Panic and phobic anxiety: Associations among neuroticism, physiological hyperarousal, anxiety sensitivity, and three phobias

Transcription

1 Anxiety Disorders 20 (2006) Panic and phobic anxiety: Associations among neuroticism, physiological hyperarousal, anxiety sensitivity, and three phobias Susan L. Longley a, *, David Watson b, Russell Noyes Jr. b, Kevin Yoder c a Johns Hopkins Bloomberg School of Public Health, 624 N. Broadway, P.O. Box 495, Baltimore, MD , USA b The University of Iowa, IA, USA c University of North Texas, TX, USA Received 30 June 2005; received in revised form 11 October 2005; accepted 3 November 2005 Abstract A dimensional and psychometrically informed taxonomy of anxiety is emerging, but the specific and nonspecific dimensions of panic and phobic anxiety require greater clarification. In this study, confirmatory factor analyses of data from a sample of 438 college students were used to validate a model of panic and phobic anxiety with six content factors; multiple scales from self-report measures were indicators of each model component. The model included a nonspecific component of (1) neuroticism and two specific components of panic attack, (2) physiological hyperarousal, and (3) anxiety sensitivity. The model also included three phobia components of (4) classically defined agoraphobia, (5) social phobia, and (6) blood-injection phobia. In these data, agoraphobia correlated more strongly with both the social phobia and blood phobia components than with either the physiological hyperarousal or the anxiety sensitivity components. These findings suggest that the association between panic attacks and agoraphobia warrants greater attention. # 2005 Elsevier Ltd. All rights reserved. Keywords: Factor structure; Anxiety; Neuroticism; Panic; Phobia * Corresponding author. address: slongley@jhsph.edu (S.L. Longley) /$ see front matter # 2005 Elsevier Ltd. All rights reserved. doi: /j.janxdis

2 S.L. Longley et al. / Anxiety Disorders 20 (2006) Anxiety is fundamentally a useful experience that has adaptive value (Marks & Nesse, 1994; Watson, 2000). Pathological symptoms of panic and phobic anxiety, however, are undeniably debilitating (Andrews & Slade, 2002; Katerndahl & Palmer, 2000; Schmidt, Lerew, & Jackson, 1997). Consequently, there is growing consensus that a taxonomy of these fear-related syndromes should account for both normal and abnormal functioning (Krueger & Piasecki, 2002; Marks & Nesse, 1994; Smoller & Tsuang, 1998; Watson, in press). An empirically informed, dimensional taxonomy has been outlined, based on the assumption that variations in personality and psychopathology fall on a continuum (Krueger, 1999; Vollebergh et al., 2001). According to this taxonomy, panic and phobic anxiety share nonspecific, trait vulnerabilities but are differentiated by unique symptoms and vulnerabilities (Krueger, 1999; Krueger, McGue, & Iancono, 2001; Mineka, Watson, & Clark, 1998). Although research in this area has addressed significant issues, relevant analyses have largely been performed on dichotomous, rather than continuous, variables (see Watson, in press). To further articulate this dimensional framework, we will perform structural analyses on continuous, self-report measures of panic and phobic anxiety. As with other forms of anxiety, a crucial issue is to validate characteristics that are specific to the fear cluster, versus those that are shared with other forms of anxiety (Barlow, 1988; Mineka et al., 1998). In this undertaking, findings from dimensional measures of panic and phobic anxiety can both augment and clarify findings based on dichotomous diagnostic categories. These self-report measures define dimensional characteristics and assess continua that span the entire range of behavior and functioning; this circumvents the arbitrary boundaries inherent in categorical diagnostic data (Smoller & Tsuang, 1998). Thus, unlike diagnostic categories, dimensional ratings fully quantify the adaptive-to-maladaptive variations in traits and symptoms (Achenbach, 2005). Moreover, numerous self-report measures of personality and psychopathology show impressive continuity across clinical and nonclinical samples (O Connor, 2002). We will, therefore, examine self-report measures that assess theoretically significant specific and nonspecific factors that comprise panic and phobic anxiety. Neuroticism is a general component that is a vulnerability factor for all forms of anxiety (Clark, Watson, & Mineka, 1994; Hayward, Killen, Kraemer, & Taylor, 2000; Lesch et al., 1996). Panic attacks are a necessary but not sufficient feature of panic disorder; these attacks have been linked to the specific factors of: (1) physiological hyperarousal the autonomic arousal symptoms of panic attacks and (2) anxiety sensitivity the tendency to fear and catastrophically misinterpret anxiety symptoms (Barlow, Brown, & Craske, 1994; Reiss, Peterson, Gursky, & McNally, 1986). The phobias have been classically subdivided into (1) agoraphobia a fear of crowded and open spaces, (2) social phobia a fear of social evaluation, and (3) specific phobia (Marks, 1970; Marks & Mathew, 1979). Confirmatory factor analysis of these constructs will allow us to address underlying taxonometric issues about the specificity of panic attacks,

