2.1. Protocol and Registration

The protocol of this study followed the preferred reporting items for systematic reviews and meta-analysis (PRISMA) guidelines (Additional file 1) [15] and is adapted from the structure provided in the Systematic Review Protocol for Animal Intervention Studies [16]. This study was registered at OSF (Registration DOI: 10.17605/OSF.IO/FNM37).

2.2. Eligibility Criteria

2.3. Search Strategy

PubMed, Web of Science, China National Knowledge Infrastructure (CNKI), and the Chinese Science and Technology Periodical Database (VIP) were searched from inception to December 20, 2020. The main terms “Polycystic ovary syndrome”, “Acupuncture”, and “Animal Experimentation”, indexed in the MeSH system, were combined [17].

2.4. Study Selection

Two independent reviewers (JY and LJZ) screened titles and abstracts for eligibility. Disagreements between reviewers were resolved by a third review (YL). Full texts were obtained and evaluated by the same reviewers using a predesigned form.

2.5. Data Extraction

Two reviewers (JY and LJZ) extracted data independently, and any controversy was resolved by discussion. The following information was recorded using a predesigned form: study design, characteristics of the included animals, characteristics of the animal model, details of the intervention, and outcome measures.

2.6. Risk of Bias Assessment

Two reviewers (YL and JJG) assessed the risk of bias using SYRCLE’s tool for assessing risk of bias (SYRCLE ROB) [18].

2.7. Data Synthesis

We performed a meta-analysis using a random-effects model with Review Manager (RevMan) 5.3. Treatment effects were summarized as the standard mean difference (SMD) with a 95% confidence interval (CI). The SMD is an evaluation of the combined effect sizes, and P values below 0.05 were considered statistically significant. The presence of heterogeneity was evaluated by I² and chi-square statistical analyses. Funnel plots were performed to evaluate publication bias if there were more than ten studies included [17].

3.1. Study Selection

A total of 384 potentially relevant studies were identified from the abovementioned four databases. After removing duplicates, 358 records remained for title and abstract screening. A total of 303 studies were excluded due to at least one of the following reasons: (1) not an animal study, (2) case report or review, or (3) not related to acupuncture. Finally, 55 studies remained after the initial reading. After full-text reading of the remaining studies, 24 studies were excluded, and 31 studies were included (Figure 1).

3.2. Study Characteristics

A total of 31 studies were included [13, 14, 19–46]. Thirteen studies were published in English, and 18 were published in Chinese. A total of 722 animals (9 studies with Wistar rats and 22 studies with SD rats) were involved, and the details of the animal models are presented in Table 1. Eleven studies induced a PCOS model by letrozole, 8 studies used dihydrotestosterone (DHT), 6 studies used dehydroepiandrosterone (DHEA), 4 studies used testosterone propionate (TP), 1 study used estradiol valerate (EV), and 1 study used prasterone sulfate. Thirty included studies randomly assigned animals to the acupuncture and control groups, and only 1 study did not report randomisation details. Twenty-six studies used electroacupuncture (EA), 9 studies used manual acupuncture (MA), and 5 of them employed two acupuncture groups, including both EA and MA. Among all the included studies, a total of 22 acupoints were involved. The frequencies of acupuncture points from high to low were as follows: SP6, 20 times; CV4 (RN4), 17 times; CV3 (RN3), 10 times; ST36, 7 times; EX-CA1, 7 times; ST29, 4 times; RN12, 3 times; ST25, 3 times; LI10, 2 times; BL23, 2 times; SP9, 1 time; ST40, 1 time; RN6, 1 time; PC6, 1 time; EX-B3, 1 time; CV12, 1 time; BL18, 1 time; BL20, 1 time; ST27, 1 time; ST28, 1 time; LR3, 1 time; and Hou Hui, 1 time. Seven studies followed the modern acupuncture pattern, in which the basic principle was two needles inserted bilaterally in the abdominal muscles, with two needles placed in each soleus and gastrocnemius hindlimb muscle in somatic segments corresponding to ovarian innervations [18, 19, 23, 24, 26, 31, 36]. The electrical frequencies used were all 2 Hz except for one study that did not mention the EA frequency, and the intensity was 0.6–3 mA. One study reported intervention with a single session; the others lasted from 2 weeks to 8 weeks.

