ABSTRACT
Assessments of the import of an association observed between an environmental chemical exposure and child neurodevelopment often focus solely on the magnitude of the effect size and its associated p-value (i.e., whether it is < 0.05). Effect size is expressed in various forms, as the difference between the mean scores of exposed and unexposed groups, the change in score per unit change in an exposure biomarker, or the change in risk (relative risk, odds ratio) associated with a particular value of the biomarker. Among the reasons cited to dismiss an effect size is that it is clinically unimportant (e.g., Kaufman 2001). This perspective fails to place the effect estimate in a public health context, however. Estimating the population burden attributable to a factor requires a metric that reflects not only the magnitude of the risk associated with the factor but also the frequency
with which the factor occurs in the population (Steenland and Armstrong 2006), a concept embodied in the environmentally attributable fraction model (Institute of Medicine 1981). Although a factor associated with a large impact would be a significant burden to a patient, it might not be a major contributor to population burden if it occurs rarely. Conversely, a factor associated with a modest but frequently occurring impact could contribute substantially to population burden.
