The Strengths and Limitations of E/e’ in Heart Failure with Preserved Ejection Fraction

In the 1980s Doppler echocardiography emerged as non-invasive alternative to cardiac catheterization for the evaluation of hemodynamics.^{1, 2} Since then, numerous indices have been advanced for estimation of left ventricular (LV) filling pressures.³ The quotient of early diastolic mitral inflow velocity and mitral annular tissue velocity (E/e’ ratio) is probably the most extensively studied of these indices. Guidelines for the diagnosis of heart failure with preserved ejection fraction (HFpEF) proposed by the European Society of Cardiology (ESC) and the recommendations for the evaluation of LV diastolic function by the American Society of Echocardiography (ASE) and the European Association of Cardiovascular Imaging (EACVI) both include the E/e’ ratio in their algorithms.^{3, 4} However, a number of recent studies have raised questions regarding the accuracy of E/e’ in patients with or at risk for HFpEF, and as such, clinical uncertainty remains as to how best to apply these data in practice.^5–11

In this issue of the Journal, Nauta and colleagues present an elegant, comprehensive systematic review and meta-analysis aimed at this important topic.¹² The authors sought to unify the available evidence to better understand correlations between guideline-recommended echocardiographic parameters and invasively-measured LV filling pressures, while also exploring their prognostic value in patients with HFpEF. The echocardiographic parameters examined included left atrial (LA) volume index, LV mass index, tricuspid regurgitation (TR) velocity, mitral annular tissue velocity, and the E/e’ ratio. HFpEF was defined by signs and symptoms of HF and normal or mildly reduced EF (EF>45%).

The authors identified nine studies that reported correlation of one or more of the echocardiographic parameters with invasive hemodynamics in HFpEF.¹² While all studies reported the correlation between E/e’ ratio and invasively-measured pulmonary capillary wedge pressure (PCWP) or LV end-diastolic pressure (EDP), very few studies reported correlations of the other guideline-recommended echocardiographic measures with invasive indices. The authors found that the correlation between E/e’ and filling pressure varied widely between studies (r=0.19–0.84), with a pooled correlation coefficient of 0.56, indicating a modest correlation. Even when restricting to simultaneous studies or those with EF≥50% alone, the correlations remained modest (r=0.59–0.62).¹²

Nauta and colleagues next evaluated the prognostic value of echocardiographic indices in patients with HFpEF using data from in eighteen studies.¹² E/e’ was associated with outcome (mortality and/or composite endpoint) in most studies, with a combined hazard ratio of 1.05 (1.03–1.06) per 1 unit increase. Sensitivity analysis showed that prognostic value was similar in patients with EF>45% and >50%. Other echocardiographic parameters were less predictive than E/e’. The authors conclude that the evidence base supporting use of echocardiography to estimate hemodynamics and stratify risk in HFpEF is relatively small, that E/e’ appears to be the most robust of the indices studied to date, but that its correlations with filling pressures and outcomes are relatively modest.¹²

The authors are to be commended on this important contribution that provides more unified insight into the strengths and weaknesses of echocardiographic parameters that are used in the evaluation of HFpEF and diastolic dysfunction in general.¹² Of the five echocardiographic indices examined, E/e’ ratio has been (by far) the most extensively studied for estimation of LV filling pressure. The authors suggest several possible explanations for the rather modest correlations between E/e’ and filling pressures, including the age dependency of both E and e’ velocities, limited accuracy of e’ velocity in subjects with regional dysfunction, effects of arrhythmias including atrial fibrillation (AF), and the presence of early stage HFpEF. However, even studies that have carefully excluded patients with these confounders and performed simultaneous imaging have often reported poor correlations between E/e’ and filling pressures.^{9, 13}

While correlative analyses are highly relevant, it is also important to consider accuracy, sensitivity, and specificity. The E/e’ ratio, despite its modest correlations,¹² displays excellent specificity for identifying patients with elevated LV filling pressure.^7–9 Thus, an elevated E/e’ may be a useful tool to rule-in heart failure in the proper clinical setting. The problem is, a normal E/e’ does not rule out HFpEF (or high filling pressures) because this index displays very poor sensitivity.^7–9 In a large (n=159), simultaneous, multicenter study from the EACVI, Lancellotti et al. observed that the sensitivities for septal and lateral E/e’ to identify high LV filling pressures were 19% and 14%, respectively, dropping to 4.8% and 2.4% in patients with EF ≥50%, respectively.⁹

