A randomized controlled trial of attention modification for social anxiety disorder

Journal of Anxiety Disorders 33 (2015) 35–44

Contents lists available at ScienceDirect

Journal of Anxiety Disorders

A randomized controlled trial of attention modification for social anxiety disorder R. Nicholas Carleton a,∗ , Michelle J.N. Teale Sapach a , Chris Oriet a , Sophie Duranceau a , Lisa M. Lix b , Michel A. Thibodeau a , Samantha C. Horswill a , Jordan R. Ubbens a , Gordon J.G. Asmundson a a b

Department of Psychology University of Regina, Regina, SK, Canada Department of Community Health Sciences, University of Manitoba, Winnipeg, MB, Canada

a r t i c l e

i n f o

Article history: Received 15 January 2015 Received in revised form 21 March 2015 Accepted 22 March 2015 Available online 20 April 2015 Keywords: Attentional bias Attention modification protocol Social anxiety disorder Randomized controlled trial Longitudinal study

a b s t r a c t Social Anxiety Disorder (SAD) models implicate social threat cue vigilance (i.e., attentional biases) in symptom development and maintenance. A modified dot-probe protocol has been shown to reduce SAD symptoms, in some but not all studies, presumably by modifying an attentional bias. The current randomized controlled trial was designed to replicate and extend such research. Participants included treatment-seeking adults (n = 108; 58% women) who met diagnostic criteria for SAD. Participants were randomly assigned to a standard (i.e., control) or modified (i.e., active) dot-probe protocol condition and to participate in-lab or at home. The protocol involved twice-weekly 15-min sessions, for 4 weeks, with questionnaires completed at baseline, post-treatment, 4-month follow-up, and 8-month follow-up. Symptom reports were assessed with repeated measures mixed hierarchical modeling. There was a main effect of time from baseline to post-treatment wherein social anxiety symptoms declined significantly (p < .05) but depression and trait anxiety did not (p > .05). There were no significant interactions based on condition or participation location (ps > .05). Reductions were maintained at 8-month follow-up. Symptom reductions were not correlated with threat biases as indexed by the dot-probe task. The modified and standard protocol both produced significant sustained symptom reductions, whether administered in-lab or at home. There were no robust differences based on protocol type. As such, the mechanisms for benefits associated with modified dot-probe protocols warrant additional research. © 2015 Elsevier Ltd. All rights reserved.

1. Introduction Social Anxiety Disorder (SAD; American Psychiatric Association, 2013) has significant lifetime (12.1%) and 12-month (7.1%) prevalence rates (Ruscio et al., 2008), and comparable rates for men and women (American Psychiatric Association, 2013). The disorder is marked by nervousness or discomfort in social situations (Antony & Swinson, 2000), resulting from fears of being evaluated, embarrassed (Weeks, Carleton, Asmundson, McCabe, & Antony, 2010; Weeks, Jakatdar, & Heimberg, 2010), or making a bad impression (Antony & Swinson, 2000). SAD typically lasts for 12 or more years (Grant et al., 2005), has low remission rates (Massion et al., 2002), and high depression comorbidity (Stein & Stein, 2008). The associated social isolation and high rates of comorbid depression may explain some of the increased suicide risk for patients with SAD

∗ Corresponding author. Tel.: +13063372473. http://dx.doi.org/10.1016/j.janxdis.2015.03.011 0887-6185/© 2015 Elsevier Ltd. All rights reserved.

relative to those without (Thibodeau, Welch, Sareen, & Asmundson, 2013), wherein 35% contemplate suicide regularly and 14% attempt suicide (Cougle, Keough, Riccardi, & Sachs-Ericsson, 2009). Models of SAD emphasize the central role of attentional processes and suggest that (1) heightened self-focused attention and (2) vigilance for external social threat cues influence the development and maintenance of SAD (e.g., Heimberg, Brozovich, & Rapee, 2010; Hofmann, 2007). People with SAD more rapidly engage with and spend more time attending to external social threat cues and emotional faces in anxiety-provoking social situations (Asmundson & Stein, 1994; Chen, Ehlers, Clark, & Mansell, 2002; Lee & Telch, 2008), and demonstrate biases in interpretation, attention, and imagery relative to non-anxious individuals (Hirsch & Clark, 2004). Together, these attentional processes may narrow attention, interfere with beneficial processing, and maintain SAD. Participants in a dot-probe protocol observe a screen on which randomly paired stimuli are presented (for ∼500 ms), one stimulus above the other. One stimulus is neutral (e.g., the word table)

36

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

and the other is associated with threat (e.g., the word snake). After the stimuli presentation, a probe appears in a location (top/bottom) corresponding to where one of the stimuli appeared. Participants press a key as quickly as possible to indicate the probe location (top/bottom). People with clinically significant anxiety respond faster when probes appear in the location of threat stimuli related to their anxiety (i.e., congruent) versus the location of neutral stimuli (i.e., incongruent), regardless of location (top/bottom). In contrast, people without clinically significant anxiety (i.e., controls) respond comparably to threat and neutral words, producing no difference in response times across congruent, incongruent, and neutral trials (i.e., two neutral words presented). Modifying attentional biases using an adapted dot-probe protocol as a treatment for social anxiety has received increasing interest (Koster, Fox, & MacLeod, 2009; MacLeod, Rutherford, Campbell, Ebsworthy, & Holker, 2002). The adapted protocol involves two conditions: (1) the Attention Control Condition (ACC), in which neutral and threat stimuli are replaced by probes with equal frequency, and (2) the Attentional Modification Condition (AMC), in which the probe consistently replaces neutral stimuli. In the AMC participants implicitly learn to direct their attention away from the threat stimulus to detect and respond to the probe, which is thought to modify the attentional bias and reduce symptoms. In one of the original studies exploring the adapted dot-probe protocol as a treatment for SAD symptoms (Schmidt, Richey, Buckner, & Timpano, 2009), participants in the AMC (n = 18), but not in the ACC (n = 18), reported a significant reduction in SAD symptoms (i.e., p < .01; d > .50). Since then, several studies have provided additional support for AMC as a method to reduce SAD symptoms (Amir et al., 2009; Amir, Taylor, & Donohue, 2011; Amir, Weber, Beard, Bomyea, & Taylor, 2008; Schmidt et al., 2009). Training attention away from threat appears more effective for reducing social anxiety than training attention toward threat, neutral, or positive stimuli (Heeren, Lievens, & Philippot, 2011; Heeren, Reese, McNally, & Philippot, 2012); however, studies have also found evidence that the AMC and ACC can both reduce symptoms, with no significant differences in those reductions between the conditions (e.g., Enock, Hofmann, & McNally, 2014; Heeren, Mogoase, McNally, Schmitz, & Philibert, 2015; Julian, Beard, Schmidt, Powers, & Smits, 2012). Much of the attention training research with social anxiety has used a static set of pre-selected word stimuli to assess for the attentional biases, but then used pictures of faces as the stimuli for modifying any such biases (Amir, Beard, Taylor, et al., 2009; Heeren et al., 2015). The intent was to avoid experimental conflation of repeated exposure to the stimuli with changes in attentional biases. That said, related research supporting the success of AMC relative to ACC for generalized anxiety disorder symptoms used words for both assessment and training (Amir, Beard, Cobb, & Bomyea, 2009), with those researchers recommending research exploring the use of only words for SAD as well (Amir et al., 2011). Furthermore, tailored stimuli, rather than a pre-selected static set, should be more salient and therein produce stronger effect sizes. Part of the appeal associated with the attention modification protocols involves the potential for wide spread cost-effective dissemination through the Internet. Indeed, such administration could be particularly suited to persons with social anxiety. At least four separate randomized controlled trials have administered the attention modification protocol remotely via the Internet to samples diagnosed with SAD found similar results (Boettcher, Berger, & Renneberg, 2012; Carlbring et al., 2012; Neubauer et al., 2013; Rapee et al., 2013); however, in all cases participants in both the AMC and ACC reported social anxiety symptom reductions of small to large effects sizes with no statistically significant differences between conditions. Furthermore, the studies did not report robust evidence of attentional biases before or after treatment, changes in

