Efficacy of Transcutaneous Auricular Vagus Nerve Stimulation for Postoperative Pain: A Systematic Review and Meta-Analysis Authors: Zheng-Chun Mao, Yue-Zhong Lv, Qi-Hong Shen, Gang Chen Published: 2026 ABSTRACT ObjectivesTranscutaneous auricular vagus nerve stimulation (taVNS) has been proposed as a non-pharmacological intervention for postoperative pain, but its effects remain unclear. We conducted this study to evaluate the efficacy and safety of taVNS for postoperative pain management.MethodsA systematic literature search was performed in PubMed, Embase, the Cochrane Library, and Web of Science. Randomized controlled trials comparing perioperative taVNS with sham stimulation implemented using inactive electrical output, electrical insulation, transient stimulation followed by deactivation, or electrode placement at presumed non-vagal auricular sites in adult surgical patients were included. The review protocol was prospectively registered in PROSPERO (CRD420251129040). Outcomes included postoperative pain scores, postoperative opioid consumption, proportion of patients requiring rescue analgesia, incidence of rebound pain, and incidence of postoperative nausea and vomiting (PONV). Standardized mean differences (SMD) or mean differences (MD) with 95% confidence intervals (CI) were calculated for continuous outcomes, and risk ratios (RR) for dichotomous outcomes, using random-effects models.ResultsSeven studies (684 patients) were included in the meta-analysis. The results showed that taVNS significantly reduced postoperative pain scores at early (MD −0.79, 95% CI −1.02 to −0.56), middle (MD −0.77, 95% CI −1.10 to −0.43), and late (MD −0.88, 95% CI −1.06 to −0.70) time points. Postoperative opioid consumption was not significantly different between the taVNS and sham groups (SMD −0.69, 95% CI −1.40 to 0.03). taVNS reduced the need for rescue analgesia and the incidence of rebound pain and was associated with a borderline lower incidence of PONV.ConclusionPerioperative taVNS may reduce postoperative pain intensity and the incidence of rescue analgesia, rebound pain, and PONV; however, it was not associated with a statistically significant reduction in postoperative opioid consumption. Given the moderate-to-very-low certainty of evidence and the exclusively Chinese origin of the included trials, larger multicentre and multinational RCTs using standardized stimulation and sham protocols are required before routine clinical adoption. Introduction Postoperative pain remains a pervasive and often inadequately managed complication after surgery. Despite advances in surgical techniques and perioperative care, up to 60–80% of surgical patients experience moderate to severe acute postoperative pain,1 which not only causes patient distress but also impairs early mobilization, prolongs hospital stay, increases healthcare costs, and may contribute to the development of chronic postsurgical pain.2 Multimodal analgesia, combining different classes of analgesics and regional anesthesia techniques, has become the cornerstone of modern postoperative pain management.3 However, even with multimodal regimens, a substantial proportion of patients still require opioids to achieve acceptable pain control. Opioids, while effective, carry a significant burden of adverse effects including nausea, vomiting, constipation, sedation, respiratory depression, and the risk of dependence and misuse.4 Non-pharmacological approaches, such as acupuncture,5 transcutaneous electrical nerve stimulation (TENS),6 and neuromodulation techniques, have been explored as adjuncts to reduce pain and opioid consumption. Among these, transcutaneous auricular vagus nerve stimulation (taVNS) has emerged as a promising, non-invasive neuromodulation technique. Preclinical studies have demonstrated that vagus nerve stimulation reduces inflammatory cytokine release,7 attenuates nociceptive transmission, and raises pain thresholds.8 In humans, taVNS has been shown to alter pain perception in experimental models and to provide analgesic effects in chronic pain conditions such as migraine,9 fibromyalgia,10 and low back pain.11 Several recent randomized controlled trials (RCTs) have evaluated the effect of taVNS on postoperative pain following various surgical procedures,12–14 but their findings remain inconsistent. A previous meta-analysis15 of auricular vagus nerve stimulation combined percutaneous and transcutaneous approaches and pooled pain outcomes measured at heterogeneous time points, leaving the specific perioperative effects of non-invasive taVNS uncertain. Rescue analgesia, rebound pain, and postoperative nausea and vomiting (PONV) were selected as secondary outcomes because they reflect clinically relevant aspects of postoperative recovery beyond pain intensity alone, while rebound pain has rarely been quantitatively evaluated in previous reviews. Therefore, we conducted a systematic review and meta-analysis using predefined postoperative time windows to evaluate the effects of taVNS on postoperative pain, analgesic requirements, rebound pain, PONV, and safety. Methods We followed the PRISMA 2020 guidelines for reporting systematic reviews and