Content uploaded by Dan J. Stein
Author content
All content in this area was uploaded by Dan J. Stein
Content may be subject to copyright.
Rethinking the duration requirement for generalized
anxiety disorder: evidence from the National
Comorbidity Survey Replication
RONALD C. KESSLER
1
*, NANCY BRANDENBURG
2
, MICHAEL LANE
1
,
PETER ROY-BYRNE
3
,PAULD.STANG
4
,DANJ.STEIN
5
AND HANS-ULRICH WITTCHEN
6
1
Department of Health Care Policy, Harvard Medical School, Boston, MA, USA ;
2
Pfizer, Inc., New York,
NY, USA;
3
Department of Psychiatry and Behavioral Sciences, University of Washington School of Medicine,
Seattle, WA, USA;
4
Department of Public Health, West Chester University, and Galt Associates, Blue Bell,
PA, USA;
5
Department of Psychiatry, University of Stellenbosch, Cape Town, South Africa;
6
Institute
of Clinical Psychology and Psychotherapy, Technical University of Dresden, Dresden, Germany
ABSTRACT
Background. The proposed revisions of the ICD and DSM diagnostic systems have led to increased
interest in evaluation of diagnostic criteria. This report focuses on the DSM-IV requirement that
episodes of generalized anxiety disorder (GAD) must persist for at least 6 months. Community
epidemiological data are used to study the implications of changing this requirement in the range
1–12 months for estimates of prevalence, onset, course, impairment, co-morbidity, associations
with parental GAD, and sociodemographic correlates.
Method. Data come from the US National Comorbidity Survey Replication (NCS-R), a US
household survey carried out during 2001–2003. Version 3.0 of the WHO Composite International
Diagnostic Interview (WMH-CIDI) was used to assess DSM-IV anxiety disorders, mood disorders,
substance disorders, and impulse-control disorders.
Results. Lifetime, 12-month, and 30-day prevalence estimates of DSM-IV GAD changed from
6
.
1%, 2
.
9%, and 1
.
8% to 4
.
2–12
.
7%, 2
.
2–5
.
5%, and 1
.
6–2
.
6% when the duratio n requirement
was changed from 6 months to 1–12 months. Cases with episodes of 1–5 months did not diff er
greatly from those with episodes of o6 months in onset, persi stence, impairment, co-morbidity,
parental GAD, or sociodemographic correlates.
Conclusions. A large number of people suffer from a GAD-like syndrome with episodes of <6
months duration. Little basis for excluding these people from a diagnosis is found in the associ-
ations examined here.
INTRODUCTION
The minimum duration requirement for gen-
eralized anxiety disorder (GAD) was 1 month
when GAD was introduced in DSM-III (APA,
1980). However, clinical studies found that
DSM-III GAD seldom occurred in the absence
of other co-morbid anxiety or mood disorders
(Breier et al. 1985; Breslau, 1985), suggesting
that GAD might better be conceptualized as a
prodrome, residual, or severi ty marker than an
independent disorder (Breslau & Davis, 1985 b;
Clayton et al. 1991 ; Noyes et al. 1992). Based on
clinical evidence that co-morbidity substantially
decreased with episode duration (Breslau &
Davis, 1985 a), GAD was retained in DSM-III-R
* Address for correspondenc e : R. C. Kessler, Ph.D., Department
of Health Care Policy, Harvard Medical School, 180 Longwood
Avenue, Boston, MA, USA 02115.
(Email: kessler@hcp.med.harvard.edu)
The views and opinions expressed in this report are those of the
authors and should not be construed to represent the views of any of
the sponsoring organizations, agencies, or U.S. Government.
Psychological Medicine, 2005, 35, 1–10. f 2005 Cambridge University Press
doi:10.1017/S0033291705004538 Printed in the United Kingdom
1
with an increased duration requirement of 6
months (APA, 1987). This change also ad-
dressed the problem of distinguishing short epi-
sodes of GAD from situational stress reactions
(Barlow & Wincze, 1998). A requirement that
‘unrealistic, hard to control worry’ be present
was also included in DSM-III-R to sharpen the
distinction between GAD and non-specific dis-
tress associated with other anxiety and mood
disorders (Barlow et al. 1986). The ICD-10
criteria for research also require a 6-month
duration (WHO, 1993), but the ICD-10 criteria
for clinical practice take a middle position on
duration by requiring GAD to last ‘ several
months’ (WHO, 1992).
This variation in required duration could
dramatically influence the number of people
classified with GAD. Contrary to the motiv-
ation for increasing the DSM-III-R duration
requirement, epidemiological data have sub-
sequently shown that GAD is not more
co-morbid than most other anxiety or mood
disorders (Kessler et al. 2001, 2004b) and that
the extremely high co-morbidity of GAD in
early clinical studies was due to a help-seeking
bias (Wittchen et al. 1994). The 6-month
duration requirement was nonetheless retained
in DSM-IV (APA, 1994), whi ch means that
episodes of shorter duration receive no diag-
nosis even if they recur over many years. It has
been suggested that such ‘orphaned’ cases of
GAD are substantial in number (Rickels &
Rynn, 2001).
Three large community epidemiological
studies examined whether episodes of GAD with
durations of less than 6 months might either
be less impairing (Maier et al. 2000) or less
co-morbid with other DSM disorders (Bienvenu
et al. 1998; Carter et al. 2001) than episodes
lasting more than 6 months. No significant dif-
ferences were found in any of these studies. One
of these three pap ers (Bienvenu et al. 1998) also
examined and found little variation in profiles of
demographic correlates of DSM-III-R GAD
with required episode duration. The current
report builds on these earlier studi es by pres-
enting data from a na tionally representative
US household sample on the relationship be-
tween required GAD episode duration in the
range 1–12 months and a wider range of corre-
lates: onset, course, impairment, co-morbidity,
parental GAD, and sociodemographics.
