Introduction

African Americans (AA) disproportionately experience poorer physical and mental health outcomes than their White counterparts (see Lewis et al., 2015). Scholars have suggested that greater exposure to racial discrimination (discrimination), a psychosocial stressor for AAs, can accelerate health deterioration by dysregulating biological stress pathways, specifically the hypothalamic–pituitary–adrenal (HPA) axis (Adam et al., 2015; Berger & Sarnyai, 2015; Chrousos, 2009; Jackson et al., 2010). Dysregulation of the HPA axis can occur following repeated and prolonged exposure to the glucocorticoid cortisol, the primary byproduct of HPA axis activation, presenting as elevated or attenuated concentrations of cortisol in response to stress and across the day (McEwen, 1998). Despite evidence that chronic psychosocial stressors such as discrimination may lead to HPA axis dysregulation, less is known about how psychological responses to discrimination (e.g., anxiety, depressive symptoms) influence HPA axis function and circulating levels of cortisol. Moreover, a few scholars have reported that AA females display attenuated cortisol reactivity in the context of discrimination compared to their male counterparts (Richman & Jonassaint, 2008). Research is limited, however, in elucidating how AA males’ and females’ psychological responses to stress may influence cortisol responses to discrimination.

In an effort to fill these gaps in the literature, our study consisted of two aims. First, guided by Clark et al. (1999) biopsychosocial model of racism and health, we examined anxiety and depressive symptoms as mediators in the relation between discrimination and circulating cortisol levels. This is a unique contribution because researchers have primarily studied the direct effect of discrimination on cortisol reactivity, despite evidence suggesting psychological functioning as a potential pathway of influence (e.g., anxiety; Vreeburg et al., 2010). Second, we assessed if the underlying psychological pathways connecting discrimination to total levels of cortisol varied by sex in light of research that supports sex as a qualifier to the discrimination-cortisol link (Adam et al., 2015).

Racial discrimination and cortisol response

Racial discrimination, a psychosocial stressor that is ubiquitous in the lives of AAs, has been defined as, “…differential treatment of members of these groups [racial outgroups] by both individuals and social institutions” (Williams & Mohammed, 2009, p. 3). Discrimination experiences have been associated with negative physiologic and psychologic outcomes (see Lewis et al., 2015). Specifically, Clark et al. (1999) biopsychosocial model of racism and health posits that AA’s experience a wide range of psychological and physiological responses to discrimination including activation of the HPA axis. In line with the notion of Allostatic load, the biopsychosocial model of racism and health suggests that routinely invoking these biological and psychological stress responses could lead to wear and tear on the body thereby influencing the genesis of disease processes (Clark et al., 1999; McEwen, 1998). Scholars, to this end, have found that discrimination can indeed set in motion physiological responses (e.g., elevated heart rate, blood pressure) which, in turn, can contribute to immediate and long-term physiological dysfunction (e.g., hypertension) and health problems (Geronimus et al., 2006; Lewis et al., 2015; Williams & Mohammed, 2013).

The HPA axis, which is activated in response to stressors perceived as uncontrollable or stressors that involve social-evaluative threat (Dickerson & Kemeny, 2004), has been frequently implicated as a key biological pathway by which discrimination leads to poorer health (Adam et al., 2015; Berger & Sarnyai, 2015; Fuller-Rowell et al., 2012; Jackson et al., 2010). Although periodic activation of the HPA axis is adaptive in the context of acute stress (i.e., readying the body for “fight or flight”), chronic and prolonged activation of the HPA axis is linked to numerous health problems (Geronimus et al., 2006). Scholars have suggested that discrimination can lead to an exaggerated physiological response both momentarily and long after a discriminatory event (Hope et al., 2015). For example, Hoggard et al. (2015) found that, among AAs, discriminatory experiences committed by a White perpetrator had a prolonged effect on cardiovascular functioning during and 24 h after the mistreatment occurred. Hoggard et al. (2015) theorize that emotionally and cognitively re-experiencing a discrimination experience or anticipating another discrimination experience may trigger the prolonged activation of the “fight or flight” response. If discrimination chronically dysregulates the HPA axis, this process may have potential long-term implications for health (Geronimus et al., 2006; Hope et al., 2015).

