Comparison of efficacy and safety of different treatment modalities for hemorrhoids: A network meta-analysis

Abstract

BACKGROUND

Hemorrhoidal disease is a prevalent anorectal condition for which a variety of interventional treatments are available; however, their comparative efficacy and complication profiles remain incompletely defined.

AIM

To compare the clinical effectiveness and safety of hemorrhoidal treatments via a network meta-analysis.

METHODS

We systematically searched major electronic databases for randomized controlled trials published up to June 2026. A Bayesian network meta-analysis was performed to compare the 20 interventions in terms of recurrence, efficacy, patient satisfaction, subsequent treatment rate, and adverse events, with subgroup analyses stratified by Goligher grade.

RESULTS

Sixty-six studies encompassing 8549 patients and 20 therapeutic modalities were included. LigaSure™ hemorrhoidectomy demonstrated a significantly lower recurrence risk than Milligan–Morgan hemorrhoidectomy [odds ratio (OR) = 0.18, 95% credible interval (95%CI): 0.03–0.94] and ranked first in effectiveness [surface under the cumulative ranking curve (SUCRA) = 98.2%], with the lowest further treatment rate (SUCRA = 4.3%), although it ranked higher for pain (82.5%) and urinary retention (75.4%). The tissue-selecting technique showed a nearly significant reduction in recurrence (OR = 0.210, 95%CI: 0.037–1.000) and improvements in safety outcomes, particularly anorectal stenosis (SUCRA = 20.0%) and constipation (11.7%). Direct current electrotherapy also showed a favorable SUCRA ranking for recurrence, but this result was derived from a very limited number of studies with small sample sizes. Rubber band ligation was associated with a markedly higher recurrence risk (OR = 6.7; SUCRA = 90.5%), as was injection sclerotherapy (SUCRA = 76.4%); both interventions had low effectiveness rankings. Subgroup analyses suggested that procedure performance might vary by Goligher grade, with LigaSure™ hemorrhoidectomy and the tissue-selecting technique showing potential advantages in severe (grade ≥ III) disease, although these findings should be interpreted with caution because of limited data.

CONCLUSION

LigaSure™ hemorrhoidectomy and the tissue-selecting technique are preferred for surgical candidates with severe disease; rubber band ligation and sclerotherapy are better suited for mild disease or for patients who are not surgical candidates. Treatment selection may be individualized according to disease severity, although the current evidence for grade-based recommendations remains exploratory.

Key Words: Hemorrhoids; Network meta-analysis; Hemorrhoidectomy; Recurrence; Complication

Core Tip: This network meta-analysis of 66 randomized controlled trials involving 8549 patients compares 20 therapeutic modalities for hemorrhoidal disease. LigaSure™ hemorrhoidectomy and the tissue-selecting technique demonstrate the most favorable balance between recurrence control and safety, particularly in severe (grade ≥ III) disease. Rubber band ligation and injection sclerotherapy, though safer, carry substantially higher recurrence risks and are best suited for mild disease or patients who are not surgical candidates. Treatment decisions should be individualized, with grade-based recommendations considered exploratory.

INTRODUCTION

Hemorrhoidal disease is one of the most prevalent anorectal disorders in adults, with a prevalence ranging from 4.4% to 25.9% and a peak incidence between 45 years and 65 years of age[1-3]. The condition is characterized by pathologic hypertrophy and distal displacement of the normal anal cushions, a process thought to be influenced by abnormal defecation habits, a low-fiber diet, pregnancy, and other contributing factors[4]. Clinically, hemorrhoidal disease presents with bleeding, prolapse, pain, pruritus, and perianal hygiene difficulties, with bleeding and prolapse being the most frequent presenting symptoms. These symptoms can impair patients’ quality of life by interfering with daily activities, social engagement, and emotional well-being, highlighting the considerable functional and psychological burden imposed by the disease[5,6].

The Goligher classification system, which grades internal hemorrhoids from I to IV according to the degree of prolapse, remains a cornerstone for guiding treatment decisions[2]. Treatment follows a stepwise algorithm based on both grade and symptom severity, with conservative and office-based interventions preferred for lower-grade disease and surgical approaches reserved for higher-grade or refractory cases[7]. The therapeutic armamentarium is notably diverse, encompassing lifestyle modifications and pharmacologic therapy, office-based procedures such as sclerotherapy and rubber band ligation, and surgical excisional techniques for advanced or complex hemorrhoidal disease[8]. In recent years, there has been a broader trend toward minimally invasive strategies such as the procedure for prolapse and hemorrhoids, Doppler-guided hemorrhoidal artery ligation, LigaSure™, and Harmonic^® devices, aimed at achieving favorable efficacy while mitigating postoperative pain, shortening recovery times, and improving perioperative outcomes[9,10]. Despite these advances, there remains a lack of consensus regarding the comparative effectiveness and safety profiles of different procedures, underscoring the need for high-quality evidence to inform clinical decision-making.

Multiple randomized controlled trials (RCTs) have directly compared various surgical and office-based interventions for hemorrhoidal disease, and a number of pairwise and network meta-analyses have sought to consolidate the evidence on their relative efficacy and safety. Nevertheless, existing meta-analyses are constrained by several notable limitations. Most have included only around 10 procedural types and therefore fail to account for recently introduced techniques, such as the tissue-selecting therapy stapler and modified ligation procedures. Furthermore, few have conducted subgroup analyses stratified by hemorrhoid severity, leaving a critical clinical question unresolved—namely, how best to tailor treatment selection to patients across different Goligher grades. In addition, the exclusion of evidence published within the past 5 years further undermines the contemporary relevance of their findings. These shortcomings collectively highlight the urgent need for an updated, comprehensive systematic evaluation that directly compares emerging therapeutic options against established conventional approaches.

In response to these evidence gaps, we performed a comprehensive network meta-analysis to systematically compare the efficacy and safety of currently available surgical and minimally invasive interventions for hemorrhoidal disease. Our study incorporates a wider array of procedural modalities than earlier syntheses, including novel techniques, and features prespecified subgroup analyses stratified by Goligher grade to assess treatment performance across distinct severity levels. The findings of this study may provide clinicians with a useful evidence-based reference for individualized treatment decision-making, pending further validation.

MATERIALS AND METHODS

Study design and search strategy

The authors have read the PRISMA 2009 Checklist, and the manuscript was prepared and revised according to the PRISMA 2009 Checklist[11]. This meta-analysis was prospectively registered with the International Prospective Register of Systematic Reviews (registration No. CRD420261444034).

We searched PubMed, EMBASE, the Cochrane Central Register of Controlled Trials, and the Web of Science Core Collection from January 1, 2010 to June 1, 2026, and supplemented these searches with searches of ClinicalTrials.gov and Google Scholar for gray literature. The search strategy combined controlled vocabulary (MeSH and Emtree) with free-text terms across 3 domains: Condition (hemorrhoids), interventions (surgical and minimally invasive procedures), and design (RCTs).

For the disease domain, we used the MeSH term “Hemorrhoids” and the free-text terms “hemorrhoid” and “piles” in title and abstract fields. For the intervention domain, to comprehensively capture the 20 therapeutic modalities prespecified in our protocol, the search strategy incorporated both controlled vocabulary and free-text terms for the following procedural categories: (1) Classical excisional techniques: Milligan–Morgan hemorrhoidectomy, Ferguson hemorrhoidectomy, semi-closed hemorrhoidectomy, and modified Park’s submucosal hemorrhoidectomy; (2) Energy-based platforms: LigaSure™ hemorrhoidectomy, harmonic scalpel hemorrhoidectomy, bipolar coagulation, radiofrequency coagulation, and laser hemorrhoidectomy; (3) Stapling and selective resection: Stapled hemorrhoidopexy and the tissue-selecting technique; (4) Arterial ligation and suture-based procedures: Hemorrhoidal artery ligation and suture ligation with mucopexy; (5) Minimally invasive non-excisional therapies: Rubber band ligation, injection sclerotherapy, infrared coagulation, and direct current electrotherapy; and (6) Other modified procedures: Modified ligation procedure for prolapsed hemorrhoids, eastern rubber band ligation and coincident dislocation for prolapse and hemorrhoid, and combined hemorrhoidal radiocoagulation.

