Agoraphobia: a review of the diagnostic classificatory position ...

DEPRESSION AND ANXIETY 27 : 113–133 (2010)
Review
AGORAPHOBIA: A REVIEW OF THE DIAGNOSTIC
CLASSIFICATORY POSITION AND CRITERIA
Hans-Ulrich Wittchen, Ph.D.,1Ã Andrew T. Gloster, Ph.D.,1 Katja Beesdo-Baum, Ph.D.,1 Giovanni A. Fava, M.D.,2
and Michelle G. Craske, Ph.D.3
The status of agoraphobia (AG) as an independent diagnostic category is reviewed
and preliminary options and recommendations for the fifth edition of The Diagnostic
and Statistical Manual (DSM-V) are presented. The review concentrates on
epidemiology, psychopathology, neurobiology, vulnerability and risk factors, clinical
course and outcome, and correlates and consequences of AG since 1990. Differences
and similarities across conventions and criteria of DSM and ICD-10 are considered.
Three core questions are addressed. First, what is the evidence for AG as a diagnosis
independent of panic disorder? Second, should AG be conceptualized as a subordinate
form of panic disorder (PD) as currently stipulated in DSM-IV-TR? Third, is there
evidence for modifying or changing the current diagnostic criteria? We come to the
conclusion that AG should be conceptualized as an independent disorder with more
specific criteria rather than a subordinate, residual form of PD as currently stipulated
in DSM-IV-TR. Among other issues, this conclusion was based on psychometric
evaluations of the construct, epidemiological investigations which show that AG can
exist independently of panic disorder, and the impact of agoraphobic avoidance upon
clinical course and outcome. However, evidence from basic and clinic validation
studies remains incomplete and partly contradictory. The apparent advantages of a
more straightforward, simpler classification without implicit hierarchies and
insufficiently supported differential diagnostic considerations, plus the option for
improved further research, led to favoring the separate diagnostic criteria for AG as a
diagnosis independent of panic disorder. Depression and Anxiety 27:113–133,
2010.
r 2010 Wiley-Liss, Inc.
Key words: agoraphobia; panic disorder; classification; diagnostic criteria;
DSM-V
1Institute of Clinical Psychology and Psychotherapy, Technische
Universitaet Dresden, Dresden, Germany
INTRODUCTION
2Clinical Psychology, University of Bologna, Bologna, Italy
T
3
his review focuses on key critical issues pertaining
Department of Psychology, University of California, Los Angeles,
California
to the definition of agoraphobia (AG) and its relation-
ship to panic disorder (PD), and its future diagnostic
This article is being co-published by Depression and Anxiety and
classificatory conceptualization in the fifth edition of
the American Psychiatric Association.
The Diagnostic and Statistical Manual (DSM-V). Starting
The authors report they have no financial relationships within the
with an historical account of various conceptualizations
past 3 years to disclose.
of AG and associated diagnostic criteria, we address
ÃCorrespondence to: Hans-Ulrich Wittchen, Department of
several core issues that are and have been controversial
Psychology, Institute of Clinical Psychology and Psychotherapy,
since DSM-III-R.[1] It should be noted that this
Technische Universita¨t Dresden, Chemnitzer StraXe 46, D-01187
discussion also occurs in the context of several other
Dresden, Germany. E-mail: wittchen@psychologie.tu-dresden.de
relevant reviews, such as the reviews on panic attack
Received for publication 13 October 2009; Accepted 30 October
(PA) and PD[2] as well as Specific Phobia,[3] which also
2009
includes a review on the relationship between AG and
DOI 10.1002/da.20646
Specific Phobia. In light of these other reviews, we
Published online in Wiley InterScience (www.interscience.wiley.
only peripherally address these disorders and their
com).

---

114
Wittchen et al.
implications for the core issues. This article was
to vulnerability and risk factors for AG to examine
commissioned by the DSM-V Anxiety, Obsessive–
evidence for AG being a separate construct using classical
Compulsive Spectrum, and Posttraumatic and Disso-
validators. In the final section, we use this evidence to
ciative Disorders Work Group. Recommendations
address the question of whether AG should be con-
provided in this article should be considered prelimin-
ceptualized as a subordinate form of PD, as currently
ary at this time; they do not necessarily reflect the final
stipulated in DSM-IV-TR, or whether it should be
recommendations or decisions that will be made for
defined as an independent disorder.
DSM-V, as the DSM-V development process is still
It should be noted upfront that changes in the DSM
ongoing. It is possible that this article’s recommenda-
nomenclature and/or the use of ICD criteria have
tions will be revised as additional data and input from
partially resulted in corresponding changes in diag-
experts and the field are obtained. It should also be
nostic assessment instruments used for research stu-
noted up front that despite consensus in the Task Force
dies. Such changes and the use of different diagnostic
there was no unanimous consensus among all advisors
criteria complicate considerably the interpretation of
to the Task Force with regard to the conclusions
research findings and the conduct of this systematic
summarized in this document.
review of AG. At the very least, results from different
studies must be interpreted in light of whether DSM or
ICD diagnostic rules were utilized. At worst, results
STATEMENT AND SIGNIFICANCE
from different studies and across different versions of
OF THE ISSUES
the manual are rendered incomparable. Given this
limitation and considering the length of manuscript,
Three core issues are addressed, most of which have
this paper could not fully appreciate in detail several,
been discussed controversially in previous revisions of the
undoubtedly important, clinical distinctions surrounding
DSM revision processes.[1] First, what is the evidence that
the expression of agoraphobia and its measurement.
AG exists independently from PD in terms of its
epidemiology and the classical validators for diagnoses?
Second, should AG be conceptualized as a subordinate
HISTORICAL PERSPECTIVE OF
form of PD as currently stipulated in DSM-IV-TR or is it
AGORAPHOBIA IN REVISIONS OF
better defined as an independent disorder, as in the
THE ICD AND DSM
International Classification of Diseases 10th revision
(ICD-10)? Finally, we also comment on how changes in
The first account of AG is credited[4] to Westphal’s
the diagnostic criteria for AG may improve precision,
classical description (1871) of the syndrome that,
reliability, and ease of administration. These questions are
during the major part of the last century, served as a
considered to be particularly relevant from the perspective
model paradigm for anxiety disorders in general.[5]
of recognition, treatment, and management, because there
From that time until the introduction of DSM-III-R,[6]
have been concerns that the residual status of AG and its
AG was frequently described and reported in the
definition might be associated with underdiagnosis and
literature as a common and distressing phobic disorder,
undertreatment. There have also been concerns that the
or as phobic neurosis in the older literature.[4,7,8]
description of AG and the relationship of AG to PA and
Internationally, AG was introduced and codified as an
PD are unnecessarily complex, making its use complicated
independent diagnostic entity and distinctive syndrome
and possibly reducing its clinical utility. Additionally,
of multiple fears in the 1970s (ICD-9),[9] where it
doubts have been expressed on whether the assumed
retains this status even today (ICD-10).[10] In the
causal role of PA and panic-like features is a valid
United States and the DSM system, AG also appeared
assumption in agreement with empirical evidence and
as the result of subdividing phobic neurosis and anxiety
whether it is needed at all. Finally, the position and
neurosis (DSM-III).[11] The DSM-III definition
definition of AG is one of the rare examples where DSM-
‘‘marked fear and avoidance of being alone or in public
IV-TR and the ICD-10 deviate from each other, leading
places from which escape might be difficult or help not
potentially to continued discrepancy in how these patients
available in case of sudden incapacitations’’ did not
are being diagnosed worldwide.
differ much from that of other phobic disorders in
The review consists of three major sections. The first
DSM-III and the ICD definition of AG. The text
section addresses the question of whether recent studies
description in DSM-III, however, already stipulated in
provide evidence that AG exists independently from PD
1980 that the diagnosis of AG is more closely linked to
and PAs. Using available evidence and selected core
PAs than to phobias. ‘‘Agoraphobia with Panic Attacks’’
publications before 1990, core questions focused on: (1)
should be coded if ‘‘the initial phase of the disorder
How frequently does PD occur with and without AG? (2)
consisted of recurrent panic attacks,’ thus leading the
How frequently does AG occur without a history of PA?
individual to develop anticipatory fear of having such
(3) What are the characteristics of PD with and without
an attack and to avoid situations associated with these
AG as well as AG without PA with regard to gender, age
attacks. Only when there was no history of PA (or the
of onset, impairment, and associated factors? In the
information was lacking), the diagnosis of ‘ Agoraphobia
second section, we review available evidence with regard
without Panic Attacks’’ was made.
Depression and Anxiety

