Establishing a nomogram for predicting the risk factors for delayed bleeding after endoscopic submucosal dissection for colorectal tumors Authors: FuCheng Bian, KunShi Li, GuangYu Bian, XiuMei Li Published: 2024 ABSTRACT AbstractBackgroundThe aim of this study is to establish a nomogram that can predict the risk factors for delayed bleeding after endoscopic submucosal dissection (ESD). This model can be used to assess the probability of delayed bleeding before ESD surgery, thereby avoiding wasting medical resources and improving patient satisfaction.MethodsThis was a retrospective study in which all patients underwent ESD surgery for colorectal tumors between August 2021 and February 2024. Patient demographics and surgical characteristics were collected. All patients were randomly divided into a training set and a testing set. Univariate and multivariate regression analyses of the training set revealed the independent risk factors for delayed bleeding after ESD. These independent risk factors were used to construct a nomogram model. This model was validated using internal validation methods such as the C-index, calibration curve, and decision curve analysis.ResultsThis study included 587 patients. The occurrence rate of delayed bleeding after ESD in the training set was 8.98%. Multivariate regression analysis revealed that the location of the lesion in the rectum, a large lesion, and a prolonged surgery time were independent risk factors for delayed bleeding after ESD. The C-index for this model was 0.89, and validation of this nomogram model demonstrated good consistency between the predicted and actual values.ConclusionMultivariate regression analysis revealed the independent risk factors for delayed bleeding after ESD, and a nomogram with a relatively consistent accuracy was established. The clinical application of this model can reduce the incidence of delayed bleeding and therefore improve patient healing.Keywords: Delayed bleeding, Endoscopic submucosal dissection, Risk factor, Nomogram Background Colorectal cancer remains one of the most common types of cancer worldwide. Additionally, a significant number of patients die from colorectal cancer each year [1]. If clinical physicians can detect precancerous colorectal lesions early, prophylactic removal of the lesions could reduce the mortality rate of affected patients [2]. Endoscopic mucosal resection (EMR) for larger lesions may result in incomplete lesions, thereby making pathological staging difficult [3]. Endoscopic submucosal dissection (ESD) allows the complete resection of lesions regardless of size or location. This minimally invasive procedure enables the removal of entire lesions and provides intact histological specimens [4]. ESD is widely performed to treat colorectal adenomas and has been proven significantly effective in treating polyposis and lateral displacement tumors [5, 6]. However, compared with ESD for gastric diseases, colorectal ESD is relatively more challenging because of the thinness of the intestinal wall and the limited operating space [7]. Therefore, ESD requires a high level of surgical skill and a long learning curve. Delayed bleeding, delayed perforation, and electrosurgical syndrome are common complications of ESD [8, 9]. Albouys et al. [8] reported that anticoagulant use, age, lesion location, lesion size, and anesthesia were independent risk factors for delayed bleeding after ESD for colorectal adenomas and that delayed bleeding occurred in 7.98% of patients after ESD. The independent risk factors for delayed bleeding identified in these studies cannot be input into a visualization tool, thus limiting their application in clinical work. This study differs from previous studies in that a nomogram model was generated on the basis of independent risk factors for delayed bleeding after ESD identified via multifactorial regression analysis. The probability of delayed bleeding was estimated by combining the risk factor scores. We validated this model to ensure its adaptability in clinical practice with the ultimate aim of reducing the incidence of delayed bleeding. Materials and methods Patients This study included 587 patients who underwent ESD at our hospital between August 2021 and February 2024. Our hospital approved this study, and written informed consent was obtained from all patients. This study included patients with only one lesion that could be removed during a single session of ESD. Patients who underwent ESD for the simultaneous removal of multiple lesions were not included in this study. All patients were free