For the study-design domain, we applied filters and free-text terms such as “randomized controlled trial,” “controlled clinical trial,” and “randomized.” In addition, we manually screened the reference lists of all included articles and relevant reviews to identify any additional potentially eligible studies not captured by the electronic searches.

Eligibility criteria and study selection

We established eligibility criteria according to the Population, Intervention, Comparator, Outcomes, and Study Design framework, as follows: (1) Population: Patients with a confirmed diagnosis of hemorrhoids based on clinical examination or endoscopy, regardless of nationality, race, sex, or age, and without restriction by Goligher grade; (2) Intervention: Any surgical or minimally invasive procedure used for the treatment of hemorrhoids; (3) Comparator: Any of the aforementioned interventions compared against one another; (4) Outcomes: Reports of at least one of the following outcomes—recurrence rate, effectiveness rate, patient satisfaction rate, rate of further treatment, or postoperative complications including pain, bleeding, urinary retention, septic complications, anal stenosis, incontinence, and constipation; and (5) Study design: RCTs, irrespective of blinding status.

We excluded: (1) Studies of non-hemorrhoid populations or those with mixed anorectal pathologies from which hemorrhoid-specific data were inseparable; (2) Non-randomized designs (e.g., observational cohorts, case series, and case reports); (3) Non-original articles (reviews, meta-analyses, editorials, and letters); (4) Basic science and animal studies; (5) Combination therapies precluding isolation of individual intervention effects; (6) Duplicate publications, retaining only the most complete or earliest dataset; (7) Studies with insufficient or unusable outcome data; (8) Comparisons within the same procedural category differing solely in technical details, formulations, or device types (e.g., different sclerosants for sclerotherapy), to preserve each network node as a distinct clinical strategy; and (9) Non-English publications.

Study selection was performed independently by 2 reviewers. All records identified through the electronic searches were imported into EndNote reference management software for duplicate removal. The 2 reviewers then independently screened titles and abstracts to exclude studies that clearly did not meet the eligibility criteria. Full texts of potentially relevant studies were retrieved and independently assessed against the predefined inclusion and exclusion criteria. Disagreements during the screening process were resolved through discussion; if consensus could not be reached, a third reviewer acted as the arbiter.

Outcomes

The primary efficacy outcome was recurrence, defined as the reappearance of hemorrhoidal signs or symptoms confirmed by clinical examination or endoscopy after treatment, including events that necessitated reintervention or additional therapy. Because the timing of recurrence assessment varied across studies (e.g., at 6 months, 1 year, or longer postoperatively), we extracted the number of events and the total number of patients from each trial according to each study’s own definition of recurrence, and these data were pooled for the network meta-analysis.

Effectiveness was defined as the proportion of patients achieving cure or treatment success according to the criteria specified in each original study. Although effectiveness definitions varied across trials, they consistently encompassed complete resolution of symptoms and sufficient shrinkage or retraction of hemorrhoidal tissue. We extracted the number of patients classified as having a “cured” or “successful” outcome according to each study’s own categorization. Satisfaction was defined as the proportion of patients reporting being “very satisfied” or the highest satisfaction category available in each study. Further treatment rate was defined as the proportion of patients requiring additional intervention—including repeat surgery, rubber band ligation, sclerotherapy, or any other supplementary procedure—following the initial treatment due to insufficient efficacy, symptom recurrence, or complications.

Safety outcomes comprised the following postoperative complications: Pain, bleeding, urinary retention, septic complications, anorectal stenosis, incontinence, and constipation. For each complication, we extracted the number of events and the total number of patients based on the definitions used in each original study. For the pain outcome, we extracted the number of events and total number of patients according to each study’s own definition of clinically significant pain complications. Because most studies reported only mean visual analog scale scores at the group level rather than individual patient-level data, we were unable to convert continuous pain measurements into binary event counts and therefore extracted the data as reported without additional transformation.

Data extraction

Data extraction was carried out independently by 2 reviewers, followed by systematic cross-verification of all extracted data. Any discrepancies between reviewers were resolved through discussion; in cases where no consensus could be reached, a third reviewer served as the final adjudicator.

We designed and pilot-tested a standardized data extraction form before use. The following data were collected from each study: (1) Study characteristics (first author, year, country, design, blinding, sample size, and follow-up); (2) Participant characteristics (age, sex, and Goligher grade); (3) Intervention details (procedure names per arm); and (4) Outcome data (event counts and patient denominators for binary outcomes). When multiple follow-up time points were reported, we extracted data from the longest available follow-up to assess long-term effects. All data were entered into Excel and cross-checked against the original articles.

Risk-of-bias assessment

The methodological quality of the included RCTs was assessed using the Cochrane Collaboration’s Risk of Bias 2.0 (RoB 2.0) tool[12], which evaluates 5 domains: Randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selection of the reported result. Each domain was rated as “low risk,” “high risk,” or “some concerns,” and an overall judgment was assigned per trial.

Statistical analysis

The network meta-analysis was performed within a Bayesian framework using the gemtc package implemented in R software (version 4.3.1). Effect estimates were calculated as odds ratios (ORs) with associated 95% credible intervals (95%CIs). For binary outcomes, we adopted a binomial likelihood with a logit link function and used a random-effects model to account for between-study heterogeneity. Non-informative prior distributions were specified for all relative effect parameters. To ensure stable and reliable estimation, we ran 4 independent Markov chains for each model, with the number of iterations determined by convergence diagnostics following a sufficient burn-in phase.

Network evidence plots were constructed using the mtc.network function from the gemtc package, wherein each node represented a distinct treatment modality and each edge corresponded to a direct pairwise comparison. Milligan–Morgan hemorrhoidectomy was prespecified as the reference comparator throughout the network, reflecting its role as the most extensively validated and conventionally accepted surgical standard. A random-effects consistency model was applied to combine both direct and indirect evidence, generating pooled ORs with 95%CIs for all possible pairwise contrasts, which were subsequently organized into a league table. To evaluate relative treatment performance, we estimated the probability of each intervention occupying each rank position and derived the surface under the cumulative ranking curve (SUCRA) as a summary measure of ranking. Convergence of the Markov chains was monitored using the potential scale reduction factor, with values < 1.05 deemed acceptable, alongside qualitative inspection of trace plots to confirm adequate mixing. To assess local inconsistency, we applied the node-splitting approach, which partitions direct and indirect evidence within each closed loop; a P value < 0.05 was taken to denote statistically significant inconsistency. Between-study heterogeneity was evaluated via the mtc.anohe function to compute the I² statistic for each direct comparison. Publication bias was explored using comparison-adjusted funnel plots for visual assessment and Egger’s regression test for quantitative evaluation, with a P value > 0.05 indicating no evidence of significant publication bias.

In addition, we performed prespecified subgroup analyses stratifying studies according to the Goligher grade of included patients into 4 subgroups: Grade ≤ II, grade ≤ III, grade ≥ II, and grade ≥ III. The network meta-analysis model was rerun separately for each subgroup to evaluate the performance of each intervention across different severity strata. Two sensitivity analyses for recurrence were performed by (1) excluding studies with an overall high risk of bias and (2) restricting the analysis to studies with low risk in the randomization process domain. To further evaluate the potential impact of clinical heterogeneity on the primary findings, a network meta-regression analysis for the recurrence outcome was performed by adjusting for 3 study-level covariates: Publication year, grade ≥ III hemorrhoids, and sample size. The adjusted ORs and 95%CIs were calculated.

