National Children’s Study Federal Advisory Committee (NCSAC) 10th Meeting
June 28 - 29, 2004
Holiday Inn Select
This meeting was held in conjunction with the National Children’s Study, which is led by a consortium of federal agency partners: the U.S. Department of Health and Human Services (including the National Institute of Child Health and Human Development [NICHD] and the National Institute of Environmental Health Sciences [NIEHS], two parts of the National Institutes of Health, and the Centers for Disease Control and Prevention [CDC]) and the U.S. Environmental Protection Agency (EPA).
Review Agenda, Meeting Goals and Objectives
Donald Mattison, M.D., NCSAC Chair, NICHD, NIH, DHHS
Dr. Mattison welcomed National Children’s Study Advisory Committee (NCSAC) members and other participants to the 10th meeting of the NCSAC. The stated goals of this meeting were to:
- Review and comprehend the background material outlining the various sampling design options
- Discuss the information and executive summary produced at the March 21–22, 2004, Sampling Design Workshop
- Answer a series of questions posed by the National Children’s Study (Study) Program Office
- Agree to a series of recommendations that will be provided to the Director, NICHD.
Additional goals of the meeting were to:
- Hear updates regarding protocol development
- Review revised hypotheses
- Determine feasibility or impact of joining the American Gene Environment Study (AGES)
- Discuss the changing role of the NCSAC and Working Groups as the Study moves into the implementation phase.
Program Office Update
Staffing, Protocol Development, Institute Updates
Peter C. Scheidt, M.D., M.P.H., National Children’s Study Director, NICHD, NIH, DHHS
Dr. Scheidt summarized Program Office events and activities since the last NCSAC meeting in March 2004. These events and activities included media coverage of the Study, additions of new Program Office staff members, protocol development, contract development, preparation of anticipated procurements, workshops, planning of upcoming workshops, and partnership development. Dr. Scheidt also reported on Study funding status, discussed proposed expansion of the Study, and reviewed development of the Study sampling design.
Media coverage from April to June 2004 included:
- “U.S. Plans Study on Environment and Kids,” by Lauran Neergaard, Associated Press, April 5, 2004. This article appeared in more than 200 outlets with an estimated total audience of more than 4.7 million.
- “Following Children to Identify Health Risks; Study Will Examine Genes, Environment,” by Rob Stein, Washington Post, April 27, 2004. This article appeared in 10 other outlets with an estimated total audience of more than 1.8 million.
- Articles featured in 12 publications with an estimated total audience of 8 million:
- American Medical News
- Chemical Policy Report
- Waco-Tribune Herald
- Education Week
- Clean Air Report
- California Educator
- St. Paul Pioneer Press
- E/The Environmental Magazine
- The Environmental Forum
- Chemical and Engineering News.
New Program Office staff members are:
- Alan R. Fleischman, M.D. (Einstein College of Medicine)
- Pediatrics, Johns Hopkins Hospital
- Neonatology, NIH, Oxford University
- The New York Academy of Medicine, Senior Vice President
- Ethics Advisor, National Children’s Study Program Office
- Cynthia Moore, M.D. (University of Tennessee), Ph.D. (genetics, University of Indiana)
- Pediatrics, University of Tennessee
- CDC, Center for Birth Defects, team leader pediatric genetics
- Program Office part-time detail
- Melanie Martinez, M.P.A. (New Mexico State University), presidential management intern (NIH)
- Health and education policy
- Program Office.
Hiring various information technology (IT) and informatics development experts is the next major need in staffing the Program Office.
Additional Program Office help is being provided by:
- Robin Toblin: Ph.D. candidate (behavioral science), research fellow; self injury hypotheses
- Mark Richardson: Emerging Leader Program in July, Federal Consortium
- Alan Simon, M.D.: pediatric resident, cost-benefit analysis
- Ruth Morley, Ph.D. (epidemiology): perinatal specimen collection, nutrition/growth hypotheses
- Ann Louise Ponsonby, Ph.D. (epidemiology): asthma hypotheses, study design
- Sophie Bous: volunteer summer student, international/national cohort studies
- Linda Maytan, D.D.S.: oral facial related hypotheses.
Dr. Scheidt summarized protocol development as follows:
- Protocol development has been increasingly content specific with Program Office assigned staff.
- Workshop and Working Group products are being incorporated.
- Data collection intervals and timelines are becoming more specific.
- The sampling approach is critical.
- A more detailed report will be prepared after the sampling approach has been determined.
Dr. Scheidt summarized development of the technical support contract (Battelle) as follows:
- Background for sampling strategies: completed
- Developmental measures for the Study: 75 percent complete
- Sample protocol for the Practice-Based Research Network (PBRN) feasibility study: completed
- Power analyses for social environment hypotheses: completed
- Review of extant databases for the Study: near completion
- Review and white paper on measurement of housing quality: underway
- Technical analysis for sampling questions: reported today
- Review of health services measures: hold for funds transfer
- Revise and update cost projections: proposed
- Estimated economic benefit of Study: completed.
According to Dr. Scheidt, designing the Study information management system (IMS) to support the goals of the Study and address stakeholder needs will occur from two perspectives: the Study perspective, using a study process model, and the system perspective, using a requirements model. The architectural design will depend on, in part, answers to the following questions:
- What are the goals of the Study?
- What Study goals can the IMS help satisfy through automation?
A consortium of contractors—Booz Allen Hamilton Inc. (lead), RTI, Battelle, Levine Fricke, and Sanz—is developing the IT system for the Study. Aspects of the system include:
- Defining IT requirements for recruitment and enrollment
- Identifying mechanisms for questionnaire data collection
- Identifying mechanisms for electronic capture of clinical data.
Dr. Scheidt listed several areas for anticipated procurements:
- Clinical/data coordinating center
- Sources sought for small business
- Request for proposal (RFP) in preparation
- Initial (vanguard) centers
- Sources sought for small business
- RFP in preparation
- Sources-sought announcement and responses received.
Implications for the “sources sought” announcements include:
- Federal procurement for small business preference required if multiple sources are capable
- Administrative decision that regional sites are necessary for community and regional support
- Range of roles for coordinating and regional centers, for example:
- Central sampling and recruitment—follow-up at regional sites
- Central sampling with regional cluster assigned to sites
- Central data collection—site-based sample and follow-up
- No exclusive single contract organization to carry out sampling, recruitment, and all data collection
- Regardless of sampling decision, if regional sites are an option and the Study is to begin in 2006, sources sought are required now.
Recently held Study workshops include:
- Addressing Rural Children in the National Children’s Study, March 2, 2004
- Sampling Design, March 21–22, 2004
- Methodologies for Capturing Day-Specific Probabilities of Conception, May 17–18, 2004
- Cancer and the National Children’s Study: Opportunities and Challenges, May 20, 2004
- Measurement of Maternal and Fetal Infection and Inflammation, May 20–21, 2004
- Methods for the Assessment of Asthma-Related Health Outcomes, May 27–28, 2004
- Gene-Environment Interaction and the Regulation of Behavior, June 2–3, 2004
- Measuring Racial/Ethnic Disparities and Racism from a Developmental Perspective, June 21–22, 2004.
Upcoming Study workshops include:
- Body Composition Measurement, October 7–8, 2004
- Ascertainment and Diagnosis of Birth Defects (in utero and postnatally)
- Measures of Neurodevelopment and Environmental Exposures
- Assessing Dietary Intakes and Patterns in Women and Young Children: Methodological Issues with Implications for the Design
- Identification of Measures for Health Care Processes and Outcomes.
Dates have not yet been determined for the latter four workshops.
Updates on Study partnership development include:
- Foundation for NIH
- Memorandum of understanding is in final legal review.
- The Program Office cosponsored a Congressional briefing with the American Chemistry Council and CEHN on May 17.
- Federal Consortium of the Study will meet in September 2004.
Dr. Scheidt described some aspects of a proposed expansion of the Study involving the National Human Genome Research Institute (NHGRI). A NHGRI working group is engaged in planning and developing the proposal.
- The current proposal includes:
- A cohort size of approximately 500,000
- Representative sample across all ages, geography, and socioeconomic status
- Small overlap with the Study
- Publications to date include:
- The Scientist, May 26, 2004
- Nature (429;2004:475).
Dr. Scheidt said that Alan Guttmacher, M.D., Deputy Director of the NHGRI, would describe details of the proposed expansion to the NCSAC in a subsequent presentation.
Dr. Scheidt noted that the Study funding status has not changed since the last NCSAC meeting:
- Startup costs for fiscal year 2005 have an estimated need of $26 million.
- The implementation costs for fiscal years 2006–2030 are estimated to be $2.7 billion over 24 years.
In a question-and-answer session, Dr. Scheidt addressed the following issues and concerns:
- The RFP process. Dr. Scheidt said that Study RFPs would follow standard operating procedures for large government procurements, which include following NIH guidance for contracts and announcements in Federal Business Opportunities (FedBizOpps.gov), The Federal Register, the Study Assembly listserv, and the Study Web site.
- Review process. Responses to RFPs will be peer-reviewed in a manner analogous to that used for the Women’s Health Initiative, in which groups and individuals from one part of the country review proposals from other parts of the country. This process is different than that used for grant reviews.
- Role of the NCSAC. During the procurement phase, the NCSAC will function in a manner similar to that of NIH council reviews and other review panels.
- Publication of Study workshop proceedings. Philip J. Landrigan, M.D., Mount Sinai School of Medicine, said he was impressed with the number of recent Study workshops that were held and suggested that a series of workshop papers be submitted to Environmental Health Perspectives for publication. Dr. Scheidt explained that such publications are either being planned or in process. Reports for all workshops appear on the Study Web site.
- Coordinating center for vanguard sites. A range of possibilities for managing vanguard sites is being considered, including management outside the federal government. Training and uniformity in procedures will be critical for a strong, consistent Study protocol. Independence of vanguard centers and supplementation of the protocol have yet to be defined. Dr. Landrigan commented that keeping Study management in central government control could help to ensure long-term stability.
- Recruitment and data collection. Boundaries have not been delineated and decisions have not yet been made regarding the details of recruitment and data collection.
- Draft Study protocol. The draft protocol will be made public with the announcements for Study RFPs. There will be no specific peer review of the draft protocol, and it will not be reviewed by the NCSAC as a group. Individual NCSAC members may comment, but there will be no collective response from the NCSAC. It was noted that the Study protocol will be “a living, breathing” document that will continue to evolve as the Study progresses over the next 25 years or so.
- Community involvement and engagement. Community involvement and engagement will be required by vanguard centers and will be an essential part of the Study process. It was noted that most potential vanguard centers presumably have established community relationships.
