A systematic review and meta-analysis of acupuncture for improving learning and memory ability in animals

Background Memory loss is the most prominent symptoms of brain aging, but there is currently no evidence-based treatment strategy. Acupuncture has been widely used in China and the effectiveness for improving learning and memory has been mentioned in previous studies. We conducted this systematic review and meta-analysis to evaluate the effectiveness of acupuncture for improving learning and memory in animal experiments. Methods We searched Pubmed, Embase, Ovid Medline(R), the China National Knowledge Infrastructure (CNKI), Chinese Science and Technology Periodical Database (VIP) and Wanfang data Information Site to collect studies published up to December 2015. Study quality for each included article was evaluated according to the CAMARADES 10-item checklist. Outcome measure is Morris water maze. A meta-analysis was conducted according to the Cochrane systematic review method by using RevMan 5.3 software. Results Forty-two studies involving 944 animals were included. The quality score of the studies ranged from 2 to 8, with a mean of 5.3. Meta-analysis results showed that 24 studies reported significant effect of acupuncture for decreasing escape latency (−3.00, 95 % CI: −3.78 ~ −2.23, P < 0.00001), 14 studies reported significant effect of acupuncture for increasing frequency of cross platform (2.57, 95 % CI: 1.92 ~ 3.22, P < 0.00001), and 7 studies reported significant effect of acupuncture for increasing time in target quadrant (2.00, 95 % CI: 1.10 ~ 2.91, P < 0.00001) compared with the control group. Conclusions These findings show acupuncture has a potential role in improving learning and memory ability in animal models, suggesting it as a candidate therapy for memory loss of aged brain.


Background
Learning and memory are the most basal and important higher nervous functions and closely related to each other [1]. Learning means accepting information from the outside environment and memory refers being able to use this information at a later date [2][3][4]. As a symptom, learning and memory impairment often appears together in some diseases, such as Alzheimer's disease, vascular dementia, diabetes, autism, and so on [5][6][7][8].
With the increasing of living pressure and changing in life style, learning and memory impairment as the important embodiment of brain dysfunction have become one of the most important factors that affect people's lives [1,2]. The treatment based on western medical science for most learning and memory impairment only relieves the symptoms and delays the progression of disease [9,10]. Moreover, it also has some side effects caused by long time treatment [11].
As a kind of economical and side-effect free natural remedies, acupuncture has been used in China widely for over 2000 years [12]. The effect of acupuncture on encephalopathy has been recognized internationally. More and more studies have been published to confirm the effectiveness of acupuncture for improving learning and memory [13][14][15][16]. However, to some extent, the small sample size makes it hard to draw firm conclusions.
Up to now, there have been no systematic reviews to analyze the effectiveness of acupuncture for improving learning and memory. Reviews based on animal data could make trails' planning more perfect, increase the odds of success of future clinical trials and assist to decide what is valuable in further research [17]. Additionally, animal experiment can make us better understand the mechanism of acupuncture on learning and memory and guide the future clinical study. Therefore, we conducted a systematic review and meta-analysis of the effectiveness of acupuncture for improving learning and memory in animal experiments to provide suggestions for future animal experiments and clinical trials.

Search strategy
The following electronic databases were searched: Pubmed, Embase, Ovid Medline(R), China National Knowledge Infrastructure (CNKI), Chinese Science and Technology Periodical Database (VIP) and Wanfang data Information Site. The publication time is from the inception of each database up to December 2015. The languages were limited to English and Chinese. Search terms consisted of two groups: intervention (acupuncture and other related terms) and object (learning and memory and other related terms). All searches were limited to animals. We combined the results of all searches and then removed the duplicates. We also tried to get additional records identified through other sources.

