaling was necessary to account for
different compartments  (e.g., milk vs. serum) and different times
(levels decline over the course of lactation-, about 20% over 6
months).  For PCBs, since the majority of cord and placenta
samples were below detection limits, we excluded these two
samples from the averaging process.  For postnatal -exposure, we
combined this average with information on the duration of

(Full  details  in  1986 paper  "effects of maternal
     The correlations between the various samples were high
 (about 0.7 among milk samples at various times and about 0.6
between milk and serum).  Although placenta and cord serum
usually had levels below the detection limits, the  measurements
we did obtain were from women who were high on their other
samples and this subgroup showed good correlation.  Thus we would
expect that our averaged prenatal measure would correlate well
with the levels in cord serum.

     Based on this, this issue is unlikely to explain differences
in the two studies.

6)   Here is_a summary  of the results to aid discussion.  How
these mesh with each other is hard for me to interpret.

     •    For Brazelton at birth:  NC found PCBs related to the
          tonicity cluster and the number of abnormal reflexes.
          MI found no relationship to PCBs,  although fish
          consumption was related to the 3 clusters named
          autonomic maturity,  number of abnormal reflexes,  and
          range of state.

     •    For Bayley at 5-6 months:   NC found deficits on the
          psychomotor scale at 6 months,  found nothing on mental
          scale.   MI saw no significant relationships;.they saw a
          tendency towards poorer performance at 5 months on a
          cluster_of Bayley items involving eye-hand coordination
          (unpublished except  mentioned in a review and  in
          personal communication).

     •    For visual recognition at  7 months:  only results  are
          MI,  saw an effect.

     •    For the Bayley at 12,  18,  24 months:  only results are
          NC,  saw psychomotor  effects.

     •    For the McCarthy at  3  years:  only  results are  NC,  saw

     •    For the McCarthy at  4  years:  NC  found nothing.  MI
          found prenatal PCBs  related to the memory and  verbal

     »     For the Beary  and Peabody  tests at 4  years:  only
          results  are MI,  saw  nothing.

     •     For the McCarthy  at  5 years:  only  results  are  NC, saw1 v

7)   Cord levels were not used in the NC study,  so there is no
impact of having 88% non-quantifiable.  Virtually all women had a
number assigned as described above.

     The MI study did use cord levels as a measure.  Of the 313
children, 115 had no usable samples, 134 had non-detectable or
non-quantifiable, and 64 had measurements.  Having 37% with no,
usable samples is worrisome.  I am a bit confused about the way
the usable samples were handled.  In the Fein 1984 paper, they
were apparently dichotomized into low (including the ND/NQs) and
high.  In other papers, nothing is stated.  Several papers give
means or medians; how the ND/NQs figured into these calculations
is not clear.

8)   I agree that the Japanese and Taiwanese poisonings are of
dubious relevance for the study of PCBs per se.   This leaves
basically the two studies mentioned.

     The NC study is focused on the general population.   (Neither
study, of course, is a random sample since both rely on
volunteers.)  It has ^about 900 children.  Extensive information
was gathered about growth and illnesses, as well  as
neurotoxicity.  The only effects attributable to PCBs were the
neurotoxic, primarily motor, ones.

     The MI study is focused on heavy fish eaters versus non-fish
eaters.  It has about 300 children.  The documented endpoints_are
growth and neurotoxicity.  PCBs effects were seen on birth weight
and later growth, as well as on visual recognition, memory, and

     Whether the different neurological effects seen are
different manifestations of the same underlying lesion is not
clear to me.' However, the difference in birth weight findings
suggests a real difference in results.  The effect seen in MI _
 (high exposure 160 grams smaller than low exposure) is of a size
which could easily be picked up by the larger NC study; yet we
saw not  even a hint of a trend.

     The difference in the study populations raises suspicions
about different congener patterns, but there is no data to
clarify  this.  If the congeners are indeed different, then
different results would be expected.  If they are not different,
it is possible that other contaminants in the Michigan fish are
playing  a role, with PCBs being a marker.

PGB Developmental Neurotoxicity in Experimental Animals

     Deliberations and outputs:  The questions laid out here  look
like a nice research agenda rather than an agenda  for our
workshop.  Based on my far  from comprehensive knowledge of  the
literature, not many of these appear answerable at this time.

1)   Nature of  the  lesion and timing of exposure:  unknown,
except that prenatal exposure appears more damaging than
postnatal exposure

2)   Mechanisms producing neurotoxic effects:  unknown

3)   Homology across species:  It's not clear that the human
studies agree with  each other, making it difficult to compare
with animals.

4)   TCDD toxic equivalents:  Based on the work by Rich Seegal
and his buddies, it's likely that a different scheme will be
needed for neurotoxicity.

5)   Congener-specific effects and impact of mixtures:  unknown

6)   Role of metabolites:  unknown

              Supplemental Pre-meeting Comments for
          Workshop  on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                    Research Triangle  Park, NC
                      September 14-15, 1992

                           Beth Gladen
               Statistics  and  Biomathematics Branch
                         Mail Drop B3-02
       National Institute of Environmental Health Sciences
                          P.O.  Box 12233
                Research Triangle Park, NC  27709

     I have briefly read the package of pre-meeting comments.
Some folks have factual questions about the NC study, and I think
matters can be facilitated if I give some answers here.

is fill-in-the-blank available?

     I've given below the answers to those specific questions I
caught.  The answers to most of these can be guessed if you keep
in mind the nature of the study.  Our mandate was to address the
question of whether the presence of PCBs in breast milk harmed
breast-fed children.  When we began planning the study in 1976,
this was unplowed territory (in humans), so we designed a broad-
brush first-look study which included, for example, extensive
information on growth and illnesses.  It happens that the only
positive results we have seen for PCBs are the transplacental
neurotoxic results.  However, the study was not narrowly focused
on these endpoints and thus was not optimized for them.


•    We do not have maternal IQ.
•    We do not have HOME scores.
•    The social class information we have is mother's and
     father's  education and occupation; since they're correlated
     anyway, we used  those for mother as covariates.
•    For the Bayley and McCarthy, we have only the  summary scales
      (2 for the Bayley and 6 for  the McCarthy); we  do not have
     the results on individual items.
•    We have maternal weight but  not height.
•    We have very limited diet information on the mother.
•    We have no specific measures on  thyroid function/-
     clinically, there are no  signs of hypothyroidism.
•    We have no information on iron status of the  children.

Availability of archived stuff

     The chromatograms went to our chemist, Jim McKinney; there
are also copies at  the contractor, Raltech (at least some, that's
where Mary Wolff got  them).  We have  a haphazard bunch of
archived samples  (some whole,  some extracts), but various freezer

 disasters and such over the years have left these in dubious

 Sample  size

      Several folks had questions related to sample size at
 various stages.   In any longitudinal study,  folks drop out and
 reappear,  so it  gets a bit confusing.   Here are some of the

      930  kids  with consent forms  signed;  of  these,  4  known  deaths
      before  5  years;  1 stillborn,  1  SIDS  at  6 weeks,  1  congenital
      heart_disease at 6 weeks,  1  car accident at  4  years
 •     915  kids  with PCB measurements  from  mom (919 with  DDE;  4
      moms with only cord blood  and/or placenta had  no usable PCB
      859  kids  with any information after  birth  (this  was reported
      as 856  or 858 in early papers,  but 3 children  who  had
      disappeared after birth  reappeared very late in  the study)


      867  with  Brazelton (of whom  one had  DDE but not  PCB
 •     788  with  Bayley  at  6 months
 •     720  with  Bayley  at  1 year
      676  with  Bayley  at  18 months
 •     670  with  Bayley  at  2 years
 •     645  with  McCarthy at 3 years
 •     628  with  McCarthy at 4 years
 •     636  with  McCarthy at 5 years
      506  sent  report  cards, 366 old  enough to be used

      (for Kim  Dietrich:   Table 1 shows 370 with at  least 1
McCarthy  but no  usable report card,  342 with at least 1 McCarthy
and at least 1 usable  report card, for a total of 712 with at
least one McCarthy).

PCB and DDE distribution

     For all children, the distribution among the various grouped
levels is:
transplacental PCB

transplacental DDE:
 breast milk PCB (amount through 1 year):
      category            number
      BOTTLE              103
      0-2                 168
      2-5                 219
      5-8                 193
      8-12                109
      12+                  43

 breast milk DDE (amount through 1 year):
      category            number
      BOTTLE              103
      0-3                 206
      3-7                 217
      7-11                156
      11-17               111
      17+                  42
      The total number for breast milk is smaller than that for
 transplacental since, for dropouts, we may not know how they were
 fed or how long they were breast fed.  By the way, Jack Moore
 says in his comments that we studied DDT and its derivatives; in
 fact, all we have is DDE.

 Alcohol and cigarette measurement

      Alcohol was indeed included as a yes/no covariate.  We asked
 how many beers, wines, liquors per week; if the answer was

 "occasional", this was coded as 0.   We asked separately for
 before and during the pregnancy.  Adding beer,  wine,  and liquor
 together gives the frequency of drinks/week (those with any
 missing excluded):

             before     during
      0         545       775
      1-7       309       115
      2-14       39         6
      15-21       5         1

      With so few women reporting more than one  drink  a day,  we
 Dust grouped as none or any based on the before-pregnancy report.

      On the subject of vices,  we had 738 non-smokers,  49  smoking
 0-0.5 pack per day before pregnancy,  46  smoking 0.5-1  pack,  and
 66 smoking more.   Of the smokers, 64 quit during pregnancy and 42
 cut back enough to move them to a lower  category.   With 82% non-
 smokers,  we again made this a yes/no covariate.

 Analyses we did

      We analyzed  PCB as unordered categories or as  a linear
 •     We did not look for interactions.   With effects limited to
      the top 5% or so of PCB exposure, trying to split  this
      further seemed to be stretching it.

 Relationships among exposure measurements

      I  enclose some graphs  which may facilitate  the discussion  of
 dose  measures.  I've  plotted all those below quantitation  limits
 as  zero,  even though  the  limits  could be high.   I've plotted
 everything  on a square root scale to  facilitate  viewing since
 they're all  skewed.   I threw out a PCB cord  blood of 410, which
 is  an order  of  magnitude  higher  than  the next highest, since it
 scrunched the graph.

      Attached are plots of  our average PCB measure versus PCB in
 cord  serum, placenta,  maternal serum  at birth,  and milk at birth.
 I also  attach plots of our  average DDE measure versus DDE in cord
blood and placenta, since this may give a notion of what the PCB
picture might have  looked like with lower detection limits.
Finally, I attach a plot of  PCB  in milk at birth versus PCB in
milk  at 6 weeks, to get at  stability over time and to some extent
at colostrum versus milk  (we don't know whether a sample is
colostrum or milk, although we know what day it was collected).


                                                                                                  -  CO
                                   -  CM
VO              ir\             ^

     O.OCQ     QLr— OJO(UC-I->CO

                   Q_ O CD
                         CO -M  CD  <-  C CO
                                                                                              -   CO




                                                                                             -  o
o. o 03
                                                                                      o o 4->

                                                                                         -  CO
                                                                                         -  (N
                                                                                          -   O
—I—I—I	1—I—r
ON        oo

  O O LJ
(N        rH

  t_ O O -M

                                                                                          -  CN

                    r^        vo

                 Q. r— Cd O (U C
co"cr 3 co
T i  •  •  •  •  i
  CM        .

 <-  O O  4->

                                                                                            -  (N


                                             f   -HH-H-
                                                 (U  0)
                                                                   (OCTDCOt-O)    t-OO-t->

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                  Associated with Exposure to PCBs
                    Research Triangle  Park, NC
                      September  14-15,  1992

                         L. Earl Gray,  Jr.
                 Developmental Reproductive Biology
                          Section (MD-72)
                 Health Effects  Research  Laboratory
               U.S. Environmental Protection Agency
                  Research  Triangle Park, NC 27711

 General comments

      It is clear from the reviews  that there  is insufficient data
 to perform a risk assessment  on the developmental  neurotoxicity
 of PCBs.   This problem exists  due  to a number  of critical data
 gaps in the PCB literature.  First,  we cannot  be sure that the
 developmental alterations reported in  humans exposed to .PCBs are
 not caused by some other toxicant.   Only experimental studies
 with laboratory animals  will be able to sort out.which congeners r
 are developmental toxicants because of the  complex exposures that
 humans receive.   Clearly,  the  SAR  relationship  between
 representative PCB congeners and developmental  toxicity needs to
 be systematically investigated. There is little data on the
 developmental toxicity of specific  PCB congeners because most of
 the animal developmental toxicology studies' have.used PCB
 mixtures.   For these  reaspns, a systematic  examination of the
 effects of different  types of PCB  congeners on  postnatal
 functional development is  needed.   Different congeners possess
 different  biological  activities.   Some have TCDD-like activity,
 others are estrogenic,,. a third  group of • PCBs are neurotoxic  to
 adults, and some PCBs alter thyroid function.   Any one,  or all of
 these mechanisms could theoretically alter mammalian development.
 An evaluation of the  functional developmental toxicity of the
 PCBs  should not. be limited to an examination of, neurotoxicity,
 but should include a  functional analysis of reproductive and
 immune function  as well.   The results  of such studies  should then
 be used guide additional human  studies.

 General Toxicity of PCBs

     _One of  the  low-level  effects of PCBs,  reported  from animals
 studxes, are their effects  on thyroid  hormone levels  (Gray et
 al.,  in press: Byrne  et al., 1987).  .One study has shown that PCB
 exposure lowers  serum thyroxine in humans as well.    Such effects
 occur in the young -adult male rat at doses 100 fold below those
 that  alter immune,  and reproductive  function (Gray et  al., in
press).  However,  the reproductive endocrine and thyroid function
 studies are  typically conducted with commercial mixtures  of  PCBs.
 In addition  to inhibiting the effects  of the reproductive   :
hormones in  the  female rat, some PCB mixtures mimic the  effects
of estrogen.

     With respect to the risk assessment of PCBs, it should be
noted that it was recently reported that the published_TEFs used
by the Agency for PCBs may be too conservative.  Certainly
additional research is needed to validate the use of TEFs for
risk assessment of dioxin-like PCB congeners and to verify their

     "Nondevelopmental neurotoxicity" is discussed in this paper.
Should we expect developmental neurotoxicity to correlate with
adult neurotoxicity, or is such a relationship unlikely because
some of the mechanisms of neurotoxicity may differ between adult
and developing animals?  For example, alklayting agents like MAM
and busulfan are potent developmental but not adult

     The carcinogenicity of the PCBs is discussed in this paper.
It has been•suggested that some of the PCB mixtures, like Aroclor
1254,- are toxic but are not carcinogenic.  Does this suggest that
different mechanisms of toxicity are responsible for cancer 'and
noncancer endpoints?

Pharmacokinetics of PCBs

     This paper states that the pharmacokinetics of the PCBs are
complex and can differ greatly from congener to congener.  Should
we focus our attention on the PCB congeners present in the
"environmental" mixtures in human and animals diets?  Which of
these persist in the environment and bioaccumulate in the food
chain?' Shouldn't this information be factored, along with other
information, into the prioritization of congeners for
developmental toxicology studies?

PCB  - SAR activity  relationships and developmental toxicity

     An important issue is raised in this paper  that is often
ignored in  discussions pf TCDD, the PCBs and other halogenated
hydrocarbons.  It seems quite likely that some of the
developmental toxicity of the PCBs is due to altered thyroid
hormone levels.  We found that serum T4 levels were reduced in
male F344 rats at low exposure levels when Aroclor 1254 was
administered by gavage for 15 weeks  (0.1 mg/kg/d)  (Gray et al.,
in press).  In our  study the effect on  serum T4  stood out  from
all  the other effects,  (immune, reproductive,  growth) being
altered at  all doses including the 0.1 mg/kg/d group.  A dose 100
fold higher was required to alter immune or kidney  function and a
dose 250: fold higher than this failed to alter testicular  and
serum T  levels.   Similar reductions  in  thyroid hormone levels
have been reported  in female SD rats  (Byrne, 1987).  The fact
that perinatal hypothyroidism  (PTU-induced)_ can  permanently alter
CNS  and  testicular  morphological and behavioral  development in
rats leads  one' to conclude that in utero and lactational exposure
to  PCBs  are likely  to adversely affect  CNS development.  Are  the
children in the NC  and MI studies hypothyroid?  In  addition to
the  effects of PCBs on thyroid hormone  levels, the  PCBs have  also

 been reported to demasculinize' sex differentiation of the male
 rat, possibly by creating a hypoandrogenic condition.  In
 contrast, the estrogenic PCB mixture Arqclor 1221 has been shown
 to alter CNS sex differentiation in the female rat-when '
 administered during the neona'tal stage of life, resulting in
 infertility and acycl'icity during adulthood.  Are homologous
 effects possible in humans?    -;:.                     .

      The two hypothesis proposed in the last two paragraphs of
 this paper should be examined in rodent studies.  First, it was
 proposed that PCBs act on the endocrine system through multiple
 mechanisms,  and secondly, it was stated that PCBs alter hormonal
 functions by acting directly as hormone agonists and antagonists
 with specific binding proteins or indirectly by altering specific
 receptor numbers.

 Developmental Neurotoxicity of PCBs in Humans

      This paper points out that disparities exist between the '
 behavioral alterations seen in the NC and MI studies.   Some of
 the factors  that might explain such differences are also
 presented.   Basically,  I feel that "perfect"  epidemiology studies
 withDefinitive results are rare.   However,  the fact that these
 studies both report that human developmental alterations are
 correlated with PCB exposure is  a  cause for concern.   It seems
 unlikely that reevaluation of these studies could eliminate
 either  the problems inherent in the data or the concerns over the
 health  effects of PCBs.  Additional human studies are  clearly
 warranted.to verify the hypotheses posed in the NC and MI

 PCB developmental neurotoxicity  in experimental animals

     Laboratory  animal  studies clearly  demonstrate that  PCBs
 (mixtures and a  few congeners) are developmental  reproductive and
 neurotoxicants.   As discussed in a number  of  the  Issue papers,
 the toxicity, pharmacokinetics, persistence and potential  for
 bioaccumulation  of  the  209  PCB congeners differ enormously.   A
 variety of biological activities of  the PCBs, including Ah-
 receptor binding, estrogenicity, alterations  of thyroid hormone
 levels  and alterations  of dopamine  in the  CNS,  have been
 described from studies  using  adult  rodents. Each  of these has  the
 potential to produce functional alterations development.  One
 strategy for testing these chemicals would be to  determine which
 PCB congener was  the most potent, with  respect  to  each of the
 above categories  and to comprehensively evaluate  the
 reproductive, neurobehavioral and  immune effects  of developmental
 exposure.  In the area  of developmental reproductive toxicology,
 PCBs produce some rather novel effects.  One of the more unusual
 alterations was reported by Sager et al.  (1983,  1987, 1991).   in
 their studies male  rats whose dams were treated with PCBs during
 lactation fertilize fewer ova when mated with untreated females
than do males raised by untreated dams, even though testicular,

and epididymal sperm counts and sperm motion parameters (measured
by CASA) were normal.  In female rats, neonatal treatment with
Aroclor 1221 accelerates reproductive aging by accelerating the
onset of persistent vaginal estrus and anovulation during middle-
age (Gellert 1978).  In addition, it is possible that the PCB
congeners that bind to the Ah receptor would alter sex
differentiation in the male rat after perinatal exposure in a
manner similar to TCDD (Mably et al., 1992 a,b,c).
DISCLAIMER.  The research described in this article has been
reviewed by the Health Effects Research Laboratory, U. S.
Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect
the views and policies of the Agency nor does mention of trade
names or commercial products constitute endorsement or
recommendation for use.

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposure to PCBs
                    Research Triangle Park, NC
                       September 14-15, 1992

                           G.  Jean Harry
        National Institute of Environmental  Health Sciences
                          P.O. Box 12233
                          Mail-Drop El-02
                 Research Triangle Park,  NC   27709

 General Toxicity of PCBs

      This issue paper presents the various  enclpoints  of  toxicity
 induced by PCB exposure and attempts to  draw the  reader's
 attention toward the interactive  nature  of  the  manifestation of
 toxicity.  The developmental  effects of  PCBs need to  be  evaluated
 within the framework of the total  system toxicity.  Effects  upon
 the immune or endocrine systems may play a  major  role in the
 developmental process and such interactions need  to be considered
 and not examined in isolation.  Much of  the non-neurotoxicity of
 PCBs is associated with the Ah receptor.  It would be of major
 importance to identify and characterize  the role  of the  Ah
 receptor mechanism in the manifestation  of  nervous system
 toxicity (whether  it  may have  a direct role or  only secondary to
 its anti-estrogenic properties).

 Pharmacokinetics of PCBs

      A significant amount of research has been  conducted
 concerning the pharmacokinetics of  PCBs  thus, establishing a
 basis_from which to evaluate differences in the developing
 organism.   The developing organism appears  to be differentially
 vulnerable to PCB  exposure  during the gestational period as
 compared to the lactational period.  Since the level of PCBs  in
 the brain is  higher following  lactational exposure this
 sensitivity may be due to individual PCBs present in  the brain or
 possibly in the organs  of the body which regulate hormonal
 balance.   The selective vulnerability of specific processes  in
 the  nervous system and the  developmental time period  of  exposure
will  play a major  role  in the manifestation of toxicity.   Given
 this  pattern  of susceptibility, it is of interest to  elucidate
 the maturation of  the pharmacokinetic process from pre-  to post-
natal periods  and  to identify any differential processing or
distribution  of individual  PCBs.  This information could be
critical  in the further 'examination of developmental effects of

PCB Structure Activity Relationships and Developmental Toxicity

     In reference  to the  second statement under the section
 "Deliberations  and Outputs", it is felt that an inter-related

approach is needed with research concerning the pharmacokinetics
and the structure activity relationships of PCBs.  PCB induced
alterations in hormonal status would play a significant role in
the process of brain.development.  By identifying chemical
structure with a specific target organ toxicity e.g., thyroid
gland followed by a subsequent effect upon the developing nervous
system one may be able to select a. chemical structure of PCB that
would perturb the nervous system in the absence pf a hormonal
influence.  If such a structure can be identified then the direct
effects on the nervous system can be examined.  Other structures
which alter hormonal homeostasis can be examined within the
framework of the existing literature concerning the role of
hormonal balance in the normal process of development.

Developmental Neurotoxicity of PCBs in Humans

     The rice oil disease in Japan and Taiwan was;associated with
intrauterine growth retardation however,.no defects were
recorded..  The syndrome was called the cola-colored baby due, to
the fact that.the babies were born with a skin discoloration
which faded within a few months. Accelerated development was
reported in four newborns with three displaying exophthalmus and
two had teeth present at birth (Funatsu et al., 1972).  This is
similar to alterations seen following gestational elevation of
epidermal growth factor  (EGF) and in fact in a limited study the.
birth weight was proportional to an alteration in the placental .
EGF receptor in PCB exposed 'pregnancies. It is an interesting
paradigm with the association to EGF that may deserve examination
in animal models.  It is especially interesting since one may
observe gestational acceleration of specific developmental  ,    .-
components followed by deficits in performance on developmental
tests.               ,           . .          _. .         ' •

PCB Developmental Neurotoxicity in Experimental Animals

     The data which is presented in the issue paper clearly,  .
presents exposure to PCBs as a major issue of concern for human ,
health.  Although, there is sufficient existing data to warrant
this concern there is a continuing need for further research at
the descriptive level.  This information is needed in order to
establish.a firm base.from which to determine the direction of
and design for further studies to elucidate underlying processes
involved in- the,neurotoxic response.  One of the major areas that
needs'.further characterization is the differential response
following gestation versus lactational exposure to PCBs;  Working
in a coordinated manner with the pharmacokinetic and structure-
activity, research activities, the susceptible periods of
vulnerability need  to be identified.  For example, is there a
specific stage during the gestational period that the brain is
most susceptible and if so, what is the ongoing developmental
process?  As mentioned in the previous sections, there'.is a
definite need to both evaluate the role of the alterations in
hormonal balance during development and to attempt to bypass -:such
alterations to allow one to examine possible direct effects of

individual PCBs on the developing nervous system.  Dr. Miller has
presented an informative review on the issue and a good
background and rationale is given for each question presented.
This should allow for productive deliberations at the workshop.

