Psychoneuroendocrinology (2011) 36, 173—181

a v a i l a b l e a t w w w. s c i e n c e d i r e c t . c o m

j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / p s y n e u e n

Cortisol reactivity to social stress in adolescents:

Role of depression severity and child maltreatment
Kate L. Harkness a,*, Jeremy G. Stewart a,1, Katherine E. Wynne-Edwards b,2

a
Department of Psychology, Queen’s University, Kingston, ON K7L 3N6, Canada
b
Faculty of Veterinary Medicine, University of Calgary, HRIC 1AC68, 3330 Hospital Drive NW, Calgary AB T2N 4N1, Canada

Received 15 December 2009; received in revised form 14 May 2010; accepted 6 July 2010

KEYWORDS Summary This study examined the hypothesis that depressed adolescents with a history of
Cortisol; childhood maltreatment will show greater cortisol reactivity to psychological stress challenge
Depression; than those without, and this relation will be moderated by level of depression severity. Seventy-
Child maltreatment; one adolescents were exposed to the Trier Social Stress Test. Salivary cortisol was assessed at
Adolescence; baseline, immediately before the challenge, after the challenge, and during an extended
Trier Social Stress Test; recovery period. Childhood maltreatment was assessed with a rigorous contextual interview
Severity and rating system. Adolescents with a history of maltreatment produced higher and more
prolonged levels of cortisol in response to the challenge than did adolescents with no maltreat-
ment, but only among those with a mild/moderate level of depression severity. Those with
moderate/severe depression exhibited a blunted cortisol response regardless of child maltreat-
ment history. These findings indicate that depression is a heterogeneous syndrome, and that both
depression severity and child maltreatment history should be considered in studies examining
biological stress reactivity.
# 2010 Elsevier Ltd. All rights reserved.

disorder (MDD) (Carroll et al., 1981; Holsboer, 2000). Further-

1. Introduction
Forty years of research has consistently documented dysre-
gulation of hypothalamic-pituitary-adrenal (HPA) axis func-
tioning in the pathophysiology of adult major depressive
* Corresponding author. Tel.: +1 613 533 2886; fax: +1 613 533 2499.
E-mail addresses: harkness@queensu.ca (K.L. Harkness),
2js4@queensu.ca (J.G. Stewart), k.wynne-edwards@ucalgary.ca
(K.E. Wynne-Edwards).
1
Tel.: +1 613 533 6003; fax: +1 613 533 2499.
2
Tel.: +1 403 210 6331; fax: +1 403 210 9550.
more, a recent meta-analysis of HPA axis function in 41
studies employing primarily adolescent samples reported
that adolescents with MDD show greater cortisol production
(or less suppression) after the dexamethasone suppression
test (DST; d = .57), as well as higher basal cortisol levels
(d = .20), in comparison to non-depressed groups (Lopez-
Duran et al., 2009).
However, the DSTand basal cortisol have been criticized as
indices of HPA axis functioning because they fail to assess the
endogenous response of the HPA axis to psychological stress
(Rao et al., 2008). Therefore, more recently investigators
have used psychological challenge paradigms such as the
Trier Social Stress Test (TSST) (Kirschbaum et al., 1993).
The TSST is one of the few available stress protocols that

0306-4530/$ — see front matter # 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.psyneuen.2010.07.006
174 K.L. Harkness et al.

