Minimally invasive endoscopic treatment of cervical radiculopathy: a comparative study of three surgical channel approaches

Abstract

Objective

To compare the clinical efficacy, safety, and surgery-related outcomes of percutaneous endoscopic surgery (PE), unilateral biportal endoscopic surgery (UBE), and large-channel endoscopic surgery (Delta) for the treatment of cervical spondylotic radiculopathy (CSR) using a network meta-analysis.

Methods

PubMed, Embase, Web of Science, the Cochrane Library, China National Knowledge Infrastructure (CNKI), Wanfang Data, and the Chinese Biomedical Literature Database (CBM) were systematically searched from database inception to March 15, 2026. Comparative studies involving patients with CSR treated with PE, UBE, and Delta were included. The primary outcomes were operative time, Visual Analog Scale score for neck pain (VAS-neck), Visual Analog Scale score for arm pain (VAS-arm), Neck Disability Index (NDI), and incidence of complications. A random-effects network meta-analysis was performed within a frequentist framework using Stata 16.0 by integrating both direct and indirect evidence. Continuous variables were expressed as mean differences (MDs) with 95% confidence intervals (CIs), and dichotomous variables were expressed as odds ratios (ORs). Consistency was assessed using the node-splitting method, and the relative effects of different surgical techniques were ranked using the surface under the cumulative ranking curve (SUCRA). Risk of bias and certainty of evidence were evaluated using the ROBINS-I tool and the GRADE approach, respectively.

Results

A total of 12 retrospective studies involving 939 patients were included. The network meta-analysis showed that Delta was associated with a shorter operative time than both PE (MD = 26.14 min, 95% CI11.81–40.47) and UBE (MD = 16.09 min, 95% CI1.98–30.21), while PE had a longer operative time than UBE. No statistically significant differences were observed among the three techniques in neck pain (VAS-neck), cervical function (NDI), or incidence of complications. No obvious statistically significant difference was observed between Delta and PE in arm pain (VAS-arm); however, the effect size was small and its clinical significance was limited. The SUCRA rankings indicated that Delta ranked first for operative time, whereas the rankings for all outcomes reflected only relative probabilities.

Conclusions

Based on the currently available retrospective evidence, the three posterior endoscopic techniques, PE, UBE, and Delta, showed generally comparable performance in pain relief, functional improvement, and safety for the treatment of CSR. Delta was associated with a shorter operative time; however, this finding was mainly derived from retrospective studies and remains subject to potential bias and uncertainty. Therefore, current evidence is insufficient to support a definite overall advantage of Delta. High-quality, multicenter, prospective randomized controlled trials are still needed to further verify the relative effects of these different surgical techniques.

Introduction

Cervical spondylotic radiculopathy (CSR) is a frequently encountered cervical degenerative disease characterized by radicular symptoms. Its pathogenesis is primarily related to mechanical compression or inflammatory irritation of the cervical nerve roots resulting from intervertebral disc degeneration, facet joint degeneration, hypertrophy of the ligamentum flavum, and osteophyte formation [1]. Clinically, patients typically present with radiating pain in the neck, shoulder, and upper extremities, often accompanied by numbness, sensory disturbances, and decreased muscle strength, which can substantially impair quality of life and work capacity [2, 3]. Epidemiological studies have demonstrated marked age- and sex-related differences in the incidence of CSR. The annual incidence has been reported to be approximately 83.2 per 100,000 individuals, with a higher prevalence in men and a peak incidence between 50 and 54 years of age [2]. The prevalence of CSR in the general population is estimated at approximately 3.5 per 1000 individuals, indicating a considerable clinical and public health burden [4]. Conservative treatment is generally recommended for patients with mild symptoms; however, when standardized conservative management fails after at least 12 weeks or symptoms markedly compromise quality of life, surgical intervention becomes an important treatment option [5].

Anterior cervical discectomy and fusion (ACDF) remains a widely accepted standard surgical option for patients with CSR. Nevertheless, ACDF may be associated with complications such as dysphagia, loss of intervertebral height, and accelerated adjacent segment degeneration, which may limit its long-term benefits in certain patients [6]. With advances in minimally invasive surgical concepts and techniques, posterior endoscopic decompression has emerged as an important alternative approach for the management of CSR.Currently, posterior endoscopic techniques used for CSR mainly include percutaneous endoscopic surgery (PE) [7], unilateral biportal endoscopic surgery (UBE) [8], and large-channel endoscopic systems that have been increasingly adopted in recent years, such as the Delta system [9]. All three techniques are performed under continuous saline irrigation and aim to achieve effective nerve root decompression while preserving cervical spinal stability. However, these techniques differ in surgical approach, visualization and decompression efficiency, extent of tissue trauma, and learning curve. Previous studies have suggested that PE is minimally invasive but limited by a relatively narrow working space [10]; UBE provides greater operative flexibility and a wider decompression range [11]; and large-channel endoscopic systems, owing to their larger working channel diameter, may be associated with improved surgical efficiency [9]. Despite these reported advantages, comparative evidence regarding the clinical efficacy, safety, and surgical efficiency of these three posterior endoscopic techniques remains inconclusive. Most available studies have focused on pairwise comparisons with inconsistent findings, while comprehensive evaluations integrating both direct and indirect evidence are still lacking [12]. In addition, conventional pairwise meta-analyses are limited in their ability to compare multiple competing interventions within a single analytical framework.

Therefore, the present study aimed to conduct a network meta-analysis to comprehensively compare the clinical efficacy, safety, and surgery-related outcomes—including operative time, postoperative pain scores, improvement in cervical function, and the incidence of complications—of PE, UBE, and Delta techniques in the treatment of CSR. By synthesizing both direct and indirect evidence and ranking treatment effects across different outcomes, this study seeks to provide more comprehensive evidence to support the rational selection of posterior endoscopic surgical approaches in clinical practice.

Methods

Study design

This study was designed as a systematic review and network meta-analysis to comprehensively compare the efficacy and safety of percutaneous endoscopic surgery (PE), unilateral biportal endoscopic surgery (UBE), and large-channel endoscopic surgery (Delta) in the treatment of cervical spondylotic radiculopathy. The design and reporting of this study strictly followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Network Meta-Analyses (PRISMA-NMA).The study protocol was not prospectively registered in PROSPERO or other public registries prior to data extraction and analysis.

Literature search strategy

A systematic literature search was conducted in PubMed, Embase, Web of Science, the Cochrane Library, China National Knowledge Infrastructure (CNKI), Wanfang Data, and the Chinese Biomedical Literature Database (CBM). The search covered the period from database inception to March 15, 2026.

