Social Support and Health: A Review of Physiological Processes Potentially Underlying Links to Disease Outcomes
Social support has been reliably related to lower rates of morbidity and mortality. An important issue concerns the physiological mechanisms by which support influences such health endpoints. In this review, I examine evidence linking social support to changes in cardiovascular, neuroendocrine, and immune function. Consistent with epidemiological evidence, social support appears to be related to more positive “biological profiles” across these disease-relevant systems. Recent research on immune-mediated inflammatory processes is also starting to provide data on more integrative physiological mechanisms potentially linking social support to health. The implications of these links, along with future research directions are discussed.
KEY WORDS:social supportphysical healthcardiovascular functionneuroendocrine functionimmune functioninflammation.
“Adequate tests of the hypothesis that social circumstances alter general susceptibility to disease in humans will not be possible, however, until data are available on physiologic mechanisms capable of mediating the relationship between social events and disease outcomes.” (Berkman and Syme, 1979, p. 203).
In1979, Berkman and Syme published the results of their seminal study linking social relationships to mortality. These researchers linked questions about the extent of peoples’ social connections to overall mortality and found that people who were less socially integrated had higher mortality rates. This paper was influential because it was able to rule out possible alternative explanations (e.g., results due to poorer initial health status) and hence provided the most compelling empirical links at the time between social relationships and mortality. Subsequent research has confirmed the reliable links between social support and better physical health outcomes (see reviews by Berkman et al., 2000; Cohen, 1988; House et al., 1988; Seeman, 1996; Uchino, 2004). Epidemiological studies indicate that individuals with low levels of social support have higher mortality rates; especially from cardiovascular disease (Berkman et al., 1992; Brummett et al.,2001; Frasure-Smith et al.,2000; Kaplan et al., 1988; Orth-Gomer et al., 1993; Rutledge et al., 2004; Williams et al., 1992). However, there is also preliminary evidence linking support to lower cancer (Ell et al., 1992; Hibbard and Pope, 1993; Welin et al., 1992), and infectious disease (Lee and Rotheram-Borus, 2001; Patterson et al., 1996) mortality.
Studies linking relationships to disease mortality represent the first wave of research on social support and health. The second wave must address a question of critical importance: what are the mechanisms by which social support influences such health outcomes? As noted by Berkman and Syme (1979), a mechanism of particular importance concerns the biological pathways potentially linking social support to health outcomes. Such data are important in efforts to (a) test the biological plausibility of the proposed links, (b) refine existing conceptual models, and (c) design appropriate interventions that consider the complexity of these links.
In this review, I will outline the main physiological pathways potentially linking social support to physical health. Social support is usually defined to include both the structure of an individual's social life (e.g., group memberships, existence of familial ties) and the more explicit functions they may serve (e.g., emotional support; see Cohen et al., 2000 for a review). I will focus on research examining these measures of social support to cardiovascular, neuroendocrine, and immune function as these are the major biological systems thought to influence disease risk. In order to place this research in its proper context, I first review major pathways by which social support may influence health at different levels of analysis.
How Does Social Support Influence Physical Health Outcomes?
The other major pathway involves psychological processes that are linked to appraisals, emotions or moods (e.g., depression), and feelings of control (Cohen, 1988; Gore, 1981; Lin, 1986). There is strong evidence linking social support to these psychological processes (Barrera, 2000), although direct evidence for their mediational role on health outcomes is lacking (House, 2001). Note that the behavioral and psychological levels are linked as each has been shown to exert an influence over the other. For instance, feelings of stress can adversely impact the practice of health behaviors (e.g., Ng and Jeffery, 2003) while health behaviors such as exercise can have beneficial effects on feelings of stress (e.g., Rejeski et al., 1992). Finally, these psychological and behavioral pathways may have reciprocal influence on social support processes. For instance, psychological distress may influence perceptions of support and contribute to negative social interactions (Alferi et al., 2001; Coyne, 1976).
