Predicting 6-week mortality after acute variceal bleeding: role of Classification and Regression Tree analysis

Area of Science:

• Hepatology
• Gastroenterology
• Clinical Prognostics

Background:

• Existing prognostic scores for acute variceal bleeding (AVB) mortality often rely on logistic regression and may have limitations.
• Clinical utility of current models can be restricted, necessitating improved predictive tools.

Purpose of the Study:

• To evaluate a novel Classification and Regression Tree (CART) analysis approach for predicting 6-week mortality in AVB patients.
• To compare the efficacy of CART analysis against established models like MELD and Child-Pugh scores.

Main Methods:

• A cohort of sixty consecutive cirrhotic patients with AVB was studied.
• Classification and Regression Tree (CART) analysis, MELD, and Child-Pugh scores were applied to predict 6-week mortality.
• Receiver operating characteristic (ROC) curves were used to assess the predictive performance of each model.

Main Results:

• The study observed 6-week rebleeding and mortality rates of 30% and 22%, respectively.
• Both Child-Pugh and MELD scores demonstrated clinical relevance in predicting 6-week mortality.
• CART analysis yielded a straightforward algorithm using albumin, bilirubin, and in-hospital rebleeding, effectively stratifying patients into low (3% mortality) and high (80% mortality) risk groups. The areas under the ROC curves (AUROC) were 0.88 for MELD, 0.84 for Child-Pugh, and 0.91 for CART.

Conclusions:

• A simple CART algorithm incorporating albumin, bilirubin, and in-hospital rebleeding provides accurate prognostic assessment for 6-week mortality following AVB.
• This CART-based approach offers a valuable tool for clinicians managing patients with acute variceal bleeding.