Introduction

Malignant biliary obstruction (MBO) is commonly caused by pancreatic cancer, cholangiocarcinoma, ampullary cancer, gallbladder cancer, and hepatocellular cancer, to name a few. MBO causes pruritus, weight loss, abdominal discomfort or pain, and obstructive jaundice, and it can be complicated by severe cholangitis, sepsis, or liver failure, all of which have a negative impact on the patient's survival and quality of life [1, 2]. Unfortunately, many patients are deemed unfit for surgery when they are diagnosed. Biliary stent placement for MBO via endoscopic retrograde cholangiopancreatography (ERCP) is widely accepted as the first-line palliative strategy, helping improve affected patients’ quality of life [3]. However, the procedure can occasionally result in serious, even fatal, adverse events, the most common and feared of which is post-ERCP pancreatitis (PEP). According to previous reports, the rate of PEP after stent insertion for MBO ranges from 1.3 to 26.8%, with the incidence of bleeding (including early bleeding and delayed bleeding) ranging between 0.5 and 5.1%, while the incidence of perforation is less common with a rate of 0.3% to 2% [4,5,6,7,8,9,10,11,12]. Notably, the management of MBO is different in malignant versus benign disorders. In MBO treatment, radical surgery will be performed if the tumor is resectable. If the tumor has metastasized or the patient’s condition does not allow for radical operation, palliative care is given instead. Palliative surgery includes cholecystogastrostomy, hepaticojejunostomy, cholecystojejunostomy, and choledochoduodenostomy. However, as preoperative imaging and interventional surgery have improved, the rate of palliative surgery has decreased. The most widely used palliative treatment is the placement of a self-expandable metal stent (SEMS) via ERCP. Other treatment options, such as EUS-guided biliary drainage and percutaneous biliary drainage, are also widely used for palliative treatment [1, 13, 14]. For benign biliary obstruction, inserting SEMS via ERCP is highly recommended. Surgery or percutaneous transhepatic biliary drainage (PTBD) may be considered for patients who are not candidates for stent placement [15, 16].

To prevent or decrease the occurrence of PEP, it is essential to recognize the risk factors of PEP. PEP risk factors include patient and procedure-related risk factors, such as female gender, history of pancreatitis, younger age, difficult cannulation, repetitive pancreatic guidewire cannulation, and pancreatic injection [17,18,19]. Since the etiology, clinical features, and therapeutic strategy of MBO differ from those of other biliopancreatic diseases, the risk factors for PEP in MBO patients may differ from those previously reported. Recently, several risk prediction models for predicting the occurrence of PEP have been developed [20,21,22,23]. Nevertheless, no prediction models exist for PEP after biliary stent placement due to MBO. As a result, we aim to build and validate a predictive model for it.

Methods

Patients

Radical surgery was conducted at our center if the tumor was resectable in MBO patients; otherwise, palliative treatment was given. We first considered biliary stent placement via ERCP for palliative treatment, while SEMS was the first choice for patients with an expected survival of more than 3 months. If these procedures were unsuccessful, PTBD would be performed instead. Chemotherapy, radiotherapy, transarterial chemoembolization (TACE), and transarterial radioembolization (TARE) could be used in conjunction with the above treatments. Patients who underwent biliary stent insertion for MBO via ERCP from January 1, 2014, to August 1, 2021 were identified retrospectively at the Second Affiliated Hospital of Harbin Medical University. MBO was diagnosed based on clinical, laboratory, radiological, and pathological examinations. All included patients underwent palliative treatment. The exclusion criteria were as follows: (1) non-native papilla, (2) ongoing acute pancreatitis, (3) failed operation, (4) age < 18, (5) replacement of stent, and (6) incomplete medical record. Before the procedure, all patients undergoing ERCP at our hospital were given rectal nonsteroidal anti-inflammatory drugs (NSAIDs). All procedures were carried out by experienced endoscopists who have completed over 200 cases in our hospital. The endoscopist was in charge of deciding which type of stent to use during the procedure. The study adhered to the Helsinki Declaration's ethical principles. The Medical Ethics Committee of the Second Affiliated Hospital of Harbin Medical University approved this data-only retrospective study. Informed consent was not required because of the retrospective nature of the study.

