Development and validation of an early prediction model for cardiac death risk in patients with light chain amyloidosis: a multicenter study.

Abstract

Background: Cardiac involvement is the primary driver of death in systemic light chain (AL) amyloidosis. However, the early prediction of cardiac death risk in AL amyloidosis remains insufficient.

Objectives: We aimed to develop a novel prediction model and prognostic scoring system that enables early identification of these high-risk individuals.

Methods: This study enrolled 235 patients with confirmed AL cardiac amyloidosis from three hospitals. Patients from the first hospital were randomly assigned to the training and internal validation sets in an 8:2 ratio, while the external validation set comprised patients from the other two hospitals. Participants were categorized into a cardiac death group and a non-cardiac death group (including survivors and those who died from other causes). Five different machine learning models were used to train model, and model performance was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis.

Results: All five models showed excellent performance on the training and internal validation sets. In external validation, both the Logistic Regression (LR) and Random Forest models achieved an area under the ROC curve of 0.873 and 0.877, respectively, and exhibited superior calibration and decision curve analysis. Considering the comprehensive performance and clinical applicability, the LR model was selected as the final prediction model. The visualization results are ultimately presented in a nomogram. Further analyses were performed on the newly identified predictors.

Conclusions: This prediction model enables early identification and risk assessment of cardiac death in patients with AL amyloidosis, exhibiting considerable predictive ability.