Skip to main content
Download PDF

Machine learning prediction model of prolonged delay to loop ileostomy closure after rectal cancer surgery: a retrospective study

World Journal of Surgical Oncology volume 23, Article number: 185 (2025) Cite this article

Abstract

Background

Delayed closure of a temporary ileostomy in patients with rectal cancer may cause psychological, physiological, and socioeconomic burdens to patients.

Purpose

This study aimed to develop and validate a machine learning-based model to predict the delayed ileostomy closure after surgery in patients with rectal cancer.

Design

A retrospective study.

Methods

LASSO regression was used for feature screening, and XGBoost was used for machine learning model construction. Model performance was assessed by receiver operating characteristic (ROC) curve analysis, calibration curve analysis, clinical decision curve analysis, sensitivity, specificity, accuracy, and F1 score. The SHAP method was used to interpretate the results of the machine learning model.

Results

A total of 442 rectal cancer patients who received a loop ileostomy were included in this study, and 305 experienced delayed closure (69%). The XGBoost model area under the ROC curve (AUC) of the training set was 0.744 (95% confidence interval [CI]: 0.686–0.806) and of the test set was 0.809 (95% CI: 0.728–0.889). The importance of each variable, in descending order was body mass index (BMI), postoperative chemotherapy, distance from tumor to anal margin, depth of tumor infiltration, neoadjuvant chemoradiotherapy, and anastomotic stenosis. The importance of SHAP variables in the model from high to low was: ‘BMI’ ‘postoperative chemotherapy’ ‘distance of the tumor from the anal verge’ ‘depth of tumor infiltration’ ‘neoadjuvant radiotherapy’ ‘anastomotic stenosis’.

Conclusion

The XGBoost machine learning model we constructed showed good performance in predicting delayed closure of loop ileostomy in rectal cancer patients. In addition, the SHAP method can help better understand the results of machine learning models.

Background

Rectal cancer (RC) is one of the most common malignant digestive tract tumors. colorectal cancer (CRC) ranks third globally in incidence and is the second leading cause of cancer-related death [1]. Total mesorectal excision (TME), proposed by Heald et al. in 1982 [2], remains the standard for radical RC resection. Its basic principle is to completely remove the mesentery surrounding the tumor at the anatomical and embryonic levels. Over time, techniques have advanced, with laparoscopic transabdominal tumor excision (laTME) becoming widely used for resectable cases [3]. After rectal cancer surgery, patients may experience postoperative complications, including anastomotic leakage, which can influence recovery, increase healthcare utilization, and potentially affect oncological outcomes [4–5]. To reduce the risk of clinical consequences association with leakage, a diverting ileostomy is often performed at the time of intestinal anastomoses [6].

However, the effectiveness of a temporary ileostomy in preventing or reduceing anastomotic leakage remains debated. Many studies support its role in preventing anastomotic leakage [7–8]. Garg et al. [9] conducted a meta-analysis of 5 randomized controlled trials including 768 patients, showing that ileostomy significantly reduced the leakage and reoperation rates. In recent decades, prophylactic stoma has been widely used in anal sphincter preservation surgery, especially for low rectal cancer [10]. Conversely, other studies suggest that ileostomy may not reduce leakage but can reduce the risk of abdominal infection [11,12,13]. Currently, no standardized guideline exist for whether patients should undergo temporary ileostomy. To reduce postoperative complications, the use of temporary preventive ileostomies has increased, leading to unnecessary stoma related complications in patients without anastomotic leakage.

Temporary ileostomy can negatively affect patients’ psychological well-being and is associated with various stoma-related complications [14]. These include intestinal obstruction, diarrhea, anastomotic leakage, incision infection, and enterocutaneous fistulae [15]. An international study involving 279 ileostomy patients reported an overall complication rate of 83% [16]. As such, a temporary ileostomy should be closed as soon as possible. Delayed closure beyond 6 months is associated with a 3.7-fold increased risk of severe intestinal dysfunction after restoring intestinal continuity [17]. In addition, the incidence of stoma-related complications increases with prolonged closure intervals [18]. Patients who undergo ileostomy closure more than 3 months after rectal surgery experience a significant decline in their quality of life [19–20] and higher healthcare cost [21]. Given the variability in closure timing, a predictive model to identify patients at risk of delayed closure would aid clinical decision making.

