Contemporary Clinical Trials 44 (2015) 70–76

Contents lists available at ScienceDirect

Contemporary Clinical Trials

journal homepage: www.elsevier.com/locate/conclintrial

Yoga for generalized anxiety disorder: design of a randomized controlled

clinical trial
Stefan G. Hofmann a,⁎, Joshua Curtiss a, Sat Bir S. Khalsa b, Elizabeth Hoge c, David Rosenfield d, Eric Bui c,
Aparna Keshaviah c, Naomi Simon c
a
Boston University, United States
b
Brigham and Women's Hospital, Harvard Medical School, United States
c
Massachusetts General Hospital, United States
d
Southern Methodist University, United States

a r t i c l e i n f o a b s t r a c t

Article history: Generalized anxiety disorder (GAD) is a common disorder associated with significant distress and interference.
Received 20 May 2015 Although cognitive behavioral therapy (CBT) has been shown to be the most effective form of psychotherapy,
Received in revised form 31 July 2015 few patients receive or have access to this intervention. Yoga therapy offers another promising, yet under-
Accepted 3 August 2015
researched, intervention that is gaining increasing popularity in the general public, as an anxiety reduction inter-
Available online 6 August 2015
vention. The purpose of this innovative clinical trial protocol is to investigate the efficacy of a Kundalini Yoga in-
Keywords:
tervention, relative to CBT and a control condition. Kundalini yoga and CBT are compared with each other in a
Clinical trial noninferiority test and both treatments are compared to stress education training, an attention control interven-
Generalized anxiety disorder tion, in superiority tests. The sample will consist of 230 individuals with a primary DSM-5 diagnosis of GAD. This
Kundalini yoga randomized controlled trial will compare yoga (N = 95) to both CBT for GAD (N = 95) and stress education
Alternative medicine (N = 40), a commonly used control condition. All three treatments will be administered by two instructors in
Cognitive behavioral therapy a group format over 12 weekly sessions with four to six patients per group. Groups will be randomized using per-
Mindfulness muted block randomization, which will be stratified by site. Treatment outcome will be evaluated bi-weekly and
at 6 month follow-up. Furthermore, potential mediators of treatment outcome will be investigated. Given the in-
dividual and economic burden associated with GAD, identifying accessible alternative behavioral treatments will
have substantive public health implications.
© 2015 Elsevier Inc. All rights reserved.

1. Introduction accessible is yoga [10,11]. Yoga, practiced in its traditional contempla-

tive practice format, is a multicomponent behavioral practice incorpo-
Generalized anxiety disorder (GAD) has been determined to be one rating physical postures and exercises, breath regulation practices,
of the most ubiquitous anxiety disorders, with one-year prevalence deep relaxation techniques, and meditation/mindfulness. In fact, a re-
rates ranging from 3% to 8% [1,2]. In light of its prevalence, significantly cent survey revealed that 8.9% of the population had used yoga as a ther-
deleterious impact on life functioning [3,4], and high comorbidity [5–7], apeutic intervention and that yoga and its component practices of deep
the identification and dissemination of efficacious treatments for GAD is breathing and meditation were among the ten most prevalent alterna-
imperative. Traditional cognitive behavior therapy (CBT) has received tive practices, as evidenced by its rapid adoption in the previous five
the most empirical support as an efficacious intervention for GAD [8]. years [11,12].
Although CBT is a gold-standard treatment for anxiety disorders, it re- Converging evidence underscores the presence of mindfulness defi-
mains inaccessible to many patients. Several barriers deter individuals cits among individuals with GAD and other emotional disorders [13,14].
from undergoing empirically supported psychotherapy, including pro- Though several mindfulness based interventions currently exist
hibitively exorbitant costs, the predominance of a face-to-face model (e.g., mindfulness based stress reduction (MBSR), mindfulness based
of delivery, the paucity of available resources to address increasing cognitive behavioral therapy (MBCT), etc.) [15], difficulties in the dis-
mental health needs, etc. [9]. Current circumstances warrant the inves- semination of such treatments undermine their potential to reduce
tigation of alternative treatment strategies to reduce this burden. A the overall burden of GAD. As yoga comprises several empirically
promising intervention that has become more widely adopted and supported components (e.g., meditation, breath regulation, physical
exercise, and relaxation techniques) [16–18], it would be profitable to
determine the efficacy of yoga in its full form. A thorough investigation
⁎ Corresponding author. of MBSR revealed that yoga exercises accounted for the largest

http://dx.doi.org/10.1016/j.cct.2015.08.003
1551-7144/© 2015 Elsevier Inc. All rights reserved.
 S.G. Hofmann et al. / Contemporary Clinical Trials 44 (2015) 70–76 71