3 720 S.L. Longley et al. / Anxiety Disorders 20 (2006) agoraphobia, and specific phobia. We will determine the latent factors that are defined and their intercorrelations. 1. General and specific components 1.1. Neuroticism Various conceptualizations of general anxiety proneness share more similarities than differences. This nonspecific component reflects individual differences in general distress; individuals high on this dimension tend to report elevated levels of stress and various negative emotions (Clark et al., 1994). Of these conceptualizations, neuroticism is perhaps the most well known and it is widely recognized to be a highly stable and heritable personality trait (Rose, Koskenvuo, Kaprio, Sarna, & Langinvaini, 1988; Smoller & Tsuang, 1998; Young, Fenton, & Lader, 1971). Neuroticism is also the first anxiety phenotype to be associated with a specific genetic locus (Lesch et al., 1996). It is now agreed that neuroticism is a nonspecific diathesis shared by the mood and anxiety disorders (Bouton, Mineka, & Barlow, 2001; Mineka et al., 1998) Panic attack Panic attacks are discrete and intense periods of autonomic arousal accompanied by fear. Some have argued that panic attacks are not limited to those with panic disorder or agoraphobia, but are common to all the anxiety disorders (Barlow, 1988). This led to separation of panic attacks from panic disorder in the current Diagnostic and Statistical Manual (DSM-IV; American Psychiatric Association [APA], 1994; Barlow et al., 1994). Our review will show, however, that the specificity of the physiological hyperarousal and anxiety sensitivity components warrants reinvestigation Physiological hyperarousal The tripartite model of Clark and Watson (1991) posited that physiological hyperarousal was characteristic of anxiety, whereas anhedonia and low positive affect were specific to depression. However, Brown, Chorpita, and Barlow (1998) modeled a latent factor of anxious arousal and found it to be specific to panic disorder and not characteristic of the anxiety disorders in general. Later, Mineka et al. (1998) proposed an integrative hierarchical model of anxiety and depression in which physiological hyperarousal was reconceptualized as a specific feature of panic attacks. Joiner et al. (1999) subsequently provided additional empirical support for the integrative model. These researchers examined the specificity of physiological hyperarousal in a large sample of Air Force Cadets. They found that symptoms of physiological hyperarousal were more strongly associated with a diagnosis of panic disorder than with diagnoses of generalized anxiety disorder

4 S.L. Longley et al. / Anxiety Disorders 20 (2006) (GAD) and major depression. Our study will examine the specificity of all 13 hyperarousal symptoms as Joiner and his colleagues assessed only Anxiety sensitivity There is abundant evidence that individuals who are sensitive to somatic hyperarousal cues are more likely to experience panic attacks. Of the commonly used measures in this area of research, the Anxiety Sensitivity Index (ASI; Peterson & Reiss, 1987), which taps fear of anxiety-related symptoms, has been the best predictor of panic attacks in both clinical and nonclinical participants (Cox & Taylor, 1999; Ehlers, 1995; Lilienfeld, Turner, & Jacobs, 1993; Maller & Reiss, 1992; Reiss & McNally, 1986; Schmidt et al., 1997; Taylor, 1995). This association remains significant even after controlling for trait anxiety and depression (Schmidt, Lerew, & Joiner, 2000; Zinbarg, Brown, Barlow, & Rapee, 2001). It is now established that the ASI is multifactorial, with most studies finding a three-factor solution (Stein, Jang, & Livesley, 1999; Stewart, Taylor, & Baker, 1997; Zinbarg, Barlow, & Brown, 1997). The factors deal with the physical, social, and mental consequences of anxiety symptoms. Several studies show that the Physical Concerns subscale which assesses fear of somatic sensations most strongly predicts physiological hyperarousal (Brown, Smits, Powers, & Telch, 2003; Zinbarg et al., 2001). We will, therefore, examine the largely unexplored relationship between the ASI Physical Concerns subscale and other measures of panic and phobic anxiety Agoraphobia The accumulated clinical evidence strongly suggests that the symptoms of agoraphobia are secondary to those of panic. Accordingly, in the DSM-IV (American Psychiatric Association, 1994) agoraphobia is subsumed under the category of panic disorder. Agoraphobia without a history of panic attacks was consigned to a residual category, with the justification that In clinical settings, almost all individuals (over 95%) who present with agoraphobia also have a current diagnosis (or history) of panic disorder (APA, 1994, p. 403). Consistent with the DSM formulation, studies with clinical samples rarely find agoraphobia without sub-threshold panic attacks (Barlow et al., 1994; Goisman et al., 1994). In contrast to findings from clinical samples, the results from two communitybased studies of young adults found that panic attacks are infrequently experienced by those with agoraphobic symptoms (Hayward, Killen, & Taylor, 2003; Wittchen, Reed, & Kessler, 1998). These data cast doubt on the link between agoraphobia and panic attacks. The convergent findings from these studies are striking given that one defined agoraphobia classically (e.g., the fear and avoidance of crowds and large open spaces) and the other used DSM-IV criteria; both studies used clinically trained interviewers to assess agoraphobia (Eaton & Keyl, 1990). We will contribute to this literature by examining the specificity of classically defined, self-reported agoraphobic symptoms vis-à-vis

5 722 S.L. Longley et al. / Anxiety Disorders 20 (2006) neuroticism and other types of panic and phobic anxiety (Chambless, Caputo, Jasin, Gracely, & Williams, 1985; Marks & Mathew, 1979; Rapee, Craske, & Barlow, 1994/1995) Ambiguity regarding the relationship between agoraphobia and other forms of panic and phobic anxiety is further suggested by the conflicting results from several large epidemiology studies. Analyses of data from more than 7000 participants in the Netherlands Mental Health Survey and Incidence Study (NEMESIS; Vollebergh et al., 2001) support the DSM-IV conceptualization of a specific link between panic and agoraphobia. In these data, tetrachoric correlations among anxiety disorder diagnoses showed that agoraphobia correlated more strongly with panic (r =.75) than with either social phobia (r =.56) or simple phobia (r =.52) (W. Vollebergh, personal communication, December 15, 2003). In contrast, similar structural analyses of data from a sample of more than 8000 participants, in the National Comorbidity Survey (NCS; Kessler et al., 1994), suggest a nonspecific link between panic and agoraphobia. In these data, tetrachoric correlations among lifetime DSM-III-R diagnoses showed that agoraphobia had nearly identical correlations with panic disorder (r =.59) simple phobia (r =.58) and social phobia (r =.54) (Krueger, 1999) Specific phobia Previous structural analyses have generally aggregated the specific phobia symptoms into an overall score (Krueger, 1999; Vollebergh et al., 2001; Zinbarg & Barlow, 1996). This approach does not allow one to examine specific phobia subtypes that may have differential relations with the components of panic and phobic anxiety. In our study, we included one well-established subtype of specific phobia, namely, blood-injection-injury phobia (Cutshall & Watson, 2004; Marks, 1970; Marks & Mathew, 1979; Watson, in press) Overview of the current research These accumulated findings indicate the need for structural analyses on data from continuous measures of these constructs. In this study, we will test a model of six correlated latent factors to clarify the existence of specific versus nonspecific components of panic and phobic anxiety. This model includes one nonspecific factor of (1) neuroticism that is hypothesized to underlie and relate to all of the other specific model components. The proposed specific panic attack components are (2) physiological hyperarousal and (3) anxiety sensitivity. In addition, we will include (4) an agoraphobic component that may or may not be specific to panic attacks. Finally, we have selected (5) social phobia and (6) one specific phobia subtype to establish benchmarks for the specificity of the panic and agoraphobia factors. The findings of Krueger (1999) indicate that social and simple phobia are strongly related to both panic attacks and agoraphobia. If agoraphobia is specifically related to panic attacks, however, it should correlate more strongly with our two panic components (i.e., the physiological