combines elements of uncontrollability and high levels of interaction, to cortisol reactivity and total cortisol expo-
social-evaluative threat. Stress tasks containing these two sure in response to the TSST. We hypothesized that (a)
components are associated with the largest HPA axis stress consistent with the adult literature, adolescents with
responses and the longest recovery times (Dickerson and MDD will show blunted cortisol reactivity and lower total
Kemeny, 2004). cortisol output in response to the TSST than non-depressed
Meta-analytic results in adults have indicated that controls, particularly among those with a severe depres-
depressed individuals show blunted cortisol reactivity in sion; (b) again, consistent with previous literature, adoles-
response to psychological challenge, particularly in older cents with a history of childhood maltreatment will show
and more severely depressed individuals (Burke et al., greater cortisol reactivity and total cortisol exposure in
2005). Only one study to our knowledge has compared response to the TSST than those without; and (c) depression
depressed versus non-depressed adolescents using the TSST severity will fully moderate the relation of childhood
(Rao et al., 2008), and it found elevated and prolonged cortisol maltreatment to cortisol reactivity: heightened cortisol
secretion in response to the TSST in depressed adolescents. In reactivity to the TSST in those with a history of child
younger children, studies have found blunted cortisol reactiv- maltreatment will only be seen in the less severely
ity in depressed children relative to controls (Luby et al., 2003, depressed adolescents, whereas blunted cortisol reactivity
2004). Several investigators have argued, however, that a will be seen in those with a history of child maltreatment in
distinction needs to be made between child and adolescent the context of severe depression. The latter hypothesis is
MDD, particularly with respect to HPA axis function (Kaufman tentative given the lack of previous research examining the
and Charney, 2003; Kaufman et al., 2001). Therefore, results relation of childhood maltreatment to HPA axis function
across these two sets of studies may not be comparable. moderated by depression severity.
Nevertheless, what is consistent across most studies using
challenge paradigms in adult and pediatric MDD is a much 2. Methods and materials
larger variability in cortisol response within the depressed
group versus the non-depressed individuals (Lopez-Duran
2.1. Participants
et al., 2009). It is this heterogeneity within MDD that may
help to account for inconsistencies in results across studies.
Participants were 71 individuals (48 females) ages 12—21
Two important sources of individual variability that have
years (M = 15.39, SD = 2.11) recruited from a mid-sized com-
consequences for the HPA axis system are (a) a history of
munity in Ontario, Canada. The ethnic diversity of our sample
childhood maltreatment, and (b) individual differences in
was consistent with that of the community (89% European
depression severity. On the one hand, research in both adult
ancestry). Our depressed sample was drawn from community
and child/adolescent samples has found that childhood mal-
mental health agencies and community advertisements seek-
treatment is associated with increased cortisol reactivity in
ing adolescents with depression. Our non-depressed sample
response to biological (e.g., corticotropin-releasing hormone
was recruited from local high schools and community adver-
[CRH]) and psychological stress challenge (De Bellis et al.,
tisements seeking healthy adolescents. Adolescents whom
1994; Heim et al., 2000, 2001, 2002; Kaufman et al., 1997;
we identified as depressed from the self-referrals were
Rao et al., 2008;). On the other hand, there is a literature in
referred for mental health treatment.
adults linking severe depression to blunted cortisol reactivity
The depressed adolescents all met Diagnostic and Statis-
response to stress challenge (Burke et al., 2005). Severe
tical Manual of Mental Disorders (DSM-IV) (American Psychia-
melancholic depression, in particular, has been linked to
tric Association, 1994) criteria for a current non-bipolar, non-
biological markers in the HPA axis (e.g., nonsuppression on
psychotic mood disorder. Exclusion criteria included the
the DST), and HPA axis parameters appear to be among the
presence of a psychotic disorder, bipolar disorder, substance
strongest discriminators of melancholic versus non-melan-
dependence, conduct disorder, developmental disability, or
cholic depression (Akil et al., 1993; Carroll et al., 1980; Dinan
medical disorder that could cause depression. Adolescents in
and Scott, 2005; Tsigos and Chrousos, 2002; Zimmerman
the non-depressed group were free of all current or past
et al., 1985).
psychiatric diagnosis.
It is noteworthy that these two lines of research reveal an
Our initial sample included 92 adolescents. Of this group,
opposing pattern of HPA function; that is, childhood mal-
12 were excluded because they met criteria for one of the
treatment is associated with heightened reactivity, and high
exclusionary diagnoses, and 9 either did not complete the
depression severity is associated with blunted reactivity.
childhood maltreatment interview, or were missing more
Childhood maltreatment and depression severity are, them-
than one cortisol sample. The resulting sample of 71 did
selves, highly correlated (Harkness and Monroe, 2002).
not differ from excluded participants in terms of age, sex,
Therefore, it is possible that failure to take both of these
ethnicity, or parental occupation status (all ps > .44). Of the
sources of heterogeneity into account in research on the HPA
21 adolescents who were excluded, 52% (n = 11) had been
axis in MDD may at least in part explain inconsistencies in
recruited from community mental health sites, and 48%
findings across studies. To date, however, there have been no
(n = 10) were from advertisements or local high schools
studies examining the relation of individual differences in
( p > .90).
depression severity to cortisol reactivity in adolescents, nor
have there been any studies in either adults or adolescents
examining the interaction of childhood maltreatment and 2.2. Assessment measures
depression severity on HPA axis function.
The goal of the present study was to examine the relation The present study was conducted in compliance with
of childhood maltreatment, depression severity, and their our University’s Health Sciences Research Ethics Board.
Cortisol reactivity to social stress in adolescents: Role of depression severity and child maltreatment
Table 1 Descriptive characteristics of the sample by depression group.
No/mild depression (n = 30)
M SD n %
Sex (female) 20 67
Age 14.73 2.30
Tanner 4.16 .82
Parental occupation 2.77 1.63
Age at first onset
Depression history
(recurrent)
Treatment (yes)
Comorbidity (yes)
Written informed consent was obtained from all adoles-
cents, and from a parent or guardian for adolescents under
18. Interviews and questionnaires were administered by
one of three graduate students in clinical psychology.
Adolescents participated in two sessions separated by
one week to reduce participant burden. Both sessions were
conducted at the same time of day. Session 1 included the
diagnostic interview and questionnaires. Session 2 included
the TSST and childhood maltreatment interview. The child-
hood maltreatment interview was conducted a full 2 h
following the TSST.
2.2.1. Diagnostic measure
All adolescents received the full child and adolescent
version of the Schedule of Affective Disorders and Schizo-
phrenia (K-SADS) (Kaufman et al., 2000) to evaluate the
presence of current and/or past DSM-IV Axis I diagnoses.
Interviews were conducted by senior graduate students in
clinical psychology who were trained to gold-standard
reliability status (Grove et al., 1981). This interview began
with the collection of basic demographic information,
including parental occupation. Occupations were subse-
quently rated by two independent judges on a 1- to 7-
point scale according to the Hollingshead Index of Social
Position (Hollingshead, 1975). Discrepancies between
raters were resolved by consensus, and the consensus
rating was used in analyses. Parent report of adolescent
symptoms was not gathered based on evidence document-
ing that parent reports of adolescent internalizing disor-
ders produce a high level of false negatives (Klein et al.,
2005).
Thirty-eight (54%) adolescents met criteria for a
current depressive disorder (n = 24 major depressive
disorder, n = 10 depressive disorder not otherwise speci-
fied, n = 4 dysthymia). Of these, 20 (53%) adolescents met
criteria for at least one comorbid Axis I disorder (see
Table 1).
2.2.2. Depression symptoms
Participants completed the BDI-II (Beck et al., 1996), a 21-
item questionnaire assessing the severity of depressive symp-
toms over the past two weeks. We compared three depres-
sion severity groups based on BDI-II scores: minimal
depression (BDI-II 0—13; n = 30), mild/moderate depression
