Pre- and post-procedure risk prediction models for post-endoscopic retrograde cholangiopancreatography pancreatitis

Artificial Intelligence in Biliopancreatic Disorders: Applications in Cross-Imaging and Endoscopy

This review explores the transformative potential of artificial intelligence (AI) in the diagnosis and management of biliopancreatic disorders. By leveraging cutting-edge techniques, such as deep learning and convolutional neural networks, AI has significantly advanced gastroenterology, particularly in endoscopic procedures, such as colonoscopy; upper endoscopy; and capsule endoscopy. These applications enhance adenoma detection rates and improve lesion characterization and diagnostic accuracy. AI's integration in cross-sectional imaging modalities, such as computed tomography and magnetic resonance imaging, has remarkable potential. Models have demonstrated high accuracy in identifying pancreatic ductal adenocarcinoma; pancreatic cystic lesions; and pancreatic neuroendocrine tumors, aiding in early diagnosis; resectability assessment; and personalized treatment planning. In advanced endoscopic procedures, such as digital single-operator cholangioscopy and endoscopic ultrasound, AI enhances anatomic recognition and improves lesion classification, with a potential for reduction in procedural variability, enabling more consistent diagnostic and therapeutic outcomes. Promising applications in biliopancreatic endoscopy include the detection of biliary stenosis, classification of dysplastic precursor lesions, and assessment of pancreatic abnormalities. This review aims to capture the current state of AI application in biliopancreatic disorders, summarizing the results of early studies and paving the path for future directions.

Current status and issues for prediction and prevention of postendoscopic retrograde cholangiopancreatography pancreatitis

Acute pancreatitis, which sometimes results in mortality, is a significant complication of endoscopic retrograde cholangiopancreatography (ERCP). Many studies have been conducted to predict and prevent post-ERCP pancreatitis (PEP), and meta-analyses have been reported that summarized these studies. However, many issues remain unresolved. Many risk factors for PEP have been reported, and it is rare for patients undergoing ERCP to have only one risk factor. The use of artificial intelligence may be important for analyzing complex and diverse risk factors. It is desirable to develop an alternative test for pancreatic enzymes that can predict the onset of PEP within 1 h after ERCP. The effectiveness of low-dose nonsteroidal anti-inflammatory drugs (NSAIDs) are controversial. Nitrate and tacrolimus are considered medications that have additional effects on NSAIDs and may be used for the prevention of PEP. Pancreatic stent placement with deliberate placement of the guidewire into the pancreatic duct may be more effective in preventing PEP. A comparison between transpancreatic sphincterotomy with deliberate guidewire placement into the pancreatic duct and needle-knife precut sphincterotomy is necessary. Early precutting is thought to be effective for the prevention of PEP, and the effectiveness of primary precut has been reported. However, the optimal timing of precut for the prevention of PEP has not been sufficiently discussed. Further research on prediction and prevention must be conducted to eliminate the mortality caused by PEP.

Impact of difficult biliary cannulation on post-ERCP pancreatitis: secondary analysis of the stent versus indomethacin trial dataset

BACKGROUND AND AIMS

Difficult biliary cannulation (DBC) is a known risk factor for developing post-ERCP pancreatitis (PEP). To better understand how DBC increases PEP risk, we examined the interplay between technical aspects of DBC and known PEP risk factors.

METHODS

This was a secondary analysis of a multicenter, randomized controlled trial comparing rectal indomethacin alone with the combination of rectal indomethacin and prophylactic pancreatic duct (PD) stent placement for PEP prophylaxis in high-risk patients. Participants were categorized into 3 groups: DBC with high preprocedure risk for PEP, DBC without high preprocedure risk for PEP, and non-DBC at high preprocedure risk for PEP.

RESULTS

In all, 1601 participants (84.1%) experienced DBC, which required a mean of 12 cannulation attempts (standard deviation, 10) and mean duration of 14.7 minutes (standard deviation, 14.9). PEP rate was highest (20.7%) in DBC with a high preprocedure risk, followed by non-DBC with a high preprocedure risk (13.5%), and then DBC without a high preprocedure risk (8.8%). Increasing number of PD wire passages (adjusted odds ratio [aOR], 1.97; 95% confidence interval [CI], 1.25-3.1) was associated with PEP in DBC, but PD injection, pancreatic sphincterotomy, and number of cannulation attempts were not associated with PEP. Combining indomethacin with PD stent placement lowered the risk of PEP (aOR, .61; 95% CI, .44-.84) in DBCs. This protective effect was evident in up to at least 4 PD wire passages.

