key: cord-0950897-7a9bkwqh
authors: Park, Jakob; Kim, Yekaterina; Pereira, Jason; Hennessey, Kerrilynn C.; Faridi, Kamil F.; McNamara, Robert L.; Velazquez, Eric J.; Hur, David J.; Sugeng, Lissa; Agarwal, Vratika
title: Understanding the role of left and right ventricular strain assessment in patients hospitalized with COVID-19
date: 2021-06-01
journal: Am Heart J Plus
DOI: 10.1016/j.ahjo.2021.100018
sha: 1ebc8be6a4bfeaf26d3facd23ccc3dfcf2efc7fe
doc_id: 950897
cord_uid: 7a9bkwqh

BACKGROUND: Coronavirus disease 2019 (COVID-19) can cause cardiac injury resulting in abnormal right or left ventricular function (RV/LV) with worse outcomes. We hypothesized that two-dimensional (2D) speckle-tracking assessment of LV global longitudinal strain (GLS) and RV free wall strain (FWS) by transthoracic echocardiography can assist as markers for subclinical cardiac injury predicting increased mortality. METHODS: We performed 2D strain analysis via proprietary software in 48 patients hospitalized with COVID-19. Clinical information, demographics, comorbidities, and lab values were collected via retrospective chart review. The primary outcome was in-hospital mortality based on an optimized abnormal LV GLS value via ROC analysis and RV FWS. RESULTS: The optimal LV GLS cutoff to predict death was −13.8%, with a sensitivity of 85% (95% CI 55–98%) and specificity of 54% (95% CI 36–71%). Abnormal LV GLS > −13.8% was associated with a higher risk of death [unadjusted hazard ratio 5.15 (95% CI 1.13–23.45), p = 0.034], which persisted after adjustment for clinical variables. Among patients with LV ejection fraction (LVEF) >50%, those with LV GLS > −13.8% had higher mortality compared to those with LV GLS < −13.8% (41% vs. 10%, p = 0.030). RV FWS value was higher in patients with LV GLS > −13.8% (−13.7 ± 5.9 vs. −19.6 ± 6.7, p = 0.003), but not associated with decreased survival. CONCLUSION: Abnormal LV strain with a cutoff of >−13.8% in patients with COVID-19 is associated with significantly higher risk of death. Despite normal LVEF, abnormal LV GLS predicted worse outcomes in patients hospitalized with COVID-19. There was no mortality difference based on RV strain.

Coronavirus disease 2019 has spread globally since December 2019 and has resulted in more than 1.9 million deaths worldwide (https://coronavirus.jhu.edu/ -last accessed: 01/09/2020). Collective understanding of the pathogenesis of COVID-19, disease course, and management has been increasing but remains limited. It is now known that cardiac injury associated with COVID-19 infection is associated with increased mortality 1 . Use of cardiac biomarkers such as troponin, NT-pro-brain natriuretic peptide (NT-proBNP), creatine phosphokinase, and presence of new right or left ventricular (RV/LV) dysfunction on echocardiogram allows identification of injury that has already occurred. However, our understanding of the impact of subclinical myocardial dysfunction on outcomes in patients with COVID-19 remains limited.

Two-dimensional (2D) speckle-tracking assessment of strain by transthoracic echocardiography (TTE) has been used as a subclinical marker of impaired myocardial function which predicts cardiovascular outcomes [2] [3] [4] . Assessment of LV strain has been used to monitor cardiotoxic effects of chemotherapy, as well to predict morbidity and mortality in heart failure 4 . RV strain imaging has been used in patients with pulmonary emboli and pulmonary hypertension 5, 6 .

Respiratory viral infections including influenza and acute respiratory distress syndrome (ARDS) have been implicated in acute myocardial injury, however, there is profound lack of data elucidating their respective roles on myocardial strain 7, 8 .

