2

Environmental Measures

This chapter begins with information on environmental measures provided in advance to workshop participants via Kwan et al. (2013, pp. 2-4), followed by two questions that panelists were asked to address. For clarity, the content from Kwan et al. (2013) is shown in block quote text throughout this workshop summary. The third section of the chapter provides highlights of the panel members’ remarks and open discussion with the audience about the two questions.

BACKGROUND ON CURRENT PLANS

The primary objective of the National Children’s Study (NCS) is to examine the relationships among exposures and outcomes that affect children’s health and development. While the NCS is considering a broad array of exposures, including characteristics of the family and neighborhood, this discussion will focus on a few exposure and outcome examples to probe some specific design questions. Current plans include, but are not necessarily limited to, collection of the following samples:

• Household dust

• Blood

• Urine

• Questionnaires on exposures and the social environment

• Placenta and cord blood at birth



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 5
2 Environmental Measures T his chapter begins with information on environmental measures provided in advance to workshop participants via Kwan et al. (2013, pp. 2-4), followed by two questions that panelists were asked to address. For clarity, the content from Kwan et al. (2013) is shown in block quote text throughout this workshop summary. The third section of the chapter provides highlights of the panel members’ remarks and open discussion with the audience about the two questions. BACKGROUND ON CURRENT PLANS The primary objective of the National Children’s Study (NCS) is to ex- amine the relationships among exposures and outcomes that affect chil- dren’s health and development. While the NCS is considering a broad array of exposures, including characteristics of the family and neigh- borhood, this discussion will focus on a few exposure and outcome examples to probe some specific design questions. Current plans include, but are not necessarily limited to, collection of the following samples: • Household dust • Blood • Urine • Questionnaires on exposures and the social environment • Placenta and cord blood at birth 5

OCR for page 5
6 DESIGN OF THE NATIONAL CHILDREN’S STUDY These samples could be tested for heavy metals, pesticide residues, semi- volatile organic compounds, and pharmaceuticals. Outcome measures include but are not limited to • Linear growth rate and body mass index as a proxy for general health • A metabolic screen of serum total protein, blood urea nitrogen, cho- lesterol, iron, and calcium as a proxy for nutrition and dietary exposure • Frequency and duration of health system encounters for respiratory illness as a proxy for pulmonary health • Timing of standard neurodevelopmental landmarks and any deviation from adjusted trajectory as a proxy for cognitive and social development The NCS also plans to use general exposure data collected at the munici- pal or neighborhood level (water quality, air quality, known industrial pollution) by either direct specimen collection or extant data collection. The current data collection plan is based on an approach that uses a core questionnaire administered at every childhood visit, plus supplemental modules to be administered to specific participants or subpopulations based on events and conditions such as age, developmental stage, and other triggers such as specific exposures or hospitalizations. Modules may be administered on a “missing by design” basis. There are at least two aspects of this missing by design approach: modules triggered by age, exposure, or specific events; and a “validation sample” ­ pproach. a Triggered modules based on age, for example, make most sense where either scientific evidence indicates that exposures only at certain ages are likely to cause health concerns or a knowledge gap exists. A validation approach may be useful, for example, when there might be two ways to measure a specific item, one inexpensive, the other expensive but more comprehensive. A smaller random sample may be assigned to have the expensive measure taken, but all respondents would provide the inexpensive measure. The data from the samples with both data might be used to establish a model to provide a correction to the inexpensive data that are available for the entire sample. This only makes sense if there is knowledge that such a model exists or could feasibly be developed and could provide an improved estimate based on the inexpensive data. In addition to questionnaires, other mo- dalities for data capture such as sound recordings, images, geographic movements, and mapping of social interactions and networks will be used. The NCS emphasizes data collections early in pregnancy and early in child development because the largest knowledge gaps, and perhaps the most critical events, occur during those time periods. Pregnancy data collections are scheduled, if possible prior to about 20 weeks of gestation and once later in pregnancy. Data collections for

