Do fear and catastrophizing about mental activities relate to fear-avoidance behavior in a community sample? An experimental study

ABSTRACT

Introduction: Healthy people often experience headache, cognitive failures, or mental fatigue. Some people even experience these symptoms on a level comparable to patients with mild spectrum brain injuries. In these individuals, the fear-avoidance model explains symptoms as a consequence of catastrophizing and fear-avoidance toward mental activities. This experimental study investigated in healthy adults whether fear-avoidance and catastrophizing about mental activities are related to fear-avoidance behavior (i.e., behavioral avoidance of mental activities) according to the fear-avoidance model.

Method: A randomized crossover within-subject design was used with two measurements and 80 participants. Participants were exposed to three demanding cognitive tasks and their simplified counterparts. Post-concussion symptoms, catastrophizing, fear-avoidance, behavioral avoidance (time spent working on cognitive tasks), exposure to mental activity, depression, heart rate, and state-trait anxiety were assessed.

Results: Significant correlations between the variables of the fear-avoidance model were found. Furthermore, catastrophizers spent less time on difficult tasks compared to easy tasks. Both catastrophizing and female sex predicted time spent on difficult tasks, whereas only female sex predicted time spent on easy tasks.

Conclusions: This study found that, according to the fear-avoidance model, catastrophizing is related to behavioral avoidance of cognitively challenging tasks in a community sample.

KEYWORDS:

• Post-concussion-like symptoms
• fear-avoidance model
• catastrophizing brain injury
• experiment

Introduction

Healthy adults often report cognitive, emotional, somatic, and behavioral complaints such as headaches, memory problems, depressed mood, apathy, and fatigue (Aaronson et al., 2003; Lee et al., 2016; Wong et al., 1994). They are not diagnosed with any physical or mental disease and therefore are considered “healthy.” However, the symptoms they report can reach a degree comparable to patients recovering from mild traumatic brain injury (mTBI) or concussion (Asken et al., 2017; Dean et al., 2012). Dean et al. (2012) named this symptom complex in healthy adults “post-concussion-like symptoms,” which shows similarities with cogniform disorder (a subtype of somatoform disorders) (Delis & Wetter, 2007). Several studies have shown the incidence of post-concussion-like symptoms in the general population (Iverson & Lange, 2003; Voormolen et al., 2019). These symptoms are influenced by cultural and sex differences (Wang et al., 2006; Zakzanis & Yeung, 2011).

Looking at an explanation for post-concussion symptoms in mTBI, prior researchers have stated that biological (e.g., injury-related characteristics such as injury severity) or psychosocial causes (e.g., personal characteristics such as history of psychological treatment) seem to be important, but fail to explain this symptom persistence on their own, and thus have suggested an integrated biopsychosocial approach for future studies (Ponsford, 2017; Scheenen et al., 2016; Silverberg et al., 2015; Theadom et al., 2016; Wäljas et al., 2015). Moreover, one of the most consistent findings has been that pre-injury mental health and early post-injury anxiety are important predictors of persistent post-concussion symptoms (Cassidy et al., 2014; Silverberg et al., 2015). Therefore, we wanted to investigate a theoretical biopsychosocial model, combining biological and psychosocial causes and centralizing the role of anxiety, known as the fear-avoidance model. This model was first developed to understand chronic pain (Vlaeyen et al., 2016). The model has been adapted and applied to various complaints in different patient populations (Cima et al., 2012; Wijenberg et al., 2020; Vlaeyen et al., 2016; Wijenberg et al., 2016), including patients who experienced mild injury with cognitive deficits as a consequence, such as post-concussion symptoms following mTBI (Wijenberg et al., 2020; Wijenberg et al., 2017). Numerous studies confirmed the potential of this adapted model in explaining the disease process leading from early benign symptoms to persistent symptoms in patients with mTBI (Wijenberg et al., 2020; Silverberg et al., 2019, 2018; Snell et al., 2020; Wijenberg et al., 2017). Although this adapted model has not yet been tested in healthy adults, the original fear-avoidance model (assessing pain) was tested and validated in healthy adults across several studies (Houben et al., 2005; Sullivan et al., 1995; Trost et al., 2011).