---

Review: Agoraphobia Review
115
Starting with DSM-III-R[6]—and unlike the ICD-9—
regarding the diagnostic criteria of PA or PD or the
AG was specifically defined as a classically conditioned
fact that both are frequently comorbid with AG. Further,
response to situations in which PAs had occurred. In
both positions agree that, by and large consistent with
response to increasing experimental and clinical
the classical literature, that PA and panic-like features
evidence,[12–17] AG in DSM-III-R was not only con-
may play a core role in the development of some but
ceptually attached to PA and PD, but was also seen
not all AG patients.[7,22] The disagreement mainly
explicitly and exclusively as a temporally secondary
surrounds whether (a) AG exists at all independent of
complication. In fact, with each consecutive DSM
initial primary PAs or panic-like features,[23–33] (b)
revision, the residual status of AG within the construct
whether PA or panic-like features are invariably
of PA and PD has been increasingly more pronounced.
causally linked to AG, or (c) whether there is any
DSM-IV-TR[18] acknowledges the relevance of AG as a
clinical utility in diagnosing AG as a separate
syndrome by describing criteria that state upfront that
disorder,[34–37] and (d) whether at all and if yes how
AG (in itself) ‘‘is not a codable disorder.’’ Instead,
to define and specify the explicit criteria for AG in a
allowances are made and codes provided for those
better way.[37–39] Further, some uneasiness has been
disorders in which AG occurs, namely PD with AG or
expressed as to whether the DSM-IV-TR conceptua-
AG without the History of Panic Disorder (AG w/o PD).
lization of AG as a residual of PD (AG without the
The latter, however, restricts the diagnosis to those
history of PD) is in line with the overarching principles
patients where AG is related to fear of developing panic-
of DSM-III and its successors to be atheoretical and
like symptoms (e.g., dizziness or diarrhea). The DSM-
descriptive unless persuasive aetiopathogenic mechan-
IV-TR formulation results in a fairly complex (12 pages)
isms have been established. Although such persuasive
differential diagnostic scheme with lengthy differential
evidence is lacking, DSM-IV-TR’s explicit diagnostic
diagnostic considerations. Unlike all other disorders, two
hierarchy rules assume such an etiological role of PA
core psychological syndromes, namely PAs (that may
and PD for AG, resulting in a near unconditional
occur in the context of many mental disorders) and AG
priority of PD over the diagnosis of AG.
(only relevant for PD and AG w/o PD), are described
To summarize, since DSM-III-R, AG is diagnosed
first before the specific diagnoses are described.
only within the context of PD (PD with AG) or as the
This procedure has deepened discrepancy between
result of PA or panic-like features (AG without a
DSM and ICD. The ICD-9 and the ICD-10 do not
history of PD). This convention is in contrast to the
assume that AG necessarily arises from PAs, and thus
ICD-10, where AG in fact takes precedence over PAs,
have retained AG as a separate disorder, independent of
and prior diagnostic conventions in DSM-III, where
PD. Furthermore, in the DSM-IV-TR, the definition
AG was defined similar to phobias.
of AG is substantially different from the definition used
Another unresolved question is how to diagnose
for other phobic disorders in that the diagnostic criteria
patients who fail to report panic-like symptoms as
for AG are tied not only to the concept of PA or PD
required by DSM-IV-TR. Assuming that such cases
but also to ‘‘panic-like symptoms’’ (AG w/o PD). This
might exist, should such cases be diagnosed as anxiety
conceptual development in DSM was based mainly on
disorder NOS or as a specific phobia in DSM? This
the observation in some studies[15,19,20] that among
issue has been identified as particularly critical by
clinical samples in research settings in the United
developers of diagnostic interviews and epidemiologi-
States, that use DSM-III-R criteria, AG patients
cal research.[26,40] The current DSM-IV-TR provides
without PAs or panic-like features appear to be
little guidance in this respect, which constitutes a
extremely rare—an observation that continues to
potential problem. Unlike phobic disorders, where
date.[20] Fava et al.[21] attributed this largely to the fact
separate criteria are specified, the current ‘ broad’’ AG
that the diagnostic criteria and assessment instruments
definition lacks specification for what constitutes
are by now so ‘ biased’’ toward the temporally primary
agoraphobic situations and cues beyond the occurrence
causal role of PAs or panic-like features that it is
or fear of panic-like symptoms. Further, unlike phobic
impossible to diagnose AG outside the context of
disorders, AG criteria do not specify the additional
primary PAs or panic-like features. Thus, to date, the
mandatory criteria for phobic disorders, such as the
implicit hierarchical criteria in DSM-IV have rendered
‘‘exposure’ criterion (B), the criterion that the person
it impossible to collect systematic data to resolve this
recognizes his fear as excessive or unreasonable (C), or
issue, if one sticks strictly to the current DSM-IV-TR
the distress and impairment criterion (E). As stipulated
definition. The lack of resolution has led to the
in the ICD-10 criteria for research,[10] AG should be
unfortunate situation of two discrepant diagnostic
diagnosed when the AG syndrome occurs in at least
criteria sets worldwide, and use of different criteria in
two out of a total of four prototypical situations to
different diagnostic interviews. This makes direct
qualify for what is called in both systems as ‘ a
comparisons of findings in this domain difficult and
characteristic cluster.’’ In response to this, most recent
in some areas impossible.
diagnostic interview versions assign the diagnoses
It is important to note that in contrast to the
whenever two or more situations are endorsed. Because
definition of AG and the assumed causal role of PA or
DSM-IV-TR lacks such a precise definition, numerous
panic-like features, there is no to little controversy
and mostly older DSM studies assigned the diagnoses
Depression and Anxiety