contribution to treatment efficacy, despite being applied less than other disorders, with the exception of concurrent post-traumatic stress disor-
MBSR components [19]. der, substance use disorders, eating disorders, significant suicidal idea-
The therapeutic benefit of CBT and yoga is likely a consequence tion, or an organic mental disorder. These exclusion criteria extend to
of differential treatment mechanisms. Extant literature attests to mal- individuals with a lifetime history of psychosis, bipolar disorder, or a de-
adaptive cognitions as mediators of treatment change in CBT for anxiety velopmental disorder.
disorders [20]. In virtue of the similarity between yoga and mindfulness To minimize confounds, participants are required to have limited
based interventions, yoga likely conveys its effect by way of changes in prior experience with CBT and yoga (i.e., no more than five yoga classes
mindfulness [21]. Basic research on yoga and its component techniques or CBT sessions within the last five years). Though concomitant and
suggests it might facilitate reductions in arousal, targeting different prior pharmacotherapy is permitted, participants must undergo appro-
mechanisms than does CBT [22–25]. Although there is evidence priate stabilization periods prior to treatment. Please refer to Table 1 for
supporting the efficacy of yoga in reducing anxiety in general [26,27], the complete catalogue of inclusion and exclusion criteria.
there are few trials of yoga for GAD, none of which are randomized con-
trolled trials. Of the currently published uncontrolled trials reporting ef- 2.2. Recruitment and screening procedures
ficacy of yoga for GAD, two employed a yoga breathing intervention [28,
29], one employed a yoga mantra practice [30], and one applied an in- Participant recruitment will occur at the Center for Anxiety and Re-
tervention combining CBT and Kundalini yoga practices [31]. The Kun- lated Disorders at Boston University (CARD) and the Center for Anxiety
dalini yoga study showed statistically significant improvements in and Traumatic Stress Disorders (CATSD) at Massachusetts General
state and trait anxiety, depression, panic, sleep and quality of life with Hospital (MGH). Recruitment strategies will include advertisements in
strong effect sizes for state and trait anxiety [31,32]. In light of the pau- the media, postings online, particular email services, and clinical refer-
city of high quality clinical trials of yoga for anxiety disorders, method- rals from patients presenting at the abovementioned clinics. At CARD
ologically rigorous research undertakings are warranted to extend our and MGH, we will post information on our respective clinic websites
current knowledge about the empirical status of yoga for GAD. and distribute study information through the MGH All-User Broadcast
This paper describes the procedures and methodology of a five-year, and the RSVP for Health program.
multi-site randomized controlled trial investigating the comparative ef- Individuals who contact study personnel in either site will undergo
ficacy of yoga, CBT, and a stress education (SE) intervention, a frequent- an initial phone screen to determine potential eligibility and interest
ly employed control condition for patients with GAD. The interventions in research participation. Data for participants who undergo the phone
will be administered over the course of twelve weeks. By employing a screening process will be de-identified and recorded in the study data-
non-inferiority test of yoga for GAD, the results of this study will inform base phone screening log. Furthermore, we will record reasons for inel-
current literature regarding its utility and efficacy. igibility and for non-participation of eligible subjects. Interested
This study is funded as a multi-site linked R01 grant R01AT007257 patients who satisfy eligibility criteria will be scheduled for an evalua-
(Hofmann) and R01AT007258 (Simon) by the National Institute of tion meeting where informed consent will be obtained.
Health/National Center for Complementary and Integrative Health
(NCCIH). ClinicalTrials.gov identifier: NCT01912287 2.3. Study design

2. Materials and methods The purpose of the proposed randomized controlled trial is to exam-
ine the efficacy of yoga for GAD relative to standard psychosocial inter-
2.1. Participants ventions. Using superiority tests, yoga (N = 95) will be compared to SE
(N = 40). Using non-inferiority tests, yoga will compared to standard
To satisfy eligibility requirements for the current study, participants CBT for GAD (N = 95). All three treatments will be administered by
must be English fluent outpatients that are 18 years of age or older with two therapists/yoga instructors/SE instructors in a group format over
a primary psychiatric diagnosis of GAD according to the criteria of the 12 weekly sessions with 4–6 patients per group. For all interventions,
DSM-5. Comorbidity is permitted with other anxiety and depressive a credible rationale for the efficacy of their assigned treatments,

Table 1
Inclusion and exclusion criteria and rationale.

Rationale

Inclusion criteria
Male or female outpatients 18 years of age or older with a primary psychiatric diagnosis of generalized anxiety disorder Population under study
CGI-severity score of 4 or higher Adequate pre-treatment severity
Off concurrent psychotropic medication for at least 2 weeks prior to initiation of randomized treatment, OR stable on Treatment confound
current medication for a minimum of 6 weeks and willing to maintain a stable dose
Willingness and ability to perform the yoga intervention and to comply with the requirements of the study protocol. Human subjects concerns
For women of childbearing potential, willingness to use a reliable form of birth control Human subjects safety

Exclusion criteria
Patients unable to understand study procedures and participate in the informed consent process. Human subjects concern
Pregnancy as assessed by a urine pregnancy test at screen for women of childbearing age Human Subjects Safety
Women who are planning to become pregnant. Human Subjects Safety
Serious medical illness or instability for which hospitalization may be likely within the next year Feasibility, subject safety
Significant current suicidal ideation or suicidal behaviors within the past 6 months (assessed with the BDI-II) Subject safety
History of head trauma causing loss of consciousness, or seizure disorder resulting in ongoing cognitive impairment Treatment confound, subject safety
Posttraumatic stress disorder, substance use disorder, eating disorder, or organic mental disorder within the past 6 months Treatment confound
Lifetime history of psychotic disorder, bipolar disorder, or developmental disorder Treatment confound
Significant personality dysfunction likely to interfere with study participation (assessed during the clinical interview) Treatment confound
Prior experience with (more than 5 Yoga classes or CBT sessions within the last 3 years) and/or current practice of Treatment confound
mind–body techniques (e.g., yoga, meditation, Tai-Chi, etc) or CBT
Concomitant psychotherapy for GAD (any psychotherapy) Treatment confound
Physical conditions that might cause injury from yoga (pregnancy, physical injuries and musculoskeletal problems) Human subjects concern and subject safety
Cognitive impairment (MOCA ≤ 21) Human subjects concern and safety
72 S.G. Hofmann et al. / Contemporary Clinical Trials 44 (2015) 70–76

including SE, is provided, with all studies presented equally throughout 2.4.1. Primary outcome measures
the study. Various psychological measures will be completed by the par- Clinical Global Impression of Severity and Improvement (CGI–S,
ticipants at each study visit. The four independent evaluator blinded CGI–I; [33]): This two part clinician administered instrument assesses
assessments occur pre-treatment (screening visit), mid-treatment overall level of illness severity and level of symptom change across the
(Week 6), post-treatment (Week 12), and at a six-month follow-up. course of treatment. The CGI–I ranges from 1 to 7, with scores less
Follow-up visits are scheduled at the termination of treatment to pre- than 4 reflecting symptom improvement and scores greater than 4 indi-
vent attrition, and further contact is maintained to ensure patients' cating symptom worsening. A score of 1 (very much improved) or 2
availability for the follow-up visit. (much improved) defines treatment response.

