Sherman 2011.

Study characteristics
Methods	Randomized controlled parallel‐group trial.
Participants	228 participants with chronic non‐specific LBP. Setting: trial run through an integrated healthcare system. 7 cohorts of classes were held in 6 different cities. Yoga classes were held at the health system facilities (exact number of facilities not stated). Country: USA. Recruitment: invitations mailed to people with primary care visits for back pain, advertisements in a health plan consumer magazine, and direct‐mail postcards. For 4 cohorts, augmentation of recruitment through outreach to general population (methods not described). Inclusion criteria: men and women aged 20–64 years with LBP. Exclusion criteria: people with back pain that was attributable to a specific cause (e.g. spondylolisthesis), or an underlying condition (e.g. pregnancy), complex (e.g. sciatica), minimally painful (i.e. rating of < 3 on a 'bothersomeness' scale of 0 to 10), or not chronic (i.e. duration < 3 months); had contraindications to the interventions (e.g. severe disc disease); had major depression; were unable to give informed consent or participate in interviews owing to mental or medical issues (e.g. dementia) or an inability to speak English; or were unable to participate in classes or unwilling to practice at home. Duration and follow‐up: interventions were provided for 12 weeks and there was an additional follow‐up at 26 weeks.
Interventions	Yoga group: (n = 92) 12 weekly 75‐min viniyoga classes. The viniyoga intervention was the same as that used in an earlier trial (Sherman 2005). Classes included breathing exercises, a set of 5–11 postures, and guided deep relaxation. 6 distinct and progressive classes were taught in pairs. Home practice: encouraged for 20 min on non‐class days. Yoga participants received a handout and CD to assist them in home practice. Exercise group: (n = 91) 12 weekly 75‐min exercise classes. The exercise intervention was adapted from the intervention used in an earlier trial (Sherman 2005). Classes included aerobic exercises, stretches, and strengthening exercises. Home practice: encouraged for 20 min on non‐class days. Exercise participants received a handout and DVD to assist them in home practice. Self‐care group: (n = 45) participants received a copy of The Back Pain Helpbook providing information on causes of back pain and advice on how to manage pain (Moore 1999). Common interventions: all intervention groups continued to have access to medical care provided by their insurance plan. Co‐interventions: no specific mention of allowed or restricted co‐interventions.
Outcomes	Back‐specific function (RMDQ) at 6, 12, and 26 weeks. Clinical improvement (dichotomous variable measuring whether LBP was improved, yes/no) at 6, 12, and 26 weeks. Other outcomes collected: bothersomeness of pain during the previous week (0 = 'not at all bothersome' and 10 = 'extremely bothersome'); 30% improvement in outcomes; 50% improvement in outcomes; very satisfied with overall care for lower back pain; days of activity restriction (not presented in study results); days in bed (not presented in study results); work loss (not presented in study results).
Notes	Adverse events: of the 87 yoga and 75 stretching class attendees, 13 in each group reported a mild or moderate adverse experience possibly related to treatment (mostly increased back pain), and 1 yoga attendee experienced a herniated disk. 1/45 participants randomized to self‐care reported increased pain after doing recommended exercises. Measurement of expectations or treatment preferences at baseline: before randomization, information on sociodemographic characteristics, back pain history, and treatment‐related beliefs collected. Table 1 of publication showed the median expectation of helpfulness for each treatment in each group: it was a median of 8 for yoga and exercise for all groups, and a median of 4 for self‐care for all groups. Preferred treatment was yoga for 26–32%, exercise for 17–22%, and other for 51–53% in intervention groups. Funding: US NIH. Quote: "Financial Disclosure: None reported. Funding/Support: This study was funded by Cooperative Agreement Number U01 AT003208 from the National Center for Complementary and Alternative Medicine (NCCAM). Discussions with several NCCAM staff influenced the study design." Additional notes: we extracted data from the online supplement to the Annals 2011 publication, choosing the adjusted 2‐step imputed data from eTable 4 for the primary analysis, and the adjusted complete‐case analysis from eTable 4 for the complete‐case sensitivity analysis.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomization used.
Allocation concealment (selection bias)	Low risk	Randomization schedule generated by statistician; inaccessible to staff.
Blinding of participants	High risk	No blinding as control interventions were conventional exercise or self‐care book; outcomes based on self‐assessment.
Blinding of personnel/providers	High risk	No blinding.
Blinding of outcome assessors	High risk	Quote: "Telephone interviews were conducted by masked interviewers at baseline and at 6, 12, and 26 weeks after randomization." Comment: however, participants were not blinded and self‐reported the outcomes.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Low attrition rates; authors also conducted a secondary analysis using a single imputation method to assess the sensitivity of the complete‐case results to loss to follow‐up. The sensitivity analysis was provided online and was consistent with the primary outcomes.
Selective reporting (reporting bias)	Low risk	Outcomes reported as per published trial protocol. Note: several mediating variables mentioned in protocol not reported in primary publication.
Group similarity at baseline	Unclear risk	Groups matched on most important indicators except that yoga group had greater back pain dysfunction.
Co‐interventions	Low risk	No between‐group differences in back pain‐related healthcare visits; medication use matched initially and decreased in yoga and exercise (active intervention) groups.
Compliance	Low risk	95% of yoga participants attended ≥ 1 class and they attended a median of 8 out of a possible 12 classes. 82% of exercise participants attended ≥ 1 class and they attended a median of 9 classes. 65% (yoga) and 59% (exercise) attended ≥ 8 classes.
ITT analysis	Low risk	Statement that ITT analysis was carried out.
Timing of outcome assessments	Low risk	Outcome assessment at set time points.
Other bias	Low risk	No other biases identified.