Machine learning (ML) is increasingly used in healthcare to assess patient risk and support decision making [22,23,24]. SHapley Additive exPlans (SHAP) is an advanced, interpretable ML framework designed to provide in-depth explanations for the predictions of any ML model [25–26]. Other stud using broader population-based datasets have found it more challenging to achieve accurate prediction [27]. ML is currently widely used in the diagnosis, treatment, and prediction of cancer recurrence, and models have practical applications [28,29,30]. Therefore, this study aimed to develop an interpretable ML model using SHAP to predict the risk of delayed closure of an ileostomy after rectal cancer surgery.

Methods

Study setting and patients

An observational prospective cohort study was conducted in a large oncology hospital in China. The data of patients who underwent TME for rectal cancer and received a temporary ileostomy at the Colorectal Surgery Department of Sun Yat-sen University Cancer Center from January 1, 2022 to December 31, 2023 were retrospectively collected and reviewed. This study was approved by the Ethics Review Committee of Sun Yat-sen University Cancer Center. This study has obtained informed consent from the participants and signed the informed consent form.

Inclusion criteria and exclusion criteria

Inclusion criteria were: (1) Diagnosed with rectal cancer by colonoscopy and pathological examination of a biopsy tissue specimen; (2) Received an ileostomy; (3) Regular follow-up at Sun Yat-sen University Cancer Center; (4) No tumor recurrence during follow-up. Exclusion criteria were: (1) Diabetes mellitus or hypertension; (2) Ongoing steroid treatment; (3) Immune or inflammatory diseases such as inflammatory bowel disease and autoimmune diseases; (4) Concomitant small intestine or colon resection during surgery; (5) Repeat or multiple surgeries during the period from rectal cancer surgery to ileostomy closure.

Feature collection and screening

Prior to data collection, we conducted a comprehensive literature review to identify potential factors that may delay ileostomy closure. Patient data were collected from the standardized database of Sun Yat-sen University Cancer Center, and included age, sex, body mass index (BMI), distance between tumor and anal margin, pathological type, depth of tumor infiltration, lymph node metastasis, distant metastases, preoperative neoadjuvant chemoradiotherapy, postoperative chemotherapy, anastomotic leakage, anastomotic stenosis, low anterior resection syndrome (LARS), complications related to ileostomy. Variables were selected by LASSO regression (n-fold = 20) using the “glmnet” package in R.

Model development and validation

The data set was partitioned before building the model: 75% of the patients were assigned to the training set using the “caret” package, and the remaining 25% were assigned to the validation set. The model was built using the “XGBoost” package, and used the train() function to optimize the parameters in the caret package, and output the optimal parameter configuration. The XGBoost model is constructed by setting the learning rate eta to 0.1, the maximum depth (max_depth) to 2, and the number of iteration rounds (i.e., enhancement rounds) to 100. Accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score, ROC curve, calibration curve, and clinical decision curve analysis (DCA) were used to evaluate the performance of the model in the training and test sets. Predicted outcome probabilities were converted to binary outcomes using a threshold of 0.5.

This model was developed to assist in predicting outcomes for patients who had already undergone prophylactic ileostomy, using clinical and pathological data available shortly after surgery. Therefore, it is not intended for preoperative use (e.g., deciding whether to defunction), but rather for postoperative decision-making regarding the timing and safety of ileostomy closure.

Visualizing data

“Shapviz” is an R package for interpreting ML model predictions, which provides visual explanations based on SHAP (SHapley Additive explanations) values. The SHAP value explains how much each feature contributes to the model’s predictions, and whether it is positive or negative. The feature importance plot was used to display the features with the greatest impact on model predictions. Feature importance is ranked based on the average absolute value of the SHAP value. Two samples were selected to create a forced plot of SHAP values ​​for predicting and interpreting the one-sample model.

Statistics analysis

Categorical data were described as frequency and composition ratio. Continuous data that conform to a normal distribution were presented as mean ± standard deviation. Non-normally distributed data were presented as frequency (percentage). The chi-square test was used for comparing categorical data, and the t-test for continuous data. A 2-sided value of P < 0.05 was considered statistically significant. R language was used for organizing, analyzing, and visualizing data.

Results

Patient characteristics

Of the 442 patients included in the analysis, 305 experienced delayed ileostomy closure (69.0%). There were no significant differences between the 2 groups in sex distribution, distant metastasis, tumor pathological type, anastomotic leakage, LARS, and other postoperative complications (all, P > 0.05). There were significant differences between the 2 groups in age, BMI, distance between the tumor and the anal verge, lymph node metastasis, depth of tumor infiltration, neoadjuvant chemoradiotherapy, postoperative chemotherapy, and anastomotic stenosis (all, P < 0.05) (Table 1).