2.4. Assessment instruments and procedures 2.4.2. Secondary outcome measures

State Trait Anxiety Inventory (STAI; [34]): The STAI is a 40-item,
All clinician-rated assessments will be undertaken by independent multiple-choice questionnaire that differentiates between the tempo-
evaluators (IEs) blind to treatment assignment. The IEs will be M.D., rary condition of “state anxiety” and the more general and long-
Ph.D., or experienced Masters level diagnosticians experienced in the standing quality of “trait anxiety.”
administration of structured clinical interviews who will receive further Beck Anxiety Inventory (BAI; [35]): This 21-item self-report inven-
training and certification for this study under the direction of an IE train- tory designed to measure severity of anxiety symptoms in psychiatric
er. Monthly inter-rater reliability checks for the assessments will be un- populations has high internal consistency and test–retest reliability.
dertaken by IEs who did not conduct the initial interview. The ratings Beck Depression Inventory—Version II (BDI-II; [36]): The BDI-II con-
will be used to calculate kappa coefficients and to facilitate supervision, tains 21 items that assess severity of depression symptoms in psychiat-
during which supervisors will discuss potential disagreement and ric populations has high internal consistency and test–retest reliability.
provide instruction to enhance inter-rater reliability. Please refer to Perceived Stress Scale (PSS; [37]): This 10-item scale is the most
Table 2 for a schedule of assessment instruments. widely used psychological instrument for measuring the perception of
As depicted in Table 2, the two Clinician Global Impressions Scales stress. It is a measure of the degree to which situations in one's life are
measuring symptom severity and symptom improvement (i.e., CGI–S appraised as stressful.
and CGI–I, respectively) will be used for the primary analyses. More spe- Quality of Life Scale (WHOQOL-BREF; [38]): This instrument as-
cifically, the primary outcome measure is the binary responder status sesses quality of life and overall satisfaction.
value determined from the CGI–I, with a score of 1 (very much im- Symptom Checklist Version 90 (SCL-90; [39]): The SCL-90 contains
proved) or 2 (much improved) indicating responder status. The self- 90 items that assess to what degree certain problems and complaints
report measures of state anxiety, depression, worry, and quality of life bothered or distressed an individual during the past week.
constitute the secondary outcome measures. For the mediator analyses, Penn State Worry Questionnaire (PSWQ; [40]): This instrument con-
we will be investigating whether self-report measures of mindfulness tains 15 items that measure trait levels of worry. The PSWQ has excel-
and metacognition andrespiratory sinus arrhythmia, mediate treatment lent psychometric properties in student, community, and clinical
outcome. samples.

Table 2
Schedule of evaluations.

Form type Measure Admin Mo. −1 Wk 0 Wks 1–5 Wk 6 Wks Wk 12–13 Wk 36–37 Reference
by (screen*) (baseline) (mid Tx) 7–11 (post Tx) (6 mo. FU)

Diagnosis & screening Informed consent IE X

Urine pregnancy test RA X
SCID/ADIS-5 IE X [63,64]
SIGH-A IE X X X X X [65]
Demographics Self X
Medical review (history and exam) MD/NP X
MoCA MD/NP X [66]
Primary outcome CGI–S/CGI–I IE X X X X X X X [33]
Secondary outcomes STAI Self X X X X [34]
BAI Self X X X X [35]
BDI-II Self X X X X [36]
PSS Self X X X X [37]
QOL Self X X X X [38]
SCL-90 Self X X X X [39]
PSWQ Self X X X X [40]
ISI Self X X [42]
PSQI Self X X [41]
ASQ Self X X X X [43]
Mediators MCQ Self X X X X [44]
FFMQ Self X X X X [13,45]
Psycho-physiological & ECG (RSA) RA X X X X
Biological stress Cortisol samples X X X X
Tx details Credibility/expectancy X X X X
Homework compliance X (weekly) X X (weekly) X X
Safety Adverse events IE X (weekly) X X (weekly) X X
Concomitant medications IE X (weekly) X X (weekly) X X
Working alliance X (weekly) X X (weekly) X X

Note: IE = independent evaluator; self = self-report; MD/NP = doctor of medicine/nurse practitioner; RA = research assistant; SCID = Structured Clinical Interview for DSM-IV TR Axis I
Disorders; ADIS-5 = Anxiety Disorders Interview Schedule-5; CGI–S/CGI–I = Clinical Global Impressions Scale-Severity/Improvement; SIGH-A = Structured Interview Guide for
the Hamilton Anxiety Rating Scale; MoCA = Montreal Cognitive Assessment; STAI = State Trait Anxiety Inventory; BAI = Beck Anxiety Inventory; BDI-II = Beck Depression Inventory;
PSS = perceived stress scale; QOL = The World Health Organization's quality of life assessment; SCL-90 = Symptom Checklist Version 90 Revised; PSWQ = Penn State Worry Question-
naire; MCQ = Meta-Cognitive Questionnaire; FFMQ = Five Facet Mindfulness Questionnaire; ECG = electrocardiogram.
 S.G. Hofmann et al. / Contemporary Clinical Trials 44 (2015) 70–76 73

