Abstract
Purpose
PPOI is one of the common complications of intraperitoneal hyperthermic chemotherapy during laparoscopic radical resection of rectal cancer, which seriously affects the prognosis of patients. The purpose of this study is to establish a prediction system for PPOI secondary to laparoscopic radical resection of rectal cancer combined with intraperitoneal hyperthermic chemotherapy with lobaplatin.
Materials and methods
Retrospectively analyzed the clinical data of 800 patients who received laparoscopic radical rectal cancer combined with lobaplatin hyperthermia and intraperitoneal chemotherapy in three Level 3 Grade A hospitals from June 1, 2014, to June 1, 2024, and determined the predictive factors through univariate, multivariate, and Lasso regression analysis. We employ eight ML algorithms, Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Extreme gradient boosting (XGB), Support Vector Machine (SVM), Multilayer Perceptron (MLP), K-Nearest Neighbor (KNN), Gaussian Naive Bayes (GNB), to train and develop ML models using a 10x cross-validation method. The performance of the model was evaluated by a variety of indicators, including the area under the receiver operating characteristic curve (ROC), calibration curve, decision curve, PR curve, and confusion matrix. In addition, model interpretation is performed through Shapley Additive Interpretation (SHAP) analysis to clarify the importance of each feature of the model and its basis for decision-making.
Results
We identified six key predictors, including Surgical bleeding, Duration of surgery, HB, WBC, ALB and adhesiolysis, and built a prediction model based on these factors. The sensitivity, specificity, positive predictive value, and negative predictive value of different models were compared. All eight models showed good predictive performance and stability, with the RF model being the optimal model. Finally, we developed a web-based calculator based on the optimal model.
Conclusions
These predictors and models were able to assess the potential for PPOI following laparoscopic radical curative rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy. Early alerts can be provided in a clinical setting, helping medical professionals make informed judgments and select the most appropriate treatment strategy (https://zw17786325639.shinyapps.io/ppoi/).
Keywords: Rectal cancer, Hyperthermic intraperitoneal chemotherapy, Loboplatin, Machine learning, Prolonged postoperative ileus, Predictive modeling
Introduction
Colorectal cancer (CRC) is one of the most common malignant tumors in the world, ranking third in incidence and fourth in mortality [1–3]. Among CRC patients, about 20% develop synchronous peritoneal metastasis (PM), and the median survival time without treatment is 5–7 months [4–6]. Rectal cancer accounts for approximately one-third of all colorectal cancers. The distal rectum lacks a serosal layer, which enables tumor cells to invade through the muscular layer and extend into the pelvic cavity at an earlier stage. These anatomical characteristics are closely associated with the unique pattern of local progression and metastasis in rectal cancer [7]. In addition to typical treatments including surgery, intravenous systemic chemotherapy, and supportive care, the combination strategy of hyperthermic intraperitoneal chemotherapy (HIPEC) has emerged as a new treatment option for the treatment of PM such as CRC, ovarian cancer, and gastric cancer [8]. However, this treatment also produces some side effects, greatly hindering the application of HIPEC in the treatment of gastrointestinal cancer. The main side effect on the digestive system is postoperative ileus (POI) [9, 10]. POI can develop into prolonged POI, known as PPOI, which is also one of the common complications after colorectal cancer surgery. However, there is controversy regarding the diagnostic criteria for PPOI. Given that the present study focused on patients undergoing laparoscopic radical resection of rectal cancer combined with lobaplatin HIPEC, we modified the PPOI definition to correlate intestinal dysfunction with HIPEC treatment. Specifically, PPOI was diagnosed when at least two of the following criteria were met within 7 days after the completion of HIPEC: nausea or vomiting for 12 h or longer without relief, intolerance to solid or semi-solid oral diet, persistent abdominal distension, failure to pass stool or flatus for 24 h or longer, and intestinal obstruction confirmed by abdominal plain film or computed tomography (CT) scan [11, 12].Loboplatin is a third-generation platinum-based antitumor drug with superior therapeutic indicators and low toxicity, and has been widely used in the treatment of colorectal cancer and the corresponding PM [13–15]. The purpose of this study is to analyze the related factors of secondary PPOI after laparoscopic radical rectal cancer surgery combined with lobaplatin hyperthermic intraperitoneal chemotherapy, and to establish a predictive model based on machine learning algorithms. The aim is to help clinicians predict PPOI early, provide timely intervention for patients with secondary PPOI after lobaplatin hyperthermic intraperitoneal chemotherapy, and ultimately improve the prognosis of these patients.
Methods
Ethics approval and consent to participate
This multicenter retrospective study was approved by the Ethics Committee of Jinzhou Central Hospital of Jinzhou Medical University (Ethics Approval No. JZZXYYLL20241204-05). The study was conducted in accordance with the Declaration of Helsinki. Informed consent was waived due to the retrospective design.
