Effects of different acupuncture methods on polycystic ovarian syndrome: a systematic review and network meta-analysis

This NMA was conducted following the PRISMA protocol [ 24 ], with strict adherence to all items including detailed reporting of search strategies, study selection criteria, data extraction procedures, risk of bias (RoB) assessment, and result synthesis. The study has been registered with PROSPERO (CRD42023494693). The inclusion and exclusion criteria of this study were rigorously developed based on the PICO framework. Inclusion criteria: (1) Participants: Women aged ≥ 18 years and diagnosed with PCOS (regardless of diagnostic criteria used in primary studies). (2) Interventions: At least one of the following acupuncture methods (any frequency or duration): body acupuncture, E-acupuncture, or ACE. (3) Comparison: Either placebo or standard treatment: routine care, western medicine, or traditional Chinese medicine (Cmedicine). (4) Outcomes: At least one of the following outcomes: hyperandrogenism, metabolic status, sexual hormone disturbance, and infertility. (5) Study design: Randomized controlled trials (RCTs). (6) Language: Studies published in English or Chinese. Hyperandrogenism was assessed by the testosterone level and Ferriman-Gallwey (FG) score (designed by Ferriman and Gallwey in the United Kingdom in 1961) [ 25 ]. The unit of testosterone was uniformly converted to nmol/L according to international standards. Metabolic parameters included body mass index (BMI, BMI=weight (kg)/heightional), waist-to-hip ratio (WHR, WHR=waist circumference (cm)/hip circumference (cm)), triglycerides, high-density lipoprotein (HDL), low-density lipoprotein (LDL), fasting blood glucose (FBG) and homeostasis model assessment of insulin resistance (HOMA-IR, HOMA-IR=(fasting insulin level (µn/mL) ×mfasting blood glucose (mmol/L)/22.5). The units of triglycerides, HDL, LDL, and FBG were mmol/L. Sexual hormone disturbance was evaluated by the luteinizing hormone (LH) level and LH/follicle-stimulating hormone (FSH) ratio. The units of LH and FSH were uniformly converted to IU/L in accordance with international standards. Infertility was assessed by the clinical pregnancy rate (defined as the presence of at least one gestational sac on ultrasound). The selection of the above outcome measures was based on clinical guidelines and clinical practical needs, which were closely associated with the pathophysiology of PCOS [ 11 ]. Testosterone, FSH, and LH were assayed during the menstrual phase, while all metabolism-related parameters were determined from fasting blood samples. Exclusion criteria: (1) No predefined acupuncture methods (body acupuncture, E-acupuncture, ACE) were used in the experimental group; (2) Non-RCTs (e.g., reviews, meta-analyses, abstracts, guidelines, consensus statements, animal experiments); (3) Studies with inaccurate or incomplete data. Incomplete data were defined as lacking relevant outcomes. Inaccurate data referred to those with logical inconsistencies, for which we would contact the authors for rectification, and those that could not be rectified would be excluded. Two reviewers (Peiru Li and Yun Lu) independently searched English databases (PubMed, Embase, CENTRAL, and Web of Science) and Chinese databases (CNKI, Wanfang, CQVIP, and CBM) by the end of 30 October, 2023. The search strategy utilized medical subject headings (MeSH) combined with free-text words including all known spellings of Polycystic Ovarian Syndrome and Acupuncture. The search strategy is displayed in Supplementary Table S1 . References of relevant studies were manually retrieved to obtain more potential studies. Studies were selected based on the predefined eligibility criteria independently by two reviewers (Peiru Li and Yun Lu). First, we imported the retrieved studies into EndNote X9 to remove duplicate publications. Titles and abstracts were examined to exclude non-eligible studies. Then the full texts were examined to ultimately determine the eligible ones. When disagreements arose between the two reviewers during title/abstract screening or full-text assessment, they first attempted to resolve them through detailed discussion. If consensus could not be reached, the third independent reviewer (Anli Weng) was consulted, whose judgment was considered final to ensure objectivity and rigor. The following data were extracted using a Cochrane data extraction form: (1) title, first author, and year of publication; (2) numbers of cases in each group, diagnostic criteria, and age; (3) interventions and controls; (4) the baseline and post-treatment time points of outcome measurement; (5) key factors for risk of bias (RoB) assessment. One reviewer (Peiru Li) extracted the data, while the other (Yun Lu) checked the accuracy. The included studies were assessed for the RoB independently by the two reviewers (Peiru Li and Yun Lu) with the Cochrane RoB 2.0 tool [ 26 ] from the following five domains: randomization process, missing outcome data, deviation from intended interventions, outcome measurement, and selective reporting. Each domain was rated as “Low”, “Some concerns”, or “High” RoB. Overall RoB was determined by the most conservative domain rating: (1) “High RoB” if any domain was rated as “High”; (2) “Some concerns” if no domains were rated as “High” but at least one domain was rated as “Some concerns”; (3) “Low RoB” if all domains were rated as “Low”. Disagreements were resolved by discussion or consultation with a third reviewer (Yongxia Wang), and results were visualized using RoB plots. Stata17 and R 4.5.2 were used for NMA [ 27 , 28 ]. NMA was performed in a Bayesian framework using random-effects models. The effect size was evaluated by mean difference (MD) for continuous variables or risk ratio (RR) with 95% credible interval (CrI) for dichotomous variables. In the network diagram, nodes represent individual interventions, and lines represent direct comparisons between two interventions. Node size indicates the total sample size of studies implementing the corresponding intervention (larger nodes = larger total sample sizes), and line thickness indicates the number of direct head-to-head comparisons between the two connected interventions (thicker lines = more comparisons). The results of NMA were based on all pairwise comparisons (direct and indirect). Local inconsistency tests were performed using the node-splitting method for the outcome measures with closed loops not formed by a single study. P 0.05 [ 29 ]. The I² test was used to evaluate statistical heterogeneity, with a value of I²>50% indicating statistically significant heterogeneity. The effects of interventions were estimated based on the surface under the cumulative ranking curve (SUCRA). A higher SUCRA value (0–100%) suggested a better treatment effect. However, SUCRA reflects a probabilistic ranking rather than absolute superiority. The results should be comprehensively interpreted in conjunction with league tables. Finally, small-study effects and potential publication bias in over 10 studies were presented in funnel plots. Calibrated funnel plots revealed an approximately symmetric distribution of symbols with identical coloration, indicating no significant publication bias.

