Journal of Gerontology: MEDICAL SCIENCES
1997, Vol. 52A, No. 2, M68-M75
Copyright 1997 by The Geromological Society of America
Salivary Cortisol Levels
and Stress Reactivity in Human Aging
Nancy Nicolson,1 Chantal Storms,2 Rudolf Ponds,1 and Jose Sulon3
'Department of Psychiatry and Neuropsychology and 2Program in Mental Health Science,
University of Maastricht, The Netherlands.
'Department of Reproductive Physiology, Faculty of Veterinary Medicine, University of Liege, Belgium.
Background. While homeostatic mechanisms are generally believed to become less efficient in the aging organism,
evidence for changes in hypothalamic-pituitary-adrenal function is inconclusive. Previous studies report higher, lower, or
unchanged basal cortisol levels in human aging. Delayed recovery of glucocorticoids to baseline following stress
exposure has been observed in aging rats, but the generalizability of these findings to humans remains unclear.
Methods. Salivary-free cortisol was measured at home and in response to a laboratory speech task in 56 healthy men
and women in three age groups (range 43-86 yr).
Results. Higher basal cortisol levels were observed in older age groups. Gender, recent life stress, and current distress
showed no relationship to basal levels. The magnitude of cortisol responses to the speech task differed by age, with the
smallest responses in the oldest group. This pattern was robust in men, with the youngest subjects (40-59 yr) showing
both the largest and the most prolonged responses. While women > 70 yr were least likely to show any response, other
analyses failed to show age effects on reactivity in women, perhaps because anticipatory baseline elevations limited
subsequent cortisol response.
Conclusions. Results indicate moderate increases in basal cortisol levels, but do not support the hypothesis that cortisol
responses to a stressor increase in magnitude or duration during normal human aging. Gender differences in stress
reactivity warrant further investigation.
T
HE hypothalamic-pituitary-adrenal (HPA) axis plays a
vital role in the regulation of physiological processes (1).
Although it is generally accepted that homeostatic mechanisms become less efficient in the aging organism, evidence
for changes in HPA function in the course of normal human
aging is mixed. In the majority of studies, basal levels of
cortisol show little change (2-11); a few studies, however,
report significantly higher (12,13) or lower (14-18) basal
cortisol in older subjects. Due to changes in circadian rhythmicity and sleep processes, cortisol secretion is particularly
likely to show age-related increases in the evening and the
early hours of the night (19,20).
Challenge tests assess the functional integrity of the feedback system at various levels of the axis. Results with
respect to human aging have recently been reviewed (21).
Adrenocorticotropic hormone (ACTH) challenge studies
have shown no significant age differences in initial cortisol
response. Following administration of dexamethasone
(DEX), a synthetic glucocorticoid, cortisol responses are
either unchanged or increased in older subjects. Corticotropin releasing hormone (CRH) infusion produces similar
ACTH and cortisol responses in older and younger subjects,
although the timing of the peak response may vary. Combined DEX/CRH challenge tests may reveal more subtle
changes; using this procedure, a recent study found agerelated increases in HPA activity, especially in women (22)..
Cortisol responses to hormonal or pharmacologic challenges are not necessarily correlated with responses to stressors, nor do all classes of stressors have the same effect. In a
M68
study in young adults, for example, males and females did
not differ in cortisol response to CRH administration or to
physical stressors, but males responded to psychosocial
stressors with cortisol elevations twice as high as those
observed in females (23). As another example, cortisol
responses to CRH challenge and to a simulated driving task
were uncorrelated in a study of healthy 70-year-olds (24).
Particularly under circumstances in which individual differences in appraisal and coping processes are brought into
play, low correlations between the functional capacity of the
HPA system and the observed response to experimental
stressors can be expected. To the extent that such stressors
can serve as models for how individuals respond to real-life
stress, they are an important complement to biological challenge tests. Surprisingly little research has been done on agerelated changes in cortisol response to acute physical or
psychological stressors in humans. Age appears to have no
effect on cortisol response to cold exposure (16) or to the
cold pressor test (25). The cortisol response to insulininduced hypoglycemia is also similar in young and elderly
subjects (5). Older patients showed increased cortisol responses to trauma in one study (26), but not in another (27).
The majority of challenge or reactivity studies have focused
on the response amplitude only. In rats, however, it appears
that aging has little effect on the rate or amplitude of the
initial glucocorticoid response, but is associated with delayed post-stress recovery to basal levels (28,29). Delayed
termination of HPA activity following stress is thought to
reflect deficits in glucocorticoid feedback regulation (30).
BASAL CORTISOL AND STRESS REACTIVITY
The generalizability of the above findings in rodents to
humans remains unclear.
The goals of the current study were twofold. The first was
to clarify whether basal free cortisol levels, measured in
saliva, change in the course of normal human aging. The
second goal was to determine whether there are age-related
differences in patterns of acute cortisol reactivity to a psychosocial stressor, a laboratory speech task. Differences in
the magnitude as well as the time course of the cortisol
response were examined. Because there is considerable
evidence for gender differences in cortisol reactivity (22,23)
and at least one report of differences in basal levels (17), the
effects of gender (both independently and in relation to age)
were specifically assessed. We also examined possible influences on basal cortisol measures of recent life stress,
psychological symptomatology, coping style, and body
mass, and the relationship between emotional state and
cortisol response to the speech task.
