pISSN 0513-5796   eISSN 1976-2437
Open Access, Peer-reviewed, Monthly
Yonsei Med J. 2025 May;66(5):277-288. English.
Published online Jan 16, 2025.
https://doi.org/10.3349/ymj.2024.0177
© Copyright: Yonsei University College of Medicine 2025
Original Article

A Comparison of Symptom Structure between Panic Disorder with and without Comorbid Agoraphobia Using Network Analysis

Joonbeom Kim,1,2,* Yumin Seo,1,3,* Seungryul Lee,1 Gayeon Lee,1 Jeong-Ho Seok,1,4 Hesun Erin Kim,1 view all
    • 1Institute of Behavioral Sciences in Medicine, Yonsei University College of Medicine, Seoul, Korea.
    • 2Department of Child Welfare, College of Human Ecology, Chungbuk National University, Cheongju, Korea.
    • 3Department of Integrative Medicine, Yonsei University College of Medicine, Seoul, Korea.
    • 4Department of Psychiatry, Gangnam Severance Hospital, Yonsei University College of Medicine, Seoul, Korea.
  • Corresponding author: Jooyoung Oh, MD, PhD. Department of Psychiatry, Gangnam Severance Hospital, Yonsei University College of Medicine, 211 Eonju-ro, Gangnam-gu, Seoul 06273, Korea. Email: ojuojuoju@yuhs.ac

    *Joonbeom Kim and Yumin Seo contributed equally to this work.
Received June 25, 2024; Revised October 07, 2024; Accepted October 08, 2024.

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (https://creativecommons.org/licenses/by-nc/4.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

AbstractGraphical AbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNotesACKNOWLEDGEMENTSDATA AVAILABILITYReferences

Abstract

Purpose

Panic disorder (PD) and PD with comorbid agoraphobia (PDA) share similar clinical characteristics but possess distinct symptom structures. However, studies specifically investigating the differences between PD and PDA are rare. Thus, the present study conducted a network analysis to examine the clinical networks of PD and PDA, focusing on panic symptom severity, anxiety sensitivity, anticipatory fear, and avoidance responses. By comparing the differences in network structures between PD and PDA, with the goal of identifying the central and bridge, we suggest clinical implications for the development of targeted interventions.

Materials and Methods

A total sample (n=147; 55 male, 92 female) was collected from the psychiatric outpatient clinic of the university hospital. We conducted network analysis to examine crucial nodes in the PD and PDA networks and compared the two networks to investigate disparities and similarities in symptom structure.

Results

The most influential node within the PD network was Anxiety Sensitivity Index-Revised (ASI-R1; fear of respiratory symptom), whereas Panic Disorder Severity Scale (PDSS5; phobic avoidance of physical sensations) had the highest influence in the PDA network. Additionally, bridge centrality estimates indicated that each of the two nodes met the criteria for “bridge nodes” within their respective networks: ASI-R1 (fear of respiratory symptom) and Albany Panic and Phobic Questionnaire (APPQ3; interoceptive fear) for the PD group, and PDSS5 (phobic avoidance of physical sensation) and APPQ1 (panic frequency) for the PDA group.

Conclusion

Although the network comparison test did not reveal statistical differences between the two networks, disparities in community structure, as well as central and bridging symptoms, were observed, suggesting the possibility of distinct etiologies and treatment targets for each group. The clinical implications derived from the similarities and differences between PD and PDA networks are discussed.

Keywords
Panic disorder; agoraphobia; network analysis; comorbidity

Panic disorder (PD) and agoraphobia are prevalent anxiety disorders, marked by distinct yet interrelated symptomatology.1, 2 PD manifests through maladaptive cognitive and behavioral changes following recurrent panic attacks (PAs), often dubbed “the most physical of mental disorders.”3 Agoraphobia, on the other hand, encompasses anxiety surrounding specific places or situations.4 Amid the emergence of coronavirus disease-2019 (COVID-19), these disorders have garnered unprecedented clinical attention. This heightened focus reflects augmented sensitivity to physical symptoms, particularly those linked to respiration, and the burgeoning fear of crowded or enclosed spaces since the onset of the pandemic. Notably, statistics from Korea revealed a 46.7% surge in PD patients and a staggering 167% increase in agoraphobia cases post-COVID-19.5

Against the backdrop of the burgeoning global burden of anxiety disorders, research on PD and agoraphobia is of paramount importance. With an estimated 28.68 million disability-adjusted life-years (DALYs) attributed to anxiety disorders worldwide in 2019, compared to 18.66 million in 1990, and an age-standardized DALYs rate of 3.60 per 1000 population in 2019, these disorders undeniably impose substantial health losses on a global scale.6 The recent surge in patient numbers and clinical attention towards PD and agoraphobia underscores the pressing need for accurate comprehension and treatment of these conditions, both from social and economic standpoints. Given the intricate interplay between these disorders, addressing this nexus has become even more pivotal.7

Previously, PAs played a central role in the development of agoraphobia until the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition era. Agoraphobia was delineated as fear or anxiety arising when escape might prove challenging during panic-like symptoms.8 However, the Diagnostic and Statistical Manual of Mental Disorders, Fifh Edition (DSM-5) broadened this scope, encompassing fear or anxiety in diverse situations, without explicit reference to PAs. Despite this shift, PD and agoraphobia remain closely intertwined, with symptoms frequently embedded in the diagnostic criteria of others.9, 10 The co-occurrence of PD and agoraphobia is associated with poor prognosis and treatment outcomes, including an increased likelihood of hospitalization and compromised social functioning.11, 12 Therefore, distinguishing PD with and without comorbid agoraphobia is clinically important but challenging.

