# Development and multicenter validation of a predictive model for malignant pleural effusion recurrence

## TL;DR

Researchers created a machine learning model to predict if lung cancer patients will experience a recurrence of malignant pleural effusion within 3 months.

## Contribution

A novel Elastic Net machine learning model using four routine clinical features for predicting MPE recurrence was developed and validated.

## Key findings

- The model achieved high performance with AUCs of 0.848 and 0.940 in external validation.
- It outperformed other machine learning approaches in predicting MPE recurrence.
- An interactive risk stratification tool was developed to classify patients into four risk groups.

## Abstract

Early prediction of malignant pleural effusion (MPE) recurrence within 3 months is essential for optimizing management in lung cancer patients. This study developed and validated a machine learning model to estimate the 3-month recurrence risk of MPE in patients with newly diagnosed lung cancer. Using data from 221 patients for model training and 237 from two external validation cohorts, the Elastic Net model—based solely on four routine clinical features (treatment regimen, alanine aminotransferase, total pleural effusion volume, and tumor diameter)—achieved excellent performance, with areas under the curve of 0.848 and 0.940 in external validation. The model significantly outperformed other machine learning approaches. An interactive risk stratification tool was further developed to classify patients into four risk groups, enabling early identification and individualized management of high-risk patients. This tool offers a practical and generalizable solution for guiding clinical decision-making.

•Developed a machine learning model to predict 3-month MPE recurrence in lung cancer patients•The Elastic Net model uses only four routine clinical features•The model achieved excellent external validation performance with AUCs of 0.848 and 0.940•An interactive risk stratification tool was developed for individualized management

Developed a machine learning model to predict 3-month MPE recurrence in lung cancer patients

The Elastic Net model uses only four routine clinical features

The model achieved excellent external validation performance with AUCs of 0.848 and 0.940

An interactive risk stratification tool was developed for individualized management

Health sciences; Medicine; Medical specialty; Oncology; Health technology