METHOD
Sample
Data come from the National Comorbidity
Survey Replication (NCS-R) (Kessler et al.
2003; Kessler & Merikangas, 2004), a face-to-
face survey (n=9282) of the US household
population conducted during 2001–2003. A
multi-stage clustered area probability sampling
design was used to select the sample (Kessler
et al. 2003, 2004a). The response rate was
70
.
9%. Recruitment began with a letter and
study fact brochure followe d by an in-person
interviewer visit to explain study aims and pro-
cedures. Verbal rather than written consent was
used to be consistent with the baseline NCS
(Kessler et al. 1994). Respondents received $50.
The human subjects committees of Harvard
Medical School and the University of Michigan
both approved these procedures.
All respondents were administered a Part I
diagnostic interview (described below), while a
probability subsample of 5692 respondents also
received a Part II interview. Part II respondents
included all who met lifetime criteria for any
Part I disorder plus a probability sample of
other respondents. The Part I sample was
weighted to adjust for differential probabilities
of selection within households and for differ-
ences in intensity of recruitment effort among
hard-to-recruit cases. The Part II sample was
additionally weighted for differential selection
probabilities from Part I. A final weight
adjusted the sample to match the 2000 census
population on geographic and sociodemo-
graphic variables. All analyses reported here
employ these weights. More complete infor-
mation on NCS-R sampling design and weight-
ing is reported elsewhere (Kessler et al. 2004 a).
Diagnostic assessment
NCS-R diagnoses are based on Version 3.0 of
the WHO Composite International Diagnostic
Interview (CIDI) (Kessler & Ustun, 2004), a
fully structured lay-administered interview that
generates diagnoses according to both ICD-10
(WHO, 1993) and DSM-IV criteria. DSM-IV
criteria are used in the current report. The
diagnoses include anxiety disorders, mood
disorders, substance disorders, and a group of
disorders that share a common feature of
difficulties with impulse control (intermittent
2 R. C. Kessler et al.
explosive disorder and three retrospectively
reported childhood-adolescent disorders –
oppositional-defiant disorder, conduct disorder,
and attention-deficit hyperactivity disorder).
Diagnostic hierarchy rules and organic ex-
clusion rules were used in making diagnoses. As
detailed elsewhere (Kessler et al. in press), blind
clinical re-interviews using the Structured
Clinical Interview for DSM-IV (SCID; First
et al. 2002) with a probability subsample of
NCS-R respondents found generally good con-
cordance between DSM-IV diagnoses based on
the WMH-CIDI and the SCID for anxiety,
mood, and substance use disorders. WMH-
CIDI diagnoses of impulse-control disorders
were not validated. Concordance between CIDI
and SCID diagnoses of GAD was good (AUC
0
.
83; OR 51
.
0, 95% CI 11
.
5–226
.
2).
As the assessment of episode duration plays a
central part in the following analyses, it is im-
portant to understand this assessment in CIDI
3.0. After inquiring about ep isodes of worry or
anxiety with associated symptoms that lasted at
least 1 month, the word ‘ episode’ is explicitly
defined: ‘ When I use the word ‘‘ episode ’’ in the
next question, I mean a time lasting 1 month or
longer when most days you were (worried/
nervous or anxious/anxious or worried) and
also had some of the other problems we just
reviewed. The episode ends when you no longer
have these feelings for a full month.’ After pro-
viding this definition, respon dents are asked
about 12-month prevalence and then about
number of lifetime episodes. Depending on
number of episodes reported, maximum dur-
ation is then assessed by asking ‘ How long
did that episode last?’ or ‘How long did the
longest of these episodes last ? ’ Responses pre-
sumably overestimate duration because episodes
are defined as not ending until there is a full
month without any symptoms rather than a
month when symptoms occurred fewer than half
the days.
Other measures
Three other sets of measures used here are role
impairment, parental GAD, and sociodemo-
graphics. Role impairment among 12-month
cases was assessed in two ways. First, respon-
dents were asked the number of days in the past
year they were ‘ totally unable ’ to carry out their
daily activities because of their GAD. Second,
respondents were administered the Sheehan
Disability Scales (Leon et al. 1997), 0–10 visual
analogue scales that ask respondents to focus on
the one month in the past year when their GAD
was most severe and to rate how much GAD
interfered with their home management, work,
social life, and personal relationships. The scales
include response options of none (0), mild (1–3),
moderate (4–6), severe (7–9), and very severe
(10). Parental history of GAD was assessed with
questions developed to expand the Family
History RDC interview (Andreasen et al. 1977)
to include GAD (Kendler et al. 1997).
Sociodemographic correlates include gender,
age at interview, race-ethnicity, education,
marital status, employment status, and region.
Analysis methods
Prevalence was estimated using the minimum
duration requirements 1, 3, 6, and 12 months as
well as using mutually exclusive subgroups with
durations of 1–2, 3–5, 6–11, and o12 months .
The actuarial method (Halli et al. 1992) was
used to calculate age-of-onset curves for mutu-
ally exclusive duration subgroups. Means were
calculated in these same subgroups for various
measures of illness severity and course. Logistic
regression analysis was used to examine vari-
ation across these subgroups in associations of
lifetime GAD with co-mor bid DSM-IV dis-
orders, parental history of GAD, and socio-
demographic variables. Standard errors and
significance tests were obtained using the Taylor
series linearization method (Wolter, 1985) im-
plemented in the
SUDAAN software package
(Research Triangle Institute, 2002) to adjust
for weighting and clustering. Multivariate sig-
nificance was evaluated using Wald x
2
tests
based on Taylor series de sign-based coefficient
variance–covariance matrices. Statistical sig-
nificance was evaluated at the 0
.
05 level with
two-sided tests.
RESULTS
Prevalence, onset, and course
Lifetime GAD prevalence estimates range from
4
.
