Background
Anxiety disorders were ranked as the second leading cause of disease burden among all mental disorders, according to a recent Global Burden Disease report [
1]. In addition to reducing quality of life and daily functioning [
2], anxiety disorders are associated with elevated risks of cardiovascular disease [
3] and premature mortality [
4]. Deficits in cognitive performance have been reported, but findings are heterogeneous and inconsistent and few studies are set in the context of primary care [
5‐
10]. There is some evidence that impairments in executive function (EF) may be of particular importance in the association between anxiety disorders and cognitive function [
11].
EF refers to several “top-down”, effortful cognitive processes needed to regulate thoughts and actions during goal-directed behaviours [
12,
13]. It is a complex cognitive concept including attention, inhibition, working memory (WM), cognitive flexibility, reasoning and problem-solving [
12]. Attention is the ability to selectively attend and focus on a task at hand [
12]. Inhibition, closely related to attention, involves being able to control one’s attention, behavior, thoughts, and/or emotions to override or suppress attention to other stimuli, and instead do what’s more appropriate or needed [
12]. Deficits in attention and inhibition have been reported in patients with panic disorder (PD) and generalized anxiety disorder (GAD) [
6,
14‐
16], though findings are inconsistent [
9,
10]. WM involves holding information in mind (maintaining) and mentally working with it (manipulation) across a shorter delay [
12]. Impairments in WM performance have been shown in patients with GAD [
7] and in induced anxiety in a non-clinical population [
17]. The evidence is mixed regarding PD, with WM deficiencies shown in one study [
18] but not in others [
10,
19]. Cognitive flexibility is being able to change perspectives spatially or interpersonally, ability to change how we think about something (think outside of the box) and flexibility to adjust to changed demands, to admit you were wrong, and to take advantage of sudden, unexpected opportunities [
12]. Research on cognitive flexibility in persons with anxiety disorders is divided with studies showing impairments in patients with PD or GAD [
6,
15,
18], but also lack of such impairments [
6,
19].
EF is also highly correlated to dimensions of the concept fluid intelligence i.e. the ability to understand relationships among components, to reason and solve problems [
12,
20]. We could not find any studies specifically analysing fluid intelligence in patients with anxiety disorders in adults, but a lack of an association of fluid intelligence and anxiety disorders (excluding specific phobia) has been shown in adolescents [
21]. Research on potential associations between anxiety severity and EF are scarce, but one study reported an inverse association between symptom severity and EF in patients with social anxiety disorder (SAD) [
8]. Associations with other anxiety disorders remain to be clarified.
Associations between anxiety disorders and EF are important to study as deficits in EF may reduce a patient’s coping abilities, affecting a patient’s ability to function socially and occupationally in everyday life [
11]. There are also clinical implications including enhanced screening and better understanding of treatment mechanisms [
11]. When investigating the relationship between anxiety and cognitive functions, it is important to also consider comorbid depression. Presence of comorbid depression may confound the results as previous studies have demonstrated an association of depression and impaired functions within the domains of EF [
22,
23]. Moreover, comorbid anxiety and depression might also represent a more severe illness state, than anxiety alone [
24].
It has been estimated that 70% of individuals seeking help for anxiety initially present in primary care [
25] and costs for mental illness within Swedish primary care are on the rise [
26]. However, the relationship between anxiety disorders and cognitive performance has been little studied in this setting. The primary aim was to investigate whether level of EF was cross-sectionally associated with severity of anxiety in patients diagnosed with anxiety disorders (PD, GAD and anxiety not otherwise specified), in a primary care setting. Association of fluid intelligence and anxiety severity was also investigated due to the high correlation of fluid intelligence and the EFs problem-solving and reasoning [
12]. Our hypothesis was that functions attributed to EFs (WM, inhibition, cognitive flexibility and attention) and fluid intelligence would be inversely associated with anxiety severity [
7,
8,
27]. Secondary aims were to investigate these associations also in models adjusted for the existence of depression and to analyse EF performance in patients diagnosed with the anxiety disorders specified above in relation to a normed population.