(BDI-II 14—25; n = 16), and moderate/severe depression
175
Moderate depression (n = 16) Severe depression (n = 25)
M SD n % M SD n %
9 56 19 76
15.69 1.58 16.00 2.00
4.23 .56 4.50 .57
4.69 1.82 3.76 1.69
12.64 3.61 13.68 1.93
4 25 7 28
8 50 8 32
6 38 13 52
(BDI-II 26—49; n = 25). The group labels are consistent with
those employed by Beck et al. (1996).3
2.2.3. Purbertal status
Pubertal status was assessed using the Tanner stages of sexual
maturation (Tanner, 1962). Self-report was based on a choice
between five illustrations of breast (females) and pubic hair
(males and females) development, as validated against phy-
sician assessment for this age group (Taylor et al., 2001).
Scores could range from 1 to 5, with higher scores represent-
ing later stages. It should be noted that none of the adoles-
cent girls were pregnant at the time of the study.
2.2.4. Childhood maltreatment
Participants were interviewed using the Childhood Experi-
ence of Care and Abuse contextual semi-structured interview
and rating system (CECA) (Bifulco et al., 1994). The CECA
interviews were conducted 2 h following the TSST, thus
allowing for full recovery of cortisol function before querying
about this emotionally charged information. The following
scales were assessed: (1) antipathy — hostility or coldness
toward the child (e.g., harsh criticism or name-calling); (2)
indifference — neglect of the child’s physical and/or emo-
tional needs (e.g., not providing adequate food or clothing;
not comforting the child when upset); (3) physical abuse —
violence toward the child (e.g., punching, hitting with an
object, or threatening with a knife); and (4) sexual abuse —
non-consensual sexual contact by any perpetrator (e.g.,
fondling, oral sex, and/or penetration). All interviews were
audiotaped.
The above scales were rated for severity (1-marked, 2-
moderate, 3-some, 4-little/none) by raters who were una-
ware of the participants’ depression status. Ratings were
anchored to the CECA manual, which includes hundreds of
examples and rating rules. CECA interviewers and raters
received extensive training and ongoing supervision in the
3
Beck et al. (1996) suggest the following cutoffs: 0—13: minimal
depression; 14—19: mild depression; 20—28: moderate depression;
and 29—63: severe depression. We did not use the exact same
groupings as Beck et al. because we did not have sufficient sample
size to divide our sample into four groupings. Therefore, we chose
three groupings that cut the moderate group in half, putting the
lower half with the mild/moderate group and the upper half with the
moderate/severe group.
176
Bedford College procedures for rating the CECA by the first
author. Inter-rater reliabilities ranged from k = .86—1.0
(Harkness et al., 2006). The severity ratings of maltreatment
using the CECA were very negatively skewed (i.e., the major-
ity of participants had ratings of 4-little/none). Therefore,
we created a composite variable — ‘‘childhood maltreat-
ment,’’ representing the presence versus absence of at least
some (level ‘3’) antipathy and/or indifference and/or phy-
sical abuse and/or sexual abuse.
2.3. Cortisol collection and Trier Social Stress
Task
Participants were asked to refrain from eating and drinking
for 1 h prior to their arrival at the lab. All saliva was
collected in previously labeled 5 ml polypropylene vials
(PGC Scientifics Corporation, MD) by passive drool (Shirt-
cliff et al., 2001) between the hours of 3 pm and 5 pm
because this is a period of low cortisol relative to the
morning (Groschl et al., 2003). All participants were medi-
cally healthy and had not experienced an acute injury in the
preceding 24 h.4
We followed precisely the procedure outlined by Krisch-
baum, Pirke, and Hellhammer (1993) in defining the time
points for salivary sample collection during the TSST. The
TSST protocol began with a 10-min rest period to allow the
participant to adjust to the research setting. Participants
then provided a baseline saliva sample (sample a). They were
then led into the experimental room and introduced to two
research assistants who the participants were told were
members of a selection committee from a human resources
department. Participants were told that they were to pre-
pare a 5-min speech that would serve as a ‘‘job interview,’’
and that they would give their speech to the selection
committee. They were told that their speech would be
videotaped. This point in time served as the onset of the
stressor. Participants were then led back into the experi-
mental room and given 10 min to prepare, after which we
collected a second sample (sample b). Participants were then
led back into the experimental room where they delivered
their speech and performed an arithmetic test, which con-
sisted of serially subtracting by 13, starting from 1022, as
quickly as possible without making any mistakes. This phase
of the TSST took approximately 20 min, after which we
collected a third sample (sample c). Therefore, sample c
was collected 30 min after the onset of the stressor, the time
point of peak cortisol reactivity. A fourth sample was col-
lected 1 h later (sample d), and a fifth and final sample was
collected another hour later (sample e). Following the TSST,
participants were fully debriefed regarding the purpose of
the TSST and the nature of the deception.
4
As would be expected, 25/41 (61%) of the depressed adolescents
in our sample were experiencing sleep disturbance (i.e., insomnia or
hypersomnia). These adolescents were not differentially distributed
across the mild/moderate (56%) and moderate/severe (64%) de-
pressed groups, x2(1) = .25, p = .62, or across those with (61%) versus
without (61%) childhood maltreatment, x2(1) < .001, p = .99. There-
fore, sleep disturbance cannot better account for our pattern of
results.
K.L. Harkness et al.
2.3.1. Hormone determinations
All saliva samples were immediately placed in a freezer for
short-term storage and eventually transported, on ice, to a
secure frozen storage ( 20 8C). Resulting supernatant were
assayed for cortisol using an enzyme-linked immunoassay
designed specifically for saliva (1-0102/1-0112; Salimetrics
LLC, State College, PA). All samples from one individual were
placed on the same plate so that inter-assay variability did
not contribute to quantification error. Each sample was
quantified in duplicate at 25 ml, and triplicate high and
low controls (4-COO1) were distributed across each plate
to track precision and accuracy. Samples that had a coeffi-
cient of variation 15% were repeated on another plate (Berg
and Wynne-Edwards, 2001). Repetition was used to reject
one of the original duplicates but was not used in analyses.
Salivary cortisol measured by this method is highly correlated
with serum cortisol (r = .96; Salimetrics validation). In all, 24
assay runs, in four batches, were conducted. The high con-
trol, measured at 1.1 mg/dL, had an intra-assay coefficient of
variation of 4.2% and an inter-assay coefficient of variation of
5.8%. The low control, measured at .1 mg/dL had an intra-
assay coefficient of variation of 6.9% and an inter-assay
coefficient of variation of 11.1%.
2.3.2. Cortisol parameters for analyses
We used sample a (‘‘rest phase’’) as participants’ baseline
against which to gauge differences across depression severity
and history of childhood maltreatment in: (a) raw baseline
cortisol concentration in mg/dL (M = .12, SD = .08,
range = .04—.49, Skew = 2.20), (b) ‘cortisol reactivity,’
defined as peak cortisol concentration (sample c) minus
baseline (sample a) divided by baseline (M = .27, SD = .71,
range = .68—3.44, Skew = 2.14), and (c) area under the
curve (AUC) relative to self (M = .07, SD = 8.65,
range = 24.25—25.63, Skew = .26). AUC is a measure of
the total cortisol output over the TSST calculated as the
sum of the area of the four trapezoids bounded by the
baseline value and framed by the cortisol concentration in
the five saliva samples from the session. Square-root trans-
formations were applied to baseline cortisol and cortisol
reactivity to normalize these variables (transformed skew
values < 1.5). The untransformed estimated marginal means
are presented in the figures for ease of interpretability.
3. Results
3.1. Participant characteristics
Demographic and clinical characteristics of the three depres-
sion severity groups are presented in Table 1. There were no
significant relations of depression severity group to sex
( p = .42) or Tanner score ( p = .19). However, groups differed
in age, at a trend, F(2, 68) = 2.80, p = .07, h2 = .08, and
parental Hollingshead index, F(2, 68) = 7.00, p < .005,
h2 = .17. Among those with a diagnosis of a depressive dis-
order, there were no significant differences between those of
moderate and severe BDI-II scores in percentage of those on a
recurrence, age at first onset, percentage of those with a
comorbid disorder, or percentage of those receiving treat-
ment (all ps > .15). Of those receiving treatment, 8 were on
an antidepressant medication. These 8 adolescents were not
Cortisol reactivity to social stress in adolescents: Role of depression severity and child maltreatment 177

