The Relationship of Agoraphobia and Panic in a Community Sample of Adolescents and Young Adults

Background  Data are presented on the prevalence of DSM-IV panic and agoraphobia in a community sample of adolescents and young adults in Munich, Germany.

Methods  A total of 3021 respondents aged 14 to 24 years were assessed with a revised version of the Composite International Diagnostic Interview (CIDI). Respondents classified as having agoraphobia without panic were subtyped by number of agoraphobic trigger situations and subjected to a clinical review.

Results  Lifetime prevalence of DSM-IV agoraphobia in the revised CIDI was higher (8.5%) than that of panic attack (4.3%) or panic disorder with (0.8%) or without (0.8%) agoraphobia. Marked differences in symptomatology, course, and associated impairments between panic disorder and agoraphobia were found. Most patients with agoraphobia reported neither full nor limited attacks or uncued paniclike experiences. Clinical review revealed that many respondents classified by the CIDI as having agoraphobia actually have specific phobia, resulting in a corrected agoraphobia prevalence of 3.5%. Number of agoraphobia trigger situations was identified as a useful way of differentiating patients with true agoraphobia from those with simple phobia. Even after correcting for overdiagnosis, however, the majority of respondents with confirmed agoraphobia were found not to have a prior history of panic.

Conclusions  The results call into question the assumed key pathogenic role of panic attacks in the onset of agoraphobia. Consistent with findings that agoraphobia without panic is rarely seen in clinical settings, we find that such patients seldom seek professional treatment.

THE RELATIONSHIP between uncued spontaneous panic attacks and agoraphobia has received considerable attention during the past 2 decades. Based on the influential work of Klein,^1,2 and supported by subsequent neurobiological and cognitive studies,^3-7 agoraphobia has been widely viewed as a secondary complication of spontaneous panic attacks or paniclike experiences. Consistent with this view, the DSM-IV⁸ classified agoraphobia under the category of panic disorder with agoraphobia and relegated the diagnosis of agoraphobia without a history of panic to a residual category.⁹ Agoraphobia without panic, in this scheme, shares the same essential features as agoraphobia with panic except that the focus of fear is on the occurrence of incapacitating or embarrassing paniclike symptoms or limited attacks rather than on full panic attacks.

Although this view has received considerable support from clinical studies,^10,11 in which agoraphobia without a history of at least subthreshold panic attacks (limited attacks) is quite rare, it has not gone unchallenged.¹² Most strikingly, epidemiological studies, with only rare exceptions,¹³ consistently find that only a minority of respondents classified as having agoraphobia report the occurrence of panic attacks before the onset of their agoraphobia.^14-19 This finding casts doubt on the assumed key pathogenic role of spontaneous panic attacks for agoraphobia. An alternative view, codified in the International Statistical Classification of Diseases, 10th Revision (ICD-10)²⁰ regards panic attacks as diagnostically relatively unspecific expressions of severe anxiety.¹²

An explanation for the apparent divergent findings in clinical and epidemiological samples might be that patients who have agoraphobia without panic are less likely to seek treatment than those with panic. Another possibility is that the epidemiological findings are not valid. This possibility is supported by clinical reappraisal studies of respondents in epidemiological surveys classified as having agoraphobia without panic,^9,11,21 which have shown that many such respondents are more accurately classified as having anxiety disorders other than agoraphobia, predominately specific phobia.

The current report presents general population data from an epidemiological survey of adolescents and young adults that provides further information on this issue. Like other epidemiological surveys, a diagnostic interview is used to generate diagnoses. However, in recognition of the complexities in diagnosing panic and agoraphobia, the interview was revised and enlarged to provide a comprehensive assessment of panic and agoraphobia and their temporal relationship.

These data come from the baseline wave of the Early Developmental Stages of Psychopathology study, a longitudinal general population survey of adolescents and young adults aged 14 to 24 years in Munich, Germany, designed to collect data on the prevalences, risk factors, comorbidities, and course of mental disorders across a 5-year period.^22,23 The sample was drawn from 1994 government registries of all residents in metropolitan Munich who were expected to be aged 14 to 24 years during the first half of 1995. Respondents aged 14 to 15 years were oversampled. The 3021 interviews were conducted face to face using a computer-assisted method in the homes of respondents during this time period. The response rate was 71%.

