BACKGROUND: Traction is used to treat low-back pain (LBP), often with other treatments. OBJECTIVES: To determine traction's effectiveness, compared to reference treatments, placebo, sham traction or no treatment for LBP. SEARCH STRATEGY: We searched CENTRAL (The Cochrane Library 2006, issue 4), MEDLINE, EMBASE, and CINAHL to October 2006, references in relevant reviews and personal files. SELECTION CRITERIA: Randomized controlled trials (RCTs) involving traction to treat acute (less than four weeks duration), sub-acute (four to 12 weeks) or chronic (more than 12 weeks) non-specific LBP with or without sciatica. DATA COLLECTION AND ANALYSIS: Study selection, methodological quality assessment and data extraction were done independently by two authors. As there were insufficient data for statistical pooling, we performed a qualitative analysis. MAIN RESULTS: We included 25 RCTs (2206 patients; 1045 receiving traction). Five trials were considered high quality.For patients with mixed symptom patterns (acute, sub-acute and chronic LBP with and without sciatica) there is: strong evidence of no statistically significant difference in outcomes between traction as a single treatment and placebo, sham or no treatment; moderate evidence that traction as a single treatment is no more effective than other treatments; limited evidence of no significant difference in outcomes between a standard physical therapy program with or without continuous traction. For LBP patients with sciatica (with acute, sub-acute or chronic pain), there is conflicting evidence in several comparisons: autotraction compared to placebo, sham or no treatment; other forms of traction compared to other treatments; different forms of traction. In other comparisons, there were no statistically significant differences; the evidence is moderate for continuous or intermittent traction compared to placebo, sham or no treatment, and limited for light versus normal force traction. AUTHORS' CONCLUSIONS: Implications for practice: The results of the available studies involving mixed groups of acute, sub-acute and chronic patients with LBP with and without sciatica were quite consistent, indicating that continuous or intermittent traction as a single treatment for LBP is not likely effective for this group. Traction for patients with sciatica cannot be judged effective at present either, due to inconsistent results and methodological problems in most studies. We conclude that traction as a single treatment for LBP is probably not effective. Implications for research: Any future research on traction for patients with LBP should distinguish between symptom pattern and duration, and should be carried out according to the highest methodological standards.