Efficacy of Pulsed Radiofrequency Combination Therapy for Postherpetic Neuralgia: A Systematic Review and Bayesian Network Meta-Analysis Authors: Beibei Li, Xiong Xiang, Ranyu Li, Peng Zhou, Jian Jia Published: 2026 ABSTRACT AbstractBackgroundPostherpetic neuralgia (PHN) lowers quality of life. However, the comparative efficacy of pulsed radiofrequency (PRF)-based combination therapies remains unclear. This study compared different PRF combination regimens for PHN.MethodsEight databases were searched up to July 25, 2025. Risk of bias was assessed using the Cochrane RoB 2 and the Newcastle‑Ottawa Scale. Pain was the primary outcome; sleep, anxiety, and depression were secondary. A Bayesian network meta‑analysis was performed in R (version 4.4.1), with effect sizes expressed as mean differences (MDs) or standardized mean differences (SMDs), alongside 95% credible intervals (CrIs) and surface under the cumulative ranking curve (SUCRA) values.ResultsSeventeen studies were included. For pain relief, PRF + radiofrequency ablation (RFA) (MD = −2.1, 95% CrI: [−2.7, −1.5], SUCRA = 0.96) was the optimal regimen at one week, while PRF+glucocorticoid (GC) was superior at around one month (MD = −2.3, 95% CrI: [−2.6, −1.9], SUCRA = 0.96) and at three months (MD = −2.3, 95% CrI: [−2.7, −2.0], SUCRA = 0.97). Regarding sleep, PRF + intravenous lidocaine infusion (SMD = −9.2, 95% CrI: [−11, −7.5], SUCRA = 0.99) was most effective at one week, whereas PRF+duloxetine (SMD = −1.5, 95% CrI: [−2.0, −1.1], SUCRA = 0.98) was superior at three months. Anxiety was most effectively alleviated by PRF + intravenous lidocaine infusion (SMD = −1.6, 95% CrI: [−2.1, −1.0], SUCRA = 0.94). PRF+duloxetine (SMD = −2.5, 95% CrI: [−3.1, −2.0], SUCRA = 0.99) was the optimal combination for mitigating depression.ConclusionPRF+RFA may be the optimal combination for pain relief at one week, and PRF+GC appears more effective at around one month and over longer follow-up. Lidocaine infusion may benefit sleep at one week and anxiety, while duloxetine may enhance longer-term sleep and alleviate depression. Additional trials are warranted.Trial NumberCRD420251155850.Keywords: Bayesian network meta-analysis, combination therapy, postherpetic neuralgia, pulsed radiofrequency Introduction Varicella-zoster virus is a ubiquitous double-stranded DNA alphaherpesvirus. Primary varicella (chickenpox) is caused by this virus; lifelong latency is subsequently established in sensory ganglia, and later reactivation leads to herpes zoster (HZ) (shingles). Clinically, HZ may be divided into prodromal, acute, and chronic stages. The prodromal stage is characterized by systemic symptoms and abnormal skin sensations before rash onset. The acute stage is marked by a painful dermatomal rash. Persistent pain after rash resolution may progress to postherpetic neuralgia (PHN).1 PHN is a common clinical complication of HZ. Although a precise definition remains unconsolidated, it is primarily defined by persistent neuropathic pain that continues after rash healing. The pain is typically described as burning, stabbing, or accompanied by abnormal itching sensations.2,3 The painful area may even extend beyond the initially affected dermatome in some affected patients.4 The risk of developing PHN among individuals with HZ varies widely, from 5% to over 30%, a difference largely attributed to varying definitions of PHN.5 A recent cross-sectional study revealed that the prevalence rate of PHN (defined as pain persisting for more than one month after rash healing) in China was 40.2%.6 The prevalence rate of PHN increases with age and is highest among individuals over 70.6 Pain severity in PHN is significantly associated with impairments in physical function, mental health, and quality of life.7,8 Accordingly, exploring and optimizing treatment strategies for PHN is a major clinical concern. Currently, no specific therapy is available for PHN in clinical practice. The therapeutic goal is to relieve pain, facilitate sleep, and enhance quality of life. Frequently employed approaches include drug therapy and intervention therapy.9 Although pharmacotherapy remains the most common and fundamental method, its efficacy and dose-related side effects limit its application in older adults. In contrast, intervention therapies, such as peripheral nerve blocks, epidural injections, and neuromodulation techniques, are attracting increasing attention and are being applied more frequently for PHN. Among these modalities, neural regulation has been shown to yield relatively pronounced analgesic effects.10 Spinal cord stimulation, a high-risk and high-cost surgical procedure, faces challenges for widespread use.11,12 Radiofrequency thermocoagulation causes irreversible damage to nerve fibers and myelin sheaths, which may exacerbate postoperative pain and lead to paresthesia, weakened muscle strength, and other adverse effects.13,14 In contrast, pulsed radiofrequency (PRF), a minimally invasive, low-cost, and repeatable targeted neuromodulation technique, is being increasingly applied in clinically treating PHN.15,16 PRF employs intermittent radiofrequency current without causing thermal destruction, and this reduces the release of pain mediators and abnormal ectopic discharges in nerves, thereby alleviating pain.17,18 According to the Levels of Evidence suggested by the Oxford Centre for