Yoga versus education for Veterans with chronic low back pain: study protocol for a randomized controlled trial

The structured and manualized yoga protocol for cLBP was initially developed by an expert panel led by the principal investigator (RBS) in 2007. It was used in a pilot study of 30 predominantly minority low-income civilian participants [40]. The protocol was further refined in 2012 for a yoga dosing study comparing one vs. two yoga classes per week [37]. This protocol was again adapted for use in a large study of 320 civilians that compared yoga, physical therapy, and education [79]. In order to adapt the yoga protocol to a Veteran population, a panel of 11 experts and stakeholders convened in November 2014. Attendees included Veterans, military leaders, yoga teachers, representatives from organizations focused on providing mind-body therapies to Veterans (There & Back Again^©, Warriors at Ease^©), and researchers with expertise in yoga, low back pain, and the VHA. Prior to the meeting, relevant materials (previous yoga for back pain protocols and literature) were distributed for the attendees to review. The full-day in-person meeting included presentations and discussions of unique aspects to teaching yoga to Veterans, class format, segment themes, relaxation exercises, breathing exercises, and specific postures. Based on this input, iterative drafts of the yoga instructor documents were revised and circulated until broad consensus was obtained on the final version.

Yoga classes contain no more than eight to ten Veterans and two yoga teachers to assure a low Veteran-to-instructor ratio (no more than five Veterans to one teacher). This arrangement maximizes effectiveness and safety, and allows flexibility in yoga teacher scheduling based on the study’s needs. Veterans are asked to attend one 75-minute class per week and receive reminder phone calls the day before each scheduled class. Attendance is taken by research staff. We provide $10 travel compensation for each yoga session. Mats and props are provided. Each class includes a yoga breathing exercise (pranayama), discussion of yoga philosophical principles, yoga postures (asanas), and a relaxation exercise (svasana). Veterans are frequently advised to proceed carefully and slowly. The difficulty of postures increases throughout the 12 weeks (Table 2). The 12-week intervention period is divided into four 3-week segments, titled Opening to Something New, Listening to Your Back, Engaging Your Power, and Bringing it Home. Posture variations, modifications, and various aids (e.g., block, chair, strap, wall) are built into the protocol to accommodate a range of abilities. Yoga participants are encouraged to practice at home for 30 minutes on days when they do not have class. To help Veterans practice at home, they receive a guidebook describing and depicting the protocol (see Additional file 1), yoga mat, strap, two blocks, and home practice instructional videos (see Additional file 2-5). Home practice videos portray Veterans in an introductory video describing the value of yoga practice in their own lives, two 30-minute videos that follow the same structure as group yoga classes, and one 20-minute extended breathing and relaxation video. Veterans are given a link to view the home practice videos online or, if requested, they receive a take-home DVD.

All yoga teachers underwent training prior to teaching study classes. A 6-hour in-person training covered the specific yoga practices suggested (e.g., postures, breathing exercises) and information on how to teach a culturally-sensitive yoga class to Veterans. Online webinars provided by Warriors at Ease^© included four 90-minute courses on cultural sensitivity, common injuries or conditions in Veterans, the role of the yoga teacher(s), and managing emotional reactions. All teachers received and follow an instructor manual (see Additional file 6), which contains detailed information on the segment themes, week-by-week yoga class structure, teaching individual postures with or without modifications, leading breathing exercises, and guiding Veterans through relaxation. To assess fidelity to the protocol, researchers observe 10 % of classes using a protocol fidelity checklist.

The Back Pain Helpbook [68] is given to individuals randomized to the education group. This book has been used successfully in previous cLBP studies for educational purposes [32, 80]. The book includes information on how back pain is influenced by posture, emotions, fear avoidance behaviors, and social support. It also encourages pain self-management strategies such as mind-body relaxation techniques, exercise programs, and medications. In addition, Veterans receive a handout after randomization suggesting specific chapters to read throughout the intervention period and brief mailed newsletters every three weeks that highlight main points from the recommended chapters (see Additional file 7). Check-in calls, which are proceeded by a newsletter mailing, are conducted by research staff every three weeks to assess progress and reinforce study retention efforts.

