Psychosocial and Endocrine Features of Chronic
First-Episode Major Depression in 8 –16 Year Olds
Ian M. Goodyer, Rebecca J. Park, and Joe Herbert
Background: Psychoendocrine processes may have a role
in explaining individual differences in the outcome of
major depression in 8 –16-year-old school children.
Methods: Salivary cortisol and dehydroepiandrosterone
(DHEA) levels at 8:00 AM and 8:00 PM, life events, and
comorbidity were assessed at presentation, 36, and 72
weeks in 47 (60%) of 78 clinically referred subjects with a
first episode of major depression. Comparisons were made
between chronic and nonchronic major depression.
Results: Chronic depression was characterized by being
older, cortisol hypersecretion at 8:00 PM at all three
assessments, increasing depression-dependent life events
over the follow-up period, and comorbid obsessive-com-
pulsive disorder (OCD) at presentation and at 36 weeks.
Chronicity may be best predicted by increasing depres-
sion-dependent events over the 72-week period. Such
events are more likely in cases with evening cortisol
hypersecretion at entry and persistent OCD. Variations in
DHEA levels were not associated with chronicity or
increasing life events.
Conclusions: During adolescence, but not childhood, the
persistence of major depression may occur through an
increase of risk for further and particular types of depres-
sion-dependent undesirable life events (personal disap-
pointments and/or dangers to the self), that are more likely
in those subjects with persisting cortisol hypersecretion
and unresolved comorbid OCD. Biol Psychiatry 2001;
50:351–357 © 2001 Society of Biological Psychiatry
Key Words: Depression, cortisol, DHEA, life events
Introduction
T he duration of first-episode major depressive disorder
(MDD) in clinically referred and community ascer-
tained adolescents is highly variable, from a few weeks to
beyond 18 months (Kovacs 1996; Lewinsohn et al 1999).
Persistence of a first episode up to 36 weeks is predicted
From the Developmental Psychiatry Section, Department of Psychiatry and
Department of Anatomy, University of Cambridge, Cambridge, UK.
Address reprint requests to Dr. I.M. Goodyer, University of Cambridge, Douglas
House, Developmental Psychiatry Section, Cambridge CB2 2AH, UK.
Received June 23, 2000; revised November 15, 2000; revised February 9, 2001;
accepted February 14, 2001.
© 2001 Society of Biological Psychiatry
by comorbid obsessive compulsive disorder (OCD),
higher self-report scores of depressive symptoms, and
onset in adolescence rather than childhood (Goodyer et al
1997a). At presentation, high cortisol/dehydroepiandros-
terone (DHEA) ratios in the evening (8:00 PM) and at
midnight (12:00 AM) are also associated with persistent
MDD in the short term (up to 36 weeks) (Goodyer et al
1996; Herbert et al 1996). This endocrine profile, together
with co-morbid OCD at entry and an increase in disap-
pointing life events during follow-up, seems to represent a
psychosocial-endocrine process specifically maintaining
the depressive phenotype up to 36 weeks (Goodyer et al
1997a). Whether this, or another, combination of clinical
predictors and psychoendocrine processes are associated
with longer term maintenance of major depression is not
known.
The objective of this study was to determine whether
chronic depression, defined as the continued presence of
DSM-IV MDD at 72 weeks—the time period by which
around 75% of first episodes are recovered (Kovacs
1996)—is characterized by a defined set of psychoendo-
crine processes and, if so, how these compare to those that
are associated with (DSM-IV) depression that remits
before 72 weeks.
To achieve this aim, chronically depressed subjects at
18 months were compared with those fully recovered.
Nondepressed but still psychiatrically ill subjects were
excluded from the analysis, as their status is ambiguous;
they could represent a recovering or a deteriorating
process.
Methods and Materials
Subjects
Full details of the ascertainment procedure and methods of
sample selection and the characteristics that predicted short-term
recovery at 36 weeks have been given elsewhere (Goodyer et al
1996; Herbert et al 1996). In brief, 360 8 –16-year-olds were
screened with a self-report questionnaire for depressed moods
and feelings (the Mood and Feelings Questionnaire [MFQ],
described below) and those with scores ⬎ 25 (the level at which
there is a high index of suspicion for major depression) were
interviewed over a consecutive period of 21 months. Of these,
0006-3223/01/$20.00
PII S0006-3223(01)01120-9
352 BIOL PSYCHIATRY
2001;50:351–357
120 subjects (28.5%) with MDD were selected for interview; 16
refused to participate, and 104 agreed to enter the study. There
were no significant differences in age, gender, or screen scores
between participants and nonparticipants. All subjects attended
normal school, and over 90% lived with their family of origin
(i.e., one or both biological parents). Seventy-eight subjects
(74%) met DSM III-R diagnostic criteria for a current episode of
major depression at entry into the study (American Psychiatric
Association 1994). Of these, 68 (87%) were reassessed at 36
weeks, and 53 (68%) agreed to reassessment at 72 weeks. The
study received full approval by the local medical ethics commit-
tee. All parents and probands gave written informed consent to
participate in all aspects of the assessments and evaluations.