- Ethical issues. Several NCSAC members suggested to Dr. Mattison that the committee should include an ethicist. Dr. Mattison explained that Dr. Fleischman will provide input on ethical issues to the Program Office and that there is an Ethics Working Group, but he agreed that the NCSAC should include an ethicist as it addresses the next stages of the Study.
Overview of Sampling Discussion
Drs. Mattison and Scheidt
Dr. Mattison acknowledged the guidance, hard work, and dedication of James J. Quackenboss, M.S., Office of Research and Development, EPA, who has been working for more than a year in planning for and facilitating analyses and discussions of the Study sampling design. Mr. Quackenboss was instrumental in planning and managing the March 21–22, 2004, Sampling Design Workshop.
Dr. Scheidt provided a chronological overview of the sampling design process:
- April 2002: Proposal and concept for analysis of sampling strategies to NCSAC; CDC, ASPE, Westat study commissioned
- October 2002: Westat report compares three models: household, office-based, center-based
- December 2002: NCSAC advises probability-based sample as feasible
- Spring 2003: Interagency Coordinating Committee (ICC) questions feasibility and cost of probability-based representative sample
- June 2003: Proposal to NCSAC for expert sampling design panel informed by commissioned papers and reviews
- Summer–fall 2003: Planning committee (composed of individuals from the NCSAC, ICC, and Program Office) commissions Battelle papers and establishes expert sampling design panel (with David A. Savitz, Ph.D., as chair)
- March 2004: Sampling Design Workshop Panel deliberates sampling strategy and proposes probability option and alternative center-based option
- Spring 2004: Program Office explores mechanisms, feasibility, and costs and commissions analyses to answer additional cost and feasibility questions
- June 28–29: NCSAC and ICC deliberate sampling options with feasibility and cost considerations.
Overview of Sampling Process
Brook Rolter, Sampling Session Facilitator
After an introduction by Dr. Mattison, Mr. Rolter defined his role as facilitator. The purpose of the sampling session facilitator was to help the NCSAC have a structured conversation in an effort to achieve its goal of finding consensus in answering the six “charge questions” (see Appendix A). Mr. Rolter explained that the information presented during the meeting’s first day would create the foundation for later discussion. To foster this discussion, Mr. Rolter would help to clarify the information presented. Mr. Rolter noted that it was important for the NCSAC to describe its expectations before answering the six charge questions. He described the “rules” of the road for facilitated discussions and presented examples of positive language. Mr. Rolter said that the decision-making continuum ranges from autocratic, autocratic with input, majority rule vote, consensus, to unanimity. He listed the following characteristics of a consensus:
- Each person feels that he or she has been heard, listened to, and understood by others.
- Each person feels that he or she can explain the decision rationale and support why the decision or solution is best.
- Each person must be willing to commit to his or her role in executing and supporting the decision.
- In those instances in which a strong, dissenting, minority opinion exists, explaining the rationale and supporting the decision includes expressing and acknowledging the dissenting perspective.
A consensus decision is:
- A proposal acceptable enough that all members can support it
- A commitment to the decision reached.
A consensus decision is not necessarily:
- A unanimous vote
- A majority vote
- Representative of everyone’s first priority.
Mr. Rolter encouraged NCSAC members to complete the survey developed by Robert T. Michael, Ph.D., University of Chicago. The written results of the survey (titled “Expressing Your Priorities for the NCS”) would be compiled for presentation and discussion later in the meeting. Mr. Rolter later used the survey results in discussing the NCSAC’s answers to the charge questions.
Summary of Sampling Design Workshop
David Savitz, Ph.D., University of North Carolina
Dr. Mattison introduced Dr. Savitz and acknowledged his successful efforts as chair of the Sampling Design Workshop Panel. Dr. Savitz thanked Dr. Mattison and all involved in the Study for the opportunity to chair the Panel. Dr. Savitz noted that the Panel members are individuals who are normally not easily engaged in such efforts. However, they all responded positively to the opportunity to participate in the workshop and worked diligently to find a consensus on the sampling design.
Dr. Savitz said that the Panel’s report was carefully crafted and that it accurately conveys the viewpoints of the Panel members. The Panel suggested a number of pilot studies and delineated an array of sampling issues. In its report, the Panel attempted to provide reasonable responses without offering too much advice. Dr. Savitz said that at this time, the Panel’s most valuable role was to offer its advice and then “disappear,” letting the decision makers do their job in the next phase of the process. He noted that a long planning process for a large study runs the risk of “using up the goodwill of and credibility with the research community.” Research, by nature, is a complicated process, and it is important to balance ambitious endeavors with feasibility. Dr. Savitz said that different issues have different importance for stakeholders. However, at some point, the planners must let go and let the implementation process move forward.
According to Dr. Savitz, the Panel did not have a preconceived idea of favoring probability-based sampling, but the cumulative weight of advantages of such an approach for the Study began to build consensus. While assessing the merits of various designs, the Panel considered two particular sampling models: probability-based surveys and biomedical recruitment. In considering the probability-based survey model, the Panel questioned how far this approach could be pushed to penetrate into the biomedical or clinical arena. In considering the research medical center recruitment model, the Panel acknowledged this approach’s success and feasibility in gathering biomedically intensive data but questioned how much further it could be pushed to do more than what it already does. In other words, could biomedically intensive studies be more truly representative or more comprehensive? In contrast, how far could probability-based sampling be pushed into clinical settings? How far could each of the models break out of the basic characteristics of their structured approaches? These are the questions that the Panel considered during its deliberations and answered in its report.
In a question-and-answer session, Dr. Savitz addressed the following issues and concerns:
- Retention. Grace LeMasters, Ph.D., University of Cincinnati, commented on what she perceived as differences in the report’s tone when comparing probability-based sampling with a center-based approach; that is, the Panel put a positive “spin” on probability-based sampling and somewhat negative spin on the center-based approaches. Dr. LeMasters noted that because Study subjects might be more inclined to identify with a particular organization, clinical center, or institute, long-term commitment to a government study might be difficult to achieve with a probability-based approach. Dr. Savitz said that the Panel recognized that long-term commitment and retention in any study are always problematic and that there may be a need for organizational identity. However, the Panel did not perceive this as an overly important issue because ultimately, in any community-based study, subjects and their families have relationships with people (doctors, nurses, interviewers, etc.), not institutions. Robert T. Michael, Ph.D., University of Chicago, described the general view that no particular sampling approach is more advantageous for retention because retention is unrelated to how a sample is drawn. Therefore, community involvement and individual relationships may be important factors for Study retention.
- Piloting of issues. Dr. LeMasters asked whether there would be any piloting regarding the various sampling design issues. Dr. Savitz explained that without piloting, researchers must rely on intuition and past experiences. There may be “gray areas” with certain research approaches, but piloting may not change the decisions that have been made. It is important to accept the gray areas, make good judgments, deal with the challenges, and move forward.
- Ongoing pilot studies. Dr. Landrigan acknowledged the strengths of the population-based model: its generalizability, its attraction to social scientists, and the ability to compare data among probability-based studies. Yet, Dr. Landrigan continues to be attracted to the medical center-based model, and he described ongoing pilot studies at three children’s environmental health research and disease prevention centers. These centers began assessing prospective cohorts in 1998–1999 as precursors to the Study. Approximately 20–25 percent of the subjects dropped out over the first year of these studies, but the study population then stabilized. The retention rate after 5 years is approximately 75 percent, with an annual attrition rate of 1 percent. The cohort is composed mostly of lower income, minority, inner-city residents. Dr. Landrigan said that reconciling the strengths of a probability-based approach with those of the medical center approach is highly commendable and may be an ideal model for the Study.
- Political issues. Dr. Landrigan briefly discussed some to the political issues regarding support and funding if major academic health centers were excluded in a probability-based national approach.
- Recruitment and follow-up. Dr. Savitz said that the Sampling Design Workshop Panel recognized that regardless of the recruitment approach, the Study would eventually have to cope with the challenges of a nationally mobile cohort, with follow-up occurring in places other than the original place of recruitment. Lucina Suarez, Ph.D., Texas Department of Health, asked whether the medical center model would have an advantage in recruitment and whether either model had an advantage in recruiting hard-to-reach populations. Dr. Savitz replied that medical centers are good at recruiting subjects that are already in the medical system, such as mothers seeking prenatal care, but may not be good at recruiting hard-to-reach populations. The Panel recognized the vital importance for the Study to include high-risk or hard-to-reach populations. Dr. Savitz said that probability-based sampling might be the best way to engage all groups of people.
- Tradeoffs of approaches. Dr. Savitz noted that proponents of the need for intensive biomedical assessments of Study subjects ultimately have to accept the high costs of such an approach. The proponents of the probability-based model firmly support the necessity of their approach to determine relationships of exposures and population-based health outcomes. These two models elicit strong and conflicting viewpoints. Unfortunately, there is no easy way to combine the best of both models. According to Dr. Savitz, the solution is to choose one of the approaches—take a base of what is certain—and creatively push the boundaries to the middle in an effort to minimize the tradeoffs. In other words, researchers should attempt to realize how medically sophisticated a probability sample can become, or conversely, how population-based a medical center sample can become. Dr. Michael stated that the application of these approaches is not an “either/or” situation. There are three important issues for either model: (1) selection of the case (that is, how a child gets enrolled in the Study), (2) the bond that the subject forms with the Study (which enhances retention), and (3) the quality of the data that are collected. The success of the Study will ultimately depend of the success of these three issues.
- Systematic approach. Judith A. Graham, Ph.D., American Chemistry Council, asked whether it would be feasible to have medical centers within certain geographical areas compete for Study participation with the stipulation that they implement probabilistic sampling within their areas. Dr. Savitz answered that at this time, the vision for Study sampling is a cluster probability sample, with the clusters not randomly distributed. To this end, the Study would need a systematic approach to integrating the medical facilities within the designated geographic areas. The challenge for a systematic approach is the ability to ensure a standardized protocol across facilities, that is, making a disorganized health care system be organized.
Sampling: Thoughts of the Program Office
Ruth Brenner, M.D., National Children’s Study Program Office, NICHD, NIH, DHHS
Dr. Brenner acknowledged the source documents and individual input for her presentation:
- Westat report (October 2002)
- Battelle report (February 2004)
- Sampling Design Workshop (March 2004)
- Comments from various Working Groups (June 2004), including Community Outreach and Communication, Study Design, and Fertility and Early Pregnancy Working Groups
- Second set of Battelle reports (June 2004)
- Ongoing discussions with the NCSAC, ICC, and Program Office staff.
Dr. Brenner described the goals of the Study, as derived from Children’s Health Act of 2000:
- Study environmental influences on child health and development
- Prospective cohort study
- Chronic and intermittent exposures
- Investigate basic mechanisms of developmental disorders and the environmental risk factors and protective factors that influence health and developmental processes
- Diverse populations of children
- Consideration of prenatal exposures.
