Article

Natural Course of Adolescent Major Depressive Disorder

in a Community Sample:
Predictors of Recurrence in Young Adults

Peter M. Lewinsohn, Ph.D. Objective: The primary purpose was to
identify factors related to the recurrence
of major depressive disorder during
Paul Rohde, Ph.D.
young adulthood (19–23 years of age) in a
community sample of formerly depressed
John R. Seeley, M.S. adolescents.
Method: A total of 274 participants with
Daniel N. Klein, Ph.D.
adolescent-onset major depressive disor-
der were assessed twice during adoles-
Ian H. Gotlib, Ph.D. cence and again after their 24th birthday.
Lifetime psychiatric information was ob-
tained from their first-degree relatives.
Adolescent predictor variables included
demographic characteristics, psychoso-
cial variables, characteristics of adoles-
cent major depressive disorder, comor-
bidity, family history of major depressive
disorder and nonmood disorder, and an-
tisocial and borderline personality disor-
der symptoms.
Results: Low levels of excessive emo-
tional reliance, a single episode of major
depressive disorder in adolescence, low
proportion of family members with recur-
rent major depressive disorder, low levels
of antisocial and borderline personality
disorder symptoms, and a positive attri-
butional style (males only) independently
predicted which formerly depressed ado-
lescents would remain free of future psy-
chopathology. Female gender, multiple
major depressive disorder episodes in ad-
olescence, higher proportion of family
members with recurrent major depres-
sive disorder, elevated borderline person-
ality disorder symptoms, and conflict with
parents (females only) independently pre-
dicted recurrent major depressive disor-
der. Comorbid anxiety and substance use
disorders in adolescence and elevated an-
tisocial personality disorder symptoms in-
dependently distinguished adolescents
who developed recurrent major depres-
sive disorder comorbid with nonmood
disorder from those who developed pure
major depressive disorder.
Conclusions: Formerly depressed ado-
lescents with the risk factors identified in
this study are at elevated risk for recur-
rence of major depressive disorder during
young adulthood and therefore warrant
continued monitoring and preventive or
prophylactic treatment.

(Am J Psychiatry 2000; 157:1584–1591)

I t is now well documented that major depressive disor-

der is one of the most common mental disorders in ado-
lescence (e.g., references 1–3). Given the high prevalence
of major depressive disorder in adolescence, examining
the natural course of episodes in this age group is clearly
important. In a previous study (4), we found that individu-
als who had experienced an episode of major depressive
disorder by age 19 were at significantly elevated risk for fu-
ture major depressive disorder, compared to participants
with adolescent-onset nonmood disorders and adoles-
cents with no disorder. In these three groups, rates of ma-
jor depressive disorder over the 5-year period from age 19
through 23 were 45.0%, 28.2%, and 18.5%, respectively. In
addition, compared to no-disorder comparison subjects,
formerly depressed adolescents had significantly higher
rates of nonmood disorder in the follow-up period (33.2%
versus 19.5%). The primary purpose of the study reported
here was to identify the factors in formerly depressed ado-
lescents that predict major depressive disorder recurrence
in young adulthood.
To our knowledge, only six studies have examined pre-
dictors of recurrence of major depressive disorder in de-
pressed children or adolescents. Five studies examined
patient groups (5–9), and one examined the offspring (6–
23 years of age) of adults treated for depression (10). The
results of these studies are inconsistent, perhaps because
of the generally small size of the study groups and the con-
siderable variation in participants’ ages, duration of the
follow-up periods, treatment settings, and assessment
methods.
The framework that guides our research is based on an
integrative, multifactorial model (11), in which depression
is conceptualized as the end result of environmentally ini-
tiated changes in behavior, affect, and cognitions. The
model distinguishes between antecedents, which occur
before the onset of the depression, and consequences,

1584 Am J Psychiatry 157:10, October 2000


which are observable during and after an episode of de-
pression and which may influence the risk of recurrence.
Antecedents examined in the study reported here included
having lived in a stressful environment during adolescence
(e.g., a nonintact household, a high level of conflict with
parents, stressful life events). The model also recognizes
individual differences (vulnerabilities and protective fac-
tors) that increase or decrease the impact of the antecedent
events and moderate the impact of the consequences of
depression on recurrence. Vulnerabilities examined in this
study included psychopathology among family members,
female gender (which has often been shown to be a risk
factor), elevated level of depression symptoms, depresso-
typic cognitions, and excessive emotional reliance. Protec-
tive factors included self-rated social competence and pos-
itive coping skills. The model, like most other current
theories of the etiology of depression, does not distinguish
between first onset and recurrent episodes of depression,
even though the results of other studies (12–14) have sug-
gested that these two types of episodes may be predicted
by different variables.
Consistent with the expectation that having experi-
enced an episode of depression creates vulnerabilities that
effect recurrence, clinical aspects of the episode of major
depressive disorder in adolescence (e.g., early onset,
longer duration, recurrence during adolescence, greater
severity, treatment utilization, and suicide attempts) were
hypothesized to be associated with recurrence. Comorbid
adolescent psychopathology was also hypothesized to be
a vulnerability for recurrence. Last, traits measuring se-
lected young adult personality disorders (i.e., antisocial
and borderline personality disorders) were posited to re-
flect stable characteristics of the person that constitute
vulnerabilities for recurrence.
This study focuses on individuals who experienced and
recovered from an episode of major depressive disorder
during adolescence, which we defined as before age 19.
On the basis of their psychiatric outcome between age 19
and 23, individuals were assigned to four mutually exclu-
sive groups: 1) those with no further disorder, 2) those with
a recurrent episode of major depressive disorder but no
comorbid mental disorder, 3) those with a recurrent epi-
sode of major depressive disorder that was accompanied
by a nonmood disorder, and 4) those with no recurrence of
major depressive disorder but who experienced a non-
mood disorder.