As shown in the PRISMA flow diagram (Fig.  1 ), a total of 6540 potentially relevant studies were identified initially. First, 2912 duplicates were removed using EndNote X9. After title/abstract screening, 2043 non-eligible studies were excluded mainly due to inappropriate study design (non-RCTs), irrelevant interventions, or no reporting of target outcomes. The remaining 216 studies underwent full-text assessment, of which 141 were further excluded for reasons including inconsistent acupuncture methods, inappropriate comparisons (e.g., lack of placebo or standard treatment), absence of key outcomes, or failure to meet criteria for RCTs. Ultimately, 59 studies were included in the NMA. Fig. 1 PRISMA flow diagram for search and selection of eligible studies included in the network meta-analysis PRISMA flow diagram for search and selection of eligible studies included in the network meta-analysis As shown in Tables 1 , 59 eligible studies [ 30 – 88 ] involving 5937 participants were published in 2005–2023, with sample sizes of 15–926. Regarding diagnostic criteria, the majority of included studies (36/59, 61.0%) adopted the Rotterdam criteria, 11 studies (18.6%) used Chinese PCOS guidelines, and four studies (6.8%) referenced Gynaecology and Obstetrics. Eight studies (13.6%) failed to specify diagnostic criteria, which might reduce the comparability and interpretability of results across studies. For interventions, body acupuncture was the most commonly used method (38 studies, 64.4%), followed by E-acupuncture (15 studies, 25.4%), and ACE (7 studies, 11.9%). All included studies were RCTs, with 56 two-arm trials and 3 three-arm trials. Table 1 Characteristics of clinical trials included in the network meta-analysis ID First author/year Sample size Age (years) diagnosis criteria Main interventions Outcome measures Treatment Control Treatment Control Treatment Control 1 YY Zhuo 2016 50 50 29 ± 5 28 ± 5 C acupuncture medicine a, l 4 XB Cai 2016 25/30 25 29.5 ± 2.3 R Eacupuncture/ACE medicine f, g,h, j,k, m 10 CL Jin 2014 33 32 29 ± 4 27 ± 5 R Eacupuncture medicine a, m 11 XZ Liu 2023 34 34 30.2 ± 2.6 31.3 ± 3.1 R acupuncture placebo a, b 12 LQ Yu 2020 36 34 30 ± 6 31 ± 6 R Eacupuncture medicine a, c,d, e,g 13 SS Wang 2023 40 40 27.55 ± 4.93 27.55 ± 4.93 C acupuncture Cmedicine a, h,j, k 14 XX Zhang 2021 32 33 18–45 C Eacupuncture medicine a, k,h 15 JM Wu 2014 30 30 20–40 C ACE medicine a, k,l, m 16 WS Lin 2018 30 30 28.70 ± 4.81 28.20 ± 4.44 C acupuncture medicine a, k,l, m 17 WW Guo 2021 30 30 28.87 ± 5.83 /27.93 ± 5.62 27.90 ± 5.51 R acupuncture medicine a, f,g, k,m 18 Y Cao 2017 28 28 31 ± 3 29 ± 5 R acupuncture medicine a, h,k, m 19 JY Wang 2020 30 30 26.46 ± 3.74 26.23 ± 3.48 N acupuncture medicine c, f,g, h,j, l 20 H Ma 2016 30 30 24 ± 3 25 ± 4 R acupuncture medicine a 22 YH Yang 2005 66 60 26 25 G&O acupuncture medicine l 23 MH Lai 2010 43 43 26.5 ± 3.0 24.9 ± 4.9 R acupuncture medicine a, b,c, d,e, f,g, h,j, k,m 24 LP Yuan 2010 30 30 20–40 R acupuncture medicine a, k,l, m 25 MH Lai 2012 60 60 26.72 ± 2.65 26.46 ± 2.72 R acupuncture medicine a, f,g, h,j, k,m 27 WW Chen 2021 40 40 26.85 ± 2.55 26.83 ± 2.56 C acupuncture medicine a, k 29 LY Shen 2018 30 30 28.8 ± 0.8 30.1 ± 0.7 R acupuncture placebo f, g,h, j,k.m 32 SH Ma 2020 60 60 29 ± 3 28 ± 3 R acupuncture medicine l 33 T Zhang 2013 30 30 29.56 ± 2.830 30.16 ± 3.579 R ACE Eacupuncture b, f,g, h,k, m 34 R Cheng 2015 20 20 24.80 ± 4.68 24.95 ± 5.00 C ACE Cmedicine f, h,j 35 GY Liu 2010 22 22 26.77 ± 3.69 27.55 ± 3.25 N ACE Cmedicine a, h 36 LL Tao 2010 20 19 / / R ACE Cmedicine c, d,e, f,g, h 37 BY Chen 2016 16 16 25.01 ± 5.23 25.23 ± 4.98 R ACE Cmedicine a, f,h, k,m 38 HW Yang 2022 30 30 29.9 ± 3.05 29.8 ± 2.87 G&O acupuncture Cmedicine a, l,m 39 QX Fang 2016 30 30 29.97 ± 4.44 29.37 ± 4.50 R acupuncture medicine a, m 41 DH Yang 2017 30 30 27 ± 5 27 ± 3 R Eacupuncture medicine a, h,k, m 42 XY Li 2019 20 20 32.25 ± 1.66 29.94 ± 2.92 R acupuncture medicine a, h,k, m 44 HJ Wei 2022 50 50 28 ± 6 27 ± 6 C acupuncture medicine k, m 45 HL Zhang 2020 20 20 29 ± 2 28 ± 3 R Eacupuncture placebo a, b,h, m 46 Y Gu 2019 39 39 26.95 ± 4.54 28.56 ± 3.98 R Eacupuncture placebo c, d,e, f,g, h,j 48 Y Peng 2017 50 50 28.58 ± 3.82 28.68 ± 3.33 R Eacupuncture placebo h, j 54 N Lei 2021 70 70 32 ± 5 31 ± 4 R acupuncture Cmedicine l, m 56 C Li 2011 30 30 24.39 ± 4.58 23.54 ± 5.33 G&O acupuncture medicine k, l,m 57 Y Wang 2022 220 220 / / R Eacupuncture placebo a, c,d, e,f, g,h, j,k, m 59 XL Yu 2023 107 106 31.13 ± 8.90 30.25 ± 9.45 C acupuncture medicine a, c,f, g,h, j,k, m 60 M Yao 2018 50 50 27.8 ± 4.8 28.2 ± 4.5 R acupuncture medicine a, b,g, h,j, k,m 61 J Yang 2015 21 20 26.85 ± 3.30 26.05 ± 3.27 R Eacupuncture medicine a, k 64 D Chen 2007 61 60 26.15 ± 3.67 26.03 ± 3.69 N acupuncture medicine a, k.m 65 FL Ren 2022 78 75 29 ± 5 27 ± 5 C acupuncture Cmedicine c, d,e, l,m 66 J Su 2015 40 40 27.65 ± 4.17 28.14 ± 5.22 R acupuncture Cmedicine a, f,g, h,k.m 67 JN Gao 2022 30 30/30 29.2 ± 4.4 29.8 ± 4.2/ 30.0 ± 4.5 R acupuncture Cmedicine/medicine l, m 68 XP He 2017 53 53 26.56 ± 4.21 C acupuncture medicine k 69 J Yue 2021 30 30 26.3 ± 4.20 27.8 ± 4.23 R acupuncture medicine i, k,l, m 70 J Yue 2020a 30 30 27.39 ± 2.68 28.78 ± 2.58 R acupuncture medicine a, k,m 71 J Yue 2020b 40 40 26.03 ± 4.38 27.45 ± 4.31 R acupuncture medicine a, h,k, l,m 72 N Li 2017 53 53 26.61 ± 8.13 