An additionally important aspect of the model concerns the proposed links to and from disease morbidity. This makes salient two aspects of this broad model. First, the links with morbidity highlight the potential role of social support in the development of certain diseases. Second, the feedback loop between morbidity and social support highlights the unique challenges faced by individuals diagnosed with disease that can impact their social network. Close network members are often called upon as sources of support after the diagnosis of disease (Bolger et al., 1996). In some cases, the strain may result in network members withdrawing in an attempt to cope with an overwhelming situation (Bolger et al., 1996). In other cases, stressful situations can provide an opportunity for growth in personal relationships (Holahan and Moos, 1990).
Of central importance to this review is that the links between social support and disease are hypothesized to be mediated through relevant physiological processes; including changes in cardiovascular, neuroendocrine, and immune function (Berkman et al., 2000; Uchino et al., 1996). I first start my review by examining the links between social support and health-relevant cardiovascular alterations, followed by relevant research on neuroendocrine and immune function. Finally, a focus on more integrative biological mechanisms is outlined by highlighting recent research on immune-mediated inflammatory processes.
Social Support and Cardiovascular Function
Most of the evidence linking social support to biological pathways has examined the cardiovascular system (Uchino et al., 1996). There is strong evidence linking social support to aspects of cardiovascular function that may confer lower risk for disease (Uchino, 2004). One of these paradigms includes conceptual links to the reactivity hypothesis of disease (Lepore, 1998). According to this perspective, individuals or situations characterized by high levels of cardiovascular reactivity (usually indexed by blood pressure or heart rate) may be related to higher risk for the development and exacerbation of cardiovascular disease. Research linking reactivity to cardiovascular disease is still on-going, but current evidence is consistent with this hypothesis (Treiber et al., 2003). Social support may thus be beneficial because it “buffers” the potentially harmful influences of stress-induced cardiovascular reactivity (Cohen and Wills, 1985).
In an early study testing this hypothesis, Kamarck et al. (1990) found that the presence of a friend was associated with lower blood pressure reactivity to an acute psychological stressor (i.e., mental arithmetic). Subsequent research has revealed complex associations as a function of such mere presence paradigms (Uchino et al., 1996). For instance, the simple presence of a friend may increase evaluation apprehension and interfere with support processes (Allen et al., 1991). Studies that directly manipulate the supportive function of relationships through verbal support appear to provide more consistent evidence for this stress-buffering hypothesis (Gerin et al., 1992; Lepore, 1995; Thorsteinnsson et al., 1998; see review by Thorsteinnsson and James, 1999).
There are several important assumptions in these laboratory reactivity studies of social support that deserve mention in predicting the development of cardiovascular disease. These studies assume that this buffering process would occur outside of the laboratory over relatively long periods of time, to reduce the potentially harmful increases in cardiovascular reactivity associated with stressful life events. I know of no such longitudinal test of these associations. However, if these assumptions are correct, one would expect for the effects of social support to be more evident in older individuals due to the cumulative influence of social support across the years. To date, we have provided evidence in several studies showing that social support is associated with lower resting blood pressure primarily in older adults (Uchino et al., 1995; Uchino et al., 1999; also see Ong and Allaire, 2005).
Social support may also be beneficial because it is associated with lower blood pressure during everyday life. This would be an important conceptual link because studies examining the prognostic value of ambulatory blood pressure suggest that it predicts future cardiovascular problems above and beyond that predicted by conventional (resting) blood pressure readings (e.g., Perloff et al., 1983). Existing studies are consistent with a link between social support and lower ambulatory blood pressure (Gump et al., 2001; Linden et al., 1993; Spitzer et al., 1992; Steptoe et al., 2000). In one illustrative study, Steptoe and colleagues (2000) found that when the role of being a parent was combined with high levels of functional support it predicted the greatest evening reduction in ambulatory systolic blood pressure compared to any of these measures in isolation. These ambulatory studies are consistent with the role of social support as a risk-reducing factor in the development of cardiovascular disease (also see Horsten et al., 1999). In future research, ambulatory protocols might also be designed to model the stress-buffering effects of support previously examined in laboratory reactivity paradigms.
More direct evidence for a role of social support on the development of cardiovascular disease can be found in recent data utilizing imaging techniques. To date, several of these studies have shown that social support predicts less underlying atherosclerosis (Angerer et al., 2000; Knox et al., 2000; Seeman and Syme, 1987; Kop et al., 2005; Wang et al., 2005). For instance, the National Heart, Lung, and Blood Institute Family Heart Study examined links between social support and carotid artery atherosclerosis using B-mode ultrasonic imaging (Knox et al., 2000). Results of this study revealed that social support was related to less underlying atherosclerosis in women at high risk for the development of heart disease.