Methods
Participants and settings
Participants for this cross-sectional study originate from the ongoing randomized controlled study Swedish Physical Fitness and Brain - Interventional Study (PHYSBI; NCT03247270; Trial Registration Date: 08/08/2017), focusing on investigating effects of an exercise intervention on symptoms of anxiety and cognitive function in patients with anxiety disorders [
28]. Individuals who sought help for anxiety issues at six primary care units in Gothenburg (Närhälsan Primary Care) and Region Halland were recruited. Potential participants were diagnosed by a study psychiatrist using the Mini International Neuropsychiatric Interview (M.I.N.I; Swedish version 7.0.0 DSM 5), a structured diagnostic interview with high reliability and validity [
29]. Patients aged 18–65 were included if diagnosed with the anxiety disorders PD (DSM 300.01) or GAD (DSM 300.02) according to M.I.N.I. In addition, patients with anxiety not otherwise specified (NOS; DSM 300.00) were also included after being diagnosed by the study psychiatrist. In order to maintain statistical power for the analyses, patients with aforementioned anxiety disorders were grouped together and denoted as patients with anxiety disorders. Patients with and without ongoing treatment with psychotropic medication were included. Individuals with ongoing psychotherapy were excluded since psychotherapy was viewed as a “commitment” in terms of time and energy which could impact adherence to the exercise intervention. Additional exclusion criteria included high suicide risk (patients with low to moderate suicide risk were included) or serious neurodevelopmental or psychotic disorders (milder cases were included) as assessed by the study general practitioner (GP). Pregnant women were not included in the current study. The study was approved by the Regional Ethics Committee in the Gothenburg, Sweden and was carried out in accordance with the Declaration of Helsinki (2013). Each participant signed a statement of informed consent after the nature of the procedures had been fully explained. For further details regarding the study methodology including sample size calculations, please see the study protocol [
28].
Assessment of anxiety severity
Severity of perceived ongoing symptoms of anxiety at baseline was self-assessed using the Swedish version (©2005 by NCS Pearson) of the clinically well-established Beck Anxiety Inventory (BAI) [
30]. BAI mainly evaluates somatic symptoms and was developed to be relatively free from contamination by depressive content [
30]. Both reliability [
31] and validity [
32] are reported to be good.
Cognitive tests
Cognitive performance was measured using Wechsler Adult Intelligence Scale 4th edition (WAIS-IV) and the Delis-Kaplan Executive Function System (D-KEFS). The cognitive tests were applied by a licensed psychologist. WAIS-IV is a battery of tests measuring intelligence and cognitive functions, standardized on a normative sample of individuals (ages 16–90) and stratified to match a Scandinavian population based on age, sex, education, ethnicity and geographic region. In the current study, we used the block design, digit span and matrix reasoning tests including scaled scores from 1 to 20 with a normed mean of 10 and standard deviation (SD) of 3 [
33]. Full WAIS-IV assessment was not employed due to the long completion time. For full information including WAIS-IV test descriptions, subtest modifications and reliability/validity statistics, see the WAIS-IV technical manual [
33].
The Delis-Kaplan Executive Function System (D-KEFS) is a standardized, non-verbal psychomotor test battery aimed at assessing EFs in individuals aged between 8 and 89 years [
34]. D-KEFS has been used in both clinical and research settings showing good reliability and validity for measuring EF [
35]. The D-KEFS normative sample was composed using the 2000 U.S. Census figures as target values [
35,
36]. The sample included over 1700 children, adolescents and adults (ages 8 to 89 years) and was based on demographic characteristics (including age, gender, socioeconomic factors) of the U.S. population [
34]. Since the normative samples were divided into age groups, every patient tested using D-KEFS is compared to a large group of individuals of the same age span. WAIS-IV tests were performed before D-KEFS in all participants.
Block design
Subjects had to replicate red and white pattern designs using three-dimensional coloured blocks. This test measures functions including visual perception and problem-solving and non-verbal reasoning [
33]. The block design test is commonly used as a measure of fluid intelligence [
37‐
39], but it involves reasoning and problem-solving which are functions also attributed to EF [
12].
Digit span
Subjects were asked to repeat a sequence of numbers read to them in order (forward), in reverse order (backward) or in ascending order of magnitude (sequencing). Digit span taps into functions including WM, attention, encoding and auditory processing. Digit span forward primarily measures short-term memory and attention [
12,
40], digit span backward measures WM and digit span sequencing captures functions such as cognitive flexibility [
33]. A combined score for all three digit span subtests (included in the WM cognitive domain) [
33], was also obtained.