It is also important to note that researchers have conceived of discrimination as a key contributor to differences in HPA axis function between AAs and White Americans (Berger & Sarnyai, 2015). Biopsychosocial theories of stress such as the attenuation hypothesis (Susman, 2006) suggest that exposure to chronic stress such as discrimination would lead to a downregulation of the HPA axis to limit the body’s exposure to cortisol. Support for this has been shown in studies that have found attenuated baseline concentrations and flatter diurnal rhythms of cortisol in AAs relative to their White peers during adolescence (DeSantis et al., 2007) and emerging adulthood (Cohen et al., 2006). Furthermore, several researchers have found that discrimination predicted flatter diurnal cortisol slopes (e.g., Adam et al., 2015) and attenuated levels of overall diurnal cortisol (Kaholokula et al., 2012). Yet, others have found no association between discrimination and diurnal cortisol rhythms in AAs (Fuller-Rowell et al., 2012; Zeiders et al., 2014). To explain these counterintuitive findings, scholars have posited that the effect of discrimination on the cortisol response may be explained by the individual’s psychological response to the discriminatory event (Adam et al., 2015).

Depressive and anxiety symptoms as mediators

Discrimination has been conceptualized as a stressor that can exact psychologic tolls (Lewis et al., 2015), yet few studies have examined psychological processes underlying the relation between discrimination and circulating levels of cortisol (see Berger & Sarnyai, 2015 for review). Along these lines, researchers have consistently found that discrimination is related to two phenomenologically distinct mental health indicators: anxiety and depressive symptoms (see Pieterse et al., 2012).

With respect to anxiety, discrimination experiences have been associated with cognitive (i.e., fear, chronic worrying) and somatic (i.e., racing heart, muscle tension) symptoms of anxiety in non-clinical samples of AAs (e.g., Lee et al., 2014). Specifically, scholars have classified discriminatory encounters as a social stressor that can evoke fear (Jones et al., 2014), trauma-related symptoms (Polanco-Roman et al., 2016), and somatic anxiety symptoms (Lee et al., 2014), which can, in turn, dysregulate the production of cortisol (Assari et al., 2015). Consistent with Susman’s (2006) attenuation hypothesis, researchers have suggested that experiencing chronic stressors such as discrimination can gradually exacerbate anxiety and attenuate cortisol reactivity to stress (Assari et al., 2015). It is, therefore, plausible that prolonged discrimination experiences generate greater symptoms of anxiety which, ultimately, attenuates circulating levels of cortisol. Although scholars have theoretically considered anxiety as a mediator in the discrimination-cortisol link (Clark et al., 1999), this research is still limited.

Scholars have also suggested that increased discrimination experiences can result in poorer self-concept, hopelessness, and anhedonia (Brown et al., 2000). In a separate study, depression was associated with cortisol patterns that reflect blunted stress reactivity and impaired stress recovery (Burke et al., 2005), although findings have been mixed between clinical and non-clinical samples (Brooks & Robles, 2009; Burke et al., 2005). For example, anhedonia, a key indicator of depression, was associated with a lower cortisol awakening response in a clinically depressed sample, but not in a non-clinical sample (Brooks & Robles, 2009). Burke et al., 2005 additionally found that laboratory-produced stress was linked to blunted cortisol reactivity among individuals diagnosed with major depressive disorder (MDD), but not in undiagnosed individuals. Thus, it is plausible that depressive symptoms would mediate the discrimination-cortisol link if the symptoms of depression are particularly severe.

