The stress–response-dampening effects of placebo

Hormones and Behavior 59 (2011) 465–472

Contents lists available at ScienceDirect

Hormones and Behavior j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / y h b e h

The stress–response-dampening effects of placebo Iris M. Balodis a,⁎, Katherine E. Wynne-Edwards b, Mary C. Olmstead c a b c

Department of Psychiatry, Yale University, New Haven, CT, USA Department of Comparative Biology and Experimental Medicine, University of Calgary, Calgary, AB, Canada Department of Psychology, Queen's University, Kingston, ON, Canada

a r t i c l e

i n f o

Article history: Received 9 September 2010 Revised 13 January 2011 Accepted 18 January 2011 Available online 24 January 2011 Keywords: Psychosocial stress Cortisol Alpha-amylase Alcohol Risk-taking Reward Placebo effect TSST

a b s t r a c t This experiment used both biological and self-report measures to examine how alcohol modifies stress responses, and to test whether the interaction between these two factors alters risk-taking in healthy young adults. Participants were divided into stress or no-stress conditions and then further divided into one of three beverage groups. The alcohol group consumed a binge-drinking level of alcohol; the placebo group consumed soda, but believed they were consuming alcohol; the sober group was aware that they were not consuming alcohol. Following beverage consumption, the stress group was subjected to the Trier Social Stress Test (TSST) while the no-stress group completed crossword puzzles; all participants subsequently completed a computerized risk-taking task. Exposure to the TSST significantly increased salivary levels of the hormone cortisol and the enzyme alpha-amylase, as well as subjective self-ratings of anxiety and tension. In the stress condition, both placebo and intoxicated groups reported less tension and anxiety, and exhibited a smaller increase in cortisol, following the TSST than did the sober group. Thus, the expectation of receiving alcohol altered subjective and physiological responses to the stressor. Neither alcohol nor stress increased risk taking, however the sober group demonstrated lower risk-taking on the computer task on the second session. These findings clearly demonstrate that the expectation of alcohol (placebo) alters subsequent physiological responses to stress. © 2011 Elsevier Inc. All rights reserved.

Introduction The idea that alcohol may alleviate stress while simultaneously increasing risk-taking behavior is a popular notion, but difficult to test in the laboratory. While stress reliably increases both physiological and subjective measures of stress (Kirschbaum et al., 1993), there is less consensus on how alcohol affects these measures. Using changes in cortisol levels as a physiological indicator of stress, alcohol intoxication may increase (Schuckit et al., 1987), decrease (Waltman et al., 1993) or have no effect (Davis and Jeffcoate, 1983) on stress responses. Factors such as the nature of the stress manipulation and whether alcohol is administered before or after the stressor may explain some of these differences but no one explanation can fully account for the ambivalent findings (Josephs and Steele, 1990; Sayette, 1993). Identifying whether alcohol intoxication alters stress reactivity has implications for understanding a number of disorders, particularly those that are characterized by risk-taking behaviors. For example, substance abusers without co-morbid psychiatric disorders and heavy social drinkers both exhibit blunted hypothalamic–pituitary–adrenal (HPA) axis responses to alcohol (Contoreggi et al., 2003; King et al., ⁎ Corresponding author at: Yale School of Medicine, Department of Psychiatry, 2 Church St. South, Suite 215F, New Haven, CT 06519, USA. E-mail address: [email protected] (I.M. Balodis). 0018-506X/$ – see front matter © 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.yhbeh.2011.01.004

2006). Individuals with Attention-Deficit-Hyperactivity Disorder and children of alcoholics both demonstrate a propensity to engage in risky behaviors and also show a blunted cortisol response following a stressor (King et al., 1998; Sorocco et al., 2006). In healthy individuals, low cortisol levels are related to punishment insensitivity and reward dependency, two characteristics that increase risk-taking (van Honk et al., 2003). An interaction between stress reactivity and alcohol intoxication may explain why the effect of alcohol on risk-taking is so difficult to establish in the lab. Some studies report increased risk-taking in intoxicated individuals (George et al., 2005; Lane et al., 2004), some report that intoxicated participants perform less impulsively (Ortner et al., 2003), whereas others report no effect of alcohol on impulsivity and risk-taking in a decision-making task (Balodis et al., 2006). Although stress levels were not specifically manipulated in these studies, testing participants in lab settings is very likely to increase hormonal markers of stress, at least in some individuals. The apparent discrepancies in results, therefore, may be explained by individual differences in stress responsivity, which then interact with behavioral measures of risk-taking. The purpose of the following experiment was to understand the relationship between alcohol intoxication, stress and risk-taking. The data were collected as part of a larger study; the first portion of that dataset examined the relationship between physiological and subjective measures of stress by comparing different methods of analyzing

466

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

biochemical changes (Balodis et al., 2010). Here, we present the second portion of the dataset examining how alcohol and stress alter performance on a risk-taking task. In a controlled laboratory setting, we tested the hypothesis that individuals who show less stress reactivity (as measured by smaller increases in cortisol, alphaamylase and self-report anxiety levels in response to a stressor) are more likely to take risks, and that this behavior is moderated by alcohol. Healthy participants were examined in order to establish the role of stress in alcohol-induced risk-taking, in the absence of any pathology. Risk-taking was assessed using a computerized task in which alcohol dose-dependently increases this maladaptive behavior (Lane et al., 2004). Changes in salivary cortisol and alpha-amylase provided a standard measure of stress reactivity. The protein alphaamylase can be detected noninvasively and acts as an index of sympathoadrenal medullary system activity, an extra-hypothalamic stress response system activating norepinephrine (Chatterton et al., 1996). Like cortisol, alpha-amylase levels increase following acute stressors and have a distinctive circadian rhythm (Rohleder et al., 2004). One study reported that alcohol intoxication decreases salivary alpha-amylase levels (Enberg et al., 2001), however, no control condition was used in this study; to date, no study has examined the interaction of stress and alcohol on alpha-amylase levels. In line with previous work (Lane et al., 2004), we predicted that alcohol would increase risk-taking in a computer-simulated task. We also hypothesized that alcohol would blunt the cortisol response to a psychosocial stressor. Given that low cortisol levels are associated with greater impulsive choice and reward dependency (Takahashi, 2004; van Honk et al., 2003), we also hypothesized an inverse relationship between stress reactivity and risk-taking. Specifically, individuals with the greatest change in biological stress markers (i.e., cortisol and alpha-amylase) would demonstrate less risk-taking following a stressor. Conversely, individuals with lower stress reactivity would demonstrate increased risk-taking and alcohol would enhance this effect.