Experimental model and study participant details
We retrospectively analyzed the clinical data of 800 patients with clinically or intraoperatively confirmed peritoneal metastasis (PM) who underwent laparoscopic radical resection of rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy (HIPEC) at Jinzhou Medical University Affiliated First Hospital Jinzhou Central Hospital and Xiangyang First People’s Hospital from June 1, 2014, to June 1, 2024. The inclusion criteria were as follows: (1) Rectal cancer confirmed pathologically by preoperative colonoscopic biopsy; (2) Abdominal CT and colonoscopy were used to determine tumor location and evaluate resectable rectal tumors; (3) No antitumor therapy prior to the surgery; (4) The first laparoscopic surgery for rectal cancer; (5) Primary rectal tumor reaches R0 resection (margin free of tumor cells); (6) Postoperative treatment of lobolatin combined with hyperthermic intraperitoneal chemotherapy; (7) All patients had confirmed peritoneal metastasis and underwent HIPEC. Exclusion criteria: (1) Difficult to tolerate surgery; (2) Suffering from dual or multiple cancers, or undergoing emergency surgery due to acute intestinal perforation, acute intestinal obstruction, etc.; (3) Distant organ metastasis to the liver, lungs, bones, etc., confirmed by preoperative imaging or postoperative pathology, as well as local tumor progression to T4b stage that is not amenable to radical resection; (5) Concurrently undergoing other chemotherapeutic regimens; (6) Having contraindications to chemotherapy; and (7) With incomplete information of relevant medical history.
Method details
All operations were performed by qualified senior surgical teams at each center, following a standardized surgical protocol and uniform HIPEC procedure across the three hospitals. After laparoscopic radical resection of rectal cancer, the input catheter was placed under the diaphragm and the output catheter was placed to the pelvis, approximately 3 L of heated saline containing lobaplatin (40 mg/m2) was injected into the peritoneal cavity at a rate of 500 ml/min and circulated for 60 min using a custom-developed high-precision body cavity thermal perfusion therapy system. During intraperitoneal chemotherapy, the temperature of the perfusate is maintained at 43 ± 0.3 °C. The perfusate is drained after HIPEC is completed. Diagnosis of PPOI was independently assessed by two senior surgeons who were blinded to patient grouping. Any disagreement was resolved by a third senior gastrointestinal surgeon to ensure diagnostic consistency.
Statistical analysis and machine learning implementation
All statistical analyses and machine learning model construction were conducted via R software (version 4.4.1, R Foundation for Statistical Computing, Vienna, Austria). Before analysis, we assessed missing data for all variables. The overall missing rate across variables was ≤ 5%, with no variables excluded from the analysis due to excessive missingness. The Kolmogorov–Smirnov test assessed the normality of continuous variables: normally distributed variables were expressed as mean ± standard deviation (SD) and compared with the Student’s t-test, while non-normally distributed variables were presented as median (interquartile range, IQR) and analyzed using the Mann–Whitney U test. Categorical variables were described as n (%) and compared with the chi-squared (χ²) test or Fisher’s exact test as appropriate. A two-sided P-value < 0.05 was defined as statistically significant.
Clinical features underwent rigorous preprocessing before model development: features with high missing data were excluded to avoid bias and preserve interpretability, and all numerical features were standardized to eliminate scale heterogeneity. A stepwise feature selection strategy was adopted to address high dimensionality and multicollinearity, and mitigate overfitting: univariate logistic regression first screened PPOI-associated clinical features, followed by multivariate logistic regression to identify independent associated factors, and finally least absolute shrinkage and selection operator (LASSO) regression to refine the optimal predictive variables for downstream model training.
Eight supervised machine learning (ML) classifiers were selected for model development: Logistic Regression (LR), Decision Tree (DT), Random Forest (RF), Extreme Gradient Boosting (XGB), Support Vector Machine (SVM), Multilayer Perceptron (MLP), K-Nearest Neighbor (KNN), and Gaussian Naive Bayes (GNB). Hyperparameters were tuned on the training set using 10-fold cross-validation. Model performance was comprehensively evaluated using AUROC, precision-recall (PR) curves, calibration curves, and confusion matrices. Decision curve analysis (DCA) assessed clinical net benefit and applicability. Shapley Additive Explanation (SHAP) analysis ranked feature importance and generated personalized model interpretability. A web-based calculator was developed based on the optimal ML model to facilitate clinical prediction of PPOI.
Postoperative care
All patients undergoing laparoscopic radical resection of rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy were managed with a standardized ERAS protocol. Oral clear liquids were initiated at 24–48 h postoperatively if bowel sounds recovered and no nausea/vomiting was present, with gradual advancement to a low-fat, low-residue diet by postoperative day 5–7 based on tolerance. Routine intraoperative nasogastric decompression was omitted; temporary intubation was only performed for refractory abdominal distension or vomiting and removed promptly. Early mobilization was enforced: bedside sitting on postoperative day 1 and assisted ambulation on day 2. For the first 7 postoperative days, patients were closely monitored for anastomotic leakage, including fever, abdominal tenderness, and increased surgical drainage.
