RESEARCH AND PRACTICE
Race and Unhealthy Behaviors: Chronic Stress, the HPA Axis,
and Physical and Mental Health Disparities Over the Life Course
James S. Jackson, PhD, Katherine M. Knight, PhD, and Jane A. Rafferty, MA
The strain of living under inhospitable environmental conditions is hypothesized to result
in physical health disparities among racial
groups.1–3 In addition, the inequities associated
with inhospitable environments—inequalities in
employment, income, and educational opportunities that favor non-Hispanic Whites over
Blacks—are hypothesized to cause not only
poorer physical health but also worse mental
health among Blacks. However, epidemiological
and clinical data show that in comparison with
non-Hispanic Whites, Blacks suffer the same or
lower rates of most major mental disorders, even
while suffering higher rates of psychological
distress.4–7 These apparently contradictory disparities in physical and mental health statuses
raise questions about the presumed relationships
among negative life conditions and stressors on
the one hand and poor physical health and
mental disorders on the other.1,3
Objectives. We sought to determine whether unhealthy behaviors play
a stress-buffering role in observed racial disparities in physical and mental
health.
Methods. We conducted logistic regressions by race on data from the first 2
waves of the Americans’ Changing Lives Survey to determine whether unhealthy
behaviors had buffering effects on the relationship between major stressors and
chronic health conditions, and on the relationship between major stressors and
meeting the criteria for major depression.
Results. Among Whites, unhealthy behaviors strengthened the relationship
between stressors and meeting major-depression criteria. Among Blacks,
however, the relationship between stressors and meeting major-depression
criteria was stronger among those who had not engaged in unhealthy behaviors
than among those who had. Among both race groups there was a positive
association between stressors and chronic health conditions. Among Blacks
there was an additional positive association between number of unhealthy
behaviors and number of chronic conditions.
Conclusions. Those who live in chronically stressful environments often cope
with stressors by engaging in unhealthy behaviors that may have protective
mental-health effects. However, such unhealthy behaviors can combine with
negative environmental conditions to eventually contribute to morbidity and
mortality disparities among social groups. (Am J Public Health. 2010;100:
933–939. doi:10.2105/AJPH.2008.143446)
RACIAL AND ETHNIC HEALTH
DISPARITIES
Compared with Americans of European descent, Black Americans have greater physical
health morbidity and mortality at every age.8
For example, Black women are twice as likely as
White women to die of hypertensive cardiovascular disease. In addition, Blacks have a lower
average life expectancy (70 years) than Whites
do (77 years), with Black men having a life
expectancy of only 66 years. Although the causes
of these differences are debated, what is notable
is how consistently these physical health disparities favor non-Hispanic Whites over Blacks.9
Conversely, psychiatric epidemiological surveys find that Blacks in noninstitutionalized
populations have lower-than-expected rates of
most major mental disorders. For example, the
Epidemiological Catchment Area study found
roughly comparable rates of mental disorders
for Blacks and Whites10; age-adjusted analyses
by gender and study site did not suggest higher
levels of lifetime or 6-month prevalence of major
depression among Blacks.11 Even more striking,
results from the National Comorbidity Study and
the recently concluded Collaborative Studies of
Psychiatric Epidemiology revealed that rates of
mental disorders, especially mood disorders,
were consistently lower for Black Americans
than they were for White Americans.6,12
We theorize that, over the life course, coping
strategies that are effective in ‘‘preserving’’ the
mental health of Blacks may work in concert
with social, economic, and environmental inequalities to produce physical health disparities
in middle age and late life.13 Exposures to stressinducing events are more accessible to one’s
consciousness than the biological degenerations
(e.g., growth in tumors, atherosclerosis, and so
on) that eventuate in physical health ailments
and chronic health conditions (e.g., heart disease,
cancer). Thus, we hypothesize that when individuals are chronically confronted with stressful conditions in daily life (e.g., poverty, crime,
poor housing), they will engage in unhealthy
May 2010, Vol 100, No. 5 | American Journal of Public Health
behaviors (e.g., smoking, alcohol use and abuse,
drug use, and overeating, especially of comfort
foods) that help to alleviate the resulting symptoms of stress.7,14 However, these same behaviors silently contribute to physical health
morbidities and early mortality. Thus, we
hypothesize that engaging in unhealthy behaviors alleviates the symptoms of stress and the
possible biological cascade to mental disorders
while simultaneously combining with the effects
of poor living conditions to contribute to the
development of physical health ailments and
chronic physical health disorders later in life.13
We believe that these unhealthy behaviors
may either block the neurologic cascade or
mask the physiological and psychological experiences of poor mental health by acting on
the hypothalamic-pituitary-adrenalcortical
(HPA) axis and related biological systems.15
These unhealthy behaviors may have salubrious
effects by helping stave off mental disorders
among some race groups, but the direct effects of
Jackson et al. | Peer Reviewed | Research and Practice | 933
RESEARCH AND PRACTICE
stressful living16 combine with the direct effects
of the unhealthy behaviors themselves to create
large physical health disparities that are unfavorable to Blacks.4
THE HYPOTHALAMIC–PITUITARY–
ADRENALCORITCAL AXIS
The physiological ‘‘stress response’’ likely
evolved to deal with acute stressors (usually
short-term, life-threatening stressors) by mobilizing energy for immediate use and suppressing nonessential systems.17 Although the stress
response is well-adapted to deal with acute
stressors, chronic activation of the system—as is
often the case for those with poor living conditions and psychological stressors—results in poor
psychological and physical health outcomes.18
Several biological systems are activated by stress,
but we focus on the HPA axis and the implications this system has for negative health behaviors that may buffer the effects of stress on
mental disorders.