English search terms consisted of combinations of disease-related and endoscopic technique–related keywords. Disease-related terms included “cervical radiculopathy,” “cervical disc herniation,” and “cervical foraminal stenosis.” Endoscopic technique–related terms included “endoscopic cervical surgery,” “percutaneous endoscopic cervical discectomy,” “endoscopic posterior cervical foraminotomy,” “biportal endoscopic surgery,” “unilateral biportal endoscopy,” “large-channel endoscopy,” and “Delta system.”

The search strategy was adapted as appropriate for each database. In addition, the reference lists of all included studies were manually screened to identify potentially relevant publications and minimize the risk of missing eligible studies. Detailed search strategies for all databases are provided in the Supplementary Appendix. Using PubMed as an example, the search strategy combined Medical Subject Headings (MeSH) terms and free-text terms related to the disease and surgical techniques using Boolean operators to ensure systematic and reproducible retrieval.

Inclusion and exclusion criteria

Inclusion criteria: 1. Participants were diagnosed with cervical spondylotic radiculopathy based on both clinical symptoms and radiological evidence; 2. Interventions included PE, UBE, or Delta endoscopic techniques; 3. At least one outcome measure was reported, including operative time, pain scores (VAS), Neck Disability Index (NDI), or complications; 4. Study designs were randomized controlled trials or observational studies.

Exclusion criteria: 1. Non-comparative studies, case reports, reviews, conference abstracts, or animal studies; 2. studies involving patients with concomitant cervical myelopathy, infection, tumor, or severe spinal deformity; 3. studies with incomplete data or from which valid data could not be extracted; and 4. duplicate publications or unpublished articles.

Outcome measures

The primary outcomes of interest included: 1. Operative time. Operative time was defined according to the original reports of the included studies. Most studies defined operative time as the duration from skin incision to skin closure, although some studies did not clearly specify the start and end points. Given the potential variability in operative time definitions and surgical workflow across studies, a random-effects model was applied in the pooled analysis to reduce the potential impact of such differences. 2. Visual Analog Scale score for neck pain (VAS-neck); 3. Visual Analog Scale score for arm pain (VAS-arm); 4. Neck Disability Index (NDI);5.Incidence of complications.For pain- and function-related outcomes (VAS-neck, VAS-arm, and NDI), data from the final follow-up were preferentially extracted. When final follow-up data were unavailable, outcomes at the longest reported follow-up time point were used. Because follow-up durations varied among the included studies, a random-effects model was employed to minimize the potential influence of differences in follow-up timing on the pooled effect estimates.

Study selection and data extraction

Two reviewers independently screened eligible studies and extracted relevant data. Any disagreements were resolved through discussion, and if consensus could not be reached, a third reviewer was consulted.The following data were extracted from each included study: first author and year of publication; study design; study population; surgical techniques (intervention and comparator); sample size (intervention/comparator); patient age (years, intervention/comparator); sex distribution (male/female, intervention/comparator); baseline Visual Analog Scale scores for neck pain (VAS-neck) and arm pain (VAS-arm); baseline Neck Disability Index (NDI); follow-up duration (months); operated cervical level(s); disease type; and reported primary outcomes.

Quality assessment

The methodological quality of the included observational studies was systematically assessed using the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool, as recommended by the Cochrane Collaboration. The ROBINS-I tool evaluates risk of bias across seven domains [13]: bias due to confounding, bias in the selection of participants, bias in classification of interventions, bias due to deviations from intended interventions, bias due to missing data, bias in measurement of outcomes, and bias in selection of the reported result.Given that all included studies were retrospective observational in design, an inherent risk of bias—particularly related to confounding and participant selection—could not be fully eliminated. Therefore, the results of this study were interpreted with methodological caution, and the included studies should not be considered equivalent to randomized controlled trials or assumed to be at low risk of bias.

Statistical analysis

Network meta-analysis was performed using a frequentist framework in Stata software (version 16.0; StataCorp, College Station, TX, USA) with the built-in network commands. The treatment network was constructed using the network setup command, and a random-effects network meta-analysis was conducted with the network meta command to synthesize direct and indirect evidence [14]. A multivariate random-effects meta-analytic model was applied, with between-study variance estimated at the network level using the mvmeta command and restricted maximum likelihood (REML).Continuous outcomes were summarized as mean differences (MDs) with 95% confidence intervals (CIs), and dichotomous outcomes were expressed as odds ratios (ORs) and reported as log ORs with 95% CIs. Because all continuous outcomes were measured on identical scales, standardized mean differences were not used. Between-study heterogeneity was assessed using both the between-study variance (τ2) and the I2 statistic, with I2 values of approximately 25%, 50%, and 75% indicating low, moderate, and high heterogeneity, respectively. Network consistency was evaluated using the node-splitting method (network sidesplit). Treatment rankings were generated using the network rank command, and the surface under the cumulative ranking curve (SUCRA) was calculated.

Comparison-adjusted funnel plots were used to assess potential publication bias and small-study effects, and Egger’s regression test was performed for outcomes including at least 10 studies. All statistical tests were two-sided, with a P value < 0.05 considered statistically significant where applicable.The transitivity assumption was considered acceptable, as included studies were generally comparable in terms of patient populations, surgical indications (disc herniation or foraminal stenosis), predominantly single-level procedures, perioperative management, and outcome assessment. Although surgeon experience and follow-up duration varied across studies, these factors were not systematically imbalanced across comparisons. Nevertheless, transitivity cannot be directly tested and represents a methodological limitation.

The certainty of evidence for major outcomes (VAS scores, NDI, and complications) was evaluated using the GRADE approach. Sensitivity analyses were conducted by excluding studies with a serious risk of bias according to the ROBINS-I assessment. Meta-regression, trial sequential analysis, and post hoc power analysis were not performed because of the limited number of studies, incomplete reporting of key covariates, and the predominantly retrospective observational nature of the included evidence.Network plots were generated to visualize the evidence structure, with node sizes proportional to sample size and edge thickness proportional to the number of direct comparisons.

Results

Study selection process

The database search conducted up to March 15, 2026 identified 1684 potentially relevant records. After 758 duplicate records were removed, 926 studies remained for title and abstract screening, of which 789 were excluded because they clearly did not meet the inclusion criteria. The full texts of the remaining 137 articles were then assessed for eligibility, and 125 studies were further excluded, mainly because of ineligible study designs, mixed reporting of different endoscopic techniques without extractable separate comparative data, or incomplete outcome data. Ultimately, 12 studies met the inclusion criteria and were included in this network meta-analysis. The detailed study selection process and the number of records retrieved from each database are shown in Fig. 1, with additional information provided in Supplementary Table S1.