---

116
Wittchen et al.
even when only one situation is endorsed and without
addressed issues of epidemiology, therapy, experimen-
additional mandatory criteria used for all phobic
tal procedures, neurobiology, psychopathology, diag-
disorders (Social Phobia or Specific Phobias).
nosis, or assessment. The DSM-IV Source Book[1] and
Beyond these core issues, revisions of DSM-III have
DSM-IV Options Book[42] were also reviewed.
added some modifications of the diagnostic specifica-
tion of AG[37] that were critically reexamined. In the
DATA COLLECTION AND ANALYSIS
DSM-III-R,[6] an AG diagnosis was assigned when the
person was anxious about having a PA that s/he
The method section of each study was examined first
restricts travel (avoidance), needs a companion to
with respect to its assessment/diagnosis of AG. Studies
travel (use of companions), or endures AG situations
that examined AG independent of PA and PD were
despite intense anxiety (distress). These criteria have
marked and selected for priority review. Given the
been criticized as lacking proof of incremental validity
paucity of such studies, quantitative analysis of results
and being overinclusive[39] because they do not restrict
was rendered impossible. As such, articles that utilized
AG to exhibiting avoidance behavior. That is, patients
hierarchical DSM-IV rules were also considered.
might be able to travel extensively even though they
Studies were further grouped according to the follow-
need a companion to do so, or they might even be able
ing thematic areas: prevalence and incidence, develop-
to travel alone while experiencing significant distress.
mental issues and age of onset, natural course and
Further, DSM-III-R delineated situational avoidance
outcome, patterns of co-morbidity and transition
as the central issue to be considered by classifying AG
between disorders, vulnerability and (causal) risk
across four different levels (none, mild, moderate,
factors, including genetic and family genetic factors,
severe). In the DSM-IV,[18,41] this is no longer the case
temperamental antecedents, cognitive and emotional
with situational avoidance on equal footing with
processing abnormalities, neural substrates and shared
distress and use of companions in making a dichot-
biomarkers, and treatment/treatment response.
omous AG diagnosis (absent, present). This has been
criticized by Schmidt and Cromer[37] as potentially
DESCRIPTION OF STUDIES/
decreasing clinical utility as well as predictive value,
METHODOLOGICAL QUALITY OF
because situational avoidance has been de-emphasized
INCLUDED STUDIES
in making AG diagnosis. Further, the reduction of
All search terms and databases revealed a total of
specification of AG from a dimensional AG 4-point
2,112 citations, 469 of which were judged relevant for
rating to a dichotomous (present/absent) approach
this review. Of these, less than 5% assessed AG
suggests that a dichotomous method of organizing
independent from present DSM hierarchical rules.
agoraphobic behaviors is superior to the more finely
Findings from studies tied to the current DSM cannot
graded assessment of phobic avoidance. However, no
inform about the independence of the diagnosis or its
data have been presented to support this. Schmidt and
relation to specific phobia or PD. The overall
Cromer[37] assumed that this decision might have
frequency of relevant articles per thematic topic was
reduced the clinical utility of an AG diagnosis.
as follows: Conceptualization (n 5 21), Relationship
Given these controversies and unresolved issues, this
between AG, PA, PD, and Specific Phobia (n 5 66),
article was commissioned by the DSM-V Anxiety,
Co-morbidity (n 5 47), Transition between disorders
Obsessive–Compulsive Spectrum, Posttraumatic, and
(n 5 21), Temperament Antecedents (n 5 19), Cogni-
Dissociative Disorders Work Group to review this
tive/Emotional Processing Abnormalities (n 5 35),
critical issue again, to explore whether new data allow a
Vulnerabilities and Risk Factors (Environmental and
resolution of these issues.
Genetic) (n 5 57), Familiality (n 5 26), Developmental
Patterns (n 5 22), Assessment Complications (n 5 27),
METHODS OF LITERATURE
Neural
Substrates
(n 5 19),
Shared
Biomarkers
REVIEW
(n 5 20), Symptom Similarity (n 5 9), and Treatment
Response (n 5 80).
SEARCH STRATEGY AND SELECTION
CRITERIA
RESULTS OF REVIEW
We searched the ‘ Web of Science,’’ ‘‘PubMed,’’ and
‘‘PsychINFO’’ databases in April 2008 for peer-
EPIDEMIOLOGICAL AND CLINICAL
reviewed articles on AG, and updated the review in
EVIDENCE FOR AG
March 2009. Reference lists of retrieved articles were
Prevalence. Since 1980s, a wide range of commu-
searched for additional studies. Only articles in English
nity studies in the United States,[43–46] Europe,[47] and
and German were considered, and over 30 search terms
the rest of world[48–52] have examined the lifetime and
were used (full list available on request). Articles on AG
12-month prevalence of PAs, PD, and AG, according
printed after 1990 were targeted, but compelling
to the criteria of DSM-III, III-R, and IV. Despite some
articles published before 1990 were also included.
variation in lifetime prevalence estimates, they almost
Articles were selected if they were peer-reviewed and
all converge on the following: (a) PAs according to
Depression and Anxiety