Diagnostic assessment was based on an expanded version of the World Health Organization's Composite International Diagnostic Interview (CIDI)²⁴ known as the Munich CIDI (M-CIDI).^23,25-27 The M-CIDI is designed to assess symptoms, syndromes, and diagnoses of 48 mental disorders along with information about onset, duration, and clinical and psychosocial severity according to DSM-IV criteria. It allows for a more detailed and comprehensive evaluation of panic attack, panic disorder, and agoraphobia than the standard CIDI by adding the following features: (1) separate 12-month and lifetime questions to assess "fearful spells" (discrete periods of intense fear or discomfort) and spontaneous "limited attacks" (attacks with 1-3 symptoms) in addition to full-blown DSM-IV panic attacks; (2) expanded questions for associated diagnostic features of panic and agoraphobia, including questions about situational-bound and predisposed panic attacks as well as paniclike experiences; (3) age-of-onset questions for key features of each syndrome to evaluate temporality; and (4) verbatim descriptions for panic and phobic experiences aimed at allowing clinical reevaluation of appropriateness of responses after completion of the interview.

Diagnoses were obtained by using the M-CIDI diagnostic algorithms. Agoraphobia without history of panic disorder as classified by the DSM-IV was subtyped for number of agoraphobic trigger situations endorsed, with respondents reporting fear of only 1 (AG1) or more (AG2+) of a total of 6 agoraphobic situations. This modification was made based on previous findings that patients who report AG1 are less likely than those reporting AG2+ to be truly agoraphobic^28,29 and is consistent with DSM-IV's stipulation that agoraphobic fears "typically involve characteristic clusters of situations . . . ."⁸(p396) The M-ClDI was administered in its computerized form using a laptop computer. The mean duration for completing the interview was 77 minutes.

Psychometric properties of the CIDI have been established in several studies.²⁷ The M-CIDI was additionally tested for test-retest reliability in a community sample.^26,28 Procedural clinical validity was examined in 60 outpatients against clinicians' best-estimate consensus diagnoses and taking into account all available information (medical records, diagnostic interview, diagnostic checklist) at the end of patients' treatments.²⁹ Test-retest reliability for lifetime diagnoses was κ=1.00 for panic disorder, κ=0.66 for agoraphobia with only one trigger situation (AG1), and κ=0.84 for agoraphobia with 2 or more trigger situations (AG2+).²⁸ Agreement with clinical consensus diagnoses (validity standard) was κ=0.74 for panic disorder, .66 for AG1, and .84 for AG2+.²⁹

The survey staff consisted of 10 clinical interviewers (clinical psychologists in postgraduate education) and 25 full-time professional health research interviewers (professional background: 2 physicians, 6 psychologists, and 5 social workers; the remainder had various other professional backgrounds). All had extensive experience in diagnostic interviewing, including use of the CIDI, from various other medical surveys (mostly pharmacoepidemiological in nature) and the developmental work of the M-CIDI. Interviewers received 2 weeks of training in both the paper-and-pencil and computerized versions. This was followed by at least 10 practice interviews that were closely monitored by our staff and additional 1-day booster sessions throughout the study.

The first author (H.-U.W.) and 2 clinical psychologists, all experienced with anxiety disorders and the Structured Clinical Interview for DSM-IV,³⁰ separately reevaluated 173 interviews of respondents reporting agoraphobic fears. The aim was a clinical consensus diagnosis that took into account all the available data from the M-CIDI interview, including the verbatim entries from the open-ended questions, as well as treatment information available for 39 subjects obtained from the records of treating mental health specialists, obtained with written consent from respondents. The clinical reviewers were instructed to explore a potential underestimation of panic and overestimation of agoraphobia by determining whether symptom data and verbatim examples clearly fit the DSM-IV agoraphobia description. They were also asked whether available treatment records supported the diagnosis, if there was any information for the presence of paniclike experiences, and whether the respondent's symptom description might be better accounted for by other DSM-IV diagnoses. In 26 cases of doubt or no consensus among the 3 evaluators, a separate Structured Clinical Interview was conducted for clarification.