CONCLUSIONS

DBC confers higher PEP risk in an additive fashion to preprocedural risk factors. PD wire passages appear to add the greatest PEP risk in DBCs, but combining indomethacin with PD stent placement reduces this risk, even with increasing PD wire passages.

Prospective validation and revision of predictive models for post-ERCP pancreatitis: focus on procedure-related factors and a novel risk stratification approach

INTRODUCTION

Post-endoscopic retrograde cholangiopancreatography pancreatitis (PEP) is the most common serious adverse event after endoscopic retrograde cholangiopancreatography (ERCP). Although retrospective models to predict PEP have shown promise, their real-world applicability remains uncertain. Thus, we used prospectively derived cohort data to validate current prediction models.

METHODS

We conducted a prospective cohort study involving patients who underwent ERCP between August 2020 and December 2023. We validated the original PEP-risk prediction models using prospective cohort data and, if necessary, refined them using logistic regression analysis.

RESULTS

Among the 1112 study participants, the original PEP-risk prediction models had limited performance. Although PEP incidence tended to increase across risk groups, the differences were mostly insignificant. Logistic regression highlighted procedural factors-total procedure time [odds ratio (OR) 1.13, 95% confidence interval (CI) 1.01-1.26 per 5 min], unintended pancreatic duct cannulation (OR 2.56, 95% CI 1.11-5.93), and pancreatic opacification (OR 2.57, 95% CI 1.19-5.58)-as independent PEP risk factors. This led to a revised model that assigned 1, 8, and 8 points to these factors. Patients stratified into low- (0-2 points), intermediate- (3-10 points), and high-risk groups (≥ 11 points) exhibited PEP incidences of 2.6% (95% CI 1.5-4.1%), 7.1% (95% CI 4.8-10.3%), and 12.6% (95% CI 8.6-17.8%), respectively.

CONCLUSIONS

We highlighted limitations of existing PEP-prediction models that necessitate refinement based on procedural variables. Our revised model accounted for the prolonged total procedure time, unintended pancreatic duct cannulation, and pancreatic opacification, offering enhanced accuracy in predicting PEP risk.

Development and validation of a machine learning-based, point-of-care risk calculator for post-ERCP pancreatitis and prophylaxis selection

BACKGROUND AND AIMS

A robust model of post-ERCP pancreatitis (PEP) risk is not currently available. We aimed to develop a machine learning-based tool for PEP risk prediction to aid in clinical decision making related to periprocedural prophylaxis selection and postprocedural monitoring.

METHODS

Feature selection, model training, and validation were performed using patient-level data from 12 randomized controlled trials. A gradient-boosted machine (GBM) model was trained to estimate PEP risk, and the performance of the resulting model was evaluated using the area under the receiver operating curve (AUC) with 5-fold cross-validation. A web-based clinical decision-making tool was created, and a prospective pilot study was performed using data from ERCPs performed at the Johns Hopkins Hospital over a 1-month period.

RESULTS

A total of 7389 patients were included in the GBM with an 8.6% rate of PEP. The model was trained on 20 PEP risk factors and 5 prophylactic interventions (rectal nonsteroidal anti-inflammatory drugs [NSAIDs], aggressive hydration, combined rectal NSAIDs and aggressive hydration, pancreatic duct stenting, and combined rectal NSAIDs and pancreatic duct stenting). The resulting GBM model had an AUC of 0.70 (65% specificity, 65% sensitivity, 95% negative predictive value, and 15% positive predictive value). A total of 135 patients were included in the prospective pilot study, resulting in an AUC of 0.74.

CONCLUSIONS

This study demonstrates the feasibility and utility of a novel machine learning-based PEP risk estimation tool with high negative predictive value to aid in prophylaxis selection and identify patients at low risk who may not require extended postprocedure monitoring.