OCR for page 5
ENVIRONMENTAL MEASURES 7 children are scheduled at birth and then every three months for the first year and every six months until five years old, for a total of 13 opportuni- ties. Seven will be face to face, including biospecimen and environmental data collection. The other six are remote collections, typically by tele- phone. Subsequent data collections have not been specifically scheduled but will be on average about every other year until 21 years old, for a t­ tal of eight additional opportunities. The visit schedule is flexible in o that children will not have assessments precisely at a given age, but within a window of several weeks around a particular age. The NCS Examples of Potential Exposures amd Outcomes table1 indicates that biospecimens of blood and urine will be collected from the mother prenatally, at birth, and when the child is 6 months and 12 months old. From ages 2 years to 5 years, blood and urine will be collected from the child. At each of these opportunities, except birth, the mother would complete a questionnaire, and household dust, among other samples, will be collected. The table Potential Environmental Exposures of Interest2 lists environmental exposures of potential interest to be measured in all NCS participants (general) and in a subset of participants (selective). Selective sampling will be based on the principle of enriching for a population more likely to have a risk of a particular exposure. For each exposure, corresponding examples of target analytes are listed. The rationale for biospecimen or environmental sample collection examples of target analytes, proposed and alternative measures, and potential health outcomes of interest are provided for each exposure type. Another column lists the preferred data sampling modality method to be used by the NCS with optional approaches for the Committee’s review and consideration in adjoining columns. The intent of the NCS is to have the highest quality biospecimen or envi­ ronmental sample available, but the NCS may not have the resources to analyze each specimen or sample for each analyte in real or near time. Consequently, processing and storage of the specimen or sample are im- portant considerations that will be based on analyte stability. The table does not represent all the specimens and samples the NCS intends to collect but is limited to those specimens and samples targeted to assess selected environmental exposures. 1To view the table, see http://www.nationalchildrensstudy.gov/research/workshops/ Pages/nationalacademyofsciencesworkshop.aspx [June 2013]. Click on National Children’s Study Examples of Potential Exposures and Outcomes. 2To view the table, see http://www.nationalchildrensstudy.gov/research/workshops/ Pages/nationalacademyofsciencesworkshop.aspx [June 2013]. Click on Potential Environ- mental Exposures of Interest.

OCR for page 5
8 DESIGN OF THE NATIONAL CHILDREN’S STUDY QUESTIONS ON ENVIRONMENTAL MEASURES Given the challenge as stated in the Children’s Health Act of 2000 to “perform complete assessments of environmental influences on chil- dren’s well-being,” does the proposed visit schedule and sample col- lection balance the complex requirements? Specifically comment on the proportion of different types of data collection—primary environmental sample collection, use of biological specimens for biomarkers of expo- sure, and use of secondary sources including retrospective analysis for environmental exposures. Considerations may include the following: 1. Are the proposed measures (biomarkers, questionnaires, physical measures) the most appropriate to assess exposures of interest? If not, what measures should be taken? 2. How should the NCS prioritize decisions regarding exposure assessments? 3. Some examples of factors to consider are a. potential public health impact of the outcome, b. technical feasibility including timing of data collection with regard to potential developmental vulnerability, c. scientific opportunity to address knowledge gaps and illuminate developmental pathways. KEY POINTS OF THE DISCUSSION Marie McCormick (Harvard University School of Public Health and the Harvard Medical School) moderated a panel that consisted of ­ ntonia A Calafat (Division of Laboratory Sciences, National Center for Environ- mental Health, Centers for Disease Control and Prevention [CDC]), Nicole Cardello Deziel (Occupational and Environmental Epidemiology Branch, National Cancer Institute), Melissa Perry (Department of Environmental and Occupational Health, George Washington University), and Linda Sheldon (National Exposure Research Laboratory, Environmental Pro- tection Agency [EPA]). The panel members focused their discussion of environmental measures and the timing of those measures by introducing the audience to the results of a workshop held in 2010 sponsored by EPA and the National Institute of Environmental Health Sciences (NIEHS) to consider exposure metrics for the NCS (Environmental Protection Agency, 2010). After Sheldon described the EPA-NIEHS workshop, a dis- cussion among panelists about the proposed measurements and schedule ensued, followed by open discussion with the audience. The panelists then addressed prioritizing decisions, again followed by open discussion with the audience.

