Autism and the Environment: Challenges and Opportunities for Research: Workshop Proceedings (2008)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Day 2 April 19, 2007 WELCOME AND INTRODUCTIONS Dr. Alan Leshner Dr. Leshner: Yesterday was a terrific day, and therefore that puts great pressure on all of us to make sure today is an equally good day. One of the things it was characteristic of is a wide variety of people sharing their views, but also listening carefully to each other. That is what characterizes a good workshop, that it is actually a workshop and not a bunch of monologues where people tell you just how it is. One of the things, since I am somewhat outside this field, that I have been struck by is how many research opportunities and how many research challenges exist. That is both good news and bad news. The bad news is that I believe that we are way behind where we ought to be, that there is a tremendous need to have a focused research agenda, but also from my perspective the resources necessary to implement that research agenda as well. I have been very impressed by the number of NIH Institute directors, the Deputy Director of the National Science Foundation, and a variety of other people who have access to resources who have spent so much time with us, and I applaud that. I applaud their leadership in helping to move this field forward. Now, just to remind everybody of the ground rules, first, speakers, there is a little clock up here. The format for the discussions is that after the 15-minute talk, we will have hopefully 3 to 5 minutes for urgent questions of the speaker. We will restrict those to the participants at the table. At the end of each 165 

166 AUTISM AND THE ENVIRONMENT session, we will have an open discussion. Again, we will give priority to people at the table, then if we have time open it to the broader audience. Then at the end of the day, we have reserved a substantial amount of time for discussion by everybody in the room. Let me now introduce Henry Falk, who is the chair of this first session on environmental epidemiology, using population-based studies to isolate the environmental causes of autism. Dr. Falk is the director of the Coordinating Center for Environmental Health and Injury Prevention at CDC. 

Session V Environmental Epidemiology—Utilizing Population-Based Studies to Isolate the Environmental Causes of Autism Dr. Falk: Thank you very much, Alan. There was a lot of discussion yesterday about issues we will be talking about this morning, so I think there is a lot of interest in these sessions. We will start with Irva Hertz-Picciotto, who has a Ph.D. in epidemiology from the University of California–Berkeley. She was on the University of North Carolina–Chapel Hill faculty for 12 years, and is now at UC–Davis Department of Public Health Sciences. Her research interests are in environmental exposures, pregnancy outcomes, and epidemiological methods. She is on the editorial boards of the American Journal of Epidemiology, Environmental Health Perspectives, and Epidemiology, and was on the scientific advisory board for the U.S. EPA. Thank you very much, Irva. ENVIRONMENTAL EPIDEMIOLOGY STUDIES: NEW TECHNIQUES AND TECHNOLOGIES TO USE EPIDEMIOLOGY TO FIND ENVIRONMENTAL TRIGGERS15 Dr. Irva Hertz-Picciotto Dr. Hertz-Picciotto: Thank you. I am going to provide an overview of environmental epidemiology and epidemiology generally. I’ll talk about different study designs and what we have learned from them, and then make some recommendations. The first couple of study designs I am going to go through somewhat quickly, because I think most of the meat is really at the end, in terms of the future for the field. Starting with focused clinical studies, these are generally self- selected populations or a group of patients in a clinic. These studies are descriptive. They usually have small numbers of subjects in them. Sometimes the hypotheses are generated a posteriori. These are the 15 Throughout Dr. Hertz-Picciotto’s presentation, she may refer to slides that can be found online at http://www.iom.edu/?id=42469. 167 

168 AUTISM AND THE ENVIRONMENT studies that have taught us about sibling recurrence, about twin concordance, male–female ratio, the comorbidities, and the genetic syndromes that seem to also go sometimes with autism, as well as seizure disorders and gastrointestinal symptoms. They have taught us about the heterogeneity of onset, including the regression phenomenon that seems to happen in a lot of cases, and the data on anthropometrics, such as head size, have come from these studies as well. The second kind of study, described in the next slide, is based on administrative databases. These are large databases that are collected for administrative purposes. The diagnosis of autism is frequently done by whoever the clinician is who happens to see that child. We have learned about perinatal factors and about time trends from these studies. There are two ways in which these studies assess exposure. When exposures are not assessed in the individual and the outcomes are summarized at the group level, for instance, by area, it is called an ecologic study; in the other design, both exposure and outcomes are assessed at the individual level. This is an example of a time trend study conducted in Denmark using an administrative database of diagnoses. It was a study looking at the removal of thimerosal from vaccines and the rates of autism before and after. What you can see from this time line, which was not necessar- ily obvious from the original paper, was that before thimerosal was removed, there was a period of time when only inpatients were in the database, and during part of the “after-removal” period, which covered all the way out to 2000 in this study, there was an interval when both outpatients and inpatients were included. This study, therefore, is not a rigorous design, because as you can see, you can’t really compare the before and after periods because of artifacts in how the database was constructed, and specifically, in how that changed over time. The next slide shows another administrative database study, quite a good one done from the Swedish Birth Registry, which was linked to their inpatient register. It gave us some information about some of the aspects of the perinatal period that seemed to be associated with higher risk for autism. From these administrative database studies, we have learned about patterns: We have learned about the age effect of the parents, obstetric complications as risk factors, and aspects of the time trends in autism. Moving along to the genetic studies, these again, like the clinical studies, are volunteer samples. The largest such study right now is the 

PROCEEDINGS 169 Autism Genetic Resource Exchange (AGRE) database. It has over 1,000 families right now. They are all multiplex families with at least two members who are affected, and they are focusing on genetics. We have learned from these studies how highly concordant monozygotic twins are, but not entirely. We have learned that this is not a condition that follows simple Mendelian inheritance. We have also seen the slide show yesterday by Isaac Pessah and others depicting the large number of chromosomes that may be involved, indicating multiple genes. But at this point those studies have focused on genes in isolation, although Clara Lajonchere has been working with me to figure out how to collect more environmental data on the AGRE families. Just a few words about studying environmental factors in autism. I think there are some misconceptions, maybe not in this room, but certainly out in the field of science and the community at large. We know from the genetic studies that about 60 to 90 percent of cases have some genetic component. What can we conclude about environment? Let’s look at these two pies. This is the sufficient causes model from epidemiology. Each pie represents a set of sufficient causes that will cause autism in at least one individual out there. It might be that A is a genetic factor, that B is another gene, that C is an environmental factor in the prenatal period, and D might be something happening at birth or postnatally, just hypothetically. Looking at the lower pie, this is another set of sufficient causes, where A and C can be substituted with some other set of events. Each set, that is, each causal pie is sufficient. So what that means is, if you take away B from either one of those pies you don’t get autism. This is gene–environment interaction. Because of gene–environment interaction, environment plays a role in 10 to 40 percent at a minimum. But notice that in this individual, say someone corresponding to the upper pie, it takes both. Let’s suppose that this particular set of sufficient causes produces 30 percent of the autism cases. That means 30 percent require genes and 30 percent require environment. Suppose the other pie corresponds to the remaining 70 percent of autism cases: Under this scenario, 100 percent of the cases require genes and 100 percent require the environment. In other words, the contribution from genes plus the contribution from environment do not have to sum to 100 percent, and they will not sum to 100 if there is any gene–environment interaction. 

170 AUTISM AND THE ENVIRONMENT There are also several environmental factors that have been associated with autism with very high relative risk. The first is congenital rubella. In the mid-1960s, the United States experienced an epidemic of rubella. Mothers who had rubella during a pregnancy and passed it transplacentally gave birth to children at a much higher risk for autism, about 10-fold higher, and that figure is based on counting only the cases that did not seem to resolve over time. Thalidomide: also, a very high relative risk. Just looking at a few other factors such as maternal age or male sex, that are not necessarily causal, but might be proxies for some causal factors, there are also some large relative risks. The fourth type of study design I will talk about is the new genera- tion of case-control studies. These are population-based studies where the diagnosis is confirmed in all of the individuals. They cover a broad range of factors. Generally speaking the exposure has been assessed retrospectively, but that is not 100 percent true. If you go back to get medical records from these individuals, you are collecting essen-tially prospective data, data that were originally written down in a prospective manner. These studies also have been collecting specimens with linkage to laboratory scientists. I’ll give an example. There aren’t very many of these studies, but you have heard yesterday a little bit about the Norwegian study, not a case-control study, but they are doing nested case-control studies within it. This is a case-control study we are conducting in California. CHARGE stands for Childhood Autism Risk from Genetics and the Environment. We currently have about 800 participating families. They include children with autism, children with developmental delay but not autism, and children from the general population. Each child with a potential diagnosis for autism is assessed with the ADOS and the ADI and other assessments are done on all children, including cognitive and adaptive development, a physical exam, medical history, and a structured interview that takes about an hour and 40 minutes, covering 12 domains. Six of them are shown here, including information about the index pregnancy, household product use, metal exposures, and so forth. Then there are some self-administered forms about comorbidities and also about treatments and services that the child receives. We collect urine, blood, hair, and we ask the mother to bring in the baby hair lock if she saved it, and many of them have. They don’t have to give all the hair; 

PROCEEDINGS 171 we can do with a few strands. We also collect specimens from siblings and parents, and then we go back to get medical records. We have them sign medical record release forms. It’s a very labor-intensive process, and we try to obtain as many of the types of medical records shown on the slide as possible. In addition, in California there is a banking of newborn blood spots on every newborn. Currently we have about 480 dried blood spots and we are applying to get more. Overall, this is the scheme that we are working with. On the left side is a panel of broad classes of exposures that we drafted as priority exposures to take a look at, and on the right side are the methods for collecting data about those exposures. Just a few examples: From blood, we can measure pesticides, and we can ask about what pesticides were used in the home. We can also link the residential information the mother provides with some databases that are also available in California, which have a record of every commercial application of a pesticide anywhere in the state. The database has geographic locations of applications, which we then can link to the residence of the mother at the time of birth, or at any other time because we collect those residential data. Measurement of metals can be conducted in blood, hair, the baby lock, the newborn blood spot. We ask about fish consumption and other household product use for metals. Another example, data on infections can be abstracted from the medical records and is collected by interview from the medical history. You get the picture. The CHARGE study is in progress. With regard to other case-control studies of this type, there is also an autism phenome project, which Sue Swedo talked about yesterday; it involves NIMH and the M.I.N.D. Institute. The CADDRE study we will be hearing about later in the session from Diana Schendel. Just briefly, some of the things that we are starting to see: In CHARGE, we are finding very different immunologic profiles in the children with autism as compared to the control groups. A wide range of immune markers appear different in the children with autism. We have also examined gene expression, and have observed a set of genes that seem to be differentially expressed in the children with autism, especially in cells from the immune system called the natural killer cells. There are also some hints now that there are perinatal factors, potentially avoidable ones, that might be linked to autism. We have looked at the metals and last year reported at the IMFAR meeting those 

172 AUTISM AND THE ENVIRONMENT results. This year I will be reporting on PBDEs; that work was funded by Cure Autism Now, which is now in the process of merging with Autism Speaks. The CHARGE study, by the way, is funded by the National Institute for Environmental Health Sciences as part of one of the children’s centers. It began in 2001, and we are now in our second 5-year period. The last study design I wanted to talk about is the prospective cohort studies. These are studies where we start with a pregnant woman, and follow her and the child forward. We have already heard a little bit about the National Children’s Study. A couple of other studies are now looking at high-risk cohorts. In particular, the pregnant women are ones who already have a child with autism and are carrying another child. Because of the high sibling recurrence rate, these are high-risk pregnancies. One of these studies, which is also part of our Children’s Center, is called MARBLES; we are also collaborating with the EARLI network, which is scheduled to be funded beginning in 2008, for which Craig Newschaffer will be the PI (principal investigator). What we are focusing on is trying to find out how early we can see biological signs of autism. Several baby sib studies to date have been focused on the early behavioral, but not biological, indicators, and have started postnatally. Our aim is to determine what are the critical time windows for environmental exposures, and what are the biomarkers that we might use to identify pregnancies and children at high risk. However, any marker is only useful if we can additionally identify what are the target issues or receptors or enzymes on which we might intervene in order to interrupt the development of autism. In conclusion then, we have had a couple of decades in which we have learned a lot about autism from psychologists and psychiatrists. We have learned how to diagnose reliably, and even some early behavioral markers. We have learned from neuroscientists about aspects of the brain and brain growth, and now is the time for the environmental epidemiol- ogy and toxicology to work together. This is an area that has not received a lot of attention. What you see in this room is what is out there, pretty much. We have in this room about 80 percent of the environmental epidemiologists and toxicologists, who are looking at autism from this perspective. And only half the scientists in this room are currently doing autism research. So this is an area that has been understudied. There are very few studies of environ- mental factors in the causation of autism. What I have shown you is 

PROCEEDINGS 173 basically about what there is. These large epidemiologic studies can be linked to all the mechanis- tic questions that were raised yesterday. We are doing genomics and there is a potential to do metabolomics and proteomics to figure out biologically, physiologically, molecularly what is different in these children. With the prospective studies we can take the hypotheses that we are getting from the case-control studies, such as the finding from the CHARGE study of differences in expression of certain genes, and ask: Do those same genes differ, and how early on do they differ? Does it start at 18 months, 12 months, 6 months, or what about in the cord blood? In other words, let’s go back to the very early stages of life and brain development. That is the state of the art and that is where we should be going. Thank you. Dr. Falk: Thank you very much, Irva. There is time maybe for one or two very quick questions. Dr. Fombonne: Your design includes data that looks at a range of exposures, but you do not include in your model any timing of expo- sures. A lot of data suggest that early exposure is important. Therefore, my question relates to how you can modify your study design to look for common exposures at early time points. Dr. Hertz-Picciotto: In the case-control studies, obviously that is the deficiency in the design of the case-control study, is that you are measuring things now, and you really want to know what happened before the diagnosis, perhaps in the prenatal or perinatal period. As I said, the medical records do provide us with early information. We have been looking at even preconception, looking at things like in vitro fertilization, medications that the mother took during pregnancy, what kind of induction or augmentation of labor happened. So there is that component. We looked at the metals so far in the concurrent blood samples, and now we are going back and taking the baby lots. We started measuring the metals in the baby lots, which represent exposures in usually the first year of life, and then the newborn blood spot, which will tell us something about right before the time of delivery. So yes, it is a problem. In the questionnaire we also asked whether we can get valid information. It is subject to how well people can remember what happened, and do cases remember better than parents who have a child who is developing quite typically, if you ask them what 

174 AUTISM AND THE ENVIRONMENT pesticides did you use around the home when you were pregnant or in the first year of the child’s life. Obviously that is a problem, and that is partly why we are moving now to also doing prospective studies, where all the information will be collected prior to the diagnosis. Dr. Fombonne: May I ask another question? There are also techniques to look at clustering around different environmental exposures. There is currently a study in California looking at increased risk of autism to exposure to different pesticides at early periods of development and gestation. They have found really interesting findings that you can look at one particular exposure and map that to autism rates. Dr. Hertz-Picciotto: Yes, that is quite a lovely analysis. That is using administrative databases and then linking them to exposure databases. Where those databases exist, I think that is a really excellent approach. One of my graduate students has been doing—it is not looking to exposures, but she is looking at spatial clustering in California, and breaking it down. But yes, those exposure databases are quite useful, and the pesticide use report one is a very interesting database. Dr. Falk: Thank you very much. The second speaker is Craig Newschaffer. He will speak on environmental exposures in autism international studies. Craig is professor and chairman of the Department of Epidemiology and Biostatistics at Drexel University School of Public Health. He had founded and directed the surgical office in development, disabilities, and epidemiology at Johns Hopkins previously. ENVIRONMENTAL EXPOSURES IN AUTISM: INTERNATIONAL STUDIES16 Dr. Craig Newschaffer Dr. Newschaffer: Thank you. I have been tasked to talk a little bit about the promise and potential of international studies and international epidemiologic studies that shed some light on environmental exposures and autism. This framework of looking at frequency by person, place, and time predates the formulation of epidemiology as a discipline. It guides us still 16 Throughout Dr. Newschaffer’s presentation, he may refer to slides that can be found online at http://www.iom.edu/?id=42470. 

PROCEEDINGS 175 in doing basic descriptive work on disease frequency. What we find, however, is just describing frequency of disease is important for assessing pharmacology burden, but very quickly we transition from just a pure description to wanting to make some inferences about risk factors or causal mechanisms. In person, place, and time, the variation of disease across those three domains is very important to analytic epidemiology. The dimension of place plays a prominent role in this, and there are a few ways that place factors into this type of thinking. For the rest of my talk—place—I am going to be focusing on national variations. As Eric brought up, we can also talk about studies of place looking at smaller geographic units, but again, for the rest of my talk I am going to be focusing on cross-national variation. The way we bring these data on place into our thinking is, there is a classical application that I will be talking about in a little bit greater detail that in theory can help us shed light on whether or not environ- mental or genetic factors may be prominent in etiology for the particular disease under study. Then there are also location-specific opportunities. We heard a lot about these yesterday, studies taking place in specific locations because there are higher or lower exposures in those particular locales. I will talk about those a little bit, too. Then lastly, the notion of place in analytic epidemiology in terms of multilevel analyses. These are the sorts of analyses where we look at variables at the individual level and also at the contextual level or level of place. A variable like socioeconomic status (SES) in modern studies of epidemiology is often looked at from a multilevel perspective. My personal socioeconomic status might influence my risk for a particular outcome, but also the SES in the area that I live might have a separate and independent effect on outcome. I think that at this point, where we are with autism epidemiology is probably not to the point where multilevel analyses are going to be prominent, but I think the first two approaches are worth talking about in greater detail. What about this classic application? The first thing that we want to do when we want to think about variation in place or variation cross- nationally in terms of helping us think about environmental versus genetic causes, we need to rule out bias. We need to make sure that variation in measures of disease frequency in one country and another country reflect underlying risk and aren’t related to other factors. 

176 AUTISM AND THE ENVIRONMENT I think all of us in the room are somewhat familiar with the discussions about time trends in autism prevalence. While we may have differing opinions on where the level of evidence on that takes us, I think we all acknowledge that in thinking about time trends in autism, there are some difficulties that emerge in trying to figure out whether or not prevalence from one time period is truly different in terms of risk, reflecting risk from another time period, or other related factors such as changing diagnosis and recognition. The same kinds of issues are going to arise in cross-national variation studies. The next thing we would do after we rule out bias is, we need to think about the extent and the magnitude of variation across countries. In general, if we see a lot of variation across countries and the variation is of great magnitude, we feel that that suggests a probable role for an environmental cost. It is theoretically possible that genetic variation could bring about large differences or multiple differences across countries, but typically when you see these patterns, they are associated with environmental risk factors. However, just looking at the extent and the size of the variation isn’t enough. What we do in epidemiology, we look to some other fairly simple designs to help us follow up on international variation patterns. One common design is to do migrant studies. In migrant studies we look at individuals who move from one nation to another and see whether or not their risk is different from individuals who stay in the home country. There are challenges here as well, because if there is a risk difference we need to determine whether that is because of external environmental factors, exogenous chemicals in the environment that they have been exposed to, or it is because they are adopting lifestyle changes of the new nation. That can be a challenge to sort those things out as well. For autism, migrant studies are going to present some challenges. If you are talking about cancer, a disease with a long induction period, individuals can move from one country to another and their risk profile can be altered. For autism, we are talking about a relatively small induction period, and there are going to be challenges to migrant studies. We also look at ethnic variation studies. There, we look at different ethnic populations in one locale and look to see if there are differences in disease frequency across those ethnic populations. If there are, this can suggest that maybe there are genetic mechanisms in play. But again, complications. There are also sociocultural factors that move with 

PROCEEDINGS 177 ethnicity, so it can be difficult to tease those effects out as well. We also have to remember that we have got to be cautious in interpreting ecologic associations based on international data. If we use resident-to-proxy exposure, we can run into some problems with inference, and I’ll give you an example of that in a minute. I just want to show you some data for prostate cancer. A lot of good data on incidence rates of prostate cancer and other cancers cross- nationally. Here we see almost a 100-fold difference from China to Northern America on incidence, but even for cancer, where the data are very good, the interpretation is somewhat complicated. We see that on the mortality side, the difference isn’t that great. We know that there are large differences in recognition of prostate cancer across countries. Prostate-specific antigen screening is much more common in developed countries than in developing countries. We also know that there are ethnic variations here in the United States, with much larger incidence rates in African Americans than in Caucasian popula- tions. But even though we see this large pattern, we suspect that there are environmental factors involved in prostate cancer. Interpreting these data is challenging. Here is an example of the problems we get when we try to use nation to proxy specific exposure. These are plots of breast cancer incidence on the y axis, and on the x axis the countries are ordered by an estimate on how much dietary fat intake there is in the country. The lower countries on the left, and higher countries in terms of fat intake on the right. This is averaged data. When we do this, we get a very nice correlation which suggests there might be an association between dietary fat intake and breast cancer. But we know from individual-level studies that there really has not been consistent evidence that this association exists. Yet the group-level studies, the national data, have sort of misled us. So we need to be cautious when we do these. Location-specific opportunities. Higher-background-exposure out- come prevalences are going to give us more power to detect associations, comparable associations. We heard examples of that yesterday. This is one reason why we might go after studies in specific locations. Here are some examples from the literature on polycyclic aromatic hydrocarbon exposure in fetal growth. The cohort at the left is from Krakow, Poland. If you just look at the bottom two rows, you see that the proportion of exposure through food intake is comparable, but the 

178 AUTISM AND THE ENVIRONMENT personal exposure measures on the moms is much higher in Krakow. That is because of airborne exposure to PAHs there. That is a high- exposure cohort. This is another example from lead. We can see a high-exposure cohort on the left from Australia. If you look at the lead concentration data on the bottom, you will see that there is much higher exposure there. You will also see that there are very few low-exposed individuals in the Australian cohort on the left. This is something that we need to be careful about if we do these kinds of studies. If the exposure is too high, we lose variation, and we might not be able to see all the effects that we want to when we reduce that variation in that way. The other reason why we do targeted studies is because of different genotype prevalences. If we suspect that genotype prevalence is higher in one nation, we might want to focus on gene-finding investigations there, because with similar logic it will give us enhanced power to identify those genes. For genotypes that are very rare in one location, if they are more common in another, it may help us find the full range of genotypes that are involved. In gene–environment interaction, the situation is a little bit more complicated, and kind of interesting. If there is gene–environment interaction, the exposure main effect may be more detectable in populations where the susceptibility is higher. So if we go into a population that has higher prevalence of susceptibility genotypes, we may be able to find the environmental exposure easier. The same logic applies for gene finding. If we move into countries where exposure levels are high, we may be able to find the susceptibility genotype more easily, and we might be able to estimate the actual interaction between the gene and the environment more easily in those situations. Now I want to talk a little bit about what we know about autism. We have been interested in international variation in autism for even greater than 25 years, as this slide implies. There are case study data going back 30 and 40 years. These are data on autistic disorder prevalence by time and also by location. If we take a quick look at this, we see that in the earlier years, there basically were limited differences across regions. There is a cluster of Japanese studies that were a little bit higher in the mid-80s, and that has been attributed to the fact that they were using DSM-III criteria and had very intense case findings. 

PROCEEDINGS 179 If we look into the more recent studies, we see a lot of variability, although it is interesting to note that in the very recent studies, variability seems to be narrowing, around 30 per 10,000. Remember, this is for autistic disorder prevalence, not the full spectrum. But if we look at patterns by region, we don’t really see anything striking. But what is striking is that these are all developed countries, for the most part. We have limited data so far in developing countries. In part through the efforts of CDC and Autism Speaks, investigators who are doing autism research internationally have been brought together in a series of symposia. What I am showing now for the remainder of my talk reflects my take on what has been presented in those symposia. What we note first is that there are now studies beginning in developing countries, as indicated by the hatched dots. The studies are of a variety of different designs. We have some freestanding prevalence surveys, some freestanding etiologic risk factor studies, but what is probably most common is this combined design, where there is some population that is going to give us prevalence estimate, but then there is also some recruitment of a non-case population so that some etiologic risk factor studies can flow from this. In this combined design, there are different ways that case finding is done in different countries. Some case-finding approaches rely on registries and service system records, and these tend to be in the higher income developing countries, and others rely on population screening, and these tend to be in the developing countries. This can cause some problems with bias. The type of case identification approaches can make for noncomparability across these studies. I am going to have to move a litter faster because I am cognizant of my time. From work that we have done on case finding, where we have been doing some screening in a developing country, China, we have come up with some issues. We found some language problems in terms of adapting our English screeners to the Chinese population, and we have come up with some cultural problems related to gestures being discour- aged and persistence being highly valued in that culture. Some objects used in common screeners don’t exist over in China. So there are some real challenges in adapting screening and case-finding techniques to developing countries. To conclude, moving forward, I think we are going to see more 

180 AUTISM AND THE ENVIRONMENT disease frequency studies in developing countries. That is going to be motivated in part by the developing countries’ concerns to characterize the burden of autism. Autism still has profound public health signifi- cance even in countries where there are other child health problems at higher prevalence than the United States. When we do this though, we need to make sure that we can maximize comparability of methods, and we can’t characterize the dif- ferences that are used across developing and developed countries. I think the initial analytic studies that we see in developing countries are going to be much like the initial studies in the United States that Irva just described. We are going to be looking at a broad array of environ- mental risk factors. I think that we are going to need stronger leads from basic science, including toxicology, clinical science, and other epidemi- ology, before there is justification for doing special focused analytic studies in particular locations where there is high exposure prevalence. I think gene-finding studies are going to go on in developing countries. They are most likely going to be primarily motivated by this notion that there might be some important functional variance that we just don’t see in the West that we might see in other populations. But I think it is important for us to remember that if gene–environment interaction is important, there is added motivation for doing these gene- finding studies in developing countries that have a very different expo- sure profile, because we might be more likely to see the genes that work in concert with that environmental exposure. In my last 24 seconds, I want to acknowledge the folks who have arranged the international epidemiology symposia, Autism Speaks and the CDC and the participating investigators there, and my colleagues on our China pilot, because I wouldn’t have gotten into the international studies business if it weren’t for all them. Thank you. Dr. Falk: Thank you. Remember that we will have time for general discussion after this session, but if there is a quick question, we have a minute or two. Dr. Martinez: Since I am not in the field, I read your recent article just published, “Summary of the Epidemiology of Autism.” One of the things that you stressed there is something that I referred to in my talk, which is the difference between ethnic groups in the United States, for which the same problems of interpretation that you have mentioned are there, and it is that precisely. 