Assessment
CURRENT MENTAL STATE. Current mental state for all
subjects with suspected MDD was assessed using a standard
interview procedure, the Children’s Schedule for Affective
Disorders, third version (Kiddie-SADS-P; Ambrosini et al 1987).
Mother and child were interviewed simultaneously but sepa-
rately. Immediately following the assessments, interviewers
compared information to determine if there were any marked
discrepancies. Symptoms rated four or more from either inter-
view were considered clinically significant. Agreement on the
presence or absence of all diagnoses for these cases was
satisfactory (␬, range for all diagnoses, 0.7– 0.85). All interview-
ers had a minimum of 5 years experience in child mental health
and completed a SADS training course that included video tapes
and at least three preliminary interviews with an already trained
interviewer. Agreement between trained and trainee interviewers
for presence and absence of diagnosis was satisfactory (␬ ⫽
0.85).
SEVERITY OF CURRENT DEPRESSION. Severity of cur-
rent depression was determined using the total score of the
self-report Mood and Feelings Questionnaire (MFQ-child form;
Costello and Angold 1988; Wood et al 1995).
DEFINITION OF OUTCOME AT 72 WEEKS. Subjects were
reassessed at 72 weeks after presentation using the same proce-
dures as had been used at entry and 36 weeks (Goodyer et al
1996). Recovery from an episode was operationally defined as
having less than two clinically significant symptoms, including
depressed mood or anhedonia, for any diagnosis for at least 8
consecutive weeks. Chronic MDD was defined as a subject
having no evidence of remission over the 72-week period, i.e.,
meeting MDD criteria at all three assessment points. Those who
met DSM-IV criteria for one or more diagnoses at reassessment
were classified as either a depressed or not a depressed case (i.e.,
meeting MDD criteria as well as any other diagnosis). Some
diagnoses (e.g., dysthymia, conduct disorder, and attention def-
icit disorder) require a duration of more than 1 year. In addition,
there is symptom overlap between some disorders (e.g., major
depression and dysthymia). In these circumstances, interviewers
systematically investigated the onset and cessation of symptoms
and their duration, to determine if diagnosis was independent and
still present. Finally, precedence of comorbid nondepressive
I.M. Goodyer et al
disorders was sought from parental interviews but using both sets
of interview information. Information from the K-SADS inter-
views at presentation was also used for duration purposes. All
diagnoses were confirmed by one of the authors (I.G.) from
mental state assessments but blind to interviewer’s diagnoses.
Psychosocial Measures
CURRENT DIFFICULTIES AND HARDSHIPS. Information
was obtained for parents at entry and at both reassessments using
interview and questionnaire procedures to provide information
on current, long-term difficulties (current employment, quality of
housing, current income, number of people living in the house,
number of children, children registered with social services,
current marital status, confiding relationships with partner and/or
another, verbal or physical conflict between partners) and friend-
ship difficulties (Goodyer et al 1997b).
RECENT LIFE EVENTS. Recent life events in the child and
family, and friendship difficulties were assessed by a semi-
structured interview procedure (Goodyer et al 1997b). Only
events that carry moderate to severe negative impact over weeks
were included. Mothers and children independently completed
the interview. This approach assesses 66 recent life events,
classified into one of four domains as follows:
1. Danger to the self: a clear expectation or occurrence of
physical threat to the child, such as illness or accidents,
involvement in a household or community disaster, or
subject of a personal attack;
2. Danger to important others: same events as above but
where the person is exposed is a parent or sibling;
3. Disappointments: the failure of a previously held set of
expectations and/or hopes to the self (includes breakdown
of boyfriend/girlfriend relationship, examination failure)
and to another (includes loss of a job, new financial
difficulties, extra-marital affair);
4. Loss: the permanent state of no further contact with a
valued other, which includes only death and permanent
separations.
