Peter T. Ellison and Emily S. Barrett
Life historical approaches to disease are attracting increasing attention within both epidemiology and evolutionary biology. The concept of life history refers to the interactions and linkages between the states of an organism at different points in its life cycle. Although an appreciation for these linkages has long been implicit within both epidemiology and evolutionary biology, the development of empirical approaches on the one hand and theoretical approaches on the other has presented formidable difficulties in data collection, analysis, and modeling. As these difficulties have been overcome, new data, theory, and speculation have rapidly accumulated.
The development of life historical approaches within epidemiology was made possible by systems of record linkages. These allow researchers to join information on disease morbidity to information on reproductive history, pre- and perinatal conditions, and familial variables available from birth records and other sources. With these linked data sets comes the ability to generate longitudinal data for individuals. Replacing cross-sectional data with linked, longitudinal data can produce very different observations and conclusions. For example, Bakketeig and Hoffman (1981) were able to demonstrate using Norwegian data that the apparent U-shaped relationship between parity and frequency of pre-term birth is an artifact of cross-sectional analysis. Substituting longitudinal data and grouping by completed family size demonstrated instead that the frequency of pre-term birth declines monotonically with parity for all levels of achieved fertility (Figure 2.1). Grouping individuals by completed family size is obviously only possible when information from later in the life cycle can be linked to the incidence of pre-term birth at an earlier stage in the life cycle.
The principal use of life historical data within epidemiology is not to correct the distortions of cross-sectional analysis, however, but to help discriminate the relative importance of genetic, developmental, and acute environmental sources of variance in morbidity and disease outcomes. In the absence of longitudinal data, only concurrent states of the organism and its environment can be analyzed as sources of this variance. This can lead to a confounding of static variables (e.g. growth status, nutritional status, reproductive status) with dynamic variables (e.g. rate of growth, energy balance, reproductive history), and so to an underestimate