of blood disorders and had normal preoperative coagulation test results. None of the patients used antiplatelet drugs, nonsteroidal anti-inflammatory drugs, or vasoactive drugs during the week prior to surgery. Patients with signs of deep submucosal cancerous lesions on optical examination and endoscopic ultrasound were excluded. Patients with neuroendocrine tumors or nontumor lesions who underwent ESD were excluded. Five hundred eighty-seven patients were ultimately included. These patients were divided into a training set (70%) and a testing set (30%) using a random grouping method. The nomogram model was constructed on the basis of the patients in the training set and then validated on the basis of the patients in the testing set. Definitions In accordance with Tajiri’s study, delayed bleeding after colorectal ESD in this study was defined as a postoperative hemoglobin level 2 g/dl lower than the preoperative hemoglobin level or persistent rectal bleeding, with or without endoscopic hemostasis, angiography, or surgery [10]. All patients were followed up 1 month after the ESD procedure. In the histological examination, R0 resection was defined as the absence of a tumor at the vertical or horizontal resection margins. Curative resection was characterized by a well-differentiated or moderately differentiated adenocarcinoma, cancer invasion depth less than 1000 μm, no lymphatic vessel invasion, and histologically complete resection. On the basis of endoscopic findings, mucosal fibrosis can be classified into three categories: F0 for no fibrosis, F1 for mild fibrosis, and F3 for severe fibrosis [11]. Tumors for ESD can be located in either the colon or rectum and classified as laterally spreading tumor (LST) or non-LST on the basis of their morphology [12]. Data collection Relevant risk factors were collected from the electronic medical records system and via a review of recent studies on delayed bleeding after ESD. These factors include general characteristics (age, gender, comorbidities), lesion characteristics (lesion size, location, and morphology), presence or absence of submucosal fibrosis, and surgery-related factors (surgical time and anesthesia method). The comorbidities considered included a history of hypertension, diabetes, smoking, and alcohol consumption. ESD procedure The ESD procedure was performed by four experienced endoscopists with more than 5 years of ESD experience. The surgery was performed under intravenous anesthesia or without anesthesia. The endoscope used during the procedure was from Fujifilm Corporation. Indigo carmine solution was injected into the submucosal layer to create a submucosal cushion. During the ESD procedure, the lesion was dissected first using a Dual knife (KD-655Q; Olympus), with the option of using a Hook knife (Olympus Medical Systems Co., Ltd.) and insulated-tip [IT] depending on the dissection situation. During surgery, carbon dioxide (CO2) was used for insufflation, and an irrigation pump was connected. Hemostatic forceps were used to manage the exposed vessels and bleeding points during the procedure, and clips were used to close the wound. Patients fasted for 24 h after surgery and were restricted to a liquid diet on the second postoperative day. Statistical analysis The SPSS 26 software (IBM Corp., Armonk, NY, USA) was used for the statistical analysis. Continuous variables are expressed as means ± SDs or medians on the basis of their distributions, whereas categorical variables are represented as frequencies or percentages. First, the randomly assigned training and testing sets were compared. The chi-square test or Fisher’s exact test was used for categorical data, whereas the paired Student’s t test or the Mann–Whitney U test was used for continuous data. Second, univariate and multivariate regression analyses of the training set were performed to identify the independent risk factors for delayed post-ESD bleeding in patients with colorectal cancer. Finally, the identified independent risk factors were used to construct a nomogram model using R software’s “rms” package (version 3.6.1). In this study, we evaluated the model’s value using the training and testing sets. The consistency index (C-index) and the area under the ROC curve (AUC) were used to assess the model’s predictive ability, with higher values indicating better predictive ability. The calibration curve compares the actual risk with the predicted risk. Decision curve analysis (DCA) was used to evaluate the utility of our model in clinical decision-making. Results This study included a total of 