It is also important to note that social support appears to play a role in the progression of diagnosed cardiovascular disease (Berkman et al., 1992; Brummett et al., 2001). One potential mechanism for this action is through the stress-buffering properties of social support. Acute stress in cardiac patients may precipitate ischemia due to hemodynamic changes in blood pressure (Rozanski et al., 1988). Thus, social support may be beneficial to cardiac patients because it decreases deleterious cardiovascular changes during stress. Imaging studies are also consistent with the possibility that social support influences the progression of clinical cardiovascular disease. Women with established coronary artery disease who were low in emotional support had faster disease progression as indexed by angiography over a subsequent three year period (Wang et al., 2005; also see Angerer et al., 2000). However, a stronger test of the above physiological pathways (as well as those detailed below) would require conceptual modeling to test if these parameters directly mediated epidemiological links between social support and disease outcomes.
Social Support and Neuroendocrine Function
In comparison to cardiovascular function, there is relatively little research linking social support to neuroendocrine function (Uchino et al., 1996). As a result, conclusions regarding such links should be considered tentative. The relative lack of such data is noteworthy because hormones and neurotransmitters mediate aspects of cardiovascular and immune function (Ader et al., 2001) and thus may shed light on how these diverse physiological systems are coordinated as a function of social support. There is some evidence that social support is associated with lower plasma and urinary catecholamine levels (Fleming et al., 1982; Seeman et al., 1994; Grewen et al., 2005). These associations are consistent with the beneficial effects of social support on cardiovascular function.
Another important hormone implicated in disease processes is cortisol which has well-documented immunosuppressive effects (Greenspan and Baxter, 1994). Potential associations between social support and cortisol may be especially important for cancer patients as the immune system is thought to be the primary mechanism of defense against malignant diseases (Dunn et al., 2002). In our earlier review, we found little evidence for a link between social support and cortisol levels (Uchino et al., 1996). Many of these early studies utilized only one plasma cortisol assessment (usually in the morning) so we speculated that advances in the measurement of cortisol may be needed to detect such an association. Turner-Cobb and colleagues (2000) utilized a more comprehensive protocol by measuring salivary cortisol over the course of three days in breast cancer patients. Results showed that social support predicted lower overall cortisol levels (Turner-Cobb et al., 2000; also see Heinrichs et al., 2003). In general, studies that have measured salivary cortisol over several time points have found more consistent links to social support (Heinrichs et al., 2003; Milagros et al., 2004; Turner-Cobb et al., 2000; but see Olga and Steptoe, 2001).
There are additional neuroendocrine processes that require increased attention in social support studies. For instance, opioids have links to both the cardiovascular and immune systems (Carr et al., 1996; McCubbin et al., 1993) although very little research has examined potential associations with social support. Another neuroendocrine hormone with promising and largely unexplored links to social support is oxytocin (Knox and Uvnas-Moberg, 1998; Taylor et al., 2000). Besides its more established physiologic role (e.g., uterine contractions during labor), oxytocin has anti-stress effects in both the brain and more peripheral physiological systems (Taylor et al., 2000). For instance, oxytocin release is associated with decreases in cortisol levels, blood pressure, sympathetic activity, and increases in parasympathetic activity (Uvnas-Moberg, 1998).
At present little research exists on oxytocin responses in humans and its potential link to social support. In one of the few human studies, perceptions of partner support were uniformly associated with higher oxytocin levels (Gwenen et al., 2005). It is also possible that oxytocin may be a primary pathway regulating the stress-buffering effects of social support on other physiological systems (Knox and Uvnas-Moberg, 1998). This hypothesis was tested by Heinrichs and colleagues (2003) who manipulated social support (via a friend) and oxytocin levels (via a nasal spray) in men undergoing acute psychological stress. Consistent with the stress-buffering hypothesis, social support was associated with lower cortisol responses. However, these support effects were especially evident if combined with the oxytocin manipulation as such individuals showed the smallest increases in cortisol during stress. These studies suggest that additional research on this and related neuroendocrine hormones may provide novel insight into biologic processes mediating social support effects.