Dr. Brenner characterized measurement requirements, as based on Study priority areas and core hypotheses:
- Some measurements will need to reflect exposures that occur prior to and around the time of conception and very early in pregnancy.
- There is a need for both community- and individual-level measurements.
- Some measurements/samples (for example, cord blood) will need to be obtained at hospitals or clinical centers; others will need to be obtained in the home and other environments to which the child is regularly exposed.
Measurements early in pregnancy are important because:
- Organogenesis, particularly neurodevelopment, occurs very early in pregnancy. Many hypotheses involve examination of very early exposures.
- Many of the exposures are transient and are important to measure during early “critical windows.”
Dr. Brenner presented a slide depicting the critical periods in human development according to Moore and Persaud, The Developing Embryo, 7th edition, 2003.
Community measures are important because a number of hypotheses try to untangle community versus individual effects of various exposures, particularly in the social and physical environments. Obtaining community measures such as characteristics of a school may be difficult if the sample is not highly clustered.
Specific examples of hypotheses to be addressed by community measures include:
- Increased collective efficacy is negatively associated with aggression and intentional injuries.
- Characteristics of the built environment, neighborhood design, and proximity of parks/playgrounds are associated with obesity and injuries.
Measures at various sites include:
- Hospitals/Study centers
- Specimens/tests: cord blood, placentas, three-dimensional ultrasounds
- Battery of examinations/tests at one time
- Environmental measures—environmental samples and observations
- Many measurements could be done either in a center or at home
- Homes may be more convenient for some families
- Others prefer to go to somewhere outside of home
- Challenge of distractions and standardization.
Multiple levels of measurement (adapted From: Lynch JW. Australas Epidemiol. 2000;7:7–15) from preconception to 21 years of age include:
- Community-level exposures
- Organizational exposures
- Social environments
- Individual characteristics
- Household environment
- Genetic characteristics
Dr. Brenner characterized the Study’s organizational structure:
- Although there is overlap between the organizational structure and the sampling design, both must be considered independently.
- The Study will involve the federal government, academic/regional centers, a central coordinating center, and possibly others.
- The Study will use a contracting mechanism with both scientific and administrative oversight from the federal government.
- The Study will include important roles for both a central coordinating center and academic/regional centers.
Advantages of an inclusive organizational structure include:
- Cumulative scientific expertise offered by inclusion of individual Study centers and coordinating center
- Coordinating center
- Centralized coordination of many of the aspects of the Study
- Tracking of data and specimens
- Recruitment and retention
- Collection of some of the data in the home
- Study centers
- Collection of clinical specimens/measurements (cord blood, placentas, newborn exam, three-dimensional ultrasounds, bone density, pulmonary function, and others)
- Local knowledge of and previous experience with the communities beneficial for recruitment and retention.
Dr. Brenner described the role of the sampling strategy and listed some of the major issues. The sampling strategy should allow the Study to:
- Facilitate collection of high-quality data to minimize measurement biases
- Maximize validity of inferences concerning exposure/outcome relationships
- Response rates
- Retention rates
- Selection biases
- Capture the diversity of the U.S. population and include a range and diversity of exposures and outcomes in United States
- Extrapolate results to all U.S. children.
Major sampling design issues include:
- Identification and selection of geographic locations
- Primary sampling units (PSUs)
- Identification and selection of individuals within the geographic area
- Timing of enrollment.
Dr. Brenner presented four Study sampling design options:
- Option 1: Nationally representative probability sample
- PSUs probability-based
- Sampling of clusters probability-based
- Option 2: Modified probability sample
- PSUs and clusters probability-based
- PSUs and clusters identified with input from the local centers
- Option 3: Modified probability sample
- PSUs not probability-based
- Clusters identified with input from centers
- Option 4: Medical center/patient-based sampling approach.
Dr. Brenner listed the details, advantages, and concerns for option 1:
- Coordinating center/statistical team would select PSUs/clusters (counties/tracts/block groups/blocks) with known probabilities; boundaries are based on Census data.
- Enrollment of Study participants could be done by either coordinating center or local center
- Work with local clinical sites to obtain clinical measurements.
- Would likely provide national estimates
- Has the capability of fulfilling the need to reflect the diversity of the United States, however defined.
- Response rates may be unacceptably low in some clusters
- Not clear whether it is feasible to assign clusters to a center with no input from the center or from the cluster
- Boundaries for the designated clusters may not conform to recognized communities or neighborhoods
- Participants’ commitment to the Study in the absence of a community commitment.
Dr. Brenner listed the details, advantages, and concerns for option 2:
- A priori designation of broad geographic areas
- Select Study centers based on ability to conduct the Study, including ability to recruit from one of the designated geographic areas
- Coordinating center/statistical team would work with the Study centers to define and then select PSUs and clusters within the primary broad geographic area
- Selection would likely be weighted toward those in proximity to the Study center, but no area has a “0” probability of selection
- These clusters could correspond to recognized communities/neighborhoods rather than Census blocks.
- Potential to provide national estimates
- No area would have a 0 probability of selection.
- Likely fulfills the need to reflect the diversity of the United States
- Allows input from the Study centers in identification of communities for potential selection, which may increase response rates.
- Coverage of a geographic area would still lead to the need to involve multiple sites of health care
- Some sites would likely be geographically distant from identified Study center.
Dr. Brenner listed the details, advantages, and concerns for option 3:
- No a priori selection of broad geographic areas.
- Award centers based on factors related to their response to the statement of work.
- Once awarded, work with centers to select neighborhoods to approach
- Define geographic source population for sample near the center
- Select clusters with input from center and known probability.
- Allows local input from the Study centers in selection of communities, which may increase response rates
- Allows selection based on ability to conduct the Study, which may increase the likelihood of successful conduct of the Study
- Could reflect the diversity of the United States, especially if specified in selection criteria
- Individuals within the cluster could be selected to be representative of selected communities.
- No ability to make national estimates
- Reflection of diversity of United States not guaranteed.
Dr. Brenner listed the details, advantages, and concerns for option 4:
- No a priori selection of broad geographic areas
- Award centers based on other factors related to their response to the statement of work
- Clusters are defined as the population that accesses facilities of enrolled centers.
- Allows selection of the “best” offerors, which may increase the likelihood of successful completion of the Study
- Could reflect the diversity of the United States, especially if specified in selection criteria
- Allows input from the Study centers in selection of participants, which may increase response rates
- Facilitates collection of clinical data and specimens.
- No ability to make national estimates—external validity
- May represent convenience sample of women receiving prenatal care—potential loss of internal validity
- Reflection of diversity of United States not guaranteed.
Approaches to Selecting Participants
There are three approaches to selecting individual participants within clusters:
- Clinic/physician office-based
- Community—defined geographic area
- Clinic/physician office/prenatal care
- Allows for self-enrollment as an additional component.
Dr. Brenner listed the advantages and concerns for the household approach:
- Facilitates preconception enrollment
- Includes those outside the medical system.
- Will need to screen a large number of homes
- Balance of cost and efficiency in screening and following all age-eligible women versus more select groups (oversampling those planning to become pregnant)
- Ability to interface with the medical community for samples that must be collected in that setting (three-dimensional ultrasounds, cord blood, placentas) is unproven.
Dr. Brenner listed the advantages and concerns for the clinic/physician office/prenatal care approach:
- Efficiency—fewer contacts per live birth
- Linkage to the medical system for collection of certain data elements (tests, specimens, and medical records).
- Miss those with no prenatal care (2 percent)
- Miss objective measurements of early exposures (15 percent).
- To capture the diversity of the United States, would need to recruit in sites that represent the sites in which women receive care. The feasibility of working with large numbers of providers is uncertain.
Dr. Brenner characterized the community recruitment approach:
- Attempt to get all or a defined sample of pregnancies in a defined geographic area
- Combine multiple methods of recruitment:
- Screening of households with targeted follow-up (at risk of pregnancy or intending pregnancy)
- Clinic/prenatal care
- Allows for self-enrollment as an additional component
- Recruit at birth to allow representation of groups missed by above strategies
- Ongoing comparisons of characteristics of participants enrolled in the Study with all live births in that area
- The challenge of this approach is that it would require strict oversight, probably from the data center, to make certain the captured population reflects the base population.
Dr. Brenner listed the possible times for enrollment:
- Early in pregnancy (less than 4 weeks)
- Prenatal care
- At delivery
- Combination of above.
The possibilities for preconception enrollment include:
- Follow all women of childbearing age in defined geographic area
- Other approaches
- Women of childbearing age within restricted age range
- Women at risk of pregnancy
- Couples planning to become pregnant
- Enrolling some, but not all of the cohort preconception
- Enrolling subsequent pregnancies among women enrolled in the Study.
Other concerns for preconception enrollment include:
- Difficult to obtain measurements around the time of conception
- Give all female participants home pregnancy tests (two) for
- Pregnancy planners
- Baseline measures
- Urine samples over fertile window
- Measures when mother knows she is pregnant
- Rest of group
- Baseline measure
- Measures when mother knows she is pregnant.
Potential scientific concerns for preconception enrollment include:
- Couples trying to become pregnant may intentionally alter their exposure profile
- Women may alter exposures once they know they are pregnant. Thus, the early exposure profile of women who know that they are pregnant early would be different from women who do not realize that they are pregnant until later in pregnancy.
The challenges of preconception enrollment include:
- The sampling option that best fulfills one Study need may be the worst option to fulfill a different Study need; for example:
- Hospital-based sampling is most efficient for collection of cord bloods and placentas.
- Household sampling prior to presentation for pregnancy is best for preconception measures.
- The community clusters approach would likely involve deliveries at multiple hospitals.
Dr. Brenner summarized the Programs Office’s current thoughts on sampling as:
- Cluster-based sampling
- Range of options for selection of PSUs, clusters, and individuals within clusters
- Each option has its own set of advantages and disadvantages
- Some options provide a better opportunity to meet the needs of the Study
- To capture early transient exposures, there will need to be some participants enrolled prior to conception.
In a question-and-answer session, Dr. Brenner addressed the following issues and concerns:
- Political considerations. Willa M. Doswell, R.N., Ph.D., University of Pittsburgh School of Nursing, asked whether the Program Office has addressed the political considerations of the different sampling approaches. In response, Dr. Scheidt said that potential Study centers would not be involved in the procurement process, that centers will be chosen based on demonstrated excellence and the ability to perform aspects of the Study, and that awards will be based on an open, competitive process. Dr. Landrigan commented on whether Study centers must be in the region that they are studying.
- Key exposures. Dr. LeMasters asked to what extent the sampling approaches would include key exposures. Dr. Brenner replied that the Study intends to stratify on many variables, not just geography. Dr. LeMasters asked whether there was any progress in determining the five most important exposures and asked specifically on what exposures the Study will stratify. Dr. Mattison emphasized that there is great consideration of exposures among the NCSAC, ICC, and Program Office.
