What Is the Evidence That the Tissue Doppler Index E/e′ Reflects Left Ventricular Filling Pressure Changes After Exercise or Pharmacological Intervention for Evaluating Diastolic Function? A Systematic Review

Background Noninvasive echocardiographic tissue Doppler assessment (E/e′) in response to exercise or pharmacological intervention has been proposed as a useful parameter to assess left ventricular (LV) filling pressure (LVFP) and LV diastolic dysfunction. However, the evidence for it is not well summarized. Methods and Results Clinical studies that evaluated invasive LVFP changes in response to exercise/other interventions and echocardiographic E/e′ were identified from PubMed, Scopus, Embase, and Cochrane Library databases. We grouped and evaluated studies that included patients with preserved LV ejection fraction (LVEF), patients with mixed/reduced LVEF, and patients with specific cardiac conditions. Overall, we found 28 studies with 9 studies for preserved LVEF, which was our primary interest. Studies had differing methodologies with limited data sets, which precluded quantitative meta‐analysis. We therefore descriptively summarized our findings. Only 2 small studies (N=12 and 10) directly or indirectly support use of E/e′ for assessing LVFP changes in preserved LVEF. In 7 other studies (cumulative N=429) of preserved LVEF, E/e′ was not useful for assessing LVFP changes. For mixed/reduced LVEF groups or specific cardiac conditions, results similar to preserved LVEF were found. Conclusions We find that there is insufficient evidence that E/e′ can reliably assess LVFP changes in response to exercise or other interventions. We suggest that well‐designed prospective studies should be conducted for further evaluation.

L eft ventricular diastolic dysfunction leading to heart failure with preserved ejection fraction (HFpEF) is a major clinical problem. [1][2][3] Elevated left ventricular filling pressure (LVFP) is often used as a clinical surrogate for impaired diastolic function in patients with preserved left ventricular ejection fraction (LVEF). 4,5 LVFP is usually measured at rest in routine clinical practice. However, changes in LVFP with exercise or other physiological intervention provide incremental information to assess diastolic function. [5][6][7][8][9][10] A direct measurement of LVFP requires an invasive intervention, which has significant risk and costs, and is therefore performed in select patients only. Echocardiography is frequently used for noninvasive evaluation of diastolic function and estimating LVFP. [4][5][6] Echocardiographic quantification of LVFP is based on E/e 0 measurement, which is the ratio of the early diastolic velocity on transmitral Doppler (E) and the early diastolic velocity of mitral valve annulus obtained from tissue Doppler (e 0 ). [4][5][6][11][12][13][14] The guidelines recommend using E/e 0 in evaluating LV diastolic function. [4][5][6]10 In research studies, E/e 0 is also used as a primary or secondary end point for assessing the treatment efficacy and quantifying changes in LVFP. [11][12][13][14][15][16][17][18][19] Despite extensive use of E/e 0 , there continues to be ongoing debate about the usefulness of E/e 0 in assessing LVFP. [20][21][22][23][24] In our recent comprehensive meta-analysis, we have found limited evidence for the use of E/e 0 under resting conditions to estimate LVFP in preserved LVEF. 20 It has been suggested that changes in E/e 0 with exercise or other physiological/pharmacologic interventions may more accurately reflect changes in the LVFP and diastolic properties. 5,6,8,10,25 Here we decided to evaluate the evidence describing the relationship of E/e 0 and LVFP in preserved LVEF with exercise or other physiological interventions. We also summarize the available evidence describing the relationship of E/e 0 and LVFP in a wider spectrum of LVEF and for specific cardiac conditions.