A weight was used to adjust for this oversampling as well as for minor discrepancies between the sample and population distributions on the cross-classification of age, sex, and neighborhood.^22,23Table 1 reports N^weighted and prevalence^weighted for more detailed information. Analysis of weighted data was carried out using the STATA software package (STATA Corp, College Station, Tex). Age-of-onset analyses were obtained using the SURVIVAL procedure. The Kaplan-Meier log rank statistic was used to test differences between hazard curve distributions. Statistical significance was evaluated at the .05 level using 2-sided tests.

Among 14- to 24-year-olds in the population (Table 1), 4.3% fulfilled lifetime criteria for DSM-IV panic attack and approximately one third of the latter fulfilled criteria for DSM-IV panic disorder. Approximately half of those with panic disorder also were classified as having agoraphobia. The prevalence of agoraphobia without a history of panic disorder was considerably higher (7.8%), with the majority (5.5%) falling into the AG1 group. The female-male odds ratio (OR) is significantly greater than 1:1 for panic attacks (2.3; 95% confidence interval [CI], 1.9-3.2), panic disorder (3.1; 95% CI, 2.4-4.7), AG1 (2.3; 95% CI, 1.4-3.6), and AG2+ (4.4; 95% CI, 2.2-7.4).

Among subjects with agoraphobia, 22% of those with AG1 and 34% with AG2+ reported at least 1 spontaneous "fearful spell or attack" "when all of a sudden [they] felt frightened, anxious, or very uneasy"; most of whom (AG1, 16.3%; AG2+, 27.4%) also had 4 or more symptoms during one of these spells, while an additional 5.6% with AG1 and 6.5% with AG2+ had "limited-symptom attacks" (Table 2). All of these respondents failed to meet diagnostic criteria for a full panic disorder because of not reporting either 4 or more symptoms, a "crescendo" of symptom onset, recurrent attacks, concerns about another attack, worry about the implications, or significant behavior change. It is noteworthy that 78% of subjects with panic disorder without agoraphobia described their panic attacks as being predominantly uncued, as compared with 58% of those with panic disorder with agoraphobia and only 27% of those with agoraphobia and a fearful spell. The latter mostly indicated at least 1 situational or predisposed attack. Respondents classified with AG1 and AG2+ with fearful spells or attacks were less likely than those with panic disorder to worry about either additional attacks (OR=0.4; 95% CI, 0.3-0.6) or the implications of these attacks (OR=0.5; 95% CI, 0.3-0.9). They were also less likely to report significant behavior changes because of the spells (OR=0.3; 95% CI, 0.2-0.7). It is also noteworthy that a substantial number of respondents reported either fearful spells (10.5%), limited-symptom attacks (3.7%), or full-blown panic attacks, but did not meet criteria for either panic disorder or agoraphobia.

Only 1 case of panic disorder without agoraphobia and 5 cases from the "neither panic disorder nor agoraphobia" group endorsed any agoraphobia stem question, indicating that subthreshold agoraphobia syndromes are rare (Table 2, bottom section). Respondents classified as having panic disorder with AG1 and AG2+ reported an average of 2.5 and 2.7, respectively, agoraphobia trigger situations. The most frequently mentioned such situations among respondents with AG1 were "unreasonably strong fears of crowds and standing in line" (31.1%), whereas about two thirds of the patients with panic disorder with agoraphobia and AG2+ reported "leaving home alone" and "using public transportation." Across groups, almost all endorsed either endurance with dread or requiring a companion when confronting the agoraphobic situations. Significant avoidance was most frequent in panic disorder with agoraphobia (59.2%) and lowest in AG1 (39.6%). Patients with panic disorder with agoraphobia reported significantly more anxiety symptoms when being or thinking about encountering agoraphobic situations (mean, 6.5) than those with AG2+ (mean, 2.9; χ² value, 15.39; P=.001) or those with AG1 (mean, 2.5; χ² value, 4.38; P=.04).