6 S.L. Longley et al. / Anxiety Disorders 20 (2006) hyperarousal and the anxiety sensitivity factors) than with either social phobia or specific phobia. The basic goal of our study is to examine the relations among these model components. In the tradition of Zinbarg and Barlow (1996), we included multiple indicators of the six specific and nonspecific components, which will allow us to model these constructs as latent factors. This study is unique, in that it focuses on the construct validity of self-report measures of theoretically significant aspects of panic and phobic anxiety. It uses data from a large, nonclinical sample of young adults. We included multiple indicators of neuroticism to test our hypothesis that it is strongly and nonspecifically related to the other factors. In light of previous evidence, however, we expect that neuroticism will correlate less strongly with symptoms of blood phobia than with the other factors in our model (see Watson, in press). We further hypothesize that anxiety sensitivity and physiological sensitivity will show greater specificity (as components of panic) and be more highly correlated with each other than with any of the assessed phobias. Finally, in light of the conflicting findings discussed earlier, we will make no prediction regarding whether agoraphobia will correlate more strongly with our panic attack components than with the other anxiety factors. 2. Method 2.1. Participants and procedure The participants were 438 young adults enrolled in an introductory psychology class. The majority were women and Caucasian. These students participated in this study in partial fulfillment of a course research exposure requirement. They were assessed in small group sessions of participants Overview of measures As noted earlier, we included multiple markers of each of the six hypothesized factors in our model neuroticism, physiological hyperarousal, anxiety sensitivity, agoraphobia, social phobia, and blood-injection phobia in our assessment battery. These measures were selected on the basis of their item content and wherever possible on prior evidence establishing that they were substantially correlated and appeared to measure the same basic construct. We subsequently report correlational and confirmatory factor analyses that evaluate the validity of our measurement model Neuroticism measures Big Five Inventory Neuroticism Scale (BFI Neuroticism; John & Srivastava, 1999). The BFI is a 44-item measure that assesses the five general dimensions of personality; its five-factor structure has been verified using factor analysis. The

7 724 S.L. Longley et al. / Anxiety Disorders 20 (2006) BFI has separate scales that assess the five personality traits of Openness to Experience, Agreeableness, Conscientiousness, Extroversion and Neuroticism. All scales have excellent validity and reliability. Items are rated on a five-point scale from very like me to very unlike me. Only the eight-item Neuroticism Scale (i.e., BFI Neuroticism) was used in this study. Penn State Worry Questionnaire (PSWQ; Meyer, Miller, Metzger, & Brooke, 1990). The PSWQ is a 16-item scale that is designed to tap chronic worry. Items are rated on a five-point scale ranging from not at all typical of me to very typical for me. Evidence from clinical and nonclinical samples suggests that the PSWQ is a reliable, unidimensional measure with good discriminant and convergent validity. Our data, however, consistently indicate that the PSWQ is very strongly correlated with measures of neuroticism (see, for example, the current Table 1). We, therefore, used the PSWQ as a second marker of Neuroticism in subsequent analyses Physiological hyperarousal measures Mood Anxiety and Symptom Questionnaire Anxious Arousal Scale (MASQ Anxious Arousal; Watson & Clark, 1991). Our participants completed a 62-item short form of the Mood and Anxiety Symptom Questionnaire, indicating the extent to which they have experienced each symptom during the past week, including today. The MASQ was constructed to test key aspects of the tripartate model of depression and anxiety proposed by Clark and Watson (1991) (see also Clark et al., 1994; Mineka et al., 1998). The MASQ short form contains two anxiety scales and two depression scales. We will only report data on the 17-item Anxious Arousal Scale (i.e., MASQ Anxious Arousal), which taps various manifestations of physiological hyperarousal (e.g., feeling dizzy or lightheaded, shortness of breath, dry mouth). Panic Attack Symptom Questionnaire (PASQ; Watson, unpublished). The 13- item PASQ is a face valid, self-report measure of the 13 DSM-IV symptoms of panic attack (e.g., chest pain or discomfort; feeling of choking; afraid that you were dying). Participants were instructed that Everyone experiences episodes of nervousness, tension, and anxiety from time-to-time. Listed below are a series of experiences that sometimes occur during episodes of anxiety. Please indicate how much you have experienced these things during the past month. Ratings were made on a five-point scale ranging from not at all to extremely. This measure was used as a second indicator of panic symptoms. As shown in Table 1, the high correlation of the PASQ with the MASQ Anxious Arousal Scale (r =.71) suggests that it is an excellent indicator of physiological arousal Anxiety sensitivity measures Anxiety Sensitivity Index (ASI; Peterson & Reiss, 1987). The ASI is a 16-item self-report measure designed to assess the fear of anxiety-related sensations and experiences. Each item assesses a concern about the negative consequences of anxiety symptoms (e.g., It scares me when I feel shaky. ). Respondents rate the