Taken together, these studies indicate that variations in the psychological response to discrimination can invoke HPA axis dysfunction (see Berger & Sarnyai, 2015, for a review). To date, however, researchers have not examined discrimination, mental health, and overall levels of cortisol mechanistically to identify potential mechanisms through which discrimination may alter HPA axis functioning. It is possible that discrimination may influence cortisol concentration through both anxiety and depressive symptoms. Some researchers have reported different patterns of association between cortisol responses to stress and depression versus anxiety symptoms in response to interpersonal stress (e.g., Powers et al., 2016). For example, depressive symptoms have been linked to less positive and negative emotional reactivity in the context of stress (Bylsma et al., 2008), whereas anxiety symptoms have been linked to higher emotional reactivity (Abelson et al., 2007; Shin & Liberzon, 2010). Associations of HPA axis functioning, depression, and anxiety symptoms are mixed in community samples. We will, therefore, test a mediation model whereby symptoms of anxiety and depression uniquely explain the link between discrimination and cortisol levels in AA emerging adults.

Sex differences

Although researchers report sex differences in the cortisol awakening response and diurnal patterns among AAs (e.g., Adam et al., 2015), Richman and Jonassaint (2008) examined sex as a moderator in the association between discrimination and circulating levels of cortisol among AAs. They found that AA females are more vulnerable than males to flatter diurnal cortisol reactivity within the context of discrimination-related stress (Richman & Jonassaint, 2008). For instance, Richman and Jonassaint (2008) revealed that during a real life racial stressor (i.e., Duke lacrosse scandal), AA females displayed a more blunted cortisol stress response pattern than AA males during a laboratory stress task. As a caveat, the Duke lacrosse scandal was an event with distinct relevance and more negative implication to AA females compared to AA males (Richman & Jonassaint, 2008). Unfortunately, the vast majority of research on discrimination and HPA axis function among AAs have treated sex as a covariate, but not as a moderator.

To understand these sex differences, scholars have postulated that sex-specific vulnerabilities to stress (e.g., depression for females) may contribute to the gendered results (Hankin & Abramson, 2001). For instance, English et al. (2014) reported that females are more vulnerable to depression in the context of race-related stress, which is a mental health indicator linked to lower cortisol responsivity (Burke et al., 2005). Alternatively, research examining sex differences in the association between discrimination and anxiety symptoms is limited and mixed. Although a few researchers have reported sex differences in the association between discrimination and anxiety symptoms (Soto et al., 2011), other scholars have found that the magnitude of this association was similar for AA males and females (McLaughlin et al., 2010). In addition, some scholars have found that anxiety symptoms moderate the physiological tolls of discrimination (Assari et al., 2015), whereas depressive symptoms do not (Chae et al., 2016). Thus, it is possible that the effect of depressive symptoms as a mediator in the discrimination-cortisol link is stronger for AA females than AA males, whereas symptoms of anxiety as a mediator of this link is similar for both sexes.

Developmental significance of emerging adulthood

Although it is well documented that discrimination experiences are pervasive and persistent in the lives of AAs (Williams & Mohammed., 2013), developmental researchers have identified emerging adulthood as a critical period for examining these issues (Arnett & Brody, 2008; Hope et al., 2015). Arnett and Brody (2008) theorized that AA emerging adults are at greater risk for confronting prejudice and discrimination as they move beyond their familial context and occupy ethnically diverse spaces (e.g., college, employment). That is, as AAs experience developmental “turning points” in their education, employment, and social life (e.g., parenthood, marriage), these turning points are often experienced within the context of institutional and interpersonal discrimination (Hope et al., 2015). For example, AA adults living in predominantly White neighborhoods reported higher levels of stress as they need to contend with implicit racialized messages such as stereotyped behavior and expectations (Stewart et al., 2009). Likewise, audit studies have found that AAs are less likely to be hired for a job than their White counterparts (see Pager & Shepherd, 2008). Thus, the culmination of these racism-related stressors in this developmental period may invoke a psychologic and physiologic toll in AAs. Not surprisingly, psychological studies of AA emerging adults have documented strong associations between discrimination and alcohol use (Hurd et al., 2014), suicide rates (Rockett et al., 2006), and perceptions of stereotype threat (Bernard et al., 2017). These studies, in tandem, suggest that emerging adulthood is a key period to examine discrimination and its psychological and physiological consequences in AAs.