To provide an integrated assessment that simultaneously considers both efficacy and safety outcomes, we performed a multi-criteria decision analysis using the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method. This approach synthesizes multiple outcomes into a single composite utility score, enabling direct comparison of interventions across benefit and harm dimensions. The analysis was based on the SUCRA values for all 11 outcomes. For each indicator, we defined the direction of preference: Higher values were preferred for effectiveness and satisfaction, whereas lower values were preferred for recurrence, further treatment rate, and all safety outcomes. The decision matrix was constructed with 19 treatment modalities (those with at least 6 non-missing outcomes) as alternatives and the 11 SUCRA values as criteria. Min–max normalization was applied to transform all criteria to a uniform direction (higher = better). Indicator weights were objectively determined using the entropy weight method. The TOPSIS method was then applied to calculate the relative closeness coefficient (composite score) for each treatment, with scores ranging from 0 to 1. Higher scores indicate better overall performance in balancing efficacy against safety.

RESULTS

Study characteristics and risk-of-bias assessment

As shown in Figure 1, of the 6896 records identified through the initial search, 66 articles met the eligibility criteria and were included in the final analysis[13-78], encompassing 8549 patients and 20 distinct treatment modalities. Hemorrhoidal disease was predominantly grade III or higher; 26 studies (39.4%) included grade II patients, while only 3 (4.5%) included grade I disease (Table 1).

Figure 1 Study selection flow diagram. The flow diagram illustrates the identification, screening, eligibility assessment, and inclusion of studies in the network meta-analysis. Of the 6896 records initially identified through database searches and other sources, 66 randomized controlled trials met the eligibility criteria and were included in the quantitative synthesis. RCT: Randomized controlled trial.

Table 1 Characteristics of included studies.

Reference	Grade of hemorrhoids	Treatment	Sample size	Recurrence	Effectiveness	Satisfaction	Further treatment	Pain	Bleeding	Urinary retention	Septic complications	Anorectal stenosis	Incontinence	Constipation
McKenzie et al[13], 2010	II	RBL vs SH	60	Reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
Sakr[14], 2010	III/IV	MM vs LigaH	84	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported
Nyström et al[15], 2010	III/IV	MM vs SH	180	Not reported	Reported	Not reported	Not reported	Reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported
Abo-hashem et al[16], 2010	III/IV	HarS vs BPC	64	Not reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Not reported	Reported	Not reported
Filingeri et al[17], 2010	IV	FH vs RFC	28	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Not reported
Sakr et al[18], 2010	III/IV	SH vs LigaH	68	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported
Giordano et al[19], 2011	II/III	SH vs HAL	52	Reported	Not reported	Reported	Reported	Reported	Reported	Reported	Not reported	Reported	Reported	Not reported
Zaher et al[20], 2011	II/III/IV	SH vs RBL	26	Reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Not reported	Reported	Not reported
Franceschilli et al[21], 2011	III/IV	MM vs LigaH	210	Reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported
Giamundo et al[22], 2011	II/III	HAL vs RBL	60	Not reported	Reported	Reported	Not reported	Not reported	Reported	Reported	Not reported	Not reported	Reported	Not reported
Infantino et al[23], 2012	III	SH vs HAL	169	Reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Reported	Not reported	Reported	Reported
Gupta et al[24], 2011	III	SLM vs HAL	48	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Not reported	Not reported	Not reported
Arslani et al[25], 2012	III	SH vs LigaH	98	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Not reported	Reported	Reported	Not reported
Filingeri et al[26], 2012	II	RBL vs CHR	90	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
Ammaturo et al[27], 2012	III	MM vs SH	79	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Not reported	Reported	Not reported
Zampieri et al[28], 2012	III/IV	HAL vs LigaH	114	Not reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Not reported	Reported	Not reported
Elmér et al[29], 2013	II/III	MM vs HAL	40	Not reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Not reported	Reported	Reported	Not reported
Kim et al[30], 2013	III	MM vs SH	122	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Not reported	Not reported	Reported	Not reported
Bulus et al[31], 2014	III/IV	FH vs HarS	151	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Reported	Not reported
Verre et al[32], 2013	III/IV	HAL vs SH	122	Not reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported
Ahmad et al[33], 2013	I/II	HAL vs IRC	245	Not reported	Reported	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
Gagloo et al[34], 2013	II/III	MM vs RBL	100	Not reported	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
De Nardi et al[35], 2014	III	MM vs HAL	50	Reported	Not reported	Reported	Reported	Not reported	Reported	Not reported	Not reported	Reported	Reported	Not reported
Talha et al[36], 2017	III/IV	MM vs HarS vs LigaH	180	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Not reported	Reported	Not reported	Not reported
Denoya et al[37], 2014	III/IV	FH vs HAL	27	Reported	Reported	Not reported	Reported	Reported	Reported	Not reported	Not reported	Reported	Reported	Not reported
Elshazly et al[38], 2015	II/III	MM vs SLM	200	Reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Not reported	Not reported	Reported	Reported
Bilgin et al[39], 2015	III/IV	SH vs HarS	99	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Not reported
Wang et al[40], 2015	III/IV	MM vs TST	480	Not reported	Reported	Reported	Not reported	Reported	Not reported	Reported	Not reported	Not reported	Reported	Reported
Lu et al[41], 2015	III/IV	MM vs SH	108	Reported	Not reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Reported
Ripetti et al[42], 2015	III/IV	MM vs SCH vs SH	180	Reported	Not reported	Not reported	Not reported	Reported	Reported	Reported	Not reported	Reported	Reported	Not reported
Aigner et al[43], 2016	III	HAL vs SLM	40	Reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported	Not reported
Zhai et al[44], 2016	III	HAL vs SLM	100	Reported	Reported	Reported	Reported	Not reported	Reported	Reported	Not reported	Reported	Reported	Not reported
Brown et al[45], 2016	II/III	HAL vs RBL	372	Reported	Reported	Not reported	Reported	Reported	Reported	Reported	Reported	Not reported	Not reported	Not reported
Leardi et al[46], 2016	II/III	SH vs HAL	100	Reported	Not reported	Reported	Reported	Reported	Reported	Not reported	Not reported	Not reported	Reported	Not reported
Watson et al[47], 2016	II/III/IV	MM vs SH	774	Reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported	Not reported	Reported
Lehur et al[48], 2016	II/III	SH vs HAL	393	Reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported
Naderan et al[49], 2017	II/III	MM vs LASER	60	Reported	Reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Not reported	Not reported	Not reported
Nikooiyan et al[50], 2016	I/II/III/IV	FH vs DCV	120	Reported	Reported	Not reported	Reported	Reported	Reported	Reported	Reported	Not reported	Reported	Reported
Tsunoda et al[51], 2017	III	HAL vs HarS	44	Not reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Not reported	Reported	Not reported	Not reported
Leung et al[52], 2017	II/III	HAL vs TST	80	Reported	Not reported	Not reported	Reported	Not reported	Not reported	Reported	Not reported	Reported	Not reported	Not reported
Nikshoar et al[53], 2018	III	FH vs IRC	40	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported
Giarratano et al[54], 2018	III/IV	SH vs HAL	100	Reported	Not reported	Reported	Reported	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Not reported
Venara et al[55], 2018	II/III	SH vs HAL	377	Reported	Not reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported
Lisi et al[56], 2019	III/IV	LigaH vs RFC	50	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Not reported	Reported	Reported	Not reported
Carvajal López et al[57], 2019	III/IV	MM vs HAL	40	Reported	Reported	Not reported	Reported	Reported	Reported	Reported	Not reported	Not reported	Not reported	Not reported
Trenti et al[58], 2019	III/IV	LigaH vs HAL	80	Not reported	Not reported	Reported	Not reported	Not reported	Reported	Reported	Reported	Not reported	Not reported	Not reported
Genova et al[59], 2019	III/IV	MM vs HAL	89	Reported	Not reported	Not reported	Reported	Not reported	Reported	Not reported	Reported	Reported	Not reported	Not reported
Poskus et al[60], 2020	II/III	MM vs SLM vs LASER	121	Reported	Reported	Not reported	Reported	Not reported	Not reported	Not reported	Reported	Reported	Not reported	Not reported
Rørvik et al[61], 2020	II/III/IV	MM vs HAL	98	Reported	Not reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported	Reported	Not reported
Abiodun et al[62], 2020	II/III	IJS vs RBL	60	Reported	Reported	Reported	Reported	Reported	Reported	Not reported	Reported	Not reported	Not reported	Not reported
Jin et al[63], 2021	III	RBL vs MM	120	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Not reported	Not reported	Not reported	Not reported
Zhai et al[64], 2022	II/III/IV	SLM vs MM	123	Reported	Not reported	Reported	Reported	Not reported	Reported	Not reported	Not reported	Not reported	Reported	Not reported
Lee et al[65], 2021	III/IV	HAL vs LigaH	80	Not reported	Reported	Reported	Reported	Reported	Reported	Reported	Not reported	Not reported	Not reported	Not reported
Huang et al[66], 2021	III/IV	MM vs TST	169	Reported	Not reported	Reported	Not reported	Not reported	Reported	Reported	Not reported	Reported	Not reported	Not reported
Song et al[67], 2022	III/IV	MPSH vs MM	186	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Reported	Not reported	Not reported
Salgueiro et al[68], 2022	I/II/III	IJS vs RBL	120	Reported	Reported	Not reported	Reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Not reported
Trenti et al[69], 2023	III/IV	HAL vs LigaH	75	Not reported	Not reported	Reported	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Reported	Not reported
Yang et al[70], 2022	II/III/IV	SH vs MLPPH	130	Not reported	Reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Not reported	Not reported
Mert[71], 2023	III/IV	LASER vs FH	100	Not reported	Reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Not reported	Reported	Not reported
Huang et al[72], 2023	III/IV	TST vs SH	306	Reported	Reported	Not reported	Not reported	Not reported	Reported	Reported	Not reported	Reported	Not reported	Not reported
Cemil et al[73], 2024	II/III	LASER vs MM	85	Not reported	Not reported	Not reported	Not reported	Not reported	Reported	Not reported	Not reported	Not reported	Reported	Not reported
Nada et al[74], 2024	III/IV	HarS vs SH	70	Reported	Not reported	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Not reported
Jin et al[75], 2024	II	LASER vs RBL	70	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Reported	Not reported	Reported	Not reported
Mustapha et al[76], 2025	II/III/IV	LigaH vs MM	52	Not reported	Reported	Reported	Reported	Not reported	Not reported	Reported	Not reported	Not reported	Reported	Not reported
Huang et al[77], 2025	III/IV	TST vs SH vs EPH	141	Not reported	Reported	Not reported	Not reported	Reported	Not reported	Reported	Not reported	Reported	Not reported	Not reported
Fernandez-Hurtado et al[78], 2025	III	HAL vs MM	50	Reported	Reported	Not reported	Reported	Not reported	Not reported	Reported	Not reported	Reported	Reported	Not reported