8 Table 1 Intercorrelations of Marker Scales: Six-Factor Model Measure Factor I Factor II Factor III Factor IV Factor V Factor VI Factor I: Neuroticism (N) 1. BFI Neuroticism 2. PSWQ.74 Factor II: Physiologic Hyperarousal 3. PSAQ MASQ Anxious Arousal Factor III: Anxiety Sensitivity 5. ASI Physical Concerns subscale BSQ Factor IV: Agoraphobia 7. APPQ Agoraphobia FQ Agoraphobia MIA Alone Scale Factor V: Social Phobia 10. APPQ Social Phobia FQ Social Phobia PSRS Social Phobia PPSC Social Anxiety Factor VI: Blood Phobia 14. FQ Blood-Injection Phobia PSRS Blood-Injection Phobia Note: N = 438. S.L. Longley et al. / Anxiety Disorders 20 (2006)

9 726 S.L. Longley et al. / Anxiety Disorders 20 (2006) degree to which they endorse these concerns on a five-point scale ranging from very little to very much. As discussed earlier, previous studies have shown that the ASI is multifactorial, with most analyses identifying a three-factor solution (Brown et al., 2003; Zinbarg et al., 1997). Based on this evidence, we used the ASI Physical Concerns subscale as an indicator of anxiety sensitivity. The eight-item ASI Physical Concerns subscale is computed by summing responses to Items 3, 4, 6, 8 11, and 14 (Zinbarg et al., 1997). Body Sensations Questionnaire (BSQ; Chambless, Caputo, Bright, & Gallagher, 1984). The BSQ is a 17-item measure that is designed to assess the fear of bodily sensations that are commonly associated with panic disorder (e.g., dizziness). Items are rated on a five-point scale that ranges from not frightened or worried about this sensation to extremely worried about this sensation. A single global score is derived from this instrument Agoraphobia symptom measures Albany Panic and Phobia Questionnaire (APPQ; Rapee et al., 1994/1995). The APPQ is a 27-item questionnaire that is designed to assess symptoms of agoraphobia and social phobia, as well as fear of activities that produce introceptive sensations. A nine-point format is used to rate how much fear (ranging from no fear to extreme fear) would be experienced in reaction to an activity or situation that occurred over the next week. The APPQ is comprised of three subscales: We used the nine-item Agoraphobia subscale (i.e., APPQ Agoraphobia) to assess agoraphobic symptoms; this subscale includes items such as going long distances from home. Fear Questionnaire (FQ; Marks & Mathew, 1979). The FQ is a 15-item inventory that asks respondents to indicate the extent to which they would avoid various situations. Each item is rated on a nine-point scale ranging from would not avoid it to always avoid it. We used the five-item Agoraphobia subscale (i.e., FQ Agoraphobia) that has items such as traveling alone, by bus or by coach or being far from home. Mobility Inventory for Agoraphobia (MIA; Chambless et al., 1985). The MIA is a 26-item scale that is designed to assess situations that typically trigger agoraphobia (e.g., airplanes, buses, theatres, classrooms etc.). Items are rated on a five-point scale that ranges from never avoid to always avoid. Higher total scales indicate greater avoidance. The MIA Avoidance When Alone (i.e., MIA Avoidance Alone) subscale was used in this study as a measure of severity of impairment Social phobia symptom measures Albany Panic and Phobia Questionnaire (APPQ; Rapee et al., 1994/1995). The APPQ was described earlier. Its Social Phobia subscale (i.e., APPQ Social Phobia) contains 10 items that tap social and evaluation anxiety (e.g., meeting strangers, giving speeches, being criticized, eating in front of others ).

10 S.L. Longley et al. / Anxiety Disorders 20 (2006) Fear Questionnaire (FQ; Marks & Mathew, 1979). The FQ also was described earlier. Its social phobia subscale (i.e., FQ Social Phobia) consists of five items that tap various social fears (e.g., being watched or stared at, being criticized ). Self-Consciousness Scales (SCS; Fenigstein, Scheier, & Buss, 1975). The SCS is a 23-item measure that yields three scores (a) Private Self- Consciousness 10 items designed to measure chronic attention to one s thoughts and feelings (e.g., I reflect about myself a lot ); (b) Public Self- Consciousness 7 items designed to measure chronic awareness and social concerns (e.g., I m concerned about what other people think of me ); and (c) Social Anxiety 6 items designed to measure anxious reactions to socially selfconscious states (e.g., It takes me a long time to overcome shyness in new situations ). The items are rated on a five-point scale ranging from very unlike me to very like me. Only the six-item Social Anxiety Scale (i.e., SCS Social Anxiety) was used in this study. Phobic Stimuli Response Scales (PSRS; Cutshall & Watson, 2004). The PSRS is a measure that is composed of scales that assess social, animal, bodily harm, blood-injection and situational phobias. We used a preliminary, 14-item version of the Social Phobia Scale from the PSRS (i.e., PSRS Social Phobia). Respondents rated the extent to which they agree with various statements (e.g., I get nervous when I know people are watching me, I often am afraid of looking foolish ) on a four-point scale that ranges from strongly disagree to strongly agree Blood phobia symptom measures (specific phobia) Fear Questionnaire (FQ; Marks & Mathew, 1979). The FQ was described earlier. Its Blood-Injection-Injury subscale (i.e., FQ Blood-Injection Phobia) consists of five items that tap various fears of blood, injection, and injury. Phobic Stimuil Response Scales (PSRS; Cutshall & Watson, 2004). The PSRS was described earlier. It includes a 10-item Blood-Injection Phobia Scale (i.e., PSRS Blood-Injection Phobia). 3. Analyses and results 3.1. Correlational analyses Table 1 presents correlations among all the assessed variables. There are clear clusters among the potential indicators. Most notably, the Table 1 findings offer initial support for our predicted model consisting of (a) Neuroticism; (b) Physiological Hyperarousal; (c) Anxiety Sensitivity; (d) Agoraphobia; (d) Social Phobia; and (e) Blood Phobia. Thus, the bold typeface in the table highlights the existence of strong correlations between (a) BFI Neuroticism Scale and the PSWQ (r =.74), (b) the MASQ Anxious Arousal Scale and the PASQ (r =.71),