Evidence-Based Medicine, PRF is recommended as a second-line treatment for PHN.19 Furthermore, PRF is recommended as an interventional regimen in both the 2016 and 2022 Chinese expert consensus guidelines on PHN treatment.20,21 Nevertheless, studies have indicated that the short-term recurrence rate after PRF alone is relatively high, and additional treatment is usually required.13 Multimodal analgesia, the most widely used strategy in pain management, targets PHN via multiple mechanisms, thereby enhancing efficacy and prolonging analgesia to achieve optimal outcomes.22 Previous research has primarily compared radiofrequency therapy with single traditional drugs or other interventional approaches. However, comparative studies on the efficacy of radiofrequency combined with various adjunctive regimens are lacking. Although PRF is recommended by current guidelines for treating PHN, the optimal adjunctive therapy remains uncertain. Differing from conventional pairwise meta-analysis, network meta-analysis (NMA) integrates direct and indirect evidence to simultaneously compare multiple PRF-based combination therapies. Therefore, this study aimed to assess their comparative efficacy and provide a reference for selecting an individualized treatment. Methods Design and Registration This study was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.23,24 The study protocol was registered with the International Prospective Register of Systematic Reviews (PROSPERO, https://www.crd.york.ac.uk/PROSPERO/) (Registration number: CRD420251155850). As this study was based on previously published clinical research, either ethical approval or informed consent was waived. Literature Search Cochrane Library, Embase, PubMed, Web of Science, SinoMed, CNKI, WanfangData, and VIP Database were systematically searched. English subject terms (eg, “postherpetic neuralgia” and “radiofrequency”) and their Chinese equivalents were used, combined with Boolean operators to construct specific search strings for each database. Detailed search strategies are provided in Table S1. A manual supplementary search of the reference lists from relevant articles was also performed to identify additional records. The search was restricted to publications in Chinese or English, with a search period covering up to July 25, 2025 (Table S1). Eligibility Criteria Inclusion Criteria (i) Participants were patients with a documented history of acute HZ and persistent pain lasting more than one month after the acute phase. The pain was characterized by intense spontaneous or paroxysmal pain within the affected nerve distribution, accompanied by skin irritation. The pain quality included descriptors such as cutting, electric shock, burning, or needling, and was associated with sensations of tightness, itching, formication, or twitching. (ii) The interventions in included studies involved PRF combined with other therapies. (iii) The control group in included studies received PRF as the foundational treatment. (iv) Included studies reported outcome indicators such as pain severity, sleep quality, and levels of anxiety or depression. (v) The study design of included literature was a randomized controlled trial (RCT) or a cohort study. Exclusion Criteria Studies were excluded if: (i) a participant diagnosis was unclear, patients were under 18 years old, or pain duration was less than one month; (ii) an intervention was radiofrequency ablation (RFA) alone and not combined with other therapies; (iii) relevant outcome data could not be extracted; or (iv) a study design was neither an RCT nor a cohort study. Literature Screening All retrieved records were imported into EndNote 21 to remove duplicates. Two investigators (Jia J and Li BB) independently screened all titles and abstracts to identify relevant literature, and then assessed the full texts of potentially eligible studies. Literature screening was performed independently and simultaneously by both investigators, with any discrepancies reconciled through discussion with a third researcher. The final screening results were compiled and summarized by a third investigator. Data Extraction Two researchers (Jia J and Li BB) independently reviewed all included studies and extracted relevant data using a pre-designed form. The extracted information included: first author, publication year, country, sample size, age, disease duration, radiofrequency parameters, intervention measures, control measures, study design, pain scores, sleep quality scores, and anxiety or depression scores. Pain scores for patients with PHN after treatment were used as the primary data source. For continuous variables, standard deviations (SDs) were directly used. If SDs were not provided, they were derived from standard errors (SEs), 95% confidence intervals (CIs), ranges, or interquartile ranges (IQRs). When IQRs were reported, the mean was taken as the IQR midpoint and SD as IQR/1.135. For outcomes measured with different scales, score directions were harmonized, and standardized mean differences (SMDs) were calculated from the extracted means and SDs. All data and calculations were independently rechecked. Treatment nodes were defined according to the main therapeutic modality and route of administration; interventions with similar clinical purposes and delivery approaches were grouped into the same node. Risk of Bias