Influence of a Suggestive Placebo Intervention on Psychobiological Responses to Social Stress

Abstract

We tested the hypothesis that a suggestive placebo intervention can reduce the subjective and neurobiological stress response to psychosocial stress. Fifty-four healthy male subjects with elevated levels of trait anxiety were randomly assigned in a 4:4:1 fashion to receive either no treatment (n = 24), a placebo pill (n = 24), or a herbal drug (n = 6) before undergoing a stress test. We repeatedly measured psychological variables as well as salivary cortisol, alpha-amylase, and heart rate variability prior to and following the stress test. The stressor increased subjective stress and anxiety, salivary cortisol, and alpha-amylase, and decreased heart rate variability (all P < .001). However, no significant differences between subjects receiving placebo or no treatment were found. Subjects receiving placebo showed increased wakefulness during the stress test compared with no-treatment controls (P < .001). Thus, the suggestive placebo intervention increased alertness, but modulated neither subjective stress and anxiety nor the physiological response to psychosocial stress.

Introduction

Placebo interventions have the potential to modulate emotional states and central nervous system activity as well as endocrine, autonomic, and immune functions. ¹ Classical conditioning and expectation are considered to be the most prominent mechanisms of the placebo effect. ² Besides, data from placebo analgesia studies suggest that a reduction of stress and anxiety plays a crucial role for the placebo analgesic response ^{1,3 –6} and may also be relevant for placebo effects in conditions other than pain.

For example, Vase et al ⁷ investigated placebo effects on pain induced by rectal balloon distension in female patients with irritable bowel syndrome. Anxiety along with desire for pain relief and expectation accounted for 58% of the placebo analgesic effect. ⁷ Morton et al ⁶ found that subjects with high dispositional optimism showed decreased anxiety after a first exposure to placebo, which subsequently predicted the placebo analgesic response in a following session. Aslaksen et al ⁴ found that placebo administration together with the suggestion that it was a potent painkiller decreased pain and stress ratings as well as the concomitant sympathetic activation. The decrease in stress after placebo administration occurred before pain was induced, indicating that the stress reduction was part of the placebo analgesic response rather than a consequence of pain reduction. Notably, the reduction in subjective stress levels before painful stimulation also predicted smaller event-related potentials in the EEG during pain stimulation. ³

Conversely, subjects with high fear of pain and subjects in whom fear of pain was induced via expectation modulation showed increased anticipatory stress levels and reduced placebo analgesic responses. ^8,9

Information that a substance will increase pain was shown to increase cortisol responses, and thus stress levels to painful stimulation. ¹⁰ Furthermore, pharmacological blockade of anxiety by the cholecystokinin antagonist proglumide abolished the nocebo response, ¹¹ indicating that anxiety at least partially mediates the nocebo response. Thus, placebo and nocebo responses are proposed to be different ends of a continuum from very little negative emotions such as stress or anxiety to extreme negative emotions. ¹

Besides studies that highlight the importance of anxiety and stress for placebo analgesic and nocebo hyperalgesic responses, data from randomized placebo-controlled trials in anxiety disorders suggest that clinically relevant placebo effects exist in generalized anxiety disorder, ¹² social anxiety disorder, ^13,14 and panic disorder. ^{14 –16} For example, a meta-analysis of 3 studies in panic disorder patients found that combining cognitive behavioral therapy with a placebo pill increased the chance of a therapeutic response by 26% compared with cognitive behavioral therapy alone. ¹⁶ Studies further showed that a sustained placebo response in social anxiety disorder was associated with attenuated stress-related amygdala activation ^13,17 and altered amygdala-frontal coactivation patterns. ¹⁸