an attentional bias over the course of treatment, or a relationship between attentional bias and symptom changes. A recent meta-analysis indicated that attentional biases were much smaller for studies administering the protocols through the Internet than for those administered in laboratories (Mogoase, David, & Koster, 2014). Nonetheless, the range of effect sizes for symptom changes was quite large for Internet studies (i.e., Hedges g = .05–.97) and for laboratory studies (i.e., Hedges g = .02–.82). The authors of the meta-analysis noted “the available evidence regarding ABM clinical utility [outside] the laboratory is currently limited” (Mogoase et al., 2014, p. 18) and recommended additional research exploring Internet administrations that include extended follow-up assessments (e.g., 4+ months). The current randomized controlled study was designed to: (1) replicate and extend the initial findings presented by Schmidt et al. (2009) and Amir et al. (2008) by using word stimuli [in line with evidence from Heeren et al. (2015), Mogoase et al. (2014), and Rapee et al. (2013)] that were idiosyncratically selected; (2) evaluate the comparative impact of participation in a laboratory setting relative to remotely at home, adding to the currently limited literature; and (3) provide extended follow-up assessments evaluating the endurance of the AMC changes (i.e., at 4 and 8 months). There were five hypothesized outcomes. First, participants completing the AMC in the laboratory were expected to report significant reductions in SAD symptoms relative to participants completing the ACC in the laboratory. Second, participants completing the AMC remotely were expected to report significant reductions in SAD symptoms relative to participants completing the ACC remotely. Third, participants completing the AMC in the laboratory were expected to report greater SAD symptom reductions relative to those participating remotely. The additional symptom reduction was expected in the laboratory condition because it was thought of as an additive exposure to social situations. Fourth, participants in the AMC were expected to maintain symptom changes at 4 and 8 months after participation relative to participants in the ACC. Fifth, participants showing the greatest change in attentional biases in AMC were expected to show the greatest SAD symptom reduction, based on initial theory suggesting that AMC protocols lead to symptom improvement via change in attentional bias (Bar-Haim, 2010; Clarke, Notebaert, & MacLeod, 2014).

2. Method 2.1. Participants Participants included treatment-seeking community members (n = 108; 45 men [Mage = 34.22; SD = 11.77] and 63 women [Mage = 36.92; SD = 14.02]) recruited through print and social media. The advertisements solicited potential participants from people who self-identified as having symptoms of social anxiety and were interested in participating in a computer treatment study. Previous research demonstrated large effect sizes for differences in reported symptom reductions between the AMC and ACC groups with relatively small samples in each group (i.e., n = 18, Schmidt et al., 2009; n = 14, Amir et al., 2009); these results were used in selecting the sample size for the current study. Interested individuals were told previous research suggested the treatment was effective in changing subconscious thought patterns and that they might experience a reduction in SAD symptoms, but reductions were not guaranteed. The purpose of comparing Internet and in-lab administrations of the dot-probe program was explained, followed by an explanation for having double-blinded treatment and control conditions. Those who agreed to participate were then administered (1) the Structured Clinical Interview for DSM-IV (SCID-I; First, Gibbon, Spitzer, & Williams, 2002) SAD

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

module, (2) the Social Interaction and Phobia Scale (Carleton et al., 2009), and (3) a condensed version of the Mini-International Neuropsychiatric Interview (MINI) screen (Sheehan et al., 1998). The entire SCID-I was not administered because comorbidity was not a focus of the current study and the researchers wanted to minimize participation burden wherever possible; as such, the M.I.N.I. screen was used to assess for exclusion criteria based on possible comorbidity and full criteria for comorbidity were not assessed. Individuals were excluded if they reported evidence of suicidal intent, substance dependence within the past 3 months, current/past schizophrenia, bipolar disorder, or were currently participating in any other therapy (e.g., cognitive behavioral therapy). Medication status was assessed by self-report and, to be eligible, participants were required to have had a stable dosage for the 3 months prior to participating and were required to maintain dosages while participating. All participants were required to meet diagnostic criteria for SAD based on the DSM-IV-TR (American Psychiatric Association, 2000) and self-identify SAD as their primary concern if possible comorbid disorders (e.g., depression) were identified by the M.I.N.I. screen. Trained clinical psychology graduate students (i.e., Doctoral candidates, Master’s candidates) or research coordinators (i.e., pre-Master’s students with baccalaureate honors degrees in psychology) under the supervision of the first author completed the telephone interviews. The first author provided the same training to all interviewers. The training included assigned readings on SAD, interactive assessments of the interviewers to ensure appropriate didactic knowledge of criteria, listening with the first author to at least three interviews conducted by experienced assessors, participating in a dual assessment with the first author, and being observed for at least one assessment by the first author. Thereafter, the first author reviewed each of the diagnostic interviews, provided supervision to the interviewers, and where appropriate, the interviewer conducted a follow-up telephone assessment interview to further ensure diagnostic criteria were met. Participants were only included if both the interviewer and the first author agreed on the diagnostic status regarding SAD. Inter-rater reliability was not assessed further; however, earlier versions of the SCID-I have adequate inter-rater reliability, ranging from .69 to 1.0 (Zanarini & Frankenburg, 2001), supporting the probability of accurate diagnostic status. Participants reported a minimum of 1 year of distressing SAD symptoms and an average of 21.07 years (SD = 14.47). Recruitment efforts were active for 11 months and resulted in 261 inquiries, 147 interviews, 118 eligible participants, and 108 participants who began the study. Most participants completed some postsecondary education (75%), graduated from high school (18%), or completed some high school but did not graduate (7%). The majority described themselves as Caucasian (86%), Asian (6%), or Aboriginal (4%), and as either single (46%), married/cohabitating (43%), or divorced (9%). 2.2. Measures Social Interaction Phobia Scale (SIPS; Carleton et al., 2009). The SIPS is a 14-item self-report measure designed to assess symptoms specific to SAD (e.g., “When mixing socially I am uncomfortable”). Each item is measured on a 5-point Likert scale, ranging from 0 (not at all) to 4 (extremely). The total score provides sufficient sensitivity and specificity for discerning clinical and nonclinical samples (i.e., the cut-off score of 21 with reports of symptoms interfering with daily activities can distinguish persons reporting clinically significant distress). Research has replicated the psychometric properties, as well as convergent and discriminant validity, of the SIPS in large, independent samples (Carleton et al., 2014; Duranceau, Peluso, Collimore, Asmundson, & Carleton, 2014; Menatti et al., in press; Reilly, Carleton, & Weeks, 2012). In those studies the SIPS correlated

37

well with the Brief Fear of Negative Evaluation Scale, Straightforward Items (Rodebaugh et al., 2004) (i.e., rs = .57–.72), the Fear of Positive Evaluation (Weeks, Heimberg, & Rodebaugh, 2008) (i.e., rs = .54–.58), the Social Phobia Inventory (Connor et al., 2000) (i.e., r = .89), and the Liebowitz Social Anxiety Scale (Baker, Heinrichs, Kim, & Hofmann, 2002) (i.e., r = .67). The SIPS was included as a measure of the symptoms associated with SAD (Weeks, Carleton, et al., 2010; Weeks, Jakatdar, et al., 2010) and served as a primary outcome measure. In the current sample, internal consistency was acceptable for the total scores at each measurement (˛ at baseline = .87; post-treatment = .89; 4-month follow-up = .90; 8-month follow-up = .89). Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977). The CES-D is a 20-item measure that assesses symptoms of depression. Items are phrased as self-statements (e.g., “I did not feel like eating”). Respondents rate how frequently each item applied to them over the course of the past week using a four point Likert scale ranging from 0 (Rarely or none of the time[less than 1 day]) to 3 (Most or all of the time[5–7 days]). The CES-D is well-supported (Carleton et al., 2013) and was included as a measure of depression symptoms per previous attention modification studies, in which the symptoms have been highly comorbid with SAD (Stein & Stein, 2008). The CES-D served as a secondary outcome measure. In the current sample, internal consistency was acceptable for the total scores at each measurement (˛ at baseline = .91; post-treatment = .87; 4-month follow-up = .92; 8-month follow-up = .93). State-Trait Anxiety Inventory (STAI; Spielberger, Gorsuch, Luschene, Vagg, & Jacobs, 1983). The STAI is a 40-item self-report measure wherein respondents rate 20 items (e.g., “I am tense”) asking about their anxiety “right now, that is, at this moment”, and 20 items (e.g., “I feel nervous and restless”) asking about their anxiety “generally”. Items are rated on 4-point Likert scale ranging from 1 (almost never/not at all) to 4 (almost always/very much so). The psychometric properties of the STAI are well established, including high reliability and good construct validity (Barnes, Harp, & Jung, 2002; Spielberger et al., 1983). The STAI trait anxiety symptoms served as a secondary outcome measure in the present study. The internal consistency was acceptable for the total trait scores at each measurement (˛ at baseline = .92; post-treatment = .93; 4-month follow-up = .93; 8-month follow-up = .94). 2.3. AMC and ACC protocol details Inquisit 3 Web Edition from Millisecond Software was used to administer the protocols to all participants. The software allowed the protocols to be initiated simply by clicking on a web link provided to each participant. The protocols used in the current study paralleled protocols from previous studies (Amir, Beard, Cobb, et al., 2009; Amir & Taylor, 2012; Schmidt et al., 2009). The stimuli comprised 64 words derived from word lists with relevant fears for individuals with SAD (e.g., embarrassed, stupid; Amir, Beard, Cobb, et al., 2009) and 64 matched neutral words (e.g., ladder, roof). Word stimuli were used in line with recent evidence suggesting words may produce larger effect sizes than pictures (Mogoase et al., 2014). Attentional bias for threat is influenced by personal relevance of stimuli in addition to emotional valence (MacLeod & Rutherford, 1992; Mogg & Bradley, 2002; Riemann & McNally, 1995). Given that SAD is a heterogeneous disorder associated with different social concerns (Bögels et al., 2010), an idiosyncratic selection procedure was used prior to each session to ensure the relevance of the particular words used during training for each participant. In line with previous ABM research using an individualized word selection procedure (Amir et al., 2009), the program required that each participant rate the emotional intensity of each threat word from very negative (−3) to very positive (+3). The 20 words rated as most