Fig. 1.

Flow diagram of the literature search and study selection process. The number of records retrieved from each database and the reasons for full-text exclusion are shown

Characteristics of included studies and methodological quality assessment

The basic characteristics of the studies included in this network meta-analysis are presented in Table 1. A total of 12 studies involving 939 patients were included. The study populations mainly consisted of patients with cervical spondylotic radiculopathy (CSR) or cervical radiculopathy (CR). The surgical techniques compared across studies included percutaneous endoscopic surgery (PE), unilateral biportal endoscopic surgery (UBE), and the large-channel endoscopic technique (Delta). All included studies were pairwise comparative studies, and no multi-arm study directly comparing all three techniques was identified.

Table 1.

Demographic, clinical, and surgical characteristics of the included studies

First author (year)	Study design	Study population	Surgical technique (intervention / control)	Sample size (intervention / control)	Age (years, intervention / control)	Sex (male/female, intervention / control)	Baseline VAS-neck	Baseline VAS-arm	Baseline NDI	Follow-up duration (months)	Operated level(s)	Disease type	Primary outcomes
Xie Guisi (2024) [15]	Retrospective	CSR	UBE / PE	23 / 21	44.4 ± 12.2 / 41.9 ± 13.8	13/10 / 13/8	3.4 ± 1.2 / 3.7 ± 0.6	4.4 ± 0.7 / 4.0 ± 0.5	23.3 ± 4.9 / 23.8 ± 4.7	3–12 (median 6.9)	C4–C7; single-level	Disc herniation / foraminal stenosis	1,2,3,4,5
Cai Zijun (2025) [16]	Retrospective	CSR	Delta / PE	32 / 35	66.37 ± 7.98 / 64.10 ± 6.22	18/14 / 18/17	NR	NR	42.04 ± 3.27 / 42.18 ± 3.56	12	C4–7; single-level (predominantly C5–6)	Disc herniation / foraminal stenosis	1,2,3,4,5
Zhang (2025) [17]	Retrospective	CSR	PE / UBE	40 / 30	52.5 ± 8.9 / 49.9 ± 7.8	17/23 / 14/16	NR	NR	NR	24	Single-level (predominantly C5–6)	Disc herniation / foraminal stenosis	1,2,3,4,5
Kang (2022) [18]	Retrospective	CR	PE / UBE	32 / 33	53.74 ± 8.50 / 52.68 ± 9.56	10/22 / 11/22	6.18 ± 0.96 / 6.21 ± 0.79	7.84 ± 0.66 / 7.92 ± 0.61	56.48 ± 6.03 / 57.31 ± 5.87	12	C4–T1; single-level	Foraminal disc herniation / stenosis	1,2,5
Wang (2023) [19]	Retrospective	CSR	UBE / PE	89 / 65	58.28 ± 11.94 / 60.10 ± 9.69	42/47 / 28/37	NR	NR	NR	NR	NR	NR	1,2,3,4,5
Kim (2022) [20]	Retrospective	CR	PE / UBE	38 / 30	55.3 ± 9.1 / 52.3 ± 12.0	27/11 / 21/9	6.5 ± 1.1 / 4.3 ± 1.6	7.4 ± 0.8 / 7.6 ± 0.7	22.3 ± 3.9 / 26.2 ± 2.5	10.0 ± 3.9 / 11.7 ± 6.4	C4–T1; single-level (predominantly C5–6)	Foraminal stenosis–related CR	1,2,3,4,5
Zhao Chendong (2023) [21]	Retrospective	CSR	Delta / UBE	25 / 25	NR	NR	NR	NR	30.80 ± 5.34 / 30.12 ± 5.36	≥ 12	C4–7; single-level (predominantly C5–6)	Foraminal stenosis–related CR	1,2,3,4,5
Yu Xiankai (2023) [22]	Retrospective	CSR	UBE / PE	32 / 30	53.37 ± 7.16 / 53.70 ± 6.35	13/19 / 15/15	NR	NR	37.78 ± 2.43 / 38.33 ± 2.35	17.30 ± 2.73	C5–6 or C6–7; single-level	Disc herniation–related CSR	1,4,5
Yan Meijun (2023) [23]	Retrospective	CSR	PE / UBE	21 / 19	43.19 ± 9.83 / 45.05 ± 10.73	11/10 / 8/11	NR	NR	NR	6	C4–7; single-level (predominantly C5–6)	Disc herniation / foraminal stenosis	1,2,3,4
Zhong (2023) [24]	Retrospective	CR	UBE / PE	36/ 33	NR	20/16 / 18/15	5.94 ± 0.83 / 6.36 ± 0.70	6.81 ± 0.86 / 6.64 ± 0.96	34.56 ± 3.42 / 34.15 ± 2.31	NR	C3–T1; single-level (predominantly C5–6, C6–7)	Disc herniation / foraminal stenosis	1,2,3,4,5
Tang (2024) [25]	Retrospective	CSR	UBE / PE	71 / 89	60.23 ± 11.34 / 57.12 ± 10.87	41/30 / 54/35	8.01 ± 0.35 / 7.99 ± 0.51	6.59 ± 1.01 / 6.92 ± 0.91	41.85 ± 5.12 / 38.47 ± 4.68	18.12 ± 5.78 / 19.24 ± 6.11	C4–7; single-level	Disc herniation / foraminal stenosis	1,2,3,4,5
Huang Pengbo (2026) [26]	Retrospective study	Cervical spondylotic radiculopathy (CSR)	PE/ UBE	47 / 43	52.71 ± 8.65 / 51.63 ± 9.24	18/29 / 15/28	5.39 ± 1.26 / 5.46 ± 1.17	7.52 ± 0.45 / 7.59 ± 0.61	40.74 ± 4.23 / 39.58 ± 4.57	12 months	C4–5, C5–6, and C6–7; single-level	CSR associated with unilateral foraminal stenosis combined with cervical disc herniation	1.2.3.4.5

CSR, cervical spondylotic radiculopathy; CR, cervical radiculopathy; NR, not reported. Primary outcomes: 1, operative time; 2, VAS-neck; 3, VAS-arm; 4, NDI; 5, complications

Sample sizes varied substantially across studies, with total sample sizes ranging from 40 to 160 patients. The mean age of participants differed among studies, but most patients were between 40 and 60 years of age. Sex distribution was reported in detail in some studies, whereas others did not provide separate sex data for each intervention group. Reporting of baseline pain and functional status, including Visual Analog Scale score for neck pain (VAS-neck), Visual Analog Scale score for arm pain (VAS-arm), and Neck Disability Index (NDI), was inconsistent across studies, and some studies did not provide baseline data, which limited the assessment of baseline comparability between intervention groups. Follow-up duration also varied considerably, ranging from 3 to 24 months or longer, and was reported in different formats, including ranges, medians, mean ± standard deviation, or minimum follow-up time. Most studies involved single-level procedures, with C5–6 and C6–7 being the most commonly treated levels. The underlying pathology mainly included cervical disc herniation and foraminal stenosis, although some studies did not provide detailed descriptions of the specific pathological type or operated level.

