Utility of the E/e’ index in ventilated patients and those with
sepsis
Imran Sunderji 1, Alan G Fraser 2
(Reply to the letter from Filippo Sanfilippo and colleagues,
ECHO-2020-0930)
1 Department of Cardiology, Castle Hill Hospital,
Hull, U.K.
2 Department of Cardiology, University Hospital of
Wales, Cardiff, U.K.
Address for correspondence :
Professor Alan G. Fraser,
University Hospital of Wales,
Heath Park,
Cardiff, CF14 4XW,
Wales, U.K.
fraserag@cf.ac.uk
Telephone: +44 (0)29 2074 5366
Fax: +44 (0)29 2074 4473
915 words
We thank Sanfilippo and his colleagues for their interest in our paper,
and for the opportunity thus afforded to comment on the E/e’ index in
critically ill patients and in those who have severe sepsis.
We agree that the E/e’ index has some utility in predicting successful
weaning from mechanical ventilation, as they have shown in their most
recent meta-analysis,1 but published studies show high
heterogeneity, there are often only small initial differences in mean
E/e’ between patients who will remain off ventilation and those who will
not, and average E/e’ values in both groups are sometimes within normal
or intermediate ranges. Earlier systematic reviews also concluded that a
higher E/e′ ratio is associated with weaning failure in ventilated
patients2 and that E/e′ (as well as other markers of
diastolic dysfunction) predicts mortality in critically ill
patients.3 In a large study of 161 patients, however,
neither E/e’ at the lateral mitral annulus nor any other
echocardiographic index predicted success in weaning.4The heterogeneity of criteria for diastolic dysfunction in these studies
is illustrated by cut-points for abnormal E/e’ varying between 8 and 12
at the lateral mitral annulus and 8 and 9.6 at the medial (septal)
annulus.3
In ventilated as in other patients, both E and e’ are
preload-dependent.5 Positive end-expiratory pressure
(PEEP) reduces both; for example PEEP of 12 cm H2O
decreased lateral e’ by 19.7% and E by 13.7%, so E/e’ was
unchanged.6 An increase in e’ when a patient is taken
off a ventilator could indicate a response to changed loading rather
than an improvement in intrinsic diastolic function. Before concluding
that observed changes in E/e’ imply corresponding changes in left
ventricular (LV) filling pressures, we should consider if E/e’ has been
validated by correlation with pulmonary capilllary wedge pressure (PCW)
measured with Swan Ganz catheters, specifically in ventilated and
critically ill patients.
In 39 patients there was no difference in E/e’ before a trial of
spontaneous breathing, between those subjects in whom it was successful
(defined as PCW remaining <18 mmHg after 60 minutes; mean
baseline E/e’ 8.0) and those in whom it was not (PCW increasing to
>18 mmHg; baseline E/e’ 7.6).7 The area
under the receiver operating characteristic curve (AUC) for E/e’ as a
guide to PCW at the end of the trial of spontaneous breathing was 0.8.
In an earlier study of patients in intensive care who were also
breathing spontaneously, E/e’ had a modest correlation with PCW
(r=0.69); a patient with E/e’ of around 10 could have a PCW ranging from
<10 to >20 mmHg.8 In other
studies of ventilated patients, the correlation of lateral E/e’ with PCW
was 0.849 and its AUC was 0.91.10Recently, Brault et al reported that the 2016 American Society of
Echocardiography and European Association of Cardiovascular Imaging
guidelines for diastolic dysfunction did not accurately assess PCW in 98
ventilated and critically ill patients, of whom 54% experienced septic
shock. The diagnostic score was indeterminate in 49% of patients,
sensitivity and specificity were both 74%, and agreement between
echocardiography and PCW was moderate (Cohen’s Kappa, 0.48). The best
echocardiographic predictor of a normal PCW was not the E/e’ ratio but a
lateral e′ >8.11
From experimental and clinical observations it is clear that severe
sepsis can depress myocardial contractile function, probably through
multiple mechanisms.12 In 40 patients with sepsis,
however, there were no significant correlations between serum
concentrations of inflammatory cytokines and measurements of e’ or
calculated E/e’.13 In another study, mortality was
predicted by the APACHE II score and mitral annular systolic excursion
(MAPSE) with an AUC of 0.88, while the E/e’ index was not selected as a
predictor in a logistic regression analysis.14Reproducibility of echocardiographic measurements in patients with
septic shock is moderate to good15 but it is difficult
to rely on single observations to guide clinical decisions.
In patients with sepsis and severe diastolic dysfunction, failure to
respond to volume replacement may be caused by impaired early diastolic
relaxation and LV suction, which cannot be detected by the E/e’ index.
In a randomised trial, an intravenous infusion of esmolol to slow the
heart rate prolonged LV filling and increased stroke volume, with a
subsequent reduction in mortality.16 In a prospective
observational study, levosimendan increased the probability of
successful weaning from ventilation, and averted any increase in
E/e’;17 that could also be explained by improved early
diastolic relaxation and filling, since levosimendan is positively
lusitropic.18 Detailed echocardiographic assessment of
ventilated patients after cardiac surgery showed that levosimendan
increased early diastolic strain rate by 30%.19 Thus
changes in E/e’ as a marker of mean PCW do not necessarily confirm a
causal relationship with any particular aspect of LV diastolic function,
while more comprehensive echocardiographic analysis of
pathophysiological mechanisms may be more informative.
These thoughts reinforce some of the conclusions that we drew in our
review. Many studies are difficult to interpret because the E/e’ index
is reported without information on changes in its individual components,
and because dichotomising patients into normal or diastolic dysfunction
(grades) loses information from multiple continuous variables that are
inter-related but may change with differing patterns according to
particular circumstances. It is unwise to use discrete cut-points
especially if they are unadjusted for age and gender, and mistaken to
conclude that LV diastolic function has changed when there are
significant differences in the E/e’ index but its mean values remain
within the normal range. The optimal assessment of diastolic dysfunction
in septic and ventilated patients requires a multiparametric approach
and we caution against over-reliance on E/e’.