This protocol was written using the “Standard Protocol Items: Recommendations for Interventional Trial” (SPIRIT) guidelines and the SPIRIT checklist is provided in additional file 2. The protocol is part of an ongoing, randomized (two condition) single center (Laureate Institute for Brain Research [LIBR], Tulsa, OK, USA) controlled trial examining multi-level predictors of response to EXP versus BA for GAD. The study is currently recruiting and is registered at the US National Institutes of Health (NIH) (ClinicalTrials.gov) #NCT02807480 (registration date: June 21, 2016). No amendments have been made to the protocol since original submission to ClinicalTrials.gov. The study is funded by the National Institute of Mental Health K23MH108707 (PI: RLA) and the William K. Warren Foundation. Interventions include 10 weeks of manualized, group-based BA or EXP therapy. Groups of participants (8-10) are randomized altogether to a therapy group (randomization conducted in blocks of four; sequence generated by RLA). Participants are kept blind to intervention condition until completion of all baseline assessments; outcome assessors are partially blinded (see further description in supplement). Primary predictor variables-of-interest are assessed using the approach-avoidance test (AAT) and the approach-avoidance conflict (AAC) task, while the primary outcome variable is the GAD-7. Secondary outcome measures include the Sheehan Disability Scale (SDS)(30), NIH Patient-Reported Outcomes Measurement Information System (PROMIS) Anxiety and Depression Scales(31), Beck Depression Inventory-II (BDI-II)(32), and Penn State Worry Questionnaire (PSWQ)(33).
The overall study protocol is represented in Figure 2. Screening assessments confirm exclusion and inclusion criteria for the study; baseline assessment includes self-report, behavioral, biological, and neuroimaging assessments. After baseline assessment, individuals are randomized to EXP or BA treatment, during which weekly self-report symptom measures are obtained. After treatment, participants repeat baseline assessments. Self-report symptom measures are repeated at three- and six- months following treatment. Research is conducted ethically in accordance with the World Medical Association Declaration of Helsinki. Research personnel trained in human subjects research obtain written informed consent from each participant prior to completing any research procedures. The consent form for the study is included in an additional file.
Participants
Projected enrollment is 100 treatment-seeking individuals meeting DSM-5 GAD(34) criteria over a 5-year period (April 2016-April 2021), recruiting from community mental health clinics and the general community through electronic and print advertisements. Participants must be 18-55 years old, have sufficient English proficiency to understand study procedures, and meet DSM-5 criteria per the Mini International Neuropsychiatric Interview (MINI 7.0) for GAD, and score >7 on the Overall Anxiety Severity and Impairment Scale (OASIS)(35). Participants are excluded for the following: (i) severe depressive symptoms (PHQ-9 score >17) and/or suicidal ideation with intent or plan, to decrease safety concerns and help ensure that GAD was the primary cause of impairment, (ii) history of substance use disorder in the past six months, (iii) meets diagnostic criteria for psychotic, bipolar, obsessive-compulsive, or eating disorders, (iv) moderate to severe traumatic brain injury or other neurocognitive disorder, (v) severe or unstable medical conditions, (vi) MRI contraindications, such as metal or metallic devices in the body, (vii) non-correctable vision or hearing problems, (viii) current use of psychotropic medications that could affect brain function (e.g., anxiolytics, antipsychotics, or mood stabilizers). Participants reporting current use of antidepressants (selective serotonin reuptake inhibitors [SSRIs]), are included as long as the dose has been stable for six weeks prior to enrollment. Inclusion/exclusion criteria are meant to decrease potential confounds while also supporting generalizability of results to GAD patient populations in the community.
Intervention
Both BA and EXP treatments consist of manualized, 10-session interventions, and are delivered in a group format for 90 minutes per week. For each, participants are provided a binder to accompany the intervention, including outlines of each session, basic descriptions of concepts, and “homework” worksheets. Brief descriptions of each intervention are provided below and in Table 1. Description of treatment compliance assessments and strategies are provided in an additional file (see additional file 1).
Behavioral Activation (BA): BA is a recognized efficacious treatment for MDD(36) and is based on the premise that negative or stressful life events can reduce one’s ability to experience reward or reinforcement (e.g., reduced social support). Depression develops and is maintained when individuals respond in ways that create additional deficits in reward or reinforcement (e.g., further isolation). The goal of BA is to identify alternative behaviors to increase in a way that increases opportunities for reward or reinforcement, particularly through naturally-reinforcing behaviors (e.g., those related to one’s values). A 10-session, structured group-based BA manual was developed by RLA and CM (with edits and revisions provided by AC), informed by previously-published BA treatment guides(29) and modified to focus on negative mood more generally, rather than solely focused on depression.
