Search strategy

Two library staff (SCL and BC) developed a search strategy for retrieving controlled trials on the efficacy of acupuncture for smoking cessation. A comprehensive literature search was conducted in two English and five Chinese databases: PubMed, the Cochrane library, China National Knowledge Infrastructure (CNKI), Chinese Biomedical Database (SinoMed), Wanfang, VIP Database and TCM online. These databases were searched from their inceptions to February 2017. In order to capture all relevant articles, search terms were intentionally broad (Appendix 1).

Inclusion/exclusion criteria

Study selection and data extraction

Two review authors (JHW and ALD) independently selected studies based on the inclusion/exclusion criteria. Data were extracted by two authors (JHW and ALD), independently, from the included studies using a self-developed electronic data extraction form. Data extracted included study location, sample size, participant characteristics, intervention details, outcome measures, and information on risk of bias. If a study had multiple intervention arms, data could be reused in different subgroup comparisons. Any disagreements were resolved by discussion with JPL.

Assessment of risk of bias in included studies

The methodological quality of the included trials was independently analyzed in detail by two authors in order to assess the strength of the evidence. Any disagreements were resolved by discussion with JPL. The Cochrane Risk of Bias tool for systematic reviews of interventions was used to evaluate the methodological quality of individual studies (Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011] http://handbook.cochrane.org.). The risk of bias in each trial was rated as low, high, or unclear, based on the following items: random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessors (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias) and other bias.

Data analysis

The measures of effectiveness were estimated by mean difference (MD) with 95% confidence interval (CI) for continuous data. We calculated the risk ratio (RR) with 95% CI for dichotomous data. Heterogeneity, which was quantitatively assessed using the Cochran Q and I² statistics to evaluate whether the pooled studies represent a homogeneous distribution of effect sizes, with I²≥75% considered as significant. RevMan 5.3.0 software (Copenhagen: The Nordic Cochrane Centre, the Cochrane Collaboration, 2014) was used for the data analyses. We performed random-effects meta-analysis for pooling data with significant heterogeneity and fixed-effects meta-analysis for pooling data without heterogeneity. We further performed subgroup analysis according to the type of experimental and control interventions if the data were available. Results were considered as statistically significant for p<0.05. We assessed publication bias using funnel plots in accordance with the Cochrane Handbook. TSA were performed by TSA 0.9.5.10 Beta software (Copenhagen: The Copenhagen Trial Unit, Center for Clinical Intervention Research, 2017) to calculate the required sample size and to detect the robustness of the result.

Ethics approval and consent to participate

The present study did not require approval of an ethics committee or consent to participate, since the extracted data were obtained from publicly available publications.

Description of studies

The literature screening flow is shown in Figure 1. Twenty-four studies that qualified for inclusion were identified in the review, yielding 27 comparisons in the intervention-level meta-analysis. Of the 24 included trials, 11 were published in Chinese^44-54, and the remainder in English^55-67. Ten were conducted in mainland China^44,46-54, 3 in France^58-60, while Taiwan^66,67, USA^56,63, UK^64,65, and Korea^57,62 each carried out 2 trials. Italy⁵⁵, Norway⁶¹ and Malta⁵⁴ each conducted 1 trial. Research in this area has been conducted over a long period since 1982⁶³. Three trials were published in the 1980s and 1990s each, seven in the 2000s and eleven in the 2010s. Nineteen studies were published in journals and five as dissertations.

Figure 1

Literature screening flow

A total of 3984 participants were included in this review, and the characteristics of 24 included trials are summarized in Table 1.

Experimental interventions

All studies in the review used a traditional approach to acupuncture with needles inserted into acupoints identified as specific for smoking cessation⁵⁴ (Fei Shu – BL13, Shen Men – HT7, etc.). The form of acupuncture, however, varied greatly. Four studies^44,50,57,62 used body acupuncture, two studies^63,66 employed auricular acupuncture, one study⁵³ adopted head acupuncture, one study⁴⁶ employed wrist-ankle acupuncture alone, two studies^56,61 used body and ear acupuncture, one study⁵⁸ adopted head and eye acupuncture, and one study⁶⁰ used body, ear and head acupuncture. Five studies^{48,49,51,52,54} combined body acupuncture and auricular acupressure, one study⁴⁷ applied body acupuncture plus moxibustion, one study⁵⁹ used body acupuncture plus nicotine gum. Additional current devices (such as electricity, seeds) were adopted to enhance the stimulation of acupuncture. One study⁶⁵ investigated body electroacupuncture, two studies applied body electroacupuncture plus auricular acupressure with seeds⁶⁴ or magnetic beads⁴⁵, and one study⁶⁷ used auricular electroacupuncture plus auricular acupressure with embedding seeds. One study⁵⁵ investigated TCM intervention consisting of pharmacopuncture, plum-blossom needling and auriculotherapy.