Matrix reasoning
Subjects had to solve a task presented in a visual format and identify patterns in designs. This test includes perceptual reasoning, non-verbal problem-solving and visuospatial ability [
33], and is usually used as a measure of fluid intelligence. However, as for the block design test, the matrix reasoning test also involves reasoning and problem-solving which also are part of EF.
Design fluency
In the current study, we used the D-KEFS design fluency test in order to assess the ability to generate a series of novel (non-repeating) and abstract designs. A rationale for choosing this test is that it simulates the cognitive chain required in daily life to generate novel responses, while maintaining focus on a desired goal [
34,
35], and involves multiple EFs including creativity, attention, inhibition and scanning and cognitive flexibility to find novel solutions.
The design fluency test is performed using a pen and paper and consists of three conditions of increasing difficulty where the subject has to connect dots and make novel shapes within a time limit [
12,
34]. In the first condition, the subject had to create novel patterns by combining filled dots with four lines without repeating previous combinations, which demands creativity in drawing new designs. The second condition was the same as the first, but now with unfilled dots instead which increases the demand also for inhibition. The third condition consists of both filled and unfilled dots where the subject had to switch between filled and unfilled dots when creating the patterns, and which adds the requirement for cognitive flexibility. The total number of correct patterns and number of total patterns within the 60 s time limit was recorded for each condition and raw scores from each subtest were converted to age-adjusted scaled scores ranging from 1 to 19 (mean 10; SD 3), before analysis [
34].
Other measures
Comorbid psychiatric disorders including major depression were diagnosed by a psychiatrist using the M.I.N.I. Severity of ongoing symptoms of depression were self-assessed with the Montgomery Åsberg Depression Rating Scale (MADRS-S) [
41]. The following variables were self-reported by study participants in a questionnaire designed by the research team: marital status, education level, years with anxiety symptoms, smoking and ongoing use of prescribed antidepressants (ATC N06) and psycholeptics/anitiepileptics (ATC N05 and N03).
Procedure
Patients were recruited at five different time points from August 2017 to September 2019. The patients were informed by their GP or primary care psychologist about the possibility of study participation and those expressing interest were contacted by the study physician for further information. Diagnoses and comorbidities were determined by a psychiatrist and the cognitive tests were administered by a psychologist.
Statistical analyses
All analyses were performed using Statistical Package for the Social Sciences (SPSS), 25.0 software (SPSS Inc., Chicago, IL). Characteristics of the whole study group are presented using descriptive statistics including number of observations, means and SD for continuous variables. Frequencies and percentages are presented for categorical variables.
Additional analyses were performed comparing characteristics, including performance on cognitive tests, of patients with or without psychotropic medication (antidepressant and/or psycholeptics/anitiepileptics), as well as patients with minimal/mild vs. medium/severe anxiety. For these, Pearson’s χ2-test was used for categorical variables and Mann-Whitney U-test was used for continuous variables. Normality was assessed graphically, and for most variables there were skewed distributions, and hence the Mann-Whitney U-test was used.
In order to investigate the relationship between severity of anxiety and cognitive functioning, multiple linear regression analyses were performed with self-reported BAI scores as an independent, continuous variable and performance scores on block design, digit span, matrix reasoning and design fluency tests as dependent variables. Different multivariable models were analysed including age, gender [
42], smoking [
43‐
45], education level and comorbid major depression (assessed through M.I.N.I.). Additional regression analyses were performed comparing patients with minimal/mild (BAI 0–16) and medium/severe anxiety (BAI 17–63) [
46].
To compare cognitive functioning in patients with anxiety disorder to an age-adjusted normed population, the Wilcoxon signed rank test with standard algorithms was used (with a normed mean of 10, and an SD of 3 for all tests) [
33,
35]. These analyses were performed since the current study design did not involve a group of persons without anxiety disorders for comparison. This method has been used previously for measuring cognitive function in young men [
47]. For these we calculated N-1, which represents the degrees of freedom, and t-values from the F(df
regression,df
residual) = F
regression eq.
P-values < 0.05 were considered statistically significant.