METHODS
Subjects
Subjects were selected in three age groups (1:40-59 yr, II:
60-69 yr, III: 5= 70 yr), with approximately equal numbers
of males and females. Sixty-four subjects from a pool of
volunteers for studies of cognitive changes in aging were
approached by telephone. Of these, four were excluded due
to diabetes, thyroid abnormalities, or use of medications that
could influence cortisol levels; two refused; and two could
not participate on the planned dates. The remaining subjects
reported being in good health.
Age group I consisted of 6 men and 10 women, ranging in
age from 43 to 58 yr (mean = 50 yr). Three subjects smoked
cigarettes. Four of the women had regular menstrual cycles,
one of whom used oral contraceptives; the remaining six
women were postmenopausal. Age group II included 10 men
and 10 postmenopausal women, ranging in age from 62 to 69
yr (mean = 65.5 yr). Four subjects smoked. Age group III
originally included 10 men and 10 postmenopausal women;
one male subject was later excluded from the analysis due to
abnormal cortisol values (see Statistical Analysis, below).
The remaining 19 subjects ranged in age from 70 to 86 yr
(mean = 77.7 yr). Two subjects in age group III smoked. In
the total sample of 55 subjects, mean educational level,
measured on an 8-point scale from primary school to higher
vocational training and university, was 3.9 (range 1-8),
with no significant age group [F(2,49) = .06, p = .95] or
gender [F( 1,49) = 1.71,/? = .20] differences. Body mass
index (BMI = kg/m2, calculated from self-reported weight
and height) ranged from 18.4 to 38.2 (mean = 24.6) and did
not differ by age group [F(2,49) = 1.22, p = .30] or by
gender [F( 1,49) = .01, p = .92].
Assessment of Basal Cortisol Levels
Subjects were visited at home, where informed consent
was obtained, questionnaires completed, and instructions
given for saliva sample collection. At prearranged times in
the morning (8 a.m.), late afternoon (4 p.m.), and late
evening (9 p.m.) on two consecutive days, subjects collected
saliva samples with a cotton dental roll, which was stored in
M69
a capped plastic vial ("Salivette," Sarstedt, Rommelsdorf,
Germany). Subjects stored samples in their home freezers
until coming to the laboratory, where uncentrifuged samples
were frozen at -20 °C until analysis.
Assessment of Stress Reactivity
In the week following basal cortisol sampling at home,
cortisol and emotional responses to an experimental stressor
were measured in the laboratory during a speech task (StressInducing Speech Task, SIST), in which subjects were asked
to prepare (10 min) and deliver (5 min) a speech concerning
their best and worst personal characteristics. Subjects were
told that their presentations would be videotaped for later
analysis by a team of psychologists. This task has been shown
to induce cortisol elevations within 15 min (31).
The SIST took place between 3 p.m. and 5 p.m. Cortisol
levels are generally fairly stable at this time of day, with no
significant change due to circadian rhythm effects within the
2-hr interval. Saliva was collected before subjects received
standardized instructions concerning the SIST (tO), immediately after the SIST (t20: 20 min after tO), and twice (t40,
t60) in a recovery period, during which subjects read neutral
magazines. At the same time points, emotional reactivity to
the SIST was assessed with 13 mood items presented as 10
cm visual analogue scales. The items nervous, anxious,
tense, irritated, dejected, worried, and disappointed formed
the negative affect (NA) scale (Cronbach's alpha = .74); the
items happy, confident, enthusiastic, satisfied, exhilarated,
and relieved formed the positive affect (PA) scale (alpha =
.81). Item scores were summed and divided by the number
of items to produce a score for each mood scale ranging from
OtolO.
Assessment of Life Stress and Psychological Characteristics
The following questionnaires were completed to assess
anxiety, depression, stress, and coping:
Anxiety: Trait anxiety was measured with the Dutch version of the State-Trait Anxiety Inventory (32).
Depression: Depressive symptomatology was assessed
with the Dutch version of the Zung self-report scale (33).
Life events: The LTE-Q is a list of 12 major life events,
such as the death of a family member, divorce, or serious
illness (34). Subjects indicated whether each type of event
had occurred in the past year and, if so, in which month.
Chronic stress: The Groningen List of Long-term Difficulties (GLLM) is a questionnaire concerning chronic
stress in domains such as work, housing, finances, physical
health, and social relations. Most of the items are scored on
4-point scales (from 1 = no difficulty to 4 = severe
difficulties), with a smaller number of problems scored as
present/absent (35).
Perceived stress: The Perceived Stress Scale (PSS, 10item version) (36), is a global measure of the extent to which
individuals appraise their lives in the past month as stressful,
reflecting aspects such as unpredictability, uncontrollability,
and overload.
Coping: Habitual coping style was measured with the
Utrecht Coping List (37). The 47 items are grouped into
seven subscales: problem-focused, palliative behaviors/
M70
NICOLSON ET AL.
distraction, avoidance, social support seeking, expression of
emotions/anger, comforting cognitions, and depressive
reaction/giving up.
age group:
Biochemical Analysis
Salivary cortisol has been shown to be highly correlated
with plasma or serum levels; it is largely unbound and thus
represents the "free," biologically active fraction of the
hormone (38). Salivary cortisol levels were determined in
duplicate by direct radioimmunoassay, using l25I-cortisol
(Farmos Diagnostica, Finland) and antiserum made against
the 3-CMO-BSA conjugate. The lower detection limit of the
assay was .33 nmol/1, with a mean intra-assay coefficient of
variation of 4.8%. Multiplication by 36.2 converts cortisol
values from nmol/1 to ng/dl.
