Systematic review and network meta-analysis: comparison of different techniques for difficult biliary cannulation Authors: Yen-Ying Wu, Sz-Iuan Shiu, Chun-Fang Tung, Hsu-En Cheng, Chung-Wang Ko Published: 2026 ABSTRACT AbstractBackground and aimDifficult biliary cannulation (DBC) prolongs endoscopic retrograde cholangiopancreatography (ERCP) manipulation and increases post-ERCP complications. Several cannulation techniques, including guidewire and cutting-based methods, have been proposed to facilitate the procedure; however, the optimum choice among these approaches remains inconclusive. We conducted a network meta-analysis to compare the techniques for DBC.MethodsThree major bibliographic databases were reviewed for enrollment of comparative trials which took place prior to January 23, 2024. We included adults diagnosed with DBC and compared two or more of seven possible interventions, including controlled group, precut papillotomy (PP), precut fistulotomy (PF), delayed precut papillotomy (dPP), delayed precut fistulotomy (dPF), transpancreatic biliary sphincterotomy (TPS), and the double-guidewire technique (DGW). Our focus was on the deep cannulation rate and any post-ERCP complications, including pancreatitis, bleeding, perforation and infection which may have occurred during the follow-up periods.ResultsSeventeen randomised controlled trials involving a total of 2,189 participants were used in the analyses. When compared with the control group, TPS was shown to be significantly superior to the other techniques in cannulation efficacy (OR 6.00, 95% CI: 1.29 ~ 27.96), while only PF was found to be significantly associated with a lower complication rate (OR 0.61, 95% CI: 0.40 ~ 0.93). Subgroup analysis also found that patients receiving PF experienced significantly less post-ERCP pancreatitis in comparison to the control group, with no differences in being seen in bleeding, perforation or infection after ERCP.ConclusionsFor DBC patients, only TPS achieved better cannulation efficacy, while PF showed a lower complication rate with statistical significance, particularly in post-ERCP pancreatitis.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12876-026-04849-8.Keywords: Difficult biliary cannulation, Double-guidewire technique, Network meta-analysis, Precut technique, Transpancreatic biliary sphincterotomy Introduction Endoscopic retrograde cholangiopancreatography (ERCP) is the main diagnostic and therapeutic procedure for pancreaticobiliary disease. Successful biliary cannulation has been achieved in more than 90% of the time [1], yet the procedure still remains technically challenging, with an increasing failure rate of up to 25% if trainees are involved in the cannulation. There are several obstacles encountered during ERCP, including variations in the size, shape and position of the ampulla of Vater, the presence of stones or strictures within the bile ducts, prior surgical interventions, congenital anomalies, as well as technical factors [2]. The latest European Society of Gastrointestinal Endoscopy (ESGE) Guidelines have defined difficult biliary cannulation (DBC) as being based upon the presence of more than 5 min or 5 cannulation attempts [3], whereas international consensus [4] has defined failure of selective biliary cannulation as being based upon it occurring within 10 min or within 5 cannulation attempts, with some reservations being felt after panel discussions. DBC not only prolongs the procedure time but also increases the risk of complications, such as post-ERCP pancreatitis (PEP), perforation or bleeding. As a result, it is crucial to have effective strategies in place to overcome these challenges. A number of innovative insertion techniques have been developed to assist with DBC, including cutting-based techniques (precut papillotomy (PP), precut fistulotomy (PF), and transpancreatic biliary sphincterotomy (TPS)), guidewire-assisted cannulation, and pancreatic stent-guided biliary cannulation. The needle-knife precut techniques can start either from the papillary orifice or the roof of the ampulla of Vater via freehand, or can be applied after pancreatic guidewire or stent insertion. Limited studies have shown better outcomes when using PF in terms of both efficacy and PEP when compared to PP [5], with no difference of PEP being seen between delayed precut techniques and persistent standard cannulation [6]. Additionally, TPS and the double-guidewire technique (DGW) can also be applied after guidewire cannulation into the pancreatic duct (PD), with Pécsi D, et al. [7] demonstrating that TPS was superior to both DGW and PP in their deep cannulation rates, with safety profiles being similar among these methods. Although TPS has a higher rate of PEP than PF based on retrospective studies, the optimum insertion technique with respect to early or delayed needle-knife precut techniques, TPS and DGW, remains elusive despite intensive investigations having been undertaken. Network meta-analysis (NMA) compares multiple treatment comparators by using direct and indirect comparisons with a preservation of randomization in