Safety of Acupuncture: Overview of Systematic Reviews

Acupuncture is increasingly used worldwide. It is becoming more accepted by both patients and healthcare providers. However, the current understanding of its adverse events (AEs) is fragmented. We conducted this overview to collect all systematic reviews (SRs) on the AEs of acupuncture and related therapies. MEDLINE and EMBASE were searched from inception to December 2015. Methodological quality of included reviews was assessed with a validated instrument. Evidence was narratively reported. Seventeen SRs covering various types of acupuncture were included. Methodological quality of the reviews was overall mediocre. Four major categories of AEs were identified, which are organ or tissue injuries (13 reviews, median: 36 cases, median deaths: 4), infections (11 reviews, median: 17 cases, median deaths: 0.5), local AEs or reactions (12 reviews, median: 8.5 cases, no deaths were reported), and other complications such as dizziness or syncope (11 reviews, median: 21 cases, no deaths were reported). Minor and serious AEs can occur during the use of acupuncture and related modalities, contrary to the common impression that acupuncture is harmless. Serious AEs are rare, but need significant attention as mortality can be associated with them. Referrals should consider acupuncturists’ training credibility, and patient safety should be a core part of acupuncture education.

Methodological quality of included reviews. Fifteen (88.2%) reviews conducted a comprehensive literature search. No review provided an a priori design to their reviews via protocol publication. Thirteen (76.5%) review provided characteristics of their included studies. No review provided lists of both included and excluded studies. Eleven (64.7%) reviews reported the presence or absence of conflict of interests on the review itself. Only one (5.9%) review assessed the scientific quality of the included primary studies. The use of appropriate statistical methods and assessment on the likelihood of publication bias were not applicable as no meta-analysis was conducted amongst all included reviews. See Table 2 for full details.

Adverse Events.
Major results are narratively presented in this section and detailed information about the cases, including age and sex, reason for treatment, training background of practitioner, site of treatment, and follow-up time and outcome can be found in Table 3. A full list of results can be found in the Appendix 2.

Discussion
This overview provided a comprehensive summary of all the adverse events and complications associated with acupuncture and related therapies that have been reported to date in published systematic reviews, with the majority coming from case reports, case series, and randomized controlled trials across 17 publications. The number of included primary studies in these publications ranged from 9 to 167. Amongst these primary studies, the number of included patients ranging from 21 to 111,692. In general, the results show that both minor and serious adverse events can occur from the use of acupuncture. Incidence rates, related confidence intervals (CIs), and p values could not be calculated because many adverse events came from case reports and many of the reviews did not include full details about the number of participants in their included studies. However, all the reviews have suggested that adverse events are rare and often minor. Although serious complications were rare, they require significant attention as mortalities are associated with these adverse events. There was insufficient data to determine which body sites or whether patient predispositions were associated with these events, but it is clear that patients can be at great risk. Practitioners should pay ample attention to risk stratifying patients based on their medical history and other relevant characteristics. Other Norheim, 1996

