Overview of Systematic Reviews and Meta-Analyses of Acupuncture for Stroke

Stroke is the second most common cause of death worldwide, as well as the leading cause of adult disability [1,2]. Nevertheless, medicine has developed few effective interventions against stroke and stroke-related disorders such as poststroke motor dysfunction and poststroke dysphagia. In China and elsewhere in East Asia, acupuncture has been a primary medical intervention for stroke and stroke-related disorders over the past 1,000 years [3], and in that part of the world, it continues to be used as a standard complementary therapy after stroke. In fact, a nationwide survey of 1,095 physicians from 247 Chinese hospitals between 1993 and 1994 found that 66% of Chinese doctors routinely used acupuncture to treat stroke patients, and 63% believed it to be effective [4]. Another more recent survey in 2007 among 202 Chinese acute stroke patients showed that 63.14% would choose acupuncture as part of medical intervention [5]. In South Korea an investigation of 304 stroke-patient respondents between 2004 and 2005 found that 54% had used complementary and alternative therapies, of which 92% had used traditional Korean medicine, primarily acupuncture [6]. Physicians and patients in Western countries and emerging economies [7,8,9,10,11,12] have also begun to accept acupuncture as an alternative therapy for a variety of neurological disorders.

Traditional Chinese medicine theory believes that the very core of disease onset is disruption of qi, the body's ‘vital energies'. Qi flows through a network of channels, called ‘meridians', running along the surface of the human body. Inserting needles at certain acupoints located along the meridians restores qi flow, thereby influencing progression of the disease [3]. In classical manual acupuncture, needles are inserted into acupoints to achieve a ‘de qi' reaction, in which the patient perceives an ache or heaviness in the area surrounding the needle, while the acupuncturist performs a biomechanical movement (grasping the needle) that restores the qi flow at the site of insertion [13]. In addition to classical manual acupuncture, various modified acupuncture therapies have entered into widespread use, including scalp acupuncture [14], warm acupuncture, electro-acupuncture, auricular acupuncture [15] and ‘xinnaokaiqiao' acupuncture [16].

The technical simplicity, noninvasiveness and lack of adverse effects of acupuncture compared to Western pharmacological and surgical approaches have made it an attractive treatment option in the West and East alike. Since the 1970s, hundreds of trials have been carried out to test the efficacy of acupuncture to treat stroke. These trials have been performed using widely differing methodologies and clinically diverse populations. To help identify real effects amid noise due to confounders, numerous systematic reviews and meta-analyses have been carried out to examine the effects of acupuncture to treat different types of stroke and stroke-related disorders at different stages of disease. These systematic reviews and meta-analyses have reported sometimes confusing and even conflicting results, which may help explain why acupuncture is still widely regarded as an ‘alternative' therapy. In an effort to definitively identify under what conditions acupuncture can be an effective stroke treatment, we performed a systematic overview of systematic reviews and meta-analyses in the literature. Our secondary goal was to identify gaps and ambiguities in the literature as a guide for future research.

We included systematic reviews and meta-analyses of randomized controlled trials (RCTs) and quasi-RCTs that evaluated the efficacy of acupuncture to treat any type of stroke or stroke-related disorders, including poststroke dysphagia, depression, hiccups, urinary incontinence, hand-shoulder syndrome and shoulder pain. Systematic reviews and meta-analyses involving any acupuncture modality were included. Systematic reviews in which studies involved other alternative therapies or drug therapy were included if these same nonacupuncture therapies were given to both the intervention (acupuncture) arm and the control arm (no acupuncture). Reviews were excluded if they: (1) included controlled clinical trials; (2) compared the efficacy of different acupuncture therapies, such that there was no nonacupuncture control group, or (3) were nonsystematic narrative reviews or overviews.

Using our prespecified inclusion criteria, we comprehensively searched the following databases with no language restrictions: the Cochrane Library (Issue 5, 2013), Ovid MEDLINE (1966 to June 2013), Cumulative Index to Nursing and Allied Health Literature (CINAHL, 1982 to June 2013), Ovid EMBASE (1980 to June 2013), Allied and Complementary Medicine (AMED, 1985 to June 2013), Chinese Biological Medicine Database (CBM, 1981 to June 2013), and China National Knowledge Infrastructure (CNKI, 1996 to June 2013). The following search terms were used: ‘stroke', ‘cerebrovascular accident', ‘isch(a)emic stroke', ‘infarction', ‘apoplexy', ‘h(a)emorrhage', ‘acupuncture', ‘acupressure', ‘auricular', ‘systematic review' and ‘meta-analysis'.