38

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

emotionally negative by each participant were used as the threat words for his/her session.1 The valence of the neutral words was not assessed, but paired for length. Participants were instructed to sit approximately 30 cm from the computer screen. Words were presented in the center of the screen, approximately 1.5 cm from one another in size 16 Arial font, for 500 ms. Word pairs for each participant were presented in a randomized order. The Millisecond software manages differences in screen presentations (e.g., display sizes, native resolutions, internet browsers) by maintaining relative positioning, with the technical details available on the company website. 2.3.1. The AMC During each session, participants completed a total of 240 trials. The trials included randomized combinations of probe type (using E or F instead of •), probe position (top or bottom), and word type (neutral or threat). The trial word presentations consisted of 80 trials with neutral words only: 2 (probe type) × 2 (probe position) × 20 (word pairs) and 160 trials with one neutral word and one threat word: 4 (probe position) × 2 (threat word position) × 20 (word pairs). For trials with one neutral word and one threat word (i.e., 66% of the trials), the probe always followed the neutral word. In other words, without specific instruction to direct their attention away from the threat word, on 66% of the trials the participant directed their attention away to detect the probe. 2.3.2. The ACC Like the AMC condition, participants completed 240 trials in each session. The ACC was similar to the AMC procedure except that for trials with one neutral word and one threat word (i.e., 66% of the trials), the probe appeared with equal frequency in the position of the threat and neutral word. In other words, participants were not consistently directing their attention away from the threat word to detect the probe. In both conditions, two thirds of trials (i.e., 160 trials) contained a threat-neutral pairing. As such, there is no difference in the number of critical trials viewed by each participant; however, setting the proportion of neutral-neutral trials at 33% resulted in the same number of trials to analyze (80) in each of the incongruent, congruent, and neutral trials. For all of the threat-neutral trials in the AMC group (160/160), the probe replaced the neutral word. In the ACC group, the probe replaced the neutral word for half of the trials (80/160). As such, there was a 2:1 ratio of active training trials for the AMC group relative to the ACC group. For comparison, in the Schmidt et al. (2009) study participants viewed 128 threat-neutral pairs. For their AMC group, on 128/128 trials attention was directed away from the threat stimulus. For the ACC group, on 64/128 trials attention was directed away from the threat stimulus. Accordingly, the 2:1 ratio of active training trials in the AMC vs. ACC group in the current study is identical to the ratio from Schmidt et al. (2009). 2.4. Procedure The current research was approved by the university research ethics board. Potential participants responded to the advertising by telephone or email, at which time participation details were provided along with an invitation to complete the telephone interview. Following the telephone interview, eligible individuals were

1 To ensure our results were not an artifact of the idiosyncratic rating procedure employed, we carried out a secondary analysis comparing (a) those who rated 70% or more of the words from Amir et al.’s (2008) word list as threatening (i.e., rated −1 or lower) as “high” in likelihood to be biased toward threatening words with (b) those who rated fewer than 70% of the words as threatening as “low”. No significant attentional bias was observed in either group, and indeed only the “low” raters showed the bias in the predicted direction.

then invited to participate and those that consented were randomly assigned, first into one of the participation modes (i.e., participating in the lab, “lab” or at home, “remote”), and then into either completing the AMC or the ACC protocol. Random assignment was based on order of presentation for participation, stratified by sex, and determined before recruitment (see CONSORT diagram, Fig. 1). Informed consent was obtained and all participants completed a web-based questionnaire package containing the aforementioned self-report measures as well as an initial ACC session to establish a baseline for the presence/absence and size of attentional biases. Participants were scheduled to complete two sessions per week for four consecutive weeks, either in the lab or at home depending on the condition to which they were randomly assigned. Participants in the remote condition were sent links to the program twice a week. If the program was not completed within 30 hr, a reminder email with a new link was sent, followed by a telephone call reminder if the session was not completed in the subsequent 20 hr. Participants in the lab condition were also sent reminder emails and telephone calls following a similar schedule if they missed an appointment. Instructions for the remote condition emphasized the importance of completing the program in a

Inquired About Participation (n=261)

Declined to Participate 1 (n=114)

Completed Interview (n=147)

Eligible (n=118)

Ineligible 2 (n=29) 3

Opted Out (n=1) Randomized (n=117) Participated with valid data (n=108)

Failed to start 4 (n=8) Data Removed 5

(n=1)

AMC in Lab (n=26)

ACC in Lab (n=31)

AMC Remote (n=25)

ACC Remote (n=26)

Lost to Posttreatment 6 (n=3)

Lost to Posttreatment 6 (n=11)

Lost to Posttreatment 6 (n=6)

Lost to Posttreatment 6 (n=6)

Lost to 4-month 7 Follow-Up (n=4)

Lost to 4-month 7 Follow-Up (n=4)

Lost to 4-month 7 Follow-Up (n=2)

Lost to 4-month 7 Follow-Up (n=3)

Lost to 8-month 8 Follow-Up (n=3)

Lost to 8-month 8 Follow-Up (n=1)

Lost to 8-month 8 Follow-Up (n=5)

Lost to 8-month 8 Follow-Up (n=3)

Fig. 1. CONSORT diagram. Notes: AMC in Lab—Attentional Modification Condition administered in the lab; AMC Remote—Attentional Modification Condition administered at participant’s home; ACC in Lab–Attentional Control Condition administered in the lab; ACC Remote—Attentional Control Condition administered at participant’s home. 1. 85 declined participation; 23 were ineligible because of residence location; 2 were calling on behalf of their children; 1 was underage; 2 were overage; 1 had insufficient English skills. 2. 5 did not have SAD as primary psychological concern; 17 reported subclinical symptoms; 4 reported recent changes in medications or psychotherapy; 2 reported current suicidal intent; 1 reported a language barrier. 3. Did not report a reason, but decided not to participate. 4. 2 were too busy to participate; 1 recently changed medications; 1 reported having a nervous breakdown, not related to participation; all others did not report a reason. 5. Data from 1 participant were excluded because of validity problems after randomization. 6. 3 were too busy to participate; 2 had unsolvable computer problems (in the remote condition); 2 had family issues; 1 could not continue due to arthritic and optic pain; 1 did not believe it worked; all others did not report a reason. 7. 1 did not believe it worked and, as such, did not want to participate; 3 moved and could not be reached; 2 were too busy to participate; all others did not report a reason. 8. 1 was too busy; all others did not report a reason.