---

Review: Agoraphobia Review
117
DSM-III-R and IV are very frequent manifestations
the community that are approximately as frequent as
with estimates of 20% or above for most samples, (b)
those that occur concurrently with PD.
lifetime rates of PD are in the range of 3–5%;
According to US studies, the situation in clinical
depending on the study, approximately 35–65% of
settings appears to differ. As reviewed by Ballenger
subjects with PD meet criteria for PD with AG,
and Fyer[1] and Barlow,[19,20] at least the diagnosis of
whereas the remaining report PD without AG, and (c)
‘‘AG without the history of PD’’ is rarely assigned in
rates for AG without history of PD are typically found
clinical practice. Faravelli et al.[54] reviewed eight
to be at least as high or even higher as those for PD.
clinical studies, seven with low sample sizes, citing
However, they also reveal a large degree of variation of
four studies with not a single case of AG without
prevalence across studies, ranging from about 1% to a
panic and four studies reporting 2–31% of PA among
high of 22% (median for older and more recent
AG patients.
studies 5 7.8%;
mediation
for
studies
since
Temporal relationship of PA, PD, and AG. Some
1990 5 3.8%), probably due to methodological factors.
studies used longitudinal data to specifically examine
AG without PD have been reported to occur in both
the question of whether AG is always related to pri-
children[53] and in the elderly.[51]
mary spontaneous PAs or at least panic-like symp-
A recent review by Faravelli et al.[54] summarized five
toms.[21,26,28,33,40,63] Consistent with other evidence
epidemiological studies that convergently found that
from similar inquiries in epidemiological[44] and clinical
46–85% of all individuals with AG do not have PAs. He
samples,[32,64,65] these explorations found no consistent
also showed, however, that in clinical samples (eight
evidence to support these assumptions. In community
studies mostly with a few dozen patients), AG without
samples, the majority of agoraphobics never experienced
PA is considerably less frequent (0–31%), with four
any PA or panic-like symptoms or psychophysiological
studies finding not a single case. In studies published
symptoms of other type that clearly preceded the onset
since 1990,[45,55,56] which used a stricter definition of
of agoraphobic avoidance. If they reported such
AG requiring at least two agoraphobic situations
symptoms, they were frequently secondary to AG onset,
consistent with ICDs and DSMs stipulation of a typical
and longitudinal evidence suggests that AG just as
‘‘cluster’’ of situations, the 12-month rates for PD with
frequently precedes PA as PA precedes AG.
and without AG was found to be 1.8% (interquartile
Mostly smaller (less than 70 patients) clinical studies
range [IR]: 0.7–2.2) and 1.3% (IR: 0.7–2.0) for AG
in the United States[17,66,67] find that in the vast
without PAs. The variation between studies has been
majority of patients with AG/PD or AG with PA,
attributed to design, methodological, and assessment
agoraphobic avoidance clearly occurs temporally sec-
variation rather than reflecting true differences; for
ondary to the PA. Ballenger and Fyer[1] concluded after
example, as a function of cultural or regional
examination whether primary PAs also affect the
effects.[47,55] An important example of such a metho-
natural course of secondary AG that, although agor-
dological factor is the different conventions used for
aphobic avoidance does not seem to be related to
the DSM-IV ‘ A’ criterion. In contrast to studies in the
variables, such as type of PA frequency or severity of
1980s with the Diagnostic Interview Schedule (DIS),
attacks, it does seem to be related to the expectation of
where even one single AG situation may qualify for a
panicking in specific AG situations. However, these
diagnosis, most recent studies using the Composite
findings do not imply that ‘‘panic expectancy is of
International Diagnostic Interview (CIDI) required
causal significance for the development of agoraphobia
two or more situations to meet criteria for DSM-IV-TR.
avoidance’’ (Ballenger and Fyer,[1] p 455).
This change led to decreases in rates of AG by
Other lines of evidence for inconsistent findings
approximately 50%. It also affected the rates for PD
come from clinical retrospective studies that used
with AG, revealing now—in contrast to older studies—
sensitive methods to detect subclinical symptomatol-
more cases of PD without AG than PD with
ogy prior to the onset of PAs. Fava et al.[64] found that
AG.[26,33,56]
the majority of 40 patients with PD with AG
To summarize, consistent with a previous review of
experienced prodromal symptoms (AG, hypochondria-
this issue by Ballenger and Fyer,[1] the prevalence of
sis, generalized anxiety) before the first PA. The
AG without a history of PD and even the prevalence of
findings were obtained with considerable methodolo-
AG even without the presence of PA and panic-like
gical precautions: careful dating of symptom onset,
symptoms (as is possible in the CIDI/ICD-10) was at
rigorous symptom definition by a reliable and validated
least as high as the combined rates of PD with and
probe suitable for prodromal and subclinical symp-
without
AG
across
all
epidemiological
stu-
toms, and delay of the interview until the acute
dies.[33,45,56–61] Thus, even when conservative defini-
disturbance has passed (to minimize distortions of
tions are utilized (i.e., strict reliance on DSM-IV-TR
recall). These methodological features may account for
criteria), as in the most recent NCS-R findings,[45,62]
the striking differences from studies performed in the
roughly 25% of all subjects with panic/AG syndromes
mid-eighties, which relied on self-rating instruments,
met AG criteria without qualifying for either PAs or
unstructured interviews, and diagnostic instead of
PD. Therefore, persuasive evidence has documented
symptomatic focus. Not surprisingly, the findings were
substantial rates of AG without PD and without PA in
confirmed by subsequent studies. Garvey et al.[68]
Depression and Anxiety