3.3. Risk of Bias within Studies

The results of the risk of bias of the included studies are summarized in Table 2. SYRCLE ROB included domains for selection bias (sequence generation, baseline characteristics, and allocation concealment), performance bias (random housing and blinding), detection bias (random outcome assessment and blinding), attrition bias, reporting bias, and other biases.

3.4. Effectiveness

The meta-analysis was not performed since the heterogeneity was significantly high. Although the heterogeneity analysis was carried out based on sensitivity analysis, the detailed reasons for the potential heterogeneity were not very certain.

3.5. Quality of Evidence Assessment (Table 3)

The GRADE recommendations for HOMA-IR, T, FINS, and BW were very low, and those for LH, LH/FSH, and FBG were low. The certainty was downgraded for the following reasons: I values that exceeded 75%; different animal modelling methods; differences in the treatment cycle and dosage; physiological and pathological differences between rodents and humans; small sample size; and the 95% CI included the value of one.

4.1. Summary of Evidence

To the best of our knowledge, this is the first systematic review summarizing the current evidence of the effects of acupuncture on PCOS in animal experiments. This finding suggests that acupuncture may play a potential role in restoring reproductive endocrine function in PCOS-like animal models.

The mechanisms of acupuncture in PCOS animal models are still unclear. It is well known that elevated sympathetic activity contributes to the development and maintenance of PCOS [47, 48]; thus, the sympathetic nervous system may offer a novel therapeutic target in treating PCOS. Normal ovulation requires three components, namely, an intact central hypothalamic-pituitary-ovarian axis, synchronized feedback signals, and normal local responses within the ovary [49]. The evidence suggests that low-frequency EA could lower sympathetic activity, and the effects may be mediated by modulation of NGF expression of sympathetic outflow to the ovaries in PCOS-like rats [50]. Zhang et al. reported that the regulation of EA in reproductive function in PCOS-like rats may not be accomplished by the hypothalamic-pituitary-ovarian axis [22, 23], while Maliqueo et al. reported that low-frequency EA significantly affects the pituitary-ovarian axis by normalizing LH secretion [32]. This contradiction is likely due to different PCOS-like rat models and the different forms of acupuncture employed and it needs to be elucidated with further experiments. It has now been demonstrated that the effect of low-frequency EA on ovarian function is mediated as a reflex response via the ovarian sympathetic nerves, and the response is controlled via supraspinal pathways [51]. It has been newly demonstrated that ovarian innervation likely plays an important role in folliculogenesis, and EA might restore PCOS pathophysiology by regulating ovarian innervation, at least partially mediated through the superior ovarian nerve. The effect of EA is based on the integrity of the nervous system [46].