2% (12-month minimum duration require-
ment) to 12
.
7% (1-month minimum duration
requirement) (Table 1). The range of prevalence
estimates is more narrow, but still substantial,
for estimates of 12-month (2
.
2–5
.
5%) and
GAD: evidence from the NCS-R 3
1-month (1
.
6–2
.
6%) prevalence. The ratio of
12-month to lifetime prevalence, an indirect in-
dicator of persistence, varies significantly with
required duration, from 51
.
1% for 12-month
minimum duration to 43
.
3% for 1-mont h
minimum duration.
Retrospective age-of-onset reports were used
to generate cumulative lifetime prevalence
curves for respondents with longest episodes
of 1–2, 3–5, 6–11, and o12 months. Although
the distributions differ significantly (x
2
3
=13
.
0,
p=0
.
005), they are substantively very similar
(Fig. 1). All have the highest slopes in the age
range between the early teens and early 30’s. All
have median ages of onset in the early to mid
20’s. All have few onsets after age 55.
Other descriptors of illness course are similar
across duration subgroups (Table 2). Mean age
of onset is somewhat earlier for shorter dur-
ations. Mean number of years in episode is sig-
nificantly higher among cases with 12-month
duration (10
.
4) than shorter durations (4
.
9–6
.
5;
F
1
=13
.
2–58
.
8, p<0
.
001), but does not differ by
episode duration in the range 1–2, 3–5, or 6–11
months (F
2
=1
.
4, p=0
.
270). Respondent reports
about number of years in an episode were div-
ided by the difference in reported ages of onset
and recency to define annual persistence. The
mean of this ratio (0
.
5–0
.
6) does not differ
substantially with episode duration, although
the difference across subg roups is statistically
significant (x
2
3
=9
.
1, p<0
.
001). The average
number of months in the past year with an
episode among 12-month cases also varies
predictably with episode duration.
Role impairment
Episode duration is unrelated to the per cent of
days respondents were totally unable to carry
out their usual daily activities because of their
GAD in the worst month of the past year
(13
.
5–18
.
1; F
3
=0
.
4, p=0
.
725) as well as to nine
of 12 nested severity responses on the Sheehan
Disability Scales (results not shown, but avail-
able on request ; x
2
3
=2
.
8–6
.
2, p=0
.
104–0
.
416),
including impaired work performance (18
.
6–
30
.
2% severe), personal relationships (23
.
7–
41
.
3% severe), and social life (32
.
5–46
.
3%
severe). The three exceptions are significant
associations of duration with severe, moderate,
and any impairment in home management
(x
2
3
=9
.
4–26
.
2, p=0
.
009–0
.
049), but these are
all non-monotonic and do not remain signifi-
cant when we compare episode durations of 1–5
versus o6 months (x
2
1
=0
.
5–1
.
5, p=0
.
218–
0
.
483).
Co-morbidity
Lifetime GAD has an elevated odds ratio (OR)
with each of the other DSM-IV disorders as-
sessed in the NCS-R across the range of GAD
duration de finitions (Table 3). Ninety per cent
of these ORs are significant at the 0
.
05 level,
with a median size of 3
.
8 and an inter-quartile
range (IQR, 25th–75th percentiles) of 2
.
8–5
.
2.
Episode duration is unrelated to size of OR in
nearly three-quarters of the comparisons. The
most dramatic exception is a very high OR
(18
.
7) of dysthymia with GAD of >12 months
duration. ORs also vary with duration of GAD
for two anxiety disorde rs [panic disorder and
post-traumatic stress disorder (PTSD)] and two
other mood disorders (bipolar I disorder and
major depressive disorder). In each of these
instances, the ORs generally increase with
longer GAD episode durations. The increases
involving the two anxiety disorders are most
evident between durations of 1–5 and o6
months, while the increases involving the two
mood disorders are monotonic across the range
of durations.
Table 1. Lifetime, 12-month, and 1-month
prevalence estimates of DSM-IV generalized
anxiety disorder by variation in required mini-
mum episode duration and in maxi mum episode
duration (n=9282)
a
Lifetime 12 months 1 month
%(
S.E.) % (S.E.) % (S.E.)
I. Required minimum
episode duration
1 month 12
.
7(0
.
5) 5
.
5(0
.
3) 2
.
6(0
.
2)
3 months 8
.
0(0
.
4) 3
.
9(0
.
2) 2
.
1(0
.
2)
6 months 6
.
1(0
.
3) 2
.
9(0
.
2) 1
.
8(0
.
2)
12 months 4
.
2(0
.
2) 2
.
2(0
.
1) 1
.
6(0
.
2)
II. Maximum
episode duration
1–2 months 4
.
7(0
.
2) 1
.
6(0
.
2) 0
.
5(0
.
1)
3–5 months 1
.
9(0
.
2) 1
.
0(0
.
1) 0
.
3(0
.
0)
6–11 months 1
.
9(0
.
2) 0
.
8(0
.
1) 0
.
2(0
.
1)
o12 months 4
.
2(0
.
2) 2
.
2(0
.
1) 1
.
6(0
.
2)
a
Defined using DSM-IV criteria with organic exclusion and di-
agnostic hierarchy rules, but varying the duration requirement for
the longest episode.
4 R. C. Kessler et al.
Parental history of GAD
We began the evaluation of parental GAD by
focusing on respondents who reported that
either none or only one of their parents had
GAD and testing whether the gender of the
disordered parent was associated with respon-
dent risk and, if so, whether this varied by the
gender of the respondent. No evidence of sig-
nificant gender specificity for parent, respon-
dent, or parent-respondent consistency was
found using any of the four minimum duration
requirements (x
2
1
=0
.
0–3
.
1, p=0
.
080–0
.
829).
We then tested whether respondent risk of
GAD differed depending on whether only one
versus both parents had a history of GAD. No
such evidence was found using any of the four
minimum duration requirements (x
2
1
=0
.
