An article have discussed the design of acupuncture studies (14). Problems in acupuncture RCTs also occurred in RCTs in other fields; such problems include insufficient sample size, high dropout rates, inadequate follow-up and randomization. However, acupuncture studies are so complex that specific methodological challenges are raised, such as sham interventions and blinding procedures.

Adequate randomization ensures that every subject has an equal opportunity to be allocated to the treatment or control groups, including suitable generation of an unpredictable randomized allocation sequence and sufficient allocation concealment (15). Larger estimates about study results were identified in non-randomized compared with randomized studies (16). Inadequate and unclear allocation concealment may overestimate treatment effects by 41 and 30%, respectively (17). Suitable methods used to generate the allocation sequences are random-number tables or computer random-number generators; by contrast, inadequate methods include alternate assignment and assignment by odd/even birth date or hospital number (15,17).

Allocation concealment is used to prevent bias from researchers that may influence the assignment of subjects until allocation. Methods such as opaque, sealed envelopes and central randomization were considered to be adequate in the process of allocation concealment (18). A previous review demonstrated that only 26% of included trials applied adequate sequence generation, and ~29% applied adequate allocation concealment methods (19). Therefore, adequate randomization methods should be implemented and should also be clearly reported in RCTs. We suggest that future large multicenter RCTs should apply central randomization, in order to reduce the bias of RCTs.

The blinding procedure is an essential part of RCTs, since it may reduce potential bias in the overall procedure (such as bias from the patients and evaluator). Previous studies (17,20) demonstrated that non-blinded trials may overestimate the treatment effects by as much as 17%. Investigators should clearly report the blinding procedure that was applied in their studies rather than describing the blinding procedure as double-blind (21).

Blinding procedures often include participant and therapist blinding. However, this is challenging in acupuncture RCTs, since the needle is manipulated by the acupuncturist and thus blinding of the acupuncturist is inappropriate (22). Blinding the patients and the evaluator is an alternative. In this process, a researcher may evaluate the blinding effect, not just report that the patients were blinded (23).

Researchers have implied that, unlike inexperienced subjects or subjects with limited acupuncture experience, experienced subjects may be able to distinguish the different appearance or sensation between real and sham acupuncture (24). For instance, a number of patients knew the exact sensation associated with acupuncture. Thus, subjects with different background knowledge and previous experience of acupuncture may induce a different result in the blinding procedure.

Other factors (24,25) also influence the blinding procedure, including the needle point (a needle in an insensitive point may have a better blinding effect), the visual impact of needling (if the subject observed the manipulation of the acupuncturist, the effect of the blinding procedure may be influenced), and the therapists (the needling differences between acupuncture therapists should be minimized).

The use of an appropriate control group was found to improve the reliability of acupuncture studies (26). However, selecting an ideal control group for acupuncture research is a challenging methodological issue (27). Questions about acupuncture cannot be answered by a single control group, since acupuncture is a complex intervention. Therefore, an appropriate control group should be selected according to the specific questions being asked in the study (28). The common categories of control groups in acupuncture studies are: Waiting-list control (no-treatment control group); standard care (such as medications); and sham interventions, which attempt to manage placebo and other nonspecific effects. The placebo effect cannot be managed in the waiting-list control and standard care groups and thus attention has been focussed on the control groups of sham interventions.

The terms ‘sham acupuncture’ and ‘placebo needle’ have not been clearly defined in clinical trials. In current research, the two terms are confusing: In certain acupuncture studies, the two terms were used to describe the non-penetrating sham needle (29,30), whereas a review (31) described ‘placebo needles’ as non-penetrating sham needles and ‘sham acupuncture’ as needling at the wrong points or non-points.

Sham intervention is a complicated control which has been often applied as a control for nonspecific or placebo effects, in order to investigate whether verum acupuncture is more effective than sham intervention (28). The benefit of this control is that the blinding procedure may be more effective compared with other methods, and the disadvantages are the potential nonspecific needling effects (28). Therefore, partly because of the non-specific needling effects, the differences in clinical effects between the real and sham acupuncture groups are unclear.

If a sham intervention was an absolutely inert procedure and its appearance was difficult to distinguish from the real acupuncture group, it could be regarded as the optimum condition in acupuncture RCTs for determining the effects of acupuncture interventions (32).

Several sham interventions, including superficial needling, needling in irrelevant true points, needling in non-points, non-penetrating sham needles, non-needling sham interventions and minimal acupuncture, were often used in acupuncture RCTs (28). Each type of sham intervention has certain advantages and disadvantages. All the aforementioned methods of sham intervention cannot be regarded as absolutely inert interventions, as they may induce certain therapeutic effects (33–36). In addition, the various manipulations and insert directions were not allowed by the device (37) in a non-penetrating sham needle.