Current study and hypotheses

To build upon the research on discrimination and HPA axis function, our first aim was to examine whether depressive and/or anxiety symptoms mediated the association between discrimination and circulating levels of cortisol. Guided by the biopsychosocial model of racism and health (Clark et al., 1999), we hypothesized that anxiety and depressive symptoms would mediate the association between discrimination and cortisol levels. Specifically, we predicted that discrimination would be associated with increased anxiety and depressive symptoms, which, in turn, would be associated with attenuated levels of cortisol. Our second aim was to explore sex differences in the mediational model derived in the first aim.

Method

Participants

The present study included participants from an ongoing larger study of 850 9th graders (92% response rate) on the effects of substance use on educational achievement, employment, and other health-related indicators such as psychological well-being (Zimmerman & Schmeelk-Cone, 2003). Participants were originally recruited from four public high schools in an urban, Midwestern city. In order to be included in the study, participants were required to have a 3.0 grade-point-average or lower at the end of 8th grade, self-reported to be AA, white, or mixed ethnicity, and could not have been diagnosed by the school to have a developmental or an emotional disability. The first wave of data was collected in the Fall of 1994 and participants, for this study, were assessed at the seventh wave of the study (2001–2002; n = 576), which occurred during early adulthood. The present study only included participants identifying as AA who had complete cortisol data at wave 7 (n = 339). We focused our analyses on data collected for AA participants during the seventh wave of the study for three reasons. First, cortisol data and other study variables of interest were collected in this wave. Second, this wave represented the transition into early adulthood, which is a key developmental period for examining discrimination and its psychological and physiological consequences (Arnett & Brody, 2008). Third, we focused on AAs because of our focus on discrimination and its implications for explaining poor health outcomes (especially related to mental health) in AAs.

Lastly, participants were excluded if they were pregnant (n = 24) or if their saliva samples had blood protein contamination levels greater than or equal to 3 mg/DL (n = 3) resulting in a final sample of 312 AA emerging adults. The average age of the sample was 22.05 (SD = 0.67) and approximately half of the sample identified as female (54.5%).

Procedure

Trained interviewers conducted 60 min face-to-face interviews with the participants at home or in the community setting. Following the interviews, participants were administered a paper-and-pencil questionnaire to collect more sensitive information regarding racial discrimination experiences, and alcohol and other drug use. The university institutional review board approved all study protocols (UM—IRB # H03-0001 309) and participants were compensated for their time.

Measures

Cortisol

We collected saliva samples for cortisol assay at three designated time points while participants were completing a paper-and-pencil questionnaire concerning sensitive information (e.g., substance use) during wave 7 interviews. Salivary cortisol reflects the free, unbound portion of cortisol and is widely used by researchers to measure HPA axis function (Kirschbaum & Hellhammer, 1989). Following consent procedures, participants rinsed their mouths with water. For each sample, participants collected saliva in their mouths for 1 min, and then expectorated slowly through a straw into a cryotube. All participants were interviewed between 9:50 AM and 6:49 PM (M = 1:30 PM, SD = 2 h 25 min). The first saliva sample (sample 1) was collected approximately 10 min following consent. The second saliva sample (sample 2) was collected at the interview mid-point, approximately 22 min following sample 1. The third and final sample (sample 3) was collected at the end of the interview, approximately 30 min following sample 2. Immediately following collection, the saliva samples were placed on ice and refrigerated until transport to a − 80° Fahrenheit freezer for storage. Saliva was only collected from participants who reported no food, drink or tobacco use in the previous hour.

Saliva samples were assayed for cortisol by a high sensitivity salivary cortisol enzyme immunoassay by Salimetrics (State College, PA). The saliva samples were thawed and centrifuged at 1500 rpm for 15 min before assay. The assay followed standard enzyme immunoassay procedures as previously described (Klimes-Dougan et al., 2001). The intra-and inter-assay coefficients of variability ranged from 3.88 to 7.12% and 6.69 to 6.88%, respectively. The lower limit of sensitivity of this assay is .007 ug/DL (Schmeelk-Cone et al., 2003).