BPC: Bipolar coagulation; CHR: Combined hemorrhoidal radiocoagulation; DCV: Direct current electrotherapy; EPH: Eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid; FH: Ferguson hemorrhoidectomy; HAL: Hemorrhoidal artery ligation; HarS: Harmonic scalpel hemorrhoidectomy; IJS: Injection sclerotherapy; IRC: Infrared coagulation; LASER: Laser hemorrhoidectomy; LigaH: LigaSure™ hemorrhoidectomy; MLPPH: Modified ligation procedure for prolapsed hemorrhoids; MM: Milligan–Morgan hemorrhoidectomy; MPSH: Modified Park’s submucosal hemorrhoidectomy; RBL: Rubber band ligation; RFC: Radiofrequency coagulation; SCH: Semi-closed hemorrhoidectomy; SH: Stapled hemorrhoidopexy; SLM: Suture ligation with mucopexy; TST: Tissue-selecting technique.

Figure 2 presents the detailed risk-of-bias evaluations across the 5 domains of the RoB 2.0 tool. Overall, the methodological quality of the included studies was satisfactory, with the majority of trials judged as low risk across most domains. Five studies (7.6%) received an overall rating of high risk of bias; the most frequent sources of concern were insufficient detail in the description of the randomization process and incomplete outcome data attributable to participant attrition in several trials.

Figure 2 Risk-of-bias assessment. A: Risk-of-bias assessments for the 66 included randomized controlled trials, showing the proportion of studies rated as “low risk,” “some concerns,” or “high risk” for each of the 5 Risk of Bias 2.0 domains: Randomization process, deviations from intended interventions, missing outcome data, outcome measurement, and selective reporting; B: Overall risk-of-bias judgment for the included studies, categorized as low risk, some concerns, or high risk; C: Traffic light plot providing a study-by-study visualization of domain-level judgments (green = low risk; yellow = some concerns; orange = high risk). The majority of studies (81.8%) were rated as having some concerns regarding risk of bias; 5 studies (7.6%) were rated as high risk, primarily due to inadequate randomization methods or substantial loss to follow-up.

Primary outcome

Primary analysis: A network for recurrence was formed by 16 of the 20 therapeutic modalities that reported this outcome (Figure 3A). Using Milligan–Morgan hemorrhoidectomy as the reference comparator, LigaSure™ hemorrhoidectomy was associated with a significantly lower risk of recurrence (OR = 0.18; 95%CI: 0.03–0.94), and the tissue-selecting technique showed a nearly significant reduction in recurrence risk (OR = 0.21; 95%CI: 0.037–1.000). Conversely, hemorrhoidal artery ligation (OR = 3.6; 95%CI: 1.7–8.5) and rubber band ligation (OR = 6.7; 95%CI: 2.1–23.0) were associated with significantly higher recurrence risks than Milligan–Morgan hemorrhoidectomy, whereas stapled hemorrhoidopexy (OR = 1.90; 95%CI: 0.96–4.10) and suture ligation with mucopexy (OR = 2.70; 95%CI: 0.96–8.20) showed a nearly significant trend toward higher recurrence (Figure 3B).

Figure 3 Network meta-analysis results for recurrence. A: Network geometry of the evidence for recurrence, with node size proportional to the number of patients and edge width proportional to the number of direct comparisons; B: Forest plot of odds ratios for recurrence, with Milligan–Morgan hemorrhoidectomy as the reference; C: Surface under the cumulative ranking curve values for recurrence (higher values indicate a greater probability of being the most recurrence-prone intervention); D: League table of pairwise comparisons, showing odds ratios and 95% credible intervals for recurrence. 95%CI: 95% credible interval; DCV: Direct current electrotherapy; FH: Ferguson hemorrhoidectomy; HAL: Hemorrhoidal artery ligation; HarS: Harmonic scalpel hemorrhoidectomy; IJS: Injection sclerotherapy; IRC: Infrared coagulation; LASER: Laser hemorrhoidectomy; LigaH: LigaSure™ hemorrhoidectomy; MM: Milligan–Morgan hemorrhoidectomy; MPSH: Modified Park’s submucosal hemorrhoidectomy; OR: Odds ratio; RBL: Rubber band ligation; RFC: Radiofrequency coagulation; SCH: Semi-closed hemorrhoidectomy; SH: Stapled hemorrhoidopexy; SLM: Suture ligation with mucopexy; TST: Tissue-selecting technique.

The ranking probabilities for recurrence across all modalities contributing to the network (Figure 3C) revealed that rubber band ligation (SUCRA, 90.5%), hemorrhoidal artery ligation (80.4%), and injection sclerotherapy (76.4%) were associated with the highest likelihood of being the most recurrence-prone interventions. By contrast, direct current electrotherapy (SUCRA = 12.3%), LigaSure™ hemorrhoidectomy (19.3%), and the tissue-selecting technique (20.9%) demonstrated the lowest probabilities of occupying the top rank for recurrence. Nevertheless, the evidence base for direct current electrotherapy was notably limited, comprising only a small number of studies, and the corresponding OR estimates were accompanied by wide CIs that included the null value. Consequently, these particular results should be viewed with appropriate caution. The league table presenting all pairwise comparisons (Figure 3D) was consistent with and reinforced the ranking analysis described above.

Convergence assessment: As shown in Supplementary Figure 1, convergence diagnostics demonstrated excellent model performance. The potential scale reduction factor for all estimated parameters had both point estimates and upper confidence limits of 1.00, and the multivariate potential scale reduction factor was also 1.00—all substantially lower than the conventional threshold of 1.05. These convergence metrics reflect negligible between-chain variability and confirm that the Markov chains achieved stable mixing, thereby supporting the robustness and dependability of the network meta-analysis results.