PD and PD with comorbid agoraphobia (PDA) share common underlying features, such as anxiety sensitivity (AS), anticipatory fear, and avoidance responses, which are pivotal targets for therapeutic interventions.13, 14 AS is primarily considered the bridging link between understanding PAs and agoraphobia.15, 16, 17 AS, also known as the “fear of anxiety itself,” refers to the trait in which one believes that experiencing anxiety and its related symptoms leads to catastrophic outcomes. Compared to other anxiety-related disorders in the PD model, AS is a panicogenic factor that plays a particularly distinctive role in triggering PAs and maintaining PD.18 A high AS increases the probability of misappraisal of internal cues and habituates an exaggerated fear of daily physical sensations, such as changes in heart rate.13, 16, 19 In line with this mechanism, elevated AS has been theorized as a predisposing factor for the onset and worsening of anxiety-related disorders,20 and evidence has revealed that AS is a stable predictor of the development and maintenance of anxiety disorders, particularly PD.19 Thus, understanding AS is crucial for shedding light on the etiology of PD and agoraphobia.

Another key feature of PD and PDA is the anticipatory fear and avoidance responses.21 Phobic situations are not limited to mere exposure to external phobic situations (e.g., using public transportation, being in open/enclosed places, standing in line), but also include fear of interoceptive symptoms (e.g., nausea or abdominal discomfort, sweating, chills, or hot flushes) that produce somatic sensations.22 According to the generic cognitive model, the appraisal of interoceptive stimuli is erroneously regarded as a sign of an impending catastrophe associated with the development of PAs.23 Misinterpretation of minor respiratory disturbances caused by CO2 as oxygen deprivation can trigger panic symptoms and induce agoraphobic avoidance. This misinterpretation leads to increased fear and activates avoidance tactics against PAs, thereby reinforcing catastrophic cognition, which reaffirms the initial misinterpretation.22, 24 Therefore, assessing and understanding anticipatory fear and avoidance responses in phobic situations may have important implications in clinical approaches to PD and agoraphobia.

In recent years, the network analysis of psychopathology has emerged as a valuable tool for understanding the complex interplay of symptoms in psychiatric disorders.18, 25 The network refers to the complex web of symptoms containing “nodes” that stand for one symptom, while “edges” represent the partial correlation coefficient between two nodes,26 and larger relations form communities (i.e., clusters or dimensions) in the network. By identifying central nodes and bridge symptoms within the network, network analysis offers insights into the underlying mechanisms of psychiatric disorders and informs the development of targeted interventions.27, 28 In addition, bridge symptoms that link two or more clusters of symptoms together, thereby activating multiple clusters simultaneously, can also be identified, which contribute to the maintenance of comorbid symptoms; hence, targeting these symptoms may mitigate the comorbid state.29 There has been an increasing number of network analysis studies on PD, but most have focused on examining the relationship between PD and other psychiatric symptoms, such as depression.30 However, research exploring the subtle relationship between PD and PDA, each of which significantly affect the severity and prognosis of the disorder, remains scarce.

The present study aimed to use network analysis to explore the clinical network structure based on panic symptom severity, particularly focusing on AS, anticipatory fear, and avoidance responses. Additionally, it sought to compare the network structures between PD and PDA to identify central and bridge symptoms. These findings are expected to offer valuable insights for developing targeted interventions tailored to the specific characteristics of each disorder, ultimately improving the treatment efficacy and reducing the overall socioeconomic burden associated with these conditions.

Participants

To conduct a network analysis comparing PD and PDA, we extracted data from all patients who complained of panic-related symptoms during their first visit to a psychiatric outpatient clinic at a university hospital in South Korea (Gangnam Severance Hospital, Seoul, Korea) between September 2020 and February 2022. Patients were excluded based on the following criteria: 1) patients who failed to complete all clinical scales on panic and agoraphobia symptoms, and 2) patients who were not diagnosed with PD or PDA according to the DSM-5. Our data were obtained during routine clinical process, and since our study was retrospectively conducted, informed consent was waived. The study design and protocol were approved by the Institutional Review Board of Yonsei University Gangnam Severance Hospital (3-2022-0465).

Measures

Panic Disorder Severity Scale

Trained clinical psychologists measured PD severity using the Korean version of the Panic Disorder Severity Scale (PDSS),31, 32 rated by trained clinical psychologists. The scale comprises seven items: frequency of PAs, panic distress, anticipatory anxiety, agoraphobic fear/avoidance, interoceptive fear/avoidance, work impairment/distress, and social impairment/distress, with higher scores indicating higher levels of panic severity. Each item is rated on a 5-point scale (0=none, 4=extreme).

Anxiety Sensitivity Index-Revised

AS and its dimensions were measured using the Korean version of the Anxiety Sensitivity Index-Revised (ASI-R),33, 34 a self-reported scale in which respondents are asked to rate each item on a 5-point scale (0=none; 4=extreme). The scale is composed of 36 items that measure four subscales: fear of respiratory symptoms, fear of publicly observable anxiety reactions, fear of cardiovascular symptoms, and fear of cognitive dyscontrol. Higher scores indicate higher levels of AS.

Albany Panic and Phobic Questionnaire

Interoceptive, agoraphobic, and social situational fears were measured using the Korean version of the Albany Panic and Phobic Questionnaire (APPQ),33, 35 a self-reported scale in which respondents are asked to rate each item on a 9-point scale (0=none; 8=extreme). The scale consists of 27 items that measure the “fear of agoraphobic situations,” “fear of social situations,” and “fear of activities that produce somatic sensations,” with higher scores indicating higher levels of fear.

Statistical analyses

Descriptive and correlations

Descriptive analyses (frequency distributions) were performed to examine the prevalence and range of demographic and panic-related clinical symptoms (PDSS, ASI-R, and APPQ). Cross-tabulations were conducted to investigate the differences between PD and PDA using independent t-tests for continuous variables and chi-squared tests for categorical variables. Additionally, Cohen's d was calculated to interpret the effect size for the t-test, where values of 0.20, 0.50, and 0.80 represented small, medium, and large effect sizes, respectively. Next, Pearson correlation analysis was conducted to examine the associations between each symptom.