In the most recent study by Li et al (2020) , abnormal right ventricular (RV) strain was associated with worse outcomes in COVID-19 patients and was found to be a predictor of mortality in these J o u r n a l P r e -p r o o f Journal Pre-proof patients 9 . We hypothesize that bi-ventricular strain assessment including LV and RV by transthoracic echocardiogram (TTE) can serve as a supplementary marker of outcomes in patients with COVID-19 infection. This study is the first to investigate the relationships between left ventricular global longitudinal strain (LV GLS) and right ventricular free wall strain (RV FWS) on inpatient mortality outcomes in patients hospitalized with COVID-19.

At our institution, a triage process was implemented for all echocardiogram requests for patients who either were diagnosed with COVID-19 or were considered as patients under investigation (PUI) to determine the appropriateness and necessity of a TTE. This triage system was implemented to prevent disease transmission from the patient to the sonographers and to conserve personal protective equipment. The factors that played an important part in the decision making included presence of shock with unknown left ventricular ejection fraction (LVEF), profound hypoxemia not explained by current pulmonary disease, concern for ACS, aortic dissection or cardiac tamponade; suspected prosthetic or native valve dysfunction. The patients were triaged to either (a) perform the study; (b) defer the study until the patient either has a negative COVID-19 test result or until point-of-care ultrasound (POCUS) is attempted; or (c) cancel the study at this time as it would not change current management. Our triage strategy has been previously published 10 . We performed a retrospective chart review of the electronic medical record (EMR) to obtain data on: patient demographics, medical history (including but not limited to comorbidities and preexisting cardiovascular conditions), and detailed clinical information at the time of the triage.

The primary outcome was the mortality during the same hospital stay. Secondary outcomes included need for the intensive-care unit (ICU), total length of stay, and days requiring mechanical ventilation. Clinical covariates including past medical history, cardiovascular risk factors, pre-existing cardiovascular comorbidities, pre-existing lung disease as well as laboratory values that were collected any time during the hospitalization (troponin T, NT-proBNP, serum creatinine) were included. In-hospital outcomes were assessed and adjudicated by two physicians from March 19 through the time of final data review on Aug 19, 2020.

Bedside transthoracic imaging was done using EPIQ 7C and iE33 ultrasound systems (Philips Medical Systems, Andover, Massachusetts). In order to decrease sonographer exposure to COVID while imaging patients, all studies were performed as focused evaluations catered to provide pertinent information needed from the echocardiogram. To decrease time at bedside, offline strain analysis was performed.

Left ventricular ejection fraction was measured by biplane Simpson method, 3-D method, and visual estimation. Right ventricular size and function were assessed using 4-chamber views.

Assessment was performed using guidelines from the American Society of Echocardiography. to account for inter-observer variability, strain analysis was performed on 15 randomly selected imaging studies by a second experienced reader. Inter-observer variability was assessed using the intraclass correlation coefficient. LV GLS was analyzed using the apical 4-chamber, 3-chamber and 2-chamber views. LV GLS was calculated as the average of the peak longitudinal strain measured in the 12 regions in the apical views. RV FWS was analyzed using the apical 4chamber views and peak average of regional segment strain (basal, mid and apical) was obtained 14 .

The statistical analyses were performed using Stata 13.1 (StataCorp LP, College Station, Texas).

Continuous data were presented as mean ± standard deviation. Categorical variables were presented as absolute and relative frequencies. Group comparisons were performed using unpaired Student's t-test for continuous variables, as well as Fisher's exact test for categorical