Inclusion criteria
They were included if the following criteria were met: (1). Subjects: Animal models of learning and memory impairment were included. (2). Interventions: Acupuncture was the main therapy and only included manual acupuncture and electroacupuncture. (3). Outcomes: Morris water maze test was the primary outcome to explore the effectiveness of acupuncture groups and the difference between control groups and acupuncture groups. The Morris maze test is arguably the preferred test for assessing learning and memory in basic research. As a classic test, it has been accepted and used widely in most related animals experiments. (4). Language: Chinese and English articles.

Exclusion criteria
They were excluded if the following criteria were met: (1) Scalp acupuncture, auricular acupuncture, moxibustion and other forms of acupuncture; (2) studies that included Chinese herbal medicine or Western medicine; (3) studies that compared different acupuncture techniques or different acupoints; (4) studies without control group; (5) duplicate publications.

Study selection and data extraction
According to the above design, one reviewer (KYH) searched those databases and listed the titles of all articles. Two evaluators (KYH and SL) assessed the eligibility of these articles and made decision on every research (inclusion or exclusion) independently. If they did not reach the same decision, the concerned articles were discussed with a third reviewer (SFL).
Two reviewers (KYH and SL) extracted data independently from each study. The database included: (1) basic information, including publication year, the first author's name and model of learning and memory impairment; (2) individual data, including the number of animals, species and weight in acupuncture group and control group; (3) information on treatment, including timing and duration for treatment and method of treatment procedure; (4) the results of Morris water maze test. If outcomes were presented at different time points, we extracted data from the last time point. Differences of extracted data were solved after discussion with a third reviewer (SFL).

Quality assessment
We evaluated the methodological quality of the included studies by a ten-item scale [12]: (1) publication in a peer-reviewed journal; (2) statements describing control of temperature; (3) random allocation to treatment or control; (4) blinded building of model; (5) assessment whether building model is successful; (6) blinded assessment of outcome;(7) use of anesthetic without significant intrinsic neuroprotective activity; (8) sample size calculation; (9) compliance with animal welfare regulations; (10) declared any potential conflict of interest. Each item of the ten-item scale was attributed to one point. Based on this, each study had a quality score from zero to ten. The higher the score is, the better the article's quality is.
Two reviewers (KYH and SL) extracted data independently and assessed study quality. Disagreements were solved after discussion with a third reviewer (SFL).

Statistical analysis
Some results of Morris water maze test including escape latency, frequency of cross platform and time in target quadrant were considered as continuous data. Standard mean difference (SMD) was given, which was an estimate of the combined effect sizes. Publication bias was assessed with a funnel plot. Moreover, to explore the impact of factors affecting the outcome measures, we analyzed the specific subgroups based on escape latency and frequency of cross platform for the following factors: manual acupuncture and electroacupuncture, articles published or unpublished, species of animals, different ways to make Alzheimer's disease (AD) model and different ways to make vascular dementia (VD) model.
The meta-analysis was performed with RevMan 5.3 software. The confidence interval (CI) was established at 95 %, and P values of less than 0.05 were considered statistically significant. For the assessment of heterogeneity, the I 2 statistic and χ 2 distribution were used.

Study inclusion
Initially, 1421 records were searched from six databases. After removing duplicates, the records were decreased to 875. Based on titles and abstracts of records, we excluded 584 papers with reasons, such as not an animal experiment, case report or review, not related to learning and memory, and so on. The 291 remaining articles were downloaded for further selection. Due to republications, not using the Morris water maze test, comparing with other forms of acupuncture or Chinese herbs, and so on, 250 articles were excluded. Eventually, 42 studies were included . The flow diagram of the study selection process is shown in Fig. 1.