meta-analyses. The study protocol was prospectively registered with PROSPERO (CRD420251129040). Search Strategy We searched PubMed, Embase, Cochrane Library, and Web of Science up to 31 May 2026 using terms covering both the intervention (“transcutaneous auricular vagus nerve stimulation”, “taVNS”, “auricular vagus nerve stimulation”) and study design (“randomized controlled trial”, “RCT”). No language restrictions were placed on the search. Reference lists of relevant reviews and included studies were also screened manually. Detailed search strategies for all databases are provided in Supplementary Material 1. Eligibility Criteria Studies were included if they met the following criteria: (1) Population: adult patients (≥ 18 years) undergoing any type of surgery; (2) Intervention: taVNS applied perioperatively (preoperative, intraoperative, or postoperative), regardless of stimulation parameters (frequency, intensity, duration, site); (3) Comparison: sham stimulation or standard care. Eligible sham procedures included inactive electrical output, electrical insulation between the electrode and skin, transient stimulation followed by device deactivation, or electrode placement at presumed non-vagal auricular sites; (4) Outcomes: at least one of the following: postoperative pain scores (visual analogue scale or numeric rating scale), postoperative cumulative opioid consumption, proportion of patients requiring rescue analgesia, incidence of rebound pain, or incidence of postoperative nausea and vomiting (PONV); (5) Study design: RCTs. We excluded crossover trials, quasi-randomized trials, observational studies, case reports, reviews, animal studies, and studies that were not transcutaneous auricular vagus nerve stimulation. Outcome Measures The primary outcome was postoperative pain intensity measured on visual analogue scale or numeric rating scale. To capture the temporal profile of pain relief, we predefined three time windows: early (2–6 h postoperatively), middle (8–24 h), and late (48–72 h or the third postoperative day). When multiple measurements fell within the same window, the one closest to the window endpoint was selected (such as 6 h for early, 24 h for middle, 72 h or day 3 for late). For studies reporting pain at several time points, such as 2, 6, 24, 48, and 72 h, we selected the 6-h, 24-h, and 72-h measurements for the early, middle, and late postoperative windows, respectively. If a study reported both resting and movement-evoked (or deep breathing) pain, we prioritised movement-evoked pain as it reflects more clinically relevant functional impairment. Secondary outcomes included: Cumulative postoperative opioid consumption, expressed as intravenous morphine milligram equivalents (MME) or, when conversion was not possible, as standardized mean difference (SMD) to allow pooling of different opioids (such as sufentanil, oxycodone) or surrogates (such as patient-controlled analgesia volume); proportion of patients requiring rescue analgesia (any additional analgesic administered due to inadequate pain control), when a study reported multiple time points, the time point with the largest number of events was selected to capture the overall rescue need; incidence of rebound pain: Studies were included only if they explicitly reported the phenomenon of “rebound pain”, regardless of the specific definition used (such as a sudden increase in pain intensity after resolution of a peripheral nerve block, NRS ≥ 7, or any other operational definition provided by the authors); incidence of PONV: we included any postoperative occurrence of nausea, vomiting, or both. If a study reported postoperative nausea or postoperative vomiting separately without a combined PONV outcome, the nausea event was used as the primary measure. Data Extraction Two authors independently assessed the titles and abstracts of all retrieved records, and subsequently evaluated the full texts of studies considered potentially eligible. Any disagreements between the two reviewers were settled through discussion, with involvement of a third reviewer when consensus could not be reached. We extracted relevant information using a pre-designed data collection form that captured the following items: first author, publication year, country, surgical procedure, sample size (taVNS vs control), patient demographics, anaesthesia type, taVNS stimulation parameters (site, frequency, pulse width, intensity, duration, timing), control details, and outcome data. For continuous outcomes, we extracted mean, standard deviation (SD), and sample size per group. If SD was not reported but median and interquartile range (IQR) were given, we converted them to mean and SD according to previous studies.16,17 If numerical data were presented only in figures (such as line graphs, bar charts, box plots without numeric labels), we first attempted to contact the corresponding author by Email to request the original data. A maximum of two follow-up emails were sent within two weeks. If no response was received within two weeks, the study was excluded from that outcome analysis rather than extracting data from the figures using digitizing software. This approach was adopted