The fear-avoidance model explains the mismatch between high severity levels of symptoms and low severity of physical injury (Vlaeyen et al., 2016). This mismatch is also present in healthy adults, explaining general symptom reactions within the normal range. Applying the model to post-concussion like symptoms specifically, the original model is adapted by changing chronic pain resulting in fear of physical activities (kinesiophobia) to post-concussion like symptoms resulting in fear of mental activities (cogniphobia). According to the adapted fear-avoidance model, post-concussion like symptoms are explained by a negative cycle of catastrophizing about symptoms, which leads to cogniphobia (i.e., fear of mental activities), which in turn leads to behavioral avoidance of cognitively demanding tasks (i.e., disuse) or depressive symptoms (see Figure 1). Previous research has shown that the adopted fear-avoidance model is relevant for explaining persistent post-concussion symptoms after TBI (Wijenberg et al., 2020; Wijenberg et al., 2017). Unfortunately, the potential and mechanism of the adapted fear-avoidance model in a non-clinical sample remain unknown.

Figure 1. The fear-avoidance model applied to post-concussion like symptoms

For this reason, an experimental study was conducted to investigate in healthy adults whether fear-avoidance and catastrophizing about mental activities are related to fear-avoidance behavior (i.e., behavioral avoidance of mental activities) according to the fear-avoidance model. The study aimed to relate fear-avoidance behavior in healthy participants to their fear-avoidance and catastrophizing thoughts. This experiment was based on the experimental paradigm and results of Vlaeyen et al. (1995). In the current study, participants were presented with cognitively challenging tasks to evoke fear-avoidance behavior. Participants were able to choose the duration of time spent working on these tasks. Thus, a manipulating variable was introduced (cognitively challenging tasks) in order to evoke features of “psychopathology” (fear-avoidance behavior) in healthy subjects, as is done in experimental psychopathology (Forsyth & Zvolensky, 2001; Jansen et al., 2010). This can contribute to the formation of theories explaining disease processes, in this instance the fear-avoidance model.

The following hypotheses were tested:

(1) We predicted that catastrophizing would be positively associated with post-concussion like symptoms and fear-avoidance, whereas disuse and depression would be positively associated with fear-avoidance and post-concussion like symptoms. This hypothesis would be supported by significant correlations consistent with the fear-avoidance model (see Figure 1). This hypothesis was not tested with the experimental manipulation, but was tested as a necessary first step to verify the interrelationships between variables of the fear-avoidance model in this population.

(2) It was hypothesized that participants with higher levels of fear-avoidance and catastrophizing thoughts (2a) show fear-avoidance behavior (i.e., behavioral avoidance) when performing a cognitively challenging task (i.e. spend less time on the task), (2b) perform worse on this task, and would be more (2 c) anxious and (2d) stressed during the task compared to participants with low levels of fear-avoidance and catastrophizing thoughts. Furthermore, it was predicted that these differences would not be seen on a task which was less cognitively challenging. This hypothesis was tested with the experimental manipulation.

(3) It was expected that participant’s level of catastrophizing and fear-avoidance, measured before experimental manipulation, would predict (3a) the time they spend on a cognitively challenging task and (3b) the performance on this task while controlling for sex, education, and age. This hypothesis was tested with the experimental manipulation.

The results of this experiment will show whether fear-avoidance and catastrophizing thoughts about cognitive tasks are related to post-concussion like symptoms in healthy adults (first hypothesis). Furthermore, it will give insight into the relationship between fear-avoidance and catastrophizing thoughts about mental activities and behavioral avoidance (second and third hypotheses).

Materials and methods

Participants

Participants were recruited between March 2017 and August 2018 by researchers of Maastricht University. Participants were recruited through personal invitation (if they provided permission to be approached for new studies of Maastricht University) or response to an advertisement. The study was advertised through (1) an online research database available for students, (2) a local community app, and (3) flyers spread across the university and public places such as the hospital and supermarkets. Individuals were eligible for the study if they (1) were between 18 and 64 years of age; (2) spoke Dutch fluently; and (3) finished pre-university education, higher vocational education, or academic education. Exclusion criteria were (1) history of any neurological disorder (including traumatic brain injury), (2) history of or current psychological and psychopharmacological treatment for depression or anxiety, (3) use of recreational drugs in the week before and during the study, and (4) unwillingness to sign informed consent. Inclusion and exclusion criteria were checked before the first and second measurements by means of self-report.

Measures

Behavioral measures

In this experiment, there were two conditions: a difficult and an easy condition, in each of which the participants had to complete three computerized tasks. As a measure of behavioral avoidance, participants could choose the time spent on each task. During all tasks, a red stop button was shown which the participants could click on to end the task.