Results
Patient characteristics
After strict screening, 800 rectal cancer patients with confirmed peritoneal metastasis were enrolled and stratified into the non-PPOI (NPPOI, n = 508, 63.5%) and PPOI (n = 292, 36.5%) groups per PPOI diagnostic criteria. Significant intergroup differences (P < 0.05) were noted in tumor stage, intraoperative blood loss, operative duration, hypertension, tumor differentiation grade, WBC, Hb, ALB, CA199, CA724, ALT, BUN, and adhesiolysis, with exact P-values provided in Table 1. The PPOI group had more advanced staging, higher hypertension/poor differentiation/adhesiolysis rates, greater intraoperative blood loss, longer operative duration, and elevated WBC.Data from Jinzhou Central Hospital was used as the training set, and data from Xiangyang No.1 People’s Hospital and The First Affiliated Hospital of Jinzhou Medical University were used as the external validation set. There were no significant differences in baseline characteristics between the training and external validation sets (all P > 0.05), confirming the comparability of the two cohorts. TNM staging in Table 1 reflects preoperative clinical staging (cTNM) by imaging, which tends to underestimate peritoneal micrometastases. Peritoneal metastasis and PCI were assessed intraoperatively.
Table 1.
Patient characteristics between PPOI and non-PPOI groups
| Characteristic | NPPOI | PPOI | P | Training | Validation | P |
|---|---|---|---|---|---|---|
| N = 508 | N = 292 | N = 560 | N = 240 | |||
| Age(years) | 68(11) | 70(10) | 0.213 | 69(10) | 69(11) | 0.258 |
| T | 0.181 | 0.342 | ||||
| 1 | 16(3.15%) | 11(3.77%) | 18(3.21%) | 9(3.75%) | ||
| 2 | 113(22.24%) | 48(16.44%) | 111(19.82%) | 50(20.83%) | ||
| 3 | 223(43.90%) | 128(43.84%) | 257(45.89%) | 94(39.17%) | ||
| 4 | 156(30.71%) | 105(35.96%) | 174(31.07%) | 87(36.25%) | ||
| N | 0.192 | 0.089 | ||||
| 0 | 292(57.48%) | 157(53.77%) | 328(58.57%) | 121(50.42%) | ||
| 1 | 176(34.65%) | 101(34.59%) | 185(33.04%) | 92(38.33%) | ||
| 2 | 40(7.87%) | 34(11.64%) | 47(8.39%) | 27(11.25%) | ||
| M | 17(3.35%) | 16(5.48%) | 0.202 | 19(3.39%) | 14(5.83%) | 0.163 |
| Staging | 0.046 | 0.422 | ||||
| 1 | 86(16.93%) | 46(15.75%) | 93(16.61%) | 39(16.25%) | ||
| 2 | 205(40.35%) | 97(33.22%) | 212(37.86%) | 90(37.50%) | ||
| 3 | 209(41.14%) | 138(47.26%) | 245(43.75%) | 102(42.50%) | ||
| 4 | 8(1.57%) | 11(3.77%) | 10(1.79%) | 9(3.75%) | ||
| Surgical bleeding(ml) | 70(80) | 185(260) | < 0.001* | 80(110) | 100(162.5) | 0.108 |
| Duration of surgery(min) | 140(50) | 180(75) | < 0.001* | 150(60) | 155(61.25) | 0.37 |
| Duration of illness(day) | 60(170) | 60(165) | 0.147 | 60(166) | 60(165) | 0.815 |
| Gender | 0.631 | 0.427 | ||||
| Male | 318(62.60%) | 177(60.62%) | 341(60.89%) | 154(64.17%) | ||
| Female | 190(37.40%) | 115(39.38%) | 219(39.11%) | 86(35.83%) | ||
| History of smoking | 258(50.79%) | 154(52.74%) | 0.647 | 292(52.14%) | 120(50.00%) | 0.632 |
| Hypertension | 140(27.60%) | 59(20.20%) | 0.026* | 141(25.18%) | 58(24.17%) | 0.83 |
| History of alcohol consumption | 144(28.30%) | 93(31.80%) | 0.334 | 170(30.36%) | 67(27.92%) | 0.543 |
| Family history | 63(12.40%) | 32(10.96%) | 0.622 | 64(11.43%) | 31(12.92%) | 0.633 |
| Diabetes | 97(19.10%) | 56(19.18%) | 0.970 | 109(19.46%) | 44(18.33%) | 0.784 |
| Diameter(cm) | 5(2.5) | 4.95(2.5) | 0.483 | 4.8(2.5) | 5(2.8) | 0.28 |