                    Pre-meeting Comments for
           Workshop on Developmental Neurotoxic  Effects
                Associated with Exposure to PCBs
                    Research Triangle Park,  NC
                      September 14-15, 1992

             Joseph L. Jacobson & Sandra W. Jacobson
                      Wayne State  University
                          71 West  Warren
                        Detroit, MI 48202
     Several hypotheses have been advanced in the pre-meeting
issues papers and comments to explain differences in the effects
associated with prenatal PCB exposure in the MI and NC studies.
One important possibility—that the two populations may have been
exposed to different mixes of congeners—cannot be tested,
because congener-specific analyses were not performed.  Given
that in both studies effects were observed only at the highest
exposure levels, it is possible that the inconsistencies
represent a form of Type II error due to the relatively low
incidence of higher exposed subjects.  In these comments, we
address three other issues relevant to this question, on which
some data are available.

Control for Potential Confounding

     In her issue paper, Dr. Prince raises questions regarding
the procedures used for selecting control variables in the MI
study.  The NC and MI studies took somewhat different approaches
to controlling for confounding.  Rogan and Gladen used a pre-
selected set of control variables; we considered a longer list of
potential confounders but controlled only for those that were
empirically related to exposure level.  Our approach was based on
the premise that a control variable cannot be the true cause of
an observed deficit unless it is related to both exposure and
outcome (Schlesselman, 1988; Jacobson et al., 1990).  Our initial
list of potential confounders consisted of variables known or
suspected to relate to the outcomes in question.  We controlled
statistically for any of these potential confounders that were
also related (even weakly at p_ < .10) to exposure.

     We do not believe that "incomplete control of potential
confounding [can] explain the positive findings from the MI
study."  To check out this possibility for the 4-year McCarthy
Memory effect,  the principal effect seen in MI but not in NC, we
reran the PCB cord serum, to McCarthy Memory regression analyses
adding six variables measuring the domains controlled for by
Gladen and Rogan (1991) that were not included in our own
analyses (Jacobson et al., 1990).  .As shown in Table 1,  the
effect of cord serum PCB level on McCarthy Memory performance was
essentially unchanged.

     One  of  the more  striking  features of our data set is the
dearth of potential confounders with which cord serum PGB level
was associated.   Table  2  lists the 24 potential confounding
variables examined in the 4-year  followup study; only three of.
these variables were  even weakly  related to cord PCB level  (Table
3).  Nevertheless, there  has been some confusion about possible
confounding  of PCB exposure with  pregnancy drinking in this
sample due to our decision to  report the birth size data in terms
of dichotomous variables  (Fein et al., 1984).  The initial
analyses  of  all of the  outcomes in this study--including birth
size—were based  on multiple regression, which treats exposure as
a continuous variable.  Because the reviewers at that time did
not seem  to  understand  the multiple regression approach, we     :
reanalyzed the data in  terms of an exposed/nonexposed dichotomy.
In retrospect, that decision was  a mistake.  Pregnancy drinking
was treated  as dichotomous in  those analyses so -that even women
who drank only one or two drinks  during pregnancy were classified
as "drinkers," which  artificially inflated the apparent
confounding  of drinking and PCB exposure in Fein et al. Table I.
The results  of the original multiple regression birth size
analyses  were subsequently published in a book chapter  (Jacobson
et al., 1985; Table 4).   (As can  be seen, maternal.weight was
included  as  a covariate in these  analyses.)

     Maternal drinking  prior to and during pregnancy/ measured in;
terms of  ounces of absolute alcohol per day (AA/day; Kuzma &
Kissinger, 1981),  was only weakly related to prenatal PCB
exposure  (Table 5).   Moreover, levels of pregnancy drinking were
exceeding low in  this sample.  Of the 313 women whose infants
were studied at birth,  none averaged more than 1.0 oz AA/day, the
usual criterion for risk  drinking, and only eight ,drank more,than
0.5 oz, the  lowest level  at which even the subtlest alcohol-
related behavioral effects have been observed (Jacobson et al.,
in^press).   Given the very low,level of pregnancy drinking in
this sample  and the fact  that  it  was controlled for wherever, even
weakly related to exposure, we do not believe that the observed
effects of PCB exposure can be attributed to prenatal alcohol.

Differences  in Assessment Procedures at 4 Years

     One  difference between the MI and NC studies relates to our
handling  of  the problem that some preschool children are
unwilling or unable to  cooperate  with the relatively structured
format of a  psychological assessment.  If an uncooperative child
scores low,  we do not know if  the low score is due to lack of
competence or his/her refusal  to  cooperate with the testing
procedure.   To deal with  this  problem, before examining the
effects of PCB exposure,  we set a criterion to identify the
uncooperative children.    Of 236 children tested on the McCarthy
Scales at 4 years, seventeen refused to respond to all or all :but
one of the items'on one or more of 17 designated McCarthy      :  •
subtests.  These.children were excluded a priori from the
analyses  of  toxic'effects  on the  grounds that their scores could;
not be assumed to  reflect  their actual competence.

     In preparation for this meeting, we went back and reanalyzed
the data to see if the McCarthy Memory effect would have been
detected if these children had been included in the regression
analyses.  As can be seen in Table 6, if these children had been
included, the PCB effect would have fallen short of conventional
levels of statistical significance, suggesting that their
exclusion had removed a source of extraneous variance.  Since
some evidence of a trend would, nevertheless, have been evident,
only a small part of the discrepancy between the MI and NC
studies would appear to be attributable to this difference in

Difference in Analytic Methodology

     It is difficult to compare exposure levels between the MI
and NC studies because of differences in analytic methodology.
The MI study used the Webb-McCall method, in which all of the
major peaks appearing in the gas chromatogram are quantified
(Fig. 1).  In NC only two peaks--125 and 146—were quantified,
and total PCB level was estimated based on those two peaks
(McKinney et al., 1986).  Because the congeners in those two
peaks constituted 26.4% of the PCBs in industrial mixtures, total
PCB level was estimated by multiplying the levels in peaks 125
and 146 by a factor of about four.  However, we found in our
human serum samples that these peaks actually constituted 40.9%
of the total PCBs, so that estimates based on them should
probably have been multiplied by a factor of only about 2.5.

     Table 7 compares the median maternal serum PCB values in the
NC and MI studies.  At first glance, the exposure levels appear
to be higher in NC.  However, if total PCBs in MI had been
estimated based only on peaks 125 and 146 as in NC, the median
maternal PCB serum values would have appeared to be similar.  In
addition, the NC values were adjusted for recovery loss based on
an estimated recovery value of 75-90% (McKinney et al.., 1986).
If the MI data are adjusted for a recovery value in the midpoint
of that range, 82.5%, the median maternal PCB value would
actually appear to be 10.1 ng/mL.  Given the differences in
analytic methodology and the number of assumptions necessary to
make this comparison, the values in Table 7 are fairly
speculative, but this analysis suggests the possibility that the
exposure levels in MI may have been marginally higher than in NC.

Demographic Characteristics of the Ml Sample

     A breakdown of socioeconomic status (SES) into five
categories for the MI sample is shown in Table 8.  The families
range broadly from upper middle to working class, but few lower
class families participated.  On the Preschool Home'Observation
for Measurement of the Environment  (HOME),  the sample.mean at 48
months was 47.7 on a 55-point scale  (SD = 4.3), slightly higher
than the mean of 42.8 (SD = 5.9) reported for a 39-month sample
characterized as "middle class"  (Gottfried & Gottfried, 1984) and

 similar to the mean of 47.8  (SD  =4.4)  in Siegel's  (1984) 60-
 month-old full-term sample.

 Study Design and Attrition Patterns

      In his pre-meeting comments, John  Moore characterizes the MI
 study as a comparison between  "a cohort of women who stated that
 they  consumed fish from Lake Michigan"  and "women who stated they
 were  rion-consumers of Lake Michigan  fish."  Although our sample
 was selected from those two groups,  the study was not intended to
 provide a comprehensive evaluation of Lake Michigan fish
 consumption but rather 'to  assess effects associated with PCB
 exposure assessed in terms of  continuous measures of PCB body
 burden.   The principal purpose of the fish consumption survey and
 recruitment strategy was to increase the incidence of higher
 level PCB exposure in the  sample.  Although Lake Michigan fish
 consumption was a major source of PCB exposure for these children
 (Jacobson et al.,  1983;  Schwartz et  al., 1983), PCBs are also
 present in a variety of other  food sources and some exposure was
 presumably from these sources  as well.

      Attrition between birth and 4 years has been described for
 the MI cohort in the literature  (Jacobson & Jacobson, 1990).  As
 can be_seen in Table 9,  the drop-outs were more likely to be
 lower in SES,  parental education, and maternal age and unmarried.
 More  highly educated parents may place  greater value on research
 on the effects of environmental  pollution and may be more
 motivated to obtain information  about their children's cognitive
 development.   Given the greater  vulnerability of low SES children
 to deficits in cognitive and attentional function from various
 sources,  including less optimal  health  care,  diet, and
 intellectual stimulation in the  home, teratogenic effects of
 prenatal PCB exposure may  be more readily detected in a more
 middle class sample.

      Table 10  provides data relating to the threat to internal
 validity posed by attrition bias.  If a teratogenically exposed
 child is ill or functioning poorly,  the parent may be
 particularly motivated to  participate in this kind of study in
 the hope of learning more  about  the  child's disability.  However,
 if a  child is  disabled but not teratologically exposed, the
 parent  is  likely to focus  his/her efforts on obtaining help
 elsewhere.   If exposed disabled  children participate at higher
 rates  than nonexposed disabled children, effects correlated with
 exposure may be a  spurious consequence of the participation and
 refusal  patterns.   As  can  be seen in Table 10,  higher exposed
 participants were  not  significantly  more likely to have had
 congenital  problems  or to  have been  lower in birthweight or ill
 or hospitalized since  birth,  nor were lower exposed non-
 participants more  likely to exhibit  such problems.  More highly
 exposed  participants  did exhibit somewhat poorer recognition
memory than nonparticipants,  but there was no corresponding
 tendency for lower exposed nonparticipants to have performed more

poorly on this test, and the participation by exposure
interaction did not approach statistical significance.

Fein, G.G., Jacobson, J.L., Jacobson, S.W., Schwartz, P.M., &
   Dowler,  J.K.   (1984).   Prenatal exposure to polychlorinated
   biphenyls:   Effects on,birth size and gestational age.   The
   Journal  of  Pediatrics,  105,  315-320.           . •,.

Gladen, 3B-.C.,  & Rogan, W.J.  (1991).  Effects of perinatal
   polychlorinated biphenyls and dichlorodiphenyl dichloroethene
   on later development.  .The Journal of Pediatrics,  119,  58-63.

Gladen, B.C.,  Rogan, W.J., Hardy, P., Thullen, J., Tingelstad,
   J.,  & Tully,  M.   (1988).   Development after exposure  to
   polychlorinated biphenyls and dichlorodiphenyl dichloroethene
   transplacentally and through human milk.  Journal of
   Pediatrics,  113,  991-995.

Gottfried.,  A.W. & Gottfried, A.E.   (1984) .  Home environment and
   cognitive development in young children of middle-
   socioceconomic-status families.  In A.W. Gottfried (Ed.),  Home
   environment and early cognitive development:   Longitudinal
   research.   Orlando,  FL:   Academic.
Hollingshead, A.B,   (1975)
   Unpublished manuscript.
  Four  factor  index of  social  status.
Yale University.
Jacobson, J.L., & Jacobson, S.W.  (1990).  Methodological issues
   in human behavioral teratology.  In C.  Rovee-Collier & L.P.
   Lipsitt  (Eds.),  Advances in infancy research (Vol.  6).
   Norwood,  NJ:   Ablex.

Jacobson, J.L., Jacobson, S.W., & Fein, G.G.  (1986).
   Intrauterine exposure to environmental  toxins:   The
   significance of  subtle behavioral effects.   In  A. Wandersman &
   R.  Hess  (Eds.),  Beyond the  individual;   Environmental
   approaches  and prevention.   New York:  Haworth,  pp. 125-137.

Jacobson, J.L., Jacobson, S.W., & Humphrey, H.E.B.  (1990).  .
   Effects  of  in utero exposure to polychlorinated biphenyls on
   cognitive functioning in young children.   The Journal of
   Pediatrics,  116,  38-45.

Jacobson, J.L., Jacobson, S.W., Sokol, R.J., Martier,  S.S., Ager,
   J.W.,  &  Kaplan-Estrin,  M.G.   (In press).   Teratogenic effects
   of alcohol  on infant  development.  Alcoholism:   Clinical and
   Experimental Research.

Jacobson, S.W., Jacobson, J.L.,  Schwartz, "P.M., & Fein, G.G.
   (1983).  Intrauterine exposure of human newborns to PCBs:
   Measures  of exposure.  In P.M. D''ltri & M. Kamrin (Eds.),
   PCBs:   Human and environmental hazards.  Boston:   Butterworth.
Kuzma, J.W., & Kissinger, D.G.
   cigarette use in pregnancy.
   Teratology,  J3.,  211-221.
 (1981).   Patterns  of  alcohol  and
Neurobehavioral Toxicology and
McKinney, J.D., Moore, L., Prokopetz, A.,& Walters, D,B.
   (1984).   Validated extraction and cleanup procedures for
   polychlorinated biphenyls  and DDE in human body fluids and
   infant formula.  Journal of the Association of Official
   Analytic  Chemists.  67.  122-129.

Schlesselman, J.  (1982). Case-control studies: Design, conduct
   analysis.   New York:  Oxford University Press'.
Schwartz, P.M., Jacobson, S.W., Fein, G.G., Jacobson, J.L., &
   Price,  H.A.  (1983).   Lake Michigan fish consumption as a
   source of polychlorinated biphenyls exposure.'  American  '
   Journal of Public Health.  73.  293-296.

Siegel, L.S.  (1984). Home environmental influences on cognitive
   development in preterm and full-term children during the first
   5 years.   In A.W. Gottfried (Ed.),  Home environment and early
   cognitive development:   Longitudinal research.   Orlando,  FL:

                         Table 1
Controlling for maternal age, gravidity,
   and examiner  (Jacobson etal., 1990)    -.16
Controlling for additional variables
   used by Gladen & Rogan (1991)
   (SES, pregnancy smoking and drinking,
   number of children, child's sex,
   duration breast-feeding)               -.16


                                          Table  2
                 Potential Confounding Variables Tested in PCS 4-Year Follow-Up Stuoy

                	Variable                            Source or Definition
           Socioenviro'nmental Influences
             Socioeconomic status
             Home Observation for Measurement
               of the Environment (HOME)
             Maternal vocabulary—Peabody Pic-
               ture Vocabulary Test-Revised
             Mother's education (years)
             Mother's age
             Marital status
             N of children
             Parity of child
             Familial stress scale
             Maternal employment
             Nursery school
           Perinatal  Influences
             Delivery complications
             Gravidity of child
           Other Toxic Exposures
             Maternal alcohol before pregnancy

             Maternal alcohol during pregnancy

             Maternal smoking before pregnancy ]
             Maternal smoking during pregnancy J
             Polybrominated biphenyls (PBBs)
           Situational Influences
             Age of child at testing
             Medications within past 24 hours

           Other Influences
             Child sex
 Hollingshead (1975) four-factor scale
 Caidwell & Bradley (1979)

 Dunn & Dunn (1981)

Highest level of stress reported for any
    of four domains:  financial, health,
    familial, or other
Hours  per week

Coded  yes if any of the following were
    present: emergency Caesarian sec-
    tion,  labor longer than 20 hours,
    placenta previa or afaruptio, tox-
    emia,  cyanosis, fetal distress, me-
    conium staining, infected cord, or
    knot in cord
Based on beer, wine, and hard liquor
    consumption reported by the mother
    and summarized in terms of ab-
    solute alcohol per day (Kuzma &
    Kissinger, 1981)
Cigarettes per day reported by the mother

In cord serum, maternal serum, breast
    milk, and child's serum  at age 4
In child's serum at age 4 years
(n child's blood at age 4 years
Antihistamines, cough syrup, or seizure
    medication coded present or absent
Jacobson,  J.L.,  5  Jacobson,  S.ff.  (1990).   Methodological  issues
    in  human  behavioral  teratology.   In C.  Rovee-Collier & L.P.
    Lipsitt  (Eds.), Advances  in infancy research  (Vol.  6).
    Norwood,  NJ:   Ablex,

                              Table 3
              THE 4-YEAR FOLLOW-UP SAMPLE (M = 146)

      Maternal age                                      r= .16*
      Gravidity                                          r= .17*
      Examiner (categorical variable tested by ANOVA)        F = 4.41 **

      *H<.05. **c<.01.
                              Table 4

Head circumference
Gestationa! age
   aAfter controlling for alcohol and caffeine consumption during pregnancy,
maternal age, maternal weight gain during pregnancy, maternal prepregnancy
weight, and sex of infant.
FROM: Jacobson, J.L., Jacobson, S.W., & Fein, G.G. (1985).  Intrauterine
   exposure to  environmental toxins: The significance of subtle behavioral
   effects. In A. Wandersman & R. Hess (Eds.), Bevond the individual:
   Environmental approaches and prevention. New York:  Haworth.

                        .   Table 5
                                 Cord serum
                                 PCB level
          fish consumption
Alcohol consumption
   Prior to pregnancy
   During pregnancy
Note.  Maternal alcohol consumption was unrelated to cord serum PCB level
      in the sample of children assessed at 4 years.
   +£<.10. *c<.05. **a<.01.
 Excluding uncooperative children   133
 Including uncooperative children    146
  aAfter controlling for examiner, mother's age, and gravidity.
  *p = .09. *p<.02.


                   Table 7


Roganetal.                            7.0 -9.1

Jacobson et al.

   Direct measurement                       5.4

   Estimated based on Peaks
      125 and 146 alone                     8.4

   Adjusted for recovery loss (based
      on estimated recovery value of 82.5%)    10.1

Values are ng/mL (parts per billion).

Major business, professional .
Medium business, semi-professional, technical
Clerical, sales
Skilled, semi-skilled
Unskilled, menial


                                   Table 9

  Demographic Characteristics and Exposure Levels of Participants and Nonparticipanis
 _	in the 4-Year Follow-Up
                  or x:
  Demographic Cliaractcrislics
   Sociocconomic status (SES)*
   . Maternal education (years)
   Paternal education (years)
   Mother's age
   Parity of child
   Marital status (% married)b
   Sex of child (% male)b
   Mother employed (%)b
  Exposure Levels
   Maternal contaminated fish
     consumption (kg/year)0
   Cord serum PCB level*
   Maternal serum PCB leveld
   Maternal milk PCB level (fat



      Note. Values .are group means, except where otherwise indicated. T-tests are used
 to compare means; chi-squares are used to evaluate contingency table data.
  t    • Hollingshead (1975) four-factor scale. b Cell frequencies taken from 2X2 con-
 tmgency tables. c Species are weighted to reflect degree of PCB contamination. * Ng/mL
 or parts per billion.
      tP < .10. *p  < .05. **p < .01. **V < .001.
                                    Table 10
 Participant and Nonparticipant Differences in Child Outcome Broken Down by Level
	'                       of Exposure
                         High Exposure9
               Low Exposure
                   Participants Nonparticipanis  Participants  Nonparticipants  Fb
Congenital medical
problems (%)
III since birth (%)
Hospitalized since
birth (%)
Birth weight (gms)
Visual recognition









4.3 n.a.

6.9 n.a.
31.0 n.a.

3833 <1
• .63 
                                                   0 Farm exposure
                                                   0 Fish exposure
               70S   84    125   146    174   203  232/244  280   332    372

                              WEBB-MCCALL PEAK NUMBER
       FIGURE 1—Webb-McCaU chromatographic elution peak values by exposure
Jacobson, J.  L.,  Humphrey, H. E. B., Jacobson,  S.  W., Schantz, S. L. 9
     Mullin,  M. D., Welch, R. Determinants  of polychlorinated biphenyls
     (PCBs),  polybrominated biphenyls  (PBBs), and dichlorodiphenyl  trichlorethane
     (DDT)  levels in the sera of young children.   American Journal  of  Public
     Health,  1989, 79,1401-1404.

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposure to PCBs
                    Research Triangle Park, NC
                       September 14-15,  1992

                          James McKinney
                  Pharmacokinetics  Branch (MD-74)
                Health Effects Research Laboratory
               U.S. Environmental Protection Agency
                 Research Triangle  Park,  NC   27711
General Toxicity of  PCBs
T, i .p        section  on  Effects  on  Endocrine Function,  it might  be
neiptul to discuss what is known with regard to effects on  the
hypothalamic-pituitary-thyroid  axis.

Developmental Neurotoxicity of  PCBs in Humans

     In _ view of the  potential of certain PCBs to act as thyroxine
antagonists  and produce mild hypothyroidism, it might be useful
to examine this possibility in  light of what is known about the
development  of children with congental hypothyroidism  (see
Birrell at al., Dev. Med. Child Neur. 25: 512-519, 1983)

                    Pre-meeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research Triangle Park, NC
                      September 14-15,  1992

                          Diane Miller
                Neurotoxieology Division (MD-74B)
                Health Effects Research  Laboratory
              U.S. Environmental Protection Agency
                Research Triangle Park,  NC  27711

General Comments

     It should be kept in mind that the term "neurotoxicity" is
an umbrella term, and there is at the present no universally
accepted definition of neurotoxicity.  The following excerpted
from Tilson (1990) should be kept in mind during the
deliberations on the developmental neurotoxicity of the PCBs.

Defining Neurotoxicity

     "At the present time, there is no commonly accepted
definition of neurotoxicity.  The Interagency Committee on;
Neurotoxicity (ICON) recently met to discuss several' of the*
definitions of neurotoxicity.  Although there is agreement that
neurotoxicity is  "any adverse effect on the structure or function
of the central or peripheral nervous system produced by a,
biological, chemical or physical agent," there is less agreement
as to what constitutes an adverse effect.  Adversity may have one
or more characteristics, including 1) side effects or overdose
 (unwanted effects), 2) decreased ability to function fully or
compensation in order to function normally, and/or 3) an
alteration that diminishes the ability  to survive, reproduce, or
adapt to the environment.  In one regulatory context, adversity
is strongly influenced by the usage of  the agent and the
conditions under which exposure occurs.  How each regulatory
agency would deal with an adverse effect depends largely on the
legislative mandate and mission of the  agency.

     Most would agree that neurotoxic effects can be permanent or
reversible and  could  be due to neuropharmacological or
neurodegenerative properties of a chemical.  One major
disagreement concerns labeling an agent as a neurotoxicant  if it
produces neurotoxic effects indirectly.  For instance, some would
not  label a chemical  as a neurotoxicant if it produced irritation
due  to direct contact with sensory receptors or effects secondary
to toxicity in  organs such as the  liver or kidney.  Others would
argue that it does not matter how  the nervous system  is affected.
What matters is that  the dose-response  curve associated with
nervous system  effects lies to the left or right of other
 indicators of toxicity.  Where the nervous system dose-response
 curve fell relative of other target-organ  toxicities would
 determine the  "label," or more likely,  the endpoint that might be

 used to set  environmental  limits.  Certainly in dealing with risk
 assessment issues,  some neurotoxic effects  (i.e., neuropathology,
 deficit in learning and memory) would receive a greater weighing
 than others  (i.e.,  irritation) in any decision-making process.
 In the 1990s as guidelines for conducting risk assessments on
 neurotoxic endpoints become a reality, agreement on terms such as
 neurotoxicity and adverse  effect will be important

 Excerpted from:

 Tilson, H.A. (1990) Perspective - Neurotoxicology in the 1990s.
 Neurotoxicol. Teratol. 12(4):293-300.