The tasks used in the difficult condition were three validated cognitive tasks used to increase the mental load of the participants. The first task was the Paced Auditory Serial Addition Task – Computerized (PASAT) (Lejuez et al., 2003). During this task, participants heard numbers, ranging from one to nine with two-second intervals. They were instructed to add the current number to the previous number. The participants received visual feedback on whether their answer was correct or not. The second task was the Distress Tolerance Test (DTT) (Nock & Mendes, 2008). For this task the 64 stimulus cards of the Wisconsin Card Sorting Test (WCST) were used (i.e., four key cards and one deck card) (Berg, 1948). The standard instructions of the WCST were presented, indicating that the participants must match the deck card with the key cards. They were not told how the cards were supposed to match. Regardless of the participants’ response, the feedback was “correct” for the first three cards and “incorrect” for the next seven. The feedback on the 11^th card was “correct”, hereafter the remaining cards were rated “incorrect.” During the third task, participants had to solve various anagrams, originally used by Vrijsen et al. (2014). Participants were allowed to make notes (if preferred) and received visual feedback. These anagrams were difficult or even unsolvable.

During the easy condition, simplified versions of the tasks in the difficult condition were used. An easier version of the PASAT was used by prolonging the interval between two consecutive numbers by 3.5 seconds. The regular version of the WCST was used instead of the DTT. During the anagram task, participants again had to solve anagrams; however, now the anagrams were easier and always solvable. These anagrams all consisted of five letters.

Psychophysiological measure

Heart rate was measured during the cognitive tasks using Brain Vision Software (Brain Vision Software, Munich, Germany). Heart rate was recorded using a standard 3-lead electrocardiogram. Electrodes were placed below the left and right collarbone and below the left lower rib. Heart rate levels were calculated by averaging interbeat intervals by detecting the R spikes (Crawford & Doherty, 2011). These values were converted to average beats per minute per condition. This was done using Matlab (The MathWorks, Natick, Massachusetts). Higher beats per minute indicate higher physiological arousal.

Self-report measures

The Rivermead Post-Concussion Symptoms Questionnaire (RPQ) measures the severity of somatic, cognitive, and emotional symptoms following traumatic brain injury (King et al., 1995). The questionnaire consists of 16 items rated on a five-point Likert scale. Total scores range from 0 to 64 with higher scores indicating a higher frequency and severity. Three or more symptoms, indicated by at least three items with an item score of 2 or higher, were used as a criterion for a disabling symptom complex (Wijenberg et al., 2017). This questionnaire was adapted for use within the healthy population as suggested by Dean et al. (2012). The question format was changed from “Compared with before the accident, do you now (i.e., over the last 24 hours) suffer from:” to “Compared with your peers, do you now (i.e., over the last 24 hours) suffer from:”. In the present study, the internal consistency was 0.82 and the test-retest reliability 0.67.

The Post-Concussion Catastrophizing Scale (PCS-CS) measures the level of catastrophizing thoughts regarding post-concussion-like symptoms (Wijenberg et al., 2017). The PCS-CS is adaptive to the RPQ; participants answer questions about the symptom complex they reported on the RPQ. If participants did not report any symptoms, the most common post-concussion symptoms “headache, cognitive problems and/or fatigue” were depicted. Participants were then asked what they think/feel when they experience a headache, cognitive problems, and/or fatigue. The questionnaire consists of 13 items rated on a five-point Likert scale. Higher scores indicate higher levels of catastrophizing and ranges between 0 and 51. In a healthy population, the cutoff score for the PCS-CS is 8 or 14, depending if the individual had “no history” or “history of psychological treatment” respectively (Wijenberg, Stapert, Rauwenhoff, Verbunt, & Van Heugten, unpublished results). The PCS-CS is an adaptation of a validated and reliable measure from the pain literature, the Pain Catastrophizing Scale (Osman et al., 2000). In the present study, the internal consistency was 0.87 and the test-retest reliability 0.65.

The Fear of Mental Activities Scale (FMA) measures the level of fear-avoidance regarding post-concussion-like symptoms and cogniphobia in mTBI patients (Wijenberg et al., 2017). The FMA is an adaptation of a validated and reliable measure from the pain literature, the Tampa Scale of Kinesiophobia (TSK) (Roelofs et al., 2007). The FMA, like the PCS-CS, is adaptive to the RPQ. The participants were required to rate each question on a four-point Likert scale. It is a 17-item questionnaire with higher scores indicating greater fear of mental activity. The scoring of four items (items 4, 8, 12, and 16) is reversed. The total score ranges between 17 and 68. According to its validation study, the four inversed items need to be removed post-hoc, resulting in a cutoff score of 15 for a heightened level of fear-avoidance in a healthy population (Wijenberg, Stapert, Rauwenhoff, Verbunt, & Van Heugten, unpublished results). The removal of the four reversed items is in line with the original validation of the TSK in the pain literature, stating that the four reversed items have been criticized as unreliable and too difficult (Goubert et al., 2004; Houben et al., 2005; Roelofs et al., 2004). In this study, after removing the four inversed items, internal consistency was 0.78 and test-retest reliability 0.71.