Pittsburg Sleep Quality Index (PSQI; [41]): This is a 19-item measure interventions. Kundalini yoga is a well-known, accessible yoga practice
of sleep quality over a one month duration. The items make up seven that includes physical postures and exercises, breathing techniques, re-
different component scores consisting of sleep quality, sleep latency, laxation exercises, and meditation practices. At study entry, subjects'
sleep duration, habitual sleep efficiency, sleep disturbances, use of current physical health, flexibility and endurance will be assessed in
sleep medication and daytime dysfunction. This measure also generates the event that adjustments and modifications of the yoga practices
a global sleep quality score. need to be made to accommodate subject limitations. This yoga inter-
Insomnia Severity Index (ISI; [42]): This 7-item instrument assesses vention will involve 12 weekly classes administered in group format
symptoms of insomnia. led by qualified and certified Kundalini Yoga instructors. The weekly in-
Affective Style Questionnaire (ASQ; [43]): The ASQ contains 20 items tervention itself entails yoga theory, philosophy and psychology, physi-
that measures individual differences in emotion regulation. This cal postures/exercises, breathing techniques, meditation, and deep
instrument demonstrates good psychometric properties and comprises relaxation practice, with the weekly group practices varying in content
three subscales: concealing (i.e., habitual attempts to conceal affect), but increasing in intensity throughout the 12-week standardized proto-
adjusting (i.e., adjusting emotions to be congruent with external col. Practices will include individual breathing practices (Sitali Breath,
goals), and tolerating (i.e., acceptance of emotions). Alternate Nostril Breathing, Pauri Kriya), full hour-long Kundalini yoga
sets/series (Basic Spinal Energy Series, Calmness and Anti-anxiety Se-
2.4.3. Mediators ries, Stress Set for Adrenals and Kidneys) and individual yoga medita-
Meta-Cognition Questionnaire (MCQ; [44]): This is a 65-item instru- tions with coordinated breathing (Segmented Breath for Anxiety,
ment used to assess meta-cognitive beliefs people have about their Tattva Balance Beyond Stress and Duality, Shabd Kriya) (accessible at
thinking styles. the website: The Yogi Bhajan Library of Teachings, http://www.
Five Facet Mindfulness Questionnaire (FFMQ; [13,45]): The FFMQ libraryofteachings.com/) Daily home practice assigned and instructed
contains 39 items that assess five components of mindfulness, including during the first treatment session will be facilitated with audio CDs to
observing, describing, acting with awareness, non-judging, and non- guide subjects throughout the 12-week intervention. The same home
reacting. This scale has evidenced good psychometric properties and practices will be used throughout in order for subjects to achieve mas-
has been validated in clinical and non-clinical samples. tery and self-efficacy. The 20-minute home practice session will begin
with physical exercises for spinal flexibility and release of tension
2.5. Treatment interventions with coordinated breathing and maintenance of mind-body awareness,
as follows: 1) seated spine flexes forward and back (1 min), 2) seated
All treatments were matched for time and attention; each treatment spinal twists (2 min), 3) shoulder shrugs (2 min), 4) neck rolls in both
entails 120 min of training and practice in group sessions let by two in- directions (2 min total), and 5) cat–cow spinal flexes on the hands
structors. Participants will receive a credible rationale for the efficacy of and knees (2 min). These exercises are followed by the “Segmented
their assigned treatment. Participants in the CBT arm will be informed of Breath for Anxiety” involving slow, rhythmic, segmented inhales in 4
the role of CBT in the efficacy of changing dysfunctional thought pat- parts followed by segmented exhales in 8 parts maintaining a 1:2 ratio
terns and beliefs, those in the yoga arm will be informed of the role of between the duration of the inhales and the duration of the exhales
yoga in changing cognitive and physical emotional and stress reactivity for 11 min.
through the postures, breath regulation and meditation components of
yoga practice, and those in the SE control will be informed of the role of 2.5.3. Stress education training
stress education in understanding the psychophysiological basis of This intervention also entails 12 weeks of group and home practice
stress and anxiety as a basis for overcoming stress and anxiety. All sessions and will function as a control for attention from instructors,
three interventions are standardized protocols with formalized treat- expectancy effects, and group support effects. In this condition,
ment manuals for consistency. participants will receive extensive information about stress and health,
but will not receive any of the active therapeutic ingredients used in
2.5.1. Cognitive behavioral therapy CBT, yoga, or other mind–body training techniques. Instead, psycho-
The 12 session CBT treatment will occur over the course of 12 ses- educational information will be provided, such as definitions of stress
sions and will be derived from the standardized protocol developed at and the stress response, the fight or flight response, physiological and
one of our centers (CARD) [46]. This intervention has received empirical psychological effects of stress, stress and performance, the negative
support in both individual and group settings [47]. This protocol stress cycle, stress and health/illness, stress and immunity, stress
comprises four primary treatment modules including cognitive buffers, stress hardiness, stress and heart disease, the role of genes
restructuring, progressive muscle relaxation, worry exposures, and and environment in health, the contribution of lifestyle behaviors such
in vivo exposure exercises. The initial sessions provide psychoeducation, as caffeine and alcohol intake and cigarette smoking, and the impor-
elucidating the cognitive behavioral model of worry and GAD. Each ses- tance of regular exercise and proper diet (without specific instructions
sion consists of a different “lesson”, which initially teaches basic infor- for exercise or dietary changes). For the daily 20 min home practice,
mation about the nature and possible function of the anxiety and subjects will listen to CDs that provide information about nutrition
worry, the negative consequences of sustained worrying, the maladap- and positive perspectives on lifestyle and emotions.
tive effects of attempting to suppress one's thoughts, the basic cognitive
errors of probability overestimation and catastrophic thinking, and 2.6. Safety protocol
adaptive strategies to successfully deal with worries. In addition, our
treatment protocol will target meta-cognitions (i.e., worrying about In the service of optimizing safety for participants in the yoga and
worrying) [48]. Subjects will also be requested to complete approxi- CBT interventions, we exclude all potential subjects with physical dis-
mately 20 min of homework daily, which will be recorded in a home- abilities or cognitive disabilities severe enough to compromise safety
work log. or ability to participate in either of the interventions. For reasons of
safety, all women of childbearing potential will be required to use a re-
2.5.2. Yoga liable form of birth control throughout the study, as there is insufficient
The yoga intervention will employ Kundalini Yoga as taught by Yogi information regarding the safety of yoga during pregnancy. Adverse
Bhajan. Dr. Sat Bir Khalsa, one of the co-investigators who will train and events will be assessed at each visit by inquiring whether any major
supervise the instructors providing the yoga intervention, is a certified change in mental or physical health occurred since the participant's pre-
Kundalini Yoga instructor and has conducted clinical trials of yoga vious visit. Furthermore, the clinical global impression of severity scale
74 S.G. Hofmann et al. / Contemporary Clinical Trials 44 (2015) 70–76