variables. An optimal cut-off value for LV GLS was calculated using receiver operating characteristic (ROC) analysis based on the maximum of sensitivity and specificity to predict mortality. The resulting binary LV GLS variable was used for univariable and multivariable stepwise (inclusion threshold p < 0.15) Cox proportional-hazards regression if the proportional J o u r n a l P r e -p r o o f Journal Pre-proof hazards assumption was not violated. For the Cox regression, the primary outcome (mortality) was used as the dependent variable; abnormal LV GLS, abnormal LV ejection fraction (LVEF ≤ 50%), abnormal RV FWS, biomarkers (troponin T ≥ 0.01 ng/ml at triage, peak NT-proBNP, peak creatinine), sex, presence of coronary artery disease, atrial fibrillation and congestive heart failure, as well as pre-existing lung disease and smoking status were used as independent variables. Mortality was further assessed using Kaplan-Meier survival estimator with log-rank and Wilcoxon-Breslow test. Inter-observer variability was reported using the intraclass correlation coefficient (ICC). For all analyses, statistical significance was defined as a two-sided p-value <0.05.

The overall mean age of the study population was 58 ± 16 years and 33% (n = 16) of subjects were female. There were no statistically significant age and sex differences across the groups of survivors and deceased. There were 17 who did not survive hospitalization. Overall, cardiovascular risk factors including hypertension and diabetes had a total prevalence of 60% (n = 29) and 38% (n = 18) respectively, whereas manifest coronary artery disease (n = 5) and congestive heart failure (n = 3) were seen at low rates (Table 1 ). The prevalence of atrial fibrillation, pre-existing lung disease and smoking was higher in the deceased cohort, however, differences were not statistically significant. The remaining baseline characteristics are shown in Table 1 .

Inter-observer variability of strain measurements by ICC showed good to near perfect correlation for all RV and LV strain measurements (Supplemental Table 1 ). ICC for RV FWS was 0.839 and for LV GLS was 0.871 (both single ICC).

RV fractional area change was significantly lower in patients with abnormal RV free wall strain (Table 2) .

Overall, there were no significant differences in mean LV ejection fraction, LV global longitudinal strain, and global circumferential strain ( Table 2) .

When using a cut-off value of LV GLS > -15% to predict death, there was no significant association between LV GLS and inpatient mortality. To determine the optimal cut-off value to maximize the sensitivity and specificity of LV GLS to predict death, ROC analysis was performed. A "new abnormal" threshold of -13.8 was calculated, with a sensitivity of 85 % (95% CI 55 -98%) and a specificity of 54 % (95% CI 36 -71%). The area under curve was 0.70 (95% CI 0.56 -0.83). For RV FWS, the cut off was -25.1, with an AUC of 0.44.

Compared with the "new normal group", those with LV GLS higher than -13.8 ("new abnormal group") had a significantly higher rate of death (41% vs. 10%, p = 0.022). Length of stay and ventilator days were numerically longer in the survival group, but not statistically significant. (Table 3) .

Among patients with normal LV EF ≥50% (n = 35), those with abnormal LV GLS (n = 14) had a significantly higher mortality than those with normal strain (43% vs. 10%, p = 0.030).

Furthermore, there rate of in-hospital mortality for patients with normal LVEF but abnormal LV GLS was similar to the mortality rate for patients with abnormal LVEF [43% (6 out of 14) vs.

38% (5 out of 13), p = 0.564] ( Table 4 ).

In univariate Cox proportional hazard regression, abnormal LV GLS was significantly associated The Kaplan-Meier estimate function is shown in Figure 2 and demonstrates a higher mortality in those patients with normal LVEF, but abnormal LV GLS > -13.8% compared to those with normal LVEF and LV GLS ≤-13.8% (p = 0.021). When comparing the former group to those with abnormal LVEF, there was no statistically significant difference in mortality (p = 0.774).

In this paper, we investigated the effects of abnormal RV and LV strain on mortality of patients hospitalized with COVID-19 and found that abnormal LV strain is associated with a significantly higher risk of in-hospital death, particularly in patients with a normal LV EF. This association persisted after adjusting for echocardiographic markers, clinical risk factors and potential confounders including elevated troponin, underlying heart and lung disease, and smoking status.

Conversely, we did not find a statistically significant association between RV strain abnormalities and survival status.