 General Toxicity of PCBs

    t  As this issue paper notes the PCBs leave a wide spectrum of
 toxicity that includes alterations in. the hepatic,  immune and
 other systems.   Consideration of the degree of systemic changes
 induced by PCB mixtures or specific congeners will  be an
 important aspect of the deliberations  concerning their
 developmental neurotoxicity.   Any or all of these systemic
 effects will need to be considered as  possible determinants  of
 altered CNS development or function.   in terms of the possible
 interactions with CNS development the  hormonal effects of PCBs
 are of  particular interest.  It should also be noted that the
 PCBs  can have many diverse endocrine effects and can interfere
 with thyroid as well as gonadal function.   PCBs have been noted
 to alter the development of the thyroid of  offspring exposed
 gestationally and through lactation of 50-500  ppm Aroclor 1254
 Consequences  included both altered structure and function of the
 thyroid (e.g. large reductions  in serum 13  and 14).   Differences
 were  apparent at birth.   The  thyroid axis effects of PCB  mixtures
 and specific  congeners  may  be  important  determinants of the  CNS
 effects of  PCBs  because adequate  function of the  thyroid  axis  is
 needed  for  appropriate  brain development.

 Collins, W.T. and  Capen, C.C.  (1980).  Fine  structural lesions
 and hormonal  alterations in thyroid glands of perinatal rats
 exposed in Utero and by the milk  to polychlorinated  biphenvls
 Amer. J. Pathol.   99:125-136.      :                           "

 Pharmacokineti.es of  PCBs

     In determining  the developmental toxicity of specific
 commercial mixtures  of  PCBs, environmental mixtures  of PCBs  and
 specific congeners or in the deliberation regarding which of the
 congeners or mixtures to evaluate, the relative distribution of
various congeners to the brain will need to be considered   For
 example, data ('see Lilienthal & Winneke, 1991 and Lilienthal et
 ai., 1990) present data indicating pre- or postnatal exposure to
a mixture of PCBs result in brain deposition data suggesting a
greater exposure of the fetus to low-chlorinated congeners during
gestation while postnatal exposure via milk is to the more highly
chlorinated PCBs.  Should information such as this be iked to

determine which of the PCB congeners would be selected for
evaluation as a developmental neurotoxicant?  It should also be
kept in mind that CMS alterations have been introduced during
development by a congener, 3,4,3',4'-tetrachlorobiphenyl that is
rapidly metabolized  (see Tilson et al., 1979), and so persistence
and neurotoxicity are not necessarily related.  Therefore,
pharmacokinetic profiles should not be used as the sole means for

Tilson, H.A., Davis, G.J. McLachlan, and Lucier, G.W. (1979).
The effects of polychlorinated biphenyls given prenatally on the -
neurobehavioral development of mice.  Environm. Res. 18:466-474.

PCB - SAR Activity Relationships and Developmental Toxicity

     As this issue paper notes PCBs can impact on many hormonal
systems and may result in either changes in circulating levels of
the hormone or alterations at the site where the hormone_binds
(e.g., change in receptor number or affinity).  The ability of
PCBs to alter hormonal systems has been mentioned in several
recent papers  (and in several of the issues papers) as a possible
means by which these compounds can alter CNS development.
However,'this is an  area that has been unexplored experimentally
for either PCB mixtures or specific congeners.  Further, there is
limited data concerning hormone profiles in human populations
exposed to PCBs.  It should be noted that the  regulation of
thyroid function in  the rat and human  is qualitatively and
quantitatively different  in human and  rodents  {Dohler et. al.,
1979).  Humans have  a specific high transport  protein with high
affinity binding for thyroid hormones.  Rodents, birds, reptiles,
amphibians and fish  do not have a transport protein.  As binding
often  slows metabolism the half-life of T4 in  humans is about 5-9
days while that in rat plasma is about 12-24 hours.  Thus, in an
athyroid rat to restore  full thyroid function  10 times more T4 is
needed than  in a human.   Such differences should be  considered in
interpretation of animal  data.

Dohler, K.D., Wong,  C.C.  and van zur Muhlen, A.  (1979) The rat as
model  for  the  study  of drug effects on thyroid function:
consideration  of methodological problems.   Pharmac.  Ther: 5 ROS-

PCB Developmental Neurotoxicity  in  Humans

      This  issue paper makes clear the  difficulty in conducting
and interpreting  epidemiological studies.   The need for well-
designed  animal  experiments concerning the  neurotoxicity  of  the
PCBs  in developing  organisms  is  clear.  In  addition the need for
further well-controlled  studies  in  exposed  human populations  is

PCB Developmental Neurotoxicity  in Experimental Animals

     The  sentence reading,  "However, several previous studies
reviewed  by  Tilson  (e.g.,  Pantaleoni et al., 1988) indicate the
prenatal  period may be  the most  sensitive period for exposure,"
should read  "As Tilson  has noted developmental abnormalities
indicative of neurotoxicity have been reported in experimental
animals with both gestational and lactational exposure.  However,
the recent work reported by Lilienthal & Winneke (1991) indicates
the prenatal period may be the most sensitive period for
DISCLAIMER.  The research described in this article has been
reviewed by the Health Effects Research Laboratory, U.S.
Environmental Protection Agency, and approved for publication.
Approval does not signify that the contents necessarily reflect
the views and policies of the Agency nor does mention of trade
names or commercial products constitute endorsement or
recommendation for use.

                    Pre-meeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research Triangle  Park, NC
                      September 14-15, 1992

                          John A. Moore
              Institute for Evaluating Health Risks
                1101 Vermont Avenue, NW,  Suite  608
                      Washington, DC  20005


     The letter of invitation stated that a main focus of the
Workshop is to focus on sufficiency of data for use in risk
assessment.  I am of the opinion that data must meet a certain
standard of quality and integrity before it can be used for risk
assessment purposes by health or regulatory organizations.  This'
is certainly not a novel position given the formal promulgation
of Good Laboratory Practices a number of years ago.  However,
this issue may need to have some consideration as a prelude for
this Workshop given that almost all the data to be reviewed arose
from studies that were not designed or conducted with risk
assessment in mind.

     Applying a modest rigor as to quality and completeness of
data, as a prerequisite for use, reveals that there is a
significant segment of the PCB literature that has little utility
risk assessment.  The Wisconsin studies on Aroclor 1016 and
Aroclor 1248 in rhesus monkeys illustrate the issue.  The results
of the basic studies are contained in the Ph. D. thesis of D.A.
Barsotti; the Aroclor 1016 and Aroclor 1254 studies were
published in Toxicology 30:31-44 (1984) and Food Cosmetic
Toxicology 14:99-103 (1976), respectively.  There are a variety
of issues that compromise the interpretation of the study; they

     1. There are no adequately matched controls given that they
     are from a group of monkeys that had been at the laboratory
     several years longer than those in the treated groups.

     2. It was stated that 5 of 8 randomly selected monkeys had
     adipose tissue concentrations of 0.69 ppm of Aroclor 1016
     before the study started.  This Aroclor is not typically
     found as part of the background burden due to the ubiquity
     of PCB in the environment.  This is either cross
     contamination or erroneous interpretation of gas

     3. The Aroclor 1016 values, of infants at birth, were given
     as 1.65 ppm for the 0.25 ppm group while that of the
     controls was 1.54 ppm.  In other words after almost a year
     on a 0.25 ppm diet, or an equal period of time on a control

      diet, some controls have a level that is essentially
      identical to the treated group.

      4. As discussed in the thesis there was a third treated
      group in the original design that was never reported in the
      published paper because of cross contamination with PBBs.

      5. GC tracings in the manuscript clearly reveal the presence
      of congeners that could only have resulted from exposure to
      a PCB other than Aroclor 1016.   These peaks probably reflect
      exposure to Aroclor 1248,  which  was also studied in the
      Wisconsin facility.

      6. The 1016 study gives a PCB value at birth of 3.37 ppm for
      the infants whose mother received 1 ppm in the diet.   The
      paper that reports the 1248 results gives a PCB value at
      birth of 2.8 ppm in infants whose mother had received 2.5
      ppm in the diet for a similar period of time.   Aroclor 1016
      is known to be well metabolized  particularly in contrast to
      Aroclor 1248 whose; higher chlorinated congeners resist
      metabolism and accumulate in lipid vet Aroclor 1248  has a
      lower reported value at birth.   This defies all that  is
      known about the absorption,  distribution and metabolism of

      The conclusion is that contamination of the study groups
repeatedly occurred.  There is evidence that exposure to Aroclor
1016  occurred prior to the study,  that infant  exposure to  Aroclor
1248  occurred in the 1016 study,  some contamination by PBBs
occurred,  there is confounding  data that  monkeys exposed  to
larger doses  of a higher chlorinated  Aroclor had body burdens
that  were lower than monkeys exposed  to  lower chlorinated
Aroclors.   I  submit that these  facts  preclude  the utilization of
these data  for risk assessment.   Should  one persist  in the use of
these studies,  the data do not  permit  one to conclude that there
was any effect due to Aroclor 1016.   In  a similar vein the use of
the Aroclor 1248  data in other  than a  weak  qualitative sense
would be a  gross  speculation, not  science.

      For the  reasons  cited above,  the  series of neurobehavioral
studxes  by  Bowman,  Schantz,  Levin, Heironimus, and other
colleagues, that were performed on the surviving  infants from
these  studies,  can not  derive a causal interpretation.

     A second illustration  of the generic concern as to the use
of data  for risk assessment  is  the data from the Michigan cohort
published by  Jacobsen  and  colleagues.   There are  flaws in the
initial  aggregation and  characterization of the cohort which
fatally  compromises any  definitive conclusion as to a causal
relation to PCBs.   Paneth  (67) published a critique of the
studies about a year ago.


     The overview provided is quite selective in the data cited.
The following narrative/ extracted from a draft review document
on PCBs by Moore, Fry, Hinton & Kimbrough, may provide a useful
supplemental perspective on some topics.

     Any discussion of the health or environmental effects of
polychlorinated biphenyls (PCBs) is potentially fraught with
problems.  It is common to talk about and analyze PCB data as if
"PCBs" were a single chemical or single mixture of chemicals.
This ignores reality and leads to generalizations that are
frequently wrong.  Aroclors were made as different chemical
mixtures, to provide a set of physical and chemical
characteristics that met certain use criteria. Knowledge that has
accrued over the years indicates that biological (toxicological)
properties are related to the number and location of chlorines
attached to the biphenyl molecule.  This seminal observation
continues to be ignored in PCB risk assessment.

     Current knowledge of the health or environmental effects of
PCBs clearly indicates that the nature and severity of effects
are dependant on the specific chlorobiphenyl or mixture of
biphenyls to which the organism is exposed.  Much of the older
literature has limited use in assessing the adverse effects of
PCBs because the identification and characterization of the PCB
studied was not well defined based on todays understanding of PCB
toxicity.  Most of the current risk assessments and quantitative
estimates used by public  or environmental health agencies
reflect judgments formed at an earlier time when there were
limited data available and the large differences in toxicity
potential of specific PCB congeners or mixtures were either
unknown or inadequately understood.  While it may be clear in
1992 that a risk assessment policy that treats all PCBs as
toxicologically equivalent is not supported by current knowledge,
some recent proposals to develop policies that incorporate the
use of congener specific data were judged to be premature due to
a general lack of congener specific data and inadequacies in some
of the data that does exist.  Despite current inadequacies with
congener specific approaches it is through this type of
consideration that an accurate qualitative and quantitative
estimate of PCB risks will finally be achieved.  The current,
challenge is to modify existing "one PCB equals all other PCBs"
approaches by  developing estimates of risk utilizing data that
best approximate the material that is causing the exposure, i.e.,
in occupational settings, the actual Aroclors that were used and
in environmental exposures, data that best match the pattern that
is being observed.

     Dermal Effects.  Frequent findings in morbidity studies were
reports of skin rashes and dryness of the skin in capacitor
workers.  These workers were also usually exposed to solvents
that would have a drying effect on the skin.  In all but two
instances  (40,65) well-characterized chloracne was not present in

 even the most heavily exposed workers.    When PCBs  were first
 made commercially in a small factory in the 1930s,  workers
 developed chloracne after the production process  was
 accelerated.   The PCBs turned dark and  had poor electrical
 qualities.   It is very likely that increased amounts  of
 chlorinated dibenzofurans were also formed during that  period.
 Chlorinated dibenzofurans are known chloracnegenic  agents  (17).
 In another study,  skin lesions were described following exposure
 to heated PCBs.   Polychlorinated dibenzofurans may  have been
 involved here as well (65).  Skin lesions mentioned  in other  ,
 studies  did not  seem to represent typical chloracne.  in most
 other worker studies,   definite chloracne was not observed,
 suggesting that  PCBs,  per se,  do not produce chloracne.  ,

      Human Immunotoxicity Studies.   There are no  studies in
 humans involving PCBs and the immune response.  Immune  effects,
 principally reduced delayed  hypersensitivity response,  decreased
 IgA & IgM and increased response to mitogens were reported as
 part of  the initial evaluation of some  patients with  Yusho or
 Yu-Cheng.   These effects moderated with time and  are  not
 considered to be the effects of major health significance in
 these patients.   Japanese researchers now believe that
 polychlorinated  dibenzofurans  are the primary causative  agents of
 Yusho and Yu-Cheng (53).   This conclusion has gained  broad
 acceptance.   These diseases  are also discussed in the congener
 specific section.

      Experimental  Immunotoxicity Data. , The  composite data do
 indicate that PCBs can selectively  modulate  responses of the
 immune system, i.e.  some indices  are depressed while  others are
 stimulated  or unaffected.  There  appears  to  be  a  clear difference
 in  species  sensitivity as  to the  dose required  to cause  an       ;
 effect;  rats  are less  sensitive than primates  (53,81).   There are
 also data which  indicate that  the higher  chlorinated  PCB mixtures
 are the  more  immunosuppressive (18).  A study has shown  that
 exposure to Aroclors  can antagonize the  immunosuppressive effects
 of  other chemicals, principally TCDD (18).  Many of the  studies
 that  found  some  depression of  the immune  response administered
 doses that  clearly would (or did) cause generalized toxicity
 (20,77,78).   Generalized illness  can result  in physiological
 changes  in  the body that, in turn,  result  in  a  depression of the .
 immune response.   Such an effect  is  regarded  as a secondary
 response rather  than a direct  effect  of a  chemical.

     Humoral  immunity, the immune component that results in the
production of antibodies that  are commonly found in blood serum,
 seems to be the most sensitive  to PCB exposure.  This is
 frequently demonstrated as a decrease in the antibody response to
an  injection  of  sheep  red blood cells.  Depressed antibody
response is not uniformly observed  even in the same experiments;
in  two separate primate studies, where there was reduced antibody
response to sheep  red blood  cells, response to either tetanus
toxoid or to pneumococcal antigen was not affected  (81,85) .

     In contrast to the number of studies that observed humoral
suppression, effects on the cell mediated component (protection
against viruses, fungi and rejection of tumors and foreign
tissue) of the immune system is not well documented.  For
example, in a recent rat study Aroclor 1254 was administered over
a large range of doses with the top levels producing evidence of
generalized toxicity (77) .  A number of assays which measure cell
mediated immunity were used with no effects observed in three of
the tests; in the fourth assay, natural killer (NK) cell response
to tumor cells, was depressed at the two top doses.  Systemic
toxicity was also seen at these dose levels.  Another means of
assessing immune  capability is to examine  resistance to an
infectious challenge.  In several of these studies, resistance
was diminished; however, the dose levels of PCB administered were
quite high which compromises interpretation for human risk
(25,58,81).  Using susceptibility to transplantable tumors to
assess immune function, an enhanced immune capability was found
in that tumor transplantability was inhibited in Aroclor 1254
treated rats  (52) .

     An extensive series of immunological studies were conducted
on a group of rhesus monkeys exposed to Aroclor 1254 over a   ,
period of 55 months. (84-7)  These studies were reasonably well
designed, seemingly executed in a sound manner, and reported that
some of the immune responses, using well accepted procedures,
were impaired; yet other classical procedures, which often show
effects in such circumstances, remained normal.  In some
instances responses that were impaired at 23 months were either
normal at 55 months or were, at best, weakly impaired. * As is
true with other data, interpretation of these experimental data
as to human risk is fraught with uncertainty.  In large part PCB
immunetoxicity remains a phenomenon in search of disease
significance, particularly at environmentally relevant doses.

     Toxic Equivalency Factors.  There have been a variety of
attempts to. aggregate PCB congeners according to toxicological or
other biological properties.  Segregation according to degree and
location of chlorination has some utility when considering
factors such as water solubility or persistence in the
environment.  However, definite limitations exist from a
toxicologic potency point of view since it was demonstrated a
number of years ago that the dose of symmetrical hexachlorinated
biphenyl congeners required to produce a systemic toxic effect
can vary.  The disease pattern produced by specific congeners was
also different  (11,64).  It was observed in these and other
studies that certain chlorinated congeners could assume a
coplanar shape; these coplanar PCBs were toxic at very low dose
levels, produced a  "dioxin like" toxic syndrome, and they also
induced the characteristic pattern of enzyme response.  Based on
a series of studies that demonstrated similarity to chlorinated
dioxins and dibenzofurans, it was proposed that Toxic Equivalency
Factors could be assigned to this select group of PCB congeners.
The EPA sponsored a workshop in 1990 to discuss the use of a TEF
approach to the assessment of risks posed by exposures to

mixtures  of PCBs.   It was  proposed that  11  PCB  congeners be
assigned  a TEF relative to that  of TCDD  which was assigned the
unit value of  1.  (4)   These 11 PCBs included 3  coplanar and  8
mono ortho congeners.   The workshop  identified data gaps,
including limitations in the accurate identification and
quantification of  these congeners,  that  hindered adoption of the
TEF scheme for PCBs at that time.

     Notwithstanding the conclusions  of  the PCB-TEF Workshop, EPA
in 1991 announced  that feasibility of incorporating Toxic
Equivalency Factors for 3  coplanar PCBs  would be part of a broad
based  initiative to re-evaluate  the risks associated with
dioxins,  i.e.,  3,3',4,4' tetrachloro;  3,3 ',4,4',5 pentachloro;
3,3',4,4',5,5'  hexachloro  biphenyl.   The inclusion of these  PCBs
in this initiative is  believed by  a number  of scientists,
including two  authors  of this review  (Moore and Kimbrough) ,  to be
ill advised from a scientific perspective.  The following
illustrate the concerns.   Recently published studies with
developing fish indicated  that the toxicity of  the PCB congeners
studied was 14-89  fold less than that  predicted by the published
TEF values (88) .   Ongoing  studies  in mice,  where magnitude and
persistence of mixed function oxidase  enzyme induction is
monitored,  show that  the TEFs are  overestimating the potency of a
number of PCB  congeners by a factor of 10-1000  (9).  Finally, in
providing public comment to EPA, General Electric cited three
instances where use of TEFs predicted  results that were grossly
in error  with  actual  data;  liver tumors  in  rats, immunology  data;
and enzyme induction.   It  would  appear that moving forward with
this approach  for  PCBs still remains  ill advised.  There is,
however,  an intellectual appeal  to  the concept within the
scientific community,  suggesting that  it will continue to be the
focus  of  research  efforts.   It is predicted that a more rigorous
scrutiny  of the types  of factors that  should be considered in
establishing a  TEF will be proposed and  implemented.  Its utility
in predicting  avian toxicity in  the environment is also being
actively  pursued.   Greater sensitivity and  specificity will  also
require application of valid models for  estimating
bioaccumulation as well as  absorption, distribution metabolism
and excretion  in species of interest,  including man.

     Yusho and  Yu-cheng.   In 1968 and  1979,  two poisoning
outbreaks  occurred in  Japan and  Taiwan, respectively
(16,54-55,63) .  Following  the ingestion of  rice oil, which had
inadvertently become contaminated with a heat transfer fluid, the'
exposed population developed chloracne, hyperpigmentation of the
skin and mucous membranes,   fatigue, digestive problems,  and other
signs  and  symptoms.  Infants born to females exposed during
pregnancy were  small, had  dark pigmented skin,  and abnormal
dentition.   In  the  1968 outbreak, polychlorinated biphenyls were
initially  identified in the heat transfer fluid.  It was assumed,
therefore, by the Japanese  investigators that PCBs had caused the
illness in  the  consumers of  the  adulterated rice oil.   Adipose
tissue levels of PCBs  found  in this population in a few samples
collected  in 1968 were  higher (76,  13, 32, and 46 ppm)  than

levels found in the general population in Japan and in the United
States.  However, when these levels where compared to levels in
workers, they were 'found to be at the lower end of the range of
levels found in occupationally exposed populations (54).   This
suggested that PCBs could not be responsible for the Yusho
poisoning, since workers with similar, or higher, body burdens in
the United States or in Japan were not suffering from these
symptoms.  Subsequently, polychlorinated dibenzofurans (PCDFs)
and polychlorinated quarterphenyIs (PCQs) were identified in the
rice oil.  More recently, chlorinated naphthalenes and
dibenzodioxins have also been found in the cooking oils.
Furthermore, the gas chromatographic pattern of the PCBs of the
Yusho patients was different from that of the general population
(55) .  The average intake per person during the intoxication
period in Taiwan was 973 itig of PCBs,  3.8 mg of PCDFs and 586 mg
PCQs.  In Japan, the average intake was 633 mg PCBs, 3.4 mg PCDFs
and 596 mg PCQs  (16,63)." It is unclear how much of the PCDDs and
chlorinated naphthalenes were ingested.  It is also unclear
whether most of the people with symptoms had ingested more than
the averages given here.  In none-of the studies were attempts
made to correlate signs and symptoms in the patients with body
burdens of these chemicals at the onset of the outbreak.   The
available scientific evidence now suggests that the poisoning
outbreak was caused primarily by the PCDFs.  Perhaps the
polychlorinated naphthalenes and PCDDs contributed as well.
Studies in rats and cynomolgous monkeys attempted to determine
the role each of these components in the toxic syndrome by
feeding the three materials separately and in combination.  The
studies showed that the  severity of immunosuppressive effects
were primarily attributable to the polychlorinated
dibenzofurans.  These findings, along with observed effects on
enzyme induction and pathological findings in liver and skin, led
•the researchers to conclude that polychlorinated dibenzofurans
were the primary causative agents of Yusho and Yu-Cheng (53).

     Human Carcinogenicity Studies.  Most mortality studies were
conducted in small groups of workers or in relatively young
worker populations with  a small number of deaths.  In all but one
study, a healthy worker  effect was noticed and the overall death
rate was lower than expected.  A healthy worker effect refers to
a common finding  that the worker population has less illness
than the general population to which it is being compared.  A
worker population typically contains a preponderance of people
that are in robust health whereas a general population will
contain all peoples including those with severe debilities or
other effects that makfe  them unfit for work.  In two studies an
increase of melanocarcinomas was noted.  However, in the first of
these two studies  (3), the number of workers was quite small and
these workers were also  exposed to other known carcinogens.  In
the other study  (76) the increased incidence of melanocarcinoma
was unrelated to length  of exposure, and the exposure assessment
performed by the authors appears to have been flawed.
Furthermore, of  6 carcinomas in males versus 2 expected, one of
the melanocarcinomas did not originate in skin but in the gall

 bladder,  one melanocarcinoma was diagnosed prior to  onset  of
 exposure,  and in at least one case,  melanocarcinoma  was  not the
 cause of  death.   In females,  2 melanocarcinomas  were observed,
 and less  than one was expected.  An  increase  in  melanocarcinomas
 was not observed in the other mortality studies.

      An increase of carcinoma of the rectum and  the  liver-was
 noted in  a study of capacitor workers in one  of  two  plants  (13).
 However,  the increase in carcinoma of the rectum was less
 pronounced in a  follow-up study,  when larger  numbers of  death
 certificates were available and the  carcinomas of the liver, upon
 review of the medical records,  did not turn out  to be all primary
 carcinomas of the liver.   These observations  made the claim of an
 increase  of these tumors  tenuous  (13).   In an additional study
 conducted in a plant in Italy (8), the overall death rate was
 increased over expected rates.   Automobile accidents contributed
 to  the increase  in deaths in  women in this group  of  workers.  The
 total number of  deaths was small  and the workers  were exposed to
 other chemicals.   Some workers  with  cancer did not have  exposure
 to  PCBs.   The results obtained in this study  are  difficult to
 interpret  in part because of  the  limited number of death
 certificates and the uncertainties about exposure to PCBs.
 Additional follow-up of this  cohort  would be  necessary to clarify
 these preliminary findings.

