Efficacy and Safety of Electroacupuncture for Pain Alleviation in Post-Total Knee Arthroplasty Patients: A Systematic Review and Meta-Analysis

Introduction

Over the past few decades, the rising prevalence of knee osteoarthritis (OA) and rheumatoid arthritis (RA) has made total knee arthroplasty (TKA) one of the most frequently performed orthopedic procedures worldwide [1,2]. While TKA effectively relieves arthritic pain and improves function, postoperative pain remains a major clinical challenge, with approximately 60% of patients experiencing severe pain after surgery [3,4]. Inadequate pain control impedes early mobilization, delays functional recovery, and increases the risk of complications such as muscle atrophy and joint stiffness, thereby prolonging hospital stays and raising healthcare costs[5, 6].

Current pain management protocols rely heavily on pharmacological interventions, particularly opioids. Despite the adoption of multimodal analgesia, opioid-related adverse effects – such as respiratory depression, nausea, and constipation – remain common, especially among elderly patients who constitute the majority of TKA recipients [7,8]. In light of the ongoing opioid crisis, there is a pressing need for effective non-pharmacological adjuncts that can reduce opioid consumption while enhancing pain control.

Electroacupuncture (EA), which combines traditional acupuncture with electrical stimulation, has gained attention as a standardized, non-pharmacological modality for postoperative pain. Recent randomized controlled trials (RCTs) have begun to examine its role in TKA [9]. For example, a 2024 trial by Huang et al reported that EA significantly reduced postoperative morphine consumption and pain scores within the first 48 hours compared to the control group[10]. Similarly, a study by Liu et al demonstrated that EA combined with conventional analgesia improved knee flexion and reduced hospital stay [11]. However, not all studies have reported consistent benefits, and some have raised questions regarding the durability of EA’s analgesic effects and its safety in older surgical populations. Although several systematic reviews have addressed acupuncture in perioperative care, none have focused specifically on EA in the context of TKA. Existing evidence remains fragmented, with variability in EA protocols, outcome measures, and control groups limiting the formation of clear clinical recommendations. Furthermore, the safety profile of EA in TKA patients has not been systematically evaluated.

To address these gaps, this study aims to conduct a systematic review and meta-analysis to evaluate the efficacy and safety of EA for pain management following TKA. By synthesizing available RCT evidence, we seek to provide robust conclusions that may guide clinical practice and inform future research directions.

Material and Methods

SEARCH STRATEGY AND DATA SOURCES:

A comprehensive literature retrieval was performed to identify RCTs evaluating the efficacy and safety of EA in the management of postoperative pain following TKA. A total of 8 electronic databases were systematically searched for all relevant RCTs investigating the application of EA in TKA patients, with the search period spanning from the establishment of each database to July 20, 2025. The databases included PubMed, Web of Science, Embase, the Cochrane Library, China National Knowledge Infrastructure (CNKI), China Biomedical Literature Database (CBM), Chinese Scientific Journal Database (VIP), and Wanfang Data. The search strategy incorporated Medical Subject Headings (MeSH) terms and keywords such as “total knee arthroplasty”, “electroacupuncture”, and “randomized controlled trial”, combined using Boolean operators to optimize retrieval sensitivity.

INCLUSION CRITERIA AND STUDY SELECTION:

Study eligibility was determined according to the PICOS framework, with each component selected to address the specific research questions while balancing methodological rigor with clinical applicability: Participants (P): Adult patients who underwent TKA. Interventions (I): EA was administered as the primary intervention, either alone or in combination with other therapies, in addition to routine care; no restrictions were imposed on needling techniques or acupoint selection. Comparisons (C): Comparators included sham interventions, no treatment, routine care, or other therapeutic modalities, but did not include electroacupuncture. Outcomes (O): Studies were required to report at least one of the following outcomes: pain intensity (assessed using the Visual Analog Scale [VAS]), levels of prostaglandin E2 (PGE-2) and β-endorphins, frequency of rescue analgesic use, or adverse events. Study Design (S): Only RCTs were included. Studies were excluded based on the following criteria: (1) Non-human studies, case reports, self-controlled studies, and non-randomized controlled trials; (2) Duplicate publications; (3) Studies where EA was administered in both the experimental and control groups, or those comparing therapeutic effects between different EA parameters (eg, acupoints or frequencies); (4) Studies without available full texts or from which data could not be extracted; (5) Studies reporting only outcomes unrelated to TKA or assessing outcomes prior to TKA.