Inconsistency assessment: Node-splitting analysis for local inconsistency is presented in Supplementary Table 1. For 19 of the 21 comparisons, P values exceeded 0.05, indicating no significant inconsistency between direct and indirect evidence across the majority of the network. Significant inconsistency was detected for the comparisons of Milligan–Morgan hemorrhoidectomy vs tissue-selecting technique (P = 0.006) and Milligan–Morgan hemorrhoidectomy vs hemorrhoidal artery ligation (P = 0.013). For both comparisons, the direction of the direct and indirect estimates was consistent, implying that the inconsistency affected precision rather than the overall direction of the rankings. However, for the comparison involving hemorrhoidal artery ligation, the direct estimate (log OR = −36.0) and indirect estimate (log OR = −1.0) diverged substantially in magnitude, a finding that may be attributed to the limited number of trials providing direct evidence for this contrast.

To further assess the impact of the identified local inconsistencies, we performed an analysis using the unrelated mean effects model, which does not assume consistency and allows direct and indirect evidence to differ. For tissue-selecting technique vs Milligan–Morgan hemorrhoidectomy, the unrelated mean effects model yielded an OR of approximately 0 (95%CI: 4.400 × 10⁻⁴² to 0.046). For hemorrhoidal artery ligation vs Milligan–Morgan hemorrhoidectomy, the unrelated mean effects model produced an OR of 19.59 (95%CI: 3.59–135.96). Although the magnitude differed, the direction of effect remained unchanged for both comparisons, supporting the robustness of the primary findings.

Heterogeneity assessment: Heterogeneity was assessed using pairwise I² statistics and the network I² statistic under the consistency model, with P values for the comparison of heterogeneity and inconsistency provided in Supplementary Table 2. Low-to-moderate heterogeneity was observed for most comparisons (I² < 50%), with substantial heterogeneity detected only for the direct comparison between hemorrhoidal artery ligation and stapled hemorrhoidopexy (I² = 80.33%). By contrast, I² values of 0% were found for comparisons involving laser hemorrhoidectomy, LigaSure™ hemorrhoidectomy, harmonic scalpel hemorrhoidectomy, and suture ligation with mucopexy, reflecting a high degree of consistency. In the consistency testing, a P value of 0.047 was obtained for the hemorrhoidal artery ligation–stapled hemorrhoidopexy loop, indicating suboptimal fit of the consistency model for this comparison; all other estimable loops had P values > 0.05. Overall, the heterogeneity across the included studies was deemed to be within an acceptable range.

Publication bias assessment: Publication bias was evaluated using a comparison-adjusted funnel plot (Figure 4) and Egger’s regression test. The funnel plot revealed a roughly symmetrical distribution of effect estimates around the overall summary effect, without obvious asymmetry. Egger’s test produced a P value of 0.11, indicating no statistically significant publication bias. Taken together, these results suggest that the stability and validity of the findings were unlikely to be substantially affected by publication bias.

Figure 4 Comparison-adjusted funnel plot for publication bias. A comparison-adjusted funnel plot is shown for the recurrence outcome in the network meta-analysis. Each point represents a direct pairwise comparison, with the effect estimate plotted against its standard error. A roughly symmetrical distribution around the overall summary effect is observed, with no overt asymmetry. Egger’s test yielded a P value of 0.110, indicating no statistically significant evidence of publication bias. DCV: Direct current electrotherapy; FH: Ferguson hemorrhoidectomy; HAL: Hemorrhoidal artery ligation; HarS: Harmonic scalpel hemorrhoidectomy; IJS: Injection sclerotherapy; IRC: Infrared coagulation; LASER: Laser hemorrhoidectomy; LigaH: LigaSure™ hemorrhoidectomy; MM: Milligan–Morgan hemorrhoidectomy; MPSH: Modified Park’s submucosal hemorrhoidectomy; RBL: Rubber band ligation; SCH: Semi-closed hemorrhoidectomy; SH: Stapled hemorrhoidopexy; SLM: Suture ligation with mucopexy; TST: Tissue-selecting technique.

Secondary outcomes

Effectiveness: When compared with Milligan–Morgan hemorrhoidectomy (Supplementary Figure 2), only LigaSure™ hemorrhoidectomy and the modified ligation procedure for prolapsed hemorrhoids were associated with ORs greater than 1 for effectiveness; however, neither estimate reached statistical significance, and both were accompanied by wide CIs. Significantly lower effectiveness relative to the reference was observed for hemorrhoidal artery ligation, infrared coagulation, and rubber band ligation. Ranking probabilities, as shown in Table 2, indicated that LigaSure™ hemorrhoidectomy (SUCRA = 98.2%), the modified ligation procedure for prolapsed hemorrhoids (83.9%), and Milligan–Morgan hemorrhoidectomy (73.3%) had the highest likelihood of being ranked first for effectiveness, whereas infrared coagulation (9.4%), rubber band ligation (16.3%), and injection sclerotherapy (22.6%) were least likely to achieve the top rank.

Table 2 Summary of surface under the cumulative ranking curve scores for all outcomes in the overall population.

Treatment	Recurrence	Effectiveness	Satisfaction	Further treatment	Pain	Bleeding	Urinary retention	Septic complications	Anorectal stenosis	Incontinence	Constipation
MM	47.1	73.3	39.4	17.4	80.5	64.3	65.1	63.7	61.9	71.2	90.4
LigaH	19.3	98.2	39.1	4.3	82.5	60.5	75.4	62.7	48.2	70.4	64.6
FH	35.2	24.3	NA	71.8	54.2	54.7	61.6	84.9	98.6	94.6	NA
SH	63.9	60.6	66.5	39.6	59.9	60.9	56.3	71.2	52.9	68.8	47.2
MPSH	28.0	NA	NA	NA	NA	60.9	NA	NA	51.1	NA	NA
BPC	NA	NA	NA	NA	NA	95.1	79.8	27.4	NA	27.6	NA
MLPPH	NA	83.9	NA	NA	NA	NA	49.2	NA	NA	NA	NA
TST	20.9	69.7	61.9	24.6	29.8	54.2	40.6	NA	20.0	51.1	11.7
EPH	NA	61.9	NA	NA	63.4	NA	37.3	NA	64.9	NA	NA
HAL	80.4	36.2	46.8	58.8	59.6	37.4	57.6	76.5	27.1	48.9	21.1
RBL	90.5	16.3	29.9	73.3	36.9	41.5	49.6	23.6	NA	45.4	NA
IJS	76.4	22.6	30.2	69.5	13.5	58.4	NA	50.8	NA	NA	NA
SLM	72.4	42.0	57.7	54.8	55.5	45.2	40.6	29.1	41.3	8.3	64.9
IRC	69.0	9.4	NA	77.3	NA	18.3	NA	NA	NA	NA	NA
HarS	24.7	NA	78.5	28.7	75.9	26.6	35.6	26.8	86.6	24.9	NA
CHR	NA	30.2	NA	NA	NA	NA	NA	NA	NA	NA	NA
RFC	46.6	NA	NA	NA	NA	95.1	49.4	NA	59.8	88.4	NA
DCV	12.3	61.4	NA	99.0	0.4	17.1	68.5	44.8	NA	44.5	NA
SCH	49.8	NA	NA	NA	37.8	35.9	28.6	NA	4.7	15.3	NA
LASER	63.6	60.0	NA	30.8	NA	23.8	4.8	38.5	42.7	40.5	NA

NA indicates not available because of insufficient data. BPC: Bipolar coagulation; CHR: Combined hemorrhoidal radiocoagulation; DCV: Direct current electrotherapy; EPH: Eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid; FH: Ferguson hemorrhoidectomy; HAL: Hemorrhoidal artery ligation; HarS: Harmonic scalpel hemorrhoidectomy; IJS: Injection sclerotherapy; IRC: Infrared coagulation; LASER: Laser hemorrhoidectomy; LigaH: LigaSure™ hemorrhoidectomy; MLPPH: Modified ligation procedure for prolapsed hemorrhoids; MM: Milligan–Morgan hemorrhoidectomy; MPSH: Modified Park’s submucosal hemorrhoidectomy; RBL: Rubber band ligation; RFC: Radiofrequency coagulation; SCH: Semi-closed hemorrhoidectomy; SH: Stapled hemorrhoidopexy; SLM: Suture ligation with mucopexy; TST: Tissue-selecting technique.

