ATDEPARTMENT OF HEALTH AND HUMAN SERVICES

 

FOOD AND DRUG ADMINISTRATION

 

CENTER FOR FOOD SAFETY AND APPLIED NUTRITION

 

 

 

 

 

 

 

 

 

 

 

FOOD ADVISORY COMMITTEE

 

METHYLMERCURY

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Tuesday, July 23, 2002

 

8:30 a.m.

 

 

 

 

Sheraton College Park Hotel

4095 Powder Mill Road

Beltsville, Maryland  20705

 

PARTICIPANTS

 

Sanford A. Miller, Ph.D., Chair

Catherine DeRoever, Executive Secretary

 

MEMBERS

 

Alex D.W. Acholonu, Ph.D.

Francis Fredrick Busta, Ph.D.

Annette Dickinson, Ph.D.

Johanna Dwyer, Ph.D.

Lawrence J. Fischer, Ph.D.

Marion H. Fuller, D.V.M.

Joseph H. Hotchkiss, Ph.D.

Lawrence N. Kuzminski, Ph.D.

Ken Lee, Ph.D.

Thomas J. Montville, Ph.D.

Robert M. Russell, M.D.

Mr. Brandon Scholz

Michael W. Shannon, M.D.

 

TEMPORARY VOTING MEMBERS

 

H. Vas Aposhian, Ph.D.

Sarah L. Friedman, M.D.

Ms. Jean M. Halloran, Consumer Representative

Margaret McBride, M.D.

Richard E. Nordgren, M.D.

Clifford Scherer, Ph.D.

 

GUEST SPEAKERS

 

Mr. Harvey Clewell

Dr. Christopher DeRosa

Ms. Caroline Smith DeWaal

Dr. Philippe Grandjean

Dr. James Heimbach

Dr. Joseph Jacobson

Dr. Penny Kris-Etherton

Dr. Gary Myers

Dr. Susan Schober

 

FDA

 

Ms. Linda Hayden

 

 

 

 

 

 

 


C O N T E N T S

 

Welcome

  Dr. Sanford Miller                                     5

 

Conflict of Interest Statement

  Catherine DeRoever                                         7

 

Opening Remarks

  Joseph A. Levitt                                           9

 

Introductions                                               19

 

National Academy of Sciences Report on

the Toxicological Effects of Methylmercury

  Dr. Joseph Jacobson                                       22

 

Questions of Clarification                                  46

 

Faroe Island Study

  Dr. Philippe Grandjean                                    51

 

Questions of Clarification                                  72

 

Seychelles Study

  Dr. Gary Myers                                        87

 

Questions of Clarification                              105

 

Agency for Toxic Substances and

Disease Registry

  Dr. Christopher DeRosa                                   117

 

Questions of Clarification                                 135

 

Consumer Messages

  Dr. Penny Kris-Etherton                                  139

 

Questions of Clarification                                 152

 

National Health and Nutrition Examination Survey

  Dr. Susan Schober                                        159

 

Questions of Clarification                                 172

 

Risk Management Strategies for Methylmercury

in Seafood - A Consumer Perspective

  Ms. Caroline Smith DeWaal                                183

 

Questions of Clarification                                 198


C O N T E N T S(Continued)

 

Fish Consumption Data and Risk Assessment Calculations

  Dr. James Heimbach                                       215

 

Questions of Clarification                             241

 

  Mr. Harvey Clewell                                       244

 

Questions of Clarification                                 259

 

Public Comment

  Dr. Rhona Applebaum, NFPA                                273

  Dr. J.H. Lee, NCPR                                       277

  Dr. Richard Fisher                                       283


P R O C E E D I N G S

Welcome and Introductions

    DR. MILLER:  Good morning.  I am Sandy Miller and I am serving as the Chairman of the Food Advisory Committee for the Center for Food Safety and Applied Nutrition.  I would like to welcome you all to this meeting, which was called in order to help the Center develop a policy for methylmercury in food.

    The Center has developed a number of questions which they want the committee to consider, and these will be discussed in just a few moments.

    Let me just go into the agenda.  Let me just see if I can get some ground rules in place.  This is a very tightly packed agenda.  If we are going to be done anytime within the next month or two, we are going to have to stick to the exact times that have been assigned by the secretary.  These generally have been determined by the speakers themselves, but in some cases, in order to finish the agenda, the times have slightly changed.

    Nevertheless, the important thing is that exactly on time, I will ask you to step down.  I will try five minutes before the end of your time to remind you that there is five minutes to go, but it is really important that we stick to the time.

    I know this is an issue of some concern and a great deal of passion to a lot of people, and it is of vital importance to us, but we want to be fair to everybody, the times must really be kept.

    Secondly, just to indicate the basis for which the committee is operating, it is the function of this committee to look at these questions which we will be asked on the basis of the science.  Our recommendations are individual recommendations to the Center to be based entirely on that issue.

    Policy determinations are complex and they involve things that are equally important to the science, but are different.  This committee is not designed to deal with those issues, so I am asking you all to try to focus your attention on the issues concerned with the science and the science only.

    Let me introduce to you Cathy DeRoever, who is the Executive Secretary of the Food Advisory Committee, who will talk about some housekeeping issues.

    MS. DeROEVER:  Thank you, Dr. Miller.

    Good morning.  I would like to welcome all of our members and our temporary voting members.  Thank you very much for being here today.  Before I do actually the administrative announcement, for the record, I want to announce that we have appointed several temporary voting members.

Conflict of Interest Statement

    The authority to appoint such members is granted to the Center Director, and I have letters for the temporary voting members that state:  By the authority granted under the Food Advisory Committee Charter, I appoint Dr. Aposhian, Dr. Friedman, Ms. Halloran, Dr. McBride, Dr. Nordgren, and Dr. Scherer as temporary voting members of the Food Advisory Committee for the July 23rd through 25th, 2002 meeting on methylmercury

    The letter is signed by the Center Director, Center for Food Safety and Applied Nutrition, U.S. Food and Drug Administration, Mr. Joseph Levitt.

    Second, also for the record, all members and temporary voting members have been screened for financial conflicts of interest.  Upon review of the FDA Form 3410, which is the financial disclosure report for special government employees, we have determined that no financial conflicts exist.

    Similarly, we have asked all our guest speakers to complete a financial interest and professional relationship certificate for guests and guest speakers to identify any potential conflicts.

    We have not received all of those forms, but for the record, there are two that I would like to mention.  Dr. Heimbach has had a relationship with the seafood industry, and Dr. Kris-Etherton, who will be speaking I believe it's tomorrow, has a relationship also with the seafood industry.

    Moving on to the administrative matter, for the people at the table, in your notebook is a menu that we are going to ask you if you would like to have lunch, we have tried to overcome some past problems we have had with respect to timing, so if you would take a moment and complete it, the staff will collect it and your lunch will be ready, hopefully, when we break.  If this works well, we will try it again for Wednesday and Thursday, but I will appreciate your feedback on that.

    With that, I turn it back to Dr. Miller.

    DR. MILLER:  Thank you, Cathy.

    To open this session of the Food Advisory Committee, Mr. Joseph Levitt, who is Director of the Center for Food Safety and Applied Nutrition, has some opening remarks.

Opening Remarks

Joseph A. Levitt

    MR. LEVITT:  Good morning.  Again, my name is Joe Levitt.  I am Director of the Center for Food Safety and Applied Nutrition.  I am pleased to welcome all of you here for a meeting of the Food Advisory Committee.

    This is my first visit with the committee since you were reconstituted.  I was on vacation when there was a meeting earlier this spring.  I welcome Dr. Miller, who is chair of our committee.  Simply no one has more experience in the broad issues facing our Center than Dr. Miller given  his past experience at the Agency and his work on many National Academy of Sciences committees.

    I also will look forward to working with the committee as a whole and its many subcommittees over the coming months and years, and in the fall, I will look forward to providing you on a day when we have a less intense agenda with an overview of our Center's activities and on engaging your advice on a number of important scientific and public health issues, which brings me to this week's meeting on FDA's Consumer Advisory regarding methylmercury and seafood consumption.

    We consider this issue to be a very important public health issue.  Indeed, I can't think of anything more important than ensuring the health of pregnant women and their unborn children.

    That is why we went to great lengths to assemble such a distinguished committee.  For those not familiar with our committee structure, we have included here members of our standing Food Advisory Committee, members of our new Subcommittee on Food Contaminants, and additional scientific experts in specialties needed for the methylmercury issue that we did not already have.

    This includes medical experts in pediatrics and neurology, as well as additional experts in developmental psychology and risk communication.

    Finally, I want to thank Jean Halloran of Consumers Union for serving as our consumer rep given that our standing consumer rep was not here.  I am not sure I see her yet, but we will thank her in advance of her arrival.

    Again, I want to thank everybody for taking the time from your very busy schedules to address this subject.

    This issue methylmercury in fish has a long history dating back to the 1970s.  This includes industrial poisonings in Japan and Iraq, major studies being conducted in geographical areas of heavy fish consumption, steps being initiated by both the FDA and the Environmental Protection Agency, as well as a number of risk analyses and data gathering exercises.

    You will hear about all of this and more over the next three days I can assure you.  The critical point of departure for this week's discussion is a report issued by the National Academy of Sciences on July 11th, 2000.  The report was actually directed to EPA under the rubric of reducing environmental pollution by this relevant to FDA regarding commercial seafood consumption was readily apparent.

    Therefore, after the National Academy of Sciences' report, we, at FDA, undertook a very extensive process to examine the risks of methylmercury in commercial seafood and to determine what advice to give consumers at the national level.

    We consulted with many of the same people and organizations that you will hear at this meeting.  We conducted a series of focus groups with consumers to examine communication style and format that any new advisory would have, and we consulted with EPA, which issues advisories to states for recreationally caught fish.

    I personally led this outreach effort and I participated in every or nearly every meeting with outside groups.  I also met regularly with our internal staff on a regular basis.  The culmination of this effort was an updated consumer advisory that FDA issued in January 2001 with a small revision in March a couple months later.

    Let me now summarize the advisory itself.  The consumer advisory was addressed to pregnant women and women of childbearing age who may become pregnant.  In short, the advisory has two main parts.  The first part says to avoid eating four kinds of fish with the highest levels of mercury, namely, shark, swordfish, king mackerel, and tilefish, avoid these completely if you are in that category.

    No. 2.  Eat 12 ounces per week of a variety of other fish including shellfish, canned fish, smaller ocean fish or farm-raised fish.  Just be sure to pick a variety of different species.

    I tend to summarize this advisory to say avoid the top four fish and eat the rest in moderation.

    Now, in issuing this advisory, FDA also put on our web site our written rationale for the advisory and data tables showing levels of methylmercury in different species of commercial fish, so the public could see how we reached the conclusion we did.

    We followed the issuance of the advisory with an extensive outreach campaign and were able to get our message into a number of newspapers, magazine, and other information outlets.

    When we issued this advisory in early 2001, it was our genuine belief that if women conscientiously followed this advisory, based on knowledge of methylmercury levels in fish and consumers' fish consumption levels, that these women would be protecting their unborn children from harm due to methylmercury.  That was our goal.

    But I can tell you, a year and a half later, the subject remains controversial and I will tell you quite candidly that a number of persons and organizations still feel that we fell somewhat short of the mark.  That is why we are here.

    We want American women to have the best advice possible and for that advice to be arrived at in an open and transparent way.  At the time that we constructed our advisory a year and a half ago, CFSAN did not have at the time the benefit of our Food Advisory Committee.  We were in the process of recharging and restructuring it, and so forth.

    So, we assembled the advisory ourselves using the best information we had and the process that I described, but now that we have reconstituted this advisory committee and it is fully functioning, we wanted to bring the issue to you.  We look forward to a full airing of the issues, ultimately focusing on whether, after everything is considered, the advisory is as strong as it needs to be to protect public health.

    Now, over the next three days, you will hear a wide range of views on this subject.  Your job is first to listen critically to the whole story that will be presented to you over the next three days.  You will hear, starting with the representative from the Academy that issued the report.  You will hear directly from a number of researchers who have conducted the key studies.

    You will hear from state and federal officials, from physicians, from consumer groups, from industry representatives.  Finally, you will hear from experts in our center who will try to explain how we arrived at the conclusions that we did.

    Then, we will stop and you will have your discussion.  We want your best thinking and advice on our advisory on whether it is adequate in its present form or whether any adjustments need to be made.

    Now, let me go through the specific charge.

    This will be circulated and everybody will have copies of this if they don't already.  It will be coming around shortly.

    [Slide.]

    The charge to the committee, I have kind of a long version and a short version.  The long version says:

    The committee is being asked to evaluate, in light of all the relevant information about potential consumption, exposures, population body burden, hazard and consumer messages, whether the Agency's Consumer Public Health Advisory on Methylmercury is adequate to protect the health of those who follow that advice.

    When I read that, I said that covers everything, but let's say it a little more simply.

    [Slide.]

    To put it more simply, does the FDA advisory provide adequate protection for pregnant women and women of childbearing age who may become pregnant?  That's the bottom line question.  If not, what changes are needed and why?  If yes, are there nevertheless enhancements to the advisory that would make it easier and more effective for women to follow it?

    Now, to help answer this general charge, we have framed it in five questions.

    The first question says:  Has the Agency adequately addressed and appropriately considered all the relevant factors and information that bear upon the elaboration of a consumer advisory on fish consumption?  Are any factors not relevant?  Are there additional factors that should be relevant?  In other words, have we considered the landscape.

    No. 2.  Focusing on the first part of the advisory, should the advisory have specifically advised pregnant women to avoid any other species not specifically mentioned, and if so, what would be the scientific rationale?

    I will tell you as you will hear, that the species most commonly mentioned would be fresh tuna is the fish you will hear a lot about, whether it ought to be included here or not.  So, we want that to be talked about.

    No. 3.  In the second part of the advisory, should the Agency issue a fish listing as an adjunct to the advisory to clarify what is mean by "variety of fish?"

    As we have gone back and looked at the advisory with hindsight, we knew what we meant, but have we provided enough information on how to eat the appropriate variety of fish, so that women are adequately protected.  We would like advice there.

    No. 4.  You will hear a lot about FDA and EPA, so we ask the question:  Should the Agency revised our advisory to make explicit that the 12 ounces per week includes all sources of fish, both recreational and commercial, so there is a better nexus?  There may be additional ways you consider how we and EPA can better be sure that we are connected, we have our web sites joined, and so forth, maybe there are additional ways.

    Finally, a subject of monitoring.  Should the Agency increase its monitoring of methylmercury in commercial fish in order to keep this advice current?  When you go through the data tables, you will see that some of the species have lots of samples, some have very few samples associated with them, and the question of monitoring and importance of that comes up, so we would like your advice on that.

    Let me conclude.  Dr. Miller asked that I describe for you two things:  Number one, why you are here, and, two, what FDA needs from you over the next three days.  I hope I have done that.

    I also hope that I have conveyed that we are truly open and indeed want your best advice whether you agree with us or not.  You will see I believe, as I did, that there is a wide range of strongly differing views about methylmercury in fish.  It is an emotionally charged issue.