OCR for page 5
ENVIRONMENTAL MEASURES 9 Proposed Measurements and Schedule Description of EPA-NIEHS Workshop Sheldon said that the charge given to the EPA-NIEHS workshop par- ticipants was to develop innovative exposure metrics and to look at the minimum amount of exposure data to collect to be able to assess three specific health linkages with exposures: air pollution and asthma, insecti- cides and neurological development, and endocrine-disrupting chemicals in reproductive endpoints. She explained that workshop organizers felt it was important to consider the minimum exposure data needed, rather than looking at all possible ways to measure exposure. Three separate workgroups conducted their work before the EPA- NIEHS workshop. Each workgroup had an epidemiologist, a toxicolo- gist, and two people who worked in exposure, ensuring cross-discipline coordination about the chemicals, the time periods of susceptibility, and, knowing that information, the feasibility of collections. About 50 people attended the 1.5-day workshop. Each workgroup presented its findings on chemicals of interest, kinds, routes and pathways of exposure, critical time windows, biological samples, environmental samples, non-measurement approaches, protocol recommendations, and research recommendations. Sheldon noted the EPA-NIEHS workshop provided background and justification for its proposals. Two important topics addressed beforehand were the definition of an exposure metric and criteria for a good exposure metric. EPA-NIEHS workshop participants defined an exposure metric to include direct measurements as well as measurements combined with other data via a model that would provide the ability to estimate an exposure. They defined a true exposure metric as one that would indi- cate biologically relevant exposure during the entire time window of susceptibility. It was further expressed that an exposure metric might be a biomarker in urine if it were related to exposure and that a biologically relevant exposure metric is one where concentration in urine provides information about the concentration at the biological target where the effect would take place. Blood lead was noted to be an excellent biologi- cally relevant exposure metric. Sheldon said defining exposure metrics and what makes a good m ­ etric provided the standard for assessing potential metrics at the EPA- NIEHS workshop. The two important criteria used were whether the mea- surement leads to biological relevance and whether a sample collected on one day provides information about exposures over the entire period of susceptibility. She said this second criterion is extremely important, especially when considering prenatal exposures. The EPA-NIEHS work- shop participants selected five time windows of exposure: first trimester, third trimester, first year after birth, years 1-4, and puberty (for endocrine

OCR for page 5
10 DESIGN OF THE NATIONAL CHILDREN’S STUDY disruption). She said they also recognized that, though important, col- lection of measurements in the first trimester may not be possible for all women, and they considered how measurements collected during the third ­ rimester might relate to exposure during the first trimester. t EPA-NIEHS workshop participants also discussed biological ­ atrices. m Blood for the mother was considered important for the first trimester and third trimester. For the child, while collecting blood may be important, only a limited amount can be collected from a newborn, and it is impor- tant to use a “blood spot” wisely. She noted the EPA-NIEHS workshop also viewed the collection of urine as important because it is easy to col- lect and is a good exposure metric, and that breast milk can be a source for measuring many exposures. According to Sheldon, EPA-NIEHS workshop participants gave house dust the highest priority among environmental samples. A single sample can be used to measure not just the concentration per square meter in the dust, but also the dust loading can be collected with a standard collection mechanism during a visit. Vacuum methods appear to be most feasible, and a protocol can address the collection of multiple analytes from a single dust sample. A dust sample can yield measurements of organics, metals, and biological agents. Shelton concluded by noting many of these recommendations are similar to the protocol proposed in Kwan et al. (2013). Discussion Among Panelists About Measurements and Schedule Perry noted minimizing the use of questionnaires, especially to col- lect environmental exposures, is important because of the burden asso- ciated with questionnaires, although a small number of questions can be used to collect information for specific key exposures that cannot be collected using a biologically relevant exposure metric. Deziel said very specific well-designed questions about pest treatments (e.g., “Do you treat for termites?”; “Do you treat for fleas and ticks?”) provide good correlation with actual measurements of the expected active pesticide ingredients in dust. However, some information cannot be collected on a questionnaire because people do not know whether, for example, they have poly­ hlorinated biphenyls (PCBs) in their home or if they have poly- c brominated diphenyl ether (PBDE) flame retardants in their televisions or couches. Dust can provide useful exposure information for those types of chemicals. Deziel echoed some of the benefits of dust measures, which she said can be very useful in providing information for chemicals for which there are no good questionnaire items. Deziel said researchers at the National Cancer Institute (NCI), Univer- sity of California at Berkeley, EPA, and elsewhere have shown that even

OCR for page 5
ENVIRONMENTAL MEASURES 11 non-persistent chemicals tend to be rather stable over time once they are in a residential environment. For example, some of the work that NCI has done resulted in observed interclass correlation coefficients of 0.6, 0.7, 0.8, and 0.9 for pesticides, polycyclic aromatic hydrocarbons, PCBs, and even some PBDE flame retardants.3 Hence, if these are the analytes of interest for the NCS, dust measurement could provide useful information about exposure during critical time periods, such as preconception and the first trimester, when the study is unlikely to actually have samples collected. Deziel said the tables and documents provided by NCS do not make the intended method of dust collection clear. Possible methods include a vacuum bag, a high-volume small surface (HVS3) sampler, a subtle dust plate, a dust wipe, and an air sample. Almost all of these methods have been considered as part of the NCS sampling protocol in the past. She stated collection of some sort of bulk dust sample is important. She noted in some of her research at NCI, she and her colleagues have compared concentrations of chemicals from a participant’s vacuum bag or vacuum canister with a more standardized vacuum approach, the HVS3, with very good correlation between the two methods for a range of chemicals, like pesticides, PCBs, and polycyclic aromatic hydrocarbons (PAHs). Hence, she said, a vacuum bag may be a feasible way to obtain useful exposure information using a method that is not too burdensome on the data col- lector or the participant. Calafat said, referring to the EPA-NIEHS workshop, a good measure- ment plan would strive for the minimum data that would provide useful information. Since not one approach will fill all information needs, it is very important to get a minimum set of environmental measures, a mini- mum set of questionnaire information, some residential data, and, last but not least, biological specimens. Different biological specimens—blood, urine, and breast milk—are appropriate or most appropriate for certain chemicals. Persistent chemicals are measured in blood. Because many of these persistent chemicals are also lipophilic and partition into fat, breast milk would be an excellent matrix for assessing postnatal exposure to some persistent chemicals. Calafat said monitoring the impact of non-persistent chemicals is a great challenge, but unfortunately the market is moving from persis- tent into non-persistent chemicals. Non-persistent chemicals metabolize quickly and, for the most part, people are exposed to them through episodic events. When a non-persistent chemical is combined with epi- sodic exposures, such as those that might be encountered through diet, and not through regular exposures, such as through the use of per- 3Later in this chapter, Deziel notes that correlations above .7 have been observed in samples taken months or even years apart.