Data Collection

We collected two types of data: (1) Patient-related data which included gender, age, hypertension, cholecystectomy, gastrectomy, diabetes, history of chronic pancreatitis, history of acute pancreatitis, total bilirubin (TBIL), direct bilirubin (DBIL), alkaline phosphatase (ALP), aspartate aminotransferase (AST), γ-glutamyl transferase (GGT), alanine aminotransferase (ALT), white blood cell (WBC), pancreatic duct (PD) diameter, common bile duct (CBD) diameter, cancer type, location of stricture, and periampullary diverticulum. (2) Procedure-related data included difficult cannulation, pancreatic injection, stent type, endoscopic sphincterotomy (EST), precut sphincterotomy, PD stenting, and double guidewire technique. Although SOD has been proposed as an independent risk factor for PEP, the disease entity is uncommon in Asian populations, so we did not include it in our study [22].

Definitions

The diagnosis of PEP conformed to Cotton’s criterion: new or worsening abdominal pain continuing for at least 24 h after ERCP, with elevated serum amylase levels three times the normal upper limit or higher [24]. Computed tomography (CT) was performed for all patients suspected of PEP to confirm the diagnosis radiologically. The revised Atlantic classification defined the severity of PEP: (1) mild: no organ dysfunction and other complications, (2) moderate: transient organ failure < 48 h or local or systemic complications, and (3) severe: persistent single or multi-organ failure > 48 h [25]. Difficult cannulation was defined as more than five cannulation attempts or a long cannulation time (> 10 min). PD dilation was defined as PD maximum diameter ≥ 4 mm in the preoperative imaging examinations (ultrasound, CT, MRCP), and CBD dilation was defined as CBD maximum diameter of ≥ 8 mm. In our study, the definition of the distal bile duct was the common bile duct located downstream of the cystic duct confluence, and the definition of the hilar bile duct was the bile duct located upstream of the cystic duct. The diagnosis of cholangitis is based on the Tokyo Guidelines 2018 [26]. ECOG criteria were used to assess the performance status of patients, patients with 0 points were assigned to group 1, patients with 1–2 points were assigned to group 2, and patients with 3–4 points were classified into group 3 [27].

Statistical Analysis

We allocated eligible patients to the development and validation cohorts at a ratio of 2:1 randomly. According to the optimal sensitivity and specificity, continuous variables were dichotomized based on the cutoff value of each indicator [28]. Categorical variables were presented as frequency and were compared between groups using the Chi-square test.

To develop the prediction model, logistic regression analysis was used to evaluate the risk factors of PEP in the development cohort. Variables with P values < 0.05 in the univariable logistic regression model were selected as covariates in the multivariable analysis. Variables with P values < 0.05 in the multivariable analysis were selected as predictors in the risk prediction model. Each predictor was assigned a score. Scores were calculated by dividing each predictor's regression coefficient by the smallest one in the model and rounding it to the nearest integer. A patient’s total risk score was calculated by adding each predictor’s score. Then, the total risk score of each patient was calculated in both the development and validation cohorts. According to the total risk score, patients were allocated to the low-risk group (0–1 points), medium-risk group (2–3 points), and high-risk group (≥ 4 points).

The area under the receiver operating characteristics (ROC) curve (AUC) was used to evaluate the discrimination of the prediction model. Calibration is another measure of prediction model performance used to test how well the predicted results match the actual results. The actual rate of PEP was calculated in the three different risk groups, respectively. The predicted rate of PEP for each group was calculated by the model regression formula and was presented as the mean predicted rate and the standard deviation (SD). Then, the predicted rate was compared with each group’s actual rate to verify the model’s calibration. Hosmer–Lemeshow (H–L) test was also used to evaluate the model’s goodness of fit. SPSS 24.0 (IBM Corp, USA) was utilized to analyze all the data. P value < 0.05 (two-sided) was considered statistically significant.

Results

Patient Characteristics

A total of 1937 patients who underwent biliary stent insertion for MBO via ERCP were considered. After excluding patients for non-native papilla (104 patients), ongoing acute pancreatitis (3 patients), failed ERCP (45 patients), replacement of stent (204 patients), and incomplete medical records (57 patients), 1524 patients were finally included. 1016 and 508 patients were randomly assigned to the development and validation cohorts, respectively. The baseline information of the included patients is shown in Tables 1, 2, and 3. The proportion of males was 55.5% in the development cohort and 53.3% in the validation cohort (P = 0.423). The proportion of patients aged > 65 was 51.7% in the development cohort and 52.0% in the validation cohort (P = 0.913). The most common cause of biliary stent placement was pancreatic cancer (72.3% and 76.2%), followed by cholangiocarcinoma (17.8% and 14.0%) and ampullary cancer (8.9% and 8.3%) in the development and validation cohorts. None of the baseline characteristics significantly differed between the development and validation cohorts.