Methodological quality was assessed using the Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool (Table 2). Most included studies were judged to have an overall serious risk of bias, mainly because of confounding inherent to retrospective, non-randomized study designs; only one study was rated as having a moderate overall risk of bias. Across the individual bias domains, bias in classification of interventions and bias due to deviations from intended interventions were generally low, whereas bias due to confounding, bias in the selection of participants, and bias in measurement of outcomes were more commonly observed.

Table 2.

Summary of ROBINS-I Risk of Bias Assessment for Included Studies

Study	Bias due to Confounding	Bias in Selection of Participants	Bias in Classification of Interventions	Bias due to Deviations from Intended Interventions	Bias due to Missing Data	Bias in Measurement of Outcomes	Bias in Selection of the Reported Result	Overall Risk of Bias
Xie Guisi [15]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Cai Zijun [16]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Zhang 17	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Kang 18	Moderate	Moderate	Low	Low	Low	Moderate	Moderate	Moderate
Tang [25]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Wang [19]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Kim [20]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Zhong [24}	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Zhao [21]	Serious	Moderate	Low	Low	Moderate	Moderate	Moderate	Serious
Yu [22]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Yan [23]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious
Huang Pengbo [26]	Serious	Moderate	Low	Low	Low	Moderate	Moderate	Serious

The certainty of evidence was evaluated using the GRADE approach (Supplementary Table S2). For operative time (12 studies), the certainty of evidence was rated as low because of the retrospective study design, serious risk of bias, and indirectness. For VAS-neck (11 studies), VAS-arm (10 studies), and NDI (11 studies), the certainty of evidence was rated as very low because of serious risk of bias and indirectness, despite some degree of precision. For complications (11 studies), the certainty of evidence was also rated as very low because of serious risk of bias, indirectness, and imprecision. Overall, the current evidence remains highly uncertain regarding whether any of the three endoscopic techniques is superior across the evaluated outcomes, highlighting the need for high-quality prospective studies for further confirmation.

Network meta-analysis results for the five outcomes

Operative time

A total of 12 retrospective observational studies were included in the network meta-analysis of operative time among the three posterior endoscopic techniques, namely the Delta technique, percutaneous endoscopic surgery (PE), and unilateral biportal endoscopic surgery (UBE) (Fig. 2). The network meta-analysis showed that operative time differed among the three techniques. Compared with Delta, PE was associated with a longer operative time (MD = 26.14 min, 95% CI11.81–40.47; P < 0.001), and UBE was also associated with a longer operative time (MD = 16.09 min, 95% CI1.98–30.21; P = 0.025). In addition, PE had a longer operative time than UBE (MD = 10.05 min, 95% CI4.39–15.71; P < 0.01) (Table 3).

Fig. 2.

Network plot of the network meta-analysis for operative time among three endoscopic techniques. legend: In the network plot, nodes represent different surgical techniques, with node size proportional to the cumulative sample size. Lines represent direct comparisons between techniques, and line thickness is proportional to the number of included studies. All networks in the present study were constructed exclusively from pairwise comparisons, and no multi-arm (≥ 3-arm) studies were included

Table 3.

League table of network meta-analysis for operative time among three endoscopic techniques

	Delta	UBE	PE
Delta	–	16.09 (1.98, 30.21), p = 0.025	26.14 (11.81, 40.47), p < 0.001
UBE	− 16.09 (− 30.21, − 1.98), p = 0.025	–	10.05 (4.39, 15.71), p < 0.001
PE	− 26.14 (− 40.47, − 11.81), p < 0.001	− 10.05 (− 15.71, − 4.39), p < 0.001	–

Statistical significance was defined as a two-sided P value < 0.05

According to the surface under the cumulative ranking curve (SUCRA) results, Delta had the highest probability of being ranked best for shorter operative time (SUCRA = 99.4%), followed by UBE (SUCRA = 50.6%), whereas PE ranked lowest (SUCRA = 0.0%) (Supplementary Table S3; Fig. 3). It should be noted that the SUCRA results represent relative ranking probabilities derived from the synthesis of direct and indirect evidence and should not be interpreted as indicating a definitive superiority or inferiority of one technique over another.

Fig. 3.

SUCRA ranking plot for operative time among three endoscopic techniques

Between-study heterogeneity for operative time was moderate (τ2 ≈ 8.74) and was considered acceptable under the random-effects model. Node-splitting analysis showed no significant inconsistency between direct and indirect evidence (all P > 0.05) (Tables 4 and 5). The comparison-adjusted funnel plot did not suggest obvious publication bias (Fig. 4). In Egger’s regression analysis, the slope was statistically significant (P = 0.017), whereas the intercept was not statistically significant (P = 0.129), suggesting the possibility of a small-study effect. Therefore, conclusions related to operative time should be interpreted with caution (Table 4).

Table 4.

Egger’s regression test for operative time

Variable	Coefficient (Coef.)	Standard Error (Std. Err.)	t value	P value	95% Confidence Interval
Slope	− 19.83	6.90	− 2.88	0.017	− 35.20 to − 4.47
Intercept (bias)	4.43	2.68	1.65	0.129	− 1.54 to 10.39

Table 5.

Consistency test results of network meta-analysis for operative time (node-splitting)

Comparison	Direct evidence	Indirect evidence	Difference	P value
Delta vs PE	25.20	26.82	− 1.62	0.916
Delta vs UBE	16.96	14.73	2.23	0.884
PE vs UBE	− 10.13	− 8.19	− 1.94	0.900

Consistency was assessed using the node-splitting method. All P values were greater than 0.05, indicating no significant inconsistency within the network

Fig. 4.

