Non-Invasive Estimation of Left Heart Filling Pressures: Another Nail in the Coffin for E/e'?

This article refers to ‘Echocardiographic estimation of left ventricular and pulmonary pressures in patients with heart failure and preserved ejection fraction: a study utilizing simultaneous echocardiography and invasive measurements’, by Y.M. Hummel et al., published in this issue on pages 1651–1660. The non-invasive diagnosis of left ventricular (LV) filling pressures is important in the management of heart failure (HF) patients with reduced as well as preserved LV ejection fraction (LVEF). To that end, a new expert consensus has proposed a diagnostic algorithm based on the ratio of the early diastolic velocities of mitral inflow and mitral annular motion (E/e'; abnormal when >14), the indexed left atrial volume (LAVi; abnormal when >34 mL/m2), and the maximal velocity of tricuspid regurgitation (TR; abnormal when >2.8 m/s).1 The recommendations have now been tested in two multicentre studies of heterogeneous populations. The first enrolled 450 patients, with a mean LVEF of 47%, including 209 with LVEF <50%. It demonstrated that the echocardiographic algorithm diagnosed elevated LV filling pressure [defined as either mean pulmonary capillary wedge pressure (PCWP) or pre-‘a’ LV diastolic pressure >12 mmHg] with an accuracy of 87%.2 The second, called Euro-Filling, was designed and conducted by the European Association of Cardiovascular Imaging (EACVI) and enrolled 159 patients undergoing elective coronary arteriography, including 39 with LVEF <50%. The new algorithm identified patients with LV end-diastolic pressure (LVEDP) ≥15 mmHg with sensitivity 75%, specificity 74%, positive predictive value 39%, and negative predictive value 93%; the area under the receiver-operating characteristic curve (C statistic) was 0.78, compared with 0.68 using the 2009 recommendations.3 The respective authors concluded that the algorithm had ‘good accuracy’ and was ‘fairly reliable’. This is the context in which Hummel et al. now report their commendably detailed retrospective analysis of 98 patients with HF and preserved LVEF (>45%), including 30% with atrial fibrillation.4 Subjects were selected if they had suspected pulmonary hypertension (for example from previously documented TR >2.8 m/s) so that they could be recruited for a trial of sildenafil to lower pulmonary vascular resistance.5 Echocardiographic measurements and invasive haemodynamic data were obtained simultaneously while patients were supine. Their C statistics for identifying LVEDP ≥15 mmHg are reported as 0.64 for E/e' and 0.76 for LAVi, compared with 0.75 for the algorithm proposed in the American Society of Echocardiography/EACVI recommendations (although their Figure 1 suggests that they did not test a range of discrimination thresholds). Importantly, the results were similar in patients in sinus rhythm and in atrial fibrillation. Perhaps their most thought-provoking finding is that E/e' was unrelated to LVEDP (r = 0.08, P = 0.45) and only weakly related to PCWP (r = 0.24, P = 0.02). When tested as single indices, LAVi and TR velocity also did not predict LVEDP (r = 0.07, P = 0.06; and r = −0.04, P = 0.73, respectively). The E/e' index was first reported in 1997 as a predictor of mean PCWP from a study of 60 patients (with r = 0.87).6 In a recent meta-analysis of data from 24 studies in patients with LVEF >50%, however, there was insufficient evidence that E/e' can reliably estimate LV filling pressure; the sensitivity of E/e' > 13 was only 36% (3–74%) while its specificity was 83% (49–100%).7 In the two recent multicentre studies, the correlations between E/e' and LVEDP were 0.652 and 0.34.3 Elsewhere, limitations of E/e' for predicting LV filling pressure have been reported in patients with HF with reduced ejection fraction8 and in subjects with hypertrophic cardiomyopathy.9 So, is it time to pronounce the death of the E/e' ratio? Probably not—but we do need to exercise caution and be smarter when we use it. The E/e' ratio is not synonymous with LV filling pressure, and it has significant limitations.10 It is just one of many indicators of different aspects of LV diastolic function, and using it as a single index above all other measurements is no more logical than defining systolic function by LVEF alone.11 In patients with quite severe diastolic dysfunction, the E/e' ratio can be uninformative (Figure 1), so a comprehensive study of indicators of preload, LV elastic recoil, early diastolic relaxation, suction, and end-diastolic pressure is more helpful. Hummel et al. have reinforced the message that LV filling pressure is estimated best by using a combination of variables; in their study, the most accurate (but still suboptimal) prediction came from a combination of left atrial reservoir function measured by strain, and LV isovolumetric relaxation time (r = 0.66).4 In patients in sinus rhythm, the duration of retrograde flow in the pulmonary veins during atrial systole is more useful when estimating LVEDP. New methods to assess LV diastolic stiffness using ultrafast plane wave imaging are also promising.12 We should also be considering whether or not estimating LV filling pressure at rest is still the most appropriate diagnostic objective in a subject with suspected HF. The e' velocity declines with age, so E/e' is affected by age and gender as well as by the degree of LV hypertrophy and the LVEF.13 Von Bibra et al. proposed that measurements of e' should be indexed to age.14 The value of e' itself as a diagnostic and prognostic discriminant in addition to LAVi and E/e' has been underscored in the recommendations on diastolic function.1 Since patients with preserved ejection fraction may have relatively normal LV function at rest, exercise echocardiography is important. The E/e' ratio is mildly correlated with LV filling pressure during handgrip stress in supine patients, but less well than at rest.15 In the study by Tan et al.,16 it did not change during exercise. This is probably explained by the fact that both E and e' are preload-dependent indices and their ratio remains unaltered during exercise. Perhaps we have over-used E/e' since it is a simple index when diastolic function is complex and diagnosis difficult. Perhaps it should be considered as a prognostic marker more than a diagnostic one.17, 18 What we need are new methods to integrate all the information from sophisticated non-invasive diagnostic tests at rest and during exercise, and to identify from large prospective studies which tools are the best for diagnosis, prognosis, and monitoring; they may be different. The application of machine learning to this field is one way in which we might achieve these results.19 Conflict of interest: none declared.