TTE during the COVID-19 pandemic is used for rapid assessment of cardiac function. However, when looking at the death rates in separate subgroups of isolated abnormal LV strain, isolated abnormal RV strain, and those with abnormal biventricular strain (44% vs. 0% vs. 39%), our data suggests mortality appears mostly driven by the left ventricle. Notably, in our study, we detected no deaths in patients with abnormal RV FWS alone when they had normal LV GLS.

In previous literature, abnormal RV strain has been shown to be a predictor for adverse events 16 , long-term all-cause mortality [17] [18] [19] in HFrEF and stress cardiomyopathy, however, data on shortterm, inpatient outcomes are lacking. In other acute conditions involving myocardial injury (e.g. myocardial infarction 20 and pulmonary embolism 21 ), strain abnormalities were shown to be largely reversible and thus effects on inpatient mortality remain unelucidated. Histopathological findings suggest that abnormal RV strain is the functional correlate for myocardial fibrosis 22 and offer a possible explanation as to why acute (possibly reversible) changes in RV strain will not necessarily influence adverse events within a relatively short time span of an acute hospitalization. Alternatively, the lack of an association between abnormal RV FWS on mortality in our study could be explained by selection bias introduced by the strict triage process at our institution that was specifically directed towards the cardiac assessment of patients whose clinical picture and symptoms appeared to be out of proportion to COVID-19-related lung disease. Hence, TTEs were preferentially performed on more acutely ill individuals in which decreased cardiac function was thought to be attributable to primary myocardial injury, as opposed to known secondary changes related to respiratory failure. Furthermore, patients with extensive lung disease tended to have worse image quality not always adequate for strain analysis.

Hypotheses to explain LV dysfunction in COVID-19 including direct myocardial damage and inflammatory cytokine storm are under active investigation 23 . The theory of direct myocardial damage finds support in a recent report Churchill et al. who found that almost half of patients with high-sensitivity troponin T (hs-TNT) levels ≥ 50 ng/mL had evidence of LV dysfunction 24 .

While Li et al. and Szekely et al. documented that mean LVEF was generally ≥ 50% and did not show statistically significant differences after comparison by survival status, myocardial strain status, or clinical severity of COVID-19, our group has observed in a prior analysis of our COVID-19 population that patients with LVEF < 50% had significantly higher mortality than those with LVEF ≥ 50% 25 .

Our most notable finding is that mortality is increased even when changes in left ventricular function in patients with COVID-19 are considered seemingly "subclinical". In our patient population with COVID-19, LVEF in both survivor and non-survivor groups was ≥50%, limiting predictive information based on LVEF alone. However, when LV GLS was also assessed, those individuals with normal LVEF but abnormal LV GLS had a mortality risk closer to those with abnormal LVEF. Our survival analysis further incorporated adjustments for other clinical parameters. While volume status has been shown by Burns et al. to acutely influence LV systolic strain 26 , the association of abnormal LV strain with mortality was independent of NT-proBNP levels at the time of TTE which was used as a surrogate for volume status and was not significantly different across our comparison groups.

LV strain traditionally has been used to predict long-term outcomes in patients with asymptomatic type 2 diabetes 27 , patients treated with anthracyclines 28 , as well as in patients with aortic stenosis with preserved ejection fraction 29 . Only little data exists on the role of abnormal J o u r n a l P r e -p r o o f LV strain in acute settings. While our population was limited to patients hospitalized with COVID-19, this study to our knowledge is the first showing that abnormal LV GLS was significantly and independently from LVEF associated with increased acute, inpatient mortality.

This might signal that the scope of abnormal LV GLS might possibly useful in other conditions in which myocardial injury might not be detected in the form of overtly reduced LVEF.

While COVID-19-related myocardial changes are found in both RV and LV, our work should incite further discussion about their potentially differential relationships with inpatient and longterm outcomes. Our work suggests that LV strain analysis might serve as an incremental predictive factor for mortality in patients with COVID-19, especially where (normal) LVEF alone would fail to unveil pertinent information on mortality.

