This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Mental Health, is properly cited. The complete bibliographic information, a link to the original publication on http://mental.jmir.org/, as well as this copyright and license information must be included.
Exposure therapy is highly effective for social anxiety disorder. However, there is room for improvement.
This is a first attempt to examine the feasibility of an arousal feedback–based exposure therapy to alleviate social anxiety symptoms in an analogue adult sample.
A randomized, pilot, proof-of-concept trial was conducted to evaluate the acceptability, safety, and preliminary efficacy of our treatment program. Sessions were administered once a week for 4 weeks (1 hour each) to an analogue sample of 50 young adults who reported at least minimal social anxiety symptoms. Participants in both intervention and waitlist control groups completed assessments for social anxiety symptoms at the baseline, week 5, and week 10.
Most participants found the intervention acceptable (82.0%, 95% CI 69.0%-91.0%). Seven (14.9%, 95% CI 7.0%-28.0%) participants reported at least one mild adverse event over the course of study. No moderate or serious adverse events were reported. Participants in the intervention group demonstrated greater improvements on all outcome measures of public speaking anxiety from baseline to week 5 as compared to the waitlist control group (Cohen
Our results indicated that it is worthwhile to proceed to a larger trial for our treatment program. This new medium of administration for exposure therapy may be feasible for treating a subset of social anxiety symptoms. Additional studies are warranted to explore its therapeutic mechanisms.
ClinicalTrials.gov NCT02493010; https://clinicaltrials.gov/ct2/show/NCT02493010
Social anxiety disorder (SAD) is one of the most prevalent anxiety disorders [
Exposure therapy is the mainstay of SAD treatment. Individuals afflicted by SAD often do not seek treatment unless it is persistent, pervasive, or accompanied by other more acute psychiatric conditions [
Prevailing technology-based exposure therapies for SAD including VRETs are therapist-mediated and thus require manual adjustment of exposure parameters to suit individual needs. Anderson et al reported that therapists modified virtual audience reactions and environments physically depending on patients’ hierarchy of fears [
Recent advances in the conceptualization of exposure therapy posit inhibitory learning as a more parsimonious theory to explain treatment effects and failure and advocate inhibitory learning techniques to optimize treatment effects among patients [
Anxiety disorders have also been treated using biofeedback-based interventions. By convention, exposure therapies and biofeedback-based interventions progress along distinct lines of research: The former facilitates desensitization to a prespecified hierarchy of anxiety-provoking situations (eg, public speaking or eating in public) by
Our study aimed to integrate biofeedback with portable hardware to enhance current technology-assisted exposure interventions for SAD. We argue that biofeedback technologies automate real-time modifications in exposure therapy as well as provide a means to sustain participant distress during exposure in accordance with inhibitory learning theory. Mechanistically different from typical biofeedback programs for anxiety, we designed and utilized a first-of-a-kind arousal feedback–based system that runs on inputs from individuals’ EEG and heart rate detection. An underlying arousal feedback–based machine learning algorithm combined EEG band powers, heart rate, and heart rate variability to influence difficulty parameters of the virtual exposure environment in real time. These dynamic changes facilitated retainment of discrepancy between participants’ actual and anticipated consequences, thereby sustaining inhibitory learning throughout the exposure session. Our combination of technologies forms an expedient means to deliver exposure therapy. Participants can be repeatedly exposed to newly learned safety associations and subjected to prolonged distress in a controlled virtual environment. Biofeedback technology helped calibrate and address individual differences in baseline anxiety and distress as well as automatize the exposure function through arousal feedback.
This was the first study to examine the viability of using biofeedback and portable technologies in delivering, personalizing, and optimizing exposure therapy in a laboratory setting. In particular, the purpose of our pilot trial was to investigate the feasibility of an arousal feedback–based exposure therapy to alleviate social anxiety symptoms, with particular emphasis on public speaking anxiety. We employed a randomized, waitlist-controlled design to evaluate the acceptability, safety, and preliminary efficacy of this treatment program. The study was conducted over a 4-week period with a 5-week follow-up of an analogue sample of 50 young adults who had public speaking anxiety complaints and reported at least minimal social anxiety symptoms.