The State-Trait Anxiety Inventory (STAI) (Spielberger, 1983) consists of two self-report scales measuring state and trait anxiety. Both subscales consist of 20 items which must be scored on a four-point Likert scale. Both scales have a maximum possible range of 20 to 80 with higher scores indicating greater anxiety. The Dutch version of the STAI and the subscales have good validity and reliability (Van der Ploeg et al., 1980). Similar to prior studies (Rossi & Pourtois, 2012), the STAI state was filled in before and after the difficult or easy tasks. A differential score was calculated, with a higher score indicating increased anxiety compared to before the tasks.

The Beck Depression Inventory-II revised (BDI) (Beck, Steer & Brown, 1996; Van der Does, 2002) is a 21-item questionnaire, which measures the level of depressive symptoms. The items are rated on a four-point Likert scale, with higher scores representing more severe depressive symptoms. The maximum possible range of the BDI is 0 to 63. The BDI has been demonstrated to have good psychometric properties in a variety of samples including healthy subjects (Wang & Gorenstein, 2013).

Regarding personal characteristics, participants were asked to fill in their age, sex, education (highest obtained educational degree), and number of hours spent on mental activity per day. Number of hours spent on mental activity per day was used as a measure for exposure to mental work, as healthy participants could not be asked to compare their current activity level to before injury. In this study, exposure to mental work represents the inverse of “disuse” indicated by number of hours spent on mental activity per day. In other words, the scores were reversed to represent disuse.

Procedure and experimental setup

For this research, a randomized cross-over within-subject design with two study arms (see Figure 2) was used. In the first study arm, participants received the difficult tasks on the first measurement, and the easy tasks on the second measurement. In the second study arm, the order of difficult tasks and easy tasks was switched across the two measurements. Participants were divided into the first or second study arm using a counterbalanced design. On the day of the experiment, no caffeine intake was allowed. Furthermore, the temperature and humidity in the lab were kept constant during the experiment, since these conditions can affect heart rate. After completing the informed consent process, participants filled in their demographic characteristics, the RPQ, PCS-CS, FMA, STAI Trait, BDI, and the STAI State. Depending on the study arm, they were then asked to complete three difficult or easy tasks after a short practice period of the PASAT and the anagrams. Participants could choose the order to maximize its resemblance with solving multiple cognitive challenging problems in everyday life and someone’s own preference to solve these different problems in a particular order. Additionally, they could stop at any moment and continue with the next task. If participants did not terminate the task, each task stopped automatically after 20 minutes. At the end of the tasks, participants were asked to fill out the STAI state again. During the tasks, heart rate was continuously measured. For the second measurement, one week after the first measurement, this procedure was repeated with the other task condition (easy or difficult). However, the STAI trait and BDI were not included. When both measurements were completed participants received course credits (university students) or € 7.50 per hour in vouchers.

Figure 2. Design of the study: randomized cross-over within subject design

Note. RPQ = Rivermead Post-Concussion Symptoms Questionnaire; PSC-CS = Post-Concussion Catastrophizing Scale; FMA = Fear of Mental Activities Scale; BDI = Beck Depression Inventory-II revised; STAI = State-Trait Anxiety Inventory; ECG = Electrocardiogram.

The experimental setup was considered valid when the following assumptions were met: participants spent less time and performed less well on the difficult tasks compared to easy tasks; the difficult tasks induced a stress reaction; age, sex, and education influenced performance in the difficult task; and the internal consistency and test-reliability of two newly-developed questionnaires, assessing catastrophizing and fear-avoidance, were sufficient.

Statistical analyses

All statistical analyses were performed using SPSS 24.0 for Windows (IBM Corp., Armonk, NY). Distributions of all variables were evaluated in terms of mean, SD, median, range, skewness, and kurtosis. No data imputation took place because there was no missing data (except for one participant who did not complete one questionnaire). Outliers and assumptions were checked. Outliers were identified according to the 3IQR rule and winsorized by the second-highest value. In case of non-normality (defined as skewness or kurtosis values outside the range of −1.0 to 1.0), multiple transformations were performed (log, square-root, or inverse). If normalization by transformation was not possible, nonparametric statistics were used. An alpha level of 0.05 was used unless otherwise stated.

