Time-Dependent Efficacy and Safety of Percutaneous Treatments for Lateral Epicondylitis: A Systematic Review and Network Meta-Analysis

Introduction

Lateral epicondylitis (LE), commonly known as “tennis elbow,” is a chronic overuse injury primarily affecting the extensor carpi radialis brevis tendon. Rather than simple acute inflammation, the underlying pathology is now understood as tendon degeneration, microtearing, and failed healing.¹ It affects approximately 1% to 3% of the general population, mostly adults aged 35 to 55 who perform repetitive wrist movements or heavy gripping. Although the majority of cases are self-limiting, approximately 10% to 20% progress to refractory chronic pain, resulting in significant occupational absenteeism and healthcare resource consumption.¹

According to current clinical practice guidelines, initial management for LE prioritizes non-operative methods like physical therapy, eccentric resistance training, bracing, and oral non-steroidal anti-inflammatory drugs (NSAIDs).² However, clinical observation suggests that conservative management alone frequently reaches a therapeutic plateau in chronic, recalcitrant cases persisting beyond three months, often characterized by a delayed onset of efficacy.³ Surgery is considered a last resort due to its invasiveness and long recovery periods.⁴ Consequently, minimally invasive injection therapies and dry needling (DN), situated between conservative management and open surgery, have emerged as focal points of recent clinical research and application due to their low invasiveness and rapid recovery profiles.³ For frontline clinicians, a major daily challenge is deciding which percutaneous treatment to offer first for chronic recalcitrant LE, and how to logically sequence options like steroids, regenerative injections, or dry needling to optimize patient recovery.

Various interventions target different aspects of tendon healing. Corticosteroids (CS) are widely used for rapid pain relief, but their potential to harm tendon tissue raises concerns about long-term safety. In contrast, regenerative therapies such as platelet-rich plasma (PRP), autologous blood (AB), and dextrose prolotherapy (DPT) aim to facilitate tissue repair and collagen regeneration through the release of growth factors or the induction of sterile inflammation.^5–8 Furthermore, hyaluronic acid (HA) improves the microenvironment for tendon gliding,⁹ botulinum toxin (BT) forces the muscle to rest via temporary paralysis,¹⁰ and DN exhibits unique therapeutic potential by mechanically releasing tight tissue bands and improving local blood flow.¹¹ Placebo (PL) serves as the baseline to measure the true effects of these treatments.²

Despite these options, finding the best treatment strategy remains difficult. First, previous meta-analyses have predominantly relied on pairwise assessments rather than providing a comprehensive head-to-head comparison of the eight mainstream interventions within a unified framework.¹² Second, existing network meta-analyses frequently exclude DN from injection protocols, rendering the comparative efficacy of physical versus biochemical stimulation inconclusive.¹² Third, therapeutic efficacy exhibits significant time dependency, exemplified by the immediate effects of CS and the delayed benefits of PRP. However, prior studies often fail to stratify outcomes by specific time intervals or overlook recent evidence for emerging therapies such as HA and DPT.^9,13 Finally, a systematic quantitative ranking regarding safety profiles and specific adverse events remains inadequate.¹⁴ To address these gaps, our study significantly expands the current evidence base by integrating 41 recent randomized controlled trials (involving over 3,200 patients) into a unified network covering eight diverse interventions.

We aim to evaluate the relative efficacy and safety of PL, CS, PRP, AB, HA, BT, DPT, and DN across short-term, mid-term, and long-term follow-ups. Based on their biological mechanisms, our a priori hypothesis was that CS would provide superior short-term relief but worse long-term outcomes compared to regenerative therapies. Ultimately, this NMA provides actionable evidence to help clinicians choose the right percutaneous treatment at the right time for patients with LE.

Material and Methods

Protocol and Registration

This NMA followed the PRISMA 2020 guidelines and the PRISMA-NMA extension (Supplementary PRISMA Checklist).^15,16 The study protocol was prospectively registered in International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD420261287903).