LITERATURE SCREENING AND DATA EXTRACTION:

Two independent reviewers conducted the literature screening process. Search results were imported into EndNote 20 (Clarivate Analytics, Philadelphia, PA, USA) for management. Following the removal of duplicate records, the reviewers initially excluded irrelevant studies through title and abstract screening. Subsequently, the full texts of the remaining studies were retrieved and evaluated against the predefined inclusion and exclusion criteria to determine final eligibility. Relevant data were extracted, including authors, year of publication, sample size, participant characteristics, details of interventions (eg, EA parameters and comparator protocols), and outcome measures. All screening and data extraction steps were performed independently by the 2 reviewers, with their results cross-checked for consistency. Discrepancies arising during the process were resolved through consensus discussion, with adjudication by a third reviewer or the corresponding author when necessary.

OUTCOMES AND DEFINITIONS:

The Visual Analog Scale (VAS) operates on the fundamental principle of quantifying pain intensity via a ruler approximately 10 centimeters in length. One side of the scale is demarcated into 10 equal segments, with the “0” endpoint denoting the absence of pain and the “10” endpoint representing the most excruciating pain imaginable. In clinical application, patients are instructed to inspect the ruler and mark the position that aligns with their subjective pain experience. Then, researchers derive numerical scores based on the marked positions to operationalize the pain level [14]. Beyond the VAS, the assessment of pain incorporates physiological biomarkers that reflect underlying pathological mechanisms. Prostaglandin E2 (PGE-2), a key inflammatory mediator, is well-established in pain literature for its role in peripheral and central sensitization. Elevated PGE-2 levels directly contribute to hyperalgesia by lowering pain thresholds through EP receptor activation and potentiation of ion channel function. Conversely, β-endorphin (β-END) functions as an endogenous opioid peptide that produces analgesia through binding to μ-opioid receptors in both central and peripheral nervous systems, inhibiting pain transmission and modulation. The inclusion of these biomarkers in the present analysis is justified by their distinct mechanistic pathways: while VAS captures the subjective pain experience, PGE-2 and β-END provide objective, complementary measures of inflammatory and endogenous analgesic activity, respectively. This multi-dimensional approach enhances the robustness of empirical findings and facilitates a more precise elucidation of EA’s analgesic mechanisms, particularly its potential dual effect on reducing pro-nociceptive mediators while enhancing endogenous pain inhibition.

ASSESSMENT OF RISK OF BIAS:

Two independent reviewers appraised the risk of bias in the included studies utilizing the Risk of Bias tool outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Version 5.0.2). This instrument evaluates 7 critical domains: random sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data, selective reporting, and other potential sources of bias. The methodological quality of the included studies was categorized into 3 tiers: low risk, unclear risk, and high risk of bias. Any discrepancies arising during the evaluation process were resolved through deliberation with a third reviewer or by consulting the corresponding author of the respective study.

CERTAINTY OF EVIDENCE:

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) framework was employed to appraise the overall quality of evidence in the present study. As specified in the GRADE guidelines, 5 core domains are subject to evaluation: risk of bias, inconsistency, indirectness, imprecision, and potential publication bias. Collectively, these domains inform the assessment of evidence quality, which is categorized into 4 hierarchical levels: high, moderate, low, and very low. To safeguard objectivity and enhance reliability, the evaluation was conducted independently by 2 researchers, with a third researcher overseeing the assessment process. Any discrepancies or disagreements arising during the evaluation were resolved through rigorous deliberation and consultation with relevant subject-matter experts.