to minimize potential bias introduced by manual graphical extraction. Risk of Bias Assessment We evaluated the methodological quality of each included study using the Cochrane ROB 2.0 tool. This instrument assesses trial quality across five domains: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. Each domain received a judgment of “low risk”, “some concerns”, or “high risk”. We defined the overall risk of bias as “low” when all domains were rated low risk; “some concerns” when at least one domain raised concerns and no domain was high risk; and “high” when any domain was high risk or when multiple domains raised substantial concerns that significantly reduced confidence in the findings. Two reviewers independently performed the quality assessment, and any disagreements were resolved by consensus. Statistical Analysis We conducted all meta-analyses using RevMan 5.4 (Nordic Cochrane Centre, Copenhagen, Denmark). Given the anticipated clinical and methodological diversity across the included studies, arising from variations in surgical type, stimulation protocol, and outcome definition, we chose a random-effects model for all pooled estimates. For pain scores, we calculated mean differences (MD) with 95% confidence intervals (CI). For opioid consumption, we used standardized mean differences (SMD) with 95% CI to accommodate different drug types and measurement units. For binary endpoints (rescue analgesia, rebound pain, and PONV), we computed risk ratios (RR) with 95% CI. All statistical tests were two-tailed, with statistical significance set at P < 0.05. Quality of Evidence We evaluated the overall certainty of the evidence for each outcome using the GRADE framework, which takes into account five domains: risk of bias, inconsistency, indirectness, imprecision, and publication bias. For each outcome, we assigned a certainty rating of high, moderate, low, or very low. Results Study Selection and Characteristics Database searching identified 234 records. After duplicates were removed and titles and abstracts had been screened, 15 full-text articles underwent detailed eligibility assessment. Eight of these were subsequently excluded for the following reasons: not a RCT (n=1),18 combined with other interventions (n=1),19 intervention not applied to the ear (n=1),20 not taVNS (n=1),21 and data not extractable (n=4).13,14,22,23 Finally, seven RCTs12,24–29 met the inclusion criteria and were included in the meta-analysis, enrolling a total of 684 patients (343 in the taVNS group and 341 in the control group). The study selection process is illustrated in the PRISMA flow diagram (Figure 1). The characteristics of the included studies are summarised in Table 1. The studies were published between 2022 and 2026 and all were conducted in China. Surgical procedures included shoulder arthroscopy,25 cesarean section,26 thoracoscopic lobectomy,28 perianal surgery,27 anterior cruciate ligament reconstruction,29 total knee arthroplasty,24 and glaucoma surgery.12 The taVNS intervention varied across studies in terms of stimulation site (left cymba conchae in most studies), frequency (ranging from 4/20 Hz alternating to 30 Hz), intensity (titrated to tingling, fixed at 1.5 mA, or 20–60 mA), and duration (from a single 30-minute session to daily sessions for up to 7 days). All included studies used a sham-stimulation control; however, the sham procedures varied across trials. These included an identical device without electrical output, electrical insulation of the electrodes, transient initial stimulation followed by device deactivation, and electrode placement at presumed non-vagal auricular sites. Study-specific sham procedures are presented in Table 1. Table 1Characteristics of Included StudiesStudy IDSurgery TypeSample Size (taVNS/Sham)AnesthesiaIntervention Details (Site, Freq, Intensity, Duration)Control DetailsPostoperative AnalgesiaXia JY 202525Shoulder Arthroscopy35/35General + brachial plexus blockLeft cymba conchae; 30 Hz, pulse width 250 μs, 30s on/off; intensity titrated to tingling; from preoperative holding until extubation + 2h on POD1 and POD2Sham (same device, no output)PCIA (Sufentanil)Xiong XY 202526Cesarean Section78/78Combined Spinal-EpiduralLeft cymba conchae; 20 Hz, pulse width 200 μs; intensity titrated to tingling; 30 min once daily on surgery day, POD1, POD2Sham (same device, no output)PCIA (sufentanil + nalbuphine)Zhang J 202528Thoracoscopic Lobectomy34 / 33GeneralLeft cymba conchae; 30 Hz, pulse width 250 μs, 30s on/off; intensity titrated to tingling; from preoperative holding until extubation + 2h on POD1 and POD2Sham (same device, no output)PCIA (Sufentanil)Yin ZY 202527Perianal Surgery45 / 45General + pudendal nerve blockBilateral tragus and cymba conchae; 30 Hz, pulse width 200 μs; intensity 20–60 mA (sub-threshold); 30 min before surgery + 30 min at 24h postoperativelySham (same device, current turned off after tingling)Oral acetaminophen + tramadol (rescue)Zhou Q 202229Anterior Cruciate Ligament Reconstruction39 / 39General + femoral nerve blockLeft cymba conchae; 30 Hz, pulse width 300 μs; intensity to maximum tolerable; 1h on / 1h off for 12h after surgerySham (electrodes on earlobe + helix tail)Femoral nerve block + PCA + rescue oral analgesicsHuang H 202524Total Knee Arthroplasty65 / 64GeneralLeft cymba conchae; 25 Hz, pulse width 300 μs; intensity 1.5 mA; 30 min three times daily from 1 day before to 7 days after surgerySham (insulated film over electrodes)PCIA (hydromorphone)Wang YZ 202612Glaucoma47 / 47GeneralLeft auricular concha; sparse-dense wave (4/20 Hz); intensity not reported (titrated to tingling threshold); 30 min twice daily from 2d preoperative to 1d postoperativeSham (electrodes placed, no output)Standard analgesia (flurbiprofen axetil + acetaminophen; no opioids)Abbreviations: taVNS, transcutaneous auricular vagus nerve stimulation; PCIA, patient-controlled intravenous analgesia; POD, postoperative day; NSAID, non-steroidal anti-inflammatory drug. Table 1Characteristics of Included StudiesStudy IDSurgery TypeSample Size (taVNS/Sham)AnesthesiaIntervention Details (Site, Freq, Intensity, Duration)Control DetailsPostoperative AnalgesiaXia JY 202525Shoulder Arthroscopy35/35General + brachial plexus blockLeft cymba conchae; 30 Hz, pulse width 250 μs, 30s on/off; intensity titrated to tingling; from preoperative holding until extubation + 2h on POD1 and POD2Sham (same device, no output)PCIA (Sufentanil)Xiong XY 202526Cesarean Section78/78Combined Spinal-EpiduralLeft cymba conchae; 20 Hz, pulse width 200 μs; intensity titrated to tingling; 30 min once daily on surgery day, POD1, POD2Sham (same device, no output)PCIA (sufentanil + nalbuphine)Zhang J 202528Thoracoscopic Lobectomy34 / 33GeneralLeft cymba conchae; 30 Hz, pulse width 250 μs, 30s on/off; intensity titrated to tingling; from preoperative holding until extubation + 2h on POD1 and POD2Sham (same device, no output)PCIA (Sufentanil)Yin ZY 202527Perianal Surgery45 / 45General + pudendal nerve blockBilateral tragus and cymba conchae; 30 Hz, pulse width 200 μs; intensity 20–60 mA (sub-threshold); 30 min before surgery + 30 min at 24h postoperativelySham (same device, current turned off after tingling)Oral acetaminophen + tramadol (rescue)Zhou Q 202229Anterior Cruciate Ligament Reconstruction39 / 39General + femoral nerve blockLeft cymba conchae; 30 Hz, pulse width 300 μs; intensity to maximum tolerable; 1h on / 1h off for 12h after surgerySham (electrodes on earlobe + helix tail)Femoral nerve block + PCA + rescue oral analgesicsHuang H 202524Total Knee Arthroplasty65 / 64GeneralLeft cymba conchae; 25 Hz, pulse width 300 μs; intensity 1.5 mA; 30 min three times daily from 1 day before to 7 days after surgerySham (insulated film over electrodes)PCIA (hydromorphone)Wang YZ 202612Glaucoma47 / 47GeneralLeft auricular concha; sparse-dense wave (4/20 Hz); intensity not reported (titrated to tingling threshold); 30 min twice daily from 2d preoperative to 1d postoperativeSham (electrodes placed, no output)Standard analgesia (flurbiprofen axetil + acetaminophen; no opioids)Abbreviations: taVNS, transcutaneous auricular vagus nerve stimulation; PCIA, patient-controlled intravenous analgesia; POD, postoperative day; NSAID, non-steroidal anti-inflammatory drug. Characteristics of Included Studies Abbreviations: taVNS, transcutaneous auricular vagus nerve stimulation; PCIA, patient-controlled intravenous analgesia; POD, postoperative day; NSAID, non-steroidal anti-inflammatory drug. Risk of Bias Assessment The results of the ROB 2.0 assessment are presented in Figure 2. Three studies25,26,28 were judged as having an overall low risk of bias. The remaining four studies were rated as having some concerns. The main reasons were: potential issues with blinding of participants due to possible sensation differences between active and sham stimulation;12,24,27 deviations from intended interventions due to non-standardized rescue analgesia protocols;12,27 and uncertainty about the measurement of the subjective pain outcome.29 No study was rated as having a high risk of bias. Effects of taVNS on Postoperative Pain Scores Postoperative pain intensity (0–10 VAS/NRS) was reported in four studies for the early, middle, and late time points. Figure 3 shows the forest plots for each time point. Early Postoperative Pain Pooled analysis of three studies showed that taVNS significantly reduced pain scores compared with sham (MD = −0.79, 95% CI −1.02 to −0.56; P < 0.01; I2 = 21%, low heterogeneity). Middle Postoperative Pain Three studies contributed to this analysis. taVNS significantly lowered pain scores (MD = −0.77, 95% CI −1.10 to −0.43; P < 0.01; I2 = 54%, moderate heterogeneity). Late Postoperative Pain Two studies reported pain within the prespecified late postoperative window of 48–72 h or on postoperative day 3. The pooled analysis showed lower pain scores in the taVNS group than in the sham group (MD = −0.88, 95% CI −1.06 to −0.70; P < 0.01; I2 = 0%, low heterogeneity). Effect on Postoperative Opioid Consumption Three studies reported opioid consumption within the first 24 h after surgery. Because different opioids and units were used (sufentanil in μg, oxycodone in mg, and PCIA solution volume in mL), we pooled the data using SMD with a random-effects model. Postoperative opioid consumption was not significantly different between the taVNS