Statistical Analysis
Data from one subject (male, 79 yr) were excluded from
the analysis because cortisol levels in all 10 samples (ranging
from 64 to 214 nmol/1) were far outside the normal range,
exceeding levels reported for Cushing's disorder. Cortisol
values were transformed to natural logarithms to normalize
the distributions, which were positively skewed in samples
collected in the laboratory, and to meet requirements of
homoscedasticity for the statistical models used.
Cortisol levels in samples collected at home at the same
time of day on two consecutive days did not significantly
differ (paired /-tests; /?-values all > .05); the mean of the two
measures was therefore used in the analysis of basal secretory patterns. Multivariate analysis of variance (MANOVA)
was used to test main and interaction effects of the factors
age (three groups: I, II, and III) and gender. Contributions of
body mass, life stress, anxiety, depression, and coping
measures to the variance in total cortisol secretion (defined
as the sum of 8 a.m., 4 p.m., and 9 p.m. levels) were
estimated, together with the effects of age as a continuous
variable, by forward stepwise multiple regression. To test
whether cortisol levels might increase exponentially in the
oldest subjects, a quadratic term for age was included in the
equation.
Cortisol peak and mean responses to the SIST were
defined as changes from prespeech (tO) baseline levels. Ageand gender-related differences in peak response were tested
with MANOVA. Age and gender effects on the time course
of responses to the SIST were tested with repeated measures
MANOVA, using Hotelling's probability estimates. The
influence of anticipatory elevation in lab baseline cortisol on
peak and mean responses was tested with forward stepwise
multiple regression.
Statistical tests were performed with SPSS, Macintosh
version. Unless otherwise noted, significance levels are based
on two-tailed tests, with/? *s .05 considered significant.
RESULTS
Basal Cortisol Levels
As shown in Figure 1, increasing age was associated with
higher salivary cortisol levels at all three times of day.
Repeated measures MANOVA revealed a statistically significant main effect for age [F(2,49) = 5.07, p = .01], but
8 am
4 pm
9 pm
time of day
Figure 1. Basal salivary cortisol levels in morning, afternoon, and
evening samples. Boxes show the median and interquartile range for each
age group (I: 40-59 yr; II: 60-69 yr; III: 2s 70 yr), with whiskers extending
from the 10th to the 90th percentile. Black symbols represent mean levels.
not for either gender [F( 1,49) = .10,/? = .75] or the Age by
Gender interaction [F(2,49) = .05, p = .95]. Time of day
had a strong effect on cortisol levels [F(2,48) = 366.31, p =£
.001], reflecting the known circadian secretory pattern. A
similar diurnal pattern was observed in all three age groups
and in both sexes: Time of Day by Age [F(4,94) = 1.11,/?
= .36], Time of Day by Gender [F(2,48) = .30,/? = .74],
and the three-way interaction term [F(4,94) = .09,/? = .99]
were all nonsignificant.
Over all subjects, age (yr) was significantly correlated
with basal cortisol at each time of day (8 a.m.: r = .30, p =s
.05;4p.m.:r = .31,/?=s .05;9p.m.: r = .39,/?^ .01). We
next examined whether the age effect might reflect differences in stress, distress, or coping measures. The three age
groups did not differ in number of recent life events or
perceived stress, but chronic difficulties tended to decrease
with age, with total scores on the GLLM Likert-scale items
averaging 23.24, 20.75, and 20.05 in age groups I, II, and
III, respectively (Kruskal-Wallis test, p = .06). No differences were found on depression, anxiety, and coping measures, with the exception of the coping style "expression of
emotions/anger," which decreased with age (means of 6.87,
5.75, and 5.16 for groups I, II, and III; Kruskal-Wallis test,
p =s .05). Next, forward stepwise multiple regression was
performed with total cortisol as the dependent variable and
age, age2, body mass index, number of recent life events,
perceived stress, chronic difficulties, depression, anxiety,
and scores on the seven coping factors as independent
variables. Of these, only age was a significant predictor of
total cortisol (R2 = .19, p =£ .001); the quadratic age term
was nonsigificant, indicating that the age effect was linear.
To test whether the same pattern held true for both sexes,
M71
BASAL CORTISOL AND STRESS REACTIVITY
we repeated the regression analysis for males and females
separately. In males, age was again the only significant
predictor of total cortisol (R2 = .23, p =s .02). In females,
both age (Beta = .533, p = .0007) and avoidant coping
(Beta = -.607, p = .0002) were significant predictors of
basal levels, together explaining 54% of the variance in total
cortisol (p ^ .0001).
area-under-the-curve measure. MANOVA results indicated
a significant effect of age on mean response [F(2,49) =
4.16, p ^ .02]; here, as with peak response, gender had no
main effect [F(l ,49) = 3.03, p ^ .09], but there was a significant Age by Gender interaction [F(2,49) = 4.81, p =£
.01]. Gender differences in the age effect are shown in
Figure 2. In men, but not in women, a pattern of decreasing
response to the SIST with age can be observed.
