Metacognitions, worry and sleep in everyday life: Studying bidirectional pathways using Ecological Momentary Assessment in GAD patients

Highlights

• We studied bidirectional pathways of worry and associated phenomena.

• The variables were investigated in everyday life of GAD patients.

• Negative metacognitions lead to heightened levels of worry – but not vice versa.

• We present evidence for a bidirectional worry–sleep quality relationship.

• The cognitive processes under study may be promising targets for treatment.

Abstract

Background

The metacognitive model of generalized anxiety disorder proposes that negative metacognitive beliefs are crucial in the maintenance of excessive worry. Furthermore, according to the cognitive model of insomnia, worry leads to problems falling or staying asleep and poor sleep quality. In order to test the assumed causal relationships, the present study examined the time-dependent course of negative metacognition and worry as well as worry and sleep quality, using Ecological Momentary Assessment (EMA).

Method

Negative metacognitions, worry and sleep were assessed by self-report questionnaires as well as EMA in 56 GAD patients who carried a portable device for 1 week and logged sleep quality, negative metacognition and worry processes four times a day.

Results

Metacognitions, worry and sleep were significantly correlated. Structural equation modeling using multilevel analyses showed a unidirectional relationship of negative metacognitions leading to prolonged worry processes and a bidirectional relationship of worry and sleep quality.

Conclusions

These findings support the theoretically derived assumptions on the relationship between negative metacognitions, worry and sleep. Implications for further research as well as clinical implications are discussed.

Introduction

Worry is a normal and everyday phenomenon experienced by most people. However, in some individuals, worrying becomes problematic and leads to significant distress and/or impairment in daily functioning. Pathological worry can be distinguished from normal worry by two key characteristics: it is excessive and it is perceived as uncontrollable (e.g., Ruscio & Borkovec, 2004). Consequently, excessive worry that is perceived as uncontrollable is the defining feature of the current diagnostic criteria for generalized anxiety disorder (GAD; American Psychiatric Association, 2013).

In his metacognitive model, Wells (1995) suggests that the perception of worrying as uncontrollable is not only a phenomenological feature of GAD, but also a key process that is causally involved in the maintenance of the disorder. The act of worrying is proposed to only become problematic when it is negatively appraised, for example, as uncontrollable. The model postulates a sequential process with positive metacognitions (i.e., “Worry helps me cope.”), initiating Type 1 worries (worries about internal or external cues; e.g., “My husband might have an accident.”), which in turn trigger negative metacognitions (i.e., “Worry is dangerous for me.”), leading to Type 2 worry (meta-worry; i.e., “If I keep worrying, it will drive me mad.”). Importantly, Type 2 worries are suggested to trigger negative emotions and vain attempts to stop worrying (i.e. thought suppression, reassurance-seeking or avoidance), both of which contribute to worry being maintained (e.g., Andor, Gerlach, & Rist, 2008). This vicious circle of negative metacognitions and continuous worry is then perceived by the individual as confirmation for the belief that worry is uncontrollable.

To date, evidence for the model assumptions mainly derives from correlational investigations. A large number of studies have shown significant and substantial bivariate associations between metacognitions and worry (for a review, see Behar et al., 2009, Wells, 2004), with negative metacognitions related to uncontrollability and danger of worrying showing the closest and most consistent connection (Cartwright-Hatton and Wells, 1997, Davis and Valentiner, 2000, Ruscio and Borkovec, 2004). Although these results provide some support for the metacognitive model, they remain silent about the precise nature of the relationship between worry and metacognitions. Additionally, measuring worry retrospectively may lead to recall biases. Global self-report was shown to only account for a small part of variance in everyday worry (Verkuil, Brosschot, & Thayer, 2007).

In order to overcome these limitations and also more closely examine the dynamic nature of the relationship between the two processes proposed by the metacognitive model, the use of Ecological Momentary Assessment (EMA), logging current states and/or experiences in real time and natural environments, is promising (e.g., Ebner-Priemer, Kubiak, & Pawlik, 2009). The use of EMA has been shown to be a beneficial addition to retrospective measurement (Ebner-Priemer & Trull, 2009).

