Participants were 27 (12 female) adults who met DSM-IV (40
) criteria for primary generalized SAD and 27 (12 female) demographically-matched healthy controls with no history of DSM-IV psychiatric disorders. Patients and controls did not differ in handedness, age, education, or ethnicity (). Among patients, current Axis-I co-morbidity included 3 with generalized anxiety disorder and 3 with specific phobia. Past Axis-I co-morbidity included 1 with past major depression, 1 with past dysthymia, and 1 with past substance abuse. Ten patients reported past (i.e., ended more than 1 year ago) non-cognitive-behavioral psychotherapy, and 8 reported past pharmacotherapy. All participants provided informed consent.
Demographic and Clinical Variables
Participants passed an MRI safety screen, and were excluded for current pharmacotherapy or psychotherapy, past cognitive-behavioral therapy, history of medical disorders or head trauma, and current psychiatric disorders other than generalized anxiety disorder, agoraphobia without a history of panic attacks, or specific phobia.
Clinical and Individual Difference Assessment
Clinical assessments were conducted by two PhD-level clinical psychologists (PG, KW) and one graduate student (TM) using the Anxiety Disorders Interview Schedule for DSM-IV
). Patients met diagnostic criteria for generalized SAD (defined as greater than moderate anxiety/fear for 5 or more distinct social situations) and healthy controls had no history of DSM-IV
Participants completed self-report measures of clinical symptoms and individual differences. As shown in , compared with controls, patients reported greater social anxiety symptoms (Liebowitz Social Anxiety Scale-Self-Report (42
)), fear of negative evaluation (Brief Fear of Negative Evaluation Scale (44
)), depressive symptoms (Beck Depression Inventory-II (45
)), negative affect, and lesser positive affect (Positive and Negative Affect Schedule (46
)) and reappraisal frequency and self-efficacy (Emotion Regulation Questionnaire (47
Participants provided information about four distinct autobiographical social situations. At the scanning session, participants were trained in reappraisal methods developed by Gross and Ochsner(49
) with two experimenter-composed social anxiety situations and instructed to either “REACT” by considering how the NSB reflected something true about themselves, or “REFRAME” by reinterpreting the NSB to down-regulate negative emotional reactions. Participants were instructed to “actively reframe the belief by thinking in a way that re-interprets the content of the belief and thereby make the belief less negative and toxic for you. For example, if the belief is “NO ONE LIKES ME”, REFRAMING may be telling yourself “That is not always true,” “Some people like me”, or “This is only a thought, not a fact.” How else might you dispute this BELIEF?”
During scanning, participants read their autobiographical social situations one sentence at a time, and after each NSB, provided a negative emotion rating using a button box and E-Prime software (Psychology Software Tools, Inc, Pittsburgh, Pennsylvania) by responding to “How negative do you feel?” (1=not at all to 5=very much).
The task consisted of five situations. The first was an experimenter-composed neutral situation about cleaning a car that was used to obtain baseline emotion ratings and fMRI BOLD signals for reading neutral statements. Then four participant-generated autobiographical social anxiety situations characterized by social anxiety, humiliation, and embarrassment were presented. For each situation, participants composed a paragraph describing the events, thoughts, and feelings and five NSBs. Experimenters modified the NSBs so that there was a set of 5 self-only (e.g., I am incompetent) and 5 self-plus-other (e.g., Others think I am not normal) (Table S1 in Supplement 1
Participants indicated their age at the time of each situation and provided ratings, on a scale of 1 (not at all) to 9 (very much), quantifying the vividness of the memory, the experience of shame at the time of the situation, as well as current shame, disturbance, avoidance, and frequency of talking about the situation.
Three situations were presented in a first run lasting 9 minutes, 21 seconds, followed by two situations in a second run of 6 minutes, 24 seconds. The sequence of 5 situations was fixed: Neutral, React NSB, Reappraise NSB, React NSB, and Reappraise NSB.
