Eight randomised controlled trials (RCTs) were included in meta-analyses (n= 614 patients). Methodological quality varied between trials, with a Jadad score ranging from 1 to 5. Blinding and concealment of allocation were inadequate in the majority of trials. The degree of confidence that acupuncture was applied appropriately ranged from 20% to 90%. All RCTs except for one had a short duration of the intervention (less than three months).
Compared with penetrating sham acupuncture, manual or electro-acupuncture was associated with a non-significant reduction in pain on the Visual Analogue Scale (weighted mean difference -0.46, 95% confidence interval (CI): -1.70 to 0.77; three RCTs).
Compared with conventional drugs, acupuncture plus moxibustion was associated with a non-significant difference in the Visual Analogue Scale response rate (relative risk 1.12, 95% CI: 0.99 to 1.28; three RCTs), pain index (weighted mean difference 1.53, 95% CI: -0.57 to 3.63; two RCTs) and joint swelling index (weighted mean difference 0.25, 95% CI: -1.31 to 1.82; two RCTs).
Statistically significant heterogeneity was not observed in the outcomes. Subgroup analyses assessing the effects of manual acupuncture on pain reduction compared with penetrating sham acupuncture were reported.
Additional outcomes for American College of Rheumatology response criteria (ACR20), Health Assessment Questionnaire (HAQ), Disease Assessment Scale (DAS) index and number of swollen/tender joints were also reported.