STATISTICAL ANALYSIS:

Meta-analysis was performed utilizing Review Manager version 5.4.0 software. For continuous variables, data were quantified using the mean difference (MD) with a 95% confidence interval (CI). The VAS scores for pain were inputted manually, with post-treatment measurements from both the intervention and control groups directly selected as statistical parameters, given that all included studies in the meta-analysis only reported VAS scores post-treatment. To compare statistical disparities between the intervention and control groups, the mean differences in PGE-2 and β-endorphin levels between pre-treatment and post-treatment were analyzed. For dichotomous variables, outcomes were evaluated using the risk ratio (RR) with a 95% CI. Heterogeneity was assessed via the Higgins test, with the I²statistic calculated; values exceeding 50% indicated substantial heterogeneity. Potential sources of heterogeneity were explored through sensitivity and subgroup analyses. Subgroup analyses focused on the following factors: 1) different combinations of EA interventions; 2) acupoint selection strategies; 3) sample size; and 4) duration of EA treatment. Sensitivity analyses were conducted by sequentially excluding each RCT to evaluate the robustness of the results. If more than 10 eligible studies were identified, funnel plots were employed to assess publication bias, supplemented by Egger’s and Begg’s tests for verification.

Results

LITERATURE SCREENING:

A total of 276 articles were initially identified through searches of 8 electronic databases. After removing 65 duplicate records, 211 studies underwent further screening. Among these, 189 were excluded based on predefined eligibility criteria, and an additional 6 studies were omitted due to inconsistencies in 11outcome measures or insufficient reporting of results. Ultimately, 16 studies were deemed eligible and included in the final meta-analysis [15–30]. The study selection procedure is summarized in Figure 1. To enhance data extraction and presentation, a detailed narrative synthesis accompanied by tabulated data has been incorporated to clarify study characteristics and outcomes. Key elements such as sample size, intervention details, and patient information are systematically outlined in Table 1, providing a structured overview that supports the interpretation and reproducibility of the meta-analysis.

LITERATURE CHARACTERISTICS:

All 16 trials included in the analysis were single-center RCTs conducted in China, with publications in both English and Chinese spanning the period 2013–2025. These 16 trials enrolled a total of 1142 patients who had undergone TKA, with sample sizes ranging from 40 to 120 participants per study (Tables 1, 2) [15–30].

Among the 16 studies, 3 selected acupoints on the contralateral (healthy) side[18, 23, 24], while 13 targeted acupoints on the ipsilateral (affected) side [15–17,19–22,25–30]. The acupoints involved included Liangqiu (ST34), Zusanli (ST36), Yanglingquan (GB34), Xuehai (SP10), Fenglong (ST40), Qiuxu (GB40), Waiqiu (GB36), Jinmen (BL63), Neimadian (EX-LE29), Diji (SP8), Yinlingquan (SP9), Shousanli (LI10), Quchi (LI11), Zhouliao (LI12), Binao (LI14), Chize (LU5), Sidu (TE9), Ashi points, Xuanzhong (GB39), Sanyinjiao (SP6), Weizhong (BL40), Futu (ST32), Kunlun (BL60), and Tiaokou (ST38). A total of 9 studies incorporated physical function exercises as part of routine care in both the intervention and control groups [19–21,24–26,28–30]. Additionally, all 16 studies administered conventional multimodal analgesia as standard care to assist in pain management. Based on an analysis of the 16 studies, the conventional multimodal analgesia regimen constitutes a well-structured perioperative pain management system. This protocol primarily employs oral celecoxib (at a standard dose of 200 mg, once or twice daily) as foundational anti-inflammatory and background analgesia [16,17,22–25,28,30,31]. Building upon this, most studies combined regional analgesic techniques, specifically including saphenous or femoral nerve blocks using ropivacaine [20,30], patient-controlled epidural analgesia (PCEA) based on ropivacaine combined with opioids [16,27,28,31], or intraoperative local infiltration analgesia with ropivacaine [23,24]. To address breakthrough pain, the regimens systematically incorporated opioids: some studies utilized patient-controlled intravenous analgesia (PCIA, commonly with sufentanil) to provide baseline opioid analgesia [18,21,23–25,30]; meanwhile, most established an on-demand rescue analgesia protocol, administering oral oxycodone/acetaminophen or injectable bucinnazine hydrochloride based on pain scores [23,26]. The specific multimodal analgesia management methods used in the other 2 studies are not available [19,29]. Together, these components form a stepwise, multimodal analgesia strategy that progresses from foundational medication and regional blocks to systemic opioid rescue. Additionally, all 16 studies administered conventional multimodal analgesia as standard care to assist in pain management; this post-surgical multimodal analgesia regimen typically included indwelling epidural analgesia pumps and oral analgesic medications. Regarding EA parameters: 14 studies utilized dilatational waveforms [15–21,23,25–30], while 2 employed continuous waveforms [22,24]. Nine studies selected an EA frequency of 2 Hz/100 Hz [15–18,23,27–30], 1 study used a frequency of 2 Hz[22], and 6 studies did not specify the frequency applied [19–21,24–26]. For needle retention time, 11 studies maintained EA for 30 minutes [15–17,19,20,24,25,27–30], 4 studies for 20 minutes [18,21–23], and 1 study for 25 minutes [26]. Outcome measures reported across the studies included: the VAS for pain, reported in 15 studies [15–21,23–30]; levels of PGE-2 and β-END, evaluated in 3 studies [17,18,30]; adverse events, documented in 9 studies [15,18,22–24,26–28,30]; and the number of additional analgesic administrations, noted in 4 studies [22,27,28,30].