In this paper, we examined the feasibility of our exposure therapy with real-time arousal detection and feedback in reducing social anxiety symptoms. Specifically, for our primary objectives, we hypothesized that the acceptability rate of our program would be high, that adverse event rate would be low, and that participants in the intervention group would show greater improvement in social anxiety symptoms between baseline and week 5 assessments than the waitlist-control group. For our secondary objectives, we hypothesized that compared to the waitlist-control group participants, the intervention group participants would demonstrate greater improvements in public speaking anxiety, fear of negative evaluation, and self-statements made during public speaking from baseline to week 5.
This was a randomized, waitlist-controlled, single-center open-label study. Participants completed assessments at baseline, week 5, and week 10. The intervention group attended intervention sessions from weeks 1 to 4, and the waitlist control group attended sessions between weeks 6 and 9. This study was approved by the Institutional Review Board of the National University of Singapore (reference code: B-14-098).
The study was conducted from May 2016 to May 2017 at Duke-NUS Medical School, Singapore. Participants were recruited on a voluntary basis through various modes of advertising including clinician referrals, posters, newspapers, social media, institutional email notices (ie, Duke-NUS Medical School, National University of Singapore), and word of mouth. Interested participants were scheduled for a written informed consent and brief screening session. Each participant was told to complete an intervention schedule once a week over a 4-week period. Participants were also instructed to complete assessments at the baseline, week 5, and week 10. Reimbursement for time and transport was provided on a prorated basis upon completion or termination of the study. All participants were told that they may or may not benefit from participating in the intervention. Data collected were anonymized, and improvements reported were not associated with any personal benefit.
Eligibility was determined after consent was obtained. The inclusion criteria were age between 21 and 35 years, literacy in English and computer skills, absence of current or previous history of neuropsychiatric disorders, and willingness to be video recorded. Eligible participants had scores of ≥31 on the Liebowitz Social Anxiety Scale (LSAS) [
The intervention was implemented using a locally developed, noninvasive portable headband (
Locally developed, noninvasive headband.
All participants completed the calibration process followed by the intervention process. During the calibration process, our system constructed a personalized arousal profile for each individual, based on unique physiological signatures (ie, EEG and photoplethysmogram patterns detected by our hardware) occurring during alternating induced periods of high and low arousal. The personalized arousal profile consisted of an adapted threshold, which was required to manipulate each individual’s exposure environment.
Our intervention consisted of 4 weekly sessions, each lasting about 60 minutes. Each session was conducted between 8.30 am and 8.30 pm on weekdays. During the intervention session, participants underwent three types of tasks: an interactive psychoeducation on screen, brief arousal control games, and arousal feedback–based speech tasks to a virtual audience. All materials were presented on the monitor of the same 14-inch laptop used to process physiological signals by the biofeedback system. Participants completed each session independently in a quiet room and were instructed to follow all instructions on the screen. The psychoeducation component delivered information about social anxiety and coping strategies and highlighted typical maladaptive thoughts and behaviors associated with particular anxiety-provoking social situations. To support the in-session learning material, participants were tested on key takeaways and tasked to identify their specific social anxiety-related concerns through pen-and-paper homework. Homework was to be completed every week before the next session. No formal assessment of homework performance was conducted.
A brief arousal control game (
In the arousal feedback–based speech task, participants were tasked to deliver six 2-minute speeches to a virtual audience (
Brief arousal control game.
Virtual audience in arousal feedback–based speech task.
The LSAS is a validated and widely used 24-item questionnaire that assesses fear or anxiety and avoidance on a variety of social interaction and performance situations [
The PSAS is a recently published 17-item self-rated questionnaire that measures cognitive, behavioral, and physiological manifestations of public speaking anxiety [
The Fear of Negative Evaluation - Brief questionnaire (FNE-B) is a validated 12-item scale that assesses one’s fear of being judged negatively by others [
The Self-Statements made during Public Speaking (SSPS) scale measures positive and negative thoughts about oneself during public speaking situations [
Acceptability was defined as a rating of 5, 6, or 7 on “How would you rate the quality of the training system?” This was measured as part of a study-specific satisfaction and immersion questionnaire (SIQ; internal consistency: α=0.85) that examined participants’ attitudes toward the arousal feedback–based intervention on a 7-point Likert-type scale (1=poor, 7=excellent) postintervention. The form included a final open-ended question to capture comments or suggestions. All participants completed the SIQ postintervention.