When an organism experiences stress, the
HPA axis response begins with the release of
corticotropin-releasing factor (CRF) from the
hypothalamus, stimulating the release of adrenocorticotropic hormone (ACTH) from the
pituitary gland. ACTH travels through the
bloodstream to stimulate the release of cortisol
from the adrenal cortex. Via a negative feedback loop, cortisol then acts on the hypothalamus and pituitary gland to shut down the
release of both CRF and ACTH.
Recent research from Dallman et al.19,20
suggests that consumption of foods that are high
in fats and carbohydrates reduces anxiety via
feedback to the HPA axis. During chronic stress,
the negative feedback loop through which
cortisol regulates further release of CRF breaks
down as glucocorticoid receptors are downregulated and the release of CRF continues. Continued release of CRF is associated with feelings
of anxiety as CRF mRNA expression in the
amygdala is increased20; consuming comfort
foods aids in the ‘‘shutdown’’ of the stress response by regulating the release of CRF. Abdominal fat deposits resulting from comfort
food consumption signal increased metabolic
energy stores, which in turn decrease the expression of CRF mRNA in the hypothalamus via
the inhibition of catecholamine production in
the nucleus of the tractus solitarius. Put more
succinctly, eating comfort food reduces anxiety
by inhibiting the release of CRF.19,20
High rates of obesity are observed in Black
populations, particularly among women,21 and
it is believed that consuming large amounts of
comfort foods may contribute to this condition.
Consuming comfort foods may be a socially
accepted, gender-appropriate way of dealing with
chronic stress among this population.3 Also,
sources of comfort foods may be more prominent in poorer and Black communities because
of the proliferation of fast-food outlets and
convenience stores in these areas. However, this
stress-reduction technique imposes a cost.
Chronic activation of the HPA axis has been
linked to type II diabetes via promotion of insulin
resistance in fat cells. In addition, consumption of
high-fat, high-carbohydrate foods is related to
stroke, cardiovascular disease, and other disorders.
Alcohol intake is believed to reduce anxiety
and relieve tension.22 Human studies reveal that
there is a positive relationship between stress
levels and negative psychological states on the
one hand and alcohol consumption on the
other.23,24 We believe that alcohol’s simultaneous elevation of dopamine and b-endorphin
levels in the brain results in a feeling of relaxation
and subjective release from stress.25 Thus, alcohol consumption activates the HPA axis, increases release of dopamine and b-endorphins,
and likely reduces feelings of stress.26
Smoking and nicotine ingestion are often
reported to result in mild euphoria, increased
energy, suppressed appetite, and a sense of
well-being.27 Nicotine is thought to reduce
stress-related anxiety,28 and researchers have
focused on how the HPA axis is affected by
tobacco use. Research in both humans and
animals has found evidence that nicotine increases levels of stress hormones,29,30 suggesting
that nicotine has an anxiogenic effect. Nicotine
has other neurologic effects, however, that may
explain the anxiolytic effects reported by individuals who use tobacco. Paradoxically, the
release of stress hormones in response to nicotine actually mediates the response of the mesolimbic dopamine system, giving rise to feelings
of relaxation, reduced anxiety, and calm.31
The same pattern of physiological responses
is also found following the use of illegal
stimulants.32 In addition, these drugs also activate the HPA axis, which may increase the
934 | Research and Practice | Peer Reviewed | Jackson et al.
allostatic load of the individual.18 Thus, although
individuals may be protected from the psychological effects of stress, they are not protected
from its physical effects.