RISK OF BIAS ASSESSMENT:

The methodological quality of the included studies was rated as low to moderate using the Cochrane Risk of Bias (ROB) Tool Version 5.3.0. Figure 2 presents a summary of the risk of bias assessment results. Among the 16 studies, 14 employed a randomized grouping approach via a random number table method, which was deemed to carry a low risk of bias [15–24,26,28–30]. However, one study utilized hospitalization numbers for randomization, leading to a high risk of bias assessment [25]; another study explicitly stated that no random allocation was performed, thus also being classified as having a high risk of bias [27].Due to the inherent characteristics of EA, blinding of EA therapists was not practicable, resulting in a high risk of performance bias across all 16 studies. None of the studies reported missing outcome data or selective reporting, which were assessed as low risks of bias. Additionally, none of the studies described blinding of outcome assessors, and no other sources of bias were explicitly addressed in any of the trials.

CERTAINTY OF EVIDENCE:

The overall quality of evidence was evaluated using the GRADE approach. As specified in the GRADE guidelines, 5 key domains are subject to assessment: risk of bias, inconsistency, indirectness, imprecision, and potential publication bias. Collectively, these domains determine the classification of evidence quality into 4 hierarchical levels: high, moderate, low, and very low. To ensure objectivity and enhance reliability, the evaluation was conducted independently by 2 researchers, with a third researcher overseeing the assessment process. Any discrepancies or disagreements arising during the evaluation were resolved through rigorous deliberation and consultation with relevant subject-matter experts.

PAIN AT REST (VAS): Meta-analysis consistently supported the primary hypothesis that EA effectively alleviates resting pain after total knee arthroplasty. A significant reduction in VAS scores was observed in the EA group across all postoperative time points. On postoperative day 1 (9 studies, 628 patients), the EA group showed significantly lower VAS scores (MD=−0.49, 95% CI [−0.62, −0.35], P<0.00001; I2=29%) (Figure 3A) [15–17,19,20,25,27–29]. Sensitivity analysis revealed that the results were relatively stable, confirming the reliability of the conclusion. Due to the small number of included studies, funnel plots were not constructed to assess publication bias. On postoperative day 3 (12 studies, 757 patients), a greater analgesic effect was observed (MD=−0.89, 95% CI [−1.11, −0.68], P<0.00001), albeit with substantial heterogeneity (I2=73%) (Figure 3B) [16–21,24,25,27–30]. Sensitivity analysis confirmed the relative stability of the results, validating the reliability of the conclusion. The statistical tests indicated no significant publication bias (Begg’s test, P=1.00; Egger’s test, P=1.00). On postoperative day 7 (11 studies, 587 patients), the benefit remained significant (MD=−0.51, 95% CI [−0.65, −0.38], P<0.00001; I2=55%) (Figure 3C) [16,19,21,23–30]. Sensitivity analysis indicated that the heterogeneity of pain at rest on postoperative day 7 decreased to 0% after excluding the study by Zhu et al [25], suggesting that this study may be a potential source of heterogeneity. The funnel plot for VAS scores at rest on postoperative day 3 and 7 is detailed in Figure 4A and 4B, with results indicating the presence of potential publication bias (Begg’s test, P=0.0127; Egger’s test, P=0.0000).