11 728 S.L. Longley et al. / Anxiety Disorders 20 (2006) (c) the ASI Physical Concerns subscale and BSQ (r =.69), (d) the APPQ Agoraphobia, FQ Agoraphobia, and MIA Avoidance Alone Scales (correlations ranged from.55 to.67), (e) the APPQ Social Phobia, FQ Social Phobia, SCS Social Anxiety, and PSRS Social Phobia Scales (correlations ranged from.61 to.75) and (f) the FQ Blood-Injection Phobia and the PSRS Blood-Injection Phobia Scales (r =.63). Table 1 also provides preliminary support for our predictions of differential relations between the six model components. We predicted that a common component which we identified as Neuroticism was shared by all the specific model components. Consistent with this prediction, our two Neuroticism markers (i.e., BFI Neuroticism and the PSWQ) had moderate correlations ranging from.28 to.49 with the markers of physiological hyperarousal, anxiety sensitivity, social phobia, and agoraphobia. As expected, our neuroticism measures correlated more weakly with indicators of blood-injection phobia (correlations ranged from.20 to.25), although these coefficients still were moderate in magnitude. It is also noteworthy that our two measures of anxiety sensitivity (the ASI and BSQ) correlated more strongly with markers of physiological hyperarousal (correlations ranged from.37 to.56; mean r =.48) than with symptoms of agoraphobia (correlations ranged from.25 to.44; mean r =.32), social phobia (correlations ranged from.25 to.47; mean r =.33), and blood-injection phobia (correlations ranged from.17 to.31; mean r =.24). This pattern supports our second hypothesis and indicates a strong, specific link between anxiety sensitivity and symptoms of panic. Finally, our markers of agoraphobia tended to be more strongly related to symptoms of social phobia (correlations ranged from.35 to.62; mean r =.47) than to either physiological hyperarousal (correlations ranged from.18 to.33; mean r =.25) or anxiety sensitivity (correlations ranged from.25 to.44; mean r =.32). These results tentatively suggest that agoraphobia does not represent a panic-specific component; we will examine this issue further in our confirmatory factor analyses. A related finding is that our blood phobia markers tended to correlate more strongly with the agoraphobia scales (correlations ranged from.23 to 43; mean r =.32) than with markers of physiological hyperarousal (correlations ranged from.13 to.20; mean r =.16), anxiety sensitivity (correlations ranged from.17 to.31; mean r =.25), and social phobia (correlations ranged from.10 to.46; mean r =.25). Once again, we will examine this link between agoraphobia and blood phobia subsequently in our confirmatory factor analyses Confirmatory factor analyses To further clarify the nature of these associations, we next conducted confirmatory factor analyses. These analyses were conducted in EQS 6.1 (Bentler, 2004) using a covariance matrix and the maximum likelihood estimation method. The primary goal of these analyses was to examine the strength of the

12 S.L. Longley et al. / Anxiety Disorders 20 (2006) associations between latent factors representing Neuroticism, Physiological Hyperarousal, Anxiety Sensitivity, Agoraphobia, Social Phobia and Blood Phobia. We initially tested our proposed six model components; each factor was defined by a minimum of two indicators: (a) Neuroticism the BFI Neuroticism Scale and the PSWQ, (b) Physiological Hyperarousal the MASQ Anxious Arousal Scale and the PASQ, (c) Anxiety Sensitivity the ASI Physical Concerns subscale and the BSQ, (d) Agoraphobia the APPQ Agoraphobia, FQ Agoraphobia, and MIA Avoidance Alone Scales, (e) Social Phobia the APPQ Social Phobia, FQ Social Phobia, SCS Social Anxiety, and PSRS Social Phobia Scales and (f) Blood Phobia the FQ Blood-Injection Phobia and PSRS Blood- Injection Phobia Scales. We considered seven different fit indices in evaluating the adequacy of the models tested. Because the overall model chi-square is sensitive to large sample size, we also examined a number of other indices. These indices are: the Bentler- Bonett normed fit index (NFI), the comparative fit index (CFI), Bollen s incremental fit index (IFI), the goodness-of-fit index (GFI), the standardized-rootmean-square residual (SRMR) and the root mean square error of approximation (RMSEA). Although there are no strict criteria for evaluating these fit indices, conventional rule-of-thumb-guidelines suggest that fit is acceptable if (a) NFI, CFI, IFI, and GFI are.90 or greater and the (b) SRMR and RMSEA are.10 or less (see Finch & West, 1997; Hu & Bentler, 1998, 1999). However, Hu and Bentler (1999) have recommended more stringent cutoffs for several of these indices, suggesting values of.95 for CFI and IFI,.08 for SRMR, and.06 for RMSEA. In interpreting our results, we will consider NFI, CFI, IFI, and GFI values of.90 or greater to indicate an adequate fit, and values of.95 or greater to represent an excellent fit. SRMR and RMSEAvalues of.10 or less will be taken as reflecting an adequate fit, with values of.06 or less representing an excellent fit. Consistent with our conceptualization, we tested a model that consisted of six correlated factors. Each factor corresponded to one of the six model components, with each noted indicator defining that factor. The fit indices for our proposed six content components are presented in the top row of Table 2 ( Model A ). Taken together, these indices demonstrate that our Model A fit the data fairly well. Table 2 Fit indices for the tested models in the confirmatory factor analyses Model df x 2 Dx 2 NFI CFI IFI GFI SRMR RMSEA Model A Model B ** Note: See text for description of models. df: degrees of freedom, x 2 : chi-squared; Dx 2 : chi-squared difference; NFI: Bentler-Bonett normed fit index; CFI: comparative fit index; IFI: Bollen s incremental fit index; GFI: goodness-of-fit index; SRMR: standardized root mean-square residual; RMSEA: root mean-square error of approximation. N = 438. ** P <.01.

