Introduction

The DSM-IV [1] classification includes clinical criteria for both panic attacks and panic disorder. The criteria for a diagnosis of panic attack are a discrete period of intense fear or discomfort, in which four or more out of thirteen (specified) panic symptoms (e.g., palpitations, sweating, trembling or shaking, feeling of choking) developed abruptly and reached a peak within ten minutes. Panic disorder requires (1) recurrent unexpected panic attacks and (2) at least one of the attacks has been followed by at least one month of persistent concern or worry about having panic attacks or its consequences and/or a significant change in behavior related to the attacks. Panic attacks may occur in the context of multiple anxiety disorders. They are considered to be amongst the most debilitating psychiatric conditions [2] and are associated with high level of mental health treatment seeking [3]. While it is known that early identification and subsequent intervention can reduce deleterious outcomes of psychiatric disorders [4], including panic disorder and panic attacks, research on risk factors for the onset of panic attacks is scarce. Some recent studies have identified panic attacks as a risk factor for other anxiety [4]–[6] and mood disorders [4], [5], independent of comorbid internalizing psychopathology. Although there is less consistency, there is some support for panic attacks to precede certain externalizing disorders, including alcohol use disorders [5], [7] and substance use [8], [9]. Hence, identifying predictors for the onset of panic attacks is an important research direction [7], [10].

The reported life-time prevalence of panic attacks when assessed by a clinical interview according to DSM criteria in youth samples ranges from 3.3% [4] to 11.6% [11]. This shows that the reported prevalence rate varies markedly across studies. Importantly, the lower prevalence rate was reported in a sample of 9–17 year olds, and the higher prevalence rate in a sample of 14–16 year olds, indicating that the prevalence rates increase with age in adolescent samples. Even higher lifetime prevalence rates are reported in studies using questionnaires instead of interviews to assess the DSM-criteria (21.4% [5] –63.3% [12]. Females typically have a higher prevalence than males [12], while no differences were found in socio-demographic characteristics between adolescents with and without panic attacks [4]. In addition, a meta-analysis [13] of the heritability of panic disorder revealed that genetic factors accounted for a large proportion of variance (43%). However, to our knowledge, no study has reported heritability for panic attacks.

As the typical age of onset of panic attacks is in late adolescence or early adulthood [14], with a peak between 15 and 19 years [15], it is crucial to examine prospective associations beginning in early adolescence. Identification of predictors of panic attacks early in development is also critical as earlier onsets are associated with increased rates of later psychopathology [16].

The few longitudinal studies on predictors of panic attacks have mostly focused on internalizing problems (emotional problems, e.g. anxiety, depression, other mood disorders). In a sample of high school students assessed over a 4-year period, negative affect [17], [18], anxiety sensitivity [5], [18], as well as separation anxiety disorder [17] were associated with an increased risk for panic attack onset in adolescents. Despite the fact that our knowledge of predictors for adolescent panic attacks is limited, no study to date has prospectively incorporated a broader range of problems, including externalizing problems (behavioral problems, e.g. conduct disorder, oppositional defiant disorder) as possible predictors. Besides internalizing problems, it is important to study other mental health problems as predictors of DSM-IV defined panic attacks since Roza et al. [19] found that both internalizing and externalizing problems in children and adolescents were predictive of anxiety disorders in young adulthood.

The aim of the current study is to extend the limited literature on predictors of panic attack onset, including both internalizing and externalizing problems in early adolescence. We hypothesized that a range of adolescent mental health problems including internalizing problems predict panic attack onset in adolescence. Furthermore, we explore if externalizing problems are part of these predictors. This study is part of the Tracking Adolescents’ Individual Lives Survey (TRAILS), a Dutch general population cohort study following 2,230 children from early adolescence (age 10–12 years) into young adulthood.

Methods

Results

Prevalence of panic attacks between T1 (age 10–12) and T4 (age 17–20) was 19.8%, and the prevalence of panic disorder 1.2%. Girls were significantly more likely than boys to have had at least one panic attack between T1 and T4 and individuals from a high SES were significantly less likely to have had at least one panic attack between T1 and T4 than individuals from a low or medium SES background (Table 1). The mean age of onset of panic attacks was 15.8 years (median 16 years) and the age of onset did not differ between girls and boys, F(1) = .837, p = .36.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sample characteristics and syndrome sum scores for the unsplit sample, and separately for subjects with and without panic attack onset between T1 and T4.

https://doi.org/10.1371/journal.pone.0051564.t001

Table 1 shows the sum scores of the CBCL and YSR scales at T1. Adolescents with at least one panic attack between T1 and T4 had higher scores for parent-rated Anxious/Depressed, Social Problems and Internalizing Problems and Somatic Complaints at age 10–12 years than adolescents without a panic attack. Adolescents with at least one panic attack between T1 and T4 had higher scores on all of the self-report YSR scales at T1 than adolescents without a panic attack.