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

quiet place, free from distractions and interruptions (to mimic the lab setting). Following four consecutive weeks of participation, all participants completed a second web-based questionnaire package containing all of the aforementioned self-report measures as well as another ACC session. Identical web-based questionnaires with no ACC sessions were then completed at 4 and 8 months posttreatment. Participants were debriefed and told which treatment condition they had been assigned to after all data collection was completed. 2.5. Analyses 2.5.1. Sample characteristics Descriptive statistics were calculated for each of the dependent variables. Analysis of variance (ANOVA) with SPSS bootstrapping (1000 samples) was used to compare demographics and self-report measures at baseline amongst participants who discontinued after baseline, discontinued after post-treatment, discontinued after 4-month follow-up, and those who completed the 8-month followup. Comparisons of demographics and self-report measures at baseline were also made between women and men on the dependent measures using independent t-tests with bootstrapping. 2.5.2. Idiosyncratic word ratings The idiosyncratic ratings of words were compared across groups and assessed over time. The assessments were conducted to (1) assess for word selection biases that might influence the results, and (2) assess whether the word ratings changed as a function of participation. Participant ratings of threat words at baseline were compared to ratings at post-treatment. The proportion of threat words that were presented at both baseline and post-treatment was also compared across each group and condition combination using significance tests. The threat word selection procedure makes interpretations based on the proportions ambiguous; as such, the associated results should be used to characterize the process rather than to interpret changes in symptoms. A Chi-square test of independence was used to compare the proportions of words selected by participants in each treatment condition (treatment vs. control) and treatment location (in lab vs. remote) to be presented at both baseline and end of treatment. ANOVAs were used to compare the following across treatment condition (treatment vs. control) and treatment location (in lab vs. remote): (1) the average baseline ratings of words presented at both baseline and end of treatment, and (2) the average baseline ratings of words presented at end of treatment but not at baseline. 2.5.3. Dot probe reaction time (RT) data Mean RTs were computed for each participant in each condition and analyzed in a mixed-model ANOVA with the betweenparticipants factors of treatment condition (treatment vs. control) and treatment location (in lab vs. remote) and the withinparticipants factors of time (baseline vs. post-treatment) and probe congruence (incongruent vs. congruent). The primary variables of interest in the current investigation were changes in self-reported symptoms; however, if such changes reflect attentional modification, then the magnitude of symptom reduction should be positively correlated with the magnitude of the reduction of attentional bias. Accordingly, a Pearson’s correlation was also conducted to clarify the relationship between change in attentional biases and the magnitude of reduction in SAD symptoms. 2.5.4. Self-reported symptom changes A random-effects model, which is a commonly used to analyze clustered data (Hedeker & Gibbons, 2006; Liu, Lu, Mogg, Mallick, & Mehrotra, 2009) was used to test whether treatment condition

39

or treatment location were associated with change in the selfreport measures over time (i.e., the SIPS, the CES-D, the STAI-T). We used a random-effects model for the analysis of change over time because this analysis does not require complete data at all measurement occasions. In other words, it uses available data at each measurement occasion and does not exclude participants who are missing data at any given measurement occasion. The questionnaire response system was set up such that participants who began any portion of a questionnaire were prompted to complete all items before proceeding if any individual item went unanswered. As such, missing data did not occur within questionnaires for any participants. The number of missing data points are indicated by the sample sizes at each level in Fig. 1. Categorical fixed-effects factors in predicting these variables included treatment condition (treatment vs. control), treatment location (in lab vs. remote), ethnicity (white vs. other), education (no high-school completion vs. high school completion vs. completing post-secondary education), marital status (married vs. not married), and gender (men vs. women). Continuous fixed-effects factors included years with SAD and age. Fixed-effects factors other than treatment condition and treatment location were added to explore potential confounders in treatment (e.g., if men are more depressed prior to treatment) and to control for differences in these pre-existing characteristics. Time was the random effect in all models (baseline, end of treatment, 4-month follow-up, 8month follow-up). Treatment condition and time interactions were included to test whether changes in scores differed between participants in the treatment vs. control conditions. Interactions between treatment location and time were included to test for changes in scores between participants who participated in the lab versus who participated in a remote location. The three-way interaction amongst treatment condition, treatment location, and time was tested to explore whether changes in scores over time were a function of both treatment location and treatment condition (e.g., whether scores decreased only in participants in the treatment condition who participated in the lab). Effect sizes of change between the time periods were calculated using Cohen’s d. 3. Results 3.1. Sample characteristics Descriptive statistics of the dependent variables for each group are presented in Table 1. There were no statistically significant differences between participants who discontinued after baseline, discontinued after post-treatment, discontinued after 4-month follow-up, and those who completed the 8-month follow-up on age, F(3, 104) = 2.08, p = .11, 2 = .06, years with significant SAD symptoms, F(3, 104) = 1.41, p = .25, 2 = .04, the SIPS total score, F(3, 104) = .97, p = .41, 2 = .03, the CES-D, F(3, 104) = 2.29, p = .08, 2 = .06, or the trait scale of the STAI, F(3, 103) = 1.41, p = .24, 2 = .04. There were no significant differences between men and women, at baseline or at the 8-month follow-up, on age, years with significant SAD symptoms, the SIPS total score, the CES-D, or the trait scale of the STAI (all ps > .05, all r2 s < .03). There were also no differences in attrition rates between men and women, 2 (3) = 1.15, p = .77, V = .10, between active and control groups, 2 (3) = 2.42, p = .49, V = .15, or between the in lab and remote location groups, 2 (3) = 1.24, p = .74, V = .11; as such, no further attrition or intentto-treat analyses were conducted. 3.2. Idiosyncratic word ratings There were no differences in the baseline average ratings of threat words between participants who completed baseline but

40

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

Table 1 Dependent variable scores for each group for each time period. Time

Treatment condition in lab (AMC)

Effect sizes (d) differences

1 (n = 26)

2 (n = 23)

3 (n = 19)

4 (n = 16)

2v1

3v1

4v1

SIPS STAI-T CES-D

36.77 (7.98) 57.81 (9.78) 26.38 (11.38)

29.39 (9.51) 54.70 (10.84) 21.35 (9.71)

31.26 (8.60) 57.37 (9.96) 25.58 (13.74)

32.56 (9.68) 56.40 (8.41) 24.19 (11.37)

1.03 0.34 0.50

0.73 0.00 0.10

0.57 0.09 0.22

Time

Control condition in lab (ACC)

Effect sizes (d) differences

1 (n = 31)

2 (n = 20)

3 (n = 16)

4 (n = 15)

2v1

3v1

4v1

SIPS STAI-T CES-D

37.19 (8.13) 57.26 (10.84) 21.68 (13.22)

28.75 (11.17) 53.65 (11.61) 19.85 (10.04)

30.13 (8.64) 53.12 (12.63) 19.81 (10.76)

29.27 (11.54) 50.13 (11.53) 20.00 (13.03)

1.04 0.33 0.14

0.87 0.38 0.14

0.97 0.66 0.13

Time

Treatment condition at remote location (AMC)

Effect sizes (d) differences

1 (n = 25)

2 (n = 19)

3 (n = 17)

4 (n = 12)

2v1

3v1

4v1

SIPS STAI-T CES-D

34.76 (11.37) 50.76 (12.17) 20.84 (10.64)

27.53 (12.82) 47.79 (8.14) 17.26 (8.06)

27.76 (13.65) 47.94 (10.05) 20.82 (11.32)

26.50 (12.19) 46.33 (11.11) 15.58 (10.23)

0.69 0.22 0.33

0.58 0.19 0.02

0.69 0.33 0.47

Time

Control condition in remote location (ACC)

SIPS STAI-T CES-D

Effect sizes (d) differences

1 (n = 26)

2 (n = 20)

3 (n = 17)

4 (n = 14)

2v1

3v1

4v1

37.77 (11.64) 56.16 (7.81) 23.65 (11.81)

29.50 (11.17) 50.15 (13.63) 22.40 (10.94)

30.47 (10.94) 51.53 (10.06) 20.35 (9.95)

28.93 (8.75) 51.86 (12.90) 23.07 (12.72)

0.71 0.77 0.11

0.63 0.59 0.28

0.76 0.55 0.05

Notes: AMC, Attentional Modification Condition; ACC, Attention Control Condition; SIPS, Social Interaction Phobia Scale; STAI-T, State-Trait Anxiety Index, Trait Form; CES-D, Centre for Epidemiological Studies Depression Scale; d, Cohen’s d; 1, baseline; 2, post-treatment; 3, 4-month follow-up; 4, 8-month follow-up.

discontinued after baseline and participants who continued to participate in the treatment, t(105) = .65, p = .49, r2 < .01. Among participants who completed baseline and post-treatment, the average concordance rates of words selected for presentation (based on ratings of emotional intensity) at both baseline and end of treatment were as follows: ACC in Lab 46%; ACC Remote 45%; AMC in Lab 44%; AMC Remote 50%. The proportions were not statistically significantly different across the four groups, 2 (1) = 0.26, p = .61, V = .04. Furthermore, there were no statistically significant differences in the average baseline ratings of words presented at both baseline and at the end of treatment, F(3, 79) = .33, p = .80, 2 = .01, nor in the average baseline ratings of words presented at the end of treatment but not at baseline, F(3, 79) = .38, p = .77, 2 = .01. Among words presented at both baseline and at the end of treatment, the average ratings were higher (i.e., less distressing) at post-treatment than at baseline, F(1, 79) = 31.86, p < .001, partial 2 = .29; however, there was no interaction across the four groups, F(3, 79) = .06, p = .98, partial 2 < .01. Among words presented only at baseline, the average ratings were higher (i.e., less distressing) at post-treatment than at baseline, F(1, 79) = 26.34, p < .001, partial 2 = .25; however, there was no interaction across the four groups, F(3, 79) = .97, p = .41, partial 2 = .04. The results indicate all participants reported the words at baseline as less distressing at post-treatment, irrespective of treatment condition or location.