587 patients who underwent ESD. None of the patients received blood transfusion therapy or underwent surgical interventions. The patients were divided into two groups: a training set (412 cases) and a testing set (175 cases) (Table 1). There were no statistically significant differences between the two groups. Therefore, the training set was used to construct a nomogram model, and the testing set was used to validate this model. In the training set, twenty-four patients experienced intermittent minor hematochezia after ESD, which resolved with conservative treatment, and thirteen patients experienced persistent rectal bleeding, which was satisfactorily treated with endoscopic hemostatic clips (including three patients who required two endoscopic treatments and one patient who required three treatments). In the training set, 3.16% of all delayed bleeding patients required endoscopic interventions because of significant bleeding. Table 1Demographic characteristics of training set and testing setCharacteristicsTraining set (412)Testing set (175)P valueGender0.45 Male16576 Female24799Age (year)58.40 ± 10.6359.13 ± 10.520.45BMI (Kg/m2)23.05 ± 3.1222.92 ± 3.430.64Lesion size (mm)20.00 ± 6.0620.98 ± 6.760.09Procedure time (min)64.50 ± 17.9365.35 ± 17.520.60Lesion location0.12 No rectum277129 Rectum13546Type of anesthesia0.50 GA17680 LA23695Morphology0.10 Non-LST23888 LST17487Fibrosis0.71 F0, F1336145 F27630Hypertension0.50 No302133 Yes11042Diabetes0.50 No302133 Yes11042Smoking0.68 No316137 Yes9638Alcohol0.48 No324133 Yes8842BMI body mass index, GA general anesthesia, LA local anesthesia, LST laterally spreading tumor Demographic characteristics of training set and testing set BMI body mass index, GA general anesthesia, LA local anesthesia, LST laterally spreading tumor In the training set, the probability of delayed bleeding after ESD was 8.98%. There was no statistically significant difference between the delayed bleeding group and the nonbleeding group in terms of age, gender, body mass index (BMI), anesthesia method, complications, or morphology. However, there were statistically significant differences between the two groups in terms of lesion size, lesion location, operation time, and morphology (Table 2). Univariate regression analysis revealed that lesion location, size, operation time, and morphology were correlated with delayed bleeding after ESD. These factors were further analyzed through multivariate regression, which revealed that three factors (lesion location, size, and operation time) are independent risk factors for delayed bleeding after ESD (Table 3), leading to the construction of a nomogram model (Fig. 1). Table 2Preoperative demographic characteristics in training setCharacteristicsNo-delayed bleedingDelayed bleedingP valueGender0.68 Male14916 Female22621Age (year)58.09 ± 10.5561.57 ± 11.040.06BMI (Kg/m2)23.10 ± 3.1222.57 ± 3.120.33Lesion size (mm)19.22 ± 5.0227.95 ± 9.32 < 0.01Procedure time (min)62.87 ± 16.1981.05 ± 25.25 < 0.01Lesion location0.01 No rectum26017 Rectum11520Type of anesthesia0.68 GA15917 LA21620Morphology0.03 Non-LST22315 LST15222Fibrosis0.60 F0, F130729 F2688Hypertension0.11 No27923 Yes9614Diabetes0.66 No27626 Yes9911Smoking0.51 No28630 Yes897Alcohol0.65 No29628 Yes799BMI body mass index, GA general anesthesia, LA local anesthesia, LST laterally spreading tumor Preoperative demographic characteristics in training set BMI body mass index, GA general anesthesia, LA local anesthesia, LST laterally spreading tumor Table 3Univariate and multivariate logistic analysis of risk factors of delayed bleeding after ESDCharacteristicsUnivariateMultivariateOR (95%CI)P valueOR (95%CI)P valueGender (female)0.87 (0.44–1.71) 0.68Age (year)1.03 (0.99–1.07)0.06BMI (Kg/m2)0.95 (0.89–1.06)0.33 Lesion size (mm)1.19 (1.13–1.24) < 0.011.17 (1.10–1.23)< 0.01Procedure time (min)1.05 (1.03–1.07)< 0.011.05 (1.03–1.07) < 0.01Lesion location2.66 (1.34–5.27)0.014.09 (1.76–9.53) < 0.01Type of anesthesia0.87 (0.44–1.71) 0.68Morphology2.15 (1.08–4.28) 0.031.01 (0.42–2.46)0.98Fibrosis1.25 (0.56–2.84)0.60Hypertension1.77 (0.88–3.58) 0.11Diabetes1.18 (0.56–2.48)0.66Smoking0.75 (0.32–1.77)0.51Alcohol1.20 (0.55–2.66)0.65ESD endoscopic submucosal dissection, BMI body mass index, GA general anesthesia, LA local anesthesia, LST laterally spreading tumor Univariate and multivariate logistic analysis of risk factors of delayed bleeding after ESD ESD endoscopic submucosal dissection, BMI body mass index, GA general anesthesia, LA local anesthesia, LST laterally spreading tumor Nomogram for predicting delayed bleeding