Satisfaction: Satisfaction data were sparse, with no statistically significant differences between interventions (Supplementary Figure 3). Harmonic scalpel hemorrhoidectomy (SUCRA = 78.5%), stapled hemorrhoidopexy (66.5%), and the tissue-selecting technique (61.9%) ranked highest for satisfaction (Supplementary Figure 3; Table 2).

Further treatment rate: Direct current electrotherapy, Ferguson hemorrhoidectomy, hemorrhoidal artery ligation, infrared coagulation, rubber band ligation, and suture ligation with mucopexy were each associated with a significantly higher risk of requiring further treatment than Milligan–Morgan hemorrhoidectomy (Supplementary Figure 4). No other interventions differed significantly from the reference procedure for this outcome. The interventions with the lowest probabilities of requiring further treatment were LigaSure™ hemorrhoidectomy (SUCRA = 4.3%), Milligan–Morgan hemorrhoidectomy (17.4%), the tissue-selecting technique (24.6%), harmonic scalpel hemorrhoidectomy (28.7%), and laser hemorrhoidectomy (30.8%) (Supplementary Figure 4; Table 2).

Safety outcomes: The interventions with the highest probability of being associated with postoperative pain were LigaSure™ hemorrhoidectomy (SUCRA = 82.5%), Milligan–Morgan hemorrhoidectomy (80.5%), harmonic scalpel hemorrhoidectomy (75.9%), the eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid (63.4%), and stapled hemorrhoidopexy (59.9%). Conversely, the interventions with the lowest probability of pain were direct current electrotherapy (0.4%), injection sclerotherapy (13.5%), the tissue-selecting technique (29.8%), rubber band ligation (36.9%), and semi-closed hemorrhoidectomy (37.8%) (Supplementary Figure 5; Table 2).

The interventions with the highest probability of postoperative bleeding were bipolar coagulation (SUCRA = 95.1%), radiofrequency coagulation (95.1%), Milligan–Morgan hemorrhoidectomy (64.3%), modified Park’s submucosal hemorrhoidectomy (60.9%), and stapled hemorrhoidopexy (60.9%). Conversely, the interventions with the lowest probability of bleeding were direct current electrotherapy (17.1%), infrared coagulation (18.3%), laser hemorrhoidectomy (23.8%), harmonic scalpel hemorrhoidectomy (26.6%), and semi-closed hemorrhoidectomy (35.9%) (Supplementary Figure 6; Table 2).

The interventions with the highest probability of urinary retention were bipolar coagulation (SUCRA = 79.8%), LigaSure™ hemorrhoidectomy (75.4%), direct current electrotherapy (68.5%), Milligan–Morgan hemorrhoidectomy (65.1%), and Ferguson hemorrhoidectomy (61.6%). Conversely, the interventions with the lowest probability of urinary retention were laser hemorrhoidectomy (4.8%), semi-closed hemorrhoidectomy (28.6%), harmonic scalpel hemorrhoidectomy (35.6%), the eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid (37.3%), and the tissue-selecting technique (40.6%) (Supplementary Figure 7; Table 2).

The interventions with the highest probability of septic complications were Ferguson hemorrhoidectomy (SUCRA = 84.9%), hemorrhoidal artery ligation (76.5%), stapled hemorrhoidopexy (71.2%), Milligan–Morgan hemorrhoidectomy (63.7%), and LigaSure™ hemorrhoidectomy (62.7%). Conversely, the interventions with the lowest probability of septic complications were rubber band ligation (23.6%), harmonic scalpel hemorrhoidectomy (26.8%), bipolar coagulation (27.4%), suture ligation with mucopexy (29.1%), and laser hemorrhoidectomy (38.5%) (Supplementary Figure 8; Table 2).

The interventions with the highest probability of anorectal stenosis were Ferguson hemorrhoidectomy (SUCRA = 98.6%), harmonic scalpel hemorrhoidectomy (86.6%), Milligan–Morgan hemorrhoidectomy (61.9%), radiofrequency coagulation (59.8%), and the eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid (54.9%). Conversely, the interventions with the lowest probability of anorectal stenosis were semi-closed hemorrhoidectomy (4.7%), the tissue-selecting technique (20.0%), hemorrhoidal artery ligation (27.1%), suture ligation with mucopexy (41.3%), and laser hemorrhoidectomy (42.7%) (Supplementary Figure 9; Table 2).

The interventions with the highest probability of incontinence were Ferguson hemorrhoidectomy (SUCRA = 94.6%), radiofrequency coagulation (88.4%), Milligan–Morgan hemorrhoidectomy (71.2%), LigaSure™ hemorrhoidectomy (70.4%), and stapled hemorrhoidopexy (68.8%). Conversely, the interventions with the lowest probability of incontinence were suture ligation with mucopexy (8.3%), semi-closed hemorrhoidectomy (15.3%), harmonic scalpel hemorrhoidectomy (24.9%), bipolar coagulation (27.6%), and laser hemorrhoidectomy (40.5%) (Supplementary Figure 10; Table 2).

The number of comparisons reporting constipation was limited. The interventions with the highest probability of constipation were Milligan–Morgan hemorrhoidectomy (SUCRA = 90.4%), suture ligation with mucopexy (64.9%), and LigaSure™ hemorrhoidectomy (64.6%). Conversely, the interventions with the lowest probability of constipation were the tissue-selecting technique (11.7%), hemorrhoidal artery ligation (21.1%), and stapled hemorrhoidopexy (47.2%) (Supplementary Figure 11; Table 2).

Subgroup analysis

The number of studies available for analysis in the grade ≤ II subgroup was limited, largely because many trials enrolled patients with mixed grade II and III disease. Consequently, only 3 interventions—stapled hemorrhoidopexy, rubber band ligation, and laser hemorrhoidectomy—had sufficient data for recurrence analysis in this subgroup. These results should therefore be interpreted with caution (Supplementary Table 3).

In the grade ≤ III subgroup, the ranking of interventions for both efficacy and safety outcomes shifted substantially compared with the overall population (Supplementary Table 4). For recurrence, the SUCRA value for Milligan–Morgan hemorrhoidectomy decreased from 47.1% in the overall population to 20.6%, suggesting a relatively lower recurrence risk in patients with less severe disease; concurrently, its SUCRA value for effectiveness increased from 73.3% to 92.2%, indicating a more pronounced efficacy advantage in this subgroup. Rubber band ligation continued to show a high recurrence risk (SUCRA = 87.5%) but performed favorably with respect to pain control (29.4%) and incontinence risk (16.1%), and it had a low rate of further treatment requirement (78.6%). Of note, stapled hemorrhoidopexy showed increased SUCRA values for anorectal stenosis (from 52.9% to 88.5%) and pain (from 59.9% to 78.2%) in this subgroup. Injection sclerotherapy demonstrated excellent pain control (SUCRA = 3.9%), and laser hemorrhoidectomy showed improved recurrence (46.4%) and effectiveness (79.4%) rankings compared with the overall population, suggesting potential promise in this patient subset. However, data for several outcomes were missing for multiple minimally invasive procedures—including LigaSure™ hemorrhoidectomy and the tissue-selecting technique—in the grade ≤ III subgroup, which limited the ability to comprehensively compare optimal strategies for patients with mild-to-moderate disease.