| History of abdominal surgery | 80(15.75%) | 38(13.01%) | 0.344 | 78(13.93%) | 40(16.67%) | 0.372 |
| Degree of differentiation | 0.002 | 0.242 | ||||
| 1 | 96(18.90%) | 78(26.71%) | 114(20.36%) | 60(25.00%) | ||
| 2 | 339(66.73%) | 158(54.11%) | 358(63.93%) | 139(57.92%) | ||
| 3 | 73(14.37%) | 56(19.18%) | 88(15.71%) | 41(17.08%) | ||
| Nerve invasion | 157(30.91%) | 94(32.19%) | 0.765 | 183(32.70%) | 68(28.30%) | 0.258 |
| Vasculoma thrombus | 184(36.20%) | 86(29.50%) | 0.061 | 193(34.46%) | 77(32.08%) | 0.568 |
| K(mmol/L) | 4.08(1.01) | 4.03(1.1) | 0.882 | 4.06(1.05) | 4.08(1.05) | 0.845 |
| Ascites | 32(6.30%) | 11(3.77%) | 0.172 | 32(5.71%) | 11(4.58%) | 0.632 |
| WBC(x109/L) | 7.46(4.14) | 9.51(4.42) | < 0.001* | 8.15(4.59) | 8.52(4.09) | 0.548 |
| PLR | 169.57(117.68) | 167.56(115.79) | 0.228 | 169.65(117.08) | 167.74(117.64) | 0.734 |
| NLR | 3.15(2.12) | 2.96(2.06) | 0.168 | 3.1(2.14) | 3.05(2.03) | 0.572 |
| Hb(g/L) | 125(29) | 118(23) | < 0.001* | 123(29) | 120.5(26) | 0.521 |
| FIB(g/L) | 3.76(1.21) | 3.7(1.48) | 0.093 | 3.76(1.26) | 3.7(1.42) | 0.795 |
| ALB(g/L) | 41(6) | 37.5(7) | < 0.001* | 40(7) | 40(7) | 0.878 |
| CEA(ng/mL) | 4.95(6.59) | 4.95(8.53) | 0.775 | 4.85(7.08) | 5.4(8.275) | 0.313 |
| CA199(U/mL) | 15.45(22.13) | 13.4(21.25) | 0.0234 | 15.05(20.53) | 15.1(24.83) | 0.748 |
| CA724(U/mL) | 5.08(5.74) | 4.2(3.63) | 0.0024 | 4.76(4.95) | 4.48(5.61) | 0.302 |
| CA125(U/mL) | 4.9(6.025) | 5.06(8.15) | 0.657 | 4.97(6.2) | 4.6(7.2) | 0.973 |
| PLT(x109/L) | 222(90.25) | 230(86.5) | 0.169 | 227(93.5) | 223(75) | 0.832 |
| AST(U/L) | 19(6) | 18(7.25) | 0.93 | 19(6) | 18(7) | 0.075 |
| ALT(U/L) | 14(9) | 13(7.25) | < 0.001* | 14(8) | 14(9.25) | 0.735 |
| TBIL(umol/L) | 11.6(8.13) | 11.25(6.4) | 0.866 | 11.55(7.6) | 11(7.825) | 0.431 |
| SCr(umol/L) | 63(22) | 66(21) | 0.645 | 63(22) | 63.5(22) | 0.479 |
| BUN(mmol/L) | 5.07(2.12) | 5.39(2.21) | 0.026 | 5.1(2.22) | 5.18(2.08) | 0.582 |
| Adhesiolysis | 62(12.20%) | 74(25.30%) | < 0.001* | 90(16.07%) | 46(19.17%) | 0.334 |
| Distance from the anus(cm) | 6.15(3.7) | 6.2(3.2) | 0.921 | 6.15(3.6) | 6.2(3.6) | 0.37 |
* represents P < 0.05
Feature selection
Firstly, univariate and multivariate logistic regression analysis was performed only in the training set data to screen the potential predictors associated with PPOI. Variables with statistical significance (P < 0.05) in the multivariate analysis were then included in the LASSO regression model for further feature selection (Fig. 1). After LASSO regularization, six variables (Surgical bleeding, Duration of surgery, WBC, Hb, ALB, and Adhesiolysis) were retained as the optimal predictors (Table 2). For continuous variables, ORs are reported per unit increase: 1-minute for duration of surgery, 1 mL for surgical bleeding, 1 × 10⁹/L for WBC, and 1 g/L for both Hb and ALB. These six factors were subsequently used to construct the machine learning prediction model for PPOI. All feature selection procedures were performed exclusively within the training set, while the validation set was kept completely independent and not involved in any step of feature selection, thereby eliminating the risk of data leakage and selection bias.
Fig. 1.
Cross-validation curve (A) and coefficient profile plot (B) for LASSO regression variable selection
Table 2.
