The authors have declared that no competing interests exist.
Conceived and designed the experiments: KDC DH DN WA EG KS. Performed the experiments: KDC DN WA. Analyzed the data: KDC DH DN WA KS. Contributed reagents/materials/analysis tools: KDC DH EG KS. Wrote the paper: KDC DH DN WA EG KS.
In cognitive theory it is hypothesized that panic attacks are provoked by catastrophic misinterpretations of bodily sensations. The aim of the present study was to investigate the ability of associated word pairs referring to catastrophic thinking (e.g. palpitations-heart attack) in producing panic attacks. Patients with PD (n = 20), patients with mixed anxiety disorders (n = 20), and a healthy control group (n = 30) participated in the present study. To enhance ecological validity we first conducted a stimulus validation experiment. Subsequently, nine suitable panic and neutral word pairs were presented in block to the participants. Anxiety levels were assessed before and after the presentation. PD patients were more anxious when reading these word pairs, compared to neutral word pairs. However, none of the participants experienced a panic attack upon reading the word pairs. From the present results it seems that catastrophic thinking is rather related to the anticipatory anxiety for panic attacks, but not necessarily with the occurrence of the panic attacks themselves.
Persons suffering from panic attacks often believe that they are having an acute medical condition and might die at any moment. Cognitive theories
The core hypotheses of this cognitive perspective can be summarized as follows:
Bias specificity: PD patients have specific catastrophic misinterpretations concerning their physical sensations.
Disorder specificity: This interpretative bias is assumed to be only specific for PD patients (in contrast with patients suffering from other anxiety disorders).
Causality: These specific catastrophic cognitions are regarded as sufficient and necessary in the production of panic attacks.
There is evidence in support of cognitive misinterpretation in PD patients. This includes the use of retrospective
Despite this rather large amount of studies, the pattern of findings only partially give support to the first two hypotheses concerning bias and disorder specificity.
The last premise of the cognitive perspective concerns the hypothesized causal role of specific catastrophic cognitions. It is hypothesized that catastrophic cognitions of bodily sensations are sufficient and necessary in evoking panic attacks in PD patients. However, to date, this has never been directly tested. There are some experimental studies which give an indication of more catastrophic misinterpretations during experimentally induced panic attacks in patients suffering from PD in contrast with controls
There is, however, one report of an uncontrolled experiment by Clark et al.
In sum, while there is evidence for catastrophic misinterpretation in PD patients, there is only equivocal support for bias and disorder specificity. Further, because of the lack of well controlled studies, the available evidence cannot entirely account for the proposed causal role of cognitions in the cognitive theory of panic attacks
The aim of the present study is to experimentally test the last hypotheses concerning the causal role of catastrophic misinterpretations of physical sensations in eliciting panic attacks in PD. We wanted to experimentally replicate the report of Clark et al.
Experiment 1 was designed to find suitable word pairs to use in our second Experiment.
The study was approved by the Maastricht University Medical Ethics Committee.
This experiment involved three groups of participants. In addition to the panic group, an anxious control group and a non-clinical control group were included in the design. Patients were recruited while seeking treatment at the Academic Anxiety Center in Maastricht. Healthy volunteers were recruited via advertisement. Inclusion criteria for all participants involved age between 18–70 years and good physical health. Exclusion criteria were the presence of dyslexia and a psychotic or dissociative disorder on axis I of DSM-IV criteria.
In the panic group, 21 patients (12 female) who met the DSM-IV criteria for PD were selected. Mean age was 42 years (
The ‘anxious control’ group consisted of 20 patients (14 female). Of these, 35% had obsessive compulsive disorder, 25% social phobia, 40% specific phobia as their primary diagnosis. Their mean age was 36 years (
The ‘healthy control’ group consisted of 30 participants (12 female). This group was matched with the panic group for sex and age. Participants had never had a psychiatric disorder. Mean age was 43 years (
The presence of possible mental disorders was evaluated by a trained clinician by means of the
Anxiety levels were measured with an
In the Paired Associated Task, two categories of words were used: 10 panic-related words and 10 neutral words. Eight threat words were derived from Clark et al.
In the pre-test screening, all questionnaires were administered after obtaining written informed consent.
In the test phase, to measure possible experimental anticipatory anxiety, participants were asked to score their anxiety level on the eVAS. After the procedure and the aim of the experiment was explained, the participants were asked to fill in the eVAS, PSL and POMS as pre-measurement.
The task was presented on a computer screen. Participants were given the instruction to read the word-pairs in silence. Word-pairs were presented separately with an inter-trial interval that ranged between 12 and 20 seconds. Presentation time was 8 seconds. The presentation order of the word pairs was randomized. After each word pair, participants were asked to score their anxiety level on the eVAS. At the end of the experiment, participants were asked again to score their anxiety level, to fill in the PSL and the POMS.