Discussion
In this cohort of primary care patients with anxiety disorders (PD, GAD and anxiety not otherwise specified), higher anxiety score was associated with lower EFs specifically related to WM (digit span) in multivariable models, after adjustment for comorbid major depression. This finding is also supported by our additional analyses showing that patients with moderate/severe anxiety symptoms scoring lower on the digit span test, compared to patients with minimal/mild symptoms.
The above result is in line with our hypothesis, that there would be a negative association of anxiety severity and measures of WM. In analyses adjusting for comorbid major depression, we observed an association of anxiety severity with scores on the digit span total and backward tests, but not with forward and sequencing tests. Digit span total is a general measure of WM and digit span backward specifically involves EF resources related to an active WM where information also is manipulated [
12,
48]. Digit span forward on the other hand primarily measures a “passive” short-term memory where information is maintained [
12] and digit span sequencing is a measure of cognitive flexibility. The inverse associations of anxiety severity and digit span total and backward scores in models adjusted for comorbid major depression, therefore indicate a relation between severity of anxiety and EFs related to WM, and not short-term memory or cognitive flexibility. Although we could find no studies investigating cross-sectional associations between anxiety level and WM performance for comparison, we note that patients with GAD (and without depression) scored lower on WM performance compared to healthy subjects [
7]. This WM deficit was connected to lower prefrontal engagement; it was suggested to represent a key component of clinical anxiety, rather than a consequence of threat. On the other hand, in a review of PD and cognitive function, limited support was found for an association of PD alone in the absence of depression and WM [
10]. Contrary to our hypothesis, we did not find an association between cognitive flexibility and anxiety severity. Lower cognitive flexibility have previously been reported for patients with PD with and without comorbid depression, but not for patients with GAD [
6]. Our results might in part be explained by the large proportion of patients with GAD (57%) included in the current study. However, younger patients (20–30 years old) with GAD without comorbid depression have demonstrated impaired cognitive flexibility [
15].
Although the block design and matrix reasoning tests are commonly used as measures of fluid intelligence, they also measure reasoning and problem-solving which are highly correlated to EFs. Our hypothesis was therefore that more severe anxiety would be associated with lower performance on these tests, but our results from analyses adjusted for comorbid major depression did not support our hypothesis. Lack of association of between anxiety symptoms and subsequent fluid intelligence (block design test) was also found in a study of Swedish twins (aged 50 or older), in models adjusted for depressive symptoms [
49]. Lower fluid intelligence may hence be more related to the depressive state and not to anxiety, as also indicated by our multivariable models not adjusted for comorbid major depression. In line with the twin study, we conclude that severity of anxiety is not associated with fluid intelligence including the related EFs problem-solving and reasoning in the current study group of primary care patients.
We did not find an association between anxiety severity and performance on the design fluency tests (measuring aspects of EF including creativity, inhibition and cognitive flexibility) and could therefore not confirm our hypothesis regarding cognitive flexibility and inhibition. Lower design fluency scores have been shown in patients with comorbid clinical depression and anxiety [
50]. However, absence of a dose-response relationship between number of comorbid anxiety disorders (as a measure of severity) and design fluency scores, after adjusting for comorbid depression, has also been reported in a large population-based sample of adults [
24]. More research is clearly needed to investigate associations of EFs and severity of anxiety.
Psychotropic medication may influence cognitive function, which may affect the results [
51,
52]. Antidepressants have been shown to have positive effects on EF in depressed patients [
51], but it remains unclear whether this is the case in patients with anxiety disorders. We did not observe any between-group differences in cognitive performance in patients with or without medication, but confounding by indication must be taken into consideration.
Cognitive function in patients with anxiety disorder compared to a normed population
Our study patients scored lower on tests on EFs related to WM and fluid intelligence compared to a normed population. Fluid intelligence is a wider concept including the ability to solve novel problems by using reasoning and not depending on accumulated knowledge such as schooling and acculturation [
20]. Fluid intelligence might thus be a factor differentiating individuals with or without a clinical anxiety disorder, while WM deficits might only be obvious among diagnosed patients with more severe anxiety symptoms.