Final predictors of PPOI
| variable | β | SE | OR | 95%CI | P |
|---|---|---|---|---|---|
| ALB | −0.09 | 0.027 | 0.91 | 0.86–0.96 | 0.001* |
| WBC | 0.25 | 0.043 | 1.28 | 1.18–1.40 | < 0.001* |
| Hb | −0.03 | 0.006 | 0.97 | 0.96–0.98 | < 0.001* |
| Duration of surgery | 0.02 | 0.003 | 1.02 | 1.02–1.03 | < 0.001* |
| Surgical bleeding | 0.01 | 0.002 | 1.01 | 1.00−1.01 | < 0.001* |
| Adhesiolysis | 0.79 | 0.310 | 2.20 | 1.20–4.03 | 0.011* |
* represents P < 0.05
Performance of developed models
The random forest (RF) model outperformed seven comparative algorithms (logistic regression, decision tree, k-nearest neighbors, XGBoost, support vector machine, multi-layer perceptron, naive Bayes) across all evaluation metrics. In the training cohort, it achieved an ROC-AUC of 0.917 (95% CI: 0.889–0.946) (Fig. 2A), and in the validation cohort, it maintained an optimal ROC-AUC of 0.858 (95% CI: 0.798–0.918) (Fig. 2B). Quantitative calibration was assessed using the Brier score, which was 0.110 in both training and validation cohorts, indicating excellent model reliability. Precision-recall curves (Fig. 2C, D), calibration curves (Fig. 2E, F), and decision curve analysis (Fig. 2G, H) consistently confirmed the RF model’s superior predictive accuracy, reliability, and clinical utility. Ten-fold cross-validation (Fig. 3) demonstrated stable performance (mean ROCAUC = 0.858, SE = 0.017), while comprehensive metrics (Fig. 4) verified its leading performance in accuracy, precision, and sensitivity. These results establish the RF model as the optimal predictive algorithm for this study. In the external validation set, the optimal random forest model achieved a sensitivity of 0.875, specificity of 0.789, positive predictive value (PPV) of 0.706, and negative predictive value (NPV) of 0.916, which further confirmed its robust generalization ability and clinical reliability.
Fig. 2.
Comparison of ROC curves for different ML algorithms. (A,B) ROC curves; (C,D) Precision-recall curves; (E,F) Calibration curves; (G,H) Decision curve analysis, in the training and validation cohorts, respectively
Fig. 3.
Ten-fold cross-validation results of different machine learning models in the training set
Fig. 4.
Prediction performance of different models
Model explainability
To enhance the clinical practicability of the predictive model, this study employed the SHAP (SHapley Additive exPlanations) method to visualize the feature importance ranking and quantify each variable’s contribution to model output. The SHAP summary plot (Fig. 5) revealed that Surgical bleeding, Duration of surgery, WBC, ALB, Hb, and Adhesiolysis were the top six most influential features for predicting PPOI. To further elucidate the directional relationships between clinical features and PPOI prediction, we constructed SHAP dependency plots for the three most influential variables (Fig. 6). These plots demonstrated that Surgical bleeding, Duration of surgery, and WBC exhibited strong monotonic relationships with model outcomes. Specifically, increased surgical blood loss, prolonged operative duration, and elevated white blood cell counts all drove positive SHAP values, confirming their robust risk-enhancing effects. Collectively, these SHAP analyses validated the reliability and interpretability of our RF-based predictive model.
Fig. 5.
SHAP summary chart
Fig. 6.
SHAP dependency plots of the top three influential clinical features for predicting PPOI. (A) Surgical bleeding; (B) Duration of surgery; (C) White blood cell count (WBC). Each plot shows the trend between the eigenvalue and the SHAP value
Establishment of a web calculator
In order to make it easier to put into clinical use, we have developed an online web calculator based on the RF model. The Web Calculator ( https://zw17786325639.shinyapps.io/ppoi/ ) enables users to enter information about the patient’s clinical indicators that can be used to predict the likelihood of PPOI. Figure 7 displays a screenshot of the web calculator interface, which provides an intuitive visualization of the prediction result and feature contributions.
Fig. 7.
Screenshot of the web calculator for predicting PPOI
Discussion
In recent years, with the development of diagnostic and treatment measures, the prognosis of colorectal cancer (CRC) has been improving. However, the recurrence rate of patients undergoing radical resection for colorectal cancer is as high as 29.9%, and the prognosis is usually worse for those with peritoneal metastasis [16].Rectal cancer is located in a narrow pelvic cavity, making it difficult to remove the primary tumor and pelvic peritoneal metastases; therefore, achieving complete cytoreduction is a challenge and increases the possibility of residual tumor, with patients with rectal cancer being more prone to peritoneal metastasis and postoperative complications [17].Currently, laparoscopic radical rectal cancer combined with oxaliplatin intraperitoneal thermal perfusion therapy has been widely used in clinical practice. Existing studies have proven the safety of combined oxaliplatin intraperitoneal thermal perfusion chemotherapy during surgery for advanced colorectal cancer, and the efficacy is considered good [18].PPOI is a common complication that hinders postoperative recovery and is a common complication after laparoscopic radical rectal cancer therapy [19, 20]. It is of great significance to predict the PPOI secondary to laparoscopic radical resection of rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy.
We developed and validated 8 machine learning models to predict PPOI secondary to laparoscopic radical resection of rectal cancer in combination with lobaplatin hyperthermic intraperitoneal chemotherapy. We found that the RF model had better prediction performance .Different from some nomogram models, we further provide a dynamic structure and establish a network calculator based on the RF model to predict the probability of PPOI secondary to laparoscopic radical resection of rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy, and improve the applicability of the model. Our analysis found that Surgical bleeding, Duration of surgery, WBC, ALB, Hb, and Adhesiolysis were key predictors. This prediction model, which estimates the probability of PPOI secondary to laparoscopic radical resection of rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy, may support clinicians in making more informed preoperative risk assessments and clinical decisions, with the potential to improve patient outcomes and quality of life.