The eVAS, PSL and POMS scores were used as main outcome variables. The Kolmogorov-Smirnov test was used to test the normality of these variables. Not all variables were normally distributed. Hence, all analyses were performed with non-parametric tests. Therefore median ( =
Non-parametric Spearman correlations were used to analyze the relationship between depressive symptoms and anxiety levels.
Statistical significance for all analyses was accepted at a significance level of
As expected, the panic group had significantly higher scores on the PAS, as compared to the anxious control group,
gender | Age | MADRS | SDS | PAS | STAI-T | |
0.57 | 42 (12.6) | 12.7 (7.4) | 47 (8) | 24 (10.4)* | 53 (10.4) | |
0.7** | 36 (13.9) | 12.6 (11.4) | 43.1 (12) | 10 (10.2)* | 48.7 (14) | |
0.4 | 43 (11.3) |
Standard deviations are presented in brackets.
Note: MADRS = Montgomery-Asberg Depression Scale; SDS = Self-rating Depression Scale; PAS = Panic and Agoraphobia Scale; STAI-T = State Trait Anxiety Inventory-Trait *significant differences between the groups and between experiment 1&2 **;
While entering the experiment, panic patients were significantly more anxious than the anxious control patients and both patient groups were significantly more anxious than the healthy volunteers, χ2(3, N = 71) = 26.36,
eVASentr | Pre-eVAS | Pos-eVAS | Pre-PSL | Pos-PSL | Pre-POM | Pos-POM | |
19 | 22 | 35 | 6 | 6 | 32 | 38 | |
6 | 5 | 4 | 1 | 1 | 32 | 27 | |
0 | 0 | 0 | 0 | 0 | 7 | 5 |
eVAS ent = eVAS entrance, Pre = before start experiment, Pos = at the end of experiment.
All participants were significantly more anxious when reading panic-relevant word pairs in contrast with neutral word pairs, χ2(1, N = 71) = 45.56,
As expected, panic patients were significantly more anxious with respect to panic word pairs compared to anxious control patients,
Inspection of the anxiety levels of the neutral word pairs (see
Neutral word pairs for panic group, anxious controls and healthy controls in Experiment 1.
In the panic word pairs (see
Panic word pairs for panic group, anxious controls and healthy controls in Experiment 2.
To assess the relationship between depressive symptoms and anxiety levels on the word pairs correlations were calculated between MADRS scores and mean differences scores between panic and neutral word pairs within the patient groups. There were no significant correlations, nor for panic patients or anxious control patients,
Experiment 2 was designed to experimentally test the causal role of catastrophic misinterpretations in eliciting panic attacks. This experiment used a sample of participants that was independent of experiment 1.
The study was approved by the Maastricht University Medical Ethics Committee.
The study involved three groups of participants: a panic group, an anxious control group and a non-clinical control group. Patients were recruited while seeking treatment at the Academic Anxiety Center in Maastricht. Healthy volunteers were recruited via advertisement. Inclusion criteria for all participants involved age between 18–70 years and good physical health. Exclusion criteria were the presence of dyslexia and a psychotic or dissociative disorder on axis I of DSM-IV criteria.
In the panic group, 20 patients (11 female) were selected who met the DSM-IV criteria for PD. Mean age was 40.7 years (
The ‘anxious control’ group consisted of 20 patients (9 female). Of these, 30% had obsessive compulsive disorder, 25% social phobia, 35% specific phobia, 10% generalized anxiety disorder as their primary diagnosis. Their mean age was 41.3 years (
The ‘healthy control’ group consisted of 30 participants (17 female). This group was matched with the panic group for sex and age. Participants had never had a psychiatric disorder. Mean age was 38.3 years (
The same questionnaires were used as in Experiment 1.
In the Paired Associated Task, two categories of word pairs were used: 9 panic-related and 9 neutral word pairs. These 18 word pairs derived from Experiment 1. Nine panic-related word pairs were considered suitable for Experiment 2. Namely, these word pairs provoked more anxiety in panic patients in contrast with healthy controls and/or other anxious controls. Only ‘headache-brain tumor’ provoked no differences in anxiety between the patient groups and between the control groups. This word associate was removed for Experiment 2. Overall all neutral word pairs were suitable for experiment 2, with the exception of ‘purchase-shopping’. This word associate provoked some anxiety in panic patients and was therefore not appropriate as neutral word. The stimuli were presented in block on a personal computer. The words appeared in 12-cm block letters on a Philips colour monitor.
In the pre-test screening, all questionnaires were administered after obtaining written informed consent.