[Personalized prediction of acute pancreatitis after endoscopic transpapillary interventions]

OBJECTIVE

To develop the personalized model for predicting the risk of acute pancreatitis after endoscopic transpapillary interventions.

MATERIAL AND METHODS

A retrospective analysis of treatment outcomes included 366 patients with benign and malignant pancreaticobiliary diseases who underwent endoscopic transpapillary interventions. Risk factors associated with patients, underlying diseases and interventions were analyzed. Logistic regression analysis was used to present the personalized model for predicting the risk of acute pancreatitis.

RESULTS

Female gender (p=0.028), age <40 years (p=0.001-0.018), calculous cholecystitis (p=0.010) and stenosis of the major duodenal papilla (p=0.008) are patient-associated risk factors of acute postoperative pancreatitis. Stenting of the main pancreatic duct and thoracic epidural analgesia reduced this risk by 6.5 and 4.6 times, respectively. We developed significant (p<0.001) regression model to determine the likelihood of acute post-manipulation pancreatitis.

CONCLUSION

Original prediction model is valuable to determine the risk of acute pancreatitis after endoscopic transpapillary interventions. This model justifies various methods to prevent this complication.

Multistep validation of a post-ERCP pancreatitis prediction system integrating multimodal data: a multicenter study

BACKGROUND AND AIMS

The impact of various categories of information on the prediction of post-ERCP pancreatitis (PEP) remains uncertain. We comprehensively investigated the risk factors associated with PEP by constructing and validating a model incorporating multimodal data through multiple steps.

METHODS

Cases (n = 1916) of ERCP were retrospectively collected from multiple centers for model construction. Through literature research, 49 electronic health record (EHR) features and 1 image feature related to PEP were identified. The EHR features were categorized into baseline, diagnosis, technique, and prevention strategies, covering pre-ERCP, intra-ERCP, and peri-ERCP phases. We first incrementally constructed models 1 to 4 incorporating these 4 feature categories and then added the image feature into models 1 to 4 and developed models 5 to 8. All models underwent testing and comparison using both internal and external test sets. Once the optimal model was selected, we conducted comparisons among multiple machine learning algorithms.

RESULTS

Compared with model 2 that incorporated baseline and diagnosis features, adding technique and prevention strategies (model 4) greatly improved the sensitivity (63.89% vs 83.33%, P < .05) and specificity (75.00% vs 85.92%, P < .001). A similar tendency was observed in the internal and external tests. In model 4, the top 3 features ranked by weight were previous pancreatitis, nonsteroidal anti-inflammatory drug use, and difficult cannulation. The image-based feature has the highest weight in models 5 to 8. Finally, model 8 used a random forest algorithm and showed the best performance.

CONCLUSIONS

We first developed a multimodal prediction model for identifying PEP with a clinical-acceptable performance. The image and technique features are crucial for PEP prediction.

Synergistic effect of independent risk factors for post-endoscopic retrograde cholangiopancreatography pancreatitis: a multicenter retrospective study in Japan

BACKGROUND/AIMS

This study aimed to examine the synergistic effect of independent risk factors on post-endoscopic retrograde cholangiopancreatography (ERCP) pancreatitis (PEP).

METHODS

This multicenter retrospective study included 1,273 patients with native papillae who underwent ERCP for bile duct stones in Japan. Independent PEP risk factors were identified using univariate and multivariate analyses. Significant risk factors for PEP in the multivariate analysis were included in the final analysis to examine the synergistic effect of independent risk factors for PEP.

RESULTS

PEP occurred in 45 of 1,273 patients (3.5%). Three factors including difficult cannulation ≥10 minutes, pancreatic injection, and normal serum bilirubin level were included in the final analysis. The incidences of PEP in patients with zero, one, two, and three factors were 0.5% (2/388), 1.9% (9/465), 6.0% (17/285), and 12.6% (17/135), respectively. With increasing risk factors for PEP, the incidence of PEP significantly increased (1 factor vs. 2 factors, p=0.006; 2 factors vs. 3 factors, p=0.033).

CONCLUSIONS

As the number of risk factors for PEP increases, the risk of PEP may not be additive; however, it may multiply. Thus, aggressive prophylaxis for PEP is strongly recommended in patients with multiple risk factors.

