Comparative effectiveness and safety of biliary drainage strategies for malignant biliary obstruction: a protocol for systematic review and network meta-analysis Authors: Lijuan Zhang, Qian Xiao, Mei Zhou, Liqun Zou Published: 2026 ABSTRACT Abstract AbstractIntroductionMalignant biliary obstruction (MBO) is a common complication of pancreaticobiliary and periampullary malignancies and is associated with substantial morbidity and reduced survival. Multiple biliary drainage strategies are available, including endoscopic, percutaneous and surgical approaches, as well as adjunctive therapies. However, direct head-to-head comparisons remain limited, and the relative effectiveness and safety of these interventions are not well established. This study aims to compare the effectiveness and safety of different biliary drainage strategies for MBO using a network meta-analysis, allowing both direct and indirect comparisons and enabling ranking of treatment strategies.Methods and analysisThis protocol is reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guidelines. We will systematically search MEDLINE via PubMed, Embase via Elsevier, Web of Science Core Collection via Clarivate, and the Cochrane Central Register of Controlled Trials via the Cochrane Library from inception to 30 June 2026. Randomised controlled trials (RCTs) and non-randomised comparative studies will be included. Interventions of interest include endoscopic retrograde cholangiopancreatography (ERCP), endoscopic ultrasound-guided biliary drainage (EUS-BD), percutaneous transhepatic biliary drainage (PTBD), surgical bypass and adjunctive therapies combined with drainage (eg, photodynamic therapy or radiofrequency ablation). Primary outcomes will include technical success, clinical success and quality of life. Secondary outcomes will include adverse events, reintervention rates, overall survival, 30-day all-cause mortality, stent patency duration and length of hospital stay. Risk of bias will be assessed using the Cochrane Risk of Bias two tool for RCTs and the ROBINS-I tool for non-randomised studies. A random-effects network meta-analysis will be conducted within a frequentist framework. Consistency and transitivity assumptions will be assessed, and treatment ranking will be estimated using the surface under the cumulative ranking curve. Subgroup analyses will be performed according to clinically relevant factors, including obstruction site, stent type, tumour type and treatment setting. The certainty of evidence will be evaluated using the GRADE approach adapted for network meta-analysis.Ethics and disseminationEthical approval is not required, as this study is based on published data. Findings will be disseminated through peer-reviewed publication.PROSPERO registration numberCRD420261373097Keywords: Systematic Review, Network Meta-Analysis, Endoscopic surgery, Hepatobiliary tumours STRENGTHS AND LIMITATIONS OF THIS STUDY This protocol follows Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols guidance and has been prospectively registered in PROSPERO. This protocol follows Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols guidance and has been prospectively registered in PROSPERO. Network meta-analysis will allow simultaneous comparison and ranking of multiple biliary drainage strategies using both direct and indirect evidence. Network meta-analysis will allow simultaneous comparison and ranking of multiple biliary drainage strategies using both direct and indirect evidence. The inclusion of both randomised and non-randomised comparative studies may improve evidence coverage, particularly for comparisons for which randomised evidence is limited. The inclusion of both randomised and non-randomised comparative studies may improve evidence coverage, particularly for comparisons for which randomised evidence is limited. The primary network meta-analysis will be restricted to randomised controlled trials, while non-randomised comparative studies will be synthesised separately in secondary analyses. The primary network meta-analysis will be restricted to randomised controlled trials, while non-randomised comparative studies will be synthesised separately in secondary analyses. Although non-randomised evidence may improve the comprehensiveness and applicability of the review, it may also be affected by confounding, selection bias and other sources of bias. Although non-randomised evidence may improve the comprehensiveness and applicability of the review, it may also be affected by confounding, selection bias and other sources of bias. Introduction Malignant biliary obstruction (MBO) is a common and life-threatening complication of pancreaticobiliary and periampullary malignancies, leading to substantial morbidity, impaired quality of life and reduced survival.1 2 Effective biliary drainage is essential not only for symptom relief but also for preventing cholangitis and facilitating subsequent oncological treatment. However, selecting the optimal drainage strategy remains challenging in clinical practice.3 4 Several biliary drainage approaches are currently available, including endoscopic retrograde cholangiopancreatography (ERCP),5–7 endoscopic ultrasound-guided biliary drainage (EUS-BD),8–11 percutaneous transhepatic biliary drainage (PTBD)12–14 and surgical bypass.15 16 Although ERCP is widely regarded as the first-line treatment, its success may be limited in patients with altered anatomy or advanced tumour infiltration. Alternative strategies, such as EUS-BD and PTBD, are increasingly used, but their comparative