---

118
Wittchen et al.
found that 28% of 32 PD patients had prodromal
features (as defined by ‘ fearful spells’’) were not at
symptoms of anxiety that lasted a median of 5 years
increased risk for subsequent PA or PD; only 11.6% of
before the occurrence of the first PA. Lelliott et al.[32]
all primary AG developed a subsequent PA and only 2.4%
reported that 70% of 57 patients with PD with AG had
developed PD.
prodromal depression, anxiety, or avoidance. Agora-
Critical methodological aspects. The failure of
phobic avoidance preceded the first PA in 23% of
epidemiological studies to support the assumption that
patients. Argyle and Roth[69]considered the sequence of
PAs and panic-like symptoms almost always play a core
the events in 56 cases of PD associated with AG. They
pathogenic role for the onset of AG has been repeatedly
found that the majority of patients (55.4%) had their
challenged on methodological grounds. Against the
onset in the same 6-month period as the onset of PD,
background of observations that AG without PA/PD is
with 19.6% of patients for whom an AG clearly
rarely seen in clinical samples,[15,19,20] methodological
preceded panic and 25% of patients with the reverse
concerns were raised that current diagnostic instruments
sequence of events. Other diagnoses, especially social
might not be able to assess panic features with sufficient
phobia and generalized anxiety disorder, however,
accuracy and validity, or more generally, that the
frequently predated the onset of panic. In the same
diagnostic interviews were not diagnostically valid.
vein, long-standing hypochondriasis was found to
Horwath et al.[25] and Goisman et al.[15] conducted
precede the onset of PAs not associated with AG.
careful clinical reappraisals that, however, were based
Perugi et al.[70] studied 126 consecutive cases of PD by
only on a few dozen patients. In response, Wittchen
means of semi-structured interviews and substantially
et al.[26] conducted a careful clinical reappraisal of all AG
replicated the findings of Fava et al.[64] Further, in
cases in the Early Developmental Stages of Psycho-
several studies,[32,70,71] the large majority of patients
pathology (EDSP) data set. The clinical appraisal of each
experienced their first PAs in public places and
case was conducted by independent clinicians who were
phobogenic situations.
blind toward the interview diagnosis. The outcome of
This considerable degree of discrepancy in findings has
this appraisal (based on clinical questions like ‘‘What are
not yet been resolved because of methodological problems
you afraid of?’ ) failed to raise any doubts that these cases
inherent in studies. But it could be at least summarized;
were AG cases, without indications of prior PAs or
there is fairly consistent evidence that PA precedes AG in
panic-like symptoms whenever the diagnosis was based
up to 50% of all individuals with AG, providing some
on at least two out of a total of five (plus ‘ other’’)
support for the assumed aetiopathogenic pathway implied
agoraphobic situations were met. The reappraisal also
in DSM-IV-TR. However, the fact that up to 50% of AG
indicated that cases meeting all criteria, but only
does not reveal indications for this pathway is suggestive
reporting one of the prototypical situations, were often
of the existence of other pathways. It should also be noted
better reclassified as specific phobia, mostly of situational
that these data are not fully consistent with the
type. Similar evidence was also provided by Faravelli
panic–agoraphobia spectrum concept[72,73] that assumes
et al.[75] and Fava et al.[76] The publication of the studies
a reciprocal relationship, meaning that either condition
in the 1990s led the CIDI criteria to be changed. They
raises the probability of the other one. The epidemiolo-
now require at least two situations to be reported before
gical evidence does not support this reciprocal relation-
assigning a diagnosis of AG; cases below this threshold
ship. A major limitation of these studies is that the
are classified as phobia NOS. It should be noted that this
majority of them rely heavily on retrospective reports
change in the diagnostic interview reduced substantially
about the onset of each condition. Further, subjects were
the rates of AG in general, as well as the rates of PD with
studied sometimes decades after onset, making retro-
AG. This algorithmic change is also responsible for the
spective appraisals highly problematic. This problem
decline in rates of AG in more recent studies since the
becomes easily apparent when examining age of onset
1990s. It should be noted though that despite the
reports for PAs and AG onset, reviewed below in a
considerable sophistication in such methodological
separate paragraph.
appraisals, some task force advisors still believe that
Prospective longitudinal investigations in well-defined
methodologically sound, and full appreciation of all
age groups might resolve such issues, but are rare. In fact,
critical concerns is still lacking.
we found only one recent prospective longitudinal multi-
How frequent is AG without panic attacks or
wave study that systematically described incidence char-
PD? Based on these revised, more stringent AG criteria
acteristics of PA, PD, and AG.[33] This prospective study
and a careful prospective longitudinal investigation about
found that 23.5% with an initial PA subsequently
the relationship of PAs and PD with AG, Wittchen et al.[33]
developed AG, as did approximately 50% developed
conducted a comprehensive analysis of the incidence
PD. This constitutes a 17-fold risk increase for developing
patterns and temporal relationship of mutually exclusive
AG and a 38-fold risk for PD. It should be noted, though
classes of various liberal definitions of panic-like features
that PAs were highly associated with all types of disorders,
(labeled fearful spells) in a sample of 3,021 subjects.
including other anxiety, mood, and substance disorders,
This longitudinal characterization (see Fig. 1)
and thus are not very specific for either AG or PD.[74] It is
revealed that the cumulative lifetime prevalence
also noteworthy that cases with temporally primary AG
of AG without any indications of even the most liberally
(without PA according to DSM-IV-TR or even panic-like
defined fearful spells is 1.5%. This rate is approximately
Depression and Anxiety