0–2
.
5,
p=0
.
111–0
.
853). Based on these results, fur-
ther evaluations included only a single vari-
able for either or bot h parents having GAD to
predict respondent GAD. Parental GAD was
Table 2. Mean age of onset, years with disorder, annual persistence among lif etime cases of DSM-IV
generalized anxiety disorder and months in episode in the past 12 months among 12-month cases by
maximum episode duration
a
Maximum episode duration
F test p
1–2 months 3–5 months 6–11 months o12 months
%(
S.E.) % (S.E.) % (S.E.) % (S.E.)
Age of onset 25
.
9(0
.
7) 26
.
9(1
.
2) 29
.
3(1
.
4) 27
.
8(0
.
7) 3
.
0* 0
.
043
Years with disorder 5
.
7(0
.
4) 4
.
9(0
.
5) 6
.
5(1
.
0) 10
.
4(0
.
5) 22
.
6* <0
.
001
Annual persistence
b
0
.
5(0
.
0) 0
.
5(0
.
0) 0
.
5(0
.
0) 0
.
6(0
.
0) 9
.
1* <0
.
001
Months in episode
c
2
.
4(0
.
1) 3
.
9(0
.
3) 4
.
7(0
.
4) 9
.
0(0
.
4) 106
.
7* <0
.
001
(n) (439) (166) (175) (412)
a
Defined using DSM-IV criteria with organic exclusion and diagnostic hierarchy rules, but varying the duration requirement for the longest
episode.
b
The ratio of number of years with an episode divided by the difference between age of onset and age of most recent episode.
c
Months in episode in the past year among 12-month cases.
* Significant difference across the four subgroups at the 0
.
05 level.
0·0
0·1
0·2
0·3
0·4
0·5
0·6
0·7
0·8
0·9
1·0
0 20406080
Onset age (years)
Cumulative probabilityof GAD
1–2 months, n = 439
3–5 months, n = 166
6–11 months, n = 175
.12 months, n = 412
χ
2*
3
= 13·0, p = 0·005
FIG. 1. Cumulative age-of-onset distribution of lifetime DSM-IV generalized anxiety disorder (GAD) by maximum episode
duration (defined using DSM-IV criteria with organic exclusion and diagnostic hierarchy rules, but varying the duration require-
ment for the longest episode). * Significantly different distributions across the four subgroups at the 0
.
05 level.
GAD: evidence from the NCS-R 5
a significant predictor of respondent GAD
both among females and among males in
each of the four duration subgroups, with
ORs in the range 2
.
3–3
.
4 among females and
3
.
5–8
.
6 among males (results not shown, but
available on request). No significant difference
was found in the ORs depending on respon-
dent episode duration either among females
(x
2
3
=2
.
0, p=0
.
563) or males ( x
2
3
=5
.
6, p=
0
.
131).
Sociodemographic predictors
Statistically significant sociodemographic pre-
dictors of broadly defined (1-month minimum
duration) lifetime GAD include being younger
than 60 years, hav ing more than a high school
education, having an ‘other’ occupational
status (mostly unemployed and disabled), and
being previously married, Non-Hispanic White,
and female (res ults not shown, but available on
Table 3. Lifetime co-morbidity (OR)
a
of other DSM-IV generalized anxiety disorder (GAD)
with other DSM-IV disorders by maximum GAD episode duration
b
Maximum episode duration
x
2
p
1–2 months 3–5 months 6–11 months o12 months
OR (95% CI) OR (95% CI) OR (95% CI) OR (95 % CI)
I. Anxiety disorders
Panic disorder 4
.
5* (3
.
1–6
.
5) 3
.
6* (2
.
1–6
.
2) 8
.
3* (5
.
2–13
.
3) 7
.
3* (4
.
9–10
.
8) 8
.
4** (0
.
039)
Agoraphobia (w/o Panic) 5
.
1* (2
.
8–9
.
3) 4
.
1* (1
.
6–10
.
4) 3
.
8* (1
.
1–13
.
3) 5
.
9* (3
.
6–9
.
8) 0
.
9(0
.
829)
Specific phobia 3
.
6* (2
.
8–4
.
6) 5
.
0* (3
.
3–7
.
4) 5
.
0* (3
.
6–7
.
0) 3
.
4* (2
.
7–4
.
3) 6
.
2(0
.
102)
Social phobia 5
.
2* (4
.
0–6
.
6) 5
.
6* (3
.
8–8
.
0) 5
.
7* (4
.
0–8
.
3) 5
.
3* (4
.
3–6
.
7) 0
.
2(0
.
971)
Post-traumatic stress disorder 3
.
5* (2
.
8–4
.
5) 3
.
3* (1
.
8–5
.
9) 6
.
4* (4
.
4–9
.
5) 6
.
5* (4
.
9–8
.
5) 30
.
2** (<0
.
001)
Obsessive–compulsive disorder 3
.
2* (1
.
3–8
.
1) 4
.
5* (1
.
2–15
.
9) 3
.
7* (1
.
2–11
.
8) 4
.
3* (2
.
2–8
.
6) 0
.
3(0
.
953)
Any anxiety disorder 5
.
5* (4
.
4–6
.
8) 5
.
1* (3
.
7–7
.
1) 8
.
6* (6
.
5–11
.
5) 5
.
7* (4
.
7–6
.
9) 9
.
2** (0
.
027)
II. Mood disorders
Major depressive disorder 6
.
3* (5
.
1–7
.
9) 8
.
2* (6
.
0–11
.
0) 10
.
7* (7
.
8–14
.
5) 12
.
4* (9
.
4–16
.
4) 16
.
0** (0
.
001)
Dysthymia 3
.
8* (2
.
6–5
.
7) 2
.
0* (1
.
0–4
.
2) 4
.
7* (2
.
8–7
.
7) 18
.
7* (13
.