(SD = 4.3) drinking occasions per month, with a mean of 5.3 (SD = 2.4) drinks consumed on these occasions. Full descriptions of drinking and drug use habits in a similar population of college students are published elsewhere (Balodis et al., 2009; Balodis et al., 2010). Exclusion criteria included any cardiovascular condition, head injury or stroke, migraines, or current use of medications that may interact with the use of alcohol. The age of participants ranged from 19 to 27 years with a mean age of 20. To ensure minimal risk of pregnancy, females were tested only during the menses phase of their menstrual cycle or if they had not had sexual intercourse since their last menstrual cycle. All participants were asked not to eat for 3 hours prior to the experiment and all were tested between the hours of 14:00 and 19:00. Course credit, or an honorarium of $15 was awarded to all participants upon completion of the experiment. In addition, participants were allowed to keep money earned on the second session of the risk-taking task. This experimental protocol was approved by the General Research Ethics Board of Queen's University. Measures Multiple measures of physiological and subjective indices of stress were acquired in order to examine similarities and differences in stress parameters. Self-report measures Profile of mood states. The Profile of Mood States (POMS; (McNair et al., 1981) is a self-report adjective checklist of 65 items, rating current mood states. The mood dimensions are measured on six subscales: tension-anxiety (0–36 raw score range), depressiondejection (0–59 raw score range), anger-hostility (0–44 raw score range), vigour-activity (0–31 raw score range), fatigue-inertia (0–28 raw score) and confusion-bewilderment (0–28). The tension-anxiety scale, which consists of such items as “tense,” “on edge” and “nervous” has been shown to be a valid measure of psychosocial stress.

Methods These data were collected as part of a larger study examining physiological and subjective responses to stress. The first part of the data set was presented in a complementary study (Balodis et al., 2010) in which we compared different ways of assessing biochemical responses to stress and how these relate to self-report measures. This analysis verified the importance of a recovery period following arrival in the laboratory, in order to establish a stable baseline prior to measuring physiological responses. The study further revealed that biochemical responses to stress are assessed, most accurately, by percent change score and Area Under the Curve (AUC) measures, particularly as these correlate with subjective measures of stress. The current study builds on these results by presenting new data on physiological (cortisol and alpha-amylase) and subjective (Profile of Mood States and Kurz-Skala Stimmung/Aktivierung) measures of stress under three conditions (sober, placebo and intoxicated). The study also presents novel findings on performance on a risk-taking task under these conditions. All measures are described in detail below.

Kurz-Skala Stimmung/Aktivierung. The Kurz-Skala Stimmung/Aktivierung or KUSTA (Wendt et al., 1985), is a visual analogue scale with a 17-point bipolar scale for mood (depression/happiness), activity (tired/fresh), and tension (irritable/relaxed) and unipolar ratings for joy, anxiety, anger and disappointment (not at all/extreme). This measure was used to assess the mood state of participants prior to and immediately following the psychosocial stressor. Manipulation check. Following the second session of the risk-taking task, participants completed a self-report measure to examine the perceived level of intoxication and to judge the effectiveness of the placebo manipulation. Physiological measures Breathalyzer. Blood alcohol concentrations were estimated through Breath Alcohol Level (BAL) using a handheld breath alcohol testing instrument: the Intoxilyzer 400D CMI, Inc., Owensboro, Kentucky. Participants blew air through a mouthpiece into a fuel cell which measured the alcohol concentration in the expired breath.

Participants Participants were 87 healthy undergraduate students (29 males, 58 females) from Queen's University, recruited through classes and a student volunteer subject pool. Eligibility criteria for the study included that the participant 1) reported no allergic reactions to alcohol; 2) had no medical conditions that contraindicated alcohol; and 3) reported consuming alcohol at least once per month. Although the minimum criteria for alcohol consumption was drinking at least once per month, actual rates reported by students were a mean of 6.08

Salivary cortisol and αlpha-amylase. Saliva samples for cortisol and alpha-amylase measurement were collected by having participants passively drool through a short straw into a polypropylene vial. No saliva flow stimulant was used. Each participant provided 3 samples, collected at approximately 75 min, 100 min (immediately preceding the TSST) and 140 min (immediately following the TSST) into the experimental session (see Fig. 1 for details). For the current experiment, the sample collected 75 min into the experimental session was used as a baseline, based on previous research

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

467

Fig. 1. Graphical representation of experimental tasks and timeline. Self-Report Measures = Barratt Impulsiveness Scale, Personal Drinking Habits, Drug Use Questionnaire, Comprehensive Effects of Alcohol Questionnaire; POMS = Profile of Mood States; KUSTA = Kurz-Skala Stimmung/Activierung; BAL = Blood Alcohol Level.