PROCEEDINGS 181 Since we have representation of an underdeveloped country in the United States today, which is from Mexico, a huge population, a very good school-based study, to try to confirm that this is due to bias as most people interpret it, would be very interesting. Dr. Newschaffer: I totally agree. He is referring to the results from the Hispanic population from the CDC surveillance study. There are a number of locations in the network in the United States where there were significant differences between the Hispanic and the White population, but there are also a number where we didn’t see significant differences. I think it is very important for us to focus and understand this, to see whether it is real or not. Dr. Falk: Thank you very much, appreciate it, Craig. The next speaker is Diana Schendel, who will be talking about the CADDRE study in environmental epidemiology. Diana is a lead health scientist in epidemiology, team lead in developmental disabilities branch of the National Center on Birth Defects and Developmental Disabilities at CDC, and serves as science liaison for CDC Center for Autism and Developmental Disabilities Research in Epidemiology, the CADDRE project, and is principal investigator for the department in metropolitan Atlanta. Thank you, Diana. ENVIRONMENTAL EPIDEMIOLOGY STUDIES: CADDRE17 Dr. Diana Schendel Dr. Schendel: Thank you. I will be describing for you today the programs at CDC that can meet the challenge of environmental epidemiology and an epidemiologic focus. These programs were initiated by a congressional mandate for CDC to establish autism surveillance and research programs following the Children’s Health Act of 2000. The programs that were initiated include the autism and developmen- tal disabilities monitoring or ADDM Network, the Centers for Autism and Developmental Disabilities Research and Epidemiology or CADDRE network, and the “learn the signs, act early” campaign. But I 17 Throughout Dr. Schendel’s presentation, she may refer to slides that can be found online at http://www.iom.edu/?id=42471. 

182 AUTISM AND THE ENVIRONMENT will be focusing my comments on the ADDM and CADDRE networks. The ADDM Network goals are to establish a comparable population- based estimate of the prevalence of autism in various regions around the country, to describe the characteristics of children who have autism, and to examine these trends over time. This map illustrates the locations of our ADDM sites. The light blue sites are those represented by our current ADDM grantees. Georgia colored in yellow represents CDC as one of the participating ADDM sites. I can’t describe for you the ADDM methodology this morning, but I would like to highlight some of its strengths. All the ADDM sites apply a common case definition, and the majority of sites also can apply a uniform case identification approach. As the map illustrated, we have multiple sites throughout the country, and consequently the ADDM Network population base represents a very large proportion of the population of the United States. At our peak, the ADDM sites repre- sented a population base of around 10 percent of 8-year-olds in the United States, and 8 years old is the target age range for our monitoring program. And of course, the ADDM Network is intended to be ongoing in order to monitor prevalence over time. This may seem like a far cry from what is needed in an environ- mental epidemiology framework, but in fact, the ADDM Network provides a very fundamental role by providing us with a much better understanding of the patterns of occurrence of autism as well as the patterns of occurrence associated with some very broad environmental factors, such as variation in geography or community and by sociodemo- graphic factors. The ADDM data can also provide some very important understand- ing of the impact of certain methodologic factors on the variation in the observed prevalence that we see. This is just one illustration of these features. This is a representation of the ADDM Network data for the study year 2002. It is the prevalence of 8-year-olds in 2002, and the overall prevalence was 6.6 per thousand. But as you can see across the different study sites, the variation of prevalence varied considerably, from a low of around 3 per thousand to a high of around 10 per thousand. These data also illustrate, though, an important component attribut- able to methodology, most likely. That is, it did seem to vary according to the type of source, where the information on the cases was being 

PROCEEDINGS 183 obtained. As you can see in the group of sites on the right-hand side of the slide, these sites had unrestricted access to both education and medical records to identify their cases with autism. For those sites, the prevalence variation tended to cluster fairly closely around between six and seven per thousand. The sites on the left of your slide had either restricted or no access to education records, and were relying only on medical records or medical sources for information, and there the variation was much broader, ranging from around three to seven per thousand. So clearly, it seems to me that the number and types of sources where you are getting informa- tion on children with autism has a considerable impact on the observed prevalence. Therefore, the ADDM data can provide very important baseline referent data that can be used for comparison with other studies, whether it be to examine trends over time or across space, which might be used, for example, in specific locales who might be looking at the impact of particular intervention or prevention programs on the prevalence of autism, or in communities who have a concern over an apparent cluster. The data in the ADDM Network can also inform public health policy by providing numbers to quantify the public health burden of autism in a given community, as well as identifying vulnerable, at-risk, or perhaps underserved populations, all of which are important reference points for developing a framework for environmental epidemiology. Finally, the ADDM data can provide clues, as Craig Newschaffer has already described, regarding potential broad environmental factors. For instance, if we see social or economic class gradients, demographic or geography gradients, or ethnic gradients, these may be markers for an associated environment factor, that is, assuming that we can be confident that the variation is not attributable to some methodologic bias. But as we have seen considerably in this workshop, environment is a term that encompasses a very complex mixture to disentangle. Typically we are identifying single or a very few number of components within this complex mixture, whether they be from the physical or biological environments, or whether they might be impacted by an individual’s lifestyle or behavior. But in reality it is much more complex, where these individual components don’t act alone, but are combined into causal pathways leading to disease. In autism we may be actually looking at multiple outcomes, multiple disease outcomes referring to the different autism subgroups, all of which 

184 AUTISM AND THE ENVIRONMENT may have different associations with a particular causal pathway. Finally, the genetic underpinnings of autism may be contributing to this variability, as we have seen, if the genetic component changes an individual’s susceptibility or resistance to a particular environmental factor. Stepping into this enormous analytic need that we have in environ- mental epidemiology is the CADDRE program, contributing a small part to helping us disentangle this complex mixture. The CADDRE program in its first funding cycle was charged with developing a collaborative epidemiologic study that was conceived at the time to be offsetting a deficit in federal programs with an explicit population-based epidemi- ologic focus. The product of that effort was to develop a study to explore early development or SEED, which I will be describing for you now. This map illustrates the locations of our CADDRE sites around the country. It includes five study sites as well as a data coordinating center site, and Georgia serves as the sixth study site for SEED. The main research areas in SEED include investigation of the autism phenotype not only to better understand the unique features of the autism phenotype, but with the objective to identify specific subgroups for etiologic investigation. Our two main etiologic domains are the roles of infection and immune function and reproductive and hormonal function in autism. We are looking at GI features, not only as a component of the autism phenotype, but as its potential role in the etiology of autism. Clearly we are looking at genetic features, especially those related to our primary etiologic domains, and we are looking at sociodemographic features, not only as their potential role in etiologic pathways in autism, but also as a part of the process of identification and diagnosis of children in the community. Some of our other areas of interest include the impact of lifestyle behaviors in pregnancy. We have been looking at select mercury exposure, such as maternal vaccine exposure and RhoGAM exposure in pregnancy, and child vaccine history. We are looking at the occupational histories of both parents during pregnancy, sleep features of children with autism as well as patterns of hospitalization and injury. To do this we have adopted a case cohort study design, and we will be using a population-based approach to identify three groups of chil- dren: a group of children with autism, a neurodevelopmentally impaired comparison group, and a group called the subcohort, which represents 

PROCEEDINGS 185 our general population. This slide simply illustrates very generally our study design. Our base population consists of children who are both born in and reside in each one of the current study areas. Within this base population, children who are being served with developmental problems in the community serve as our source of children for the case and NIC groups, and the subcohort are drawn randomly from our base population. This slide illustrates our identification process for our case and NIC children. We are casting a broad net to identify these children. In other words, our eligible diagnosis includes not only autism, but a broad array of other related developmental problems. This broad net will include both children with an autism diagnosis as well as those with one or more of these other developmental conditions, and all of these children will be administered an autism screen. Children who fall above or who achieve a score above our cutoff for the autism screen will be designated a possible case, and those who fall below the cutoff will be randomly selected to serve as our NIC group participants. I might also add that the target age range for children in this study is children ages 24 to 60 months. Then our subcohort, as I said, was randomly selected from our base population, most of whom are typically developing, but nevertheless all of these candidates will be also given the autism screen. So for the purposes of enrollment, they are considered members of the subcohort, but if the child scores above the cutoff, that child will be sent through the data collection comparable to a case in order to have a full clinical evaluation. When we went to estimate how many children we might be able to identify through this process, we applied a very conservative prevalence estimate of about three per thousand to our estimated base population of about 485,000 children in our study areas, which gave us an estimated total number of children with autism of 1,550. Following potential losses due to inability to contact or ineligibility for the study or for refusal to participate, we hope to enroll around 900 children and families with autism. In following the possibility that not all of these families will complete the full data collection, we hope at the end to have at least 650 children and families with autism in our study. Since we intend to enroll the NIC and subcohort groups in a one-to-one ratio with cases, that means we will be enrolling around 2,700 families and getting full data collection, we hope, on 1,950 of these families. 

186 AUTISM AND THE ENVIRONMENT Based on this projected sample size, we did a calculation of the minimal detectable risk we might be able to detect with adequate power across a variety of scenarios of different exposure prevalences in different autism subgroup sizes. As you can see over a broad range of exposure prevalences, we have the ability to detect risks of between 1.5 to 2, although clearly at the most rare exposures and smaller subgroups, the power of our study declines accordingly. Our data collection includes a variety of interviews and self- administered questionnaires, including a 7-day stool diary and 3-day diet history. We will be doing extensive medical abstraction, similar to what was described for the CHARGE study, including preconceptional records for certain providers of the mothers, if they had psychiatric or immune dysfunction or hormonal dysfunction, and we have 3-year postnatal medical record examinations of the children. The children will be given a full clinical evaluation and a brief physical exam, and we will be collecting biologics from both the child and both parents, including hair from the child. We hope with the biologics to be able to measure both proteins and genes as well as mercury in the hair samples. Our list of possible candidates for analysis is not fixed at the current time in order to accommodate new discoveries in the meantime until we are ready for analysis. But the data collection protocols, of course, will accommodate a variety of analytes to be tested. We have also included in consent to allow the retention of sample after analysis for further testing in the future. Clearly these analytes that we choose to measure will primarily address our key etiologic domains in terms of brain function or immune or hormonal function, but we also hope to measure epigenetic or parent- of-origin effects, all of which we have learned may modify or be modified by the environment. Given these strengths, we believe SEED has a role in environmental epi of autism. Certainly since it is the only multisite study planned to date representing diverse communities and populations, it certainly will increase the generalizability of our study to the U.S. Since we are following a population-based approach to identify our subjects, this will enhance the representativeness of our sample and reduce biases due to subject selection. Since we chose to have two comparison groups, an affected and unaffected group, we hope this will enhance our ability to detect biases 

PROCEEDINGS 187 due to differential recall between parents with and without an affected child, and therefore reduce our exposure misclassification. Since all sites will be applying a uniform protocol, we can pool data across sites and therefore give us the largest sample size of a study planned to date, which clearly will enhance our study power and permit phenotypic subtyping for etiologic analysis, which although that does reduce the sample size we would be using, we hope it reduces imprecision due to outcome misclassification. Finally, with multiple research domains, we can look at multiple causal pathways or look at multiple points within a single pathway, and with some overlap of other studies that have been described, we can replicate analyses from prior studies. So we think the primary role of the SEED study for an environ- mental epidemiology perspective is that it simply gives us a much better understanding of the role of a variety of broadly environmental and genetic factors in ASD which can serve as a referent for studies that might be focused on more specific toxicological factors. We have the potential to expand the array of environmental factors that we have data on through data linkages or added data collection. Clearly it can also serve as both a hypothesis-testing and hypothesis-generating study and inform data collection and analysis in other studies to come. In terms of a recommendation, we have lots of needs, but probably the greatest need is the fact that a single study can’t do it all. The data collection that is needed and the analytic approaches are simply too diverse to be accommodated by a single study. So perhaps one area we might focus on is data pooling or coordination across studies. By coordination I don’t mean centralized control, but perhaps arriving at some consensus on certain data elements and analysis which can improve the strength of meta-analysis or data pooling for de novo analysis. Thank you. Dr. Falk: Thank you very much. We do have several minutes, if there are any questions at the moment. Dr. Beaudet: I’d just like to ask, does the CDC have any initiative to try to understand if the incidence of autism changed from 1970 to 2000? It seems like if it did change and we are at some new plateau, it would be nice to know that it did change over that interval. Dr. Schendel: We do, although we don’t have the data sources here in the United States that would address that question. But we are working with investigators in Scandinavia and Denmark, which do have these 

188 AUTISM AND THE ENVIRONMENT historic registers, and we are attempting to use those registers to investigate the impact of a variety of administrative changes in diagnosis or referral of children in those areas, to see what that impact might have had on the change over time. Dr. Beaudet: But it seems to me like I could interpret your answer to say there is no effort to answer this question for the U.S. population. Dr. Schendel: We don’t have the data. We only started the ADDM Network beginning with the 2000 study year. Dr. Beaudet: But there are some things that could be looked at retrospectively. If terbutaline administration is a high-risk factor, it seems to me that by retrospective study could be elucidated. Dr. Schendel: Where would you get the autism prevalence data? Dr. Beaudet: I guess I would be thinking about some kind of case-control study of autism with controls in terms of exposure to terbutaline in utero. I’m not an epidemiologist, so feel free to point out my ignorance. Dr. Schendel: That certainly would give you a risk estimate, but it is not going to give you an estimate of the number of children who might have had the disease in the general population. Dr. Falk: This might be an interesting issue to further explore during the general discussion session. Mr. Blaxill: Dr. Schendel, there is a lot of confusion about rates and trends and comparability of various statistics. Why would you make the choice to lump all of the autism spectrum disorders together in a single measure, as opposed to distinguishing between the different subclassifi- cations? That strikes me as a very unfortunate choice. Dr. Schendel: Well, that is true, there are components within the autism spectrum. Even though we do have the DSM-IV criteria, which may give us some differences within the spectrum of different subtypes, it is still not clear how that may be well implemented in this kind of study to where we are confident that the subtype that we have assigned a particular child based on behavioral data reported in the record would be reliable. So to be conservative, we have chosen to lump the diagnoses across the spectrum. Although we do collect a variety of behavioral data, as I said, it may be difficult and would have to be validated, with the ability to subtype. Dr. Leshner: So let me ask you a question about that. If you did that in your study, but if the database were widely available, could other 

PROCEEDINGS 189 investigators see if they could parse it out? Dr. Schendel: The data are there yes, to the extent that these data are recorded in the medical records of these children. We aren’t examining these children. We collect what information is available, and that information is not perfect. So while we feel confident that within the broad spectrum we can identify a child who may or may not have autism, it is much more challenging to take that information and reliably subtype. So is the direction we have taken. But the data we have are there. Dr. Leshner: I’m just asking whether it would be possible for other people to access that database. Dr. Schendel: You could. The data are available. It would have to be validated through a clinical validation study to confirm whether or not you can make a correct judgment call on that data. Dr. Newschaffer: Just to clarify for people who might be listening and confused, there are two studies. They have now transitioned to talking about the ADDM surveillance study, which is records based, not the SEED study, which is going to involve direct examination of kids. Dr. Schendel: Right, you’re right. Maybe that is the confusion. The SEED study will have the clinical validation. Dr. Leshner: I was thinking of CADDRE. Dr. Newschaffer: CADDRE will be doing extensive subtyping. Dr. Schendel: Thanks, Greg, I didn’t catch that confusion. The SEED study will have clinical validation of all the children, and we will be able to do subtyping based on the SEED study. Dr. Newschaffer: We are taking everybody in the spectrum in, but we will be able to subtype, including conventional and also nonconventional ways, by incorporating other information on medical presentation and comorbidities and symptomatology. So it will be ex- tensively subtyped. Dr. Falk: One more question from Sallie. Ms. Bernard: But Dr. Newschaffer, the SEED study is not going to give us prevalence, is that right? Dr. Newschaffer: Correct. Dr. Schendel: Right. Ms. Bernard: So knowing the subtypes—we lost it on ADDM, and if we are getting it through SEED it is really not delivering what we need to get. Dr. Newschaffer: That’s right, yes. 

190 AUTISM AND THE ENVIRONMENT Ms. Bernard: My other question for you, since we are not getting the information that we need from our U.S.-based studies, are you getting subtypes in your international studies, and are you getting different age groups, including into the older population, since we are just focusing on very young children in the United States? Dr. Newschaffer: First of all, they are not my international studies. If we look at the international studies across the board, there is a lot of heterogeneity in terms of what they are doing. Some are looking at different age groups; some, especially those that are based on more developed countries might be able to do subtyping. In developing countries where they are doing screening and have to come up with a way to validate diagnosis, none of them have gotten far enough. I suspect they will face some challenges in terms of doing subtyping. So I think it is going to be a mix for awhile as these international studies develop. Dr. Susser: Can I just ask that you redirect us to this question in the discussion period, because I think it is important. Dr. Falk: Yes, I was just going to suggest that. Thank you, Ezra. We will come back to this discussion for sure. The last speaker in this session is Allen Wilcox. He will be speaking about prenatal and perinatal exposures. Allen is the senior investigator in the epidemiology branch at NIEHS, where he has worked since 1979. He was chief of that branch for about 10 years and serves as editor-in-chief of the Journal of Epidemiology. He is also past president of most of the major epidemiological societies in the United States. Thank you, Allen. PRENATAL AND PERINATAL EXPOSURES18 Dr. Allen Wilcox Dr. Wilcox: Thank you for the invitation to be here. I am not an autism researcher. I thought maybe it was a mistake when I first got this invitation. But I am glad to be here. I have learned a lot. I am going to start by diverging from my text. The reason is, I would like to address a question that came up yesterday that I think is an 18 Throughout Dr. Wilcox’s presentation, he may refer to slides that can be found online at http://www.iom.edu/?id=42472. 

PROCEEDINGS 191 important one, about the categorization of autism, and looking for categories of phenotype that might help us understand causation better. There were a lot of discussions about ways to do that, using molecular biology, biomarkers, signs and symptoms, but no one mentioned epidemiology. So I would like to introduce the idea that epidemiology can actually help construct categories of this outcome that would be useful for unraveling its cause. I am going to present an analogy from another area of epidemiology, birth defects, and specifically oral facial clefts. This type of birth defect comes in what might seem at first to be a bewildering variety of manifestations: it can be only a cleft lip, it can be only a cleft palate, and it can be various combinations of both of those things. What we have come to understand is that these are actually two different birth defects. One is cleft palate by itself, and the other is cleft lip with or without cleft palate. The first recognition of that was by family studies by epidemiologists who showed that a family who had one child with a cleft lip had a 25-fold risk of having a second child with that same defect, but their risk of having a child with a cleft palate was by comparison much, much less—much closer to the general population. The converse is also true. The families with cleft palate have an enormous risk of having another baby with a cleft palate, but the risk of having a baby with cleft lip is much closer to the general background level. Since that time, we have come to understand the embryonic origins of these two defects and why they are distinct. Understanding that they are distinct has led epidemiologists to study these two defects as separate categories. This in turn has been very important for understanding their causation. I can give you an example. We did a study looking at folic acid and how it might prevent cleft lip or palate. We found that for cleft lip or palate, there was a 40 percent reduction with folic acid supplement, whereas for cleft palate there was no association. If we had combined all of these birth defects, all of these facial defects into one category, we would have had a much harder time identifying this association. There is an analogy in autism. You have the opportunity in multiplex families, in studies looking at recurrence of autism, to see whether there are different categories of autism that tend to cluster together in families. If there are, these might be used to create categories of autism for etiologic studies. I’ll just put that on the table for your consideration. 

192 AUTISM AND THE ENVIRONMENT Now back to our regularly scheduled program. I want to discuss the possibility of prenatal and perinatal origins for autism. I am not discussing the important questions of treatment. My comments have entirely to do with the discovery of causes that are presentable. The fact is, we know very little about causes of autism in general, and of prenatal causes in particular. But I would like to go through a number of arguments why looking in the prenatal period might be a productive line of research for autism. I am also going to talk a little bit about the difficulty of doing these kinds of studies, and finally, I will say something about the opportunities for doing better than we have done in the past. Plausibility. Some of these things are very familiar to you all, and have already been touched on by various speakers. We know the fetal nervous system is particularly susceptible to toxins at a low level of exposure that would not affect adults. The exposure to organic mercury at Minamata is a prime example of an exposure that had little or no effect on adults, and yet had devastating effects on the fetuses who were exposed. Thalidomide was a very important discovery for epidemiologists and clinicians because the exposure was not merely benign to the mother, it was actually prescribed to the mother. This severe birth defect (pho- comelia) was the result. And as Irva Hertz-Picciotto has mentioned, phocomelia is associated with autism. So we know the fetal nervous system is particularly susceptible to toxins. We also know that the fetal brain has a very long period of susceptibility. This figure is from an embryology textbook by Moore and Persaud. Along the top you see the whole sequence of prenatal events from conception through embryonic and fetal life, ending with the term infant. Each one of these colored bars represents a particular organ system and the period of time during which it is vulnerable to being disrupted. The red bar indicates the period of time when there can be structural malformations produced by exposures, and the yellow bar is the period of time in which functional problems can be introduced. You may have trouble seeing it, but the top bar is the central nervous system. As you can see, it is the only organ system that is vulnerable from the earliest embryonic life through delivery—and beyond, as we now know. Does this mean the fetal brain is fragile? I think we have an 

PROCEEDINGS 193 important responsibility in our public health role to find the right balance in our communications to the public. On one hand, we know there are thousands of chemicals that are potential neurotoxins in our environment. We know the fetal brain is very susceptible. On the other hand, we know that the brain is, in the language used yesterday, plastic. It is robust. It can manage to accommodate insults without being damaged. The fact is that most babies develop normally, despite all the things that their mothers may worry about. As experts in this field, we have to find the right balance, not to frighten the general public, but to create an awareness of serious issues that we need to address scientifically. A third point about prenatal exposures and autism is the interesting association of autism with difficulties of pregnancy. Much of this evidence is quite old. The studies go back to a time when research methods may not have been as good as we would expect today. But there is a sense that pregnancies that eventually result in an autistic child have some signs during the pregnancy that things are not exactly right. These are nonspecific, and I certainly would not want to suggest that the signs and symptoms that you see on this list “cause” autism. However, they may share some common cause with autism. These studies provide one more sign that the roots of autism, or at least of some cases of autism, may go back into the prenatal period. Whenever you hear epidemiologists talk, they will always tell you how difficult their work is—I am going to be no exception. There are some things about autism that make it particularly difficult for the kind of studies we would like to carry out. Autism is a rare outcome. Autism is hard to define. The exposures that may be important could also be rare, and they are often hard to measure. An example from yesterday was the discussion of the “hit-and-run” exposure, that is, a fluky exposure at a crucial time. How do we ever go back and reconstruct that? The retrospective reconstruction of exposures, as a number of the previous speakers have commented on, is always difficult. This is one reason prospective studies are particularly useful—they can gather exposure information in more detail, and we can then try to associate those exposures with the emergence of the disease later. The difficulty with prospective studies is that, for a rare disease, those prospective studies have to be very large. Until recently, this was not considered practical. However, the new wave of large prospective pregnancy studies offers new opportunities. We have heard some wonderful presentations this morning about 

194 AUTISM AND THE ENVIRONMENT promising studies of various creative designs. I am going to talk about the one that I happen to know best, which is the Norwegian cohort. You have heard this mentioned several times already in this meeting. In the cohort, the mothers are enrolled by week 18 of pregnancy, with a target population of 100,000 mothers and pregnancies by the end of 2007. I think they are close to that target. A great advantage of doing this kind of study in a medically well-organized country is that there are many registries and routine mechanisms of collecting health data which, when combined with universal identification numbers of all citizens, allow information to be assembled in a way that would be difficult and expensive in the United States. The plan in the Norway cohort is to follow these offspring through age 6, but of course longer would be better. This kind of a cohort study is a very ambitious undertaking, even in a place like Norway. It is expensive to do, and it has no immediate product. So there has to be some sense of building an infrastructure that can be used later for useful things. Let me tell you about my own institute’s participation in the study. This is one of those stories that starts out, “there were these two guys in a bar.” I was in Oslo having a beer with Per Magnus, who was one of the principals in the Norwegian cohort. This was 6 or 7 years ago, when the study was just on the verge of going into the field. The Norwegians had plunged into the study, with no clear idea of how they would pay for its completion. My conversation with Per Magnus went along these lines: You have great opportunities to measure pregnancies and pregnancy outcomes, but you’re not paying much attention to environmental exposures. Environmental research was not the Norwegians’ expertise, not something that they had much experience with. So I suggested that I would try to get my institute to provide some support to this cohort, specifically to pay for the collection of biological specimens during pregnancy that we could use to measure exposures. Specimens collected during the time when the fetus was actually developing could be stored and later assayed for toxicants that might be relevant to particular poor outcomes. I went back to my institute, and the leadership of the institute thought this was a great idea. We were able to provide money for the Norwegians to collect urine and blood during pregnancy. Blood is being stored as whole blood and plasma. We are prepared to assay at least this list of potential environmental exposures. To measure some of these 