OCR for page 5
12 DESIGN OF THE NATIONAL CHILDREN’S STUDY sonal care products, the issue of variability arises. The concentrations of these chemicals as measured in urine have tremendous variability, and a measure taken on a given day may not reflect exposure during the critical window of susceptibility or reflect exposure in the future. She said no single solution currently exists to identify episodic exposures to c ­ hemicals that metabolize quickly. One approach would be to collect as many samples as possible, as well as information about when the sample was taken and the time of the last urination. All could be important when analyzing the data. Calafat said her laboratory is analyzing data from the initial phase of the Vanguard pilot consisting of samples for about 500 women. It was a nationally representative sample, and the urine samples were analyzed for a suite of different chemicals including some phenols, phthalates, and metals. Despite the variability, she said her lab is already seeing very important differences in concentrations of some of these chemicals, depending on the demographics of the population. Just because these measures are variable, she noted, does not mean that they are not useful. In some cases, these chemicals are very prevalent in the environment and could be introduced into the sample through the materials used for collec- tion of the samples; thus, prescreening of materials might be considered to make sure this is not an issue. She noted prescreening would be very important if the goal is to measure metals. Perry stated the proposal in Kwan et al. (2013) for the collection of biological matrices, to include urine, blood, blood from the infant, and cord blood, seems to be well put together and well timed to the extent that windows of susceptibility are understood. She acknowledged the chal- lenges associated with identifying and collecting information from the prenatal cohort, although the prenatal period represents a critical ­ indow w of exposure. She said the proposed protocol does not prominently feature collection of breast milk, which, although challenging to collect, is impor- tant to consider. She agreed with Sheldon that the collection of a single blood spot for a newborn and planning how to use it are important, and she also concurred that house dust is a very important biologically rel- evant matrix of exposure. Perry noted the proposed sequencing chronology of measurements— possibly first trimester, third trimester, at birth, and then going forward— looks like a logical and a well-thought-out trajectory of sampling over time. However, Deziel said assessments of the appropriateness of the timing and method of sample collection usually depend on the research question of interest, while the current NCS proposal contains no spe- cific research questions. She characterized the NCS proposal instead as a broad-based approach to collect a lot of detailed information to address numerous future research questions. Given this model, the repeated dust,

OCR for page 5
ENVIRONMENTAL MEASURES 13 blood, and urine measurements proposed seem appropriate and strong metrics to collect. Deziel also observed that Kwan et al. (2013) did not mention GPS measurements or obtaining a good residential history, both of which she said are important. She said with GPS coordinates at a residence and answers to a few questions about how long the individual has lived in that residence, the growing number of rich publicly available databases can provide useful information about exposure. Examples she cited included EPA’s Toxics Release Inventory (TRI), National Air Toxics Assessments (NATA) databases,4 and pesticide-use databases maintained by some states. These databases may provide a way to get exposure information during critical time windows when samples were not collected. Open Discussion About Measurements and Schedule Naihua Duan (Columbia University) asked how well first-­rimester t exposure can be assessed through retrospective recall to fill in miss- ing data if needed. Perry replied recall depends on the contaminant. She reminded the audience of Deziel’s comments that, for some more u ­ biquitous invisible compounds, no one is fully aware as to when he or she is being exposed. At the same time, one can imagine remembering a pesticide event or using a paint or solvent during the first trimester of pregnancy. Questions of this type might provide information about prenatal exposures during a very specific period such as during the first trimester, especially if a woman is asked immediately postpartum. How- ever, questionnaires may not assess persistence of exposure. Deziel added many of the studies in which she is involved look at samples collected months or even years apart and still see interclass correla- tion coefficients of 0.7 or higher for many pesticides, PAHs, and PCBs. The repeatability of these samples will depend on the physical and chemical properties of the chemical and consistency of use. Over a several-year time- frame, she and her colleagues observed a single sample may be representa- tive of a period of months or years for some chemicals. They are also look- ing at questions to determine how well people recall pesticide use during different time periods of pregnancy and how well that recall correlates with dust collected. However, she thinks relying on recall could be challenging. Sheldon noted one issue brought up in the EPA-NIEHS workshop was relocation: Every time a person moves, samples need to be recollected at the new site. At a minimum, it is important to know whether (and how long) the person was in the environment where samples are collected. 4These databases can be found at the following links: http://www.epa.gov/tri and http:// www.epa.gov/nata [June 2013].

