FINAL REPORT: Validation of Test Methods for Assessing Neurodevelopment in Children1
Investigations of adverse health effects in developing organisms that are attributable to neurotoxicant exposure require tests that accurately measure behavioral change across a wide range of exposure levels. Developmental consequences of exposures to high dosages of neurotoxicants are almost always severe and usually affect many domains of developmental function. Assessing such effects usually can be accomplished using standard tests of neurological and general cognitive ability. Exposures at lower dosages however, are usually expected to cause less severe, subtler effects on development, effects that may or may not be global, but may be detectable only in specific domains.
There is a considerable literature on approaches to assessing neurotoxicity (e.g., Fiedler, et al., 1996; Jacobson & Jacobson, 1996) and various strategies have been examined for their efficacy in identifying neurobehavioral effects, including paper and pencil tests and computerized batteries (cf. Williamson, 1996).
Much of the work in this field has focused on adults whose exposures took place in the workplace, rather than on children whose exposures were developmental. For example, the Neurobehavioral Evaluation System (NES) developed by Baker & Letz (1985), is a well-known computerized software system was designed for field use with adults. There are few reports of its use with children (see Otto, et al., 1996). Work specifically initiated for evaluating children has included development of the Pediatric Environmental Neurobehavioral Test Battery (PENTB), a screening battery for field use (Amler & Gilbertini, 1996). A more extensive assessment protocol was proposed by Dahl, et al., 1996). The general approach in developing these tools, whether for children or adults, began with the assumption that deficient performance on tests resulted from the neurotoxicant in question and the test, in turn, was assumed to be accurately measuring the effect of that specific neurotoxicant. This approach is not consistent with basic psychometric principles; it implies attribution of a predicted effect without evidence that the effect can actually be distinguished from a non-effect by the endpoint measure. Although the approach is widely employed and may be acceptable when gross effects are being assessed, it is questionable when exposures occur at low levels and in situations in which evidence of subtle effects is sought. Absence of effects cannot be interpreted as an indication of no risk under such circumstances.
An alternative strategy would be to select a battery of tests based on their sensitivity and specificity in predicting test performance decrements assumed to be caused by exposure to a neurotoxicant. Reliance on sensitivity and specificity has not occurred often in human behavioral neurotoxicological research; they have been recommended as the best parameters to gauge the capability of a test or task to detect a rare event (cf. Hrudey & Leiss, 2003). Sensitivity is defined as the conditional probability of detecting a true positive while specificity is the conditional probability of detecting a true negative (Hrudey & Leiss, 2003). True (actual) risk must be known in order to determine both of these parameters. Likewise, the behaviors measured should be known to reflect true risk. Such data are generally not available for human application. White and colleagues (1994) underscored the importance of considering both sensitivity and specificity in test selection for neurotoxicogical studies. Fiedler (1996) points out more refined studies of sensitive and specific behavioral measures are needed.
Sources of information for selecting candidate measures include experimental data from animal studies and clinical or epidemiological studies in humans. Such data can provide guides to functional endpoints that may not have been studied in depth, and provide more comprehensive biological measures of exposure in target tissues, complemented by morphological indices. A second source consists of data from children with pre-diagnosed developmental disabilities resulting from CNS damage that can serve as a reference population to validate the battery.
The strategy followed in the present study was to blend these two approaches, using animal data to develop, refine and extend the scope of tests suitable for humans, and then assess their sensitivity and specificity in populations already identified to be at risk for the presumed deficits. The resulting data could then be used to assist investigators and assessors in selecting specific tests and tasks with empirically determined detection capacities for use in neurotoxicity research and risk assessment.
The aim of the present study was to evaluate the sensitivity and specificity of a battery of neurodevelopmental tests for detection of outcomes representing the types of subtle neurodevelopmental deficits in children caused by exposure to common developmental neurotoxicants. These projected tests fell into three categories:
- Neuropsychological tests with already established psychometric properties not previously used to study developmental neurotoxicity;
- Electrophysiological and behavioral tests of sensory and motor functioning spanning a broader range of indices than those used in other investigations of children exposed to developmental neurotoxicants; and
- Adaptations of performance tasks used previously only in animals or not yet applied to assess developmental neurotoxicity in children.
The project studied children at risk for the kinds of performance deficits these tests endeavor to measure, but who themselves had no known exposure to neurotoxicants: Neonatal Intensive Care Unit (NICU) graduates, known to be at risk for both major and mild anomalies in perception, motor functioning, learning, memory and cognition (Hack, et al., 1994; Saigal, et al., 1991). The intent was to develop a battery comprised of tests and tasks that predicted these deficits. Those tests and tasks that demonstrated a high probability of predicting risk status were incorporated into a final battery that might serve as a second tier to instruments such as the PENTB, and which might prove useful both in field studies and research. While different neurotoxicants may have different mechanisms of insult and therefore, different signature effects on behavior in children, a battery designed to encompass a broad spectrum of functions may be useful for assessing the developmental neurotoxicity of exposure to a range of chemicals hypothesized to cause compromised neurodevelopment.