differentially distributed between the mild/moderate
and moderate/severe depression groups (25% vs. 12%;
x2[1] = 1.16, p = .28) or between the groups with versus
without a history of maltreatment (12% vs. 11%; x2[1] =
.003, p = .96).5
Twenty-six adolescents reported a history of childhood
maltreatment (37%). Therefore, the present sample con-
sisted of the following groups: (a) no/minimal depression
with no maltreatment (n = 27); (b) no/minimal depression
with a history of maltreatment (n = 3); (c) mild/moderate
depression with no maltreatment (n = 8); (d) mild/moderate
depression with a history of maltreatment (n = 8); (e) mod-
erate/severe depression with no maltreatment (n = 10); and
(f) moderate/severe depression with a history of maltreat-
ment (n = 15).
Participants with versus without a history of childhood
maltreatment did not differ significantly in terms of sex, age,
Tanner score, or parental Hollingshead index (all ps > .15).
Among the clinically depressed adolescents, there were no
significant relations of childhood maltreatment to depression
history status, age at first onset, or treatment status (all
ps > .35). However, those with a history of maltreatment
were significantly more likely to have a comorbid diagnosis
(61% vs. 28%; x2[1] = 4.45, p < .05).6
The specific types of maltreatment assessed in our sample
included physical abuse (n = 7), sexual abuse (n = 6), and
emotional maltreatment (n = 24). As is typical in maltreated
samples, most adolescents reported more than one form of
maltreatment. We found no evidence for a differential dis-
tribution across mild/moderate and moderate/severe
depression of physical abuse (19% vs. 8%, x2[1] = 1.05,
p = .30), sexual abuse (19% vs. 12%, x2[1] = .36, p = .55), or
emotional abuse (50% vs. 56%, x2[1] = .14, p = .71). The
average age at start of maltreatment was 7.96 (SD = 4.89)
and at end was 14.81 (SD = 2.40), with an average duration of
11.58 years (SD = 5.46). The presence versus absence of
current maltreatment was not differentially distributed
across the mild/moderate and moderate/severe depression
groups (x2[1] = .17, p = .68).
(depression group: minimal depression, mild/moderate
depression, moderate/severe depression)  2 (childhood mal-
treatment: presence versus absence) repeated-measures Ana-
lysis of Covariance (RMANCOVA) controlling for age and
parental occupation status. There was a 2-way interaction
of time by depressed group, Wilks’ l = .72, F(8, 120) = 2.66,
p < .01, h2 = .16. This was qualified by a significant 3-way
interaction of time by depressed group by childhood maltreat-
ment in the quadratic trend, F(2, 63) = 3.39, p < .05, h2 = .10.
Among those with no history of maltreatment, the three
depression groups did not differ in terms of the shape of their
cortisol curves over time, F(2, 40) = 1.18, p = .32, h2 = .06 (see
Fig. 1a). In contrast, among those with a history of maltreat-
ment, there remained a significant time by depression group
interaction, F(2, 21) = 3.72, p < .05, h2 = .26 (see Fig. 1b). As
hypothesized, the mild/moderate depressed group with mal-
treatment showed the strongest reactivity to the stressor,
demonstrating a significant quadratic trend in the curve over
time, F(1, 7) = 8.50, p < .05, h2 = .57. In contrast, and also as
hypothesized, the moderate/severe depressed group with
maltreatment showed no reactivity to the stressor, demon-
strating a significant downward linear trend in the curve over
time, F(1, 14) = 24.67, p < .001, h2 = .64.
3.3. Between-group cortisol parameters
As displayed in Table 2, baseline cortisol concentration was
significantly positively correlated with age, and cortisol
reactivity was significantly positively correlated with paren-
tal Hollingshead scores (i.e., related to lower parental occu-