OCR for page 5
14 DESIGN OF THE NATIONAL CHILDREN’S STUDY Duan noted the next workshop session would include remarks by Irwin Garfinkel (Columbia University) about the importance of the first- born. Duan asked whether much is known about mobility, especially as a family prepares for the birth of the first child, and any information on how mobility might compromise the persistence of an exposure. McCormick replied, in her experience, a young couple, particularly for the first-born, is likely to move into a house because they have a child. Deziel added this question highlights the importance of taking a residential history to learn the relevant timeframe of the sample. Michael Bracken (Yale University) said his studies show 30 percent of families move within seven years after a birth. Greg Duncan (University of California, Irvine) noted in one of his national studies, mobility rates were just under 20 percent per year. He said in general, rates are higher among low-income than high-income families and are higher among younger families. Most mobility is local within a county but still is a change of household residence. Duncan said one of the design options under consideration is to recruit some of the children from a hospital or prenatal care providers, and then have subsequent births to the original mothers become part of the sample. The advantage of including subsequent births is that pre­ conception and very early prenatal information on exposures can be obtained for these subsequent births. A problem is that they are all second and higher priority births, so there is no preconception and early prenatal information about first births. He asked the panel about different hypoth- esized effects of exposures for first births versus subsequent births and the importance of the distinction. He also asked about the importance of obtaining exposure information on first births quite early in the pre­ conception period. Perry noted even though collecting pregnancy and prenatal data is challenging, the data are critically important for generating information about in utero exposure related to new findings about prenatal bases of adult disease. She said one flaw she sees with foregoing prenatal sampling in anticipation of collecting information on the second- or third-born is the fact that individuals who have children with health problems may not go on to conceive and reproduce again. If the study relies solely on second births for prenatal and pregnancy information, there would be no information to study first births affected by an immediate or chronic disease. She went on to say that there has been active discussion about preconception and how challenging it is to collect preconception data, but perhaps “missing by design” approaches or validation studies, where the NCS could collect preconception data from subsamples of women, would help fill the data gap. She stressed the importance of information about exposures in infancy and early development as they relate to predicting adult disease. Noting, however, that if the NCS can collect early expo-

OCR for page 5
ENVIRONMENTAL MEASURES 15 sure data, the study would be positioned to accomplish something very significant, because few studies have realized a large sample with early exposure information. Deziel said by recruiting the second child, not only is there unbalanced ­ exposure information on these siblings, but also the exposures of these siblings will be correlated. As a result, they would have to be analyzed separately or through different statistical techniques. She expressed doubt this approach would provide sufficient statistical power to analyze pre­ natal and preconception periods. Calafat said there could be differences between the first-born and sub- sequent children in terms of exposure to persistent pollutants. Jean Kerver (Michigan State University) said the interpregnancy interval would also have to be considered. A woman could have decreased nutrient stores after her first birth if she does not have time to build them up before the second birth. Decreased nutrient stores would have an impact on prenatal biomarkers such as vitamin D or other fat-soluble vitamins, which Kerver said would be a big consideration in going to a design that would elimi- nate first births by design. Garfinkel asked the panelists to sharpen the question about the effects of exposure. He said he understands the prevalence of exposures may differ and there may not even be second births if the first child has health issues. He noted the key question for him is if the biological effects of exposures differ for first and second births. Nigel Paneth (Michigan State University) emphasized much of what is important in child health, particularly in many neurodevelopmental disabilities, birth defects, and preterm births, has been determined by birth. He said postnatal environmental measurements are irrelevant to the causation of those central components, which puts the onus squarely on the prenatal period. He described an analysis in which he is involved of MOBAND, a combined Danish and Norwegian birth cohort that totals 200,000. All have prenatal collections with prenatal blood and urine, but none has prenatal environmental house exposures. The possibility of collecting prenatal exposure environmental information is unique to the potential NCS. Kerver said she would add diet back into the conversation and expressed concern that Kwan et al. (2013) stated that “a metabolic screen of serum total protein, BUN, cholesterol, iron, and calcium” may be used “as a proxy for nutrition and dietary exposure.” She said those measures are not valid estimates of, and cannot be used as a proxy for, nutritional status or dietary exposure. Additionally, she said it is important to collect dietary exposure, like other exposures, in real time and prenatally. Sheldon identified two reasons to collect dietary information: nutri- tion and exposure to contaminants. Her work has shown much variability