demonstrating that a recovery period, following arrival in the laboratory, is necessary to establish a stable baseline for physiological measures (Balodis et al., 2010). The samples were capped, labeled and frozen at −20 °C in a non-self-defrosting freezer. Assays were completed using an expanded range high sensitivity salivary cortisol enzyme immunoassay kit (Salimetrics, State College, PA) and a salivary α-amylase assay kit (Salimetrics, State College, PA) to determine free cortisol and alpha-amylase concentrations in the samples. Behavioral measures Risk-taking task. The Risk-Taking Task is a computer task on which intoxicating levels of alcohol have been shown to dose-dependently increase risky choices (Lane et al., 2004). The task consists of a twochoice procedure in which participants choose between two panel buttons, associated with risky and non-risky payoffs. The letters C and A appear near the middle of the screen, with a question mark under the A. This latter option represents the risky choice, associated with gains or losses of $0.25, $0.50, $0.75 or $1.00 equiprobable on each trial. A box under the letter C displays “$0.02” and is considered the non-risky option, with each trial producing a gain of $0.02. After reading the instructions for the task, participants completed a 50-trial training session. Following the psychosocial stressor, participants completed a 100-trial session of the task. To increase motivation on the task, participants were informed that they could keep the money they earn on the second (100-trial) session of the task. Procedures Alcohol administration This study followed the protocol of alcohol administration previously used in our laboratory to examine the effects of alcohol on computer tasks (Balodis et al., 2007, 2006). Participants were weighed at the beginning of the session so that each participant would receive the same dose of alcohol. The intoxicated group received 3 alcoholic drinks consisting of a 2:1 ratio of Fresca soda to Vodka (40% alcohol), so as to raise their blood-alcohol level (BAL) to 0.08%, the legal limit in Ontario. A 76 kg participant, therefore, would receive 169 mL of vodka, mixed with 337 mL of Fresca, a total volume of approximately ½ L. Participants in the placebo group were told that they were receiving alcohol, although their drinks consisted only of Fresca soda mixed with flattened tonic water. The glasses of the placebo group were also rimmed with alcohol, so as to convincingly smell of alcohol when lifted to drink, without affecting the BAL of the participants. The control group received 3 drinks of Fresca soda and was informed that they were not receiving alcohol. Participants watched two episodes of ‘The Simpsons’ (totaling ~45 min) while consuming the drinks. The appropriate alcohol mixture was divided into 3 glasses, each consumed at 10–15 min intervals. Prior to

beginning the task, individuals in the placebo and intoxicated groups had their BAL measured using a Breathalyser. All participants in a session were assigned to the same beverage group: sober, placebo or intoxicated. A graphical representation of the procedure and all of the experimental tasks is presented in Fig. 1.

Psychosocial stressor Psychosocial stress was induced by having participants perform the Trier Social Stress Test (TSST) in which the individual is asked to perform tasks in front of an audience (Kirschbaum et al., 1993). A total of 51 participants participated in this part of the experiment [n = 15 sober (5 male), n = 12 placebo (2 male) and n = 24 (4 male) intoxicated]. In the first stage of the task, the participant entered a different room in which a panel of 3 people was introduced to them. The participants were told that they had 10 min to prepare a 5-min mock-job talk for a position as a research assistant. To increase anticipatory stress, the participants were told that their speech would be videotaped and that one of the panel members, trained to monitor nonverbal behavior, would later examine body language and do a voice frequency analysis of the taped session. During the 10-min preparatory time, participants were provided with a pen and paper to outline their talk, however, just before giving the speech, any written notes prepared were taken by one of the panel members and the individual was told that their written concepts would be compared to the actual content of the presentation. Participants were then asked to stand on an ‘X’ marked on the floor and deliver the speech with the video camera set to record mode and the LCD monitor rotated around so that the participant could also see their own face in the frame. The participant faced the panel members who each had a clipboard and took notes while the speech was delivered. If the participant finished before the 5 min were up, they were informed by the panel member holding a stopwatch that they still had X min left. If the participant had nothing left to say, following a 10 second pause, the panel members began to ask prompting questions such as “What personality characteristics are important for working in a lab environment?” or “How have your grades been this past year?” Following the presentation (time + 15 min), participants were told that they would be required to perform one more task to verify their alertness, but not to worry because the task was very easy and most people had no problems with it. The mental arithmetic component lasted 5 min in which participants were asked to serially subtract prime numbers from four digit numbers (e.g. 1223 by 17) as quickly and as accurately as possible. If the participant made a mistake, they were corrected and asked to start again from a different number. After 5 min, participants were thanked and sent back to the laboratory. The TSST lasted 20 min and reliably activates HPA axis and Sympathetic Adrenal Medullary axis activity, including an increase in salivary cortisol, and alpha-amylase levels in humans (Kirschbaum et al., 1993; Nater et al., 2005).

468

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

In the no-stress control condition, the experimental protocol was identical, with the exception that these individuals (n = 36) completed crossword puzzles for 20 min, instead of the TSST. In this group, n = 9 (4 males) were in the sober group, n = 9 (2 males) were in the placebo group and n = 18 (12 males) were in the intoxicated group.

Results Manipulation check

Data was analyzed using SPSS version 16.0 for Macintosh computers. Statistical significance was set at a probability of p b 0.05. The dependent measures were analyzed as follows: POMS and KUSTA ratings Scores from the POMS and KUSTA were correlated with cortisol levels as well as with scores on the risk-taking task. Salivary cortisol and alpha-amylase Cortisol and alpha-amylase response to stress are reported as the percent change as well as in the area under the curve in order to minimize the effect of individual differences and to focus on the changes in these levels (Balodis et al., 2010). Physiological measures were analyzed using planned comparisons with beverage group (intoxicated, placebo, sober) and stress condition (stress vs. no stress) as the between-subjects factors. Percent change. Percent change in physiological measures was calculated using the following formula:   sampleðiF1Þ FsampleðiÞ sampleðiÞ