A functional-anatomical relationship between placebo analgesia and emotional processing was highlighted by Petrovic et al. ¹⁹ The authors demonstrated that, after prior conditioning with an anxiolytic drug, placebo administration modulated the emotional perception of affective pictures toward reduced unpleasentness. Using event-related functional magnetic resonance imaging data, the authors further demonstrated that the changes in emotional perception were accompanied by activity changes in brain regions such as the amygdala, the rostral anterior cingulate cortex, and the lateral orbitofrontal cortex. This pattern showed marked overlap with the modulatory network involved in placebo analgesia. ²⁰

It therefore appears plausible that placebo interventions aimed to reduce stress and anxiety will affect neurobiological systems involved in emotional and stress processing, for example, the hypothalamus-pituitary-adrenal axis, the sympathetic-adrenal-medullary axis, or the parasympathetic nervous system. To our knowledge, no study so far has tested directly whether placebo administration together with the information that it was a potent drug to reduce stress and anxiety modulates acute psychological as well as hypothalamus-pituitary-adrenal axis, sympathetic-adrenal-medullary axis, and parasympathetic responses to experimentally induced psychosocial stress. Therefore, we applied the Trier Social Stress Test (TSST), one of the most prominent laboratory stress paradigms, to test the hypothesis that a suggestive placebo intervention attenuates the psychological, neuroendocrine and autonomic stress response. Besides psychometric assessments, we repeatedly measured salivary cortisol reflecting activation of the hypothalamus-pituitary-adrenal axis, ^21,22 salivary alpha-amylase (sAA) reflecting activation of the sympathetic-adrenal-medullary axis, ^23,24 and heart rate variability (HRV) reflecting parasympathetic activation ²⁵ prior to and following the stress test.

Methods

Subjects

Based on a priori sample size calculation, we estimated that 24 subjects in the placebo group and 24 subjects in the no-treatment group would be needed to detect a difference in increase in salivary cortisol of 10 ± 12 nmol/L between the groups at a 2-sided significance level of P < .05, with a power of 80%. The estimation of the effect of the placebo intervention was based on the data by Fries et al, ²⁶ who found a difference in cortisol response to the TSST between verum (alprazolam) and placebo of approximately 20 nmol/L. Thus, we estimated that the difference between our placebo intervention and no treatment would be half as large as the difference between drug and placebo in the study bt Fries et al. ²⁶

Fifty-four healthy male subjects were recruited at local universities in Berlin. Women were excluded from the study because cortisol responses may vary with menstrual cycle phase and intake of oral contraceptives ²⁷ and because significant placebo analgesic responses were observed in men, but not in women in a number of previous studies. ^3,4 Eligible volunteers filled in a health questionnaire and the 20 item trait anxiety subscale of Spielberger’s State-Trait-Anxiety Inventory. Subjects with trait anxiety scores of more than 33 were enrolled. This cutoff was found to be the median for trait anxiety in a previous study involving young healthy males recruited in Berlin. ²⁸ We decided to recruit subjects with elevated trait anxiety because we assumed that such subjects would be characterized by a higher desire for anxiety relief, possibly increasing their proneness to respond to a placebo intervention aimed to reduce stress and anxiety. This assumption was based on findings from placebo analgesia studies, where desire for pain relief was shown to contribute significantly to the variance in placebo responses. ^7,29 In addition, higher levels of psychopathology (composite score of depression, anxiety for pain and neuroticism) were associated with higher placebo analgesic responses in chronic back pain patients. ³⁰ Exclusion criteria were a history or presence of physical or psychiatric disease, current medication, drug abuse, obesity or underweight, familiarity with the TSST, or current participation in another study. Subjects were instructed to abstain from alcohol and caffeinated beverages 24 hours before the experiment. An allowance of €25 (approximately USD30) was paid for participation. Our institutional ethics committee approved the study, and written informed consent was obtained from all subjects.

Study Protocol

Subjects were randomly assigned to 3 groups in a 4:4:1 fashion. In the first group (n = 24), subjects took part in the TSST without any active or placebo medication (no-treatment group). The second group (n = 24) received a placebo tablet, and the third group (n = 6) received a tablet containing 45 mg of standardized St. John’s wort extract. The verum group was introduced by vote of the Charité Ethics Committee in order to reduce the deceptive element of the study. This also allowed for double blinding the placebo and the verum arm of the study.