OCR for page 5
16 DESIGN OF THE NATIONAL CHILDREN’S STUDY day to day in dietary exposure to contaminants for any particular indi- vidual and shows a study can only get the extremes. How to Prioritize Decisions The panel went on to address the second question: “How should the NCS prioritize decisions regarding exposure assessments?” Discussion Among Panelists About Prioritization Sheldon started the discussion by saying when she looked at the cur- rent NCS strategy she thought hypotheses to test were needed. As one starts to look at different hypotheses and what is known about environ- mental pollution, she said, key elements of the strategy are reflected in the sampling scheme, measurements, and operational feasibility. This leads to almost the same strategy as the current proposal. She said the health outcomes being targeted should be kept in mind. The important questions become what can be measured and how well measures indicate exposures that are persistent or persistent for the time window of interest. A sample collected in one day has to represent an entire period. She said it is important to focus on groups of chemicals that are going to have a reasonable correlation or interclass correlation with different time periods. She further observed that there are new models for air pollution and three relevant environments: the home, the community, and the ambient environment. For some air pollutants, available databases and model- ing can retrospectively determine a person’s exposure over various time periods. Knowing the time window of interest and the exposures to moni- tor during that window are absolutely critical for prioritizing decisions. Whether measuring exposure for epidemiology, exposure analysis, or another purpose, there is often a tendency to focus on things already looked at and with existing knowledge about how to measure them. New analytical techniques in the future may allow for screening of tens of thou- sands of chemicals in matrices like house dust. As technologies improve, they may help in the analysis of archived samples. Perry expanded on the idea of using new technologies, referring to active progress in the United States and other countries in determining a way to consolidate multiple exposure measurements within one small sample of blood or urine. She reminded the audience that future analytic opportunities are dependent on success in collecting blood samples and cord blood samples. She suggested the use of personal monitoring. It is now feasible for individuals to wear personal monitors, monitor their physical activity, upload it online or in real time, and see personally cus-

OCR for page 5
ENVIRONMENTAL MEASURES 17 tomized data. She suggested NCS might consider inexpensive monitors that participants could wear to monitor indoor air quality or nutrition intake. She underlined the importance of being fully aware of technology trends in order to use them at an appropriate time. Sheldon followed up, saying many groups with National Science Foundation support are looking at this topic. As an example, a group with which she is working at North Carolina State University is using nanomaterials to generate power for monitors to collect the pollutant and a physiological response. When this technology is fully developed, a per- son would get a monitor on a patch and no further visits by staff would be needed, which would save on the cost of repeated visits. The key is having electronic transmission for long-term monitoring. Open Discussion About Prioritization Dorr Dearborn (Case Western Reserve University) commented on the availability of personal monitoring devices. He said with NCS funding, his group developed eight different residential air quality monitoring parameters that could be wirelessly downloaded and could detect when a cigarette was lit or gas cooking stove turned on. These devices are not the size that comes to mind when the term “personal monitors” is used, but they can also collect air particulates with laser light scattering. The researchers are almost ready to put them into some NCS participant homes to gain field experience with how they work. They could easily add a microphone and collect sound to get some sense of the source and nature of the sound on a continuous basis or put photocells into a child’s bedroom to track nocturnal light exposure. These ideas are feasible now, and they are seeking input on the value of adding the sound and/or light monitoring. Perry observed the technology for small cameras and other technol- ogy exists, but privacy issues are very real. It is not clear to what extent participants would be willing to use these devices. She noted people now seem to be willing to wear monitors for physical activity and have their data uploaded, and perhaps NCS respondents would also be willing to use these devices. The technology may be relevant to monitoring envi- ronmental contaminants and exposures as well as nutrition and physical activity. Bracken noted the NCS has not commissioned systematic reviews of the use of different technologies. He suggested systematic literature reviews on what has been done in other cohorts might provide more information than relying on the limited Vanguard data. He provided examples from his work. In one of his studies, they gave women monitors to wear in pregnancy during three different weeks, all using randomized

