In recent years, the importance of heart failure with preserved ejection fraction has been increasingly recognized, and the demand for a simple, noninvasive, bedside test for validating this diagnosis has dramatically increased. Echocardiography is the natural candidate technique for this task, but the echocardiographic detection of left ventricular (LV) diastolic dysfunction has been fraught with differing concepts and confusing terminology, including the question of what exactly constitutes diastolic dysfunction. At the bottom of this problem lies the inability to obtain noninvasively the LV diastolic pressure-volume relationship, because echocardiography cannot measure absolute pressures. Many echocardiographic signs and patterns of diastolic dysfunction have been described over the years, but the multifactorial nature, in particular the load dependency, of echocardiographically obtainable parameters of diastolic LV physiology has dogged efforts to find a simple way to assess diastolic function. The problem therefore has been restated in a simpler, although perhaps simplistic, way: how can we diagnose elevated filling pressures by echocardiography, regardless of the complex pathophysiology (which involves the respective roles of active relaxation, chamber compliance, interventricular dependence, etc)? The group of Nagueh et al, in which the senior author of the article by Dokainish et al in this issue of the Journal was a coworker, has pioneered the use of a novel parameter to estimate “filling pressure,” namely, the ratio of peak early transmitral blood flow velocity (E) and peak early diastolic tissue velocity at the base of the left ventricle (e′). They studied the correlation of E/e′ with LV filling pressures in several clinical scenarios and found reasonable correlations in patients with a wide range of ejection fractions. However, the reported relations have always been far from perfect and too weak to base clinical decisions solely on this parameter. Moreover, a recent study in patients who in a substantial proportion had wide QRS complexes and mechanical dyssynchrony indicated that the correlation is further degraded in such circumstances.
Elegant experimental work from Otto Smiseth’s group in Norway has further refined our understanding of the determinants of e′. They identified 3 main LV myocardial factors for e′:
- •
the rate of LV relaxation;
- •
“lengthening load,” the load the myocardium experiences during relaxation, a parameter akin, but not identical, to preload; and
- •
restoring forces, which, similar to the restoring force of a compressed spring, reflect the preceding systolic contraction.
Thus, the magnitude of e′ is highly multifactorial, and the relation of E/e′ to LV filling pressures is not a hewn-in-granite physical law but an empirical relationship that fortuitously can be used clinically in many cases but fails in others (a similar example would be the inverse relation of transmitral pressure half time to stenotic mitral orifice area).
The study of Dokainish et al in this issue illustrates these difficulties and contributes to a better understanding of the role of E/e′ and other echocardiographic parameters in the diagnosis of elevated LV pressures. The study defined elevated filling pressure as a LV pre-A pressure > 15 mm Hg. Several closely correlated pressure parameters have been used to characterize LV diastolic pressure levels: mean pulmonary capillary pressure, mean left atrial pressure, mean LV diastolic pressure, pre-A pressure (pressure before the onset of the atrial or A wave in the LV pressure tracing), or LV end-diastolic pressure (LVEDP; LV pressure at the onset of LV isovolumic contraction or the peak of the R wave on electrocardiography, well after the A pressure). A mean pulmonary capillary pressure > 12 mm Hg or an LVEDP > 16 mm Hg is considered elevated. Although LVEDP is the most widely used parameter, pre-A pressure has been shown to be closer to mean left atrial pressure than LVEDP in case of a stiff ventricle at end-diastole and therefore was chosen as the target parameter in this study. Pre-A pressures were invasively measured in 122 patients with preserved ejection fractions (>50%) undergoing coronary angiography, with a high incidence of significant coronary disease (65%) and associated risk factors (hypertension in 88%, diabetes in 55%), and several echocardiographic parameters were analyzed regarding their ability to predict elevated pre-A pressure. Pressure and echocardiographic measurements were performed within 20 minutes, not simultaneously, as would be ideal, but by and large, one would expect patients with diastolic dysfunction to demonstrate reasonably stable hemodynamics whether supine on the catheterization table or a few minutes later in a left lateral decubitus position for echocardiography.
The limited utility of E/e′ for individual patient assessment is immediately evident from Figure 1 A in Dokainish et al’s report, in which E/e′ varied between 7 and 24 at the directly measured pre-A pressure of 15 mm Hg (the proposed cutoff to distinguish normal and elevated filling pressure), with a modest correlation coefficient of 0.63. Left atrial volume and peak systolic pulmonary arterial pressure were then added to E/e′ in a stepwise fashion to predict elevated pre-A pressures. These two parameters, taken singularly, were even weaker predictors of filling pressures, as shown in Figures 1 B and 1C. Nevertheless, the combination of E/e′ and left atrial volume increased diagnostic accuracy markedly in the patients whose E/e′ ratios were in the “gray zone” of 8 to 13, whereas it did not improve diagnostic accuracy in patients whose E/e′ ratios were already high (>13). The presence of (1) E/e′ > 13 or (2) E/e′ of 8 to 13, together with left atrial volume > 31 mL/m 2 , predicted pre-A pressures > 15 mm Hg with satisfactory sensitivity of 87% and specificity of 88%. Considering calculated peak systolic pulmonary pressure in addition to E/e′ and left atrial volume index did not further improve accuracy. Thus, by appropriately stratifying patients using E/e′, the additional diagnostic information of left atrial volume could be “leveraged” to increase overall diagnostic accuracy considerably. This procedure, which has been formalized in statistics as “classification and regression tree” analysis, harnesses Reverend Bayes’s deceptively simple yet enormously powerful insight that posttest probability is modified by pretest likelihood. Indeed, the data presented by Dokainish et al nicely confirm the clinical utility of the diagnostic pathway outlined in the recently published joint guidelines of the American Society of Echocardiography and the European Society of Echocardiography that call for an integrative approach using several parameters to estimate LV filling pressures.
