Development and internal validation of a prediction model for severe postoperative complications (Thoracic Morbidity and Mortality Grade ≥ III) after radical esophagectomy.
- 1Department of Cardiothoracic Surgery, Affiliated Hospital of Putian University, Putian, China.
Abstract
Severe complications after radical esophagectomy remain a major clinical concern, highlighting the need for early postoperative risk stratification. Existing prediction models rely largely on preoperative factors, and models incorporating early postoperative inflammatory biomarkers to predict severe complications remain limited. This study aimed to develop and internally validate a nomogram integrating baseline characteristics and postoperative day (POD) 1 inflammatory biomarkers to predict severe postoperative complications after radical esophagectomy. Clinical data from 606 consecutive patients (2019-2025) were analyzed. Severe complications were defined as Thoracic Morbidity and Mortality (TM&M) grade ≥ III within 30 days after surgery. Candidate predictors, measured at baseline and from venous blood collected at 06:00 AM on POD 1, included demographics, surgical variables, and inflammatory biomarkers. Least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation was embedded within a 1,000-bootstrap internal validation framework to avoid data leakage, followed by multivariable logistic regression. Severe complications occurred in 178 patients (29.4%). Age [adjusted odds ratio (OR) =1.04], operative time (OR =1.04 per 10-min increase), and non-middle tumor location (OR =1.58) were independent predictors (all P<0.05). Female sex (OR =0.60), neoadjuvant therapy (OR =0.52), and POD 1 lymphocyte-to-monocyte ratio (D1-LMR, OR =0.76) were protective factors. The nomogram yielded an apparent AUC of 0.685 (95% confidence interval: 0.635-0.735) and a bootstrap optimism-corrected AUC of 0.672. At the optimal cutoff of 0.28 determined by the Youden index, sensitivity was 68.5%, specificity 63.4%, positive predictive value 45.2%, and negative predictive value 82.1%. Calibration was excellent (calibration slope =0.942, intercept =0.018, Brier score =0.184). Adding D1-LMR significantly improved predictive performance over baseline clinical models (ΔAUC =0.034, net reclassification improvement =0.158, P=0.008). Decision curve analysis confirmed positive net clinical benefits across threshold probabilities of 2-45%. Integrating baseline features and D1-LMR, this nomogram provides an exploratory early risk-stratification tool on POD 1 to optimize perioperative management and intensive care unit resource allocation.
Related
- Machine Learning–Based Prediction of Major Postoperative Complications After Noncardiac Surgery Under General Anesthesia Using Routinely Available Electronic Health Record Data: Retrospective Multicenter Development and Validation Study
- Development and validation of a nomogram for predicting anastomotic leakage after esophagectomy: integrating baseline characteristics and early postoperative biomarkers.
- Development and internal validation of a machine learning prediction model for postoperative hemorrhage following modified radical mastectomy for breast cancer
- Development and internal validation of a clinical prediction model for postoperative venous thromboembolism after neurosurgical intervention in patients with traumatic brain injury
- From risk factors to individualized prediction: development and internal validation of a machine learning model for early identification of postoperative pneumonia after lung cancer surgery.
BibTeX
@article{liu2026development,
title = {{Development and internal validation of a prediction model for severe postoperative complications (Thoracic Morbidity and Mortality Grade ≥ III) after radical esophagectomy.}},
author = {Qiuxia Liu and Hao Chen and Xunlin Chen and Qiang Zhang and Xianlin Yin and Pengfei Chen},
journal = {Journal of thoracic disease},
year = {2026},
month = sep,
doi = {10.21037/jtd-2026-1418},
url = {https://europepmc.org/article/MED/42996693},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Development and internal validation of a prediction model for severe postoperative complications (Thoracic Morbidity and Mortality Grade ≥ III) after radical esophagectomy.
AU - Qiuxia Liu
AU - Hao Chen
AU - Xunlin Chen
AU - Qiang Zhang
AU - Xianlin Yin
AU - Pengfei Chen
PY - 2026
DA - 2026/09/28
JO - Journal of thoracic disease
DO - 10.21037/jtd-2026-1418
UR - https://europepmc.org/article/MED/42996693
AB - Severe complications after radical esophagectomy remain a major clinical concern, highlighting the need for early postoperative risk stratification. Existing prediction models rely largely on preoperative factors, and models incorporating early postoperative inflammatory biomarkers to predict severe complications remain limited. This study aimed to develop and internally validate a nomogram integrating baseline characteristics and postoperative day (POD) 1 inflammatory biomarkers to predict severe postoperative complications after radical esophagectomy. Clinical data from 606 consecutive patients (2019-2025) were analyzed. Severe complications were defined as Thoracic Morbidity and Mortality (TM&M) grade ≥ III within 30 days after surgery. Candidate predictors, measured at baseline and from venous blood collected at 06:00 AM on POD 1, included demographics, surgical variables, and inflammatory biomarkers. Least absolute shrinkage and selection operator (LASSO) regression with 10-fold cross-validation was embedded within a 1,000-bootstrap internal validation framework to avoid data leakage, followed by multivariable logistic regression. Severe complications occurred in 178 patients (29.4%). Age [adjusted odds ratio (OR) =1.04], operative time (OR =1.04 per 10-min increase), and non-middle tumor location (OR =1.58) were independent predictors (all P<0.05). Female sex (OR =0.60), neoadjuvant therapy (OR =0.52), and POD 1 lymphocyte-to-monocyte ratio (D1-LMR, OR =0.76) were protective factors. The nomogram yielded an apparent AUC of 0.685 (95% confidence interval: 0.635-0.735) and a bootstrap optimism-corrected AUC of 0.672. At the optimal cutoff of 0.28 determined by the Youden index, sensitivity was 68.5%, specificity 63.4%, positive predictive value 45.2%, and negative predictive value 82.1%. Calibration was excellent (calibration slope =0.942, intercept =0.018, Brier score =0.184). Adding D1-LMR significantly improved predictive performance over baseline clinical models (ΔAUC =0.034, net reclassification improvement =0.158, P=0.008). Decision curve analysis confirmed positive net clinical benefits across threshold probabilities of 2-45%. Integrating baseline features and D1-LMR, this nomogram provides an exploratory early risk-stratification tool on POD 1 to optimize perioperative management and intensive care unit resource allocation.
SN - 2072-1439
SN - 2077-6624
ER -