In the subgroup of patients with grade ≥ II hemorrhoids, the SUCRA rankings for various interventions also differed from those in the primary analysis (Supplementary Table 5). In terms of overall performance, LigaSure™ hemorrhoidectomy ranked favorably for both effectiveness (SUCRA = 76.0%) and satisfaction (98.9%), while maintaining a low recurrence risk (14.6%). Milligan–Morgan hemorrhoidectomy performed well in terms of effectiveness (83.6%). Of note, rubber band ligation continued to show a high recurrence risk (80.8%) in patients with grade ≥ II disease, and injection sclerotherapy exhibited the highest recurrence risk (84.5%) and the lowest effectiveness (11.3%), suggesting that these 2 approaches should be used with caution in this population. The tissue-selecting technique demonstrated excellent urinary retention control (SUCRA = 3.5%) and may serve as an alternative for patients at risk of urinary retention. Hemorrhoidal artery ligation did not show a clear efficacy advantage (44.2%) in patients with more severe disease, being outperformed by both LigaSure™ hemorrhoidectomy and Milligan–Morgan hemorrhoidectomy. Compared with the overall population, both LigaSure™ hemorrhoidectomy and Milligan–Morgan hemorrhoidectomy showed superior or more stable efficacy and safety profiles in patients with grade ≥ II disease, whereas injection sclerotherapy and rubber band ligation had more prominent shortcomings in recurrence control in this more severely affected group.

In the subgroup of patients with grade ≥ III hemorrhoids (Supplementary Table 6), the tissue-selecting technique (SUCRA = 11.2%) and LigaSure™ hemorrhoidectomy (22.1%) were associated with the lowest recurrence risks, whereas hemorrhoidal artery ligation (90.9%), suture ligation with mucopexy (77.3%), and rubber band ligation (73.4%) had relatively higher recurrence risks. For effectiveness, LigaSure™ hemorrhoidectomy maintained the highest ranking (96.7%), followed by the tissue-selecting technique (66.5%) and the eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid (56.6%). With respect to pain, Milligan–Morgan hemorrhoidectomy (89.2%) and the eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid (71.4%) were associated with a higher probability of pain, whereas hemorrhoidal artery ligation (18.6%) and the tissue-selecting technique (38.3%) were associated with a lower probability. For bleeding, laser hemorrhoidectomy (0.4%) had the lowest risk, whereas radiofrequency coagulation (94.7%) and bipolar coagulation (95.9%) had the highest risks. For urinary retention, laser hemorrhoidectomy (1.0%) had the lowest risk, whereas bipolar coagulation (80.0%) and LigaSure™ hemorrhoidectomy (78.3%) had the highest risks. For anorectal stenosis, semi-closed hemorrhoidectomy (3.2%) and the tissue-selecting technique (16.2%) had the lowest risks, whereas Ferguson hemorrhoidectomy (98.8%) and harmonic scalpel hemorrhoidectomy (81.8%) had the highest risks. For incontinence, suture ligation with mucopexy (27.8%) and semi-closed hemorrhoidectomy (14.6%) had lower risks, whereas stapled hemorrhoidopexy (75.0%) and LigaSure™ hemorrhoidectomy (73.9%) had higher risks. For constipation, the tissue-selecting technique (27.3%) had the lowest risk, whereas Milligan–Morgan hemorrhoidectomy (89.3%) had the highest risk.

Sensitivity analysis

To assess the robustness of the primary findings, we performed 2 sensitivity analyses for recurrence. First, we excluded the 5 studies rated as having an overall high risk of bias and repeated the network meta-analysis (Supplementary Figure 12). Second, we restricted the analysis to studies rated as having low risk in the randomization process domain of the RoB 2.0 tool (Supplementary Figure 13). In both sensitivity analyses, the SUCRA rankings and the relative ordering of interventions remained largely consistent with those of the primary analysis. These results confirm that our primary findings were not materially influenced by studies with methodological concerns, supporting the overall stability and reliability of the network meta-analysis estimates.

Meta-regression analysis

To further address clinical heterogeneity across studies, we performed a network meta-regression analysis adjusting for publication year, grade ≥ III disease, and sample size (Supplementary Figure 14). The adjusted results remained largely consistent with the primary analysis: LigaSure™ hemorrhoidectomy (adjusted OR = 0.34; 95%CI: 0.10–0.91) and the tissue-selecting technique (adjusted OR = 0.31; 95%CI: 0.12–0.73) continued to show significantly lower recurrence risks compared with Milligan–Morgan hemorrhoidectomy, whereas rubber band ligation (adjusted OR = 5.71; 95%CI: 3.49–9.41) and stapled hemorrhoidopexy (adjusted OR = 1.93; 95%CI: 1.51–2.48) maintained significantly higher risks. Although some effect estimates were slightly attenuated after adjustment, the direction and overall ranking of interventions remained substantially unchanged.

Multi-criteria decision analysis

The multi-criteria decision analysis results are presented in Supplementary Figure 15. The 2-dimensional scatter plot further illustrated the distribution of each intervention across efficacy and safety dimensions. LigaSure™ hemorrhoidectomy and Milligan–Morgan hemorrhoidectomy were positioned in the upper-left quadrant, indicating high efficacy but with a safety burden. The tissue-selecting technique and harmonic scalpel hemorrhoidectomy were located in the upper-right quadrant. By contrast, rubber band ligation and injection sclerotherapy clustered in the lower region, characterized by acceptable safety but markedly inferior efficacy. Among the treatment modalities evaluated, LigaSure™ hemorrhoidectomy achieved the highest composite utility score (69.9%), followed by Milligan–Morgan hemorrhoidectomy (65.1%) and the tissue-selecting technique (60.7%).

DISCUSSION

To our knowledge, this network meta-analysis represents the most comprehensive synthesis of evidence on hemorrhoidal disease interventions to date, encompassing 66 RCTs, 8549 patients, and 20 distinct procedural modalities identified through a systematic search spanning 2010 to 2026. Our investigation advances the field in several important ways that distinguish it from previous meta-analyses[9,79-81]. The scope of interventions covered in our network substantially exceeds that of earlier syntheses. The classic network meta-analysis by Simillis et al[79], published in 2015, included 11 surgical procedures, and the analysis by Du et al[80] in 2019 included 14 modalities; however, neither incorporated the tissue-selecting technique, which has been widely adopted in China and other regions in recent years. Our analysis advances the evidence base in 3 key respects. First, we expanded the intervention network to include not only the tissue-selecting technique but also emerging and modified procedures—specifically, the eastern rubber band ligation and coincident of dislocation for prolapse and hemorrhoid, as well as the modified ligation procedure for prolapsed hemorrhoids—thereby addressing important gaps left by earlier meta-analyses that excluded these novel modalities. Second, our study is the first, to our knowledge, to perform systematic subgroup analyses stratified by Goligher grade. While prior network meta-analyses have typically pooled patients across severity levels despite the fundamentally different therapeutic rationales for grade II vs grade III–IV disease, our stratified approach revealed that the relative performance of interventions can shift considerably according to disease severity. This finding underscores the risk of obscuring true differential treatment effects when pooling across grades. Third, we extended the scope of outcome assessment beyond the conventional endpoints—recurrence, pain, bleeding, urinary retention, anorectal stenosis, and incontinence—to incorporate effectiveness, patient satisfaction, and further treatment rate. This broader analytic framework provides a more clinically nuanced and patient-centered perspective than previous syntheses, which have predominantly focused on recurrence and complication profiles.

