Nomogram to predict postoperative infectious complications after surgery for colorectal cancer: a retrospective cohort study in China

Interpretable machine learning-based clinical prediction model for predicting lymph node metastasis in patients with intrahepatic cholangiocarcinoma

OBJECTIVE

Prediction of lymph node metastasis (LNM) for intrahepatic cholangiocarcinoma (ICC) is critical for the treatment regimen and prognosis. We aim to develop and validate machine learning (ML)-based predictive models for LNM in patients with ICC.

METHODS

A total of 345 patients with clinicopathological characteristics confirmed ICC from Jan 2007 to Jan 2019 were enrolled. The predictors of LNM were identified by the least absolute shrinkage and selection operator (LASSO) and logistic analysis. The selected variables were used for developing prediction models for LNM by six ML algorithms, including Logistic regression (LR), Gradient boosting machine (GBM), Extreme gradient boosting (XGB), Random Forest (RF), Decision tree (DT), Multilayer perceptron (MLP). We applied 10-fold cross validation as internal validation and calculated the average of the areas under the receiver operating characteristic (ROC) curve to measure the performance of all models. A feature selection approach was applied to identify importance of predictors in each model. The heat map was used to investigate the correlation of features. Finally, we established a web calculator using the best-performing model.

RESULTS

In multivariate logistic regression analysis, factors including alcoholic liver disease (ALD), smoking, boundary, diameter, and white blood cell (WBC) were identified as independent predictors for LNM in patients with ICC. In internal validation, the average values of AUC of six models ranged from 0.820 to 0.908. The XGB model was identified as the best model, the average AUC was 0.908. Finally, we established a web calculator by XGB model, which was useful for clinicians to calculate the likelihood of LNM.

CONCLUSION

The proposed ML-based predicted models had a good performance to predict LNM of patients with ICC. XGB performed best. A web calculator based on the ML algorithm showed promise in assisting clinicians to predict LNM and developed individualized medical plans.

The effect of surgical starting time on elective colorectal cancer surgery: A propensity score matching analysis

The purpose of the current study is to analyze whether surgical starting time affects the short-term outcomes of elective colorectal cancer (CRC) surgery. We retrospectively collected CRC patients who underwent elective surgery from Jan 2008 to Jan 2021 in a single clinical center. The effect of surgical starting time (morning surgery vs afternoon surgery, day surgery vs night surgery) on elective CRC surgery was analyzed using propensity score matching (PSM). A total of 6783 patients were included in the current study. There were 5751 patients in day surgery group and 1032 patients in night surgery group, and there were 2920 patients in morning surgery group and 2831 patients in afternoon surgery group. After 1:1 ratio PSM, there were no significant difference in terms of the baseline information (P > .05). Day surgery group had longer operation time (P = .000) and longer hospital stay (P = .029) than night surgery group after PSM. Morning surgery group had longer operation time than afternoon surgery group before PSM (P = .000) and after PSM (P = .000). Univariate and multivariate analysis of the total of 6783 patients were conducted to find predictors of complications, and found that night surgery was a predictor of major complications (P = .002, OR = 1.763, 95% CI = 1.222-2.543) but not a predictor of overall complications (P = .250, OR = 1.096, 95% CI = 0.938-1.282). Night surgery is a predictor of major complications after elective CRC surgery, therefore, surgeons should be careful when operating at night.

Prediction of postoperative infectious complications in elderly patients with colorectal cancer: a study based on improved machine learning

BACKGROUND

Infectious complications after colorectal cancer (CRC) surgery increase perioperative mortality and are significantly associated with poor prognosis. We aimed to develop a model for predicting infectious complications after colorectal cancer surgery in elderly patients based on improved machine learning (ML) using inflammatory and nutritional indicators.

METHODS

The data of 512 elderly patients with colorectal cancer in the Third Affiliated Hospital of Anhui Medical University from March 2018 to April 2022 were retrospectively collected and randomly divided into a training set and validation set. The optimal cutoff values of NLR (3.80), PLR (238.50), PNI (48.48), LCR (0.52), and LMR (2.46) were determined by receiver operating characteristic (ROC) curve; Six conventional machine learning models were constructed using patient data in the training set: Linear Regression, Random Forest, Support Vector Machine (SVM), BP Neural Network (BP), Light Gradient Boosting Machine (LGBM), Extreme Gradient Boosting (XGBoost) and an improved moderately greedy XGBoost (MGA-XGBoost) model. The performance of the seven models was evaluated by area under the receiver operator characteristic curve, accuracy (ACC), precision, recall, and F1-score of the validation set.

