A systematic review of randomised clinical trials was conducted to assess the effect of spinal manipulative therapy on clinically relevant outcomes in patients with chronic low back pain. Databases searched included EMBASE, CINAHL, MEDLINE and PEDro. Methodological assessment of the trials was performed using the PEDro scale. Where there was sufficient homogeneity, a meta-analysis was conducted. Nine trials of mostly moderate quality were included in the review. Two trials were pooled comparing spinal manipulative therapy and placebo treatment, and two other trials were pooled comparing spinal manipulative therapy and non-steroidal anti-inflammatory drugs (NSAIDs). Spinal manipulative therapy reduced pain by 7mm on a 100mm visual analogue scale (95% CI 1 to 14) at one month follow-up when compared with placebo treatment, and by 14mm (95% CI -11 to 40) when compared with NSAIDs. Spinal manipulative therapy reduced disability by 6 points (95% CI 1 to 12) on a 100-point disability questionnaire when compared with NSAIDs. It is concluded that spinal manipulation does not produce clinically worthwhile decreases in pain compared with sham treatment, and does not produce clinically worthwhile reductions in disability compared with NSAIDs for patients with chronic low back pain. It is not clear whether spinal manipulation is more effective than NSAIDs in reducing pain of patients with chronic low back pain.