Previous trials to assess the efficacy of lumbar traction for back pain have been methodologically flawed. To avoid these shortcomings, we conducted a randomised controlled trial in which high-dose traction was compared with sham traction. The sham traction was given with a specially developed brace that tightens in the back during traction. To the patient, the experience is that of traction. The patients and outcome assessor were blinded for the assigned treatment. 151 patients with at least six weeks of non-specific low back pain were randomised. Intention to treat analysis showed no differences between the groups on all outcome measures (patients' global perceived effect, severity of main complaints, functional status and pain); all 95% confidence intervals included the value zero. The number of withdrawals from treatment, loss to follow-up, and protocol deviations was low. Consequently, the per-protocol analysis showed results similar to the intention to treat analysis. Subgroup analyses did not show any group for which traction might seem promising. Our data do not support the claim that traction is effective for patients with low back pain.