This design allowed us to address three questions. First,
what factors predict staying well (group 1 versus groups 2,
3, and 4). Second, what factors predict major depressive
disorder recurrence (groups 1 and 4 versus groups 2 and
3). Third, what factors predict pure (i.e., noncomorbid)
major depressive disorder versus major depressive disor-
der with a co-occurring nonmood disorder (group 2 ver-
sus group 3).
Am J Psychiatry 157:10, October 2000
LEWINSOHN, ROHDE, SEELEY, ET AL.
Method
Participants
Probands. Participants were a subset of individuals from the Or-
egon Adolescent Depression Project. Participants were originally
randomly selected from nine senior high schools in western Ore-
gon. A total of 1,709 adolescents (ages 14–18; mean age at initial
assessment=16.6 years, SD=1.2) completed the initial assessment
(time 1), which consisted of an interview and questionnaires, be-
tween 1987 and 1989. Approximately 1 year later (time 2), 1,507
participants (88.2%) participated in a reassessment that used the
same interview questions and questionnaires (mean interval be-
tween time 1 and time 2=13.8 months, SD=2.3) (additional details
are provided elsewhere [2]).
As probands from the Oregon Adolescent Depression Project
reached their 24th birthday, participants with a history of major
depressive disorder at time 2 (N=360), those with a history of non-
mood disorder at time 2 (N=284), and a subset of those with no
history of mental disorder at time 2 (N=457) were invited to par-
ticipate in a time 3 interview. The no-disorder comparison group
was representative of the entire group of participants with no
mental disorder at time 2 (N=863) in age and gender within age;
all participants with nonwhite ethnicity were invited to partici-
pate in the time 3 assessment.
Of the 1,101 young adults selected for time 3 interview, 940 par-
ticipated (85.4%). Of those participants, 57.2% were female, 89.0%
were white, 34.1% were married, 96.8% had graduated from high
school, and 31.4% had a bachelor’s degree or a higher educational
level. Their average age at time 3 was 24.2 years (SD=0.6). Women
were more likely than men to complete the time 3 assessments
(88.9% versus 81.0%) (χ2=13.55, df=1, N=1,101, p<0.001). Differ-
ences in time 3 participation as a function of other demographic
characteristics or time 2 diagnostic status were nonsignificant.
A total of 315 time 3 participants had experienced an episode of
major depressive disorder before age 19. Forty-one of these par-
ticipants were deselected to better clarify the predictors of major
depressive disorder recurrence between age 19–23: 22 were expe-
riencing the major depressive disorder episode at the time of their
19th birthday, three met criteria for bipolar disorder after age 19,
three met criteria for dysthymia but not major depressive disor-
der after age 19, and 13 met criteria for adjustment disorder with
depressed mood but not for major depressive disorder after age
19. Thus, the reference group for the present study consisted of
274 participants who had experienced and recovered from an ep-
isode of major depressive disorder by age 19.
Family members. Lifetime psychiatric information was ob-
tained from the first-degree relatives over the age of 13 (biological
parents, full siblings) of the Oregon Adolescent Depression
Project participants in the time 3 assessment. To supplement the
direct interviews, informant psychiatric data were collected from
probands or another first-degree relative. Our goal was to collect
diagnostic data from two sources for each family member. Diag-
nostic information was available for the relatives of 234 of the 274
probands with a history of major depressive disorder during ado-
lescence (85.4%). Of the 761 first-degree relatives for whom diag-
nostic information was collected, direct interviews were obtained
for 442 (58.1%); data for 145 (19.1%) of the family members were
based on a single source.
After a thorough description of the study, written informed
consent was obtained from Oregon Adolescent Depression
Project probands (and their guardians, if applicable) and from the
family members.
Diagnostic Interviews
Oregon Adolescent Depression Project probands. Pa r t i c i -
pants were interviewed at time 1 with a version of the Schedule
1585
COURSE OF ADOLESCENT DEPRESSION
for Affective Disorders and Schizophrenia for School-Age Chil-
dren (K-SADS) that combined features of the epidemiologic ver-
sion (15) and the present episode version and included additional
items to derive DSM-III-R diagnoses. At time 2 and time 3, partic-
ipants were interviewed with the Longitudinal Interval Follow-Up
Evaluation (16), which elicited detailed information about the
course of psychiatric symptoms and disorders since the previous
interview. Time 3 interviews were conducted by telephone, and
time 3 diagnoses were made by using DSM-IV criteria.
Most diagnostic interviewers had an advanced degree in clini-
cal or counseling psychology or social work, and all were exten-
sively trained before data collection. Interrater reliability for life-
time diagnoses in a randomly selected subsample at time 1 (N=
233) was moderate to excellent: kappa=0.86 for major depressive
disorder, and kappa=0.76–0.89 for nonmood disorder categories.
For a randomly selected subsample at time 3 (N=178), interrater
reliability was excellent for the primary categories of major de-
pressive disorder (kappa=0.87) and nonmood disorder (kappa=
0.82).
Family members. Parents and siblings were interviewed with
either the Structured Clinical Interview for DSM-III-R, Nonpa-
tient Version (SCID-NP) (17) or the K-SADS employed in the time
1 proband assessment, modified for collection of DSM-IV criteria.
Family history data were collected by using the revised Family In-
formant Schedule and Criteria (18), which is based on the Family
History Research Diagnostic Criteria (19), modified for DSM-IV
criteria. Using all available data, the project’s four senior diag-
nosticians (P.R., D.N.K., P.M.L., and Nicholas Allen), who were
blind to proband diagnoses, derived best-estimate diagnoses (20)
for relatives.
