According to some authors, acupuncture carry out its analgesic effects by interacting with the hypothalamic-pituitary-adrenal axis and the endogenous opioid system, known to be important mediators of the stress response to pain (5). Several studies demonstrated that acupuncture activate the release of opioid in the central nervous system (6–10). Gao et al. (6) report that “Release of these peptides corresponds to long-lasting activation of ascending sensory tracks, thereby relieving an array of pain conditions.” Furthermore, using a migraine rat model, researchers observed that the expression of CGRP in the trigeminal ganglion, which is considered a migraine trigger factor (11), could be depressed by the electro-acupuncture treatment (12). In addition, acupuncture was demonstrated to restore the descending pain modulatory system, impaired in patients affected by migraine, by decreasing the resting state of functional connectivity between the rostral anterior cingulate cortex/medial prefrontal cortex and the periaqueductal gray, which is correlated with the intensity of pain during the attacks (13).

To understand the mechanisms of action of acupuncture, as well as its local, remote, and long-term effects, a key may be the reaction to needles insertion, called “de qi” and widely judged as an important factor affecting the therapeutic effect of acupuncture. “De qi” includes both a typical needling sensation, sensed by the patient, and a “needle grasp” noticed by the provider: the acupuncturist perceives tearing and augmented resistance to additional movement of the needle. Langevin hypothesized that “needle grasp is due to mechanical coupling between the needle and connective tissue with winding of tissue around the needle during needle rotation and needle manipulation transmits a mechanical signal to connective tissue cells via mechano-transduction” (14).

In conclusion, although the mechanism of action of acupuncture doesn't have yet a definite explanation, acupuncture has a scientific basis built on a series of studies that have been conducted over the last decades: there are neurobiological models that could elucidate how acupuncture obtain its therapeutic effect in several clinical settings.

The efficacy of acupuncture for migraine has been confirmed by several clinical trials and a series of systematic reviews, including a Cochrane Review (15, 16). Consequently, acupuncture is becoming a more accepted form of integrative medicine in the Western countries for the prevention and treatment of migraine (17) and is administered for migraine prophylaxis and chronic pain treatment all over the world (15, 18): migraine and other headaches were the primary indications for acupuncture treatment in 9.9% of US patients (17).

In 2009, a Cochrane review concluded that “there is consistent evidence that acupuncture provides additional benefit to treatment of acute migraine attacks only or to routine care. Available studies suggested that acupuncture is at least as effective as, or possibly more effective than, prophylactic drug treatment, and has fewer adverse effects” (15); after 7 years, the update of the same Cochrane review suggested the prevalence of headaches is reduced by combining acupuncture with usual care; it also suggested that there is an effect over sham and that acupuncture may be at least as effective as treatment with prophylactic drugs and it is associated with fewer adverse effects (16).

Since 2016, several new studies have appeared to support the efficacy of acupuncture in migraine prevention (16); acupuncture can be recognized as a valid treatment in Western countries if its specific effects are clearly understood.

The objective of this systematic review was to assess the efficacy and safety of acupuncture for the prophylaxis of episodic or chronic migraine in adult patients when compared to pharmacological treatment.

We included randomized-controlled trials published in western languages that compared any treatment involving needle insertion (with or without manual or electrical stimulation) at acupuncture points, pain points or trigger points, described as acupuncture with any pharmacological prophylaxis in adult (≥18 years) participants with chronic and episodic migraine with or without aura according to the criteria of the International Headache Society (1). Studies on patients with cluster headache or tension-type headache were excluded. We also excluded studies that evaluated acupuncture at specific “microsystems” (e.g., scalp or ear acupuncture), although we included trials using micro-system points in addition to body acupuncture; evaluated other methods of stimulating acupuncture points without needle insertion, for example, acupressure, laser stimulation, or transcutaneous electrical stimulation; injected fluids at acupuncture or trigger points.

Primary outcomes were:

Secondary outcomes assessed were:

Cochrane Database of Systematic Reviews (CENTRAL), Embase, MEDLINE, and ClinicalTrial.gov were searched for eligible studies. Literature search was performed using free text and Mesh terms from inception up to 14 May 2020 without language restriction. The detailed search strategy is reported in the Supplementary Material 1.

Two authors independently screened articles retrieved via the search strategy from the title and abstracts. Potentially relevant studies were acquired in full text and assessed for final inclusion independently by two authors. Any disagreement was discussed with a third author. Two review authors independently extracted data from the studies We extracted the following information: number and characteristics of participants: mean age, % female, duration of disease in years, chronic vs. episodic migraine; details of acupuncture treatments: number of sessions, number of acupoints, achievement of de-chi (an irradiating feeling considered to indicate effective needling), duration of treatment in weeks; type of drugs received by participant of the control group, length of follow-up after the end of treatment, types of outcomes assessed, country where the study was conducted.

Two authors independently assessed risk of bias according to the criteria set out in the Cochrane Handbook for Systematic Reviews of Interventions (24). The following criteria were considered: sequence generation and allocation concealment (selection bias), blinding of participants and providers (performance bias), blinding of outcome assessors (detection bias), incomplete outcome data (attrition bias), and selective outcome reporting (reporting bias). Disagreement between reviewers was resolved by discussion.

