Prediction of mortality after esophagectomy: A comprehensive analysis of various risk scores in a national esophageal center

TL;DR

This study compares nine risk models to predict 30-day mortality after esophagectomy, finding that O-POSSUM and Charlson scores perform best.

Contribution

The study validates and compares nine risk scores for predicting mortality after esophagectomy in a national center.

Findings

O-POSSUM and Charlson scores showed highest predictive accuracy with AUCs of 0.832 and 0.806.

PER and Fuchs models had moderate predictive ability, while ASA and Philadelphia scores were less useful.

Predictive performance varied significantly across patient subgroups.

Abstract

Esophagectomy remains the cornerstone treatment for esophageal cancer but is associated with significant perioperative morbidity and mortality, even in specialized centers. Accurate preoperative risk assessment is crucial to improve patient outcomes, and various predictive models are available for risk stratification. This study aimed to validate and compare the performance of nine established predictive models in forecasting 30-day mortality following esophagectomy in a high-volume esophageal cancer center.

We retrospectively analyzed of 101 patients who underwent esophagectomy between January 2020 and December 2023 was performed. Clinicopathological characteristics and mortality data were obtained. The predictive accuracy of nine risk models, including the Esophageal-POSSUM (O-POSSUM), Charlson Comorbidity Index (Charlson), Postoperative Estimation of Risk (PER), and Fuchs scores, was assessed using logistic regression, Hosmer-Lemeshow tests for calibration, and the area under the receiver operating characteristic curve (AUC) for discrimination. Mann-Whitney U tests were used to evaluate significant differences between survivors and non-survivors.

The 30-day mortality rate was 8.91 %. The O-POSSUM and Charlson scores demonstrated the highest predictive accuracy with AUCs of 0.832 and 0.806, respectively. The PER and Fuchs models also showed significant associations with mortality but with moderate predictive ability. Models such as the American Society of Anesthesiologists (ASA) and Philadelphia scores demonstrated limited predictive utility. Significant differences in predictive performance were noted across patient subgroups.

The O-POSSUM and Charlson scores were reliable tools for predicting 30-day mortality after esophagectomy. Other models require further validation and refinement. Tailoring risk assessment models in specific clinical settings may enhance their predictive accuracy and contribute to improved patient outcomes.

Unlabelled Image

Figures