Cortisol values were log transformed to reduce the skewness of the distribution. To make the repeated measurements of salivary cortisol more easily interpretable, we calculated area under the curve with respect to ground (AUCG; Pruessner et al., 2003). AUCG is a measure of total cortisol levels across the three sampling times (Fekedulegn et al., 2007). Cortisol AUCG was not associated with doctor visits in the past year or lifetime diagnosis of asthma, chronic bronchitis/emphysema, diabetes, or high blood pressure/hypertension. In females, use of a birth control pill or Depo Provera shot was not associated with cortisol AUCG.

Perceived racial discrimination

Participants’ experience with perceived interpersonal racial discrimination was measured using the 20-item Daily Life Experience scale (DLE; Harrell, 1997) at wave 7 in the paper and pencil questionnaire at the end of the interview. The DLE assessed the frequency of racial hassles experienced during the past 12 months on a 6-point Likert-type scale ranging from “never happened to me” to “once a week or more”. The participants’ scores on the DLE were averaged to capture overall levels of perceived racial discrimination over the past year (αmales = .96, αfemales = .94). Previous research demonstrated that the DLE is a reliable and a valid measure of discrimination for AA young adults (e.g., Neblett and Carter, 2012).

Depressive symptoms

The frequency of depressive symptoms were measured by 6-items from the Brief Symptom Inventory (Derogatis & Spencer, 1982) during the face-to-face interview portion of the study at wave 7. Response options on the 5-point Likert-type scale ranged from “never” to “very often,” and the respondents’ scores were averaged. Sample items include “feeling blue (or sad)” and “feeling no interest in things.” This scale demonstrated good reliability in the study sample (αmales = .81, αfemales = .93) and responses ranged from 0 to 4.

Anxiety symptoms

The frequency of anxiety symptoms were measured by 6-items from the Brief Symptom Inventory (Derogatis & Spencer, 1982) during the face-to-face interview portion of the study at wave 7. Response options on the 5-point Likert-type scale ranged from “never” to “very often,” and the respondents’ scores were averaged. Sample items include “feeling fearful” and “feeling tense or keyed up.” This scale demonstrated good reliability in the study sample (αmales = .78, αfemales = .86) and responses ranged from 0 to 2.50.

Covariates

In wave 1, participants self-reported their sex. Socio-economic status was based on current employment status (i.e., employed full- or part-time vs unemployed), educational attainment (i.e., did not graduate high school, graduated high school, or attending/completed higher-education), and financial situation (i.e., difficulty affording basic necessities) on a 3-point Likert-type scale ranging from “never” to “a lot” at wave 7. We also controlled for the start time of saliva collection to take into account the diurnal decline in cortisol. The participant’s body mass index (BMI) was calculated using the participant’s self-reported weight at wave 7 and height at wave 6 (Mage = 21). Although the participant’s height was not available at wave 7, literature posits that most individuals complete the process of physical maturation by age 20 (Loesch et al., 1995). Lastly, participants’ self-reported stress level during the past month was accounted for with the 10-item Perceived Stress Scale asking during the face-to-face interview portion of the study. Participants responded on a 5-point Likert-type scale ranging from “never” to “very often” (α = .84).

Analytic approach

To evaluate the study aims, we employed path analysis as implemented in Mplus version 7.2 (Muthén & Muthén, 2012). In all models, full information maximum likelihood (Arbuckle, 1996) was used to generate likelihood functions for only data that is available for each case. In addition, the model fit indicators examined were the comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square error of approximation (RMSEA), and the Standardized Root Mean Square Residual (SRMR). Confidence intervals were obtained using bias-corrected bootstrapping for all model parameters to account for non-normality of the distribution of indirect effects (Dearing & Hamilton, 2006).