A key issue in the diagnostic evaluation of DSM-IV agoraphobia is whether subjects are afraid of suddenly occurring paniclike experiences when thinking about or being about to enter trigger situations. The M-CIDI evaluated this issue by presenting respondents with a list of paniclike symptoms and cognitions and asking separate questions about fear of incapacitation, fear of embarrassment, and fear of having no help available. At least 1 such fear was acknowledged by only 18.9% of AG1 cases, 23.2% of AG2+ cases, and 58.6% of patients having panic disorder with agoraphobia. The most frequently endorsed specific concerns across these groups were being afraid of panic attack (n=36), fainting (n=6), dizziness (n=7), suffocation (n=4), losing control (n=4), developing embarrassing symptoms (urination or vomiting; n=3), or heart problems and fear of dying (n=2). Most AG1 (81%) and AG2+ (76%) cases not endorsing any such paniclike internal symptoms and sensations mentioned 1 or more external threat–related cognition such as unreasonably strong fear of being attacked or mugged, getting lost, catastrophic events that might happen (accidents, collapsing of bridges or tunnels, or being trapped in a crowd during a mass event [eg, Oktoberfest]) as the focus of their agoraphobic cognitions.

The cumulative hazard rates for age of first onset in Figure 1 (A) reveal that fearful spells and limited attacks are overall more frequent than full DSM-IV panic attacks and more often begin in childhood (P<.001). Full DSM-IV panic attacks are rare before the age of 15 years and show considerable increases in incidence after that age, with some indications for continued increased risk for females but not for males. Part B of the figure demonstrates considerably earlier onsets for agoraphobia (mean age with AG1, 12.6 years; mean age with AG2+, 12.9 years) compared with panic disorder with (mean age, 14.5 years) and without (mean age, 16.2 years; P<.001) agoraphobia. The cumulative hazards function for AG1 cases without fearful spells in part C of the figure differs markedly from agoraphobia with fearful spells, with the latter displaying cumulative hazards characteristics similar to panic disorder. It is noteworthy that across the 4 agoraphobia groups almost identical low mean ages of onset were found, ranging from 12.5 to 13.1 years.

Among respondents with agoraphobia, those with paniclike experiences (AG2+/fearful spells) reveal more similarities to panic disorder than those without panic (AG2+), whereas AG1 cases differ significantly in most variables considered (Table 3). Compared with panic disorder, AG2+ cases reveal only 2 significant differences: (1) they are more likely to report fearful spells secondary to agoraphobia (OR=6.4); (2) AG2+ cases without fearful spells are less likely to report severe psychosocial impairments. In contrast, the 2 AG1 groups are less likely than the panic disorder group to be currently symptomatic, to report persistence of agoraphobic trigger avoidance, to have comorbid depression, and to be severely disabled; they also are less likely to report repeated inability to work in the past month. Those with agoraphobia are more likely than those with panic disorder to report an onset of their first fearful spell or attack secondary to their agoraphobia onset (OR=6.4).

As shown in the last 2 rows of Table 3, subjects with fearful spells or panic disorder—as compared with those with neither agoraphobia nor paniclike experiences—are much more likely to seek professional help and receive treatment than those not having either panic or agoraphobia. Almost all respondents with panic disorder and AG2+/fearful spells report seeking professional help. Treatment rates among the remaining groups were considerably lower but still significantly elevated, except for the AG1 group.

Consistent with the low threshold for evaluating panic that we used, no indications were found for underestimation of panic attacks in the clinical reappraisal of agoraphobia. Furthermore, review of verbatim descriptions of fears in the open-ended questions of AG2+ cases, together with available treatment records, failed to raise any doubts that these were true cases of agoraphobia. The situation was different, though, for the 173 respondents classified with AG1, where the clinical reappraisal unequivocally confirmed agoraphobia in only 13.9% of cases. The vast majority of M-CIDI AG1 cases (Table 4) were diagnosed as having specific phobias mostly of the situational type, such as using the subway system at night or crossing certain bridges. Interestingly, many respondents related this fear to a particular bridge that had been temporarily closed for truck traffic during the time of the study. In some cases situational phobia was confounded with generalized fears about darkness. One culturally bound phobia found among the AG1 cases was triggered by experiences in the Oktoberfest crowd and generalized to other such situations later on. Other less frequent diagnoses were separation anxiety disorder, social phobia, and anxiety due to a general medical condition. The latter diagnosis was assigned (1) to a woman with an exaggerated fear that crowds and walking the dog during pregnancy could lead to dangerous complications and (2) to a subject with exaggerated agoraphobic fears related to a neurologic condition. Only 2 M-CIDI cases of AG1 were classified in the clinical review as normal, nonphobic anxiety.