(CGI–S), clinical global impression of improvement (CGI–I), Beck De- The mediation analysis includes 2 independent variables (i.e., the
pression Inventory (BDI-II), and a suicide risk assessment will be admin- contrast of yoga vs. SE and the contrast of CBT vs. SE) and 3 mediators:
istered to evaluate any potential worsening symptoms biweekly from mindfulness, vagal tone, and maladaptive cognitions. The 3 mediators
baseline to termination of treatment. will be included in the growth curve model as additional simultaneous
Any participant with a CGI–I of 5 or greater at any time will undergo predictors of response. Significance of mediated pathways will be deter-
weekly assessment by a study clinician including clinical and safety as- mined using bias-corrected bootstrap mediation analysis [51]. We pre-
sessments, CGI–S and CGI–I ratings, and suicidality assessment. To en- dict that changes in mindfulness will mediate changes in outcome
sure participants are able to benefit from some form of treatment, we over time for patients in the yoga condition, and changes in cognitions
will discontinue participants who endorse ‘much worse’ (i.e., N 5) on will mediate changes in CBT. Plus, changes in RSA and cortisol may
the CGI–I for 2 consecutive weeks or who develop significant suicidality also mediate changes in outcome across conditions.
at any point. Appropriate referrals will be made to ensure they receive To address missing data, we will use pattern mixture modeling
care from certified mental health providers. In addition, if the BDI-II [52,53] and re-conduct the analyses to identify potential missing data
suicidality item (i.e., item number 9) exceeds 1 or the CSS suicide check- patterns (e.g., no missing data, early dropouts, late dropouts, FU drop-
list exceeds 2, the patient will be assessed weekly to monitor suicidality outs, etc.) and determine whether missingness influences our findings.
and receive any indicated clinical intervention.
2.9. Sample size
2.7. Fidelity monitoring
The sample will consist of 230 patients with a primary DSM-5 diag-
All of the group intervention sessions will be audio recorded. We will nosis of GAD. We will compare yoga (N = 95) to both CBT for GAD
rate 20% of the recordings for adherence and fidelity to ensure that indi- (N = 95) and stress education (SE) (N = 40), a commonly used atten-
viduals receive the form of treatment appropriate to their assigned tion control condition [54]. To directly appraise our hypothesis that
group. Failure to satisfy minimum standards (i.e., receives an adherence yoga will have comparable short-term and long-term efficacy with
or competence rating below the certification standard for 2 consecutive that of CBT, we will adopt a non-inferiority test framework for the pri-
sessions (≤5 for adherence, ≤3 for competence)), will be redressed with mary outcome, which is the binary responder status value determined
supplemental training provided by the supervising clinician. Further- from the CGI–I, with a score equal to or less than 2 indicating responder
more, the next two sessions will be reviewed to appraise whether status. To address potential attrition, we will conduct an intent-to-treat
they meet certification standards. In the unlikely event therapists do analysis. The non-inferiority margin will be half the difference in re-
not meet these standards, they will be replaced. sponse rates between the active comparison (CBT) and control condi-
tion. In accord with prior literature, we estimated response rates for
2.8. Data analytic strategy our particular interventions (i.e. 57.9% in CBT and 22.2% in control).
Thus, our non-inferiority margin will be set at half the difference be-
Randomization will be stratified by site. Treatment groups will be tween 57.9 and 22.2, or 17.85%. Using the power analysis program
assigned using a permuted block randomization for each site [49]. One PASS 12 [55], we determined that the number of subjects required to
block will be generated for each site. Conservatively assuming an aver- obtain .80 power for the non-inferiority test comparing CBT and yoga
age of 4 participants per treatment group, the block will consist of 12 is 95 participants in each of the 2 conditions.
Yoga groups, 12 CBT groups, and 5 SE groups to achieve the appropriate For the superiority tests on responder status (a greater proportion of
allocation ratio of participants to treatment conditions. We will use the participants will respond to treatment in CBT and in Yoga than in SE),
Sealed Envelope program to generate the randomization. After all par- we will use GLMM with a logistic linking function. We performed vari-
ticipants in a potential treatment group have completed their baseline ous Monte Carlo simulations to determine the smallest percentage dif-
assessment, the study coordinator will inform the study statistician ference between the treatments and the control group that we could
that the next group is ready for randomization, and the statistician detect with .80 power. Assuming a sample size of 230 and 25% missing
will then inform the treatment coordinator of the randomization for data, we found that we could detect a difference in response rate as little
that group. as 22% between the treatments and the control (e.g., if the control re-
Our primary analyses will use mixed-effects regression models sponse rate was 22.2%, we have .80 power to detect a significant differ-
(MRMs) with a logistic linking function, a general linear mixed model ence between either treatment condition and the control condition if
(GLMM) analysis, as our primary outcome measure (response to treat- the treatment condition response rate is 44.2% or higher). This differ-
ment, defined as CGI–I ≤ 2) is dichotomous. Analyses of secondary ence is equivalent to an effect size of ω = .24, smaller than a medium
outcomes will be performed using MRM. MRM and GLMM easily ac- effect size (a medium effect size for proportions is ω = .30, and a
commodate nesting of repeated observations within subjects, include small effect size is ω = .10).
all subjects with at least one assessment, and are the preferred method For the superiority tests (CBT vs. SE and Yoga vs. SE) on the second-
to analyze longitudinal data [50]. Since these analyses will include all ary outcomes (which are the continuous self-report measures present-
subjects who are randomized and complete at least one assessment, it ed in Table 2), we used the mixed-effects regression models (MRM)
is an intent-to-treat analysis. Also, since subjects will undergo treatment power analysis program PinT 2.12 [56]. We assumed 25% missing
in groups, our MRM and GLMM models will include 3 levels: repeated data, and used data from our recent CBT trials [57] to estimate the var-
measurements (level 1) nested within subjects (level 2) who will be iances and covariances needed for PinT. PinT indicated greater than .80
nested within their treatment cohort (level 3). power to detect an effect size as small as d = 0.27, between a small
The three treatment conditions (yoga, CBT, and SE) will be coded (d = 0.20) and medium (d = 0.50) effect size.
using 2 dummy variable contrasts, with SE as the “reference” treatment.
The first dummy variable will contrast yoga vs. SE, and the second will 3. Discussion
contrast CBT to SE. Time will be centered at week 12 and coded as
weeks since baseline. In MRM or GLMM analyses in which this Time The study will examine the efficacy of Kundalini Yoga for GAD as
centered variable is used, the significance of the dummy variable compared to CBT and a psychological attention control condition. Extant
contrasts (yoga vs. SE and CBT vs. SE) provides the test of the differences literature suggests that CBT conveys its therapeutic effect, in part,
between these treatment conditions at post-treatment. A similar ap- through modification of maladaptive cognitions [20]. Yoga entails the
proach can be used to test differences between treatment conditions cultivation of physical and psychological health through a variety of
at follow-up. techniques including physical posture exercises, breathing exercises,
 S.G. Hofmann et al. / Contemporary Clinical Trials 44 (2015) 70–76 75