6–25
.
8) 85
.
4** (<0
.
001)
Bipolar I disorder 4
.
0* (2
.
5–6
.
4) 5
.
7* (3
.
6–9
.
1) 6
.
9* (4
.
3–11
.
0) 9
.
0* (5
.
8–14
.
1) 17
.
1** (0
.
001)
Bipolar II disorder 4
.
4* (1
.
4–14
.
0) 1
.
7(0
.
2–13
.
7) 4
.
8(0
.
6–38
.
2) 3
.
5(0
.
7–16
.
7) 0
.
3(0
.
960)
Any mood disorder 6
.
3* (5
.
0–7
.
9) 7
.
9* (6
.
0–10
.
5) 10
.
6* (7
.
6–14
.
6) 14
.
3* (10
.
7–19
.
3) 20
.
2** (<0
.
001)
III. Substance disorders
Alcohol abuse or dependence 2
.
2* (1
.
8–2
.
7) 1
.
9* (1
.
3–2
.
8) 2
.
1* (1
.
3–3
.
2) 2
.
1* (1
.
7–2
.
6) 0
.
3(0
.
960)
Drug abuse or dependence 2
.
1* (1
.
6–2
.
9) 1
.
9(0
.
9–3
.
9) 2
.
0* (1
.
2–3
.
1) 2
.
5* (1
.
8–3
.
5) 1
.
6(0
.
668)
Any substance disorder 2
.
0* (1
.
7–2
.
5) 1
.
9* (1
.
3–2
.
7) 2
.
1* (1
.
4–3
.
2) 2
.
1* (1
.
7–2
.
6) 0
.
4(0
.
942)
IV. Impulse-control disor ders
Attention defici t disorder 2
.
6* (1
.
8–3
.
9) 3
.
1* (1
.
5–6
.
3) 3
.
5* (1
.
9–6
.
4) 3
.
6* (2
.
3–5
.
7) 1
.
5(0
.
690)
Oppositional defiant disorder 2
.
5* (1
.
8–3
.
6) 3
.
8* (2
.
4–6
.
1) 1
.
8* (1
.
0–3
.
1) 2
.
8* (1
.
6–4
.
8) 5
.
0(0
.
174)
Intermittent explosive disorder 4
.
3* (3
.
3–5
.
6) 3
.
0* (1
.
7–5
.
2) 3
.
4* (2
.
0–5
.
8) 3
.
8* (2
.
8–5
.
2) 2
.
4(0
.
503)
Conduct disorder 1
.
7* (1
.
1–2
.
7) 2
.
4* (1
.
1–5
.
3) 2
.
2* (1
.
3–3
.
9) 2
.
8* (1
.
9–4
.
2) 5
.
8(0
.
123)
Any impulse disorder 3
.
5* (2
.
6–4
.
7) 3
.
8* (2
.
4–6
.
1) 2
.
7* (1
.
8–4
.
0) 3
.
4* (2
.
4–4
.
8) 2
.
3(0
.
506)
V. Any disorder
Exactly one disorder 0
.
9(0
.
7–1
.
2) 1
.
2(0
.
7–2
.
1) 1
.
3(0
.
8–1
.
9) 0
.
8(0
.
5–1
.
0) 4
.
7(0
.
197)
Exactly two disorders 2
.
0* (1
.
4–2
.
7) 2
.
0* (1
.
1–3
.
5) 2
.
1* (1
.
3–3
.
5) 2
.
4* (2
.
0–3
.
0) 1
.
9(0
.
602)
Three or more disorders 6
.
8* (5
.
6–8
.
2) 5
.
9* (4
.
4–8
.
0) 7
.
9* (5
.
1–12
.
1) 10
.
2* (7
.
9–13
.
5) 8
.
8** (0
.
033)
(n)
c
(4945) (4674) (4682) (4921)
a
The odds ratios (ORs) were estimated in logistic regression equations with a dummy variable for GAD predicting a dummy variable for
the co-morbid disorder, controlling for age of respondent with dummy variables for the following age groups: five-year intervals between
25–29 and 60–64 and o65 years. The age controls were used to adjust for spurious associations between GAD and other disorders due to age
differences in lifetime risk of all these disorders, yielding an approximation of the weighted average of the within-year ORs across the full age
range of the sample.
b
All disorders were defined using DSM-IV criteria with organic exclusion and diagnostic hierarchy rules. The definition of GAD varied the
duration requirement for the longest episode.
c
The samples used to estimate each of the four equations include respondents with no lifetime GAD (n=4510) plus those in one and only
one of the four GAD episode duration subgroups of 1–2 months (n =435), 3–5 months (n=164), 6–11 months (n=172), or o12 months
(n=411). For example, the 4945 respondents used to estimate the equation for GAD with a dura tion of 1–2 months include the 4510 with no
lifetime GAD plus the 435 having a history of GAD with maximum duration of 1–2 months.
* Significant at the 0
.
05 level, two-sided test.
** Significant difference of ORs across the four subgroups at the 0
.
05 level.
6 R. C. Kessler et al.
request). Although all these predictors are sig-
nificant, the ORs are uniformly modest in mag-
nitude (1
.
3–2
.
3). Nested analyses (results not
shown, but available on request) showed that
four of these six predictors are not significantly
related to any of the successively more restric-
tive defin itions of GAD: education (x
2
3
=
0
.
2–2
.
7, p=0
.
448–0
.
975), employment status
(x
2
4
=2
.
9–5
.
6, p=0
.
231–0
.
577), race-ethnicity
(x
2
3
=1
.
4–5
.
7, p=0
.
129–0
.
699), and gender
(x
2
1
=0
.
1–0
.
2, p=0
.
648–0
.