13 730 S.L. Longley et al. / Anxiety Disorders 20 (2006) Specifically, NFI, CFI, IFI, SRMR, and RMSEA all suggested an acceptable fit, whereas the GFI fell just short of an adequate fit. Although Model A generally appeared to be reasonable, our inspection of the standardized residual matrix indicated that the fit could be significantly improved by correlating the error terms that accounted for the strong relationships between the two APPQ Scales (APPQ Agoraphobia and APPQ Social Phobia, r =.62) and the two FQ Scales (FQ Agoraphobia and FQ Social Phobia; r =.61). This revised model ( Model B ) was identical to our Model A, except that we included these two correlated error terms. As expected, this Model B fit the data significantly better than Model A [Dx 2 = ; P <.01]. Moreover, as is shown in the second row of Table 2, all of the other fit indices now are in the acceptable range, with the SRMR indicating an excellent fit and the CFI, IFI, NFI, GFI, and RMSEA suggesting an adequate fit. We will, therefore, use the Model B for all subsequent analyses. Table 3 shows the factor loadings of the indicators in Model B. The most noteworthy aspect of this table is that each of the scales was a strong indicator of its target dimension. For example, the loadings were strong across the six content Table 3 Confirmatory factor analysis: Model B factor loadings Factor/marker/measure Factor I II III IV V VI Factor I: Neuroticism 1. BFI Neuroticism Penn State Worry Questionnaire (PSWQ).83 Factor II: Physiological Hyperarousal 3. MASQ Anxious Arousal Scale Panic Attack Symptoms Questionnaire (PASQ).91 Factor III: Anxiety Sensitivity 5. ASI Physical Concerns subscale Body Sensations Questionnaire (BSQ).79 Factor IV: Agoraphobia 7. FQ Agoraphobia APPQ Agoraphobia MIA Avoidance alone.72 Factor V: Social Phobia 10. APPQ Social Phobia FQ Social Phobia PPSC Social Anxiety PSRS Social Phobia.82 Factor VI: Blood Phobia 14. FQ Blood-Injection Phobia PSRS Blood-Injection Phobia.98 Note: N = 438.

14 factors that represent Neuroticism (range.83.89), Physiological Hyperarousal (range.83.91), Anxiety Sensitivity (range.79.91), Agoraphobia (range.72.81), Social Phobia (range.77.86) and Blood Phobia (.64.98). Thus, we see clear evidence of six well-defined content factors Factor intercorrelations S.L. Longley et al. / Anxiety Disorders 20 (2006) Table 4 reports the standardized correlations between the six content factors in Model B. Supporting our first hypothesis, Neuroticism emerged as a general, nonspecific predictor of anxiety-related symptoms. Indeed, Neuroticism had strong and very similar correlations with Physiological Hyperarousal (.58), Social Phobia (.56), Anxiety Sensitivity (.47), and Agoraphobia (.52). Also as predicted, Neuroticism had a more moderate association with the Blood Phobia factor (r =.25). Consistent with our second prediction, Physiological Hyperarousal and Anxiety Sensitivity can be identified as panic-specific components of the model. Thus, these two factors were very strongly linked (r =.64) in our data. Note, moreover, that the Anxiety Sensitivity factor correlated much more strongly with the Physiological Hyperarousal factor than with any other factor (correlations ranged from.35 to.47). It is particularly striking that this correlation of the Anxiety Sensitivity and the Physiological Hyperarousal factors was also stronger than with any of the phobia factors (e.g., Agoraphobia (r =.45), Social Phobia (r =.42) or Blood Phobia (r =.33)). These results further establish a specific link between anxiety sensitivity and symptoms of panic attacks. Finally, we did not make any specific predictions about agoraphobia. One very striking aspect of our data, however, was the very strong association (r =.70) between the Agoraphobia and Social Phobia factors. Indeed, Agoraphobia correlated much more strongly with Social Phobia than with either Physiological Hyperarousal (r =.38) or Anxiety Sensitivity (r =.45). It also is noteworthy that Blood Phobia was more strongly correlated with Agoraphobia (r =.48) than with any other factor (range r =.25.38). Thus, our data offer no support for the idea Table 4 Confirmatory factor analysis: intercorrelations of the six factor Factor/marker Factor I II III IV V VI Factor I: Neuroticism Factor II: Physiological Hyperarousal.58 Factor III: Anxiety Sensitivity Factor IV: Agoraphobia Factor V: Social Phobia Factor VI: Blood Phobia Note: N = 438.

15 732 S.L. Longley et al. / Anxiety Disorders 20 (2006) Table 5 Model B with constraints added to correlations between factors Model Equality constraints Model x 2 Dx 2 from Model df P-value Model B Model B No equality constraints Model 1 AG, SC, AG, AS <.005 and AG, PH Model 2 AG, SC and AG, PH <.005 Model 3 AG, SC and AG, AS <.005 Model 4 AG, BP, AG,AS <.025 and AG, PH Model 5 AG, BP and AG, PH <.025 Model 6 AG, BP and AG, AS <.01 Note:,: equality constraint imposed between two factors; AG: Agoraphobia factor; SC: Social Phobia factor; AS: Anxiety Sensitivity factor; PH: Physiological Hyperarousal factor; BP: Blood Phobia factor, N = 438. that agoraphobia can be specifically linked to panic attacks, as opposed to other types of anxiety Testing the significance of the Agoraphobia and Social Phobia correlations Next, we tested whether key latent factor correlations in our model differed significantly from one another. The top row of Table 5 displays the fit of the Model B depicted in Tables 2 4; this model previously has been specified without any equality constraints. We then examined whether the latent Agoraphobia factor correlated more strongly with Social Phobia than with Physiological Hyperarousal and Anxiety Sensitivity. We tested this by forcing these factor correlations to be equal and then examining the effect of these constraints on model fit (Long, 1983). In Model 1, we constrained all three of these correlations to be equal (i.e., we forced Agoraphobia to correlate equally with Social Phobia, Physiological Hyperarousal, and Anxiety Sensitivity). In comparison to Model B, with no equality constraints, Model 1 resulted in a significantly worse fit to the data. The difference in x 2 ( = ) for 2 df exceeds the critical value of at a =.005. We, therefore, rejected the null hypothesis that these three parameters were equal in the Model 1. Accordingly, the correlation between the latent Agoraphobia and Social Phobia factors is significantly greater than at least one of these other correlations. We then examined this issue in greater detail by comparing specific pairs of latent factor correlations (Long, 1983). In Model 2, we constrained the Agoraphobia Social Phobia correlation to be equal to the Agoraphobia Physiological Hyperarousal correlation; next, in Model 3, we forced it to be equal to the Agoraphobia Anxiety Sensitivity correlation. As is shown in the corresponding rows of Table 5, both of these constraints led to a significant