3.3. Dot probe task RT data RTs on the attentional bias task were screened for outliers by excluding trials with RTs that were greater than two standard deviations from the mean RT for each individual participant on each day. The outlier analysis was scaled to each individual’s mean response time, rather than implementing a general cut-off (e.g., 2000 ms), which could eliminate legitimate observations from slower participants. This resulted in exclusion of data from 2.37% of trials. Trials with errors (4.42%) were also excluded from analyses.

Data were available from both baseline and post-treatment for 82 participant completers. Response times post-treatment were overall faster (M = 618 ms; SE = 25.7) than baseline (M = 720 ms; SE = 32.9), leading to a significant main effect of time, F(1, 78) = 20.6, MSE = 40151, p < .001, 2p = .21. No other significant main effects or significant interactions were observed, all Fs < 2.41, all ps > .12. Across the entire sample of 82 participant completers, responses were 3 ms faster on congruent probe trials than on incongruent probe trials baseline, an effect that did not differ across the ACC and AMC groups and was not statistically greater than zero, t(81) = .76, p > .45. Based on initial theory suggesting that the AMC leads to symptom reduction by modifying an initial bias toward threat stimuli (Clarke et al., 2014), the magnitude of the change in bias from pre- to post-treatment should be correlated with the magnitude of reduction in SAD symptoms. For each participant, the change in bias toward threat words was determined by computing the average difference in response times on incongruent and congruent trials from pre- to post-treatment. The change in total SIPS scores over time was also computed, and Pearson’s correlations were computed among these three change scores, stratified by treatment type (ACC vs. AMC). Negative correlations (i.e., smaller post-treatment biases accompanied by larger symptom reductions) were expected in the AMC condition if attentional modification was associated with symptom reduction; no correlation should have been observed in the ACC condition. Contrasting expectations, the observed correlations were near zero in the AMC group (i.e., SIPS, r < .01, p = .97) and negative, but non-significant, in the ACC group (i.e., SIPS, r = −.26, p = .10).

3.4. Self-reported symptom changes 3.4.1. SAD symptoms The time factor was statistically significant in the randomeffects model (p < .0001) predicting SIPS scores over time. Paired contrasts demonstrated that SIPS scores at post-treatment were

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

significantly lower than SIPS scores at baseline (MD = −7.12, SE = 1.82, p < .001), that SIPS scores at the 4-month follow-up were significantly lower than SIPS scores at baseline (MD = −6.50, SE = 1.94, p < .001), and that SIPS scores at the 8-month follow-up were significantly lower than SIPS scores at baseline (MD = −7.03, SE = 2.08, p < .001). Without taking into account condition status, the current findings suggest that participants experienced reductions in SIPS scores, and that these reductions were maintained after 8 months. The interaction of treatment condition and time was not statistically significant (p > .50), suggesting that changes over time in SIPS scores did not differ between the treatment conditions. The interaction of treatment location and time was not statistically significant (p > .90), suggesting that changes over time in SIPS scores did not differ between individuals who participated in the lab and those who participated in a remote location. Moreover, the three-way interaction of treatment location, treatment condition, and time was not statistically significant (p > .80). None of the other fixed factors (i.e., ethnicity, education, marital status, gender, years with SAD, age) were statistically significant predictors of SIPS scores (ps > 30). SIPS total scores for each of the conditions at each time period are reported in Table 1. Changes in scores over time across the conditions and time periods were medium to large (Cohen’s d ranging from 0.57 to 1.04). Effect sizes of change at the 8-month follow-up (0.57 to 0.97) were comparable to those demonstrated immediately after treatment (0.69 to 1.04). 3.4.2. Trait anxiety (STAI-T scores) The time factor was statistically significant in the randomeffects model (p < .01) predicting STAI-T scores over time. Paired contrasts demonstrated that STAI-T scores at post-treatment were significantly lower than STAI-T scores at baseline (MD = −4.43, SE = 1.96, p < .05). STAI-T scores at the 4-month follow-up were not statistically significantly lower than STAI-T scores at baseline (MD = −3.07, SE = 2.06, p > .10), and scores at the 8-month followup were not statistically significantly lower than STAI-T scores at baseline (MD = −2.61, SE = 2.21, p > .20). The results suggest that participants experienced reductions during treatment that were not sustained over the 4- and 8-month follow-ups, irrespective of condition. The interaction of treatment condition and time was not statistically significant (p > .30), the interaction of treatment location and time was not statistically significant (p > .90), and the three-way interaction of treatment location, treatment condition, and time was not statistically significant (p > .10). No other fixedeffect factors (i.e., ethnicity, education, marital status, gender, years with SAD, age) were statistically significant predictors of STAI-T scores over time (ps > .30). 3.4.3. Depression symptoms (CES-D scores) The time factor was not statistically significant in the random-effects model (p > .20) predicting CES-D scores over time, suggesting participants did not experience reductions in depression symptoms during the study. The interaction of treatment condition and time was not statistically significant (p > .10), the interaction of treatment location and time was not statistically significant (p > .80), and the three-way interaction of treatment location, treatment condition, and time was not statistically significant (p > .80). The level of education factor was a statistically significant predictor of CES-D scores (p < .05). The paired contrasts indicated that participants who did not complete high school reported greater CES-D scores (10.91, SE = 4.03, p < .01) than participants who completed education greater than high school. Participants who completed high school did not report different CES-D scores relative to those who received education greater than high-school (MD = −2.20, SE = 2.60, p > .30). No other fixed-effect

41

factors (i.e., ethnicity, marital status, gender, years with SAD, age) were statistically significant predictors of CES-D scores (ps > .30).

4. Discussion The current study was designed to replicate and extend previous findings that supported using AMC, but not ACC, to reduce SAD symptoms (Amir et al., 2008; Schmidt et al., 2009). The study also addresses recommendations from a recent meta-analysis that more research assessing remote administration of AMC is needed (Mogoase et al., 2014). The current design used word stimuli (in line with Rapee et al., 2013) that were idiosyncratically selected to ensure individual relevance, evaluated the comparative impact of participation in a laboratory setting relative to remotely at home, and provided extended follow-up assessments evaluating the endurance of the AMC changes (i.e., at 4 and 8 months). There were five specific hypotheses that were tested. First, in line with prior reports (Amir, Beard, Taylor, et al., 2009; Boettcher et al., 2012; Carlbring et al., 2012; Neubauer et al., 2013), participants completing the AMC in the laboratory were expected to report significant SAD symptom reductions relative to participants completing the ACC in the laboratory. The hypothesis was not supported as participants in the AMC in the laboratory did report significant reductions in SAD symptoms, but not significantly more than those in the ACC. Second, participants who completed the AMC remotely were expected to report significant reductions in SAD symptoms relative to participants who completed the ACC remotely. The hypothesis was not supported, as participants completing the AMC remotely did report significant reductions in SAD symptoms, but not significantly more than those in the ACC. Third, participants completing the AMC in the laboratory were expected to report greater SAD symptom reductions than those in the remote condition because attending the laboratory to participate was conceptualized as an additive component of exposure. The hypothesis was not supported. Participants completing the active condition in the laboratory and those participating remotely reported comparable reductions in SAD symptoms. The effect sizes for the reductions were large and comparable to some of the effect sizes from the active conditions in previous anxiety sample research (Mogoase et al., 2014). Fourth, participants in the AMC were expected to maintain symptom changes at 4- and 8-month follow-ups relative to participants in the ACC. Again, the hypothesis was not supported, because both AMC and ACC conditions maintained symptom reductions. The final hypothesis was that participants showing the greatest change in attentional biases in the AMC would show the greatest SAD symptom reduction. The hypothesis was also not supported. No significant correlation was observed between the magnitude of the attentional bias to threat stimuli remaining post-treatment and changes in SIPS total scores in the AMC; moreover, response times overall were no faster on congruent probe trials than on incongruent trials, suggesting participants in this study were not strongly biased toward threat stimuli. Contrasting some prior findings (Amir et al., 2008; Schmidt et al., 2009), the current results did not indicate a differential effect between the AMC and ACC on SAD symptoms; however, the current results are in line with other research investigating the efficacy of an Internet-delivered attention modification protocol (Boettcher et al., 2012; Enock et al., 2014; Heeren et al., 2015). In line with the recent Mogoase et al. (2014) meta-analysis, the symptom reductions appear to be relatively uninfluenced by change in attentional bias. Indeed, the meta-analysis indicated an absence of evidence for “direct or indirect effect of the preexistent [attentional bias] on the change in symptoms” (Mogoase et al., 2014, p. 18). The reductions also appear to be relatively uninfluenced by participation location