Assessment Risk of bias was independently assessed by two investigators (Jia J and Li BB) using the Cochrane RoB 2.25 Any disagreements were reconciled through consultation with a third researcher. This tool, widely used in systematic reviews and meta-analyses, assesses five domains (randomization process, deviations from intended interventions, missing outcome data, measurement of outcomes, and selection of reported results), contributing to an overall risk of bias judgment of low risk, some concerns, or high risk. For cohort studies, the Newcastle-Ottawa scale (NOS)26 was applied, covering eight items across three dimensions (selection, comparability, and outcome assessment); each item was rated “YES” (1 point), “NO”, or “NI”, with total scores from 0 to 9 (7–9 = high quality, 5–6 = moderate quality, 0–4 = low quality). Evidence quality was assessed using the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach (http://www.gradeworkinggroup.org) at the outcome level; however, formal network GRADE or CINeMA frameworks were not used for individual comparisons. Overall, the risk-of-bias assessment relied on the evaluation of the quality of the randomization process, deviations from intended interventions, missing outcome data, or selection of reported results. GRADE assesses evidence quality by evaluating risk of bias and consistency across studies. Any discrepancies were reconciled through discussion with a third investigator. All studies were included regardless of their risk of bias or methodological quality, and potential bias was explored through sensitivity analyses by excluding studies one by one. Statistical Analysis The model for this NMA was constructed using the “GeMTC” package in R (version 4.4.1). All statistical analyses were performed under a Bayesian random-effects model, supported by the “GeMTC” and “rjags” packages. Study design was not included as an adjustment covariate, and therefore a sensitivity analysis excluding cohort studies was performed to assess the robustness of the findings. The transitivity assumption was evaluated by comparing the distribution of key potential effect modifiers (patient characteristics, disease duration, PRF parameters, adjunctive treatment protocols, outcome measurement scales, and follow-up durations) across treatment comparisons. The posterior distributions of parameters were simulated via the Markov chain Monte Carlo (MCMC) method, as the Bayesian approach provided a probability distribution for each effect estimate. Model parameters were set with 25,000 tuning iterations, followed by 250,000 simulation iterations (thinning factor = 10) to generate the final Bayesian model. Model consistency was validated using both consistency and inconsistency models. For closed loops in the network diagram, node‑splitting was applied to assess local inconsistency between direct and indirect evidence. Model convergence was diagnosed using the Brooks‑Gelman‑Rubin plot, together with visual inspection of trace and density plots. Continuous outcomes were synthesized as mean differences (MDs) or SMDs with 95% credible intervals (CrIs), and statistical significance was declared when the 95% CrI excluded zero. Treatment rankings were estimated using the surface under the cumulative ranking curve (SUCRA), where a higher value indicates a greater probability of being ranked higher. SUCRA values were interpreted alongside effect estimates, 95% CrIs, study quality, and network structure, as ranking probabilities do not directly reflect evidence certainty or robustness. Publication bias and small-study effects for pain-related outcomes at three follow-up periods were explored via funnel plots and Egger’s regression test. Network geometry plots were drawn to illustrate treatment relationships, and the numbers of direct comparisons and closed loops were summarized to characterize network structure. Results Literature Search In total, 1,644 records were retrieved from eight databases. After 881 duplicate records were removed, the titles and abstracts of the remaining 763 studies were assessed, and 65 were excluded. The full texts of the remaining 698 studies were further assessed. Of these, 66 were excluded because the study population did not consist of patients with PHN, 347 were removed due to inappropriate interventions, 173 were not RCTs or cohort studies, 58 were available only as abstracts, 4 were animal studies, and 2 were not published in Chinese or English. Accordingly, 48 studies met the initial criteria. After full‑text review by two researchers (Jia J and Li BB), 23 studies were excluded for inappropriate design, 2 for unavailable full texts, 4 for duplication, and 2 for not reporting relevant data. Ultimately, 17 studies (13 RCTs and 4 cohort studies) were included. Literature screening is illustrated in Figure S1 and Table S2. Characteristics of Included Studies The 17 included studies27–43 involved 1,481 individuals with PHN (737 in control groups and 744 in experimental groups). Baseline characteristics are summarized in Table S2. Comparisons were made between PRF alone and PRF combined with 13 different modalities: RFA, traditional Chinese medicine (TCM), local injection, chemical neurolysis, nerve block, glucocorticoid (GC), ozonated autohaemotherapy (O3‑AHT), local platelet-rich plasma (PRP) injection at the nerve root, intravenous lidocaine