The effect of time-varying left ventricular ejection fraction on clinical outcome in critically ill COVID-19 patients

Background Although right ventricular (RV) dysfunction and acute cor pulmonale are associated with adverse outcomes in critically ill patients with COVID-19, the prognostic significance of left ventricular (LV) systolic dysfunction remains insufficiently characterized. In particular, dynamic changes in cardiac function during intensive care has seldom been studied. We therefore investigated the association between time-varying LV systolic function, assessed by LV ejection fraction (LVEF) and intensive care unit (ICU) mortality in patients with COVID-19, taking into account right ventricular function and forward flow. Methods This retrospective observational cohort study combined three ICU data repositories comprising 808 patients with 1614 echocardiographic examinations. Time-varying LV systolic function (LVEF ≤ 50%, 51–70%, >70%) was the primary exposure, and ICU mortality was the outcome. Multivariable Cox models with time-varying covariates, including, acute cor pulmonale, left ventricular outflow tract velocity time integral and key clinical characteristics were used to assess associations between LV systolic function phenotypes and outcome. Results At baseline, 27% of patients had abnormal LV systolic function (LVEF ≤ 50%) and 17% had hyperdynamic LV function (LVEF > 70%). The proportion of patients with normal LV systolic function increased over time. Abnormal LV systolic function was most prevalent at the first echocardiographic examination (27%) and least prevalent at the fifth examination (17%), whereas hyperdynamic LV function became more frequent during ICU stay, peaking at the 4 th echocardiographic examination. Abnormal (aHR; 1.174, 95% CI: 0.901-1.531) or hyperdynamic (aHR: 1.041, 95% CI: 0.785-1.379) LV systolic dysfunction was not independently associated with ICU mortality. Conclusion In critically ill patients with COVID-19, LV systolic dysfunction, including both abnormal and hyperdynamic LV phenotypes, were common and dynamic during ICU stay. LV systolic dysfunction was not independently associated with ICU mortality when modelled longitudinally and adjusted for RV dysfunction, forward flow and other clinical covariates. These findings suggest that the prognostic value LVEF phenotypes in severe COVID-19 may be limited. Further studies incorporating more sensitive measures of LV function and biomarkers of myocardial injury may be required to delineate subtle changes not captured by LVEF.