---

Review: Agoraphobia Review
119
Figure 1. Modified from Wittchen et al., 2008.[33] The prevalence and mutually exclusive combinations of fearful spells (FS), DSM-IV-
TR panic attacks (PA), DSM-IV-TR panic disorder (PD), and Agoraphobia (AG) as assessed with the M-CIDI in the community.
Numbers indicate the number of cases and the weighted prevalence estimate. ÃThe definition of agoraphobia is only partially (namely
for subsets of those with PD with AG, as well as AG w/o PD (but with PA)) consistent with the DSM-IV-TR criteria of either PD with
AG or ‘‘Agoraphobia without history of PD.’’ AG/FS cases meet all AG criteria, but despite fearful spells no ‘‘panic-like symptoms.’’ AG
w/o FS report not even fearful spells. The latter two groups would not be diagnosed as AG according to DSM-IV-TR. ÃÃSubjects with
fearful spells (FS) have acknowledged the CIDI question: ‘‘Have you ever had an attack of fear or panic, when all of a sudden you felt
frightened, anxious or uneasy? Some people call this a panic attack.’’ They failed, however, meeting the criteria of PA, for example by not
reporting any or only an insufficient number of symptoms or by denying that the attack occurred out of the blue.
the same rate with which PD without AG occurs. In
differences. According to retrospective cross-sectional
addition to AG without fearful spells (1.5%), 1.3% of
community studies covering ages 181, all three groups
cases report fearful spells after the onset of AG. The
reveal a mean age of onset of 21–23 years with overall
strict DSM-IV-TR definition of AG without a history of
strikingly similar age of onset curves.[56] As a function
PD is only met by 0.6%, whereas the rate of PD with
of different sample composition, some studies reported
AG is 1.9%. Along with other information and
slightly higher ages (23–36).[48,78,79] The studies con-
transitional analyses, this study concluded that AG is a
sistently show that two thirds of all PD cases develop
clinically significant disorder that may exist indepen-
before age 35, with a substantial incidence risk in late
dently from PA and PD in a substantial proportion of
adolescence and rarely in childhood.[80] There are some
subjects. This exploration also shed light on a second
indications for a bimodal high risk distribution for PAs,
problem inherent in DSM-IV criteria and diagnosis of
with one peak around ages 15–19 and another one at
PD and AG, namely that the subjects in the population
higher ages (35–50).[59,81–83] Some studies suggest that
without PA and PD remain undiagnosed because of the
the presence of AG seems to affect the age of onset of
DSM stipulation that AG syndromes occur in response
PD with some inconsistent indications for a later age of
to panic-like symptoms. As documented in this study, a
onset.[35,84] Overall, age of onset characteristics of AG
substantial number of subjects fail to report or
without PA are less well established. Among those
experience any such symptoms.
specifically addressing AG without PAs, a slightly later
Gender and age of onset of PA, PD, and AG. A
mean onset of AG was shown, ranging from 25–29
stronger female preponderance was found for AG
years of age,[59,78,83,85] with some indications for a
without panic than for PD.[47,77] To our knowledge,
bimodal distribution (second high incidence period
there are no studies that have reported significant
after age 40) regarding onset risk. This is only partially
age  gender interaction differences between PD and
consistent with other findings that indicate that AG
AG. Age of onset characteristics of PAs as well as PD
may occur as early as childhood in greater frequency
with and without AG are well studied and suggests little
than PD.[26,53]
Depression and Anxiety

---

120
Wittchen et al.
To summarize, despite some differences between AG
and PD/AG are noticeably more frequently unem-
without PAs and PD with AG, the size of the
ployed and disabled than individuals without AG.[26,45]
differences and the extent of evidence is not sufficient
Co-morbidity patterns. Few studies exist that
to conclude that age of onset characteristics of AG
allow for direct comparisons of co-morbidity patterns
differ substantially from those observed for PD.
in community samples. There is agreement, however,
Impairment. The few available studies,[33,62,86]
that PD and AG both are rarely seen in pure forms and
which have directly compared differences in impair-
both are significantly associated with many other
ment and disability between PD with and without AG
diagnoses, including other anxiety, mood, substance,
and AG (without PAs or panic-like features), consis-
and somatoform disorders.[54,56,62,90–92] Direct com-
tently find greatest impairment for those with PD/AG
parisons reveal higher co-morbidity rates with depres-
and lowest impairment for those with PAs without
sive disorders for PD/AG (52%) and AG with PA
meeting criteria for either disorder, with PD (without
(52.3%) than those who have AG without PA (33.1%).
AG) and AG (without PAs or panic-like features) in
Thus, PD/AG and AG with PA are similarly strong
between. It is noteworthy that PD without AG has
predictors for increased depression risk. Co-morbidity
significantly fewer impairments than PD/AG in all
with other anxiety disorders was found to be in the
available indices (i.e., role functioning, work produc-
same range for all groups (49–64%), with no significant
tivity, disability days, Panic Disorder Severity Scale,
differences, though AG without PA appeared to be
Sheehan Disability Scale) and does not differ from the
more comorbid with other phobic disorders. In terms
impairment observed for AG without panic.[33] It
of transitions from one syndrome to another, a number
should also be highlighted that these findings are
of authors[26,74,93] have highlighted that temporally
consistent with findings suggesting that the degree of
primary PAs are a sensitive marker of subsequent
AG seems to be a more potent determinant of disability
psychopathology (over 90% developed at least one
than number and severity of PAs.[87] This seems to
mental disorder), but not necessarily specific for PD,
suggest that (a) with the exception of PA, all three
AG, or other anxiety disorders.[1] Similarly, high
conditions are quite impairing disorders, with PD/AG
associations were found for mood disorders, psychotic
cases revealing the most severe expressions, (b) AG
disorders, and substance use disorders. In contrast, AG
without PAs is as impairing as PD without AG, and (b)
is more closely and specifically linked to anxiety
that AG in PD patients considerably worsens function-
disorders (highest probability) and secondary depres-
ing and degree of disability.
sion.[89,91,94]
Helpseeking and treatment. Few studies have
Summarizing the available evidence from epidemio-
examined help-seeking rates for all three conditions
logical studies in the community with regard to
considered. Wang et al.[88] reported for the United
associated factors in PD, PD/AG, and AG, there seems
States that PD (41.2%) and AG without PD (42.1%)
to be considerable evidence that AG without PAs and
rank among the mental disorders having most fre-
even panic-like features exists and reveals similar
quently received ‘‘at least minimally adequate treat-
impairment and disability findings as PD without
ment.’’ However, AG without PD differs substantially
AG. Overall PD/AG seems to be the most impairing
from PD patients in that they received less treatment
conditions. Beyond some indications for minor differ-
by psychiatrists and the general health care sector.
ences in the other factors considered, there is little
Similar results were reported from a comprehensive
persuasive evidence for the existence of major differ-
analysis of 14 European Union studies.[47] The lower
ences between AG without PAs and PD. Co-morbidity
rates for AG cases in psychiatric and general health
analyses, however, reveal that AG appears to be more
care are consistent with clinical observations that such
specifically linked to other anxiety disorders and
patients are rarely seen in specialized care.[15] More
phobias in particular, whereas PA and PD are
detailed analyses by Nocon[89] in Germany further
associated with a broader spectrum of comorbid
reveals that professional help-seeking among those
disorders.
with PD/AG and those with AG with PA occurred
significantly more often because of panic problems
than AG problems. It should be noted, however, that
such patterns depend heavily on characteristics of the
VULNERABILITY AND RISK FACTORS
health care system. For example, AG without PA in
In the next section, we will consider evidence for
Germany is predominantly seen by psychotherapists,
differences between PD and AG without PA in
whereas PD is more frequently seen by psychiatrists.
vulnerability and risk factors as well as selected clinical
Kessler et al.[45] reported similar indications for the
aspects.
United States, although they were not statistically
Genetic and familial factors. There is consider-
significant.
able evidence for the familial aggregation of PD from
Socio-demographic correlates. There is no con-
various types of clinical and family studies. Studies
sistent evidence that salient correlates differ across the
fairly consistently show higher rates of PD in all first-
diagnostic groups, with one noteworthy excep-
degree relatives[95–97] of PD patients as compared to
tion. Namely, individuals who have AG without PA
controls. Hayward et al.[98] suggest that parental
Depression and Anxiety

