The effects of ear acupressure therapy on obstetric and gynecological pain in women: a systematic review and meta-analysis

1. Inclusion and exclusion criteria

In this study, the selection criteria were based on the PICO-SD framework (Population, Intervention, Comparison, Outcome, and Study Design) [11]. The population (P) was women; the intervention (I) included ear acupressure, acupuncture, or auriculotherapy; the control group (C) consisted of usual obstetric care or placebo therapy; the outcome (O) was pain; and the study design (SD) was randomized controlled trials (RCTs).

The inclusion criteria consisted of studies reported in either Korean or English, with at least two measurements taken before and after the intervention. For effect size calculation, the first measurement taken at the endpoint of the intervention, rather than intermediate measurements during the intervention, was used. Only studies that provided sample size, mean, and standard deviation in the results were selected to ensure accurate effect size calculations.

The exclusion criteria were studies involving participants under the age of 19, studies where the intervention site was not the ear, studies that did not report pain as an outcome variable, studies not published in Korean or English, studies where the full text was unavailable, and single-group studies without a control group.

2. Search strategy and data sources

The literature search was conducted across seven electronic databases: PubMed, Cochrane, EMBASE-OVID, CINAHL, Web of Science, Scopus, and RISS between June 19, 2024, and July 19, 2024. The search terms included MeSH terms such as ‘women,’ ‘Acupuncture, Ear,’ and ‘pain,’ along with relevant natural language terms such as ‘ear pressing’, ‘auricular pressing’, ‘auriculoacupressure’. Filters were applied to limit results to studies involving humans, published in Korean or English, RCTs, and full-text available articles published in each database, either online or offline, by July 31, 2024. The search was performed by a librarian and one researcher, and another researcher cross-checked the search results.

The meta-analysis for this study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 checklist(accessed on June 30, 2024, from https://prisma-statement.org/PRISMAStatement/Checklist.aspx). The search protocol was registered on PROSPERO, the International Prospective Register of Systematic Reviews (registration No. CRD42024598617), on October 18, 2024 (https://www.crd.york.ac.uk/prospero).

3. Data extraction

The bibliographic data of selected studies were organized in Excel, with each study assigned a serial number and consolidated in a sheet by author, publication year, title, and journal. Duplicate records were filtered by title and author. Titles, abstracts, and full texts were reviewed to select studies based on inclusion and exclusion criteria, with exclusion reasons noted (e.g., language, target population, study design, and statistical data). Consistency was ensured through cross-checking at each screening stage.

The selected studies were organized in Excel, with each study assigned a serial number and documented by author, publication year, title, and journal. Duplicates were filtered by title and author. Titles, abstracts, and full texts were reviewed based on inclusion and exclusion criteria, with exclusion reasons (e.g., language, target population, study design) recorded and cross-checked for consistency.

Eligible studies were documented in a coding book, detailing authorship, publication year, institutional review board (IRB) approval, funding status, participant characteristics (sample size, group distribution, pain type), research design, and intervention specifics (e.g., stimulus type, ear acupressure tool, acupoint locations, control intervention, session details, outcome measures, and quality scores).

In the meta-analysis, the experimental group consisted of participants receiving ear acupressure, and the control group included those receiving no intervention or standard care. Sample sizes were calculated by summing participants across groups. Data were independently coded by two researchers, with discrepancies resolved through review and consensus. For meta-analysis, the experimental group included those receiving ear acupressure, and control groups receiving no intervention or standard care. Sample size was calculated by summing participants from both groups. Data were independently coded by two researchers, with discrepancies resolved through review and consensus.

4. Quality assessment

The quality of the selected studies was assessed using the Revised Checklist for Randomized Controlled Trials from the Joanna Briggs Institute (JBI) [12]. The JBI Checklist for Randomized Controlled Trials consists of 13 items, divided into two categories: risk of bias to assess internal validity (10 items) and statistical conclusion validity(3 items). Each item scored 1 point for ‘clear,’ and 0 points for ‘unclear,’ ‘no,’ or ‘not applicable.’