Network estimation techniques

Network analyses were conducted using the R software (v.4.2.2.; R Foundation for Statistical Computing, Vienna, Austria).36 The polychoric network for all nodes was calculated based on the Gaussian graphical model (GGM) with the graphic least absolute shrinkage and selection operator (LASSO) and the Extended Bayesian Information Criterion (EBIC) model using the qgraph package version 1.6.7. The qgraph package uses two estimators: the EBICglasso algorithm for regularized GGM estimation and the ggmModSelect algorithm for potentially non-regularized model research. Both estimations are primarily based on the graphical LASSO (glasso) algorithm in combination with EBIC model selection, EBICglasso estimation is a commonly used procedure for network analysis in psychopathological research.26 In this context, unregularized estimators are preferred for discovering individual edges (specificity) rather than focusing solely on the strongest edges (precision), particularly when the sample size is less than 300.37 This study examined the polychoric differences between PDSS, ASI-R, and APPQ symptoms in the psychopathology of PD and PDA status in terms of specificity. Therefore, we used a non-regularized ggmModSelect algorithm for calculations, as focusing on specificity may have clinical implications.

Community detection and bridge centrality

The network community structure and bridge nodes were evaluated using the EGAnet and networktools packages,29, 38 respectively. First, the Exploratory Graph Analysis (EGA) was conducted to detect communities in both PD and PDA. Communities are clusters of highly connected variables, equivalent to factors in the context of factor analysis.38 However, EGA produces more robust and accurate results than traditional latent factor analysis when based on edges.38 Thus, we used the Walktrap algorithm based on random walks, that is, steps from one node to another randomly chosen in the network, to calculate the modularity and explore the number of compositions of individual symptoms in each group. To estimate the median number of communities, 1000 bootstrap samples were generated using the bootEGA function. To quantify the importance of individual nodes in the network, one- and two-step expected influences (EI1 and EI2, respectively) were computed using the qgraph package. This approach is regarded as more accurate for calculating influential nodes than traditional centrality indices, such as strength, closeness, and betweenness.39 Furthermore, as EI1 does not consider information regarding the expected influence of the connected node, we only reported metrics as EI2 to account for both immediate and secondary influences on the network through its neighbors.39 Next, we identified important nodes that had the strongest connections and acted as bridges between detected communities by calculating the one- and two-step bridge expected influence (BEI1 and BEI2, respectively) using the bridge function.29 Edge-weight metrics from ggmModSelect were used to identify bridge nodes based on the centrality indices. The BEI is calculated by summing a node's edge weights, but only the edges that connect nodes from one community to another are counted. Tests for significant differences between the BEI values of the individual nodes were conducted using bootstrap analysis.

To further ensure the validity of centrality, a case-dropping bootstrap procedure was implemented using the bootnet package.40 The stability of all centralities was estimated by calculating correlation stability (CS) coefficients. The CS indicates the proportion of cases that can be dropped while maintaining a 95% probability that 10000 bootstrapped correlations between true and resampled centrality indices are greater than 0.25, with a preferred value above 0.5. The predictability, which reflects how well all neighboring nodes predict a specific node, was estimated using the mgm package.41

Network comparison

A network comparison test (NCT) was performed to investigate the differences in the network structures between PD and PDA using the NetworkComparisonTest package.41 NCT utilizes permutations to explore the differences in network structure between groups by testing invariance in global strength (the sum of all edge weights), network structure (e.g., distributions of edge weights), and specific edge values between two networks.

Sample characteristics and comparisons by group

A total of 709 outpatients were initially screened by psychiatrists and divided into PD and PDA groups according to the DSM-5. A total of 561 patients were excluded based on the criteria, leaving 147 patients included in the final analysis. Table 1 presents the characteristics of the participants and comparisons by group. Of the total sample, 62.6% (n=92) were female and the mean age was 36.3 years. In the PDA group, the majority of participants were female (74.6%, n=44), and the mean age (34.3±13.4 years) did not significantly differ between the two groups. The PDA group showed more severe symptoms compared to the PD group in almost all areas. APPQ1 (agoraphobia) had the largest effect size (Cohen's d=2.76) between the groups, followed by ASI-R2 (fear of publicly observable anxiety reaction), APPQ3 (interoceptive fear), and ASI-R4 (fear of cognitive dyscontrol), whereas neither PDSS1 (panic frequency) nor PDSS5 (phobic avoidance of physical sensation) showed statistical significance.

Table 1
Sample Characteristics and Comparisons of PDSS, ASI-R, and APPQ by Group

Correlations of the variables of interest

Bivariate correlation analysis was conducted for the PDSS, ASI-R, and APPQ. As illustrated in Fig. 1, all correlations were statistically significant in the predicted direction, except between ASI-R3 (fear of cardiovascular symptom) and PDSS4 (phobic avoidance of situations) (r=0.11, p>0.05). As shown in Fig. 1, ASI-R1 (fear of respiratory symptom) and ASI-R3 (fear of cardiovascular symptom) were highly correlated (r=0.73). However, none of the correlations exceeded the cut-off for multicollinearity (r>0.84).

Fig. 1
Correlations among variables of interests. All correlation coefficients were statistically significant in the positive direction, except for the correlation between ASI-R3 (fear of cardiovascular symptom) and PDSS4 (phobic avoidance of situations). PDSS, Panic Disorder Severity Scale; ASI-R, Anxiety Sensitivity Index-Revised; APPQ, Albany Panic and Phobic Questionnaire.