We found that LV GLS with an optimal cutoff threshold of -13.8 % was able to predict survival status in patients hospitalized with COVID-19. Due to the nature of ROC analysis, this is expectedly less negative than the pooled normal value from a meta-analysis using healthy individuals which is reported to be -19.5% (95% CI -20.4 --18.9%) 30 , and less negative than the 5 th percentile in a healthy general population (-18.4%) 31 . Our threshold ranges between the mean LV GLS of patients with acute pulmonary embolism (-16%) 32 and those with stress cardiomyopathy (-11.6%) 33 , or acute decompensated heart failure (-9.8%) 34 . Based on our review of literature, this is the first report to provide a LV GLS cutoff value for presumed acute myocardial injury in the setting of COVID-19.

Patients with this "new abnormal" cut-off for LV GLS showed a mean RV FWS of -13.7 % which is comparable to a value of -13.1% found in patients with acute decompensated heart failure associated with adverse events in 36-month follow-up 34 . This was also less negative than -23.3%, a lower limit of normal of 3-segment RV FWS originating from a reference group of 276 healthy volunteers 35 .

Li et al. report that in a cohort of 120 patients admitted for COVID-19, the mean 3-segment RV FWS was -18.9% in non-survivors 9 , while our corresponding group showed a mean of -16.5%.

In contrast to Li et al., neither data segmentation into quartiles nor sensitivity analysis for RV FWS was able to predict survival outcomes in patients with COVID-19 with statistical significance.

To our knowledge, only one other study to this date has compared the effects of both RV and LV strain on clinical outcomes. Cameli et al. showed that in patients with HFrEF referred for heart transplant, RV FWS (AUC 0.87) with a cutoff of -15% was a better predictor of adverse outcome events than LV GLS (AUC 0.26) with a cutoff of -8.1% 36 . However, our study population with COVID-19 is significantly different in the regard that the majority of patients (35 out of 48) had normal ejection fraction, while the same cannot be said of an advanced heart failure population.

In the latter population, RV parameters gradually gain importance in the prediction of mortality as LV parameters including strain and EF lose their clinical utility as disease progresses. In contrast, in a study population like ours in which most patients are presumed to have subclinical myocardial damage, we showed that LV strain can uncover these subclinical differences. Overall, our work is the first to compare RV and LV strain parameters in an acute setting, more specifically when an acute myocardial injury (here due to is suspected, in a population with comparably low prevalence of manifest cardiovascular comorbidities.

We acknowledge that one of the biggest limitations of this study is the relatively low number of diagnostic echocardiograms, due to the significant number of studies excluded due to image quality and focused nature of TTEs in this patient population. Precision of results from adjusted analyses were limited by our cohort size and thus low numbers of events. Furthermore, our study might be subject to selection bias due to the careful selection of patients appropriate for TTE by our triage criteria. The lack of long-term follow-up precludes any conclusions on post-discharge mortality, although all patients included had a final disposition at the time of study. Our work should be considered hypothesis generating and further research including prospective work should be initiated. Importantly, more widespread use of strain imaging in patients with COVID-19 is feasible since it requires no added scanning time and may enable future research including the incorporation of longitudinal follow-up data of serial echocardiograms, using myocardial strain in treatment monitoring, as well as its use in lung disease.

J o u r n a l P r e -p r o o f

In patients hospitalized with COVID-19, abnormal LV strain is an early marker for increased risk of death independent of LVEF. This association persists after further adjustment for biomarkers, clinical risk factors including underlying heart and lung disease, as well as smoking status. Conversely, such association was not present with echocardiographic RV strain markers in our study.