Each situation consisted of (a) an instruction to react or reappraise (6s), (b) 16 sentences (3 seconds each) in white font against a black background describing the situation, (c) 10 NSBs (9 seconds each) embedded in the unfolding story in uppercase letters that flashed 9 times (850 milliseconds on + 150 milliseconds off), and (d) a negative emotion rating after each NSB (3 seconds) (). NSBs were flashed to maintain attention, and appeared in white font for Neutral and React trials and green for Reappraisal trials.
Components and structure of one autobiographical social situation trial.
Imaging was performed on a GE 3-T Signa magnet with a T2*-weighted gradient echo spiral-in/out pulse sequence (51
) and a custom-built quadrature “dome” elliptical bird-cage head-coil (GE Healthcare, Wisconsin). 630 functional volumes were obtained from 22 axial slices (repetition time=1500 milliseconds, echo time=30 milliseconds, flip angle=60°, field of view=22 cm, matrix=64×64, resolution=3.438 mm2
× 4.5 mm). High-resolution anatomical scans were acquired using fast spin-echo SPGR (.85942
× 1.5 mm; field of view=22 cm, frequency encoding=256).
fMRI Data Preprocessing
Using AFNI (52
) software, preprocessing included volume registration, motion correction, 4 mm3
isotropic gaussian spatial-smoothing, high-pass filtering (.011 Hz), and linear detrending. No volumes demonstrated motion in excess of ±0.6 mm. There was no evidence of stimulus-correlated motion between condition-specific reference functions and x, y, z motion parameters (all P
fMRI Statistical Analysis
Using 3dDeconvolve, multiple-regression included parameters to remove mean, linear and quadratic trends, and motion-related variance in the blood oxygen level-dependent (BOLD) signal. Regressors (convolved with the gamma variate model (53
) of the hemodynamic response function) were used to examine early (first 2 time points; 0–3s) and late (last 2 time points; 6–9s) BOLD responses for each of the 3 conditions (Neutral, React, Reappraise). Linear contrasts compared early versus late responses to test the hypothesis of linear decrease of emotional reactivity and increases of regulatory responses during the 9s trials. This method of investigating linear change in BOLD response over time within a trial has been used successfully in prior studies of emotion reactivity and reappraisal (27
) and cognitive appraisal (54
). BOLD signal intensity was computed as percentage of signal change, an effect size measure [(MR signal per voxel per time point / mean MR signal in that voxel for the entire functional run) × 100]. BOLD signal time series are relative to the neutral condition.
Individual brain maps were resampled to 3.438 mm3
and converted to Talairach atlas space (55
) and second-level group statistical maps were produced according to a random-effects model. To correct for multiple comparisons, AlphaSim, a Monte Carlo simulation bootstrapping program, was used to protect against false positives (56
). This method uses a voxel-wise and cluster volume joint-probability threshold to establish a cluster-wise false positive cluster detection level. Statistical thresholds consisted of a voxel-wise P
<.005 and cluster volume >162 mm3
(4 voxels × 3.438 mm3
) to protect against false-positive cluster detection at P
<.01 for between-group contrasts, and voxel-wise P
<.001 and cluster volume higher than 162 mm3
to protect against false-positive cluster detection at P
<.005 for within-group early versus late contrasts.
Functional connectivity (FC) analysis was seeded to a group-level left dorsal amygdala activation common to patients and controls during React NSBs. The group-level left amygdala functional region of interest was transformed from Talairach atlas space to the native brain space of each participant. Participant-specific left amygdala time series were used as a regressor of whole-brain BOLD response in a multiple-regression model that included parameters to remove variance related to mean, linear and quadratic trends, head movement, and whole-brain average signal intensity at each time point. Within-group t-tests examined FC patterns during Reappraise NSBs. The resultant t-maps were thresholded using a joint-probability method consisting of voxel-wise P<.001 and cluster volume>162 mm3 to protect against false positive cluster detection at cluster-wise P<.005.