A total sample size of 41 participants was required to yield a precision (width of 95% CI) of approximately 12% in the proportion of participants who provide positive feedback on acceptability, assuming the true proportion is approximately 80%. Assuming an attrition rate of approximately 20%, a total sample size of 50 subjects was required. We simultaneously evaluated the preliminary efficacy of the training system to determine whether a larger-scale trial is warranted, by using Simon’s randomized selection design [
Randomization was performed in a 1:1 allocation ratio, using blocks of 10 with permuted subblocks of sizes 4 and 6, via a password-protected Web-based program. Block size was determined by the study statistician and not made known to clinical investigators or site personnel until after study closure.
Acceptability analysis was based on all enrolled subjects. Safety analyses were conducted for treated participants who received at least one intervention session. Efficacy analyses were intention-to-treat (ITT) and involved all randomized participants, with per protocol (PP) analysis conducted as sensitivity analysis. Acceptability analyses were rated by pooling responses on the SIQ question “How would you rate the quality of the training system?” from both intervention and waitlist control groups after receiving treatment. Missing acceptability assessment was imputed as “not acceptable.” Further complete-case sensitivity analyses of primary and secondary efficacy endpoints analyzed all participants with baseline assessments, accounting for missing data using a mixed-effects model with random subject intercepts, adjusted for group, time, and group-time interaction and with restricted maximum likelihood estimation.
Data for participants that found the training system acceptable and for whom the training system was safe were presented as Wilson score CIs. Preliminary efficacy evaluation was conducted using Cohen
Supplementary analyses compared median change and adjusted mean change (
A total of 72 participants were screened, of whom 22 were excluded. Fifty were recruited and randomized into the intervention group (n=25) or waitlist control group (n=25). The majority of participants were female (n=37, 74%) and Chinese (n=42, 84%), with a mean age of 25.6 years. Baseline characteristics were similar between the intervention and waitlist control groups of the ITT population (
A majority (n=44, 88%) of the participants received all four intervention sessions; in addition, 45 (90%) completed week 5 assessments and 44 (88%) completed week 10 assessments. There were five (10%) withdrawals initiated by participants due to their inability to commit to the study and one (2%) by the investigator due to an ear condition that interfered with hardware administration. Three withdrawals occurred before intervention, and three occurred at weeks 1, 2, and 3. The
Demographic and baseline characteristics.
Characteristic | Intervention (n=25) | Waitlist control (n=25) | Total (N=50) | |
Age (years), mean (SD) | 24.2 (3.23) | 27.0 (4.19) | 25.6 (3.96) | |
Female, n (%) | 17 (68.0) | 20 (80.0) | 37 (74.0) | |
Chinese | 20 (80) | 22 (88) | 42 (84) | |
Other | 5 (20) | 3 (12) | 8 (16) | |
Secondary education | 0 (0) | 1 (4.0) | 1 (2.0) | |
Preuniversity | 0 (0) | 2 (8.0) | 2 (4.0) | |
Currently in/graduated from university | 25 (100) | 22 (88.0) | 47 (94.0) | |
BDI-IIa total score, mean (SD) | 12.0 (7.52) | 10.0 (9.10) | 11.0 (8.32) | |
AUDITb total score, mean (SD) | 1.1 (1.59) | 1.9 (2.22) | 1.5 (1.95) | |
LSASc total score, mean (SD) | 68.8 (20.63) | 69.7 (21.61) | 69.3 (20.91) | |
PSASd total score, mean (SD) | 67.0 (4.53) | 68.4 (5.92) | 67.7 (5.27) | |
FNE-Be total score, mean (SD) | 45.6 (7.08) | 43.6 (6.65) | 44.6 (6.88) | |
SSPSf total score, mean (SD) | 26.5 (5.67) | 24.1 (7.35) | 25.4 (6.57) |
aBDI-II: Beck Depression Inventory (2nd edition).
bAUDIT: Alcohol Use Disorders Identification Test.
cLSAS: Liebowitz Social Anxiety Scale.
dPSAS: Public Speaking Anxiety Scale.
eFNE-B: Fear of Negative Evaluation - Brief questionnaire.
fSSPS: Self-Statements made during Public Speaking scale.