In this study, an association was concluded between Surgical bleeding and PPOI, which is consistent with the findings of Mohsen Alhashemi et al. and Maria et al. [21–24]. Surgical bleeding is associated with a higher risk of PPOI secondary to laparoscopic radical resection of rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy as follows: increased surgical bleeding may trigger an elevated sympathetic and hormonal stress response, which in turn inhibits gastrointestinal activity; It can also be used as an indicator of intestinal injury and can directly be associated with the development of PPOI [25–27]. As the duration of surgery increases, anesthesia is associated with prolonged duration of intestinal paralysis, which may delay the recovery of postoperative bowel function and the occurrence of PPOI. Using a prospective study of 327 patients with risk factor analysis, 27% of whom developed PPOI, Vather et al. identified a number of independent risk factors, including prolonged duration of surgery. In a retrospective analysis, the duration of surgery greater than three hours was considered an independent predictor of the development of PPOI [28].After CRC surgery, inflammatory factors such as WBC are associated with the onset and development of PPOI [29]. Kalff et al. observed that minor surgical manipulation of the rat small intestine resulted in significant leakage of leukocytes into the muscularis propria, demonstrating for the first time that the postoperative cellular inflammatory response ( extravasation of leukocytes and their secretions) directly led to a decrease in the function of annular smooth muscle in the small intestine after surgery [30].A study primarily in patients with chronic gastrointestinal diseases found sufficient data to suggest that the inflammatory response associated with WBC plays a key role in the development of PPOI [31].Preoperative hemoglobin insufficiency, or preoperative anemia, can lead to dysfunction of multiple organs and systems, affecting blood circulation in the intestinal wall, destroying the intestinal mucosal barrier, leading to intestinal wall muscle dysfunction, loss of intestinal peristalsis, increasing the risk of intestinal spasm, and increasing the incidence of postoperative PPOI [32]. Second, perioperative transfusions may be present for anemia, and multiple studies have shown that perioperative transfusions suppress immune function and are associated with tumor recurrence and metastasis, and transfusions can also promote the progression of PPOI by inducing inflammatory responses and immunosuppression [25, 33–36].Hypoproteinemia may contribute to intestinal edema, thus affecting the recovery of intestinal function [37]. In one study, a progressive increase in the risk of postoperative complications was demonstrated for every 0.5 mg/dl decrease in albumin in patients after colorectal surgery [38]. Mechanical separation of intestinal adhesions can damage the intestinal serous membrane and muscle layer, induce local inflammation and activate the neuromuscular network, and form an inflammatory response in the intestinal muscle layer, which is the core mechanism of PPOI [39]. Surgical traction and ischemia-reperfusion can activate sympathetic nerves, trigger systemic inflammation, and further inhibit intestinal peristalsis. When adhesiolysis is combined with HIPEC, hyperthermic chemotherapy will aggravate the damage to the peritoneum and intestinal mucosa, form a secondary blow with mechanical injury, and the superposition of inflammation reduces the tolerance of intestinal trauma and significantly increases the incidence of PPOI.
Laparoscopic radical rectal cancer surgery combined with lobaplatin hyperthermic intraperitoneal chemotherapy is a high-risk factor for postoperative intestinal obstruction, therefore, for rectal cancer patients, it is especially important to comprehensively evaluate the patient’s physical condition, especially Surgical bleeding, Duration of surgery, WBC, ALB, Hb, and Adhesiolysis for high-risk patients. The model we developed differs from the linear models used in traditional studies in that it maximizes the utilization of clinical parameters and significantly improves the accuracy of diagnosis. Based on the RF model, we built a concise, intuitive, and dynamic online application that calculates the probability of PPOI occurring. It is helpful to weigh the potential risk of intestinal obstruction after laparoscopic radical surgery for rectal cancer combined with lobaplatin hyperthermic intraperitoneal chemotherapy, and to individualize the most suitable treatment plan for this patient.
Although the predictive model we developed showed good accuracy and was visualized by an online calculator, making it convenient for rapid clinical evaluation, there are several limitations that should be addressed. First, the present study enrolled a highly selected cohort of patients with isolated peritoneal metastasis, PCI < 10, and eligibility for radical resection plus HIPEC, which represents a clinically indicated population for HIPEC rather than an unselected routine cohort. This strict selection ensures clinical relevance but may limit the direct generalizability to unselected patients in daily practice. Second, the AUC decreased from 0.917 in the training set to 0.858 in the validation set, suggesting potential overfitting, which may be exacerbated by the use of multiple complex machine learning models. Third, the patient selection pathway, clinical indications for HIPEC, and staging logic were based on institutional standard protocols and intraoperative assessment; future studies are warranted to provide more detailed and standardized reporting of these processes to further improve reproducibility and generalizability.
Conclusion
In summary, we constructed an ML-based prediction model, which all had good performance and could predict PPOI secondary to laparoscopic radical resection of rectal cancer combined with intraperitoneal hyperthermic chemotherapy with lobaplatin based on predictor variables. In addition, we build online calculators to translate research findings into clinical practice.
Acknowledgements
Not applicable.
