Introduction

Colorectal polyps are protruding growths resulting from the excessive proliferation of the intestinal mucosa into the intestinal lumen, which can develop into colorectal cancer through the adenoma-carcinoma sequence pathway1. The prevalence of this condition rises with age, and research indicates that it can reach as high as 12.95% among individuals aged over 40 years2. Endoscopic surgery remains the primary treatment modality for colorectal polyps, However, it is associated with adverse events including bleeding, perforation, and postoperative coagulation syndrome3. Among these, bleeding is particularly common, and postoperative bleeding is classified into immediate postoperative bleeding (IPPB) and DPPB. DPPB is characterized by bleeding occurring within 30 days post-surgery, necessitating additional visits to the emergency room or hospitalization4.The incidence is between 0.3% and 2.8%5,6,7. Clinicians have paid significant attention to the insidious onset of DPPB due to its potential to be life-threatening, prolong hospital stays, and escalate medical costs, particularly in cases of heavy bleeding. Identifying potential independent risk factors for DPPB through screening and conducting individualized assessments for patients are crucial steps in preventing DPPB.

Previous research has identified several risk factors linked to the development of DPPB. However, discrepancies exist between studies regarding the controversy surrounding certain factors, while others have not been sufficiently investigated7,8,9,10,11,12,13. Some studies utilize risk factors and static Nomogram risk maps to achieve quantitative assessment13,14. However, clinicians still encounter challenges in implementing the assessment process. Due to its simplicity, convenience, maneuverability, and high accuracy, the dynamic nomogram finds broad utility in predicting acute respiratory distress syndrome following extracorporeal circulation and community-acquired pneumonia complicating acute kidney injury15,16,17. However, there is a scarcity of studies addressing this issue in the ___domain of DPPB. This study aims to identify risk factors for DPPB, develop a dynamic nomogram for predicting DPPB risk post-colorectal polyp surgery, aid clinicians in identifying high-risk patients, enabling individualized diagnosis and treatment. Moreover, it seeks to implement preoperative interventions and postoperative prophylaxis to mitigate DPPB occurrence.

Patients and methods

Patient data collection

Informed consent was not required for this study because this study is retrospective research. All protocols received approval from the Ethics Committee of the First Hospital of Lanzhou University and were conducted in accordance with its regulations and guidelines.

The study population comprised patients who underwent endoscopic colorectal polypectomy at the First Hospital of Lanzhou University between January 2020 and March 2023. Data collection encompassed three main aspects: (1) Baseline Characteristics: age, gender, body mass index (BMI), smoking, drink, hypertension, diabetes, chronic liver disease, chronic kidney disease, use of direct oral anticoagulants (DOACs), and history of gastrointestinal surgeries. (2) Polyp and Surgical Details: the number, diameter, ___location, morphology, and pathological staging of the polyps, as well as the use of prophylactic hemoclips, Operative modality and operator experience. (3) Preoperative Examination Parameters: PLT, RBC, HGB, HCT, MCV, MCH, MCHC, WBC, L, N, M, CRP, systemic immune-inflammation index (SII), systemic inflammation response index (SIRI), systemic immune-inflammation response index (SIIRI), TC, TG, HDL-C, and fecal occult blood (FOB). A reference questionnaire was utilized to assess smoking and drink habits. Polyp ___location was classified into three categories: right semicolon (ileocecal junction, ascending colon, and transverse colon), left semicolon (descending colon, sigmoid colon, and rectum) and mixed (indicating polypectomy performed in both the right and left colon). Polyp diameter was estimated by the operator based on the opening width of the biopsy forceps. Polyp pathological staging was divided into adenomatous polyps (tubular adenomas, tubular villous adenomas, and choriocarcinomas) and non-adenomatous polyps (hyperplastic polyps and serrated polyps). Surgical procedures encompassed APC, EMR, ESD, and CSP. Polyps were divided into sessile polyps and pedunculated polyps by morphologic diagnosis. Operator experience was categorized based on the number of gastroenterostomy units operated within a five-year period.