Reactivity to the Speech Task
Cortisol responses. — Table 1 summarizes cortisol responses measured in the laboratory. Prespeech baseline was
elevated in comparison to 4 p.m. home level in 32 of the 55
subjects (means for the entire sample of 4.39 nmol/1 in the
laboratory vs 3.46 nmol/1 at home; Wilcoxon test, p =s .05),
probably due to the novelty of the laboratory setting or to
anticipatory anxiety. Neither lab baseline nor anticipatory
response (change in lab baseline relative to home level),
however, differed significantly by age or by gender.
The average peak response represented a 37% increase
over laboratory baseline and a 77% increase over home
level. Peak response differed by age [F(2,49) = 3.80, p =s
.03]; furthermore, although gender had no main effect
[F(l,49) = 2.89, p =s .10], there was a significant interaction between age and gender [F(2,49) = 5.85, p *£ .005].
Post-hoc examination of the age effect (one-way analysis of
variance) indicated that, while no two groups were statistically significantly different from one another, the oldest
group had the lowest cortisol peaks relative to the two
younger groups combined (based on separate variance estimates, 48.6 df, f-value = -2.54, p = .014).
Using a more stringent definition of "response" [peak
increment of > 2.76 nmol/1 (39)], we found no difference in
the percentage of responders in each age group (Chi-square
= .62, p = .73). The percentage of complete nonresponders (subjects with no increase in cortisol levels above lab
baseline), on the other hand, increased with age (Chi-square
= 6.63,/?=s .04).
The mean cortisol response is a measure that combines
both magnitude and duration; based on three samples obtained at equal (20 min) time intervals, it approximates an
Time course. — The results of the repeated measures
MANOVA confirmed age [F(2,49) = 4.16, p ^ .02] and age
by gender [F(2,49) = 4.81, p ^ .01] differences in the
overall cortisol response, with no significant effect of gender
alone [F(l,49) = 3.03, p = .09]. With respect to the time
course of the response, a significant effect of time [t20, t40,
t60; F(2,48) = 4.43, p ^ .02] and a Gender by Time
interaction effect [F(2,48) = 5.66, p =S .006] were found.
Age, in contrast, had no significant effect on the temporal
patterning of the cortisol response [F(4,94) = .42,/? = .80],
either alone or in interaction with gender [F(4,94) = 1.16,
p = . 34]. In Figure 3, it is evident that the cortisol response to
the SIST tended to peak later in men than in women. With a
late response defined as the observed cortisol peak at t40 or
t60, 14 of the 25 men compared to 7 of the 30 women showed
a late response (Chi-square = 6.17, p =s .05).
Effects of anticipatory cortisol elevation. — Next, we
evaluated the possibility that anticipatory changes in cortisol
laboratory baseline, as described above, may have influenced the subsequent response to the SIST, possibly
explaining or interacting with age effects. Across all subjects, a larger anticipatory response was associated with a
smaller mean SIST response (r = -.35, p =s .01); peak
response also tended to be lower, although not significantly
so (r = -.25, p < . 10). Home levels at 4 p.m., on the other
hand, showed no association with mean (r = -.06) or peak
(r = .03) responses.
Since MANOVA results had already shown that age
effects on cortisol responses were modified by gender,
forward stepwise regressions were performed separately for
male and female subsamples. Here, a new variable (Antici-
Table 1. Cortisol Response (nmol/1) to a Speech Task by Age Group and Gender
Nonresponset
n
Age Group
Subjects
n
tO Baseline
Mean (SD)
Peak*
Mean (SD)
1: 40-59 yr
Male
Female
6
10
3.24 (.75)
4.86(2.53)
3.33(1.79)
0.23 (2.40)
3
1
0
5
II: 60-69 yr
Male
Female
10
10
4.04(1.65)
3.55(1.30)
1.14(2.54)
2.34 (3.20)
2
3
4
1.
III: 3=70yr
Male
Female
9
10
6.24(7.17)
4.14(1.34)
1.64(3.17)
0.09(1.14)
2
1
4
8
Total
55
4.39(3.27)
1.32(2.63)
12
22
•Maximum cortisol change relative to baseline.
t P e a k > 2.76 nmol/1.
t P e a k < 0 nmol/1.
High Response!
n
NICOLSON ET AL.
M72
Age group:
-
4-
E 2i•
9 0
1
J
—1
C
.2-2-4-
-6
Ml
Mil
age group
Mill
t0
t20
t40
time (min)
B
Figure 2. Cortisol response to the speech task (SIST), averaged over t20,
t40, and t60, in each age group, plotted separately for men, A, and for
women, B. See legend for Figure 1 for further explanation.
patory response x Age) was computed to assess differential
effects of initial elevations on later response as a function of
age. In men, anticipatory responses had no effect on subsequent response to the SIST (mean or peak), either alone or in
interaction with age. Age remained a significant determinant
of mean cortisol response (R2 = .26, p «£ .01) in this
analysis but had no significant effect on peak response.
A different picture emerged for women. Here, only the
anticipatory response had a significant negative effect,
which did not differ as a function of age, on both mean
response (R2 = .23, p =S .01) and peak response (R2 = . 19,
p =s .05). Cortisol elevation in anticipation of the speech task
thus appears to have limited subsequent response.