and sham groups (SMD −0.69, 95% CI −1.40 to 0.03, I2 = 87%, high heterogeneity) (Figure 4). Effect on Rescue Analgesia Requirement Three studies12,25,29 reported the number of patients requiring rescue analgesia within the postoperative period. We extracted the time point with the largest number of events for each study. Pooled analysis showed that taVNS significantly reduced the need for rescue analgesia (RR = 0.43, 95% CI 0.28 to 0.68; P < 0.01; I2 = 0%) (Figure 5). Effect on Rebound Pain Two studies28,29 explicitly reported the incidence of rebound pain after peripheral nerve block resolution, using clinically relevant definitions. The pooled RR was 0.37 (95% CI 0.21 to 0.64; P < 0.01; I2 = 0%), indicating that taVNS reduces the occurrence of rebound pain (Figure 6). Effect on PONV Five studies12,24,25,27,29 reported the incidence of PONV. The pooled RR showed lower occurrence in the taVNS group (RR = 0.63, 95% CI 0.40 to 0.99; P = 0.05; I2 = 6%) (Figure 7). Safety No serious adverse events related to taVNS were reported across the seven included studies. Mild and transient local skin discomfort was mentioned in a few studies, all of which resolved without intervention. Publication Bias Publication bias was not assessed by funnel plot because fewer than 10 studies were included. GRADE The overall certainty of evidence ranged from moderate to very low. The certainty was rated as moderate for early postoperative pain, rescue analgesia, and PONV; low for middle and late postoperative pain and rebound pain; and very low for postoperative opioid consumption. Early postoperative pain was downgraded for risk of bias. Middle postoperative pain was downgraded for risk of bias and inconsistency, whereas late postoperative pain was downgraded for risk of bias and imprecision. Opioid consumption was downgraded for risk of bias, substantial inconsistency, and imprecision. Rescue analgesia and PONV were downgraded for risk of bias, and rebound pain was downgraded for risk of bias and imprecision. Publication bias could not be formally assessed because fewer than 10 studies were available for each outcome; therefore, no additional downgrade was applied solely on this basis. Detailed GRADE assessments are presented in Table 2. Table 2GRADE Assessment of Certainty of EvidenceOutcomeEffect (95% CI)I2CertaintyDowngradePain (early)MD −0.79 (−1.02 to −0.56)21%⨁⨁⨁◯ ModerateRisk of biasPain (middle)MD −0.77 (−1.10 to −0.43)54%⨁⨁◯◯ LowRisk of bias; inconsistencyPain (late)MD −0.88 (−1.06 to −0.70)0%⨁⨁◯◯ LowRisk of bias, imprecisionOpioid consumptionSMD −0.69 (−1.40 to 0.03)87%⨁◯◯◯ Very LowRisk of bias; inconsistency; imprecisionRescue analgesiaRR 0.43 (0.28 to 0.68)0%⨁⨁⨁◯ ModerateRisk of biasRebound painRR 0.37 (0.21 to 0.64)0%⨁⨁◯◯ LowRisk of bias; imprecisionPONVRR 0.63 (0.40 to 0.99)6%⨁⨁⨁◯ ModerateRisk of biasAbbreviations: MD, mean difference; SMD, standardized mean difference; RR, risk ratio; CI, confidence interval; PONV, postoperative nausea and vomiting. Table 2GRADE Assessment of Certainty of EvidenceOutcomeEffect (95% CI)I2CertaintyDowngradePain (early)MD −0.79 (−1.02 to −0.56)21%⨁⨁⨁◯ ModerateRisk of biasPain (middle)MD −0.77 (−1.10 to −0.43)54%⨁⨁◯◯ LowRisk of bias; inconsistencyPain (late)MD −0.88 (−1.06 to −0.70)0%⨁⨁◯◯ LowRisk of bias, imprecisionOpioid consumptionSMD −0.69 (−1.40 to 0.03)87%⨁◯◯◯ Very LowRisk of bias; inconsistency; imprecisionRescue analgesiaRR 0.43 (0.28 to 0.68)0%⨁⨁⨁◯ ModerateRisk of biasRebound painRR 0.37 (0.21 to 0.64)0%⨁⨁◯◯ LowRisk of bias; imprecisionPONVRR 0.63 (0.40 to 0.99)6%⨁⨁⨁◯ ModerateRisk of biasAbbreviations: MD, mean difference; SMD, standardized mean difference; RR, risk ratio; CI, confidence interval; PONV, postoperative nausea and vomiting. GRADE Assessment of Certainty of Evidence Abbreviations: MD, mean difference; SMD, standardized mean difference; RR, risk ratio; CI, confidence interval; PONV, postoperative nausea and vomiting. Discussion This systematic review and meta-analysis found that perioperative taVNS was associated with lower postoperative pain scores, reduced rescue analgesia and rebound pain, and a borderline lower incidence of PONV, but not with significantly lower postoperative opioid consumption. No serious adverse events related to taVNS were reported. The certainty of evidence was moderate to very low, mainly due to risk of bias and heterogeneity. Our findings differ from a recent meta-analysis by Duff et al15 which included 12 studies and reported no significant reduction in acute postoperative pain with aVNS. Several factors may explain this discrepancy. First, Duff et al combined percutaneous (pVNS) and transcutaneous (taVNS) approaches, whereas our review focused exclusively on taVNS. Second, their analysis pooled pain outcomes from heterogeneous assessment time points, while we used predefined time windows. Third, we included more recent trials published after their search date. Importantly, the favourable safety profile of taVNS observed in our analysis was also reflected in the previous meta-analysis, where no serious adverse events were reported. Preclinical studies have shown that vagal activation suppresses inflammatory cytokine release and