807). The association
of respondent age with lifetime GAD, in com-
parison, varies significantly depending on epi-
sode duration, with increasingly higher
prevalence in the middle age group s (30 –44 and
45–59) relative to either the younger (18–29) or
older (o60) age groups as the duration re-
quirement increases. The association of marital
status with lifetime GAD also varies by episode
duration, with risk decreasing among the never
married and increasing among the previously
married relative to the married as the duration
requirement increases.
DISCUSSION
The analyses reported here have two import-
ant limitations. First, diagnoses were based on
fully structured lay interviews rather than clin-
ician interviews, although clinical reappraisal
showed good concordance with the SCID in a
random subsample of cases. Second, duration
was assessed retrospectively with a single ques-
tion. No data are available on the accuracy of
these retrospective reports. Given the impor t-
ance of episode duration for these analyses, it is
noteworthy that respondent age at interview
and retrospectively reported age of onset are
both significantly related to reported episode
duration. These associations raise the possi-
bility of either recall failure among older re-
spondents with a history of short episodes or
progression from shorter to longer episodes
with age. The cross-sectional NCS-R design is
incapable of capturing either of these dynamic
possibilities.
Within the context of these limitations, we
found substantive ly plausible variation in esti-
mates of prevalence by ep isode duration.
Measures of onset and course, in comparison,
were quite comparable across subsamples
defined in terms of episode duration, although
respondents with episodes that persisted for
>12 months reported more years in episode
than respondents with episodes in the 1–11
months range. In addition, consistent with
previous research, we found that episode dur-
ation was largely unrelated to most measures of
impairment, co-morbidity, and sociodemo-
graphics (Bienvenu et al. 1998; Maier et al. 2000 ;
Carter et al. 2001). In cases where duration
was related to impairment, the associations were
non-monotonic. In cases where duration was
related to co-morbidity, the ORs generally in-
creased with GAD episode duration, but in
ways that were inconsistent across the range of
co-morbid disorders (i.e. the 1–5 v. o6 mo nth
for co-morbid panic disorder and PTSD com-
pared to 1–11 v. o12 months for co-morbid
dysthymia and associations increasing in
roughly linear fashion with duration for other
co-morbid mood disorders). We also found that
episode duration is not significantly related to
parental history of GAD. This finding is
indirectly consistent with evidence from twin
research that duration of GAD in one co-twin
is unrelated to risk of GAD in the other
co-twin (Kendler et al. 1994). We also found
that episode duration is not strongly related to
most sociodemographic correlates with the
notable exception of age. In the case of age, as
noted above, the possibility exists of either
age-related differential recall failure or an
association of age with progression from
shorter to longer episodes. We are incapable
of distinguishing these possibilities with the
cross-sectional NCS-R data.
These results support the view that a large
number of people exist with a clinically signifi-
cant GAD-like syndrome that is characterized
by episodes of <6 months duration (Rickels &
Rynn, 2001). In saying this, the issue of recur-
rence is one of obvious importance. We found
that cases with exclusively short episodes typi-
cally recur over a number of years and that the
average number of years with an episode is
equal to that of cases with episodes that last 6–11
months. In light of this fact, it might be useful
for future studies to consider whether episode
recurrence should play a more important part
than it currently does as a defining feature of
GAD. This might be especially useful among
cases with exclusively short episodes. It is note-
worthy in this regard that the average number of
GAD: evidence from the NCS-R 7
months in an episode in the past year among
NCS-R respondents with 12-month prevalence
whose longest lifetime episodes were in the
range 1–2 months was above the upper end of
this range (2
.
4 months). This might be inter-
preted as meaning that respondents with short
episodes sometimes have recurrences within a
single year. Another plausible possibility, of
course, is that these short episodes represent
fluctuations in severity among chronic sub-
threshold cases. Although we are unable to
distinguish these two possibilities in the NCS-R
data, an attempt to do so should be made in
future epidemiological studies.
A question can be raised whether reverting to
a 1-month duration requirement would make
sense clinically in that treatment might not be
necessary for cases with such short durations.
This concern should be less, however, for
maintenance medication of chronic-recurrent
cases. It is not clear from existing trials that
currently available treatment would be effective
in preventing recurrence of cases with short
episode durations, as maintenance therapy trials
have been carried out exclusively with cases of
longer duration (Pollack et al. 2003). Even if
subsequent empirical research shows that cur-
rently available therapies are not effective in this
way we should not, however, reify this failure by
defining the problem out of existence in the
upcoming DSM and ICD revisions. Instead,
cases with short duration should be included in
the diagnost ic system in recognition of their
clinical significance.
It is not clear whether reverting to a 1-month
GAD duration requirement would lead to the
same problem as in DSM-I II of distinguishing
GAD from situational stress reactions. The re-
quirement that the worry must be excessive and
hard to control was added to DSM-III-R for
precisely this purpose. However, excessiveness is
more difficult to assess when focusing only on
short-term reactions. In addition, duration is
one dimension by which excessive worry is as-
sessed – as indicated either by worry persisting
even after the objective life situation that led to
the worry resolves or by the extreme worry that
often occurs at the onset of an objectively worri-
some life situation not decreasing with time in
the way it would normally be expected to do.
A deeper issue that drives the concern about
distinguishing between GAD and adjustment
disorder is that a DSM-IV diagnosis of GAD
requires worry to be excessive in relation to
objective life situations. Were it not for this
requirement, the issue of distinguishing
GAD from adjustment disorder would be of
considerably less interest, as adjustment dis-
order would be merely a residual category.
However, as it is now patients who meet all
other criteria for GAD cannot be diagnosed
with GAD if they are exposed to a n ongoing
life situation that is so objectively worrisome
that most people in the situation meet the
other criteria for GAD. Adjustment disorder
is the default diagnosis unless the patient
meets criteria for some other relevant disorder
(e.g. PTSD). The 6-month GAD episode
duration requirement is thought to be import-
ant in a situation of this type to reduce con-
fusion between GAD and adjustment disorder.