Emotional responses. — To determine whether subgroup
differences in cortisol response to the SIST may have
reflected different patterns of emotional response to the task,
we first tested the significance of changes in negative (NA)
and positive affect (PA) during the SIST with repeated
measures MANOVA. Over all subjects, significant changes
in NA were observed over the four assessments [F(3,47) =
4.99, p =s .004]. NA tended to peak immediately after the
speech performance and to decline to its lowest level by the
t60
time (min)
Figure 3. Median cortisol responses to the speech task (SIST) in each age
group, plotted separately for men, A, and for women, B. Data are plotted as
changes from 4 p.m. home levels.
end of the recovery period (means of 1.35, 1.88, 1.38, and
1.17 at tO, t20, t40, and t60, respectively). For PA, changes
over time approached significance [F(3,47) = 2.31, p =
.09], with the lowest ratings prior to the SIST, rising after
the task and then remaining fairly stable throughout the
recovery period (means of 4.46, 5.07, 5.07, and 5.20 at tO,
t20, t40, and t60, respectively). Individual differences in
overall NA levels were unrelated to age [F(2,49) = 1.48,
p = .24], gender [F(l,49) = .28, p = .60], or their interaction [F(2,49) = .10, p = .90]. Similar nonsignificant
results were obtained for PA. More importantly, withinsubject patterns of change in NA and PA over time did not
differ significantly in males vs females or among the three
age groups, nor was the three-way (Time x Age X Gender)
interaction statistically significant. Thus, age or gender differences in emotional response to the SIST do not appear to
explain the effects of these variables on cortisol responses.
Indeed, we found no evidence for an association between
emotional state and cortisol response. Spearman correlations
between mean levels of NA during the SIST with peak (rho
BASAL CORTISOL AND STRESS REACTIVITY
= .09) and mean (rho = .18) cortisol responses were nonsignificant, as were those between mean PA with peak (rho
= .08) and mean (rho = .06) cortisol responses.
DISCUSSION
Basal Cortisol
This study provides evidence for age-related increases in
basal cortisol levels, remaining within the normal range, in
both males and females. This finding is consistent with the
results of some recent studies, but not with many others that
have reported no effect of aging on basal HPA activity. A
number of features of the current design may explain why
age effects were demonstrated. Firstly, it appears that age
differences in HPA activity are more likely to be found when
samples include individuals older than 70 yr (40); our sample
covered a relatively wide age range, with one-third of the
subjects older than 70. However, while post-hoc comparisons showed a significant difference between the > 70 yr
group and the combined younger groups only, regression
analysis suggested that the age effect on basal cortisol was
more or less linear over the entire sample. Secondly, discrepancies in the results of previous studies may be due to
differences in the way cortisol was measured. Depending on
whether cortisol is determined in plasma, urine, or saliva
samples and whether total levels, free levels, or metabolites
are assessed, different results may be obtained, even in the
same subjects (41,42). In particular, total cortisol levels can
be misleading if, for example, there are individual differences in level or activity of cortisol binding globulin. This,
in addition to the evidence that it is the biologically active
hormone fraction, makes free cortisol the measure of choice.
Salivary free cortisol has the additional advantage of easy,
stress-free measurement; this enabled us to obtain repeated
samples at different times of day in the home environment,
which may have provided a more reliable estimate of basal
levels.
We found no difference in basal cortisol levels between
males and females at any age. With the exception of reports
of higher 4 p.m. salivary cortisol levels in men (43) and
higher cortisol in old women than in old men (17), there
appears to be little evidence for significant gender differences in basal levels in healthy subjects.
Basal cortisol levels did not appear to be influenced by
previous stress exposure or psychological distress. An avoidant coping style, characterized by a tendency to avoid
stressful situations and let events take their own course (37),
was associated with significantly lower basal cortisol levels
in women. We have no explanation for this finding and
suggest that it be interpreted with caution, given the number
of independent variables in the regression equation relative
to the number of subjects. Two additional factors should be
considered with respect to possible psychosocial influences.
Firstly, subjects in this study were healthy volunteers, with
limited variability in measured life events or psychological
symptomatology. Secondly, only recent life events, chronic
difficulties, and perceived stress were assessed, whereas
stress exposure over the lifetime might be expected to have
greater cumulative effects on HPA axis regulation. Although
M73
difficult to conduct, prospective studies of the effects of
chronic stress in aging humans would be informative.
What are the clinical implications of increased basal
cortisol secretion with aging? In the current study, observed
elevations in cortisol with age were moderate, with total
levels in the oldest group on average 38% higher than those
in the youngest group. Clearly, age-related changes are not
necessarily pathological. Mild cortisol elevation could, for
example, simply reflect normal changes in sleep patterns
with aging. On the other hand, it has been postulated (20,29)
that age-related changes in HPA activity may ultimately be
markers of dysfunction in higher central nervous centers,
including the hippocampus, hypothesized to be involved in
the cognitive decline often observed in human aging. In
addition to evidence from animal models, results from studies in humans point to possible detrimental effects of even
slight increases in cortisol levels over time. For example, in
a recent study in which basal cortisol levels were prospectively monitored over several years in healthy elderly, a
pattern of increasing levels over time, within the normal
range, was associated with a decline in specific cognitive
functions (44). Such findings underscore the need for longitudinal studies in both healthy and patient populations.