Information about recent events was obtained 1) at presentation
and referred retrospectively to the previous 12 months before
presentation; 2) at second reassessment, retrospectively for the
36-week follow-up period; and 3) at third reassessment, retro-
spectively for the 36 –72-week follow-up period. At entry,
mother-child agreement on the presence of events was good
across the age range (␬ ⫽ 0.9). The agreement between respon-
dents on the presence of negative impact on the child (yes/no)
was also high (␬ ⫽ 0 85). Objective panel ratings showed good
agreement between each other for the assignment of negative
impact on the child (␬ ⫽ 0 85). Two-week test–retest reliability
(n ⫽ 20 of the sample) for parent and child identifying recent
events was good (␬ ⫽ 0.8 parents, 0.85 children). Event
occurrence discrepancies between mother and child were few,
and none occurred where events were confined to those judged as
carrying negative impact. In a clinical population it is assumed
that events that are concurrent with disorder are not independent
Chronic Depression In School-Aged Children
of illness-related behavior, except for losses that involve death of
others where the subject was absent.
Salivary Cortisol and DHEA Levels
Saliva assays that were obtained at two time points (8:00 AM and
8:00 PM) over 2 consecutive days, within 2–7 days of interview
at presentation, were used in this analysis. The same procedure
was repeated at 36- and 72-week follow-up periods. Cortisol and
DHEA were assayed separately from each sample, and mean
concentrations of cortisol and DHEA were derived for each time
point. The molar cortisol/DHEA (C/D) ratio was also calculated
for each time point. Cortisol was measured by enzyme-linked
immunosorbent assay and DHEA by a validated radioimmuno-
asssay (Goodyer et al 1996). There is a good correlation between
plasma and salivary concentrations for both steroids (r ⫽ 0.6 for
cortisol, 0.9 for DHEA [Goodyer et al 1996]). Salivary levels
represent about 5% of those in the blood for both hormones:
these are similar to those in the cerebrospinal fluid (Guazzo et al
1996).
Statistical Procedures
Repeated measures analysis of variance (ANOVA) on trans-
formed data were used to explore differences between variables
at baseline and follow-up. Backwards stepwise logistic regres-
sion was used to characterize a best-fit model for predicting
chronic depression. Univariate statistics were also used with
exact test reported.
Results
Characteristics at Baseline and Follow-Up at 72
Weeks
At 72 weeks, 53 (68%; 20 boys, 33 girls; mean age 13.4
[SD ⫽ 2.25] years) of the original sample of 78 were
reassessed for psychiatric disorder. Twenty-seven subjects
(10 boys, 17 girls) had recovered, 6 were nondepressed but
had another psychiatric disorder and were therefore
dropped from further analyses, and 20 (6 boys, 14 girls)
met criteria for chronic disorder (i.e., met DSM-IV criteria
for MDD at all three assessments with no in-between
episode of remission) Telephone or personal contact was
made with 10 of the 15 subjects who had been reassessed
at 36 weeks but declined 72-week follow-up, to obtain
clarification of the reasons for refusal for a second
follow-up. The other families were untraceable. The moth-
ers of all 10 subjects reported that they no longer wished
to participate because their child was recovered. We
ascertained from the general practitioner that none had
been reviewed by any child mental health services, or
received any further medical appointments related to their
original complaints. Thus, the final follow-up group is
likely to contain nearly all cases of chronic first-episode
major depression from the original consecutive cohort of
BIOL PSYCHIATRY 353
2001;50:351–357
78 subjects. This gives an estimated prevalence rate of
25% for chronic major depression from this clinical
population. A comparison between the follow-up sub-
sample (n ⫽ 47) and the total cohort at presentation
(Goodyer et al 1997b) revealed that there were no signif-
icant differences at presentation in any of the following:
severity of depression (mean MFQ scores); mean age at
presentation; duration of depression; proportion of males
and females; and baseline or 36-week hormone concen-
trations. There were no significant differences between the
sample at entry and follow-up on any of the social
variables measured.
Effects of Age On Psychosocial and Endocrine
Variables
The effects of age on the independent variables of interest
were examined before comparisons were made between
the outcome groups (recovered (n ⫽ 27) vs. chronic (n ⫽
20) depression). All subjects were divided into younger
(⬍13 years, 5 months, n ⫽ 22) and older groups (ⱖ13 yrs,
6 months, n ⫽ 25). Older subjects had significantly higher
cortisol levels at 8:00 PM at entry [F(1,44) ⫽ 4.97, p ⬍
.05], reported more negative life events over the 72-week
follow-up period (exact test, p ⬍ .05), and higher mood
and feelings scores at 36 weeks [F(1,45) ⫽ 4.99, p ⬍ .05].
There were no associations between age and any other
hormone measures (cortisol or DHEA at 8:00 AM or 8:00
PM) or self-report at any assessment point (entry, 36, or 72
weeks). Because of the effects of age on the independent
variables of interest, age is used as a covariate where
appropriate in the group comparisons.