Outcome Variable: Psychopathology
in Early Adulthood
The primary outcome variable used in this study was the de-
velopment of another episode of major depressive disorder dur-
ing early adulthood. The 274 formerly depressed participants
were categorized into four groups: 1) no disorder, consisting of
87 participants (31.8%) who experienced neither major depres-
sive disorder nor a nonmood disorder from age 19 through 23;
2) major depressive disorder, consisting of 58 participants
(21.2%) who had a recurrence of major depressive disorder by
age 24 but no nonmood disorder during the 19–23 age period;
3) major depressive disorder plus nonmood disorder, consisting
of 67 participants (24.5%) who had a recurrence of major depres-
sive disorder and experienced a nonmood disorder during the
19–23 age period; and 4) nonmood disorder, consisting of 62 par-
ticipants (22.6%) who experienced a nonmood disorder but did
not have a recurrence of major depressive disorder during the
19–23 age period. Of the 129 participants with a nonmood disor-
der during the 19–23 age period, 99 (76.7%) had a diagnosis of
substance use disorder, 44 (34.1%) anxiety disorder, three (2.3%)
oppositional defiant disorder, three (2.3%) conduct disorder, two
(1.6%) eating disorder, one (0.8%) schizophreniform disorder,
and one (0.8%) somatoform disorder. The majority of these non-
mood disorder episodes (58.1%) were new disorders that began
after age 19 (i.e., did not represent the continuation of a non-
mood disorder that began before age 19 and persisted into the
19–23 age period). Twenty-two of the 274 participants (8.0%)
were experiencing a current major depressive disorder episode
at the time 1 assessment; differences in diagnostic outcome as a
function of time 1 current major depressive disorder status were
nonsignificant (χ2=3.39, df=3, N=274, n.s.).
Predictor Variables
Demographic characteristics. Four dichotomous time 1 vari-
ables were included: 1) gender; 2) family intactness (living with
versus not living with both biological parents); 3) highest parental
1586
education within the household (bachelor’s degree versus less
than a bachelor’s degree), which was included as a proxy measure
of socioeconomic status; and 4) ethnicity (white or nonwhite).
Psychosocial variables. An extensive battery of psychosocial
measures was administered at time 1 to all participants. All mea-
sures had been previously shown to possess very good psycho-
metric properties (21, 22)). Variables were standardized and
scored such that higher values reflected more problematic func-
tioning.
Current depression was assessed by using the 20-item National
Institute of Mental Health Center for Epidemiologic Studies De-
pression Scale (CES-D Scale) (23) (alpha=0.89; time 1–time 2 r=
0.61) (df=1505, p<0.001 for all correlations). Negative cognitions
were assessed with 27 items on self-reinforcement, likelihood of
future positive events, dysfunctional attitudes, and perceived
control over one’s life (alpha=0.81; time 1–time 2 r=0.61). Attribu-
tional style was assessed with the 48-item Kastan Attributional
Style Questionnaire for Children (N. Kaslow, R.L. Tanenbaum,
M.E.P. Seligman, unpublished 1978 questionnaire) (alpha=0.63;
time 1–time 2 r=0.55). Self-esteem was assessed with nine items
on physical appearance and general self-esteem (alpha=0.81;
time 1–time 2 r=0.62). Excessive emotional reliance was assessed
with 10 items on the extent to which the individual desires exces-
sive support and approval from others and is interpersonally sen-
sitive (alpha=0.83; time 1–time 2 r=0.54). Self-rated social compe-
tence was assessed with 12 items (alpha=0.85; time 1–time 2 r=
0.64). Coping skills were assessed with 17 items (alpha=0.76, time
1–time 2 r=0.55). Social support from friends was assessed with 13
items (alpha=0.81, time 1–time 2 r=0.60). Social support from
family was assessed with 22 items (alpha=0.86, time 1–time 2 r=
0.64). Conflict with parents was assessed with the 45-item Issues
Checklist scale (24) (alpha=0.81; time 1–time 2 r=0.51). Daily has-
sles were assessed with 20 items from the Unpleasant Events
Schedule (25) (alpha=0.79; time 1–time 2 r=0.55). Major life
events were assessed by using questions about 11 negative life
events involving self or significant other (parent, sibling, other
relative, close friend) during the past year (alpha=0.78; time 1–
time 2 r=0.52). Academic problems were assessed with nine items
from the questionnaire and the time 1 K-SADS interview (e.g.,
lifetime occurrence of school expulsion or suspension, truancy,
repeating a grade in school, most recent grade point average).
Physical illness at time 1 was assessed by using 90 items on phys-
ical symptoms, number of sick days, and physician visits during
the past year. Cigarette use was assessed in the time 1 K-SADS in-
terview with a question about whether the participant had been a
daily cigarette smoker.
Adolescent major depressive disorder. T h e f o l l ow i n g s i x
characteristics of adolescent major depressive disorder were in-
cluded in the analysis: 1) age at onset of first major depressive dis-
order episode before age 13 years (N=47 of 274, 17.2%) versus at
age 13 or older, 2) duration of the first major depressive disorder
episode of 2–8 weeks (N=136 of 274, 49.6%) versus 9 or more
weeks, 3) single major depressive disorder episode before age 19
(N=207 of 274, N=75.5%) versus two or more episodes, 4) severity
of the worst episode indicated by the presence of five to seven
DSM-IV symptoms (N=175 of 274, 63.9%) versus eight or nine
symptoms, 5) treatment utilization (outpatient or inpatient treat-
ment or medications) for major depressive disorder during ado-
lescence (N=58 of 274, 21.2%) versus no treatment, and 6) history
of a suicide attempt during adolescence (N=62 of 274, 22.6%) ver-
sus no attempt.
