The Trier Social Stress Test: Principles and practice

1. Introduction

Psychological stress can be defined as psychological tension or strain that is difficult to manage or endure (Colman, 2001). Although chronic (extended) stress or traumatic stress is detrimental to an organism's health, acute (short-term) stress is important for responding to threatening stimuli (McEwen, 2007), though the acute stress response can itself be dysregulated by chronic stress, such as when caring for a relative with dementia (e.g. Aschbacher et al., 2006). The impact of stress is visible across multiple psychological and physiological domains (McEwen, Gray and Nasca, 2015a), and chronic stress is considered a major contributor to neuropsychiatric diseases. Although much of this information comes from preclinical studies, the translational relevance is clear, especially when one considers the ubiquity of potential sources of stress in modern society. For example, the European Union report on “Psychological risks in Europe: Prevalence and strategies for prevention” indicated that a quarter of workers experience stress all or nearly all of their working time (Eurofund & EU-OSHA, 2014). There is also evidence that perceived stress in the United States population has increased over a number of decades (Cohen and Janicki-Deverts, 2012). In this context, a better understanding of the neurobiology of stress, as well as stress resilience (McEwen et al., 2015b, Russo et al., 2012), in human subjects is becoming increasingly relevant for individual and population health. Although some features of this field focus on chronic naturalistic stressors, such as unemployment (Dettenborn et al., 2010, Gallagher et al., 2016) or caring for a relative with dementia (Clipp and George, 1993, Kim and Schulz, 2008), valuable insights have also been derived from laboratory studies based on acute stress paradigms (c.f. Bali and Jaggi, 2015; for a review).

Researchers wishing to examine the neurobiological response to acute stress in humans are in need of a reliable and valid acute stressor-one which can robustly induce an acute stress response under experimentally controlled conditions. The Trier Social Stress Test (TSST; Kirschbaum et al., 1992, Kirschbaum et al., 1993) is an ecologically valid stressor, based on the stress induced by public speaking. It incorporates social evaluation and unpredictability, by obliging the person to speak in front of an unresponsive audience and completing a surprise mental arithmetic test. Hundreds of research studies have employed the TSST in order to examine the impact of acute stress on human neurobiology.

In the following review we will outline the methodology of the TSST, offering practical advice on employing this technique, and discussing advantages as well as potential pitfalls and challenges of the approach. We will discuss how age and gender can moderate the effects of the TSST (including modifications of the TSST for younger populations). The impact of genetic factors on acute stress response as assessed with the TSST will be outlined (including both genotype and epigenetic factors). Finally, we will discuss potential future directions for research in this area.

2. Nature of the stressors

To study acute stress in humans, it is critical for researchers to have a valid and reliable method for inducing acute stress under controlled conditions. The TSST was developed over two decades ago with the primary aim to design such a task, which would produce a consistent hypothalamic-pituitary-adrenal (HPA) axis response in humans (Kirschbaum et al., 1993). Activation of the HPA axis represents the core neuroendocrine response to stress in humans and other mammals, and dysfunction of this axis is associated with a range of physical and mental health disorders (Pariante and Lightman, 2008). Other experimental stress procedures, including cognitive tasks such as the Stroop, public speaking tasks (without the other aspects of the TSST), noise exposure, emotion induction or pain induction, tend to produce variable, or no, HPA axis responses in humans. A meta-analysis of a number of laboratory-based acute psychological stress procedures, including the TSST, determined that the key psychological elements in motivated performance tasks that produce the greatest HPA axis response are a combination of social-evaluative threat and uncontrollability (Dickerson and Kemeny, 2004). The key procedural stages of the TSST incorporate each of these elements (see Fig. 1). In the standard TSST procedure, participants take part in a role-playing scenario, the most common feature of which is imagining they have applied for a job of their choice (or in some cases a pre-specified job), and after a short preparation period, must present to a panel of ‘committee members’ why they believe they are the best candidate for the particular position. Following the speech task, they are then given a surprise mental arithmetic task (c.f. Allen et al., 2014). Performing each task in front of a panel of committee members introduces the element of social-evaluative threat which is enhanced by their performance being video and voice recorded (purportedly for further analysis), the participant being informed that the committee are experts in behaviour analysis, and the committee members being trained to withhold any type of social engagement or positive feedback; they do not smile or nod, and stick closely to scripted instructions and responses. The TSST incorporates a number of elements of uncontrollability throughout the procedure; participants do not know what their task is until 3–10 min (depending on the protocol used) prior to doing it; they have a very short period of time to prepare for the task; the mental arithmetic component is completely unexpected; and the committee members are non-responsive to any attempts at social engagement.