after ESD. Each risk factor was assigned a point, which was summed to give a total point that corresponded to the probability of the hazard on the bottom row of the figure according to the total point In this study, we validated the model using internal validation methods. The C-index of the training set was 0.89 (CI: 0.87–0.94), indicating that the model has high accuracy. ROC curves were drawn for both the training and testing sets, with calculated AUC areas of 0.89 and 0.87, respectively, indicating the good discriminatory ability of the model (Fig. 2). Decision curve analysis (DCA) revealed that when the threshold was between 1% and 70%, the net benefit rate of the model constructed in this study increased, indicating better clinical utility (Fig. 3). Calibration curves demonstrated good consistency between the actual and predicted probabilities of delayed bleeding after ESD (Fig. 4). Comparison of the area under the receiver operating characteristic curve between nomogram independent predictors in the training set A and the testing set B Comparison of decision curve analyses between the training set A and the testing set B Comparison of calibration curves between the training set A and the testing set B A well-performing model can be conveniently applied in clinical settings. The model in this study calculates a total score by summing the corresponding scores from independent risk factors. The probability of delayed bleeding after ESD is subsequently determined on the basis of the total score. For example, a 68-year-old male patient with a rectal lesion measuring 30 mm who underwent a 90-min surgery had scores of 23, 50, and 47, respectively, in the model. The total combined score was 120, resulting in a 63% probability of delayed bleeding after ESD. Discussion In a global context, as the number of colorectal adenoma patients undergoing ESD increases annually, the incidence of postoperative complications is also increasing. Previous studies have shown that the incidence of delayed bleeding after ESD ranges from 4.7 to 12% [8, 13–16]. In this large single-center retrospective study, we analyze the risk factors for delayed bleeding after ESD for patients with colorectal adenomas. In the present study, we identified the independent risk factors for delayed bleeding after ESD, such as lesion location, lesion size, and duration of surgery, with a probability of delayed bleeding in the training set at 8.98%, similar to previous studies. Furthermore, in this study, we identified the independent risk factors for delayed bleeding through multifactorial regression analysis and developed the first nomogram model. Validation confirmed the model’s efficiency in clinical applications, ultimately aiming to effectively reduce the likelihood of delayed bleeding postoperatively. The location of colorectal adenomas has always been a relevant risk factor for delayed bleeding after ESD [7, 8, 17, 18], especially when the lesion is in the rectum [19], which is consistent with previous research findings and the current study. However, Suzuki et al. [20] discovered that 4.3% (14/325) of lesions experienced delayed bleeding after ESD, indicating that lesions in the caecum are an independent risk factor. In 2018, Seo et al. [17] retrospectively analyzed data from 5 centers, reported that lesions in the rectosigmoid colon, a tumor size ≥ 30 mm, and antiplatelet are independent risk factors for post-ESD bleeding, and established a model on the basis of these three risk factors. A study conducted in 2023 revealed that the risk of delayed bleeding after ESD was 7.98%, with a multifactorial analysis indicating five independent risk factors for post-ESD bleeding, including lesions in the rectum [8]. The primary reasons for delayed bleeding are anatomical factors in the rectum, characterized by several rich arterial blood supplies [21]. Furthermore, the submucosal layer in the rectosigmoid colon is thick, and incomplete coagulation of vessels during the handling process due to the thickness of this layer or difficulty in adequately coagulating vessels may increase the risk of delayed bleeding. Finally, lesions in the rectosigmoid colon can lead to iatrogenic ulcers due to stimulation from the remaining hard-formed stool [22]. The results from the latest study indicate that delayed bleeding postoperatively is a significant complication following colorectal ESD. Delayed bleeding is correlated with the size of the lesion [18]. Our research identified a large lesion as an independent risk factor for delayed bleeding after ESD, which is