Comparison-adjusted funnel plot for operative time among three endoscopic techniques. Legend: A = Delta; B = PE; C = UBE

Neck pain (VAS-neck)

A total of 11 retrospective observational studies were included in the network meta-analysis evaluating the effects of the Delta technique, percutaneous endoscopic surgery (PE), and unilateral biportal endoscopic surgery (UBE) on improvement in neck pain, as measured by the Visual Analog Scale score for neck pain (VAS-neck) (Fig. 5). The network meta-analysis showed no statistically significant differences among the three techniques in terms of neck pain improvement. Compared with Delta, PE showed no statistically significant difference in VAS-neck improvement (MD = 0.19, 95% CI− 0.11 to 0.50; P = 0.218), and UBE likewise showed no statistically significant difference compared with Delta (MD = 0.16, 95% CI− 0.15 to 0.46; P = 0.310). Similarly, no significant difference was observed between UBE and PE (MD = 0.03, 95% CI− 0.10 to 0.17; P > 0.05) (Tables 5 and 6).

Fig. 5.

Network plot of the network meta-analysis for neck pain (VAS-neck) among three endoscopic techniques. Legend: In the network plot, nodes represent different surgical techniques, and node size is proportional to the cumulative sample size. Lines indicate direct comparisons between techniques, with line thickness proportional to the number of included studies. The network was constructed exclusively from pairwise comparisons, and no multi-arm (≥ 3-arm) studies were included

Table 6.

League table of network meta-analysis for neck pain (VAS-neck) among three endoscopic techniques

	Delta	UBE	PE
Delta	–	0.16 (− 0.15, 0.46), p = 0.310	0.19 (− 0.11, 0.50), p = 0.218
UBE	− 0.16 (− 0.46, 0.15), p = 0.310	–	0.03 (− 0.10, 0.17), p > 0.05
PE	− 0.19 (− 0.50, 0.11), p = 0.218	− 0.03 (− 0.17, 0.10), p > 0.05	–

Data are presented as mean difference (MD) with 95% confidence intervals. Negative MD values indicate that the treatment shown in the column is associated with greater improvement in neck pain than that shown in the row

According to the surface under the cumulative ranking curve (SUCRA) results, Delta had the highest probability of ranking best for neck pain improvement (SUCRA = 86.8%), followed by UBE (SUCRA = 41.8%), whereas PE ranked lowest (SUCRA = 21.5%) (Supplementary Table S4; Fig. 6). Because none of the pairwise comparisons reached statistical significance, these ranking results reflect only the relative probabilities derived from the synthesis of direct and indirect evidence and should not be interpreted as indicating a definitive difference in efficacy among the three techniques.

Fig. 6.

SUCRA ranking plot for neck pain (VAS-neck) among three endoscopic techniques

Between-study heterogeneity for the VAS-neck outcome was low to moderate overall (τ2 ≈ 0.03) and was considered acceptable under the random-effects model. Node-splitting analysis showed no significant inconsistency between direct and indirect evidence, and none of the comparisons reached statistical significance (all P > 0.05) (Tables 7 and 8). The comparison-adjusted funnel plot did not suggest obvious asymmetry (Fig. 7). However, Egger’s regression analysis showed that both the slope (P = 0.031) and the intercept (P = 0.044) were statistically significant, suggesting the presence of a potential small-study effect. Therefore, the results related to VAS-neck should be interpreted with caution (Table 7).

Table 7.

Egger’s regression test for neck pain (VAS-neck)

Variable	Coefficient (Coef.)	Standard Error (Std. Err.)	t value	P value	95% Confidence Interval
Slope	− 0.35	0.14	− 2.56	0.031	− 0.65 to − 0.04
Intercept (bias)	3.04	1.30	2.34	0.044	0.10 to 5.97

Table 8.

Consistency test results of network meta-analysis for neck pain (VAS-neck) (node-splitting)

Comparison	Direct evidence	Indirect evidence	Difference	P value
Delta vs PE	0.26	0.10	0.16	0.631
Delta vs UBE	0.08	0.24	− 0.16	0.631
PE vs UBE	− 0.02	− 0.18	0.16	0.631

Consistency was assessed using the node-splitting method. All P values were greater than 0.05, indicating no significant inconsistency within the network

Fig. 7.

Comparison-adjusted funnel plot for neck pain (VAS-neck) among three endoscopic techniques. Legend: A = Delta; B = PE; C = UBE

Arm pain (VAS-arm)

A total of 10 retrospective observational studies were included in the network meta-analysis evaluating the effects of the Delta technique, percutaneous endoscopic surgery (PE), and unilateral biportal endoscopic surgery (UBE) on improvement in arm pain, as measured by the Visual Analog Scale score for arm pain (VAS-arm) (Fig. 8). The network meta-analysis showed that the differences among the three techniques in improving arm pain were generally small. Compared with Delta, PE was associated with a slightly higher VAS-arm score (MD = 0.20, 95% CI0.02–0.38; P = 0.028), suggesting slightly less improvement in arm pain in the PE group; however, the absolute effect size was small, and its clinical relevance remains uncertain. Compared with Delta, no statistically significant difference was observed for UBE (MD = 0.16, 95% CI− 0.03 to 0.34; P = 0.094). In addition, no statistically significant difference was found between UBE and PE (MD = 0.04, 95% CI− 0.03 to 0.12; P > 0.05) (Table 9).

Fig. 8.

Network plot of the network meta-analysis for arm pain (VAS-arm) among three endoscopic techniques. Legend: In the network plot, nodes represent different surgical techniques, and node size is proportional to the cumulative sample size. Lines represent direct comparisons between techniques, with line thickness proportional to the number of included studies. All networks in the present study were constructed exclusively from pairwise comparisons, and no multi-arm (≥ 3-arm) studies were included

Table 9.

League table of network meta-analysis for arm pain (VAS-arm) among three endoscopic techniques

	Delta	UBE	PE
Delta	–	0.16 (− 0.03, 0.34), p = 0.094	0.20 (0.02, 0.38), p = 0.028
UBE	− 0.16 (− 0.34, 0.03), p = 0.094	–	0.04 (− 0.03, 0.12), p > 0.05
PE	− 0.20 (− 0.38, − 0.02), p = 0.028	− 0.04 (− 0.12, 0.03), p > 0.05	–

Data are presented as mean difference (MD) with 95% confidence intervals. Negative MD values indicate that the treatment shown in the column is associated with greater improvement in arm pain than that shown in the row

According to the surface under the cumulative ranking curve (SUCRA) results, Delta had the highest probability of ranking best for improvement in arm pain (SUCRA = 96.9%), followed by UBE (SUCRA = 46.2%), whereas PE ranked lowest (SUCRA = 6.9%) (Supplementary Table S5; Fig. 9). Because the effect sizes across comparisons were small and most comparisons did not reach statistical significance, these ranking results reflect only relative probabilities and should not be interpreted as indicating a definitive difference in efficacy among the three techniques.