× 100

Area under the curve. Area under the curve (AUC) was calculated according to the formulas described by Pruessner and colleagues (Pruessner et al., 2003), for both area under the curve with respect to the ground (AUCG) as well as area under the curve with respect to intensity (AUCI) and were applied once for the area defined by the baseline, post-stressor, and post-task samples for both alpha-amylase and cortisol. Risk-taking task Participants completed the Risk-Taking Task prior to drinking and the difference score between this session and the second session (following drinking and the stressor) was used to assess the effects of stress and alcohol on risk-taking behavior. Risk-taking on the task was measured on a trial by trial basis by calculating each trial (N) in which the risky option was chosen and reinforced, followed by another selection from the risky deck in which the response was penalized (N+1). The number of penalized responses following the initial loss was calculated for the five subsequent trials as an index of how willing a participant is to continue to risk losing money. The percent total risk taking score was used for the first session of the task, which had 50 trials. This allowed participants to become familiar with the task, but shortened the total experimental timeline. Other behavioral measures included the total number of choices made from the risky deck, overall response rate and earnings on the task. The number of risky choices was used as the dependent variable in a 2 (group) × 2 (session) repeated-measures ANOVA, examining whether the mean number of risks taken was significantly higher following the stressor. This analysis was also used to determine whether there was a main effect of alcohol or stress on risk-taking, or an interaction between the two. Drug effects questionnaire and manipulation check Subjective ratings examining perceived intoxication effects were analyzed with one-way ANOVAs with the subjective rating as the

The intoxicated group had a mean BAL of 0.092 (SD = 0.04), therefore the experimental manipulation to raise the BAL of participants to 0.08 was successful. Seventy–four individuals completed the Manipulation Check; these data were used in three separate one-way ANOVAs to confirm the effectiveness of the placebo manipulation. Sober, placebo and intoxicated groups differed significantly in their ratings of feeling sober or drunk [F(2, 71) = 42.54, p b 0.001)], in their estimates of bottles of beer consumed [F(2, 71) = 57.69, p b 0.001)], and in their ratings of alcohol content in the drinks [F(2, 71) = 30.84, p b 0.001)]. A Bonferroni post-hoc comparison showed that the beverage groups differed in a stepwise fashion: the placebo group always rated themselves significantly higher than the sober group, but not as high as the intoxicated group (p b 0.05). Alcohol effects Subjective effects Eighty-four individuals completed the POMS and the KUSTA. Subjective mood ratings on the POMS revealed, in multiple ANOVAs, significant differences between the 3 beverage groups on the subscales of Vigour [F(2, 81)=12.82, pb 0.001], Confusion [F(2, 81)=6.84, pb 0.01], Friendliness [F(2, 81)=3.2, p b 0.05], and Elation [F(2, 81)=15.39, pb 0.001]. A post-hoc multiple comparison showed that the sober and placebo groups differed from the intoxicated group, in that the latter reported significantly greater levels of Vigour (M=18.51, sober=12.41; placebo = 11.77), Friendliness (M = 21.2; sober = 18.14; placebo= 17.95) and Elation (M = 14.27; sober = 9.36; placebo = 8.95). Placebo (M=3.34) and intoxicated (M=4.2) groups reported significantly higher levels of confusion than the sober group (M=0.91). These results are depicted in Fig. 2. Subjective mood ratings on the KUSTA revealed, in multiple ANOVAs, a difference between the 3 beverage groups only on the subscale of anxiety [F(2, 81) = 3.13, p = 0.049]. A post-hoc test

25

* POMS Subjective Ratings

Statistical analyses

Percent change =

dependent factor and either stress condition and/or beverage group as a between subject factor.

20

* *

15

Sober Placebo Intoxicated

10

5

*

0 Vigour

Confusion Friendliness

Elation

Fig. 2. Subjective ratings of the 3 experimental groups on the POMS subscales of Vigour, Confusion, Friendliness and Elation. The intoxicated (n = 41) group reported significantly higher levels of Vigour, Friendliness and Elation as compared to the sober (n = 22) and placebo (n = 21) groups. The sober group reported significantly lower levels on the Confusion subscale as compared to the placebo and intoxicated groups. *p b 0.05. POMS = Profile of Mood States.

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

showed that only the placebo group had slightly lower ratings (M = 12.52) than the intoxicated group (M = 14.66; p = 0.03).

Physiological effects Two one-way ANOVAs were conducted in order to assess the acute effect of alcohol on the absolute concentrations of cortisol and alphaamylase, immediately following drink consumption (approximately at 75 min). These analyses revealed no significant differences in baseline cortisol levels between the sober, placebo or alcohol groups [F(2, 80) = 0.04, p = 0.96) or alpha-amylase levels [F(2, 80) = 0.26, p = 0.77) at this timepoint. That is, there were no statistically significant differences between sober, placebo and intoxicated groups on this measure following drink consumption. A timeline depicting cortisol and alpha-amylase levels across the 3 sampling points is depicted in Fig. 3.

Stressor effects A detailed description of the subjective and physiological effects of the TSST is provided elsewhere (Balodis et al., 2010). In brief, the stress manipulation induced robust increases in the subjective reports of anxiety, depression, anger and tension on the POMS and KUSTA questionnaires. Both cortisol and alpha-amylase levels increased significantly following the TSST with a 41% and 71% change from baseline respectively (Balodis et al., 2010). In contrast, cortisol and alpha-amylase decreased 14.8% and 0.3% in the no stress control group during the same period (Balodis et al., 2010).