3.2. Within-group cortisol reactivity over time

Square-root transformed group cortisol concentrations (in mg/

dL) at each time point were analyzed using a 5 (time)  3

5
None of the adolescents in our sample were on an anxiolytic
medication. In analyses including only the depressed adolescents,
results we report below were all robust when we controlled for
antidepressant medication usage. Results are available from the
authors.
6
In analyses including only the depressed adolescents, all effects
were robust when controlling for the presence versus absence of a
comorbid diagnosis. In terms of the specific comorbidities, all effects
were robust when controlling for the presence versus absence of
social phobia and alcohol/substance abuse, in separate models. The
frequencies of the remaining comorbidities were too small to merit
including these comorbidities as covariates. However, all results
were robust in separate models in which the individuals with each
of these comorbidities were excluded. Therefore, there is no evi-
dence that our pattern of results is driven by the individuals with any Figure 1 Cortisol concentrations (in mg/dL) across time strati-
of these comorbidities. Results are available from the authors. fied by childhood maltreatment history and depression severity.
178 K.L. Harkness et al.

Table 2 Relation of cortisol parameters to demographic and

clinical characteristics of the sample.

Cortisol Cortisol AUC

baseline reactivity
Sex .02 .11 .15
Age .38 ** .08 .05
Tanner stages .06 .03 .12
Hollingshead index .02 .32 ** .16
Age at first onset .001 .10 .01
Depression history .08 .21 .11
Treatment status .10 .22 .002
Comorbidity .04 .38 * .16
*
p < .05.
**
p < .01.
Figure 3 Mean cortisol reactivity (in mg/dL) by childhood
pation status). In addition, among the clinically depressed maltreatment history and depression severity.
adolescents, those with a comorbid diagnosis had higher
cortisol reactivity than those without. Below we report on depressed adolescents with maltreatment had a significantly
a series of ANCOVA models examining the effects of depres- lower level of reactivity than those with no/minimal depres-
sion severity group, childhood maltreatment, and their inter- sion, F(1, 63) = 5.57, p < .05, h2 = .08, and than those with
action to cortisol baseline, reactivity, and AUC, controlling mild/moderate depression, F(1, 63) = 13.06, p < .005,
for age and parental Hollingshead index. h2 = .17 (see Fig. 3). Among those with no childhood mal-
treatment, however, the three depression groups did not
3.3.1. Cortisol baseline differ significantly (all ps > .07).
The interaction of depression severity group and childhood It is important to note that the moderate/severe and
maltreatment was significant, F(2, 63) = 3.82, p < .05, mild/moderate groups with maltreatment did not differ in
h2 = .11. Adolescents with a history of childhood maltreatment terms of their cortisol baseline, t(21) = .66, p = .52.
had a significantly higher cortisol baseline than those without, Further, in the moderate/severe depressed and maltreated
but only among those with moderate/severe depression, F(1, group, baseline cortisol did not correlate significantly with
63) = 5.18, p < .05, h2 = .08 (see Fig. 2). Indeed, among those cortisol reactivity, r = .02, p = .94. Therefore, the blunted
with a history of maltreatment, the moderately/severely reactivity seen in the moderate/severe depressed group with
depressed, F(1, 63) = 7.84, p < .01, h2 = .11, and the maltreatment likely does not represent a ceiling effect.
mildly/moderately depressed, F(1, 63) = 5.76, p < .05,
h2 = .08, adolescents had a significantly higher baseline than 3.3.3. Area under the curve
those with no/minimal depression. Among those with no child- The interaction of depression severity group and childhood
hood maltreatment, however, the three depression groups did maltreatment was significant, F(2, 63) = 3.41, p < .05,
not differ significantly (all ps > .25). h2 = .10. Adolescents with a history of childhood maltreat-
ment had a significantly greater AUC than those without, but
3.3.2. Cortisol reactivity only among those with mild/moderate depression, F(1,
The interaction of depression severity group and childhood 63) = 10.42, p < .005, h2 = .14 (see Fig. 4). Among those with
maltreatment approached significance, F(2, 63) = 2.59, no childhood maltreatment, however, the three depression
p = .08, h2 = .08. As hypothesized, the moderate/severe groups did not differ significantly (all ps > .08).