Exposure Therapy (EXP): EXP is a recognized efficacious strategy for the treatment of anxiety disorders. EXP is based on the premise that anxiety arises perceived threat associated with discrete cues or contexts, whether from direct or indirect/vicarious experience or informational transmission of perceived threat. Anxiety is thought to be maintained by avoidance behavior, preventing corrective learning. EXP guides individuals to decrease avoidance and experience anxiety-provoking situations or cues in a safe environment, allowing for inhibitory learning or habituation. The 10-session, structured, group-based EXP manual was based on a previous group-based anxiety treatment manual(37) developed by MGC, but modified further by MGC and RLA (with edits and revisions provided by KWT and AC), to focus on exposure strategies only (without cognitive restructuring), and inhibitory learning rather than habituation only(38)
Therapist Training and Treatment Fidelity: Each EXP and BA group is delivered by two co-therapists; a licensed doctoral or master’s level clinician with either another licensed clinician or a therapist-in-training (i.e., clinical psychology post-doctoral fellows or graduate students). Each therapist completes in-person or online workshops (e.g. Behavioral Tech, LLC, https://behavioraltech.org, Centre for Research on Eating Disorders at Oxford, https://credo-oxford.com), reads articles and manuals related to each treatment(29, 38), and watches videos of previous therapy sessions. Each therapy session is video and audio-recorded and at least 20% of sessions will be randomly selected for fidelity ratings. Skill acquisition and fidelity is assessed using the Quality of Behavioral Activation Scale (Q-BAS) for BA (Dimidjian, Hubley, Martell, Herman-Dunn, Dobson, 2012, unpublished measure) and a fidelity form created for the EXP treatment by RLA in consultation with MGC and KWT. Fidelity ratings will be provided by experts in each therapy (KWT; CM) or their trainees. Each therapist attends weekly consultation and supervision with the PI and/or consultants.
Data Collection:
All interview-based assessments (e.g., MINI) are administered by experienced, blinded examiners, trained to high levels of inter-rater reliability (kappa > 0.80). Self-report data are collected electronically using Research Electronic Data Capture (REDCap)(39). Study consent records are stored in a locked records room at LIBR. Study data records and blood/urine/biological samples are assigned code numbers and are not individually identifiable. REDCap servers are housed in a local data center at LIBR and all web-based information transmission is encrypted.
Measures
Self-report, behavioral, and neuroimaging measures included in the protocol are listed in Table 2 (refer to Figure 2 for timing of measures). Below are descriptions of the behavioral and neuroimaging tasks serving as primary predictors of interest. The remaining tasks are described in an additional file (see additional file 1).
Approach-Avoidance Task (AAT)(40): This task assesses behavioral avoidance tendencies. Participants are shown a picture of an emotional face (happy, angry, or neutral) framed by a blue or yellow border and instructed to pull a joystick (approach) when the border is one color and push away (avoid) when it is the other (counterbalanced). The picture zooms out and in accordingly. Mean response latency for push is subtracted from pull (e.g., angry pull - angry push) to obtain an avoidance bias score.
Approach Avoidance Conflict (AAC) Task(23, 24) (Figure 3): This task probes decision-making processes during approach-avoidance conflict. On each trial, the subject decides between two outcomes, represented on each side of a runway. A cloud indicates that a negative affective image/sound pair will be shown while a sun indicates that a positive image/sound pair will be shown (e.g., from International Affective Picture System [IAPS] and International Affective Digitized Sounds system [IADS](41, 42)). The amount of red in a rectangle indicates the amount of money awarded for each option (2,4,6 cents). For conflict trials, negative stimuli are paired with reward. Thus, the same behavior leads to both affective punishment and reward. For non-conflict ‘approach’ trials, both outcomes include positive affective stimuli, but only one offers reward. For ‘avoid’ trials, neither outcome offers reward, but one involves a negative affective image. For each trial, the subject moves the avatar, knowing the probability of each outcome (10-90%) depends on their end position. Behavioral variables include approach behavior (end avatar position) and response time (RT) for initial button press.
During behavioral tasks, the BioPac MP150 system and AcqKnowledge software (BIOPAC instrumentation; Lehigh, Pennsylvania) are used to collect galvanic skin conductance, heart rate (electrocardiogram), and respiration rate (respiration transducer). MRI data are collected on a GE MR750 3T MRI scanner. Analysis of Functional NeuroImages (AFNI)(43) is used for processing of MRI data. Electroencephalography (EEG) is simultaneously recorded during MRI scans using a 31-electrode cap attached to an MRI-compatible BrainAmp MR Plus amplifier. Blood samples for plasma, serum, and peripheral blood mononuclear cells (PBMCs) are collected at baseline and post-treatment to quantify biomarkers for future exploratory analyses. See additional file 1 for further description of MRI, EEG, and blood biomarker methods.
Analysis Procedures
The characteristics of all measures will be examined for missing data and deviation from normality prior to subsequent analyses. Baseline demographic characteristics and attrition data will be contrasted between treatment groups and analyses adjusted to account for potential confounders. For the first aim, we will test the hypotheses that approach and conflict arbitration behavior and neural responses will explain significant variance in baseline symptoms above and beyond avoidance-related behavior and neural responses. Approach and avoidance behavior are defined by AAT bias scores; conflict arbitration is defined by RT during AAC conflict trials. For brain responses, we will focus on the AAC task and extracted percent signal change (PSC) from a priori ROIs: (i) Approach: left caudate (reward versus no reward outcome); (ii) Avoidance: right amygdala (negative versus positive affective outcome), and (iii) Conflict: right dlPFC (conflict versus non-conflict decisions). We will use Huber robust regression with baseline GAD-7 as the dependent variable (DV), and approach, avoidance, and conflict measures as independent variables (IV).