relaxation strategies, and meditation practice. As mindfulness based in- [3] J. Ormel, M. VonKorff, T.B. Ustun, S. Pini, A. Korten, T. Oldehinkel, Common mental
disorders and disability across cultures: results from the WHO Collaborative Study
terventions comprise several of these components [58], we hypothesize on Psychological Problems in General Health Care, JAMA 272 (1994) 1741–1748.
that increases in mindfulness will function as a core treatment mecha- [4] W.H. Schonfeld, C.J. Verboncoeur, S.K. Fifer, R.C. Lipschutz, D.P. Lubeck, D.P.
nism in yoga. Furthermore, we are evaluating the presence of biological Buesching, The functioning and well-being of patients with unrecognized anxiety
disorders and major depressive disorder, J. Affect. Disord. 43 (1997) 105–119.
mechanisms in yoga, namely vagal tone. The chief pathological process [5] O. Brawman-Mintzer, N. Emmanuel, M.P. Jarrell, J.C. Ballenger, Psychiatric comor-
underling GAD is worry. Several studies have established that excessive bidity in patients with generalized anxiety disorder, Am. J. Psychiatry 2 (1993) 4.
worry entails reduced autonomic flexibility as a result of low cardiac [6] T.A. Brown, D.H. Barlow, M.R. Liebowitz, The empirical basis of generalized anxiety
disorder, Am. J. Psychiatry 151 (1994) 1272–1280.
vagal tone [59,60]. Assessment of this biomarker is accomplished by re- [7] T.A. Brown, L.A. Campbell, C.L. Lehman, J.R. Grisham, R.B. Mancill, Current and life-
cording respiratory sinus arrhythmia, which reflects rhythmic varia- time comorbidity of the DSM-IV anxiety and mood disorders in a large clinical sam-
tions in heart rate that occur at the frequency of respiration. Because ple, J. Abnorm. Psychol. 110 (2001) 585.
[8] S.G. Hofmann, J.A. Smits, Cognitive-behavioral therapy for adult anxiety disorders: a
yoga exercises emphasize breathing techniques, we predict that symp-
meta-analysis of randomized placebo-controlled trials, J. Clin. Psychiatry 69 (2008)
tom improvement will also be a consequence of changes in vagal tone. 621.
By ascertaining the chief mechanisms responsible for treatment effica- [9] A.E. Kazdin, S.L. Blase, Rebooting psychotherapy research and practice to reduce the
cy, the results of the current study can facilitate the development of burden of mental illness, Perspect. Psychol. Sci. 6 (2011) 21–37.
[10] B. Tsui, Yoga magazine gets boost from spirituality surge, Advert. Age 71 (2000) 20.
more parsimonious, effective treatments for GAD. [11] T.C. Clarke, L.I. Black, B.J. Stussman, P.M. Barnes, R.L. Nahin, Trends in the use of com-
Adopting a randomized controlled trial design provides a stringent plementary health approaches among adults: United States, 2002–2012, National
experimental framework to examine our hypotheses. Moreover, the health statistics reports 2015, pp. 1–16.
[12] P.M. Barnes, B. Bloom, R.L. Nahin, Statistics NCfH, Complementary and Alternative
data analytic and methodological strategies used in the current study Medicine Use among Adults and Children: United States, 2007, US Department of
include prospective efforts to address potential limitations related to at- Health and Human Services, Centers for Disease Control and Prevention, National
trition. To retain sufficient power for group comparisons, our initial Center for Health Statistics, Hyattsville, MD, 2008.
[13] J. Curtiss, D.H. Klemanski, Factor analysis of the five facet mindfulness questionnaire
sample size was estimated under the assumption of 25% missing data, in a heterogeneous clinical sample, J. Psychopathol. Behav. 36 (2014) 683–694.
and our complementary use of mixed-effects regression models and [14] J. Curtiss, D.H. Klemanski, Teasing apart low mindfulness: Differentiating deficits in
pattern mixture modeling will foster greater confidence in the validity mindfulness and in psychological flexibility in predicting symptoms of generalized
anxiety disorder and depression, J. Affect. Disord. 166 (2014) 41–47.
of our results. Certain limitations do warrant mention, however. It
[15] J.D. Herbert, E.M. Forman, Acceptance and Mindfulness in Cognitive Behavior
would be desirable to assess our proposed mechanisms more frequent- Therapy: Understanding and Applying the New Therapies, John Wiley & Sons, 2011.
ly, which would provide a more robust test of mediation. As these vari- [16] G.L. Stonerock, B.M. Hoffman, P.J. Smith, J.A. Blumenthal, Exercise as treatment for
anxiety: systematic review and analysis, Ann. Behav. Med. (2015) 1–15.
ables are measured at pre-, mid-, and post-treatment, the current study
[17] R.P. Brown, P.L. Gerbarg, F. Muench, Breathing practices for treatment of psychiatric
might be unable to discern whether changes in the mediator variables and stress-related medical conditions, Psychiatr. Clin. N. Am. 36 (2013) 121–140.
precede improvement in symptoms [61]. Though our mediational hy- [18] G.M. Manzoni, F. Pagnini, G. Castelnuovo, E. Molinari, Relaxation training for
potheses represent only a secondary aim, we recommend that future re- anxiety: a ten-years systematic review with meta-analysis, BMC Psychiatry 8
(2008) 41.
search measure mediator variables at more frequent time points to [19] J. Carmody, R.A. Baer, Relationships between mindfulness practice and levels of mind-