Figure 2 Mean cortisol baseline (in mg/dL) by childhood mal- Figure 4 Total cortisol exposure (in mg/dL) by childhood mal-
treatment history and depression severity. treatment history and depression severity.
Cortisol reactivity to social stress in adolescents: Role of depression severity and child maltreatment
4. Discussion
The present study is the first to our knowledge to examine the
effect of a history of childhood maltreatment on HPA axis
reactivity as moderated by depression severity. Consistent
with hypotheses, a history of childhood maltreatment was
associated with significantly higher cortisol reactivity and
total cortisol exposure (i.e., AUC) to a psychosocial stress
challenge. However, an intriguing differential association
between childhood maltreatment and HPA axis function
emerged when the sample was stratified by depression sever-
ity. Only those with a mild/moderate level of depression
severity (i.e., BDI-II between 14 and 25) showed increased
cortisol secretion in response to the TSST. In marked contrast,
adolescents with a moderate/severe level of depression
(i.e., BDI-II > 25) showed significant blunting of the cortisol
response, particularly among those with a history of mal-
treatment. Indeed, cortisol levels decreased in a linear
fashion across the sampling period in moderate/severe
depression with a history of maltreatment. This pattern of
results cannot be better accounted for by socio-economic
status, age, pubertal status, or gender. Further, the mild/
moderate and moderate/severe depression groups did not
differ in baseline cortisol concentration, anti-depressant
medication status, or the presence of a comorbid disorder,
including post-traumatic stress disorder.
Our results are consistent with, and expand upon, the
literature in adult MDD. Specifically, Burke et al. (2005) also
reported blunted stress reactivity and impaired stress recov-
ery in response to psychological challenge in MDD compared to
controls, and described a similarly flat and unresponsive pat-
tern of cortisol secretion, particularly in severe depression.
Our results suggest that this pattern generalizes to an adoles-
cent sample, and may be specific to severe depression. Future
research is now needed to determine whether a blunted HPA
axis is specific to particular symptoms, or qualitatively distinct
subtypes (e.g., melancholic depression) of MDD.
In addition, our results are consistent with previous lit-
erature examining the effect of childhood maltreatment on
HPA axis function. On the one hand, greater cortisol reactiv-
ity in challenge studies has been observed in adult samples
with a history of abuse (Heim et al., 2000, 2001, 2002), as
well as in adolescent samples with a history of adversity more
generally (e.g., early parental loss, life-threatening illness,
and abuse) (Rao et al., 2008). On the other hand, however,
studies in adolescents employing the TSST, and examining
maltreatment experiences similar to those assessed in the
present study, have reported blunted cortisol reactivity
(Carpenter et al., 2007; De Bellis et al., 1994). We suggest
based on our results that differences in the severity of the
depressed groups examined in these previous studies may
help to reconcile these conflicting findings.
The complete dissociation of cortisol response between
the mild/moderate and moderate/severe depressed adoles-
cents with maltreatment suggests that the neurobiological
stress response systems may be functioning very differently
in these two groups. Previous research has suggested that
individuals at genetic risk for depression, who would be prone
to developing a more severe manifestation of the syndrome,
express resistance (i.e., desensitization) of glucocorticoid
receptors as a trait (Holsboer et al., 1995; Modell et al.,
1998). In these individuals, high levels of CRH release as a
179
result of the stress of childhood maltreatment, in the context
of this reduced glucocorticoid negative feedback, would
ultimately lead to a downregulation in CRH receptors on
ACTH-producing cells in the anterior pituitary. As such, this
may suggest a mechanism to explain why individuals with a
severe, potentially genetically-mediated depression, and a
history of chronic adversity would be more likely to show a
blunted cortisol response to acute stress.
In contrast, maltreated individuals with mild/moderate
depressive symptoms may be experiencing the typical HPA
abnormalities that have been well documented in the con-
text of stress. That is, these individuals are able to maintain
an increased secretion of glucocorticoids to an acute stressor
despite negative feedback from glucocorticoid and mireral-
corticoid receptors. This is achieved by increased expression
of CRH and vasopressin in the paraventricular nucleus,
decrease in glucocorticoid and mineralcorticoid receptors
and hypertrophy of the adrenal glands (see Checkley, 1996
for a review). The fact that some studies have found higher
levels of CRH in the cerebrospinal fluid of depressed indivi-
duals versus healthy controls (e.g., Catalán et al., 1998),
whereas others have found lower CSF CRH levels in depres-
sion (e.g., Geracioti et al., 1997), also supports the proposal
that there may be two different groups of depressed indivi-
duals who may prone, respectively, to a normal versus a
blunted cortisol response to stress. Future studies examining
a broad array of neurohormonal indicators of HPA axis func-
tion are now required to more firmly determine the exact
mechanism that accounts for the dissociation in cortisol
response to stress in mild/moderate versus moderate/severe
depressed individuals with a history of maltreatment.
The present results should be interpreted in light of the
following limitations. First, the results should be replicated
with a larger sample. Nevertheless, it is important to note
that we obtained medium to large effect sizes for all com-
parisons, indicating the robustness of our effects. In parti-
cular, we had a small number of adolescents in the no/
minimal depression group who had a history of childhood
maltreatment. These adolescents had the lowest mean base-
line cortisol level. However, this result is difficult to interpret
given the size of this group. While this group did not repre-
sent the crucial comparison in the present study, future
studies confirming and expanding upon the results presented
here should nevertheless strive to include a higher number of
healthy adolescents with maltreatment histories.
Future research is also needed to stratify the sample by
socio-economic status, age, and gender. Lower socio-eco-
nomic status significantly predicted higher cortisol reactivity