27.61 ± 8.03 G&O acupuncture medicine a, m 73 N Li 2016 14 14 / / N acupuncture medicine a, h,k 74 L Li 2014 49 51 26.2 ± 2.1 25.2 ± 1.8 R acupuncture medicine a, b,c, d,e, f,g, h,j, k,m E4 Johansson 2013 16 16 28.4 ± 3.1 27.9 ± 3.2 R Eacupuncture placebo b, c,d, e,f, g,h, j,k, m E5 H Chang 2023 458 468 27.97 ± 3.33 27.87 ± 3.33 R acupuncture placebo b, c,d, e,f, h E6 Nekooi 2022 48 48 32.6 ± 3.4 32.8 ± 2.7 N acupuncture placebo b E7 QD Wen 2022 114 114/114 25.0–31.0 24.0–29.0/25.0–30.0 R acupuncture Placebo/medicine a, b,f, g,h, k E11 HX Dong 2022 27 27 23.3 ± 2.7 22.3 ± 2.4 N Eacupuncture placebo b, h,j, k, m E19 Jedel 2010 33 34 29.7 ± 4.3 30.2 ± 4.7 R Eacupuncture placebo a, b,h, k,m E20 HX Dong 2021 27 27 23.3 ± 2.7 22.3 ± 2.4 N Eacupuncture placebo c, d,e, f,g E22 Stener 2009 9 6 / / R Eacupuncture placebo a, b,c, d,e, f,g, h,j, k,m E26 Pastore 2011 40 44 28.0 ± 6.3 26.5 ± 5.8 N acupuncture placebo k, m Abbreviations : C Chinese guidelines for PCOS, G&O Gynaecology and Obstetrics, R Rotterdam criteria, N Not mentioned, Eacupuncture Electroacupuncture, ACE Acupoint catgut embedding, Cmedicine Chinese herbs, Tre Treatment, Con control a, Testosterone levels; b, Ferriman-Gallwey score; c, triglyceride; d, high-density lipoprotein; e, Low-density lipoprotein; f, fasting blood-glucose; g, homeostasis model assessment of insulin resistance; h, Body Mass Index, BMI; j,waist hip rate, WHR; k, LH/FSH ratio(luteinizing hormone(LH), follicle-stimulating hormone(FSH); l, pregnancy rate; m, luteinizing hormone(LH) level Characteristics of clinical trials included in the network meta-analysis 28.87 ± 5.83 /27.93 ± 5.62 29.8 ± 4.2/ 30.0 ± 4.5 Abbreviations : C Chinese guidelines for PCOS, G&O Gynaecology and Obstetrics, R Rotterdam criteria, N Not mentioned, Eacupuncture Electroacupuncture, ACE Acupoint catgut embedding, Cmedicine Chinese herbs, Tre Treatment, Con control a, Testosterone levels; b, Ferriman-Gallwey score; c, triglyceride; d, high-density lipoprotein; e, Low-density lipoprotein; f, fasting blood-glucose; g, homeostasis model assessment of insulin resistance; h, Body Mass Index, BMI; j,waist hip rate, WHR; k, LH/FSH ratio(luteinizing hormone(LH), follicle-stimulating hormone(FSH); l, pregnancy rate; m, luteinizing hormone(LH) level Fifteen studies [ 30 , 33 , 36 , 37 , 43 , 45 , 46 , 50 , 54 , 57 , 65 , 78 , 81 , 83 , 85 ] were rated as a high RoB for bias arising from randomization process (7) [ 43 , 45 , 46 , 54 , 57 , 78 , 81 ], due to deviations from intended interventions (6) [ 30 , 33 , 36 , 37 , 50 , 85 ], and the measurement of the outcome (3) [ 65 , 81 , 83 ]. Thirty studies [ 31 , 32 , 35 , 38 , 39 , 41 , 42 , 47 , 51 , 55 , 56 , 58 , 59 , 61 – 64 , 66 , 67 , 69 – 71 , 73 – 77 , 82 , 86 , 87 ] were rated as a low RoB, and the remaining fourteen studies [ 34 , 40 , 44 , 48 , 49 , 52 , 53 , 60 , 68 , 72 , 79 , 80 , 86 , 88 ] as some concerns for inadequate information and indistinct reporting (Fig. 2 ). Fig. 2 The results of the risk of bias assessment The results of the risk of bias assessment The effects of six interventions on the testosterone levels (lower testosterone levels indicate better clinical improvement) were assessed in 36 studies [ 30 , 32 – 40 , 42 , 44 – 47 , 52 , 54 – 58 , 60 , 65 – 69 , 71 , 75 – 79 , 83 , 85 , 87 ] (3219 participants) (Fig. 3 A). Compared with medicine, acupuncture greatly lowered testosterone levels in PCOS patients (MD = 0.69, 95% CrI [0.35, 1.03]) (Fig. 3 B). E-acupuncture (MD = 0.05, 95% CrI: [−0.70,0.08]) and ACE (MD = 0.89, 95% CrI [−0.15,1.94]) decreased testosterone levels compared with medicine, with no statistically significant differences (Fig. 3 B). ACE was the most recommended intervention for reducing testosterone levels (Fig. 3 C). Fig. 3 The results of the network meta-analysis for the testosterone levels. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the testosterone levels. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The effects of five interventions on the FG score (lower FG scores indicate better clinical improvement) were assessed in thirteen studies [ 33 , 34 , 50 , 60 , 67 , 79 – 85 , 87 ] (1986 participants) (Fig. 4 A). E-acupuncture was significantly effective in reducing FG scores in PCOS patients compared with placebo (MD: 1.52; 95% CrI [0.50, 2.53]) (Fig. 4 B), and it was also the preferred choice for reducing FG scores (SUCRA = 84.9%) (Fig. 4 C). In summary, acupuncture-based interventions may ameliorate hyperandrogenism-related clinical manifestations in PCOS patients. Specifically, acupuncture can significantly reduce testosterone levels, and E-acupuncture can significantly improve hirsutism. Fig. 4 The results of the network meta-analysis for the FG score. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the FG score. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The effects of six interventions on BMI (lower BMI indicates better clinical improvement) were evaluated in 31 studies [ 31 , 35 , 36 , 40 , 41 , 44 , 46 , 48 , 50 – 54 , 57 , 58 , 60 – 62 , 65 – 67 , 71 , 76 , 78 – 81 , 83 – 85 , 87 ] (3617 participants) (Fig. 5 A). Compared with medicine, acupuncture (MD: 0.92; 95% CrI [0.27, 1.57]) and ACE (MD: 1.74; 95% CrI [0.34, 3.13]) significantly decreased BMI in PCOS patients (Fig. 5 B). ACE was the most preferred option for lowering BMI (SUCRA = 90.9%) (Fig. 5 C). Fig. 5 The results of the network meta-analysis for the BMI. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the BMI. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value Sixteen studies [ 31 , 35 , 41 , 44 , 46 , 48 , 51 , 61 , 62 , 65 – 67 , 79 , 80 , 84 , 87 ] (1658 participants) described the effects of six interventions on WHR (Fig. 6 A) (lower WHR indicates better clinical improvement). Acupuncture (MD: 0.05; 95% CrI [0.02, 0.08]), E-acupuncture (MD: 0.06; 95% CrI [0.02, 0.11]), and ACE (MD: 0.09; 95% CrI [0.04, 0.14]) significantly decreased WHR in PCOS patients compared with medicine (Fig. 6 B). ACE was identified as the optimal choice for decreasing WHR (SUCRA = 95.3%) (Fig. 6 C). Fig. 6 The results of the network meta-analysis for the WHR. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the WHR. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value Thirteen studies [ 34 , 41 , 44 , 53 , 61 , 65 , 66 , 70 , 79 – 81 , 86 , 87 ] (2226 participants) described the effects of six interventions on triglyceride levels (Fig. 7 A) (lower triglyceride levels indicate better clinical improvement). Acupuncture (MD = 0.14, 95% CrI [−0.08,0.36]) tended to reduce triglyceride levels compared with medicine, with no statistically significant difference (Fig. 7 B). Acupuncture was the most recommended intervention for reducing triglyceride levels (SUCRA = 79.3%) (Fig. 7 C). Fig. 7 The results of the network meta-analysis for the triglyceride level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the triglyceride level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value Eleven studies [ 34 , 44 , 53 , 61 , 65 , 70 , 79 – 81 , 86 , 87 ] (1993 participants) reported the effects of six interventions on HDL levels (Fig. 8 A) (higher HDL levels indicate better clinical improvement). Compared with medicine, acupuncture (MD=−0.05, 95% CrI [−0.26,0.16]), E-acupuncture (MD=−0.04, 95% CrI: [−0.35,0.27]), and ACE (MD=−0.04, 95% CrI [−0.53,0.46]) tended to increase HDL levels, but no statistically significant differences were identified (Fig. 8 B). Acupuncture was possibly the preferred choice for increasing HDL levels (SUCRA = 54.8%) (Fig. 8 C). Fig. 8 The results of the network meta-analysis for the HDL level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the HDL level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value Six interventions were analyzed for their effects on LDL levels (lower LDL levels indicate better clinical improvement) in 11 studies [ 34 , 44 , 53 , 61 , 65 , 70 , 79 – 81 , 86 , 87 ] (1993 participants) (Fig. 9 A). Acupuncture (MD = 0.01, 95% CrI [−0.40,0.43]) and E-acupuncture (MD = 0.016, 95% CrI [−0.33,0.64]) tended to lower LDL-C levels compared with medicine, with no statistically significant differences (Fig. 9 B). E-acupuncture might be the optimal choice to reduce LDL levels (SUCRA = 75.2%) (Fig. 9 C). Fig. 9 The results of the network meta-analysis for the LDL level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the LDL level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value Six interventions were analyzed for their effects on FBG (lower FBG levels indicate better clinical improvement) in 20 studies [ 31 , 39 , 41 , 44 , 46 , 48 , 50 , 51 , 53 , 54 , 61 , 65 , 66 , 71 , 79 – 81 , 83 , 86 , 87 ] (2917 participants) (Fig. 10 A). Acupuncture (MD = 0.14, 95% CrI [−0.14,0.43]), E-acupuncture (MD = 0.17, 95% CrI [−0.24,0.59]), and ACE (MD = 0.17, 95% CrI [−0.31,0.64]) could reduce blood glucose levels compared with medicine, with no statistically significant differences (Fig. 10 B). E-acupuncture was identified as the preferred choice for reducing FBG levels (SUCRA = 64.6%) (Fig. 10 C). Fig. 10 The results of the network meta-analysis for the FBG level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the FBG level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The effects of six interventions on HOMA-IR (lower values indicate better insulin sensitivity) were evaluated in 19 studies [ 31 , 34 , 39 , 41 , 44 , 46 , 48 , 50 , 53 , 61 , 65 – 67 , 71 , 79 , 80 , 83 , 86 , 87 ] (2089 participants) (Fig. 11 A). Medicine significantly reduced HOMA-IR compared with acupuncture (MD: −2.34, 95% CrI [−3.66, −1.02]), E-acupuncture (MD: −2.52, 95% CrI [−3.26, −1.78]), and ACE (MD: −2.59, 95% CrI [−3.93, −1.25]) (Fig. 11 B). Medicine was recommended for lowering HOMA-IR (SUCRA = 100.0%) (Fig. 11 C). Collectively, ACE was the optimal intervention for improving BMI and WHR, and it also positively regulated lipid profiles (reduced TG and LDL, and elevated HDL) and FBG, but the differences had no statistical significance. In contrast, medicine was significantly superior in reducing HOMA-IR. Fig. 11 The results of the network meta-analysis for the HOMA-IR. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the HOMA-IR. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value A total of 37 studies [ 31 , 32 , 37 – 40 , 44 – 46 , 48 , 50 , 54 – 60 , 63 – 67 , 69 – 72 , 74 – 77 , 79 , 80 , 84 , 85 , 87 , 88 ] (3214 participants) assessed the effects of six interventions on LH levels (Fig. 12 A). Acupuncture (MD: 1.23; 95% CrI [0.41, 2.05]) and Cmedicine (MD: 1.81; 95% CrI [0.19, 3.42]) were more effective than medicine in reducing LH levels in PCOS patients (Fig. 12 B). ACE was the best intervention in reducing LH levels (SUCRA = 73.3%) (Fig. 12 C). Fig. 12 The results of the network meta-analysis for the LH level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the LH level. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value Thirty-six studies [ 31 , 35 – 40 , 44 – 48 , 50 , 54 , 57 – 59 , 64 – 69 , 71 , 73 – 76 , 78 – 80 , 83 – 85 , 87 , 88 ] (3242 participants) assessed the effects of six interventions on the LH/FSH ratio (Fig. 13 A). Acupuncture (MD = 0.13, 95% CrI [−0.02, 0.28]), E-acupuncture (MD = 0.21, 95% CrI [−0.10, 0.52]), and ACE (MD = 0.33, 95% CrI [−0.03, 0.68]) tended to reduce the LH/FSH ratio compared with medicine, and the observed differences did not achieve statistical significance (Fig. 13 B). ACE was considered the first recommendation for reducing the LH/FSH ratio (SUCRA = 80.9%) (Fig. 13 C). In conclusion, ACE was the most recommended intervention for ameliorating sex hormone disturbance. Fig. 13 The results of the network meta-analysis for the LH/FSH ratio. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the LH/FSH ratio. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The effects of four interventions on the pregnancy rate were assessed in 14 studies [ 30 , 37 , 38 , 41 , 43 , 45 , 49 , 55 , 63 , 64 , 70 , 72 , 74 , 76 ] (1229 participants) (Fig. 14 A). Compared with medicine, acupuncture achieved a significantly higher pregnancy rate (RR = 1.33; 95% CrI [1.09, 1.62]) (Fig. 14 B). This indicated that acupuncture might increase the probability of pregnancy by approximately 33% relative to medicine, which has potential clinical significance for infertile PCOS patients. Furthermore, acupuncture acted as the first recommendation for increasing the pregnancy rate (SUCRA = 76.6%) (Fig. 14 C). Fig. 14 The results of the network meta-analysis for the pregnancy rate. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The results of the network meta-analysis for the pregnancy rate. A Network diagram of eligible comparisons. B The league table for the relative effects of all treatments. C The SUCRA value The consistency test was performed, and P  ≥ 0.05 indicated good consistency among the included studies. For global consistency, the results revealed that all P-values were ≥ 0.05, except for HOMA-IR. No local inconsistency was detected via the node-splitting method, except for HOMA-IR (medicine vs. ACE) (Table S2). Different outcomes had significant differences in heterogeneity. Outcomes with high heterogeneity (I²>50%) mainly included testosterone (72.50%−99.90%), FG scores (58.00%−99.80%), BMI (58.10%−99.80%), WHR (62.00%−78.70%), triglycerides (88.50%), HDL (88.60%), LDL (51.00%−51.70%), FBG (71.90%−91.40%), HOMA-IR (53.50%−96.90%), and LH/FSH (59.40%−93.70%). Outcomes with low or no significant heterogeneity (I²≤50%) were relatively scarce, primarily under certain intervention comparisons (Table S3). Sensitivity analysis was conducted by excluding studies with a single-group sample size of less than 20 to explore the robustness of the results, with a total of three studies removed [ 78 , 80 , 87 ]. The results remained essentially consistent, except for the league table results for testosterone and LH, which were inconsistent with the original findings (Supplementary Figure S1). Moreover, the funnel plot revealed that studies with identical intervention comparisons were approximately symmetrically distributed around the pooled effect size with no obvious asymmetry. This suggests that publication bias was unlikely to exert a significant impact on the results (Supplementary Figure S2).

No single acupuncture method is the optimal method for all indicators in PCOS patients. Clinically, clinicians may select acupuncture methods based on patients’ primary clinical concerns (e.g., ACE for weight management and testosterone reduction, E-acupuncture for hirsutism and blood glucose control, and body acupuncture for improving pregnancy rate). These findings remain to be further verified by rigorously designed studies with longer follow-ups due to limitations of existing clinical studies and evidence.

This NMA comprehensively compared three common acupuncture methods (body acupuncture, E-acupuncture, and ACE) for treating PCOS across four core outcome domains (hyperandrogenism, metabolic status, sexual hormone disturbance, and infertility). The main finding is that no single acupuncture method was universally superior. Instead, each method exhibited distinct advantages in improving specific outcomes, reflecting the complex, multifaceted pathophysiology of PCOS and highlighting the need for personalized treatment strategies. The SUCRA value could present the most recommended intervention. However, management strategies of PCOS should also be based on a careful consideration of pairwise comparison results in clinical practice. Previous systematic reviews and meta-analyses also showed that ACE significantly lowers serum testosterone levels, consistent with the results in this paper [ 89 ]. As recommended by the guideline, testosterone should be preferably detected to assess hyperandrogenism in PCOS diagnosis [ 11 ]. CYP17, a key enzyme in androgen synthesis expressed in the ovary and adrenal glands, plays a critical role in androgen overproduction. Ovarian hormone disorders can trigger excessive release of hypothalamic gonadotropin-releasing hormone (GnRH), leading to elevated luteinizing hormone and subsequent androgen excess [ 90 ]. A relevant animal experiment showed