In this study, all 16 studies included in the analysis were RCTs, with a mean quality score of 10.19 ± 1.72. Among the items, ‘Q5. Were those delivering the treatment blind to treatment assignment?’ was not clearly reported in any of the studies. ‘Q4. Were participants blind to treatment assignment?’ was reported clearly in seven studies (Study ID: 3, 4, 5, 6, 7, 9, 10). Similarly, ‘Q2. Was allocation to treatment groups concealed?’ was reported clearly in nine studies (Study ID: 3, 4, 5, 6, 7, 9, 10, 11, 14), as was ‘Q7. Were outcome assessors blind to treatment assignment?’ (Study ID: 1, 2, 5, 6, 7, 8, 9, 10, 14). The detailed results are presented in Appendix 1, 2.

5. Statistical analysis

The overall effect size of the studies included in the analysis was calculated, and the validity of the meta-analysis was evaluated using MIX 2.0 Pro Ver. 2.0.1.6 (BiostatXL) [13]. The effect sizes of individual studies and the pooled overall effect size are expressed as Hedge’s g. Hedges's g is a statistical measure used to assess the effect size in inferential statistics. It is typically calculated using the square root of the mean square error derived from an analysis of variance that tests for differences between two groups. The interval estimates were calculated as 95% confidence intervals (CI), and the weights for each effect size were derived using the inverse of variance [14]. The pooled overall effect size is estimated using a random effects model, which adjusts the weights to account for both within-study variability and between-study heterogeneity [15].

The heterogeneity among the included studies was assessed using Cochrane’s Q-statistics and Higgins' I², which represents the proportion of variance across studies due to heterogeneity [16]. A p-value of less than 0.05 in the Q-statistics or an I² value exceeding 50% was interpreted as indicating significant heterogeneity [17]. To identify the factors contributing to heterogeneity in studies on OBGY pain, subgroup analysis, meta-regression, and exclusion sensitivity analyses were conducted [18]. Meta-regression analysis is a statistical technique that involves the regression of summary estimates from individual studies on one or more covariates. In this study, we used a fixed-effect linear meta-regression model, predicated on the assumption that all heterogeneity between studies could be adequately explained by the covariates included in the regression model [19]. Additionally, publication bias was assessed using funnel plots, trim-and-fill plots, Begg’s test, Egger’s regression, and the trim-and-fill method to adjust and validate the overall effect [20].

1. Characteristics of the included studies

In this study, the selection of target studies for analysis followed a three-step process in accordance with the PRISMA guidelines, as depicted in Appendix 3. Among the 16 studies included, 11 were published in Asia and 11 were published after 2019. A total of 15 studies had obtained IRB approval, and 11 studies had received funding. All studies employed a RCT design, and five studies included more than 100 participants. Regarding the types of pain addressed, seven studies focused on labor pain during normal delivery, four studies on cesarean section, two studies on surgical procedures such as episiotomy, and three studies on other types of breastfeeding pain, pregnancy-related back pain, and premenstrual symptoms.

In terms of intervention methods, 13 studies used acupressure, while three studies used acupuncture. Regarding intervention tools, seven studies utilized seeds, and three studies used needles. Ten studies applied fewer than five stimulus points to reduce pain, and four studies used alternative or placebo interventions for the control group. The duration of the intervention was more than two days in six studies, and six studies included more than two sessions. The intervention time per session was more than 75 seconds in four studies. For outcome measurement, two studies evaluated the long-term effect of the intervention. The quality assessment scores were 10 or higher in 11 studies (Table 1, 2).

Table 1.

Descriptive Summary of the Included Studies

Table 2.

Subgroup Analysis Regarding Ear Acupressure on Obstetric and Gynecological Pain by Study Characteristics

2. The effect of ear acupressure therapy on OBGY pain

The pooled overall effect size of the 16 studies on ear acupressure therapy for OBGY pain was Hedge’s g = −1.41 (95% CI: −1.96 to −0.86). According to Brydges' [21] criteria for interpreting effect sizes, this result indicates a large effect (Figure 1). In the heterogeneity test, the Q-statistics was 343.03 (p < .001), and Higgins' I² was 95.6%, indicating high heterogeneity among the studies. To explore the sources of heterogeneity, subgroup analysis, and meta-regression analysis were conducted.

Figure 1.

The effects of ear acupressure intervention on obsteric and gynecological pain. N = Number; ES = Effect size; CI = Confidence interval.