Network structure and centrality

The estimated network and detected communities based on the 13 clinical symptoms in each PD and PDA sample are presented in Table 2 and Fig. 2. The CS coefficients for the expected influence of both networks were 0.36 and 0.31, respectively, which were above the acceptable cut-off score of 0.25. Within the PD network, the edge weight between ASI-R1 and ASI-R3 (fear of respiratory symptom-fear of cardiovascular symptom) was the strongest, followed by edges APPQ1-APPQ2 (agoraphobia-social phobia), ASI-R3-APPQ3 (fear of cardiovascular symptom-interoceptive fear), PDSS4-PDSS5 (phobic avoidance of situations-phobic avoidance of physical sensation), and PDSS2-ASI-R1 (distress during panic-fear of respiratory symptom) (Fig. 3). In contrast, the edge between PDSS2-PDSS3 (distress during panic- panic-focused anticipatory anxiety) was the strongest in the PDA, followed by ASI-R2-APPQ2 (fear of publicly observable anxiety reaction-social phobia), ASI-R1-ASI-R3 (fear of respiratory symptom-fear of cardiovascular symptom), PDSS3-PDSS7 (panic-focused anticipatory anxiety-impairment in social functioning), and PDSS1-APPQ1 (panic frequency-agoraphobia). The node predictability index ranged from 31.7% to 77.2% and from 34.3% to 66.7% in each network, with an average of approximately 49.0%, accounting for half of the variance that could be explained by neighboring nodes.

Fig. 2
Network structure and expected influence of PDSS, ASI-R, and APPQ in panic disorder (A) and panic disorder comorbid with agoraphobia (B). Detected communities were represented with the same colors and shapes. Transparent blue dots represent the bridge symptoms that connect individual nodes and facilitate the spread of comorbidity. Blue line indicates a positive correlation, whereas red line indicates a negative correlation. PDSS, Panic Disorder Severity Scale; ASI-R, Anxiety Sensitivity Index-Revised; APPQ, Albany Panic and Phobic Questionnaire.

Fig. 3
Bootstrapped 95% confidence intervals of estimated edge weights for both panic disorder (A) and panic disorder comorbid with agoraphobia (B) in the estimated networks. Confidence intervals are represented by the gray area. A positive edge weight indicates a positive relationship, while a negative edge weight signifies a negative relationship. The absolute value of each edge weight represents the magnitude of the effect. PDSS, Panic Disorder Severity Scale; ASI-R, Anxiety Sensitivity Index-Revised; APPQ, Albany Panic and Phobic Questionnaire.

Table 2
Descriptive Statistics of Measurement Items by Group and Group Differences

The EI metrics were calculated for each network to determine the most influential nodes. The most influential node within the PD network was ASI-R1 (fear of respiratory symptom, EI=2.49), followed by PDSS2 (distress during panic, EI=2.26), and PDSS5 (phobic avoidance of physical sensation, EI=1.91). In contrast, PDSS5 (phobic avoidance of physical sensations, EI=2.29) had the greatest influence on the PDA network, followed by PDSS2 (distress during panic, EI=2.14) and ASI-R3 (fear of cardiovascular symptom, EI=1.98). Next, the CS coefficients for the bridge expected influence of both networks were 0.38 and 0.33, respectively, which were also above the acceptable cut-off score. Bridge centrality estimates using BEI2s indicated that each of two nodes had the highest BEI2s in their respective communities (Z-score ≥1) and were thus identified as bridge nodes in the individual networks; ASI-R1 (fear of respiratory symptom), APPQ3 (interoceptive fear), PDSS5 (phobic avoidance of physical sensation), and APPQ1 (panic frequency).

Community detection and comparison of PD versus PDA

The communities detected in each PD and PDA sample are shown in Fig. 2. In the PD network, 23 of the 91 possible edges (25.2%) were not zero, and the EGA revealed two communities. The first community comprised the ASI-R and the APPQ (“anxiety sensitivity with panic and phobia”), while the second community solely consisted of PDSS (“panic symptom severity”). In the PDA network, 21 of the 91 possible edges (23.1%) were not zero, and four dimensions with mixed PDSS, ASI-R, and APPQ scores were detected: 1) diagnostic symptoms of PD, 2) anticipatory fear, 3) avoidance behavior, and 4) social anxiety. The first dimension was composed mainly of diagnosis-related symptoms, such as panic frequency, anticipatory anxiety, and impairment in work functioning. The “anticipatory fear” included fear of interoceptive symptoms and cognitive dyscontrol, and the third “avoidance behavior” dimension comprised phobic avoidance of situations and physical sensations, impairment in social functioning, and the severity of agoraphobic symptoms. Lastly, “social anxiety” dimension comprised fear of publicly observable anxiety reaction and social phobia. When examining the statistical differences between networks using NCT, we found no significant differences in the overall network structure (network structure maximum=0.35, p=0.58) or global strength (global strength=5.86, p=0.52).

The current study examined the intricate relationships among panic-related symptoms in patients with PD and PDA using network analysis. While the two groups' network structures and overall global strength did not significantly differ, suggesting a strong similarity between the conditions, notable variations were observed in community structure. These differences could be important when considering the underlying mechanisms and tailoring clinical interventions for each condition. To ensure the robustness of these findings, highly reliable and validated psychometric scales were employed. Specifically, the Korean version of the ASI-R demonstrated excellent internal consistency (Cronbach's alpha: 0.967), the APPQ showed strong reliability (Cronbach's alpha: 0.920), and the PDSS presented an acceptable level of consistency (Cronbach's alpha: 0.871). The strong reliability of these measures further reinforces the validity of the observed network patterns and differences.

The PD network was composed of two communities. One community consisted solely of the PDSS, and the combined ASI-R and APPQ cluster formed the other community. In other words, the former focused on diagnosing PD and directly determining its severity through specific items, whereas the latter focused on items related to anxiety levels and avoidance behaviors. Within the symptom network structure of PD, fear of respiration (ASI-R1) and distress during panic (PDSS2) have emerged as key factors that maintain and aggravate PD symptoms, given their high EI measure. Moreover, PD severity was associated with respiratory fear (ASI-R1) and interoceptive fear (APPQ3). Symptoms of respiratory fear (ASI-R1) and interoceptive fear (APPQ3) emerged as potential risk factors contributing to the escalation of PD severity.