Fig. 9.

SUCRA ranking plot for arm pain (VAS-arm) among three endoscopic techniques

Between-study heterogeneity for the VAS-arm outcome was very low, with τ2 close to 0, indicating good consistency across studies. Node-splitting analysis showed no significant inconsistency between direct and indirect evidence (all P > 0.05) (Tables 10 and 11). The comparison-adjusted funnel plot did not suggest obvious publication bias (Fig. 10). Egger’s regression analysis showed that neither the slope (P = 0.135) nor the intercept (P = 0.120) was statistically significant, indicating no clear evidence of a small-study effect (Table 10).

Table 10.

Egger’s regression test for arm pain (VAS-arm)

Variable	Coef	Std. Err	t	P	95% CI
Slope	− 0.17	0.10	− 1.66	0.135	− 0.41 to 0.07
Intercept	1.58	0.91	1.74	0.120	− 0.52 to 3.67

Table 11.

Consistency test results of network meta-analysis for arm pain (VAS-arm) (node-splitting)

Comparison	Direct	Indirect	Difference	P
Delta vs PE	0.25	0.12	0.13	0.483
Delta vs UBE	0.08	0.21	− 0.13	0.483
PE vs UBE	− 0.04	− 0.17	0.13	0.483

All P > 0.05, indicating no inconsistency

Fig. 10.

Comparison-adjusted funnel plot for arm pain (VAS-arm) among three endoscopic techniques. Legend: A = Delta; B = PE; C = UBE

Neck disability index (NDI)

A total of 11 retrospective observational studies were included in the network meta-analysis evaluating the effects of the Delta technique, percutaneous endoscopic surgery (PE), and unilateral biportal endoscopic surgery (UBE) on improvement in cervical function, as measured by the Neck Disability Index (NDI) (Fig. 11). The network meta-analysis showed no statistically significant differences among the three techniques in terms of NDI improvement, indicating that their effects on functional recovery were generally comparable. Compared with Delta, PE showed no statistically significant difference in NDI improvement (MD = 0.35, 95% CI− 1.04 to 1.74; P = 0.621), and UBE likewise showed no significant difference compared with Delta (MD =  − 0.14, 95% CI− 1.54 to 1.26; P = 0.848). Similarly, no statistically significant difference was observed between UBE and PE (MD =  − 0.49, 95% CI− 1.13 to 0.16; P >  0.05) (Table 12).

Fig. 11.

Network plot of the network meta-analysis for Neck Disability Index (NDI) among three endoscopic techniques. Legend: Nodes represent surgical techniques, with node size proportional to cumulative sample size. Lines denote direct comparisons, and line thickness reflects the number of studies. The network consisted solely of pairwise comparisons, with no multi-arm (≥ 3-arm) studies

Table 12.

Leaguetable of network meta-analysis for neck disability index (NDI) among three endoscopic techniques

	Delta	UBE	PE
Delta	–	− 0.14 (− 1.54, 1.26), p = 0.848	0.35 (− 1.04, 1.74), p = 0.621
UBE	0.14 (− 1.26, 1.54), p = 0.848	–	− 0.49 (− 1.13, 0.16), p > 0.05
PE	− 0.35 (− 1.74, 1.04), p = 0.621	0.49 (− 0.16, 1.13), p > 0.05	–

Data are presented as mean difference (MD) with 95% confidence intervals. Negative MD values indicate that the treatment shown in the column is associated with greater improvement in cervical function than that shown in the row

According to the surface under the cumulative ranking curve (SUCRA) results, PE had the highest probability of ranking best for NDI improvement (SUCRA = 81.2%), followed by Delta (SUCRA = 43.8%), whereas UBE ranked lowest (SUCRA = 25.0%) (Supplementary Table S6; Fig. 12). Because none of the pairwise comparisons reached statistical significance, these ranking results reflect only the relative distribution of treatment effects and should not be interpreted as indicating a definitive difference in efficacy among the three techniques.

Fig. 12.

SUCRA ranking plot for Neck Disability Index (NDI) among three endoscopic techniques

Between-study heterogeneity for the NDI outcome was moderate (τ2 ≈ 0.89), indicating a certain degree of variability across studies, but it was considered acceptable overall under the random-effects model. Node-splitting analysis showed no significant inconsistency between direct and indirect evidence (all P > 0.05) (Tables 13 and 14). The comparison-adjusted funnel plot did not suggest obvious publication bias (Fig. 13). Egger’s regression analysis showed that neither the slope (P = 0.700) nor the intercept (P = 0.297) was statistically significant, indicating no evidence of a small-study effect (Table 13).

Table 13.

Egger’s regression test for neck disability index (NDI)

Variable	Coef	Std. Err	t	P	95% CI
Slope	0.13	0.32	0.40	0.700	− 0.60 to 0.86
Intercept	− 1.17	1.06	− 1.11	0.297	− 3.57 to 1.23

Table 14.

Consistency test results of network meta-analysis for neck disability index (NDI) (node-splitting)

Comparison	Direct	Indirect	Difference	P
Delta vs PE	0.06	0.73	− 0.67	0.660
Delta vs UBE	0.20	− 0.47	0.67	0.660
PE vs UBE	− 0.53	0.14	− 0.67	0.660

P > 0.05, indicating no inconsistency

Fig. 13.

Comparison-adjusted funnel plot for Neck Disability Index (NDI) among three endoscopic techniques. Legend: A = Delta; B = PE; C = UBE

Complications

A total of 11 retrospective observational studies were included in the network meta-analysis evaluating the incidence of complications among the three posterior endoscopic techniques, namely the Delta technique, percutaneous endoscopic surgery (PE), and unilateral biportal endoscopic surgery (UBE) (Fig. 14). The network meta-analysis showed no statistically significant differences among the three techniques in terms of complication rates. Compared with Delta, PE showed no statistically significant difference in complication risk (log OR = 0.82, 95% CI− 1.15 to 2.79; P = 0.413), corresponding to an OR of approximately 2.27. Likewise, no significant difference was observed between UBE and Delta (log OR = 0.80, 95% CI− 1.19 to 2.80; P = 0.430), corresponding to an OR of approximately 2.23. Similarly, no statistically significant difference was found between UBE and PE (log OR = 0.02, 95% CI− 0.60 to 0.63; P > 0.05), corresponding to an OR of approximately 1.02 (Table 15).

Fig. 14.