The subgroup analysis revealed that studies published after 2019 (Hedge’s g = −1.70, 95% CI: −2.34 to −1.06) and those conducted in Asia (Hedge’s g = −1.38, 95% CI: −1.99 to −0.77) demonstrated statistically significant pain reduction effects. Studies with fewer than 100 participants (Hedge’s g = −1.90, 95% CI: −2.68 to −1.13) also showed significant results. The type of pain was another influential factor, with normal delivery (Hedge’s g = −1.92, 95% CI: −2.93 to −0.92) and cesarean section (Hedge’s g = −0.96, 95% CI: −1.51 to −0.40) yielding different effect sizes. Intervention methods also played a role; acupressure as the intervention type resulted in Hedge’s g = −1.28 (95% CI: −1.81 to −0.75). When needles were used as the ear acupressure tool, the effect size was Hedge’s g = −0.44 (95% CI: −0.81 to −0.08). Studies using no intervention or routine obstetric care as the control group intervention achieved Hedge’s g = −1.86 (95% CI: −2.52 to −1.20). Additionally, studies with session durations of less than 75 seconds or not reported showed Hedge’s g = −1.56 (95% CI: −2.18 to −0.94) (Table 2).

A univariate meta-regression was performed to assess the potential impact of study characteristics on effect size. The results indicated that studies published after 2019 (Z = −10.29, p < .001) and those conducted in Asia (Z = 3.11, p = .002) had statistically significant effects. The absence of IRB approval (Z = 10.52, p < .001) and funding (Z = 3.73, p < .001) were also associated with significant findings. Studies with fewer than 100 participants (Z = 8.54, p < .001) and those targeting pain types such as normal delivery (Z = −6.81, p < .001), cesarean section (Z = −2.83, p = .005), Episiotomy (Z = 11.29, p < .001) and other pain conditions(Z = −3.91, p < .001) exhibited significant influences. Acupressure as the intervention type (Z = −3.90, p < .001) and the use of tools other than needles or seeds, such as beads or magnetic plates (Z = −2.46, p = .014), were also significant factors. Fewer than five stimulus points (Z = −10.07, p < .001) and control groups receiving no intervention or routine obstetric care (Z = 10.68, p < .001) further contributed to the results. Intervention durations exceeding two days (Z = 2.23, p = .026) and fewer than two sessions or unreported session numbers (Z = −8.29, p < .001) were also influential. Longer session times exceeding 75 seconds (Z = 4.64, p < .001) and outcome measurements not conducted immediately after the intervention (Z = −11.03, p < .001) impacted the findings. Finally, studies with quality assessment scores of 10 or higher (Z = 4.27, p < .001) demonstrated statistically significant effects (Table 3).

Table 3.

Meta-Regression Analysis Evaluating Ear Acupressure on Obstetric and Gynecological Pain

The exclusion sensitivity test [22] evaluated the combined effect size and statistical significance after reducing weights. The Hedge’s g ranged from −1.54 to −1.15, indicating a large effect size, with a 95% CI between −2.13 to −1.66 and −0.99 to −0.64, demonstrating statistically significant sensitivity to exclusion (Appendix 4).

3. Publication bias

To assess publication bias, funnel plot, and trim-and-fill plot analyses were performed. The individual effect sizes of the 16 studies included in this research (blue circles) appeared skewed to the left, indicating the presence of publication bias (Figure 2-A). The trim-and-fill plot suggested that six additional studies (white circles) should be added (Figure 2-B).

Figure 2.

(A) Funnel plot of ear acupressure intervention on obsteric and gynecological pain. (B) Trim and fill plot of ear acupressure intervention on obsteric and gynecological pain. Precision = 1/standard error; 0.05; limit line = 95% confidecne limit.

Further analysis using the trim-and-fill method [23] confirmed that six studies were missing, and the adjusted pooled effect size of the 22 studies was −0.25 (95% CI: −0.35 to −0.15). Even after adjusting for publication bias, the effect size for pain reduction remained statistically significant.

Additionally, both Begg's test for rank correlation and Egger's regression test showed significant results for pain reduction, confirming that while publication bias was present, it did not alter the statistical significance of the results. Therefore, the findings were deemed acceptable despite the presence of publication bias (Appendix 5).