PAIN AT MOTION: The efficacy of EA was further demonstrated during physical activity, a key outcome for functional recovery. On postoperative day 1 (6 studies, 330 patients), the EA group reported lower pain scores (MD=−0.54, 95% CI [−0.76, −0.33], P<0.00001) (Figure 5A) [15,16,24,27–29]. Sensitivity analysis confirmed the relative stability of the results, validating the reliability of the conclusion. Due to the small number of included studies, funnel plots were not constructed to assess publication bias. On postoperative day 3 (6 studies, 290 patients), the EA group showed significantly lower VAS scores (D=−0.56, 95% CI [−0.69, −0.42], P<0.00001; I2=0%) [16,24,27–30] (Figure 5B). Sensitivity analysis revealed that the results were relatively stable, confirming the reliability of the conclusion. Due to the small number of included studies, funnel plots were not constructed to assess publication bias. On postoperative day 7 (6 studies, 290 patients), the EA group reported lower pain scores (MD=−0.57, 95% CI [−0.69, −0.44], P<0.00001; I2=0%) [16,24,27–30] (Figure 5C). Sensitivity analysis indicated that the results were relatively stable, confirming the reliability of the conclusion. Due to the small number of included studies, funnel plots were not constructed to assess publication bias.

THE LEVELS OF PGE-2 AND β-END: To explore potential mechanisms, biochemical marker levels were assessed. Analysis of 3 studies (170 patients) revealed that EA intervention led to a significant decrease in the pro-inflammatory mediator PGE-2 (MD=−33.15, 95% CI [−37.29, −29.01], P<0.00001; I2=0%) (Figure 6A) and a concurrent increase in the endogenous opioid β-END (MD = 54.75, 95% CI [47.08, 62.42], P<0.00001; I2=58%) (Figure 6B) [15,16,28]. These results provide biological plausibility for the observed analgesic effects of EA. Sensitivity analysis confirmed the relative stability of the results, validating the reliability of the conclusion. Due to the small number of included studies, funnel plots were not constructed to assess publication bias.

SUPPLEMENTARY ANALGESIA AND ADVERSE EVENTS: Analysis of clinical utility and safety yielded favorable outcomes. Regarding supplementary analgesic use, pooled data from 4 studies (n=250) demonstrated that patients receiving EA required significantly fewer additional analgesic doses compared to the control group (RR=0.46, 95% CI [0.25, 0.85], P=0.01), with no heterogeneity (I2=0%) [22,27,28,30] (Figure 7). This indicates that EA contributes to improved postoperative pain control, reducing the reliance on rescue medication. Due to the small number of included studies, funnel plots were not constructed to assess publication bias. In terms of safety, analysis of 9 studies (n=583) showed that EA was associated with a significantly lower risk of adverse effects (RR=0.45, 95% CI [0.27, 0.75], P=0.002; I2=21%) (Figure 8) [15,18,22–24,26–28,30]. The reported adverse events, such as minor bleeding, needle bending, dizziness, and nausea, were predominantly mild in nature. This finding, combined with the reduction in analgesic use, underscores the favorable safety profile of EA in the context of postoperative recovery. Due to the small number of included studies, funnel plots were not constructed to assess publication bias.

SUBGROUP ANALYSIS:

Subgroup analyses were performed for VAS scores at rest on postoperative days 3 and 7 to investigate whether the efficacy of EA was influenced by specific study or intervention characteristics. The pre-specified stratification variables were: (1) EA combination, examining whether EA was used as a standalone therapy or in conjunction with other analgesics, to assess its independent and adjunctive value; (2) Acupoint selection, categorized as local points, distant points, or contralateral (healthy) side points, to explore the importance of needling location; (3) Sample size, dichotomized as above or below the median, to evaluate the impact of study size on the effect estimate; and (4) Treatment duration, to determine if a longer intervention period was associated with superior outcomes. The results, detailed in Table 3, demonstrated the consistent analgesic effect of EA across most subgroups. For VAS at rest on postoperative day 3, EA showed statistically significant superiority over the control group in all subgroups defined by the 4 categorical indicators (all P<0.05). Similarly, on postoperative day 7, a significant benefit of EA was maintained in subgroups of different EA combinations, sample sizes, and treatment durations (all P<0.05). However, in the subgroup where acupuncture was applied exclusively to the “contralateral (healthy) side”, no statistically significant difference was found between the EA and control groups on day 7 (P=0.31), although a trend favoring EA was still observed. This suggests that needling location, particularly avoiding the affected limb, can influence the sustainability of the analgesic effect.