PROCEEDINGS 195 things, for example heavy metals, we have to collect specimens in a very careful way in order to avoid contamination after the sample is collected. But we have done this. The samples are now in storage. I notice that a number of these exposures were on Irva Hertz-Picciotto’s list of hypotheses that she had for autism. More recently, Ezra Susser was in my office talking about this great project that he was involved with, in which they planned to identify autistic kids in the Norway cohort. He and I immediately recognized that this was a perfect partnership, in which we could provide biological specimens for measuring specific exposures and they would find the cases. So we made plans on the spot to combine our biological samples with their quite complicated and complete study of autism in Norway, all within the framework of this cohort study. So my pitch here is that these cohort studies provide a structure that can be useful in ways that we don’t even anticipate when they are started. Thank you very much. Dr. Falk: Thank you, Allen. If there is a quick question for Allen we’ll take that now. If not, we will go into the general discussion. So perhaps we will move to the general discussion. Questions or discussion points? Ms. Bono: I have several over the course of several speakers. For the ADDM surveillance, Diane Schendel, you were talking about, you were going to be looking at trends. But then Art Beaudet asked a question about birth cohorts. I know that you have looked at the 8-year-olds back to the 1992 and 1994 birth cohorts, but I’m not quite sure much further back. Then you were saying that those data weren’t available. I am concerned about that, because you are looking at education in, but in the education sources there have been a couple of studies of the 1989 and 1992 birth cohorts, where autism jumped. So if that is the case, we need to be going back further and looking at those years and seeing what changed in the environment at that point. For the SEED study, I was wondering if there were going to be any nonvaccinated in the study. I see that you have got 1,950 enrolled children by the end, but if you are looking at exposures, are there going to be any kids with no exposures to vaccines? Because that is certainly an issue that we need to deal with. So that is number two. Number three, the mercury and hair samples, I just want to remind everyone that there is a baby hair study, where the one thing that they noticed in the study is that the children that had autism had very little 

196 AUTISM AND THE ENVIRONMENT metals in their hair. That was the determining feature of those children. So once again, that theory of the nonexcreters. You may not see mercury in the hair of the kids that go on to develop autism, and that may be something you want to think about. I think those were the main points I wanted to make. Dr. Schendel: At CDC we have for the 1996 study, we have children born from 1986 to 1993, children ages 3 to 10 in 1996. But that is just limited to—it is not the higher cohorts of the ADDM Network. The ADDM Network, our first study year was 1992 births for children aged 8 in 2000, and we are moving forward. I don’t know whether the ADDM Network has discussed going back further and getting earlier cohorts, but currently they are moving on to the 2004 and 2006 study cohorts. Ms. Bono: It would seem to me, if the jump was earlier, then we need to go back earlier. We need to make that a priority in this room, that if the jump in autism was in 1989 and 1992 or perhaps even earlier, that late 1980s jump, that is when we need to go back and figure out what happened in the environment, because then the trend just continued. Dr. Falk: One of several comments we have had in terms of looking over a longer period of time. Ms. Bono: Right, it is of huge importance in time trends. So to look now at a prevalence, it is very important; we still want to look at these kids, but we also need to go back and look in time. Dr. Falk: Thank you. There are two more questions. Do you want to say anything about the other questions? Dr. Schendel: We are not specifically targeting children with or without vaccines. It is children against everybody within the study areas, some which may and some which may not have vaccine history. Ms. Bono: That begs the question of whether we need to look at that, because that is certainly an issue that we have heard over and over, and mercury is implicated throughout this entire workshop. Dr. Schendel: Maybe in the CHARGE study. Dr. Hertz-Picciotto, you have a specific psychological focus. Perhaps you could describe what CHARGE is doing. Dr. Hertz-Picciotto: We have been trying to reconstruct all sources of mercury exposures in the children in CHARGE, using diet, using use of household products like nasal sprays, vaccines, RhoGAM, and something else I cannot remember right now. 

PROCEEDINGS 197 So we are trying to ascertain all the sources of mercury in the child’s environment and then trying to correlate that with whether they have autism or not, and also related to the blood mercury levels. Most of the children—since the mercury was taken out of most of the vaccines, although it is still in some vaccines, there is a difficulty in ascertaining whether people have been exposed to thimerosal in vaccines, in particular influenza vaccine, which may or may not have thimerosal, depending on where they are getting it. Although the medical records are supposed to record lot and batch numbers, we have not always found that in the medical records, for one thing. The other problem is that people who now get these flu vaccines at their local mall, there may be no record that we can determine. So in other words, we may be able to ascertain the vaccines they have gotten, but we may not always be able to determine—some of them we can exclude thimerosal because thimerosal is no longer used in those, some of them we can’t always determine the dose the child would have had. So this is what we have been running up against in trying to do that analysis at this point in time. Ms. Bono: I just wanted to also comment regarding the limitations of that. Can you tease out non-exposure to vaccines, a zero-exposure category? Also, there is an overwhelming viral load as well in these children. So if there is a way to tease out all exposures to virus through vaccines and naturally, I think it is important to include too. Then also, we still don’t know the amount of trace exposure to thimerosal with vaccines, because it is used in the processing. We know there are low-level exposures, and we don’t know that that is safe either. So these are all things that I would love to see outlined in the study Ms. Redwood: I wanted to follow up on that same thing. I guess I am a little concerned with the study, in that if you are asking the question of vaccine exposure, there is not going to be enough variability. As you know, they are mandated by law; children have to have them to get into daycare centers and schools. As we saw in the VSD data when the first analysis went through, there was a zero-category exposure. If you are looking at all the children being exposed, you are not going to have enough variability. You need a zero-category exposure, which we had in the VSD data, and we saw statistically significant associations. But then the entrance criteria were changed to where all children had to be vaccinated to be in the study, and it took out the zero-level exposures and the statistical significance 

198 AUTISM AND THE ENVIRONMENT decreased. So I would like to see appear a vaccinated–unvaccinated study. I think as an advocate from the community, that is what we hear from parents over and over and over again. Dr. Schendel: I appreciate that, and I certainly understand there is a definite concern. When we planned the SEED study, it was at the same time that CHARGE was being planned, and there were other activities investigating vaccine safety at CDC. So we chose not to focus on that particular etiology, rather go at the main etiologies I described. So I think what you are saying is important, and I think it should be responded to. As I said in my last slide, we need multiple studies. SEED will address one area, and we need complementary studies to look at the others. Ms. Redwood: But that is mercury and aluminum. If you look at the full vaccine load, you are looking now specifically at mercury. That will include aluminum. Over 90 percent of the flu vaccines this year administered to pregnant women, infants, and children contain thimerosal. Maybe that will help you in your analysis. Dr. Newschaffer: Something else to consider in terms of the age of entry for a retrospective study is, the older the kids are, self-report information, even medical record information, becomes increasingly compromised. So for other sources of mercury exposure like diet or questionnaires and the types of things that Dr. Hertz-Picciotto used, there is a trade-off when you get older kids in with recall. That also affects other exposures that are of interest, too, where that is our prime source. Dr. Hertz-Picciotto: In regard to the issue of children who weren’t vaccinated, that is true; very few children in this country are not vaccinated at all. Recently, the NIH held a workshop specifically on the vaccine database that CDC has acquired, which is a better way to look at the vaccine question than either the CHARGE or the CADDRE SEED study, because it is large enough to have children without vaccines. We did make some recommendations about improved studies coming out of that large database, the vaccine safety database. So as I understand it, there are some studies that are now being done, and there were recommendations for other studies to be done. I think that is probably the best place for that particular hypothesis to be addressed. I think your point is well taken, that when there is a limited amount of variability in exposure, then there is a limitation in how much you can 

PROCEEDINGS 199 learn from that. Those data in one of Dr. Newschaffer’s slides were making that very point. So thank you. Dr. Schwartz: I had a potential suggestion here. As a nonepidemi- ologist, I am learning a lot about these epidemiologic studies that are ongoing, and some that have been completed in this area. But it is a little bit unclear to me exactly what the strengths and the weaknesses are, what the complementary and overlapping areas of goals are for each one of these studies, and what the gaps might be that need to be filled by studies that have not yet been funded or solicited. I’m just wondering whether there should be an effort to try to look at all of the studies. Maybe that is what is going on in the matrix, the autism matrix. I don’t really know, but I guess I am throwing this out there as a question. Do we need an effort to try to look at these studies that are ongoing and have been completed as a way of trying to get a sense of where we are and what needs to be done? Dr. Hertz-Picciotto: First of all, none of them is completed. In CHARGE we have our first 5 years of data, which include as I said about 800 subjects. You did fund us for another 5 years, and we do expect to get about another 800 subjects. So we will have about 1,600 at the end. Then the CADDRE study will probably be slightly larger than that. I think Dr. Schendel’s point about trying to look at ways to combine data across these studies may be a very good idea. We haven’t really talked about that directly, although in the planning stage we did share some of our instruments so that there would be some similarities in the data that are collected. As you can see, the designs are very, very similar. So I think that is a good point about potentially getting more statistical power, particularly if we are looking at the gene–environment interactions question, which does require large sample sizes. Dr. Schwartz: But I am really asking a different question here. As someone who gets to weigh in or at least provide an opinion about what should or shouldn’t be funded, it is always difficult for me to know what is being done, as opposed to what is the new research that is being proposed to be done and how that complements, overlaps, or meets an area that is not being addressed. And if it is not being addressed, it all of a sudden takes a higher priority in terms of funding. Is there some attempt to do that among these ongoing studies that haven’t yet come to fruition? Art Beaudet pointed out a very important point, which was, what is the incidence prevalence over the past 20 years, and how has that changed? 

200 AUTISM AND THE ENVIRONMENT Dr. Newschaffer: There is another point that you need to consider in this. I know that you are very cognizant of this, being in the position that you are in. It is not only the questions that aren’t being addressed; it is the ability of the data and the study designs to answer those questions validly and accurately. That has to be part of the matrix as well. Some of these questions, they are real fundamental challenges to doing that. Dr. Hertz-Picciotto: Let me give some specific examples. I think we are going to have a very good ability, and we have already started looking at those data, and there are some hypotheses that are arising, some associations that we are seeing, to look at some of the perinatal factors when we can get the medical records, and we are working very hard to beef that part of the project up right now, because it is not easy to always get these records. But I think that is an area that is going to be well covered. I think between us and the CDC study, that is an area—not that we have all the answers now, but I think in the next 5 years we are going to get a lot of information on the prenatal and periconception and perideliv- ery part of the period. Those are good data. Now, when it comes to environmental exposures, as you know, the problem is how do you reconstruct exposures that happened in the past. And questionnaires are one of the tools we have. Some things can be remembered well and some things maybe not so well, and maybe more variable and more subject to the state of mind the person who is trying to remember is in. What can we measure? In California we have the blood spot that can be used to measure metal levels. We can also measure cytokines. We cannot measure any transcriptome, we can’t measure many of the other—we can’t measure cell-specific functions in immunology, for example. So the retrospective studies have those certain limitations in terms of prenatal and early postnatal exposures that fundamentally limit case control studies. Those will be addressed in the Norwegian cohort, the National Children’s Study, and the high-risk pregnancy cohorts of mothers in the MARBLES (Markers of Autism Risk in Babies— Learning Early Signs) study in the early network. The problem is there are so many things that are so open, that we don’t exactly know where to go at this point. Dr. Schendel: I guess I want to reiterate the point that I made at the beginning of my presentation. When we conceived the CADDRE 

PROCEEDINGS 201 program in 2000–2001, as I said, it was still an evident deficit in the number of federally funded projects with a specific epidemiologic focus. With the exception of CHARGE, there were no others. So we started from ground zero essentially, and we have tried to plan a study that was, we thought, capable of addressing some specific questions, but yet broadly, broadly, broadly, broadly conceived. I do think there is a continuing deficit of these kinds of studies. The Norwegian cohort, although that is not a U.S. study, will fill that. The National Children’s Study will fill it, but not any one study can do it. So I think to advance the kinds of questions that you are proposing here, given the challenges in data collection, we do need a more concentrated effort and expansion of these particular types of studies, because one study cannot do it all. Dr. Falk: Dr. Schwartz, let me try and summarize just for a minute. David and I have had similar kinds of conversations lately over various issues from Hurricane Katrina to global environmental health. The issue for us often is what can CDC do, what can NIH do, where are the gaps? I think since these studies were designed perhaps 6, 7, 8, 10 years ago, maybe it is worthwhile, as David Schwartz is suggesting to think about what is covered by the current group of studies and what is perhaps missing, and some suggestions to that. That might be worth maybe coming out of this afternoon’s discussion with some suggestions about that. So maybe we can all think about the current studies and in the afternoon discussion we will come to that. Dr. Leshner: Very briefly, if you can, what I was going to suggest is, listening to this conversation and again not knowing a whole lot about the subject, this is clearly an unresolved gap and one that ought to get included in what we talked about. Ms. Bernard: I want to bring up gaps. So you are saying wait until this afternoon to talk about gaps in the research? Dr. Leshner: Well, we have had a long discussion about one gap, but I think there are other gaps that we ought to be talking about more this afternoon. I’m afraid of getting started on another topic and then being unable to give anybody a break. 



Session VI Technology and Infrastructure Needs for Future Research Dr. Falk: This session will be focused on technology and infrastruc- ture needs for future research. The very first speaker is going to be Larry Needham from the CDC Environmental Health Lab. He will be talking about body burden measures. Dr. Needham is the chief of the Organic Analytical Toxicology Branch of the National Center for Environmental Health at CDC. He served at CDC for over 30 years in the area of assessing human exposure to environmental chemicals through biomonitoring, authored or coauthored over 400 publications dealing with multiple chemicals, and has been very much involved and is past president of the International Society for Exposure Analysis. CDC ENVIRONMENTAL HEALTH LAB— BODY BURDEN MEASURES19 Dr. Larry Needham Dr. Needham: Today I will emphasize exposure to stressors, particularly environmental chemicals that may have the potential to lead to autism spectrum disorders. In 1989, the National Research Council first set forth the concept of the exposure–effect continuum, which traces an environmental chemical or some other stressor from its source into the environment, then to human exposure, and finally to a possible effect. Exposure means contact of the stressor with the individual. The portion of that exposure dose that gets into the body is termed the “internal dose.” Downstream from that, a particular amount may go to a target organ, and a portion of that may then become the biologically effective dose. After various pharmacody- namic processes have taken place in the body, adverse health outcomes may or may not occur. 19 Throughout Dr. Needham’s presentation, he may refer to slides that can be found online at http://www.iom.edu/?id=42473. 203 

204 AUTISM AND THE ENVIRONMENT Thus far in this meeting we have talked about various environmental stressors. These can be roughly categorized as chemicals, infectious agents, or social factors. By chemicals, we mean environmental chemicals; so-called “occupational” chemicals (these are often the same as environmental chemicals, but generally are found at higher concentra- tions in people); pharmaceuticals; personal-care products; diet; and household products. The individual stressors that we have discussed include genetics and personal characteristics such as age, nutritional status, health status, and sex, and many others. In addition to the environmental stressors and individual stressors, we have talked about autism spectrum disorders occurring in a popula- tion mostly from the confluence of environmental stressors and individual stressors that affect the fetus through the mother or affect the infant directly. Within such a population, after various pharmacodynamic processes have occurred, there may be a range of effects, which may (1) involve no autism spectrum disorders at all; (2) may lead to the development of autism spectrum effect biomarkers, but not to any autism spectrum disorder; or (3) in fact, may lead to cases of autism spectrum disorders. I’ll come back to this slide in a few minutes. These next two slides show the effects of multiple stressors in a series of experiments conducted at the University of Pittsburgh. This slide shows an aquarium, a tank. In that tank researchers put 10 liters of water and 10 tadpoles. Adjacent to that tank was another tank in which a predator (a stressor) may or may not have been present. Researchers added to the water in the first tank the pesticide carbaryl (another stressor) at two concentrations in acetone and just acetone as well, all with and without the predator next door. Very quickly, the results of that experiment: They found an almost complete survival rate of the tadpoles when only water was in the tank as well as when there was only water in the tank and the predator was in the second tank, as shown here by black circles. With acetone control, there was also almost complete survival of the tadpoles regardless of whether or not the second tank held the predator. When there were high concentrations of the pesticide carbaryl in the first tank, they found that the tadpoles survived for about 5 days, and then almost all of the tadpoles died, whether or not the predator was in the second tank. At lower concentrations of carbaryl, however, there was a difference. This slide shows the effect of multiple stressors. With just 

PROCEEDINGS 205 the carbaryl at 0.05 milligrams per liter, there was about a 50 percent survival rate after 10 days, but with the carbaryl at 0.05 milligrams per liter and the predator, there was almost a complete demise of the tadpoles. This very simple experiment showed the effect of multiple stressors. This slide shows the categories of various chemicals that people may be exposed to. It is taken from a 2006 publication on recent EPA data. I don’t know the year of the EPA data. What it shows is that in commerce in the United States, there are 82,000 chemicals. About 40 percent of those are polymers; many polymers, following exposure, are not absorbed by the body. So even in those situations in which people may be exposed to polymers, but the polymers are not absorbed, there will be no adverse health effect. In addition to the 82,000 industrial chemicals, there are about 8,600 food additives, about 3,400 cosmetic ingredients, between 1,800 and 2,000 pharmaceuticals, and about 1,000 active ingredients in pesticides. That’s about 100,000 chemicals. So where do we start in assessing human exposure to chemicals that may be linked to autism spectrum disorders? At CDC we are interested in biomonitoring, but we also recognize that biomonitoring is not a stand-alone tool, so we also use questionnaire data and historical information. We also especially need pharmacokinetic data so we can interpret the results of biomonitoring, and we should not underestimate the importance of environmental monitoring. In an exposure assessment, we try to combine all three of these approaches with calibrated and validated models to assess exposure to environmental chemicals. About every 2 years, CDC also publishes its National Report on Human Exposure to Environmental Chemicals. This is a copy of the Third Report released in 2005. Human samples for the Fourth Report are being analyzed in our laboratory now, and the actual document is slated for release in 2008. Some of the chemicals that we have measured are the very chemicals that we talked about in this meeting. These chemicals include metals, PCBs, dioxins, furans, and organochlorine pesticides. Organophosphorus pesticides have been mentioned here as well as polycyclic aromatic hydrocarbons and phthalates. Some of the newer chemicals in the report include the brominated flame retardants, which include the polybromi- nated diphenyl ethers; perfluorinated chemicals; triclosan, a bactericide, 

206 AUTISM AND THE ENVIRONMENT which is added to many personal-care products; sunscreen agents; and parabens, which are used as preservatives in foods and in other products. Choosing the appropriate matrix, such as blood or urine, for biomonitoring depends on the chemical, but the choice is also population dependent, involving such demographic characteristics as age, race, and health status. Broadly speaking, there are two primary classes of environmental chemicals. The first is those that are persistent, having half-lives in years. These include dioxins; PCBs; some of the polybromi- nated diphenyl ethers; some of the fluorinated chemicals, such as PFOS and PFOA; organochlorine insecticides; and lead in various biological stores. For these chemicals, we can perform the measurement now and get a good indication of what the levels may have been years ago if the exposure was somewhat continuous. At the other extreme are nonpersistent chemicals, which have half- lives in minutes or hours. These chemicals include phthalates; contempo- rary pesticides such as organophosphorus pesticides, carbamates, and pyrethroids; and volatile organic compounds. Again, where do we start? This slide was prepared for the National Children’s Study. I was the federal co-chair on the Exposure Work Group. We were looking at exposures from preconception to 21 years of age. For studying environmental causes of autism, we are concerned not only about exposures that occur during preconception, but also exposures that occur during the three trimesters of pregnancy, as well as those that occur in infants and toddlers. The point here is that the biological matrix that one would use or that is available for assessing exposure during these different periods varies a great deal. The Work Group assembled a series of slides to examine various classes of chemicals. This slide depicts monitoring persistent organic chemicals at various life stages. On the left-hand side, we listed the different biological matrixes that might be available to us. Then we looked at various life stages, for example, at adult preconception, each trimester during the fetal period, age zero to 1 year, 2 to 3 years, 4 to 11 years, and so forth, and then ranked the potential for monitoring persistent organic chemicals in these various matrixes at these various life stages. For example, for adult preconception, we would prefer to assess exposure to persistent organic chemicals in whole blood, serum, or plasma. If those matrixes weren’t available, we could still get a good indication of what the levels were in the mother by analyzing the cord 

PROCEEDINGS 207 blood. If we are looking at exposures during the fetal period, in general we can’t get fetal matrixes, so to assess exposure to persistent organic chemicals during this critical time, we can use cord blood or we can use maternal blood, either during pregnancy or after the mother has given birth. Clearly, there are various ways to use biomonitoring during these various life stages for assessing exposure to persistent organic chemicals. We have a similar series of slides that include nonpersistent organic chemicals, metals, VOCs, and other chemicals. These slides are included in the publication that is referenced at the bottom of this slide. Again, where do we start? We could begin by listing those chemicals among this list of 100,000 chemicals that (1) are absorbed by body, (2) are increasing in the environment, (3) are increasing in concentrations or in prevalence of detection in people, (4) are neuroreactive, (5) cause oxidative stress, or (6) have structures similar to chemicals that have been linked to autism spectrum disorder. In listing some of the chemicals that have been linked by various groups to autism spectrum disorders, there is no obvious relation among their chemical structures. Thus, using the similarity of structure for selecting potentially active chemicals may not be a fruitful means of getting to the chemicals of concern; perhaps, we should look at some of these other potential starting points to begin our search for chemicals that may be involved in autism spectrum disorder causes. We also need to gain information on what chemicals to measure in the body by looking at epidemiological studies. Today we have heard about the nested case-control studies, the sibling and twin studies, and the National Children’s Study. CDC is also heavily involved with several of the national children’s centers that are funded by EPA and NIEHS, again looking at exposure assessment and effects. In addition to doing this targeted analysis—one advantage of targeted analysis is greater sensitivity―there are also advantages for casting a broader net. One way to do this, which is in a research mode now, is getting improved resolution by using something like 2-D gas chromatography or HPLC, and then coupling that process with mass spectrometry. Doing so allows us to acquire full-scan information for chemical identification on all the chemicals that are coming out of the chromatographic system as opposed to selecting the chemicals to measure prior to the analysis. What we would do is take a biological sample and do minimal workup. We don’t want to remove any chemicals, neither exogenous nor 

208 AUTISM AND THE ENVIRONMENT endogenous. Then we would use high-resolution chromatography, such as 2-D gas chromatography, and obtain a full-scan mass spectra for all the chemicals in that sample. This technique allows us not only to look at individual chemicals on a semi-quantitative basis, but also to look at patterns of chemicals. For example, we could potentially look at people with autism and see whether their chemical patterns differ from people who do not have autism. We also can gain much from case-control epidemiological studies by working backward. By that I mean, instead of working from environ- mental stressors down to health effects and if we could come up with some biomarkers of effect in autism cases and work backward on this continuum, then it might lead us to various environmental stressors. I also want to address how concentrations of chemicals in a particular matrix reflect body burden. This slide shows three routes of exposure and absorption: ingestion, inhalation, and dermal contact. There are very good equilibrium data for concentrations of persistent lipophilic chemicals, such as dioxins and PBDEs, between the lipids in the blood and lipids in adipose tissue and in the lipids in breast milk of lactating women. For lipophilic compounds, persistent chemicals, we can estimate the amount of body burden by doing blood measurements. For lead, the correlation between blood and bone concentrations is a little more difficult and depends on factors such as age. For fluorinated chemicals, like PFOS and PFOA, which are persistent, there is not an equilibrium between the concentrations in blood and fat because those chemicals are not lipophilic. But there are some data on the equilibrium between blood levels and protein in the liver, so in general we can estimate the body burden for persistent chemicals. For nonpersistent chemicals, making estimates is much more difficult. There are some modeling efforts that are going on now to help ascertain some of the body burden measurements for nonpersistent chemicals, but going from a urine concentration to the total amount of a chemical in the body is difficult. In conclusion, biomonitoring techniques are available for assessing human exposure to environmental chemicals; however, we need your help in determining the chemicals or classes of chemicals to measure. I will entertain some questions. Thank you. Dr. Falk: Thank you. We have a few minutes for specific questions for Larry. 