effectiveness and safety remain uncertain. In addition, adjunctive therapies, including photodynamic therapy17 and radiofrequency ablation,18 19 have been introduced to improve stent patency and clinical outcomes, further adding to the complexity of treatment selection. Despite the availability of multiple treatment options, current evidence is insufficient to support clear treatment selection.20 21 Most existing studies rely on pairwise comparisons, often involving small sample sizes and heterogeneous patient populations. Furthermore, clinically relevant factors, such as obstruction site, stent type, tumour type and treatment setting, vary across studies and may influence treatment outcomes, yet have not been evaluated within a unified analytical framework. As a result, clinical decision-making often relies on local expertise rather than robust comparative evidence. Network meta-analysis (NMA) provides a robust analytical framework that enables simultaneous comparison of multiple interventions by integrating both direct and indirect evidence.22 This approach allows estimation of relative treatment effects and ranking of available strategies, which is particularly valuable in complex clinical settings such as MBO, where multiple competing interventions exist but direct comparisons are limited. A previous network meta-analysis compared ERCP, PTBD and EUS-BD across 17 studies identified up to October 2019.23 That review focused on these three broad intervention categories and primarily evaluated technical success, clinical success, adverse events and reintervention.23 Building on this work, the present systematic review and network meta-analysis aims to update the evidence through June 2026 and compare the relative effectiveness and safety of a broader range of biliary drainage strategies for MBO, including surgical bypass and drainage combined with adjunctive local therapies. It will also evaluate additional patient-important outcomes, including quality of life, overall survival, 30-day all-cause mortality, stent patency duration and length of hospital stay, and assess whether the comparative effectiveness and safety of biliary drainage strategies differ according to obstruction site, stent type, tumour type and treatment setting. RCTs and non-randomised comparative studies will be synthesised separately. Methods Patient and public involvement Patients and the public were not involved in the design, conduct, reporting or dissemination of this study. Study protocol and registration This protocol is reported in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) guidelines24 and has been registered in PROSPERO (CRD420261373097). This study is planned to be conducted from April 2026 to September 2026. The literature search will be performed from database inception to 30 June 2026, followed by study selection, data extraction, risk-of-bias assessment, data synthesis and manuscript preparation. Eligibility criteria Participants We will include studies involving adults (≥18 years) with MBO confirmed by imaging, histopathology, or clinical diagnosis, caused by pancreaticobiliary or periampullary malignancies. Interventions and comparators Eligible studies will evaluate biliary drainage strategies, including ERCP, EUS-BD, PTBD, surgical bypass and adjunctive therapies combined with drainage (eg, photodynamic therapy or radiofrequency ablation). Comparisons among any of these interventions will be considered to enable construction of a connected treatment network. Outcomes Studies must report at least one of the following outcomes. Primary outcomes (1) Technical success will be defined as successful placement of the drainage device; (2) clinical success will be defined according to the definitions used in the included studies, most commonly resolution of biliary obstruction or a reduction in serum bilirubin levels after drainage. We will extract the exact bilirubin threshold and timing of assessment used in each study. When definitions differ across studies, clinical success will be analysed according to the study-reported definition, and variations in bilirubin reduction thresholds and assessment time points will be explored through subgroup or sensitivity analyses where sufficient data are available. (3) Quality of life will be assessed using validated measurement instruments. Secondary outcomes (1) Adverse events will be defined as procedure-related complications, including cholangitis, pancreatitis, infection, bleeding and perforation; (2) reintervention will be defined as any additional procedure required due to treatment failure or complications; (3) overall survival will be defined as the time from intervention to death from any cause; (4) 30-day all-cause mortality will be defined as death from any cause within 30 days after the procedure; (5) stent patency duration will be defined as the time from stent placement to stent dysfunction or occlusion; (6) length of hospital stay will be defined as the duration of hospitalisation associated with the index procedure. Study design Randomised controlled trials (RCTs) and non-randomised comparative studies will be included. Eligible non-randomised studies will include prospective and retrospective cohort studies and other comparative studies evaluating at least two eligible biliary drainage strategies. Case reports, case series without a comparator, reviews, editorials, letters, conference abstracts without sufficient data and animal studies will be excluded. Although RCTs generally provide the most internally valid estimates of comparative treatment