CERTAINTY OF EVIDENCE:

Table 4 summarizes the certainty of evidence for various outcomes. Two outcomes – VAS scores at rest on postoperative day 3 and β-END levels – were rated as having low certainty, while the evidence for VAS scores at rest on postoperative day 7 was assessed as very low. The remaining 7 outcomes were assigned a moderate level of certainty, including VAS scores at rest on postoperative day 1, VAS scores at motion on postoperative days 1, 3, and 7, PGE-2 levels, and the number of additional analgesics administered.

Discussion

LIMITATION:

This study is subject to several limitations. Firstly, the literature search, although not formally restricted by language, ultimately identified relevant publications only in English and Chinese, with a predominant inclusion of trials conducted in China (15 out of 16). This geographical and linguistic concentration may limit the generalizability of the findings and potentially introduce a positive bias, as regional practices and publication tendencies could influence outcomes. Secondly, significant clinical heterogeneity was observed, partly attributable to the co-interventions of physical function exercises and pain management in both experimental and control groups across most studies. Variations in the intensity and frequency of these exercises, coupled with differences in the selection and combination of acupoints, complicate the interpretation of the isolated effect of EA. Methodological concerns also exist, as several trials failed to provide adequate descriptions of allocation concealment, undermining the assessment of their methodological rigor. Finally, a critical constraint is the absence of long-term follow-up data in all included studies, which precludes any evaluation of the sustained effects of EA beyond the immediate intervention period. These limitations highlight several specific avenues for future investigation. To address the issues of generalizability and potential bias, subsequent research should prioritize large-scale, multi-regional RCTs that enroll diverse patient demographics from different healthcare systems. Future studies must explicitly report and standardize co-interventions (eg, physical exercise protocols) to better isolate the specific contribution of EA. Furthermore, there is a pressing need to explore the long-term efficacy and sustainability of EA, which requires trials incorporating extended follow-up periods of at least 6 to 12 months. Finally, to enhance the validity of future findings, researchers should adhere to rigorous methodological standards, clearly describing random sequence generation, allocation concealment, and blinding procedures, as per the CONSORT guidelines. Investigating the optimal EA parameters (eg, intensity, frequency, and acupoint selection) for specific patient subgroups also represents a crucial direction for optimizing this therapeutic approach.

Conclusions

Based on the synthesis of available evidence, the outcomes, though limited by methodological quality, strongly suggest that EA is an effective and safe adjunctive therapy for pain management after TKA. The results indicate that EA is associated with reduced postoperative pain scores, diminished requirement for supplemental analgesics, and modulation of relevant biochemical markers, all within a context of a favorable short-term safety profile characterized primarily by mild adverse events. The consistent beneficial effects across the first postoperative week support the practical integration of EA into standardized, multimodal analgesic pathways for TKA patients. This integration can be operationalized by initiating EA early in the postoperative period, utilizing a protocol based on common effective parameters (eg, dilatational wave, local acupoints), and coordinating its delivery within the interdisciplinary surgical team. Such an approach holds promise for enhancing pain management while potentially reducing opioid reliance. However, these conclusions must be interpreted in light of notable limitations. The overall certainty of evidence is low to moderate, constrained by the methodological limitations of the included trials – particularly concerns regarding blinding – and their exclusive origin within a single country, which may affect generalizability. Significant heterogeneity in EA application parameters, including stimulation characteristics and acupoint selection, was observed. The subgroup analysis indicating that contralateral needling might be less effective for sustained analgesia at 1 week after surgery further underscores that the optimal EA protocol remains to be established. Therefore, while EA demonstrates potential as a non-pharmacological component of multimodal analgesia that could help reduce opioid consumption, its definitive efficacy, long-term safety, and standardized application parameters require validation through rigorously designed, multinational RCTs.