Network plot of the network meta-analysis for complications among three endoscopic techniques. Legend: In the network plot, nodes represent different surgical techniques, with node size proportional to the cumulative sample size. Lines denote direct comparisons between techniques, and line thickness reflects the number of contributing studies. The evidence network in this study consisted solely of pairwise comparisons, with no multi-arm (≥ 3-arm) trials included

Table 15.

League table of network meta-analysis for complications among three endoscopic techniques

	Delta	UBE	PE
Delta	–	0.80 (− 1.19, 2.80), p = 0.430	0.82 (− 1.15, 2.79), p = 0.413
UBE	− 0.80 (− 2.80, 1.19), p = 0.430	–	0.02 (− 0.60, 0.63), p > 0.05
PE	− 0.82 (− 2.79, 1.15), p = 0.413	− 0.02 (− 0.63, 0.60), p > 0.05	–

Data are presented as log odds ratios (log OR) with 95% confidence intervals. Negative log OR values indicate that the treatment shown in the column is associated with a lower complication risk than the treatment shown in the row. For clinical interpretability, corresponding OR values may be obtained by exponentiating the log OR estimates

According to the surface under the cumulative ranking curve (SUCRA) results, Delta had the highest probability of ranking best for a lower risk of complications (SUCRA = 78.7%), followed by UBE (SUCRA = 36.9%) and PE (SUCRA = 34.4%) (Supplementary Table S7; Fig. 15). Because none of the pairwise comparisons reached statistical significance and the overall number of complication events was low, these ranking results reflect only relative probability distributions and should not be interpreted as indicating a definitive safety advantage of one technique over another.

Fig. 15.

SUCRA ranking plot for complications among three endoscopic techniques

Between-study heterogeneity for the complications outcome was very low (τ2 close to 0), indicating overall consistency across studies. Node-splitting analysis showed no significant inconsistency between direct and indirect evidence, with all P values greater than 0.05 (Tables 16 and 17). The comparison-adjusted funnel plot did not suggest obvious publication bias (Fig. 16). Egger’s regression analysis showed that neither the slope (P = 0.731) nor the intercept (P = 0.666) was statistically significant, indicating no evidence of a small-study effect (Table 16).

Table 16.

Egger’s regression test for complications

Variable	Coef	Std. Err	t	P	95% CI
Slope	− 0.22	0.62	− 0.35	0.731	− 1.62 to 1.18
Intercept	0.28	0.62	0.45	0.666	− 1.12 to 1.68

Table 17.

Consistency test results of network meta-analysis for complications (node-splitting)

Comparison	Direct	Indirect	Difference	P
Delta vs PE	0.63	1.17	− 0.54	0.797
Delta vs UBE	1.14	0.60	0.54	0.797
PE vs UBE	− 0.03	0.51	− 0.54	0.797

Consistency was assessed using the node-splitting method. All P values were greater than 0.05, indicating no significant inconsistency within the network.

Fig. 16.

Comparison-adjusted funnel plot for complications among three endoscopic techniques. Legend: A = Delta; B = PE; C = UBE

Sensitivity analysis

According to the ROBINS-I risk-of-bias assessment, only the five studies with a low risk of bias were selected for further analysis in the sensitivity analysis. After sensitivity analysis, the directions of effect for the major outcomes, including operative time, Visual Analog Scale score for neck pain (VAS-neck), Visual Analog Scale score for arm pain (VAS-arm), Neck Disability Index (NDI), and complications, were generally consistent with those of the primary analysis. Only slight changes in the magnitude of the effect estimates were observed, with no substantial differences in the direction of the results or the statistical conclusions.

These findings suggest that, under the current body of evidence, the main results of this study are relatively robust to the inclusion of studies at high risk of bias. However, because all included studies were retrospective observational studies, and some outcomes were based on relatively small sample sizes with the potential for small-study effects, the sensitivity analysis indicates a certain degree of robustness of the primary conclusions, while the overall strength of evidence remains limited.

Discussion

This network meta-analysis was based on retrospective observational evidence and suggested that relative differences may exist among the three posterior endoscopic techniques with respect to operative time. Within the body of evidence included in the present study, the Delta technique was associated with a shorter operative time and showed statistically significant differences compared with both PE and UBE. However, because all included studies were non-randomized comparative studies and the network did not contain any three-arm direct comparisons, this finding should be regarded as an association rather than definitive evidence that one technique is overall superior to the other two. Although Delta ranked first in the surface under the cumulative ranking curve (SUCRA) analysis, SUCRA reflects only relative ranking probabilities rather than absolute superiority; therefore, in the absence of high-quality direct comparative evidence, these findings should be interpreted with caution.Previous studies have mostly relied on pairwise comparisons to evaluate operative time. Some studies have suggested that UBE may be more efficient than PE [19, 27], whereas others have found no significant difference between the two techniques [12, 28]. With the increasing adoption of large-channel endoscopic systems, some studies have proposed that these systems may offer better surgical efficiency than conventional endoscopic techniques [9, 29]. However, the currently available evidence is still limited by relatively small sample sizes and single-comparison study designs. By incorporating PE, UBE, and Delta into a unified network meta-analysis framework, the present study provides a more systematic comparison of these three techniques to a certain extent; nevertheless, the interpretation of these findings remains constrained by the limitations inherent to retrospective study designs and indirect comparisons.

In terms of clinical efficacy, no statistically significant differences were observed among the three surgical techniques with respect to neck pain relief, as measured by the Visual Analog Scale score for neck pain (VAS-neck), or improvement in cervical function, as assessed by the Neck Disability Index (NDI). This finding is generally consistent with previous studies, which have suggested that, once adequate neural decompression is achieved, different posterior endoscopic techniques may provide comparable benefits in pain relief and functional recovery [12, 17, 19]. From a perioperative perspective, posterior endoscopic surgery is generally characterized by limited tissue trauma, less disruption of the posterior musculoligamentous structures, and relatively precise decompression targeting, which is in some respects consistent with the concept of Enhanced Recovery After Surgery (ERAS) [30]. However, the present study did not directly evaluate postoperative recovery speed, anesthesia-related outcomes, or perioperative burden; therefore, these potential advantages still require confirmation in prospective studies.Regarding arm pain (VAS-arm), a small statistical difference was observed between Delta and PE in the present study; however, the absolute effect size was small and the confidence interval was close to the null value, suggesting limited clinical relevance. In addition, no statistically significant difference was observed between Delta and UBE. Therefore, this finding is more appropriately interpreted as a limited statistical difference rather than clear evidence of a meaningful clinical advantage. Relief of radiating arm pain is often closely related to the adequacy of nerve root decompression. Previous studies have suggested that a larger working channel may facilitate foraminal enlargement and removal of compressive structures [21, 31]. However, no clear difference was observed between Delta and UBE in the present study, suggesting that once sufficient decompression is achieved, the clinical benefits of different techniques may be generally comparable. Variations among studies may also be attributable to differences in follow-up duration, pain assessment methods, and baseline disease severity [9].Operative time is an important indicator of surgical workflow efficiency and may be influenced by multiple factors, including instrument design, operative technique, and surgeon-related characteristics. The Delta system uses a larger working channel, which can accommodate a wider range of conventional surgical instruments and provide a larger operative space. Theoretically, this may facilitate procedures such as laminar drilling, ligamentum flavum resection, and foraminal enlargement, thereby influencing decompression efficiency and operative time. Previous studies on large-channel or non-coaxial endoscopic systems have likewise indicated that instrument compatibility and degree of operative freedom are important determinants of surgical efficiency [9, 32].