Because no single index robustly reflects diastolic function and hemodynamics, the EACVI/ASE consensus guidelines have developed algorithms that incorporate multiple indices, in the interest of providing complementary information.³ This approach relies on 4 indices: e’ velocity, E/e’ ratio, tricuspid regurgitation velocity, and left atrial volume. As discussed by the Nauta and colleagues,¹² recent data validating these guidelines have also raised questions regarding diagnostic accuracy. Of the five studies described in Table 4 of their review, the sensitivity for the algorithmic approach varied extremely widely, between 34% and 87%^{8–11, 14} Pooling together the data from these 5 studies (n=1233), the weighted average sensitivity to identify high filling pressures is only 61%. Does anyone really want to tell the patient in clinic presenting with dyspnea that they are 61% certain that the patient does not have heart failure?

It is important to remember that sensitivity (not specificity) is necessary to rule out disease. We often employ less expensive and less invasive tests (like echocardiography) for this purpose. But if the less invasive test cannot exclude disease, then one must seriously question how it is best used. In contrast, the same studies reviewed by Nauta and colleagues showed a more consistent and favorable specificity for the ASE/EACVI algorithm (86% pooling the studies), so perhaps this is the superior application.¹²

Many patients with HFpEF have normal filling pressure when they are assessed at rest, only to develop profound increases in filling pressures during physiological stresses like exercise.^{15, 16} Only 1 study using simultaneous catheterization-echocardiographic evaluation at rest and during exercise in patients being evaluated for exertional dyspnea (EF≥50%) has been published to date.⁸ This study again demonstrated low sensitivities for the ASE/EACVI algorithm³ (34%) and for the ESC guidelines⁴ (60%) to diagnose HFpEF. However it must also be considered that resting invasive assessment was also poorly sensitive, since 44% of the patients in that study developed increases in filling pressure exclusively during exercise. Addition of E/e’ during exercise improved the sensitivity compared to resting assessment alone, but at the cost of a decrease in specificity and reduction in the number of patients with diagnostic quality images due to poorer signal quality during exercise.⁸

Thus, as suggested by Nauta and colleagues, exercise echocardiography may also be useful in the evaluation of HFpEF.¹² However, given the wide variability in findings that have been observed in studies evaluating patients at rest, it would seem essential that additional studies be performed to reproduce the currently available data.⁸ Ideally, such studies would be multicenter and involve simultaneous assessment, with blinded echocardiographic and hemodynamic interpretation, and definitive ascertainment of HFpEF or non-cardiac dyspnea (NCD) status by the gold standard of invasive exercise hemodynamics.^{15, 16}

One striking fact raised by the review of Nauta and colleagues is indeed the relatively small evidence base in HFpEF.¹² Indeed, there may be high risk for publication bias in this field. Neutral studies may be perceived by authors, reviewers and journal editors as less valuable, which may then deter publication. This should clearly not be the case, because negative data are just as valuable as positive findings. To explore the reproducibility of exercise echocardiography in follow up to our recent study,⁸ we reviewed the charts from a different series of 206 consecutive patients with invasive verification of diagnosis (134 HFpEF, 72 NCD) who had also undergone noninvasive cardiopulmonary exercise testing.¹⁶ From this cohort, we observed that 48 of the patients (35 with HFpEF) had undergone exercise echocardiography as part of their clinical evaluation outside of research and prior to referral to the catheterization laboratory. In contrast to our earlier data,⁸ we found that only 10/35 of patients with invasively-proven HFpEF in this “real world” clinical setting were correctly identified by exercise E/e’ (sensitivity 29%, AUC 0.64, B. Borlaug, personal communication). Of course, this observation from a selected cohort in a separate study is biased, retrospective and predicated on a relatively small sample size, but the observation makes the point plainly: what is demonstrable in rigorously-controlled research settings needs to be translatable to real world practice and reproducible in multiple centers. The data from Nauta and colleagues emphasizes how further study is required to validate the current evidence base that guides the use of exercise echocardiography in the evaluation of HFpEF.¹²

The association between E/e’ and adverse outcomes from the current data is another important finding.¹² Since E/e’ tracks with both hemodynamics and prognosis, it raises the question whether it might be an intermediate endpoint for clinical trials. The E/e’ ratio does not change with as much dynamic range as filling pressures do,⁸ but a recent study demonstrated reduction in E/e’ with treatment, which was correlated with improved exercise capacity.¹⁷ In contrast, other studies have failed to identify any changes in E/e’ with medical therapy, even as alternative measures of congestion such as natriuretic peptides or LA volume improved.¹⁸ Thus, more data on the utility of diastolic indices (including E/e’) as endpoints would also be of great interest.