Control interventions

Nine studies^51,57,61-67 used non-acupuncture points or points considered to have no specific effect on smoking cessation as a sham control intervention. NRT of known effect was adopted as a control intervention in four studies^44,48,50,54. One study had a no-treatment control⁵³. Five studies^45-47,49,52 compared different acupuncture therapies with each other.

Nineteen studies had two arms. Five studies had more than one control group and therefore qualified for more than one comparison. One study had three arms: acupuncture, nicotine gum and no intervention⁵⁸. The data of acupuncture compared with nicotine gum or no intervention were analyzed in separate pooled groups. Another study with three arms compared true TCM with true TCM combined with counselling or sham TCM combined with counseling, which were handled as two comparisons in different pooled analyses⁵⁵. Two studies had four arms^59,60 and one had three arms⁵⁶, but only one comparison with data that could be pooled analyzed in every study was adopted in the meta-analysis.

Risk of bias assessment

The overall quality of evidence was low for all outcomes. The risk of bias was evaluated as unclear in most studies, except for selective reporting, because of lack of detail in the reports. The assessments of risk of bias for included studies are given in Figures 2 and 3. Eleven studies (11/24) described the random sequence generation and were judged as low risk of bias. Only two trials (2/24) reported allocation concealment while detailed information was deficient in the other 22 trials. There were 7 trials (7/24) that described blinding of participants and personnel clearly being judged to be ‘Low’ in this domain, and 11 trials (11/24) were judged to be ‘High’ because of the huge differences of interventions between groups. The majority of trials (21/24) provided no details related to blinding of outcome assessment, except for two (2/24) judged to be ‘Low’ and one (1/24) judged to be ‘High’. We considered 21 trials (21/24) as low risk of bias for incomplete outcome data according to the reports of the drop-outs or intention-to-treat analysis. Selective reporting was judged by the consistence between the outcome measures described in the method section and the actual outcomes in the result section, due to the deficiency of protocol information of all trials. All trials were assessed as low risk of bias for selective reporting.

Figure 2

Risk of bias summary

Figure 3

Risk of bias graph

Effects of interventions

Based on the type of experimental and control interventions, we conducted meta-analyses for 9 different comparisons. Every outcome was reported in three subgroups: short-term, mid-term and long-term follow-up, where data were available (Table 2).

Funnel plot analysis

Since there were less than five studies in all metaanalyses, we did not use funnel plots to assess publication bias. It is possible that the results of this review are at risk of publication bias, because we identified a trial registered in Chinese Clinical Trial Registry (ChiCTR-TRC-13003544) that had not been published since completion.

Trial sequential analysis

TSA⁶⁸ was conducted to estimate the required sample size for a systematic review based on primary outcome of different comparisons and to detect the robustness of the result. The estimate of required information size (RIS) was based on type I error (α=5%), type II error (β=20%), relative risk reduction (RRR=20.0%) and cessation rate in control group. The TSA results are given in Appendix 2.

TSA illustrated that the cumulative Z-curve (blue curve) of acupuncture versus acupuncture combined with counselling/educational smoking cessation program/moxibustion in the long-term going across the traditional boundary of 5% significance (green curve) and crossing the monitoring boundaries (red inward sloping curves). This means that although the cumulative sample size did not meet expectations, no more trials were needed and a positive conclusion was reached in advance. By adjusting the random error, we can conclude that acupuncture combined with counselling/educational smoking cessation program/ moxibustion results in a higher rate of long-term abstinence rate than acupuncture alone. However, the cumulative Z-curve (blue curve) of other comparisons or follow-up terms whether or not across the traditional boundary of 5% significance (green curves) did not cross the monitoring boundaries (red inward sloping curves) and RIS boundary (red vertical line), which was needed to obtain firm evidence controlling for the risk of random error. TSA of these included trials suggests that high quality RCTs are required to confirm possible intervention effects.

Adverse events

Of the included 24 trials, there were 13 trials^{45-47,49-53,58-61,67} (13/24; 54.17%) that did not give any information on adverse events. And no serious adverse events occurred in the remaining 11 trials (11/24; 45.83%). Four trials^54,57,62,64 (4/24; 16.67%) reported that no adverse events occurred. Another four trials^48,55,56,63 (4/24; 16.67%) reported infrequent minor bleeding or bruising upon needle removal. Non-serious adverse effects including fainting, pain, weeping, anorexia, or headache were observed in association with electroacupuncture in one trial⁶⁵ (1/24; 4.17%). One trial⁶⁶ (1/24; 4.17%) reported side effects during the period of auricular acupuncture therapy including tenderness sensation, feeling, dizziness, minor bleeding, and nausea sensation. In one trial⁴⁴ (1/24; 4.17%) with acupuncture, a participant experienced dizziness as he had not eaten breakfast.

