In the Aftermath of the National Children’s Study: Is Large Birth Cohort Data Still a Priority? | Pediatrics | JAMA Pediatrics | JAMA Network
[Skip to Navigation]
March 2017

In the Aftermath of the National Children’s Study: Is Large Birth Cohort Data Still a Priority?

Author Affiliations
  • 1Nuffield Department of Obstetrics and Gynecology, The George Institute for Global Health, University of Oxford, Oxford, United Kingdom
  • 2Norwegian Institute of Public Health, Oslo, Norway
  • 3The Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark
JAMA Pediatr. 2017;171(3):214-215. doi:10.1001/jamapediatrics.2016.3968

The 2014 decision to stop the US National Children’s Study (NCS)1 brings to the forefront questions about what has been lost and how studies such as this might still be important almost 20 years after initiation in 2000. The rationale then was clear.2 Little progress had been made in the previous decades in understanding the causes of many major childhood disorders, and there was insufficient evidence available to confidently mount interventions to prevent many of them. A lack of evidence from cohort studies with prospective data had left a major evidence gap in childhood disease etiology, in stark contrast to efforts involving successful research on adult diseases where cohort studies were a central component.

Important pediatric conditions for which prospective data might be critical included birth defects, childhood cancer, type 1 diabetes, and autism.2 The exposures of interest for these conditions embraced infections of the mother and infant, environmental chemicals, nutrition, and growth of the fetus. Environmental factors, such as infections and many chemicals, could not be measured validly with retrospective data because parents’ recall of these had been shown to be affected by bias.3,4 Analysis of biospecimens collected after disease onset failed to provide data on what was happening during the key time when disease was evolving. Therefore, the field of pediatric disease prevention could not move forward without prospective data of this kind.

The barrier had been the cost and effort required to conduct cohort studies of the size needed to provide adequate power to investigate important pediatric diseases that, relative to adult diseases, are uncommon. Only very large cohort studies, with at least 100 000 participants, would be sufficiently powered to have a chance of providing reliable evidence of an association with potential risk factors on the less common serious diseases of interest.

Nonetheless, the proposition that this obstacle must be overcome if progress was to be made in understanding pediatric disease causation was clearly gaining strong support toward the end of the 20th century. As well as the decision to start the NCS, Danish and Norwegian birth cohort studies were launched in the period leading up to 2000. They benefited from the comprehensive sampling frames for health studies and health data linkage that those nations possess. Partly as a consequence of this, they were able to enroll and measure mothers in pregnancy and their infants and collect data on their 100 000 participants for what they perceived as acceptable costs, around US$20 million.

The NCS faced bigger cost hurdles because of the lack of such infrastructure. In addition, its planners argued that prenatal recruitment was necessary to avoid potential selection bias associated with failure to include mothers who miscarried or experienced a stillbirth and decided to sample from the only feasible sampling base in the United States that would include all potential new mothers: the household. Importantly, this would have enabled home-based collection of exposure data that the other more modestly funded international cohorts were unable to incorporate. Analyses conducted later by the NCS team revealed that this sampling approach would inflate the costs over an alternative scenario involving sampling of prenatal clinicians and birth hospitals by a ratio of 30:1. However, it seems that the lower-cost approach was not seen as sufficiently attractive an option to salvage the study in the minds of those who undertook the final review. It has been reported that when the NCS was stopped, it had already cost more than $1 billion, and completion of a successful study within a fundable budget was judged not to be feasible.

In contrast, more than 15 years after their initiation, the Danish National Birth Cohort and Norwegian Mother and Child Cohort Study are well on the way to providing important evidence on the conditions they set out to investigate. More than 800 articles have been published from data emerging from the 2 cohorts. Key findings include studies on the importance of maternal chronic diseases in pregnancy5 to child health and evidence that folic acid intake by the mother periconceptionally might be related to occurrence of autism spectrum disorder.6

For some diseases, in particular cancer, type 1 diabetes, and cerebral palsy, it was apparent at the planning stage of these cohorts that even 100 000 participants would be marginal for providing the required power. Consequently, the Danish National Birth Cohort and Norwegian Mother and Child Cohort Study have participated in an initiative to pool data with some older studies that have relevant data to obtain the necessary power. These include the Jerusalem Perinatal Study, the National Collaborative Perinatal Project, Avon Longitudinal Study of Parents and Children, and the Tasmanian Infant Health Survey, as well as newer, large birth cohorts such as the Japan Environment and Children’s Study and others just commencing. This initiative, the International Childhood Cancer Cohort Consortium, only includes follow-up for cancer, but conceivably, in time, other conditions will be examined using the same strategy in the future.