Authors’ contributions
W.Z. (Wei Zhang) and W.Z. (Weibin Zhang) contributed to the conception and design of the study. N.Y., F.C., Q.X., Y.L., J.X., C.G., Z.Y., D.M., T.Z., and Z.B. were involved in the collection of the clinical data. W.Z. (Wei Zhang) and B.C. performed the statistical analyses and wrote the first draft of the manuscript. B.C. and N.Y. revised the manuscript critically for important intellectual content. All authors read and approved the final submitted version of the manuscript. W.Z. (Wei Zhang) and B.C. are co-first authors with equal contributions to the manuscript.
Funding
This research received no external funding.
Data availability
The datasets generated and/or analyzed during the current work are not publicly available because of patient confidentiality, but are available from the corresponding author upon reasonable request. Data requests should be directed to the corresponding author’s contact information.
Declarations
Ethics approval and consent to participate
This retrospective study was approved by the Ethics Committee of Jinzhou Central Hospital of Jinzhou Medical University (Ethics Approval No. JZZXYYLL20241204-05). The study was conducted in accordance with the 1964 Declaration of Helsinki. Informed consent was waived by the ethics committee due to the retrospective nature of this study.
Consent for publication
Not applicable. No identifiable individual data or images are included in this article.
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.
Wei Zhang and Baitong Che contributed equally to this work.
References
- 1.Chen W, et al. Cancer statistics in China, 2015. CA Cancer J Clin. 2016;66(2):115–32. [DOI] [PubMed] [Google Scholar]
- 2.Liao CK, et al. Prognostic value of the C-reactive protein to albumin ratio in colorectal cancer: an updated systematic review and meta-analysis. World J Surg Oncol. 2021;19(1):139. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Wang W, et al. Comparison of non-schistosomal colorectal cancer and schistosomal colorectal cancer. World J Surg Oncol. 2020;18(1):149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Gamboa AC, et al. Optimal Surveillance Frequency After CRS/HIPEC for Appendiceal and Colorectal Neoplasms: A Multi-institutional Analysis of the US HIPEC Collaborative. Ann Surg Oncol. 2020;27(1):134–46. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Klaver CEL, et al. Adjuvant hyperthermic intraperitoneal chemotherapy in patients with locally advanced colon cancer (COLOPEC): a multicentre, open-label, randomised trial. Lancet Gastroenterol Hepatol. 2019;4(10):761–70. [DOI] [PubMed] [Google Scholar]
- 6.Razenberg LG, et al. Bevacizumab in Addition to Palliative Chemotherapy for Patients With Peritoneal Carcinomatosis of Colorectal Origin: A Nationwide Population-Based Study. Clin Colorectal Cancer. 2016;15(2):e41–6. [DOI] [PubMed] [Google Scholar]
- 7.Chen N, et al. LASP1 induces colorectal cancer proliferation and invasiveness through Hippo signaling and Nanog mediated EMT. Am J Transl Res. 2020;12(10):6490–500. [PMC free article] [PubMed] [Google Scholar]
- 8.Verwaal VJ, et al. 8-year follow-up of randomized trial: cytoreduction and hyperthermic intraperitoneal chemotherapy versus systemic chemotherapy in patients with peritoneal carcinomatosis of colorectal cancer. Ann Surg Oncol. 2008;15(9):2426–32. [DOI] [PubMed] [Google Scholar]
- 9.Mehta SS, et al. Complications of Cytoreductive Surgery and HIPEC in the Treatment of Peritoneal Metastases. Indian J Surg Oncol. 2016;7(2):225–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sluiter NR, et al. Metachronous Peritoneal Metastases After Adjuvant Chemotherapy are Associated with Poor Outcome After Cytoreduction and HIPEC. Ann Surg Oncol. 2018;25(8):2347–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Liang WQ, et al. Preoperative albumin levels predict prolonged postoperative ileus in gastrointestinal surgery. World J Gastroenterol. 2020;26(11):1185–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Vather R, Trivedi S, Bissett I. Defining postoperative ileus: results of a systematic review and global survey. J Gastrointest Surg. 2013;17(5):962–72. [DOI] [PubMed] [Google Scholar]
- 13.Ali I, et al. Platinum compounds: a hope for future cancer chemotherapy. Anticancer Agents Med Chem. 2013;13(2):296–306. [DOI] [PubMed] [Google Scholar]
- 14.Feng L, et al. [Safety evaluation of intraoperative peritoneal chemotherapy with Lobaplatin for advanced colorectal cancers]. Zhonghua Wei Chang Wai Ke Za Zhi. 2015;18(10):1006–10. [PubMed] [Google Scholar]
- 15.Shang A, et al. Effect and safety of intraoperative intraperitoneal chemotherapy on patients suffering from colorectal cancer. World J Surg Oncol. 2021;19(1):84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Obrand DI, Gordon PH. Incidence and patterns of recurrence following curative resection for colorectal carcinoma. Dis Colon Rectum. 1997;40(1):15–24. [DOI] [PubMed] [Google Scholar]
