Estimation of health effects of prenatal methylmercury exposure using structural equation models

BackgroundObservational studies in epidemiology always involve concerns regarding validity, especially measurement error, confounding, missing data, and other problems that may affect the study outcomes. Widely used standard statistical techniques, such as multiple regression analysis, may to some extent adjust for these shortcomings. However, structural equations may incorporate most of these considerations, thereby providing overall adjusted estimations of associations. This approach was used in a large epidemiological data set from a prospective study of developmental methyl-mercury toxicity.ResultsStructural equation models were developed for assessment of the association between biomarkers of prenatal mercury exposure and neuropsychological test scores in 7 year old children. Eleven neurobehavioral outcomes were grouped into motor function and verbally mediated function. Adjustment for local dependence and item bias was necessary for a satisfactory fit of the model, but had little impact on the estimated mercury effects. The mercury effect on the two latent neurobehavioral functions was similar to the strongest effects seen for individual test scores of motor function and verbal skills. Adjustment for contaminant exposure to poly chlorinated biphenyls (PCBs) changed the estimates only marginally, but the mercury effect could be reduced to non-significance by assuming a large measurement error for the PCB biomarker.ConclusionsThe structural equation analysis allows correction for measurement error in exposure variables, incorporation of multiple outcomes and incomplete cases. This approach therefore deserves to be applied more frequently in the analysis of complex epidemiological data sets.

[1]  D. Ruppert,et al.  Measurement Error in Nonlinear Models , 1995 .

[2]  M. Browne,et al.  Alternative Ways of Assessing Model Fit , 1992 .

[3]  Roberta F. White,et al.  Neurobehavioral deficits associated with PCB in 7-year-old children prenatally exposed to seafood neurotoxicants. , 2001, Neurotoxicology and teratology.

[4]  K J Roghmann,et al.  Pathways of lead exposure in urban children. , 1997, Environmental research.

[5]  A. Satorra,et al.  Corrections to test statistics and standard errors in covariance structure analysis. , 1994 .

[6]  S. Griffin,et al.  Calculating the interindividual geometric standard deviation for use in the integrated exposure uptake biokinetic model for lead in children. , 1999, Environmental health perspectives.

[7]  Ronald Schoenberg,et al.  Pseudo maximum likelihood estimation and a test for misspecification in mean and covariance structure models , 1989 .

[8]  P. Grandjean,et al.  Validity of mercury exposure biomarkers , 2002 .

[9]  Nicole A. Lazar,et al.  Statistical Analysis With Missing Data , 2003, Technometrics.

[10]  B. Muthén,et al.  Robust inference using weighted least squares and quadratic estimating equations in latent variable modeling with categorical and continuous outcomes , 1997 .

[11]  D. Rubin,et al.  Statistical Analysis with Missing Data. , 1989 .

[12]  P Grandjean,et al.  Impact of maternal seafood diet on fetal exposure to mercury, selenium, and lead. , 1992, Archives of environmental health.

[13]  K. Jöreskog.  A General Method for Estimating a Linear Structural Equation System. , 1970 .

[14]  Roberta F. White,et al.  Cognitive deficit in 7-year-old children with prenatal exposure to methylmercury. , 1997, Neurotoxicology and teratology.

[15]  A. Satorra,et al.  Measurement Error Models , 1988 .

[16]  N. Keiding,et al.  Methylmercury exposure biomarkers as indicators of neurotoxicity in children aged 7 years. , 1999, American journal of epidemiology.

[17]  H. Duivenvoorden,et al.  The relative importance of prenatal PCB exposure, feeding type, and parental characteristics for cognitive and motor development in healthy children studied from 3 to 84 months of age , 2000 .

[18]  B. Muthén.  A general structural equation model with dichotomous, ordered categorical, and continuous latent variable indicators , 1984 .

[19]  Roberta F. White,et al.  Consequences of exposure measurement error for confounder identification in environmental epidemiology , 2003, Statistics in medicine.

[20]  B. Muthén.  Latent variable modeling in heterogeneous populations , 1989 .