RESULTS

Five hundred twelve cases were included in this study; 125 cases (24%) had postoperative infectious complications. Postoperative infectious complications were notably associated with 10 items features: American Society of Anesthesiologists scores (ASA), operation time, diabetes, presence of stomy, tumor location, NLR, PLR, PNI, LCR, and LMR. MGA-XGBoost reached the highest AUC (0.862) on the validation set, which was the best model for predicting postoperative infectious complications in elderly patients with colorectal cancer. Among the importance of the internal characteristics of the model, LCR accounted for the highest proportion.

CONCLUSIONS

This study demonstrates for the first time that the MGA-XGBoost model with 10 risk factors might predict postoperative infectious complications in elderly CRC patients.

Impact of visceral obesity on infectious complications after resection for colorectal cancer: a retrospective cohort study

OBJECTIVES

To explore the impact of visceral obesity (VO) measured by preoperative abdominal computed tomography (CT) on postoperative infectious complications for colorectal cancer (CRC) patients and establish a predictive model.

METHODS

Patients who underwent resection for colorectal cancer between January 2015 and January 2021 were enrolled in this study. All patients were measured for body mass index (BMI) and visceral fat area (VFA) preoperatively. Infectious complications were compared between the different groups according to BMI and VO categories. Univariate and multivariate logistic regression were used to analyze whether VO was an independent risk factor for postoperative infectious complications. According to the results of logistic regression, six machine learning approaches were used to establish predictive models and perform internal validation. The best-performing model was interpreted by the SHAPley Additive exPlanations value.

RESULTS

Approximately 64.81% of 520 patients had VO. VO was significantly connected with postoperative infectious complications (P < 0.001), coronary heart disease (P = 0.004), cerebral infarction (P = 0.001), hypertension (P < 0.001), diabetes (P < 0.001), and fatty liver (P < 0.001). The rates of wound infection (P = 0.048), abdominal or pelvic infection (P = 0.006), and pneumonia (P = 0.008) increased obviously in patients with VO. Compared to the low BMI group, a high BMI was found to be significantly associated with hypertension (P=0.007), fatty liver (P＜0.001), and a higher rate of postoperative infection (P=0.003). The results of logistic regression revealed that VO (OR = 2.01, 95% CI 1.17 ~ 3.48, P = 0.012), operation time ≥ 4 h (OR = 2.52, 95% CI 1.60 ~ 3.97, P < 0.001), smoking (OR = 2.04, 95% CI 1.16 ~ 3.59, P = 0.014), ostomy (OR = 1.65, 95% CI 1.04 ~ 2.61, P = 0.033), and chronic obstructive pulmonary disease (COPD) (OR = 2.23, 95% CI 1.09 ~ 4.57, P = 0.029) were independent risk factors. The light gradient boosting machine (LGBM) model displayed the largest area under the receiver operating characteristic curve (AUC) (0.74, 95% CI 0.68 ~ 0.81).

CONCLUSIONS

In this study, VO was superior to BMI in evaluating the influence of obesity on metabolic comorbidities and postoperative infectious complications in colorectal cancer patients.

Development and validation of a nomogram for predicting metachronous peritoneal metastasis in colorectal cancer: A retrospective study

BACKGROUND

Peritoneal metastasis (PM) after primary surgery for colorectal cancer (CRC) has the worst prognosis. Prediction and early detection of metachronous PM (m-PM) have an important role in improving postoperative prognosis of CRC. However, commonly used imaging methods have limited sensitivity to detect PM early. We aimed to establish a nomogram model to evaluate the individual probability of m-PM to facilitate early interventions for high-risk patients.

AIM

To establish and validate a nomogram model for predicting the occurrence of m-PM in CRC within 3 years after surgery.

METHODS

We used the clinical data of 878 patients at the Second Hospital of Jilin University, between January 1, 2014 and January 31, 2019. The patients were randomly divided into training and validation cohorts at a ratio of 2:1. The least absolute shrinkage and selection operator (LASSO) regression was performed to identify the variables with nonzero coefficients to predict the risk of m-PM. Multivariate logistic regression was used to verify the selected variables and to develop the predictive nomogram model. Harrell’s concordance index, receiver operating characteristic curve, Brier score, and decision curve analysis (DCA) were used to evaluate discrimination, distinctiveness, validity, and clinical utility of this nomogram model. The model was verified internally using bootstrapping method and verified externally using validation cohort.