      In the aggregate,  the scientific evidence does  not
 demonstrate that  PCBs are carcinogenic in humans.  Additional
 studies are_needed to examine this problem further,  since the
 latency period may not have been  long enough  in some instances,
 and the number of workers in  other studies was small.

      Experimental Carcinogenicity Data.   There has never been a
 systematic series of long term  studies  performed  on  the  range of
 commercial mixtures  that  were commonly used in this  country.
 Thus,  our  ability to clearly  distinguish between  Aroclors as to
 dose,  incidence and severity  of liver pathology,   including
 tumors, is compromised.,

     A consistent carcinogenic  response was seen  from  studies in
 rats_fed a_PCB mixture Containing 60%  chlorine.   The data on the
 Carcinogenicity, of  lower  chlorinated mixtures of  PCBs  is less
 clear.  There  are fewer studies, a number  of which are flawed in
 design  and execution.   However, it is  clear from  the aggregate
 data that  the  magnitude of tumor response  in rodents differs
between the commercial  mixtures ,that have been tested.  A recent
 re-evaluation  of  five  key PCB studies  in rats illustrates this
point  (24).  In this  re-evaluation all liver tissue  from each of
 the studies was examined  by a group of nationally recognized
pathologists using the  tumor  criteria promulgated by the federal
government's National  Toxicology Program.  Consistency of tumor
diagnoses  across  all studies provided added confidence that the
differences between  the 'studies was likely due to inherent
differences in the PCB mixtures studied.  A comparison of each
study result showed a consistent liver tumor incidence of about


80% response with PCB mixtures having 60% chlorination.  In
contrast a statistically significant increase in liver tumors was
not seen in studies of PCB mixtures with either 54% or 42%

     There are other studies that reported a benign liver tumor
response in rats fed Japanese PCB mixtures with either 48% or 54%
chlorination.  Interpretation of these positive results from a
human risk point of view is difficult since the doses at which '
tumors were observed exceeded (often grossly) a maximum tolerated
dose.  The studies with the 42% chlorine PCB mixture were
negative, as were the lower two doses (500 & 100 ppm) of the rats
fed the 48% chlorine mixture.  However,  the small sample size and
duration of the studies limits the power of the finding '(27-51) .

     Studies in mice with Aroclor 1254 resulted in benign and
malignant liver tumors.  The dose administered clearly exceeded
the MTD  (50) .  Another study in mice, using Japanese PCB mixtures
with 54% chlorination caused benign and malignant liver tumors at
the highest dose tested (500 ppm), doses of 250 and 100 ppm were
negative.  In these same experiments mice treated with PCBs
containing either 48% or 42% chlorine did not develop liver
tumors at doses as high'as 500 ppm.  (26).  A compilation of the
carcinogenicity data has been provided by Silberhorn, Glauert and
Robertson (75).

     The traditional  (and current) EPA policy is to view all PCB
mixtures as carcinogens with equal potency. '• EPA derives its •
quantitative estimates of cancer potency from the rat study of a
PCB mixture with 60% chlorine that yielded the highest incidence
of tumors; this value is then used as a surrogate for all PCBs.
In July, 1991 the results of a;'pathology re-evaluation of five
key rat studies, including those used by EPA for quantisation of
cancer risk,  was submitted to EPA with a recommendation that its
traditional policy, which is not supported by science, be
reconsidered.  EPA was also urged to employ procedures that
utilize all available study results in estimating cancer
potency.  Incorporating any one of the recommendations would have
resulted in a decrease in the current estimate of cancer
potency.  EPA acknowledged receipt of the material, praised the
quality of the pathology review,' and stated it was considering
the recommendations.  No further action has been announced.

     The tumor response (or any other toxic effect) is not
believed to be caused through mutagenic mechanisms, i.e., the
PCBs are not held to be'genotoxic  (89).  While there is some
disagreement among scientists as to the significance of certain
tumor responses - in rodent studies for predicting risk to humans,
there is broad international agreement that a chemical that
causes tumors in rodents and can cause genetic damage; i.e., is
genotoxic, constitutes a1 human risk.

     A number of investigators have attempted  to assess the
ability of specific congeners to affect the carcinogenic process

 by utilizing a laboratory tumor model in rodents.   This  model
 permits identification of chemicals that may initiate the
 carcinogenic process,  promote the expression of  induced  cells  or
 act as complete carcinogens.   In these systems rats are  given  a
 low dose of a known carcinogen which,  by itself  will not result
 in the development of  a liver tumor.   If however,  the rat
 receives continual doses of a second chemical, which by  itself
 can not cause tumors to develop,  and a tumor develops this  second
 chemical is said to "promote" the development of a tumor incited
 by^exposure to the first chemical.   PCBs are viewed by many
 scientists  as acting  as a tumor promoter in rodent studies.

     As would be predicted from studies on TCDD, the coplanar
 ("dioxin like")  PCB congeners are effective promoters in these
 tumor  models.   There are very large differences  in the doses at
 which  promotion occurs (15).   There are no data  that suggests
 that PCB congeners are complete carcinogens in these systems;
 studies to  determine if they  are initiators have not been
 performed.   Some congeners that are not "dioxin  like"  have  also
 promoted tumor growth.   These congeners do induce  a different
 form of mixed function oxidase that also has been  found  to  be
 associated  with tumor  promotion (14) .   It has been postulated
 that promotion may be  related to mitogenic stimulation which
 occurs concurrently with,  not as a  result of, induction  of  the
 mixed  function oxidase enzymes (19).                  .        .

     _There  is  interest in determining  the accuracy of congener
 specific results,  obtained from these  initiation/promotion  test
 protocols,  in quantitatively  predicting the liver  tumors  response
 in long term studies with Aroclors.  If tumor model  results
 correctly identify such congeners within a PCB mixture,  and their
 relative potency,  they would  then have predictive  value  for
 "mixtures"   found in the environment.   There is  also  intrinsic
 value  in identifying those congeners that are not  active  in these
 cancer models  since they currently  are treated as  part of the
 total  PCB that  is  assumed to  cause  a maximum tumor response.

     There  are  two major assumptions associated with  the utility
 of these data  for  human risk  assessment.   First  is  an  assumption
 that the mechanism by which commercial  Aroclor mixtures cause
 liver  tumors in  rodent  livers  is  through a "promotion" type of
mechanism.   There  is a  plausible  basis  for this assumption that
would  be accepted  by many.  Second, if  the results  in  such a
 system are  held  to have relevance for  estimating human risk, of
 cancer positive  and negative  results should have utility.  These  .
 systems  are not  well understood outside  of  the cancer  research
community,  and results  may not  be quickly or broadly accepted.


     The following narrative,  extracted  from a draft review
document  on PCBs by Moore, Fry, Hinton &  Kimbrough, may provide a
useful  supplemental perspective.

     Taylor et al.  (79-80) found a clinically insignificant
decrease in birth weights of the offspring from female capacitor
workers.  This is the only study of reproductive effects in PCB

     Many studies demonstrate adverse reproductive effects from
PCB exposure. There are differences in species and strain
sensitivity, with the greatest sensitivity, from a dose
perspective, seen in mink and rhesus monkey.   The nature of the
developmental toxicity response varies with species with mice
likely to produce malformations, rats having decreased fetal
weight or resorptions, and rhesus monkeys decreased in utero
survival.  In general, adverse effects are produced at lower dose
levels with the higher chlorinated mixtures.

     Many of the effects occur at doses where maternal toxicity
was observed.  When PCB mixtures were found to produce structural
(teratogenic) malformations it is quite plausible that these
effects are caused by congeners that have "dioxin like"
properties, i.e., the coplanar and possibly the mono-ortho
coplanar moieties.  A number of PCB congeners have been
systematically examined in mouse teratology studies (60-62).  The
classic malformations caused by TCDD, cleft palate and kidney
hydronephrosis, were observed with 3,3',4,4' tetra,
3.,3' ,4,4' ,5,5' hexa, and 2,2',3,3',4,4' hexachlorobiphenyl,
whereas no effects were seen with the symmetrical 2,3,5; 2,3,6;
2,4,5; 2,4,6 hexachloro, the 4,4' dichloro, or the 3,3',5,5'
tetrachloro congeners.  There is a study that showed 2,3,3'4,4',5
hexachlorobiphenyl could enhance the incidence of cleft palate in
TCDD treated mice; 2,2',4,4',5,5' hexachlorobiphenyl was without
effect  (10).  On the other hand, exposure to Aroclor 1254
antagonized the teratogenicity of TCDD (23) .  While these latter
studies demonstrate, in laboratory environments, that there are
interactions between PCBs and TCDD the public health relevance is
not currently understood.  Chlorinated dibenzofurans,  which cause
"dioxin like" developmental toxicity in animals, are believed to
be the primary cause of the developmental toxicity observed in
Yusho and Yu-Cheng.

     There are several studies in rhesus monkeys dealing with
reproduction and development. In these studies, adult females
were exposed to either Aroclor 1254, 1248, or 1016 for a period
ranging from 6 months to two years.  In general, the ability to
conceive was not consistently impaired even when there were signs
of general toxicity.  In the Aroclor 1254 study, all monkeys
failed to deliver live young  (2); with Aroclor 1248 all but one
monkey at the 5.0 ppm dose aborted, resorbed their conceptus or
delivered a stillborn baby; only 5 of 8 mothers delivered live
young at the 2.5 ppm level  (6).  The live young were reported to
be small at birth and developed signs of toxicity after two
months of nursing; three died in the first year.  The same
laboratory did a further study with Aroclor 1248 using doses of
1.0 or  0.5 ppm; conception was normal, there was some fetal loss,
and babies were small at birth  (1) .  Studies with Aroclor 1016

 did not impair fertility and resulted in  100%  live births.   The
 author's reported a reduced birth weight  in  the  1.0 ppm group.
 There was no statistically significant difference between this
 group and the controls at weaning.   The utility  of the  Aroclor
 1248 and 1016 studies  for risk assessment is highly suspect  due
 to  the issues of cross contamination with different PCBs and
 other halogenated aromatic hydrocarbons.   The  studies with
 Aroclor 1254,  which were performed at a different laboratory, are
 not associated with the issues of misdosihg  and  cross
 contamination.   Failure of the rhesus monkey to  maintain
 pregnancy to term may  reflect a sensitivity  of certain  fetal
 stages to PCBs or an impairment of placental function.   From a
 cross species extrapolation point of view the  latter would appear
 to  merit investigation since the endocrinology of pregnancy  in
 the rhesus monkey differs from that of humans.   If the  effect in
 rhesus monkey is due to disruption of placental  hormone
 regulation the relevance to human risk may be  diminished.

      Despite there being major flaws in these  rhesus studies,
 general trends are clear;  the higher chlorinated mixtures are
 more toxic as manifested by maternal toxicity, failure  to carry
 pregnancies  to term, and postnatal  toxicity  as a result  of
 nursing.   The mixture  with the lowest degree of  chlorination,
 Aroclor 1016,  was arguably associated with,  at best, a  very  weak
 effect.  _The toxic effects that develop in female rhesus monkeys
 of  breeding  age do not affect the major endocrine parameters that
 characterize the menstrual cycle (83)  or  the ability to  conceive
 upon mating.

      A likely explanation for the toxicity that  develops in  the
 nursing rhesus  is the  qualitative and quantitative nature of the
 PCB consumed.   For example,  in the  Aroclor 1016  study,  the mother
 and infant PCB burdens were generally equivalent at birth; at
 weaning,  the maternal  level was unchanged or marginally
 increased, yet  the infant  level had increased  7-8 fold  (7).  The
 array of  congeners to  which the rhesus  infants were exposed  are
 probably  different than that to which a nursing  human infant
 would be  exposed.   In  the  Aroclor 1016  study (7)  there was clear
 evidence  from the gas  chromatograms  that  the congeners  in rhesus
 milk represented those from the store in maternal adipose (which
 would occur  in  humans)  and congeners  which are present  in the
 Aroclor and  not stored by  the mother.   Presumably the latter
 congeners  are  those which  are metabolized  and  secreted by
 adults.   The toxicologic  significance,  if  any,  of this difference
 in  the laboratory study:and real  world  is  not known.


      The  following narrative,  extracted from a draft review
 document  on  PCBs  by Moore,  Fry, Hint on  & Kimbrough,  may provide a
useful  supplemental perspective.

      Two  groups of  researchers  have reported on studies that
attempted  to determine whether  PCBs affect infant development.

One group established a cohort of women who stated that they
consumed fish from Lake Michigan and used women who stated they
were non-consumers of Lake Michigan fish for comparison
(28-33,72) .  The. other researchers enrolled a group of women in
North Carolina and evaluated the observed effects in children by
relating them to body burdens calculated for the mothers
(21-22,68-71).  Both groups of researchers attempted to determine
whether neurobehavioral effects resulted solely from in utero
exposure or from in uterp exposure and consumption of
PCB-containing human milk.

     The Michigan studies have flaws which fatally confound their
interpretation.  A most basic issue is the degree to which it is
appropriate to .contrast the results observed in the control and
exposed populations and conclude that any observed differences
are due to PCB exposure-  For example, there was only a modest
correlation between estimates of PCB consumption and measured
maternal PCB levels; there was essentially no correlation between
estimated consumption and serum PCB levels in infants; there is a
very weak correlation between milk levels and fish consumption;-
maternal serum PCB levels in the control and exposed mothers
overlapped.             •.:....

     -The studies are based on a group of infants of which 242 .are
considered exposed due to their mothers' stated recollection as .
to amount, type and frequency of consumption of fish .from Lake
Michigan.  Seventy-one children, whose mothers denied eating Lake
Michigan fish a're not considered "exposed" and  are included as a
comparison group.  The data, which are found in two papers
(28,72), are not tabulated in such a way that it can be • ..
determined in how many of the 71 "^controls" (non-fish .eaters) and
of the 242 "exposed" (fish eaters)  PCBs were actually measured.
Slightly less than 200 sera were analyzed.  At least some of. the,
samples must have originated from non-fish eaters.   It is   •  ,
possible that PCBs were not determined or not measurable in more
than, a third of each of the two groups.  It is also not clarified
in the papers how the non-detectable levels and the
non-quantifiable levels are dealt with in the various
correlations made with behavioral effects.  The accuracy of .these
exposure criteria are important since the papers report
associations between "adverse affect" and either PCB ingestion
history or levels in serum.

     Aside from the uncertainties introduced in the exposure ,
assessment, the behavioral tests used in these studies need to be
examined as well.  These tests are not well-standard!zed and have
many confounders.  It is also not clear from the paper by
Schwartz et al. (72) whether some of the infants with low birth
weights were actually premature infants and how prematurity would
affect the results of behavioral tests.  Another paper (28)
stated that cord serum values were not available for 36.7% of the
infants and only for 48% of the "highest fish eaters."  They were
available on 65% of the remainder of the sample, apparently
including fish eaters.  Here the range of the birth weight is

given as 909.1 grams to 5028.9 grains.  Obviously premature
infants were included in this cohort.  A thorough critique of the
Michigan cohort has been provided by Paneth  (20).

     A second cohort of children was established independently in
North Carolina by Dr. Walter Rogan and coworkers (21-22,68-71)
between 1978 and 1982 to measure exposure to polychlorinated
biphenyls, DDT  (1,1,1 trichloro 2,2-bis-(p-chlorophenyl)-ethane)
and its derivatives, mainly DDE (1,l-dichloro-2,2-bis-
(p-chlorophenyl) ethylene), and to evaluate health outcomes.  In
the North Carolina cohort, polychlorinated biphenyls and DDT and
its derivatives were measured in human milk at term.  The results
of these measurements were used to calculate the prenatal
exposure of the infants.   In these calculations it was assumed
that the body weight of the mother is 60 kg; that 25% of her
weight is fat; that the only means of excretion are pregnancy or
lactation; that the daily  dose was constant over a lifetime; and
that the mother is primiparous.  Based on these assumptions, the
authors calculated that the daily dose associated with any given
level in breast milk is the body burden at that level divided by
the age of the mother in days  (82). The authors then carried
these assumptions further by using the crude data in which the
outcome variable   (percent abnormal on the Brazelton assessment
or means on the Baileys) is arrayed by level in milk fat and
estimate a NOAEL by inspection.  For these data, that level
appears to be about 3.4 ppm (mg/kg) in breast milk fat.  For
visual recognition memory, the equivalent level is 1.0 ppm in fat
of breast milk.

     In the evaluation of  the infants from the North Carolina
study, a physical examination was performed at the time of
delivery, and the Brazelton Neonatal Behavioral Assessment Scales
(BNBAS) were administered  in the presence of the parents by staff
trained at centers certified by.members of Dr. Brazelton's
staff.  The BNBAS were performed in 51% of the infants during the
first week of life, in 20% during the second week of life, and in
16% during the third week.  Some investigators think that the'
BNBAS should only be administered during the first three days of
life.  If the analysis is confined to data obtained from infants
whose examination was given at day three or earlier, the sample
size is more than halved.  However, even with the smaller sample
size the effect of PCB and DDE on hyporeflexia remained
significant while the effect of PCBs on tonicity no longer
achieved significance.

     The BNBAS yields results of 27 behavioral scales, and 20
reflexes.  In order to better deal with these results they were
summarized into clusters by Rogan et al. (68) based on the
clusters, defined and summarized by Jacobson et al.  (32).  For
the BNBAS the only cluster scores significantly affected by PCBs
or DDE were the tonicity and reflex scores.  The tonicity cluster
score was affected by PCBs.  Higher PCB levels were associated
with less muscle tone and activity.  Both chemicals were
associated with hyporeflexia or hyperreflexia.  However, the

number of  infants  in which this was noted was small.  A total of
49 infants had mothers with PCB milk levels above 3.5 ppm  (mg/kg)
42.2 % of  these had hyporeflexia or hyperreflexia.  This
represents 21 children out of a total of 866 children that were
examined or 2.4%.  However, in Rogan et al. (69) it is stated
that 856 children  participated.  Thus, depending on the number of
children that actually participated, this percentage may be
slightly higher or lower.  The authors also noted that infants
who were delivered by emergency cesarean section or with low
forceps or suction had greater numbers of abnormal reflexes but
this did not change the significance of the positive association
with PCBs or DDE.  Furthermore, mothers with higher PCB blood
levels were older, and were more likely to have consumed one or
more alcoholic beverages per week.  It is not stated in the paper
whether alcohol consumption by the mother was positively
associated with abnormal reflexes in the infant.  There was also
a positive association between higher DDE (but not PCB) levels
and smoking.  It was not stated whether this was further
examined.'  None of the observed differences were still evident at
ages 3, 4, or 5 in these children (21).

     Since not all women donated all samples in these studies
(69), using a single sample such as milk at birth would eliminate
many women.  Therefore, values from all samples from a given
woman were combined and expressed as an estimated amount of
chemical in milk at birth.  Levels below the quantitation limits
were_changed to an estimated amount.  It was assumed that the
chemical levels had a log normal distribution.  Furthermore,
levels were multiplied by a scale factor to make them all
comparable to milk at birth.  This scale factor is the median
ratio of birth milk samples to the sample in question.  It
adjusts for the declining milk values over time and also for the
fact that serum levels are substantially lower than milk levels.
All available levels were averaged.   It was not clarified in the
papers whether all milk samples were actually milk samples, or
whether some of these samples represented colostrum.  Colostrum
fat usually has higher levels of chlorinated organic chemicals
than the fat of mature human milk (66).  Thus, colostrum would
predict higher body burdens and introduce  uncertainty into the
exposure assessment if many of the early samples represented
colostrum.  This and other variations in the concentration of
halogenated organic chemicals in human milk are also summarized
in a book edited by Jensen and Slorach (39).


     The following narrative,  extracted from a draft review
document on PCBs by Moore, Fry,  Hinton & Kimbrough,  may provide a
useful supplemental perspective.

     A review of human studies and studies in experimental
animals that were  in the published literature through 1988 have
been conducted by  Tilson et al.  (82),   This  review article
concluded that developmental exposure to PCB results in

persistent neurobehaviqral alterations that can be observed
across species and,in the absence of generalized toxicity.  The
most consistent finding was hyperactivity; effects on higher
cognitive processes, or learning, was also observed across
species.  Their section on quantitative assessment is quite
speculative; their summary of human studies is also less critical
than that which is. summarized in this review.        -

     This review focused primarily on papers that were published -
after the Tilson et al review.  As pointed out in the individual
reviews of these papers, all the experimental studies reporting
adverse neurobehavioral effects have problems with the design and
execution of the studies. The measurements used to determine the
effect on behavior have also not been standardized or have not
been used on large numbers of control animals to determine the
range of variation in a larger population.  The reported
neurobehavioral effects seem to occur at relatively low levels of
exposure, particularly in monkeys (12,57-58,73-74).

     In a developmental neurotoxicity study altered
neurobehavioral performance was seen in rats whose mothers
received 30 ppm Clophen A30; the 5 ppm group was similar to
controls (57).  In a subsequent study,in rats receiving 30 ppm
Clophen A30  (58), the same neurobehavioral parameters were
altered  (active avoidance learning and retention of a visual
discrimination task).  Using cross fostering techniques it'was
found that prenatal exposure could result in neurobehavioral
changes; postnatal exposure caused no behavioral changes.

     The non-human primates used in the study by Seegal et al.
(74) were obtained from two different sources.  They were either
obtained as feral animals or from a breeding colony.  It is not
clear whether they were obtained at the same time, how similar
they were,  or whether the non-human primates from the different
sources were equally divided among the groups.  Feral animals may
react differently to stress than animals from a breeding colony.
The monkeys were given bread soaked in honey for a four-week
training period, and then they continued to receive a piece of
bread daily soaked with honey, corn oil and PCBs.  It is not
clear whether the controls received the bread with the honey and
the corn oil.  The number of monkeys in the different groups was
not equal.   There were 4 controls and 2 or 3 animals per exposed
groups.  No information is given about the age and sex of the

     Apparently the animals were anesthetized every two weeks for
examination, weighing and blood collection for the measurement of
PCBs, although the PCB blood determinations are not reported in
the paper.   It is also not clear whether the control animals were
also anesthetized, nor is it stated what the anesthetic was.
However, at the end of the experiment the animals were euthanized
with sodium pentobarbital.  It was,  therefore,  assumed by the
reviewers,  that this drug was also used for the intermittent

anesthesias. The repeated anesthesias could have affected the
level of catecholamines measured in the brain by investigators.

     In taking the brain samples, 2 mm punches were taken from
different parts of the brain.  The number of punches from the
different brain structures varied from 2-3 to 6-8,  because some
of the regions are very small.   The average weight of these
pieces of tissue was 8 mg and dopamine was quantified on a
wet-weight basis.  It is difficult to biopsy exactly the same
area of brain in each animal.  If the locations where the punches
were taken was not exactly the same, particularly if only two
animals were used in a group, it could have introduced a
variation in the concentration of the amount of dopamine found in
the tissue.  It would be useful, therefore, to review the results
of each individual animal.

     Loss of moisture during processing of the brain may not have
been uniform.  More information on the sequence in which the
brain fragments of these animals were processed and the sequence
in which the dopamine was measured would perhaps alleviate some
of these concerns.  Dopamine was determined in the supernatant
from the brain punches of the cells.  Some insight into how good
and how variable the recovery is from tissue would be helpful to
evaluate whether the differences that are shown between the
groups are.real differences or whether they are part, of the
experimental variation, particularly since there does not seem to
be a dose response relationship.  The concentrations of
norepinephrine and epinephrine were apparently normal in the
affected animals.  This is difficult to understand since the
animals were dosed over a long period of time and since dopamine
is an intermediate in the conversion of tyrosine to epinephrine
and norepinephrine.

     In addition, in this paper the results of tissue culture
studies are also reported.  In these studies a 0.1% solution of
DMSO in water was used as a solvent for the PCBs and the
supernatant was analyzed for catecholamines.  It is unlikely that
the PCBs were in solution and penetrated the cells.  It would be
useful to know what the day-to-day variations are in the
analytical method used for the measurement of dopamine and
whether the controls were interspersed with the exposed tissue
culture extracts.  It also would be useful to know how much
variation occurs in the recovery of dopamine in the supernatant.
However, this information is not given in the article.