Inclusion and exclusion criteria

Inclusion criteria: (1) Diagnosis of colorectal polyps confirmed by endoscopic examination and pathology; (2) DPPB diagnosis adhering to the European Society of Endoscopy criteria. Exclusion criteria: (1) Concurrent hematologic disorders; (2) History of inflammatory bowel disease or familial adenomatous polyposis; (3) Severe cardiac, pulmonary, or cerebral conditions, or tumors in other locations; (4) Gastrointestinal bleeding from alternate etiologies; (5) Abnormal preoperative coagulation profiles; (6) Current use of antiplatelet or anticoagulant medications, including clopidogrel; (7) Presence of IPPB; (8) Incomplete clinical or pathological data.

Research design

This study collected data from 1591 patients, with 1544 ultimately meeting the inclusion and exclusion criteria. These individuals were divided into the training set (N = 1081) and the internal validation set (N = 463) at a 7:3 ratio. (Fig. 1).

Fig. 1
figure 1

Flowchart of the enrollment process.

Full size image

Statistical analysis

Numerical variables were assessed using either the t-test or Mann–Whitney U-test, while categorical variables underwent analysis using either the chi-square test or Fisher's exact test. Initial variable screening employed lasso regression. Subsequently, all variables with non-zero coefficients were incorporated into multifactorial logistic regression to identify independent DPPB risk factors. R software facilitated the creation of Nomogram and supported dynamic network interactions. ROC analysis gauged the predictive model's discrimination, with the Area Under the Curve (AUC) calculated for assessment. Calibration curves evaluated model calibration, and DCA analyzed the predictive model's clinical utility. A two-sided test was utilized, with statistical significance set at P < 0.05.Data were analyzed using IBM SPSS Statistics Version 25.0 (IBM Corp, Armonk, NK, USA) and R version 4.1.3 (http://www.r-project.org).

Results

Over the period from January 2020 to March 2023, our study center conducted a total of 1591 endoscopic colon polypectomies. Among these cases, DPPB was observed in 38 patients (2.39%), aligning with findings from prior research. All individuals experiencing DPPB underwent repeat colonoscopy to identify the bleeding site and received appropriate treatment. Specifically, two patients required a second endoscopic hemostasis, while the remainder achieved successful hemostasis with a single intervention.

Univariate analysis

Baseline information of included patients

Statistically significant differences were observed between patients in the DPPB group and those without DPPB across three factors: drink, diabetes, and DOCAs (P < 0.05), while no significant differences were found for other variables (Table 1).

Table 1 Baseline characteristics of study patients.
Full size table

Analysis of polyp and surgery-related factors

Significant differences were found between patients with DPPB and those without DPPB regarding polyp diameter, polyp ___location, operative modality, and the prophylactic hemoclips (P < 0.05). No other statistically significant differences were observed (Table 2).

Table 2 Analysis of polyps and surgery related factors.
Full size table

Analysis of preoperative tests

Significant differences were observed between patients with DPPB and those without DPPB in PLT, HGB, HCT, TC, HDL-L and FOB(P < 0.05), while no other statistically significant differences were found (Table 3).

Table 3 Factors associated with relevant preoperative tests.
Full size table

The results of univariate analysis revealed associations between the occurrence of DPPB and the following factors: drink, diabetes, DOACs, polyp ___location and diameter, use of prophylactic hemoclips, operative modality, as well as levels of PLT, HGB, HCT, TC, HDL-C, and FOB.

Factor selection for the predictive model and dynamic nomogram developing

The Lasso regression initially screened 13 variables. Shrinkage coefficients for these variables were plotted (Fig. 2), and cross-validated graphs were generated (Fig. 2). From cross-validation results, seven variables (Drink, DOACs, Polyp Location, Polyp Diameter, prophylactic hemoclips, HGB, FOB) exhibited non-zero coefficients in the Lasso regression analysis, minimizing the model error to 0.0048 (Fig. 2). Furthermore, the correlation heatmap indicated no multicollinearity among these seven factors (Fig. 2). Subsequently, multifactorial logistic analysis (Table 4) identified five independent predictors (Drink, DOACs, Polyp Location, Polyp Diameter, Prophylactic hemoclips). These five risk factors were then used to construct visualized nomogram (Fig. 3) and dynamic interactions, accessible via a web page. (http://ldyy-xhnj.shinyapps.io/dynnomapp) (Fig. 3).