16 S.L. Longley et al. / Anxiety Disorders 20 (2006) decrement in model fit. Accordingly, we can conclude that Agoraphobia correlated more strongly with Social Phobia than with both Physiological Hyperarousal and Anxiety Sensitivity in our data Testing the significance of Agoraphobia and Specific Phobia correlations We also tested whether the strong correlation between the latent Agoraphobia and Blood Phobia factors was significantly greater than the correlations of the Agoraphobia factor with either Physiological Hyperarousal or Anxiety Sensitivity. Our general approach was identical to the analyses we outlined above (Long, 1983). In Model 4, we forced the Agoraphobia factor to correlate equally with Blood Phobia, Physiological Hyperarousal, and Anxiety Sensitivity. In comparison to the Model B, with no equality constraints, Model 4 resulted in a significantly worse fit to the data; specifically, the difference in the x 2 value ( = 8.308) for 2 df exceeded the critical value of at a =.025. Accordingly, the correlation between Agoraphobia and Blood Phobia is significantly greater than at least one of these other correlations. We then examined this issue in greater detail by comparing specific pairs of correlations. In Model 5, we constrained the Agoraphobia Blood Phobia correlation to be equal to the Agoraphobia Physiological Hyperarousal correlation. Then, in Model 6, we forced the Agoraphobia Blood Phobia correlation to be equal to the Agoraphobia Anxiety Sensitivity correlation. The results of these analyses are shown in the corresponding rows of Table 5, which indicates that both of these constraints led to a significant decrement in model fit. Accordingly, we can conclude that Agoraphobia correlated more strongly with Blood Phobia than with both Physiological Hyperarousal and Anxiety Sensitivity in our data. 4. Discussion Our structural analyses of data from self-report measures contributes to the larger body of research aimed at clarifying the taxonomy of anxiety. These analyses provide strong empirical support for the proposed six general and specific model components of panic and phobic anxiety. Consistent with prior theory and research, we demonstrated that neuroticism is the common underlying factor shared by all the specific components of panic and phobic anxiety. This was shown by the fact that the Neuroticism factor had strong and nearly equal correlations with the panic, social phobic and agoraphobic specific factors (see Table 4). As predicted, the Blood Phobia factor was more weakly related to the Neuroticism factor. Thus, neuroticism represents an underlying vulnerability factor that is shared by the anxiety disorders and the constructs assessed by our self-report measures (see also Krueger, 1999; Krueger et al., 2001; Mineka et al., 1998).

17 734 S.L. Longley et al. / Anxiety Disorders 20 (2006) Our results also are consistent with and extend existing research (e.g., Lilienfeld et al., 1993; Schmidt et al., 1997, 2000; Taylor, 1995) by establishing a specific link between anxiety sensitivity and the physiological symptoms of panic attack listed in DSM-IV. Specifically, our confirmatory factor analyses indicated that latent Anxiety Sensitivity factor correlated more strongly with Physiological Hyperarousal (r =.64) than with either Agoraphobia (r =.45), Social Phobia (r =.41), or Blood Phobia (r =.33). Thus, whereas, Neuroticism represents a general nonspecific component, physiological hyperarousal and anxiety sensitivity can be viewed as specific to panic attacks. Another goal of our study was to examine the link between panic attacks and agoraphobia. This investigation was motivated by the inconsistent findings that had been reported previously. Our review indicates that clinical and epidemiological studies have produced discrepant findings. On the one hand, studies with clinical samples rarely find agoraphobia without sub-threshold panic attacks (Barlow et al., 1994; Goisman et al., 1994; Lelliott, Marks, McNamee, & Tobena, 1989). In contrast, two community-based studies of young adults have found that respondents assessed with agoraphobia rarely report the occurrence of panic attacks (Hayward et al., 2003; Wittchen et al., 1998). Both of these studies used clinically trained interviewers and screened out social and specific phobia cases, thereby addressing concerns about the methods used in previous epidemiological studies (Barlow et al., 1994). In our study, the use of multiple symptom level indicators gave us a unique opportunity to investigate these relations using confirmatory factor analysis. Our method of using multiple self-report indicators is in keeping with the tradition of Zinbarg and Barlow (1996). This study, however, is unique in using CFA to focus exclusively on panic and phobic anxiety constructs. It is noteworthy, therefore, that we found no evidence of a specific link between panic attacks and agoraphobia. Indeed, Agoraphobia actually correlated more strongly with our Social Phobia (r =.70) and Blood Phobia (r =.48) factors than with Physiological Hyperarousal (r =.38) or Anxiety Sensitivity (r =.45). One potential explanation of these results is that our data reflect the limitations of self-report agoraphobia measures. This argument is not without merit, as some of our model indicators showed questionable discriminant validity (see Table 1). Most notably, Table 1 indicates that the APPQ and FQ Agoraphobia Scales were strongly related to the Social Phobia measures from these same instruments (correlations =.62 and.61, respectively). We were able to account for these strong within-instrument correlations in our analyses by correlating the relevant error terms in our measurement model. In light of these strong correlations, we encourage further work to improve the self-report assessment of agoraphobia, with the goal of enhancing the discriminant validity of these measures Implications of this research Our study contributes to an expanding area of research that links personality and psychopathology. This body of work has addressed important taxonometric

