diseases involving immune or inflammatory components. For example, catecholamines are known to influence several aspects of leukocyte function (Sanders and Straub, 2002; Kavelaars, 2002), and polymorphisms in genes encoding their alpha—and beta—adrenergic receptors are associated with differential incidence of asthma, parasitic infections, and cardiovascular disease (Ramsay et al., 1999; Ulbrecht et al., 2000; Ukkola et al., 2001; Weiss, 2005; Thakkinstian et al., 2005; Lanfear et al., 2005). Glucocorticoids, another physiological system exquisitely sensitive to the psychosocial environment, play a key role in regulating inflammatory gene expression (Webster et al., 2002), and polymorphisms in the glucocorticoid receptor gene (NR3C1) have been linked to cardiovascular and autoimmune disease (Lin et al., 1999; Ukkola et al., 2001; Jiang et al., 2001; Dobson et al., 2001; van Rossum et al., 2002; Lin et al., 2003).
Mediating and moderating variables often are inferred in human studies through multivariate statistical analysis (Baron and Kenny, 1986). A moderating variable is one that changes the way an independent variable is related to a dependent variable (e.g., sex differences in the relationship between reported symptoms and risk for cardiac disease). A mediating variable is one that statistically accounts for the association between an independent and dependent variable in a study (e.g., cortisol levels may be a better predictor of disease onset than feelings of stress). However, the disease process may be mediated by autonomic tone, not measured in the study, and not cortisol itself.
In the animal literature, however, these terms have more stringent criteria. Studies demonstrate “mediation” only when a hypothesized intermediate factor has been experimentally manipulated to block the effects of some upstream influence on a downstream outcome within a transitive causal chain. “Moderation” is reserved for cases in which one variable is experimentally manipulated to alter the causal effect of a second manipulated variable on an observed outcome. A statistical interaction is not sufficient. The strongest evidence for genetic moderation comes from studies in which both genes and environment are experimentally manipulated, but few studies meet this criterion. Based on this fact alone, it can be concluded that much remains to be learned about the interaction between genes and the social environment in the context of immune system function and disease.
Behavioral ecologists have elegantly and dramatically revealed that we cannot expect human studies to reveal monolithic or simple linear relation-