BACKGROUND: Non-specific low back pain is a common, potentially disabling condition usually treated with self-care and non-prescription medication. For chronic low back pain, current guidelines state that exercise therapy may be beneficial. Yoga is a mind-body exercise sometimes used for non-specific low back pain. OBJECTIVES: To assess the effects of yoga for treating chronic non-specific low back pain, compared to no specific treatment, a minimal intervention (e.g. education), or another active treatment, with a focus on pain, function, and adverse events. SEARCH METHODS: We searched CENTRAL, MEDLINE, Embase, five other databases and four trials registers to 11 March 2016 without restriction of language or publication status. We screened reference lists and contacted experts in the field to identify additional studies. SELECTION CRITERIA: We included randomized controlled trials of yoga treatment in people with chronic non-specific low back pain. We included studies comparing yoga to any other intervention or to no intervention. We also included studies comparing yoga as an adjunct to other therapies, versus those other therapies alone. DATA COLLECTION AND ANALYSIS: Two authors independently screened and selected studies, extracted outcome data, and assessed risk of bias. We contacted study authors to obtain missing or unclear information. We evaluated the overall certainty of evidence using the GRADE approach. MAIN RESULTS: We included 12 trials (1080 participants) carried out in the USA (seven trials), India (three trials), and the UK (two trials). Studies were unfunded (one trial), funded by a yoga institution (one trial), funded by non-profit or government sources (seven trials), or did not report on funding (three trials). Most trials used Iyengar, Hatha, or Viniyoga forms of yoga. The trials compared yoga to no intervention or a non-exercise intervention such as education (seven trials), an exercise intervention (three trials), or both exercise and non-exercise interventions (two trials). All trials were at high risk of performance and detection bias because participants and providers were not blinded to treatment assignment, and outcomes were self-assessed. Therefore, we downgraded all outcomes to 'moderate' certainty evidence because of risk of bias, and when there was additional serious risk of bias, unexplained heterogeneity between studies, or the analyses were imprecise, we downgraded the certainty of the evidence further.For yoga compared to non-exercise controls (9 trials; 810 participants), there was low-certainty evidence that yoga produced small to moderate improvements in back-related function at three to four months (standardized mean difference (SMD) -0.40, 95% confidence interval (CI) -0.66 to -0.14; corresponding to a change in the Roland-Morris Disability Questionnaire of mean difference (MD) -2.18, 95% -3.60 to -0.76), moderate-certainty evidence for small to moderate improvements at six months (SMD -0.44, 95% CI -0.66 to -0.22; corresponding to a change in the Roland-Morris Disability Questionnaire of MD -2.15, 95% -3.23 to -1.08), and low-certainty evidence for small improvements at 12 months (SMD -0.26, 95% CI -0.46 to -0.05; corresponding to a change in the Roland-Morris Disability Questionnaire of MD -1.36, 95% -2.41 to -0.26). On a 0-100 scale there was very low- to moderate-certainty evidence that yoga was slightly better for pain at three to four months (MD -4.55, 95% CI -7.04 to -2.06), six months (MD -7.81, 95% CI -13.37 to -2.25), and 12 months (MD -5.40, 95% CI -14.50 to -3.70), however we pre-defined clinically significant changes in pain as 15 points or greater and this threshold was not met. Based on information from six trials, there was moderate-certainty evidence that the risk of adverse events, primarily increased back pain, was higher in yoga than in non-exercise controls (risk difference (RD) 5%, 95% CI 2% to 8%).For yoga compared to non-yoga exercise controls (4 trials; 394 participants), there was very-low-certainty evidence for little or no difference in back-related function at three months (SMD -0.22, 95% CI -0.65 to 0.20; corresponding to a change in the Roland-Morris Disability Questionnaire of MD -0.99, 95% -2.87 to 0.90) and six months (SMD -0.20, 95% CI -0.59 to 0.19; corresponding to a change in the Roland-Morris Disability Questionnaire of MD -0.90, 95% -2.61 to 0.81), and no information on back-related function after six months. There was very low-certainty evidence for lower pain on a 0-100 scale at seven months (MD -20.40, 95% CI -25.48 to -15.32), and no information on pain at three months or after seven months. Based on information from three trials, there was low-certainty evidence for no difference in the risk of adverse events between yoga and non-yoga exercise controls (RD 1%, 95% CI -4% to 6%).For yoga added to exercise compared to exercise alone (1 trial; 24 participants), there was very-low-certainty evidence for little or no difference at 10 weeks in back-related function (SMD -0.60, 95% CI -1.42 to 0.22; corresponding to a change in the Oswestry Disability Index of MD -17.05, 95% -22.96 to 11.14) or pain on a 0-100 scale (MD -3.20, 95% CI -13.76 to 7.36). There was no information on outcomes at other time points. There was no information on adverse events.Studies provided limited evidence on risk of clinical improvement, measures of quality of life, and depression. There was no evidence on work-related disability. AUTHORS' CONCLUSIONS: There is low- to moderate-certainty evidence that yoga compared to non-exercise controls results in small to moderate improvements in back-related function at three and six months. Yoga may also be slightly more effective for pain at three and six months, however the effect size did not meet predefined levels of minimum clinical importance. It is uncertain whether there is any difference between yoga and other exercise for back-related function or pain, or whether yoga added to exercise is more effective than exercise alone. Yoga is associated with more adverse events than non-exercise controls, but may have the same risk of adverse events as other back-focused exercise. Yoga is not associated with serious adverse events. There is a need for additional high-quality research to improve confidence in estimates of effect, to evaluate long-term outcomes, and to provide additional information on comparisons between yoga and other exercise for chronic non-specific low back pain.