Furthermore, respiratory symptoms within the PD network were strongly correlated with other symptoms and played a central role in the manifestation of the disorder and network activation. Moreover, they demonstrated significant associations with panic severity symptoms within the community. Therefore, respiratory symptoms appeared to be pivotal indicators of PD. This result is consistent with the previous literature regarding the relationship between respiratory symptoms and other panic symptoms. Respiratory abnormalities, namely greater breath-to-breath variability, irregular respiratory frequency, excessive sighing, and tidal volume irregularity,42, 43 are common symptoms in patients with PD during PAs. The suffocation false-alarm theory (SFA) postulates that PAs occur when the brain misinterprets the amount of air available and sends a false alarm to a suffocation alarm system.44 According to SFA, PAs are the result of an oversensitive fear network that is closely related to the central nucleus of the amygdala (CNA).42 CNA increases respiratory rate, leading to an increased risk of false suffocation alarms that eventually trigger PAs.44 Thus, along with previous literature, our results also showed that managing respiratory symptoms is crucial for patients with PD, highlighting the role of breathing exercises and exposure therapy for PAs.

In contrast, community detection identified four clusters in the PDA network. More intricate communities within the PDA network may suggest that the symptoms observed in PDA are interconnected in a more complex manner, indicating the possibility of numerous comorbid pathways. This suggests potential heterogeneity between the two groups.45 One community mainly consisted of PD diagnostic symptoms, and the other three communities represented subsets of anxiety-related symptoms of PD and PDA: anticipatory fear, avoidance response, and social anxiety. The central symptoms of the PDA network were phobic avoidance of physical sensation (PDSS5) and distress during panic (PDSS2), whereas symptoms closely associated with the manifestation of other communities were agoraphobia (APPQ1) and phobic avoidance of physical sensation (PDSS5).

It is intriguing and worth noting that within the PDA network, PDSS5 emerges as a central and bridging symptom, despite not exhibiting significantly higher scores than other items in the PDSS. This suggests that severity, represented as an individual item score, may not be actively expressed by patients in clinical settings, potentially leading to oversight. Despite this, it displayed a stronger association with other indicators. Given its highest EI measure within the PDA network, avoidance behavior related to physical symptoms may play a primary role in exacerbating or maintaining a wide range of symptoms within PD, including symptoms of PAs, anticipatory anxiety, and social and occupational dysfunctions. Moreover, it provides a predictive link to the manifestation of symptoms such as anticipatory anxiety and impaired social functioning in other communities. Therefore, despite not attaining high scores on clinical assessments, it should be regarded as a significant marker that requires careful attention and targeting. This finding can be regarded as a valuable outcome that confirms the utility of this study and network analysis, highlighting its clinical implications.

These results showed that the PD and PDA networks share many common characteristics, suggesting underlying similarities in the factors contributing to PD and PDA. In both networks, despite having different structures, central and bridge symptoms were consistently observed within the anxiety-related communities. In other words, while PAs are the hallmark symptoms of PD, it is one's perspective and coping strategy, rather than the frequency or intensity of the attacks themselves, that determine the severity and maintenance of the disorder. This reaffirms the role of cognitive and behavioral coping mechanisms in the treatment of PD, specifically the development and implementation of these mechanisms through cognitive and behavioral therapy (CBT). CBT is an evidence-based treatment for PD and PDA, demonstrating superior therapeutic outcomes compared to other treatment approaches.46 It is a well-established therapy that yields successful treatment outcomes, whether administered individually or in group settings via the Internet or through self-guided interventions.

Moreover, in both groups, fear and distress experienced by patients during PAs contributed to the severity of the disorder and perpetuated other symptoms. The EI centrality of the nodes identified distress during panic (PDSS2) as the core symptom with a high centrality value in both groups. The average PDSS in the PD group was mild, whereas that in the PDA group was moderate. Regardless of the objective severity of symptoms, there is a possibility that the distress experienced subjectively by patients plays a significant role in the manifestation of the disorder.47, 48 Therefore, interventions that aim to alleviate the distress associated with panic attacks and help patients cope should take precedence in any disorder involving PAs. This finding further supports the rationale of previous studies that have examined the efficacy and effectiveness of CBT in PD. Particularly, it underscores the crucial role of basic techniques, such as breathing exercises, muscle relaxation, and interoceptive exposure, during the treatment process.

These findings further reinforce the significance and effectiveness of exposure therapy in treating PD and PDA, as they shed light on the interpretation of anxiety and coping strategies for distress regarding PAs. Exposure training, which involves exposing patients to anxiety-inducing stimuli to eliminate fear responses, is a fundamental aspect of panic treatment.3 Various techniques have been developed, ranging from exposure training that focuses on specific symptoms of PA to imagery exposure, in which patients imagine fearful situations in a therapy setting, and in vivo exposure therapy, which involves directly confronting feared real-life situations. Extensive research has been conducted over an extended period, providing evidence for the effectiveness of these techniques in treating PD and PDA.49

However, the findings of this study also indicate that distinctive characteristics may exist between the two clinical groups, suggesting a different intervention approach. First, our findings regarding EI measures suggest that the primary targets of PD and PDA in clinical intervention may differ. In our network model of PD, the fear of respiratory symptoms played a significant role in maintaining or exacerbating disorders and other symptoms. Therefore, among the various techniques comprising panic-CBT, mastering basic techniques related to bodily sensations, such as breathing and relaxation training, can be considered an essential and prioritized element of treatment based on our findings. However, in cases of PDA, the primary cause that perpetuates the disorder and causes distress for patients appears to be avoidance behavior resulting from the fear of these symptoms. Thus, in their treatment, it may be crucial to eliminate agora avoidance, as it is a prerequisite for achieving proper therapeutic effects. The results of this study suggest that PD and PDA can be classified as distinct symptom systems, aligning with previous research findings. According to a study,50 PD is more of a distress disorder, related to emotional disorders, whereas PDA is more of a fear disorder, related to phobias. This indicates that PD and PDA represent different psychopathological constructs. Therefore, the findings of the present study support the conceptualization of PD and PDA as separate diagnostic entities, reinforcing their independent classification in the DSM-5.