Search Strategy

A comprehensive systematic search was performed across four major electronic databases: PubMed, Embase, The Cochrane Library, and Web of Science, covering the period from inception to September 2025. The search strategy employed a combination of Medical Subject Headings (MeSH) terms and free-text keywords relevant to the condition and interventions, including “Lateral epicondylitis”, “Tennis elbow”, “Platelet-rich plasma”, “Corticosteroids”, “Dry needling”, “Botulinum toxin”, and “Autologous blood”. No restrictions were imposed regarding language or publication date. To ensure literature saturation, the reference lists of eligible studies and relevant review articles were manually screened. The detailed search strings for each database are provided in Appendix 1 (Tables S1.1–S1.4).

Eligibility Criteria

Eligibility criteria were established a priori based on the PICOS framework. The population included adult patients (aged >18 years) with a confirmed clinical or radiological diagnosis of lateral epicondylitis (LE). Eligible interventions encompassed eight specific percutaneous therapies: PL, CS, PRP, AB, HA, BT, DPT, and DN. Studies involving standalone physical therapy, extracorporeal shock wave therapy, surgical interventions, or multimodal combinations that precluded the isolation of injection effects were excluded. Comparisons were restricted to randomized head-to-head trials between any of the aforementioned interventions. The primary outcome was pain intensity, assessed exclusively via the Visual Analog Scale (VAS). Secondary outcomes included functional disability, evaluated using validated scales such as the DASH and PRTEE, pain-free grip strength (PFGS), and maximum grip strength (GS). Safety endpoints included the incidence of adverse events and complications. Regarding study design, only RCTs were included; observational studies, reviews, case reports, and animal models were excluded.

Data Extraction and Quality Assessment

Data Selection and Categorization

Two independent reviewers (Y.X. and W.L.) screened titles, abstracts, and full-text articles to identify eligible studies, resolving any discrepancies through consultation with a senior reviewer (Z.Y). From each included trial, data regarding study characteristics, participant demographics, intervention protocols, and clinical outcomes were extracted. To account for time-dependent variations in therapeutic efficacy, outcome data were stratified into three distinct intervals: short-term (≤ 1 month), mid-term (> 1 month to ≤ 3 months), and long-term (> 6 months). In instances where studies reported multiple data points within a single interval, priority was assigned to the data point closest to the primary endpoint of that interval (4 weeks for short-term, 12 weeks for mid-term, and 24 weeks for long-term) to maximize consistency and minimize heterogeneity arising from varying follow-up durations.

Statistical Handling and Evidence Appraisal

For trials with multiple intervention arms belonging to the same treatment node, data were pooled into a single arm to prevent unit-of-analysis errors. Continuous outcomes reported as medians with ranges or interquartile ranges were converted to means and standard deviations (SDs) using the method described by Wan et al.¹⁷ For trials reporting only change-from-baseline scores, post-intervention means were calculated by subtracting the mean improvement from the baseline mean, with missing SDs imputed using baseline variability as recommended by the Cochrane Handbook.¹⁸ All grip strength data were standardized to kilograms to ensure comparability. Finally, the methodological quality of the included RCTs was appraised using the Cochrane Risk of Bias tool 2.0 (RoB 2),¹⁹ while the certainty of evidence for network estimates was graded using the Confidence in Network Meta-Analysis (CINeMA) framework^20,21 based on domains of within-study bias, reporting bias, indirectness, imprecision, heterogeneity, and incoherence.

Statistical Analysis

Network Meta-Analysis and Outcome Measures

A frequentist NMA was performed using Stata 19.0 (StataCorp, College Station, TX, USA),²² with network plots generated to visualize the geometry of participant numbers and direct comparisons.²³ Treatment effects were stratified by outcome metric: Mean Differences (MD) were calculated for uniform measures, specifically pain intensity and grip strength, whereas Standardized Mean Differences (SMD) were employed for functional disability to synthesize data derived from heterogeneous scales such as the DASH and PRTEE.¹⁸ To estimate the hierarchy of efficacy, Surface Under the Cumulative Ranking Curve (SUCRA) values were calculated,²⁴ where higher percentages indicate a greater likelihood of an intervention being the superior treatment.

Assessment of Validity and Robustness

Statistical heterogeneity was quantified by estimating the common between-study variance (τ²).²⁵ Global and local inconsistency were rigorously evaluated using the design-by-treatment interaction model²⁶ and the node-splitting method, respectively.²⁷ The transitivity assumption was assessed by comparing the distribution of clinical and methodological effect modifiers across studies. Regarding safety, a quantitative synthesis was precluded by inconsistent reporting standards; thus, adverse events were summarized qualitatively. Finally, to ensure the robustness of the primary findings, sensitivity analyses were conducted across all time intervals by excluding studies classified as having a high risk of bias.