- Preconception enrollment. Dr. Landrigan suggested that prepregnancy recruitment might skew the representativeness of the Study because the preconception cohort would be a very select group.
Technical Background Relating to Sampling Issues
Warren Strauss, Sc.M., Battelle
Mr. Strauss discussed the status of research on various sampling questions for the Study. At this time, additional sampling design questions being researched by Battelle include:
- Sample size of women needed
- Alternative sampling frames for recruitment of women
- Recruitment cost estimates for various approaches
- Comparison of highly clustered designs with geographically dispersed designs
- Effects of mobility on Study cohort over time as a function of the number of clusters
- Exploration and characterization of validation samples
- Justification for preconception and early pregnancy measures.
Approaches to Sample Size of Women
Mr. Strauss characterized the approaches to Study sample size as follows:
- Three recruitment approaches:
- Physician’s office
- Full community
- Four pregnancy/contraception status categories of women:
- Currently pregnant (CP)
- Currently seeking pregnancy (CSP)
- Current contraception users (CCUs)
- Other (non-CCU, non-CSP)
- Two target enrolled populations:
- Two levels of average initial recruitment rates: 80 and 60 percent
- Three recruitment periods: 3, 4 and 5 years
- Two levels of monthly attrition rates:
- CP/CSP = .0015
- CCU/Other = .0035
- Other rates:
- 94 percent live birth enrollment
- 95 percent contact
- 10 percent vacancy.
Mr. Strauss presented a pregnancy transition model that depicted the complex interactions among CP, CSP, CCUs, and Other. The key element of this model was the constantly changing status of women. In discussing the results of a 4-year recruitment period, Mr. Strauss compared the three recruitment approaches with regard to:
- Pregnancy status of women targeted
- Number of households/women initially targeted
- Number of women contacted
- Number of women enrolled and followed.
Mr. Strauss then compared the number of women and resulting number of live births available for preconception measures by sampling approach.
The key findings were:
- Targeting CP/CSP women only results in larger numbers of women contacted but 75 percent fewer enrolled and followed.
- Use of 60 percent initial enrollment rate raises estimated numbers of women by approximately one third.
- Decreasing live birth enrollment rate from 94 to 80 percent increases estimates by 20 percent.
- Full community approach results are similar to household approach results.
- A 3-year recruitment period requires more women (7–35 percent).
- A 5-year recruitment period requires fewer women (4–19 percent).
Methods for Efficiently Identifying Age-Eligible (15–44 years) Women to Enroll in the Study
One efficient method would be to combine multiple datasets that are likely to contain current names and addresses of age-eligible women, including:
- Drivers license/state ID
- IRS or Social Security Administration data
- Medicaid records
- Local school records
- Recent birth records.
The advantages of this method are:
- Combining data from multiple sources reduces likelihood of excluding any particular subpopulation.
- When used in conjunction with a household sampling approach, this method could lead to an approximate 43 percent reduction in the number of households to contact.
The limitations of this approach are:
- Gaining access to these data could prove to be difficult, costly, and time-consuming.
- Merging records from different sources could be challenging and error prone; determining which records are most recent could also be problematic.
- These two factors could erode the 43 percent gain in efficiency.
- This method could exclude women who recently moved unless the Study tracks them down.
Mr. Strauss listed methods for efficiently identifying women at high risk of pregnancy:
- Capitalize on recent birth records: Women who recently gave birth to their first child have a 44 percent chance of being pregnant when contacted or of conceiving a child during a 4-year follow-up period.
- These women are more likely to seek pregnancy and give birth.
- Yield: Over a 4-year follow-up period, there would be one birth in 2.3 women (compared with one birth in 4.3 women in general population of age-eligible women).
- Access to birth records
- Introduces potential biases into the Study
- Could be mitigated by using this strategy for a portion of recruitment strategy (with another component of recruitment enrolling only women with no previous children).
The following eight recruitment approaches were explored:
- National probability-based sample (NPBS) of households in 250 PSUs
- Enroll all women (1)
- Enroll CP/CSP women only (2)
- Full community approach (25 regions, four communities each)
- Enroll all women (3)
- Enroll CP/CSP women only (4)
- Physician’s office approach (50 and 100 PSUs)
- Enroll all women: 80,000 CP; 10,000 CSP; 10,000 CCU/Other
- 50 PSUs (5)
- 100 PSUs (6)
- Enroll CP/CSP women only: 80,000 CP; 20,000 CSP
- 50 PSUs (7)
- 100 PSUs (8).
The following cost model assumptions were made:
- A 4-year recruitment period with all births in the 48-month window from months 4–51
- A 12-month enrollment period except for CP women in physician’s office approach
- An 80 percent average initial enrollment rate
- A 95 percent contact rate requires six visits in household and physician office approaches, four visits in full community approach
- Constant hourly labor rates: $125 for physicians, $100 for principal investigators, $40 for senior staff members, $25 for junior staff members, $20 for recruiters.
Recruitment cost categories include:
- Initial enrollment
- Labor cost to contact all targeted households/women
- Travel expenses for recruiting trips
- 100 miles per trip in household and physician’s office approaches
- 50 miles per trip in full community approach
- Labor cost for periodic check-in with enrolled women
- Check-in with CCU/Other women once every 6 months
- Check-in with pregnant women three times during pregnancy
- Pregnancy test kits for women seeking pregnancy
- Other operating costs
- Labor cost of principal investigator, Study coordinator, and physicians
- Advertising expenses
- $25,000 for household and physician’s office models
- $100,000 for full-community models.
Initial cost estimates for the eight previously described recruitment approaches are:
Mr. Strauss characterized this task as follows:
- Investigated designs with:
- Number of subjects per cluster
- Focus on relationships between exposure (X) and health outcome (Y)
- Investigated models for both continuous and binary responses (Y)
- Results presented as a function of
- ρY|X = within-cluster correlation in Y after adjusting for X
- λ = within-cluster correlation in X.
Clustering results for continuous Y and X are:
- Relative design effects and power compared with a 250-cluster design
- Design effects assess size of variance for the relationship between outcome and exposure (slope estimate, ß Y|X)
- Power relative to the specific slope estimate when power of a 250-cluster design reaches 80 percent
- Results shown for ρY|X = 0.01
- Allow λ to range from 0 to 1
- Similar results hold for other values of ρY|X
- Most cases, relative design effects are same or smaller.
Mr. Strauss showed graphs depicting distributions by number of clusters for:
- Relative design effects for continuous Y and X
- Power for continuous Y and X.
Mr. Strauss presented results for binary Y and X (logistic regression model), presenting odds ratios detectable with 80 percent power.
The major conclusions for clustering are:
- If focus of analysis is estimating relationship between outcomes (Y) and exposures (X) and
λ <1 (focus on individual-level exposures that vary within cluster)
- There is surprisingly little loss in statistical efficiency when moving from a design with lots of clusters (250) to a design with few clusters (50)
- This does not hold for community-level exposures
- When λ = 1, designs with more clusters have more power
- However, it is very informative to look at community-level exposures based on the percentage of between-cluster variability in outcomes (Y) that it is able to explain.
Questions concerning cohort mobility include:
- Where, when, and how much will the cohort move?
- Is mobility mitigated by selecting larger number of clusters initially?
- How many subjects will need to be “covered” by “new” data collection units?
Mr. Strauss provided an example of cohort mobility showing dispersion at 1, 5, 10, and 15 years.
Mr. Strauss explained that estimating cohort mobility began with the following activities:
- Projected cohort mobility for 50 realizations of a geographically clustered design with counties selected probabilistically from 8 strata
- Evaluated designs that start with selection of largest county in each state and then select next largest counties across all states
- Evaluated designs with different numbers of counties selected initially (50, 100, 250)
Mr. Strauss presented projections on cohort mobility based on the number of original counties. These projections included:
- Number of subjects living in original counties
- Number of subjects in counties within 25 miles of original counties
- Number of subjects residing in counties
- 25–50 miles from original counties
- 50–100 miles from original counties
- More than 100 miles from original counties
- Number of counties more than 100 miles and with more than 25 subjects.
Mr. Strauss provided the following conclusions about cohort mobility:
- Procedures for tracking and following movers needed
- Significant portion of cohort will move over the course of the Study
- Significant portion will move out of the original counties
- Might be more cost-efficient to begin with fewer counties and establish mobile collection units as needed over the course of the Study
- Further work to assist Study planners includes:
- Compute the number of additional regions that need to be “covered” in order to maintain collection for 95 percent (or greater percentage) of cohort
- Project cohort mobility once original counties are selected.
The conceptual model for sample validation has the following assumptions:
Let: Y be the health outcome of interest,
X 1 be the “gold-standard” measure of exposure, and
X 2 be a less precise measure of exposure.
X 1 is measured on a small subset of the cohort, whereas Y and X 2 are measured on the entire cohort. The idea is to leverage the information contained in the small validation sample that contains X 1 to draw inferences on the effect of “true” exposure (X 1) on outcome (Y).
There are three general methods for selecting the subset of Study participants that are in the validation sample (that is, have X 1):
- Outcome-dependent sampling (depending on Y)
- Covariate-dependent sampling (depending on X 2)
- Random sampling (no information).
Using validation samples in the Study will:
- Provide a statistical basis to correct for bias and error in exposure assessment when investigating relationships
- Allow the Study to use less detailed measures of exposure for most of the cohort while preserving the ability to assess the impact of “true” exposure on disease (assume “true” exposure can be measured on subset)
- Offer potential for:
- Substantial cost savings when detailed exposure assessment is expensive and reasonable surrogate measures can be implemented at a lower cost (for example, passive air sampler and questionnaire versus continuous/active samplers for pesticides in indoor air)
- A more feasible Study, especially when applied to preconception or periconception exposures; temporal variability/bias from early gestation to later in pregnancy is just another source of error that can be addressed using this methodology.
Results from previous work on validation samples (Battelle, Harvard, EPA) indicate that optimal designs depend on characteristics of exposure and health outcome of interest. Other general conclusions are:
- It is difficult to identify a single optimal design.
- For each hypothesis of interest, optimally designed substudies should be considered and investigated.
- Well-designed substudies can efficiently characterize exposure-health outcome relationships using a fraction of the Study cohort.
- “Detailed” exposure information for all subjects may not be necessary to characterize effect of exposure on health outcome.
Factors to consider when using validation samples include:
- Exposure period
- Variability of exposure measurements
- Exposure metric related to disease (average, acute)
- Continuous or binary outcome
- Longitudinal outcomes
- Prevalence of outcome
- Variability of outcome measurements
- Strength of exposure-outcome relationship.
Mr. Strauss listed several new questions for validation samples:
- What is the sample size of the validation sample that is necessary to support Study inferences?