Table 1 Comparison of basic information between case group and control group
Full size table

Feature selection

A total of 14 variables were included in the analysis: LASSO regression was used for the selection of variables due to the fact that too many variables may lead to overfitting of the model, as well as covariance problems (Fig. 1). By choosing lambda values that were 1 standard deviation from the minimum lambda value (within 1 standard error of the minimum value), we identified the 7 most predictive variables (Fig. 2). After screening out the 7 variables with LASSO regression, we fitted a stepwise multifactor logistic regression based on the variables to screen them again. Ultimately, age was excluded because it was not an independent risk factor. Finally, 6 independent variables showed the strongest associations with the dependent variables, ensuring model simplicity and mitigating overfitting problems. The 6 variables were BMI, tumor distance from the anal verge, depth of tumor infiltration, neoadjuvant radiotherapy, postoperative chemotherapy, and anastomotic stenosis (Table 2). Using multivariate logistic regression, we determined that all of these variables were independent risk factors for delayed ileostomy closure.

Fig. 1
figure 1

LASSO coefficient path diagram

Full size image
Fig. 2
figure 2

LASSO regularization path diagram, where the model has fewer features. while maintaining predictive performance when the regularization parameter value is set to one standard deviation

Full size image
Table 2 Multivariate logistics regression of model variables
Full size table

Model performance

The performance evaluation metrics of the model are shown in Table 2. The XGBoost model had high area under the ROC curve (AUC) for predicting delayed ileostomy closure in the training set and the validation set: training set AUC = 0.744, 95% confidence interval (CI): 0.686–0.806; test set AUC = 0.809, 95% CI: 0.728–0.889 (Fig. 3A-B). The model had high sensitivity, but low specificity in the training and validation sets: PPV (training set: 0.753, 95% CI: 0.701–0.805; validation set: 0.782, 95% CI: 0.705–0.859), NPV (training set: 0.698, 95% CI: 0.56–0.835; validation set: 0.773, 95% CI: 0.598–0.948). The F1 score of the training set was 0.836, and of the validation set was 0.856) (Table 3).

Fig. 3
figure 3

Receiver operating characteristic curve of the model on the training set(A). Receiver operating characteristic curve of the model validation set(B)

Full size image
Table 3 Evaluation of XGBoost model performance in the training set and testing set
Full size table

We assessed the accuracy of the XGBoost model in predicting the probability of delayed ileostomy closure by analyzing the calibration curves and clinical decision curves of the training and validation sets. The calibration curves for both the training and validation sets showed a good fit, indicating that the model predictions are highly consistent with the actual incidence (Fig. 4A-B). The clinical decision curves of the training and validation sets showed that the model has a good degree of clinical utility (Fig. 5A-B).

Fig. 4
figure 4

Calibration curve of the model training set(A). Calibration curve of the model validation set (B)

Full size image
Fig. 5
figure 5

Clinical decision curve of the model on the training set(A). Clinical decision curve of the model validation set(B)

Full size image

Interpretation of the model

The SHAP summary plot of the model explains the effect of each variable on the model (Fig. 6). The order of importance of each variable is, in descending order, BMI > postoperative chemotherapy > distance of the tumor from the anal verge > depth of tumor infiltration > neoadjuvant radiotherapy > anastomotic stenosis. SHAP force diagrams are commonly used to explain how an XGBoost model assesses patient individualized variable contributions. We used the SHAP force plot to explain the level of variable contribution for 2 patients in this study. The colors indicate the contribution of each variable, with blue indicating that the feature is a negative predictor variable (arrow to the left, SHAP value decreases), and red indicating that the feature is a positive predictor variable (arrow to the right, SHAP value increases). The length of the color bar indicates the strength of the contribution, and E [f (x)] indicates the SHAP reference value, which is the average of the model predictions. For the “true positive” patient group, the XGBoost model predicted a delayed closure SHAP value of 0.837, exceeding the baseline value and indicating the occurrence of delayed closure (Fig. 7A). For the group of patients classified as “true negative”, the XGBoost model predicted a delayed closure value of 3.12, which did not exceed the reference value thus indicating that no delayed closure occurred (Fig. 7B).

Fig. 6
figure 6

Importance chart for SHAP variables

Full size image
Fig. 7
figure 7

SHAP force plot for “true positive” (A). SHAP force plot for “true negative” (B)

Full size image

Discussion

There is currently no consensus on when a temporary loop ileostomy should be closed. Generally, closure is performed at 8–12 weeks after the operation, depending on the situation [18, 31]. In China, patients with a temporary ileostomy usually undergo examinations such as fiber colonoscopy, pelvic magnetic resonance imaging (MRI), defecography, and anorectal pressure measurement to evaluate tumor activity and perianal function before ileostomy closure. The aforementioned examinations take about 1 week. For this study, we defined delayed ileostomy closure as closure ≥ 90 days after the index operation based on prior literature suggesting that 3 months is a widely accepted threshold for planned closure [32]. To our knowledge, this is the first study to develop a SHAP-interpretable XGBoost model to predict the occurrence of temporary ileostomy closure after rectal cancer surgery.