To test if the condition manipulation (difficult versus easy tasks) was successful, differences in state anxiety (post-pre) and physiological arousal (average beats per minute) were compared between conditions using a paired sample t-test. A lower score (indicative for less induced anxiety or less arousal) was expected after the easy tasks compared to the difficult tasks.

To test hypothesis 1 (the interrelationships between variables of the fear-avoidance model in healthy adults) correlations between “symptoms,” “catastrophizing,” “fear-avoidance,” “depression,” and “exposure to mental activity,” assessed during the first measurement, were examined. Correlations of the fear-avoidance model were computed by Pearson correlations (in case of normality) or Spearman correlations (in case of non-normality).

To test hypothesis 2 (the effect of catastrophizing, fear-avoidance, and condition on behavioral avoidance, performance, induced anxiety, and stress) four three-way mixed ANOVAs were performed with time spent on tasks, performance, state-anxiety, and heart rate as the dependent variable, respectively. Time spent on the three cognitive tasks together per condition was calculated by the average of the standardized time spent per cognitive task. For all ANOVAs, high and low catastrophizers or fear-avoiders were derived from their respective median splits. Catastrophizing (low/high) and fear-avoidance (low/high) were considered between-subject factors, whereas condition (difficult/easy) was considered a within-subject factor. Besides normality and outliers, the assumptions of homogeneity of variance (assessed by Levene’s test for equality of variance) and sphericity (assessed by Mauchly’s test of sphericity) were checked.

To test hypothesis 3 (predicting performance and time spent on a cognitive task with catastrophizing and fear-avoidance, while controlling for sex, education, and age) two backward multiple regression analyses were performed (separately for time and performance in each condition). Dummy variables were created for categorical variables (catastrophizer, fear-avoider, sex, and education). The quantitative variable (age) was centered (subtracting the median due to non-normality) and its quadratic term was added. The least significant variable with a threshold of α = 0.1 was removed stepwise. In this process, dummy variables were treated as a block. For both the full model as the final model, assumptions of independence of observations (assessed by Durbin-Watson statistics), linearity (assessed by visual inspection), homoscedasticity of residuals (assessed by visual inspection), multicollinearity (assessed by the variance inflation factor and correlations), outliers (assessed by studentized residuals, Cook’s distance and centered leverage), and normality of residuals (assessed by visual inspection) were checked.

Results

Sample characteristics

A total of 80 adults participated in the study. All participants completed both conditions. One participant ended the first session prematurely and did not complete the STAI State following the easy tasks. Participants had a mean age of 30.5 years (SD = 14.4) and 66.3% were female. Despite the exclusion criterion of a low level of education, three participants indicated a low education level as their highest obtained degree (3.8%). Since their scores fell within the range of the rest of the sample, they were not excluded from the analyses. The average number of hours spent on mental activities per day was 7.1 hours (SD = 3.0). Participant characteristics are shown in Table 1.

Table 1. Personal characteristics (N = 80)

Characteristics	Values	Range
Sex, % female (n)	66.3 (53)
Mean age in years (SD)	30.5 (14.4)	18–64
Education level, % (n)		3–7
Low (level 4–5)	3.8 (3)
Medium (level 6–7)	50.0 (40)
High (level 8)	46.3 (37)
Mean hours spent on mental activities per day (SD)	7.1 (3.0)	1–13

Note. SD = Standard Deviation.

Experimental manipulation check

To check if the manipulation of induced stress as a result of task difficulty was successful, differences in state anxiety (post-pre) and physiological arousal (average beats per minute) were compared between conditions. Results show that participants reported a significantly higher increase in state anxiety after the difficult tasks compared to the easy tasks (p < .01, D = 0.69). After removal of one outlier, the results show that participants did not differ in heart rate for the difficult tasks compared to the easy tasks (p = .44).

Hypothesis 1: The fear-avoidance model in healthy adults

Table 2 shows the level of post-concussion like symptoms, catastrophizing, fear-avoidance, exposure to mental activity, and depression. Next to the reliability indexes, the percentages of participants fulfilling the criterion of post-concussion syndrome or having a heightened level of catastrophizing, fear-avoidance, and depression are reported. The distributions of all variables were skewed; for normalization, winsorizing, and/or different transformations across the variables were needed. Therefore, although Pearson correlation analyses revealed similar results, nonparametric statistics were chosen.