     An additional study from this laboratory was conducted in
rats (73).  The treated rats received 500 and 1,000 ppm Aroclor
1254 mixed into their diet for 30 days.  These are rather high
doses,  and the exposed rats gained less weight than the controls
during the study period.  No dose response relationship was
demonstrated.  In other words,  the rats receiving the 500 ppm and
those receiving twice as much had the same quantitative loss of
catecholamines.  Again, questions can be raised about day-to-day
variations in the analytical method by which the catecholamines

were determined, and how the decreased weight gain may have
affected the results.  It is also curious that the PCB levels in
the brain for some isomers at the higher dose were either the
same or less than the PCB levels at the lower dose.

     In an earlier primate study female rhesus monkeys, eight per
group, were fed 0, 0.25, and 1.0 ppm of Aroclor 1016 in the diet
for about seven months prior to breeding, during pregnancy and
through weaning of their young.  This study is discussed in the
reproductive and developmental toxicity section (6).  Additional
information on this study was gleaned from the thesis of D.A.
Barsotti (5).  It was stated in the thesis that another dose
group, not mentioned in the paper, received a lower dose of PCBs
(0.025 ppm) and was contaminated with polybrominated biphenyls.
No mention of this group was made, therefore, in the paper.  In
addition, review of the gas chromatograms depicted in the thesis
indicates that monkeys in the other test groups also had received
higher chlorinated biphenyls not present in the commercial
mixture of Aroclor 1016.  These findings confound any
interpretation of effects attributable to Aroclor 1016.
Neurobehavioral tests were performed on the four year old
offspring of the monkeys that had received 1 ppm (56).   Some mild
effects were noted on the highest dosed group when that group was
compared to the group receiving the lower dose, but not the
controls.  Since these monkeys received exposures to other
chemicals,  such as highly chlorinated PCBs and brominated
biphenyls,  these studies cannot be used to derive a tolerable
daily intake or to develop a quantitative risk assessment.

     It is difficult to interpret the various primate results.  A
well-designed study with larger groups of animals should be
conducted to get a better idea of the effects of PCBs on subhuman
primates.  At the same time, it is also not clear how predictive
results obtained in subhuman primates are for humans.  In many of
these studies the neurological studies were performed on
offspring that had developed signs of clinical toxicity during
infancy, or were in groups where other infants died due to toxic
effects in either themselves or their mother.  It appears that
humans are less sensitive to PCBs than rhesus monkeys.   This may
be due to differences in lipid metabolism, toxicokinetics of
PCBs, differences in hormone levels, fetal development  and PCB
excretion in milk (73-74).  In the different subhuman primate
studies, the controls frequently differed in age and other
parameters from the experimental groups.

     In summary, the studies on offspring raise interesting
questions about potential neurological effects of PCBs; however,
additional studies need to be done to confirm these results and
to elucidate these findings further.

     Finally, there are species differences among non human
primates as to PCB toxicity, with the rhesus monkey more
sensitive than the cynomolgous monkey and the squirrel  monkey
anecdotally reported as being somewhat insensitive.

Understanding the bases for these differences could also be  of
priority relevance in extrapolation' of data for human risk.


1.   Allen JR, Barsotti DA, Lambrecht LK, Van Miller JP. .
Reproductive effects of halogenated aromatic hydrocarbons on
nonhuman primates. Ann NY Acad Sci. 1979;320:419-425.

2.   Arnold DL, Mes J, Bryce F, et al. A pilot study on the
effects of Aroclor 1254 ingestion by rhesus and cynomolgus
monkeys as a model for human ingestion of PCBs.  Fd Chem Toxicol.
1990;28;847-857.                                  .  ,

3.   Bahn AK, Rosenwaike J, Herrman N, Grover P, Stellman J,
O'Leary K.  Melanoma after exposure to PCBs.  N Engl J Med.

4.   Barnes D, Alford-Stevens A, Birnbaum L, Kutz F, Wood W,
Patton D.  Toxicity equivalency factors for PCBs?  Quality
Assurance: Good Practice, Regulation, and Law. 1991;!:70-81.

5.   Barsotti DA.  Gross, clinical and reproductive effects of
polychlorinated biphenyls  (PCBs) in the rhesus monkey.
University of Wisconsin; 1980. Thesis.        .

6.   Barsotti DA, Marlar RJ, Allen JR.  Reproductive dysfunction
in rhesus monkeys exposed to low .levels of polychlorinated
biphenyls  (Aroclor 1248). Food Cosinet Toxicol.  1976'; 14:99-103 .

7.   Barsotti DA, Van Miller JP.  Accumulation of a commercial
polychlorinated biphenyl mixture (Aroclor 1016) in adult rhesus
monkeys and their nursing infants.  Toxicol. 1984;30:31-44.

8.   Bertazzi PA, Riboldi L, Pesatori A, Radice L, Zocchetti C.
Cancer mortality of capacitor manufacturing workers.  Am J
Industr Med. 1987;11:165-176.

9.   Birnbaum LS. Public statement and personal communication.
Washington, DC: U.S. Environmental Protection Agency; 1992.

10.  Birnbaum LS, Weber H, Harris MW, Lamb JL, McKinney JM.
Toxic interaction of specific polychlorinated biphenyls and
2,3,7,8 tetrachlorodibenzo-p-dioxin: Incidence of cleft palate in
mice.  Toxicol Appl Pharamacol. 1985;77:292-302.

11.  Biocca M, Gupta BN, Chae K, McKinney JD, Moore JA. Toxicity
of selected symmetrical hexachlorobiphenyl isomers in the mouse.
Toxicol Appl Pharmacol. 1981;58:461-474.

12.  Bowman R, Heironimus M.  Hypoactivity in adolescent monkeys
perinatally exposed to PCBs and hyperactive as juveniles.
Neurobehav Toxicol and Teratol. 1981;3:15-18.

13.  Brown DP.  Mortality of workers exposed to polychlorinated
biphenyls  - an update.  Arch Environ Health. 1987;42:333-339.

14.  Buchmann A, Kunz W, Wolf CR, Oesch F, Robertson LW.
Polychlorinated biphenyls, classified as either phenobarbital- or
3-methylcholanthrene-type inducers of cytochrome P-450, are. both
hepatic tumor- promoters in diethylnitorsamine-induced rats.
Cancer Letters. 1986;32:243-253.

15.  Buchmann A, Ziegler S, Wolf A, Robertson LW, Durham SK,
Schwartz M.  Effects of polychlorinated biphenyls in rat liver:
correlation between primary subcellular effects and promoting
activity.  Toxicol Appl Pharmacol. 1991;ll:454-468.

16.  Chen PH, Wang CK, Rappe C, Nygren M.  Polychlorinated
biphenyls, dibehzofurans and quaterphenyls in toxic,,,rice-bran oil
and in the blood and tissues of patients with PCB poisoning
(Yu-cheng) in Taiwan.  Environ Health Persp. 1985;59:59-65.

17.  Colombi A, Maroni M, Feripli A, et al.  Increase in urinary
porphyrin excretion in workers exposed to polychlorinated
biphenyls. J Appl Toxicol. 1982;2:118-121.

18.  Davis D, Safe S.  Dose response irnmunotoxicities of
commercial polychlorinated biphenyls (PCBs) and their interaction
with 2,3,7,8 TCDD.  Toxicol Letters. 1989;48:35-43.

19.  Demi E, Oesterle D.  Sex-dependant promoting effect of
polychlorinated biphenyls on enzyme-altered islands induced by
diethylnitrosamine in rat liver. Carcinogenesis. 1982;3(12):
1449-1453.             '    .

20.  Exon J, Talcott P, Koller L. Effect of lead, polychlorinated
biphenyls, and cyclophosphamide on rat natural killer cells,
interleukin @, and antibody synthesis.  Fund 'and Appl Toxicol.

21.  Gladen B, Rogan W.  Effect of perinatal polychlorinated
biphenyls and dichlorobiphenyl dichloroethene on later
development.  J Pediatr. 1991;119:58-63.

22.  Gladen B,., Rogan W, Hardy P, Thullen J, Tingelstad J, Tully
M.  Development after exposure to polychlorinated biphenyls and
dichlorodiphenyl dichloroethene transplacentally and through
human milk.  J Pediatr. 1988;113:991-995.

23.  Haake J, Safe S, Mayura-K, Phillips T.  Aroclor 1254 as an
antagonist of the teratogenicity of 2,3,7,8
tetrachlorodibenzo-p-dioxin.  Toxicol Letters. 1987;38:299-306.

24.  Institute for Evaluating Health Risks.  Reassessment of
liver findings in five PCB studies in rats.' (IEHR  Report, July
1991).                               ,     .

25.  Imanishi.                   ..-..'•

26.  Ito N, Nagasaki H, Aral M, Makiura M, Sugihara S, Hirao K.
Histopathologic  studies on liver tumorigenesis induced in mice by
technical polychlorinated biphenyls and its promoting effect on
liver tumors induced by benzene hexachloride. JNCI.

27.  Ito N, Nagasaki H, Makiura S, Arai M.  Histopathological
studies on liver tumorigenesis in rats treated with
polychlorinated  biphenyls. GANN. 1974;65:545-547.

28.  Jacobson J, Fein G, Jacobson S, Schwartz P, Dowler J.  The
transfer of polychlorinated biphenyls (PCBs) and polybrominated
biphenyls across the human placenta and into maternal milk. Am J
Public Health. 1984,-74:378-379.

29.  Jacobson J, Fein G, Schwartz P, Dowler J.  Perinatal
exposure to an environmental toxin: a test of the multiple
effects model.   .Devel Psych. 1984;20:523-532.

30.  Jacobson J, Jacobson S.  Effects of exposure to PCBs and
related compounds on growth and activity in children.   Neurotox
and Teratol. 1990,• 12:319-326.

31.  Jacobson J, Jacobson S, Humphrey H, et al.  Determinants of
polychlorinated  biphenyls (PCBs), polybrominated biphenyls  (PBBs)
and dichlorodiphenyl trichloroethane (DDT) levels in the sera of
young children.  Am J Public Health. 1989/79:1401-1404.

32.  Jacobson JL, Fein GG, Jacobson SW,  Schwartz PM.  Factors and
clusters for the Brazelton scale.  An investigation of the
dimensions of neonatal behavior.  Devel Psycho!. 1984;20:339-353.

33.  Jacobson S, Fein G, Jacobson J, Schwartz P, Dowler J.  The
effect of intrauterine PCB exposure on visual recognition memory.
Child Devel. 1985;56:853-860 .

34.  James MO, Little PJ.  Polyhalogenated biphenyls and
phenobarbital: evaluation as inducers of drug metabolizing
enzymes in the sheepshead, Archosargus probatocephalus.  Chem
Biol Interact 1981;36:229-248.

35.  Jan J, Malnersic S.  Determination of PCB and PCT residues
in fish by tissue acid hydrolysis and destructive cleanup of the
extract.  Bull Environ Contam Toxicol 1978;19:772-780.

36.  Jefferies DJ,  Parslow JLF.  Effect of one polychlorinated
biphenyl on size and activity.of the gull thyroid. Bull Environ
Contam and Toxicol. 1972;8:306-310.

37.  Jefferies DJ,  Parslow JLF.  Thyroid changes in PCB-dosed
guillemots and their indication of one of the mechanisms of
action of these  materials. Environ Pollution. 1976,• 10:293-311.

38.  Jensen AA. In: Halogenated biphenyls, terphenyls,
napthalenes., dibenzodioxins, and related products.  Kimbrough RK,
Jensen AA, eds.  2nd ed.  Amsterdam: Elsevier; 1989;345-375 .

39.  Jensen AA, Slorach S. A ed. Chemical contaminants in human
milk.  Boca Raton: CRC Press, 1991.

40.  Jones J, Alden H.  An acneform dermatergosis.  Arch Dermatol
Syphilol. 1936;33:1022-1034.

41.  Kaattari SL.  Stress and immunity in fish.  Lecture 7.
First Nordic Symposium on Fish Immunology.  Tromso, Norway,

42.  Kalmaz EV, Kalmaz GD.  Transport, distribution and toxic
effects of polychlorinated biphenyls in ecosystems: review.
Ecol Model 1979; 6 .-223-251.
43.  Kammann U, Knickmeyer R, Steinhart H.  Distribution_of
polychlorobiphenyls and hexachlorobenzene in different tissues of
the dab  (Limanda limanda L.) in relation to lip id polarity.  Bull
Environ  Contain Toxicol 1990; 45: 552 -559 .

44.  Kannan N, Tanabe S, Tatsukawa R.  Toxic potential of
non-ortho and mono-ortho coplanar PCBs in commercial PCB
preparations: 2,3,7,8-T4CDD toxicity equivalence factors
approach.  Bull Environ Contam Toxicol 1988a;41:267-276.

45.  Kannan N, Tanabe S, Tatsukawa R.  Potentially hazardous
residue  of non-ortho chlorine substituted coplanar PCBs  in human
adipose  tissue. 'Arch Environ Health 1988b;43:11-14.

46.  'Kerr GR, Scheffler GM, Waisman HA.  Growth arid development
of infant M. Mulatta fed a  standardized diet. Growth.

47.  Kimbrough RD.  Chronic toxicity of halogenated biphenyls and
related  compounds in animals and health effects in humans.  In:
Toxicology of Halogenated Hydrocarbons: Health and Ecological
Effects.  Khan MHQ, Stanton RH  (eds),  New York:Pergammon Press,

48.  'Kimbrough RD.  Consumption of  fish: benefits and perceived
risk.  J Toxicol Environ Health. 1991;33:81-91.

49.  Kimbrough RD.  Human health effects of polychlorinated
biphenyls and polybrominated biphenyls.  Annu Rev Pharmacol.
1987 ,-27:87-111.

50.  Kimbrough RD,  Linder RE.  Induction of adenofibrosis and
hepatomas of the liver  in BALB/cJ mice by polychlorinated
biphenyls  (Aroclor  1254). JNCI. 1974;53:2.

 51.   Kimura NT, Baba T.  Neoplastic changes in the rat liver
 induced by polychlorinated biphenyl. GANN.  1973;64:105-108.

 52.   Kirkyliet NI, Kimeldorf DJ.  Antitumor activity of a
 polychlorinated biphenyl mixture,  aroclor 1254, in rats
 inoculated with Walker 256 carcinosarcoma cells.   JNCI.

 53.   Kunita N, Hori S, Obana H,  et al.   Biological effect of
 PCBs,  PCQs and PCDFs present in the oil causing Yusho and
 Yu-Cheng.  Environ Health Perspect. 1985;59:79-84.

 54.  _Kuratsune M,  Yoshimura T,  Matsuzaka J,  Yamaguchi A.   Yusho,
 a poisoning caused by rice oil  contaminated with polychlorinated
 biphenyls.  HSMHA Health Reports.  1971;86:1083-1091.

 55.   Kuratsune M.   Yusho with reference to  Yu-Cheng.  In:
 Halogenated biphenyls, terphenyls, napthalenes, dibenzodioxins
 and  related products.   Kimbrough RD,  Jensen AA, eds.   2nd ed.
 Amsterdam; 1989:381-396.

 56.   Levin DE, Schantz SL,  Bowman  ER.   Delayed spatial
 alternation deficits resulting  from perinatal  PCB  exposure in
 monkeys .    Arch Toxi col. 1988 ,-62:267-273.

 57.   Lilienthal H,  Neuf M,  Munoz C,  Winneke  G.  Behavioral
 effects of pre- and postnatal exposure  to a  mixture of low
 chlorinated PCBs in rats.   Fund and Applied Toxicol.

 58.  Lilienthal H,  Winneke  G.   Sensitive period for behavioral
 toxicity of polychlorinated biphenyls:  determination  by cross
 fostering  in rats.   Fund and Applied Toxicol.  1991,-17:368-375.

 59.  Loose LD,  Silkworth JB,  Pittman KA,  Benttz KF, Mueller W.
 Impaired host  resistance to endotoxin and malaria  in
polychlorinated biphenyl- and hexachlorobenzene-treated mice.
 Inf Immun.  1987b;20(1):30-35.

 60.  Marks  TA,  Kimmel  GL, Staples  RE.   Influence of symmetrical
biphenyl isomers on  embryo  and fetal development in mice.
 Toxicol Appl Pharmacol.  1981;61:269-276.

61.  Marks  TA,  Kimmel  GL, Staples  RE.   Influence of symmetrical
biphenyl isomers on  embryo  and fetal development in mice,  II.
Comparison  of  4,4'Dichlorobiphenyl, 3,3',4,4'tetrachlorobiphenyl,
3,3',5,5' tetrachlorobiphenyl and  3,3',4,4'  tetramethyl biphenyl.
Fund Appl  Toxicol. 1989;13:681-693.

62.  Marks  TA,  Staples RE.  Teratogenic evaluation of the
symmetrical isomers of hexachlorobi-phenyl (HCB) in the mouse
Toxicol Appl Pharmacol.  1980;52:A66.

63.  Masuda Y.  Health status of Japanese and Taiwanese after
exposure to contaminated rice oil.  Environ Health Persp.

64.  McKinney JD, Chae K, Gupta BM, Moore JA, Goldstein JA.
Toxicological assessment of hexachlorobiphenyl isomers and
2,3,7,8-tetra-chlorodibenzofuran in chicks.  1. Relationship of
chemical parameters. , Toxicol Appl Pharmacol. 1976; 36: 65-80 .

65.  Meigs J, Albom J, Kartin B.  Chloracne from an unusual
exposure to aroclor.  JAMA. 1954;154:1417-1418.

66.  Miller HJ, Cucos S, Wassermann D, Wassermann M.
Organochlorine insecticides and polychlorinated biphenyls in
human milk. Toxicol and. Occupational Med. Deichmann WB, ed.
Elsevier/North-Holland, Amsterdam; 1979:359.

67.  Paneth N.  Human reproduction after eating PCB-contaminated
fish.  Health and Environ Digest. 1991;5:4-6.

68.  Rogan W, Gladen B, McKinney J, et al.  Neonatal effects of
transplacental exposure to PCBs and DDE.  J Pediatr.
1986; 109 :335-.341.

69.  Rogan W, Gladen B, McKinney J, et al.  Polychlorinated
biphenyls and dichlorodiphenyl dichloroethene  (DDE) in human
milk: effects of maternal factors and previous lactation.  Am J
Public Health. 1986;76:172-177.

70.  Rogan W, Gladen B, McKinney J, et al.  Polychlorinated
biphenyls  (PCBs) and dichlorodiphenyl' dichloroethene  (DDE) in
human milk: effects on growth, morbidity and duration of
lactation.  Am J Public Health. 1987;77:1294-1297.

71.  Rogan W, Gladen B, Wilcox A.  Potential reproductive and
postnatal morbidity from exposure  to polychlorinated biphenyls:
epidemiologic considerations. Environ Health Persp.

72.   Schwartz P, Jacobson S,  Fein  G/ Jacobson  J,  Price H.  Lake
Michigan fish consumption as  a  source of polychlorinated
biphenyls  in, human  cord  serum, maternal  serum, and milk.  Am J
Public Health.  1983;73:293-296.

73.   Seegal R,  Bush B, Brosch K.   Sub-chronic  exposure of  the
adult rat  to  Aroclor  1254 yields  regionally-specific  changes in
central  dopaminergic  function.  Neurotox.  1991;12:55-66.

74.   Seegal R,  Bush B,  Shain  W.   Light  chlorinated
ortho-substituted PCB  congeners decrease dopamine in  nonhuman
primate  brain and tissue culture.  Toxicol  and Appl Pharmacol.

 75.   Silberhorn EM, Glauert HP, Robertson LW.  Carcinogenicity of
 polyhalogenated biphenyls: PCBs and PBBs.  Critical Reviews in
 Toxicol. 1990;20:6:440-496.

 76.   Sinks T, Smith AB, Steele G,  Rinsky R,  Watkins K.   A
 retrospective cohort mortality study of workers at an electric
 capacitor plant utilizing polychlorinated biphenyls.
 Occupational epidem. Sakuri H, Okazaki I,  Omae K.  eds.
 Proceedings of the 7th International Symposium on Epidemiology in
 Occupational Health.  Tokyo,  Japan 11 - 13 Oct. 1989 pp 113 -
 119.   Excerta Medica, Amsterdam 1990 and NIOSH, Health Hazard
 Evaluation Report HETA 89-116-2094.

 77.   Smialowicz R,  Andrews J,  Riddle M,  Rogers R,  Leubke R,
 Copeland C.  Evaluation of the immunotoxicity of low-level PCB
 exposure in the rat.  Toxicology.  1989;56:197-211.

 78.   Talcott PA,  Koller LD,  Exon JH.  The  effect of lead and
 polychlorinated biphenyl exposure  on rat natural killer cell
 toxicity.  Int J Iirtmunopharmac. 1985;7 :2 :255-261.

 79.   Taylor PR,  Lawrence CE,  Hwang HL,  Paulson LS.
 Polychlorinated biphenyls:  Influence on birthweight and
 gestation.  Am J Public Health.  1984;74:1153-1154.

 80.   Taylor PR,  Stelman JM,  Lawrence CE.   The relation  of
 polychlorinated biphenyls to  birth weight  and gestational age in
 the offspring of occupationally exposed  mothers.  Am J Epidemiol.

 81.   Thomas P,  Hinsdill R. Effect of polychlorinated biphenyls
 on the  immune responses of rhesus  monkeys  and mice.   Toxicol Appl
 Pharmacol.  1978,-44:41-51.

 82.   Tilson H,  Jacobson J,  Rogan W.   Polychlorinated biphenyls
 and the developing  nervous system: cross-species comparisons.
 Neurotox and Teratol.  1990,• 12:239-248.

 83.   Truelove J,  Tanner J,  Langlois  R, Arnold D, Mes  J.   Effect
 of polychlorinated  biphenyls on several  endocrine reproductive
 parameters  in the female rhesus monkey.  Arch Environ Contam and
 Toxicol. 1990;19:939-943.

 84.   Tryphonas H, Hayward S, O'Grady L,  et al.   Immunotoxicity
 studies  of  PCB  (Aroclor 1254)  in the adult rhesus  (Macaca
mulatta) monkey - Preliminary  Report.  Int J Immunopharmac.

 85.   Tryphonas H, Luster ML, Schiffman G, et al.  Effect  of
 chronic  exposure of  PCB (Aroclor 1254) on specific and non
 specific immune parameters in  the  rhesus (Macaca mulatta) monkey.
Fund Appl Toxicol.  1991/16:773-786.

86.  Tryphonas H, Luster, ML, Whitet KL, et al.  Effects of PCS
(Aroclor 1254) on non-specific immune parameters in rhesus
(Macaca mulatta) monkeys. Int J. Immunopharm.  1991;13:639-648.

87.  Van Wagenen G, Catchpole HR.  Physical .growth of the rhesus
monkey.  Am J Phys Anthropol. 1956;14:245-273.
88.  Warshaw R, Fischbein A, Thornton J, Miller A, Selikoff IJ.
Decrease in vital capacity in PCB-exposed workers in a capacitor
manufacturing facility.  NY Acad of Sci. 1979;320:277-283.

89.  Zeiger E. Genetic toxicity. In: Halogenated Biphenyls,
Terphenyls, Naphthalenes, Dibenzodioxins and Related Products.
Kimbrough RD, Jensen AA, eds.  2nd ed. Amsterdam: Elsevier;

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposure to PCBs
                    Research Triangle Park, NC
                       September 14-15,  1992

                        Richard E.  Peterson
                        School of Pharmacy
                      University of Wisconsin
                     425 North Charter Street
                        Madison, WI 53706

 General Toxicity of  PCBs

 Pharmacokinetics of  PCBs

 PCB Structure Activity Relationships and Developmental Toxicity


     1.  Goal #1 is unrelated to  the neurotoxic  focus of  this
 workshop.   It deals with  SAR for PCBs and  related dioxins on
 developmental toxicity  in general.  It  is  recommended that Goal
 #1  be  dropped.