    There is also a long history of scientific debate about this issue that will not likely end with this meeting although it would be nice.  Our collective challenge, therefore, during the next three days will be to rise above any such divisiveness.  We need to do what is best for the American consumer, in this case, American women and their offspring.  They certainly deserve no less.

    Thank you very much.  I will try personally to stay for as much of the meeting as I can although I am sure a couple times I will get pulled out for different issues.

    Again, thank you for your time.  You will have a fascinating three days, I can assure you, but most importantly, we hope that you will help us move and advance this issue in a way that will be best for American women and their children.

    Thank you again very, very much, and thank you, Dr. Miller, for chairing the meeting.

    DR. MILLER:  Thank you, Joe.

Introductions

    Before we proceed, let me take this opportunity of having the various members of the committee introduce themselves, at least for the record, so we know you are here.

    We will begin with Dr. Scherer.

    DR. SCHERER:  Cliff Scherer, Cornell University, Department of Communications.  My specialty is risk communication.

    DR. NORDGREN:  Dick Nordgren.  I am a pediatric neurologist from Dartmouth Medical School.

    DR. McBRIDE:  Margaret McBride.  I am a pediatric neurologist from Rochester, New York, and Akron, Ohio.

    DR. FRIEDMAN:  Sarah Friedman from the National Institute of Child Health & Development, one of the NIH institutes.  I am a developmental psychologist.

    DR. RUSSELL:  Rob Russell.  I am Director of the Human Nutrition Research Center at Tufts.

    DR. MONTVILLE:  Tom Montville, Professor of Food Science from Rutgers, The State University of New Jersey.

    DR. FULLER:  Marion Fuller.  I am the Director of Food Safety for the Florida Department of Agriculture and Consumer Services.

    DR. FISCHER:  I am Larry Fischer, Director of the Institute for Environmental Toxicology at Michigan State University.

    DR. HOTCHKISS:  Joe Hotchkiss from the Department of Food Science at Cornell University.

    DR. LEE:  Ken Lee, Ohio State University, Department of Food Science, Professor and Chair.

    DR. KUZMINSKI:  I am Larry Kuzminski, a retired Vice President of Technology from Ocean Spray, had previous positions to Ocean Spray that included officer positions with the Kellogg Company and tenure professorship at University of Massachusetts.

    DR. MILLER:  I am Sandy Miller and I am associated with the Center for Food Nutrition Policy at Virginia Tech University.

    MS. DeROEVER:  Catherine DeRoever, FDA.

    DR. BUSTA:  Frank Busta.  I am a Professor Emeritus, Department of Food Science and Nutrition, at the University of Minnesota.

    DR. ACHOLONU:  Alex Acholonu.  I am from Alcorn State University in Mississippi.  I am a Professor of Biology and my specialty is epidemiology of diseases.

    DR. DICKINSON:  Annette Dickinson, Vice President for Scientific and Regulatory Affairs with the Council for Responsible Nutrition.

    DR. DWYER:  Johanna Dwyer, Assistant Administrator for Human Nutrition, Agricultural Research Service, USDA.

    DR. SHANNON:  I am Michael Shannon.  I am a pediatrician and toxicologist at Children's Hospital and Harvard Medical School in Boston.

    DR. APOSHIAN:  I an Vas Aposhian from the Department of Molecular & Cellular Biology and the Department of Pharmacology at the University of Arizona.  My research interests have for many years been the toxicology of heavy metals including mercury and arsenic.

    DR. MILLER:  Thank you all.  There are still a couple of members of the committee that have not yet arrived.  When they do, we will have them introduce themselves.

    Let me make a request, that when you speak, try to speak into the microphone since there is a record of this meeting being kept.

    Our first speaker is Dr. Joseph Jacobson, a member of the National Academy, Committee on Methylmercury.  Dr. Jacobson is from Wayne State University.

    Dr. Jacobson.

National Academy of Sciences Report on

the Toxicological Effects of Methylmercury

Dr. Joseph Jacobson

    DR. JACOBSON:  First of all, I want to apologize.  I am not quite sure how I got to the 20 minutes, but in reviewing this morning, I really am going to need 30.  We are a little ahead, so hopefully, that won't be a problem for your schedule.

    DR. MILLER:  Okay, as long as we stay within the schedule, the exact moment.

    DR. JACOBSON:  I am a developmental psychologist and I am going to be giving you an overview of the history behind the constituting of the NAS panel, as well as some of the logic and thinking that went into the process of the conclusions that we reached in our report.

    Vas Aposhian was a member of the panel.  He is here, so he can correct me if I get some of the details wrong.

    [Slide.]

    Obviously, everyone here I am sure is familiar with the fact that prenatal exposure to methylmercury can have very serious developmental consequences for the central nervous system was first established in the Minamata incident in Japan, that led to some of the more severe deficits that were seen at the very heavy levels of exposure in that population.

    [Slide.]

    And then, of course, the second famous mass poisoning was in Iraq in the early 1970s when seed grain that had been contaminated with the methylmercury fungicide was used to bake bread because there was drought and the infant who were born to mothers who ate the contaminated bread while they were pregnant showed very similar severe neurological problems.

    One important difference between the two episodes as that in Iraq, a group of researchers from the University of Rochester went in and did systematic assessments of a large number of the infants who were exposed, very systematic developmental assessments, and so we had, not just the qualitative descriptions, but also some reasonably semi-quantitative data that risk assessment could be based on.

    [Slide.]

    EPA, when it did, not the most recent risk assessment, but the one before that, used the Iraqi data as the basis for the risk assessment, and the developmental endpoints that they used were developmental milestones - age of walking, age of talking, which were affected by the very heavy exposure levels in this population.

    The EPA Iraqi risk assessment was the first to use a benchmark dose analysis for purposes of risk assessment. Prior to that, the method used was the NOAEL, the No Observed Adverse Effect Level method, which tended to be based on animal studies where different groups of animals would be exposed at different levels, and the lowest level at which no adverse effect was seen was the one that was used for the EPA reference dose, the reference dose being an estimate of the average daily intake at which you wouldn't find adverse effects.

    When EPA and other agencies began to move to human data, we got the increasing popularity of the benchmark dose statistical assessment, which doesn't look for discrete groups, and, of course, in human exposures, you don't get discrete groups, people tend to be exposed over a broad range of exposures, and the benchmark dose analysis uses the full range of exposures and the outcomes associated with those exposures to arrive at a statistically driven estimate of the level where you might not see an adverse effect.

    To do the benchmark dose analysis, you have to start out by taking a cutoff.  Well, first of all, you have to start out by picking an endpoint, and I will talk later about the choice of endpoints that the NAS Panel considered based on the data that were before us, but you have to pick an endpoint.

    Once you have that endpoint, you have to pick a cutoff, and the cutoff represents the level at which you are saying the child is doing very poorly and we become very concerned.  On an IQ test, we will often pick the level of 70, the borderline for mental retardation, and we will talk about 70 as the cutoff, that is, 70 as the level of poor performance that we are trying to prevent an increased incidence of.

    So, we take an endpoint that methylmercury increases the incidence in the population, an endpoint where methylmercury makes it more likely that we are going to get a bad effect, and we pick a cutoff, and we say we want to make sure that we do not appreciably increase the number of children who are performing at that level just by virtue of the fact that they were exposed to methylmercury.

    The benchmark response is our criterion for how much of an increase we are willing to tolerate.  Let's say we are willing to tolerate a deficit of 70, an IQ deficit of 70, and we are willing to tolerate an increase of 1 percent or 2 percent.

    Typically, we don't want there to be enough methylmercury exposure in the population that the incidence of mental retardation is increased by 5 percent or 10 percent.  So, the benchmark response is our decision.

    These are all policy driven, what level of performance is the cutoff, what level of performance are we really trying to prevent an increased incidence of, what benchmark response are we willing to tolerate how much of an increased incidence of poor performance are we willing to tolerate, and once we have made those decisions, we use the dose-response data from our studies to plot a dose-response line, usually, it is done as a straight line, and that dose- response data lets us determine the dose, the level of exposure at which we get that increased incidence that we are very, very eager to prevent.

    So, this is a statistically driven analysis.  We use the full range of the dose-response data to derive a benchmark dose, and then we set 95 percent confidence limits around that dose, and the lower 95th percentile is called the BMDL, the lower limit of the benchmark dose.

    That is the point of departure that EPA used in its Iraqi risk assessment, as well as in the more recent risk assessment to derive the RfD.  What I am leaving out of this is once you get the BMDL, once you get the lowest level at which we expect to see a deficit in the normal population, then, you add the uncertainty factors that I am sure this group will be considering in more detail later on.

    [Slide.]

    As I said, the Iraqi data were used in the initial methylmercury assessment by EPA, but there were several problems with using the Iraqi data, and those included the fact that the developmental milestones, age of walking, age of talking, are fairly gross endpoints.  They are not very sensitive, and they are not very predictive of how a child is going to do later on.

    They were used because they were the best that was available using human data.  The alternative was to extrapolate from animal data, but humans and animals will often metabolize metals differently, and so the feeling is that if we go with human data, there are some advantages to that.

    The other major disadvantage with the Iraqi data was that the exposure was so high, that there were very few individuals in that sample who were exposed in the range at which we get exposure in the general population.

    So, we were plotting a dose-response curve and then extrapolating down to apply to our population in a range at which there were very few datapoints.

    So, although from some perspectives, the Iraqi risk assessment represented a real advance from a scientific point of view, there were really very serious problems in using it to make inferences for contemporary exposure, the other being, of course, that an acute exposure from seed grain is not necessarily going to have the same kind of damage as a chronic exposure from fish over a long period of time.

    As a result, NIEHS, the National Institute of Environmental Health Sciences, funded two very large and very well-designed, prospective longitudinal studies of prenatal methylmercury exposure that started during the early 1990s.

    One was in the Seychelles Island in the Indian Ocean, the other in the Faroe Islands in the North Sea.  Those locations were chosen because they had populations where there were people who ate a lot of fish, and so you could get--and it is always optimal in these kinds of samples--to take a population where there is a broad range of exposures, in other words, you are going to get the clearest picture if you can see the full dose-response curve.

    There still was good overlap with the exposures that we get in the U.S. population, but there was a broader range, and so that made those two populations optimal.

    [Slide.]

    The Seychelles Islands was the first to report effects, and the first effect that they reported came from a pilot study.  It's actually rather large by my standards for a pilot study, and this was 217 children who were assessed at age 5, and they actually, mainly had very low exposed children and heavily exposed children.

    They assessed 9 developmental endpoints, and I am showing them 3 of them here, but actually, they found statistically significant associations between prenatal methylmercury exposure and adverse outcome on 4 of their developmental endpoints.

    One was the General Cognitive Index, which is like an IQ score for pre-schoolchildren from the McCarthy scales of children's abilities, which is an IQ-type test for pre-schoolchildren.  Another was on the Perceptual Performance subtest of the McCarthy, visual-spatial function, preschool language, and Auditory Comprehension was the fourth one that is not shown.

    What was unusual about this report was that the investigators themselves, after reporting the data, tended to discount it, and they discounted it on two grounds - one, that there were 4 outliers, which when they were dropped from the analysis, the results were no longer statistically significant, and the other was that they had not measured a full range of potential confounding variables, particularly social class, and had not controlled for them.

    The Panel reviewing these data raised questions about whether they should have been discounted.  Dropping outliers, not all statisticians agree that is the best way to handle outliers.  A lot of statisticians feel they should be recoded.  The social class, we did not think was necessarily such a serious problem here, because methylmercury tends to be more concentrated in more expensive fish, and, if anything, it is likely that the more middle class children got the heavy exposures.

    But again, this is a pilot study, and these are tentative data, and no one would do a risk assessment based on these data, but they were the first indication from this more recent series of studies of adverse effects.

    [Slide.]

    However, when the Seychelles group, which is the University of Rochester group, assessed the main study, the full cohort several years later, at the same age, that is, at age 5 1/2, they found no evidence of adverse effect.

    [Slide.]

    I have a slide here to demonstrate there just was no relationship between exposure and outcome.  I mean, you know, you can look and look and look, but there is nothing going on in these data of this second main cohort.

    [Slide.]

    The endpoints they assessed tended to be relatively global endpoints, that is, again the McCarthy General Cognitive Index, preschool language tests, tests of academic achievement as opposed to what we call narrow band, domain-specific tests where a test might look in great detail at one aspect of cognitive function, such as sustained attention or vocabulary or visual-spatial reasoning, and so forth.

    Here, the Seychelles group tended to use more global tests and saw no adverse effect.

    [Slide.]

    They also used maternal hair as their measure of prenatal exposure.  Mercury is excreted in the hair, and so if the mother has long hair, which most of the women in this population do, and you get a sample of her hair after she delivers--and hair we know grows at a rate of about 1.1 cm per month--you can take the hair, you can estimate when during pregnancy the mercury was excreted into the hair, and get a very accurate retrospective picture of the mercury intake during pregnancy.

    [Slide.]

    So, as I said, Seychelles, age 5 1/2, main study, no evidence of adverse effect, not even a hint or a suggestion in the data.

    By contrast, we had the Faroe study at age 7, which did not use global tests, they used more narrow band, domain-specific tests, and they reported I think it's out of 20 outcomes, they found adverse effects on 8 of them, and these were actually in many domains.

    Even though they were domain-specific tests, adverse effects were found across the full range of cognitive and neuromotor functioning including fine motor function, finger tapping, sustained attention, short-term memory as assessed in the digit span test, vocabulary as assessed in the Boston Naming test, and verbal learning and memory as assessed in the California Verbal Learning test.

    The publication of these two studies, as I said, both well designed, both very large samples, exposure levels, very similar, presented quite a quandary both to scientists and regulators - how is it possible that you can conduct two large-scale comprehensive, state-of-the-art studies and come up with such absolutely contradictory findings.

    One of the first responses to the appearance of these two sets of contradictory data was that NIEHS and EPA, and other agencies, ATSDR, Chris DeRosa was involved in this who is here today, convened a workshop in Raleigh, North Carolina in 1998, where there were 50 scientists assembled.  Larry Fischer was one of the scientists on that panel.  We spent 2 1/2 days intensively scrutinizing these two studies, and we brought in the investigators, and, boy, I would not have liked to have been in their shoes for the close questioning, detail by detail, that these people underwent.

    But we learned a lot.  We learned a lot about the studies at that time, and the question, as I said, from a scientific point of view, is how can you have two large-scale studies come up with absolutely contradictory findings.

    [Slide.]

    At the end of the 2 1/2 days, having considered the data, the panel concluded that there were several differences in exposure and design that could explain how one study concluded there were adverse effects and the other study concluded no adverse effects.

    One had to do with biomarker of exposure.  As I mentioned, the Seychelles study used maternal hair mercury. I didn't mention the Faroe study, in their initial reports, used cord blood mercury.  Now, cord blood mercury reflects intake during pregnancy during the last, I think it is 12 weeks or the latter part of pregnancy, and that is the period when we get a lot of neuronal proliferation differentiation and brain development that might well be involved in the kind of endpoints that were being assessed in these studies.

