Saper 2009.

Study characteristics
Methods	Randomized controlled parallel‐group trial.
Participants	30 participants with moderate‐to‐severe chronic non‐specific LBP. Setting: study run from 2 community health centers. Yoga classes held at 1 of 2 community health centers. Country: USA. Recruitment: self‐referral and healthcare practitioner referral through flyers in the health centers and surrounding community, radio, and newspaper advertisements, and presentations and e‐mails to healthcare providers. Letters sent by providers to participants identified from community health center electronic medical records as seen in last 2 years with an LBP diagnosis. Inclusion criteria: men and women aged 18–64 years with current LBP persisting for ≥ 12 weeks, and mean LBP intensity for 2 weeks before enrolment rated ≥ 4 on a 0‐ to 10‐rating scale; sufficient understanding of English to follow class instructions and complete surveys. Exclusion criteria: yoga use in previous year; new pain medicine or other LBP treatments started within the previous month or anticipated to begin in next 6 months; pregnancy; back surgery in previous 3 years; non‐muscular pathologies (e.g. spinal canal stenosis, spondylolisthesis, infection, malignancy, fracture); severe or progressive neurologic deficits; sciatica pain ≥ back pain; active substance or alcohol abuse; serious systemic disease, medical, or psychiatric comorbidities precluding yoga practice; active or planned worker's compensation, disability, or personal injury claims; inability to attend classes at times and location offered.
Interventions	Yoga group: (n = 15) 12 weekly 75‐min hatha yoga classes divided into 4 × 3‐week themed segments followed a standardized protocol in which each class began and ended with Savasana relaxation and breathing exercises, and included a selection from 22 or 23 other yoga postures depending on class. Home practice: 30 min/day strongly encouraged. Provided participants with audio CD of protocol; portable CD player; handbook describing and depicting the exercises; and yoga mat, strap, and block. Usual care group: (n = 15) participants continued usual care and were offered the yoga intervention after the 26‐week follow‐up. Common interventions: both groups continued to receive usual medical care and medications, and received a copy of The Back Pain Helpbook (Moore 1999). Co‐interventions: both groups were discouraged from beginning any new back pain treatments during study. Duration and follow‐up: interventions provided for 12 weeks and there was an additional follow‐up at 26 weeks.
Outcomes	Back‐specific function (RMDQ) at 6, 12, and 26 weeks. Mean pain for the previous week (0 = no pain to 10 = worst possible pain) at 6, 12, and 26 weeks. Global improvement (dichotomized into improved vs no change or worse) at 12 weeks. Other outcomes collected: SF‐36 Mental Health component (not reported in results); SF‐36 Physical Health component (not reported in results); changes in medication use at 6 and 12 months.
Notes	Adverse events: 1 yoga participant reported transient worsening of LBP that improved after discontinuing yoga. No other significant adverse events reported. Measurement of expectations or treatment preferences at baseline: none. Unpublished data: Dr Robert Saper e‐mailed endpoint data for pain and back‐related function at 6 weeks to LSW on 4 November 2014. Funding: US NIH. Quote: "Dr Saper is supported by a Career Development Award (K07 AT002915‐04) from the National Center for Complementary and Alternative Medicine (NCCAM), National Institutes of Health (NIH), Bethesda, Maryland. Dr Phillips is supported by a Mid‐career Investigator Award (5K24AT000589‐08) from NCCAM, NIH. NCCAM had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript for submission."
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Used computer‐generated permuted block to generate sequence.
Allocation concealment (selection bias)	Low risk	Opaque, sequentially numbered envelopes prepared by a biostatistician with no contact with participants.
Blinding of participants	High risk	No blinding as used waiting list control; outcomes based on self‐assessment.
Blinding of personnel/providers	High risk	No blinding.
Blinding of outcome assessors	High risk	Quote: "All study participants met in person with unblinded research staff members to complete paper questionnaires at baseline, 6, and 12 weeks." Comment: participants were not blinded and self‐reported the outcomes, with the assistance of unblinded study staff.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Low attrition rates (3% at 12 weeks; 23% at 26 weeks) but reasons unclear and rate differed between groups.
Selective reporting (reporting bias)	Unclear risk	Outcomes reported as per protocol on ClinicalTrials.gov. However, SF‐36 was in protocol and in paper methods, but only lack of statistical significance was reported in results.
Group similarity at baseline	Low risk	Groups matched on most important indicators.
Co‐interventions	Low risk	Use of non‐study treatments by yoga group was 27% and control group was 40% (P = 0.7) up to 12 weeks and by yoga group was 87% and control group 100% between 12 and 26 weeks.
Compliance	Low risk	Quote: "Yoga participants attended a median of 8 classes (range 0–12)."
ITT analysis	Low risk	All randomized participants were analyzed in the group to which they were randomized and an ITT analysis with LOCF imputation for 1 missing yoga participant was carried out at 12 weeks and 7 missing yoga participants at 26 weeks.
Timing of outcome assessments	Low risk	Outcome assessment at set time points.
Other bias	Low risk	No other biases identified.