Training and validation sets were used to evaluate the performance of the XGBoost model, and the AUC of the training set for predicting delayed closure was 0.744 (95% CI: 0.686–0.806) and the AUC of the validation set was 0.809 (95% CI: 0.728–0.889). Thus, the model has the potential to assist in the management of patients with a temporary ileostomy by predicting the risk of delayed closure. Additionally, in order to achieve visualization of the data and results, we introduced the SHAP method to attempt to interpretate the model constructed in this study. The order of importance of each variable is, in descending order, BMI > postoperative chemotherapy > distance of the tumor from the anal verge > depth of tumor infiltration > neoadjuvant radiotherapy > anastomotic stenosis.

Previous studies have revealed that temporary ileostomy closure timing is influenced by various factors, and that there are risk factors associated with delayed closure. In this study, SHAP identified BMI as the most important variable, followed by postoperative chemotherapy and distance of the tumor from the anal verge as the second and third most important variables. Obesity has always been a risk factor for many diseases, and studies have shown that obesity (BMI ≥ 30.0 kg/m2) can cause various stoma related complications, thereby affecting the timely closure of the stoma [33–34]. Our results showed that a BMI ≥ 24.0 kg/m2 is a significant risk factor for delayed closure.

Postoperative radiotherapy and chemotherapy are considered a risk factors for delayed ileostomy closure [34,35,36]. Compared to postoperative adjuvant chemotherapy, little study has been done to examine the impact of neoadjuvant chemoradiotherapy on the delay of ileostomy closure. den Dulk et al. [35] concluded that neoadjuvant radiochemotherapy significantly delays ileostomy closure, probably because preoperative radiochemotherapy increases the risk of postoperative anastomotic complications in rectal cancer patients, which delays the closure of the ileostomy. In this study, the proportion of patients who underwent neoadjuvant therapy and postoperative adjuvant chemotherapy was significantly higher in the delayed closure group, and the LASSO regression analysis showed that postoperative adjuvant chemotherapy and neoadjuvant radiotherapy were risk factors for delayed ileostomy closure. Ileostomy closure is considered by most patients, families, and physicians to be an elective, non-essential procedure, and therefore is often considered unimportant. As such, during postoperative adjuvant chemotherapy, most patients and treating physicians will agree to delay the timing of ileostomy closure for the sake of continuity of chemotherapy. However, a recent study has shown that delaying stoma closure beyond 6 months is associated with nearly a threefold increase in the risk of reoperation [37]. In addition, postoperative chemotherapy can cause patients to become weak, making it difficult for them to undergo surgery in a short period of time and delaying the return of the stoma.

According to reports, 36% of rectal anastomotic leaks can lead to the formation of anastomotic leakage channels, delaying the reintroduction of prophylactic stomas and even causing them to evolve into permanent stomas [38–39]. However, in our study there was no difference in the occurrence of anastomotic leakage between the 2 groups, which may be due to an insufficient number of patients. However, a factor related to anastomotic leakage, the distance between the tumor and the anal margin, was shown to be an independent factor for delayed closure. Research has shown that large tumors near the anus are predictive for irreversible or delayed reversal of a shunt stoma [40]. This may be due to the tumor being closer to the anal margin, which may result in greater tension at the anastomotic site and poor local healing, leading to decisions to delay closure. It may also be due to the fact that with the improvement of surgical techniques and treatment, anastomotic leaks are easy to recognize early and thus controlled without related complications. It is noteworthy that anastomotic stenosis was found to be an independent risk factor in this study, a finding different than in some prior studies. Rectal anastomotic stenosis is the second most serious complication related to anastomosis after rectal cancer resection, and a recent meta-analysis reported that the incidence of anastomotic stenosis after rectal cancer surgery is 17% (95% CI: 13-21%) [41]. Anastomotic stenosis can also delay ileostomy closure and increase the risk of stoma-related complications [13].