METHODS
We analyzed data from the first 2 waves
(collected in 1986 and 1989) of the Americans’
Changing Lives study, conducted by the Survey
Research Center, Institute for Social Research,
University of Michigan. The Americans’
Changing Lives study is a multistage, stratified,
area-probability sample of noninstitutionalized
people aged 25 years and older residing in
the 48 contiguous United States. Data were
collected in face-to-face interviews in wave 1
and largely by telephone in wave 2. In the
initial pool of respondents, there was an oversampling of Blacks and people aged 60 years or
older. The full wave 1 sample included 3617
respondents, for a 68% response rate; the full
wave 2 sample included 2867 individuals,
83% of wave 1 respondents who were still alive
in 1989. For the current analyses, our starting
sample included Blacks (n = 874) and Whites
(n =1906) who responded at both waves.33
Measures
Because we conceptualized our model as
a series of processes that unfold over time, we
used predictor measures from variables collected at wave 1, and we used outcome measures from wave 2 assessments. The wave 2
major-depression measure represented an algorithm that includes survey questions; the
lead-in question was a modified version of the
stem question from Diagnostic Interview
Schedule version III-R and assessed whether
the respondent has ever had a time in their life
lasting an entire week when they felt sad, blue,
or depressed most of the time, or when they
lost interest in all things.34 Respondents who
replied yes to this screening question were asked
additional questions regarding duration of episodes, timing of episodes, and symptoms needed
to ascertain whether a person met DSM-II-R
criteria for major depression. The final measure
incorporates all of these items and indicates
whether the respondent had met the criteria for
major depression since wave 1 data collection.35
The depression-assessment questions were
not asked at wave 1, but we wanted to predict
American Journal of Public Health | May 2010, Vol 100, No. 5
RESEARCH AND PRACTICE
change in depression status, so we included as
a control a dichotomous version of the Center for Epidemiological Studies–Depression
Scale (CES-D) that was assessed at wave 1. The
CES-D is a measure of depressive symptoms,
an 11-item measure scaled to 20 by multiplying
the sum by 1.818. A CES-D score of 16 or
higher suggests that criteria for clinical depression have been met.36 Several studies
have shown that Blacks have more depressionrelated symptoms than do Whites but do not
meet criteria for major depression on instruments based on the Diagnostic Interview Schedule or the Composite International Diagnostic
Interview; therefore, using wave 1 CES-D scores
represents a conservative strategy for assessing
new cases between waves 1 and 2.
The physical health measure represented
a count of the total number of chronic health
conditions that respondents reported experiencing within the prior year. The list presented
to respondents included 10 conditions: arthritis/rheumatism, lung disease, hypertension,
heart attack or heart trouble, diabetes, cancer
or any malignant tumor, stroke, broken or
fractured bone, foot problems, and urinary
incontinence. Because of slight skewness of this
count measure, we modeled a 2-level version
of this variable that was split at the median
(respondents at or below the median were
coded as 0, and those above the median were
coded as 1).
Sex and region were both 2-level variables,
with 1= males and 1= residing in the South.
Respondent’s age and highest level of education were measured in years. Employment
status and occupation were 2-level variables,
with 1= employed and 1= blue-collar job. Poverty was a size-adjusted measure of household
income, calculated by dividing total household income by the official US poverty threshold37 corresponding to the size of the household.
Values above 1.0 represented households
where income exceeds officially defined needs.
Respondents were asked if they had experienced any of 9 stressful events within given
time frames: serious physical attack, lifethreatening illness, or accidental injury at any
point in one’s life; moved to a new residence,
involuntarily lost job (excluding retirement),
robbery or burglary, or other upsetting event
within the prior 3 years; and providing care
to a friend or family member who needs
assistance or having an injury or other sudden
crisis within the prior year. The stressors
measure represented the total count of these 9
events that the respondent had experienced.
Unhealthy behavior represented a count of
negative health behaviors, including smoking
cigarettes (currently or ever), drinking alcohol
(ever), and being obese, defined by having
a body mass index ([BMI; defined as weight in
kilograms divided by height in meters squared])
of 30 kg/m2 or more.