18 S.L. Longley et al. / Anxiety Disorders 20 (2006) issues and advanced an understanding of anxiety. In particular, these studies demonstrate the central role of neuroticism as a trait that contributes to all forms of anxiety and is a major source of the systematic comorbidity among such disorders (Mineka et al., 1998; Watson, 1999). Consistent with this conceptualization, in our data the Neuroticism factor had strong and nearly equal correlations with the panic and phobia factors, a notable exception was the correlation with the Blood Phobia factor. Taken together, these findings suggest that neuroticism is a common underlying component not only for the fear-related syndromes we studied, but for all forms of anxiety. Our research also suggests that the opportunity to better understand agoraphobia may have been prematurely curtailed when, in the DSM-IV, it was largely redefined in terms of panic. Indeed, our findings indicate that the association between panic attacks and agoraphobia merits reconsideration. Moreover, we emphasize that our results are broadly consistent with a growing body of work reviewed previously that has failed to demonstrate a specific link between panic attacks and agoraphobia. Based on these findings, it seems reasonable to conclude that this specific link has not yet been clearly established. Given our results, we believe that further studies of this link are badly needed. Our study also highlights the need to consider more carefully the position of blood-injection phobia within the anxiety taxonomy. Past structural analyses have considered blood-injection phobia to be one of the fear-related syndromes. It has been argued, however, that this phobia is uniquely defined by both disgust and fear and, as such, it should be considered separately from the other specific phobias (Thyer, Himle, & Curtis, 1985; Woody & Teachman, 2000). In this regard, our data show that the Neuroticism factor had a notably weaker correlation with the Blood Phobia factor (r =.25) than with the other specific model components (correlations ranged from.47 to.58). These differing relationships demonstrate that the Blood Phobia factor shares substantially less common variance with Neuroticism than the other fear-related syndromes we assessed. Based on these findings we suggest that the common practice of aggregating all specific phobia symptoms together should be reconsidered in future structural analyses Limitations of this research Finally, we would like to acknowledge some notable limitations of our research. First, the measures we used in our study do not strictly correspond to current DSM criteria. Accordingly, our results are based on self-report instruments and are not directly comparable to those based on the prevailing diagnostic criteria. For example, these self-report measures make no distinctions between the cued panic attacks associated with the phobias and the spontaneous panic attacks associated with panic disorder. Agoraphobia was defined in accord with the self-report measures we used, rather than as it is defined in the DSM-IV. Moreover, self-report measures rely on retrospective data and the respondents recollection of these events

19 736 S.L. Longley et al. / Anxiety Disorders 20 (2006) may be inaccurate. These issues may limit our ability to make definitive generalizations about the nature of the link between panic attack and agoraphobia. Second, generalizing our findings to other samples should be done cautiously. This cross-sectional study used an undergraduate sample and as is not atypical of such samples consists primarily of females and Caucasians. Moreover, our unselected, nonclinical sample may or may not yield findings comparable to those obtained from clinical samples. Clearly, in the future it will be important to replicate our model and findings using additional samples across a diverse array of participants. In conclusion, we have developed a model with six content factors to examine general and specific components within the panic domain. Although our study has the limitations we have noted, our work represents the first attempt to use multiple self-report indicators to conduct structural analyses to clarify the nature of the associations between panic and phobic anxiety components, exclusively. We did find a specific link between the panic attack components of physiological hyperarousal and anxiety sensitivity, but not between these components and agoraphobia. The six-factor model we examined in this study extends previous work in this area and provides a coherent theoretical/structural foundation upon which future research can build. References Achenbach, T. M. (2005). Empirically based assessment and taxonomy: application to clinical research. Psychological Assessment, 7, American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders. (4th ed.). Washington, DC: Author. Andrews, G., & Slade, T. (2002). Agoraphobia without a history of panic disorder may be a part of the panic disorder syndrome. Journal of Nervous & Mental Disease, 90, Barlow, D. H. (1988). Anxiety and its disorders: the nature and treatment of anxiety and panic. New York: Guilford Press. Barlow, D. H., Brown, T. A., & Craske, M. G. (1994). Definitions of panic attacks and panic disorder in the DSM-IV: implications of research. Journal of Abnormal Psychology, 103, Bentler, P. M. (2004). EQS 6 structural equation program manual. Encino, CA: Multivariate Software, Inc. Bouton, M., Mineka, S., & Barlow, D. H. (2001). A modern learning theory perspective on panic disorder. Psychological Review, 108, Brown, M., Smits, J., Powers, M. B., & Telch, M. J. (2003). Differential sensitivity of the three ASI factors in predicting panic disorder patients subjective and behavioral responses to hyperventilation challenge. Journal of the Anxiety Disorders, 17, Brown, T. A., Chorpita, B. F., & Barlow, D. H. (1998). Structural relationships among dimensions of the DSM-IV anxiety and mood disorders and dimensions of negative affect, positive affect and autonomic arousal. Journal of Abnormal Psychology, 107, Chambless, D. L., Caputo, G. C., Bright, P., & Gallagher, R. (1984). Assessment of fear in agoraphobics: the Body Sensation Questionnaire and the Agoraphobic Cognitions Questionnaire. Journal of Consulting and Clinical Psychology, 52, Chambless, D. L., Caputo, G. C., Jasin, S. E., Gracely, E. J., & Williams, C. (1985). The mobility inventory for Agoraphobia. Behaviour Research and Therapy, 23,