effects, randomised evidence may remain limited or unavailable for some eligible interventions and clinically distinct treatment settings in MBO. Non-randomised comparative studies may therefore provide complementary evidence on interventions and settings that are underrepresented in RCTs, as well as additional information on longer-term outcomes and routine clinical practice. Their inclusion may improve the comprehensiveness and applicability of the review. However, because non-randomised studies are more susceptible to confounding and selection bias, RCTs and non-randomised comparative studies will be synthesised separately. Search strategy A comprehensive search will be conducted from database inception to 30 June 2026 in MEDLINE via PubMed, Embase via Elsevier, Web of Science Core Collection via Clarivate and the Cochrane Central Register of Controlled Trials via the Cochrane Library. No language restrictions will be applied. No study design filter will be applied at the search stage to avoid excluding potentially eligible non-randomised comparative studies. The search strategy will combine Medical Subject Headings (MeSH) and free-text terms related to MBO and biliary drainage techniques. Reference lists of included studies and relevant reviews will also be screened to identify additional eligible studies. The full search strategies for all databases, including search terms, Boolean operators and filters, are provided in online supplemental file 1. Study selection All retrieved records will be imported into EndNote X9, and duplicates will be removed. Two reviewers (LJZ and QX) will independently screen titles and abstracts, followed by full-text assessment of potentially eligible studies (figure 1). Disagreements will be resolved through discussion or consultation with a third reviewer. PRISMA flow diagram of study selection. PRISMA, Preferred Reporting Items for Systematic Review and Meta-Analysis. Data extraction and management Two reviewers (LJZ and QX) will independently extract data using a standardised, predesigned data extraction form. Extracted information will include study characteristics, patient characteristics, intervention details and outcome data. For clinical success, the exact definition, bilirubin reduction threshold and timing of outcome assessment will be extracted from each study. For multi-arm studies, all relevant intervention arms will be included to preserve the network structure. Where necessary, corresponding authors will be contacted to obtain missing or unclear data. Any discrepancies will be resolved through discussion or consultation with a third reviewer. Risk of bias assessment The risk of bias of included studies will be assessed independently by two reviewers (QX and MZ). For RCTs, the Cochrane Risk of Bias two tool25 will be applied at the level of the specific outcome result. For non-randomised comparative studies, ROBINS-I26 will similarly be applied at the level of the specific outcome result. Risk-of-bias assessments will be conducted for the following prespecified key outcomes: technical success, clinical success, quality of life, adverse events, reintervention, overall survival and stent patency duration. These outcomes comprise all primary outcomes and the key secondary outcomes. Any discrepancies will be resolved through discussion or consultation with a third reviewer. Statistical analysis All statistical analyses will be performed using R software (R Foundation for Statistical Computing, Vienna, Austria). Pairwise meta-analyses will be conducted using the meta package, and frequentist network meta-analyses will be conducted using the netmeta package. For each outcome, the primary network meta-analysis will be restricted to RCTs. A random-effects model will be used to account for anticipated clinical and methodological heterogeneity. A network meta-analysis will be performed only when the RCT evidence forms a connected network and the transitivity assumption is considered plausible. When a connected RCT network cannot be constructed, the available RCT evidence will be synthesised using pairwise meta-analysis or presented narratively, as appropriate. Non-randomised comparative studies will be synthesised separately in secondary analyses and will not be pooled with RCTs within the same treatment network. Depending on the availability, clinical comparability and connectivity of the evidence, non-randomised studies will be synthesised using pairwise meta-analysis, a separate network meta-analysis or narrative synthesis. When both adjusted and unadjusted effect estimates are reported, appropriately adjusted estimates will be preferred. For dichotomous outcomes, treatment effects will be expressed as ORs with 95% CIs. For time-to-event outcomes, including overall survival and stent patency duration, HRs with 95% CIs will be used as the preferred effect measure. If HRs are not reported directly, they will be extracted or estimated, where feasible, from available information such as Kaplan–Meier curves, log-rank statistics, P values, CIs or numbers at risk. When HRs cannot be obtained reliably, the corresponding outcomes will be synthesised narratively rather than pooling medians as equivalent treatment-effect estimates. For continuous outcomes measured using the same scale and reported in the same unit across studies, mean differences with 95% CIs will be used. Standardised mean differences, calculated as Hedges’ g, will be used when studies measure the same outcome domain using different scales. The geometry of each treatment network will be