Psychiatric comorbidity during adolescence. Four catego-
ries of comorbid psychiatric disorders occurring before age 19
were examined: 1) dysthymia (N=20 of 274, 7.3%), 2) anxiety dis-
orders (N=74 of 274, 27.0%), 3) substance use disorder (N=80 of
274, 29.2%), and 4) conduct disorder/oppositional defiant disor-
Am J Psychiatry 157:10, October 2000

der (N=35 of 274, 12.7%). For the 143 participants with nonmood
comorbidity during adolescence, the major depressive disorder
episode was categorized as either primary (N=42, 29.4%) or sec-
ondary (N=101, 70.6%) on the basis of the age at onset of the two
disorders; primary major depressive disorder could not be deter-
mined for eight participants due to simultaneous onset with the
nonmood disorder.
Family history of psychopathology. Three variables were in-
cluded as indications of the degree of psychopathology in the
family: 1) the proportion of first-degree relatives (parents and sib-
lings) with major depressive disorder, 2) the proportion of rela-
tives with recurrent major depressive disorder, and 3) the propor-
tion of relatives with nonmood disorder.
Personality disorders. The two most frequent and relevant
DSM-IV personality disorders (antisocial and borderline) were
assessed by using relevant portions of the Personality Disorder
Examination (26). Personality Disorder Examination dimensional
scores (i.e., summation of partial and full symptom criteria) were
used in the analyses. The interrater reliability of the Personality
Disorder Examination dimensional scores was excellent (intrac-
lass correlation coefficients>0.80).
Statistical Analyses
To address the three primary questions posed in this study
(What are the predictors of staying well, of major depressive dis-
order recurrence, and of pure depression?), three multiple logistic
regression analyses were conducted with each potential predictor
variable. The first multiple logistic regression analysis compared
the no-disorder control group to the remaining three outcome
groups (i.e., major depressive disorder, major depressive disorder
plus nonmood disorder, and nonmood disorder). The second
multiple logistic regression analysis identified the predictors of
major depressive disorder recurrence by comparing participants
who experienced major depressive disorder (i.e., the groups with
major depressive disorder and with major depressive disorder
plus nonmood disorder) with those who did not (i.e., the groups
with no disorder and with nonmood disorder). The third multiple
logistic regression analysis compared predictors of pure major
depressive disorder versus comorbid major depressive disorder
plus nonmood disorder. To test for possible moderating effects of
gender, gender was entered in the model in a second block within
each multiple logistic regression analysis, and the predictor-by-
gender interaction term was entered in the third block. By using
variables that emerged as significant univariate predictors of the
specific contrast, three summary multiple logistic regression
models with the likelihood ratio backward selection procedure
were conducted to identify unique predictors (i.e., variables that
remained significant when controlling for other variables in the
model).
The strength of the associations is illustrated by the odds ratio
with 95% confidence interval (CI). For the examined contrasts
and for interactions with gender, with alpha set to 0.05, statistical
power was sufficient (>0.80) to detect medium effect sizes or
larger (Cohen’s w > 0.30). Although numerous tests were per-
formed, thereby inflating the type I error rate, statistical signifi-
cance was based on p<0.05 to identify predictors of potential clin-
ical relevance.
Results
Univariate Predictors
Adolescent predictor variables with one or more signifi-
cant contrasts are shown in Table 1. The first four columns
of Table 1 illustrate the percentage or mean value of the
variable in the four diagnostic outcome groups. The last six
Am J Psychiatry 157:10, October 2000
LEWINSOHN, ROHDE, SEELEY, ET AL.
columns of Table 1 present odds ratios and CIs for the three
contrasts.
Gender interactions. Two variables interacted signifi-
cantly with gender. Although nonsignificant as a main ef-
fect, attributional style had a significant interaction with
gender in the first contrast, predicting staying well (χ2=4.83,
df=1, N=274, p<0.05). When the results were examined sep-
arately by gender, positive attributional style predicted no-
disorder status between age 19 and 23 for male participants
(odds ratio=2.21, 95% CI=1.20–4.09) but not female partici-
pants (odds ratio=1.06, 95% CI=0.79–1.42). In the second
contrast, conflict with parents had a significant gender in-
teraction in predicting recurrent major depressive disorder
(χ2=5.47, df=1, N=274, p<0.05), having a significant effect
for female participants (odds ratio=1.38, 95% CI=1.01–1.87)
but not for male participants (odds ratio=0.71, 95% CI=
0.43–1.16).
Multivariate Predictors
To examine whether the variables distinguishing the out-
come groups remained significant when controlling for the
effects of the other variables, summary multiple logistic re-
gression analyses using the likelihood ratio backward selec-
tion procedure were conducted for each of the three con-
trasts. The variables with unique and independent main
effects retained in the final model for each contrast are indi-
cated in Table 1.
Of the variables predicting which formerly depressed ad-
olescents remained free of future psychopathology, low lev-
els of excessive emotional reliance, a single (as opposed to
multiple) major depressive disorder episode in adoles-
cence, a low proportion of family members with recurrent
major depressive disorder, low rates of antisocial and bor-
derline personality disorder symptoms, and the interaction
of attributional style and gender were retained in the first
contrast. The solution correctly classified 78.0% of the par-
ticipants (sensitivity=77.4%).
In the second contrast, female gender, multiple major de-
pressive disorder episodes in adolescence, a higher propor-
tion of family members with recurrent major depressive
disorder, elevated rates of borderline personality disorder
symptoms, and the interaction of conflict with parents and
gender were retained as predictive of recurrent major de-
pressive disorder in young adulthood. The solution cor-
rectly classified 72.0% of participants (sensitivity=72.6%).
Although female gender was retained as an independent
predictor of recurrent major depressive disorder, examin-
ing the gender composition of the four outcome groups in-
dicated that gender differences between no-disorder, major
depressive disorder, and major depressive disorder plus
nonmood disorder groups were not statistically significant,
suggesting that formerly depressed young men made up a
significantly higher proportion of the pure nonmood disor-
der group.