Many prior studies have examined risk factors for delayed ileostomy closure after surgery for rectal cancer, but almost all have been limited by using multiple linear regression for model construction [34, 42–43]. Machine learning is a technique that uses computer algorithms to improve model accuracy and precision, enabling better handling of nonlinear relations between multiple variables [25]. Unlike previous studies, we first used LASSSO regression for variable screening. This feature selection method has good interpretability, and can automatically select important features related to the target variable to avoid overfitting of the model [44]. Then, XGBoost ML was used to construct the model, which can improve the accuracy and generalization ability of models, and handle high-dimensional data and complex classification problems [45]. The SHAP method was used to interpret the model, and visualize the results. Multi-center clinical trial data is needed to further evaluate and clarify the accuracy and precision of this predictive model. The model could be integrated into clinical workflows to stratify patients based on their predicted risk of delayed ileostomy closure. For those identified as high-risk, clinicians might consider early postoperative imaging, more frequent follow-up, or prioritizing them for timely stoma closure when appropriate. Applying the model in this way may help support more proactive patient management and optimize use of clinical resources.

Another important consideration is the risk of stoma permanence associated with defunctioning stomas, which may occur even in patients initially intended for reversal [46]. In addition, the presence of a stoma, particularly high-output ileostomies, has been associated to an increased risk of acute kidney injury. These reveal the need for careful monitoring and early closure where appropriate [47].

Limitations

This study has various limitations to consider. First, we did not use systematic retrieval to include as many risk variables as possible. The study focused on the limited parameters we are familiar with, without considering more potential influencing factors and subgroups, such as diabetes and hypertension. Second, our XGBoost model was trained and validated using data from a single center without external or temporal validation, so its generalizability is unknown. While the current results demonstrate good internal performance, it remains unclear how well the model would perform across different institutions, patient populations, or care settings. We acknowledge this limitation and plan to conduct prospective external validation in future studies, ideally across multiple centers with more diverse patient cohorts. Third, the sample size used is relatively small. Although model performance was evaluated through training and validation sets, sample size may limit the generalization ability of the results. Finally, although the use of SHAP improves the interpretability of the model, comprehension of ML models remains a challenge for persons not experienced with ML.

Conclusion

The XGBoost ML model we built helps predict delayed closure of a temporary ileostomy after rectal cancer surgery. In addition, the SHAP method is able to explain the results generated by the ML model, improving the interpretability of the model and helping to provide explanations for the source of the predicted results for each patient. This model may help to avoid delayed ileostomy closure after rectal cancer surgery, and help practitioners better manage a temporary ileostomy.

Data availability

No datasets were generated or analysed during the current study.

References

  1. Bray F, Ferlay J, Soerjomataram I, Siegel RL, Torre LA, Jemal A. Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA CANCER J CLIN. 2018.

  2. Heald RJ, Husband EM, Ryall RDH. The mesorectum in rectal cancer surgery—the clue to pelvic recurrence? Br J Surg. 1982;69:613–6.

    Article  CAS  PubMed  Google Scholar 

  3. You YN, Hardiman KM, Bafford A, Poylin V, Francone TD, Davis K et al. The American Society of Colon and Rectal Surgeons Clinical Practice Guidelines for the Management of Rectal Cancer. Dis Colon Rectum [Internet]. 2020;63. Available from: https://journals.lww.com/dcrjournal/fulltext/2020/09000/the_american_society_of_colon_and_rectal_surgeons.6.aspx

  4. Kverneng Hultberg D, Svensson J, Jutesten H, Rutegård J, Matthiessen P, Lydrup M-L et al. The Impact of Anastomotic Leakage on Long-term Function After Anterior Resection for Rectal Cancer. Dis Colon Rectum [Internet]. 2020;63. Available from: https://journals.lww.com/dcrjournal/fulltext/2020/05000/the_impact_of_anastomotic_leakage_on_long_term.10.aspx

  5. Keller DS, Talboom K, Van Helsdingen CPM, Hompes R. Treatment modalities for anastomotic leakage in rectal Cancer surgery. Clin Colon Rectal Surg. 2021;34:431–8.

    Article  PubMed  PubMed Central  Google Scholar 

  6. CHAND M, NASH GF, TALBOT RW. Timely closure of loop ileostomy following anterior resection for rectal cancer. Eur J Cancer Care (Engl). 2008;17:611–5.

    CAS  PubMed  Google Scholar 

  7. Blok RD, Stam R, Westerduin E, Borstlap WAA, Hompes R, Bemelman WA, et al. Impact of an institutional change from routine to highly selective diversion of a low anastomosis after TME for rectal cancer. Eur J Surg Oncol. 2018;44:1220–5.