Table 1 Demographic parameters of patients in the development and validation cohorts
Full size table
Table 2 Co-morbidities of patients in the development and validation cohorts
Full size table
Table 3 Laboratory parameters of patients in the development and validation cohorts
Full size table

Incidence of PEP and Its Severity

A total of 108 patients developed PEP among the 1524 patients who underwent ERCP, with a rate of 7.1%. Of which, 66 (6.5%) patients had PEP in the development cohort, and 42 (8.3%) had PEP in the validation cohort. According to the revised Atlantic classification, 60 (5.9%) patients had mild, 4 (0.4%) had moderate, and 2 (0.2%) had severe PEP in the development cohort, and 37 (7.3%) patients had mild, 3 (0.6%) had moderate, and 2 (0.4%) had serve PEP in the validation cohort.

Developing the Risk Prediction Model

Univariate analysis was performed to ascertain the potential risk factors related to PEP in the development cohort. In the univariable analysis, age ≤ 65, acute pancreatitis history, the absence of PD dilation, nonpancreatic cancer, difficult cannulation, pancreatic injection, and double guidewire technique were associated with the development of PEP (Table 4). These seven risk factors were included in our multivariate analysis, revealing that acute pancreatitis history (OR = 4.517, 95% CI 1.667–12.245), absence of PD dilation (OR = 2.813, 95% CI 1.519–5.208), nonpancreatic cancer (OR = 2.218, 95% CI 1.210–4.066), difficult cannulation (OR = 5.807, 95% CI 3.237–10.417), and pancreatic injection (OR = 4.365, 95% CI 1.653–11.524) were independent risk factors for the occurrence of PEP (Table 5). As a result, these five independent risk factors were chosen as predictors in the model.

Table 4 Univariate logistic regression analysis in the development cohort
Full size table
Table 5 Multivariate logistic regression analysis in the development cohort and scoring system
Full size table

Scores were allocated to each predictor based on their β coefficient. The β coefficient and scores are shown in Table 3. The total risk score of each patient in the development and validation cohorts was calculated by adding up the scores of each predictor. Based on their total risk scores, patients were classified into three different groups: low-risk (0–1 points), medium-risk (2–3 points), and high-risk (≥ 4 points). Furthermore, each patient's predicted rate was calculated using the model formula and assigned to each patient in both the development and validation cohorts.

Performance of the Model

The H–L test showed good fitness for the model in both development cohort (χ2 = 8.35, P = 0.423) and validation cohort (χ2 = 2.17, P = 0.714). The model in the development cohort had an AUC of 0.810 (95% CI 0.751–0.868), while the model in the validation cohort had an AUC of 0.781 (95% CI 0.703–0.858) (Fig. 1).

Fig. 1
figure 1

ROC curves in the A development cohort and B validation cohort

Full size image

In both the development and validation cohorts, the mean predicted rate, SD, and the actual rate of PEP were calculated for the three different risk groups. In the development cohort, the actual rates of PEP were 1.53%, 9.12%, and 36.36% in the low-risk, medium-risk, and high-risk groups, respectively. The predicted rates of PEP were 1.83% ± 0.92%, 8.55% ± 3.67%, and 36.85% ± 15.51% in the low-risk, medium-risk, and high-risk groups, respectively. In the validation cohort, the actual rates of PEP were 4.17%, 8.75%, and 41.67% in the low-risk, medium-risk, and high-risk groups, respectively. Meanwhile, the predicted rates of PEP were 3.32% ± 2.49%, 11.53% ± 7.77%, and 36.63% ± 15.07% in the low-risk, medium-risk, and high-risk groups, respectively (Fig. 2). Therefore, in each risk group, the actual rate was consistent with the predicted rate, indicating that the model was well calibrated.

Fig. 2
figure 2

The actual and predicted rates PEP in the A development cohort and B validation cohort

Full size image

In the development cohort, the incidence of PEP was higher in the medium-risk (OR = 6.540, 95% CI 3.050–13.652, P < 0.001) and high-risk (OR = 36.762, 95% CI 16.068–84.108, P < 0.001) groups compared to the low-risk group. Similarly, in the validation cohort, the incidence of PEP was higher in the medium-risk (OR = 2.205, 95% CI 1.011–4.813, P = 0.047) and high-risk (OR = 16.429, 95% CI 6.921–38.999, P < 0.001) groups compared to the low-risk group (Table 6).

Table 6 PEP rates in different groups
Full size table

Discussion

Biliary stent insertion is an important palliative treatment for MBO. PEP is the most common complication, and its incidence is about 1.3% to 26.8% [4,5,6]. In this study, the overall incidence of PEP was 7.1%, which is within the above-reported range.

