Re-analysis of acupuncture trials with sham interventions based on data from the Cochrane Review

[Objective] Recent acupuncture trials with sham control groups have failed to show significant beneficial effects of real acupuncture. As a result, the specific effects of acupuncture were discounted and assumed to be placebo and/or expectation effects. However, the sham acupuncture methods used in these studies appeared to be physiologically active. Thus the influence of the different types of sham interventions on the results were re-analysed by subgroup meta-analysis. [Methods] Clinical acupuncture trials using sham interventions were selected from the Cochrane Database of Systematic Reviews. The outcome measures for the real and sham acupuncture groups from each study were collected. For subgroup analysis, sham interventions were classified into 5 types: Type A, needling to non-acupuncture points; Type B, superficial needling; Type C, non-penetration; Type D, electro-acupuncture to non-acupuncture points; Type E, electro-acupuncture without current. Cochrane Review Manager 5.3 software was used for analysis. Standardized mean difference and risk ratio were used to analyse continuous and dichotomous data, respectively. A random model was applied to all analyses. The heterogeneity of analysed studies was indicated as I 2 , and the overall effect of pooled data was shown as Z (P value). Forest and funnel plots were made for each subgroup analysis based on the different types of sham acupuncture and pain or non-pain diseases. [Results] A total of 116 studies were selected from the Cochrane Database of Systematic Reviews (from 1974 to 2013). The overall effects of real acupuncture were significantly more beneficial than the sham acupuncture groups in both meta-analyses of continuous and dichotomous data sets. The standardized mean difference and 95% confidence intervals were -0.34 [-0.44, -0.23] (Z=6.03, P<0.00001) and for risk ratio were 0.80 [0.73, 0.87] (Z=5.23, P<0.00001). Subgroup analysis based on sham acupuncture (Types A to E) and outcomes (pain or non-pain) demonstrated that half of the data sets were statistically significant although they were not related to the type of sham acupuncture and disease. The heterogeneity of the data sets was relatively high. [Conclusion] Results from meta-analysis of the pooled data indicated highly significant differences between real and sham acupuncture groups; however, we should interpret these results with caution, as heterogeneity of the trials was relatively high. Despite large variations, effect sizes of non-penetrating sham acupuncture did not generally exceed those of penetrating sham acupuncture. These results suggest that non-penetrating sham acupuncture, similar to Japanese acupuncture procedures, may produce some physiological activity.