Skip to main content

Multivariate analyses of joint effects of prenatal alcohol exposure and socioeconomic factors on a battery of neurocognitive tests at 21 years of age

This talk addresses some statistical challenges in a study of the relationship between prenatal alcohol exposure and neurocognitive functioning in a sample of 419 Seattle-area 21-year-olds who had been examined systematically in respect of alcohol behavioral teratology since midway through their mothers' pregnancies in 1974-75. A five-hour laboratory examination of 17 neurocognitive tests administered at a mean age of 21.5 years yielded a total of 164 outcome scores. Prenatal alcohol exposure was quantified by a previously published combination of thirteen measures of self- reported maternal alcohol use optimally predicting offspring deficits at earlier ages.

A concern for possible effects of loss-to-follow-up in this longitudinal study let to our study of outcomes in terms of their joint prediction by prenatal alcohol and occupational status of the household (mid-pregnancy). It has been convincingly argued that analyses of neuropsychological outcomes in a context like this should adjust for parental IQ. In the absence of measures of parental IQ we chose to adjust all outcomes for offspring IQ in accord with a path analytic argument. Patterns of relationship of IQ-adjusted outcomes with prenatal alcohol and occupational status were analyzed by principal components analysis and summarization in terms of odds ratios of extreme outcomes. Our results identify one nonlinear bivariate prediction function that represents certain neurocoginitive deficits only at lower occupational levels and only at the highest levels of exposure. A Spatial-Visual Reasoning Task (SVRT) demonstrated the clearest alcohol-related deficits in this context and the result of this multivariate analysis is well- summarized by an odds ratio of deficit on one single speed-accuracy tradeoff score.

The confounding of socioeconomic factors with the effect of prenatal alcohol exposure on neurocognitive functioning in this sample of adults cannot be resolved by the usual methods of multiple regression. Models must consider the effects of alcohol and occupational status jointly, rather than "adjusting" the effect of either for the other. The surprisingly focused finding of an effect of prenatal alcohol exposure on spatial-visual reasoning, a test not in standard neuropsychological assess, emt batteries, suggests further study of this task as a teratogenic outcome.