- Is this a function of the size of the Study cohort (that is, a 10 percent sample)?
- Is this a function of the number of (X 1, X 2) pairs necessary to accurately establish the relationship between “true” and “surrogate” measures of exposure to use in the statistical errors-in-variables correction?
- How is the power to assess relationships between exposure and health outcome affected by the strength of the relationship between X 1 and X 2?
- How does the approach to selecting participants into the validation sample affect the previous two questions?
Mr. Strauss suggested that answers to these questions should be assessed using design effects as a function of ρX1,X2, n X1, and X 1 sampling strategy, where:
- Reference design is one for which entire cohort has X 1 measures
- Simulations correspond to a cohort size of 10,000 with a disease prevalence of 0.025.
Mr. Strauss presented two graphs that plotted design effects against portion of variability in X 2 explained by X 1. The first graph showed effects when the validation sample size is 10 percent of the cohort; the second graph showed effects when the validation sample size is 5 percent of the cohort.
The general conclusions for validation samples are:
- Size of validation sample is dictated by the strength of the relationship between X 1 and X 2
- Use of an outcome-dependent design for the selection of subjects into the validation sample leads to extremely efficient analysis; design effects below 2, even when ρX1,X2 is small (<0.3)
- Use of random sampling for the selection of subjects into the validation sample leads to slightly less efficient designs; design effects still below 2 when ρX1,X2 is reasonably large (greater than 0.5).
Justification for Periconception Measures
For current priority outcomes and “core hypotheses” (10 of 21), the case for periconception measures is based on the critical nature of the developmental period shortly after conception combined with:
- Exposures that are very transient (for example, nonpersistent pesticides)
- Biological/physical measures that are time-dependent (micronutrient levels), or
- Retrospective measures that are subject to error or recall bias.
Another key reason for periconception measures is to allow the Study to serve as a “resource for future studies” by examining critical windows of development that are difficult to study retrospectively, in particular in relation to the etiology of birth defects and other adverse child outcomes.
There are several supporting factors for using validation samples to extend power of limited preconception measures:
- Validation sampling may allow the Study to reduce the number of participants measured preconception.
- This process involves obtaining “gold standard” preconception or periconception measures on a representative subset of the cohort and less precise, less costly surrogate measures (postconception) on the entire cohort.
- By building the relationship between the gold standard and surrogate measures in the representative subset, preconception effects can be studied across the cohort using combination of gold standard and surrogate measures.
- For example, a gold standard could be measurements of a pesticide in urine within 2 weeks of conception; a surrogate could be measurement of a pesticide metabolite and questionnaire information 3 months after conception.
- Depending on prevalence of exposures and outcomes and adequacy of surrogates, validation subsampling may provide sufficient power while significantly reducing costs and burden of preconception sampling.
In a question-and-answer session, Mr. Strauss addressed the following issues and concerns:
- Live birth enrollment rates. Donald J. Dudley, M.D., University of Texas Health Sciences Center at San Antonio, noted that Mr. Strauss had used the 94 percent rate for live birth enrollment, which was derived from the Westat report. Dr. Dudley said that other research results indicate that an 80 percent enrollment weight might be more realistic.
- Other rates. Dr. LeMasters suggested that the various rates in Mr. Strauss’s report might be overly optimistic. She cited a 60 percent contact rate and a 40 percent recruitment rate for a 5-year study of a cohort of 900.
- Subject age range. Dr. Graham said that the Study might use subjects from an age range narrower than the one Mr. Strauss assumed (15–44 years).
- Preconception enrollment. Dr. Graham questioned the costs and feasibility of preconception enrollment.
- Emigration. Dr. Landrigan asked whether the Study had considered the possibility of subject moving out of the United States. Are there plans to track and continue data for these subjects? Mr. Strauss replied that there were no plans to track subjects outside the United States, but the Study may include recent immigrants.
Presentation of ICC’s Questions
Dr. Scheidt noted the following about the “charge questions” (see Appendix A):
- The questions are based on perceived challenges and difficulties with the proposed nationally representative sample.
- Realistic planning requires that answers and details be addressed.
- These are some of the questions those involved in the planning of the Study face.
- The questions may not available or even answerable.
- The Study is interested in the NCSAC’s thoughts. Possible responses include:
- Requested facts, opinions
- Additional sources process
- Feasibility study
- No answer
- The outcome is an approach, not a final decision.
- Consideration of the sampling design is an evolving process.
Dr. Scheidt listed the following “charge questions”:
- Logistical and cost feasibility of scientifically intensive measures with a probability sample
- Probability sample at what cost? (in terms of other elements)
- If preconception enrollment is too costly, what portion (if any) would suffice?
- Are there alternative (innovative) approaches to meet Study goals?
- What size and number of clusters are necessary for efficiency versus representiveness?
- What attributes are important to assure diversity?
Discussion of Sampling Strategy
Facilitated by Mr. Rolter
Mr. Rolter described the NCSAC’s discussion of the sampling strategy as a method to explore the potential ways to make a group decision. He reiterated the concepts of consensus decision-making and discussed some of the conceptions and misconceptions about this process. Basically, reaching consensus is a matter of expressing priorities. According to Mr. Rolter, the four sampling design options are on a continuum of decreasing probabilities. All options have known probabilities, and different weights can be assigned to different probabilities.
The 13 NCSAC members present at the meeting agreed to vote on preference of the four sampling options before continuing any further discussion. Two members abstained from voting, and members were allowed to vote for more than one option. The initial voting (day 1) tally was as follows:
- Option 1: 5 votes
- Option 2: 8 votes
- Option 3: 0 votes
- Option 4: 0 votes.
After the discussion, the NCSAC members voted again; the second voting (day 2) tally was as follows:
- Option 1: 6 votes
- Option 2: 7 votes
- Option 3: 0 votes
- Option 4: 0 votes.
The NCSAC members agreed that option 4 was not viable. They noted that options 1 and 2 were similar, with the primary difference being “input from local centers” in option 2 versus no mention of centers in option 1. There was concern about the definition of “input” and which entity is actually defining and choosing the sample (that is, with the PSUs not being probability-based in option 3). Option 2 allows for community input, and the PSUs can be geographically distant from the centers. The basic difference between options 1 and 2 is who gets to identify/choose the PSUs and clusters.
Dr. Scheidt explained that with either option 1 or 2, the Study centers would not be selecting the clusters. Although the centers could help identify communities and sampling units, the Study’s statistical team would determine the actual sampling.
The NCSAC agreed to the following draft consensus statement regarding the sampling options:
- Of the four sampling strategy options, the NCSAC prefers options 1 and 2, but with a preference to have the choice of samples and cases be determined nationally versus by the centers.
- Points of distinction/clarification desired between the two options are:
- What is the definition of “center input” from the centers in option 2 (the issue of “cherry picking” versus “making the study richer”)?
- Can the local centers weight the samples?
- How can the definition of “center input” be made precise enough to control or avoid the “slippery slope”?
- How will community involvement be obtained?
Additional discussion themes were:
- The NCSAC prefers adopting a sampling approach that is based on national probability core.
- A specific, prioritized set of hypotheses would help inform the discussion and be the foundation for many of the tradeoff decisions before the Study.
After the voting and subsequent discussion, the NCSAC agreed to address the six charge questions (see Appendix A) with draft consensus statements.
Charge Question 1
NCSAC members agreed that this question was substantially different than the other five questions (that is, question 1 involves an operational framework, whereas questions 2–6 involve conceptual frameworks). The NCSAC agreed upon the following draft consensus statements regarding charge question 1:
- The basic question is what can the Study find acceptable (that is, what can they “live with”) with respect to sampling design (between options 1 and 2).
- The hypotheses will determine the bottom line for sampling and integration of Study centers.
- Issues of cost, feasibility, and community engagement require further consideration.
Charge Question 2
The NCSAC agreed upon the following draft consensus statements regarding charge question 2:
- There is general agreement that some kind of nationally representative sample should be a core element of the Study.
- Where tradeoffs have to be made, they should be made around specific hypotheses but preserve the core structure of sampling.
- To operationalize, the Study must agree upon the hypotheses to be addressed and prioritize them.
Charge Question 3
The NCSAC agreed upon the following consensus statements regarding charge question 3:
- The NCSAC proposes that a proportion of couples/women be enrolled prior to conception.
- That proportion should be selected based on the hypotheses proposed by the Fertility and Early Pregnancy Working Group or some selection of those hypotheses such that the recruitment of preconception couples/women allows those hypotheses to be addressed.
- The Study should not use a “snapshot” of individuals/couples who know they are attempting pregnancy to set the proportion. The proportion recruited should be based on some consideration of the hypotheses.
Charge Question 4
The NCSAC agreed upon the following consensus statements regarding charge question 4:
- There is some degree of enthusiasm for pilot studies that would address cost or ease or improve particular assays, but information on retention rates appears to be adequate. There is not much need for additional work, if any.
- There is still some degree of enthusiasm despite the recommendation of the Sampling Design Workshop Panel to explore alternative sampling strategies, with the caveat that, given that a group of savvy statisticians and design experts have examined the issues and have not been able to propose an alternative, the yield might be low.
- The NCSAC suggests a working meeting with principal investigators who have experience with major longitudinal children’s studies or studies of pregnant women to determine and discuss lessons learned.
Charge Question 5
The NCSAC agreed upon the following consensus statements regarding charge question 5:
- Representative sampling, if done with considerable thought and effort, is consistent with clustering for measuring community attributes and consistent with cost-efficiency. This is not an “either or” situation; the Study would just need to work really, really hard to get both.
- To the extent that communities or neighborhoods can be built into the sample in a way that preserves the ability of the Study to talk to the broader nation or other studies, there is a degree of comfort in the NCSAC with the recognition that this approach allows efficiency in measuring exposures and in characterizing outcomes.
- Where sampling loses that ability to extrapolate more broadly, the NCSAC would ask that it be viewed primarily in the context of testing exposure hypotheses relationships or other special study design issues such as measure validation.
Charge Question 6
The NCSAC agreed upon the following consensus statements regarding charge question 6:
- Use a NPBS as much as possible.
- Hypotheses need to be prioritized and then used as the foundation to change the sampling as needed to accomplish the hypotheses via weighting, subsampling, oversampling, or other techniques.
- There is a need to identify how to define and determine the weighting and other techniques.
- Determine the most appropriate NPBS based on the hypotheses with input from the Study Design Working Group.
Update of NHGRI’s AGES
Alan Guttmacher, M.D., Deputy Director, National Human Genome Research Institute, NIH, DHHS
Dr. Guttmacher provided an update on the progress of AGES (American Gene Environment Study). AGES planners are developing a broad outline of the study to submit to NIH Director Elias A. Zerhouni, M.D. Planners are also estimating costs of this large prospective cohort study of gene-environment interactions. Five subgroups are currently working on this effort, which would attempt to survey a representative sample of as many as 500,000 participants from all geographic, racial, ethnic, and socioeconomic groups defined in the most recent U.S. Census. A 500,000-person cohort would generate great sampling power, but there are cost-benefit issues with a cohort of this size. The study would include genetic, phenotypic, and environmental measures and would attempt to incorporate evolving technologies.