PROCEEDINGS 209 Ms. Bono: Thank you, Larry. I liked your chemicals in the fourth report that you are looking at blood data. How are you going to be measuring metals? Dr. Needham: That is in another group. What they do there is lead, cadmium, and mercury are measured all the way down to 1 year of age in blood. Then there is a suite of metals left that are measured in urine. Ms. Bono: I am concerned about that, especially with this group of kids. You might be able to get it on the population basis, but again blood doesn’t show metals with these kids, or urine, unless you use a provoca- tive agent. I know you mentioned just now protein in liver, which I am not really familiar with that much, or even bone. That gets into more invasive types of things, so there is a limitation there. Dr. Needham: What we talked about are nonexcreters. So what you are saying there is that neither the blood level nor the urine level may reflect what the body burden is. Ms. Bono: Just based on thousands and thousands of these kids and what we are seeing, unless you use a provocative agent. Also, you had a great slide on where to start, and you had the list of chemicals that cause oxidative stress and several things. I think it is important to add to that list what chemicals are injected, because we have a problem right now with the mercury, aluminum, and other additives in vaccines. There are all sorts of other additives. So when we look at the children’s exposures, we also need to look at what is injected as well, because that is going right into their bodies. Dr. Needham: That would be a route of exposure in addition to inhalation, ingestion, and dermal contact. It would be what is actually injected into the body. Ms. Bono: Correct. You need to have that injection, very important. Thanks. Dr. Needham: Thank you. Dr. Herbert: I want to make a comment. I would like to get back to the discussion, but I want to link it to your slide on working backward from effect markers. I think it is important to think about how we determine what those effect markers may be, the ones that we think are pertinent. This also ties into my thoughts about the distinction between saying that it is possible that these various substances may affect the organism in utero or during the perinatal period, versus that that is the only time 

210 AUTISM AND THE ENVIRONMENT that they are going to affect the organism. There is also the question of body burdens. These chemicals may come into the organism at a critical period and affect neuronal develop- ment, but then they may hang around and keep giving their gift and affect the ongoing ambient environment. That sets up a whole different kind of class of effects that would be missed if you are biased entirely toward looking at altered neuronal development. Whereas some of the variability in the presentation of these children day to day is related to the borderline chemical immunological status that can be affected even by some of the transient exposures that occur substantially postnatally. So I think that the idea of the effect marker needs to be expanded to take account of that kind of pathophysiology. Dr. Needham: Okay. I think what you also mentioned was the potential for persistent chemicals especially, but also for nonpersistent chemicals because a portion of those are still sequestered. For example, the persistent chemicals may go into the fat and then leach into the blood, and what you have is continuous exposure. Dr. Herbert: Right, over life course. Dr. Falk: I might just add briefly that the biannual reports are from a representative sample of the United States, and it is a snapshot in a sense. But the laboratory also works at any given time on 60 or 70 different studies with various groups, some of those that Dr. Needham has mentioned, to fill in some of the additional kinds of information over and above the baseline national report. Dr. Needham: We do collaborate with other investigators with federal agencies, like NIEHS, David Schwartz and his folks, and EPA and others as well, academia. Dr. Falk: Thank you very much, Larry. The next speaker is David Walt, who is going to be speaking about personalized environmental sensors. David is professor of chemistry at Tufts University and Howard Hughes Medical Institute Professor. He served as chairman of the chemistry department from 1989 to 1996. He has been executive editor of Applied Biochemistry and Biotechnology, and is the statistical founder and director of Illumina. Thank you very much, David. 

PROCEEDINGS 211 PERSONALIZED ENVIRONMENTAL SENSORS20 Dr. David Walt Dr. Walt: Thank you, it is a pleasure to be here. What I am tasked to do is tell you what is possible, not necessarily what you are doing today. I categorize measurements in two general categories: first, assessing what is there, and second, discovering what needs to be measured. This latter category will be discussed a little bit later in terms of being able to identify what needs to be measured and what is technically feasible in our ability to make measurements. We have a number of different ways of assessing any kind of clinical problem. We can measure the environment for what is present. We have the possibility of using personal sensors or dosimeters for measuring drug doses as well as dosimeters for environmental exposure. During this talk, as we transition toward the future of measurement technologies, we will see a trend toward the molecular level, looking at single molecular lesions, developing nanodosimeters that can circulate in the bloodstream for identifying body burden, and we will also see how we can use molecular methods for presymptomatic and early diagnosis. This slide shows an environmental sensor that we developed in my laboratory that has been deployed on a buoy. This buoy is just off the coast of Massachusetts, right off Martha’s Vineyard. I want to point out that there is a sensor on the buoy that is making measurements. You can see occasional spikes in the data. These spikes are not noise; they are actual spikes of something that is being measured in the environment. If you take a sample of seawater, or for that matter just take a sample of blood and make a single point measurement, you are going to miss all this variation. These data underscore the importance of being able to measure things continuously. The ability to measure continuously is what defines sensors. We are not talking about sensors when we talk about making a single measure- ment. We are talking about things that can make measurements continuously. From what I have heard here today, the ability to make measurements continuously is essential. This next slide shows a sensor used for military applications. It is detecting things that are in the air. These are continuous sensors. They 20 Throughout Dr. Walt’s presentation, he may refer to slides that can be found online at http://www.iom.edu/?id=42474. 

212 AUTISM AND THE ENVIRONMENT are making spectroscopic measurements of things that are in the air path. This system is pretty large, is deployed on a truck, and requires lots of power. If we are interested in environmental exposure and we want to know what is being released, some of these systems are going to have to be implemented in both urban and rural environments to get a handle on what is there, what is their temporal variation, and what the concentra- tions are in a variety of urban and rural settings. Large, complicated, power-hungry instruments may be one way to get to these answers in the near term and should be considered. The next slide shows analytical test strips. Such strips are a low-tech method. One can do pregnancy tests this way. We have been performing clinical measurements of renal patients who spit in a tube, dip test strips in the tube, and measure various components in saliva that previously have been measured in the blood. It is a relatively noninvasive method for detecting whether these patients need to come into the clinic and have dialysis. We are all familiar with glucose tests. This type of measurement relates to one of the points that I made earlier, that is, it is a single measurement. If you eat three meals a day and your blood sugar is at a hyperglycemic or a hypoglycemic level and you take your blood measurement three times a day, you are probably going to miss these extremes unless you just happen to sample at precisely the right time. This next slide shows what people really want to address the glucose monitoring problem. This system is commercially available and is called a GlucoWatch. This kind of technology can be brought to bear on a wide variety of substances. I am not going to talk much about it, but essen- tially what it does is, it causes the patient to sweat, and there is a little sensor on the underside of the watch that measures glucose continuously. It doesn’t do a great job of measuring glucose accurately but it gives a trend, and it tells the patient when they need to do something, such as take insulin or eat something. There is a wide range of things we heard in the previous talk that can be measured with laboratory-based instruments. These instruments are getting better and better. They are getting much more sensitive. We can’t ignore the fact that while these instruments make measurements of single time point samples, they have the ability to measure lots of things. Sensors and new measurement capabilities are enabled by new materials. I just want to illustrate a couple of these. 

PROCEEDINGS 213 This next slide shows colorful solutions of things called quantum dots. They are tiny little balls with nanometer dimensions. They are presumably inert, although we don’t want to use these things in vivo right now because their toxicity has not been fully tested. As a research tool, however, they can be of immense value. As can be seen, quantum dots glow in many different colors. They can be incredibly powerful labels. Here is an example of three different kinds of cancer cells that were injected locally into a rat, and then three different kinds of nanosensors with different specificity to these three different tumor types were injected in the rat’s tail vein. The nanosensors went through the circulatory system and localized very precisely at these three distinct locations. This next slide depicts molecular science. One can now analyze molecules at the single-cell level. These examples show some of the capabilities that are being brought to bear using nanotechnology. The next slide shows another application of nanotechnology. These triangles are nanogold prisms attached to a surface. One can create immunoassays with these materials that have the ability to measure as few as 600 molecules. This capability is unprecedented in terms of the level of sensitivity achievable with traditional immunoassay techniques. Let me now talk a little bit about taking the laboratory and bringing it closer to the patient or to the study subject. These devices are called total sensing systems. This next slide shows a technology that I think will genuinely enable the field of autism research, and that is new sequencing methods. These sequencing methods can sequence literally an entire human genome in a few weeks for about $100,000. Compare that to $2 billion for a human genome 6 years ago. This trend is going to continue as we move toward $1,000 per genome. If you don’t need a whole genome sequence you can measure lots of sequence very cheaply for hundreds of dollars a sample, and you can get tremendous amounts of information about variability, literally tissue-to-tissue variability, as well as even cell- to-cell variability. This next slide shows a new technology that has been developed over the last decade called lab on a chip, also referred to as microfluidics. All of the capabilities of an entirely analytical laboratory are being integrated onto a substrate about the size of a microscope slide. You are probably all familiar with these slides, about 3 centimeters by about 12 centime- 

214 AUTISM AND THE ENVIRONMENT ters. These chips can perform many different analyses simultaneously with incredibly small sample volumes, and collect a large amount of data from each sample. Other promising technologies are listed in this next slide, for example, miniature mass spectrometers. We heard about mass specs already. Now, mass specs are getting to the stage where they can be taken out into the field and perhaps even left unattended to be able to perform analyses on a minute-by-minute basis. Now let me finish the technology side before getting into needs and opportunities by talking a little bit about two technical advances. One advance is the ability to measure single molecules and the second advance is the ability to measure lots of things simultaneously using array technology. Arrays can perform lots of experiments and make many measure- ments simultaneously—a capability that is important in any field, particularly the clinical field. What I have heard today is a lot of bias; that is, you know what you want to measure. I don’t think you necessar- ily want to work that way. You want to measure as many things as you can and use these new technologies to discover things that might lead you to find some new markers—you want to pinpoint some new leads in these kinds of diseases. Technologies exist that enable us to look for rare events, to isolate and identify things that don’t happen often, but when they do happen they lead to a problem. Here is just a very small picture of a type of array that we use in my laboratory. The array itself is about 2 millimeters in diameter. All the spots on this array would take up the entire front of this room. This image is a very tiny section of the array. We can measure hundreds of thousands of different sensors simultaneously. We are not the only ones who can do these types of experiments. This example is where technol- ogy is today. You can measure thousands of things every time you take a sample and you don’t need a large sample. A single drop of blood will enable you to measure thousands of different analytes in that sample. So the take-home message is that you can multiplex. You can and should want to measure lots of things in every sample. Here is another platform. This platform measures thousands of things in nearly a hundred samples at once. Each of these little sensors is able to mate with a well in a 96-well microtiter plate with a different sample in each well, so you can measure lots of samples and measure thousands of things simultaneously. 

PROCEEDINGS 215 This slide shows some work that we have recently done in the laboratory, measuring single enzyme molecules. I’m not going to tell you how we did this experiment as it was recently published. Each of these little red dots is a single enzyme molecule. We measured the kinetics of each of these enzyme molecules. What would conventional biochemistry predict? It would predict that every enzyme is identical. In fact, every enzyme is not identical. This graph shows the kinetic traces of single enzyme molecules. Why are they all different? Some possible reasons are that there are amino acid substitutions in the proteins and because even identical proteins adopt different conformations. These features give rise to an entire diversity of activities in a population of enzymes. We never see this variation when we look at bulk ensemble measurements. We only see them when we look at single molecules. We talk about lesions when we discuss disease. There are a lot of lesions here. Every one of these enzyme molecules is different from every other enzyme molecule. Some of these lesions may lead to disease, while others may simply be in the normal range of biological variation. The next slide shows a new technology that has recently been published in Science. It is able to look at a single lesion in a particular sequence of DNA with very high specificity—a single molecular lesion! Here is an example of a technology that can detect mutations at the single-cell level. What are the challenges and opportunities? We have to decide what we need to measure. But what I’ve tried to tell you is that you don’t need to decide all that much, because you can measure a lot of things. I would rather err on the side of measuring lots of things by spending a little more money, and then worrying about what to measure after we have measured a lot of things and got some leads. Sample collection—I am not going to belabor this point as there are epidemiologists in the audience and they know more about this topic than me. But it is important to remember that when you take small samples you have to worry if the thing that you want to measure is actually in that sample. Sample types—we can and probably should move toward much easier types of samples to collect. Better sensor sensitivity is still important and is worth investing in. It is also important to expand the ability to look at many things by using these new kinds of techniques. We still need increased data processing 

216 AUTISM AND THE ENVIRONMENT capability and data integration. As we collect more data we are going to need to process it and get information out of the data. Then there are some ancillary issues such as power for driving these devices, including new battery technology. We also need to consider new deployment strategies such as putting sensors in cell phones, and we need to think about system integration. It is important to specify the analytes. We can consider using surrogate measurements for the analytes that we cannot measure today. For example, there are things that can be measured easily that might indicate that there is something present that cannot be measured readily. Some sensors are available that can be used today for monitoring. Continuous measurement capability is going to be essential, where you can measure things continuously, rather than take samples and send them to laboratories. Nanotechnology is going to advance measurement science—it is unavoidable. We are getting better and better at being able to measure things, and because of this ability, it will enable lots of fields, including the field of autism research. I’d be happy to answer any questions. Dr. Falk: Thank you. Any specific questions? We will take two or three. Dr. Wilcox: David, I just want to emphasize a point that you made. In our cleft study we recently used an aluminum panel through CDER to look at 300 genes of 5 SNPs each, and it was very successful, but there is good news and bad news in there. The good news is, we now have genetic information on 1,200 SNPs, and the bad news is that we now have genetic information on 1,200 SNPs. Our capacity to use all this information that is now possible to generate is way behind the technology to produce it. I know you mentioned data integration there, but I think we can’t understate how important it is for the informatics to catch up with this, and we are not there yet. Dr. Walt: Absolutely. Dr. Pessah: When you referred to thousands of analytes, were you referring to proteins, DNA, or also small molecules? Dr. Walt: Yes, protein, DNA, and small molecules. I think it is critical that we address all three of those classes. Those are the three important classes. 

PROCEEDINGS 217 Dr. Pessah: So I wonder, when you can go to measuring with validation and with high accuracy and sensitivity hundreds of thousands of analytes, could you reduce the sample population? If we believe that autism is on the individual level, can you gain important information by not going out and getting 2,000 individuals into the study, but maybe 5 individuals in the study? Dr. Walt: I leave that to others to answer. Dr. Pessah: It’s a thought. Dr. Hertz-Picciotto: It depends what the variability is in the population in those 12,000 or 20,000, whatever order of magnitude we are at. If there is low variability and we are looking for some signal, it is going to be extremely strong, then yes, maybe five cases and five controls. But most of what we are looking at is probably not going to jump out of this quite in that way. Dr. Walt: Or at least, that is what we think. Dr. Spence: Given the urgency question that keeps coming up here, how much does it cost, and how quickly is the price coming down? Dr. Walt: That is a very good question. One of the things that turns out to be a little bit counterintuitive is, as things get smaller they get cheaper, because the materials cost goes to virtually nothing. When you talk about nano materials, when we make sensors, we can make about a million sensor arrays with a gram of material. A gram of material is a powder in a bottle about like this. That is because we are using microsen- sors. When you get to nanosensors you can make billions of sensors from a gram. So I am simplifying it a little bit. The supporting instrumentation that goes with reading it out still has to be developed to a level that matches the capabilities of the small sizes of the micro and nano systems. Dr. Falk: Let’s open this now for discussion of both papers. Dr. Schendel: I just wanted to follow up with that. For these technologies, how do you know that you have a sensitive and specific measure? I know in some technologies, you may get signals, but you are not exactly sure that the signal is what you think it is. Dr. Walt: Obviously for any analytical test, you have to test the specificity and the sensitivity. You do that with conventional methods of first testing it in buffer and then adding interference and then going to your relevant samples like blood and urine and whatever and seeing if 

218 AUTISM AND THE ENVIRONMENT the measurements that you are taking with those matrixes are any different from those that are in pure buffer. But I think a more important and more relevant answer has to do with just the advantages of making thousands of measurements. The fact is, when you make thousands of measurements, and this speaks to some of the study that you did with CDER, one data point does not give you an answer. It is looking at the pattern of response that helps you identify what the overall process or the overall change that is going on in that sample is. So the example is, there are lots of correlations typically when you make thousands of measurements. One thing may go up, another thing may go down, and if you make a number of measurements, you might find that if this one doesn’t go up at the same time this one is going down, you know that you have made an incorrect measurement. So there are correlations within those thousands of analytes that you are measuring that get you away from needing to be as specific as you would be if you were measuring just one thing or two things. Dr. Needham: His question is related to the last question. As far as sensitivity goes, as you know, David, oftentimes the analyte present in urine or blood is picograms per milliliter or nanograms per gram and so forth. One thing that you showed was improved sensitivity as a need. So what kind of sensitivity is available for measuring these small molecules? Dr. Walt: The small molecules, you are talking about organic molecules? Dr. Needham: Right. Dr. Walt: Organic molecules, still the gold standard I would say is GC mass spec, which is a laboratory-based type of system. But once you have identified which of those organic molecules in laboratory-based settings turn out to be interesting, then you can spend the time and energy to design sensors that can address either that class of molecules or specifically that molecule with conventional kinds of measurements. In terms of the sensitivities, as I have showed you, Richard Van Duyne, who is at Northwestern, is using these gold nano particles, these little nano prisms that are on surfaces. He just gave a seminar at Tufts the other day, and he presented data that showed that he can get to some- where on the order of about 10 to 12 molecules of sensitivity for small organic molecules. 

PROCEEDINGS 219 So this is not science fiction at this point, but it is not something that is in the general public domain for routine kinds of measurements. Dr. Beaudet: I just wanted to point out that the kinds of techniques that were displayed there in the arrays are the kind of techniques that are being used for array-comparative genomic hybridization. These things are in the clinic. They are not in research, they are in the clinic on routine use, and they are not significantly more expensive than other genetic tests. And it is diagnosing tons of stuff. Dr. Schwartz: I just wanted to point out that what David presented is part of what we call at the NIH the exposure biology program, which is part of a bigger initiative called the genes and environment initiative. It is a trans-NIH initiative that all the Institutes are contributing to, that Dr. Zerhouni supports, and also Secretary Leavitt helped develop. Francis Collins at the Genome Institute is leading the genetics program within the genes and environment initiative, and NIEHS is leading the exposure biology program, but it is really a trans-NIH program to develop sensors, both personalized environmental sensors and biological responses to environmental forms of stress, that allow us to then take those measurements and move them quickly into studies of populations and individuals that are at risk of being exposed. So NIH is investing in this. The question I have for you, David, is how is industry investing in this, and what are the roadblocks to have industry invest in this area? It seems to me to be something that would be an obvious opportunity for industry. Dr. Walt: I’ll have to give you a generic answer as opposed to specific examples, but the generic answer is that there are a lot of companies that are exploring new measurement technologies. Many of them are doing things that probably could be relevant to this community, but wouldn’t be. So those companies need to be enticed to be able to address problems that have a relatively narrow focus, but that they have to pay attention to those problems because they provide in many cases non-equity funding that enables them to leverage their existing funds, and to do essentially the same thing that they are in business to do, except to focus on a problem that is of interest to a particular community. So I would say that the biggest problem with getting industry involved is that they have to be enticed to be distracted from their mainstream opportunities. 

220 AUTISM AND THE ENVIRONMENT Dr. Goldstein: I’d like to ask the same question of federal funding, NIH funding, this new initiative which is gigantic and has obvious implications to cancer and heart disease and whatever, how will the autism research community be able to be early in line to be able to use these new approaches? How do we position ourselves? I think it raises lots of issues that we might go into later in terms of recruitment of talent to make use of this technology and committing to autism research. Dr. Schwartz: This gets at Tom Insel’s question from yesterday. You have 200 assays, but you have 49,000 chemicals out there, how are you going to measure—are these 200 assays good enough at this point to move forward with? Should we wait for the 49,000? Several years from now, hopefully within the next 5 years, maybe within the next 10 years, we will have assays for several thousand as a result of this initiative. Those assays, as soon as they develop, will be made available to investigators. However, like any assay, even the genomic assays and the genetic variation assays, the SNP assays, come with a price. It costs investigators money to actually do the studies. But we are in the market of trying to support investigators to use those tools as much as possible to apply to the research. As part of the genes and environment initiative, there is a component in 2009 and 2010 to use these assays in genetic studies, so that we are looking at gene and environment as opposed to just environmental exposures or genetic susceptibility factors. So if things go the way we plan them to go, in 2008–2009 we will be putting out RFAs to provide funds for investigators to use these environmental sensors in genetic studies of complex diseases. The genetic component of the genes and environment initiative put out an RFA for any complex disease, so I anticipate that autism will be one of the diseases that has at least investigators applying for. But that is not the limit of the types of studies. Even if autism isn’t part of the initial studies that get funded in the genes and environment initiative, an autism study could certainly come in to apply for funds to use these environmental sensors as they become available to look more carefully and critically at populations to see if these environmental exposures are associated with the development or progression of autism. So I think it is a very open playing field right now, and will remain an open playing field in terms of allowing investigative groups to use these tools to be developed. 

PROCEEDINGS 221 Is that what you were asking, Gary? Dr. Goldstein: I think so. The field has been able to attract really top geneticists who use most top technology to be studying autism. I think we have to anticipate not only the money to do these studies, but the talent to plan and interpret these studies, whether our field has antici- pated them, who is going to take advantage of them. The obvious application this whole 2 days is all about gene– environment interactions which seem perfect. But I think the point was made that probably every environmental researcher in the country is in this room right now, and there are very few of them. So the balance is how we can participate in this. Dr. Schwartz: I think you are raising a really great point, which is, do we need to think about training as part of what we are considering here? Do we need to think more broadly about the pipeline of investiga- tors that could undertake this work? Dr. Walt: I want to address both of your comments. Getting to a more specific answer, David, to your question, the companies that are exploring some of these really new technologies are focused primarily on cancer, and maybe on heart disease, because they are huge markets. The ability to detect early stages of CVD, for example, would revolutionalize the treatment and care of that disease, and early detection of cancer may or may not affect outcomes, but at least there is a perception that if you can detect it earlier you can do something about it, and you are going to make a difference. How do we entice companies that have these technologies to now spend their time looking at environmental exposure for autism and other diseases? Again, it is a tough question. The genetics tools that are available are in fact not particularly relevant. It is an example of technology that has now become mature and is available for any disease. You decide you want to study autism, it is there. There are whole genotyping arrays; there are whole gene expression arrays. They cost you a few hundred dollars in experiment, maybe a little bit more, but not that much more, and the price has gone way down compared to what it was 5 years ago. So the question is really, how do we draw in those new technologies that can jump-start the studies of a disease without having to wait for them to get mature? There is one answer, and that is money. It is to get their interest diverted from something where they have to compete with a lot of other people to something where there is going to be a market, but 

222 AUTISM AND THE ENVIRONMENT it will enable the technology to develop. It is just that they are going to develop that technology devoted to solving an autism problem as opposed to cancer, and then they apply that to cancer once they are a year down the road. Ms. Singer: Are these technologies being designed to produce data that are compatible with all of the databases that we are investing so heavily in, and the NDAR and the IAN projects and all of the parent registries? Or is it going to be data that are in isolation? Dr. Walt: I think certainly on the side of the genetic databases, and now there are protein databases, there are issues with validation that are being addressed. On the genetic side they have to be compatible. On the protein side, there are similar kinds of movements that have happened over the last 5 years with the genetics to make sure that new technologies that come into play are somehow correlated to those existing databases. So I think that those are issues that need to be addressed. It is an important issue, but I would say that the technology infrastructure is just not there yet to have that be as mature of a question as it should be. Dr. Falk: We have a few more minutes before the break, but this maybe goes to Gary Goldstein as well as to David Schwartz, but in terms of training needs for people working with the instrumentation, but going back to Alan’s opening comment, maybe also training needs for people using all the data and informatics and so on. What exactly do you see as the main needs in terms of the kind of people that need to be trained? Dr. Goldstein: To move quickly, my sense is that the successes in genetics have not been taking people interested in autism and training them to become geneticists, but it has been by creating the incentives and the interest of the leadership in genetics to tackle these problems. At this moment, in the last few years, the very top of this field has now begun to work on autism. Otherwise we are talking about a 10-year feed-in of young people who might work out. So I think it is a question of how we get the talent that exists already to be convinced that it is worthwhile of their time to work in the autism area. That will also—these people will train people, and there will be a follow-through for the future. But I think the first step for those who want to see something happen in the next 5 years is to take people with these talents and convince them to work in autism. Dr. Newschaffer: I think the bioinformatics issues are generalizable across diseases. The data don’t know whether it is a CVD or an autism dataset. The training in bioinformatics to address the data reduction 

PROCEEDINGS 223 issues that Alan highlighted can be advanced-disease independent in many ways, I think. I don’t know how to do it, but I think it can be. Dr. Schwartz: I would just ask the mental health community how many psychiatrists or psychologists are trained in environmental sciences to address issues that are relevant to environmental causes of psychiatric and behavioral problems. I think there is an opportunity in that. I see that as a deficiency, and probably an opportunity to think about developing programs that could address that need in a very clear, scientific, programmatic way. Dr. Insel: If that was an actual question, David, the number is probably less than zero. But in terms of the more general question of what do we do, going back to Gary Goldstein’s issue, how is any of this relevant to autism, what I am hearing is something that is so similar to where we were maybe a decade ago in genetics. At that point, we were spending an awful lot on trying to genotype relatively small samples of many, many different disorders. Then about 5 years ago we said, enough of this, we are not getting any answers. Let’s just collect the DNA and bank it from as many people as we can, put more of our effort into tool development, getting the capacity to be able to do things like whole genome association, and when that capacity finally arrives, we can then move very quickly. So I think we are probably in the same framework here. The question that Gary Goldstein asked was the question that I had after looking at the NIEHS website describing the exposure biology program. It is very difficult to tell from there where anybody who has a specific interest in a given disorder will get into that. It is all about capacity building. Until you answered the question, I didn’t understand the overall plan. It is just like genetics. You decided, let’s build the tools and then when we are ready in 2009 probably, we will be able to then use them. But I think now, to go back to Gary’s issue, what should we be doing currently is collecting the samples, so that when the tools are there, we don’t have to start. We will have the biorepositories. I think the discussion to have here is, what should we be collecting, and how do you have this in a way that will be most useful? Obviously it is a guess. If we knew the answer, we wouldn’t have to be having the discussion. But I do think we need to think very carefully, so that we cannot lose 2 years while people are out building the tools that we need. Dr. Falk: I think that is a really excellent comment. Do you have 

224 AUTISM AND THE ENVIRONMENT any last comment? Dr. Colamarino: Along those lines, it was something I was thinking about from the earlier session. One of the take-home messages from yesterday’s discussion was the crucial importance of phenotyping. That may indeed be a bottleneck, but not just phenotyping across all levels, but particularly getting at the biologics, what I like to call the wet stuff. So I was pretty impressed by the fact that, although we didn’t discuss it in any depth as a program, there is this Phenome Project that is working with the NIMH Intramural Program and with the M.I.N.D. Institute, and now I am hearing a lot more about the CADDRE as they are getting up and going, and there seems to be a question of biologics. That ties into what Tom Insel was just saying, which is banking. I was really curious, listening to both of your ends, what are the efforts to coordinate those and make sure that the samples are collected in a somewhat analogous fashion, so that when these tools are ready to go, we can make use of them? So is there cross talk? Dr. Schendel: We certainly are talking, yes. We are talking, and we are banking. As I say, the collection protocols that we have developed in CADDRE hopefully will accommodate a variety of analytes to be tested. Since the phenome project is on the threshold, I think there is ample opportunity. Dr. Swedo: And phenome grew out of CHARGE. So what Irva Hertz-Picciotto has been doing in her study is the basis for what we are doing. We have added a few things, but absolutely are using the same methodology and collection. So it won’t matter whether they are from NIH intramural, M.I.N.D., or UC–Davis. They should all be comparable. Dr. Spence: I think one thing to mention, though, is that with the genetics and the repository, that is a renewable sample because they are cell lines. So we have to think about the fact that these, once they are used, they are gone. I assume that we can’t—I don’t know of any biologic way of reproducing protein in from a blood spot. So I think we have to think about that when we think about banking. It is not quite the same as the repository, where we can keep getting the cell lines and keep getting the cell lines. Dr. Insel: And the genes don’t change, so what you want to think about now is, are we missing something, is there some critical window that we need to have samples, and what those samples should be, that we will have wished we had collected 3 years from now. 