OCR for page 5
18 DESIGN OF THE NATIONAL CHILDREN’S STUDY nested subgroups. Although monitoring an entire cohort might be too expensive, he said it can be done in randomly selected subgroups. In one study cohort, subgroups of women wore monitors to measure environ- mental tobacco smoke; in a second cohort, subgroups wore monitors to measure electromagnetic field exposure, as documented in papers that have been out for 15 years now. A wealth of data is available on how to actively monitor pregnant women throughout their pregnancies. Duan called personal monitoring a very promising technology. He described statistical issues associated with personal monitoring data. First, a stream of personal monitoring data over a period of time would be the ideal way to address the variability question that Calafat raised earlier. It might provide information about variation in the short term and variation over a period of time. Second, personal monitoring provides an automated way to sample a person’s exposure across different activity patterns, rather than measuring the environment at a certain place or the residence. Deziel said although she supports consideration of these new tech- nologies, specifying how they would be used and the research questions they would address would be important to achieve a balance between costs and feasibility. Duan added the technologies are not all entirely new. The EPA has conducted a variety of personal monitoring studies for several decades, including the Total Human Exposure Study. The tech- nology is advancing, but there is a history of its use. Learning from what has been done is a good idea. Perry agreed the technologies are not new, but people seem to be in an era of greater ubiquity and involvement in embracing them. Sheldon observed the technology has to be inexpensive and easy enough to use so that repeated visits to the home are not required. One exciting idea is a self-powered, bandage-type monitor, which would make personal monitoring more affordable. If a device is cheap enough, a par- ticipant can throw it away when the measurement is complete. She said it is important to consider the cost of deploying and undeploying instru- ments in a large study. Calafat noted the NCS will collect a large number of specimens, both environmental and biological, and store them for future analysis. A key activity will be to document exactly how the samples were collected. In the future, these samples may be used to test new chemicals (or bio- logical or genetic information). She said the tradeoff between the cost of collecting and storing samples versus future use must be kept in mind. She noted if the cost becomes very high, one potential option would be to collect and/or store samples for a subset of the participants, much as is done within the National Health and Nutrition Examination Survey (NHANES). NHANES is a survey conducted by CDC that examines the

OCR for page 5
ENVIRONMENTAL MEASURES 19 general population, but most chemicals and some other biomarkers are only measured in a subset of the population that can be made representa- tive of the whole United States. Bracken asked about sample storage, collection, and stability, par- ticularly at the point of collection, saying samples are almost certainly going to be looked at for assessing environmental exposure in terms of gene-environment interaction or epigenetics. He stated it is crucial these samples are preserved for later decades, although many are not stable and deteriorate over time. Deziel said exposure metrics might be prioritized based on simple descriptive statistics, such as the percent detection or the range of vari- ability in the population, so there is adequate statistical power to look at questions of interest. She pointed to the 1.5 to 2 years of pilot data avail- able to inform the Main Study, but there is now a very short turnaround between the pilot and the Main Study. She urged the NCS to mine those data as quickly as possible so the pilot can really inform the prioritization scheme and related decisions. Sara McLanahan (Princeton University) asked about the tradeoff between the first trimester and the third trimester of pregnancy and the relative importance of measuring exposures during those two periods. Perry responded the answer depends on the outcome of interest, because the fetus goes through various stages of development during the three trimesters. If chromosomal abnormalities are of interest, data from the preconception phase are most important. If neurodevelopmental out- comes are most important, third-trimester data are most important. Even knowing the outcome of interest, she said, the precise timing at which measures are needed is not well known. Paneth commented one of the questions the panel was asked to address concerned potential public health impact of the outcome and noted some of the conditions that might be related to prenatal exposures are not very frequent in the population. Congenital heart defects, cerebral palsy, and type-1 diabetes, for example, are prevalent at less than 1 per- cent. According to the data in National Children’s Study (2012), the power in a sample of 100,000, even with 25 percent exposure, will barely pick up an odds ratio of 2. If the sample is cut down to 40,000 exposures and does not include measurements on all subjects, then the public health impact of what could be done with the NCS would be proportionately reduced. McCormick added NCS ought to be very explicit about which analyses (conditions) its data will and will not support. There may be conditions of high salience, such as autism, that may not be able to be addressed. Such limitations are important to state upfront. Carol Henry (George Washington University) said she was not sure that the panel has yet come to grip with priorities. The EPA-NIEHS