that ACE can down-regulate the expression of luteinizing hormone receptors and CYP17 mRNA [ 91 ], suggesting a potential mechanistic pathway. However, this finding is derived from animal models and may have translational limitations, so it cannot be directly generalized to humans. Meanwhile, the sensitivity analysis indicated that the league table results for testosterone were inconsistent with the previous findings, which may be attributed to the reduced impact of E-acupuncture on the outcomes due to the small-study effect. Caution should be exercised when considering this result in clinical practice. Hirsutism, often a sign of androgen excess, is defined as excessive growth of body hair in a typical male pattern in a female, which is assessed by FG scores. Over 85% of people with hirsutism suffer from PCOS and idiopathic hyperandrogenism [ 92 ]. Jing Zhou et al. argued that testosterone levels can be reduced by E-acupuncture [ 93 ]. Therefore, the decrease in FG scores in the E-acupuncture group may result from the decline in androgen. However, Yajie Ge found that some PCOS patients typically present with hirsutism but no abnormal changes in androgen levels, indicating that the mechanism of E-acupuncture for treating hirsutism does not depend solely on androgen reduction but may involve other pathways. This requires further exploration. In addition, this NMA revealed that ACE outperformed others in reducing BMI and WHR. According to relevant data, 30%−70% of PCOS patients are accompanied by overweight/obesity and visceral obesity [ 94 ]. It has been reported that ACE is superior to other acupuncture methods in weight loss [ 95 , 96 ]. A potential mechanism may be that the insertion and embedding of absorbable catgut sutures induce local tissue responses, which could lead to adipocyte death, fat liquefaction, and reduced adipocyte volume. Additionally, ACE may moderately increase local temperature, potentially enhancing basal metabolic rate and energy consumption [ 97 ]. However, these mechanisms are based on limited evidence and require further validation in rigorous clinical and basic studies. Besides, this NMA showed that acupuncture was considered the most recommended intervention in lowering triglyceride and elevating HDL levels. Contemporary research suggests that adiponectin is closely related to lipoprotein metabolism, especially HDL and triglyceride, which can induce an increase in HDL and a decrease in LDL [ 98 ]. Acupuncture plays a role in lowering lipids by elevating adiponectin levels [ 99 , 100 ]. Sterol-regulatory element binding proteins (SREBPs) (SREBP1a, SREBP1c, and SREBP2) are crucial for regulating lipid metabolism and are also key connection points of various metabolic diseases [ 101 ]. For LDL, E-acupuncture ranked first in reducing LDL levels in PCOS patients. E-acupuncture can also ameliorate hyperlipidemia by inhibiting hepatic SREBP-2 expression in rats [ 102 ], enabling E-acupuncture to reduce LDL levels. For fasting blood glucose, E-acupuncture was also identified as the optimal choice in reducing FBG levels in PCOS patients. E-acupuncture can also promote the secretion of insulin to lower blood glucose in animals by activating cholinergic nerves and stimulating the release of β-endorphin and other endogenous opioid peptides [ 103 – 105 ]. Notably, the mechanism derived from animal experiments needs to be verified in human studies due to potential species differences. Therefore, we look forward to more rigorous, large-scale human studies with standardized protocols and long-term follow-up in the future. In this way, these proposed mechanisms can be further explored and confirmed, and the translation of preclinical insights into effective clinical strategies can be facilitated. This paper demonstrated that medicine was the most recommended intervention in reducing HOMA-IR. In contrast, Liu et al. have proved in a meta-analysis that acupuncture achieves a greater mean reduction in HOMA-IR vs. sham or medicine [ 106 ]. Inconsistent with our findings, another study identified E-acupuncture as the most effective intervention for lowering HOMA-IR values in women with PCOS-related IR based on SUCRA values [ 107 ] Such a discrepancy may be attributed to heterogeneity in the detection methods among studies. The 2023 PCOS Guideline has mentioned that although IR is considered a key pathophysiological factor in PCOS, routinely available measures are so inaccurate that clinical measurement of IR is not recommended [ 11 ]. Remarkably, no unified diagnostic criteria for IR have been established to date. More importantly, there is a lack of specific diagnostic criteria for IR applicable to patients with PCOS. Thus, the degree of IR may vary among the subjects included in our study. Metformin, a recommended drug for IR in the 2023 PCOS Guideline [ 11 ], has been widely prescribed to IR sufferers in current clinical applications. Research suggests that the effect of metformin on IR depends most possibly on direct and indirect effects on mediators [ 108 ]. Meanwhile, it should not be overlooked that HOMA-IR exhibited both global and local inconsistencies. They may stem from several interconnected factors. First, variability in baseline participant characteristics across the included studies could be a primary contributor. Patients with PCOS exhibit substantial heterogeneity in IR