PROCEEDINGS 225 Dr. Falk: We must now end this session, but first Alan Leshner has a few words. Dr. Leshner: The next session is extremely important, and I would like you to think about what you are going to say before you say it. Drs. Martinez and Pessah have agreed to chair it. The idea here is to try to articulate briefly as many research gaps for as much feed-in toward the research agenda as we can get in. That is not to say we haven’t done a lot, but in order to make sure that we have a full discussion, let’s not have everybody say the same thing, just as a ground rule. If somebody makes a point, let’s not beat it to death, but keep moving. I want to repeat the comment that I made when we began. My own belief is that the research agenda needs to be as broad as possible and cover many items. From my own point of view, nothing should be off the table, but I would like to make sure that in the brief time we are going to have for that discussion, that we do in fact cover it. 



Session VII Future Research Directions—Discussion with Workshop Speakers Dr. Leshner: Dr. Martinez and Dr. Pessah will moderate this session, but the purpose of it is to identify research tasks. So do one of you want to start? Dr. Pessah: We were quite surprised that this would be the topic that we would have to moderate. What we tried to do, since it is a broad topic and we want to keep it somewhat logical in progression, was to identify the major problems that were brought up during the discussion. For example, how do we identify a study population, what should go into a study population, what are the comparison groups? So this necessarily involves epidemiologists, which need to identify a priori assessment tools, what sorts of medical record abstractions are needed, what sorts of questionnaires might be needed to get the most information that might in fact lead to an understanding of environmental exposures, even though questionnaires have their own limitations. Then most importantly, to define a rational and doable stra- tegy for biological sampling. One of the major hurdles is, how do you collect samples that are consistent not just across studies, but with-in a study? Samples collected at one period in time need to be comparable to samples collected 3 years later in the same study. So this is a very important issue. I am going to let Fernando take us through the first part, where the discussion has led to the initiation of case-control studies. This is CHARGE. APP has a case-control design. We have also talked about MARBLES and the National Children’s Study, which is more of a longitudinal assessment, and it is exposure based, and then a cross- sectional design. We haven’t talked much about that sort of approach. Then, are there other approaches that we can engineer that would give us even more information or more accurate information? Dr. Martinez: Just very briefly, because we discussed this yester- day, I think it would be very important for people when they come up with ideas for environmental studies to think about them in the frame- work of which type of design would be appropriate for the specific exposures they are interested in. If the exposures can be relatively easily determined, I am going to give an example, duration of pregnancy could be an exposure, certainly 227 

228 AUTISM AND THE ENVIRONMENT the case-control approach is the best. You take a group of cases and a group of controls and determine the difference retrospectively of this easily understandable exposure, or relatively easily understandable. For most exposures that are not easily ascertained retrospectively, the longitudinal approach is the best one. In other words, you start by defining the exposure and then you determine which is going to be the proportion of subjects with a certain exposure that is going to develop the condition or not. The one that we have talked very little bit about is the cross-sectional design, which is particularly useful for suspected exposures that can be ascertained at the time in which you also ascertain the outcome. I was talking about it before, and I give it only as an example, is Hispanic and non-Hispanic. That is something that could be easily done, quote-unquote easily, through a cross-sectional analysis. You take a group of subjects self-defined as Hispanic, or you can even do it by genetics, because there are now markers for ethnicity, and you determine by an objective measure what proportion of subjects have some type of autism-related disorder and what proportion do not within those particular groups. There are other designs that we could talk about, but these are the three main ones. So it would be very important when we think about what we are proposing what way you think would be the best way to do it, and some of the epidemiologists here can help with that. Dr. Pessah: So we are open for discussion. Dr. Schendel: I just wanted to suggest that another opportunity might be occupational exposures, in populations with farm workers or other highly exposed families. They may be another group of individuals to do an exposure cohort type of model that can be used in this kind of analysis. Dr. Fombonne: I’d like to come back to the issue of the natural experiments that we raised yesterday. One way to look at it is with the development of international studies in autism, to try to identify populations who have different traits, while surveying with the same methods. I gave a preliminary communication last year. One population we have found so far where it seems that autism is somewhat absent. That is the population which is Inuit, living in Northern Quebec, where we have a population of 10,000, captive population in 14 villages which are constantly surveyed by our services, and we have found no cases 

PROCEEDINGS 229 whatsoever, despite very intensive efforts to find some. So I just want to present that evidence as preliminary. I would like to confirm it, but I have some other parts of Canada with similar informa- tion, too. If so, that would be very interesting to follow through in terms of environmental exposure and also genetic backgrounds. In that case, the population is highly exposed to mercury and also highly exposed to PCBs and other kinds of neurotoxicants. So it is just to give an example of opportunities which might develop. Other opportunities would be to look at plausible sources of environmental exposure. Here we have pesticides or mercury as an example, and to find situations where there is change over time, or geographical change in the rates of exposure, and try to relate that to risk of autism. As I said this morning, there is an elegant study done in California where they use time-clustering and space-clustering methods to look at risk of autism in association with pesticide exposure. It is a very interesting finding that they have. So these methods should be followed thoroughly, because they are already cheap and easy to implement, provided that they are guided by some kind of theoretical model. I want to come back to the issue of the design. The case-control design to look at pregnancy is what we don’t want to have. The prospective provides a very wonderful vehicle to look prospectively at a range of exposures. Using a fishing expedition, they are going to get samples and look at a range of possible things, and that is fine. But I think they are also limited in the fact that they started to survey pregnancy at age 18 weeks, which if you were to devise a model in terms of the timing of exposure regarding specific exposure for autism, you would try to look at what is happening before. What we know from environmental exposure, like valproic acid, thalidomide exposure, were all in the first weeks of gestation, when it all led to autism. So I think there is a critical task for epidemiologists, to try to sample the possible exposure during the first weeks of gestation. The way to do that would be to revisit the first sibling studies, where you can have a high participation rate from families, you have an increased risk of outcome which is quite substantial, and you could look also at phenotypes, which are not only the full-blown autism, but also a variation on the same thing. I think that would be a way to enrich these studies by focusing the research on the biology. 

230 AUTISM AND THE ENVIRONMENT Dr. Newschaffer: Just to follow up on what Eric said, the MARBLES study that started in California, we have in Maryland a pilot study already funded. The early network which has been mentioned, but not really explained in any depth here, will be funded as an autism center of excellence beginning in FY 2008, and it is exactly that design. It is recruiting moms who already have a child with autism at the start of a subsequent pregnancy. It will still be a challenge to get those very, very early samples, but that is what we are going to attempt, following them through pregnancy with repeated sampling, following the child with repeated sampling until 36 months. So it is a high-risk cohort design. The pilot studies have already begun, and the full-blown study, which will attempt to recruit 1,200 pregnancies with a thousand kids followed up through 36 months of a 10-year period, funding will begin in 2008 and recruitment will begin in 2009. That is going to be done in four centers around the United States. Dr. Schwartz: Let me ask a really simple question. Are the mental health folks comfortable with the way the phenotyping is being done in the environmentally driven studies, and are the environmental folks comfortable with the way the environmental phenotyping is being done in the mental health-driven studies? Dr. Martinez: Who is going to speak for each of those? Who are the phenotypers? Dr. Schwartz: What are the phenotypes? Dr. Martinez: No, who are the phenotypers? My experience in CHARGE, we have had several individuals, experts in their own fields, come into the study with really no expectations. The ones who stand out are the molecular biologists who, when we discuss autism, they say there is no way that you will be able to detect a signature for autism using transcriptional analysis. In fact, they were right. When we started to look at the transcrip- tional profiles of 25, 50 kids, comparing them to DD and to general population, there was no real clear trend, there was just noise. Because we have Sallie Rogers and Sallie Ozonoff, and we have the very, very thorough psychometric analysis, the ADI, the ADOS, and other instruments, we could then subphenotype them into early onset, regression, spectrum versus full-blown autism. That really made a huge difference for the molecular biologists. So I think in that respect, we could do better, but it clearly shows that if you have very well defined stratification, then it improves the 

PROCEEDINGS 231 molecular biology and the interpretation of the results. Dr. Herbert: I would like to expand David Schwartz’s question. There are the psychologists, the mental health people, there are the environmental people and there are the medical people, the people dealing with the medical illnesses of the children. The question of how those are going to be characterized also needs to be addressed. It is one of the areas that has been least well developed in the field. I think that is important at the level of physical examination, and it is also important at the level of distinguishing between biomarkers of exposure and biomarkers of effect. I think it is important with biological sampling and with all of the questions that we are asking here, what is the impact on the individual patient, because that may have impact on interpreting the levels of exposure relating to the thresholds that may be impacted by the redox status or a variety of other things. So I think that the middle needs to be addressed, and not just the behavioral and the environmental. Dr. Schendel: I’d like to tag onto those two points. My first thought that I had before we got into this last issue of what is missing in the phenotyping fortified what Craig Newschaffer was saying and others. What we can do with these different types of study designs for either ongoing or existing studies, is augment the data collection for particular environmental characteristics and add to the data collection. For instance, what we could do in the SEED study, and what is done in CHARGE, is to go back and get residential histories into data linkages with environmental datasets, identified at a very crude level, perhaps, but to some extent some sort of household exposures or residential neighbor- hood exposures, preconceptually, so being creative and getting additional environmental data that might be augmented for existing or ongoing studies. You can’t do all of this data collection on any one study, which feeds into my second point. When you are doing the phenotyping, feasibly there is a limit to what the individuals may be able to tolerate, both the cases and the comparison group that you are bringing forward. So I think you do need to keep that in mind. Some studies may focus specifically on a phenotypic assessment to get as much detail as possible, while others may do sufficient data collection, for instance what CHARGE and CADDRE are doing, to do an adequate phenotyping, but perhaps without the mirror imaging and so on that are also beneficial. 

232 AUTISM AND THE ENVIRONMENT So going back to the idea of coordination across studies, I think it would be important to come up with a core protocol, a core phenotype protocol, which then could be used to connect information across different studies in more creative analyses, and taking advantage of the wealth of data that is being collected across studies. Dr. Insel: Some of that is actually in place. There is an attempt to have a core assessment battery. That is part of what NDARS tried to lay out, and to make it very easy by having all of the tools in one place available on a laptop, so you can take them and run with them. I think David’s question still needs to be drilled into a little bit, and maybe we can have a more thorough discussion of that. Before we get there, I am still stuck on that first line up here about biological sampling. If I understood what Eric Fombonne was saying, in those rare cases where you have an environmental exposure that appears to be causative, as we sometimes say about certain genetic mutations that are so predominant that the environment doesn’t count, and there are certain environmental exposures where the genetics don’t count. So if we had a couple of those, just as we do in genetics, you would want to use those as your beachhead, to start off with. Indeed, if it is the case, which I hadn’t heard before in the discussion, that the ones we have, valproic acid, thalidomide, whatever else, which you could say would be hard-core discoveries, happen very, very early, it seems to me that ought to change the nature of the discussion about biological sampling. If we had had this meeting 2 years ago, we would probably be talking about what samples should be collected in the second year of life, or maybe toward the end of the first year of life. Most of our discussions have been about prenatal sampling, meaning second, third trimester. But if it really is first trimester, then we need a very different kind of discussion about sampling, as you are suggesting. Am I correctly understanding that proposal? Dr. Fombonne: This model is for the fourth or fifth week of exposure. All the data that we know about valproic acid and thalidomide suggest exposure during a very narrow time window, which is around the 24th, 25th days of gestation. It is also consistent with neuropathology findings, like Margaret Bowman and Thomas Kemper have also related that back to the first trimester of the pregnancy, most certainly. So there is converging evidence that the first trimester is probably a critical time point. 

PROCEEDINGS 233 Dr. Pessah: I’d like to speak to this as a toxicologist, because those models certainly are interesting models of autism, but they don’t really portray the full spectrum of autism. One could imagine that an early hit could devastate the developing system in such a way that one of the endpoints is going to be autism, but there are also significantly other teratogenic effects that these models produce. We know that we can have autism without teratogenesis. So although they serve as a valuable model, I would like to point out, for example, to the valproate model, where it has been linked to HOX-1, and now there is some linkage to the HOX-1 gene, but it is a downregula- tion of the HOX-1 product, whereas with valproate, I know there are some data out there now where it increases in response to valproate. So you have to be very careful with limiting the spectrum to a very early point in development. It may be relevant, but it doesn’t fully explain the full spectrum. Dr. Susser: This is something that I have done a lot of work in. I agree with Eric Fombonne, in the sense that we should get samples as early as possible in pregnancy. That doesn’t mean that only very early is important, but we have suggestive evidence that very early is one of the important time periods. So we should be aiming for that. There are indirect ways to get that evidence prior to the National Child Study, where it is actually built into it, but that will be quite a long time before that actually happens. One indirect way is to use the first prenatal visit as an indirect proxy for what is going on in early preg- nancy. It is not perfect, but it is something that we can do. We have, for example, information at 17 weeks, what the mother was taking at 5 weeks, and you have information from blood samples from the mothers about things that may not change that much between 5 and 18 weeks. It is indirect, but it is still much better than anything else we have. The second thing we can do is something that Dr. Fombonne also alluded to, which is similar to what we did with the famine study, which is seek populations who have been exposed early in pregnancy to some of the toxicants that we are concerned about. There are such studies, but I think Dr. Schwartz’s question is relevant here, because those kind of populations, I don’t believe have yet been used to study autism and those kind of outcomes. So that is another way that one could solve the problem. I would agree with you that the focus shouldn’t be exclusively early 

234 AUTISM AND THE ENVIRONMENT pregnancy, but that should be an important focus. Dr. Beaudet: I think that a biological sample should be collected as early as possible from the father, if you are going to collect biological samples. We know things about advanced paternal age, and there is plenty of potential for other paternally originated abnormalities. Dr. Martinez: Should we go to the next point, because endopheno- type seems to be very important? Dr. Pessah: Yes. One of the major points that was brought up is endophenotypes and better stratifying autism, if anybody would like to comment. The idea would be that you have to have communication between the epidemiological design and how you are going to substratify the cases based on very detailed diagnoses and maybe medical exams as well. One of the things that I think is a gap is, we haven’t paid a huge amount of attention to comorbidities, including what seizure disorder might tell us about substratification, or cardiovascular problems. Some of the genes are expressed both in terms of the cardiovascular system. If you want to go out there, nobody ever collects data on family history of different kinds of cancer. Given that PTEN and Met and some of the other genes we have heard about are very, very highly involved in human carcinogenesis, is there any link there? Ms. Bono: We talked yesterday, one of the ideas was to data mine some of the practices. I know that every time I have taken my son in, they have asked, heart disease in the family, cancer in the family? They ask all sorts of autoimmune disorders in the family. So again from yesterday, I think we need a formal mechanism where we can send investigators into these practices and data mine what is there from the kids. That leads into the recovery study as well, what can we learn from the recovery period. Dr. Martinez: But in that case it would be extremely useful to have some kind of instrument. Ms. Bono: Right, a formal mechanism. Dr. Martinez: That has to be not too complicated, either, because many of these practices are extremely busy practices. I am telling you my experience in asthma. We have done this in our work in longitudinal studies, and the best way is for those who are experts in the field—this is something that David Schwartz was asking for—to define a minimal set of ascertainments that need to be done in a very large number of potential cases. I don’t get the feeling, I may be completely wrong, that 

PROCEEDINGS 235 such a thing exists. It would be very good for some consensus to be created as to the development of such kinds of tools. Ms. Bono: I agree with that, and to not count on the practitioners to be able to do it. They are just too busy. Dr. Martinez: Right, but you could develop even some automated systems that can be accessed through the Web, in which you just fill out one form. That is what we have in asthma. It is an idea that could be developed for autism, too. I know it is a complicated disease, but believe me, asthma is also complicated. Dr. Swedo: Is Paul Law here this time? There you are, Paul. Do you want to take the mike and respond to that for IAN? Participant: The IAN project is an online environment where families are invited to come and provide information that they can provide accurately based on their own experiences with autism in general, everything from the school system to their diagnostic workup, the day they were diagnosed, and so on and so forth. One of the more valuable parts of what we are collecting is the treatments that parents say that they are on. It seems like it is pretty accurate—we try to do a lot of things to assess the quality of the data that are coming in, but I think the treatments that families have their kids on, they know this pretty well and are able to comment on it pretty well. It is very interesting. There is clearly a different practice. There are different camps in terms of the way families are being managed. I think that capturing data from the clinical environment as well would be a really good thing. It is going to give you higher quality data than what we are able to capture on IAN. The nice thing about IAN is, you create a website, the next day it is available to everybody, and you have some very dedicated data collectors coming in by the four thousands to provide information. A lot of noise issues can be dealt with, with this enormous sample size. Literally we have had over 3,000 people join in 2 weeks. Dr. Spence: Can you describe what the general questionnaires are? Are you doing a medical history? You talked about the treatments, but in terms of thinking about environment, are you doing an exposures questionnaire, that kind of thing? Participant: Yes, I think there are opportunities there as well, but you always need to be cognizant of the limitations. The content in IAN right now is wafer thin, because we wanted to make sure we didn’t 

236 AUTISM AND THE ENVIRONMENT overburden families. It was mainly designed around making sure that we had enough information to help researchers be able to identify subjects. We wanted to at least be able to do that very well, create a very efficient matching system where researchers can identify subjects in the areas. We can e-mail them, and we can try to illuminate this thing that is slowing down autism research. It seems like every researcher I talk to has problems getting subjects. If there is somebody who has solved that, let me know. But there is a huge amount of potential. I get 10 e-mails a day from mothers wanting a pregnancy history to be added that is extensive, or immunization histories to be added. In general, families are very interested in providing more data. I think for certain types of studies this could be at least good preliminary data to build off of. It is certainly not going to compete with nice, sound epidemiological population-based studies, but it is one of these things that we can do really quickly. Like the terbutaline thing, we can have that done in a couple of months with a sample size of 4,000. We can really move things along pretty quickly for certain types of questions. Dr. Martinez: I think you are referring to family-based methods. That’s fine, I think it is very useful. The caregiver base would also be important and perhaps could have a different type of role. Since I am involved in that field, I have to say, for example, that in the field of cystic fibrosis, there are very interesting developments. Some of the things that we have found out about cystic fibrosis were based on a large epidemiologic study which is caregiver based. For example, the importance of diabetes in the prognosis of cystic fibrosis was ascertained through these kind of methods. So the development of tools for standardized acquisition of information from caregivers would be a very interesting development. Dr. Law: I’ll make one comment about that, and I’ll get down. One of my degrees is in medical informatics. If you look at private practices or community-based practices, less than 10 percent or so have any kind of electronic data capture happening in their practices. So if we are going to try to capture all these clinical data, we are going to have to act very strategically and try to solve the problem together. It probably needs to be a Web-based system, and we need to think about electronic health records, creative ways of addressing that issue and capturing it. Dr. Leshner: Can I just make a suggestion? I am a little bit nervous that we are spending way too much time relative to what we have 

PROCEEDINGS 237 designing specific studies. So if we could identify the content gaps, I think that would be very helpful. Dr. Pessah: Just content gaps. Anybody else want to comment? Dr. Beaudet: Just a brief point. I think it is really important for those of us who are studying autism at different levels, in my case genomics, to have the appropriate controls available. So far, although I can use autism-unaffected sibs from families with autism, I don’t consider them to be the best controls. There is no national repository I can go to to get children in that age group. I can get adults, but I want to compare against my children controls where there is no autism in the family. Dr. Needham: One gap that we have talked about somewhat as far as biological samples go is what kind of chemicals to try to measure. I was just wondering, if oxidative stress is a potential pathway, is there an in vitro or some kind of test that we can use for various chemicals to test if those chemicals have the capacity to cause oxidative stress, and use those perhaps as biomarkers of exposure? Dr. Noble: Sure. That is the kind of work that I talked about yesterday. You can run as many chemicals as we can grow cells. We have sensitivities, mercury, four parts per billion, lead, low ranges, ethanol, 17 millimolar, low end of the clinical range, everything we look at. If you look at progenitor cells taken from the right time in the developing nervous system, you have these extraordinary sensitivities and the capacity for high throughput in multiple endpoints. Dr. Needham: Are you saying that any chemical can cause oxidative stress? Dr. Noble: No, I’m saying that any chemical that causes oxidative stress will have a set of well-defined endpoints that can be analyzed. Dr. Needham: So can you test a large array of chemicals, say 1,000 chemicals, and say that 500 of those cause oxidative stress and 500 do not? Dr. Noble: Yes, absolutely. We do that all the time for various purposes. Dr. Needham: I’d like to see that. Dr. Pessah: I just want to add one gap here that is relevant before I go on. That is the gap which deals with the cost of extracting quality information from medical records. We need to put more money into that. These studies regardless of design really rely on the data that you extract, 

238 AUTISM AND THE ENVIRONMENT and there is no easy way to do it. It is all done by hand at this point. So we are going to go on. Exposure assessment. There was a huge, huge issue about what we measure, what is our sensitivity, and how do we integrate large datasets on exposure assessment. This would involve exposure experts, analytical chemists, engineers who are developing new technologies, and of course through the new NIH genome environment initiative. We need to get new technologies in place, and we need to do it as soon as possible. One of the things that I think we need to identify gaps is in terms of prioritizing which chemicals and which physical agents that are accurate or predictive of the environment in which the study population is living or has lived in. This would be different, I would imagine, for agricultural communities such as those in the Central Valley of California versus urban communities, let’s say in New York City. We want to be able to optimize environmental samples. What does that mean? How do we stabilize across studies so that these environ- mental samples have the same information that we haven’t changed in the process of either collection or storage of the samples? There are no real guidelines at this point. I think we need to set up guidelines. I think there are some guidelines maybe through the CDC, but are those the guidelines that everyone uses for the studies that have been mentioned that are in progress or just beginning. Implementations and development of biosensors. I think this was a perfect lead-in for this discussion, because I think the technology has progressed to the point where very soon we will be able to make thousands of measurements simultaneously. Could that impact the study design, the fact that we would be able to measure everything, or everything we think at that point in time is important, rather than having to make choices? So I want to open that up. Dr. Noble: One thing that I don’t see on there, and I want to put this in perspective. I live in multiple medical communities. We work on spinal cord injury, we work on cancer, we work with developmental disabilities. In all of these communities that I work in, I have to say that I have never heard patient stories like the excreter phenotype story, like the recovery stories that I am learning from talking to patients here. If I hear those stories once or twice, I have the standard reaction of a scientist of being an anecdote. But I am hearing too many similarities. This issue with the excreter phenotype, I have to say I find it both fascinating and worrisome in this trial design. If this idea is right, that at 

PROCEEDINGS 239 least some subset of these kids are accumulating body burdens because they cannot excrete, they will never be picked up with that design. So I think there needs to be a part of this where somehow we are getting improved access to the ongoing patient experiments that are going to go on, no matter what we decide in this room. They are going to go on, they have to go on, because that is what the parents need to do. Somehow we have to be willing to solve this scientific question of, are there kids who when they are treated with chelation therapy or antioxi- dants or whatever is being done, are they pouring out heavy metals? As a scientist, I would really like to know that. Dr. Insel: Why couldn’t you do that in a challenge approach? Dr. Noble: Right, and that would be incorporated here. That is what chelation is, it is a challenge approach, right? Ms. Bono: I think what you are saying, though, is the difficulty of measuring when you are pulling these samples. They are not done provocatively. Trying to get the chelation study off the ground, as we are all hoping to, but pulling the data samples without a provocative agent may not produce any information. Dr. Noble: That is what I am worried about, from what I am hearing. Dr. Swedo: I think you speak to a bigger issue, Mark Noble, and I hope we can let Isaac Pessah finish this. I see this as being one area, that is, are there population differences for children who end up with autism versus those who don’t end up with autism, and the increased evaluation one can do to those groups. But I still am not satisfied that the popula- tion-based approach is the only way to go. I again go back to the example of leukemia and diabetes and other things, where you start with the patients, and you start looking for differences within those kids, so you would use a clinic-based approach as well. Dr. Martinez: That is why at the very beginning we said that there are several different approaches. Many of these are longitudinal, the ones we are talking about now, but the case-control approach has its advantages. Dr. Swedo: I don’t even know that it would necessarily have to be a case-control. Dr. Martinez: For example, it could be that cases are the ones that are improving and controls are the ones that are not improving. These are mainly epidemiological. We can add to this that is much more a case control type than the type we have talked about. 