In healthy humans, the TSST has been shown in a multitude of studies to activate the HPA axis as measured via salivary, plasma or serum cortisol and adrenocorticotrophic hormone (ACTH) (Kudielka et al., 2007). The TSST reliably induces a two-to-three fold increase in cortisol levels in approximately 70–80% of study participants (Dickerson and Kemeny, 2004, Frisch et al., 2015, Kudielka et al., 2007). In addition, the TSST has been employed to examine the effects of acute psychosocial stress on a range of neurobiological and other measures including immune parameters (Campisi et al., 2012, Izawa et al., 2013a, Izawa et al., 2013b, Slavich et al., 2010, Yamakawa et al., 2009), the sympathetic nervous system (El-Sheikh et al., 2008, Gold et al., 2004, Gordis et al., 2006, Graeff et al., 2003, Het et al., 2009, Jezova et al., 2004), and the cardiovascular system (Childs et al., 2006, Rimmele et al., 2007). The TSST has also proven crucial in understanding the demographic and psychological characteristics that moderate the neurobiology of the acute stress response in humans, such as age and gender (see sections 4, 5 below), personality factors such as neuroticism (Oswald et al., 2006), individual appraisals of the stress environment (Gaab et al., 2005) and social status (Gruenewald et al., 2006). See Allen et al. (2014) for an extensive review.

The TSST has also progressed our understanding of the aberrant neurobiological mechanisms which may underlie stress-related disorders. For example, individuals with a diagnosis of depression have been shown to exhibit elevated proinflammatory cytokine levels (Fagundes et al., 2013, Pace et al., 2006) in response to the TSST, as well as an altered HPA response to the TSST, evident in a higher net increase in cortisol compared to participants at risk of depression (Dienes et al., 2013), although this effect may be moderated by gender, with depression being associated with a higher cortisol response in females specifically (Chopra et al., 2009). Although findings in this area are not always consistent (Young et al., 2000), they do lend support to key neurobiological theories of depression, including those highlighting the role of immune activation (e.g. Maes, 1995, Leonard, 2000) and the HPA axis (e.g. Pariante and Lightman, 2008). Attenuated cortisol responses to the TSST have been reported in panic disorder (Petrowski et al., 2010, Petrowski et al., 2013) and individuals who have experienced early traumatic experiences and received a psychiatric diagnosis, have elevated HPA axis, cardiovascular (Heim et al., 2000) and plasma interleukin 6 (Carpenter et al., 2007) responses to the TSST. People with irritable bowel syndrome, a stress-related brain-gut axis disorder (Eisenstein, 2016, Grayson, 2016, Kennedy et al., 2012) have exhibited altered HPA axis activation in response to the TSST (Kennedy et al., 2014, Suárez-Hitz et al., 2012).

In summary, the TSST combines the key elements of social-evaluative threat and uncontrollability to produce a consistent and robust physiological and psychological stress response in humans. The reliable, robust and modifiable nature of the acute stress response produced by the TSST has led to this procedure being considered the gold standard in human experimental stress research and becoming a mainstay in stress laboratories worldwide.

3. Practical issues and adapted versions

As noted, the fact that the TSST induces a consistent and reliable physiological and psychological stress response in the majority of participants is a major advantage of this procedure. The experimental conditions during the TSST are highly standardized, and controllability is enhanced due to the availability of an extensive literature on the various environmental, psychological and physiological factors which can affect the acute stress response to the TSST. This controllability also allows an investigator to measure multiple psychological and biological outcomes prior to, during, and after the acute stress procedure. Moreover, despite being an experimental, simulated procedure, the ability to sell one's qualities in an interview scenario is a common experience for many individuals at some point in their life, which lends ecological validity to the TSST. Finally, the TSST requires little specialist equipment (only a microphone, video camera etc.) and as such, the main costs associated with the procedure will accrue depending on the biological readouts which the investigator wishes to measure.