With respect to the first study aim, we tested a parallel mediation model with depressive and anxiety symptoms as mediators in the association between discrimination and total levels of cortisol (see Fig. 1). In all models, we controlled for participants’ age, sex, BMI, current employment status, the start time of saliva collection, educational attainment, financial situation, and perceived stress. Due to the strong correlation between anxiety and depressive symptoms, we also tested the mediating role of anxiety and depressive symptoms separately while controlling for symptoms of the other disorder.

Fig. 1
figure 1

Mediation of discrimination to cortisol responsivity via symptoms of anxiety and depression. Note Comparative fit index (CFI) = 0.98; Tucker-Lewis index (TLI) = 0.97; root mean square error of approximation (RMSEA) = 0.04; standardized root mean square residual (SRMR) = 0.03; χ2 = 27.22, df = 19, p = 0.10. Solid lines indicate significant pathways. Dashed lines indicate non-significant pathways. Direct effects were estimated with maximum likelihood estimation. Confidence intervals for direct and indirect effects were estimated with bias corrected bootstrapping. Not shown are direct effects of sex, age, employment status, educational attainment, difficulty affording basic necessities, stress level, time of first saliva sample, and BMI. Correlations between covariates were modeled but not drawn

Full size image

To evaluate the second aim, we tested multi-group models to assess sex differences in the parallel mediation model parameters. With the exception of sex, all covariates from prior models were controlled for in the multi-group models. A Chi square likelihood ratio test (LRT) was used to determine whether stratifying the sample by sex altered model fit. To assess sex differences in the path coefficients, we constrained the coefficients to equality and freed one constraint at a time. After freeing a constraint, we assigned a new parameter that assessed the difference between the freed path coefficients by sex using bias corrected bootstrapping.

Results

Descriptive statistics and bivariate correlations are presented in Tables 1 and 2. Notably, anxiety symptoms and total cortisol levels were inversely associated, whereas depressive symptoms were not associated with cortisol levels. In addition, discrimination experience was associated with more symptoms of anxiety and depression. Notably, rates of discrimination, depressive symptoms, and anxiety symptoms in our sample were comparable to AA samples from larger community or nationally-representative samples (Hudson et al., 2016; Lee & Neblett, 2017; Williams et al., 2012).

Table 1 Descriptive statistics
Full size table
Table 2 Bivariate correlations
Full size table

Consistent with our hypotheses, greater exposure to discrimination was associated with more anxiety (b = 0.14, 95% CI [0.08, 0.22]) and depression symptoms (b = 0.15, 95% CI [0.05, 0.24]). Yet, only anxiety symptoms (b = − 2.70, 95% CI [− 4.94, − 0.64]) were associated with attenuated levels of cortisol. Further, a significant indirect effect from discrimination experience to cortisol levels was observed through anxiety symptoms (b = − 0.38, 95% CI [− 0.88, − 0.10]), but not through depressive symptoms (b = 0.09, 95% CI [− 0.18, 0.50]). As shown in Fig. 1, anxiety symptoms, but not depressive symptoms, mediated the association between discrimination experience and cortisol levels. We then tested whether anxiety and depressive symptoms separately mediated the association between discrimination and cortisol. Notably, anxiety mediated the association between discrimination and cortisol after controlling for depressive symptoms (b = − 0.17, 95% CI [− 0.49, − 0.02]; see Fig. 2), whereas depressive symptoms did not mediate the discrimination-cortisol link (b = − 0.03; 95% CI [− 0.05, 0.25]; see Fig. 3). Further, in all of the models examined thus far, depressive symptoms were not associated with cortisol. For this reason, to test the second study aim, we elected to examine depressive symptoms as a covariate to the model rather than as a mediator in the discrimination-cortisol link.