In the test phase, when entering the experiment, patients scored their anxiety level on the eVAS. After more information was given about the experiment participants were asked to fill in the PSL, POMS and eVAS. The task was presented on a computer screen. Before each block of word pairs the participants were asked to fill in the eVAS ( = pre-eVAS). Further, participants were given the instruction to read the word-pairs. Word-pairs were presented in block (instead separately) in line with the anecdotecal report of Clark et al.
First, the difference between pre- and post-eVAS was calculated for each block of word–pairs ( = delta-eVAS). The median delta-eVAS ( =
The pattern of results for the questionnaires (see
gender | Age | MADRS | SDS | PAS | STAI-T | |
0.55 | 41 (13.5) | 15 (7.6) | 47 (11.2) | 25 (10.3)* | 51 (10.7) | |
0.45 | 41 (15) | 10.6 (9.1) | 41 (10.6) | 4.5 (8)* | 46 (15) | |
0.6 | 38 (12.5) |
Standard deviations are presented in brackets.
Note: MADRS = Montgomery-Asberg Depression Scale; SDS = Self-rating Depression Scale; PAS = Panic and Agoraphobia Scale; STAI-T = State Trait Anxiety Inventory-Trait. *significant differences between the groups and between experiment 1&2 **;
While entering the experiment all patients were significantly more anxious than the healthy controls, χ2(3, N = 69) = 11.21,
eVAS-entr | Pre-eVAS | Pos-eVAS | Pre-PSL | Pos-PSL | Pre-POM | Pos-POM | |
19 | 23 | 18 | 3 | 1 | 28 | 29 | |
15 | 12 | 13 | 1 | 3 | 27 | 24 | |
1 | 1 | 0 | 0 | 0 | 9 | 9 |
eVAS ent = eVAS entrance, Pre = before start experiment, Pos = at the end of experiment.
There were significant group differences both for panic, χ2 (2, N = 70) = 14.5,
Overall delta-eVAS values were rather low and following Clark‘s definition of a panic attack as “the sudden increase in anxiety reaching at least 50 on a 100 point scale” (p155)
Correlations between MADRS scores and mean differences scores between panic and neutral word pairs within the patient groups were not significant different,
To enhance the ecological validity of our study we first conducted a stimulus validation experiment. The results of this led to the selection of a number word pairs that appeared to induce anxiety. Next, we presented these anxiety inducing word pairs repeatedly to panic disorder patients, under the hypothesis that they would trigger panic attacks. Indeed, PD patients were more anxious when reading these word pairs, compared to neutral word pairs. Furthermore, regardless of the presentation of word associates patients were also more anxious and reported more symptoms than control participants. Both these observations confirm the ecological validity of the experimental setup. However, none of the participants experienced a panic attack upon reading the word pairs. Second, when comparing the anxiety response between PD patients and participants with other anxiety disorders, no significant differences were found. These findings are in contrast with a previous report in which 83% of the PD patients were reported to have a panic attack after reading similar word pairs
To explain this discrepancy the following issues may be taken into account: (1) severity of PD, (2) interference of co-morbid depressive complaints, (3) sample size and (4) physical symptom reports. An accurate comparison is difficult however, since the previous report does not contain complete information in this regard. First, in the present study, panic diagnosis and severity was assessed using standardized scales as well as clinical diagnosis by structured interview. All PD patients had at least moderate to severe panic disorder, in clear contrast with the ‘anxious control’ group. Further, there were no differences in experimental anxiety, trait anxiety and depressive symptoms between the patients, indicating that the panic patients had substantially higher levels of symptomatology related to panic and agoraphobia. The overall disorder severity of both patient groups is comparable based on the scores on the STAI-T. Moreover the panic patients in the present study seem also very comparable with former panic studies. In a recent study by Kircher et al.
Some limitations of the present study need to be taken into account. First, it is important to acknowledge that one of the limitations of the cognitive model of panic is that it can be difficult to test. For instance, Roth, Willhelm and Petit
In sum, despite the use of carefully selected anxiety inducing word pairs, appropriate sample size, presence of an anxious control group and controlling for depressive symptoms, we were not able to find support for the ‘causality hypothesis’ in cognitive theory to explain the role of catastrophic cognitions in the occurrence of panic attacks. However, this does not mean that catastrophic thinking does not play a role in the broader concept of panic disorder, for this condition is characterized by anticipatory anxiety and avoidance behavior besides panic attacks. It is very likely that this kind of thinking plays a central role in the anticipatory anxiety that characterizes panic disorder patients. On the basis of phenomenological and neurobiological research Bouton et al
In conclusion, on the basis of our results it appears unlikely that catastrophic thinking has a central role in the immediate development of panic attacks. Anxiety can however contribute to the evolution of sporadic panic attacks towards PD through mechanisms such as interoceptive conditioning