     2.  Goal #2 is fine with respect to  the focus on
 neurotoxicity,  but is vague as to  kinds  of  PCB  and PCDD  congeners
 to  be  tested,  types of developmental neurotoxic responses to be
 assessed,  and strain and  species of animals to  be used.
     3. Goal  #3  suffers  from the  same weakness as Goal #1.
too broad and needs  to be  focussed on neurotoxicity.

It is
     The data base on developmental neurotoxicity of the PCBs is
only beginning to emerge and there is a general lack of
understanding on mechanisms of the various effects.

     The Issue section focuses on potential mechanisms of PCB
developmental neurotoxicity and is linked only weakly to the SAR
goals of the proposal.  Furthermore, the hypothesized mechanisms
of PCB developmental neurotoxicity are not linked to specific
responses which is problematic.  Because this critical linkage is
missing it is difficult to comment on the validity of each
hypothesized mechanism because one does not know which of the
various developmental neurotoxic responses reported in the
literature each mechanism pertains to.

     It_is not clear whether the SAR goals of the proposal will
be met if mechanisms of the various neurotoxic effects fail to be
elucidated_in a timely fashion.  In lieu of this possibility,
consideration might be given to basing interim SARs for PCB

 congeners  on sensitive  developmental neurotoxic responses  in  a
 prototype  rodent  species.

      A potential  site of  action  of perinatal  PCB exposure  that
 was  not mentioned is  the  cholinergic nervous  system.   It
 undergoes  rapid development  during perinatal  brain  growth  and is
 involved in such  functions as memory and  learning which have  been
 reported to be affected by perinatal PCB  exposure.  Furthermore,'
 Eriksson (1988) and Eriksson et  al.:. (1991)  found that  exposure  of
'neonatal NMRI mice to 3,3', 4,4'-TCB altered cholinergic
 muscarinic receptor number in the hippbcampal area  of  the  brain.

 Deliberations and Outputs

      1. SAR studies are needed to fully understand  PCB-induced
 developmental neurotoxicity  in laboratory animals and  humans.
 However, there is not yet a  sufficient data base on developmental
 neurotoxic effects of a series of individual  PCB congeners at the
 whole animal level to make a SAR analysis, of  existing  data

      Effects on cellular  dopamine content of  exposing  PC12 cells
 to graded  concentrations  of  individual PCB congeners has been
 investigated by Seegal  and coworkers  (Shain et al., 1991). Their
 findings provide  considerable new insight into the  SAR for this
 particular effect and reveal that ortho-  or ortho-,para-chlorine-
 substituted congeners are the most potent in  reducing!cellular
 dopamine content  with non-ortho-substituted,  dioxin-like
 congeners  being the least potent.  However, a cause and effect
 linkage at the whole  animal  level between effects of ortho- or
 ortho-,para-chlorine-substituted congeners on (1) reductions  in
 dopamine content  in certain  brain nuclei  and  (2) behavioral
 changes in the same animal have  yet to be demonstrated.
 Furthermore, a finding  which complicates  our  understanding of the
 SAR  for PCB effects on  dopamine  at the whole  animal level  is  that
 in utero exposure of  CD-I mice to 3,3',4,4'-TCB  (a  dioxin-like
 PCB  congener which was  not active in reducing dopamine content  in
 PC12 cells) nevertheless  decreased both striatal dopamine  levels
 and  dopamine receptor binding and produced a  neurological
 syndrome consisting of  elevated  levels of motor activity and
 spinning behavior- (Agrawal et al.,  1981).

      2. Based on  information presented in the Issue section,  the
 high degree of enthusiasm for. a  thyroid hormone mechanism  of  PCB
 developmental toxicity  seems unfounded.   Background information
 on PCBs binding to prealbumin, PCB  exposure decreasing
 circulating thyroid hormone  levels, and thyroid hormones being  of
 critical importance for normal brain development was given.  Yet
 specific developmental  neurotoxic effects of  the PCBs  that have
 been reported in  the  literature  and are considered  to  be due  to
 an alteration in  thyroid  status  were not  described.  Also  the
 profile of biochemical  effects caused by  altered thyroid status
 in the brains of  developing  laboratory animals '(Ford and Cramer,
 1977) have, to this reviewers knowledge,  not  yet been  shown to

 occur following perinatal PCB exposure.   In the absence of such
 information the hypothesized thyroid hormone mechanism of PCB
 developmental neurotoxicity seems no more likely than any of  the
 other hypothesized mechanisms.   Why is it receiving so much
 attention?  Is unpublished information available that could not
 be presented in the report?  What are the thyromimetic properties
 of the PCBs?  What developmental neurotoxic responses of PCBs in
 laboratory mammals mimic those caused by thyroid hormones?

      3.  Since the focus of this workshop is developmental
 neurotoxicity it is assumed that non-Ah receptor,  PCB-specific
 binding proteins will be searched for in brains of laboratory

 Developmental Neurotoxicity of PCBs in Humans

 PCB Developmental Neurotoxicity in Experimental Animals


      1.   Goal #1 is the most important of the three goals.  It
 will evaluate the available data base on PCB developmental
 neurotoxicity,  identify information gaps,  and recommend future
 research to eliminate such gaps.

      2.   Goal #2 is to evaluate the role of structure activity
 relationships in PCB developmental neurotoxicity.   It is
 recommended that Goal #2 be dropped from this proposal  because it
 is covered in the McKinney proposal.             •   .  "

      3.   Goal #3 is to determine if maternal  factors  play a
 significant role in PCB developmental neurotoxicity.   It  is an
 important goal  but should be assigned a  lower priority  than Goal


      A major issue is how to evaluate the existing  data base  for
 information gaps.   It is recommended that  this  be done  in a
 species-specific manner.   For each species  developmental
 neurotoxic effects should be separated into categories based  on
 (1)  PCB congener or mixture,  (2)  animal  strain,  (3) time  of
 exposure  during development,  and  (4)  dose  level.  This approach
 should reveal information gaps  and may rectify  apparent
 discordance in  the data base of  each species.

      An issue which was  not  mentioned, but  may prove useful in
 comparing the profile of  developmental neurotoxic responses
 between species,  is  to provide  information  on the critical
 periods of  CNS  development  for  the  expression of certain
 responses  in  various  species.   For  example, sexual
 differentiation of the CNS  as reflected by  the ability to express
 feminine  sexual  behavior  in  adulthood  is imprinted earlier during
perinatal  development  in  the monkey  than the rat.

     An issue which was alluded to is whether alterations in
sexual behavior that have been reported to occur in male rats
after perinatal exposure to TCDD (Mably et al.,  1992)  will also
occur after perinatal exposure to PCB congeners  that are Ah
receptor agonists.  At the present time there is no information
on whether perinatal exposure to coplanar PCBs alters sexual
differentiation of the CNS.  This is an important information gap
which must be filled to fully comprehend the developmental
neurotoxic effects of the PCBs.

     The issue of maternal factors playing a significant role in
the developmental neurotoxicity of the PCBs was not discussed.
Is this because there is no data on maternal effects of PCB
exposure during pregnancy and lactation?  If so this information
gap should be so identified.

     Also among the maternal effects possible an issue that
should be considered is which types of effects are the most
likely cause of the developmental neurotoxicity observed.  These
types of effects should then be assigned the highest priority for
future research.

     An issue which needs to be deliberated is the definition of
developmental neurotoxicity.  Is a change in the concentration of
a neurotransmitter or an alteration in the number of steroid
hormone receptors in a certain brain region considered evidence
of developmental neurotoxicity?  I would argue that such effects,
in isolation, are not neurotoxic that they must be linked to
quantitative changes in behavior before they can be considered as

Deliberations and Outputs

     1.  Alterations in striatal dopaminergic and hippocampal
cholinergic  systems were reported in mice exposed prenatally to
3,3',4,4'-TCB.  Both effects were associated with motor
alterations.                                                ,

     To my knowledge PCB-induced alterations in neural growth
factors at the  perinatal stage of development have not been

     Other questions which should be deliberated at this  time are
whether perinatal  exposure to  PCBs  affects plasma hormone
concentrations, hormone receptor  levels  in the brain, or
responsiveness  of  the  brain to hormones  during early development?
If perinatal exposure  to PCBs  is  found  in the future to  affect
sexual differentiation of  the  CNS such  information  could provide
insight  into the  mechanism.

      2.   All of the questions  on  PCBs with TCDD-like properties
assume that  these coplanar PCB congeners will alter sexual
differentiation of the CNS.  However, alterations  in sexual
behavior  of  adult male rats exposed to  a single maternal  dose  of

 TCDD on day 15 of gestation is a very recent discovery and no 'PCB
 congener or PCB mixture has yet been tested and shown to produce
 this effect.  Also it has yet to be determined whether the size
 of the sexually dimorphic nucleus of the rat brain is affected by
 perinatal exposure to any Ah receptor agonist including TCDD
 Furthermore, it is not known if effects of perinatal exposure to
 TCDD_on CNS sexual differentiation in male rats occurs in other
 species.  Thus, the main point is that the data base on altered
 CNS sexual differentiation due to perinatal PCB exposure is
 nonexistent, but in view of the recent findings with TCDD (Mably
 et al., 1992)  it merits future investigation.

      3. In male rats exposed to a single maternal dose of TCDD (1
 M-g/kg,  po)  on day 15 of gestation there is a statistically
 significant delay in the onset of puberty (Bjerke and Peterson,
 unpublished results).  However,  it is not known if PCB congeners
 that are Ah receptor agonists produce a similar effect.   Future
 consideration might be :given to assessing the onset of puberty in
 male and female laboratory animals exposed perinatally to PCBs.
 It_may  show that the onset of puberty is also delayed in these
 animals and that it is a PCB-sensitive developmental landmark.

      4. The problem of testing a large number of PCB congeners in
 order to evaluate the developmental neurotoxicity of the PCBs is
 difficult because of the time,  expense,  and large numbers of
 animals required.   If mammals  are used the research must be
 focussed to reduce cost.   In this context,  consensus might be
 sought  on an experimental design that  will permit dose-related
 effects of  a PCB mixture to be determined for a  variety  of
 developmental  neurotoxic endpoints  in  a prototype species
 possibly the rat.   The design  of the first study might be to
 determine dose-related neurotoxic effects in  offspring who are
 (1)  born to parents  with steady  state  body burdens of the PCB
 mixture and (2)  continuously exposed to the PCB  mixture  from
 conception  through completion  of  the study.   The parameters to
 measure could  be  determined from the existing data base  on
 neurotoxic  effects associated with perinatal  exposure of  rats to
 PCBs or TCDD.  Analysis  of  all dose-related effects  upon
 completion  of  the  study  should reveal which ones  are  the  most
 sensitive to PCB mixture  exposure.   Once  this  "developmental
 neurotoxicity profile" is obtained subsequent  studies in  the
 prototype species  could'focus on  determining,  for  the most highly
 sensitive effects,  (1) critical periods of  exposure,  (2)  SARs,
 and  (3)  mechanisms involved.

     Obviously other experimental designs and approaches  are
possible.  The issue is not whether the one above  is best.
Rather the issue is whether the existing data base for any one
 species on 'PCB developmental neurotoxicity is sufficiently strong
to clearly define dose response relationships for  the various
developmental neurotoxic endpoints.  If such dose  response
information is not available it would be prudent to obtain it for
at least one prototype species using an experimental design that

is the general consensus of workshop participants.  This
descriptive dose response information can then be used to focus
subsequent SAR and mechanistic investigations on those responses
that are the most sensitive and potentially relevant to human

     Strategies for dealing with the problem of predicting the
developmental neurotoxicity of PCB congeners that act by
different mechanisms is difficult.  For those developmental
neurotoxic responses that are discovered in the future to occur
through an Ah receptor mechanism the TCDD toxic equivalents
approach would seem appropriate.  For responses which are not Ah
receptor mediated other strategies 'based on SARs for the
responses and in depth knowledge of their mechanisms might be
considered.  The suggestion to consider a thyroxine equivalents
approach is not appealing because of the paucity of evidence
relating alterations in circulating thyroid hormone levels or
cellular thyroid status to changes in the behavior of laboratory-
animals perinatally exposed to PCBs.  If future research
establishes such a cause and effect linkage this approach may be
warranted but not at the present time.

     5. The question of whether toxic potency of a PCB congener
changes if exposure to it occurs in a complex mixture format is
probably not known for developmental neurotoxicity endpoints.
However, it seems at least potentially possible for a mixture of
congeners to influence biotransformation rate and/or tissue
distribution of another PCB congener.  The question is whether
such effects would ever occur at the low concentrations of
complex PCB mixtures found in exposed populations.

     6. If certain hydroxylated PCB metabolites have
antiestrogenic activity  (see issue paper by McKinney) they have
the potential to alter sexual differentiation of the CNS in the
rat.  However to do so they would have to accumulate to
sufficiently high concentrations in the brain to elicit an
antiestrogenic response.

                      Premeeting Comments  for
           Workshop on Developmental Neurotoxic Effects
                  Associated with Exposure to PCBs
                     Research Triangle  Park,  NC
                       September 14-15, 1992

                          Mary M.  Prince
                      Risk Assessment Program
                 Division of Standards Development
                            and Technology
                National  Institute of  Occupational
                         Safety and Health
                   4676 Columbia  Parkway  (C-15)
                       Cincinnati, OH  45226

 General Toxicity of PCBs

      In evaluating PCB developmental toxicity,  a systematic
 review of the dose levels associated with other health outcomes
 is  important for comparing mechanisms  of  toxicity,  as  well as
 issues related to tissue  distribution, metabolism,  and cogener-
 specific toxicity.  This  paper briefly reviews  the  PCB-related
 effects on non-developmental neurotoxicity,  immunotoxicity,
 endocrine function,  and carcinogenicity.   A comparison of  the
 dose  levels associated with these various outcomes  could be more
 closely delineated in tabular form.

      The_section  on  endocrine function summarizes the  results of
 two studies (Miller  et  al,  1978;  Truelove et al, 1990)   It  is
 not clear why the Tryphonas et al (1991)  study  was  cited in  this
 section.   Unless  there  is some mechanism  common to  PCB effects on
 immunotoxicity and endocrine function,  the last sentence does not
 appear to add anything  to the discussion  of  these data.

      The two studies  investigating the effect of PCBs  on
 endocrine function are  inconsistent  with  respect to their
 conclusions  regarding the association  between PCB exposure and
 endocrine function.   In situations where  studies show  disparate
 results,  a comparison and evaluation of the  study design,
 exposure  regiment, and analyses of the studies  would be  valuable
 for weighing the  scientific  evidence for  a PCB-related effect  on
 endocrine function.   The  interpretation of these disparate
 results  should be  discussed  further  during the  workshop.

     It would  also be important to discuss whether there are
overlapping mechanisms of toxicity common to different endpoints
and whether different congeners or PCB mixtures are related to
either different organ system  effects and/or to different
functional domains of the neurologic system.

PCB Developmental Neurotoxicity in Experimental Animals

     A major concern addressed in this paper is the mechanism(s)
by which a particular congener or a complex mixture of PCBs
affects nervous system development.  In terms of assessing
strength of evidence, it would be important to first assess which
PCB congeners are likely to affect developmental neurotoxicity
(based on chemical and pharmocokinetic considerations).  For
instance, it was stated that there is confusion regarding whether
dibenzofurans contribute to the neurotoxicity of PCB mixtures.
This suggests that one must first characterize the exposure more
precisely in order to relate relevant outcome measures to
exposure.  The strength of evidence is only as good as our
understanding of PCB exposure characteristics and toxicity.  This
particular data gap should be a primary question of concern.

     It is also important to determine whether the animal studies
that examine PCB developmental toxicity are relevant to human
exposures in terms of the PCB mixtures used in these experiments
and congener-specific toxicity.  This should be carefully
evaluated before drawing conclusions about PCB developmental
neurotoxicity in humans from animal experiments.

     Once these questions are addressed, the deliberations can
then concentrate on the issues regarding mechanism of
neurobehavioral toxicity, pharmacokinetics, congener-specific
patterns of toxicity, and inter-species homology of
neurobehavioral changes.

                     Pre-meeting Comments for
           Workshop on Developmental Neurotoxic Effects
                 Associated with Exposures to PCBs
                    Research Triangle Park, NC
                       September 14-15, 1992

                           Walter  Rogan
                       National Institute of
                   Environmental Health Sciences
                      P.O. Box  12233  (A3-05)
                 Research Triangle Park,  NC  27709
 Developmental Neurotoxicity in Humans

      These comments are arranged by the set of questions.
 However,  I think that dismissing the Yu-Cheng children out of
 hand because they are too complicated is a mistake.


      1.  Is there incomplete control of confounding,  especially in
 MI  data?   There is no doubt residual confounding,  but  I at least
 cannot  come up with a plausible confounder that should give the
 results  seen.   The usual suspects have been,  for the most  part,
 addressed.   Selection and selective drop out  is a more likely
 possibility,  but this is very difficult to assess; the other
 inherent  problem with anyone evaluated as much and as  often as
 these kids is multiple comparisons.

      2. Are there differences in exposures between MI  and  NC?  To
 the degree that the body burden in the women  in the  MI study come
 directly  from the fish they ate,  and to the degree that whatever
 the pattern is,  is lost as  those fish contribute to  general food
 contamination,  there is a difference in pattern.   I  think  it-
 would be  at least reflected in the highest chlorinated congeners,
 since they accumulate best,  and are largely what the total PCBs
 measurement measures.

      3. How different are the analytic methods?  Jim McKinney is
 apparently  available to this  effort,  and he developed  the
 analytic  methods  for NC.  Basically,  NC quantitated  on 2 peaks,
MI  on 6.  We asked Mary Wolf  at  Mt.  Sinai  New  York to  look at
 some  of our chromatograms once,  and  see how they would be  read
using Webb-McCall.   Although  strictly you  can't  do this, the
numbers come out  fairly stably at  one half; i.e.,  for  a given
chromatogram, the McKinney method  gives  a  value  twice what Webb-
McCall gives.  This means that NC  is  seeing motor impairment  at a
lower level than  the papers say  (if Webb-McCall  is "true").

     4. Could different congeners produce different results?
This is the same as question 2.   It is certainly possible that
the congener responsibility for the memory finding at 4 years is
something in the fish, although for some reason this lacks
plausibility to me as an explanation.

     5. Could the difference in estimating prenatal exposure
produce other differences?  Putting aside issues of noise near
the limit of detection, the levels measured in different matrices
should be linear transformations of each-other.  The NC method
smooths the set of data available'on any woman; the MI method
uses one determination, in a matrix that does not have much PCB
in it.  My suspicion is that the exposure estimation in MI is
less stable than NC; the neonatal findings might be more
congruent if the data were analyzed using more data from each

     6. Why are different domains affected?  If we assume that
the Brazelton findings are congruent, then'the big difference is
the four year memory finding.  I have never done the power
calculations, but I suspect that MI could not detect a 5 point
motor scale effect with only 200 or so kids.  NC did not do
Fagin's test, and assessed only reported "hyperactivity" not

     7. Do the low values in cord blood affect the validity of
the study?  I don't see how they could.  Suppose 'that NC did not
do cord bloods at all — would that affect validity?  MI uses
cord serum values as an exposure estimate, but they could do this
with other numbers that they have on women.  The censoring should
affect power, but censoring per se should not affect validity.

     8. What is the big picture?  We've tried to answer this in
print several times.  'There is no question that PCB and compounds
like them are toxic to the developing nervous system.  There is
ample laboratory evidence for this.  There'is  (as far as I know)
universally accepted data from the high dose dirty episodes in
Asia that humans are not immune to this class of chemicals.  The
fact that there is even a-strong suspicion that something
detectable is going on in the general population using relatively
crude measures of exposure and incredibly blunt measures of brain
function is scary.

General Toxicity of PGBs

     1. How do the levels for developmental neurotoxicity compare
with the cancer numbers?  We've done these calculations crudely.
If you accept the NC data as published, and you use very non-
conservative assumptions, you can get within an order of
magnitude of the cancer "safe" level.

     2. What about immunotoxicity?- Absent wiping out a certain
stem cell that was responsible for giving rise to a particular
clone that would protect you from something lethal, there must be

thresholds for measurable changes in immune function.  On the
other hand, the developing thymus is very sensitive to these
agents, and the neonatal or fetal thymus may have committed cells
that, if wiped out, lead to long lasting impairment.

     3. Do the same congener's that produce adult neurotoxicity
also produce developmental neurotoxicity?  Since we don't know
the mechanism of the neurotoxicity, we can't answer this.  The
fact that certain congeners concentrate in the brain may be
completely irrelevant.  The motor findings in the kids are as
consistent with thyroid impairment or even anterior horn cell
dysfunction as they are with a central mechanism.

                    Pre-meeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with Exposure to PCBs
                   Research Triangle Park, NC
                      September 14-15, 1992

                         Susan Zi.  Schantz*
                      University of Illinois
               Institute for Environmental Studies
                      1101 W. Peabody Drive
                        Urbana, IL  61801

General Toxicity of PCBs (Carole Braverman)

No pre-meeting comments.

Pharmacokinetics of PCBs (Linda Birnbaum)

No pre-meeting comments.

PCB Structure Activity Relationships and Developmental Toxicity
(James McKinney)

     My comments on the role of altered thyroid function  in
mediating CNS effects  (see Diane Miller, below) also  relate to
this issue.

Developmental Neurotoxicity of PCBs in Humans  (Mary Prince)

     Most of the issues raised by Dr. Prince can be better
addressed by other workshop participants.  However, I would like
to make a comment with regard to the issue of  comparability of
animal and human data.  I believe it is crucial that  '
investigators conducting animal and human studies  (myself
included) begin doing a better job of selecting behavioral tests
that facilitate extrapolation from animals to  humans  and  vice
versa.  We also need to make better use of existing data  (animal
and human) in our selection of behavioral tests.

     Better communication may be all that is necessary to
accomplish this goal.  Here are a couple of examples  to
illustrate my point.  Monkeys exposed to PCBs  perinatally show
deficits in discrimination-reversal learning.  Discrimination-
reversal learning is a cognitive task that has been extensively
studied in terms of the brain areas and functions  involved.
Human tests that measure behavioral functions  similar to  those
assessed in discrimination-reversal learning exist and their
similarities to discrimination-reversal learning have been
discussed in the literature.  These include the Wisconsin Card
Sorting Test and the Stroop Test.  Both tests  are  easy to
administer and have been used extensively in clinical
neuropsychology.  They could readily be adapted for use in
ongoing epidemiological studies of PCB exposure.   Discriminatipn-

*See also Bowman and Schantz Supplemental Pre-meeting Comments immediately following
Bowman Pre-meeting Comments.

 reversal  learning can also be done  in  rodents, which would  allow
 for  cross species comparisons for consistency of  effects.

      From the opposite point-of-view,  the Fagan Visual
 Recognition Test  appears  to be a  sensitive  indicator of  PCB
 effects in human  infants.   Excellent animal' models of this  task
 are  available and could be used to  document whether the  same
 deficits  are present  in controlled  animal experiments.   If  we can
 document  similar  deficits  on readily comparable tasks in animals
 and  humans the job of risk assessment  will  become much easier.

 PCB  Developmental Neurotoxicity in  Experimental Animals  (Diane

      Dr.  Miller has raised a number of very important issues that
 I believe highlight the need for  further research on the
 neurotoxicity of  PCBs.  One important  issue is sensitive periods
 for  exposure.   Several investigators including the Jacobsons and
 Lilienthal and Winneke have published  data  which  suggest that
 cognitive deficits are mediated by  prenatal PCB exposure.
 Postnatal lactational exposure to PCBs appears to be unrelated to
 cognitive outcome in  both  the human and animal studies.  However,
 we must remember  that the  data set  upon which these conclusions
 are  based is  very incomplete.

      Brain structures develop at  different  times  and at  different
 rates.  Some  areas are relatively mature at birth, whereas  others
 undergo considerable  development  after birth.  The cognitive
 tests used in the above mentioned studies were fairly simple and
 global in nature.'_  Tests specifically  designed to assess the
 functional integrity  of late developing brain structures were not
 used.  For example, one area of the brain that does not  become
 functionally  mature'until  well  after birth  is the dorsolateral
 prefrontal cortex.  In a series of  elegant  studies, Patricia
 Goldman-Rakic and her colleagues  have  shown that  this brain area
 is critically 'important in regulating  short-term
 (representational)  memory  for spatial  information and have  shown
 that  certain  behavioral tasks  such  as  spatial delayed response
 and  delayed spatial alternation can be used to detect damage to
 this area.

