Verum versus sham acupuncture for the treatment of migraine.

Dear Editor, We read with great interest the recently published article entitled “Verum versus sham manual acupuncture for migraine: a systematic review of randomised controlled trials” by Yang et al. The authors examined the efficacy of verum acupuncture compared with sham acupuncture for the treatment of migraine. However, in our opinion, they should have exercised more caution when drawing conclusions from the results of their study, which had several drawbacks, as follows. Firstly, as the database search only included articles published up to 31 May 2014, several more recently published relevant studies will unfortunately have been missed. Secondly, the quality of the studies included was likely overestimated. The authors stated that they judged the risk of performance bias as low in all of the studies, including that related to the use of sham interventions. However, we believe that this interpretation is incorrect, because it is also important in studies that include patients undergoing sham acupuncture as the control group to test whether or not blinding was successful. Only one of the 10 included primary trials clearly reported that patients were successfully blinded. One report stated that patients were not in fact blind and none of the others reported blinding success. In addition, although the authors reported having assessed the risk of selection bias, their explanations were not convincing. They stated “although none of the study protocols was published a priori, risk of selective reporting bias was deemed to be low” but did not explain how they came to regard the other individual bias domains as low risk. Thirdly, the information provided regarding the questionnaire scores reported in the study by Li and Sun, data from which was used in the meta-analysis, appears to be incorrect. Having reviewed the dissertation detailing the primary trial by Li and Sun, we noticed that the data presented were actually those obtained at 1 month post-treatment, and not 3 months as stated. In conclusion, for the aforementioned reasons, the published conclusions of Yang et al should be interpreted with great caution.