Search Strategy and Study Selection
Original clinical studies that evaluated LVFP by using echocardiographic E/e 0 and invasive techniques were screened from PubMed, Scopus, Embase, and Cochrane Library databases to September 2016 using a number of search strategies (Figure). Specific search terms and full-text studies excluded after evaluation are listed in Tables S1 and   S2. Clinical studies (in English) that reported changes in E/e 0 and invasively measured LVFP attributable to physiologic and/or pharmacologic or other therapeutic intervention and/ or repeated serial measurements in the adult subjects (age >18 years) with any LVEF and clinical conditions were included. References of important studies were also reviewed for comprehensive search. LVFP measurements included LV end diastolic pressure, LV pre-A wave pressure, LV mean diastolic pressure, left atrial pressure, and pulmonary capillary wedge pressure (PCWP) obtained during the left or right heart catheterization or from a permanently implantable cardiac pressure monitoring system. Only studies that utilized transthoracic echocardiographic pulsed-wave tissue Doppler imaging for E/e 0 measurements at interventricular septum (E/e 0 septal ), lateral mitral annulus (E/e 0 lateral ), and/or mean of septal and lateral values (E/e 0 mean ) were selected. The studies were included if they reported at least 1 of the following data sets: (1) E/e 0 and LVFP values at baseline and after intervention; (2) changes in E/e 0 and LVFP values because of intervention; (3) assessment of correlation between E/e 0 and LVFP postintervention, alone, or combined with baseline; (4) assessment of correlation between changes in E/e 0 and LVFP with intervention; and (5) the diagnostic accuracy of either postintervention E/e 0 values or postintervention changes in E/e 0 to predict elevated LVFP or LVFP changes.

Patient Cohorts and Study Analysis
Included studies were grouped and analyzed based on patient cohorts. The first group was for studies that included patients with LVEF ≥50%, including HFpEF patients, but without a substantial number of moderate-to-severe valvular heart disease, hypertrophic or restrictive cardiomyopathy, congenital heart disease, acute coronary syndromes, septic shock, cardiac transplant, and atrial fibrillation. This group was our primary interest. Other groups were for studies that included patients with reduced/mixed LVEF, and for studies that included patients grouped with specific cardiac conditions (eg, cardiac transplants). Overall, we found 28 studies: 9 studies 24,26-33 for our primary interest, and 19 studies 25,34-51 for secondary interest (Figure). One study included a data set for primary and secondary interest. 30 Since most of the studies were single center with differing methodologies with many reporting only a limited data set, we chose descriptive methodologies to summarize the results.