Social Support and Immune Function
A final physiological pathway by which social support may influence physical health is via the immune system. Indeed, some of the strongest associations we found in our earlier meta-analytic review were for a link between social support and better immune function, especially in older adults (Uchino et al., 1996). This association is particularly noteworthy because the risk of cancer increases with age and infectious diseases are a leading cause of death in older individuals (Effros and Walford, 1987).
A number of recent studies continue to suggest that social support is related to better immune function (Dixon et al., 2001; Esterling et al., 1996; Lutgendorf et al., 2005; Miyazaki et al., 2005). Of particular note are studies that have examined the link between social support and natural killer cell activity (e.g., Esterling et al., 1996; Lutgendorf et al., 2005) because natural killer cells are thought to play an important “surveillance” role in cancer (Abbas and Lichtman, 2003). This association was tested by Lutgendorf and colleagues (2005) in ovarian cancer patients. Replicating prior research (Levy et al., 1990), results showed that social support predicted greater natural killer cell activity in peripheral blood. Importantly, natural killer cell activity in tumor-infiltrating lymphocytes was also greater for individuals with high levels of social support.
Another patient population of importance given links between social support and immune function is HIV+ individuals. A few studies have reported an association between social support and helper T-cell counts in HIV+ men (Persson et al., 1994; Theorell et al., 1995). In one long-term study, researchers examined this link across a five year period (Theorell et al., 1995). Results showed that social support became a more powerful predictor of helper T-cells as time progressed (years four and five). For instance, during year five of the study, individuals with high social support showed a −37% change in helper T-cells compared to a −64% change in helper T-cells for individuals low in support. However, several studies have failed to find an association between social support and aspects of immune function in HIV+ men (Goodkin et al., 1992; Perry et al., 1992). The state of this literature makes firm conclusions difficult, but suggests potential moderators may need greater attention in future studies.
It is also important to note that existing research is consistent with the biological significance of the links between social support and immunity. For instance, individuals lower in support were less likely to clinically seroconvert to a Hepatitus B vaccination (Glaser et al., 1992). Social support has also been correlated with antibody titers following influenza vaccination (Moynihan et al., 2004; Pressman et al., 2005). These results are consistent with data by Cohen et al. (1997) who found that individuals with more diverse social networks (i.e., relationships from a variety of domains such as work, home, church) were less likely to develop clinical colds following inoculation.
Integrative Mechanisms: Social Support and Immune-Mediated Inflammation
Consistent with epidemiological findings, much of the research conducted above suggests a role for social support across a number of diseases (i.e., cardiovascular disorders, cancer, infectious diseases). As a result, research that examines more integrative mechanisms can provide a basis for understanding the general health risks associated with poor social support. Recent research on immune-mediated inflammation provides one possible avenue for a more integrative approach. Traditionally, immune processes have been linked to infectious diseases and cancer. Cardiovascular disease can now be added to that list of disease processes with an immunologic component (Libby, 2002; Ross, 1999).
Immune processes are implicated in just about every stage of atherosclerosis (Libby, 2002). One of the earliest events in the atherogenic process is endothelial damage (Ross, 1999). At early stages of damage, the endothelium begins to express adhesion molecules such as vascular adhesion molecule-1 that help in the binding of immune cells to the vasculature. Monocytes and T-lymphocytes are then recruited to sites of inflammation and migrate into vessel walls via various cytokines/chemokines (e.g., monocyte chemoattractant protein-1) that are released from vascular cell walls (Charo and Taubman, 2004). Once inside vessel walls, these immune cells proliferate and release a variety of growth factors (e.g., platelet-derived growth factors) and cytokines (e.g., interleukin-1) characteristic of the inflammatory response (Libby, 2002). Immune events at later stages can lead to the rupture of such plaques. For instance, macrophages are common in vulnerable plaques and can produce enzymes (e.g., metalloproteinases) that degrade the fibrous cap, while T-lymphocytes release interferon-γ that can impede collagen formation by smooth muscle cells (Libby, 2002).