In the third contrast, comorbid anxiety disorder and sub-
stance use disorder in adolescence and elevated levels of
1587
COURSE OF ADOLESCENT DEPRESSION

TABLE 1. Characteristics of Subjects With a History of Major Depressive Disorder in Adolescence by Diagnostic Outcome at
Age 19–23 Years and Comparisons of Diagnostic Outcome Groups
Diagnostic Outcome at Age 19–23 Years
Planned Contrasts
3. Major
Depressive Groups 1 and 4
2. Major Disorder Plus Group 1 Versus Versus Group 2 Versus
1. No Depressive Nonmood 4. Nonmood Groups 2–4 Groups 2 and 3 Group 3
Characteristic Disorder Disorder Disorder Disorder Odds Odds Odds
in Adolescence (N=87) (N=58) (N=67) (N=62) Ratio 95% CI Ratio 95% CI Ratio 95% CI
Female gender (%) 71.3 86.2 73.1 50.0 0.92 0.53–1.61 2.29*a 1.33–3.95 0.44 0.17–1.09
Psychosocial
variablesb
Emotional reliance
(mean z score) –0.21 0.01 0.27 0.00 1.38*a 1.06–1.79 1.31* 1.03–1.67 1.29 0.91–1.84
Conflict with parents
(mean z score) –0.21 0.05 0.10 0.14 1.38* 1.05–1.81 1.15 0.91–1.47 1.07 0.73–1.56
Daily hassles
(mean z score) –0.19 0.12 0.11 0.03 1.34* 1.02–1.75 1.24 0.97–1.57 0.98 0.70–1.39
Academic problems
(mean z score) –0.20 –0.31 0.23 0.33 1.36* 1.04–1.77 0.96 0.75–1.22 1.80* 1.21–2.67
Cigarette use (%) 23.0 19.0 43.3 33.9 1.62 0.90–2.91 1.24 0.74–2.09 3.26* 1.44–7.37
Characteristics of
depressive episodes
Duration of first
episode ≥9 weeks (%) 40.2 58.6 55.4 50.8 1.81* 1.08–3.03 1.64* 1.01–2.66 0.88 0.43–1.79
≥2 episodes (%) 11.5 36.2 27.7 24.6 3.20*a 1.54–6.65 2.30*a 1.29–4.05 0.67 0.31–1.45
Severity of worst
episode consisted of
eight or nine
symptoms of DSM-
III-R major depres-
sive disorder (%) 23.0 44.8 38.5 40.0 2.33* 1.30–4.15 1.65 1.00–2.74 0.77 0.37–1.58
Suicide attempt (%) 14.9 20.7 31.3 25.8 2.02* 1.03–3.96 1.48 0.84–2.62 1.75 0.77–3.97
Comorbidity
Anxiety disorder (%) 21.8 13.8 40.3 32.3 1.49 0.82–2.71 1.09 0.64–1.87 4.22*a 1.73–10.29
Substance use
disorder (%) 14.9 10.3 49.3 45.2 3.18* 1.64–6.15 1.19 0.71–2.01 8.41* a 3.18–22.22
Conduct disorder/
oppositional
defiant disorder (%) 6.9 5.2 20.9 19.4 2.48 0.99–6.21 1.15 0.56–2.33 4.84* 1.32–17.82
Primary major
depressive disorder
in subjects with
nonmood
comorbidity (%) 28.6 45.5 18.0 34.9 0.95 0.38–2.37 1.34 0.65–2.75 3.80* 1.26–11.48
Family history of
psychopathology
(mean proportion of
first-degree relatives
with disorder)
Major depressive
disorder 0.27 0.33 0.44 0.28 2.65* 1.02–6.92 3.59* 1.50–8.63 2.90 0.87–9.66
Recurrent major
depressive disorder 0.11 0.23 0.27 0.14 7.42*a 1.89–29.11 9.46*a 2.92–30.61 1.64 0.43–6.25
Nonmood disorder 0.45 0.46 0.59 0.51 2.00 0.85–4.69 1.70 0.77–3.75 3.77* 1.13–12.55
Axis II psychopathologyc
Borderline (mean
score out of a
possible 18) 0.20 0.98 2.25 0.70 2.73*a 1.73–4.31 1.80*a 1.42–2.28 1.34* 1.09–1.65
Antisocial (mean
score out of a
possible 44) 0.87 1.00 4.11 3.89 1.41*a 1.20–1.67 1.04 0.97–1.11 1.51*a 1.22–1.86
a Variable was a unique and independent predictor in the summary multiple logistic regression.
b Measured with an extensive battery of psychosocial measures (21, 22).
c Measured with the relevant sections of the Personality Disorder Examination (27).
*p<0.05.

borderline personality disorder symptoms were retained as nonmood disorder from those who developed pure major
unique measures distinguishing adolescents who devel- depressive disorder in early adulthood. The solution cor-
oped recurrent major depressive disorder comorbid with rectly classified 77.2% of participants (sensitivity=80.0%).

1588 Am J Psychiatry 157:10, October 2000


Discussion
Formerly depressed adolescents who developed psychi-
atric problems during young adulthood were character-
ized during adolescence by more severe depressive epi-
sodes (e.g., longer episode duration, multiple episodes,
greater number of symptoms, history of suicide attempts);
by elevated stress, conflict, and interpersonal depen-
dency; and, for males only, by a negative attributional
style. These individuals also had an elevated rate of sub-
stance use disorder during adolescence, and, as young
adults, they showed traits associated with borderline and
antisocial personality disorder. In addition, their parents
and siblings were more likely to have experienced both
single-episode and recurrent major depressive disorder.
