A New Approach to Design and Data Analysis for CPP-Based Studies that Make Use of Sera, and Experience with Recent Laboratory Analysis of CPP Sera

A NEW APPROACH TO DESIGN AND DATA ANALYSIS

FOR CPP-BASED STUDIES THAT MAKE USE OF SERA, AND EXPERIENCE

WITH RECENT LABORATORY ANALYSIS OF CCP SERA

Matthew P.

Longnecker*, John Brock, Haibo Zhou, Mark Klebanoff.

Epidemiology Branch, National Institute of Environmental

Health Sciences, NIH National Center for Environmental

Health, CDC, Department of Biostatistics, University of North

Carolina, Chapel Hill, National Institute of Child Health and

Human Development

Using the Collaborative

Perinatal Project (CPP) data set and serum repository, the

authors are testing a number of hypotheses about health

effects of in utero exposure to background levels of

persistent environmental contaminants. For example, maternal

serum level of polychlorinated biphenyls is being used to

estimate in utero exposure to these suspected

neurotoxins. Outcomes of interest are findings on neonatal

neurologic exam, Bayley Scale of Infant Development, and WISC

IQ.

In addition to making use of

the relatively new case-base study design, we are also

developing statistical methods that allow us to use an

efficient sampling scheme for continuous outcomes. For

example, in studying IQ as an outcome, besides analyzing

blood from a random sample of the cohort, we are also

analyzing specimens from a random sample of subjects whose IQ

is either above or below one standard deviation of average.

For a given total number of specimens analyzed this

“oversampling” of subjects in the tails of the

outcome distribution increases our power to detect

associations relative to simple random sampling alone. The

gain in statistical efficiency is marked when the relation

under study is not completely linear. A proper statistical

method for analyzing data sampled in this manner is being

developed and will be discussed; ignoring this type of

sampling in the analysis will cause bias in parameter

estimation.

To work with these older

serum specimens, we have had to modify the laboratory

analytical techniques to work around the frequent finding of

precipitates in the stored specimens. In addition, we will

report the findings from analysis of the specimens for

sodium, which reflects the amount of desiccation.