To detect possible placebo effects, comparisons were planned between the placebo and the no treatment group. Subjects were told that they would take part in a study on the “effects of an anxiolytic intervention on the stress response,” and that they would either receive placebo, St. John’s wort, or no treatment. The likelihood to be in either group was not disclosed.

All laboratory sessions took place at 11 am or 3 pm. After arrival, a first saliva sample was collected using Salivette tubes (Sarstedt; Nuernbrecht, Germany). Subjects in the placebo group were asked about their prior experience with St. John’s wort (no experience, having heard or read about it, or previous personal use of it). Subjects in the placebo and active medication group were then told a cover story about the beneficial effects of St. John’s wort. In detail, the suggestions were delivered by FZ-V wearing a white coat and introduced as the principal investigator of the study. The wording of the suggestion was as follows:

You may have heard about St. John’s wort, one of the most extensively investigated herbal drugs in the world. St. John’s wort has been shown to be a very effective treatment for depression, anxiety, and other stress-related complaints. It is well tolerated and hardly causes side effects. We have recently found that a single treatment with St. John’s wort effectively reduced acute stress and anxiety. Subjects treated with St. John’s wort commonly felt very calm and relaxed, but still concentrated and alert during a stress test.

We decided to use a herbal drug in the study to reduce the likelihood of increased anxiety and stress associated with the anticipation of side effects. This assumption was based on the experiences of 2 previous studies investigating the effects of suggestive placebo interventions aimed to decrease blood pressure, which found different results if a homoeopathic or a conventional drug was used. ^31,32

After delivery of the suggestions, subjects were asked to rate their expectation regarding the effectiveness of a single treatment with St. John’s wort to reduce stress and anxiety by means of a visual analogue scale (0-100 mm). Thereafter, the respective tablets were administered.

The recording session started with a 30-minute resting period. During the first 20 minutes, subjects were seated in a comfortable, high-backed chair and were allowed to read light magazines. After 20 minutes, subjects were asked to stand up to avoid orthostatic reactions potentially interfering with autonomic measures. ³³ Subjects then completed the state subscale of Spielberger’s State-Trait-Anxiety Inventory (score range from 20 to 80), ³⁴ and the Multidimensional Mood State Questionnaire, ³⁵ comprising the 3 dimensions “wakefulness/tiredness,” “good/bad mood,” and “calmness/restlessness.” Each of the 3 subscales’ score ranged from 4 to 20. Additionally, subjects rated subjective stress by means of a visual analogue scale (0-100 mm). Two minutes prior to the TSST, a second saliva sample was collected.

A detailed protocol of the TSST was described elsewhere. ³⁶ In brief, the subject was led to the TSST room where he was instructed to stand behind a microphone in front of a committee, consisting of one man and one woman wearing white coats, as well as a video camera. The experimenter instructed the subject to deliver a 5-minute speech for a job application, for which he had 5 minutes to prepare, and that a second task would follow. Subjects were informed that the committee was trained in behavioral observation, and that the whole session would be voice- and video-recorded. After the speech, a 5-minute mental arithmetic task had to solved, where the subject had to count backward from 2043 in 17-steps as accurately and quickly as possible. In case of miscalculations, the subject was asked to restart from 2043. The committee did not provide further feedback, and behaved in a cold and reserved manner. The TSST was shown to reliably induce increases of salivary cortisol ^21,22 and sAA. ^23,24

Immediately after the TSST and after further 10, 20, and 30 minutes, additional saliva samples were collected. While standing for 10 minutes after termination of the TSST, subjects again rated their perceived stress by means of visual analogue scale, and completed the state subscale of Spielberger’s State-Trait-Anxiety Inventory and the Multidimensional Mood State Questionnaire. Subjects were also asked about their subjective perception of whether they had received placebo or active medication (believed placebo, believed active medication, or do not know). During the last 20 minutes of the session, subjects were again allowed to sit down.