Cortisol Reactivity
The magnitude of total and peak cortisol responses to the
speech task varied in the three age groups, with the oldest
subjects (> 70 yr) tending to show the smallest responses.
These results contrast with those of Gotthardt et al. (45),
who found greater cortisol responses to a cognitive challenge
in older (> 60 yr) vs younger (< 30 yr) subjects. The current
finding is also unexpected in light of results of recent studies
showing increased reactivity to hormonal (22) and pharmacologic (46) challenges in older subjects. While the ageassociated elevation in basal cortisol measured at home
could theoretically result in diminished acute reactivity, as
has been suggested in depression (45), we found no relationship between 4 p.m. home levels and either peak or mean
response to the SIST. However, anticipatory elevation in
cortisol levels prior to the speech task, in relation to home
levels at the same time of day, complicates the interpretation
of the results. While the pattern of decreasing cortisol
response to the speech task with increasing age was confirmed for men, in whom anticipation appeared to have a
negligible effect, women with high anticipatory elevation
had blunted cortisol responses to the speech task. It is
impossible to assess the extent to which anticipatory responses may have obscured or confounded the results of
previous laboratory stress reactivity studies, since home
basal levels have very rarely been measured [see (47) for an
illuminating exception]. Without these, one is unable to
determine whether lengthier acclimation procedures do indeed have the desired effect of stabilizing baseline cortisol.
Despite the statistically significant age effect on reactivity, it is important to note the marked individual differences,
particularly in the oldest group; while the majority (12 of the
19 subjects, including 8 of the 10 women) showed no poststress cortisol elevation, high (> 2.76 nmol/1) peak responses were observed in three individuals, including the
oldest subject (male, 86 yr). Another elderly male (79 yr)
M74
NIC OLSON ET AL.
was excluded from the analysis due to consistent, abnormally high cortisol levels. Increasing heterogeneity in HPA
activity (either basal or stimulated), even in the absence of
changes in mean levels, may reflect vulnerability toward
dysfunction in human aging (24,48).
Age effects interacted with gender; in particular, men in
the youngest age group (40-59 yr) showed relatively large
cortisol responses, which also tended to peak later in time,
and the oldest women, as mentioned above, were the least
likely to show any response to the SIST. Greater cortisol
reactivity to psychosocial stress in males compared to females has previously been reported in young adults (23) and
in human neonates (49), but not, to our knowledge, in
middle-aged or older subjects. In other studies, healthy
elderly women have been reported to show larger cortisol
responses to CRH administration (17) and to a combined
dexamethasone suppression/CRH-stimulation test (22) than
men. Gender effects therefore warrant further investigation.
In addition, possible effects of estrogen replacement on
salivary cortisol responses to stress should be examined (5052); we did not have sufficiently reliable information on
estrogen treatment to address this issue.
With respect to the hypothesis that aging is associated
with delayed post-stress recovery of cortisol to basal levels,
small or absent responses to the SIST in many of the subjects
preclude definite conclusions about response duration. In
this regard, it is important to note that the amplitude and
duration of the responses in the youngest group of men were
pronounced, similar to those observed in previous studies
employing the same stressor (31). Nevertheless, different
types of psychosocial challenges may be needed to reveal
age differences in HPA system recovery from acute stress. In
a pilot study of ACTH and cortisol responses to a driving
simulation test, for example, healthy older subjects did show
a prolonged post-stress elevation of cortisol levels compared
to younger subjects [Seeman et al., cited in (21)], while
Gotthardt et al. (45) found only slightly delayed cortisol
recovery in old vs young subjects following a cognitive
challenge. In addition, more frequent cortisol sampling
throughout stressor exposure and recovery may be necessary
to resolve subtle age-related differences.
Although age and gender may have influenced the subjective experience of the speech task, we found no differences
in emotional responses or in trait characteristics that would
support this hypothesis. Future studies with measures of
cortisol and ACTH responses to CRH challenge and psychosocial stressors in the same subjects may help clarify whether
age-related differences in stress reactivity reflect functional
changes in the HPA axis and, if so, at what level.
Conclusion
In summary, our study provides evidence for agedependent increases in basal HPA activity, as reflected in
salivary cortisol secretion; the mechanisms underlying such
changes and their significance in healthy human subjects
remain to be clarified. The hypothesized delay in recovery of
cortisol to basal levels following acute stress was not supported. In general, older subjects showed decreased rather
than increased cortisol responses to an acute stressor. Because gender modified both the relationship between age and
cortisol response magnitude and the time course of the
response, future studies of HPA reactivity in aging should
examine the responses of both males and females.
ACKNOWLEDGMENTS
This study was supported in part by the Institute of Brain and Behavior,
University of Maastricht.
Address correspondence to Dr. Nancy Nicolson, Department of Psychiatry and Neuropsychology, Social Psychiatry and Psychiatric Epidemiology
Section, University of Maastricht, P.O. Box 616, 6200 MD Maastricht,
The Netherlands. E-mail: [email protected]
REFERENCES
1. Munck A, Guyre PM, Holbrook NJ. Physiological functions of glucocorticoids in stress and their relation to pharmacological actions.
EndocrRev 1984;5:25-44.