Artificial intelligence in a prediction model for postendoscopic retrograde cholangiopancreatography pancreatitis

OBJECTIVES

In this study we aimed to develop an artificial intelligence-based model for predicting postendoscopic retrograde cholangiopancreatography (ERCP) pancreatitis (PEP).

METHODS

We retrospectively reviewed ERCP patients at Nagoya University Hospital (NUH) and Toyota Memorial Hospital (TMH). We constructed two prediction models, a random forest (RF), one of the machine-learning algorithms, and a logistic regression (LR) model. First, we selected features of each model from 40 possible features. Then the models were trained and validated using three fold cross-validation in the NUH cohort and tested in the TMH cohort. The area under the receiver operating characteristic curve (AUROC) was used to assess model performance. Finally, using the output parameters of the RF model, we classified the patients into low-, medium-, and high-risk groups.

RESULTS

A total of 615 patients at NUH and 544 patients at TMH were enrolled. Ten features were selected for the RF model, including albumin, creatinine, biliary tract cancer, pancreatic cancer, bile duct stone, total procedure time, pancreatic duct injection, pancreatic guidewire-assisted technique without a pancreatic stent, intraductal ultrasonography, and bile duct biopsy. In the three fold cross-validation, the RF model showed better predictive ability than the LR model (AUROC 0.821 vs. 0.660). In the test, the RF model also showed better performance (AUROC 0.770 vs. 0.663, P = 0.002). Based on the RF model, we classified the patients according to the incidence of PEP (2.9%, 10.0%, and 23.9%).

CONCLUSION

We developed an RF model. Machine-learning algorithms could be powerful tools to develop accurate prediction models.

Predictive models in EUS/ERCP

Predictive models (PMs) in endoscopic retrograde cholangiopancreatography (ERCP) and endoscopic ultrasound (EUS) have the potential to improve patient outcomes, enhance diagnostic accuracy, and guide therapeutic interventions. This review aims to summarize the current state of predictive models in ERCP and EUS and their clinical implications. To be considered useful in clinical practice a PM should be accurate, easy to perform, and may consider objective variables. PMs in ERCP estimate correct indication, probability of success, and the risk of developing adverse events. These models incorporate patient-related factors and technical aspects of the procedure. In the field of EUS, these models utilize clinical and imaging data to predict the likelihood of malignancy, presence of specific lesions, or risk of complications related to therapeutic interventions. Further research, validation, and refinement are necessary to maximize the utility and impact of these models in routine clinical practice.

Improved fluoroscopy-guided biopsies in the diagnosis of indeterminate biliary strictures: a multi-center retrospective study

To evaluate the diagnostic accuracy of improved fluoroscopy-guided biopsies for indeterminate biliary strictures (IBDS). A multi-center retrospective study was performed. Patients with IBDS who underwent digital single-operator cholangioscopy (DSOC) and improved fluoroscopy-guided biopsies procedures were included. The individual sensitivity, specificity, and accuracy were analyzed. A total of 67 patients were enrolled in this multi-center retrospective study. The DSOC and improved fluoroscopy-guided biopsies procedures were successfully performed in all cases (100%). The sensitivity, specificity, and accuracy values were 83.3%, 89.5%, and 85.1% for DSOC visual impression; 95.8%, 94.7%, and 95.5% for improved fluoroscopy-guided biopsies procedures, respectively. The sensitivity and accuracy of improved fluoroscopy-guided biopsies were significantly higher compared with DSOC visual impression. Four patients (6.0%, 4/67) occurred adverse events after the procedures. Improved fluoroscopy-guided biopsies had a high diagnostic accuracy of IBDS diagnosis.

The efficacy and safety of laparoscopic common bile duct exploration and cholecystectomy for the treatment of difficult common bile duct stones combined with gallstones: a multicenter retrospective study

PURPOSE

To analyze the benefits of laparoscopic common bile duct exploration and laparoscopic cholecystectomy (LCBDE + LC) versus endoscopic retrograde cholangiopancreatography and/or endoscopic sphincterotomy following laparoscopic cholecystectomy (ERCP/EST + LC) for difficult common bile duct stones combined with gallstones.

METHODS

A retrospective analysis of consecutive patients with difficult common bile duct stones combined with gallstones in three hospitals from January 2016 to January 2021 was performed.