Vasovagal reaction
Continued potential areas of improvement include enforcing stricter sterile needle practices, improving patient education about common and/or serious risks, and enhancing practitioner recognition of acute complications. Better communication should exist between physicians treating complications and the practitioner that administered the acupuncture, so that practitioners can become more cognizant of issues that can arise from their practice. The methodological quality of the included reviews was mediocre. The majority of the included reviews conducted a comprehensive literature search. Most reviews also included the characteristics of their included studies and stated any conflicts of interest. However, no studies provided an a priori design, and only one study 6 thoroughly assessed the scientific quality of included studies, which might be caused by lack of appropriate methodological quality assessment tools for case reports 23 . Regardless, the reviews provided a tantamount of information on the existence of adverse events in the literature across studies from around the world. A major limitation of the presented information was that no causality could be determined. In the reviews that commented on the scientific quality of included studies, concern was raised regarding the ability to ascertain that acupuncture resulted in the adverse outcome 8,10,11,14,22 . One review classified reported cases on a causality scale, and only a minority of adverse events were classified as certainly caused by acupuncture 19 . Future studies need to be more rigorous in their assessment of causality, and document their means of determining causality. Ideally, prospective cohort studies or randomized controlled trials should be reporting all the adverse events that occur during their investigations, as these provide the best evidence for causality. For rarer adverse events, case-control studies would be the ideal design. Another limitation is that a significant number of adverse events were not followed up. Although most complications were minor, practitioners and researchers should still follow up with the patient so that meaningful and definitive data can be derived. It is improper to assume that minor complications resolve spontaneously, or that major complications result in long-term implications. More attention needs to paid on the documentation and follow-up of all adverse events that occur during a study. Furthermore, a standardized template should be developed in the near future so that practitioners around the world can use it to track and report complications for research and clinical purposes.
Due to the diversity in study designs (e.g. case report, case series, case control studies, cohort studies, and clinical trials), populations, and data collection methods included in the identified systematic reviews, none of them conducted a meta-analysis to generate a pooled incidence rate with CI 24 . Although systematic reviews on adverse events are recommended to summarize evidence in a qualitative manner, a quantitative estimation of the upper limit of the 95% CI for the probability of the adverse events will help the clinicians to estimate what the worst-case scenario could be 24 . Future systematic reviews on this topic should use available statistical method 25 to provide such a quantitative estimation. For primary studies, case control studies is the preferred method give the rare occurrence of adverse events.
Another concern raised by most reviews was the issue of underreporting 10,11,13,15,16,[19][20][21] . Often, only medically interesting findings are reported as many case reports are published by the physicians treating them 16 . Minor, less significant adverse events are often not published. Some reviews found that no adverse events were reported by acupuncturists 12 , raising the question of whether some practitioners are even aware of complications in their patients. Many journals restrict the type of publications that are accepted, and thus publication bias can also limit the number of adverse events reported 20 . This is a systemic issue, and regulatory bodies around the world need to create a convenient platform for which practitioners providing acupuncture and physicians treating complications can report known adverse events and the surrounding circumstances which can help with epidemiological and clinical research. In addition to providing a comprehensive summary, this overview serves as an important step towards furthering the knowledge, safety, and application of acupuncture. The overview may inform practitioners around the world about and modify the way they practice acupuncture, given that many acupuncturists may not be fully aware of the full breadth and depth of risk their treatments can pose. Usage of acupuncture is increasing worldwide, and more acupuncturists are being trained to match the demand. Thus, the need to pay careful attention towards the risks of acupuncture is becoming increasingly paramount as more patients become subject to it.

Methods
Inclusion criteria. Any systematic review (SR) that summarized adverse effects of acupuncture and related therapies, including electro-acupuncture, cupping, moxibustion, laser acupuncture, indwelling needles, dry needling, bee venom acupuncture, acupuncture point injection and acupressure, were considered eligible for this overview. To be included, the SR must have a primary objective of identifying adverse events instead of investigating its treatment efficacy or effectiveness. We had no restriction for the type of patients included, as long as they received acupuncture or related therapies for the management of any diseases or symptoms. We did not set  any restriction on the control treatment as long as adverse effects of acupuncture were reported. However, SRs on adverse effects specifically caused by injected drugs through acupoint injections were excluded.
Literature search. MEDLINE and EMBASE were searched from their inception to December 2015.
Published search filters related to SR 26,27 and adverse effects 28 were used during the literature search, in addition to search keywords for acupuncture and related therapies. Details on search strategies as well as the retrieved results from the electronic databases could be found in Appendix 1.
Literature selection, data extraction and assessments of the methodological quality. Literature search was conducted by one researcher, and retrieved results were equally distributed to 4 pairs of trained research assistants. Each pair of assistants independently screened and evaluated the eligibility of citations that were assigned to them, and extracted data from the included reviews using a standardized, piloted template. The template was designed according to the requirement of the PRISMA harm checklist 29 . Disagreements were resolved via discussion and consensus within each pair. A senior researcher was consulted when disagreement was unresolvable.
Methodological quality of included SRs were evaluated with the validated Methodological Quality of Systematic Reviews (AMSTAR) 30 instrument by two researchers independently. It includes 11 items, with each item being assessed as yes, no, cannot answer, or not applicable based on information provided by the SRs. Disagreements between assessors were discussed to reach consensus. A third reviewer was consulted if necessary.
Data synthesis. Adverse effects of acupuncture and related therapies and their outcomes were narratively reported according to each adverse event. Protocol of this overview has been registered in PROSPERO (http:// www.crd.york.ac.uk/PROSPERO/printPDF.php?RecordID=43943&UserID=6569).