Assessment of Salivary Cortisol and Salivary Alpha-Amylase

To determine cortisol and sAA levels, the saliva samples were sent to Dresden (Germany) to Professor Kirschbaum’s laboratory. Free cortisol levels were measured using an immunoassay (IBL, Hamburg, Germany), and sAA levels were measured using a quantitative enzyme kinetic method as described in detail elsewhere. ³⁷ Inter- and intracoefficients of variation were both less than 10%.

Assessment of Heart Rate and Heart Rate Variability

Heart rate was continuously recorded using a Polar heart rate monitor (Polar RS800CX). Data from the Polar device were transferred to a personal computer using the Polar electro interface. Prior to HRV analysis, the measured RR interval data were preprocessed for artifacts using the Polar Precision Performance software that came with the heart rate monitor. Heart rate and HRV were analyzed in 5-minute intervals using Kubios HRV version 2.0 software. ³⁸ The root mean square of differences between adjacent normal RR intervals (RMSSD) was calculated as a time domain measure of HRV.

Statistical Analyses

Analyses were performed using SPSS Statistics 21 software (IBM Corp, Armonk, NY). Baseline characteristics of subjects in the placebo and the no treatment group were compared using t tests and chi-square tests. Psychological and physiological responses to the TSST were analyzed using analyses of variance for repeated measures with the different time points of psychological and physiological measurements as within-subjects factor, and group (placebo vs no treatment) as between-subjects factor. Greenhouse-Geisser corrections were applied if appropriate. P < .05 in 2-sided tests was considered statistically significant.

Results

Subjects

Characteristics of subjects in the placebo group, no treatment group and verum group are shown in Table 1. There were no significant differences with respect to age, body mass index, education, trait anxiety, basal cortisol, and basal sAA. The average of subjects’ ratings regarding the expected effectiveness of St. John’s Wort to reduce stress and anxiety was 47.9% ± 17.75%, indicating that our cover story induced expectations of at least moderate effectiveness. The majority of subjects (n = 22, 91.7%) in the placebo group had no prior experience with St. John’s wort, 2 subjects had heard or read about it, and none of the subjects had previously used St. John’s wort. After the TSST, 2 subjects believed that they had received active medication, 9 subjects (37.5%) believed that they had received placebo, and 13 subjects (54.2%) answered that they did not know.

Table 1.

Characteristics of Subjects in the Placebo, No-Treatment Group, and Verum Group, and Change of Psychological Measures from Baseline.

	No Treatment (n = 24)	Placebo (n = 24)	P Value (Placebo vs No Treatment)	Verum (n = 6)
Demographic characteristics
Age (years), mean ± SD	28.6 ± 7.5	25.5 ± 5.6	0.06	22.5 ± 2.2
High school graduate, n (%)	15 (62)	20 (83)	0.08	6 (100)
Body mass index (kg/m2), mean ± SD	23.3 ± 2.9	22.4 ± 2.3	0.26	24.23 ± 1.9
Psychological measures at baseline
Trait Anxiety, mean ± SD	43.88 ± 10.1	41.70 ± 6.6	0.39	39.3 ± 7.5
State Anxiety, mean ± SD	39.17 ± 7.12	37.7 ± 6.18	0.75	38.3 ± 7.00
MDMQ Wakefulness, mean ± SD	31.54 ± 6.97	28.54 ± 5.95	0.24	30.66 ± 4.54
MDMQ Calmness, mean ± SD	31.62 ± 6.46	28.54 ± 5.94	0.83	28.50 ± 6.25
MDMQ Mood, mean ± SD	31.33 ± 4.92	32.37 ± 3.94	0.37	29.66 ± 3.93
Stress VAS, mean ± SD	18.87 ± 19.22	19.58 ± 13.12	0.30	8.5 ± 10.19
Psychological measures after stress
State Anxiety, mean ± SD	55.82 ± 9.79	54.95 ± 8.14	0.75	58.17 ± 12.79
MDMQ Wakefulness, mean ± SD	28.33 ± 7.19	30.62 ± 5.11	0.39	28.33 ± 4.67
MDMQ Calmness, mean ± SD	19.58 ± 6.51	19.41 ± 5.81	0.99	19.33 ± 6.71
MDMQ Mood, mean ± SD	21.62 ± 7.21	22.29 ± 7.74	0.76	19.83 ± 6.85
Stress VAS, mean ± SD	66.41 ± 20.36	66.66 ± 20.27	0.95	69.33 ± 23.60