Table 3 shows the data collection schedule. Our measures include core patient-oriented outcomes for the cLBP trials, including those recommended by the Report of the NIH Task Force on Research Standards for Chronic Low Back Pain [73]. Change in pain intensity and back-related function are the two co-primary outcomes at 12 weeks [63, 71]. Secondary back pain outcomes include pain medicine use in the previous week, including medication class and dosage [81]; health-related quality of life (SF-12) [82]; pain interference [83]; overall improvement (7-point Likert scale, 0 = extremely worsened to 6 = extremely improved); and patient satisfaction with cLBP treatment (5-point Likert scale, 1 = very satisfied to 5 = very dissatisfied) [84]. Psychological secondary outcomes include post-traumatic stress symptoms (PCL-C) [85] and depression symptoms measured using the 9-item Patient Health Questionnaire (PHQ-9) [86]. The Pittsburgh Sleep Quality Index (PSQI) is also administered [87]. Exploratory neuropsychological outcomes include post-concussive symptoms measured using the Neurobehavioral Symptom Inventory (NSI) [88], 7-item Generalized Anxiety Disorder questionnaire (GAD-7) [89], Coping Strategies Questionnaire (CSQ) [90], and Pain Self-Efficacy Questionnaire (PSEQ) [91]. Exploratory social outcomes include marital/family functioning measured using the Dyadic Adjustment Scale (DAS) [92], employment status (Work Productivity and Activity Impairment questionnaire) [93], and housing status. NIH PROMIS® (Patient Reported Outcomes Measurement Information System) tools [94] for outcomes of interest, such as pain, function, depression, anxiety, and sleep, are used to assess how well they correlate with the DVPRS, RMDQ, PHQ-9, GAD-7, and PSQI. At baseline, we also gather standard sociodemographic information and possible covariates, including race, ethnicity, income, education, military service history, medical comorbidities, back pain history, expectation of helpfulness for the different interventions [95], and preference of treatment assignment.

We use a multi-method approach to collect cost data at baseline, 6 weeks, 12 weeks, and 24 weeks. Since our interventions may influence other common comorbidities of cLBP (e.g., depression [52]), we measure both back-related utilization and total medical utilization. Direct medical costs are measured and consist of the cost of (1) implementing the interventions themselves and (2) ongoing medical utilization before, during, and after the intervention. Intervention implementation costs (e.g., non-study-specific staff hours, materials, facility use) are captured from study records and valued at their actual costs. Ongoing total VA medical utilization including visits, hospitalizations, tests, radiology, and medications are taken directly from the electronic medical record system. Direct medical costs are valued at their actual costs to the VA. Any non-VA medical utilization (e.g., chiropractors, out-of-pocket back-related expenses) is obtained from a cost questionnaire completed by the Veteran with the primary initial recall period being the past 6 weeks and is valued at the reported actual price paid by Veterans. Indirect costs (i.e., lost productivity) for employed Veterans will be calculated as the number of lost productive hours (as reported in the Work Productivity and Activity Impairment questionnaire) multiplied by a national average cost of employment for each Veteran’s general job category. Lost productivity costs for those not in the work force are assumed to be captured by their report of overall quality of life [96]. Lost productivity costs for those looking for work will be considered in the sensitivity analyses.

Separate and conjoint interviews are conducted by a trained qualitative interviewer with the Veteran and his or her partner at two time points – prior to participation in yoga and approximately four weeks after ending classes – in order to maximize the opportunity to authentically discuss the impact of cLBP and yoga on family relationships. Interviews are conducted with each member of the couple individually followed by a short conjoint interview. We encourage couples to attend interviews at the same time. However, if this is not possible, we conduct the two interviews within approximately 72 hours. Semistructured interview guides incorporate the following domains: the Veteran’s perception of the impact of cLBP on himself or herself and the partner or family in terms of mood and functioning; the partner’s perception of the impact of the Veteran’s cLBP on the Veteran’s mood and functioning; the partner’s perception of the Veteran’s cLBP on her/his own and overall family mood and functioning; the attitude of the Veteran and partner toward yoga; the Veteran and partner’s perception of change in the Veteran’s cLBP; the Veteran and partner’s perception of any possible change in psychological concerns including PTSS, depression, and anxiety after yoga; and the Veteran and partner’s perception of any change in overall family mood and functioning after yoga classes (see Additional file 4).