This sham intervention-induced effect may result in the difference between verum and sham group not being significant (38). In addition, these controls were often investigated in inexperienced subjects (29,39), so the blinding of this method should be considered when the subjects have had extensive experience in acupuncture.

Designing an appropriate control requires understanding the traditional acupuncture theory and the modern mechanism of acupuncture, and should match the specific objective of the study. To date, no sham intervention can be regarded as an ideal control and this question should be further investigated by methodologists.

Positive results were often produced from studies that applied a non-acupuncture control group, and the results were usually negative in studies that used sham acupuncture or mock transcutaneous electrical nerve stimulation (40). Vickers et al (41) found that potent placebo effects were contributing to the total effects of acupuncture. Generally, placebo implied an inert substance or procedure and a placebo effect was an essential component of all therapeutic interventions (42). Intervention should be divided into specific and placebo (non-specific) aspects in order to design a placebo-controlled procedure (43).

Specific elements was considered to be responsible for the specific therapeutic actions, and other aspects that influenced the effects were placebo elements, such as the patient expectations and the communication between therapists and subjects (43). A previous study considered that the placebo effect was derived from factors included in a social, cultural and clinical context, rather than from the sham interventions per se (44). Other studies (42,45,46) demonstrated that a placebo effect was a complexity resulting from psychological, social and cultural background factors, as well as real physiological responses. Notably, one study (47) demonstrated that the placebo effect had a ‘dose-dependent response’: Although the intervention was sham, if people received more care, their outcome would be improved.

The psychological mechanisms (42) of the placebo effect include many aspects associated with the procedure of acupuncture, including expectations, memory, motivation, cognitive and conditioning mechanisms.

If the major mechanisms of placebo responsiveness could be identified, strategies aimed at minimizing placebo effects could be developed, so that the true effect of an intervention can be revealed in trials (48). The factors influencing placebo effects in acupuncture studies included frequent communication between acupuncturist and subject, the ‘ritual’ of acupuncture, the expectations of patients and certain potential physiological effects caused by the procedures of acupuncture (49).

Systematic research was conducted to explore how to minimize the placebo effect. Researchers successfully reduced the placebo effect by decreasing the interaction between therapists and participants (47). Other studies demonstrated that strong placebo effects could significantly influence the outcome of subjective symptoms and this significant influence may be confined to subjective outcomes (50). Therefore, in future experiments, multiple outcome measurements should be selected, including subjective and objective outcomes.

Although the placebo effect is complex, research demonstrated that the placebo effect from sham acupuncture was stronger compared with the effect from a placebo pill (50). These results may be due to more complex medical interventions having a higher placebo effect than medication (51).

Given the potential physiological effects of sham interventions and the powerful placebo effects, the results of acupuncture RCTs may be undervalued in numerous studies (52). Although sham acupuncture exerted a powerful placebo response and uncertain physiological mechanisms, evidence also demonstrated a specific effect from acupuncture which was better than the ‘placebo effect’ (53).

One of the least investigated areas of acupuncture research may be acupuncture administration. Researchers have to increase the effectiveness of drugs to prove that their effect is greater than placebo control; however, this has not been the case in acupuncture trials. An insufficient acupuncture treatment group may also provide potential bias regarding study outcomes. The results of acupuncture efficacy based on insufficient acupuncture treatment groups can be compared with formulating conclusions on the pharmaceutical efficacy of a drug with an inadequate dose (54). It is important to emphasize that acupuncture is not a straightforward needling intervention. The evaluation is therefore far more challenging than that of a pharmacological agent.

When the results of two studies about acupuncture on hypertension are compared (10,55), the data may demonstrate that the effects of active or verum acupuncture groups are different. Macklin et al (10) revealed that active acupuncture groups provided no greater benefit than sham acupuncture, while Flachskampf et al (55) reported a clear-cut effect on blood pressure in the active treatment group. The setting of acupuncture administration was different in the two trials. The acupuncture procedure in the trial by Flachskampf et al (55) was that patients received 22 sessions in 6 weeks, with 5 sessions/week conducted in the first 2 weeks and 3 sessions/week conducted in the following 4 weeks. Each session lasted 30 min after the needle was inserted into the points. The needling points were selected according to the Chinese type of hypertension in Traditional Chinese Medicine theory, and the acupuncture details, including angle, depth and type of manipulation of needling, were consistent with typical prescriptions. By contrast, in the trial that obtained a negative result (10), the principles of diagnoses and treatment were consistent with the description of Cheng (56). The treatment procedure consisted of ≤12 sessions, generally twice-weekly, over 6–8 weeks with each session providing a 30-min treatment. Therefore, it can be concluded that the difference in the acupuncture administration setting may induce certain potential bias.

Although there were a variety of factors that may account for the improvement in the acupuncture group, two principal features should be emphasized: Acupoint specificity and manipulation of acupuncture.