     We have  reported that monkeys  exposed  to PCB mixtures
perinatally are profoundly impaired on delayed spatial
 alternation.   Studies  are  currently underway in our lab  to
 determine whether rats  perinatally  exposed  to various,
 structurally  diverse  PCB congeners  are also impaired on  thi:s
 task.  Before  we  conclude  that lactational  PCB exposure  is
without effect, we  need to do additional fostering studies in
 rodents to document that delayed  spatial alternation and other
 specific  cognitive  behaviors regulated by late developing brain
areas (e.g. prefrontal  cortex, hippocampus,  cerebellum)  are not
 impaired  by postnatal  PCB  exposure.   In addition,  we need to
 employ specific cognitive  tasks (e.g.   Piaget's AB task)  that will
address the same  issue  in human infants and children.

     We also need to give some thought to how we can better model
human pre- and postnatal, exposure in rodent species.   At birth
the rat brain is roughly'equivalent developmentally to the human
brain at the end of the second trimester.  The first postnatal
week is roughly equivalent to the third trimester in human
infants.  Thus, PCB-exposed rats that are fostered at birth are
not exposed for the full period of brain development covered by  .
human prenatal exposure.  Conversely, PCB-exposed rats that are
not fostered receive greater exposure during the latter one third
of this period because considerably more PCS is transferred via
lactation.  Workshop participants should consider how these
problems influence/hinder interpretation of. already published
studies and discuss experimental procedures that could be used to
deal with these problems in future studies.

     Another very important issue raised by Dr. Miller is the
role that endocrine effects of PCBs may play in mediating CMS
effects.  This is a potentially very important area of research
that has received very little emphasis.  Recent studies by
Peterson and colleagues have demonstrated that perinatal exposure
to 2,3,7,8-TCDD alters the sexual behavior of male rats in
adulthood.  The rats are both feminized and demasculinized and
the effects appear to be the result of an altered sexual
differentiation of the brain.  Dioxin-like PCB'congeners have not
yet been evaluated.for similar effects and the morphology of
sexually dimorphic brain areas has not been evaluated in the
TCDD-exposed rats.  Also,  very little work has been done to
document whether changes in other sexually dimorphic behaviors -
such as activity level and spatial learning strategies also
exist.  A preliminary study we conducted in .collaboration with
Peterson et al. suggested that some of these other sexually
dimorphic behaviors may also be altered in TCDD-exposed rats.
Our study also suggested that female rats may be affected,
highlighting the need to evaluate both sexes in future studies.

     It has been known for some time that exposure to PCB
mixtures affects thyroid function.  PCBs decrease serum thyrpxine
 (T4) levels, leaving triiodothyronine  (T3) unchanged, except at
very high levels of exposure.  A number of complex and
interacting mechanisms appear to account for these effects.  PCBs
inhibit serum  transport of T4 by competing with it for binding to
the carrier protein, they increase biliary excretion of T4 by
inducing hepatic UDP-GT and they appear to act directly on the
thyroid gland  causing a decrease in the colloid droplet-lysosome
interaction necessary for secretion of thyroid hormones.  Very
recent studies in our laboratory have shown that perinatal
exposure to certain individual PCB congeners that are prevalent
in human tissue dramatically reduces serum T4 levels in the pups.
One congener,  245,3'4'-pentachlorobiphenyl  (PCB 118) reduced the
serum T4 levels to below the detection limit of the assay.  This
very profound  'suppression of thyroid hormone levels occurred at a
dose that did  not reduce litter size or pup survival, did not
produce any signs of maternal toxicity and did not appear to
cause hypothyroidism in the dams.             .  •     . .  .

      Another congener,  245,2'4'5'-hexachlorobiphenyl  (PCB  153)
 caused a  significant, but  less  dramatic  decrease  in serum  T4.  A
 dose^effect  relationship was  observed  for both  congeners.  These
 findings  could be of great significance  because thyroid hormones
 are_critical for normal brain development and hypothyroidism
 during the perinatal period is  known to  cause permanent cognitive
 deficits  in  animals  and in humans  (cretinism).  They  are also
 important because the congeners studied, particularly PCB  153,
 are  prevalent in human  tissue.  Studies  are currently underway in
 our  lab to determine whether  these  congeners cause cognitive
 deficits  in  the pups.   The role.that altered thyroid  function may
 play in mediating cognitive effects is unclear  at this point, but
 will be the  focus of future studies in our lab.  We hope to look
 at various neurochemical and morphological endpoints  (e.g.
 choline acetyl transferase activity; dendritic  branching)  to
 determine if CNS changes characteristic  of perinatal
 hypothyroidism are present in PCB-exposed rats.  We also hope:to
 investigate  the extent  to  which concurrent T4 treatment will
 protect against cognitive  deficits  and neurochemical  and
 morphological changes associated with perinatal PCB exposure.

      It is important to note  that mechanisms underlying cognitive
 deficits  seen following PCB exposure are likely to be complex and
 multifaceted.   It is unlikely that  altered endocrine  function
 will  account for all of the observed effects.   Seegal et al.'s
 findings  suggest that a number of PCB congeners, particularly
 those that are lightly  chlorinated  and ortho-substituted, may act
 directly  to  alter brain dopamine content. We are currently in the
 process of evaluating one  of these  congeners, 24,4'-
 trichlorobiphenyl (PCB  28)  for behavioral effects.  In contrast
 to the other congeners,  we  have evaluated, thyroid function does
 not  appear to be altered by PCB 28.  In  addition to our
 behavioral testing and  assessments  of thyroid function,  we will
 be measuring dopamine levels and turnover, dopamine receptor
 binding and  tyrosine hydroxylase activity in specific brain
 regions from rats exposed  to each of the three congeners
 currently under study.   We hope these studies evaluating
 behavioral changes,  thyroid function and neurochemical function
 in the same  litters  of  PCB-exposed  animals will begin to shed
 some  light on the interrelationships between these different
 toxic  endpoints.

      It is becoming  increasingly clear that TEFs based on AHH
 induction are unlikely  to  be sufficient  for estimating neurotoxic
 risk  from PCB exposure.   Other structure-activity relationships
 related to thyroid and  dopamine endpoints are likely to be
 equally or more  important  in estimating neurotoxic risk and I
believe we need  to focus more of our energy on clarifying
 structure-activity relationships for the thyroid and dopamine
 effects and on determining  the role of these effects in mediating
 cognitive and behavioral deficits.

                     Pre-Meeting  Comments  for
          Workshop on Developmental Neurotoxic Effects
                Associated with  Exposure  to PCBs
                    Research Triangle  Park, NC
                      September 14-15, 1992
                           Rich Seegal
          Wadsworth Center for Laboratories and Research
               New York State Department of Health
                   Empire  State Plaza  -  Box 509
                     Albany, New York 12201
     These comments are based both on thoughts derived from my
research as well as those gathered from selected reading of
comments of other participants.  At least two major issues need
to be answered in order to:  (i) estimate the neurological risk to
humans following perinatal exposure to PCBs and (ii) determine
the mechanisms by which polychlorinated biphenyls (PCBs) alter
central nervous system (CNS) function.

     First, it is essential to determine if the active agent(s)
responsible for the observed changes in CNS function are the
parent congeners or their metabolites.
     We  (Seegal, Shain and Bush) have begun to address this
problem.  Using both in-vivo and in-vitro techniques we have
demonstrated that exposure of the non-human primate to complex
mixtures of PCB congeners results in significant decreases in
brain concentrations of dopamine and accumulation of a small
number of ortho-substituted congeners.  Exposure of cells in
culture to these congeners, either alone or in combination,
resulted in significant decreases in cellular dopamine content
suggesting that the parent congeners are responsible for the in-
vivo decreases in dopamine.  Further work in culture (manuscript
in preparation), using radiolabelled 2,5,2'5' has demonstrated
significant decreases in cellular dopamine content in the absence
of detectable amounts of metabolites of 2,5,2'5'.  These results
indicate that metabolism of a putative neurotoxic congener may
not be required for reductions in cellular dopamine.

     What is needed?  Although considerable data has accumulated
demonstrating that PCB metabolites are readily excreted in urine
and bile, that few metabolites are detected in brain and that
little cytochrome P450 activity is present, the majority of this
information has been gathered in the adult animal. Given the
absence of a mature blood-brain-barrier, the developing fetus
could be at neurological risk due to exposure to PCB metabolites
generated by the exposed dam.  This potential problem,  as well as
further investigation of the generality of the above mentioned
in-vitro findings need to be determined.  Furthermore,  given the
tenet that the developing organism is most sensitive to exposure
to neurotoxic agents, behavioral and neurochemical endpoints need

 to be gathered following exposure to a more complete  range of
 doses_of both readily metabolizable dioxin-like and ortho-
 substituted congeners as well as non-metabolizable  congeners.   In
 addition,  determination of possible interactions between these
 classes of congeners should be determined.

      Measurement of brain concentrations  of congeners using glass
 capillary gas chromatography with electron  capture  detection as
 well  as mass spectrometric analysis of congeners and  metabolites
 will  aid in determining the relationships between these  agents
 and(the above mentioned endpoints.   The lack of commercially-
 available metabolites (for analytical standards and experimental
 exposure)  as well as radiolabelled congeners and metabolites (to
 increase analytical sensitivity)  needs to be rectified.   Indeed,
 this  lack of a wide spectrum of these agents has impeded the
 ability to vigorously pursue this important question.

      Secondly,  it is of the utmost  importance to determine the
 mechanisms by which perinatal exposure to either the  parent
 compound or its daughters alter central nervous system function.
 These effects could be direct (e.g.  inhibition  of tyrosine
 hydroxylase by either the parent  or daughter products) or
 indirect by altering the concentrations of  endogenous
 neurohormones.   An example of a direct effect of PCBs  has been
 shown by us when we demonstrated  that exposure  of
 pheochromocytoma (PC12)  cells in  culture  to either  commercial
 mixtures of PCBs or individual congeners  decreased  cellular
 dopamine content by inhibiting the  synthetic capability  of
 tyrosine_hydroxylase,  the rate limiting enzyme  in the  synthesis
 of  dopamine.   These in-vitro findings need  to be confirmed ±xi-
 vivo  in both the developing and adult organism.

      Additionally,  perinatal exposure to  PCBs may induce  profound
 effects  on brain function by altering the hormonal  status  of the
 animal  or  human during development.   Such mechanisms can  be
 classified as  indirect and,  as suggested  in the  papers provided
 by  Dr. McKinney,  may include changes  in estrogenic  and thyroid
 function.   At present  there is insufficient  data  to indicate that
 changes  in either endocrine system  is  solely responsible  for
 inducing the  complete  spectrum of behavioral  and  neurochemical
 changes  seen  following perinatal  exposure to  PCBs.  Nevertheless,
 data  we  have  gathered  using both  dioxin-like and  ortho-
 substituted congeners  administered perinatally  suggest that  some
 of  the dopaminergic  effects  may be correlated with  changes in
 estrogenic  function.   Indeed,  there is  a large and growing
 literature  describing  the  relationship  between estrogen and
 dopamine.   The  relationships  between  changes in brain dopamine
 content  and the known  estrogenic, anti-estrogenic and
 thyromimetic effects of ortho  and dioxin-like PCBs are shown
below in the following table.

                 PCB Exposure:   Ortho vs. Co-planar
l^e •
aSeegal, Toxicologist 12:320 (1992)
bAgrawal et al., Toxicol. Lett. 7:417-424 (1981)
°Seegal et al., Toxicology 66:145-163 (1991)
dTilson et al., Environ. Res. 18:466-474 (1979)
 Chishti and Seegal, Toxicologist 12:320 (1992)
eKorach et al., Mol. Pharmacol. 33:120-126 (1988)
'Gierthy etal., Biochem. Biophys. Res. Comm. 157:515-520 (1988)
 Safe et al., Pharmacol. Toxicol. 69:400-409 (1991)
gvan den Berg et al., Toxicol. Lett. 41:77-86 (1988)

      These relationships,  which must  be confirmed by additional
experiments, not only demonstrate  a possible causative link
between the estrogenic activity of certain PCB congeners and
neurochemical  function,  but also provide support for the argument
that  the neurological effects of PCBs differ based  on the
structure of the congener.  Thus,  Dr. Barnes' rhetorical question
of whether it  is possible to distinguish between the neurological
actions of dioxin-like and non-dioxin-like PCBs may be partially
answered by examining differences  in  neurochemical  function
following perinatal and adult exposure to these two classes of
PCBs.   Further evidence of differences in activity  between these
two major classes of  congeners is  provided by data  from a study
examining the  effects of intrastriatal injection of individual
PCB congeners -into adult rats.  As suggested by Tilson et al.,
the adult  animal is insensitive to exposure to the  dioxin-like
congeners,  but, as we have recently demonstrated, responds to
ortho-substituted PCBs by decreases in striatal dopamine       '

                          Strlatol Dopamlne ConcentraUons
                              24 hrs Post-Injection
                          Aroclor 2,4.4' 3.4.3-.41 3A,53'A'
What is needed?   First,  greater collaboration between
investigators  concerned with the developmental neuro.toxic effects
of PCBs.  It is  difficult for a single laboratory to carry out
the large number of  behavioral,  neurochemical and endocrine
studies needed to answer the questions posed in the position
papers.  For example,  as questioned by Peterson,  are changes in
neurotransmitter content or receptor levels sufficient to be
considered neurotoxic?  If changes in the development or
expression of  behaviors are needed to be considered neurotoxic,
there are few  laboratories that are capable of conducting both
the biochemical  and  behavioral studies needed to truly define a
'neurotoxic' response-following perinatal exposure to PCBs.  I
suggest that EPA provide a funding mechanism (e.g. an RFP) for an
inter-laboratory,  multi-year project to begin to answer the
questions raised in  the position papers,  the participants'
comments and those that will emerge from the meeting.

     Secondly, efforts should be made to standardize the test
animal.  Are behavioral and neurochemical measurements obtained
in the rat comparable  to those seen in non-human primates and

ultimately the human?  Given the financial and ethical cost of
using non-human primates, additional effort should be devoted to
determining the relationship between behavioral and neurochemical
changes in the non-human primate and rodent.  If the rat is an
appropriatetmodel for assessing the behavioral and neurochemical
effects of perinatal exposure to PCBs EPA should define a
standard rat strain for use by researchers investigating these
problems. In addition, the periods of exposure to PCBs should be
standardized.  Should female rats be exposed to PCBs prior to
mating in order to achieve steady-state levels of PCBs?  This
approach would more closely approximate the human exposure

     Thirdly, additional effort should focus on the neurological
effects of perinatal exposure to complex environmental mixtures
of putative environmental neurotoxicants,  including laboratory
studies using material derived from contaminated wildlife (e.g
Daly's rodent studies).  Finally, we should not consider that
PCBs and related halogenated hydrocarbons exist in an
'environmental vacuum'.  For example, at least on a superficial
level, the behavioral effects in humans and the neurochemical
effects in laboratory animals following exposure to lead are
similar to those seen following exposure to PCBs.  Do halogenated
hydrocarbons interact neurologically with other known or
suspected neurotoxicants?

                     Pre-meeting Comments for
          Workshop on Developmental Neurotoxic Effects
                Associated with. Exposure to PCBs
                    Research Triangle Park, NC
                      September 14-15, 1992
                           Anne Sweeney
                     Program  in Epidemiology
                    College of Human Medicine
                    Michigan  State University
                        A206 East Fee Hall
                   East Lansing/ MI 48824-1316
General Toxicity of PCBs

     This set of papers provided a review of PCB toxicity
primarily in terms of immunotoxicity, carcinogenesis,   ,. : .
mutagenesis, and genotoxicity, mainly derived from animal
studies.  With regard to the three key points of deliberation,
presented in this issue paper, I submit the following comments:.;
     A.  How do dose levels associated with developmental
neurotoxicity compare to dose levels associated with 10-4
cancer risk?
     First, it should be noted that the dose levels associated
with cancer risk are based on data from animal risk assessment  .
 (generally linearized) models, and the difficulties in
extrapolating to,humans are,widely acknowledged.1  Data  :
currently available from,studies of human PCB/PBB exposure do not
provide information on congener-specific effects of PCBs, •    '•.,•-.
although these data will be forthcoming.  Prospective studies of
perinatal PCB exposure  (preferably initiated prior to conception)
utilizing congener-specific analyses are required to address this
issue.  These studies may enable us to understand the conflicting
results reported previously regarding PCB exposure
transplacentally and through lactation on birthweight,:
gestational age, and subsequent neurodevelopmental.effects in   :
these offspring.   ;    ,        ,   .       :

     B.  Is there a threshold for immunotoxicity* and how do the
dose levels associated with this endpoint compare with dose
levels associated with developmental neurotoxicity?

     Again, congener-specific analyses should be performed to
answer these questions;  Moreover, it is important to determine
if there is any relationship between immunosuppression and a
quantifiable adverse human health outcome rather than designating
immunosuppression itself as the endpoint.                .     '

      C.   Is  there  a  correlation between congeners implicated  in
adult neurotoxieity  and congeners responsible for developmental

      This important  issue  should figure prominently in the design
of  future prospective  studies.  Seegal et al 2 have indicated
that  the  lightly chlorinated  congeners were more effective at
decreasing dopamine  concentration than the more highly
chlorinated  congeners.  It appears that the highly chlorinated
PCBs  represented the greatest proportion of body burden in the
human studies of perinatal exposure to PCBs which evaluated women
from  the  general population  (with no occupational exposure).
This  would indicate  that a method to measure current continuous
low-level exposure to  PCBs should be developed and incorporated
into  the  study design.  I  am  here thinking in terms of measures
other than invasive  procedures for serum testing, which may be
unacceptable during  pregnancy.

Some  specific comments  on  the individual papers

      1.   There are interesting discussions pertaining to the
animal models and  the  role of PCBs/PBBs as initiators and/or
promoters in carcinogenesis.  While conditions to assess these
mechanisms may be  easily   controlled in animal models, there
needs to  be  some discussion on how this could be evaluated in
population-based human  studies, or in occupational settings,
where multiple exposures (which may include an initiating
agent/s)  may be present before, during, and after exposure to

      2.   The example of the hairless mouse model (Silberhorn et
al, p. 461)  supports the necessity of evaluating genetic
susceptibility in  human studies of PCB exposure-adverse health
outcomes.  If only a subset of the population is susceptible to
the effects  of PCBs  (e.g.,  why do only 13% of cigarette smokers
develop lung cancer?) then this would dilute any observed effect.

     3.   With regard to the human epidemiologic studies of
PCBs/PBBs and cancer (Silberhorn et al, pp.477-81),  it is not
indicated whether  the cancer  mortality rates presented were age-
adjusted  or  not, which could  have great impact on the
interpretation of  these findings,  particularly when rates were
calculated over 20-36 year intervals with changing denominators.

     In addition to  these  studies being few in number, they
suffer from  inadequate sample sizes,  crude measures of exposure,
and lack  or  inadequate control for confounding variables.   I
would like to focus  on the common lament that these studies lack
the power to detect  associations between exposure to PCBS/PBBs
and adverse health outcomes (especially with a rare disease such
as cancer).  I would like  to  see a discussion on the feasibility
of establishing exposure registries,  based on specific criteria
for exposure.  These discussions should include methods of "case"
ascertainment and data collection and management procedures,

which if standardized could provide 1) consistency and 2) power
to future epidemiologic evaluations of these exposures and health

Developmental Neurotoxicity of PCBs in Humans

     This issue paper very nicely presents the dilemma of the
conflicting results from studies assessing perinatal exposure to
PCBs and developmental neurotoxicity in human offspring.  While
this endpoint is the major outcome of interest, effects on
birthweight and gestational age are also alluded to/ and may
themselves be related to subsequent development in these

     Although the highly exposed populations of Yusho and Yucheng
are discussed, the primary concern is the effect of chronic low
level PCB exposure in the general population.  Following are
comments on some of the specific questions raised in this issue

     Question 1.  Before this question is addressed, the group's
knowledge could benefit from a discussion of the sampling
technique employed in the Michigan study, specifically:

     a).  Why was the study designed with a case/control ratio of
1/3?  This is a serious compromise of study power for no apparent

     b).  Attrition rates were high in both cases (29%) and
controls  (38%) , and this was higher among women of low
socioeconomic status.  Also, it should be noted that this
attrition was higher in the "unexposed" group, which already
under-represented the control women.

     c).  In the later study of the 4 year children, what was the
refusal rate by exposure status?

     Question 2.  The exposure levels reported among "cases" in
the first Michigan study of prenatal PCB exposure and birthweight
was a mean of 6.1 ppb  (+.3.7) in maternal serum; among  "controls"
the mean PCB level was 4.1 ppb  (+_ 2.7), indicating a substantial
overlap in exposure.  Given the skewness of these PCB data, it
would be helpful to see the median levels in these groups as

     Again, in the later study of the 4 year old children, 75.4%
 (n=236) of the original cohort participated, but no mention is
given of their exposure status as defined in the first  study.
How many of these children were from the "exposed" group?

     Of these 236, 62%  (n=146) have cord serum levels available;
of the  172 mothers who breast fed, 120  (70% provided breast milk
samples.  Finally, only 178  (75%) of the children provided serum

samples at 4 years of age; an.additional 27 at 5 years.
these missing'data impact on the observed results?
     Question 5.  Since the North Carolina study also collected
data on cord serum PCB levels/ as well as maternal serum levels,
can these analyses be re-run to examine these specimens and
compare the results with the findings in the Michigan study?

     Question 6.  In the later studies evaluating children at 4
years of age, both the Michigan and the North Carolina studies
used the McCarthy Scales of Children's Abilities and found
conflicting results.

     One point to consider is the timing of exposure to PCBs, in
that the women in the Michigan study who consumed Great Lakes
fish during pregnancy may have a very different PCB profile,than
women exposed to the "normal background" levels during pregnancy.
Could these intermittent "bolus" exposures to PCBs (contaminated
fish meals) during organogenesis have a different effect on the
developing fetus than continuous, low dose exposures?

     Please refer to the comments under Paper #1 for suggestions
for discussion on how to approach these methodological issues in
future epidemiologic studies.
1. Klassen CD.  In; Cass'arett and Doull's Toxicology, 3rd edition,
Macmillan; New York, 1986: pp.28-32.

2.   Seegal  R.,   et  al.   Lightly  chlorinated ortho-substituted
congeners  decrease dopamine  in non-human  primate brain  and in
tissue culture.  Toxicol Appl Pharmacol 1990;106:136-44.





                   U.S. Environmental Protection Agency

                   EXPOSURE TO PCBs

                   September 14-15, 1992
                   Research Triangle Park, NC

                   FINAL AGENDA
8:00 AM

9:00 AM




11:00 AM


11:45 AM

Registration/Check-in at the U.S. Environmental Protection Agency (U.S. EPA)

Linda Birnbaum, Health Effects Research Laboratory, U. S. EPA


General Toxicity/Exposure
Michael Bolger, Center for Food Safety and Nutrition, U.S. Food and
Drug Administration

Linda Birnbaum


Structure-Activity Relationships
James McKinney, Health Effects Research Laboratory, U.S. EPA

Risk Assessment for Developmental Toxicity
Gary Kimmel, Reproductive and Developmental Toxicology Branch,





 Human Developmental Neurotoxicity Data
 Co-chairs:     Maty Ptince, National Institute for Occupational Safety and Health
              Jane Adams, University of Massachusetts

 Animal Developmental Neurotoxicity Data
 Co-chairs:     Diane Miller, Health Effects Research Laboratory, U.S. EPA
              Carole Kimmel, Reproductive and Developmental Toxicology
              Branch, U.S. EPA '

 Discussions in both workgroups will focus on criteria for assessing study data,
 relevant pharmacokinetic and structure-activity considerations, and whether the
 available data are sufficient or insufficient, for risk assessment.