Interviewers obtain informed consent from the Veteran and partner prior to the interviews. Each individual receives $25 for each 60–90 minute interview visit. Interviews take place at a location convenient to the couple.

If a couple completes a pre-intervention interview but declines to complete a second interview, we attempt to replace that couple with another couple.

Per-protocol analysis will also be conducted similarly. Adherence to the yoga protocol will be defined as attending at least 9 of the 12 yoga classes. For education, adherence will be defined as reading at least three-fourths of the book by self-report.

Our hypothesis is that yoga will be superior compared to education for reducing PTSS in the subset of participants with comorbid PTSS (baseline PCL-C ≥30). We will first compare PCL-C scores between groups for the subset of participants who qualify as having PTSS at baseline using methods similar to those described above for primary outcome analyses. We will also examine the whole study sample and the subset of participants with lower baseline PCL-C scores. We stratify for this variable during randomization and thus anticipate similar numbers and baseline characteristics between the two study groups. We will also explore whether a change in PTSS is an effect modifier for our primary outcomes of pain and function. We will similarly compare yoga to education for the subgroup of participants with moderate or greater depressive symptoms (PHQ-9 ≥10) [97] and those with less depressive symptoms. Exploratory subset analyses will also be conducted for Veterans scoring moderate or higher on anxiety symptoms (GAD-7 ≥10) and sleep difficulties (PSQI >5). Similarly, an exploratory subset analysis will be conducted for participants with post-concussive symptoms, defined as scoring ≥20 on the NSI [98].

Since back pain in older Veterans may be different structurally compared to younger Veterans, we will also complete exploratory subset analyses of Veterans ≤45 and >45 years old to assess if age confounds the response to yoga.

To analyze intermediate-term outcomes, we will examine changes from baseline to week 6 using methodology similar to the primary 12-week analyses. Additional subgroup longitudinal analyses can be done using only data from those participants who were 12-week completers. We will also compare the change in pain medication usage over time using a generalized-estimating equation approach or nonlinear mixed effects model to account for the repeated measures of a dichotomous outcome. This approach will parallel the longitudinal analyses for continuous outcomes described above.

All cost-effectiveness analyses will use the intention-to-treat principle. Quality-adjusted life-years (QALYs) will be calculated from the results of the SF-12 based on an algorithm developed by Brazier et al. [99] For the Veteran perspective, we will compare Veterans' incremental out-of-pocket costs (e.g., transportation, over-the-counter medications, and co-payments) to their incremental QALY impacts. For the VHA perspective, incremental direct intervention costs will be compared to ongoing direct medical utilization costs. This cost-benefit analysis will address whether, economically, yoga should be a treatment option for the VA population. For the society perspective, we will conduct a cost-utility analysis by comparing the incremental societal costs for each treatment arm (i.e., direct medical, nonmedical, and productivity costs) to the incremental change in QALYs [100]. We will use bootstrap methods to calculate confidence intervals [101] and perform one-way sensitivity analyses to determine the robustness of our estimates with different assumptions used to value productivity [102]. Combined non-back pain and back pain-related utilization will be used for all of these analyses. We will also perform sensitivity analyses using only back pain-related utilization.

Qualitative interviews will be downloaded from digital recorders, transcribed, and cleaned by qualitative research staff. Given the exploratory nature of the interviews, codes will be developed both inductively and deductively using techniques from Grounded Theory [103]. Initially, three research team members will read transcripts from three dyads (six interviews) to identify content categories using the process of line-by-line coding. In order to establish reliability, the team will then confer to discuss codes, condense overlapping areas, and consider alternative possibilities with the goal of developing a coding schema to be used with all interviews. Subsequently, two team members will read each of the remaining pairs of transcripts and continue to conduct focused coding using the decided-upon schema. The team will confer every four transcripts to insure that the codes remain relevant and accurate given the research questions. Where necessary, coding categories will be changed and condensed. This process of line-by-line and focused coding will ultimately lead to axial coding, where larger thematic categories and relationships between them are identified. Matrices will be utilized to compare and contrast code categories and individual cases. The software program NVivo (QSR^©, Melbourne, Australia) will be used for qualitative data management and analysis.