in the present sample, and significant sex and age differences
in cortisol reactivity have been reported in previous samples
(Chopra et al., 2009; Gunnar et al., 2009; Kelly et al., 2008).
While our results were robust when controlling for these
demographic characteristics, this does not preclude the
possibility that the pattern reported here may be further
moderated by age, sex, or socio-economic status.
Our small sample also precluded examination of our
effects separately for different types of abuse (physical
versus sexual versus emotional). As noted earlier, we found
no evidence for a differential distribution across mild/mod-
erate and moderate/severe depression of physical abuse,
sexual abuse, or emotional abuse. Therefore, the dissocia-
tion of HPA axis function found between these two groups
180
cannot be better accounted for by a differential distribution
of maltreatment type across mild/moderate and moderate/
severe depression. Nevertheless, this is an important area for
future research as different types of abuse may have differ-
ent neurobiological consequences.
Second, our definition of the depression groups was based
on self-report. Nevertheless, all adolescents in our mild/
moderate and moderate/severe depression groups met DSM-
IV criteria for a unipolar mood disorder based on the K-SADS
interview. Further, the BDI is a highly reliable measure
(a = .95 in the present sample) that correlates significantly
with clinician-rated severity scales (Steer et al., 1987).
Third, our study relied on retrospective self-reports of child-
hood maltreatment, which may be subject to bias. However,
it is unclear how differential biases to recall childhood mal-
treatment would lead to the specific pattern of results seen
here given that the mild/moderate and moderate/severe
depressed groups did not differ significantly in the proportion
who reported this history (50% vs. 60%; x2[1] = .40, p = .53).
Further, recent reports have confirmed the validity of retro-
spective reports of maltreatment using the CECA in the study
of depression (Brown et al., 2007). Fourth, in the present
study we did not collect data regarding non-psychotropic
medications that may have affected cortisol function, nor did
we provide urine screenings for substance use. However, we
do note that our results were robust when controlling for self-
reported alcohol/substance abuse. Finally, future research
should employ more frequent cortisol assessment points to
provide a more fine-grained analysis of cortisol recovery.
The present study also had a number of strengths, includ-
ing the use of an ecologically valid challenge paradigm and
the use of rigorous, structured interviews to determine
clinical diagnoses and childhood maltreatment history. In
summary, the present findings suggest that HPA axis dysre-
gulation in adolescent MDD is specific to those with a history
of maltreatment, and shows an opposing pattern of response
based on the severity of depression. These results have
significant implications for our understanding of the neu-
roendocrinological pathology of MDD, and suggest that indi-
vidual differences in both depression severity and
maltreatment history should be considered in all future
studies examining HPA axis reactivity in MDD.
Role of funding sources
Funding for this study was provided by a grant from the Ontario
Mental Health Foundation (Harkness, Wynne-Edwards). The
OMHF had no further role in study design; in the collection,
analysis and interpretation of data; in the writing of the
report; and in the decision to submit the paper for publication.
Conflicts of interest
Dr. Harkness, Mr. Stewart, and Dr. Wynne-Edwards report no
biomedical financial interests or potential conflicts of interest.
Contributions
Dr. Harkness designed the study, supervised data collection,
and wrote the manuscript. Mr. Stewart aided in conducting
K.L. Harkness et al.
the diagnostic and childhood maltreatment interviews and
undertook the statistical analyses. Dr. Wynne-Edwards wrote
the study protocol and supervised the cortisol assays.
Acknowledgements
We gratefully acknowledge the technical of Lea Bond in
performing the cortisol assays. We would also like to
acknowledge Eric Bulmash, Alexandra Sutherland, and Naza-
nin Alavi who performed the diagnostic and maltreatment
interviews, and Shannon Coyle, Lindsey Lytle, Emma Dargie,
and Lindsay Delima for their help with data coding and data
management.
References
Akil, H., Haskett, R.F., Young, E.A., Grunhaus, L., Kotun, J., Wein-
berg, V., et al., 1993. Multiple HPA profiles in endogenous
depression: effect of age and sex on cortisol and beta-endorphin.
Biol. Psychiatry 33, 73—85.
American Psychiatric Association, 1994. Diagnostic and Statistical
Manual of Mental Disorders, 4th ed. Author, Washington, DC.
Beck, A.T., Brown, G., Steer, R.A., 1996. Beck Depression Inventory II
Manual. The Psychological Corporation, San Antonio, TX.
Berg, S.J., Wynne-Edwards, K.E., 2001. Changes in testosterone,
cortisol and estradiol levels in men becoming fathers. Mayo Clin.
Proc. 76, 582—592.
Bifulco, A., Brown, G., Harris, T., 1994. Childhood Experience of Care
and Abuse (CECA): a retrospective interview measure. J. Child
Psychol. Psychiatry 31, 1419—1435.
Brown, G.W., Craig, T.K.J., Harris, T.O., Handley, R.V., Harvey, A.L.,
2007. Validity of retrospective measures of maltreatment and
depressive episodes using the Childhood Experience of Care and
Abuse (CECA) instrument — a life course study of adult chronic
depression — 2. J. Affect. Dis. 103, 217—224.
Burke, H.M., Davis, S.C., Otte, C., Mohr, D.C., 2005. Depression and
psychological response to stress: a meta-analysis. Psychoneur-
oendocrinology 30, 846—856.
Carpenter, L.L., Carvalho, J.P., Tyrka, A.R., Wier, L.M., Mello, A.F.,
Mello, M., et al., 2007. Decreased adrenocorticotropic hormone
and cortisol responses to stress in healthy adults reporting signif-
icant childhood maltreatment. Biol. Psychiatry 62, 1080—1087.
Carroll, B.J., Feinberg, M., Greden, J.F., Haskett, R.F., Steiner, J.M.,
Tarika, J., 1980. Diagnosis of endogenous depression. J. Affect.
Dis. 2, 177—194.
Carroll, B.J., Feinberg, M., Greden, J.F., Tarika, J., Albala, A.A.,
Haskett, R.F., et al., 1981. A specific laboratory test for diagnosis
of melancholia — standardization, validation, and clinical utility.
Arch. Gen. Psychiatry 38, 15—22.
Catalán, R., Gallart, J.M., Castellanos, J.M., Galard, R., 1998.
Plasma corticotropin-releasing factor in depressive disorders.
Biol. Psychiatry 44, 15—20.
Checkley, S., 1996. The neuroendocrinology of depression and chron-
ic stress. Br. Med. Bull. 52, 597—617.
Chopra, K.K., Ravindran, A., Kennedy, S.H., Mackenzie, B., Mat-
thews, S., Anisman, H., et al., 2009. Sex differences in hormonal