a) 0.35 0.3

cortisol nmol/l

No Stress - Sober

0.25

No Stress - Placebo

0.2

No Stress - Intoxicated Stress - Sober

0.15

Stress - Placebo

0.1 Stress - Intoxicated

Alcohol–stressor interactions Subjective effects A 3 (beverage group) × 2 (stress condition) × 2 (session) repeatedmeasures ANOVA revealed a significant main effect of stress condition on anxiety ratings on the POMS anxiety subscale [F(1,76) = 9.24, p b 0.01)]. There was no main effect of beverage group [F(2,76) = 0.54, p N 0.05)], but a trend for a stress condition by beverage group interaction [F(2,76) = 2.74, p = 0.07)]. There was a significant withinsubjects effect of session [F(1,76) = 34.71, p b 0.01)] as well as a session by stress condition interaction [F(1,76) = 40.44, p b 0.01)]. There was, however, no session by beverage group interaction [F(2,76) = 2.15, p N 0.05)], but there was a 3-way session by stress by beverage group interaction [F(2,76) = 3.75, p b 0.05)]. Pairwise comparisons revealed that this interaction could be accounted for by beverage group differences in the stress condition post-TSST. Specifically, the sober group reported higher anxiety on the POMS than the intoxicated group (p b 0.05), however, the placebo group did not differ significantly in their anxiety ratings when compared to the sober or the intoxicated groups (p N 0.05; Fig. 4a). In order to verify that these effects were not related to gender differences, the same repeated-measures ANOVA was conducted in the female sample only. This 3 (beverage group) × 2 (stress condition) × 2 (session) repeated-measures ANOVA produced equivalent results to the analysis which included males: there was a significant main effect of stress condition on anxiety ratings on the POMS anxiety subscale [F(1,50) =7.80, p b 0.01)]. There was no main effect of beverage group [F(2,50) = 0.99, p N 0.05)], and no stress condition by beverage group interaction [F(1,50) =1.86, p = 0.17)]. There was a significant within-subjects effect of session [F(1,50) = 30.17, p b 0.01)] as well as a session by stress condition interaction [F(1,50) = 27.64, p b 0.01)]. There was, however, no session by beverage group interaction [F(2,50) =2.35, p N 0.05)], but the 3-way session by stress by beverage group interaction was significant [F(2,50) = 3.47, p b 0.05)]. Pairwise comparisons revealed that this interaction could be accounted for by beverage group differences in the stress condition post-TSST. Specifically, the sober group reported higher anxiety on the POMS than the intoxicated group (p b 0.05), however, the placebo group did not differ significantly in their anxiety ratings, when compared to the sober or the intoxicated groups. A 3 (beverage group) × 2 (stress condition) × 2 (session) repeatedmeasures ANOVA revealed a significant main effect of stress condition [F(1,76) = 9.42, p b 0.01)] as well as a significant beverage group effect

0.05 16 Baseline

C2

14

C3

b)

Alpha-amylase U/ml

250 No Stress - Sober

200

No Stress - Placebo

150

No Stress - Intoxicated

Subjective Rating

0

469

a)

*

b)

c)

*

12 10 Sober Placebo Intoxicated

8 6 4

Stress - Sober

2

Stress - Placebo

0

100 Stress - Intoxicated

50

0

Baseline

A2

A3

Fig. 3. Timeline depicting absolute cortisol (a) and alpha-amylase (b) levels for the 3 experimental groups in the stress and no-stress conditions. There were no significant group differences between experimental groups in absolute levels of cortisol or alphaamylase at baseline. Physiological measures are depicted in absolute values in order to demonstrate changes over time.

POMS Anxiety KUSTA Anxiety KUSTA Tension Subscale Subscale Subscale Fig. 4. Subjective ratings of anxiety and tension following the TSST in the 3 experimental groups. a) on the POMS Anxiety subscale, the sober group (n = 14) reported significantly higher ratings of anxiety than the intoxicated group (n = 24); the placebo group (n = 12) did not differ significantly in their anxiety ratings from either the sober or the intoxicated groups. b) there were no statistically significant differences in reports of anxiety on the KUSTA Anxiety subscale. c) on the KUSTA Tension subscale, the sober group reported significantly higher levels of tension than either the placebo or intoxicated groups. POMS = Profile of Mood States; KUSTA = Kurz-Skala Stimmung/ Activierung.

470

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

on KUSTA anxiety subscale ratings [F(2,76) = 3.60, p b 0.05)]. Pairwise comparisons showed that the beverage group effect was driven by the sober group, who reported higher anxiety ratings, when compared to the alcohol group (p b 0.05), but not the placebo group (p N 0.05; Fig. 4b). There was, however, no significant interaction between stress condition and beverage group [F(2,76) = 0.23, p N 0.05)]. There was a significant main within-subjects effect of session [F(1,76) = 9.70, p b 0.01)] as well as a significant session by stress condition interaction [F(1,76) = 21.03, p b 0.01)]. However, there was no significant 2-way session by beverage group interaction [F(2,76) = 0.20, p N 0.05)], nor any 3-way session by stress condition by beverage group interactions [F(2,76) = 0.99, p N 0.05)]. In order to verify that these effects were again not driven by gender differences, the same repeated-measures ANOVA was conducted in the female sample only. This 3 (beverage group) × 2 (stress condition)× 2 (session) repeated-measures ANOVA produced equivalent results to the analysis which included males: there was a significant main effect of stress condition [F(1,49) = 12.12, p b 0.01)] as well as a significant beverage group effect on KUSTA anxiety subscale ratings [F(2,49) = 4.07, p b 0.05)]. Pairwise comparisons showed that the beverage group effect was driven by the sober group, who reported higher anxiety ratings, when compared to the alcohol group (p b 0.05), but not the placebo group (p N 0.05). There was, however, no significant interaction between stress condition and beverage group [F(2,49) = 0.40, p N 0.05)]. There was a significant main within-subjects effect of session [F(1,49)= 4.74, p b 0.05)] as well as a significant session by stress condition interaction [F(1,49) = 12.77, p b 0.01)]. However, there was no significant 2-way session by beverage group interaction [F(2,49) = 0.04, p N 0.05)], nor any 3-way session by stress condition by beverage group interactions [F(2,49) = 0.74, p N 0.05)].