For the second aim, we will test the hypotheses that approach-related and conflict arbitration behavior and neural responses will predict treatment response above and beyond avoidance-related behavior and neural responses. We will use linear mixed models (LMEs) with random subject-level intercepts and slopes, GAD-7 scores across the 10 sessions as DVs; baseline GAD-7 as covariate; and approach, avoidance, and conflict measures as IVs. The main effect of intervention type and its interaction with IVs will be included to determine treatment main effects and moderating effects. We will determine the best set of IVs using the Lasso method(44) and utilize functional linear models(45) to model on-parametric symptom trajectories as needed.
For the third aim, we will test the hypothesis that the degree to which conflict arbitration abilities increase with treatment will positively relate to functional improvement from pre- to post-treatment. We will use LMEs to test main and interaction effects between intervention type and change in AAC conflict arbitration in predicting trajectories of GAD-7 scores over the 10 sessions. We will employ the asymmetric distribution of product of coefficients test (versus Baron and Kenny methods) due to the greater power and more appropriate Type 1 error rate it affords(46).
In addition, we are collecting data from other measures for exploratory analyses. For such analyses, we will explore (i) data reduction methods to derive multi-level factors associated with approach, avoidance, and conflict arbitration processes, and (ii) use of random forest machine learning, which is particularly appropriate with a large ratio of predictors to participants(47), to identify predictors of treatment outcome.
Sample size and Power Analysis
Previous research suggests large effects for fMRI predictors (i.e., r = .60-.75) and medium-large effects for behavioral predictors (r = .30-.47) of intervention outcomes(48, 49). For this study, we aim to recruit 100 participants, which with 20% attrition would allow for complete longitudinal data for 80 participants (i.e., ~40/intervention). LMEs will include all participants with any post-baseline assessments. Thus, we anticipate approximately 50 participants per intervention for Aims 2/3. For relationships between individual predictors and DVs, we estimate having 80% power to detect medium to large effects (r = .27 for N = 100; r = .37 for N = 50; α = .05). In a model with three predictors (approach, avoidance, conflict arbitration), we also estimate having 80% power to detect medium to large effects (η2 = .11 for N = 100; η2 = .24 for N = 50).
Design Considerations
We considered an alternative design where we examined predictors of EXP response compared to an attentional control intervention. We instead decided to identify unique predictors of two theoretically-divergent behavioral therapies because (i) the current protocol was not meant to test intervention efficacy as compared to “placebo” (as efficacy has been established in previous research), (ii) variability in outcomes would be greatest for efficacious, as opposed to placebo, interventions, thus enhancing statistical power for prediction, and (iii) identifying unique predictors for two therapies would be most clinically meaningful. We also considered using more unified cognitive-behavioral interventions(50). However, interventions that simultaneously target multiple processes (e.g., both cognitive and behavioral strategies) would have made it more difficult to identify predictors relating to the specific therapeutic target. The goal here is to foster individualized, precision targeting of treatments, rather than applying more broadly targeted treatments to all patients.
Ethics
Study approval was obtained from Western Institutional Review Board (WIRB) protocol #20151232. Any protocol modifications will be made to records on ClinicalTrials.gov and communicated to study investigators, WIRB, and funding organizations as required.
Gender/minority/pediatric inclusion for research: Planned study enrollment is reflective of Tulsa county population demographics and is described further in an additional file (see additional file 1). We will not exclude subjects based on sex, gender, race, or ethnicity. Children are not included due to our initial focus on understanding neurobehavioral predictors of exposure therapy for adults suffering from GAD. The variability introduced by developmental changes could reduce sensitivity to detect hypothesized relationships.
Safety: The risk for adverse events is minimal. The study is a clinical trial, but is not a Phase III trial, involves only one site, is not blinded, and does not employ high risk interventions or vulnerable populations. The interventions employed are known to be efficacious for the treatment of anxiety or depression. The PI (RLA), in collaboration with mentor and LIBR scientific director (MPP), are performing the monitoring function for the study. Any unanticipated adverse events will be reported immediately to LIBR Human Protection Administrator and to the Western IRB. Any adverse events will be included in the annual IRB report.
Each week prior to session, participants complete questionnaires assessing symptom severity and suicidal ideation. As needed, a therapist meets with the participant individually to assess risk and provide referrals or identify emergency services. LIBR is situated on campus with an inpatient psychiatric hospital (Laureate Psychiatric Clinic and Hospital), which has a 24-hour clinical assessment department.
Dissemination: Results will be shared with the scientific and health professional community through presentation at national and international scientific meetings and publication in scientific journals. The full protocol and statistical code used for data analysis will be provided with resulting publications. To disseminate results to the general public, all final peer reviewed manuscripts will be submitted to the digital archive PubMed Central in compliance with the NIH Public Access Policy. The PI will maintain a local website where lay summaries of scientific results will be provided, with appropriate links to scientific presentations and publications. The PI will ensure the summary of results information is submitted to ClinicalTrials.gov.