Acute-phase biomarker correlation networks and myopericardial involvement on follow-up cardiac magnetic resonance imaging after COVID-19

Abstract Myopericardial involvement (MPI) is frequently reported on cardiac magnetic resonance imaging (MRI) after COVID-19, but the mechanisms linking acute SARS-CoV-2 infection to these findings are unclear, and individual circulating biomarkers may not capture the interacting biological processes involved. We investigated whether the correlation structure among circulating biomarkers during acute COVID-19, rather than isolated biomarker concentrations, is associated with MPI on follow-up cardiac MRI. Of 180 patients hospitalized with COVID-19 and evidence of myocardial injury or thrombotic risk who were prospectively enrolled, 37 (20.6%) attended the follow-up visit with a completed cardiac MRI and had paired serum from acute hospitalization (T1) and follow-up (T2); reasons for non-inclusion of the remaining 143 patients are shown in a participant flow diagram, and non-included patients were older and had a higher in-hospital mortality. Thirty-one serum biomarkers reflecting cardiac injury, inflammation and immune activation, matrix remodeling and oxidative stress were measured at both time points, and cardiac MRI was performed a median of 3.7 months (range 0.6–11.8) after discharge. Eighteen patients (48.6%) had MPI, comprising pericardial involvement in 13 and non-ischemic myocardial late gadolinium enhancement in 6; no patient had myocardial edema, and none fulfilled the updated Lake Louise criteria for acute myocarditis. Individual biomarker concentrations at T1 did not differ between groups after correction for multiple testing. In exploratory correlation-network analyses, correlations spanning different functional biomarker categories were stronger in patients with MPI in all six inter-category blocks, whereas three of four intra-category blocks shifted in the opposite direction; this coordinated pattern was significant by permutation testing ( p = 0.013) and was absent at follow-up ( p = 0.99), when the two networks had converged. The proportion of significant correlations spanning functional categories was higher in the affected group (44.4% versus 14.3%) and this direction was preserved in all 37 leave-one-patient-out replicates. Comparisons of individual biomarker pairs did not reach significance after correction for multiple testing, a procedure with approximately 7% power at this sample size, so the difference between groups is one of network organization rather than of any single correlation. Because no cardiac MRI was performed during the acute phase and no pre-infection imaging was available, the follow-up abnormalities cannot be confirmed as new, as persistent, or as directly related to COVID-19. These findings are hypothesis-generating and require confirmation in larger cohorts with acute-phase imaging and clinical outcome data.