Physiological effects In order to minimize the effect of individual differences and to focus on physiological changes in response to the stressor, cortisol and alpha-amylase were analyzed using the percent change from baseline and area under the curve (Balodis et al., 2010). A 2 (stress condition) × 3 (beverage group) ANOVA examined the percent change in cortisol from pre- to post-stressor. This analysis yielded a significant main effect of stress condition [F(1, 80) = 12.26, p b 0.01] and a main effect of beverage group [F(2, 80) = 4.94, p b 0.01], but no significant interaction between the two [F(2, 80) = 2.49, p N 0.05]. Beverage group differences were driven by the sober group exhibiting a 55.33% increase in cortisol levels, while the placebo and alcohol groups displayed an overall decrease of −2.76 and −3.53% respectively. Furthermore, as shown in Fig. 5a, in the no stress condition, the sober group differed significantly from the placebo and alcohol groups in the percentage change in cortisol, who did not significantly differ from each other. In contrast, a 2-way ANOVA with stress and beverage group as fixed factors examining AUCG for cortisol levels immediately following the stressors did not produce a significant main effect of stress condition [F(1, 80) = 1.49, p N 0.05], beverage group [F(2, 80) = 0.72, p N 0.05] or an interaction between the two [F(2, 80) = 1.89, p N 0.05]. To verify that this cortisol difference was not driven by gender effects, another 2 (stress condition) × 3 (beverage group) ANOVA examined the percent change in cortisol from pre- to post-stressor in female participants only. This analysis also yielded a significant main effect of stress condition [F(1, 51) = 5.09, p b 0.05] and a main effect of beverage group [F(2, 51) = 4.10, p b 0.05], but no significant interaction between the two [F(2, 51) = 1.05, p N 0.05]. Beverage group differences were driven by the sober females exhibiting a 58.36% increase in cortisol levels, while the placebo and alcohol groups displayed an overall decrease of −6.76 and −11.80% respectively. Pairwise comparisons in the stress group showed that the sober group had significantly higher cortisol levels than both the alcohol and the

Fig. 5. Cortisol and alpha-amylase reactivity following the TSST between stress groups and in the 3 experimental groups. a) In the stress condition, the sober group (n = 14) demonstrated a significant increase in cortisol levels following the stressor, while the placebo (n = 12) and the intoxicated group (n = 24) did not. b) During the stress condition, there were no significant differences in alpha-amylase levels between the 3 experimental groups. TSST = Trier Social Stress Test.

placebo groups (p b 0.05), again the latter two groups did not significantly differ from one another (p N 0.05). In contrast to cortisol, there were no group differences in alphaamylase responses to stress (see Fig. 5b). This was confirmed by a 2 (stress condition)×3 (beverage group) ANOVA which yielded a significant main effect of stress condition [F(1, 76)=11.84, pb 0.01] but not beverage group [F(2, 76)=0.32, pN 0.01] and no interaction between the two [F(2, 76)=1.26, pN 0.05]. Similarly, analysis of the AUCG for alpha-amylase revealed a significant main effect of stress condition [F(1, 76)=6.57, pb 0.05], but not beverage group [F(2, 76)=0.22, pN 0.05] and no stress condition and beverage group interaction [F(2, 80)=0.71, pN 0.05]. Risk-taking A 3 (beverage group) × 2 (stress) × 2 (session) repeated-measures ANOVA revealed no main effect of session on risk-taking [F(1,77) = 0.21, p = 0.65] in that mean risk-taking scores were the same on both sessions across groups (see Fig. 6). There was no significant main effect of stress on risk-taking [F(1, 77) = 0.59, p = 0.45], and no significant stress x session interaction [F(1, 77) = 0.42, p = 0.52]. Also, there was no main effect of beverage group [F(2, 77) = 0.15, p = 0.87], there was, however, a significant interaction between group and session [F(2, 77) = 3.66, p = 0.03]; a Bonferroni-adjusted post-hoc comparison showed that, collapsed across stress conditions, the sober group demonstrated significantly lower risk-taking scores on the second session (p = 0.03). The 3-way interaction between session, stress and group was not significant [F(2, 77) = 0.27, p = 0.77]. Cortisol, alpha-amylase and risk taking Correlation coefficients were calculated for task performance and cortisol levels, indicating no significant relationship between percent change in cortisol levels and risk-taking score (r = 0.08, p N 0.05).

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

Fig. 6. The effects of stress and intoxication on risk-taking in a computer task. Risktaking was measured by calculating each trial (N) in which the risky option was chosen and reinforced, followed by another selection from the risky deck in which the response was penalized (N + 1). The number of penalized responses following the initial loss was calculated for the five subsequent trials. There was no difference in risk-taking scores between sessions or between groups. Note: At session 1, all groups were sober and had not yet experienced stress.

Pearson correlation coefficients were calculated for task performance and alpha-amylase levels, indicating no significant relationship between percent change in alpha-amylase levels and risk-taking score or the amount of money won on the task (p N 0.05). Discussion In this study, alcohol intoxication was associated with an increase in Vigour, Friendliness and Elation, as measured by the POMS. This is consistent with other studies showing a general increase in self-report measures of stimulation on the ascending limb of the Blood Alcohol Curve (Soederpalm et al., 2002). The stress manipulation was also effective in that self-reports of anxiety, tension and depression, as well as cortisol and alpha-amylase levels, increased significantly following the psychosocial stressor. Both alcohol and the expectation of receiving alcohol decreased subjective and physiological responses to stress, although not all measures were affected equally. Finally, beverage group differences were revealed on the task, whereby sober individuals, regardless of stress condition, demonstrated lower risktaking on the computer task. As predicted, alcohol reduced both subjective and physiological responses to stress; surprisingly, the expectation of consuming alcohol reproduced many of these effects. Specifically, compared to sober participants, placebo and intoxicated groups reported lower anxiety levels, and cortisol increases following the stressor were dampened in both of these groups. As in our previous studies (Balodis et al., 2006; Ortner et al., 2003), the protocol for administering alcohol controlled for the expectancy effect of the drug: although individuals in the placebo group did not rate themselves as intoxicated as the alcohol group, they clearly believed that they had consumed alcohol. The placebo manipulation in the current study consists of an elaborate procedure in order to convince the participants that they are consuming alcohol. This includes initially having the participant agree to take a cab home, weighing the participant, mixing the drinks with vodka bottles (filled with flattened tonic water) in front of the participant, rimming the glass with alcohol, asking the participant to blow into a breathalyzer multiple times and having the experimenter write down false BALs on a sheet of paper. The blunted cortisol response to stress in the placebo group is a striking indication that the procedure is effective. This effect has not been reported previously but most experimental studies do not include two control conditions (i.e., sober and placebo) (e.g. Soederpalm, 2002; Zimmermann et al., 2006). Interestingly, alcohol did not reduce alpha-amylase responses to stress in any group supporting our previous contention that the two biochemical measures yield independent measures of physiological