Study characteristics
The 42 included studies involved 944 rats. The total animal number in control groups is 470 and the number in acupuncture groups is 474. 36 studies of all have mentioned specific weight of rats. The rats' weight ranged from 160 to 320 g in 29 studies. The rats' weight was around 20 g in 3studies and more than 320 g in 4 studies. The age of animals was different and mentioned concretely in 20 studies. It ranged from new-born to 24month-old. 1 study used new-born rats; 10 studies used 2-4 months old rats; 6 studies used 6-9 months old rats; 3 studies used aged rats (more than 12 months old). Different subtests of the Morris water maze test were used in these studies: 41 studies with 912 animals reported data as escape latency, 18 studies with 406 animals reported data as frequency of cross platform and 12 studies using 255 animals reported data as time in target quadrant. The rat species included Sprague-Dawley (SD) rats, Wistar rats and AKR rats. Eighteen out of the 42 studies (42.9 %, n = 377) were AD models. Fourteen studies (33.3 %, n = 309) were VD models. And the 10 remaining studies (23.8 %) used other models. The main characteristics of the 42 studies are shown in Table 1.

Description of acupuncture regime
Varied acupuncture techniques were used in terms of selection of acupuncture-points, manipulation or stimulation methods ( Table 1). The most commonly used acupoints, which have been used by four or more studies, were GV20 (baihui), ST36 (zusanli), GV14 (dazhui), BL23 (shenshu), BL17 (geshu) and CV17 (danzhong). The frequency of acupuncture was mostly once per day. Animals received acupuncture treatment 1 to 30 min per session. The course of acupuncture treatment ranged from 7 to 60 days. The average duration of acupuncture was 22.4 days (SD = 10.8). 13 studies used manual acupuncture, and the rest 29 studies chose electroacupuncture. 5 of 13 studies stated detailed operating methods of manual acupuncture. The operating method of manual acupuncture was mostly twirling reinforcing. 28 of 29 studies stated detailed operating parameters of electroacupuncture and only one ignored related descriptions. 20 of 28 studies used continuous waves. The frequency of continuous wave is from 0.5Hz to 150Hz. The current density of continuous wave is from 0.5 mA to 30 mA. The rest 8 studies used disperse-dense waves. The frequency of disperse wave is from 1Hz to 80Hz and the frequency of dense wave is from 15Hz to 100Hz. The current density of continuous wave is from 1 mA to 3 mA.

Description of control interventions
Eighteen of included studies used some interventions in control groups (Table 1). Control interventions consisted of western medicine and sham acupuncture. Western medicine was adopted in 9 studies and 9 experiments used sham acupuncture. Medication was administered for similar treatment duration as acupuncture. Types of control medication consisted of donepezil (3 studies), nimodipine (3 studies), fluoxetine (1 study), madopar (1 study) and piracetam (1 study). Puncturing points lateral to acupoints is the way of sham acupuncture in 8 experiments. Not using electroacupuncture apparatus is the other way of sham acupuncture in 1 experiment which adopted electroacupuncture as the intervention way.

Study quality and publication bias
The score of the study quality was ranged from 2 to 8 out of a total 10 points. Concretely, one study got 2 points; three studies got 3; twelve studies got 4; eleven studies got 5; seven studies got 6; three studies got 7 and five studies got 8 points. Five studies were not published because they were Master's or Ph.D thesis. Twentyseven studies mentioned control of temperature, including control of the room or water temperature. Nineteen studies adopted blinded building of model and seventeen mentioned assessment whether building model was successful. Random allocation to control group or acupuncture group and blinded assessment of outcome were described in 42 and 12 studies. No study reported inducing significant intrinsic neuroprotective activity because of anesthetic. Only one study described the sample size calculation. Fourteen studies reported statement of potential conflict of interests and eleven reported compliance with animal welfare regulations. The study quality and publication bias evaluation are shown in Table 2.