Factors and Processes Associated with Chronic
Depression
AGE AND GENDER. A comparison between chronic
depression (cMDD) and recovery (REC) at 72 weeks
showed that cMDD was significantly associated with older
age at entry [F(1,45) ⫽ 18.02, p ⬍ .0001], but not gender
(17 [63%] girls REC group vs. 14 [70%] cMDD group
[␹2(1) ⫽ 0.25, ns]). Among the 20 cases of cMDD, 17
(85%; 12 girls, 5 boys) were older than 13 years, 6 months
at entry (odds ratio ⫽ 2.14, p ⫽ .006, 95% confidence
interval [CI], 1.24 –3.68).
SEVERITY. The mean MFQ (SD) scores at entry, 36,
and 72 weeks were compared for the two outcome groups
using repeated measures ANOVA with age as a covariate.
The findings are shown in Table 1.
The differential progression in MFQ scores between the
two groups shown in Table 1 is confirmed as significant
by the repeated measures ANOVA. Of the 47 subjects, 40
entered the analysis due to missing data for 7 subjects.
354 BIOL PSYCHIATRY I.M. Goodyer et al
2001;50:351–357

Table 1. Mood and Feelings Questionnaire (MFQ) Scores ⫾ Table 2. Mean (SD) Salivary Cortisol Levels (ng/mL) at Each
SD at Entry and During Follow-Up Assessment
Recovered (REC) Depressed (cMDD) Cortisol (8:00 AM) Cortisol (8:00 PM)
Assessment
MFQ scores (n ⫽ 27) (n ⫽ 20)
period REC cMDD REC cMDD
At entry 28.8 ⫾ 16.2 33.3 ⫾ 15.7
Entry 2.31 (1.39) 3.28 (2.50) 0.43 (0.41) 4.0 (2.68)a
36 weeks 18.1 ⫾ 16.1 34.5 ⫾ 15.2a
(n ⫽ 26) (n ⫽ 20) (n ⫽ 26) (n ⫽ 16)
72 weeks 12.9 ⫾ 12.1 32.6 ⫾ 16.8b
36 weeks 2.39 (1.15) 2.73 (1.60) 0.47 (0.51) 0.71 (0.54)a
a
p ⫽ .001; bp ⫽ .0001 (Univariate comparison). (n ⫽ 25) (n ⫽ 17) (n ⫽ 24) (n ⫽ 16)
72 weeks 2.11 (1.26) 2.91 (2.39) 0.34 (0.40) 0.90 (0.93)b
(n ⫽ 23) (n ⫽ 17) (n ⫽ 23) (n ⫽ 17)
There is a main effect of group [cMDD ⬎ REC, F(1,40) ⫽ a
p ⬍ .05; bp ⬍ .005 (univariate comparison of cortisol 8:00 PM REC vs.
6.19, p ⫽ .02] and a group ⫻ time interaction (group cMDD).
[cMDD ⬎ REC] by time [entry, 36, 72 weeks] Pillais REC, recovered subjects; cMDD, persistent major depression at 72 weeks.
exact F ⫽ 4.14, p ⫽ .023). The 95% confidence limits of
MFQ scores for the cMDD group at 72 weeks were 25– 41,
indicating that all subjects still had clinically significant DHEA. A second repeated measures ANOVA was
self-report depression scores (⬎25 [Wood et al 1995]) carried out for log transformed DHEA levels (8:00 AM,
compared with 8 –18 for the REC group. 8:00 PM) with age as a covariate. There was no main effect
of group (cMDD vs. REC) [F(1,26) ⫽ 0.14, ns] but there
COMORBIDITY. At 72-week follow-up, 19(95%) of was a significant main effect of time of day on DHEA
the 20 cMDD cases remained comorbid for one or more levels (8:00 AM vs. 8:00 PM) [F(1,26) ⫽ 189.69, p ⬍ .001].
diagnosis; the range was 1– 4 (median 3). A diagnosis of There was also a significant multivariate effect of time of
OCD at entry was significantly associated with persistent assessment (entry, 36, and 72 weeks, with levels dimin-
depression at 72 weeks (exact test [one tailed], p ⫽ .029). ishing over time, Pillais exact F ⫽ 8.92, p ⫽ .001) on
Seven (78%) of the 9 subjects with OCD at entry had DHEA levels. The means and standard deviations of the
persistent depression at 72 weeks. All cases of comorbid DHEA data at entry, 36, and 72 weeks are shown in Table
depressive obsessional disorder re-assessed at 36 weeks 3.