consistent with the findings of previous studies [7, 8, 17, 23, 24]. However, a study in 2020 revealed that hypertension and the use of hot biopsy forceps were independent risk factors for delayed bleeding, with multifactorial analysis indicating that lesion size is not a risk factor [13]. This might be due to the relatively small sample size. In a recent study involving 940 patients, 75 (7.98%) experienced delayed bleeding, and 95 (10.1%) experienced perforation [8]. It was also found that a large lesion is an independent risk factor for delayed bleeding. The study revealed a positive correlation between the probability of delayed bleeding and lesion size [17]. This is because larger lesions result in larger post-ESD ulcers, potentially prolonging the healing time and increasing the blood vessel exposure time, thereby increasing the likelihood of delayed bleeding after ESD. In this study, we found that the independent risk factor for delayed bleeding after ESD was procedure time (OR: 1.05 CI: 1.03–1.07). A study by Yamamoto et al. [25] involving 398 lesions in Japan revealed that the surgical time in patients who experienced delayed bleeding after ESD was significantly longer than that in who did not experience such bleeding after ESD (P = 0.006), which is consistent with the results of this study (P < 0.01). Although regression analysis revealed that fibrosis is not an independent risk factor for delayed bleeding, severe fibrosis can increase the difficulty of submucosal injection and prolong the operation time [26]. Related studies have also revealed that longer operation times are more likely to result in delayed bleeding [7]. Another study in Japan revealed that arterial bleeding occurring three or more times during ESD and lesions located in the rectum are independent risk factors for delayed bleeding after the operation, and the operation time in the delayed bleeding group was significantly longer than that in the nonbleeding group [27]. Therefore, the prolongation of operation time is likely due to the management of large blood vessels and frequent changes in surgical instruments during the operation. In this study, we established a nomogram model on the basis of a relatively large sample size. The model was first applied to delayed postoperative bleeding after colorectal ESD. The model is able to predict the risk of postoperative bleeding during the perioperative period. The size and location of the lesion were assessed preoperatively, and the operative time was shortened as much as possible during surgery to reduce the incidence of intraoperative bleeding. Application of this model can enhance the perioperative management of patients and reduce the incidence of delayed bleeding after ESD. This study has several limitations. First, like most studies, this was a retrospective study, lacking support from prospective studies. Second, this was a single-center study, with a lack of diversity in the included patients and a lack of validation from external data. Thus, multicenter studies are needed as soon as possible. Conclusions Our study revealed that lesions located in the rectum, large lesions, and a prolonged surgery time are independent risk factors for delayed bleeding after colorectal ESD. On the basis of these three independent risk factors, we developed a nomogram predictive model that can predict the probability of delayed bleeding after surgery, thus reducing the need for repeat endoscopic treatment in patients. Author contributions Author FC B and XM L designed the study. Author FC B and KS L collected the clinical data and conducted the statistical analysis. Author FC B and GY B wrote the manuscript. Author FC B, KS L and XM L revised the manuscript; All authors critically read the manuscript to improve intellectual content. All authors read and approved the final manuscript. Data availability The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request. Declarations Ethics approval and consent to participate This study was performed in line with the principles of the Declaration of Helsinki. Approval was granted by the Daqing oil field General Hospital. Consent for publication All participants agreed with the data and publication of the manuscript. The authors affirm that human research participants provided informed consent for publication of the images. 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. FuCheng Bian, KunShi Li and GuangYu Bian contributed equally to this work. References Associated Data Data Availability Statement The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.