---

Review: Agoraphobia Review
121
history of PD/AG may have a core role in the
have been surprisingly few studies that examined
development of PAs in offsprings. However, one should
differences between PD, PD/AG, and AG without
caution this conclusion because PA and PD also
panic. A notable exception is a study by Garvey and
increase the morbidity risk of offsprings for a whole
Noyes[129] that directly and specifically examined whether
broad range of other disorders.[99,100] Nevertheless, the
PD and AG are variants of the same disorder or distinct
handful of studies that accounted for other disor-
diseases by laboratory measures. They examined 91 AG
ders[95,101–104] suggest that there is at least some
patients, without specifying though the existence of
specificity for PD/AG. Only a few studies examined
PA and 24 PD patients in terms of levels of the urinary
whether there is a differential familial aggregation of
lysomal enzyme NAG that has been discussed to
PA/PD and AG. Harris et al.[105] found that the
be a marker of serotonin binding and metabolism.
increased risk of relatives of agoraphobic patients is
This study revealed that NAG levels were signifi-
not confined to AG (33% vs. 15% in controls), but also
cantly lower in patients with PD as compared to AG,
for PD (32%) and other phobias. In contrast, Smeraldi
providing limited support that PD and AG may be
et al.[106] reported an increased risk only for AG and
distinct illnesses.
not for PD. Further, Tsuang et al.[107] reported that AG
Family climate. Most studies in patients with PD
among those with PA is familial, yet AG in twins is not
and AG describe the family climate or child rearing
associated with increased PD aggregation, suggesting
behavior as being characterized by reduced warmth and
that AG and panic liability are not positioned on an
increased overprotection.[89,130–133] This relationship
agoraphobia-panic continuum. Nocon et al.[108] found
is best established for PD,[132,134,135] but this factor
that PD and AG aggregate in families; AG without PD
has been rarely considered specifically in studies with
is not familial but it might enhance the familial
AG without PAs. Some indication for specificity has
transmission of PD. Both parental AG and PD
been reported by Kendler et al.[136] who showed that
similarly increase the risk in offspring to develop any
AG, but not PD, in females was significantly asso-
anxiety disorder.[109]
ciated with parental lack of warmth, overprotection,
Heritability has been estimated to be 43–48% for
and autoritarism.
PD[110–112] and 61% for AG.[112,113] Kendler et al.[113]
Critical life events. There is evidence that negative
estimated for phobias among females a high heritability
events in childhood (e.g., separation, death of parent,
(40–60%), suggesting a ‘‘phobia proneness’’ with AG
etc.) are associated with both AG and PD.[114,137–141]
revealing the strongest and most specific associa-
Increased rates of critical life events have been associated
tions.[114,115]
with the onset of various disorders, with little evidence
Molecular genetic strategies have been used extensively
for diagnostic specificity.[141,142] However, few studies
in PD (reviewed in[89]) and have identified a wide
suggest that early death and separation was associated
range of candidate genes and suggested mechanisms
with PD, while only death (not separation) was
related to neurobiological and pharmacologic targets,
associated with AG.[114,143] Similarly, there is no
most of which remain controversial: Locus coeruleus
evidence that AG and PD nor any other anxiety disorder
(NTRK3),
dopamine
(D4DR,
DAT),
Serotonin
reveal differences in the structure and frequency of
(5HT-1A, 5HT-2C, 5HTTLPR), Katechoalminsystem
life events over the life span; all conditions were
(MAO-A, COMT), Neuropeptide (NPY Y1, Y2, Y5),
increased when significant life events were (retrospec-
Adenosinreceptors (ADORA1, ADORA2a), and CCK
tively) reported.[144]
(e.g., CCKAR).[116,117] In addition, there is a wide
Temperamental antecedents and personality. Similar
range of studies that conducted genomewise scans
to all anxiety disorders, behavioral inhibition and neurotic
without any clear replicated candidates. More impor-
disposition (i.e., neuroticism, negative affect, anxiety sensi-
tantly, there are only very few studies that specifically
tivity) are associated with AG, phobic disorders, PD, and a
separate PD and AG. Politi et al.[118] reported no
range of other conditions. Overall, there is little convergent
association between Glyoxalas-1-Polymorphisms and
evidence that these dispositional measures are diagnostically
PD/AG, unless they exclude AG cases. Rothe et al.[116]
specific. Behavioral inhibition has been shown cross-
reported differential findings with regard to the 5HT
sectionally and prospectively to be associated with many
Transporter polymorphism for PD with and PD
anxiety
disorders,[140,145–149]
including
PD
and
without AG. Hettema et al.[119] reported association
AG.[140,142,150,151] Similarly, neuroticism[79,152] and negative
with the COMT-gene for females in PD and AG.
affect[153–155] confirmed this association with little to no
To summarize, there is little evidence for diagnosti-
evidence for differences. It should be noted, however, that
cally specific genetic mechanisms in either PD or AG;
specific tests between PD, PD/AG, and AG without PAs
nor is there sufficient evidence for different mechanism
were not conducted.
between them.
A notable exception is anxiety sensitivity, or the trait
Other neurobiological factors. Although the psy-
to disposition to believe that symptoms of anxiety are
chophysiology and neuroendocronological mechanism of
harmful, as measured by the Anxiety Sensitivity Index
panic-agoraphobia are well studied (see reviews:[120–122]),
(ASI). A review by Hirshfeld-Becker et al.[156] suggests
including studies in the context of panic provocation
that the ASI is a specific predictor for PD but not for
tests (see reviews by[123–128]), to our knowledge there
other anxiety disorders. However, Hayward and
Depression and Anxiety

