Eleven RCTs (n=2,821, with 1,155 receiving acupuncture and 1,660 controls) were included in the review. There was a slight discrepancy between the total number of patients reported in the text and tables (total quoted is the figure given in the text). Sample sizes ranged from 30 to 1,039 participants. RCTs scored between four and 10 on the validity assessment, with seven (possibly eight) scoring six or more. Experience of acupuncturists was adequate in five RCTs, but adequacy assessments were partially unblinded.
Acupuncture compared with waiting list (four RCTs): clinically relevant short-term improvements in pain and function were reported by patients receiving acupuncture, SMD -0.96 (95% CI: -1.21, -0.70) compared with patients on the waiting list, SMD: -0.93 (95% CI: -1.16, -0.69).
Acupuncture compared with usual care (two RCTs): clinically relevant short-term improvements in pain and function were reported by patients receiving acupuncture, SMD: -0.62 (95% CI: -0.75, -0.49) compared with patients receiving usual care SMD: -0.56 (95% CI: -0.69, -0.43). These improvements were maintained at six months for pain, SMD: -0.52 (95% CI: -0.66, -0.39) and function, SMD: -0.45 (95% CI: -0.59, -0.32).
Clinically irrelevant improvements were reported in the acupuncture groups compared to sham groups for pain and function outcomes in the short term (seven RCTs) and long term (three RCTs). There was evidence of significant heterogeneity in the short term for RCTs comparing acupuncture (I2=66%) with sham (I2=69%).
Sensitivity analyses on patient blinding and funding source significantly altered the results. SMD for function was no longer statistically significant and was borderline significant for pain (two RCTs).
Where reported, adverse events were similar for acupuncture and control groups, but pooling could not be undertaken due to heterogeneity. Funnel plots and the Egger test indicated potential small-study bias, but the Egger test results were not statistically significant.