Table 2. Descriptive statistics of the fear-avoidance model in healthy adults (N = 80)

Variables	Mean (SD)	Range	%I
Post-concussion like symptoms (RPQ)	6.21 (6.10)	0–28	33.8
Catastrophizing (PCS-CS)	6.24 (5.63)	0–26	27.5
Fear-avoidance (FMA)	15.90 (3.81)	13–36	45.0
Depression (BDI)	5.24 (4.56)	0–21	16.3
Exposure to mental activity (inverse of “disuse”)	7.11 (3.04)	1–13	NA

Note. SD = Standard Deviation; %I = Percentage of participants with an impairment as defined by the cutoff scores (see Method section); RPQ = Rivermead Post-Concussion Symptoms Questionnaire; PCS-CS = Post-Concussion Symptom Catastrophizing Scale; FMA = Fear of Mental Activity Scale; BDI = Beck’s Depression Inventory; NA = Not applicable.

The correlation analyses revealed that all correlations specific to the fear-avoidance model were significant (p < .05), except for the correlations involving exposure to mental activity (see Figure 3). Catastrophizing was positively associated with post-concussion like symptoms (r = 0.31) and fear-avoidance (r = 0.46). In contrast to exposure to mental activity, depression was positively associated with fear-avoidance (r = 0.31) and post-concussion like symptoms (r = 0.57).

Figure 3. Spearman correlations of the fear-avoidance model

Note. BDI = Beck Depression Inventory-II revised; RPQ = Rivermead Post-Concussion Symptoms Questionnaire; PSC-CS = Post-Concussion Catastrophizing Scale; FMA = Fear of Mental Activities Scale.

Hypothesis 2a: The effect of catastrophizing and fear-avoidance on behavioral avoidance

A three-way mixed ANOVA was run to understand the effects of catastrophizing and fear-avoidance on behavioral avoidance (as measured by time spent on cognitive tasks). Time scores were not normally distributed per cell and required square-root transformations. There were no outliers and assumptions of homogeneity of variance and sphericity were met.

There were no significant three-way or two-way interactions (p > .05), but a significant within-subjects effect of condition (F = 7.68; p < .01; partial η² = .09) and a significant between-subjects effect of catastrophizing (F = 4.90; p = .03; partial η² = .06) was found. Post-hoc analyses revealed that participants spent more time on easy tasks and that catastrophizers stopped sooner with the difficult tasks.

Hypothesis 2b: The effect of catastrophizing and fear-avoidance on task performance

A three-way mixed ANOVA was run to understand the effects of catastrophizing and fear-avoidance on task performance (measured as percentage correct on the PASAT). Performance scores were not normally distributed and a logarithmic transformation was needed in the easy condition. There was one outlier in the easy condition and therefore winsorized. Assumptions of homogeneity of variance and sphericity were met.

There were no significant three-way or two-way interactions (p > .05), but a significant within-subjects effect of condition (F = 546.49; p < .01; partial η2 = .88) was found. Post-hoc analyses revealed that participants performed better on the easy tasks compared to the difficult tasks.

Hypothesis 2c: The effect of catastrophizing and fear-avoidance on self-report anxiety

A three-way mixed ANOVA was run to understand the effects of catastrophizing and fear-avoidance on state anxiety (difference score of anxiety before and after the tasks). Anxiety scores were not normally distributed; therefore, a log transformation was performed. There was one outlier in the easy condition and one in the difficult condition. These outliers were winsorized. The assumptions of homogeneity of variance and sphericity were met.

There were no significant three-way or two-way interactions (p > .05), but there was a significant within-subjects effect of condition (F = 31.33; p < .01; partial η2 = .30). Post-hoc analyses revealed that participants had increased state anxiety following the difficult tasks compared to the easy tasks.

Hypothesis 2d: The effect of catastrophizing and fear-avoidance on heart rate

A three-way mixed ANOVA was run to understand the effects of catastrophizing and fear-avoidance on heart rate (measured as average beats per minute per condition). Heart rate was normally distributed in both conditions. There were no outliers and the assumptions of homogeneity of variance and sphericity were met.

There were no significant three-way interactions, two-way interactions, or main effects (p > .05).