individual trials [8–10]. One recent NMA [11] aimed at determining the optimal technique was conducted, with the results confirming that TPS was superior to other interventions both in increasing biliary cannulation rates and decreasing PEP rates. However, Facciorusso A, et al. did not compare PP and PF separately, and the distribution of prophylactic pancreatic stents differed in more than half of the enrolled articles, which may have potentially resulted in the risk of bias regarding the domains of deviations from the intended interventions. In the present study, we conducted an NMA to compare the relative efficacy and safety among 7 insertion techniques for patients with DBC. Furthermore, we performed a subgroup analysis of pancreatitis, bleeding, perforation, infection and death during or after ERCP in order to determine the safety of the various insertion techniques. Materials and methods Search strategy and selection criteria We searched electronic databases, including PubMed, Embase and the Cochrane Central Register of Controlled Trials without involving any language restrictions, while also performing a manual literature search of bibliographies in retrieved articles and published reviews found in eligible publications prior to January 23, 2024, without the enrollment of any studies published in abstract form. This study was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) 2020 statement [12] (Table S1), with the protocol being registered in PROSPERO (CRD 42023403090). We collaborated with a librarian (S-YH) to expand the available vocabulary surrounding the different insertion techniques for DBC. A detailed description of the search strategies is provided in Table S2. Ethical approval and informed consent from the participants were not necessary as there were no data on individual participants. We included randomized comparative trials which had evaluated the efficacy and safety of interventions in adults (aged ≥ 18 years) receiving ERCP due to any biliary indication and who also had a DBC, which was defined as requiring more than five minutes while attempting cannulation of the ampulla of Vater, or more than one unintended PD cannulation or five unintended biliary duct cannulations [3]. We included reports that compared seven interventions, with the conventional technique being defined as continuous standard cannulation after randomizing participants with DBC, PP, PF, delayed PP (dPP), delayed PF (dPF), TPS and DGW. Reports that involved surgically altered anatomy, including Billroth II gastrectomy or Roux-en-Y anastomosis, prior biliary sphincterotomy, gastric outlet obstruction, and pregnancy were all excluded. Outcome measures For the clinical outcome parameters, we determined the deep cannulation rate of intention to treat (ITT) analysis and post-ERCP complications (pancreatitis, bleeding, perforation, infection and death) of per-protocol analysis at follow-up periods both during and after completion of interventions from the enrolled studies. We performed subgroup analysis of pancreatitis, bleeding, perforation, infection and death during or after ERCP in order to determine the safety of insertion techniques for DBC. We defined successful achievement of deep cannulation as the occurrence of selective biliary cannulation during the ERCP procedure. PEP was judged by endoscopists objectively if abdominal pain was present owing to acute pancreatitis where there was an elevation of either serum amylase or lipase at a level of more than two times the upper normal limit after ERCP. Post-ERCP bleeding was defined as there being any clinical evidence of bleeding, reduced hemoglobin, or need for a blood transfusion, endoscopic intervention, or hospitalization after the endoscopic procedures. Perforation was defined as the presence of air, fluid or contrast media leakage either in the abdominal or retroperitoneal cavity during ERCP procedures or after an imaging examination thereafter. Data extraction and quality assessment Two investigators (S-SI and W-YY) independently screened the titles and abstracts for eligibility, and full texts were assessed to clarify the eligibility status of each article. All discrepancies were discussed and resolved during consultation with a third investigator (T-CF). Two reviewers (S-SI and W-YY) independently extracted the data, which was then checked by a third investigator (T-CF). The following variables were extracted: year of publication, country of study, inclusion criteria, participants’ characteristics, details of comparative interventions, and outcome measurements. We contacted any authors for details when encountering missing data. Two investigators (S-SI and W-YY) evaluated the risk of bias of all studies independently and assessed the quality of the articles included in the analysis using version 2 of the Cochrane tool for assessing Risk of Bias in randomised trials (RoB 2.0 tool) [13, 14]. Disagreements were discussed until a consensus was reached, with a third investigator (T-CF) being consulted when necessary. Data synthesis and statistical analyses The NMA, which consisted of both