severity, BMI, and hormonal profiles at baseline; imbalances in these variables between direct and indirect comparison groups for HOMA-IR may have undermined the consistency of pooled results. Second, differences in interventions might play a critical role. Variations in the dosage, administration frequency, and treatment duration of the medicine and ACE across studies could lead to different effects on insulin sensitivity, thereby generating discrepancies between direct and indirect evidence. For sexual hormone disturbance, ACE was regarded as the preferred intervention in reducing LH levels and the LH/FSH ratio. Excessive androgen secretion results in negative feedback to the hypothalamus-pituitary-ovary axis, causing GnRH release rhythm disorders and increasing LH levels [ 109 ]; it also leads to an imbalance of the LH/FSH ratio [ 110 ]. In this paper, ACE was also the most preferred measure in lowering testosterone levels. Therefore, an ACE-induced decrease in testosterone levels in PCOS may indirectly lead to a decline in LH levels and LH/FSH ratio. Moreover, blood omentin-1 levels in PCOS patients are obviously lower than in healthy people, which is negatively correlated with serum LH levels [ 111 ]. A clinical trial found that ACE can improve sexual hormone disturbance in PCOS patients by increasing omentin-1 [ 112 ]. Therefore, decreases in LH levels and LH/FSH ratio in the ACE group might also be associated with the effect of ACE on omentin-1. However, it should be emphasized that the identified association between Omentin-1 and the LH/FSH ratio does not establish a direct causal mediating relationship. Additionally, the sensitivity analysis demonstrated inconsistency between the league table results and the original results for LH, which might be explained by the small-study effect. Thus, careful consideration is required for clinicians in clinical application. Acupuncture was possibly the preferred choice in increasing the pregnancy rate in PCOS patients. Chen et al. confirmed that acupuncture has a positive effect on the pregnancy rate [ 113 ], consistent with our findings. In terms of mechanism, acupoint stimulation inhibits GnRH and subsequent excessive LH release [ 15 ], contributing to normal ovulation. Meanwhile, acupuncture can improve the environment for conception by increasing the expression of progesterone and estrogen receptors on the endometrium and increasing endometrial thickness and proliferation [ 114 ]. However, potential confounding factors (e.g., duration of infertility, age) may exist across trials, which could contribute to heterogeneity in pregnancy rate outcomes. Therefore, the effect of acupuncture should be interpreted with awareness of these potential confounding elements. Unfortunately, significant differences in heterogeneity were observed across the outcomes in this study. The potential reasons may be summarized as follows: First, significant variations in acupuncture methods among included studies, including differences in acupoint selection, treatment frequency and duration, and acupuncture techniques. Second, diverse baseline characteristics of participants, such as differences in disease severity, age distribution, and comorbidities, may lead to differential responses to acupuncture. Third, inconsistencies in outcome measurement methods and assessment criteria, such as the use of different detection kits for endocrine indicators. Additionally, potential confounding factors such as acupuncturist experience levels may have further contributed to the observed heterogeneity. This is the first NMA comparing the effects of different acupuncture methods on hyperandrogenism, metabolic disorder, sexual hormone disturbance, and infertility in PCOS patients. The NMA design has prominent methodological strengths: It enables the inclusion of a wide range of acupuncture interventions and establishes a comprehensive comparative framework that integrates both direct comparisons and indirect comparisons. This not only increases the statistical power by synthesizing more evidence but also rigorously ranks the efficacy of different acupuncture methods. Moreover, SUCRA values provide intuitive evidence for the selection of the optimal acupuncture for PCOS patients. However, there were several limitations. First, the accuracy and applicability of our findings may be affected by the small sample size and the limited number of studies. Specifically, the FG score (only 13 studies, 1986 participants) and pregnancy rate (14 studies, 1229 participants) had fewer included studies and smaller sample sizes, which may reduce the statistical precision of these comparisons. Second, it should be noted that the pooled results based on RCTs may be affected by variations in study design, such as inconsistent acupoint selection and variable treatment durations, which means the results should be interpreted cautiously. Third, the combination of different acupuncture methods was not involved such as E-acupuncture combined with ACE due to the limited number of studies. Fourth, high heterogeneity was detected, and heterogeneity in methodology may undermine the reliability of metabolic and hormonal-related outcomes. Finally, this NMA was confined to English- and Chinese-language studies, bringing about selectivity bias. Consequently, further thorough and high-quality studies are needed to verify our findings.

