Trait anxiety moderates the impact of performance pressure on salivary cortisol in everyday life

Summary

Stress and negative affective states are associated with cortisol in everyday life. However, it remains unclear what types of stressors and which affective states yield these associations, and the effect of trait anxiety is unknown. This study investigates the associations of specific task-related stressors and negative affective states in everyday life with salivary cortisol, and explores the mediating and moderating role of state negative affect and trait anxiety, respectively.

Salivary cortisol, subjective stress, and state negative affect were measured three times a day on 2 days in 71 participants in everyday life, using a handheld computer to collect self-reports and time stamps and an electronic device to monitor saliva sampling compliance. Stress measures comprised the experience of performance pressure and failure during daily tasks; measures of negative affect comprised worn-out, tense, unhappy, and angry. Effects were tested using multilevel fixed-occasion models.

Momentary performance under pressure was related to higher momentary cortisol measures, while mean task failure was related to lower daily cortisol concentrations. The association of performance pressure with cortisol varied between subjects, and this variation was explained by trait anxiety, yielding stronger associations in participants scoring high on trait anxiety. No evidence was found for a mediating role of state negative affect.

These results describe the well-documented associations of everyday stressors and affect with salivary cortisol more precisely, suggesting that performance pressure is a significant condition related to short-term changes in cortisol. Subjects scoring high on trait anxiety seem to process stress-relevant information in a way that amplifies the association of performance pressure with reactions of the hypothalamus–pituitary–adrenal axis.

Introduction

In humans, stress results from uncontrollable demands, which may trigger changes in physiological, emotional, cognitive, and behavioural states. Two prominent elements of the stress response are the activation of the hypothalamus–pituitary–adrenal (HPA) axis, with an increase in its end product cortisol, and negative emotional reactions, experienced as negative affective states like tension or anger. In the absence of specific physiological challenges, relevant stimulus information is processed in cortical and limbic central nervous system areas before endocrine changes are triggered (Feldman et al., 1995, López et al., 1999, Buijs and Van Eden, 2000, Herman et al., 2003). Thus, psychosocial stressors from everyday life, e.g. momentary work-related or social stressors, are expected to trigger both emotional and endocrine stress responses, and the strength of the response associations may depend on individual differences in the processing of emotional information. However, in studies that reported results of a correlation analysis the covariation of endocrine and psychological stress indicators was found to be small (e.g. al'Absi et al., 1997, Buchanan et al., 1999, Cohen et al., 2000).

The investigation of the relative importance of specific stressors and the role of particular affective states and traits may help us to understand psychoendocrine responses to stressors. Given the links between HPA activity and various bodily and mental disorders (e.g. Chrousos & Gold, 1992), specific psychoendocrine associations may also help to understand the role of subjective stress in health and disease processes. This study investigates the effects of specific task-related demands and negative affective states on salivary cortisol secretion in everyday life. In addition, the impact of trait anxiety on the relationship between task-related demands and salivary cortisol will be studied.

A relatively new field of research is the investigation of associations between stressful events and cortisol responses using synchronous measures in a naturalistic design, and making use of suitable methods to measure cortisol and self-reports synchronously, as well as suitable statistical methods to test the effects of interest. Three studies (Van Eck et al., 1996, Smyth et al., 1998, Peeters et al., 2003) used paper–pencil diaries and instructed the participants to report every stress-relevant events that took place since the last measure. These reports were aggregated, yielding an event score, but the influence of specific events or stressors was not tested. All three studies found positive effects of stress-relevant events on synchronous (Van Eck et al., 1996, Peeters et al., 2003) or 20 min lagged (Smyth et al., 1998) salivary cortisol measures. The fourth study (Hanson et al., 2000) used a handheld computer and asked the participants about specific momentary demands (being interrupted and experiencing time pressure, respectively), as well as satisfaction and acknowledgement resulting from their work, but found no effect of these stressors on synchronous salivary cortisol measures. Thus, while stressful events in everyday life seem to have an impact on salivary cortisol, the type of event that is most relevant is unknown. Because a meta-analysis demonstrated the effect of motivated performance situations on cortisol in laboratory situations convincingly (Dickerson & Kemeny, 2004), cortisol responses in everyday life may be expected to be most reliably elicited in motivated performance situations, i.e. tasks where an important goal is threatened.