were in the cMDD group at 72 weeks. Such cases account For the REC group, paired t tests showed that there were
for around 25% of first-episode major depressions in this significantly lower DHEA levels (8:00 AM) at 72 weeks
age range (Herbert et al 1996). No other co-morbid compared with entry [t(16) ⫽ 2 .83, p ⬍ .02], and at 72
diagnoses were associated with chronicity. compared with 36 weeks [t(15) ⫽ 2.32, p ⬍ .05]. DHEA
at 8:00 PM revealed that in the REC group there were also
Adrenal Steroids and Chronic Depression significantly lower levels at 72 weeks compared with entry
CORTISOL. A repeated measures ANOVA was car-
[t(17) ⫽ 3.87, p ⬍ .001], and at 72 compared with 36
ried out with group (cMDD vs. REC) as the between- weeks [t(15) ⫽ 3.84, p ⬍ .002].
subject factor and cortisol (log transformed) (8:00 AM,
8:00 PM) and time of assessment (entry, 36, and 72 weeks)
as within-subject factors, with age as covariate. Thirty-five Table 3. Mean (SD) DHEA Levels (ng/mL) at Each
of the 47 subjects were entered into the analysis, because Assessment
of missing hormone data points in 12 subjects. There was Clinical status at 72 weeks follow-up
a significant main effect of group [F(1,32) ⫽ 4.16, p ⫽
DHEA 8:00 AM DHEA 8:00 PM
.05] and time of day for cortisol (8:00 AM, 8:00 PM) Assessment
[F(1,33) ⫽ 150.74, p ⫽ .0001] but not time of assessment period REC cMDD REC cMDD
(entry, 36, and 72 weeks). There were no significant Entry 0.34 (0.18)a
0.36 (0.17) 0.18 (0.09) c
0.20 (0.08)
interactions. One-way ANOVAs comparing REC and (n ⫽ 21) (n ⫽ 18) (n ⫽ 22) (n ⫽ 18)
cMDD groups revealed that evening cortisol was signifi- 36 weeks 0.39 (0.31)b 0.34 (0.22) 0.19 (0.08)d 0.20 (0.10)e
cantly higher in the cMDD compared with the REC group (n ⫽ 20) (n ⫽ 16) (n ⫽ 20) (n ⫽ 15)
72 weeks 0.26 (0.20)a,b 0.26 (0.17) 0.11 (0.80)c,d 0.16 (0.10)e
at all three time points [entry: F(1,44) ⫽ 4.23, p ⬍ .05; 36 (n ⫽ 19) (n ⫽ 17) (n ⫽ 19) (n ⫽ 18)
weeks: F(1,38) ⫽ 4.26, p ⬍ .05; 72 weeks: F ⫽
Within-REC group at 8:00 AM: ap ⫽ .012 (entry vs. 72 weeks); bp ⫽ .035 (36
9.28(1,38), p ⬍ .005]. There were no differences in vs. 72 weeks).
morning cortisol at any time point (p ⬎ .05). The means Within-REC group at 8:00 PM: cp ⫽ .001 (entry vs. 72 weeks); dp ⫽ .013 (36
vs. 72 weeks).
and standard deviations of the cortisol data at entry, 36, Within-cMDD group at 8:00 PM: ep ⫽ .035 (36 vs. 72 weeks).
and 72 weeks are shown in Table 2. REC, recovered subjects; cMDD, persistent major depression at 72 weeks.
Chronic Depression In School-Aged Children
Cortisol/DHEA Ratios
A further repeated measures ANOVA was carried out with
the cortisol/DHEA ratios (8:00 AM and 8:00 PM, arsine
transformation). There were no significant multivariate
effects for group (REC vs. cMDD) or time (entry, 36, and
72 weeks).
Psychosocial Factors and Chronic Depression
ONGOING DIFFICULTIES. Three ongoing difficulties
at presentation (lack of maternal confiding relations,
family dysfunction, and friendship difficulties) have been
associated with nonrecovery at 36 weeks, although not
specifically with major depression (Goodyer et al 1997b).
None of the ongoing adversities or friendship difficulties
assessed at presentation or at 36 weeks were associated
with cMDD at 72 weeks. In addition, no other long-term
hardship or adversity at presentation was associated with
cMDD.
RECENT LIFE EVENTS. The number of undesirable
events within each category over the 72-week follow-up
period were summed. There was a significant association
with cMDD group for multiple (⬎3) disappointing events
occurring over the 72-week follow-up (REC 5/24 [21%]
vs. cMDD 13/19 [68%] exact test [one tailed], p ⬍ .005).