Results

Search results

A total of 1,511 records were identified from the initial database search. After removing duplicates and screening titles and abstracts, 345 full-text articles were assessed for eligibility. Ultimately, 41 RCTs^28–68 met our inclusion criteria for the NMA (Figure 1).

Figure 1 PRISMA flow diagram of the study selection process.

Study Characteristics

A total of 41 RCTs (N=3,285) were included, with sample sizes ranging from 17 to 396. Participants had a mean age of 33.8–53.4 years and a relatively balanced gender distribution. Symptom duration varied widely (1.5 to >30 months). CS and PRP were the most frequently investigated interventions, followed by AB, DPT, and DN, while BT and HA were less common. Follow-up durations ranged from 1.5 to 24 months, enabling assessment across short-, mid-, and long-term time points. Detailed characteristics and the complete list of extracted data are summarized in Appendix 2 (Table S2.1) and Appendix 3, respectively.

Quality Assessment and Validity of Assumptions

Methodological Quality of Included Studies

Risk of bias assessment using RoB 2.0 identified 12 studies (29.3%) as “low risk,” 6 (14.6%) as “high risk,” and 23 (56.1%) as having “some concerns” (Figure 2; Appendix 4 and Table S4.1). While randomization processes were generally robust, the “measurement of the outcome” domain frequently raised concerns due to potential detection bias arising from subjective patient-reported outcomes in unblinded trials. Additionally, the “selection of the reported result” was often graded as “some concerns” due to the absence of pre-registered protocols or statistical analysis plans. Detailed risk of bias contributions and overall risk of bias across different treatment comparisons for all clinical outcomes are provided in Appendix 5 (Figures S5.1–S5.14).

Figure 2 Overall risk of bias presented as percentage of each risk of bias item across all included studies. Green = Low risk, Red = High risk, Yellow = Some concerns.

Statistical Consistency and Heterogeneity

Regarding quantitative statistical validity, the global inconsistency analysis using the design-by-treatment interaction model revealed no significant inconsistency across all evaluable networks (P > 0.05 for all outcomes), as detailed in Appendix 6 (Table S6.1). Additionally, the node-splitting analysis found no significant discrepancies between direct and indirect evidence in any closed loops (P > 0.05, Table S6.2–S6.8). The estimated between-study heterogeneity (τ²) ranged from 0.08 to 0.97, suggesting acceptable levels of heterogeneity within the network. Specifically, assessments for short-term outcomes (χ²=3.12–13.98, P > 0.05), mid-term outcomes (χ²=2.52–10.31, P > 0.05), and long-term pain intensity (χ²=12.85, P=0.46) consistently indicated that the NMA models fit the data well. Note that consistency assessments for long-term functional status and grip strength were precluded by the sparse network geometry and absence of closed loops.

Assessment of Clinical and Methodological Assumptions

The assessment of transitivity (Appendix 5, Table S5.1) confirmed that key clinical characteristics, such as mean age, disease duration, and sex distribution, were broadly balanced across comparisons, supporting the plausibility of the transitivity assumption. Furthermore, visual inspection of comparison-adjusted funnel plots (Appendix 7, Figures S7.1–S7.7) revealed a generally symmetric distribution, suggesting no significant small-study effects or reporting bias within the network.

Network Meta-Analysis

Short-Term Pain Intensity

This analysis included 24 RCTs (n=1,473). The network geometry presented a closed-loop structure with CS and PL as central nodes, supported by robust direct evidence (Figure 3A). According to the forest plot (Figure 3B) and summary rankings (Table 1; Appendix 8, Figure S8.1 and Table S8.1), CS was the only intervention significantly superior to PL (MD −1.62, 95% CI −2.52 to −0.72) and ranked highest (SUCRA 88.9%). DN and DPT showed substantial but non-significant benefits. CINeMA assessment indicated that evidence certainty was generally “Low” due to within-study bias and heterogeneity (Appendix 5 and Table S5.2). The league table revealed no statistically significant differences among active interventions, suggesting limited short-term disparity among primary treatments (Appendix 9 and Table S9.1).