    So, it was plausible to say maybe the Faroe's group was measuring mercury at the right time to see the problem.  However, subsequently, the Faroe's group, they also had the hair samples, they just hadn't analyzed them yet, analyzed the maternal hair and looked at it in relation to these developmental endpoints, and found the same adverse effects as they had found with cord blood, so the first theory which seemed to explain the differences kind of fell by the wayside.

    The second was that global tests were used in the Seychelles, domain-specific tests in the Faroes.  In retrospect, that doesn't seem all that convincing to me, because the Faroes group used the domain-specific tests, but found adverse effects across a broad range of domains, so the global tests, in my view, should have picked up.  The narrow band tests are important if the deficit is only in one or two narrow domains, but it is clear from the Faroe's data that it isn't.

    So, in retrospect, that one, I think we should have not put as much stock in as we did at the time.  The age point, 5 1/2 years, is not a good time to do developmental testing because it is a period of rapid brain growth, rapid cognitive reorganization, and relatively subtle neurotoxic effects are not likely to show up.

    The Faroe's group, I think picked a better age point, 7.  Once children have gotten past that developmental period, they usually perform in ways that are more stable and reliable over time.

    Then, a fourth difference was the source of exposure.  Both populations ate a lot of fish, but in the Faroe's, they also get a lot of methylmercury exposure from whale meat, and the difference in the fish, it is a low level chronic exposure, the woman is eating a little bit every day.

    The whale meat, the methylmercury is much more concentrated, and they will beach the whale, there will be a big party, there will be splurging on whale meat, so to speak, and you could get some very heavy doses that perhaps could explain why you might see the adverse effects in the Seychelles, and not in the Faroes.

    The fifth difference, that is not up there, is that the Faroes were heavily exposed to PCBs, which is a ubiquitous environmental contaminant, and there are two concerns about PCBs that you have to understand.

    One is that it is possible, since you get PCBs from fish and methylmercury from fish, it is possible that we have confounding.  It is possible that the same children who get heavy methylmercury exposures, also get heavy PCB exposures, and that where you think you are measuring methylmercury, you are really measuring PCBs.  That is one possible problem with the PCB exposure.

    A second possible problem is that there may be synergism between the PCB and methylmercury exposure, that is, being exposed to methylmercury when you are also at the same time being exposed to PCBs may make the methylmercury more toxic.  There is no good mechanism that has been hypothesized as to why the synergism would exist, but it certainly is possible in theory.

    So, basically, these differences between the two cohorts allowed the 50 scientists in the Raleigh meeting to come up with an intellectually very satisfying understanding about why one well-designed study would lead to one set of results and another two, a very different set of results, but it presented no relief to the regulators who these are the data they have, and they have to one way or another make sense of them and come up with some criteria and recommendations.

    Around this time, EPA was eager to set up some new rules to regulate emissions from coal-fired utility plants in the Midwest, and methylmercury is a major pollutant from that source.

    They tried to issue the regulations and were blocked by congressmen from West Virginia and Ohio, and then the resulting battle in Congress led to a directive to NAS to convene an expert panel to try to look more carefully at the data and from a regulator's point of view, do a better job than the Raleigh group and come up with something that could be useful from a regulatory point of view.

    [Slide.]

    That is how the NAS panel was convened with representatives from a broad range of relevant disciplines, epidemiology, developmental psychology, statistics, methylmercury chemistry, and so forth.

    [Slide.]

    One key difference in our deliberations, in the deliberations of the NAS panel, over what had come before in the Raleigh panel, was that in the NAS panel, we considered the results from an earlier study, from a study that was conducted in the late 1980s in New Zealand, which the Raleigh panel was instructed not to pay attention to because it wasn't published.

    Well, after the Raleigh panel meeting, some of it became published, and we considered it in the NAS panel.  Although it is not as large a sample, I think there were about 230 children, not as well designed in terms of controlling for confounders.  There was some control, but not as comprehensive.  It is actually quite a good epidemiological study as these studies go.

    What is interesting about it is that in terms of the sources of exposure and the design of the study, it is very similar to the Seychelles study, that is, the methylmercury was measured in the mother's hair, the children were assessed at about the same age, it was age 6 rather than age 5 1/2, the developmental endpoints were very many of the same global IQ and achievement tests, and so forth.

    The difference was that the New Zealand study found adverse effects using the same exposure measures, the same research design, essentially the same research design as the Seychelles, they saw adverse effects.

    In epidemiological studies, oftentimes what you have to do is go with the weight of the evidence.

    Actually, let me put that off, because that point will be a little bit easier to make toward the end.

    [Slide.]

    So, basically, when we included the Seychelles pilot data and the New Zealand data, these differences that we thought had explained why we are seeing effects in one study, and not in the other, fell by the wayside.

    We are seeing adverse effects with cord blood, mercury, and maternal hair mercury in multiple studies.  We are seeing them on global tests, not just on narrow band tests.  We are seeing them at age 5 1/2 and 6, and we are seeing them in populations where the exposure is just from fish.

    PCBs, we don't know what the PCB exposures are in New Zealand, but there is no reason to expect there to be particularly high levels of exposure in New Zealand.

    DR. MILLER:  Dr. Jacobson, we have about five more minutes.

    DR. JACOBSON:  I will just try to say very briefly what the other slide was going to show, which has to do with the confounding of PCBs and methylmercury in the Faroe study.  There were four endpoints which we shown on the slide where methylmercury affected the endpoint and PCBs did not.  There were four where both methylmercury and PCBs affected the endpoint, and it was very difficult to tease apart that difference, and if people have more questions about it, I will try to explain that in greater detail.

    Basically, once we see that these factors do not explain the differences between the two studies, you are left with the question, who could explain the difference between the two studies.

    Well, one conclusion that we came up with on the panel was that we think that there is an issue of power. Now, when you have samples of 700 and 900 children, it is kind of funny to talk about power, because that would seem to be an adequate sample size to detect anything.

    [Slide.]

    There are two factors.  First of all, much of the effect is going to be seen above 15 parts per million in maternal hair.  We have got lots of cases at the lower end, but when you think about power, power is going to be weakened in these studies--this is the Seychelles data--even in a study of 700 children when you have a limited number of children whose exposure is in the upper end of the distribution where most of the effect is going to be seen, so there may have been a power problem from that point of view.

    [Slide.]

    Then, we took the raw regression coefficients that had been reported in the studies and transformed the standardized regression coefficients to try to get a sense for the magnitude of the effects, and what we find is that the magnitude of a lot of these effects is very, very small, so even very large samples may find it difficult to detect some of these very small effects.

    [Slide.]

    The other issue in an epidemiological study is there are many uncontrolled factors, there are many possible unmeasured confounders.  We find this in the lead literature, the PCB literature, these other exposures where many studies have been done.  You can have a well-designed study that fails to detect an effect that is seen in study after study.

    The reasons are probably that in many populations or in any given sample, there may well be unmeasured factors that you are unable to control for because it doesn't occur to you that there could be confounders, and that is basically why you have to go with the weight of the evidence.

    The basic conclusion in the NAS panel was even though one very well-designed study clearly failed to find effects, one very well-designed study did, one quite well-designed study did, and even the Seychelles pilot gave some indication, so the weight of the evidence seemed to be pretty clearly in the direction of adverse effect.

    [Slide.]

    When you do the benchmark analysis, you have to figure out, the way the methodology works is you have to choose a developmental endpoint that is going to be your guide, that is going to be the one that you do your statistic analyses based on.

    Some people argue it should be the most sensitive endpoint, that is, the endpoint at which you see an effect at the lowest level of exposure.  If we had recommended the most sensitive endpoint to EPA, we would have gone with the New Zealand data because effects were seen at the lowest levels in the New Zealand study.

    We felt that since it was not a well-designed study, it had not had as extensive peer review, it was not as large a sample, that we would do better going with the Faroe study, which did show adverse effects at somewhat higher levels of exposure.

    Based on methylmercury maternal hair, the endpoint at which we saw effects at the lowest doses was the Boston Naming Test, it's a vocabulary test, so that was the endpoint that we decided to go with in terms of recommendation for EPA for its risk assessment benchmark dose computation.

    Thank you.

    DR. MILLER:  Thank you very much.

    Comments, questions from the committee?  Yes, Dr. Russell.

Questions of Clarification

    DR. RUSSELL:  I was wondering, is another possible explanation for the differences that the diets eaten in these various areas differ in other components that could affect the absorption or the bioavailability, if you will, of the methylmercury?  In other words, if the rest of the diet that was eaten with fish is vastly different from place to place, could that affect the amount that is absorbed?

    DR. JACOBSON:  In principle, yes.  I am not sure and I don't think there is any really good theory sketching out what components of the diet that would be.

    There has been research on selenium.  I am pretty sure--well, actually, I am not sure about the Faroes--how is the selenium levels in the Faroes, are they fairly high?  I am pretty sure they are fairly high in the Seychelles.  Maybe we will that off until Philippe has his talk.

    It is certainly plausible.  Then, of course, you have to ask yourself would those particular nutrients be particularly high in the U.S. diet, but, yes, that is something that obviously should be considered and something that we don't have good comprehensive data on.

    DR. MILLER:  Dr. Fischer.

    DR. FISCHER:  Joe, when you chose the Boston Naming Test as the test that was most sensitive to the effects, and calculated a benchmark dose using that, why did you pick a single test instead of picking a score of a group of tests, either those tests that showed an effect or maybe even the whole neurological analysis, a score?

    In other words, it seems to me what you are doing there is picking data from a single test and using it to regulate, when, in fact, you had a whole lot of data that you just seemed to not use.

    DR. JACOBSON:  It is traditional in risk assessment to go with the most sensitive endpoint because from my understanding, the philosophy is we want to protect, so we want to err in the direction of caution, we want to pick the most sensitive endpoint.

    I was particularly comfortable with taking the Boston Naming Test because vocabulary is actually a very, very good surrogate for overall IQ, and if you had to pick a specific test that would be likely to be predictive of how the child would do later on, you would do best with the Boston Naming Test.

    When we looked at the cord blood measure in the Faroe study, there was another test that was actually more sensitive, which was the Sustained Attention Test.  I wasn't comfortable going with that one, first of all, because data had been collected only on half the cohort.  Secondly, it doesn't have very good predictive validity to a broad range of other aspects of function, the way the vocabulary did.

    I would philosophically see nothing wrong with developing a composite measure, and we considered it, but we went in this direction to follow what is traditionally done in the field, that is, to err in the direction of caution.

    DR. FISCHER:  So, scientifically, you would have picked a group of scores, right?

    DR. JACOBSON:  Normally, but I can't make a strong scientific argument that a group of scores would have been any better than vocabulary, because vocabulary is such a robust predictor of a broad range of effects.

    DR. FISCHER:  Then, why do the other tests?

    DR. JACOBSON:  Scientists do the other tests because they want to get a comprehensive understanding of what is going on.  A risk assessor may or may not take all of that information.  You know, formulas are developed, procedures and approaches are developed for various reasons, and as I said, I think the rationale here is to try to get maximal protection.

    DR. MILLER:  Dr. Dwyer.

    DR. DWYER:  Are there any other confounders that come to your mind?

    DR. JACOBSON:  My sense was that these groups did an excellent job of controlling for confounders, and we actually looked at that in great detail in the Raleigh meeting, and were very impressed with it.

    We put the Faroes people, we raked them over the coals, we made them construct an urban/rural variable, and they seemed to pass all the tests.  It seemed no matter what you controlled for, the effects were still there.  Sometimes the effects looked a little weaker, sometimes a little stronger, but my sense is all the things I can think of were controlled for.

    DR. MILLER:  Other questions?

    DR. KUZMINSKI:  Please correct me if I am wrong, but the question is along the same line as Dr. Fischer's, and that in your presentation here today, you have outlined the three studies and the differences and the parameters, but from what I have read in the pre-read material--again, correct me if I am wrong--the Academy committee did not consider the results of the Seychelles study in the deliberation towards recommendations to the EPA on the RfD.

    Am I interpreting that correctly?

    DR. JACOBSON:  Well, we actually did two exercises.  We did one exercise where we took the data from the three studies and integrated them.  This was a statistical exercise, which is kind of averaging along the lines that Dr. Fischer was recommending, and that analysis would have led to a set of recommendations that were surprisingly similar to those that we finally did make.

    But again, we felt that it was appropriate to follow some of the protocols of the way risk assessment has traditionally been done, and typically, what is done is a single study is selected, the best study is selected, the one that appears to be most valid, most sound, most solid, and then within that study, the most sensitive endpoint is selected.

    So, in our final recommendations, we went with what we considered to be the traditional risk assessment approach.  However, as I said, the other exercise did lead us to a surprisingly similar conclusion.

    DR. MILLER:  Thank you very much.

    There are two speakers now dealing with the two principal studies.  The first, Dr. Philippe Grandjean from Odense University to talk about the Faroe Islands study.

    Dr. Grandjean.

Faroe Islands Study

    DR. GRANDJEAN:  Thank you.  I am very pleased to be here.  I have previously worked with colleagues in the U.S. EPA at the ATSDR, at the European Commission.  I am very pleased to be here with FDA now to tell you about our experience in the Faroe Islands.

    [Slide.]

    What I am going to do today is try to look at that research and see what can we learn from that experience if we go beyond the results as such.

    [Slide.]

    Let me just give you a brief overview of why we chose doing research in the Faroe Islands.  It is a rainy place in the North Atlantic between Iceland and Norway.  It's rather cool in summer, but the winters are mild.  The reason it is interesting to us environmental epidemiologists is that people in the Faroes have this tradition of eating pilot whale.

    They chase pods into shallow bays, pods that come near the coast, and for hundreds and hundreds of years, the tradition has been that they kill of pods of these small whales that are nonendangered, it's a sustainable use of the species, so they get this extra supply of proteins and fatty acids, essential fatty acids and vitamins, but unfortunately, it has turned out that the meat and the blubber are contaminated with methylmercury and PCBs respectively.

    [Slide.]

    The reason for doing the research in the Faroes is that it is almost like a natural experiment because the pods do not come in regularly, and when they come in, when they come near the coast, you can't be sure if they will be near one island or another island, therefore, the communities are exposed to methylmercury or PCBs on an irregular basis.  You can't choose it, so it depends on availability of the whale meat.

    At the same time, these people eat a lot of fish, they eat fish for dinner three times a week on the average. They eat lots of cod, halibut, salmon, the types of fish that you would normally eat when you are in the North Atlantic.

    It is a homogeneous population.  It is a wealthy, developed, industrialized country with scandinavian background, with what you would call socialized medicine, equal access to social support and health care.  In studies we have conducted there, we have had a high participation rate.

    [Slide.]

    Now, this study has been in international collaboration between Faroes and Danish researchers, researchers from the U.S., Sweden, Japan primarily.  So far we have looked at three cohorts, and here, I am going to talk about, first of all, Cohort 1, I will talk a little bit about Cohort 2, and we have some preliminary findings on Cohort 3.

    [Slide.]

    First of all, the conclusions so far from these studies have been that we see neurobehavioral adverse effects associated with developmental methylmercury exposure.  We see that that exposure is also associated with increased blood pressure, poorer heart rate control, and also decreased growth of all physical growth postnatally, and we see that the prenatal exposure is much more closely associated with these adverse effects than the postnatal.