There was also an association between one or more
dangerous events to the self and chronic depression (REC
0/24 vs. cMDD 7/19 [37%], exact test [one tailed] p ⬍
.005). The reported occurrence of these two types of
events (multiple disappointments, one or more dangers to
the self) were significantly associated with each other
(exact test p ⬍ .02). There were no differences between
groups for the number of subjects reporting loss events.
Predicting Chronic Depression
A preliminary exploration to predict chronic first-episode
depression was made from all psychosocial and hormone
variables occurring from entry through 72 weeks, age at
entry, and persistent OCD. A backwards stepwise logistic
regression estimated that the likelihood of having chronic
depression was best predicted by recurrent undesirable life
events (multiple [⬎3] disappointments and/or one or more
dangerous events to the self) over the 72 weeks (OR ⫽
2.86, p ⫽ .027, 95% CI ⫽ 1.113–7.27), together with age
at entry (OR ⫽ 2.44, p ⫽ .008, 95% CI ⫽ 1.27– 4.69).
Further exploration showed that these recurrent undesir-
able life events (when taken as the dependent response
variable: yes ⫽ occurred, no ⫽ did not occur) were
themselves best predicted by the additive effects of higher
evening cortisol at entry (OR ⫽ 11.9, p ⫽ .034, 95% CI ⫽
1.20 –117.21) and persistent OCD (OR ⫽ 7.1, p ⫽ .047,
95% CI ⫽ 1.03– 49.33). There were no univariate associ-
BIOL PSYCHIATRY 355
2001;50:351–357
ations between number or type of life events and level of
hormones at any assessment point.
Discussion
Chronic depression, as defined in this study, occurred in
an estimated 25% of this clinic-referred sample, a figure
consistent with previous reports of clinical populations
followed in the medium term (Kovacs 1996). The subjects
with chronic depression met full DSM-IV criteria at all
three assessments and did not enter remission or recovery
between assessment times. They are therefore a clear
sub-group of patients with chronic illness rather than
recurrent or relapsing disorders. In this clinical population,
older subjects with co-morbid OCD may be more at risk
for chronicity as defined in this article. Obsessive-com-
pulsive disorder that remains beyond the first 36 weeks of
an episode appears a potentially useful clinical marker for
about one third of chronic major depressions in this
adolescent age range. The interview procedures suggested
that depressive disorder preceded the onset of OCD in
these cases, but the methods used prevent firm conclusions
regarding the precedence of co-existing disorders.
The effects of age on the liability of having chronic
MDD are marked, with 85% of such cases being older than
13 years, 6 months at entry. Subjects in this older age
group (13 years, 6 months to 17 years at entry) are also
significantly more likely to show evening cortisol hyper-
secretion, regardless of outcome group status. It may be
that this age-related difference is associated with the
liability for chronicity in adolescents compared with chil-
dren with the same affective condition. The finding that
cortisol dysregulation is a correlate of age may account for
negative findings in cortisol secretory patterns in previous
studies that have focused exclusively on childhood depres-
sion or in studies with a young adolescent group (Kutcher
and Sokolov 2001). Higher evening cortisol may also have
been obscured in previous childhood studies if no distinc-
tion was made between cases of first-episode MDD with
and without accompanying dysthymia, as this affective
sub-type appears to be specifically associated with
evening cortisol hypersecretion in this age range (Herbert
et al 1996). The current findings are consistent with the
observations that ageing is associated with poorer homeo-
static adaptations of cortisol secretion to stress in rodents
and primates throughout the life span (Bergendahl et al
2000; Gust et al 2000; Van Cauter et al 2000; Veldhuis
2000). Whether there are specific age-mediated effects
(for example, as a consequence of rises in gonadal
hormones) on the liability for cortisol hypersecretion
during adolescence requires further study. The current
findings did not measure pubertal status nor examine the
effects of gonadal steroids on course of disorder, prevent-
356 BIOL PSYCHIATRY
2001;50:351–357
ing any firm conclusions regarding the mechanisms relat-
ing to age-mediated effects on course in this age range.
Interestingly, there was no effect of age on DHEA
levels, which is surprising given the known increases in
DHEA over this age range in normal children and adoles-
cents (Goodyer et al 2000a) (Kroboth et al 1999). In fact,
DHEA levels are lowered over the 72 weeks in both
chronic and recovered groups, suggesting a potentially
negative effect of an episode of MDD on circulating levels
over time. This cannot be confirmed without longitudinal
comparisons with a group of normal adolescents matched
for age and gender and a fully recovered clinical group
with no residual depressive symptoms.
By contrast, evening cortisol is higher at entry and
remains so in those who become chronically depressed.