Surprisingly, our results showed that overall design fluency test performance of patients with anxiety were higher than the normed mean for the population. This result may be interpreted as patients with anxiety have better EF. There are sporadic reports of an association between anxiety disorders and “creativity” such as a creative occupation [
24]. However, a more plausible explanation for our results may involve the normative sample for the design fluency test. The D-KEFS normative sample was composed using the 2000 U.S. Census figures as target values [
35], but cross-national differences in cognitive function exist [
53]. One possible explanation for our results is that Swedish D-KEFS means might be higher than U.S. means. WAIS-IV, on the other hand, was normed against a Scandinavian population [
33]. Another possibility is the presence of a “Flynn effect” (the globally observed rise in intelligence test scores over time) since the construction of the D-KEFS normative sample in 2000 [
54].
Limitations
Several limitations are acknowledged. The cross-sectional design of the study excludes causal inferences between anxiety severity and EFs. Deficits in EF can be a consequence of, but may also increase vulnerability to the development and maintenance of anxiety disorder. The relationship may also be bidirectional. The causal link between anxiety disorder and EF has not been well established and longitudinal studies will be required to elucidate this relationship. The choice of measure for anxiety symptoms might impact the results. We chose to use BAI since it is well established within Swedish primary care and minimizes influences of depressive content. Although BAI was developed to capture both somatic and cognitive aspects of anxiety, it focuses primarily on somatic symptoms (Beck 1998). A different measure such as the commonly used State-Trait Anxiety Inventory (STAI) might yield different results. However, STAI was not the most appropriate for the current study given that the original purpose for study participation was to longitudinally investigate the effect of physical exercise on anxiety symptoms over time. STAI assesses both symptoms associated to situations or events and symptoms associated to more stable personal traits. Given the intent to capture longstanding traits, STAI is less suitable to detect change than BAI (Julian 2011). Another limitation is the lack of a non-clinical control group for comparison of measured variables. However, we did perform analyses comparing cognitive function in patients with anxiety disorder, to normed populations. Study participants were recruited on a voluntary basis to take part in an exercise intervention study, which might have produced a selection bias regarding cognitive functioning. Sample size calculations for the original longitudinal RCT study were based on the effect of exercise interventions on symptoms of anxiety and depression [
55]. These included two intervention groups and one control group. Specific sample size calculations for the current study, assessing baseline anxiety severity and cognitive functions, were not performed. However, the current study pooled the three groups of patients (two intervention groups and one control group), which should increase the statistical power. The order in which cognitive tests were performed could also affect the test results, given that the WAIS-IV tests were consistently performed before the D-KEFS test. One might speculate that patients with more severe anxiety might experience greater fatigue towards the end of the test procedure (which took approximately 35–45 min) than patients with milder anxiety. However, the last test was design fluency, which showed no association with anxiety severity. The observed association might also be influenced by unmeasured variables affecting both anxiety severity and cognitive function.
Conclusions and implications
In primary care patients with anxiety (PD, GAD and anxiety not otherwise specified) anxiety severity is negatively associated with EFs related to WM after adjustment of major depression. The current study has implications for the understanding of executive behavioural control in primary care patients with the above-specified anxiety disorder. Characterization of cognitive function in patients with anxiety may facilitate the development of more individualized treatment strategies, that could include interventions to improve specific cognitive domains when indicated. Such individualized strategies could have clinical implications for treatment compliance, symptom reduction, coping mechanisms, as well as overall daily functioning for primary care patients with anxiety disorders. For example, EF may predict treatment response to cognitive behaviour therapy in anxious older adults [
56]. Moreover, research on EF may advance the understanding of the psychopathology of anxiety and identify vulnerability factors. Research has shown that EF impairments can elevate the impact of repetitive negative thoughts (including worry and rumination) on the development of anxiety disorder [
57]. The potential impact of physical exercise on anxiety severity and tests of cognitive performance in primary care patients with anxiety disorders will be presented in a future publication.
Acknowledgements
The authors would like to thank Annika Appelgren and Jennifer Rudolph at Närhälsan Sisjön, Region Västra Götaland for support and encouragement. Special thanks are given to Birgitta Johansson at the Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Ulla Karilampi, Emelie Delphin and Stefan Wiktorsson at the Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Sweden and Linus Schiöler at the Department of Public Health and Community Medicine, Institute of Medicine, Sahlgrenska Academy, University of Gothenburg.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.