Myoglobin-centred multimarker phenotyping of chronic heart failure: results of an unsupervised cluster analysis

The aim – to identify laboratory-pathophysiological phenotypes of chronic heart failure (CHF) using unsupervised cluster analysis of plasma levels of myoglobin (Myo), N-terminal pro-B-type natriuretic peptide (NT-proBNP), superoxide dismutase 1 (SOD1), haem oxygenase-1 (HO-1), soluble transferrin receptor 1 (sTfR1), and vasodilator-stimulated phosphoprotein (VASP), without using predefined grouping variables, and to assess the association of the identified phenotypes with clinical-etiological patient cohorts.Materials and methods. This prospective, single-center observational study included 123 patients with CHF of various etiologies, including combined and isolated aortic and mitral regurgitation, non-valvular CHF, and postoperative thromboembolic complications. Biomarker concentrations were measured using enzyme-linked immunosorbent assays. Following appropriate logarithmic transformation, the data were standardized as Z-scores. Hierarchical cluster analysis using Ward’s method and squared Euclidean distance, together with k-means clustering, was performed. Agreement between the clustering solutions was assessed using the adjusted Rand index.Results and discussion. Four laboratory-pathophysiological phenotypes were identified: the Myo–NT-proBNP injury and hemodynamic-overload phenotype (n=28), the antioxidant-adaptive phenotype (n=32), the low multimarker-burden phenotype (n=47), and the VASP–sTfR1 vascular-metabolic signaling phenotype (n=16). Agreement between Ward’s method and k-means clustering was moderate (ARI=0.434). The distribution of clinical-etiological cohorts across the phenotypes was statistically significant (Pearson’s χ2 = 79.371; df=12; Monte Carlo exact p<0.001; Cramer’s V = 0.464), but did not correspond to a «one etiology – one phenotype» model. The strongest association was observed between non-valvular CHF and the Myo–NT-proBNP injury and hemodynamic-overload phenotype: 22 patients from this cohort belonged to the injury and hemodynamic-overload phenotype. The hypothesized distinct thrombotic-injury profile was not confirmed.Conclusions. In CHF, myoglobin should be assessed not in isolation but as part of a multimarker panel reflecting hemodynamic stress, antioxidant response, tissue iron availability, and vascular-metabolic signaling. The etiology of CHF is associated with biomarker profiles but does not uniquely determine them. These pilot findings support the potential of multimarker phenotyping for personalized risk stratification in patients with CHF and require external validation.