Construction of a metabolism-malnutrition-inflammation prognostic risk score in patients with heart failure with preserved ejection fraction: a machine learning based Lasso-Cox model.

Background: Metabolic disorder, malnutrition and inflammation are involved and interplayed in the mechanisms of heart failure with preserved ejection fraction (HFpEF). We aimed to construct a Metabolism-malnutrition-inflammation score (MIS) to predict the risk of death in patients with HFpEF.

Methods: We included patients diagnosed with HFpEF and without infective or systemic disease. 20 biomarkers were filtered by the Least absolute shrinkage and selection operator (Lasso)-Cox regression. 1000 times bootstrapping datasets were generated to select biomarkers that appeared above 95% frequency in repetitions to construct the MIS.

Results: Among 1083 patients diagnosed with HFpEF, 342 patients (31.6%) died during a median follow-up period of 2.5 years. The MIS was finally constructed based on 6 biomarkers, they were albumin (ALB), red blood cell distribution width-standard deviation (RDW-SD), high-sensitivity C-reactive protein (hs-CRP), lymphocytes, triiodothyronine (T3) and uric acid (UA). Incorporating MIS into the basic predictive model significantly increased both discrimination (∆C-index = 0.034, 95% CI 0.013-0.050) and reclassification (IDI, 6.6%, 95% CI 4.0%-9.5%; NRI, 22.2% 95% CI 14.4%-30.2%) in predicting all-cause mortality. In the time-dependent receiver operating characteristic (ROC) analysis, the mean area under the curve (AUC) for the MIS was 0.778, 0.782 and 0.772 at 1, 3, and 5 years after discharge in the cross-validation sets. The MIS was independently associated with all-cause mortality (hazard ratio: 1.98, 95% CI [1.70-2.31], P < 0.001).

Conclusions: A risk score derived from 6 commonly used inflammatory, nutritional, thyroid and uric acid metabolic biomarkers can effectively identify high-risk patients with HFpEF, providing potential individualized management strategies for patients with HFpEF.