    The preliminary results from the follow-up at age 14 of Cohort 1 are in agreement with the results that we saw at age 7.  This is not published, so I can only give you the preliminary results.

    [Slide.]

    When you do research on this area, there are a couple of very crucial issues.  One issue that Dr. Jacobson touched upon was the validity of the exposure estimate.  If you do a regression analysis to assess the effects of an exposure with regard to some effects, your basic assumption is that the exposure is measured without error, it is a precise measure, but there is no such thing as a precise exposure measurement because what you would like to know is how much methylmercury is there at the target, let's say, of some particular part of the brain.

    We will never know that, so anything else that we are using is a proxy.  That means we are going to have an underestimation of the true effects of methylmercury.  Now, these are the exposure biomarkers that we have used, and I will talk a little bit about those.

    [Slide.]

    First of all, you have to have as precise a result as possible from the layout, and I will show you briefly on the next slide what I mean when I say that.  The other issue here is that the timing of the sample has to relate to the toxicokinetics of behavior of methylmercury in the body, and you also have to consider the characteristics of the specimen, and particularly that is a problem with hair.

    Hair varies a lot between people, and hair structure or hair treatment even varies a lot, and that causes uncertainty.  Finally, the bottom line is obviously the predictive validity, which one correlates the best with the outcomes that you are looking at.

    [Slide.]

    So, here are the issues in regard to the laboratory validity.  I think we have done as well as we can possibly do with modern atomic absorption techniques and supporting methods.  The chemists told me that the imprecision of the microanalysis, it should be better than 5 percent.

    So, being confident environmental epidemiologists, we thought, oh, wonderful, we have an exposure estimate which has an imprecision of about 5 percent, that is great, we are in a fabulous position here.

    Well, I will tell you a little bit down the road that this was a naive assumption.  These exposure biomarkers are not all that precise.

    [Slide.]

    Here is one issue, though, you have to consider, and that is a variability of exposure.  Now, the colleagues in Rochester looked at hair from the Faroes that they chopped into segments, and we followed up on that issue, and we have a total of 21 long hair samples where the average was above 10 ppm that we had chopped into these segments, and these are the three that showed the greatest variability with a coefficient of variation of about 25 percent.

    Actually, you can see there is only one of them where there is a definite clinal tendency.  It is only 1 out of 21 samples, but anyway, this kind of temporal variability will mean that you will have an imprecision associated with just about any exposure biomarkers that you choose, simply because there is variability during pregnancy.

    Now, we have chopped this into segments of 1.5 or 1.1 cm simply because the mercury half-life is about 45 days, so each segment corresponds to a half-life.  That is the reason for doing this.  We have also looked at profiles with colleagues in New York, trying to see if there were shorter term variabilities.

    We compared hairs from the Seychelles and the Faroes.  I have reservations about the technique that they use, because the results, in my view, were not all that reliable, but the conclusion from that study was that the profiles were indistinguishable between the Faroe and the Seychelles.

    [Slide.]

    The first biomarker we looked at in this regard was the long hair sample of 8 or 9 cm from the women, that the hair was taken at parturition, at the time of childbirth, and this graph essentially shows you the contribution of mercury exposure during pregnancy and before actually to the hair mercury concentration in that particular sample we obtained at childbirth, which is here indicated as Week 40.

    There is a lag time because some of the most recently observed methylmercury will still be in the hair root, so it will not have made it into the maternal hair at the point where we sampled the hair, but then you can see that this type of a sample will overestimate the importance of the first and second trimesters of pregnancy.

    We then, for about 600 of these kids, we were able to obtain from the remaining hair sample, if there was any hair sample or remaining after the first analysis, we were able to cut off the proximal 2 cm hair sample, the one closest to the root, and you can see from this representation that that would better reflect the end of the second and the early part of the third trimester methylmercury exposure.

    [Slide.]

    When we compare them as predictors of the outcomes that Dr. Jacobson also focused on, the finger tapping, the attention, the Bender/Gestalt, the Boston Naming, and the California Verbal Learning Test, you can see that there is perhaps a slight tendency that the proximal hair sample is better than the long hair sample, but these small differences are by no means significant, also because we are losing power as we get from the 900 to the 600.

    [Slide.]

    This slide shows you the correlation between the two hair mercury concentration measures.  I have indicated the ones with open circles where the coefficient of variation is more than 25 percent.  This is about 10 percent of the samples where the coefficient of variation based on those two measurements only was large.

    So those must have been the individuals where the mother had a variable methylmercury exposure during pregnancy.  Now, the interesting thing is then going to be what happens if we remove the 10 percent of the kids who had a variable methylmercury exposure during pregnancy.

    [Slide.]

    This is what we did.  We used the third exposure biomarker, namely, the cord blood measure as the independent judge.  Is there any difference between stable and variable or mercury exposure, and indeed the bottom line is that if you remove the ones with variable methylmercury exposure, you get an increased power.  It is easier to see the mercury exposure simply because you eliminate one source of imprecision.

    So, on the other hand, this also indicates to us that variable exposure or peak exposure cannot be the reason for our seeing that there is an association with the cord blood measure.

    [Slide.]

    We also did a separate analysis, let's say, a neutral statistical analysis to see what are the uncertainties involved here if we have absolutely no other assumptions other than the three exposure measures must be in some way a measure of some sort of true mercury exposure that we don't know.

    Each of them has to be a reflection associated with some error indicated by the epsilon on the right of this equation.  In order to solve this equation, you have to have three sets of equations, and then you can do a factor analysis.

    So, we used the cord blood measure, the long hair mercury measure, and there is a third independent variable, the dietary questionnaire information, namely, how often have you eaten pilot whale for dinner during the pregnancy. So, when we fed that into the computer and calculated the overall epsilon for each of those three biomarkers, then, this is what we find.

    [Slide.]

    We set the loading factor for the cord blood to 1. You can see that the two other parameters are less good indicators of the true mercury exposure defined as the best, let's say, background that can be calculated from this imprecise information that we have here.

    But the important thing is that the coefficient of variation is, number one, 30 percent for the cord blood mercury.  This is much more than what the chemists told us, much more than the 5 percent.  The other important information is here that hair mercury is much more imprecise than the cord blood is.

    This is not based on any toxicokinetic information.  It is simply based on the concrete numbers for each child with regard to these three exposure variables only, nothing else.

    [Slide.]

    So how come a hair measure is more imprecise than a blood measure?  Well, there are some issues involved here like I referred to before, that hair is not just hair, it varies between people, and there are several issues that you might want to consider, and it was actually done in the NAS report.  They looked at this very carefully.

    [Slide.]

    I am showing you again this hair curve because I wanted to compare with the cord blood.  The cord blood is, of course, obtained at the only time you can obtain it, at the time of parturition.

    [Slide.]

    Here, you actually have a representation of the last trimester.  If we want to consider cord blood in regard to the predictive validity, you have to think of the windows of vulnerability here - is it important to have a representation of the last trimester?

    People who work in developmental psychology, like Dr. Jacobson, would say the third trimester is certainly very important with regard to the programming of the brain.

    [Slide.]

    These are the results that we published in 1997, but what I have done here is to compute the regression coefficients as percent of the standard deviation for each outcome variable associated with a doubling of the mercury exposure.

    The doubling is obviously you take the regression coefficient for the log transformation and multiply by 0.301, and then you divide by the standard deviation that is present in percent.  You can do it yourself from the regression coefficients, but this way you can actually compare the results from the different domains, and you can see that it is really attention and language that appear to be the most sensitive.

    Here, we are in a way going beyond the psychometric properties of each of these tests.  It looks like attention and language are the modalities that are affected the most.

    The important thing is here, blood is a much better risk indicator than is hair.

    [Slide.]

    Conclusions on these issues is that the cord blood is the best risk predictor, but it is still imprecise.  We still have to consider this 30 percent imprecision, and I will get back to that.  The more imprecision we have, the more we underestimate the true extent of the effect.  There is nothing new in this, this is in perfect accordance with everything we have learned in the past from lead and many other situations.

    [Slide.]

    Now, if we go back to the regression coefficients for blood, we can actually do a sensitivity analysis and adjust for the 30 percent imprecision, and this is what I have done here.  I am giving you the regression coefficients before, adjusted, and then you can see the increase.

    Overall, it looks like for each doubling of the mercury exposure, you lose something that corresponds to about 10 percent of the standard deviation, perhaps a little more for attention and language, about 10 percent for these sensitive modalities.

    [Slide.]

    Now, I am going to talk a little bit about the outcome variables, because that is a second issue that you have to consider.

    In regression analysis, you do take into account that they are imprecise, but there are psychometric issues that are important because some tests are very useful and some tests have a lot of noise involved like they may depend on the child's motivation or the testing situation or the testing situation, the training of the tester, or whatever.  Many variables may play a role here, so you have to consider these tests very carefully and also the age of testing.

    [Slide.

    These are the criteria that we use for the selection of clinical tests.  Unfortunately, the Faroes is a scandinavian society, so what we did was to apply tests that are also applied in Denmark and Norway and other countries like that, also the United States, but we did use internal age standardization and we piloted the tests, we translated them, of course, and made sure that they functioned in that society before we went ahead.

    [Slide.

    Now, let me just point out a couple of concerns that we have about outcomes like this.  The first issue I wanted to bring out is that it is important that a test has as many possible outcomes as possible meaning that if you do the clinical tests of catching a ball, it is something that pediatric neurologists do, they throw a ball in the clinic and then the child has to catch it a distance of 10 feet, and the ball has to have a particular size.  Either the child catches the ball or the child fumbles or the child doesn't catch the ball, so you have only three outcomes.

    This is a very simple test, and it shows an association with mercury in the right direction, but the p-value doesn't prove anything, but if you use tests that reflect attention and which have an increased number of possible outcomes, then, the digits bend forward score is better and the computer assisted reaction time is even better than that simply because it is better to dissociate within the patterns of gray, because we are looking for some subtle, we are not looking for sick kids, we are looking for something subtle.

    [Slide.

    The second issue here is that some of these tests are I would say they are complicated to do the same way every time.  They have to be done in an extremely standardized way, and you sometimes run into trouble if you have technicians or nurses do the test, and we have emphasized that we wanted the highest possible expertise, board-certified staff to administer the tests, and we saw a clear example when the similarities tests simply could not fit into the neuropsychologist's time allotment.

    We had to move the test to another examination station where a technician did it, and when we looked at the kids that the neuropsychologist had examined, there was a mercury effect, but the result that we published in 1997 was the bottom line where we used the results from both stations and adjusted for the examiner and then you don't see a mercury effect.

    The question is if it is not more reliable to look at, even if it is a smaller number, than to look at the kids that were examined by the neuropsychologist.

    [Slide.

    Let me just say a few words about the outcomes at age 14.  These results have not even been submitted for publication yet.  We are still grinding confounder adjustments through the computer, but let me give you one which is reasonably simple to explain and one of the tests that Dr. Jacobson also mentioned, the reaction time measure using the NES continuous performance test.

    We actually used a revised version with animal silhouettes because at age 7 we were not sure that all the kids all knew the letters--and that is a standard version for adults--equally well, but we were quite sure that they all knew animal silhouettes.  So, instead of the letters, we used five different animals, and the kids had to react to the.  You see this cat on the screen here.

    So, the only difference between 7 and 14 years is that at 14 years, the test was extended to last for 10 minutes rather than 4, which was the time we used at 7 years.  Now, this was administered by the neuropsychologist.

    [Slide.

    These are the results.  These are the correlation coefficients that are not adjusted for confounders, but what you see here is that cord blood is still a significant predictor of the outcome 14 years later, and it is actually better than maternal hair and also better than the postnatal measures.

    At age 14, it turns out that the kids' exposure at that time actually correlates pretty well with what the mother had 14 years before, possibly an indication that dietary habits are quite stable within families at particular locations.

    So, when we do a multiple regression analysis, we try to leave in as many predictors of mercury exposure as possible.  It turns out that the cord blood microconcentration is the only one that remains.

    [Slide.

    The important thing with this slide is that the beta for age is almost the same as the beta for mercury, which means that if you increase the mercury exposure by a factor of 10, then, the child has a result which would have been similar to the situation had the child been one year younger.

    In other words, if you increase the mercury exposure by 10-fold, then, the effect is similar to losing one year of development.

    [Slide.

    Now, let me just say a few words the Cohort 2.

    DR. MILLER:  Dr. Grandjean, you have five more minutes.

    DR. GRANDJEAN:  Okay.  I will run through this quickly.

    Cohort 2.  These results were published in the Journal of Pediatrics.  These are the results for age 2 weeks.

    [Slide.

    This is PCB.  Dr. Jacobson talked about this.  We have looked into PCB.

    [Slide.

    And we have validated the cord tissue PCB and even if we assume that there is a large imprecision in the PCB measure, PCB is not a significant determinant of the outcome in this study.

    [Slide.

    These are the results from the paper published last year that shows that PCB has an effect in this population, but only in the kids who have a high mercury exposure at the same time.

    [Slide.

    These are results of brain stem auditory-evoked potentials from two locations, the Faroes and Madeira, and we see that the evoked potentials increase in latency, and the increase is similar in the Faroes and in Madeira.

    [Slide.

    The results of brain stem auditory-evoked potentials used for calculation of benchmark doses are very parallel to the results that the National Academy came up with.  You can see that the results for Madeira and the Faroes are quite similar to the neuropsychological benchmark doses published by the National Academy.

    [Slide.

    These are the blood pressure results.  You have the publication, so I won't dwell into that.  I will just say that this is an effect which is seen below the current reference dose that the EPA has decided upon.  We don't know the implications yet, but I am just saying this is an effect which is seen in very low exposure levels.

    [Slide.

    These are unpublished data on Cohort 2 where we show the weight at age 18 months adjusted for confounders. We see that kids with the highest mercury exposure actually weigh about 1 kilogram less than kids with the lowest exposure.

    [Slide.

    The bottom line of all of this is how do we translate this to public health.  I have already said that you can compare this to the age, the effect of age on development, and the result that we see is that for every time you double the mercury exposure, the child loses some months in its development.

    Ten percent of the standard deviation had this been an IQ, it would have been 1.5 points of IQ, so our results would translate to a loss of about 1.5 IQ points every time you double the exposure.

    Thank you very much.

    DR. MILLER:  Thank you.

    Questions from the committee?  Dr. Dwyer.

Questions of Clarification

    DR. DWYER:  I was wondering if there would be any effect of alcohol on absorption of methylmercury.

    DR. GRANDJEAN:  The Faroes is a very traditional society where men drink, but women don't, and it is a very small percentage of women who have at all touched alcohol during pregnancy.  The Faroes have the lowest alcohol use in I think all of European countries simply because it is more traditional.

    DR. RUSSELL:  I wonder if you would clarify one thing for me.  Is the level in codfish caught in the Faroe Islands likely to be the same as caught in Cape Cod, for example, or another geographical location?

    DR. GRANDJEAN:  I don't know what the level is here, but in the Faroes, the mercury content of cod is about 0.1 ppm.  Does that make sense?