Table 1 Summary of SUCRA Rankings and Clinical Efficacy Across Time Intervals

Figure 3 Network geometry and forest plots for short-term outcomes. (A) Network plot for short-term pain intensity. Node size is proportional to the total number of patients, and line thickness corresponds to the number of trials. (B) Forest plot for short-term pain intensity compared with placebo. Diamonds represent the mean difference (MD) with 95% confidence intervals (CIs). The left-pointing arrow indicates that a lower score favors the active treatment. (C) Network plot for short-term physical function. (D) Forest plot for short-term physical function compared with placebo. Diamonds represent the standardized mean difference (SMD) with 95% CIs. The left-pointing arrow indicates that a lower score favors the active treatment. (E) Network plot for short-term grip strength. (F) Forest plot for short-term grip strength compared with placebo. Diamonds represent the MD with 95% CIs. The right-pointing arrow indicates that a higher value favors the active treatment. Abbreviations: AB, Autologous Blood; BT, Botulinum Toxin; CS, Corticosteroids; DN, Dry Needling; DPT, Dextrose Prolotherapy; HA, Hyaluronic Acid; PL, Placebo; PRP, Platelet-Rich Plasma.

Short-Term Physical Function

In 23 studies (n=1,341), the network structure relied heavily on CS and PRP connections, with interventions like BT and DN anchored primarily by single-arm connections to PL (Figure 3C). According to the forest plot (Figure 3D) and SUCRA analysis (Table 1; Figure S8.2 and Table S8.2), DN emerged as the most promising intervention, achieving the highest ranking (SUCRA 93.8%) and exhibiting the most favorable efficacy trend (SMD −0.92, 95% CI −1.88 to 0.04), although this comparison did not reach statistical significance. DN was followed by CS (SUCRA 70.9%) and DPT (SUCRA 68.6%). The 95% CIs for all active interventions crossed the null line compared with PL. The overall evidence quality was graded as “Low” to “Very Low” due to serious imprecision (Appendix 5 and Table S5.3). The league table showed that DN was significantly superior to AB (SMD −1.09, 95% CI −2.09 to −0.09) in head-to-head comparisons (Appendix 9 and Table S9.2). No statistically significant differences were observed among the remaining pairwise comparisons.

Short-Term Grip Strength

Among 12 studies (n=710), the network geometry was centered around a triangular loop involving AB, CS, and PL (Figure 3E). As shown in Table 1, Figure S8.3, Table S8.3, and the forest plot (Figure 3F), AB (SUCRA 81.1%) and CS (SUCRA 73.1%) ranked highest but lacked statistical significance against PL. Conversely, BT was the sole intervention to demonstrate statistical significance against placebo, significantly reducing grip strength (MD −6.66, 95% CI −13.28 to −0.04) and ranking lowest (SUCRA 8.1%), raising potential safety concerns. CINeMA assessment showed that evidence quality was limited to “Low” or “Very Low” due to imprecision and bias (Appendix 5 and Table S5.4). The league table further revealed that BT was significantly inferior to AB (MD −10.22, 95% CI −18.94 to −1.51) and CS (MD −9.23, 95% CI −16.85 to −1.62) (Appendix 9 and Table S9.3). No statistically significant differences were observed among the remaining pairwise comparisons.

Mid-Term Pain Intensity

Across 29 RCTs (n=2,373), the network geometry was characterized by balanced connections among CS, PL, and PRP (Appendix 10, Figure S10.1). Forest plots (Appendix 10, Figure S10.1) indicated a significant shift in efficacy compared to short-term results: CS no longer showed a statistically significant difference compared with placebo (MD −0.65, 95% CI −1.52 to 0.22) and dropped in ranking (Table 1; Figure S8.4 and Table S8.4). DPT demonstrated the most robust pain reduction (MD −1.73, 95% CI −2.85 to −0.60; SUCRA 87.4%), followed by AB (MD −1.19, 95% CI −2.34 to −0.05; SUCRA 65.1%), both showing statistical superiority over PL. Notably, PRP also demonstrated statistical superiority over placebo (MD −0.96, 95% CI −1.90 to −0.02), although it ranked fifth (SUCRA 53.8%).Conversely, while DN ranked third (SUCRA 64.3%) (Table 1; Figure S8.4 and Table S8.4), it failed to reach statistical significance (MD −1.19, 95% CI −2.59 to 0.21) due to wide confidence intervals. CINeMA assessment results showed that evidence certainty for DPT and AB vs. PL was upgraded to “Moderate” (Appendix 5 and Table S5.5). Head-to-head comparisons in the league table confirmed that DPT was significantly superior to CS (MD −1.08, 95% CI −2.15 to −0.01), supporting its status as a preferred mid-term intervention (Appendix 9 and Table S9.4). No statistically significant differences were observed among the remaining pairwise comparisons.

