A total of eight RCTs (n=896) were included in this review (11 publications) of which five reported sufficient data to be included in meta-analyses. Sample sizes ranged from small pilot studies (n=10) to large multi-centre studies (n=409). Overall the methodological quality was reported as relatively high with all studies scoring between 3 and 4 out of 5. In all studies where blinding of patients was assessed, this was reported as having been successful. Publication bias did not seem to be present based on the funnel plot and Egger test.
There was no significant difference between active and sham acupuncture in the number of headache days per month during treatment, WMD -2.93 (95% CI: -7.49, 1.64; five studies). Statistically significant heterogeneity was noted (X2=119, p<0.00001). The forest plot was inspected visually and one outlier study removed (details not given). Re-analysis produced a non-significant WMD of -1.37 (95% CI: -2.93, 0.18) without significant statistical heterogeneity.
There was a significant difference in favour of active acupuncture in headache days per month at long-term follow up, WMD -1.83 (95% CI: -3.01, -0.64; four studies) without significant heterogeneity.
There was no significant difference between active and sham acupuncture in the headache intensity during treatment, WMD -7.24 (95% CI: -18.46, 3.99; three studies), however, there was a significant difference in favour of active acupuncture at long-term follow up, WMD -3.64 (95% CI: -6.55, -.073; four studies). There was no statistically significant heterogeneity noted in either of these analyses.
Three trials reported adverse events data. No severe adverse events related to the active acupuncture were observed, but one severe exacerbation of headache occurred in the sham group. Adverse events (headache exacerbation/trigger, haematoma, dizziness) occurred in 16 per cent to 17 per cent of the active groups, and in four per cent to 17 per cent of the control groups where reported.