This low rate of clinical confirmation has a substantial effect on our initial prevalence estimate of agoraphobia. The intial prevalence estimate of 7.8% for agoraphobia drops to 3.5%.

We investigated the claim that agoraphobia is nosologically strongly related to panic, with primary spontaneous panic attacks or paniclike symptoms underlying the vast majority of cases with agoraphobia. Although previous epidemiological studies have failed to confirm this assumption, recent research suggests that this was because the lay-administered structured diagnostic interviews used in these studies were invalid. We administered a comprehensive series of structured and open-ended questions about agoraphobic symptoms to address this problem. The young age of the sample minimized the possibility of recall bias that has plagued previous studies in this area. However, some important limitations remain. These include the fact that the results cannot be generalized to people older than 24 years, that clinical reappraisal interviewers were administered only to a subset of respondents, and that we were not able to examine patterns in subsamples defined on the basis of potentially important specifying variables such as sex or social class due to lack of statistical power.

Within the context of these limitations, we found, consistent with previous epidemiological studies,^14-19,31 that the lifetime prevalence of M-CIDI/DSM-IV agoraphobia without a history of panic disorder (7.8%) is much higher than the prevalence of either panic disorder without agoraphobia (0.8%) or panic disorder with agoraphobia (0.8%). Unlike previous epidemiological studies, we investigated whether these results were due to classification errors. Although this clinical reappraisal found that agoraphobia was overdiagnosed by the M-CIDI, we were unable to document any evidence of panic or paniclike experiences in the vast majority of patients with clinically confirmed agoraphobia. Furthermore, many patients with confirmed agoraphobia with panic reported that their first panic experiences occurred only after the onset of agoraphobia.

Our failure to demonstrate a consistent temporality pattern of primary panic in patients with agoraphobia and paniclike experiences and the lack of any such panic signs in the majority of patients with agoraphobia are both consistent with the findings by Lelliot et al¹² that aversive conditioning of panic is involved in only a minority of people with agoraphobia. These results also raise the question of what those with agoraphobia actually fear. In contrast to the observations of Goisman et al⁹ and Barlow et al¹⁰ in clinical samples, the most frequently mentioned fears in our general population sample were unreasonably strong, threat-related anticipations of being attacked on public transportation, getting lost during travel or in a crowd, or some catastrophic event (such as mass panic). Furthermore, in contrast to clinical observation,³² psychophysiological symptoms were only rarely mentioned in this age group as the focus of fear, even though the M-CIDI explicitly asked about fears of such "internal" sensations.

The most plausible interpretation of why these results differ from those in clinical studies is that patients who have agoraphobia with panic are considerably more likely than those who have agoraphobia without panic to seek help from mental health professionals, leading to an overrepresentation of the former relative to the latter in clinical samples compared with general population samples. This intepretation is consistent with the strong effect of panic on help-seeking behavior in our own data, with 90% of those with agoraphobia who have at least limited attacks reporting that they contacted a health professional because of their condition, compared with only 7% to 37% of those with agoraphobia and without any paniclike symptoms.

Our findings call for a rethinking of current ideas about the triggering experiences that promote agoraphobia. While panic is clearly important in many cases, it is not the causal trigger for the majority of young persons with agoraphobia. This calls into question the notion that panic and agoraphobia are part of a single disease spectrum, with agoraphobia without panic at the low end and panic disorder with agoraphobia at the other end.⁹ Taking into account other findings about the relative unspecificity of DSM-IV panic attacks²³ and differential patterns of course in early- and late-onset panic and agoraphobia,^33,34 a seemingly more plausible scenario is that diverse pathogenic pathways exist for both panic and agoraphobia. It thus remains for future research, necessarily with community samples, to provide information on the prototypic triggering experiences of agoraphobia without panic.

Accepted for publication August 11, 1998.

Corresponding author: Hans-Ulrich Wittchen, PhD, Department of Clinical Psychology, Max Planck Institute of Psychiatry, Kraepelinstr 2, Munich 80804, Germany.