42

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

(i.e., in lab or at home) or the use of idiosyncratic word stimuli (as opposed to static or pictorial stimuli). Symptom reductions with effect sizes comparable to any previous ABM study (Mogoase et al., 2014) appear achievable with remote administration and without measurable change in attention processes; however, the variability across this and previous studies (Boettcher et al., 2012; Carlbring et al., 2012; Heeren et al., 2012, 2015; Mogoase et al., 2014; Neubauer et al., 2013; Rapee et al., 2013) suggests there are still substantial gaps in understanding the mechanisms associated with symptom changes subsequent to attention modification treatment protocols. The current results support the specificity of attention modification protocols in that SAD symptoms were targeted, altered, and changes were maintained over time, whereas changes in trait anxiety were transient and there were no changes in depression symptoms. That relative specificity may serve as support against a generalized placebo effect as the mechanism of change behind attention modification protocols. Despite the specificity and effect size, in the absence of a waitlist control the change in symptoms may be the result of the passage of time; however, a recent study that included a waitlist condition demonstrated similar effect size reductions for social anxiety symptoms in both the AMC (i.e., ds = .71–1.06) and ACC (ds = .60–.90) conditions and non-significant reductions with small effect sizes (i.e., ds = .17–.23) in the waitlist condition (Enock et al., 2014). In any case, the absence of differences between the AMC and ACC combined with the presence of specific symptom improvements might be explained most parsimoniously by expectancy (i.e., placebo) effects; however, an alternative explanation, though more complex, involves considering the entire protocol as providing exposure and increasing attentional control (Klumpp & Amir, 2010). Explanations of the current results are also complicated by the fact that no evidence was found for a robust attentional bias associated with social threat stimuli at baseline. In line with a recent meta-analysis (Mogoase et al., 2014), the current results also provided no evidence of a relationship between change in attentional bias and the large symptom reductions. Given the longstanding precedent research indicating the robust (if occasionally small) nature of attentional biases (Asmundson & Stein, 1994; Chen et al., 2002; Lee & Telch, 2008; Mattia, Heimberg, & Hope, 1993; Mogoase et al., 2014) and research providing evidence of a reduction in symptoms despite absent changes in attentional biases (e.g., Boettcher et al., 2012; Heeren et al., 2015), there may be important considerations going forward with respect to protocol design (e.g., stimuli presentation type, stimuli spacing, presentation speed) and analytic decisions (e.g., comparisons of specific RT pairs). Theories underlying attention bias modification protocols would benefit from being further explored and refined. Attentional control rather than change in attentional bias per se may be responsible for symptom change (Heeren et al., 2015; Klumpp & Amir, 2010); however, additional studies would be needed to support this claim. Future research could test this hypothesis by assessing symptom change following an attention modification procedure solely comprised of neutral stimuli (e.g., shapes of different colors, letters of the alphabet, numbers; Bar-Haim, 2010). Several limitations of the current study provide directions for future research. First, symptom measurements were limited to selfreport. Future research should explore whether reported changes in symptoms are reflected in behavioral changes (see Heeren et al., 2015). Second, expectancy effects were not directly controlled for in that participants were not asked whether they expected the treatment to reduce their symptoms; nevertheless, the very nature of a randomized controlled trial suggests against the utility of such a question while simultaneously reducing expectancy effects. Third, there was an average SIPS score reduction of 7.83 from baseline to post-treatment across all groups, representing an average

reduction of 21%. Whether the reduction is clinically significant is open to debate (Jacobson & Truax, 1991); however, the intervention appears innocuous, inexpensive, timely, and can be made broadly accessible. Nonetheless, in the absence of changes specific to the active condition or the passage of time (Enock et al., 2014), additional research should be conducted to determine what mechanisms (e.g., a placebo effect) beyond the passage of time are driving the modest symptom reduction. Fourth, there were uncontrolled environmental conditions for participants who completed the protocols remotely. Those participants may have been exposed to varying amounts of external distraction that may have impacted results; that said, such distractions could reasonably be expected to reduce the benefits for the remote condition and no such reductions were observed. Fifth, using words rather than faces may have mitigated differences or reduced the magnitude of attentional biases. Unlike some studies (Amir et al., 2008; Schmidt et al., 2009), the current study used the same type of stimuli to measure biases at baseline and post-treatment, as well as to modify the biases. Using the same type of stimuli should have facilitated detecting changes in biases from baseline to post-treatment; however, participants may have habituated to the stimuli, diminishing the salience of threat vs. neutral stimuli. Future research should directly compare the relative impact of using different types of stimuli. Sixth, although the potential threat words were individually selected by each participant rating, the neutral words were not because of a programmatic limitation. There is no way to know whether the neutral words were neutral. Future research could also rate neutral words to maximally tailor the stimuli to each participant. Seventh, the current sample sizes per condition at posttreatment (i.e., n = 19–23) were slightly smaller than the average for similar research (i.e., n = 10–57; mean ∼28.6; Mogoase et al., 2014). Larger samples may yield differences between conditions; however, that possibility is unlikely given the small actual differences in responses between conditions (see Table 1). Eighth, future studies could also include a waitlist control to control for the passage of time (i.e., Enock et al., 2014). Attentional bias for threat can be divided into three distinct components (i.e., facilitated attention, difficulty in disengagement, avoidance) occurring at different stages of the cognitive processing sequence (i.e., automatic, strategic; Cisler & Koster, 2010). Evidence suggests that reducing difficulty in disengagement specifically may be most beneficial to SAD patients (Amir et al., 2003; Buckner et al., 2010; Heeren et al., 2012). As such, future research should compare the impact of using different training procedures (i.e., re-training attention toward threat vs. away from threat) at different stages of the cognitive processing sequence (i.e., using different stimuli presentation times). There may also be important considerations regarding stimuli sizes and relative positions, all of which warrant additional research. Finally, the use of idiosyncratic stimuli may yet have impacted the current results in a way not previously accounted for; therefore, future research should compare idiosyncratic stimuli and static stimuli. The current study is the first study to directly compare in lab and remote conditions. The current study is also the first to demonstrate that both active and control condition groups, in lab and remotely, can experience symptom reductions in the absence of an identified attentional bias at baseline and with no concomitant change in attentional bias after treatment. Furthermore, those symptom reductions were comparable in magnitude to each other and to reductions in any previous research using an ABM protocol. In summary, using the attention modification protocol with idiosyncratic word stimuli, irrespective of the condition or location, provided sustained and substantial reductions in SAD symptoms. The symptom reductions also appeared unrelated to the presence of, or changes in, attention biases. The current results indicate that, if research is to continue with ABM protocols, more emphasis must

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

be placed on understanding the underlying mechanisms responsible for the disparate results reported in the current literature.

Acknowledgements The authors would like to thank the Saskatchewan Health Research Foundation for making this research possible through a New Investigator Establishment Grant and a New Investigator Equipment Grant (SHRF2456) awarded to R. N. Carleton.

