Plasma proteomics improves risk prediction in heart failure and reveals unique biology in chronic chagas cardiomyopathy
PLOS Neglected Tropical Diseases, 2026
Patané J., Giugni F., Rosa R., Marcondes-Braga F., Mansur A., Pereira A., Krieger J.
| Disease area | Application area | Sample type | Products |
|---|---|---|---|
CVD Infectious Diseases | Pathophysiology Patient Stratification | Plasma |  Olink Explore 3072/384 |
Abstract
Background
Chronic Chagas cardiomyopathy (CCC) remains a major cause of heart failure (HF)–related mortality in Latin America and is increasingly recognized as a global health concern. Prognostic models developed in non-Chagas populations often perform poorly in CCC, highlighting the need for etiology-specific risk stratification.
Methodology/principal findings
We applied high-throughput plasma proteomics to evaluate 2-year mortality risk in CCC compared with other HF etiologies. Baseline plasma from 1,212 adults with heart failure with reduced ejection fraction (HFrEF; LVEF <50%) was analyzed to quantify 734 circulating proteins. CCC was confirmed in 191 participants (16%) by dual Trypanosoma cruzi serology. Two-year mortality was higher in CCC than in the overall HF cohort (26% vs. 16%, p < 0.01). Feature-selection methods identified a nine-protein panel (P9: C1QA, CCL4, REN, EGLN1, COL9A1, GP1BA, ITM2A, CNPY2, NT-proBNP) that improved risk classification compared with NT-proBNP alone, increasing F1-macro by 20% (0.674 vs. 0.560) and integrated time-dependent discrimination for 2-year mortality (iAUC) by 6%. Performance gains varied by HF etiology. Improvements were greatest in hypertensive (+40%) and ischemic (+21%) HF, whereas in CCC the P9 panel underperformed NT-proBNP alone (−16%), suggesting distinct underlying disease biology. External validation in the UK Biobank confirmed generalizability: compared with NT-proBNP, P9 improved F1-macro by 18% and iAUC by 7.4%, reaching an F1-macro of 0.612 in the highest-risk tertile. Pathway enrichment identified 14 CCC-specific pathways, mainly related to fibrosis, integrin signaling, immune dysregulation, and impaired protein trafficking. Exploratory analyses also highlighted potential pathway-linked therapeutic targets consistent with distinct CCC mechanisms.
Conclusions/significance
The P9 proteomic panel improved mortality risk prediction beyond NT-proBNP and the MAGGIC clinical score across most HF etiologies and showed consistent performance in an independent population-based cohort. In contrast, in CCC P9 underperformed NT-proBNP alone, highlighting the distinct biological features of this disease. These findings underscore the limitations of universal biomarker models in CCC and support the need for etiology-specific risk stratification strategies.