It is instructive to compare the ways in which
the issues described in the last paragraph play
out in the diagnosis of GAD versus the diagnosis
of major depression. The important distinction
is that the latter diagnosis, unlike GAD, does
not hinge on whether dysphoria or anhedonia
is considered excessive. Indeed, a distinct ion is
made between endogenous and reactive cases of
major depression in order to indicate the role
played by objective life situations in episode
onset. Because of this greater diagnostic inclus-
iveness, concerns have not been raised about
increasing the 2-week major depress ion mini-
mum duration requirement to reduce confusion
with adjustment disorder. It might be useful to
consider why excessive and non-excessive cases
of dysphoria and anhedonia are both included
in the diagnosis of major depression, while only
excessive cases are included in the diagnosis of
GAD. The fact that normal grief reactions are
excluded from a diagnosis of major depression
might be taken to imply that other types of
stress reactions are considered not normal (i.e.
excessive), but such an interpretation goes well
beyond the conventional interpretation of the
diagnostic criteria for major depression. While a
thorough consideration of the rationale for this
lack of parallel between the treatment of the
excessiveness requirement for major depress ion
and GAD is outside the scope of the current
paper, it is clear from this brief comparison that
the legi timacy of the excessiveness requirement
for a diagnosis of GAD needs to be addressed in
8 R. C. Kessler et al.
future research in order to consider the full
implications of changing the current GAD
duration requirement.
ACKNOWLEDGMENTS
The National Comorbidity Survey Replication
(NCS-R) is supported by NIMH (U01-
MH60220) with supplemental support from
NIDA, the Substance Abuse and Mental Health
Services Administration (SAMHSA), the Robert
Wood Johnson Foundation (RWJF; Grant
044708), and the John W. Alden Trust. Ad-
ditional support for the preparation of this
report was provided by Pfizer, Inc. Collaborating
investigators include Ronald C. Kessler (Prin-
cipal Investigator, Harvard Medical School),
Kathleen Merikangas (Co-Principal Invest i-
gator, NIMH), James Anthony (Michigan State
University), William Eaton (The Johns Hopkins
University), Meyer Glantz (NIDA), Doreen
Koretz (Harvard University), Jane McLeod
(Indiana Universit y), Mark Olfson (Columbia
University College of Physicians and Surgeons),
Harold Pincus (University of Pittsburgh), Greg
Simon (Group Health Cooperative), Michael
Von Korff (Group Heal th Cooperative), Philip
Wang (Harvard Medical School), Kenneth
Wells (UCLA), Elaine Wethington (Cornell
University), and Hans-Ulrich Wittchen (Max
Planck Institute of Psychiatry).
A complete list of NCS publications and the
full text of all NCS-R instruments can be found
at http://www.hcp.med.harvard.edu/ncs. Send
correspondence to NCS@hcp.med.harvard.edu.
The NC S-R is carried out in conjunction with
the World Health Organization World Mental
Health (WMH) Survey Initiative. We thank the
staff of the WMH Data Collection and Data
Analysis Coordination Centres for assistance
with instrumentation, fieldwork, and consu l-
tation on data analysis. These activities were
supported by the John D. and Catherine T.
MacArthur Foundation, the Pfizer Foundation,
the US Public Health Service (R13-MH066849,
R01-MH069864, and R01 DA016558), the Pan
American Health Organization, Eli Lilly and
Company, and GlaxoSmithKline. A complete
list of WMH publications and instruments can
be found at http://www.hcp.med.harvard.edu/
wmhcidi.
The authors ap preciate the helpful comments
of Kathleen Merikangas and Bedirhan Ustun
on earlier drafts.
DECLARATION OF INTEREST
None.
NOTE
Supplementary information accompanies this
paper on the Journal’s website (http ://journals.
cambridge.org).
REFERENCES
APA (1980). Diagnostic and Statistical Manual of Mental Disorders
(3rd edn). American Psych iatric Association: Washington, DC.
APA (1987). Diagnostic and Statistical Manual of Mental Disorders
(3rd edn revised). American Psychiatric Association: Washington,
DC.
APA (1994). Diagnostic and Statistical Manual of Mental Disorders
(4th edn). American Psychiatric Association : Washington, DC.
Andreasen, N. C., Endicott, J., Spitzer, R. L. & Winokur, G. (1977).
The family history method using diagnostic criteria : reliability and
validity. Archives of General Psychiatry 34, 1229–1235.
Barlow, D. H., Blanchard, E. B., Vermilyea, J. A., Vermilyea, B. B.
& DiNardo, P. A. (1986). Generalized anxiety and generalized
anxiety disorder: description and reconceptualization. American
Journal of Psychiatry 143, 40–44.
Barlow, D. H. & Wincze, J. (1998). DSM-IV and beyond : what is
generalized anxiety disorder? Acta Psychiatrica Scandinavica
(Suppl.) 393, 23–29.
Bienvenu, O. J., Nestadt, G. & Eaton, W. W. (1998). Characteriz ing
generalized anxiety: temporal and symptomatic thresholds.
Journal of Nervous and Mental Disease 186, 51–56.
Breier, A., Charney, D. S. & Heninger, G. R. (1985). The diagnostic
validity of anxiety disorders and their relationship to d epressive
illness. American Journal of Psychiatry 142, 787–797.
Breslau, N. (1985). Depressive symptoms, major depression and
generalized anxiety: a comparison of self-reports on CES-D and
results from diagnostic interviews. Psychiatry Research 15, 219–
229.
Breslau, N. & Davis, G. C. (1985 a). DSM-III generalized anxiety
disorder: an empirical investigation of more stringent criteria.
Psychiatry Research 15, 231–238.
Breslau, N. & Davis, G. C. (1985 b). Further evidence on the doubtful
validity of generlized anxiety disorder. Psychiatry Research 16,
177–179.
Carter, R. M., Wittchen, H. U., Pfister, H. & Kessler, R. C. (2001).
