All procedures in the study were performed according to the principles of the Declaration of Helsinki and by the guidelines of Strengthening the Reporting of Surgical Cohort Studies (STROCSS). All patients signed an informed consent form, which was approved by the Ethics Committee of the Second Hospital of Hebei Medical University (2019-R209). To protect the privacy of our patients, all data were anonymized by removing sensitive personal information.

A total of 407 hospitalized patients who underwent LPD at the Department of Hepatobiliary Surgery in the Second Hospital of Hebei Medical University between January 2020 and August 2023 for periampullary carcinoma were selected as the objects. The exclusion criteria are: patients with (1) incomplete clinical data; (2) hematological diseases; (3) malignant tumors of other organs; (4) autoimmune disease; (5) long-term use of non-steroidal drugs; (6) non-cooperation with postoperative treatment (Figure 1). This was done to keep the excellent population homogeneous and to allow straightforward interpretation of results.

The data included demographics, medical comorbidities, treatment-related, laboratory biomarkers, and postoperative complications. The demographic data includes age, gender (male or female), severe addiction (alcohol or smoking), BMI. Comorbidities included hypertension, diabetes mellitus, coronary heart disease, hepatitis, liver cirrhosis, pancreatitis, previous abdominal surgery. Treatment-related variables included percutaneous transhepatic cholangial drainage (PTCD), blood loss during the surgery, transfusion and operation time. Preoperative laboratory rusults included red blood cell count (RBC), white blood cell count (WBC), platelet count (PLT), albumin (Alb), total bilirubin (TBIL); alanine transaminase (ALT), aspartate transaminase (AST), fibrinogen (FIB), international normalized ratio (INR), creatinine (Cr), prothrombin time (PT). They activated partial thromboplastin time (APTT). Postoperative complications included clinically relevant postoperative pancreatic fistula (CR-POPF), bile leak, intra-abdominal infection (IAI), delayed gastric emptying (DGE), ileusand chyle leak.

ISGPS defined POPF as an amylase level in the abdominal drainage fluid that exceeds three times the upper limit of the average serum value for 3 days or more after surgery. ISGPS further categorized POPF into grades A to C based on the severity of the condition. Grade A represents biochemical fistulae, grade B indicates POPF requiring clinical treatment or intervention, and grade C signifies POPF combined with organ failure or life-threatening situations necessitating surgical intervention. Grades B and C POPF are collectively called clinically relevant POPF (CR-POPF) due to their significant association with prognosis (27). Three days after pancreaticoduodenectomy (PD), the bilirubin concentration in the drainage fluid exceeds three times the concentration in the serum. In such cases, interventional drainage or surgical intervention is necessary to address bile accumulation or diagnose cholestatic peritonitis, indicating a bile leak (28). Intra-abdominal infection (IAI) is characterized by symptoms such as chills, high fever, abdominal distension, and intestinal paralysis, persisting for more than 24 h. Laboratory tests show a significant increase in white blood cell count, hypoproteinemia, and anemia, while imaging studies reveal the presence of fluid in the abdominal cavity. Diagnosis can be confirmed by aspirating pus or detecting bacteria (29). In accordance with the ISGPS, delayed gastric emptying (DGE) is the inability to resume a regular diet by the end of the first week after surgery, including patients requiring prolonged nasogastric tube usage (30). The ISGPS defines chyle leak as the discharge of chyle-like fluid from drains, drain orifices, or wounds occurring on or after the third day following surgery if the concentration of triglycerides exceeds 1.2 mmol/L regardless of the volume (31).

Pneumoperitoneum was created through a routine 5-hole approach. Subsequently, the operation was divided into two distinct phases: resection and reconstruction. The resection phase included the following steps: (1) performing a Kocher incision, (2) severing the stomach, (3) dissecting the hepatoduodenal ligament, (4) amputating the pancreas, (5) severing the jejunum, (6) dissecting the superior mesenteric vein-portal vein system, and (7) dissecting the superior mesenteric artery-celiac trunk system. The arterial or venous approach selection during surgery is contingent upon the tumor's positioning in relation to the blood vessels. After the resection, a 5-cm specimen was extracted through an incision located beneath the xiphoid process of the upper abdomen, and pneumoperitoneum was reinstated. The digestive tract reconstruction was accomplished using the Child method, which involved implementing various techniques, including the end-to-end mucosal-to-mucosal anastomosis method for pancreatic enterostomy. In this technique, the placement of the pancreatic duct stent was determined by intermittently or continuously suturing it, considering the diameter of the duct. Choledochojejunostomy was performed using an end-to-side anastomosis technique—the anastomosis procedure involved a linear cutting occluder, followed by manual suturing of the joint opening. Two standard abdominal drainage tubes were placed during the surgery. One tube was positioned above the pancreaticojejunostomy and drained through the left abdominal wall, while the other tube was placed below the choledochojejunostomy and drained through the right abdominal wall. No additional drainage methods were employed in this particular case.