Fig. 2
figure 2

Discrimination frequency predicts cortisol responsivity via symptoms of anxiety, controlling for depressive symptoms. Note CFI = 0.93; TLI = 0.90; RMSEA = 0.06; SRMR = 0.04; χ2 = 38.17, df = 18, p < 0.05. Solid lines indicate significant pathways. Dashed lines indicate non-significant pathways. Direct effects were estimated with Maximum Likelihood Estimation. Confidence intervals for direct and indirect effects were estimated with bias corrected bootstrapping. Direct pathway from discrimination frequency to cortisol responsivity not significant (B = − 0.10, SE = 0.77, p = 0.90, β = − 0.01, Bootstrapped 95% CI [− 1.58, 1.56]). Not shown are direct effects of sex, age, employment status, educational attainment, difficulty affording basic necessities, stress level, time of first saliva collection, and BMI. Correlations between covariates were modeled but not drawn

Full size image
Fig. 3
figure 3

Discrimination frequency predicts cortisol responsivity via symptoms of depression, controlling for symptoms of anxiety. Note CFI = 0.94; TLI = 0.91; RMSEA = 0.06; SRMR = 0.05; χ2 = 38.21, df = 18, p < 0.05. Solid lines indicate significant pathways. Dashed lines indicate non-significant pathways. Direct effects were estimated with Maximum Likelihood Estimation. Confidence intervals for direct and indirect effects were estimated with bias corrected bootstrapping. Direct pathway from discrimination frequency to cortisol responsivity not significant (B = − 0.08, SE = 0.78, p = 0.92, β = − 0.01, Bootstrapped 95% CI [− 1.55, 1.59]). Not shown are direct effects of sex, age, employment status, educational attainment, difficulty affording basic necessities, stress level, time of first saliva collection, and BMI. Correlations between covariates were modeled but not drawn

Full size image

Our multi-group mediation model fit equally well for AA males and females (see Table 3). We compared a multi-group model with no equality constraints across males and females, and a model with equality constraints on the path coefficients. The Chi square LRT was not significant suggesting that stratifying the sample by sex and invoking equality constraints does not contribute to alterations in model fit. In addition, no sex differences were found in the direct and indirect path coefficients from discrimination to cortisol level.

Table 3 Sex differences in the path model
Full size table

Discussion

Our findings lend support to Clark et al. (1999 biopsychosocial model of racism and health by demonstrating that anxiety symptoms mediate the association between discrimination and total levels of cortisol. Consistent with the attenuation hypothesis (Susman, 2006) and prior empirical work (Assari et al., 2015), prolonged interpersonal discrimination encounters seem to be perceived as stressful and threatening by AAs, an emotional response that can exact physiologic tolls if routinely experienced. Although scholars have already documented the association between discrimination and cortisol dysregulation (e.g., Adam et al., 2015), this study is the first that examines anxiety symptoms as a mediator of the association. Findings suggest that anxiety symptoms that stem from prolonged discrimination encounters (e.g., being followed or hearing a racial epithet) may routinely activate the HPA axis over time, and may help explain the process by which racial discrimination contributes to racial health disparities.

Notably, although discrimination predicted more depressive symptoms, depressive symptoms did not mediate the association between discrimination and cortisol levels. This finding extends the current research on discrimination and health by indicating that different psychological responses to discrimination may exact different physiological tolls. In the present study, the different findings for anxiety and depressive symptoms may be attributed to the fact that anxiety symptoms (i.e., fear, hyperarousal, and tension) are more likely to invoke hyperarousal in physiologic stress systems (e.g., cardiovascular reactivity) which may be more proximal to circulating levels of cortisol than depressive symptoms (Assari et al., 2015). This finding is also consistent with empirical studies that have found that discrimination threatens one’s sense of control which fosters a sense of anxiety in AAs (Lee et al., 2014) and also is known to affect cortisol (Dickerson & Kemeny, 2004). Lastly, studies have reported that depressed patients display greater attenuations in cortisol reactivity to stress than non-clinical samples (see Burke et al., 2005). Given that our study participants were from a non-clinical, community sample, it may be that discrimination had less of an influence on cortisol concentration via depressive symptoms. Together, this study demonstrates that different psychological responses to discrimination can uniquely invoke physiological systems.