In addition to these automated searches in databases, we manually searched four Chinese journals relevant to acupuncture (from 1980 to June 2013): Acupuncture Research, Chinese Acupuncture and Moxibustion, Journal of Clinical Acupuncture and Moxibustion, and Shanghai Journal of Acupuncture and Moxibustion.

We identified eligible reviews using a two-step selection process. First, two investigators (J.Z. and D.W.) independently screened the titles and abstracts to exclude reviews that obviously did not meet the prespecified criteria. Second, the same investigators read the full text of the potentially eligible reviews and discarded those failing to meet the inclusion criteria. Disagreements were resolved by discussion and consensus.

The same two authors extracted the following data from included reviews: countries of the included trials, number of trials included, patient demographic and clinical data, main outcomes and conclusions. The quality of the included reviews was assessed using the Overview Quality Assessment Questionnaires (OQAQ). This instrument evaluates the scientific quality of a systematic review based on the answers to nine questions, generating an overall quality score ranging from 1 to 7 [17]. A score of 6-7 indicates a review of ‘good' quality, 3-5 indicates a review of ‘moderate' quality, and 0-3 a review of ‘poor' quality.

Data from reviews were quoted in the form of standardized mean difference (SMD), weighted mean difference (WMD), odds ratio (OR) or relative risk (RR), depending on what the review authors reported. Whenever possible, meta-analysis results were also reported with 95% confidence intervals (CI).

Initially, 10,739 potentially relevant records were identified through electronic and manual searches. After screening titles and abstracts, 10,702 records were excluded and the remaining 37 reviews were read in full. Finally, 24 systematic reviews were included in this overview: 4 (16.7%) Cochrane systematic reviews and 20 (83.3%) non-Cochrane systematic reviews (fig. 1).

Table 1 summarizes the key findings of all systematic reviews. Among the 24 reviews, 18 (75%) were conducted in China, 4 (16.7%) in Korea, 2 (8.3%) in the UK, 1 (4.2%) in Canada and 1 (4.2%) in Sri Lanka. The 24 reviews included 366 RCTs and quasi-RCTs, with each systematic review comprising 3-72 RCTs. Three reviews (12.5%) analyzed acupuncture as an acute stroke intervention [18,19,20], 6 (25%) analyzed it as an approach to stroke rehabilitation [21,22,23,24,25,26], and the remaining 15 (62.5%) analyzed it as an intervention to treat various stroke-related disorders. Of these 15 reviews, 4 focused on poststroke motor dysfunction [27,28,29,30], 4 on poststroke dysphagia [31,32,33,34], 2 on poststroke depression [35,36], and 1 review on each of poststroke hiccup [37], poststroke urinary incontinence [38], shoulder-hand syndrome [29], shoulder pain [39] and apoplectic aphasia [40]. The OQAQ quality of 10 reviews was rated as ‘good', 6 were rated as ‘moderate' and 8 as ‘poor'.

Three reviews [18,19,20], comprising 27 RCTs and quasi-RCTs, examined the efficacy of acupuncture therapy as an intervention for acute stroke. All 3 reviews conducted meta-analyses. The Cochrane systematic review by Zhang et al. [20] investigated both acute ischemic and hemorrhagic stroke, while the two non-Cochrane systematic reviews focused on acute ischemic stroke.

Two reviews assessed the efficacy of manual and electronic acupuncture against conventional care as the control [19], or against sham acupuncture or no treatment as the control [20]. The primary outcome in both reviews was death or dependency/disability at the end of treatment/follow-up, and in both cases there was no significant difference between the acupuncture and control arms. Nevertheless, Zhang et al. [20] discovered a borderline significant trend favoring acupuncture, with slightly lower rates of mortality and dependency at the end of long-term follow-up (15-22 months). The secondary outcome in both reviews was an improvement in neurological deficit scores. Both studies found acupuncture to be associated with significantly greater improvements than the control treatment: SMD 1.17, 95% CI 0.30-2.04 [20] and WMD 3.49, 95% CI 2.00-4.99 [19].

In the third review, Wang et al. [18] meta-analyzed 8 RCTs and compared the efficacy of scalp acupuncture with conventional medication. Acupuncture was associated with significantly greater improvement in neurological deficit scores (WMD 3.89, 95% CI 2.43-5.36) as well as significantly higher rates of clinical efficacy (RR 1.23, 95% CI 1.11-1.37).