Main findings

The goal of this systematic review was to evaluate the evidence regarding the effect of transcutaneous acupuncture on smoking cessation. We evaluated 24 RCTs with 3984 participants published from 1982 to February 2017. These studies varied considerably with regard to relevant participant characteristics, interventions, and outcome measures. The quality of evidence was low due to the overall high or unclear risk of bias. We classified comparisons based on the types of experimental and control interventions, which possibly explains why no significant statistical heterogeneity existed, except for one pooled analysis comparing acupuncture in combination with auricular acupressure to sham acupuncture plus sham auricular acupressure on the outcome of daily cigarette consumption (heterogeneity of 95%).

From the results of meta-analysis and TSA, the conclusion can be drawn that acupuncture plus counselling/education/moxibustion is more effective for abstinence rate at long-term than acupuncture alone. Apart from this, on the basis of the available evidence, the TSA indicates that it is not possible to draw any firm conclusions with regard to the other comparisons. However, grouping studies by the nature of the comparisons and intervention types yielded the following trends: 1) acupuncture may be more effective than no intervention/waiting list at short- and long-term follow-up; 2) real acupuncture does not appear to be superior to sham acupuncture in every stage of follow-up; 3) acupuncture shows a similar effect to NRT on smoking cessation in this review for whatever periods and indicators; 4) secondary outcomes are consistent with the primary outcome, sometimes secondary outcomes vary more sensitively than the primary outcome; and 5) only one trial⁴⁸ with 60 participants explored the cost-effectiveness of the interventions between two groups. This study demonstrated that acupuncture plus auricular acupressure with beads was more economical than nicotine paste.

Results between different comparisons, follow-up periods and outcome measures were not always consistent. With regard to sham intervention, most outcome measures showed no significant differences between real acupuncture and sham acupuncture, whereas acupuncture in combination with auricular acupressure was superior to sham acupuncture in combination with sham auricular acupressure This may point to a possible synergistic action between acupuncture and auricular acupuncture. Involving follow-up periods, acupuncture plus counselling/ educational smoking cessation program/moxibustion was more effective than acupuncture on cessation rate for the short-term and long-term, but equal for the mid-term. Referring to outcome measures, in the comparison between acupuncture and sham acupuncture, some outcome indicators showed no significant differences including cessation rate, NWS, BDI, BAI, and craving for cigarette, while others showed significant differences including exhaled CO level, serum cotinine, and daily cigarette consumption. This inconsistency may illustrate the complex and changing process of smoking cessation and interventions, as well as the relatively exploratory nature of these secondary outcome measures.

Limitations of the systematic review

There were several limitations in this systematic review. Firstly, the methodological quality of the evidence was generally poor as the lack of sufficient information about the random sequence generation, allocation concealment, and blinding in most included studies. Secondly, we only included trials that were published in Chinese or English and this could have excluded potentially important studies. In addition, it is possible that the results of this review are at risk of publication bias. There were no funnel plots in this review because there were less than 10 eligible studies per comparison. Finally, participants who dropped out or lost follow-up due to moving or changing contact information were not counted in the efficacy analysis, which may have affected the results.

Comparison with other systematic reviews

Several systematic reviews have been performed to evaluate the efficacy of acupoint stimulation for smoking cessation. However, none of them evaluated the efficacy of transdermal acupuncture by needles specifically. This review focused on penetrating acupuncture and we evaluated all outcome measures of included trials for three follow-up periods.

The evidence of this paper shows that secondary outcomes are not always consistent with the primary outcome, sometimes secondary outcomes vary more sensitively than the primary outcome. That was a warning to pay attention not only to abstinence rate but also to the secondary outcome measures, in order to evaluate the efficacy of smoking cessation measures earlier and more fully. Acupuncture appears to be equivalent to NRT for smoking cessation for whatever periods and indicators. These findings have not been reported in previous systematic reviews.

Acupuncture may be more effective than no intervention/waiting list in the short-term and long-term, which is inconsistent with a Cochrane review in 2014³⁷. Our findings suggest that real acupuncture appears not to be superior to sham acupuncture in every stage of follow-up. This is consistent with the result of previous systemic reviews^38,39.

One meta-analysis⁶⁹ indicated that multi-modality treatment, especially acupuncture combined with smoking cessation education or other interventions, can help smokers to quit smoking during treatment and to avoid relapse after treatment. This conclusion is consistent with our results.

Future research

In order to confirm the effect of acupuncture on smoking cessation, future trials should improve the quality of methodology, enlarge the sample size, extend the follow-up period, and increase secondary outcome indicators especially laboratory indicators. The current evidence supports the notion that further studies should focus on acupuncture combined with other TCM therapies and other currently recommended interventions as part of a complex intervention⁷⁰, which may help smokers to quit or avoid relapse more effectively. Acupuncture may be satisfactory for smoking cessation, if we can give smokers the appropriate therapy package according to the stage of smoking cessation⁷¹ and the specific factors that promote or hinder quitting.