Hypothesis 3a: The prediction of behavioral avoidance

A backward multiple regression was run to understand whether personal characteristics (e.g., age, sex, education, or heightened level of catastrophizing or fear-avoidance) could predict behavioral avoidance (measured as time spent on cognitive tasks). Assumptions of multicollinearity were not met and led to the removal of the quadratic term of age. For all cases with a low education level (score 4 or 5, N = 3), leverage values surpassed the threshold of .2 and were removed.

In the difficult task, the final model revealed that female sex and a heightened level of catastrophizing significantly predicted behavioral avoidance (F(2,74) = 6.10, p < .01), accounting for 14.1% of the variation in behavioral avoidance. The prediction equation was “average standardized time spent on difficult tasks = .07 + −.17*Catastrophizing (−1 = low catastrophizing; 1 = high catastrophizing) + −.23*Sex (−1 = male; 1 = female),” meaning that if you are a catastrophizer or female, you spent less time on the difficult task.

In the easy task, time spent on the cognitive task was normalized after log transformation. The final model revealed that sex significantly predicted behavioral avoidance (F(1,75) = 4.018, p < .05), accounting for 5.1% of the variance in behavioral avoidance. The prediction equation was “natural logarithm of average standardized time spent on easy tasks = .08 + −.09*Sex (−1 = male; 1 = female),” meaning that if you are a female, you spent less time on the easy task.

Hypothesis 3b: The prediction of task performance

A backward multiple regression was run to understand whether personal characteristics (e.g., age, sex, education, or having a heightened level of catastrophizing or fear-avoidance) could predict task performance on the PASAT. Assumptions of multicollinearity were not met and led to the removal of the quadratic term of age. For cases with a low education level (score 4 or 5, N = 3), leverage values surpassed the threshold of .2 and were removed.

In contrast to the easy task where no significant predictors were found, the final model of the difficult task revealed that age, sex, and education level significantly predicted task performance (F(4,72) = 5.497, p < .01), accounting for 19.1% of the variance in task performance. The prediction equation was “percentage correct on difficult PASAT = 56.33 + −.52*Age (Age-23) + −5.16*Sex (−1 = male; 1 = female) + 15.83*Education (1 = higher vocational education; 0 = other),” meaning that if you are older or female or did not complete higher vocational education, you score lower on the difficult task.

Discussion

To investigate whether fear-avoidance and catastrophizing about mental activities are related to behavioral avoidance in a healthy population, an experimental study was conducted with 80 healthy adults engaging in cognitively challenging tasks. Participants were exposed to cognitive tasks and could choose the exposure time to these tasks as a proxy for behavioral avoidance.

Vlaeyen et al. (1995) experimentally tested the fear-avoidance model by letting people with lower back pain lift heavy bags. The current experiment was an adaptation to this experimental paradigm to fit the fear-avoidance model assessing post-concussion like symptoms and cogniphobia.

Although only tested post-hoc, the manipulation of the experiment was successful. That is, participants spent more time and performed better on the easy tasks; difficult tasks induced more anxiety compared to the easy tasks; age, sex, and education predicted performance in the difficult task; and internal consistencies and test-retest reliabilities ranged from sufficient to good.

In line with the fear-avoidance model and results of Vlaeyen et al. (1995), the correlations specific to the fear-avoidance model were significant, meaning that the levels of post-concussion-like symptoms, fear-avoidance, catastrophizing, and depression were significantly correlated. Furthermore, participants who reported higher levels of catastrophizing chose shorter exposure time to difficult tasks, suggestive of behavioral avoidance. Furthermore, we found that higher levels of catastrophizing predicted time spent on the difficult tasks. Additionally, female sex also predicted behavioral avoidance regardless of difficulty, in line with the established link between catastrophizing and female sex (Leung, 2012).

Despite these promising results, several findings were not in line with our hypotheses. No significant correlations were found between “exposure to mental activity” and “fear-avoidance” or “post-concussion like symptoms.” Furthermore, fear-avoidance was not related to time spent on the tasks, performance on a task, induced anxiety, and heart rate. Comparison with previous research is limited due to a lack of prior studies investigating the role of disuse concerning cogniphobia or post-concussion symptoms. However, multiple chronic pain studies found similar results, including significant correlations with depression and catastrophizing and not with disuse and fear-avoidance. The non-significant results could also be due to a lack of validity of the measures used for disuse and fear-avoidance (Cook et al., 2006; Verbunt et al., 2010; Wideman et al., 2013). The inverse of disuse was measured by means of one non-validated question (i.e., “How many hours a day do you currently spend performing mental activities such as writing, working on the computer, reading, and participating in a meeting?”). The measure of fear-avoidance originated from the TSK questionnaire, which received some criticism, especially regarding its validity (Lundberg et al., 2011). Future studies are warranted to develop a more ecologically valid measure, such as the photograph series of daily activities (PHODA) scale assessing perceived harmfulness of daily activities (Leeuw et al., 2007). Furthermore, we did not find any effect on our psychophysiological measure (heart rate). This was also the case in the original experiment by Vlaeyen et al. (1995). Furthermore, factors affecting heart rate (e.g., participants’ body weight, medication use, exercise prior to experiment, and the weather) were not considered.