    Article  CAS  PubMed  Google Scholar 

  8. Pisarska M, Gajewska N, Małczak P, Wysocki M, Witowski J, Torbicz G, et al. Defunctioning ileostomy reduces leakage rate in rectal cancer surgery - systematic review and meta-analysis. Oncotarget. 2018;9:20816–25.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Garg PK, Goel A, Sharma S, Chishi N, Gaur MK. Protective diversion stoma in low anterior resection for rectal cancer: A Meta-Analysis of randomized controlled trials. Visc Med. 2019;35:156–60.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Luglio G, Pendlimari R, Holubar SD, Cima RR, Nelson H. Loop ileostomy reversal after Colon and rectal surgery: A single institutional 5-Year experience in 944 patients. Arch Surg. 2011;146:1191–6.

    Article  PubMed  Google Scholar 

  11. Chapman WC, Subramanian M, Jayarajan S, Makhdoom B, Mutch MG, Hunt S, et al. First, do no harm: rethinking routine diversion in Sphincter-Preserving rectal Cancer resection. J Am Coll Surg. 2019;228:547–e5568.

    Article  PubMed  PubMed Central  Google Scholar 

  12. Gu W, Wu S. Meta-analysis of defunctioning stoma in low anterior resection with total mesorectal excision for rectal cancer: evidence based on thirteen studies. World J Surg Oncol. 2015;13:9.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Danielsen AK, Park J, Bock D, Haglind E. Early closure of a temporary ileostomy in patients with rectal Cancer. Ann Surg. 2016;265.

  14. Ye H, Huang S, Yu J, Zhou Q, Xi C, Cao L, et al. Comparison of the clinical outcomes of skin Bridge loop ileostomy and traditional loop ileostomy in patients with low rectal cancer. Sci Rep. 2021;11:9101.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  15. Jabbal IS, Spaulding AC, Lemini R, Borkar SR, Stanek K, Colibaseanu DT. Temporary vs. permanent Stoma: factors associated with the development of complications and costs for rectal cancer patients. Int J Colorectal Dis. 2022;37:823–33.

    Article  PubMed  Google Scholar 

  16. Mehboob A, Perveen S, Iqbal M, Moula Bux K, Waheed A. Frequency and Complications of Ileostomy. Cureus [Internet]. 2020 [cited 2024 Sep 27]; Available from: https://www.cureus.com/articles/41166-frequency-and-complications-of-ileostomy

  17. Hughes DL, Cornish J, Morris C, On behalf of the LARRIS Trial Management Group. Functional outcome following rectal surgery—predisposing factors for low anterior resection syndrome. Int J Colorectal Dis. 2017;32:691–7.

    Article  PubMed  Google Scholar 

  18. Podda M, Coccolini F, Gerardi C, Castellini G, Wilson MSJ, Sartelli M, et al. Early versus delayed defunctioning ileostomy closure after low anterior resection for rectal cancer: a meta-analysis and trial sequential analysis of safety and functional outcomes. Int J Colorectal Dis. 2022;37:737–56.

    Article  PubMed  PubMed Central  Google Scholar 

  19. Park J, Danielsen AK, Angenete E, Bock D, Marinez AC, Haglind E, et al. Quality of life in a randomized trial of early closure of temporary ileostomy after rectal resection for cancer (EASY trial). Br J Surg. 2018;105:244–51.

    Article  CAS  PubMed  Google Scholar 

  20. Shimizu H, Yamaguchi S, Ishii T, Kondo H, Hara K, Takemoto K, et al. Who needs diverting ileostomy following laparoscopic low anterior resection in rectal cancer patients? Analysis of 417 patients in a single Institute. Surg Endosc. 2020;34:839–46.

    Article  PubMed  Google Scholar 

  21. van den Hout WB, van den Brink M, Stiggelbout AM, van de Velde CJH, Kievit J. Cost-effectiveness analysis of colorectal cancer treatments. Eur J Cancer. 2002;38:953–63.

    Article  PubMed  Google Scholar 

  22. Schwalbe N, Wahl B. Artificial intelligence and the future of global health. Lancet. 2020;395:1579–86.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  23. Kolker E, Özdemir V, Kolker E. How healthcare can refocus on its Super-Customers (Patients, n = 1) and customers (Doctors and Nurses) by leveraging lessons from Amazon, Uber, and Watson. OMICS J Integr Biol. 2016;20:329–33.

    Article  CAS  Google Scholar 

  24. Lei L, Zhang S, Yang L, Yang C, Liu Z, Xu H, et al. Machine learning-based prediction of delirium 24 h after pediatric intensive care unit admission in critically ill children: A prospective cohort study. Int J Nurs Stud. 2023;146:104565.