Second, the PDA network exhibited a more intricate symptom structure than that of the PDA network. This highlights the need for a multi-dimensional treatment approach.51 Instead of symptoms being highly interconnected, having multiple symptom structures may indicate that PDA symptoms are composed of more tightly connected subgroups, necessitating a symptom-based approach based on patterns exhibited by individuals. Third, the distinction between the PD and PDA networks was also evident in the fact that, unlike PD, the PDA network exhibited a division within the panic symptom distress scale. Notably, agoraphobia was related to phobic avoidance of situations and physical sensations (PDSS4 & 5) along with impairment in social functioning (PDSS7). The emergence of PDSS5 (phobic avoidance of physical sensations) as a key node in PDA populations distinct from the PD network indicates the chronic development of sensation-related fear. This chronic fear contributes to notable impairment across different aspects of life, as evidenced by the fact that PDSS7 is part of the same symptoms' community, in which nodes specifically address social functioning impairment. This community structure implies that the PDA group exhibited a more pronounced impairment in social functioning attributed to phobic avoidance compared to the PD group. As social functioning is related to various domains of life, individuals with PDA may experience more severe impairments than those with PD.52 Moreover, earlier research indicated that phobic avoidance stands out as the most reliable predictor of reduced improvement during PDA treatment.53 Consequently, our study underscores the importance of monitoring phobic avoidance and social functioning as crucial targets in the treatment of PDA.

This study examined the characteristics of PD and PDA symptom structures and compared them using a network analysis. These results demonstrated the utility of node- and symptom-centric network analyses in understanding and treating mental disorders. Rather than understanding PDA as a result of merely adding agoraphobic features to PD symptoms, our results shed light on more intricate and interactive dynamics among the symptoms. These findings suggest that treatment tailored to the specific symptoms of the disorder could have a significant impact on treatment outcomes.

Furthermore, the symptom-based approach to treating PD emphasizes the need for individualized and evidence-based approaches. However, the application of these techniques to patients requires specialized professionals, which may limit their time and cost-effectiveness. Recently, with advancements in virtual reality (VR) technology, a promising alternative that can address these limitations and deliver evidence-based therapy has emerged. Virtual reality exposure therapy (VRET) has been successfully utilized in the clinical context and has been found to have an efficacy comparable to that of in vivo exposure. Given that the flexibility and scalability of VR content could integrate diverse forms of content, such as psychoeducation, interoceptive exposure, and other modularized contents, the CBT format across VRET enables consistency in the delivery of evidence-based and personalized treatment.

Despite its novel findings and clinical implications, this study had several limitations. First, due to the relatively small sample size, the results should be interpreted with caution. There may have been a loss of statistical power, which often leads to difficulties in finding true differences in centrality and in estimating more stable and accurate networks.40 Moreover, the small sample size may have influenced the results of the NCTs. Thus, future studies should analyze data from a larger sample to estimate more stable networks and examine whether the network structures are different. Furthermore, there are a few additional suggestions that consider the characteristics of the PD and PDA samples. Given that the prevalence rates of PD and agoraphobia are high in women,54 future studies should compare the network structures of PD and PDA symptoms according to sex. Although this study did not include depressive symptoms as a measure, PD is known to be highly comorbid with depressive symptoms, in severe cases. To delineate the symptom-level relationship between depression, panic symptoms, and agoraphobic symptoms, it is highly recommended to add depressive symptoms to PD and PDA networks in future studies. Second, this study was cross-sectional in nature; therefore, a causal relationship between the symptoms could not be assumed. We discovered the most influential nodes in each network; therefore, future studies are needed not only to confirm whether the results can be replicated but also to determine how they influence other nodes using longitudinal data. In addition, a previous study that analyzed a longitudinal data network indicated that their results demonstrated within-person differences.55 Hence, a longitudinal design may help clinicians understand the changes in the pattern of each individual's symptom structure, thereby allowing for more precise individualized treatment. Lastly, one notable limitation of this study is the inability to conduct sub-analyses on gender differences within each group using NCT. Due to the relatively small sample sizes in the subgroups (PD male: 40, PD female: 48, PDA male: 15, PDA female: 44), we encountered convergence issues in the network models, particularly within the smaller PDA subgroups. As a result, we were unable to assess potential gender differences in network characteristics, such as centrality and bridge symptoms. Future research with larger and more balanced samples will be necessary to explore these differences in greater detail.

In conclusion, this study examined panic-related symptoms in patients with PD and PDA using network analysis. The two groups had different structures, as well as central and bridge symptoms, suggesting that adopting different approaches for each disorder could be beneficial and may be required in a clinical setting. The findings also revealed similarities in the factors contributing to PD and PDA, highlighting the role of cognitive and behavioral coping mechanisms in treatment. Distress during panic was identified as the core symptom in both groups. These findings support the effectiveness of exposure therapy and the importance of addressing avoidance behaviors. Personalized treatment and VR technology have been suggested as potential approaches for improving treatment outcomes.

AbstractGraphical AbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNotesACKNOWLEDGEMENTSDATA AVAILABILITYReferences
Notes

The authors have no potential conflicts of interest to disclose.

AUTHOR CONTRIBUTIONS:

  • Conceptualization: Joonbeom Kim, Yumin Seo, and Jooyoung Oh.

  • Data curation: Hesun Erin Kim and Joonbeom Kim.

  • Formal analysis: Joonbeom Kim and Jooyoung Oh.

  • Funding acquisition: Jooyoung Oh and Jeong-Ho Seok.

  • Investigation: Yumin Seo and Hesun Erin Kim.

  • Methodology: Joonbeom Kim, Yumin Seo, and Gayeon Lee.

  • Project administration: Jooyoung Oh.

  • Resources: Hesun Erin Kim and Seungryul Lee.

  • Software: Joonbeom Kim.

  • Supervision: Jooyoung Oh and Jeong-Ho Seok.

  • Validation: Yumin Seo, Seungryul Lee, and Gayeon Lee.

  • Visualization: Joonbeom Kim.