Mid-Term Physical Function

In 24 studies (n=1,414), the network geometry was densely connected around CS, PL, and PRP (Appendix 10 and Figure S10.2). According to the forest plot (Appendix 10 and Figure S10.2) and summary rankings (Table 1; Figure S8.5 and Table S8.5), DPT was the sole intervention significantly superior to PL (SMD −1.23, 95% CI −2.17 to −0.28) and achieved the highest ranking (SUCRA 94.5%), establishing dominance in functional recovery. CINeMA assessment indicated that while DPT vs. PL evidence was “Very Low”, comparisons involving other agents contained “Low” certainty evidence (Appendix 5 and Table S5.6). Head-to-head comparisons in the league table further consolidated the advantage of DPT, which was significantly superior to both CS (SMD −1.43, 95% CI −2.40 to −0.46) and HA (SMD −1.31, 95% CI −2.55 to −0.07) (Appendix 9 and Table S9.5). No statistically significant differences were observed among the remaining pairwise comparisons.

Mid-Term Grip Strength

Analysis of 14 studies (n=805) showed a highly connected structure (Appendix 10, Figure S10.3). According to the forest plot (Appendix 10, Figure S10.3) and Table 1 (Figure S8.6 and Table S8.6), AB (SUCRA 94.4%) and DPT (SUCRA 78.6%) ranked highest, though neither statistically exceeded PL. BT remained the lowest-ranked (SUCRA 12.3%) with a persistent negative trend (MD −6.97, 95% CI −15.45 to 1.51). However, it is important to note that no intervention demonstrated a statistically significant difference compared with placebo in the forest plot analysis, as all 95% confidence intervals crossed the null line. CINeMA assessment showed that the comparison between AB and BT was supported by “Moderate” quality evidence (Appendix 5 and Table S5.7). League table analysis revealed that AB was significantly superior to CS (MD 12.50, 95% CI 3.28 to 21.72), HA (MD 12.31, 95% CI 1.09 to 23.53), and BT (MD 15.97, 95% CI 4.43 to 27.51) (Appendix 9 and Table S9.6). No statistically significant differences were observed among the remaining pairwise comparisons.

Long-Term Pain Intensity

Analysis of 16 studies (n=1,254) demonstrated dense direct evidence among CS, PL, and PRP (Appendix 10, Figure S10.4). Forest plots revealed a marked divergence in outcomes: BT was the sole intervention significantly superior to PL (MD −1.16, 95% CI −2.02 to −0.30) and ranked first (SUCRA 97.8%) (Table 1; Figure S8.7 and Table S8.7). Conversely, CS exhibited a significant rebound effect, with pain scores exceeding PL (MD 1.06, 95% CI 0.56 to 1.57) and ranking lowest (SUCRA 2.1%). CINeMA assessment showed that evidence certainty was “Low” for BT vs. PL, “High” for PRP/AB/DN vs. PL, and “Moderate” for CS vs. PL (Appendix 5 and Table S5.8). League table analysis indicated BT was significantly superior to CS (MD −2.23, 95% CI −3.22 to −1.23) and AB (MD −1.16, 95% CI −2.26 to −0.07). CS was significantly inferior to PRP (MD 1.23, 95% CI 0.89 to 1.56), AB (MD 1.06, 95% CI 0.47 to 1.66), and DN (MD 1.29, 95% CI 0.71 to 1.88). Additionally, DN demonstrated statistical superiority over DPT (MD −0.92, 95% CI −1.65 to −0.19) (Appendix 9 and Table S9.7).

Long-term functional and grip strength analyses were not performed due to data scarcity.