Fig. 2
figure 2

(a) Lasso coefficient profile plot. (b) Tuning parameter (λ) selection in the LASSO model used tenfold cross-validation via minimum criteria. (c) Correlation between all variables in the LASSO model visualized by the diameters of the spots.

Full size image
Table 4 Multiple logistic regression analysis for risk factors of delayed post-polypectomy bleeding.
Full size table
Fig. 3
figure 3

(a) The nomogram constructed from the full model was used to predict DPPB.(b) A dynamic nomogram is deployed on the shinyapps.io platform http://ldyy-xhnj.shinyapps.io/dynnomapp.Users can calculate the probability of a poor prognosis for DPPB patients by submitting five related traits online via mobile phone or compute.

Full size image

Evaluation of the model

ROC curves were employed to assess the discriminative abilities of the model (Fig. 4). The AUC values for the model were 0.936, 0.796, and 0.865 in the training, testing and entire cohorts, respectively, surpassing those of the individual variables. Calibration curves were utilized to confirm the model's consistency (Fig. 4), demonstrating that the predictions from the column-line plot model aligned well with actual observations across all cohorts. Additionally, DCA curves illustrated the net clinical gain (Fig. 4), indicating favorable clinical net gains across all cohorts. Moreover, the overall model exhibited superior performance compared to the individual variable model across all threshold probabilities from 0–100%.

Fig. 4
figure 4

(a) ROC curves depicting the occurrence of DPPB for individual risk factors and the entire model in the training cohort. (b) ROC curves depicting the occurrence of DPPB for individual risk factors and the entire model in the testing cohort. (c) ROC curves depicting the occurrence of DPPB for individual risk factors and the entire model in the entire cohort. (d) Calibration curves in the training cohort. (e) Calibration curves in the testing cohort. (f) Calibration curves in the entire cohort. (g) DCA curves illustrating the occurrence of DPPB for individual risk factors and the entire model in the training cohort. (h) DCA curves illustrating the occurrence of DPPB for individual risk factors and the entire model in the testing cohort. (i) DCA curves illustrating the occurrence of DPPB for individual risk factors and the entire model in the entire cohort.

Full size image

Ethics statement

Informed consent was not required for this study because this study is retrospective research. All protocols received approval from the Ethics Committee of the First Hospital of Lanzhou University and were conducted in accordance with its regulations and guidelines.

Discussion

DPPB represents a significant adverse event following colorectal polypectomy, with most cases occurring within 1 to 6 days postoperatively3. This analysis identified several independent risk factors for DPPB, including DOACs, polyp diameter, polyp ___location, prophylactic hemoclips, and drink. Furthermore, it developed a dynamic nomogram prediction model for DPPB, demonstrating robust predictive efficacy. This model facilitates individualized clinical diagnosis and treatment, thereby enhancing clinical decision-making.

This study identified DOACs as a significant risk factor for the development of DPPB following colorectal polypectomy, aligning with findings from previous studies10,18,19.Whereas Yu et al. found no significant increase in gastrointestinal bleeding (GIB) after polypectomy in patients taking DOACs in a retrospective analysis of the Clinformatics Data Mart Database20.Current guidelines are unclear on the discontinuation time and postoperative recovery time for taking DOACs21,22,23. The occurrence of DPPB might relate to inadequate perioperative discontinuation time or early postoperative medication resumption. Future investigations are necessary to delve into the underlying mechanisms and ascertain the optimal perioperative dosage and timing for medication resumption in high-risk patients.