infusion, high-intensity laser therapy (HILT), buprenorphine transdermal patch, local ozone (O3) injection at the nerve root, and oral duloxetine. Mean patient age ranged from 57 to 72 years, and pain duration exceeded one month in all cases. In three trials, maximum pain duration was ≤3 months.28,34,40 In another three, it was ≤6 months.35,37,43 All included studies reported pain scores. Pain scores were reported in all studies, sleep quality scores in eight trials, and anxiety or depression scores in five trials. The included studies consisted of 13 RCTs and 4 cohort studies. Risk of Bias Results Risk of bias for the 13 RCTs was independently assessed using the Cochrane RoB 2, with data processing and chart generation performed in Excel. In the domain of the randomization process, two studies did not describe a random allocation method and were judged as having some concerns. Regarding deviations from intended interventions, one study did not meet blinding criteria. However, upon further assessment, it was determined that this was primarily due to the study context and was unlikely to affect outcomes or interventions. Accordingly, that study was also rated as having some concerns. In the domain of missing outcome data, all or nearly all outcome data were available across all RCTs, and hence, they were assessed to be at low risk. For measurement of outcomes, all RCTs reported clear methods for outcome measurement and were judged as having a low risk of bias. In the domain of selection of reported results, one study did not specify a pre-established analysis plan, preventing an assessment of bias in the reporting of results from data analysis. Accordingly, that study was rated as having some concerns. Risk of bias assessments are detailed in Figures S2 and S3. The four cohort studies were evaluated using the NOS, with overall scores ranging from 8 to 9, indicating high quality. Two studies scored 8 (mainly due to insufficient follow‑up duration), while the other two met all criteria and scored 9. Details are provided in Table S3. Using the GRADE approach, the certainty of evidence for the pain outcome derived from RCTs was rated as moderate. The risk of bias was judged as serious because some studies did not clearly report randomization methods or implement blinding. Inconsistency, indirectness, imprecision, and other considerations were not serious; therefore, the overall certainty was downgraded from high to moderate. For sleep, anxiety, and depression outcomes from RCTs, the certainty was rated as high (Table S4A). For cohort studies, evidence certainty starts at low, and no upgrading factors were identified for any outcome. Accordingly, overall certainty for pain, sleep, anxiety, and depression was rated as low (Table S4B). These ratings reflected outcome-level certainty and should not be interpreted as certainty estimates for each individual network comparison. Sensitivity Analysis Excluding Cohort Studies Sensitivity analyses were performed by excluding the four cohort studies and restricting the NMA to RCTs only. After this exclusion, the PRF+RFA node was no longer present because its supporting evidence was derived exclusively from cohort studies. The overall trends of the remaining treatment effects were generally consistent with the primary analysis. Detailed results are provided in Supplementary Tables S5–S7 and Figures S4–S6. Postoperative Pain Score Regarding pain at one week after treatment, 16 studies involving 1,373 patients with PHN were included. Compared with PRF alone, PRF+RFA (MD = −2.1, 95% CrI: [−2.7, −1.5], SUCRA = 0.96) showed the highest SUCRA value in pain relief at one week after treatment. However, this finding was based on a cohort study and was therefore considered low-certainty evidence. Other regimens that showed favorable effects (ranked by SUCRA) were: PRF+GC (MD = −2.0, 95% CrI: [−2.3, −1.6], SUCRA = 0.94), PRF + oral duloxetine (MD = −1.5, 95% CrI: [−2.0, −1.1], SUCRA = 0.79), PRF + buprenorphine transdermal patch (MD = −1.4, 95% CrI: [−2.0, −0.79], SUCRA = 0.71), PRF + local O3 injection at the nerve root (MD = −1.3, 95% CrI: [−1.9, −0.79], SUCRA = 0.69), PRF + nerve block (MD = −1.3, 95% CrI: [−1.7, −0.94], SUCRA = 0.68), PRF + intravenous lidocaine infusion (MD = −0.97, 95% CrI: [−1.5, −0.50], SUCRA = 0.51), PRF + chemical neurolysis (MD = −0.84, 95% CrI: [−1.2, −0.48], SUCRA = 0.43), PRF + local PRP injection at the nerve root (MD = −0.81, 95% CrI: [−0.94, −0.69], SUCRA = 0.41), PRF+TCM (MD = −0.76, 95% CrI: [−0.98, −0.53], SUCRA = 0.37), and PRF + intradermal injection (ID) (MD = −0.45, 95% CrI: [−0.69, −0.20], SUCRA = 0.22). However, for PRF+O3-AHT (MD = 0.12, 95% CrI: [−1.1, 1.3], SUCRA = 0.14) and PRF+HILT (MD = 0.96, 95% CrI: [−0.02, 1.9], SUCRA = 0.01), no statistically significant difference in reducing pain at one week after treatment was observed. The forest plot is detailed in Figure S7a. Although apparent closed loops were present, they arose solely from multi-arm studies and did not provide independent direct and indirect evidence, and therefore local inconsistency could not be assessed. The network plot is presented in Figure S7b, and the league table is provided in Table S8. In terms of pain at around one month after treatment, 16 studies covering 1,387 individuals suffering from PHN were included. In comparison with PRF alone, PRF + GC injection