RESULTS

LASSO regression analysis identified six potential risk factors with nonzero coefficients. Multivariate logistic regression confirmed the risk factors to be independent. Based on the results of two regression analyses, a nomogram model was established. The nomogram included six predictors: Tumor site, histological type, pathological T stage, carbohydrate antigen 125, v-raf murine sarcoma viral oncogene homolog B mutation and microsatellite instability status. The model achieved good predictive accuracy on both the training and validation datasets. The C-index, area under the curve, and Brier scores were 0.796, 0.796 [95% confidence interval (CI) 0.735-0.856], and 0.081 for the training cohort and 0.782, 0.782 (95%CI 0.690-0.874), and 0.089 for the validation cohort, respectively. DCA showed that when the threshold probability was between 0.01 and 0.90, using this model to predict m-PM achieved a net clinical benefit.

CONCLUSION

We have established and validated a nomogram model to predict m-PM in patients undergoing curative surgery, which shows good discrimination and high accuracy.

Preoperative Immunocite-Derived Ratios Predict Surgical Complications Better when Artificial Neural Networks Are Used for Analysis-A Pilot Comparative Study

We aimed to comparatively assess the prognostic preoperative value of the main peripheral blood components and their ratios-the systemic immune-inflammation index (SII), neutrophil-to-lymphocyte ratio (NLR), lymphocyte-to-monocyte ratio (LMR), and platelet-to-lymphocyte ratio (PLR)-to the use of artificial-neural-network analysis in determining undesired postoperative outcomes in colorectal cancer patients. Our retrospective study included 281 patients undergoing elective radical surgery for colorectal cancer in the last seven years. The preoperative values of SII, NLR, LMR, and PLR were analyzed in relation to postoperative complications, with a special emphasis on their ability to accurately predict the occurrence of anastomotic leak. A feed-forward fully connected multilayer perceptron network (MLP) was trained and tested alongside conventional statistical tools to assess the predictive value of the abovementioned blood markers in terms of sensitivity and specificity. Statistically significant differences and moderate correlation levels were observed for SII and NLR in predicting the anastomotic leak rate and degree of postoperative complications. No correlations were found between the LMR and PLR or the abovementioned outcomes. The MLP network analysis showed superior prediction value in terms of both sensitivity (0.78 ± 0.07; 0.74 ± 0.04; 0.71 ± 0.13) and specificity (0.81 ± 0.11; 0.69 ± 0.03; 0.9 ± 0.04) for all the given tasks. Preoperative SII and NLR appear to be modest prognostic factors for anastomotic leakage and overall morbidity. Using an artificial neural network offers superior prognostic results in the preoperative risk assessment for overall morbidity and anastomotic leak rate.

Effect of the systemic immune-inflammation index on postoperative complications and the long-term prognosis of patients with colorectal cancer: a retrospective cohort study

BACKGROUND

Colorectal cancer (CRC) is one of the most common malignant tumors of the digestive tract. Surgery is the main way to cure CRC, but the postoperative complication rate and recurrence rate remain high. The systemic immune-inflammation (SII) index reflects a patient's systemic inflammatory state and immune state. Postoperative recurrence and the occurrence of complications are closely related to the inflammatory state and immune state. Thus, the SII index may have some value in predicting postoperative complications and the long-term prognosis of CRC patients, but relevant studies are currently lacking. The present study sought to examine the effect of the SII index on the postoperative complications and long-term prognosis of patients with CRC.

METHODS

From January 2014 to January 2017, the data of 440 patients with CRC who had been admitted to the Affiliated Hospital of Guangdong Medical University were retrospectively collected, and the patients were equally divided into the high and the low SII groups according to their preoperative SII index levels. The postoperative complication rate and postoperative progression-free survival (PFS) and mortality between the 2 groups were compared.

RESULTS

Compared to the low SII group, the incidence of postoperative infection in the high SII group was significantly increased (15.45% vs. 9.09%, P=0.042), mortality was significantly increased at 5 years postoperatively (20.91% vs. 7.27%, P<0.001), and PFS was significantly shortened (P<0.001). The SII index had certain predictive value for postoperative infection in CRC patients, and the area under the curve (AUC) was 0.645 [95% confidence interval (CI): 0.559-0.731, P=0.001]. The SII index also had certain predictive value for the progression of CRC patients within 5 years of surgery, and the AUC was 0.670 (95% CI: 0.610-0.729, P<0.001). Additionally, the SII index had certain predictive value for death within 5 years of surgery in patients with CRC, and the AUC was 0.660 (95% CI: 0.593-0.726, P<0.001). CRC patients with postoperative infection had a significantly shorter PFS period than those who did not develop postoperative infection (P=0.029).

CONCLUSIONS

The SII index has certain predictive value for the diagnosis of postoperative infectious complications and the long-term prognosis of CRC patients.