attenuates nociceptive transmission in the spinal cord and brainstem.30 In humans, experimental studies have demonstrated that taVNS raises pain thresholds to thermal and mechanical stimuli.31 Recent clinical trials have reported that taVNS reduces acute pain after orthopaedic, thoracic, and abdominal surgeries, which aligns with our pooled results.24,26,28 The reduction in pain scores of approximately 0.8, while statistically significant, is below the commonly cited minimal clinically important difference of 1.0–1.5 for acute postoperative pain.32 Nevertheless, the reduction in rescue analgesia requirement and the borderline reduction in PONV may indicate potential improvements in postoperative recovery, although these findings should be interpreted cautiously. The absence of a significant reduction in opioid consumption may be explained by the small number of studies, high heterogeneity, and variability in opioid types and reporting units across studies. Although the reduction in opioid consumption did not reach statistical significance, the direction of effect favoured taVNS. Future studies with larger samples and standardized outcome measures are needed to determine whether taVNS confers a clinically meaningful opioid-sparing effect. The reduction in rebound pain is a finding that has not been specifically quantified in prior meta-analyses of taVNS for acute postoperative pain; it supports the concept that taVNS may prevent the exaggerated pain response after nerve block resolution, possibly by reducing central sensitisation.29,33 Compared with other non-pharmacological interventions, taVNS directly targets auricular regions associated with vagal afferents; however, direct comparative evidence is lacking, and standardized stimulation protocols have not yet been established. The certainty of evidence ranged from moderate to very low. Moderate-certainty evidence supported early pain, rescue analgesia, and PONV. Low-certainty evidence supported middle pain, late pain, and rebound pain. Very low-certainty evidence supported opioid consumption, primarily due to a combination of risk of bias, substantial heterogeneity, and inconsistency in outcome measures. This is not unexpected given the diversity of surgical procedures (orthopaedic, thoracic, perianal, obstetric, and ophthalmic surgery), different taVNS protocols (frequency ranging, variable duration and timing), and different definitions and units for opioid consumption (such as PCIA volume). Subgroup analyses by surgery type or stimulation parameters were not performed due to the limited number of studies, but the direction of effect was consistent across most studies (favouring taVNS). taVNS was associated with a lower incidence of PONV, although the borderline statistical significance warrants cautious interpretation. This effect may be related to vagal modulation of autonomic and gastrointestinal function, although the underlying mechanism remains uncertain.34,35 However, this finding was based on only five trials and should be interpreted cautiously. Several limitations should be acknowledged. First, the number of included studies is small, and four studies had some concerns in risk of bias, mainly due to incomplete blinding (sensation differences between active and sham may have unblinded participants) and non-standardized rescue analgesia protocols. Second, substantial heterogeneity existed for some outcomes, which could not be fully explored due to limited study numbers. In addition, the sham procedures varied across studies and included inactive electrical output, electrical insulation, transient stimulation followed by deactivation, and stimulation at presumed non-vagal auricular sites. These differences may have affected the success of participant blinding and may also have introduced different degrees of physiological activity in the control groups. Third, the follow-up period was short, so the effects on long-term outcomes such as chronic postsurgical pain remain unknown. Fourth, publication bias could not be assessed by funnel plot because fewer than 10 studies were included. Fifth, the optimal stimulation parameters (frequency, intensity, duration, timing, ear laterality) have not been established, and the studies used variable protocols, making it difficult to recommend a standardized regimen. Sixth, all seven included trials were conducted in China. Differences in patient characteristics, perioperative care pathways, surgical and anaesthetic practices, and the implementation of taVNS may limit the generalizability of the findings to other countries and healthcare settings. Seventh, trial registries such as ClinicalTrials.gov and the Chinese Clinical Trial Registry were not searched; therefore, unpublished or ongoing studies may have been missed. Conclusions Perioperative taVNS may reduce postoperative pain, rescue analgesia, rebound pain, and PONV, but not opioid consumption. Given the limited certainty and generalizability of the evidence, larger multicentre and multinational RCTs using standardized taVNS and sham protocols, as well as standardized opioid-consumption measures, are needed. Figure 1PRISMA 2020 flow diagram of study selection.The