Polycystic ovary syndrome (PCOS) is a prevalent endocrine condition characterized by polycystic ovarian morphology, hyperandrogenism, and irregular menstruation in women [ 1 ]. It affects approximately 10%−13% of women globally [ 2 ]. Most of the patients present with difficulty in monthly ovulation, resulting in infertility, with an infertility rate up to 70%−80% [ 3 ]. In addition, PCOS is frequently complicated by insulin resistance (IR) and metabolic syndrome, which elevate the risks of type 2 diabetes and cardiovascular diseases. This imposes a substantial burden on both individual health and healthcare systems [ 4 – 6 ]. Evidence from systematic reviews and meta-analyses have demonstrated that women with PCOS exhibit a threefold elevation of risk of type 2 diabetes compared with those without PCOS [ 7 ]. In America, the additional total economic burden of PCOS on account of pregnancy-related and long-term morbidity was estimated to be $4.3 billion in 2020 [ 6 ]. Given the severe clinical complications and heavy disease burden of PCOS, developing effective management strategies for PCOS is therefore an urgent clinical priority. Clinical management of PCOS is highly individualized based on phenotypic characteristics and patient needs [ 8 ]. For obese/overweight phenotypes, lifestyle modifications (diet and exercise) are first-line interventions [ 9 ]. For hyperandrogenic phenotypes (with hirsutism/acne), combined oral contraceptives are preferred [ 8 , 10 ]. For infertile phenotypes without other infertility factors, letrozole is recommended as first-line ovulation induction therapy [ 11 ]. However, it has been reported that combined oral contraceptives not only exacerbate IR but also lead to a higher risk of inflammatory and coagulation disorders in PCOS women [ 12 ]. Notably, these risks vary with the patient’s individual characteristics (such as age, and metabolic status). Combined oral contraceptives are still considered an acceptable first-line intervention for hyperandrogenic phenotypes in many clinical guidelines due to their significant therapeutic effects on hirsutism and acne [ 11 ]. Meanwhile, the safety profile of letrozole in pregnancy has not been fully clarified [ 13 ]. Therefore, the vast majority of modern treatments for PCOS are symptomatically oriented and associated with some adverse effects, necessitating safe therapeutic approaches with comprehensive efficacy. Acupuncture is a complementary treatment method that involves the stimulation of specific acupoints to modulate key physiological pathways relevant to PCOS: It regulates the hypothalamic-pituitary-ovarian axis to improve hormonal balance, and enhances insulin sensitivity by modulating insulin signaling pathways, achieving dual-directional regulation of nerves innervating skin, muscles, and reproductive organs [ 14 ]. Various acupuncture-based techniques have been used in the clinical management of PCOS, such as body acupuncture (manual stimulation of acupoints with needles, with short-term effects), electro-acupuncture (E-acupuncture, needle stimulation combined with low-frequency electric current, enhanced and sustained stimulation), and acupoint catgut embedding (ACE, absorbable catgut implanted into acupoints to provide prolonged stimulation over 1–2 weeks) [ 15 , 16 ]. Meanwhile, acupuncture is widely recognized as a safe clinical treatment since serious adverse events extremely rarely occur. For instance, one retrospective study has estimated that the rate of significant complications is 0–1.1.1/10,000 [ 17 ]. Similarly, synthesized evidence from 11 systematic reviews also suggested that acupuncture caused no serious adverse events, and only mild, non-harmful adverse events were documented in individual cases [ 18 ]. Traditional meta-analyses comparing acupuncture with medicine have been conducted. However [ 19 ], no comparisons among different acupuncture techniques have been conducted. Furthermore, these studies only focused on a single phenotypic characteristic of PCOS, such as metabolic disturbances [ 20 ] or infertility [ 21 ]. High-quality comparative evidence regarding the clinical efficacy of these different acupuncture techniques for PCOS is still lacking in available studies. Hence, the most appropriate acupuncture method in clinical practice remains to be determined. Network meta-analysis (NMA) synthesizes both direct evidence (from head-to-head comparisons of acupuncture techniques) and indirect evidence (from common controls such as placebo or medicine) to calculate relative effects. NMA compensates for the lack of head-to-head randomized trials of different acupuncture techniques. This enables comprehensive ranking of all interventions, which is not feasible in traditional pairwise meta-analysis, providing more robust evidence for clinical decisions [ 22 , 23 ]. Therefore, this study aims to compare the effects of different acupuncture methods on PCOS patients via an NMA, thereby offering scientific evidence-based guidance for clinical decisions.

Supplementary Material 1 Supplementary Material 1