    DR. RUSSELL:  I guess what I am getting at is whether there is large geographical distribution, wide variation in mercury levels in a particular species of fish depending on the geography of where it is caught in the Atlantic or the Pacific.

    DR. GRANDJEAN:  I am not an expert.  All I can say is that the main source of exposure is pilot whale, which overlaps with swordfish and tilefish, et cetera, but the average is higher.  It is about 2 ppm.  The Faroese eat it both as chunks, as steaks, and they also eat it like pemmican, like little slices, and that exposure is sort of diluted because they eat a lot of fish at the same time, so they may have some peak exposures now and then, but then they also have a background from the pemmican and the fish.

    DR. LEE:  I was wondering if you can give me a feel for how much the mercury in the hair comes from dietary versus environmental exposure.

    DR. GRANDJEAN:  What do you mean when you say environmental?

    DR. LEE:  Well, for example, polluted water or smoke.  I mean if I am being exposed to mercury via polluted water, will it get into my hair?  What kind of exposure can I expect from that?

    DR. GRANDJEAN:  It is possible that the hair and also the blood may contain some inorganic.  In the speciation that we have done, more than 90 percent is methylmercury, and not inorganic mercury.

    Methylmercury would, as far as I know, come from marine food or freshwater fish only.  I am not aware of any other important sources except perhaps if there is some internal methylation of inorganic mercury, but I think it would be an extremely small contribution.

    DR. LEE:  So, you are saying all the mercury that I would find in my hair would be from food sources?

    DR. GRANDJEAN:  I would think so, but there is a possibility which has been seen in various instances that hair might absorb inorganic in particular mercury from outside sources.

    DR. FISCHER:  Dr. Grandjean, tell me or tell us, knowing the exposure to PCBs of this population that we have studied, would you expect that the levels of PCBs would allow a contribution of those substances to the effects that you are attributing to methylmercury?

    DR. GRANDJEAN:  It is a very good question and perhaps Dr. Jacobson might also contribute to this.  The difference between the evidence that we have on PCB and the evidence we have on methylmercury is that the PCB studies that have been carried out in North Carolina, with the Great Lakes, and in the Netherlands, have not been adjusted for methylmercury, so we don't really know what the methylmercury contribution to the PCB associated or attributive effects might have been.

    In the Faroes at least, we have measured the PCB both in Cohort 1 and Cohort 2, and we have been able to adjust for PCB, and it is very hard to see what the PCB contribution is in these cohorts because the mercury effect is so strong, so we were not able to discern any clear PCB effect.

    It doesn't mean that PCB is not neurotoxic in the Faroes.  It simply means that it is less neurotoxic than methylmercury at the levels that we have in that population.

    DR. FISCHER:  And the method you use to adjust for the PCBs, would you explain that to us?

    DR. GRANDJEAN:  We did regression analysis that have been published.  We also have a paper in press where we used structural equation analysis, and the structural equation analysis indicates that even if we assume that our PCB measure is vastly imprecise, it never reaches a level of statistical significance of 0.05.  It simply doesn't become significant, but mercury is.

    DR. DWYER:  Have you ever looked at meconium in the fetus?

    DR. GRANDJEAN:  No.

    DR. MILLER:  Dr. Friedman.

    DR. FRIEDMAN:  Is there a reason why you looked only at cognitive outcomes, cognitive/achievement outcomes, and not at social/emotional outcomes?

    DR. GRANDJEAN:  We looked at the Child Behavior Checklist, but we had difficulty translating it into Faroese.

    DR. FRIEDMAN:  What was that measure that was mentioned?

    DR. JACOBSON:  Child Behavior Checklist.

    DR. FRIEDMAN:  CGCL, okay.

    DR. GRANDJEAN:  Which is the standard measure, and it is only currently being standardized into the language of Danish, and we tried to translate it and pilot it in Faroese, and the mothers simply had great difficulty answering these questions perhaps because of linguistic problems and perhaps because of cultural problems.

    This is a test that has been I think developed in New Hampshire?

    DR. JACOBSON:  In Vermont, but it has been standardized in the Netherlands and many European countries.

    DR. GRANDJEAN:  Anyway, we could not apply it.  We tried and we failed, and it was, in our hands, a very unreliable instrument and therefore we did not dare to go ahead with this.  I think you are right, that it is an aspect that should be looked into, but I think that one should not look at that aspect in the Faroes population without having something that is standardized, and we don't.

    DR. APOSHIAN:  Dr. Grandjean, would you say something about the amount of selenium in the diet in the Faroe Islands, Seychelles Islands in New Zealand, please?

    DR. GRANDJEAN:  We measured selenium in cord blood and the average concentration is I think about 30 or 40 percent higher than in the Danish population, clearly because the Faroes depend so heavily on seafood.

    We also looked at the mercury-selenium ratio as a predictor of these outcomes because both mercury and selenium were measured in cord blood, and the mercury-selenium ratio was not a better predictor than the mercury concentration as such.

    So, it looks like selenium deficiency does not explain the effects, nor does high intakes of selenium protect against mercury toxicity.  We have done the same exercise in Cohort 2 with essential fatty acids especially docosa-hexaenoic acid, and DHA prolongs pregnancy and mercury has no effect on birth weight.

    It could be that the DHA intake protects against effects on birth weight by methylmercury, because we have been unable to see any effects.  Likewise, we have not seen effects on visual-evoked potentials in the Faroes, perhaps because DHA protects the visual part of the brain.

    These are research issues.  I can't make any conclusions, but it is speculation that some of these essential nutrients in seafood are perhaps modifying, however, only slightly the mercury toxicity that we see.

    DR. RUSSELL:  I think in the Faroe Islands, the dietary fiber levels are fairly low because of a lack of a large amount of fruit and vegetables.

    Is there any evidence at all that dietary fiber alters the bioavailability?  It does for other metals.  I have no idea about mercury.

    DR. GRANDJEAN:  I think the difference between methylmercury and other metals is that methylmercury is almost completely absorbed in the gut, and any interference due to dietary fibers would have a very small impact simply because of the lipophilic character of the methylmercury species.  Whether it might affect the bacterial environment, so that it might affect the methylation or demethylation in the gut, I can't say, but I am not able to make any judgment.  I would think, if anything, it would only have a minor effect.

    DR. KUZMINSKI:  Dr. Grandjean, is there, in your data, any way of backing out the effects of the whale consumption and isolating only the fish consumption, because in the comparison of the Faroese consumption of fish and whale compared to U.S. consumption, that seems to come out as the one big difference?

    DR. GRANDJEAN:  I think had the Faroese have exposure through drinking water, you would pose the same question.  We look at methylmercury as the toxic species, and it doesn't matter from where it comes.  I may be wrong, but sometimes a methylmercury concentration in the whale meat is like 0.5 or 0.8 ppm, which you would not consider high.  Sometimes it is as much as 3 ppm, and people who eat the whale meat don't know.  It is only after the methylmercury has been absorbed from fish or whole or shellfish, wherever it comes from, that we can detect it, and then we measure the blood or the hair or whatever.

    I would have sincere disagreements with you if you felt the methylmercury from whale meat would have any different effects from methylmercury from fish.  It is the same species.

    DR. KUZMINSKI:  Where I was headed, this is intuitively, was trying to ascertain an effect just due to fish consumption and not whale consumption.  It is not the whale mercury being different from fish mercury, no.

    DR. GRANDJEAN:  I don't think we can do it.  I also don't think that I understand what the scientific basis would be, but even if one would do it, I don't think it is possible to do simply because people who eat a lot of whale meat also eat a lot of fish, so it is very difficult to sort out where it comes from.

    DR. RUSSELL:  I have one final question on the diet.  Can you give us some feel for the percentage of calories coming in as fat in the Faroe Islanders?  In other words, is it more or less comparative to American fat intake or is it higher because of the blubber and the lack of fruit and vegetables?

    DR. GRANDJEAN:  I think fat intake may be more relevant because of the lipophilic character of the methylmercury.  People who would eat a lot of blubber might have a higher relative fat intake than people who eat less.

    The overall average in the Faroes has not been calculated.  I would assume it is similar to scandinavian levels, which are similar to U.S.

    DR. McBRIDE:  Do you have any information on birth weight?  Fatter babies might store mercury and have an exposure after birth.  Do you have any information on birth weight effects and their variables?

    DR. GRANDJEAN:  Number one, the Faroese have one of the highest birth weights in the world, and our interpretation is that they have such a high intake of essential fatty acids from seafood, especially DHA, and we have also been able to show that the high the DHA intake, the longer the duration of pregnancy.

    I mean most of these women have a pregnancy duration of 40 weeks, some of them even 41, so it is an effect of prolonged gestation.  When we tried to figure out if there is a mercury or a PCB effect, because this has been seen in some studies that these toxicants might affect birth weight, we don't see anything, but when we look at postnatal growth, we see that the prenatal mercury exposure affects the postnatal weight gain.

    DR. McBRIDE:  But does birth weight affect the outcome on your psychological tests?

    DR. GRANDJEAN:  It would have been if we had used kids who were also pre-term.  All of these kids were normal full term.

    DR. McBRIDE:  But I am not thinking of pre-term effects, I am really thinking of birth weight.

    DR. GRANDJEAN:  No, birth weight does not have an effect.  I mean all of these birth weights are above  2,500 grams, and we looked at it.  It doesn't affect anything.

    DR. McBRIDE:  So, I mean you looked at birth weight compared to outcome.

    DR. GRANDJEAN:  Yes, birth weight was included with other risk factors like previous history of skull trauma, history of meningitis, neonatal jaundice.  We looked at all these factors, and we didn't find an effect.

    DR. FRIEDMAN:  There is no way in your study, which is a kind of a natural history study, to disassociate the prenatal effects from later effects, right?  That is, the children continued to have high consumption of the same foods that their mothers had.

    Is there a way to know whether this could be reversible if the mothers had high consumption and then, after birth, it stopped, would the children look the same later on?  This is all hypothetical, but I guess we are talking a lot here about prenatal effects, and I am not sure if those are really just prenatal effects of cumulative effects over time.

    DR. GRANDJEAN:  There are two things here.  After the child gets born, the mercury content in the child would drop way down, because the supply from the mother would cease except for those kids who get breast fed.

    We see at age 12 months that there is a very clear association between the hair mercury concentration of the child and the duration of breast feeding, so those who have been breast fed for a long time have more methylmercury in the hair because the mother continues to contribute.

    However, the hair mercury concentration at 12 months is only about 20 percent of the average of the hair mercury concentration of the mother, so there is less mercury coming through human milk.  We can still see it, but it's less.

    Now, after weaning, it is going to take some time before a child start eating whale meat.  Usually, they will get other kinds of food before the mother will introduce fish or whale meat.  At age 7, I don't remember the percentage of children who had started eating whale meat, but it was not a majority.

    So, it is only when you get up to age 14 that they are closer to eating the adult diet.  We have two issues involved here, namely, one, we have a scale of mercury, potential mercury exposures, and then we also have a scale of susceptibility that we have to take into account.

    It is going to be very difficult to figure out how these two different factors play a role.  The only thing we can do is to do multiple regression analysis and also structural equation analysis, and the cord blood measure is way, way, way strongest predictor of these outcomes.

    DR. MILLER:  One more question and then we have got to move on.

    DR. APOSHIAN:  Did the Faroe Island studies separate the effects of breast feeding and no breast feeding as far as the domain results were concerned?  As a confounding factor, in other words.

    DR. GRANDJEAN:  We did two things here.  We looked at milestone achievement during the first year of life, and it is very clear that there is an advantage to being breast fed that overrides the possible adverse effects of getting methylmercury from breast milk.

    We have also looked at the outcomes at age 7, and there is an advantage associated with having been breast fed for a long time.  There were very few mothers who didn't breast feed at all, so I can say is there is an association with the duration of breast feeding, and the duration of breast feeding is not associated at all with the exposure level, neither mercury nor PCB, there is no association here, so it is not really a confounder, but there is a small advantage that we can see.

    This is a paper which is going to be submitted very shortly, and I don't quite remember if there was a difference in the domains.

    DR. APOSHIAN:  Along those same lines, and this may not be a fair question, maybe I should wait for Dr. Myers, is there a difference in the length of breast feeding of a child in the Seychelles Islands versus the Faroe Islands, do you know?

    DR. GRANDJEAN:  All I can say is that kids in the Faroes are being breast fed much longer than kids in scandinavia.  They do not live up to the World Health Organization recommendation, as nobody does, so breast feeding is the rule in the Faroes, and we see an advantage associated with it, and the duration is not associated with the exposure level.

    DR. MILLER:  Thank you very much.

I will take this opportunity, another member of the committee has shown up, Ms. Halloran from Consumers Union. Welcome.

    The next speaker is Dr. Gary Myers of the University of Rochester to talk about the Seychelles study.

Seychelles Study

Dr. Gary Myers

    DR. MYERS:  Thank you very much for inviting us here to present our study.

    I would like to go through the Seychelles study with you and then answer whatever questions I can.

    [Slide.

    This is the hypothesis that we have been addressing in the Seychelles study - whether or not prenatal exposure to methylmercury from maternal fish consumption during pregnancy can adversely affect children's developmental outcomes.

    [Slide.

    This hypothesis actually came out of work that we did in Iraq, which was mentioned earlier.  This is just one of the graphs from a publication that Dr. Cox was the first author in back in 1989, and this one looks at the frequency of retarded walking, and as was mentioned earlier, the endpoints in Iraq were somewhat less sophisticated that they were in our Seychelles study and as they have been in other studies since that time.

    Retarded walking was simply defined as walking before or after 18 months of age.  When one plots the abnormals versus the normals and does this hockey stick plot, if you will look at the top of the graph, there are all these little pluses, those are individual cases where the child had an abnormality of walking.  Along the bottom, the pluses are all individual cases of where the child was normal in walking.

    As you can see, there are a couple of things that come out of the graph.  The first is that if you project this lying downward, it looks like you might have effects down around 10 to 20 parts per million.  These are concentrations in maternal hair, which is, in fact, the biomarker which has been used by every other study, studying this issue, and for reasons which I will try to address briefly in a moment.

    So, that is one thing.  You see that down around  10 to 20 parts per million, one might expect to have some effects.  The other thing that you see is that when you look at the top, there isn't any data or very little or very few points of data below about 50 parts per million.

    [Slide.

    Following that, we actually proposed this hypothesis that these lower levels in fish might actually have some sort of adverse effect, and we looked at the literature and we came to some interesting conclusions.

    First, it seemed pretty clear that the fetal brain was much more sensitive than the postnatal brain to the effects of mercury.

    The second was that it looked like from the neuropathological studies and all of the other clinical things that had been done previously, and which have, in fact, been done since, that the effect really should be global.  We couldn't see a reason why it would be domain specific from our review of the literature.

    We decided that if there were going to be any effects from the consumption of fish at these low exposures, they would probably be subtle effects.  We wouldn't expect any of the major things that were seen at Minamata.

    Just an aside about Minamata, one of the interesting things about Minamata is there were either serious affected children or they were non-affected children.  Nobody really described this spectrum of decreasing morbidity.  Whether that was because it wasn't studied or because it didn't occur has never been clear, but it still hasn't been described from the Japanese experience.

