The study was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) for systematic review protocols (38) and meta-analysis checklist (39). This review protocol has been registered in PROSPERO (ID: CRD42023426305).

An electronic literature search was performed using the following databases for articles published in English and Chinese as of May 2023: PubMed, Web of Science, Cochrane Library, Embase, China National Knowledge Infrastructure, Wanfang Data, China Science and Technology Journal Database, and Chinese Biomedical Database. Medical Search Heading (MeSH) terms and their synonyms (free text) were combined as search terms using the Boolean operators “AND” and “OR.” The following MeSH terms were used: “Parkinson Disease,” “Acupuncture,” “Acupressure,” “Massage,” “Cupping Therapy,” “moxibustion,” “acupoint,” “acupoint injection,” “auricular acupuncture,” “scraping therapy,” and “randomized controlled trial.” The detailed search strategy is shown in the box. For more details in Supplementary material 3.

Search items in PubMed.

Two independent authors (WH and HZ) revised the titles and abstracts of the searched papers to determine suitable studies. Then, the two authors revised the full texts of the retrieved reports independently. Any conflicts between authors were solved by the third author (ZW).

Two independent authors (WH and HZ) extracted data. The following data were extracted in a predefined data collection form: first author, year of publication, language, sample size, demographic data of participants, baseline characteristics of patients, diagnostic criteria, inclusion and exclusion criteria for participants, experimental and control interventions, course of treatment, frequency, location of the study, outcomes, and safety. Any conflicts between authors were solved by the third author (ZW).

The Cochrane risk-of-bias tool (ROB 2) for RCTs was used to assess potential bias in the included studies (40). This risk-of-bias tool consists of the following seven domains: (I) random sequence generation, (II) allocation concealment, (III) blinding of participants and personnel, (IV) blinding of outcome assessment, (V) incomplete outcome data, (VI) selective reporting, and (VII) other bias (e.g., we assessed trials with no reported monitoring of self-acupressure procedures as having a high risk of compliance bias). Each RCT was categorized as having a low, high, or unclear risk of bias in each domain.

Review Manager (RevMan, version 5.4; The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark) and STATA (version 17.0) software were used to quantitatively analyze the included studies, as follows. (I) Combination of effects: continuous outcomes (HAMD, MMSE, PDSS, MoCA, PSQI, and PAC-QoL) were evaluated using weighted mean differences (WMDs), whereas dichotomous outcomes (rates of improvement) were assessed using risk ratios (RRs). The 95% confidence interval (CI) was evaluated for all effect sizes, with a 95% CI excluding the point of no effect, indicating significance. When the median and first and third quartiles were provided, the mean and standard deviation were estimated using the formula developed by Wan et al. (41). (II) Heterogeneity test: the Q-test was performed to assess heterogeneity among studies. If p > 0.10, the results of multiple similar studies were considered homogeneous. If p > 0.10 and I² ≥ 0 and ≤ 50%, a fixed effects model was used for the integrative analysis of studies. If p ≤ 0.10 or I² > 50%, the results of multiple similar studies were considered heterogeneous, and a sensitivity analysis was performed (42). Articles were removed sequentially to observe changes in heterogeneity, WMDs, and RRs. If heterogeneity was altered after the removal of an article, it was considered a source of heterogeneity, and the underlying reason was analyzed. In contrast, if heterogeneity remained unaltered, a random effects model was used for a more conservative evaluation of the intervening effects. Subgroup analyses were performed based on acupoint methods (single acupuncture versus single moxibustion versus acupuncture combined with other therapies). When ≥10 studies were included in the meta-analysis, Egger’s test was used to assess publication bias, with p < 0.05 indicating significance (43).

Although the studied specimens were human, this is a secondary analysis of published articles and does not involve ethical issues.

From the eight databases, 1711 articles were selected. Of these, 731 articles were removed as duplicates. Of the remaining 980 articles, many articles were excluded because they failed to meet the inclusion criteria. Finally, 60 full-text articles that met the eligibility criteria were selected. According to the inclusion and exclusion criteria, 42 articles were eventually included in the review (44–85). The study selection process is illustrated in Figure 1.