References American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders (4th text revision ed.). Washington, DC: Author. American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Washington, DC: Author. Amir, N., Beard, C., Cobb, M., & Bomyea, J. (2009). Attention modification program in individuals with generalized anxiety disorder. Journal of Abnormal Psychology, 118, 28–33. http://dx.doi.org/10.1037/a0012589 Amir, N., Beard, C., Taylor, C. T., Klumpp, H., Elias, J., Burns, M., et al. (2009). Attention training in individuals with generalized social phobia: A randomized controlled trial. Journal of Consulting and Clinical Psychology, 77, 961–973. http://dx.doi.org/ 10.1037/a0016685 Amir, N., Elias, J., Klumpp, H., & Przeworski, A. (2003). Attentional bias to threat in social phobia: facilitated processing of threat or difficulty disengaging attention from threat? Behaviour Research and Therapy, 41, 1325–1335. Amir, N., & Taylor, C. T. (2012). Combining computerized home-based treatments for generalized anxiety disorder: An attention modification program and cognitive behavioral therapy. Behavior Therapy, 43, 546–559. http://dx.doi.org/10.1016/j. beth.2010.12.008 Amir, N., Taylor, C. T., & Donohue, M. C. (2011). Predictors of response to an attention modification program in generalized social phobia. Journal of Consulting and Clinical Psychology, 79, 533–541. http://dx.doi.org/10.1037/a0023808 Amir, N., Weber, G., Beard, C., Bomyea, J., & Taylor, C. T. (2008). The effect of a single-session attention modification program on response to a public-speaking challenge in socially anxious individuals. Journal of Abnormal Psychology, 117, 860–868. http://dx.doi.org/10.1037/a0013445 Antony, M. M., & Swinson, R. P. (2000). The shyness and social anxiety workbook: Proven techniques for overcoming your fears. Oakland, CA: New Harbinger Publications. Asmundson, G. J. G., & Stein, M. B. (1994). Selective processing of social threat in patients with generalized social phobia: Evaluation using a dot-probe paradigm. Journal of Anxiety Disorders, 8, 107–117. http://dx.doi.org/10.1016/ 0887-6185(94)90009-4 Baker, S. L., Heinrichs, N., Kim, H.-J., & Hofmann, S. G. (2002). The Liebowitz social anxiety scale as a self-report instrument: A preliminary psychometric analysis. Behaviour Research and Therapy, 40, 701–715. http://dx.doi.org/10.1016/S00057967(01)00060-2 Barnes, L. L. B., Harp, D., & Jung, W. S. (2002). Reliability generalization of scores on the Spielberger State-Trait Anxiety Inventory. Educational and Psychological Measurement, 62, 603–618. Bar-Haim, Y. (2010). Research Review: attention bias modification (ABM): A novel treatment for anxiety disorders. Journal of Child Psychology and Psychiatry, 51, 859–870. http://dx.doi.org/10.1111/J.1469-7610.2010.02251.X Boettcher, J., Berger, T., & Renneberg, B. (2012). Does a pre-treatment diagnostic interview affect the outcome of internet-based self-help for social anxiety disorder? A randomized controlled trial. Behavioural and Cognitive Psychotherapy, 40, 513–528. http://dx.doi.org/10.1017/S1352465812000501 Bögels, S. M., Alden, L., Beidel, D. C., Clark, L. A., Pine, D. S., Stein, M. B., et al. (2010). Social anxiety disorder: Questions and answers for the DSM-V. Depression and Anxiety, 27, 168–189. http://dx.doi.org/10.1002/da.20670 Buckner, J. D., Maner, J. K., & Schmidt, N. B. (2010). Difficulty disengaging attention from social threat in social anxiety. Cognitive Therapy & Research, 34, 99–105. http://dx.doi.org/10.1007/s10608-008-9205-y Carlbring, P., Apelstrand, M., Sehlin, H., Amir, N., Rousseau, A., Hofmann, S. G., et al. (2012). Internet-delivered attention bias modification training in individuals with social anxiety disorder—A double blind randomized controlled trial. BMC Psychiatry, 12, 66. http://dx.doi.org/10.1186/1471-244X-12-66 Carleton, R. N., Collimore, K. C., Asmundson, G. J. G., McCabe, R. E., Rowa, K., & Antony, M. M. (2009). Refining and validating the Social Interaction Anxiety Scale and the Social Phobia Scale. Depression and Anxiety, 26, E71–E81. http://dx.doi.org/ 10.1002/da.20480 Carleton, R. N., Thibodeau, M. A., Teale, M. J., Welch, P. G., Abrams, M. P., Robinson, T., et al. (2013). The center for epidemiologic studies depression scale: A review with a theoretical and empirical examination of item content and factor structure. PLoS ONE, 8, e58067. http://dx.doi.org/10.1371/journal.pone.0058067 Carleton, R. N., Thibodeau, M. A., Weeks, J. W., Teale Sapach, M. J., McEvoy, P. M., Horswill, S. C., et al. (2014). Comparing short forms of the Social Interaction Anxiety Scale and the Social Phobia Scale. Psychological Assessment, 26, 1116–1126. http://dx.doi.org/10.1037/a0037063

43

Chen, Y. P., Ehlers, A., Clark, D. M., & Mansell, W. (2002). Patients with generalized social phobia direct their attention away from faces. Behaviour Research and Therapy, 40, 677–687. Cisler, J. M., & Koster, E. H. (2010). Mechanisms of attentional biases towards threat in anxiety disorders: An integrative review. Clinical Psychology Review, 30, 203–216. http://dx.doi.org/10.1016/j.cpr.2009.11.003 Clarke, P. J. F., Notebaert, L., & MacLeod, C. (2014). Absence of evidence or evidence of absence: Reflecting on therapeutic implementations of attentional bias modification. BMC Psychiatry, 14, 8. http://dx.doi.org/10.1186/1471-244X-14-8 Connor, K. M., Davidson, J. R., Churchill, L. E., Sherwood, A., Foa, E., & Weisler, R. H. (2000). Psychometric properties of the Social Phobia Inventory (SPIN): New self-rating scale. British Journal of Psychiatry, 176, 379–386. Cougle, J. R., Keough, M. E., Riccardi, C. J., & Sachs-Ericsson, N. (2009). Anxiety disorders and suicidality in the National Comorbidity Survey-Replication. Journal of Psychiatric Research, 43, 825–829. http://dx.doi.org/10.1016/j.jpsychires.2008. 12.004 Duranceau, S., Peluso, D. L., Collimore, K. C., Asmundson, G. J. G., & Carleton, R. N. (2014). “Social Interaction Phobia Scale”: Propriétés psychométriques de la version franc¸aise. Canadian Journal of Behavioral Science/Revue Canadienne des Sciences du Comportement, 46, 406–413. http://dx.doi.org/10.1037/a0033036 Enock, P. M., Hofmann, S. G., & McNally, R. J. (2014). Attention bias modification training via smartphone to reduce social anxiety: A randomized, controlled multi-session experiment. Cognitive Therapy and Research, 38, 200–216. http:// dx.doi.org/10.1007/s10608-014-9606-z First, M. B., Gibbon, M., Spitzer, R. L., & Williams, J. B. W. (2002). Structured clinical interview for DSM-IV axis I disorders—Patient edition. New York, NY: New York State Psychiatric Institute, Biometrics Research Department. Grant, B. F., Hasin, D. S., Stinson, F. S., Dawson, D. A., Patricia Chou, S., June Ruan, W., et al. (2005). Co-occurrence of 12-month mood and anxiety disorders and personality disorders in the US: Results from the national epidemiologic survey on alcohol and related conditions. Journal of Psychiatric Research, 39, 1–9. http:// dx.doi.org/10.1016/j.jpsychires.2004.05.004 Hedeker, D., & Gibbons, R. D. (2006). Longitudinal data analysis. Hoboken, NJ: Wiley & Sons. Heeren, A., Lievens, L., & Philippot, P. (2011). How does attention training work in social phobia: Disengagement from threat or re-engagement to non-threat? Journal of Anxiety Disorders, 25, 1108–1115. http://dx.doi.org/10.1016/j.janxdis. 2011.08.001 Heeren, A., Mogoase, C., McNally, J. D., Schmitz, A., & Philibert, P. (2015). Does attention bias modification improve attentional control? A double-blind randomized experiment with individuals with social anxiety disorder. Journal of Anxiety Disorders, 29, 35–42. http://dx.doi.org/10.1016/j.janxdis.2014.10.007 Heeren, A., Reese, H. E., McNally, R. J., & Philippot, P. (2012). Attention training toward and away from threat in social phobia: Effects on subjective, behavioral, and physiological measures of anxiety. Behaviour Research and Therapy, 50, 30–39. http://dx.doi.org/10.1016/j.brat.2011.10.005 Heimberg, R. G., Brozovich, F. A., & Rapee, R. M. (2010). A cognitive-behavioral model of social anxiety disorder: Update and extension. In: S. G. Hofmann, & P. M. DiBartolo (Eds.), Social anxiety: Clinical, developmental, and social perspectives. New York, NY: Elsevier. Hirsch, C. R., & Clark, D. M. (2004). Information-processing bias in social phobia. Clinical Psychology Review, 24, 799–825. http://dx.doi.org/10.1016/j.cpr.2004.07. 005 Hofmann, S. G. (2007). Cognitive factors that maintain social anxiety disorder: A comprehensive model and its treatment implications. Cognitive Behaviour Therapy, 36, 193–209. http://dx.doi.org/10.1080/16506070701421313 Jacobson, N. S., & Truax, P. (1991). Clinical significance: A statistical approach to defining meaningful change in psychotherapy research. Journal of Consulting and Clinical Psychology, 59, 12–19. Julian, K., Beard, C., Schmidt, N. B., Powers, M. B., & Smits, J. A. (2012). Attention training to reduce attention bias and social stressor reactivity: An attempt to replicate and extend previous findings. Behaviour Research and Therapy, 50, 350–358. http://dx.doi.org/10.1016/j.brat.2012.02.015 Klumpp, H., & Amir, N. (2010). Preliminary study of attention training to threat and neutral faces on anxious reactivity to a social stressor in social anxiety. Cognitive Therapy and Research, 34(3), 263–271. http://dx.doi.org/10.1007/s10608-0099251-0 Koster, E. H., Fox, E., & MacLeod, C. (2009). Introduction to the special section on cognitive bias modification in emotional disorders. Journal of Abnormal Psychology, 118, 1–4. http://dx.doi.org/10.1037/a0014379 Lee, H. J., & Telch, M. J. (2008). Attentional biases in social anxiety: An investigation using the inattentional blindness paradigm. Behaviour Research and Therapy, 46, 819–835. Liu, G., Lu, K., Mogg, R., Mallick, M., & Mehrotra, D. V. (2009). Should baseline be a covariate or dependent variable in analyses of change from baseline in clinical trials? Statistics in Medicine, 28, 2509–2530. http://dx.doi.org/10.1002/sim.3639 MacLeod, C., & Rutherford, E. M. (1992). Anxiety and the selective processing of emotional information: Mediating roles of awareness, trait and state variables, and personal relevance of stimu. Behaviour research and therapy, 30(5), 479–491. MacLeod, C., Rutherford, E., Campbell, L., Ebsworthy, G., & Holker, L. (2002). Selective attention and emotional vulnerability: Assessing the causal basis of their association through the experimental manipulation of attentional bias. Journal of Abnormal Psychology, 111, 107–123. http://dx.doi.org/10.1037/0021-843X.111. 1.107 Massion, A. O., Dyck, I. R., Shea, M. T., Phillips, K. A., Warshaw, M. G., & Keller, M. B. (2002). Personality disorders and time to remission in generalized anxiety