Variables that independently contributed to the predic-
tion of psychopathology in young adulthood included
emotional reliance, multiple depressive episodes, a family
history of recurrent major depressive disorder, borderline
and antisocial personality disorder symptoms, and attri-
butional style (for males). Excessive emotional reliance is
a component of the broader construct of interpersonal de-
pendency (27), which has been implicated theoretically in
the etiology of depression, especially in the psychoana-
lytic literature (e.g., references 28, 29). We previously re-
ported that excessive emotional reliance predicted de-
pression onset in adolescents (22) and that emotional
reliance remained elevated after major depressive disor-
der recovery (30). The results reported here indicate that
emotional reliance is also predictive of the recurrence of
psychopathology, particularly of major depressive disor-
der, underscoring both the centrality of emotional reli-
ance to depression theory and its potential explanatory
breadth to other disorders.
Variables that specifically predicted major depressive
disorder recurrence included multiple depressive epi-
sodes in adolescence, a family history of recurrent major
depressive disorder, borderline personality disorder
symptoms, and, for females only, increased conflict with
parents. Gender was retained in the multivariate solution,
but this finding appeared to be due to the greater likeli-
hood that formerly depressed male adolescents would de-
velop pure nonmood disorders in young adulthood. These
findings suggest that clinical characteristics, both of the
proband and of the first-degree relatives, are among the
strongest predictors of major depressive disorder recur-
rence.
There was no evidence that adolescent comorbidity
acted as a risk factor for pure major depressive disorder in
young adulthood. The presence of nonmood disorders in
adolescence predicted nonmood disorders in young
adulthood, generally comorbid with recurrent major de-
pressive disorder. Contrary to previous research (e.g., ref-
erences 7, 10), comorbid dysthymia in the present study
failed to predict major depressive disorder recurrence.
Unlike previous research on the impact of double depres-
Am J Psychiatry 157:10, October 2000
LEWINSOHN, ROHDE, SEELEY, ET AL.
sion (i.e., major depressive disorder superimposed on a
preexisting dysthymia), adolescent major depressive dis-
order and dysthymia in the present study occurred largely
at different periods. Additional research is needed to dis-
entangle whether concurrent comorbidity of major de-
pressive disorder and dysthymia specifically confers a risk
for major depressive disorder recurrence.
Familial psychopathology was a significant predictor in
all three contrasts. Most significantly, the proportion of
family members with recurrent major depressive disorder
predicted major depressive disorder recurrence in the off-
spring, with and without a nonmood disorder. These find-
ings are consistent with previous research indicating that
the presence of major depressive disorder in family mem-
bers significantly increases the likelihood of major depres-
sive disorder recurrence in adults (e.g., references 31–34).
The results reported here extend previous research by in-
dicating that recurrent depression breeds true from parent
to offspring.
Both elevated antisocial and borderline personality dis-
order symptoms were significant in multiple contrasts.
Several adult studies have suggested strong comorbidity
between depression and borderline personality disorder
symptoms (e.g., references 35–38) and a small number of
studies of adult patients have shown that axis II pathology,
especially cluster B disorders (i.e., antisocial, borderline,
narcissistic, histrionic), increases the likelihood of depres-
sion recurrence in treated adult patients (e.g., references
39–41). Our findings on axis II psychopathology need to be
qualified in two respects. First, rates of personality disor-
ders were low, so elevated dimensional scores were gener-
ally in the subthreshold range. Second, the data on axis II
psychopathology were obtained at time 3. Therefore, the
direction of these associations cannot be determined, al-
though previous research supports the hypothesis that ad-
olescent depression precedes the onset of personality dis-
order symptoms (42).
The majority of psychosocial variables assessed during
adolescence were not found to predict recurrence of psy-
chopathology, including negative cognitions, self-esteem,
social skills, coping, life events, social support, and ele-
vated depression symptoms (as measured by the CES-D
Scale). The most plausible explanation for these nonsig-
nificant findings is that their prognostic impact in previ-
ous studies was time limited. Thus, the extent to which a
variable’s predictive ability diminishes as a function of
time needs to be explored in future research. A second ex-
planation is that the assessment of these variables in the
design of the study reported here may have occurred be-
fore, during, or after the adolescent major depressive dis-
order episode.
A depressotypic attributional style predicted future psy-
chopathology in formerly depressed young men. There is
a large literature, conducted primarily with adults, exam-
ining the role of depressotypic cognitions in predicting re-
currence. The findings of these studies are mixed, with
1589
COURSE OF ADOLESCENT DEPRESSION
some investigations reporting significant results (e.g., ref-
erence 43) and others reporting nonsignificant effects
(e.g., reference 41). Important issues that need to be ad-
dressed in future research concern whether depressotypic
cognitions are measured at intake or at the end of treat-
ment, whether results obtained with adults are replicated
in samples of adolescents and children, and whether the
cognitions specifically predict depression.
A number of limitations of the present study should be
acknowledged. Given the absence of previous research in
this area, a stringent correction for type I error was not ap-
plied, and the results need to be considered preliminary
pending independent replication. Second, although a rel-
atively large number of predictors were examined, some
potentially important categories of predictors (e.g., bio-
logical measures, childhood physical and sexual abuse)
were not represented. Third, although our sample was
community based, which reduced a number of selection
biases, the sample selection may also limit the applicabil-
ity of our findings to clinical practice. Offsetting these lim-
itations were several noteworthy strengths of this study,
including the large sample size, prospective design, rigor-
ous diagnostic interviews, and assessment of a broad array
of predictors, including direct interview data on family
history of psychopathology. Overall, the findings support
arguments by Warner et al. (10), Ilardi et al. (41), and
Goodyer et al. (44) that the predictors of depression onset,
duration (persistence), and recurrence may differ in im-
portant ways.
