Indicators and designs in genetic epidemiology: separating heredity and environment

The problems of separating highly confounded hereditary and environmental factors are reviewed from an epidemiological point of view. Once the fact rather than the appearance of familial aggregation is established, one can search for and analyze strong indicators of either genetic or environmental effects, and ultimately attempt their separation by strong design. Indicators of genetic effects are classified according to the presence or absence of family recurrence. In the presence of family recurrence, three analytic approaches are available: segregation analysis, linkage, and heritability estimates. In the absence of family recurrence, biological markers and endogamous groups can be used. Indicators of environmental effects are also classified according to the presence or absence of family recurrence of a disorder. In the presence of family recurrence, three environmental indicators are considered: non-Mendelian clustering, such as time of onset versus age in family clusters, and sex clusters; cohabitational effects; and maternal transmission. In the absence of family recurrence, environmental indicators include secular trends, migration, time and place clusters, family size, and birth order. Designs that aim to separate heredity and environment include twin studies, fixed clusters that include a variety of degrees of relatedness (particularly the family set-design), and comparisons of separately reared relatives. The strengths and weaknesses of twin studies and family set designs are reviewed. Separately reared relatives provide the most cogent tests of genetic and environmental hypotheses. Among these, separated twin pairs, half-sibs, and relatives separated through institutional placement or adoption are considered. Adoption strategies are illustrated by the Danish adoption studies of schizophrenia, criminality, and alcoholism, and these studies are reanalyzed from the perspective of epidemiological case-control and cohort studies. Finally, the potential uses of multiple family settings as they occur in the general population are briefly discussed.