A number of earlier studies used EMA to assess worry in everyday life and investigate its association with other relevant processes (e.g., heart rate variability, trait worry or sleep; Brosschot et al., 2007, Takano et al., 2014, Verkuil et al., 2007). However, to our knowledge, only one earlier study has used EMA to assess the relationship between negative metacognitions regarding uncontrollability and worry. In this study, non-clinical participants first filled in a self-report questionnaire of negative metacognitions and then recorded upcoming worry several times a day during the following week (Thielsch, Andor, & Ehring, in press). Results showed that negative metacognitions significantly and substantially predicted worry recorded in the following week. The first aim of the current study was to extend these findings in two important ways. First, the current study focussed on individuals suffering from GAD instead of non-clinical participants. Secondly, not only worry but also negative metacognitions regarding uncontrollability of worrying were assessed using EMA, allowing to more closely study the lagged relationship between these variables. Based on the metacognitive model, a bidirectional relationship between the two constructs was expected as perceived uncontrollability is assumed to increase worry, whereby excessive worry in turn should maintain the belief that worry is uncontrollable and dangerous.

As described earlier, the metacognitive model suggests that excessive worrying leads to a number of affective, physiological, cognitive and behavioral consequences that further contribute to the maintenance of worry. One particular process that has been suggested to be related to excessive worry in this way is impaired sleep. Examining the relationship between worry and sleep appears especially relevant as sleep disturbance is among the diagnostic criteria for GAD. Furthermore, GAD has been identified as the disorder with highest comorbidity of sleeping problems/insomnia among the anxiety disorders (e.g., Monti & Monti, 2000). These findings provide indirect evidence for a close link between worry and sleep, although more specific research on the association has been carried out as well. In her cognitive model of insomnia, Harvey (2002) suggests that worrisome thinking during the day and at bedtime leads to arousal and distress that interferes with sleep onset and/or quality. Decreased sleep quality, in turn, is suggested to lead to increased worrying by triggering selective attention and a biased perception of the sleeping deficit (see also Jansson & Linton, 2006). Many of the specific predictions generated by the cognitive model of insomnia have been empirically tested, whereby their explanatory power could be verified (for a review, see Harvey, 2005). Correlational studies measuring sleep quality as well as cognitive activity at daytime and in bed (e.g., Harvey, 2000) document the relationship of worry and sleeping problems, whereas experimental manipulations indicate a causal nature for worry impairing sleep (i.e., increasing cognitive activity before bed and thereby decreasing sleep quality; Tang & Harvey, 2004) as well as poor sleep quality causing worry (provoking worry by providing false feedback on poor sleep; Neitzert Semler & Harvey, 2005).

Only a few studies have used EMA to assess the association between worry and sleep quality online and in the natural environment (Takano et al., 2014, Wicklow and Espie, 2000) and also found significant associations. To our knowledge, only one study, an unpublished doctoral dissertation, also tested the assumption of a bidirectional relationship between worry and sleep. From this study in high worriers, evidence emerges only for one direction of the relationship: worry at daytime predicted increased sleep disturbance that night, whereas sleep quality did not predict worry the following day (McGowan, 2014). However, the study focussed on a non-clinical undergraduate student sample. The second aim of the current study therefore was to more closely examine the relationship between worry and sleep using EMA in individuals suffering from GAD.

In order to test the mutually maintaining relationships between negative metacognitions regarding uncontrollability and worry on the one hand, and worry and sleep on the other, a sample of GAD patients completed measures of these variables for 7 days on a total of four measurements per day. The following hypotheses, derived from the metacognitive model of GAD and the cognitive model of insomnia, were tested.

Previous perceived uncontrollability significantly contributes to the amount of current worry, while simultaneously controlling for the effect of previous worry (Hypothesis 1a). Previous worry significantly contributes to the amount of current uncontrollability, while simultaneously controlling for the effect of previous uncontrollability (Hypothesis 1b).

Previous day's mean worry score significantly predicts the amount of poor sleep quality in the following night, while simultaneously controlling for the effect of prior sleep quality (Hypothesis 2a). Poor sleep quality in the night before significantly contributes to the mean amount of worry the following day, while simultaneously controlling for the effect of prior day's average worry score (Hypothesis 2b).