One-year prevalence of subthreshold and threshold DSM-IV gen-
eralized anxiety disorder in a nationally representative sample.
Depression Anxiety 13, 78–88.
Clayton, P. J., Grove, W. M., Coryell, W., Keller, M., Hirschfeld, R.
& Fawcett, J. (1991). Follow-up and family study of anxious
depression. American Journal of Psychiatry 148, 1512–1517.
First, M. B., Spitzer, R. L., Gibbon, M. & Williams, J. B. W.
(2002). Structured Clinical Interview for DSM-IV-TR Axis I
Disorders, Research Versio n, Non-Patient Edition (SCID-I/NP ).
Biometrics Research, New York State Psychiatric Institute : New
York.
Halli, S. S., Rao, K. V. & Halli, S. S. (1992). Advanced Techniques of
Population Analysis. Plenum : New York.
Kendler, K. S., Davis, C. G. & Kessler, R. C. (1997). The familial ag-
gregation of common psychiatric and substance use disorders in
GAD: evidence from the NCS-R 9
the National Comorbidity Survey : a family history study. British
Journal of Psychiatry 170 , 541–548.
Kendler, K. S., Neale, M. C., Kessler, R. C., Heath, A. C. & Eaves,
L. J. (1994). Clinical characteristics of familial generalized anxiety
disorder. Anxiety 1, 186–191.
Kessler, R. C., Berglund, P., Chiu, W. T., Demler, O., Heeringa, S.,
Hiripi, E., Jin, R., Pennell, B. E., Walters, E. E., Zaslavsky, A. &
Zheng, H. (2004 a). The US National Comorbidity Survey
Replication (NCS-R) : design and field procedures. International
Journal of Methods in Psychiatric Research 13, 69–92.
Kessler, R. C., Berglund, P., Demler, O., Jin, R., Koretz, D.,
Merikangas, K. R., Rush, A. J., Walters, E. E. & Wang, P. S.
(2003). The epidemiology of major depressive disorder: results
from the National Comorbidity Survey Replication (NCS-R).
Journal of the American Medical Association 289, 3095–3105.
Kessler, R. C., Berglund, P. A., Demler, O., Jin, R. & Walters, E. E.
(in press). Lifetime prevalence and age-of-onset distributions of
DSM-IV disorders in the National Comorbidity Survey
Replication (NCS-R). Archives of General Psychiatry.
Kessler, R. C., Keller, M. B. & Wittchen, H.-U. (2001). The
epidemiology of generalized anxiety disorder. Psychiatric Clinics
of North America 24, 19–40.
Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes,
M., Eshleman, S., Wittchen, H.-U. & Kendler, K. S. (1994).
Lifetime and 12-month prevalence of DSM-III-R psychiatric
disorders in the United States: results from the Natio nal Co-
morbidity Survey. Archives of General Psychiatry 51, 8–19.
Kessler, R. C. & Merikangas, K. R. (2004). The National Co-
morbidity Survey Replication (NCS-R) : background and aims.
International Journal of Methods in Psychiatric Research 13, 60–68.
Kessler, R. C. & Ustun, T. B. (2004). The World Mental Health
(WMH) Survey Initiative Version of the World Health
Organization (WHO) Composite International Diagnostic
Interview (CIDI). International Journal of Methods in Psychiatric
Research 13, 93–121.
Kessler, R. C., Walters, E. E. & Wittchen, H.-U. (2004b).
Epidemiology. In Generalized Anxiety Disorder: Advances in
Research and Practice (ed. R. G. Heimberg, C. L. Turk and D. S.
Mennin), pp. 29–50. Guilford Press: New York.
Leon, A. C., Olfson, M., Portera, L., Farber, L. & Sheehan, D. V.
(1997). Assessing psychiatric impairment in primary care with the
Sheehan Disability Scale. International Journal of Psychiatry in
Medicine 27, 93–105.
Maier, W., Gansicke, M., Freyberger, H. J., Linz, M., Heun, R. &
Lecrubier, Y. (2000). Generalized anxiety disorder (ICD-10) in
primary care from a cross-cultural perspective: a valid diagnostic
entity? Acta Psychiatrica Scandinavica 101, 29–36.
Noyes Jr., R., Woodman, C., Garvey, M. J., Cook, B. L., Suelzer, M.,
Clancy, J. & Anderson, D. J. (1992). Generalized anxiety disorder
vs. panic disorder. Distinguishing characteristics and patterns
of comorbidity. Journal of Nervous and Mental Disease 180,
369–379.
Pollack, M. H., Meoni, P., Otto, M. W., Simon, N. & Hackett, D.
(2003). Predictors of outcome following venlafaxine extended-
release treatment of DSM-IV generalized anxiety disorder: a
pooled ana lysis of short- and long-ter m studies. Journal of Clinical
Psychopharmacology 23, 250–259.
Research Triangle Institute (2002). SUDAAN : Professional Software
for Survey Data Analysis [computer program], (Version 8.01).
Research Triangle Park, NC.
Rickels, K. & Rynn, M. (2001). Overview and clinical presentation of
generalized anxiety disorder. Psychiatric Clinics of North America
24, 1–17.
Wittchen, H.-U., Zhao, S., Kessler, R. C. & Eaton, W. W. (1994).
DSM-III-R generalized anxiety disorder in the National
Comorbidity Survey. Archives of General Psychiatry 51, 355–
364.
Wolter, K. M. (1985). Introduction to Variance Estimation. Springer-
Verlag: New York.
WHO (1992). The ICD-10 Classification of Mental and Behavioral
Disorders: Clinical Descriptions and Diagnostic Guidelines. WHO :
Geneva.
WHO (1993). The ICD-10 Classification of Mental and Behavioral
Disorders: Diagnostic Criteria for Research. WHO: Geneva.
10 R. C. Kessler et al.