471

responses to stress that might reflect activity along different mechanistic pathways (Balodis et al., 2010). The alcohol expectancy effect in our study could reflect a conditioned response to cues predicting intoxication because all of the participants regularly consumed alcohol. These appear limited to anxiolytic, not positive reinforcing, effects of alcohol in that the groups differed only on subjective ratings of anxiety and tension, but not on happiness or elation following the psychosocial stressor. Regardless of the interpretation, the potency of alcohol expectancy emphasizes that both sober (i.e., no expectation of receiving alcohol) and placebo groups are necessary controls in studies examining the effects of alcohol on subjective and physiological responses to alcohol. Unlike our measures of stress reactivity, alcohol intoxication did not alter performance on the risk-taking task, contradicting a previous report using the same task (Lane et al., 2004). The main methodological differences between the two studies are that Lane and colleagues used a within-subjects design across several alcohol doses, and tested a non-college sample on different days. In addition the two studies used different testing times (14:00–19:00 in our study versus 9:15 in the Lane et al. study), social environments (2–3 participants drinking together in our study versus 1 participant drinking alone in the Lane et al. study), and remuneration. On the behavioral task, the sober group demonstrated a significant decrease in risk-taking from the first to second session. It is possible that the performance in this group was influenced by the addition of the monetary reinforcer, as participants were told that they could keep winnings on the second session. Nevertheless it appears that repeated testing alters task performance, an effect that is masked by alcohol and the expectation of alcohol intoxication. These differences in risktaking also support the idea that alcohol and placebo have similar behavioral responses that differ from the sober group. Another difference between the two studies is that we used a response option variable ratio (VR) of 4, whereby participants were required to make an average of 4 responses to complete each trial. We chose this option because it provided an average inter-trial interval of 4 seconds and maintained the interest of participants. Lane's study used a VR-25 schedule which perhaps provided participants with more time to reflect on their wins or losses as well as their subsequent choice. Alcohol may have altered performance under these conditions of increased cognitive load, revealing an alcohol-induced increase in risk-taking in the Lane et al. study.

Limitations and future directions The current study is limited by several factors, including a small sample size in some groups as well as unbalanced gender samples. Sex differences are noted in stress, alcohol as well as risk-taking measures, as well as in the interaction of many of these variables (Fox et al., 2009; Franco et al., 2010; Maestripieri et al., 2010; Urban et al., 2010). Consistent with other studies, we observed higher cortisol responses to the stressor in males, but the small number of male participants in our sample precluded an analysis of sex differences (Childs et al., 2010; Childs et al., 2010) Additionally, we tried to test females during the same stage of their menstrual cycle, but did not control for oral contraceptive use which may alter cortisol levels (Kirschbaum et al., 1999). Nevertheless, since all measures were within individual proportionate changes, effects on baseline cortisol were not directly involved in analyses. In addition, these stress–response dampening effect of placebo in both physiological and subjective measures were evident with and without the inclusion of males in the sample. In the future, it would be interesting to examine subjective measures of affect at more timepoints, including prior to alcohol administration, as well as following the risk-taking task. This information will help to determine whether anxiety and other mood variables are significantly altered by different types of winnings on the task. While we have discussed how intra-individual differences in stress levels may affect risk taking, future

472

I.M. Balodis et al. / Hormones and Behavior 59 (2011) 465–472

studies could also examine how specific personality traits, such as impulsivity, may mediate stress responses and risk-taking behaviors. Conclusions In this experiment, we provide a characterization of alcohol's effect on multiple measures of stress. Using the TSST, a paradigm that reliably increases stress, we demonstrated that a high dose of alcohol produces a stress-dampening effect on some, but not all, subjective and physiological measures of stress. We also report a powerful placebo effect that emulates some of the anxiolytic and hormonal effects of alcohol. The stress–response-dampening effects evidenced on cortisol reactivity in the current study have significant implications as cortisol acts as a feedback measure to cortico-limbic processing (Lovallo, 2006). Therefore any alterations in cortisol activity over time could have profound influences on motivation and subsequent decision-making. As this is the same neural circuit that is engaged by stress, drugs of abuse, and general reward-processing, our characterization of intoxicated responding in healthy individuals is a first step in understanding the cognitive and physiological mechanisms mediating adaptations to stress. These findings have implications for putting into context stress–reward interactions in psychiatric populations, including individuals with pathological gambling or other addictions, who demonstrate alterations in corticosteroid hormone reactivity as well as difficulties in behavioral adaptation (Contoreggi et al., 2003; Meyer et al., 2004). Acknowledgments This research was supported by the Natural Sciences and Engineering Research Council of Canada and the Ontario Problem Gambling Research Centre. These funding agencies had no further role in the study design, data collection, analysis or interpretation of data, writing of the report, or the decision to submit the paper for publication. The authors would like to thank Scott Lane for the use of his risk-taking task. Many thanks to Lea Bond for her technical help with the assays. The authors would like to extend a special thanks to Sheila Dillon-Leitch, Hayley Walkden and Vivian Lee for their immense contribution of time and energy in conducting the TSST and collecting the experimental data. References Balodis, I.M., et al., 2007. Intact preference conditioning in acute intoxication despite deficient declarative knowledge and working memory. Alcohol. Clin. Exp. Res. 31, 1800–1810. Balodis, I.M., et al., 2006. Instructional cues modify performance on the Iowa Gambling Task. Brain Cogn. 60, 109–117. Balodis, I.M., et al., 2010. The other side of the curve: examining the relationship between pre-stressor physiological responses and stress reactivity. Psychoneuroendocrinology 35, 1363–1373. Balodis, I.M., et al., 2009. Binge drinking in undergraduates: relationships with sex, drinking behaviors, impulsivity, and the perceived effects of alcohol. Behav. Pharmacol. 20, 518–526. Chatterton Jr., R.T., et al., 1996. Salivary alpha-amylase as a measure of endogenous adrenergic activity. Clin. Physiol. 16, 433–448. Childs, E., et al., 2010. Cardiovascular, hormonal, and emotional responses to the TSST in relation to sex and menstrual cycle phase. Psychophysiology 47, 550–559.