Abbreviations: MDMQ, Multidimensional Mood State Questionnaire; SD, standard deviation; VAS, visual analogue scale.

Physiological Measures

The TSST induced significant increases in salivary cortisol—time effect, F(4, 94.54) = 21.34, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.32—and sAA—time effect, F(4, 104.29) = 26.62, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.38. The TSST also induced significant increases in heart rate—time effect, F(14, 113.88) = 85.31, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.70—and decreases in RMSSD—time effect, F(14, 170.59) = 30.18, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.46. Thus, the TSST induced the expected increases of hypothalamus-pituitary-adrenal axis and sympathetic-adrenal-medullary axis activity, as well as decreases in parasympathetic activation within the whole study sample.

However, we found no significant time × treatment (placebo vs no treatment) interaction effects for cortisol, F(4, 94.54) = 0.72, P = .58, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.02; sAA, F(4, 104.29) = 0.36; P = .84, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.008; heart rate, F(14, 113.88) = 0.68, P = .79, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.02; and RMSSD, F(14, 170.59) = 0.77, P = .71, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.02 (Figure 1), indicating that the placebo intervention did not modulate hypothalamus-pituitary-adrenal axis activation, sympathetic-adrenal-medullary axis activation, or parasympathetic activation.

Figure 1.

Salivary cortisol (A), salivary alpha-amylase (B), heart rate (C), and heart rate variability (D) in subjects receiving placebo and no-treatment controls. TSST, Trier Social Stress Test; RMSSD, root mean square of differences between adjacent normal RR intervals. Error bars represent standard errors of the means.

The 6 subjects in the St. John's wort group showed no significant differences regarding cortisol or sAA responses to the TSST compared with subjects in the placebo group (data not shown).

Psychological Measures

The TSST induced significant increases of state anxiety—time effect, F(1, 45) = 222.2, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.83—and subjective stress—time effect, F(1, 46) = 186.18, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.80 (Table 1). Again, no significant time × condition interaction effects regarding state anxiety and subjective stress were observed—F(1, 45) = 0.07, P = .79, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.002 and F(1, 46) = 0.004, P = .95, ${\mathit{\eta }}_{\mathrm{p}}^2$ < .001, respectively. Thus, the placebo intervention did not modulate anxiety and subjective stress responses to the TSST.

Calmness—time effect, F(1, 46) = 118.37, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.72—and mood—time effect, F(1, 46) = 93.73; P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.67—significantly decreased in response to the stressor, while wakefulness was not significantly changed—time effect, F(1, 46) = 0.71, P = .40, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.02 (Table 1). No significant time × treatment effects were found for calmness and mood—F(1, 46) = 0.15, P = .70, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.03 and F(1, 46) = 0.34, P = .86, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.001, respectively (Table 1).

However, a significant time × condition effect was found for wakefulness. Wakefulness increased in the placebo group, and decreased in the no treatment group—F(1, 46) = 15.68, P < .001, ${\mathit{\eta }}_{\mathrm{p}}^2$ = 0.25 (Table 1).