Community involvement and consent of subgroups would be important. Dr. Guttmacher noted that the study would not be hypothesis-driven but “hypothesis-informed,” serving as a resource to address future hypotheses. NHGRI issued a request for information (RFI) in May 2004, posing 14 questions to other large cohort studies. Planners are trying to determine whether other cohorts can be incorporating into AGES and, if so, the best way to use these resources. To date, the RFI has generated more than 100 responses.
In describing the relationship between AGES and the Study, Dr. Guttmacher suggested a possible sharing of personnel. The teams for the two studies would also keep each other informed on status, progress, and issues. Dr. Guttmacher said that the Study would not incur any additional costs because of AGES. One idea that Dr. Guttmacher proposed was the sharing of a subsample of, for example, 8,000 newborns. The AGES team has not yet determined at what age the study will begin assessments in children. The AGES and Study teams need to begin discussion on areas of mutual benefit and how the two studies might work together.
In a question-and-answer session, Dr. Guttmacher addressed the following issues and concerns:
- Community engagement
- Overall goal and objectives of AGES
- Etiological associations of disease development.
“Expressing Your Priorities” Survey Results: NCSAC Study Priorities
The survey results were compiled and presented by Mr. Rolter.
Discussion of Positive Health Framework
Deborah A. Phillips, Ph.D., NCSAC Member, Georgetown University
Dr. Phillips described her presentation as an update from the Healthy Development Ad Hoc Working Group (Neal Halfon, M.D., M.P.H.; Paul Wise, M.D.; Deborah Phillips, Ph.D.). The group held a meeting in San Francisco, California, on May 4, 2004. The participants included:
- Drs. Mattison and Scheidt
- Drs. Halfon, Phillips, and Wise
- Margaret Bridges, Ph.D., University of California, Berkeley
- Frank Oberklaid, Dch, M.B., B.S., FRACP, University of Melbourne
- Michael Regalado, M.D., Cedars Sinai Medical Center
- Michael Seid, Ph.D., Rand Corp
- Melissa Wake, Director, Center for Community Child Health, Victoria, Australia.
The group also consulted with Heather Joshi, Director, British Millennium Cohort Study.
The group has not yet finished a formal set of hypotheses for positive health, but Dr. Phillips described the task as 90 percent complete.
Dr. Phillips characterized the discussion on positive health with the following list:
- Support for school readiness as organizing framework for young children
- Maps onto other national studies of children (Australian, British)
- Critical need to identify who succumbs, who dodges the bullet, who thrives despite risk/exposures
- Consider measurement branching strategies (go deeper on X for subset of children)
- Critical mediational role of the home caretaking environment, including family interactions, parental response to stress, family conflict, consistent routines
- Recommend workshop on measurement of caretaking environment
- Recommend workshop on time-use measures for both caretaking environment and exposures in the home.
Dr. Phillips presented composites of recent definitions of positive health (World Health Organization Ottawa Charter and Institute of Medicine, 2004):
- “Health is the extent to which an individual or group is able, on the one hand, to develop and realize aspirations and satisfy needs, and on the other hand, to develop the capacities that allow them to change and cope with the environment.”
- “The extent to which individual children or groups of children are able or enabled to:
- Develop and realize their potential
- Satisfy their needs, and
- Develop the capacities that allow them to interact successfully with their biological, physical, and social environments.”
Dr. Phillips listed characteristics of sentinel capabilities:
- A strong consensus that they relate directly to essential elements of positive health
- Reflect capacity to transform potential into adaptive functioning
- Relate to critical transition points in childhood
- Evidence exists to suggest important genetic and environmental determinants
- Relevant measurement tools available
- Developmental pathways can have significant impact on current and future health.
Dr. Phillips described a proposed sentinel capability for the Study and presented the draft core hypotheses for this capability.
School readiness is a measure of a child’s capacity to respond to challenges and opportunities presented by entering the formal education process. School readiness involves:
- Physical/motor development
- Cognitive development
- Social and emotional development
- Language development
- Executive/regulatory/adaptive functioning.
Draft Core Hypothesis(Adapted from the Healthy Development Ad Hoc Working Group)
Familial—especially parenting—processes, social, and neurotoxic environmental exposures, in combination with genetic and biologically based individual differences in children (for example, stress reactivity, birth weight/gestational age, temperament, integrity of sensory and motor systems), exert profound influences—both positive and negative—on children’s healthy development conceptualized as:
- An important outcome in and of itself, linked to virtually every exposure examined in the Study
- A source of significant individual differences in functioning among children with known serious health conditions (some arising from exposures)
- A significant component of susceptibility considered in terms of how developmental trajectories carrying forward into adulthood are affected by potentially detrimental exposures during the childhood and adolescent years.
School readiness will be influenced by the character of family work patterns.
- Specific aim: The Study will identify the components, patterns, types, and consistency of work by both parents (and/or guardians) on the child-rearing/parenting received and the development of specific components of school readiness.
- Genetic and biological variation in the children will influence the impact of parental work patterns on school readiness.
- Preservation of key child-rearing activities and behaviors (for example, consistent warmth/nurturance, cognitive and language stimulation, stable family routines) will have a protective influence on patterns of school readiness in the context of parental work, perhaps especially in the context of low and/or unstable income.
- Characteristics of alternative child care environments (notably their amount, quality, and type) during the infant, toddlers, and preschool and school-age years, will mediate and modify the role of parental work in altering school readiness trajectories.
- Exposure to media will influence school readiness and will be related to parental work patterns.
School readiness will be influenced by the prevalence of parental depressive illness.
- Specific aim: The Study will identify the timing, intensity, and duration of parental depression on the child-rearing/parenting received and the development of specific components of school readiness.
- Genetic and biological variation in the children will influence the impact of parental depression on school readiness.
- Preservation of key child-rearing activities and behaviors (for example, consistent warmth/nurturance, cognitive and language stimulation, stable family routines) will have a protective influence on patterns of school readiness in the context of parental depression, perhaps especially in the context of single parenthood or marital conflict.
- Characteristics of alternative child care environments (notably their amount, quality, and type) during the infant, toddlers, and preschool and school-age years will mediate and modify the role of parental depression in altering school readiness trajectories.
School readiness is influenced by the timing, levels, intensity, type, and duration of exposures to specific and interacting neurotoxins, including lead, mercury or other heavy metals, and pesticides.
- Specific aim: The Study will identify the timing, levels, intensity, and duration of neurotoxic exposures on the child-rearing/parenting received and the development of specific components of school readiness.
- The level and duration of exposure to neurotoxins will influence school readiness.
- The developmental timing of exposures to neurotoxins will influence school readiness.
- Genetic and biological variation in the children will influence the impact of neurotoxic exposures on school readiness.
Dr. Phillips mentioned a second proposed sentinel capability for the Study but noted that the draft core hypotheses for this capability are not yet fully developed.
This capability will address positive health behaviors in adolescence, including patterns of avoidance of tobacco, alcohol, and illicit drugs.
Proposed next steps are:
- July 6: Completed hypothesis draft; send to Working Groups for review (neurodevelopment, social environment, Study design, others?)
- Mid-July: Battelle documents on measurement of school readiness completed
- July 23: Feedback due to Healthy Development Ad Hoc Working Group
- July 30: Final draft completed and sent out for formal review by NCSAC
- September meeting: Follow-up discussion of healthy development hypotheses
- October: Workshop on measurement of caregiving environment.
In a question-and-answer session, Dr. Phillips addressed the following issues and concerns:
- Age of school readiness. Loretta Jones, M.A., Healthy African American Families, noted that children are often ready for school by age 5. However, because of disparities in community resources, not all schools are ready to receive children at age 5. Dr. Phillips replied that the Study would be measuring school environments.
- Age of self-regulatory functioning. Dr. Doswell questioned the view that 5-year-old children need self-regulatory functioning that requires them to sit for hours at a time in a school desk. She commented that sitting down is an artificial concept for young school children. Dr. Doswell noted that there are issues of personal and environmental fit with regard to developing appropriate impulse control.
- Parenting and caregiving environment. Dr. Doswell asked whether the Study would have the capacity to assess family resources independent of income. Because parental neglect of children can transcend socioeconomic status, measures should include family structure, family routine, and “quality time.”
- Impact of health services. Dr. Dudley suggested that the Healthy Development Ad Hoc Working Group somehow address the impact of health services availability on health outcomes, perhaps by considering materials developed by the Health Services Working Group.
- Time-use diaries. Dr. Michael commented that time-use diaries should be considered. Dr. Graham agreed that time-use diaries are essential for assessing exposures to chemicals, media, and built environments. Dr. Mattison noted that there are automated methods to track children’s activities and behaviors. Dr. Mattison said that healthy development is being considered as a sixth thematic area for the Study. Although there is an increasingly diminished opportunity for including new Study hypotheses, the opportunity to consider more hypotheses remains.
Review and Discussion of Working Group Hypotheses
Prior to the meeting, NCSAC members were assigned to review and discuss hypotheses submitted by various Study Working Groups. Reviewers were asked to address the hypotheses from the following perspectives.
- Relevance to a national study of environmental impacts on health of children at various life stages
- Contemporary nature of hypotheses and analytic approach proposed
- Benefit of including hypothesis in the Study
- Cost of including the hypothesis in the Study (number of environmental samples needed, number of biological samples needed, information needed from individuals and families)
- Rating for each hypothesis (one of three groups):
- Recommend—the Study should include this as a core hypothesis.
- Consider—although it may not meet the criteria of a core hypothesis, this proposal should be given consideration if sufficient resources are available.
- Not recommended at this time—this should not be considered as core or potential secondary hypothesis at the present time.
Social Environment and Children’s Health and Development Hypotheses
Linda M. Burton, Ph.D., Pennsylvania State University, noted that much of the material developed and submitted by the Social Environment Working Group (SEWG) has already been endorsed by the NCSAC. In this presentation, Dr. Burton reviewed new information reflecting the SEWG’s responses to NCSAC requests during its March 2004 meeting, in which the NCSAC asked for the SEWG’s opinion of the set of core hypotheses developed by the ICC. The new information includes an integrated power analysis assessing the sample sizes needed to test specific hypotheses and an eighth hypothesis concerning media effects. Dr. Burton explained that because the social environment and the different domains specified in the SEWG’s materials are strongly related to health outcomes of interest to the Study, these social environment domains should be centrally included in the core hypotheses. Not only do social environmental factors represent a fundamental cause of certain health outcomes, but they may also serve as mediating factors. The Study may be able to detect numerous interactions among social environments and other environments that are critical to understanding child health outcomes. The SEWG’s hypotheses have particular relevance to policy makers and may have ramifications with regard to future interventions.