flowchart illustrates the PRISMA study selection process. It begins with identification, where 234 records are identified from databases and none from registers. Before screening, 78 duplicate records are removed, with no records marked ineligible by automation tools or removed for other reasons. During screening, 156 records are screened and 141 are excluded after screening titles and abstracts. Reports sought for retrieval total 15, with none not retrieved. Reports assessed for eligibility also total 15, with 8 excluded for reasons such as data not extractable (4), intervention not applied to the ear (1), combined with another intervention (1), not a randomized controlled trial (1) and not taVNS (1). Finally, 7 studies are included in the review.A PRISMA flowchart of study selection process from identification to inclusion. Figure 2Risk of bias assessment of included studies using the Cochrane RoB 2 tool. D1, bias arising from the randomization process; D2, bias due to deviations from intended interventions; D3, bias due to missing outcome data; D4, bias in measurement of the outcome; D5, bias in selection of the reported result. Green “+” indicates low risk of bias; yellow “?” in individual domains and yellow “!” in the overall assessment indicate some concerns.Table showing risk of bias domains for various studies with low risk and some concerns. Figure 3Forest plot of postoperative pain scores. Comparison of taVNS versus sham for postoperative pain intensity at early (2–6 h), middle (8–24 h), and late (48–72 h) time points. Data were pooled using a random-effects model.Forest plot with three sections: early pain, middle pain and late pain. Left table columns are Study or Subgroup; taVNS with Mean, SD, Total; Control with Mean, SD, Total; Weight; and Mean Difference with 95 percent confidence interval. The x-axis is labeled Mean Difference, with tick labels at minus 1, minus 0.5, 0, 0.5 and 1. The y-axis lists studies and subtotals. Early pain studies: Wang YZ 2026 mean difference minus 1.06, 95 percent confidence interval minus 2.12 to 0.00; Yin ZY 2025 mean difference minus 0.68, 95 percent confidence interval minus 0.88 to minus 0.48; Zhang J 2025 mean difference minus 1.00, 95 percent confidence interval minus 1.37 to minus 0.63. Subtotal mean difference minus 0.79, 95 percent confidence interval minus 1.02 to minus 0.56. Middle pain studies: Wang YZ 2026 mean difference minus 0.71, 95 percent confidence interval minus 1.15 to minus 0.27; Yin ZY 2025 mean difference minus 0.97, 95 percent confidence interval minus 1.17 to minus 0.77; Zhang J 2025 mean difference minus 0.36, 95 percent confidence interval minus 0.95 to 0.23. Subtotal mean difference minus 0.77, 95 percent confidence interval minus 1.10 to minus 0.43. Late pain studies: Xiong XY 2025 mean difference minus 0.90, 95 percent confidence interval minus 1.28 to minus 0.52; Yin ZY 2025 mean difference minus 0.87, 95 percent confidence interval minus 1.08 to minus 0.66. Subtotal mean difference minus 0.88, 95 percent confidence interval minus 1.06 to minus 0.70. A vertical reference line is at 0 and each section includes a diamond for the subtotal.A forest plot comparing taVNS and control pain intensity across early, middle and late time points.Abbreviations: MD, mean difference; CI, confidence interval; taVNS, transcutaneous auricular vagus nerve stimulation; VAS, visual analogue scale; NRS, numeric rating scale. Figure 4Forest plot of postoperative opioid consumption within the first 24 hours. Comparison of taVNS versus sham for cumulative opioid consumption. Because different opioids and units were used across studies, data were pooled using standardized mean difference (SMD) with a random-effects model. Substantial heterogeneity was observed (I2 = 87%).Forest plot summary: Studies compared taVNS and control groups. Xia JY 2025: taVNS mean 36.94, SD 6.68, n=35; control mean 45.26, SD 6.83, n=35; weight 32%; difference -1.22, CI -1.73 to -0.71. Xiong XY 2025: taVNS mean 53.35, SD 3.78, n=78; control mean 58, SD 6.04, n=78; weight 35.4%; difference -0.92, CI -1.25 to -0.59. Zhang JY 2025: taVNS mean 33, SD 13, n=34; control mean 32, SD 9, n=33; weight 32.6%; difference 0.09, CI -0.39 to 0.57. Total: taVNS n=147, control n=146; weight 100%; difference -0.69, CI -1.40 to 0.03. Heterogeneity: tau superscript 2=0.34, chi superscript 2=15.93, df=2, P=0.0003, I superscript 2=87%. Overall effect: Z=1.89, P=0.06. Plot: x-axis ticks -2 to 2; captions favor experimental/control. Study points: -1.22, -0.92, 0.09; vertical line at 0. Diamond summary: -1.40 to 0.03, center -0.69.A forest plot comparing taVNS versus control for postoperative opioid consumption, with mixed effects.Abbreviations: SMD, standardized mean difference; CI, confidence interval; taVNS, transcutaneous auricular vagus nerve stimulation. Figure 5Forest plot of rescue analgesia requirement. Comparison of taVNS versus sham for the proportion of patients requiring rescue analgesia postoperatively. Data were pooled using risk ratio (RR) with a random-effects model.Forest plot comparing taVNS vs control for rescue analgesia. X-axis: Risk Ratio (0.01, 0.1, 1, 10, 100). Y-axis lists