Echocardiographic analyses of population-based postCOVID cohorts in Germany

A. Hamza (Rendsburg)¹, J. Thiele (Kiel)², N. E. El Mokhtari (Rendsburg)³, J. Frank (Kiel)⁴, J. Heyckendorf (Kiel)⁵, S. Störk (Würzburg)⁶, C. Morbach (Würzburg)⁷, P. U. Heuschmann (Würzburg)⁸, A. Dempfle (Kiel)⁹, S. Schreiber (Kiel)⁵, D. Frank (Kiel)^4
1Schön Klinik Rendsburg Innere Medizin - Kardiologie Rendsburg, Deutschland; ²University Hospital Schleswig-Holstein (UKSH), Campus Kiel Department of Internal Medicine III Kiel, Deutschland; ³Schön Klinik Rendsburg Klinik für Innere Medizin I -Kardiologie, Pneumologie, Nephrologie Rendsburg, Deutschland; ⁴Universitätsklinikum Schleswig-Holstein Innere Medizin III mit den Schwerpunkten Kardiologie, Angiologie und internistische Intensivmedizin Kiel, Deutschland; ⁵University Medical Center Schleswig-Holstein Department of Internal Medicine I Kiel, Deutschland; ⁶Universitätsklinikum Würzburg Deutsches Zentrum für Herzinsuffizienz/DZHI Würzburg, Deutschland; ⁷Universitätsklinikum Würzburg Medizinische Klinik I, Kardiologie Würzburg, Deutschland; ⁸Universitätsklinikum Würzburg Institut für klinische Epidemiologie und Biometrie Würzburg, Deutschland; ⁹Christian-AlbrechtsUniversität zu Kiel / University Hospital Schleswig-Holstein Institute of Medical Statistics Kiel, Deutschland

Background:
The question whether SARS-CoV-2 infection leaves structural traces on the heart remains debated. Typically, reports of post-COVID cardiac involvement come from hospitalised or symptom-selected patients as well as imaging studies without matched pre-pandemic controls. These circumstances make it difficult to separate viral effects from referral bias and polulation background prevalence. Population-based data on cardiac structure after predominantly mild infections are scarce. We therefore examined whether standardised, observer-independent echocardiographic analyses differ between a large post-COVID cohort and a pre-pandemic population sample.

Methods:
We prospectively enrolled participants approximately 6-9 months after a PCR-confirmed SARS-CoV-2 infection at Kiel, Berlin and Würzburg university hospitals (COVIDOM cohort). The population-based STAAB cohort (Würzburg) which was recruited before 2019 served as controls. All participants underwent standard transthoracic echocardiography (TTE) according to current guidelines. In addition, left ventricular end-diastolic diameter (LVEDD), posterior wall thickness (PWD) and interventricular septal thickness (IVD) were assessed using an in-house AI trained automated reading analysis tool. Mean IVD values were compared between the COVIDOM and STAAB cohorts by regression analyses adjusted for age, sex, BMI, height, eGFR, hypertension, type 2 diabetes, previous heart failure, coronary heart disease and centre.

Results:
We analysed 1829 COVIDOM participants (age range 27 to 83, mean age 49.14 (SD 14.02), 1002 (54.78%) female) and 4965 STAAB participants (age range 29 to 81, mean age 54.88 (SD 11.81), 2604 (52.45%) female). Centrally AI-measured PWD and LVEDD showed significant differences between COVIDOM centers, pointing to relevant assessor effects and were thus not compared to STAAB cohort. Adjusted mean IVS was minimally higher in the COVIDOM cohort than in the STAAB cohort (mean difference 0.11 mm, p=0.017).

Conclusion:
Six to nine months after mostly mild SARS-CoV-2 infection, left ventricular wall thickness and chamber size were essentially comparable to a pre-pandemic population cohort. The adjusted septal difference of 0.11 mm is extremely unlikely to be clinically meaningful. The between-center variability in AI-derived PWD and LVEDD also cautions against pooling automated measurements across sites without prior harmonisation. Taken together, these data argue against relevant systematic structural cardiac remodelling at the population level after non-severe COVID-19.

Keywords:
COVID-19, transthoracic echocardiography; artificial intelligence; LVEDD; posterior wall thickness; interventricular septal thickness;