Importantly, there are preliminary data linking social support to immune-mediated inflammatory processes. One cytokine of recent interest is interleukin-6 given its diverse physiologic role in inflammation as well as the adaptive immune response (Hawkleyet al., in press; Kiecolt-Glaser et al., 2002). For instance, interleukin-6 is a potent stimulator of C-reactive protein and fibrinogen production from the liver (Libby, 2002). To date, three studies found that social support was related to lower levels of interleukin-6 (Costanzo et al., 2005; Friedman et al., 2005; Lutgendorf et al., 2000). Kiecolt-Glaser and colleagues (2005) also found in a laboratory study that supportive interactions were associated with a stronger acute increase in inflammatory cytokines (tumor necrosis factor-α, interleukin-1, interleukin-6) at the site of blister wounds, as well as faster wound healing compared to a conflict interaction. This study highlights the importance of considering the disease context and time course (e.g., acute versus more chronic inflammation) in links between social support and disease processes.
Despite these preliminary and promising links, there are several complexities in examining inflammatory processes that require discussion (Hawkleyet al., in press). First, cytokines can activate many different cell types and hence are associated with diverse biologic effects (Abbas and Lichtman, 2003). For instance, although evidence does exists on the pro-inflammatory influences of interleukin-6 (e.g., via induction of C-reactive protein, see Verma et al., 2004); interleukin-6 also appears to have important anti-inflammatory effects (e.g., via activation of the hypothalamic-pituitary-adrenal axis, see Barton, 1997; Hawkleyet al., in press). As a result, Hawkley and colleagues (in press) caution researchers in interpreting the results of correlational studies because in some circumstances it is possible that elevated levels of interleukin-6 may be co-activated in an attempt to control inflammation.
A second complexity related to links between social support and inflammation that will require increased attention relates to the conditions under which normal control points can be disrupted (Hawkleyet al., in press). One important control mechanism is activation of the hypothalamic-pituitary-adrenal axis that tends to suppress immunity (Munck et al., 1984). However, psychosocial processes such as chronic stress can also lead to a state of glucocorticoid resistance (Miller et al., 2002; Hawkleyet al., in press). For instance, pro-inflammatory cytokines such as tumor necrosis factor-α can modulate glucocorticoid receptors on cytokine producing cells (Webster et al., 2001). The cytokine macrophage migration inhibitory factor is also receiving attention as an important mechanism of glucocorticoid resistance (Baugh and Donnelly, 2003; Hawkleyet al., in press).
Implications and Future Directions
Researchers have long argued for the importance of understanding the biological mechanisms linking social support to physical health outcomes (Berkman and Syme, 1979; Broadhead et al., 1983; Cassell, 1976). The research reviewed above represents the tremendous progress researchers have made to date in providing answers to this question. There is ample evidence linking social support to aspects of the cardiovascular, neuroendocrine, and immune system which is consistent with the beneficial role of social support across different diseases (Uchino, 2004). Recent research on immune-mediated inflammatory processes is also providing a more integrative perspective on modeling how these biological changes are coordinated in ways that lower disease risk.
Despite these promising links to health-relevant biological processes, there are several issues that will require increased attention in this literature. One issue concerns the need for more data on the stage of disease potentially impacted by social support (i.e., development and/or progression of diagnosed disease). One perspective that may be helpful in this regard is a lifespan or successful aging approach (Rowe and Kahn, 1987). This perspective highlights the importance of linking social support to disease processes throughout the lifespan; with appropriate attention to developmental processes that impact social functioning (Antonucci and Akiyama, 1987; Carstensen, 1992). For instance, there is evidence that some variance in perceptions of social support may be established early in life (Sarason et al., 1986). Graves and colleagues (1998) found that ratings of early parental closeness were associated with a greater number of close contacts providing social support, as well as increased group participation thirty years later. Such data suggest that early social processes can cast “long shadows” that become more evident from a lifespan perspective.