Results
Studies in LVEF ≥50% With or Without HFpEF Table 1 summarizes study details and results for the 9 studies that included participants with preserved LVEF (≥50%), including HFpEF patients (see Table S3 for more details). All studies, except 1, 29 had a prospective design and all studies, except 1, 32 simultaneously measured echocardiographic and hemodynamic variables. Most of these studies had a low sample size (median N = 22 with interquartile range of . Three of these studies had subjects perform exercise stress echocardiography using a supine bicycle 29,31 or passive and then active leg-raise 33 for evaluating patients with suspected HFpEF. There was an increase in invasive LVFP but no consistent relationship for the changes in E/e 0 postintervention in these 3 studies. Talreja et al 31 found that E/e 0 provides a reliable estimation of PCWP with exercise in a small study of 12 patients. Based on their scatterplot, 31 we estimated that stress E/e 0 septal >15 predicts PCWP ≥20 mm Hg with sensitivity of 83% and specificity of 100%. Maeder et al 29 found decreased E/e 0 septal with exercise and no correlation between poststress E/e 0 septal and PCWP. In the largest exercise study in patients with exertional dyspnea (N=181), Choi et al 33 recorded no change of E/e 0 septal despite a significant elevation of LV end diastolic pressure with passive and active leg raise.
In another set of studies, authors performed stress echocardiography using differing pharmacological interventions 27,30,32 or body position change 24 that resulted in significant decrease of LVFP (Table 1). Only in 1 small study, 32 authors reported the decrease of group average E/e 0 lateral in response to decreased LVFP for 10 HFpEF patients. However, this study did not provide any individual data for further analysis. Interestingly, in another publication from the same group 45 (Table 2), authors reported no correlation between individual changes of E/e 0 and LVFP for a combined cohort of 10 HFpEF and 15 heart failure with reduced ejection fraction (HFrEF) patients. In studies 24,27,30 with a total of 179 HFpEF and/or coronary artery disease patients, there were no significant changes in E/e 0 values despite reduced LVFP. Furthermore, in these studies there was no significant correlation between postintervention values of E/e 0 and LVFP or between individual changes in E/e 0 and LVFP.
In 2 other studies, participants underwent preload changes leading to lower LVFP caused by low body negative pressure and increase in LVFP by saline infusion. 26,28 Both studies found that E/e 0 cannot reliably track changes in LVFP in healthy people 26,28 and in HFpEF patients. 26 Studies in Reduced or Mixed LVEF Table 2 summarizes study details and results for the 10 studies that included participants with mixed or reduced LVEF (see Table S4 for more details). In the study of Burgess et al, 25 which included 37 unselected patients with varying LVEF, authors reported a significant correlation (r=0.59) between E/e 0 septal and LV mean diastolic pressure during single-leg supine exercise. They reported high AUC value (0.89) for exercise E/e 0 septal to predict an elevation of LV mean diastolic pressure >15 mm Hg. 25 In their reports for the same patient cohort, E/e 0 septal >13 had sensitivity of %70% and specificity of %95% for estimating elevated LV mean diastolic pressure >15 mm Hg. 25,44 In another study of 22 patients, 46 mean E/e 0 lateral increased with preload stress. However, on detailed analysis, E/e 0 lateral increase was observed in only a small subset of patients (N=6). No correlation of E/e 0 and LVFP or diagnostic value of E/e 0 lateral was reported. 46 In another study in patients with reduced LVEF (N=40), authors reported a significant correlation between exercise E/e 0 and LVFP and a paradoxical decrease of exercise E/e 0 values despite LVFP elevation. 40 In 4 studies, investigators used different pharmacological agents to decrease LVFP and measured corresponding changes in E/e 0 (Table 2). 30   In 2 studies, the investigators performed serial measurements using implanted hemodynamic measurement devices (Table 2). 41,50 In 1 study of 15 patients with chronic heart failure with reduced ejection fraction, the investigators found high diagnostic values of E/e 0 mean , E/e 0 lateral , or E/e 0 septal to predict the elevated mean left atrial pressure (≥15 mm Hg). In their study, E/e 0 lateral and E/e 0 mean had no correlation and E/e 0 septal had only modest correlation (r=0.46) with mean left atrial pressure on serial measurements. 41 In another study of 5 patients with reduced LVEF, the investigators found no correlation between E/e 0 and LVFP and no significant diagnostic value of E/e 0 to detect elevated LVFP on serial measurements. 50 Table 3 summarizes study details and results for the 10 studies that included participants with specific cardiac conditions (see Table S5 for more details). In 3 studies, cardiac transplant patients were evaluated. 37-39 In 1 study of 14 transplant patients, serial measurements revealed an excellent correlation between changes in E/e 0 mean and changes in PCWP. 39 In contrast, in another study with a larger cohort (N=57), there was no difference in the E/e 0 values postexercise despite changes in PCWP. 37 In another study with similarly large sample cohort, the investigators found low predictive power of exercise E/e 0 for identifying elevated PCWP and a modest correlation for the E/e 0 -PCWP and DE/e 0 -DPCWP. 38 In patients with severe aortic stenosis (N=28) 49 and in patients with recent myocardial infarction (N=61), 48 the investigators concluded that E/e 0 does not reflect exerciseinduced changes in PCWP. 48,49 Three studies measured E/e 0 and PCWP before and after cardiovascular surgery. 34,35,47 In 1 study, a strong correlation between E/e 0 lateral and PCWP was noted before and 30 days after cardiac surgery (coronary artery bypass grafting or aortic valve replacement) (N=52, LVEF 40AE17%). 34 Interestingly in these patients, E/e 0 septal increased whereas PCWP decreased after surgery. 34 In a study of hypertrophic cardiomyopathy (N=17), ethanol-induced septal infarction caused changes in PCWP of either direction, which strongly correlated with changes in E/e 0 lateral . 35 In patients with chronic thromboembolic pulmonary hypertension (with E<A, NYHA class III-IV, and preserved LVEF, N=61), both PCWP and E/e 0 increased following pulmonary thromboendarterectomy. 47 Another study reported a strong correlation between individual changes in PCWP and E/e 0 mean in 9 patients with differing LVEF in the intensive care unit following 2 days of treatment with diuretics and/or inotropes. 43 However, in 51 Table 3. patients with decompensated heart because of advanced systolic HF, no correlation was found between changes in E/e 0 mean and PCWP. 51