better establish temporal precedence. fulness, medical and psychological symptoms and well-being in a mindfulness-based
Although there are several empirically supported behavioral and stress reduction program, J. Behav. Med. 31 (2008) 23–33.
[20] S.G. Hofmann, Cognitive mediation of treatment change in social phobia, J. Consult.
pharmacological interventions that ameliorate symptoms of GAD [62], Clin. Psychol. 72 (2004) 392.
there may be a distinct opportunity to contribute to the provision [21] T. Gard, N. Brach, B.K. Hölzel, J.J. Noggle, L.A. Conboy, S.W. Lazar, Effects of a yoga-
of care by exploring alternative treatments. Given the great deal of based intervention for young adults on quality of life and perceived stress: the
potential mediating roles of mindfulness and self-compassion, J. Posit. Psychol. 7
overlap between yoga and other empirically supported interventions (2012) 165–175.
(i.e., mindfulness based interventions), there may be potential for [22] O. Bhatnagar, A. Ganguly, V. Anantharaman, Influence of Yoga training on thermo-
yoga to have a substantive impact on the clinical and economic burden regulation, Indian J. Med. Res. 67 (1978) 844.
[23] M. Gharote, A psychophysiological study of the effects of short-term yogic training
posed by GAD. The current research endeavor represents an innovative on adolescent high school boys, Yoga Mimamsa 11 (1971) 92–99.
effort to address clinically and scientifically relevant questions related to [24] S. Joseph, K. Sridharan, S. Patil, M. Kumaria, W. Selvamurthy, N. Joseph, et al., Study
the treatment of GAD. In addition, it will provide a rigorous test of the of some physiological and biochemical parameters in subjects undergoing yogic
training, Indian J. Med. Res. 74 (1981) 120.
efficacy of yoga for a prevalent mental health condition.
[25] U. Ray, S. Mukhopadhyaya, S. Purkayastha, V. Asnani, O. Tomer, R. Prashad, et al., Ef-
fect of yogic exercises on physical and mental health of young fellowship course
trainees, Indian J. Physiol. Pharmacol. 45 (2001) 37.
Conflict of interest [26] A.W. Li, C. Goldsmith, The effects of yoga on anxiety and stress, Altern. Med. Rev. 17
(2012) 21–35.
[27] N. Chugh-Gupta, F. Baldassarre, B. Vrkljan, A systematic review of yoga for state anx-
Dr. Hofmann receives compensation for his work as an advisor from iety: considerations for occupational therapy, Rev. Can. Ergo. 80 (2013) 150–170.
the Palo Alto Health Sciences and Otsuka America Pharmaceutical, Inc., [28] S. Doria, A. de Vuono, R. Sanlorenzo, F. Irtelli, C. Mencacci, Anti-anxiety efficacy of
Sudarshan Kriya yoga in general anxiety disorder: A multicomponent, yoga based,
and for his work as a Subject Matter Expert from John Wiley & Sons, breath intervention program for patients suffering from generalized anxiety disor-
Inc. and SilverCloud Health, Inc. He also receives royalties and payments der with or without comorbidities, J. Affect. Disord. 184 (2015) 310–317.
for his editorial work from various publishers. Dr. Khalsa is a paid con- [29] M.A. Katzman, M. Vermani, P.L. Gerbarg, R.P. Brown, C. Iorio, M. Davis, et al., A mul-
ticomponent yoga-based, breath intervention program as an adjunctive treatment
sultant for three nonprofit yoga organizations and is engaged in the in patients suffering from generalized anxiety disorder with or without comorbidi-
practice and promotion of yoga. Dr. Simon receives compensation for ties, Int. J. Yoga 5 (2012) 57.
her work as a consultant for Pfizer Pharmaceuticals and Massachusetts [30] V. Dhansoia, H. Bhargav, K. Metri, Immediate effect of mind sound resonance tech-
nique on state anxiety and cognitive functions in patients suffering from generalized
General Hospital Psychiatry Academy. Dr. Hoge, Dr. Bui, Dr. Rosenfield,
anxiety disorder: A self-controlled pilot study, Int. J. Yoga 8 (2015) 70.
Ms. Keshaviah, and Mr. Curtiss do not report any potential conflict of [31] M.K. Khalsa, J.M. Greiner‐Ferris, S.G. Hofmann, S.B.S. Khalsa, Yoga‐enhanced cogni-
interest. tive behavioural therapy (Y‐CBT) for anxiety management: a pilot study, Clin.
Psychol. Psychother. 22 (2014) 364–371.
[32] S.A. Saeed, D.J. Antonacci, R.M. Bloch, Exercise, yoga, and meditation for depressive
and anxiety disorders, Am. Fam. Physician 81 (2010) 981–986.
References [33] M. Berk, F. Ng, S. Dodd, T. Callaly, S. Campbell, M. Bernardo, et al., The validity of the
CGI severity and improvement scales as measures of clinical effectiveness suitable
[1] B.J. Sadock, V.A. Sadock, Kaplan and Sadock's Synopsis of Psychiatry: Behavioral for routine clinical use, J. Eval. Clin. Pract. 14 (2008) 979–983.
Sciences/clinical Psychiatry, Lippincott Williams & Wilkins, 2011. [34] C.D. Spielberger, R.L. Gorsuch, R.E. Lushene, State-trait anxiety inventory, Consulting
[2] R.C. Kessler, P. Berglund, O. Demler, R. Jin, K.R. Merikangas, E.E. Walters, Lifetime Psychologists Press, Palo Alto. Cal, 1970.
prevalence and age-of-onset distributions of DSM-IV disorders in the National [35] A.T. Beck, N. Epstein, G. Brown, R.A. Steer, An inventory for measuring clinical
Comorbidity Survey Replication, Arch. Gen. Psychiatry 62 (2005) 593–602. anxiety: psychometric properties, J. Consult. Clin. Psychol. 56 (1988) 893.
76 S.G. Hofmann et al. / Contemporary Clinical Trials 44 (2015) 70–76