The baseline information of the 42 studies included in this systematic review and meta-analysis is summarized in Supplementary Table 1. Of the 3,120 patients with PD, 1556 were assigned to the experimental group and 1,564 were assigned to the control group, with 1743 male patients and 1,377 female patients. Of the 42 studies, 24 were randomized into groups using the random number table method (45–49, 52–78, 81–85), two were randomized according to the order of treatment (63, 73), two were grouped by computer random sampling (69, 72), one (50) was grouped by convenient sampling selection, one was grouped by numerical random sampling (64), and one was grouped by envelope (85). The remaining 11 studies (51, 55–59, 65, 68, 77, 78, 81) did not mention the specific randomization method. Of the 42 studies, none mentioned allocation concealment. Given the specificity of acupuncture procedures, the studies displayed a high risk of bias in terms of the blinding of participants and personnel; although one study (85) used a single-blinding method, no other studies mentioned the blinding method that was used. All 42 studies reported outcomes, and no studies demonstrated attrition bias. In summary, the risk of bias in random sequence generation was low. The risk of bias of allocation concealment was moderate. The risks of bias of blinding of participants and personnel and outcome assessment were high. The risks of bias of missing outcome data and selective reporting were low. The methodological quality of studies is shown in detail in Figure 2. A summary of the basic characteristics of the included studies is shown in Supplementary Table 2. All of the studies mentioned sample size. 41 studies mentioned the proportion of female in their research, and women were more than men in only five studies. 37 of 42 studies mentioned the disease duration, and 33 studies had patients with disease of less than 10 years.

Thirteen RCTs reported the use of HAMD for assessing depression status (58, 59, 61, 65, 66, 73, 74, 76, 78–81, 84). HAMD scores were significantly better in patients treated with acupoint therapy combined with CM or NDT than in those treated with CM or NDT alone. A random effects model was used for the analysis (WMD = −2.14; 95% CI, −2.9 to −1.38; Z = 5.54; p < 0.00001). Heterogeneity was observed among the studies (chi-squared = 74.59; df = 12 [p < 0.00001]; I² = 83.9%) (Figure 7). The results of a sensitivity analysis were not significantly different from the aforementioned results.

In the subgroup analysis, 10 RCTs (59, 65, 66, 73, 74, 76, 78, 79, 81, 84) reported comparisons between a combination of acupoint therapy with drugs versus drugs alone, and 3 RCTs (58, 61, 80) reported comparisons between acupoint therapy combined with other therapies versus other therapies alone. The fixed effects model results suggested strong heterogeneity in the 10 studies. After deleting two studies, heterogeneity still existed; a random effects model was therefore used for the analysis (WMD = −2.1; 95% CI, −3.15 to −1.05; Z = 3.91; p < 0.00001; heterogeneity test: chi-squared = 73.68, df = 9 [p = 0], I² = 87.8%) (Supplementary Figure 10). The HAMD scores of patients treated with a combination of acupoint therapy and drugs were better than those of patients treated with drugs alone.

Acupoint therapy combined with other therapies was more effective than other therapies alone in improving HAMD scores (WMD = −2.26; 95% CI, −2.77 to −1.75; Z = 8.71; p < 0.00001; heterogeneity test: chi-squared = 0.9, df = 2 [p > 0.05], I² = 0%) (Supplementary Figure 11). Heterogeneity was not assessed because only three studies reported comparisons between acupoint therapy combined with other therapies versus other therapies alone. Additionally, a meta-regression analysis of different acupoint therapies using “type of therapy” as an ordinal variable revealed no differences between acupuncture and moxibustion or acupoint sticking (Supplementary Figure 12).

A funnel plot was drawn to investigate publication bias; its symmetry indicated no publication bias (Supplementary Figure 13). Similarly, Egger’s test (p > 0.05) indicated no publication bias (Supplementary Figure 14).

Two RCTs (45, 46) reported the use of MMSE, which assesses intellectual status and cognitive function. MMSE scores appeared to be better in patients treated with a combination of acupoint therapy and other therapies than in those treated with other therapies alone, but this effect was not significant. A random effects model was used for the analysis (WMD = 2.33; 95% CI, −0.99 to 5.65; Z = 1.37; p > 0.05). Heterogeneity was observed among studies (chi-squared = 14.45; df = 1 [p = 0.0001]; I² = 93%) (Figure 8). The results of the sensitivity analysis were not significantly different from the aforementioned results.

Eight RCTs (45, 59, 68, 69, 71, 75, 77, 80) reported the use of MoCA, which assesses cognitive function. MoCA scores were significantly better in patients treated with a combination of acupoint therapy with other therapies than in those treated with other therapies alone. A random effects model was used for the analysis (WMD = 2.73; 95% CI, 0.63 to 4.84; Z = 2.54; p < 0.00001). Heterogeneity was observed among studies (chi-squared = 179.26; df = 7 [p < 0.0001]; I² = 96%) (Figure 9). Furthermore, meta-regression analysis was conducted using “type of therapy” as an ordinal variable, there were no differences between acupuncture and acupoint sticking (Supplementary Figure 15).