studies and total. Columns: Study/Subgroup; taVNS Events/Total; Control Events/Total; Weight; Risk Ratio M-H Random 95% CI. Studies: Wang YZ 2026 (taVNS 4/47, control 11/47, weight 17.4%, risk ratio 0.36, CI 0.12-1.06), Xia JY 2025 (taVNS 6/35, control 16/35, weight 30.1%, risk ratio 0.38, CI 0.17-0.85), Zhou Q 2022 (taVNS 10/39, control 20/39, weight 52.5%, risk ratio 0.50, CI 0.27-0.93). Total: taVNS 121, control 121, weight 100%, pooled risk ratio 0.43, CI 0.28-0.68. Total events: taVNS 20, control 47. Plot markers match risk ratios; diamond spans 0.28-0.68, centered at 0.43.A forest plot comparing rescue analgesia requirement between taVNS and control, with pooled reduction.Abbreviations: RR, risk ratio; CI, confidence interval; taVNS, transcutaneous auricular vagus nerve stimulation. Figure 6Forest plot of rebound pain incidence. Comparison of taVNS versus sham for the incidence of rebound pain after peripheral nerve block resolution. Data were pooled using risk ratio (RR) with a random-effects model.Forest plot with study table and risk ratio plot. Columns: Study/Subgroup; taVNS Events/Total; Control Events/Total; Weight; Risk Ratio, M H, Random, 95% CI. Row 1: Zhang J 2025, taVNS 6/34, Control 19/33, Weight 49.1%, Risk Ratio 0.31, 95% CI 0.14-0.67. Row 2: Zhou Q 2022, taVNS 7/39, Control 16/39, Weight 50.9%, Risk Ratio 0.44, 95% CI 0.20-0.94. Total: taVNS 73, Control 72, Weight 100%, pooled Risk Ratio 0.37, 95% CI 0.21-0.64. Total events: taVNS 13, Control 35. Heterogeneity: tau superscript 2=0.00; chi superscript 2=0.40, df=1, P=0.53; I superscript 2=0%. Overall effect: Z=3.58, P=0.0003. Right plot x-axis: Risk Ratio, M H, Random, 95% CI, logarithmic scale 0.01-100, ticks at 0.01, 0.1, 1, 10, 100. Vertical line at 1. Square markers at 0.31 (0.14-0.67) and 0.44 (0.20-0.94). Diamond marker at 0.37 (0.21-0.64). Bottom: Favours experimental (left), Favours control (right).A forest plot comparing rebound pain incidence between taVNS and sham, with risk ratios below 1.Abbreviations: RR, risk ratio; CI, confidence interval; taVNS, transcutaneous auricular vagus nerve stimulation. Figure 7Forest plot of postoperative nausea and vomiting (PONV) incidence. Comparison of taVNS versus sham for the incidence of PONV. Data were pooled using risk ratio (RR) with a random-effects model.Forest plot with study table and risk ratio plot. Columns: Study/Subgroup; taVNS Events/Total; Control Events/Total; Weight; Risk Ratio, M-H, Random, 95% CI. Studies: Huang H 2025 (taVNS 5/65, control 13/64, weight 20.4%, risk ratio 0.38, CI 0.14-1.00), Wang YZ 2026 (5/47 vs 4/47, 12.6%, 1.25, 0.36-4.37), Xia JY 2025 (3/35 vs 10/35, 13.6%, 0.30, 0.09-1.00), Yin ZY 2025 (5/45 vs 6/45, 15.8%, 0.83, 0.27-2.54), Zhou Q 2022 (10/39 vs 13/39, 37.5%, 0.77, 0.38-1.54). Total: taVNS 231, control 230, weight 100%, pooled risk ratio 0.63, CI 0.40-0.99. Total events: 28 taVNS, 46 control. Heterogeneity: Tau superscript 2=0.02; Chi superscript 2=4.27, df=4, P=0.37; I superscript 2=6%. Overall effect: Z=1.99, P=0.05. Right plot x-axis: Risk Ratio, M-H, Random, 95% CI, scale 0.01, 0.1, 1, 10, 100. Bottom labels: Favours experimental/control. Each study is a square with a horizontal CI line; pooled effect is a diamond at 0.63, spanning 0.40-0.99.A forest plot comparing postoperative nausea and vomiting incidence between taVNS and control, mostly below 1.Abbreviations: RR, risk ratio; CI, confidence interval; taVNS, transcutaneous auricular vagus nerve stimulation; PONV, postoperative nausea and vomiting. PRISMA 2020 flow diagram of study selection. Risk of bias assessment of included studies using the Cochrane RoB 2 tool. D1, bias arising from the randomization process; D2, bias due to deviations from intended interventions; D3, bias due to missing outcome data; D4, bias in measurement of the outcome; D5, bias in selection of the reported result. Green “+” indicates low risk of bias; yellow “?” in individual domains and yellow “!” in the overall assessment indicate some concerns. Forest plot of postoperative pain scores. Comparison of taVNS versus sham for postoperative pain intensity at early (2–6 h), middle (8–24 h), and late (48–72 h) time points. Data were pooled using a random-effects model. Forest plot of postoperative opioid consumption within the first 24 hours. Comparison of taVNS versus sham for cumulative opioid consumption. Because different opioids and units were used across studies, data were pooled using standardized mean difference (SMD) with a random-effects model. Substantial heterogeneity was observed (I2 = 87%). Forest plot of rescue analgesia requirement. Comparison of taVNS versus sham for the proportion of patients requiring rescue analgesia postoperatively. Data were pooled using risk ratio (RR) with a random-effects model. Forest plot of rebound pain incidence. Comparison of taVNS versus sham for the incidence of rebound pain after peripheral nerve block resolution. Data were pooled using risk ratio (RR) with a random-effects model. Forest plot of postoperative nausea and vomiting (PONV) incidence. Comparison of taVNS versus sham for the incidence of PONV. Data were pooled using risk ratio (RR) with a random-effects model.