The diameter of polyps has long been acknowledged as an independent risk factor for DPPB. In a retrospective study, Lu et al. discovered a 4.035-fold higher risk of hemorrhage associated with polyps larger than 10 mm compared to those smaller than 10 mm13. Bae and colleagues discovered that the risk of delayed post-polypectomy bleeding (DPPB) escalated by 11.6% with each 1 mm increment in polyp size8. In our investigation, polyps in the DPPB group exhibited a larger diameter compared to those in the non-DPPB group. This phenomenon is likely attributable to the increased vascularization associated with larger polyps, thereby enhancing the susceptibility to vascular injury during manipulation14.

Polyps located in the right semicolon were an independent risk factor for DPPB in this study, consistent with previous studies8,9,13. A plausible explanation for this phenomenon is the susceptibility of the thin, high-tension wall of the cecum to vascular injury within the deeper submucosal layers during manipulation. Additionally, the influx of digestive enzymes and bile acids from the ileal fluid into the cecum could dissolve blood clots that cover post-polypectomy trauma, thereby elevating the risk of DPPB9,24, Moreover, the constrained space and challenging maneuverability within the right half of the colon invariably result in mucosal damage when polyps are abundant.

The association between prophylactic hemoclips and the onset of DPPB remains contentious. Interestingly, in our investigation, the utilization of prophylactic hemoclips was associated with a heightened incidence of DPPB, aligning with findings reported by Okinawa et al25. Several findings indicate that the prophylactic use of tourniquets is not correlated with the incidence of DPPB26,27. While the majority of studies advocate for the use of prophylactic hemoclips to decrease the incidence of DPPB28. Potential explanations include ineffective clamping positions or dislodgement of titanium clips29. Furthermore, the decision to clip relies on the discretion of the overseeing endoscopist. In cases where the risk of DPPB is deemed substantial, there may be a heightened reliance on titanium clamps, resulting in inevitable mucosal damage and an escalated DPPB risk. Limited research exists on this topic, and future inquiries could delve deeper into evaluating the efficacy of titanium clamping in wound closure for DPPB prevention, as well as its susceptibility to individual variables within endoscopic practice.

Previous correlation studies that included drink did not identify statistically significant associations7,12,30,Our study identified drink as an independent risk factor for DPPB. Some studies have demonstrated that chronic heavy alcohol consumption can impair the intestinal mucosal barrier, resulting in heightened intestinal permeability and microbial translocation31.Nonetheless, the precise mechanism remains elusive. Asquith et al. observed in an animal experiment that chronic drinking (CD) could substantially disturb mucosal homeostasis via microRNA expression32. Barr et al. identified region-specific alterations in the microbial composition of mucosal tissues in individuals with chronic alcohol consumption. These alterations included a reduction in beneficial bacteria, such as Brucella. in the jejunum and Ruminococcaceae. in the ileum, and an elevation in inflammation-associated bacteria, such as Porphyromonas and Prevotella33. The inflammatory response in the intestines exacerbates pathological damage34. Hence, we hypothesized that the onset of DPPB might correlate with inflammatory responses, impairment of the intestinal mucosal barrier, and alterations in intestinal flora due to alcohol consumption. Future investigations are warranted to elucidate the relationship and mechanisms linking DPPB and alcohol consumption.

One of the strengths of this study is establishing a dynamic nomogram prediction model with strong predictive performance based on DPPB risk factors. The identified predictive parameters are routine in the examination of patients with colorectal polyps, making the model user-friendly and widely applicable. Accessible via computers, cell phones, and other electronic devices, it serves as a convenient navigational tool for preoperative tendency diagnosis, personalized therapeutic measures, and surgical modality selection. Additionally, the study identifies alcohol consumption as an independent risk factor for DPPB, warranting further investigation in future studies.

However, this study has several limitations. Firstly, it is a retrospective single-center study, necessitating future multicenter and prospective studies to validate the predictive accuracy and clinical utility of dynamic columnar mapping. Secondly, the study primarily included multiple small polyps and did not assess the impact of specific locations on DPPB occurrence.

Conclusion

This study developed a straightforward and effective online prediction model for DPPB, providing clinicians with a tool to identify high-risk patients and enabling personalized diagnosis and treatment.