On average emotion induction procedures in the laboratory produce equivocal results. On the basis of 16 studies included in a meta-analysis, positive as well as negative effects were observed, which balanced to a zero net effect (Dickerson & Kemeny, 2004). In contrast, all four field studies that tested the association of stressor occurrence and state negative affect in everyday life (Van Eck et al., 1996, Smyth et al., 1998, Hanson et al., 2000, Peeters et al., 2003) reported positive associations of state negative affect with salivary cortisol. All of the studies used scales to assess negative affect, thus combining a number of adjectives into factors. While this procedure is psychometrically sound and yields high measurement reliability, a test of how much specific affective states contribute to the overall association with cortisol is overlooked. Because different affective states involve different patterns of central nervous system activation (Phan et al., 2002) and the secretion of cortisol is influenced by signal generators located at several levels of the central nervous system (e.g. Buijs & Van Eden, 2000), some negative affective states may in particular be associated with HPA responses. Thus, this study investigates which kind of affective state contributes most to the impact of negative affective states on synchronously measured salivary cortisol in everyday life.

The construct of trait anxiety describes stable individual differences in the proneness to experience negative affective states like tension, nervousness, low self-confidence and worries in different situations. Given the positive associations of state negative affect with cortisol in field studies reported above, a positive association of trait anxiety with cortisol would be expected. However, results are equivocal: trait anxiety has been shown to be associated with blunted plasma and salivary cortisol levels during a laboratory public speaking situation (Jezova et al., 2004), a steeper decline in salivary cortisol over the day (Vedhara et al., 2003), and higher salivary cortisol levels in everyday life (Van Eck et al., 1996). Looking at closely related constructs like neuroticism or negative affectivity, results on associations with cortisol are also mixed. Studies found no association with neuroticism (Kirschbaum et al., 1992a, Schommer et al., 1999), positive associations of morning salivary cortisol with neuroticism (Portella et al., 2005), but reduced plasma cortisol point measures in the afternoon (LeBlanc & Ducharme, 2005), and blunted responses to a laboratory mental stress test (Phillips et al., 2005). Trait negative affect as measured by aggregated daily state negative affect measures was positively correlated to salivary cortisol (Polk et al., 2005), and negative affectivity was associated with higher cortisol responses to a mental arithmetic task (Habra et al., 2003), while two field studies found no influence of trait negative affect on salivary cortisol (Smyth et al., 1998, Hanson et al., 2000). All of these studies tested main effects, so they omitted the opportunity to test if the association between the perceived severity of the stress situation and the cortisol response varied with the level of trait anxiety or negative affectivity, e.g. if trait anxiety moderated the stressor-response association.

To summarise, salivary cortisol measured in everyday life is expected to vary as a function of state negative affect and stressor experience. This study investigates if salivary cortisol is associated with (1) real-life tasks in which (a) an important goal is threatened, i.e. performance under pressure to succeed, and (b) a task-related demand has not been met, i.e. task failure; (2) specific negative affective states; and (3) trait anxiety. Subsequently, variability in the relationship between stressors and cortisol will be tested and the role of affect in the explanation of this slope variance will be evaluated. The aim is to decide on the relative importance of a mediating effect of state negative affect and a moderating effect of trait anxiety. A state negative affect variable serves as a mediator if it explains the relationship between stressor and cortisol secretion; this is tested by including the affect predictor in a model where cortisol is regressed on performance pressure or task failure. Trait anxiety serves as a moderator if it influences the strength of the relationship between stressor and cortisol secretion; this is tested by including the stressor x trait anxiety interaction in the model.

Because cortisol measures are known to be affected by high intensity physical activity (Jacks et al., 2002), sex and age (Van Cauter et al., 1996, Seeman et al., 2001, Kudielka et al., 2004, Kajantie and Phillips, in press), nicotine (Wilkins et al., 1982, Kirschbaum et al., 1992b), caffeine (Lovallo et al., 2005), and having a meal (Quigley and Yen, 1979, Follenius et al., 1982), statistical analyses controlled for these influences. In addition, statistical analyses controlled for possible effects of thyroid or estrogen replacement medication, and use of oral contraceptives.