RESULTS

ERCP/EST + LC contributed to reducing postoperative drainage time. However, LCBDE + LC showed a higher rate of complete clearance, along with lower postoperative hospital stays, expenses and incidence of postoperative hyperamylasemia, pancreatitis, re-operation and recurrence. In addition, LCBDE + LC showed safe and feasible performance in the elderly and patients with previous upper abdominal surgery.

CONCLUSION

It is an effective and safe method for LCBDE + LC for difficult common bile duct stones combined with gallstones.

A Risk Prediction Model for Post-endoscopic Retrograde Cholangiopancreatography Pancreatitis After Stent Insertion for Malignant Biliary Obstruction: Development and Validation

OBJECTIVES

Pancreatitis is the most common complication of post-endoscopic retrograde cholangiopancreatography (ERCP). There are currently no prediction models, particularly for post-ERCP pancreatitis (PEP) after biliary stent placement due to malignant biliary obstruction (MBO). To that end, we aim to develop and validate a predictive model for PEP.

METHODS

We retrospectively analyzed the data of patients who underwent ERCP for biliary stent placement due to MBO at the Second Affiliated Hospital of Harbin Medical University from January 1, 2014 to August 31, 2021. The eligible patients were randomly allocated to the development and validation cohorts. A prediction model was built using the development cohort, and the model's effect was validated using a validation cohort.

RESULTS

A total of 1524 patients were enrolled, including 1016 in the development cohort and 508 in the validation cohort, with an overall PEP rate of 7.1%. The model's predictors included acute pancreatitis history, the absence of pancreatic duct dilation, nonpancreatic cancer, difficult cannulation, and pancreatic injection. The area under the curve (AUC) in the development cohort was 0.810, and the incidence of PEP in the low-risk, medium-risk, and high-risk groups was 1.53%, 9.12%, and 36.36%, respectively. Meanwhile, the AUC of the validation cohort was 0.781, and the incidence of PEP in the low-risk, medium-risk, and high-risk groups was 4.17%, 8.75%, and 41.67%, respectively.

CONCLUSIONS

This study was the first to build and validate a risk prediction model, especially for PEP after biliary stent placement due to MBO. Moreover, this model might assist clinicians in identifying high-risk patients and help implement preventive measures in a more timely manner.

Prevention of post-ERCP pancreatitis: current strategies and novel perspectives

Endoscopic retrograde cholangiopancreatography (ERCP) is an advanced endoscopic procedure that might lead to severe adverse events. Post-ERCP pancreatitis (PEP) is the most common post-procedural complication, which is related to significant mortality and increasing healthcare costs. Up to now, the prevalent approach to prevent PEP consisted of employing pharmacological and technical expedients that have been shown to improve post-ERCP outcomes, such as the administration of rectal nonsteroidal anti-inflammatory drugs, aggressive intravenous hydration, and the placement of a pancreatic stent. However, it has been reported that PEP originates from a more complex interaction of procedural and patient-related factors. Appropriate ERCP training has a pivotal role in PEP prevention strategy, and it is not a chance that a low PEP rate is universally considered one of the most relevant indicators of proficiency in ERCP. Scant data on the acquisition of skills during the ERCP training are currently available, although some efforts have been recently done to shorten the learning curve by way of simulation-based training and demonstrate competency by meeting technical requirements as well as adopting skill evaluation scales. Besides, the identification of adequate indications for ERCP and accurate pre-procedural risk stratification of patients might help to reduce PEP occurrence regardless of the endoscopist's technical abilities, and generally preserve safety in ERCP. This review aims at delineating current preventive strategies and highlighting novel perspectives for a safer ERCP focusing on the prevention of PEP.