The clinical implications of this study are straightfor-
ward. Clinicians are routinely faced with decisions about
what to do with youngsters with a history of major depres-
sive disorder. We previously demonstrated that all individ-
uals with a history of major depressive disorder during
adolescence are at very high risk for major depressive dis-
order recurrence and for other psychiatric disorders during
young adulthood (4). In the study reported here, roughly
one-fourth of formerly depressed adolescents experienced
subsequent pure major depressive disorder, one-fourth ex-
perienced comorbid major depressive disorder, and one-
fourth remained free from depression recurrence but expe-
rienced a nonmood disorder. Earlier and better identifica-
tion of the most salient risk factors is the first step in avoid-
ing a protracted course of disorder. Depressed female
adolescents (especially those who experience high conflict
with parents), adolescents with multiple major depressive
disorder episodes, and adolescents with a family history of
recurrent depression are at particularly high risk for de-
pression recurrence and should be closely monitored or re-
ceive continued prophylactic treatment. The association
between recurrent depression and borderline personality
disorder features indicates that individuals with recurrent
depression may warrant additional treatment focused on
emerging axis II psychopathology.
1590
Received June 14, 1999; revisions received Jan. 27 and Apr. 28,
2000; accepted May 12, 2000. From the Oregon Research Institute;
the Department of Psychology, State University of New York, Stony
Brook; and the Department of Psychology, Stanford University, Stan-
ford, Calif. Address reprint requests to Dr. Lewinsohn, Oregon Re-
search Institute, 1715 Franklin Blvd., Eugene, OR 97403-1983.
This research was supported in part by NIMH grants MH-40501,
MH-50522, and MH-52858.
References
1. Fleming JE, Offord DR: Epidemiology of childhood depressive
disorders: a critical review. J Am Acad Child Adolesc Psychiatry
1990; 29:571–580
2. Lewinsohn PM, Hops H, Roberts RE, Seeley JR, Andrews JA: Ad-
olescent psychopathology, I: prevalence and incidence of de-
pression and other DSM-III-R disorders in high school students.
J Abnorm Psychol 1993; 102:133–144
3. Reinherz HZ, Giaconia RM, Lefkowitz ES, Pakiz B, Frost AK:
Prevalence of psychiatric disorders in a community population
of older adolescents. J Am Acad Child Adolesc Psychiatry 1993;
32:369–377
4. Lewinsohn PM, Rohde P, Klein DN, Seeley JR: Natural course of
adolescent major depressive disorder, I: continuity into young
adulthood. J Am Acad Child Adolesc Psychiatry 1999; 38:56–63
5. Emslie GJ, Rush AJ, Weinberg WA, Gullion CM, Rintelmann J,
Hughes CW: Recurrence of major depressive disorder in hospi-
talized children and adolescents. J Am Acad Child Adolesc Psy-
chiatry 1997; 36:785–792
6. Harrington RC, Fudge H, Rutter M, Pickles A, Hill J: Adult out-
comes of childhood and adolescent depression, I: psychiatric
status. Arch Gen Psychiatry 1990; 47:465–473
7. Kovacs M, Feinberg TL, Crouse-Novack MA, Paulauskas SL,
Finkelstein R: Depressive disorders in childhood, I: a longitudi-
nal prospective study of characteristics and recovery. Arch Gen
Psychiatry 1984; 41:229–237
8. McCauley E, Myers K, Mitchell J, Calderon R, Schloredt K, Trader
R: Depression in young people: initial presentation and clinical
course. J Am Acad Child Adolesc Psychiatry 1993; 32:714–722
9. Rao U, Ryan ND, Birmaher B, Dahl RE, Williamson DE, Kaufman
J, Rao R, Nelson B: Unipolar depression in adolescents: clinical
outcome in adulthood. J Am Acad Child Psychiatry 1995; 34:
566–578
10. Warner V, Weissman MM, Fendrich M, Wickramaratne P,
Moreau D: The course of major depression in the offspring of
depressed parents: incidence, recurrence, and recovery. Arch
Gen Psychiatry 1992; 49:795–801
11. Lewinsohn PM, Hoberman H, Teri L, Hautzinger M: An integra-
tive theory of depression, in Theoretical Issues in Behavior
Therapy. Edited by Reiss S, Basin R. San Diego, Academic Press,
1985, pp 331–359
12. Post RM, Weiss SB, Leverich GS: Recurrent affective disorder:
roots in developmental neurobiology and illness progression
based on changes in gene expression. Dev Psychopathol 1994;
6:781–813
13. Monroe SM, Rohde P, Seeley JR, Lewinsohn PM: Life events and
depression in adolescence: relationship loss as a prospective
risk factor for first onset of major depressive disorder. J Ab-
norm Psychol 1999; 108:606–614
14. Lewinsohn PM, Allen NB, Seeley JR, Gotlib IH: First onset versus
recurrence of depression: differential processes of psychoso-
cial risk. J Abnorm Psychol 1999; 108:483–489
15. Orvaschel H, Puig-Antich J, Chambers WJ, Tabrizi MA, Johnson
R: Retrospective assessment of prepubertal major depression
with the Kiddie-SADS-E. J Am Acad Child Psychiatry 1982; 21:
392–397
Am J Psychiatry 157:10, October 2000

16. Keller MB, Lavori PW, Friedman B, Nielsen E, Endicott J, Mc-
Donald-Scott P, Andreasen NC: The Longitudinal Interval Fol-
low-Up Evaluation: a comprehensive method for assessing out-