This is not, however, associated with increased severity
(as measured by the MFQ) at entry. Neither is it a direct
function of age. A previous article (Rao et al 1996)
suggested that higher evening plasma cortisol at entry was
associated with recurrence in adolescents over a 7-year
period; but there is no previous endocrine data on chronic
episodes of MDD in this age range. The current findings
suggest that cortisol hypersecretion at the time of presen-
tation may exert effects on the risk for chronicity, as
opposed to recurrence, of first-episode MDD in young
people. We have previously reported that cortisol hyper-
secretion in MDD is specifically associated with co-
morbid dysthymia (Herbert et al 1996). Childhood dythy-
mia is a known precursor for recurrent MDD and bipolar
disorder (Kovacs 1996). It may be that chronicity is a
consequence of preceding dysthymia, together with
evening cortisol hypersecretion present during an episode
of MDD. Cortisol hypersecretion is also now known to
precede the onset of major depression in adolescents, so in
some cases may exert effects on chronicity independent of
current or recent psychopathology (Goodyer et al 2000b).
These findings may be directly related to those experimen-
tal observations showing “endangerment” of the brain, par-
ticularly the hippocampus to adverse events (e.g., neurotox-
ins) in the presence of high cortisol (Sapolsky 1994). Also
relevant is the recent confirmation that elevated cortisol in
adults can impair declarative memory, suggesting a poten-
tially important modulating role for cortisol in social
cognitive function (Newcomer et al 1999).
In the sample discussed here, chronic depression is
associated with both subsequent dangerous life events to
the self together with further personal disappointments.
This augments our previous findings that only personal
disappointments were associated with persistent depres-
sion at 36 weeks (Goodyer et al 1997a). This widening of
negative event type in the latter course of a first episode
may be a critical social component in the risk for chronic-
ity. Dangerousness to the self is a latent psychological
I.M. Goodyer et al
inference from events with differing social characteristics,
consisting of personal accidents, physical threat, self-
harm, and violence to the subject by others. The frequency
of each of these sub-types is too small to determine which
characteristics of danger to the self are specifically related
to chronicity.
The preliminary logistic analyses for predicting cMDD
suggest that it is further adverse events in adolescent, as
opposed to childhoood, cases occurring as a function of
the disorder, that are most likely to predict chronicity. The
occurrence of these “depression-dependent” events is in-
creased in subjects with persistent OCD and cortisol
hypersecretion at entry. As already noted, raised cortisol
may impair episodic (autobiographical) memory pro-
cesses, and OCD may result in increased rumination over
the implications of negative events. These latter two
factors may operate together by first distorting the salience
of recall, increasing the negative bias toward the self (“it’s
my fault”) and then promoting amplification of such a bias
over the meaning of disappointing or dangerous events to
the self via a recursive rumination process. The absence of
a direct effect of cortisol hypersecretion on risk for
chronicity supports a psychoendrocine process of this type
in the maintenance of disorder.
This was a naturalistic study, and treatment was neither
systematically assessed nor controlled for, but at follow-up
treatment history during the episode was recorded. There
were no differences in the types of treatment (therapies
recorded included cognitive behavior therapy, psychother-
apy, family therapy, group therapy, antidepressants, sup-
portive measures, and parent counseling) provided accord-
ing to comorbidity at presentation or duration of episode.
The possibility that different treatment regimens may be
needed for chronic compared to remitting depressions in
this age range requires further investigation.
The sample size and attrition rate restrict the strength of
conclusions that can be drawn from these findings. Nev-
ertheless, they suggest that chronic depressive disorder
arising in a cohort experiencing their first episode of major
depression is more likely in adolescents compared with
children and in those with co-morbid OCD. This chronic
course may be a consequence of an increasing and
widening set of negative interpersonal events, arising in
the context of an adverse endocrine milieu, which may
modulate the personal appraisal of some types of recent
social experiences, perhaps those which in particular are
brought about in part by the depressed young person’s
own illness-related behavior.
Supported by a project grant from the Wellcome Trust. This work was
completed within the MRC (UK) co-operative in Brain, Behavior and
Neuropsychiatry.
Chronic Depression In School-Aged Children
References
Ambrosini PJ, Metz C, Prabucki K, Chi-Lee J (1987): Videotape
reliability of the third revised edition of the K-Sads. J Am
Acad Child Adolesc Psychiatry 28:723–728.
American Psychiatric Association (1994): Diagnostic and Sta-
tistical Manual For Mental and Behavioural Disorders, 4th
edition. Washington, DC: American Psychiatric Association.
Bergendahl M, Iranmanesh A, Mulligan T, Veldhuis JD (2000)
Impact of age on cortisol secretory dynamics basally and as
driven by nutrient-withdrawal stress. J Clin Endocrinol
Metab 85:2203–2214.