    The other thing is that we thought that peer analysis was really an excellent way of looking at exposure and, in fact, we have subsequently looked at neuropathology in relationship to hair mercury concentrations, and they correlate better than fetal blood in our pathological specimens, and there is some evidence--and Dr. Clarkson will be down in a day or two and perhaps speak more eloquently to this--that the transport mechanism into hair for methylmercury is much more akin to what happens in mercury getting into the brain.  Of course, the brain is the target organ that we are all worried about.

    The last thing is that actually, you ought to be able to detect these effects early on.  After all, the Iraq study, the average age of the children studied in Iraq was 30 months, so waiting five or six years didn't really make sense to us.

    When Dr. Marsh talked me into going out to Seychelles and starting this study, he told me, he said if you don't find anything in six months, you probably won't find anything.  It turned out that wasn't necessarily true.

    [Slide.

    Anyway, we looked for a population that had large consumption of fish.  We actually started several studies before we ended up in Seychelles, and there were a variety of reasons why the other studies didn't work out, but it is very difficult to set up these studies.  Dr. Grandjean is fortunate to have the Faroes, and we are fortunate to have the Seychelles.  There are many places in the world where these sorts of detailed studies simply can't be done.

    For those of you who are not familiar with the Seychelles, it is where the three red lines come together there.

    [Slide.

    These are some of the characteristics of Seychelles, and I will be glad to elaborate on them later if anybody wants to ask, but they have a high fish consumption. Dr. Grandjean presented some evidence earlier that in Faroes, they have three fish meals a week.

    Well, when I started the study there to gather the main cohort, we asked a question how many fish meals do you eat a week, and they told us 12 was the average, so between 10 and 12 fish meals a week is the average in Seychelles.

    It has been a socialistic state for quite some time, for the last 30 years.  They have free universal health care.  The infant mortality is lower than in the U.S. They have a 98 percent immunization rate, which is better than the U.S.  You wouldn't want to get sick there, but actually, the preventive care is excellent.  Free universal education.  All the child start in the educational system at age 3 1/2, and it goes on from there up into the teenage years.

    They have really very limited poverty.  There is a social structure, but it's very compressed.  There is literally no malnutrition on the island, and they quite low levels of other sorts of contaminants.

    [Slide.

    This is downtown Victoria, which is really the only major city on the island.

    [Slide.

    This is just the fish market.  People go the market every day and buy fish or they buy it on the roadsides or the beaches, but the people eat large quantities of quite fresh fish.

    [Slide.

    We have looked at PCBs in Seychelles, and they are really below detectable limits.  We have looked at lead, and lead levels are quite low.  We haven't actually measured PCBs, but we are told that the levels of pesticide exposure are substantially below the FAO Codex Alimentarius reference levels for a problem.

    [Slide.

    So, as best we can tell, the other sorts of exposures that one might be exposed to are very low levels.

    Let's go back to that slide a second.

    We measured prenatal exposure in maternal hair.  This gives you really a very excellent index of exposure for the whole pregnancy as opposed to just the final trimester. We have not measured cord blood in Seychelles.  The exposure averaged about 7 parts per million and ranged from below 1 part per million up to about 27 parts per million.

    [Slide.

    This gives you some idea.  We adopted a longitudinal design to the study which I will show you in a moment, but this gives you some idea of the numbers of children that have been seen at each one of these ages, so we have actually examined these children on five occasions now.

    We have been able to maintain quite a substantial number of the cohort.  It is a small island, there is not a great deal of the population that emigrates, and it is easy to find them.

    We have excluded individuals for various reasons.  Among the 39, the major reason for exclusion was that we really couldn't recapitulate their mercury exposure. When we looked at the hair samples that we had, we simply couldn't recapitulate their exposure, but we have also removed from the cohort, a few individuals who had perinatal seizures, intraventricular hemorrhage, substantial head trauma, and other things that are known to be highly correlated with abnormal children's development.

    [Slide.

    We looked at a variety of covariates in our population.  We have looked at socioeconomic status, IQ.  We have been to every one of the homes in Seychelles and assessed their home environment with the Caldwell Bradley assessment of the home.

    We have looked at maternal smoking and alcohol. They are extremely low in Seychelles.  We have looked at the medical history of the mothers, and we have also looked carefully at the language spoken in the home.  About 98 percent of the people there speak Creole, so the vast majority of them actually have Creole.

    [Slide.

    This is simply one of our testing rooms and one of our testers administering the KBIT for maternal intelligence.

    [Slide.

    This is a typical home in Seychelles.  The two women on the left are the nurses who were in the home doing the home environment, the HOME Scale.

    [Slide.

    We have looked at a number of covariates that affect the children.  We have looked at gender obviously, because that is a significant thing.  We have looked at hearing in the children.  We have looked at their health history, their birth weight, gestational age, birth order, length of breast feeding, and a variety of other things.

    [Slide.

    This is the design of the study.  We started the main cohort.  We did have a pilot cohort, as Dr. Jacobson mentioned earlier.  The pilot cohort was done originally by myself before we started the main cohort, and then I went to Seychelles and lived there for a year to enroll the children at the six-month evaluation, so I did all of the Denvers and Fagans and neurological examinations at that point in time.

    Subsequently, we have used a battery of Seychelles professionals who have done the testing for us.  We have now examined the children, over 700 children at each one of these five points, and the list of test is shown there.

    I did put down in the corner there, there has been a double-blind study just to remind myself to mention to you that from the beginning, we have never shared mercury levels with anyone in Seychelles, nor with any of the investigators who are clinically looking at the data.  So, it has been double-blind from day one, which goes back to about 1987. The Seychelles have been very cooperative with that.

    They have reviewed all of the data that we have published and made their own decisions about their choices in terms of regulation.

    [Slide.

    We have felt that the most important thing is not so much the tester, although we have used professionals, but doing reliability on the testers to be sure that, in fact, they are reliable.  So, we have used two types of reliability.

    First, we have used what we call the gold standard.  The gold standard is one of our psychologists from Rochester, a Ph.D. psychologist, who goes to Seychelles, sits down with the tester, and actually scores the test while the tester is administering the test.

    Then, we have used interscore reliability, which is each week we have two of our testers score the same child independently, and we have compared them.  We have looked carefully at those statistics, and they have had very high correlation.

    [Slide.

    This is Dr. Davidson here on the right, one of our nurses, is simply doing a gold standard.

    [Slide.

    When we looked at our results during infancy, this is what we found.  We did find the expected effects of covariates - maternal intelligence, birth weight, and other things.  We had most r-squareds for our study.  This is consistent with what has been found in most developmental studies.

    We did not find any adverse associations between the prenatal exposure and any of our endpoints at 6, 19, and 29 months.  We did one association between an item from the infant behavior record and methylmercury, and at 29 months, on the infant behavior record, there was a decrease in the examiner's scoring of the activity level, and it was present only in boys.

    We have been confused as to how to interpret that. It is a very subjective endpoint, the infant behavior record, and we are not sure whether it is better for boys to be less active or more active.

    [Slide.

    We like data.  We like to see the points and share them with people, and let them know what, in fact, the data looks like.  This is the visual recognition memory on the top and the visual attention on the bottom.  This is data from the Fagan test, which is I am told by psychologists felt to be one of the premier tests for intelligence at these early ages.

    This is at the 6-month examination, and there was no association with mercury within the range we have been studying.

    [Slide.

    This is data from the 29-month examination.  This is the metal developmental index from the Bailey, at 19 months on the top and at 29 months on the bottom, and that is on the left.  On the right is this infant behavior record.  You can see that in girls, the slope is flat, and in boys, it tails off as one goes to higher mercury levels. That is the association that we are not sure whether it is a beneficial or adverse association.

    [Slide.

    When we get up into early childhood, again, we find the effects that we would expect from a variety of different covariates.  Again, most r-squareds.  Again, no association between exposure and the endpoint.  We did find one beneficial association

    [Slide.

    This is the McCarthy GCI adjusted for covariates, and these are partial residual plots.  Again, we like data. This is the 66th month examination all plotted against prenatal methylmercury exposure, and there is no significance there.

    [Slide.

    This is the PLS language score adjusted for covariates at 66 months.  There is an association, but it seems to be a positive association here, we are not sure what to make of that, but no adverse association.

    [Slide.

    This is the Woodcock-Johnson applied problems, which is the mathematical part of it.  Again, no association at 66 months.

    [Slide.

    This is the Bender.  What you see here is that this is the errors on the Bender test.  What you see is that males are flat and in females, there is a slight downward slope meaning fewer errors.  This is one of the ones where when it goes down, there are fewer errors, so that is a beneficial effect.

    [Slide.

    At 107 months, we examined the children again. This is the nine year evaluation, which we are in the process of publishing.  Again, we looked at socioeconomic status, maternal IQ, age, family status, health history, and the home environment.

    [Slide.

    For the child, we did use postnatal methylmercury, age of testing, gender, hearing level, and examiner.  We have included examiner in these analyses, as well.

    [Slide.

    This is just one of the tests.  I just have a couple pictures of the tests.  This is finger tapping.  It is how many taps you do in a certain period of time.

    [Slide.

    This is a grooved pegboard using either preferred or non-preferred, and these are little pegs that have a little notch in them, and you have to fit them into holes. It is how quickly one can do the test.

    [Slide.

    This is a picture completion.  It is just a series of pictures, and you have to put them together in a story from the WISC.

    [Slide.

    This is from the Woodcock-Johnson test.

    DR. MILLER:  Dr. Myers, you have five more minutes.

    DR. MYERS:  Okay.

    [Slide.

    When we looked at the results from the 107th month evaluation, again, we found the expected associations with covariates.  Again, modest R-squareds.  Out of the 21 endpoints that we examined, we found 1 adverse association and 1 beneficial association.

    [Slide.

    This is the Connors Teacher Rating Scale, the hyperactivity index from it.  We had every teacher of all of our main cohort children evaluate the children on this scale, and the line goes down, which is a beneficial effect.

    [Slide.

    This is our adverse effect here.  This is the grooved pegboard, and it is of the children using their non-preferred hand, and what you see is that in females, the slope is essentially negative, it is not significant, or flat and nonsignificant.

    In males, one sees that there is a slight upward slope here, and that means that it takes them longer to do the grooved pegboard, so that would be an adverse effect.

    So, out of 21 endpoints, there is 1 adverse effect we found at 9 years of age.

    [Slide.

    This is just a graph to show you the effect of covariates on the various tests here.  We found a lot of associations with covariates, so we have every confidence that the tests are working there.  They are picking up other things that we know affect child development.

    [Slide.

    So, our outcome so far is that we found a lot of associations between predictors and between covariates of the endpoints at every age.  We have not found an association with mercury exposure from the fish consumption.

    I didn't mention, but the people in Seychelles do not eat whales at all, they don't eat sea mammals, it is purely fish consumption.  All of these associations were in the direction that one expects.  We have only found one adverse association in our five evaluations.

    That raises the question of how do you interpret it.  Well, it is good to remember that we were the group that originally raised the issue from Iraq and proposed the hypothesis, so we like to think that, you know, our studies have become more sophisticated since that time rather than that we have lost our ability scientifically.

    So, we look at other interpretations.  One interpretation may be that the exposure level is simply below the toxic threshold.  Another possibility, though, that we are currently exploring in Seychelles is that there is neurotoxicity, but somehow it is modified at these levels.  Either there is something beneficial about fish that overcomes whatever the toxic effect is, or maybe there is something else in fish that is mitigating that, either fatty acids or selenium or something else.

    We are looking currently at a new cohort that is being studied very carefully from the time the women are first pregnant, looking at these nutritional factors.  The other possibility is that perhaps there is toxicity, but it simply doesn't occur until much later in life.

    [Slide.

    We like to think that the Seychelles is a bit of a sentinel population for the U.S.  One reason is that the source of exposure is about the same as what we have here in the U.S.  It is really open ocean fish.

    The second reason is that we have looked carefully at the mercury content of the fish in Seychelles, and it is very similar to what is on the market here in the U.S., but, in fact, the women's hair levels are between 10 and 20 times those of the U.S. levels.

    So, we like to think that perhaps Seychelles could serve as a sentinel population for the U.S.

    I would be happy to answer questions.

    DR. MILLER:  Thank you.

    Comments, questions from the committee?  Dr. Friedman.

Questions of Clarification

    DR. FRIEDMAN:  Dr. Myers, I was wondering whether--I am sure you looked, but only didn't present--at interactions.  Do children of poor quality home environments show different outcomes relative to the effects of--

    DR. MYERS:  We have included socioeconomic status in all of our analyses.

    DR. FRIEDMAN:  You controlled for everything.

    DR. MYERS:  Yes.

    DR. FRIEDMAN:  But I am asking whether if you took out the control of the HOME, for example, and then looked at the children who are high on the HOME versus low on the HOME, would you find the same relationship holding in the two extreme groups.

    DR. MYERS:  We have looked at social effects, and there are some.  I am not very good at this.  Dr. Cox, who is in the audience, may recall the social effects better than I.

    Chris, the question has to do with the social effects on the outcomes.

    DR. FRIEDMAN:  The interaction.

    DR. COX:  I will do my best.  As I heard it, the term was interaction, so the way I would interpret that is to ask whether the effects of mercury are modified by levels of other variables, for example, socioeconomic status.

    Is that the question?

    DR. FRIEDMAN:  I guess so.  Let me try to phrase it.  I realize there are many things that operate to produce the outcomes that we see on psychological testing, and what we are trying to do with statistical analysis is to clean out the effects of variables that we are not interested in and ask if this was an experimental design, would mercury have an effect, but as I said before, we realize different things work together, and it may be that a high-quality family environment and high-quality out of home environment actually work against the ill effects of mercury.

    DR. COX:  That is modify the effect.

    DR. FRIEDMAN:  Right.

    DR. COX:  So, we are both saying the same thing.  Do levels of whatever factor you might want to look at change or modify the association between mercury and outcome, right?

    That is I think what is usually meant by the term "interaction."  Because of results in a study done in Canada, we felt there was some evidence for an interaction between gender and mercury, and that interaction was included in all the models that we looked at, and you saw an example of that here.

    Dr. Davidson, who unfortunately couldn't be here today, was also interested in interactions between mercury and other variables - socioeconomic factors, maternal age, what have you.  One problem was looking at such interactions, there is a very long list, and one can look at a great many interactions, and it is difficult.  You have to be careful, we all have to keep our hats on.

    I am a statistician, I am not a toxicologist, I am not a developmental psychologist, but my sense is that it is difficult to know what interactions one ought to look at.

    We have, however, done, to get to the answer to your question, we have done and published some analyses looking at interaction effects.  I can give you a reference to a paper if you are interested.

    I didn't know that that question would come up, so I can't summarize the results very well for you, but my own view is we didn't find anything that was very consistent, but we did find some evidence for differential effects.  It is hard for me to know what it means.

    I think that kind of question is very difficult. There has been some work done in the lead literature, as well, looking for interaction effects.  It is tough.  So, I don't think, besides gender, I don't think we have any very consistent evidence for modification of mercury effects by other variables that would be from the Seychelles study.

    DR. FRIEDMAN:  Thank you.