at the nerve root (MD = −2.3, 95% CrI: [−2.6, −1.9], SUCRA = 0.96) yielded the highest SUCRA value in pain relief at approximately one month after treatment. This regimen was followed by PRF+RFA (MD = −2.2, 95% CrI: [−2.8, −1.7], SUCRA = 0.95), PRF + intravenous lidocaine infusion (MD = −1.8, 95% CrI: [−2.3, −1.3], SUCRA = 0.84), PRF + oral duloxetine (MD = −1.3, 95% CrI: [−1.8, −0.86], SUCRA = 0.69), PRF + local O3 injection at the nerve root (MD = −1.2, 95% CrI: [−1.8, −0.58], SUCRA = 0.63), PRF+TCM (MD = −1.2, 95% CrI: [−1.4, −0.94], SUCRA = 0.62), PRF + buprenorphine transdermal patch (MD = −1.1, 95% CrI: [−1.7, −0.58], SUCRA = 0.59), PRF + chemical neurolysis (MD = −0.95, 95% CrI: [−1.3, −0.61], SUCRA = 0.50), and PRF + local PRP injection at the nerve root (MD = −0.83, 95% CrI: [−0.95, −0.71], SUCRA = 0.42). PRF + nerve block showed a trend toward lower pain scores at around one month (MD = −0.5, 95% CrI: [−1.0, 0.00], SUCRA = 0.31). In contrast, PRF+ID (MD = −0.10, 95% CrI: [−0.35, 0.16], SUCRA = 0.19) and PRF+O3-AHT (MD = 0.14, 95% CrI: [−1.1, 1.4], SUCRA = 0.16) showed no statistically significant differences compared with PRF alone. PRF+HILT was associated with a higher pain score than PRF alone at around one month (MD = 1.2, 95% CrI: [0.28, 2.1], SUCRA = 0.01), suggesting a less favorable effect for this outcome. The forest plot is detailed in Figure S8a. Although apparent closed loops existed, they arose solely from multi-arm studies and did not provide independent direct and indirect evidence, and therefore local inconsistency could not be evaluated. The network plot is provided in Figure S8b, and the league table is presented in Table S9. Concerning pain longer-term follow-up (generally ≥3 months) after treatment, 12 studies involving 1,113 patients with PHN were included. Compared with PRF alone, PRF + GC injection at the nerve root (MD = −2.3, 95% CrI: [−2.6, −2.0], SUCRA = 0.97) showed the highest SUCRA value in pain alleviation at intermediate- to longer-term follow-up. This combination was followed by PRF+RFA (MD = −2.2, 95% CrI: [−2.8, −1.6], SUCRA = 0.94), PRF + buprenorphine transdermal patch (MD = −1.5, 95% CrI: [−2.0, −0.94], SUCRA = 0.78), PRF+TCM (MD = −1.3, 95% CrI: [−1.6, −0.99], SUCRA = 0.72), PRF + oral duloxetine (MD = −1.2, 95% CrI: [−1.7, −0.74], SUCRA = 0.68), PRF + local PRP injection at the nerve root (MD = −0.9, 95% CrI: [−1.0, −0.77], SUCRA = 0.53), PRF + local O3 injection at the nerve root (MD = −0.81, 95% CrI: [−1.5, −0.12], SUCRA = 0.48), and PRF + chemical neurolysis (MD = −0.62, 95% CrI: [−0.94, −0.30], SUCRA = 0.37). PRF + nerve block (MD = −0.5, 95% CrI: [−0.95, −0.05], SUCRA = 0.31) and PRF+ID (MD = −0.33, 95% CrI: [−0.62, −0.04], SUCRA = 0.23). In contrast, for PRF+O3-AHT (MD = −0.52, 95% CrI: [−1.8, 0.75], SUCRA = 0.35) and PRF +HILT (MD = 0.34, 95% CrI: [−0.59, 1.3], SUCRA = 0.05), no statistically significant difference in mitigating pain at intermediate- to longer-term follow-up was noted. The forest plot is presented in Figure S9a. Although apparent closed loops were present, they arose solely from multi-arm studies and did not provide independent direct and indirect evidence. Therefore, local inconsistency could not be assessed. The network plot is provided in Figure S9b, and the league table is presented in Table S10. For pain-related outcomes at the three time points, the networks included 15, 15, and 14 direct comparisons, respectively, and apparent loops arose only from multi-arm studies and did not constitute independent closed loops. Funnel plots for pain-related outcomes were not significantly asymmetric, and Egger’s tests were non-significant (all P > 0.05). However, publication bias could not be excluded because several comparisons included few studies (Figure S10). Other outcomes were reported in too few studies for reliable assessment. Postoperative Sleep Score Regarding sleep quality at one week after treatment, 7 studies involving 644 individuals with PHN were included. Compared with PRF alone, PRF + intravenous lidocaine infusion (SMD = −9.2, 95% CrI: [−11, −7.5], SUCRA = 0.99) yielded the highest SUCRA value in enhancing sleep quality at one week after treatment. This unusually large estimate should be interpreted cautiously, as it may partly reflect differences in scale variability and limited sample size. Other regimens ranked by SUCRA were: PRF + oral duloxetine (SMD = −2.5, 95% CrI: [−3, −2], SUCRA = 0.86), PRF+HILT (SMD = −1.2, 95% CrI: [−1.7, −0.72], SUCRA = 0.65), PRF+ID (SMD = −0.97, 95% CrI: [−1.5, −0.48], SUCRA = 0.54), PRF + chemical neurolysis (SMD = −0.92, 95% CrI: [−1.4, −0.43], SUCRA = 0.51), and PRF + local PRP injection at the nerve root (SMD = −0.44, 95% CrI: [−0.77, −0.11], SUCRA = 0.30). In contrast, PRF+TCM (SMD = 0.16, 95% CrI: [−0.21, 0.53]) showed no significant difference. The forest plot is presented in Figure S11a. As the network structure did not form closed loops, an inconsistency test could not be performed. The network plot is provided in Figure S11b, and the league table is presented in Table S11. In terms of sleep quality at intermediate- to longer-term follow-up, five studies covering 630 patients with PHN were included. In comparison with PRF alone, PRF + oral duloxetine (SMD = −1.5, 95% CrI: [−2, −1.1], SUCRA = 0.98) showed the highest SUCRA value in enhancing sleep quality at