2. Hunt GE, Johnson GFS, Caterson ID. The effect of age on cortisol and
plasma dexamethasone concentrations in depressed patients and controls. J Affect Disord 1989; 17:21-32.
3. Barton RN, Horan MA, Weijers JWM, Sakkee AN, Roberts NA, van
Bezooijen CFA. Cortisol production rate and the urinary excretion of 17hydroxycorticosteroids, free cortisol, and 6B-hydroxycortisol in healthy
elderly men and women. J Gerontol Med Sci 1993;48:M213-8.
4. Blichert-Toft B, Blichert-Toft M. Adrenocortical function in the aged,
assessed by the rapid corticotrophin test (Synacthen®). Acta Endocrinol 1970;64:410-4.
5. Cartlidge NEF, Black MM, Hall MRP, Hall R. Pituitary function in the
elderly. Gerontol Clin 1970; 12:65-70.
6. Jensen HK, Blichert-Toft M. Serum corticotropin, plasma cortisol and
urinary excretion of 17-ketogenic steroids in the elderly (age group:
66-94 years). Acta Endocrinol 1971 ;66:25-34.
7. Murray D, Wood PJ, Moriarty J, Clayton BE. Adrenocortical function
in old age. J Clin Exp Gerontol 1981 ;3:255-68.
8. Olsson T, Viitanen M, Hagg E, et al. Hormones in "young" and
"old" elderly: pituitary-thyroid and pituitary-adrenal axes. Gerontology 1989,35:144-52.
9. Waltman C, Blackman MR, Chrousos GP, Riemann C, Harman SM.
Spontaneous and glucocorticoid-inhibited adrenocorticotropic hormone and cortisol secretion are similar in healthy young and old men. J
Clin Endocrinol Metab 1991 ;73:495-502.
10. West CD, Brown H, Simons EL, Carter DB, Kumagai LF, Englert E.
Adrenocortical function and cortisol metabolism in old age. J Clin
Endocrinol Metab 1961;21:1197-207.
11. Pavlov EP, Harman SM, Chrousos GP, Loriaux DL, Blackman MR.
Responses of plasma adrenocorticotropin, cortisol, and dehydroepiandrosterone to ovine corticotropin-releasing hormone in healthy aging
men. J Clin Endocrinol Metab 1986;62:767-72.
12. Armanini D. Corticosteroid receptors in lymphocytes: a possible
marker of brain involution? J Steroid Biochem Mol Biol 1994;49:
429-34.
13. Rozenberg S, Bosson D, Peretz A, Caufriez A, Robyn C. Serum levels
of gonadotrophins and steroid hormones in the post-menopause and
later life. Maturitas 1988;10:215-24.
14. Drafta D, Schindler AE, Stroe E, Neacsu E. Age-related changes of
plasma steroids in normal adult males. J Steroid Biochem
1982;17:683-7.
15. Maes M, Calabrese J, Lee M, Meltzer HY. Effects of age on spontaneous cortisolaemia of normal volunteers and depressed patients. Psychoneuroendocrinology 1994; 19:79-84.
16. Wagner JA, Horvath SM, Kitagawa K, Bolduan NW. Comparisons of
blood and urinary responses to cold exposures in young and older men
and women. J Gerontol 1987;42:173-9.
17. Greenspan SL, Rowe JW, Maitland LA, McAloon-Dyke M, Elahi D.
The pituitary-adrenal glucocorticoid response is altered by gender and
disease. J Gerontol Med Sci 1993;48:M72-7.
18. Sharma M, Palacios-Bois J, Schwartz G, et al. Circadian rhythms of
melatonin and cortisol in aging. Biol Psychiat 1989;25:305-19.
19. Copinschi G, Van Cauter E. Effects of ageing on modulation of
hormonal secretions by sleep and circadian rhythmicity. Horm Res
1995;43:20-4.
BASAL CORTISOL AND STRESS REACTIVITY
20. Dodt C, Theine KJ, Uthgenannt D, Born J, Fehm HL. Basal secretory
activity of the hypothalamo-pituitary-adrenocortical axis is enhanced
in healthy elderly. An assessment during undisturbed night-time sleep.
Eur J Endocrinol 1994; 131:443-50.
21. Seeman TE, Robbins RJ. Aging and hypothalamic-pituitary-adrenal
response to challenge in humans. Endocr Rev 1994; 15:233-60.
22. Heuser U, Gotthardt U, Schweiger U, et al. Age-associated changes of
pituitary-adrenocortical hormone regulation in humans: importance of
gender. Neurobiol Aging 1994; 15:227-31.
23. Kirschbaum C, Wiist S, Hellhammer D. Consistent sex differences in
cortisol responses to psychological stress. Psychosom Med
1992;54:648-57.
24. Seeman TE, Berkman LF, Gulanski BI, et al. Self-esteem and neuroendocrine response to challenge: MacArthur studies of successful
aging. J Psychosom Res 1995;39:69-84.
25. Casale G, Pecorini M, Cuzzoni G, de Nicola P. Beta-endorphin and
cold pressor test in the aged. Gerontology 1985;31:101-5.
26. Desai D, March R, Watters J. Hyperglycemia after trauma increases
with age. J Trauma 1989;29:719-23.