Childs, E., et al., 2010. Cardiovascular, hormonal, and emotional responses to the TSST in relation to sex and menstrual cycle phase. Psychophysiology 47, 550–559. Contoreggi, C., et al., 2003. Stress hormone responses to corticotropin-releasing hormone in substance abusers without severe comorbid psychiatric disease. Biol. Psychiatry 54, 873–878. Davis, J.R., Jeffcoate, W.J., 1983. Lack of effect of ethanol on plasma cortisol in man. Clin Endocrinol (Oxf). 19, 461–466. Enberg, N., et al., 2001. Saliva flow rate, amylase activity, and protein and electrolyte concentrations in saliva after acute alcohol consumption. Oral Surg. Oral Med. Oral Pathol. Oral Radiol. Endod. 92, 292–298. Fox, H.C., et al., 2009. Sex-specific dissociations in autonomic and HPA responses to stress and cues in alcohol-dependent patients with cocaine abuse. Alcohol Alcohol. 44, 575–585. Franco, C., et al., 2010. Male gamblers have significantly greater salivary cortisol before and after betting on a horse race, than do female gamblers. Physiol. Behav. 99, 225–229. George, S., et al., 2005. The acute effect of alcohol on decision making in social drinkers. Psychopharmacology (Berl). 182, 160–169. Josephs, R.A., Steele, C.M., 1990. The two faces of alcohol myopia: attentional mediation of psychological stress. J. Abnorm. Psychol. 99, 115–126. King, A., et al., 2006. Attenuated cortisol response to alcohol in heavy social drinkers. Int. J. Psychophysiol. 59, 203–209. King, J.A., et al., 1998. Attention-deficit hyperactivity disorder and the stress response. Biol. Psychiatry 44, 72–74. Kirschbaum, C., et al., 1999. Impact of gender, menstrual cycle phase, and oral contraceptives on the activity of the hypothalamus–pituitary–adrenal axis. Psychosom. Med. 61, 154–162. Kirschbaum, C., et al., 1993. The ‘Trier Social Stress Test’—a tool for investigating psychobiological stress responses in a laboratory setting. Neuropsychobiology 28, 76–81. Lane, S.D., et al., 2004. Alcohol effects on human risk taking. Psychopharmacology (Berl). 172, 68–77. Lovallo, W.R., 2006. Cortisol secretion patterns in addiction and addiction risk. Int. J. Psychophysiol. 59, 195–202. Maestripieri, D., et al., 2010. Between- and within-sex variation in hormonal responses to psychological stress in a large sample of college students. Stress 13, 413–424. McNair, D.M., et al. (Ed.), 1981. Manual for the Profile of Mood States. Educational and Industrial Testing Service, San Diego. Meyer, G., et al., 2004. Neuroendocrine response to casino gambling in problem gamblers. Psychoneuroendocrinology 29, 1272–1280. Nater, U.M., et al., 2005. Human salivary alpha-amylase reactivity in a psychosocial stress paradigm. Int. J. Psychophysiol. 55, 333–342. Ortner, C.N., et al., 2003. Alcohol intoxication reduces impulsivity in the delaydiscounting paradigm. Alcohol Alcohol. 38, 151–156. Pruessner, J.C., et al., 2003. Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 28, 916–931. Rohleder, N., et al., 2004. Psychosocial stress-induced activation of salivary alphaamylase: an indicator of sympathetic activity? Ann. NY Acad. Sci. 1032, 258–263. Sayette, M.A., 1993. An appraisal-disruption model of alcohol's effects on stress responses in social drinkers. Psychol. Bull. 114, 459–476. Schuckit, M.A., et al., 1987. Plasma cortisol levels following ethanol in sons of alcoholics and controls. Arch. Gen. Psychiatry 44, 942–945. Soederpalm, A.H., et al., 2002. Effects of stress and alcohol on subjective state in humans. Alcohol. Clin. Exp. Res. 26, 818–826. Sorocco, K.H., et al., 2006. Blunted hypothalamic–pituitary–adrenocortical axis responsivity to stress in persons with a family history of alcoholism. Int. J. Psychophysiol. 59, 210–217. Takahashi, T., 2004. Cortisol levels and time-discounting of monetary gain in humans. NeuroReport 15, 2145–2147. Urban, N.B., et al., 2010. Sex differences in striatal dopamine release in young adults after oral alcohol challenge: a positron emission tomography imaging study with [(1)(1)C]raclopride. Biol. Psychiatry 68, 689–696. van Honk, J., et al., 2003. Low cortisol levels and the balance between punishment sensitivity and reward dependency. NeuroReport 14, 1993–1996. Waltman, C., et al., 1993. The effects of mild ethanol intoxication on the hypothalamic– pituitary–adrenal axis in nonalcoholic men. J. Clin. Endocrinol. Metab. 77, 518–522. Wendt, G., et al., 1985. KUSTA (Kurz-Skala Stimmung/Aktivierung): a daily self-rating scale for depressive patients. Pharmacopsychiatry 18, 118–122. Zimmermann, U.S., Buchmann, A., Steffin, B., Dieterie, C., Uhr, M., 2006. Alcohol administration acutely inhibits ghrelin secretion in an experiment involving psychosocial stress. Addict. Biol. 12, 17–21.