presented using network plots. Before conducting a network meta-analysis, the transitivity assumption will be assessed by comparing the distributions of prespecified potential effect modifiers across treatment comparisons. These effect modifiers will include obstruction site, tumour type, disease stage, treatment setting, stent type, drainage route, baseline bilirubin level and performance status. Their distributions will be summarised by treatment comparison in tables and assessed for important clinical or methodological imbalance. Transitivity will be considered separately for each outcome because the relevance and distribution of effect modifiers may differ across outcomes. If a substantial imbalance in potential effect modifiers is identified, its possible impact will be explored using subgroup analysis or network meta-regression where sufficient data are available. Comparisons considered clinically or methodologically incompatible will not be included in the same network and will instead be synthesised separately or described narratively. Statistical inconsistency between direct and indirect evidence will be assessed using a global design-by-treatment interaction approach and local node-splitting analyses. Both the magnitude and uncertainty of inconsistency estimates will be considered, rather than relying solely on statistical significance. Treatment ranking will be estimated using the surface under the cumulative ranking curve and will be interpreted alongside the corresponding relative-effect estimates, their uncertainty and the certainty of evidence. Where sufficient data are available, subgroup analyses will be conducted according to obstruction site (distal vs hilar/proximal), stent type (metal vs plastic), tumour type (pancreatic vs biliary or periampullary malignancies), treatment setting (primary drainage vs rescue drainage following failed ERCP), bilirubin reduction threshold and timing of clinical-success assessment. Sensitivity analyses for the RCT evidence will include leave-one-out analyses and exclusion of studies judged to be at high risk of bias. For non-randomised evidence, sensitivity analyses will include exclusion of studies judged to be at serious or critical risk of bias and restriction to studies reporting adjusted effect estimates. When at least 10 studies contribute to an eligible pairwise meta-analysis, funnel plots and effect-measure-appropriate tests for funnel-plot asymmetry will be considered. For network meta-analyses with a sufficient number of studies and a clinically meaningful ordering of interventions, comparison-adjusted funnel plots will be used to explore possible small-study effects. Funnel-plot asymmetry will not be regarded as proof of publication bias. The certainty of evidence will be evaluated using the GRADE approach adapted for network meta-analysis.27 Discussion This study will synthesise the available evidence on the comparative effectiveness and safety of biliary drainage strategies for MBO. Given the range of available interventions and the limited direct comparisons between them, a network meta-analysis provides a practical approach to integrate existing evidence and establish a treatment hierarchy. A key strength of this study is the application of network meta-analysis, which enables simultaneous comparison of multiple interventions within a single analytical framework. In addition, the inclusion of clinically relevant factors, such as obstruction site, stent type, and tumour type, may enhance the clinical interpretability of the findings and support more individualised treatment decisions. This study will include both randomised and non-randomised comparative studies to improve evidence coverage and permit the evaluation of comparisons that may not be represented in the RCT network. RCTs and non-randomised studies will be synthesised in separate evidence networks where feasible, with primary conclusions and treatment rankings based on the RCT evidence. Risk of bias will be assessed using design-specific tools, and findings from non-randomised evidence will be interpreted cautiously in view of the potential for confounding and selection bias. This approach is intended to balance methodological rigour with evidence comprehensiveness. Several limitations should be considered. Differences in patient populations, tumour characteristics and procedural techniques may introduce heterogeneity and affect the transitivity assumption required for network meta-analysis. The inclusion of non-randomised comparative studies may improve evidence comprehensiveness and network connectivity, but it may also introduce confounding, selection bias and imbalance in baseline characteristics; therefore, findings driven mainly by non-randomised evidence will be interpreted with caution. Variability in outcome definitions, particularly the bilirubin reduction threshold and timing used to define clinical success, may also contribute to uncertainty. In addition, some comparisons may be informed by a limited number of studies, resulting in sparse networks and reduced precision. Despite these limitations, this study is likely to provide clinically relevant insights into the relative effectiveness and safety of different biliary drainage strategies. The findings may support more informed clinical decision-making and contribute to the standardisation of comparative evaluation in this field. Supplementary material Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2026-121406) Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. References Supplementary Materials