Nomogram to predict permanent stoma in rectal cancer patients after sphincter-saving surgery

BACKGROUND

Approximately 20 percent of patients with a tumour localized in the low rectum still encounter the possibility of requiring permanent stoma (PS), which can cause drastic changes in lifestyle and physical perceptions.

AIM

To determine the risk factors for PS and to develop a prediction model to predict the probability of PS in rectal cancer patients after sphincter-saving surgery.

METHODS

A retrospective cohort of 421 rectal cancer patients who underwent radical surgery at Taipei Medical University Hospital between January 2012 and December 2020 was included in this study. Univariate and multivariate analyses were performed to identify the independent risk factors for PS. A nomogram was developed according to the independent risk factors obtained in the multivariate analysis. The performance of the nomogram was assessed using a receiver operating characteristic curve and a calibration curve.

RESULTS

The PS rate after sphincter-saving surgery was 15.1% (59/391) in our study after a median follow-up of 47.3 mo (range 7–114 mo). Multivariate logistic regression analysis demonstrated that local recurrence, perirectal abscess, anastomosis site stenosis, perineural invasion, tumor size and operative time were independent risk factors for PS. These identified risk factors were incorporated into the nomogram, and the concordance index of this model was 0.903 (95%CI: 0.851-0.955). According to the calibration curves, the nomogram represents a perfect prediction model.

CONCLUSION

Several risk factors for PS after sphincter-saving surgery were identified. Our nomogram exhibited perfect predictive ability and will improve a physician’s ability to communicate the benefits and risks of various treatment options in shared decision making.

The Effect of Complications on Oncological Outcomes of Colorectal Cancer Patients After Primary Surgery: A Propensity Score Matching Analysis

Purpose

The purpose of this study is to explore the oncologic outcomes of complications on colorectal cancer (CRC) patients who underwent primary surgery using a propensity score matching (PSM) analysis.

Methods

A retrospective study was conducted from Jan 2011 to Jan 2020 in a clinical center. The overall survival (OS) and disease-free survival (DFS) were compared among the no complications group, the major complications group and the minor complications group.

Results

A total of 4250 CRC patients who underwent radical primary surgery were included in the current study. Among them, 927 (21.8%) patients suffered complications. After 1:1 ratio PSM, there were 98 patients in the major complications group and in the minor complications group, and 911 patients in the overall complications group and in the no complications group. There was no significant difference in terms of baseline information after PSM (p>0.05). Complications were independent predictors of OS (p=0.000, HR=1.693, 95% CI=1.476-1.941) and DFS (p=0.000, HR=1.555, 95% CI=1.367-1.768). In terms of specific tumor stage, the no complications group had better OS on all stages (p=0.006) and stage III (p=0.003) CRC than the complications group after PSM. Furthermore, the no complications group had better DFS on all stages (p=0.005) and stage III (p=0.021) CRC than the complications group after PSM. However, there was no significant difference between the minor complications group and the major complications group in different tumor stages (p>0.05).

Conclusion

Complications were associated with poor prognosis of CRC and surgeons should be cautious of the adverse events.

Role of the triad of procalcitonin, C-reactive protein, and white blood cell count in the prediction of anastomotic leak following colorectal resections

Purpose

The enhanced recovery after surgery (ERAS) program expedites patient recovery after major surgery. This study aimed to investigate the role of the triad of procalcitonin (PCT), C-reactive protein (CRP), and white blood cells (WBC) trajectories as a predictive biomarker for the anastomotic leak (AL) after colorectal surgery.

Method

Patients who had colorectal anastomosis were prospectively included. Postoperative clinical and laboratory parameters and outcomes were collected and analyzed. The 5-day trajectories of PCT, CRP, and WBC were evaluated. Based on the trajectory of the three biomarkers, we compared patients with and without AL as detected during the first 30 days after surgery using the area under receiver operator characteristic curves (AUC) for logistic estimation.

Results

This study included 205 patients, of whom 56% were men and 43.9% were women with a mean age of 56.4 ± 13.1 years. Twenty-two patients (10.7%) had AL; 77.3% underwent surgery, and 22.7% were treated with drainage and antibiotics. Procalcitonin was the best predictor for AL compared to CRP and WBC at three days postoperatively (AUC: 0.84, 0.76, 0.66, respectively). On day 5, a cutoff value of 4.93 ng/mL for PCT had the highest sensitivity, specificity, and negative predictive value. The predictive power of PCT was substantially improved when combined with either CRP or WBC, or both (AUC: 0.92, 0.92, 0.93, respectively).

Conclusion

The 5-day trajectories of combined CRP, PCT, and WBC had a better predictive power for AL than the isolated daily measurements. Combining the three parameters may be a reliable predictor of early patient discharge, which would be highly beneficial to ERAS programs.