Effectiveness
Forty-one studies reported the impact of acupuncture on decreasing escape latency compared with the control group (p < 0.05 or p < 0.01). Twenty-four of them provided detailed data to show significant effectiveness of acupuncture compared with the control group (n = 494, SMD −3.00, 95 % CI: −3.78~−2.23, P < 0.00001; heterogeneity χ 2 = 185.09, I 2 = 88 %, Fig. 2). The remaining seventeen studies did not provide detailed data and just showed the data demonstrated in a graphical form.   Eighteen studies reported the impact of acupuncture on increasing frequency of cross platform compared with the control group (p < 0.05 or p < 0.01). Fourteen of them provided detailed data to show significant effectiveness of acupuncture compared with the control group (n = 317, SMD 2.57, 95 % CI: 1.92~3.22, P < 0.00001; heterogeneity χ 2 = 52.81, I 2 = 75 %, Fig. 3). The remaining four studies just showed the data demonstrated in the graphical form and failed for meta-analysis.
Eleven studies reported the impact of acupuncture on increasing time in target quadrant compared with the control group (p < 0.05 or p < 0.01). Seven of them provided detailed data to show significant effectiveness of acupuncture compared with the control group (n = 149, SMD 2.00, 95 % CI: 1.10~2.91, P < 0.00001; heterogeneity χ 2 = 28.18, I 2 = 79 %, Fig. 4). The remaining five studies just showed the data demonstrated in the graphical form.

Signaling pathways
Several different signaling pathways were investigated to gain a better understanding of the underlying mechanism of acupuncture in the amelioration of learning and memory impairment. 39 of 42 included studies got detailed descriptions about possible mechanisms. It can be found that reduction nerve apoptosis and necrosis and suppression of oxidative stress are the main signaling pathways. A summary of proposed mechanism is shown in Table 3.

Assessment of bias
The funnel plot was approximately symmetric for the effect of acupuncture on escape latency, frequency of cross platform and time in target quadrant (Fig. 5). Nevertheless, publication bias was still considered to be potential.
In the subgroup analysis of time in target quadrant, the efficacy of manual acupuncture was better (SMD 2.94, 95 % CI: −0.67~6.55, Fig. 8a). There were no unpublished articles using time in target quadrant as outcome indexes. According to different varieties of experimental animals, SD rats were more sensitive to acupuncture than Wister rats and AKR rats (SMD2.33, 95 % CI: 0.77~3.89, Fig. 8b). Compared with AD models, VD models got more improvement of time in target quadrant (SMD 3.36, 95 % CI: 1.75 4.96, Fig. 8c). By analyzing different ways making models of AD, we found that AD models caused by D-gal and Aβ were more sensitive to acupuncture (SMD 2.92, 95 % CI: 1.72~4.13, Fig. 8d). 4-VO was the only way to be included causing VD models and using time in target quadrant as one of outcome indexes. Different weight and age of rats were included in these studies. We found that rats whose weight ranges from 18 to 35 g (SMD 3.10, 95 % CI: −0.166 .37, Fig. 7e) and rats whose age ranges from 6 to 9 months old (SMD 2.31, 95 % CI: 0.65~3.98, Fig. 7f ) were more sensitive to acupuncture for improvement of time in target quadrant.
These results were mostly inconsistent in three subgroup analyses. It may be caused by studies of low quality, publication bias or other reasons.