J o u r n a l P r e -p r o o f 

Association of Cardiac Injury with Mortality in Hospitalized Patients with COVID-19 in Wuhan, China

Impaired systolic function by strain imaging in heart failure with preserved ejection fraction

The Functional Role of Longitudinal, Circumferential, and Radial Myocardial Deformation for Regulating the Early Impairment of Left Ventricular Contraction and Relaxation in Patients With Cardiovascular Risk Factors: A Study With Two-Dimensional Strain Im

Association of left ventricular strain with 30-day mortality and readmission in patients with heart failure

Role of serial quantitative assessment of right ventricular function by strain in pulmonary arterial hypertension

Transient depression of myocardial function after influenza virus infection: A study of echocardiographic tissue imaging

The Right Ventricle in ARDS

Prognostic Value of Right Ventricular Longitudinal Strain in Patients with COVID-19

Inpatient Transthoracic Echocardiography During the Coronavirus Disease

Noninvasive Myocardial Strain Measurement by Speckle Tracking Echocardiography

Two-dimensional strain-A Doppler-independent ultrasound method for quantitation of regional deformation: Validation in vitro and in vivo

Age-and sex-based reference limits and clinical correlates of myocardial strain and synchrony: The framingham heart study

Right Ventricular Longitudinal Strain Reproducibility Using Vendor-Dependent and Vendor-Independent Software

The Spectrum of Cardiac Manifestations in Coronavirus Disease 2019 (COVID-19) -a Systematic Echocardiographic Study

Right ventricular response to intensive medical therapy in advanced decompensated heart failure

Right Ventricular Longitudinal Strain Measures Independently Predict Chronic Heart Failure Mortality

Right ventricular global longitudinal strain provides prognostic value incremental to left ventricular ejection fraction in patients with heart failure

Prevalence and prognostic role of right ventricular involvement in stress-induced cardiomyopathy

Right Ventricular Function After Acute Myocardial Infarction Treated With Primary Percutaneous Coronary Intervention (from the Glycometabolic Intervention as Adjunct to Primary Percutaneous Coronary J o u r n a l P r e -p r o o f Intervention in ST-Segment Elevation Myocardial Infarction III Trial)

Reversible Right Ventricular Regional Non-Uniformity Quantified by Speckle-Tracking Strain Imaging in Patients With Acute Pulmonary Thromboembolism

RV longitudinal deformation correlates with myocardial fibrosis in patients with end-stage heart failure

COVID-19 and Cardiovascular Disease

Echocardiographic Features of COVID-19 Illness and Association with Cardiac Biomarkers

Left Ventricular Systolic Function and Inpatient Mortality in Patients Hospitalized with Coronavirus Disease 2019 (COVID-19)

Left ventricular strain and strain rate: Characterization of the effect of load in human subjects

Rate Imaging in Asymptomatic Patients With Type 2 Diabetes Mellitus

Myocardial Strain Is Associated with Adverse Clinical Cardiac Events in Patients Treated with Anthracyclines

Apical four-chamber longitudinal left ventricular strain in patients with aortic stenosis and preserved left ventricular ejection fraction: Analysis related with flow/gradient pattern and association with outcome

Normal ranges of left ventricular strain: A meta-analysis

Left ventricular strain and strain rate in a general population

Reversible left ventricular regional non-uniformity quantified by speckle-tracking displacement and strain imaging in patients with acute pulmonary embolism

Characteristics and Outcomes of Patients With Takotsubo Syndrome: Incremental Prognostic Value of Baseline Left Ventricular Systolic Function

Incremental Prognostic Value of Right Ventricular Strain in Patients With Acute Decompensated Heart Failure

Sex-and method-specific reference values for right ventricular strain by 2-dimensional speckle-tracking echocardiography

Comparison of right versus left ventricular strain analysis as a predictor of outcome in patients with systolic heart failure referred for heart transplantation

Conceptualization, Formal analysis, Investigation, Writing -Original draft, review & editing. Yekaterina Kim: Conceptualization, Investigation, Writing -Original draft, Data curation

Lissa Sugeng: Conceptualization, supervision, writing -review and editing