Defining histamine H2 receptor antagonist response in critically ill patients with heart failure: a machine learning cluster analysis.

Abstract

Background: Recent studies showed histamine H2 receptor antagonists (H2RAs) exposure was associated with reduced mortality in heart failure (HF) patients. However, specific HF patients who are sensitive to H2RAs exposure or not are yet to be further defined.

Aim: This study aimed to identify HF patient characteristics that may benefit from H2RAs exposure.

Method: Neural network-based variational autoencoders and Gaussian Mixture Model (GMM) clustering methods were employed to classify the critically ill patients with HF exposed to H2RAs based on Medical Information Mart for Intensive Care III and IV databases. Binary logistic and multivariable Cox regression analysis based on propensity score matching (PSM) were employed to estimate the association between H2RAs exposure of each cluster and all-cause mortality of included patients.

Results: A total of 9,585 H2RAs users among 23,855 included HF patients were identified into four clusters according to GMM clustering: cluster 1 (combined with acute kidney failure, septic shock, and pneumonia), cluster 2 (combined with atrial fibrillation), cluster 3 (combined with coronary artery disease (CAD) and/or had higher urine output), and cluster 4 (co-administered with calcium-antagonists). The cluster 3 patients were significantly associated with reduced all-cause mortality compared with non-H2RAs users after PSM, which were further validated in 14,280 HF patients from the large multi-center electronic intensive care unit Collaborative Research Database (eICU-CRD).

Conclusion: Histamine H2 receptor antagonist exposure was more sensitive in HF patients who were combined with CAD. Furthermore, male HF patients or those with higher urine output were also sensitive to H2RAs exposure.