    Article  PubMed  Google Scholar 

  25. Ali S, Akhlaq F, Imran AS, Kastrati Z, Daudpota SM, Moosa M. The enlightening role of explainable artificial intelligence in medical & healthcare domains: A systematic literature review. Comput Biol Med. 2023.

  26. Kuo C-Y, Kuo L-J, Lin Y. Artificial intelligence based system for predicting permanent stoma after sphincter saving operations. Sci Rep. 2023;13:16039.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  27. Back E, Häggström J, Holmgren K, Haapamäki MM, Matthiessen P, Rutegård J, et al. Permanent stoma rates after anterior resection for rectal cancer: risk prediction scoring using preoperative variables. Br J Surg. 2021;108:1388–95.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  28. Wei W, Li Y, Huang T. Using machine learning methods to study colorectal Cancer tumor Micro-Environment and its biomarkers. Int J Mol Sci. 2023;24:11133.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  29. Li Y, Wu X, Yang P, Jiang G, Luo Y. Machine learning for lung Cancer diagnosis, treatment, and prognosis. Genomics Proteom Bioinf. 2022;20:850–66.

    Article  Google Scholar 

  30. Zuo D, Yang L, Jin Y, Qi H, Liu Y, Ren L. Machine learning-based models for the prediction of breast cancer recurrence risk. BMC Med Inf Decis Mak. 2023;23:276.

    Article  Google Scholar 

  31. Hallböök O, Matthiessen P, Leinsköld T, Nyström P-O, Sjödahl R. Safety of the temporary loop ileostomy. Colorectal Dis. 2002;4:361–4.

    Article  PubMed  Google Scholar 

  32. Chow A, Tilney HS, Paraskeva P, Jeyarajah S, Zacharakis E, Purkayastha S. The morbidity surrounding reversal of defunctioning ileostomies: a systematic review of 48 studies including 6,107 cases. Int J Colorectal Dis. 2009;24:711–23.

    Article  PubMed  Google Scholar 

  33. Chun LJ, Haigh PI, Tam MS, Abbas MA. Defunctioning Loop Ileostomy for Pelvic Anastomoses: Predictors of Morbidity and Nonclosure. Dis Colon Rectum [Internet]. 2012;55. Available from: https://journals.lww.com/dcrjournal/fulltext/2012/02000/defunctioning_loop_ileostomy_for_pelvic.10.aspx

  34. Sier MF, Van Gelder L, Ubbink DT, Bemelman WA, Oostenbroek RJ. Factors affecting timing of closure and non-reversal of temporary ileostomies. Int J Colorectal Dis. 2015;30:1185–92.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  35. den Dulk M, Smit M, Peeters KC, Kranenbarg EM-K, Rutten HJ, Wiggers T, et al. A multivariate analysis of limiting factors for stoma reversal in patients with rectal cancer entered into the total mesorectal excision (TME) trial: a retrospective study. Lancet Oncol. 2007;8:297–303.

    Article  Google Scholar 

  36. Lordan JT, Heywood R, Shirol S, Edwards DP. Following anterior resection for rectal cancer, defunctioning ileostomy closure May be significantly delayed by adjuvant chemotherapy: a retrospective study. Colorectal Dis. 2007;9:420–2.

    Article  CAS  PubMed  Google Scholar 

  37. Munshi E, Segelman J, Matthiessen P, Park J, Rutegård M, Sjöström O, et al. Increased risk of postoperative complications after delayed stoma reversal: a multicenter retrospective cohort study on patients undergoing anterior resection for rectal cancer. Int J Colorectal Dis. 2025;40:36.

    Article  PubMed  PubMed Central  Google Scholar 

  38. van Koperen PJ, van der Zaag ES, Omloo JMT, Slors JFM, Bemelman WA. The persisting presacral sinus after anastomotic leakage following anterior resection or restorative proctocolectomy. Colorectal Dis. 2011;13:26–9.

    Article  PubMed  Google Scholar 

  39. Waterland P, Goonetilleke K, Naumann DN, Sutcliff M, Soliman F. Defunctioning ileostomy reversal rates and reasons for delayed reversal: does delay impact on complications of ileostomy reversal? A study of 170 defunctioning ileostomies. J Clin Med Res. 2015;7:685–9.

    Article  PubMed  PubMed Central  Google Scholar 

  40. Fukui R, Nozawa H, Hirata Y, Kawai K, Hata K, Tanaka T, et al. Low preoperative maximum squeezing pressure evaluated by anorectal manometry is a risk factor for non-reversal of diverting stoma. Langenbecks Arch Surg. 2021;406:131–9.