Discussion
The major findings of our study are that there is lack of robust clinical evidence to support the use of E/e 0 in response to physiological and/or pharmacological intervention to estimate LVFP changes and LV diastolic dysfunction. Furthermore, most of the studies are single center with limited sample size with nonuniform study methodologies and data reporting that does not allow for quantitative meta-analysis of the studies. Invasive LVFP measurements (primarily LV end diastolic pressure or PCWP as its surrogate) in response to altered physiological conditions provide incremental information about the LV function and stiffness. 4 In proper context, it can be extremely useful in diagnosing diastolic dysfunction. 4 Since some studies 5,35,43,52,53 have suggested that echocardiographic E/e 0 can be used to estimate LVFP quantitatively/ semiquantitatively, there has been tremendous interest in evaluating changes in E/e 0 to physiological and/or pharmacological interventions as a surrogate to changes in LVFP and therefore its potential use in assessing LV diastolic function. 5,6,[11][12][13][14][15][16][17][18][19] Recent meta-analysis has demonstrated that E/e 0 measurements at rest have limited diagnostic accuracy in evaluating LVFP in patients with preserved LVEF. 20 In the present systematic review, we again noted absence of meaningful correlation (where reported) between E/e 0 and LVFP at rest in preserved LVEF (Table S3). In contrast, for the reduced or mixed LVEF group, a stronger correlation between E/e 0 and LVFP is reported (Table S4), which may be related to a wider range of E/e 0 and LVFP (for instance, see Figure 4 in Nagueh et al 54 ). However, other factors may also be playing an important role as Manouras et al 30 demonstrated a higher correlation for the reduced LVEF group compared to preserved LVEF despite similar LVFP and E/e 0 range of values in the 2 cohorts (see Table 3 and Figure 4, Manouras et al 30 ). It is interesting to note that the recent American Society of Echocardiography guidelines propose a consensus-based approach consisting of multiple parameters for evaluating diastolic function in preserved LVEF. 10 Regarding posthemodynamic changes induced by exercise or physiological interventions, we note that there is no significant correlation between E/e 0 and LVFP in preserved LVEF cohorts (Table S3). Moreover, most studies demonstrated worse correlation in mixed or reduced LVEF cohorts after exercise or physiological interventions (Table S4). A similar trend was also noticed in patients with specific cardiac conditions (Table S5).
For evaluating the relationship of change in E/e 0 to changes in LVFP in response to exercise or physiological/ pharmacological intervention, we find that there are only 2 studies with limited sample size that directly 31 or indirectly 32 support the use of E/e 0 for the assessment of LVFP changes in HFpEF patients. In 12 patients, Talreja et al 31 found a promising diagnostic value of specific exercise E/e 0 cutoff (>15) to predict elevation of exercise PCWP (>20 mm Hg). Efstratiadis et al 32 reported concordant reduction of E/e 0 and LVFP following nesiritide infusion in 10 patients. Seven other studies 24,26-30,33 (cumulative N=429) found that E/e 0 does not reliably reflect changes in LVFP in response to physiological or pharmacological intervention in preserved EF. For studies that evaluated mixed LVEF groups, results similar to those of the preserved LVEF group were noted. Only 1 study 25 demonstrated a clinically meaningful relationship and diagnostic characteristics of E/e 0 in estimating elevated LVFP with exercise and reduced exercise capacity. No consistent trends were found in other studies with mixed groups. Also, in specific cardiac conditions we did not find consistent trends across the studies. In the present study we did not evaluate the prognostic value or the pathognomonic mechanisms that may be attributed to the lack of reported relationships with exercise or other interventions of changes in E/e 0 and LVFP. It is well recognized 7 that LVFP may increase in diastolic dysfunction on invasive measurements. However, E/e 0 measurements did not demonstrate a predictable relationship, which may be attributable to the small sample size of individual studies with relatively heterogeneous LV mechanics. This requires further exploration in future studies.
A number of guidelines/tools such as STARD 55 and QUADAS 56 have been developed for evaluating diagnostic test accuracy studies. As evident from our data tables, because of a limited number of studies, limited sample size, and nonuniform methodologies and data reporting, performing such an analysis would not substantially alter our results. Here we are unable to quantify effects of publication bias due to lack of consistent findings and limited studies. However, this is unlikely to affect the overall conclusions.
In summary, our review indicates that there is inadequate evidence for using E/e 0 for estimating LVFP changes in response to exercise/other physiological interventions. Welldesigned prospective multicenter studies are required for evaluation and validation before recommending it for clinical and research purposes.

Sources of Funding
The study was supported by a National Institutes of Health National Heart, Lung, and Blood Institute grant R01-HL104018. The funding organizations did not have any role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

New search
PubMed (