Linet, Martha S.
From the Division of Cancer Epidemiology and Genetics, National Cancer Institute, Bethesda, MD.
Submitted 31 March 2008; accepted 8 April 2008; posted 20 May 2008.
Supported by the Intramural Research Program of the National Cancer Institute, National Institutes of Health.
Editors’ note: Related articles appear on pages 532 and 534.
Correspondence: Martha S. Linet, MD, Radiation Epidemiology Branch, Division of Cancer Epidemiology & Genetics, National Cancer Institute 6120 Executive Blvd., EPS Room 7048, Bethesda, MD. E-mail: firstname.lastname@example.org.
The report by Divan et al linking early exposures to cell phones with behavioral problems in young children, published in this issue of Epidemiology, provides an opportunity to consider pursuit of high-risk hypotheses that could open new areas of understanding in contrast with the more common assessment of lower-risk leads. Other key considerations for epidemiologists are the requirements for selecting plausible risk factors while remaining alert to serendipitous discovery, for validating proxy measures of complex disease outcomes and exposures, and for pursuing replication of unexpected results in independent settings. In the face of unexpected findings, research consortia provide opportunities for pursuing exploratory and follow-up studies of high-risk hypotheses. Reviewers and editors also play major roles.
There is a need to identify preventable causes of neurodevelopmental disorders. Normal central nervous system (CNS) development involves many precisely integrated steps from the prenatal period through adolescence, which renders the system highly vulnerable to developmental injury.1 The timing of specific exposures in key neurodevelopmental events,2 and children’s physiologic characteristics and behaviors3,4 may explain children’s increased susceptibility. Epidemiologic and laboratory data have linked neurologic and behavioral problems in children with in utero or postnatal exposure to methylmercury, lead, ethyl alcohol, tobacco, ionizing radiation to the brain, and other agents.5 There are, no doubt, additional neurotoxins among more recently introduced chemical and nonchemical exposures. A National Academy of Sciences committee has estimated that 3% of developmental disabilities arise directly from environmental exposures and an additional 25% from gene-environment interactions.6 Most neurodevelopmental disorders, however, are unexplained.
The report by Divan et al7 on early exposures to cell phones and behavioral problems in young children provides an opportunity to consider the shoals and whirlpools that impair navigation by epidemiologists. How might one balance pursuit of high-risk novel hypotheses that might open new areas of understanding (but more likely will lead to a null result), with lower-risk leads that are more likely to be true but increase our understanding of etiology only marginally?
Plausibility is important for decision-making, but one should always remain alert to serendipitous discovery. Prevalence of ADHD is higher in boys than in girls and in children in low-income than in high income families, and so plausible postnatal exposures are more likely in boys than girls, and in children from families with lower incomes than those with higher incomes.8 Serendipity might lead to the discovery of a completely unknown mechanism of action such as a powerful indirect effect. “Plausibility,” as one of the Bradford Hill criteria for causation, specifies that the postulated exposure-disease association should agree with known pathologic processes.9 Even so, exposure-disease relationships that do not fit the established biologic processes should not necessarily be dismissed, since unexpected statistical associations could lead to new biologic understanding.
Do standardized diagnostic questionnaires administered to mothers provide sufficient information for diagnostic assessment of complex pediatric disorders? Probably, although this could be explored further. The prevalence of ADHD estimated in community-based US studies ranges 10-fold depending on the proxy sources of diagnostic data (eg, parental report, medical record review, or history of medication use) and the various versions of the Diagnostic and Statistical Manual of Mental Disorders used to define diagnostic criteria.10,11 Since the best estimate of population-based ADHD prevalence in the United States was derived using the structured National Institute of Mental Health Diagnostic Interview Schedule for Children-IV to mothers,11 it seems reasonable that Divan et al used a structured diagnostic interview.
In the event of an unexpected finding of potential public health importance, what are the responsibilities of the investigators? It can be helpful to explore the validity of exposure and outcome information. For Divan et al,7 one modest goal might be to compare diagnostic information from the Strengths and Difficulties Questionnaire with data from a second “gold standard” diagnostic instrument in a subset of the target population. Similarly, intensive diary measurement of cell phone use could be carried out in a subset to validate questionnaire data. As always, alternative explanations such as recall bias12,13 and selection bias14,15 should be evaluated.