One Cochrane systematic review [24] and 5 non- Cochrane systematic reviews [21,22,23,25,26] assessed the efficacy of acupuncture for poststroke rehabilitation. All reviews focused on ischemic and hemorrhagic stroke, and the authors placed no restrictions on the disease stage. Five of these reviews assessed the efficacy of manual or electronic acupuncture against various control treatments including sham acupuncture, conventional rehabilitation, medication or Chinese herbs. Quantitative analyses were impossible because the included studies reported different outcomes, which spanned physical assessment, disability and motor assessments, and other validated stroke assessment scales. Nevertheless, three reviews tried to perform meta-analysis on subgroups to reduce the bias caused by heterogeneity [22,23,24]. The Cochrane review of Wu et al. [24] included subacute and chronic stroke patients and meta-analyzed changes in neurological deficit scores on the Chinese Stroke Recovery Scale. They found that acupuncture was associated with a much higher probability of a deficit reduction (OR 6.55, 95% CI 1.89-22.76, p = 0.04), but the heterogeneity was high (I² = 68%). Kong et al. [23] meta-analyzed subgroups who received acupuncture or sham acupuncture, and found no significant difference in either functionality during activities of daily living (ADL) or in changes in neurological deficit scores. Wu et al. [22] transformed the diverse outcome measurements of 38 trials into binary variable outcomes and pooled the data. They found acupuncture to be associated with significantly better outcomes than sham or no acupuncture (OR 4.33, 95% CI 3.09-6.08), leading them to conclude that acupuncture is likely to be effective for improving poststroke rehabilitation.

Among the 6 reviews assessing the efficacy of acupuncture for poststroke rehabilitation, only the review by Lee et al. [21] assessed the efficacy of scalp acupuncture. Those authors found that scalp acupuncture, when used as an adjunct to conventional treatment, was associated with higher rates of clinical efficacy and higher ADL functionality. Those authors concluded that the evidence for acupuncture as a useful part of poststroke rehabilitation was weak but positive.

Four reviews focused on poststroke motor dysfunction, and all of them performed meta-analyses. These reviews included studies on both hemorrhagic and ischemic stroke patients at different stages of disease [27,28,29,30]. Sze et al. [29] analyzed 14 RCTs that compared motor impairment and disability in patients treated with manual acupuncture alone, manual acupuncture with rehabilitation, or rehabilitation alone. A significant clinical benefit due to acupuncture was found only for severity of disability after treatment (OR 12.5, 95% CI 4.3-36.2). Liu et al. [28] analyzed 23 RCTs comparing various acupuncture modalities (manual, electro, scalp, auricular) with various control treatments, which the review did not describe clearly. Acupuncture proved superior to control treatments for motor recovery, assessed using the Traditional Chinese Medicine Efficacy Evaluation (OR 2.44, 95% CI 1.42-4.19) or the Fugl-Meyer Assessment (OR 12.38, 95% CI 7.18-17.58), as well as for ADL functionality, measured using the Barthel Index (OR 17.30, 95% CI 16.05-18.05). Li et al. [30] analyzed 9 RCTs that compared three acupuncture modalities (electro, scalp, manual) with conventional medication in terms of muscle tone improvement of hemiplegic limbs and neurological deficit scores at the end of treatment. Acupuncture was associated with significantly better muscle tone (OR 2.45, 95% CI 1.62-3.72) and improvement in neurological deficit score (OR 2.90, 95% CI 1.98-4.26). The fourth review, by Qi et al. [27], also found acupuncture to be associated with significantly greater improvement in ADL functionality, as measured by the Barthel Index (WMD 8.70, 95% CI 5.43-11.97). However, none of the 4 reviews made firm conclusions about the efficacy of acupuncture due to the low quality of included trials or the small number of study participants.