27. Stoner HB, Frayn KN, Barton RN, et al. The relationships between
plasma substrates and hormones and the severity of injury in 227
recently injured patients. Clin Sci 1979;56:563-73.
28. Odio M, Brodish A. Age-related adaptation of pituitary-adrenocortical
responses to stress. Neuroendocrinology 1989;49:382-8.
29. Sapolsky RM, Krey LC, McEwen BS. The neuroendocrinology of
stress and aging: the glucocorticoid cascade hypothesis. Endocr Rev
1986;7:284-301.
30. de Kloet ER. Brain corticosteroid receptor balance and homeostatic
control. Front Neuroendocrinol 1991; 12:95—164.
31. van Eck MM, Nicolson NA, Berkhof H, Sulon J. Individual differences in cortisol responses to a laboratory speech task and their
relationship to responses to stressful daily events. Biol Psychol
1996;43:69-84.
32. van der Ploeg HM, Defares PB, Spielberger CD. Zelf-beoordelingsvragenlijst STAI versie DY. Lisse, The Netherlands: Swets &
Zeitlinger, 1979.
33. Dijkstra P. De zelf-beoordelingsschaal voor depressie van Zung. In:
van Praag HM, Rooymans HGM, eds. Stemming en Ontstemming.
Amsterdam: Erven Bonn, 1974:98-120.
34. Brugha T, Bebbington P, Tennant C, Hurry J. The list of threatening
experiences: a subset of 12 life event categories with considerable longterm contextual threat. Psychol Med 1985; 15:189-94.
35. Hendriks AAJ, Ormel J, van de Willige G. Langdurige moeilijkheden
gemeten volgens zelfbeoordelingsvragenlijst en semi-gestructureerd
interview. Gedrag en Gezondheid 1990; 18:273-83.
36. Cohen S, Williamson GM. Perceived stress in a probability sample of
the United States. In: Spacepan S, Oskamp S, eds. The social psychology of health. Newbury Park, CA: Sage Publications, 1988:31-67.
37. Schreurs PJG, van de Willige G. De Utrechtse Copinglijst (UCL).
Lisse, The Netherlands: Swets & Zeitlinger, 1988.
M75
38. Vining RF, McGinley RA. Hormones in saliva. Crit Rev Clin Lab Sci
1986;23:95-145.
39. Kirschbaum C, Hellhammer DH. Salivary cortisol in psychobiological
research: an overview. Neuropsychobiology 1989;22:150-69.
40. Sapolsky RM, Plotsky PM. Hypercortisolism and its possible neural
bases. Biol Psychiat 1990;27:937-52.
41. Touitou Y, Sulon J, Bogdan A, et al. Adrenal circadian system in
young and elderly human subjects: a comparative study. J Endocrinol
1982;93:201-10.
42. Touitou Y, Sulon J, Bogdan A, Reinberg A, Sodoyez J, DemeyPonsart E. Adrenocortical hormones, ageing and mental condition:
Seasonal and circadian rhythms of plasma 18-hydroxy-11 -deoxycorticosterone, total and free cortisol and urinary corticosteroids. J Endocrinol 1983;96:53-64.
43. Brandtstadter J, Baltes-Gotz B, Kirschbaum C, Hellhammer D. Developmental and personality correlates of adrenocortical activity as indexed by salivary cortisol: observations in the age range of 35 to 65
years. J Psychosom Res 1991 ;35:173-85.
44. Lupien S, Lecours AR, Lussier I, Schwartz G, Nair NPV, Meaney MJ.
Basal cortisol levels and cognitive deficits in human aging. J Neurosci
1994; 14:2893-903.
45. Gotthardt U, Schweiger U, Fahrenburg J, Lauer CJ, Holsboer F,
Heuser I. Cortisol, ACTH, and cardiovascular response to a cognitive
challenge paradigm in aging and depression. Am J Physiol
1995;268:R865-73.
46. Peskind ER, Raskind MA, Wingerson D, et al. Enhanced hypothalamic-pituitary-adrenal axis responses to physostigmine in normal
aging. J Gerontol Med Sci 1995;5OA:M114-20.
47. Larson MC, Gunnar MR, Hertsgaard L. The effects of morning naps,
car trips, and maternal separation on adrenocortical activity in human
infants. Child Dev 1991;62:362-72.
48. Sapolsky R, Armanini M, Packan D, Tombaugh G. Stress and glucocorticoids in aging. Endocrinol Metab Clin North Am 1987; 16:
965-80.
49. Davis M, Emory E. Sex differences in neonatal stress reactivity. Child
Dev 1995 ;66:14-27.
50. Collins A, Hanson V, Eneroth P, Hagenfeldt K, Lundberg U,
Frankenhaeuser M. Psychophysiological stress responses in postmenopausal women before and after hormonal replacement therapy. Hum
Neurobiol 1982;1:153-9.
51. Lindheim SR, Legro RS, Bernstein L, et al. Behavioral stress responses in premenopausal and postmenopausal women and the effects
of estrogen. Am J Obstet Gynecol 1992;167:1831-6.
52. Kirschbaum C, Pirke KM, Hellhammer DH. Preliminary evidence for
reduced cortisol responsivity to psychological stress in women using
oral contraceptive medication. Psychoneuroendocrinology 1995;
20:509-14.
Received January 8, 1996
Accepted July 26, 1996