Chair:       Elaine Francis- Pesticides/Toxics Team, U.S. EPA

Summary of Workgroup Discussions and Comparison of Data

Human Data               -       Mary Prince

Animal Data                      Carole Kimmel

Summaries will focus on specific types of effects'that can be compared between
animal and human data.



Chair:        Elaine Francis
             Michael Bolger
             Carole Braverman
             John Brown
             Carole Kimmel
             John Moore
             Mary Prince



Discussions will include determination of sufficiency or insufficiency of the data
base as a .whole, .and the impact of pharmacokinetics and structure-activity
relationships on developmental toxicity data. Also, the discussions will consider
the relationship of developmental neurotoxic effects to other PCB toxicity.


Chair:        Hugh Tilson, Health Effects Research  Laboratory, U.S. EPA

Summary of Workgroup Discussions

Human Data                     Jane Adams
                    Animal Data
                                 Diane Miller
Chair:    >••   Hugh Tilson    .    .       ,  .

Panelists:      Jane Adams ,
              Earl Gray
              Joseph Jacobson
              Diane Miller
              Deborah Rice                ,
              Walter Rogan

Discussions will focus on research needs that can be of benefit in addressing
uncertainties in the data base.'                  ,

Carole Kimmel               .

                                       Panel Members
 Dr. Jane Adams
 University of Massachusetts

 Dr. Donald Barnes
 U.S. Environmental Protection Agency

 Dr. David Bellinger
 Children's Hospital

 Dr. Robert Bowman
 Wisconsin Regional Primate
 Research Center
 University of Wisconsin

 Dr. Theo Colborn
 World Wildlife Fund

 Dr. Eric Dewailly
 Center Hospital of Laval University

 Dr. Kim Dietrich
 University of Cincinnati College of Medicine

 Dr. Beth Gladen
 National Institute of
 Environmental Health Sciences

 Dr. L. Earl Gray
 Developmental Toxicology Division
 U.S. Environmental Protection Agency

 Dr. Steve Hamilton
 General Electric

 Dr. G. Jean Harry
 National Institute of
 Environmental Health Sciences

 Dr. Joseph L. Jacobson
 Wayne State University

 Dr. Sandra W. Jacobson
 Wayne State University

 Dr. Ruth E. Little
National Institute of
Environmental Health Sciences
 Dr. James McKinney
 Health Effects Research Laboratory
 U.S. Environmental Protection Agency

 Dr. Diane Miller
 Health Effects Research Laboratory
 U.S. Environmental Protection Agency

 Dr. John A. Moore
 Institute for Evaluating Health Risks

 Dr. Richard E. Peterson
 University of Wisconsin

 Dr. Mary Prince
 National Institute for Occupational
 Safry and Health

 Dr. Deborah C. Rice
 Heath and Welfare Canada
 Banting Building, Ross Avenue

 Dr. Walter Rogan
 National Institute of
 Environmental Health Sciences

 Dr. Susan Schantz
 University of Illinois

 Dr. Richard Seegal
 New York State Department of Health
 Wadsworth Center  for
 Laboratories and Research

 Dr. Anne Mellody Sweeney
 College of Human Medicine
Michigan State University

Dr. Gerhard Winneke
University of Dusseldorf

           APPENDIX C


                    U.S. Environmental Protection Agency

                    EXPOSURE TO PCBs

                    Research Triangle Park, NC
                    September 14-15, 1992

                    PARTICIPANT LIST
Jane Adams
Assistant Professor of Psychology
Department of Psychology
University of Massachusetts - Boston
100 Morrissey Boulevard
Boston, MA 02125-3393
Fax: 617-265-7173

Cynthia Alley
Center for Environmental Health
Environmental Health and Injury Control
Centers for Disease Control
1600 Clifton Road, F-17
Atlanta, GA 30333
Fax: 404-488-4609

Donald G. Barnes
Staff Director, Science Advisory Board
U.S. Environmental Protection Agency
401 M Street, SW (A-101)
Washington, DC 20460
Fax: 202-260-9232

David Bellinger
Assistant Professor of Neurology
Neuroepidemiology Unit
Gardner House, Room 457
Harvard Medical School/Children's Hospital
300 Longwood Avenue
Boston, MA 02115
Fax: 617-735-7940
*Linda S. Birnbaum
Director, Environmental Toxicology Division
Health Effects Research Laboratory (MD-66)
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Fax: 919-541-4324

*Michael Bolger
Center for Food Safety and Applied Nutrition
U.S. Food and Drug Administration
200 C Street, SW (HFF-156)
Washington, DC  20204
Fax: 202-205-4758

Robert E. Bowman
Professor Emeritus, Psychology
Harlow Primate Laboratory
University of Wisconsin - Madison
22 North Charter Street
Madison, WI  53715
Fax: 608-262-6020

*Carole Braverman
U.S. Environmental Protection Agency
77 West Jackson Boulevard
Chicago, IL 60604-3507
Fax: 312-353-4788
  indicates workshop technical panel member

 John F. Brown, Jr.
 Manager, Health Research
 Chemical Research Center
 General Electric
 Corporate Research and Development
 P.O. Box 8, Building Kl
 Schenectady, NY  12301-0008
 Fax: 518-387-5604

 Theo Colborn
 Senior Fellow
 World Wildlife Fund
 W. Alton-Jones Foundation
 1250 24th Street, NW
 Washington, DC 20037-1175
 Fax: 202-293-9345

 Eric Dewailly
 Director of Environmental Health Service
 Community Health Department
 2050 Boulevard St. Cyrille Quest
 Sainte Foy, Quebec G1V 2K8
 418-687-1090 (x222)
 Fax: 418-681-5635

 Kim N. Dietrich
 Associate Professor
 Department of Environmental Health
 College of Medicine
 University of Cincinnati
 Cincinnati, OH 45267-0056
 Fax: 513-558-4838

 *Elaine Z.  Francis
 Chief, Pesticides/Toxics Team
 Office of Technology Transfer and
 Regulatory Support
 U.S. Environmental  Protection Agency
401 M Street, SW (H-8105)
Washington, DC 20460
Fax: 202-260-6932
 Beth Gladen
 .Mathematical Statistician
 Statistics and Biomathematics Branch
 National Institute of
 Environmental Health Sciences
 P.O. Box 12233 (B3-02)
 Research Triangle Park, NC  27709
 Fax: 919-541-4311

 L. Earl Gray, Jr.
 Section Chief
 Developmental Reproductive Biology
 Section (MD-72)
 Health Effects Research Laboratory
 U.S. Environmental Protection Agency
 Research Triangle Park, NC  27711
 Fax: 919-541-5138

 Jean Harry
 National Institute of
 Environmental Health Sciences
 P.O. Box 12233 (Wl-02)
 Research Triangle Park, NC 27709
 Fax: 919-541-0295   '             :;

 Joseph L.Jacobson
 Department of Psychology
 Wayne State  University
 71 West Warren
 Detroit, MI 48202
 Fax: 313-577-7636

 Sandra W. Jacobson
 Senior Research  Scientist
 Department of Psychology
 Wayne State University
 71 West Warren    ,                  '
 Detroit, MI  48202
Fax:  313-577-7636
"indicates workshop technical panel member

"Carole A. Kimmel
Reproductive and Developmental Toxicology
U.S. Environmental Protection Agency  .
401 M Street, SW  (RD-689)
Washington, DC 20460
Fax: 202-260-3803

*Gary L. Kimmel
Developmental Toxicologist
Reproductive and Developmental Toxicology
U.S. Environmental Protection Agency
401 M Street, SW  (RD-689)
Washington, DC  20460       .
Fax: 202-260-3803

James McKinney
Chief, Pharmacokinetics Branch (MD-74)
Health  Effects Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Fax: 919-541-5394

Diane Miller
Research Toxicologist
Neurotoxicology Division (MD-74B)
Health Effects Research Laboratory
U.S.  Environmental Protection Agency
 Research Triangle Park, NC  27711
Fax:  919-541-2631

John A. Moore
 Institute for Evaluating
 Health Risks
 1101 Vermont Avenue, NW - Suite 608
 Washington, DC 20005
 Fax:  202-289-8530
Richard E. Peterson
School of Pharmacy
University of Wisconsin
425 North Charter Street
Madison, WI 53706
Fax: 608-265-3316

Mary M. Prince
Risk Assessment Program
Division of Standards Development
and Technology
National Institute of Occupational
Safety and Health
4676 Columbia Parkway (C-15)
Cincinnati, OH 45226
Fax: 513-533-8588

Deborah C. Rice
Toxicology Research Division
Health Protection Branch
Health and Welfare Canada
Banting Building - Ross Avenue
Tunney's Pasture
Ottawa, Ontario K1A OL2
Fax: 613-957-1907

Walter Rogan
National Institute of
Environmental Health Sciences
P.O. Box 12233 (A3-05)
Research Triangle Park, NC  27709
Fax:  919-541-2511 or 7887

Susan L. Schantz
Assistant Professor of Toxicology
 Institute for Environmental Studies
 University of Illinois at  Urbana-Champaign
 1101 West Peabody Drive
 Urbana, IL 61801
 Fax: 217-333-8046 •
 "indicates workshop technical panel member

  Richard Seegal
  Research Scientist
  Wadsworth Center for
  Laboratories and Research
  New York State Department of Health
  Empire State Plaza - Box 509
  Albany, NY 12201
  Fax: 518-474-8590

  Anne Sweeney
  Assistant Professor
  Program in Epidemiology
  College of Human Medicine
  Michigan State University
  A-206 East Fee Hall
  East Lansing, MI 48824-1316
  Fax: 517-336-1130

  *Hugh A. Tilson
 Neurotoxicology Division (MD-74B)
 Health Effects Research Laboratory
 U.S. Environmental Protection Agency
 Research Triangle Park, NC 27711
 Fax: 919-541-4849

 Gerhard Winneke
 Medical Institute of
 Environmental Hygiene
 University of Dusseldorf
 AuFm Hennekamp 50
 W-4000 Dusseldorf
 Fax: 011-49-211-3190-910
"Indicates workshop technical panel member

                   U.S. Environmental Protection Agency

                   EXPOSURE TO PCBs

                   Research Triangle Park, NC
                   September 14-15, 1992

                   OBSERVER LIST
Melvin E. Anderson
Instructor in Medicine
Duke University Medical Center
2200 West Main Street - Suite B-200
Durham, NC  27710
Fax: 919-541-5394 '

Stan Barone
Project Scientist
Neurotoxicology Section
ManTech Environmental
Research Triangle Park, NC 27711
Fax: 919-541-4849

George Becking
Team Leader, IPCS/IRRU
World  Health Organization
P.O. Box 12233  (EC-07)
Research Triangle Park, NC 27709
Fax: 91.9-541-2712

John Cicmanec
Environmental Criteria and
Assessment Office
U.S. Environmental Protection Agency
26 West Martin Luther King Drive
Cincinnati, OH  45268
Fax: 513-569-7916
John Dougherty     '
Manager of Environmental Compliance
Heavy Duty Electric     :
P.O. Box 268
Goldsborough, NC  27530
Fax: 919-580-3248

E.G. Farnsworth
General Electric Company
3135 Easton Turnpike
Fairfield, CT  06431

John Festa
American  Paper Institute
1250 Connecticut Avenue, NW
Washington, DC  20036

Tim Fitzpatrick
Industrial  Hygienist
The Aluminum Company of America - Alcoa
1501 Alcoa Building
Pittsburgh, PA 15219-1850
Fax: 412-553-3835

Ellen Goldey
Neurotoxicology Division (74B)
Health Effects Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
919.541-4229   "                 .

  S.B. Hamilton
  Manager, Environmental
  Science and Technology
  General Electric Company
  3135 Easton Turnpike
  Fairfield, CT  06431
  Fax: 203-373-2683

  Thomas Hill
  General Electric Company
  3135 Easton Turnpike
  Fairfield, CT  06431
  Fax: 203-373-2683

  David Hoel
  Director, Division of Biometry and
  Risk Assessment (A3-02)
  National Institute for
  Environmental Health Sciences
  P.O. Box 12233
  Research Triangle Park, NC 27709
  Fax: 919-541-7887

 Jan Johannessen
 Research Biologist
 U.S. Food and  Drug Administration
 8301 Muirkirk  Road (HFF-162)
 Laurel, MD  20708

 Robert Kaley
 Director of Environmental Affairs
 Monsanto Company
 800 North Lindbergh Boulevard
 St.  Louis, MO  63167
 Fax: 314-694-6858

 Prasada Ruo Kodavandi
 Visiting Scientist
 Health Effects Research Laboratory
 U.S. Environmental Protection Agency
Research Triangle Park, NC  27711
  Edward D. Levin
  Assistant Professor of Psychiatry
  Medical Center
  Duke University
  Box 3412
  Durham, NC  27710
  Fax: 919-286-6825

  Barbara McElgunn
  Research Service Committee
  Learning Disabilities Association of America
  74 Holmcrest Trail
  West Hill, Ontario  MIC 1V5
  Fax: 416-283-0344

  Audrey R. McMahon
  Research Service Center                .
  Learning Disabilities Association of America
 2991 Princeton Pike
 Lawrenceville, NJ  08648
 Fax: 609-771-4214

 John Schell
 TERRA,  Inc.
 325 John  Knox Road
 Atrium Building - Suite 201
 Tallahassee, FL 32303
 Fax: 904-422-0333

 Mark Stanton
 Environmental Health Scientist
 Neurotoxicology Division (74B)
 Health Effects Research Laboratory
 U.S. Environmental Protection Agency
Research Triangle Park, NC  27711
Fax: 919-541-4849

Thomas R. Ward
Neurotoxicology Division (74B)
Health Effects Research Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC  27711

Suzanne Wuerthele
Regional Toxicologist
U.S. Environmental Protection Agency
999 18th Street - Suite 500
Denver, CO 80202

Mei Lin Yu
Survey Research Associates
100 Capitola Drive - Suite 30]
Durham, NC 27713





                                     Richard Seegal

                     Wadsworth Center for Laboratories and Research
                          New York State Department of Health
                                       Albany, NY
       In Dick Peterson's terminology, the approaches I have suggested are "bottom-up" (e.g.,
mechanistic in nature). Given the complex nature of PCBs and the knowledge that PCB
congeners differ in their neurotoxic "potency," it is essential for a fine tuning of the risk of
human exposure to complex mixtures of PCBs to continue investigation of the mechanisms by
which PCBs alter central nervous system function. Yet do we wait for a final determination
before determining a preliminary risk assessment? On the other hand, it  is also important to
validate the lexicological changes induced by perinatal exposure to PCBs  (e.g., changes in brain
neurotransmitter concentrations, thyroid and steroid hormone concentrations) by correlating
these changes with alterations in function (e.g., a "top down" approach).

       Finally, it is important to resolve differences in the development of nervous system
function seen in the studies of Jacobson et al. and Rogan et al.  Are the differences in
neurological  development simply methodological in nature or do they reflect differences in
exposure to mixtures of putative environmental contaminants (e.g., contaminants in Lake
Michigan fish other than PCBs)?  Given the relevance assigned to the perinatal non-human
primate behavioral studies by the "animal" group, it would be feasible, and highly productive, to
contrast the behavioral (and neurochemical?) consequences of perinatal exposure to: (i) mixtures
of PCB congeners found in Lake Michigan fish and (ii) the same congeners plus the known and
suspected neurotoxic agents found in the fish.


                                      Gerhard Winneke

      Department of Psychophysiology. Medical Institute of Environmental Hygiene. University of
                      Diisseldorf1. Hennekamp 50. Dusseldorf. FR Germany.
                                      E. Rudy Boersma

    Department of Obstetrics and Gynecology. State University Groningen. Oostersingel. Groningen.
                                      The Netherlands.
                                       Pieter J. Sauer

   Division of Neonatology. Sophia Children's Hospital. Erasmus University. Gordelweg. Rotterdam.
                                      The Netherlands.


        Epidemiological findings on neuromotor and neurobehavioral developmental deficit associated
 with early exposure to PCBs and related compounds (Jacobson et al., 1985, 1990; Rogan et al, 1986-
 Gladen et al., 1988), although generally consistent, nevertheless present some uncertainties The most
 prominent of these include uncertain causality, namely insufficient control of potentially important
 confounders (e.g. Hg, Pb), differences in terms of the spectrum of effects, namely neuromotor vs
 cognitive, and inconsistency regarding the persistence of deficit The prospective study to be described
 now might help to clarify some of these gaps of knowledge. The following description is based on a
 transnational study proposal submitted to the Commission of the European Communities (DG XII) for
 funding within the forthcoming RTD Programme in me field of environment, area Environment and
 Health.                   .      .

       This study expands on an ongoing clinical research program in the Netherlands, in which 200
newborns each were (and still are being) sampled at the Department of Obstetrics and Gynecology of
the State University Groningen (E.R. Boersma) and at the Sophia Children's Hospital of the Erasmus
University Rotterdam (P.J. Sauer). PCB-levels are measured in maternal blood, cord blood and in
maternal milk. First reports on this program are available (Kaam et al., 1991; Koopman-Esseboom et
    | Additional participants from the University of Dusseldorf are Eberhard Schmidt, University Child
Clinic, Dept. of General Pediatrics, Neonatology and Nutrition, and Hans-J. Steingriiber, Institute of
Medical Psychology.

al., 1992; Huisman et al., 1992). This program will be amplified within the present study proposal by
(1) extending the observation period beyond age 18 until age 42 months in part of the Dutch sample,
and (2) by adding extreme groups of between 200 to 300 newboms from the Dusseldorf area with
high and low cord blood PCB-concentrations, respectively. The Dutch and the German contributions
are expected to be complementary hi terms of the covered observation period, the general sampling
strategy, and a partly different spectrum of outcome measures. A core protocol covering independent,
dependent and confounding variables, training efforts, as well as ring-tests for quality assurance serve
to harmonize the Dutch and the German elements of the-program.

       Sampling Strategies.  The Dutch and the planned German procedures differ in this respect.
Both in Groningen (Boersma) and in Rotterdam (Sauer) approximately 200 pregnant women each
volunteered to participate in having breast milk samples collected and in having their babies examined
for subsequent neurodevelopment until the age of 18 months. In terms of PCB-levels in cord blood
and in breast milk, this sample presumably represents intermediate levels of PCB body burden. The
Dusseldorf sample, on the other hand, is planned to be sampled as the upper and lower 10% of 1500
to 2000 cord blood samples within  12 to 18 months; cutoff points based on available frequency
distributions will have to be defined to approximately meet the desired proportions.
PCB-concentrations in breast milk spot samples will be measured at 10 days and 4 weeks of age.

        Independent Variables. The basic unifying variable is the PCB cord blood level. According
to Ballschmitter & Zell (1980), the  three congeners 138,153,180 (x 1.64) will be determined by
GC/MS, whereas in breast milk, three additional congeners (28, 52,  101) will be measured. To ensure
comparability, ring tests will be organized between the participating analytical laboratories in
Dusseldorf'(Institute of Environmental Hygiene), Kiel (Laboratory of Environmental Toxicology),
Zeist (TNO), and Wageningen (Agricultural University).  In addition to PCB-concentrations in cord
blood, PCB-coneentrations in breast- land bottle milk and infant formula are measured in the Dutch
samples in a systematic manner, whereas—due to culture-specific differences hi nursing habits—such.
information will be available in only about 40% of the German sample. Whereas the Dutch groups
collect 24 hour-samples, the German data will be based on spot breast milk samples.

        Dependent Variables. The  agreed upon study protocol includes both essential as well as
optional variables. Essential variables include the early assessment of sensorimotor and neurological
status according to Touwen et al.(1980) and Prechtl (1977) to be done at 10 days of age, the Bayley
Scales of Infant Development and the "Visual Recognition Memory Test" (Pagan & McGrath,  1981) to
be given at the age of 7 months, the neurodevelopmental status according to Touwen (1979) and,
again, the Bayley scales at 18 months of age. The Dutch groups will reassess about half of their cohort
at about 42 months of age using, among others, the Dutch adaptation of the Kaufman Assessment
Battery for Children (Kaufman & Kaufman, 1983) and the Dutch version of the Reynell
Developmental Language Scales (Reynell, 1977). Optional variables, to be introduced in the German
sample, include the digitized analysis of spontaneous infant vocalizations, to be recorded under
standardized conditions at the age of 6 and 12 months according to the methods developed at the
Institute of Medical Psychology, University of Dusseldorf (Bisping et al., 1990). Confounding
Variables and Criteria for Inclusion: Only 1st and 2nd births of families whose native language is
Dutch or German, respectively, with a 1st Apgar of at least 7 are or will be considered. Important
confounders to be considered are: (1) The "Antenatal Obstetric Optimaliry Score" (Touwen et al.,
 1980), (2)  length of gestation, (3) maternal age, (4) socioeconomic status, (5) the adapted HOME-scale

  (Caldwell & Bradley, 1984), (6) birth weight, (7) anthropometric growth, (8) nursing habits and
  supplementary bottle feeding, and (9) Pb and Hg in blood. This listmight be modified in the course of
  further developments.                                         '

         Temporal Structure.  The study is planned for three years starting early in 1993. Within this
  period the Dutch groups will extend the follow-up.of about half of their cohort until 42 months of age,
  whereas the German cohort will be studied until age 18 months;  continuation of the project is planned.'

 REFERENCES                                                                             ''"'<•

 Ballschmitter, K., Zell, M. (1980) Analysis of polychlorinated biphenyls (PCB) by glass capillary gas
 chromatography. Fresenius Z. Anal. Chem. 302, 20-31.

 Bisping, R., Steingruber, HJ. et al. (1990) Adult's tolerance of cries:  An experimental investigation
 of acoustic features.    Child Develop. 61, 1218-1229.

 Caldwell, B.M., Bradley, R.H. (1984) Home observation for measurement of the environment.
 Administration Manual.—Little Rock: University of Arkansas.

 Pagan, J.F., McGrath, S.K. (1981) Infant recognition memory and later intelligence. Intelligence

 Gladen, B.C., Rogan, WJ.  et al. (1988) Development after exposure to polychlorinated biphenyls and
 dichlorodiphenyl dichloroethene transplacentally and through human milk. J. Pediatr. 113, 991-995.

 Huisman, M., Koopman-Esseboom, C. et al. (1992) Risk assessment of fetal exposure to
 polychlorobiphenyls (PCB's). Paper presented at 13th European Congress of Perinatal Medicine, Mav

 Jacobson, J.L., Jacobson, S.W. et al. (1984) Prenatal exposure to an environmental toxin: A test of the
 multiple effects model. Develop. Psychol. 20, 523-532.

 Jacobson, J.L., Jacobson, S.W., Humphrey, H.E.B. (1990) Effects of in utero exposure to
 polychlorinated biphenyls and related contaminants on cognitive functioning in young children  J
 Pediat. 116, 38-45.

 Kaam, A.H.L.C.,  Koopman-Esseboom, C.  et al. (1991) Polychloorbifenylen (PCB's) in moedermelk,
 vetweefsel, plasma en navelstrengbloed; gehalten en correlaties.
 Ned. Tijdschr. Geneeskd. 135,  1399-1403.

 Kaufinan, A.S., Kauflnan, A.L. (1983) Kaufinan Assessment Battery for Childrea American Guidance
 Service. Minnesota: Circle Press.

 Koopman-Esseboom, C., Weisglas-Kuperus, N. et al. (1992) Effects of PCB's and dioxins during
 pregnancy and breast feeding on growth and development of newborn  infants; a study design. In:
 Skakkebaek, RE. et al.(eds.) Impact of the environment on reproductive  health. Enviroa Health
Perspect. (in press)

Reynell, J. (1969) Reynell Developmental Language Scales. Experimental Editions. Windsor N.E.F.R.

Rogan, WJ. et al. (1986) Neonatal effects of transplacental expossure to PCB's and DDE. J. Pediatr.
109, 335-341.

Touwen, B.C.L. et al.(1980) Obstetrical condition and neonatal neurological mobidity. An analysis
with the help of the optimality concept Early Human Develop. 4, 207-228.

Touwen, B.C.L. (1979) Examination of the child with minor neurological dysfunction. Clin. Develop.
Med. SIMP with Heinemann Medical. Lippincott

»U.S.GOVERNMENTPRINTINGOFF1CE 1993 .75 0 - 002/-60 1 98