intermediate- to longer-term follow-up. This combination regimen was followed by PRF + chemical neurolysis (SMD = −1.1, 95% CrI: [−1.6, −0.64], SUCRA = 0.80), PRF+HILT (SMD = −0.88, 95% CrI: [−1.4, −0.41], SUCRA = 0.67), and PRF+ID (SMD = −0.63, 95% CrI: [−1.1, −0.16], SUCRA = 0.53). However, PRF +O3-AHT (SMD = 0.2, 95% CrI: [−0.08, 0.47], SUCRA = 0.20) and PRF + local PRP injection at the nerve root (SMD = −0.23, 95% CrI: [−0.64, 0.18], SUCRA = 0.32) did not show statistically credible differences from PRF alone for enhancing sleep quality at intermediate- to longer-term follow-up. The forest plot is provided in Figure S12a. Since the network structure did not form closed loops, an inconsistency test was infeasible. The network plot is presented in Figure S12b, and the league table is provided in Table S12. Postoperative Anxiety Score Three studies involving 257 individuals with PHN were included for anxiety assessment. Compared with PRF alone, PRF + intravenous lidocaine infusion (SMD = −1.6, 95% CrI: [−2.1, −1.0], SUCRA = 0.94) yielded the highest SUCRA value in anxiety alleviation within the available network. This combination regimen was followed by PRF+TCM (SMD = −1.3, 95% CrI: [−1.9, −0.64], SUCRA = 0.81). In contrast, PRF+ID (SMD = −0.41, 95% CrI: [−0.88, 0.06], SUCRA = 0.44) and PRF+HILT (SMD = −0.13, 95% CrI: [−0.58, 0.32], SUCRA = 0.23) did not show statistically credible differences from PRF alone for anxiety alleviation at around one month after treatment. The forest plot is presented in Figure S13a. Because the network structure did not form closed loops, an inconsistency test could not be implemented. The network plot is provided in Figure S13b, and the league table is presented in Table S13. Postoperative Depression Score For depression assessment, four studies involving 360 patients with PHN were included. In comparison with PRF alone, PRF + oral duloxetine (SMD = −2.5, 95% CrI: [−3.1, −2.0], SUCRA = 0.99) showed the highest SUCRA value in mitigating depression. In contrast, the following combination regimens demonstrated no statistically significant difference in alleviating depression over one month after treatment, comprising PRF + intravenous lidocaine infusion (MD = 0.51, 95% CrI: [0.02, 1.0], SUCRA = 0.02), PRF+ID (MD = −0.07, 95% CrI: [−0.52, 0.38], SUCRA = 0.44), PRF+HILT (MD = −0.18, 95% CrI: [−0.63, 0.27], SUCRA = 0.55), and PRF+TCM (MD = −0.34, 95% CrI: [−0.92, 0.25], SUCRA = 0.66). The forest plot is provided in Figure S14a. The network structure did not form closed loops; hence, an inconsistency test was infeasible. The network plot is presented in Figure S14b, and the league table is provided in Table S14. A summary of the highest-ranked interventions, effect estimates, SUCRA values, and certainty of evidence across clinical outcomes and follow-up time points is provided in Supplementary Table S15. Adverse events were inconsistently reported, precluding quantitative synthesis. Potential safety concerns included nerve injury after RFA, neurological complications from chemical neurolysis, and injection-related or systemic adverse effects of GCs. Discussion This NMA included 17 studies involving 1,481 patients with PHN. For pain-related outcomes, PRF+RFA showed the highest SUCRA value at one week after treatment, whereas PRF+GC yielded the highest SUCRA value at around one month and at intermediate- to longer-term follow-up. For sleep quality, PRF + intravenous lidocaine infusion showed the highest SUCRA value at one week after treatment, while PRF + oral duloxetine yielded the highest SUCRA value at intermediate- to longer-term follow-up. For anxiety, PRF + intravenous lidocaine infusion showed the highest SUCRA value. For depression, PRF + oral duloxetine yielded the highest SUCRA value. These rankings, however, should be interpreted with caution, as several intervention nodes were based on single small studies, and a high SUCRA value does not necessarily reflect high-certainty evidence of clinical superiority. In this NMA, three commonly reported follow-up time points (at one week, around one month, and intermediate‑ to longer‑term) were selected for pooled analysis of PHN pain scores. PRF+RFA ranked highest for short-term analgesia (at one week), whereas PRF+GC appeared more effective for pain reduction at around one month and at intermediate- to longer-term follow-up. However, PRF+RFA was no longer present in the RCT-only sensitivity analysis, as its supporting evidence was derived exclusively from observational studies, and accordingly, this ranking should be interpreted with caution. RFA rapidly alleviates pain by reducing abnormal discharges in targeted nerves through thermal injury.44 GC mitigates local inflammatory responses and neural edema, potentially improving changes related to inflammatory demyelination,45–47 thereby reducing pain and neural sensitization.45,46,48,49 GC may also prolong the effect of PRF and may remain favorable at intermediate and longer-term follow-up. Nevertheless, the irreversible damage to nerve fibers and myelin sheaths caused by RFA may exacerbate pain, paresthesia, muscle weakness, and other adverse effects after treatment.13,14 Chemical neurolysis and GC injections may also carry neurological or injection-related risks. However, comparative safety could not be assessed because adverse events were poorly reported. Systemic