Sensitivity Analysis

Sensitivity analysis excluding studies with a high risk of bias confirmed the robustness of the primary findings (Appendix 11 and Tables S11.1–S11.3). In the short-term, CS maintained the highest SUCRA ranking (85.5%) and remained the only intervention significantly superior to PL (MD −1.58, 95% CI −2.50 to −0.65). For mid-term outcomes, DPT retained both its dominant ranking (SUCRA 88.4%) and statistical significance, whereas PRP and AB lost significance within the high-quality subset. In the long-term, BT consistently ranked first (SUCRA 92.7%) with sustained significance (MD −0.93). Notably, DN achieved a higher ranking (SUCRA 84.7%) but did not reach statistical significance. Overall, these analyses validated the time-dependent efficacy profiles of CS, DPT, and BT.

Safety Analysis

Safety outcomes were qualitatively synthesized from the included studies (Appendix 12 and Table S12.1). Overall, no severe or life-threatening adverse events were reported. BT was uniquely associated with functional motor deficits, including transient digit extensor weakness, lag, and paresis,^38,50,58 which corroborates the reduced grip strength observed in the efficacy analysis. CS injections were frequently linked to local dermatological complications, specifically skin atrophy and hypopigmentation,^48,63 alongside reports of post-injection pain flare and higher recurrence rates.^54,59 Regarding regenerative therapies, PRP and AB were primarily characterized by post-injection pain and local discomfort, which were often reported as significantly higher compared to CS.^53,66 Other interventions, including DPT, HA, and DN, demonstrated favorable safety profiles, with adverse events limited to minor, transient injection site pain.

Discussion

Temporal Efficacy and Pathophysiological Mechanisms

This study synthesized evidence across multiple timeframes, revealing distinct time-dependent efficacy profiles for percutaneous treatments in lateral epicondylitis (LE). Corticosteroids (CS) provided superior pain relief predominantly in the short term (less than four weeks), but this efficacy diminished rapidly, often accompanied by a notable risk of symptom rebound in the mid- and long-term periods. This pharmacological action fundamentally mismatches the chronic pathology of LE, which histopathological evidence confirms is a degenerative “angiofibroblastic tendinosis” rather than an acute inflammatory “tendinitis.”⁶⁹ Mechanistically, while suppressing inflammation, CS also exerts deleterious anti-proliferative effects, including the inhibition of tenocyte viability and the suppression of collagen synthesis, thereby compromising the biomechanical integrity of the tendon.^69,70

In contrast, regenerative therapies (PRP, AB, DN, DPT) demonstrated a slower onset but sustained clinical benefits exceeding 12 weeks.^14,71 This delayed but robust efficacy corroborates the time required for biological tissue repair and remodeling, supporting a paradigm shift in treatment strategy from symptomatic suppression to structural restoration.

Mechanistic Insights into Regenerative Interventions

The sustained superiority of PRP, AB, DN, and DPT aligns with their biological potential to reactivate the stalled healing process in chronic tendinosis. Biological agents such as PRP and AB are theorized to initiate a reparative cascade through the degranulation of alpha-granules, releasing supraphysiological concentrations of growth factors including PDGF, TGF-β, and VEGF.^14,72 These mediators stimulate chemotaxis, tenocyte proliferation, and angiogenesis, thereby facilitating the synthesis of extracellular matrix and structural collagen repair.¹⁴ However, the clinical consistency of PRP is limited by heterogeneity in preparation protocols, and its statistical significance appeared sensitive to risk of bias exclusions.⁷³

Similarly, physical and chemical modalities act as catalysts for tissue remodeling. DN is proposed to disrupt chronic fibrotic tissue and calcifications through mechanical stimulation.⁷¹ Furthermore, it may improve local microcirculation and induce local twitch responses (LTR), which are thought to modulate pain transmission.^71,74 DPT utilizes a hyperosmolar dextrose solution to induce an osmotic gradient, triggering a controlled pro-inflammatory response that stimulates the influx of macrophages to restart the dormant healing cycle.^75,76 While our findings suggest DPT and DN are promising options, clinical recommendations should be tempered in accordance with the certainty of current evidence. Notable uncertainties remain regarding their optimal application in clinical practice. Both modalities exhibit substantial protocol heterogeneity across studies, including variations in needle gauge, manipulation techniques, and duration for DN, as well as differing dextrose concentrations, injection volumes, and frequencies for DPT. Additionally, while the material costs for DPT and DN are relatively low, their widespread availability and generalizability in primary care settings are constrained by the necessity for specialized practitioner training and operator experience.