In recent years, several risk prediction models have been established for predicting the incidence of PEP [20,21,22,23]. Nonetheless, there are no prediction models, especially for PEP after biliary stent placement due to MBO. Therefore, it is necessary to establish a risk prediction model for PEP after biliary stent placement due to MBO to facilitate the risk stratification of patients undergoing this procedure and allow physicians to implement timely preventive measures for high-risk patients. The present study is the first to build and validate a risk prediction model for PEP after biliary stent placement due to MBO. Acute pancreatitis history, the absence of PD dilation, nonpancreatic cancer, difficult cannulation, and pancreatic injection were all identified as risk factors for this procedure. Consistent with the literature, pancreatic injection, acute pancreatitis history, and difficult cannulation are well-known risk factors for PEP [17]. We also found that the absence of PD dilation and nonpancreatic cancer were associated with PEP.

The reason why PD dilation was associated with PEP could be that PD dilation is often characterized by PD hypertension. As a result, the patients may have a high tolerance for increased PD pressure and thus be resistant to PEP [29]. Nonpancreatic cancer has been reported as an independent risk factor for PEP after biliary stent insertion [30, 31], consistent with our findings. In most pancreatic cancer patients, the tumor invaded and dilated the PD. Furthermore, the exocrine function of the pancreas may decrease due to parenchymal atrophy caused by tumor compression. Studies have shown that the volume of pancreatic parenchyma is strongly associated with the occurrence of PEP [32, 33]. This suggests that pre-ERCP graphical evaluation might be useful for predicting PEP after biliary stent insertion for MBO.

Many endoscopists prefer to perform EST before inserting a biliary stent. EST is thought to make biliary stent deployment easier and may lower the risk of PEP by reducing tension at the PD opening [34, 35]. However, whether EST prior to biliary stent placement can prevent PEP is debatable. A meta-analysis that included three randomized controlled trials (RCTs) showed that EST prior to stent placement might effectively reduce the occurrence of PEP [36]. In addition, a retrospective study also indicated that EST could prevent PEP in patients with biliary tumors, especially when transpapillary biopsy and intraductal ultrasound were performed [37]. In contrast, some studies reached the opposite conclusions. For instance, Kato et al. revealed that EST prior to biliary stent insertion could not decrease the occurrence of PEP for patients with distal MBO without PD involvement [6]. Moreover, a recent systematic review and meta-analysis by Sofi et al. reported that biliary stent placement without EST was not related to a higher risk of PEP in patients with distal biliary tract obstruction and PD involvement [38]. Our findings were similar to the two previous studies in that we did not find that EST prior to biliary stent placement can prevent PEP. This could be because thermal injury induced by EST can cause edema of the peripapillary tissue, compressing the PD orifice, which may counteract its effect of reducing tension in the PD opening [39]. Given the risks of EST, such as bleeding and perforation, we do not recommend EST for patients with MBO before biliary stent placement.

According to previous studies, SEMS is more likely to cause PEP than plastic stent (PS) because of its larger diameter and higher axial force compressing the opening of the PD, resulting in the obstruction of pancreatic juice outflow [4, 29, 40]. However, the relationship between stent types and PEP remains controversial. Martinez et al. found no difference in PEP rates when using SEMS versus PS for MBO in a recent large retrospective study of 1136 patients [41]. Moreover, in a retrospective study covering the national population in Korea, there was also no significant difference in the incidence of PEP between metal stents and PS [42]. In our study, we found no significant difference in the occurrence of PEP between SEMS and PS and between covered self-expandable metal stent (cSEMS) and uncovered self-expandable metal stent (uSEMS) as well; with many previous studies also confirming our conclusions [43,44,45]. As a result, we recommend SEMS over PS for the palliative treatment of unresectable MBO because of its longer patency time, lower stent dysfunction, and lower reintervention rates [3].

Nonetheless, our study had several limitations. First, since this was a single-center retrospective study with a relatively small sample size, the generalizability of this risk prediction model may be limited. Therefore, the validity of this model still needs to be verified by large-scale, multicenter prospective studies. Second, since this is a single-center study, the risk prediction model lacked external validation. We also hope that the findings of this study can be replicated in populations outside of China. Third, because of our hospital's current situation and the limitations of retrospective studies, we did not include operator experience or papilla type in the study. Another limitation of our study is that we did not include the outcomes of patients who had PEP.

Despite the limitations mentioned above, this is the first risk prediction model for PEP after biliary stent placement for MBO. Importantly, all risk factors for PEP included in the risk prediction model are routinely available clinical data. As a result, this predictive model can assist clinicians in identifying high-risk patients so that preventive measures can be implemented to reduce the occurrence of PEP.