OCR for page 5
20 DESIGN OF THE NATIONAL CHILDREN’S STUDY workshop described by Sheldon recognized the NCS cannot be all things to all people, instead choosing three health outcomes to look at and to better understand the correlations between environmental exposures and health outcomes. If those three outcomes were emphasized in the NCS and data collection started, there could be progress in understanding them. Monitoring those three outcomes would be critical to the success of the NCS. Perry said in her comparison of the current plan to the EPA- NIEHS workshop, she determined that based on the chronology, includ- ing the prenatal with attempts at preconception, and the sampling plan over time, the NCS would have the matrices necessary and would be prepared to analyze the exposures of interest as they pertained to those three specific outcomes. Perry also noted that she did not want to dismiss the question about the prevalence of outcomes and whether or not the study will be ade- quately powered given rare events. In fact, she noted, there may not be an adequate number of actual diseases or precursors to disease to be able to study them. As an example, she cited her own studies on chro- mosomal abnormalities in sperm, which are potential precursors to con- genital abnormalities if that sperm is successful in fertilization. A number of predisease indices from DNA adducts5 to chromosomal abnormalities could be identified. At the same time, there are genetic and epigenetic mechanistic studies that blood in particular is going to afford and that would provide insight into mechanisms without having the critical mass of required cases. McCormick said she is uncomfortable when people talk about neuro- development as a specific outcome because neurodevelopment covers a fairly large number of relatively rare conditions. The study is unlikely to have the power to look at individual conditions. Edward Sondik (National Center for Health Statistics) asked about geographic diversity in the sample: using the original 105 primary sam- pling units (PSUs) versus using a smaller number of PSUs. He asked about the tradeoffs associated with a more clustered sample in terms of the diversity of geographic experience in the country. Calafat said the response depends on the chemical of interest. If the chemical is an agri- cultural pesticide, for example, the study would want to ensure agricul- tural areas are covered in addition to non-agricultural areas. She noted NHANES samples about 15 localities every year, yet they obtain represen- tative data for the whole U.S. population. The NCS will have a tradeoff among number of PSUs, cost, and the impact on outcomes and exposures 5From Wikipedia, “a DNA adduct is a piece of DNA covalently bonded to a (cancer- causing) chemical. This process could be the start of a cancerous cell, or carcinogenesis. DNA adducts in scientific experiments are used as biomarkers of exposure.”

OCR for page 5
ENVIRONMENTAL MEASURES 21 of interest. She added there may be differences in the use of some particu- lar chemicals by demographics, such as phthalates. If a chemical related to exposure in a residential use is driven by socioeconomic status and there are not enough localities to cover that particular chemical, it could be a problem. However, a more ubiquitous chemical may make a larger number of sites less relevant. Sheldon commented that community-level environmental expo- sures are important: air pollution, water pollution, and soil. Perry said in environmental exposure assessment, a wealth of information shows that environmental exposures are not uniformly distributed and differ- ent subpopulations are affected and exposed in different ways. People living in public housing, for example, are more likely to be exposed to pesticides and a variety of other chemicals and fumigants, as are people living in proximity to Superfund sites; thus, the importance of representa- tive sampling when it comes to patterns of exposure is obvious. Kerver also expressed concern that reducing the number of PSUs would result in a reduction in the variation in food intake, which would reduce variation in estimates of both nutrient intake and exposure to pesticides through food among sampled women. Roderick Little (University of Michigan, Ann Arbor) commented that one way of increasing power is potentially to increase the variability of the predictive variables of interest. He noted the proposed design includes a sample of about 10,000 births for “something else,” the something else not yet specified. He asked the panel about the promise in focusing some of the 10,000 births on areas where it might increase the variability of the predictive variable, which might improve the power. Perry noted that Kwan et al. (2013) covered two useful ideas: miss- ing by design and the validation approach. She said the sample of 10,000 could come into play, for example, to study children in the Salinas Valley or children of farm workers who may be excessively exposed to ambient pesticides and food residue pesticides. A missing-by-design study could take advantage of the fact that exposure is much higher in that subpopu- lation. Considering two different measurement mechanisms—one more expensive, more invasive, and perhaps more precise; the other easier, cheaper, but perhaps not as reliable—would be a perfect scenario for a validation study. She also said these ideas might be used for preconcep- tion opportunities given the challenges of recruitment. Calafat noted exposure for some chemicals is quite homogeneous, depending on how exposure is defined. If exposure is coming from use of personal care products that are also environmental chemicals, exposure is much more ubiquitous around the whole nation. She also noted many of the exposures proposed to be collected in the NCS are from indoor envi- ronments. There may be differences in exposure between populations in

OCR for page 5
22 DESIGN OF THE NATIONAL CHILDREN’S STUDY urban and suburban environments, because the population in suburban areas is more likely to be spread out versus an urban population where people share more exposures. In an urban environment, even though a person is not using a particular product or chemical, a close neighbor may be.