Dr. Burton said that the SEWG materials emphasized the priority areas of the Study, including families and households, socioeconomic status, neighborhoods and communities, formal institutions, public policy, and media. The SEWG’s hypotheses specifically focus on the outcomes of asthma and obesity. These hypotheses effectively integrated both social environmental and biological factors. As for the new information, the SEWG clearly described the power issues related to sample sizes necessary to test the hypotheses, and their justifications were based on multi-leveled, hierarchical approaches. That is, the SEWG’s hypotheses would allow the Study to investigate multiple pathways to specific outcomes. Dr. Burton said that because the variables are multi-dimensional, they require the large sample size of the Study and will allow investigators to assess interactions among factors. The SEWG hypotheses would allow the Study to focus on the critical domains in the social environment, including issues of rural versus urban environments, variability in socioeconomic status, and race and ethnicity. According to Dr. Burton, the SEWG’s hypotheses could be easily integrated with hypotheses from other Working Groups and the SEWG is willing to collaborate with other Working Groups to facilitate this integration.
Dr. Phillips began her review by declaring her strong support of the proposed media hypotheses, which she described as highly relevant because of the ubiquitous nature of modern media exposures. These exposures may have significant developmental effects, both positive and negative, depending on the content of the media, the nature of the exposure, and a range of variables such as parental modulation. Dr. Phillips cited some new data from the Kaiser Family Foundation on early childhood exposure to media, but noted that, overall, little is known about media effects on a range of health outcomes. Some of these outcomes may be simultaneously affected by other exposures such as neurotoxins, which may relate to outcomes such as attentional deficits. Dr. Phillips emphasized the importance of assessing how the constellation of common media exposures affects childhood development and the need to include such an assessment in the Study. Dr. Phillips mentioned recent research on tragic events such as 9/11 and the Challenger incident indicating the role of the media as a source of exacerbated stress reactions in children. Because of its contemporary nature, Dr. Phillips described the media hypotheses as “extremely timely.” Media exposure is also a keen issue for parents and policymakers. However, assessing media effects over the duration of the Study may prove challenging (that is, a “moving target”) because of ever-evolving and increasingly portable technologies. According to Dr. Phillips, the benefits of the media hypotheses include:
- Added capacity for the Study to assess developmental consequences of multiple constellations of exposures
- Consideration of the positive or beneficial effects of media exposure under some circumstances
- Potential to contribute to the Study’s capacity to begin to understand the processes underlying other exposures and outcomes such as obesity, violence, and use of tobacco and other harmful substances that are correlated to media exposure
- Increased likelihood that an exposure affecting non-at-risk children and of concern to a wide spectrum of parents and policymakers is included in the Study.
Dr. Phillips said that the cost of adding media exposure questions to the Study would be minimal but that gathering time-use data might incur some additional costs (for example, home visits to assess media exposure or electronic data collection tools such as Arbitron). Because cost is not a big deterrent, Dr. Phillips highly recommended that the media hypotheses be added to the Study.
In a question-and-answer session, Dr. Phillips addressed the following issues and concerns:
- Protective and resiliency factors
- Learning and school readiness
- Diversity of media types
- Cultural considerations
- Role of media in prosocial behavior
- Literacy issues diaries and time-use records
- Integration of racism measures into the media hypotheses.
Built Environment and Physical Activity in Children Hypothesis
Dr. Mattison (filling in for primary reviewer Dr. Landrigan) summarized Dr. Landrigan’s review of this hypothesis. The physical environment hypothesis addresses the relevant health outcome of obesity but may not require a large sample size to study. The hypothesis would be greatly strengthened by including measures such as serum levels of endocrine disrupters (for example, phthalates). The hypothesis is “state-of-the-art,” and it addresses an endpoint of great and growing importance (that is, obesity). Because anthropomorphic data would be gathered as an integral part of the Study, the cost of including the hypothesis would be low. Therefore, this hypothesis should be included in the Study’s core hypotheses.
Dr. Michael (secondary reviewer) said that his conclusion of this hypothesis was the same as Dr. Landrigan’s. He described the approach as practical, sensible, and reasonable. The hypothesis would be inexpensive, impose a low burden, and be easy to measure. Subjects’ access to built environments such as playgrounds and school yards would be relatively easy to assess and could be correlated with children’s physical activity and outcomes such as obesity. Dr. Michael questioned the Physical Environment Working Group’s assumptions that the Study would collect data on body mass index and diets. Because the Working Group also assumes the collection of data on physical activity, Dr. Michael suggested that time-use measurements might be of value. While a sample size of 100,000 might not be necessary for this hypothesis, its inclusion could provide insightful information and a valuable dataset for future analysis.
Child Maltreatment Hypotheses
David C. Bellinger, Ph.D., Boston Children’s Hospital, described these hypotheses as previously submitted and reviewed. The hypotheses address important issues but raise concerns regarding feasibility, particularly because of reporting requirements, confidentiality, and privacy issues. For example, reporting requirements may negatively affect retention. Dr. Bellinger noted that the Injury Working Group had revised and expanded their original submissions, which now included two basic hypotheses. The first hypothesis addresses the relationship between marital and/or relationship violence and subsequent physical abuse of children and includes several subhypotheses on mediating multi-level risk factors and the likelihood of injury. The second hypothesis addresses the impact of physical and sexual abuse on cortisol levels, stress reactivity, the likelihood of post-traumatic stress disorder, conduct disorder and other psychiatric disorders, and impaired development. Several subhypotheses involve dose-response effects, aspects of resiliency, protective factors, and effects of interventions. Dr. Bellinger, citing incidences of child physical and sexual abuse, reported that the Study’s sample size would be appropriate for these hypotheses. In their write-up, the Injury Working Group adequately responded to the concerns of the ICC and NCSAC by citing the successful efforts of several ongoing studies. The success of these studies provides assurance that adequate procedures can be implemented to minimize the effects of consent and reporting issues on Study retention. Given the importance of child maltreatment in American society and the resulting costs—in terms of morbidity and dollars—Dr. Bellinger recommended that these hypotheses be included as a central element of the Study.
Dr. Doswell (secondary reviewer) was not as enthusiastic about the hypotheses, citing her reservations about the proposed markers (neighborhood factors such as high rates of crime, violence, joblessness, poverty, and high family mobility). Dr. Doswell described the hypotheses as operating from a deficit model. Because of this, she suggested that the Study assess family violence and child maltreatment across the socioeconomic spectrum. This would be a novel approach and would expand what is currently described in the scientific literature. Dr. Doswell expressed reservations about enrolling victims of chronic physical abuse because of issues regarding imminent danger. She also questioned whether these hypotheses would need to be examined in a very large sample such as the Study. Dr. Doswell described the proposal to investigate genetic vulnerability as underdeveloped. She noted that the linkage between cortisol levels and sexual and/or physical abuse does not require a large sample size. Finally, because Dr. Doswell was concerned about the quality and duration of interventions described in the hypotheses, she would not recommend this subhypothesis.
Unintentional Injury Hypothesis
Dr. Suarez described this hypothesis as lacking specificity; she said that nothing in the material is explicitly defined or well developed. In her opinion, a sample size of 100,000 is not needed for this hypothesis. Therefore, Dr. Suarez said she could not recommend this proposal at this time.
New Unintentional Injury Hypothesis
Dr. LeMasters described this hypothesis about risk-taking behavior as undeveloped. She noted that neither exposure nor outcome measurements are described. Therefore, Dr. LeMasters said she could not recommend this proposal at this time and suggested that the Injury Working Group submit more details on this hypothesis. The meeting participants discussed the following issue regarding risk-taking behaviors and unintentional injuries:
- Incidence of morbidity and mortality due to unintentional injury
- Prevention strategies
- Sample size required to assess causative factors
- Genetic variations in neurotransmitters (for example, serotonin polymorphism) and behaviors
- Environmental determinants of injury outcomes
- Relative inexpensiveness of data collection
- Aspects of parenting behaviors, parental supervision, and childcare environments.
Subsequent to this discussion, the NCSAC agreed to send this hypothesis to the Program Office for further development and to engage other Working Groups to assist with this effort.
Gene-Environment Interaction Hypothesis
Dr. Dudley explained that this hypothesis was submitted by Sarah S. Knox, Ph.D., NICHD, NIH, DHHS, who is currently working in the Program Office. The hypothesis is related to discussions during Gene-Environment Interaction and the Regulation of Behavior Workshop that was held in June 2004. Dr. Dudley described the hypothesis as highly significant, particularly with regard to the current prevalence of depression in the adolescent population. Including this hypothesis would create an opportunity to assess an important cutting-edge scientific area (that is, gene polymorphism and environmental interactions on behavioral outcomes). The proposed research has the potential to uncover new information and fills a key area that has not been studied. Because the proposal did not include a cost estimate, power calculation, or sample size, Dr. Dudley suggested that the hypothesis be included in the Study in some manner but that it not be considered as a core hypothesis.
P. Barry Ryan, Ph.D., Emory University Rollins School of Public Health, explained that the hypothesis offered no rationale as to why it should be included in the Study. Dr. Ryan thought that the proposal would be better served in a smaller pilot study. He said that the hypothesis needed further development.
After a brief discussion, NCSAC members agreed that the gene-environment interaction hypothesis could serve a critical role in the Study and that it could serve as a “placeholder” for an important class of hypotheses. They recommended that the Program Office continue to develop this hypothesis. Dr. Knox said that she would complete the power calculations.
Report from Joint ICC/NCSAC Executive Committee
Judith A. Graham, Ph.D., NCSAC Member and Chair of the Executive Committee, American Chemistry Council
Dr. Graham said that most of the committee’s recent efforts have focused on sampling issues in order to prepare for the June 2004 NCSAC meeting. Other recent committee activities include:
- Developing guidance and focus for the next NCSAC meeting
- Preparing a status report to Duane F. Alexander, M.D., Director, NICHD, NIH, DHHS
- Exploring methods to improve marketing of the Study
- Prioritizing Study hypotheses
- Evaluating outcomes from the workshop on racism and determining how to best incorporate them into the Study
- Considering operational issues for Study implementation
- Focusing on the upcoming challenges
- Working with the Program Office to identify areas that could benefit from NCSAC advice.
Discussion Regarding Future of NCSAC
Facilitated by Dr. Mattison
Dr. Mattison explained that the future of the NCSAC is not yet determined and that its role would continue to evolve as the Study moves from the planning stage to implementation. During this transition, the types of activities and input from the various Working Groups will diminish, and the Working Groups’ duties will finish when the Study exposures and outcomes have been determined. The NCSAC will continue its advisory role, particularly with regard to protocol development, initiation of vanguard centers, and ethics of recruitment. Future issues include NSCAC membership, constitution, and expectations.