UBE separates the viewing portal from the working portal, thereby overcoming, to some extent, the limitations of the coaxial channel used in percutaneous endoscopic surgery (PE) and providing a wider decompression range and greater operative flexibility. Accordingly, some studies have suggested that UBE may be more efficient than PE in terms of operative performance [18]. However, UBE requires the establishment and maintenance of two independent portals. In the cervical spine, where the muscular layer is relatively thin and the anatomical space is limited, this may increase technical difficulty and, to some extent, prolong operative time [17]. This may partly explain why UBE did not outperform Delta with respect to operative time in the present study.By contrast, PE is characterized by a smaller working channel and a more restricted operative space, with greater dependence on specialized endoscopic instruments. When wider decompression or more complex procedures are required, the technical difficulty may increase, thereby prolonging operative time. Similar findings have been reported in previous comparative studies of PE and UBE [12]. In addition, differences in operative time across studies may also be related to patient selection, extent of decompression, surgeon experience, and stage of the learning curve [33]. By synthesizing data across multiple studies, the present analysis reduced, to some extent, the influence of individual study results on the overall estimate and provided a more systematic comparison of operative time among different endoscopic techniques.In terms of safety, no significant differences were observed among PE, UBE, and Delta in the incidence of complications. Because the original studies provided limited information on the types and severity of complications, complications could only be analyzed as a pooled outcome. In addition, because complication events were relatively infrequent, the corresponding confidence intervals were wide, which may have reduced the statistical power to detect true differences. The learning curve may also have influenced the safety outcomes. Previous studies have shown that both PE and UBE may be affected by limitations related to restricted operative space or the need for coordination between two portals, whereas large-channel endoscopic systems, although more similar to conventional open surgery in terms of visualization and instrument handling, may also influence the surgical learning process in theory [9, 17]. However, inconsistent reporting of surgeon experience and learning-curve stage in the original studies limited further analysis. These findings are consistent with previous evidence suggesting that precise decompression under direct endoscopic visualization may help reduce the risk of nerve and dural injury [9, 17].

This study has several limitations. First, all included studies were retrospective observational studies, and no high-quality randomized controlled trials were available; therefore, the overall certainty of evidence was generally low. In addition, clinical heterogeneity in surgical indications, surgeon experience, stage of the learning curve, extent of decompression, and follow-up duration may have influenced the pooled results [12]. Inconsistencies in the timing of pain and functional assessments across studies may also have affected the pooled estimates. Third, all comparisons were based on two-arm studies, with no direct three-arm comparisons, which limited the robustness of direct evidence within the network. Although no significant inconsistency was detected, indirect comparisons still have inherent methodological limitations. Fourth, all included studies were conducted in Asian populations, predominantly in China, which may limit the generalizability of the findings. Finally, definitions of operative time were not fully consistent across studies, and baseline covariates were incompletely reported, which precluded further subgroup analyses or meta-regression.Based on the currently available retrospective observational evidence, PE, UBE, and Delta appear to provide generally comparable outcomes in terms of pain relief, functional improvement, and complication risk for patients with cervical radiculopathy. Differences in operative time were observed among the three techniques, with Delta being associated with a shorter operative time. However, this conclusion was derived from retrospective comparative evidence and remains subject to the limitations of small-study effects and indirect comparisons; therefore, it should not be regarded as definitive evidence of the overall superiority of any single technique. Future multicenter, large-scale, prospective randomized controlled trials are needed to further evaluate the relative effects of different posterior endoscopic techniques and to clarify their most appropriate indications across various clinical settings.

Conclusions

The results of this network meta-analysis indicate that the three posterior endoscopic techniques—percutaneous endoscopic surgery (PE), unilateral biportal endoscopic surgery (UBE), and the Delta technique—show generally comparable performance in terms of pain relief, functional improvement, and complication risk for cervical radiculopathy. Among the evaluated outcomes, Delta was associated with a shorter operative time and showed only a very small difference in arm pain compared with PE. However, the effect sizes of these differences were small, and most of the available evidence was derived from retrospective observational studies; therefore, their clinical significance and the stability of these findings remain uncertain. These results do not support the conclusion that Delta is overall superior to PE or UBE.Because the currently available evidence is derived mainly from retrospective studies conducted in Asian populations, and high-quality, multicenter, prospective randomized controlled trials are lacking, these conclusions should be interpreted with caution. Further multicenter, large-scale, well-designed randomized controlled trials conducted across different regions and healthcare systems are needed to clarify the relative effects of different posterior endoscopic techniques and to define their optimal indications.

Author contributions

Yingxuan Rong drafted the initial manuscript.Yingxuan Rong and Hongjian Zheng jointly conducted the literature screening, data extraction, statistical analysis, and figure preparation.As the corresponding author, Hongjian Zheng supervised the study design and oversaw the entire research process. Junjie Chen also provided some statistical assistance for this study.

Funding

This network meta-analysis was conducted based on previously published literature and received no funding from governmental agencies, research institutions, or other funding bodies.

Data availability

All data used in this study were obtained from publicly available published literature. No patient-level data or clinical interventions were involved; therefore, there were no associated ethical concerns. The data were rigorously screened and standardized, ensuring reliable sources and controlled quality.

Declarations

Ethics approval and consent to participate

This study was conducted in strict accordance with the ethical principles governing medical research and was reviewed and approved by the Ethics Committees of Zhongshan Orthopedic Hospital and The Affiliated Shunde Hospital of Jinan University.

Consent to publish

Not applicable.

Informed consent

Informed consent was not required for this study because all data were obtained from publicly available published literature and databases, and no direct patient involvement or collection of personal identifiable information was involved.

Human or animal rights

Not applicable.

Competing interests

The authors declare no competing interests.