Consolidated Standards of Reporting Trials (CONSORT) diagram. AUDIT: Alcohol Use Disorders Identification Test; LSAS: Liebowitz Social Anxiety Scale; PSAS: Public Speaking Anxiety Scale.
Most participants (82.0%, 95% CI 69.0%-91.0%) found the training system acceptable. The remaining, including 6 missing assessments, were classified as “not acceptable.”
Seven (14.9%, 95% CI 7.0%-28.0%) participants reported at least one adverse event over the course of the study. There were a total of eight mild adverse events including eye strain (n=1), itch on forehead and scalp (n=1), headache (n=3), and dizziness (n=3). No moderate or serious adverse events were reported.
Mean total efficacy scores (SD), mean change scores (SD), and effect sizes of differences in mean changes between weeks 0 and 5. Change in total score = week 0 total score – week 5 total score. A positive change in total score indicates improvement (a reduction in symptoms/scores).
Measures | Group | Cohen |
|||||||
Intervention | Waitlist | ||||||||
Week 0 | Week 5 | Change | Week 0 | Week 5 | Change | ||||
LSASa | 68.8 (20.63) | 67.7 (20.87) | 1.5 (20.54) | 69.7 (21.61) | 72.2 (21.32) | –0.8 (14.55) | 0.13 (–0.47 to 0.72) | ||
PSASb | 67.0 (4.53) | 56.9 (9.83) | 10.2 (8.56) | 68.4 (5.92) | 67.7 (7.49) | 1.3 (3.07) | 1.39 (0.72 to 2.05) | ||
FNE-Bc | 45.6 (7.08) | 43.0 (9.20) | 3.0 (8.10) | 43.6 (6.65) | 44.5 (7.51) | –0.9 (4.38) | 0.61 (0.00 to 1.22) | ||
SSPSd | 26.5 (5.67) | 21.2 (8.24) | 6.2 (8.42) | 24.1 (7.35) | 24.1 (8.61) | –0.0 (7.46) | 0.79 (0.17 to 1.41) |
aLSAS: Liebowitz Social Anxiety Scale.
bPSAS: Public Speaking Anxiety Scale.
cFNE-B: Fear of Negative Evaluation - Brief questionnaire.
dSSPS: Self-Statements made during Public Speaking scale.
Unadjusted and adjusted results for efficacy measures were qualitatively the same as results for Cohen
The waitlist control group received the intervention from weeks 6 to 9, with mean changes in efficacy measures between weeks 5 and 10 ranging from 2.2 to 6.5. Corresponding mean changes in efficacy measures between weeks 5 and 10 of the intervention group ranged from 0.9 to 10.0. Pooled pre- and postintervention changes of both arms showed improvement (changes in mean point estimates of 2.6-8.2) on all efficacy outcomes (
Pooled pre- and postintervention efficacy scores. The pooled pre- and post- intervention change score is calculated as the sum of change scores of both arms, where the intervention arm change score is scores of week 0 – week 5, and waitlist control arm change score is score of week 5 – week 10; a positive change indicates improvement (a reduction in symptoms).
Measures | Mean change (SD) | ||
LSASb | 4.0 (21.66) | .23 | |
PSASc | 8.2 (7.50) | <.001 | |
FNE-Bd | 2.6 (6.59) | .01 | |
SSPSe | 5.5 (7.80) | <.001 |
a
bLSAS: Liebowitz Social Anxiety Scale.
cPSAS: Public Speaking Anxiety Scale.
dFNE-B: Fear of Negative Evaluation - Brief questionnaire.
eSSPS: Self-Statements made during Public Speaking scale.
The study results indicated that the arousal feedback–based exposure therapy was acceptable and safe. Improvements in the PSAS, FNE-B, and SSPS scores, which were sustained over a follow-up 5-week period, suggested that our intervention might be efficacious in alleviating adult public speaking anxiety. Our findings did not provide strong support for the efficacy of our intervention in reducing overall social anxiety symptoms on the LSAS. Caution is also needed when interpreting the difference in change scores, as the randomized selection design was only meant to identify intervention that is worthy of further research instead of providing confirmation of efficacy. Overall, our preliminary findings indicated that it is worthwhile to proceed with a larger trial.