We analyzed dichotomous outcomes by calculating the risk ratio (RR) for each trial with the uncertainty in each result being expressed with a 95% confidence interval (CI). We analyzed continuous outcomes by calculating the mean difference (MD) with 95% CI when the studies used the same instrument for assessing the outcome. We used the standardized mean difference (SMD) when the studies used different instruments. We interpreted SMD values with the classification proposed by Cohen et al. (25) where an effect size of 0.2 means a small effect, 0.5 means a medium effect, 0.8 means a large effect.

As we supposed a certain degree of heterogeneity among studies, due to treatment schedules, way in assessing response criteria, risk of bias and other factors which may have affected direction and magnitude of treatment effect, we pooled data used the random effect model for each outcome.

Seeking statistical heterogeneity among studies, the Cochrane Q-test was performed, with a significant threshold of alpha = 0.1 and inconsistency among studies was quantified by the I-squared statistic (24); an I square >70% was judged a significant heterogeneity.

Results are depicted in all figures as conventional meta-analysis forest plots. RevMan 5.3 was used for producing forest plot figures (26).

We planned to use visual inspection of funnel plots (plots of the effect estimate from each study against the sample size or effect standard error) to indicate possible publication bias if there were at least 10 studies included in the meta-analysis.

Although the STRICTA (Standards for Reporting Interventions in Controlled Trials of Acupuncture) recommendations describe the components of acupuncture procedures (27) better outcomes appear to be associated with a greater numbers of needles and treatment sessions (28) and on the other hand, an insufficient dose of acupuncture may be an obstacle to good patient care (29).

When starting this study, the vast heterogeneity among the contributions available in the literature has soon become evident. The extremely different ways of administering acupuncture made it almost impossible to compare the outcomes of the trials. Consequently, we have decided to establish standard criteria for comparison of data. By introducing the concept of adequate dose of acupuncture already expressed by other authors (30), we have considered the following three parameters:

The “de qi” response, that is to say the sensation from needling experienced by the patient, may be reported as numbness (A-beta fiber activation) or as aching, dull, heavy, and warm sensation (A-delta or C fiber activation) (31).

The concept of dose-intensity has thus been introduced and used to group the studies according to the intensity of acupuncture based on the following criteria:

Acupuncture was judged as at low intensity of only one criterion was met; on medium intensity if two criteria were met; high intensity if all the three criteria were met. Subgroup analyses was performed for intensity of acupuncture.

We assessed the overall quality of the evidence for the primary outcomes using the five GRADE domains (study limitations, consistency of effect, imprecision, indirectness, and publication bias) according to the GRADE approach (32).

Based on the above domains, the GRADE system uses the following criteria to grade the evidence:

High: we are very confident that the true effect lies close to that of the estimate of the effect.

Moderate: we are moderately confident in the effect estimate: the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

Low: our confidence in the effect estimate is limited: the true effect may be substantially different from the estimate of the effect.

Very low: we have very little confidence in the effect estimate: the true effect is likely to be substantially different from the estimate of effect.

The existing evidence was summarized in a “Summary of Findings” table that provides key information about the magnitudes of relative and absolute effects of the interventions, the amount of available evidence and the certainty of available evidence (33).

Six studies were judged at low risk of selection bias because both the methods for random sequence generation and allocation concealment was appropriate; one study (44) followed ad adequate method for random sequence generation but did not provide information about concealment of allocation. The remaining two studies (31, 40) were judged at unclear risk for selection bias because they did not provide any information about methods followed to generate the random sequence and to conceal the allocation. All but three studies were judged at high risk of both performance and detection bias because they were open label; Two studies (40, 42) used the double-blind double dummy approach and was judged at low risk for both the domains. Three studies were judged at high of attrition bias (38, 40, 41) because of the high number of subjects who dropped out from studies and no attempt to perform an intention to treat analysis. One study (31) did not provide information about subjects dropped out from each group. The study protocol was available only for two studies (31, 42) and the outcomes reported in the final publication coincided with the outcomes listed in the protocol; for all the remaining studies the protocol was not available and they were judged at unclear risk of selective outcome reporting (Figure 2).

The two most recent SRs that addressed this topic were Linde et al. (16) and Zhang et al. (18).

Linde et al. is a Cochrane systematic review, well-conducted, but not updated since 2016, and Zhang et al. did not pool data across studies.

The strength of our review relies in a comprehensive bibliographic search on several databases without time restriction and in the rigor of the methodology that followed the highest standards as recommended by Cochrane (24).

Our review has some limitations. We limited our inclusion criteria to studies published in western languages due to our inability to translate studies published in Chinese or other eastern languages. Given the widespread use of acupuncture in Eastern countries and particularly in China, we probably missed some studies that made our comparison of interest. An overview of systematic reviews recently published described 14 systematic reviews that assess the efficacy of acupuncture against sham acupuncture or pharmacologic prophylaxis; we retrieved the full text of such reviews to look for trials which we could have missed. Unfortunately, eight of these reviews were written in Chinese and included primary studies written in Chinese as well so we were unable to evaluate the included studies.

Furthermore, a limitation of some of the included studies is the short duration of treatment (4 weeks) and the short duration of follow up, as the effect of the medication might not be developed yet and acupuncture treatment usually last for several months. Finally, we were unable to visually inspect funnel plot for the presence of possible publication bias because if <10 studies are included in meta-analysis, the funnel plot in considered uninformative (24).

The major flaws of most retrieved studies were the lack of blinding, that poses the efficacy results at high risk of performance and detection bias. Studies that adopt a double blind, double dummy design could provide unbiased estimates of efficacy results, though, due to the nature of the intervention, a double-blind trial is hard to be conducted.