Innovative adaptations of the original TSST have been introduced which confer relative advantages. For example, a TSST for groups (TSST-G) has been introduced which increases the throughput of study participants, and has been shown to produce a significant HPA axis and autonomic nervous system response (Boesch et al., 2014, von Dawans et al., 2011). A virtual reality version of the TSST (TSST-VR) has also been designed (Kotlyar et al., 2008) which removes any variability in data due to interactions between the study participant and committee members. However, it seems the type of VR technology utilized is important in producing a robust stress response, as an earlier version of the TSST-VR did not significantly increase cortisol levels (Kotlyar et al., 2008), whereas a more recent version, using newer technology, induced a measurable neuroendocrine response (Jordanova et al., 2007). Although the standard TSST is not well-suited to brain imaging paradigms, the Montreal Imaging Stress Task (MIST) has been developed, which can be combined with PET and fMRI (Dedovic et al., 2005). Participants must complete an arithmetic task; in the stress condition, these are manipulated to be slightly above the participant's level of mathematical skill. However, the MIST does not include a job interview task component.

The recently developed socially-evaluated cold pressor test (SECPT; Schwabe et al., 2008), is a viable alternative to the TSST. It is a shorter procedure (stress exposure lasts approximately 3 min) and requires only one committee member. In addition, the SECPT produces a significant HPA axis response, in addition to increasing other stress parameters such as blood pressure and subjective stress (Schwabe et al., 2008) and has recently been adapted for groups (Minkley et al., 2014), which further increases the throughput of study participants, and has been indicated to induce heightened HPA axis responsivity across repeated visits. However, as the SECPT involves thermal pain as a stressor, it is not possible to separate the psychologically and physically stressful aspects of the procedure, and it arguably lacks the ecological validity of the TSST. As such, the TSST is currently the best available experimental stress procedure for investigating the acute psychosocial stress response in humans.

4. Impact of age

5. Studies in males and females

A number of stress-related disorders have a higher prevalence in females than in males, such as generalised anxiety disorder (e.g. Baxter et al., 2013), major depression (e.g. Ferrari et al., 2013) and irritable bowel syndrome (e.g. Lovell and Ford, 2012). There has been much interest in gender differences in acute stress response, and the TSST has frequently been used to assess such gender differences (Eisenberger et al., 2007, Uhart et al., 2006). A recent large study (N = 798) of young adults aged 18 found that males had consistently higher plasma ACTH and salivary cortisol than females, and females not on oral contraceptives had higher ACTH than those taking these contraceptives. Males were more likely than females to be reactive responders, who have increased cortisol in response to the event, and less likely than females to be non-responders, whose cortisol levels do not increase. However, a similar level of males and females were anticipatory responders, who have high baseline HPA activity levels, suggesting an anticipatory mounting of a response (Herbison et al., 2016). It should be noted that non-response may not simply be due to a lack of subjective stress appraisal, and can occur in the context of conditions associated with chronic stress (e.g. adverse early life events; Elzinga et al., 2008; panic disorder; Petrowski et al., 2010). Time series analysis indicated that plasma ACTH and salivary cortisol in males had longer peak latencies, higher post-stress peaks and more intense post-peak decline than females, which seems to suggest that although male HPA axis activity peaks at a higher level, such activity recovers to baseline levels to a similar extent as in females (Lopez-Duran et al., 2014).

In contrast to findings on ACTH and cortisol, there was not a moderating effect of gender on DHEA or DHEA-S response to the TSST, although DHEA changes were associated with changes in cortisol and ACTH (Lennartsson et al., 2012a). A lack of gender differences in salivary alpha-amylase has been demonstrated (Hidalgo et al., 2012), and there were unclear findings for a gender difference in prolactin response (Lennartsson and Jonsdottir, 2011). Although women had higher salivary immunoglobulin A than men at baseline, both genders showed similar increase and recovery in this immune marker (Birkett et al., 2016).

Sex hormones have also been examined in TSST research examining acute stress and gender. This has been examined directly, by assessing hormone levels, as well as indirectly by scheduling women to perform tests at different phases of the menstrual cycle. When women were tested in the follicular phase of their menstrual cycle (when progesterone levels are most similar to men) men still had greater ACTH and cortisol following the TSST. Testosterone negatively correlated with salivary cortisol response, and progesterone negatively with salivary ACTH and cortisol response (Stephens et al., 2016), although another study found that testosterone and estradiol were elevated by the TSST, as were ACTH and cortisol (Lennartsson et al., 2012b). Another study found that women in the follicular phase, but not the luteal phase, showed a significant increase in cortisol in response to the TSST, and cortisol levels were only associated with impaired emotional retrieval during the follicular phase (Maki et al., 2015) The changes in stress response across the menstrual cycle were not significant in another study (Herbison et al., 2016). Women with premenstrual syndrome have been found to have blunted cortisol reactivity throughout the menstrual cycle compared to control women (Huang et al., 2015). Other research has found that the higher cortisol response to the TSST in males is reduced by controlling for sex hormone levels (Juster et al., 2016). It has been argued that epinephrine may play a greater role in hemodynamic reactivity, given that epinephrine reactivity but not estradiol predicted cardiac and vascular resistance indices (Gordon and Girdler, 2014).