One Cochrane review [33] and 3 non-Cochrane systematic reviews [31,32,34] examined poststroke dysphagia. The Cochrane review by Geeganage et al. [33] compared the efficacy of deep-swallowing manual or electro-acupuncture therapy with that of routine medication and rehabilitation for treating patients with acute or subacute stroke. Although the treatments did not differ significantly for the primary outcome of death or dependency/disability, acupuncture was associated with significantly less severe dysphagia (OR 0.24, 95% CI 0.13-0.46) based on a post-treatment water-swallowing test and bulbar function. Long and Wu [32] analyzed 72 RCTs involving 6,134 patients with various types of stroke in which medicine and/or rehabilitation training plus acupuncture was compared with medicine and/or rehabilitation only using the video-fluoroscopic swallowing study (VFSS) test and the water-swallowing test as outcomes. Data for these outcomes were converted into clinical efficacy rates in order to allow quantitative synthesis of data from all trials. Meta-analysis showed that acupuncture was associated with significant improvement in the severity of dysphagia (OR 5.17, 95% CI 4.18-6.38) when VFSS and water-swallowing test results were combined. This clinical benefit was even greater when only water-swallowing test results were meta-analyzed (OR 5.57, 95% CI 4.21-7.38). Long and Wu [32] verified this result in a subgroup analysis of 4 higher-quality RCTs (OR 2.34, 95% CI 1.34-4.07). An earlier systematic review similarly reported that acupuncture was associated with significantly better VFSS and water-swallowing test results than were medication, rehabilitation training or Chinese herbs [34]. Wong et al. [31] made only qualitative comparisons in their review comparing the combination of acupuncture with swallowing training against either medication with training or medicine alone. Those authors concluded that acupuncture together with conventional rehabilitation might have a positive effect on dysphagia after stroke.

The remaining reviews included in our overview focused on other stroke-related disorders: 2 examined poststroke depression [35,36] and single reviews examined poststroke hiccup [37], poststroke urinary incontinence [38], shoulder-hand syndrome [41], shoulder pain [39] and apoplectic aphasia [40]. Data in only one review on poststroke depression [36] and one on poststroke hiccup [37] suggested that acupuncture was superior to control treatments, but the authors in those studies did not draw any firm conclusions about the therapeutic efficacy of acupuncture due to the low quality of included trials or the small number of study participants.

This overview identified a considerable number of systematic reviews, giving us a comprehensive picture of controlled assessments of acupuncture as a possible treatment for stroke and related disorders. Most studies focused on acupuncture as an intervention for stroke rehabilitation. Although most of the reviews we included did not wish to draw firm conclusions due to the small size of the included trials or their low methodological quality, our overview identified relatively convincing evidence for acupuncture as a treatment for stroke-related dysphagia.

Based on the ability of acupuncture to increase the flow of qi at points of needle insertion, practitioners of traditional Chinese medicine have historically held it to be particularly effective for ameliorating poststroke neurological impairment and dysfunction. Our overview supports this belief by providing evidence that acupuncture can be effective for stroke rehabilitation and stroke-related dysfunction based on assessments of neurological impairment, global neurological deficit, ADL functionality or water-swallowing ability. On the other hand, acupuncture was not associated with significantly better rates of death or dependency/disability than control treatments in any of the reviews we included.

It is worth noting that death or dependency/disability remain the primary outcomes accepted by most researchers and clinicians for assessing the efficacy of acupuncture. Other outcomes are less widely accepted because of potential measurement bias. For example, the clinical benefit of acupuncture for treating poststroke dysphagia may be partly attributable to the fact that the VFSS test and water-swallowing test are more subjective than measures of death and disability. Therefore, it would be valuable to develop more precise and objective tools that can be widely recognized for assessing the efficacy of acupuncture to treat stroke.

Our overview is constrained by limitations in the included systematic reviews. These limitations occur at the level of individual studies: only 13.5% of acupuncture trials included in Cochrane systematic reviews show acceptable quality in terms of randomization, allocation concealment, blinding and low rate of loss to follow-up [42]. In fact, many of the systematic reviews in our overview did not draw any firm conclusions because of methodological flaws in the included trials. In addition, many of the systematic reviews themselves failed to prespecify their inclusion and exclusion criteria according to the ‘PICO' acronym. This may have led the authors to include inappropriate studies, causing significant heterogeneity in the dataset and increasing the risk of misleading conclusions according to the ‘garbage in, garbage out' principle. One aspect of this heterogeneity is that many systematic reviews included different types of stroke (ischemic and hemorrhagic) at different stages.

In conclusion, the current evidence suggests that acupuncture is not effective for death or dependency/disability. However, acupuncture may be effective for treating poststroke neurological impairment and dysfunction, particularly poststroke dysphagia, although these findings require verification in rigorous RCTs. Future studies should place more emphasis on the efficacy of acupuncture for treating poststroke neurological impairment and dysfunction and on the development of more precise tools for assessing these outcomes.

The authors have nothing to disclose.

J.Z. and D.W. contributed equally to this study.