[36] A.T. Beck, R.A. Steer, G.K. Brown, Beck Depression Inventory-II, Psychological Corpo- [53] C.K. Enders, Missing not at random models for latent growth curve analyses,
ration, San Antonio. Tx, 1996. Psychol. Methods 16 (2011) 1.
[37] S. Cohen, T. Kamarck, R. Mermelstein, A global measure of perceived stress, J. Health [54] E.A. Hoge, E. Bui, L. Marques, C.A. Metcalf, L.K. Morris, D.J. Robinaugh, et al., Random-
Soc. Behav. 28 (1983) 385–396. ized controlled trial of mindfulness meditation for generalized anxiety disorder:
[38] H. Herrman, H. Schofield, B. Murphy, Z. Metelko, S. Szabo, M. Pibernik-Okanovic, effects on anxiety and stress reactivity, J. Clin. Psychiatry 74 (2013) 786.
et al., Development of the World Health Organization WHOQOL-BREF quality of [55] J. Hintze, PASS 12, NCSS, LLC, Kaysville: Utah, USA, 2013.
life assessment, Psychol. Med. 28 (1998) 551–558. [56] T.A. Snijders, Power and sample size in multilevel linear models, Encyclopedia of
[39] L.R. Derogatis, SCL-90-R: Administration, Scoring and Procedures Manual for the R statistics in behavioral science2005.
(evised) Version and Other Instruments of the Psychopathology Rating Scale Series, [57] J.R. Stern, S.B.S. Khalsa, S.G. Hofmann, A yoga intervention for music performance
Clinical Psychometric Research, 1992. anxiety in conservatory students, Med. Probl. Perform. Art. 27 (2012) 123.
[40] T.J. Meyer, M.L. Miller, R.L. Metzger, T.D. Borkovec, Development and validation of [58] J. Kabat-Zinn, Full Catastrophe Living: Using the Wisdom of Your Body and Mind to
the penn state worry questionnaire, Behav. Res. Ther. 28 (1990) 487–495. Face Stress. Pain, and Illness, 2008.
[41] D.J. Buysse, C.F. Reynolds, T.H. Monk, S.R. Berman, D.J. Kupfer, The Pittsburgh Sleep [59] S.G. Hofmann, D.A. Moscovitch, B.T. Litz, H.-J. Kim, L.L. Davis, D.A. Pizzagalli, The
Quality Index: a new instrument for psychiatric practice and research, Psychiatry worried mind: Autonomic and prefrontal activation during worrying, Emotion 5
Res. 28 (1989) 193–213. (2005) 464.
[42] C.H. Bastien, A. Vallières, C.M. Morin, Validation of the Insomnia Severity Index as an [60] S.G. Hofmann, S.M. Schulz, S. Heering, F. Muench, L.F. Bufka, Psychophysiological
outcome measure for insomnia research, Sleep Med. 2 (2001) 297–307. correlates of generalized anxiety disorder with or without comorbid depression,
[43] S.G. Hofmann, T.B. Kashdan, The affective style questionnaire: development and Int. J. Psychophysiol. 78 (2010) 35–41.
psychometric properties, J. Psychopathol. Behav. 32 (2010) 255–263. [61] A.E. Kazdin, Mediators and mechanisms of change in psychotherapy research, Annu.
[44] M.M. Spada, C. Mohiyeddini, A. Wells, Measuring metacognitions associated with Rev. Clin. Psychol. 3 (2007) 1–27.
emotional distress: Factor structure and predictive validity of the metacognitions [62] K. Mitte, Meta-analysis of cognitive-behavioral treatments for generalized anxiety
questionnaire 30, Personal. Individ. Differ. 45 (2008) 238–242. disorder: a comparison with pharmacotherapy, Psychol. Bull. 131 (2005) 785.
[45] R.A. Baer, G.T. Smith, J. Hopkins, J. Krietemeyer, L. Toney, Using self-report assess- [63] M.B. First, R.L. Spitzer, Miriam Gibbon, Janet B.W. Williams, Structured Clinical
ment methods to explore facets of mindfulness, Assessment 13 (2006) 27–45. Interview for DSM-IV-TR Axis I Disorders, Research Version, Patient Edition with
[46] R.E. Zinbarg, M. Craske, D.H. Barlow, Mastery of Your Anxiety and Worry: Therapist Psychotic Screen (SCID-I/PW/PSY SCREEN), Biometrics Research, New York State
Guide, 2nd ed. Oxford University Press, New York, NY, 2006. Psychiatric Institute, New York, 2002.
[47] R.J. DeRubeis, P. Crits-Christoph, Empirically supported individual and group psy- [64] T.A. Brown, D.H. Barlow, Anxiety and Related Disorders Interview Schedule for
chological treatments for adult mental disorders, J. Consult. Clin. Psychol. 66 DSM-5 (ADIS-5 L): Lifetime Version, Client Interview Schedule, Oxford University
(1998) 37–52. Press, 2014.
[48] A. Wells, Metacognitive Therapy for Anxiety and Depression, Guilford press, 2011. [65] M.K. Shear, J. Vander Bilt, P. Rucci, J. Endicott, B. Lydiard, M.W. Otto, et al., Reliability
[49] J. Efird, Blocked randomization with randomly selected block sizes, Int. J. Environ. and validity of a structured interview guide for the Hamilton Anxiety Rating Scale
Res. Public Health 8 (2010) 15–20. (SIGH‐A), Depress. Anxiety 13 (2001) 166–178.
[50] R.M. Hamer, P.M. Simpson, Last observation carried forward versus mixed models in [66] Z.S. Nasreddine, N.A. Phillips, V. Bédirian, S. Charbonneau, V. Whitehead, I. Collin,
the analysis of psychiatric clinical trials, Am. J. Psychiatry 166 (2009) 639–641. et al., The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cog-
[51] M.S. Fritz, D.P. MacKinnon, Required sample size to detect the mediated effect, nitive impairment, J. Am. Geriatr. Soc. 53 (2005) 695–699.
Psychol. Sci. 18 (2007) 233–239.
[52] D. Hedeker, R.D. Gibbons, Longitudinal Data Analysis, John Wiley & Sons, 2006.