up on previously surveyed samples, such as those for other National Longitudinal Surveys, on young women, mature women, young men, and older men, even though some of the surveys have been previously discontinued. Two other longitudinal studies—the Coronary Artery Risk Development in Young Adults study and the Atherosclerosis Risk in Communities study—include, by design, samples of whites and black, and may therefore be particularly useful for looking at racial differences.
Short of cohort studies, various other approaches can be taken to analyze life course influences. Surveys of adult health could be enhanced with information about childhood experiences. Retrospective questions could be used, though they are subject to recall bias, and responses can be distorted because of current status. Data can sometimes be obtained from administrative records matched with individual survey records. For instance, birth records contain information on birthweight that could be added to a number of ongoing surveys. Similarly, school records may contain information of past illnesses or health-related events (prolonged absences, personality reports, etc.). Test scores and examination results could also be useful to roughly characterize the school environment of survey respondents. A further possible approach is to extend existing data sets by linking them with other data sources. All these approaches require interdisciplinary teams and close collaboration among funding agencies, some of which may support work on early childhood and others on adult health and mortality. There are also obviously difficult issues of methodology and data quality, problems with releasing sufficient data to allow matching, and issues of confidentiality that must be attended to in each case.
Studies could also focus on specific stages of the life course to document the operation of mechanisms that are hypothesized to mediate the effects of early experiences on adult health. If deprivation alters key physiological parameters during early childhood, it should be possible to verify this effect without a longitudinal study. Uncovering such effects is crucial to any argument that early experiences affect adult health through physiological programming or imprinting. Studies of these mechanisms may require invasive procedures (such as extensive and expensive blood tests, magnetic resonance imaging, etc.) but could be completed relatively quickly. For instance, studies could focus on metabolic variables among younger individuals exposed or not exposed to known sources of stress to determine allostatic load and to assess whether physiological plasticity of response can be lost at early ages, and, if so, with what consequences.
Studies that combine data sets covering different stages in the life course would be useful to draw inferences across life stages. With properly designed estimation (and simulation) procedures, composite data sets would allow reasonable inferences about a range of early exposure effects and about their importance relative to contemporaneous determinants of health.