  • Writing—original draft: Joonbeom Kim and Yumin Seo.

  • Writing—review & editing: all authors.

  • Approval of final manuscript: all authors.

The authors would like to express our deepest gratitude to Professor Jae-Jin Kim and Eunjoo Kim for their support of this paper.

This work was supported by the Korea Medical Device Development Fund grant funded by the Korean government (the Ministry of Science and ICT, the Ministry of Trade, Industry and Energy, the Ministry of Health & Welfare, the Ministry of Food and Drug Safety) (Project Number: 1711174512, RS-2022-00140408).

The data that support the findings of this study are available from Gangnam Severance Hospital, but restrictions apply to their availability. These data were used under license for the current study and are not publicly available.

AbstractGraphical AbstractINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONNotesACKNOWLEDGEMENTSDATA AVAILABILITYReferences

References

    1. Bandelow B, Michaelis S. Epidemiology of anxiety disorders in the 21st century. Dialogues Clin Neurosci 2015;17:327–335.
    1. Martin P. The epidemiology of anxiety disorders: a review. Dialogues Clin Neurosci 2003;5:281–298.
    1. Barlow DH. In: Anxiety and its disorders: the nature and treatment of anxiety and panic. 2nd ed. New York: Guilford Press; 2002.
    1. Association Psychiatric Association. Diagnostic and statistical manual of mental disorders (DSM-5). 5th ed. Arlington(VA): American Psychiatric Association Publishing; 2013.
    1. Health Insurance Review & Assessment Service. Analysis of depression and anxiety disorder clinical statistics over the last five years (2017–2021). [accessed on 2024 April 12].
      Available at: https://www.hira.or.kr/bbsDummy.do?pgmid=HIRAA020041000100&brdScnBltNo=4&brdBltNo=10627&pageIndex=1.
    1. Yang X, Fang Y, Chen H, Zhang T, Yin X, Man J, et al. Global, regional and national burden of anxiety disorders from 1990 to 2019: results from the global burden of disease study 2019. Epidemiol Psychiatr Sci 2021;30:e36
    1. Inoue K, Kaiya H, Hara N, Okazaki Y. A discussion of various aspects of panic disorder depending on presence or absence of agoraphobia. Compr Psychiatry 2016;69:132–135.
    1. Asmundson GJ, Taylor S, Smits JA. Panic disorder and agoraphobia: an overview and commentary on DSM-5 changes. Depress Anxiety 2014;31:480–486.
    1. Craske MG, Simos G. Panic disorder and agoraphobia. In: Simos G, Hofmann SG, editors. CBT for anxiety disorders: a practitioner book. 1st ed. Chichester: John Wiley & Sons; 2013. pp. 3-24.
    1. Starcevic V, Uhlenhuth EH, Kellner R, Pathak D. Patterns of comorbidity in panic disorder and agoraphobia. Psychiatry Res 1992;42:171–183.
    1. Barlow DH, DiNardo PA. The nature of anxiety and panic disorders. In: Barlow DH, editor. Anxiety and its disorders. 2nd ed. New York: Guilford Press; 1991. pp. 390-417.
    1. Bruce SE, Yonkers KA, Otto MW, Eisen JL, Weisberg RB, Pagano M, et al. Influence of psychiatric comorbidity on recovery and recurrence in generalized anxiety disorder, social phobia, and panic disorder: a 12-year prospective study. Am J Psychiatry 2005;162:1179–1187.
    1. Clark DM. A cognitive approach to panic. Behav Res Ther 1986;24:461–470.
    1. McNally RJ. Anxiety sensitivity and panic disorder. Biol Psychiatry 2002;52:938–946.
    1. Kang EH, Kim B, Choe AY, Lee JY, Choi TK, Lee SH. Panic disorder and health-related quality of life: the predictive roles of anxiety sensitivity and trait anxiety. Psychiatry Res 2015;225:157–163.
    1. Naragon-Gainey K. Meta-analysis of the relations of anxiety sensitivity to the depressive and anxiety disorders. Psychol Bull 2010;136:128–150.
    1. Shin J, Park DH, Ryu SH, Ha JH, Kim SM, Jeon HJ. Clinical implications of agoraphobia in patients with panic disorder. Medicine (Baltimore) 2020;99:e21414
    1. Cha EJ, Hong S, Park DH, Ryu SH, Ha JH, Jeon HJ. A network analysis of panic symptoms in relation to depression and anxiety sensitivity in patients with panic disorder. J Affect Disord 2022;308:134–140.
    1. Taylor S. Anxiety sensitivity. In: Abramowitz JS, Blakey SM, editors. Clinical handbook of fear and anxiety: maintenance processes and treatment mechanisms. Washington, DC: American Psychological Association; 2020. pp. 65-80.
    1. Asnaani A, Tyler J, McCann J, Brown L, Zang Y. Anxiety sensitivity and emotion regulation as mechanisms of successful CBT outcome for anxiety-related disorders in a naturalistic treatment setting. J Affect Disord 2020;267:86–95.
    1. White KS, Brown TA, Somers TJ, Barlow DH. Avoidance behavior in panic disorder: the moderating influence of perceived control. Behav Res Ther 2006;44:147–157.
    1. De Cort K, Hermans D, Noortman D, Arends W, Griez EJ, Schruers KR. The weight of cognitions in panic: the link between misinterpretations and panic attacks. PLoS One 2013;8:e70315
    1. Beck AT, Haigh EA. Advances in cognitive theory and therapy: the generic cognitive model. Annu Rev Clin Psychol 2014;10:1–24.
    1. Ohst B, Tuschen-Caffier B. Catastrophic misinterpretation of bodily sensations and external events in panic disorder, other anxiety disorders, and healthy subjects: a systematic review and meta-analysis. PLoS One 2018;13:e0194493