It is also important to keep in mind that biological mechanisms represent just one level of analysis in the complex links between relationships and physical health (see Figure 1). There is a pressing need to also identify processes and mechanisms impacting associations between social support and health at other levels of analysis. Although not depicted in Figure 1, more macro social-cultural factors can influence support-seeking as well as the effectiveness of support (Dressler, 1994; Taylor et al., 2005; see Uchino, 2004). For instance, social networks are normally composed of individuals with a similar socioeconomic background (Lin, 1982). As a result, the broader social networks of individuals low in socioeconomic status may be similarly taxed and thus be limited sources of support (Krause, 2001). It is also important to note that the biological mechanisms reviewed earlier can serve as intermediate endpoints by which to examine other levels of analysis (see Figure 1). For instance, researchers can model if links between social support and ambulatory blood pressure are mediated by lower perceptions of stress. Such research can inform epidemiological studies that track social support, relevant multilevel mechanisms (including biological pathways), and disease morbidity/mortality over relatively long periods of time.
Another important consideration in my view is the need for a more comprehensive view on social relationships and health that incorporates both positive and negative features (Uchino et al., 2001). It is clear that social ties may serve as significant sources of stress or become models for deviant or unhealthy behaviors (Burg and Seeman, 1994; Wills and Yaeger, 2003). Research also indicates that many of our close relationships are characterized by both helpful and upsetting qualities (Fincham and Linfield, 1997; Uchino et al., 2001). In fact, the effect sizes associated with social support and health may actually be larger than it appears due to the practice of ignoring the co-occurring negative feelings that exist in many of our close relationships. Future research that takes into account this complexity in relationships can serve to inform and refine existing theoretical models of social support processes (Newsom et al., 2003; Uchino et al., 2001).
Of course, the application of this research to interventions should also receive priority. Several studies are consistent with the health benefits of social support in both focused (e.g., Levine et al., 1979) and more general psychosocial interventions (Linden et al., 1996; Rozanski et al., 1999). One of the most comprehensive social support interventions to date was the multi-site Enhancing Recovery in Coronary Heart Disease (ENRICHD) clinical trial (The ENRICHD Investigators, 2003). In this intervention, myocardial infarction patients with low social support and/or depression were randomized to cognitive-behavioral therapy that individually addressed problems related to depression/social support. Results of this study revealed a statistically significant increase in social support after six months for participants given the intervention. However, after about twenty-nine months there were no differences in survival between the support and usual care groups.
The results of the ENRICHD trial were surprising in that it was inconsistent with the large body of epidemiological and clinical/laboratory studies linking support to beneficial health outcomes. Why was the support intervention not associated with differences in survival? There are a few possibilities discussed by the authors. One explanation relates to the effect size associated with the intervention (The ENRICHD Investigators, 2003). At the six month period there was approximately a 9% difference in support changes that favored the intervention. However, there was no longer a significant difference between the intervention and usual care groups at the forty-two month follow-up (due mostly to unexpected improvements in social support for the usual care group). Thus, the initial support differences may have been too small or not maintained over a long enough period of time to influence the main cardiovascular outcomes. In fact, as discussed by the authors, it is unknown how much of a change in social support may be needed or how long an intervention is necessary to influence cardiac outcomes that have a long-term etiology. It should be noted that post-hoc analyses of these data have shown beneficial influences for some patient subgroups (Burg et al., 2005; Schneiderman et al., 2004). Overall, the results of the ENRICHD trial highlight the challenges in translating prior social support research to interventions that attempt to modify equally complex disease processes.
The complexity in designing support interventions highlights the need to focus on a more basic question: “what is social support?” There is presently no universally agreed upon definition of social support. In some cases it refers to aspects of the social network (groups, familial ties), at other times to specific behaviors (e.g., emotional or informational support), and sometimes to our perceived availability of support resources that may be shaped early in life. Answers to the question of “what is social support” will likely require complex analyses of childhood social experiences, personality, and close relationships over the lifespan (Badr et al., 2001; Pierce et al., 2000; Smith and Gallo, 2001). Such an analysis may be necessary if we are to understand the development and maintenance of strong support networks that then facilitate health-relevant psychological, behavioral, and biologic processes. Due to the complexity of these issues, an interdisciplinary perspective will be crucial, and if successful, could highlight multiple potential entry points for intervention.
Preparation for this review was generously supported by grant number R01 HL68862 from the National Heart, Lung, and Blood Institute.