Discussion Regarding AGES
Facilitated by Dr. Mattison
Dr. Mattison said that he hoped that there would be a continuous, open, and interactive dialogue between the NHGRI and the Study. Dr. LeMasters emphasized the potential public good that could develop from the cooperation and partnership of AGES and the Study. Although Dr. Fleischman expressed some doubt about the ability of sincere collaboration between AGES and the Study, he suggested that the Study remain open to alternate strategies of working with AGES. Dr. Mattison noted that the Study seeks a broad perspective on development, whereas the AGES approach seems more static. The Study has substantial differences and unique perspectives that would not be captured by AGES.
NCSAC Open Forum
Facilitated by Dr. Mattison
Dr. Mattison invited public comment and opened to floor to general discussion of the Study. Doris B. Haire, American Foundation for Maternal and Child Health, mentioned two areas of research that could be of potential interest to the Study:
- Long-term effects of drugs given to women during labor and delivery
- Long-term effects of ultrasound, which could be registered as part of a child’s birth record.
Dr. Graham reiterated the NCSAC’s availability to continue advising the Study.
Donald R. Mattison, M.D., NCSAC Chair, NICHD, NIH, DHHS
Jan L. Leahey, NCSAC Executive Secretary, NICHD, NIH, DHHS
David C. Bellinger, Ph.D., Boston Children’s Hospital
Linda M. Burton, Ph.D., Pennsylvania State University
*George P. Daston, Ph.D., Proctor and Gamble Company
Willa M. Doswell, R.N., Ph.D., University of Pittsburgh School of Nursing
Donald J. Dudley, M.D., University of Texas Health Sciences Center at San Antonio
*Barbara R. Foorman, Ph.D., University of Texas Health Science Center at Houston
Judith A. Graham, Ph.D., American Chemistry Council
*Fernando A. Guerra, M.D., M.P.H., San Antonio Metropolitan Health District
Loretta Jones, M.A., Healthy African American Families
*Shiriki Kumanyika, Ph.D., M.P.H., University of Pennsylvania School of Medicine
Philip J. Landrigan, M.D., Mount Sinai School of Medicine
Grace LeMasters, Ph.D., University of Cincinnati
*Roderick Joseph A. Little, Ph.D., University of Michigan
Robert T. Michael, Ph.D., University of Chicago
Deborah A. Phillips, Ph.D., Georgetown University
P. Barry Ryan, Ph.D., Emory University Rollins School of Public Health
*M. Anne Spence, Ph.D., University of California, Irvine Medical Center
Lucina Suarez, Ph.D., Texas Department of Health
*Daniel J. Swartz, Children’s Environmental Health Network
*Barry S. Zuckerman, M.D., Boston University School of Medicine
* Did not attend
*Amy Branum, M.S.P.H., National Center for Health Statistics, CDC, DHHS
Adolfo Correa, M.D., Ph.D., National Center on Birth Defects and Developmental Disabilities, CDC, DHHS
Sarah A. Keim, M.A., NICHD, NIH, DHHS
Woodie Kessel, M.D., M.P.H., Office of the Secretary, DHHS
*Carole A. Kimmel, Ph.D., National Center for Environmental Assessment, EPA
Mark Klebanoff, M.D., M.P.H., NICHD, NIH, DHHS
Pauline Mendola, Ph.D., Office of Research and Development, EPA
*Sheila A. Newton, Ph.D., NIEHS, NIH, DHHS
James J. Quackenboss, M.S., Office of Research and Development, EPA
Peter C. Scheidt, M.D., M.P.H., NICHD, NIH, DHHS
Kenneth C. Schoendorf, M.D., M.P.H., National Center for Health Statistics, CDC, DHHS
Sherry G. Selevan, Ph.D., Office of Research and Development, EPA
Marshalyn Yeargin-Allsopp, M.D., National Center on Birth Defects and Developmental Disabilities, CDC, DHHS
* Did not attend
Observers and Other Participants
Janet Aker, The Blue Sheet, FDC Reports, Inc.
Teneshia G. Alston, NICHD, NIH, DHHS
Christine A. Bachrach, Ph.D., NICHD, NIH, DHHS
Marion J. Balsam, M.D., NICHD, NIH, DHHS
Adelaide Barnes, B.A., NICHD, NIH, DHHS
Arthur M. Bennett, M.E.A., B.E.E., NICHD, NIH, DHHS
Sophie Bous, NICHD, NIH, DHHS
Ruth A. Brenner, M.D., M.P.H., NICHD, NIH, DHHS
Margo Brinkley, M.S., RTI International
Clarice Brown, CODA, Inc.
Leni Buff, NICHD, NIH, DHHS
Audrey Burwell, M.S., Office of the Secretary, DHHS
Kimberly Caraballo, NICHD, NIH, DHHS
Katherine Colbath, American Academy of Pediatrics
Barry E. Collins, Ph.D., Healthy African American Families
Lauren Pindzola Courtney, M.P.H., RTI International
Ben P. Daughtry, R.Ph., MBA, FACHE, Dynport Vaccine Company, LLC
Elizabeth A. Davis, NICHD, NIH, DHHS
Linda L. Dimitropoulos, Ph.D., RTI International
Terence Dwyer, M.D., M.P.H., NICHD, NIH, DHHS
Brenda Ecken, B.S.N., M.Ed., Booz Allen Hamilton Inc.
Jonas H. Ellenberg, Ph.D., Westat
Alan R. Fleischman, M.D., NICHD, NIH, DHHS
David G. Forvendel, B.S., RTI International
Alexa Fraser, Ph.D., Westat
Warren Galke, Ph.D., NICHD, NIH, DHHS
Peter J. Gergen, M.D., M.P.H., NICHD, NIH, DHHS
Alan E. Guttmacher, M.D., National Human Genome Research Institute, NIH, DHHS
Doris B. Haire, American Foundation for Maternal and Child Health
Patrick Hemming, American Academy of Pediatrics
Anne E. Imrie, Science Applications International Corporation
Carole Johnson, CompTech Associates, Inc.
Raffael Jovine, Ph.D., Booz Allen Hamilton Inc.
Sarah S. Knox, Ph.D., NICHD, NIH, DHHS
William Lyman, Ph.D., Wayne State University
Shara Marrero, NICHD, NIH, DHHS
Melanie Martinez, NICHD, NIH, DHHS
Elizabeth Meltzer, Society for Research in Child Development
John R. Menkedick, M.S., Battelle Memorial Institute
Maurice Owens, Ph.D., Science Applications International Corporation
Haluk Ozkaynak, Ph.D., M.S., Office of Research and Development, EPA
Sherri L. Park, NICHD, NIH, DHHS
Bobbie Peterson, Altarum
James H. Raymer, Ph.D., RTI International
Jerry D. Rench, Ph.D., RTI International
J. Brook Rolter
Mike Rozendaal, Iowa Foundation for Medical Care
Lee Salamone, American Chemistry Council
David A. Savitz, Ph.D., University of North Carolina
Mary Jean Schmitt, Sun Microsystems, Inc.
Kathy Schneider, Ph.D., Iowa Foundation for Medical Care
Julie A. Schoenborn, M.B.A., Brady Corporation
Angela L. Sharpe, M.G., Consortium of Social Science Associations
Ann M. Smith, B.A., Michigan State University
Peter A. Soyka, Native American Management Services, Inc.
Kathleen A. Stralka, M.S., Science Applications International Corporation
Warren J. Strauss, Sc.M., Battelle Memorial Institute
Karen A. Studwell, J.D., American Psychological Association
Robin Toblin, NICHD, NIH, DHHS
Ann M. Vinup, Learning Disabilities Association of America
Diane K. Wagener, Ph.D., RTI International
Roy Whitmore, Ph.D., RTI International
Edward Tin Wong, Ph.D., Science Applications International Corporation
I hereby certify that, to the best of my knowledge, the foregoing minutes are accurate and complete.
| August 25, 2004
Donald Mattison, M.D.
National Children’s Study
Federal Advisory Committee
Appendix A: Charge Questions to the National Children’s Study Advisory Committee Regarding Sampling Design
The National Children’s Study Advisory Committee (NCSAC) has recommended that the sampling strategy for the Study be a probability-based representative sample. The Interagency Coordinating Committee (ICC) and the Program Office have seriously considered this recommendation and have identified some significant challenges in the implementation of this strategy to meet the goals of the National Children’s Study. In addition, we asked Battelle to develop white papers on several key issues related to sampling, which served as the background for discussions at the Sampling Design Workshop. The Workshop Panel provided its findings to the NCSAC in its report, which identified two major types of design options. The ICC and Program Office have confronted the following issues and questions with planning specifics of the sampling strategy. Understanding that these questions may not be fully answerable at this stage or without additional information, we are interested in the NCSAC’s thoughts and responses to help us as much as possible with the practicalities of the sample design.
1. Given that extensive clinical exams and sample collections are needed over many years for answering core hypotheses (for example, cord blood, placenta, newborn exams, annual/biannual/periodic developmental and neurological exams) some modifications or compromise of a strict nationally representative probability-based sample appear to be necessary. How does the NCSAC envision recruiting and retaining the sample with these relatively intense measurements and with sufficient logistical feasibility and cost constraints?
2. Given that the overriding priority of the Study is to assess exposure-outcome relationships (versus prevalence estimates) and that a probability-based sample will likely add some measure of additional costs above that of a non- or less-representative sample, what is the relative importance of allocating additional costs for (1) nationally or (2) locally representative samples relative to collecting detailed measures or design elements? Stated differently, if there are tradeoffs between national representativeness and measures to test core hypotheses due to cost constraints, what would be a reasonable price to pay for national representativeness?
3. Given that there may be substantial costs and burden associated with preconception enrollment and follow-up of women, is it appropriate to focus some proportion of this effort on women who are intending to become pregnant? What proportion of the Study population should be recruited before pregnancy?
4. Considering the reports developed by Westat and Battelle and the report of the Sampling Design Workshop Panel, are there alternative and/or innovative sampling strategies and techniques for achieving the recommended representativeness with acceptable costs and retention rates that have not been considered or proposed previously?
5. Evaluation of neighborhood or community variables (for example, exposures, built environment, and social factors) longitudinally over 21 years appears to be increasingly efficient with greater degrees of clustering. What is the appropriate balance of clustering for measuring community attributes and cost efficiency with representativeness?
6. Given that it is important for the Study to describe environmental and social influences that are measured at the geographic cluster-level, what information (factors) and approach should be used to assure that the broad diversity and the necessary attributes of the sample are achieved (for example, racial/ethnic characteristics, urban/rural make-up, and proximity to identifiable pollutant sources)?