come in prospective longitudinal studies. Arch Gen Psychiatry
1987; 44:540–548
17. Spitzer RL, Williams JBW, Gibbon M, First MB: The Structured
Clinical Interview for DSM-III-R (SCID), I: history, rationale, and
description. Arch Gen Psychiatry 1992; 49:624–629
18. Mannuzza S, Fyer AJ: Family Informant Schedule and Criteria
(FISC), July 1990 revision. New York, New York State Psychiatric
Institute, Anxiety Disorders Clinic, 1990
19. Endicott J, Andreasen N, Spitzer RL: Family History Research Di-
agnostic Criteria, 3rd ed. New York, New York State Psychiatric
Institute, Biometrics Research, 1978
20. Leckman JF, Sholomskas D, Thompson D, Belanger A, Weiss-
man MM: Best estimate of lifetime psychiatric diagnosis: a
methodological study. Arch Gen Psychiatry 1982; 39:879–883
21. Andrews JA, Lewinsohn PM, Hops H, Roberts RE: Psychometric
properties of scales for the measurement of psychosocial vari-
ables associated with depression in adolescence. Psychol Rep
1993; 73:1019–1046
22. Lewinsohn PM, Roberts RE, Seeley JR, Rohde P, Gotlib IH, Hops
H: Adolescent psychopathology, II: psychosocial risk factors for
depression. J Abnorm Psychol 1994; 103:302–315
23. Radloff LS: The CES-D Scale: a self-report depression scale for
research in the general population. J Applied Psychol Measure-
ment 1977; 1:385–401
24. Robin AL, Weiss JG: Criterion-related validity of behavioral and
self-report measures of problem-solving communication skills
in distressed and nondistressed parent-adolescent dyads. Be-
havioral Assessment 1980; 2:339–352
25. Lewinsohn PM, Mermelstein RM, Alexander C, MacPhillamy DJ:
The Unpleasant Events Schedule: a scale for the measurement
of aversive events. J Clin Psychol 1985; 41:483–498
26. Loranger AW, Susman VL, Oldham JM, Russakoff M: The Per-
sonality Disorder Examination (PDE) Manual. Yonkers, NY, DV
Communications, 1988
27. Hirschfeld RMA, Klerman GL, Lavori P, Keller MB, Griffith P,
Coryell W: Premorbid personality assessments of first onset of
major depression. Arch Gen Psychiatry 1989; 46:345–350
28. Blatt S, Quinlan D, Chevron E, McDonald C, Zuroff D: Depen-
dency and self-criticism: psychological dimensions of depres-
sion. J Consult Clin Psychol 1982; 50:113–124
29. Hirschfeld RMA, Klerman GL, Chodoff P, Korchin S, Barrett J:
Dependency–self-esteem–clinical depression. J Am Acad Psy-
choanal 1976; 4:373–388
Am J Psychiatry 157:10, October 2000
LEWINSOHN, ROHDE, SEELEY, ET AL.
30. Rohde P, Lewinsohn PM, Seeley JR: Are adolescents changed by
an episode of major depression? J Am Acad Child Adolesc Psy-
chiatry 1994; 33:1289–1298
31. Bland RC, Newman SC, Orn H: Recurrent and nonrecurrent de-
pression: a family study. Arch Gen Psychiatry 1986; 43:1085–
1089
32. Gonzales LR, Lewinsohn PM, Clarke GN: Longitudinal follow-up
of unipolar depressives: an investigation of predictors of re-
lapse. J Consult Clin Psychol 1985; 53:461–469
33. Merikangas KR, Wicki W, Angst J: Heterogeneity of depression:
classification of depressive subtypes of longitudinal course. Br
J Psychiatry 1994; 164:342–348
34. Weissman MM, Gammon GD, John K, Merikangas KR, Warner V,
Prusoff BA, Sholomskas D: Children of depressed parents: in-
creased psychopathology and early onset of major depression.
Arch Gen Psychiatry 1987; 44:847–853
35. Farmer R, Nelson-Gray RO: Personality disorders and depres-
sion: hypothetical relations, empirical findings, and method-
ological considerations. Clin Psychol Rev 1990; 10:453–475
36. Akiskal HS, Hirschfeld RMA, Yerevanian BI: The relationship of
personality to affective disorders. Arch Gen Psychiatry 1983;
40:801–810
37. Oldham JM, Skodol AE, Kellman HD, Hyler SE, Doidge N, Ros-
nick L, Gallagher PE: Comorbidity of axis I and axis II disorders.
Am J Psychiatry 1995; 152:571–578
38. Skodol AE, Stout RL, McGlashan TH, Grilo CM, Gunderson JG,
Shea MT, Morey LC, Zanarini MC, Dyck IR, Oldham JM: The co-
occurrence of mood and personality disorders: a report from
the Collaborative Longitudinal Personality Disorders Study
(CLPS). Depress Anxiety 1999; 10:175–182
39. Peselow ED, Fieve RR, DiFiglia C: Personality traits and re-
sponse to desipramine. J Affect Disord 1992; 24:209–216
40. Pfohl B, Coryell W, Zimmerman M, Stangl D: Prognostic validity
of self-report and interview measures of personality disorder
in depressed inpatients. J Clin Psychiatry 1987; 48:468–472
41. Ilardi SS, Craighead WE, Evans DD: Modeling relapse in unipo-
lar depression: the effects of dysfunctional cognitions and per-
sonality disorders. J Consult Clin Psychol 1997; 65:381–391
42. Cohen P: Childhood risks for young adult symptoms of person-
ality disorder: method and substance. Multivariate Behavioral
Res 1996; 31:121–148
43. Simons AD, Murphy GE, Levine JL, Wetzel RD: Cognitive therapy
and pharmacotherapy for depression: sustained improvement
over one year. Arch Gen Psychiatry 1986; 43:43–48
44. Goodyer IM, Herbert J, Secher SM, Pearson J: Short-term out-
come of major depression, I: comorbidity and severity at pre-
sentation as predictors of persistent disorder. J Am Acad Child
Adolesc Psychiatry 1997; 36:179–187
1591