The definition of intraluminal post-pancreatectomy hemorrhage (IPPH) adhered strictly to the guidelines set forth by the International Study Group of Pancreatic Surgery (ISGPS). The ISGPS classification system distinguishes between intraluminal and extraluminal types of hemorrhage based on the location of the bleeding. Within the category of IPPH, various manifestations (e.g., pancreatic surface hemorrhage, peri-anastomotic ulcers, erosions, and biliary hemorrhage) are encompassed (31). The identification of the hemorrhage site was accomplished by gastroscopy or cesarean section.

Data were analyzed using SPSS 27.0 (IBM). Measurements were expressed as x ± s. In normal distribution, the independent sample t-test was utilized to compare the two groups, while the rank sum test of two samples was employed for non-normal distribution. Counting data was analyzed using either the χ² test or the Fisher exact test. Multivariate logistic regression analysis was performed to identify independent risk factors for predicting IPPH among variables with P < 0.10. Subsequently, a nomogram was developed based on these factors using R Programming Language V.4.2.1 (R Foundation). The accuracy, sensitivity and specificity of the nomogram were assessed by the area under the curve (AUC) using receiver operating characteristic (ROC) analysis. Besides, decision curve analysis (DCA) was utilized to investigate the clinical predictive significance of different nomogram variables about the occurrence of IPPH.

A total of 326 patients out of 407 who underwent LPD were deemed eligible for inclusion criteria in this study. Among the enrolled participants, 200 (61.35%) were male, and 126 (38.65%) were female, with a mean age of 59.76 ± 11.24. After LPD, a complication rate of 62.88% (205/326) was observed. Among these patients, 23 were confirmed to have IPPH with a prevalence rate of 7.06% (23/326). IPPH involved the surface of pancreas in 8 (34.78%,8/23), the surface of gastrointestinal anastomosis in 14 (60.87%,14/23), and ulcer of choledochojejunostomy complicated with erosive hemorrhage in 1 (4.34%, 1/23).

After univariate analysis, seven potential predictors were identified in all variables. Then, this study used multivariate regression analysis to examine the independent risk factors of IPPH (Tables 1, 2). The findings indicate that advanced age (OR = 1.065, 95% CI = 1.001–1.133, P = 0.045), low Fib level (OR = 0.485, 95% CI = 0.242–0.972, P = 0.041), low Alb level (OR = 0.840, 95% CI = 0.739–0.956), and CR POPF (OR = 4.300, 95% CI = 1.347–13.722) were associated with an increased incidence of IPPH. IAI (OR = 6.347, 95% CI = 1.454–27.716, P = 0.014) was also identified as a significant factor. The differences in other factors were not statistically significant (P > 0.05).

The prediction nomogram was constructed based on multivariate analysis findings (Figure 2), offering a more user-friendly and practical instrument for clinical implementation. The area under the curve (AUC) was determined to be 0.861 (95% CI = 0.783–0.939), indicating a substantial degree of sensitivity (82.6%) and specificity (82.2%), thereby attesting to the discriminatory capacity of the model (Figure 3). Furthermore, the c-index and Brier score were computed as 0.862 and 0.052, respectively.

When utilizing the nomogram, the designated variable axis was employed to position the level of each variable. Subsequently, a vertical line was drawn along the axis from the corresponding point to ascertain the risk score. This procedure was iterated for each variable, culminating in the calculation of a cumulative score. The ultimate sum was then located on the “total points” axis, and a vertical line was drawn to intersect the probability axis, thereby yielding the projected probability of experiencing IPPH.

To validate the net benefit for IPPH of patients who underwent LPD, a decision curve analysis (DCA) and a Hosmer-Lemeshow good of fit test were performed on the prediction model. The Hosmer-Lemeshow X2 statistics of the calibration curve (Figure 4) illustrated a favorable consistency between the probability of predicting postoperative intraluminal hemorrhage and the actual probability of occurrence in patients undergoing LPD. As could be seen in Figure 5, it was clear suggested that the model enhanced the net benefit of the “treat all” or “no treatment” scenario when the threshold probability was between 19% and 100%. The performance of the DCA showed that the model can guide clinical practice well.