In a fear conditioning study, the TSST was used to assess the impact of stress on differentiation of a conditioned stimulus from a stimulus unpaired with the unconditioned stimulus in the nucleus accumbens (for early acquisition) and in the amygdala and anterior cingulate (for late acquisition); differentiation was reduced in men but increased in women taking oral contraceptives (Merz et al., 2013). Women performed better than men at a verbal learning test in a no-stress control, but men performed equally well for a TSST condition (Espin et al., 2013). Differences in emotional cognition may underlie gender differences in TSST response-females who read genuine negative news stories showed a greater cortisol response to the TSST than those who read neutral reports, but this effect was not evident in men (Marin et al., 2012). Furthermore, females displayed better recall of negative news than their male counterparts.

Given its psychosocial nature, it is plausible that gender socialisation could play a role in the neurobiological response to the TSST, with reduced differentiation in sex roles dampening difference in psychological response to the TSST (Dedovic et al., 2009). The gender of the TSST panellists is also relevant; there is evidence in young adults that a cortisol increase only occurred when the TSST panellists were of the opposite gender (Duchesne et al., 2012). During the TSST, men engage in more displacement behaviours, i.e. behaviours that do not have an apparent relevance to the situation, and so may divert attention from the stressful stimulus (Troisi, 2002). Within men, displacement was associated with reduced self-reported stress, fewer cognitive errors and a trend towards lower heart rate (Mohiyeddini, Bauer and Semple, 2013a). In contrast, high levels of public self-consciousness in healthy women were associated with displacement behaviour increasing levels of reported stress and poorer cognitive performance (Mohiyeddini, Bauer and Semple, 2013b).

There is no consistent evidence to suggest that there are sex differences in children/adolescents in their reactivity to the TSST-C. Kudielka and Kirschbaum (2005) reported no difference in cortisol reactivity to the TSST-C between male and female children, and Sumter et al. (2010) also reported no sex difference in cortisol responses to the Leiden PST, a variant of the TSST-C, although others have shown differences at some ages. For example, differences observed between 13 year old boys and girls are not evident in 11 year old boys and girls (e.g. Gunnar et al., 2009, Klimes-Dougan et al., 2001). The inconsistencies may be attributed to sex and age differences in sensitivity to social evaluation (Gunther Moor et al., 2010) and these effects of developmental stage on stress reactivity warrant further study, with particular attention to the meaning of task demands, pubertal stage and psychosocial development.

There is some evidence that sexual orientation modulates the endocrine response to acute stress, and that this effect is sex-dependent. Whereas lesbian or bisexual women had greater cortisol reactivity post-stressor, homosexual or bisexual men had lower cortisol concentrations throughout testing (Juster et al., 2015). Participants were exposed to a TSST modified to ensure maximisation of sex differences in stress response (the panel were placed behind a one-way mirror; this is associated with reduced HPA axis response compared to standard TSST in heterosexual females but not males). However, another study did not find an impact of sexual orientation on salivary cortisol in men (Jacobson et al., 2016). Sexual orientation may exert a greater influence in regions where social stigma towards homosexuality is particularly high; structural stigma (operationalised using factors such as level of exclusion of homosexual people from social institutions) was associated with a blunted response to the TSST in homosexual and bisexual young adults (Hatzenbuehler and McLaughlin, 2014).

6. Genetics and the TSST

7. Limitations of the TSST

As with any experimental procedure, there are practical limitations to using the TSST. One of the major limitations of the TSST is that there is a high degree of habituation of the HPA axis response with repeated exposures (Pruessner et al., 1997, Schommer et al., 2003). As such, it has limitations when applied to intervention studies in which it is necessary to determine a pre- and post-treatment HPA axis response, as it is difficult to parse the treatment and habituation effect. Measures of autonomic nervous system show less habituation (Gerra et al., 2001, Jönsson et al., 2010, Schommer et al., 2003, von Kanel et al., 2006) and may be more suitable for intervention studies. However, it is important to note that modifications should be made to the speech and mental arithmetic task to be performed during each repeated TSST session (e.g. requiring participants to perform serial subtraction on a different set of numbers) to minimize any habituation effects (Kirschbaum et al., 1995, Petrowski et al., 2012, Schommer et al., 2003).