240 AUTISM AND THE ENVIRONMENT Ms. Redwood: I just had a real quick comment. One of the things that we have been focusing on is the heterogeneity of autism. My concern when I hear about these chelation studies is that we are not doing biomarkers ahead of time to identify the children who actually have a heavy metal burden. There is a test available, it is a urinary porphyrin test. I would say that we first try to delineate the heterogeneity of autism and then provide targeted treatments. If you take every child with autism and enroll them in a chelation study without first identifying the children that have body burdens of metals, then the ones who do respond, it is not going to be significant because the other ones that didn’t have a body burden of metals to begin with. Dr. Martinez: I think we should not discuss, as Dr. Leshner has asked, the details of different studies. We have a general agreement that some type of therapeutic approach needs to be tested, and I think that is what we should say. Why don’t you please continue, because otherwise we won’t finish. Dr. Hertz-Picciotto: Can we go back to that last? I did want to say something about exposure assessment. Dr. Martinez: Yes, go ahead. Dr. Hertz-Picciotto: There are a couple of things that I wanted to just mention. First of all, in terms of the chemicals and the prioritization, there is a whole class of compounds, and it has been alluded to in a couple of slides that no one has started looking at, at all, as far as I know. We haven’t started, and I don’t know if CDC is planning to, in the cosmetics. Now, cosmetics are primarily used by women, and they don’t usually just stop at the point when they become pregnant. So that is a whole area that I think does need to be addressed in this field. The other aspect, in terms of how we have been prioritizing, I think most people have been thinking in terms of direct neurotoxins. I’m not sure who it was yesterday who gave a list of the ways we might think about direct toxins as well as factors that indirectly affect the develop- ment of the central nervous system. In particular the hypothesis around inflammation has come up in the literature in a number of different contexts. The work from our center seems to suggest immunologic dysregulation going on. So I think we need to think in terms of toxins to the immune system as being suspect agents that ought to be on the list. 

PROCEEDINGS 241 Then in addition to the issue of biological sampling, I just wanted to raise the issue of general environmental sampling. Diana Schendel referred to existing databases that are out there for exposure assessment in general, but we also might be thinking in terms of this whole issue of sensors in the home. Indoor air is a lot more contaminated than outdoor air. Most of what EPA does is outdoor air monitoring. So a lot of the chemicals that are generated by our textiles and other home products that we use for cleaning agents and so forth have a great deal of toxicity and ought to be on the list. Dr. Pessah: We should probably move on. Dr. Hertz-Picciotto: But anyway, household sampling, not just biological. Dr. Insel: Same slide, last line, where you say analytical methods, I would include some way of using the genome as a sensor, whether it is epigenetic changes or DNA damage. Dr. Pessah: Yes, that comes right here. Mr. Blaxill: Do you guys have a guide to the framework, so that we can be trying to absorb it while we are talking about it? Dr. Pessah: I’m sorry? Mr. Blaxill: The framework is being progressively revealed, and it is hard to—in your head without knowing what the framework is. Dr. Pessah: I should have had it all planned out ahead of time in terms of a diagram. Dr. Hertz-Picciotto: We should have let him finish presenting before we started picking it apart. Dr. Martinez: Everybody wants to say something. Dr. Schendel: Just real quick about the exposure assessment. One gap is the understanding of the exposure response in pregnancy specifically of the woman who is exposed to a particular compound or chemical, how the body is processing that exposure. Larry Needham had in his slides how it is being shunted to different compartments in the body. Those are typically done on nonpregnant subjects, I would imagine. I think the pregnant state would be an important gap to fill. Dr. Martinez: Why don’t you finish the whole thing fast? Dr. Pessah: Okay. Clearly to try to understand genetic differences between some groups of autistic children and the general population, we can now do this much more efficiently with the technology that has matured. Five hundred thousand single-nucleotide polymorphism chips are now available. If you want to look at particular sections of chromo- 

242 AUTISM AND THE ENVIRONMENT somes like long-arm chromosome 10 or short-arm chromosome 7, you can do so with a nimble gene array, which allows you to get very high resolution information about low-copy repeats and so forth. Epigenetic information is a bit more arduous and time-consuming, but one can see if there are major differences in methylation in any one of these designs. Metabolomic, proteomic, and lipomics are the -omics profiling. There has been this amazing advance in how many analytes we can measure at once. I know David in Bruce Hammock’s lab can give you every single intermediate in the methionine pathway all in one shot. The question is, are the samples stored properly so you are not just looking at variation, which goes back to sample quality. Toxicologists and cell biologists, we have real difficulty collecting brain biopsies for cellular studies, mechanistic studies. It is just never going to happen in this disorder. However, I want to point out that many of the candidate genes that have been reported to date are expressed in the immune system. So we need to focus a bit more on the immune system as a biomarker or a target in autism. Here I have given you some examples, where one could immune profile from primary cells. I think that cell lines, Epstein-Barr trans- formed B lymphocytes are very useful, but they have their limitations. So primary cells give us more specific information, and will probably be more predictive, especially when you are trying to subphenotype. Intracellular cytokines can now be measured, so you can measure cell activation parameters for intact cells. You can do antigen recall and see if there are problems with antigen recall, and you can also challenge with environmental agents, endotoxins being one, but you can also, once you have identified candidate xenobiotics that are of high probability, you can then apply them to a cell-based system to see how the signaling response of immune cells differs. And of course, one can look for autoantibodies to brain proteins, which sometimes we forget. If the mother has autoantibodies in circulation during gestation, the IgGs do cross the placental barrier and can have an effect. It is a genetic problem which is transformed environmentally within the developing fetus. One could also use immune cells, primary immune cells, to do a detailed analysis of signaling cascades. Here, we don’t have to identify which kids have the mutations ahead of time, because these signaling cascades converge on common denominators. 

PROCEEDINGS 243 I have just named a few of them here, and we can talk about them. Met goes through Grb2, Gsb1; it impacts pI3 kinase, which then can influence all of the targets of pI3 kinase that influence cell growth and proliferation. Cav1.2, it is part of a large complex that regulates local calcium signals in cells, which is incredibly important for just about every cell process we can think of. It is also regulated by pI3 kinase. So we can look for commonalities in cell signaling that is aberrant in autism cases without knowing the mutations ahead of time. Animal models. I think there has been some negative press here about, you can’t do a single-gene mutation in a mouse because it won’t be representative of the condition. But I would like to bring up for discussion for transgene X knock-in and knock-out mice are in fact useful, especially if you can time when the mutation occurs, and look at specific endpoints in signaling abnormalities that may be relevant to what you discovered in the cell models that you did get from autistic kids out at the periphery. Here you can look at the immune system and the nervous system. Finally, one can humanize the mice in a very different way. If there are roles of autoantibodies for autism susceptibility risk, one could humanize the mice by exposing them to the human IgGs and see if they develop neurodevelopmental defects. Other models include the PKU model, which is not an early model. Poly-IC, I think we have talked about the early models, poly-IC models that Paul Patterson has initiated work on, and oxidative stress models. We can test those in animals thoroughly. Data validation, integration, and modeling are absolutely essential if you are going to put an integrated effort together. Finally, and this leads to the next round of talks this afternoon, is the randomized controlled trials of novel therapeutics and innovative intervention. Chelation is one that has been discussed quite a bit here, but also what about antioxidants, vitamin supplementation, what about DHA, some of the biological markers that are oxidative stress involve lipid metabolism. How can we bring these into innovative clinical trials? Then finally, combinatorial therapies. Dr. Martinez: Opinions? Dr. Noble: One of the things that I did not hear, and I understand its complexity, is the understanding of toxicant synergies. Having recently felt that our mechanistic analysis allowed us to explore this, I was horrified by how little knowledge we have in the literature about this, yet 

244 AUTISM AND THE ENVIRONMENT here we are very interested in multiple exposure paradigms. So I think that is a biological question that I think is addressable, and I think it is critical to understand. From what we are learning, it is also going to teach us a lot about the underlying mechanisms of susceptibility. From what Gil said in regard to sample collection, one of the notes that I took was the critical aspects of the timing of sample collection, and whether the individual has had a meal beforehand. If you are going to do metabolic profiling, you have got to take into account that metabolic profiles change, and circadian rhythms. They change according to what we need to change according to when we ate. So I think that has to be built into the trial design. One of the things that we have not discussed yet that I would love to hear information on, if not in this setting, then eventually, in the pediatric neurology community, where there are so many kids that need some kind of treatment, one of the points that I have been making repeatedly in the meetings that we have is, tell me what I am supposed to repair. I am a stem cell biologist. In spinal cord injury, I know what I am supposed to repair. I can design experiments. I know what I am going for. So I am a total newcomer to this area. Before I came to this meeting I read widely. One of the papers that many of you know about is this mirror neuron study from Rama Chandra’s lab, which is so intriguing because of the role we think mirror neurons have in enabling recognition of the other as self in a study. That needs to be repeated, it needs to be extended, but it raises the possibility that kids with autism don’t have mirror neurons. Well, if I know that as a guy who is trying to do repair, it helps me to think about it. I need to know a lot more like that. It is part of why larger studies are so important. The last note I have here I have mentioned a couple of times, and I just want to come back and emphasize. What I am hearing from the patients in this setting is unlike any other setting I have ever been in. This is one of—in more than 25 years of doing clinically related neurological research, this is the first meeting where I am coming away from it saying that the patient populations are doing experiments that we need to pay attention to. They are telling us a lot that concerns me about trial designs. I have been talking to a number of the parents and saying you do the chelation therapy, and without doing secretin and it doesn’t work so well. I think 

PROCEEDINGS 245 we have to draw that information out. Dr. Schwartz: I just had one more item to add on the list, which is data sharing and general access to data, both the genetic and the phenotyping data. I think it is something that we should consider. Dr. Martinez: That is a crucial point in all areas of research, but I have the feeling that particularly in this area it is crucial. Dr. Leshner: Does this community have any implied at least policies around data sharing and database sharing? Dr. Martinez: That used to be a difficult question. Dr. Insel: It depends what you mean by the community. Dr. Leshner: Or the database. In the ACEs, that is, the autism centers of excellence, it is a requirement that all the data goes into NDAR. But as you are hearing, there are many different communities which are not at this point integrated together. As David Schwartz suggests, that is an opportunity. Dr. Martinez: And it is also a challenge. This seems to be a need. I will give again the example of cystic fibrosis. There is a central foundation that directs the availability of data, and it has been extraordi- narily useful, extraordinarily useful for us in the CF community. In any event, just as an example. One of the things that could come out of here is precisely what David Schwartz has just asked for, which is mechanisms of data sharing. It would be crucial. Dr. Beaudet: The Autism Treatment Network, their entire vision was to emulate the CF center concept, both from research and care and so on. Although I don’t know what its latest merger implications are. Dr. Colamarino: I was just going to say that, at least from the perspective of what Cure Autism Now (now merging with Autism Speaks) has been doing over the last 10 years, by building a database with the genetic information, it is all open access; it is all forced sharing. The problem is integrating it on a larger level, which is supposed to be attempted by NDAR, with respect to the medical records. It was based on CF. And there is the Autism Treatment Network, but they are just getting off the ground, which is partially the reason they are not there yet. The idea was to create standardized medical assessments. All that information would be put into a communal database that could be mined by whomever. But they are just getting off the ground, which is partially the reason they are not. Dr. Spence: I think the opportunity for sharing is great, because 

246 AUTISM AND THE ENVIRONMENT people have been recreating the wheel a lot. This is something that there is a great medical history that the ATN is using based on the AGRE history that we have developed for NIMH. But then there is also CHARGE, and the ACE centers are going to be using the CHARGE medical history. Dr. Colamarino: We use that one. Dr. Spence: So there is some overlap. I think that is something we can work on as a community. But I think providing those resources to other groups, for instance, that we can then—what I really wanted to say was, would it be helpful for us to sit for 5 minutes and have everybody mention one study that is ongoing that we know of that is collecting data, and then get in the room representatives from those studies so that we can talk about this sharing issue. Dr. Martinez: I don’t know. Dr. Leshner, I think one thing that could be done, I don’t know if it can be organized, but some kind of a questionnaire or some document would need to come from everybody who has those type of studies, that could be given to the organizing committee. I don’t think we have time today to go through that. Mr. Blaxill: I think you were asking us to think about gaps, so I have two categories of gaps. But first I want to endorse what Tom Insel said, which is this notion of the integration. The communities are not well integrated, and I think that is a problem. I think actually the environment is one of the reasons—the relative lack of effort on the environmental side is one of the reasons there is lack of integration in this. It would be helpful if this discussion leads to more integration. But there ought to be some progress on that front. In terms of gaps, I would list two categories, and they are very different. On the one hand, I am thinking more about yesterday. There were a bunch of discussions about mechanisms, all of which raised questions about biomarkers. Dr. Insel, you asked the question about biomarkers. Biomarkers presume a model of what you are measuring and why you are concerned about it. I think there is a critical gap. There is a real question about what is going on, what brings autism together, what makes it a coherent disease entity, if indeed it is one. I think there is a bunch of problems in that. There are hypotheses in different levels of scale, at the level of the topology of the network of the brain to tissue and pervasive tissue growth, or at the molecular level with signaling and redox balance and that sort of thing. I think that is a gap. As a consumer of the science, I would love to 

PROCEEDINGS 247 have a clear idea of what my child has, and why she is like Ms. Bernard’s child and why she is different, some kind of mechanism that clarifies some of the complexity and gets to final common pathways in a meaningful way. I think success in that kind of work would—we would end up renaming autism. We would have another name for it that would be much more specific as to what the mechanism is. That would be a satisfying outcome and some distance from that. The integration of some of the different systems, the interdisciplinary teaming, multidisciplinary teaming, I think is critical to make progress there. I think you all are doing more of that, which I think is great. The other gap that is a different category is, I hear the enthusiasm among the professional epidemiologists for prospective studies. I know that from a methodologic standpoint you can control it, you can do it right, you can do it the way you want to. I think there is a gap in terms of high-quality retrospective studies. Recognizing that there is imperfect information, natural experiments; I’d love to look at the Inuit population. You could learn a lot from that. You could also triage some things about genetic versus environment. Carefully done, that could be helpful. The Amish community that comes up. They all have a lifestyle. Maybe we all need to follow the Eskimo lifestyle. The Amish have a lifestyle. Is that a risk factor or a protective factor, or are there other things going on genetically? Are there protective Amish genes? I don’t know. There are reports that the Amish have low rates of autism. That would be interest- ing. Going back to the original cases first discovered by Dr. Leo Kanner, Martha Herbert mentioned that all of the 11 Kanner cases had somatic symptoms. A lot of them lived in the same place. Was that a signal or just an accident of the diagnostic pathway? I think there is something to learn from that. I think we need to face up to the question of trend. That raises retrospective problems, but I think it is critically important strategically. I obviously have a strong position on that, I have tried to articulate that. I think we need to put that behind us. My personal bias is, if you have a 10-fold increase and there is a hypothesis that it is artifactual, the burden of proof is on the person making the hypothesis. We ought to test those theories and get them behind us, because we have been delayed for a long time by the confusion, and maybe a reassuring sense that everything is fine and there is not a problem. I think it is a strategic 

248 AUTISM AND THE ENVIRONMENT question when you talk about the environment. That can only be done retrospectively. There are going to be imperfections when you do that. It is not as clean as a prospective study, but there are lots of natural experiments there and information that you can draw on. Dr. Herbert: One more retrospective study that could be done is documenting retrospectively claims of parents that their child has recovered. There is a registry online, and the campaign is starting to get more people to register their kids. In the interest of disclosure, I have an IRB to do this, but if we could go back and find kids who did have reliable diagnosis and follow them up and reliably assess them, that would have very important implications for the neurobiological underpinnings of our hypotheses about what it is that we need to measure. Dr. Martinez: As a trained epidemiologist, I have to say that this goes back to what we were saying before. In order to be able to know what is characteristic of the children who get better, you need to know what is characteristic of those who don’t get better. So you need good information from a large set of patients. Once again, the availability of tools that will allow you to assess that, so that you have it at one point, and then 10 years later you can assess retrospectively, based on how kids got along the way, who had a certain characteristic at the beginning and who didn’t, would be crucial. Dr. Noble: I agree with that, but it is important to recognize in the context of neurological disorders, how unusual it is to have these kinds of improvements. You don’t see this with spinal cord injury. You don’t see this with cerebral palsy. As a scientist I would like to have what you have, but also as a scientist if I had what Martha Herbert is trying to collect, I would be in there trying to get interesting data out of it. Dr. Swedo: That speaks to the point that I was going to make. I think there is a huge gap in our definitions. Definitions of what is autism, we heard that in Mark Noble’s comments, but also earlier when we were trying to talk about these epidemiologic studies. We continue to change the definition of autism spectrum disorders. We need to get back to more continuous variables, because our discrete categories are not working very well. The second issue is definition of recovered. I was speaking with some of the practitioners and folks in the audience yesterday; recovered 

PROCEEDINGS 249 to me is very different than recovered to some other folks. I would want to know that we have a standard definition of response and nonresponse. Dr. Martinez: This word is coming back very often: standard. The other one that comes is collaboration between groups, which is the other side of the coin of standard. Ms. Bernard: I think there is a gap in tracking individuals with autism as they age. Some of these people become more severe and sometimes they get less severe, even if they don’t fully recover, and there could be an environmental component to that. Do they have ongoing exposures? Do some of the differences in degree of severity that we see or onset of comorbid conditions, does it have to do with something in their environment? If we could get rid of later exposures, maybe it didn’t trigger the autism, but it changes the course of disease. I think we need to look at that. In addition in terms of gaps, for the epidemiology I think it is very important that we get a handle on subtypes, and that we don’t just report rates of autism spectrum disorder as 1 in 150, or 1 in 100 for New Jersey, but that we actually determine whether that is Asperger’s or PDD-NOS. Some of our current studies are not able to answer that with their methodology. There was one other. I can’t remember what it is. I’ll think of it. Dr. Schwartz: I just wanted to get back to a point that Tom Insel raised before which is related to acquisition that could be used 5 years from now, 3, 5 years from now. It relates to the new measures that may become available. The question is, should we think critically about developing a biobank for samples across these different epidemiologic studies that could be centrally located, available to investigators, very similar to what NIMH has set up with the genetic resources. Dr. Newschaffer: Another issue in the class of biomarkers, and I don’t think this is a repeat, biomarkers of exposure susceptibility. We have talked about efforts in genetic markers of exposure susceptibility, but I wonder if there are gaps in the area, and the toxicologists can help me, in terms of phenotypic markers, other biomarkers of exposure susceptibility. As an epidemiologist, if I am going to study exposure effects, I would really like to know who is susceptible and who is not susceptible to those exposures. I get a sense that there are some gaps there. Dr. Susser: Just a very quick point. I am just afraid that we might 

250 AUTISM AND THE ENVIRONMENT miss the forest for the trees. I agree with all of the trees that we have talked about, but I feel that we do need to tackle the question of the time trend and what has happened, whether it has been stable or whether it has been increasing over the period in question. It is not impossible to do. All the things we are talking about are difficult to do, but that is one that I think we ought to take on in terms of gaps. We haven’t done it yet. Dr. Martinez: Any other thoughts? Ms. Bono: I agree with that about the time trend data. The CDC said this morning that it was going to be difficult, but we are dealing with something that is very important. We have to find out what changed in the environment and when, and try to figure it out. As I made notes throughout the last couple of days, I have identified several gaps. I’m not quite sure, Isaac, if it is in your things that you have mentioned—methylation and looking more into that. That is a fruitful area, so we need to make sure that those studies are ongoing. We didn’t have up on the chart, although I know a lot of people have talked about it, the toxokinetics of mercury, the transport mechanisms. We need a better chelator for the blood–brain barrier. So I would love to see something like that. We need to understand the transport mechanisms in the body and how it gets into the brain, how it can possibly come back out. With methylation comes detox support. I saw you had an antioxidant protocol, so perhaps that is methylation/detox in some of that. We identified this morning through the exposures we cannot identify, or we haven’t yet, the nonvaccinated kids. I think that is an important gap that we aren’t testing. When we see the large population studies, so much of the population is vaccinated. So we need to try and find those kids. Mark Blaxill mentioned the Amish, but there are also several medical practices across the country. There are education records, the waivers. We can go in and find those kids and be able to see what is going on there. Lastly, gastrointestinal studies. So much is going on in the guts of these kids. There are failing enzyme systems, there is bacterial dysbiosis, there are gut pathogens and virus, fungus. Certain phenotypes have this, and can remarkably get better if these things are addressed. So I would like to see a gut study. Dr. Wilcox: I haven’t heard any discussion about what happens to these kids as adults. It seems to me that the natural course of the disease into adulthood ought to be information that would help us understand the 

PROCEEDINGS 251 etiology as well as progression. Mr. Blaxill: The Leo Kanner natural experiment would be—they are 70 years old, the first 11 cases. Some of them are still alive. So that is one way to use that kind of information. I think, Sallie Bernard, you raised that point as well. Dr. Martinez: We are almost at the time, right? I knew you were going to tell me that. Dr. Leshner: I have decided that during the discussion session later, we are going to go around this table, everybody is getting one minute, literally one minute to identify a residual gap, not design the experiment, not design the instrument, content, what is the gap. You may pass, you may not pass, you can do it however you want, but you get one minute. I am really nervous that we are losing something, that we are not capturing something. So that will be the first exercise. Dr. Martinez: I have the opposite feeling, that we have too many things on the table, but that’s okay. Dr. Leshner: For me you can’t have too many things on the table. I think this is sufficiently understudied. Anyway, this was a good session. You can have one more minute. Dr. Martinez: Let me try to summarize as an outsider the areas that I feel you guys and ladies have talked about. I think there are issues that can be measured in individuals with the disease. I think that is the first area. I think we all have said that needs to be measured from the time of conception—of course, we cannot measure them in the individual him- or herself, but at least in the mother. Eric Fombonne has stressed many times the need for this to be from the time of conception, because things can happen in the very first month that are crucial. These things that can be measured are phenotypic or clinical, genotypic, genomic, and metabolomic and all of the other things that we have talked about, and could be predictors of disease and consequence of disease. So those two areas need to be distinguished, I think. The second is what can we measure in the environment? Everybody has said a lot about that. Since everybody is going to say something in one minute, it would be good that you would try to tell us which of these areas that I am trying to identify are interesting, or even other areas. I think that these two measures—because in the end, science is measuring, these two areas which we can measure, I think we will need to concentrate ourselves in trying to understand. 

252 AUTISM AND THE ENVIRONMENT The third area that has been stressed, which is potential models, are models relevant, which models? These models could be in silico models. Everybody has said the importance of bioinformatics, but also it would be important to know animal models, are they relevant, which animal models, which are you interested in? Mark was talking about mechanisms. Unfortunately in humans it is tough to study mechanisms, but that is what animal models are good for. So perhaps that could be something that could be added to the discussion. So those three I think are the main areas that people need to talk about. Then this can be transformed into specific studies. But I don’t think it is necessary for us to discuss here the details of each specific study. Dr. Leshner: That was really well summarized. Thank you. 