Nine RCTs (44, 59, 60, 62, 64, 66, 82, 83, 85) reported the use of PDSS, which assesses cognitive function. PDSS scores were significantly better in patients treated with a combination of acupoint therapy with drugs than in those treated with drugs alone. A random effects model was used for the analysis (WMD = 15.87; 95% CI, 6.27 to 25.47; Z = 3.24; p = 0.001). Heterogeneity was observed among studies (chi-squared = 170.54; df = 8 [p < 0.0001]; I² = 95.3%) (Figure 10). Meta-regression analysis using “type of therapy” as an ordinal variable was conducted, there were no differences between acupuncture and acupoint sticking (Supplementary Figure 16). The funnel plot test for publication bias is shown (Supplementary Figure 17). Egger’s bias test (p > 0.05) indicated no publication bias (Supplementary Figure 18).

Six RCTs reported the use of PSQI, which assesses sleep quality. A random effects model revealed that sleep quality was significantly better in patients treated with a combination of acupoint therapy with other therapies than in those treated with other therapies alone (WMD = −3.28; 95% CI, −4.55 to −2.00; Z = 5.03; p < 0.00001). Heterogeneity was observed among studies (chi-squared = 101.91; df = 5 [p < 0.0001]; I² = 95%) (Figure 11). The results of the sensitivity analysis were not significantly different from the aforementioned results. When meta-regression analysis was conducted used “type of therapy” as an ordinal variable, the result showed no difference between acupuncture and acupoint sticking (Supplementary Figure 19). The funnel plot test for publication bias is shown (Supplementary Figure 20). Egger’s bias test indicated no publication bias (Supplementary Figure 21).

Five RCTs reported the use of PAC-QoL, which assesses quality of life. A random effects model revealed that PAC-QoL was significantly better in patients treated with a combination of acupoint therapy with other therapies than in those treated with other therapies alone (WMD = −19.48; 95% CI, −30.47 to −8.49; Z = 3.47; p = 0.0005). Heterogeneity was observed among studies (chi-squared = 149.55; df = 4 [p < 0.00001]; I² = 97.3%) (Figure 12). Meta-regression analysis using “type of therapy” as an ordinal variable,” showed there were no differences between treatments (Supplementary Figure 22). The results of the sensitivity analysis were not significantly different from the aforementioned results. Egger’s bias test indicated no publication bias (Supplementary Figure 23). The results of the meta-analysis for all outcome indicators were summarized in Supplementary Table 3.

Safety conditions were reported in 12 RCTs (44, 64, 65, 66, 69, 71, 74, 75, 78, 79, 82, 85) and adverse effects were not mentioned in 30 studies. Of the 12 studies that mentioned adverse effects, seven trials (21, 65, 69, 70, 78, 79, 82) reported a negative association with acupoint therapies; the other five trials (64, 66, 71, 75, 85) reported no such association. The observed adverse effects were mainly gastrointestinal symptoms (such as discomfort, vomiting, and nausea), subcutaneous hematoma, headache, or were related to the insertion of needles. All of these adverse effects were mild and showed signs of reversibility.

To our knowledge, our study is the first to comprehensively examine acupoint therapy for PD-NMS, although some symptoms (e.g., dizziness, bladder urgency, and apathy) were not reviewed because of a lack of original data. Our study combined data from 42 eligible studies that covered an array of common PD-NMS; our findings provide potential evidence for several non-pharmaceutical therapies for PD-NMS. Previous studies have mostly focused on motor symptoms (97, 98) or have reviewed the effects of complementary therapies (including acupuncture) for motor and non-motor symptoms but did not provide a quantitative analysis (99). Therefore, our study represents a comprehensive assessment of the currently available evidence regarding PD-NMS.

Given that acupoint therapy is a manipulated intervention, blinding for practitioners is hard to implement in most studies (even in studies outside of PD) (100). Consequently, concerns about the risk of bias in our study mostly existed in the domains of blinding—particularly in participant and personnel blinding—and weakened the strength of the evidence, and thus of our conclusions. Furthermore, in the present study, we included acupoint sticking therapy as an intervention method, and confirmed that acupoint sticking combined with conventional care was more effective for PD-NMS than conventional care alone. Some of the pooled results also continued to exhibit high levels of heterogeneity and were unstable in the sensitivity analyses, despite restricting the criteria for study enrollment and performing subgroup analyses. Another major limitation of the study was that all trials were conducted in China; this may affect the generalizability of our findings. In addition, there were no studies on Guasha, cupping, or acupressure interventions in PD-NMS, so we were unable to assess the efficacy of these treatments for PD-NMS. Our findings on the effectiveness of acupoint therapy in PD-NMS should therefore be interpreted with caution.

Some results from the present study may provide guidance for clinical applications. First, for patients with PD-related depression or decreased quality of life or cognition, acupoint therapy may be a good choice. Second, acupoint therapy combined with CM may be suitable for treating PD-related depression, cognitive function, insomnia, and constipation. Third, acupoint sticking therapy may be used to treat constipation.