Analyses
The c2 test and t test were conducted to
assess differences between Blacks and Whites
on all measures included in the analyses.
Parallel logistic regression analyses were run
within race groups to examine key segments of
the framework. Specifically, the analyses examined the direct and moderating effects of
stressors and unhealthy behaviors on meeting
DSM-III criteria for depression, and on being in
the higher of a 2-category measure of chronic
conditions experienced within the prior 12
months. Moderating effects were assessed
by including an interaction term for stressors · unhealthy behavior. The interaction
terms included a mean-centered version of the
stressors variable; when using a continuous
variable as an interaction term, a mean-centered construction of the variable reduces
collinearity between the interaction term and
the main effect.38 A Wald test of difference was
calculated to confirm the differential magnitude
of the stressors · unhealthy behavior interaction
in predicting depression.
RESULTS
Table 1 presents a comparative overview of
the study variables for Blacks and Whites. The
results suggest that there were no race differences in the percentage of respondents meeting
DSM-III criteria for depression (13.0% of
Blacks versus 12.5% of Whites); however,
there was a greater percentage of Blacks than
Whites with a high level of chronic conditions
(46.6% of Blacks versus 36.4% of Whites).
There were a number of race differences in the
control and socioeconomic measures, and in
the stressors and unhealthy behaviors measures. Overall, the data reveal that compared
with Whites, Blacks lived in more precarious
May 2010, Vol 100, No. 5 | American Journal of Public Health
socioeconomic circumstances but had
experienced fewer major life stressors (1.4
versus 1.8) and had engaged in slightly fewer
unhealthy behaviors (1.3 versus 1.3, with
rounding).
Table 2 presents the logistic regression
models predicting depression, run separately
for Blacks and Whites. The models included
demographic control measures, indicators of
socioeconomic status, a count of stressful life
events (stressors), and the interaction between
stressors and unhealthy behaviors. The key
finding here is that the direction of the interaction term among Blacks (odds ratio
[OR] = 0.81; 95% confidence interval
[CI] = 0.67, 0.97) is opposite the direction of
the interaction term among Whites (OR =1.11;
95% CI = 0.98, 1.25). The interaction term for
Whites is only significant at the trend level
(P < .10).
Figure 1 depicts the moderating effect of
unhealthy behaviors for Blacks and Whites
separately. The relationship between stressors
and meeting criteria for depression varied by
the level of unhealthy behaviors; however, the
direction of the effect of poor health behavior
was strikingly different for Blacks and Whites.
Among Blacks, we found that the relationship
between stressors and meeting criteria for depression was stronger among those who had
engaged in none of the unhealthy behaviors
than among those who had engaged in unhealthy behaviors (Figure 1a). In contrast,
among Whites, there was an increasingly more
positive association between stressors and
meeting criteria for depression at higher levels
of unhealthy behaviors, and unhealthy behavior exacerbated the relationship between
stressors and depression (Figure 2b).
In Table 3 , we present 2 statistical models
for both Blacks and Whites, predicting a 2-level
measure of chronic conditions. Model 1 includes all measures, with the exception of the
stressors · unhealthy behaviors interaction, and
model 2 includes all measures. Those results
suggest that among both Blacks and Whites
there was a positive association between
stressors and the higher of the 2 levels of
chronic conditions. Among Blacks there was an
additional positive and independent relationship between number of unhealthy behaviors
and chronic conditions. Importantly, we did
not find evidence of unhealthy behaviors
Jackson et al. | Peer Reviewed | Research and Practice | 935
RESEARCH AND PRACTICE
environments—to the disparities in mortality and
physical health morbidity observed between
Black and non-Hispanic White populations.13
TABLE 1—Sample Characteristics, by Race/Ethnicity: Americans’ Changing Lives Survey,
1986 and 1989
Black
White
t or c2
Limitations
Dependent measures (wave 2)
Meets DSM-III depression criteria, % (no.)
13.0 (869)
0.68
46.6
36.4
26.1*
35.0 (871)
32.2
21.4 (1900)
37.8
58.3*
8.2*
52.4 (16.8)
53.7 (17.2)
24–96
25–91
56.1
Has 2-level chronic conditions, %
12.5 (1896)
32.5
139.0*
12.8*
Control measures (wave 1)
Meets CES-D depression criteria,a % (no.)
Male, %
Age, y
–1.9*
Mean (SD)
Range
Residence in South, %
Socioeconomic measures (wave 1)
Education, no.