direct and indirect comparisons, was conducted under the frequentist model and generalized linear mixed models [15]. We undertook mixed treatment comparisons of NMA using a random-effect model, which was based upon the mvmeta command, with the restricted maximum likelihood approach applied in the Stata program (version 16, Stata, College Station, TX) [16], in order to investigate therapeutic efficacy and safety. The pooled odds ratios (ORs) and 95% confidence intervals (CIs) were reported for binary variables and for dichotomous variables of safety, where we noted zero events. Imputation for zero cell counts of 0.5 was performed. Heterogeneity among the enrolled studies was evaluated by the τ statistic. Additionally, we evaluated any potential inconsistencies between direct and indirect evidence within the NMA using the design­by­treatment interaction model [17], the loop-specific approach, and the side-splitting method [18, 19]. Statistical significance was set at 5% for all analyses. We also calculated the probabilities of each treatment being at each rank for each intervention and outcome. We obtained a treatment hierarchy using both the Surface Under the Cumulative Ranking curve (SUCRA) and mean ranks [20]. Comparison-adjusted funnel plots [21] and Egger’s tests were used to examine potential small-study bias after treatments were placed in order from oldest to newest. Results The detailed search strategy is summarized in Fig. 1. Initially, we screened 2,060 abstracts and reviewed 85 full-text articles independently after the exclusion of 1,975 studies which were not relevant to our topic. Ultimately, we included 17 articles for qualitative and quantitative analysis. A summary of the 2,189 participants is provided in the Supplementary Data section. PRISMA 2020 flow diagram Characteristics of the included studies The methodology and characteristics of the study design and patient outcomes from the 17 randomized controlled studies are summarized in Tables S3, S4 and S5. Among these studies, sample sizes ranged from 40 to 375 (median 73), while the ranges for age, and percentage of male gender, and cirrhosis were 50.5–70.3, 30.8–64.9, and 0.0-57.7, respectively. More than half of the studies were performed in Western countries (58.8%), with the remaining studies originating from either Japan, Thailand, China, Korea or India. Ten trials (58.8%) enrolled cases involving maneuver of DBC > 10 min when attempting cannulation of the ampulla of Vater, whereas three trials [22–24] involved periods of more than 20 min. Ten trials included maneuvers defined as > 3 unintended PD cannulations, while only one trial registered > 10 unintended biliary duct cannulations [25]. Control group (30.7%), PP (19.8%), and PF (18.8%) were the most common interventions randomly assigned in our review, followed by DGW (14.3%), TPS (13.9%), dPP (1.5%), and dPF (1.0%). The Cochrane Collaboration’s Risk of Bias assessment is shown in Table S6. Some concerns were common in overall bias due to half of the studies (52.9%) not thoroughly describing both the allocation sequence method and randomization process. High risk in overall bias was shown in two studies owing to the trial personnel or the participants being aware of their interventions during the blinding process, or the use of either a non-randomization process or a non-allocated sequence. Taken together, the aforementioned procedural issues resulted in risk of bias regarding the domains of randomization, and deviations from the intended interventions. Outcome parameters The geometry for network comparisons of the deep cannulation rate of DBC and overall post-ERCP complication rates is shown in Fig. 2A and B, while the forest plot of NMA for efficacy is shown in Fig. 3A. When compared with the conventional technique (reference treatment), only TPS (OR 6.00, 95% CI: 1.29–27.96) showed significantly greater efficacy, while dPF (OR 2.88, 95% CI: 0.08-101.07), PP (OR 2.76, 95% CI: 0.73–10.41), DGW (OR 2.00, 95% CI: 0.52–7.66), and PF (OR 1.37, 95% CI: 0.42–4.45) showed non-significantly greater efficacy, with dPP (OR 0.42, 95% CI: 0.03–6.07) showing lower efficacy without significance. The medians (IQRs) for the deep cannulation rates (%) of the conventional technique, PP, PF, dPP, dPF, TPS, and DGW were 72.3 (IQR, 56.9–88.6), 79.6 (IQR, 73.4–87.2), 81.0 (IQR, 75.7–90.8), 81.3, 95.0, 86.0 (IQR, 77.9–94.1), and 70.3 (IQR, 58.5–79.4), respectively. Moreover, only TPS showed a significantly higher efficacy when compared to the conventional technique in indirect comparisons of NMA, as summarized in Table S7A. Network comparisons regarding the geometry of different techniques for difficult biliary cannulation regarding the deep cannulation rate (Figure 2A) and overall Post-ERCP complication rate (Figure 2B). Abbreviations: C, conventional technique; PP, precut papillotomy; PF, precut fistulotomy; dPP, delayed precut papillotomy; dPF, delayed precut fistulotomy; TPS, transpancreatic biliary sphincterotomy; DGW, double-guidewire technique Network forest plot regarding the indirect comparisons (ORs) of Different techniques