    1. Borsboom D, Cramer AO. Network analysis: an integrative approach to the structure of psychopathology. Annu Rev Clin Psychol 2013;9:91–121.
    1. Epskamp S, Cramer AO, Waldorp LJ, Schmittmann VD, Borsboom D. qgraph: network visualizations of relationships in psychometric data. J Stat Softw 2012;48:1–18.
    1. Fried EI, van Borkulo CD, Cramer AO, Boschloo L, Schoevers RA, Borsboom D. Mental disorders as networks of problems: a review of recent insights. Soc Psychiatry Psychiatr Epidemiol 2017;52:1–10.
    1. Vierl L, Juen F, Benecke C, Hörz-Sagstetter S. Exploring the associations between psychodynamic constructs and psychopathology: a network approach. Personal Ment Health 2023;17:40–54.
    1. Jones PJ, Ma R, McNally RJ. Bridge centrality: a network approach to understanding comorbidity. Multivariate Behav Res 2021;56:353–367.
    1. Kim HJ, Lee SH, Pae C. Gender differences in anxiety and depressive symptomatology determined by network analysis in panic disorder. J Affect Disord 2023;337:94–103.
    1. Kim JB. [Development of the Korean version of the panic disorder severity scale]. Korean J Psychopathol 2001;10:140–151.
      Korean.
    1. Shear MK, Brown TA, Barlow DH, Money R, Sholomskas DE, Woods SW, et al. Multicenter collaborative panic disorder severity scale. Am J Psychiatry 1997;154:1571–1575.
    1. Kim JH, Yu BH, Oh KS, Yang JC, Kim Y, Lee SY, et al. [A validation study of Korean anxiety sensitivity index-revised (ASI-R)]. J Korean Neuropsychiatr Assoc 2004;43:54–61.
      Korean.
    1. Taylor S, Cox BJ. Anxiety sensitivity: multiple dimensions and hierarchic structure. Behav Res Ther 1998;36:37–51.
    1. Rapee RM, Craske MG, Barlow DH. Assessment instrument for panic disorder that includes fear of sensation-producing activities: the Albany panic and phobia questionnaire. Anxiety 1994-1995;1:114–122.
    1. R Core Team. R: a language and environment for statistical computing; 2020 [Software]. [accessed on 2024 March 6].
      Available at: https://cran.r-project.org/bin/windows/base/old/.
    1. Isvoranu AM, Epskamp S. Which estimation method to choose in network psychometrics? Deriving guidelines for applied researchers. Psychol Methods 2023;28:925–946.
    1. Golino HF, Epskamp S. Exploratory graph analysis: a new approach for estimating the number of dimensions in psychological research. PLoS One 2017;12:e0174035
    1. Robinaugh DJ, Millner AJ, McNally RJ. Identifying highly influential nodes in the complicated grief network. J Abnorm Psychol 2016;125:747–757.
    1. Epskamp S, Borsboom D, Fried EI. Estimating psychological networks and their accuracy: a tutorial paper. Behav Res Methods 2018;50:195–212.
    1. van Borkulo CD, van Bork R, Boschloo L, Kossakowski JJ, Tio P, Schoevers RA, et al. Comparing network structures on three aspects: a permutation test. Psychol Methods 2023;28:1273–1285.
    1. Nardi AE, Freire RC, Zin WA. Panic disorder and control of breathing. Respir Physiol Neurobiol 2009;167:133–143.
    1. Wilhelm FH, Gevirtz R, Roth WT. Respiratory dysregulation in anxiety, functional cardiac, and pain disorders: assessment, phenomenology, and treatment. Behav Modif 2001;25:513–545.
    1. Klein DF. Testing the suffocation false alarm theory of panic disorder. Anxiety 1994;1:1–7.
    1. Javed MA, Younis MS, Latif S, Qadir J, Baig A. Community detection in networks: a multidisciplinary review. J Netw Comput Appl 2018;108:87–111.
    1. Gould RA, Ott MW, Pollack MH. A meta-analysis of treatment outcome for panic disorder. Clin Psychol Rev 1995;15:819–844.
    1. Casey LM, Oei TP, Newcombe PA. An integrated cognitive model of panic disorder: the role of positive and negative cognitions. Clin Psychol Rev 2004;24:529–555.
    1. Story TJ, Craske MG. Responses to false physiological feedback in individuals with panic attacks and elevated anxiety sensitivity. Behav Res Ther 2008;46:1001–1008.
    1. Clum GA, Clum GA, Surls R. A meta-analysis of treatments for panic disorder. J Consult Clin Psychol 1993;61:317–326.
    1. Greene AL, Eaton NR. Panic disorder and agoraphobia: a direct comparison of their multivariate comorbidity patterns. J Affect Disord 2016;190:75–83.
    1. Christensen AP, Golino H. Estimating the stability of psychological dimensions via bootstrap exploratory graph analysis: a Monte Carlo simulation and tutorial. Psych 2021;3:479–500.
    1. Eguchi M, Noda Y, Nakano Y, Kanai T, Yamamoto I, Watanabe N, et al. Quality of life and social role functioning in Japanese patients with panic disorder. J Nerv Ment Dis 2005;193:686–689.
    1. Porter E, Chambless DL. A systematic review of predictors and moderators of improvement in cognitive-behavioral therapy for panic disorder and agoraphobia. Clin Psychol Rev 2015;42:179–192.
    1. Tello AB, Yonkers KA. Gender differences in the clinical courses of panic disorder. Psychiatr Times 2004;4:18
    1. Chambon M, Kammeraad WG, van Harreveld F, Dalege J, Elberse JE, van der Maas HLJ. Understanding change in COVID-19 vaccination intention with network analysis of longitudinal data from Dutch adults. NPJ Vaccines 2022;7:114
Publication Types
Original Article
MeSH Terms
Agoraphobia*
Ambulatory Care Facilities
Anxiety
Comorbidity*
Female
Humans
Male
Panic
Panic Disorder*
Sensation
Metrics
Share
Figures
slide
slide
slide

1 / 3

Tables
slide
slide

1 / 2

ORCID IDs
Funding Information
PERMALINK