Across multiple core efficacy outcomes, LigaSure™ hemorrhoidectomy emerged as one of the consistently highest-ranked interventions. For recurrence, it demonstrated a statistically significant risk reduction compared with Milligan–Morgan hemorrhoidectomy and achieved the second-highest SUCRA ranking, surpassed only by direct current electrotherapy—an intervention for which the evidence base was notably limited, warranting cautious interpretation. In terms of effectiveness, LigaSure™ hemorrhoidectomy attained the top rank, and it also ranked first for the lowest rate of further treatment. These consistent findings carry meaningful clinical implications, particularly given the procedure’s favorable balance of efficacy and safety. The earlier network meta-analysis by Simillis et al[79] concluded that energy-based platforms (LigaSure™ and Harmonic) were superior to conventional excisional techniques in reducing postoperative pain and shortening hospital stay but showed no significant difference from conventional surgery in terms of recurrence rates. Our study, based on a broader and more contemporary dataset, further revealed the efficacy advantage of LigaSure™ hemorrhoidectomy over Milligan–Morgan hemorrhoidectomy, suggesting that this procedure may be challenging the established notion that conventional excisional hemorrhoidectomy confers the lowest recurrence risk. However, the advantages of LigaSure™ hemorrhoidectomy were accompanied by several safety concerns. Our analysis showed that postoperative pain and urinary retention remain important complications requiring active management with this technique. This finding appears to diverge somewhat from earlier reports suggesting that LigaSure™ hemorrhoidectomy reduces postoperative pain[79]. Several factors may contribute to this apparent discrepancy. First, our study employed a binary definition of postoperative pain, whereas the majority of previous investigations have used a continuous visual analog scale. These 2 measurement approaches differ fundamentally in their sensitivity to detect pain variations and in the threshold at which pain is considered clinically meaningful, potentially influencing the comparability of findings across studies[82,83]. Second, the indications for LigaSure™ hemorrhoidectomy may have expanded in recent years to include more complex hemorrhoidal cases[84,85], potentially leading to higher reported event rates for pain. Thus, when considering LigaSure™ hemorrhoidectomy, clinicians should carefully weigh its efficacy benefits against the need for robust postoperative pain management and urinary retention prophylaxis.

The tissue-selecting technique is a modified stapled procedure based on the principle of the procedure for prolapse and hemorrhoids. Its core advantage lies in selectively resecting prolapsed mucosa while preserving normal mucosal bridges, thereby reducing the risk of anorectal stenosis and other complications associated with circumferential stapling[77,86]. Our study showed that the tissue-selecting technique was associated with a nearly significant reduction in recurrence risk compared with Milligan–Morgan hemorrhoidectomy, ranking third in SUCRA for recurrence. In terms of safety, the tissue-selecting technique performed particularly well: It ranked second lowest for anorectal stenosis, behind only semi-closed hemorrhoidectomy, and had the lowest ranking for constipation, with low rankings also observed for pain and urinary retention. These findings are highly consistent with the conclusions of Zhang et al[87], indicating that the tissue-selecting technique was superior with respect to urinary retention, fecal incontinence, anorectal stenosis, and recurrence. Our study extends this evidence by further demonstrating, through subgroup analysis, that the recurrence risk for the tissue-selecting technique decreased in the grade ≥ III subgroup, suggesting that its recurrence control may be particularly pronounced in patients with severe disease. For patients with grade III–IV hemorrhoids, especially those who place a high value on postoperative anal function preservation, the tissue-selecting technique may represent an attractive option that balances low recurrence with good safety, although further high-quality studies are needed to confirm this observation. Given that the tissue-selecting technique is relatively easy to perform, with a short operative time and rapid recovery, it is well positioned for broader adoption across more medical centers[40,88].

Rubber band ligation and injection sclerotherapy are among the most commonly employed office-based minimally invasive interventions for hemorrhoidal disease, both domestically and internationally. Their widespread adoption is attributable to several practical advantages, including technical simplicity, low procedural cost, and the absence of anesthesia requirements[89,90]. Jin et al[91] found that rubber band ligation and injection sclerotherapy were associated with significantly higher recurrence rates than excisional procedures, albeit with less postoperative pain and fewer complications. Our study extends this evidence across a broader spectrum of disease severity and, through subgroup analyses, provides preliminary insights that, while exploratory, may help inform treatment selection across different severity strata. Rubber band ligation was found to be associated with a significantly higher recurrence risk than Milligan–Morgan hemorrhoidectomy in the overall population, and injection sclerotherapy was also associated with a high recurrence risk; both interventions ranked lowest for effectiveness. These patterns were even more pronounced in the grade ≥ II and grade ≥ III subgroups, where recurrence risks were further elevated and effectiveness rankings declined. These findings indicate that the utility of rubber band ligation and injection sclerotherapy as definitive treatments for grade II or higher disease is limited, with high recurrence and low effectiveness rates potentially necessitating additional interventions and increasing healthcare costs. Nevertheless, the clinical value of these procedures should not be dismissed. They remain appropriate as first-line options for mild grade I–II disease, as therapeutic alternatives for patients with severe hemorrhoids who cannot tolerate surgery, and as temporary hemostatic measures for acute bleeding. For patients with grade II or higher disease who desire curative treatment and can tolerate surgery, more effective procedures—including LigaSure™ hemorrhoidectomy and the tissue-selecting technique—should be prioritized. However, it should be noted that follow-up durations varied considerably across the included trials, with office-based procedures such as rubber band ligation and sclerotherapy generally having shorter follow-up periods than surgical interventions. This disparity may influence the detection and reporting of late recurrences, and caution is warranted when comparing recurrence rates between procedures with inherently different follow-up trajectories.

Based on the above findings, a stratified, grade-based strategy may be considered, although the current subgroup evidence remains exploratory. For patients with grade I–II disease primarily presenting with bleeding and no significant prolapse, rubber band ligation or injection sclerotherapy may be appropriate first-line choices, provided patients are adequately counseled about the higher recurrence risk and possible need for further treatment. For patients with grade II–III disease who have curative treatment goals and are able to tolerate surgery, LigaSure™ hemorrhoidectomy and the tissue-selecting technique may be preferred based on the available evidence. LigaSure™ hemorrhoidectomy currently appears to offer the best overall efficacy and recurrence control, while the tissue-selecting technique provides distinctive safety advantages—particularly for anorectal stenosis, constipation, and pain. These options may facilitate individualized treatment selection based on disease characteristics and patient preferences. For grade III–IV disease, circumferential prolapse, or recurrent hemorrhoids, Milligan–Morgan and Ferguson hemorrhoidectomies remain dependable choices, though rigorous perioperative pain control and complication prevention are essential. However, definitive recommendations await confirmation from future adequately powered studies.

Several limitations of this study should be acknowledged. First, follow-up durations varied considerably across the included trials, with office-based procedures generally having shorter follow-up periods than surgical interventions, which may influence the detection and reporting of late recurrence events. Second, the definitions of secondary outcomes such as effectiveness and satisfaction were not standardized across the original studies and were subject to the investigators’ own criteria, potentially introducing heterogeneity in outcome measurement. Third, for certain interventions with limited direct evidence—such as direct current electrotherapy—the SUCRA rankings should be interpreted with caution because these estimates were derived from a small number of studies with small sample sizes. It should also be noted that anesthesia protocols varied across the included studies, which may have influenced early postoperative pain and urinary retention rates, and this should be considered when interpreting the corresponding safety outcomes. Furthermore, technical variations within the same procedure may exist across studies and could theoretically affect the precision of effect estimates. This issue is inherent to surgical meta-analyses and should be considered when interpreting our findings.

CONCLUSION

In this network meta-analysis—one of the most comprehensive procedural syntheses to date—20 therapeutic modalities for hemorrhoidal disease were systematically compared with respect to efficacy and safety. The findings demonstrate that different interventions exhibit distinct profiles across recurrence control, symptom relief, and complication risks, reinforcing the need for individualized clinical decision-making that balances efficacy against safety. With regard to efficacy, LigaSure™ hemorrhoidectomy emerged as the most consistently advantageous procedure, ranking highest or among the highest for recurrence control, effectiveness, and reduction of further treatment requirements. The tissue-selecting technique offered a compelling minimally invasive alternative, characterized by low recurrence risk and an excellent safety profile. Although direct current electrotherapy showed a favorable SUCRA ranking for recurrence, this result was derived from a very limited number of studies with small sample sizes and should therefore be interpreted with caution. With regard to safety, office-based procedures such as rubber band ligation and injection sclerotherapy were associated with lower risks of postoperative pain and complications. However, the substantially higher recurrence rates observed with these procedures compared with excisional techniques should prompt caution in patients with grade II or higher disease. Subgroup analyses provided preliminary evidence suggesting that the performance of individual procedures may vary according to Goligher grade, although these findings are exploratory and require confirmation in future studies specifically designed to compare treatments across severity strata. Further high-quality, head-to-head RCTs are warranted to strengthen the evidence base and to guide the refinement of clinical practice algorithms.