Discussion
To our knowledge, this is the first systematic review and meta-analysis to explore the efficacy of acupuncture for improving learning and memory in animal experiments with the results of Morris water maze test as the outcome measures. This systematic review and meta-analysis indicated that acupuncture could reduce time of escape latency, decrease frequency of cross platform and increase the time in target quadrant in animal model of learning and memory impairment. It suggests that acupuncture has a potential role in improving learning and memory impairment in animal models.
This review made a more detailed description of the acupuncture treatment procedure, including acupoint selection, stimulation methods, and treatment duration. The variation in the acupuncture protocol might contribute to the heterogeneity in treatment outcome between studies. The most common acupoints in acupuncture treatment for improving learning and memory were GV20 (baihui), ST36 (zusanli), GV14 (dazhui), BL23 (shenshu), BL17 (geshu) and CV17 (danzhong). GV20 and GV14 are Governor Vessel acupoints. BL23 and BL17 are Bladder Meridian of Foot-Taiyang acupoints. This result suggests the possible importance of Governor Vessel and Bladder Meridian of Foot-Taiyang for learning and memory. About one-third of included studies used manual acupuncture, and twothird used electroacupuncture. We have found that continuous waves, 2-4Hz of frequency and 1-2 mA of current density are the most commonly used stimulus parameters of electroacupuncture and twirling slowly is commonly used in manual acupuncture groups. Thus it can be seen that mild stimulation which means the reinforcing method in Traditional Chinese Medicine theory is adopted widely for improving learning and memory.
In the acupuncture study on animal model, setting sham acupuncture as control is very important and difficult. The importance of sham acupuncture is that it can help to clear the effectiveness of acupuncture after excluding placebo effects. While popular sham ways are mainly skin penetration. Except for placebo effects, it can also produce biological effects. Therefore, the true effect of acupuncture may be underestimated when compared to sham acupuncture [60].
Acupuncture has been known as an effective therapy for learning and memory impairment appearing in Alzheimer's disease, vascular dementia, and so on   [5][6][7][8]. The mechanism of acupuncture on improving learning and memory ability remains unclear. It is recognized that learning and memory are associated with cerebral cortex and hippocampus closely [40].
Modern medical research suggests that acupuncture may reduce nerve apoptosis and necrosis to protect cortex and hippocampus neurons through different aspects, which include decreasing the level of cytokines in hippocampus [41], adjusting the neurotrophic factors and cholinergic system [42,47], inhibiting the expression of matrix metalloproteinase(MMP) [45], and so on. Additionally, it has been suggested that acupuncture can increase the activity of superoxide dismutase (SOD) and decrease the level of malondialdehyde (MDA) in brain to improve the antioxidant capacity and reduce brain tissue damage caused by free radical [58]. There are some limitations to this review. Firstly, our search only included Chinese and English articles and excluded those articles published in other languages. At the same time, we only included manual acupuncture and electroacupuncture and excluded some special acupuncture therapies, such as scalp acupuncture, auricular acupuncture, abdominal acupuncture, and so on. It may cause selective bias. Secondly, the total number of studies and the total sample size were too small for reliable. We have performed comprehensive literature search (six databases) and conducted extensive searches through other sources. But no more studies were found. Thirdly, articles which reported negative results may not be popular to publish so that the effectiveness of published articles would be better than those unpublished. Therefore, the effectiveness of acupuncture for improving learning and memory may be overstated. Fourthly, the quality of included studies was very low, so that it had significant impacts on the outcomes of the meta-analysis.
Based on the above limitations, more other language articles and special acupuncture treatment means should be included in the future systematic. Meanwhile, control of temperature, random allocation to treatment or control, blinded building of model, assessment successful model building and blinded assessment of outcome should be pay attention to in the future animal or clinical studies. Not only positive results, but also negative should be reported in the future animal or clinical studies.
In addition, some implications are also brought out after analyzing affecting factors of outcome indexes. Manual acupuncture showed more effective to escape latency and time in target quadrant than electroacupuncture. And they had roughly the same impact on frequency of cross platform. But in the present study, only 13 out of 42 studies performed manual acupuncture and the rest performed electroacupuncture. Because electroacupuncture is easier to control, standard and objectively measure than manual acupuncture [12], so it is used widely in clinical and experimental researches. Therefore, it remains unknown which means of acupuncture is more effective and convenient for people's learning and memory impairment. Other impact factors do not show any universal regularity through histogram analysis. It need be solved by higher quality studies, lower publication bias, and so on.

Conclusion
In animal model, acupuncture has a potential role in improving learning and memory ability. But it is still ambiguous that which stimulating mode (manual acupuncture or electroacupuncture) is more effective. Low quality of studies and larvaceous publication bias may reduce persuasiveness of positive results and should be solved in the future.