Session VIII Public–Private Partnerships Dr. Leshner: This afternoon we have two parts to our discussions. The purpose of the first session of the afternoon is to talk about public–private partnerships and to try and see whether there are things we can do that we cannot do independently, to try to see if there are unique opportunities here. In a minute I am going to ask each of the people listed on the agenda to make 5 minutes of comments. They can do that from their seats, but the light will be blinking nonetheless. Then we will have some discussion of that topic, but then I would like to reserve the last 45 minutes or so for a residual discussion of the gap issues and things like that, to make sure that we have captured everybody’s major thoughts. We don’t want your minor thoughts, but your major thoughts. Let me just say that I think this kind of an event that we have had yesterday and today is important for a variety of reasons, but it is an example of a lesson many of us have been learning over and over, that is, of the importance of engaging the public in the deliberations of the scientific community and the utility of it. It is not just about who pays—that old adage, he who pays eats—but it is about taking advantage of the experience and expertise that people who live with these issues bring to the table. I believe, and actually editorialized in Science magazine and had no effect about this issue, that it is important to provide the public with an opportunity to help shape the research agenda in a positive sense of shaping. An obvious question is how do you do that? There are a variety of mechanisms that people have used. NIH has long had a distinguished history of bringing the public into research agenda-shaping kinds of exercises, but this is an example of one, I hope. By having the major funders in the room and the people who get to articulate the research agenda, I think we have a special opportunity, and I am very pleased that we have been able to have as much interaction as we have had. The topic of public–private partnerships tracks with that, and is a part of it. So with that little bit of context, Sallie Bernard, you’re on. 253 

254 AUTISM AND THE ENVIRONMENT MS. SALLIE BERNARD SafeMinds I made the observation last night when we were leaving for dinner that the last time I was down at the IOM building, I was actually outside protesting. Obviously this meeting is very different, and I think we can attribute that to this whole process being a true example of a public– private partnership, from the planning of the meeting to the meeting itself, and hopefully after the meeting with your leadership, and Dr. Leshner in guiding the meeting, and the tone set by Dr. Raub, we really have a public–private partnership right before us in this meeting. I think this is an example of what happens when the advocacy community, scientists, and government come together. That is how we are going to get our environmental agenda in effect, by continuing the path that we have started down. When you look at public–private partnerships, I see them happening in two ways, at least as it comes to promoting an environmental agenda in autism. One is bringing together people with diverse experiences, expertise, and perspectives in critical thinking, and you need the public and the private sides in order to do that. So it is really a human capital idea. The second component is a funding component, but you also need to bring people together in forums like this, but it carries across to other activities. For example, if there is a large study that is being designed, bring in the advocate community to help you design and implement and interpret the results of that project. Things like being involved in the autism strategic plan which we just spoke about earlier today, bringing advocates into that role, making sure that they have a prominent position on the Interagency Autism Coordi- nating Committee. This is an example of public–private partnership. I think that the organization I represent here, which is SafeMinds, we have certainly tried to historically have a role in those types of activities, and bringing the environmental side of the gene–environment equation to the forefront of autism research. The second area in public–private is what projects are you actually funding in the research arena? Where SafeMinds has played a role in that is to take a new high-risk hypothesis, and certainly one that is not popular, and people would prefer not to study it, which is the role of 

PROCEEDINGS 255 mercury and vaccines and thimerosal in autism, and broader than that, the environmental side, bringing it to the forefront in autism research. To get that off the ground, since the year 2000 we have funded studies on thimerosal and mercury in mouse models, primates, looking at what goes in the brain and other tissues, behavior in primates, cell culture studies, a baby hair study that we did. These are all very small-scale studies, but they help to set the stage, provide a platform from which larger things can come. Now we are at the stage where we would like to see these types of ideas go to the next level, and for that we need the public side of the public–private partnership. If you are looking at the data and the data say that there is a gene– environment interaction that is involved in autism, and you look at what has happened with autism research, all of the money except for maybe a little tiny proportion has gone to the gene side. So we need to rebalance that. In fact, we need an overcorrection, because we need more money on the environmental side now to compensate for the fact that for the last 15 years, as Laura Bono said at the beginning yesterday, nothing has been done on the environmental side. So we would like to see a lot more of that happen right away. The last thing I would like to point out is, when we walk out of this room we are going to get a lot of push-back in implementing the ideas that we are all going to go around and have today. We need a mechanism to help us move forward, to make sure that the ideas that we come up with today get implemented. I hope that turns into another public–private partnership and we put that mechanism in place. DR. HENRY FALK Coordinating Center for Environmental Health and Injury Prevention, Centers for Disease Control and Prevention Thank you very much. This has been a really good meeting, in the sense of having an opportunity to listen and exchange information. I want to thank Bill Raub and the organizers for inviting me. Much of my comments will be within the context of public–private partnership, because I don’t think there is any potential for success without that. At CDC, in the last several years the terrorism part of the agency’s budget has gone way up, and everybody worries that the bulk of our other public health activities are under a lot of stress. We have taken to 

256 AUTISM AND THE ENVIRONMENT using the terms that there are “urgent threats” that we have to deal with and but also “urgent realities.” These are Dr. Gerberding’s words. I think “urgent realities” resonates with me in this setting. Autism certainly fits. So, what can we do in this situation that would be helpful and positive, maintaining an open mind and attitude and what can CDC scientists do to help, particularly in the environmental area? As you heard from Larry Needham and others, there are potential ways in which we can assist. One of the things that I heard over the last day and a half is there are real limits on the amount of work that has been done related to the environmental aspects of autism. I think environment has been underrep- resented; perhaps also aspects of epidemiology are underrepresented. I heard a number of issues that are not yet fully addressed on the epidemiology, but certainly more on the environment. When we look at different ways to address the environment, to me there is always the issue of how do you find the right clues. We don’t at the moment have the hook into what are the key environmental issues. I always see this as two very distinct ways of going about hunting for clues. One is, hit it with everything you’ve got. When we talk about things like the National Children’s Study, we use multiple endpoints and multiple risk factors, we get everything we can into that study to be as efficient as possible, and as broad as possible, and we hope that as David Schwartz says, you look at as much as you can, and you hope that you catch all the right clues. So that is one approach. Larry Needham in the lab has been involved in working with the National Children’s Study. We would like to participate in ways that we can contribute to that approach. I would also like to emphasize the second approach. Early in my career I spent a lot of time with Dr. Robert Miller, who headed the childhood cancer epidemiology program at NIH. He would do things like catch me at a meeting in Japan and say, just look at all those people. The Americans have stripes in their ties going this way and the Japanese have stripes going in the other direction. Somehow you have to really look carefully at the problem to find the clues. This second aspect requires looking intensively at the cases themselves and, for example, using case-control studies. You have to look for anything uneven in the data, and for geographic discrepancies. You need to look at more severe and at less severe cases and see what can you learn from them. You look at isolated cases that occur someplace 

PROCEEDINGS 257 else and see what you can learn from those. If there are groups in the population that are not being vaccinated, study them. Whatever the issue may be, search for unusual features and see whether you can learn something interesting that then provides you a broader clue. One is a macro approach and the other is a micro approach. We need to do both. I also think that there are many issues here that people have expressed or feel strongly about that are potentially amenable to analysis, and we should try to do as many of those as possible. Whether that relates to chelation or other ideas, if there is some way, for example, of actually analyzing issues like excretion, equilibrium, and metal distribution in children with autism, we should look into how we can best do that. How to go forward. I think there are lots of good ideas that have been put on the table in the course of the last day and a half. My sense is, we should try to capture the moment so that we not only have these good ideas, but that we figure out a good mechanism, as Sallie Bernard said, to follow up on these good ideas, for example, through a smaller workgroup of people here, to say how many of these good ideas can we put into practice and how can we do that? But I think certainly on the environmental focus and on some of the epidemiologic issues, there is potential for good collaboration between CDC and NIH within the government, as well as with private groups and others on the outside. Somebody asked Phil Landrigan yesterday how the lead program made so much progress. There were some key moments in there, one of which was how to utilize data to get government agencies to move forward, for example, how to get HUD (Department of Housing and Urban Development) engaged to remove lead from housing. So, we should think about how to seize a moment like this and go forward in terms of follow-up to this meeting. We hope to contribute to doing that. DR. GARY GOLDSTEIN Kennedy Krieger Institute Thank you. I am a founding and volunteer board member of Autism Speaks, which is now about 2 years old. I thought I would take my time to tell you a little bit about what we are doing. 

258 AUTISM AND THE ENVIRONMENT The object here was to raise awareness of autism in the general community, not only so we pick up cases early, but so the general public would support the research and advocacy efforts of this community. I think it is off to a good start. In the past 2 years, three other organizations funding research in autism, AGRE, NAAR, and CAM, those people in this community would know those, have all come under—have merged in, which is pretty unusual in my experience, of nonprofits merging, and have come in under what is now Autism Speaks. This foundation has the objective to do as well in fundraising as the March of Dimes or Muscular Dystrophy Association or whatever, and have enough money to augment and influence and create progress in understanding and treating autism, preventing autism. I will give you some examples of the infrastructure things that have historically and more recently have been funded or are going to be funded. They are things like the Baby Sibs Consortium, which is a group of about six to eight different centers now looking at the subsequent children born to a family that has one child with autism. There are lots of interesting observations that can be made. But the idea of bringing this group together and sharing data. Second is something called the Autism Tissue Project. This is one to encourage—when a child with autism dies, to encourage a donation of brain tissue that can be banked and used for histologic study. We greatly lack careful and detailed studies of the anatomy of autism. A third is the AGRE Project, which is the multiplex family project of collecting genetic material and making it available to investigators widely, with the idea that again, the data are shared. Another one is the IAN Project that you heard about before (http://www.ianproject.org). I would encourage you to look at this. It is a community site and it is a research site of online research. One big goal of that is to match and provide potential subjects for studies that someone is appropriate for and would like to participate in. The Autism Treatment Network is under discussion right now. This is benchmarked to the successes of the Cystic Fibrosis Network. We are at a much earlier stage. We would also like to have the ability that every child in this country who has autism can reach a medical clinic by car. We in Baltimore at the Kennedy Krieger Institute are continually getting referrals, people who are flying in from different cities for what seems to be very simple evaluations. They should be able to get them in their own 

PROCEEDINGS 259 community, but they can’t because there is no one there to provide them, or the waiting list is so enormous. So one of our goals is to have treatment centers, medical centers, in children’s hospitals hopefully that can evaluate children within a car drive, an hour car drive, and that these people share standards of evaluation, standards of care and their observations are collected and centralized in a data management system. These are some of the infrastructural things we are funding that I think will be available for research, rather than specific projects. We do have projects. We have pilot grants, fellowship grants, and most recently augmentation grants. CDC is funding a project, and it looks like something more can be done with an augmentation grant, that we would be open to that, or the NIH has a funded NIH grant, or a grant that is a near miss. We are finding now the grants that have outstanding scores, priority scores of 140, not getting funded. I think autism is not in a separate pile. We would be open to helping that investigator keeping his or her investigation going via a new R01 or a continuation grant that just didn’t get funded until they have time to go through the 9-month, year process to get that grant reconsidered. So these are some of the things we are doing. Then we are looking at gaps. I have tried to do a portfolio analysis. I was able to collect the amount of funding from each of the institutes directed to autism projects. These were projects that had autism in the title. I could see where the money was going roughly. The gaps that we talked about. One is, if there is a role for immu- nologic mechanisms in either the treatments or the prevention or the cause of autism. I only could find one grant. I found 0.6 of the NIH budget going to the immunology of autism. NIH doesn’t plan what it spends, this is just in response to what is submitted. That is what was submitted and that is what was funded. Lots of money going to gene searches, defining autism in children. Very little to drug studies, but then there are no targets, so maybe that is why that is the case, but very little, and very little going to the whole toxicant things that we have been talking about, very, very little. So as we looked at the gaps, these are gaps we see. Are we going to help fill them? Are we going to partner with different agencies to help fill those gaps? Are they appropriate gaps to work on right now? That is what Autism Speaks is about. 

260 AUTISM AND THE ENVIRONMENT DR. TOM INSEL National Institute of Mental Health My interest in being here was generated through the evaluation we did for the IACC autism research matrix. Just to back up for a minute, there are five NIH Institutes that contribute to autism research. For the most part, all served on the IACC that was enacted under the 2002 Children’s Health Act and which included 12 federal agencies as well as public members. At Congress’s request, in 2003 we developed an autism research matrix which was a set of short-term and long-term goals, both high and low risk, over a 10-year period. In 2006 we reviewed the matrix, including fairly broad input in our analysis. Most recently, the matrix was out for public comment in a phase that ended January 16. In putting those comments together, our expert reviewers, the people on the IACC itself, and the public all gave us very much the same message: You have done some things very well, such as building capacity in certain areas, particularly genetics and some tissue banks, and you have built up the field in a number of ways. However, you have not paid enough attention to environmental factors. So going forward, I think it is fair to say that is going to be an area of increased interest at NIH. I’m sure Dr. David Schwartz, who is the expert in this area, will fill in more of the details. In terms of the public–private partnership, I think it is important to realize that autism is different from many other areas of biomedical research in that we do have a very significant private investment here. Autism Speaks is aiming to have about $60 million a year to invest in research. The Simons Foundation has already put in about $30 million and just announced a new RFA last week for another $15 million investment this year. This is in addition to the $108 million NIH is spending. It is really more than just an exchange of ideas; there is a real opportunity here to put in money that will synergize efforts and perhaps serve as a catalyst for certain things we want to accomplish. The question is—and this goes back to what Sallie Bernard started with—how do you divide this up? At the intellectual level, there is an opportunity here for a much better partnership. You heard a lot about this from Laura Bono at the very beginning of the meeting, such as the expertise that families can bring to this discussion and to this partnership. Then the question becomes, how do you facilitate that, how do we make 

PROCEEDINGS 261 sure that we are involving families as partners as well as these other private groups? A second question arising from Sallie Bernard’s division of this into intellectual and infrastructural or resource-driven money, is how we can make sure that we are working together and not at cross-purposes. Is there a way that in many of these areas we can divide and make major progress, and say Autism Speaks will take on this challenge, CDC takes on a different challenge, NIMH addresses another part, and by following the expertise, get the best science done? I bring this up because there is a lot of energy in this room and a lot of enthusiasm for opportunities on environmental factors, and this has been great to hear. That is what I wanted to get out of this; where are the opportunities, where do we want to move, how are we going to make this happen? We haven’t talked much about the challenge of peer review and what happens when many of these ideas, some of which we talked about here, go to peer review. Innovative and cutting-edge science, particularly ideas for which there are limited pilot data, often face significant challenges through the peer-review process. So we have this tension between the sense of urgency that many people in this room feel and the culture of science, which ensures the rigor we expect, but can also make it very difficult for us to move quickly and sometimes makes it more difficult to move in new ways. So here is my suggestion. As we think about public–private partnerships, we think not only about opportunities and who will do what, but also in terms of the kind of science we want to undertake. As Dr. Gary Goldstein was mentioning, there may be an opportunity here for particular areas, which might struggle in peer review, to be picked up, at least for the pilot phase, by Autism Speaks or the Simons Foundation or by some other group before they maybe get taken to a larger scale and get funded by NIH. I don’t know any other way to do some of the most innovative projects, especially at the earliest phases. So I think that is the public health challenge, balancing urgency and rigor. We just have to get our heads together to figure out how to make that happen. 

262 AUTISM AND THE ENVIRONMENT MS. LYN REDWOOD National Autism Association First I want to thank Dr. Raub with the deepest gratitude for listening to the parents and taking our concerns to heart and making this important meeting a reality. I want to thank Dr. Schwartz and Dr. Insel and the staff and the panel participants for taking away time to commit themselves to this important research. I hope this will be the first of many meetings in an effort to get to the bottom of what is making our kids sick and what we can do to help them now. This is a mock newspaper article. If tomorrow we woke up and 1 in every 150 children were reported missing by their families, what would happen? It would be a national crisis. Communities and federal agencies around the country would band together to form search parties in an all- out effort to find the lost children. I am asking today that our federal agencies move into crisis mode and rethink the way they conduct research by reaching out to the persons who are the most knowledgeable about this disorder—parents. Parents are confronted on a daily basis with the needs of their child. They are the ones who are most knowledgeable about their illness. They also have the highest level of commitment to finding help, because anything less is just not acceptable. Parents are able to add perspective, passion, and urgency to this discussion, assuring that the human dimension of the disease is incorporated into scientific considerations, program policy, the investment strategy, and research focus. With PubMed available online, parents can now access research that in the past was only available at medical school libraries. This has tremendously increased their knowledge base of the disease and has leveled the playing field between parents, physicians, and researchers. An article appearing in this month’s issue of the Archives of Pediatrics and Adolescent Medicine documents the important role of parents in autism research. Parents have organized research funding, they have constructed clinical research networks, they have popularized empirical-based treatments, they have suggested new avenues for research, and they have anticipated shifts in the understanding of autism. The article concluded that the existence of partnerships with parents is a critical component of future research and treatment programs. I would like to outline very briefly four different mechanisms that are available now to be able to partner. These include autism advisory 

PROCEEDINGS 263 boards, integration panels, the establishment of shared research invent- tories, and community-based and -driven research initiatives. In the Combating Autism Act, Congressman Barton called for the creation of an autism advisory board to provide public feedback and interaction with the NIH Interagency Autism Coordinating Committee. He went on to state that public participation is necessary to emphasize the human side of autism research, and to ensure that federal resources are used widely. An idea to embrace consumer participation was also recommended in 1993 by the Institute of Medicine in a report solicited by the Department of Defense in the development of a congressionally mandated research program for breast cancer. In this program, consumer participation occurs at both levels of peer review, scientific and programmatic. Consumers read proposals, they present their opinions, and they have full voting privileges on the committee. A recent published review of the program found consumer participation a very positive and successful aspect of the program, and suggested that it be used as a model for those who desire to work in partnerships on critical health issues. Parents are not waiting for the double-blind, placebo-control studies to determine if a simple intervention such as changing a child’s diet, administering supplements in an effort to support the child’s nutritional status, or reduced oxidative stress is beneficial. Parents are leading the science beyond educational or behavioral interventions because they recognize that their children have medical problems which present with behavioral manifestations. Addressing these medical issues often results in improved overall function for the child. The Autism Research Institute reports that over a thousand parents have completed questionnaires regarding their child’s recovery from autism. In addition, several large clinical practices that utilize a biomedical approach to treating autism have offered to open up their practices for data mining to identify historical, physical, and clinical information which could provide clues to autism’s etiology, heterogene- ity, and effective treatments. In summary, we are faced with an urgent public health crisis that demands immediate attention and action from our federal agencies and the allocation of resources necessary to respond rapidly and effectively. This change in paradigm dictates a shift in the focus of autism research away from an exclusively genetic model to one that investigates the role of environmental factors, an abandonment of the traditional bed-to-bench 

264 AUTISM AND THE ENVIRONMENT approach to research that is predominantly investigator driven, to one that embraces a sense of urgency and direction infused by parents. Parents as stakeholders need to have their status elevated and voices heard. We need mechanisms where input and ideas are actively solicited and formally embraced by our federal agencies, one where stakeholders are able to guide research in a direction which offers the most promise in an effort to find meaningful ways to help improve the lives of those suffering with autism now and to prevent its occurrence in the future. We need to devise ways to partner, share ideas and resources, and work together in the areas of policy, science, and research. Thank you. Dr. Leshner: The last speaker on the panel is David Schwartz. DR. DAVID SCHWARTZ National Institute of Environmental Health Sciences My view is pretty simple. Autism is a complex public health problem that we are not going to solve from any one perspective. Therefore, it seems to me that partnerships, including scientific partnerships, public– private partnerships, partnerships with families, and partnerships with patients are essential to being able to understand this complex problem. In fact, I think over the past couple of days, what is clear to me is that without understanding the genetics component of autism, we are not going to be able to make headway in the environmental component, and without understanding the environmental component of autism, it is going to be harder to make sense of the underlying genetics. That is probably why 15 of the 23 pairs of chromosomes have loci that have been linked to autism. There must be lots of different causes of autism, lots of different phenotypes, and I think the environment and other factors can help us understand the causes of these different types of this disease. What are the fundamental bases or ground rules of a partnership? The ground rules of a partnership are that everyone wins from the partnership, that we all make each other better by being part of this partnership. It seems to me that autism and this problem that we are dealing with is just the kind of mission-oriented problem that would benefit from input from several different vantage points and individual parties or groups working together to solve this complex problem. 

PROCEEDINGS 265 Plus, I think it would be a lot of fun. Over the past day and a half, I have enjoyed time thinking about this problem with everyone around the table. It has been enjoyable to me to consider the ideas that people brought forward in the context of scientific questions. Now, I don’t want to reiterate the scientific gaps, because I think we are going to get to that later. What I want to do is bring up three infrastructure issues that I think are important to consider in a partner- ship. One is, I think this idea of a biobank that cuts across studies might be a helpful way of creating transparency, creating access to data, and creating access to samples both today as well as 5 years from now that could be helpful in solving the problem. A second infrastructure need that seems to me to be apparent is, when I think back to the early 1980s when I was taking care of patients with AIDS, we were all running around giving different types of therapies in uncontrolled ways. It seems like therapy for autism is at risk of a similar type of poorly coordinated, poorly studied problem. A clinical trials network would ultimately benefit children with autism, if patients, families, investigators, and clinicians embraced this approach. I believe we could make the same progress that we’ve made with AIDS if we develop and support a clinical trials network for autism. AIDS patients were leaders in the AIDS clinical trials network and I think the same could be done with families with autistic children. The third infrastructure need that I think is important to consider is training in this area. While I think that experienced, established individuals could contribute to this area, I think we need to think about the next generation. So I would want to think more about the possibility of developing environmental sciences programs that could inform the behavioral sciences, cross-training individuals so that they could develop new knowledge in this important area of research. DISCUSSION Dr. Leshner: Let me open the discussion to people at the table and see what your reactions and your thoughts are. Tom Insel answered my question that I had for Dr. Goldstein, which was going to be about how much money are we talking about from a variety of organizations. It would be really interesting and perhaps as a next step coming out of this 

266 AUTISM AND THE ENVIRONMENT workshop to have serious sitdowns among the various funders about how to do a better job coordinating and parsing out who does what, because we are talking serious money. They are not trivial amounts of money coming out of the private sector, whereas in some other domains there really is trivial money coming out. So there is something worth coordinating. Dr. Insel: There is a conspicuous absence though in this discussion, which is pharma. In almost any other area of medicine, if we were talking about such an effort we would be talking about pharma, which represents double the investment of the NIH every year in R&D (research and development). So one might ask whether that is also something that could also be brought into the discussion. Dr. Leshner: Can I just ask out of ignorance, do pharmaceutical companies do research on autism? Dr. Insel: No. According to them, there are no targets. They don’t think the market is big enough. Participant: And it is children. Dr. Leshner: Comments? Dr. Spence: I just want to say, pharmaceutical companies are running a few trials, a little. I don’t think they spend the kind of money you guys are talking about, but I don’t want it to go on record as saying they are not doing anything. Existing psychiatric drugs are being tested in autistic individuals. Dr. Leshner: Since people aren’t jumping to say something about this public–private thing, I would like to challenge the group, that what we have seen yesterday and today is actually a very good example of a public–private partnership, and the beginning of a process of conversa- tion and a model for perhaps how to have these kinds of interactions. I really think we need to give Bill Raub a great deal of credit for having pushed this as hard as he did, which was hard, but well as always. But I would say that since we have the making of a public–private partnership, if we don’t have one going forward, shame on us. I don’t actually know exactly what I am talking about in terms of how to make sure that it has momentum and continues, but at a minimum we have the base for it, and maybe a next step to assure that it partly continues is to do two things. One is to try to get the funders together, because that is always a good way to get things moving. The second is to have some discussions among the organized groups of things like what the scientific criteria 

PROCEEDINGS 267 would be for acceptable kinds of studies. I think that is something where there is some either disagreement or confusion among the various groups, about what actually is a fundable study and what is a doable study and what will have the credibility or not the credibility. Again, that needs to be a two-way discussion that goes on between the people who want an array of new kinds of research done. It may have been Tom Insel who raised the issue of peer review. I can tell you as a journal publisher that 96 percent of the studies that come to Science magazine get turned down, but an awful lot of this kind of research would have a terrible time getting through normal peer review processes. I have to ask, where do you move the line, where do you risk moving the line? I can promise you, from my own experience in the drug addiction business, if you lower the bar, the bar will fall. So we will have to be very, very careful, because it is a very fragile bar, much more fragile than people may think. So I just offer that as unsolicited advice. Having said that, let’s open up for the around-the-table gap business, what you didn’t say at lunch. Alison will start. Ms. Singer: I’m going to do two in 30 seconds each. I want to talk a little bit more about the time trends data that Mark Blaxill brought up before, and say that this continues to be a dark cloud that distracts us from the real issues that we have to solve. To Tom Insel’s point earlier about what will we be sorry about when we look back 5 years, we need to do the retrospective studies, and we also need to start to gather the data now in the ADDM studies, looking at the subtypes, what percentage of the kids are Asperger’s, what percent- age are autism, and what percentage of the kids are PDD-NOS, so that we can use that to inform the science and also use that for planning purposes so that we can plan for treatment for all of the kids. The treatment protocols across the life span for a person with lower function- ing life-span autism will be dramatically different from the treatment protocols for someone with Asperger’s. I also want to say that we need to focus on studies that yield actionable information for parents. All across America there are thousands of minitrials going on every day in our homes, where we are experimenting on our kids based on anecdote, and we deserve better. We deserve better than anecdote, we deserve evidence. I would urge when we are looking at the list of 80,000 chemicals, that we focus on those that are still in our environment, that are still being used, where the data that we glean from the studies will result in 

268 AUTISM AND THE ENVIRONMENT actionable information. In my last 10 seconds I will say by way of an example, I don’t think that many parents or pregnant women are still prescribed thalidomide, but they are still prescribed terbutaline. So I would say, let’s try to prioritize in a way that helps parents to have information that can help to prevent more children from being diagnosed with autism. Dr. Colamarino: I don’t think this will come as any surprise. I would say biomarkers, biomarkers of exposure, biomarkers that are predictive, biomarkers of outcome, especially treatment outcome, and biomarkers of change across treatment lifetime. I would qualify that as saying physiologic biomarkers, measurable things. Dr. Needham: And biomarkers, but in addition to biomarkers, and this has been touched upon also, is a need for biobanks. But importantly in that is the correct procurement and storage of the specimens. There would be nothing worse than taking samples, and 5 years from now the samples are neither taken nor stored correctly. So we need to talk to each other about how to take and store those samples. Also, we need to look at other stressors, certainly environmental chemicals come to mind, but we also need to look at such stressors as infectious agents and also the effects of the psychosocial. There are many stressors that may play a role in autism. Ms. Bernard: I agree with everything that has been said, so I won’t spend time on those. I would also like to add the recovery study, which we could do with videotapes of when they were autistic to verify that they really were, and then when they are not recovered and have a blinded—we could do a study like that and we could do it right away. I think we also should do a study of cases of unusual onset, because they give us clues. We talked about first trimester or whatever, but there are also cases in the literature more than those of people getting autism when they are teenagers, when they are in their late childhood, from viruses or other infectious agents and from mercury exposure. We should look at those, and maybe it gives us information on windows of vulnerability. We should also look at adults, not just with biomarkers as you said, Sophia, but also phenotypes and how that changes, and what do teens and adults get that the children don’t have. I would also have us consider when we do studies of low-dose and paradoxical effects, because that seems to be something that affects studies when we are looking at dose response, to consider those. 