- 17.Sugarbaker PH. Update on the prevention of local recurrence and peritoneal metastases in patients with colorectal cancer. World J Gastroenterol. 2014;20(28):9286–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Shan L, et al. Lobaplatin suppresses proliferation and peritoneal metastasis of colorectal cancer in a preclinical model. Biomed Pharmacother. 2018;108:486–91. [DOI] [PubMed] [Google Scholar]
- 19.Baratti D, et al. Past, present and future of adjuvant HIPEC in patients at high risk for colorectal peritoneal metastases. Eur J Surg Oncol. 2020;46(5):737–9. [DOI] [PubMed] [Google Scholar]
- 20.Tsuyoshi H, et al. Hyperthermic intraperitoneal chemotherapy (HIPEC) for gynecological cancer. J Obstet Gynaecol Res. 2020;46(9):1661–71. [DOI] [PubMed] [Google Scholar]
- 21.Alhashemi M, et al. Incidence and predictors of prolonged postoperative ileus after colorectal surgery in the context of an enhanced recovery pathway. Surg Endosc. 2019;33(7):2313–22. [DOI] [PubMed] [Google Scholar]
- 22.Charalambides M, et al. A systematic review of the literature assessing operative blood loss and postoperative outcomes after colorectal surgery. Int J Colorectal Dis. 2022;37(1):47–69. [DOI] [PubMed] [Google Scholar]
- 23.Guo F, et al. Nomogram for predicting prolonged postoperative ileus after laparoscopic low anterior resection for rectal cancer. World J Surg Oncol. 2023;21(1):380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Wang X, Li T. Postoperative pain pathophysiology and treatment strategies after CRS + HIPEC for peritoneal cancer. World J Surg Oncol. 2020;18(1):62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Cata JP, et al. Inflammatory response, immunosuppression, and cancer recurrence after perioperative blood transfusions. Br J Anaesth. 2013;110(5):690–701. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Huang DD, et al. Prediction of Prolonged Postoperative Ileus After Radical Gastrectomy for Gastric Cancer: A Scoring System Obtained From a Prospective Study. Med (Baltim). 2015;94(51):e2242. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Vather R, Bissett IP. Risk factors for the development of prolonged post-operative ileus following elective colorectal surgery. Int J Colorectal Dis. 2013;28(10):1385–91. [DOI] [PubMed] [Google Scholar]
- 28.Millan M, et al. Risk factors for prolonged postoperative ileus after colorectal cancer surgery. World J Surg. 2012;36(1):179–85. [DOI] [PubMed] [Google Scholar]
- 29.Blomberg J, et al. Albumin and C-reactive protein levels predict short-term mortality after percutaneous endoscopic gastrostomy in a prospective cohort study. Gastrointest Endosc. 2011;73(1):29–36. [DOI] [PubMed] [Google Scholar]
- 30.Kalff JC, et al. Surgically induced leukocytic infiltrates within the rat intestinal muscularis mediate postoperative ileus. Gastroenterology. 1999;117(2):378–87. [DOI] [PubMed] [Google Scholar]
- 31.Yang X, et al. Effect of remote ischemic preconditioning in patients undergoing laparoscopic colorectal cancer surgery: a randomized controlled trial. Scand J Gastroenterol. 2023;58(6):634–42. [DOI] [PubMed] [Google Scholar]
- 32.Venara A, et al. Postoperative ileus: Pathophysiology, incidence, and prevention. J Visc Surg. 2016;153(6):439–46. [DOI] [PubMed] [Google Scholar]
- 33.Cardenas MA, et al. Differentiation fate of a stem-like CD4 T cell controls immunity to cancer. Nature. 2024;636(8041):224–32. [DOI] [PubMed] [Google Scholar]
- 34.Gunka I, et al. Impact of blood transfusions on survival and recurrence in colorectal cancer surgery. Indian J Surg. 2013;75(2):94–101. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Tarantino I, et al. Blood transfusion does not adversely affect survival after elective colon cancer resection: a propensity score analysis. Langenbecks Arch Surg. 2013;398(6):841–9. [DOI] [PubMed] [Google Scholar]
- 36.Wang Z, et al. Machine learning-based model for CD4(+) conventional T cell genes to predict survival and immune responses in colorectal cancer. Sci Rep. 2024;14(1):24426. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Kim Y, et al. Peri-operative Inflammatory Marker as a Predictive Factor for Prolonged Post-operative Ileus After Gastrectomy for Gastric Cancer. J Neurogastroenterol Motil. 2021;27(4):588–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Bendersky V, et al. Determining the Optimal Quantitative Threshold for Preoperative Albumin Level Before Elective Colorectal Surgery. J Gastrointest Surg. 2017;21(4):692–9. [DOI] [PubMed] [Google Scholar]
- 39.Venara A, et al. Incidence and Risk Factors for Severity of Postoperative Ileus After Colorectal Surgery: A Prospective Registry Data Analysis. World J Surg. 2020;44(3):957–66. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated and/or analyzed during the current work are not publicly available because of patient confidentiality, but are available from the corresponding author upon reasonable request. Data requests should be directed to the corresponding author’s contact information.