Other hypotheses have also been reported. EA stimulates the development and maturation of eggs in PCOS-like rats by increasing the level of stem cell factor and reducing the level of TNF-α responsible for follicular fluid [26]. EA could increase P450arom and decrease P450c17α as well as the expression levels of their mRNA in ovarian tissues in PCOS-like rats. These effects may thereby change the local ovarian environment of excessive androgen and improve the reproductive, endocrine, and metabolic disorders associated with PCOS [28]. Adiponectin reduces androstenedione synthesis in human theca cells [52]. EA stimulated the ovarian adiponectin system in rats with letrozole-induced PCOS, and the effect does not seem to be mediated by modulation of sympathetic activity [32]. Oxidative stress is now recognized to play a central role in the pathophysiology of PCOS [53]. Zheng et al. reported that EA decreases serum malondialdehyde levels and increases superoxide dismutase levels, hence improving the oxidative stress status of PCOS-like rats [33]. EA not only regulates abnormal glucose and lipid metabolism in PCOS-like rats but also increases glucose transporter 1 and glucose transporter 4 expression in ovarian tissue, which may alleviate insulin resistance [35]. EA increases angiogenesis in the antral follicles of PCOS-like rats, which favours follicle maturation, and ovulation is suggested as being one of the mechanisms involved in the effects of EA on PCOS [54]. It has also been demonstrated that EA regulates androgen receptor and Connexin 43 (which plays an important role in the process of oocyte meiosis and follicular selection) in PCOS-like rats; however, further studies are needed to clarify whether this is one of the mechanisms involved in the effects of EA on PCOS [14]. It has recently been recognized that autophagy is involved in the occurrence and development of PCOS [55]. Huang et al. demonstrated that EA inhibits autophagy in ovarian tissue through the PI3K/AKT pathway [44]. Sterol regulatory element binding protein-1 (SREBP1) is a key gene in lipid metabolism regulation. Peng et al. suggested that EA regulates SREBP1 expression, thereby improving insulin resistance, mitochondrial dysfunction, and oxidative stress in PCOS-like animals [13]. These studies provide novel insights into the mechanisms of EA in PCOS; however, further studies are needed to confirm the findings.

EA and MA are both widely used in clinical practice; interestingly, studies on PCOS animal models demonstrate that the mechanism of their action is not identical. Feng et al. demonstrated that EA regulates neuroendocrine and reproductive functions through the endogenous opioid receptor system and manual stimulation by regulating steroid hormone receptors [25]. Johansson et al. reported that MA has a greater effect on glucose tolerance than EA [27]. EA reduces the weight of the subcutaneous fat depot, increases the weight of the soleus muscle, and affects the expression of genes and proteins related to the insulin signaling pathway in the soleus skeletal muscle, while MA improves systemic glucose tolerance and affects gene expression in mesenteric adipose tissue [27]. Benrick et al. further reported that EA has stronger effects on glucose uptake than MA and that it induces more pronounced changes in molecular pathways and improves insulin sensitivity more rapidly, and both EA and MA are equally effective during the poststimulation period [29]. The underlying mechanism of the different actions remains to be elucidated [29].

Different acupoint protocols are employed in studies, and several studies have investigated the different actions. Xu et al. demonstrated that there was no significant difference in body weight when stimulating ST36, SP6, and CV4 separately or in combination. Electrostimulation with ST36 or CV4 alone significantly decreased the T level compared with stimulation with SP6 or their combination. Expression of androgen receptor decreased significantly in the SP6 and CV4 stimulation groups [39]. ST36 significantly improves hormone levels and the expression of receptors in ovarian tissue, but it does not reduce the number of growing follicles. CV4 can regulate follicular development and hormone levels but has no obvious effect on the expression of gonadotropin receptors. The combination group had no evident advantage compared with the single acupoint group [41]. It suggests that the effect of EA at multiple points may not be better than that of a certain empiric point. Whether there is an antagonistic effect on the therapeutic effect of different acupoints is worthy of further study.

4.2. Limitations

The positive findings of acupuncture in PCOS animals should be interpreted with great caution since there were several limitations in the present systematic review. First, the substantial heterogeneity should be taken into consideration. Although we performed sensitivity analysis and subgroup analysis, the reason for the generated high heterogeneity was not identified. The difference between acupuncture protocols and different methods used to induce the PCOS model might contribute to the high heterogeneity. Second, the result of the risk of bias assessed by SYRCLE ROB indicated the methodological weakness of the included studies. A majority of the included studies did not report details of sequence generation, baseline characteristics, allocation concealment, randomizing, blinding, or random outcome assessment, which impaired the power of the evidence generated from the present systematic review.

The main implications for further research are as follows: first, the design of the experiment should follow SYRCLE to minimize the risk of bias; second, the report of acupuncture on PCOS animals should follow the Standards for Reporting Interventions in Controlled Trials of Acupuncture (STRICTA) to prompt better quality reporting on acupuncture interventions and help the interpretation and analysis and enable research replications.