The 6 subjects in the St. John’s wort group showed no significant differences regarding increases in anxiety or subjective stress after the TSST compared with subjects in the placebo group.

Discussion

Contrary to our a priori hypothesis, the suggestive placebo intervention had no impact on state anxiety, subjective stress or the physiological response to the psychosocial stressor in males with elevated trait anxiety. Subjects showed a typical pattern of increases in state anxiety and subjective stress, accompanied by neuroendocrine and autonomic arousal, which was independent of allocation to the placebo or no treatment group.

Subjects in the placebo intervention group showed an increase in wakefulness compared with no treatment, possibly indicating that the intervention modulated subjective arousal. This finding is in line with previous studies that have demonstrated significant effects of placebo interventions on subjective and physiological arousal. ^39,40 Our finding could, however, also be due to chance given the multiple testing.

Previous studies showed that reductions of stress and anxiety accompany and predict placebo analgesic responses. ^3,4,6,7 In placebo analgesia, the modulation of negative emotions results from manipulating the subjects’ expectations toward decreased pain, while in our study, the suggestions targeted stress and anxiety directly. Thus, findings from placebo analgesia studies may not translate into other paradigms such as laboratory stress. One explanation for this difference may be that the occasional use of painkillers is widespread, and previous treatment experiences were shown to shape the placebo response. ⁴¹ In contrast, the use of anxiolytic drugs is much less common in healthy subjects, and none of the subjects in the placebo group had previously used St. John’s wort.

Our null findings also appear to be conflicting with the observation of substantial placebo effects in randomized controlled trials in anxiety disorder patients. ^{12,14 –16} Possible explanations include differences in the study sample (anxiety disorder patients vs healthy subjects with elevated trait anxiety), duration of treatment (single vs continuous treatment), and outcome parameters (chronic or chronic-intermittent anxiety symptoms vs acute stress-induced anxiety). Importantly, the studies in anxiety disorder patients did not comprise a no-treatment control group. Thus, “true” placebo effects cannot be differentiated from other factors such as regression to the mean or the natural course of the disease. ⁴²

Darragh et al ⁴³ described significantly enhanced recovery of subjective stress and HRV after a serial subtraction task in subjects who had received a placebo nasal spray together with the suggestion that the spray contained serotonin. Compared with the TSST, the serial subtraction task is likely to be a less powerful inductor of stress because the element of social evaluation is largely missing. Thus, we cannot rule out the possibility that the TSST was a too stressful task to allow for discrimination of psychobiological responses of subjects receiving placebo and controls. Interestingly, Fries et al ²⁶ found that pretreatment with benzodiazepine alprazolam attenuated the hypothalamus-pituitary-adrenal axis response to the TSST, but had also no effect on state anxiety. These results suggest that the TSST is a very powerful inductor of anxiety that appears to be relatively robust even against well-established anxiolytic pharmacological interventions.

Strengths of our study include the use of the TSST as one of the most vigorous paradigms for psychosocial stress, and the simultaneous assessment of psychological measures together with assessment of markers of different biological systems involved in stress regulation.

On the other hand, our study faces several limitations. First, only healthy male subjects with elevated trait anxiety were enrolled, and the findings may thus not be generalizable to women, older subjects, and subjects with clinical manifestations of stress or anxiety. We cannot fully rule out the possibility that the recruitment of individuals with elevated trait anxiety may have led to a ceiling effect. However, based on findings from placebo analgesia studies, ^7,29,30 we assumed that higher anxiety individuals would rather be characterized by a higher likelihood to respond to a placebo intervention.

Furthermore, the introduction of the active medication group may have resulted in subjects’ uncertainty about whether receiving placebo or active medication. This may have reduced subjects’ expectations of anxiety relief and thus the placebo response. ⁴⁴

In conclusion, the suggestive placebo intervention increased subjective alertness, but did not modulate subjective stress and anxiety or the physiological response to a powerful psychosocial stressor.