Although the TSST is a highly controlled procedure, and committee members will normally be well-trained prior to an experiment beginning, nuances in interactions between a specific participant and committee member cannot be controlled, as with any social interaction, and may introduce some variability to the data. In addition, dependent on the resources available to the investigator, there may be limitations on the number of participants who can be tested each day/week, as the procedure is quite labour-intensive. Depending on the requirements of a particular study, the adaptations mentioned previously can help to overcome some of these limitations; for example the TSST for groups can increase throughput and the TSST-VR can reduce variability in participant/committee interactions.

By combining a number of stressful components, the TSST induces moderate stress in a majority of participants more reliably than its constituent components (e.g. public speaking) in isolation (Kirschbaum et al., 1993). However, this leads to a difficulty in disentangling the specific role of different aspects of the stressor in inducing stress. This is a problem for a number of methods of inducing acute stress (e.g. the socially evaluated cold pressor includes both physical pain and social evaluation). Markers of stress which can be measured throughout TSST exposure, and which respond in a more time-locked manner to a stressor than cortisol (e.g. heart rate) will help in determining which aspect of the TSST is inducing a greater deal of stress for the participant.

8. Conclusion and perspectives

The TSST has allowed for a deeper understanding of the neurobiology of the acute stress response under both normal and pathological conditions, as well as how this is moderated by factors such as gender, age and genetics. Numerous adaptations of the TSST have been adapted that have allowed for acute stress to be assessed in various contexts. Nonetheless, given its extensive demonstration of acute stress across multiple laboratories, the TSST remains the gold standard of assessment of acute stress under laboratory conditions. Enhanced consistency of methodology across laboratories will ensure greater comparability of data, and increased consistency of outcomes assessed will allow for more powerful analyses to be conducted, for example through meta-analyses.

The TSST provides a flexible platform for employing technological advances in the assessment of physiology (e.g. in human epigenetics). There are already outcomes of the TSST that are under-explored; performance during the TSST (e.g. verbal performance on the interview speech, non-verbal behaviour during the interview speech) has, with some exceptions, generally not received much research interest, despite being a potential source of rich behavioural data. Nonetheless, a method such as the consensual assessment technique (Amabile, 1983) could be used to assess the interview speeches (or stories told in the TSST-C) on criteria such as creativity. Future research employing the TSST within longer-term longitudinal studies will allow for a deeper understanding of the impact of ageing upon the stress process.

Study of the neurobiology of the acute stress response may help us to better understand how chronic stress impacts upon the acute stress response. Related to this, the TSST provides an experimental platform to investigate the neurobiology of chronic stress-related disorders such as generalised anxiety disorder, major depression and irritable bowel syndrome, including maintaining factors for stress-related disorders, as well as the efficacy of pharmacological or psychological approaches to managing stress and alleviating stress-related disorder. For example, mindfulness-based interventions have been shown to lead to attenuated psychological responses to acute stress in patients with generalised anxiety disorder (Hoge et al., 2013), and partially-remitted depression (Britton et al., 2012), as well as in healthy volunteers (Creswell et al., 2014), although the latter study observed a heightened HPA axis response to TSST following meditation, particularly in those with low dispositional mindfulness.

Although the neurobiological study of acute stress in humans has developed tremendously in recent decades with advances in neuroscience and its cognate disciplines, full elucidation of processes such as acute stress responses is likely to take many decades of further work and technological advance. A controlled, effective acute stressor is invaluable in this process, and we believe that the TSST has many advantages in this regard when used in an informed way in specific contexts. Equally, it should be recognized that data from the TSST may need to be paired with information from other sources (e.g. naturalistic stressors, SECPT) for maximum insight. To better understand causal mechanisms in the neurobiology of stress, integrating the observations from TSST in human populations with preclinical data will be crucial. The neurobiological processes involved in the TSST response in health and disease depend heavily on the impact of the environment across the lifespan, and are influenced by a variety of factors. The role of the gut microbiome therein is currently underappreciated. Given evidence for changes in the composition of the gut microbiota in stress-related disorder, such as depression (Jiang et al., 2015, Kelly et al., 2016, Naseribafrouei et al., 2014) and irritable bowel syndrome (Jeffery et al., 2012), it is of interest if different microbial compositions within individuals may be predictive of altered reactivity to the TSST. The implications of improving our knowledge in this area, for both stress vulnerability and resilience, are clear. This strategy offers benefits for both the maintenance of health and for the management of stress-related neuropsychiatric disorders.