Even the International Childhood Cancer Cohort Consortium program, working as it is with a participant pool of approximately 400 000 mothers and infants, among whom 670 cancers have already occurred, struggles to find power to investigate all the exposures of interest. While the numbers available for follow-up investigation in this consortium seem large compared with those available for the early articles published from landmark adult cohort studies, such as Framingham,7 in the era of big data, their modest size concerns many critics. Possibly, the focus on precision rather than avoidance of bias has become counterproductive8 to the solution of childhood disease puzzles.

How necessary is it to provide evidence from a source with prospective data? It remains as important now for major childhood diseases as it was 20 years ago because most of the conditions of interest then, such as cancer and birth defects, are waiting on the key pieces in the puzzle that only prospective data and biospecimen analysis can provide. Advances in techniques for capturing environmental influences on human biology, such as epigenomics and the use of organism-specific DNA in the examination of infectious disease exposures, have opened up new horizons for determining the contribution of relevant exposures.

Can the United States contribute, given its niche disadvantages for data collection that seem to have priced it out of contributing to the large-scale population-based data collection required? The National Institutes of Health clearly thinks it can to some extent and has created a new grant program, Environmental Influences on Child Health Outcomes, that seeks to enhance the quality of evidence through grants to improve exposure measures and to pool the data already available from existing smaller cohorts in the United States. The effort, which may involve up to 50 000 participants in pooled samples, will contribute to the understanding of causes of the more common conditions identified by the NCS, such as obesity, airways disease, and poor intellectual development, but will not provide the power to substantially inform efforts to identify causes of a number of the rare but important childhood diseases. While the US scientific community may not be ready for a new NCS, the United States can play a major part in pushing forward global collaborative efforts, such as International Childhood Cancer Cohort Consortium, that emanated from the NCS and where the involvement of the National Institutes of Health gave the credibility needed to launch such ambitious quests.

Back to top
Article Information

Corresponding Author: Terence Dwyer, MPH, MD, MB, BS, AO, The George Institute for Global Health, Oxford Martin School, University of Oxford, 34 Broad St, Oxford OX1 3BD, United Kingdom (

Published Online: January 9, 2017. doi:10.1001/jamapediatrics.2016.3968

Conflict of Interest Disclosures: None reported.

Kaiser  J.  Biomedical research: ambitious children’s study meets disappointing end.  Science. 2014;346(6216):1441.PubMedGoogle ScholarCrossref
Olsen  J, Melbye  M, Olsen  SF,  et al.  The Danish National Birth Cohort: its background, structure and aim.  Scand J Public Health. 2001;29(4):300-307.PubMedGoogle ScholarCrossref
Roman  E, Simpson  J, Ansell  P, Lightfoot  T, Smith  A.  Infectious proxies and childhood leukaemia: findings from the United Kingdom Childhood Cancer Study (UKCCS).  Blood Cells Mol Dis. 2009;42(2):126-128.PubMedGoogle ScholarCrossref
Schüz  J, Spector  LG, Ross  JA.  Bias in studies of parental self-reported occupational exposure and childhood cancer.  Am J Epidemiol. 2003;158(7):710-716.PubMedGoogle ScholarCrossref
Zhu  Y, Olsen  SF, Mendola  P,  et al.  Growth and obesity through the first 7 y of life in association with levels of maternal glycemia during pregnancy: a prospective cohort study.  Am J Clin Nutr. 2016;103(3):794-800.PubMedGoogle ScholarCrossref
Surén  P, Roth  C, Bresnahan  M,  et al.  Association between maternal use of folic acid supplements and risk of autism spectrum disorders in children.  JAMA. 2013;309(6):570-577.PubMedGoogle ScholarCrossref
Dawber  TR, Meadors  GF, Moore  FE  Jr.  Epidemiological approaches to heart disease: the Framingham Study.  Am J Public Health Nations Health. 1951;41(3):279-281.PubMedGoogle ScholarCrossref
Hill  AB.  The environment and disease: association or causation?  Proc R Soc Med. 1965;58:295-300.PubMedGoogle Scholar