    DR. MILLER:  Dr. Nordgren.

    DR. NORDGREN:  This may be unrelated, but a recent area of controversy is the subject of autism, and I was going to ask also Dr. Grandjean, did you find any increased level of autism on the Seychelles or in your cohorts?

    DR. MYERS:  Well, the answer is no, but I qualify it by saying we have not done tests for autism.  In fact, we are in the process now of putting together a proposal to do that very study with detailed tests looking for autism in Seychelles to see if there is an increased incidence.

    Just from our casual experience, we have not recognized it, if there is, but that doesn't mean that it is not there.

    MS. HALLORAN:  Do you have information on the types of fish that were commonly eaten and what the mercury levels were, the range commonly seen in those species, and whether there is any season variation?  I am trying to get at whether there might have been peak exposures or it's a very constant exposure.

    DR. MYERS:  We have looked at hair analyses longitudinally, and there are season variations, but they are not like what we saw in Iraq.  They are down in the low range, and if they go up 30 percent, they go from 6 to 7 or 8 or 9, so it is within a very narrow range basically even though there are seasonal differences that we have seen.

    We have quite a bit of information about the mercury content of fish in Seychelles.  The majority of the fish eaten, probably the commonest species is the species called Karong, and the mercury concentrations in it are below a half part per million, down around 0.3.

    They eat a lot of reef fish, and the reef fish, some of those are some of the lowest concentrations that we have ever recorded in fish.

    DR. MILLER:  Can we let Dr. Dwyer ask a question?

    DR. DWYER:  Thank you.  Just two perhaps silly ones.  One is do you have any information on fetal wastage, and secondly, were there other prespecified interactions that you looked at.

    DR. MYERS:  As far as fetal wastage, we don't really have any information on fetal wastage, so I can't really provide anything on that.  We have thought about looking at it, but we have not done it yet.

    As far as other interactions, the only interaction that we have consistently had in our analyses is that for gender.  I think that is correct, isn't it, Chris?

    DR. COX:  Yes.

    DR. APOSHIAN:  Dr. Myers, as I remember from the Raleigh White House Conference and from the NIEHS, the recommendation was made that both groups standardize the neurobehavior tests or neuropsychological tests being used.

    The 107-month study that you say is in press or is about to be written, either one, does that have the same tests that were done in the Faroe Islands?  I wasn't quite certain about your abbreviations.  For example, the Boston Naming Test, was that done?

    DR. MYERS:  Actually, the tests are almost identical.  The Faroes were good enough to share their test battery with us, and we looked at the tests and decided what to do, and there is a great deal of overlap in the testing.

    DR. APOSHIAN:  The second question I wanted to ask was everyone, of course, is concerned about the difference between the Faroe Islands and Seychelles Islands.  Nowhere have I seen anything take into consideration racial, genetic makeups.  For example, for N-acetyltransferase, as I am certain you know, but let me just say for the others, that the American population is quite different from the Chinese mainland population, which is also different from the Egyptian population.

    Has any attempt been made to do genetic markings or genetic marker tests to see whether the differences between the Faroe Islands and the Seychelles Islands studies on methylmercury are due to difference in genetic makeup?

    DR. MYERS:  I don't know whether that has been done in the Faroes.  We have not done it in Seychelles.  As part of our current study, studying the nutritional components, we are measuring some DNA things, but not a wide range of them.

    I think there are a lot of differences between the Faroes and the Seychelles.  One is in cold water, one is in warm water, one is predominantly scandinavian, one is predominantly African in origin.  Their diets are vastly different really.  The exposure is different.  There are a whole series of things that differentiate the two populations in my mind.

    DR. MILLER:  Dr. Lee.

    DR. LEE:  Dr. Myers, your last slide indicated that we are eating some of the same seafoods here in the United States, yet, the maternal hair there is about 10 to 20 times the mercury than U.S. samples.

    Do you attribute all of that 10- to 20-fold difference to diet?

    DR. MYERS:  As far as we can tell, it is dietary, yes.  They literally eat fish twice a day in Seychelles.  Even given a choice, they love fish.

    DR. LEE:  So, that is corroborated by your direct measurements of the mercury in the diet?

    DR. MYERS:  No, that is not corroborated by mercury measurements in the diet.  We are currently doing that in a new cohort.  The main cohort that I have just been talking about, this was examined longitudinally five times over nine years.  That cohort was enrolled at six months of age, so we did not do things prospectively in that cohort.

    We had prenatal exposure because we had been collecting hair at antenatal clinics for a long time, so we had quite a bit of prenatal exposure data on all of these women, but we didn't have other things.  We didn't measure nutritional factors during pregnancy, which we are doing with this cohort.

    DR. MILLER:  Dr. Fischer.

    DR. FISCHER:  Looking for differences between the two studies, have you looked at the differences in preparation of the fish in each case?  I have no idea how the whale meat is eaten, for example.  Anyway, the possibility exists that there could be something there.

    DR. MYERS:  My only experience with whale was when I attended a conference there, and they eat it, like Dr. Grandjean said, as pemmican, you know, just a chunk of blubber and you put it in your mouth, and I think it is an acquired taste myself.

    [Laughter.]

    DR. MYERS:  As far as food preparation in Seychelles, you know, they are eating fish twice a day, they are eating a great variety of fish, and really it is prepared in multiple different ways, and we are actually looking at preparation and other things for fish at the current time, but in the main cohort, we did not do that.

    DR. MILLER:  Just the one question.  Hopefully, as the Chair, maybe I could ask the last question before getting on.

    Why don't you ask your question and then I will hopefully get to my naive one.

    DR. ACHOLONU:  The last speaker, Dr. Grandjean, made reference to the fact that the concentration of methylmercury drops after a child is born, and you have said that the toxicity of methylmercury may show later in life.

    What I would like to know is, is the concentration cumulative in the person, the concentration of methylmercury, is it cumulative in the person?

    DR. MYERS:  Well, there is a half-life to it, and there is excretion, but it is slow excretion.  We, as in the Faroes, have measured hair levels in the children, and they are generally fairly low until the children get older.

    It is excreted some in breast milk.  Chris, do you want to answer that?

    DR. COX:  Well, I think Dr. Clarkson will be here, you could ask him, but I believe the half-time is 50 days.

    DR. MYERS:  That was mentioned earlier, that is correct, 45 to 50 days.

    DR. MILLER:  Just one question that kind of puzzles me.  Given actually in both populations, but given in the Seychelles that you have a population that is compressed socially, I think you said, that this is a relatively isolated community, at least in terms of its dietary sources, what explains the 27 times variation that you found I think in your hair?

    DR. MYERS:  That's a good question.  I am not sure we have an answer.  We have assumed that it is dietary and related to the species of fish that is being eaten and favorite.  People have different favorites, and there are fish that have higher concentrations.  Mackerel and bekin, which is--I have forgotten the name--barracuda, barracuda, shark do have higher concentrations.

    So, if you eat enough of the higher concentration fish, that could explain everything, but we have not studied it specifically.

    DR. MILLER:  Thank you very much.

    We are going to call a break now, 10 minutes, please, and be back about 10:15.

    [Break.]

    DR. MILLER:  The next speaker is Dr. Christopher DeRosa from Centers for Disease Control to talk about the recommendations from the Agency for Toxic Substances and Disease Registry.

Agency for Toxic Substances and Disease Registry

Dr. Christopher DeRosa

    DR. DeROSA:  I would like to thank FDA for the opportunity to share with you some of our Agency's perspectives regarding the Agency's views on mercury and related compounds.  I would also like to acknowledge my colleagues John Risher and Dennis Jones, who assisted me in preparing some of the materials for today's meeting.

    Today, what I would like to do is provide you a brief background of our agency.  We are affiliated with the Centers for Disease Control, but we are actually one of eight independent agencies of the U.S. Public Health Service within the Department of Health and Human Services, and we are the primary health agency or the primary federal agency dealing with the implementation of the health mandates of Superfund.

    It is not totally incorrect to affiliate me with CDC because our administrator is also the director of CDC, as was currently announced, Dr. Julie Gerberding.

    I would also like to provide a chronology of some of our key activities over the years, talk about the rationale for the position we took in our toxicological profile, a document that we published in 1999, an update of two previous toxicological profiles, and then some insight as to our future activities and some current ongoing activities that may have a bearing as this dialogue at this meeting will have as we go forward with a reassessment of methylmercury, as well as other forms of mercury.

    [Slide.

    Among the health mandates that we have under the Superfund or CERCLA legislation are to prepare toxicological profiles.  These are documents that appear and publicly reviewed, articulating what we know in the broad areas of exposure, toxicity, and epidemiology.

    They attempt to provide health guidance for methylmercury, as well as other compounds that identify what the adverse effects are that we need to be concerned about and at what level those effects might be a concern.

    These are based on a list of priority pollutants that we prepare in conjunction with our colleagues at EPA on a biannual basis.  It lists in priority order the 275 substances most frequently encountered at waste sites, on controlled hazardous waste sites, based on the probability of human exposure, toxicity, and frequency of occurrence at those sites.

    So, we have prepared 250 profiles addressing some 1,000 chemicals, and they are inclusive of mercury, which is No. 3 on our priority list of 275 chemicals.

    We also are mandated to initiate a research program, an applied research program to address what we don't know.  I think it is important that the profiles, in addition to setting forth what we do know, also address what we don't know, which is sometimes a challenging effort in terms of identifying mechanisms to fill those data gaps, but we have identified 200 priority data needs, and they are currently addressing those in cooperation with our colleagues at EPA, NIEHS, and through some grants mechanisms, as well.

    Once a site becomes listed on what is referred to by EPA as an NPL or National Priorities List site, we are required by law to prepare a public health assessment on the health hazards associated with that site.

    This is based on four general avenues of information - health outcome data that the community might provide, what the community concerns are, environmental monitoring data provided by EPA, as well as the information contained in our toxicological profiles prepared on those substances that might be encountered at a given site.

    Finally, we are required to update those profiles at intervals not to exceed three years.  Our first profile on mercury was released in 1989, and we have had subsequent updates on two occasions since then.

    [Slide.

    This slide is really some of what we knew and when we knew it in terms of the documentation regarding mercury.  I mentioned the first profile in 1989.  In 1993, we updated that profile using the Iraqi study to develop the chronic MRL of 1 microgram per kilogram per day, similar to EPA's current value.

    We had convened an expert panel to discuss a benchmark dose approach for methylmercury, but we are advised that the Iraqi study at that time had been somewhat overextended and overinterpreted and that it would be better that we wait for the outcome of the information coming out of the Seychelles.

    We convened a second expert panel in '95 to address the issue of bioavailability.  This is working intermittently, which is worse than not at all, and then we began consideration of the update of our profile from '93 based on the publication in 1995 of some of the data coming out of the Seychelles.

    So, we initiated the update in '97, and the next slide is a continuation of that.

    [Slide.

    We involved in that process an extensive peer review process including an expert panel review of the post-public comments that we had on the draft that was released for public comment.  We had representatives of EPA and other federal agencies, but importantly, from the Faroes and the Seychelles there to further vet some of the data that they had developed to date.

    I am sure there is an activation point on this one, too, but I can't quite see it, but at any rate, the point being that one of the things that we were strongly reliant on was this workshop that was referenced.  It was a workshop initiated by the President's Office of Science and Technology Policy.

    It was one which brought together four expert panels in some broad disciplinary areas to really dig into the critical data sets, not only in the Seychelles and Faroes, but some of the work that Donna Mergler [ph] in Canada had done, as well as others.

    So, we followed that very closely and we used that as a key basis for some of the decisions we made in releasing the mercury profile to the public.  I would like to just mention for a second that our mandate is one of getting information in the hands of citizens, so that they can make informed decisions about their health.

    In addition to methylmercury, elemental mercury is an issue from the standpoint of emergency response, which we also have responsibility for under the National Contingency Plan.  It is the number one agent that is involved in emergency responses at our agency.

    We also have concerns about the salts of mercury, also dimethylmercury, which was responsible for the unfortunate death of a researcher at Dartmouth because it is used as a calibrating agent in some instrumentation.

    More recently, since the publication of the profile, the question of the safety of vaccines has come into play based on the use of thimerosol ethylmercury as a preservative in batch vaccines, and then finally, more recently, we have become very concerned about the use and misuse of chelation therapy by a number of individuals who are profiting at the expense of both physiologically and financially of the people they are treating in some instances.

    The next slide really talks a little bit about what you have already heard, and I am not going to spend a lot of time on this because it has already been discussed at length.

    [Slide.

    I would say that two key issues with respect to the Seychelles that have been cited is this issue of lack of an effect although we now have seen that there are effects. The reason that lack of an effect may be of concern to some is that it may suggest that your protocol was not vigorous enough to detect what you were looking for, there may be some other issue that you need to be concerned about, but they have, in fact, demonstrated some enhanced performance on some of the tests and, as we just saw, one report of an adverse effect in addition to the one at 29 months in the boys.

    Then, the missing domain-specific endpoints that had not yet been assessed at 66 months was something that we took into consideration in our treatment of uncertainty, and we relied in part on the Faroes data to help us deal with that.

    In terms of the Faroe Island, another excellent study, we have the benefit of two very fine epidemiologic studies.  There is the issue of the type of and duration of exposure.  By that, I mean the whale blubber being consumed perhaps once to twice a month at relatively high levels, 3 parts per million as opposed to 0.3 parts per million, which is characteristic of fish sold commercially in this country.

    Then, the concurrent exposure to PCBs and other persistent organic pollutants.  PCBs are at levels 10 times higher than in the U.S. population in the Faroes, and at three times the level of FDA's tolerable daily intake.

    Other persistent organic pollutants are also a concern because PCBs are generally considered to be a marker for other POPs.  For example, DDT is present at a level, in terms of exposure, five times our health guidance value for that compound.

    This has already been referenced in terms of the work of Todd Kelstrom [ph] and his colleagues done in New Zealand, and the sensitivity to outliers.  The initial report, the initial analysis was not significant until one of the most highly exposed individuals, who showed no adverse effects or associations, was deleted, and then we did see the association in some areas become positive.  So, that is just the issue of the sensitivity to one outlier, and the question is, is the outlier relevant statistically, is it relevant biologically, and I think as geneticist Dijanski [ph] said many years ago, "Treasure your outliers or your exceptions."  That is a significant issue.

    [Slide.

    Now, going ahead to the workshop held in '98 in North Carolina, you can see a very distinguished panel dealing with the confounders and variables issue including Dr. Jacobson, who is here with us today.

    I would like to share with you just some of the findings that we have really centered on as we went forward in trying to bring our document to closure, because I think it illustrates some of the deliberative process and some of the key concerns that we had as we attempted to deal with this issue.

    [Slide.

    This is the first of one of the findings dealing with PCBs.  When PCBs and mercury are included together in the model that was used to analyze the outcomes, one of the outcomes is specifically related to mercury exposure, one of the four that had been reported as positive.  For the other three, which included the Boston Naming Test, both PCB and mercury effects fall show of conventional levels of statistical significance.  So, I think that that is a key finding that we focused on.

    [Slide.