The intervention was safe and acceptable. Majority of the few complaints concerned prolonged use of hardware rather than treatment material and software. Young adults today are mobile- and technologically savvy but not yet accustomed to biofeedback technologies. Thus, discomfort with system usage was not unforeseeable. Regarding the treatment material, study participants indicated that “simulations did not feel real” and “having real life audience would be helpful” in the feedback comments. Although the levels of immersive exposure experience may not reach those elicited by
Our study recruited an analogue adult sample and did not include formal clinical diagnoses. Nonetheless, participants enrolled in this study were not unlikely to have SAD, as indicated by baseline scores of ≥31 on the LSAS [
Another more probable explanation for the negative result found on the LSAS was that our intervention helped specifically in addressing public speaking anxiety, which constitutes a subset of SAD symptoms. Effect sizes of improvements on secondary outcome measures, ie, the PSAS, FNE-B, and SSPS scale, ranged from moderate to large (ie, Cohen
Our findings further concurred with extant literature indicating that SAD should be differentiated and treated according to a more severe generalized subtype or a less severe nongeneralized subtype encompassing public speaking anxiety [
Our exposure therapy integrated a key feature of inhibitory learning (ie, distress tolerance) to enhance treatment effects. However, we did not maximize violations of participants’ fear-based expectancies for harm or provide multiple contexts to facilitate the decontextualization of newly learned inhibitory associations [
Set against traditional habituation-based exposure therapies, exposure treatments based on the inhibitory learning model do not necessitate fear reduction during exposure to produce posttreatment fear extinction. Although habituation models suggest that fearful associations (eg, “people hate to hear me speak”) must be eliminated altogether for treatment efficacy, inhibitory learning models postulate that successful exposure occurs even when fearful associations are not eliminated. Although described initially as different theories of exposure therapy [
Some limitations restricted the generalizability of our study findings to patients with SAD, including the recruitment of an analogue subclinical adult sample and a lack of objective outcome measures (eg, measuring performance and arousal during speech to a real audience). This study employed self-rated measurement tools that could be confounded by participant bias or motivation to alleviate social anxiety. However, this was unlikely, given the differential outcomes of the overall social anxiety vis-à-vis specific public speaking anxiety measures. Nonetheless, a replication study investigating the effects of arousal feedback–based exposure therapy using a clinically representative sample and objective assessment tools should be conducted. Future research should also consider conducting active-control studies to tease out the differential effects between arousal feedback–based exposure therapy and therapist-mediated exposure therapy.
In conclusion, it is worthwhile to proceed to a larger trial. This pilot proof-of-concept study is a first attempt to establish the acceptability, safety, and potential efficacy of an arousal feedback–based exposure therapy for an analogue adult sample in order to reduce a subset of social anxiety symptoms. Our findings contribute to a growing body of literature on incorporating technology into mental health care services to improve treatment accessibility. Technology-assisted exposure therapies were previously found to be more cost-effective and amenable to therapists or clients than CBT for SAD [
Statistical analysis.
CONSORT‐EHEALTH checklist (V 1.6.1).
Alcohol Use Disorders Identification Test
Beck Depression Inventory (2nd ed)
cognitive behavioral therapy
electroencephalogram
Fear of Negative Evaluation - Brief questionnaire
internet-delivered cognitive behavioral therapy
intention to treat
Liebowitz Social Anxiety Scale
per protocol
Public Speaking Anxiety Scale
social anxiety disorder
satisfaction and immersion questionnaire
Self-Statements made during Public Speaking
virtual reality exposure therapy
The authors would like to thank all participants for their commitment and participation in this study. The authors would also like to acknowledge Min Quan Heo for assisting with implementation of the study. This work was supported by the National Medical Research Council [NMRC/CTGIITE/0002/2015] that was granted to TSL.
XL, TSL, YC, LL, HZ, ZC, CW, RK, and CG received grants from the National Medical Research Council Singapore during the study. HZ, ZC, CW, and CG have a patent Closed Loop Virtual Reality Exposure Therapy for Social Anxiety Disorder Based on Arousal Detection pending. RK reports intellectual property in related areas that have been signed to Neeuro Private Limited. RK holds ownership rights in Cennerv Pharma and is the Chief Executive Officer of Rush Health System and Chairman of National Medical Research Council, Ministry of Health Singapore. RK is also on the CHS Board of Directors and Singhealth Board of Directors.