Current approaches and questions yet to be resolved for the prophylaxis of post-endoscopic retrograde cholangiopancreatography pancreatitis

Prophylaxis is important for post-endoscopic retrograde cholangiopancreatography (ERCP) pancreatitis (PEP), which is the most common and serious complication of ERCP. Although the current guidelines include independent patient- and procedure-related risk factors for PEP and available PEP prophylactic measures, the synergistic effect of these risk factors on PEP should also be considered, given that patients often harbor multiple risk factors. Furthermore, a combination of prophylactic measures is often selected in clinical practice. However, established methods estimating the synergistic effect of independent risk factors on PEP incidence are lacking, and evidence on the impact of combining prophylactic measures on PEP should be discussed. Selection of appropriate candidate patients for ERCP is also important to reduce the incidence of PEP associated with unnecessary ERCP. ERCP indications in patients with asymptomatic common bile duct stones (CBDSs) and in those with suspected CBDSs with no imaging-based evidence of stones are controversial. Further studies are warranted to predict the synergistic effect of independent risk factors on PEP, determine the best prophylactic PEP measures, and identify appropriate candidates for ERCP in patients with asymptomatic CBDSs and those with suspected CBDSs.

The assessment of the ASGE-grading system of ERCP: a large-sample retrospective study

BACKGROUND AND AIMS

The American Society for Gastrointestinal Endoscopy (ASGE) has developed a complexity-grading system for endoscopic retrograde cholangiopancreatography (ERCP) to predict technical success and adverse events. This study aimed to assess the association between the degree of difficulty for ERCP and the rates of success and adverse event, in turn demonstrating the validity and practicality of this system.

METHODS

ERCP procedures performed in the First Affiliated Hospital of Nanchang University from January 2011 to December 2020 were retrospectively reviewed. Procedural success and adverse events were recorded based on difficulty level according to the ASGE-grading system.

RESULTS

A total of 20,652 ERCP procedures performed during the study period were analyzed, including 1908 procedures considered grade 1(9.2%), 10,170 procedures considered grade 2 (49.2%), 7764 procedures considered grade 3 (37.6%), 810 procedures considered grade 4 (3.9%). The overall success rate increased from 92.8% in 2011-2015 to 94.0% in 2016-2020, while the distribution of procedures and the incidence of complications showed little variation. The success rate revealed a significantly decreasing trend with increasing difficulty (ranging from 55.6 to 98.6%), mainly for biliary diseases. In addition, the difficulty scale was not associated with any differences in the rate of adverse event, except for the pancreatitis for grade 1 procedures, which had a low incidence.

CONCLUSIONS

The ASGE-grading system can help predict the success rate of ERCP procedures but showed poor performance in predicting adverse events. Further exploration may be required to improve the grading system by adjusting or including certain clinical parameters, and to validate the system for extrapolation to other endoscopy units.

[Comparison of Clinical Performance and Safety between Domestic New Pull-type Triple-lumen Sphincterotome and Conventional Sphincterotome: A Prospective Multicenter Trial]

Background/Aims

Sphincterotomes are essential for endoscopic sphincterotomy (EST) and can also be used for cannulation in ERCP. A domestic new pull-type sphincterotome (Optimos™, Taewoong, Goyang, Korea) provides acceptable technical feasibility and safety, but there are no comparison results. Thus, this study compared the clinical performance and safety of Optimos™ sphincterotome to a conventional sphincterotome (CleverCut3™, Olympus, Tokyo, Japan) in patients who underwent ERCP.

Methods

From April 2021 to July 2021, a randomized prospective comparative study was conducted on 104 consecutive patients who underwent ERCP in three medical centers. The primary endpoint was the clinical performance and safety of sphincterotomes during ERCP.

Results

One hundred and four patients were assigned randomly to the Optimos™ group (n=51) or CleverCut3™ group (n=53). All demographic characteristics did not differ between the groups except the BMI. The technical success rate for cannulation, performance of EST, and total procedure time were similar in the two groups. The adverse events did not differ, even though two cases of post-ERCP pancreatitis occurred in CleverCut3™. On the other hand, in questionnaire analysis, CleverCut™ showed a better user's convenience (median [interquartile range] 4.0 [3.0-4.0] vs. 3.0 [3.0-4.0], p=0.013) and manipulability (median [interquartile range], 4.0 [3.0-4.0] vs. 3.0 [3.0-4.0], p=0.039) than Optimos™, even though the other profiles did not reveal any differences.

Conclusions

New domestic pull-type sphincterotome can offer comparable clinical performance and safety profiles to conventional sphincterotome, but it needs refinements to increase the user's convenience and manipulability. Further improvement and innovation will be required to advance domestic medical devices.