    Again dealing with this issue of PCBs is that it is likely that both of these contaminants adversely affect these three outcomes, but the relative contributions cannot be determined given their co-occurrence in the population.

    [Slide.

    Finally, regarding the concurrent PCB exposure, the Statistics and Design Expert Panel determined that the best method to deal with this would be to study a population where exposure to PCBs is not an issue.

    [Slide.

    This is simply a listing of those individuals who served on that, and these people have published extensively in this field and are recognized obviously as experts in the field.

    [Slide.

    Turning now to a little bit about the health guidance value, well, let me just back up and talk about why we use the Seychelles cohort as the primary study, but we also relied heavily on the Faroes study, as well.

    With the exception of two things, these issues, these attributes are all linked to both the Faroes study, as well as the Seychelles study.  The two that I would like to bring out, that I think relate specifically to the Seychelles, is the issue of the pattern of exposure and the levels of exposure over time.

    Obviously,  there are 10 to 20 times higher levels in the Seychelles, but it is not because the fish is more contaminated, it is because they eat more fish.  They have the highest per capita consumption of fish in the world.

    Then, the issue of confounding factors, we felt that there were fewer personal and lifestyle confounding factors, that it is a relatively pristine environment, and those levels of some of the other contaminants of concern were shared by Dr. Myers.  As he also pointed out, there is basic health and education infrastructure that is really quite remarkable in my mind, having had the opportunity to visit and observe some of the activities of the researchers there, and then again this issue of confounding for a number of other factors.

    [Slide.

    Turning to the issue of health guidance, there are a number of different terms used for health guidance factors.  They are very much analogous in the way that these different terms are derived.  Our agency uses the term "minimal risk level," which is analogous to the reference dose, the tolerable daily intake, and previously referred to as the "acceptable daily intake."

    So, it is an estimate of exposure that is thought to be without significant risk of an adverse health outcome over a given route and duration of exposure in addition to deriving chronic oral exposures, chronic inhalation exposure guidance values, we also deal with acute, defined as less than 14 days, and intermediate, 15 days to a year in our documents.

    [Slide.

    This is the generic formula that we used to derive a minimal risk level.  If it is not visible, it is probably because there is a rule that you have to have at least one slide that no one can see when you present.  This probably is not the only one, but at any rate, the MRL is simply derived operationally in a very straightforward way, analogous to the reference dose.

    You identify a toxicity benchmark, a no observed adverse effect level, or a low observed adverse effect level, or a benchmark dose, as was mentioned earlier, divided by an uncertainty factor whose magnitude is inversely related to our confidence in the database.  The larger the uncertainty factor, the less our confidence in the database.

    DR. MILLER:  Dr. DeRosa, you have 5 minutes more.

    DR. DeROSA:  Thanks.

    [Slide.

    In the derivation of the MRL, the issue is that you have mercury ingested by the mothers, the offspring of the group of concern.  Hair mercury levels are measured in the mothers, and you have to have hair-to-blood ratio in order to calculate a daily intake based on the blood concentration.

    Fortunately, we have some very good human data to provide this ratio.  The point of central tendencies of about 14 studies is about 250 to 1.

    [Slide.

    This just shows you some of the calculus that goes into identifying what the dietary intake is.  It is related to this issue of the fraction of the daily intake that is actually taken up by the blood, which is defined by what is absorbed, and then what portion of what is absorbed actually makes its way into the bloodstream.  It also is related to the blood volume, which is about 7 percent.  That is about standard, about the second trimester of pregnancy.

    This is what we have the concentration in the blood is based on what we see in the hair and the ratio that I just mentioned.

    [Slide.

    This just lays out the mathematics of this and the concentration in the blood that equates to dietary intake in milligrams per kilogram per day was divided by an uncertainty factor of 4.5, providing us with an MRL of 3 micrograms or 0.003 milligram per kilogram per day.

    [Slide.

    Some people say that I posed for this, but I did not pose for this slide.  How certain are we about what we know?

    [Slide.

    This attempts to lay out--this may be one of those other illegible slides--the standard factors of 10 that are typically used in deriving an MRL, 1 to 10 for human variability, 1 to 10 for animal to human extrapolation, 1 to 10 for extrapolation from an NOAEL to a LOAEL, a modifying factor to adjust for scientific quality of the database lack missing datasets perhaps, and 1 to extrapolate across duration.

    Some have said that we do this because we have 10 fingers, but these are biologically distributed phenomena that we are looking at, generally speaking, so dividing by 10, you encompass 95 percent of the variability.  So, it does have some basis in biological science.

    [Slide.

    This slide simply sorts out our treatment of uncertainty.  Because we have human subjects being assessed, there is no species-to-species extrapolation factor other than 1, because we used the NOAEL identified in the Seychelles study.  We used a 1 because of the issue of lifetime or long-term study over multiple generations or exposure over multiple generations we use 1.

    In terms of human variability, we used a factor of 3.  This is a factor of 1.5 for pharmacokinetics, which we had determined through some modeling that I believe Harvey Clewell or Kenny Krump [ph] did for us, and then we have the World Health Organization Steering Committee on Risk Assessment pointed out, and I participated as a member of the Steering Committee, that these are equally determined by pharmacodynamics and pharmacokinetics.  Kinetics is how it gets there, dynamics is what does it do once it gets to the target.

    So, we added these two components of this, and it sometimes has been said that we multiplied uncertainty factors.  We did not do that.  We added these two components of this uncertainty factor and then multiplied it by a modifying factor of 1.5 to account for our concern regarding domain-specific effects out of the Seychelles.

    [Slide.

    In summary, we have MRLs for multiple forms of mercury.  The critical study was based upon some of the things that I have just mentioned and that have been touched upon elsewhere.

    [Slide.

    Some ongoing activities.  We are just again following the science, where it leads in terms of the data coming out of the Seychelles and the Faroes.  We have developed an interaction profile on the contaminants typically found in contaminated fish in this country in cooperation with EPA's Office of Research and Development.

    We are engaged in a study in the Czech and Slovak Republics to look at perinatal exposures to persistent organic pollutants.  We are looking at thimerosol in some rodent studies to identify the comparative kinetics of thimerosol and methylmercury , and it turns out that while methylmercury has a half-life of from 45 to 50 days, thimerosol ethylmercury has a half-life about one-fifth of that.

    We are planning a chelation workshop to come around this issue of chelation.  The mercury document also served as the basis for WHO's International Assessment document recently released.  John Risher, who is here today, is the author of that document.  We are participating in the OSTP-CNR Working group.

    I think there is one more slide and I will be finished here.

    [Slide.

    Obviously, there is a lot of work going on, not only on methylmercury, but other forms of mercury.  We are committed to a continuing evaluation and understanding insofar as we can reduce uncertainty and provide some improved guidance to the public.

    We will be looking carefully at the deliberations of bodies, such as this, as we go forward with the update of our tox profile later this year.

    Thank you.

    DR. MILLER:  Thank you.

    Questions, comments?

Questions of Clarification

    DR. HOTCHKISS:  I just wanted to make sure that I understood.  You went through your MRL pretty quickly.  Your kind of bottom line number was 0.3 mcg/kg/day?

    DR. DeROSA:  It was 0.3 mcg/kg/day, and it was 0.0003 mg/kg/day.

    DR. HOTCHKISS:  Then, I got that number, and that is based on a number of factors, but, in essence, on one or more studies during the outcome of exposure during pregnancy, is that correct?

    DR. DeROSA:  Yes, the two studies that were pivotal there were the Seychelles, we used the mean hair level and the highest quartile studied in the Seychelles of I think it was 15.3 ppm in maternal hair, converted that to a blood level, then used that blood level to back-calculate a daily intake, and that daily intake was then divided by the uncertainty factor of 4.5.

    DR. HOTCHKISS:  And the 0.3 mcg/kg/day, can you tell me when your agency publicly released that amount and what the response to that has been?  Is it published in the scientific literature?

    DR. DeROSA:  The 0.3?

    DR. HOTCHKISS:  Yes, and the rationale for it.

    DR. DeROSA:  Yes, the rationale and the overall evaluation of the database is in the toxicological profile on mercury.  That is about a 750-page document.  It went through some extensive peer review, as well as public comment period.

    We have had a range of different comments on that. There is a wide spectrum of opinion about what the health guidance value for mercury should be, but I think that more important than what the divergence of what that opinion is, is that there is no disagreement that methylmercury is a neurotoxicant of the first degree, the one that we have to be concerned about minimizing exposures to, and that while we continue to espouse the benefit of fish as a component of the diet, that because mercury is bad, if you had consumed fish with highly contaminated levels of mercury, there may be implications for your health depending on the time of your exposure.

    DR. HOTCHKISS:  Thank you.

    DR. MILLER:  Dr. Nordgren.

    DR. NORDGREN:  I was wondering, would it be possible for you to make copies of the last few slides on how you determined this?

    DR. DeROSA:  Yes, we can make those available to you.

    DR. NORDGREN:  I think that is kind of crucial to what we are trying to do here.

    DR. DeROSA:  Right.

    DR. MILLER:  Other questions or comments?

    If not, we thank you very much.  It looks like we are right on schedule.

    We will now break for lunch.  Please be back here at 1 o'clock.  We will begin at 1 o'clock precisely whether it's just me or anybody else.

    [Whereupon, at 11:50 a.m., the proceedings were recessed, to be resumed at 1:00 p.m.]

AFTERNOON PROCEEDINGS

[1:00 p.m.]

    DR. MILLER:  I call the committee to order.

    Our first speaker this afternoon is Dr. Penny Kris-Etherton of Penn State University, who will be talking about consumer messages.

    Dr. Kris-Etherton.

Consumer Messages

Dr. Penny Kris-Etherton

    DR. KRIS-ETHERTON:  The topics I am going to cover are shown on this slide.  I will talk about fish recommendations from professional organizations and government agencies, such as FDA.

    I am going to just give a real brief truncated version of a talk that I give communicating fish recommendations from both American Heart Association, other professional groups, and FDA, and then we will talk a little bit about effective risk communication principles.

    Then, I am going to present a consumer research model developed by the International Food Information Council for communicating food and nutrition messages effectively.  Then, we will sum it up.

    [Slide.

    A number of professional groups have made recommendations for fish consumption.  American Dietetic Association recommends eating two to three fish meals per week to decrease risk of cardiovascular disease.  In the late 1990s, ADA published a position paper on women's health and nutrition, and this recommended consuming fish two to three times a week.

    [Slide.

    In the year 2000, the American Heart Association released their revised dietary guidelines, and I had the distinct privilege of serving on the Nutrition Committee that developed these food-based dietary recommendations. They differed from other dietary recommendations that were nutrient based in terms of specific percent recommendations.

    So, American Heart Association recommends two servings of fish per week to confer cardioprotective effects.

    Then, USDA, Department of Health and Human Services, in their dietary guidelines as presented in the Food Guide Pyramid, encourages two to three servings of fish weekly.

    [Slide.

    In terms of the fish recommendations made by American Heart Association, as I noted, in the 2000 dietary guidelines, a fish recommendation was made, and that was published in 2000.  There was a science advisory published in 1996 entitled, "Fish Consumption, Fish Oil, Lipids, and Risk of Coronary Heart Disease."  Well, a lot has happened since 1996 with respect to fish and health benefits in terms of heart disease, so there is another science advisory in the pipeline, and I am not at liberty to tell you what it says except to say that at least it is in the pipeline, it is going to be published pretty soon, and the American Heart Association took the position of looking at health benefits with respect to heart disease and safety issues with respect to environmental contamination.

    So, they balanced a consumer message on the basis of health and risk.  That will be coming out somewhat soon.

    [Slide.

    As I said, I am going to give just a truncated version of a little talk that I give on fish and cardioprotective effects of omega-3 fatty acids.  I have given this talk many times and, as a cardiovascular nutritionist, I am deeply committed to the health benefits of omega-3 fatty acids.

    You are going to hear a detailed discussion on nutrition issues tomorrow from Dr. Bill Connor, and this is just sort of the tip of the iceberg of what I present.

    [Slide.

    The exact quote from the HA Dietary Guidelines is shown on this slide.  "Because of increased evidence for the cardiovascular benefits of fish, particularly fatty fish, consumption of at least two fish servings per week is now recommended."

    What is the science evidence that led to this recommendation?

    [Slide.

    Well, this particular slide just shows the many, the multiple cardioprotective effects of omega-3 fatty acids in fish, and in particular, I will show you some evidence that shows decreased incidence of sudden death, reduced arrhythmias, antiplatelet effects which protects against thrombosis, marked triglyceride lowering such that omega-3's are used by some physicians in hypertriglyceridemic patients, reduced coronary disease, morbidity and mortality, and what we know is that both alpha-linolenic acid, the plant derived source of omega-3 fatty acids, and marine-based omega-3 fatty acids, EPA and DHA, have cardioprotective effects, higher intakes, about 900 mg/day of EPA and DHA given as a fish oil supplement may benefit patients with coronary disease.

    Tomorrow, you are going to hear about striking neurological benefits in terms of the brain, the vasculature, eyes, for fetuses, infants, and young children. So, again, many, many health benefits of omega-3 fatty acids.

    [Slide.

    Here are some of the epidemiologic evidence that we have looked at that we took into account when we made the recommendation for two servings of fish per week.  This is from the Physicians Health Study Follow-up from the Harvard Group.  It was published in 1998.

    What you see here is that with one to two servings of fish per week and more, risk of sudden death is cut in half.

    [Slide.

    In a very well-known study, the DART study, which stands for Diet and Reinfarction Trial, a secondary intervention study, it was shown that men with heart disease who were given fish advice to consume between 6 and 12 ounces of fish per week had a much greater survival rate, as you can see, than men who got no fish advice.

    In fact, for people who didn't want to eat fish, they were given a fish oil supplement of about 1 gram per day, and those individuals have the same greater survival rate as did individuals who ate fish, showing that the cardioprotective effects are due to omega-3 fatty acids.

    [Slide.

    So, given all this, and given the FDA advice, I think that we have to put recommendations for fish consumption in perspective, in the proper context, so that people realize benefits and risks associated with fish consumption.

    This is a slide that Bill Harris put together, and this is what we tell people.  For pregnant women and women who may became pregnant, the risk for CVD is very low, the risk for methylmercury toxicity is very high, and so the recommendations for fish consumption are to avoid shark, king mackerel, tilefish, swordfish, consume no more than 12 ounces per week of fish low in methylmercury, and select a variety of fish low in mercury and PCBs.

    [Slide.

    To deal with other population groups, men under 45 and premenopausal women, they have a moderate risk of heart disease, risk of methylmercury is pretty low, and the fish message for them is consume at least two servings of fish, preferably fatty fish, per week, consume a variety of fish, follow state and federal advisories, and for men greater than 45 years of age and postmenopausal women, who have a high risk of CVD and low risk of methylmercury toxicity, we give them the same recommendations as we do for the younger cohorts - consume at least two servings of fish, eat a variety of fish, and follow state and federal advisories.

    So, this is one way I think of balancing the health message with a risk message, and that is, telling people health benefits and what FDA is recommending.

    [Slide.

    Let's talk a little bit about effective communication strategies for consumers.  I just want to say two things real quickly on the side.