Besides these unexpected findings, there are several limitations to consider. First, participants reported several other reasons than cogniphobia to terminate cognitive tasks (e.g., fatigue, boredom, lack of effort, frustration with difficult or non-solvable tasks). These reasons could have decreased the time spent on the cognitive tasks regardless of their level of catastrophizing or fear-avoidance. However, reasons for terminating the tasks were not retrieved systematically. Secondly, participants were rewarded, potentially influencing their time spent on the tasks, despite their level of catastrophizing or fear-avoidance. However, this seems unlikely as participants were informed beforehand that the experiment would take a fixed period (set at the maximum duration) and participants did not know that the experiment would terminate after their termination of the cognitive tasks. Thirdly, the absence of effects regarding fear-avoidance could also be related to the community sample in this study. Although symptom levels can be found on a comparable level in patient populations, it is stated that when hypothesized pathological processes are evoked in healthy participants a similar subclinical pathology originates, however to smaller extents (Jansen et al., 2010). This effect might have been too small to become meaningful. It is therefore important to also administer this experimental study in patient populations with conditions associated with unexpected persistent cognitive symptoms such as mTBI to assess its generalizability. Furthermore, taking into account the limitations of stepwise selection, future research is needed to cross-validate the final models. Fourthly, the exclusion criterion related to absence of recreational drugs within one week before- and during testing was checked before the start of each measurement by the use of self-report. Self-report may potentially influence the results through the inclusion of unsuitable participants. However, since this study has a within-subject design it is unlikely that this influenced the results. Fifth, no pilot testing occurred to check the experimental manipulation. However, posthoc analyses were performed in order to test the manipulation, which appeared to be successful. Finally, one could argue that the common established interrelationships of the fear-avoidance model across various patient and healthy populations limit its specificity. One of the main elements of the fear-avoidance model is the prominent and evident role of psychological factors. Previous research has shown that psychological factors such as personality, coping, and psychological vulnerability are important to consider in various patient populations (Pincus et al., 2002; Silverberg et al., 2015; Van Mierlo et al., 2014) and healthy adults (Bernstein et al.; Moroz & Dunkley, 2019; Riskind & Alloy, 2006). Future studies should also evaluate other potential explanations or models, such as an integrative model explaining medically unexplained symptoms (Brown, 2004), biopsychosocial model of overuse (Hasenbring & Verbunt, 2010), anxiety (Fear et al., 2009), depression (Iverson, 2006; Wang et al., 2006), somatization (Harris et al., 2009), alexithymia (Leising et al., 2009), or low self-efficacy (Zenger et al., 2013), to understand if our findings regarding the fear-avoidance model are of additive value compared to these alternative explanations.

Despite some unexpected findings and lack of significant results concerning fear-avoidance, the results of this study extend our current knowledge on the fear-avoidance model applied to post-concussion (like) symptoms revealing its presence in healthy adults and highlighting the importance of catastrophizing due to its demonstrated relationship with behavioral avoidance. This biopsychosocial approach adds to our understanding why low levels of injury severity (i.e., benign injuries) can lead to high levels of symptom reporting which cannot be explained based on strictly biological models. Moreover, even in the absence of injury in a non-clinical sample, the normal range of daily complaints such as headaches are related to psychosocial variables. As post-concussion-like symptoms occur within the healthy population and likely become more severe in combination with disease or injury (such as mTBI), the results support further (experimental) investigation into the fear-avoidance model and its associated treatment in patient populations.

Disclosure of interest

No potential conflict of interest was reported by the author(s).

Acknowledgments

We thank all participants for their participation. Furthermore, we would like to thank Myrthe Dingemanse, Noëlle Laros, and Veerle van Gils for helping collect the data. Lastly, we thank Charlotte Southcombe, Ashley K. Smith Watts, PhD, and the INS Global Engagement Committee Research Editing and Consulting Program for their English language editing service.