medications (eg, duloxetine, buprenorphine patch, and intravenous lidocaine) serve as supplementary strategies for patients with comorbid neuropathic pain or specific populations. In this analysis, however, their analgesic effects appeared less favorable than those of targeted interventional combinations, consistent with previous evidence and current recommendations that prioritize systemic pharmacotherapy and reserve interventional approaches for selected patients with refractory PHN.50 Compared with some previous reports, this study found that the relative benefits of several PRF combination therapies were more apparent at intermediate and longer-term follow-up than in the short term, possibly reflecting the delayed onset of interventions oriented toward repair or regeneration.51 This suggests that long-term pain control in PHN may require strategies that address both neural repair and microenvironment improvement. In contrast, therapies such as PRF combined with O3, local drug injections, and HILT demonstrated unstable therapeutic effects or showed limited overall benefits across different time points. This may be related to the fact that PHN is primarily characterized by nerve injury and central sensitization, or influenced by variations in dosage parameters, treatment frequency, and pain phenotypes of patients, leading to insufficient consistency in the evidence. PRF+TCM may provide sustained benefit across follow-up phases, whereas the effects of strategies such as nerve block may diminish over time.50 Therefore, PRF+GC may represent one of the more effective regimens for pain relief at intermediate- to longer-term follow-up, though this finding should be interpreted cautiously given the limited precision of some comparisons. In clinical practice, targeted combination regimens may be considered as a core option, while systemic medications may serve as adjunctive, alternative, or bridging strategies in selected patients. Regarding sleep quality, PRF + intravenous lidocaine infusion ranked highest at one week after treatment, whereas PRF + oral duloxetine appeared most effective at intermediate- to longer-term follow-up after treatment. The very large SMD for short-term sleep improvement may be influenced by scale dispersion and small-study effects, and should not be interpreted as a proportionally large clinical benefit. PRF + chemical neurolysis also showed a favorable estimate at intermediate- to longer-term follow-up. The effects of different PRF combination therapies on sleep quality appeared to vary across follow-up phases, possibly reflecting the evolution of pain from an early nociceptive component toward chronic neuropathic features, as well as a shift in the drivers of sleep disturbance—from acute stress and severe pain to persistent pain and emotional comorbidities.52,53 Previous studies have suggested that intravenous lidocaine may reduce the requirements for perioperative sedative medications,54–56 with a mechanism potentially involving interactions with excitatory or inhibitory neurotransmitter or receptor systems in the central nervous system. Nevertheless, some research indicates that although intravenous lidocaine reduces pain-related sleep interference metrics with statistical significance, the effect size might be insufficient to translate into clinically perceptible improvement.57 Moreover, systematic studies on its independent impact on sleep quality or architecture are lacking, so reproducibility needs further validation. In contrast, duloxetine, a first‑line neuropathic pain medication,58 simultaneously modulates mood and alleviates pain. The sleep-improving effects of duloxetine typically emerge approximately two weeks after administration and may become more pronounced with longer follow-up (eg, 8 weeks).59–61 This may help explain why duloxetine ranked highest for longer-term sleep enhancement in this study. Consistent with the pain outcome analysis, PRF + chemical neurolysis showed a favorable estimate for sustained analgesia, possibly contributing to its relatively good performance at intermediate- to longer-term follow-up. Conversely, PRF combined with TCM, PRP, or O3 was not significantly superior in enhancing sleep quality at any follow-up phase, possibly due to longer onset windows, insufficient analgesic intensity, or limited effects on mood or sleep regulatory pathways. In summary, PRF + oral duloxetine ranked highest for sleep improvement, though the available evidence remains limited. In terms of anxiety alleviation, PRF + intravenous lidocaine infusion ranked highest, which may be related to its central neuromodulatory and sedative effects. However, lidocaine exerts limited direct modulatory effects on monoaminergic neurotransmitter systems (eg, 5-hydroxytryptamine [5-HT] and norepinephrine) associated with depression. Accordingly, intravenous lidocaine did not appear effective for optimizing depression outcomes in this study. For depression mitigation, PRF + oral duloxetine ranked highest. Previous studies have shown that duloxetine effectively regulates mood and alleviates depression by increasing brain 5‑HT and norepinephrine levels,62 and may offer comprehensive benefits, especially in patients with chronic pain accompanied by depression or sleep disorders.60 In contrast, PRF combined with HILT, local injection, or TCM acts primarily on peripheral