44

R.N. Carleton et al. / Journal of Anxiety Disorders 33 (2015) 35–44

disorder, social phobia, and panic disorder. Archives of General Psychiatry, 59, 434–440. http://dx.doi.org/10.1001/archpsyc.59.5.434 Mattia, J. I., Heimberg, R. G., & Hope, D. A. (1993). The revised Stroop color-naming task in social phobics. Behaviour Research and Therapy, 31, 305–313. http://dx. doi.org/10.1016/0005-7967(93)90029-T Menatti, A. R., Weeks, J. W., Carleton, R. N., Morrison, A. S., Heimberg, R. G., Hope, D. A., et al. (2015). The Social Interaction Phobia Scale: Continued support for the psychometric validity of the SIPS using clinical and non-clinical samples. Journal of Anxiety Disorders, 32, 46–55. http://dx.doi.org/10.1016/j.janxdis.2015.03.003 Mogg, K., & Bradley, B. P. (2002). Selective orienting of attention to masked threat faces in social anxiety. Behaviour Research and Therapy, 40, 1403–1414. http:// dx.doi.org/10.1016/S0005-7967(02)00017-7 Mogoase, C., David, D., & Koster, E. H. W. (2014). Clinical efficacy of attentional bias modification procedures: An updated meta-analysis. Journal of Clinical Psychology, 70, 1133–1157. http://dx.doi.org/10.1002/Jclp.22081 Neubauer, K., von Auer, M., Murray, E., Petermann, F., Helbig-Lang, S., & Gerlach, A. L. (2013). Internet-delivered attention modification training as a treatment for social phobia: A randomized controlled trial. Behaviour Research and Therapy, 51, 87–97. http://dx.doi.org/10.1016/j.brat.2012.10.006 Radloff, L. S. (1977). The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385–401. http:// dx.doi.org/10.1177/014662167700100306 Rapee, R. M., Macleod, C., Carpenter, L., Gaston, J. E., Frei, J., Peters, L., et al. (2013). Integrating cognitive bias modification into a standard cognitive behavioural treatment package for social phobia: A randomized controlled trial. Behaviour Research and Therapy, 51, 207–215. http://dx.doi.org/10.1016/j.brat.2013.01.005 Reilly, A. R., Carleton, R. N., & Weeks, J. W. (2012). Psychometric evaluation of the Social Interaction Phobia Scale. Anxiety, Stress, and Coping, 25, 529–542. http:// dx.doi.org/10.1080/10615806.2011.598150 Riemann, B. C., & McNally, R. J. (1995). Cognitive processing of personally relevant information. Cognition and Emotion, 9, 325–340. http://dx.doi.org/10.1080/ 02699939508408970 Rodebaugh, T. L., Woods, C. M., Thissen, D. M., Heimberg, R. G., Chambless, D. L., & Rapee, R. M. (2004). More information from fewer questions: The factor structure and item properties of the original and brief fear of negative evaluation

scale. Psychological Assessment, 16, 169–181. http://dx.doi.org/10.1037/10403590.16.2.169 Ruscio, A. M., Brown, T. A., Chiu, W. T., Sareen, J., Stein, M. B., & Kessler, R. C. (2008). Social fears and social phobia in the USA: Results from the National Comorbidity Survey Replication. Psychological Medicine, 38, 15–28. http://dx.doi.org/10.1017/ S0033291707001699 Schmidt, N. B., Richey, J. A., Buckner, J. D., & Timpano, K. R. (2009). Attention training for generalized social anxiety disorder. Journal of Abnormal Psychology, 118, 5–14. http://dx.doi.org/10.1037/a0013643 Sheehan, D. V., Lecrubier, Y., Harnett-Sheehan, K., Amorim, P., Janavs, J., Weiller, E., et al. (1998). The Mini-International Neuropsychiatric Interview (M.I.N.I.): The development and validation of a structured diagnostic psychiatric interviewfor DSM-IV and ICD-10. Journal of Clinical Psychiatry, 59, 22–33. Spielberger, C. D., Gorsuch, R. L., Luschene, R. E., Vagg, P. R., & Jacobs, G. A. (1983). Manual for the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press. Stein, M. B., & Stein, D. J. (2008). Social anxiety disorder. Lancet, 371, 1115–1125. Thibodeau, M. A., Welch, P. G., Sareen, J., & Asmundson, G. J. G. (2013). Anxiety disorders are independently associated with suicide ideation and attempts: Propensity score matching in two epidemiological samples. Depression and Anxiety, 30, 947–954. http://dx.doi.org/10.1002/Da.22203 Weeks, J. W., Carleton, R. N., Asmundson, G. J. G., McCabe, R. E., & Antony, M. M. (2010). “Social Anxiety Disorder Carved at its Joints”: Evidence for the taxonicity of social anxiety disorder. Journal of Anxiety Disorders, 24, 734–742. http://dx.doi. org/10.1016/j.janxdis.2010.05.006 Weeks, J. W., Heimberg, R. G., & Rodebaugh, T. L. (2008). The Fear of Positive Evaluation Scale: Assessing a proposed cognitive component of social anxiety. Journal of Anxiety Disorders, 22, 44–55. http://dx.doi.org/10.1016/j.janxdis.2007.08.002 Weeks, J. W., Jakatdar, T. A., & Heimberg, R. G. (2010). Comparing and contrasting fears of positive and negative evaluation as facets of social anxiety. Journal of Social and Clinical Psychology, 29, 68–94. http://dx.doi.org/10.1521/jscp.2010.29. 1.68 Zanarini, M. C., & Frankenburg, F. R. (2001). Attainment and maintenance of reliability of axis I and II disorders over the course of a longitudinal study. Comprehensive Psychiatry, 42, 369–374. http://dx.doi.org/10.1053/comp.2001.24556