for difficult biliary cannulation regarding the deep cannulation rate (Figure 3A) and overall Post-ERCP complication rate (Figure 3B). Abbreviations: CI, confidence interval. The abbreviations of therapies are listed in Figure 2 In this NMA, none of the cannulation techniques for DBC, except PF, were significantly associated with a worse post-ERCP complication rate when compared with the conventional technique (Fig. 3B). Only PF (OR 0.61, 95% CI: 0.40–0.93) showed significantly lower complications when compared to the control group. Medians (IQRs) for the post-ERCP complication rates (%) of the conventional technique, PP, PF, dPP, dPF, TPS, and DGW were 12.7 (IQR, 6.2–19.4), 16.5 (IQR, 8.3–23.1), 12.6 (IQR, 9.7–16.7), 0, 5.0, 13.0 (IQR, 5.9–21.6), and 19.8 (IQR, 5.9–34.8), respectively. We also performed subgroup analyses on pancreatitis, bleeding, perforation and infection during the follow-up period (Figures S1A-S1D). In contrast to the conventional technique, only PF (OR 0.35, 95% CI: 0.17–0.71) had a significant protective effect on post-ERCP pancreatitis (Figure S2A). As for post-ERCP bleeding, perforation and infection, we observed that there were no statistical differences among the different cannulation modalities for DBC (Figures S2B-S2D). The League Table is shown in Tables S7A-S7F, while the SUCRA probability is presented in Figures S3A-S3F. Funnel plot revealed visual symmetry (Figures S4A-S4F), while the Egger’s test showed no significant publication bias (Figures S5A-S5F). The League Table for post-ERCP pancreatitis for the different techniques for DBC is shown in Table S7C and confirms that when compared to PP, PF (OR 0.34, 95% CI: 0.13–0.88) also showed significantly lower complications, while DGW (OR 4.50, 95% CI: 1.74–11.66) was associated with a significant increase in complications, which is in contrast to PF. Simultaneously, we used several methods to assess any inconsistencies between the direct and indirect evidence (see Tables S8-S10). There was no significant inconsistency when using the loop-inconsistency model. Using the design­by­treatment inconsistency, we found a statistically significant inconsistency in the outcome of the deep cannulation rate (P < 0.01), as well as the application of the side-splitting method, both in the comparison of C versus dPP (P < 0.05) and PP versus dPP (P < 0.05). In addition, the result of sensitivity analysis after excluding one study [26], which involved comparisons among C, PP and dPP, was also distributed similarly in their ranking to the deep cannulation rate. Both TPS (OR 8.28, 95% CI: 1.81–37.99) and PP (OR 4.53, 95% CI: 1.13–18.13) showed significantly higher insertion efficacy when compared with the conventional technique. There was also a significant inconsistency when using the design­by­treatment model (P < 0.05), loop-inconsistency model (P < 0.05), and side-splitting method in the comparison of C versus PP (P < 0.01). Discussion In this NMA, we comprehensively compared 7 cannulation techniques in order to clarify their relative efficacy and safety in patients experiencing difficult biliary cannulation. We demonstrated that TPS was significantly superior to the other techniques in cannulation efficacy when compared with the control group, while PF was significantly associated with a lower complication rate. In addition, PF was associated with significantly less post-ERCP pancreatitis, and there was also no difference in bleeding, perforation or infection after ERCP. Therefore, our NMA suggests that only TPS achieved better cannulation efficacy for difficult biliary cannulation, while PF showed a lower complication rate, particularly in post-ERCP pancreatitis, which was seen as statistically significant. When encountering difficult biliary manipulation, it is common for the guidewire to accidentally pass into the PD. In such instances, the application of TPS and DGW may be subsequently considered. TPS [4] is performed by making a small incision directed at 11 o’clock through the PD in order to expose the septum between the PD and common bile duct (CBD), particularly when confronting a small and flat papilla. DGW left a guidewire in PD in order to straighten the S-shaped distal CBD and occupy the PD to help facilitate the cannulation with the second guidewire. A traditional meta-analysis [27] reported that TPS resulted in a significantly lower PEP rate than DGW, with the deep cannulation rate being 93.3% in the TPS group and 79.4% in the DGW group. Another retrospective study [28] summarized the real-life experience of a referral center, which demonstrated that DGW displayed a higher success rate and similar safety profile when compared to TPS, although the analysis included one study in abstract form and 78.2% of its patients had malignant biliary stricture. The latest NMA [11] confirmed that TPS was superior to DGW as its biliary cannulation rate was higher and the PEP rate was lower. In our study, only TPS showed significantly greater efficacy when compared with the conventional technique, while DGW (OR 1.16, 95% CI: 0.64–2.09) was non-significantly associated