Regardless of whether other investigators pursue these findings, the original investigators should themselves seek opportunities to confirm their results in an independent setting. A good precedent was set by Robison and colleagues,12 who had reported a link between maternal marijuana use during pregnancy and increased risk of acute myeloid leukemia in offspring. Fifteen years later, these investigators published a second high-quality study in which they were unable to confirm their original finding.13
Editors and reviewers also have responsibilities in the face of unexpected findings. Reviewers should provide critical and constructive feedback, and editors should serve as gatekeepers by rejecting unsound research. But as most authors know, persistence, responsiveness, and flexibility often improve the authors’ chances of publishing their papers—even from inadequate studies.
Epidemiologists should not shy away from undertaking investigations of complex disorders that present diagnostic challenges. Still, priorities must be established to enable comprehensive assessment of key hypotheses. In the specific example, the need to reduce the enormous gaps in our current understanding of the etiology of children’s neurodevelopmental disorders must be balanced by fiscal and temporal constraints imposed by the scope of the effort needed for such studies.
The development of research consortia will improve our ability to conduct studies that assess high-risk novel hypotheses or potentially contentious issues. Consortia could permit parallel exploratory or follow-up studies in 2 or more populations. The rapidly expanding genome-wide association studies carried out by consortia provide a framework for new study designs, analytic strategies, bioinformatics approaches, and input from multiple experienced investigators. Navigating the roiling waters between the urgent need to evaluate new chemical and physical agents, and the continuous need for improvements in epidemiologic study quality, will test us in ways that echo Odysseus’ voyage through the straits guarded by Scylla and Charybdis.
1. Rodier PM. Environmental causes of central nervous system maldevelopment. Pediatrics. 2004;113:1076–1083.
2. Rice D, Barone S Jr. Critical periods of vulnerability for the developing nervous system: evidence from human and animal models. Environ Health Perspect. 2000;108(Suppl 3):511–533.
3. Moya J, Bearer CR, Etzel RA. Children’s behavior and physiology and how it affects exposure to environmental contaminants. Pediatrics. 2004;113(4 Suppl):996–1006.
4. Xue J, Zartarian V, Moya J, et al. A meta-analysis of children’s hand-to-mouth frequency data for estimating nondietary ingestion exposure. Risk Anal. 2007;27:411–420.
5. Grandjean P, Landrigan PJ. Developmental neurotoxicity of industrial chemicals. Lancet. 2006;368:2167–2178.
6. National Research Council. Scientific frontiers in developmental toxicology and risk assessment. Washington DC: National Academy Press; 2000.
7. Divan HA, Kheifets L, Obel C, Olsen J. Prenatal and postnatal exposure to cell phone use and behavioral problems in children. Epidemiology. 2008;19:523–529.
8. Scahill L, Schwab-Stone M, Merikangas KR, et al. Psychosocial and clinical correlates of ADHD in a community sample of school age children. J Am Acad Child Adolesc Psychiatry. 1999;38:976–984.
9. Hill BA. The environment and disease: association or causation? Proc Royal Soc Med. 1965;58:295–300.
10. Millichap JG. Etiologic classification of attention-deficit/hyperactivity disorder. Pediatrics. 2008;121:e358–e365.
11. Froehlich TE, Lanphear BP, Epstein JN, et al. Prevalence, recognition, and treatment of attention-deficit/hyperactivity disorder in a national sample of U.S. children. Arch Pediatr Adolesc Med. 2007;161:857–864.
12. Robison LL, Buckley JD, Daigle AE, et al. Maternal drug use and risk of childhood nonlymphocytic leukemia among offspring. An epidemiologic investigation implicating marijuana (a report from the Children’s Cancer Study Group). Cancer. 1989;63:1904–1911.
13. Trivers KF, Mertens AC, Ross JA, et al. Parental marijuana use and risk of childhood acute myeloid leukaemia: a report from the Children’s Cancer Group (United States and Canada). Paediatr Perinat Epidemiol. 2006;20:110–118.
14. Hartge P. Participation in population studies. Epidemiology. 2006;17:252–254.
15. Hatch EE, Kleinerman RA, Linet MS, et al. Do confounding or selection factors of residential wiring codes and magnetic fields distort findings of electromagnetic fields studies? Epidemiology. 2000;11:189–198.
© 2008 Lippincott Williams & Wilkins, Inc.