nerves or local tissues, with limited direct impact on central emotion‑regulation circuits. Consequently, no statistically significant differences were observed for HILT or local injection in anxiety or depression, whereas TCM favored lower anxiety scores but did not show significant benefit for depression. Additionally, factors such as sample size, intervention duration, and scale sensitivity may affect the detection of differences. Statistical significance should be interpreted alongside clinical relevance. Reductions of approximately 2 points or 30% on 0–10 pain scales are generally considered clinically meaningful, whereas the clinical significance of SMD-based secondary outcomes remains less certain. In comparison with previous reviews,63–66 this study has several notable strengths. First, this NMA enabled the simultaneous comparison of multiple PRF-based combination regimens, going beyond conventional pairwise analyses. Second, pain, sleep, anxiety, and depression were evaluated, providing a multidimensional assessment. Third, outcomes were examined across different follow‑up phases, offering insight into temporal variations in treatment performance. Several limitations should be acknowledged. First, both RCTs and cohort studies were included without formal adjustment for study design. Although the RCT-only sensitivity analysis yielded generally consistent results, PRF+RFA was removed from the network, and residual confounding from observational studies could not be excluded. Second, the sparse network and single-study nodes limited the robustness of the estimates, and therefore SUCRA rankings should be interpreted alongside effect sizes and 95% CrIs, rather than as definitive evidence of superiority. Certainty was assessed at the outcome level, not for individual network comparisons, and no formal network GRADE or CINeMA assessment was performed. Third, variations in PRF parameters, adjunctive protocols, outcome scales, follow-up durations, and broad intervention categories may have introduced clinical and methodological heterogeneity. Although transitivity was assessed by comparing key potential effect modifiers, residual violations could not be excluded. Fourth, publication bias and limited generalizability remain possible, as most studies originated from one country, several treatment comparisons included few studies, and variability in adjunctive interventions and PRF regimens may limit applicability to broader clinical settings. Fifth, limited long-term follow-up restricted assessment of sustained efficacy and recurrence. Adverse events were inadequately reported, preventing comparative safety assessment, and therefore efficacy rankings should not be interpreted as overall treatment superiority. Overall, these findings should support hypothesis generation and individualized decision-making, rather than being considered definitive evidence for a single optimal regimen. Conclusion Different PRF combination therapies may have distinct relative probabilities of benefit across symptom domains in patients with PHN. PRF+GC appears effective for pain reduction, PRF + oral duloxetine is potentially beneficial for longer-term sleep and depression outcomes, and PRF + intravenous lidocaine infusion was superior in anxiety relief. Nevertheless, these rankings reflect relative probabilities rather than definitive superiority and should be interpreted with caution due to sparse treatment networks, small sample sizes, clinical heterogeneity, and insufficient safety reporting. Although the findings may assist in individualized, symptom-oriented treatment selection, they do not support the establishment of any single optimal regimen. Before specific PRF combination therapies can be recommended for routine clinical practice, large‑scale, multicenter, high‑quality RCTs with standardized protocols are still warranted. Funding Statement This research was not supported by any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Abbreviations PHN, postherpetic neuralgia; PRF, pulsed radiofrequency; RFA, radiofrequency ablation; O3-AHT, ozonated autohaemotherapy; PRP, platelet-rich plasma; HILT, high-intensity laser therapy; ID, intradermal injection; TCM, traditional Chinese medicine; GC, glucocorticoid; O3, ozone; 5-HT, 5-hydroxytryptamine. Data Sharing Statement All data generated or analyzed in this study are included in this published article and its supplementary information files. Author Contributions All authors contributed to the study conception and design. Writing - original draft preparation: [Beibei Li, Xiong Xiang]; Writing - review and editing: [Beibei Li, Xiong Xiang]; Conceptualization: [Beibei Li, Xiong Xiang; Ranyu Li; Peng Zhou; Jian Jia]; Methodology: [Beibei Li, Xiong Xiang; Ranyu Li; Peng Zhou; Jian Jia]; Formal analysis and investigation: [Beibei Li, Xiong Xiang; Ranyu Li; Peng Zhou; Jian Jia]; Funding acquisition: [Jian Jia]; Resources: [Jian Jia]; Supervision: [Jian Jia]. All authors commented on previous versions of the manuscript and reviewed and approved the final version. All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work. Disclosure The authors report no conflicts of interest in this work. References Associated Data Data Availability Statement All data generated or analyzed in this study are included in this published article and its supplementary information files.