with a worse post-ERCP complication rate when compared to the conventional technique. The ESGE [3, 29] recommends prophylactic pancreatic stenting in patients who have undergone inadvertent guidewire insertion; however, only one study [30] enrolled in our meta-analysis performed prophylactic pancreatic stenting in all patients, with a 2.9% rate of PEP resulting in each group. Yoo YW et al. [26] did not employ any pancreatic stents, yet PEP occurred in 38.2% of the DGW group and in 10.8% of the TPS group, while Kylänpää L et al. [31] adopted prophylactic pancreatic stenting in less than 10% of patients and their PEP result was 14.8% throughout their whole cohort. All of the above indicates that our results are based on heterogenous evidence owing to a performance bias, and indirectness among the enrolled trials. Needle-knife precut techniques are usually executed when performing difficult cannulation without involuntary insertion to PD, and PF spares the orifice of the ampulla of Vater, resulting in reduced PD-related thermal injury, as well as subsequent PEP. The ESGE [3] recommends PF as the preferred method for precutting owing to its use significantly decreasing the PEP rate, and that the selection options between PP and PF on the basis of major duodenal papilla anatomy are equally safe and efficient for patients with DBC [32]. One meta-analysis [6] confirmed that early use of the precut sphincterotomy technique exhibited a reduced risk of PEP when compared with persistent standard cannulation, while Tang Z et al. concluded there was a decreased PEP rate particularly for experienced endoscopists using PF earlier on for patients with DBC [33]. In our analysis, there was no difference among the early precutting groups, late precutting groups, and persistent standard cannulation in terms of the deep cannulation rate. Moreover, only early PF showed significantly lower overall complications when compared to the standard technique, which was mainly attributed to post-ERCP pancreatitis being seen in our subgroup analysis. There are several limitations in this meta-analysis. First, we did not include the pancreatic stent-assisted technique within our seven interventions due to the various execution rates seen in prophylactic pancreatic stenting among the published studies, and the use of prophylactic pancreatic stenting in unintentional insertion is recommended in the ESGE guidelines. A randomized crossover trial [34] enrolled in the latest NMA [11] reported that nine patients had crossed over from the DGW group to the pancreatic stent-assisted group, thus indicating the possibility of selection bias of difficult papilla in the DGW group and interobserver variability among endoscopists. Second, we did not perform meta-regression according to the prevalence of malignancy for the purposes of examining the impact of patient variables on trial outcomes due to missing characteristics. Additionally, the prevalence of malignancy ranged from 0 to 57.7%, which may have contributed to selection bias having originated from heterogeneity in the sample population. Finally, some concerns were common in overall bias due to a lack of detailed information on the allocation sequence method or the randomization process used in approximately half of the trials. Although we discovered significant inconsistencies in the outcome of deep cannulation rates upon using the design­by­treatment inconsistency and the side-splitting methods, the sensitivity analysis after excluding one study showed similar results of ranking in deep cannulation rates. All of the above indicates that our results are based on low-quality evidence, which accounts for both the inconsistency and indirectness among the enrolled trials. Conclusions In this NMA regarding difficult biliary cannulation, we suggest that TPS be initially recommended in patients who experienced inadvertent guidewire insertion into the PD based on the low-quality evidence. Additionally, PF is a safer procedure for patients who had a failed cannulation to PD, as well as for those at a high risk of post-ERCP pancreatitis. Supplementary Information Supplementary Material 1. Acknowledgements The authors gratefully acknowledge the contribution provided by the Evidence-based Practice and Policymaking Committee, Taichung Veterans General Hospital, Taichung, Taiwan. Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. Provenance and peer review Not commissioned; externally peer reviewed. Authors' contributions S-SI, T-CF and W-YY designed the meta-analysis, with input from all listed authors. S-SI, W-YY, and C-HE contributed to data acquisition and drafted the article. S-SI, W-YY, T-CF, and K-CW contributed to data analysis and interpretation. All authors performed critical revisions of the manuscript and approved the final draft of the article. Funding None reported. Data availability No datasets were generated or analysed during the current study. Declarations Ethics approval and consent to participate Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement No datasets were generated or analysed during the current study.