Although the effect of discrimination on cortisol level is not a prerequisite for statistical mediation (see Zhao et al., 2010), it should be noted that, in contrast to other researchers (Adam et al., 2015), we found no evidence for this direct effect. Empirical support for the association between discrimination and HPA axis functioning has been mixed for AAs, with some scholars reporting no association between discrimination and the cortisol awakening response (Zeiders et al., 2014) or the diurnal rhythm in cortisol (Fuller-Rowell et al., 2012). Scholars have found, however, that psychological (i.e., anxiety, depression; e.g., Powers et al., 2016) and social (i.e., poverty, discrimination; e.g., Adam et al., 2015) factors can directly influence cortisol output and reactivity in AAs. Our findings, however, provide support for a mechanistic biopsychosocial approach (Berger & Sarnyai, 2015; Clark et al., 1999) in understanding the physiological consequences of discrimination. Specifically, we highlight that discrimination exacts psychologic tolls, which, in turn, attenuates overall cortisol concentrations.

Lastly, as hypothesized, anxiety symptoms mediated the association between discrimination and cortisol levels for AA males and females. Although researchers have found that AA females have flatter diurnal cortisol rhythms in relation to AA males (Cohen et al., 2006), our findings suggest that the magnitude of direct and indirect associations linking discrimination to circulating levels of cortisol through anxiety symptoms is the same for AA males and females. Thus, in contrast to studies that reveal sex differences in racism and its psychophysiological consequences (see Lewis et al., 2015), our results suggest that chronically being on the receiving end of interpersonal racism elevates risk for fear, hyperarousal, and tension, regardless of sex, which then dysregulates the HPA axis and cortisol level. Accordingly, although certain psychological responses to discrimination may implicate physiological stress responses differently for AA males and females (e.g., substance use; Hurd et al., 2014), our findings suggest that discrimination attenuates cortisol concentration through anxiety similarly for AA males and females. To this end, an important consideration in studies of discrimination and health, when using a biopsychosocial frame of reference, is to examine the role of sex in the pathways of influence.

Limitation and future directions

We note several limitations to our findings. First, given the cross-sectional design of the present study, the study results are correlational and causal inferences cannot be drawn. Nevertheless, our study was the first to examine a mediational model and provides compelling evidence that future longitudinal studies are warranted. Second, the interview protocol was not designed to evoke acute psychological stress. Several scholars have, however, reported attenuations in the cortisol response to minor stressors such as disclosing sensitive and personal information (Aiyer et al., 2014; Graber & Sontag, 2009) and suggested that cortisol production in this context associates with contextually relevant variables (e.g., violence exposure; Aiyer et al.,). We also were unable to schedule interviews at the same time of day to account for the diurnal rhythm of cortisol; however, our results did not change when we controlled for the time of first saliva sample collection in our analyses. Third, our measure of discrimination was limited to interpersonal race-related experiences (e.g., being overlooked or ignored because of race). Discrimination experiences are, however, multidimensional (see Williams & Mohammed, 2013) and can occur in different contexts (e.g., interpersonal, institutional) for different reasons (e.g., race, sex). Thus, to build on the findings of the current study, future studies could examine the combined and unique effects of race-based compared to non-race-based discrimination, and discrimination experiences at the individual, interpersonal or cultural level (Williams & Mohammed, 2013).

Conclusion

These limitations notwithstanding, our study contributes to the research on racial health disparities in several important ways. First, guided by the biopsychosocial model of racism and health (Clark et al., 1999), this study highlights salient social (i.e., discrimination) and specific psychological (i.e., anxiety symptoms) risk factors that may influence circulating levels of cortisol among AA emerging adults. As researchers have implicated discrimination as a contributor to racial health disparities, our study advances this understanding by examining discrimination and its psychological risk factors. Second, our findings are unique and significant because we examined sex differences for the effect of discrimination on total cortisol levels through anxiety symptoms. Thus, findings from our study highlight the need for future research to consider whether other psychological consequences of discrimination (e.g., substance use, trauma) mediate the discrimination-cortisol link similarly for AA males and females.