Continuous-time survival analyses were performed to examine whether problems scores at age 10–12 years predicted first onsets of panic attacks during adolescence. The hazard ratios (HR) with 95% confidence interval (CI) adjusted for gender and SES are shown in Table 2 (parent-report) and Table 3 (self-report).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Results of univariate and multivariate models of survival analysis, predicting the onset of panic attacks with standardized CBCL scores and adjusted for gender and SES.

https://doi.org/10.1371/journal.pone.0051564.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Results of univariate and multivariate models of survival analysis, predicting the onset of panic attacks with standardized YSR scores and adjusted for gender and SES.

https://doi.org/10.1371/journal.pone.0051564.t003

The CBCL Anxious/Depressed, Social Problems, Somatic Complaints, Internalizing, and Total Problems scores significantly predicted panic attack onset between T1 and T4 (all ps<.01, HR range from 1.15–1.17). As the individual problem scales share variance, we performed multivariate analyses to identify the unique contribution of each CBCL syndrome scale. With all eight syndrome scales in one model, only parent-reported Social Problems (HR = 1.19, 95%CI = 1.01–1.41) at age 10–12 years independently predicted panic attack onset (p<.05). In a final model, we examined the broad-band Internalizing and Externalizing problem scores simultaneously. In this model, only Internalizing Problems predicted panic attack onset (HR = 1.15, 95%CI = 1.02–1.28, p<.05).

Each of the eight YSR syndrome scales, as well as the Internalizing, Externalizing and Total Problems scales significantly predicted panic attack onset between T1 and T4. When the eight subscales were entered into a multivariate model, only the Social Problems (HR = 1.26, 95%CI = 1.06–1.49, p<.01) and Thought Problems (HR = 1.15, 95%CI = 1.01–1.30, p<.05) at age 10–12 years scales independently predicted panic attack onset. When examining Internalizing and Externalizing Problems simultaneously, both Internalizing Problems (HR = 1.14, 95%CI = 1.01–1.30, p<.05) and Externalizing Problems remained significant (HR = 1.19, 95%CI = 1.04–1.36, p<.01).

Discussion

While having panic attacks increases risk for psychopathology, little is known about predictors of panic attacks. In this study, we measured panic attacks and tested the power of a broad range of both parent- and self-reported problems in 10 to 12 year-old children assessed with the CBCL and YSR to predict the onset of panic attacks across adolescence.

We found a life-time prevalence rate of panic attacks of 19.8%, which is higher than the few previous studies that have assessed panic attacks with a structured interview according to DSM criteria. We suspect that the age range of our study influenced this finding. The other population samples were younger, and lifetime prevalence increases relative to the age of the sample (prevalence 3.3% at age range 9–17 years [4]; prevalence 11.6% at age range 14–16 years [11]; prevalence 19.8% at age range 17–20 years). As the peak onset of panic attacks is between 15 and 19 years, it follows that the prevalence rate of our study sample is higher than in the previous studies.

The parent-reported CBCL scales Anxious/Depressed, Social Problems, Somatic Complaints, Internalizing, and Total Problems scores predicted first onset of panic attack in adolescence. Only the Social Problems and Somatic Complaints syndrome scale predicted panic attack onset when we controlled for shared variance of the scales. All eight self-reported YSR syndrome scales, as well as the Internalizing, Externalizing and Total Problems scales predicted the onset of panic attacks. After controlling for shared variance among the eight syndrome scales and among the Internalizing and Externalizing scales in the multivariate analyses, we found that the Social Problems, Thought Problems and Externalizing scales predicted adolescent panic attack onset.

At first sight, a broad range of problem perceptions are predictive of panic attack onset. This is consistent with Hayward et al.’s suggestion that there are different pathways that lead to adolescent panic attacks [17]; however, their focus was on pathways within the internalizing spectrum of problems. Our findings extend this beyond internalizing problems to include externalizing problems and also a more wide-ranging, general perception of problematic emotions and behavior. All types of problems reported by adolescents on the YSR were risk indicators of later panic attacks. Yet, when we look closer, only a few subscales were independent risk factors. Social Problems was the most consistent risk factor for panic attacks.