872
1904
10.4 (3.7)
0–17
12.3 (3.0)
0–17
Mean (SD)
1.6 (1.8)
3.0 (2.5)
0.10–17.0
Mean y (SD)
Range, y
Poverty ratioc
-16.7*
Range
0.1–15.9
Employed,d %
52.2
54.4
1.2
Blue-collar occupation, %
31.7
18.6
58.3*
Stressor and unhealthy behavior measures (wave 1)
Stressors
Mean (SD)
1.4 (1.3)
1.8 (1.4)
–6.8*
Range
0–6
0–8
1.3 (0.9)
1.3 (0.8)
0–3
0–3
Unhealthy behaviorsb
Mean (SD)
Range
–2.0*
Note. CES-D = Center for Epidemiological Studies—Depression Scale; DSM-III = Diagnostic and Statistical Manual of Mental
Disorders, Revised Third Edition. Unless otherwise noted, for Blacks, n = 874; for Whites, n = 1906. For differences in race
ethnicity, the t test was used for continuous variables and the c2 test for dichotomous variables.
a
An 11-item measure scaled to 20 by multiplying the sum by 1.818. A CES-D score of 16 or higher represents meeting the
criteria for clinical depression.
b
Unhealthy behaviors include smoking (current or ever), drinking (ever), and being obese (defined as having a body mass
index [weight in kilograms divided by height in meters squared] of 30 or higher). Reported difference in count of unhealthy
behaviors reflects rounding; the mean for Blacks is 1.2746, and the mean for Whites is 1.3410.
c
Lower scores indicate more impoverished status.
d
For Whites, n = 1905.
*P < .05.
moderating the effect of stressors on physical
health among Blacks or Whites.
DISCUSSION
Many Black Americans live in chronically
precarious and difficult environments.1,3 These
environments produce stressful living conditions,
and often the most easily accessible options for
addressing stress are various unhealthy behaviors (e.g., smoking, drinking, drug use, and so on).
As we have noted, these behaviors may alleviate
the symptoms of stress through the same mechanisms that are hypothesized to contribute to
some mental disorders: the HPA axis.15 What is
certain is that negative health behaviors, such as
smoking, overeating (especially comfort foods),
drinking alcohol, and drug use also have direct
and debilitating effects on physical health. Thus,
although these activities have the effect of alleviating or masking the ostensible symptoms of
stress, they contribute—along with difficult living
936 | Research and Practice | Peer Reviewed | Jackson et al.
Our analysis was limited by the fact that it
was based on data not collected to test the
specific hypotheses. Thus, we have had to make
compromises in operationalizing the variables
in the statistical models. For example, our
measure of stressors is a simple count of
negative life events and does not include any
assessment of perceived social or psychological
stress. Another limitation is that BMI is not
a direct assessment of overeating as conceptualized in the framework. A better test would
include direct measures of overeating comfort
foods, because a person could have a high body
mass index for reasons not related to overeating comfort foods (genetic, hormonal, and so
on).
Also, because we lacked the same measure
of depression at waves 1 and 2, we had to use
the CES-D with suggested cutoffs for depressive disorder, to feel more confident we were
looking at ‘‘new’’ cases from wave 1 to wave
2. It would have been optimal if we had been
able to use the same DSM measure at both
points in time. Because the CES-D is based on
a total symptom count, it tends to overstate
the likelihood of meeting criteria for depression
and is not based upon gateway symptoms, as
is the case with DSM criteria for major depression. In addition, the analyses in this article
are only relevant for noninstitutionalized populations. We do not know what the effects
would be if we included, for example, incarcerated Blacks and Whites. Although these
limitations may need to be addressed in future
studies, their effect is to make our test of the
hypothesis in this paper more conservative,
because in all cases the limitations dampen the
statistical assessment of the hypothesized relationships. Thus, we have more confidence
that the observed statistical relationships reflect
likely significant associations among the variables in the models.
Conclusions
A full understanding of racial disparities in
physical and mental health requires a consideration of the life course, socioeconomic status,
culture, and gender.1,3,4 Descriptively, health
American Journal of Public Health | May 2010, Vol 100, No. 5
RESEARCH AND PRACTICE
TABLE 2—Results of Logistic Regressions Predicting DSM-III Depression at Wave
2 Among Blacks (n = 864) and Whites (n = 1887): Americans’ Changing Lives Survey,
1986 and 1989
Blacks, OR (95% CI)
Whites, OR (95% CI)
Control measures (wave 1)
Male