    Article  PubMed  Google Scholar 

  41. He F, Yang F, Chen D, Tang C, Woraikat S, Xiong J, et al. Risk factors for anastomotic stenosis after radical resection of rectal cancer: A systematic review and meta-analysis. Asian J Surg. 2024;47:25–34.

    Article  PubMed  Google Scholar 

  42. Chau TC-Y, Nguyen H, Robertson IK, Harvey X, Tan B, Tan M, et al. Factors affecting timing of loop ileostomy closure: a regional centre’s experience with 106 patients. ANZ J Surg. 2024;94:193–8.

    Article  PubMed  Google Scholar 

  43. Kim MJ, Kim YS, Park SC, Sohn DK, Kim DY, Chang HJ, et al. Risk factors for permanent stoma after rectal cancer surgery with temporary ileostomy. Surgery. 2016;159:721–7.

    Article  PubMed  Google Scholar 

  44. Xie Y, Shi H, Han B. Bioinformatic analysis of underlying mechanisms of Kawasaki disease via weighted gene correlation network analysis (WGCNA) and the least absolute shrinkage and selection operator method (LASSO) regression model. BMC Pediatr. 2023;23:90.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  45. Zhang Z, Jung C. GBDT-MO: Gradient-Boosted decision trees for multiple outputs. IEEE Trans Neural Netw Learn Syst. 2021;32:3156–67.

    Article  PubMed  Google Scholar 

  46. Holmgren K, Häggström J, Haapamäki MM, Matthiessen P, Rutegård J, Rutegård M. Defunctioning stomas May reduce chances of a stoma-free outcome after anterior resection for rectal cancer. Colorectal Dis. 2021;23:2859–69.

    Article  PubMed  Google Scholar 

  47. Rutegård M, Häggström J, Back E, Holmgren K, Wixner J, Rutegård J, et al. Defunctioning loop ileostomy in anterior resection for rectal cancer and subsequent renal failure: nationwide population-based study. BJS Open. 2023;7:zrad010.

    Article  PubMed  PubMed Central  Google Scholar 

Download references

Acknowledgements

We would like to thank the patients, doctors, and nurses who participated in this study.

Author information

Author notes

  1. Jianying Liu and Mengxiao Jiang contributed equally to this work.

Authors and Affiliations

  1. Department of Colorectal Surgery, State Key Laboratory of Oncology in South China, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guangzhou, 510060, P.R. China

    Jianying Liu, Yonglan Ge, Xia Yang, Wenhao Zhou, Meichun Zheng & Baojia Luo

  2. Department of Urinary Surgery, State Key Laboratory of Oncology in South China, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guangzhou, 510060, P.R. China

    Mengxiao Jiang, Xiaoping Chen & Zongxin Zheng

  3. Nursing Department of Huangpu Yard, State Key Laboratory of Oncology in South China, Guangdong Provincial Clinical Research Center for Cancer, Sun Yat-sen University Cancer Center, Guangzhou, 510060, P.R. China

    Huiting Zhang

Authors
  1. Jianying Liu
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Mengxiao Jiang
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Xiaoping Chen
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Yonglan Ge
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Zongxin Zheng
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Xia Yang
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Wenhao Zhou
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Huiting Zhang
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Meichun Zheng
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Baojia Luo
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Jianying Liu: Writing– review & editing, Writing– original draft, Methodology, Data curation, Conceptualization. Mengxiao Jiang: Investigation, Methodology, Software, Data curation. Yonglan Ge: Investigation, Methodology, Data curation. Zongxin Zheng, Xia Yang & Xiaoping Chen: Data collection, Data input, and verification. Wenhao Zhou: Research and Design Consulting. Huiting Zhang & Meichun Zheng: Visualization, Validation, Conceptualization. Baojia Luo: Visualization, Validation, Methodology, Formal analysis.

Corresponding authors

Correspondence to Huiting Zhang, Meichun Zheng or Baojia Luo.

Ethics declarations

Ethics approval and consent to participate

This study was approved by the Ethics Review Committee of Sun Yat-sen University Cancer Center (SL-B2024-578-01).

Competing interests

The authors declare no competing interests.

Conflict of interest

The authors declare that they have no known competing financial interests or personal relationships that might influence the work reported in this paper.

Sources of funding

This study has not attained any funding.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Supplementary Material 1

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Liu, J., Jiang, M., Chen, X. et al. Machine learning prediction model of prolonged delay to loop ileostomy closure after rectal cancer surgery: a retrospective study. World J Surg Onc 23, 185 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12957-025-03843-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s12957-025-03843-w

Keywords

World Journal of Surgical Oncology

ISSN: 1477-7819

Contact us