This cross-informant consistent result offers broader generalizability and greater theoretical significance than if derived from just one source. The Social Problems scale comprises items such as “acts too young for age,” “too dependent,” “does not get along with other kids,” “gets teased a lot,” “not liked by other kids,” “poorly coordinated or clumsy,” and “prefers being with younger kids”. The long-term consequences of problems with peer relations are supported by findings in other longitudinal research, among which TRAILS: Adolescents who reported to be bullied at age 10–12 years reported persistently higher scores of anxiety symptoms, including symptoms of panic over adolescence, regardless of the continuation of victimization at later ages [31]. In another sample, Roza et al. [19] reported a unique prospective relation between the CBCL Social Problems scale (4–16 years) and lifetime incidence of anxiety disorders (among which panic disorder) assessed 14 years later.

The association found between Social Problems and onset of panic attacks may be the result of a downward spiral starting with poor social skills and difficulties in peer relations, which in turn can lead to even lower self-confidence, and feelings of lack of control and helplessness. Yet, it may also be an expression of genetic transmission of vulnerability from parents to their children, or gene-environment interaction. An adolescent who reports panic attacks is likely to have parents who experience panic attacks or panic disorder [13], [32], [33]. Cross-sectional studies show that adults with panic attacks have problems with relations with others [34]–[36], which may lead to transmission of less developed social skills to their offspring, eventually resulting in social problems in children.

The current study has several strengths and weaknesses. The main strength is that we used a large adolescent sample representative of the general population of adolescents, covering the age range from 10 to 20 years. Also, our predictor variables were assessed prospectively, which avoids selective recall bias. Furthermore, we used information from multiple informants and we assessed panic attacks with strictly defined criteria. The main weakness of this study is that our data were censored at age 17–20 (T4) and consequentially, they did not extend through the entire peak risk period. Generalization of our findings to panic attack onset beyond adolescence will need to be confirmed in future studies. Second, we relied on a snap-shot assessment of clinical symptoms rather than relying on information about longitudinal course of symptoms. It is possible that incorporating symptom course may result in stronger effects. Lastly, our analyses focused on clinical scales to forecast risk for future panic attacks. Additional domains of risk, including social functioning, stress reactivity, anxiety sensitivity [37], and family history could be plausibly related to the onset of panic attacks or panic disorder. Future work in these areas could be interesting. Due to the novelty of our findings, there are limited direct implications for clinical interventions. In the context of prevention, however, there are some potential areas of clinical relevance. First, as associations are found using predictors in early adolescence, early assessment of these indicators can be important for effective preventative intervention. Second, predictors of panic attacks are not limited to the internalizing spectrum only. Assessment of risk for panic attacks is warranted among youth with externalizing problems as well. Thus, by recognizing early risk for panic attacks, problems associated with panic may be averted. Importantly, additional research on predictors of panic attacks is needed before specific clinical intervention recommendations can be formulated.

An important direction for future research is to focus on identifying pathways and mechanisms that explain how early Internalizing, Externalizing and Social Problems increase the risk of panic attack onset in adolescence. In this, an important first step is to identify specific aspects of Social Problems that predict panic attacks. Furthermore, information about panic attacks and panic disorder from first degree relatives should be included to advance understanding of how panic attacks may be transmitted across generations, including genetic and environmental means. Lastly, it is relevant to examine factors that predict and explain the progression from panic attacks to panic disorder. To conclude, we identified childhood Social Problems as consistent independent risk factor of adolescent panic attacks. In addition, we find that the risk markers of developing adolescent panic attacks are diverse and extend beyond problems in the internalizing spectrum to include externalizing problems and a broad-spectrum problems perception.

Acknowledgments

This research is part of the TRacking Adolescents’ Individual Lives Survey (TRAILS). Participating centers of TRAILS include various departments of the University Medical Center and University of Groningen, the Erasmus University Medical Center Rotterdam, the University of Utrecht, the Radboud Medical Center Nijmegen, and the Parnassia Bavo group, all in the Netherlands. We are grateful to all adolescents, their parents and teachers who participated in this research and to everyone who worked on this project and made it possible.