responses to a social stressor in chronic major depression. Psy-
choneuroendocrinology 34, 1235—1241.
De Bellis, M.D., Chrousos, G.P., Dorn, L.D., Burke, L., Helmers, K.,
Kling, M.A., et al., 1994. Hypothalamic-pituitary-adrenal axis
dysregulation in sexually abused girls. J. Clin. Endocr. Metab. 78,
249—255.
Dickerson, S.S., Kemeny, M.E., 2004. Acute stressors and cortisol
responses: a theoretical integration and synthesis of laboratory
research. Psychol. Bull. 130, 355—391.
Cortisol reactivity to social stress in adolescents: Role of depression severity and child maltreatment
Dinan, T.G., Scott, L.V., 2005. Anatomy of melancholia: focus on
hypothalamic-pituitary-adrenal axis overactivity and the role of
vasopressin. J. Anat. 207, 259—264.
Geracioti, T.D., Loosen, P.T., Orth, D.N., 1997. Low cerebrospinal
fluid corticotropin-releasing hormone concentrations in eucorti-
solemic depression. Biol. Psychiatry 42, 165—174.
Groschl, M., Rauh, M., Dorr, H.G., 2003. Circadian rhythm of salivary
cortisol, 17 alpha-hydroxyprogesterone and progesterone in
healthy children. Clin. Chem. 49, 1688—1691.
Grove, W., Andreasen, N., McDonald-Scott, P., Keller, M., Shapiro, R.,
1981. Reliability studies of psychiatric diagnosis. Arch. Gen.
Psychiatry 38, 408—413.
Gunnar, M.R., Wewerka, S., Frenn, K., Long, J.D., Griggs, C., 2009.
Developmental changes in hypothalamus-pituitary-adrenal activ-
ity over the transition to adolescence: normative changes and
associations with puberty. Dev. Psychopathol. 21, 69—85.
Harkness, K.L., Bruce, A.E., Lumley, M.N., 2006. Childhood adversity
and the sensitization to stressful life events in adolescent de-
pression. J. Abnorm. Psychol. 115, 730—741.
Harkness, K.L., Monroe, S.M., 2002. Childhood adversity and the
endogenous versus nonendogenous distinction in women with
major depression. Am. J. Psychiatry 159, 387—393.
Heim, C., Newport, D.J., Heit, S., Graham, Y.P., Wilcox, M., Bonsall,
R., et al., 2000. Pituitary-adrenal and autonomic responses to
stress in women after sexual and physical abuse in childhood.
JAMA 284, 592—597.
Heim, C., Newport, D.J., Bonsall, R., Miller, A.H., Nemeroff, C.B.,
2001. Altered pituitary-adrenal axis responses to provocative
challenge tests in adult survivors of childhood abuse. Am. J.
Psychiatry 158, 575—581.
Heim, C., Newport, D.J., Wagner, D., Wilcox, M.M., Miller, A.H.,
Nemeroff, C.B., 2002. The role of early adverse experience and
adulthood stress in the prediction of neuroendocrine stress reac-
tivity in women: a multiple regression analysis. Depress. Anxiety
15, 117—125.
Hollingshead, A.B., 1975. Four factor index of social position. Un-
published manuscript, Yale University, New Haven, CT
Holsboer, F., 2000. The corticosteroid receptor hypothesis of depres-
sion. Neuropsychopharmacology 23, 477—501.
Holsboer, F., Lauer, C.J., Schreiber, W., Krieg, J.C., 1995. Altered
hypothalamic-pituitary-adrenocortical regulation in healthy sub-
jects at high familial risk for affective disorders. Neuroendocri-
nology 62, 340—347.
Kaufman, J., Birmaher, B., Brent, D.A., Ryan, N.D., Rao, U., 2000. K-
SADS-PL. J. Am. Acad. Child Psychiatry 39, 1208—11208.
Kaufman, J., Birmaher, B., Perel, J., Dahl, R.E., Moreci, P., Nelson,
B., et al., 1997. The corticotropin-releasing hormone challenge
in depressed abused, depressed nonabused, and normal control
children. Biol. Psychiatry 42, 669—679.
Kaufman, J., Charney, D., 2003. The neurobiology of child and
adolescent depression: current knowledge and future directions.
In: Cicchetti, D., Walker, E.F. (Eds.), Neurodevelopmental
Mechanisms in Psychopathology. Cambridge University Press,
Cambridge, UK, pp. 461—490.
181
Kaufman, J., Martin, A., King, R.A., Charney, D., 2001. Are child-,
adolescent-, and adult-onset depression one and the same disor-
der. Biol. Psychiatry 49, 980—1001.
Kelly, M.M., Tyrka, A.R., Anderson, G.M., Price, L.H., Carpenter, L.L.,
2008. Sex differences in emotional and physiological responses to
the Trier Social Stress Test. J. Behav. Ther. Exp. Psychiatry 39,
87—98.
Kirschbaum, C., Pirke, K.M., Hellhammer, D.H., 1993. The Trier
Social Stress Test — a tool for investigating psychobiological stress
responses in a laboratory setting. Neuropsychobiology 28, 76—81.
Klein, D.N., Dougherty, L.R., Olino, T.M., 2005. Toward guidelines for
evidence-based assessment of depression in children and adoles-
cents. JCCAP 34, 412—432.
Lopez-Duran, N.L., Kovacs, M., George, C.J., 2009. Hypothalamic-
pituitary-adrenal axis dysregulation in depressed children and
adolescents: a meta-analysis. Psychoneuroendocrinology 34,
1272—1283.
Luby, J.L., Heffelfinger, A., Mrakotsky, C., Brown, K., Hessler, M.,
Spitznagel, E., 2003. Alterations in stress cortisol reactivity in
depressed preschoolers relative to psychiatric no-disorder com-
parison groups. Arch. Gen. Psychiatry 60, 1248—1255.
Luby, J.L., Mrakotsky, C., Heffelfinger, A., Brown, K., Spitznagel, E.,
2004. Characteristics of depressed preschoolers with and without
anhedonia: evidence for a melancholic depressive subtype in
young children. Am. J. Psychiatry 168, 1998—2004.
Modell, S., Lauer, C.J., Schreiber, W., Huber, J., Krieg, J.C., Hols-
boer, F., 1998. Hormonal response pattern in the combined DEX-
CRH test is stable over time in subjects at high familial risk for
affective disorders. Neuropsychopharmacology 18, 253—262.
Rao, U., Hammen, C., Ortiz, L.R., Chen, L.A., Poland, R.E., 2008.
Effects of early and recent adverse experiences on adrenal
response to psychosocial stress in depressed adolescents. Biol.
Psychiatry 64, 521—526.
Shirtcliff, E.A., Granger, D.A., Schwartz, E.B., Curran, M.J., 2001.
Use of salivary biomarkers in biobehavioral research: cotton-
based sample collection methods can interfere with salivary
immunoassay results. Psychoneuroendocrinology 26, 165—173.
Steer, R.A., Beck, A.T., Riskind, J.H., Brown, G., 1987. Relationship
between the Beck Depression Inventory and the Hamilton Psychi-
atric Rating Scale for Depression in depressed outpatients. J.
Psychopathol. Behav. 9, 327—339.
Tanner, J.M., 1962. Growth and Adolescence, 2nd ed. Thomas,
Springfield, IL.
Taylor, S.J., Whincup, P.H., Hindmarsh, P.C., Lampe, F., Odoki, K.,
Cook, D.G., 2001. Performance of a new pubertal self-assessment
questionnaire: a preliminary study. Paediatr. Perinat. Epidemiol.
15, 88—94.
Tsigos, C., Chrousos, G.P., 2002. Hypothalamic-pituitary-adrenal
axis, neuroendocrine factors and stress. J. Psychosom. Res. 53,
865—871.
Zimmerman, M., Coryell, W., Pfohl, B.M., 1985. Importance of
diagnostic thresholds in familial classification: dexamethasone
suppression test and familial subtypes of depression. Arch. Gen.
Psychiatry 42, 300—304.