Costello EJ, Angold A (1988): Scales to assess child and
adolescent depression: Checklists, screens and nets. J Am
Acad Child Adolesc Psychiatry 27:726 –737.
Goodyer IM, Herbert J, Altham PME (1997a): Short term
outcome of major depression. III. A high cortisol/DHEA ratio
and subsequent disappointing life events predict persistent
depression. Psychol Med 28:265–273.
Goodyer IM, Herbert J, Altham PME, Pearson J, Secher S, Shiers
S (1996): Adrenal secretion during major depression in 8 to
16 year olds. I: Altered diurnal rhythms in salivary cortisol
and dehydroepiandrosterone (DHEA) at presentation. Psy-
chol Med 26:245–256.
Goodyer IM, Herbert J, Tamplin A, Altham PME (2000a): First
episode major depression in adolescents: Affective, cognitive,
and endocrine characteristics of risk status and predictors of
onset. Br J Psychiatry 176:142–149.
Goodyer IM, Herbert J, Tamplin A, Altham PME (2000b):
Recent life events, cortisol dehydroepiondrosterone and the
onset of major depression in high risk adolescents. Br J
Psychiatry 177:499 –504.
Goodyer IM, Herbert J, Tamplin A, Secher S, Pearson J (1997b):
Short term outcome of major depression:II. Life events,
family dysfunction and friendship difficulties as predictors of
persistent disorder. J Am Acad Child and Adolesc Psychiatry
36:474 – 480.
Guazzo EP, Kirkpatrick PJ, Goodyer IM, Shiers S, Herbert J
(1996): Cortisol, d0ehydroepandrosterone (DHEA), and
DHEA sulfate in the cerebrospinal fluid of man: Relation to
blood levels and the effects of age. J Clin Endocrinol Metab
81:3951–3960.
Gust DA, Wilson ME, Stocker T, Conrad S, Plotsky PM, Gordon
BIOL PSYCHIATRY 357
2001;50:351–357
TP (2000): Activity of the hypothalamic-pituitary-adrenal
axis is altered by aging and exposure to social stress in female
rhesus monkeys. J Clin Endocrinol Metab 85:2556 –2563.
Herbert J, Goodyer IM, Altham PME, Pearson J, Secher S, Shiers
S (1996): Adrenal secretion and major depression in 8 to 16
year olds, II. Influence of co-morbidity at presentation.
Psychol Med 26:257–263.
Kovacs M (1996): Presentation and course of major depressive
disorder during childhood and later years of the lifespan. J Am
Acad Child Adolesc Psychiatry 35:705–715.
Kroboth PD, Salek FS, Pittenger AL, Fabian TJ, Frye RF (1999):
DHEA and DHEA-S: A review. J Clin Pharmacol 39:327–
348.
Kutcher S, Sokolov S (2001): Adolescent depression: Neuroen-
docrine aspects. In: Goodyer I, editor. The Depressed Child
and Adolescent: Developmental and Clinical Perspectives.
Cambridge, UK: Cambridge University Press, 195–215.
Lewinsohn PM, Allen NB, Seeley JR, Gotlib IH (1999): First
onset versus recurrence of depression: Differential processes
of psychosocial risk. J Abnorm Psychol 108:483– 489.
Newcomer JW, Selke G, Melson AK, Hershey T, Craft S,
Richards K, Alderson AL (1999): Decreased memory perfor-
mance in healthy humans induced by stress-level cortisol
treatment. Arch Gen Psychiatry 56:527–533.
Rao U, Dahl RE, Ryan ND, Bimaker B, Williamson DE, Giles
DE et al (1996): The relationship between longitudinal
clinical course and sleep and cortisol changes in adolescent
depression. Biol Psychiatry 40:474 – 484.
Sapolsky RM (1994): The physiological relevance of glucocor-
ticoid endangerment of the hippocampus. Ann N Y Acad Sci
746:294 –304.
Van Cauter E, Leproult R, Plat L (2000): Age-related changes in
slow wave sleep and REM sleep and relationship with growth
hormone and cortisol levels in healthy men [see comments].
JAMA 284:861– 868.
Veldhuis JD (2000): Nature of altered pulsatile hormone release
and neuroendocrine network signalling in human ageing:
Clinical studies of the somatotropic, gonadotropic, cortico-
tropic and insulin axes. Novartis Found Symp 227:163–185;
discussion: 185–189.
Wood A, Kroll L, Moore A, Harrington RC (1995): Properties of
the mood and feelings questionnaire: A research note. J Child
Psychol Psychiatry 36:327–334.