The discussion of physical interventions in tendinopathy is further strengthened by recent evidence on percutaneous electrolysis. As an emerging regenerative and minimally invasive option, percutaneous electrolysis delivers a direct galvanic current through an acupuncture needle to induce a localized, controlled inflammatory response. Recent literature highlights its clinical utility, particularly when combined with eccentric exercise and stretching.⁷⁷ This reinforces the relevance of biologically oriented, multimodal approaches in the comprehensive management of lateral epicondylitis.

Safety Profiles and Functional Implications

Regarding safety, BT presents distinct challenges despite its efficacy in reducing resting pain.¹² Its mechanism of chemical denervation inevitably leads to dose-dependent extensor muscle weakness and reduced grip strength.⁷⁸ Given these functional impairment risks, the risk-benefit profile of BT is less favorable compared with other interventions, particularly for patients requiring manual dexterity.⁶⁵ In contrast, HA demonstrated a favorable safety profile with dual benefits of lubrication and analgesia, offering a viable alternative for patients intolerant to CS or blood-derived products.^9,75

Comparison with Previous Evidence

Our findings partially align with the Cochrane review by Karjalainen et al,⁷⁹ which expressed skepticism regarding the clinical utility of AB and PRP due to insufficient evidence for their superiority over placebo Consistent with this view, PRP in our analysis did not demonstrate statistical superiority over DPT. However, our study advances current knowledge by incorporating recent evidence on DPT and DN not fully covered in the Cochrane review. This inclusion fills a critical gap by establishing the mid-term dominance of DPT and the long-term functional benefits of DN. Recent specialized reviews^80,81 have further enriched this field by providing in-depth evaluations of individual modalities like dry needling. Building on these insights, our NMA extends the current understanding by positioning these interventions within a broader hierarchy of eight major therapies and integrating objective grip strength to complement subjective scales. Regarding BT, although we corroborate the analgesic efficacy reported by Dong et al,⁸² our clinical interpretation adopts a more cautious stance While previous analyses have often prioritized pain reduction as the primary measure of success, our study highlights the potential trade-offs, specifically the negative trend in grip strength associated with BT. Consequently, a more conservative application of BT might be considered, ensuring that potential functional loss is carefully weighed against analgesic benefits. Furthermore, the inclusion of recent RCTs such as those by Uygur et al^28,63 supports the role of DN as a non-pharmacological strategy for long-term functional recovery, a finding that our analysis further characterizes within the hierarchy of major percutaneous options.

Limitations

Several limitations of this study should be acknowledged. First, substantial clinical heterogeneity exists among the included trials, particularly regarding leukocyte content in PRP preparation, DN techniques, and CS dosages.⁷³ Furthermore, although we originally aimed to investigate the impact of injection guidance techniques, a planned subgroup analysis comparing ultrasound-guided versus palpation-guided administration was precluded by the insufficient number of trials and inconsistent reporting methods. Second, subjective outcomes including pain and function scores are susceptible to placebo effects, introducing potential bias in non-blinded comparisons.⁷⁶ Finally, data scarcity imposed specific constraints on our long-term and secondary analyses. The limited number of studies reporting functional status and grip strength beyond six months prevented the execution of a network meta-analysis for these outcomes, restricting our conclusions primarily to pain intensity in the long term. Similarly, the scarcity of high-quality RCTs with follow-ups exceeding one year limits the precision of long-term recurrence estimates.

Conclusion

The optimal management of LE depends on the disease stage. Corticosteroids may be most appropriate for the short-term control of acute and severe pain. For chronic and recalcitrant cases, DPT emerges as a viable first-line consideration given its favorable efficacy, safety, and cost-effectiveness, while PRP and DN serve as valuable options for long-term management. HA offers a reasonable alternative for patients unsuitable for these modalities. Conversely, BT warrants cautious application due to potential functional impairment risks. Future research should prioritize the standardization of PRP and DN protocols.