---

122
Wittchen et al.
Wilson[157] find that the ASI also predicts AG without
coping strategies are associated with higher levels of
the presence of PA.
anxious responding and increased distress in res-
To summarize, the distinction between PD, PD/AG,
ponse to bodily sensations.[168,170] Thus, it may be
and AG without PA has been rarely specifically addressed
that underlying cognitive or psychological factors
in studies on vulnerability and risk factors. Due to design
(e.g., coping style) dispose certain PD patients to
and assessment problems, there is no conclusive evidence
develop AG. If this is the case, these underlying factors
that reliably informs us about differences among these
may help account for the clinical utility associated with
conditions. The same, however, also applies to the direct
situational avoidance.
examination of differences between other anxiety and
Clinical course and outcome. The clinical course
mood conditions. Increased research on diagnosis-
of PD and AG in clinical and epidemiological samples
specific vulnerability, risk factors and possible interac-
have been reported as being chronically persistent (AG)
tions may reveal specific risks.
and chronically recurrent (PD).[171] Emmelkamp and
Clinical-phenomenological
aspects. There
is
Wittchen[94] found that AG without panic ranks among
convergent evidence from studies that AG—measured
the most persistent disorders over a period of 10 years
with various instruments—is a reliable construct that
follow-up, with a homotypic continuity greater than
appears in virtually all taxometric investigations as one
the one found for PD. They also reported that in
of the major classes.[38,87,158–162] It has been repeatedly
comparison to all other phobias, complete remissions
associated with social role impairments and clinical and
are rare;[77] none of the initial AG cases assessed in that
cognitive correlates.[163–166] The consistency with
study achieved complete remission.[94] This is true,
which this has been found in the context of PD led
despite considerable variations in severity and various
some to suggest that AG is an indicator of severity for
degrees of syndromal shifts that might occur as with
PD.[56,167] Using AG indicators from various sources,
regard to comorbid disorders, in particular the
Slade et al.[87] examined the latent structure of AG in
occurrence of depression. Regarding predictors of
patients with PD and PD/AG and a community
long-term outcome of PD, the presence of severe AG
sample, and identified an underlying dimensional
has been the most consistent finding.[32,66,81,172,173] AG
structure that suggests AG should be best conceptua-
severity was shown to reduce the chance of full
lized as a continuum of avoidance. Although this study
remission, to increase risk of relapse, and to enhance
did not include a group of AG without PA, it adds to
chronicity. In the longest follow-up study of PD with
our knowledge of AG as an important construct.
AG treated by exposure (2–14 years), the presence of
Another interesting exploration by Schmidt and
residual AG was a strong predictor of relapse into
Cromer[37] has suggested that an AG specifier provides
panic.[174] Additional factors that significantly contri-
meaningful information regarding the expression of
bute to chronicity and relapse are comorbid depression,
PD, specifically with regard to social functioning
personality disorders, and high scores on dispositional
impairment and total distress experienced.
measures. It is noteworthy though that there are no
An important finding from this report[37] is the clear
studies that specifically compared PD, PD/AG, and
suggestion that the highest level of clinical utility might
AG patients in this respect.
be achieved by reverting to a dimensional measure that
Treatment. A large body of research suggests that
is specific to situational avoidance (vs. distress or use of
various forms of CBT and antidepressives are highly
companions). Among the different outcomes that were
effective in treating PD/AG, including long-term
assessed, avoidance, and in particular the dimensional
efficacy.[175] There is some controversy whether all
measure of avoidance, consistently explained the most
these treatments are equally effective and whether
variance beyond that accounted for by overall severity,
combined
treatments
provide
additional
bene-
the three core panic variables (frequency, intensity, and
fits.[175,176] Some meta-analyses suggest that PD/AG
worry) and DSM-IV AG diagnosis. These data provide
improvements are greater with CBT alone than with
some further indirect support for the importance of
pharamacotherapy alone or combined with psycholo-
AG, particularly in terms of greater clinical utility.
gical treatment, yet this conclusion has been criticized
Although this exploration does not address the
as being flawed.[177] There is an impressive body of
independent diagnostic status of AG, it suggests that
evidence (reviewed by[4]) on the efficacy of exposure in
evaluating the level of situational avoidance would also
treating AG. It is difficult, however, to extrapolate from
avoid some of the criticism of the expanded criteria
pre-DSM-III literature whether samples were predo-
AG,[39] and might also improve diagnostic reliability[168]
minantly associated with panic or not. Similar con-
toward the use of avoidance (vs. distress and use of
siderations apply to the efficacy of pharmacological
companions). Although situational avoidance appears
treatment (particularly, imipramine).[4] Homework
to have important clinical utility in the context of a PD
exposure treatment directed to agoraphobic avoidance
diagnosis, it is not clear why this is the case. Feldner
has been used to treat PD with AG. In the London–
et al.[169] found that patients with PD utilize more
Toronto study,[178] homework exposure targeted agor-
avoidance-based
coping
strategies
and
ulti-
aphobic avoidance, and pharmacological placebo
mately view these strategies as more effective, when
was found to be significantly more effective than
compared to controls. Moreover, avoidant-based
alprazolam and a psychological placebo (relaxation)
Depression and Anxiety

---