PROCEEDINGS 269 Then something that is related to research, push maybe a little heavier on the precautionary principle, and look at what are the top substances out there that could be causing harm and raise the alarm that we really need to do something about all these things in our world that have a high possibility of making our kids sick. Dr. Goldstein: I’d just like to say that we are learning from the IAN Project that more than 80, probably 85, percent of parents have never before participated in a research trial. If you contrast that with what goes on in the successes of childhood cancer, where over 95 percent of the children are in research trials, we are almost the reverse. And these are people interested. So I think there is an enormous difference between what is going on in terms of testing anything, whether it is diagnostic tools or whatever. Then I would add that I think this would be a good time to begin to explore the role of the immune system as it relates to autism. Ms. Redwood: I think it is really important to go into the database of the children, as Sallie Bernard said, that have recovered, but also the ones that are currently undergoing treatment. I think there is a wealth of information there with regard to biomarkers to be able to better identify the heterogeneity of this disease. I also want to put a plug in for—and I missed this in one of my slides—having public and private agencies to develop some type of method of coordination. As we have heard, the private sources are matching the public sources now, and may very well exceed the public’s resources in terms of autism research. So we need to coordinate those efforts so there is not duplication in the areas that Gary has mentioned that have been underfunded in the past. Historically, environmental research comes to the forefront. So again, I am echoing things that have been said previously, but I think they are very important. Dr. Schwartz: I feel like I have made too many suggestions, but I’ll make one another, which is the data access, coordination, and sharing policy among autism investigators, as well as a vehicle to make that happen. Dr. Schendel: Since one of the driving forces behind this meeting is our hypothesis that environmental factors contribute to the changing prevalence of autism, one of the things that we lack is an understanding of the features of individuals who were born in the 1980s. So we might want to consider a descriptive study of individuals in 

270 AUTISM AND THE ENVIRONMENT their late teens and early 20s, and if it was population based you might even get a better handle on prevalent characteristics. Dr. Newschaffer: I’m pretty confident that most of my personal high-priority areas have been addressed or will be addressed. I want to go to a couple of areas. First, in relation to the topic I was asked to speak upon, which is the international studies, especially in developing countries, I think we need to work on culturally robust, easy-to-implement screeners. I think these may have to rely more heavily on child observation and language and interview, because the combination of language and culture make it very difficult to do screeners that are effective in developing countries. That is one area. Second, I want to drill down a little bit in the area of retrospective studies for time trend. Maybe we need to involve folks from other disciplines, like medical sociology. It is very hard to figure out how to evaluate empirically diagnostic tendency retrospectively. You can’t look at what is written down, you can’t look at records, you can’t ask people what they think they were doing 20 years ago. When you examine the older kids, which is a nice idea, the phenotype has changed, and if you assess them now, you are assessing them through the current lens, not the lens of what evaluation was 20 years ago. So perhaps we need different disciplines to think about this, to help us come up with valid ways to do this kind of study. Dr. Hertz-Picciotto: We were talking about that at dinner last night. It is definitely an issue that needs to be addressed, including the possibility of people having been institutionalized and therefore having undergone heavy medications along the way. So it is a difficult question, but we do need to figure out if this is a feasible thing, to go back and see the autistic adults, if they are out there. The other issue that hasn’t been brought up and hasn’t really been talked about, I think there might be something that can be learned from looking at the other things that have been increasing in this same period of time, if there is a true increase, and there may well be. I am thinking of asthma as one, which we heard a bit about, and also the obesity epidemic. I think those may point us in the directions of immunologic and metabolic things, that perhaps there is some common- ality here. Obviously very big differences among those things, but something that we may be able to learn from that. 

PROCEEDINGS 271 Ms. Bono: I think I said some of mine earlier, so I am just going to run down them very quickly, and I’ll stay within a minute. I agree with Alison Singer and several others that time trends are very important, what changed in the environment and when. The phenotype is very important, biomarkers that Sophia Colamarino mentioned, the clinical data mining Lyn Redwood mentioned, recovery studies that Sallie Bernard has talked about. We discussed this morning, and I am going to bring up again, we need a vaccinated and unvaccinated study. It is very important to the community to be able to move forward. We have to look at that, we have to move beyond that. We need a mercury toxokinetic study, especially because so many kids are getting better. They do have heavy metal problems, and we need a chelator that will cross the blood–brain barrier. Currently, and this is maybe where pharma can get involved, we have chelators that were created in the 1940s. I know we can do better. It is 70 years later. I’m sure we can do better. We need methylation detox types of studies. We need anti- inflammatory immune studies, which I think Gary Goldstein brought up. We need gastrointestinal studies, including the enzyme systems and bacterial gut viruses. I love that Dr. Insel brought up peer review. I think that is a big area we need to discuss, especially about scoring proposals, based on the gaps that we are seeing here; put the priority on the studies that we need. And bottom line, which I know we all are thinking about now, that is helping the most kids as quickly as possible. If we keep that in mind, then I think we are going in the right direction. Dr. James: I’d like to address the issue of biomarkers and the biorepository, which we have all alluded to. There are several epidemi- ologic studies that are now going on collecting biological samples, but I think there are important practical considerations in the details that are absolutely essential for sample collection. I think we need to establish standardized protocols for sample collection. For example, for our glutathione redox ratio, I can’t use your samples. It has to be fasting, it has to be at the same time of day, dumb little details that are going to make all the difference in being able to have meaningful data that we can compare between studies. Another big area that is a problem for me is age-matched control samples. If we want to find out what is unique about autism with these 

272 AUTISM AND THE ENVIRONMENT biomarkers, we have got to have something to compare it to, and they are very hard to come by. I don’t know how to approach that, whether we can have a control repository at the same time. The other practical problem with the repository, especially metabolic biomarkers, is that they degrade over time. They have to be stored properly in the same way. So I would just make a plea for standardized protocol for collection. It will take a small working group to establish those criteria. Dr. Noble: Three things that haven’t been said. On the immune system, I think Gary Goldstein is right. I think one can make the prediction that these individuals are in TH-2 imbalance. They will have low glutathione levels. It is easily studied in T-cell populations. They will have memory cell defects. There are a lot of data from the AIDS literature on the effects of glutathione decreases on those areas. It might be useful surrogates, if not functional. Isaac Pessah said we can’t get access to the nervous system. I have been thinking about that. What we can get access to is neural crest stem cells. They are very easily grown from skin biopsies. That has been very well worked out. That may be something for us to look at. Lastly, I made the statement earlier that this is unlike any other neurological disorder that I have been interested in in terms of these recovery stories. That was a misstatement. This sounds like a biochemi- cal disorder. That is where we get those kinds of turnarounds. When we get the right drugs into an individual who has a biochemical disorder, we have these effects. So I am starting to now go in that direction in my thinking for awhile. Dr. Goldstein: In terms of neural tissue, we do nasal biopsies in Rett syndrome, and can get growing neurons from nasal biopsies. Dr. Noble: So you and I should talk about what would be the best to look at. Dr. Beaudet: I guess I am the genetic representative. There seems to be a certain sense that maybe genetics is somehow competing for resources with the environment considerations. I would just make a couple of points about the genetic situation. There are dramatic advances going on at this moment. They are not at all the outcome of the money that was invested. They do with discovery of de novo defects, and all this big investment has been in inherited kinds of abnormalities. I think this group that have de novo defects are by and large mentally 

PROCEEDINGS 273 retarded and dysmorphic, and they can be identified to a great extent and moved apart from some of the other studies, because I think they will make any other study more complicated and contaminated if it is not done. Another thing which I haven’t pushed particularly is epigenetic sampling, which is very difficult because of the tissue specificity. I have a sense there is probably a group of patients where the etiology is mostly genetic, and another group where it is mostly environmental with some genetic susceptibility component. I think it would be helpful to separate these. Then an area where I don’t really have any expertise, I still feel like we need to know whether the incidence has changed between 1970 and 2000. Even if it is stable now, we need to know if it changed. Mr. Blaxill: I want to talk a little bit about the burden of proof on time trends. I would make the suggestion that given the increases that we have seen, the notion that the reported increases are an artifact is a hypothesis, and it is a testable hypothesis. I’ll just take California as an example, because there is a pretty good surveillance system there, better than other parts of the country. A child born in California in the early 1980s had less than a 5 in 10,000 chance of becoming autistic. By the late 1990s, that rate was closer to 40 for 10,000, so that is roughly a 10-fold increase in about 15 years. The notion of that increase being artifactual has been tested in a lot of natural experiments. There is a hypothesis of diagnostic substitution that has been tested and falsified. There is the hypothesis of diagnostic expansion, that somehow we are changing the quality of the diagnoses. The interesting thing about California is that the registry is for a full syndrome, it doesn’t include the broader spectrum, so that theory of expansion doesn’t hold. The M.I.N.D. Institute has done a quality control check across decades. There are problems with those kinds of studies, but they didn’t uncover any different diagnostic quality in birth cohorts from the 1980s or the 1990s. And the surveillance system has been in place. It has changed, any administrative system changes, but it has been in place since the 1970s, unlike some of the educational records in the 1990s. So the notion of diagnostic expansion is not supported. The only remaining hypothesis, or what I like to call the hidden horde hypothesis, that somehow hundreds if not thousands of children escaped the service systems, and if we looked for 25-year-old Califor- nian young men and women with autism, we would find them in large 

274 AUTISM AND THE ENVIRONMENT numbers somewhere. That is an interesting hypothesis. I would suggest people ought to prove it before we start accepting the notion that the increases are artifactual. So all you can say from that is that there is a lot of evidence that suggests that in California the increases are real. California, when you compare other databases to the rest of the country, they don’t look that different, so again that is inductive reasoning, but you could argue that is a pretty useful database for the United States. Then I would ask the question in terms of studies, I think we should pursue studies to clarify uncertainties, but I would urge us to consider changing the burden of proof. Rather than saying the burden of proof is to demonstrate that all this is real, I would say the burden of proof is to demonstrate that it is artifactual. If that is the case, we ought to think about changing our official narrative, because the expression of doubt about the increases creates the sense that we have a mystery and a puzzle, and no sense of urgency. The recognition of the reality changes the entire dynamic. I think a lot of us are saying we need to treat autism as an emergency, and that is what all the data points to. Dr. Susser: I agree that we should really tackle the time trends problem, and that we should trace the course of autism over an individ- ual’s life span, not just over the first few years. So I second those two points that have already been made. The only point that I would like to add is that I would like to extend what Art Beaudet said, but in a different way. I think one thing that we have learned from all these genetic studies is that there is a sizeable proportion of kids with autism that do have de novo genetic defects. That is really important, and that is new knowledge. I think what that should lead us to do is to look for the causes of those de novo defects. They are probably going to be environmental. I think we should focus there when we are thinking about environmental studies of autism. Dr. Landrigan: On the first day of science class, we are generally taught to beware of the study with the n of 1, and for good reason. But to summarize from yesterday, I think we heard several instances where the n of 1 study, to the extent that it is a hypothesis-generating process, can be extraordinarily valuable, especially in the face of a disease of this complexity. But also, it would be a way of ensuring that the engagement of the 

PROCEEDINGS 275 parents and the treating clinicians and the scientific community come together. The problem is that peer review will be deaf on most of those kinds of studies seen one at a time. But if we could create an infrastruc- ture, direct the peer review to the mental level in terms of something that fosters those opportunities, and then make it easier somehow, a lower energy barrier, whereby opportunistic examinations of that sort, testing a variety of the hypotheses we had here, I think it could be seminal in terms of shaping the more deliberate and longer term research protocol. Dr. Herbert: That is really great, thanks. I want to mention three quick points. I think there needs to be support at the infrastructure level. Sallie Bernard talked about makeup funding for environment. I think there needs to be makeup funding for parents and integrative practitioners to come together and figure out what it is that they need to convey about the disease phenomenology and the methods of measurement that should be fed into the biomarker development and study design development process. The second thing is, there has been a lot of talk about having some kind of a biomarker consensus meeting or think tank to come up with standard operating procedures and standardized measurement. I think that should be fast tracked. Finally, with regard to interesting existing scientists and companies in studying autism, which is not a huge population, there are a variety of conditions which have overlapping biochemistry and pathophysiology with regard to inflammation and oxidative stress, various neurodegenera- tive diseases across the life span, obesity, diabetes, and work has been done in those domains. People who have been working in those domains, that could be recruited to this effort so that it wouldn’t have to start from scratch in autism. Dr. Pessah: A major gap in our knowledge is the neurodevelopmen- tal toxicity of some major priority pollutants that fall below the radar screen of the EPA. These are non-dioxin-like. They have tremendous potency toward certain signaling systems that may be very relevant to autism and other neurodevelopmental disorders. I think we need to promote applications that address low-level exposures that try to understand specific mechanisms that are relevant to autism. I think we can also use cell samples from autistic kids, preferably primary cells, to try to validate or further understand how these mechanisms are hypersensitive or insensitive to certain mechanisms. 

276 AUTISM AND THE ENVIRONMENT Dr. Spence: I have nothing new. Everything that everybody else said, and one other thing I stole from the father of one of my patients. He said, we ought to be able to look at the brain and decide if there are toxic elements. He said, can you look at the brain and see if there is mercury in them? I said, not with the technology that I know about. But I think the imaging technology is getting better, so maybe there are ways of development of technology to look at toxic exposures directly in the nervous system or in other systems. Dr. Swedo: This is the advantage of being at the end of the table; we can go fast. I would like to suggest that medical and genetic workups be done on every child who is suspected of having autism. Typically they are sent off to the waiting list for the developmental clinics without having had an EEG done, or other very basic workup. If the child wasn’t able to walk, we wouldn’t be sending them off for a psychologist to evaluate them. We need hypothesis-generating research. We heard about the NM-1 studies, but there are some case series that have been ignored. I would like to see genetics added to all of the epidemiologic studies that we have heard about this morning. If they aren’t already being collected, genetic samples should be, and would certainly get at that question of genetic susceptibility to environmental risk. Common measures across the patient population to allow compari- sons. Common measures of assessment, of recovery, of response, but also as we have heard, of biologic measures. I would like to propose a plan that we have follow-up evaluations for the IAN database. I know that Paul Law and the folks at Autism Speaks are working on ways that those individuals who can’t currently participate in research rapidly could. I will finish by echoing Gary Goldstein’s plea that we move very quickly to a cystic fibrosis or cancer-line treatment network. I know that the Autism Treatment Network and others have that as a goal, so they would need additional support to make that happen. Dr. Zimmerman: I would like to make two strategy points. First, I think there is a base for both a theoretical approach to this, but also an empirical approach. The theoretical approach is, all the scientists are here, and finding a neurobiological answer to why we have this problem. Second, I think the empirical approach would be a meta-analysis of all the treatment paradigms that have been tried to date, including biomedical, dietary, nutritional, home-based, or relationship-based 

PROCEEDINGS 277 treatments, because I think there is a probabilistic and causal relationship there. Dr. Coetzee: I am like Alan, I don’t have a particular stake in this disease. I come from the multiple sclerosis world. In thinking about what I have observed in the last couple of days, there are two areas. One is, I think the cystic fibrosis model is particu- larly powerful for another area, and that is the foundation’s approach to companies, in terms of contracting with them to do research on a population of 30,000. There is not a big market for cystic fibrosis drugs, but they have managed to invest $200 million over the last 6 years in company research. So I think that is a powerful model for the private sector to stimulate pharmaceutical companies to be engaged in this process. It is not the usual big players, but it is the small innovator companies that will come out of the young people who get recruited to be researchers in autism research and other areas. I would also suggest that we think about looking at other diseases that may have an adult onset, but that have complex genetics that have environmental triggers and environmental factors for which we don’t yet know causes. I think we don’t want to necessarily reinvent wheels that may have already been invented elsewhere. Dr. Cohen: I come here as a member of the Forum. This has all been very new to me. I would like to thank you all for an extraordinary 2-day education. From that admittedly uninformed perspective, two things struck me. One was the absence of imaging in the discussion. I was going to say fMRI, except Martha Herbert last night advised me that there were other directions to go. The second thing I missed was a deeper discussion of viral patho- gens. Like many of you who have been struck by a lot of the anecdotal conversations, that struck me as well, some of the conversa- tions about the kids with severe GI distress that responds to antivirals, just around the table, the very late onset with possible viral etiology. While I wouldn’t necessarily say that that is my lead hypothesis, it certainly strikes me as a significant element that has to be considered. Dr. Fombonne: Two ideas. One is to go back to the idea of setting up a study of discordant twins. I think the rates of twins are going up, and that would be an opportunity to look at not only environmental risk factors, but also maybe product risk factors, what makes the outcome 

278 AUTISM AND THE ENVIRONMENT different in that case. Second, it follows up on the last point. I think we should probably look at environmental exposure which might increase the rate of mutations in germ cell lines of fathers in particular. That would be something which would also be to follow up. Dr. Insel: There is not a lot left at the end of the row here, but let me give you one from Ian Lipkin which hasn’t been mentioned today, although David Schwartz started to get there, using microbiomics approaches, that is, this new way that Ian Lipkin described of looking at all the potential microbes through sequencing, not through culture. I think intuitively that is a great opportunity in this area for a lot of reasons. It fits in with the immune story, it fits in with potential time trends. There is just a whole range of things that would help you understand what is going on, and it is completely untapped. It is now doable. It wasn’t doable 6 months ago. But this is a place one might go. Two others. Mentioned yesterday, not today so much, Sue Swedo started by saying there are plenty of genetic diseases, Mendelian diseases, that have autism as part of the story. If you had anything else that increased the risk as much as Fragile X, you would want to know why. It seems to me that is also an opportunity. The last thing. Several people have mentioned international efforts. I do think, besides finding these natural experiments and special cohorts, there are a lot of things going on in other countries that we need to be more aware of. I was recently at a national autism meeting in The Netherlands. Someone came up to me and said, I have been collecting CSF on several hundred people who now have autism. Would that be of any interest to anybody? So there are opportunities out there that we need to be looking beyond our own borders to try to exploit. Dr. Leshner: That was pretty spectacular. I’m not quite sure what to do now other than to open it up to the broader floor with the same rule. You get one minute, no more, from the audience. Line up at the microphones. You have got to yell, we are collecting all of this. Name, where you are from, and no speech. A one- minute opportunity. 

Session IX Discussion with Meeting Participants and Audience Participant: I am from George Washington University. One of the things that has been maybe just hinted at here but hasn’t really been addressed is the possibility—and also, this would involve parents and clinicians as well, is the use of response to medications, to phenotype individuals for further study. As the parent of a child with Asperger’s syndrome, he has been thrown bucketloads of drugs, most of which have absolutely no positive effects, lots of negative effects. But if we can tap that database and sort out those kids that may respond well to antidepressants or antipsychotics, those are probably good biological phenotypes that can be useful in genetic or genomic or other studies as well as metabolomic studies. Participant: Dr. Nancy O’Hara. Two points I would like to follow up on from the panel. One, the University of Maryland has just partnered with Autism Research Institute to start to collect and bank tissue samples, not just brain samples, from children and adults with autism. I think that is very important when you look at environmental factors. Second, Dr. Sidney Baker, who I think is one of the most brilliant clinical minds in this research, is launching a beta trial of metagenesis, which is a collaborative technological tool to collect data from parents. Ten thousand parents will be receiving letters next Friday. I offer that to the board to look at that as a means, not of scientific rigor, but as a means of data collection, to start to look at what bio- markers, lab data, treatment protocols might be out there. If you don’t receive a letter as a parent or members of the board, you may go to http://autism.com/ to look at that. I think that is something we can use to start collecting some of these data. Participant: Richard Deth from Northeastern. There is a meeting at the same time as we are meeting here of the think tank from DAN and the Autism Research Institute. I want to bring greetings from them, because there has been some back and forth and there will continue to be back and forth. They represent an organization of resources for cooperation in these kinds of relationships, especially parent based as opposed to fundraising. Different organizations have different foci and so forth. 279 

280 AUTISM AND THE ENVIRONMENT That group is really anxious—you will probably receive something official from them, offering some cooperation in whatever way possible from that group. The other thing is, Dr. Insel, you pointed out the importance of private funding because the R01 process is fraught with problems. I, for example, had an R21 where the primary reviewer just cut and pasted the thimerosal statement from the FDA Web site instead of reviewing my grant. So this is really important for the environmental factors that other people step up and never get through the NIH peer review system. Participant: Mary McKenna, University of Maryland. We are talking here about environmental and inflammatory factors and how they impact on metabolism in the developing brain. I just wanted to point out, related to what he just said, that brain metabolism studies have difficulty already in study sections. There is no home for them. They get bounced from study section to study section. I think that is an issue that needs to be dealt with. Participant: Alaina Fournier, Department of Health and Human Services. I have been working with Dr. Raub for the last 3 months in preparation on this topic for this meeting and future work with this topic, but now I am on a different detail at the NCI (National Cancer Institute) Office of Liaison Activities. I would like to suggest using the NCI model of getting consumer advocates involved in the scientific process. They have several different programs that have been long, positive, and successful in getting the consumer advocates involved in speaking with the director about the research agenda as well as the peer review process, and just keeping open lines of communication between the two groups. Participant: I didn’t know she was going to talk, but this falls perfectly with what I was going to mention. I have had this conversation with Dr. Raub as well. I think it is very important to remember that just as our kids have varying—display different aspects that are all different, it is also the way our community is. We have a lot of different viewpoints. It is a wide spectrum. The point I would like to make about the partnership is, I hope that— and it sounds like from what she just said that the partnership isn’t going to be—the determining factor isn’t going to be that we have millions of dollars to bring to the table to be able to have our voices heard at the table. 

PROCEEDINGS 281 So I think that it would be very important to remember that—to bring all the different advocacy groups to the table so that we all have a place and a voice, because these are all our kids. So I think that is very important to remember. Participant: Hi, I’m Beth Roy with Social and Scientific Systems. I have been the director of the Pediatric AIDS Clinical Trial Group Operations Center, and I was very interested in hearing about the possibility of looking at a model like that, that I think was successful in bringing together public–private partnerships, the involvement of consumers, and dealing with a very complex disease that was very multidisciplinary. So I just think it would be a good model to look at. Participant: Jim Moody with SafeMinds. To paraphrase a famous quote, money is the mother’s milk of science. When this country has faced a crisis in the past, the Manhattan Project in World War II, the opportunity to land a man on the moon in the 1960s with NASA, and more recently responding to crises like bird flu, SARS, even the recent pet food crisis, there has been a political response which helped drive the scientific response. So one thing that would I think help would be if this panel of distinguished scientists could urge upon national government the same kind of commitment that has called forth the sense of national urgency at the political level, and that begins at the presidential level, to declare a national emergency, to marshal all of the resources that we can bring to bear on this crisis. At the present rate of increase, which I think Mark Blaxill said was 10-fold since 1990, in two or three generations every child will be born with autism or related neurological deficit or some sort. These kids now are the canaries in the mine, and there needs to be the strongest possible social response to that. America is full of the top scientists in the world. We can solve any problem if we put our hearts and minds to it. The cost alone of the epidemic would justify that, if not the moral imperative. Thank you. Participant: A quick question to all of you here from NIH and the Institutes. How is the so-called funding from the Combat Autism Act recently passed by Congress supposed to be directed or targeted? Dr. Leshner: I can answer that. It has to be appropriated. It has not been appropriated. It was just authorized. Participant: A quick comment. I am with the Environmental Protection Agency. My name is Mark Corrales. I just wanted to offer 

282 AUTISM AND THE ENVIRONMENT myself as an unofficial contact. I am in the Administrator’s office. I am familiar with the various datasets and offices. EPA is a big place. It is sometimes hard to find the right information or people. So I just wanted to offer myself as a contact, if I can help, in the spirit of partnerships, data sharing, and so on. Dr. Leshner: I think we have gotten to the moment. It feels like I don’t want it to end. One thing that I would like everybody to do, because I think it has been really terrific, and we have been so well behaved and not applauded anyone, we need to take a moment. I think this has been spectacular, and I hope I am right. I would repeat the comment I made fairly glibly before, that is, this is a very important start, and if we don’t do some- thing, then shame on us. Thank you very much.