Bayesian Methods for Correcting Misclassification: An Example from Birth Defects Epidemiology

MacLehose, Richard F.a,b; Olshan, Andrew F.c,d; Herring, Amy H.d,e; Honein, Margaret A.f; Shaw, Gary M.g; Romitti, Paul A.h; the National Birth Defects Prevention Study

doi: 10.1097/EDE.0b013e31818ab3b0
Methods: Original Article

Background: Cleft lip with or without cleft palate (CL/P) and cleft palate only (CPO) are common congenital malformations. Numerous epidemiologic studies have shown an increased risk for orofacial clefts among children whose mothers smoked during early pregnancy; however, there is concern that the results of these studies may have been biased because of exposure misclassification. The purpose of this study is to use previous research on the reliability of self-reported cigarette smoking to produce corrected point estimates (and associated credible intervals) of the effect of maternal smoking on children's risk of clefts.

Methods: We accounted for misclassification using 4 Bayesian models that made different assumptions about the sensitivity and specificity of self-reported maternal smoking data. We used results from previous studies to specify the prior distributions for sensitivity and specificity of reporting and used Markov chain Monte Carlo algorithms to calculate the posterior distribution of the effect of maternal smoking on children's risk for CL/P and CPO.

Results: After correcting for potential sources of misclassification in data from the National Birth Defects Prevention Study, we found an increased risk of CL/P among children born to mothers who smoked during early pregnancy (posterior odds ratio [OR] = 1.6, 95% credible interval = 1.1–2.2). The posterior effect of smoking on CPO provided less evidence of effect (posterior OR = 1.1, 95% credible interval = 0.7–1.7).

Conclusion: Our results lend some credibility to the hypothesis that periconceptional maternal smoking increases the risk of a child being born with CL/P. The results concerning CPO provide no overall evidence of effect, although the estimates were relatively imprecise. We suggest that future research should emphasize validity studies, especially those of differential reporting, rather than replicating existing analyses of the relationship between maternal smoking and clefts. We discuss how our approach is also applicable to evaluating misclassification in a wide range of exposure-outcome scenarios.

From the aDivision of Biostatistics, University of Minnesota, Minneapolis, MN; bDivision of Epidemiology and Community Health, University of Minnesota, Minneapolis, MN; cDepartment of Epidemiology, University of North Carolina School of Public Health, Chapel Hill, NC; dCarolina Population Center, University of North Carolina, Chapel Hill, NC; eDepartment of Biostatistics, University of North Carolina School of Public Health, Chapel Hill, NC; fNational Center on Birth Defects and Developmental Disabilities, Centers for Disease Control and Prevention, Atlanta, GA; gMarch of Dimes, California Research Division, Oakland, CA; and hDepartment of Epidemiology, University of Iowa College of Public Health, Iowa City, IA.

Submitted 14 November 2007; accepted 29 April 2008.

The first author completed much of this work while in the Biostatistics Branch at the National Institute of Environmental Health Sciences, Research Triangle Park, NC.

The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the Centers for Disease Control and Prevention.

Supported in part by the Intramural Research Program of the NIH, National Institute of Environmental Health Sciences and by grants from the Centers for Disease Control and Prevention (U50/CCU422096) and the National Institute of Environmental Health Sciences (P30ES10126).

Supplemental material for this article is available with the online version of the journal at www.epidem.com; click on “Article Plus.”

Correspondence: Richard F. MacLehose, PhD, 1300 S. 2nd St, Suite 300, Minneapolis, MN 55454. E-mail: mac10029@umn.edu.

© 2009 Lippincott Williams & Wilkins, Inc.